def test_appkeroutputparser(datadir):
from akrr.parsers.akrrappkeroutputparser import AppKerOutputParser, total_seconds
import xml.etree.ElementTree as ElementTree
parser = AppKerOutputParser(name='test',
version=1,
description="Test the resource deployment",
url='http://xdmod.buffalo.edu',
measurement_name='test')
parser.add_common_must_have_params_and_stats()
assert len(parser.mustHaveParameters) == 2
parser.add_must_have_parameter("Must have parameter 1")
parser.add_must_have_parameter("Must have parameter 2")
parser.add_must_have_parameter("Must have parameter 3")
parser.add_must_have_statistic("Must have statistic 1")
parser.add_must_have_statistic("Must have statistic 2")
parser.add_must_have_statistic("Must have statistic 3")
assert "Must have parameter 1" in parser.mustHaveParameters
assert "Must have parameter 2" in parser.mustHaveParameters
assert "Must have parameter 3" in parser.mustHaveParameters
assert "Must have statistic 1" in parser.mustHaveStatistics
assert "Must have statistic 2" in parser.mustHaveStatistics
assert "Must have statistic 3" in parser.mustHaveStatistics
# parse common parameters and statistics
parser.parse_common_params_and_stats(str(datadir / 'appstdout'),
str(datadir / 'stdout'),
str(datadir / 'stderr'),
str(datadir / 'gen.info'),
resource_appker_vars={
'resource': {
'name': 'HPC-Cluster'
},
'app': {
'name': 'test'
}
})
# set statistics
if parser.wallClockTime is not None:
parser.set_statistic("Wall Clock Time",
total_seconds(parser.wallClockTime), "Second")
# check resulting xml
xml_text_out = parser.get_xml()
xml_out = ElementTree.fromstring(xml_text_out)
params = xml_out.find(".//parameters")
stats = xml_out.find(".//statistics")
assert len(
params.find(".//parameter[ID='RunEnv:Nodes']").find('value').text) > 5
assert stats.find(".//statistic[ID='Network scratch directory accessible']"
).find('value').text == '1'
assert stats.find(".//statistic[ID='Task working directory exists']").find(
'value').text == '1'
assert float(
xml_out.find(".//statistic[ID='Wall Clock Time']").find(
'value').text) == 2.0
assert xml_out.find('./exitStatus/completed').text == "false"
def process_appker_output(appstdout=None,
stdout=None,
stderr=None,
geninfo=None,
proclog=None,
resource_appker_vars=None):
"""
Process test appkernel output.
"""
# set App Kernel Description
parser = AppKerOutputParser(name='test',
version=1,
description="Test the resource deployment",
url='http://xdmod.buffalo.edu',
measurement_name='test')
# set obligatory parameters and statistics
# set common parameters and statistics
parser.add_common_must_have_params_and_stats()
# set app kernel custom sets
parser.add_must_have_statistic('Wall Clock Time')
parser.add_must_have_statistic('Shell is BASH')
# parse common parameters and statistics
parser.parse_common_params_and_stats(appstdout, stdout, stderr, geninfo,
resource_appker_vars)
# set statistics
if parser.wallClockTime is not None:
parser.set_statistic("Wall Clock Time",
total_seconds(parser.wallClockTime), "Second")
# read output
lines = []
if os.path.isfile(stdout):
fin = open(stdout, "rt")
lines = fin.readlines()
fin.close()
# process the output
parser.set_statistic('Shell is BASH', 0)
j = 0
while j < len(lines):
if lines[j].count("Checking that the shell is BASH") > 0 and lines[
j + 1].count("bash") > 0:
parser.set_statistic('Shell is BASH', 1)
j += 1
if __name__ == "__main__":
# output for testing purpose
print(("parsing complete:", parser.parsing_complete()))
parser.print_params_stats_as_must_have()
print((parser.get_xml()))
# return complete XML otherwise return None
return parser.get_xml()
def process_appker_output(appstdout=None,
stdout=None,
stderr=None,
geninfo=None,
proclog=None,
resource_appker_vars=None):
# set App Kernel Description
parser = AppKerOutputParser(name='quantum_espresso',
version=1,
description="Quantum ESPRESSO (PWSCF)",
url='http://www.quantum-espresso.org',
measurement_name='Quantum_ESPRESSO')
# set obligatory parameters and statistics
# set common parameters and statistics
parser.add_common_must_have_params_and_stats()
# set app kernel custom sets
parser.add_must_have_parameter('App:Version')
parser.add_must_have_parameter('Input:Number of Atoms per Cell')
parser.add_must_have_parameter('Input:Number of Atomic Types')
parser.add_must_have_parameter('Input:Number of Electrons')
parser.add_must_have_statistic('Wall Clock Time')
parser.add_must_have_statistic('User Time')
parser.add_must_have_statistic("Per-Process Dynamical Memory")
parser.add_must_have_statistic("Time Spent in Program Initialization")
parser.add_must_have_statistic("Time Spent in Electron Energy Calculation")
parser.add_must_have_statistic("Time Spent in Force Calculation")
# This statistic probably was working for a different set of inputs, optional now
# parser.add_must_have_statistic("Time Spent in Stress Calculation")
# This statistic probably was working for a different set of inputs, optional now
# parser.add_must_have_statistic("Time Spent in Potential Updates "\
# "(Charge Density and Wavefunctions Extrapolations)")
# parse common parameters and statistics
parser.parse_common_params_and_stats(appstdout, stdout, stderr, geninfo,
resource_appker_vars)
# read output
lines = []
if os.path.isfile(appstdout):
fin = open(appstdout, "rt")
lines = fin.readlines()
fin.close()
# process the output
parser.successfulRun = False
j = 0
while j < len(lines):
m = re.match(r'^\s+Program PWSCF\s+([\w.]+)\s+starts', lines[j])
if m:
parser.set_parameter("App:Version", m.group(1).strip())
m = re.match(r'^\s+number of atoms/cell\s*=\s*([\d.]+)', lines[j])
if m:
parser.set_parameter("Input:Number of Atoms per Cell",
m.group(1).strip())
m = re.match(r'^\s+number of atomic types\s*=\s*([\d.]+)', lines[j])
if m:
parser.set_parameter("Input:Number of Atomic Types",
m.group(1).strip())
m = re.match(r'^\s+number of electrons\s*=\s*([\d.]+)', lines[j])
if m:
parser.set_parameter("Input:Number of Electrons",
m.group(1).strip())
m = re.match(r'^\s+per-process dynamical memory:\s*([\d.]+)\s*Mb',
lines[j])
if m:
parser.set_statistic("Per-Process Dynamical Memory",
(m.group(1).strip()), "MByte")
m = re.match(r'^\s+init_run\s+:\s*([\d.]+)s CPU', lines[j])
if m:
parser.set_statistic("Time Spent in Program Initialization",
(m.group(1).strip()), "Second")
m = re.match(r'^\s+electrons\s+:\s*([\d.]+)s CPU', lines[j])
if m:
parser.set_statistic("Time Spent in Electron Energy Calculation",
(m.group(1).strip()), "Second")
m = re.match(r'^\s+forces\s+:\s*([\d.]+)s CPU', lines[j])
if m:
parser.set_statistic("Time Spent in Force Calculation",
(m.group(1).strip()), "Second")
m = re.match(r'^\s+stress\s+:\s*([\d.]+)s CPU', lines[j])
if m:
parser.set_statistic("Time Spent in Stress Calculation",
(m.group(1).strip()), "Second")
m = re.match(r'^\s+update_pot\s+:\s*([\d.]+)s CPU', lines[j])
if m:
parser.set_statistic(
"Time Spent in Potential Updates (Charge Density and Wavefunctions Extrapolations)",
float(m.group(1).strip()), "Second")
m = re.match(r'^\s+PWSCF\s+:(.+CPU.+)', lines[j])
if m:
run_times = m.group(1).strip().split(',')
for run_time in run_times:
v = run_time.split()
if len(v) > 1:
if v[0].lower().find("m") >= 0:
m = re.match(r'^([0-9]+)m([0-9.]+)s', v[0])
sec = float(m.group(1)) * 60.0 + float(m.group(2))
else:
m = re.match(r'^([0-9.]+)s', v[0])
sec = float(m.group(1))
if v[1].upper().find("CPU") >= 0:
parser.set_statistic("User Time", sec, "Second")
if v[1].upper().find("WALL") >= 0:
parser.set_statistic("Wall Clock Time", sec, "Second")
if re.match(r'^\s+JOB DONE', lines[j]):
parser.successfulRun = True
j += 1
if __name__ == "__main__":
# output for testing purpose
print("parsing complete:", parser.parsing_complete(True))
if hasattr(parser, 'successfulRun'):
print("successfulRun", parser.successfulRun)
parser.print_params_stats_as_must_have()
print(parser.get_xml())
# return complete XML overwize return None
return parser.get_xml()
def process_appker_output(appstdout=None,
stdout=None,
stderr=None,
geninfo=None,
resource_appker_vars=None):
# set App Kernel Description
if resource_appker_vars is not None and 'app' in resource_appker_vars and 'name' in resource_appker_vars[
'app']:
ak_name = resource_appker_vars['app']['name']
else:
ak_name = 'unknown'
# initiate parser
parser = AppKerOutputParser(name=ak_name)
# set obligatory parameters and statistics
# set common parameters and statistics (App:ExeBinSignature and RunEnv:Nodes)
parser.add_common_must_have_params_and_stats()
# set app kernel custom sets
# parser.add_must_have_parameter('App:Version')
parser.add_must_have_statistic('Wall Clock Time')
# parse common parameters and statistics
parser.parse_common_params_and_stats(appstdout, stdout, stderr, geninfo,
resource_appker_vars)
if hasattr(parser, 'appKerWallClockTime'):
parser.set_statistic("Wall Clock Time",
total_seconds(parser.appKerWallClockTime),
"Second")
# Here can be custom output parsing
# #read output
# lines=[]
# if os.path.isfile(appstdout):
# fin=open(appstdout,"rt")
# lines=fin.readlines()
# fin.close()
#
# #process the output
# parser.successfulRun=False
# j=0
# while j<len(lines):
# m=re.search(r'My mega parameter\s+(\d+)',lines[j])
# if m:parser.set_parameter("mega parameter",m.group(1))
#
# m=re.search(r'My mega parameter\s+(\d+)',lines[j])
# if m:parser.set_statistic("mega statistics",m.group(1),"Seconds")
#
# m=re.search(r'Done',lines[j])
# if m:parser.successfulRun=True
#
# j+=1
if __name__ == "__main__":
# output for testing purpose
print("Parsing complete:", parser.parsing_complete(verbose=True))
print("Following statistics and parameter can be set as obligatory:")
parser.print_params_stats_as_must_have()
print("\nResulting XML:")
print(parser.get_xml())
# return complete XML otherwise return None
return parser.get_xml()
def process_appker_output(appstdout=None, stdout=None, stderr=None, geninfo=None, proclog=None,
resource_appker_vars=None):
# set App Kernel Description
parser = AppKerOutputParser(
name='ior',
version=1,
description="IOR (Interleaved-Or-Random) Benchmark",
url='http://freshmeat.net/projects/ior',
measurement_name='IOR'
)
app_vars = None
if resource_appker_vars is not None and 'app' in resource_appker_vars:
app_vars = resource_appker_vars['app']
# set obligatory parameters and statistics
# set common parameters and statistics
parser.add_common_must_have_params_and_stats()
# set app kernel custom sets
parser.add_must_have_parameter('App:Version')
if app_vars is None or (
app_vars is not None and 'testHDF5' in app_vars and
app_vars['testHDF5'] is True):
parser.add_must_have_parameter('HDF Version')
parser.add_must_have_parameter('HDF5 Collective N-to-1 Test File System')
parser.add_must_have_parameter('HDF5 Independent N-to-1 Test File System')
parser.add_must_have_parameter('HDF5 N-to-N Test File System')
if app_vars is None or (
app_vars is not None and 'testMPIIO' in app_vars and
app_vars['testMPIIO'] is True):
parser.add_must_have_parameter('MPIIO Collective N-to-1 Test File System')
parser.add_must_have_parameter('MPIIO Independent N-to-1 Test File System')
parser.add_must_have_parameter('MPIIO N-to-N Test File System')
if app_vars is None or (
app_vars is not None and 'testPOSIX' in app_vars and
app_vars['testPOSIX'] is True):
parser.add_must_have_parameter('POSIX N-to-1 Test File System')
parser.add_must_have_parameter('POSIX N-to-N Test File System')
if app_vars is None or (
app_vars is not None and 'testNetCDF' in app_vars and
app_vars['testNetCDF'] is True):
parser.add_must_have_parameter('Parallel NetCDF Collective N-to-1 Test File System')
parser.add_must_have_parameter('Parallel NetCDF Independent N-to-1 Test File System')
parser.add_must_have_parameter('Parallel NetCDF Version')
parser.add_must_have_parameter('Per-Process Data Size')
parser.add_must_have_parameter('Per-Process I/O Block Size')
parser.add_must_have_parameter('RunEnv:Nodes')
parser.add_must_have_parameter('Transfer Size Per I/O')
if app_vars is None or (
app_vars is not None and 'testHDF5' in app_vars and
app_vars['testHDF5'] is True):
parser.add_must_have_statistic('HDF5 Collective N-to-1 Read Aggregate Throughput')
parser.add_must_have_statistic('HDF5 Collective N-to-1 Write Aggregate Throughput')
parser.add_must_have_statistic('HDF5 Independent N-to-1 Read Aggregate Throughput')
parser.add_must_have_statistic('HDF5 Independent N-to-1 Write Aggregate Throughput')
parser.add_must_have_statistic('HDF5 N-to-N Read Aggregate Throughput')
parser.add_must_have_statistic('HDF5 N-to-N Write Aggregate Throughput')
if app_vars is None or (
app_vars is not None and 'testMPIIO' in app_vars and
app_vars['testMPIIO'] is True):
parser.add_must_have_statistic('MPIIO Collective N-to-1 Read Aggregate Throughput')
parser.add_must_have_statistic('MPIIO Collective N-to-1 Write Aggregate Throughput')
parser.add_must_have_statistic('MPIIO Independent N-to-1 Read Aggregate Throughput')
parser.add_must_have_statistic('MPIIO Independent N-to-1 Write Aggregate Throughput')
parser.add_must_have_statistic('MPIIO N-to-N Read Aggregate Throughput')
parser.add_must_have_statistic('MPIIO N-to-N Write Aggregate Throughput')
if app_vars is None or (
app_vars is not None and 'testPOSIX' in app_vars and
app_vars['testPOSIX'] is True):
parser.add_must_have_statistic('POSIX N-to-1 Read Aggregate Throughput')
parser.add_must_have_statistic('POSIX N-to-1 Write Aggregate Throughput')
parser.add_must_have_statistic('POSIX N-to-N Read Aggregate Throughput')
parser.add_must_have_statistic('POSIX N-to-N Write Aggregate Throughput')
if app_vars is None or (
app_vars is not None and 'testNetCDF' in app_vars and
app_vars['testNetCDF'] is True):
parser.add_must_have_statistic('Parallel NetCDF Collective N-to-1 Read Aggregate Throughput')
parser.add_must_have_statistic('Parallel NetCDF Collective N-to-1 Write Aggregate Throughput')
parser.add_must_have_statistic('Parallel NetCDF Independent N-to-1 Read Aggregate Throughput')
parser.add_must_have_statistic('Parallel NetCDF Independent N-to-1 Write Aggregate Throughput')
parser.add_must_have_statistic('Number of Tests Passed')
parser.add_must_have_statistic('Number of Tests Started')
parser.add_must_have_statistic('Wall Clock Time')
parser.completeOnPartialMustHaveStatistics = True
# parse common parameters and statistics
parser.parse_common_params_and_stats(appstdout, stdout, stderr, geninfo, resource_appker_vars)
if hasattr(parser, 'appKerWallClockTime'):
parser.set_statistic("Wall Clock Time", total_seconds(parser.appKerWallClockTime), "Second")
# read output
lines = []
if os.path.isfile(appstdout):
fin = open(appstdout, "rt")
lines = fin.readlines()
fin.close()
# process the output
# find which version of IOR was used
ior_output_version = None
j = 0
while j < len(lines) - 1:
# IOR RELEASE: IOR-2.10.3
m = re.match(r'^#\s+IOR RELEASE:\s(.+)', lines[j])
if m:
ior_output_version = 20
# IOR-3.2.0: MPI Coordinated Test of Parallel I/O
# IOR-3.3.0+dev: MPI Coordinated Test of Parallel I/O
m = re.match(r'^IOR-([3-9])\.([0-9])+\.[0-9]\S*: MPI Coordinated Test of Parallel I/O', lines[j])
if m:
ior_major = int(m.group(1))
ior_minor = int(m.group(2))
if ior_major >=3:
if ior_minor >= 3:
ior_output_version = 33
elif ior_minor >= 2:
ior_output_version = 32
else:
ior_output_version = 30
j += 1
if ior_output_version is None:
print("ERROR: unknown version of IOR output!!!")
