Intel® oneAPI Math Kernel Library
Ask questions and share information with other developers who use Intel® Math Kernel Library.
7032 Discussions

MKL Sparse Solver and maximum number of unknowns



I have been using MKL for its sparse solver capabilities. My code is a finite element analysis solver. My code solves piezoelectric problems. Each element has 20 nodes and each node has 4 degrees of freedom. If am looking at a 10,000 element problem we are looking at (20*10000*4) 800K number of unknowns.

I am running in Windows 32 bit application so I am limited to the 2 Gigs limit imposed by Microsoft. Based on this limit what is the maximum number of unknowns (degrees of freedom) I can solve for using the MKL sparse solver, anyone looked at this?

Any help on this will be greatly appreciated.

Marcias Martinez

0 Kudos
1 Reply
New Contributor III
I use MKL for finite element work as well. The amount of memory needed is dependent on the sparse structure of the matrix, not necessarily the number of unknowns.

As an alternative to MKL, I would suggest also looking at TAUCS for an out-of-core sparse solver for positive definite symmetric matrices. I have used that on problems that use far more than 2G.
0 Kudos