def attributes():
from taucmdr.model.project import Project
from taucmdr.model.target import Target
from taucmdr.model.measurement import Measurement
from taucmdr.cf.platforms import DARWIN, HOST_OS, CRAY_CNL
def _encourage_except_baseline(lhs, lhs_attr, lhs_value, rhs):
if isinstance(rhs, Measurement):
if not rhs['baseline']:
lhs_name = lhs.name.lower()
rhs_name = rhs.name.lower()
LOGGER.warning("%s = %s in %s recommends True in %s", lhs_attr,
lhs_value, lhs_name, rhs_name)
return {
'projects': {
'collection': Project,
'via': 'applications',
'description': 'projects using this application'
},
'name': {
'primary_key': True,
'type': 'string',
'description': 'application configuration name',
'unique': True
},
'openmp': {
'type': 'boolean',
'description': 'application uses OpenMP',
'default': False,
'argparse': {
'flags': ('--openmp', )
},
'rebuild_required': True
},
'pthreads': {
'type': 'boolean',
'description': 'application uses pthreads',
'default': False,
'argparse': {
'flags': ('--pthreads', )
},
'rebuild_required': True
},
'python': {
'type': 'boolean',
'description': 'application uses Python',
'default': False,
'argparse': {
'flags': ('--python', )
},
'rebuild_required': True
},
'tbb': {
'type': 'boolean',
'description': 'application uses Thread Building Blocks (TBB)',
'default': False,
'argparse': {
'flags': ('--tbb', )
},
'rebuild_required': True
},
'mpi': {
'type': 'boolean',
'default': False,
'description': 'application uses MPI',
'argparse': {
'flags': ('--mpi', )
},
'compat': {
True: _encourage_except_baseline
},
'rebuild_required': True
},
'caf': {
'type': 'boolean',
'default': False,
'description': 'application uses Coarray Fortran',
'argparse': {
'flags': ('--caf', )
},
'compat': {
True: (_encourage_except_baseline,
Measurement.require('source_inst', 'never'))
},
'rebuild_required': True
},
'cuda': {
'type': 'boolean',
'default': False,
'description': 'application uses NVIDIA CUDA',
'argparse': {
'flags': ('--cuda', )
},
'compat': {
True: Target.require('cuda_toolkit')
},
'rebuild_required': True
},
'opencl': {
'type': 'boolean',
'default': False,
'description': 'application uses OpenCL',
'argparse': {
'flags': ('--opencl', )
},
'compat': {
True:
(Target.require('cuda_toolkit'), _encourage_except_baseline)
},
'rebuild_required': True
},
'shmem': {
'type': 'boolean',
'default': False,
'description': 'application uses SHMEM',
'argparse': {
'flags': ('--shmem', )
},
'rebuild_required': True
},
'mpc': {
'type': 'boolean',
'default': False,
'description': 'application uses MPC',
'argparse': {
'flags': ('--mpc', )
},
'rebuild_required': True
},
'max_threads': {
'type': 'integer',
'description': 'maximum number of threads',
'argparse': {
'flags': ('--max-threads', ),
'metavar': 'threads',
'nargs': '?',
'const': 25
},
'rebuild_required': True
},
'linkage': {
'type': 'string',
'default': 'static' if HOST_OS is CRAY_CNL else 'dynamic',
'description': "application linkage",
'argparse': {
'flags': ('--linkage', ),
'metavar': '<linkage>',
'choices': ('static', 'dynamic')
},
'compat': {
'static': Target.exclude('host_os', DARWIN)
},
'rebuild_required': True
},
}
def attributes():
from taucmdr.model.project import Project
from taucmdr.model.target import Target
from taucmdr.model.measurement import Measurement
from taucmdr.cf.platforms import DARWIN, HOST_OS, CRAY_CNL
return {
'projects': {
'collection': Project,
'via': 'applications',
'description': 'projects using this application'
},
'name': {
'primary_key': True,
'type': 'string',
'description': 'application configuration name',
'unique': True
},
'openmp': {
'type': 'boolean',
'description': 'application uses OpenMP',
'default': False,
'argparse': {'flags': ('--openmp',)},
'rebuild_required': True
},
'pthreads': {
'type': 'boolean',
'description': 'application uses pthreads',
'default': False,
'argparse': {'flags': ('--pthreads',)},
'rebuild_required': True
},
'tbb': {
'type': 'boolean',
'description': 'application uses Thread Building Blocks (TBB)',
'default': False,
'argparse': {'flags': ('--tbb',)},
'rebuild_required': True
},
'mpi': {
'type': 'boolean',
'default': False,
