Horton

RE: Popcount optimization using AVX512

Amonson, Paul D <paul.d.amonson@intel.com>

View thread
From: "Amonson, Paul D" <paul.d.amonson@intel.com>
To: Nathan Bossart <nathandbossart@gmail.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: 2024-03-28T22:03:04Z
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.

> -----Original Message-----
> From: Nathan Bossart <nathandbossart@gmail.com>
> Sent: Thursday, March 28, 2024 2:39 PM
> To: Amonson, Paul D <paul.d.amonson@intel.com>
> 
> * The latest patch set from Paul Amonson appeared to support MSVC in the
>   meson build, but not the autoconf one.  I don't have much expertise here,
>   so the v14 patch doesn't have any autoconf/meson support for MSVC, which
>   I thought might be okay for now.  IIUC we assume that 64-bit/MSVC builds
>   can always compile the x86_64 popcount code, but I don't know whether
>   that's safe for AVX512.

I also do not know how to integrate MSVC+Autoconf, the CI uses MSVC+Meson+Ninja so I stuck with that.
 
> * I think we need to verify there isn't a huge performance regression for
>   smaller arrays.  IIUC those will still require an AVX512 instruction or
>   two as well as a function call, which might add some noticeable overhead.

Not considering your changes, I had already tested small buffers. At less than 512 bytes there was no measurable regression (there was one extra condition check) and for 512+ bytes it moved from no regression to some gains between 512 and 4096 bytes. Assuming you introduced no extra function calls, it should be the same.

> I forgot to mention that I also want to understand whether we can actually assume availability of XGETBV when CPUID says we support AVX512:

You cannot assume as there are edge cases where AVX-512 was found on system one during compile but it's not actually available in a kernel on a second system at runtime despite the CPU actually having the hardware feature.

I will review the new patch to see if there are anything that jumps out at me.

Thanks,
Paul