软件资源

部分软件资源

GNU
Intel
NVIDIA
AMD
Python
Golang
Open MPI
Tcl/TK
计算软件
VASP6 GPU 编译
Quantum Espresso
NAMD
OOMMF
mumax
支撑软件
Singularity Image
gnuplot
OVITO
Vim
Zsh

GNU

GNU Compiler Collection

Name	Path	Module
GNU Compiler Collection (GCC) 12.1.0 & GNU Binutils 2.38	/fs00/software/gcc/12.1.0	gcc/12.1.0
GNU Compiler Collection (GCC) 11.3.0 & GNU Binutils 2.36.1	/fs00/software/gcc/11.3.0	gcc/11.3.0
GNU Compiler Collection (GCC) 10.5.0 & GNU Binutils 2.34	/fs00/software/gcc/10.5.0	gcc/10.5.0
GNU Compiler Collection (GCC) 9.5.0 & GNU Binutils 2.32	/fs00/software/gcc/9.5.0	gcc/9.5.0
GNU Compiler Collection (GCC) 8.5.0 & GNU Binutils 2.30	/fs00/software/gcc/8.5.0	gcc/8.5.0
GNU Compiler Collection (GCC) 7.5.0 & GNU Binutils 2.28.1	/fs00/software/gcc/7.5.0	gcc/7.5.0
GNU Compiler Collection (GCC) 6.5.0 & GNU Binutils 2.26.1	/fs00/software/gcc/6.5.0	gcc/6.5.0
GNU Compiler Collection (GCC) 5.4.0	/fs00/software/gcc/5.4.0	gcc/5.4.0

GNU Make

Name	Path	Module
GNU Make 4.3	/fs00/software/make/4.3	make/4.3
GNU Make 4.2.1	/fs00/software/make/4.2.1	make/4.2.1
GNU Make 4.2	/fs00/software/make/4.2	make/4.2

GNU Scientific Library

Name	Compiler	Path	Module
GNU Scientific Library (GSL) 2.7.1	GCC 12.1.0	/fs00/software/gsl/2.7.1-gcc12.1.0	gsl/2.7.1-gcc12.1.0
GNU Scientific Library (GSL) 2.5	GCC 8.3.0	/fs00/software/gsl/2.5-gcc8.3	gsl/2.5-gcc8.3

GNU C Library

Name	Compiler	Path	Module
GNU C Library (glibc) 2.36	GCC 12.1.0	/fs00/software/glibc/2.36-gcc12.1.0	glibc/2.36-gcc12.1.0
GNU C Library (glibc) 2.30	GCC 9.2.0	/fs00/software/glibc/2.30-gcc9.2.0	glibc/2.30-gcc9.2.0

GNU Binutils

Name	Compiler	Path	Module
GNU Binutils 2.38	GCC 12.1.0	/fs00/software/binutils/2.38-gcc12.1.0	binutils/2.38-gcc12.1.0
GNU Binutils 2.27	GCC 5.4.0	/fs00/software/binutils/2.27-gcc5.4.0	binutils/2.27-gcc5.4.0

Intel

Intel oneAPI

Name	Path	MODULEPATH
Intel oneAPI Base Toolkit 2024.0.1 Intel HPC Toolkit 2024.0.1	/fs00/software/intel/oneapi2024.0	/fs00/software/modulefiles/oneapi/2024.0

Intel Parallel Studio

Name	Path	Module
Intel Parallel Studio XE 2020 Update 2 Cluster Edition	/fs00/software/intel/ps2020u2	ips/2020u2
Intel Parallel Studio XE 2019 Update 5 Cluster Edition	/fs00/software/intel/ps2019u5	ips/2019u5
Intel Parallel Studio XE 2018 Update 4 Cluster Edition	/fs00/software/intel/ps2018u4	ips/2018u4
Intel Parallel Studio XE 2017 Update 6 Cluster Edition	/fs00/software/intel/ps2017u6	ips/2017u6
Intel Parallel Studio XE 2017 Update 2 Cluster Edition	/fs00/software/intel/ps2017u2	ips/2017u2
Intel Parallel Studio XE 2016 Update 4 Cluster Edition	/fs00/software/intel/ps2016u4	ips/2016u4
Intel Parallel Studio XE 2015 Update 6 Cluster Edition	/fs00/software/intel/ps2015u6	ips/2015u6
Intel Cluster Studio XE 2013 Service Pack 1 (SP1) Update 1	/fs00/software/intel/cs2013sp1u1	ics/2013sp1u1
Intel Cluster Studio XE 2013	/fs00/software/intel/cs2013	ics/2013
Intel Parallel Studio XE 2011 SP1 Update 3	/fs00/software/intel/ps2011sp1u3	ips/2011sp1u3

