Solve sparse linear system for complex numbers

[GitHubOpenIssue]

[asistradition]

The MKL sparse QR solver doesn't support complex floats, only reals: https://www.intel.com/content/www/us/en/docs/onemkl/developer-reference-c/2024-0/sparse-qr-routines.html

[GitHubOpenIssue]

Is it true? pyMKL is a Python wrapper of Intel MKL routines and solves sparce linear systems with support for complex numbers.

I'm a bit confused. Maybe the last MKL version erased it?

[asistradition]

The pardiso solver supports complex floats. Wrapping it for python wouldn't be terribly difficult but the interface is complex compared to the QR solver.

[GitHubOpenIssue]

Thank you for your replay,

Yes but I need it for electromagnetc field simulation where fields are complex numbers. That is the reason.

---

[asistradition]

Aight I have added a wrapper for pardisoinit and pardiso in d12521d2eac15d826ee6eb172fdf0d7b9f7b6a66 and they can be called as follows:

from sparse_dot_mkl.solvers import pardiso, pardisoinit

pt, iparm = pardisoinit(11, single_precision=True)

X, pt, perm, error = pardiso(
    A,    # CSR MATRIX
    B,    # Dense Matrix
    pt,
    11,  # mtype
    iparm,
    13   # phase
)

I can't promise every pardiso option will work; I've only added tests for real & complex nonsymmetric A.