Use coo_array/ndarray with MKL

[bentrainer]

Hi, this may be a stupid question because I just started to learn C++ 😭.

I want to achieve a simple task using nanobind to call mkl_sparse_z_trsm in Python so I could solve Ax=b where A is a sparse upper triangular matrix (scipy.sparse.coo_array) while b is a dense matrix (np.ndarray).

I have not dig into coo_array but stuck at the b part with foo(nb::ndarray<>& b). Because it's an complex array, I don't know how to convert it into MKL_Complex16 type even though I think the numpy also use mkl to store the underlying data. I tried to pass the b.data() into the mkl function but got:

argument of type "nanobind::ndarray<>::Scalar *" (aka "void *") is incompatible with parameter of type "MKL_Complex16 *" (aka "_MKL_Complex16 *")

What is the correct way to convert it to a MKL_Complex16 pointer?

BTW, the reason I want to try this rather than simply go spsolve_triangular or spsolve from scipy.sparse.linalg is because they are way slow compared to MATLAB. I want to figure out where the performance goes...

[wjakob]

Can you use a static_cast?