Chromium Code Reviews
chromiumcodereview-hr@appspot.gserviceaccount.com (chromiumcodereview-hr) | Please choose your nickname with Settings | Help | Chromium Project | Gerrit Changes | Sign out
(44)

Side by Side Diff: tools/auto_bisect/bisect_perf_regression_test.py

Issue 764733005: Lower "confidence score" required in order to not abort, and refactor test. (Closed) Base URL: https://chromium.googlesource.com/chromium/src.git@master
Patch Set: Created 6 years ago
Use n/p to move between diff chunks; N/P to move between comments. Draft comments are only viewable by you.
Jump to:
View unified diff | Download patch
« no previous file with comments | « tools/auto_bisect/bisect_perf_regression.py ('k') | no next file » | no next file with comments »
Toggle Intra-line Diffs ('i') | Expand Comments ('e') | Collapse Comments ('c') | Show Comments Hide Comments ('s')
OLDNEW
1 # Copyright 2014 The Chromium Authors. All rights reserved. 1 # Copyright 2014 The Chromium Authors. All rights reserved.
2 # Use of this source code is governed by a BSD-style license that can be 2 # Use of this source code is governed by a BSD-style license that can be
3 # found in the LICENSE file. 3 # found in the LICENSE file.
4 4
5 import os 5 import os
6 import re 6 import re
7 import shutil 7 import shutil
8 import sys 8 import sys
9 import unittest 9 import unittest
10 10
(...skipping 10 matching lines...) Expand all
21 CLEAR_NON_REGRESSION = [ 21 CLEAR_NON_REGRESSION = [
22 # Mean: 30.223 Std. Dev.: 11.383 22 # Mean: 30.223 Std. Dev.: 11.383
23 [[16.886], [16.909], [16.99], [17.723], [17.952], [18.118], [19.028], 23 [[16.886], [16.909], [16.99], [17.723], [17.952], [18.118], [19.028],
24 [19.552], [21.954], [38.573], [38.839], [38.965], [40.007], [40.572], 24 [19.552], [21.954], [38.573], [38.839], [38.965], [40.007], [40.572],
25 [41.491], [42.002], [42.33], [43.109], [43.238]], 25 [41.491], [42.002], [42.33], [43.109], [43.238]],
26 # Mean: 34.76 Std. Dev.: 11.516 26 # Mean: 34.76 Std. Dev.: 11.516
27 [[16.426], [17.347], [20.593], [21.177], [22.791], [27.843], [28.383], 27 [[16.426], [17.347], [20.593], [21.177], [22.791], [27.843], [28.383],
28 [28.46], [29.143], [40.058], [40.303], [40.558], [41.918], [42.44], 28 [28.46], [29.143], [40.058], [40.303], [40.558], [41.918], [42.44],
29 [45.223], [46.494], [50.002], [50.625], [50.839]] 29 [45.223], [46.494], [50.002], [50.625], [50.839]]
30 ] 30 ]
31
31 # Regression confidence: ~ 90% 32 # Regression confidence: ~ 90%
32 ALMOST_REGRESSION = [ 33 ALMOST_REGRESSION = [
33 # Mean: 30.042 Std. Dev.: 2.002 34 # Mean: 30.042 Std. Dev.: 2.002
34 [[26.146], [28.04], [28.053], [28.074], [28.168], [28.209], [28.471], 35 [[26.146], [28.04], [28.053], [28.074], [28.168], [28.209], [28.471],
35 [28.652], [28.664], [30.862], [30.973], [31.002], [31.897], [31.929], 36 [28.652], [28.664], [30.862], [30.973], [31.002], [31.897], [31.929],
36 [31.99], [32.214], [32.323], [32.452], [32.696]], 37 [31.99], [32.214], [32.323], [32.452], [32.696]],
37 # Mean: 33.008 Std. Dev.: 4.265 38 # Mean: 33.008 Std. Dev.: 4.265
38 [[34.963], [30.741], [39.677], [39.512], [34.314], [31.39], [34.361], 39 [[34.963], [30.741], [39.677], [39.512], [34.314], [31.39], [34.361],
39 [25.2], [30.489], [29.434]] 40 [25.2], [30.489], [29.434]]
