|
From: | Alois Schloegl |
Subject: | Re: Compiling octave with MSVC |
Date: | Mon, 23 Oct 2006 11:40:45 +0200 |
User-agent: | Mozilla Thunderbird 0.9 (Windows/20041103) |
Hi Micheal,your result is about speed. Could you also check the accuracy of the results?
I suggest this test: http://www.dpmi.tu-graz.ac.at/~schloegl/matlab/fermat/ Download these files into a local directory http://www.dpmi.tu-graz.ac.at/~schloegl/matlab/fermat/fermat.m http://www.dpmi.tu-graz.ac.at/~schloegl/matlab/fermat/flttest.m http://www.dpmi.tu-graz.ac.at/~schloegl/matlab/fermat/flttest1000.m and run flttest and flttest1000. Please send the resulting *.txt files. Alois address@hidden wrote:
Here are some numbers. I ran the benchmark script in regular cygwin package (current version is 2.1.73) and in 2 MSVC-compiled 2.9.9+ (actually, CVS versions) packages. Results are attached.Both MSVC versions do not rely on external libraries, like an optimizedATLAS or so. One version is compiled with full debug, no optimization. The other with no debug, full speed optimization (/O2 flag).Michael.------------------------------------------------------------------------ > Have you tried the Octave2.m benchmark from sciviews.org to > see how an MSVC compiled Octave compares to a cygwin/gcc > compiled one? Does this link against ATLAS, FFTW, etc?
[Prev in Thread] | Current Thread | [Next in Thread] |