public inbox for [email protected]  

help / color / mirror / Atom feed

[PATCH v7 3/3] Optimize hash function
50+ messages / 3 participants
[nested] [flat]

* [PATCH v6 3/3] Optimize hash function
@ 2021-01-20 22:06 Heikki Linnakangas <[email protected]>
  0 siblings, 0 replies; 50+ messages in thread

From: Heikki Linnakangas @ 2021-01-20 22:06 UTC (permalink / raw)

---
 src/backend/utils/resowner/resowner.c | 24 ++++++++++++++++++------
 src/include/common/hashfn.h           | 15 +++++++++++++++
 2 files changed, 33 insertions(+), 6 deletions(-)

diff --git a/src/backend/utils/resowner/resowner.c b/src/backend/utils/resowner/resowner.c
index 1db1e7cf14b..72e4d8a784c 100644
--- a/src/backend/utils/resowner/resowner.c
+++ b/src/backend/utils/resowner/resowner.c
@@ -163,15 +163,27 @@ static void ReleaseAuxProcessResourcesCallback(int code, Datum arg);
  *	  INTERNAL ROUTINES														 *
  *****************************************************************************/
 
+/*
+ * Hash function for value+kind combination.
+ */
 static inline uint32
 hash_resource_elem(Datum value, ResourceOwnerFuncs *kind)
 {
-	Datum		data[2];
-
-	data[0] = value;
-	data[1] = PointerGetDatum(kind);
-
-	return hash_bytes((unsigned char *) &data, 2 * SIZEOF_DATUM);
+	/*
+	 * Most resources types store a pointer in 'value', and pointers are
+	 * unique all on their own. But some resources store plain integers (Files
+	 * and Buffers as of this writing), so we want to incorporate the 'kind'
+	 * in the hash too, otherwise those resources will collide a lot. But
+	 * because there are only a few resource kinds like that - and only a few
+	 * resource kinds to begin with - we don't need to work too hard to mix
+	 * 'kind' into the hash. Just add it with hash_combine(), it perturbs the
+	 * result enough for our purposes.
+	 */
+#if SIZEOF_DATUM == 8
+	return hash_combine64(murmurhash64((uint64) value), (uint64) kind);
+#else
+	return hash_combine(murmurhash32((uint32) value), (uint32) kind);
+#endif
 }
 
 static void
diff --git a/src/include/common/hashfn.h b/src/include/common/hashfn.h
index c634cc067a1..009ffbbdd38 100644
--- a/src/include/common/hashfn.h
+++ b/src/include/common/hashfn.h
@@ -101,4 +101,19 @@ murmurhash32(uint32 data)
 	return h;
 }
 
+/* 64-bit variant */
+static inline uint64
+murmurhash64(uint64 data)
+{
+	uint64		h = data;
+
+	h ^= h >> 33;
+	h *= 0xff51afd7ed558ccd;
+	h ^= h >> 33;
+	h *= 0xc4ceb9fe1a85ec53;
+	h ^= h >> 33;
+
+	return h;
+}
+
 #endif							/* HASHFN_H */
-- 
2.29.2


--------------F30A5443A5D524A5CBC51AE7
Content-Type: text/x-patch; charset=UTF-8;
 name="v2-0001-resownerbench.patch"
Content-Transfer-Encoding: 7bit
Content-Disposition: attachment;
 filename="v2-0001-resownerbench.patch"



^ permalink  raw  reply  [nested|flat] 50+ messages in thread

* [PATCH v6 3/3] Optimize hash function
@ 2021-01-20 22:06 Heikki Linnakangas <[email protected]>
  0 siblings, 0 replies; 50+ messages in thread

From: Heikki Linnakangas @ 2021-01-20 22:06 UTC (permalink / raw)

---
 src/backend/utils/resowner/resowner.c | 24 ++++++++++++++++++------
 src/include/common/hashfn.h           | 15 +++++++++++++++
 2 files changed, 33 insertions(+), 6 deletions(-)

diff --git a/src/backend/utils/resowner/resowner.c b/src/backend/utils/resowner/resowner.c
index 1db1e7cf14b..72e4d8a784c 100644
--- a/src/backend/utils/resowner/resowner.c
+++ b/src/backend/utils/resowner/resowner.c
@@ -163,15 +163,27 @@ static void ReleaseAuxProcessResourcesCallback(int code, Datum arg);
  *	  INTERNAL ROUTINES														 *
  *****************************************************************************/
 
+/*
+ * Hash function for value+kind combination.
+ */
 static inline uint32
 hash_resource_elem(Datum value, ResourceOwnerFuncs *kind)
 {
-	Datum		data[2];
-
-	data[0] = value;
-	data[1] = PointerGetDatum(kind);
-
-	return hash_bytes((unsigned char *) &data, 2 * SIZEOF_DATUM);
+	/*
+	 * Most resources types store a pointer in 'value', and pointers are
+	 * unique all on their own. But some resources store plain integers (Files
+	 * and Buffers as of this writing), so we want to incorporate the 'kind'
+	 * in the hash too, otherwise those resources will collide a lot. But
+	 * because there are only a few resource kinds like that - and only a few
+	 * resource kinds to begin with - we don't need to work too hard to mix
+	 * 'kind' into the hash. Just add it with hash_combine(), it perturbs the
+	 * result enough for our purposes.
+	 */
+#if SIZEOF_DATUM == 8
+	return hash_combine64(murmurhash64((uint64) value), (uint64) kind);
+#else
+	return hash_combine(murmurhash32((uint32) value), (uint32) kind);
+#endif
 }
 
 static void
diff --git a/src/include/common/hashfn.h b/src/include/common/hashfn.h
index c634cc067a1..009ffbbdd38 100644
--- a/src/include/common/hashfn.h
+++ b/src/include/common/hashfn.h
@@ -101,4 +101,19 @@ murmurhash32(uint32 data)
 	return h;
 }
 
+/* 64-bit variant */
+static inline uint64
+murmurhash64(uint64 data)
+{
+	uint64		h = data;
+
+	h ^= h >> 33;
+	h *= 0xff51afd7ed558ccd;
+	h ^= h >> 33;
+	h *= 0xc4ceb9fe1a85ec53;
+	h ^= h >> 33;
+
+	return h;
+}
+
 #endif							/* HASHFN_H */
-- 
2.29.2


--------------F30A5443A5D524A5CBC51AE7
Content-Type: text/x-patch; charset=UTF-8;
 name="v2-0001-resownerbench.patch"
Content-Transfer-Encoding: 7bit
Content-Disposition: attachment;
 filename="v2-0001-resownerbench.patch"



^ permalink  raw  reply  [nested|flat] 50+ messages in thread

* [PATCH v6 3/3] Optimize hash function
@ 2021-01-20 22:06 Heikki Linnakangas <[email protected]>
  0 siblings, 0 replies; 50+ messages in thread

From: Heikki Linnakangas @ 2021-01-20 22:06 UTC (permalink / raw)

---
 src/backend/utils/resowner/resowner.c | 24 ++++++++++++++++++------
 src/include/common/hashfn.h           | 15 +++++++++++++++
 2 files changed, 33 insertions(+), 6 deletions(-)

diff --git a/src/backend/utils/resowner/resowner.c b/src/backend/utils/resowner/resowner.c
index 1db1e7cf14b..72e4d8a784c 100644
--- a/src/backend/utils/resowner/resowner.c
+++ b/src/backend/utils/resowner/resowner.c
@@ -163,15 +163,27 @@ static void ReleaseAuxProcessResourcesCallback(int code, Datum arg);
  *	  INTERNAL ROUTINES														 *
  *****************************************************************************/
 
+/*
+ * Hash function for value+kind combination.
+ */
 static inline uint32
 hash_resource_elem(Datum value, ResourceOwnerFuncs *kind)
 {
-	Datum		data[2];
-
-	data[0] = value;
-	data[1] = PointerGetDatum(kind);
-
-	return hash_bytes((unsigned char *) &data, 2 * SIZEOF_DATUM);
+	/*
+	 * Most resources types store a pointer in 'value', and pointers are
+	 * unique all on their own. But some resources store plain integers (Files
+	 * and Buffers as of this writing), so we want to incorporate the 'kind'
+	 * in the hash too, otherwise those resources will collide a lot. But
+	 * because there are only a few resource kinds like that - and only a few
+	 * resource kinds to begin with - we don't need to work too hard to mix
+	 * 'kind' into the hash. Just add it with hash_combine(), it perturbs the
+	 * result enough for our purposes.
+	 */
+#if SIZEOF_DATUM == 8
+	return hash_combine64(murmurhash64((uint64) value), (uint64) kind);
+#else
+	return hash_combine(murmurhash32((uint32) value), (uint32) kind);
+#endif
 }
 
 static void
diff --git a/src/include/common/hashfn.h b/src/include/common/hashfn.h
index c634cc067a1..009ffbbdd38 100644
--- a/src/include/common/hashfn.h
+++ b/src/include/common/hashfn.h
@@ -101,4 +101,19 @@ murmurhash32(uint32 data)
 	return h;
 }
 
+/* 64-bit variant */
+static inline uint64
+murmurhash64(uint64 data)
+{
+	uint64		h = data;
+
+	h ^= h >> 33;
+	h *= 0xff51afd7ed558ccd;
+	h ^= h >> 33;
+	h *= 0xc4ceb9fe1a85ec53;
+	h ^= h >> 33;
+
+	return h;
+}
+
 #endif							/* HASHFN_H */
-- 
2.29.2


--------------F30A5443A5D524A5CBC51AE7
Content-Type: text/x-patch; charset=UTF-8;
 name="v2-0001-resownerbench.patch"
Content-Transfer-Encoding: 7bit
Content-Disposition: attachment;
 filename="v2-0001-resownerbench.patch"



^ permalink  raw  reply  [nested|flat] 50+ messages in thread

* [PATCH v7 3/3] Optimize hash function
@ 2021-01-20 22:06 Heikki Linnakangas <[email protected]>
  0 siblings, 0 replies; 50+ messages in thread

From: Heikki Linnakangas @ 2021-01-20 22:06 UTC (permalink / raw)

---
 src/backend/utils/resowner/resowner.c | 24 ++++++++++++++++++------
 src/include/common/hashfn.h           | 15 +++++++++++++++
 2 files changed, 33 insertions(+), 6 deletions(-)

diff --git a/src/backend/utils/resowner/resowner.c b/src/backend/utils/resowner/resowner.c
index d62b1495dd3..6189d051072 100644
--- a/src/backend/utils/resowner/resowner.c
+++ b/src/backend/utils/resowner/resowner.c
@@ -163,15 +163,27 @@ static void ReleaseAuxProcessResourcesCallback(int code, Datum arg);
  *	  INTERNAL ROUTINES														 *
  *****************************************************************************/
 
+/*
+ * Hash function for value+kind combination.
+ */
 static inline uint32
 hash_resource_elem(Datum value, ResourceOwnerFuncs *kind)
 {
-	Datum		data[2];
-
-	data[0] = value;
-	data[1] = PointerGetDatum(kind);
-
-	return hash_bytes((unsigned char *) &data, 2 * SIZEOF_DATUM);
+	/*
+	 * Most resources types store a pointer in 'value', and pointers are
+	 * unique all on their own. But some resources store plain integers (Files
+	 * and Buffers as of this writing), so we want to incorporate the 'kind'
+	 * in the hash too, otherwise those resources will collide a lot. But
+	 * because there are only a few resource kinds like that - and only a few
+	 * resource kinds to begin with - we don't need to work too hard to mix
+	 * 'kind' into the hash. Just add it with hash_combine(), it perturbs the
+	 * result enough for our purposes.
+	 */
+#if SIZEOF_DATUM == 8
+	return hash_combine64(murmurhash64((uint64) value), (uint64) kind);
+#else
+	return hash_combine(murmurhash32((uint32) value), (uint32) kind);
+#endif
 }
 
 static void
diff --git a/src/include/common/hashfn.h b/src/include/common/hashfn.h
index c634cc067a1..009ffbbdd38 100644
--- a/src/include/common/hashfn.h
+++ b/src/include/common/hashfn.h
@@ -101,4 +101,19 @@ murmurhash32(uint32 data)
 	return h;
 }
 
+/* 64-bit variant */
+static inline uint64
+murmurhash64(uint64 data)
+{
+	uint64		h = data;
+
+	h ^= h >> 33;
+	h *= 0xff51afd7ed558ccd;
+	h ^= h >> 33;
+	h *= 0xc4ceb9fe1a85ec53;
+	h ^= h >> 33;
+
+	return h;
+}
+
 #endif							/* HASHFN_H */
-- 
2.30.1


--------------C6A03312D0F99F9183A0DF46--





^ permalink  raw  reply  [nested|flat] 50+ messages in thread

* [PATCH v7 3/3] Optimize hash function
@ 2021-01-20 22:06 Heikki Linnakangas <[email protected]>
  0 siblings, 0 replies; 50+ messages in thread

From: Heikki Linnakangas @ 2021-01-20 22:06 UTC (permalink / raw)

---
 src/backend/utils/resowner/resowner.c | 24 ++++++++++++++++++------
 src/include/common/hashfn.h           | 15 +++++++++++++++
 2 files changed, 33 insertions(+), 6 deletions(-)

diff --git a/src/backend/utils/resowner/resowner.c b/src/backend/utils/resowner/resowner.c
index d62b1495dd3..6189d051072 100644
--- a/src/backend/utils/resowner/resowner.c
+++ b/src/backend/utils/resowner/resowner.c
@@ -163,15 +163,27 @@ static void ReleaseAuxProcessResourcesCallback(int code, Datum arg);
  *	  INTERNAL ROUTINES														 *
  *****************************************************************************/
 
+/*
+ * Hash function for value+kind combination.
+ */
 static inline uint32
 hash_resource_elem(Datum value, ResourceOwnerFuncs *kind)
 {
-	Datum		data[2];
-
-	data[0] = value;
-	data[1] = PointerGetDatum(kind);
-
-	return hash_bytes((unsigned char *) &data, 2 * SIZEOF_DATUM);
+	/*
+	 * Most resources types store a pointer in 'value', and pointers are
+	 * unique all on their own. But some resources store plain integers (Files
+	 * and Buffers as of this writing), so we want to incorporate the 'kind'
+	 * in the hash too, otherwise those resources will collide a lot. But
+	 * because there are only a few resource kinds like that - and only a few
+	 * resource kinds to begin with - we don't need to work too hard to mix
+	 * 'kind' into the hash. Just add it with hash_combine(), it perturbs the
+	 * result enough for our purposes.
+	 */
+#if SIZEOF_DATUM == 8
+	return hash_combine64(murmurhash64((uint64) value), (uint64) kind);
+#else
+	return hash_combine(murmurhash32((uint32) value), (uint32) kind);
+#endif
 }
 
 static void
diff --git a/src/include/common/hashfn.h b/src/include/common/hashfn.h
index c634cc067a1..009ffbbdd38 100644
--- a/src/include/common/hashfn.h
+++ b/src/include/common/hashfn.h
@@ -101,4 +101,19 @@ murmurhash32(uint32 data)
 	return h;
 }
 
+/* 64-bit variant */
+static inline uint64
+murmurhash64(uint64 data)
+{
+	uint64		h = data;
+
+	h ^= h >> 33;
+	h *= 0xff51afd7ed558ccd;
+	h ^= h >> 33;
+	h *= 0xc4ceb9fe1a85ec53;
+	h ^= h >> 33;
+
+	return h;
+}
+
 #endif							/* HASHFN_H */
-- 
2.30.1


--------------C6A03312D0F99F9183A0DF46--





^ permalink  raw  reply  [nested|flat] 50+ messages in thread

* [PATCH v6 3/3] Optimize hash function
@ 2021-01-20 22:06 Heikki Linnakangas <[email protected]>
  0 siblings, 0 replies; 50+ messages in thread

From: Heikki Linnakangas @ 2021-01-20 22:06 UTC (permalink / raw)

---
 src/backend/utils/resowner/resowner.c | 24 ++++++++++++++++++------
 src/include/common/hashfn.h           | 15 +++++++++++++++
 2 files changed, 33 insertions(+), 6 deletions(-)

diff --git a/src/backend/utils/resowner/resowner.c b/src/backend/utils/resowner/resowner.c
index 1db1e7cf14b..72e4d8a784c 100644
--- a/src/backend/utils/resowner/resowner.c
+++ b/src/backend/utils/resowner/resowner.c
@@ -163,15 +163,27 @@ static void ReleaseAuxProcessResourcesCallback(int code, Datum arg);
  *	  INTERNAL ROUTINES														 *
  *****************************************************************************/
 
+/*
+ * Hash function for value+kind combination.
+ */
 static inline uint32
 hash_resource_elem(Datum value, ResourceOwnerFuncs *kind)
 {
-	Datum		data[2];
-
-	data[0] = value;
-	data[1] = PointerGetDatum(kind);
-
-	return hash_bytes((unsigned char *) &data, 2 * SIZEOF_DATUM);
+	/*
+	 * Most resources types store a pointer in 'value', and pointers are
+	 * unique all on their own. But some resources store plain integers (Files
+	 * and Buffers as of this writing), so we want to incorporate the 'kind'
+	 * in the hash too, otherwise those resources will collide a lot. But
+	 * because there are only a few resource kinds like that - and only a few
+	 * resource kinds to begin with - we don't need to work too hard to mix
+	 * 'kind' into the hash. Just add it with hash_combine(), it perturbs the
+	 * result enough for our purposes.
+	 */
+#if SIZEOF_DATUM == 8
+	return hash_combine64(murmurhash64((uint64) value), (uint64) kind);
+#else
+	return hash_combine(murmurhash32((uint32) value), (uint32) kind);
+#endif
 }
 
 static void
diff --git a/src/include/common/hashfn.h b/src/include/common/hashfn.h
index c634cc067a1..009ffbbdd38 100644
--- a/src/include/common/hashfn.h
+++ b/src/include/common/hashfn.h
@@ -101,4 +101,19 @@ murmurhash32(uint32 data)
 	return h;
 }
 
+/* 64-bit variant */
+static inline uint64
+murmurhash64(uint64 data)
+{
+	uint64		h = data;
+
+	h ^= h >> 33;
+	h *= 0xff51afd7ed558ccd;
+	h ^= h >> 33;
+	h *= 0xc4ceb9fe1a85ec53;
+	h ^= h >> 33;
+
+	return h;
+}
+
 #endif							/* HASHFN_H */
-- 
2.29.2


--------------F30A5443A5D524A5CBC51AE7
Content-Type: text/x-patch; charset=UTF-8;
 name="v2-0001-resownerbench.patch"
Content-Transfer-Encoding: 7bit
Content-Disposition: attachment;
 filename="v2-0001-resownerbench.patch"



^ permalink  raw  reply  [nested|flat] 50+ messages in thread

* [PATCH v7 3/3] Optimize hash function
@ 2021-01-20 22:06 Heikki Linnakangas <[email protected]>
  0 siblings, 0 replies; 50+ messages in thread

From: Heikki Linnakangas @ 2021-01-20 22:06 UTC (permalink / raw)

---
 src/backend/utils/resowner/resowner.c | 24 ++++++++++++++++++------
 src/include/common/hashfn.h           | 15 +++++++++++++++
 2 files changed, 33 insertions(+), 6 deletions(-)

diff --git a/src/backend/utils/resowner/resowner.c b/src/backend/utils/resowner/resowner.c
index d62b1495dd3..6189d051072 100644
--- a/src/backend/utils/resowner/resowner.c
+++ b/src/backend/utils/resowner/resowner.c
@@ -163,15 +163,27 @@ static void ReleaseAuxProcessResourcesCallback(int code, Datum arg);
  *	  INTERNAL ROUTINES														 *
  *****************************************************************************/
 