Intel Distribution for Python

Name	Path
Intel Distribution for Python 2.7 2019 Update 5	/fs00/software/intel/ps2019u5/intelpython2
Intel Distribution for Python 3.6 2019 Update 5	/fs00/software/intel/ps2019u5/intelpython3
Intel Distribution for Python 2.7 2018 Update 3	/fs00/software/intel/python2018u3/intelpython2
Intel Distribution for Python 3.6 2018 Update 3	/fs00/software/intel/python2018u3/intelpython3
Intel Distribution for Python 2.7 2017 Update 3	/fs00/software/intel/python2017u3/intelpython2
Intel Distribution for Python 3.5 2017 Update 3	/fs00/software/intel/python2017u3/intelpython3

NVIDIA

CUDA Toolkit

Name	Path	Module
CUDA Toolkit 12.3.1	/fs00/software/cuda/12.3.1	cuda/12.3.1
CUDA Toolkit 12.0.0	/fs00/software/cuda/12.0.0	cuda/12.0.0
CUDA Toolkit 11.8.0	/fs00/software/cuda/11.8.0	cuda/11.8.0
CUDA Toolkit 11.2.0	/fs00/software/cuda/11.2.0	cuda/11.2.0
CUDA Toolkit 10.2.89	/fs00/software/cuda/10.2.89	cuda/10.2.89
CUDA Toolkit 10.1.243	/fs00/software/cuda/10.1.243	cuda/10.1.243
CUDA Toolkit 10.0.130	/fs00/software/cuda/10.0.130	cuda/10.0.130
CUDA Toolkit 9.2.148	/fs00/software/cuda/9.2.148	cuda/9.2.148
CUDA Toolkit 9.0.176 with Patch 3	/fs00/software/cuda/9.0.176	cuda/9.0.176
CUDA Toolkit 8.0 GA2 8.0.61 with Patch 2	/fs00/software/cuda/8.0.61	cuda/8.0.61

cuDNN

Name	CUDA	Path	Module
cuDNN v8.9.7.29	12.x	/fs00/software/cudnn/8.9.7.29-cuda12	cudnn/8.9.7.29-cuda12
cuDNN v8.9.7.29	11.x	/fs00/software/cudnn/8.9.7.29-cuda11	cudnn/8.9.7.29-cuda11
cuDNN v8.7.0.84	11.x	/fs00/software/cudnn/8.7.0.84-cuda11	cudnn/8.7.0.84-cuda11
cuDNN v8.7.0.84	10.2	/fs00/software/cudnn/8.7.0.84-cuda10	cudnn/8.7.0.84-cuda10
cuDNN v8.1.1.33	11.2	/fs00/software/cudnn/11.2-v8.1.1.33	cudnn/11.2-v8.1.1.33
cuDNN v8.2.2.26	10.2	/fs00/software/cudnn/10.2-v8.2.2.26	cudnn/10.2-v8.2.2.26
cuDNN v7.6.5.32	10.2	/fs00/software/cudnn/10.2-v7.6.5.32	cudnn/10.2-v7.6.5.32
cuDNN v7.6.4.38	10.1	/fs00/software/cudnn/10.1-v7.6.4.38	cudnn/10.1-v7.6.4.38
cuDNN v7.6.5.32	10.0	/fs00/software/cudnn/10.0-v7.6.5.32	cudnn/10.0-v7.6.5.32
cuDNN v7.1.4	9.2	/fs00/software/cudnn/9.2-v7.1.4	cudnn/9.2-v7.1.4
cuDNN v7.1.4	9.0	/fs00/software/cudnn/9.0-v7.1.4	cudnn/9.0-v7.1.4
cuDNN v7.0.5	8.0	/fs00/software/cudnn/8.0-v7.0.5	cudnn/8.0-v7.0.5
cuDNN v6.0	8.0	/fs00/software/cudnn/8.0-v6.0	cudnn/8.0-v6.0
cuDNN v5.1	8.0	/fs00/software/cudnn/8.0-v5.1	cudnn/8.0-v5.1

HPC SDK

Name	Path	MODULEPATH
HPC SDK 23.11	/fs00/software/nvhpc/23.11	/fs00/software/nvhpc/23.11/modulefiles
HPC SDK 22.11	/fs00/software/nvhpc/22.11	/fs00/software/nvhpc/22.11/modulefiles
HPC SDK 21.3	/fs00/software/nvhpc/21.3	/fs00/software/nvhpc/21.3/modulefiles
HPC SDK 20.9	/fs00/software/nvhpc/20.9	/fs00/software/nvhpc/20.9/modulefiles

HPC-X

Name	CUDA	Path	MODULEPATH
HPC-X 2.17.1	12.x	/fs00/software/hpcx/2.17.1-cuda12	/fs00/software/hpcx/2.17.1-cuda12/modulefiles

