pg.ddx.io pgsql-hackers@postgresql.org mailing list archive
help / color / mirror / Atom feedremove pg_spin_delay() from atomics code
4+ messages / 2 participants
[nested] [flat]
* remove pg_spin_delay() from atomics code
@ 2026-05-05 17:52 Nathan Bossart <nathandbossart@gmail.com>
2026-05-27 06:37 ` Re: remove pg_spin_delay() from atomics code Corey Huinker <corey.huinker@gmail.com>
0 siblings, 1 reply; 4+ messages in thread
From: Nathan Bossart @ 2026-05-05 17:52 UTC (permalink / raw)
To: pgsql-hackers
While inspecting what 32-bit optimizations we had, I noticed this line in
atomics/arch-x86.h:
#elif defined(_MSC_VER) && defined(__x86_64__)
...which led me to rediscover what Thomas Munro had found a while back [0],
i.e., we aren't using the right x86_64 macro for MSVC in lots of places.
Upon further inspection, all of this pg_spin_delay() stuff seems to have
been unused since it was added in commit b64d92f1a5, so I'm guessing it's
just an implementation artifact that didn't get cleaned up.
Here's a patch to remove that dead code.
[0] https://postgr.es/m/CA%2BhUKGL8Hs-phHPugrWM%3D5dAkcT897rXyazYzLw-Szxnzgx-rA%40mail.gmail.com
--
nathan
From 588685e23134c1916ff9640b7c9eca155434e346 Mon Sep 17 00:00:00 2001
From: Nathan Bossart <nathan@postgresql.org>
Date: Tue, 5 May 2026 12:37:50 -0500
Subject: [PATCH v1 1/1] remove pg_spin_delay()
---
src/include/port/atomics.h | 5 ---
src/include/port/atomics/arch-x86.h | 54 -----------------------------
src/include/port/atomics/generic.h | 6 ----
3 files changed, 65 deletions(-)
diff --git a/src/include/port/atomics.h b/src/include/port/atomics.h
index d8b1d20fe60..c50d95d29e2 100644
--- a/src/include/port/atomics.h
+++ b/src/include/port/atomics.h
@@ -154,11 +154,6 @@
#define pg_read_barrier() pg_read_barrier_impl()
#define pg_write_barrier() pg_write_barrier_impl()
-/*
- * Spinloop delay - Allow CPU to relax in busy loops
- */
-#define pg_spin_delay() pg_spin_delay_impl()
-
/*
* pg_atomic_init_flag - initialize atomic flag.
*
diff --git a/src/include/port/atomics/arch-x86.h b/src/include/port/atomics/arch-x86.h
index bd6f4f56ca2..b317cf7a188 100644
--- a/src/include/port/atomics/arch-x86.h
+++ b/src/include/port/atomics/arch-x86.h
@@ -76,60 +76,6 @@ typedef struct pg_atomic_uint64
} pg_atomic_uint64;
#endif /* __x86_64__ */
-#endif /* defined(__GNUC__) || defined(__INTEL_COMPILER) */
-
-#if !defined(PG_HAVE_SPIN_DELAY)
-/*
- * This sequence is equivalent to the PAUSE instruction ("rep" is
- * ignored by old IA32 processors if the following instruction is
- * not a string operation); the IA-32 Architecture Software
- * Developer's Manual, Vol. 3, Section 7.7.2 describes why using
- * PAUSE in the inner loop of a spin lock is necessary for good
- * performance:
- *
- * The PAUSE instruction improves the performance of IA-32
- * processors supporting Hyper-Threading Technology when
- * executing spin-wait loops and other routines where one
- * thread is accessing a shared lock or semaphore in a tight
- * polling loop. When executing a spin-wait loop, the
- * processor can suffer a severe performance penalty when
- * exiting the loop because it detects a possible memory order
- * violation and flushes the core processor's pipeline. The
- * PAUSE instruction provides a hint to the processor that the
- * code sequence is a spin-wait loop. The processor uses this
- * hint to avoid the memory order violation and prevent the
- * pipeline flush. In addition, the PAUSE instruction
- * de-pipelines the spin-wait loop to prevent it from
- * consuming execution resources excessively.
