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

History

1642 lines (1418 loc) · 71.6 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

438

439

440

441

442

443

444

445

446

447

448

449

450

451

452

453

454

455

456

457

458

459

460

461

462

463

464

465

466

467

468

469

470

471

472

473

474

475

476

477

478

479

480

481

482

483

484

485

486

487

488

489

490

491

492

493

494

495

496

497

498

499

500

501

502

503

504

505

506

507

508

509

510

511

512

513

514

515

516

517

518

519

520

521

522

523

524

525

526

527

528

529

530

531

532

533

534

535

536

537

538

539

540

541

542

543

544

545

546

547

548

549

550

551

552

553

554

555

556

557

558

559

560

561

562

563

564

565

566

567

568

569

570

571

572

573

574

575

576

577

578

579

580

581

582

583

584

585

586

587

588

589

590

591

592

593

594

595

596

597

598

599

600

601

602

603

604

605

606

607

608

609

610

611

612

613

614

615

616

617

618

619

620

621

622

623

624

625

626

627

628

629

630

631

632

633

634

635

636

637

638

639

640

641

642

643

644

645

646

647

648

649

650

651

652

653

654

655

656

657

658

659

660

661

662

663

664

665

666

667

668

669

670

671

672

673

674

675

676

677

678

679

680

681

682

683

684

685

686

687

688

689

690

691

692

693

694

695

696

697

698

699

700

701

702

703

704

705

706

707

708

709

710

711

712

713

714

715

716

717

718

719

720

721

722

723

724

725

726

727

728

729

730

731

732

733

734

735

736

737

738

739

740

741

742

743

744

745

746

747

748

749

750

751

752

753

754

755

756

757

758

759

760

761

762

763

764

765

766

767

768

769

770

771

772

773

774

775

776

777

778

779

780

781

782

783

784

785

786

787

788

789

790

791

792

793

794

795

796

797

798

799

800

801

802

803

804

805

806

807

808

809

810

811

812

813

814

815

816

817

818

819

820

821

822

823

824

825

826

827

828

829

830

831

832

833

834

835

836

837

838

839

840

841

842

843

844

845

846

847

848

849

850

851

852

853

854

855

856

857

858

859

860

861

862

863

864

865

866

867

868

869

870

871

872

873

874

875

876

877

878

879

880

881

882

883

884

885

886

887

888

889

890

891

892

893

894

895

896

897

898

899

900

901

902

903

904

905

906

907

908

909

910

911

912

913

914

915

916

917

918

919

920

921

922

923

924

925

926

927

928

929

930

931

932

933

934

935

936

937

938

939

940

941

942

943

944

945

946

947

948

949

950

951

952

953

954

955

956

957

958

959

960

961

962

963

964

965

966

967

968

969

970

971

972

973

974

975

976

977

978

979

980

981

982

983

984

985

986

987

988

989

990

991

992

993

994

995

996

997

998

999

1000

# Redistribution and use in source and binary forms, with or without modification, are

# permitted provided that the following conditions are met:

# 1. Redistributions of source code must retain the above copyright notice, this list of

# conditions and the following disclaimer.

# 2. Redistributions in binary form must reproduce the above copyright notice, this list of

# conditions and the following disclaimer in the documentation and/or other materials provided

# with the distribution.

# THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND ANY EXPRESS

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

# MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE

# COPYRIGHT HOLDER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL,

# EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE

# GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED

# AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING

# NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED

# OF THE POSSIBILITY OF SUCH DAMAGE.

from __future__ import print_function

import functools

import shutil

import statistics

import sys

import os

import re

import shlex

import subprocess

from abc import ABC, abstractproperty

from contextlib import contextmanager

from os.path import join

from datetime import datetime

from pathlib import Path

import mx

import mx_benchmark

import mx_polybench

import mx_sdk_benchmark

from mx_benchmark import StdOutRule, java_vm_registry, OutputCapturingVm, GuestVm, VmBenchmarkSuite, AveragingBenchmarkMixin, bm_exec_context

from mx_graalpython_bench_param import HARNESS_PATH

from mx_util import StageName

# ----------------------------------------------------------------------------------------------------------------------

# the graalpython suite

# ----------------------------------------------------------------------------------------------------------------------

SUITE = mx.suite("graalpython")

DIR = Path(__file__).parent.resolve()

# ----------------------------------------------------------------------------------------------------------------------

# constants

# ----------------------------------------------------------------------------------------------------------------------

VM_NAME_GRAALPYTHON = "graalpython"

VM_NAME_GRAALPYTHON_SVM = "graalpython-svm"

GROUP_GRAAL = "Graal"

SUBGROUP_GRAAL_PYTHON = "graalpython"

PYTHON_VM_REGISTRY_NAME = "Python"

CONFIGURATION_DEFAULT = "default"

CONFIGURATION_CUSTOM = "custom"

CONFIGURATION_INTERPRETER = "interpreter"

CONFIGURATION_NATIVE_INTERPRETER = "native-interpreter"

CONFIGURATION_DEFAULT_MULTI = "default-multi"

CONFIGURATION_INTERPRETER_MULTI = "interpreter-multi"

CONFIGURATION_NATIVE_INTERPRETER_MULTI = "native-interpreter-multi"

