I have a scientific program written in Fortran77 (about 220 lines without comments) with
library also in Fortran77 (about 5000 lines without comments). I want to
use this program on GPU card with CUDA acceleration. I would prefer to
avoid translation of this source code into C language. My question is how
to do it in the best way, to achieve the best performance? I will be
grateful for your help and guidance.
I have a scientific program written in Fortran77 (about 220 lines without comments) with
library also in Fortran77 (about 5000 lines without comments). I want to
use this program on GPU card with CUDA acceleration. I would prefer to
avoid translation of this source code into C language. My question is how
to do it in the best way, to achieve the best performance? I will be
grateful for your help and guidance.
You probably want to investigate the PGI fortran compiler for CUDA. It will probably be the easiest way to get some Gpu acceleration into a legacy fortran codebase, although the performance may not be quite as good as porting the data parallel parts to C by hand.
You probably want to investigate the PGI fortran compiler for CUDA. It will probably be the easiest way to get some Gpu acceleration into a legacy fortran codebase, although the performance may not be quite as good as porting the data parallel parts to C by hand.