| Age | Commit message (Collapse) | Author |
|---|
|
The hex_encode() and hex_decode() functions serve as the workhorses
for hexadecimal data for bytea's text format conversion functions,
and some workloads are sensitive to their performance. This commit
adds new implementations that use routines from port/simd.h, which
testing indicates are much faster for larger inputs. For small or
invalid inputs, we fall back on the existing scalar versions.
Since we are using port/simd.h, these optimizations apply to both
x86-64 and AArch64.
Author: Nathan Bossart <nathandbossart@gmail.com>
Co-authored-by: Chiranmoy Bhattacharya <chiranmoy.bhattacharya@fujitsu.com>
Co-authored-by: Susmitha Devanga <devanga.susmitha@fujitsu.com>
Reviewed-by: John Naylor <johncnaylorls@gmail.com>
Discussion: https://postgr.es/m/aLhVWTRy0QPbW2tl%40nathan
|
|
Presently, the SIMD implementation of this function uses unsigned
saturating subtraction to find bytes less than or equal to the
given value, which is a workaround for the lack of unsigned
comparison instructions on some architectures. However, Neon
offers vminvq_u8(), which returns the minimum (unsigned) value in
the vector. This commit adds a Neon-specific implementation that
uses vminvq_u8() to optimize vector8_has_le() on AArch64.
In passing, adjust the SSE2 implementation to use vector8_min() and
vector8_eq() to find values less than or equal to the given value.
This was the only use of vector8_ssub(), so it has been removed.
Reviewed-by: John Naylor <johncnaylorls@gmail.com>
Discussion: https://postgr.es/m/aNHDNDSHleq0ogC_%40nathan
|
|
Per discussion, this compiler suite is no longer maintained, and
it has not been able to compile PostgreSQL since at least PostgreSQL
17.
This removes all the remaining support code for this compiler.
Note that the Solaris operating system continues to be supported, but
using GCC as the compiler.
Reviewed-by: Andres Freund <andres@anarazel.de>
Reviewed-by: Tom Lane <tgl@sss.pgh.pa.us>
Discussion: https://www.postgresql.org/message-id/flat/a0f817ee-fb86-483a-8a14-b6f7f5991b6e%40eisentraut.org
|
|
Since commit f5e0186f865 (Raise C requirement to C11), we effectively
require at least GCC version 4.7, so checks for older versions can be
removed.
Reviewed-by: Andres Freund <andres@anarazel.de>
Reviewed-by: Tom Lane <tgl@sss.pgh.pa.us>
Discussion: https://www.postgresql.org/message-id/flat/a0f817ee-fb86-483a-8a14-b6f7f5991b6e%40eisentraut.org
|
|
Remove conditionally-compiled code for smaller Datum widths,
and simplify comments that describe cases no longer of interest.
I also fixed up a few more places that were not using
DatumGetIntXX where they should, and made some cosmetic
adjustments such as using sizeof(int64) not sizeof(Datum)
in places where that fit better with the surrounding code.
One thing I remembered while preparing this part is that SP-GiST
stores pass-by-value prefix keys as Datums, so that the on-disk
representation depends on sizeof(Datum). That's even more
unfortunate than the existing commentary makes it out to be,
because now there is a hazard that the change of sizeof(Datum)
will break SP-GiST indexes on 32-bit machines. It appears that
there are no existing SP-GiST opclasses that are actually
affected; and if there are some that I didn't find, the number
of installations that are using them on 32-bit machines is
doubtless tiny. So I'm proceeding on the assumption that we
can get away with this, but it's something to worry about.
(gininsert.c looks like it has a similar problem, but it's okay
because the "tuples" it's constructing are just transient data
within the tuplesort step. That's pretty poorly documented
though, so I added some comments.)
Author: Tom Lane <tgl@sss.pgh.pa.us>
Reviewed-by: Peter Eisentraut <peter@eisentraut.org>
Discussion: https://postgr.es/m/1749799.1752797397@sss.pgh.pa.us
|
|
Solaris has never bothered to add "const" to the second argument
of PAM conversation procs, as all other Unixen did decades ago.
This resulted in an "incompatible pointer" compiler warning when
building --with-pam, but had no more serious effect than that,
so we never did anything about it. However, as of GCC 14 the
case is an error not warning by default.
To complicate matters, recent OpenIndiana (and maybe illumos
in general?) *does* supply the "const" by default, so we can't
just assume that platforms using our solaris template need help.
What we can do, short of building a configure-time probe,
is to make solaris.h #define _PAM_LEGACY_NONCONST, which
causes OpenIndiana's pam_appl.h to revert to the traditional
definition, and hopefully will have no effect anywhere else.