+/*
+ * Hash function for value+kind combination.
+ */
 static inline uint32
 hash_resource_elem(Datum value, ResourceOwnerFuncs *kind)
 {
-	Datum		data[2];
-
-	data[0] = value;
-	data[1] = PointerGetDatum(kind);
-
-	return hash_bytes((unsigned char *) &data, 2 * SIZEOF_DATUM);
+	/*
+	 * Most resources types store a pointer in 'value', and pointers are
+	 * unique all on their own. But some resources store plain integers (Files
+	 * and Buffers as of this writing), so we want to incorporate the 'kind'
+	 * in the hash too, otherwise those resources will collide a lot. But
+	 * because there are only a few resource kinds like that - and only a few
+	 * resource kinds to begin with - we don't need to work too hard to mix
+	 * 'kind' into the hash. Just add it with hash_combine(), it perturbs the
+	 * result enough for our purposes.
+	 */
+#if SIZEOF_DATUM == 8
+	return hash_combine64(murmurhash64((uint64) value), (uint64) kind);
+#else
+	return hash_combine(murmurhash32((uint32) value), (uint32) kind);
+#endif
 }
 
 static void
diff --git a/src/include/common/hashfn.h b/src/include/common/hashfn.h
index c634cc067a1..009ffbbdd38 100644
--- a/src/include/common/hashfn.h
+++ b/src/include/common/hashfn.h
@@ -101,4 +101,19 @@ murmurhash32(uint32 data)
 	return h;
 }
 
+/* 64-bit variant */
+static inline uint64
+murmurhash64(uint64 data)
+{
+	uint64		h = data;
+
+	h ^= h >> 33;
+	h *= 0xff51afd7ed558ccd;
+	h ^= h >> 33;
+	h *= 0xc4ceb9fe1a85ec53;
+	h ^= h >> 33;
+
+	return h;
+}
+
 #endif							/* HASHFN_H */
-- 
2.30.1


--------------C6A03312D0F99F9183A0DF46--





^ permalink  raw  reply  [nested|flat] 50+ messages in thread

* [PATCH v7 3/3] Optimize hash function
@ 2021-01-20 22:06 Heikki Linnakangas <[email protected]>
  0 siblings, 0 replies; 50+ messages in thread

From: Heikki Linnakangas @ 2021-01-20 22:06 UTC (permalink / raw)

---
 src/backend/utils/resowner/resowner.c | 24 ++++++++++++++++++------
 src/include/common/hashfn.h           | 15 +++++++++++++++
 2 files changed, 33 insertions(+), 6 deletions(-)

diff --git a/src/backend/utils/resowner/resowner.c b/src/backend/utils/resowner/resowner.c
index d62b1495dd3..6189d051072 100644
--- a/src/backend/utils/resowner/resowner.c
+++ b/src/backend/utils/resowner/resowner.c
@@ -163,15 +163,27 @@ static void ReleaseAuxProcessResourcesCallback(int code, Datum arg);
  *	  INTERNAL ROUTINES														 *
  *****************************************************************************/
 
+/*
+ * Hash function for value+kind combination.
+ */
 static inline uint32
 hash_resource_elem(Datum value, ResourceOwnerFuncs *kind)
 {
-	Datum		data[2];
-
-	data[0] = value;
-	data[1] = PointerGetDatum(kind);
-
-	return hash_bytes((unsigned char *) &data, 2 * SIZEOF_DATUM);
+	/*
+	 * Most resources types store a pointer in 'value', and pointers are
+	 * unique all on their own. But some resources store plain integers (Files
+	 * and Buffers as of this writing), so we want to incorporate the 'kind'
+	 * in the hash too, otherwise those resources will collide a lot. But
+	 * because there are only a few resource kinds like that - and only a few
+	 * resource kinds to begin with - we don't need to work too hard to mix
+	 * 'kind' into the hash. Just add it with hash_combine(), it perturbs the
+	 * result enough for our purposes.
+	 */
+#if SIZEOF_DATUM == 8
+	return hash_combine64(murmurhash64((uint64) value), (uint64) kind);
+#else
+	return hash_combine(murmurhash32((uint32) value), (uint32) kind);
+#endif
 }
 
 static void
diff --git a/src/include/common/hashfn.h b/src/include/common/hashfn.h
index c634cc067a1..009ffbbdd38 100644
--- a/src/include/common/hashfn.h
+++ b/src/include/common/hashfn.h
@@ -101,4 +101,19 @@ murmurhash32(uint32 data)
 	return h;
 }
 
+/* 64-bit variant */
+static inline uint64
+murmurhash64(uint64 data)
+{
+	uint64		h = data;
+
+	h ^= h >> 33;
+	h *= 0xff51afd7ed558ccd;
+	h ^= h >> 33;
+	h *= 0xc4ceb9fe1a85ec53;
+	h ^= h >> 33;
+
+	return h;
+}
+
 #endif							/* HASHFN_H */
-- 
2.30.1


--------------C6A03312D0F99F9183A0DF46--





^ permalink  raw  reply  [nested|flat] 50+ messages in thread

* [PATCH v7 3/3] Optimize hash function
@ 2021-01-20 22:06 Heikki Linnakangas <[email protected]>
  0 siblings, 0 replies; 50+ messages in thread

From: Heikki Linnakangas @ 2021-01-20 22:06 UTC (permalink / raw)

---
 src/backend/utils/resowner/resowner.c | 24 ++++++++++++++++++------
 src/include/common/hashfn.h           | 15 +++++++++++++++
 2 files changed, 33 insertions(+), 6 deletions(-)

diff --git a/src/backend/utils/resowner/resowner.c b/src/backend/utils/resowner/resowner.c
index d62b1495dd3..6189d051072 100644
--- a/src/backend/utils/resowner/resowner.c
+++ b/src/backend/utils/resowner/resowner.c
@@ -163,15 +163,27 @@ static void ReleaseAuxProcessResourcesCallback(int code, Datum arg);
  *	  INTERNAL ROUTINES														 *
  *****************************************************************************/
 
+/*
+ * Hash function for value+kind combination.
+ */
 static inline uint32
 hash_resource_elem(Datum value, ResourceOwnerFuncs *kind)
 {
-	Datum		data[2];
-
-	data[0] = value;
-	data[1] = PointerGetDatum(kind);
-
-	return hash_bytes((unsigned char *) &data, 2 * SIZEOF_DATUM);
+	/*
+	 * Most resources types store a pointer in 'value', and pointers are
+	 * unique all on their own. But some resources store plain integers (Files
+	 * and Buffers as of this writing), so we want to incorporate the 'kind'
+	 * in the hash too, otherwise those resources will collide a lot. But
+	 * because there are only a few resource kinds like that - and only a few
+	 * resource kinds to begin with - we don't need to work too hard to mix
+	 * 'kind' into the hash. Just add it with hash_combine(), it perturbs the
+	 * result enough for our purposes.
+	 */
+#if SIZEOF_DATUM == 8
+	return hash_combine64(murmurhash64((uint64) value), (uint64) kind);
+#else
+	return hash_combine(murmurhash32((uint32) value), (uint32) kind);
+#endif
 }
 
 static void
diff --git a/src/include/common/hashfn.h b/src/include/common/hashfn.h
index c634cc067a1..009ffbbdd38 100644
--- a/src/include/common/hashfn.h
+++ b/src/include/common/hashfn.h
@@ -101,4 +101,19 @@ murmurhash32(uint32 data)
 	return h;
 }
 
+/* 64-bit variant */
+static inline uint64
+murmurhash64(uint64 data)
+{
+	uint64		h = data;
+
+	h ^= h >> 33;
+	h *= 0xff51afd7ed558ccd;
+	h ^= h >> 33;
+	h *= 0xc4ceb9fe1a85ec53;
+	h ^= h >> 33;
+
+	return h;
+}
+
 #endif							/* HASHFN_H */
-- 
2.30.1


--------------C6A03312D0F99F9183A0DF46--





^ permalink  raw  reply  [nested|flat] 50+ messages in thread

* [PATCH v6 3/3] Optimize hash function
@ 2021-01-20 22:06 Heikki Linnakangas <[email protected]>
  0 siblings, 0 replies; 50+ messages in thread

From: Heikki Linnakangas @ 2021-01-20 22:06 UTC (permalink / raw)

---
 src/backend/utils/resowner/resowner.c | 24 ++++++++++++++++++------
 src/include/common/hashfn.h           | 15 +++++++++++++++
 2 files changed, 33 insertions(+), 6 deletions(-)

diff --git a/src/backend/utils/resowner/resowner.c b/src/backend/utils/resowner/resowner.c
index 1db1e7cf14b..72e4d8a784c 100644
--- a/src/backend/utils/resowner/resowner.c
+++ b/src/backend/utils/resowner/resowner.c
@@ -163,15 +163,27 @@ static void ReleaseAuxProcessResourcesCallback(int code, Datum arg);
  *	  INTERNAL ROUTINES														 *
  *****************************************************************************/
 
+/*
+ * Hash function for value+kind combination.
+ */
 static inline uint32
 hash_resource_elem(Datum value, ResourceOwnerFuncs *kind)
 {
-	Datum		data[2];
-
-	data[0] = value;
-	data[1] = PointerGetDatum(kind);
-
-	return hash_bytes((unsigned char *) &data, 2 * SIZEOF_DATUM);
+	/*
+	 * Most resources types store a pointer in 'value', and pointers are
+	 * unique all on their own. But some resources store plain integers (Files
+	 * and Buffers as of this writing), so we want to incorporate the 'kind'
+	 * in the hash too, otherwise those resources will collide a lot. But
+	 * because there are only a few resource kinds like that - and only a few
+	 * resource kinds to begin with - we don't need to work too hard to mix
+	 * 'kind' into the hash. Just add it with hash_combine(), it perturbs the
+	 * result enough for our purposes.
+	 */
+#if SIZEOF_DATUM == 8
+	return hash_combine64(murmurhash64((uint64) value), (uint64) kind);
+#else
+	return hash_combine(murmurhash32((uint32) value), (uint32) kind);
+#endif
 }
 
 static void
diff --git a/src/include/common/hashfn.h b/src/include/common/hashfn.h
index c634cc067a1..009ffbbdd38 100644
--- a/src/include/common/hashfn.h
+++ b/src/include/common/hashfn.h
@@ -101,4 +101,19 @@ murmurhash32(uint32 data)
 	return h;
 }
 
+/* 64-bit variant */
+static inline uint64
+murmurhash64(uint64 data)
+{
+	uint64		h = data;
+
+	h ^= h >> 33;
+	h *= 0xff51afd7ed558ccd;
+	h ^= h >> 33;
+	h *= 0xc4ceb9fe1a85ec53;
+	h ^= h >> 33;
+
+	return h;
+}
+
 #endif							/* HASHFN_H */
-- 
2.29.2


--------------F30A5443A5D524A5CBC51AE7
Content-Type: text/x-patch; charset=UTF-8;
 name="v2-0001-resownerbench.patch"
Content-Transfer-Encoding: 7bit
Content-Disposition: attachment;
 filename="v2-0001-resownerbench.patch"



^ permalink  raw  reply  [nested|flat] 50+ messages in thread

* [PATCH v7 3/3] Optimize hash function
@ 2021-01-20 22:06 Heikki Linnakangas <[email protected]>
  0 siblings, 0 replies; 50+ messages in thread

From: Heikki Linnakangas @ 2021-01-20 22:06 UTC (permalink / raw)

---
 src/backend/utils/resowner/resowner.c | 24 ++++++++++++++++++------
 src/include/common/hashfn.h           | 15 +++++++++++++++
 2 files changed, 33 insertions(+), 6 deletions(-)

diff --git a/src/backend/utils/resowner/resowner.c b/src/backend/utils/resowner/resowner.c
index d62b1495dd3..6189d051072 100644
--- a/src/backend/utils/resowner/resowner.c
+++ b/src/backend/utils/resowner/resowner.c
@@ -163,15 +163,27 @@ static void ReleaseAuxProcessResourcesCallback(int code, Datum arg);
  *	  INTERNAL ROUTINES														 *
  *****************************************************************************/
 
+/*
+ * Hash function for value+kind combination.
+ */
 static inline uint32
 hash_resource_elem(Datum value, ResourceOwnerFuncs *kind)
 {
-	Datum		data[2];
-
-	data[0] = value;
-	data[1] = PointerGetDatum(kind);
-
-	return hash_bytes((unsigned char *) &data, 2 * SIZEOF_DATUM);
+	/*
+	 * Most resources types store a pointer in 'value', and pointers are
+	 * unique all on their own. But some resources store plain integers (Files
+	 * and Buffers as of this writing), so we want to incorporate the 'kind'
+	 * in the hash too, otherwise those resources will collide a lot. But
+	 * because there are only a few resource kinds like that - and only a few
+	 * resource kinds to begin with - we don't need to work too hard to mix
+	 * 'kind' into the hash. Just add it with hash_combine(), it perturbs the
+	 * result enough for our purposes.
+	 */
+#if SIZEOF_DATUM == 8
+	return hash_combine64(murmurhash64((uint64) value), (uint64) kind);
+#else
+	return hash_combine(murmurhash32((uint32) value), (uint32) kind);
+#endif
 }
 
 static void
diff --git a/src/include/common/hashfn.h b/src/include/common/hashfn.h
index c634cc067a1..009ffbbdd38 100644
--- a/src/include/common/hashfn.h
+++ b/src/include/common/hashfn.h
@@ -101,4 +101,19 @@ murmurhash32(uint32 data)
 	return h;
 }
 
+/* 64-bit variant */
+static inline uint64
+murmurhash64(uint64 data)
+{
+	uint64		h = data;
+
+	h ^= h >> 33;
+	h *= 0xff51afd7ed558ccd;
+	h ^= h >> 33;
+	h *= 0xc4ceb9fe1a85ec53;
+	h ^= h >> 33;
+
+	return h;
+}
+
 #endif							/* HASHFN_H */
-- 
2.30.1


--------------C6A03312D0F99F9183A0DF46--





^ permalink  raw  reply  [nested|flat] 50+ messages in thread

* [PATCH v7 3/3] Optimize hash function
@ 2021-01-20 22:06 Heikki Linnakangas <[email protected]>
  0 siblings, 0 replies; 50+ messages in thread

From: Heikki Linnakangas @ 2021-01-20 22:06 UTC (permalink / raw)

---
 src/backend/utils/resowner/resowner.c | 24 ++++++++++++++++++------
 src/include/common/hashfn.h           | 15 +++++++++++++++
 2 files changed, 33 insertions(+), 6 deletions(-)

diff --git a/src/backend/utils/resowner/resowner.c b/src/backend/utils/resowner/resowner.c
index d62b1495dd3..6189d051072 100644
--- a/src/backend/utils/resowner/resowner.c
+++ b/src/backend/utils/resowner/resowner.c
@@ -163,15 +163,27 @@ static void ReleaseAuxProcessResourcesCallback(int code, Datum arg);
  *	  INTERNAL ROUTINES														 *
  *****************************************************************************/
 
+/*
+ * Hash function for value+kind combination.
+ */
 static inline uint32
 hash_resource_elem(Datum value, ResourceOwnerFuncs *kind)
 {
-	Datum		data[2];
-
-	data[0] = value;
-	data[1] = PointerGetDatum(kind);
-
-	return hash_bytes((unsigned char *) &data, 2 * SIZEOF_DATUM);
+	/*
+	 * Most resources types store a pointer in 'value', and pointers are
+	 * unique all on their own. But some resources store plain integers (Files
+	 * and Buffers as of this writing), so we want to incorporate the 'kind'
+	 * in the hash too, otherwise those resources will collide a lot. But
+	 * because there are only a few resource kinds like that - and only a few
+	 * resource kinds to begin with - we don't need to work too hard to mix
+	 * 'kind' into the hash. Just add it with hash_combine(), it perturbs the
+	 * result enough for our purposes.
+	 */
+#if SIZEOF_DATUM == 8
+	return hash_combine64(murmurhash64((uint64) value), (uint64) kind);
+#else
+	return hash_combine(murmurhash32((uint32) value), (uint32) kind);
+#endif
 }
 
 static void
diff --git a/src/include/common/hashfn.h b/src/include/common/hashfn.h
index c634cc067a1..009ffbbdd38 100644
--- a/src/include/common/hashfn.h
+++ b/src/include/common/hashfn.h
@@ -101,4 +101,19 @@ murmurhash32(uint32 data)
 	return h;
 }
 
+/* 64-bit variant */
+static inline uint64
+murmurhash64(uint64 data)
+{
+	uint64		h = data;
+
+	h ^= h >> 33;
+	h *= 0xff51afd7ed558ccd;
+	h ^= h >> 33;
+	h *= 0xc4ceb9fe1a85ec53;
+	h ^= h >> 33;
+
+	return h;
+}
+
 #endif							/* HASHFN_H */
-- 
2.30.1


--------------C6A03312D0F99F9183A0DF46--





^ permalink  raw  reply  [nested|flat] 50+ messages in thread

* [PATCH v6 3/3] Optimize hash function
@ 2021-01-20 22:06 Heikki Linnakangas <[email protected]>
  0 siblings, 0 replies; 50+ messages in thread

From: Heikki Linnakangas @ 2021-01-20 22:06 UTC (permalink / raw)

---
 src/backend/utils/resowner/resowner.c | 24 ++++++++++++++++++------
 src/include/common/hashfn.h           | 15 +++++++++++++++
 2 files changed, 33 insertions(+), 6 deletions(-)

diff --git a/src/backend/utils/resowner/resowner.c b/src/backend/utils/resowner/resowner.c
index 1db1e7cf14b..72e4d8a784c 100644
--- a/src/backend/utils/resowner/resowner.c
+++ b/src/backend/utils/resowner/resowner.c
@@ -163,15 +163,27 @@ static void ReleaseAuxProcessResourcesCallback(int code, Datum arg);
  *	  INTERNAL ROUTINES														 *
  *****************************************************************************/
 
+/*
+ * Hash function for value+kind combination.
+ */
 static inline uint32
 hash_resource_elem(Datum value, ResourceOwnerFuncs *kind)
 {
-	Datum		data[2];
-
-	data[0] = value;
-	data[1] = PointerGetDatum(kind);
-
-	return hash_bytes((unsigned char *) &data, 2 * SIZEOF_DATUM);
+	/*
+	 * Most resources types store a pointer in 'value', and pointers are
+	 * unique all on their own. But some resources store plain integers (Files
+	 * and Buffers as of this writing), so we want to incorporate the 'kind'
+	 * in the hash too, otherwise those resources will collide a lot. But
+	 * because there are only a few resource kinds like that - and only a few
+	 * resource kinds to begin with - we don't need to work too hard to mix
+	 * 'kind' into the hash. Just add it with hash_combine(), it perturbs the
+	 * result enough for our purposes.
+	 */
+#if SIZEOF_DATUM == 8
+	return hash_combine64(murmurhash64((uint64) value), (uint64) kind);
+#else
+	return hash_combine(murmurhash32((uint32) value), (uint32) kind);
+#endif
 }
 
 static void
diff --git a/src/include/common/hashfn.h b/src/include/common/hashfn.h
index c634cc067a1..009ffbbdd38 100644
--- a/src/include/common/hashfn.h
+++ b/src/include/common/hashfn.h
@@ -101,4 +101,19 @@ murmurhash32(uint32 data)
 	return h;
 }
 
+/* 64-bit variant */
+static inline uint64
+murmurhash64(uint64 data)
+{
+	uint64		h = data;
+
+	h ^= h >> 33;
+	h *= 0xff51afd7ed558ccd;
+	h ^= h >> 33;
+	h *= 0xc4ceb9fe1a85ec53;
+	h ^= h >> 33;
+
+	return h;
+}
+
 #endif							/* HASHFN_H */
-- 
2.29.2


--------------F30A5443A5D524A5CBC51AE7
Content-Type: text/x-patch; charset=UTF-8;
 name="v2-0001-resownerbench.patch"
Content-Transfer-Encoding: 7bit
Content-Disposition: attachment;
 filename="v2-0001-resownerbench.patch"



^ permalink  raw  reply  [nested|flat] 50+ messages in thread

* [PATCH v6 3/3] Optimize hash function
@ 2021-01-20 22:06 Heikki Linnakangas <[email protected]>
  0 siblings, 0 replies; 50+ messages in thread

