Can an MPI task be moved from a CPU to a GPU in oneAPI Fortran

Intel® Moderncode for Parallel Architectures

Support for developing parallel programming applications on Intel® Architecture.

Can an MPI task be moved from a CPU to a GPU in oneAPI Fortran

1,778 Views

Hi,

I am presently running Fortran with MPI under WIN10 to calculate typically 5000 matrix elements using 60 threads during each iteration step. I can't use OpenMP! Will it be possible, in the new version of Fortran, to move this calculation into a GPU with, lets say, 5000 cores. This should mean that each core could calculate one matrix element. As I understand it this should require that each core in the GPU has a reserved memory area or am I wrong? The amount of data allocated to each core for the calculations is rather small. Will the memory data transfer to the GPU and/or the MPI overhead to collect the calculated matrix elements kill this use of the GPU?

Best regards

Anders S

Link Copied

0 Replies

Community support is provided Monday to Friday. Other contact methods are available here.

Intel does not verify all solutions, including but not limited to any file transfers that may appear in this community. Accordingly, Intel disclaims all express and implied warranties, including without limitation, the implied warranties of merchantability, fitness for a particular purpose, and non-infringement, as well as any warranty arising from course of performance, course of dealing, or usage in trade.

For more complete information about compiler optimizations, see our Optimization Notice.