Then we can use that same symbol to control whether we include
"const" in the declaration of pam_passwd_conv_proc().
Bug: #18995
Reported-by: Andrew Watkins <awatkins1966@gmail.com>
Author: Tom Lane <tgl@sss.pgh.pa.us>
Discussion: https://postgr.es/m/18995-82058da9ab4337a7@postgresql.org
Backpatch-through: 13
|
|
Commit e2809e3a101 added code to a header which assigns a pointer
to void to a pointer to unsigned char. This causes build errors for
extensions written in C++. Fix by adding an explicit cast.
Reviewed-by: Tom Lane <tgl@sss.pgh.pa.us>
Discussion: https://postgr.es/m/CANWCAZaCq9AHBuhs%3DMx7Gg_0Af9oRU7iAqr0itJCtfmsWwVmnQ%40mail.gmail.com
Backpatch-through: 18
|
|
We stopped defining IOV_MAX on non-Windows systems in 75357ab94, on
the assumption that every non-Windows system defines it in <limits.h>
as required by X/Open. GNU Hurd, however, doesn't follow that
standard either. Put back the old logic to assume 16 if it's
not defined.
Author: Michael Banck <mbanck@gmx.net>
Co-authored-by: Christoph Berg <myon@debian.org>
Reviewed-by: Tom Lane <tgl@sss.pgh.pa.us>
Discussion: https://postgr.es/m/6862e8d1.050a0220.194b8d.76fa@mx.google.com
Discussion: https://postgr.es/m/6846e0c3.df0a0220.39ef9b.c60e@mx.google.com
Backpatch-through: 16
|
|
Broken by commits 7fe2f67c7c9f, 81f287dc923f and bf1119d74a79. Backpatch
to 18, same as the offending commits.
Backpatch-through: 18
|
|
When querying NUMA status of pages in shared memory, we need to touch
the memory first to get valid results. This may trigger valgrind
reports, because some of the memory (e.g. unpinned buffers) may be
marked as noaccess.
Solved by adding a valgrind suppresion. An alternative would be to
adjust the access/noaccess status before touching the memory, but that
seems far too invasive. It would require all those places to have
detailed knowledge of what the shared memory stores.
The pg_numa_touch_mem_if_required() macro is replaced with a function.
Macros are invisible to suppressions, so it'd have to suppress reports
for the caller - e.g. pg_get_shmem_allocations_numa(). So we'd suppress
reports for the whole function, and that seems to heavy-handed. It might
easily hide other valid issues.
Reviewed-by: Christoph Berg <myon@debian.org>
Reviewed-by: Bertrand Drouvot <bertranddrouvot.pg@gmail.com>
Discussion: https://postgr.es/m/aEtDozLmtZddARdB@msg.df7cb.de
Backpatch-through: 18
|
|
Oversight in commit 3c6e8c123.
|
|
Make sure that function declarations use names that exactly match the
corresponding names from function definitions in a few places. These
inconsistencies were all introduced during Postgres 18 development.
This commit was written with help from clang-tidy, by mechanically
applying the same rules as similar clean-up commits (the earliest such
commit was commit 035ce1fe).
|
|
This moves/renames some of the functions defined in pg_numa.c:
* pg_numa_get_pagesize() is renamed to pg_get_shmem_pagesize(), and
moved to src/backend/storage/ipc/shmem.c. The new name better reflects
that the page size is not related to NUMA, and it's specifically about
the page size used for the main shared memory segment.
* move pg_numa_available() to src/backend/storage/ipc/shmem.c, i.e. into
the backend (which more appropriate for functions callable from SQL).
While at it, improve the comment to explain what page size it returns.
* remove unnecessary includes from src/port/pg_numa.c, adding
unnecessary dependencies (src/port should be suitable for frontent).
These were either leftovers or unnecessary thanks to the other changes
in this commit.
This eliminates unnecessary dependencies on backend symbols, which we
don't want in src/port.
Reported-by: Kirill Reshke <reshkekirill@gmail.com>
Reviewed-by: Andres Freund <andres@anarazel.de>
https://postgr.es/m/CALdSSPi5fj0a7UG7Fmw2cUD1uWuckU_e8dJ+6x-bJEokcSXzqA@mail.gmail.com
|
|
Add basic NUMA awareness routines, using a minimal src/port/pg_numa.c
portability wrapper and an optional build dependency, enabled by
--with-libnuma configure option. For now this is Linux-only, other
platforms may be supported later.
