Solving real unsymmetric system

[Aakash_Dwivedi]

Dear Intel Community,

I am new to MKL library. I was trying to calculate the solution of a real-unsymmetric system. Unfortunately, I'm not able to obtain expected result. I have used the following parameter setting. Could someone please give a feedback.

m_size = 0;

m_initialized = false;

bool symmetric = FALSE;

m_msglvl = 0; // No output

m_type = 11; // isSemidefinite ? -2 : 2;  //LDLT or LLT

memset(m_iparm, 0, sizeof(m_iparm));

m_factorizationIsOk = false;

m_iparm[0] = 1;   // No solver default

m_iparm[1] = 3;   // use Metis for the ordering

m_iparm[2] = 1;   // Numbers of processors, value of OMP_NUM_THREADS

m_iparm[3] = 0;   // No iterative-direct algorithm

m_iparm[4] = 0;   // No user fill-in reducing permutation

m_iparm[5] = 0;   // Write solution into x

m_iparm[6] = 0;   // Not in use

m_iparm[7] = 2;   // Max numbers of iterative refinement steps

m_iparm[8] = 0;   // Not in use

m_iparm[9] = 10;  // Perturb the pivot elements with 1E-13

m_iparm[10] = symmetric ? 0 : 1; // Use nonsymmetric permutation and scaling MPS

m_iparm[11] = 0;  // Not in use

m_iparm[12] = symmetric ? 0 : 1;  // Maximum weighted matching algorithm is switched-off (default for symmetric).

m_iparm[13] = 0;  // Output: Number of perturbed pivots

m_iparm[14] = 0;  // Not in use

m_iparm[15] = 0;  // Not in use

m_iparm[16] = 0;  // Not in use

m_iparm[17] = -1; // Output: Number of nonzeros in the factor LU

m_iparm[18] = -1; // Output: Mflops for LU factorization

m_iparm[19] = 0;  // Output: Numbers of CG Iterations

m_iparm[20] = 0;  // 1x1 pivoting

m_iparm[26] = 0;  // No matrix checker

m_iparm[27] = 0; //double

m_iparm[34] = 1;  // C indexing

m_iparm[59] = 1;  // Automatic switch between In-Core and Out-of-Core modes

Adding to above, I believe that the algorithm used in PARDISO for solving real-unsymmetric system is LU decomposition, which can solve both symmetric as well as unsymmetric matrices. So, do I need to switch to Cholesky setting or can I use LU method for symmetric case too?