CONFIGURATION_NATIVE = "native"

CONFIGURATION_NATIVE_JIT2 = "native-jit2"

CONFIGURATION_UNCACHED = "interpreter-uncached"

CONFIGURATION_NATIVE_MULTI = "native-multi"

CONFIGURATION_SANDBOXED = "sandboxed"

CONFIGURATION_SANDBOXED_MULTI = "sandboxed-multi"

PYTHON_JAVA_EMBEDDING_VM_REGISTRY_NAME = "PythonJavaDriver"

CONFIGURATION_JAVA_EMBEDDING_MULTI = "java-driver-multi-default"

CONFIGURATION_JAVA_EMBEDDING_MULTI_SHARED = "java-driver-multi-shared"

CONFIGURATION_JAVA_EMBEDDING_INTERPRETER_MULTI = "java-driver-interpreter-multi"

CONFIGURATION_JAVA_EMBEDDING_INTERPRETER_MULTI_SHARED = "java-driver-interpreter-multi-shared"

DEFAULT_ITERATIONS = 10

BENCH_BGV = 'benchmarks-bgv'

BENCHMARK_DEBUG_ARGS = (

# These first two are not /too/ bad for runtime

# '--python.PoisonNativeMemoryOnFree=true',

# '--python.SampleNativeMemoryAllocSites=true',

# These below can be *extremely* heavy

# '--python.TraceNativeMemory=true',

# '--python.TraceNativeMemoryCalls=true',

# '--log.python.level=FINER',

)

# ----------------------------------------------------------------------------------------------------------------------

# utils

# ----------------------------------------------------------------------------------------------------------------------

def is_sandboxed_configuration(conf):

return conf in (CONFIGURATION_SANDBOXED, CONFIGURATION_SANDBOXED_MULTI)

@contextmanager

def environ(env):

def _handle_var(key_value):

(k, v) = key_value

if v is None:

del os.environ[k]

else:

os.environ[k] = str(v)

if env:

prev_env = {v: os.getenv(v) for v in env}

list(map(_handle_var, list(env.items())))

else:

prev_env = None

try:

yield

finally:

if prev_env:

list(map(_handle_var, list(prev_env.items())))

# ----------------------------------------------------------------------------------------------------------------------

# the vm definitions

# ----------------------------------------------------------------------------------------------------------------------

class AbstractPythonVm(OutputCapturingVm, ABC):

def __init__(self, name, config_name, options=None):

super().__init__()

self._name = name

self._config_name = config_name

self._options = options

@property

def options(self):

return self._options

def name(self):

return self._name

def config_name(self):

return self._config_name

@abstractproperty

def interpreter(self):

return None

def post_process_command_line_args(self, args):

return args

def run_vm(self, args, out=None, err=None, cwd=None, nonZeroIsFatal=False, env=None):

cmd = [self.interpreter] + args

cmd = mx.apply_command_mapper_hooks(cmd, self.command_mapper_hooks)

mx.logv(shlex.join(cmd))

return mx.run(

cmd,

out=out,

err=err,

cwd=cwd,

nonZeroIsFatal=nonZeroIsFatal,

env=env,

)

class AbstractPythonIterationsControlVm(AbstractPythonVm):

def __init__(self, name, config_name, options=None, iterations=None):

super().__init__(name, config_name, options=options)

try:

self._iterations = int(iterations)

except:

self._iterations = None

def override_iterations(self, requested_iterations):

return self._iterations if self._iterations is not None else requested_iterations

def _override_iterations_args(self, args):

_args = []

i = 0

while i < len(args):

arg = args[i]

_args.append(arg)

if arg == '-i':

_args.append(str(self.override_iterations(int(args[i + 1]))))

i += 1

return _args

def run_vm(self, args, *splat, **kwargs):

args = self._override_iterations_args(args)

return super().run_vm(args, *splat, **kwargs)

class CPythonVm(AbstractPythonIterationsControlVm):

def __init__(self, config_name, options=None, virtualenv=None, iterations=0):

super().__init__("cpython", config_name, options=options, iterations=iterations)

self._virtualenv = virtualenv

def override_iterations(self, requested_iterations):

# CPython has no JIT right now, just a quickening interpreter

return min(requested_iterations, 3)

@property

def interpreter(self):

if venv := self._virtualenv:

path = os.path.join(venv, 'bin', 'python')

mx.log(f"Using CPython from virtualenv: {path}")

elif python3_home := mx.get_env('PYTHON3_HOME'):

path = os.path.join(python3_home, 'python')

mx.log(f"Using CPython from PYTHON3_HOME: {path}")

elif path := shutil.which('python'):

mx.log(f"Using CPython from PATH: {path}")

else:

assert sys.implementation.name == 'cpython', "Cannot find CPython"

path = sys.executable

mx.log(f"Using CPython from sys.executable: {path}")

return path

def run_vm(self, args, *splat, **kwargs):

for idx, arg in enumerate(args):

if "--vm.Xmx" in arg:

mx.warn(f"Ignoring {arg}, cannot restrict memory on CPython.")

args = args[:idx] + args[idx + 1 :]

break

return super().run_vm(args, *splat, **kwargs)

class PyPyVm(AbstractPythonIterationsControlVm):

def __init__(self, config_name, options=None, virtualenv=None, iterations=0):

super().__init__("pypy", config_name, options=options, iterations=iterations)

def override_iterations(self, requested_iterations):

