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

## User-defined stopping test in PARDISO

Beginner
100 Views
Hi,

I have tried to find an answer at this forum and documentation, but I didn't find it.

Can I use my own stopping test in PARDISO (iterative solver)? For example, I can do this at RCI ISS (cg_mrhs_stop_crt.f file in examples).

The second part of my question is - Can I change matrix and right hand side coefficients after each internal iteration?

Thanks,

Nikolai

Employee
100 Views
Dear Nikolai,

> Can I use my own stopping test in PARDISO (iterative solver)?

If we are talking about preconditioned CGS or CG from PARDISO, the answer is yes but the convergence criteria is different. Please take a look at the description of iparm(4). Here is a part ofthe description:
"iparm(4) has the form iparm(4)= 10*L+K.
...

Value L:

The value L controls the stopping criterion of the Krylow-Subspace iteration:

epsCGS = 10-L is used in the stopping criterion

||dxi|| / ||dxi|| < epsCGS

with ||dxi|| = ||inv(L*U)*ri|| and ri is the residue at iteration i of the preconditioned Krylow-Subspace iteration."

> The second part of my question is - Can I change matrix and right hand side coefficients after each internal iteration?

Yes, you can but please take into account the strategy of the PARDISO CGS/CG solver

"Strategy: A maximum number of 150 iterations is fixed with the assumption that the iteration will converge before consuming half the factorization time. Intermediate convergence rates and residue excursions are checked and can terminate the iteration process. If phase =23, then the factorization for a given A is automatically recomputed in cases where the Krylow-Subspace iteration failed, and the corresponding direct solution is returned. Otherwise the solution from the preconditioned Krylow-Subspace iteration is returned. Using phase =33 results in an error message (error =-4) if the stopping criteria for the Krylow-Subspace iteration can not be reached. More information on the failure can be obtained from iparm(20)."

All the best
Sergey