From: Heikki Linnakangas @ 2021-01-20 22:06 UTC (permalink / raw)

---
 src/backend/utils/resowner/resowner.c | 24 ++++++++++++++++++------
 src/include/common/hashfn.h           | 15 +++++++++++++++
 2 files changed, 33 insertions(+), 6 deletions(-)

diff --git a/src/backend/utils/resowner/resowner.c b/src/backend/utils/resowner/resowner.c
index 1db1e7cf14b..72e4d8a784c 100644
--- a/src/backend/utils/resowner/resowner.c
+++ b/src/backend/utils/resowner/resowner.c
@@ -163,15 +163,27 @@ static void ReleaseAuxProcessResourcesCallback(int code, Datum arg);
  *	  INTERNAL ROUTINES														 *
  *****************************************************************************/
 
+/*
+ * Hash function for value+kind combination.
+ */
 static inline uint32
 hash_resource_elem(Datum value, ResourceOwnerFuncs *kind)
 {
-	Datum		data[2];
-
-	data[0] = value;
-	data[1] = PointerGetDatum(kind);
-
-	return hash_bytes((unsigned char *) &data, 2 * SIZEOF_DATUM);
+	/*
+	 * Most resources types store a pointer in 'value', and pointers are
+	 * unique all on their own. But some resources store plain integers (Files
+	 * and Buffers as of this writing), so we want to incorporate the 'kind'
+	 * in the hash too, otherwise those resources will collide a lot. But
+	 * because there are only a few resource kinds like that - and only a few
+	 * resource kinds to begin with - we don't need to work too hard to mix
+	 * 'kind' into the hash. Just add it with hash_combine(), it perturbs the
+	 * result enough for our purposes.
+	 */
+#if SIZEOF_DATUM == 8
+	return hash_combine64(murmurhash64((uint64) value), (uint64) kind);
+#else
+	return hash_combine(murmurhash32((uint32) value), (uint32) kind);
+#endif
 }
 
 static void
diff --git a/src/include/common/hashfn.h b/src/include/common/hashfn.h
index c634cc067a1..009ffbbdd38 100644
--- a/src/include/common/hashfn.h
+++ b/src/include/common/hashfn.h
@@ -101,4 +101,19 @@ murmurhash32(uint32 data)
 	return h;
 }
 
+/* 64-bit variant */
+static inline uint64
+murmurhash64(uint64 data)
+{
+	uint64		h = data;
+
+	h ^= h >> 33;
+	h *= 0xff51afd7ed558ccd;
+	h ^= h >> 33;
+	h *= 0xc4ceb9fe1a85ec53;
+	h ^= h >> 33;
+
+	return h;
+}
+
 #endif							/* HASHFN_H */
-- 
2.29.2


--------------F30A5443A5D524A5CBC51AE7
Content-Type: text/x-patch; charset=UTF-8;
 name="v2-0001-resownerbench.patch"
Content-Transfer-Encoding: 7bit
Content-Disposition: attachment;
 filename="v2-0001-resownerbench.patch"



^ permalink  raw  reply  [nested|flat] 50+ messages in thread

* [PATCH v6 3/3] Optimize hash function
@ 2021-01-20 22:06 Heikki Linnakangas <[email protected]>
  0 siblings, 0 replies; 50+ messages in thread

From: Heikki Linnakangas @ 2021-01-20 22:06 UTC (permalink / raw)

---
 src/backend/utils/resowner/resowner.c | 24 ++++++++++++++++++------
 src/include/common/hashfn.h           | 15 +++++++++++++++
 2 files changed, 33 insertions(+), 6 deletions(-)

diff --git a/src/backend/utils/resowner/resowner.c b/src/backend/utils/resowner/resowner.c
index 1db1e7cf14b..72e4d8a784c 100644
--- a/src/backend/utils/resowner/resowner.c
+++ b/src/backend/utils/resowner/resowner.c
@@ -163,15 +163,27 @@ static void ReleaseAuxProcessResourcesCallback(int code, Datum arg);
  *	  INTERNAL ROUTINES														 *
  *****************************************************************************/
 
+/*
+ * Hash function for value+kind combination.
+ */
 static inline uint32
 hash_resource_elem(Datum value, ResourceOwnerFuncs *kind)
 {
-	Datum		data[2];
-
-	data[0] = value;
-	data[1] = PointerGetDatum(kind);
-
-	return hash_bytes((unsigned char *) &data, 2 * SIZEOF_DATUM);
+	/*
+	 * Most resources types store a pointer in 'value', and pointers are
+	 * unique all on their own. But some resources store plain integers (Files
+	 * and Buffers as of this writing), so we want to incorporate the 'kind'
+	 * in the hash too, otherwise those resources will collide a lot. But
+	 * because there are only a few resource kinds like that - and only a few
+	 * resource kinds to begin with - we don't need to work too hard to mix
+	 * 'kind' into the hash. Just add it with hash_combine(), it perturbs the
+	 * result enough for our purposes.
+	 */
+#if SIZEOF_DATUM == 8
+	return hash_combine64(murmurhash64((uint64) value), (uint64) kind);
+#else
+	return hash_combine(murmurhash32((uint32) value), (uint32) kind);
+#endif
 }
 
 static void
diff --git a/src/include/common/hashfn.h b/src/include/common/hashfn.h
index c634cc067a1..009ffbbdd38 100644
--- a/src/include/common/hashfn.h
+++ b/src/include/common/hashfn.h
@@ -101,4 +101,19 @@ murmurhash32(uint32 data)
 	return h;
 }
 
+/* 64-bit variant */
+static inline uint64
+murmurhash64(uint64 data)
+{
+	uint64		h = data;
+
+	h ^= h >> 33;
+	h *= 0xff51afd7ed558ccd;
+	h ^= h >> 33;
+	h *= 0xc4ceb9fe1a85ec53;
+	h ^= h >> 33;
+
+	return h;
+}
+
 #endif							/* HASHFN_H */
-- 
2.29.2


--------------F30A5443A5D524A5CBC51AE7
Content-Type: text/x-patch; charset=UTF-8;
 name="v2-0001-resownerbench.patch"
Content-Transfer-Encoding: 7bit
Content-Disposition: attachment;
 filename="v2-0001-resownerbench.patch"



^ permalink  raw  reply  [nested|flat] 50+ messages in thread

* [PATCH v7 3/3] Optimize hash function
@ 2021-01-20 22:06 Heikki Linnakangas <[email protected]>
  0 siblings, 0 replies; 50+ messages in thread

From: Heikki Linnakangas @ 2021-01-20 22:06 UTC (permalink / raw)

---
 src/backend/utils/resowner/resowner.c | 24 ++++++++++++++++++------
 src/include/common/hashfn.h           | 15 +++++++++++++++
 2 files changed, 33 insertions(+), 6 deletions(-)

diff --git a/src/backend/utils/resowner/resowner.c b/src/backend/utils/resowner/resowner.c
index d62b1495dd3..6189d051072 100644
--- a/src/backend/utils/resowner/resowner.c
+++ b/src/backend/utils/resowner/resowner.c
@@ -163,15 +163,27 @@ static void ReleaseAuxProcessResourcesCallback(int code, Datum arg);
  *	  INTERNAL ROUTINES														 *
  *****************************************************************************/
 
+/*
+ * Hash function for value+kind combination.
+ */
 static inline uint32
 hash_resource_elem(Datum value, ResourceOwnerFuncs *kind)
 {
-	Datum		data[2];
-
-	data[0] = value;
-	data[1] = PointerGetDatum(kind);
-
-	return hash_bytes((unsigned char *) &data, 2 * SIZEOF_DATUM);
+	/*
+	 * Most resources types store a pointer in 'value', and pointers are
+	 * unique all on their own. But some resources store plain integers (Files
+	 * and Buffers as of this writing), so we want to incorporate the 'kind'
+	 * in the hash too, otherwise those resources will collide a lot. But
+	 * because there are only a few resource kinds like that - and only a few
+	 * resource kinds to begin with - we don't need to work too hard to mix
+	 * 'kind' into the hash. Just add it with hash_combine(), it perturbs the
+	 * result enough for our purposes.
+	 */
+#if SIZEOF_DATUM == 8
+	return hash_combine64(murmurhash64((uint64) value), (uint64) kind);
+#else
+	return hash_combine(murmurhash32((uint32) value), (uint32) kind);
+#endif
 }
 
 static void
diff --git a/src/include/common/hashfn.h b/src/include/common/hashfn.h
index c634cc067a1..009ffbbdd38 100644
--- a/src/include/common/hashfn.h
+++ b/src/include/common/hashfn.h
@@ -101,4 +101,19 @@ murmurhash32(uint32 data)
 	return h;
 }
 
+/* 64-bit variant */
+static inline uint64
+murmurhash64(uint64 data)
+{
+	uint64		h = data;
+
+	h ^= h >> 33;
+	h *= 0xff51afd7ed558ccd;
+	h ^= h >> 33;
+	h *= 0xc4ceb9fe1a85ec53;
+	h ^= h >> 33;
+
+	return h;
+}
+
 #endif							/* HASHFN_H */
-- 
2.30.1


--------------C6A03312D0F99F9183A0DF46--





^ permalink  raw  reply  [nested|flat] 50+ messages in thread

* [PATCH v7 3/3] Optimize hash function
@ 2021-01-20 22:06 Heikki Linnakangas <[email protected]>
  0 siblings, 0 replies; 50+ messages in thread

From: Heikki Linnakangas @ 2021-01-20 22:06 UTC (permalink / raw)

---
 src/backend/utils/resowner/resowner.c | 24 ++++++++++++++++++------
 src/include/common/hashfn.h           | 15 +++++++++++++++
 2 files changed, 33 insertions(+), 6 deletions(-)

diff --git a/src/backend/utils/resowner/resowner.c b/src/backend/utils/resowner/resowner.c
index d62b1495dd3..6189d051072 100644
--- a/src/backend/utils/resowner/resowner.c
+++ b/src/backend/utils/resowner/resowner.c
@@ -163,15 +163,27 @@ static void ReleaseAuxProcessResourcesCallback(int code, Datum arg);
  *	  INTERNAL ROUTINES														 *
  *****************************************************************************/
 
+/*
+ * Hash function for value+kind combination.
+ */
 static inline uint32
 hash_resource_elem(Datum value, ResourceOwnerFuncs *kind)
 {
-	Datum		data[2];
-
-	data[0] = value;
-	data[1] = PointerGetDatum(kind);
-
-	return hash_bytes((unsigned char *) &data, 2 * SIZEOF_DATUM);
+	/*
+	 * Most resources types store a pointer in 'value', and pointers are
+	 * unique all on their own. But some resources store plain integers (Files
+	 * and Buffers as of this writing), so we want to incorporate the 'kind'
+	 * in the hash too, otherwise those resources will collide a lot. But
+	 * because there are only a few resource kinds like that - and only a few
+	 * resource kinds to begin with - we don't need to work too hard to mix
+	 * 'kind' into the hash. Just add it with hash_combine(), it perturbs the
+	 * result enough for our purposes.
+	 */
+#if SIZEOF_DATUM == 8
+	return hash_combine64(murmurhash64((uint64) value), (uint64) kind);
+#else
+	return hash_combine(murmurhash32((uint32) value), (uint32) kind);
+#endif
 }
 
 static void
diff --git a/src/include/common/hashfn.h b/src/include/common/hashfn.h
index c634cc067a1..009ffbbdd38 100644
--- a/src/include/common/hashfn.h
+++ b/src/include/common/hashfn.h
@@ -101,4 +101,19 @@ murmurhash32(uint32 data)
 	return h;
 }
 
+/* 64-bit variant */
+static inline uint64
+murmurhash64(uint64 data)
+{
+	uint64		h = data;
+
+	h ^= h >> 33;
+	h *= 0xff51afd7ed558ccd;
+	h ^= h >> 33;
+	h *= 0xc4ceb9fe1a85ec53;
+	h ^= h >> 33;
+
+	return h;
+}
+
 #endif							/* HASHFN_H */
-- 
2.30.1


--------------C6A03312D0F99F9183A0DF46--





^ permalink  raw  reply  [nested|flat] 50+ messages in thread

* [PATCH v6 3/3] Optimize hash function
@ 2021-01-20 22:06 Heikki Linnakangas <[email protected]>
  0 siblings, 0 replies; 50+ messages in thread

From: Heikki Linnakangas @ 2021-01-20 22:06 UTC (permalink / raw)

---
 src/backend/utils/resowner/resowner.c | 24 ++++++++++++++++++------
 src/include/common/hashfn.h           | 15 +++++++++++++++
 2 files changed, 33 insertions(+), 6 deletions(-)

diff --git a/src/backend/utils/resowner/resowner.c b/src/backend/utils/resowner/resowner.c
index 1db1e7cf14b..72e4d8a784c 100644
--- a/src/backend/utils/resowner/resowner.c
+++ b/src/backend/utils/resowner/resowner.c
@@ -163,15 +163,27 @@ static void ReleaseAuxProcessResourcesCallback(int code, Datum arg);
  *	  INTERNAL ROUTINES														 *
  *****************************************************************************/
 
+/*
+ * Hash function for value+kind combination.
+ */
 static inline uint32
 hash_resource_elem(Datum value, ResourceOwnerFuncs *kind)
 {
-	Datum		data[2];
-
-	data[0] = value;
-	data[1] = PointerGetDatum(kind);
-
-	return hash_bytes((unsigned char *) &data, 2 * SIZEOF_DATUM);
+	/*
+	 * Most resources types store a pointer in 'value', and pointers are
+	 * unique all on their own. But some resources store plain integers (Files
+	 * and Buffers as of this writing), so we want to incorporate the 'kind'
+	 * in the hash too, otherwise those resources will collide a lot. But
+	 * because there are only a few resource kinds like that - and only a few
+	 * resource kinds to begin with - we don't need to work too hard to mix
+	 * 'kind' into the hash. Just add it with hash_combine(), it perturbs the
+	 * result enough for our purposes.
+	 */
+#if SIZEOF_DATUM == 8
+	return hash_combine64(murmurhash64((uint64) value), (uint64) kind);
+#else
+	return hash_combine(murmurhash32((uint32) value), (uint32) kind);
+#endif
 }
 
 static void
diff --git a/src/include/common/hashfn.h b/src/include/common/hashfn.h
index c634cc067a1..009ffbbdd38 100644
--- a/src/include/common/hashfn.h
+++ b/src/include/common/hashfn.h
@@ -101,4 +101,19 @@ murmurhash32(uint32 data)
 	return h;
 }
 
+/* 64-bit variant */
+static inline uint64
+murmurhash64(uint64 data)
+{
+	uint64		h = data;
+
+	h ^= h >> 33;
+	h *= 0xff51afd7ed558ccd;
+	h ^= h >> 33;
+	h *= 0xc4ceb9fe1a85ec53;
+	h ^= h >> 33;
+
+	return h;
+}
+
 #endif							/* HASHFN_H */
-- 
2.29.2


--------------F30A5443A5D524A5CBC51AE7
Content-Type: text/x-patch; charset=UTF-8;
 name="v2-0001-resownerbench.patch"
Content-Transfer-Encoding: 7bit
Content-Disposition: attachment;
 filename="v2-0001-resownerbench.patch"



^ permalink  raw  reply  [nested|flat] 50+ messages in thread

* [PATCH v7 3/3] Optimize hash function
@ 2021-01-20 22:06 Heikki Linnakangas <[email protected]>
  0 siblings, 0 replies; 50+ messages in thread

From: Heikki Linnakangas @ 2021-01-20 22:06 UTC (permalink / raw)

---
 src/backend/utils/resowner/resowner.c | 24 ++++++++++++++++++------
 src/include/common/hashfn.h           | 15 +++++++++++++++
 2 files changed, 33 insertions(+), 6 deletions(-)

diff --git a/src/backend/utils/resowner/resowner.c b/src/backend/utils/resowner/resowner.c
index d62b1495dd3..6189d051072 100644
--- a/src/backend/utils/resowner/resowner.c
+++ b/src/backend/utils/resowner/resowner.c
@@ -163,15 +163,27 @@ static void ReleaseAuxProcessResourcesCallback(int code, Datum arg);
  *	  INTERNAL ROUTINES														 *
  *****************************************************************************/
 
+/*
+ * Hash function for value+kind combination.
+ */
 static inline uint32
 hash_resource_elem(Datum value, ResourceOwnerFuncs *kind)
 {
-	Datum		data[2];
-
-	data[0] = value;
-	data[1] = PointerGetDatum(kind);
-
-	return hash_bytes((unsigned char *) &data, 2 * SIZEOF_DATUM);
+	/*
+	 * Most resources types store a pointer in 'value', and pointers are
+	 * unique all on their own. But some resources store plain integers (Files
+	 * and Buffers as of this writing), so we want to incorporate the 'kind'
+	 * in the hash too, otherwise those resources will collide a lot. But
+	 * because there are only a few resource kinds like that - and only a few
+	 * resource kinds to begin with - we don't need to work too hard to mix
+	 * 'kind' into the hash. Just add it with hash_combine(), it perturbs the
+	 * result enough for our purposes.
+	 */
+#if SIZEOF_DATUM == 8
+	return hash_combine64(murmurhash64((uint64) value), (uint64) kind);
+#else
+	return hash_combine(murmurhash32((uint32) value), (uint32) kind);
+#endif
 }
 
 static void
diff --git a/src/include/common/hashfn.h b/src/include/common/hashfn.h
index c634cc067a1..009ffbbdd38 100644
--- a/src/include/common/hashfn.h
+++ b/src/include/common/hashfn.h
@@ -101,4 +101,19 @@ murmurhash32(uint32 data)
 	return h;
 }
 
+/* 64-bit variant */
+static inline uint64
+murmurhash64(uint64 data)
+{
+	uint64		h = data;
+
+	h ^= h >> 33;
+	h *= 0xff51afd7ed558ccd;
+	h ^= h >> 33;
+	h *= 0xc4ceb9fe1a85ec53;
+	h ^= h >> 33;
+
+	return h;
+}
+
 #endif							/* HASHFN_H */
-- 
2.30.1


--------------C6A03312D0F99F9183A0DF46--





^ permalink  raw  reply  [nested|flat] 50+ messages in thread

* [PATCH v6 3/3] Optimize hash function
@ 2021-01-20 22:06 Heikki Linnakangas <[email protected]>
  0 siblings, 0 replies; 50+ messages in thread

From: Heikki Linnakangas @ 2021-01-20 22:06 UTC (permalink / raw)

---
 src/backend/utils/resowner/resowner.c | 24 ++++++++++++++++++------
 src/include/common/hashfn.h           | 15 +++++++++++++++
 2 files changed, 33 insertions(+), 6 deletions(-)

diff --git a/src/backend/utils/resowner/resowner.c b/src/backend/utils/resowner/resowner.c
index 1db1e7cf14b..72e4d8a784c 100644
--- a/src/backend/utils/resowner/resowner.c
+++ b/src/backend/utils/resowner/resowner.c
@@ -163,15 +163,27 @@ static void ReleaseAuxProcessResourcesCallback(int code, Datum arg);
  *	  INTERNAL ROUTINES														 *
  *****************************************************************************/
 
+/*
+ * Hash function for value+kind combination.
+ */
 static inline uint32
 hash_resource_elem(Datum value, ResourceOwnerFuncs *kind)
 {
-	Datum		data[2];
-
-	data[0] = value;
-	data[1] = PointerGetDatum(kind);
-
-	return hash_bytes((unsigned char *) &data, 2 * SIZEOF_DATUM);
+	/*
+	 * Most resources types store a pointer in 'value', and pointers are
+	 * unique all on their own. But some resources store plain integers (Files
+	 * and Buffers as of this writing), so we want to incorporate the 'kind'
+	 * in the hash too, otherwise those resources will collide a lot. But
+	 * because there are only a few resource kinds like that - and only a few
+	 * resource kinds to begin with - we don't need to work too hard to mix
+	 * 'kind' into the hash. Just add it with hash_combine(), it perturbs the
+	 * result enough for our purposes.
+	 */
+#if SIZEOF_DATUM == 8
+	return hash_combine64(murmurhash64((uint64) value), (uint64) kind);
+#else
+	return hash_combine(murmurhash32((uint32) value), (uint32) kind);
+#endif
 }
 
 static void
diff --git a/src/include/common/hashfn.h b/src/include/common/hashfn.h
index c634cc067a1..009ffbbdd38 100644
--- a/src/include/common/hashfn.h
+++ b/src/include/common/hashfn.h
@@ -101,4 +101,19 @@ murmurhash32(uint32 data)
 	return h;
 }
 
