graalpython/mx.graalpython/mx_graalpython_bisect.py at master · oracle/graalpython · GitHub

History

437 lines (385 loc) · 17.9 KB

Raw

100

101

102

103

104

105

106

107

108

109

110

111

112

113

114

115

116

117

118

119

120

121

122

123

124

125

126

127

128

129

130

131

132

133

134

135

136

137

138

139

140

141

142

143

144

145

146

147

148

149

150

151

152

153

154

155

156

157

158

159

160

161

162

163

164

165

166

167

168

169

170

171

172

173

174

175

176

177

178

179

180

181

182

183

184

185

186

187

188

189

190

191

192

193

194

195

196

197

198

199

200

201

202

203

204

205

206

207

208

209

210

211

212

213

214

215

216

217

218

219

220

221

222

223

224

225

226

227

228

229

230

231

232

233

234

235

236

237

238

239

240

241

242

243

244

245

246

247

248

249

250

251

252

253

254

255

256

257

258

259

260

261

262

263

264

265

266

267

268

269

270

271

272

273

274

275

276

277

278

279

280

281

282

283

284

285

286

287

288

289

290

291

292

293

294

295

296

297

298

299

300

301

302

303

304

305

306

307

308

309

310

311

312

313

314

315

316

317

318

319

320

321

322

323

324

325

326

327

328

329

330

331

332

333

334

335

336

337

338

339

340

341

342

343

344

345

346

347

348

349

350

351

352

353

354

355

356

357

358

359

360

361

362

363

364

365

366

367

368

369

370

371

372

373

374

375

376

377

378

379

380

381

382

383

384

385

386

387

388

389

390

391

392

393

394

395

396

397

398

399

400

401

402

403

404

405

406

407

408

409

410

411

412

413

414

415

416

417

418

419

420

421

422

423

424

425

426

427

428

429

430

431

432

433

434

435

436

437

# DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.

# The Universal Permissive License (UPL), Version 1.0

# Subject to the condition set forth below, permission is hereby granted to any

# person obtaining a copy of this software, associated documentation and/or

# data (collectively the "Software"), free of charge and under any and all

# copyright rights in the Software, and any and all patent rights owned or

# freely licensable by each licensor hereunder covering either (i) the

# unmodified Software as contributed to or provided by such licensor, or (ii)

# the Larger Works (as defined below), to deal in both

# (a) the Software, and

# (b) any piece of software and/or hardware listed in the lrgrwrks.txt file if

# one is included with the Software each a "Larger Work" to which the Software

# is contributed by such licensors),

# without restriction, including without limitation the rights to copy, create

# derivative works of, display, perform, and distribute the Software and make,

# use, sell, offer for sale, import, export, have made, and have sold the

# Software and the Larger Work(s), and to sublicense the foregoing rights on

# either these or other terms.

# This license is subject to the following condition:

# The above copyright notice and either this complete permission notice or at a

# minimum a reference to the UPL must be included in all copies or substantial

# portions of the Software.

# THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR

# IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,

# FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE

# AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER

# LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,

# OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE

# SOFTWARE.

import abc

import argparse

import json

import os

import shlex

import types

from pathlib import Path

import mx

def print_line(l):

print('=' * l)

SUITE = mx.suite('graalpython')

GIT = SUITE.vc

DIR = Path(SUITE.vc_dir).absolute()

GRAAL_DIR = DIR.parent / 'graal'

VM_DIR = GRAAL_DIR / 'vm'

GRAAL_ENTERPRISE_DIR = DIR.parent / 'graal-enterprise'

VM_ENTERPRISE_DIR = GRAAL_ENTERPRISE_DIR / 'vm-enterprise'

SUITE_MAPPING = {

GRAAL_DIR: VM_DIR,

GRAAL_ENTERPRISE_DIR: VM_ENTERPRISE_DIR,

}

DOWNSTREAM_REPO_MAPPING = {

DIR: GRAAL_DIR,

GRAAL_DIR: GRAAL_ENTERPRISE_DIR,

}

RESULTS_JSON_PATH = 'bisect-benchmark-result.json'

def get_commit(repo_path, ref='HEAD'):

if repo_path:

return GIT.git_command(repo_path, ['rev-parse', ref], abortOnError=True).strip()

def get_message(repo_path: Path, commit):

return GIT.git_command(repo_path, ['log', '--format=%s', '-n', '1', commit]).strip()

def run_bisect_benchmark(repo_path: Path, bad, good, callback, good_result=None, bad_result=None):

commits = GIT.git_command(

repo_path,

['log', '--first-parent', '--format=format:%H', '{}^..{}'.format(good, bad)],

abortOnError=True,

).splitlines()

if not commits:

raise RuntimeError("No merge commits found in the range. Did you swap good and bad?")

