Databases
StackOverflow2013
Postsdbo
PO
 

Note that there are some explanatory texts on larger screens.

plurals
  1. PO
    primarykey
    data
    text
    <p>I have a few comments that you might find useful.</p> <p>In general I try to avoid accessing the thread number directly (thread_id). Instead of defining arrays with a size equal to the number of threads you should define your variables in the parallel block (this automatically makes them private). Then instead of looping over the number of threads after the parallel block you should use an atomic, critical, or single approach (I'm not sure which is best here). Something like this.</p> <pre><code>Point* farthestPoint; //int distance = *distance_sub[0]; //is this a bug shouldn't distance be a double? double distance = 0 #pragma omp parallel private { Point farthest_sub = input[0]; double distance_sub = 0; #pragma omp for nowait for (int index = 1; index &lt; input.size(); index++){ // for loop code } #pragma omp critical { if (distance &lt; distance_sub){ farthestPoint = farthest_sub; } } #pragma omp barrier //next part of code } </code></pre> <p>One problem you might have now is false sharing. Each thread is trying to write to arrays in the same cache line (e.g. in the array distance_sub[num_threads]). I'm not sure what OpenMP does when you declare the values inside the the parallel block but I suspect that it's more likely to allocate the values to avoid false sharing.</p> <p>Another comment you, should try and avoid calling OpenMP so many times, especially on a small number of elements. There is an overhead with OpenMP. I would try and get as much in one parallel block as possible using a barrier or whatever.</p> <p>Also, in your code you have <code>int distance = *distance_sub[0]</code> is this a bug? Shouldn't distance be a double?</p> <p>Lastly, a pedantic point. I doubt you have 8 cores. You probably have 4 cores and 8 hardware threads due to Intel hyper threading. The distinction can be important.</p>
    singulars
    1. This table or related slice is empty.
    1. POParallelizing QuickHull: OpenMP gives small speedup whilst TBB gives negative speedup
      singulars
      1. PTQuestion
    1. PTAnswer
    1. This table or related slice is empty.
    1. This table or related slice is empty.
    plurals
    1. This table or related slice is empty.
    1. This table or related slice is empty.
    1. POParallelizing QuickHull: OpenMP gives small speedup whilst TBB gives negative speedup
      singulars
      1. PTQuestion
    1. This table or related slice is empty.
    1. VO
      singulars
      1. PO
      1. This table or related slice is empty.
      1. VTUpMod
    2. VO
      singulars
      1. PO
      1. This table or related slice is empty.
      1. VTAcceptedByOriginator
    1. COthanks, I got a another 2x speedup because of this!
      singulars
      1. PO
      1. USMadsen
    2. COGlad I could help. So you're getting a 4x speedup now? I think that's the best you can expect with 4 cores (unless you make other optimizations to your single threaded code such as using SIMD). You might be able to get a bit more from hyper-threading. Have you tried schedule(dynamic)? Anyway, if you think my answer is the best one feel free to mark it as one.
      singulars
      1. PO
      1. This table or related slice is empty.
 

Querying!

 
Guidance
A row detail

Detail views are divided into sections. All the information in the data section comes from columns in the selected row. The other sections display data from other, related rows.

Related data can be related in a to-one or a to-many fashion. Captions of data related in a to-many fashion link to a list view showing a filtered view of the table.

Try moving around until you find a non-empty to-many entry and click on the label to get to one. You can move back to the root by clicking on the database name in the header.

SQuiL has stopped working due to an internal error.

If you are curious you may find further information in the browser console, which is accessible through the devtools (F12).

Reload