1) I am NOT a dev (of ACML etc), but this is where the ACML website sends me to ask questions. (The GitHub pages are not appropriate for my following questions, as I need to ask a common question of all of these three libraries.) If this is not the right place for my question, could you kindly point me in the right direction? Thanks.
2) I have an AMD Phenom(tm) II X6 1100T Processor, which has a K10 microarchitecture. My OS is Ubuntu 14.04. I need optimized BLAS, lapack libraries, on top of which I'll install numpy, scipy (python 2.7) and MUMPS. (I need the sequential, single core versions). I'll be inverting a lot of very big sparse matrices. (This installation set is necessary, as I'll be running a specific simulation tool: Quantum transport simulations made easy - Kwant .) Should I go with CBLAS+ACML, or BLIS+libflame, or is OpenBLAS better? None of these libraries mention the K10 architecture.
3) I have been following Speeding up the solvers — eofs 0.5.0 documentation (and its many variants on the internets) for installing ACML, but after many tries, I can't get numpy to work with ACML. Essentially, I get the error described here: Numpy-discussion - Problems when using ACML with numpy . I've been trying to install ACML on the assumption that it ought to provide better optimization for K10. This assumption is somewhat baseless: The others make no mention of K10 when they discuss microarchitectures. ACML doesn't either, but it doesn't mention any microarchitecture at all. It just talks about instruction sets and compilers. Assuming my assumption is the correct one, is there a good guideline to install ACML for my machine?
Thank you all for your concern and time.