| OLD | NEW |
|---|
| (Empty) |
| 1 #!/usr/bin/python | |
| 2 | |
| 3 # Copyright (c) 2011, the Dart project authors. Please see the AUTHORS file | |
| 4 # for details. All rights reserved. Use of this source code is governed by a | |
| 5 # BSD-style license that can be found in the LICENSE file. | |
| 6 | |
| 7 import datetime | |
| 8 import math | |
| 9 from matplotlib.font_manager import FontProperties | |
| 10 import matplotlib.pyplot as plt | |
| 11 import optparse | |
| 12 import os | |
| 13 from os.path import dirname, abspath | |
| 14 import platform | |
| 15 import shutil | |
| 16 import subprocess | |
| 17 import time | |
| 18 import traceback | |
| 19 import sys | |
| 20 | |
| 21 TOOLS_PATH = os.path.join(dirname(dirname(dirname(abspath(__file__))))) | |
| 22 sys.path.append(TOOLS_PATH) | |
| 23 import utils | |
| 24 | |
| 25 """This script runs to track performance and correctness progress of | |
| 26 different svn revisions. It tests to see if there a newer version of the code on | |
| 27 the server, and will sync and run the performance tests if so.""" | |
| 28 | |
| 29 DART_INSTALL_LOCATION = os.path.join(dirname(abspath(__file__)), | |
| 30 '..', '..', '..') | |
| 31 V8_MEAN = 'V8 Mean' | |
| 32 FROG_MEAN = 'frog Mean' | |
| 33 COMMAND_LINE = 'commandline' | |
| 34 V8 = 'v8' | |
| 35 FROG = 'frog' | |
| 36 V8_AND_FROG = [V8, FROG] | |
| 37 CORRECTNESS = 'Percent passing' | |
| 38 COLORS = ['blue', 'green', 'red', 'cyan', 'magenta', 'black'] | |
| 39 GRAPH_OUT_DIR = 'graphs' | |
| 40 | |
| 41 BROWSER_PERF = 'browser-perf' | |
| 42 TIME_SIZE = 'code-time-size' | |
| 43 CL_PERF = 'cl-results' | |
| 44 BROWSER_CORRECTNESS = 'browser-correctness' | |
| 45 | |
| 46 SLEEP_TIME = 200 | |
| 47 VERBOSE = False | |
| 48 HAS_SHELL = False | |
| 49 if platform.system() == 'Windows': | |
| 50 # On Windows, shell must be true to get the correct environment variables. | |
| 51 HAS_SHELL = True | |
| 52 | |
| 53 """First, some utility methods.""" | |
| 54 | |
| 55 def run_cmd(cmd_list, outfile=None, append=False): | |
| 56 """Run the specified command and print out any output to stdout. | |
| 57 | |
| 58 Args: | |
| 59 cmd_list: a list of strings that make up the command to run | |
| 60 outfile: a string indicating the name of the file that we should write | |
| 61 stdout to | |
| 62 append: True if we want to append to the file instead of overwriting it""" | |
| 63 if VERBOSE: | |
| 64 print ' '.join(cmd_list) | |
| 65 out = subprocess.PIPE | |
| 66 if outfile: | |
| 67 mode = 'w' | |
| 68 if append: | |
| 69 mode = 'a' | |
| 70 out = open(outfile, mode) | |
| 71 if append: | |
| 72 # Annoying Windows "feature" -- append doesn't actually append unless you | |
| 73 # explicitly go to the end of the file. | |
| 74 # http://mail.python.org/pipermail/python-list/2009-October/1221859.html | |
| 75 out.seek(0, os.SEEK_END) | |
| 76 p = subprocess.Popen(cmd_list, stdout = out, stderr = subprocess.PIPE, | |
| 77 shell=HAS_SHELL) | |
| 78 output, not_used = p.communicate(); | |
| 79 if output: | |
| 80 print output | |
| 81 return output | |
| 82 | |
| 83 def time_cmd(cmd): | |
| 84 """Determine the amount of (real) time it takes to execute a given command.""" | |
| 85 start = time.time() | |
| 86 run_cmd(cmd) | |
| 87 return time.time() - start | |
| 88 | |
| 89 def sync_and_build(): | |
| 90 """Make sure we have the latest version of of the repo, and build it. We | |
| 91 begin and end standing in DART_INSTALL_LOCATION. | |
| 92 | |
| 93 Returns: | |
| 94 err_code = 1 if there was a problem building.""" | |
| 95 os.chdir(DART_INSTALL_LOCATION) | |
| 96 #Revert our newly built minfrog to prevent conflicts when we update | |
| 97 run_cmd(['svn', 'revert', os.path.join(os.getcwd(), 'frog', 'minfrog')]) | |
| 98 | |
| 99 run_cmd(['gclient', 'sync']) | |
| 100 | |
| 101 # On Windows, the output directory is marked as "Read Only," which causes an | |
| 102 # error to be thrown when we use shutil.rmtree. This helper function changes | |
| 103 # the permissions so we can still delete the directory. | |
| 104 def on_rm_error(func, path, exc_info): | |
| 105 os.chmod(path, stat.S_IWRITE) | |
| 106 os.unlink(path) | |
| 107 # TODO(efortuna): building the sdk locally is a band-aid until all build | |
