v4-0002-Use-Generation-memory-contexts-to-store-tuples-in.patch

text/plain

Filename: v4-0002-Use-Generation-memory-contexts-to-store-tuples-in.patch
Type: text/plain
Part: 1
Message: Re: Use generation context to speed up tuplesorts

Patch

Format: format-patch
Series: patch v4-0002
Subject: Use Generation memory contexts to store tuples in sorts
File+
src/backend/utils/sort/tuplesort.c 3 3
From d628112d6ac7cc5c6956fa8975c36b9c1e6060d9 Mon Sep 17 00:00:00 2001
From: David Rowley <drowley@postgresql.org>
Date: Wed, 23 Mar 2022 02:24:00 +1300
Subject: [PATCH v4 2/2] Use Generation memory contexts to store tuples in
 sorts

The general useage pattern when we store tuples in tuplesort.c is that
we store a series of tuples one by one then either perform a sort or spill
them to disk.  In the common case there is no pfreeing of already stored
tuples.  For the common case since we do not individually pfree tuples we
have very little need for aset.c memory allocation behavior which
maintains freelists and always rounds allocation sizes up to the next
power of 2 size.

Here we unconditionally switch to using the Generation context type,
which does not round allocation sizes up to the next power of 2.  This
results in more dense memory allocations and on average uses around 25%
less memory.  This can result in some fairly good performance improvements
for sorting.  Performance improvements are fairly significant in cases
where we no longer spill to disk due to the more compact memory
allocation. However, performnace improvements are also quite promising in
other cases as well.

Author: David Rowley
Discussion: https://postgr.es/m/CAApHDvoH4ASzsAOyHcxkuY01Qf++8JJ0paw+03dk+W25tQEcNQ@mail.gmail.com
---
 src/backend/utils/sort/tuplesort.c | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/src/backend/utils/sort/tuplesort.c b/src/backend/utils/sort/tuplesort.c
index 086e948fca..df4f5ad6c5 100644
--- a/src/backend/utils/sort/tuplesort.c
+++ b/src/backend/utils/sort/tuplesort.c
@@ -845,9 +845,9 @@ tuplesort_begin_batch(Tuplesortstate *state)
 	 * in the parent context, not this context, because there is no need to
 	 * free memtuples early.
 	 */
-	state->tuplecontext = AllocSetContextCreate(state->sortcontext,
-												"Caller tuples",
-												ALLOCSET_DEFAULT_SIZES);
+	state->tuplecontext = GenerationContextCreate(state->sortcontext,
+												  "Caller tuples",
+												  ALLOCSET_DEFAULT_SIZES);
 
 	state->status = TSS_INITIAL;
 	state->bounded = false;
-- 
2.32.0