that’s how I read it too in general, unless you want squeeze the maximum performance you can of a specific application, then you will need to benchmark each implementation, (including other options such as mkl) and see in practice which is superior for you.
Meh I don’t use it enough to warrant the effort of benchmarking. I just wanted a really general “if you don’t know what you want, your should probably use X”, and it appears that I should just use blas-openblas.
yes definitely, I am also going to just install blas-openblas and forget about it, however I just added the part about benchmarking for people that run research/ machine-learning stuff regularly , figuring out which implementation is faster for their use case is worth it.
that’s how I read it too in general, unless you want squeeze the maximum performance you can of a specific application, then you will need to benchmark each implementation, (including other options such as mkl) and see in practice which is superior for you.
Meh I don’t use it enough to warrant the effort of benchmarking. I just wanted a really general “if you don’t know what you want, your should probably use X”, and it appears that I should just use
blas-openblas.yes definitely, I am also going to just install blas-openblas and forget about it, however I just added the part about benchmarking for people that run research/ machine-learning stuff regularly , figuring out which implementation is faster for their use case is worth it.