'description': 'application uses MPI',
'argparse': {'flags': ('--mpi',)},
'compat': {True: Measurement.require('mpi', True)},
'rebuild_required': True
},
'cuda': {
'type': 'boolean',
'default': False,
'description': 'application uses NVIDIA CUDA',
'argparse': {'flags': ('--cuda',)},
'compat': {True: Target.require('cuda_toolkit')},
'rebuild_required': True
},
'opencl': {
'type': 'boolean',
'default': False,
'description': 'application uses OpenCL',
'argparse': {'flags': ('--opencl',)},
'compat': {True: (Target.require('cuda_toolkit'),
Measurement.encourage('opencl', True))},
'rebuild_required': True
},
'shmem': {
'type': 'boolean',
'default': False,
'description': 'application uses SHMEM',
'argparse': {'flags': ('--shmem',)},
'rebuild_required': True
},
'mpc': {
'type': 'boolean',
'default': False,
'description': 'application uses MPC',
'argparse': {'flags': ('--mpc',)},
'rebuild_required': True
},
'select_file': {
'type': 'string',
'description': 'specify selective instrumentation file',
'argparse': {'flags': ('--select-file',),
'metavar': 'path'},
'compat': {True: Measurement.exclude('source_inst', 'never')},
'rebuild_required': True
},
'linkage': {
'type': 'string',
'default': 'static' if HOST_OS is CRAY_CNL else 'dynamic',
'description': "application linkage",
'argparse': {'flags': ('--linkage',),
'metavar': '<linkage>',
'choices': ('static', 'dynamic')},
'compat': {'static': Target.exclude('host_os', DARWIN)},
'rebuild_required': True
},
}
def attributes():
from taucmdr.model.project import Project
from taucmdr.model.target import Target
from taucmdr.model.measurement import Measurement
from taucmdr.cf.platforms import DARWIN, HOST_OS, CRAY_CNL
return {
'projects': {
'collection': Project,
'via': 'applications',
'description': 'projects using this application'
},
'name': {
'primary_key': True,
'type': 'string',
'description': 'application configuration name',
'unique': True
},
'openmp': {
'type': 'boolean',
'description': 'application uses OpenMP',
'default': False,
'argparse': {
'flags': ('--openmp', )
},
'rebuild_required': True
},
'pthreads': {
'type': 'boolean',
'description': 'application uses pthreads',
'default': False,
'argparse': {
'flags': ('--pthreads', )
},
'rebuild_required': True
},
'tbb': {
'type': 'boolean',
'description': 'application uses Thread Building Blocks (TBB)',
'default': False,
'argparse': {
'flags': ('--tbb', )
},
'rebuild_required': True
},
'mpi': {
'type': 'boolean',
'default': False,
'description': 'application uses MPI',
'argparse': {
'flags': ('--mpi', )
},
'compat': {
True: Measurement.encourage('mpi', True)
},
'rebuild_required': True
},
'cuda': {
'type': 'boolean',
'default': False,
'description': 'application uses NVIDIA CUDA',
'argparse': {
'flags': ('--cuda', )
},
'compat': {
True: Target.require('cuda_toolkit')
},
'rebuild_required': True
},
'opencl': {
'type': 'boolean',
'default': False,
'description': 'application uses OpenCL',
'argparse': {
'flags': ('--opencl', )
},
'compat': {
True: (Target.require('cuda_toolkit'),
Measurement.encourage('opencl', True))
},
'rebuild_required': True
},
'shmem': {
'type': 'boolean',
'default': False,
'description': 'application uses SHMEM',
'argparse': {
'flags': ('--shmem', )
},
'rebuild_required': True
},
'mpc': {
'type': 'boolean',
'default': False,
'description': 'application uses MPC',
'argparse': {
'flags': ('--mpc', )
},
'rebuild_required': True
},
'select_file': {
'type': 'string',
'description': 'specify selective instrumentation file',
'argparse': {
'flags': ('--select-file', ),
'metavar': 'path'
},
'compat': {
bool: (Measurement.exclude('source_inst', 'never'),
Application.discourage('linkage', 'static'))
},
'rebuild_required': True
},
'linkage': {
'type': 'string',
'default': 'static' if HOST_OS is CRAY_CNL else 'dynamic',
'description': "application linkage",
'argparse': {
'flags': ('--linkage', ),
'metavar': '<linkage>',
'choices': ('static', 'dynamic')
},
'compat': {
'static': Target.exclude('host_os', DARWIN)
},
'rebuild_required': True
},
}
def attributes():
"""Construct attributes dictionary for the measurement model.
