Viewing: github-redirect.dependabot.com

Repository navigation

More items

Blame

History

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

"""Print a summary of specialization stats for all files in the

default stats folders.

"""

from __future__ import annotations

# NOTE: Bytecode introspection modules (opcode, dis, etc.) should only

# be imported when loading a single dataset. When comparing datasets, it

# could get it wrong, leading to subtle errors.

import argparse

import collections

from collections.abc import KeysView

from dataclasses import dataclass

from datetime import date

import enum

import functools

import itertools

import json

from operator import itemgetter

import os

from pathlib import Path

import re

import sys

import textwrap

from typing import Any, Callable, TextIO, TypeAlias

RawData: TypeAlias = dict[str, Any]

Rows: TypeAlias = list[tuple]

Columns: TypeAlias = tuple[str, ...]

RowCalculator: TypeAlias = Callable[["Stats"], Rows]

# TODO: Check for parity

if os.name == "nt":

DEFAULT_DIR = "c:\\temp\\py_stats\\"

else:

DEFAULT_DIR = "/tmp/py_stats/"

SOURCE_DIR = Path(__file__).parents[2]

TOTAL = "specialization.hit", "specialization.miss", "execution_count"

def pretty(name: str) -> str:

return name.replace("_", " ").lower()

def _load_metadata_from_source():

def get_defines(filepath: Path, prefix: str = "SPEC_FAIL"):

with open(SOURCE_DIR / filepath) as spec_src:

defines = collections.defaultdict(list)

start = "#define " + prefix + "_"

for line in spec_src:

line = line.strip()

if not line.startswith(start):

continue

line = line[len(start) :]

name, val = line.split()

defines[int(val.strip())].append(name.strip())

return defines

import opcode

return {

"_specialized_instructions": [

op for op in opcode._specialized_opmap.keys() if "__" not in op # type: ignore

"_stats_defines": get_defines(

Path("Include") / "cpython" / "pystats.h", "EVAL_CALL"

"_defines": get_defines(Path("Python") / "specialize.c"),

}

def load_raw_data(input: Path) -> RawData:

if input.is_file():

with open(input, "r") as fd:

data = json.load(fd)

data["_stats_defines"] = {int(k): v for k, v in data["_stats_defines"].items()}

data["_defines"] = {int(k): v for k, v in data["_defines"].items()}

return data

elif input.is_dir():

stats = collections.Counter[str]()

for filename in input.iterdir():

with open(filename) as fd:

for line in fd:

try:

key, value = line.split(":")

except ValueError:

print(

f"Unparsable line: '{line.strip()}' in {filename}",

file=sys.stderr,

)

continue

# Hack to handle older data files where some uops

# are missing an underscore prefix in their name

if key.startswith("uops[") and key[5:6] != "_":

key = "uops[_" + key[5:]

stats[key.strip()] += int(value)

stats["__nfiles__"] += 1

data = dict(stats)

data.update(_load_metadata_from_source())

return data

else:

raise ValueError(f"{input} is not a file or directory path")

def save_raw_data(data: RawData, json_output: TextIO):

json.dump(data, json_output)

@dataclass(frozen=True)

class Doc:

text: str

doc: str

def markdown(self) -> str:

return textwrap.dedent(

f"""

{self.text}

{self.doc}

</details>

"""

)

class Count(int):

def markdown(self) -> str:

return format(self, ",d")

@dataclass(frozen=True)

class Ratio:

num: int

den: int | None = None

percentage: bool = True

def __float__(self):

if self.den == 0:

return 0.0

elif self.den is None:

return self.num

else:

return self.num / self.den

def markdown(self) -> str:

if self.den is None:

return ""

elif self.den == 0:

if self.num != 0:

return f"{self.num:,} / 0 !!"

return ""

elif self.percentage:

return f"{self.num / self.den:,.01%}"

else:

return f"{self.num / self.den:,.02f}"

class DiffRatio(Ratio):

def __init__(self, base: int | str, head: int | str):

if isinstance(base, str) or isinstance(head, str):

super().__init__(0, 0)

else:

super().__init__(head - base, base)

class OpcodeStats:

"""

Manages the data related to specific set of opcodes, e.g. tier1 (with prefix

"opcode") or tier2 (with prefix "uops").

