This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
Re: [DOC PATCH]: PR target/34702: Document limitations of recip sequences for x86
- From: Gerald Pfeifer <gerald at pfeifer dot com>
- To: Uros Bizjak <ubizjak at gmail dot com>
- Cc: GCC Patches <gcc-patches at gcc dot gnu dot org>, Dominique Dhumieres <dominiq at lps dot ens dot fr>
- Date: Tue, 8 Jan 2008 22:44:29 +0100 (CET)
- Subject: Re: [DOC PATCH]: PR target/34702: Document limitations of recip sequences for x86
- References: <5787cf470801080200gc2a307aqd1b7a920e35b437e@mail.gmail.com>
On Tue, 8 Jan 2008, Uros Bizjak wrote:
> 2008-01-08 Uros Bizjak <ubizjak@gmail.com>
>
> PR target/34702
> * doc/invoke.texi (i386 and x86-64 Options) [mrecip]: Document
> limitations of reciprocal sequences on x86 targets.
This looks good from my doc maintainer perspective (with two minor
comments below), but it would be good to get this approved by an x86
maintainer as well. :-)
> @opindex mrecip
> This option will enable GCC to use RCPSS and RSQRTSS instructions (and their
> -vectorized variants RCPPS and RSQRTPS) instead of DIVSS and SQRTSS (and their
> -vectorized variants). These instructions will be generated only when
> -@option{-funsafe-math-optimizations} is enabled.
> +vectorized variants RCPPS and RSQRTPS) with additional Newton-Rhapson step
Should this be "with an additional...step" or "with additional...steps"?
^^^^ ^^^
> +Note that while the throughput of the sequence is higher than the throughput
> +of the non-reciprocal instruction, the precision of the sequence can be
> +decreased up to 2 ulp.
I believe this should read "decreased by up to 2 ulp"?
Gerald