[Trilinos-Users] zoltan tests.
Jason Slemons
slemons at cray.com
Thu Oct 15 16:19:40 MDT 2009
Hello Karen,
Assuming that BOOL=TRUE and BOOL=ON do the same thing I think we have pretty similar cmake scripts. Some differences I see are here:
-D Trilinos_EXTRA_LINK_FLAGS:STRING="-framework vecLib" \
-D Zoltan_ENABLE_EXAMPLES:BOOL=ON \
-D MPIEXEC_MAX_NUMPROCS:STRING=11 \
-D Trilinos_VERBOSE_CONFIGURE:BOOL=ON \
Also I do have that directory you mentioned:
trilinos-10.0-Source/pgi.BUILD/packages/zoltan/test/CMakeFiles/Zoltan_ZDriveMainCopyFiles.dir> ls
DependInfo.cmake cmake_clean.cmake depend.make
build.make depend.internal progress.make
I have a lot of TPLS in my cmake script but here is everything else:
cmake\
...(TPL STUFF)
-D CMAKE_CXX_FLAGS:STRING="-DHAVE_MPI_CXX -DMPICH_IGNORE_CXX_SEEK"\
-D CMAKE_CXX_COMPILER_FLAGS:STRING="-Wl,-Bstatic"\
-D Trilinos_EXTRA_LINK_FLAGS:STRING="-Wl,-Bstatic"\
-D Trilinos_ENABLE_Rythmos:BOOL=OFF\
-D Trilinos_ENABLE_Amesos:BOOL=ON\
-D Trilinos_ENABLE_AztecOO:BOOL=ON\
-D Trilinos_ENABLE_Zoltan:BOOL=ON\
-D Trilinos_ENABLE_Teuchos:BOOL=ON\
-D Trilinos_ENABLE_Sundance:BOOL=OFF\
-D Trilinos_ENABLE_ALL_OPTIONAL_PACKAGES:BOOL=OFF\
-D Trilinos_ENABLE_TESTS:BOOL=OFF\
-D Amesos_ENABLE_TESTS:BOOL=ON\
-D Zoltan_ENABLE_TESTS:BOOL=ON\
-D Epetra_ENABLE_TESTS:BOOL=ON\
-D Trilinos_ENABLE_ALL_PACKAGES:BOOL=ON\
-D Trilinos_ENABLE_Fortran:BOOL=OFF\
-D TPL_ENABLE_MPI:BOOL=ON \
-D MPI_BASE_DIR:PATH="${MPICH_DIR}"\
-D MPI_EXEC:FILEPATH="aprun"\
-D BUILD_SHARED_LIBS:BOOL=OFF\
-D CMAKE_VERBOSE_MAKEFILE:BOOL=TRUE\
-D CMAKE_CXX_COMPILER:STRING="${ASYNCPE_DIR}/bin/CC"\
-D CMAKE_C_COMPILER:STRING="${ASYNCPE_DIR}/bin/cc"\
-D CMAKE_Fortran_COMPILER:STRING="${ASYNCPE_DIR}/bin/ftn"\
-D DART_TESTING_TIMEOUT:STRING=60 \
-D CMAKE_INSTALL_PREFIX:PATH=${base_dir}/trilinos.${PE_ENV}.install.$$ \
-D MPI_EXEC_NUMPROCS_FLAG:STRING=-n \
A separate issue I've been having most of the time is that I try running ctest as soon as the make was over but my tests always fail('no tests were found' or 'couldn't find executables'). So I delete CMakeCache.txt and re-ran the cmake script and the second time around (the folder you mention is still there with the same files in it and), ctest works and the tests pass (with the exception of the zoltan tests im writing about here). what can I do to check and make sure my tests are going to be found and used the first time?
Thanks,
_Jason
-----Original Message-----
From: Devine, Karen D [mailto:kddevin at sandia.gov]
Sent: Thursday, October 15, 2009 2:34 PM
To: Jason Slemons; trilinos-users at software.sandia.gov
Subject: Re: [Trilinos-Users] zoltan tests.
1. Regarding the <mca issue....in your output below, the zoltan test is
checking to see whether or not it is a valid option. Clearly it isn't for
aprun, so the test decides not to use it; the output
"24: DEBUG: mpiexec /usr/bin/aprun" shows that only your aprun command is
actually being used. So that part of the output looks fine to me. The
problem with the tests not running appears to be, rather, that the test
directories are not being copied, which I suspect is similar to your
ctest_zoltan.pl problem.