"""

def __init__(self, data: dict[str, Any], defines, specialized_instructions):

self._data = data

self._defines = defines

self._specialized_instructions = specialized_instructions

def get_opcode_names(self) -> KeysView[str]:

return self._data.keys()

def get_pair_counts(self) -> dict[tuple[str, str], int]:

pair_counts = {}

for name_i, opcode_stat in self._data.items():

for key, value in opcode_stat.items():

if value and key.startswith("pair_count"):

name_j, _, _ = key[len("pair_count") + 1 :].partition("]")

pair_counts[(name_i, name_j)] = value

return pair_counts

def get_total_execution_count(self) -> int:

return sum(x.get("execution_count", 0) for x in self._data.values())

def get_execution_counts(self) -> dict[str, tuple[int, int]]:

counts = {}

for name, opcode_stat in self._data.items():

if "execution_count" in opcode_stat:

count = opcode_stat["execution_count"]

miss = 0

if "specializable" not in opcode_stat:

miss = opcode_stat.get("specialization.miss", 0)

counts[name] = (count, miss)

return counts

@functools.cache

def _get_pred_succ(

self,

) -> tuple[dict[str, collections.Counter], dict[str, collections.Counter]]:

pair_counts = self.get_pair_counts()

predecessors: dict[str, collections.Counter] = collections.defaultdict(

collections.Counter

)

successors: dict[str, collections.Counter] = collections.defaultdict(

collections.Counter

)

for (first, second), count in pair_counts.items():

if count:

predecessors[second][first] = count

successors[first][second] = count

return predecessors, successors

def get_predecessors(self, opcode: str) -> collections.Counter[str]:

return self._get_pred_succ()[0][opcode]

def get_successors(self, opcode: str) -> collections.Counter[str]:

return self._get_pred_succ()[1][opcode]

def _get_stats_for_opcode(self, opcode: str) -> dict[str, int]:

return self._data[opcode]

def get_specialization_total(self, opcode: str) -> int:

family_stats = self._get_stats_for_opcode(opcode)

return sum(family_stats.get(kind, 0) for kind in TOTAL)

def get_specialization_counts(self, opcode: str) -> dict[str, int]:

family_stats = self._get_stats_for_opcode(opcode)

result = {}

for key, value in sorted(family_stats.items()):

if key.startswith("specialization."):

label = key[len("specialization.") :]

if label in ("success", "failure") or label.startswith("failure_kinds"):

continue

elif key in (

"execution_count",

"specializable",

) or key.startswith("pair"):

continue

else:

label = key

result[label] = value

return result

def get_specialization_success_failure(self, opcode: str) -> dict[str, int]:

family_stats = self._get_stats_for_opcode(opcode)

result = {}

for key in ("specialization.success", "specialization.failure"):

label = key[len("specialization.") :]

val = family_stats.get(key, 0)

result[label] = val

return result

def get_specialization_failure_total(self, opcode: str) -> int:

return self._get_stats_for_opcode(opcode).get("specialization.failure", 0)

def get_specialization_failure_kinds(self, opcode: str) -> dict[str, int]:

def kind_to_text(kind: int, opcode: str):

if kind <= 8:

return pretty(self._defines[kind][0])

if opcode == "LOAD_SUPER_ATTR":

opcode = "SUPER"

elif opcode.endswith("ATTR"):

opcode = "ATTR"

elif opcode in ("FOR_ITER", "GET_ITER", "SEND"):

opcode = "ITER"

elif opcode.endswith("SUBSCR"):

opcode = "SUBSCR"

for name in self._defines[kind]:

if name.startswith(opcode):

return pretty(name[len(opcode) + 1 :])

return "kind " + str(kind)

family_stats = self._get_stats_for_opcode(opcode)

def key_to_index(key):

return int(key[:-1].split("[")[1])

max_index = 0

for key in family_stats:

if key.startswith("specialization.failure_kind"):

max_index = max(max_index, key_to_index(key))

failure_kinds = [0] * (max_index + 1)

