I want to transfer part of my Intel Fortran code to an Nvidia GTX3070 GPU to speed up computations. What are my options? I have not yet moved to OneAPI and I am running Windows for graphic output reasons. Use of MPI is also compulsory.
Intel oneAPI supports Intel accelerators. There is no support for Nvidia GPUs.
There are other vendors taking oneAPI DPC++ to Nvidia GPUs, but I'm not aware of anything like that for Fortran.
Thanks for your informative answer! Will the new ALchemist GPUs support DP FP calculations or just SP FP calculations?
I really don't know the answer to that. Did you google for information? Seems like there's a lot of speculation about the product.
Are you aware of Intel Innovation? It's an Intel developer's conference being held October 27-28. It's virtual and free. The time zone may be tough for you, but I expect replays will be available.