NCCL

Name	CUDA	Path	Module
NCCL 2.19.3	12.3	/fs00/software/nccl/2.19.3-cuda12.3	nccl/2.19.3-cuda12.3
NCCL 2.16.2	12.0	/fs00/software/nccl/2.16.2-cuda12.0	nccl/2.16.2-cuda12.0
NCCL 2.16.2	11.8	/fs00/software/nccl/2.16.2-cuda11.8	nccl/2.16.2-cuda11.8
NCCL 2.16.2	11.0	/fs00/software/nccl/2.16.2-cuda11.0	nccl/2.16.2-cuda11.0
NCCL v2.5.6	10.2	/fs00/software/nccl/10.2-v2.5.6	nccl/10.2-v2.5.6
NCCL v2.4.8	10.1	/fs00/software/nccl/10.1-v2.4.8	nccl/10.1-v2.4.8

TensorRT

Name	CUDA	cuDNN	Path	Module
TensorRT 8.6.1.6	12.0		/fs00/software/tensorrt/8.6.1.6-cuda12.0	tensorrt/8.6.1.6-cuda12.0
TensorRT 8.6.1.6	11.8		/fs00/software/tensorrt/8.6.1.6-cuda11.8	tensorrt/8.6.1.6-cuda11.8
TensorRT 8.5.2.2	11.8	8.6	/fs00/software/tensorrt/8.5.2.2-cuda11.8-cudnn8.6	tensorrt/8.5.2.2-cuda11.8-cudnn8.6
TensorRT 8.5.2.2	10.2	8.6	/fs00/software/tensorrt/8.5.2.2-cuda10.2-cudnn8.6	tensorrt/8.5.2.2-cuda10.2-cudnn8.6
TensorRT 8.2.0.6	11.4	8.2	/fs00/software/tensorrt/8.2.0.6-cuda11.4-cudnn8.2	tensorrt/8.2.0.6-cuda11.4-cudnn8.2
TensorRT 8.2.0.6	10.2	8.2	/fs00/software/tensorrt/8.2.0.6-cuda11.4-cudnn8.2	tensorrt/8.2.0.6-cuda10.2-cudnn8.2

AMD

AMD Optimizing C/C++ Compiler

Name	Path	Module
AMD Optimizing C/C++ Compiler 2.3.0 (AOCC)	/fs00/software/aocc/2.3.0	aocc/2.3.0
AMD Optimizing C/C++ Compiler 2.1.0 (AOCC)	/fs00/software/aocc/2.1.0	aocc/2.1.0
AMD Optimizing C/C++ Compiler 2.0.0 (AOCC)	/fs00/software/aocc/2.0.0	aocc/2.0.0

AMD Optimizing CPU Libraries

Name	Path	Module
AMD Optimizing CPU Libraries 2.2 (AOCL)	/fs00/software/aocl/2.2	aocl/2.2
AMD Optimizing CPU Libraries 2.0 (AOCL)	/fs00/software/aocl/2.0	aocl/2.0

Python

请使用者自行解决License问题，本中心概不负责！

Anaconda

Name	Path	Module
Anaconda 3 (Python3) Latest	/fs00/software/anaconda/3	anaconda/3
Anaconda 2 (Python2) Latest	/fs00/software/anaconda/2	anaconda/2
Anaconda 5.0.1 (Python 3.6)	/fs00/software/anaconda/3-5.0.1	anaconda/3-5.0.1
Anaconda 5.0.1 (Python 2.7)	/fs00/software/anaconda/2-5.0.1	anaconda/2-5.0.1
Anaconda 3.4.1 (Python 3.6)	/fs00/software/anaconda/3-3.4.1	anaconda/3-3.4.1
Anaconda 3.4.1 (Python 2.7)	/fs00/software/anaconda/2-3.4.1	anaconda/2-3.4.1

Golang

Name	Path	Module
Golang 1.21.6	/fs00/software/golang/1.21.6	golang/1.21.6
Golang 1.19.5	/fs00/software/golang/1.19.5	golang/1.19.5
Golang 1.18.10	/fs00/software/golang/1.18.7	golang/1.18.10
Golang 1.17.13	/fs00/software/golang/1.17.13	golang/1.17.13
Golang 1.16.15	/fs00/software/golang/1.16.15	golang/1.16.15
Golang 1.15.15	/fs00/software/golang/1.15.15	golang/1.15.15