40 ] 41 ]
42
41 # Regression confidence: ~ 98% 43 # Regression confidence: ~ 98%
42 BARELY_REGRESSION = [ 44 BARELY_REGRESSION = [
43 # Mean: 28.828 Std. Dev.: 1.993 45 # Mean: 28.828 Std. Dev.: 1.993
44 [[26.96], [27.605], [27.768], [27.829], [28.006], [28.206], [28.393], 46 [[26.96], [27.605], [27.768], [27.829], [28.006], [28.206], [28.393],
45 [28.911], [28.933], [30.38], [30.462], [30.808], [31.74], [31.805], 47 [28.911], [28.933], [30.38], [30.462], [30.808], [31.74], [31.805],
46 [31.899], [32.077], [32.454], [32.597], [33.155]], 48 [31.899], [32.077], [32.454], [32.597], [33.155]],
47 # Mean: 31.156 Std. Dev.: 1.980 49 # Mean: 31.156 Std. Dev.: 1.980
48 [[28.729], [29.112], [29.258], [29.454], [29.789], [30.036], [30.098], 50 [[28.729], [29.112], [29.258], [29.454], [29.789], [30.036], [30.098],
49 [30.174], [30.534], [32.285], [32.295], [32.552], [32.572], [32.967], 51 [30.174], [30.534], [32.285], [32.295], [32.552], [32.572], [32.967],
50 [33.165], [33.403], [33.588], [33.744], [34.147], [35.84]] 52 [33.165], [33.403], [33.588], [33.744], [34.147], [35.84]]
51 ] 53 ]
54
52 # Regression confidence: 99.5% 55 # Regression confidence: 99.5%
53 CLEAR_REGRESSION = [ 56 CLEAR_REGRESSION = [
54 # Mean: 30.254 Std. Dev.: 2.987 57 # Mean: 30.254 Std. Dev.: 2.987
55 [[26.494], [26.621], [26.701], [26.997], [26.997], [27.05], [27.37], 58 [[26.494], [26.621], [26.701], [26.997], [26.997], [27.05], [27.37],
56 [27.488], [27.556], [31.846], [32.192], [32.21], [32.586], [32.596], 59 [27.488], [27.556], [31.846], [32.192], [32.21], [32.586], [32.596],
57 [32.618], [32.95], [32.979], [33.421], [33.457], [34.97]], 60 [32.618], [32.95], [32.979], [33.421], [33.457], [34.97]],
58 # Mean: 33.190 Std. Dev.: 2.972 61 # Mean: 33.190 Std. Dev.: 2.972
59 [[29.547], [29.713], [29.835], [30.132], [30.132], [30.33], [30.406], 62 [[29.547], [29.713], [29.835], [30.132], [30.132], [30.33], [30.406],
60 [30.592], [30.72], [34.486], [35.247], [35.253], [35.335], [35.378], 63 [30.592], [30.72], [34.486], [35.247], [35.253], [35.335], [35.378],
61 [35.934], [36.233], [36.41], [36.947], [37.982]] 64 [35.934], [36.233], [36.41], [36.947], [37.982]]
62 ] 65 ]
63 66
64 # Regression confidence > 95%, taken from: crbug.com/434318 67 # Regression confidence > 95%, taken from: crbug.com/434318
65 # Specifically from Builder android_nexus10_perf_bisect Build #1198 68 # Specifically from Builder android_nexus10_perf_bisect Build #1198
66 MULTIPLE_VALUES = [ 69 MULTIPLE_VALUES = [
67 [ 70 [
68 [18.916000,22.371000,8.527000,5.877000,5.407000,9.476000,8.100000, 71 [18.916, 22.371, 8.527, 5.877, 5.407, 9.476, 8.100, 5.334,
69 5.334000,4.507000,4.842000,8.485000,8.308000,27.490000,4.560000, 72 4.507, 4.842, 8.485, 8.308, 27.490, 4.560, 4.804, 23.068, 17.577,
70 4.804000,23.068000,17.577000,17.346000,26.738000,60.330000,32.307000, 73 17.346, 26.738, 60.330, 32.307, 5.468, 27.803, 27.373, 17.823,
71 5.468000,27.803000,27.373000,17.823000,5.158000,27.439000,5.236000, 74 5.158, 27.439, 5.236, 11.413],
72 11.413000 75 [18.999, 22.642, 8.158, 5.995, 5.495, 9.499, 8.092, 5.324,
73 ], 76 4.468, 4.788, 8.248, 7.853, 27.533, 4.410, 4.622, 22.341, 22.313,
74 [18.999000,22.642000,8.158000,5.995000,5.495000,9.499000,8.092000, 77 17.072, 26.731, 57.513, 33.001, 5.500, 28.297, 27.277, 26.462,
