--- Comment #6 from ronis at ronispc dot chem dot mcgill dot ca 2009-07-07
19:45 ---
Turns out that the problem isn't in lapack/gfortran at all; rather the tests
link to blas (here atlas) and one of my blas libs was corrupted. I
rebuilt/reinstalled atlas and the problem goes away, eve
--- Comment #5 from ubizjak at gmail dot com 2009-07-07 18:57 ---
Unfortunately, it is nearly impossible to debug your problem without the
testcase, since the problem can not be analyzed.
Can you at least run the breaking case through a gdb? It will show invalid
instruction at the point
--- Comment #4 from ronis at ronispc dot chem dot mcgill dot ca 2009-07-07
18:45 ---
As per comment #1, I compiled with '-O -pipe' (lapack added -fimplicit-none).
I get the same behavior.
--
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=40673
--- Comment #3 from ronis at ronispc dot chem dot mcgill dot ca 2009-07-07
18:41 ---
As per comment 2, here's my /proc/cupinfo. As to finding a minimized testcase,
this is nontrivial as I don't know if the problem is in a lapack routine or in
the testsuite routine itself. The package
--- Comment #2 from ubizjak at gmail dot com 2009-07-07 17:32 ---
Please add (minimized) testcase and all other info, as explained in [1].
Also, please post /proc/cpuid.
[1] http://gcc.gnu.org/bugs.html
--
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=40673
--- Comment #1 from kargl at gcc dot gnu dot org 2009-07-07 17:28 ---
What happens if you compile with only '-O -pipe'
--
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=40673