parser.successfulRun = False
total_number_of_tests, tests_passed = process_ior_output_v33(parser, lines)
if app_vars is not None and 'doAllWritesFirst' in app_vars:
if app_vars['doAllWritesFirst']:
# i.e. separate read and write
total_number_of_tests = total_number_of_tests // 2
else:
# by default separate read and write
total_number_of_tests = total_number_of_tests // 2
parser.set_statistic('Number of Tests Passed', tests_passed)
parser.set_statistic('Number of Tests Started', total_number_of_tests)
if __name__ == "__main__":
# output for testing purpose
print("parsing complete:", parser.parsing_complete(verbose=True))
parser.print_params_stats_as_must_have()
parser.print_template_for_pytest()
print(parser.get_xml())
# return complete XML overwize return None
return parser.get_xml()
def process_appker_output(appstdout=None,
stdout=None,
stderr=None,
geninfo=None,
resource_appker_vars=None):
# set App Kernel Description
parser = AppKerOutputParser(
name='gromacs_micro',
version=1,
description="GROMACS: micro-benchmark for testing purposes",
url='http://www.gromacs.org/',
measurement_name='GROMACS-Micro')
# set obligatory parameters and statistics
# set common parameters and statistics
parser.add_common_must_have_params_and_stats()
# set app kernel custom sets
parser.add_must_have_parameter('App:Version')
parser.add_must_have_statistic('Simulation Speed')
parser.add_must_have_statistic('Wall Clock Time')
# parse common parameters and statistics
parser.parse_common_params_and_stats(appstdout, stdout, stderr, geninfo,
resource_appker_vars)
# read output
lines = []
if os.path.isfile(appstdout):
fin = open(appstdout, "rt")
lines = fin.readlines()
fin.close()
# process the output
successful_run = False
j = 0
while j < len(lines):
m = re.search(r'^GROMACS:\s+ gmx mdrun, version\s+(\S+)$', lines[j])
if m:
parser.set_parameter("App:Version", m.group(1))
m = re.search(r'^Performance: \s+([0-9.]+)', lines[j])
if m:
parser.set_statistic("Simulation Speed", float(m.group(1)),
"ns/day")
m = re.search(r'^ \s+Time: \s+([0-9.]+) \s+([0-9.]+)', lines[j])
if m:
parser.set_statistic("Wall Clock Time", m.group(2), "Second")
parser.set_statistic("Core Clock Time", m.group(1), "Second")
m = re.match(r'^GROMACS reminds you', lines[j])
if m:
successful_run = True
j += 1
parser.successfulRun = successful_run
if __name__ == "__main__":
# output for testing purpose
print("parsing complete:", parser.parsing_complete())
parser.print_params_stats_as_must_have()
print(parser.get_xml())
# return complete XML overwize return None
return parser.get_xml()
def process_appker_output(appstdout=None,
stdout=None,
stderr=None,
geninfo=None,
resource_appker_vars=None):
# set App Kernel Description
parser = AppKerOutputParser(
name='nwchem',
version=1,
description="NWChem: Northwest Computational Chemistry Package",
url='http://www.emsl.pnl.gov/docs/nwchem',
measurement_name='NWChem')
# set obligatory parameters and statistics
# set common parameters and statistics
parser.add_common_must_have_params_and_stats()
# set app kernel custom sets
parser.add_must_have_parameter('App:Version')
parser.add_must_have_parameter('App:Branch')
parser.add_must_have_parameter('Input:File')
parser.add_must_have_statistic('Wall Clock Time')
parser.add_must_have_statistic('User Time')
parser.add_must_have_statistic("Global Arrays 'Create' Calls")
parser.add_must_have_statistic("Global Arrays 'Destroy' Calls")
parser.add_must_have_statistic("Global Arrays 'Get' Calls")
parser.add_must_have_statistic("Global Arrays 'Put' Calls")
parser.add_must_have_statistic("Global Arrays 'Accumulate' Calls")
parser.add_must_have_statistic("Global Arrays 'Get' Amount")
parser.add_must_have_statistic("Global Arrays 'Put' Amount")
parser.add_must_have_statistic("Global Arrays 'Accumulate' Amount")
# parse common parameters and statistics
parser.parse_common_params_and_stats(appstdout, stdout, stderr, geninfo,
resource_appker_vars)
# read output
lines = []
if os.path.isfile(appstdout):
fin = open(appstdout, "rt")
lines = fin.readlines()
fin.close()
# process the output
j = 0
while j < len(lines):
m = re.search(
r'Northwest Computational Chemistry Package \(NWChem\) (.+)',
lines[j])
if m:
parser.set_parameter("App:Version", m.group(1).strip())
m = re.search(r'nwchem branch *=(.+)', lines[j])
if m:
parser.set_parameter("App:Branch", m.group(1).strip())
m = re.search(r'input\s+= (.+)', lines[j])
if m:
parser.set_parameter("Input:File", m.group(1).strip())
m = re.search(r'Total times\s+cpu:\s+([0-9.]+)s\s+wall:\s+([0-9.]+)s',
lines[j])
if m:
parser.set_statistic("Wall Clock Time",
m.group(2).strip(), "Second")
parser.set_statistic("User Time", m.group(1).strip(), "Second")
# GA Statistics for process 0
# ------------------------------
#
# create destroy get put acc scatter gather read&inc
# calls: 521 521 6.28e+05 6.45e+04 6.78e+05 0 0 0
# number of processes/call 1.05e+00 1.36e+00 1.03e+00 0.00e+00 0.00e+00
# bytes total: 7.33e+09 4.35e+08 1.53e+09 0.00e+00 0.00e+00 0.00e+00
# bytes remote: 5.74e+09 1.31e+08 1.09e+09 0.00e+00 0.00e+00 0.00e+00
# Max memory consumed for GA by this process: 47428032 bytes
if re.search(r'GA Statistics for process', lines[j]):
if re.match(r'^calls', lines[j + 4]):
v = lines[j + 4].strip().split()
parser.set_statistic("Global Arrays 'Create' Calls",
"%.0f" % float(v[1]), "Number of Calls")
parser.set_statistic("Global Arrays 'Destroy' Calls",
"%.0f" % float(v[2]), "Number of Calls")
parser.set_statistic("Global Arrays 'Get' Calls",
"%.0f" % float(v[3]), "Number of Calls")
parser.set_statistic("Global Arrays 'Put' Calls",
"%.0f" % float(v[4]), "Number of Calls")
parser.set_statistic("Global Arrays 'Accumulate' Calls",
"%.0f" % float(v[5]), "Number of Calls")
v = lines[j + 6].strip().split()
parser.set_statistic("Global Arrays 'Get' Amount",
(float(v[2])) / 1048576.0, "MByte")
parser.set_statistic("Global Arrays 'Put' Amount",
(float(v[3])) / 1048576.0, "MByte")
parser.set_statistic("Global Arrays 'Accumulate' Amount",
(float(v[4])) / 1048576.0, "MByte")
# NWChem can be optionally compiled with PAPI, and it will
# report some GLOPS at the end
# thus here it is optional
m = re.search(r'Aggregate GFLOPS \(Real_time\):\s+([0-9.]+)', lines[j])
if m:
parser.set_statistic(
"Floating-Point Performance (Wall Clock Time)",
1000.0 * float(m.group(1).strip()), "MFLOP per Second")
m = re.search(r'Aggregate GFLOPS \(Proc_time\):\s+([0-9.]+)', lines[j])
if m:
parser.set_statistic("Floating-Point Performance (User Time)",
1000.0 * float(m.group(1).strip()),
"MFLOP per Second")
j += 1
if __name__ == "__main__":
# output for testing purpose
print("parsing complete:", parser.parsing_complete())
parser.print_params_stats_as_must_have()
print(parser.get_xml())
# return complete XML overwize return None
return parser.get_xml()
def process_appker_output(appstdout=None,
stdout=None,
stderr=None,
geninfo=None,
proclog=None,
resource_appker_vars=None):
"""
Process test appkernel output.
"""
# set App Kernel Description
parser = AppKerOutputParser(name='test',
version=1,
description="Test the resource deployment",
url='http://xdmod.buffalo.edu',
measurement_name='test')
# set obligatory parameters and statistics
# set common parameters and statistics
parser.add_common_must_have_params_and_stats()
# set app kernel custom sets
# parser.add_must_have_statistic('Wall Clock Time')
# parser.add_must_have_statistic('Shell is BASH')
# parse common parameters and statistics
parser.parse_common_params_and_stats(appstdout, stdout, stderr, geninfo,
resource_appker_vars)
# set statistics
if parser.wallClockTime is not None:
parser.set_statistic("Wall Clock Time",
total_seconds(parser.wallClockTime), "Second")
# read output
lines = []
if os.path.isfile(stdout):
fin = open(stdout, "rt")
lines = fin.readlines()
fin.close()
# process the output
parser.set_statistic('Shell is BASH', 0)
j = 0
while j < len(lines):
if lines[j].count("Checking that the shell is BASH") > 0 and lines[
j + 1].count("bash") > 0:
parser.set_statistic('Shell is BASH', 1)
j += 1
# process proc log
if proclog is not None:
os_start = None
os_first_login = None
os_start_shutdown = None
os_terminated = None
with open(proclog, "rt") as fin:
for line in fin:
m = re.search(
"Starting OpenStack instance \(([0-9]+)-([0-9]+)-([0-9]+) ([0-9]+):([0-9]+):([0-9]+)\)",
line)
if m:
os_start = datetime.datetime(int(m.group(1)),
int(m.group(2)),
int(m.group(3)),
int(m.group(4)),
int(m.group(5)),
int(m.group(6)))
m = re.search(
"OpenStack Instance should be up and running \(([0-9]+)-([0-9]+)-([0-9]+) ([0-9]+):([0-9]+):([0-9]+)\)",
line)
if m:
os_first_login = datetime.datetime(int(m.group(1)),
int(m.group(2)),
int(m.group(3)),
int(m.group(4)),
int(m.group(5)),
int(m.group(6)))
m = re.search(
"Shutting down OpenStack instance \(([0-9]+)-([0-9]+)-([0-9]+) ([0-9]+):([0-9]+):([0-9]+)\)",
line)
if m:
os_start_shutdown = datetime.datetime(
int(m.group(1)), int(m.group(2)), int(m.group(3)),
int(m.group(4)), int(m.group(5)), int(m.group(6)))
m = re.search(
"OpenStack Instance should be down and terminated \(([0-9]+)-([0-9]+)-([0-9]+) ([0-9]+):([0-9]+):([0-9]+)\)",
line)
if m:
os_terminated = datetime.datetime(int(m.group(1)),
int(m.group(2)),
int(m.group(3)),
int(m.group(4)),
int(m.group(5)),
int(m.group(6)))
# googlecloud
m = re.search(
"Starting Google Cloud instance \(([0-9]+)-([0-9]+)-([0-9]+) ([0-9]+):([0-9]+):([0-9]+)\)",
line)
if m:
os_start = datetime.datetime(int(m.group(1)),
int(m.group(2)),
int(m.group(3)),
int(m.group(4)),
int(m.group(5)),
int(m.group(6)))
m = re.search(
"Google Cloud Instance should be up and running \(([0-9]+)-([0-9]+)-([0-9]+) ([0-9]+):([0-9]+):([0-9]+)\)",
line)
if m:
os_first_login = datetime.datetime(int(m.group(1)),
int(m.group(2)),
int(m.group(3)),
int(m.group(4)),
int(m.group(5)),
int(m.group(6)))
m = re.search(
"Shutting down Google Cloud instance \(([0-9]+)-([0-9]+)-([0-9]+) ([0-9]+):([0-9]+):([0-9]+)\)",
line)
if m:
os_start_shutdown = datetime.datetime(
int(m.group(1)), int(m.group(2)), int(m.group(3)),
int(m.group(4)), int(m.group(5)), int(m.group(6)))
m = re.search(
"Google Cloud Instance should be down and terminated \(([0-9]+)-([0-9]+)-([0-9]+) ([0-9]+):([0-9]+):([0-9]+)\)",
line)
if m:
os_terminated = datetime.datetime(int(m.group(1)),
int(m.group(2)),
int(m.group(3)),
int(m.group(4)),
int(m.group(5)),
int(m.group(6)))
if os_start is not None and os_first_login is not None:
parser.set_statistic('Cloud Instance, Start Time to Login',
total_seconds(os_first_login - os_start))
if os_start_shutdown is not None and os_terminated is not None:
parser.set_statistic(
'Cloud Instance, Shut Down Time',
total_seconds(os_terminated - os_start_shutdown))
# log.info("OpenStack Instance should be up and running (%s)"
# log.info("Shutting down OpenStack instance (%s)" % datetime.datetime.today().strftime("%Y-%m-%d %H:%M:%S"))
# log.info("OpenStack Instance should be down and terminated (%s)"
if __name__ == "__main__":
# output for testing purpose
print(("parsing complete:", parser.parsing_complete()))
parser.print_params_stats_as_must_have()
print((parser.get_xml()))
# return complete XML otherwise return None
return parser.get_xml()
def process_appker_output(appstdout=None,
stdout=None,
stderr=None,
geninfo=None,
resource_appker_vars=None):
# set App Kernel Description
parser = AppKerOutputParser(
name='mpi-tile-io',
version=1,
description="MPI-Tile-IO Benchmark",
url='http://www.mcs.anl.gov/research/projects/pio-benchmark',
measurement_name='MPI-Tile-IO')
# set obligatory parameters and statistics
# set common parameters and statistics
parser.add_common_must_have_params_and_stats()
# set app kernel custom sets
parser.add_must_have_parameter('2D Collective Read Test File System')
parser.add_must_have_parameter('2D Collective Write Test File System')
parser.add_must_have_parameter('2D HDF5 Collective Read Test File System')
parser.add_must_have_parameter('2D HDF5 Collective Write Test File System')
parser.add_must_have_parameter('2D Independent Read Test File System')
parser.add_must_have_parameter('2D Independent Write Test File System')
parser.add_must_have_parameter('2D Per-Process Data Topology')
parser.add_must_have_parameter('2D Per-Process Ghost Zone')
parser.add_must_have_parameter('2D Per-Process Memory')
parser.add_must_have_parameter('2D Process Topology')
parser.add_must_have_parameter('3D Collective Read Test File System')
parser.add_must_have_parameter('3D Collective Write Test File System')
parser.add_must_have_parameter('3D HDF5 Collective Read Test File System')
parser.add_must_have_parameter('3D HDF5 Collective Write Test File System')
parser.add_must_have_parameter('3D Independent Read Test File System')
parser.add_must_have_parameter('3D Independent Write Test File System')
parser.add_must_have_parameter('3D Per-Process Data Topology')
parser.add_must_have_parameter('3D Per-Process Ghost Zone')
parser.add_must_have_parameter('3D Per-Process Memory')
parser.add_must_have_parameter('3D Process Topology')
parser.add_must_have_parameter('App:ExeBinSignature')
parser.add_must_have_parameter('HDF Version')
# parser.add_must_have_parameter('MPI-IO Hints')
parser.add_must_have_parameter('RunEnv:Nodes')
parser.add_must_have_statistic(
'2D Array Collective Read Aggregate Throughput')
parser.add_must_have_statistic(
'2D Array Collective Write Aggregate Throughput')
parser.add_must_have_statistic(
'2D Array HDF5 Collective Read Aggregate Throughput')
parser.add_must_have_statistic(
'2D Array HDF5 Collective Write Aggregate Throughput')
parser.add_must_have_statistic(
'2D Array Independent Read Aggregate Throughput')
parser.add_must_have_statistic(
'2D Array Independent Write Aggregate Throughput')
parser.add_must_have_statistic(
'3D Array Collective Read Aggregate Throughput')
parser.add_must_have_statistic(
'3D Array Collective Write Aggregate Throughput')
parser.add_must_have_statistic(
'3D Array HDF5 Collective Read Aggregate Throughput')
parser.add_must_have_statistic(
'3D Array HDF5 Collective Write Aggregate Throughput')
parser.add_must_have_statistic(
'3D Array Independent Read Aggregate Throughput')
parser.add_must_have_statistic(
'3D Array Independent Write Aggregate Throughput')
parser.add_must_have_statistic('File Close Time (2D Data Collective Read)')
parser.add_must_have_statistic(
'File Close Time (2D Data Collective Write)')
parser.add_must_have_statistic(
'File Close Time (2D Data HDF5 Collective Read)')
parser.add_must_have_statistic(
'File Close Time (2D Data HDF5 Collective Write)')
parser.add_must_have_statistic(
'File Close Time (2D Data Independent Read)')
parser.add_must_have_statistic(
'File Close Time (2D Data Independent Write)')
parser.add_must_have_statistic('File Close Time (3D Data Collective Read)')
parser.add_must_have_statistic(
'File Close Time (3D Data Collective Write)')
parser.add_must_have_statistic(
'File Close Time (3D Data HDF5 Collective Read)')
parser.add_must_have_statistic(
'File Close Time (3D Data HDF5 Collective Write)')
parser.add_must_have_statistic(
'File Close Time (3D Data Independent Read)')
parser.add_must_have_statistic(
'File Close Time (3D Data Independent Write)')
parser.add_must_have_statistic('File Open Time (2D Data Collective Read)')
parser.add_must_have_statistic('File Open Time (2D Data Collective Write)')
parser.add_must_have_statistic(
'File Open Time (2D Data HDF5 Collective Read)')
parser.add_must_have_statistic(
'File Open Time (2D Data HDF5 Collective Write)')
parser.add_must_have_statistic('File Open Time (2D Data Independent Read)')
parser.add_must_have_statistic(
'File Open Time (2D Data Independent Write)')
parser.add_must_have_statistic('File Open Time (3D Data Collective Read)')
parser.add_must_have_statistic('File Open Time (3D Data Collective Write)')
parser.add_must_have_statistic(
'File Open Time (3D Data HDF5 Collective Read)')
parser.add_must_have_statistic(
'File Open Time (3D Data HDF5 Collective Write)')
parser.add_must_have_statistic('File Open Time (3D Data Independent Read)')
parser.add_must_have_statistic(
'File Open Time (3D Data Independent Write)')
parser.add_must_have_statistic('Wall Clock Time')
# parse common parameters and statistics
parser.parse_common_params_and_stats(appstdout, stdout, stderr, geninfo,
resource_appker_vars)
if hasattr(parser, 'appKerWallClockTime'):
parser.set_statistic("Wall Clock Time",
parser.appKerWallClockTime.total_seconds(),
"Second")
# read output
lines = []
if os.path.isfile(appstdout):
fin = open(appstdout, "rt")
lines = fin.readlines()
fin.close()
# process the output
# The parameters mapping table
# The result mapping table
pm = {
"processesTopology": {
're': r"^# processes topology:(.+)",
'refun': re.match,
'val': None
},
"localDatasetTopology": {
're': r"^# local dataset topology:(.+)element",
'refun': re.match,
'val': None
},
"localMemoryUsage": {
're': r"^# local dataset memory usage:(.+)byte",
'refun': re.match,
'val': None
},
"datasetGhostZone": {
're': r"^# local dataset ghost zone:(.+)",
'refun': re.match,
'val': None
},
"mpiIOhints": {
're': r"^# mpiio hints:(.+)",
'refun': re.match,
'val': None
},
"maxFileOpenTime": {
're': r"^# Open:.+?max=(\S+)",
'refun': re.match,
'val': None
},
"maxFileCloseTime": {
're': r"^# Close:.+?max=(\S+)",
'refun': re.match,
'val': None
},
"collectiveIO": {
're': r"^# collective IO:(.+)",
'refun': re.match,
'val': None
},
"testFileName": {
're': r"^# filename:(.+)",
'refun': re.match,
'val': None
},
"fileSystem": {
're': r"^# filesystem:(.+)",
'refun': re.match,
'val': None
},
"hdf5Version": {
're': r"^# HDF5 Version:(.+)",
'refun': re.match,
'val': None
},
}
parser.successfulRun = False
j = -1
while j < len(lines) - 1:
for k, v in pm.items():
m = v['refun'](v['re'], lines[j])
if m:
v['val'] = m.group(1).strip()
m = re.match(r'^# (.+?)bandwidth:(.+)bytes', lines[j])
if m:
read_or_write = m.group(1).strip()
io_bandwidth = m.group(2).strip()
