Databases
StackOverflow2013
Postsdbo
POC and OpenMP: pointer to shared read-only data slows down execution
 

Note that there are some explanatory texts on larger screens.

plurals
  1. POC and OpenMP: pointer to shared read-only data slows down execution
    primarykey
    data
    text
    <p>This is my situation, I have an array of pointers that point to arrays of some data... Let's say:</p> <pre><code> Data** array = malloc ( 100 * sizeof(Data*)); for(i = 0; i &lt; 100; i++) array[i] = malloc (20 * sizeof(Data); </code></pre> <p>Inside a parallel region, I make some operations that use that data. For instance:</p> <pre><code> #pragma omp parallel num_threads(4) firstprivate(array) { function(array[0], array[omp_get_thread_num()]; } </code></pre> <p>The first parameter is read-only but is the same along all threads... </p> <p>The problem is that if I use as the first parameter a diferent block of data, i.e.: array[omp_get_thread_num()+1], each function lasts 1seg. But when I use the same block of data, array[0], <em>each</em> function call lasts 4segs.</p> <p>My theory is that there is no way to know if the array[0] will be changed or not by the funciton so each thread asks for a copy and invalidate the copies that other threads have and that should explain the delay...</p> <p>I tried to make a local copy of array[0] like this:</p> <pre><code> #pragma omp parallel num_threads(4) firstprivate(array) { Data* tempData = malloc(20 * sizeof(Data)); memcpy(tempData,array[0], 20*sizeof(Data)); function(tempData, array[omp_get_thread_num()]; } </code></pre> <p>But I get the same result... It's like the thread doesn't 'release' the Data block so other threads could use it...</p> <p>I have to note that the first parameter is not always array[0] so I can't use firstprivate(array[0]) in the pragma line...</p> <p>Questions are:</p> <ul> <li>Am I doing something wrong?</li> <li>Is there a way to 'release' a shared block of memory so other threads could use it?</li> </ul> <p>It was very difficult try to make me understand so if you need further information, please let me know!</p> <p>Thanks in advance... Javier</p> <p>EDIT: I can't change the function declaration because it comes inside a library! (ACML)</p>
    singulars
    1. This table or related slice is empty.
    1. This table or related slice is empty.
    1. PTQuestion
    1. USJavi Ortiz
    1. USJavi Ortiz
    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. VO
      singulars
      1. POC and OpenMP: pointer to shared read-only data slows down execution
      1. This table or related slice is empty.
      1. VTUpMod
    1. COyou say the first arg is read-only, but then you say theres no way to say if it was modified. how many times is `function` called in total by your program? if you do memset many times, runtime will certainly be high.
      singulars
      1. POC and OpenMP: pointer to shared read-only data slows down execution
      1. USjitihsk
    2. COI "know" that the arg will not be changed because the function is part of the ACML lib and this is how it works... What I tried to say is that the compiler/program doesn't know if it will be changed or not and that's why the other copies are being 'invalidated'
      singulars
      1. POC and OpenMP: pointer to shared read-only data slows down execution
      1. USJavi Ortiz
    3. COyou could try this. declare data** outside the parallel region, but put `for(i = 0; i < 100; i++) array[i] = malloc (20 * sizeof(Data);` inside the par region so that every thread mallocs its own data blocks. of course `i<100` will change to something like `i<100/omp_get_num_threads()`. I am still not sure why case 2 did not work, but am interested to know the result of this.
      singulars
      1. POC and OpenMP: pointer to shared read-only data slows down execution
      1. USjitihsk
 

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