Horton

Re: Popcount optimization using AVX512

Nathan Bossart <nathandbossart@gmail.com>

View thread
From: Nathan Bossart <nathandbossart@gmail.com>
To: "Amonson, Paul D" <paul.d.amonson@intel.com>
Cc: Tom Lane <tgl@sss.pgh.pa.us>, David Rowley <dgrowleyml@gmail.com>, Andres Freund <andres@anarazel.de>, Alvaro Herrera <alvherre@alvh.no-ip.org>, "Shankaran, Akash" <akash.shankaran@intel.com>, Noah Misch <noah@leadboat.com>, Matthias van de Meent <boekewurm+postgres@gmail.com>, "pgsql-hackers@lists.postgresql.org" <pgsql-hackers@lists.postgresql.org>
Date: 2024-03-30T20:03:29Z
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. Fix __attribute__((target(...))) usage.

  2. Use __attribute__((target(...))) for AVX-512 support.

  3. Fix code for probing availability of AVX-512.

  4. Optimize visibilitymap_count() with AVX-512 instructions.

  5. Optimize pg_popcount() with AVX-512 instructions.

  6. Inline pg_popcount() for small buffers.

  7. Avoid function call overhead of pg_popcount() in syslogger.c.

  8. Refactor code for setting pg_popcount* function pointers.

  9. Inline pg_popcount{32,64} into pg_popcount().

  10. Remove MSVC scripts

  11. Use ARMv8 CRC instructions where available.

  12. Use Intel SSE 4.2 CRC instructions where available.

Attachments

I used John Naylor's test_popcount module [0] to put together the attached
graphs (note that the "small arrays" one is semi-logarithmic).  For both
graphs, the X-axis is the number of 64-bit words in the array, and Y-axis
is the amount of time in milliseconds to run pg_popcount() on it 100,000
times (along with a bit of overhead).  This test didn't show any
regressions with a relatively small number of bytes, and it showed the
expected improvements with many bytes.

There isn't a ton of use of pg_popcount() in Postgres, but I do see a few
places that call it with enough bytes for the AVX512 optimization to take
effect.  There may be more callers in the future, though, and it seems
generally useful to have some of the foundational work for using AVX512
instructions in place.  My current plan is to add some new tests for
pg_popcount() with many bytes, and then I'll give it a few more days for
any additional feedback before committing.

[0] https://postgr.es/m/CAFBsxsE7otwnfA36Ly44zZO+b7AEWHRFANxR1h1kxveEV=ghLQ@mail.gmail.com

-- 
Nathan Bossart
Amazon Web Services: https://aws.amazon.com