+/* 64-bit variant */
+static inline uint64
+murmurhash64(uint64 data)
+{
+	uint64		h = data;
+
+	h ^= h >> 33;
+	h *= 0xff51afd7ed558ccd;
+	h ^= h >> 33;
+	h *= 0xc4ceb9fe1a85ec53;
+	h ^= h >> 33;
+
+	return h;
+}
+
 #endif							/* HASHFN_H */
-- 
2.29.2


--------------F30A5443A5D524A5CBC51AE7
Content-Type: text/x-patch; charset=UTF-8;
 name="v2-0001-resownerbench.patch"
Content-Transfer-Encoding: 7bit
Content-Disposition: attachment;
 filename="v2-0001-resownerbench.patch"



^ permalink  raw  reply  [nested|flat] 50+ messages in thread

* [PATCH v6 3/3] Optimize hash function
@ 2021-01-20 22:06 Heikki Linnakangas <[email protected]>
  0 siblings, 0 replies; 50+ messages in thread

From: Heikki Linnakangas @ 2021-01-20 22:06 UTC (permalink / raw)

---
 src/backend/utils/resowner/resowner.c | 24 ++++++++++++++++++------
 src/include/common/hashfn.h           | 15 +++++++++++++++
 2 files changed, 33 insertions(+), 6 deletions(-)

diff --git a/src/backend/utils/resowner/resowner.c b/src/backend/utils/resowner/resowner.c
index 1db1e7cf14b..72e4d8a784c 100644
--- a/src/backend/utils/resowner/resowner.c
+++ b/src/backend/utils/resowner/resowner.c
@@ -163,15 +163,27 @@ static void ReleaseAuxProcessResourcesCallback(int code, Datum arg);
  *	  INTERNAL ROUTINES														 *
  *****************************************************************************/
 
+/*
+ * Hash function for value+kind combination.
+ */
 static inline uint32
 hash_resource_elem(Datum value, ResourceOwnerFuncs *kind)
 {
-	Datum		data[2];
-
-	data[0] = value;
-	data[1] = PointerGetDatum(kind);
-
-	return hash_bytes((unsigned char *) &data, 2 * SIZEOF_DATUM);
+	/*
+	 * Most resources types store a pointer in 'value', and pointers are
+	 * unique all on their own. But some resources store plain integers (Files
+	 * and Buffers as of this writing), so we want to incorporate the 'kind'
+	 * in the hash too, otherwise those resources will collide a lot. But
+	 * because there are only a few resource kinds like that - and only a few
+	 * resource kinds to begin with - we don't need to work too hard to mix
+	 * 'kind' into the hash. Just add it with hash_combine(), it perturbs the
+	 * result enough for our purposes.
+	 */
+#if SIZEOF_DATUM == 8
+	return hash_combine64(murmurhash64((uint64) value), (uint64) kind);
+#else
+	return hash_combine(murmurhash32((uint32) value), (uint32) kind);
+#endif
 }
 
 static void
diff --git a/src/include/common/hashfn.h b/src/include/common/hashfn.h
index c634cc067a1..009ffbbdd38 100644
--- a/src/include/common/hashfn.h
+++ b/src/include/common/hashfn.h
@@ -101,4 +101,19 @@ murmurhash32(uint32 data)
 	return h;
 }
 
+/* 64-bit variant */
+static inline uint64
+murmurhash64(uint64 data)
+{
+	uint64		h = data;
+
+	h ^= h >> 33;
+	h *= 0xff51afd7ed558ccd;
+	h ^= h >> 33;
+	h *= 0xc4ceb9fe1a85ec53;
+	h ^= h >> 33;
+
+	return h;
+}
+
 #endif							/* HASHFN_H */
-- 
2.29.2


--------------F30A5443A5D524A5CBC51AE7
Content-Type: text/x-patch; charset=UTF-8;
 name="v2-0001-resownerbench.patch"
Content-Transfer-Encoding: 7bit
Content-Disposition: attachment;
 filename="v2-0001-resownerbench.patch"



^ permalink  raw  reply  [nested|flat] 50+ messages in thread

* [PATCH v6 3/3] Optimize hash function
@ 2021-01-20 22:06 Heikki Linnakangas <[email protected]>
  0 siblings, 0 replies; 50+ messages in thread

From: Heikki Linnakangas @ 2021-01-20 22:06 UTC (permalink / raw)

---
 src/backend/utils/resowner/resowner.c | 24 ++++++++++++++++++------
 src/include/common/hashfn.h           | 15 +++++++++++++++
 2 files changed, 33 insertions(+), 6 deletions(-)

diff --git a/src/backend/utils/resowner/resowner.c b/src/backend/utils/resowner/resowner.c
index 1db1e7cf14b..72e4d8a784c 100644
--- a/src/backend/utils/resowner/resowner.c
+++ b/src/backend/utils/resowner/resowner.c
@@ -163,15 +163,27 @@ static void ReleaseAuxProcessResourcesCallback(int code, Datum arg);
  *	  INTERNAL ROUTINES														 *
  *****************************************************************************/
 
+/*
+ * Hash function for value+kind combination.
+ */
 static inline uint32
 hash_resource_elem(Datum value, ResourceOwnerFuncs *kind)
 {
-	Datum		data[2];
-
-	data[0] = value;
-	data[1] = PointerGetDatum(kind);
-
-	return hash_bytes((unsigned char *) &data, 2 * SIZEOF_DATUM);
+	/*
+	 * Most resources types store a pointer in 'value', and pointers are
+	 * unique all on their own. But some resources store plain integers (Files
+	 * and Buffers as of this writing), so we want to incorporate the 'kind'
+	 * in the hash too, otherwise those resources will collide a lot. But
+	 * because there are only a few resource kinds like that - and only a few
+	 * resource kinds to begin with - we don't need to work too hard to mix
+	 * 'kind' into the hash. Just add it with hash_combine(), it perturbs the
+	 * result enough for our purposes.
+	 */
+#if SIZEOF_DATUM == 8
+	return hash_combine64(murmurhash64((uint64) value), (uint64) kind);
+#else
+	return hash_combine(murmurhash32((uint32) value), (uint32) kind);
+#endif
 }
 
 static void
diff --git a/src/include/common/hashfn.h b/src/include/common/hashfn.h
index c634cc067a1..009ffbbdd38 100644
--- a/src/include/common/hashfn.h
+++ b/src/include/common/hashfn.h
@@ -101,4 +101,19 @@ murmurhash32(uint32 data)
 	return h;
 }
 
+/* 64-bit variant */
+static inline uint64
+murmurhash64(uint64 data)
+{
+	uint64		h = data;
+
+	h ^= h >> 33;
+	h *= 0xff51afd7ed558ccd;
+	h ^= h >> 33;
+	h *= 0xc4ceb9fe1a85ec53;
+	h ^= h >> 33;
+
+	return h;
+}
+
 #endif							/* HASHFN_H */
-- 
2.29.2


--------------F30A5443A5D524A5CBC51AE7
Content-Type: text/x-patch; charset=UTF-8;
 name="v2-0001-resownerbench.patch"
Content-Transfer-Encoding: 7bit
Content-Disposition: attachment;
 filename="v2-0001-resownerbench.patch"



^ permalink  raw  reply  [nested|flat] 50+ messages in thread

* [PATCH v6 3/3] Optimize hash function
@ 2021-01-20 22:06 Heikki Linnakangas <[email protected]>
  0 siblings, 0 replies; 50+ messages in thread

From: Heikki Linnakangas @ 2021-01-20 22:06 UTC (permalink / raw)

---
 src/backend/utils/resowner/resowner.c | 24 ++++++++++++++++++------
 src/include/common/hashfn.h           | 15 +++++++++++++++
 2 files changed, 33 insertions(+), 6 deletions(-)

diff --git a/src/backend/utils/resowner/resowner.c b/src/backend/utils/resowner/resowner.c
index 1db1e7cf14b..72e4d8a784c 100644
--- a/src/backend/utils/resowner/resowner.c
+++ b/src/backend/utils/resowner/resowner.c
@@ -163,15 +163,27 @@ static void ReleaseAuxProcessResourcesCallback(int code, Datum arg);
  *	  INTERNAL ROUTINES														 *
  *****************************************************************************/
 
+/*
+ * Hash function for value+kind combination.
+ */
 static inline uint32
 hash_resource_elem(Datum value, ResourceOwnerFuncs *kind)
 {
-	Datum		data[2];
-
-	data[0] = value;
-	data[1] = PointerGetDatum(kind);
-
-	return hash_bytes((unsigned char *) &data, 2 * SIZEOF_DATUM);
+	/*
+	 * Most resources types store a pointer in 'value', and pointers are
+	 * unique all on their own. But some resources store plain integers (Files
+	 * and Buffers as of this writing), so we want to incorporate the 'kind'
+	 * in the hash too, otherwise those resources will collide a lot. But
+	 * because there are only a few resource kinds like that - and only a few
+	 * resource kinds to begin with - we don't need to work too hard to mix
+	 * 'kind' into the hash. Just add it with hash_combine(), it perturbs the
+	 * result enough for our purposes.
+	 */
+#if SIZEOF_DATUM == 8
+	return hash_combine64(murmurhash64((uint64) value), (uint64) kind);
+#else
+	return hash_combine(murmurhash32((uint32) value), (uint32) kind);
+#endif
 }
 
 static void
diff --git a/src/include/common/hashfn.h b/src/include/common/hashfn.h
index c634cc067a1..009ffbbdd38 100644
--- a/src/include/common/hashfn.h
+++ b/src/include/common/hashfn.h
@@ -101,4 +101,19 @@ murmurhash32(uint32 data)
 	return h;
 }
 
+/* 64-bit variant */
+static inline uint64
+murmurhash64(uint64 data)
+{
+	uint64		h = data;
+
+	h ^= h >> 33;
+	h *= 0xff51afd7ed558ccd;
+	h ^= h >> 33;
+	h *= 0xc4ceb9fe1a85ec53;
+	h ^= h >> 33;
+
+	return h;
+}
+
 #endif							/* HASHFN_H */
-- 
2.29.2


--------------F30A5443A5D524A5CBC51AE7
Content-Type: text/x-patch; charset=UTF-8;
 name="v2-0001-resownerbench.patch"
Content-Transfer-Encoding: 7bit
Content-Disposition: attachment;
 filename="v2-0001-resownerbench.patch"



^ permalink  raw  reply  [nested|flat] 50+ messages in thread

* [PATCH v6 3/3] Optimize hash function
@ 2021-01-20 22:06 Heikki Linnakangas <[email protected]>
  0 siblings, 0 replies; 50+ messages in thread

From: Heikki Linnakangas @ 2021-01-20 22:06 UTC (permalink / raw)

---
 src/backend/utils/resowner/resowner.c | 24 ++++++++++++++++++------
 src/include/common/hashfn.h           | 15 +++++++++++++++
 2 files changed, 33 insertions(+), 6 deletions(-)

diff --git a/src/backend/utils/resowner/resowner.c b/src/backend/utils/resowner/resowner.c
index 1db1e7cf14b..72e4d8a784c 100644
--- a/src/backend/utils/resowner/resowner.c
+++ b/src/backend/utils/resowner/resowner.c
@@ -163,15 +163,27 @@ static void ReleaseAuxProcessResourcesCallback(int code, Datum arg);
  *	  INTERNAL ROUTINES														 *
  *****************************************************************************/
 
+/*
+ * Hash function for value+kind combination.
+ */
 static inline uint32
 hash_resource_elem(Datum value, ResourceOwnerFuncs *kind)
 {
-	Datum		data[2];
-
-	data[0] = value;
-	data[1] = PointerGetDatum(kind);
-
-	return hash_bytes((unsigned char *) &data, 2 * SIZEOF_DATUM);
+	/*
+	 * Most resources types store a pointer in 'value', and pointers are
+	 * unique all on their own. But some resources store plain integers (Files
+	 * and Buffers as of this writing), so we want to incorporate the 'kind'
+	 * in the hash too, otherwise those resources will collide a lot. But
+	 * because there are only a few resource kinds like that - and only a few
+	 * resource kinds to begin with - we don't need to work too hard to mix
+	 * 'kind' into the hash. Just add it with hash_combine(), it perturbs the
+	 * result enough for our purposes.
+	 */
+#if SIZEOF_DATUM == 8
+	return hash_combine64(murmurhash64((uint64) value), (uint64) kind);
+#else
+	return hash_combine(murmurhash32((uint32) value), (uint32) kind);
+#endif
 }
 
 static void
diff --git a/src/include/common/hashfn.h b/src/include/common/hashfn.h
index c634cc067a1..009ffbbdd38 100644
--- a/src/include/common/hashfn.h
+++ b/src/include/common/hashfn.h
@@ -101,4 +101,19 @@ murmurhash32(uint32 data)
 	return h;
 }
 
+/* 64-bit variant */
+static inline uint64
+murmurhash64(uint64 data)
+{
+	uint64		h = data;
+
+	h ^= h >> 33;
+	h *= 0xff51afd7ed558ccd;
+	h ^= h >> 33;
+	h *= 0xc4ceb9fe1a85ec53;
+	h ^= h >> 33;
+
+	return h;
+}
+
 #endif							/* HASHFN_H */
-- 
2.29.2


--------------F30A5443A5D524A5CBC51AE7
Content-Type: text/x-patch; charset=UTF-8;
 name="v2-0001-resownerbench.patch"
Content-Transfer-Encoding: 7bit
Content-Disposition: attachment;
 filename="v2-0001-resownerbench.patch"



^ permalink  raw  reply  [nested|flat] 50+ messages in thread

* [PATCH v6 3/3] Optimize hash function
@ 2021-01-20 22:06 Heikki Linnakangas <[email protected]>
  0 siblings, 0 replies; 50+ messages in thread

From: Heikki Linnakangas @ 2021-01-20 22:06 UTC (permalink / raw)

---
 src/backend/utils/resowner/resowner.c | 24 ++++++++++++++++++------
 src/include/common/hashfn.h           | 15 +++++++++++++++
 2 files changed, 33 insertions(+), 6 deletions(-)

diff --git a/src/backend/utils/resowner/resowner.c b/src/backend/utils/resowner/resowner.c
index 1db1e7cf14b..72e4d8a784c 100644
--- a/src/backend/utils/resowner/resowner.c
+++ b/src/backend/utils/resowner/resowner.c
@@ -163,15 +163,27 @@ static void ReleaseAuxProcessResourcesCallback(int code, Datum arg);
  *	  INTERNAL ROUTINES														 *
  *****************************************************************************/
 
+/*
+ * Hash function for value+kind combination.
+ */
 static inline uint32
 hash_resource_elem(Datum value, ResourceOwnerFuncs *kind)
 {
-	Datum		data[2];
-
-	data[0] = value;
-	data[1] = PointerGetDatum(kind);
-
-	return hash_bytes((unsigned char *) &data, 2 * SIZEOF_DATUM);
+	/*
+	 * Most resources types store a pointer in 'value', and pointers are
+	 * unique all on their own. But some resources store plain integers (Files
+	 * and Buffers as of this writing), so we want to incorporate the 'kind'
+	 * in the hash too, otherwise those resources will collide a lot. But
+	 * because there are only a few resource kinds like that - and only a few
+	 * resource kinds to begin with - we don't need to work too hard to mix
+	 * 'kind' into the hash. Just add it with hash_combine(), it perturbs the
+	 * result enough for our purposes.
+	 */
+#if SIZEOF_DATUM == 8
+	return hash_combine64(murmurhash64((uint64) value), (uint64) kind);
+#else
+	return hash_combine(murmurhash32((uint32) value), (uint32) kind);
+#endif
 }
 
 static void
diff --git a/src/include/common/hashfn.h b/src/include/common/hashfn.h
index c634cc067a1..009ffbbdd38 100644
--- a/src/include/common/hashfn.h
+++ b/src/include/common/hashfn.h
@@ -101,4 +101,19 @@ murmurhash32(uint32 data)
 	return h;
 }
 
+/* 64-bit variant */
+static inline uint64
+murmurhash64(uint64 data)
+{
+	uint64		h = data;
+
+	h ^= h >> 33;
+	h *= 0xff51afd7ed558ccd;
+	h ^= h >> 33;
+	h *= 0xc4ceb9fe1a85ec53;
+	h ^= h >> 33;
+
+	return h;
+}
+
 #endif							/* HASHFN_H */
-- 
2.29.2


--------------F30A5443A5D524A5CBC51AE7
Content-Type: text/x-patch; charset=UTF-8;
 name="v2-0001-resownerbench.patch"
Content-Transfer-Encoding: 7bit
Content-Disposition: attachment;
 filename="v2-0001-resownerbench.patch"



^ permalink  raw  reply  [nested|flat] 50+ messages in thread

* [PATCH v7 3/3] Optimize hash function
@ 2021-01-20 22:06 Heikki Linnakangas <[email protected]>
  0 siblings, 0 replies; 50+ messages in thread

From: Heikki Linnakangas @ 2021-01-20 22:06 UTC (permalink / raw)

---
 src/backend/utils/resowner/resowner.c | 24 ++++++++++++++++++------
 src/include/common/hashfn.h           | 15 +++++++++++++++
 2 files changed, 33 insertions(+), 6 deletions(-)

diff --git a/src/backend/utils/resowner/resowner.c b/src/backend/utils/resowner/resowner.c
index d62b1495dd3..6189d051072 100644
--- a/src/backend/utils/resowner/resowner.c
+++ b/src/backend/utils/resowner/resowner.c
@@ -163,15 +163,27 @@ static void ReleaseAuxProcessResourcesCallback(int code, Datum arg);
  *	  INTERNAL ROUTINES														 *
  *****************************************************************************/
 
+/*
+ * Hash function for value+kind combination.
+ */
 static inline uint32
 hash_resource_elem(Datum value, ResourceOwnerFuncs *kind)
 {
-	Datum		data[2];
-
-	data[0] = value;
-	data[1] = PointerGetDatum(kind);
-
-	return hash_bytes((unsigned char *) &data, 2 * SIZEOF_DATUM);
+	/*
+	 * Most resources types store a pointer in 'value', and pointers are
+	 * unique all on their own. But some resources store plain integers (Files
+	 * and Buffers as of this writing), so we want to incorporate the 'kind'
+	 * in the hash too, otherwise those resources will collide a lot. But
+	 * because there are only a few resource kinds like that - and only a few
+	 * resource kinds to begin with - we don't need to work too hard to mix
+	 * 'kind' into the hash. Just add it with hash_combine(), it perturbs the
+	 * result enough for our purposes.
+	 */
+#if SIZEOF_DATUM == 8
+	return hash_combine64(murmurhash64((uint64) value), (uint64) kind);
+#else
+	return hash_combine(murmurhash32((uint32) value), (uint32) kind);
+#endif
 }
 
 static void
diff --git a/src/include/common/hashfn.h b/src/include/common/hashfn.h
index c634cc067a1..009ffbbdd38 100644
--- a/src/include/common/hashfn.h
+++ b/src/include/common/hashfn.h
@@ -101,4 +101,19 @@ murmurhash32(uint32 data)
 	return h;
 }
 