downstream_repo_path = DOWNSTREAM_REPO_MAPPING.get(repo_path)

results = [None] * len(commits)

if good_result is None and bad_result is None:

bad_index = 0

good_index = len(commits) - 1

bad_result = results[bad_index] = callback(repo_path, bad)

downstream_bad = get_commit(downstream_repo_path)

good_result = results[good_index] = callback(repo_path, good)

downstream_good = get_commit(downstream_repo_path)

if not good_result.bound_is_valid(bad_result):

raise RuntimeError(

"Didn't detect a regression: "

f"'good' value ({good_result}) is worse than or same as than 'bad' value ({bad_result})"

)

if not good_result.bound_is_significant(bad_result, 0.03):

raise RuntimeError(

"Didn't detect a regression: "

f"less that 3% difference between 'good' value ({good_result}) and 'bad' value ({bad_result})"

)

else:

bad_index = -1

good_index = len(commits)

downstream_bad = None

downstream_good = None

while True:

index = bad_index + ((good_index - bad_index) // 2)

if index in [bad_index, good_index]:

assert good_index - bad_index == 1

break

commit = commits[index]

result = results[index] = callback(repo_path, commit)

if result.is_good(good_result, bad_result):

good_index = index

downstream_good = get_commit(downstream_repo_path)

else:

bad_index = index

downstream_bad = get_commit(downstream_repo_path)

subresults = {}

if downstream_bad and downstream_good and downstream_bad != downstream_good:

GIT.update_to_branch(repo_path, commits[good_index])

subresult = run_bisect_benchmark(downstream_repo_path, downstream_bad, downstream_good, callback, good_result,

bad_result)

subresults[bad_index] = subresult

return BisectResult(repo_path, commits, results, good_index, bad_index, subresults)

class BisectResult:

def __init__(self, repo_path: Path, commits, results, good_index, bad_index, dependency_results):

self.repo_path = repo_path

self.commits = commits

self.results = results

self.good_index = good_index

self.bad_index = bad_index

self.dependency_results = dependency_results

@property

def repo_name(self):

return self.repo_path.name

@property

def good_commit(self):

if 0 <= self.good_index < len(self.commits):

return self.commits[self.good_index]

@property

def bad_commit(self):

if 0 <= self.bad_index < len(self.commits):

return self.commits[self.bad_index]

def visualize(self, level=1):

level_marker = '=' * level

out = ["{} {}".format(level_marker, self.repo_name)]

for index, (commit, value) in enumerate(zip(self.commits, self.results)):

if value is not None:

out.append(f"{level_marker} {commit} {value} {get_message(self.repo_path, commit)}")

if self.dependency_results and index in self.dependency_results:

out.append(self.dependency_results[index].visualize(level + 1))

return '\n'.join(out)

def summarize(self):

if self.bad_commit and self.good_commit:

for dependency_result in self.dependency_results.values():

summary = dependency_result.summarize()

if summary:

return summary

return ("Detected bad commit in {} repository:\n{} {}"

.format(self.repo_name, self.bad_commit, get_message(self.repo_path, self.bad_commit)))

return ''

def to_dict(self):

return {

'repo_name': self.repo_name,

'repo_path': str(self.repo_path),

'commits': self.commits,

'results': [result.to_dict() if result is not None else None for result in self.results],

'good_index': self.good_index,

'bad_index': self.bad_index,

'good_commit': self.good_commit,

'bad_commit': self.bad_commit,

'dependency_results': {

str(index): dependency_result.to_dict()

for index, dependency_result in self.dependency_results.items()

}

class BenchmarkResult(abc.ABC):

def __init__(self, value, unit=None):

self.value = value

self.unit = unit

def __str__(self):

return f'{self.value:6.6} {self.unit}'

def bound_is_valid(self, bad_result):

return self.is_good(self, bad_result) or not bad_result.is_good(self, bad_result)

def bound_is_significant(self, bad_result, epsilon):

avg = (self.value + bad_result.value) / 2

if not avg:

return False

diff = abs(self.value - bad_result.value)

return diff / avg >= epsilon

@abc.abstractmethod

def is_good(self, good_result, bad_result):

pass

def to_dict(self):

return {

'kind': type(self).__name__,

'value': self.value,

'unit': self.unit,

'display': str(self),

}

class LowerIsBetterResult(BenchmarkResult):

