安装编译环境:
Cluster简况:xeon 5420, 12G ram, 千兆交换机。
CentOS5.2+mkl 10.1.0.015+openmpi 1.3.2+ifort 11.0.074
注:1, vasp作者在makefile里说用openmpi比用mpich2要更快。
2, 按网上的说法, 对于vasp4.6, 采用intel的ifc编译器,BLAS采用Kazushige Goto's BLAS,LAPACK采用lapack_double.o 的编译得到vasp计算速度最快。vasp作者在makefile里也说用Goto's BLAS更快。
3, 按intel官网上 后面的评论,用sequential mkl libraries (mkl_sequential)比用 multithreaded libraries (mkl_intel_thread)编译的要快2-10倍,但是我觉得他可能编译的是串行版本,并行的应该是multithreaded libraries更快些,但是因为我不用串行版本,所以没有比较。
编译得到intel的libfftw3xf_intel.a (不知道有没有人对比过不同fft对速度的影响)
cd /opt/intel/mkl/10.1.0.015/interfaces/fftw3xf
make libem64t compiler=intel
=====================================================================
1: 编译openmpi
cd /home/msemsi/share/openmpi-1.3.2
./configure CC=icc CXX=icpc F77=ifort FC=ifort
这里编译器最好与你后面编译vasp的编译器一致,否则容易编译出错。
then
make & make install
openmpi的设置可以参考相关资料,注意把ifort, mkl, icc, openmpi的lib加到INCLUDE路径中,否则也可能出错。
比如我的系统上
echo $INCLUDE
/opt/intel/mkl/10.1.0.015/include:/opt/intel/cce/10.1.018/include:/opt/intel/Compiler/11.0/074/include:/usr/local/include:/usr/include:/include:/opt/intel/mkl/10.1.0.015/include
=====================================================================
vasp编译过程:
2: Install VASP lib
mkdir /home/msemsi/share/vasp/vasp.5.lib/
cd /home/msemsi/share/vasp/vasp.5.lib/
tar xvfz /home/msemsi/share/vasp/vasp.5.lib.tar
cd vasp.5.lib/
cp makefile.linux_ifc_P4 makefile
vim makefile
前几行改成这样:
CPP = /opt/intel/cce/10.1.018/bin/icc -E -P -C $*.F >$*.f
CC= /opt/intel/cce/10.1.018/bin/icc
FC= /opt/intel/Compiler/11.0/074/bin/intel64/ifort
CFLAGS = -O
OFLAGS = -O3 -align -xT
FFLAGS = -I/opt/intel/mkl/10.1.0.015/include/fftw -FR -lowercase -assume byterecl $(OFLAGS)
FREE = -FR
DOBJ后面的就不用改了,其实保险起见,只要把原始makefile里CPP,FC改一下就行了
Then
make
=====================================================================
3: Compile VASP parallel general-k verision by openmpi
cd /home/msemsi/share/vasp/vasp.5.2
tar xvfz /home/msemsi/share/vasp/vasp.5.2.tar
cd vasp.5.2
cp makefile.linux_ifc_P4 makefile
修改makefile如下(并行版本),
.SUFFIXES: .inc .f .f90 .F
#-----------------------------------------------------------------------
# Makefile for Intel Fortran compiler for Pentium/Athlon/Opteron
# bases systems
# we recommend this makefile for both Intel as well as AMD systems
# for AMD based systems appropriate BLAS and fftw libraries are
# however mandatory (whereas they are optional for Intel platforms)
#
# The makefile was tested only under Linux on Intel and AMD platforms
# the following compiler versions have been tested:
# - ifc.7.1 works stable somewhat slow but reliably
# - ifc.8.1 fails to compile the code properly
# - ifc.9.1 recommended (both for 32 and 64 bit)
# - ifc.10.1 partially recommended (both for 32 and 64 bit)
# tested build 20080312 Package ID: l_fc_p_10.1.015
# the gamma only mpi version can not be compiles
# using ifc.10.1
#
# it might be required to change some of library pathes, since
# LINUX installation vary a lot
# Hence check ***ALL*** options in this makefile very carefully
#-----------------------------------------------------------------------
#
# BLAS must be installed on the machine
# there are several options:
# 1) very slow but works:
# retrieve the lapackage from
# and compile the blas routines (BLAS/SRC directory)
# please use g77 or f77 for the compilation. When I tried to
# use pgf77 or pgf90 for BLAS, VASP hang up when calling
# ZHEEV (however this was with lapack 1.1 now I use lapack 2.0)
# 2) more desirable: get an optimized BLAS
#
# the two most reliable packages around are presently:
# 2a) Intels own optimised BLAS (PIII, P4, PD, PC2, Itanium)
#
http://developer.intel.com/software/products/mkl/# this is really excellent, if you use Intel CPU's
#
# 2b) probably fastest SSE2 (4 GFlops on P4, 2.53 GHz, 16 GFlops PD,
# around 30 GFlops on Quad core)
# Kazushige Goto's BLAS
#
#
#
#-----------------------------------------------------------------------
# all CPP processed fortran files have the extension .f90
SUFFIX=.f90
#-----------------------------------------------------------------------
# fortran compiler and linker
#-----------------------------------------------------------------------
FC=/opt/intel/Compiler/11.0/074/bin/intel64/ifort
# fortran linker
FCL=$(FC)
#-----------------------------------------------------------------------
# whereis CPP ?? (I need CPP, can't use gcc with proper options)
# that's the location of gcc for SUSE 5.3
#
# CPP_ = /usr/lib/gcc-lib/i486-linux/2.7.2/cpp -P -C
#
# that's probably the right line for some Red Hat distribution:
#
# CPP_ = /usr/lib/gcc-lib/i386-redhat-linux/2.7.2.3/cpp -P -C
#
# SUSE X.X, maybe some Red Hat distributions:
CPP_ = ./preprocess <$*.F | /usr/bin/cpp -P -C -traditional >$*$(SUFFIX)