# can output data ?
if pm['processesTopology']['val'] and pm['collectiveIO']['val']:
# construct the label
label = ''
dim = '2D'
m = re.search(r'\d+x\d+x\d', pm['processesTopology']['val'])
if m:
dim = '3D'
if pm['hdf5Version']['val']:
label += 'HDF5 '
parser.set_parameter("HDF Version",
pm['hdf5Version']['val'])
m = re.search(r'yes', pm['collectiveIO']['val'], re.I)
if m:
label += 'Collective '
else:
label += 'Independent '
m0 = re.search(r'read', read_or_write, re.I)
m1 = re.search(r'write', read_or_write, re.I)
if m0:
label += 'Read'
elif m1:
label += 'Write'
else:
label += read_or_write[0].upper() + read_or_write[1:]
parser.set_statistic(
"%s Array %s Aggregate Throughput" % (dim, label),
"%.2f" % (float(io_bandwidth) / 1024.0 / 1024.0),
"MByte per Second")
if pm["maxFileOpenTime"]['val']:
parser.set_statistic(
"File Open Time (%s Data %s)" % (dim, label),
pm["maxFileOpenTime"]['val'], "Second")
if pm["maxFileCloseTime"]['val']:
parser.set_statistic(
"File Close Time (%s Data %s)" % (dim, label),
pm["maxFileCloseTime"]['val'], "Second")
parser.set_parameter("%s Process Topology" % (dim, ),
pm["processesTopology"]['val'])
if pm["localMemoryUsage"]['val']:
parser.set_parameter(
"%s Per-Process Memory" % (dim, ),
float(pm["localMemoryUsage"]['val']) / 1024.0 / 1024.0,
"MByte")
if pm["localDatasetTopology"]['val']:
parser.set_parameter(
"%s Per-Process Data Topology" % (dim, ),
pm["localDatasetTopology"]['val'], "Element")
if pm["datasetGhostZone"]['val']:
parser.set_parameter("%s Per-Process Ghost Zone" % (dim, ),
pm["datasetGhostZone"]['val'])
if pm["mpiIOhints"]['val']:
parser.set_parameter("MPI-IO Hints",
pm["mpiIOhints"]['val'])
# $benchmark->set_parameter( "${dim} ${label} Test File", $testFileName ) if( defined($testFileName) )
if pm["fileSystem"]['val']:
parser.set_parameter(
"%s %s Test File System" % (dim, label),
pm["fileSystem"]['val'])
parser.successfulRun = True
pm["processesTopology"]['val'] = None
pm["localDatasetTopology"]['val'] = None
pm["localMemoryUsage"]['val'] = None
pm["datasetGhostZone"]['val'] = None
pm["mpiIOhints"]['val'] = None
# pm["readOrWrite"]['val']=None
pm["collectiveIO"]['val'] = None
# pm["IObandwidth"]['val']=None
pm["maxFileOpenTime"]['val'] = None
pm["maxFileCloseTime"]['val'] = None
pm["testFileName"]['val'] = None
pm["fileSystem"]['val'] = None
pm["hdf5Version"]['val'] = None
j += 1
if __name__ == "__main__":
# output for testing purpose
print("parsing complete:", parser.parsing_complete(verbose=True))
parser.print_params_stats_as_must_have()
print(parser.get_xml())
# Print out missing parameters for debug purpose
parser.parsing_complete(verbose=True)
# return complete XML overwize return None
return parser.get_xml()
def process_appker_output(appstdout=None,
stdout=None,
stderr=None,
geninfo=None,
proclog=None,
resource_appker_vars=None):
# set App Kernel Description
parser = AppKerOutputParser(name='bundle',
version=1,
description='bundled tasks',
url='https://xdmod.ccr.buffalo.edu',
measurement_name='BUNDLE')
parser.add_must_have_parameter('RunEnv:Nodes')
parser.add_must_have_statistic('Wall Clock Time')
parser.add_must_have_statistic("Success Rate")
parser.add_must_have_statistic("Successful Subtasks")
parser.add_must_have_statistic("Total Number of Subtasks")
# set obligatory parameters and statistics
# set common parameters and statistics
parser.parse_common_params_and_stats(appstdout, stdout, stderr, geninfo,
resource_appker_vars)
if hasattr(parser, 'wallClockTime'):
parser.set_statistic("Wall Clock Time",
total_seconds(parser.wallClockTime), "Second")
# check the status of subtasks
# resource_appker_vars['taskId']=self.task_id
# resource_appker_vars['subTasksId']=self.subTasksId
success_rate = 0.0
total_subtasks = 0
successful_subtasks = 0
try:
db, cur = akrr.db.get_akrr_db()
for subTaskId in resource_appker_vars['subTasksId']:
cur.execute(
'''SELECT instance_id,status FROM akrr_xdmod_instanceinfo
WHERE instance_id=%s ;''', (subTaskId, ))
raw = cur.fetchall()
instance_id, status = raw[0]
success_rate += status
successful_subtasks += status
success_rate /= len(resource_appker_vars['subTasksId'])
total_subtasks = len(resource_appker_vars['subTasksId'])
cur.close()
del db
except:
print(traceback.format_exc())
parser.set_statistic("Success Rate", success_rate)
parser.set_statistic("Successful Subtasks", successful_subtasks)
parser.set_statistic("Total Number of Subtasks", total_subtasks)
# if successfulSubtasks==totalSubtasks:
if __name__ == "__main__":
# output for testing purpose
print("parsing complete:", parser.parsing_complete(verbose=True))
parser.print_params_stats_as_must_have()
print(parser.get_xml())
# return complete XML overwize return None
return parser.get_xml()
def process_appker_output(appstdout=None,
stdout=None,
stderr=None,
geninfo=None,
proclog=None,
resource_appker_vars=None):
# set App Kernel Description
parser = AppKerOutputParser(name='imb',
version=1,
description="Intel MPI Benchmarks",
url='http://www.intel.com/software/imb',
measurement_name='Intel MPI Benchmarks')
# set obligatory parameters and statistics
# set common parameters and statistics
parser.add_common_must_have_params_and_stats()
# set app kernel custom sets
parser.add_must_have_parameter('App:MPI Thread Environment')
parser.add_must_have_parameter('App:MPI Version')
parser.add_must_have_parameter('App:Max Message Size')
parser.add_must_have_statistic('Max Exchange Bandwidth')
parser.add_must_have_statistic(
"Max MPI-2 Bidirectional 'Get' Bandwidth (aggregate)")
parser.add_must_have_statistic(
"Max MPI-2 Bidirectional 'Get' Bandwidth (non-aggregate)")
parser.add_must_have_statistic(
"Max MPI-2 Bidirectional 'Put' Bandwidth (aggregate)")
parser.add_must_have_statistic(
"Max MPI-2 Bidirectional 'Put' Bandwidth (non-aggregate)")
parser.add_must_have_statistic(
"Max MPI-2 Unidirectional 'Get' Bandwidth (aggregate)")
parser.add_must_have_statistic(
"Max MPI-2 Unidirectional 'Get' Bandwidth (non-aggregate)")
parser.add_must_have_statistic(
"Max MPI-2 Unidirectional 'Put' Bandwidth (aggregate)")
parser.add_must_have_statistic(
"Max MPI-2 Unidirectional 'Put' Bandwidth (non-aggregate)")
parser.add_must_have_statistic('Max PingPing Bandwidth')
parser.add_must_have_statistic('Max PingPong Bandwidth')
parser.add_must_have_statistic('Max SendRecv Bandwidth')
parser.add_must_have_statistic('Min AllGather Latency')
parser.add_must_have_statistic('Min AllGatherV Latency')
parser.add_must_have_statistic('Min AllReduce Latency')
parser.add_must_have_statistic('Min AllToAll Latency')
parser.add_must_have_statistic('Min AllToAllV Latency')
parser.add_must_have_statistic('Min Barrier Latency')
parser.add_must_have_statistic('Min Broadcast Latency')
parser.add_must_have_statistic('Min Gather Latency')
parser.add_must_have_statistic('Min GatherV Latency')
# parser.add_must_have_statistic("Min MPI-2 'Accumulate' Latency (aggregate)")
# parser.add_must_have_statistic("Min MPI-2 'Accumulate' Latency (non-aggregate)")
parser.add_must_have_statistic('Min MPI-2 Window Creation Latency')
parser.add_must_have_statistic('Min Reduce Latency')
parser.add_must_have_statistic('Min ReduceScatter Latency')
parser.add_must_have_statistic('Min Scatter Latency')
parser.add_must_have_statistic('Min ScatterV Latency')
parser.add_must_have_statistic('Wall Clock Time')
# parse common parameters and statistics
parser.parse_common_params_and_stats(appstdout, stdout, stderr, geninfo,
resource_appker_vars)
if hasattr(parser, 'appKerWallClockTime'):
parser.set_statistic("Wall Clock Time",
total_seconds(parser.appKerWallClockTime),
"Second")
# Intel MPI benchmark suite contains three classes of benchmarks:
#
# Single-transfer, which needs only 2 processes
# Parallel-transfer, which can use as many processes that are available
# Collective, which can use as many processes that are available
# The parameters mapping table
params = {
"MPI Thread Environment": ["MPI Thread Environment", "", ""],
"MPI Version": ["MPI Version", "", ""],
"Maximum message length in bytes":
["Max Message Size", "MByte", "<val>/1024/1024"]
}
# The result mapping table
metrics = {
"PingPing": ["PingPing Bandwidth", "MByte per Second", "max"],
"PingPong": ["PingPong Bandwidth", "MByte per Second", "max"],
"Multi-PingPing": ["PingPing Bandwidth", "MByte per Second", "max"],
"Multi-PingPong": ["PingPong Bandwidth", "MByte per Second", "max"],
"Sendrecv": ["SendRecv Bandwidth", "MByte per Second", "max"],
"Exchange": ["Exchange Bandwidth", "MByte per Second", "max"],
"Allreduce": ["AllReduce Latency", "us", "min"],
"Reduce": ["Reduce Latency", "us", "min"],
"Reduce_scatter": ["ReduceScatter Latency", "us", "min"],
"Allgather": ["AllGather Latency", "us", "min"],
"Allgatherv": ["AllGatherV Latency", "us", "min"],
"Gather": ["Gather Latency", "us", "min"],
"Gatherv": ["GatherV Latency", "us", "min"],
"Scatter": ["Scatter Latency", "us", "min"],
"Scatterv": ["ScatterV Latency", "us", "min"],
"Alltoall": ["AllToAll Latency", "us", "min"],
"Alltoallv": ["AllToAllV Latency", "us", "min"],
"Bcast": ["Broadcast Latency", "us", "min"],
"Barrier": ["Barrier Latency", "us", "min"],
"Window": ["MPI-2 Window Creation Latency", "us", "min"],
"Multi-Unidir_Get":
["MPI-2 Unidirectional 'Get' Bandwidth", "MByte per Second", "max"],
"Multi-Unidir_Put":
["MPI-2 Unidirectional 'Put' Bandwidth", "MByte per Second", "max"],
"Multi-Bidir_Get":
["MPI-2 Bidirectional 'Get' Bandwidth", "MByte per Second", "max"],
"Multi-Bidir_Put":
["MPI-2 Bidirectional 'Put' Bandwidth", "MByte per Second", "max"],
"Unidir_Get":
["MPI-2 Unidirectional 'Get' Bandwidth", "MByte per Second", "max"],
"Unidir_Put":
["MPI-2 Unidirectional 'Put' Bandwidth", "MByte per Second", "max"],
"Bidir_Get":
["MPI-2 Bidirectional 'Get' Bandwidth", "MByte per Second", "max"],
"Bidir_Put":
["MPI-2 Bidirectional 'Put' Bandwidth", "MByte per Second", "max"],
"Accumulate": ["MPI-2 'Accumulate' Latency", "us", "min"]
}
# read output
lines = []
if os.path.isfile(appstdout):
fin = open(appstdout, "rt")
lines = fin.readlines()
fin.close()
# process the output
parser.successfulRun = False
aggregate_mode = None
metric = None
j = -1
while j < len(lines) - 1:
j += 1
m = re.search(r'All processes entering MPI_Finalize', lines[j])
if m:
parser.successfulRun = True
m = re.match(r'^# Benchmarking\s+(\S+)', lines[j])
if m:
if m.group(1) in metrics:
metric = m.group(1)
continue
m = re.match(r'^#\s+MODE:\s+(\S+)', lines[j])
if m and metric and aggregate_mode is None:
aggregate_mode = m.group(1)
continue
m = re.match(r'^# (.+): (.+)', lines[j])
if m: # benchmark parameters
param = m.group(1).strip()
if param in params:
val = m.group(2).strip()
v = params[param][2]
if v.find('<val>') >= 0:
val = get_float_or_int(val)
val = eval(v.replace('<val>', 'val'))
parser.set_parameter("App:" + params[param][0],
str(val) + " ", params[param][1])
continue
m = re.match(r'^\s+([1-9]\d*)\s+\d+', lines[j])
if m and metric: # this effectively skips the first line of result, which has #bytes = 0
results = []
while m:
numbers = lines[j].split()
results.append(
float(numbers[-1]
)) # tokenize the line, and extract the last column
j += 1
if j < len(lines):
m = re.match(r'^\s+([1-9]\d*)\s+\d+', lines[j])
if lines[j].count('IMB_init_buffers_iter') > 0:
break
else:
break
metric_name = metrics[metric][0]
if aggregate_mode:
metric_name += " (" + aggregate_mode.lower() + ")"
if len(results) > 0:
if metrics[metric][1] == 'us':
statname = metrics[metric][2][0].upper(
) + metrics[metric][2][1:] + " " + metric_name
statval = eval(metrics[metric][2] + "(results)")
parser.set_statistic(statname, statval * 1e-6, "Second")
else:
statname = metrics[metric][2][0].upper(
) + metrics[metric][2][1:] + " " + metric_name
statval = eval(metrics[metric][2] + "(results)")
parser.set_statistic(statname, statval, metrics[metric][1])
aggregate_mode = None
metric = None
if parser.get_parameter("App:MPI Thread Environment") is None:
parser.set_parameter("App:MPI Thread Environment", "")
if __name__ == "__main__":
# output for testing purpose
print("parsing complete:", parser.parsing_complete(verbose=True))
parser.print_params_stats_as_must_have()
print(parser.get_xml())
# Print out missing parameters for debug purpose
parser.parsing_complete(verbose=True)
# return complete XML overwize return None
return parser.get_xml()
def process_appker_output(appstdout=None,
stdout=None,
stderr=None,
geninfo=None,
proclog=None,
resource_appker_vars=None):
# set App Kernel Description
parser = AppKerOutputParser(name='akrr network check',
version=1,
description="network benchmarking",
url='http://www.xdmod.org',
measurement_name='akrr network check')
# set obligatory parameters and statistics
# set common parameters and statistics
parser.add_must_have_parameter('App:ExeBinSignature')
parser.add_must_have_statistic('Ping, Mean')
parser.add_must_have_statistic('Secure Copy Bandwidth (in), Mean')
parser.add_must_have_statistic('Secure Copy Bandwidth (out), Mean')
parser.add_must_have_statistic('WGet Bandwidth, Mean')
parser.add_must_have_statistic('Wall Clock Time')
# set app kernel custom sets
parser.add_must_have_statistic('Wall Clock Time')
# parse common parameters and statistics
parser.parse_common_params_and_stats(appstdout, stdout, stderr, geninfo,
resource_appker_vars)
if hasattr(parser, 'wallClockTime') and parser.wallClockTime is not None:
parser.set_statistic("Wall Clock Time",
total_seconds(parser.wallClockTime), "Second")
if hasattr(
parser,
'appKerWallClockTime') and parser.appKerWallClockTime is not None:
parser.set_statistic("Wall Clock Time",
total_seconds(parser.appKerWallClockTime),
"Second")
# read output
lines = []
if os.path.isfile(appstdout):
fin = open(appstdout, "rt")
lines = fin.readlines()
fin.close()
# process the output
successful_run = False
j = 0
start = None
while j < len(lines):
if lines[j].strip() == "AKRR Network Check Results:":
start = j
if lines[j].strip() == "Done":
end = j
j += 1
if start is not None and end is not None:
r = json.loads(" ".join(lines[(start + 1):end]))
successful_run = True
if 'ping' in r:
count = 0
ping = 0.0
for k, v in r['ping'].items():
if v is None:
successful_run = False
else:
ping += float(v['rtt_avg'])
count += 1
parser.set_statistic("Ping, Mean", ping / count, "ms")
if 'wget' in r:
count = 0
val = 0.0
for k, v in r['wget'].items():
if v is None:
successful_run = False
else:
val += float(v['bandwidth'])
count += 1
parser.set_statistic("WGet Bandwidth, Mean", val / count, "MB/s")
if 'scp' in r:
count = 0
val_in = 0.0
val_out = 0.0
for k, v in r['scp'].items():
if v is None:
successful_run = False
else:
val_in += float(v['bandwidth_ingress'])
val_out += float(v['bandwidth_egress'])
count += 1
parser.set_statistic("Secure Copy Bandwidth (in), Mean",
val_in / count, "MB/s")
parser.set_statistic("Secure Copy Bandwidth (out), Mean",
val_out / count, "MB/s")
parser.successfulRun = successful_run
if __name__ == "__main__":
# output for testing purpose
print("parsing complete:", parser.parsing_complete())
parser.print_params_stats_as_must_have()
print(parser.get_xml())
# return complete XML overwize return None
return parser.get_xml()
def push_to_db_raw(self, cur, task_id, time_finished):
print("Pushing to DB")
import xml.etree.ElementTree as ElementTree
import xml.dom.minidom
result_file = os.path.join(self.taskDir, "result.xml")
jobfiles_dir = os.path.join(self.taskDir, "jobfiles")
(batch_job_dir, stdout_file, stderr_file, appstdout_file,
taskexeclog_file) = self.get_result_files()
# sanity check
fin = open(result_file, "r")
content = fin.read()
fin.close()
if content[0] != '<':
# need to reformat
i0 = content.find("<rep:report")
i1 = content.find("</rep:report>")
fout = open(result_file, "w")
content = fout.write("<?xml version='1.0'?>\n" +
content[i0:i1 + len("</rep:report>")] + "\n")
fout.close()
try:
tree = ElementTree.parse(result_file)
tree.getroot()
except:
self.status_info = "(1)==Resulting XML file content==\n" + content + \
"\n==Previous status==" + self.status + \
"\n==Previous status info==" + self.status_info + "\n" + \
traceback.format_exc()
self.status = "Cannot process final XML file"
self.write_error_xml(result_file, cdata=True)
instance_id = task_id
message = None
stderr = None
job_id = None
cur.execute(
"""SELECT instance_id,collected,committed,resource,executionhost,reporter,
reporternickname,status,message,stderr,body,memory,cputime,walltime,job_id
FROM akrr_xdmod_instanceinfo WHERE instance_id=%s""", (task_id, ))
raw = cur.fetchone()
if raw is not None: # .i.e. not new entry
(instance_id, collected, committed, resource, executionhost,
reporter, reporternickname, status, message, stderr, body, memory,
cputime, walltime, job_id) = raw
if self.RemoteJobID is not None:
job_id = self.RemoteJobID
else:
collected = time_finished
committed = datetime.datetime.today().strftime("%Y-%m-%d %H:%M:%S")
resource = self.resourceName
executionhost = self.resource.get('__regexp__', self.resourceName)
reporter = self.appName
reporternickname = akrr.util.replace_at_var_at(
self.app['nickname'],
[self.resource, self.app, self.resourceParam, self.appParam])
if hasattr(self, "RemoteJobID"):
job_id = self.RemoteJobID
# Process XML file
try:
tree = ElementTree.parse(result_file)
root = tree.getroot()
except:
self.status_info = "(2)==Resulting XML file content==\n" + content + \
"\n==Previous status==" + self.status + \
"\n==Previous status info==" + self.status_info
self.status = "Cannot process final XML file"
root = None
memory = 0.0
cputime = 0.0
walltime = 0.0
if root is not None:
try:
t = root.find('exitStatus').find('completed').text
if t.upper() == "TRUE":
completed = True
else:
completed = False
if completed:
root.find('body').find('performance').find(
'benchmark').find('statistics')
self.status = "Task was completed successfully."