Open MPI

Name	Compiler	Path	Module
Open MPI 4.1.2	GNU Compiler Collection (GCC) 11.2.0	/fs00/software/openmpi/4.1.2-gcc11.2.0	openmpi/4.1.2-gcc11.2.0
Open MPI 3.1.2	GNU Compiler Collection (GCC) 8.2.0	/fs00/software/openmpi/3.1.2-gcc8.2.0	openmpi/3.1.2-gcc8.2.0
Open MPI 1.10.0	Intel C++ Compiler XE 15.0 Update 3 & Fortran Compiler XE 15.0 Update 3	/fs00/software/openmpi/1.10.0-iccifort-15.0.3	openmpi/1.10.0-iccifort-15.0.3
Open MPI 1.10.0	GNU Compiler Collection (GCC) 5.2.0	/fs00/software/openmpi/1.10.0-gcc-5.2.0	openmpi/1.10.0-gcc-5.2.0
Open MPI 1.10.5	GNU Compiler Collection (GCC) 5.4.0	/fs00/software/openmpi/1.10.5-gcc5.4.0	openmpi/1.10.5-gcc5.4.0

Tcl/TK

Tcl/Tk

Name	Compiler	Path	Module
Tcl/Tk 8.6.12	GNU Compiler Collection (GCC) 11.2.0	/fs00/software/tcl/8.6.12-gcc11.2.0	tcl/8.6.12-gcc11.2.0
Tcl/Tk 8.6.6	Intel Parallel Studio XE 2017 Update 2	/fs00/software/tcl/8.6.6-ips2017u2	tcl/8.6.6-ips2017u2
Tcl/Tk 8.6.4		/fs00/software/tcl/8.6.4	tcl/8.6.4
Tcl/Tk 8.6.4	Intel Parallel Studio XE 2016 Update 2	/fs00/software/tcl/8.6.4-ips2016u2	tcl/8.6.4-ips2016u2
Tcl/Tk 8.6.4	Intel Parallel Studio XE 2016 Update 2	/fs00/software/tcl/8.6.4-ips2016u2-avx2	tcl/8.6.4-ips2016u2-avx2

计算软件

cd请使用者自行解决License问题，本中心概不负责！

FFTW 3.3.7

/fs00/software/fftw/3.3.7-iccifort-17.0.6-* 依赖iccifort/17.0.6

FFTW 3.3.8

/fs00/software/fftw/3.3.8-ips2019u5 依赖ips/2019u5

LAMMPS 11Aug17

/fs00/software/lammps/11Aug17

OpenFOAM® v1806

/fs00/software/openfoam/v1806-ips2017u6

source /fs00/software/openfoam/v1806-ips2017u6/OpenFOAM-v1806/etc/bashrc

P4vasp 0.3.29

/fs00/software/p4vasp/0.3.29

Modulefile: p4vasp/0.3.29

Phonopy 1.11.2

/fs00/software/phonopy/1.11.2

Quantum ESPRESSO 5.2.0 & 6.1

/fs00/software/qe/5.2.0-ips2015u3/ 依赖ips/2015u3

/fs00/software/qe/6.1-ips2017u2/ 依赖ips/2017u2

ShengBTE

/fs00/software/shengbte 依赖 iccifort/15.0.3 openmpi/1.10.0-iccifort-15.0.3

Siesta 3.2-pl-5

/fs00/software/siesta/3.2-pl-5 依赖ips/2017u6

thirdorder 1.0.2 04d3f46feb78

/fs00/software/thirdorder/1.0.2

Modulefile: thirdorder/1.0.2 依赖 anaconda/2-4.3.1 spglib/1.9.9

TBPLaS

/fs00/software/tbplas

Modulefile:

oneapi/2024.0/compiler/2024.0.2
oneapi/2024.0/ifort/2024.0.2
oneapi/2024.0/mkl/2024.0

VASP6 GPU 编译

VASP6 GPU（N卡）编译实例

王勇 (孙建课题组)

人工微结构科学与技术协同创新中心高性能计算中心

根据vasp官方给出的信息，以后vasp gpu版本会着重开发openacc标准的版本，之前老cuda版本会逐渐被废弃，因此我们以vasp6.2 的openacc版本结合官方指导进行编译。官方指导链接

https://www.vasp.at/wiki/index.php/OpenACC_GPU_port_of_VASP

编译器：

Openacc gpu Vasp6官方给出编译器意见为使用NVIDIA HPC-SDK或者 PGI's Compilers & Tools (version >=19.10)。在此vasp官方建议使用NVIDIA HPC-SDK且版本号最好为20.9，因为之后的版本可能会有一些对于vasp的bug。

关于NVIDIA HPC-SDK 20.9的安装，见

https://developer.nvidia.com/nvidia-hpc-sdk-209-downloads

相关主页，安装流程十分简单，使用wget下载，一键式安装，对于不连接外网的节点，可以手动下载tarball压缩包，上传后本地解压，在此不再赘述，安装前请使用nvidia-smi命令查看本地硬件驱动和兼容的cuda版本（version >=10.0），确保匹配，必要时候需升级硬件驱动。

