Score-P/7.1-cpeCray-22.08 (Score-P-7.1-cpeCray-22.08.eb)
To access module help and find out for which stacks and partitions the module is
installed, use module spider Score-P/7.1-cpeCray-22.08.
EasyConfig:
easyblock = 'Bundle'
local_zlib_version = '1.2.12'
local_CubeLib_version = '4.6'
local_CubeWriter_version = '4.6'
local_OPARI2_version = '2.0.6'
local_OTF2_version = '2.3'
local_libunwind_version = '1.6.2'
local_CubeLib_checksum = '36eaffa7688db8b9304c9e48ca5dc4edc2cb66538aaf48657b9b5ccd7979385b'
local_CubeWriter_checksum = '99fe58ce7ab13061ebfbc360aedaecc28099a30636c5269a42c0cbaf57149aa8'
local_OPARI2_checksum = '55972289ce66080bb48622110c3189a36e88a12917635f049b37685b9d3bbcb0'
local_OTF2_checksum = '36957428d37c40d35b6b45208f050fb5cfe23c54e874189778a24b0e9219c7e3'
local_ScoreP_checksum = '98dea497982001fb82da3429ca55669b2917a0858c71abe2cfe7cd113381f1f7'
name = 'Score-P'
version = '7.1'
homepage = 'https://www.score-p.org'
whatis = [
"Description: Score-P measurement infrastructure is a highly scalable and easy-to-use "
"tool suite for profiling, event tracing, and online analysis of HPC "
"applications."
]
description = """
The goal of Score-P is to simplify the analysis of the behavior of high
performance computing software and to allow the developers of such software to
find out where and why performance problems arise, where bottlenecks may be
expected and where their codes offer room for further improvements with respect
to the run time. A number of tools have been around to help in this respect, but
typically each of these tools has only handled a certain subset of the questions
of interest. A software developer who wanted to have a complete picture of his
code therefore was required to use a multitude of programs to obtain the desired
information. Most of these utilities work along similar principles. The first
step is usually an instrumentation of the code to be investigated. Next, the
instrumented programs are executed and write out (often very large amounts of)
performance data. These data are then finally graphically displayed and analyzed
after the end of the program run. In certain special cases, the visualization
and analysis of the program behavior is also done while the program is running.
Score-P supports the following parallel programming paradigms:
- Multi-process paradigms:
- MPI
- SHMEM
- Thread-parallel paradigms:
- OpenMP
- Pthreads
And possible combinations from these including simple serial programs.
This module is a bundle that includes Score-P as well as its depedencies:
CubeLib, CubeWriter, OPARI2 and OTF2.
"""
usage = """
The first step is to instruments your code. This is done by using the Score-P
command that needs to be prefixed to all the compile and link commands usually
employed to build the application. Thus, an application executable app that is
normally generated from the two source files src1.f90 and src2.f90 via the
command:
ftn src1.f90 src2.f90 -o app
will now be built by:
scorep ftn src1.f90 src2.f90 -o app
It's not necessary to use the instrumenter all source files. You can for example
only include those containing MPI code. However, when invoking the linker, the
instrumenter must always be used.
Once your application has been instrumented, you can start a profiling
experiment by setting the SCOREP_ENABLE_PROFILING environment variable to true
and disabling tracing by setting SCOREP_ENABLE_TRACING to false. In addition you
can define the memory that can be used by Score-P as well as the output
directory where to store the profiling data.
export SCOREP_ENABLE_PROFILING=true
export SCOREP_ENABLE_TRACING=false
export SCOREP_TOTAL_MEMORY=100M
export SCOREP_EXPERIMENT_DIRECTORY=<output-dir>
After running your instrumented application, you can use the scorep-score tool
to obtain insights into the application execution and potential execution
overhead due to Score-P
scorep-score <output-dir>/profile.cubex
For a more detailed report you can use the -r option.
scorep-score -r <output-dir>/profile.cubex
As a second measurement step, you can perform a tracing experiment by setting
the value of the SCOREP_ENABLE_TRACING environment variable to true.
export SCOREP_ENABLE_TRACING=true
export SCOREP_ENABLE_PROFILING=false
export SCOREP_TOTAL_MEMORY=274M
export SCOREP_EXPERIMENT_DIRECTORY=<output-dir>
The optimal memory for the tracing experiment can be determined using data from
a previous profiling experiment and the scorep-score tool:
$ scorep-score <output-dir>/profile.cubex
(hint: When tracing set SCOREP_TOTAL_MEMORY=274MB to avoid intermediate
flushes or reduce requirements using USR regions filters.)
After the execution of your application, you can use the trace file located at
<output-dir>/traces.otf2. This file can be visualized using tools like Vampir.
It may be desirable to reduce the amount of generated tracing data in order to
minimize overheads. This can be done by filtering out small routines that are
called often. You are invited to read the Score-P user manual for more
information.
