- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Email to a Friend
- Report Inappropriate Content

As you know the dposv is a function that solve A.X = B while the A is a symmetric matrix.

The dposv only needs the upper or lower part of symmetric matrix as it can be observed in this example.

Now I have to solve a 33000x33000 double precision matrix which takes over 8.1 GB of memory.

Can making a 33000x33000 array be avoided so I don't have to pass the zero members?

Is there any alternative solving function that provides that?

I know there are solvers out there that can solve this type of problems very efficiently; Please help me to make most of MKL.

- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Email to a Friend
- Report Inappropriate Content

If your matrix is banded, you may use DPBSV from Lapack. If the matrix has a more general sparseness pattern, MKL offers you a fine solution: Pardiso, which you may call directly, or through the DSS interface. For details, see the MKL manuals. There are several small example problem source codes that are included in MKL in the examples/solverf and examples/solverc directories.

Link Copied

- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Email to a Friend
- Report Inappropriate Content

If your matrix is banded, you may use DPBSV from Lapack. If the matrix has a more general sparseness pattern, MKL offers you a fine solution: Pardiso, which you may call directly, or through the DSS interface. For details, see the MKL manuals. There are several small example problem source codes that are included in MKL in the examples/solverf and examples/solverc directories.

- Subscribe to RSS Feed
- Mark Topic as New
- Mark Topic as Read
- Float this Topic for Current User
- Bookmark
- Subscribe
- Printer Friendly Page