A built-in SQL function pg_numa_available() allows checking NUMA
support, i.e. that the server was built/linked with the NUMA library.
The main function introduced is pg_numa_query_pages(), which allows
determining the NUMA node for individual memory pages. Internally the
function uses move_pages(2) syscall, as it allows batching, and is more
efficient than get_mempolicy(2).
Author: Jakub Wartak <jakub.wartak@enterprisedb.com>
Co-authored-by: Bertrand Drouvot <bertranddrouvot.pg@gmail.com>
Reviewed-by: Andres Freund <andres@anarazel.de>
Reviewed-by: Álvaro Herrera <alvherre@alvh.no-ip.org>
Reviewed-by: Tomas Vondra <tomas@vondra.me>
Discussion: https://postgr.es/m/CAKZiRmxh6KWo0aqRqvmcoaX2jUxZYb4kGp3N%3Dq1w%2BDiH-696Xw%40mail.gmail.com
|
|
The previous implementation of CRC32C on x86 relied on the native
CRC32 instruction from the SSE 4.2 extension, which operates on
up to 8 bytes at a time. We can get a substantial speedup by using
carryless multiplication on SIMD registers, processing 64 bytes per
loop iteration. Shorter inputs fall back to ordinary CRC instructions.
On Intel Tiger Lake hardware (2020), CRC is now 50% faster for inputs
between 64 and 112 bytes, and 3x faster for 256 bytes.
The VPCLMULQDQ instruction on 512-bit registers has been available
on Intel hardware since 2019 and AMD since 2022. There is an older
variant for 128-bit registers, but at least on Zen 2 it performs worse
than normal CRC instructions for short inputs.
We must now do a runtime check, even for builds that target SSE
4.2. This doesn't matter in practice for WAL (arguably the most
critical case), because since commit e2809e3a1 the final computation
with the 20-byte WAL header is inlined and unrolled when targeting
that extension. Compared with two direct function calls, testing
showed equal or slightly faster performance in performing an indirect
function call on several dozen bytes followed by inlined instructions
on constant input of 20 bytes.
The MIT-licensed implementation was generated with the "generate"
program from
https://github.com/corsix/fast-crc32/
Based on: "Fast CRC Computation for Generic Polynomials Using PCLMULQDQ
Instruction" V. Gopal, E. Ozturk, et al., 2009
Co-authored-by: Raghuveer Devulapalli <raghuveer.devulapalli@intel.com>
Co-authored-by: Paul Amonson <paul.d.amonson@intel.com>
Reviewed-by: Nathan Bossart <nathandbossart@gmail.com>
Reviewed-by: Andres Freund <andres@anarazel.de> (earlier version)
Reviewed-by: Matthew Sterrett <matthewsterrett2@gmail.com> (earlier version)
Tested-by: Raghuveer Devulapalli <raghuveer.devulapalli@intel.com>
Tested-by: David Rowley <<dgrowleyml@gmail.com>> (earlier version)
Discussion: https://postgr.es/m/BL1PR11MB530401FA7E9B1CA432CF9DC3DC192@BL1PR11MB5304.namprd11.prod.outlook.com
Discussion: https://postgr.es/m/PH8PR11MB82869FF741DFA4E9A029FF13FBF72@PH8PR11MB8286.namprd11.prod.outlook.com
|
|
On Red Hat 9 systems (or similar), the packaged gcc targets x86-64-v2,
but clang does not. This has caused build failures in the wake of
commit e2809e3a1 when building --with-llvm.
The most expedient fix is to use the same function attributes for
the inlined function as we do for the global function.
Reported-by: Kyotaro Horiguchi <horikyota.ntt@gmail.com> (plus members skimmer and bumblebee)
Diagnosed-by: Tom Lane <tgl@sss.pgh.pa.us>
Tested-by: Todd Cook <cookt@blackduck.com>
Discussion: https://postgr.es/m/CANWCAZZSxs3a1YRKehkgk2OHKbrVn+xZ+AWW8Co2R_f70NqqmA@mail.gmail.com
|
|
pg_crc32c.h now has a simplified copy of the loop in pg_crc32c_sse42.c
suitable for inlining where possible.
This may slightly reduce contention for the WAL insertion lock,
but that hasn't been tested. The motivation for this change is avoid
regressing for a future commit that will use a function pointer for
non-constant input in all x86 builds.
While it's technically possible to make a similar change for Arm and
LoongArch, there are some questions about how inlining should work
since those platforms prefer stricter alignment. There are also no
immediate plans to add additional implementations for them.
