Gentoo Archives: gentoo-science

From: Andrea Arteaga <andyspiros@×××××.com>
To: gentoo-science@l.g.o
Subject: Re: [gentoo-science] Atlas giving wrong results on matrix multiplication
Date: Mon, 19 Sep 2011 15:17:38
Message-Id: CAPtnn5F=ue3Ctn_D_P4Eb6dZd8cHvRCdH1HcjkF8KWPh+yQWug@mail.gmail.com
In Reply to: [gentoo-science] Atlas giving wrong results on matrix multiplication by "Xavier Fernández i Marín"
1 Dear Xavier,
2
3 thank you for your email and sorry for the late reply.
4 In theory it should be useless to provide optimization flags when
5 compiling ATLAS, because it automatically uses and tests many flags,
6 but I could be wrong.
7
8 How did you install ATLAS? From the science overlay? I suggest that
9 you update it. Now in the overlay we have sci-libs/atlas-3.9.49
10 (~keyworded) [BTW, I will probably bump the version to 3.9.51 today].
11 Could you install it and redo the test? Otherwise, could you shortly
12 explain how to reproduce the problem, so that I can try?
13
14 Cheers
15 Andrea Arteaga
16
17
18 2011/9/14 Xavier Fernández i Marín <xfim.ll@×××××.com>:
19 > Dear all,
20 >
21 > I recently submitted a question to the R-help mailing list, because I
22 > noticed that matrix multiplication was giving me trouble (basically, that
23 > there were some times when a result was NaN):
24 > http://tolstoy.newcastle.edu.au/R/e15/help/11/09/7165.html
25 >
26 >
27 > As it is mentioned in the end of the thread, the solution was to recompile
28 > R against the reference blas and lapack libraries, and avoid the atlas
29 > that I was using.
30 >
31 > So it seems that the problem is not R-related, but atlas-related, and
32 > linked to gentoo, I would like to share the problem with you, just in case
33 > somebody has any clue on what is wrong with atlas.
34 >
35 > Although I have firstly used a very aggressive cflags to compile atlas
36 > (-Ofast), I suspected that the problem was that, but I recompiled atlas
37 > using -O2 and the problem still presists with atlas, but is solved with
38 > reference.
39 >
40 > I am using atlas-3.9.23-r4, R-2.13.1 and gcc-4.6.1.

Replies

Subject Author
Re: [gentoo-science] Atlas giving wrong results on matrix multiplication "Xavier Fernández i Marín" <xfim.ll@×××××.com>