安装HPC-SDK 20.9的过程中会询问安装路径，我们以路径为/usr/software/nv-hpcsdk为例，在安装过程中指定此路径为例以后，需要设定环境变量：

export NVARCH=`uname -s`_`uname -m`;
export NVCOMPILERS=/usr/software/nv-hpcsdk #修改此处为安装路径
export PATH=$NVCOMPILERS/$NVARCH/20.9/compilers/bin:$PATH
export MANPATH=$MANPATH:$NVCOMPILERS/$NVARCH/20.9/compilers/man
export LD_LIBRARY_PATH=$NVCOMPILERS/$NVARCH/20.9/compilers/lib/:$LD_LIBRARY_PATH
export PATH=$NVCOMPILERS/$NVARCH/20.9/comm_libs/mpi/bin/:$PATH

上述可以每次使用gpu vasp6时手动添加在任务脚本，也可直接写在bashrc，但是不建议直接写在bashrc，可能会和intel版本的 mpirun冲突，造成其他之前软件的运行问题。

依赖库：

安装完NVIDIA HPC-SDK 20.9以后，还需要将软件运行需要的依赖库整理好，主要有CUDA Toolkit, QD, NCCL, 以及FFTW, 前三项直接包含在HPC-SDK，不需要单独安装。

对于FFTW，最好不要用nvhpc-sdk的编译器进行安装，如果设置了上面安装完hpc-sdk的环境变量，请先用GNU或者intel的编译器环境变量进行覆盖，不然可能会导致计算效率问题，可以自己编译，也可以直接使用集群内已经安装好的版本，路径为/fs00/software/fftw/3.3.8-ips2019u5

编译：

准备好编译器和依赖库以后，就可以进行编译了，进入vasp6.2根目录，

cp arch/makefile.include.linux_nv_acc makefile.include`

可以使用

which nvfortran | awk -F /compilers/bin/nvfortran '{ print $$1 }'`

查看nvfortran是否为nv-hpc-sdk的路径，如有问题，可以重新添加一遍环境变量。

（注：最近vasp官方又在官网添加了新的openacc+openmp混编版本的makefile.include.linux_nv_acc+omp+mkl来解决nccl库限制openacc版本只能单进程运行的问题，通过openmp来提升单进程多线程的运算效率，但目前没有太多相关测试的数据来支撑openmp混编会提升很多并行速度，所以在此我仍以旧版本的makefile.include.linux_nv_acc为例，同时官方也在积极改进这个问题，后续应该也可以多进程运行）

makefile.include的内容需要修改几处，包括确认编译器的位置，明确依赖库的路径等。修改后的makefile.include如下（需要注意和修改的位置后面有注释）：

#Precompiler options

CPP_OPTIONS= -DHOST=\"LinuxPGI\" \
         -DMPI -DMPI_BLOCK=8000 -DMPI_INPLACE -Duse_collective \
         -DscaLAPACK \
         -DCACHE_SIZE=4000 \
         -Davoidalloc \
         -Dvasp6 \
         -Duse_bse_te \
         -Dtbdyn \
         -Dqd_emulate \
         -Dfock_dblbuf \
         -D_OPENACC \
         -DUSENCCL -DUSENCCLP2P

CPP        = nvfortran -Mpreprocess -Mfree -Mextend -E $(CPP_OPTIONS) $*$(FUFFIX)  > $*$(SUFFIX)

FC         = mpif90 -acc -gpu=cc60,cc70,cc80,cuda11.0
FCL        = mpif90 -acc -gpu=cc60,cc70,cc80,cuda11.0 -c++libs

FREE       = -Mfree

FFLAGS     = -Mbackslash -Mlarge_arrays

OFLAG      = -fast

DEBUG      = -Mfree -O0 -traceback

#Specify your NV HPC-SDK installation, try to set NVROOT automatically
NVROOT     =$(shell which nvfortran | awk -F /compilers/bin/nvfortran '{ print $$1 }')
#or set NVROOT manually
#NVHPC      ?= /opt/nvidia/hpc_sdk
#NVVERSION  = 20.9
#NVROOT     = $(NVHPC)/Linux_x86_64/$(NVVERSION)

#Use NV HPC-SDK provided BLAS and LAPACK libraries
BLAS       = -lblas
LAPACK     = -llapack

BLACS      =
SCALAPACK  = -Mscalapack

CUDA       = -cudalib=cublas,cusolver,cufft,nccl -cuda

LLIBS      = $(SCALAPACK) $(LAPACK) $(BLAS) $(CUDA)

#Software emulation of quadruple precsion
QD         = $(NVROOT)/compilers/extras/qd                    #注意并修改这里
LLIBS      += -L$(QD)/lib -lqdmod -lqd
INCS       += -I$(QD)/include/qd

#Use the FFTs from fftw
FFTW       = /fs00/software/fftw/3.3.8-ips2019u5              #修改fftw路径至本地安装路径
LLIBS      += -L$(FFTW)/lib -lfftw3
INCS       += -I$(FFTW)/include

