[Scipy-tickets] [SciPy] #1170: integrate.quad crashes with SIGILL

SciPy Trac scipy-tickets@scipy....
Sun Jun 6 10:36:20 CDT 2010


#1170: integrate.quad crashes with SIGILL
----------------------------------+-----------------------------------------
 Reporter:  Joe                   |       Owner:  rgommers
     Type:  defect                |      Status:  new     
 Priority:  normal                |   Milestone:  0.8.0   
Component:  Official binaries     |     Version:  0.7.0   
 Keywords:  scipy integrate quad  |  
----------------------------------+-----------------------------------------

Comment(by rgommers):

 The arpack file above contains sse instructions for both the 2.5 and 2.6
 superpacks.

 In numpy/tools/win32build/misc/ a similar script to Pauli's can be found:
 x86analysis.py
 There's a note in it that detection may not be completely reliable - it
 can give false positives for code that is only executed after CPU
 detection.

 Tested the blas/lapack/atlas binaries in svn/vendor as well, they are
 fine. Setuptools/distribute seems to not have anything to do with this.
 f2py however may be using sse instructions by default, at least that's
 what I got from some old mailing list threads (google "f2py sse sse2",
 first hit). Arpack does contain Fortran code so this may be it.

 SSE goes back as far as 1999 though, so some SSE instructions will
 probably be OK no matter what CPU one has. Should be fixed though, but
 this is not super urgent.

-- 
Ticket URL: <http://projects.scipy.org/scipy/ticket/1170#comment:16>
SciPy <http://www.scipy.org>
SciPy is open-source software for mathematics, science, and engineering.


More information about the Scipy-tickets mailing list