+/* 64-bit variant */
+static inline uint64
+murmurhash64(uint64 data)
+{
+	uint64		h = data;
+
+	h ^= h >> 33;
+	h *= 0xff51afd7ed558ccd;
+	h ^= h >> 33;
+	h *= 0xc4ceb9fe1a85ec53;
+	h ^= h >> 33;
+
+	return h;
+}
+
 #endif							/* HASHFN_H */
-- 
2.30.1


--------------C6A03312D0F99F9183A0DF46--





^ permalink  raw  reply  [nested|flat] 50+ messages in thread

* [PATCH v7 3/3] Optimize hash function
@ 2021-01-20 22:06 Heikki Linnakangas <[email protected]>
  0 siblings, 0 replies; 50+ messages in thread

From: Heikki Linnakangas @ 2021-01-20 22:06 UTC (permalink / raw)

---
 src/backend/utils/resowner/resowner.c | 24 ++++++++++++++++++------
 src/include/common/hashfn.h           | 15 +++++++++++++++
 2 files changed, 33 insertions(+), 6 deletions(-)

diff --git a/src/backend/utils/resowner/resowner.c b/src/backend/utils/resowner/resowner.c
index d62b1495dd3..6189d051072 100644
--- a/src/backend/utils/resowner/resowner.c
+++ b/src/backend/utils/resowner/resowner.c
@@ -163,15 +163,27 @@ static void ReleaseAuxProcessResourcesCallback(int code, Datum arg);
  *	  INTERNAL ROUTINES														 *
  *****************************************************************************/
 
+/*
+ * Hash function for value+kind combination.
+ */
 static inline uint32
 hash_resource_elem(Datum value, ResourceOwnerFuncs *kind)
 {
-	Datum		data[2];
-
-	data[0] = value;
-	data[1] = PointerGetDatum(kind);
-
-	return hash_bytes((unsigned char *) &data, 2 * SIZEOF_DATUM);
+	/*
+	 * Most resources types store a pointer in 'value', and pointers are
+	 * unique all on their own. But some resources store plain integers (Files
+	 * and Buffers as of this writing), so we want to incorporate the 'kind'
+	 * in the hash too, otherwise those resources will collide a lot. But
+	 * because there are only a few resource kinds like that - and only a few
+	 * resource kinds to begin with - we don't need to work too hard to mix
+	 * 'kind' into the hash. Just add it with hash_combine(), it perturbs the
+	 * result enough for our purposes.
+	 */
+#if SIZEOF_DATUM == 8
+	return hash_combine64(murmurhash64((uint64) value), (uint64) kind);
+#else
+	return hash_combine(murmurhash32((uint32) value), (uint32) kind);
+#endif
 }
 
 static void
diff --git a/src/include/common/hashfn.h b/src/include/common/hashfn.h
index c634cc067a1..009ffbbdd38 100644
--- a/src/include/common/hashfn.h
+++ b/src/include/common/hashfn.h
@@ -101,4 +101,19 @@ murmurhash32(uint32 data)
 	return h;
 }
 
+/* 64-bit variant */
+static inline uint64
+murmurhash64(uint64 data)
+{
+	uint64		h = data;
+
+	h ^= h >> 33;
+	h *= 0xff51afd7ed558ccd;
+	h ^= h >> 33;
+	h *= 0xc4ceb9fe1a85ec53;
+	h ^= h >> 33;
+
+	return h;
+}
+
 #endif							/* HASHFN_H */
-- 
2.30.1


--------------C6A03312D0F99F9183A0DF46--





^ permalink  raw  reply  [nested|flat] 50+ messages in thread

* [PATCH v8 3/3] Optimize hash function
@ 2021-01-20 22:06 Heikki Linnakangas <[email protected]>
  0 siblings, 0 replies; 50+ messages in thread

From: Heikki Linnakangas @ 2021-01-20 22:06 UTC (permalink / raw)

---
 src/backend/utils/resowner/resowner.c | 24 ++++++++++++++++++------
 src/include/common/hashfn.h           | 15 +++++++++++++++
 2 files changed, 33 insertions(+), 6 deletions(-)

diff --git a/src/backend/utils/resowner/resowner.c b/src/backend/utils/resowner/resowner.c
index d62b1495dd3..6189d051072 100644
--- a/src/backend/utils/resowner/resowner.c
+++ b/src/backend/utils/resowner/resowner.c
@@ -163,15 +163,27 @@ static void ReleaseAuxProcessResourcesCallback(int code, Datum arg);
  *	  INTERNAL ROUTINES														 *
  *****************************************************************************/
 
+/*
+ * Hash function for value+kind combination.
+ */
 static inline uint32
 hash_resource_elem(Datum value, ResourceOwnerFuncs *kind)
 {
-	Datum		data[2];
-
-	data[0] = value;
-	data[1] = PointerGetDatum(kind);
-
-	return hash_bytes((unsigned char *) &data, 2 * SIZEOF_DATUM);
+	/*
+	 * Most resources types store a pointer in 'value', and pointers are
+	 * unique all on their own. But some resources store plain integers (Files
+	 * and Buffers as of this writing), so we want to incorporate the 'kind'
+	 * in the hash too, otherwise those resources will collide a lot. But
+	 * because there are only a few resource kinds like that - and only a few
+	 * resource kinds to begin with - we don't need to work too hard to mix
+	 * 'kind' into the hash. Just add it with hash_combine(), it perturbs the
+	 * result enough for our purposes.
+	 */
+#if SIZEOF_DATUM == 8
+	return hash_combine64(murmurhash64((uint64) value), (uint64) kind);
+#else
+	return hash_combine(murmurhash32((uint32) value), (uint32) kind);
+#endif
 }
 
 static void
diff --git a/src/include/common/hashfn.h b/src/include/common/hashfn.h
index c634cc067a1..009ffbbdd38 100644
--- a/src/include/common/hashfn.h
+++ b/src/include/common/hashfn.h
@@ -101,4 +101,19 @@ murmurhash32(uint32 data)
 	return h;
 }
 
+/* 64-bit variant */
+static inline uint64
+murmurhash64(uint64 data)
+{
+	uint64		h = data;
+
+	h ^= h >> 33;
+	h *= 0xff51afd7ed558ccd;
+	h ^= h >> 33;
+	h *= 0xc4ceb9fe1a85ec53;
+	h ^= h >> 33;
+
+	return h;
+}
+
 #endif							/* HASHFN_H */
-- 
2.30.2


--------------846F221A8A016B7AA8E567B3--





^ permalink  raw  reply  [nested|flat] 50+ messages in thread

* [PATCH v7 3/3] Optimize hash function
@ 2021-01-20 22:06 Heikki Linnakangas <[email protected]>
  0 siblings, 0 replies; 50+ messages in thread

From: Heikki Linnakangas @ 2021-01-20 22:06 UTC (permalink / raw)

---
 src/backend/utils/resowner/resowner.c | 24 ++++++++++++++++++------
 src/include/common/hashfn.h           | 15 +++++++++++++++
 2 files changed, 33 insertions(+), 6 deletions(-)

diff --git a/src/backend/utils/resowner/resowner.c b/src/backend/utils/resowner/resowner.c
index d62b1495dd3..6189d051072 100644
--- a/src/backend/utils/resowner/resowner.c
+++ b/src/backend/utils/resowner/resowner.c
@@ -163,15 +163,27 @@ static void ReleaseAuxProcessResourcesCallback(int code, Datum arg);
  *	  INTERNAL ROUTINES														 *
  *****************************************************************************/
 
+/*
+ * Hash function for value+kind combination.
+ */
 static inline uint32
 hash_resource_elem(Datum value, ResourceOwnerFuncs *kind)
 {
-	Datum		data[2];
-
-	data[0] = value;
-	data[1] = PointerGetDatum(kind);
-
-	return hash_bytes((unsigned char *) &data, 2 * SIZEOF_DATUM);
+	/*
+	 * Most resources types store a pointer in 'value', and pointers are
+	 * unique all on their own. But some resources store plain integers (Files
+	 * and Buffers as of this writing), so we want to incorporate the 'kind'
+	 * in the hash too, otherwise those resources will collide a lot. But
+	 * because there are only a few resource kinds like that - and only a few
+	 * resource kinds to begin with - we don't need to work too hard to mix
+	 * 'kind' into the hash. Just add it with hash_combine(), it perturbs the
+	 * result enough for our purposes.
+	 */
+#if SIZEOF_DATUM == 8
+	return hash_combine64(murmurhash64((uint64) value), (uint64) kind);
+#else
+	return hash_combine(murmurhash32((uint32) value), (uint32) kind);
+#endif
 }
 
 static void
diff --git a/src/include/common/hashfn.h b/src/include/common/hashfn.h
index c634cc067a1..009ffbbdd38 100644
--- a/src/include/common/hashfn.h
+++ b/src/include/common/hashfn.h
@@ -101,4 +101,19 @@ murmurhash32(uint32 data)
 	return h;
 }
 
+/* 64-bit variant */
+static inline uint64
+murmurhash64(uint64 data)
+{
+	uint64		h = data;
+
+	h ^= h >> 33;
+	h *= 0xff51afd7ed558ccd;
+	h ^= h >> 33;
+	h *= 0xc4ceb9fe1a85ec53;
+	h ^= h >> 33;
+
+	return h;
+}
+
 #endif							/* HASHFN_H */
-- 
2.30.1


--------------C6A03312D0F99F9183A0DF46--





^ permalink  raw  reply  [nested|flat] 50+ messages in thread

* [PATCH v7 3/3] Optimize hash function
@ 2021-01-20 22:06 Heikki Linnakangas <[email protected]>
  0 siblings, 0 replies; 50+ messages in thread

From: Heikki Linnakangas @ 2021-01-20 22:06 UTC (permalink / raw)

---
 src/backend/utils/resowner/resowner.c | 24 ++++++++++++++++++------
 src/include/common/hashfn.h           | 15 +++++++++++++++
 2 files changed, 33 insertions(+), 6 deletions(-)

diff --git a/src/backend/utils/resowner/resowner.c b/src/backend/utils/resowner/resowner.c
index d62b1495dd3..6189d051072 100644
--- a/src/backend/utils/resowner/resowner.c
+++ b/src/backend/utils/resowner/resowner.c
@@ -163,15 +163,27 @@ static void ReleaseAuxProcessResourcesCallback(int code, Datum arg);
  *	  INTERNAL ROUTINES														 *
  *****************************************************************************/
 
+/*
+ * Hash function for value+kind combination.
+ */
 static inline uint32
 hash_resource_elem(Datum value, ResourceOwnerFuncs *kind)
 {
-	Datum		data[2];
-
-	data[0] = value;
-	data[1] = PointerGetDatum(kind);
-
-	return hash_bytes((unsigned char *) &data, 2 * SIZEOF_DATUM);
+	/*
+	 * Most resources types store a pointer in 'value', and pointers are
+	 * unique all on their own. But some resources store plain integers (Files
+	 * and Buffers as of this writing), so we want to incorporate the 'kind'
+	 * in the hash too, otherwise those resources will collide a lot. But
+	 * because there are only a few resource kinds like that - and only a few
+	 * resource kinds to begin with - we don't need to work too hard to mix
+	 * 'kind' into the hash. Just add it with hash_combine(), it perturbs the
+	 * result enough for our purposes.
+	 */
+#if SIZEOF_DATUM == 8
+	return hash_combine64(murmurhash64((uint64) value), (uint64) kind);
+#else
+	return hash_combine(murmurhash32((uint32) value), (uint32) kind);
+#endif
 }
 
 static void
diff --git a/src/include/common/hashfn.h b/src/include/common/hashfn.h
index c634cc067a1..009ffbbdd38 100644
--- a/src/include/common/hashfn.h
+++ b/src/include/common/hashfn.h
@@ -101,4 +101,19 @@ murmurhash32(uint32 data)
 	return h;
 }
 
+/* 64-bit variant */
+static inline uint64
+murmurhash64(uint64 data)
+{
+	uint64		h = data;
+
+	h ^= h >> 33;
+	h *= 0xff51afd7ed558ccd;
+	h ^= h >> 33;
+	h *= 0xc4ceb9fe1a85ec53;
+	h ^= h >> 33;
+
+	return h;
+}
+
 #endif							/* HASHFN_H */
-- 
2.30.1


--------------C6A03312D0F99F9183A0DF46--





^ permalink  raw  reply  [nested|flat] 50+ messages in thread

* [PATCH v7 3/3] Optimize hash function
@ 2021-01-20 22:06 Heikki Linnakangas <[email protected]>
  0 siblings, 0 replies; 50+ messages in thread

From: Heikki Linnakangas @ 2021-01-20 22:06 UTC (permalink / raw)

---
 src/backend/utils/resowner/resowner.c | 24 ++++++++++++++++++------
 src/include/common/hashfn.h           | 15 +++++++++++++++
 2 files changed, 33 insertions(+), 6 deletions(-)

diff --git a/src/backend/utils/resowner/resowner.c b/src/backend/utils/resowner/resowner.c
index d62b1495dd3..6189d051072 100644
--- a/src/backend/utils/resowner/resowner.c
+++ b/src/backend/utils/resowner/resowner.c
@@ -163,15 +163,27 @@ static void ReleaseAuxProcessResourcesCallback(int code, Datum arg);
  *	  INTERNAL ROUTINES														 *
  *****************************************************************************/
 
+/*
+ * Hash function for value+kind combination.
+ */
 static inline uint32
 hash_resource_elem(Datum value, ResourceOwnerFuncs *kind)
 {
-	Datum		data[2];
-
-	data[0] = value;
-	data[1] = PointerGetDatum(kind);
-
-	return hash_bytes((unsigned char *) &data, 2 * SIZEOF_DATUM);
+	/*
+	 * Most resources types store a pointer in 'value', and pointers are
+	 * unique all on their own. But some resources store plain integers (Files
+	 * and Buffers as of this writing), so we want to incorporate the 'kind'
+	 * in the hash too, otherwise those resources will collide a lot. But
+	 * because there are only a few resource kinds like that - and only a few
+	 * resource kinds to begin with - we don't need to work too hard to mix
+	 * 'kind' into the hash. Just add it with hash_combine(), it perturbs the
+	 * result enough for our purposes.
+	 */
+#if SIZEOF_DATUM == 8
+	return hash_combine64(murmurhash64((uint64) value), (uint64) kind);
+#else
+	return hash_combine(murmurhash32((uint32) value), (uint32) kind);
+#endif
 }
 
 static void
diff --git a/src/include/common/hashfn.h b/src/include/common/hashfn.h
index c634cc067a1..009ffbbdd38 100644
--- a/src/include/common/hashfn.h
+++ b/src/include/common/hashfn.h
@@ -101,4 +101,19 @@ murmurhash32(uint32 data)
 	return h;
 }
 
+/* 64-bit variant */
+static inline uint64
+murmurhash64(uint64 data)
+{
+	uint64		h = data;
+
+	h ^= h >> 33;
+	h *= 0xff51afd7ed558ccd;
+	h ^= h >> 33;
+	h *= 0xc4ceb9fe1a85ec53;
+	h ^= h >> 33;
+
+	return h;
+}
+
 #endif							/* HASHFN_H */
-- 
2.30.1


--------------C6A03312D0F99F9183A0DF46--





^ permalink  raw  reply  [nested|flat] 50+ messages in thread

* [PATCH v7 3/3] Optimize hash function
@ 2021-01-20 22:06 Heikki Linnakangas <[email protected]>
  0 siblings, 0 replies; 50+ messages in thread

From: Heikki Linnakangas @ 2021-01-20 22:06 UTC (permalink / raw)

---
 src/backend/utils/resowner/resowner.c | 24 ++++++++++++++++++------
 src/include/common/hashfn.h           | 15 +++++++++++++++
 2 files changed, 33 insertions(+), 6 deletions(-)

diff --git a/src/backend/utils/resowner/resowner.c b/src/backend/utils/resowner/resowner.c
index d62b1495dd3..6189d051072 100644
--- a/src/backend/utils/resowner/resowner.c
+++ b/src/backend/utils/resowner/resowner.c
@@ -163,15 +163,27 @@ static void ReleaseAuxProcessResourcesCallback(int code, Datum arg);
  *	  INTERNAL ROUTINES														 *
  *****************************************************************************/
 
+/*
+ * Hash function for value+kind combination.
+ */
 static inline uint32
 hash_resource_elem(Datum value, ResourceOwnerFuncs *kind)
 {
-	Datum		data[2];
-
-	data[0] = value;
-	data[1] = PointerGetDatum(kind);
-
-	return hash_bytes((unsigned char *) &data, 2 * SIZEOF_DATUM);
+	/*
+	 * Most resources types store a pointer in 'value', and pointers are
+	 * unique all on their own. But some resources store plain integers (Files
+	 * and Buffers as of this writing), so we want to incorporate the 'kind'
+	 * in the hash too, otherwise those resources will collide a lot. But
+	 * because there are only a few resource kinds like that - and only a few
+	 * resource kinds to begin with - we don't need to work too hard to mix
+	 * 'kind' into the hash. Just add it with hash_combine(), it perturbs the
+	 * result enough for our purposes.
+	 */
+#if SIZEOF_DATUM == 8
+	return hash_combine64(murmurhash64((uint64) value), (uint64) kind);
+#else
+	return hash_combine(murmurhash32((uint32) value), (uint32) kind);
+#endif
 }
 
 static void
diff --git a/src/include/common/hashfn.h b/src/include/common/hashfn.h
index c634cc067a1..009ffbbdd38 100644
--- a/src/include/common/hashfn.h
+++ b/src/include/common/hashfn.h
@@ -101,4 +101,19 @@ murmurhash32(uint32 data)
 	return h;
 }
 
+/* 64-bit variant */
+static inline uint64
+murmurhash64(uint64 data)
+{
+	uint64		h = data;
+
+	h ^= h >> 33;
+	h *= 0xff51afd7ed558ccd;
+	h ^= h >> 33;
+	h *= 0xc4ceb9fe1a85ec53;
+	h ^= h >> 33;
+
+	return h;
+}
+
 #endif							/* HASHFN_H */
-- 
2.30.1


--------------C6A03312D0F99F9183A0DF46--





^ permalink  raw  reply  [nested|flat] 50+ messages in thread

* [PATCH v7 3/3] Optimize hash function
@ 2021-01-20 22:06 Heikki Linnakangas <[email protected]>
  0 siblings, 0 replies; 50+ messages in thread

From: Heikki Linnakangas @ 2021-01-20 22:06 UTC (permalink / raw)

---
 src/backend/utils/resowner/resowner.c | 24 ++++++++++++++++++------
 src/include/common/hashfn.h           | 15 +++++++++++++++
 2 files changed, 33 insertions(+), 6 deletions(-)

diff --git a/src/backend/utils/resowner/resowner.c b/src/backend/utils/resowner/resowner.c
index d62b1495dd3..6189d051072 100644
--- a/src/backend/utils/resowner/resowner.c
+++ b/src/backend/utils/resowner/resowner.c
@@ -163,15 +163,27 @@ static void ReleaseAuxProcessResourcesCallback(int code, Datum arg);
  *	  INTERNAL ROUTINES														 *
  *****************************************************************************/
 
+/*
+ * Hash function for value+kind combination.
+ */
 static inline uint32
 hash_resource_elem(Datum value, ResourceOwnerFuncs *kind)
 {
-	Datum		data[2];
-
-	data[0] = value;
-	data[1] = PointerGetDatum(kind);
-
-	return hash_bytes((unsigned char *) &data, 2 * SIZEOF_DATUM);
+	/*
+	 * Most resources types store a pointer in 'value', and pointers are
+	 * unique all on their own. But some resources store plain integers (Files
+	 * and Buffers as of this writing), so we want to incorporate the 'kind'
+	 * in the hash too, otherwise those resources will collide a lot. But
+	 * because there are only a few resource kinds like that - and only a few
+	 * resource kinds to begin with - we don't need to work too hard to mix
+	 * 'kind' into the hash. Just add it with hash_combine(), it perturbs the
+	 * result enough for our purposes.
+	 */
+#if SIZEOF_DATUM == 8
+	return hash_combine64(murmurhash64((uint64) value), (uint64) kind);
+#else
+	return hash_combine(murmurhash32((uint32) value), (uint32) kind);
+#endif
 }
 
 static void
diff --git a/src/include/common/hashfn.h b/src/include/common/hashfn.h
index c634cc067a1..009ffbbdd38 100644
--- a/src/include/common/hashfn.h
+++ b/src/include/common/hashfn.h
@@ -101,4 +101,19 @@ murmurhash32(uint32 data)
 	return h;
 }
 
