Multithreaded multiplication of large non-reflex, non-square matrices in C / C ++

Question

All,

I am looking for recommendations for C or C ++ libraries (preferably open source) that use multi-threaded methods to multiply large, non-square (e.g. 65536xn, where n <65536) non-sparse matrices. Thanks.

- &&

Answer 1

Intel MKL (not open source)

AMD ACML (free but not open source)

GOTO BLAS (free for academic use, IIRC, and you get the source, but not "open source")

ATLAS BLAS (open source)

Answer 2

Intel Software people seem to be discussing this topic well.