Intel® Math Kernel Library

misunderstanding of pardiso_64

Dear sir or madam:

Hi, I need to use mkl pardiso to help me solve problem in 64bit linux system. Because I use 64bit-integer (ILP64 interface layer), I need to make use the function, compile command and library to link is correct. However, I get a little confuse about pardiso_64 function.

From the website

SVD runs very slow and sometimes doesn't return for matrices which cannot converge

Hi Guys

I have been struggling with the MKL SVD computation recently. So I really appreciate if anyone can help out.

Basically I am using MKL SVD to calculate the first two eigenvectors for a matrix. I use the LAPACK_dgesvd method in the MKL to do the job. It runs fairly fast for matrix which can converge. But for those who can't converge, MKL SVD runs extremely slow and sometimes it never return. I am not sure what is going on behind the scene. Maybe MKL just tries exhaustively when the matrix can't converge?

My questions is:

MKL numpy/scipy not passing tests


I've been pulling my hair out for a few days trying to get a build of numpy/scipy against the MKL to pass all the scipy tests.  I've ran out of ideas to try, so any ideas on what to try next would be greatly appreciated!  My guess is that I'm missing an ifort flag, but I've tried a bunch of different combinations without improving the situation.

My system is running 64bit Ubuntu 13.04 with a Intel Core i7-4700MQ CPU with Python 2.7.6.  I've been using icc and ifort versions 14.0.2, and MKL version 11.1.2.  I've been building Numpy 1.8.0 and Scipy 0.13.3.

Pardiso test example fail

I am trying to debug my implementation of PARDISO, but unfortunately I cannot get even a simple example to work.

I created a 6x6 matrix with 16 non-zero elements and used the mkl ddnscsr routine to obtain my acsr, ia, and ja vectors to feed into the PARDISO routine. Whenever i plug things into the function call nothing happens; my code just stops running after the call to PARDISO. 

Here is my code:

double *tempB;

Progress bar for FEAST eigensolver

Does anyone know if it is possible to determine the progress of an eigenvalue analysis using the Feast solver?

I use it to calculate the natural frequencies of buildings and bridges and, for medium size structures, getting 40 eigenvalues is sometimes taking 15 or 20 minutes with absolutely no indication of how far through the solution it is. It is really disconcerting having to wait that long with no feedback at all.

I would like to be able to display a progress bar or, failing that, just show which iteration it is working on.

VML for safe evaluation of function


         I plan to use VML to evalaute the following functions:

y = x / (exp(x) - 1)   and dy/dx =  1/(exp(x)-1)  - x exp(x)/(exp(x)-1)^2

I have to make sure that, i do not get NAN or Infinity at x =0 or x << 1

Any suggestions on how to write the subroutine.

for y i use, expm1 and code works fine i.e, limit x->0 y = 1

using expm1 in dy/dx i am getting, infinity for small x  say 1D-20.





Subscribe to Intel® Math Kernel Library