75 5.324000,4.468000,4.788000,8.248000,7.853000,27.533000,4.410000, 78 5.009, 27.361, 5.130, 10.955]
76 4.622000,22.341000,22.313000,17.072000,26.731000,57.513000,33.001000,
77 5.500000,28.297000,27.277000,26.462000,5.009000,27.361000,5.130000,
78 10.955000
79 ]
80 ], 79 ],
81 [ 80 [
82 [18.238000,22.365000,8.555000,5.939000,5.437000,9.463000,7.047000, 81 [18.238, 22.365, 8.555, 5.939, 5.437, 9.463, 7.047, 5.345, 4.517,
83 5.345000,4.517000,4.796000,8.593000,7.901000,27.499000,4.378000, 82 4.796, 8.593, 7.901, 27.499, 4.378, 5.040, 4.904, 4.816, 4.828,
84 5.040000,4.904000,4.816000,4.828000,4.853000,57.363000,34.184000, 83 4.853, 57.363, 34.184, 5.482, 28.190, 27.290, 26.694, 5.099,
85 5.482000,28.190000,27.290000,26.694000,5.099000,4.905000,5.290000, 84 4.905, 5.290, 4.813],
86 4.813000 85 [18.301, 22.522, 8.035, 6.021, 5.565, 9.037, 6.998, 5.321, 4.485,
87 ], 86 4.768, 8.397, 7.865, 27.636, 4.640, 5.015, 4.962, 4.933, 4.977,
88 [18.301000,22.522000,8.035000,6.021000,5.565000,9.037000,6.998000, 87 4.961, 60.648, 34.593, 5.538, 28.454, 27.297, 26.490, 5.099, 5,
89 5.321000,4.485000,4.768000,8.397000,7.865000,27.636000,4.640000, 88 5.247, 4.945],
90 5.015000,4.962000,4.933000,4.977000,4.961000,60.648000,34.593000, 89 [18.907, 23.368, 8.100, 6.169, 5.621, 9.971, 8.161, 5.331, 4.513,
91 5.538000,28.454000,27.297000,26.490000,5.099000,5,5.247000,4.945000 90 4.837, 8.255, 7.852, 26.209, 4.388, 5.045, 5.029, 5.032, 4.946,
92 ], 91 4.973, 60.334, 33.377, 5.499, 28.275, 27.550, 26.103, 5.108,
93 [18.907000,23.368000,8.100000,6.169000,5.621000,9.971000,8.161000, 92 4.951, 5.285, 4.910],
94 5.331000,4.513000,4.837000,8.255000,7.852000,26.209000,4.388000, 93 [18.715, 23.748, 8.128, 6.148, 5.691, 9.361, 8.106, 5.334, 4.528,
95 5.045000,5.029000,5.032000,4.946000,4.973000,60.334000,33.377000, 94 4.965, 8.261, 7.851, 27.282, 4.391, 4.949, 4.981, 4.964, 4.935,
96 5.499000,28.275000,27.550000,26.103000,5.108000,4.951000,5.285000, 95 4.933, 60.231, 33.361, 5.489, 28.106, 27.457, 26.648, 5.108,
97 4.910000 96 4.963, 5.272, 4.954]
98 ],
99 [18.715000,23.748000,8.128000,6.148000,5.691000,9.361000,8.106000,
100 5.334000,4.528000,4.965000,8.261000,7.851000,27.282000,4.391000,
101 4.949000,4.981000,4.964000,4.935000,4.933000,60.231000,33.361000,
102 5.489000,28.106000,27.457000,26.648000,5.108000,4.963000,5.272000,
103 4.954000
104 ]
105 ] 97 ]
106 ] 98 ]
107 99
108 # Default options for the dry run 100 # Default options for the dry run
109 DEFAULT_OPTIONS = { 101 DEFAULT_OPTIONS = {
110 'debug_ignore_build': True, 102 'debug_ignore_build': True,
111 'debug_ignore_sync': True, 103 'debug_ignore_sync': True,
112 'debug_ignore_perf_test': True, 104 'debug_ignore_perf_test': True,
113 'debug_ignore_regression_confidence': True, 105 'debug_ignore_regression_confidence': True,
114 'command': 'fake_command', 106 'command': 'fake_command',
115 'metric': 'fake/metric', 107 'metric': 'fake/metric',
116 'good_revision': 280000, 108 'good_revision': 280000,
117 'bad_revision': 280005, 109 'bad_revision': 280005,
118 } 110 }
119 111
120 # This global is a placeholder for a generator to be defined by the testcases 112 # This global is a placeholder for a generator to be defined by the test cases
121 # that use _MockRunTest 113 # that use _MockRunTests.