OBJECTS    = fftmpiw.o fftmpi_map.o fftw3d.o fft3dlib.o

#Redefine the standard list of O1 and O2 objects
SOURCE_O1  := pade_fit.o
SOURCE_O2  := pead.o

#For what used to be vasp.5.lib
CPP_LIB    = $(CPP)
FC_LIB     = nvfortran
CC_LIB     = nvc
CFLAGS_LIB = -O
FFLAGS_LIB = -O1 -Mfixed
FREE_LIB   = $(FREE)

OBJECTS_LIB= linpack_double.o getshmem.o

#For the parser library
CXX_PARS   = nvc++ --no_warnings

#Normally no need to change this
SRCDIR     = ../../src
BINDIR     = ../../bin

此外编译前和每次提交任务前请清除其他环境并重新设置一遍环境变量：

module load ips/2019u5  #这是为了我们使用集群安装好的fftw找到依赖的路径
export NVARCH=`uname -s`_`uname -m`;
export NVCOMPILERS=/usr/software/nv-hpcsdk
export PATH=$NVCOMPILERS/$NVARCH/20.9/compilers/bin:$PATH
export MANPATH=$MANPATH:$NVCOMPILERS/$NVARCH/20.9/compilers/man
export LD_LIBRARY_PATH=$NVCOMPILERS/$NVARCH/20.9/compilers/lib/:$LD_LIBRARY_PATH
export PATH=$NVCOMPILERS/$NVARCH/20.9/comm_libs/mpi/bin/:$PATH

/usr/software/nv-hpcsdk为nv-hpc-sdk的安装路径确认好以上所有设置后，就可以使用

make std gam ncl

进行编译了，注意，由于openacc版本的原因，直接取消了编译make gpu的方式，编译得到的vasp_std之类的版本直接可以使用GPU进行计算。

其他注意事项：

1.由于nccl库的限制，openacc的gpu版本只能单进程运行.

2.INCAR中NCORE在openacc版本中只能设为1.

3.INCAR中NSIM和KPAR参数需要对不同体系进行测试来达到效率最大化，一般来说，KPAR和所使用的GPU也就是进程数一致，NSIM需要设置的比在cpu上更大，具体请自行测试.

更多相关问题可以见GPU官方指导

也可以去VASP论坛提问

Quantum Espresso

直接使用Singularity

例：使用qe7.1版本，输入文件为当前目录下的FeO_test.in，申请1个GPU提交到723090ib队列，使用pw.x

#BSUB -J FeO_test
#BSUB -q 723090ib
#BSUB -gpu num=1

module load singularity/latest

export OMP_NUM_THREADS="$LSB_DJOB_NUMPROC"
SINGULARITY="singularity run --nv /fs00/software/singularity-images/ngc_quantum_espresso_qe-7.1.sif"
${SINGULARITY} pw.x < FeO_test.in > FeO_test.out

NAMD

NAMD 2.12 (2016-12-22)

/fs00/software/namd/2.12

例：输入文件为当前目录下的in.conf，申请48个核提交到e5v3ib队列

#BSUB -n 48
#BSUB -q e5v3ib

input=in.conf
#bindir=/fs00/software/namd/2.12/verbs/
bindir=/fs00/software/namd/2.12/ibverbs/

nodefile=nodelist
echo "group main" > $nodefile
for i in `echo $LSB_HOSTS`
do
	echo "host $i" >> $nodefile
done

${bindir}charmrun ++remote-shell ssh ++nodelist $nodefile +p$LSB_DJOB_NUMPROC ${bindir}namd2 $input

OOMMF

The Object Oriented MicroMagnetic Framework (OOMMF) 1.2 alpha 6

/fs00/software/oommf/12a6-tcl8.6.4-ips2016u2 依赖tcl/8.6.4-ips2016u2和ips/2016u2
/fs00/software/oommf/12a6-tcl8.6.4-ips2016u2-avx2 依赖tcl/8.6.4-ips2016u2-avx2和ips/2016u2

例：输入文件为当前目录下的sample.mif，申请72个核提交到6140ib队列

#BSUB -q 6140ib
#BSUB -n 72

module load ips/2016u2
module load tcl/8.6.4-ips2016u2-avx2

oommfin=sample.mif

oommfrun=/fs00/software/oommf/12a6-tcl8.6.4-ips2016u2-avx2/oommf.tcl

OOMMF_HOSTPORT=`tclsh $oommfrun launchhost 0`
export OOMMF_HOSTPORT
tclsh $oommfrun mmArchive
tclsh $oommfrun boxsi -numanodes auto -threads $LSB_DJOB_NUMPROC $oommfin
tclsh $oommfrun killoommf all

