On Mon, Aug 11, 2025, Cherry, Kendall via AMBER wrote:
>
>I am trying to install Amber on one of our lab computers now, rather than
>using it as a module as I have thus far. I am running into some CUDA/GNU
>compatibility issues when running ./run_cmake for pmemd24. I have followed
>the instructions in the manual exactly and have been troubleshooting how
>to downgrade the version of GCC I am using (at least temporarily, or in a
>conda environment) so that I can get this working.
It's generally easier (at least for me!) to upgrade the Nvidia SDK than to
downgrade compilers. You can have several versions of CUDA (by default
installed into /usr/local, but other locations can be used.) Then just make
a symlink (e.g. ln -s /usr/local/cuda/cuda-11.8 /usr/local/cuda) to choose
the one you want.
>-- The C compiler identification is GNU 11.4.0
>-- Found CUDA: /usr/local/cuda-11.4 (found version "11.4")
>-- CUDA version 11.4 detected
>--
>************************************************************
>Error: Incompatible CUDA and GNU versions
> 11.4.0
>
>See https://stackoverflow.com/questions/6622454/cuda-incompatible-with-my-gcc-version
The linke above shows you which compilers and CUDA versions are compatible.
Downloading CUDA 11.8 (or later) should work.
...good luck...dac
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Tue Aug 12 2025 - 10:30:01 PDT