| 108 # platform SDKs are hosted in Google storage. Pull from https://sandbox. | |
| 109 # google.com/storage/?arg=dart-dump-render-tree#dart-dump-render-tree%2Fsdk | |
| 110 # eventually. | |
| 111 # TODO(efortuna): Currently always building ia32 architecture because we don't | |
| 112 # have test statistics for what's passing on x64. Eliminate arch specification | |
| 113 # when we have tests running on x64, too. | |
| 114 shutil.rmtree(os.path.join(os.getcwd(), | |
| 115 utils.GetBuildRoot(utils.GuessOS(), 'release', 'ia32')), | |
| 116 onerror=on_rm_error) | |
| 117 lines = run_cmd([os.path.join('.', 'tools', 'build.py'), '-m', 'release', | |
| 118 '--arch=ia32', 'create_sdk']) | |
| 119 | |
| 120 for line in lines: | |
| 121 if 'BUILD FAILED' in lines: | |
| 122 # Someone checked in a broken build! Just stop trying to make it work | |
| 123 # and wait to try again. | |
| 124 print 'Broken Build' | |
| 125 return 1 | |
| 126 return 0 | |
| 127 | |
| 128 def ensure_output_directory(dir_name): | |
| 129 """Test that the listed directory name exists, and if not, create one for | |
| 130 our output to be placed. | |
| 131 | |
| 132 Args: | |
| 133 dir_name: the directory we will create if it does not exist.""" | |
| 134 dir_path = os.path.join(DART_INSTALL_LOCATION, 'tools', 'testing', | |
| 135 'perf_testing', dir_name) | |
| 136 if not os.path.exists(dir_path): | |
| 137 os.mkdir(dir_path) | |
| 138 print 'Creating output directory ', dir_path | |
| 139 | |
| 140 def has_new_code(): | |
| 141 """Tests if there are any newer versions of files on the server.""" | |
| 142 os.chdir(DART_INSTALL_LOCATION) | |
| 143 results = run_cmd(['svn', 'st', '-u']) | |
| 144 for line in results: | |
| 145 if '*' in line: | |
| 146 return True | |
| 147 return False | |
| 148 | |
| 149 def get_browsers(): | |
| 150 browsers = ['ff', 'chrome'] | |
| 151 if platform.system() == 'Darwin': | |
| 152 browsers += ['safari'] | |
| 153 if platform.system() == 'Windows': | |
| 154 browsers += ['ie'] | |
| 155 return browsers | |
| 156 | |
| 157 def get_versions(): | |
| 158 return V8_AND_FROG | |
| 159 | |
| 160 def get_benchmarks(): | |
| 161 return ['Mandelbrot', 'DeltaBlue', 'Richards', 'NBody', 'BinaryTrees', | |
| 162 'Fannkuch', 'Meteor', 'BubbleSort', 'Fibonacci', 'Loop', 'Permute', | |
| 163 'Queens', 'QuickSort', 'Recurse', 'Sieve', 'Sum', 'Tak', 'Takl', 'Towers', | |
| 164 'TreeSort'] | |
| 165 | |
| 166 def get_os_directory(): | |
| 167 """Specifies the name of the directory for the testing build of dart, which | |
| 168 has yet a different naming convention from utils.getBuildRoot(...).""" | |
| 169 if platform.system() == 'Windows': | |
| 170 return 'windows' | |
| 171 elif platform.system() == 'Darwin': | |
| 172 return 'macos' | |
| 173 else: | |
| 174 return 'linux' | |
| 175 | |
| 176 def upload_to_app_engine(): | |
| 177 """Upload our results to our appengine server.""" | |
| 178 # TODO(efortuna): This is the most basic way to get the data up | |
| 179 # for others to view. Revisit this once we're serving nicer graphs (Google | |
| 180 # Chart Tools) and from multiple perfbots and once we're in a position to | |
| 181 # organize the data in a useful manner(!!). | |
| 182 os.chdir(os.path.join(DART_INSTALL_LOCATION, 'tools', 'testing', | |
| 183 'perf_testing')) | |
| 184 shutil.rmtree(os.path.join('appengine', 'static', 'graphs'), | |
| 185 ignore_errors=True) | |
| 186 shutil.copytree('graphs', os.path.join('appengine', 'static', 'graphs')) | |
| 187 shutil.copyfile('index.html', os.path.join('appengine', 'static', | |
| 188 'index.html')) | |
| 189 run_cmd(['../../../third_party/appengine-python/1.5.4/appcfg.py', 'update', | |
| 190 'appengine/']) | |
| 191 | |
| 192 class TestRunner(object): | |
| 193 """The base clas to provide shared code for different tests we will run and | |
| 194 graph.""" | |
| 195 | |
| 196 def __init__(self, result_folder_name, platform_list, v8_and_or_frog_list, | |
| 197 values_list): | |
| 198 """Args: | |
| 199 result_folder_name the name of the folder where a tracefile of | |
| 200 performance results will be stored. | |
| 201 platform_list a list containing the platform(s) that our data has been | |
