Horton

Re: Popcount optimization using AVX512

Nathan Bossart <nathandbossart@gmail.com>

View thread
From: Nathan Bossart <nathandbossart@gmail.com>
To: "Shankaran, Akash" <akash.shankaran@intel.com>
Cc: Tom Lane <tgl@sss.pgh.pa.us>, David Rowley <dgrowleyml@gmail.com>, Ants Aasma <ants.aasma@cybertec.at>, Alvaro Herrera <alvherre@alvh.no-ip.org>, "Amonson, Paul D" <paul.d.amonson@intel.com>, Andres Freund <andres@anarazel.de>, Noah Misch <noah@leadboat.com>, Matthias van de Meent <boekewurm+postgres@gmail.com>, "pgsql-hackers@lists.postgresql.org" <pgsql-hackers@lists.postgresql.org>, "Devulapalli, Raghuveer" <raghuveer.devulapalli@intel.com>
Date: 2024-04-18T19:53:46Z
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.

On Thu, Apr 18, 2024 at 06:12:22PM +0000, Shankaran, Akash wrote:
> Good find. I confirmed after speaking with an intel expert, and from the intel AVX-512 manual [0] section 14.3, which recommends to check bit27. From the manual:
> 
> "Prior to using Intel AVX, the application must identify that the operating system supports the XGETBV instruction,
> the YMM register state, in addition to processor's support for YMM state management using XSAVE/XRSTOR and
> AVX instructions. The following simplified sequence accomplishes both and is strongly recommended.
> 1) Detect CPUID.1:ECX.OSXSAVE[bit 27] = 1 (XGETBV enabled for application use1).
> 2) Issue XGETBV and verify that XCR0[2:1] = '11b' (XMM state and YMM state are enabled by OS).
> 3) detect CPUID.1:ECX.AVX[bit 28] = 1 (AVX instructions supported).
> (Step 3 can be done in any order relative to 1 and 2.)"

Thanks for confirming.  IIUC my patch should be sufficient, then.

> It also seems that step 1 and step 2 need to be done prior to the CPUID OSXSAVE check in the popcount code.

This seems to contradict the note about doing step 3 at any point, and
given step 1 is the OSXSAVE check, I'm not following what this means,
anyway.

I'm also wondering if we need to check that (_xgetbv(0) & 0xe6) == 0xe6
instead of just (_xgetbv(0) & 0xe0) != 0, as the status of the lower half
of some of the ZMM registers is stored in the SSE and AVX state [0].  I
don't know how likely it is that 0xe0 would succeed but 0xe6 wouldn't, but
we might as well make it correct.

[0] https://en.wikipedia.org/wiki/Control_register#cite_ref-23

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