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

Pardiso error

velvia
Beginner
513 Views

Hi,

I am using Pardiso, from the MKL that ships with icpc 16.0.0 on Mac OSX.

I have already computed a LU factorization, and I want to use Pardiso as an iterative solver using the previous LU factorization as a preconditionner. For that, I've set iparm[3] = 21 for 2 digits of accuracy. But pardiso_64 returns an error -4 and iparm[19] contains -18.

I don't understand what's going on as the last digit of iparm[19] should be either 1, 2, 3, 4 or 5 according to the MKL documentation : https://software.intel.com/en-us/node/521691

0 Kudos
3 Replies
Gennady_F_Intel
Moderator
513 Views

It looks like the issue. Could you give us the example to check the problem on our side? 

0 Kudos
Sheng_L_
Beginner
513 Views

The error code says you have zero pivot, I changed element[2] = -1 to 1, it runs ok.

0 Kudos
Anton_A_Intel
Employee
513 Views

Hi,

In spite of Pardiso returns error=-4 and iparm[19]=-18 the solution is correct for new matrix with element[0] = 2.1 Could you check this behavior on larger matrix on your side or sent it to us also? It seem that only error output is incorrect in such way of pardiso using.  

Best regards, Anton

0 Kudos
Reply