| 202 run on. (command line, firefox, chrome, etc) | |
| 203 v8_and_or_frog_list a list specifying whether we hold data about Frog | |
| 204 generated code, plain JS code (v8), or a combination of both. | |
| 205 values_list a list containing the type of data we will be graphing | |
| 206 (benchmarks, percentage passing, etc)""" | |
| 207 self.result_folder_name = result_folder_name | |
| 208 # cur_time is used as a timestamp of when this performance test was run. | |
| 209 self.cur_time = str(time.mktime(datetime.datetime.now().timetuple())) | |
| 210 self.browser_color = {'chrome': 'green', 'ie': 'blue', 'ff': 'red', | |
| 211 'safari':'black'} | |
| 212 self.values_list = values_list | |
| 213 self.platform_list = platform_list | |
| 214 self.revision_dict = dict() | |
| 215 self.values_dict = dict() | |
| 216 self.color_index = 0 | |
| 217 for platform in platform_list: | |
| 218 self.revision_dict[platform] = dict() | |
| 219 self.values_dict[platform] = dict() | |
| 220 for f in v8_and_or_frog_list: | |
| 221 self.revision_dict[platform][f] = dict() | |
| 222 self.values_dict[platform][f] = dict() | |
| 223 for val in values_list: | |
| 224 self.revision_dict[platform][f][val] = [] | |
| 225 self.values_dict[platform][f][val] = [] | |
| 226 if V8 in v8_and_or_frog_list: | |
| 227 self.revision_dict[platform][V8][V8_MEAN] = [] | |
| 228 self.values_dict[platform][V8][V8_MEAN] = [] | |
| 229 if FROG in v8_and_or_frog_list: | |
| 230 self.revision_dict[platform][FROG][FROG_MEAN] = [] | |
| 231 self.values_dict[platform][FROG][FROG_MEAN] = [] | |
| 232 | |
| 233 def get_color(self): | |
| 234 color = COLORS[self.color_index] | |
| 235 self.color_index = (self.color_index + 1) % len(COLORS) | |
| 236 return color | |
| 237 | |
| 238 def style_and_save_perf_plot(self, chart_title, y_axis_label, size_x, size_y, | |
| 239 legend_loc, filename, platform_list, v8_and_or_frog_list, values_list, | |
| 240 should_clear_axes=True): | |
| 241 """Sets style preferences for chart boilerplate that is consistent across | |
| 242 all charts, and saves the chart as a png. | |
| 243 | |
| 244 Args: | |
| 245 size_x: the size of the printed chart, in inches, in the horizontal | |
| 246 direction | |
| 247 size_y: the size of the printed chart, in inches in the vertical direction | |
| 248 legend_loc: the location of the legend in on the chart. See suitable | |
| 249 arguments for the loc argument in matplotlib | |
| 250 filename: the filename that we want to save the resulting chart as | |
| 251 platform_list: a list containing the platform(s) that our data has been | |
| 252 run on. (command line, firefox, chrome, etc) | |
| 253 values_list: a list containing the type of data we will be graphing | |
| 254 (performance, percentage passing, etc) | |
| 255 should_clear_axes: True if we want to create a fresh graph, instead of | |
| 256 plotting additional lines on the current graph.""" | |
| 257 if should_clear_axes: | |
| 258 plt.cla() # cla = clear current axes | |
| 259 for platform in platform_list: | |
| 260 for f in v8_and_or_frog_list: | |
| 261 for val in values_list: | |
| 262 plt.plot(self.revision_dict[platform][f][val], | |
| 263 self.values_dict[platform][f][val], | |
| 264 color=self.get_color(), label='%s-%s-%s' % (platform, f, val)) | |
| 265 | |
| 266 plt.xlabel('Revision Number') | |
| 267 plt.ylabel(y_axis_label) | |
| 268 plt.title(chart_title) | |
| 269 fontP = FontProperties() | |
| 270 fontP.set_size('small') | |
| 271 plt.legend(loc=legend_loc, prop = fontP) | |
| 272 | |
| 273 fig = plt.gcf() | |
| 274 fig.set_size_inches(size_x, size_y) | |
| 275 fig.savefig(os.path.join(GRAPH_OUT_DIR, filename)) | |
| 276 | |
| 277 def add_svn_revision_to_trace(self, outfile): | |
| 278 """Add the svn version number to the provided tracefile.""" | |
| 279 def search_for_revision(svn_info_command): | |
| 280 p = subprocess.Popen(svn_info_command, stdout = subprocess.PIPE, | |
| 281 stderr = subprocess.STDOUT, shell = HAS_SHELL) | |
| 282 output, _ = p.communicate() | |
| 283 for line in output.split('\n'): | |
| 284 if 'Revision' in line: | |
| 285 run_cmd(['echo', line.strip()], outfile) | |
| 286 return True | |
| 287 return False | |
| 288 | |
| 289 if not search_for_revision(['svn', 'info']): | |