self.status_info = "Done"
except:
completed = None
try:
t = root.find('xdtas').find('batchJob').find('status').text
if t.upper() == "ERROR":
error = True
else:
error = False
except:
error = None
status = 0
if completed is not None:
if completed:
status = 1
if completed is not None:
if completed is True:
body_et = root.find('body').find('performance')
body = ElementTree.tostring(body_et)
xml = xml.dom.minidom.parseString(body)
body = xml.toprettyxml()
body = body.replace("""<?xml version="1.0" ?>\n""", "")
statistics_et = root.find('body').find('performance').find(
'benchmark').find('statistics')
for e in statistics_et:
if e.find('ID').text == 'Memory':
memory = float(e.find('value').text)
if e.find('ID').text == 'Wall Clock Time':
walltime = float(e.find('value').text)
else:
# i.e. error caught by parser on resource
try:
message = "ERROR: error caught by parser"
stderr = root.find('exitStatus').find(
'errorMessage').text
body = """<xdtas>
<batchJob>
<status>Error</status>
<errorCause>ERROR: error caught by parser</errorCause>
<reporter>%s</reporter>
<errorMsg>%s</errorMsg>
</batchJob>
</xdtas>
""" % (reporter, stderr)
except:
message = "ERROR: error caught by parser"
stderr = None
body = """<xdtas>
<batchJob>
<status>Error</status>
<errorCause>ERROR: error caught by parser</errorCause>
<reporter>%s</reporter>
<errorMsg></errorMsg>
</batchJob>
</xdtas>
""" % reporter
elif completed is None:
if error is not None: # i.e. xml with error generated afterwards
if error is True:
# fin=open(resultFile,"r")
# body=fin.read()
# fin.close
message = root.find('xdtas').find('batchJob').find(
'errorCause').text
stderr = root.find('xdtas').find('batchJob').find(
'errorMsg').text
body_et = root.find('xdtas')
body = ElementTree.tostring(body_et)
else:
raise IOError("This condition should not happens")
else:
raise IOError("This condition should not happens")
else:
raise IOError("This condition should not happens")
# memory=None
# cputime=None
# walltime=None
else: # i.e. xml was not parsed
status = 0
message = self.status
stderr = self.status_info
body = """<xdtas>
<batchJob>
<status>Error</status>
<errorCause>Unknown error</errorCause>
<reporter>AKRR Server</reporter>
<errorMsg>"Cannot process final XML file</errorMsg>
</batchJob>
</xdtas>
"""
message = akrr.util.clean_unicode(message)
stderr = akrr.util.clean_unicode(stderr)
body = akrr.util.clean_unicode(body)
# Get Nodes
nodes = None
nodes_file_name = os.path.join(jobfiles_dir, "nodes")
if os.path.isfile(nodes_file_name):
parser = AppKerOutputParser()
parser.parse_common_params_and_stats(
geninfo=os.path.join(batch_job_dir, "gen.info"))
nodes = ";"
for line in parser.nodesList:
line = line.strip()
nodes += "%s;" % line
if len(nodes.strip().strip(';')) == 0:
nodes = None
if raw is not None: # .i.e. new entry
print("Updating")
cur.execute(
"""UPDATE akrr_xdmod_instanceinfo
SET instance_id=%s,collected=%s,committed=%s,resource=%s,executionhost=%s,reporter=%s,
reporternickname=%s,status=%s,message=%s,stderr=%s,body=%s,memory=%s,cputime=%s,walltime=%s,job_id=%s,nodes=%s
WHERE instance_id=%s""",
(instance_id, collected, committed, resource, executionhost,
reporter, reporternickname, status, message, stderr, body,
memory, cputime, walltime, job_id, nodes, instance_id))
else:
cur.execute(
"""INSERT INTO akrr_xdmod_instanceinfo
(instance_id,collected,committed,resource,executionhost,reporter,reporternickname,status,message,stderr,body,memory,cputime,walltime,job_id,nodes)
VALUES (%s,%s,%s,%s,%s,%s,%s,%s,%s,%s,%s,%s,%s,%s,%s,%s)""",
(instance_id, collected, committed, resource, executionhost,
reporter, reporternickname, status, message, stderr, body,
memory, cputime, walltime, job_id, nodes))
if appstdout_file is not None:
fin = open(appstdout_file, "r")
appstdout_file_content = fin.read()
fin.close()
else:
appstdout_file_content = "Does Not Present"
if stdout_file is not None:
fin = open(stdout_file, "r")
stdout_file_content = fin.read()
fin.close()
else:
stdout_file_content = "Does Not Present"
if stderr_file is not None:
fin = open(stderr_file, "r")
stderr_file_content = fin.read()
fin.close()
else:
stderr_file_content = "Does Not Present"
if taskexeclog_file is not None:
fin = open(taskexeclog_file, "r")
taskexeclog_file_content = fin.read()
fin.close()
else:
taskexeclog_file_content = "Does Not Present"
cur.execute(
"""INSERT INTO akrr_errmsg
(task_id,appstdout,stderr,stdout,taskexeclog)
VALUES (%s,%s,%s,%s,%s)""",
(instance_id, appstdout_file_content, stderr_file_content,
stdout_file_content, taskexeclog_file_content))
def process_appker_output(appstdout=None,
stdout=None,
stderr=None,
geninfo=None,
resource_appker_vars=None):
# set App Kernel Description
if resource_appker_vars is not None and 'app' in resource_appker_vars and 'name' in resource_appker_vars[
'app']:
akname = resource_appker_vars['app']['name']
else:
akname = 'unknown'
# initiate parser
parser = AppKerOutputParser(name=akname)
# set obligatory parameters and statistics
# set common parameters and statistics (App:ExeBinSignature and RunEnv:Nodes)
parser.add_common_must_have_params_and_stats()
# set app kernel custom sets
parser.add_must_have_parameter('App:ExeBinSignature')
parser.add_must_have_parameter('App:Version')
parser.add_must_have_parameter('Number of Darts Throws')
parser.add_must_have_parameter('Number of Rounds')
parser.add_must_have_parameter('RunEnv:Nodes')
parser.add_must_have_statistic('Darts Throws per Second')
parser.add_must_have_statistic('Time for PI Calculation')
parser.add_must_have_statistic('Wall Clock Time')
# parse common parameters and statistics
parser.parse_common_params_and_stats(appstdout, stdout, stderr, geninfo)
if hasattr(parser, 'appKerWallClockTime'):
parser.set_statistic("Wall Clock Time",
parser.appKerWallClockTime.total_seconds(),
"Second")
# Here can be custom output parsing
# read output
lines = []
if os.path.isfile(appstdout):
fin = open(appstdout, "rt")
lines = fin.readlines()
fin.close()
# process the output
parser.successfulRun = False
j = 0
while j < len(lines):
m = re.search(r'version:\s+(.+)', lines[j])
if m:
parser.set_parameter('App:Version', m.group(1))
m = re.search(r'number of throws at dartboard:\s+(\d+)', lines[j])
if m:
parser.set_parameter('Number of Darts Throws', m.group(1))
m = re.search(r'number of rounds for dartz throwing\s+(\d+)', lines[j])
if m:
parser.set_parameter('Number of Rounds', m.group(1))
m = re.search(r'Time for PI calculation:\s+([0-9.]+)', lines[j])
if m:
parser.set_statistic("Time for PI Calculation", m.group(1),
"Seconds")
m = re.search(r'Giga Darts Throws per Second \(GDaPS\):\s+([0-9.]+)',
lines[j])
if m:
parser.set_statistic("Darts Throws per Second", m.group(1),
"GDaPS")
m = re.search(r'Giga Darts Throws per Second', lines[j])
if m:
parser.successfulRun = True
j += 1
if __name__ == "__main__":
# output for testing purpose
print(("Parsing complete:", parser.parsing_complete(verbose=True)))
print("Following statistics and parameter can be set as obligatory:")
parser.print_params_stats_as_must_have()
print("\nResulting XML:")
print((parser.get_xml()))
# return complete XML otherwise return None
return parser.get_xml()
def process_appker_output(appstdout=None,
stdout=None,
stderr=None,
geninfo=None,
proclog=None,
resource_appker_vars=None):
# set App Kernel Description
parser = AppKerOutputParser(
name='namd',
version=1,
description="NAMD: Scalable Molecular Dynamics Package",
url='http://www.ks.uiuc.edu/Research/namd/',
measurement_name='NAMD')
# set obligatory parameters and statistics
# set common parameters and statistics
parser.add_common_must_have_params_and_stats()
# set app kernel custom sets
parser.add_must_have_parameter('App:Version')
parser.add_must_have_parameter('Input:Coordinate File')
parser.add_must_have_parameter('Input:Number of Angles')
parser.add_must_have_parameter('Input:Number of Atoms')
parser.add_must_have_parameter('Input:Number of Bonds')
parser.add_must_have_parameter('Input:Number of Dihedrals')
parser.add_must_have_parameter('Input:Number of Steps')
parser.add_must_have_parameter('Input:Structure File')
parser.add_must_have_parameter('Input:Timestep')
parser.add_must_have_statistic('Memory')
parser.add_must_have_statistic('Molecular Dynamics Simulation Performance')
parser.add_must_have_statistic('Wall Clock Time')
# parse common parameters and statistics
parser.parse_common_params_and_stats(appstdout, stdout, stderr, geninfo,
resource_appker_vars)
# read output
lines = []
if os.path.isfile(appstdout):
fin = open(appstdout, "rt")
lines = fin.readlines()
fin.close()
# process the output
successful_run = False
j = 0
while j < len(lines):
m = re.match(r'^Info: NAMD ([0-9a-zA-Z.]+)', lines[j])
if m:
parser.set_parameter("App:Version", m.group(1))
m = re.match(r'^Info: TIMESTEP\s+([0-9.]+)', lines[j])
if m:
parser.set_parameter("Input:Timestep",
m.group(1) + "e-15", "Second per Step")
m = re.match(r'^Info: NUMBER OF STEPS\s+([0-9.]+)', lines[j])
if m:
parser.set_parameter("Input:Number of Steps", m.group(1))
m = re.match(r'^Info: COORDINATE PDB\s+(.+)', lines[j])
if m:
parser.set_parameter("Input:Coordinate File", m.group(1))
m = re.match(r'^Info: STRUCTURE FILE\s+(.+)', lines[j])
if m:
parser.set_parameter("Input:Structure File", m.group(1))
m = re.match(
r'^Info: Running on ([0-9.]+) processors, ([0-9.]+) nodes, ([0-9.]+) physical nodes.',
lines[j])
if m:
parser.set_parameter("App:NCores", m.group(1).strip())
parser.set_parameter("App:NNodes", m.group(3).strip())
if re.match(r'^Info: STRUCTURE SUMMARY', lines[j]):
j += 1
for k in range(25):
if re.match(r'^Info: \*\*\*\*\*', lines[j]):
break
m = re.match(r'^Info:\s+([0-9]+)\s+ATOMS\n', lines[j])
if m:
parser.set_parameter("Input:Number of Atoms", m.group(1))
m = re.match(r'^Info:\s+([0-9]+)\s+BONDS\n', lines[j])
if m:
parser.set_parameter("Input:Number of Bonds", m.group(1))
m = re.match(r'^Info:\s+([0-9]+)\s+ANGLES\n', lines[j])
if m:
parser.set_parameter("Input:Number of Angles", m.group(1))
m = re.match(r'^Info:\s+([0-9]+)\s+DIHEDRALS\n', lines[j])
if m:
parser.set_parameter("Input:Number of Dihedrals",
m.group(1))
j += 1
if re.search(r'Info: Benchmark time:', lines[j]):
m = re.search(r' ([0-9.]+) days/ns', lines[j])
if m:
parser.set_statistic(
"Molecular Dynamics Simulation Performance",
str(1.0e-9 / float(m.group(1))), "Second per Day")
m = re.match(
r'^WallClock:\s+([0-9.]+)\s+CPUTime:\s+([0-9.]+)\s+Memory:\s+([0-9.]+)',
lines[j])
if m:
parser.set_statistic("Wall Clock Time", m.group(1), "Second")
parser.set_statistic("Memory", m.group(3), "MByte")
successful_run = True
m = re.match(r'^End of program', lines[j])
if m:
successful_run = True
j += 1
parser.successfulRun = successful_run
if __name__ == "__main__":
# output for testing purpose
print("parsing complete:", parser.parsing_complete())
parser.print_params_stats_as_must_have()
print(parser.get_xml())
# return complete XML overwize return None
return parser.get_xml()
def process_appker_output(appstdout=None, stdout=None, stderr=None, geninfo=None, resource_appker_vars=None):
# set App Kernel Description
parser = AppKerOutputParser(
name='enzo',
version=1,
description="Enzo: an Adaptive Mesh Refinement Code for Astrophysics",
url='http://enzo-project.org',
measurement_name='Enzo'
)
# set obligatory parameters and statistics
# set common parameters and statistics
parser.add_common_must_have_params_and_stats()
# set app kernel custom sets
parser.add_must_have_parameter('App:Version')
parser.add_must_have_statistic('Wall Clock Time')
# parse common parameters and statistics
parser.parse_common_params_and_stats(appstdout, stdout, stderr, geninfo, resource_appker_vars)
# read output
lines = []
if os.path.isfile(appstdout):
fin = open(appstdout, "rt")
lines = fin.readlines()
fin.close()
parser.set_parameter("App:Version", "unknown")
# process the output
successful_run = False
j = 0
while j < len(lines):
m = re.match(r'^Mercurial Branch\s+(\S+)', lines[j])
if m:
branch = m.group(1)
revision = ""
if j + 1 < len(lines):
m = re.match(r'^Mercurial Revision\s+(\S+)', lines[j + 1])
if m:
revision = m.group(1)
parser.set_parameter("App:Version", "Branch:" + branch + " Revision:" + revision)
m = re.match(r'^Time\s*=\s*([0-9.]+)\s+CycleNumber\s*=\s*([0-9]+)\s+Wallclock\s*=\s*([0-9.]+)', lines[j])
if m:
parser.set_statistic("Final Simulation Time", m.group(1), "Enzo Time Unit")
parser.set_statistic("Total Cycles", m.group(2))
parser.set_statistic("Wall Clock Time", m.group(3), "Second")
successful_run = True
m = re.match(r'^Successful run, exiting.', lines[j])
if m:
successful_run = True
# performance
m = re.match(r'^Cycle_Number\s+([0-9]+)', lines[j])
if m:
j += 1
performance_metrics = {}
while j < len(lines):
if lines[j].strip() != "":
v = lines[j].strip().split()
if v[0] not in performance_metrics:
performance_metrics[v[0]] = float(v[1])
else:
performance_metrics[v[0]] += float(v[1])
else:
if j + 1 < len(lines):
m = re.match(r'^Cycle_Number\s+([0-9]+)', lines[j + 1])
if m:
pass
else:
break
else:
break
j += 1
metric = "CommunicationTranspose"
if metric in performance_metrics:
parser.set_statistic("Communication Transpose Time", performance_metrics[metric], "Second")
metric = "ComputePotentialFieldLevelZero"
if metric in performance_metrics:
parser.set_statistic("Gravitational Potential Field Computing Time", performance_metrics[metric],
"Second")
metric = "EvolvePhotons"
if metric in performance_metrics:
parser.set_statistic("Radiative Transfer Calculation Time", performance_metrics[metric], "Second")
metric = "Group_WriteAllData"
if metric in performance_metrics:
parser.set_statistic("All Data Group Write Time", performance_metrics[metric], "Second")
metric = "Level_00"
if metric in performance_metrics:
parser.set_statistic("All Grid Level 00 Calculation Time", performance_metrics[metric], "Second")
metric = "Level_01"
if metric in performance_metrics:
parser.set_statistic("All Grid Level 01 Calculation Time", performance_metrics[metric], "Second")
metric = "Level_02"
if metric in performance_metrics:
parser.set_statistic("All Grid Level 02 Calculation Time", performance_metrics[metric], "Second")
metric = "RebuildHierarchy"
if metric in performance_metrics:
parser.set_statistic("Grid Hierarchy Rebuilding Time", performance_metrics[metric], "Second")
metric = "SetBoundaryConditions"
if metric in performance_metrics:
parser.set_statistic("Boundary Conditions Setting Time", performance_metrics[metric], "Second")
metric = "SolveForPotential"
if metric in performance_metrics:
parser.set_statistic("Poisson Equation Solving Time", performance_metrics[metric], "Second")
metric = "SolveHydroEquations"
if metric in performance_metrics:
parser.set_statistic("Hydro Equations Solving Time", performance_metrics[metric], "Second")
metric = "Total"
if metric in performance_metrics:
parser.set_statistic("Total Time Spent in Cycles", performance_metrics[metric], "Second")
j += 1
parser.successfulRun = successful_run
if __name__ == "__main__":
# output for testing purpose
print(("parsing complete:", parser.parsing_complete()))
parser.print_params_stats_as_must_have()
print((parser.get_xml()))
# return complete XML overwize return None
return parser.get_xml()
def process_appker_output(appstdout=None,
stdout=None,
stderr=None,
geninfo=None,
proclog=None,
resource_appker_vars=None):
# set App Kernel Description
parser = AppKerOutputParser(name='hpcg',
version=1,
description="HPCG Benchmark",
url='http://www.hpcg-benchmark.org/index.html',
measurement_name='HPCG')
# set obligatory parameters and statistics
# set common parameters and statistics
parser.add_common_must_have_params_and_stats()
# set app kernel custom sets
parser.add_must_have_parameter('App:ExeBinSignature')
parser.add_must_have_parameter('App:Version')
parser.add_must_have_parameter('Input:Distributed Processes')
parser.add_must_have_parameter('Input:Global Problem Dimensions Nx')
parser.add_must_have_parameter('Input:Global Problem Dimensions Ny')
parser.add_must_have_parameter('Input:Global Problem Dimensions Nz')
parser.add_must_have_parameter('Input:Local Domain Dimensions Nx')
parser.add_must_have_parameter('Input:Local Domain Dimensions Ny')
parser.add_must_have_parameter('Input:Local Domain Dimensions Nz')
parser.add_must_have_parameter('Input:Number of Coarse Grid Levels')
parser.add_must_have_parameter('Input:Threads per processes')
parser.add_must_have_parameter('RunEnv:CPU Speed')
parser.add_must_have_parameter('RunEnv:Nodes')
parser.add_must_have_statistic('Floating-Point Performance, Raw DDOT')
parser.add_must_have_statistic('Floating-Point Performance, Raw MG')
parser.add_must_have_statistic('Floating-Point Performance, Raw SpMV')
parser.add_must_have_statistic('Floating-Point Performance, Raw Total')
parser.add_must_have_statistic('Floating-Point Performance, Raw WAXPBY')