def is_good(self, good_result, bad_result):

threshold = (good_result.value + bad_result.value) / 2

return self.value < threshold

class HigherIsBetterResult(BenchmarkResult):

def is_good(self, good_result, bad_result):

threshold = (good_result.value + bad_result.value) / 2

return self.value > threshold

class WorksResult(BenchmarkResult):

def is_good(self, good_result, bad_result):

return self.value == 0

def bound_is_significant(self, bad_result, epsilon):

return True

def __str__(self):

return "works" if self.value == 0 else "doesn't work"

def filter_benchmark_results(docs, benchmark_name):

exact_docs = [x for x in docs if x['benchmark'] == benchmark_name]

if exact_docs:

return exact_docs

return [x for x in docs if x['benchmark'].endswith(f'.{benchmark_name}')]

def _bisect_benchmark(argv, bisect_id):

default_metric = 'time'

if 'BISECT_BENCHMARK_CONFIG' in os.environ:

import configparser

cp = configparser.ConfigParser()

cp.read(os.environ['BISECT_BENCHMARK_CONFIG'])

sec = cp['bisect-benchmark']

args = types.SimpleNamespace()

args.bad = sec['bad']

args.good = sec['good']

args.build_command = sec['build_command']

args.benchmark_command = sec['benchmark_command']

args.benchmark_metric = sec.get('benchmark_metric', default_metric)

args.benchmark_name = sec.get('benchmark_name', None)

args.enterprise = sec.getboolean('enterprise', False)

args.no_clean = sec.getboolean('no_clean', False)

args.rerun_with_commands = sec.get('rerun_with_commands')

else:

parser = argparse.ArgumentParser()

parser.add_argument('bad', help="Bad commit for bisection")

parser.add_argument('good', help="Good commit for bisection")

parser.add_argument('build_command', help="Command to run in order to build the configuration")

parser.add_argument('benchmark_command',

help="Command to run in order to run the benchmark. Output needs to be in mx's format")

parser.add_argument('benchmark_name',

help="Filters the results to choose only benchmarks of given name. "

"Useful if the benchmark command runs multiple benchmarks.")

parser.add_argument('--rerun-with-commands',

help="Re-run the bad and good commits with this benchmark command(s) "

"(multiple commands separated by ';')")

parser.add_argument(

'--benchmark-metric', default=default_metric,

help=(

"Which result metric should be used for comparisons (metric.name in the result json). "

"A special value 'WORKS' can be used to consider only the success of the benchmark command."

)

parser.add_argument('--enterprise', action='store_true', help="Whether to checkout graal-enterprise")

parser.add_argument('--no-clean', action='store_true', help="Do not run 'mx clean' between runs")

args = parser.parse_args(argv)

def checkout(repo_path: Path, commit):

if 'CI' in os.environ:

# discard any changes made by the previous build before checkout

GIT.run(['git', 'reset', '--hard', 'HEAD'], nonZeroIsFatal=True, cwd=repo_path)

GIT.update_to_branch(repo_path, commit)

if repo_path == DIR:

mx.run_mx(['sforceimports'], suite=str(DIR))

if args.enterprise:

# Keep the Graal revision selected by the current bisection point. The

# enterprise suite imports Graal, so its sforceimports would otherwise

# replace that revision with the one recorded in graal-enterprise.

graal_commit = get_commit(GRAAL_DIR)

if repo_path.name != 'graal-enterprise':

mx.run_mx(['--quiet', 'checkout-downstream', 'vm', 'vm-enterprise', '--no-fetch'],

suite=str(VM_ENTERPRISE_DIR))

mx.run_mx(['--dy', 'substratevm-enterprise-gcs', 'sforceimports'], suite=str(VM_ENTERPRISE_DIR))

GIT.update_to_branch(GRAAL_DIR, graal_commit)

debug_str = f"debug: {SUITE.name}={get_commit(SUITE.vc_dir)} graal={get_commit(GRAAL_DIR)}"

if args.enterprise:

debug_str += f" graal-enterprise={get_commit(GRAAL_ENTERPRISE_DIR)}"

print(debug_str)

def fetch_jdk():

import mx_fetchjdk

if args.enterprise:

fetch_args = [

'--configuration', str(GRAAL_ENTERPRISE_DIR / 'common.json'),

'--jdk-binaries', str(GRAAL_ENTERPRISE_DIR / 'ci' / 'jdk-binaries.json'),

'labsjdk-ee-latest',

]

else:

fetch_args = [

'--configuration', str(GRAAL_DIR / 'common.json'),

'labsjdk-ce-latest',

]

