Commit Graph

Select branches

Hide Pull Requests

fix-extelemwise-in-combine-ops

gh-pages

jit-hook

keren/assert

keren/improve-hook

keren/insert-slice-other-nonzero

keren/perf-debug

keren/v100-perf-regression

master

phil/fused-attention-perf-fixup

phil/mma-v1-is-row-debug

phil/swizzle-bug-repro

port-fma

rocm

#10

#100

#1000

#1001

#1002

#1004

#1006

#1007

#1008

#101

#1010

#1012

#1013

#1013

#1014

#1018

#1019

#102

#1020

#1020

#1025

#1027

#1028

#1029

#103

#1030

#1033

#1034

#1036

#1037

#1038

#1039

#104

#1042

#1043

#1043

#105

#106

#107

#108

#109

#11

#11

#110

#111

#112

#114

#116

#118

#119

#120

#121

#123

#124

#125

#126

#127

#128

#129

#13

#131

#132

#133

#134

#135

#136

#137

#138

#139

#140

#141

#142

#143

#144

#145

#146

#147

#148

#149

#15

#151

#152

#158

#164

#165

#167

#168

#172

#173

#178

#179

#18

#180

#185

#186

#188

#19

#190

#192

#193

#195

#198

#199

#20

#200

#203

#204

#205

#207

#209

#212

#219

#22

#222

#224

#225

#228

#23

#23

#231

#24

#240

#249

#250

#251

#253

#255

#256

#257

#258

#259

#260

#261

#268

#271

#272

#273

#276

#279

#28

#280

#281

#282

#283

#285

#286

#287

#288

#291

#292

#293

#294

#295

#296

#297

#298

#299

#3

#300

#301

#302

#303

#304

#305

#306

#307

#308

#309

#311

#312

#313

#314

#317

#318

#320

#324

#326

#331

#336

#337

#338

#342

#344

#345

#346

#347

#348

#349

#35

#350

#351

#356

#357

#358

#361

#362

#367

#368

#372

#373

#374

#377

#379

#38

#380

#381

#382

#383

#386

#387

#388

#390

#391

#392

#393

#394

#395

#396

#397

#399

#40

#400

#401

#403

#406

#407

#408

#409

#41

#413

#414

#415

#417

#418

#420

#421

#422

#423

#424

#425

#426

#427

#428

#430

#431

#432

#436

#438

#439

#440

#442

#444

#445

#446

#447

#448

#449

#45

#450

#451

#453

#455

#456

#457

#458

#462

#463

#464

#467

#468

#469

#470

#471

#473

#474

#478

#481

#482

#483

#484

#485

#487

#488

#490

#491

#492

#493

#495

#499

#500

#501

#502

#503

#505

#507

#510

#513

#514

#515

#516

#519

#52

#520

#522

#523

#524

#526

#527

#528

#53

#531

#533

#534

#535

#538

#538

#539

#541

#545

#546

#547

#548

#549

#551

#552

#553

#555

#556

#557

#559

#560

#561

#562

#564

#565

#567

#569

#57

#570

#571

#572

#575

#575

#577

#578

#579

#58

#582

#587

#588

#59

#590

#595

#598

#60

#600

#601

#602

#604

#606

#607

#608

#61

#614

#614

#617

#62

#623

#63

#632

#636

#637

#644

#65

#650

#651

#652

#653

#654

#655

#657

#658

#66

#660

#661

#662

#663

#664

#665

#666

#667

#668

#669

#670

#671

#672

#678

#68

#682

#683

#684

#685

#689

#69

#691

#692

#693

#694

#696

#697

#699

#7

#70

#700

#701

#702

#703

#704

#706

#708

#709

#71

#710

#711

#712

#715

#716

#718

#722

#724

#726

#727

#728

#729

#73

#732

#733

#735

#736

#738

#739

#740

#742

#746

#747

#749

#75

#750

#751

#752

#753

#754

#755

#757

#758

#759

#764

#765

#766