Reviewed-by: Nathan Bossart <nathandbossart@gmail.com>
Reviewed-by: Raghuveer Devulapalli <raghuveer.devulapalli@intel.com>
Discussion: https://postgr.es/m/CANWCAZZEiTzhZcuwTiJ2=opiNpAUn1vuDRu1N02z61AthwRZLA@mail.gmail.com
Discussion: https://postgr.es/m/CANWCAZYRhLHArpyfV4uRK-Rw9N5oV5HMkkKtBehcuTjNOMwCZg@mail.gmail.com
|
|
This commit introduces SVE implementations of pg_popcount{32,64}.
Unlike the Neon versions, we need an additional configure-time
check to determine if the compiler supports SVE intrinsics, and we
need a runtime check to determine if the current CPU supports SVE
instructions. Our testing showed that the SVE implementations are
much faster for larger inputs and are comparable to the status
quo for smaller inputs.
Author: "Devanga.Susmitha@fujitsu.com" <Devanga.Susmitha@fujitsu.com>
Co-authored-by: "Chiranmoy.Bhattacharya@fujitsu.com" <Chiranmoy.Bhattacharya@fujitsu.com>
Co-authored-by: "Malladi, Rama" <ramamalladi@hotmail.com>
Reviewed-by: John Naylor <johncnaylorls@gmail.com>
Reviewed-by: Kirill Reshke <reshkekirill@gmail.com>
Discussion: https://postgr.es/m/010101936e4aaa70-b474ab9e-b9ce-474d-a3ba-a3dc223d295c-000000%40us-west-2.amazonses.com
Discussion: https://postgr.es/m/OSZPR01MB84990A9A02A3515C6E85A65B8B2A2%40OSZPR01MB8499.jpnprd01.prod.outlook.com
|
|
This reverts commit 8e993bff5326b00ced137c837fce7cd1e0ecae14.
It causes various build failures on the buildfarm, to be investigated.
Discussion: https://postgr.es/m/CWZBBRR6YA8D.8EHMDRGLCKCD%40neon.tech
|
|
This commit introduces Neon implementations of pg_popcount{32,64},
pg_popcount(), and pg_popcount_masked(). As in simd.h, we assume
that all available AArch64 hardware supports Neon, so we don't need
any new configure-time or runtime checks. Some compilers already
emit Neon instructions for these functions, but our hand-rolled
implementations for pg_popcount() and pg_popcount_masked()
performed better in testing, likely due to better instruction-level
parallelism.
Author: "Chiranmoy.Bhattacharya@fujitsu.com" <Chiranmoy.Bhattacharya@fujitsu.com>
Reviewed-by: John Naylor <johncnaylorls@gmail.com>
Discussion: https://postgr.es/m/010101936e4aaa70-b474ab9e-b9ce-474d-a3ba-a3dc223d295c-000000%40us-west-2.amazonses.com
|
|
This macro protects x86_64-specific code, and a subsequent commit
will introduce AArch64-specific versions of that code. To prevent
confusion, let's rename it to clearly indicate that it's for
x86_64. We should likely move this code to its own file (perhaps
merging it with the AVX-512 popcount code), but that is left as a
future exercise.
Reviewed-by: "Chiranmoy.Bhattacharya@fujitsu.com" <Chiranmoy.Bhattacharya@fujitsu.com>
Reviewed-by: John Naylor <johncnaylorls@gmail.com>
Discussion: https://postgr.es/m/010101936e4aaa70-b474ab9e-b9ce-474d-a3ba-a3dc223d295c-000000%40us-west-2.amazonses.com
|
|
Remove setlocale() and _configthreadlocal() as fallback strategy on
systems that don't have uselocale(), where ECPG tries to control
LC_NUMERIC formatting on input and output of floating point numbers. It
was probably broken on some systems (NetBSD), and the code was also
quite messy and complicated, with obsolete configure tests (Windows).
It was also arguably broken, or at least had unstated environmental
requirements, if pgtypeslib code was called directly.
Instead, introduce PG_C_LOCALE to refer to the "C" locale as a locale_t
value. It maps to the special constant LC_C_LOCALE when defined by libc
(macOS, NetBSD), or otherwise uses a process-lifetime locale_t that is
allocated on first use, just as ECPG previously did itself. The new
replacement might be more widely useful. Then change the float parsing
and printing code to pass that to _l() functions where appropriate.
Unfortunately the portability of those functions is a bit complicated.
First, many obvious and useful _l() functions are missing from POSIX,
though most standard libraries define some of them anyway. Second,
although the thread-safe save/restore technique can be used to replace
the missing ones, Windows and NetBSD refused to implement standard
uselocale(). They might have a point: "wide scope" uselocale() is hard
to combine with other code and error-prone, especially in library code.
