agora inbox for [email protected]  
help / color / mirror / Atom feed
Missing headers Windows NT port
267+ messages / 3 participants
[nested] [flat]

* Missing headers Windows NT port
@ 1998-09-10 18:21 Joost Kraaijeveld <[email protected]>
  1998-09-10 18:54 ` Re: [HACKERS] Missing headers Windows NT port Bruce Momjian <[email protected]>
  0 siblings, 1 reply; 267+ messages in thread

From: Joost Kraaijeveld @ 1998-09-10 18:21 UTC (permalink / raw)
  To: Pgsql-Hackers <[email protected]>

Hi,

I am experiencing missing header files. I am not suprised but I would like
to have any missing header file. If I have them I could attempt to write
/find them on my system. Can I access them throught the internet or is
someone kind enought to send them to me on request (Hugo??)?

At this moment I miss: <sys/ipc.h> and <sys/un.h>

TIA

Joost




^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* Re: [HACKERS] Missing headers Windows NT port
  1998-09-10 18:21 Missing headers Windows NT port Joost Kraaijeveld <[email protected]>
@ 1998-09-10 18:54 ` Bruce Momjian <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Bruce Momjian @ 1998-09-10 18:54 UTC (permalink / raw)
  To: Joost Kraaijeveld <[email protected]>; +Cc: [email protected]

> Hi,
> 
> I am experiencing missing header files. I am not suprised but I would like
> to have any missing header file. If I have them I could attempt to write
> /find them on my system. Can I access them throught the internet or is
> someone kind enought to send them to me on request (Hugo??)?
> 
> At this moment I miss: <sys/ipc.h> and <sys/un.h>
> 
> TIA
> 
> Joost
> 
> 
> 

Here they are from BSDI.

-- 
Bruce Momjian                          |  830 Blythe Avenue
[email protected]              |  Drexel Hill, Pennsylvania 19026
  +  If your life is a hard drive,     |  (610) 353-9879(w)
  +  Christ can be your backup.        |  (610) 853-3000(h)

/*-
 * Copyright (c) 1996 Berkeley Software Design, Inc. All rights reserved.
 * The Berkeley Software Design Inc. software License Agreement specifies
 * the terms and conditions for redistribution.
 *
 * BSDI ipc.h,v 2.4 1996/08/21 17:42:01 bostic Exp 
 */

/*
 * Copyright (c) 1988 University of Utah.
 * Copyright (c) 1990, 1993
 *	The Regents of the University of California.  All rights reserved.
 * (c) UNIX System Laboratories, Inc.
 * All or some portions of this file are derived from material licensed
 * to the University of California by American Telephone and Telegraph
 * Co. or Unix System Laboratories, Inc. and are reproduced herein with
 * the permission of UNIX System Laboratories, Inc.
 *
 * This code is derived from software contributed to Berkeley by
 * the Systems Programming Group of the University of Utah Computer
 * Science Department.
 *
 * Redistribution and use in source and binary forms, with or without
 * modification, are permitted provided that the following conditions
 * are met:
 * 1. Redistributions of source code must retain the above copyright
 *    notice, this list of conditions and the following disclaimer.
 * 2. Redistributions in binary form must reproduce the above copyright
 *    notice, this list of conditions and the following disclaimer in the
 *    documentation and/or other materials provided with the distribution.
 * 3. All advertising materials mentioning features or use of this software
 *    must display the following acknowledgement:
 *	This product includes software developed by the University of
 *	California, Berkeley and its contributors.
 * 4. Neither the name of the University nor the names of its contributors
 *    may be used to endorse or promote products derived from this software
 *    without specific prior written permission.
 *
 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
 * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
 * SUCH DAMAGE.
 *
 *	@(#)ipc.h	8.4 (Berkeley) 2/19/95
 */

/*
 * SVID compatible ipc.h file
 */
#ifndef _SYS_IPC_H_
#define _SYS_IPC_H_

#include <sys/cdefs.h>

struct ipc_perm {
	uid_t	uid;	/* user id */
	gid_t	gid;	/* group id */
	uid_t	cuid;	/* creator user id */
	gid_t	cgid;	/* creator group id */
	mode_t	mode;	/* r/w permission */
	u_long	seq;	/* sequence # (to generate unique msg/sem/shm id) */
	key_t	key;	/* user specified msg/sem/shm key */
};

