This is the mail archive of the
libc-alpha@sourceware.org
mailing list for the glibc project.
Re: [PATCH 1/N] x86_64 vectorization support: vectorized math functions addition to Glibc
- From: Jakub Jelinek <jakub at redhat dot com>
- To: Andi Kleen <andi at firstfloor dot org>
- Cc: Rich Felker <dalias at libc dot org>, Matthew Fortune <Matthew dot Fortune at imgtec dot com>, "Joseph S. Myers" <joseph at codesourcery dot com>, Andrew Senkevich <andrew dot n dot senkevich at gmail dot com>, libc-alpha <libc-alpha at sourceware dot org>, "igor dot zamyatin at intel dot com" <igor dot zamyatin at intel dot com>, "Melik-Adamyan, Areg" <areg dot melik-adamyan at intel dot com>
- Date: Fri, 12 Sep 2014 09:42:52 +0200
- Subject: Re: [PATCH 1/N] x86_64 vectorization support: vectorized math functions addition to Glibc
- Authentication-results: sourceware.org; auth=none
- References: <CAMXFM3t=ppndDUBzHzSus7xyuF5hTaLFZ5b273jD39NtddSvsw at mail dot gmail dot com> <Pine dot LNX dot 4 dot 64 dot 1409101549490 dot 12853 at digraph dot polyomino dot org dot uk> <6D39441BF12EF246A7ABCE6654B0235320F09D65 at LEMAIL01 dot le dot imgtec dot org> <20140911210246 dot GN23797 at brightrain dot aerifal dot cx> <87a9655rnu dot fsf at tassilo dot jf dot intel dot com>
- Reply-to: Jakub Jelinek <jakub at redhat dot com>
On Thu, Sep 11, 2014 at 10:33:41PM -0700, Andi Kleen wrote:
> Rich Felker <dalias@libc.org> writes:
> >
> > This really seems like something the compiler should be doing --
> > translating parallelizable calls to the standard math functions into
> > calls to special simd versions (
>
> Of course gcc already supports that. Even in two different flavours.
>
> Not sure why the patch doesn't implement one of those ABIs though.
Because GCC supports even another flavor, which presumably the patches
implement. The two you are mentioning are for compatibility with existing
math libraries. The third one is used by #pragma omp declare simd
functions and Cilk+ elemental functions. So, to use those you don't
need any extra gcc support, glibc headers could just add
#if defined(__OPENMP) && __OPENMP >= 201307
#pragma omp declare simd
#endif
on the prototypes (of course maybe with some clauses if needed).
IMHO entry points which take arbitrary length vectors in memory plus counts
are generally useful, you can do whatever you want under the hood, but
entrypoints with arguments directly in vector registers are useful too, you
don't have to use IFUNC for that, the compiler knows what version it
prefers, and if the body of the vectorized loop does more than just
compute a single transcendental function in a tight loop often will be more
handy than having to spill all the vectors to memory and pass it that way.
Jakub