Hi there, I get a desktop machine having two Intel Core2 processors, to utilise the mutli-processor machine with OpenMP, I wrote a benckmark code to verify the supportness of OpenMP in Visual Studio 2005.
When I increase the numbe of parallel thread from One to Two, I saw the improvement in performance which is in term of the running time nearly decreased by half; however when I continue increase the number of threads to 4, I didn't see much improvement, the perofrmance seems stay constantly at the level of 2 parallel thread.
In Solaris system, when I submitted a such parallel code, I can specify how many cpu resouce I want for executing it, does anybody know do I need to do that in VS 2005 and how?
There is the code
for (int loop = 0 ; loop < 10; loop++)
{
time[loop] = 0.0;
__int64 ctr1 = 0, ctr2 = 0, freq = 0;
QueryPerformanceCounter((LARGE_INTEGER *)&ctr1);
int tid, i, j;
omp_set_num_threads(NUM_THREAD);
#pragma omp parallel shared(A,B,D) private(tid,i,j)
{
tid = omp_get_thread_num();
for (j = tid; j < MAXLOOP; j += NUM_THREAD)
{
for(i = 0; i < MAX; i++)
D [ i ] = A [ i ] * A [ i ] + B [ i ] * B [ i ] + A [ i ] * B [ i ] ;
}
}
QueryPerformanceCounter((LARGE_INTEGER *)&ctr2);
QueryPerformanceFrequency((LARGE_INTEGER *)&freq);
time[loop] = (ctr2 - ctr1) * 1.0 / freq;
AverageExTime += time[loop];
}
AverageExTime = AverageExTime/10;