We build the attributes in a function so that classes like ``taucmdr.module.project.Project`` are
fully initialized and usable in the returned dictionary.
Returns:
dict: Attributes dictionary.
"""
from taucmdr.model.project import Project
from taucmdr.model.target import Target
from taucmdr.model.application import Application
from taucmdr.cf.platforms import HOST_OS, DARWIN, IBM_CNK
from taucmdr.cf.compiler.mpi import MPI_CC, MPI_CXX, MPI_FC
from taucmdr.cf.compiler.shmem import SHMEM_CC, SHMEM_CXX, SHMEM_FC
def _merged_profile_compat(lhs, lhs_attr, lhs_value, rhs):
if isinstance(rhs, Application):
if not (rhs['mpi'] or rhs['shmem']):
lhs_name = lhs.name.lower()
rhs_name = rhs.name.lower()
raise IncompatibleRecordError(
"%s = %s in %s requires either mpi = True or shmem = True in %s"
% (lhs_attr, lhs_value, lhs_name, rhs_name))
def _discourage_callpath(lhs, lhs_attr, lhs_value, rhs):
if isinstance(rhs, Measurement):
if rhs.get('callpath', 0) > 0:
lhs_name = lhs.name.lower()
rhs_name = rhs.name.lower()
LOGGER.warning(
"%s = %s in %s recommends against callpath > 0 in %s",
lhs_attr, lhs_value, lhs_name, rhs_name)
def _ensure_instrumented(lhs, lhs_attr, lhs_value, rhs):
if isinstance(rhs, Measurement):
if rhs.get('source_inst') == 'never' and rhs.get(
'compiler_inst') == 'never':
lhs_name = lhs.name.lower()
rhs_name = rhs.name.lower()
raise IncompatibleRecordError(
"%s = %s in %s requires source_inst and compiler_inst are not both 'never' in %s"
% (lhs_attr, lhs_value, lhs_name, rhs_name))
return {
'projects': {
'collection': Project,
'via': 'measurements',
'description': "projects using this measurement"
},
'name': {
'primary_key': True,
'type': 'string',
'unique': True,
'description': "measurement configuration name",
},
'baseline': {
'type': 'boolean',
'default': False,
'description':
"completely disable all instrumentation and measure wall clock time via the OS",
'argparse': {
'flags': ('--baseline', ),
'group': 'instrumentation'
},
'compat': {
True: (Measurement.require('profile', 'tau'),
Measurement.require('trace', 'none'))
}
},
'profile': {
'type': 'string',
'default': 'tau',
'description': "generate application profiles",
'argparse': {
'flags': ('--profile', ),
'group': 'output format',
'metavar': '<format>',
'nargs': '?',
'choices': ('tau', 'merged', 'cubex', 'none'),
'const': 'tau'
},
'compat': {
'cubex': Target.exclude('scorep_source', None),
'merged': _merged_profile_compat
},
'rebuild_required': True
},
'trace': {
'type': 'string',
'default': 'none',
'description': "generate application traces",
'argparse': {
'flags': ('--trace', ),
'group': 'output format',
'metavar': '<format>',
'nargs': '?',
'choices': ('slog2', 'otf2', 'none'),
'const': 'otf2'
},
'compat': {
'otf2': Target.exclude('libotf2_source', None),
lambda x: x != 'none': _discourage_callpath
},
'rebuild_required': True
},
'sample': {
'type': 'boolean',
'default': HOST_OS not in (DARWIN, IBM_CNK),
'description':
"use event-based sampling to gather performance data",
'argparse': {
'flags': ('--sample', ),
'group': 'instrumentation'
},
'compat': {
True:
(Target.require('binutils_source'),
Target.exclude('binutils_source',
None), Target.encourage('libunwind_source'),
Target.discourage('libunwind_source',
None), Target.exclude('host_os', DARWIN),
Measurement.exclude('baseline', True))
}
},
'source_inst': {
'type': 'string',
'default': 'never' if HOST_OS is not DARWIN else 'automatic',
'description':
"use hooks inserted into the application source code to gather performance data",
'argparse': {
'flags': ('--source-inst', ),
'group': 'instrumentation',
'metavar': 'mode',
'nargs': '?',
