xmjiao / petsc4m-legacy Goto Github PK

Thanks for your great work! What you did is exactly what I'm looking for!
I have added your work into my work, and I can also reproduce the example you provided in readme. But it runs quite slowly.
The following is how I build the program.
cd petsc4m;
startup;
build_petsc('-I/data2/DG/RGQ/ThirdParty/mpich/include','-omp'); %Here, the path to the head file of mpich is appointed
%explicitly, or the program will hault with error that mpi.h can't be found;
%'-omp' is also included since my program runs on a cluster with 72
% cores, and the environment variable OMP_NUM_THREADS is set to
% be 72.
init_petsc;

I run a test case after initialization of the program. The time elapsed is 232.30 seconds for the direct solver provided by matlab,
while it takes 1478.16 seconds when use SuperLU as a direct solver. And the time is even much longer for other solvers provided
in the readme file.
Do you have any idea about this?
Thanks in advance!
Best regards!