Welcome to Intel® Software Network Quick Login | Join | Help |
Search in Intel® Software Network Forums
in Go

Parallel Programming with IMSL?

Last post 07-12-2008, 6:13 PM by liubing84. 3 replies.
Sort Posts: Previous Next
 07-11-2008, 7:37 PM 30258949  

Parallel Programming with IMSL?

My operation environment is:  Windows XP, Microsoft Visual Studio .net 2005 plus Intel Visual Fortran 10.1. I have also installed IMSL 5.0 and MKL 10.0. In my parallel program, (exactly, within OpenMP parallel dirctives) I need to call functions to calculate A*x=b(a linear equation).  However, I do not understand the corresponding description  in the documentation, and it is not clear according to my operation environment. Would someone please help me and give some instructions? 

 Thank you for your kind attention!

 
 07-12-2008, 7:09 AM 30258961 in reply to 30258949  

Re: Parallel Programming with IMSL?

As MKL implements the standard BLAS functionality, standard BLAS documents are useful.  You could use either the original fortran 66/77 or the fortran 90/95 setup.  When you choose MKL thread linkage, MKL automatically implements parallelism.   If you are controlling parallelism with your own OpenMP directives, and don't want MKL to start threads itself, use the sequential library link.
Your question is potentially book length.
 
 07-12-2008, 8:58 AM 30258964 in reply to 30258961  

Re: Parallel Programming with IMSL?

If part of your question is regarding the 'how' to solve systems of linear equations using the MKL (i'll leave the parallel part out of this posting):

MKL has functions for LU factorization (getrs). Unless I’m going totally off-the-plot, you should be able to decompose a matrix into a product of a lower triangular matrix and an upper triangular matrix then use the LU factorization of self to solve the linear system Ax = b.

Given a matrix A, an LU factorization should factor A as follows:

PA = LU;

Where P is its permutation matrix, L is its lower triangular matrix with ones on the diagonal, and U is its upper triangular matrix.

Once an LU factorization is formed, it can be reused to solve for different right-hand sides, to compute: inverses, condition numbers, etc.

I don't know if this is part of what you were asking, but I figured it would not hurt to chime in some details (as long as i've not totally lost it) :)



Ben
 
 07-12-2008, 6:13 PM 30258973 in reply to 30258961  

Re: Parallel Programming with IMSL?

Thank you very much! I will have a try!  ^_^
 
View as RSS news feed in XML

Shortcuts


Tags For This Post

...

Community Tags

...