Hello Benjamin In my home, the cpu of computer is celeron-M so that SSE2 version of ATLAS is selected. In that case,
a=[1 2 3 4 5]+j*[5 4 3 2 1]; c=a'*a; does not cause crash on octave 3.2.3 / mingw32. Thus the problem seems to rely on the ATLAS. Regards Tatsuro --- Benjamin Lindner wrote: > > > >> Jaroslav Hajek wrote: > >>> On Mon, Nov 2, 2009 at 1:30 PM, Benjamin Lindner <[email protected]> > >>> wrote: > >>>>>> On Mon, Nov 2, 2009 at 9:10 AM, Nit > >>>>>> Nit > >>>>>> wrote: > >>>>>>> Hello Benjamin, Tatsuro and other octave developers. > >>>>>>> > >>>>>>> Thanks for your work on Octave/mingw32. > >>>>>>> > >>>>>>> I have installed the binary with all octaveforge > >>>>>> packages (problematic > >>>>>>> packages do not auto-loaded) on my XP SP3 machine and > >>>>>> encountered a crash > >>>>>>> when performing the following simple operation: > >>>>>>> > >>>>>>> a=[1 2 3 4 5]+j*[5 4 3 2 1]; > >>>>>>> c=a'*a; > >>>>>>> > >>>>>>> I do not know whether this is mingw32 only issue or > >>>>>> not. > >>>>>> I reckon it is. At least I don't see this on linux. > >>>>> neither on cygwin > >>>> Yeah, seems a mingw issue. I see it too. > >>>> > >>>> Interestingly, the following does not crash: > >>>> > >>>> a=[1 2 3 4 5]+j*[5 4 3 2 1]; > >>>> b=a'; > >>>> c=b*a; > >>>> > >>>> but shouldn't it do so? I mean this is the same as > >>>> > >>>> c=a'*a > >>>> > >>>> which crashes? > >>>> > >>> It isn't. The first one calls ZGEMM, the second one resolves to ZHERK. > >>> > >> Does that mean it's maybe a lapack problem? The one call crashing, the > >> other not? But the lapack version did not change from octave/mingw 3.2.2 > >> to 3.2.3. That would indeed be strange. > >> > >> benjamin > >> > > Hello > > > > For my build of octave 3.2.3 on mingw by gcc-4.4.0 does not crash, > > > > a=[1 2 3 4 5]+j*[5 4 3 2 1]; > > c=a'*a; > > > > > > > > *********** > > GNU Octave, version 3.2.3 > > : > > Octave was configured for "i686-pc-mingw32". > > : > > octave:1> a=[1 2 3 4 5]+j*[5 4 3 2 1]; > > octave:2> c=a'*a; > > octave:3> c > > c = > > > > 26 + 0i 22 - 6i 18 - 12i 14 - 18i 10 - 24i > > 22 + 6i 20 + 0i 18 - 6i 16 - 12i 14 - 18i > > 18 + 12i 18 + 6i 18 + 0i 18 - 6i 18 - 12i > > 14 + 18i 16 + 12i 18 + 6i 20 + 0i 22 - 6i > > 10 + 24i 14 + 18i 18 + 12i 22 + 6i 26 + 0i > > ********************* > > > > I am using the lapack 3.2.1 while the lapack 3.1.1 seem to be used in > your build. > > But I do not know whether the difference in lapack version is related > to this issue. > > > > As far as I remember, you used the lapack 3.1.1 on octave 3.2.2 build. > > On octave 3.2.2 built by you > > > > a=[1 2 3 4 5]+j*[5 4 3 2 1]; > > c=a'*a; > > > > does not cause crash. > > > > HMMM! > > > > This seem to be curios problem. > > On octave-3.2.3 with gcc-4.4.0 using reference blas/lapack it does not > crash. > Using an SSE3 atlas/blas/lapack library set (from the installer) it crashes. > > Nit, could you test, if your examples still crashes if you select the > reference atlas libraries during install? > > benjamin > -------------------------------------- GyaO! - Anime, Dramas, Movies, and Music videos [FREE] http://pr.mail.yahoo.co.jp/gyao/ ------------------------------------------------------------------------------ Come build with us! The BlackBerry(R) Developer Conference in SF, CA is the only developer event you need to attend this year. Jumpstart your developing skills, take BlackBerry mobile applications to market and stay ahead of the curve. Join us from November 9 - 12, 2009. Register now! http://p.sf.net/sfu/devconference _______________________________________________ Octave-dev mailing list [email protected] https://lists.sourceforge.net/lists/listinfo/octave-dev