'choices': ('automatic', 'manual', 'never'),
'const': 'automatic'
},
'compat': {
lambda x: x in ('automatic', 'manual'):
(Target.exclude('pdt_source',
None), Measurement.exclude('baseline', True))
},
'rebuild_required': True
},
'compiler_inst': {
'type': 'string',
'default': 'never',
'description':
"use compiler-generated callbacks to gather performance data",
'argparse': {
'flags': ('--compiler-inst', ),
'group': 'instrumentation',
'metavar': 'mode',
'nargs': '?',
'choices': ('always', 'fallback', 'never'),
'const': 'always'
},
'compat': {
lambda x: x in ('always', 'fallback'):
(Target.require('binutils_source'),
Target.exclude('binutils_source',
None), Target.require('libunwind_source'),
Target.exclude('libunwind_source',
None), Target.exclude('host_os', DARWIN),
Measurement.exclude('baseline', True))
},
'rebuild_required': True
},
'mpi': {
'type': 'boolean',
'default': False,
'description':
'use MPI library wrapper to measure time spent in MPI methods',
'argparse': {
'flags': ('--mpi', )
},
'compat': {
True: (Target.require(MPI_CC.keyword),
Target.require(MPI_CXX.keyword),
Target.require(MPI_FC.keyword),
Application.require('mpi', True),
Measurement.exclude('baseline', True))
},
'rebuild_required': True
},
'openmp': {
'type': 'string',
'default': 'ignore',
'description':
'use specified library to measure time spent in OpenMP directives',
'argparse': {
'flags': ('--openmp', ),
'metavar': 'library',
'choices': ('ignore', 'opari', 'ompt'),
'nargs': '?',
'const': 'ompt'
},
'compat': {
'opari': (Application.require('openmp', True),
Measurement.exclude('baseline', True)),
'ompt': (Application.require('openmp', True),
Measurement.exclude('baseline', True))
},
'rebuild_required': True
},
'cuda': {
'type': 'boolean',
'default': False,
'description': 'measure cuda events via the CUPTI interface',
'argparse': {
'flags': ('--cuda', )
},
'compat': {
True: (Target.require('cuda_toolkit'),
Application.require('cuda', True),
Measurement.exclude('baseline', True))
},
},
'shmem': {
'type': 'boolean',
'default': False,
'description':
'use SHMEM library wrapper to measure time spent in SHMEM methods',
'argparse': {
'flags': ('--shmem', )
},
'compat': {
True: (Target.require(SHMEM_CC.keyword),
Target.require(SHMEM_CXX.keyword),
Target.require(SHMEM_FC.keyword),
Application.require('shmem', True),
Measurement.exclude('baseline', True))
},
'rebuild_required': True
},
'opencl': {
'type': 'boolean',
'default': False,
'description': 'measure OpenCL events',
'argparse': {
'flags': ('--opencl', )
},
'compat': {
True: (Target.require('cuda_toolkit'),
Application.require('opencl', True),
Measurement.exclude('baseline', True))
},
},
'callpath': {
'type': 'integer',
'default': 100,
'description': 'maximum depth for callpath recording',
'argparse': {
'flags': ('--callpath', ),
'group': 'data',
'metavar': 'depth',
'nargs': '?',
'const': 100
}
},
'unwind_depth': {
'type': 'integer',
'default': 0,
'description': 'Record callstack to specified depth',
'argparse': {
'flags': ('--unwind-depth', ),
'group': 'data',
'metavar': 'depth',
'nargs': '?',
'const': 10
},
'compat': {
(lambda unwind: unwind > 0):
Measurement.exclude('sample', False)
}
},
'io': {
'type': 'boolean',
'default': False,
'description': 'measure time spent in POSIX I/O calls',
'argparse': {
'flags': ('--io', )
},
'compat': {
True: Measurement.exclude('baseline', True)
},
},
'heap_usage': {
'type': 'boolean',
'default': False,
'description': 'measure heap memory usage',
'argparse': {
'flags': ('--heap-usage', ),
'group': 'memory'
},
'compat': {
True: Measurement.exclude('baseline', True)
},
},
'system_load': {
'type': 'boolean',