2. I'd need to see your configuration file to try to understand why
ctest_zoltan.pl is not being copied for you as it is for me when I build.
The file packages/zoltan/test/CMakeLists.txt instructs Cmake to copy
ctest_zoltan.pl to the appropriate place in the build directory. It also
creates, in the build directory, the subdirectory
packages/zoltan/test/CMakeFiles/Zoltan_ZDriveMainCopyFiles.dir. Do you have
that directory in your build directory? The following script works for me;
how does yours differ?
cmake \
-D
CMAKE_INSTALL_PREFIX:FILEPATH="/Users/kddevin/code/Trilinos_v10/Obj_cmake" \
-D Trilinos_ENABLE_Fortran:BOOL=OFF \
-D TPL_ENABLE_MPI:BOOL=ON \
-D CMAKE_C_FLAGS:STRING="-m64 -g" \
-D CMAKE_CXX_FLAGS:STRING="-m64 -g" \
-D CMAKE_VERBOSE_MAKEFILE:BOOL=ON \
-D MPIEXEC_MAX_NUMPROCS:STRING=11 \
-D Trilinos_ENABLE_ALL_PACKAGES:BOOL=OFF \
-D Trilinos_ENABLE_EXAMPLES:BOOL=ON \
-D Trilinos_VERBOSE_CONFIGURE:BOOL=ON \
-D Trilinos_ENABLE_Zoltan:BOOL=ON \
-D Trilinos_EXTRA_LINK_FLAGS:STRING="-framework vecLib" \
-D Zoltan_ENABLE_EXAMPLES:BOOL=ON \
-D Zoltan_ENABLE_TESTS:BOOL=ON \
-D Zoltan_ENABLE_ParMETIS:BOOL=ON \
-D ParMETIS_INCLUDE_DIRS:FILEPATH="/Users/kddevin/code/ParMETIS3_1" \
-D ParMETIS_LIBRARY_DIRS:FILEPATH="/Users/kddevin/code/ParMETIS3_1" \
.. |& tee OUTPUT.CMAKE
On 10/15/09 12:39 PM, "Jason Slemons" <slemons at cray.com> wrote:
> I had a number of tests fail for the same reason in zoltan(see the output of
> ctest VV in the attachment), it appears that ctest_zoltan.pl is in the wrong
> place. I am using pgi 9.0.4, and I typed ctest from my build
> directory(pgi.BUILD; which is a child of trilinos-10.0-Source dir). Here¹s is
> the error:
>
> test 24
> Start 24: Zoltan_ch_simple_zoltan_parallel
> 24: Test command: /opt/cmake/2.8.0-rc3/bin/cmake -DTEST_CONFIG= -P
> /lus/nid00008/slemo\
> ns/trilinos/trilinos-10.0-Source/pgi.BUILD/packages/zoltan/test/ch_simple/Zolt
> an_ch_si\
> mple_zoltan_parallel.cmake
> 24: Test timeout computed to be: 60
> 24:
> 24:
> XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX
> XX
> 24:
> 24: Advanced Test: Zoltan_ch_simple_zoltan_parallel
> 24:
> 24: Running test commands: TEST_0
> 24:
> 24:
> ==============================================================================
> ==
> 24:
> 24: TEST_0
> 24:
> 24: Running: "/usr/bin/perl" "../ctest_zoltan.pl" "4" "Zoltan" "1"
> "/usr/bin/aprun"
> 24:
> 24:
> ------------------------------------------------------------------------------
> --
> 24:
> 24: Can't open perl script "../ctest_zoltan.pl": No such file or directory
> 24:
> 24:
> ------------------------------------------------------------------------------
> --
> 24:
> 24: TEST_0: Return code = 2
> 24: TEST_0: Pass criteria = Return code
> 24: TEST_0: Result = FAILED
> 24:
> 24:
> ==============================================================================
> ==
> 24:
> 24: OVERALL FINAL RESULT: TEST FAILED
> 24:
> 24:
> XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX
> XX
> 24:
> 24/39 Test #24: Zoltan_ch_simple_zoltan_parallel ........***Failed Error
> regular expr\
> ession found in output. Regex=[FAILED] 0.10 sec
>
>
> So I put a copy of ctest_zoltan.pl in
> Otrilinos-10.0-Source/pgi.BUILD/packages/zoltan/test/¹ (as opposed to
> Otrilinos-10.0-Source/packages/zoltan/test/¹ which I believe is where it was)
> and now it Opasses¹(though essentially fails I think) with a strange error
> about having passed unrecognized options Omca¹ to aprun. The output is below:
>
> 24: Test command: /opt/cmake/2.8.0-rc3/bin/cmake -DTEST_CONFIG= -P
> /lus/nid00008/slemo\
