[Opm] about GPU accelerate

[Markus Blatt]

Hi Lei Ting,

On Wed, Jul 21, 2021 at 09:33:19AM +0800, LEI TING wrote:
> Hi, thanks for the great work in OPM.
> I want to know if the OPM team has a plan to support GPU acceleration?

actually we already support using the GPU for solving the linear systems.

just pass the option --accelerator-mode=cuspare for CUDA or --accelerator-mode=opencl
for OpenCL. There are a few more options concerning the GPU just look at the output of
flow --help.

Unfortunately, this is still  a bit experimental. Your mileage may vary depending on your
GPU. One still cannot mix MPI and the GPU and hence if you have a powerful CPU a parallel run
might be preferable.

HTH

MArkus
-- 

Markus Blatt
CTO @ OPM-OP AS, Heyerdahlsvei 12b, 0777 Oslo, Norway
https://opm-op.com | +4916097590858