#-----------------------------------------------------------------------
# possible options for CPP:
# NGXhalf charge density reduced in X direction
# wNGXhalf gamma point only reduced in X direction
# avoidalloc avoid ALLOCATE if possible
# PGF90 work around some for some PGF90 / IFC bugs
# CACHE_SIZE 1000 for PII,PIII, 5000 for Athlon, 8000-12000 P4, PD
# RPROMU_DGEMV use DGEMV instead of DGEMM in RPRO (depends on used BLAS)
# RACCMU_DGEMV use DGEMV instead of DGEMM in RACC (depends on used BLAS)
#-----------------------------------------------------------------------
#CPP = $(CPP_) -DHOST=\"LinuxIFC\" \
-Dkind8 -DCACHE_SIZE=12000 -DPGF90 -Davoidalloc -DNGXhalf \
# -DRPROMU_DGEMV -DRACCMU_DGEMV
#-----------------------------------------------------------------------
# general fortran flags (there must a trailing blank on this line)
# byterecl is strictly required for ifc, since otherwise
# the WAVECAR file becomes huge
#-----------------------------------------------------------------------
FFLAGS = -FR -lowercase -assume byterecl
#-----------------------------------------------------------------------
# optimization
# we have tested whether higher optimisation improves performance
# -axK SSE1 optimization, but also generate code executable on all mach.
# xK improves performance somewhat on XP, and a is required in order
# to run the code on older Athlons as well
# -xW SSE2 optimization
# -axW SSE2 optimization, but also generate code executable on all mach.
# -tpp6 P3 optimization
# -tpp7 P4 optimization
#-----------------------------------------------------------------------
# ifc.9.1, ifc.10.1 recommended
OFLAG=-O3
OFLAG_HIGH = $(OFLAG)
OBJ_HIGH =
OBJ_NOOPT =
DEBUG = -FR -O0
INLINE = $(OFLAG)
#-----------------------------------------------------------------------
# the following lines specify the position of BLAS and LAPACK
# VASP works fastest with the libgoto library
# so that's what we recommend
#-----------------------------------------------------------------------
# mkl.10.0
# set -DRPROMU_DGEMV -DRACCMU_DGEMV in the CPP lines
#BLAS=-L/opt/intel/mkl100/lib/em64t -lmkl -lpthread
# even faster for VASP Kazushige Goto's BLAS
#
# parallel goto version requires sometimes -libverbs
BLAS= /lib64/libgoto_penrynp-r1.26.so
# LAPACK, simplest use vasp.5.lib/lapack_double
LAPACK= ../vasp.5.lib/lapack_double.o
# use the mkl Intel lapack
#LAPACK= -lmkl_lapack
#-----------------------------------------------------------------------
#LIB = -L../vasp.5.lib -ldmy \
../vasp.5.lib/linpack_double.o $(LAPACK) \
$(BLAS)
# options for linking, nothing is required (usually)
LINK =