+/* 64-bit variant */
+static inline uint64
+murmurhash64(uint64 data)
+{
+	uint64		h = data;
+
+	h ^= h >> 33;
+	h *= 0xff51afd7ed558ccd;
+	h ^= h >> 33;
+	h *= 0xc4ceb9fe1a85ec53;
+	h ^= h >> 33;
+
+	return h;
+}
+
 #endif							/* HASHFN_H */
-- 
2.30.1


--------------C6A03312D0F99F9183A0DF46--





^ permalink  raw  reply  [nested|flat] 50+ messages in thread

* [PATCH v7 3/3] Optimize hash function
@ 2021-01-20 22:06 Heikki Linnakangas <[email protected]>
  0 siblings, 0 replies; 50+ messages in thread

From: Heikki Linnakangas @ 2021-01-20 22:06 UTC (permalink / raw)

---
 src/backend/utils/resowner/resowner.c | 24 ++++++++++++++++++------
 src/include/common/hashfn.h           | 15 +++++++++++++++
 2 files changed, 33 insertions(+), 6 deletions(-)

diff --git a/src/backend/utils/resowner/resowner.c b/src/backend/utils/resowner/resowner.c
index d62b1495dd3..6189d051072 100644
--- a/src/backend/utils/resowner/resowner.c
+++ b/src/backend/utils/resowner/resowner.c
@@ -163,15 +163,27 @@ static void ReleaseAuxProcessResourcesCallback(int code, Datum arg);
  *	  INTERNAL ROUTINES														 *
  *****************************************************************************/
 
+/*
+ * Hash function for value+kind combination.
+ */
 static inline uint32
 hash_resource_elem(Datum value, ResourceOwnerFuncs *kind)
 {
-	Datum		data[2];
-
-	data[0] = value;
-	data[1] = PointerGetDatum(kind);
-
-	return hash_bytes((unsigned char *) &data, 2 * SIZEOF_DATUM);
+	/*
+	 * Most resources types store a pointer in 'value', and pointers are
+	 * unique all on their own. But some resources store plain integers (Files
+	 * and Buffers as of this writing), so we want to incorporate the 'kind'
+	 * in the hash too, otherwise those resources will collide a lot. But
+	 * because there are only a few resource kinds like that - and only a few
+	 * resource kinds to begin with - we don't need to work too hard to mix
+	 * 'kind' into the hash. Just add it with hash_combine(), it perturbs the
+	 * result enough for our purposes.
+	 */
+#if SIZEOF_DATUM == 8
+	return hash_combine64(murmurhash64((uint64) value), (uint64) kind);
+#else
+	return hash_combine(murmurhash32((uint32) value), (uint32) kind);
+#endif
 }
 
 static void
diff --git a/src/include/common/hashfn.h b/src/include/common/hashfn.h
index c634cc067a1..009ffbbdd38 100644
--- a/src/include/common/hashfn.h
+++ b/src/include/common/hashfn.h
@@ -101,4 +101,19 @@ murmurhash32(uint32 data)
 	return h;
 }
 
+/* 64-bit variant */
+static inline uint64
+murmurhash64(uint64 data)
+{
+	uint64		h = data;
+
+	h ^= h >> 33;
+	h *= 0xff51afd7ed558ccd;
+	h ^= h >> 33;
+	h *= 0xc4ceb9fe1a85ec53;
+	h ^= h >> 33;
+
+	return h;
+}
+
 #endif							/* HASHFN_H */
-- 
2.30.1


--------------C6A03312D0F99F9183A0DF46--





^ permalink  raw  reply  [nested|flat] 50+ messages in thread

* [PATCH v7 3/3] Optimize hash function
@ 2021-01-20 22:06 Heikki Linnakangas <[email protected]>
  0 siblings, 0 replies; 50+ messages in thread

From: Heikki Linnakangas @ 2021-01-20 22:06 UTC (permalink / raw)

---
 src/backend/utils/resowner/resowner.c | 24 ++++++++++++++++++------
 src/include/common/hashfn.h           | 15 +++++++++++++++
 2 files changed, 33 insertions(+), 6 deletions(-)

diff --git a/src/backend/utils/resowner/resowner.c b/src/backend/utils/resowner/resowner.c
index d62b1495dd3..6189d051072 100644
--- a/src/backend/utils/resowner/resowner.c
+++ b/src/backend/utils/resowner/resowner.c
@@ -163,15 +163,27 @@ static void ReleaseAuxProcessResourcesCallback(int code, Datum arg);
  *	  INTERNAL ROUTINES														 *
  *****************************************************************************/
 
+/*
+ * Hash function for value+kind combination.
+ */
 static inline uint32
 hash_resource_elem(Datum value, ResourceOwnerFuncs *kind)
 {
-	Datum		data[2];
-
-	data[0] = value;
-	data[1] = PointerGetDatum(kind);
-
-	return hash_bytes((unsigned char *) &data, 2 * SIZEOF_DATUM);
+	/*
+	 * Most resources types store a pointer in 'value', and pointers are
+	 * unique all on their own. But some resources store plain integers (Files
+	 * and Buffers as of this writing), so we want to incorporate the 'kind'
+	 * in the hash too, otherwise those resources will collide a lot. But
+	 * because there are only a few resource kinds like that - and only a few
+	 * resource kinds to begin with - we don't need to work too hard to mix
+	 * 'kind' into the hash. Just add it with hash_combine(), it perturbs the
+	 * result enough for our purposes.
+	 */
+#if SIZEOF_DATUM == 8
+	return hash_combine64(murmurhash64((uint64) value), (uint64) kind);
+#else
+	return hash_combine(murmurhash32((uint32) value), (uint32) kind);
+#endif
 }
 
 static void
diff --git a/src/include/common/hashfn.h b/src/include/common/hashfn.h
index c634cc067a1..009ffbbdd38 100644
--- a/src/include/common/hashfn.h
+++ b/src/include/common/hashfn.h
@@ -101,4 +101,19 @@ murmurhash32(uint32 data)
 	return h;
 }
 
+/* 64-bit variant */
+static inline uint64
+murmurhash64(uint64 data)
+{
+	uint64		h = data;
+
+	h ^= h >> 33;
+	h *= 0xff51afd7ed558ccd;
+	h ^= h >> 33;
+	h *= 0xc4ceb9fe1a85ec53;
+	h ^= h >> 33;
+
+	return h;
+}
+
 #endif							/* HASHFN_H */
-- 
2.30.1


--------------C6A03312D0F99F9183A0DF46--





^ permalink  raw  reply  [nested|flat] 50+ messages in thread

* [PATCH v6 3/3] Optimize hash function
@ 2021-01-20 22:06 Heikki Linnakangas <[email protected]>
  0 siblings, 0 replies; 50+ messages in thread

From: Heikki Linnakangas @ 2021-01-20 22:06 UTC (permalink / raw)

---
 src/backend/utils/resowner/resowner.c | 24 ++++++++++++++++++------
 src/include/common/hashfn.h           | 15 +++++++++++++++
 2 files changed, 33 insertions(+), 6 deletions(-)

diff --git a/src/backend/utils/resowner/resowner.c b/src/backend/utils/resowner/resowner.c
index 1db1e7cf14b..72e4d8a784c 100644
--- a/src/backend/utils/resowner/resowner.c
+++ b/src/backend/utils/resowner/resowner.c
@@ -163,15 +163,27 @@ static void ReleaseAuxProcessResourcesCallback(int code, Datum arg);
  *	  INTERNAL ROUTINES														 *
  *****************************************************************************/
 
+/*
+ * Hash function for value+kind combination.
+ */
 static inline uint32
 hash_resource_elem(Datum value, ResourceOwnerFuncs *kind)
 {
-	Datum		data[2];
-
-	data[0] = value;
-	data[1] = PointerGetDatum(kind);
-
-	return hash_bytes((unsigned char *) &data, 2 * SIZEOF_DATUM);
+	/*
+	 * Most resources types store a pointer in 'value', and pointers are
+	 * unique all on their own. But some resources store plain integers (Files
+	 * and Buffers as of this writing), so we want to incorporate the 'kind'
+	 * in the hash too, otherwise those resources will collide a lot. But
+	 * because there are only a few resource kinds like that - and only a few
+	 * resource kinds to begin with - we don't need to work too hard to mix
+	 * 'kind' into the hash. Just add it with hash_combine(), it perturbs the
+	 * result enough for our purposes.
+	 */
+#if SIZEOF_DATUM == 8
+	return hash_combine64(murmurhash64((uint64) value), (uint64) kind);
+#else
+	return hash_combine(murmurhash32((uint32) value), (uint32) kind);
+#endif
 }
 
 static void
diff --git a/src/include/common/hashfn.h b/src/include/common/hashfn.h
index c634cc067a1..009ffbbdd38 100644
--- a/src/include/common/hashfn.h
+++ b/src/include/common/hashfn.h
@@ -101,4 +101,19 @@ murmurhash32(uint32 data)
 	return h;
 }
 
+/* 64-bit variant */
+static inline uint64
+murmurhash64(uint64 data)
+{
+	uint64		h = data;
+
+	h ^= h >> 33;
+	h *= 0xff51afd7ed558ccd;
+	h ^= h >> 33;
+	h *= 0xc4ceb9fe1a85ec53;
+	h ^= h >> 33;
+
+	return h;
+}
+
 #endif							/* HASHFN_H */
-- 
2.29.2


--------------F30A5443A5D524A5CBC51AE7
Content-Type: text/x-patch; charset=UTF-8;
 name="v2-0001-resownerbench.patch"
Content-Transfer-Encoding: 7bit
Content-Disposition: attachment;
 filename="v2-0001-resownerbench.patch"



^ permalink  raw  reply  [nested|flat] 50+ messages in thread

* [PATCH v7 3/3] Optimize hash function
@ 2021-01-20 22:06 Heikki Linnakangas <[email protected]>
  0 siblings, 0 replies; 50+ messages in thread

From: Heikki Linnakangas @ 2021-01-20 22:06 UTC (permalink / raw)

---
 src/backend/utils/resowner/resowner.c | 24 ++++++++++++++++++------
 src/include/common/hashfn.h           | 15 +++++++++++++++
 2 files changed, 33 insertions(+), 6 deletions(-)

diff --git a/src/backend/utils/resowner/resowner.c b/src/backend/utils/resowner/resowner.c
index d62b1495dd3..6189d051072 100644
--- a/src/backend/utils/resowner/resowner.c
+++ b/src/backend/utils/resowner/resowner.c
@@ -163,15 +163,27 @@ static void ReleaseAuxProcessResourcesCallback(int code, Datum arg);
  *	  INTERNAL ROUTINES														 *
  *****************************************************************************/
 
+/*
+ * Hash function for value+kind combination.
+ */
 static inline uint32
 hash_resource_elem(Datum value, ResourceOwnerFuncs *kind)
 {
-	Datum		data[2];
-
-	data[0] = value;
-	data[1] = PointerGetDatum(kind);
-
-	return hash_bytes((unsigned char *) &data, 2 * SIZEOF_DATUM);
+	/*
+	 * Most resources types store a pointer in 'value', and pointers are
+	 * unique all on their own. But some resources store plain integers (Files
+	 * and Buffers as of this writing), so we want to incorporate the 'kind'
+	 * in the hash too, otherwise those resources will collide a lot. But
+	 * because there are only a few resource kinds like that - and only a few
+	 * resource kinds to begin with - we don't need to work too hard to mix
+	 * 'kind' into the hash. Just add it with hash_combine(), it perturbs the
+	 * result enough for our purposes.
+	 */
+#if SIZEOF_DATUM == 8
+	return hash_combine64(murmurhash64((uint64) value), (uint64) kind);
+#else
+	return hash_combine(murmurhash32((uint32) value), (uint32) kind);
+#endif
 }
 
 static void
diff --git a/src/include/common/hashfn.h b/src/include/common/hashfn.h
index c634cc067a1..009ffbbdd38 100644
--- a/src/include/common/hashfn.h
+++ b/src/include/common/hashfn.h
@@ -101,4 +101,19 @@ murmurhash32(uint32 data)
 	return h;
 }
 
+/* 64-bit variant */
+static inline uint64
+murmurhash64(uint64 data)
+{
+	uint64		h = data;
+
+	h ^= h >> 33;
+	h *= 0xff51afd7ed558ccd;
+	h ^= h >> 33;
+	h *= 0xc4ceb9fe1a85ec53;
+	h ^= h >> 33;
+
+	return h;
+}
+
 #endif							/* HASHFN_H */
-- 
2.30.1


--------------C6A03312D0F99F9183A0DF46--





^ permalink  raw  reply  [nested|flat] 50+ messages in thread

* [PATCH v7 3/3] Optimize hash function
@ 2021-01-20 22:06 Heikki Linnakangas <[email protected]>
  0 siblings, 0 replies; 50+ messages in thread

From: Heikki Linnakangas @ 2021-01-20 22:06 UTC (permalink / raw)

---
 src/backend/utils/resowner/resowner.c | 24 ++++++++++++++++++------
 src/include/common/hashfn.h           | 15 +++++++++++++++
 2 files changed, 33 insertions(+), 6 deletions(-)

diff --git a/src/backend/utils/resowner/resowner.c b/src/backend/utils/resowner/resowner.c
index d62b1495dd3..6189d051072 100644
--- a/src/backend/utils/resowner/resowner.c
+++ b/src/backend/utils/resowner/resowner.c
@@ -163,15 +163,27 @@ static void ReleaseAuxProcessResourcesCallback(int code, Datum arg);
  *	  INTERNAL ROUTINES														 *
  *****************************************************************************/
 
+/*
+ * Hash function for value+kind combination.
+ */
 static inline uint32
 hash_resource_elem(Datum value, ResourceOwnerFuncs *kind)
 {
-	Datum		data[2];
-
-	data[0] = value;
-	data[1] = PointerGetDatum(kind);
-
-	return hash_bytes((unsigned char *) &data, 2 * SIZEOF_DATUM);
+	/*
+	 * Most resources types store a pointer in 'value', and pointers are
+	 * unique all on their own. But some resources store plain integers (Files
+	 * and Buffers as of this writing), so we want to incorporate the 'kind'
+	 * in the hash too, otherwise those resources will collide a lot. But
+	 * because there are only a few resource kinds like that - and only a few
+	 * resource kinds to begin with - we don't need to work too hard to mix
+	 * 'kind' into the hash. Just add it with hash_combine(), it perturbs the
+	 * result enough for our purposes.
+	 */
+#if SIZEOF_DATUM == 8
+	return hash_combine64(murmurhash64((uint64) value), (uint64) kind);
+#else
+	return hash_combine(murmurhash32((uint32) value), (uint32) kind);
+#endif
 }
 
 static void
diff --git a/src/include/common/hashfn.h b/src/include/common/hashfn.h
index c634cc067a1..009ffbbdd38 100644
--- a/src/include/common/hashfn.h
+++ b/src/include/common/hashfn.h
@@ -101,4 +101,19 @@ murmurhash32(uint32 data)
 	return h;
 }
 
+/* 64-bit variant */
+static inline uint64
+murmurhash64(uint64 data)
+{
+	uint64		h = data;
+
+	h ^= h >> 33;
+	h *= 0xff51afd7ed558ccd;
+	h ^= h >> 33;
+	h *= 0xc4ceb9fe1a85ec53;
+	h ^= h >> 33;
+
+	return h;
+}
+
 #endif							/* HASHFN_H */
-- 
2.30.1


--------------C6A03312D0F99F9183A0DF46--





^ permalink  raw  reply  [nested|flat] 50+ messages in thread

* [PATCH v6 3/3] Optimize hash function
@ 2021-01-20 22:06 Heikki Linnakangas <[email protected]>
  0 siblings, 0 replies; 50+ messages in thread

From: Heikki Linnakangas @ 2021-01-20 22:06 UTC (permalink / raw)

---
 src/backend/utils/resowner/resowner.c | 24 ++++++++++++++++++------
 src/include/common/hashfn.h           | 15 +++++++++++++++
 2 files changed, 33 insertions(+), 6 deletions(-)

diff --git a/src/backend/utils/resowner/resowner.c b/src/backend/utils/resowner/resowner.c
index 1db1e7cf14b..72e4d8a784c 100644
--- a/src/backend/utils/resowner/resowner.c
+++ b/src/backend/utils/resowner/resowner.c
@@ -163,15 +163,27 @@ static void ReleaseAuxProcessResourcesCallback(int code, Datum arg);
  *	  INTERNAL ROUTINES														 *
  *****************************************************************************/
 
+/*
+ * Hash function for value+kind combination.
+ */
 static inline uint32
 hash_resource_elem(Datum value, ResourceOwnerFuncs *kind)
 {
-	Datum		data[2];
-
-	data[0] = value;
-	data[1] = PointerGetDatum(kind);
-
-	return hash_bytes((unsigned char *) &data, 2 * SIZEOF_DATUM);
+	/*
+	 * Most resources types store a pointer in 'value', and pointers are
+	 * unique all on their own. But some resources store plain integers (Files
+	 * and Buffers as of this writing), so we want to incorporate the 'kind'
+	 * in the hash too, otherwise those resources will collide a lot. But
+	 * because there are only a few resource kinds like that - and only a few
+	 * resource kinds to begin with - we don't need to work too hard to mix
+	 * 'kind' into the hash. Just add it with hash_combine(), it perturbs the
+	 * result enough for our purposes.
+	 */
+#if SIZEOF_DATUM == 8
+	return hash_combine64(murmurhash64((uint64) value), (uint64) kind);
+#else
+	return hash_combine(murmurhash32((uint32) value), (uint32) kind);
+#endif
 }
 
 static void
diff --git a/src/include/common/hashfn.h b/src/include/common/hashfn.h
index c634cc067a1..009ffbbdd38 100644
--- a/src/include/common/hashfn.h
+++ b/src/include/common/hashfn.h
@@ -101,4 +101,19 @@ murmurhash32(uint32 data)
 	return h;
 }
 
+/* 64-bit variant */
+static inline uint64
+murmurhash64(uint64 data)
+{
+	uint64		h = data;
+
+	h ^= h >> 33;
+	h *= 0xff51afd7ed558ccd;
+	h ^= h >> 33;
+	h *= 0xc4ceb9fe1a85ec53;
+	h ^= h >> 33;
+
+	return h;
+}
+
 #endif							/* HASHFN_H */
-- 
2.29.2


--------------F30A5443A5D524A5CBC51AE7
Content-Type: text/x-patch; charset=UTF-8;
 name="v2-0001-resownerbench.patch"
Content-Transfer-Encoding: 7bit
Content-Disposition: attachment;
 filename="v2-0001-resownerbench.patch"



^ permalink  raw  reply  [nested|flat] 50+ messages in thread

* [PATCH v6 3/3] Optimize hash function
@ 2021-01-20 22:06 Heikki Linnakangas <[email protected]>
  0 siblings, 0 replies; 50+ messages in thread

From: Heikki Linnakangas @ 2021-01-20 22:06 UTC (permalink / raw)

---
 src/backend/utils/resowner/resowner.c | 24 ++++++++++++++++++------
 src/include/common/hashfn.h           | 15 +++++++++++++++
 2 files changed, 33 insertions(+), 6 deletions(-)

diff --git a/src/backend/utils/resowner/resowner.c b/src/backend/utils/resowner/resowner.c
index 1db1e7cf14b..72e4d8a784c 100644
--- a/src/backend/utils/resowner/resowner.c
+++ b/src/backend/utils/resowner/resowner.c
@@ -163,15 +163,27 @@ static void ReleaseAuxProcessResourcesCallback(int code, Datum arg);
  *	  INTERNAL ROUTINES														 *
  *****************************************************************************/
 