#767

#769

#77

#774

#775

#776

#777

#78

#780

#782

#784

#785

#786

#788

#789

#790

#791

#792

#794

#796

#797

#798

#799

#80

#800

#801

#803

#804

#805

#809

#81

#812

#814

#815

#816

#817

#818

#819

#82

#820

#821

#822

#823

#825

#826

#827

#829

#83

#830

#831

#833

#834

#835

#836

#837

#838

#839

#840

#841

#842

#843

#844

#845

#847

#848

#849

#850

#851

#852

#853

#854

#856

#857

#858

#859

#86

#862

#863

#864

#867

#868

#869

#87

#872

#873

#874

#875

#876

#877

#878

#879

#88

#880

#881

#883

#885

#886

#887

#887

#888

#889

#89

#890

#890

#894

#896

#897

#898

#899

#90

#901

#902

#903

#904

#906

#907

#908

#909

#91

#910

#912

#913

#914

#915

#916

#917

#918

#92

#920

#921

#922

#923

#924

#925

#926

#927

#928

#929

#93

#930

#931

#933

#936

#937

#938

#939

#94

#941

#943

#944

#945

#946

#947

#947

#948

#95

#951

#952

#953

#956

#957

#958

#959

#96

#960

#961

#962

#963

#964

#966

#968

#969

#97

#970

#971

#972

#973

#975

#976

#977

#978

#979

#980

#982

#982

#983

#985

#987

#988

#990

#991

#993

#994

#995

#996

#997

#998

#999

isaac

legacy-backend

v0.1

v0.2.3

v0.4

v1.0

v1.1

v1.1.1

v1.1.2

Mono Color

• f20f48a255 Move Jokeren 2022-12-06 13:29:29 -08:00
• 3eff110fbc Restore Jokeren 2022-12-06 13:28:43 -08:00
• 5f85b79718 Merge branch 'triton-mlir' into keren/insert-slice-other-nonzero Jokeren 2022-12-06 13:25:20 -08:00
• bab7338965 Fix Jokeren 2022-12-06 13:24:50 -08:00
• 74f3d7a80f Fix Jokeren 2022-12-06 12:53:25 -08:00
• 115cd3ac47 [FRONTEND] Added reshape as an alias for view (for now) (#956) Philippe Tillet 2022-12-06 09:57:05 -08:00
• 532e10cf87 [FRONTEND][BACKEND] Clean-up transpositions (#953) Philippe Tillet 2022-12-06 09:32:13 -08:00
• 16e973edf2 [BACKEND] Fix dependency analysis in pipeline (#946) Keren Zhou 2022-12-06 09:08:55 -08:00
• b539e031e8 Add test Jokeren 2022-12-05 23:38:54 -08:00
• 46fa29496c Init Jokeren 2022-12-05 23:18:13 -08:00
• 9490252261 [FRONTEND] Support alternative install locations of system libdevice.10.bc (#951) Crutcher Dunnavant 2022-12-05 19:41:44 -08:00
• e419781978 [Triton-MLIR][BACKEND] Make mmav1 works on basic cases (#944) Yan Chunwei 2022-12-06 10:57:08 +08:00
• 189491727a [FRONTEND] Extract and unify @builtin/@extern (#913) Crutcher Dunnavant 2022-12-05 14:59:41 -08:00
• e0072d210a [FRONTEND] Propagate mypy types through @jit, @builtin, etc (#915) Crutcher Dunnavant 2022-12-05 14:41:02 -08:00
• 2fa17588f7 [FRONTEND] Expand __init__ * imports, add __all__ (#912) Crutcher Dunnavant 2022-12-05 14:22:55 -08:00
• e057c65cf0 [BACKEND] Porting the legacy heuristic rule in assigning shared layout for A/B of MMAv1 (#948) goostavz 2022-12-06 03:30:23 +08:00
• 99c7e0e008 [BUILD] Change default build type (#945) Philippe Tillet 2022-12-03 17:47:33 -08:00
• f2fcaeabf3 [BACKEND] Support dot op when the output is mma encoding and allowtf32 is true (#937) Keren Zhou 2022-12-03 11:14:12 -08:00