#-----------------------------------------------------------------------
# fft libraries:
# VASP.5.2 can use fftw.3.1.X ()
# since this version is faster on P4 machines, we recommend to use it
#-----------------------------------------------------------------------
#FFT3D = fft3dfurth.o fft3dlib.o
# alternatively: fftw.3.1.X is slighly faster and should be used if available
#FFT3D = fftw3d.o fft3dlib.o /opt/libs/fftw-3.1.2/lib/libfftw3.a
#=======================================================================
# MPI section, uncomment the following lines until
# general rules and compile lines
# presently we recommend OPENMPI, since it seems to offer better
# performance than lam or mpich
#
# !!! Please do not send me any queries on how to install MPI, I will
# certainly not answer them !!!!
#=======================================================================
#-----------------------------------------------------------------------
# fortran linker for mpi
#-----------------------------------------------------------------------
FC=/usr/local/bin/mpif90
FCL=$(FC)
#-----------------------------------------------------------------------
# additional options for CPP in parallel version (see also above):
# NGZhalf charge density reduced in Z direction
# wNGZhalf gamma point only reduced in Z direction
# scaLAPACK use scaLAPACK (usually slower on 100 Mbit Net)
#-----------------------------------------------------------------------
CPP = $(CPP_) -DMPI -DHOST=\"LinuxIFC\" -DIFC \
-Dkind8 -DCACHE_SIZE=16000 -DPGF90 -Davoidalloc -DNGZhalf \
-DMPI_BLOCK=8000 -DRPROMU_DGEMV -DRACCMU_DGEMV
#-----------------------------------------------------------------------
# location of SCALAPACK
# if you do not use SCALAPACK simply leave that section commented out
#-----------------------------------------------------------------------
#BLACS=$(HOME)/archives/SCALAPACK/BLACS/
#SCA_=$(HOME)/archives/SCALAPACK/SCALAPACK
#SCA= $(SCA_)/libscalapack.a \
# $(BLACS)/LIB/blacsF77init_MPI-LINUX-0.a $(BLACS)/LIB/blacs_MPI-LINUX-0.a $(BLACS)/LIB/blacsF77init_MPI-LINUX-0.a
SCA=
#-----------------------------------------------------------------------
# libraries for mpi
#-----------------------------------------------------------------------
LIB = -L../vasp.5.lib -ldmy \
../vasp.5.lib/linpack_double.o \
-L/opt/intel/mkl/10.1.0.015/lib/em64t \
-lmkl_em64t -lguide -lpthread -lm \
$(BLAS)
#LINK = -static
# FFT: fftmpi.o with fft3dlib of Juergen Furthmueller
#FFT3D = fftmpi.o fftmpi_map.o fft3dfurth.o fft3dlib.o
# alternatively: fftw.3.1.X is slighly faster and should be used if available
FFT3D = fftmpi.o fftmpi_map.o fftw3d.o fft3dlib.o \
