From bug-octave-request at bevo dot che dot wisc dot edu Wed Dec 18 10:22:54 2002 Subject: RE: 0*NaN within matrix multiplication From: "John W. Eaton" To: "Lippert, Ross A." Cc: "Paul Kienzle" , "Schloegl Alois" , Date: Wed, 18 Dec 2002 10:22:07 -0600 On 18-Dec-2002, Lippert, Ross A. wrote: | My version of octave hasn't been updated in a while, but I am just | using whatever blas octave comes with. Is ATLAS encumbered such | that it cannot be included in octave as its blas? No, but it is very slow to build from scratch, so I'm not sure that it is the best default version of the BLAS to use with Octave's source distribution. I would guess that most people get binary ATLAS distributions to avoid the build time. But including binary versions of ATLAS in Octave's source package doesn't make much sense to me, and requiring that you go find one before you can build Octave seems bad as well (maybe you don't care about speed or the boundary cases like [0,0]*[NaN;Inf] but you do want to get Octave working quickly enough to solve some problems). The configure script has checks in the following order: BLAS_LIBS environment variable BLAS linked by default (happens on some supercomputers) ATLAS Apple vecLib framework (Mac OS X) Alpha CXML library Alpha DXML library (now called CXML, see above) Sun Performance library SCSL library (SGI/Cray Scientific Library) SGIMATH library IBM ESSL library (requires generic BLAS lib, too) Generic BLAS library If none of these are found, then we use the reference implementation in Fortran. jwe ------------------------------------------------------------- Octave is freely available under the terms of the GNU GPL. Octave's home on the web: http://www.octave.org How to fund new projects: http://www.octave.org/funding.html Subscription information: http://www.octave.org/archive.html -------------------------------------------------------------