Horton

Re: Popcount optimization using AVX512

Nathan Bossart <nathandbossart@gmail.com>

View thread
From: Nathan Bossart <nathandbossart@gmail.com>
To: David Rowley <dgrowleyml@gmail.com>
Cc: "Amonson, Paul D" <paul.d.amonson@intel.com>, Andres Freund <andres@anarazel.de>, Alvaro Herrera <alvherre@alvh.no-ip.org>, "Shankaran, Akash" <akash.shankaran@intel.com>, Noah Misch <noah@leadboat.com>, Tom Lane <tgl@sss.pgh.pa.us>, Matthias van de Meent <boekewurm+postgres@gmail.com>, "pgsql-hackers@lists.postgresql.org" <pgsql-hackers@lists.postgresql.org>
Date: 2024-03-18T21:29:19Z
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 Tue, Mar 19, 2024 at 10:27:58AM +1300, David Rowley wrote:
> On Tue, 19 Mar 2024 at 10:08, Nathan Bossart <nathandbossart@gmail.com> wrote:
>> On Tue, Mar 19, 2024 at 10:02:18AM +1300, David Rowley wrote:
>> > The only thing I'd question in the patch is in pg_popcount_fast(). It
>> > looks like you've opted to not do the 32-bit processing on 32-bit
>> > machines. I think that's likely still worth coding in a similar way to
>> > how pg_popcount_slow() works. i.e. use "#if SIZEOF_VOID_P >= 8".
>> > Probably one day we'll remove that code, but it seems strange to have
>> > pg_popcount_slow() do it and not pg_popcount_fast().
>>
>> The only reason I left it out was because I couldn't convince myself that
>> it wasn't dead code, given we assume that popcntq is available in
>> pg_popcount64_fast() today.  But I don't see any harm in adding that just
>> in case.
> 
> It's probably more of a case of using native instructions rather than
> ones that might be implemented only via microcode.  For the record, I
> don't know if that would be the case for popcntq on x86 32-bit and I
> don't have the hardware to test it. It just seems less risky just to
> do it.

Agreed.  Will send an updated patch shortly.

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