Hello,
I would like to use Amber22 on H100 cards, which have CUDA 12.2. I managed to compile it by adding this:
set(SM90FLAGS -gencode arch=compute_90,code=sm_90)
to the CudaConfig.cmake, and changing the accepted cuda version from >12.2 to >12.3. Everything looks ok in the tests, for what I can tell.
However, I had to disable QUICK. Do you know when CUDA 12.2 will be fully supported and tested?
Best,
Elisa Pieri
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Fri Sep 22 2023 - 09:30:02 PDT
