OSDN -- 开发和下载开源软件
简体中文翻译状态 translation status for zh
  • R/O
  • HTTP
  • SSH
  • HTTPS
Fork

提交

标签
No Tags

Frequently used words (click to add to your profile)

javac++androidlinuxc#windowsobjective-ccocoa誰得qtpythonphprubygameguibathyscaphec計画中(planning stage)翻訳omegatframeworktwitterdomtestvb.netdirectxゲームエンジンbtronarduinopreviewer

Commit MetaInfo

修订版314ab5863865ce512193a6b2e5458205be056d9f (tree)
时间2013-07-24 16:36:28
作者Mikiya Fujii <mikiya.fujii@gmai...>
CommiterMikiya Fujii

Log Message

README is updated for MPI-parallelization. #31588

git-svn-id: https://svn.sourceforge.jp/svnroot/molds/trunk@1416 1136aad2-a195-0410-b898-f5ea1d11b9d8

更改概述

差异

--- a/doc/README.txt
+++ b/doc/README.txt
@@ -29,49 +29,71 @@
2929
3030 ==============================================================================
3131 REQUIREMENTS:
32- MolDS requires c++ mpi compiler that is wrapping Intel (icpc) or GNU (g++) and boost-libraries.
33- Valid versions of the wrapped c++ compilers are icpc 12.0.4(MkL 10.3 update 4), g++ 4.4, or later
34- because the MolDS is implemented with openMP 3.0.
35- To compile MolDS with GNU, furthermore, openBLAS (version 0.2.5 or later) is also required.
36-
37- To get and install the boost-libraries, see the HP:<http://www.boost.org/>.
38- The version of the boost would be no problem if 1.46.0 or later is used.
39- Especially, the boost-libraries should be builded with MPI
40- because MolDS needs boost_mpi-library(i.e. -lboost_mpi).
41-
42- To get and install the openBLAS-libraries, see the HP:<http://xianyi.github.com/OpenBLAS/>.
43- Note that "USE_OPENMP = 1" should be set for the installation of the opneBLAS.
44- Furthermore, "INTERFACE64 = 1" is also needed when you install the openBLAS into 64-bits machines
32+ -Compilers:
33+ MolDS requires c++ mpi compiler (e.g. Intel MPI or Open MPI)
34+ that is wrapping Intel (icpc with MKL) or GNU (g++) c++ compiler.
35+ Valid versions of the mpi compilers are Intel MPI 4.0.2, Open MPI 1.4.5, or later.
36+ Valid versions of the wrapped c++ compilers are icpc 12.0.4(MkL 10.3 update 4),
37+ g++ 4.4, or later because the MolDS is implemented with openMP 3.0.
38+
39+ -Boost C++ Libraries
40+ Boost C++ Libraries builded with MPI is needed.
41+ To get and install the Boost, see the HP:<http://www.boost.org/>.
42+ The version of the boost would be no problem if 1.46.0 or later is used.
43+ Especially, the Boost should be builded with MPI
44+ because MolDS needs boost_mpi-library(i.e. -lboost_mpi).
45+
46+ -Linear Algebra Packages (i.e. BLAS and LAPACK)
47+ MolDS needs a linear algebra package. In the current implementation of MolDS,
48+ MKL (Intel's Math Kernel Library) or OpenBLAS is assumed as the linear algebra package
49+ for the Intel or GNU compilers, respectively.
50+ See also the section of compilers about the version of the MKL.
51+ To get and install the OpenBLAS-libraries, see the HP:<http://xianyi.github.com/OpenBLAS/>.
52+ The version of the OpenBLAS would be no problem if 0.2.5 or later is used.
53+ Note that "USE_OPENMP = 1" should be set for the installation of the opneBLAS.
54+ Furthermore, "INTERFACE64 = 1" is also needed when you install the OpenBLAS into 64-bits machines
4555
4656 ==============================================================================
4757 COMPILE(using GNUmake):
4858 In the "src" directory in the MolDS package.
4959
50- Case i) Using Intel mpi c++ compiler (mpiicpc)
60+ Case i) The Intel mpi compiler (mpiicpc) which is wrapping the Intel c++ compiler (icpc)
5161 Change the "BOOST_TOP_DIR" in Makefile to the top directory of the
52- boost-libraries in your systems.
62+ Boost C++ Libraries in your systems.
5363
5464 To compile MolDS on 32 bits machine,
5565 $ make INTEL=32
5666
5767 To compile MolDS on 64 bits machine,
5868 $ make INTEL=64
69+
70+ Case ii) The openMPI compiler (mpicxx) which is wrapping the Intel c++ compiler (icpc)
71+ Change the "BOOST_TOP_DIR" in Makefile to the top directory of the
72+ Boost C++ Libraries in your systems.
73+
74+ To compile MolDS on 32 bits machine,
75+ $ make INTEL=32 CC=mpicxx
5976
60- Case ii) Using GNU c++ compiler (mpicxx)
77+ To compile MolDS on 64 bits machine,
78+ $ make INTEL=64 CC=mpicxx