+/*
+ * Hash function for value+kind combination.
+ */
 static inline uint32
 hash_resource_elem(Datum value, ResourceOwnerFuncs *kind)
 {
-	Datum		data[2];
-
-	data[0] = value;
-	data[1] = PointerGetDatum(kind);
-
-	return hash_bytes((unsigned char *) &data, 2 * SIZEOF_DATUM);
+	/*
+	 * Most resources types store a pointer in 'value', and pointers are
+	 * unique all on their own. But some resources store plain integers (Files
+	 * and Buffers as of this writing), so we want to incorporate the 'kind'
+	 * in the hash too, otherwise those resources will collide a lot. But
+	 * because there are only a few resource kinds like that - and only a few
+	 * resource kinds to begin with - we don't need to work too hard to mix
+	 * 'kind' into the hash. Just add it with hash_combine(), it perturbs the
+	 * result enough for our purposes.
+	 */
+#if SIZEOF_DATUM == 8
+	return hash_combine64(murmurhash64((uint64) value), (uint64) kind);
+#else
+	return hash_combine(murmurhash32((uint32) value), (uint32) kind);
+#endif
 }
 
 static void
diff --git a/src/include/common/hashfn.h b/src/include/common/hashfn.h
index c634cc067a1..009ffbbdd38 100644
--- a/src/include/common/hashfn.h
+++ b/src/include/common/hashfn.h
@@ -101,4 +101,19 @@ murmurhash32(uint32 data)
 	return h;
 }
 
+/* 64-bit variant */
+static inline uint64
+murmurhash64(uint64 data)
+{
+	uint64		h = data;
+
+	h ^= h >> 33;
+	h *= 0xff51afd7ed558ccd;
+	h ^= h >> 33;
+	h *= 0xc4ceb9fe1a85ec53;
+	h ^= h >> 33;
+
+	return h;
+}
+
 #endif							/* HASHFN_H */
-- 
2.29.2


--------------F30A5443A5D524A5CBC51AE7
Content-Type: text/x-patch; charset=UTF-8;
 name="v2-0001-resownerbench.patch"
Content-Transfer-Encoding: 7bit
Content-Disposition: attachment;
 filename="v2-0001-resownerbench.patch"



^ permalink  raw  reply  [nested|flat] 50+ messages in thread

* [PATCH v7 3/3] Optimize hash function
@ 2021-01-20 22:06 Heikki Linnakangas <[email protected]>
  0 siblings, 0 replies; 50+ messages in thread

From: Heikki Linnakangas @ 2021-01-20 22:06 UTC (permalink / raw)

---
 src/backend/utils/resowner/resowner.c | 24 ++++++++++++++++++------
 src/include/common/hashfn.h           | 15 +++++++++++++++
 2 files changed, 33 insertions(+), 6 deletions(-)

diff --git a/src/backend/utils/resowner/resowner.c b/src/backend/utils/resowner/resowner.c
index d62b1495dd3..6189d051072 100644
--- a/src/backend/utils/resowner/resowner.c
+++ b/src/backend/utils/resowner/resowner.c
@@ -163,15 +163,27 @@ static void ReleaseAuxProcessResourcesCallback(int code, Datum arg);
  *	  INTERNAL ROUTINES														 *
  *****************************************************************************/
 
+/*
+ * Hash function for value+kind combination.
+ */
 static inline uint32
 hash_resource_elem(Datum value, ResourceOwnerFuncs *kind)
 {
-	Datum		data[2];
-
-	data[0] = value;
-	data[1] = PointerGetDatum(kind);
-
-	return hash_bytes((unsigned char *) &data, 2 * SIZEOF_DATUM);
+	/*
+	 * Most resources types store a pointer in 'value', and pointers are
+	 * unique all on their own. But some resources store plain integers (Files
+	 * and Buffers as of this writing), so we want to incorporate the 'kind'
+	 * in the hash too, otherwise those resources will collide a lot. But
+	 * because there are only a few resource kinds like that - and only a few
+	 * resource kinds to begin with - we don't need to work too hard to mix
+	 * 'kind' into the hash. Just add it with hash_combine(), it perturbs the
+	 * result enough for our purposes.
+	 */
+#if SIZEOF_DATUM == 8
+	return hash_combine64(murmurhash64((uint64) value), (uint64) kind);
+#else
+	return hash_combine(murmurhash32((uint32) value), (uint32) kind);
+#endif
 }
 
 static void
diff --git a/src/include/common/hashfn.h b/src/include/common/hashfn.h
index c634cc067a1..009ffbbdd38 100644
--- a/src/include/common/hashfn.h
+++ b/src/include/common/hashfn.h
@@ -101,4 +101,19 @@ murmurhash32(uint32 data)
 	return h;
 }
 
+/* 64-bit variant */
+static inline uint64
+murmurhash64(uint64 data)
+{
+	uint64		h = data;
+
+	h ^= h >> 33;
+	h *= 0xff51afd7ed558ccd;
+	h ^= h >> 33;
+	h *= 0xc4ceb9fe1a85ec53;
+	h ^= h >> 33;
+
+	return h;
+}
+
 #endif							/* HASHFN_H */
-- 
2.30.1


--------------C6A03312D0F99F9183A0DF46--





^ permalink  raw  reply  [nested|flat] 50+ messages in thread

* [PATCH v7 3/3] Optimize hash function
@ 2021-01-20 22:06 Heikki Linnakangas <[email protected]>
  0 siblings, 0 replies; 50+ messages in thread

From: Heikki Linnakangas @ 2021-01-20 22:06 UTC (permalink / raw)

---
 src/backend/utils/resowner/resowner.c | 24 ++++++++++++++++++------
 src/include/common/hashfn.h           | 15 +++++++++++++++
 2 files changed, 33 insertions(+), 6 deletions(-)

diff --git a/src/backend/utils/resowner/resowner.c b/src/backend/utils/resowner/resowner.c
index d62b1495dd3..6189d051072 100644
--- a/src/backend/utils/resowner/resowner.c
+++ b/src/backend/utils/resowner/resowner.c
@@ -163,15 +163,27 @@ static void ReleaseAuxProcessResourcesCallback(int code, Datum arg);
  *	  INTERNAL ROUTINES														 *
  *****************************************************************************/
 
+/*
+ * Hash function for value+kind combination.
+ */
 static inline uint32
 hash_resource_elem(Datum value, ResourceOwnerFuncs *kind)
 {
-	Datum		data[2];
-
-	data[0] = value;
-	data[1] = PointerGetDatum(kind);
-
-	return hash_bytes((unsigned char *) &data, 2 * SIZEOF_DATUM);
+	/*
+	 * Most resources types store a pointer in 'value', and pointers are
+	 * unique all on their own. But some resources store plain integers (Files
+	 * and Buffers as of this writing), so we want to incorporate the 'kind'
+	 * in the hash too, otherwise those resources will collide a lot. But
+	 * because there are only a few resource kinds like that - and only a few
+	 * resource kinds to begin with - we don't need to work too hard to mix
+	 * 'kind' into the hash. Just add it with hash_combine(), it perturbs the
+	 * result enough for our purposes.
+	 */
+#if SIZEOF_DATUM == 8
+	return hash_combine64(murmurhash64((uint64) value), (uint64) kind);
+#else
+	return hash_combine(murmurhash32((uint32) value), (uint32) kind);
+#endif
 }
 
 static void
diff --git a/src/include/common/hashfn.h b/src/include/common/hashfn.h
index c634cc067a1..009ffbbdd38 100644
--- a/src/include/common/hashfn.h
+++ b/src/include/common/hashfn.h
@@ -101,4 +101,19 @@ murmurhash32(uint32 data)
 	return h;
 }
 
+/* 64-bit variant */
+static inline uint64
+murmurhash64(uint64 data)
+{
+	uint64		h = data;
+
+	h ^= h >> 33;
+	h *= 0xff51afd7ed558ccd;
+	h ^= h >> 33;
+	h *= 0xc4ceb9fe1a85ec53;
+	h ^= h >> 33;
+
+	return h;
+}
+
 #endif							/* HASHFN_H */
-- 
2.30.1


--------------C6A03312D0F99F9183A0DF46--





^ permalink  raw  reply  [nested|flat] 50+ messages in thread

* [PATCH v6 3/3] Optimize hash function
@ 2021-01-20 22:06 Heikki Linnakangas <[email protected]>
  0 siblings, 0 replies; 50+ messages in thread

From: Heikki Linnakangas @ 2021-01-20 22:06 UTC (permalink / raw)

---
 src/backend/utils/resowner/resowner.c | 24 ++++++++++++++++++------
 src/include/common/hashfn.h           | 15 +++++++++++++++
 2 files changed, 33 insertions(+), 6 deletions(-)

diff --git a/src/backend/utils/resowner/resowner.c b/src/backend/utils/resowner/resowner.c
index 1db1e7cf14b..72e4d8a784c 100644
--- a/src/backend/utils/resowner/resowner.c
+++ b/src/backend/utils/resowner/resowner.c
@@ -163,15 +163,27 @@ static void ReleaseAuxProcessResourcesCallback(int code, Datum arg);
  *	  INTERNAL ROUTINES														 *
  *****************************************************************************/
 
+/*
+ * Hash function for value+kind combination.
+ */
 static inline uint32
 hash_resource_elem(Datum value, ResourceOwnerFuncs *kind)
 {
-	Datum		data[2];
-
-	data[0] = value;
-	data[1] = PointerGetDatum(kind);
-
-	return hash_bytes((unsigned char *) &data, 2 * SIZEOF_DATUM);
+	/*
+	 * Most resources types store a pointer in 'value', and pointers are
+	 * unique all on their own. But some resources store plain integers (Files
+	 * and Buffers as of this writing), so we want to incorporate the 'kind'
+	 * in the hash too, otherwise those resources will collide a lot. But
+	 * because there are only a few resource kinds like that - and only a few
+	 * resource kinds to begin with - we don't need to work too hard to mix
+	 * 'kind' into the hash. Just add it with hash_combine(), it perturbs the
+	 * result enough for our purposes.
+	 */
+#if SIZEOF_DATUM == 8
+	return hash_combine64(murmurhash64((uint64) value), (uint64) kind);
+#else
+	return hash_combine(murmurhash32((uint32) value), (uint32) kind);
+#endif
 }
 
 static void
diff --git a/src/include/common/hashfn.h b/src/include/common/hashfn.h
index c634cc067a1..009ffbbdd38 100644
--- a/src/include/common/hashfn.h
+++ b/src/include/common/hashfn.h
@@ -101,4 +101,19 @@ murmurhash32(uint32 data)
 	return h;
 }
 
+/* 64-bit variant */
+static inline uint64
+murmurhash64(uint64 data)
+{
+	uint64		h = data;
+
+	h ^= h >> 33;
+	h *= 0xff51afd7ed558ccd;
+	h ^= h >> 33;
+	h *= 0xc4ceb9fe1a85ec53;
+	h ^= h >> 33;
+
+	return h;
+}
+
 #endif							/* HASHFN_H */
-- 
2.29.2


--------------F30A5443A5D524A5CBC51AE7
Content-Type: text/x-patch; charset=UTF-8;
 name="v2-0001-resownerbench.patch"
Content-Transfer-Encoding: 7bit
Content-Disposition: attachment;
 filename="v2-0001-resownerbench.patch"



^ permalink  raw  reply  [nested|flat] 50+ messages in thread

* [PATCH v7 3/3] Optimize hash function
@ 2021-01-20 22:06 Heikki Linnakangas <[email protected]>
  0 siblings, 0 replies; 50+ messages in thread

From: Heikki Linnakangas @ 2021-01-20 22:06 UTC (permalink / raw)

---
 src/backend/utils/resowner/resowner.c | 24 ++++++++++++++++++------
 src/include/common/hashfn.h           | 15 +++++++++++++++
 2 files changed, 33 insertions(+), 6 deletions(-)

diff --git a/src/backend/utils/resowner/resowner.c b/src/backend/utils/resowner/resowner.c
index d62b1495dd3..6189d051072 100644
--- a/src/backend/utils/resowner/resowner.c
+++ b/src/backend/utils/resowner/resowner.c
@@ -163,15 +163,27 @@ static void ReleaseAuxProcessResourcesCallback(int code, Datum arg);
  *	  INTERNAL ROUTINES														 *
  *****************************************************************************/
 
+/*
+ * Hash function for value+kind combination.
+ */
 static inline uint32
 hash_resource_elem(Datum value, ResourceOwnerFuncs *kind)
 {
-	Datum		data[2];
-
-	data[0] = value;
-	data[1] = PointerGetDatum(kind);
-
-	return hash_bytes((unsigned char *) &data, 2 * SIZEOF_DATUM);
+	/*
+	 * Most resources types store a pointer in 'value', and pointers are
+	 * unique all on their own. But some resources store plain integers (Files
+	 * and Buffers as of this writing), so we want to incorporate the 'kind'
+	 * in the hash too, otherwise those resources will collide a lot. But
+	 * because there are only a few resource kinds like that - and only a few
+	 * resource kinds to begin with - we don't need to work too hard to mix
+	 * 'kind' into the hash. Just add it with hash_combine(), it perturbs the
+	 * result enough for our purposes.
+	 */
+#if SIZEOF_DATUM == 8
+	return hash_combine64(murmurhash64((uint64) value), (uint64) kind);
+#else
+	return hash_combine(murmurhash32((uint32) value), (uint32) kind);
+#endif
 }
 
 static void
diff --git a/src/include/common/hashfn.h b/src/include/common/hashfn.h
index c634cc067a1..009ffbbdd38 100644
--- a/src/include/common/hashfn.h
+++ b/src/include/common/hashfn.h
@@ -101,4 +101,19 @@ murmurhash32(uint32 data)
 	return h;
 }
 
+/* 64-bit variant */
+static inline uint64
+murmurhash64(uint64 data)
+{
+	uint64		h = data;
+
+	h ^= h >> 33;
+	h *= 0xff51afd7ed558ccd;
+	h ^= h >> 33;
+	h *= 0xc4ceb9fe1a85ec53;
+	h ^= h >> 33;
+
+	return h;
+}
+
 #endif							/* HASHFN_H */
-- 
2.30.1


--------------C6A03312D0F99F9183A0DF46--





^ permalink  raw  reply  [nested|flat] 50+ messages in thread

* [PATCH v6 3/3] Optimize hash function
@ 2021-01-20 22:06 Heikki Linnakangas <[email protected]>
  0 siblings, 0 replies; 50+ messages in thread

From: Heikki Linnakangas @ 2021-01-20 22:06 UTC (permalink / raw)

---
 src/backend/utils/resowner/resowner.c | 24 ++++++++++++++++++------
 src/include/common/hashfn.h           | 15 +++++++++++++++
 2 files changed, 33 insertions(+), 6 deletions(-)

diff --git a/src/backend/utils/resowner/resowner.c b/src/backend/utils/resowner/resowner.c
index 1db1e7cf14b..72e4d8a784c 100644
--- a/src/backend/utils/resowner/resowner.c
+++ b/src/backend/utils/resowner/resowner.c
@@ -163,15 +163,27 @@ static void ReleaseAuxProcessResourcesCallback(int code, Datum arg);
  *	  INTERNAL ROUTINES														 *
  *****************************************************************************/
 
+/*
+ * Hash function for value+kind combination.
+ */
 static inline uint32
 hash_resource_elem(Datum value, ResourceOwnerFuncs *kind)
 {
-	Datum		data[2];
-
-	data[0] = value;
-	data[1] = PointerGetDatum(kind);
-
-	return hash_bytes((unsigned char *) &data, 2 * SIZEOF_DATUM);
+	/*
+	 * Most resources types store a pointer in 'value', and pointers are
+	 * unique all on their own. But some resources store plain integers (Files
+	 * and Buffers as of this writing), so we want to incorporate the 'kind'
+	 * in the hash too, otherwise those resources will collide a lot. But
+	 * because there are only a few resource kinds like that - and only a few
+	 * resource kinds to begin with - we don't need to work too hard to mix
+	 * 'kind' into the hash. Just add it with hash_combine(), it perturbs the
+	 * result enough for our purposes.
+	 */
+#if SIZEOF_DATUM == 8
+	return hash_combine64(murmurhash64((uint64) value), (uint64) kind);
+#else
+	return hash_combine(murmurhash32((uint32) value), (uint32) kind);
+#endif
 }
 
 static void
diff --git a/src/include/common/hashfn.h b/src/include/common/hashfn.h
index c634cc067a1..009ffbbdd38 100644
--- a/src/include/common/hashfn.h
+++ b/src/include/common/hashfn.h
@@ -101,4 +101,19 @@ murmurhash32(uint32 data)
 	return h;
 }
 
+/* 64-bit variant */
+static inline uint64
+murmurhash64(uint64 data)
+{
+	uint64		h = data;
+
+	h ^= h >> 33;
+	h *= 0xff51afd7ed558ccd;
+	h ^= h >> 33;
+	h *= 0xc4ceb9fe1a85ec53;
+	h ^= h >> 33;
+
+	return h;
+}
+
 #endif							/* HASHFN_H */
-- 
2.29.2


--------------F30A5443A5D524A5CBC51AE7
Content-Type: text/x-patch; charset=UTF-8;
 name="v2-0001-resownerbench.patch"
Content-Transfer-Encoding: 7bit
Content-Disposition: attachment;
 filename="v2-0001-resownerbench.patch"



^ permalink  raw  reply  [nested|flat] 50+ messages in thread

* [PATCH v6 3/3] Optimize hash function
@ 2021-01-20 22:06 Heikki Linnakangas <[email protected]>
  0 siblings, 0 replies; 50+ messages in thread

From: Heikki Linnakangas @ 2021-01-20 22:06 UTC (permalink / raw)

---
 src/backend/utils/resowner/resowner.c | 24 ++++++++++++++++++------
 src/include/common/hashfn.h           | 15 +++++++++++++++
 2 files changed, 33 insertions(+), 6 deletions(-)

diff --git a/src/backend/utils/resowner/resowner.c b/src/backend/utils/resowner/resowner.c
index 1db1e7cf14b..72e4d8a784c 100644
--- a/src/backend/utils/resowner/resowner.c
+++ b/src/backend/utils/resowner/resowner.c
@@ -163,15 +163,27 @@ static void ReleaseAuxProcessResourcesCallback(int code, Datum arg);
  *	  INTERNAL ROUTINES														 *
  *****************************************************************************/
 
+/*
+ * Hash function for value+kind combination.
+ */
 static inline uint32
 hash_resource_elem(Datum value, ResourceOwnerFuncs *kind)
 {
-	Datum		data[2];
-
-	data[0] = value;
-	data[1] = PointerGetDatum(kind);
-
-	return hash_bytes((unsigned char *) &data, 2 * SIZEOF_DATUM);
+	/*
+	 * Most resources types store a pointer in 'value', and pointers are
+	 * unique all on their own. But some resources store plain integers (Files
+	 * and Buffers as of this writing), so we want to incorporate the 'kind'
+	 * in the hash too, otherwise those resources will collide a lot. But
+	 * because there are only a few resource kinds like that - and only a few
+	 * resource kinds to begin with - we don't need to work too hard to mix
+	 * 'kind' into the hash. Just add it with hash_combine(), it perturbs the
+	 * result enough for our purposes.
+	 */
+#if SIZEOF_DATUM == 8
+	return hash_combine64(murmurhash64((uint64) value), (uint64) kind);
+#else
+	return hash_combine(murmurhash32((uint32) value), (uint32) kind);
+#endif
 }
 
 static void
diff --git a/src/include/common/hashfn.h b/src/include/common/hashfn.h
index c634cc067a1..009ffbbdd38 100644
--- a/src/include/common/hashfn.h
+++ b/src/include/common/hashfn.h
@@ -101,4 +101,19 @@ murmurhash32(uint32 data)
 	return h;
 }
 