- */
-#if defined(__GNUC__) || defined(__INTEL_COMPILER)
-#define PG_HAVE_SPIN_DELAY
-static inline void
-pg_spin_delay_impl(void)
-{
- __asm__ __volatile__(" rep; nop \n");
-}
-#elif defined(_MSC_VER) && defined(__x86_64__)
-#define PG_HAVE_SPIN_DELAY
-static __forceinline void
-pg_spin_delay_impl(void)
-{
- _mm_pause();
-}
-#elif defined(_MSC_VER)
-#define PG_HAVE_SPIN_DELAY
-static __forceinline void
-pg_spin_delay_impl(void)
-{
- /* See comment for gcc code. Same code, MASM syntax */
- __asm rep nop;
-}
-#endif
-#endif /* !defined(PG_HAVE_SPIN_DELAY) */
-
-
-#if defined(__GNUC__) || defined(__INTEL_COMPILER)
-
#define PG_HAVE_ATOMIC_TEST_SET_FLAG
static inline bool
pg_atomic_test_set_flag_impl(volatile pg_atomic_flag *ptr)
diff --git a/src/include/port/atomics/generic.h b/src/include/port/atomics/generic.h
index fd64b6cbd86..daa772e9a6d 100644
--- a/src/include/port/atomics/generic.h
+++ b/src/include/port/atomics/generic.h
@@ -28,12 +28,6 @@
# define pg_write_barrier_impl pg_memory_barrier_impl
#endif
-#ifndef PG_HAVE_SPIN_DELAY
-#define PG_HAVE_SPIN_DELAY
-#define pg_spin_delay_impl() ((void)0)
-#endif
-
-
/* provide fallback */
#if !defined(PG_HAVE_ATOMIC_FLAG_SUPPORT) && defined(PG_HAVE_ATOMIC_U32_SUPPORT)
#define PG_HAVE_ATOMIC_FLAG_SUPPORT
--
2.50.1 (Apple Git-155)
Attachments:
[text/plain] v1-0001-remove-pg_spin_delay.patch (3.7K, ../../afouZUH_eUkIj4i4@nathan/2-v1-0001-remove-pg_spin_delay.patch)
download | inline diff:
From 588685e23134c1916ff9640b7c9eca155434e346 Mon Sep 17 00:00:00 2001
From: Nathan Bossart <nathan@postgresql.org>
Date: Tue, 5 May 2026 12:37:50 -0500
Subject: [PATCH v1 1/1] remove pg_spin_delay()
---
src/include/port/atomics.h | 5 ---
src/include/port/atomics/arch-x86.h | 54 -----------------------------
src/include/port/atomics/generic.h | 6 ----
3 files changed, 65 deletions(-)
diff --git a/src/include/port/atomics.h b/src/include/port/atomics.h
index d8b1d20fe60..c50d95d29e2 100644
--- a/src/include/port/atomics.h
+++ b/src/include/port/atomics.h
@@ -154,11 +154,6 @@
#define pg_read_barrier() pg_read_barrier_impl()
#define pg_write_barrier() pg_write_barrier_impl()
-/*
- * Spinloop delay - Allow CPU to relax in busy loops
- */
-#define pg_spin_delay() pg_spin_delay_impl()
-
/*
* pg_atomic_init_flag - initialize atomic flag.
*
diff --git a/src/include/port/atomics/arch-x86.h b/src/include/port/atomics/arch-x86.h
index bd6f4f56ca2..b317cf7a188 100644
--- a/src/include/port/atomics/arch-x86.h
+++ b/src/include/port/atomics/arch-x86.h
@@ -76,60 +76,6 @@ typedef struct pg_atomic_uint64
} pg_atomic_uint64;
#endif /* __x86_64__ */
-#endif /* defined(__GNUC__) || defined(__INTEL_COMPILER) */
-
-#if !defined(PG_HAVE_SPIN_DELAY)
-/*
- * This sequence is equivalent to the PAUSE instruction ("rep" is
- * ignored by old IA32 processors if the following instruction is
- * not a string operation); the IA-32 Architecture Software
- * Developer's Manual, Vol. 3, Section 7.7.2 describes why using
- * PAUSE in the inner loop of a spin lock is necessary for good
- * performance:
- *
- * The PAUSE instruction improves the performance of IA-32
- * processors supporting Hyper-Threading Technology when
- * executing spin-wait loops and other routines where one
- * thread is accessing a shared lock or semaphore in a tight
- * polling loop. When executing a spin-wait loop, the
- * processor can suffer a severe performance penalty when
- * exiting the loop because it detects a possible memory order
- * violation and flushes the core processor's pipeline. The
- * PAUSE instruction provides a hint to the processor that the
- * code sequence is a spin-wait loop. The processor uses this
- * hint to avoid the memory order violation and prevent the
- * pipeline flush. In addition, the PAUSE instruction
- * de-pipelines the spin-wait loop to prevent it from
- * consuming execution resources excessively.
