Install Amber with GPU support

Overview

Teaching: 5 min min
Exercises: 0 min

Questions

• How to install Amber with GPU support to Palmetto

Objectives

AMBER_GPU

1. Request a GPU node
2. Load necessary modules with cuda and anaconda:

$ module load cuda/10.2.89-gcc/8.3.1   anaconda3/2019.10-gcc/8.3.1

1. Modify the run_cmake file:

$ cmake $AMBER_PREFIX/amber20_src \
    -DCMAKE_INSTALL_PREFIX=$AMBER_PREFIX/amber20 \
    -DCOMPILER=GNU  \
    -DMPI=TRUE -DCUDA=TRUE -DINSTALL_TESTS=TRUE \
    -DDOWNLOAD_MINICONDA=FALSE -DMINICONDA_USE_PY3=FALSE \
    2>&1 | tee  cmake.log

Once the CMakeCache file created, modify the CMakeCache to include: -luuid in the following line:

X11_SM_LIB:FILEPATH=/usr/lib64/libSM.so:-luuid

The execute the compiler command:

./run_cmake
make -j16
make install

AMBER MPI NO GPU

1. Request a regular node with mpi processors
2. Load necessary modules with cuda and anaconda:

$ module load anaconda3/2019.10-gcc/8.3.1 cmake/3.17.3-gcc/8.3.1 openmpi/3.1.5-gcc/7.1.0-ucx gcc/7.1.0 parallel-netcdf/1.12.1-gcc/8.3.1-cuda10_2

1. configure and compile:

$ ./configure -mpi gnu
$ make -j16
$ make install

Key Points

• amber, gpu