topic Hello Stefano, in Intel® oneAPI Math Kernel Library

MKL Pardiso (version 11.2.3): wrong output of phase 331 with multiple rhs and Schur complement enabled

zampini__stefano — Mon, 12 Oct 2015 16:35:21 GMT

Hi,

I recently started using MKL_PARDISO. I noticed that phase 331 gives the wrong result if you want to solve for multiple right hand sides using the Schur complement feature.

Attached a code to reproduce the problem. I just copied the example you provide with mkl distribution for the Schur complement and added multiple rhs.

I'm using composer_xe_2015.3.187, with MKL 11.2.3

Stefano

Just a clarification about

zampini__stefano — Mon, 12 Oct 2015 18:55:17 GMT

Just a clarification about the issue.

Since the 3 rhs I'm using are all equal to one, I'm expecting that the output of phase 331 will be the same for x[0...7],x[8...15] and x[16...23]. Instead, the numbers in x[8...15] and x[16...23] (that are the same) are not in the same order as in x[0..7].

Hello Stefano,

asd__asdqwe — Tue, 13 Oct 2015 08:51:00 GMT

Hello Stefano,

Here are some problems in your piece of code :

You need to set iparm(36) to 2 instead of 1
If you want the same results from phase = 33, you need to call phase = 331, then 332 (or dsytrs as done in the MKL example), then 333.

Here is a program that returns the same results.