[CPU blog][FUT] Still no library that is optimal on all processors

eGospodarka.pl › Grupy › pl.comp.programming › [CPU blog][FUT] Still no library that is optimal on all processors › [CPU blog][FUT] Still no library that is optimal on all processors

Path: news-archive.icm.edu.pl!agh.edu.pl!news.agh.edu.pl!newsfeed2.atman.pl!newsfeed.
atman.pl!news.task.gda.pl!newsfeed.icp.pl!news.icpnet.pl!.POSTED!not-for-mail
From: n...@m...invalid
Newsgroups: pl.comp.lang.c,pl.comp.programming
Subject: [CPU blog][FUT] Still no library that is optimal on all processors
Followup-To: pl.comp.programming
Date: Sat, 21 Apr 2012 20:23:01 +0200
Organization: ICP News Server
Lines: 21
Message-ID: <jmutu1$2he$2@news.icpnet.pl>
NNTP-Posting-Host: 95.108.122.248
Mime-Version: 1.0
Content-Type: text/plain; charset=ISO-8859-2
Content-Transfer-Encoding: 7bit
X-Trace: news.icpnet.pl 1335032577 2606 95.108.122.248 (21 Apr 2012 18:22:57 GMT)
X-Complaints-To: a...@i...pl
NNTP-Posting-Date: Sat, 21 Apr 2012 18:22:57 +0000 (UTC)
User-Agent: Mozilla/5.0 (Windows NT 5.2; rv:7.0.1) Gecko/20110929 Thunderbird/7.0.1
Xref: news-archive.icm.edu.pl pl.comp.lang.c:298863 pl.comp.programming:196908
[ ukryj nagłówki ]
Agner Fog:
> Choosing the most efficient function library can be a nightmare to a
> programmer. I have tried to calculate the cosine function with
> different libraries and compare the calculation time. The best
> version is 19 times faster than the worst!
>
> AMD have now updated their math libraries and added CPU dispatching.
> There are two versions of code in AMD's LIBM library: One for the
> SSE2 instruction set and one for AVX and FMA4. Intel processors will
> run the inferior SSE2 branch because they don't have the FMA4
> instruction set. The incompatibility between Intel's and AMD's FMA
> instructions is another scandal, which I have discussed in this blog
> post. The AMD library does not check the CPU brand name as Intel
> libraries do. It only checks for the FMA4 instructions which are not
> supported by Intel processors, although - quite ironically - they
> were designed by Intel. It will be possible to run the better branch
> on Intel processors if Intel decides to support the FMA4 instruction
> set in the future.
-- <http://agner.org/optimize/blog/read.php?i=209>
--
FUT: p.c.p