'default': False,
'description': 'measure system load',
'argparse': {
'flags': ('--system-load', ),
'group': 'data'
},
'compat': {
True: Measurement.exclude('baseline', True)
},
},
'memory_alloc': {
'type': 'boolean',
'default': False,
'description':
'record memory allocation/deallocation events and detect leaks',
'argparse': {
'flags': ('--memory-alloc', ),
'group': 'memory'
},
'compat': {
True: Measurement.exclude('baseline', True)
},
},
'metrics': {
'type': 'array',
'default': ['TIME'],
'description':
'a space separated list of performance metrics to gather, e.g. TIME PAPI_FP_INS',
'argparse': {
'flags': ('--metrics', ),
'group': 'data',
'metavar': '<metric>'
},
'compat': {
lambda metrics: bool(
len([met for met in metrics if 'PAPI' in met])):
(Target.require('papi_source'),
Target.exclude('papi_source', None)),
lambda x: x != ['TIME']:
Measurement.exclude('baseline', True)
},
'rebuild_required': True
},
'keep_inst_files': {
'type': 'boolean',
'default': False,
'description': "don't remove instrumented files after compilation",
'argparse': {
'flags': ('--keep-inst-files', ),
'group': 'instrumentation'
},
'compat': {
True: Measurement.exclude('source_inst', 'never')
}
},
'reuse_inst_files': {
'type': 'boolean',
'default': False,
'description':
'reuse and preserve instrumented files after compilation',
'argparse': {
'flags': ('--reuse-inst-files', ),
'group': 'instrumentation'
},
'compat': {
True: Measurement.exclude('source_inst', 'never')
}
},
'comm_matrix': {
'type': 'boolean',
'default': False,
'description': 'record the point-to-point communication matrix',
'argparse': {
'flags': ('--comm-matrix', )
},
'compat': {
True: Measurement.exclude('baseline', True)
},
},
'throttle': {
'type': 'boolean',
'default': True,
'description': 'throttle lightweight events to reduce overhead',
'argparse': {
'flags': ('--throttle', )
},
},
'throttle_per_call': {
'type': 'integer',
'default': 10,
'description':
'lightweight event duration threshold in microseconds',
'argparse': {
'flags': ('--throttle-per-call', ),
'metavar': 'us',
'nargs': '?',
'const': 10
},
},
'throttle_num_calls': {
'type': 'integer',
'default': 100000,
'description': 'lightweight event call count threshold',
'argparse': {
'flags': ('--throttle-num-calls', ),
'metavar': 'count',
'nargs': '?',
'const': 100000
},
},
'sampling_period': {
'type': 'integer',
'default': 5000,
'description': 'default sampling period in microseconds',
'argparse': {
'flags': ('--sampling-period', ),
'metavar': 'us',
'nargs': '?',
'const': 5000
},
'compat': {
True: (Measurement.require('sample', True))
},
},
'track_memory_footprint': {
'type': 'boolean',
'default': False,
'description': 'track memory footprint',
'argparse': {
'flags': ('--track-memory-footprint', ),
'group': 'memory'
},
'compat': {
True: Measurement.exclude('baseline', True)
},
},
'metadata_merge': {
'type': 'boolean',
'default': True,
'description': 'merge metadata of TAU profiles',
'argparse': {
'flags': ('--metadata-merge', )
},
},
'callsite': {
'type': 'boolean',
'default': False,
'description': 'record event callsites',
'argparse': {
'flags': ('--callsite', )
},
'compat': {
True:
(Target.require('binutils_source'),
Target.exclude('binutils_source',
None), Target.encourage('libunwind_source'),
Target.discourage('libunwind_source',
None), Target.exclude('host_os', DARWIN),
Measurement.exclude('baseline', True))
}
},
'select_file': {
'type': 'string',
'description': 'specify selective instrumentation file',
'argparse': {
'flags': ('--select-file', ),
'metavar': 'path'
},
'compat': {
bool: (_ensure_instrumented,
Application.discourage('linkage', 'static'),
