cpython/Tools/pybench/pybench.py

955 lines
31 KiB
Python
Executable File

#!/usr/local/bin/python -O
""" A Python Benchmark Suite
"""
# Note: Please keep this module compatible to Python 2.6.
#
# Tests may include features in later Python versions, but these
# should then be embedded in try-except clauses in the configuration
# module Setup.py.
#
from __future__ import print_function
# pybench Copyright
__copyright__ = """\
Copyright (c), 1997-2006, Marc-Andre Lemburg (mal@lemburg.com)
Copyright (c), 2000-2006, eGenix.com Software GmbH (info@egenix.com)
All Rights Reserved.
Permission to use, copy, modify, and distribute this software and its
documentation for any purpose and without fee or royalty is hereby
granted, provided that the above copyright notice appear in all copies
and that both that copyright notice and this permission notice appear
in supporting documentation or portions thereof, including
modifications, that you make.
THE AUTHOR MARC-ANDRE LEMBURG DISCLAIMS ALL WARRANTIES WITH REGARD TO
THIS SOFTWARE, INCLUDING ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND
FITNESS, IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY SPECIAL,
INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES WHATSOEVER RESULTING
FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN AN ACTION OF CONTRACT,
NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF OR IN CONNECTION
WITH THE USE OR PERFORMANCE OF THIS SOFTWARE !
"""
import sys, time, operator, platform
from CommandLine import *
try:
import cPickle
pickle = cPickle
except ImportError:
import pickle
# Version number; version history: see README file !
__version__ = '2.1'
### Constants
# Second fractions
MILLI_SECONDS = 1e3
MICRO_SECONDS = 1e6
# Percent unit
PERCENT = 100
# Horizontal line length
LINE = 79
# Minimum test run-time
MIN_TEST_RUNTIME = 1e-3
# Number of calibration runs to use for calibrating the tests
CALIBRATION_RUNS = 20
# Number of calibration loops to run for each calibration run
CALIBRATION_LOOPS = 20
# Allow skipping calibration ?
ALLOW_SKIPPING_CALIBRATION = 1
# Timer types
TIMER_TIME_TIME = 'time.time'
TIMER_TIME_CLOCK = 'time.clock'
TIMER_SYSTIMES_PROCESSTIME = 'systimes.processtime'
# Choose platform default timer
if sys.platform[:3] == 'win':
# On WinXP this has 2.5ms resolution
TIMER_PLATFORM_DEFAULT = TIMER_TIME_CLOCK
else:
# On Linux this has 1ms resolution
TIMER_PLATFORM_DEFAULT = TIMER_TIME_TIME
# Print debug information ?
_debug = 0
### Helpers
def get_timer(timertype):
if timertype == TIMER_TIME_TIME:
return time.time
elif timertype == TIMER_TIME_CLOCK:
return time.clock
elif timertype == TIMER_SYSTIMES_PROCESSTIME:
import systimes
return systimes.processtime
else:
raise TypeError('unknown timer type: %s' % timertype)
def get_machine_details():
if _debug:
print('Getting machine details...')
buildno, builddate = platform.python_build()
python = platform.python_version()
if sys.maxunicode == 65535:
# UCS2 build (standard)
unitype = 'UCS2'
else:
# UCS4 build (most recent Linux distros)
unitype = 'UCS4'
bits, linkage = platform.architecture()
return {
'platform': platform.platform(),
'processor': platform.processor(),
'executable': sys.executable,
'implementation': getattr(platform, 'python_implementation',
lambda:'n/a')(),
'python': platform.python_version(),
'compiler': platform.python_compiler(),
'buildno': buildno,
'builddate': builddate,
'unicode': unitype,
'bits': bits,
}
def print_machine_details(d, indent=''):
l = ['Machine Details:',
' Platform ID: %s' % d.get('platform', 'n/a'),
' Processor: %s' % d.get('processor', 'n/a'),
'',
'Python:',
' Implementation: %s' % d.get('implementation', 'n/a'),
' Executable: %s' % d.get('executable', 'n/a'),
' Version: %s' % d.get('python', 'n/a'),
' Compiler: %s' % d.get('compiler', 'n/a'),
' Bits: %s' % d.get('bits', 'n/a'),
' Build: %s (#%s)' % (d.get('builddate', 'n/a'),
d.get('buildno', 'n/a')),
' Unicode: %s' % d.get('unicode', 'n/a'),
]
joiner = '\n' + indent
print(indent + joiner.join(l) + '\n')
### Test baseclass
class Test:
""" All test must have this class as baseclass. It provides
the necessary interface to the benchmark machinery.