79+
80+
81+ Case iii) The openMPI compiler (mpicxx) which is wrapping the GNU c++ compiler (g++)
6182 Change the "BOOST_TOP_DIR" in "Makefile_GNU" to the top directory of the
62- boost-libraries in your systems.
83+ Boost C++ Libraries in your systems.
6384 Change the "OPENBLAS_TOP_DIR" in "Makefile_GNU" to the top directory of the
64- boost-libraries in your systems.
85+ OpneBLAS in your systems.
6586
6687 Then, just type:
6788 $ make -f Makefile_GNU
6889
69- For both case, the compile succeeded if you could fine "MolDS.out" in the "src" directory.
70- Type "$ make clean" when you wanna clean the compilation.
90+ For all case, the compile succeeded if you could fine "MolDS.out" in the "src" directory.
91+ If you want to clean the compilation, type
92+ $ make clean
7193 If you want to compile MolDS in debug-mode,
7294 -g, -rdynamic(for function names in backtrace) and -DMOLDS_DBG should be added to CFLAGS,
73- that is, hit the following command:
74- $make CFLAGS="-O0 -g -rdynamic -DMOLDS_DBG"
95+ namely, hit the following command:
96+ $ make CFLAGS="-O0 -g -rdynamic -DMOLDS_DBG"
7597
7698 ==============================================================================
7799 CARRY OUT MolDS:
@@ -82,8 +104,31 @@ CARRY OUT MolDS:
82104 or
83105 $ ./MolDS.out input.in
84106
85- For the calculations with multiple processes(n) by MPI:
86- $ mpirun -np n MolDS.out input.in
107+ For the calculations with muliple threads, type
108+ $ export OMP_NUM_THREADS=n1
109+ $ ./MolDS.out input.in
110+ , where n1 is the number of threads.
111+
112+ For the calculations with multiple processes by MPI:
113+ $ mpirun -np n2 MolDS.out input.in
114+ , where n2 after the "-np" is the number of process.
115+
116+ For the calculations with muliple threads and muliple processes, type
117+ $ export OMP_NUM_THREADS=n1
118+ $ mpirun -np n2 MolDS.out input.in
119+ , where n1 is the number of cores of each node and n2 is the number of nodes.
120+
121+ In the multiple processes calculations, process-0 can only output results.
122+ If you want to get all output from the all process,
123+ -DMOLDS_DBG should be added to CFLAGS at the compilation.
124+ Then, make only one process on each node and output results to
125+ node unique file (e.g. local file system of each node.),
126+ namely,
127+ $ make CFLAGS="-DMOLDS_DBG"
128+ $ export OMP_NUM_THREADS=n1
129+ $ mpirun -np n2 MolDS.out input.in > /localFileSyste/output.dat
130+ , where n1 is the number of cores of each node and n2 is the number of nodes.
131+
87132 ==============================================================================
88133 SAMPLE and TEST
89134 See files in "test" directories for sample files.
@@ -99,44 +144,50 @@ SAMPLE and TEST
99144
100145 $ ruby Test_Of_MolDS.rb test1.in test2.dat test3 ...
101146
147+ Note that this test script needs at least 4 cores.
148+
102149 ==============================================================================
103150 CAPABILITIES:
104151
105- Electronic state and molecular dynamics:
106- | HF | CIS | MD(gs) | MD(es) | MC(gs) | MC(es) | RPMD(gs) | RPMD(es) | Optimize(gs) | Optimize(es) | Frequencies(gs) |
107- ---------|-----|-----|--------|--------|--------|--------|----------|----------|--------------|--------------|-----------------|
108- CNDO2 | OK | -- | -- | -- | OK | -- | -- | -- | -- | -- | -- |
109- ---------|-----|-----|--------|--------|--------|--------|----------|----------|--------------|--------------|-----------------|
110- INDO | OK | -- | -- | -- | OK | -- | -- | -- | -- | -- | -- |
111- ---------|-----|-----|--------|--------|--------|--------|----------|----------|--------------|--------------|-----------------|
112- ZINDO/S | OK | OK | OK | OK | OK | OK | OK | OK | OK | OK | -- |
113- ---------|-----|-----|--------|--------|--------|--------|----------|----------|--------------|--------------|-----------------|
114- MNDO | OK | OK | OK | OK | OK | OK | OK | OK | OK | OK | OK |
115- ---------|-----|-----|--------|--------|--------|--------|----------|----------|--------------|--------------|-----------------|
116- AM1 | OK | OK | OK | OK | OK | OK | OK | OK | OK | OK | OK |