for key in family_stats:

if not key.startswith("specialization.failure_kind"):

continue

failure_kinds[key_to_index(key)] = family_stats[key]

return {

kind_to_text(index, opcode): value

for (index, value) in enumerate(failure_kinds)

if value

}

def is_specializable(self, opcode: str) -> bool:

return "specializable" in self._get_stats_for_opcode(opcode)

def get_specialized_total_counts(self) -> tuple[int, int, int]:

basic = 0

specialized_hits = 0

specialized_misses = 0

not_specialized = 0

for opcode, opcode_stat in self._data.items():

if "execution_count" not in opcode_stat:

continue

count = opcode_stat["execution_count"]

if "specializable" in opcode_stat:

not_specialized += count

elif opcode in self._specialized_instructions:

miss = opcode_stat.get("specialization.miss", 0)

specialized_hits += count - miss

specialized_misses += miss

else:

basic += count

return basic, specialized_hits, specialized_misses, not_specialized

def get_deferred_counts(self) -> dict[str, int]:

return {

opcode: opcode_stat.get("specialization.deferred", 0)

for opcode, opcode_stat in self._data.items()

if opcode != "RESUME"

}

def get_misses_counts(self) -> dict[str, int]:

return {

opcode: opcode_stat.get("specialization.miss", 0)

for opcode, opcode_stat in self._data.items()

if not self.is_specializable(opcode)

}

def get_opcode_counts(self) -> dict[str, int]:

counts = {}

for opcode, entry in self._data.items():

count = entry.get("count", 0)

if count:

counts[opcode] = count

return counts

class Stats:

def __init__(self, data: RawData):

self._data = data

def get(self, key: str) -> int:

return self._data.get(key, 0)

@functools.cache

def get_opcode_stats(self, prefix: str) -> OpcodeStats:

opcode_stats = collections.defaultdict[str, dict](dict)

for key, value in self._data.items():

if not key.startswith(prefix):

continue

name, _, rest = key[len(prefix) + 1 :].partition("]")

opcode_stats[name][rest.strip(".")] = value

return OpcodeStats(

opcode_stats,

self._data["_defines"],

self._data["_specialized_instructions"],

)

def get_call_stats(self) -> dict[str, int]:

defines = self._data["_stats_defines"]

result = {}

for key, value in sorted(self._data.items()):

if "Calls to" in key:

result[key] = value

elif key.startswith("Calls "):

name, index = key[:-1].split("[")

label = f"{name} ({pretty(defines[int(index)][0])})"

result[label] = value

for key, value in sorted(self._data.items()):

if key.startswith("Frame"):

result[key] = value

return result

def get_object_stats(self) -> dict[str, tuple[int, int]]:

total_materializations = self._data.get("Object inline values", 0)

total_allocations = self._data.get("Object allocations", 0) + self._data.get(

"Object allocations from freelist", 0

)

total_increfs = (

self._data.get("Object interpreter mortal increfs", 0) +

self._data.get("Object mortal increfs", 0) +

self._data.get("Object interpreter immortal increfs", 0) +

self._data.get("Object immortal increfs", 0)

)

total_decrefs = (

self._data.get("Object interpreter mortal decrefs", 0) +

self._data.get("Object mortal decrefs", 0) +

self._data.get("Object interpreter immortal decrefs", 0) +

self._data.get("Object immortal decrefs", 0)

)

result = {}

for key, value in self._data.items():

if key.startswith("Object"):

if "materialize" in key:

den = total_materializations

elif "allocations" in key:

den = total_allocations

elif "increfs" in key:

den = total_increfs

elif "decrefs" in key:

den = total_decrefs

else:

den = None

label = key[6:].strip()

label = label[0].upper() + label[1:]

result[label] = (value, den)

return result

def get_gc_stats(self) -> list[dict[str, int]]:

gc_stats: list[dict[str, int]] = []

for key, value in self._data.items():

if not key.startswith("GC"):

continue

n, _, rest = key[3:].partition("]")

name = rest.strip()

gen_n = int(n)

while len(gc_stats) <= gen_n:

gc_stats.append({})