/* common mode bits */
#define	IPC_R		000400	/* read permission */
#define	IPC_W		000200	/* write/alter permission */
#define	IPC_M		010000	/* permission to change control info */

/* SVID required constants (same values as system 5) */
#define	IPC_CREAT	001000	/* create entry if key does not exist */
#define	IPC_EXCL	002000	/* fail if key exists */
#define	IPC_NOWAIT	004000	/* error if request must wait */

#define	IPC_PRIVATE	(key_t)0 /* private key */

#define	IPC_RMID	0	/* remove identifier */
#define	IPC_SET		1	/* set options */
#define	IPC_STAT	2	/* get options */

#ifdef KERNEL
/* Macros to convert between ipc ids and array indices or sequence ids */
#define	IPCID_TO_IX(id)		((id) & 0xffff)
#define	IPCID_TO_SEQ(id)	(((id) >> 16) & 0xffff)
#define	IXSEQ_TO_IPCID(ix,perm)	(((perm.seq) << 16) | (ix & 0xffff))

__BEGIN_DECLS
int	ipcperm __P((struct ucred *,struct ipc_perm *,int));
__END_DECLS
#else 
/* Doesn't really belong but its traditionally here. */
#ifndef _POSIX_SOURCE
__BEGIN_DECLS
key_t	ftok __P((const char *, int));
__END_DECLS
#endif 
#endif 
#endif 

/*	BSDI un.h,v 2.6 1996/10/11 16:02:48 pjd Exp	*/

/*
 * Copyright (c) 1982, 1986, 1993
 *	The Regents of the University of California.  All rights reserved.
 *
 * Redistribution and use in source and binary forms, with or without
 * modification, are permitted provided that the following conditions
 * are met:
 * 1. Redistributions of source code must retain the above copyright
 *    notice, this list of conditions and the following disclaimer.
 * 2. Redistributions in binary form must reproduce the above copyright
 *    notice, this list of conditions and the following disclaimer in the
 *    documentation and/or other materials provided with the distribution.
 * 3. All advertising materials mentioning features or use of this software
 *    must display the following acknowledgement:
 *	This product includes software developed by the University of
 *	California, Berkeley and its contributors.
 * 4. Neither the name of the University nor the names of its contributors
 *    may be used to endorse or promote products derived from this software
 *    without specific prior written permission.
 *
 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
 * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
 * SUCH DAMAGE.
 *
 *	@(#)un.h	8.3 (Berkeley) 2/19/95
 */

#ifndef _SYS_UN_H_
#define _SYS_UN_H_

/*
 * Definitions for LOCAL IPC domain.
 */
struct	sockaddr_un {
	u_char	sun_len;		/* sockaddr len excluding null */
	u_char	sun_family;		/* AF_LOCAL */
	char	sun_path[104];		/* path name (gag) */
};

#define	LOCAL_CREDS	0x0001		/* get/set struct fcred structure */

#ifdef KERNEL
struct unpcb;

int	uipc_usrreq __P((struct socket *so, int req, struct mbuf *m,
		struct mbuf *nam, struct mbuf *control));
int	unp_attach __P((struct socket *so));
int	unp_bind __P((struct unpcb *unp, struct mbuf *nam, struct proc *p));
int	unp_connect __P((struct socket *so, struct mbuf *nam, struct proc *p));
int	unp_connect2 __P((struct socket *so, struct socket *so2));
void	unp_detach __P((struct unpcb *unp));
void	unp_discard __P((struct file *fp));
void	unp_disconnect __P((struct unpcb *unp));
void	unp_dispose __P((struct mbuf *m));
void	unp_drop __P((struct unpcb *unp, int errno));
void	unp_gc __P((void));
void	unp_mark __P((struct file *fp));
void	unp_scan __P((struct mbuf *m0, void (*op) __P((struct file *))));
void	unp_shutdown __P((struct unpcb *unp));
#else /* !KERNEL */