122 _MockResultsGenerator = (x for x in []) 114 _MockResultsGenerator = (x for x in [])
123 115
116
117 def _MockRunTests(*args, **kwargs):
118 _, _ = args, kwargs
119 return _FakeTestResult(_MockResultsGenerator.next())
120
121
124 def _FakeTestResult(values): 122 def _FakeTestResult(values):
125 result_dict = {'mean': 0.0, 'std_err': 0.0, 'std_dev': 0.0, 'values': values} 123 result_dict = {'mean': 0.0, 'std_err': 0.0, 'std_dev': 0.0, 'values': values}
126 success_code = 0 124 success_code = 0
127 return (result_dict, success_code) 125 return (result_dict, success_code)
128 126
129 127
130 def _MockRunTests(*args, **kwargs):
131 _, _ = args, kwargs
132 return _FakeTestResult(_MockResultsGenerator.next())
133
134
135 def _GetBisectPerformanceMetricsInstance(options_dict): 128 def _GetBisectPerformanceMetricsInstance(options_dict):
136 """Returns an instance of the BisectPerformanceMetrics class.""" 129 """Returns an instance of the BisectPerformanceMetrics class."""
137 opts = bisect_perf_regression.BisectOptions.FromDict(options_dict) 130 opts = bisect_perf_regression.BisectOptions.FromDict(options_dict)
138 return bisect_perf_regression.BisectPerformanceMetrics(opts, os.getcwd()) 131 return bisect_perf_regression.BisectPerformanceMetrics(opts, os.getcwd())
139 132
140 133
141 def _GetExtendedOptions(improvement_dir, fake_first, ignore_confidence=True): 134 def _GetExtendedOptions(improvement_dir, fake_first, ignore_confidence=True):
142 """Returns the a copy of the default options dict plus some options.""" 135 """Returns the a copy of the default options dict plus some options."""
143 result = dict(DEFAULT_OPTIONS) 136 result = dict(DEFAULT_OPTIONS)
144 result.update({ 137 result.update({
(...skipping 201 matching lines...) Expand 10 before | Expand all | Expand 10 after
346 339
347 def testBisectImprovementDirectionSucceeds(self): 340 def testBisectImprovementDirectionSucceeds(self):
348 """Bisects with improvement direction matching regression range.""" 341 """Bisects with improvement direction matching regression range."""
349 # Test result goes from 0 to 100 where lower is better 342 # Test result goes from 0 to 100 where lower is better
350 results = _GenericDryRun(_GetExtendedOptions(-1, 100)) 343 results = _GenericDryRun(_GetExtendedOptions(-1, 100))
351 self.assertIsNone(results.error) 344 self.assertIsNone(results.error)
352 # Test result goes from 0 to -100 where higher is better 345 # Test result goes from 0 to -100 where higher is better
353 results = _GenericDryRun(_GetExtendedOptions(1, -100)) 346 results = _GenericDryRun(_GetExtendedOptions(1, -100))
354 self.assertIsNone(results.error) 347 self.assertIsNone(results.error)
355 348
356 @mock.patch('bisect_perf_regression.BisectPerformanceMetrics.' 349 def _CheckAbortsEarly(self, results):
357 'RunPerformanceTestAndParseResults', _MockRunTests) 350 """Returns True if the bisect job would abort early."""
358 def testBisectStopsOnDoubtfulRegression(self):
359 global _MockResultsGenerator 351 global _MockResultsGenerator
360 _MockResultsGenerator = (rs for rs in CLEAR_NON_REGRESSION) 352 _MockResultsGenerator = (r for r in results)
361 results = _GenericDryRun(_GetExtendedOptions(0, 0, False)) 353 bisect_class = bisect_perf_regression.BisectPerformanceMetrics
362 confidence_warnings = [x for x in results.warnings if x.startswith( 354 original_run_tests = bisect_class.RunPerformanceTestAndParseResults
363 '\nWe could not reproduce the regression')] 355 bisect_class.RunPerformanceTestAndParseResults = _MockRunTests
364 self.assertGreater(len(confidence_warnings), 0)
365 356
366 _MockResultsGenerator = (rs for rs in ALMOST_REGRESSION) 357 try:
367 results = _GenericDryRun(_GetExtendedOptions(0, 0, False)) 358 _GenericDryRun(_GetExtendedOptions(0, 0, False))
368 confidence_warnings = [x for x in results.warnings if x.startswith( 359 except StopIteration:
369 '\nWe could not reproduce the regression')] 360 # If StopIteration was raised, that means that the next value after
370 self.assertGreater(len(confidence_warnings), 0) 361 # the first two values was requested, so the job was not aborted.