• 8edfe813a5 [FRONTEND][BACKEND] Added trans instruction; made flash attention bwd pass work (#943) Philippe Tillet 2022-12-03 09:58:24 -08:00
• 4d64589b22 [Triton-MLIR][Backend] Fix the definition of MmaEncodingAttr v1, and the output sequence of DotConversion in MMAv1 (#941) goostavz 2022-12-03 21:12:48 +08:00
• 8650b4d1cb [DRIVER] Fix typos (#939) legacy-backend Yang Hau 2022-12-03 03:13:46 +08:00
• 521ff9ad74 [TRITON-MLIR][FRONTEND]fix scf.if to run through layernorm tutorial (#938) donproc 2022-12-02 17:45:29 +08:00
• c280ebda1b [Triton-MLIR][BACKEND] Fix the membar pass to add missing barriers caused by scf.for (#933) Keren Zhou 2022-12-01 11:54:18 -08:00
• 9def1bcebf [TRITON-MLIR][FRONTEND]minor fix to run through atomic_cas test (#925) donproc 2022-12-01 21:43:26 +08:00
• 7d90a07d0b [Triton-MLIR][BACKEND] Refactor decompose insert_slice_async (#929) Keren Zhou 2022-11-30 10:07:34 -08:00
• 6461254fb5 [BACKEND] Make flash attention forward pass work (#928) Philippe Tillet 2022-11-30 11:13:24 +01:00
• 4e6a8209ed [Triton-MLIR] Two fixes on allocation and backend related with MMA v1 (#930) goostavz 2022-11-30 17:27:26 +08:00
• 9bb54402b3 [FRONTEND][BACKEND] Small fixes to multiple_of, num_programs, axisinfo; enable block-sparse tests (#927) Philippe Tillet 2022-11-29 20:00:34 +01:00
• 66c36c4378 [BACKEND] Fixed bounds-wrapping issues (#926) Philippe Tillet 2022-11-29 17:56:45 +01:00
• 661be523c0 [Triton-MLIR][BACKEND] Minor fixes of shared memory in ReduceOpConversion (#924) Qingyi Liu 2022-11-29 11:50:31 +08:00
• c87fbf886e [Triton-MLIR][BACKEND] Remove static and unnamed namespace in Utility.h (#923) Yan Chunwei 2022-11-29 09:06:06 +08:00
• dfc8e7fb95 Fix keren/perf-debug Jokeren 2022-11-28 13:47:13 -08:00
• 2f9aef1132 Fix Jokeren 2022-11-28 13:00:26 -08:00
• f605d95b82 unroll_2 Jokeren 2022-11-28 12:59:05 -08:00
• b378118647 c64 Jokeren 2022-11-28 12:19:52 -08:00
• cfcf042e55 Init Jokeren 2022-11-28 11:55:41 -08:00
• 0c1d4d764e [Triton-MLIR][Backend] support MMA v1 in ConvertLayout (#922) goostavz 2022-11-28 16:10:30 +08:00
• 9d31998a9d [Triton-MLIR][BACKEND] Add argmin / argmax implementation for ReduceOp (#918) Qingyi Liu 2022-11-28 14:59:27 +08:00
• 04ec5deb41 [Triton-MLIR][BACKEND] decouple the dot code (#921) Yan Chunwei 2022-11-28 13:30:27 +08:00
• 630dc315ee [Triton-MLIR] uncomment the UT in test_gemm that has already been fixed (#920) goostavz 2022-11-28 11:23:20 +08:00
• 35c9ec1103 [Triton-MLIR][Backend] Fix number of warps and threads per warp when matrices are small (#917) Keren Zhou 2022-11-26 12:30:38 -08:00
• ee098d0341 Merge branch 'master' into keren/improve-hook Jokeren 2022-11-25 15:04:59 -08:00
• f63be0e9b5 [TRITON-MLIR][BACKEND]support atomic_cas (#914) donproc 2022-11-25 12:02:08 +08:00
• feef58ee8a Pass fn to CompiliedKernel Jokeren 2022-11-24 14:22:35 -08:00
• 153aecb339 [Triton-MLIR][BACKEND] insert_slice_async on GPUs < sm80 (#908) Keren Zhou 2022-11-24 14:05:54 -08:00
• f98aed1258 [Triton-MLIR][RUNTIME] Add /usr/bin/ptxas as a search path (#909) Crutcher Dunnavant 2022-11-24 10:49:16 -08:00