gc_stats[gen_n][name] = value

return gc_stats

def get_optimization_stats(self) -> dict[str, tuple[int, int | None]]:

if "Optimization attempts" not in self._data:

return {}

attempts = self._data["Optimization attempts"]

created = self._data["Optimization traces created"]

executed = self._data["Optimization traces executed"]

uops = self._data["Optimization uops executed"]

trace_stack_overflow = self._data["Optimization trace stack overflow"]

trace_stack_underflow = self._data["Optimization trace stack underflow"]

trace_too_long = self._data["Optimization trace too long"]

trace_too_short = self._data["Optimization trace too short"]

inner_loop = self._data["Optimization inner loop"]

recursive_call = self._data["Optimization recursive call"]

low_confidence = self._data["Optimization low confidence"]

unknown_callee = self._data["Optimization unknown callee"]

executors_invalidated = self._data["Executors invalidated"]

return {

Doc(

"Optimization attempts",

"The number of times a potential trace is identified. Specifically, this "

"occurs in the JUMP BACKWARD instruction when the counter reaches a "

"threshold.",

): (attempts, None),

Doc(

"Traces created", "The number of traces that were successfully created."

): (created, attempts),

Doc(

"Trace stack overflow",

"A trace is truncated because it would require more than 5 stack frames.",

): (trace_stack_overflow, attempts),

Doc(

"Trace stack underflow",

"A potential trace is abandoned because it pops more frames than it pushes.",

): (trace_stack_underflow, attempts),

Doc(

"Trace too long",

"A trace is truncated because it is longer than the instruction buffer.",

): (trace_too_long, attempts),

Doc(

"Trace too short",

"A potential trace is abandoned because it is too short.",

): (trace_too_short, attempts),

Doc(

"Inner loop found", "A trace is truncated because it has an inner loop"

): (inner_loop, attempts),

Doc(

"Recursive call",

"A trace is truncated because it has a recursive call.",

): (recursive_call, attempts),

Doc(

"Low confidence",

"A trace is abandoned because the likelihood of the jump to top being taken "

"is too low.",

): (low_confidence, attempts),

Doc(

"Unknown callee",

"A trace is abandoned because the target of a call is unknown.",

): (unknown_callee, attempts),

Doc(

"Executors invalidated",

"The number of executors that were invalidated due to watched "

"dictionary changes.",

): (executors_invalidated, created),

Doc("Traces executed", "The number of traces that were executed"): (

executed,

None,

Doc(

"Uops executed",

"The total number of uops (micro-operations) that were executed",

): (

uops,

executed,

}

def get_optimizer_stats(self) -> dict[str, tuple[int, int | None]]:

attempts = self._data["Optimization optimizer attempts"]

successes = self._data["Optimization optimizer successes"]

no_memory = self._data["Optimization optimizer failure no memory"]

builtins_changed = self._data["Optimizer remove globals builtins changed"]

incorrect_keys = self._data["Optimizer remove globals incorrect keys"]

return {

Doc(

"Optimizer attempts",

"The number of times the trace optimizer (_Py_uop_analyze_and_optimize) was run.",

): (attempts, None),

Doc(

"Optimizer successes",

"The number of traces that were successfully optimized.",

): (successes, attempts),

Doc(

"Optimizer no memory",

"The number of optimizations that failed due to no memory.",

): (no_memory, attempts),

Doc(

"Remove globals builtins changed",

"The builtins changed during optimization",

): (builtins_changed, attempts),

Doc(

"Remove globals incorrect keys",

"The keys in the globals dictionary aren't what was expected",

): (incorrect_keys, attempts),

}

def get_jit_memory_stats(self) -> dict[Doc, tuple[int, int | None]]:

jit_total_memory_size = self._data["JIT total memory size"]

jit_code_size = self._data["JIT code size"]

jit_trampoline_size = self._data["JIT trampoline size"]

jit_data_size = self._data["JIT data size"]

jit_padding_size = self._data["JIT padding size"]

jit_freed_memory_size = self._data["JIT freed memory size"]