> ns/trilinos/trilinos-10.0-Source/pgi.BUILD/packages/zoltan/test/ch_simple/Zolt
> an_ch_si\
> mple_zoltan_parallel.cmake
> 24: Test timeout computed to be: 60
> 24:
> 24:
> XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX
> XX
> 24:
> 24: Advanced Test: Zoltan_ch_simple_zoltan_parallel
> 24:
> 24: Running test commands: TEST_0
> 24:
> 24:
> ==============================================================================
> ==
> 24:
> 24: TEST_0
> 24:
> 24: Running: "/usr/bin/perl" "../ctest_zoltan.pl" "4" "Zoltan" "1"
> "/usr/bin/aprun"
> 24:
> 24:
> ------------------------------------------------------------------------------
> --
> 24:
> 24: CTEST_FULL_OUTPUT
> 24: 4Zoltan1/usr/bin/aprun
> 24: DEBUG: package Zoltan
> 24: /usr/bin/aprun: unrecognized option `--mca'
> 24: Usage: aprun [-D level][-L node_list][-N pes][-S pes][-T]
> 24: [-a arch][-b][-cc cpu_list][-cp file][-F flag][-q][-m size]
> 24: [-P wfd[,rfd]][-sl numa_node_list][-sn numa_nodes][-ss]
> 24: [-t sec][-d depth] -n width command [options...]
> 24: [: -n width cmd2 [: -n width cmd3 [: ...] ] ]
> 24:
> 24: -a arch Architecture type (e.g. xt or x2)
> 24: -b Bypass application transfer to compute node
> 24: -cc cpu_list Cpu binding list or keyword
> 24: ([cpu#[,cpu# | cpu1-cpu2] | x]...] | keyword)
> 24: -cp file Cpu binding placement filename
> 24: -D level Debug level (1, 2, 4, 8, and 15)
> 24: -d depth Depth of each processor (number of threads)
> 24: -F flag exclusive or share node resources flag
> 24: -L node_list Manual placement list (node[,node | node1-node2]...)
> 24: -m size Memory limit in megabytes
> 24: (default node memory/number of processors)
> 24: K|M|G suffix supported (16 == 16M == 16 megabytes)
> 24: Add an 'h' suffix to request per PE huge page memory
> 24: (XT only)
> 24: Add an 's' to the 'h' suffix to make the per PE huge
> 24: page memory size strict (required) (XT only)
> 24: -N pes PEs per node
> 24: -n width Number of processors needed
> 24: -P pipes Write[,read] pipes (not applicable for general use)
> 24: -q Quiet mode; suppress aprun non-fatal messages
> 24: -S pes PEs per NUMA node
> 24: -sl numa_node_list List of NUMA nodes to use
> 24: (numa_node[,numa_node | numa_node1-numa_node2]...)
> 24: -sn numa_nodes Number of NUMA nodes needed per compute node
> 24: -ss Strict memory containment per NUMA node
> 24: -T Use synchronous TTY
> 24: -t sec Time limit in seconds (default unlimited)
> 24: : Separate binaries for MPMD mode
> 24: (Multiple Program, Multiple Data)
> 24: DEBUG: mpiexec /usr/bin/aprun
> 24: DEBUG Dir
> /lus/nid00008/slemons/trilinos/trilinos-10.0-Source/pgi.BUILD/packages/\
> zoltan/test/ch_simple dirname simple
> 24: DEBUG Outfilebase: ; Dropbase:
> 24: Test simple: 0 out of 0 tests PASSED.
> 24:
> 24:
> ------------------------------------------------------------------------------
> --
> 24:
> 24: TEST_0: Return code = 0
> 24: TEST_0: Pass criteria = Return code
> 24: TEST_0: Result = PASSED
> 24:
> ==============================================================================
> ==
> 24:
> 24: OVERALL FINAL RESULT: TEST PASSED
> 24:
> 24:
> XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX
> XX
> 24:
> 24/39 Test #24: Zoltan_ch_simple_zoltan_parallel ........ Passed 0.19 sec
>
> Like I said these are my only zoltan failures. I think ctest_zoltan should
> just be moved to the right place, but I also have no idea where the Omca¹
> option is coming from to aprun.
>
> Has anyone seen this or know how to fix the Omca¹ problem?
> Best,
> -Jason
>
>
More information about the Trilinos-Users
mailing list