This is the mail archive of the
libc-alpha@sourceware.org
mailing list for the glibc project.
Re: [PATCH RFC] Imporve 64bit memset performance for Haswell CPU with AVX2 instruction
- From: OndÅej BÃlka <neleai at seznam dot cz>
- To: "H.J. Lu" <hjl dot tools at gmail dot com>
- Cc: Ling Ma <ling dot ma dot program at gmail dot com>, GNU C Library <libc-alpha at sourceware dot org>, Richard Henderson <rth at twiddle dot net>, Andreas Jaeger <aj at suse dot com>, Liubov Dmitrieva <liubov dot dmitrieva at gmail dot com>, Ling Ma <ling dot ml at alibaba-inc dot com>
- Date: Thu, 19 Jun 2014 21:12:38 +0200
- Subject: Re: [PATCH RFC] Imporve 64bit memset performance for Haswell CPU with AVX2 instruction
- Authentication-results: sourceware.org; auth=none
- References: <20140515201458 dot GA24885 at domone dot podge> <CAOGi=dNmn2bPfB65VoXUGjQ7t6RLVJ2hj2QDarrUjZV75kTbDA at mail dot gmail dot com> <20140530113041 dot GB26528 at domone dot podge> <CAOGi=dPdWegEo1s8=wG4WzOANaQ3x=boLFitQ_wBp+Xf+hxexQ at mail dot gmail dot com> <CAMe9rOqv5RYK1MO2M098n3o50-KmmZJuvsvMmXqkBBt0g3OY_g at mail dot gmail dot com> <CAOGi=dMNyzckY8s3uF0qRpKuqUwHHhzQeyy-j29ydLNn_s9Bog at mail dot gmail dot com> <20140605163224 dot GA8041 at domone dot podge> <CAOGi=dN-kC5tZ3ZMhjijGqK+3ePVMrOsT1M2EOhOnmhWMW7kpg at mail dot gmail dot com> <CAOGi=dNw3AcPJPRj60-VFNdAMCtNgv71OWvguXt-XRHYra0u5Q at mail dot gmail dot com> <CAMe9rOoYMFKp-s2P+nP7gyP5mkM5yUjy+Vq3Hsc+0SuD=7w_3A at mail dot gmail dot com>
On Wed, Jun 18, 2014 at 09:47:11AM -0700, H.J. Lu wrote:
> On Tue, Jun 10, 2014 at 6:52 AM, Ling Ma <ling.ma.program@gmail.com> wrote:
> > In this patch as gziped attachment, we take advantage of HSW memory
> > bandwidth, manage to reduce miss branch prediction by avoiding using
> > branch instructions and
> > force destination to be aligned with avx & avx2 instruction.
> >
> > The CPU2006 403.gcc benchmark indicates this patch improves performance
> > from 26% to 59%.
> >
> > This version accept Ondra's comments and avoid branch instruction to
> > cross 16byte-aligned code.
>
> Any feedback? I'd like to check it in before 2.20 code freeze.
>
As I said before its ok with fixed formatting, you could commit it if
you wish.