parser.add_must_have_statistic('Floating-Point Performance, Total')
parser.add_must_have_statistic('Memory Bandwidth, Read')
parser.add_must_have_statistic('Memory Bandwidth, Total')
parser.add_must_have_statistic('Memory Bandwidth, Write')
parser.add_must_have_statistic('Setup Time')
parser.add_must_have_statistic('Wall Clock Time')
# parse common parameters and statistics
parser.parse_common_params_and_stats(appstdout, stdout, stderr, geninfo,
resource_appker_vars)
if hasattr(parser, 'appKerWallClockTime'):
parser.set_statistic("Wall Clock Time",
total_seconds(parser.appKerWallClockTime),
"Second")
# get path to YAML file
# read data
lines = []
if os.path.isfile(appstdout):
fin = open(appstdout, "rt")
lines = fin.readlines()
fin.close()
# Parse YAML lines because YAML is often malformed
yaml_lines = []
# get yaml lines from appstdout
bool_in_yaml_section = False
for line in lines:
if re.match(r"^====== .+\.yaml End ======", line):
break
if bool_in_yaml_section:
yaml_lines.append(line)
if re.match(r"^====== .+\.yaml Start ======", line):
bool_in_yaml_section = True
from pprint import pprint
import yaml
# fix some issues with yaml
if re.search(r"After confirmation please upload results from the YAML",
yaml_lines[-1]):
yaml_lines.pop()
if re.search(r"You have selected the QuickPath option", yaml_lines[-1]):
yaml_lines.pop()
yaml_text = "".join(yaml_lines)
yaml_text = re.sub(r"^ {6}HPCG 2\.4 Rating \(for historical value\) is:",
" HPCG 2.4 Rating (for historical value) is:",
yaml_text,
flags=re.M)
results_yaml = yaml.load(yaml_text)
# Set Parameters
# App version
app_version_list = []
for ver in [x for x in results_yaml.keys() if re.search("version", x)]:
app_version_list.append(ver + " " + str(results_yaml[ver]))
app_version = ", ".join(app_version_list)
parser.set_parameter('App:Version', app_version)
# Problem size
parser.set_parameter(
'Input:Number of Coarse Grid Levels',
results_yaml['Multigrid Information']['Number of coarse grid levels'])
parser.set_parameter(
'Input:Global Problem Dimensions Nx',
results_yaml['Global Problem Dimensions']['Global nx'])
parser.set_parameter(
'Input:Global Problem Dimensions Ny',
results_yaml['Global Problem Dimensions']['Global ny'])
parser.set_parameter(
'Input:Global Problem Dimensions Nz',
results_yaml['Global Problem Dimensions']['Global nz'])
parser.set_parameter('Input:Local Domain Dimensions Nx',
results_yaml['Local Domain Dimensions']['nx'])
parser.set_parameter('Input:Local Domain Dimensions Ny',
results_yaml['Local Domain Dimensions']['ny'])
parser.set_parameter('Input:Local Domain Dimensions Nz',
results_yaml['Local Domain Dimensions']['nz'])
parser.set_parameter(
'Input:Distributed Processes',
results_yaml['Machine Summary']['Distributed Processes'])
parser.set_parameter(
'Input:Threads per processes',
results_yaml['Machine Summary']['Threads per processes'])
if "cpu_speed" in parser.geninfo:
ll = parser.geninfo["cpu_speed"].splitlines()
cpu_speed_max = 0.0
for l in ll:
m = re.search(r'([\d.]+)$', l)
if m:
v = float(m.group(1).strip())
if v > cpu_speed_max:
cpu_speed_max = v
if cpu_speed_max > 0.0:
parser.set_parameter("RunEnv:CPU Speed", cpu_speed_max, "MHz")
# Set Statistics
parser.successfulRun = results_yaml['Reproducibility Information'][
'Result'] == 'PASSED'
parser.set_statistic('Setup Time',
results_yaml['Setup Information']['Setup Time'],
'Seconds')
parser.set_statistic('Memory Bandwidth, Read',
results_yaml['GB/s Summary']['Raw Read B/W'], 'GB/s')
parser.set_statistic('Memory Bandwidth, Write',
results_yaml['GB/s Summary']['Raw Write B/W'], 'GB/s')
parser.set_statistic('Memory Bandwidth, Total',
results_yaml['GB/s Summary']['Raw Total B/W'], 'GB/s')
parser.set_statistic(
'Floating-Point Performance, Total',
results_yaml['__________ Final Summary __________']
['HPCG result is VALID with a GFLOP/s rating of'], 'GFLOP/s')
parser.set_statistic('Floating-Point Performance, Raw DDOT',
results_yaml['GFLOP/s Summary']['Raw DDOT'],
'GFLOP/s')
parser.set_statistic('Floating-Point Performance, Raw WAXPBY',
results_yaml['GFLOP/s Summary']['Raw WAXPBY'],
'GFLOP/s')
parser.set_statistic('Floating-Point Performance, Raw SpMV',
results_yaml['GFLOP/s Summary']['Raw SpMV'],
'GFLOP/s')
parser.set_statistic('Floating-Point Performance, Raw MG',
results_yaml['GFLOP/s Summary']['Raw MG'], 'GFLOP/s')
parser.set_statistic('Floating-Point Performance, Raw Total',
results_yaml['GFLOP/s Summary']['Raw Total'],
'GFLOP/s')
if __name__ == "__main__":
# output for testing purpose
print("parsing complete:", parser.parsing_complete(verbose=True))
parser.print_params_stats_as_must_have()
print(parser.get_xml())
# return complete XML otherwise return None
return parser.get_xml()
def process_appker_output(appstdout=None,
stdout=None,
stderr=None,
geninfo=None,
resource_appker_vars=None):
# set App Kernel Description
parser = AppKerOutputParser(name='hpcc',
version=1,
description="HPC Challenge Benchmarks",
url='http://icl.cs.utk.edu/hpcc/',
measurement_name='xdmod.benchmark.hpcc')
# set obligatory parameters and statistics
# set common parameters and statistics
parser.add_common_must_have_params_and_stats()
# set app kernel custom sets
parser.add_must_have_parameter('App:Version')
parser.add_must_have_parameter('Input:DGEMM Problem Size')
parser.add_must_have_parameter('Input:High Performance LINPACK Grid Cols')
parser.add_must_have_parameter('Input:High Performance LINPACK Grid Rows')
parser.add_must_have_parameter(
'Input:High Performance LINPACK Problem Size')
parser.add_must_have_parameter('Input:MPI Ranks')
parser.add_must_have_parameter('Input:MPIRandom Problem Size')
parser.add_must_have_parameter('Input:OpenMP Threads')
parser.add_must_have_parameter('Input:PTRANS Problem Size')
parser.add_must_have_parameter('Input:STREAM Array Size')
parser.add_must_have_parameter('RunEnv:CPU Speed')
parser.add_must_have_parameter('RunEnv:Nodes')
parser.add_must_have_statistic(
'Average Double-Precision General Matrix Multiplication (DGEMM) Floating-Point Performance'
)
parser.add_must_have_statistic("Average STREAM 'Add' Memory Bandwidth")
parser.add_must_have_statistic("Average STREAM 'Copy' Memory Bandwidth")
parser.add_must_have_statistic("Average STREAM 'Scale' Memory Bandwidth")
parser.add_must_have_statistic("Average STREAM 'Triad' Memory Bandwidth")
parser.add_must_have_statistic(
'Fast Fourier Transform (FFTW) Floating-Point Performance')
parser.add_must_have_statistic('High Performance LINPACK Efficiency')
parser.add_must_have_statistic(
'High Performance LINPACK Floating-Point Performance')
parser.add_must_have_statistic('High Performance LINPACK Run Time')
parser.add_must_have_statistic('MPI Random Access')
parser.add_must_have_statistic('Parallel Matrix Transpose (PTRANS)')
parser.add_must_have_statistic('Wall Clock Time')
# parse common parameters and statistics
parser.parse_common_params_and_stats(appstdout, stdout, stderr, geninfo,
resource_appker_vars)
if parser.appKerWallClockTime is not None:
parser.set_statistic("Wall Clock Time",
total_seconds(parser.appKerWallClockTime),
"Second")
# Intel MPI benchmark suite contains three classes of benchmarks:
#
# Single-transfer, which needs only 2 processes
# Parallel-transfer, which can use as many processes that are available
# Collective, which can use as many processes that are available
# The parameters mapping table
params = {
"CommWorldProcs": ["MPI Ranks", "", ""],
"HPL_N": ["High Performance LINPACK Problem Size", "", ""],
"HPL_nprow": ["High Performance LINPACK Grid Rows", "", ""],
"HPL_npcol": ["High Performance LINPACK Grid Cols", "", ""],
"PTRANS_n": ["PTRANS Problem Size", "", ""],
"MPIRandomAccess_N":
["MPIRandom Problem Size", "MWord", "val/1024/1024"],
"STREAM_VectorSize": ["STREAM Array Size", "MWord", ""],
"DGEMM_N": ["DGEMM Problem Size", "", ""],
"omp_get_num_threads": ["OpenMP Threads", "", ""],
}
# The result mapping table
metrics = {
"HPL_Tflops": [
"High Performance LINPACK Floating-Point Performance",
"MFLOP per Second", "val*1e6"
],
"HPL_time": ["High Performance LINPACK Run Time", "Second", ""],
"PTRANS_GBs":
["Parallel Matrix Transpose (PTRANS)", "MByte per Second", "val*1024"],
"MPIRandomAccess_GUPs":
["MPI Random Access", "MUpdate per Second", "val*1000"],
"MPIFFT_Gflops": [
"Fast Fourier Transform (FFTW) Floating-Point Performance",
"MFLOP per Second", "val*1000"
],
"StarDGEMM_Gflops": [
"Average Double-Precision General Matrix Multiplication (DGEMM) Floating-Point Performance",
"MFLOP per Second", "val*1000"
],
"StarSTREAM_Copy": [
"Average STREAM 'Copy' Memory Bandwidth", "MByte per Second",
"val*1024"
],
"StarSTREAM_Scale": [
"Average STREAM 'Scale' Memory Bandwidth", "MByte per Second",
"val*1024"
],
"StarSTREAM_Add": [
"Average STREAM 'Add' Memory Bandwidth", "MByte per Second",
"val*1024"
],
"StarSTREAM_Triad": [
"Average STREAM 'Triad' Memory Bandwidth", "MByte per Second",
"val*1024"
]
}
# read output
lines = []
if os.path.isfile(appstdout):
fin = open(appstdout, "rt")
lines = fin.readlines()
fin.close()
# process the output
parser.successfulRun = False
result_begin = None
hpl_tflops = None
num_cores = None
values = {}
j = -1
while j < len(lines) - 1:
j += 1
m = re.search(r'End of HPC Challenge tests', lines[j])
if m:
parser.successfulRun = True
m = re.match(r'^Begin of Summary section', lines[j])
if m:
result_begin = 1
continue
m = re.match(r'^(\w+)=([\w.]+)', lines[j])
if m and result_begin:
metric_name = m.group(1).strip()
values[metric_name] = m.group(2).strip()
if metric_name == "HPL_Tflops":
hpl_tflops = float(values[metric_name])
if metric_name == "CommWorldProcs":
num_cores = int(values[metric_name])
m = re.match(r'^Running on ([0-9.]+) processors', lines[j])
if m:
num_cores = int(m.group(1).strip())
if hpl_tflops is None or num_cores is None:
parser.successfulRun = False
hpcc_version = None
mhz = None
theoretical_gflops = None
if "VersionMajor" in values and "VersionMinor" in values and "VersionMicro" in values:
hpcc_version = values["VersionMajor"] + "." + values[
"VersionMinor"] + "." + values["VersionMicro"]
if "VersionRelease" in values:
hpcc_version += values["VersionRelease"]
if hpcc_version:
parser.set_parameter("App:Version", hpcc_version)
for k, v in params.items():
if k not in values:
continue
val = values[k]
if v[2].find('val') >= 0:
# if convertion formula is used, then first set val variable and then eval the formula
val = get_float_or_int(values[k])
val = eval(v[2])
units = v[1] if [1] != "" else None
parser.set_parameter("Input:" + v[0], val, units)
for k, v in metrics.items():
if k not in values:
continue
val = values[k]
if v[2].find('val') >= 0:
# if convertion formula is used, then first set val variable and then eval the formula
val = get_float_or_int(values[k])
val = eval(v[2])
units = v[1] if [1] != "" else None
parser.set_statistic(v[0], val, units)
if "cpu_speed" in parser.geninfo:
ll = parser.geninfo["cpu_speed"].splitlines()
cpu_speed_max = 0.0
for l in ll:
m = re.search(r'([\d.]+)$', l)
if m:
v = float(m.group(1).strip())
if v > cpu_speed_max:
cpu_speed_max = v
if cpu_speed_max > 0.0:
parser.set_parameter("RunEnv:CPU Speed", cpu_speed_max, "MHz")
mhz = cpu_speed_max
if resource_appker_vars is not None:
if 'resource' in resource_appker_vars and 'app' in resource_appker_vars:
if 'theoreticalGFlopsPerCore' in resource_appker_vars['app']:
resname = resource_appker_vars['resource']['name']
if resname in resource_appker_vars['app'][
'theoreticalGFlopsPerCore']:
theoretical_gflops = resource_appker_vars['app'][
'theoreticalGFlopsPerCore'][resname] * num_cores
print("theoreticalGFlops", resname, theoretical_gflops)
if theoretical_gflops is None and mhz is not None:
# Most modern x86 & POWER processors are superscale and can issue 4 instructions per cycle
theoretical_gflops = mhz * num_cores * 4 / 1000.0
if theoretical_gflops and hpl_tflops:
# Convert both to GFlops and derive the Efficiency
percent = (1000.0 * hpl_tflops / theoretical_gflops) * 100.0
parser.set_statistic("High Performance LINPACK Efficiency",
"%.3f" % percent, "Percent")
if __name__ == "__main__":
# output for testing purpose
print("parsing complete:", parser.parsing_complete(verbose=True))
parser.print_params_stats_as_must_have()
print(parser.get_xml())
# return complete XML overwize return None
return parser.get_xml()
def process_appker_output(appstdout=None,
stdout=None,
stderr=None,
geninfo=None,
proclog=None,
resource_appker_vars=None):
# set App Kernel Description
parser = AppKerOutputParser(
name='gamess',
version=1,
description=
"Gamess: General Atomic and Molecular Electronic Structure System",
url='http://www.msg.ameslab.gov',
measurement_name='Gamess')
# set obligatory parameters and statistics
# set common parameters and statistics
parser.add_common_must_have_params_and_stats()
# set app kernel custom sets
parser.add_must_have_parameter('App:Version')
parser.add_must_have_statistic('Wall Clock Time')
parser.add_must_have_statistic('User Time')
parser.add_must_have_statistic('Time Spent in MP2 Energy Calculation')
parser.add_must_have_statistic(
'Time Spent in Restricted Hartree-Fock Calculation')
# parse common parameters and statistics
parser.parse_common_params_and_stats(appstdout, stdout, stderr, geninfo,
resource_appker_vars)
# read output
lines = []
if os.path.isfile(appstdout):
fin = open(appstdout, "rt")
lines = fin.readlines()
fin.close()
# process the output
start_time = None
end_time = None
mp2_energy_calculation_time = 0.0
rhf_calculation_time = 0.0
efficiency = None
j = 0
while j < len(lines):
m = re.search(r'GAMESS VERSION = ([^*]+)', lines[j])
if m:
parser.set_parameter("App:Version", m.group(1).strip())
m = re.search(
r'PARALLEL VERSION RUNNING ON\s*([\d.]+) PROCESSORS IN\s*([\d.]+) NODE',
lines[j])
if m:
parser.set_parameter("App:NCores", m.group(1).strip())
parser.set_parameter("App:NNodes", m.group(2).strip())
m = re.search(r'EXECUTION OF GAMESS BEGUN (.+)', lines[j])
if m:
start_time = parser.get_datetime_local(m.group(1).strip())
m = re.search(r'EXECUTION OF GAMESS TERMINATED NORMALLY (.+)',
lines[j])
if m:
end_time = parser.get_datetime_local(m.group(1).strip())
if re.search(r'DONE WITH MP2 ENERGY', lines[j]):
j += 1
m = re.search(r'STEP CPU TIME=\s*([\d.]+)', lines[j])
if m:
mp2_energy_calculation_time += float(m.group(1).strip())
if re.search(r'END OF RHF CALCULATION', lines[j]):
j += 1
m = re.search(r'STEP CPU TIME=\s*([\d.]+)', lines[j])
if m:
rhf_calculation_time += float(m.group(1).strip())
m = re.search(r'TOTAL WALL CLOCK TIME.+CPU UTILIZATION IS\s+([\d.]+)',
lines[j])
if m:
efficiency = float(m.group(1).strip())
j += 1
if start_time and end_time:
walltime = total_seconds(end_time - start_time)
if walltime >= 0.0:
parser.set_statistic('Wall Clock Time', str(walltime), "Second")
if efficiency:
parser.set_statistic("User Time",
str((0.01 * efficiency * walltime)),
"Second")
parser.set_statistic("Time Spent in MP2 Energy Calculation",
str(mp2_energy_calculation_time), "Second")
parser.set_statistic("Time Spent in Restricted Hartree-Fock Calculation",
str(rhf_calculation_time), "Second")
if "attemptsToLaunch" in parser.geninfo:
parser.set_statistic("Attempts to Launch",
parser.geninfo['attemptsToLaunch'])
else:
parser.set_statistic("Attempts to Launch", 1)
if __name__ == "__main__":
# output for testing purpose
print(("parsing complete:", parser.parsing_complete()))
parser.print_params_stats_as_must_have()
print((parser.get_xml()))
# return complete XML overwize return None
return parser.get_xml()
def process_appker_output(appstdout=None,
stdout=None,
stderr=None,
geninfo=None,
proclog=None,
resource_appker_vars=None):
# initiate parser
parser = AppKerOutputParser(name='mdtest')
# set obligatory parameters and statistics
# set common parameters and statistics (App:ExeBinSignature and RunEnv:Nodes)
parser.add_common_must_have_params_and_stats()
# set app kernel custom sets
parser.add_must_have_parameter('RunEnv:Nodes')
parser.add_must_have_parameter('Arguments (single directory per process)')
parser.add_must_have_parameter('Arguments (single directory)')
parser.add_must_have_parameter(
'Arguments (single tree directory per process)')
parser.add_must_have_parameter('Arguments (single tree directory)')
parser.add_must_have_parameter(
'files/directories (single directory per process)')
parser.add_must_have_parameter('files/directories (single directory)')
parser.add_must_have_parameter(
'files/directories (single tree directory per process)')
parser.add_must_have_parameter('files/directories (single tree directory)')
parser.add_must_have_parameter('tasks (single directory per process)')
parser.add_must_have_parameter('tasks (single directory)')
parser.add_must_have_parameter('tasks (single tree directory per process)')
parser.add_must_have_parameter('tasks (single tree directory)')
parser.add_must_have_statistic(