The tests must set .rounds to a value high enough to let the
test run between 20-50 seconds. This is needed because
clock()-timing only gives rather inaccurate values (on Linux,
for example, it is accurate to a few hundreths of a
second). If you don't want to wait that long, use a warp
factor larger than 1.
It is also important to set the .operations variable to a
value representing the number of "virtual operations" done per
call of .run().
If you change a test in some way, don't forget to increase
its version number.
"""
### Instance variables that each test should override
# Version number of the test as float (x.yy); this is important
# for comparisons of benchmark runs - tests with unequal version
# number will not get compared.
version = 2.1
# The number of abstract operations done in each round of the
# test. An operation is the basic unit of what you want to
# measure. The benchmark will output the amount of run-time per
# operation. Note that in order to raise the measured timings
# significantly above noise level, it is often required to repeat
# sets of operations more than once per test round. The measured
# overhead per test round should be less than 1 second.
operations = 1
# Number of rounds to execute per test run. This should be
# adjusted to a figure that results in a test run-time of between
# 1-2 seconds.
rounds = 100000
### Internal variables
# Mark this class as implementing a test
is_a_test = 1
# Last timing: (real, run, overhead)
last_timing = (0.0, 0.0, 0.0)
# Warp factor to use for this test
warp = 1
# Number of calibration runs to use
calibration_runs = CALIBRATION_RUNS
# List of calibration timings
overhead_times = None
# List of test run timings
times = []
# Timer used for the benchmark
timer = TIMER_PLATFORM_DEFAULT
def __init__(self, warp=None, calibration_runs=None, timer=None):
# Set parameters
if warp is not None:
self.rounds = int(self.rounds / warp)
if self.rounds == 0:
raise ValueError('warp factor set too high')
self.warp = warp
if calibration_runs is not None:
if (not ALLOW_SKIPPING_CALIBRATION and
calibration_runs < 1):
raise ValueError('at least one calibration run is required')
self.calibration_runs = calibration_runs
if timer is not None:
timer = timer
# Init variables
self.times = []
self.overhead_times = []
# We want these to be in the instance dict, so that pickle
# saves them
self.version = self.version
self.operations = self.operations
self.rounds = self.rounds
def get_timer(self):
""" Return the timer function to use for the test.
"""
return get_timer(self.timer)
def compatible(self, other):
""" Return 1/0 depending on whether the test is compatible
with the other Test instance or not.
"""
if self.version != other.version:
return 0
if self.rounds != other.rounds:
return 0
return 1
def calibrate_test(self):
if self.calibration_runs == 0:
self.overhead_times = [0.0]
return
calibrate = self.calibrate
timer = self.get_timer()
calibration_loops = range(CALIBRATION_LOOPS)
# Time the calibration loop overhead
prep_times = []
for i in range(self.calibration_runs):
t = timer()
for i in calibration_loops:
pass
t = timer() - t
prep_times.append(t)
min_prep_time = min(prep_times)
if _debug:
print()
print('Calib. prep time = %.6fms' % (
min_prep_time * MILLI_SECONDS))
# Time the calibration runs (doing CALIBRATION_LOOPS loops of
# .calibrate() method calls each)
for i in range(self.calibration_runs):
t = timer()
for i in calibration_loops:
calibrate()
t = timer() - t
self.overhead_times.append(t / CALIBRATION_LOOPS
- min_prep_time)
# Check the measured times
min_overhead = min(self.overhead_times)
max_overhead = max(self.overhead_times)
if _debug:
print('Calib. overhead time = %.6fms' % (
min_overhead * MILLI_SECONDS))
if min_overhead < 0.0:
raise ValueError('calibration setup did not work')
if max_overhead - min_overhead > 0.1:
raise ValueError(
'overhead calibration timing range too inaccurate: '
'%r - %r' % (min_overhead, max_overhead))
def run(self):
""" Run the test in two phases: first calibrate, then
do the actual test. Be careful to keep the calibration
timing low w/r to the test timing.