# PyPy warms up much faster, half should be enough

return int(requested_iterations / 2)

@property

def interpreter(self):

if home := mx.get_env("PYPY_HOME"):

exe = join(home, "bin", "pypy3")

else:

try:

exe = subprocess.check_output("which pypy3", shell=True).decode().strip()

except OSError:

mx.abort("PYPY_HOME is not set!")

mx.log(f"PyPy {exe=}")

return exe

def run_vm(self, args, *splat, env=None, **kwargs):

env = env or os.environ.copy()

xmxArg = re.compile("--vm.Xmx([0-9]+)([kKgGmM])")

pypyGcMax = "8GB"

for idx, arg in enumerate(args):

if m := xmxArg.search(arg):

args = args[:idx] + args[idx + 1 :]

pypyGcMax = f"{m.group(1)}{m.group(2).upper()}B"

mx.log(f"Setting PYPY_GC_MAX={pypyGcMax} via {arg}")

break

else:

mx.log(

f"Setting PYPY_GC_MAX={pypyGcMax}, use --vm.Xmx argument to override it"

)

env["PYPY_GC_MAX"] = pypyGcMax

return super().run_vm(args, *splat, env=env, **kwargs)

class GraalPythonVm(AbstractPythonIterationsControlVm):

def __init__(self, config_name, options=None, virtualenv=None, iterations=None, extra_polyglot_args=None):

super().__init__(VM_NAME_GRAALPYTHON, config_name, options=options, iterations=iterations)

self._extra_polyglot_args = extra_polyglot_args or []

def override_iterations(self, requested_iterations):

# If we are native and without JIT, half as many iterations should be enough

if self.launcher_type == "native" and "interpreter" in self.config_name():

return int(requested_iterations / 2)

else:

return requested_iterations

@property

@functools.lru_cache

def launcher_type(self):

if mx.dependency("GRAALPY_NATIVE_STANDALONE", fatalIfMissing=False):

return "native"

else:

return "jvm"

@property

@functools.lru_cache

def interpreter(self):

if self.config_name().startswith(CONFIGURATION_CUSTOM):

home = mx.get_env("GRAALPY_HOME")

if not home:

mx.abort("The custom benchmark config for graalpy is to run with a custom GRAALPY_HOME locally")

launcher = join(home, "bin", "graalpy")

mx.log(f"Using {launcher} based on GRAALPY_HOME environment for custom config.")

return launcher

from mx_graalpython import graalpy_standalone

launcher = graalpy_standalone(self.launcher_type, build=False)

mx.log(f"Using {launcher} based on enabled/excluded GraalPy standalone build targets.")

return launcher

def post_process_command_line_args(self, args):

return self.get_extra_polyglot_args() + args

def extract_vm_info(self, args=None):

out_version = subprocess.check_output([self.interpreter, '--version'], universal_newlines=True)

# The benchmark data goes back a ways, we modify the reported dims for

# continuity with the historical queries

graalvm_version_match = re.search(r"$([^$]+ ((?:\d+\.?)+)).*\)", out_version)

if not graalvm_version_match:

mx.log(f"Using {out_version} as platform version string input")

graalvm_version_match = [out_version, out_version, out_version]

dims = {

'guest-vm': self.name(),

'guest-vm-config': self.config_name(),

'host-vm': 'graalvm-' + ('ee' if 'Oracle GraalVM' in out_version else 'ce'),

'host-vm-config': self.launcher_type,

"platform.graalvm-edition": 'EE' if 'Oracle GraalVM' in out_version else 'CE',

"platform.graalvm-version": graalvm_version_match[2],

"platform.graalvm-version-string": graalvm_version_match[1],

}

self._dims = dims

def run(self, *args, **kwargs):

code, out, dims = super().run(*args, **kwargs)

dims.update(self._dims)

is_uncached_config = self.config_name().startswith(CONFIGURATION_UNCACHED)

if ("forced uncached interpreter: True" not in out) == is_uncached_config:

mx.abort(f"ERROR: benchmark config '{CONFIGURATION_UNCACHED}' not consistent with what runtime reported.")

return code, out, dims

def get_extra_polyglot_args(self):

return ["--experimental-options", "-snapshot-startup", "--python.MaxNativeMemory=%s" % (2**34), *self._extra_polyglot_args]

class GraalPythonJavaDriverVm(GuestVm):

def __init__(self, config_name=CONFIGURATION_DEFAULT, cp_suffix=None, distributions=None, cp_prefix=None,

host_vm=None, extra_vm_args=None, extra_polyglot_args=None):

super().__init__(host_vm=host_vm)

self._config_name = config_name

self._distributions = distributions or ['GRAALPYTHON_BENCH']

self._cp_suffix = cp_suffix

self._cp_prefix = cp_prefix

self._extra_vm_args = extra_vm_args

self._extra_polyglot_args = extra_polyglot_args if isinstance(extra_polyglot_args, list) else []

def name(self):

return VM_NAME_GRAALPYTHON

def config_name(self):

return self._config_name

def hosting_registry(self):

return java_vm_registry

def get_classpath(self):

cp = []

if self._cp_prefix:

cp.append(self._cp_prefix)