• ace7d28736 [Triton-MLIR][RUNTIME] Fix ir metadata lookup bug (#910) Crutcher Dunnavant 2022-11-24 00:27:23 -08:00
• b688f7b7b8 [Triton-MLIR] add_volta_warpsPerTile (#907) ben-zhang-609 2022-11-24 09:44:29 +08:00
• 8925c2cd11 [TRITON-MLIR][BACKEND]AtomicRMWOp supports scalar (#903) donproc 2022-11-23 15:59:09 +08:00
• 2e33352419 [Triton-MLIR] Fix side effects (#906) Keren Zhou 2022-11-22 23:29:18 -08:00
• 037f9efa95 [Triton-MLIR][BACKEND] Fix wpt overflow issue in mma v2 (#904) Yan Chunwei 2022-11-23 11:27:15 +08:00
• 07786dc932 [Triton-MLIR] Add compute capability (#902) ben-zhang-609 2022-11-23 03:08:23 +08:00
• 2afebcd79b [Triton-MLIR][Backend] Remove unnecessary barriers (#901) Keren Zhou 2022-11-22 10:03:29 -08:00
• 136668bac3 [Triton-MLIR][BACKEND] tiny code cleanup (#899) Yan Chunwei 2022-11-21 16:00:46 +08:00
• 04b852e031 [Triton-MLIR] Fix warnings and variable names (#898) Keren Zhou 2022-11-20 22:25:27 -08:00
• 85cccfb81f [BUILD] Fix compilation problems in the release build (#897) Keren Zhou 2022-11-20 21:40:36 -08:00
• 23f71daa27 [OPTIMIZER] Fixed up order of shared layouts (#881) Philippe Tillet 2022-11-21 06:25:02 +01:00
• 44f577984d Fix format double substitution bug: {i} => {{i}} (#886) Crutcher Dunnavant 2022-11-20 11:44:42 -08:00
• 4d64ffb5fe [FRONTEND] Handle for loops with negative constant steps (#896) Philippe Tillet 2022-11-20 11:37:38 +01:00
• 6c5f646f4e [WIP][Triton-MLIR] Prefetch pass fixup (#873) Keren Zhou 2022-11-19 19:57:16 -08:00
• e8994209f4 [Triton-MLIR][Backend]fix mma-v2 transpose error (#888) Yan Chunwei 2022-11-20 11:29:09 +08:00
• 8a5647782d [Triton-MLIR][Testing]Fix tests warning, with small code clean-up (#894) Jun Yang 2022-11-19 22:33:59 +08:00
• afaf59b0c9 [TRITON-MLIR][BACKEND] Atomic support mask (#889) donproc 2022-11-19 19:57:19 +08:00
• 46fd581b0a Merge pull request #29 from ROCmSoftwarePlatform/parse_amdgcn_from_rocminfo rocm rsanthanam-amd 2022-11-18 12:53:25 -06:00
• 8cc448d92e Changes to eliminate the need for the MI_GPU_ARCH environment variable. Rohit Santhanam 2022-11-18 12:58:51 +00:00
• dab4855bdf [TESTING] Added infrastructure for executing TTGIR program and test for layout conversions (#885) Philippe Tillet 2022-11-18 07:46:45 +01:00
• 9ea6135eb5 [Triton-MLIR][Backend] Some cleanup in getMultiDimIndex/getLinearIndex (#880) goostavz 2022-11-18 09:19:21 +08:00
• 0e4691e6dd [FRONTEND] Fix ExternLibrary(format=) bug; type annotate build_extern.py (#883) Crutcher Dunnavant 2022-11-17 09:45:30 -08:00
• 5eee738df7 [Triton-MLIR][FRONTEND] [BACKEND] fix atomics (#879) donproc 2022-11-16 12:25:15 +08:00
• 37f5846280 [Triton-MLIR][Backend] Minor fix for allocation and backend in handling tt.ptr tensors (#878) goostavz 2022-11-15 18:08:07 +08:00
• a22ff39017 [Triton-MLIR][BACKEND] Refine/add codegen for get_promgram_id and get_num_programs Op (#877) Yan Chunwei 2022-11-15 15:45:24 +08:00
• 4c4159c6fa [Triton-MLIR] Add ex2.approx implementation for ExpOp and fix smem allocation for ReduceOpConversion (#875) Qingyi Liu 2022-11-15 09:27:32 +08:00
• c28cfd821b [Triton-MLIR][Backend] Fix convert_layout blocked->shared in non-default order (#876) goostavz 2022-11-15 09:02:46 +08:00
• 1eedaf7bec [Triton-MLIR][BACKEND] adapt DotOp layout for FMADot (#872) Yan Chunwei 2022-11-14 16:56:30 +08:00
• 516a241234 [Triton-MLIR] Fix some typos (#874) Chenggang Zhao 2022-11-14 10:15:53 +08:00
• f40c63fb03 [Triton-MLIR][OPTIMIZER] Cleaned up swizzling (#869) Philippe Tillet 2022-11-10 12:05:46 -08:00
• 2aa538ec2e [BACKEND] Added support for mma layouts in reductions (#863) Philippe Tillet 2022-11-10 09:58:07 -08:00
• 57fd1864a7 [Triton-MLIR] Support FP8 (#864) Chenggang Zhao 2022-11-10 15:53:06 +08:00
• 4946167241 [Triton-MLIR] tt.dot operands now must have DotOperand layout; also added prefetch pass prototype (#712) Da Yan 2022-11-10 13:57:27 +08:00
• 8832e32683 [Triton-MLIR][BACKEND] Refine ptxbuilder (#867) Yan Chunwei 2022-11-10 13:41:52 +08:00
• 4640023d9b [Triton-MLIR][Backend]add atomic rmw without mask (#842) donproc 2022-11-10 08:15:58 +08:00
• 0c87360657 [Triton-MLIR][Backend] Port FMADot conversion for DotOp (#844) Yan Chunwei 2022-11-09 12:57:50 +08:00
• de5b84c476 [Triton-MLIR][Backend] Fix mma<v2> int8 precision error (#850) Yan Chunwei 2022-11-09 12:23:43 +08:00
• e517b58d59 [Triton-MLIR] Minor fixes to enable fused-softmax and layer-norm tutorials (#835) Qingyi Liu 2022-11-09 10:18:56 +08:00
• 2da71b2aaa [Triton-MLIR] Increase block size K to completely eliminate shared memory bank conflicts (#862) Keren Zhou 2022-11-08 17:39:23 -08:00
• 080b4addf8 [Triton-MLIR][Backend] Fix the order in linear/delinear and a few bugs in reduce conversion (#851) goostavz 2022-11-09 02:10:09 +08:00
• 303790da88 [BUILD] use Python Var In Tests (#859) Ian Bearman 2022-11-08 09:44:19 -08:00
• 137344946f [OPTIMIZER] Fix the load-mask issue with the pipeline pass (#857) Da Yan 2022-11-09 01:29:53 +08:00
• 976cf12af1 [OPTIMIZER] Fixed memory coalescing (#847) Philippe Tillet 2022-11-07 06:22:18 -08:00
• b6f15e214b [FRONTEND] Fixed up type cast in atomics codegen (#853) Philippe Tillet 2022-11-07 05:46:24 -08:00
• 84ad215268 [Triton-MLIR] Enable libdevice for ptx backend when has external functions. (#848) ben-zhang-609 2022-11-07 16:01:50 +08:00
• fdd59900f7 [Triton-MLIR] Replace triton.extract_slice with tensor.extract_slice and support more general tensor slicing (#837) Keren Zhou 2022-11-06 22:59:03 -08:00
• a4ff0c362c [FRONTEND] Fix issues with atomics (#849) Philippe Tillet 2022-11-06 20:52:11 -08:00
• d767919bc1 [OPTIMIZER] Not using MMA on FP32 when allowTF32 is false port-fma Phil Tillet 2022-11-04 23:16:28 -07:00
• 0d7e753227 [TESTING] use torch.int for autotuning cache (#840) Natalia Gimelshein 2022-11-04 18:05:16 -07:00
• b6dbe959f0 [RUNTIME] Re-vamped cache so users can manually patch IR / ptx / cubin files (#845) Philippe Tillet 2022-11-04 10:57:29 -07:00
• b39cc56f93 up Superjomn 2022-11-04 18:04:20 +08:00
• db64477153 Merge remote-tracking branch 'origin/triton-mlir' into port-fma Superjomn 2022-11-04 17:43:54 +08:00
• 95d8e383cb add testing Superjomn 2022-11-04 17:43:09 +08:00
• 1ed6ee34ba finish coding Superjomn 2022-11-04 16:54:05 +08:00