/opt/intel/mkl/10.1.0.015/lib/em64t/libfftw3xf_intel.a
#-----------------------------------------------------------------------
# general rules and compile lines
#-----------------------------------------------------------------------
BASIC= symmetry.o symlib.o lattlib.o random.o
SOURCE= base.o mpi.o smart_allocate.o xml.o \
constant.o jacobi.o main_mpi.o scala.o \
asa.o lattice.o poscar.o ini.o xclib.o xclib_grad.o \
radial.o pseudo.o mgrid.o gridq.o ebs.o \
mkpoints.o wave.o wave_mpi.o wave_high.o \
$(BASIC) nonl.o nonlr.o nonl_high.o dfast.o choleski2.o \
mix.o hamil.o xcgrad.o xcspin.o potex1.o potex2.o \
metagga.o constrmag.o cl_shift.o relativistic.o LDApU.o \
paw_base.o egrad.o pawsym.o pawfock.o pawlhf.o paw.o \
mkpoints_full.o charge.o dipol.o pot.o \
dos.o elf.o tet.o tetweight.o hamil_rot.o \
steep.o chain.o dyna.o sphpro.o us.o core_rel.o \
aedens.o wavpre.o wavpre_noio.o broyden.o \
dynbr.o rmm-diis.o reader.o writer.o tutor.o xml_writer.o \
brent.o stufak.o fileio.o opergrid.o stepver.o \
chgloc.o fast_aug.o fock.o mkpoints_change.o sym_grad.o \
mymath.o internals.o dimer_heyden.o dvvtrajectory.o vdwforcefield.o \
hamil_high.o nmr.o force.o \
pead.o subrot.o subrot_scf.o pwlhf.o gw_model.o optreal.o davidson.o \
electron.o rot.o electron_all.o shm.o pardens.o paircorrection.o \
optics.o constr_cell_relax.o stm.o finite_diff.o elpol.o \
hamil_lr.o rmm-diis_lr.o subrot_cluster.o subrot_lr.o \
lr_helper.o hamil_lrf.o elinear_response.o ilinear_response.o \
linear_optics.o linear_response.o \
setlocalpp.o wannier.o electron_OEP.o electron_lhf.o twoelectron4o.o \
ratpol.o screened_2e.o wave_cacher.o chi_base.o wpot.o local_field.o \
ump2.o bse.o acfdt.o chi.o sydmat.o
INC=
vasp: $(SOURCE) $(FFT3D) $(INC) main.o
rm -f vasp
$(FCL) -o vasp main.o $(SOURCE) $(FFT3D) $(LIB) $(LINK)
makeparam: $(SOURCE) $(FFT3D) makeparam.o main.F $(INC)
$(FCL) -o makeparam $(LINK) makeparam.o $(SOURCE) $(FFT3D) $(LIB)
zgemmtest: zgemmtest.o base.o random.o $(INC)
$(FCL) -o zgemmtest $(LINK) zgemmtest.o random.o base.o $(LIB)
dgemmtest: dgemmtest.o base.o random.o $(INC)
$(FCL) -o dgemmtest $(LINK) dgemmtest.o random.o base.o $(LIB)
ffttest: base.o smart_allocate.o mpi.o mgrid.o random.o ffttest.o $(FFT3D) $(INC)
$(FCL) -o ffttest $(LINK) ffttest.o mpi.o mgrid.o random.o smart_allocate.o base.o $(FFT3D) $(LIB)
kpoints: $(SOURCE) $(FFT3D) makekpoints.o main.F $(INC)
$(FCL) -o kpoints $(LINK) makekpoints.o $(SOURCE) $(FFT3D) $(LIB)
clean:
-rm -f *.g *.f *.o *.L *.mod ; touch *.F
main.o: main$(SUFFIX)
$(FC) $(FFLAGS) $(DEBUG) $(INCS) -c main$(SUFFIX)
xcgrad.o: xcgrad$(SUFFIX)
$(FC) $(FFLAGS) $(INLINE) $(INCS) -c xcgrad$(SUFFIX)