| 290 if not search_for_revision(['git', 'svn', 'info']): | |
| 291 run_cmd(['echo', 'Revision: unknown'], outfile) | |
| 292 | |
| 293 def write_html(self, delimiter, rev_nums, label_1, dict_1, label_2, dict_2, | |
| 294 cleanFile=False): | |
| 295 """Adds an html table to the webpage to display the data values. This method | |
| 296 will be removed when we have a nicer way to display data values.""" | |
| 297 #TODO(efortuna): fix this. | |
| 298 return | |
| 299 #TODO(efortuna): Take this method out when have finalized where the data is | |
| 300 # going to be displayed. | |
| 301 f = '' | |
| 302 out = '' | |
| 303 if cleanFile: | |
| 304 f = open('template.html') | |
| 305 else: | |
| 306 shutil.copy('index.html', 'temp.html') | |
| 307 f = open('temp.html') | |
| 308 out = open('index.html', 'w') | |
| 309 inTable = False | |
| 310 for line in f.readlines(): | |
| 311 if not inTable: | |
| 312 out.write(line) | |
| 313 if delimiter in line: | |
| 314 inTable = not inTable | |
| 315 if inTable: | |
| 316 out.write('<table border="1"> <tr> <td> svn revision </td>') | |
| 317 for revision in rev_nums: | |
| 318 out.write('<td>%d</td>' % revision) | |
| 319 out.write('</tr>\n<tr><td> %s</td>' % label_1) | |
| 320 for perf in dict_1: | |
| 321 out.write('<td>%f</td>' % perf) | |
| 322 out.write('</tr>\n<tr><td> %s</td>' % label_2) | |
| 323 for perf in dict_2: | |
| 324 out.write('<td>%f</td>' % perf) | |
| 325 out.write('</tr> </table>') | |
| 326 | |
| 327 def calculate_geometric_mean(self, platform, frog_or_v8, svn_revision): | |
| 328 """Calculate the aggregate geometric mean for V8 and frog benchmark sets, | |
| 329 given two benchmark dictionaries.""" | |
| 330 geo_mean = 0 | |
| 331 for benchmark in get_benchmarks(): | |
| 332 geo_mean += math.log(self.values_dict[platform][frog_or_v8][benchmark][ | |
| 333 len(self.values_dict[platform][frog_or_v8][benchmark]) - 1]) | |
| 334 | |
| 335 mean = V8_MEAN | |
| 336 if frog_or_v8 == FROG: | |
| 337 mean = FROG_MEAN | |
| 338 self.values_dict[platform][frog_or_v8][mean] += \ | |
| 339 [math.pow(math.e, geo_mean / len(get_benchmarks()))] | |
| 340 self.revision_dict[platform][frog_or_v8][mean] += [svn_revision] | |
| 341 | |
| 342 def run(self): | |
| 343 """Run the benchmarks/tests from the command line and plot the | |
| 344 results.""" | |
| 345 plt.cla() # cla = clear current axes | |
| 346 os.chdir(DART_INSTALL_LOCATION) | |
| 347 ensure_output_directory(self.result_folder_name) | |
| 348 ensure_output_directory(GRAPH_OUT_DIR) | |
| 349 self.run_tests() | |
| 350 os.chdir(os.path.join('tools', 'testing', 'perf_testing')) | |
| 351 | |
| 352 # TODO(efortuna): You will want to make this only use a subset of the files | |
| 353 # eventually. | |
| 354 files = os.listdir(self.result_folder_name) | |
| 355 | |
| 356 for afile in files: | |
| 357 if not afile.startswith('.'): | |
| 358 self.process_file(afile) | |
| 359 | |
| 360 self.plot_results('%s.png' % self.result_folder_name) | |
| 361 | |
| 362 class PerformanceTest(TestRunner): | |
| 363 """Super class for all performance testing.""" | |
| 364 def __init__(self, result_folder_name, platform_list, platform_type): | |
| 365 super(PerformanceTest, self).__init__(result_folder_name, | |
| 366 platform_list, get_versions(), get_benchmarks()) | |
| 367 self.platform_list = platform_list | |
| 368 self.platform_type = platform_type | |
| 369 | |
| 370 def plot_all_perf(self, png_filename): | |
| 371 """Create a plot that shows the performance changes of individual benchmarks | |
| 372 run by V8 and generated by frog, over svn history.""" | |
| 373 for benchmark in get_benchmarks(): | |
| 374 self.style_and_save_perf_plot( | |
| 375 'Performance of %s over time on the %s' % (benchmark, | |
| 376 self.platform_type), 'Speed (bigger = better)', 16, 14, 'lower left', | |
| 377 benchmark + png_filename, self.platform_list, get_versions(), | |
| 378 [benchmark]) | |
| 379 | |
| 380 def plot_avg_perf(self, png_filename): | |
| 381 """Generate a plot that shows the performance changes of the geomentric mean | |
| 382 of V8 and frog benchmark performance over svn history.""" | |
| 383 (title, y_axis, size_x, size_y, loc, filename) = \ | |
