Maybe it would be nice of me to say that (I think) I resolved the issue :)
sudo lshw -C display #discovered no Nvidia drivers
apt search nvidia-driver #looked up for linux supported nvidia drivers
sudo apt install nvidia-driver-450 #installed the latest one
sudo reboot #reboot for final impementation
nvidia-smi #drivers are now ready and shown, GPU's communicate with CUDA
And now pmemd.cuda works.
If anybody will have some precautionary advice, I would be glad to
listen to you.
If not, then I'll just try to see if I can somehow benchmark the
installation and make sure it's all fine and smooth.
Cheers to you all :)
Luka Bilić
Citiram Luka Bilić <Luka.Bilic.irb.hr>:
> Dear all,
>
> I've tried to install Amber20 using Cmake.
> Unfortunately pmemd.cuda doesn't recognize GPU's.
>
> The operating system is Ubuntu 20
> GPU's are 3x Nvidia 2080Ti
>
>
> Before Cmake installation I did these:
> sudo apt-get update
> sudo apt-get upgrade
> sudo apt-get install build-essential
> sudo snap install cmake --classic
> #this installed cmake version 3.18.4
>
> sudo apt-get install -y bison
> sudo apt-get install -y flex
> sudo apt-get install intel-mkl-full
> sudo apt-get install libnetcdf-dev libnetcdff-dev
> sudo apt-get install libbz2-dev
> sudo apt install libboost-all-dev
>
> sudo apt install nvidia-cuda-toolkit
> #this installed cuda 10.1
>
> By engaging ./run_cmake I found issues regarding gcc-9 and g++-9
> This I "fixed" by:
>
> sudo update-alternatives --config gcc #gcc8 chosen
> sudo update-alternatives --config g++ #g++8 chosen
>
> However I left out gfortran-9 (that I did not change to lower version)
>
> My run_cmake was this:
>
> # Assume this is Linux:
>
> cmake $AMBER_PREFIX/amber20_src \
> -DCMAKE_INSTALL_PREFIX=$AMBER_PREFIX/amber20 \
> -DCOMPILER=GNU \
> -DMPI=TRUE -DCUDA=TRUE -DINSTALL_TESTS=TRUE \
> -DDOWNLOAD_MINICONDA=TRUE -DMINICONDA_USE_PY3=TRUE \
> 2>&1 | tee cmake.log
>
> This results in installation that doesn't seem to have anything wrong
> (on the first glance) There were some fortran warnings though.
> However, running pmemd.cuda results with this:
>
> ~/AMBER/amber20/bin/pmemd.cuda -O -i mdtest.in -o mdtest.out -p
> diaC-wat_rHPA.parm7 -c diaC-wat_rHPA.crd -r diaC-wat_rHPA-CUDA.crd -x
> diaC-wat_rHPA-CUDA.mdcrd -inf mdinfoCUDA
>
> cudaGetDeviceCount failed no CUDA-capable device is detected
>
> My guess is that I might need new nvidia drivers?
> However, with the detailed setup given some of you with more
> experience could provide me with better solutions on how to install
> amber20.
>
> I would like MPI, CUDA (and CUDA MPI) installation so as to be able to
> use maximum of these 3x 2080Ti and CPU.
>
> Thank you in advance.
> Yours Luka Bilić
>
>
>
>
> _______________________________________________
> 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 Fri Oct 30 2020 - 04:00:02 PDT