xcspin.o: xcspin$(SUFFIX)
$(FC) $(FFLAGS) $(INLINE) $(INCS) -c xcspin$(SUFFIX)
makeparam.o: makeparam$(SUFFIX)
$(FC) $(FFLAGS)$(DEBUG) $(INCS) -c makeparam$(SUFFIX)
makeparam$(SUFFIX): makeparam.F main.F
#
# MIND: I do not have a full dependency list for the include
# and MODULES: here are only the minimal basic dependencies
# if one strucuture is changed then touch_dep must be called
# with the corresponding name of the structure
#
base.o: base.inc base.F
mgrid.o: mgrid.inc mgrid.F
constant.o: constant.inc constant.F
lattice.o: lattice.inc lattice.F
setex.o: setexm.inc setex.F
pseudo.o: pseudo.inc pseudo.F
poscar.o: poscar.inc poscar.F
mkpoints.o: mkpoints.inc mkpoints.F
wave.o: wave.inc wave.F
nonl.o: nonl.inc nonl.F
nonlr.o: nonlr.inc nonlr.F
$(OBJ_HIGH):
$(CPP)
$(FC) $(FFLAGS) $(OFLAG_HIGH) $(INCS) -c $*$(SUFFIX)
$(OBJ_NOOPT):
$(CPP)
$(FC) $(FFLAGS) $(INCS) -c $*$(SUFFIX)
fft3dlib_f77.o: fft3dlib_f77.F
$(CPP)
$(F77) $(FFLAGS_F77) -c $*$(SUFFIX)
.F.o:
$(CPP)
$(FC) $(FFLAGS) $(OFLAG) $(INCS) -c $*$(SUFFIX)
.F$(SUFFIX):
$(CPP)
$(SUFFIX).o:
$(FC) $(FFLAGS) $(OFLAG) $(INCS) -c $*$(SUFFIX)
# special rules
#-----------------------------------------------------------------------
# these special rules are cummulative (that is once failed
# in one compiler version, stays in the list forever)
# -tpp5|6|7 P, PII-PIII, PIV
# -xW use SIMD (does not pay of on PII, since fft3d uses double prec)
# all other options do no affect the code performance since -O1 is used
fft3dlib.o : fft3dlib.F
$(CPP)
$(FC) -FR -lowercase -O2 -c -xT -unroll0 -vec_report3 $*$(SUFFIX)
fft3dfurth.o : fft3dfurth.F
$(CPP)
$(FC) -FR -lowercase -O1 -c $*$(SUFFIX)
radial.o : radial.F
$(CPP)
$(FC) -FR -lowercase -O1 -c $*$(SUFFIX)
symlib.o : symlib.F
$(CPP)
$(FC) -FR -lowercase -O1 -c $*$(SUFFIX)
symmetry.o : symmetry.F
$(CPP)
$(FC) -FR -lowercase -O1 -c $*$(SUFFIX)
wave_mpi.o : wave_mpi.F
$(CPP)
$(FC) -FR -lowercase -O1 -c $*$(SUFFIX)
wave.o : wave.F
$(CPP)
$(FC) -FR -lowercase -O1 -c $*$(SUFFIX)
dynbr.o : dynbr.F
$(CPP)
$(FC) -FR -lowercase -O1 -c $*$(SUFFIX)
asa.o : asa.F
$(CPP)
$(FC) -FR -lowercase -O1 -c $*$(SUFFIX)
broyden.o : broyden.F
$(CPP)
$(FC) -FR -lowercase -O2 -c $*$(SUFFIX)
us.o : us.F
$(CPP)
$(FC) -FR -lowercase -O1 -c $*$(SUFFIX)
LDApU.o : LDApU.F
$(CPP)
$(FC) -FR -lowercase -O2 -c $*$(SUFFIX)
Then
make
Test:
cd /home/msemsi/share/vasp
rm -rf bench.Hg
mkdir bench.Hg
cd bench.Hg
tar xvf /home/msemsi/share/vasp/bench.Hg.tar.gz
mpirun -np 4 ../vasp.5.2/vasp
=======================================================================================
4.编译针对Gamma点计算的版本,采用下述设置,官方文档认为可提高一倍速度。
串行:CPP = ... cpp ... -DNGXhalf -DwNGXhalf ...
