Skip to content

CentOS 7

  • Contributed by ZhuoChen Yu @ Jilin University. Thanks!

Installation

Installation in clusters without nvhpc modules

$ wget https://developer.download.nvidia.com/hpc-sdk/24.5/nvhpc_2024_245_Linux_x86_64_cuda_multi.tar.gz
$ tar xpzf nvhpc_2024_245_Linux_x86_64_cuda_multi.tar.gz
$ nvhpc_2024_245_Linux_x86_64_cuda_multi/install
>>>Press enter to continue...
>>>1
>>>${HOME}/nvidia/hpc_sdk
$ export PATH=${HOME}/nvidia/hpc_sdk/Linux_x86_64/24.5/compilers/bin:$PATH

$ cd gRASPA-main/src_clean
$ cp ../NVC_COMPILE ./ && vim NVC_COMPILE
Set NVC_COMPILE as followings: use '--gcc-toolchain=' to set gcc path. 
#!/bin/bash
rm nvc_main.x
module load gcc-11 #Load gcc which version is later than 8.5
CXX="nvc++"

LINKFLAG="-L/usr/lib64"

NVCFLAG="-O3 -std=c++20 -gpu=cc86 -Minline -fopenmp -cuda -stdpar=multicore  --gcc-toolchain=gcc"
$CXX $NVCFLAG $LINKFLAG -c axpy.cu

$CXX $NVCFLAG $LINKFLAG -c main.cpp

$CXX $NVCFLAG $LINKFLAG -c read_data.cpp

$CXX $NVCFLAG $LINKFLAG -c data_struct.cpp

$CXX $NVCFLAG $LINKFLAG -c VDW_Coulomb.cu

$CXX $NVCFLAG $LINKFLAG main.o read_data.o axpy.o data_struct.o VDW_Coulomb.o -o nvc_main.x

$ chmod 777 ./NVC_COMPILE
$ ./NVC_COMPILE

Installation on CentOS 7/RedHat 7

  1. install gcc (eg. installation of gcc-11) 
    $ sudo yum install devtoolset-11-gcc*
$ source /opt/rh/devtoolset-11/enable
  2. install nvhpc
  3. install gRASPA 
    $ cd gRASPA-main/src_clean
$ cp ../NVC_COMPILE ./ && vim NVC_COMPILE

Set NVC_COMPILE as followings: 

#!/bin/bash
rm nvc_main.x
CXX="nvc++"

LINKFLAG="-L/usr/lib64"

NVCFLAG="-O3 -std=c++20 -gpu=cc86 -Minline -fopenmp -cuda -stdpar=multicore  --gcc-toolchain=gcc"
$CXX $NVCFLAG $LINKFLAG -c axpy.cu

$CXX $NVCFLAG $LINKFLAG -c main.cpp

$CXX $NVCFLAG $LINKFLAG -c read_data.cpp

$CXX $NVCFLAG $LINKFLAG -c data_struct.cpp

$CXX $NVCFLAG $LINKFLAG -c VDW_Coulomb.cu

$CXX $NVCFLAG $LINKFLAG main.o read_data.o axpy.o data_struct.o VDW_Coulomb.o -o nvc_main.x


$ chmod 777 ./NVC_COMPILE
$ ./NVC_COMPILE