"""
test = self.test
timer = self.get_timer()
# Get calibration
min_overhead = min(self.overhead_times)
# Test run
t = timer()
test()
t = timer() - t
if t < MIN_TEST_RUNTIME:
raise ValueError('warp factor too high: '
'test times are < 10ms')
eff_time = t - min_overhead
if eff_time < 0:
raise ValueError('wrong calibration')
self.last_timing = (eff_time, t, min_overhead)
self.times.append(eff_time)
def calibrate(self):
""" Calibrate the test.
This method should execute everything that is needed to
setup and run the test - except for the actual operations
that you intend to measure. pybench uses this method to
measure the test implementation overhead.
"""
return
def test(self):
""" Run the test.
The test needs to run self.rounds executing
self.operations number of operations each.
"""
return
def stat(self):
""" Return test run statistics as tuple:
(minimum run time,
average run time,
total run time,
average time per operation,
minimum overhead time)
"""
runs = len(self.times)
if runs == 0:
return 0.0, 0.0, 0.0, 0.0
min_time = min(self.times)
total_time = sum(self.times)
avg_time = total_time / float(runs)
operation_avg = total_time / float(runs
* self.rounds
* self.operations)
if self.overhead_times:
min_overhead = min(self.overhead_times)
else:
min_overhead = self.last_timing[2]
return min_time, avg_time, total_time, operation_avg, min_overhead
### Load Setup
# This has to be done after the definition of the Test class, since
# the Setup module will import subclasses using this class.
import Setup
### Benchmark base class
class Benchmark:
# Name of the benchmark
name = ''
# Number of benchmark rounds to run
rounds = 1
# Warp factor use to run the tests
warp = 1 # Warp factor
# Average benchmark round time
roundtime = 0
# Benchmark version number as float x.yy
version = 2.1
# Produce verbose output ?
verbose = 0
# Dictionary with the machine details
machine_details = None
# Timer used for the benchmark
timer = TIMER_PLATFORM_DEFAULT
def __init__(self, name, verbose=None, timer=None, warp=None,
calibration_runs=None):
if name:
self.name = name
else:
self.name = '%04i-%02i-%02i %02i:%02i:%02i' % \
(time.localtime(time.time())[:6])
if verbose is not None:
self.verbose = verbose
if timer is not None:
self.timer = timer
if warp is not None:
self.warp = warp
if calibration_runs is not None:
self.calibration_runs = calibration_runs
# Init vars
self.tests = {}
if _debug:
print('Getting machine details...')
self.machine_details = get_machine_details()
# Make .version an instance attribute to have it saved in the
# Benchmark pickle
self.version = self.version
def get_timer(self):
""" Return the timer function to use for the test.
"""
return get_timer(self.timer)
def compatible(self, other):
""" Return 1/0 depending on whether the benchmark is
compatible with the other Benchmark instance or not.
"""
if self.version != other.version:
return 0
if (self.machine_details == other.machine_details and
self.timer != other.timer):
return 0
if (self.calibration_runs == 0 and
other.calibration_runs != 0):
return 0
if (self.calibration_runs != 0 and
other.calibration_runs == 0):
return 0
return 1
def load_tests(self, setupmod, limitnames=None):
# Add tests
if self.verbose:
print('Searching for tests ...')