# Awkward way to suppress the confirmation prompt

ci = 'CI' in os.environ

if not ci:

os.environ['CI'] = '1'

try:

return mx_fetchjdk.fetch_jdk(fetch_args)

finally:

if not ci:

del os.environ['CI']

def checkout_and_build(repo_path, commit):

checkout(repo_path, commit)

os.environ['JAVA_HOME'] = fetch_jdk()

build_command = shlex.split(args.build_command)

if not args.no_clean:

try:

clean_command = build_command[:build_command.index('build')] + ['clean', '--all', '--aggressive']

retcode = mx.run(clean_command, nonZeroIsFatal=False)

if retcode:

print("Warning: clean command failed")

except ValueError:

pass

retcode = mx.run(build_command, nonZeroIsFatal=False)

if retcode:

raise RuntimeError("Failed to execute the build command for {}".format(commit))

def benchmark_callback(repo_path: Path, commit, bench_command=args.benchmark_command):

checkout_and_build(repo_path, commit)

retcode = mx.run(shlex.split(bench_command), nonZeroIsFatal=False)

if args.benchmark_metric == 'WORKS':

return WorksResult(retcode)

if retcode:

raise RuntimeError("Failed to execute benchmark for {}".format(commit))

with open('bench-results.json') as f:

data = json.load(f)

docs = [x for x in data['queries'] if x['metric.name'] == args.benchmark_metric]

if args.benchmark_name:

docs = filter_benchmark_results(docs, args.benchmark_name)

if not docs:

raise RuntimeError(f"Couldn't find specified metric {args.benchmark_metric!r} in the results")

if len(docs) > 1:

print("WARNING: found multiple results for the metric, picking the last")

names = set([x['benchmark'] for x in docs])

if len(names) != 1:

print(f"WARNING: found multiple results with different benchmark name attributes: {names}. "

"Use benchmark_name option to filter specific benchmark name.")

doc = docs[-1]

result_class = HigherIsBetterResult if doc.get('metric.better', 'lower') == 'higher' else LowerIsBetterResult

return result_class(doc['metric.value'], doc['metric.unit'])

bad = get_commit(DIR, args.bad)

good = get_commit(DIR, args.good)

result = run_bisect_benchmark(DIR, bad, good, benchmark_callback)

visualization = result.visualize()

summary = result.summarize()

print()

print(visualization)

print()

print(summary)

with open(RESULTS_JSON_PATH, 'w', encoding='utf-8') as result_file:

json.dump({

'bisect_id': bisect_id,

'summary': summary,

'visualization': visualization,

'result': result.to_dict(),

'build_url': os.environ.get('BUILD_URL'),

}, result_file, indent=2, sort_keys=True)

result_file.write('\n')

if args.rerun_with_commands:

print('\n\nRerunning the good and bad commits with extra benchmark commands:')

repo_path = DIR

current_result = result

while current_result.subresults and current_result.bad_index in current_result.subresults:

downstream_repo_path = DOWNSTREAM_REPO_MAPPING.get(repo_path)

next_result = current_result.subresults[current_result.bad_index]

if not next_result.good_commit or not next_result.bad_commit:

print(f"Next downstream repo {downstream_repo_path.name} does not have both good and bad commits")

break

print(f"Recursing to downstream repo: {downstream_repo_path.name}, commit: {current_result.bad_commit}")

checkout(downstream_repo_path, current_result.bad_commit)

current_result = next_result

repo_path = downstream_repo_path

for commit in [current_result.good_commit, current_result.bad_commit]:

print_line(80)

print("Commit: {}".format(commit))

checkout_and_build(repo_path, commit)

for cmd in args.rerun_with_commands.split(";"):

print_line(40)

mx.run(shlex.split(cmd.strip()), nonZeroIsFatal=False)

def bisect_benchmark(argv):

initial_branch = GIT.git_command(DIR, ['rev-parse', '--abbrev-ref', 'HEAD']).strip()

initial_commit = GIT.git_command(DIR, ['log', '--format=%s', '-n', '1']).strip()

bisect_id = f'{initial_branch}: {initial_commit}'

try:

_bisect_benchmark(argv, bisect_id)

finally:

GIT.update_to_branch(DIR, initial_branch)

graalpython/mx.graalpython/mx_graalpython_bisect.py at master · oracle/graalpython · GitHub

Uh oh!

Expand file tree

Breadcrumbs

mx_graalpython_bisect.py

Latest commit

History

Breadcrumbs

mx_graalpython_bisect.py

File metadata and controls