并行:CPP = ... cpp ... -DNGZhalf -DwNGZhalf ... #其实只需要添加-DwNGZhalf,前面的-DwNGZhalf在原始makefile中就有
touch *.F
make vasp
=======================================================================================
5: Compile VASP-NEB with openmpi
cd /home/msemsi/share/vasp/vtstcode
tar xvfz .. /vasp.5.2.tar
wget
tar xvfz vtstcode.tar.gz
mv vtstcode/* .
Almost the same Makefile as the step 3 "Compile VASP parallel general-k verision by openmpi", except to add between
steep.o and chain.o
with
dimer.o dynmat.o neb.o lanczos.o instanton.o sd.o cg.o qm.o lbfgs.o bfgs.o fire.o opt.o
Then
make
============================================================================================================================================
参考的网上资料:
Ref1:
Be warned!!!!! The makefile on this web page will give a very slow VASP executable! It would be nice if INTEL support would post an updated version of this makefile.
VASP needs to use the sequential mkl libraries (mkl_sequential), not the multithreaded libraries (mkl_intel_thread).
Linking with mkl_intel_thread will generate a VASP executable which will run 2 to 10 times slower, if the environmental variable OMP_NUM_THREADS is not set to 1
另外如果-O2 参数编译的vasp速度不比-O3编译的快多少,为安全起见,尽量使用-O2来编译。
Ref2: VASP安裝編譯fft3dlib.F的問題和加入Core 2 Duo變數
本研究室共串聯4台個人PC,配備如下:
CPU:Intel Core 2 Duo E6600
complier: IFC 9.1.040
library: MKL 8.1
使用vasp.4.6的makefile.linux_ifc_P4串平行編譯時,若遇到以下兩問題的解決方式:
問題1:
從VASP-FTP下載vasp.4.6原始碼時,FTP上會看到有分壓縮和未壓縮2種版本
修改比較makefile.linux_ifc_P4會發現同樣版本但makefile內容卻有差!
在第88行,其差異如下~
壓縮版本:FFLAGS = -FR -lowercase -assume byterecl
未壓縮版本:FFLAGS = -FR -lowercase
使用壓縮版本進行編譯,若遇到以下問題:
mpif90 -FR -lowercase -assume byterecl-FR -O0 -c main.f90
ifort: Command line error: Unrecognized keyword 'byterecl_FR' for option '-assume'
make: *** [main.o] Error 1
解決方式1:
在第285行,會發現在主程式的物件檔編寫中少了一個空格,造成error產生
$(FC) $(FFLAGS)$(DEBUG) $(INCS) -c main$(SUFFIX)
改成
$(FC) $(FFLAGS) $(DEBUG) $(INCS) -c main$(SUFFIX)
解決方式2:(不建議)
只要將變數"-assume byterecl"去掉,改成FFLAGS = -FR -lowercase即可!
問題2:
串平行編譯時,若遇到FFTW的fft3dlib.F出現以下問題:
fortcom: Error: fft3dlib.f90, line 1627: Sharing of a DO termination statement by more than one DO statement is an obsolescent feature in Fortran 95. Use an END DO or CONTINUE statement for each DO statement. [20]
20 CONTINUE
---^
fortcom: Error: fft3dlib.f90, line 1704: The computed GOTO statement is an obsolescent feature in Fortran 95.
GOTO (10,50,90,130,170,210,250),IGO
------^
fortcom: Error: fft3dlib.f90, line 2625: The computed GOTO statement is an obsolescent feature in Fortran 95.
GOTO (10,50,90,130,170,210,250),IGO
------^
fortcom: Error: fft3dlib.f90, line 3531: The computed GOTO statement is an obsolescent feature in Fortran 95.
GOTO (10,50,90,130,170,210,250),IGO
------^
fortcom: Error: fft3dlib.f90, line 4064: The computed GOTO statement is an obsolescent feature in Fortran 95.
GOTO (1010,1050,1090,1130,1170,1210,1250),IGO
------^
compilation aborted for fft3dlib.f90 (code 1)
make: *** [fft3dlib.o] Error 1
主因為fft3dlib.F主要撰寫語法為F77,但IFC是以F95語法去讀取,所以多少會出現警告訊息!