print('--------------------------------------')
for testclass in setupmod.__dict__.values():
if not hasattr(testclass, 'is_a_test'):
continue
name = testclass.__name__
if name == 'Test':
continue
if (limitnames is not None and
limitnames.search(name) is None):
continue
self.tests[name] = testclass(
warp=self.warp,
calibration_runs=self.calibration_runs,
timer=self.timer)
l = sorted(self.tests)
if self.verbose:
for name in l:
print(' %s' % name)
print('--------------------------------------')
print(' %i tests found' % len(l))
print()
def calibrate(self):
print('Calibrating tests. Please wait...', end=' ')
sys.stdout.flush()
if self.verbose:
print()
print()
print('Test min max')
print('-' * LINE)
tests = sorted(self.tests.items())
for i in range(len(tests)):
name, test = tests[i]
test.calibrate_test()
if self.verbose:
print('%30s: %6.3fms %6.3fms' % \
(name,
min(test.overhead_times) * MILLI_SECONDS,
max(test.overhead_times) * MILLI_SECONDS))
if self.verbose:
print()
print('Done with the calibration.')
else:
print('done.')
print()
def run(self):
tests = sorted(self.tests.items())
timer = self.get_timer()
print('Running %i round(s) of the suite at warp factor %i:' % \
(self.rounds, self.warp))
print()
self.roundtimes = []
for i in range(self.rounds):
if self.verbose:
print(' Round %-25i effective absolute overhead' % (i+1))
total_eff_time = 0.0
for j in range(len(tests)):
name, test = tests[j]
if self.verbose:
print('%30s:' % name, end=' ')
test.run()
(eff_time, abs_time, min_overhead) = test.last_timing
total_eff_time = total_eff_time + eff_time
if self.verbose:
print(' %5.0fms %5.0fms %7.3fms' % \
(eff_time * MILLI_SECONDS,
abs_time * MILLI_SECONDS,
min_overhead * MILLI_SECONDS))
self.roundtimes.append(total_eff_time)
if self.verbose:
print(' '
' ------------------------------')
print(' '
' Totals: %6.0fms' %
(total_eff_time * MILLI_SECONDS))
print()
else:
print('* Round %i done in %.3f seconds.' % (i+1,
total_eff_time))
print()
def stat(self):
""" Return benchmark run statistics as tuple:
(minimum round time,
average round time,
maximum round time)
XXX Currently not used, since the benchmark does test
statistics across all rounds.
"""
runs = len(self.roundtimes)
if runs == 0:
return 0.0, 0.0
min_time = min(self.roundtimes)
total_time = sum(self.roundtimes)
avg_time = total_time / float(runs)
max_time = max(self.roundtimes)
return (min_time, avg_time, max_time)
def print_header(self, title='Benchmark'):
print('-' * LINE)
print('%s: %s' % (title, self.name))
print('-' * LINE)
print()
print(' Rounds: %s' % self.rounds)
print(' Warp: %s' % self.warp)
print(' Timer: %s' % self.timer)
print()
if self.machine_details:
print_machine_details(self.machine_details, indent=' ')
print()
def print_benchmark(self, hidenoise=0, limitnames=None):
print('Test '
' minimum average operation overhead')
print('-' * LINE)
tests = sorted(self.tests.items())
total_min_time = 0.0
total_avg_time = 0.0
for name, test in tests:
if (limitnames is not None and
limitnames.search(name) is None):
continue
(min_time,
avg_time,
total_time,
op_avg,
min_overhead) = test.stat()
total_min_time = total_min_time + min_time
total_avg_time = total_avg_time + avg_time
print('%30s: %5.0fms %5.0fms %6.2fus %7.3fms' % \
(name,
min_time * MILLI_SECONDS,
avg_time * MILLI_SECONDS,
op_avg * MICRO_SECONDS,
min_overhead *MILLI_SECONDS))
print('-' * LINE)
print('Totals: '
' %6.0fms %6.0fms' %
(total_min_time * MILLI_SECONDS,
total_avg_time * MILLI_SECONDS,
))
print()
def print_comparison(self, compare_to, hidenoise=0, limitnames=None):
# Check benchmark versions
if compare_to.version != self.version:
print('* Benchmark versions differ: '
'cannot compare this benchmark to "%s" !' %
compare_to.name)
print()
self.print_benchmark(hidenoise=hidenoise,
limitnames=limitnames)
return
# Print header
compare_to.print_header('Comparing with')