/* actual length of an initialized sockaddr_un */
#define SUN_LEN(su) \
	(sizeof(*(su)) - sizeof((su)->sun_path) + strlen((su)->sun_path))
#endif /* KERNEL */

#endif /* !_SYS_UN_H_ */



Attachments:

  [text/plain] /usr/include/sys/ipc.h (4.0K, ../../[email protected]/2-%2Fusr%2Finclude%2Fsys%2Fipc.h)
  download | inline:
/*-
 * Copyright (c) 1996 Berkeley Software Design, Inc. All rights reserved.
 * The Berkeley Software Design Inc. software License Agreement specifies
 * the terms and conditions for redistribution.
 *
 * BSDI ipc.h,v 2.4 1996/08/21 17:42:01 bostic Exp 
 */

/*
 * Copyright (c) 1988 University of Utah.
 * Copyright (c) 1990, 1993
 *	The Regents of the University of California.  All rights reserved.
 * (c) UNIX System Laboratories, Inc.
 * All or some portions of this file are derived from material licensed
 * to the University of California by American Telephone and Telegraph
 * Co. or Unix System Laboratories, Inc. and are reproduced herein with
 * the permission of UNIX System Laboratories, Inc.
 *
 * This code is derived from software contributed to Berkeley by
 * the Systems Programming Group of the University of Utah Computer
 * Science Department.
 *
 * Redistribution and use in source and binary forms, with or without
 * modification, are permitted provided that the following conditions
 * are met:
 * 1. Redistributions of source code must retain the above copyright
 *    notice, this list of conditions and the following disclaimer.
 * 2. Redistributions in binary form must reproduce the above copyright
 *    notice, this list of conditions and the following disclaimer in the
 *    documentation and/or other materials provided with the distribution.
 * 3. All advertising materials mentioning features or use of this software
 *    must display the following acknowledgement:
 *	This product includes software developed by the University of
 *	California, Berkeley and its contributors.
 * 4. Neither the name of the University nor the names of its contributors
 *    may be used to endorse or promote products derived from this software
 *    without specific prior written permission.
 *
 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
 * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
 * SUCH DAMAGE.
 *
 *	@(#)ipc.h	8.4 (Berkeley) 2/19/95
 */

/*
 * SVID compatible ipc.h file
 */
#ifndef _SYS_IPC_H_
#define _SYS_IPC_H_

#include <sys/cdefs.h>

struct ipc_perm {
	uid_t	uid;	/* user id */
	gid_t	gid;	/* group id */
	uid_t	cuid;	/* creator user id */
	gid_t	cgid;	/* creator group id */
	mode_t	mode;	/* r/w permission */
	u_long	seq;	/* sequence # (to generate unique msg/sem/shm id) */
	key_t	key;	/* user specified msg/sem/shm key */
};

/* common mode bits */
#define	IPC_R		000400	/* read permission */
#define	IPC_W		000200	/* write/alter permission */
#define	IPC_M		010000	/* permission to change control info */

/* SVID required constants (same values as system 5) */
#define	IPC_CREAT	001000	/* create entry if key does not exist */
#define	IPC_EXCL	002000	/* fail if key exists */
#define	IPC_NOWAIT	004000	/* error if request must wait */

#define	IPC_PRIVATE	(key_t)0 /* private key */

#define	IPC_RMID	0	/* remove identifier */
#define	IPC_SET		1	/* set options */
#define	IPC_STAT	2	/* get options */

#ifdef KERNEL
/* Macros to convert between ipc ids and array indices or sequence ids */
#define	IPCID_TO_IX(id)		((id) & 0xffff)
#define	IPCID_TO_SEQ(id)	(((id) >> 16) & 0xffff)
#define	IXSEQ_TO_IPCID(ix,perm)	(((perm.seq) << 16) | (ix & 0xffff))

__BEGIN_DECLS
int	ipcperm __P((struct ucred *,struct ipc_perm *,int));
__END_DECLS
#else 
/* Doesn't really belong but its traditionally here. */
#ifndef _POSIX_SOURCE
__BEGIN_DECLS
key_t	ftok __P((const char *, int));
__END_DECLS
#endif 
#endif 
#endif 

