Auto-suggest helps you quickly narrow down your search results by suggesting possible matches as you type.

Showing results for

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

Prapti_N_

Beginner

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

06-17-2014
08:52 AM

40 Views

Pardiso L and U factors

I have been trying to figure out a way to extract the L and U factors of a matrix using pardiso. I read somewhere that this is not possible, but I just wanted to check with people here to see if there is a way to do it.

Thanks for the help!

Link Copied

3 Replies

mecej4

Black Belt

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

06-17-2014
09:10 AM

40 Views

Pardiso is an example of a "black-box" routine. You are trying to peek into the black-box, with some expectations of how it should work, and you seek to do things with Pardiso that it was not intended to do.

Do you have a definite purpose for the L and U factors, supposing that they were obtainable? How do you propose to handle the question of permutation?

Prapti_N_

Beginner

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

06-17-2014
09:37 AM

40 Views

Ying_H_Intel

Employee

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

06-18-2014
11:16 PM

40 Views

Hi Prapti N.

You can get the permuation vector or diag element i.e using pardiso_getdiag , or phase 332 to get D when A=LDL*.

As mecej4 mentioned, you can't get L and U directly from pardiso, but what kind of operation you hope to do with the L and U?

For example, you can use phase 331 to L*y=b and phase=333 do Ux=y. Where you can specify b and y.

Best Regards,

Ying

A real symmetric positive definite matrix A(mtype= 2) is factored by Intel MKL PARDISO as A= L*LT

. In this case the solution of the system A*x=bcan be found as sequence of substitutions: L*y=b(forward

substitution, phase=331) and L*x=y(backward substitution, phase=333).

A real nonsymmetric matrix A(mtype= 11) is factored by Intel MKL PARDISO as A= L*U. In this case the

solution of the system A*x=bcan be found by the following sequence: L*y=b(forward substitution, phase

=331) and U*x=y(backward substitution, phase=333).

Solving a system with a real symmetric indefinite matrix A(mtype= -2) is slightly different from the cases

above. Intel MKL PARDISO factors this matrix as A=LDLT , and the solution of the system A*x=bcan be

calculated as the following sequence of substitutions: L*y=b(forward substitution, phase=331), D*v=y

(diagonal substitution, phase=332), and finally L

T

*x=v(backward substitution, phase=333). Diagonal

substitution makes sense only for symmetric indefinite matrices (mtype= -2, -4, 6). For matrices of

other types a solution can be found as described in the first two examples.

CAUTION

For more complete information about compiler optimizations, see our Optimization Notice.