Pardiso Output = -1.#IND00

[rditz]

I am currently using PARDISO SPARSE MATRIX SOLVER for Finite Element Method problems.I am working on Microsoft Visual C++ 2010.I am getting the output as -1.#IND00 in the following cases-

1)When i increase the coefficient matrix elements by the order of 10e5 or more form(10e-1 to 10e+4).The solver gives correct solution in the former case.

2)When I increase the size of my coefficient matrix .

Note that in both the cases Pardiso shows no error during the input phase.It just shows all the solution variables as -1.#IND00.

Any quick and useful help will be appreciated.

Thank you.

[Alexander_K_Intel2]

Hi,

It would be great if you provide small testcase to reproduce this issue on our side. Maybe there are some problems in matrix structure, maybe some input parameters like iparm are incorrect, and maybe condition number of input matrix is really huge. To choose correct answer we need to run testcase.

With best regards,

Alexander Kalinkin

[rditz]

Hi, sorry but I am new to this.What do you exactly mean by testcase ? Should I post my code or should I send the .cpp file.Also sparse matrix solver is a part of a bigger FEM code.Will it be possible for you to pinpoint the error ?

Also I think there cannot be a mistake with iparm as it is solving correctly for lesser(magnitudewise) coefficients of A matrix.

Thank you for the reply.

[Alexander_K_Intel2]

Hi,

You could send .cpp example or file with ia, ja, a, rhs, iparm and correct solution arrays. Any variant would be convenientfor us if we reproduce your issue.

With best regards,

Alexander Kalinkin

[Gennady_F_Intel]

please use this post to see how attach the code to the forum.

[rditz]

Hey thanks for the reply.As I said earlier Pardiso solver is part of a larger code.I have attached the code and 3 txt files.Open each files through freopen.Note the output of each file.

1) 3bricks......smaller magnitude coefficient + small size of matrix
2) 3bricks_diff._coefficients...larger magnitude coefficient + small size of matrix
3) Increased_size_of_matrix...smaller magnitude coefficient + larger size of matrix

Once again thanks for the reply and the help