Hi Filip,
That change should be all you need. That will compile for SM80 which appears to work fine on RTX3080 and 3090 and passes all the tests. There is also an SM86 option which I have not had a chance to test the difference against SM80 yet. I suspect it would be minimal difference though without corresponding optimization within the AMBER code itself.
All the best
Ross
> On Oct 6, 2020, at 08:26, David A Case <david.case.rutgers.edu> wrote:
>
> On Tue, Oct 06, 2020, Filip Fratev wrote:
>>
>> Can you also help me with an advice how to compile with CUDA 11.1? I can
>> do that only by Cuda 11.0.
>
> Try this: on line 67 of cmake/CudaConfig.cmake, change
>
> elseif(${CUDA_VERSION} VERSION_EQUAL 11.0)
>
> to
>
> elseif(${CUDA_VERSION} VERSION_GREATER_EQUAL 11.0)
>
> We're not 100% sure that this will work, but it's worth a try. Ross is
> running some tests of some other possible changes.
>
> ....dac
>
>
> _______________________________________________
> AMBER mailing list
> AMBER.ambermd.org
> http://lists.ambermd.org/mailman/listinfo/amber
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Tue Oct 06 2020 - 11:00:02 PDT