  [text/plain] /usr/include/sys/un.h (3.2K, ../../[email protected]/3-%2Fusr%2Finclude%2Fsys%2Fun.h)
  download | inline:
/*	BSDI un.h,v 2.6 1996/10/11 16:02:48 pjd Exp	*/

/*
 * Copyright (c) 1982, 1986, 1993
 *	The Regents of the University of California.  All rights reserved.
 *
 * Redistribution and use in source and binary forms, with or without
 * modification, are permitted provided that the following conditions
 * are met:
 * 1. Redistributions of source code must retain the above copyright
 *    notice, this list of conditions and the following disclaimer.
 * 2. Redistributions in binary form must reproduce the above copyright
 *    notice, this list of conditions and the following disclaimer in the
 *    documentation and/or other materials provided with the distribution.
 * 3. All advertising materials mentioning features or use of this software
 *    must display the following acknowledgement:
 *	This product includes software developed by the University of
 *	California, Berkeley and its contributors.
 * 4. Neither the name of the University nor the names of its contributors
 *    may be used to endorse or promote products derived from this software
 *    without specific prior written permission.
 *
 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
 * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
 * SUCH DAMAGE.
 *
 *	@(#)un.h	8.3 (Berkeley) 2/19/95
 */

#ifndef _SYS_UN_H_
#define _SYS_UN_H_

/*
 * Definitions for LOCAL IPC domain.
 */
struct	sockaddr_un {
	u_char	sun_len;		/* sockaddr len excluding null */
	u_char	sun_family;		/* AF_LOCAL */
	char	sun_path[104];		/* path name (gag) */
};

#define	LOCAL_CREDS	0x0001		/* get/set struct fcred structure */

#ifdef KERNEL
struct unpcb;

int	uipc_usrreq __P((struct socket *so, int req, struct mbuf *m,
		struct mbuf *nam, struct mbuf *control));
int	unp_attach __P((struct socket *so));
int	unp_bind __P((struct unpcb *unp, struct mbuf *nam, struct proc *p));
int	unp_connect __P((struct socket *so, struct mbuf *nam, struct proc *p));
int	unp_connect2 __P((struct socket *so, struct socket *so2));
void	unp_detach __P((struct unpcb *unp));
void	unp_discard __P((struct file *fp));
void	unp_disconnect __P((struct unpcb *unp));
void	unp_dispose __P((struct mbuf *m));
void	unp_drop __P((struct unpcb *unp, int errno));
void	unp_gc __P((void));
void	unp_mark __P((struct file *fp));
void	unp_scan __P((struct mbuf *m0, void (*op) __P((struct file *))));
void	unp_shutdown __P((struct unpcb *unp));
#else /* !KERNEL */

/* actual length of an initialized sockaddr_un */
#define SUN_LEN(su) \
	(sizeof(*(su)) - sizeof((su)->sun_path) + strlen((su)->sun_path))
#endif /* KERNEL */

#endif /* !_SYS_UN_H_ */

^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread

* [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation.
@ 2020-06-12 02:38 Andres Freund <[email protected]>
  0 siblings, 0 replies; 267+ messages in thread

From: Andres Freund @ 2020-06-12 02:38 UTC (permalink / raw)

---
 src/include/portability/instr_time.h | 68 ++++++++++++++++++++++++----
 1 file changed, 60 insertions(+), 8 deletions(-)

diff --git a/src/include/portability/instr_time.h b/src/include/portability/instr_time.h
index fc058d548a8..8b2f9a2e707 100644
--- a/src/include/portability/instr_time.h
+++ b/src/include/portability/instr_time.h
@@ -83,7 +83,9 @@
 #define PG_INSTR_CLOCK	CLOCK_REALTIME
 #endif
 
+/* time in baseline cpu cycles */
 typedef int64 instr_time;
+
 #define NS_PER_S INT64CONST(1000000000)
 #define US_PER_S INT64CONST(1000000)
 #define MS_PER_S INT64CONST(1000)
@@ -95,17 +97,67 @@ typedef int64 instr_time;
 