if self._cp_suffix:

cp.append(self._cp_suffix)

return cp

def with_host_vm(self, host_vm):

return self.__class__(config_name=self._config_name, distributions=self._distributions,

cp_suffix=self._cp_suffix, cp_prefix=self._cp_prefix, host_vm=host_vm,

extra_vm_args=self._extra_vm_args, extra_polyglot_args=self._extra_polyglot_args)

def launcher_class(self):

return 'com.oracle.graal.python.benchmarks.JavaBenchmarkDriver'

def run(self, cwd, args):

extra_polyglot_args = self.get_extra_polyglot_args()

host_vm = self.host_vm()

cp = self.get_classpath()

jhm = mx.dependency("mx:JMH_1_21")

cp_deps = mx.classpath_entries(filter(None, [

mx.distribution('GRAALPYTHON_BENCH', fatalIfMissing=True),

mx.distribution('TRUFFLE_RUNTIME', fatalIfMissing=True),

mx.distribution('TRUFFLE_ENTERPRISE', fatalIfMissing=False),

jhm,

mx.dependency("sdk:LAUNCHER_COMMON")

] + jhm.deps))

cp += [x.classpath_repr() for x in cp_deps]

java_args = ['-cp', ':'.join(cp)] + [self.launcher_class()]

out = mx.TeeOutputCapture(mx.OutputCapture())

code = host_vm.run_java(java_args + extra_polyglot_args + args, cwd=cwd, out=out, err=out)

out = out.underlying.data

dims = host_vm.dimensions(cwd, args, code, out)

return code, out, dims

def get_extra_polyglot_args(self):

return ["--experimental-options", "--python.MaxNativeMemory=%s" % (2**34), *self._extra_polyglot_args]

class GraalPythonPolyBenchVm(GuestVm):

HOSTED_INSTANCE = "GraalPythonPolyBenchVm.hosted="

STAGED_PROGRAM = "staged-benchmark.py"

# Staging rewrites a benchmark into STAGED_PROGRAM. Benchmarks listed here also need data files copied from

# GRAALPYTHON_POLYBENCH_BENCHMARKS to keep their original sibling-file lookups working.

STAGED_SIBLING_RESOURCES = {

"warmup/pyflate-fast.py": ("graalpython-pyflate-benchmark-resource.tar.gz",),

}

JAVA_MODULE_ARGS_WITH_VALUE = (

"--add-exports",

"--add-modules",

"--add-opens",

"--add-reads",

"--enable-native-access",

)

LAUNCHER_ARG_PREFIXES = (

"--python.",

"--engine.",

"--vm.",

"--inspect",

"--log.",

"--experimental-options",

)

PYTHONPATH_LAUNCHER_ARG = "--python.PythonPath"

STAGING_INCOMPATIBLE_ARG_PREFIXES = (

"--vm.",

"--inspect",

)

def __init__(self, config_name=CONFIGURATION_DEFAULT, host_vm=None, extra_launcher_args=None):

super().__init__(host_vm=host_vm)

self._config_name = self.canonical_config_name(config_name)

self._extra_launcher_args = list(extra_launcher_args or [])

self._guest_run_on_java_home_value = None

self._guest_run_on_java_home_set = False

def name(self):

return VM_NAME_GRAALPYTHON

def config_name(self):

return self._config_name

@staticmethod

def canonical_config_name(config_name):

return config_name if config_name else CONFIGURATION_DEFAULT

def hosting_registry(self):

return java_vm_registry

def with_host_vm(self, host_vm):

hosted_name = (

f"{self.HOSTED_INSTANCE}{self.name()}:{self.config_name()}@{host_vm.name()}:{host_vm.config_name()}"

)

if not bm_exec_context().has(hosted_name):

hosted_instance = self.__class__(

self.config_name(), host_vm=host_vm, extra_launcher_args=self._extra_launcher_args

)

bm_exec_context().add_context_value(hosted_name, mx_benchmark.ConstantContextValue(hosted_instance))

hosted_instance = bm_exec_context().get(hosted_name)

if self._guest_run_on_java_home_set:

hosted_instance.set_run_on_java_home(self._guest_run_on_java_home_value)

return hosted_instance

def set_run_on_java_home(self, value):

self._guest_run_on_java_home_value = value

self._guest_run_on_java_home_set = True

host = self.host_vm()

if host is not None and hasattr(host, "set_run_on_java_home"):

host.set_run_on_java_home(value)

def run_on_java_home(self):

if self._guest_run_on_java_home_set:

return self._guest_run_on_java_home_value

host = self.host_vm()

if host is not None and hasattr(host, "run_on_java_home"):

return host.run_on_java_home()

return None

def extract_vm_info(self, args=None):

self._host_vm_for_execution().extract_vm_info(args)

def polybench_staging_run_args(self, run_args):

return [arg for arg in run_args if not self._is_staging_incompatible_arg(arg)]

@classmethod

def _is_launcher_arg(cls, arg):

return arg.startswith(cls.LAUNCHER_ARG_PREFIXES)

