Closed dajuno closed 8 years ago
Hello @dajuno, you raise a very valid concern. The easiest thing for me was to use Intel MKL libraries at first, and that's why I chose it over other alternatives. I accept advice, guidance or even pull requests to make FEniCS + OpenBLAS a reality :)
Here's the other reason: OpenBLAS has been available in Anaconda for just one month.
I am about to make it but I need to build petsc4py against NumPy 1.9 and there's an issue blocking me: https://github.com/conda/conda-build/issues/648
Great, thanks! :)
By the way, I just saw this: https://twitter.com/bryvdv/status/660672659752022016 "the MKL optimizations for NumPy available free to all"
Hello,
I just installed fenics with conda through your repository. I see you included the commercial MKL library Vendor: Continuum Analytics, Inc. Package: mkl Message: trial mode expires in 30 days <--- !!
while the fenics source installer uses ATLAS+LAPACK, and so do (of course) the versions included in the Linux distributions. conda numpy uses openblas.
Why have you chosen not to use the free packages over the MKL demo? Can I build fenics without it, using conda and your recipes?
Thanks!