return {

Doc(

"Total memory size",

"The total size of the memory allocated for the JIT traces",

): (jit_total_memory_size, None),

Doc(

"Code size",

"The size of the memory allocated for the code of the JIT traces",

): (jit_code_size, jit_total_memory_size),

Doc(

"Trampoline size",

"The size of the memory allocated for the trampolines of the JIT traces",

): (jit_trampoline_size, jit_total_memory_size),

Doc(

"Data size",

"The size of the memory allocated for the data of the JIT traces",

): (jit_data_size, jit_total_memory_size),

Doc(

"Padding size",

"The size of the memory allocated for the padding of the JIT traces",

): (jit_padding_size, jit_total_memory_size),

Doc(

"Freed memory size",

"The size of the memory freed from the JIT traces",

): (jit_freed_memory_size, jit_total_memory_size),

}

def get_histogram(self, prefix: str) -> list[tuple[int, int]]:

rows = []

for k, v in self._data.items():

match = re.match(f"{prefix}\\[([0-9]+)\\]", k)

if match is not None:

entry = int(match.groups()[0])

rows.append((entry, v))

rows.sort()

return rows

def get_rare_events(self) -> list[tuple[str, int]]:

prefix = "Rare event "

return [

(key[len(prefix) + 1 : -1].replace("_", " "), val)

for key, val in self._data.items()

if key.startswith(prefix)

]

class JoinMode(enum.Enum):

# Join using the first column as a key

SIMPLE = 0

# Join using the first column as a key, and indicate the change in the

# second column of each input table as a new column

CHANGE = 1

# Join using the first column as a key, indicating the change in the second

# column of each input table as a new column, and omit all other columns

CHANGE_ONE_COLUMN = 2

# Join using the first column as a key, and indicate the change as a new

# column, but don't sort by the amount of change.

CHANGE_NO_SORT = 3

class Table:

"""

A Table defines how to convert a set of Stats into a specific set of rows

displaying some aspect of the data.

"""

def __init__(

self,

column_names: Columns,

calc_rows: RowCalculator,

join_mode: JoinMode = JoinMode.SIMPLE,

self.columns = column_names

self.calc_rows = calc_rows

self.join_mode = join_mode

def join_row(self, key: str, row_a: tuple, row_b: tuple) -> tuple:

match self.join_mode:

case JoinMode.SIMPLE:

return (key, *row_a, *row_b)

case JoinMode.CHANGE | JoinMode.CHANGE_NO_SORT:

return (key, *row_a, *row_b, DiffRatio(row_a[0], row_b[0]))

case JoinMode.CHANGE_ONE_COLUMN:

return (key, row_a[0], row_b[0], DiffRatio(row_a[0], row_b[0]))

def join_columns(self, columns: Columns) -> Columns:

match self.join_mode:

case JoinMode.SIMPLE:

return (

columns[0],

*("Base " + x for x in columns[1:]),

*("Head " + x for x in columns[1:]),

)

case JoinMode.CHANGE | JoinMode.CHANGE_NO_SORT:

return (

columns[0],

*("Base " + x for x in columns[1:]),

*("Head " + x for x in columns[1:]),

) + ("Change:",)

case JoinMode.CHANGE_ONE_COLUMN:

return (

columns[0],

"Base " + columns[1],

"Head " + columns[1],

"Change:",

)

def join_tables(self, rows_a: Rows, rows_b: Rows) -> tuple[Columns, Rows]:

ncols = len(self.columns)

default = ("",) * (ncols - 1)

data_a = {x[0]: x[1:] for x in rows_a}

data_b = {x[0]: x[1:] for x in rows_b}

if len(data_a) != len(rows_a) or len(data_b) != len(rows_b):

raise ValueError("Duplicate keys")

# To preserve ordering, use A's keys as is and then add any in B that

# aren't in A

keys = list(data_a.keys()) + [k for k in data_b.keys() if k not in data_a]

rows = [

self.join_row(k, data_a.get(k, default), data_b.get(k, default))

for k in keys

]

if self.join_mode in (JoinMode.CHANGE, JoinMode.CHANGE_ONE_COLUMN):

rows.sort(key=lambda row: abs(float(row[-1])), reverse=True)

columns = self.join_columns(self.columns)

return columns, rows

def get_table(

self, base_stats: Stats, head_stats: Stats | None = None

) -> tuple[Columns, Rows]:

if head_stats is None:

rows = self.calc_rows(base_stats)

return self.columns, rows

else:

rows_a = self.calc_rows(base_stats)

rows_b = self.calc_rows(head_stats)

cols, rows = self.join_tables(rows_a, rows_b)

return cols, rows

class Section:

"""

A Section defines a section of the output document.