print('Test '
' minimum run-time average run-time')
print(' '
' this other diff this other diff')
print('-' * LINE)
# Print test comparisons
tests = sorted(self.tests.items())
total_min_time = other_total_min_time = 0.0
total_avg_time = other_total_avg_time = 0.0
benchmarks_compatible = self.compatible(compare_to)
tests_compatible = 1
for name, test in tests:
if (limitnames is not None and
limitnames.search(name) is None):
continue
(min_time,
avg_time,
total_time,
op_avg,
min_overhead) = test.stat()
total_min_time = total_min_time + min_time
total_avg_time = total_avg_time + avg_time
try:
other = compare_to.tests[name]
except KeyError:
other = None
if other is None:
# Other benchmark doesn't include the given test
min_diff, avg_diff = 'n/a', 'n/a'
other_min_time = 0.0
other_avg_time = 0.0
tests_compatible = 0
else:
(other_min_time,
other_avg_time,
other_total_time,
other_op_avg,
other_min_overhead) = other.stat()
other_total_min_time = other_total_min_time + other_min_time
other_total_avg_time = other_total_avg_time + other_avg_time
if (benchmarks_compatible and
test.compatible(other)):
# Both benchmark and tests are comparable
min_diff = ((min_time * self.warp) /
(other_min_time * other.warp) - 1.0)
avg_diff = ((avg_time * self.warp) /
(other_avg_time * other.warp) - 1.0)
if hidenoise and abs(min_diff) < 10.0:
min_diff = ''
else:
min_diff = '%+5.1f%%' % (min_diff * PERCENT)
if hidenoise and abs(avg_diff) < 10.0:
avg_diff = ''
else:
avg_diff = '%+5.1f%%' % (avg_diff * PERCENT)
else:
# Benchmark or tests are not comparable
min_diff, avg_diff = 'n/a', 'n/a'
tests_compatible = 0
print('%30s: %5.0fms %5.0fms %7s %5.0fms %5.0fms %7s' % \
(name,
min_time * MILLI_SECONDS,
other_min_time * MILLI_SECONDS * compare_to.warp / self.warp,
min_diff,
avg_time * MILLI_SECONDS,
other_avg_time * MILLI_SECONDS * compare_to.warp / self.warp,
avg_diff))
print('-' * LINE)
# Summarise test results
if not benchmarks_compatible or not tests_compatible:
min_diff, avg_diff = 'n/a', 'n/a'
else:
if other_total_min_time != 0.0:
min_diff = '%+5.1f%%' % (
((total_min_time * self.warp) /
(other_total_min_time * compare_to.warp) - 1.0) * PERCENT)
else:
min_diff = 'n/a'
if other_total_avg_time != 0.0:
avg_diff = '%+5.1f%%' % (
((total_avg_time * self.warp) /
(other_total_avg_time * compare_to.warp) - 1.0) * PERCENT)
else:
avg_diff = 'n/a'
print('Totals: '
' %5.0fms %5.0fms %7s %5.0fms %5.0fms %7s' %
(total_min_time * MILLI_SECONDS,
(other_total_min_time * compare_to.warp/self.warp
* MILLI_SECONDS),
min_diff,
total_avg_time * MILLI_SECONDS,
(other_total_avg_time * compare_to.warp/self.warp
* MILLI_SECONDS),
avg_diff
))
print()
print('(this=%s, other=%s)' % (self.name,
compare_to.name))
print()
class PyBenchCmdline(Application):
header = ("PYBENCH - a benchmark test suite for Python "
"interpreters/compilers.")
version = __version__
debug = _debug
options = [ArgumentOption('-n',
'number of rounds',
Setup.Number_of_rounds),
ArgumentOption('-f',
'save benchmark to file arg',
''),
ArgumentOption('-c',
'compare benchmark with the one in file arg',
''),
ArgumentOption('-s',
'show benchmark in file arg, then exit',
''),
ArgumentOption('-w',
'set warp factor to arg',
Setup.Warp_factor),
ArgumentOption('-t',
'run only tests with names matching arg',
''),
ArgumentOption('-C',
'set the number of calibration runs to arg',
CALIBRATION_RUNS),
SwitchOption('-d',
'hide noise in comparisons',
0),
SwitchOption('-v',
'verbose output (not recommended)',
0),
SwitchOption('--with-gc',
'enable garbage collection',
0),
SwitchOption('--with-syscheck',
'use default sys check interval',
0),
ArgumentOption('--timer',
'use given timer',
TIMER_PLATFORM_DEFAULT),
]
about = """\
The normal operation is to run the suite and display the
results. Use -f to save them for later reuse or comparisons.