| 384 ('Geometric Mean of benchmark %s performance' % self.platform_type, | |
| 385 'Speed (bigger = better)', 16, 5, 'center', 'avg'+png_filename) | |
| 386 clear_axis = True | |
| 387 for platform in self.platform_list: | |
| 388 self.style_and_save_perf_plot(title, y_axis, size_x, size_y, loc, | |
| 389 filename, [platform], [V8], [V8_MEAN], clear_axis) | |
| 390 clear_axis = False | |
| 391 self.style_and_save_perf_plot(title, y_axis, size_x, size_y, loc, | |
| 392 filename, [platform], [FROG], [FROG_MEAN], clear_axis) | |
| 393 | |
| 394 def plot_results(self, png_filename): | |
| 395 self.plot_all_perf(png_filename) | |
| 396 self.plot_avg_perf('2' + png_filename) | |
| 397 | |
| 398 | |
| 399 class CommandLinePerformanceTest(PerformanceTest): | |
| 400 """Run performance tests from the command line.""" | |
| 401 | |
| 402 def __init__(self, result_folder_name): | |
| 403 super(CommandLinePerformanceTest, self).__init__(result_folder_name, | |
| 404 [COMMAND_LINE], 'command line') | |
| 405 | |
| 406 def process_file(self, afile): | |
| 407 """Pull all the relevant information out of a given tracefile. | |
| 408 | |
| 409 Args: | |
| 410 afile: The filename string we will be processing.""" | |
| 411 f = open(os.path.join(self.result_folder_name, afile)) | |
| 412 tabulate_data = False | |
| 413 revision_num = 0 | |
| 414 for line in f.readlines(): | |
| 415 if 'Revision' in line: | |
| 416 revision_num = int(line.split()[1]) | |
| 417 elif 'Benchmark' in line: | |
| 418 tabulate_data = True | |
| 419 elif tabulate_data: | |
| 420 tokens = line.split() | |
| 421 if len(tokens) < 4 or tokens[0] not in get_benchmarks(): | |
| 422 #Done tabulating data. | |
| 423 break | |
| 424 v8_value = float(tokens[1]) | |
| 425 frog_value = float(tokens[3]) | |
| 426 if v8_value == 0 or frog_value == 0: | |
| 427 #Then there was an error when this performance test was run. Do not | |
| 428 #count it in our numbers. | |
| 429 return | |
| 430 benchmark = tokens[0] | |
| 431 self.revision_dict[COMMAND_LINE][V8][benchmark] += [revision_num] | |
| 432 self.values_dict[COMMAND_LINE][V8][benchmark] += [v8_value] | |
| 433 self.revision_dict[COMMAND_LINE][FROG][benchmark] += [revision_num] | |
| 434 self.values_dict[COMMAND_LINE][FROG][benchmark] += [frog_value] | |
| 435 f.close() | |
| 436 | |
| 437 self.calculate_geometric_mean(COMMAND_LINE, FROG, revision_num) | |
| 438 self.calculate_geometric_mean(COMMAND_LINE, V8, revision_num) | |
| 439 | |
| 440 def run_tests(self): | |
| 441 """Run a performance test on our updated system.""" | |
| 442 os.chdir('frog') | |
| 443 self.trace_file = os.path.join('..', 'tools', 'testing', 'perf_testing', | |
| 444 self.result_folder_name, 'result' + self.cur_time) | |
| 445 run_cmd(['python', os.path.join('benchmarks', 'perf_tests.py')], | |
| 446 self.trace_file) | |
| 447 os.chdir('..') | |
| 448 | |
| 449 | |
| 450 class BrowserPerformanceTest(PerformanceTest): | |
| 451 """Runs performance tests, in the browser.""" | |
| 452 | |
| 453 def __init__(self, result_folder_name): | |
| 454 super(BrowserPerformanceTest, self).__init__( | |
| 455 result_folder_name, get_browsers(), 'browser') | |
| 456 | |
| 457 def run_tests(self): | |
| 458 """Run a performance test in the browser.""" | |
| 459 | |
| 460 os.chdir('frog') | |
| 461 run_cmd(['python', os.path.join('benchmarks', 'make_web_benchmarks.py')]) | |
| 462 os.chdir('..') | |
| 463 | |
| 464 for browser in get_browsers(): | |
| 465 for version in get_versions(): | |
| 466 self.trace_file = os.path.join('tools', 'testing', 'perf_testing', | |
| 467 self.result_folder_name, | |
| 468 'perf-%s-%s-%s' % (self.cur_time, browser, version)) | |
| 469 self.add_svn_revision_to_trace(self.trace_file) | |
| 470 file_path = os.path.join(os.getcwd(), 'internal', 'browserBenchmarks', | |
| 471 'benchmark_page_%s.html' % version) | |
| 472 run_cmd(['python', os.path.join('tools', 'testing', 'run_selenium.py'), | |
| 473 '--out', file_path, '--browser', browser, | |
| 474 '--timeout', '600', '--perf'], self.trace_file, append=True) | |
| 475 | |
| 476 def process_file(self, afile): | |
| 477 """Comb through the html to find the performance results.""" | |
| 478 parts = afile.split('-') | |
| 479 browser = parts[2] | |
