pardiso_schur.c with MKL_PARDISO

[bopeng]

Hi,

I am using 2021.04 oneAPI release.

/opt/intel/oneapi/mkl/latest/examples/c/sparse_directsolvers/source/pardiso_schur.c

currently the setting are:

MKL_INT perm[8] = {0, 0, 0, 0, 0, 0, 1, 1};

It assumes the last two row and col 2x2 will be think as A22.

If I change that to

    MKL_INT perm[8] = {1, 1, 0, 0, 0, 0, 0, 0};

I thought it will do reordering for me, but after change that, the calculated results are wrong.

Can you point out what's wrong?

@Kirill_V_Intel 

I am an employee of Intel, ping me directly if any need.

Thanks so much!

-------------------------------------------------------------------------

The reason I am doing this:

My own matrix is

    A11=1482x1482, A22=78281x78381

If I want to reuse this pardiso_schur.c, S=A22-A21 * A11-1 * A12 will be 78281x78281

in such situation, S*xl=bl need to solve 78281x78281, it needs huge computation time.

[bopeng · Accepted solution]

Applying following patch will make it works, fixed, thanks for your support!

===============================================================================

- MKL_INT perm[8] = {0, 0, 0, 0, 0, 0, 1, 1};
+ MKL_INT perm[8] = {1, 1, 0, 0, 0, 0, 0, 0};
MKL_INT ipiv[2];
MKL_INT n_schur = 2; /* Schur complement solution size */
/* -------------------------------------------------------------------- */
@@ -191,7 +191,7 @@ int main (void)
printf("info after LAPACKE_dsytrf = " IFORMAT " \n", info);
return 1;
}
- info = LAPACKE_dsytrs(matrix_order,uplo,n_schur,nrhs,schur,n_schur,ipiv,&b[n-n_schur],nrhs);
+ info = LAPACKE_dsytrs(matrix_order,uplo,n_schur,nrhs,schur,n_schur,ipiv,&b[0],nrhs);

[VidyalathaB_Intel]

Hi,

Glad to know that your issue is resolved and thanks for letting us know.

As the issue is resolved we are closing this thread. Please post a new question if you need any additional information from Intel as this thread will no longer be monitored.

Have a Nice Day!

Regards,

Vidya.