Measurement.discourage('throttle', True))
},
'rebuild_required': True
},
'update_nightly': {
'type': 'boolean',
'default': False,
'description': 'Download newest TAU nightly',
'argparse': {
'flags': ('--update-nightly', )
},
'compat': {
bool: (Target.require('tau_source', 'nightly'))
},
},
'ptts': {
'type': 'boolean',
'default': False,
'description': 'enable support for PTTS',
'argparse': {
'flags': ('--ptts', )
},
},
'ptts_post': {
'type': 'boolean',
'default': False,
'description':
'skip application sampling and post-process existing PTTS sample files',
'argparse': {
'flags': ('--ptts-post', )
},
'compat': {
bool: (Measurement.require('ptts', True))
},
},
'ptts_sample_flags': {
'type': 'string',
'default': '', #None,
'description': 'flags to pass to PTTS sample_ts command',
'argparse': {
'flags': ('--ptts-sample-flags', ),
'metavar': 'sample_flags'
},
'compat': {
bool: (Measurement.require('ptts', True))
},
},
'ptts_restart': {
'type': 'boolean',
'default': False,
'description':
'enable restart support within PTTS, allowing application to continue running and be reinstrumented after stop',
'argparse': {
'flags': ('--ptts-restart', )
},
'compat': {
bool: (Measurement.require('ptts', True))
},
},
'ptts_start': {
'type': 'string',
'default': '', #None,
'description': 'address at which to start a PTTS sampling region',
'argparse': {
'flags': ('--ptts-start', )
},
'compat': {
bool: (Measurement.require('ptts', True))
},
},
'ptts_stop': {
'type': 'string',
'default': '', #None,
'description': 'address at which to stop a PTTS sampling region',
'argparse': {
'flags': ('--ptts-stop', )
},
'compat': {
bool: (Measurement.require('ptts', True))
},
},
'ptts_report_flags': {
'type': 'string',
'default': '', #None,
'description': 'flags to pass to PTTS report_ts command',
'argparse': {
'flags': ('--ptts-report-flags', )
},
'compat': {
bool: (Measurement.require('ptts', True))
},
},
'extra_tau_options': {
'type': 'string',
'description':
'forcibly add to the TAU_OPTIONS environment variable (not recommended)',
'rebuild_on_change': True,
'argparse': {
'flags': ('--extra-tau-options', ),
'nargs': '+',
'metavar': '<option>'
},
'compat': {
bool: (Measurement.discourage('extra_tau_options'),
Measurement.exclude('baseline', True))
}
},
'mpit': {
'type': 'boolean',
'default': False,
'description': 'MPI-T profiling interface',
'argparse': {
'flags': ('--mpit', )
},
},
'force_tau_options': {
'type': 'array',
'description':
"forcibly set the TAU_OPTIONS environment variable (not recommended)",
'rebuild_on_change': True,
'argparse': {
'flags': ('--force-tau-options', ),
'nargs': '+',
'metavar': '<option>'
},
'compat': {
bool: (Measurement.discourage('force_tau_options'),
Measurement.exclude('extra_tau_options'),
Measurement.exclude('baseline', True))
}
},
'tag': {
'type': 'string',
'description': "Narrow search for tags",
'default': '', #None,
'argparse': {
'flags': ('--tag', ),
'nargs': '+',
'metavar': '<tags>'
}
},
'extra_tau_options': {
'type': 'array',
'description':
"append extra options to TAU_OPTIONS environment variable (not recommended)",
'rebuild_on_change': True,
'argparse': {
'flags': ('--extra-tau-options', ),
'nargs': '+',
'metavar': '<option>'
},
'compat': {
bool: (Measurement.discourage('extra_tau_options'),
Measurement.exclude('force_tau_options'),
Measurement.exclude('baseline', True))
}
}
}
def attributes():
"""Construct attributes dictionary for the measurement model.
We build the attributes in a function so that classes like ``taucmdr.module.project.Project`` are
fully initialized and usable in the returned dictionary.
Returns:
dict: Attributes dictionary.