mumax

mumax 3.10

/fs00/software/mumax/3.10-cuda11.0/

例：输入文件为当前目录下的sample.mx3，申请1个GPU提交到723090ib队列

#BSUB -q 723090ib
#BSUB -gpu num=1

mx3in=sample.mx3

module load cuda/11.2.0
/fs00/software/mumax/3.10-cuda11.0/mumax3 $mx3in

支撑软件

请使用者自行解决License问题，本中心概不负责！

AWS CLI v2

Name	Path	Module
AWS CLI current	/fs00/software/aws-cli/v2/current	aws-cli/current
AWS CLI 2.9.6	/fs00/software/aws-cli/v2/2.9.6	aws-cli/2.9.6

bbcp

Name	Path	Module
bbcp 14.04.14.00.1	/fs00/software/bbcp/14.04.14.00.1	bbcp/14.04.14.00.1

Boost

Name	Path	Module
Boost 1.72.0	/fs00/software/boost/1.72.0	boost/1.72.0
Boost 1.58.0	/fs00/software/boost/1.58.0	boost/1.58.0

CMake

Name	Path	Module
CMake 3.23.2	/fs00/software/cmake/3.23.2/	cmake/3.23.2
CMake 3.16.3	/fs00/software/cmake/3.16.3/	cmake/3.16.3
CMake 3.11.4	/fs00/software/cmake/3.11.4/	cmake/3.11.4

Git

Name	Path	Module
Git 2.38.1	/fs00/software/git/2.38.1	git/2.38.1

Grace

Name	Path	Module
Grace 5.1.25	/fs00/software/grace/5.1.25	grace/5.1.25

HDF5

Name	Path	Module
HDF5 1.10.5	/fs00/software/hdf5/1.10.5	hdf5/1.10.5

libpng

Name	Path	Module
libpng 1.5.26	/fs00/software/libpng/1.5.26	libpng/1.5.26

jq

Name	Path	Module
jq 1.7	/fs00/software/jq/1.7	jq/1.7

Libxc

Name	Compiler	Path	Module
Libxc 5.2.2	GNU Compiler Collection (GCC) 11.2.0	/fs00/software/libxc/5.2.2	libxc/5.2.2

libzip

Name	Path	Module
libzip 1.6.1	/fs00/software/libzip/1.6.1	libzip/1.6.1

NetCDF-C

Name	Path	Module
NetCDF-C 4.7.0	/fs00/software/netcdf/c-4.7.0	netcdf/c-4.7.0

PCRE

Name	Path	Module
PCRE 8.39	/fs00/software/pcre/8.39	pcre/8.39

Qt

Name	Path	Module
Qt 5.11.1	/fs00/software/qt/5.11.1	qt/5.11.1

Spglib (OpenMP)

Name	Compiler	Path	Module
Spglib 1.9.9		/fs00/software/spglib/1.9.9	spglib/1.9.9
Spglib 1.9.0	GNU Compiler Collection (GCC) 5.2.0	/fs00/software/spglib/1.9.0-gcc5.2.0	spglib/1.9.0-gcc5.2.0

tmux

Name	Path	Module
tmux 3.3a	/fs00/software/tmux/3.3a	tmux/3.3a

zlib

Name	Path	Module
zlib 1.2.11	/fs00/software/zlib/1.2.11	zlib/1.2.11

Singularity Image

/fs00/software/singularity-images/

有丰富的官方容器镜像包

gnuplot

Name	Path	Module
gnuplot 5.2.7	/fs00/software/gnuplot/5.2.7	gnuplot/5.2.7
gnuplot 5.2.2	/fs00/software/gnuplot/5.2.2	gnuplot/5.2.2
gnuplot 5.0.6	/fs00/software/gnuplot/5.0.6	gnuplot/5.0.6
gnuplot 5.0.1	/fs00/software/gnuplot/5.0.1	gnuplot/5.0.1

OVITO

Name	Path	Module
OVITO 3.7.12	/fs00/software/ovito/3.7.12	ovito/3.7.12
OVITO 2.9.0	/fs00/software/ovito/2.9.0	ovito/2.9.0

Vim

Name	Path	Module
Vim 9.0.1677	/fs00/software/vim/9.0.1677	vim/9.0.1677
Vim 8.2.0488	/fs00/software/vim/8.2.0488	vim/8.2.0488
Vim 8.1	/fs00/software/vim/8.1	vim/8.1

Zsh

Name	Path	Module
Zsh latest	/fs00/software/zsh/latest	zsh/latest
Zsh 5.8	/fs00/software/zsh/5.8	zsh/5.8