 #define INSTR_TIME_SET_ZERO(t)	((t) = 0)
 
-static inline instr_time pg_clock_gettime_ns(void)
+#include <x86intrin.h>
+#include <cpuid.h>
+
+/*
+ * Return what the number of cycles needs to be multiplied with to end up with
+ * seconds.
+ *
+ * FIXME: The cold portion should probably be out-of-line. And it'd be better
+ * to not recompute this in every file that uses this. Best would probably be
+ * to require explicit initialization of cycles_to_sec, because having a
+ * branch really is unnecessary.
+ *
+ * FIXME: We should probably not unnecessarily use floating point math
+ * here. And it's likely that the numbers are small enough that we are running
+ * into floating point inaccuracies already. Probably worthwhile to be a good
+ * bit smarter.
+ *
+ * FIXME: This would need to be conditional, with a fallback to something not
+ * rdtsc based.
+ */
+static inline double __attribute__((const))
+get_cycles_to_sec(void)
 {
-	struct timespec tmp;
+	static double cycles_to_sec = 0;
 
-	clock_gettime(PG_INSTR_CLOCK, &tmp);
+	/*
+	 * Compute baseline cpu peformance, determines speed at which rdtsc advances
+	 */
+	if (unlikely(cycles_to_sec == 0))
+	{
+		uint32 cpuinfo[4] = {0};
 
-	return tmp.tv_sec * NS_PER_S + tmp.tv_nsec;
+		__get_cpuid(0x16, cpuinfo, cpuinfo + 1, cpuinfo + 2, cpuinfo + 3);
+		cycles_to_sec = 1 / ((double) cpuinfo[0] * 1000 * 1000);
+	}
+
+	return cycles_to_sec;
+}
+
+static inline instr_time pg_clock_gettime_ref_cycles(void)
+{
+	/*
+	 * The rdtscp waits for all in-flight instructions to finish (but allows
+	 * later instructions to start concurrently). That's good for some timing
+	 * situations (when the time is supposed to cover all the work), but
+	 * terrible for others (when sub-parts of work are measured, because then
+	 * the pipeline stall due to the wait change the overall timing).
+	 */
+#if 0
+	unsigned int aux;
+	int64 tsc = __rdtscp(&aux);
+
+	return tsc;
+#else
+
+	return __rdtsc();
+#endif
 }
 
 #define INSTR_TIME_SET_CURRENT(t) \
-	(t) = pg_clock_gettime_ns()
+	(t) = pg_clock_gettime_ref_cycles()
 
 #define INSTR_TIME_ADD(x,y) \
 	do { \
@@ -123,13 +175,13 @@ static inline instr_time pg_clock_gettime_ns(void)
 	} while (0)
 
 #define INSTR_TIME_GET_DOUBLE(t) \
-	((double) (t) / NS_PER_S)
+	((double) (t) * get_cycles_to_sec())
 
 #define INSTR_TIME_GET_MILLISEC(t) \
-	((double) (t) / NS_PER_MS)
+	((double) (t) * (get_cycles_to_sec() * MS_PER_S))
 
 #define INSTR_TIME_GET_MICROSEC(t) \
-	((double) (t) / NS_PER_US)
+	((double) (t) * (get_cycles_to_sec() * US_PER_S))
 
 #else							/* !HAVE_CLOCK_GETTIME */
 
-- 
2.25.0.114.g5b0ca878e0


--wn4ncs637ccpvbhb--





^ permalink  raw  reply  [nested|flat] 267+ messages in thread


end of thread, other threads:[~2020-06-12 02:38 UTC | newest]

Thread overview: 267+ messages (download: mbox mbox.gz follow: Atom feed)
-- links below jump to the message on this page --
1998-09-10 18:21 Missing headers Windows NT port Joost Kraaijeveld <[email protected]>
1998-09-10 18:54 ` Bruce Momjian <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>
2020-06-12 02:38 [PATCH v1 2/2] WIP: Use cpu reference cycles, via rdtsc, to measure time for instrumentation. Andres Freund <[email protected]>

This inbox is served by agora; see mirroring instructions
for how to clone and mirror all data and code used for this inbox