'Directory creation (single directory per process)')
parser.add_must_have_statistic('Directory creation (single directory)')
parser.add_must_have_statistic(
'Directory creation (single tree directory per process)')
parser.add_must_have_statistic(
'Directory creation (single tree directory)')
parser.add_must_have_statistic(
'Directory removal (single directory per process)')
parser.add_must_have_statistic('Directory removal (single directory)')
parser.add_must_have_statistic(
'Directory removal (single tree directory per process)')
parser.add_must_have_statistic('Directory removal (single tree directory)')
parser.add_must_have_statistic(
'Directory stat (single directory per process)')
parser.add_must_have_statistic('Directory stat (single directory)')
parser.add_must_have_statistic(
'Directory stat (single tree directory per process)')
parser.add_must_have_statistic('Directory stat (single tree directory)')
parser.add_must_have_statistic(
'File creation (single directory per process)')
parser.add_must_have_statistic('File creation (single directory)')
parser.add_must_have_statistic(
'File creation (single tree directory per process)')
parser.add_must_have_statistic('File creation (single tree directory)')
parser.add_must_have_statistic('File read (single directory per process)')
parser.add_must_have_statistic('File read (single directory)')
parser.add_must_have_statistic(
'File read (single tree directory per process)')
parser.add_must_have_statistic('File read (single tree directory)')
parser.add_must_have_statistic(
'File removal (single directory per process)')
parser.add_must_have_statistic('File removal (single directory)')
parser.add_must_have_statistic(
'File removal (single tree directory per process)')
parser.add_must_have_statistic('File removal (single tree directory)')
parser.add_must_have_statistic('File stat (single directory per process)')
parser.add_must_have_statistic('File stat (single directory)')
parser.add_must_have_statistic(
'File stat (single tree directory per process)')
parser.add_must_have_statistic('File stat (single tree directory)')
parser.add_must_have_statistic(
'Tree creation (single directory per process)')
parser.add_must_have_statistic('Tree creation (single directory)')
parser.add_must_have_statistic(
'Tree creation (single tree directory per process)')
parser.add_must_have_statistic('Tree creation (single tree directory)')
parser.add_must_have_statistic(
'Tree removal (single directory per process)')
parser.add_must_have_statistic('Tree removal (single directory)')
parser.add_must_have_statistic(
'Tree removal (single tree directory per process)')
parser.add_must_have_statistic('Tree removal (single tree directory)')
parser.add_must_have_statistic('Wall Clock Time')
# parse common parameters and statistics
parser.parse_common_params_and_stats(appstdout, stdout, stderr, geninfo,
resource_appker_vars)
if hasattr(parser, 'appKerWallClockTime'):
parser.set_statistic("Wall Clock Time",
total_seconds(parser.appKerWallClockTime),
"Second")
# Here can be custom output parsing
# read output
lines = []
if os.path.isfile(appstdout):
fin = open(appstdout, "rt")
lines = fin.readlines()
fin.close()
# process the output
testname = ""
parser.successfulRun = False
j = 0
while j < len(lines):
m = re.match(r'^#Testing (.+)', lines[j])
if m:
testname = " (" + m.group(1).strip() + ")"
m = re.match(r'^SUMMARY.*:', lines[j])
if m:
j = j + 3
while j < len(lines):
m = re.match(
r'([A-Za-z0-9 ]+):\s+[0-9.]+\s+[0-9.]+\s+([0-9.]+)\s+([0-9.]+)',
lines[j])
if m:
parser.set_statistic(
m.group(1).strip() + testname, m.group(2),
"Operations/Second")
else:
break
j = j + 1
m = re.search(r'finished at', lines[j])
if m:
parser.successfulRun = True
m = re.match(r'^Command line used:.+mdtest\s+(.+)', lines[j])
if m:
parser.set_parameter("Arguments" + testname, m.group(1).strip())
m = re.search(r'([0-9]+) tasks, ([0-9]+) files/directories', lines[j])
if m:
parser.set_parameter("tasks" + testname, m.group(1).strip())
parser.set_parameter("files/directories" + testname,
m.group(2).strip())
j = j + 1
# parser.set_parameter("mega parameter",m.group(1))
#
# m=re.search(r'My mega parameter\s+(\d+)',lines[j])
# if m:parser.set_statistic("mega statistics",m.group(1),"Seconds")
#
# m=re.search(r'Done',lines[j])
# if m:parser.successfulRun=True
#
# j+=1
if __name__ == "__main__":
# output for testing purpose
print("Parsing complete:", parser.parsing_complete(verbose=True))
print("Following statistics and parameter can be set as obligatory:")
parser.print_params_stats_as_must_have()
print("\nResulting XML:")
print(parser.get_xml())
# return complete XML otherwise return None
return parser.get_xml()
def process_appker_output(appstdout=None,
stdout=None,
stderr=None,
geninfo=None,
resource_appker_vars=None):
# set App Kernel Description
parser = AppKerOutputParser(
name='lammps',
version=1,
description=
"LAMMPS: Large-scale Atomic/Molecular Massively Parallel Simulator",
url='http://lammps.sandia.gov',
measurement_name='LAMMPS')
# set obligatory parameters and statistics
# set common parameters and statistics
parser.add_common_must_have_params_and_stats()
# set app kernel custom sets
parser.add_must_have_parameter('App:Version')
parser.add_must_have_parameter('Input:Number of Atoms')
parser.add_must_have_parameter('Input:Number of Steps')
parser.add_must_have_parameter('Input:Timestep')
parser.add_must_have_statistic('Molecular Dynamics Simulation Performance')
parser.add_must_have_statistic('Per-Process Memory')
parser.add_must_have_statistic('Time Spent in Bond Potential Calculation')
parser.add_must_have_statistic('Time Spent in Communication')
parser.add_must_have_statistic(
'Time Spent in Long-Range Coulomb Potential (K-Space) Calculation')
parser.add_must_have_statistic('Time Spent in Neighbor List Regeneration')
parser.add_must_have_statistic(
'Time Spent in Pairwise Potential Calculation')
parser.add_must_have_statistic('Wall Clock Time')
# parse common parameters and statistics
parser.parse_common_params_and_stats(appstdout, stdout, stderr, geninfo,
resource_appker_vars)
# read output
lines = []
if os.path.isfile(appstdout):
fin = open(appstdout, "rt")
lines = fin.readlines()
fin.close()
# process the output
parser.successfulRun = False
wall_clock_time = None
simulation_units = None
num_steps = None
step_size = None
j = 0
while j < len(lines):
m = re.match(r'^LAMMPS\s+\(([\w ]+)\)', lines[j])
if m:
parser.set_parameter("App:Version", m.group(1).strip())
m = re.match(r'^Memory usage per processor = ([\d.]+) Mbyte', lines[j])
if m:
parser.set_statistic("Per-Process Memory",
m.group(1).strip(), "MByte")
m = re.match(r'^Loop time of ([\d.]+) on', lines[j])
if m:
parser.successfulRun = True
wall_clock_time = float(m.group(1).strip())
parser.set_statistic("Wall Clock Time", wall_clock_time, "Second")
m1 = re.search(r'(\d+) atoms', lines[j])
if m1:
parser.set_parameter("Input:Number of Atoms",
m1.group(1).strip())
m = re.match(r'^units\s+(\w+)', lines[j])
if m:
simulation_units = m.group(1).strip().lower()
m = re.match(r'^run\s+(\d+)', lines[j])
if m:
num_steps = int(m.group(1).strip())
parser.set_parameter("Input:Number of Steps", num_steps)
m = re.match(r'^timestep\s+([\d.]+)', lines[j])
if m:
step_size = float(m.group(1).strip())
m = re.match(r'^Pair\s+time.+= ([\d.]+)', lines[j])
if parser.successfulRun and m:
parser.set_statistic(
"Time Spent in Pairwise Potential Calculation",
m.group(1).strip(), "Second")
m = re.match(r'^Bond\s+time.+= ([\d.]+)', lines[j])
if parser.successfulRun and m:
parser.set_statistic("Time Spent in Bond Potential Calculation",
m.group(1).strip(), "Second")
m = re.match(r'^Kspce\s+time.+= ([\d.]+)', lines[j])
if parser.successfulRun and m:
parser.set_statistic(
"Time Spent in Long-Range Coulomb Potential (K-Space) Calculation",
m.group(1).strip(), "Second")
m = re.match(r'^Neigh\s+time.+= ([\d.]+)', lines[j])
if parser.successfulRun and m:
parser.set_statistic("Time Spent in Neighbor List Regeneration",
m.group(1).strip(), "Second")
m = re.match(r'^Comm\s+time.+= ([\d.]+)', lines[j])
if parser.successfulRun and m:
parser.set_statistic("Time Spent in Communication",
m.group(1).strip(), "Second")
j += 1
if parser.successfulRun and num_steps and simulation_units != "lj":
# The default value for $stepSize is (see http://lammps.sandia.gov/doc/units.html):
#
# 0.005 tau for $simulationUnits eq "lj"
# 1e-15 second for $simulationUnits eq "real" or "metal"
# 1e-18 second for $simulationUnits eq "electron"
# 1e-8 second for $simulationUnits eq "si" or "cgs"
# If $simulationUnits is (see http://lammps.sandia.gov/doc/units.html)
#
# "lj", the unit for $stepSize is tau
# "real" or "electron", the unit for $stepSize is 1e-15 second
# "metal", the unit for $stepSize is 1e-12 second
# "si" or "cgs", the unit for $stepSize is second
# The default $simulationUnits is "lj"
#
# We ignore "lj" since "lj" is unitless.
if step_size is None:
if simulation_units == "real":
step_size = 1.0
if simulation_units.find("electron") >= 0 or simulation_units.find(
"metal") >= 0:
step_size = 0.001
if simulation_units.find("si") >= 0 or simulation_units.find(
"cgs") >= 0:
step_size = 1.0e-8
step_size_in_sec = step_size
if step_size:
if simulation_units.find("electron") >= 0 or simulation_units.find(
"real") >= 0:
step_size_in_sec = step_size * 1.0e-15
if simulation_units == "metal":
step_size_in_sec = step_size * 1.0e-12
if step_size_in_sec:
parser.set_parameter("Input:Timestep", step_size_in_sec,
"Second per Step")
parser.set_statistic(
"Molecular Dynamics Simulation Performance",
1.0e-9 * (1.0e9 * step_size_in_sec * num_steps) /
(wall_clock_time / 86400.0), "Second per Day")
if __name__ == "__main__":
# output for testing purpose
print("parsing complete:", parser.parsing_complete())
parser.print_params_stats_as_must_have()
print(parser.get_xml())
# return complete XML overwize return None
return parser.get_xml()
def process_appker_output(appstdout=None, stdout=None, stderr=None, geninfo=None, resource_appker_vars=None):
# set App Kernel Description
parser = AppKerOutputParser(
name='graph500',
version=1,
description="Graph500 Benchmark",
url='http://www.Graph500.org',
measurement_name='Graph500'
)
# set obligatory parameters and statistics
# set common parameters and statistics
parser.add_common_must_have_params_and_stats()
# set app kernel custom sets
parser.add_must_have_parameter('App:Version')
parser.add_must_have_parameter('Edge Factor')
parser.add_must_have_parameter('Input File')
parser.add_must_have_parameter('Number of Roots to Check')
parser.add_must_have_parameter('Number of Edges')
parser.add_must_have_parameter('Number of Vertices')
parser.add_must_have_parameter('Scale')
parser.add_must_have_statistic('Harmonic Mean TEPS')
parser.add_must_have_statistic('Harmonic Standard Deviation TEPS')
parser.add_must_have_statistic('Median TEPS')
parser.add_must_have_statistic('Wall Clock Time')
# parse common parameters and statistics
parser.parse_common_params_and_stats(appstdout, stdout, stderr, geninfo, resource_appker_vars)
if parser.appKerWallClockTime is not None:
parser.set_statistic("Wall Clock Time", total_seconds(parser.appKerWallClockTime), "Second")
elif parser.wallClockTime is not None:
parser.set_statistic("Wall Clock Time", total_seconds(parser.wallClockTime), "Second")
# read output
lines = []
if os.path.isfile(appstdout):
fin = open(appstdout, "rt")
lines = fin.readlines()
fin.close()
# process the output
parser.successfulRun = True
num_of_errors = 0
j = 0
while j < len(lines):
m = re.match(r'^Graph500 version:\s+(.+)', lines[j])
if m:
parser.set_parameter("App:Version", m.group(1).strip())
m = re.match(r'ERROR:\s+(.+)', lines[j])
if m:
num_of_errors += 1
m = re.match(r'^Reading input from\s+(.+)', lines[j])
if m:
parser.set_parameter("Input File", m.group(1))
m = re.match(r'^SCALE:\s+(\d+)', lines[j])
if m:
parser.set_parameter("Scale", m.group(1))
m = re.match(r'^edgefactor:\s+(\d+)', lines[j])
if m:
parser.set_parameter("Edge Factor", m.group(1))
m = re.match(r'^NBFS:\s+(\d+)', lines[j])
if m:
parser.set_parameter("Number of Roots to Check", m.group(1))
m = re.match(r'^median_TEPS:\s+(\d[0-9.e+]+)', lines[j])
if m:
parser.set_statistic("Median TEPS", m.group(1), "Traversed Edges Per Second")
m = re.match(r'^harmonic_mean_TEPS:\s+(\d[0-9.e+]+)', lines[j])
if m:
parser.successfulRun = True
parser.set_statistic("Harmonic Mean TEPS", m.group(1), "Traversed Edges Per Second")
m = re.match(r'^harmonic_stddev_TEPS:\s+(\d[0-9.e+]+)', lines[j])
if m:
parser.set_statistic("Harmonic Standard Deviation TEPS", m.group(1), "Traversed Edges Per Second")
m = re.match(r'^median_validate:\s+([\d.]+)\s+s', lines[j])
if m:
parser.set_statistic("Median Validation Time", m.group(1), "Second")
m = re.match(r'^mean_validate:\s+([\d.]+)\s+s', lines[j])
if m:
parser.set_statistic("Mean Validation Time", m.group(1), "Second")
m = re.match(r'^stddev_validate:\s+([\d.]+)\s+s', lines[j])
if m:
parser.set_statistic("Standard Deviation Validation Time", m.group(1), "Second")
j += 1
if num_of_errors > 0:
parser.successfulRun = False
if parser.get_parameter('Scale') is not None and parser.get_parameter('Edge Factor') is not None:
scale = int(parser.get_parameter('Scale'))
edgefactor = int(parser.get_parameter('Edge Factor'))
parser.set_parameter("Number of Vertices", 2 ** scale)
parser.set_parameter("Number of Edges", edgefactor * 2 ** scale)
if __name__ == "__main__":
# output for testing purpose
parser.parsing_complete(True)
print("parsing complete:", parser.parsing_complete())
parser.print_params_stats_as_must_have()
print(parser.get_xml())
# return complete XML overwize return None
return parser.get_xml()
def process_appker_output(appstdout=None,
stdout=None,
stderr=None,
geninfo=None,
proclog=None,
resource_appker_vars=None):
# set App Kernel Description
parser = AppKerOutputParser(name='hpcg',
version=1,
description="HPCG Benchmark",
url='http://www.hpcg-benchmark.org/index.html',
measurement_name='HPCG')
# set obligatory parameters and statistics
# set common parameters and statistics
parser.add_common_must_have_params_and_stats()
# set app kernel custom sets
parser.add_must_have_parameter('App:ExeBinSignature')
parser.add_must_have_parameter('App:Version')
parser.add_must_have_parameter('Input:Distributed Processes')
parser.add_must_have_parameter('Input:Global Problem Dimensions Nx')
parser.add_must_have_parameter('Input:Global Problem Dimensions Ny')
parser.add_must_have_parameter('Input:Global Problem Dimensions Nz')
parser.add_must_have_parameter('Input:Local Domain Dimensions Nx')
parser.add_must_have_parameter('Input:Local Domain Dimensions Ny')
parser.add_must_have_parameter('Input:Local Domain Dimensions Nz')
parser.add_must_have_parameter('Input:Number of Coarse Grid Levels')
parser.add_must_have_parameter('Input:Threads per processes')
parser.add_must_have_parameter('RunEnv:CPU Speed')
parser.add_must_have_parameter('RunEnv:Nodes')
parser.add_must_have_statistic('Floating-Point Performance, Raw DDOT')
parser.add_must_have_statistic('Floating-Point Performance, Raw MG')
parser.add_must_have_statistic('Floating-Point Performance, Raw SpMV')
parser.add_must_have_statistic('Floating-Point Performance, Raw Total')
parser.add_must_have_statistic('Floating-Point Performance, Raw WAXPBY')
parser.add_must_have_statistic('Floating-Point Performance, Total')
parser.add_must_have_statistic('Memory Bandwidth, Read')
parser.add_must_have_statistic('Memory Bandwidth, Total')
parser.add_must_have_statistic('Memory Bandwidth, Write')
parser.add_must_have_statistic('Setup Time')
parser.add_must_have_statistic('Wall Clock Time')
# parse common parameters and statistics
parser.parse_common_params_and_stats(appstdout, stdout, stderr, geninfo,
resource_appker_vars)
if hasattr(parser, 'appKerWallClockTime') and getattr(
parser, 'appKerWallClockTime') is not None:
parser.set_statistic("Wall Clock Time",
total_seconds(parser.appKerWallClockTime),
"Second")
# get path to YAML file
# read data
lines = []
if os.path.isfile(appstdout):
fin = open(appstdout, "rt")
lines = fin.readlines()
fin.close()
# older version stores results in yaml
# Parse YAML lines because YAML is often malformed
yaml_lines = []
# get yaml lines from appstdout
bool_in_yaml_section = False
for line in lines:
if re.match(r"^====== .+\.yaml End ======", line):
break
if bool_in_yaml_section:
yaml_lines.append(line)
if re.match(r"^====== .+\.yaml Start ======", line):
bool_in_yaml_section = True
# newer in summary txt
# txt "====== HPCG-Benchmark_3.1_2020-09-23_17-54-20.txt Start ======"
txt_lines = []
bool_in_txt_section = False
for line in lines:
if re.match(r"^====== HPCG-Benchmark.*\.txt End ======", line):
break
if bool_in_txt_section:
txt_lines.append(line)
if re.match(r"^====== HPCG-Benchmark.*\.txt Start ======", line):
