Databases
StackOverflow2013
Postsdbo
POPerformance loss from parallelization
 

Note that there are some explanatory texts on larger screens.

plurals
  1. POPerformance loss from parallelization
    primarykey
    data
    text
    <p>I've modified a raytracer I wrote a while ago for educational purposes to take advantage of multiprocessing using OpenMP. However, I'm not seeing any profit from the parallelization.</p> <p>I've tried 3 different approaches: a task-pooled environment (the <code>draw_pooled()</code> function), a standard OMP parallel nested <code>for</code> loop with image row-level parallelism (<code>draw_parallel_for()</code>), and another OMP parallel <code>for</code> with pixel-level parallelism (<code>draw_parallel_for2()</code>). The original, serial drawing routine is also included for reference (<code>draw_serial()</code>).</p> <p>I'm running a 2560x1920 render on an Intel Core 2 Duo E6750 (2 cores @ 2,67GHz each w/Hyper-Threading) and 4GB of RAM under Linux, binary compiled by gcc with libgomp. The scene takes an average of:</p> <ul> <li>120 seconds to render in series,</li> <li>but 196 seconds (<strong>sic!</strong>) to do so in parallel in 2 threads (the default - number of CPU cores), regardless of which of the three particular methods above I choose,</li> <li>if I override OMP's default thread number with 4 to take HT into account, the parallel render times drop to 177 seconds.</li> </ul> <p>Why is this happening? I can't see any obvious bottlenecks in the parallel code.</p> <p><strong>EDIT:</strong> Just to clarify - the task pool is <strong>only one of the implementations</strong>, please do read the question - scroll down to see the parallel <code>for</code>s. Thing is, they are just as slow as the task pool!</p> <pre><code>void draw_parallel_for(int w, int h, const char *fname) { unsigned char *buf; buf = new unsigned char[w * h * 3]; Scene::GetInstance().PrepareRender(w, h); for (int y = 0; y &lt; h; ++y) { #pragma omp parallel for num_threads(4) for (int x = 0; x &lt; w; ++x) Scene::GetInstance().RenderPixel(x, y, buf + (y * w + x) * 3); } write_png(buf, w, h, fname); delete [] buf; } void draw_parallel_for2(int w, int h, const char *fname) { unsigned char *buf; buf = new unsigned char[w * h * 3]; Scene::GetInstance().PrepareRender(w, h); int x, y; #pragma omp parallel for private(x, y) num_threads(4) for (int xy = 0; xy &lt; w * h; ++xy) { x = xy % w; y = xy / w; Scene::GetInstance().RenderPixel(x, y, buf + (y * w + x) * 3); } write_png(buf, w, h, fname); delete [] buf; } void draw_parallel_for3(int w, int h, const char *fname) { unsigned char *buf; buf = new unsigned char[w * h * 3]; Scene::GetInstance().PrepareRender(w, h); #pragma omp parallel for num_threads(4) for (int y = 0; y &lt; h; ++y) { for (int x = 0; x &lt; w; ++x) Scene::GetInstance().RenderPixel(x, y, buf + (y * w + x) * 3); } write_png(buf, w, h, fname); delete [] buf; } void draw_serial(int w, int h, const char *fname) { unsigned char *buf; buf = new unsigned char[w * h * 3]; Scene::GetInstance().PrepareRender(w, h); for (int y = 0; y &lt; h; ++y) { for (int x = 0; x &lt; w; ++x) Scene::GetInstance().RenderPixel(x, y, buf + (y * w + x) * 3); } write_png(buf, w, h, fname); delete [] buf; } std::queue&lt; std::pair&lt;int, int&gt; * &gt; task_queue; void draw_pooled(int w, int h, const char *fname) { unsigned char *buf; buf = new unsigned char[w * h * 3]; Scene::GetInstance().PrepareRender(w, h); bool tasks_issued = false; #pragma omp parallel shared(buf, tasks_issued, w, h) num_threads(4) { #pragma omp master { for (int y = 0; y &lt; h; ++y) { for (int x = 0; x &lt; w; ++x) task_queue.push(new std::pair&lt;int, int&gt;(x, y)); } tasks_issued = true; } while (true) { std::pair&lt;int, int&gt; *coords; #pragma omp critical(task_fetch) { if (task_queue.size() &gt; 0) { coords = task_queue.front(); task_queue.pop(); } else coords = NULL; } if (coords != NULL) { Scene::GetInstance().RenderPixel(coords-&gt;first, coords-&gt;second, buf + (coords-&gt;second * w + coords-&gt;first) * 3); delete coords; } else { #pragma omp flush(tasks_issued) if (tasks_issued) break; } } } write_png(buf, w, h, fname); delete [] buf; } </code></pre>
    singulars
    1. This table or related slice is empty.
    1. This table or related slice is empty.
    1. PTQuestion
    1. USIneQuation
    1. USIneQuation
    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
    2. PO
      singulars
      1. PTAnswer
    3. PO
      singulars
      1. PTAnswer
    1. This table or related slice is empty.
    1. COAll parallelizing is going to result in some overhead. Sometimes the overhead is more than the cost of just doing a sequential call. There don't have to be bottlenecks for the parallel code to perform worse.
      singulars
      1. POPerformance loss from parallelization
      1. USNominSim
    2. COI understand that, however, the tracing of a single pixel is quite an expensive operation. It did occur to me that the tasks might be too finely grained (i.e. rendering one pixel may not be expensive enough to justify the threading overhead), but after making the granularity more coarse (switch to one image row per thread, as opposed to one pixel per thread) the performance doesn't improve.
      singulars
      1. POPerformance loss from parallelization
      1. USIneQuation
    3. COAre you sure your OpenMP code is corrent? I've used quite a lot of OpenMP but I've never seen it used that way... I've never seen anything like: #pragma omp parallel for num_threads(4) for (int x = 0; x < w; ++x) Shouldn't that be more like: int x; #pragma omp parallel for private(x) for(x = 0; x < w; ++x)
      singulars
      1. POPerformance loss from parallelization
      1. USMFH
 

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