openmp in CUDA openmp support in the host code

Accelerated Computing CUDA CUDA Programming and Performance
1

hi, everyone!

I want to support the openmp features in the host code. However, I failed.
I got the error “This application has failed to start because vompd.dll was not found.”. How to solve this problem? thanks.
Here is a brief description on what I have done:
(1) OS: winXP. IDE: VS2005.
(2) I have included “omp.h”
(3) I have added “/openmp” in the nvcc compilation.
(4) The code is as followed:
{
#pragma omp parallel for
for( i=0; i<numThread; i++ )
{
//do something.
}
}

2

Have you enabled OpenMP in your project’s configuration ?

3

You are right. We need this configuration.

Originally, the option is listed in the configuration of “C/C++” in the property page.

However, I can not find “C/C++” in the .cu project, because we use custem building. I think we need a manual way to handle this.

4

Use the following flags when compiling a .cu file to use OpenMP:
–host-compilation=C++ -Xcompiler /openmp
That’s what cudaOpenMP SDK sample uses (available in Windows). You’ll have to provide the proper gcc flag after -Xcompiler for Linux (not sure what that is, but gcc 4.2 and later does support OpenMP).

Paulius