On Tue, Jun 22, 2010 at 5:27 PM, Dr. David Kirkby <david.kir...@onetel.net> wrote: > On 06/22/10 05:45 PM, Robert Bradshaw wrote: >> >> On Jun 22, 2010, at 12:43 AM, David Kirkby wrote: >> >>> Someone commented the other day that he could not understand why BLAS >>> was in Sage when we had ATLAS. This was in response to my concern >>> about linbox not being able to find BLAS >>> >>> http://trac.sagemath.org/sage_trac/ticket/9101 >>> >>> Robert Bradshaw did a list of what he considered the most important >>> parts of Sage, and includes BLAS in that. I have just checked and see >>> there is no self-test package for BLAS. >>> >>> So I'm wondering if we need BLAS or not. >> >> Me too. FWIW, that was one of them that I was unsure of, but figured >> better safe than sorry and assumed it was used somewhere in the linear >> algebra, so I put it on the important list. > > I suspect it is currently used, but perhaps we can remove it. That will > probably mean changing a few packages which currently link with BLAS, to > force them to link with ATLAS instead. Unfortunately, many of the packages > which link to blas have lots of platform specific code, so things behaving > differently on different platforms. Testing that sort of thing becomes > difficult when you don't have access to all the platforms.
I can tell you the *reason* BLAS is in Sage. Josh Kantor and I included numpy and scipy in Sage, and in order to do that, we needed a BLAS and used the Fortran BLAS (the subject of this thread) and *also* the GSL reference cblas. ATLAS wasn't included into Sage until at least a year later. -- William -- To post to this group, send an email to sage-devel@googlegroups.com To unsubscribe from this group, send an email to sage-devel+unsubscr...@googlegroups.com For more options, visit this group at http://groups.google.com/group/sage-devel URL: http://www.sagemath.org