362 return False
363 finally:
364 bisect_class.RunPerformanceTestAndParseResults = original_run_tests
371 365
372 @mock.patch('bisect_perf_regression.BisectPerformanceMetrics.' 366 # If the job was aborted, there should be a warning about it.
373 'RunPerformanceTestAndParseResults', _MockRunTests) 367 assert [w for w in results.warnings
374 def testBisectContinuesOnClearRegression(self): 368 if 'could not reproduce the regression' in w]
375 global _MockResultsGenerator 369 return True
376 _MockResultsGenerator = (rs for rs in CLEAR_REGRESSION)
377 with self.assertRaises(StopIteration):
378 _GenericDryRun(_GetExtendedOptions(0, 0, False))
379 370
380 _MockResultsGenerator = (rs for rs in BARELY_REGRESSION) 371 def testBisectStopsOnClearUnclearRegression(self):
381 with self.assertRaises(StopIteration): 372 self.assertTrue(self._CheckAbortsEarly(CLEAR_NON_REGRESSION))
382 _GenericDryRun(_GetExtendedOptions(0, 0, False))
383 373
384 _MockResultsGenerator = (rs for rs in MULTIPLE_VALUES) 374 def testBisectStopsOnClearUnclearRegression(self):
385 with self.assertRaises(StopIteration): 375 self.assertFalse(self._CheckAbortsEarly(ALMOST_REGRESSION))
386 _GenericDryRun(_GetExtendedOptions(0, 0, False)) 376
377 def testBisectStopsOnClearUnclearRegression(self):
378 self.assertFalse(self._CheckAbortsEarly(CLEAR_REGRESSION))
379
380 def testBisectStopsOnClearUnclearRegression(self):
381 self.assertFalse(self._CheckAbortsEarly(BARELY_REGRESSION))
382
383 def testBisectStopsOnClearUnclearRegression(self):
384 self.assertFalse(self._CheckAbortsEarly(MULTIPLE_VALUES))
387 385
388 def testGetCommitPosition(self): 386 def testGetCommitPosition(self):
389 cp_git_rev = '7017a81991de983e12ab50dfc071c70e06979531' 387 cp_git_rev = '7017a81991de983e12ab50dfc071c70e06979531'
390 self.assertEqual(291765, source_control.GetCommitPosition(cp_git_rev)) 388 self.assertEqual(291765, source_control.GetCommitPosition(cp_git_rev))
391 389
392 svn_git_rev = 'e6db23a037cad47299a94b155b95eebd1ee61a58' 390 svn_git_rev = 'e6db23a037cad47299a94b155b95eebd1ee61a58'
393 self.assertEqual(291467, source_control.GetCommitPosition(svn_git_rev)) 391 self.assertEqual(291467, source_control.GetCommitPosition(svn_git_rev))
394 392
395 def testGetCommitPositionForV8(self): 393 def testGetCommitPositionForV8(self):
396 bisect_instance = _GetBisectPerformanceMetricsInstance(DEFAULT_OPTIONS) 394 bisect_instance = _GetBisectPerformanceMetricsInstance(DEFAULT_OPTIONS)
(...skipping 209 matching lines...) Expand 10 before | Expand all | Expand 10 after
606 ], (None, 0)), 604 ], (None, 0)),
607 ] 605 ]
608 self._SetupRunGitMock(try_cmd) 606 self._SetupRunGitMock(try_cmd)
609 bisect_perf_regression._BuilderTryjob( 607 bisect_perf_regression._BuilderTryjob(
610 git_revision, bot_name, bisect_job_name, patch) 608 git_revision, bot_name, bisect_job_name, patch)
611 609
612 610
613 if __name__ == '__main__': 611 if __name__ == '__main__':
614 unittest.main() 612 unittest.main()
615 613
OLDNEW
« no previous file with comments | « tools/auto_bisect/bisect_perf_regression.py ('k') | no next file » | no next file with comments »

Powered by Google App Engine
This is Rietveld 408576698