@classmethod

def _is_staging_incompatible_arg(cls, arg):

return arg.startswith(cls.STAGING_INCOMPATIBLE_ARG_PREFIXES)

def _launcher_args(self):

run_args = bm_exec_context().get("bm_suite_args")

launcher_args = list(self._extra_launcher_args)

launcher_args += [arg for arg in self.bmSuite.runArgs(run_args) if self._is_launcher_arg(arg)]

return launcher_args

def _host_vm_for_execution(self):

host = self.host_vm()

if host is None:

mx.abort("GraalPy PolyBench guest VM requires a host VM; none was provided.")

if not isinstance(host, mx_sdk_benchmark.NativeImageVM):

mx.abort(f"GraalPy PolyBench guest VM requires NativeImageVM host; got {host.__class__.__name__}.")

if host.pgo_sampler_only or host.pgo_use_perf:

mx.abort("GraalPy PolyBench native launcher staging supports instrumentation PGO, but not PGO sampling.")

return host

def prepare_stages(self, bm_suite, bm_suite_args):

return self._host_vm_for_execution().prepare_stages(bm_suite, bm_suite_args)

def run(self, cwd, args):

host = self._host_vm_for_execution()

self.extract_vm_info(args)

args = host.post_process_command_line_args(args)

host.bmSuite = self.bmSuite

host.stages_info = self.bmSuite.stages_info

out = mx.TeeOutputCapture(mx.OutputCapture())

host.config = mx_sdk_benchmark.NativeImageBenchmarkConfig(host, self.bmSuite, args)

host.stages = mx_sdk_benchmark.StagesContext(

host, out, out, False, os.path.abspath(cwd if cwd else os.getcwd())

)

host.config.output_dir.mkdir(parents=True, exist_ok=True)

host.config.config_dir.mkdir(parents=True, exist_ok=True)

host.config.image_build_reports_directory.mkdir(parents=True, exist_ok=True)

code = self._run_single_stage(host)

output = out.underlying.data

return code, output, host.dimensions(cwd, args, code, output)

def _run_single_stage(self, host):

stage = host.stages_info.current_stage.stage_name

if stage == StageName.INSTRUMENT_IMAGE:

return self._run_stage_instrument_image(host)

if stage == StageName.INSTRUMENT_RUN:

return self._run_stage_instrument_run(host)

if stage == StageName.IMAGE:

return self._run_stage_image(host)

if stage == StageName.RUN:

return self._run_stage_run(host)

mx.abort(f"Unknown stage {stage}")

def _run_stage_instrument_image(self, host):

with host.get_stage_runner() as runner:

exit_code = self._build_standalone(host, runner, instrumented=True)

if exit_code == 0:

self._move_image_build_stats_file(host, StageName.INSTRUMENT_IMAGE)

return exit_code

def _run_stage_image(self, host):

with host.get_stage_runner() as runner:

exit_code = self._build_standalone(host, runner, instrumented=False)

if exit_code == 0:

self._move_image_build_stats_file(host, StageName.IMAGE)

return exit_code

def _run_stage_instrument_run(self, host):

launcher = self._launcher(host, instrumented=True)

profile_vm_args = [

f"--vm.XX:ProfilesDumpFile={host.config.profile_path}",

f"--vm.XX:ProfilesLCOVFile={host.config.output_dir / 'graalpy.info'}",

]

with host.get_stage_runner() as runner:

exit_code = self._stage_harness(host, runner)

if exit_code != 0:

return exit_code

with self._graal_python_vm_args(*profile_vm_args):

with self._graalpy_launcher_env():

return runner.execute_command(

host, [str(launcher)] + self._profile_launcher_args(host) + [str(self._staged_program(host))]

)

def _run_stage_run(self, host):

with host.get_stage_runner() as runner:

exit_code = self._stage_harness(host, runner)

if exit_code != 0:

return exit_code

with self._graalpy_launcher_env():

return runner.execute_command(

host,

[str(self._launcher(host, instrumented=False))]

+ self._run_launcher_args(host)

+ [str(self._staged_program(host))],

)

def _stage_harness(self, host, runner):

# Native stable PolyBench shares image stages across benchmarks, but the staged harness is benchmark-specific.

staging_args = [

"--stage-to-language",

"Python",

"--stage-to-file",

str(self._staged_program(host)),

"--log-staged-program",

"True",

]

self._staged_program(host).parent.mkdir(parents=True, exist_ok=True)

image_run_args = self._polybench_staging_run_args(host, runner.config.image_run_args)

java_args = (

runner.config.classpath_arguments

+ runner.config.modulepath_arguments

+ runner.config.system_properties

+ self._java_staging_vm_args(runner.config.image_vm_args or [])

+ runner.config.executable

+ image_run_args

+ staging_args

)

exit_code = self._execute_setup_command(host, runner, host.generate_java_command(java_args))

if exit_code == 0:

self._copy_staged_sibling_resources(host)

return exit_code

def _copy_staged_sibling_resources(self, host):

# A few staged Python benchmarks load data files relative to __file__. Keep this targeted so the staging

# directory only grows for benchmarks that are known to need sibling resources.

benchmark = bm_exec_context().get("benchmark")

resources = self.STAGED_SIBLING_RESOURCES.get(benchmark, ())

if not resources:

return

source_dir = Path(mx.dependency("GRAALPYTHON_POLYBENCH_BENCHMARKS").get_output()) / Path(benchmark).parent

destination_dir = self._staged_program(host).parent

for resource in resources:

source = source_dir / resource

if not source.is_file():

mx.abort(f"Required staged benchmark resource does not exist: {source}")

shutil.copy2(source, destination_dir / resource)

