Re:FastGPT: Faster than PyTorch in 300 lines of Fortran
by jiehong on Wednesday March 15, 2023 @06:40AM
(#35166187)
The article mentions requiring help to use the GPU for compute: how easy is using the GPU for compute in Fortran? (Total newbie in Fortran)
[ Parent
| on HN ]
Re:FastGPT: Faster than PyTorch in 300 lines of Fortran
by Koshkin on Wednesday March 15, 2023 @08:56AM
(#35167586)
Just to chime in here: Intel's Fortran compiler (part of their oneAPI ecosystem) "provides CPU and GPU offload support."
https://www.intel.com/content/www/us/en/developer/tools/onea...
Re:FastGPT: Faster than PyTorch in 300 lines of Fortran
by certik on Wednesday March 15, 2023 @03:08PM
(#35173219)
I don't have much GPU experience myself. As the sibling comment said, there are Fortran compilers that can offload to GPU, there is also Cuda Fortran. There is OpenMP offloading. I think LLVM can also target it somehow, and I would like to support it in LFortran, a compiler that we are developing. In general I am hoping people more experienced with GPU would be interested in helping out.