topic Weird solution from PARDISO in Intel® oneAPI Math Kernel Library

Weird solution from PARDISO

asd__asdqwe — Mon, 17 Nov 2014 14:22:52 GMT

Hello,

I'm trying to solve a symmetric indefinite system with PARDISO but the returned solution is somewhat strange, it's made of only large values (like 3.53e+18). I'm attaching a reproducer, I would really appreciate some insight into this issue.

Thank you in advance.

It may be the case that the

mecej4 — Mon, 17 Nov 2014 18:12:02 GMT

It may be the case that the matrix is nearly singular. Can you check?

That is right, but it has a

asd__asdqwe — Mon, 17 Nov 2014 20:19:49 GMT

That is right, but it has a really low rank deficiency of 1. For other test cases with similar spectral properties, I have an actual solution, so I was wondering what is happening here.