Thanks for the explanation Dave,
I understand that the intel seems to be a pain to keep up with ... If
there are no expected gains, I'll probably also give up on it as I have
now a functional compilation. Nevertheless, if I do try it the nvcc flag
and it works I'll keep it posted.
Best
Vlad
On 6/5/23 13:58, David A Case wrote:
> On Mon, Jun 05, 2023, Vlad Cojocaru via AMBER wrote:
>>
>> I am trying to compile Amber 22 (AT 23) on a remote HPC machine using
>> the Intel compiler (21.7.1 or 2022). intel MPI 2021.7 and CUDA 11.6
>> but I am getting the error below. The same compilation works fine
>> with GCC 11.2 and openmpi 4.1.
>
> There is not much to be gained by doing this, so developers spend
> little time
> worrying about it. With CUDA (especially pmemd.cuda), almost all of the
> execution is done on the GPU, so there is essentially no benefit to
> compiling the CPU part with some other compiler.
>
> You could create a shell script in your PATH that calls nvcc with the
> '-allow-unsupported-compiler' flag. Or, edit cmake/CudaConfig.cmake
> to add
> this flag to the CUDA_NVCC_FLAGS variable. I haven't tried either of
> these
> myself.
>
> [Aside: even for CPU compilations, the advantages of using an Intel
> compiler
> really only show up in specialized calculations. Each update from Intel
> seems to add some new incompatibility, so keeping up with that sort of
> thing
> is a growing job.]
>
> ...good luck...dac
>
--
Dr. Vlad Cojocaru (PD, Habil.)
----------------------------------------------
Guest Scientist / Fmr. Project Group Leader
Max Planck Institute for Molecular Biomedicine
Röntgenstrasse 20, 48149 Münster, Germany
----------------------------------------------
Email: vlad.cojocaru.mpi-muenster.mpg.de
http://www.mpi-muenster.mpg.de/43241/cojocaru
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Mon Jun 05 2023 - 07:30:02 PDT
