def run(self, accessor, opts):
"""Ask fake Graphite Web for points.
See command.CommandBase.
"""
# import here to make sure everything is setup.
os.environ['DJANGO_SETTINGS_MODULE'] = 'graphite.settings'
accessor.connect()
from django.conf import settings as django_settings
# Disable carbon link (enabled by default)
django_settings.CARBONLINK_HOSTS = []
# Make sure logging goes to stderr
django_settings.LOG_FILE_INFO = '-'
django_settings.LOG_FILE_EXCEPTION = '-'
django_settings.LOG_FILE_CACHE = '-'
django_settings.LOG_FILE_RENDERING = '-'
from graphite import util as graphite_util
from biggraphite.plugins import graphite
settings = bg_utils.settings_from_confattr(opts, prefix='')
metadata_cache = bg_utils.cache_from_settings(accessor, settings)
metadata_cache.open()
if opts.profile == "flamegraph":
flamegraph.start_profile_thread(fd=open("./perf.log", "w"))
finder = graphite.Finder(
directories=[],
accessor=accessor,
metadata_cache=metadata_cache
)
time_start = graphite_util.timestamp(opts.time_start)
time_end = graphite_util.timestamp(opts.time_end)
output_csv = opts.output_csv
results = finder.fetch(opts.patterns, time_start, time_end)
for i, result in enumerate(results):
# Change the output to match something that display_metrics
# can work with.
metric = _FakeMetric(result['path'])
time_start, time_end, step = result['time_info']
points = []
for i, v in enumerate(result['values']):
v = 0 if v is None else v
points.append((time_start + step * i, v))
result = (
points,
time_start,
time_end,
step
)
if not opts.no_output:
self._display_metric(metric, result, output_csv)
def make_flamegraph(filterx=None):
import flamegraph
import os
perf_log_path = os.path.join(os.path.dirname(__file__), "../tmp/perf.log")
flamegraph.start_profile_thread(fd=open(perf_log_path, "w"),
filter=filterx,
interval=0.001)
def make_flamegraph(filterx=None):
import flamegraph
perf_log_path = os.path.join(
os.path.dirname(__file__),
"/home/racecar/racecar-ws/src/simulator/headless_simulator/tmp/perf2.log"
)
flamegraph.start_profile_thread(fd=open(perf_log_path, "w"),
filter=filterx,
interval=0.001)
def start_flamegraphing():
# TODO: codify my workflow for profiling graphics vs simulation code.
# Basically, profiling graphics is done by using dev tools of whatever
# browser (in Firefox, F12 > Performance > Start Recording Performance)
# And profiling python simulation code is done by running this function,
# then processing PERF_FILE with https://github.com/brendangregg/FlameGraph
# The command is, without appropriate paths, is:
# dos2unix PERF_FILE && flamegraph.pl PERF_FILE > orbitx-perf.svg
# Where flamegraph.pl is from that brendangregg repo.
import flamegraph
flamegraph.start_profile_thread(fd=open(PERF_FILE, 'w'),
filter=r'(simthread|MainThread)')
def run(self, accessor, opts):
"""Ask fake Graphite Web for points.
See command.CommandBase.
"""
# import here to make sure everything is setup.
os.environ["DJANGO_SETTINGS_MODULE"] = "graphite.settings"
accessor.connect()
from django.conf import settings as django_settings
# Disable carbon link (enabled by default)
django_settings.CARBONLINK_HOSTS = []
# Make sure logging goes to stderr
django_settings.LOG_FILE_INFO = "-"
django_settings.LOG_FILE_EXCEPTION = "-"
django_settings.LOG_FILE_CACHE = "-"
django_settings.LOG_FILE_RENDERING = "-"
from graphite import util as graphite_util
from biggraphite.plugins import graphite
settings = bg_settings.settings_from_confattr(opts, prefix="")
metadata_cache = bg_cache_factory.cache_from_settings(accessor, settings)
metadata_cache.open()
if opts.profile == "flamegraph":
flamegraph.start_profile_thread(fd=open("./perf.log", "w"))
finder = graphite.Finder(
directories=[], accessor=accessor, metadata_cache=metadata_cache
)
time_start = graphite_util.timestamp(opts.time_start)
time_end = graphite_util.timestamp(opts.time_end)
output_csv = opts.output_csv
results = finder.fetch(opts.patterns, time_start, time_end)
for i, result in enumerate(results):
# Change the output to match something that display_metrics
# can work with.
metric = _FakeMetric(result["path"])
time_start, time_end, step = result["time_info"]
points = []
for i, v in enumerate(result["values"]):
v = 0 if v is None else v
points.append((time_start + step * i, v))
result = (points, time_start, time_end, step)
if not opts.no_output:
self._display_metric(metric, result, output_csv)
def main():
population = Population(150)
try:
blocks = sf.blocks('inputs/hello_world.wav', blocksize=512, overlap=32)
generations = tqdm(range(1))
for generation in generations:
block = next(blocks)
block = np.pad(block, (0, 512 - len(block)), 'constant')
fitness = population.evaluate(block)
population.reproduce(generation)
population.speciate(generation)
generations.set_postfix({
'G': '{:.4e}'.format(fitness[0]),
'D': '{:.4e}'.format(fitness[1])
})
except AssertionError:
pass
print(population)
print(population.enc_best_gen, population.enc_best_fit)
print(population.dec_best_gen, population.dec_best_fit)
print(population.crit_best_gen, population.crit_best_fit)
print()
import flamegraph
flamegraph.start_profile_thread(fd=open("./perf.log", "w"))
main()
def make_flamegraph(filterx=None):
import flamegraph, os
perf_log_path = os.path.join(os.path.dirname(__file__), "../tmp/perf.log")
flamegraph.start_profile_thread(fd=open(perf_log_path, "w"),
filter=filterx,
interval=0.001)
"""
Example usage of flamegraph.
To view a flamegraph run these commands:
$ python example.py
$ flamegraph.pl perf.log > perf.svg
$ inkview perf.svg
"""
import time
import sys
import flamegraph
def foo():
time.sleep(.1)
bar()
def bar():
time.sleep(.05)
if __name__ == "__main__":
flamegraph.start_profile_thread(fd=open("./perf.log", "w"))
N = 10
for x in xrange(N):
print "{}/{}".format(x, N)
foo()
import flamegraph
import time
flamegraph.
flamegraph.start_profile_thread(fd=open("./graphs/perf_{}.log".format(time.time()), "w"))
DATASET = 'ng20'
H = 10
USED_SPARSE_MATRIX = scipy.sparse.lil_matrix
PERF_LOG = './perf.log'
# Warm up primes cache
primes.get_highest_prime_range()
# Get dataset
X, Y = dataset_helper.get_concept_map_for_dataset(DATASET, graphs_only=True)
# Convert to adj
X = graph_helper.convert_graphs_to_adjs_tuples(X, copy=True)
# Same label for all nodes
adj, labels = zip(*X)
labels = [[1] * len(x) for x in labels]
X = list(zip(adj, labels))
# Start profiling
thread = flamegraph.start_profile_thread(fd=open(PERF_LOG, "w"))
start = time()
# Calculate fast_wl
phi_lists, new_label_lookups, new_label_counters = fast_wl.transform(
X, h=H, used_matrix_type=USED_SPARSE_MATRIX)
# Convert to sparse lil_matrix for "fancy" indexing (needed for other pipeline elements)
phi_lists = [phi.tolil().T for phi in phi_lists]
print(time_utils.seconds_to_human_readable(time() - start))
