Is SparseMatrix use nativeMKL to speed up in v4.7?

[tmp-coder]

hi, all

i do a test on sparse matrix multiply, (rowCount, colCount) = (1000,1000), and nonzero count = 22328, and sparse matrix is here https://paste.ubuntu.com/p/c82q42yKsf/

while result show mathnet.numerics is slower than scipy about 5 times, scipy run 0.007s, while mathnet run 0.036s

code

var mat = GenerateRandomSparseMatrix(nr, nc, true);
           
 var elapsedtime = Utils.MeasureTime<Matrix<double>>(() => (mat.Transpose().Multiply(mat)));
  Console.WriteLine("elapsed time = {0}s", elapsedtime);

py

mat = pd.read_csv(filePath,header=None).values

mat = sps.csr_matrix(mat)
print("elapsed = {}s",utils.measure_time(lambda : mat.T * mat))

[cdrnet]

No, not yet unfortunately (only dense).

[cdrnet]

There is a sparse solver native provider now - but the other sparse operations are still not covered.