[Numpy-discussion] Why using cblas in umath_test ?

David Cournapeau david@ar.media.kyoto-u.ac...
Wed Mar 4 07:13:54 CST 2009


    I re-enabled umath tests (to test generalized ufuncs), to fix
remaining issues, but I think there is something fundamentally wrong
with it: it assumes cblas is available, which is not true. It happens to
work on (some) Linux and mac os X only because those platforms provide
cblas and blas in the same libraries (Atlas, Accelerate framework).
    Is there a rationale for using cblas at all ? Why not using straight
C functions - it is not like we care about speed for tests, right ? Or
am I missing something ?



More information about the Numpy-discussion mailing list