Add funcionality to set/get total number of threads (wrapper for mkl_set_num_threads and mkl_get_num_threads)

[VinceNeede]

This should give to the user the possibility to set the global number of threads, in particular this is the only way to set LAPACK threads (see this discussion on MKL forum, and JuliaLinearAlgebra/libblastrampoline#151). Also a section for Threading control is added in the Readme.

Solves #174

[ViralBShah]

@VinceNeede May I give you commit access here to be able to make new releases and such? It is a pretty slow moving package, but hopefully it doesn't need to wait on me or anyone else.

[hersle]

Can a new release be tagged? It would be nice to access this :)

It confused me that MKL.{set/get}_num_threads mentioned in the README is not available in the latest release.

[ViralBShah]

Releasing in #181. Happy to give folks here commit access to help maintain the package. Please let me know.

[VinceNeede]

I'm sorry for the late answer. Unfortunately I don't feel confident enough for this, yet. Still thank you for the offer!

[ViralBShah]

Totally understand. Note that MKL.jl is a low-risk package to learn the ropes, and it is always possible to fix things or undo them.

[ViralBShah]

@hersle Can I give you commit access for managing this repo? You seemed to be ahead of me in the registration process :-)

[hersle]

@ViralBShah Sure. I can only help with small stuff like that, though, at least for now :)

Thanks both for the PR and swift reply!

[ViralBShah]

@ViralBShah Sure. I can only help with small stuff like that, though, at least for now :)

That's the best help! Invited you.

[m-schwendt]

If I understand this correctly then mkl_set_num_threads should now set the number of threads used by MKL at runtime, right? It doesn't seem to work, though. The only way still seems to set the environmental variable MKL_NUM_THREADS. At least that was the only way I could get different threads when trying it on two different HPC clusters (using Julia 1.11.6 and MKL 0.9.0).

[VinceNeede]

I cannot confirm. It works for me with Julia v1.11.6 and MKL v0.9.0. Could you please open a separate issue? Also note that for Blas and Lapack threads you can now use BLAS.set_num_theads(nthreads) directly from LinearAlgebra

[m-schwendt]

@VinceNeede Sorry, you are right. I accidentally used MKL.BLAS.set_num_threads instead of MKL.set_num_threads.