| 480 version = parts[3] | |
| 481 f = open(os.path.join(self.result_folder_name, afile)) | |
| 482 lines = f.readlines() | |
| 483 line = '' | |
| 484 i = 0 | |
| 485 revision_num = 0 | |
| 486 while '<div id="results">' not in line and i < len(lines): | |
| 487 if 'Revision' in line: | |
| 488 revision_num = int(line.split()[1]) | |
| 489 line = lines[i] | |
| 490 i += 1 | |
| 491 | |
| 492 if i >= len(lines) or revision_num == 0: | |
| 493 # Then this run did not complete. Ignore this tracefile. | |
| 494 return | |
| 495 | |
| 496 line = lines[i] | |
| 497 i += 1 | |
| 498 results = [] | |
| 499 if line.find('<br>') > -1: | |
| 500 results = line.split('<br>') | |
| 501 else: | |
| 502 results = line.split('<br />') | |
| 503 for result in results: | |
| 504 name_and_score = result.split(':') | |
| 505 if len(name_and_score) < 2: | |
| 506 break | |
| 507 name = name_and_score[0].strip() | |
| 508 score = name_and_score[1].strip() | |
| 509 if version == V8: | |
| 510 bench_dict = self.values_dict[browser][V8] | |
| 511 else: | |
| 512 bench_dict = self.values_dict[browser][FROG] | |
| 513 bench_dict[name] += [float(score)] | |
| 514 self.revision_dict[browser][version][name] += [revision_num] | |
| 515 | |
| 516 f.close() | |
| 517 self.calculate_geometric_mean(browser, version, revision_num) | |
| 518 | |
| 519 def write_html(self, delimiter, rev_nums, label_1, dict_1, label_2, dict_2, | |
| 520 cleanFile=False): | |
| 521 #TODO(efortuna) | |
| 522 pass | |
| 523 | |
| 524 | |
| 525 class BrowserCorrectnessTest(TestRunner): | |
| 526 def __init__(self, test_type, result_folder_name): | |
| 527 super(BrowserCorrectnessTest, self).__init__(result_folder_name, | |
| 528 get_browsers(), [FROG], [CORRECTNESS]) | |
| 529 self.test_type = test_type | |
| 530 | |
| 531 def run_tests(self): | |
| 532 """Run a test of the latest svn revision.""" | |
| 533 system = get_os_directory() | |
| 534 suffix = '' | |
| 535 if platform.system() == 'Windows': | |
| 536 suffix = '.exe' | |
| 537 for browser in get_browsers(): | |
| 538 current_file = 'correctness%s-%s' % (self.cur_time, browser) | |
| 539 self.trace_file = os.path.join('tools', 'testing', | |
| 540 'perf_testing', self.result_folder_name, current_file) | |
| 541 self.add_svn_revision_to_trace(self.trace_file) | |
| 542 dart_sdk = os.path.join(os.getcwd(), utils.GetBuildRoot(utils.GuessOS(), | |
| 543 'release', 'ia32'), 'dart-sdk') | |
| 544 run_cmd([os.path.join('.', 'tools', 'testing', 'bin', system, | |
| 545 'dart' + suffix), os.path.join('tools', 'test.dart'), | |
| 546 '--component=webdriver', | |
| 547 '--browser=%s' % browser, '--frog=%s' % os.path.join(dart_sdk, 'bin', | |
| 548 'frogc'), '--froglib=%s' % os.path.join(dart_sdk, 'lib'), '--report', | |
| 549 '--timeout=20', '--progress=color', '--mode=release', '-j1', | |
| 550 self.test_type], self.trace_file, append=True) | |
| 551 | |
| 552 def process_file(self, afile): | |
| 553 """Given a trace file, extract all the relevant information out of it to | |
| 554 determine the number of correctly passing tests. | |
| 555 | |
| 556 Arguments: | |
| 557 afile: the filename string""" | |
| 558 browser = afile.rpartition('-')[2] | |
| 559 f = open(os.path.join(self.result_folder_name, afile)) | |
| 560 revision_num = 0 | |
| 561 lines = f.readlines() | |
| 562 total_tests = 0 | |
| 563 num_failed = 0 | |
| 564 expect_fail = 0 | |
| 565 for line in lines: | |
| 566 if 'Total:' in line: | |
| 567 total_tests = int(line.split('Total: ')[1].split()[0]) | |
| 568 if 'will be skipped' in line: | |
| 569 total_tests -= int(line.split()[1]) | |
| 570 if 'we should fix' in line: | |
| 571 expect_fail += int(line.split()[1]) | |
| 572 if 'Revision' in line: | |
| 573 revision_num = int(line.split()[1]) | |
| 574 if '--- TIMEOUT ---' in line or 'FAIL:' in line or 'PASS' in line: | |
| 575 # (A printed out 'PASS' indicates we incorrectly passed a negative | |
| 576 # test.) | |
| 577 num_failed += 1 | |
| 578 | |
| 579 self.revision_dict[browser][FROG][CORRECTNESS] += [revision_num] | |
| 580 self.values_dict[browser][FROG][CORRECTNESS] += [100.0 * | |
| 581 (((float)(total_tests - (expect_fail + num_failed))) /total_tests)] | |
| 582 f.close() | |
| 583 | |
| 584 def plot_results(self, png_filename): | |
