Horton

Re: Improve CRC32C performance on SSE4.2

Nathan Bossart <nathandbossart@gmail.com>

View thread
From: Nathan Bossart <nathandbossart@gmail.com>
To: Tom Lane <tgl@sss.pgh.pa.us>
Cc: John Naylor <johncnaylorls@gmail.com>, Andy Fan <zhihuifan1213@163.com>, "Devulapalli, Raghuveer" <raghuveer.devulapalli@intel.com>, Jesper Pedersen <jesperpedersen.db@gmail.com>, Tomas Vondra <tomas@vondra.me>, "pgsql-hackers@lists.postgresql.org" <pgsql-hackers@lists.postgresql.org>, "Shankaran, Akash" <akash.shankaran@intel.com>
Date: 2025-06-23T15:15:26Z
Lists: pgsql-hackers

Commits

Same data as JSON: GET /api/v1/messages/:b64id/commits the thread's linked commits as JSON, with link sources. API reference →

  1. Include _mm512_zextsi128_si512() in AVX-512 configure probes.

  2. Properly fix AVX-512 CRC calculation bug

  3. Workaround code generation bug in clang

  4. Compute CRC32C using AVX-512 instructions where available

  5. Inline CRC computation for small fixed-length input on x86

  6. Be more paranoid in configure's checks for CRC and POPCNT intrinsics.

On Mon, Jun 23, 2025 at 11:10:45AM -0400, Tom Lane wrote:
> Nathan Bossart <nathandbossart@gmail.com> writes:
>> On Mon, Jun 23, 2025 at 10:51:21AM -0400, Tom Lane wrote:
>>> This has broken the build completely on my RHEL8 x86_64 box,
>>> with gcc 8.5.0:
> 
>> Unfortunately, this will probably require more than replacing
>> _mm512_castsi512_si128 with _mm512_zextsi512_si128 because the latter
>> doesn't exist.
> 
> I was imagining just including _mm512_zextsi128_si512() in the
> code being tested during configure, so that we fall back to
> the non-AVX-512 code if the compiler is too old to have it.
> I don't really feel a need to work harder than that.

Sorry, my note wasn't clear.  Right now, the configure test uses
_mm512_castsi512_si128(), so we can't just do a simple s/cast/zext.  We'll
need to make a slightly bigger modification to the test to make sure the
zext intrinsics are understood.  I agree that we needn't work any harder
than that.

-- 
nathan