"""
from taucmdr.model.project import Project
from taucmdr.model.target import Target
from taucmdr.model.application import Application
from taucmdr.cf.platforms import HOST_OS, DARWIN, IBM_CNK
from taucmdr.cf.compiler.mpi import MPI_CC, MPI_CXX, MPI_FC
from taucmdr.cf.compiler.shmem import SHMEM_CC, SHMEM_CXX, SHMEM_FC
def _merged_profile_compat(lhs, lhs_attr, lhs_value, rhs):
if isinstance(rhs, Application):
if not (rhs['mpi'] or rhs['shmem']):
lhs_name = lhs.name.lower()
rhs_name = rhs.name.lower()
raise IncompatibleRecordError("%s = %s in %s requires either mpi = True or shmem = True in %s" %
(lhs_attr, lhs_value, lhs_name, rhs_name))
def _discourage_callpath(lhs, lhs_attr, lhs_value, rhs):
if isinstance(rhs, Measurement):
if rhs.get('callpath', 0) > 0:
lhs_name = lhs.name.lower()
rhs_name = rhs.name.lower()
LOGGER.warning("%s = %s in %s recommends against callpath > 0 in %s",
lhs_attr, lhs_value, lhs_name, rhs_name)
return {
'projects': {
'collection': Project,
'via': 'measurements',
'description': "projects using this measurement"
},
'name': {
'primary_key': True,
'type': 'string',
'unique': True,
'description': "measurement configuration name",
},
'profile': {
'type': 'string',
'default': 'tau',
'description': "generate application profiles",
'argparse': {'flags': ('--profile',),
'group': 'output format',
'metavar': '<format>',
'nargs': '?',
'choices': ('tau', 'merged', 'cubex', 'none'),
'const': 'tau'},
'compat': {'cubex': (Target.exclude('scorep_source', None),
Application.require('mpi', True),
Measurement.require('mpi', True)),
'merged': _merged_profile_compat},
},
'trace': {
'type': 'string',
'default': 'none',
'description': "generate application traces",
'argparse': {'flags': ('--trace',),
'group': 'output format',
'metavar': '<format>',
'nargs': '?',
'choices':('slog2', 'otf2', 'none'),
'const': 'otf2'},
'compat': {'otf2': Target.exclude('libotf2_source', None),
lambda x: x != 'none': _discourage_callpath}
},
'sample': {
'type': 'boolean',
'default': HOST_OS not in (DARWIN, IBM_CNK),
'description': "use event-based sampling to gather performance data",
'argparse': {'flags': ('--sample',),
'group': 'instrumentation'},
'compat': {True: (Target.require('binutils_source'),
Target.exclude('binutils_source', None),
Target.encourage('libunwind_source'),
Target.discourage('libunwind_source', None),
Target.exclude('host_os', DARWIN))}
},
'source_inst': {
'type': 'string',
'default': 'never' if HOST_OS is not DARWIN else 'automatic',
'description': "use hooks inserted into the application source code to gather performance data",
'argparse': {'flags': ('--source-inst',),
'group': 'instrumentation',
'metavar': 'mode',
'nargs': '?',
'choices': ('automatic', 'manual', 'never'),
'const': 'automatic'},
'compat': {'automatic': Target.exclude('pdt_source', None)},
'rebuild_required': True
},
'compiler_inst': {
'type': 'string',
'default': 'never',
'description': "use compiler-generated callbacks to gather performance data",
'argparse': {'flags': ('--compiler-inst',),
'group': 'instrumentation',
'metavar': 'mode',
'nargs': '?',
'choices': ('always', 'fallback', 'never'),
'const': 'always'},
'compat': {lambda x: x in ('always', 'fallback'):
(Target.require('binutils_source'),
Target.exclude('binutils_source', None),
Target.require('libunwind_source'),
Target.exclude('libunwind_source', None),
Target.exclude('host_os', DARWIN))},
'rebuild_required': True
},
'link_only': {
'type': 'boolean',
'default': False,
'description': "don't instrument, only link the TAU library to the application",
'argparse': {'flags': ('--link-only',),
'group': 'instrumentation'},
'rebuild_required': True
},
'mpi': {
'type': 'boolean',
'default': False,
'description': 'use MPI library wrapper to measure time spent in MPI methods',
'argparse': {'flags': ('--mpi',)},
'compat': {True:
(Target.require(MPI_CC.keyword),
Target.require(MPI_CXX.keyword),
Target.require(MPI_FC.keyword))},
'rebuild_required': True
},
'openmp': {
'type': 'string',
'default': 'ignore',
