How to use cudaDeviceSetFlags in an OpenACC program?

Accelerated Computing HPC Compilers Legacy PGI Compilers
#1

Hi,

I was curious if/how I could call “cudaDeviceSetFlags” in an OpenACC program (to set the cudaDeviceScheduleBlockingSync, cudaDeviceMapHost, etc. flags for instance).

Thanks.

#2

Hi stw,

Sure, you can do that.

In Fortran, just include “use cudafor” and compile with “-Mcuda”. See: https://www.pgroup.com/resources/docs/19.10/x86/cuda-fortran-prog-guide/index.htm#dm-cudasetdeviceflags

With C/C++, include “cuda_runtime.h” and compile with “-Mcuda”.

Hope this helps,
Mat

#3

Thank you very much.