Databases
StackOverflow2013
Postsdbo
POhow to use orphaned for loop in OpenMP?
 

Note that there are some explanatory texts on larger screens.

plurals
  1. POhow to use orphaned for loop in OpenMP?
    primarykey
    data
    text
    <p><strong>SOLVED: see EDIT 2 below</strong></p> <p>I am trying to parallelise an algorithm which does some operation on a matrix (lets call it blurring for simplicity sake). Once this operation has been done, it finds the biggest change between the old and new matrix (max of absolute difference between old and new matrix on a per element basis). If this maximum difference is above some threshold, then do another iteration of the matrix operation.</p> <p>So my main program has the following loop:</p> <pre><code>converged = 0; for( i = 1; i &lt;= iteration_limit; i++ ){ max_diff = update( &amp;data_grid ); if( max_diff &lt; tol ) { converged = 1; break; } } </code></pre> <p><code>update( &amp;data_grid )</code> then calls the actual implementation of the blurring algorithm. The blurring algorithm then iterates over the matrix, it is this loop that I am trying to parallelise:</p> <pre><code>for( i = 0; i &lt; width; i++ ) { for( j = 0; j &lt;= height; j++ ) { g-&gt;data[ update ][ i ][ j ] = ONE_QUARTER * ( g-&gt;data[ update ][ i + 1 ][ j ] + g-&gt;data[ update ][ i - 1 ][ j ] + g-&gt;data[ update ][ i ][ j + 1 ] + g-&gt;data[ update ][ i ][ j - 1 ] + ); diff = fabs( g-&gt;data[ old ][ i ][ j ] - g-&gt;data[ update ][ i ][ j ] ); maxdiff = maxdiff &gt; diff ? maxdiff : diff; } } </code></pre> <p>I could just stick a parallel region inside <code>update(&amp;data_grid)</code> but that would mean threads would be created and destroyed on each iteration which I am trying to avoid.:</p> <pre><code>#pragma omp parallel for private(i, j, diff, maxdg) shared(width, height, update, g, dg, chunksize) default(none) schedule(static, chunksize) </code></pre> <p>I have 2 copies of the grid and write the new answer in the "other one" on every iteration by switching <code>old</code> and <code>update</code> between <code>0</code> and <code>1</code>.</p> <p><strong>EDIT:</strong></p> <p>So I've made an orphaned omp for loop as per Jonathan Dursi's suggestion, but for some reason, can't seem to find the maximum value between threads...</p> <p>Here is my "outer" code:</p> <pre><code> converged = 0; #pragma omp parallel shared(i, max_iter, g, tol, maxdg, dg) private(converged) default(none) { for( i = 1; i &lt;= 40; i++ ){ maxdg = 0; dg = grid_update( &amp;g ); printf("[%d] dg from a single thread: %f\n", omp_get_thread_num(), dg ); #pragma omp critical { if (dg &gt; maxdg) maxdg = dg; } #pragma omp barrier #pragma omp flush printf("[%d] maxdg: %f\n", omp_get_thread_num(), maxdg); if( maxdg &lt; tol ) { converged = 1; break; } } } </code></pre> <p>And the result:</p> <pre><code> [11] dg from a single thread: 0.000000 [3] dg from a single thread: 0.000000 [4] dg from a single thread: 0.000000 [5] dg from a single thread: 0.000000 [0] dg from a single thread: 0.166667 [6] dg from a single thread: 0.000000 [7] dg from a single thread: 0.000000 [8] dg from a single thread: 0.000000 [9] dg from a single thread: 0.000000 [15] dg from a single thread: 0.000000 [10] dg from a single thread: 0.000000 [1] dg from a single thread: 0.166667 [12] dg from a single thread: 0.000000 [13] dg from a single thread: 0.000000 [14] dg from a single thread: 0.000000 [2] maxdg: 0.000000 [3] maxdg: 0.000000 [0] maxdg: 0.000000 [8] maxdg: 0.000000 [9] maxdg: 0.000000 [4] maxdg: 0.000000 [5] maxdg: 0.000000 [6] maxdg: 0.000000 [7] maxdg: 0.000000 [1] maxdg: 0.000000 [14] maxdg: 0.000000 [11] maxdg: 0.000000 [15] maxdg: 0.000000 [10] maxdg: 0.000000 [12] maxdg: 0.000000 [13] maxdg: 0.000000 </code></pre> <p><strong>EDIT 2:</strong> Made some mistakes with the private/shared classifiers and forgot a barrier. This is the correct code:</p> <pre><code> #pragma omp parallel shared(max_iter, g, tol, maxdg) private(i, dg, converged) default(none) { for( i = 1; i &lt;= max_iter; i++ ){ #pragma omp barrier maxdg=0; /*#pragma omp flush */ dg = grid_update( &amp;g ); #pragma omp critical { if (dg &gt; maxdg) maxdg = dg; } #pragma omp barrier /*#pragma omp flush*/ if( maxdg &lt; tol ) { converged = 1; break; } } } </code></pre>
    singulars
    1. PO
      singulars
      1. PTAnswer
    1. This table or related slice is empty.
    1. PTQuestion
    1. USlms
    1. USlms
    plurals
    1. This table or related slice is empty.
    1. This table or related slice is empty.
    1. This table or related slice is empty.
    1. PO
      singulars
      1. PTAnswer
    1. This table or related slice is empty.
    1. COAre you sure your OMP implementation actually destroys threads after parallel region? Normally I'd expect it uses a threadpool to avoid this.
      singulars
      1. POhow to use orphaned for loop in OpenMP?
      1. This table or related slice is empty.
    2. COFrom what I can understand it does, in any case however I'd like to know how to apply orphaned directives to this situation.
      singulars
      1. POhow to use orphaned for loop in OpenMP?
      1. USlms
    3. COMost implementations actually use a thread pool to reduce the overhead of destroying and getting threads again for each new parallel region. That said, there is still a fair amount of overhead each time you enter a parallel region - though the first region is usually going to be the most expensive (since it creates the threads). Note - that it depends on the implementation as to when the threads are really first obtained - it could be before the user code gets control, or the first parallel region, or ...
      singulars
      1. POhow to use orphaned for loop in OpenMP?
      1. USejd
 

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