117- ---------|-----|-----|--------|--------|--------|--------|----------|----------|--------------|--------------|-----------------|
118- AM1-D | OK | OK | OK | OK | OK | OK | OK | OK | OK | OK | OK |
119- ---------|-----|-----|--------|--------|--------|--------|----------|----------|--------------|--------------|-----------------|
120- PM3 | OK | OK | OK | OK | OK | OK | OK | OK | OK | OK | OK |
121- ---------|-----|-----|--------|--------|--------|--------|----------|----------|--------------|--------------|-----------------|
122- PM3-D | OK | OK | OK | OK | OK | OK | OK | OK | OK | OK | OK |
123- ---------|-----|-----|--------|--------|--------|--------|----------|----------|--------------|--------------|-----------------|
124- PM3/PDDG | OK | OK | OK | OK | OK | OK | OK | OK | OK | OK | OK |
152+ -Electronic state and molecular dynamics
153+ | HF | CIS | MD(gs) | MD(es) | MC(gs) | MC(es) | RPMD(gs) | RPMD(es) | Optimize(gs) | Optimize(es) | Frequencies(gs) |
154+ ---------|-----|-----|--------|--------|--------|--------|----------|----------|--------------|--------------|-----------------|
155+ CNDO2 | OK | -- | -- | -- | OK | -- | -- | -- | -- | -- | -- |
156+ ---------|-----|-----|--------|--------|--------|--------|----------|----------|--------------|--------------|-----------------|
157+ INDO | OK | -- | -- | -- | OK | -- | -- | -- | -- | -- | -- |
158+ ---------|-----|-----|--------|--------|--------|--------|----------|----------|--------------|--------------|-----------------|
159+ ZINDO/S | OK | OK | OK | OK | OK | OK | OK | OK | OK | OK | -- |
160+ ---------|-----|-----|--------|--------|--------|--------|----------|----------|--------------|--------------|-----------------|
161+ MNDO | OK | OK | OK | OK | OK | OK | OK | OK | OK | OK | OK |
162+ ---------|-----|-----|--------|--------|--------|--------|----------|----------|--------------|--------------|-----------------|
163+ AM1 | OK | OK | OK | OK | OK | OK | OK | OK | OK | OK | OK |
164+ ---------|-----|-----|--------|--------|--------|--------|----------|----------|--------------|--------------|-----------------|
165+ AM1-D | OK | OK | OK | OK | OK | OK | OK | OK | OK | OK | OK |
166+ ---------|-----|-----|--------|--------|--------|--------|----------|----------|--------------|--------------|-----------------|
167+ PM3 | OK | OK | OK | OK | OK | OK | OK | OK | OK | OK | OK |
168+ ---------|-----|-----|--------|--------|--------|--------|----------|----------|--------------|--------------|-----------------|
169+ PM3-D | OK | OK | OK | OK | OK | OK | OK | OK | OK | OK | OK |
170+ ---------|-----|-----|--------|--------|--------|--------|----------|----------|--------------|--------------|-----------------|
171+ PM3/PDDG | OK | OK | OK | OK | OK | OK | OK | OK | OK | OK | OK |
125172
126173 "OK", "Sch", and "--" mean available, shceduled, and non-scheduled methods, respectively.
127174 "gs" and "es" mean ground and excited states, respectively.
128175 i.e., MD(gs) and MD(es) mean Born-Oppenheimer Molecular Dynamics on ground and excited states, respectively.
129176
130- Elements:
131- CNDO2 | H, Li, C, N, O, and S
132- INDO | H, Li, C, N, and O
133- ZINDO/S | H, C, N, O, and S
134- MNDO | H, C, N, O, and S
135- AM1 | H, C, N, O, and S
136- AM1-D | H, C, N, O, and S
137- PM3 | H, C, N, O, and S
138- PM3-D | H, C, N, O, and S
139- PM3/PDDG | H, C, N, O, and S
177+ -Elements
178+ CNDO2 | H, Li, C, N, O, and S
179+ INDO | H, Li, C, N, and O
180+ ZINDO/S | H, C, N, O, and S
181+ MNDO | H, C, N, O, and S
182+ AM1 | H, C, N, O, and S
183+ AM1-D | H, C, N, O, and S
184+ PM3 | H, C, N, O, and S
185+ PM3-D | H, C, N, O, and S
186+ PM3/PDDG | H, C, N, O, and S
187+
188+ -Parallelization
189+ Open MP parallelization: everywhere in MolDS
190+ MPI parallelization: CIS is only parallelized with MPI.
140191
141192 ==============================================================================
142193 HOW TO WRITE INPUT:
@@ -158,6 +209,7 @@ HOW TO WRITE INPUT:
158209 THEORY_END
159210
160211 -options
212+ Write below options in SCF-directive.
161213 "max_iter", "rms_density", "damping_thresh", "damping_weight",
162214 "diis_num_error_vect", "diis_start_error", "diis_end_error",
163215 "vdW", "vdW_s6", and "vdW_d" are prepared as options.