"""

def __init__(

self,

title: str = "",

summary: str = "",

part_iter=None,

comparative: bool = True,

doc: str = "",

self.title = title

if not summary:

self.summary = title.lower()

else:

self.summary = summary

self.doc = textwrap.dedent(doc)

if part_iter is None:

part_iter = []

if isinstance(part_iter, list):

def iter_parts(base_stats: Stats, head_stats: Stats | None):

yield from part_iter

self.part_iter = iter_parts

else:

self.part_iter = part_iter

self.comparative = comparative

def calc_execution_count_table(prefix: str) -> RowCalculator:

def calc(stats: Stats) -> Rows:

opcode_stats = stats.get_opcode_stats(prefix)

counts = opcode_stats.get_execution_counts()

total = opcode_stats.get_total_execution_count()

cumulative = 0

rows: Rows = []

for opcode, (count, miss) in sorted(

counts.items(), key=itemgetter(1), reverse=True

cumulative += count

if miss:

miss_val = Ratio(miss, count)

else:

miss_val = None

rows.append(

(

opcode,

Count(count),

Ratio(count, total),

Ratio(cumulative, total),

miss_val,

)

return rows

return calc

def execution_count_section() -> Section:

return Section(

"Execution counts",

"Execution counts for Tier 1 instructions.",

[

Table(

("Name", "Count:", "Self:", "Cumulative:", "Miss ratio:"),

calc_execution_count_table("opcode"),

join_mode=JoinMode.CHANGE_ONE_COLUMN,

)

doc="""

The "miss ratio" column shows the percentage of times the instruction

executed that it deoptimized. When this happens, the base unspecialized

instruction is not counted.

""",

)

def pair_count_section(prefix: str, title=None) -> Section:

def calc_pair_count_table(stats: Stats) -> Rows:

opcode_stats = stats.get_opcode_stats(prefix)

pair_counts = opcode_stats.get_pair_counts()

total = opcode_stats.get_total_execution_count()

cumulative = 0

rows: Rows = []

for (opcode_i, opcode_j), count in itertools.islice(

sorted(pair_counts.items(), key=itemgetter(1), reverse=True), 100

cumulative += count

rows.append(

(

f"{opcode_i} {opcode_j}",

Count(count),

Ratio(count, total),

Ratio(cumulative, total),

)

return rows

return Section(

"Pair counts",

f"Pair counts for top 100 {title if title else prefix} pairs",

[

Table(

("Pair", "Count:", "Self:", "Cumulative:"),

calc_pair_count_table,

)

comparative=False,

doc="""

Pairs of specialized operations that deoptimize and are then followed by

the corresponding unspecialized instruction are not counted as pairs.

""",

)

def pre_succ_pairs_section() -> Section:

def iter_pre_succ_pairs_tables(base_stats: Stats, head_stats: Stats | None = None):

assert head_stats is None

opcode_stats = base_stats.get_opcode_stats("opcode")

for opcode in opcode_stats.get_opcode_names():

predecessors = opcode_stats.get_predecessors(opcode)

successors = opcode_stats.get_successors(opcode)

predecessors_total = predecessors.total()

successors_total = successors.total()

if predecessors_total == 0 and successors_total == 0:

continue

pred_rows = [

(pred, Count(count), Ratio(count, predecessors_total))

for (pred, count) in predecessors.most_common(5)

]

succ_rows = [

(succ, Count(count), Ratio(count, successors_total))

for (succ, count) in successors.most_common(5)

]

yield Section(

opcode,

f"Successors and predecessors for {opcode}",

[

Table(

("Predecessors", "Count:", "Percentage:"),

lambda *_: pred_rows, # type: ignore

Table(

("Successors", "Count:", "Percentage:"),

lambda *_: succ_rows, # type: ignore

)

return Section(

"Predecessor/Successor Pairs",

"Top 5 predecessors and successors of each Tier 1 opcode.",

iter_pre_succ_pairs_tables,

comparative=False,

doc="""

This does not include the unspecialized instructions that occur after a

specialized instruction deoptimizes.

