topic Pardiso produces the wrong output in Intel® oneAPI Math Kernel Library

Pardiso produces the wrong output

jbertels01 — Fri, 09 Sep 2011 11:05:03 GMT

Hi there,
Ich have a problem with the sparse solver Pardiso. I don't know whether I don't understand the parameters of Pardiso or whether it is not working with my system. My problem ist that Pardiso produces the wrong output for my input. I also tried the example program pardiso_sym_f90.f90and I thinkI did not work. Maybe it worked but I don't understand the output.It produced

The solution of the system is

x( 1 ) = -4.186020128680939E-002

x( 2 ) = -3.413124159279098E-003

x( 3 ) = 0.117250376805018

x( 4 ) = -0.112639579923180

x( 5 ) = 2.417224446137141E-002

x( 6 ) = -0.107633340356223

x( 7 ) = 0.198719673273585

x( 8 ) = 0.190382963551205
Is this the right solution?
I also tried to call Pardiso myself in the file test.f90.
There I get the following output:

The array ia:

1 2 3 4 5 6

The array a:

(1.000000,0.0000000E+00) (1.000000,0.0000000E+00) (1.000000,0.0000000E+00)

(1.000000,0.0000000E+00) (1.000000,0.0000000E+00)

The array ja:

1 2 3 4 5

The array x:

(0.0000000E+00,1.875000)

(0.0000000E+00,0.0000000E+00)

(0.0000000E+00,1.875000)

(0.0000000E+00,0.0000000E+00)

(0.0000000E+00,0.0000000E+00)

If I got the ja, ia and a arrays right, I don't understad the output.
I'm using Visual Studio 2008and the MKL 10.3 under Windows XP Professional SP2 .

Pardiso produces the wrong output

mecej4 — Fri, 09 Sep 2011 15:55:52 GMT

My remarks apply only to the example program pardiso_sym_f90.f90. The solution that it gives for x agrees with what I get from Matlab.

Why do you think that there is something is wrong? The matrix A is present in the source code, with the upper triangular part neatly formatted. Compute the product A.x and see if it produces a vector whose elements are all = 1.

What do you mean by "it did not work"?

Pardiso produces the wrong output

Alexander_K_Intel2 — Fri, 09 Sep 2011 16:44:22 GMT

Hi,

Just compare your example with pardiso_unsym_complex_f: I've replaced your complex type to complex*16 and set each complex element as pair of numbers (like (1.0d0, 2.0d0)) - after your updated example work correctly.

With best regards,

Alexander Kalinkin

Pardiso produces the wrong output

jbertels01 — Mon, 12 Sep 2011 07:41:01 GMT

Hi,
OK pardiso_sym_f90.f90 works correct (of course). I made a mistake in testing it.
When I changed the type complexin test.f90 into complex*16 it gave me the right output. But how can I use single precision? In my application I only need single precision. Why is it not working with the type complex resp. where is my mistake?

best regards
Julian Bertels

Pardiso produces the wrong output

Alexander_K_Intel2 — Mon, 12 Sep 2011 09:41:03 GMT

Hi,

To use pardiso with single precision please change value of iparm(28) to 1.

With best regards,

Alexander Kalinkin