hi,
I would like to implement the multiGPUs calculations using CUDA Fortran+MPI. Can I use PGI community edition to do this ? Thanks
Neo
hi,
I would like to implement the multiGPUs calculations using CUDA Fortran+MPI. Can I use PGI community edition to do this ? Thanks
Neo
Yes. You create an openmp section of code on the host,
where each thread assigns a GPU and starts a program on it.
Then all the GPUs run simultaneously, instead of sequentially.
dave