topic Re: problem with dgetrf, LU-decomposion in Intel® oneAPI Math Kernel Library

problem with dgetrf, LU-decomposion

shapovalov — Thu, 12 Mar 2009 11:00:33 GMT

I want to calculate LU-decomposition of square matrix. I call function dgetrf() but returned result is uncorrect.

double A[100,100]; int Size=100; int *ipiv = new int[100]; int info;
/*...*/
dgetrf(&Size,&Size,A,&Size,ipiv,&info);

I think, matrix A beforedgetrf()must be = P*L*U after dgetrf(), where P can be obtained from ipiv. Please,check my code.

Re: problem with dgetrf, LU-decomposion

TimP — Thu, 12 Mar 2009 13:51:52 GMT

You don't give enough context to test whether your A looks like a Fortran double subscripted array; in general, it will not. You must refer to the reference source code for the interpretation of ipiv. As you don't give anything to check, there are several other points on which you could make mistakes.

Re: problem with dgetrf, LU-decomposion

Gennady_F_Intel — Sat, 14 Mar 2009 19:05:18 GMT

Oleg, there are some Lapack examples where you can find how these routines calling the right way.
Please see: exampleslapacksource ...dgeconx.f, dgerfsx.f, dgetrfx.f, dgetrix.f, dgetrsx.f
--Gennady