3 This test suite is a *supplement* to other test suites, including the
4 original MPICH testsuite, the Intel testsuite, and the IBM MPI test suite
5 (or test suites derived from that test, including the MPI C++ tests).
7 Building the Test Suite
8 =======================
9 In many cases, configure will find the MPI implementation
10 automatically. In some cases, it will need some help. For example:
12 For IBM MPI, where the compilation commands are not mpicc and mpif77 etc.:
14 ./configure CC=xlc MPICC=mpcc F77=xlf MPIF77=mpxlf CXX=xlC \
15 MPICXX="mpCC -cpp" F90=xlf90 MPIF90=mpxlf90 \
19 (or the _r versions of the compilers)
21 If mpicc and friends are not in your default path (and you do not want to
22 add them), you can specify the path with --with-mpi=<path>. For example,
23 if they are in /usr/local/mympi/bin, use
25 ./configure --with-mpi=/usr/local/mympi
27 (configure will append the bin to the path that you give).
29 You may need to add MPI_SIZEOF_OFFSET=8 .
31 The option "-cpp" is needed for at least some versions of mpCC to define the
32 C++ bindings of the MPI routines.
34 For implementations that do not implement all of MPI-2, there are --disable
35 options, including --disable-spawn and --disable-cxx. To restrict tests to
36 just what is defined in the MPI specification, use --enable-strictmpi .
38 The script that runs the tests assumes that the MPI implementation
39 supports mpiexec; you should consider this the first test of the implementation.
43 The following environment variables will modify the behavior of the tests
45 MPITEST_DEBUG - if set, output information for debugging the test suite
46 MPITEST_VERBOSE - if set to an integer value, output messages whose
47 level is at least that value (0 is a good choice here)
48 MPITEST_RETURN_WITH_CODE - Set the return code from the test programs based on
49 success or failure, with a zero for success and one
50 for failure (value must be yes, YES, true, or TRUE to
52 MPITEST_THREADLEVEL_DEFAULT - Set the default thread level. Values are
53 multiple, serialized, funneled, and single.
57 For systems that run applications through a batch system, the option "-batch"
58 to the runtests script will create a script file that can be edited and
59 submitted to the batch system. The script checktests can be run to
60 summarize the results.
62 Specifically, (assuming the bash shell, and that the directory "btest", a
63 subdirectory of the test suite directory, is used for running the tests):
65 export MPITEST_BATCHDIR=`pwd`/btest
66 runtests -batch -tests=testlist
67 ... edit btest/runtests.batch to make it a value batch submissions script
68 ... run that script and wait for the batch job to complete
69 cd btest && ../checktests
71 If a program other than mpiexec is used in the batch form to run programs, then
72 specify that to runtests:
74 runtests -batch -mpiexec=aprun -tests=testlist
76 (Here, aprun is the command used on Cray XE6 systems.)
78 Note that some programs that are used to run MPI programs add extra output,
79 which can confuse any tool that depends on clean output in STDOUT. Since
80 such unfortunate behavior is common, the option -ignorebogus can be given
83 cd btest && ../checktests --ignorebogus
85 Controlling the Tests that are Run
86 ==================================
87 The tests are actually built and run by the script "runtests". This script
88 can be given a file that contains a list of the tests to run. This file has
89 two primary types of entries:
91 directories: Enter directory and look for the file "testlist".
92 Recursively run the contents of that file
93 program names: Build and run that program
95 Lines may also be commented out with "#".
97 The simplest program line contains the name of the program and the number of
98 MPI processes to use. For example, the following will build the
99 program sendrecv1 and run it with 4 processes:
103 In addition, the program line can contain key=value pairs that provide
104 special information about running the test. For example,
106 sendflood 8 timeLimit=600
108 says to build and run the program sendflood with 8 MPI processes and
109 permit the test to run for 600 seconds (by default, at least for
110 MPICH, the default timelimit is 180 seconds). Other key=value pairs
111 can be used to select whether a program should be run at all,
112 depending on the abilities of the MPI implementation (this is
113 particularly important for Fortran programs, since preprocessor
114 support for Fortran is a non-standard extension to the Fortran
115 language, and there are some compilers that would not accept Fortran
116 programs that used the preprocessor).
118 The most important key=value pairs are:
121 timeLimit=n : Use a timelimit of n seconds
123 arg=string : Run the program with string as an argument to the program
125 mpiexecarg=string : Run the program with string as an argument to mpiexec
127 env=name=value : Run the program with environment variable "name" given the
130 mpiversion=x.y : Build and run the program only if the MPI version is at
131 least x.y. For example,
133 distgraph1 4 mpiversion=2.2
135 will build and run distgraph1 with 4 MPI processes only
136 if the MPI version is at least 2.2.
138 strict=bool : If bool is false, only build and run the program if
139 --enable-strictmpi was not used in configuring the test suite.
140 That is, a line such as
142 neighb_coll 4 strict=false
144 Says that this test is not valid for a strict MPI implementation;
145 it contains extensions to the standard, or in the case of some
146 MPICH development, MPIX routines
148 resultTest=proc : This is used to change the way in which the success or
149 failure of a test is evaluated. proc is one of several
150 Perl subroutines defined within the runtest program. These
151 are primarily used within the testsuite for tests programs
152 exit with expected status values or that timeouts are