postgresql/src/backend/utils
David Rowley 99c754129d Fix performance regression in tuplesort specializations
697492434 added 3 new qsort specialization functions aimed to improve the
performance of sorting many of the common pass-by-value data types when
they're the leading or only sort key.

Unfortunately, that has caused a performance regression when sorting
datasets where many of the values being compared were equal.  What was
happening here was that we were falling back to the standard sort
comparison function to handle tiebreaks.  When the two given Datums
compared equally we would incur both the overhead of an indirect function
call to the standard comparer to perform the tiebreak and also the
standard comparer function would go and compare the leading key needlessly
all over again.

Here improve the situation in the 3 new comparison functions.  We now
return 0 directly when the two Datums compare equally and we're performing
a 1-key sort.

Here we don't do anything to help the multi-key sort case where the
leading key uses one of the sort specializations functions.  On testing
this case, even when the leading key's values are all equal, there
appeared to be no performance regression.  Let's leave it up to future
work to optimize that case so that the tiebreak function no longer
re-compares the leading key over again.

Another possible fix for this would have been to add 3 additional sort
specialization functions to handle single-key sorts for these
pass-by-value types.  The reason we didn't do that here is that we may
deem some other sort specialization to be more useful than single-key
sorts.  It may be impractical to have sort specialization functions for
every single combination of what may be useful and it was already decided
that further analysis into which ones are the most useful would be delayed
until the v16 cycle.  Let's not let this regression force our hand into
trying to make that decision for v15.

Author: David Rowley
Reviewed-by: John Naylor
Discussion: https://postgr.es/m/CA+hUKGJRbzaAOUtBUcjF5hLtaSHnJUqXmtiaLEoi53zeWSizeA@mail.gmail.com
2022-04-22 16:02:15 +12:00
..
activity pgstat: Use correct lock level in pgstat_drop_all_entries(). 2022-04-16 14:44:58 -07:00
adt Disallow infinite endpoints in generate_series() for timestamps. 2022-04-20 18:08:23 -04:00
cache Remove extraneous blank lines before block-closing braces 2022-04-13 19:16:02 +02:00
error Remove extraneous blank lines before block-closing braces 2022-04-13 19:16:02 +02:00
fmgr Fix various typos and spelling mistakes in code comments 2022-04-11 20:49:41 +12:00
hash Update copyright for 2022 2022-01-07 19:04:57 -05:00
init Rethink method for assigning OIDs to the template0 and postgres DBs. 2022-04-21 16:23:15 -04:00
mb Update copyright for 2022 2022-01-07 19:04:57 -05:00
misc Remove extraneous blank lines before block-closing braces 2022-04-13 19:16:02 +02:00
mmgr Remove extraneous blank lines before block-closing braces 2022-04-13 19:16:02 +02:00
resowner Remove extraneous blank lines before block-closing braces 2022-04-13 19:16:02 +02:00
sort Fix performance regression in tuplesort specializations 2022-04-22 16:02:15 +12:00
time Fix incorrect logic in HaveRegisteredOrActiveSnapshot(). 2022-04-16 16:04:50 -04:00
.gitignore Rearrange makefile rules for running Gen_fmgrtab.pl. 2018-05-03 17:54:18 -04:00
errcodes.txt Update copyright for 2022 2022-01-07 19:04:57 -05:00
Gen_dummy_probes.pl Update copyright for 2022 2022-01-07 19:04:57 -05:00
Gen_dummy_probes.pl.prolog Update copyright for 2022 2022-01-07 19:04:57 -05:00
Gen_dummy_probes.sed Update copyright for 2022 2022-01-07 19:04:57 -05:00
Gen_fmgrtab.pl Update copyright for 2022 2022-01-07 19:04:57 -05:00
generate-errcodes.pl Update copyright for 2022 2022-01-07 19:04:57 -05:00
Makefile Correct Makefile dependencies for catalog scripts 2022-02-14 09:07:09 +07:00
probes.d Update copyright for 2022 2022-01-07 19:04:57 -05:00
README.Gen_dummy_probes Tweak generation of Gen_dummy_probes.pl 2021-05-11 20:02:02 -04:00

# Generating dummy probes

If Postgres isn't configured with dtrace enabled, we need to generate
dummy probes for the entries in probes.d, that do nothing.

This is accomplished in Unix via the sed script `Gen_dummy_probes.sed`. We
used to use this in MSVC builds using the perl utility `psed`, which mimicked
sed. However, that utility disappeared from Windows perl distributions and so
we converted the sed script to a perl script to be used in MSVC builds.

We still keep the sed script as the authoritative source for generating
these dummy probes because except on Windows perl is not a hard requirement
when building from a tarball.

So, if you need to change the way dummy probes are generated, first change
the sed script, and when it's working generate the perl script. This can
be accomplished by using the perl utility s2p.

s2p is no longer part of the perl core, so it might not be on your system,
but it is available on CPAN and also in many package systems. e.g.
on Fedora it can be installed using `cpan App::s2p` or
`dnf install perl-App-s2p`.

The Makefile contains a recipe for regenerating Gen_dummy_probes.pl, so all
you need to do is once you have s2p installed is `make Gen_dummy_probes.pl`
Note that in a VPATH build this will generate the file in the vpath tree,
not the source tree.