On Wed, Jul 31, 2024 at 01:52:54PM -0700, Andres Freund wrote:
> On 2024-07-30 22:12:18 -0500, Nathan Bossart wrote:
>> As I started on this, I remembered why I needed it. The file
>> pg_popcount_avx512_choose.c is compiled without the AVX-512 flags in order
>> to avoid inadvertently issuing any AVX-512 instructions before determining
>> we have support. If that's not a concern, we could still probably remove
>> the XSAVE check.
>
> I think it's a valid concern - but isn't that theoretically also an issue with
> xsave itself? I guess practically the compiler won't do that, because there's
> no practical reason to emit any instructions enabled by -mxsave (in contrast
> to e.g. -mavx, which does trigger gcc to emit different instructions even for
> basic math).
Yeah, this crossed my mind. It's certainly not the sturdiest of
assumptions...
> I think enabling options like these on a per-translation-unit basis isn't
> really a scalable approach. To actually be safe there could only be a single
> function in each TU and that function could only be called after a cpuid check
> performed in a separate TU. That a) ends up pretty unreadable b) requires
> functions to be implemented in .c files, which we really don't want for some
> of this.
Agreed.
> I think we'd be better off enabling architectural features on a per-function
> basis, roughly like this:
>
> [...]
>
> /* FIXME: Should be gated by configure check of -mavx512vpopcntdq -mavx512bw
> support */
> pg_enable_target("avx512vpopcntdq,avx512bw")
> uint64_t
> pg_popcount_avx512(const char *buf, int bytes)
> ...
I remember wondering why the CRC-32C code wasn't already doing something
like this (old compiler versions? non-gcc-like compilers?), and I'm not
sure I ever discovered the reason, so out of an abundance of caution I used
the same approach for AVX-512. If we can convince ourselves that
__attribute__((target("..."))) is standard enough at this point, +1 for
moving to that.
--
nathan