bool_in_txt_section = True
if len(yaml_lines) > 5:
process_yaml(yaml_lines, parser)
else:
process_txt(txt_lines, parser)
if __name__ == "__main__":
# output for testing purpose
print("parsing complete:", parser.parsing_complete(verbose=True))
parser.print_params_stats_as_must_have()
print(parser.get_xml())
# return complete XML otherwise return None
return parser.get_xml()
def process_appker_output(appstdout=None, stdout=None, stderr=None, geninfo=None, proclog=None,
resource_appker_vars=None):
# set App Kernel Description
parser = AppKerOutputParser(
name='amber',
version=1,
description="Amber: Assisted Model Building with Energy Refinement",
url='http://ambermd.org',
measurement_name='Amber'
)
# set obligatory parameters and statistics
# set common parameters and statistics
parser.add_common_must_have_params_and_stats()
# set app kernel custom sets
parser.add_must_have_parameter('App:Version')
parser.add_must_have_parameter('Input:Coordinate File')
parser.add_must_have_parameter('Input:Number of Angles')
parser.add_must_have_parameter('Input:Number of Atoms')
parser.add_must_have_parameter('Input:Number of Bonds')
parser.add_must_have_parameter('Input:Number of Dihedrals')
parser.add_must_have_parameter('Input:Number of Steps')
parser.add_must_have_parameter('Input:Structure File')
parser.add_must_have_parameter('Input:Timestep')
parser.add_must_have_statistic('Molecular Dynamics Simulation Performance')
parser.add_must_have_statistic('Time Spent in Direct Force Calculation')
parser.add_must_have_statistic('Time Spent in Non-Bond List Regeneration')
parser.add_must_have_statistic('Time Spent in Reciprocal Force Calculation')
parser.add_must_have_statistic('Wall Clock Time')
# parse common parameters and statistics
parser.parse_common_params_and_stats(appstdout, stdout, stderr, geninfo, resource_appker_vars)
# read output
lines = []
if os.path.isfile(appstdout):
fin = open(appstdout, "rt")
lines = fin.readlines()
fin.close()
# process the output
parser.successfulRun = False
num_steps = 0
step_size = 0
j = 0
while j < len(lines):
m = re.search(r'Amber\s+([0-9a-zA-Z]+)\s+SANDER\s+20[0-9]+', lines[j])
if m:
parser.set_parameter("App:Version", "SANDER " + m.group(1))
m = re.match(r'^\|\s+PMEMD implementation of SANDER, Release\s+([0-9.]+)', lines[j])
if m:
parser.set_parameter("App:Version", "PMEMD " + m.group(1))
m = re.match(r'^\|\s+INPCRD:\s+(\S+)', lines[j])
if m:
parser.set_parameter("Input:Coordinate File", m.group(1))
m = re.match(r'^\|\s+PARM:\s+(\S+)', lines[j])
if m:
parser.set_parameter("Input:Structure File", m.group(1))
if re.search(r'CONTROL\s+DATA\s+FOR\s+THE\s+RUN', lines[j]):
j += 2
for k in range(256):
if re.match(r'^-----------------------------', lines[j]):
break
m = re.search(r'nstlim\s+=\s+([0-9]+)', lines[j])
if m:
num_steps = int(m.group(1).strip())
parser.set_parameter("Input:Number of Steps", num_steps)
m = re.search(r'dt\s+=\s+([0-9.]+)', lines[j])
if m:
step_size = 1000.0 * float(m.group(1).strip())
parser.set_parameter("Input:Timestep", step_size * 1e-15, "Second per Step")
j += 1
if re.search(r'RESOURCE\s+USE', lines[j]):
j += 2
num_bonds = 0
num_angles = 0
num_dihedrals = 0
for k in range(256):
if re.match(r'^-----------------------------', lines[j]):
break
m = re.search(r'NATOM\s+=\s+([0-9]+)', lines[j])
if m:
parser.set_parameter("Input:Number of Atoms", m.group(1).strip())
m = re.search(r'NBONH\s+=\s+([0-9]+)', lines[j])
if m:
num_bonds += int(m.group(1).strip())
m = re.search(r'NBONA\s+=\s+([0-9]+)', lines[j])
if m:
num_bonds += int(m.group(1).strip())
m = re.search(r'NTHETH\s+=\s+([0-9]+)', lines[j])
if m:
num_angles += int(m.group(1).strip())
m = re.search(r'NTHETA\s+=\s+([0-9]+)', lines[j])
if m:
num_angles += int(m.group(1).strip())
m = re.search(r'NPHIH\s+=\s+([0-9]+)', lines[j])
if m:
num_dihedrals += int(m.group(1).strip())
m = re.search(r'NPHIA\s+=\s+([0-9]+)', lines[j])
if m:
num_dihedrals += int(m.group(1).strip())
j += 1
if num_bonds > 0:
parser.set_parameter("Input:Number of Bonds", num_bonds)
if num_angles > 0:
parser.set_parameter("Input:Number of Angles", num_angles)
if num_dihedrals > 0:
parser.set_parameter("Input:Number of Dihedrals", num_dihedrals)
if re.search(r'PME Nonbond Pairlist CPU Time', lines[j]):
j += 2
for k in range(20):
m = re.search(r'Total\s+([\d.]+)', lines[j])
if m:
parser.set_statistic("Time Spent in Non-Bond List Regeneration", m.group(1), "Second")
break
j += 1
if re.search(r'PME Direct Force CPU Time', lines[j]):
j += 2
for k in range(20):
m = re.search(r'Total\s+([\d.]+)', lines[j])
if m:
parser.set_statistic("Time Spent in Direct Force Calculation", m.group(1), "Second")
break
j += 1
if re.search(r'PME Reciprocal Force CPU Time', lines[j]):
j += 2
for k in range(20):
m = re.search(r'Total\s+([\d.]+)', lines[j])
if m:
parser.set_statistic("Time Spent in Reciprocal Force Calculation", m.group(1), "Second")
break
j += 1
m = re.match(r'^\|\s+Master Total wall time:\s+([0-9.]+)\s+seconds', lines[j])
if m:
parser.set_statistic("Wall Clock Time", m.group(1), "Second")
parser.successfulRun = True
# calculate the performance
simulation_time = step_size * num_steps * 0.000001 # measured in nanoseconds
if simulation_time > 0.0:
parser.set_statistic("Molecular Dynamics Simulation Performance",
1.e-9 * simulation_time / (float(m.group(1)) / 86400.0), "Second per Day")
j += 1
if __name__ == "__main__":
# output for testing purpose
print("parsing complete:", parser.parsing_complete())
parser.print_params_stats_as_must_have()
print(parser.get_xml())
# return complete XML overwize return None
return parser.get_xml()
def process_appker_output(appstdout=None,
stdout=None,
stderr=None,
geninfo=None,
resource_appker_vars=None):
# set App Kernel Description
parser = AppKerOutputParser(
name='wrf',
version=1,
description="Weather Research and Forecasting Model",
url='http://www.wrf-model.org',
measurement_name='WRF')
# set obligatory parameters and statistics
# set common parameters and statistics
parser.add_common_must_have_params_and_stats()
# set app kernel custom sets
parser.add_must_have_parameter('App:Version')
parser.add_must_have_parameter('Input:Grid Resolution')
parser.add_must_have_parameter('Input:Simulation Length')
parser.add_must_have_parameter('Input:Simulation Start Date')
parser.add_must_have_parameter('Input:Timestep')
parser.add_must_have_parameter('RunEnv:Nodes')
parser.add_must_have_parameter('WRF Dynamical Solver')
# parser.add_must_have_statistic('Average Floating-Point Performance')
parser.add_must_have_statistic('Average Simulation Speed')
parser.add_must_have_statistic('Mean Time To Simulate One Timestep')
parser.add_must_have_statistic('Output Data Size')
# parser.add_must_have_statistic('Peak Floating-Point Performance')
parser.add_must_have_statistic('Peak Simulation Speed')
parser.add_must_have_statistic('Time Spent on I/O')
parser.add_must_have_statistic('Wall Clock Time')
# parse common parameters and statistics
parser.parse_common_params_and_stats(appstdout, stdout, stderr, geninfo,
resource_appker_vars)
# read output
lines = []
if os.path.isfile(appstdout):
fin = open(appstdout, "rt")
lines = fin.readlines()
fin.close()
# process the output
io_size = None
wall_clock_time = None
iteration_wall_clock_time = []
sim_time_per_iteration = None
dx = None
dy = None
flops_conversion = None
j = 0
while j < len(lines):
m = re.search(r'XDMOD\*\*\*SIZE OF CURRENT DIR BEFORE WRF RUN\s*(\d+)',
lines[j])
if m:
io_size = int(m.group(1).strip())
m = re.search(r'XDMOD\*\*\*SIZE OF CURRENT DIR AFTER WRF RUN\s*(\d+)',
lines[j])
if m and io_size:
parser.set_statistic("Output Data Size",
(int(m.group(1).strip()) - io_size) / 1024.0 /
1024.0, "MByte")
m = re.search(r'XDMOD\*\*\*WRF RUN BEGINS HERE --(.+)', lines[j])
if m:
wall_clock_time = parser.get_datetime_local(m.group(1).strip())
m = re.search(r'XDMOD\*\*\*WRF RUN HAS FINISHED --(.+)', lines[j])
if m and wall_clock_time:
wall_clock_time = parser.get_datetime_local(
m.group(1).strip()) - wall_clock_time
parser.set_statistic("Wall Clock Time",
wall_clock_time.total_seconds(), "Second")
if lines[j].find('XDMOD***RESULT OF rsl.out.0000 BEGINS') >= 0:
# the output from MPI rank #0
io_time = None
while j < len(lines):
if lines[j].find('XDMOD***RESULT OF rsl.out.0000 ENDS') >= 0:
break
m = re.search(
r'Timing for processing restart file.+?:\s+(\d\S+)',
lines[j], re.I)
if m:
if io_time is None:
io_time = 0.0
io_time += float(m.group(1).strip())
m = re.search(r'Timing for Writing.+?:\s+(\d\S+)', lines[j],
re.I)
if m:
if io_time is None:
io_time = 0.0
io_time += float(m.group(1).strip())
m = re.search(
r'Timing for main: time.+?on domain.+?:\s+(\d\S+)',
lines[j], re.I)
if m:
iteration_wall_clock_time.append(float(m.group(1).strip()))
m = re.search(r'WRF NUMBER OF TILES.+?(\d+)', lines[j])
if m:
omp_threads = int(m.group(1).strip())
if omp_threads > 1:
parser.set_parameter("Number of OpenMP Threads",
omp_threads)
m = re.match(r'^\s+WRF V(\S+) MODEL', lines[j])
if m:
parser.set_parameter("App:Version", m.group(1).strip())
j += 1
parser.set_statistic("Time Spent on I/O", io_time, "Second")
if re.search('XDMOD\*\*\*RESULT OF wrfout.+?BEGINS',
lines[j]) is not None:
# the output file's header (netCDF dump)
io_time = None
while j < len(lines):
if re.search('XDMOD\*\*\*RESULT OF wrfout.+?ENDS',
lines[j]) is not None:
break
m = re.search(r':DX = (\d+)', lines[j], re.I)
if m:
dx = float(m.group(1).strip()) * 0.001 # in meters
m = re.search(r':DY = (\d+)', lines[j], re.I)
if m:
dy = float(m.group(1).strip()) * 0.001 # in meters
m = re.search(r':DT = (\d+)', lines[j], re.I)
if m:
sim_time_per_iteration = float(
m.group(1).strip()) # in seconds
parser.set_parameter("Input:Timestep",
sim_time_per_iteration,
"Second per Step")
m = re.search(r':SIMULATION_START_DATE = "(.+?)"', lines[j],
re.I)
if m:
parser.set_parameter("Input:Simulation Start Date",
(m.group(1).strip()))
m = re.search(r':GRIDTYPE = "(.+?)"', lines[j], re.I)
if m:
solver = m.group(1).strip()
if solver == 'C':
solver = 'Advanced Research WRF (ARW)'
if solver == 'E':
solver = 'Nonhydrostatic Mesoscale Model (NMM)'
parser.set_parameter("WRF Dynamical Solver", solver)
m = re.search(r'Timing for Writing.+?:\s+(\d\S+)', lines[j],
re.I)
if m:
if io_time is None:
io_time = 0.0
io_time += float(m.group(1).strip())
m = re.search(
r'Timing for main: time.+?on domain.+?:\s+(\d\S+)',
lines[j], re.I)
if m:
iteration_wall_clock_time.append(float(m.group(1).strip()))
m = re.search(r'WRF NUMBER OF TILES.+?(\d+)', lines[j])
if m:
omp_threads = int(m.group(1).strip())
if omp_threads > 1:
parser.set_parameter("Number of OpenMP Threads",
omp_threads)
m = re.match(r'^\s+WRF V(\S+) MODEL', lines[j])
if m:
parser.set_parameter("App:Version", m.group(1).strip())
j += 1
if dx and dy:
if (dx - int(dx)) * 1000 < 0.1 and (
dy - int(dy)
) * 1000 < 0.1: # back compatibility with output format
parser.set_parameter("Input:Grid Resolution",
"%.0f x %.0f" % (dx, dy), "km^2")
else:
parser.set_parameter("Input:Grid Resolution",
str(dx) + " x " + str(dy), "km^2")
m = re.search(r'XDMOD\*\*\*FLOATING-POINT PERFORMANCE CONVERSION',
lines[j])
if m:
flops_conversion = lines[j + 1].strip()
j += 1
if wall_clock_time:
parser.successfulRun = True
else:
parser.successfulRun = False
if len(iteration_wall_clock_time) > 0 and sim_time_per_iteration:
parser.set_parameter("Input:Simulation Length",
(len(iteration_wall_clock_time)) *
sim_time_per_iteration / 3600.0, "Hour")
iteration_wall_clock_time = sorted(iteration_wall_clock_time)
iteration_wall_clock_time.pop()
t = 0.0
min_t = iteration_wall_clock_time[0]
for tt in iteration_wall_clock_time:
t += tt
t = t / len(iteration_wall_clock_time)
parser.set_statistic("Mean Time To Simulate One Timestep", t, "Second")
parser.set_statistic("Average Simulation Speed",
sim_time_per_iteration / t,
"Simulated Second per Second")
parser.set_statistic("Peak Simulation Speed",
sim_time_per_iteration / min_t,
"Simulated Second per Second")
if flops_conversion:
flops_conversion = flops_conversion.replace("$", "").replace(
"gflops=", "")
gflops = eval(flops_conversion, {'T': t})
parser.set_statistic("Average Floating-Point Performance",
1000.0 * gflops, "MFLOP per Second")
gflops = eval(flops_conversion, {'T': min_t})
parser.set_statistic("Peak Floating-Point Performance",
1000.0 * gflops, "MFLOP per Second")
if __name__ == "__main__":
# output for testing purpose
parsing_complete = parser.parsing_complete(True)
print("parsing complete:", parsing_complete)
if hasattr(parser, 'successfulRun'):
print("successfulRun", parser.successfulRun)
parser.print_params_stats_as_must_have()
print(parser.get_xml())
# return complete XML overwize return None
return parser.get_xml()
def process_appker_output(appstdout=None,
stdout=None,
stderr=None,
geninfo=None,
resource_appker_vars=None):
# set App Kernel Description
parser = AppKerOutputParser(
name='charmm',
version=1,
description="CHARMM: Chemistry at Harvard Macromolecular Mechanics",
url='http://www.charmm.org',
measurement_name='CHARMM')
# set obligatory parameters and statistics
# set common parameters and statistics
parser.add_common_must_have_params_and_stats()
# set app kernel custom sets
parser.add_must_have_parameter('App:Version')
parser.add_must_have_parameter('Input:Number of Angles')
parser.add_must_have_parameter('Input:Number of Atoms')
parser.add_must_have_parameter('Input:Number of Bonds')
parser.add_must_have_parameter('Input:Number of Dihedrals')
parser.add_must_have_parameter('Input:Number of Steps')
parser.add_must_have_parameter('Input:Timestep')
parser.add_must_have_statistic('Molecular Dynamics Simulation Performance')
parser.add_must_have_statistic('Time Spent in External Energy Calculation')
parser.add_must_have_statistic('Time Spent in Integration')
parser.add_must_have_statistic('Time Spent in Internal Energy Calculation')
parser.add_must_have_statistic('Time Spent in Non-Bond List Generation')
parser.add_must_have_statistic(
'Time Spent in Waiting (Load Unbalance-ness)')
parser.add_must_have_statistic('User Time')
parser.add_must_have_statistic('Wall Clock Time')
# parse common parameters and statistics
parser.parse_common_params_and_stats(appstdout, stdout, stderr, geninfo,
resource_appker_vars)
# read output
lines = []
if os.path.isfile(appstdout):
fin = open(appstdout, "rt")
lines = fin.readlines()
fin.close()
# process the output
parser.successfulRun = False
wall_clock_time = 0.0
num_steps = 0
step_size = 0.0
time_breakdown_columns = None
num_atoms = 0
num_bonds = 0
num_angles = 0
num_dihedrals = 0
j = 0
while j < len(lines):
m0 = re.search(r'\s+Chemistry at HARvard Macromolecular Mechanics',
lines[j])
m1 = re.search(r'\sVersion\s+([\da-zA-Z]+)', lines[j + 1])
if m0 and m1:
parser.set_parameter("App:Version", m1.group(1).strip())
if re.search(r'Summary of the structure file counters', lines[j]):
j += 1
for k in range(256):
if re.search(r'CHARMM>', lines[j]):
break
m = re.search(r'Number of atoms\s+=\s+(\d+)', lines[j])
if m:
num_atoms += int(m.group(1).strip())
m = re.search(r'Number of bonds\s+=\s+(\d+)', lines[j])
if m:
num_bonds += int(m.group(1).strip())
m = re.search(r'Number of angles\s+=\s+(\d+)', lines[j])
if m:
num_angles += int(m.group(1).strip())
m = re.search(r'Number of dihedrals\s+=\s+(\d+)', lines[j])
if m:
num_dihedrals += int(m.group(1).strip())
j += 1
if re.search(r'<MAKGRP> found', lines[j]):
j += 1
for k in range(256):
if re.search(r'NUMBER OF DEGREES OF FREEDOM', lines[j]):
break
m = re.search(r'NSTEP\s+=\s+(\d+)', lines[j])
if m:
num_steps = int(m.group(1).strip())
parser.set_parameter("Input:Number of Steps", num_steps)
if re.search(r'TIME STEP\s+=', lines[j]):
m = re.search(r'([\d\-Ee.]+)\s+PS', lines[j])
if m:
step_size = 1000.0 * float(m.group(1).strip())
parser.set_parameter("Input:Timestep",
step_size * 1e-15,
"Second per Step")
j += 1
if re.search(r'NORMAL TERMINATION BY NORMAL STOP', lines[j]):
parser.successfulRun = True
if re.search(r'JOB ACCOUNTING INFORMATION', lines[j]):
parser.successfulRun = True
j += 1
for k in range(256):
if j > len(lines) - 1:
break
m = re.search(r'ELAPSED TIME:\s*([\d.]+)\s*MINUTES', lines[j])
if m:
wall_clock_time = 60.0 * float(m.group(1).strip())
parser.set_statistic("Wall Clock Time", wall_clock_time,
"Second")
m = re.search(r'CPU TIME:\s*([\d.]+)\s*MINUTES', lines[j])
if m:
parser.set_statistic("User Time",
60.0 * float(m.group(1).strip()),
"Second")
m = re.search(r'ELAPSED TIME:\s*([\d.]+)\s*SECONDS', lines[j])
if m:
wall_clock_time = float(m.group(1).strip())
parser.set_statistic("Wall Clock Time", wall_clock_time,
"Second")
m = re.search(r'CPU TIME:\s*([\d.]+)\s*SECONDS', lines[j])
if m:
parser.set_statistic("User Time",
m.group(1).strip(), "Second")
j += 1
if j > len(lines) - 1:
break
if re.search(r'Parallel load balance \(sec', lines[j]):