+/* 64-bit variant */
+static inline uint64
+murmurhash64(uint64 data)
+{
+	uint64		h = data;
+
+	h ^= h >> 33;
+	h *= 0xff51afd7ed558ccd;
+	h ^= h >> 33;
+	h *= 0xc4ceb9fe1a85ec53;
+	h ^= h >> 33;
+
+	return h;
+}
+
 #endif							/* HASHFN_H */
-- 
2.29.2


--------------F30A5443A5D524A5CBC51AE7
Content-Type: text/x-patch; charset=UTF-8;
 name="v2-0001-resownerbench.patch"
Content-Transfer-Encoding: 7bit
Content-Disposition: attachment;
 filename="v2-0001-resownerbench.patch"



^ permalink  raw  reply  [nested|flat] 50+ messages in thread

* [PATCH v10 3/3] Optimize hash function
@ 2021-10-18 10:59 Heikki Linnakangas <[email protected]>
  0 siblings, 0 replies; 50+ messages in thread

From: Heikki Linnakangas @ 2021-10-18 10:59 UTC (permalink / raw)

---
 src/backend/utils/resowner/resowner.c | 24 ++++++++++++++++++------
 src/include/common/hashfn.h           | 15 +++++++++++++++
 2 files changed, 33 insertions(+), 6 deletions(-)

diff --git a/src/backend/utils/resowner/resowner.c b/src/backend/utils/resowner/resowner.c
index 35e824d4534..118bd2b34c9 100644
--- a/src/backend/utils/resowner/resowner.c
+++ b/src/backend/utils/resowner/resowner.c
@@ -169,15 +169,27 @@ static void ReleaseAuxProcessResourcesCallback(int code, Datum arg);
  *	  INTERNAL ROUTINES														 *
  *****************************************************************************/
 
+/*
+ * Hash function for value+kind combination.
+ */
 static inline uint32
 hash_resource_elem(Datum value, ResourceOwnerFuncs *kind)
 {
-	Datum		data[2];
-
-	data[0] = value;
-	data[1] = PointerGetDatum(kind);
-
-	return hash_bytes((unsigned char *) &data, 2 * SIZEOF_DATUM);
+	/*
+	 * Most resources types store a pointer in 'value', and pointers are
+	 * unique all on their own.  But some resources store plain integers
+	 * (Files and Buffers as of this writing), so we want to incorporate the
+	 * 'kind' in the hash too, otherwise those resources will collide a lot.
+	 * But because there are only a few resource kinds like that - and only a
+	 * few resource kinds to begin with - we don't need to work too hard to
+	 * mix 'kind' into the hash.  Just add it with hash_combine(), it perturbs
+	 * the result enough for our purposes.
+	 */
+#if SIZEOF_DATUM == 8
+	return hash_combine64(murmurhash64((uint64) value), (uint64) kind);
+#else
+	return hash_combine(murmurhash32((uint32) value), (uint32) kind);
+#endif
 }
 
 /*
diff --git a/src/include/common/hashfn.h b/src/include/common/hashfn.h
index c634cc067a1..009ffbbdd38 100644
--- a/src/include/common/hashfn.h
+++ b/src/include/common/hashfn.h
@@ -101,4 +101,19 @@ murmurhash32(uint32 data)
 	return h;
 }
 
+/* 64-bit variant */
+static inline uint64
+murmurhash64(uint64 data)
+{
+	uint64		h = data;
+
+	h ^= h >> 33;
+	h *= 0xff51afd7ed558ccd;
+	h ^= h >> 33;
+	h *= 0xc4ceb9fe1a85ec53;
+	h ^= h >> 33;
+
+	return h;
+}
+
 #endif							/* HASHFN_H */
-- 
2.30.2


--------------AF226508B1039DE40796DE51--





^ permalink  raw  reply  [nested|flat] 50+ messages in thread

* Real config values for bytes needs quotes?
@ 2023-02-27 08:32 Kyotaro Horiguchi <[email protected]>
  2023-03-02 10:54 ` Re: Real config values for bytes needs quotes? Peter Eisentraut <[email protected]>
  0 siblings, 1 reply; 50+ messages in thread

From: Kyotaro Horiguchi @ 2023-02-27 08:32 UTC (permalink / raw)
  To: [email protected]

Hello.

I found it frustrating that the line "shared_buffers = 0.1GB" in
postgresql.conf postgresql.conf was causing an error and that the
value required (additional) surrounding single quotes.  The attached
patch makes the parser accept the use of non-quoted real values
followed by a unit for such variables. I'm not sure if that syntax
fully covers the input syntax of strtod, but I beieve it is suffucient
for most use cases.

Is the following a correct English sentense?

Do you folks think this makes sense?


regards.

-- 
Kyotaro Horiguchi
NTT Open Source Software Center


Attachments:

  [text/x-patch] guc-accept-real-values-follows-by-unit.patch (518B, ../../[email protected]/2-guc-accept-real-values-follows-by-unit.patch)
  download | inline diff:
diff --git a/src/backend/utils/misc/guc-file.l b/src/backend/utils/misc/guc-file.l
index 41d62a9f23..6727f1952f 100644
--- a/src/backend/utils/misc/guc-file.l
+++ b/src/backend/utils/misc/guc-file.l
@@ -76,7 +76,7 @@ UNIT_LETTER		[a-zA-Z]
 INTEGER			{SIGN}?({DIGIT}+|0x{HEXDIGIT}+){UNIT_LETTER}*
 
 EXPONENT		[Ee]{SIGN}?{DIGIT}+
-REAL			{SIGN}?{DIGIT}*"."{DIGIT}*{EXPONENT}?
+REAL			{SIGN}?{DIGIT}*("."{DIGIT}*{EXPONENT}?|{EXPONENT}){UNIT_LETTER}*
 
 LETTER			[A-Za-z_\200-\377]
 LETTER_OR_DIGIT [A-Za-z_0-9\200-\377]


^ permalink  raw  reply  [nested|flat] 50+ messages in thread

* Re: Real config values for bytes needs quotes?
  2023-02-27 08:32 Real config values for bytes needs quotes? Kyotaro Horiguchi <[email protected]>
@ 2023-03-02 10:54 ` Peter Eisentraut <[email protected]>
  2023-03-03 05:38   ` Re: Real config values for bytes needs quotes? Kyotaro Horiguchi <[email protected]>
  0 siblings, 1 reply; 50+ messages in thread

From: Peter Eisentraut @ 2023-03-02 10:54 UTC (permalink / raw)
  To: Kyotaro Horiguchi <[email protected]>; [email protected]

On 27.02.23 09:32, Kyotaro Horiguchi wrote:
> I found it frustrating that the line "shared_buffers = 0.1GB" in
> postgresql.conf postgresql.conf was causing an error and that the
> value required (additional) surrounding single quotes.  The attached
> patch makes the parser accept the use of non-quoted real values
> followed by a unit for such variables. I'm not sure if that syntax
> fully covers the input syntax of strtod, but I beieve it is suffucient
> for most use cases.

This seems sensible to fix.  If you're not sure about the details, write 
some test cases. :)






^ permalink  raw  reply  [nested|flat] 50+ messages in thread

* Re: Real config values for bytes needs quotes?
  2023-02-27 08:32 Real config values for bytes needs quotes? Kyotaro Horiguchi <[email protected]>
  2023-03-02 10:54 ` Re: Real config values for bytes needs quotes? Peter Eisentraut <[email protected]>
@ 2023-03-03 05:38   ` Kyotaro Horiguchi <[email protected]>
  0 siblings, 0 replies; 50+ messages in thread

From: Kyotaro Horiguchi @ 2023-03-03 05:38 UTC (permalink / raw)
  To: [email protected]; +Cc: [email protected]

At Thu, 2 Mar 2023 11:54:10 +0100, Peter Eisentraut <[email protected]> wrote in 
> On 27.02.23 09:32, Kyotaro Horiguchi wrote:
> > I found it frustrating that the line "shared_buffers = 0.1GB" in
> > postgresql.conf postgresql.conf was causing an error and that the
> > value required (additional) surrounding single quotes.  The attached
> > patch makes the parser accept the use of non-quoted real values
> > followed by a unit for such variables. I'm not sure if that syntax
> > fully covers the input syntax of strtod, but I beieve it is suffucient
> > for most use cases.
> 
> This seems sensible to fix.  If you're not sure about the details,
> write some test cases. :)

Thanks. I initially intended to limit the change for REAL to accept
units following a value. However, actually I also modified it to
accept '1e1'.

man strtod says it accepts the following format.

> The  expected  form  of the (initial portion of the) string is optional
> leading white space as recognized by isspace(3), an optional plus ('+')
> or  minus  sign  ('-')  and then either (i) a decimal number, or (ii) a
> hexadecimal number, or (iii) an infinity, or (iv) a NAN (not-a-number).
>
> A decimal number consists of a nonempty sequence of decimal digits pos‐
> sibly  containing  a  radix character (decimal point, locale-dependent,
> usually '.'), optionally followed by a  decimal  exponent.   A  decimal
> exponent  consists  of  an  'E' or 'e', followed by an optional plus or
> minus sign, followed by a nonempty  sequence  of  decimal  digits,  and
> indicates multiplication by a power of 10.

It is written in regexp as
'\s*[-+]?(\.\d+|\d+(\.\d*)?)([Ee][-+]?\d+)?'.  The leading whitespace
is unnecessary in this specific usage, and we also need to exclude
INTERGER from this set of matches. Therefore, it should be modified to
'[-+]?((\.\d+|\d+\.\d*)([Ee][-+]?\d+)?|\d+([Ee][-+]?\d+))'.

It is translated to the following BNF notation. UNIT_LETTER is also
needed.

{SIGN}?(("."{DIGIT}+|{DIGIT}+"."{DIGIT}*){EXPONENT}?|{DIGIT}+{EXPONENT}){UNIT_LETTER}*

The attached patch applies aforementioned change to guc-file.l and
includes tests for values in certain patters.

regards.

-- 
Kyotaro Horiguchi
NTT Open Source Software Center


Attachments:

  [text/x-patch] v1-0001-Widen-unquoted-value-acceptance-in-configuration-.patch (2.8K, ../../[email protected]/2-v1-0001-Widen-unquoted-value-acceptance-in-configuration-.patch)
  download | inline diff:
From 9a413b791364152d44838d329f0efe3cc7db4d22 Mon Sep 17 00:00:00 2001
From: Kyotaro Horiguchi <[email protected]>
Date: Fri, 3 Mar 2023 13:39:26 +0900
Subject: [PATCH v1] Widen unquoted value acceptance in configuration file

Numeric configuration parameters can accept real numbers like '10.3',
but the values cannot be followed by a unit if they are not quoted. In
addition, certain formats like '1e4' are not accepted unless they are
not quoted. This commit changes the configuration file parser to
accept all of these values without the need for quotes.
---
 src/backend/utils/misc/guc-file.l             |  2 +-
 src/test/modules/test_misc/t/003_check_guc.pl | 41 +++++++++++++++++++
 2 files changed, 42 insertions(+), 1 deletion(-)

diff --git a/src/backend/utils/misc/guc-file.l b/src/backend/utils/misc/guc-file.l
index 41d62a9f23..ca4add9710 100644
--- a/src/backend/utils/misc/guc-file.l
+++ b/src/backend/utils/misc/guc-file.l
@@ -76,7 +76,7 @@ UNIT_LETTER		[a-zA-Z]
 INTEGER			{SIGN}?({DIGIT}+|0x{HEXDIGIT}+){UNIT_LETTER}*
 
 EXPONENT		[Ee]{SIGN}?{DIGIT}+
-REAL			{SIGN}?{DIGIT}*"."{DIGIT}*{EXPONENT}?
+REAL			{SIGN}?(("."{DIGIT}+|{DIGIT}+"."{DIGIT}*){EXPONENT}?|{DIGIT}+{EXPONENT}){UNIT_LETTER}*
 
 LETTER			[A-Za-z_\200-\377]
 LETTER_OR_DIGIT [A-Za-z_0-9\200-\377]
diff --git a/src/test/modules/test_misc/t/003_check_guc.pl b/src/test/modules/test_misc/t/003_check_guc.pl
index e9f33f3c77..053fbe336d 100644
--- a/src/test/modules/test_misc/t/003_check_guc.pl
+++ b/src/test/modules/test_misc/t/003_check_guc.pl
@@ -106,4 +106,45 @@ foreach my $param (@sample_intersect)
 	);
 }
 
+# check if GUC parser accepts some unquoted values accepted by strtod()
+my $conffile = $node->data_dir . "/postgresql.conf";
+my $conf = slurp_file($conffile);
+## without units
+my $lines =
+  join("\n", map {"checkpoint_completion_target = $_"}
+	   ('0.1', '1.', '.1', '1e-1', '.1e0', '1.e-01'));
+## with units
+$lines .= "\n" .
+  join("\n", map {"work_mem = ${_}MB"}
+	   ('0.3', '1.', '.3', '+3e-1', '-.3e0', '1.e01'));
+$node->append_conf('postgresql.conf', $lines);
+
+$node->reload();
+## has the last value been read?
+ok ($node->poll_query_until('postgres',
+			"SELECT setting::int = 10240 FROM pg_settings WHERE name = 'work_mem'"),
+	'all variations of real number parameter values passed');
+
+# check if parser correctly rejects some invalid patterns
+my $n = 0;
+foreach my $val ('.', '.e3')
+{
+	$n++ if (test_one_value($node, $val, $conf));
+}
+ok($n == 2, "parser succesfully rejected invalid values");
+
 done_testing();
+
+sub test_one_value
+{
+  my ($node, $value, $base_conf) = @_;
+
+  open(my $fh, '>', $conffile) || die "failed to open $conffile";
+  print $fh $conf . "\nwork_mem=$value\n";
+  close($fh);
+
+  my $logstart = -s $node->logfile;
+  $node->reload();
+  return $node->wait_for_log('syntax error in file ".*postgresql.conf"',
+							 $logstart);
+}
-- 
2.31.1



^ permalink  raw  reply  [nested|flat] 50+ messages in thread

end of thread, other threads:[~2023-03-03 05:38 UTC | newest]

Thread overview: 50+ messages (download: mbox mbox.gz follow: Atom feed)
-- links below jump to the message on this page --
2021-01-20 22:06 [PATCH v6 3/3] Optimize hash function Heikki Linnakangas <[email protected]>
2021-01-20 22:06 [PATCH v6 3/3] Optimize hash function Heikki Linnakangas <[email protected]>
2021-01-20 22:06 [PATCH v6 3/3] Optimize hash function Heikki Linnakangas <[email protected]>
2021-01-20 22:06 [PATCH v7 3/3] Optimize hash function Heikki Linnakangas <[email protected]>
2021-01-20 22:06 [PATCH v7 3/3] Optimize hash function Heikki Linnakangas <[email protected]>
2021-01-20 22:06 [PATCH v6 3/3] Optimize hash function Heikki Linnakangas <[email protected]>
2021-01-20 22:06 [PATCH v7 3/3] Optimize hash function Heikki Linnakangas <[email protected]>
2021-01-20 22:06 [PATCH v7 3/3] Optimize hash function Heikki Linnakangas <[email protected]>
2021-01-20 22:06 [PATCH v7 3/3] Optimize hash function Heikki Linnakangas <[email protected]>
2021-01-20 22:06 [PATCH v6 3/3] Optimize hash function Heikki Linnakangas <[email protected]>
2021-01-20 22:06 [PATCH v7 3/3] Optimize hash function Heikki Linnakangas <[email protected]>
2021-01-20 22:06 [PATCH v7 3/3] Optimize hash function Heikki Linnakangas <[email protected]>
2021-01-20 22:06 [PATCH v6 3/3] Optimize hash function Heikki Linnakangas <[email protected]>
2021-01-20 22:06 [PATCH v6 3/3] Optimize hash function Heikki Linnakangas <[email protected]>
2021-01-20 22:06 [PATCH v6 3/3] Optimize hash function Heikki Linnakangas <[email protected]>
2021-01-20 22:06 [PATCH v7 3/3] Optimize hash function Heikki Linnakangas <[email protected]>
2021-01-20 22:06 [PATCH v7 3/3] Optimize hash function Heikki Linnakangas <[email protected]>
2021-01-20 22:06 [PATCH v6 3/3] Optimize hash function Heikki Linnakangas <[email protected]>
2021-01-20 22:06 [PATCH v7 3/3] Optimize hash function Heikki Linnakangas <[email protected]>
2021-01-20 22:06 [PATCH v6 3/3] Optimize hash function Heikki Linnakangas <[email protected]>
2021-01-20 22:06 [PATCH v6 3/3] Optimize hash function Heikki Linnakangas <[email protected]>
2021-01-20 22:06 [PATCH v6 3/3] Optimize hash function Heikki Linnakangas <[email protected]>
2021-01-20 22:06 [PATCH v6 3/3] Optimize hash function Heikki Linnakangas <[email protected]>
2021-01-20 22:06 [PATCH v6 3/3] Optimize hash function Heikki Linnakangas <[email protected]>
2021-01-20 22:06 [PATCH v6 3/3] Optimize hash function Heikki Linnakangas <[email protected]>
2021-01-20 22:06 [PATCH v7 3/3] Optimize hash function Heikki Linnakangas <[email protected]>
2021-01-20 22:06 [PATCH v7 3/3] Optimize hash function Heikki Linnakangas <[email protected]>
2021-01-20 22:06 [PATCH v8 3/3] Optimize hash function Heikki Linnakangas <[email protected]>
2021-01-20 22:06 [PATCH v7 3/3] Optimize hash function Heikki Linnakangas <[email protected]>
2021-01-20 22:06 [PATCH v7 3/3] Optimize hash function Heikki Linnakangas <[email protected]>
2021-01-20 22:06 [PATCH v7 3/3] Optimize hash function Heikki Linnakangas <[email protected]>
2021-01-20 22:06 [PATCH v7 3/3] Optimize hash function Heikki Linnakangas <[email protected]>
2021-01-20 22:06 [PATCH v7 3/3] Optimize hash function Heikki Linnakangas <[email protected]>
2021-01-20 22:06 [PATCH v7 3/3] Optimize hash function Heikki Linnakangas <[email protected]>
2021-01-20 22:06 [PATCH v7 3/3] Optimize hash function Heikki Linnakangas <[email protected]>
2021-01-20 22:06 [PATCH v6 3/3] Optimize hash function Heikki Linnakangas <[email protected]>
2021-01-20 22:06 [PATCH v7 3/3] Optimize hash function Heikki Linnakangas <[email protected]>
2021-01-20 22:06 [PATCH v7 3/3] Optimize hash function Heikki Linnakangas <[email protected]>
2021-01-20 22:06 [PATCH v6 3/3] Optimize hash function Heikki Linnakangas <[email protected]>
2021-01-20 22:06 [PATCH v6 3/3] Optimize hash function Heikki Linnakangas <[email protected]>
2021-01-20 22:06 [PATCH v7 3/3] Optimize hash function Heikki Linnakangas <[email protected]>
2021-01-20 22:06 [PATCH v7 3/3] Optimize hash function Heikki Linnakangas <[email protected]>
2021-01-20 22:06 [PATCH v6 3/3] Optimize hash function Heikki Linnakangas <[email protected]>
2021-01-20 22:06 [PATCH v7 3/3] Optimize hash function Heikki Linnakangas <[email protected]>
2021-01-20 22:06 [PATCH v6 3/3] Optimize hash function Heikki Linnakangas <[email protected]>
2021-01-20 22:06 [PATCH v6 3/3] Optimize hash function Heikki Linnakangas <[email protected]>
2021-10-18 10:59 [PATCH v10 3/3] Optimize hash function Heikki Linnakangas <[email protected]>
2023-02-27 08:32 Real config values for bytes needs quotes? Kyotaro Horiguchi <[email protected]>
2023-03-02 10:54 ` Re: Real config values for bytes needs quotes? Peter Eisentraut <[email protected]>
2023-03-03 05:38   ` Re: Real config values for bytes needs quotes? Kyotaro Horiguchi <[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