topic please give us the reproducer in Intel® oneAPI Math Kernel Library

Inconsistency finding Eigenvalues for sparse matrices with MKL

Delgado_Schumann__An — Mon, 17 Jun 2019 13:23:21 GMT

Hi,

I am trying to calculate the smallest real eigenvalues of a general problem A*V = B*V*D with sparse matrices using MKL and I have been having four problems when comparing to MATLAB:

1. I am always missing at least one of the smallest eigenvalues in MKL in comparison to MATLAB.

For example

2. Using the same matrix most of the eigenvalues I do get are almost Identical between MATLAB and MKL, but which of the eigenvalues I am missing is somewhat random.

Same System, different results

3. The smallest eigenvalue tend to be a several orders of magnitudes different between MATLAB and MKL (in some unfrequent cases not only the first)

MATLAB both Graphs: -0.000200

MKL first graph: 380057.918617

MKL second graph: 364053.320270

4. For some sparse matrices describing a very similar system with similar sparcity but different non-zero values I get no eigenvalues on MKL but several hundreds on MATLAB

I attached all data needed to reproduce my results in two Zips docs

Could somebody help me understand what I might be doing wrong or what alternative could I use to solve the problem

Thanks a lot

Andrés Delgado

Andres, is that 32 or 64-bit

Gennady_F_Intel — Wed, 19 Jun 2019 12:29:57 GMT

Andres, is that 32 or 64-bit case? Did you link with the threaded version of mkl? is OpenMP or tbb?

please give us the reproducer

Gennady_F_Intel — Wed, 19 Jun 2019 14:29:57 GMT

please give us the reproducer shows how to you call mkl's routines

Hello Gennady,

Delgado_Schumann__An — Wed, 19 Jun 2019 15:11:54 GMT

Hello Gennady,

We compile in x64. We do not want to run in parallel.

Following is the code used to call feast algorithm:

   _putenv("OMP_DYNAMIC=FALSE");
   _putenv("MKL_DYNAMIC=FALSE");
   mkl_set_num_threads(1);

MKL_INT n = jDof;

   // CSR Matrix A
   double* a = toArray(env, jVals_A);
   MKL_INT* ia = toArray(env, jRows_A);
   MKL_INT* ja = toArray(env, jCols_A);

   // CSR Matrix B
   double* b = toArray(env, jVals_B);
   MKL_INT* ib = toArray(env, jRows_B);
   MKL_INT* jb = toArray(env, jCols_B);

   // search interval & subspace size for solution of eigenvalue problem
   double emin = jEmin;
   double emax = jEmax;
   MKL_INT m0 = jM0;

   // solution variables
   double* e = new double[m0];
   double* x = new double[m0*n];
   MKL_INT* m = new MKL_INT[1];
   double* res = new double[m0];

   // Eigen solver control parameters
   MKL_INT fpm[128];
   MKL_INT loop, info;
   double epsout;

   /* -------------------------------------------------------------------- */
   /* .. Setup eigenvalue solver control parameters. */
   /* -------------------------------------------------------------------- */
   feastinit(fpm);

   fpm[0] = 1;
   fpm[26] = 0;              /* Check input matrices */
   fpm[27] = 0;              /* Check whether B matrix is positive definite */

   info = 0;                 /* Initialize info flag */
   const char* uplo = "F";   /* full matrices are provided*/

   dfeast_scsrgv(uplo, &n, a, ia, ja, b, ib, jb, fpm, &epsout, &loop, &emin, &emax, &m0, e, x, m, res, &info);

thank you for your interest in trying to help us

Best regards

Andrés

Hello everybody,

Delgado_Schumann__An — Mon, 15 Jul 2019 12:43:52 GMT

Hello everybody,

do you have updates on my issue? I need a solution urgently.

Best regards

Andrés