Just enrolled in a R Data science course. While setting up the environment thought why use the slow inbuilt BLAS when I can use Intel MKL.
Last time it was a great experience in python with the intel’s conda packages.
Upon digging a bit deeper there seems to be some weird problems with it.
Octave gives wrong result when using mkl shared libraries as default libblas.
https://savannah.gnu.org/bugs/index.php?58926
https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=921207
There also seems to be issues when using MKL with TBB in R. Some people recommended to use GCC OpenMP instead.
https://github.com/eddelbuettel/mkl4deb/commit/c6d8d843346c71cd1ad0703c635f7d481127f950
Just about then Level1 released the video. So I guess maybe you guys here know more about it.
For now I’ll just use the “slow” BLAS. I’m just starting out, probably won’t NEED that extra performance.