- */
-#if defined(__GNUC__) || defined(__INTEL_COMPILER)
-#define PG_HAVE_SPIN_DELAY
-static inline void
-pg_spin_delay_impl(void)
-{
- __asm__ __volatile__(" rep; nop \n");
-}
-#elif defined(_MSC_VER) && defined(__x86_64__)
-#define PG_HAVE_SPIN_DELAY
-static __forceinline void
-pg_spin_delay_impl(void)
-{
- _mm_pause();
-}
-#elif defined(_MSC_VER)
-#define PG_HAVE_SPIN_DELAY
-static __forceinline void
-pg_spin_delay_impl(void)
-{
- /* See comment for gcc code. Same code, MASM syntax */
- __asm rep nop;
-}
-#endif
-#endif /* !defined(PG_HAVE_SPIN_DELAY) */
-
-
-#if defined(__GNUC__) || defined(__INTEL_COMPILER)
-
#define PG_HAVE_ATOMIC_TEST_SET_FLAG
static inline bool
pg_atomic_test_set_flag_impl(volatile pg_atomic_flag *ptr)
diff --git a/src/include/port/atomics/generic.h b/src/include/port/atomics/generic.h
index fd64b6cbd86..daa772e9a6d 100644
--- a/src/include/port/atomics/generic.h
+++ b/src/include/port/atomics/generic.h
@@ -28,12 +28,6 @@
# define pg_write_barrier_impl pg_memory_barrier_impl
#endif
-#ifndef PG_HAVE_SPIN_DELAY
-#define PG_HAVE_SPIN_DELAY
-#define pg_spin_delay_impl() ((void)0)
-#endif
-
-
/* provide fallback */
#if !defined(PG_HAVE_ATOMIC_FLAG_SUPPORT) && defined(PG_HAVE_ATOMIC_U32_SUPPORT)
#define PG_HAVE_ATOMIC_FLAG_SUPPORT
--
2.50.1 (Apple Git-155)
^ permalink raw reply [nested|flat] 4+ messages in thread
* Re: remove pg_spin_delay() from atomics code
2026-05-05 17:52 remove pg_spin_delay() from atomics code Nathan Bossart <nathandbossart@gmail.com>
@ 2026-05-27 06:37 ` Corey Huinker <corey.huinker@gmail.com>
2026-05-27 17:43 ` Re: remove pg_spin_delay() from atomics code Nathan Bossart <nathandbossart@gmail.com>
0 siblings, 1 reply; 4+ messages in thread
From: Corey Huinker @ 2026-05-27 06:37 UTC (permalink / raw)
To: Nathan Bossart <nathandbossart@gmail.com>; +Cc: pgsql-hackers
>
>
> Here's a patch to remove that dead code.
+1
The patch removes all 0 usages of pg_spin_delay, and all mentions of
pg_spin_delay.
All the associated comments were removed, no more, no less.
Deleted code is debugged code.
^ permalink raw reply [nested|flat] 4+ messages in thread
* Re: remove pg_spin_delay() from atomics code
2026-05-05 17:52 remove pg_spin_delay() from atomics code Nathan Bossart <nathandbossart@gmail.com>
2026-05-27 06:37 ` Re: remove pg_spin_delay() from atomics code Corey Huinker <corey.huinker@gmail.com>
@ 2026-05-27 17:43 ` Nathan Bossart <nathandbossart@gmail.com>
2026-06-30 15:58 ` Re: remove pg_spin_delay() from atomics code Nathan Bossart <nathandbossart@gmail.com>
0 siblings, 1 reply; 4+ messages in thread
From: Nathan Bossart @ 2026-05-27 17:43 UTC (permalink / raw)
To: Corey Huinker <corey.huinker@gmail.com>; +Cc: pgsql-hackers
On Wed, May 27, 2026 at 02:37:56AM -0400, Corey Huinker wrote:
> +1
>
> The patch removes all 0 usages of pg_spin_delay, and all mentions of
> pg_spin_delay.
>
> All the associated comments were removed, no more, no less.
>
> Deleted code is debugged code.
Thanks for looking. FWIW I'm planning to wait until v20 opens for
development before committing this. That's probably unnecessary, but I see
no advantage to rushing.
--
nathan
^ permalink raw reply [nested|flat] 4+ messages in thread
* Re: remove pg_spin_delay() from atomics code
2026-05-05 17:52 remove pg_spin_delay() from atomics code Nathan Bossart <nathandbossart@gmail.com>
2026-05-27 06:37 ` Re: remove pg_spin_delay() from atomics code Corey Huinker <corey.huinker@gmail.com>
2026-05-27 17:43 ` Re: remove pg_spin_delay() from atomics code Nathan Bossart <nathandbossart@gmail.com>
@ 2026-06-30 15:58 ` Nathan Bossart <nathandbossart@gmail.com>
0 siblings, 0 replies; 4+ messages in thread
From: Nathan Bossart @ 2026-06-30 15:58 UTC (permalink / raw)
To: Corey Huinker <corey.huinker@gmail.com>; +Cc: pgsql-hackers
Committed.
--
nathan
^ permalink raw reply [nested|flat] 4+ messages in thread
end of thread, other threads:[~2026-06-30 15:58 UTC | newest]
Thread overview: 4+ messages (download: mbox mbox.gz follow: Atom feed)
-- links below jump to the message on this page --
2026-05-05 17:52 remove pg_spin_delay() from atomics code Nathan Bossart <nathandbossart@gmail.com>
2026-05-27 06:37 ` Corey Huinker <corey.huinker@gmail.com>
2026-05-27 17:43 ` Nathan Bossart <nathandbossart@gmail.com>
2026-06-30 15:58 ` Nathan Bossart <nathandbossart@gmail.com>
This inbox is served by DDX for PostgreSQL; see mirroring instructions
for how to clone and mirror all data and code used for this inbox