""",

)

def specialization_section() -> Section:

def calc_specialization_table(opcode: str) -> RowCalculator:

def calc(stats: Stats) -> Rows:

DOCS = {

"deferred": 'Lists the number of "deferred" (i.e. not specialized) instructions executed.',

"hit": "Specialized instructions that complete.",

"miss": "Specialized instructions that deopt.",

"deopt": "Specialized instructions that deopt.",

}

opcode_stats = stats.get_opcode_stats("opcode")

total = opcode_stats.get_specialization_total(opcode)

specialization_counts = opcode_stats.get_specialization_counts(opcode)

return [

(

Doc(label, DOCS[label]),

Count(count),

Ratio(count, total),

)

for label, count in specialization_counts.items()

]

return calc

def calc_specialization_success_failure_table(name: str) -> RowCalculator:

def calc(stats: Stats) -> Rows:

values = stats.get_opcode_stats(

"opcode"

).get_specialization_success_failure(name)

total = sum(values.values())

if total:

return [

(label.capitalize(), Count(val), Ratio(val, total))

for label, val in values.items()

]

else:

return []

return calc

def calc_specialization_failure_kind_table(name: str) -> RowCalculator:

def calc(stats: Stats) -> Rows:

opcode_stats = stats.get_opcode_stats("opcode")

failures = opcode_stats.get_specialization_failure_kinds(name)

total = opcode_stats.get_specialization_failure_total(name)

return sorted(

[

(label, Count(value), Ratio(value, total))

for label, value in failures.items()

if value

key=itemgetter(1),

reverse=True,

)

return calc

def iter_specialization_tables(base_stats: Stats, head_stats: Stats | None = None):

opcode_base_stats = base_stats.get_opcode_stats("opcode")

names = opcode_base_stats.get_opcode_names()

if head_stats is not None:

opcode_head_stats = head_stats.get_opcode_stats("opcode")

names &= opcode_head_stats.get_opcode_names() # type: ignore

else:

opcode_head_stats = None

for opcode in sorted(names):

if not opcode_base_stats.is_specializable(opcode):

continue

if opcode_base_stats.get_specialization_total(opcode) == 0 and (

opcode_head_stats is None

or opcode_head_stats.get_specialization_total(opcode) == 0

continue

yield Section(

opcode,

f"specialization stats for {opcode} family",

[

Table(

("Kind", "Count:", "Ratio:"),

calc_specialization_table(opcode),

JoinMode.CHANGE,

Table(

("Success", "Count:", "Ratio:"),

calc_specialization_success_failure_table(opcode),

JoinMode.CHANGE,

Table(

("Failure kind", "Count:", "Ratio:"),

calc_specialization_failure_kind_table(opcode),

JoinMode.CHANGE,

)

return Section(

"Specialization stats",

"Specialization stats by family",

iter_specialization_tables,

)

def specialization_effectiveness_section() -> Section:

def calc_specialization_effectiveness_table(stats: Stats) -> Rows:

opcode_stats = stats.get_opcode_stats("opcode")

total = opcode_stats.get_total_execution_count()

(

basic,

specialized_hits,

specialized_misses,

not_specialized,

) = opcode_stats.get_specialized_total_counts()

return [

(

Doc(

View remainder of file in raw view

Terms
Privacy
Security
Status
Community
Docs
Contact
Manage cookies
Do not share my personal information

Navigation Menu

Repository navigation

Files

summarize_stats.py

summarize_stats.py

Navigation Menu

Repository navigation

FilesExpand file tree

Breadcrumbs

summarize_stats.py

Latest commit

History

Breadcrumbs

summarize_stats.py

File metadata and controls

Footer

Footer navigation

Files