Horton

Re: Improve CRC32C performance on SSE4.2

Tomas Vondra <tomas@vondra.me>

View thread
From: Tomas Vondra <tomas@vondra.me>
To: Nathan Bossart <nathandbossart@gmail.com>
Cc: Andy Fan <zhihuifan1213@163.com>, John Naylor <johncnaylorls@gmail.com>, "Devulapalli, Raghuveer" <raghuveer.devulapalli@intel.com>, "pgsql-hackers@lists.postgresql.org" <pgsql-hackers@lists.postgresql.org>, "Shankaran, Akash" <akash.shankaran@intel.com>
Date: 2025-06-14T14:01:38Z
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 6/14/25 15:56, Nathan Bossart wrote:
> On Sat, Jun 14, 2025 at 03:47:33PM +0200, Tomas Vondra wrote:
>> I suggest you try with a newer gcc, perhaps 13.4. There's been a bunch
>> of fixes related to AVX512 since 13.0, chances are this was already
>> fixed. I don't see this failure on 14.3.1.
> 
> From the config.log, it looks like Andy is using clang:
> 
> 	configure:3998: clang --version >&5
> 	clang version 18.1.6 (https://gitee.com/mirrors/llvm-project.git 1118c2e05e67a36ed8ca250524525cdb66a55256)
> 
> And I see -O0 used, too, which would match the existing report [0],
> although that report is for clang 19.1.7.
> 

Ah, I got confused by this:

-----------------
Found candidate GCC installation:
/usr/local/bin/../lib/gcc/x86_64-pc-linux-gnu/13.3.0
Selected GCC installation:
/usr/local/bin/../lib/gcc/x86_64-pc-linux-gnu/13.3.0
-----------------

> I'm also genuinely curious why folks are using -O0...
> 
> [0] https://postgr.es/m/CAE-ML%2B-OV6p9uvCFBcSQjZUEh__y0h-KjN%2BBseyGJHt7u8EP%2Bw%40mail.gmail.com
> 

I personally use -O0 to get better backtraces (without values optimized
out), better valgrind reports, etc.


-- 
Tomas Vondra