Luckily they have the _l() functions we want so far anyway. So we have
to be prepared for both ways of doing things:
1. In ECPG, use strtod_l() for parsing, and supply a port.h replacement
using uselocale() over a limited scope if missing.
2. Inside our own snprintf.c, use three different approaches to format
floats. For frontend code, call libc's snprintf_l(), or wrap libc's
snprintf() in uselocale() if it's missing. For backend code, snprintf.c
can keep assuming that the global locale's LC_NUMERIC is "C" and call
libc's snprintf() without change, for now.
(It might eventually be possible to call our in-tree Ryū routines to
display floats in snprintf.c, given the C-locale-always remit of our
in-tree snprintf(), but this patch doesn't risk changing anything that
complicated.)
Author: Thomas Munro <thomas.munro@gmail.com>
Reviewed-by: Peter Eisentraut <peter@eisentraut.org>
Reviewed-by: Tristan Partin <tristan@partin.io>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/CWZBBRR6YA8D.8EHMDRGLCKCD%40neon.tech
|
|
The default of 128kB is unchanged, but the upper limit is changed from
32 blocks to 128 blocks, unless the operating system's IOV_MAX is too
low. Some other RDBMSes seem to cap their multi-block buffer pool I/O
around this number, and it seems useful to allow experimentation.
The concrete change is to our definition of PG_IOV_MAX, which provides
the maximum for io_combine_limit and io_max_combine_limit. It also
affects a couple of other places that work with arrays of struct iovec
or smaller objects on the stack, so we still don't want to use the
system IOV_MAX directly without a clamp: it is not under our control and
likely to be 1024. 128 seems acceptable for our current usage.
For Windows, we can't use real scatter/gather yet, so we continue to
define our own IOV_MAX value of 16 and emulate preadv()/pwritev() with
loops. Someone would need to research the trade-offs of raising that
number.
NB if trying to see this working: you might temporarily need to hack
BAS_BULKREAD to be bigger, since otherwise the obvious way of "a very
big SELECT" is limited by that for now.
Suggested-by: Tomas Vondra <tomas@vondra.me>
Discussion: https://postgr.es/m/CA%2BhUKG%2B2T9p-%2BzM6Eeou-RAJjTML6eit1qn26f9twznX59qtCA%40mail.gmail.com
|
|
Since commit f78667bd91, we've used __attribute__((target(...)))
instead of extra compiler flags for AVX-512 support, but this
comment still says that we put the code in a separate file because
it might require extra compiler flags. Let's just remove that part
of the comment.
|
|
It is not clear why these were originally added. One hypothesis is
that an ancient version of MinGW didn't define them. In any case,
they appear to now be superfluous, so let's remove them. If
nothing else, the buildfarm might offer us clues to their origins.
Reviewed-by: Thomas Munro
Discussion: https://postgr.es/m/Z4chOKfnthRH71mw%40nathan
|
|
Add various widely useful "IWYU pragma" annotations, such as
- Common header files such as c.h, postgres.h should be "always_keep".
- System headers included in c.h, postgres.h etc. should be considered
"export".
- Some portability headers such as getopt_long.h should be
"always_keep", so they are not considered superfluous on some
platforms.
- Certain system headers included from portability headers should be
considered "export" because the purpose of the portability header is
to wrap them.
- Superfluous includes marked as "for backward compatibility" get a
formal IWYU annotation.
- Generated header included in utils/syscache.h is marked exported.
This is a very commonly used include and this avoids lots of
complaints.
Discussion: https://www.postgresql.org/message-id/flat/9395d484-eff4-47c2-b276-8e228526c8ae@eisentraut.org
|
|
Change our ftruncate() macro to use the 64-bit variant of chsize(), and
add a new macro to redirect lseek() to _lseeki64().
Back-patch to all supported releases, in preparation for a bug fix.
Tested-by: Davinder Singh <davinder.singh@enterprisedb.com>
Discussion: https://postgr.es/m/CAKZiRmyM4YnokK6Oenw5JKwAQ3rhP0YTz2T-tiw5dAQjGRXE3Q%40mail.gmail.com
|
|
Backpatch-through: 13
|
|
Redefine our exact width types with standard C99 types and macros,
including int64_t, INT64_MAX, INT64_C(), PRId64 etc. We were already
using <stdint.h> types in a few places.
One complication is that Windows' <inttypes.h> uses format strings like
"%I64d", "%I32", "%I" for PRI*64, PRI*32, PTR*PTR, instead of mapping to
other standardized format strings like "%lld" etc as seen on other known
systems. Teach our snprintf.c to understand them.
