This is the mail archive of the libc-alpha@sourceware.org mailing list for the glibc project.

Index Nav:	[Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav:	[Date Prev] [Date Next]	[Thread Prev] [Thread Next]
Other format:	[Raw text]

Re: [PATCH] x86-64: Add sinf with FMA

From: Arjan van de Ven <arjan at linux dot intel dot com>
To: Nix <nix at esperi dot org dot uk>, "H.J. Lu" <hjl dot tools at gmail dot com>
Cc: GNU C Library <libc-alpha at sourceware dot org>
Date: Fri, 8 Dec 2017 08:32:20 -0800
Subject: Re: [PATCH] x86-64: Add sinf with FMA
Authentication-results: sourceware.org; auth=none
References: <20171204180905.GA31592@gmail.com> <87a7yt5ht6.fsf@esperi.org.uk> <cecb8f10-170f-f616-def4-e11f167a52fb@linux.intel.com> <CAMe9rOrgOXYh9pAkW3LYsU+=TAhsNhoTBKN7_cbObH4pE8UNFA@mail.gmail.com> <87609h5h6j.fsf@esperi.org.uk>

On 12/8/2017 8:16 AM, Nix wrote:

Yeah, I guess I was wondering if the speedup was actually due to FMA or
some other microarchitectural variation which might not be present on
older FMA-capable processors.


I'm not aware of any system where a FMA would be slower than discrete mul + add...
Even for HSW/BDW they're faster. Now on SKL the performance increase of FMA is *more*
(FMA is 4 cycles, but so are individual add and mul .. so it's really 2x gain)
but that's more upside, not a downside for others.

References:
- [PATCH] x86-64: Add sinf with FMA
  - From: H.J. Lu
- Re: [PATCH] x86-64: Add sinf with FMA
  - From: Nick Alcock
- Re: [PATCH] x86-64: Add sinf with FMA
  - From: Arjan van de Ven
- Re: [PATCH] x86-64: Add sinf with FMA
  - From: H.J. Lu
- Re: [PATCH] x86-64: Add sinf with FMA
  - From: Nix

Index Nav:	[Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav:	[Date Prev] [Date Next]	[Thread Prev] [Thread Next]