解決方式:
將第343行
$(FC) -FR -lowercase -O1 -tpp7 -xW -prefetch- -unroll0 -e95 -vec_report3 -c $*$(SUFFIX)
去掉"-e95"變數改成
$(FC) -FR -lowercase -O1 -tpp7 -xW -prefetch- -unroll0 -vec_report3 -c $*$(SUFFIX)
因為"-e95"變數會把F95編譯F77程式碼產生的警告(warning)改成錯誤(error)型式輸出,造成編譯強迫停止無法忽略跳過!
順道一提,若使用Intel Core 2 Duo的處理器,可將以下的-xW變數改成-xT得到優化
第102行:OFLAG=-O3 -xW -tpp7
第343行:$(FC) -FR -lowercase -O1 -tpp7 -xW -prefetch- -unroll0 -vec_report3 -c $*$(SUFFIX)
改成
第102行:OFLAG=-O3 -xT -tpp7
第343行:$(FC) -FR -lowercase -O1 -tpp7 -xT -prefetch- -unroll0 -vec_report3 -c $*$(SUFFIX)
编译器和数学库对VASP的计算速度的影响
一、所采用的机器、编译器、数学库、vasp程序和计算的例子
1. 机器是2.4 MHz的CPU、512M的内存,操作系统Redhat7.3的单机。
2. VASP:4.6版本
3. Fotran编译:PGI公司的pgf90 3.1版本;Intel公司的ifc 6.1版本。
4. 数学库:
BLAS:blas.tgz;ATLAS (atlas3.6.0_Linux_P4SSE2.tar.gz); Kazushige Goto's BLAS (libgoto_p4_512-r0.94.so);MKL 5.2 version
LAPACK:ATLAS (atlas3.6.0_Linux_P4SSE2.tar.gz);MKL 5.2 version;vasp.4.lib/lapack_double.o
注释:blas.tgz从下载。
Atlas3.6.0_Linux_P4SSE2.tar.gz 从
下载。
Libgoto_p4_512-r0.94.so从
下载。
5. 计算的例子:采用PBE-GGA的PAW势(Nb_pv, Sn_d 和 C)优化Nb2SnC的几何结构,主要的输入文件如下:
-------INCAR---------------------
SYSTEM = Nb2SnC
ENCUT = 500
ISMEAR = 1 ; SIGMA = 0.18
ISTART = 0 ; ICHARG = 2
GGA = PE
EDIFF = 1E-5 ; EDIFFG = -1E-2
NSW = 60; IBRION = 2
ISIF = 3 ; POTIM = 0.1
PREC= Accurate ; NBANDS = 60
-------POSCAR---------------------
Nb2SnC:
3.241000000000000
0.0000000000000000 -1.0110073892063070 0.0000000000000000
0.8755580824825159 0.5055036946031537 0.0000000000000000
0.0000000000000000 0.0000000000000000 4.2888295708784200
4 2 2
Direct
0.3333333333333357 0.6666666666666643 0.0828382368328892
0.6666666666666643 0.3333333333333357 0.5828382368328890
0.6666666666666643 0.3333333333333357 0.9171617631671110
0.3333333333333357 0.6666666666666643 0.4171617631671108
0.3333333333333357 0.6666666666666643 0.7500000000000000
0.6666666666666643 0.3333333333333357 0.2500000000000000
0.0000000000000000 0.0000000000000000 0.0000000000000000
0.0000000000000000 0.0000000000000000 0.5000000000000000
-------KPOINTS---------------------
auto
0
Gamma
11 11 5
0.0 0.0 0.0
--------------------------------------
二、编译VASP时的主要参数设置
先编译blas.tgz,编译的步骤:
先建立一个blas目录,把blas.tgz拷贝到blas目录下,用命令tar xzvf blas.tgz解压blas.tgz,
得到很多.f的源文件。对pgf90,采用pgf90 –c *.f 编译这些源文件,得到很多.o的文件,然
后把.o文件用命令ar -rv blas.a *.o压缩成blas.a文件。对ifc,使用ifc –c *.f,其他的命令类
似。
Blas.a, libgoto_p4_512-r0.6.so, 从atlas3.6.0_Linux_P4SSE2.tar.gz解压得到的libatlas.a,
libf77blas.a, libcblas.a, liblapack.a等库文件都放在vasp.4.6的上一级目录下的lib目录中。
重要的参数设置,就算makefile文件中的ATLASHOME, BLAS, LAPACK。
采用pgf90 编译VASP时,OFLAG = -O2 -tp p6
1.
