Hi everyone,
I’m using OpenMP to run 2 threads in 2GPU’s (580GTX). The parallel section is inside of a loop and it works fine. However, in some iterations, the code runs in series. I don’t know why this is happening.
About 20% of iterations, the parallel section runs in series.
Does anyone know what could be happening?
Thank you External Image
[...]
for(int i = 0; i < nsteps; ++i){
#pragma omp parallel num_threads(2)
{
#pragma omp sections
{
#pragma omp section
{
cutilSafeCall( cudaSetDevice(1) );
task1();
// MemcpyToHost --> h_data[1]
}
#pragma omp section
{
cutilSafeCall( cudaSetDevice(0));
task0();
// MemcpyToHost --> h_data[0]
// ThreadSync
}
}
}
[...]
[...]
}
[...]