Available timers:
time.time
time.clock
systimes.processtime
Examples:
python2.1 pybench.py -f p21.pybench
python2.5 pybench.py -f p25.pybench
python pybench.py -s p25.pybench -c p21.pybench
"""
copyright = __copyright__
def main(self):
rounds = self.values['-n']
reportfile = self.values['-f']
show_bench = self.values['-s']
compare_to = self.values['-c']
hidenoise = self.values['-d']
warp = int(self.values['-w'])
withgc = self.values['--with-gc']
limitnames = self.values['-t']
if limitnames:
if _debug:
print('* limiting test names to one with substring "%s"' % \
limitnames)
limitnames = re.compile(limitnames, re.I)
else:
limitnames = None
verbose = self.verbose
withsyscheck = self.values['--with-syscheck']
calibration_runs = self.values['-C']
timer = self.values['--timer']
print('-' * LINE)
print('PYBENCH %s' % __version__)
print('-' * LINE)
print('* using %s %s' % (
getattr(platform, 'python_implementation', lambda:'Python')(),
' '.join(sys.version.split())))
# Switch off garbage collection
if not withgc:
try:
import gc
except ImportError:
print('* Python version doesn\'t support garbage collection')
else:
try:
gc.disable()
except NotImplementedError:
print('* Python version doesn\'t support gc.disable')
else:
print('* disabled garbage collection')
# "Disable" sys check interval
if not withsyscheck:
# Too bad the check interval uses an int instead of a long...
value = 2147483647
try:
sys.setcheckinterval(value)
except (AttributeError, NotImplementedError):
print('* Python version doesn\'t support sys.setcheckinterval')
else:
print('* system check interval set to maximum: %s' % value)
if timer == TIMER_SYSTIMES_PROCESSTIME:
import systimes
print('* using timer: systimes.processtime (%s)' % \
systimes.SYSTIMES_IMPLEMENTATION)
else:
print('* using timer: %s' % timer)
print()
if compare_to:
try:
f = open(compare_to,'rb')
bench = pickle.load(f)
bench.name = compare_to
f.close()
compare_to = bench
except IOError as reason:
print('* Error opening/reading file %s: %s' % (
repr(compare_to),
reason))
compare_to = None
if show_bench:
try:
f = open(show_bench,'rb')
bench = pickle.load(f)
bench.name = show_bench
f.close()
bench.print_header()
if compare_to:
bench.print_comparison(compare_to,
hidenoise=hidenoise,
limitnames=limitnames)
else:
bench.print_benchmark(hidenoise=hidenoise,
limitnames=limitnames)
except IOError as reason:
print('* Error opening/reading file %s: %s' % (
repr(show_bench),
reason))
print()
return
if reportfile:
print('Creating benchmark: %s (rounds=%i, warp=%i)' % \
(reportfile, rounds, warp))
print()
# Create benchmark object
bench = Benchmark(reportfile,
verbose=verbose,
timer=timer,
warp=warp,
calibration_runs=calibration_runs)
bench.rounds = rounds
bench.load_tests(Setup, limitnames=limitnames)
try:
bench.calibrate()
bench.run()
except KeyboardInterrupt:
print()
print('*** KeyboardInterrupt -- Aborting')
print()
return
bench.print_header()
if compare_to:
bench.print_comparison(compare_to,
hidenoise=hidenoise,
limitnames=limitnames)
else:
bench.print_benchmark(hidenoise=hidenoise,
limitnames=limitnames)
# Ring bell
sys.stderr.write('\007')
if reportfile:
try:
f = open(reportfile,'wb')
bench.name = reportfile
pickle.dump(bench,f)
f.close()
except IOError as reason:
print('* Error opening/writing reportfile')
except IOError as reason:
print('* Error opening/writing reportfile %s: %s' % (
reportfile,
reason))
print()
if __name__ == '__main__':
PyBenchCmdline()