ATLASHOME= ../lib
BLAS= -L$(ATLASHOME) -lf77blas -latlas
LAPACK= ../vasp.4.lib/lapack_atlas.o -L$(ATLASHOME) -llapack -lcblas
2.
ATLASHOME= ../lib
BLAS= -L$(ATLASHOME) -lf77blas -latlas
LAPACK= ../vasp.4.lib/lapack_double.o
3.
BLAS= ../blas/blas.a
LAPACK= ../vasp.4.lib/lapack_double.o
4.
BLAS= ../lib/libgoto_p4_512-r0.94.so
LAPACK= ../vasp.4.lib/lapack_double.o
采用ifc编译VASP时,OFLAG=-O3 -xW -tpp7
FFT3D = fft3dfurth.o fft3dlib.o
5.
ATLASHOME= ../lib/
BLAS= -L$(ATLASHOME) -lf77blas -latlas
LAPACK= ../vasp.4.lib/lapack_double.o
6.
ATLASHOME= ../lib/
BLAS= -L$(ATLASHOME) -lf77blas -latlas
LAPACK= ../vasp.4.lib/lapack_atlas.o -L $(ATLASHOME) -llapack -lcblas
7.
BLAS= ../blas/blasifc.a
LAPACK= ../vasp.4.lib/lapack_double.o
8.
BLAS= ../lib/libgoto_p4_512-r0.94.so
LAPACK= ../vasp.4.lib/lapack_double.o
9.
BLAS=-L/opt/intel/mkl/lib/32 -lmkl_p4 -lguide -lpthread
LAPACK= -lmkl_lapack
10.
BLAS=-L/opt/intel/mkl/lib/32 -lmkl_p4 -lguide -lpthread
LAPACK= ../vasp.4.lib/lapack_double.o
11.
BLAS=-L/opt/intel/mkl/lib/32 -lmkl_p4 -lguide -lpthread
LAPACK= -L/opt/intel/mkl/lib/32 -lmkl_lapack
三、计算测试得到的数据
Serial Total CPU Time User Time Elapsed time
(min) (min) (min)
1 102.541 98.016 102.611
2 102.457 97.894 102.520
3 130.943 127.696 131.015
4 92.880 89.603 92.956
5 79.399 74.624 79.538
6 78.909 74.272 78.983
7 76.977 73.700 77.022
8 68.589 65.276 68.660
9 70.478 67.092 70.544
10 69.902 66.548 69.955
11 70.450 67.082 70.493
注释:Serial指的是在编译时,主要参数设置的情况,见第二部分中的说明。Total CPU Time
指的是计算所花的总的时间。单位是分钟。
四、结论
采用intel的ifc编译器,BLAS采用Kazushige Goto's BLAS,LAPACK采用lapack_double.o
的编译得到vasp计算速度最快。
采用pgf90,blas.tgz和lapack_double.o来编译得到的vasp的计算速度最慢。
对采用pgf90,BLAS采用Kazushige Goto's BLAS,编译得到的vasp的计算速度提
高9%~29%。但是对采用ifc时,BLAS采用Kazushige Goto's BLAS,编译得到的vasp的
计算速度提高10%~13%。
采用pgf90,BLAS采用ATLAS时,不同的LAPACK库对计算速度影响非常小。
对ifc,BLAS采用同样的数学库时,不同的LAPACK库对计算速度的影响也是非常小。
总的一点,采用intel的ifc来编译要比采用PGI的pgf90来编译得到的vasp计算速度快,
大约快14%~47%,与采用的数学库有关。
安装编译环境: