cuda-python/docs/module/nvrtc.html at main · Python-Repository-Hub/cuda-python

History

1067 lines (1036 loc) · 67.8 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

<!doctype html>

<title>nvrtc - CUDA Python 11.6.0 documentation</title>

<style>

body {

--color-code-background: #f8f8f8;

--color-code-foreground: black;

}

body[data-theme="dark"] {

--color-code-background: #202020;

--color-code-foreground: #d0d0d0;

}

@media (prefers-color-scheme: dark) {

body:not([data-theme="light"]) {

--color-code-background: #202020;

--color-code-foreground: #d0d0d0;

}

</style></head>

<body>

document.body.dataset.theme = localStorage.getItem("theme") || "auto";

</script>

<title>Contents</title>

stroke-width="1.5" stroke-linecap="round" stroke-linejoin="round">

</svg>

</symbol>

stroke-width="2" stroke-linecap="round" stroke-linejoin="round" class="feather-menu">

</svg>

</symbol>

<title>Expand</title>

stroke-width="2" stroke-linecap="round" stroke-linejoin="round" class="feather-chevron-right">

</svg>

</symbol>

<title>Light mode</title>

stroke-width="1.5" stroke-linecap="round" stroke-linejoin="round" class="feather-sun">

</svg>

</symbol>

stroke-width="1.5" stroke-linecap="round" stroke-linejoin="round" class="icon-tabler-moon">

</svg>

</symbol>

<title>Auto light/dark mode</title>

stroke-width="1.5" stroke-linecap="round" stroke-linejoin="round" class="icon-tabler-shadow">

</svg>

</symbol>

</svg>

<div class="visually-hidden">Hide navigation sidebar</div>

</label>

<div class="visually-hidden">Hide table of contents sidebar</div>

</label>

<div class="visually-hidden">Toggle site navigation sidebar</div>

</label>

</div>

<a href="../index.html"><div class="brand">CUDA Python 11.6.0 documentation</div></a>

</div>

<div class="visually-hidden">Toggle Light / Dark / Auto color theme</div>

</button>

</div>

<div class="visually-hidden">Toggle table of contents sidebar</div>

</label>

</div>

</header>

</div>

CUDA Python 11.6.0 documentation

</a><form class="sidebar-search-container" method="get" action="../search.html" role="search">

</form>

Contents:

<li class="toctree-l1"><a class="reference internal" href="../install.html">Installation</a></li>

<li class="toctree-l1"><a class="reference internal" href="../overview.html">Overview</a></li>

<li class="toctree-l1"><a class="reference internal" href="../motivation.html">Motivation</a></li>

<li class="toctree-l1"><a class="reference internal" href="../conduct.html">Code of Conduct</a></li>

<li class="toctree-l1"><a class="reference internal" href="../contribute.html">Contributing</a></li>

<li class="toctree-l1 has-children"><a class="reference internal" href="../release.html">Release Notes</a><input class="toctree-checkbox" id="toctree-checkbox-1" name="toctree-checkbox-1" role="switch" type="checkbox"/><label for="toctree-checkbox-1"><div class="visually-hidden">Toggle child pages in navigation</div><svg><use href="#svg-arrow-right"></use></svg></label><ul>

</ul>

</li>

<li class="toctree-l1 current has-children"><a class="reference internal" href="../api.html">CUDA Python API Reference</a><input checked="" class="toctree-checkbox" id="toctree-checkbox-2" name="toctree-checkbox-2" role="switch" type="checkbox"/><label for="toctree-checkbox-2"><div class="visually-hidden">Toggle child pages in navigation</div><svg><use href="#svg-arrow-right"></use></svg></label><ul class="current">

<li class="toctree-l2"><a class="reference internal" href="cudart.html">cudart</a></li>

<li class="toctree-l2 current current-page"><a class="current reference internal" href="#">nvrtc</a></li>

</ul>

</li>

</ul>

</div>

</aside>

<div class="visually-hidden">Toggle Light / Dark / Auto color theme</div>

</button>

</div>

<div class="visually-hidden">Toggle table of contents sidebar</div>

</label>

</div>

<h1>nvrtc<a class="headerlink" href="#nvrtc" title="Permalink to this headline">¶</a></h1>

<h2>Error Handling<a class="headerlink" href="#error-handling" title="Permalink to this headline">¶</a></h2>

NVRTC defines the following enumeration type and function for API call error handling.

enum cuda.nvrtc.nvrtcResult(value)<a class="headerlink" href="#cuda.nvrtc.nvrtcResult" title="Permalink to this definition">¶</a></dt>

<dd>The enumerated type nvrtcResult defines API call result codes.

NVRTC API functions return nvrtcResult to indicate the call result.

<dt class="field-odd">Member Type</dt>

</dd>

</dl>

Valid values are as follows:

NVRTC_SUCCESS<a class="headerlink" href="#cuda.nvrtc.nvrtcResult.NVRTC_SUCCESS" title="Permalink to this definition">¶</a></dt>

NVRTC_ERROR_OUT_OF_MEMORY<a class="headerlink" href="#cuda.nvrtc.nvrtcResult.NVRTC_ERROR_OUT_OF_MEMORY" title="Permalink to this definition">¶</a></dt>

NVRTC_ERROR_PROGRAM_CREATION_FAILURE<a class="headerlink" href="#cuda.nvrtc.nvrtcResult.NVRTC_ERROR_PROGRAM_CREATION_FAILURE" title="Permalink to this definition">¶</a></dt>

NVRTC_ERROR_INVALID_INPUT<a class="headerlink" href="#cuda.nvrtc.nvrtcResult.NVRTC_ERROR_INVALID_INPUT" title="Permalink to this definition">¶</a></dt>

NVRTC_ERROR_INVALID_PROGRAM<a class="headerlink" href="#cuda.nvrtc.nvrtcResult.NVRTC_ERROR_INVALID_PROGRAM" title="Permalink to this definition">¶</a></dt>

NVRTC_ERROR_INVALID_OPTION<a class="headerlink" href="#cuda.nvrtc.nvrtcResult.NVRTC_ERROR_INVALID_OPTION" title="Permalink to this definition">¶</a></dt>

NVRTC_ERROR_COMPILATION<a class="headerlink" href="#cuda.nvrtc.nvrtcResult.NVRTC_ERROR_COMPILATION" title="Permalink to this definition">¶</a></dt>

NVRTC_ERROR_BUILTIN_OPERATION_FAILURE<a class="headerlink" href="#cuda.nvrtc.nvrtcResult.NVRTC_ERROR_BUILTIN_OPERATION_FAILURE" title="Permalink to this definition">¶</a></dt>

NVRTC_ERROR_NO_NAME_EXPRESSIONS_AFTER_COMPILATION<a class="headerlink" href="#cuda.nvrtc.nvrtcResult.NVRTC_ERROR_NO_NAME_EXPRESSIONS_AFTER_COMPILATION" title="Permalink to this definition">¶</a></dt>

NVRTC_ERROR_NO_LOWERED_NAMES_BEFORE_COMPILATION<a class="headerlink" href="#cuda.nvrtc.nvrtcResult.NVRTC_ERROR_NO_LOWERED_NAMES_BEFORE_COMPILATION" title="Permalink to this definition">¶</a></dt>

NVRTC_ERROR_NAME_EXPRESSION_NOT_VALID<a class="headerlink" href="#cuda.nvrtc.nvrtcResult.NVRTC_ERROR_NAME_EXPRESSION_NOT_VALID" title="Permalink to this definition">¶</a></dt>

NVRTC_ERROR_INTERNAL_ERROR<a class="headerlink" href="#cuda.nvrtc.nvrtcResult.NVRTC_ERROR_INTERNAL_ERROR" title="Permalink to this definition">¶</a></dt>

</dd></dl>

cuda.nvrtc.nvrtcGetErrorString(result: nvrtcResult)<a class="headerlink" href="#cuda.nvrtc.nvrtcGetErrorString" title="Permalink to this definition">¶</a></dt>

<dd>nvrtcGetErrorString is a helper function that returns a string describing the given nvrtcResult code, e.g., NVRTC_SUCCESS to <cite>“NVRTC_SUCCESS”</cite>. For unrecognized enumeration values, it returns <cite>“NVRTC_ERROR unknown”</cite>.

<dt class="field-odd">Parameters</dt>

<dt>resultnvrtcResult</dt><dd>CUDA Runtime Compilation API result code.

</dd>

</dl>

</dd>

<dt class="field-even">Returns</dt>

<dt>nvrtcResult</dt><dd>Message string for the given nvrtcResult code.

</dd>

<dt>None</dt><dd>None

</dd>

</dl>

</dd>

</dl>

</dd></dl>

</div>

<h2>General Information Query<a class="headerlink" href="#general-information-query" title="Permalink to this headline">¶</a></h2>

NVRTC defines the following function for general information query.

cuda.nvrtc.nvrtcVersion()<a class="headerlink" href="#cuda.nvrtc.nvrtcVersion" title="Permalink to this definition">¶</a></dt>

<dd>nvrtcVersion sets the output parameters <cite>major</cite> and <cite>minor</cite> with the CUDA Runtime Compilation version number.

<dt class="field-odd">Returns</dt>

<dt>nvrtcResult</dt><dd>NVRTC_SUCCESS

NVRTC_ERROR_INVALID_INPUT

</dd>

<dt>majorint</dt><dd>CUDA Runtime Compilation major version number.

</dd>

<dt>minorint</dt><dd>CUDA Runtime Compilation minor version number.

</dd>

</dl>

</dd>

</dl>

</dd></dl>

cuda.nvrtc.nvrtcGetNumSupportedArchs()<a class="headerlink" href="#cuda.nvrtc.nvrtcGetNumSupportedArchs" title="Permalink to this definition">¶</a></dt>

<dd>nvrtcGetNumSupportedArchs sets the output parameter <cite>numArchs</cite> with the number of architectures supported by NVRTC. This can then be used to pass an array to nvrtcGetSupportedArchs to get the supported architectures.

<dt class="field-odd">Returns</dt>

<dt>nvrtcResult</dt><dd>NVRTC_SUCCESS

NVRTC_ERROR_INVALID_INPUT

</dd>

<dt>numArchsint</dt><dd>number of supported architectures.

</dd>

</dl>

</dd>

</dl>

</dd></dl>

cuda.nvrtc.nvrtcGetSupportedArchs()<a class="headerlink" href="#cuda.nvrtc.nvrtcGetSupportedArchs" title="Permalink to this definition">¶</a></dt>

<dd>nvrtcGetSupportedArchs populates the array passed via the output parameter <cite>supportedArchs</cite> with the architectures supported by NVRTC. The array is sorted in the ascending order. The size of the array to be passed can be determined using nvrtcGetNumSupportedArchs.

<dt class="field-odd">Returns</dt>

<dt>nvrtcResult</dt><dd>NVRTC_SUCCESS

NVRTC_ERROR_INVALID_INPUT

</dd>

<dt>supportedArchsList[Int]</dt><dd>sorted array of supported architectures.

</dd>

</dl>

</dd>

</dl>

</dd></dl>

</div>

<h2>Compilation<a class="headerlink" href="#compilation" title="Permalink to this headline">¶</a></h2>

NVRTC defines the following type and functions for actual compilation.

class cuda.nvrtc.nvrtcProgram(*args, **kwargs)<a class="headerlink" href="#cuda.nvrtc.nvrtcProgram" title="Permalink to this definition">¶</a></dt>

<dd>Methods

</colgroup>

<tbody>

<tr class="row-odd"><td>getPtr()</td>

<td>Get memory address of class instance</td>

</tr>

</tbody>

</table></div>

</dd></dl>

cuda.nvrtc.nvrtcCreateProgram(char *src, char *name, int numHeaders, list headers, list includeNames)<a class="headerlink" href="#cuda.nvrtc.nvrtcCreateProgram" title="Permalink to this definition">¶</a></dt>

<dd>nvrtcCreateProgram creates an instance of nvrtcProgram with the given input parameters, and sets the output parameter <cite>prog</cite> with it.

<dt class="field-odd">Parameters</dt>

<dt>srcbytes</dt><dd>CUDA program source.

</dd>

<dt>namebytes</dt><dd>CUDA program name. <cite>name</cite> can be <cite>NULL</cite>; <cite>“default_program”</cite> is

used when <cite>name</cite> is <cite>NULL</cite> or “”.

</dd>

<dt>numHeadersint</dt><dd>Number of headers used. <cite>numHeaders</cite> must be greater than or equal

to 0.

</dd>

<dt>headerslist</dt><dd>Sources of the headers. <cite>headers</cite> can be <cite>NULL</cite> when <cite>numHeaders</cite>

is 0.

</dd>

<dt>includeNameslist</dt><dd>Name of each header by which they can be included in the CUDA

program source. <cite>includeNames</cite> can be <cite>NULL</cite> when <cite>numHeaders</cite> is

0.

</dd>

</dl>

</dd>

<dt class="field-even">Returns</dt>

<dt>nvrtcResult</dt><dd>NVRTC_SUCCESS

NVRTC_ERROR_OUT_OF_MEMORY

NVRTC_ERROR_PROGRAM_CREATION_FAILURE

NVRTC_ERROR_INVALID_INPUT

NVRTC_ERROR_INVALID_PROGRAM

</dd>

<dt>prognvrtcProgram</dt><dd>CUDA Runtime Compilation program.

</dd>

</dl>

</dd>

</dl>

See also

<dt><a class="reference internal" href="#cuda.nvrtc.nvrtcDestroyProgram" title="cuda.nvrtc.nvrtcDestroyProgram"><code class="xref py py-obj docutils literal notranslate">nvrtcDestroyProgram</code></a></dt><dd></dd>

</dl>

</div>

</dd></dl>

cuda.nvrtc.nvrtcDestroyProgram(prog)<a class="headerlink" href="#cuda.nvrtc.nvrtcDestroyProgram" title="Permalink to this definition">¶</a></dt>

<dd>nvrtcDestroyProgram destroys the given program.

<dt class="field-odd">Parameters</dt>

<dt>progAny</dt><dd>CUDA Runtime Compilation program.

</dd>

</dl>

</dd>

<dt class="field-even">Returns</dt>

<dt>nvrtcResult</dt><dd>NVRTC_SUCCESS

NVRTC_ERROR_INVALID_PROGRAM

</dd>

<dt>None</dt><dd>None

</dd>

</dl>

</dd>

</dl>

See also

<dt><a class="reference internal" href="#cuda.nvrtc.nvrtcCreateProgram" title="cuda.nvrtc.nvrtcCreateProgram"><code class="xref py py-obj docutils literal notranslate">nvrtcCreateProgram</code></a></dt><dd></dd>

</dl>

</div>

</dd></dl>

cuda.nvrtc.nvrtcCompileProgram(prog, int numOptions, list options)<a class="headerlink" href="#cuda.nvrtc.nvrtcCompileProgram" title="Permalink to this definition">¶</a></dt>

<dd>nvrtcCompileProgram compiles the given program.

<dt class="field-odd">Parameters</dt>

<dt>progAny</dt><dd>CUDA Runtime Compilation program.

</dd>

<dt>numOptionsint</dt><dd>Number of compiler options passed.

</dd>

<dt>optionslist</dt><dd>Compiler options in the form of C string array. <cite>options</cite> can be

<cite>NULL</cite> when <cite>numOptions</cite> is 0.

</dd>

</dl>

</dd>

<dt class="field-even">Returns</dt>

<dt>nvrtcResult</dt><dd>NVRTC_SUCCESS

NVRTC_ERROR_OUT_OF_MEMORY

NVRTC_ERROR_INVALID_INPUT

NVRTC_ERROR_INVALID_PROGRAM

NVRTC_ERROR_INVALID_OPTION

NVRTC_ERROR_COMPILATION

NVRTC_ERROR_BUILTIN_OPERATION_FAILURE

</dd>

<dt>None</dt><dd>None

</dd>

</dl>

</dd>

</dl>

</dd></dl>

cuda.nvrtc.nvrtcGetPTXSize(prog)<a class="headerlink" href="#cuda.nvrtc.nvrtcGetPTXSize" title="Permalink to this definition">¶</a></dt>

<dd>nvrtcGetPTXSize sets <cite>ptxSizeRet</cite> with the size of the PTX generated by the previous compilation of <cite>prog</cite> (including the trailing <cite>NULL</cite>).

<dt class="field-odd">Parameters</dt>

<dt>progAny</dt><dd>CUDA Runtime Compilation program.

</dd>

</dl>

</dd>

<dt class="field-even">Returns</dt>

<dt>nvrtcResult</dt><dd>NVRTC_SUCCESS

NVRTC_ERROR_INVALID_INPUT

NVRTC_ERROR_INVALID_PROGRAM

</dd>

<dt>ptxSizeRetint</dt><dd>Size of the generated PTX (including the trailing <cite>NULL</cite>).

</dd>

</dl>

</dd>

</dl>

See also

<dt><a class="reference internal" href="#cuda.nvrtc.nvrtcGetPTX" title="cuda.nvrtc.nvrtcGetPTX"><code class="xref py py-obj docutils literal notranslate">nvrtcGetPTX</code></a></dt><dd></dd>

</dl>

</div>

</dd></dl>

cuda.nvrtc.nvrtcGetPTX(prog, char *ptx)<a class="headerlink" href="#cuda.nvrtc.nvrtcGetPTX" title="Permalink to this definition">¶</a></dt>

<dd>nvrtcGetPTX stores the PTX generated by the previous compilation of <cite>prog</cite> in the memory pointed by <cite>ptx</cite>.

<dt class="field-odd">Parameters</dt>

<dt>progAny</dt><dd>CUDA Runtime Compilation program.

</dd>

<dt>ptxbytes</dt><dd>Compiled result.

</dd>

</dl>

</dd>

<dt class="field-even">Returns</dt>

<dt>nvrtcResult</dt><dd>NVRTC_SUCCESS

NVRTC_ERROR_INVALID_INPUT

NVRTC_ERROR_INVALID_PROGRAM

</dd>

<dt>None</dt><dd>None

</dd>

</dl>

</dd>

</dl>

See also

<dt><a class="reference internal" href="#cuda.nvrtc.nvrtcGetPTXSize" title="cuda.nvrtc.nvrtcGetPTXSize"><code class="xref py py-obj docutils literal notranslate">nvrtcGetPTXSize</code></a></dt><dd></dd>

</dl>

</div>

</dd></dl>

cuda.nvrtc.nvrtcGetCUBINSize(prog)<a class="headerlink" href="#cuda.nvrtc.nvrtcGetCUBINSize" title="Permalink to this definition">¶</a></dt>

<dd>nvrtcGetCUBINSize sets <cite>cubinSizeRet</cite> with the size of the cubin generated by the previous compilation of <cite>prog</cite>. The value of cubinSizeRet is set to 0 if the value specified to <cite>-arch</cite> is a virtual architecture instead of an actual architecture.

<dt class="field-odd">Parameters</dt>

<dt>progAny</dt><dd>CUDA Runtime Compilation program.

</dd>

</dl>

</dd>

<dt class="field-even">Returns</dt>

<dt>nvrtcResult</dt><dd>NVRTC_SUCCESS

NVRTC_ERROR_INVALID_INPUT

NVRTC_ERROR_INVALID_PROGRAM

</dd>

<dt>cubinSizeRetint</dt><dd>Size of the generated cubin.

</dd>

</dl>

</dd>

</dl>

See also

<dt><a class="reference internal" href="#cuda.nvrtc.nvrtcGetCUBIN" title="cuda.nvrtc.nvrtcGetCUBIN"><code class="xref py py-obj docutils literal notranslate">nvrtcGetCUBIN</code></a></dt><dd></dd>

</dl>

</div>

</dd></dl>

cuda.nvrtc.nvrtcGetCUBIN(prog, char *cubin)<a class="headerlink" href="#cuda.nvrtc.nvrtcGetCUBIN" title="Permalink to this definition">¶</a></dt>

<dd>nvrtcGetCUBIN stores the cubin generated by the previous compilation of <cite>prog</cite> in the memory pointed by <cite>cubin</cite>. No cubin is available if the value specified to <cite>-arch</cite> is a virtual architecture instead of an actual architecture.

<dt class="field-odd">Parameters</dt>

<dt>progAny</dt><dd>CUDA Runtime Compilation program.

</dd>

<dt>cubinbytes</dt><dd>Compiled and assembled result.

</dd>

</dl>

</dd>

<dt class="field-even">Returns</dt>

<dt>nvrtcResult</dt><dd>NVRTC_SUCCESS

NVRTC_ERROR_INVALID_INPUT

NVRTC_ERROR_INVALID_PROGRAM

</dd>

<dt>None</dt><dd>None

</dd>

</dl>

</dd>

</dl>

See also

<dt><a class="reference internal" href="#cuda.nvrtc.nvrtcGetCUBINSize" title="cuda.nvrtc.nvrtcGetCUBINSize"><code class="xref py py-obj docutils literal notranslate">nvrtcGetCUBINSize</code></a></dt><dd></dd>

</dl>

</div>

</dd></dl>

cuda.nvrtc.nvrtcGetNVVMSize(prog)<a class="headerlink" href="#cuda.nvrtc.nvrtcGetNVVMSize" title="Permalink to this definition">¶</a></dt>

<dd>nvrtcGetNVVMSize sets <cite>nvvmSizeRet</cite> with the size of the NVVM generated by the previous compilation of <cite>prog</cite>. The value of nvvmSizeRet is set to 0 if the program was not compiled with <cite>-dlto</cite>.

<dt class="field-odd">Parameters</dt>

<dt>progAny</dt><dd>CUDA Runtime Compilation program.

</dd>

</dl>

</dd>

<dt class="field-even">Returns</dt>

<dt>nvrtcResult</dt><dd>NVRTC_SUCCESS

NVRTC_ERROR_INVALID_INPUT

NVRTC_ERROR_INVALID_PROGRAM

</dd>

<dt>nvvmSizeRetint</dt><dd>Size of the generated NVVM.

</dd>

</dl>

</dd>

</dl>

See also

<dt><a class="reference internal" href="#cuda.nvrtc.nvrtcGetNVVM" title="cuda.nvrtc.nvrtcGetNVVM"><code class="xref py py-obj docutils literal notranslate">nvrtcGetNVVM</code></a></dt><dd></dd>

</dl>

</div>

</dd></dl>

cuda.nvrtc.nvrtcGetNVVM(prog, char *nvvm)<a class="headerlink" href="#cuda.nvrtc.nvrtcGetNVVM" title="Permalink to this definition">¶</a></dt>

<dd>nvrtcGetNVVM stores the NVVM generated by the previous compilation of <cite>prog</cite> in the memory pointed by <cite>nvvm</cite>. The program must have been compiled with -dlto, otherwise will return an error.

<dt class="field-odd">Parameters</dt>

<dt>progAny</dt><dd>CUDA Runtime Compilation program.

</dd>

<dt>nvvmbytes</dt><dd>Compiled result.

</dd>

</dl>

</dd>

<dt class="field-even">Returns</dt>

<dt>nvrtcResult</dt><dd>NVRTC_SUCCESS

NVRTC_ERROR_INVALID_INPUT

NVRTC_ERROR_INVALID_PROGRAM

</dd>

<dt>None</dt><dd>None

</dd>

</dl>

</dd>

</dl>

See also

<dt><a class="reference internal" href="#cuda.nvrtc.nvrtcGetNVVMSize" title="cuda.nvrtc.nvrtcGetNVVMSize"><code class="xref py py-obj docutils literal notranslate">nvrtcGetNVVMSize</code></a></dt><dd></dd>

</dl>

</div>

</dd></dl>

cuda.nvrtc.nvrtcGetProgramLogSize(prog)<a class="headerlink" href="#cuda.nvrtc.nvrtcGetProgramLogSize" title="Permalink to this definition">¶</a></dt>

<dd>nvrtcGetProgramLogSize sets <cite>logSizeRet</cite> with the size of the log generated by the previous compilation of <cite>prog</cite> (including the trailing <cite>NULL</cite>).

Note that compilation log may be generated with warnings and

informative messages, even when the compilation of <cite>prog</cite> succeeds.

<dt class="field-odd">Parameters</dt>

<dt>progAny</dt><dd>CUDA Runtime Compilation program.

</dd>

</dl>

</dd>

<dt class="field-even">Returns</dt>

<dt>nvrtcResult</dt><dd>NVRTC_SUCCESS

NVRTC_ERROR_INVALID_INPUT

NVRTC_ERROR_INVALID_PROGRAM

</dd>

<dt>logSizeRetint</dt><dd>Size of the compilation log (including the trailing <cite>NULL</cite>).

</dd>

</dl>

</dd>

</dl>

See also

<dt><a class="reference internal" href="#cuda.nvrtc.nvrtcGetProgramLog" title="cuda.nvrtc.nvrtcGetProgramLog"><code class="xref py py-obj docutils literal notranslate">nvrtcGetProgramLog</code></a></dt><dd></dd>

</dl>

</div>

</dd></dl>

cuda.nvrtc.nvrtcGetProgramLog(prog, char *log)<a class="headerlink" href="#cuda.nvrtc.nvrtcGetProgramLog" title="Permalink to this definition">¶</a></dt>

<dd>nvrtcGetProgramLog stores the log generated by the previous compilation of <cite>prog</cite> in the memory pointed by <cite>log</cite>.

<dt class="field-odd">Parameters</dt>

<dt>progAny</dt><dd>CUDA Runtime Compilation program.

</dd>

<dt>logbytes</dt><dd>Compilation log.

</dd>

</dl>

</dd>

<dt class="field-even">Returns</dt>

<dt>nvrtcResult</dt><dd>NVRTC_SUCCESS

NVRTC_ERROR_INVALID_INPUT

NVRTC_ERROR_INVALID_PROGRAM

</dd>

<dt>None</dt><dd>None

</dd>

</dl>

</dd>

</dl>

See also

<dt><a class="reference internal" href="#cuda.nvrtc.nvrtcGetProgramLogSize" title="cuda.nvrtc.nvrtcGetProgramLogSize"><code class="xref py py-obj docutils literal notranslate">nvrtcGetProgramLogSize</code></a></dt><dd></dd>

</dl>

</div>

</dd></dl>

cuda.nvrtc.nvrtcAddNameExpression(prog, char *name_expression)<a class="headerlink" href="#cuda.nvrtc.nvrtcAddNameExpression" title="Permalink to this definition">¶</a></dt>

<dd>nvrtcAddNameExpression notes the given name expression denoting the address of a global function or device/__constant__ variable.

The identical name expression string must be provided on a subsequent

call to nvrtcGetLoweredName to extract the lowered name.

<dt class="field-odd">Parameters</dt>

<dt>progAny</dt><dd>CUDA Runtime Compilation program.

</dd>

<dt>name_expressionbytes</dt><dd>constant expression denoting the address of a global function or

device/__constant__ variable.

</dd>

</dl>

</dd>

<dt class="field-even">Returns</dt>

<dt>nvrtcResult</dt><dd>NVRTC_SUCCESS

NVRTC_ERROR_NO_NAME_EXPRESSIONS_AFTER_COMPILATION

</dd>

<dt>None</dt><dd>None

</dd>

</dl>

</dd>

</dl>

See also

<dt><a class="reference internal" href="#cuda.nvrtc.nvrtcGetLoweredName" title="cuda.nvrtc.nvrtcGetLoweredName"><code class="xref py py-obj docutils literal notranslate">nvrtcGetLoweredName</code></a></dt><dd></dd>

</dl>

</div>

</dd></dl>

cuda.nvrtc.nvrtcGetLoweredName(prog, char *name_expression)<a class="headerlink" href="#cuda.nvrtc.nvrtcGetLoweredName" title="Permalink to this definition">¶</a></dt>

<dd>nvrtcGetLoweredName extracts the lowered (mangled) name for a global function or device/__constant__ variable, and updates lowered_name to point to it. The memory containing the name is released when the NVRTC program is destroyed by nvrtcDestroyProgram. The identical name expression must have been previously provided to nvrtcAddNameExpression.

<dt class="field-odd">Parameters</dt>

<dt>prognvrtcProgram</dt><dd>CUDA Runtime Compilation program.

</dd>

<dt>name_expressionbytes</dt><dd>constant expression denoting the address of a global function or

device/__constant__ variable.

</dd>

</dl>

</dd>

<dt class="field-even">Returns</dt>

<dt>nvrtcResult</dt><dd>NVRTC_SUCCESS

NVRTC_ERROR_NO_LOWERED_NAMES_BEFORE_COMPILATION

NVRTC_ERROR_NAME_EXPRESSION_NOT_VALID

</dd>

<dt>lowered_namebytes</dt><dd>initialized by the function to point to a C string containing the

lowered (mangled) name corresponding to the provided name

expression.

</dd>

</dl>

</dd>

</dl>

See also

<dt><a class="reference internal" href="#cuda.nvrtc.nvrtcAddNameExpression" title="cuda.nvrtc.nvrtcAddNameExpression"><code class="xref py py-obj docutils literal notranslate">nvrtcAddNameExpression</code></a></dt><dd></dd>

</dl>

</div>

</dd></dl>

</div>

<h2>Supported Compile Options<a class="headerlink" href="#supported-compile-options" title="Permalink to this headline">¶</a></h2>

NVRTC supports the compile options below. Option names with two preceding dashs (<code class="docutils literal notranslate">--</code>) are long option names and option names with one preceding dash (<code class="docutils literal notranslate">-</code>) are short option names. Short option names can be used instead of long option names. When a compile option takes an argument, an assignment operator (<code class="docutils literal notranslate">=</code>) is used to separate the compile option argument from the compile option name, e.g., <code class="docutils literal notranslate">"--gpu-architecture=compute_60"</code>. Alternatively, the compile option name and the argument can be specified in separate strings without an assignment operator, .e.g, <code class="docutils literal notranslate">"--gpu-architecture"</code> <code class="docutils literal notranslate">"compute_60"</code>. Single-character short option names, such as <code class="docutils literal notranslate">-D</code>, <code class="docutils literal notranslate">-U</code>, and <code class="docutils literal notranslate">-I</code>, do not require an assignment operator, and the compile option name and the argument can be present in the same string with or without spaces between them. For instance, <code class="docutils literal notranslate">"-D=<def>"</code>, <code class="docutils literal notranslate">"-D<def>"</code>, and <code class="docutils literal notranslate">"-D <def>"</code> are all supported.

The valid compiler options are:

<ul>

<li>Compilation targets

<ul>

<li><code class="docutils literal notranslate">--gpu-architecture=<arch></code> (<code class="docutils literal notranslate">-arch</code>)

Specify the name of the class of GPU architectures for which the input must be compiled.

<li>Valid <code class="docutils literal notranslate"><arch></code>s:

<ul>

<li><code class="docutils literal notranslate">compute_35</code></li>

<li><code class="docutils literal notranslate">compute_37</code></li>

<li><code class="docutils literal notranslate">compute_50</code></li>

<li><code class="docutils literal notranslate">compute_52</code></li>

<li><code class="docutils literal notranslate">compute_53</code></li>

<li><code class="docutils literal notranslate">compute_60</code></li>

<li><code class="docutils literal notranslate">compute_61</code></li>

<li><code class="docutils literal notranslate">compute_62</code></li>

<li><code class="docutils literal notranslate">compute_70</code></li>

<li><code class="docutils literal notranslate">compute_72</code></li>

<li><code class="docutils literal notranslate">compute_75</code></li>

<li><code class="docutils literal notranslate">compute_80</code></li>

</ul>

</li>

<li>Default: <code class="docutils literal notranslate">compute_52</code></li>

</ul>

</li>

</ul>

</li>

<li>Separate compilation / whole-program compilation

<ul>

<li><code class="docutils literal notranslate">--device-c</code> (<code class="docutils literal notranslate">-dc</code>)

Generate relocatable code that can be linked with other relocatable device code. It is equivalent to –relocatable-device-code=true.

</li>

<li><code class="docutils literal notranslate">--device-w</code> (<code class="docutils literal notranslate">-dw</code>)

Generate non-relocatable code. It is equivalent to <code class="docutils literal notranslate">--relocatable-device-code=false</code>.

</li>

<li><code class="docutils literal notranslate">--relocatable-device-code={true|false}</code> (<code class="docutils literal notranslate">-rdc</code>)

Enable (disable) the generation of relocatable device code.

<li>Default: <code class="docutils literal notranslate">false</code></li>

</ul>

</li>

<li><code class="docutils literal notranslate">--extensible-whole-program</code> (<code class="docutils literal notranslate">-ewp</code>)

Do extensible whole program compilation of device code.

<li>Default: <code class="docutils literal notranslate">false</code></li>

</ul>

</li>

</ul>

</li>

<li>Debugging support

<ul>

<li><code class="docutils literal notranslate">--device-debug</code> (<code class="docutils literal notranslate">-G</code>)

Generate debug information.

</li>

<li><code class="docutils literal notranslate">--generate-line-info</code> (<code class="docutils literal notranslate">-lineinfo</code>)

Generate line-number information.

</li>

</ul>

</li>

<li>Code generation

<ul>

<li><code class="docutils literal notranslate">--ptxas-options</code> <options> (<code class="docutils literal notranslate">-Xptxas</code>)</li>

<li><code class="docutils literal notranslate">--ptxas-options=<options></code>

Specify options directly to ptxas, the PTX optimizing assembler.

</li>

<li><code class="docutils literal notranslate">--maxrregcount=<N></code> (<code class="docutils literal notranslate">-maxrregcount</code>)

Specify the maximum amount of registers that GPU functions can use. Until a function-specific limit, a higher value will generally increase the performance of individual GPU threads that execute this function. However, because thread registers are allocated from a global register pool on each GPU, a higher value of this option will also reduce the maximum thread block size, thereby reducing the amount of thread parallelism. Hence, a good maxrregcount value is the result of a trade-off. If this option is not specified, then no maximum is assumed. Value less than the minimum registers required by ABI will be bumped up by the compiler to ABI minimum limit.

</li>

<li><code class="docutils literal notranslate">--ftz={true|false}</code> (<code class="docutils literal notranslate">-ftz</code>)

When performing single-precision floating-point operations, flush denormal values to zero or preserve denormal values. <code class="docutils literal notranslate">--use_fast_math</code> implies <code class="docutils literal notranslate">--ftz=true</code>.

<li>Default: <code class="docutils literal notranslate">false</code></li>

</ul>

</li>

<li><code class="docutils literal notranslate">--prec-sqrt={true|false}</code> (<code class="docutils literal notranslate">-prec-sqrt</code>)

For single-precision floating-point square root, use IEEE round-to-nearest mode or use a faster approximation. <code class="docutils literal notranslate">--use_fast_math</code> implies <code class="docutils literal notranslate">--prec-sqrt=false</code>.

<li>Default: <code class="docutils literal notranslate">true</code></li>

</ul>

</li>

<li><code class="docutils literal notranslate">--prec-div={true|false}</code> (<code class="docutils literal notranslate">-prec-div</code>)

For single-precision floating-point division and reciprocals, use IEEE round-to-nearest mode or use a faster approximation. <code class="docutils literal notranslate">--use_fast_math</code> implies <code class="docutils literal notranslate">--prec-div=false</code>.

<li>Default: <code class="docutils literal notranslate">true</code></li>

</ul>

</li>

<li><code class="docutils literal notranslate">--fmad={true|false}</code> (<code class="docutils literal notranslate">-fmad</code>)

Enables (disables) the contraction of floating-point multiplies and adds/subtracts into floating-point multiply-add operations (FMAD, FFMA, or DFMA). <code class="docutils literal notranslate">--use_fast_math</code> implies <code class="docutils literal notranslate">--fmad=true</code>.

<li>Default: <code class="docutils literal notranslate">true</code></li>

</ul>

</li>

<li><code class="docutils literal notranslate">--use_fast_math</code> (<code class="docutils literal notranslate">-use_fast_math</code>)

Make use of fast math operations. <code class="docutils literal notranslate">--use_fast_math</code> implies <code class="docutils literal notranslate">--ftz=true</code> <code class="docutils literal notranslate">--prec-div=false</code> <code class="docutils literal notranslate">--prec-sqrt=false</code> <code class="docutils literal notranslate">--fmad=true</code>.

</li>

<li><code class="docutils literal notranslate">--extra-device-vectorization</code> (<code class="docutils literal notranslate">-extra-device-vectorization</code>)

Enables more aggressive device code vectorization in the NVVM optimizer.

</li>

<li><code class="docutils literal notranslate">--modify-stack-limit={true|false}</code> (<code class="docutils literal notranslate">-modify-stack-limit</code>)

On Linux, during compilation, use <code class="docutils literal notranslate">setrlimit()</code> to increase stack size to maximum allowed. The limit is reset to the previous value at the end of compilation. Note: <code class="docutils literal notranslate">setrlimit()</code> changes the value for the entire process.

<li>Default: <code class="docutils literal notranslate">true</code></li>

</ul>

</li>

<li><code class="docutils literal notranslate">--dlink-time-opt</code> (<code class="docutils literal notranslate">-dlto</code>)

Generate intermediate code for later link-time optimization. It implies <code class="docutils literal notranslate">-rdc=true</code>.

Note: when this is used the nvvmGetNVVM API should be used, as PTX or Cubin will not be generated.

</li>

</ul>

</li>

<li>Preprocessing

<ul>

<li><code class="docutils literal notranslate">--define-macro=<def></code> (<code class="docutils literal notranslate">-D</code>)

<code class="docutils literal notranslate"><def></code> can be either <code class="docutils literal notranslate"><name></code> or <code class="docutils literal notranslate"><name=definitions></code>.

<ul>

Predefine <code class="docutils literal notranslate"><name></code> as a macro with definition <code class="docutils literal notranslate">1</code>.

</li>

The contents of <code class="docutils literal notranslate"><definition></code> are tokenized and preprocessed as if they appeared during translation phase three in a <code class="docutils literal notranslate">#define</code> directive. In particular, the definition will be truncated by embedded new line characters.

</li>

</ul>

</li>

<li><code class="docutils literal notranslate">--undefine-macro=<def></code> (<code class="docutils literal notranslate">-U</code>)

Cancel any previous definition of <code class="docutils literal notranslate"><def></code>.

</li>

<li><code class="docutils literal notranslate">--include-path=<dir></code> (<code class="docutils literal notranslate">-I</code>)

Add the directory <code class="docutils literal notranslate"><dir></code> to the list of directories to be searched for headers. These paths are searched after the list of headers given to nvrtcCreateProgram.

</li>

<li><code class="docutils literal notranslate">--pre-include=<header></code> (<code class="docutils literal notranslate">-include</code>)

Preinclude <code class="docutils literal notranslate"><header></code> during preprocessing.

</li>

<li><code class="docutils literal notranslate">--no-source-include</code> (<code class="docutils literal notranslate">-no-source-include</code>) The preprocessor by default adds the directory of each input sources to the include path. This option disables this feature and only considers the path specified explicitly.</li>

</ul>

</li>

<li>Language Dialect

<ul>

<li><code class="docutils literal notranslate">--std={c++03|c++11|c++14|c++17}</code> (<code class="docutils literal notranslate">-std={c++11|c++14|c++17}</code>)

Set language dialect to C++03, C++11, C++14 or C++17

</li>

<li><code class="docutils literal notranslate">--builtin-move-forward={true|false}</code> (<code class="docutils literal notranslate">-builtin-move-forward</code>)

Provide builtin definitions of <code class="docutils literal notranslate">std::move</code> and <code class="docutils literal notranslate">std::forward</code>, when C++11 language dialect is selected.

<li>Default: <code class="docutils literal notranslate">true</code></li>

</ul>

</li>

<li><code class="docutils literal notranslate">--builtin-initializer-list={true|false}</code> (<code class="docutils literal notranslate">-builtin-initializer-list</code>)

Provide builtin definitions of <code class="docutils literal notranslate">std::initializer_list</code> class and member functions when C++11 language dialect is selected.

<li>Default: <code class="docutils literal notranslate">true</code></li>

</ul>

</li>

</ul>

</li>

<li>Misc.

<ul>

<li><code class="docutils literal notranslate">--disable-warnings</code> (<code class="docutils literal notranslate">-w</code>)

Inhibit all warning messages.

</li>

<li><code class="docutils literal notranslate">--restrict</code> (<code class="docutils literal notranslate">-restrict</code>)

Programmer assertion that all kernel pointer parameters are restrict pointers.

</li>

<li><code class="docutils literal notranslate">--device-as-default-execution-space</code> (<code class="docutils literal notranslate">-default-device</code>)

Treat entities with no execution space annotation as <code class="docutils literal notranslate">device</code> entities.

</li>

<li><code class="docutils literal notranslate">--device-int128</code> (<code class="docutils literal notranslate">-device-int128</code>)

Allow the <code class="docutils literal notranslate">__int128</code> type in device code. Also causes the macro <code class="docutils literal notranslate">CUDACC_RTC_INT128</code> to be defined.

</li>

<li><code class="docutils literal notranslate">--optimization-info=<kind></code> (<code class="docutils literal notranslate">-opt-info</code>)

Provide optimization reports for the specified kind of optimization. The following kind tags are supported:

<li><code class="docutils literal notranslate">inline</code> : emit a remark when a function is inlined.</li>

</ul>

</li>

<li><code class="docutils literal notranslate">--version-ident={true|false}</code> (<code class="docutils literal notranslate">-dQ</code>)

Embed used compiler’s version info into generated PTX/CUBIN

<li>Default: <code class="docutils literal notranslate">false</code></li>

</ul>

</li>

<li><code class="docutils literal notranslate">--display-error-number</code> (<code class="docutils literal notranslate">-err-no</code>)

Display diagnostic number for warning messages. (Default)

</li>

<li><code class="docutils literal notranslate">--no-display-error-number</code> (<code class="docutils literal notranslate">-no-err-no</code>)

Disables the display of a diagnostic number for warning messages.

</li>

<li><code class="docutils literal notranslate">--diag-error=<error-number></code>,… (<code class="docutils literal notranslate">-diag-error</code>)

Emit error for specified diagnostic message number(s). Message numbers can be separated by comma.

</li>

<li><code class="docutils literal notranslate">--diag-suppress=<error-number></code>,… (<code class="docutils literal notranslate">-diag-suppress</code>)

Suppress specified diagnostic message number(s). Message numbers can be separated by comma.

</li>

<li><code class="docutils literal notranslate">--diag-warn=<error-number></code>,… (<code class="docutils literal notranslate">-diag-warn</code>)

Emit warning for specified diagnostic message number(s). Message numbers can be separated by comma.

</li>

</ul>

</li>

</ul>

</div>

</article>

</div>

View remainder of file in raw view

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

FilesExpand file tree

nvrtc.html

Latest commit

History

nvrtc.html

File metadata and controls