This removes a lot of configure clutter, and should also allow 64-bit
numbers and other standard types to be used in localized messages
without casting.
Reviewed-by: Peter Eisentraut <peter@eisentraut.org>
Discussion: https://postgr.es/m/ME3P282MB3166F9D1F71F787929C0C7E7B6312%40ME3P282MB3166.AUSP282.PROD.OUTLOOK.COM
|
|
The code in question (pg_preadv() and pg_pwritev()) has been around
for a while, but commit 15c9ac3629 moved it to a header file. If
third-party code that includes this header file is built with
-Wsign-compare on a system without preadv() or pwritev(), warnings
ensue. This commit fixes said warnings by casting the result of
pg_pread()/pg_pwrite() to size_t, which should be safe because we
will have already checked for a negative value.
Author: Wolfgang Walther
Discussion: https://postgr.es/m/16989737-1aa8-48fd-8dfe-b7ada06509ab%40technowledgy.de
Backpatch-through: 17
|
|
macOS doesn't have posix_fadvise(), but fcntl() with the F_RDADVISE
command does the same thing.
Some related documentation has been generalized to not mention
posix_advise() specifically anymore.
Reviewed-by: Thomas Munro <thomas.munro@gmail.com>
Discussion: https://www.postgresql.org/message-id/flat/0827edec-1317-4917-a186-035eb1e3241d%40eisentraut.org
|
|
Use gmtime_r() and localtime_r() instead of gmtime() and localtime(),
for thread-safety.
There are a few affected calls in libpq and ecpg's libpgtypes, which
are probably effectively bugs, because those libraries already claim
to be thread-safe.
There is one affected call in the backend. Most of the backend
otherwise uses the custom functions pg_gmtime() and pg_localtime(),
which are implemented differently.
While we're here, change the call in the backend to gmtime*() instead
of localtime*(), since for that use time zone behavior is irrelevant,
and this side-steps any questions about when time zones are
initialized by localtime_r() vs localtime().
Portability: gmtime_r() and localtime_r() are in POSIX but are not
available on Windows. Windows has functions gmtime_s() and
localtime_s() that can fulfill the same purpose, so we add some small
wrappers around them. (Note that these *_s() functions are also
different from the *_s() functions in the bounds-checking extension of
C11. We are not using those here.)
On MinGW, you can get the POSIX-style *_r() functions by defining
_POSIX_C_SOURCE appropriately before including <time.h>. This leads
to a conflict at least in plpython because apparently _POSIX_C_SOURCE
gets defined in some header there, and then our replacement
definitions conflict with the system definitions. To avoid that sort
of thing, we now always define _POSIX_C_SOURCE on MinGW and use the
POSIX-style functions here.
Reviewed-by: Stepan Neretin <sncfmgg@gmail.com>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Thomas Munro <thomas.munro@gmail.com>
Discussion: https://www.postgresql.org/message-id/flat/eba1dc75-298e-4c46-8869-48ba8aad7d70@eisentraut.org
|
|
Previously we had a fallback implementation that made a harmless system
call, based on the assumption that system calls must contain a memory
barrier. That shouldn't be reached on any current system, and it seems
highly likely that we can easily find out how to request explicit memory
barriers, if we've already had to find out how to do atomics on a
hypothetical new system.
Removed comments and a function name referred to a spinlock used for
fallback memory barriers, but that changed in 1b468a13, which left some
misleading words behind in a few places.
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Suggested-by: Andres Freund <andres@anarazel.de>
Discussion: https://postgr.es/m/721bf39a-ed8a-44b0-8b8e-be3bd81db748%40technowledgy.de
Discussion: https://postgr.es/m/3351991.1697728588%40sss.pgh.pa.us
|
|
Previously we had a fallback implementation of pg_compiler_barrier()
that called an empty function across a translation unit boundary so the
compiler couldn't see what it did. That shouldn't be needed on any
current systems, and might not even work with a link time optimizer.
Since we now require compiler-specific knowledge of how to implement
atomics, we should also know how to implement compiler barriers on a
hypothetical new system.
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Suggested-by: Andres Freund <andres@anarazel.de>
Discussion: https://postgr.es/m/721bf39a-ed8a-44b0-8b8e-be3bd81db748%40technowledgy.de
Discussion: https://postgr.es/m/3351991.1697728588%40sss.pgh.pa.us
|
|
Modern versions of all relevant architectures and tool chains have
atomics support. Since edadeb07, there is no remaining reason to carry
code that simulates atomic flags and uint32 imperfectly with spinlocks.