| 585 first_time = True | |
| 586 for browser in get_browsers(): | |
| 587 self.style_and_save_perf_plot('Percentage of language tests passing in ' | |
| 588 'different browsers', '% of tests passed', 8, 8, 'lower left', | |
| 589 png_filename, [browser], [FROG], [CORRECTNESS], first_time) | |
| 590 first_time = False | |
| 591 | |
| 592 | |
| 593 class CompileTimeAndSizeTest(TestRunner): | |
| 594 """Run tests to determine how long minfrog takes to compile, and the compiled | |
| 595 file output size of some benchmarking files.""" | |
| 596 def __init__(self, result_folder_name): | |
| 597 super(CompileTimeAndSizeTest, self).__init__(result_folder_name, | |
| 598 [COMMAND_LINE], [FROG], ['Compiling on Dart VM', 'Bootstrapping', | |
| 599 'minfrog', 'swarm', 'total']) | |
| 600 self.failure_threshold = {'Compiling on Dart VM' : 1, 'Bootstrapping' : .5, | |
| 601 'minfrog' : 100, 'swarm' : 100, 'total' : 100} | |
| 602 | |
| 603 def run_tests(self): | |
| 604 os.chdir('frog') | |
| 605 self.trace_file = os.path.join('..', 'tools', 'testing', 'perf_testing', | |
| 606 self.result_folder_name, self.result_folder_name + self.cur_time) | |
| 607 | |
| 608 self.add_svn_revision_to_trace(self.trace_file) | |
| 609 | |
| 610 suffix = '' | |
| 611 if platform.system() == 'Windows': | |
| 612 suffix = '.exe' | |
| 613 elapsed = time_cmd([os.path.join('..', utils.GetBuildRoot(utils.GuessOS(), | |
| 614 'release', 'ia32'), 'dart' + suffix), os.path.join('.', 'minfrogc.dart'), | |
| 615 '--out=minfrog', 'minfrog.dart']) | |
| 616 run_cmd(['echo', '%f Compiling on Dart VM in production mode in seconds' | |
| 617 % elapsed], self.trace_file, append=True) | |
| 618 elapsed = time_cmd([os.path.join('.', 'minfrog'), '--out=minfrog', | |
| 619 'minfrog.dart', os.path.join('tests', 'hello.dart')]) | |
| 620 if elapsed < self.failure_threshold['Bootstrapping']: | |
| 621 #minfrog didn't compile correctly. Stop testing now, because subsequent | |
| 622 #numbers will be meaningless. | |
| 623 return | |
| 624 size = os.path.getsize('minfrog') | |
| 625 run_cmd(['echo', '%f Bootstrapping time in seconds in production mode' % | |
| 626 elapsed], self.trace_file, append=True) | |
| 627 run_cmd(['echo', '%d Generated checked minfrog size' % size], | |
| 628 self.trace_file, append=True) | |
| 629 | |
| 630 run_cmd([os.path.join('.', 'minfrog'), '--out=swarm-result', | |
| 631 '--compile-only', os.path.join('..', 'samples', 'swarm', | |
| 632 'swarm.dart')]) | |
| 633 swarm_size = 0 | |
| 634 try: | |
| 635 swarm_size = os.path.getsize('swarm-result') | |
| 636 except OSError: | |
| 637 pass #If compilation failed, continue on running other tests. | |
| 638 | |
| 639 run_cmd([os.path.join('.', 'minfrog'), '--out=total-result', | |
| 640 '--compile-only', os.path.join('..', 'samples', 'total', | |
| 641 'client', 'Total.dart')]) | |
| 642 total_size = 0 | |
| 643 try: | |
| 644 total_size = os.path.getsize('total-result') | |
| 645 except OSError: | |
| 646 pass #If compilation failed, continue on running other tests. | |
| 647 | |
| 648 run_cmd(['echo', '%d Generated checked swarm size' % swarm_size], | |
| 649 self.trace_file, append=True) | |
| 650 | |
| 651 run_cmd(['echo', '%d Generated checked total size' % total_size], | |
| 652 self.trace_file, append=True) | |
| 653 os.chdir('..') | |
| 654 | |
| 655 def process_file(self, afile): | |
| 656 """Pull all the relevant information out of a given tracefile. | |
| 657 | |
| 658 Args: | |
| 659 afile: is the filename string we will be processing.""" | |
| 660 f = open(os.path.join(self.result_folder_name, afile)) | |
| 661 tabulate_data = False | |
| 662 revision_num = 0 | |
| 663 for line in f.readlines(): | |
| 664 tokens = line.split() | |
| 665 if 'Revision' in line: | |
| 666 revision_num = int(line.split()[1]) | |
| 667 else: | |
| 668 for metric in self.values_list: | |
| 669 if metric in line: | |
| 670 num = tokens[0] | |
| 671 if num.find('.') == -1: | |
| 672 num = int(num) | |
| 673 else: | |
| 674 num = float(num) | |
| 675 self.values_dict[COMMAND_LINE][FROG][metric] += [num] | |
| 676 self.revision_dict[COMMAND_LINE][FROG][metric] += [revision_num] | |
| 677 | |
| 678 if revision_num != 0: | |