j += 1
# grab the column headers from the output, e.g.
#
# Parallel load balance (sec.):
# Node Eext Eint Wait Comm List Integ Total
# 0 205.5 6.4 1.2 31.2 23.2 2.8 270.4
# 1 205.2 7.3 1.1 31.2 23.3 3.2 271.2
# 2 205.2 7.7 0.6 32.3 23.3 3.2 272.3
# 3 205.2 7.8 0.6 32.1 23.3 3.3 272.3
# PARALLEL> Average timing for all nodes:
# 4 205.3 7.3 0.9 31.7 23.3 3.1 271.6
time_breakdown_columns = lines[j].strip().split()
if re.search(r'PARALLEL>\s*Average timing for all nodes',
lines[j]) and time_breakdown_columns:
j += 1
time_breakdown = lines[j].strip().split()
if len(time_breakdown_columns) == len(time_breakdown):
for k in range(len(time_breakdown)):
if time_breakdown_columns[k] == "Eext":
parser.set_statistic(
"Time Spent in External Energy Calculation",
time_breakdown[k], "Second")
if time_breakdown_columns[k] == "Eint":
parser.set_statistic(
"Time Spent in Internal Energy Calculation",
time_breakdown[k], "Second")
if time_breakdown_columns[k] == "Wait":
parser.set_statistic(
"Time Spent in Waiting (Load Unbalance-ness)",
time_breakdown[k], "Second")
if time_breakdown_columns[k] == "List":
parser.set_statistic(
"Time Spent in Non-Bond List Generation",
time_breakdown[k], "Second")
if time_breakdown_columns[k] == "Integ":
parser.set_statistic("Time Spent in Integration",
time_breakdown[k], "Second")
j += 1
if num_atoms > 0:
parser.set_parameter("Input:Number of Atoms", num_atoms)
if num_bonds > 0:
parser.set_parameter("Input:Number of Bonds", num_bonds)
if num_angles > 0:
parser.set_parameter("Input:Number of Angles", num_angles)
if num_dihedrals > 0:
parser.set_parameter("Input:Number of Dihedrals", num_dihedrals)
if wall_clock_time > 0.0 and num_steps > 0 and step_size > 0.0:
# $stepSize is in femtoseconds
# $wallClockTime is in seconds
parser.set_statistic("Molecular Dynamics Simulation Performance",
(1e-6 * step_size * num_steps) /
(wall_clock_time / 86400.0) * 1e-9,
"Second per Day")
if __name__ == "__main__":
# output for testing purpose
print("parsing complete:", parser.parsing_complete())
parser.print_params_stats_as_must_have()
print(parser.get_xml())
# return complete XML overwize return None
return parser.get_xml()
def push_to_db_raw(self, cur, task_id, time_finished):
import xml.etree.ElementTree
log.info("Pushing to DB")
result_file = os.path.join(self.taskDir, "result.xml")
jobfiles_dir = os.path.join(self.taskDir, "jobfiles")
(batch_job_dir, stdout_file, stderr_file, appstdout_file, taskexeclog_file) = self.get_result_files()
# sanity check
fin = open(result_file, "r")
content = fin.read()
fin.close()
if content[0] != '<':
# need to reformat
i0 = content.find("<rep:report")
i1 = content.find("</rep:report>")
fout = open(result_file, "w")
content = fout.write("<?xml version='1.0'?>\n" + content[i0:i1 + len("</rep:report>")] + "\n")
fout.close()
try:
tree = xml.etree.ElementTree.parse(result_file)
tree.getroot()
except Exception as e:
log.exception("Got exception in push_to_db_raw, during xml read: %s", e)
self.set_method_to_run_next(
None,
"Cannot process final XML file",
"(1)==Resulting XML file content==\n" + content +
"\n==Previous status==" + self.status +
"\n==Previous status info==" + self.status_info + "\n" +
traceback.format_exc())
self.write_error_xml(result_file, cdata=True)
instance_id = task_id
status = None
message = None
stderr = None
job_id = None
cur.execute("""SELECT instance_id,collected,committed,resource,executionhost,reporter,
reporternickname,status,message,stderr,body,memory,cputime,walltime,job_id
FROM akrr_xdmod_instanceinfo WHERE instance_id=%s""", (task_id,))
raw = cur.fetchone()
if raw is not None: # .i.e. not new entry
(instance_id, collected, committed, resource, executionhost, reporter, reporternickname, status, message,
stderr, body, memory, cputime, walltime, job_id) = raw
if hasattr(self, "RemoteJobID"):
job_id = self.RemoteJobID
else:
collected = time_finished
committed = datetime.datetime.today().strftime("%Y-%m-%d %H:%M:%S")
resource = self.resourceName
executionhost = self.resource.get('__regexp__', self.resourceName)
reporter = self.appName
# reporternickname="%s.%d"%(self.appName,self.resourceParam['ncpus'])
reporternickname = akrr.util.replace_at_var_at(self.app['nickname'],
[self.resource, self.app, self.resourceParam, self.appParam])
if hasattr(self, "RemoteJobID"):
job_id = self.RemoteJobID
# Process XML file
try:
tree = xml.etree.ElementTree.parse(result_file)
root = tree.getroot()
except Exception as e:
log.exception("Got exception in push_to_db_raw, during xml read: %s", e)
self.status_info = "(2)==Resulting XML file content==\n" + content + \
"\n==Previous status==" + self.status + \
"\n==Previous status info==" + self.status_info
self.status = "Cannot process final XML file"
root = None
memory = 0.0
cputime = 0.0
walltime = 0.0
completed = None
log.debug("%s %s", root, status)
if root is not None:
try:
t = root.find('exitStatus').find('completed').text
log.debug('exitStatus:completed %s', t)
if t.strip().upper() == "TRUE":
completed = True
else:
completed = False
if completed:
root.find('body').find('performance').find('benchmark').find('statistics')
self.set_method_to_run_next(None, "Task was completed successfully.", "Done")
except Exception as e:
log.exception("Got exception in push_to_db_raw: %s", e)
log.debug("completedstatus %s %s", completed, status)
error = None
try:
t = root.find('xdtas').find('batchJob').find('status').text
if t.upper() == "ERROR":
error = True
else:
error = False
except Exception as e:
log.exception("Got exception in push_to_db_raw: %s", e)
status = 0
if completed is not None:
if completed:
status = 1
log.debug("completedstatus %s %s", completed, status)
if completed is not None:
if completed:
body_et = root.find('body').find('performance')
body = xml.etree.ElementTree.tostring(body_et, encoding="unicode")
import xml.dom.minidom
xml = xml.dom.minidom.parseString(body)
body = xml.toprettyxml(indent=" ")
body = body.replace("""<?xml version="1.0" ?>\n""", "")
body = akrr.util.strip_empty_lines(body)
statistics_et = root.find('body').find('performance').find('benchmark').find('statistics')
for e in statistics_et:
if e.find('ID').text == 'Memory':
memory = float(e.find('value').text)
if e.find('ID').text == 'Wall Clock Time':
walltime = float(e.find('value').text)
else:
# i.e. error caught by parser on resource
try:
message = "ERROR: error caught by parser"
stderr = root.find('exitStatus').find('errorMessage').text
body = """<xdtas>
<batchJob>
<status>Error</status>
<errorCause>ERROR: error caught by parser</errorCause>
<reporter>%s</reporter>
<errorMsg>%s</errorMsg>
</batchJob>
</xdtas>
""" % (reporter, stderr)
except Exception as e:
log.exception("Got exception in push_to_db_raw: %s", e)
message = "ERROR: error caught by parser"
stderr = None
body = """<xdtas>
<batchJob>
<status>Error</status>
<errorCause>ERROR: error caught by parser</errorCause>
<reporter>%s</reporter>
<errorMsg></errorMsg>
</batchJob>
</xdtas>
""" % reporter
elif completed is None:
if error is not None: # i.e. xml with error generated afterwards
if error:
# fin=open(resultFile,"r")
# body=fin.read()
# fin.close
message = root.find('xdtas').find('batchJob').find('errorCause').text
stderr = root.find('xdtas').find('batchJob').find('errorMsg').text
body_et = root.find('xdtas')
body = xml.etree.ElementTree.tostring(body_et)
else:
raise IOError("This condition should not happens")
else:
raise IOError("This condition should not happens")
else:
raise IOError("This condition should not happens")
# memory=None
# cputime=None
# walltime=None
else: # i.e. xml was not parsed
status = 0
message = self.status
stderr = self.status_info
body = """<xdtas>
<batchJob>
<status>Error</status>
<errorCause>Unknown error</errorCause>
<reporter>AKRR Server</reporter>
<errorMsg>"Cannot process final XML file</errorMsg>
</batchJob>
</xdtas>
"""
message = akrr.util.clean_unicode(message)
stderr = akrr.util.clean_unicode(stderr)
body = akrr.util.clean_unicode(body)
# Get Nodes
nodes = None
nodes_filename = os.path.join(jobfiles_dir, "gen.info")
if os.path.isfile(nodes_filename):
parser = AppKerOutputParser()
parser.parse_common_params_and_stats(geninfo=nodes_filename)
if hasattr(parser, 'geninfo') and 'node_list' in parser.geninfo:
nodes_list = parser.geninfo['node_list'].split()
nodes = ";"
for line in nodes_list:
line = line.strip()
nodes += "%s;" % line
if len(nodes.strip().strip(';')) == 0:
nodes = None
internal_failure_code = 0
if 'masterTaskID' in self.taskParam and appstdout_file is None:
internal_failure_code = 10004
log.debug("completedstatus", completed, status)
if raw is not None: # .i.e. new entry
print("Updating")
cur.execute("""UPDATE akrr_xdmod_instanceinfo
SET instance_id=%s,collected=%s,committed=%s,resource=%s,executionhost=%s,reporter=%s,
reporternickname=%s,status=%s,message=%s,stderr=%s,body=%s,memory=%s,cputime=%s,walltime=%s,job_id=%s,nodes=%s,
internal_failure=%s
WHERE instance_id=%s""",
(instance_id, collected, committed, resource, executionhost, reporter, reporternickname, status,
message, stderr, body, memory, cputime, walltime, job_id, nodes, internal_failure_code,
instance_id))
else:
cur.execute("""INSERT INTO akrr_xdmod_instanceinfo
(instance_id,collected,committed,resource,executionhost,reporter,reporternickname,status,message,stderr,body,memory,
cputime,walltime,job_id,nodes,internal_failure)
VALUES (%s,%s,%s,%s,%s,%s,%s,%s,%s,%s,%s,%s,%s,%s,%s,%s,%s)""",
(instance_id, collected, committed, resource, executionhost, reporter, reporternickname, status,
message, stderr, body, memory, cputime, walltime, job_id, nodes, internal_failure_code))
if appstdout_file is not None:
fin = open(appstdout_file, "r", encoding="utf-8")
appstdout_file_content = fin.read()
fin.close()
else:
appstdout_file_content = "Does Not Present"
if stdout_file is not None:
fin = open(stdout_file, "r", encoding="utf-8")
stdout_file_content = fin.read()
fin.close()
else:
stdout_file_content = "Does Not Present"
if stderr_file is not None:
fin = open(stderr_file, "r", encoding="utf-8")
stderr_file_content = fin.read()
fin.close()
else:
stderr_file_content = "Does Not Present"
if taskexeclog_file is not None:
fin = open(taskexeclog_file, "r", encoding="utf-8")
taskexeclog_file_content = fin.read()
fin.close()
else:
taskexeclog_file_content = "Does Not Present"
assert isinstance(appstdout_file_content, str)
assert isinstance(stdout_file_content, str)
assert isinstance(stderr_file_content, str)
assert isinstance(taskexeclog_file_content, str)
cur.execute("""SELECT task_id
FROM akrr_errmsg WHERE task_id=%s""", (task_id,))
raw = cur.fetchall()
cur.execute("""SELECT @@max_allowed_packet""")
max_allowed_packet = cur.fetchall()[0][0]
if (len(appstdout_file_content) + len(stderr_file_content) + len(stdout_file_content) + len(
taskexeclog_file_content)) > 0.9 * max_allowed_packet:
print("WARNING: length of files exceed max_allowed_packet will trancate files")
if len(appstdout_file_content) > 0.2 * max_allowed_packet:
appstdout_file_content = appstdout_file_content[:int(0.2 * max_allowed_packet)]
appstdout_file_content += \
"\nWARNING: File was trancated because it length of files exceed max_allowed_packet\n"
if len(stderr_file_content) > 0.2 * max_allowed_packet:
stderr_file_content = stderr_file_content[:int(0.2 * max_allowed_packet)]
stderr_file_content += \
"\nWARNING: File was trancated because it length of files exceed max_allowed_packet\n"
if len(stdout_file_content) > 0.2 * max_allowed_packet:
stdout_file_content = stdout_file_content[:int(0.2 * max_allowed_packet)]
stdout_file_content += \
"\nWARNING: File was trancated because it length of files exceed max_allowed_packet\n"
if len(taskexeclog_file_content) > 0.2 * max_allowed_packet:
taskexeclog_file_content = taskexeclog_file_content[:int(0.2 * max_allowed_packet)]
taskexeclog_file_content += \
"\nWARNING: File was trancated because it length of files exceed max_allowed_packet\n"
appstdout_file_content = akrr.util.clean_unicode(appstdout_file_content)
stderr_file_content = akrr.util.clean_unicode(stderr_file_content)
stdout_file_content = akrr.util.clean_unicode(stdout_file_content)
taskexeclog_file_content = akrr.util.clean_unicode(taskexeclog_file_content)
if len(raw) > 0: # .i.e. updating existing entry
print("Updating", raw)
cur.execute("""UPDATE akrr_errmsg
SET appstdout=%s,stderr=%s,stdout=%s,taskexeclog=%s
WHERE task_id=%s""",
(appstdout_file_content, stderr_file_content, stdout_file_content,
taskexeclog_file_content, instance_id))
else:
cur.execute("""INSERT INTO akrr_errmsg
(task_id,appstdout,stderr,stdout,taskexeclog)
VALUES (%s,%s,%s,%s,%s)""",
(instance_id, appstdout_file_content, stderr_file_content, stdout_file_content,
taskexeclog_file_content))