@staticmethod

def _java_staging_vm_args(image_vm_args):

return [arg for arg in image_vm_args if not arg.startswith("-H")]

@staticmethod

def _execute_setup_command(host, runner, command):

mx.log("Running: ")

mx.log(" ".join(command))

write_output = False

runner.exit_code = mx.run(

command,

out=runner.stdout(write_output),

err=runner.stderr(write_output),

cwd=host.stages.cwd,

nonZeroIsFatal=False,

env=host.config.bm_suite.get_stage_env(),

)

return runner.exit_code

def _build_standalone(self, host, runner, instrumented):

enterprise = host.graalvm_edition == "ee"

use_product_profile = getattr(host, "product_profile", False) and enterprise

if instrumented:

pgo_profile = ""

elif host.pgo_instrumentation:

pgo_profile = host.config.bm_suite.get_pgo_profile_for_image_build(host.config.profile_path)

else:

pgo_profile = None

if pgo_profile is not None:

pgo_profile = str(pgo_profile)

# GRAALPY_PGO_PROFILE is a GraalPy build switch:

# unset: let libpythonvm_build_args auto-select the CI generated PGO profile;

# "": suppress auto-selection and, for benchmark-local PGO, build an instrumented image instead;

# path: build the final image with that explicit profile.

# GRAALPY_REQUIRE_PGO_PROFILE is separate: product-ee sets it to fail if the CI profile cannot be downloaded.

env_pgo_profile = pgo_profile if pgo_profile is not None else (None if use_product_profile else "")

env_require_pgo_profile = "true" if use_product_profile else None

from mx_graalpython import (

BUILD_NATIVE_IMAGE_WITH_ASSERTIONS,

GITHUB_CI,

_graalpy_launcher,

run_mx,

set_env,

)

standalone_dist = "GRAALPY_NATIVE_STANDALONE"

mx_args = ["-p", SUITE.dir, "--env", "native-ee" if enterprise else "native-ce"]

mx_args.append("--extra-image-builder-argument=-Ob" if GITHUB_CI else "--extra-image-builder-argument=-g")

if pgo_profile is not None:

if not enterprise:

mx.abort("PGO is only supported on enterprise NI")

if pgo_profile:

mx_args.append(f"--extra-image-builder-argument=--pgo={pgo_profile}")

mx_args.append("--extra-image-builder-argument=-H:+UnlockExperimentalVMOptions")

mx_args.append("--extra-image-builder-argument=-H:+PGOPrintProfileQuality")

else:

mx_args.append("--extra-image-builder-argument=--pgo-instrument")

mx_args.append("--extra-image-builder-argument=-H:+UnlockExperimentalVMOptions")

mx_args.append("--extra-image-builder-argument=-H:+ProfilingLCOV")

elif BUILD_NATIVE_IMAGE_WITH_ASSERTIONS and not use_product_profile:

mx_args.append("--extra-image-builder-argument=-ea")

mx_args.append("--extra-image-builder-argument=-J-ea")

for arg in self._mx_extra_image_builder_args(self._image_build_args(host)):

mx_args.append(f"--extra-image-builder-argument={arg}")

with set_env(GRAALPY_PGO_PROFILE=env_pgo_profile, GRAALPY_REQUIRE_PGO_PROFILE=env_require_pgo_profile):

write_output = host.stages_info.should_produce_datapoints()

exit_code = run_mx(

mx_args + ["build", "-f", "--only", f"libpythonvm,{standalone_dist}"],

nonZeroIsFatal=False,

out=runner.stdout(write_output),

err=runner.stderr(write_output),

)

runner.exit_code = exit_code

if exit_code != 0:

return exit_code

standalone_home = self._platform_mxbuild_output(standalone_dist)

libpythonvm_home = self._platform_mxbuild_output("libpythonvm")

debug_info = libpythonvm_home / "libpythonvm.so.debug"

if debug_info.exists():

shutil.copy(debug_info, standalone_home / "lib")

destination = self._standalone_home(host, instrumented=instrumented)

if destination.exists():

shutil.rmtree(destination)

shutil.copytree(standalone_home, destination, symlinks=True)

launcher = destination / "bin" / _graalpy_launcher()

if not launcher.is_file():

mx.abort(f"GraalPy launcher was not built at expected path: {launcher}")

self._copy_native_image_artifact(host, destination, instrumented=instrumented)

return exit_code

def _copy_native_image_artifact(self, host, standalone_home, instrumented):

libpythonvm = standalone_home / "lib" / mx.add_lib_suffix(mx.add_lib_prefix("pythonvm"))

if not libpythonvm.is_file():

mx.abort(f"GraalPy native image artifact was not built at expected path: {libpythonvm}")

destination = host.config.instrumented_image_path if instrumented else host.config.image_path

if destination.exists():

destination.unlink()

shutil.copy2(libpythonvm, destination)

def _image_build_args(self, host):

stage_name = host.stages_info.current_stage.stage_name

args = [

f"-H:BuildOutputJSONFile={host.config.get_build_output_json_file(stage_name)}",

"-H:+CollectImageBuildStatistics",

]

args += host.config.system_properties

args += self._standalone_image_vm_args(host.config.image_vm_args or [])

if host.gc:

args += ["--gc=" + host.gc, "-H:+SpawnIsolates"]

if host.native_architecture:

args.append("-march=native")

if host.future_defaults_all:

args.append("--future-defaults=all")

if host.optimization_level:

args.append("-" + host.optimization_level)

if host.is_quickbuild:

args.append("-Ob")

args += host.config.extra_image_build_arguments

return mx_sdk_benchmark.svm_experimental_options(args)