或从镜像站下载编译安装

Environment Modules

添加 Environment Modules 至 Zsh

echo "source /fs00/software/modules/latest/init/profile.sh" >> ~/.zshrc

命令行提示符

因登录时需判断命令行提示符是否为$ ，因此建议保持 Bash 为默认 Shell，登录后通过下列命令切换至 Zsh

module load zsh/latest && exec zsh

如设置 Zsh 为默认 Shell 或登录会自动切换至 Zsh，则必须更改 Zsh 的命令行提示符（环境变量PS1）以$ 结尾（$+空格），否则登录会失败！

Oh My Zsh

安装

在本地克隆后获取安装脚本再进行安装

git clone https://mirror.nju.edu.cn/git/ohmyzsh.git
cd ohmyzsh/tools
REMOTE=https://mirror.nju.edu.cn/git/ohmyzsh.git sh install.sh

切换已有 Oh My Zsh 至镜像源

git -C $ZSH remote set-url origin https://mirror.nju.edu.cn/git/ohmyzsh.git
git -C $ZSH pull

升级

omz update

主题

Powerlevel10k

安装

git clone --depth=1 https://mirror.nju.edu.cn/git/powerlevel10k.git ${ZSH_CUSTOM:-$HOME/.oh-my-zsh/custom}/themes/powerlevel10k

编辑 ~/.zshrc 的主题 ZSH_THEME="powerlevel10k/powerlevel10k"

升级

cd ${ZSH_CUSTOM:-$HOME/.oh-my-zsh/custom}/themes/powerlevel10k && git pull

256色终端

开启 xterm 终端的256色以便正确显示提示颜色

echo "export TERM=xterm-256color" >> ~/.zshrc

永久在左侧显示主机名

在集群中经常需要登录到各个节点，因此最好在左侧永久显示主机名防止误操作编辑 ~/.p10k.zsh

将POWERLEVEL9K_RIGHT_PROMPT_ELEMENTS中的context剪切到POWERLEVEL9K_LEFT_PROMPT_ELEMENTS的第一项
注释typeset -g POWERLEVEL9K_CONTEXT_{DEFAULT,SUDO}_{CONTENT,VISUAL_IDENTIFIER}_EXPANSION=这一行
更改typeset -g POWERLEVEL9K_CONTEXT_PREFIX=为''

插件

zsh-autosuggestions

安装

git clone https://mirror.nju.edu.cn/git/zsh-autosuggestions ${ZSH_CUSTOM:-~/.oh-my-zsh/custom}/plugins/zsh-autosuggestions

编辑 ~/.zshrc 在 plugins= 中添加 zsh-syntax-highlighting

plugins=( ... zsh-autosuggestions)

升级

cd ${ZSH_CUSTOM:-~/.oh-my-zsh/custom}/plugins/zsh-autosuggestions && git pull

zsh-syntax-highlighting

安装

git clone https://mirror.nju.edu.cn/git/zsh-syntax-highlighting.git ${ZSH_CUSTOM:-~/.oh-my-zsh/custom}/plugins/zsh-syntax-highlighting

编辑 ~/.zshrc 在 plugins= 最后添加 zsh-syntax-highlighting

plugins=( [plugins...] zsh-syntax-highlighting)

升级

cd ${ZSH_CUSTOM:-~/.oh-my-zsh/custom}/plugins/zsh-syntax-highlighting && git pull

软件资源

GNU

GNU Compiler Collection

GNU Make

GNU Scientific Library

GNU C Library

GNU Binutils

Intel

Intel oneAPI

Intel Parallel Studio

Intel Distribution for Python

NVIDIA

CUDA Toolkit

cuDNN

HPC SDK

HPC-X

NCCL

TensorRT

AMD

AMD Optimizing C/C++ Compiler

AMD Optimizing CPU Libraries

Python

Anaconda

Golang

Golang

Open MPI

Open MPI

Tcl/TK

Tcl/Tk

计算软件

FFTW 3.3.7

FFTW 3.3.8

LAMMPS 11Aug17

OpenFOAM® v1806

P4vasp 0.3.29

Phonopy 1.11.2

Quantum ESPRESSO 5.2.0 & 6.1

ShengBTE

Siesta 3.2-pl-5

thirdorder 1.0.2 04d3f46feb78

TBPLaS

VASP6 GPU 编译

VASP6 GPU（N卡） 编译实例

编译器：

依赖库：

编译：

其他注意事项：

Quantum Espresso

直接使用Singularity

NAMD

NAMD 2.12 (2016-12-22)

OOMMF

The Object Oriented MicroMagnetic Framework (OOMMF) 1.2 alpha 6

mumax

mumax 3.10

支撑软件

AWS CLI v2

bbcp

Boost

CMake

Git

Grace

HDF5

libpng

jq

Libxc

libzip

NetCDF-C

PCRE

Qt

Spglib (OpenMP)

tmux

zlib

Singularity Image

gnuplot

gnuplot

OVITO

OVITO

Vim

Vim

VASP6 GPU（N卡）编译实例