ASU-KE / RC

Research Computing public repository for anything related to HPC
2 stars 0 forks source link

Amber #9

Closed NilaBlueshirt closed 2 months ago

NilaBlueshirt commented 5 months ago

Contact email

John.Kazan@asu.edu

ASURITE

ikazan

Software Name

Amber

Software version

22

Notes

Spack only has v20 and has cuda conflicts. The user needs MPI and cuda.

Cluster

Phoenix

NilaBlueshirt commented 4 months ago
#!/bin/bash
# 2024-05-23 by Nil with help from Alan
# the tar balls were copied from Sol amber/22v3
# install guide: https://ambermd.org/doc12/Amber24.pdf
# the clean_build script was edited to leave this one out

# the GCC compiler and cuda-12.4 didn't work
# need cuda<12 and gcc<=11, so the selected cuda module is made by
# cuda@11.8.0%gcc@11.2.0+allow-unsupported-compilers
module load cmake-3.27.7-tt \
            intel-oneapi-compilers-2023.2.1-5l \
            intel-oneapi-mkl-2023.2.0-me \
            intel-oneapi-mpi-2021.10.0-gcc-12.3.0 \
            cuda-11.8.0-gcc-11.2.0-66 \
            mamba/latest

amber_home=/packages/apps/build/amber/amber22_v3/amber22_src

cd $amber_home
cd build

# amber 22 has miniconda on Sol, and it's the default option
# added mkl
./configure_cmake.py  --prefix /packages/apps/amber/22_v3 \
                      --compiler INTEL \
                      --mpi \
                      --cuda \
                      -mkl \
                      -mkl-multithreaded \
                      --no-gui

# after this step, edit the run_cmake script:
#   cmake $AMBER_PREFIX/amber22_src \
#   -DCMAKE_INSTALL_PREFIX=/packages/apps/amber/22_v3 \
#   -DCOMPILER=INTEL  \
#   -DMPI=TRUE -DCUDA=TRUE -DINSTALL_TESTS=TRUE \
#   -DDOWNLOAD_MINICONDA=TRUE \
./run_cmake

make install
NilaBlueshirt commented 4 months ago

source /packages/apps/amber/22_v3/amber.sh

NilaBlueshirt commented 3 months ago

Testing codes: https://github.com/John-Kazan/MDGuide

NilaBlueshirt commented 2 months ago
#!/bin/bash
# step 3 - add the MPI support
module load cmake-3.26.5-gcc-11.2.0-ed \
            gcc-8.5.0-gcc-11.2.0-kn \
            intel-oneapi-mkl-2023.2.0-me \
            openblas-0.3.24-5f \
            openmpi/4.1.5 \
            cuda-11.8.0-gcc-11.2.0-66
# edit the run_cmake script:
#   -DCMAKE_INSTALL_PREFIX=/packages/apps/amber/24 \
#   -DMPI=TRUE -DCUDA=TRUE       
export CUDA_HOME=/packages/apps/spack/21.2/opt/spack/x86_64_v3/gcc-11.2.0/cuda-11.8.0-66eeijo
./run_cmake
make install