64 bit atomics are still emulated with spinlocks, if needed, for now.
Any modern compiler capable of implementing C11 <stdatomic.h> must have
the underlying operations we need, though we don't require C11 yet. We
detect certain compilers and architectures, so hypothetical new systems
might need adjustments here.
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Tom Lane <tgl@sss.pgh.pa.us> (concept, not the patch)
Reviewed-by: Andres Freund <andres@anarazel.de> (concept, not the patch)
Discussion: https://postgr.es/m/3351991.1697728588%40sss.pgh.pa.us
|
|
A later change will require atomic support, so it wouldn't make sense
for a hypothetical new system not to be able to implement spinlocks.
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Tom Lane <tgl@sss.pgh.pa.us> (concept, not the patch)
Reviewed-by: Andres Freund <andres@anarazel.de> (concept, not the patch)
Discussion: https://postgr.es/m/3351991.1697728588%40sss.pgh.pa.us
|
|
They spell it "strtok_s" there.
There are currently no uses, but some will be added soon.
Reviewed-by: Kyotaro Horiguchi <horikyota.ntt@gmail.com>
Reviewed-by: David Steele <david@pgmasters.net>
Discussion: https://www.postgresql.org/message-id/flat/79692bf9-17d3-41e6-b9c9-fc8c3944222a@eisentraut.org
|
|
This code wanted to ensure that the 'exchange' variable passed to
pg_atomic_compare_exchange_u64 has correct alignment, but apparently
platforms don't actually require anything that doesn't come naturally.
While messing with pg_atomic_monotonic_advance_u64: instead of using
Max() to determine the value to return, just use
pg_atomic_compare_exchange_u64()'s return value to decide; also, use
pg_atomic_compare_exchange_u64 instead of the _impl version; also remove
the unnecessary underscore at the end of variable name "target".
Backpatch to 17, where this code was introduced by commit bf3ff7bf83bc.
Reported-by: Alexander Lakhin <exclusion@gmail.com>
Discussion: https://postgr.es/m/36796438-a718-cf9b-2071-b2c1b947c1b5@gmail.com
|
|
This old CPU architecture hasn't been produced in decades, and
whatever instances might still survive are surely too underpowered
for anyone to consider running Postgres on in production. We'd
nonetheless continued to carry code support for it (largely at my
insistence), because its unique implementation of spinlocks seemed
like a good edge case for our spinlock infrastructure. However,
our last buildfarm animal of this type was retired last year, and
it seems quite unlikely that another will emerge. Without the ability
to run tests, the argument that this is useful test code fails to
hold water. Furthermore, carrying code support for an untestable
architecture has costs not to be ignored. So, remove HPPA-specific
code, in the same vein as commits 718aa43a4 and 92d70b77e.
Discussion: https://postgr.es/m/3351991.1697728588@sss.pgh.pa.us
|
|
This tracks the position of WAL that's been fully copied into WAL
buffers by all processes emitting WAL. (For some reason we call that
"WAL insertion"). This is updated using atomic monotonic advance during
WaitXLogInsertionsToFinish, which is not when the insertions actually
occur, but it's the only place where we know where have all the
insertions have completed.
This value is useful in WALReadFromBuffers, which can verify that
callers don't try to read past what has been inserted. (However, more
infrastructure is needed in order to actually use WAL after the flush
point, since it could be lost.)
The value is also useful in WaitXLogInsertionsToFinish() itself, since
we can now exit quickly when all WAL has been already inserted, without
even having to take any locks.
|
|
Commit 792752af4e added infrastructure for using AVX-512 intrinsic
functions, and this commit uses that infrastructure to optimize
visibilitymap_count(). Specificially, a new pg_popcount_masked()
function is introduced that applies a bitmask to every byte in the
buffer prior to calculating the population count, which is used to
filter out the all-visible or all-frozen bits as needed. Platforms
without AVX-512 support should also see a nice speedup due to the
reduced number of calls to a function pointer.
Co-authored-by: Ants Aasma
Discussion: https://postgr.es/m/BL1PR11MB5304097DF7EA81D04C33F3D1DCA6A%40BL1PR11MB5304.namprd11.prod.outlook.com
|
|
Presently, pg_popcount() processes data in 32-bit or 64-bit chunks
when possible. Newer hardware that supports AVX-512 instructions
can use 512-bit chunks, which provides a nice speedup, especially
for larger buffers. This commit introduces the infrastructure
required to detect compiler and CPU support for the required
AVX-512 intrinsic functions, and it adds a new pg_popcount()
implementation that uses these functions. If CPU support for this
optimized implementation is detected at runtime, a function pointer
is updated so that it is used by subsequent calls to pg_popcount().