@classmethod

def _mx_extra_image_builder_args(cls, args):

# GraalPy standalone builds receive these through mx --extra-image-builder-argument, whose suite-side

# filtering keeps option-like args and drops bare operands. Use the equivalent --option=value form for

# Java module options that NativeImageVM keeps as space-separated pairs.

normalized_args = []

i = 0

while i < len(args):

arg = args[i]

if arg in cls.JAVA_MODULE_ARGS_WITH_VALUE:

if i + 1 >= len(args):

mx.abort(f"Missing value for native-image argument {arg}")

normalized_args.append(f"{arg}={args[i + 1]}")

i += 2

else:

normalized_args.append(arg)

i += 1

return normalized_args

@classmethod

def _standalone_image_vm_args(cls, args):

# These module-access arguments come from the Java PolyBench launcher command. They are valid for building

# that launcher image, but not for the separate GraalPy standalone/libpythonvm image build.

standalone_args = []

i = 0

while i < len(args):

arg = args[i]

if arg in cls.JAVA_MODULE_ARGS_WITH_VALUE:

i += 2

elif any(arg.startswith(f"{prefix}=") for prefix in cls.JAVA_MODULE_ARGS_WITH_VALUE):

i += 1

else:

standalone_args.append(arg)

i += 1

return standalone_args

def _move_image_build_stats_file(self, host, stage):

stats_file = self._platform_mxbuild_output("libpythonvm") / "reports" / "image_build_statistics.json"

destination = host.config.get_image_build_stats_file(stage)

if not stats_file.is_file():

# Stable PolyBench may revisit an image stage after the shared GraalPy image was already built.

if destination.is_file():

mx.log(f"Reusing image build statistics file from earlier stage execution: {destination}")

return

mx.abort(f"Could not find image build statistics file at expected path: {stats_file}")

if destination.exists():

destination.unlink()

mx.move(stats_file, destination)

def _extra_native_run_args(self, _host, prefix):

vm_args = self.bmSuite.vmArgs(bm_exec_context().get("bm_suite_args"))

return mx_sdk_benchmark.parse_prefixed_args(prefix, vm_args)

@staticmethod

def _platform_mxbuild_output(name):

# The native standalone may be removed from mx's active dependency graph after build; its artifacts still

# live in the platform-dependent mxbuild output directory.

return Path(SUITE.dir) / "mxbuild" / f"{mx.get_os()}-{mx.get_arch()}" / name

def _polybench_staging_run_args(self, host, image_run_args):

extra_run_args = self._extra_native_run_args(host, "-Dnative-image.benchmark.extra-run-arg=")

if extra_run_args and image_run_args[-len(extra_run_args):] == extra_run_args:

return image_run_args[:-len(extra_run_args)]

return image_run_args

def _profile_launcher_args(self, host):

extra_profile_args = self._extra_native_run_args(host, "-Dnative-image.benchmark.extra-profile-run-arg=")

if not extra_profile_args:

extra_profile_args = self._extra_native_run_args(host, "-Dnative-image.benchmark.extra-run-arg=")

return self._launcher_args() + host.config.extra_jvm_args + extra_profile_args

def _run_launcher_args(self, host):

return self._launcher_args() + host.config.extra_jvm_args + self._extra_native_run_args(

host, "-Dnative-image.benchmark.extra-run-arg="

)

def _launcher_pythonpath(self):

prefix = f"{self.PYTHONPATH_LAUNCHER_ARG}="

for arg in reversed(self._launcher_args()):

if arg.startswith(prefix):

return arg[len(prefix):]

return None

@contextmanager

def _graalpy_launcher_env(self):

pythonpath = self._launcher_pythonpath()

if pythonpath is None:

yield

return

from mx_graalpython import set_env

# GraalPy launcher reads PYTHONPATH after generic launcher option parsing, so keep it aligned

# with the explicit PolyBench launcher option while running the staged native standalone.

with set_env(PYTHONPATH=pythonpath):

yield

@contextmanager

def _graal_python_vm_args(self, *new_args):

from mx_graalpython import set_env

existing = os.environ.get("GRAAL_PYTHON_VM_ARGS")

value = "\v".join(([existing] if existing else []) + list(new_args))

with set_env(GRAAL_PYTHON_VM_ARGS=value):

yield

def _staged_program(self, host):

return host.config.output_dir / "graalpython" / self.STAGED_PROGRAM

def _standalone_home(self, host, instrumented):

return host.config.output_dir / ("graalpy-instrumented" if instrumented else "graalpy")

def _launcher(self, host, instrumented):

from mx_graalpython import _graalpy_launcher

return self._standalone_home(host, instrumented=instrumented) / "bin" / _graalpy_launcher()

