gecko-dev/servo/etc/memory_reports_over_time.py

150 строки
4.9 KiB
Python

#!/usr/bin/env python
# Copyright 2018 The Servo Project Developers. See the COPYRIGHT
# file at the top-level directory of this distribution.
#
# Licensed under the Apache License, Version 2.0 <LICENSE-APACHE or
# http://www.apache.org/licenses/LICENSE-2.0> or the MIT license
# <LICENSE-MIT or http://opensource.org/licenses/MIT>, at your
# option. This file may not be copied, modified, or distributed
# except according to those terms.
import json
import os
import sys
import tempfile
import webbrowser
def extract_memory_reports(lines):
in_report = False
report_lines = []
times = []
for line in lines:
if line.startswith('Begin memory reports'):
in_report = True
report_lines += [[]]
times += [line.strip().split()[-1]]
elif line == 'End memory reports\n':
in_report = False
elif in_report:
if line.startswith('|'):
report_lines[-1].append(line.strip())
return (report_lines, times)
def parse_memory_report(lines):
reports = {}
parents = []
last_separator_index = None
for line in lines:
assert(line[0] == '|')
line = line[1:]
if not line:
continue
separator_index = line.index('--')
if last_separator_index and separator_index <= last_separator_index:
while parents and parents[-1][1] >= separator_index:
parents.pop()
amount, unit, _, name = line.split()
dest_report = reports
for (parent, index) in parents:
dest_report = dest_report[parent]['children']
dest_report[name] = {
'amount': amount,
'unit': unit,
'children': {}
}
parents += [(name, separator_index)]
last_separator_index = separator_index
return reports
def transform_report_for_test(report):
transformed = {}
remaining = list(report.items())
while remaining:
(name, value) = remaining.pop()
transformed[name] = '%s %s' % (value['amount'], value['unit'])
remaining += map(lambda (k, v): (name + '/' + k, v), list(value['children'].items()))
return transformed
def test_extract_memory_reports():
input = ["Begin memory reports",
"|",
" 154.56 MiB -- explicit\n",
"| 107.88 MiB -- system-heap-unclassified\n",
"End memory reports\n"]
expected = ([['|', '| 107.88 MiB -- system-heap-unclassified']], ['reports'])
assert(extract_memory_reports(input) == expected)
return 0
def test():
input = '''|
| 23.89 MiB -- explicit
| 21.35 MiB -- jemalloc-heap-unclassified
| 2.54 MiB -- url(https://servo.org/)
| 2.16 MiB -- js
| 1.00 MiB -- gc-heap
| 0.77 MiB -- decommitted
| 1.00 MiB -- non-heap
| 0.27 MiB -- layout-thread
| 0.27 MiB -- stylist
| 0.12 MiB -- dom-tree
|
| 25.18 MiB -- jemalloc-heap-active'''
expected = {
'explicit': '23.89 MiB',
'explicit/jemalloc-heap-unclassified': '21.35 MiB',
'explicit/url(https://servo.org/)': '2.54 MiB',
'explicit/url(https://servo.org/)/js': '2.16 MiB',
'explicit/url(https://servo.org/)/js/gc-heap': '1.00 MiB',
'explicit/url(https://servo.org/)/js/gc-heap/decommitted': '0.77 MiB',
'explicit/url(https://servo.org/)/js/non-heap': '1.00 MiB',
'explicit/url(https://servo.org/)/layout-thread': '0.27 MiB',
'explicit/url(https://servo.org/)/layout-thread/stylist': '0.27 MiB',
'explicit/url(https://servo.org/)/dom-tree': '0.12 MiB',
'jemalloc-heap-active': '25.18 MiB',
}
report = parse_memory_report(input.split('\n'))
transformed = transform_report_for_test(report)
assert(sorted(transformed.keys()) == sorted(expected.keys()))
for k, v in transformed.items():
assert(v == expected[k])
test_extract_memory_reports()
return 0
def usage():
print('%s --test - run automated tests' % sys.argv[0])
print('%s file - extract all memory reports that are present in file' % sys.argv[0])
return 1
if __name__ == "__main__":
if len(sys.argv) == 1:
sys.exit(usage())
if sys.argv[1] == '--test':
sys.exit(test())
with open(sys.argv[1]) as f:
lines = f.readlines()
(reports, times) = extract_memory_reports(lines)
json_reports = []
for (report_lines, seconds) in zip(reports, times):
report = parse_memory_report(report_lines)
json_reports += [{'seconds': seconds, 'report': report}]
with tempfile.NamedTemporaryFile(delete=False) as output:
thisdir = os.path.dirname(os.path.abspath(__file__))
with open(os.path.join(thisdir, 'memory_chart.html')) as template:
content = template.read()
output.write(content.replace('[/* json data */]', json.dumps(json_reports)))
webbrowser.open_new_tab('file://' + output.name)