On 4/26/11 12:57 PM, Jason Grout wrote:
On 4/25/11 5:13 PM, Rob Beezer wrote:
On Apr 25, 2:56 pm, John H Palmieri<jhpalmier...@gmail.com> wrote:
I can answer part of this: on Mac OS X, we don't install ATLAS at
all, we
just rely on the system's version. I can't answer the rest of your
questions.
Thanks, John. So I wonder if somebody could exercise the Mac system
ATLAS/LAPACK with a Fortran or C example? The FORTRAN example (link
above) might even be good enough to provide evidence if the second
unitary matrix has a row (or column) of all zeros.
Short update. Karl-Dieter and I exercised his system this morning. I
think we've traced the problem back to a false assumption numpy makes
when using a lapack function. Instead, I think numpy should use a
different lapack function in certain cases. See the trac ticket for
details.
Another short update: it appears that while there is still a false
assumption from numpy about the return value of the lapack function,
that is not what is biting us here (the lapack function docs were
ambiguous and I interpreted it wrongly).
See http://icl.cs.utk.edu/lapack-forum/viewtopic.php?f=2&t=2402
The lapack people are looking for an example code and system details
exhibiting the error. We're working on the trac ticket to provide this.
So it seems that we actually found a bug in the Accelerate framework
on certain systems.
Thanks,
Jason
--
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
