Path: news.mathworks.com!not-for-mail
From: "Duncan Po" <Duncan.Po@mathworks.com>
Newsgroups: comp.soft-sys.matlab
Subject: Re: GOTO BLAS and multiple threads
Date: Wed, 25 Feb 2009 00:39:02 +0000 (UTC)
Organization: The MathWorks Inc
Lines: 6
Message-ID: <go23v6$qqt$1@fred.mathworks.com>
References: <g7seu6$4du$1@news2.rz.uni-karlsruhe.de> <g7udqa$59s$1@news2.rz.uni-karlsruhe.de> <gmi1sd$df3$1@fred.mathworks.com> <gnvcfq$l28$1@fred.mathworks.com> <go13eh$d93$1@fred.mathworks.com> <go1mia$m73$1@fred.mathworks.com> <go1ok0$djk$1@fred.mathworks.com>
Reply-To: "Duncan Po" <Duncan.Po@mathworks.com>
NNTP-Posting-Host: webapp-05-blr.mathworks.com
Content-Type: text/plain; charset="ISO-8859-1"
Content-Transfer-Encoding: 8bit
X-Trace: fred.mathworks.com 1235522342 27485 172.30.248.35 (25 Feb 2009 00:39:02 GMT)
X-Complaints-To: news@mathworks.com
NNTP-Posting-Date: Wed, 25 Feb 2009 00:39:02 +0000 (UTC)
X-Newsreader: MATLAB Central Newsreader 379004
Xref: news.mathworks.com comp.soft-sys.matlab:520583


Most references point to Goto BLAS may be slightly faster than Intel MKL, which is the default choice of MATLAB. You should expect you may be able to get some speedup, but not that much.

For instance, here is a comparison:
http://www.google.com/url?sa=U&start=1&q=http://www.lr.tudelft.nl/live/pagina.jsp%3Fid%3D608b3015-5195-438a-90fc-c30c2252a066%26lang%3Den%26binary%3D/doc/blas_lapack.pdf&ei=mJKkSYK6J5W6twff_7HMBA&usg=AFQjCNFyz_skF2zrp8kZrUXJVN4NnHOo5A

Looking forward to see your results.