# ----------------------------------------------------------------------------------------------------------------------

# the benchmark definition

# ----------------------------------------------------------------------------------------------------------------------

python_vm_registry = mx_benchmark.VmRegistry(PYTHON_VM_REGISTRY_NAME, known_host_registries=[java_vm_registry])

python_java_embedding_vm_registry = mx_benchmark.VmRegistry(PYTHON_JAVA_EMBEDDING_VM_REGISTRY_NAME, known_host_registries=[java_vm_registry])

class PythonBaseBenchmarkSuite(VmBenchmarkSuite, AveragingBenchmarkMixin):

def __init__(self, name, benchmarks):

super().__init__()

self._checkup = 'GRAALPYTHON_BENCHMARKS_CHECKUP' in os.environ

self._name = name

self._benchmarks = benchmarks

self._graph_dump_time = None

self._benchmarks_graphs = {}

def has_graph_tags(self, benchmark):

has_tag = False

self._benchmarks_graphs[benchmark] = set()

benchmark_path = join(self._bench_path, "{}.py".format(benchmark))

with open(benchmark_path, 'r') as fp:

text = fp.read()

for l in text.split('\n'):

if '# igv:' in l:

s = l.replace('# igv:', '').strip()

if s:

self._benchmarks_graphs[benchmark].add(s)

has_tag = True

return benchmark if has_tag else None

def has_graph_option(self, bmSuiteArgs):

if "--graph" in bmSuiteArgs:

import time

self._graph_dump_time = time.time()

os.mkdir(join(BENCH_BGV, '%d' % self._graph_dump_time))

bmSuiteArgs.remove('--graph')

return True

return False

def add_graph_dump_option(self, bmSuiteArgs, benchmark):

if self._graph_dump_time and self._benchmarks_graphs.get(benchmark):

dump_opt = '-Dgraal.Dump='

dump_path = join(BENCH_BGV, '%d' % self._graph_dump_time, 'run')

dump_path_opt = '-Dgraal.DumpPath=' + dump_path

if dump_opt not in bmSuiteArgs:

return [dump_opt, dump_path_opt]

return [dump_path_opt]

return []

def post_run_graph(self, benchmark, host_vm_config, guest_vm_config):

graph_tags = self._benchmarks_graphs.get(benchmark)

if self._graph_dump_time and graph_tags:

repo_hash = SUITE.vc.tip(SUITE.dir, 'master').strip()[:7]

name_format = '{h}.{g_vm}.{h_vm}.{s}.{b}.{t}.bgv'

dump_path = join(BENCH_BGV, '%d' % self._graph_dump_time, 'run')

selected_bgv = {}

num_tags = len(graph_tags)

for f in sorted(os.listdir(dump_path), reverse=True):

selected = False

if num_tags == 0 or not f.endswith('.bgv'):

os.remove(join(dump_path, f))

continue

for t in graph_tags:

if t in selected_bgv:

continue

if t in f:

graph_tags.remove(t)

selected = True

num_tags -= 1

selected_bgv[t] = f

break

if not selected:

os.remove(join(dump_path, f))

bench_bgv_dir = join(BENCH_BGV, '%d' % self._graph_dump_time)

for t in selected_bgv:

old_name = selected_bgv[t]

new_name = name_format.format(

h=repo_hash,

g_vm=guest_vm_config,

h_vm=host_vm_config,

s=self._name,

b=benchmark,

t=t)

print('rename: %s to %s' % (old_name, new_name))

os.rename(join(dump_path, old_name), join(bench_bgv_dir, new_name))

if graph_tags:

print('The following tags did not correspond to any bgv file:')

print('\n'.join(graph_tags))

raise FileNotFoundError

def benchmarkList(self, bmSuiteArgs):

if self.has_graph_option(bmSuiteArgs):

return [b for b in self._benchmarks.keys() if self.has_graph_tags(b)]

return list(self._benchmarks.keys())

def benchmarks(self):

raise FutureWarning('the benchmarks method has been deprecated for VmBenchmarkSuite instances, '

'use the benchmarkList method instead')

def successPatterns(self):

return [

re.compile(r"^### iteration=(?P<iteration>[0-9]+), name=(?P<benchmark>[a-zA-Z0-9.\-_]+), duration=(?P<time>[0-9]+(\.[0-9]+)?$)", re.MULTILINE), # pylint: disable=line-too-long

re.compile(r"^@@@ name=(?P<benchmark>[a-zA-Z0-9.\-_]+), duration=(?P<time>[0-9]+(\.[0-9]+)?$)", re.MULTILINE), # pylint: disable=line-too-long

]

def failurePatterns(self):

return [

# lookahead pattern for when truffle compilation details are enabled in the log

re.compile(r"^(?!(\[truffle\])).*Exception")

]

def group(self):

return GROUP_GRAAL

def name(self):

return self._name

def subgroup(self):

return SUBGROUP_GRAAL_PYTHON

@staticmethod

def with_branch_and_commit_dict(d):

"""

We run our benchmark from the graalpython directories, but with other

suites as primary suites in the CI, so we potentially want to update

View remainder of file in raw view

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

Uh oh!

Expand file tree

Breadcrumbs

mx_graalpython_benchmark.py

Latest commit

History

Breadcrumbs

mx_graalpython_benchmark.py

File metadata and controls