| 679 for metric in self.values_list: | |
| 680 self.revision_dict[COMMAND_LINE][FROG][metric].pop() | |
| 681 self.revision_dict[COMMAND_LINE][FROG][metric] += [revision_num] | |
| 682 # Fill in 0 if compilation failed. | |
| 683 if self.values_dict[COMMAND_LINE][FROG][metric][-1] < \ | |
| 684 self.failure_threshold[metric]: | |
| 685 self.values_dict[COMMAND_LINE][FROG][metric] += [0] | |
| 686 self.revision_dict[COMMAND_LINE][FROG][metric] += [revision_num] | |
| 687 | |
| 688 f.close() | |
| 689 | |
| 690 def plot_results(self, png_filename): | |
| 691 self.style_and_save_perf_plot('Compiled minfrog Sizes', | |
| 692 'Size (in bytes)', 10, 10, 'center', png_filename, [COMMAND_LINE], | |
| 693 [FROG], ['swarm', 'total', 'minfrog']) | |
| 694 self.write_html('bar', self.revision_dict[COMMAND_LINE][FROG]['minfrog'], | |
| 695 'minfrog size', self.values_dict[COMMAND_LINE][FROG]['minfrog'], '', []) | |
| 696 | |
| 697 self.style_and_save_perf_plot('Time to compile and bootstrap', | |
| 698 'Seconds', 10, 10, 'center', '2' + png_filename, [COMMAND_LINE], [FROG], | |
| 699 ['Bootstrapping', 'Compiling on Dart VM']) | |
| 700 self.write_html('baz', | |
| 701 self.revision_dict[COMMAND_LINE][FROG]['Bootstrapping'], | |
| 702 'Bootstrapping', self.values_dict[COMMAND_LINE][FROG]['Bootstrapping'], | |
| 703 'Compiling on Dart VM', | |
| 704 self.values_dict[COMMAND_LINE][FROG]['Compiling on Dart VM']) | |
| 705 | |
| 706 | |
| 707 def parse_args(): | |
| 708 parser = optparse.OptionParser() | |
| 709 parser.add_option('--command-line', '-c', dest='cl', | |
| 710 help = 'Run the command line tests', | |
| 711 action = 'store_true', default = False) | |
| 712 parser.add_option('--size-time', '-s', dest = 'size', | |
| 713 help = 'Run the code size and timing tests', | |
| 714 action = 'store_true', default = False) | |
| 715 parser.add_option('--language', '-l', dest = 'language', | |
| 716 help = 'Run the language correctness tests', | |
| 717 action = 'store_true', default = False) | |
| 718 parser.add_option('--browser-perf', '-b', dest = 'perf', | |
| 719 help = 'Run the browser performance tests', | |
| 720 action = 'store_true', default = False) | |
| 721 parser.add_option('--forever', '-f', dest = 'continuous', | |
| 722 help = 'Run this script forever, always checking for the next svn ' | |
| 723 'checkin', action = 'store_true', default = False) | |
| 724 parser.add_option('--verbose', '-v', dest = 'verbose', | |
| 725 help = 'Print extra debug output', action = 'store_true', default = False) | |
| 726 | |
| 727 args, ignored = parser.parse_args() | |
| 728 if not (args.cl or args.size or args.language or args.perf): | |
| 729 args.cl = args.size = args.language = args.perf = True | |
| 730 return (args.cl, args.size, args.language, args.perf, args.continuous, | |
| 731 args.verbose) | |
| 732 | |
| 733 def run_test_sequence(cl, size, language, perf): | |
| 734 # The buildbot already builds and syncs to a specific revision. Don't fight | |
| 735 # with it or replicate work. | |
| 736 if sync_and_build() == 1: | |
| 737 return # The build is broken. | |
| 738 if size: | |
| 739 CompileTimeAndSizeTest(TIME_SIZE).run() | |
| 740 if cl: | |
| 741 CommandLinePerformanceTest(CL_PERF).run() | |
| 742 if language: | |
| 743 BrowserCorrectnessTest('language', BROWSER_CORRECTNESS).run() | |
| 744 if perf: | |
| 745 BrowserPerformanceTest(BROWSER_PERF).run() | |
| 746 | |
| 747 # TODO(efortuna): Temporarily disabled until you make a safe way to provide | |
| 748 # your username/password for the uploading process. | |
| 749 #upload_to_app_engine() | |
| 750 | |
| 751 def main(): | |
| 752 global VERBOSE | |
| 753 (cl, size, language, perf, continuous, verbose) = parse_args() | |
| 754 VERBOSE = verbose | |
| 755 if continuous: | |
| 756 while True: | |
| 757 if has_new_code(): | |
| 758 run_test_sequence(cl, size, language, perf) | |
| 759 else: | |
| 760 time.sleep(SLEEP_TIME) | |
| 761 else: | |
| 762 run_test_sequence(cl, size, language, perf) | |
| 763 | |
| 764 if __name__ == '__main__': | |
| 765 main() | |
| OLD | NEW |
|---|
Chromium Code Reviews
Issue 9706088:
Some changes to performance test script. (Closed)
Base URL: http://dart.googlecode.com/svn/branches/bleeding_edge/dart/