Most of the existing in-tree calls to pg_popcount() should benefit
from these instructions, and calls with smaller buffers should at
least not regress compared to v16. The new infrastructure
introduced by this commit can also be used to optimize
visibilitymap_count(), but that is left for a follow-up commit.
Co-authored-by: Paul Amonson, Ants Aasma
Reviewed-by: Matthias van de Meent, Tom Lane, Noah Misch, Akash Shankaran, Alvaro Herrera, Andres Freund, David Rowley
Discussion: https://postgr.es/m/BL1PR11MB5304097DF7EA81D04C33F3D1DCA6A%40BL1PR11MB5304.namprd11.prod.outlook.com
|
|
If there aren't many bytes to process, the function call overhead
of the optimized implementation isn't worth taking, so instead we
inline a loop that consults pg_number_of_ones in that case. If
there are many bytes to process, we accept the function call
overhead because the optimized versions are likely to be faster.
The threshold at which we use the optimized implementation is set
to the smallest amount of data required to use special popcount
instructions.
Reviewed-by: Alvaro Herrera, Tom Lane
Discussion: https://postgr.es/m/20240402155301.GA2750455%40nathanxps13
|
|
This commit simplifies pg_lfind32() a bit by moving the standard
one-by-one linear search code to an inline helper function.
Reviewed-by: Tom Lane
Discussion: https://postgr.es/m/20240327013616.GA3940109%40nathanxps13
|
|
The newly-introduced "one_by_one" label produces -Wunused-label
warnings when building without SIMD support. To fix, move the
label into the SIMD section of this function.
Oversight in commit 7644a7340c.
Reported-by: Tom Lane
Discussion: https://postgr.es/m/3189995.1711495704%40sss.pgh.pa.us
|
|
This commit improves the performance of pg_lfind32() in many cases
by modifying it to process the remaining "tail" of elements with
SIMD instructions instead of processing them one-by-one. Since the
SIMD code processes a large block of elements, this means that we
will process a subset of elements more than once, but that won't
affect the correctness of the result, and testing has shown that
this helps more cases than it regresses. With this change, the
standard one-by-one linear search code is only used for small
arrays and for platforms without SIMD support.
Suggested-by: John Naylor
Reviewed-by: John Naylor
Discussion: https://postgr.es/m/20231129171526.GA857928%40nathanxps13
|
|
On some systems, calls to pg_popcount{32,64} are indirected through
a function pointer. This commit converts pg_popcount() to a
function pointer on those systems so that we can inline the
appropriate pg_popcount{32,64} implementations into each of the
pg_popcount() implementations. Since pg_popcount() may call
pg_popcount{32,64} several times, this can significantly improve
its performance.
Suggested-by: David Rowley
Reviewed-by: David Rowley
Discussion: https://postgr.es/m/CAApHDvrb7MJRB6JuKLDEY2x_LKdFHwVbogpjZBCX547i5%2BrXOQ%40mail.gmail.com
|
|
Add PGDLLIMPORT to pg_popcount32/64. In passing, fix a typo.
Diagnosis by Masahiko Sawada, patch by David Rowley
Per buildfarm members drongo and fairywren
Discussion: https://postgr.es/m/CAD21AoAMm1mQd%3Dw4PrfrKK%3DOMP8j8%3D7ntJRPF8%2B%3D10iUuvwiCA%40mail.gmail.com
Discussion: https://postgr.es/m/CAApHDvov7724UrD1Ug0D1eV%2B9Pd_x5VEQmw-6HVG9w1WdCxXPA%40mail.gmail.com
|
|
The first argument of vshrq_n_s8 needs to be a signed vector type,
but it was passed unsigned. Clang is more lax with conversion, but
gcc needs a cast.
Fix by me, tested by Masahiko Sawada
Per buildfarm members splitfin, batta, widowbird, snakefly, parula,
massasauga
Discussion: https://postgr.es/m/20240306074106.mg6w4koohdlworbs%40alap3.anarazel.de
|
|
vector8_min - helper for emulating ">=" semantics
vector8_highbit_mask - used to turn the result of a vector
comparison into a bitmask
Masahiko Sawada
Reviewed by Nathan Bossart, with additional adjustments by me
Discussion: https://postgr.es/m/CAFBsxsHbBm_M22gLBO%2BAZT4mfMq3L_oX3wdKZxjeNnT7fHsYMQ%40mail.gmail.com
|