*** Note for Fortran Code compiled with the Cray compiler ***
- Score-P may instrument correctly OpenMP code if the -homp flag is used to
enable OpenMP. Use the '-fopenmp' flag to enable OpenMP instead.
- Invalid pointer errors can occur when running code compiled with the Cray
Fortran compiler. The cause of this error is the use of the tcmalloc
library for memory allocation. tcmalloc is used by default when producing
dynamically linked executables, hence the error. Use the flag
'-hsystem_alloc' when compiling in order to get rid of these errors.
"""
docurls = [
'https://perftools.pages.jsc.fz-juelich.de/cicd/scorep/tags/scorep-7.1/html/'
]
software_license_urls = [
'https://scalasca.org/scalasca/front_content.php?idart=1094',
'https://opensource.org/licenses/BSD-3-Clause'
]
toolchain = {'name': 'cpeCray', 'version': '22.08'}
toolchainopts = {'openmp': True, 'usempi': True, 'pic': True}
builddependencies = [
('buildtools', '%(toolchain_version)s', '', True),
]
dependencies = [
('zlib', local_zlib_version),
('libunwind', local_libunwind_version),
('papi', EXTERNAL_MODULE),
('cray-openshmemx', EXTERNAL_MODULE),
]
default_easyblock = 'ConfigureMake'
local_ScoreP_configopts = ' --with-machine-name=LUMI --enable-shared '
local_ScoreP_configopts += ' --with-libunwind=$EBROOTLIBUNWIND '
local_ScoreP_configopts += ' --with-cubelib=%(installdir)s/bin '
local_ScoreP_configopts += ' --with-cubew=%(installdir)s/bin '
local_ScoreP_configopts += ' --with-otf2=%(installdir)s/bin '
local_ScoreP_configopts += ' --with-opari2=%(installdir)s/bin '
local_ScoreP_configopts += ' --with-papi-header=$CRAY_PAPI_PREFIX/include --with-papi-lib=$CRAY_PAPI_PREFIX/lib '
components = [
('CubeLib', local_CubeLib_version, {
'source_urls' : ['https://apps.fz-juelich.de/scalasca/releases/cube/%(version_major_minor)s/dist'],
'sources' : [SOURCELOWER_TAR_GZ],
'checksums' : [local_CubeLib_checksum],
'configopts' : '--enable-shared',
'start_dir' : '%(namelower)s-%(version)s',
}),
('CubeWriter', local_CubeWriter_version, {
'source_urls' : ['https://apps.fz-juelich.de/scalasca/releases/cube/%(version_major_minor)s/dist'],
'sources' : ['cubew-%(version)s.tar.gz'],
'checksums' : [local_CubeWriter_checksum],
'start_dir' : 'cubew-%(version)s',
}),
('OPARI2', local_OPARI2_version, {
'source_urls' : ['http://perftools.pages.jsc.fz-juelich.de/cicd/%(namelower)s/tags/%(namelower)s-%(version)s'],
'sources' : [SOURCELOWER_TAR_GZ],
'checksums' : [local_OPARI2_checksum],
'start_dir' : '%(namelower)s-%(version)s',
}),
('OTF2', local_OTF2_version, {
'source_urls' : ['http://perftools.pages.jsc.fz-juelich.de/cicd/%(namelower)s/tags/%(namelower)s-%(version)s'],
'sources' : [SOURCELOWER_TAR_GZ],
'checksums' : [local_OTF2_checksum],
'configopts' : '--enable-shared',
'start_dir' : '%(namelower)s-%(version)s',
}),
(name, version, {
'easyblock' : 'ConfigureMake',
'source_urls' : ['http://perftools.pages.jsc.fz-juelich.de/cicd/scorep/tags/scorep-%(version)s'],
'sources' : ['scorep-%(version)s.tar.gz'],
'checksums' : [local_ScoreP_checksum],
'configopts' : local_ScoreP_configopts,
'start_dir' : 'scorep-%(version)s',
})
]
sanity_check_paths = {
'files': ['bin/scorep', 'include/scorep/SCOREP_User.h'],
'dirs': [],
}
sanity_check_commands = ['scorep-config --help']
modextravars = {
'EBROOTCUBELIB': '%(installdir)s',
'EBVERSIONCUBELIB': local_CubeLib_version,
'EBROOTCUBEWRITER': '%(installdir)s',
'EBVERSIONCUBEWRITER': local_CubeWriter_version,
'EBROOTOPARI2': '%(installdir)s',
'EBVERSIONOPARI2': local_OPARI2_version,
'EBROOTOTF2': '%(installdir)s',
'EBVERSIONOTF2': local_OTF2_version,
}
modluafooter = """
extensions( "CubeLib/%(cubelib)s, CubeWriter/%(cubewriter)s, " ..
"OPARI2/%(opari2)s, OTF2/%(otf2)s")
""" % {
'cubelib' : local_CubeLib_version,
'cubewriter' : local_CubeWriter_version,
'opari2' : local_OPARI2_version,
'otf2' : local_OTF2_version,
}
moduleclass = 'perf'