'description': 'use specified library to measure time spent in OpenMP directives',
'argparse': {'flags': ('--openmp',),
'metavar': 'library',
'choices': ('ignore', 'opari', 'ompt')},
'compat': {'opari':
(Application.require('openmp', True),
Measurement.discourage('source_inst', 'never')),
'ompt': Application.require('openmp', True)},
'rebuild_required': True
},
'cuda': {
'type': 'boolean',
'default': False,
'description': 'measure cuda events via the CUPTI interface',
'argparse': {'flags': ('--cuda',)},
'compat': {True: Target.require('cuda_toolkit')},
},
'shmem': {
'type': 'boolean',
'default': False,
'description': 'use SHMEM library wrapper to measure time spent in SHMEM methods',
'argparse': {'flags': ('--shmem',)},
'compat': {True:
(Target.require(SHMEM_CC.keyword),
Target.require(SHMEM_CXX.keyword),
Target.require(SHMEM_FC.keyword))},
'rebuild_required': True
},
'opencl': {
'type': 'boolean',
'default': False,
'description': 'measure OpenCL events',
'argparse': {'flags': ('--opencl',)},
'compat': {True: (Target.require('cuda_toolkit'),
Application.require('opencl'))},
},
'callpath': {
'type': 'integer',
'default': 100,
'description': 'maximum depth for callpath recording',
'argparse': {'flags': ('--callpath',),
'group': 'data',
'metavar': 'depth',
'nargs': '?',
'const': 100}
},
'io': {
'type': 'boolean',
'default': False,
'description': 'measure time spent in POSIX I/O calls',
'argparse': {'flags': ('--io',)},
},
'heap_usage': {
'type': 'boolean',
'default': False,
'description': 'measure heap memory usage',
'argparse': {'flags': ('--heap-usage',),
'group': 'memory'},
},
'memory_alloc': {
'type': 'boolean',
'default': False,
'description': 'record memory allocation/deallocation events and detect leaks',
'argparse': {'flags': ('--memory-alloc',),
'group': 'memory'},
},
'metrics': {
'type': 'array',
'default': ['TIME'],
'description': 'performance metrics to gather, e.g. TIME, PAPI_FP_INS',
'argparse': {'flags': ('--metrics',),
'group': 'data',
'metavar': '<metric>'},
'compat': {lambda metrics: bool(len([met for met in metrics if 'PAPI' in met])):
(Target.require('papi_source'),
Target.exclude('papi_source', None))},
'rebuild_required': True
},
'keep_inst_files': {
'type': 'boolean',
'default': False,
'description': "don't remove instrumented files after compilation",
'argparse': {'flags': ('--keep-inst-files',),
'group': 'instrumentation'},
'compat': {True: Measurement.exclude('source_inst', 'never')}
},
'reuse_inst_files': {
'type': 'boolean',
'default': False,
'description': 'reuse and preserve instrumented files after compilation',
'argparse': {'flags': ('--reuse-inst-files',),
'group': 'instrumentation'},
'compat': {True: Measurement.exclude('source_inst', 'never')}
},
'comm_matrix': {
'type': 'boolean',
'default': False,
'description': 'record the point-to-point communication matrix',
'argparse': {'flags': ('--comm-matrix',)}
},
'throttle': {
'type': 'boolean',
'default': True,
'description': 'throttle lightweight events to reduce overhead',
'argparse': {'flags': ('--throttle',)},
},
'throttle_per_call': {
'type': 'integer',
'default': 10,
'description': 'lightweight event duration threshold in microseconds',
'argparse': {'flags': ('--throttle-per-call',),
'metavar': 'us',
'nargs': '?',
'const': 10},
},
'throttle_num_calls': {
'type': 'integer',
'default': 100000,
'description': 'lightweight event call count threshold',
'argparse': {'flags': ('--throttle-num-calls',),
'metavar': 'count',
'nargs': '?',
'const': 100000},
},
'metadata_merge': {
'type': 'boolean',
'default': True,
'description': 'merge metadata of TAU profiles',
'argparse': {'flags': ('--metadata-merge',)},
},
'callsite': {
'type': 'boolean',
'default': False,
'description': 'record event callsites',
'argparse': {'flags': ('--callsite',)},
'compat': {True: (Target.require('binutils_source'),
Target.exclude('binutils_source', None),
Target.encourage('libunwind_source'),
Target.discourage('libunwind_source', None),
Target.exclude('host_os', DARWIN))}
},
}