0 Replies Latest reply on Mar 11, 2011 10:23 AM by wpoely68

    OpenMP + VTune

    wpoely68

      Hello,

       

      I've got a question about understanding the results from VTune. I have got a program that runs over a loop in which it has to calculate eigenvectors and eigenvalues for a lot of matrices. To make it runs faster, i have used OpenMP to run the loop over the matrices in parallel. It works but not as good as i expected. Now, i have used VTune to get an idea why it is slower then expected but i have trouble understanding the results. For example, if i run a "lock and wait" and look at "thread concurrency", i see the openmp threads but according to VTune, only 1 thread is running and the other are waiting. But all of them are using cpu time. What is happening? Why are they using cpu time but not running according to VTune? It's a for loop that runs in parallel, it should all just work...

       

      Thanks in advance,

       

      Ward