[Numpy-discussion] Use OpenBLAS for the binary releases?

[Sturla Molden]

On 19.11.2012 18:42, Dag Sverre Seljebotn wrote:

> Even on CPUs that are not directly supported, this is at least better
> than reference BLAS.
>
> (On our AMD CPUs, which are too new to have a separate OpenBLAS
> implementation, the implementations for older AMD CPUs still outperform
> at least Intel MKL, because MKL does so poorly on these -- although ACML
> beats them both by a factor 2. And of course on supported CPUs
> (everything Intel and older AMD) OpenBLAS is wonderful.

Indeed.

I have a laptop with a quadcore AMD Phenom II, but ACML consistently 
segfaults when used with MinGW (gcc or gfortran).

With another set of compilers (Microsoft C and Absoft Fortran), ACML 
usually works, but sometimes decides my CPU is not supported, and just 
terminates the process.

I'm not sure NumPy even builds against ACML, e.g. it has a non-standard 
C interface for BLAS and LAPACK. Also, it wouldn't work with GNU 
compilers on Windows. Dependency on Intel or PGI run-time DLLs would 
also make it useless for the binary superpacks.

OTOH, OpenBLAS works wonderfully :)


Sturla