Commit Graph

Select branches

Hide Pull Requests

aashaka/py_changes

aashaka/test_memreg

aashaka/vllm_changes

anantharamus/bcast-with-recvbuff-exchange

anantharamus/broadcast-amd

anantharamus/broadcast-amd-bcastfix

anantharamus/broadcast-amd-fuse-two-stage

angie/libclang

binyli/GB200

binyli/add-uts

binyli/conn

binyli/fifo

binyli/local-channel

binyli/mnnvl

binyli/nccl-api

binyli/py-lint

binyli/rocm-6.3

binyli/sglang

binyli/signal-wait

binyli/switch-channel

binyli/tile-api

binyli/unique-qp-and-gid-index

caiorocha/4_nodes_allreduce

caiorocha/adding_copilot_skill

caiorocha/auto_barrier

caiorocha/dsl_fuse_op

caiorocha/dsl_instance_tmp

caiorocha/fix_channel_order

caiorocha/fix_er_tol

caiorocha/fix_get_operation_msccllang

caiorocha/ltp

caiorocha/msccllang

caiorocha/multi_node_allgather

caiorocha/nvls_brodcast

caiorocha/op_dag

caiorocha/pilot_test

caiorocha/support_tbg_pipeline

chhwang/2-node-ag

chhwang/alltoall_support

chhwang/atomic-add

chhwang/benchmark

chhwang/blackwell

chhwang/code-coverage

chhwang/comments

chhwang/ctx-default-stream

chhwang/custom-testing-framework

chhwang/docker

chhwang/ib-gather

chhwang/ib-only-tests

chhwang/ibcfg

chhwang/ibv-link

chhwang/ipc-handle

chhwang/local-channel

chhwang/macro

chhwang/minor

chhwang/new-atomic-add

chhwang/rocm-py

chhwang/sema-in-conn

chhwang/switch-chan

chhwang/test

chhwang/tests

chhwang/warnings

copilot/sub-pr-671-again

copilot/sub-pr-698-again

copilot/sub-pr-726

docs

ekow-dev

mahdieh/GB200_nvloptimized

mahdieh/bcast-lowinitoverhead

mahdieh/cuda-arch-variant-autodetect

mahdieh/gb200_1node

mahdieh/mnnvls_support

mahdieh/nvls_multinode

mahdieh/sendrecv-debug

mahdieh/switchchannel_test

main

olli/autosetup-v2

pash-msft/rocm-6.0.0

perf-debug_merge_local_channel_fix_destruction

qinghuazhou/all_to_all_v

qinghuazhou/all_to_all_v_dsl

qinghuazhou/all_to_all_v_dsl_debug_data_validation

qinghuazhou/allreduce_nvls_size_alignment

qinghuazhou/alltoallv_kernel

qinghuazhou/alltoallv_kernel_moe_integration

qinghuazhou/alltoallv_kernel_moe_integration_dynamic_allocation_multinode

qinghuazhou/alltoallv_kernel_moe_integration_dynamic_allocation_multinode_reset

qinghuazhou/alltoallv_kernel_multinode

qinghuazhou/expert_parallel

qinghuazhou/expert_parallel_gb200

qinghuazhou/expert_parallel_ll_opt

qinghuazhou/fifo-test-multi-gpu

qinghuazhou/fifo-test-multi-gpu-write-data

qinghuazhou/group_management

qinghuazhou/local-channel-fifo-test-multi-gpu

qinghuazhou/local-channel_fix_merge_main

qinghuazhou/nccl-rccl-integration-ci-test

qinghuazhou/nccl-rccl-integration-dlopen-force-fallback-ops

qinghuazhou/support-execution-plan-for-channel-cache-disabling

qinghuazhou/test_cuMem

release/v0.6.0

rjsouza/sglang-tests

saemal/py_mscclpp_cleanup

saemal/py_mscclpp_pack128

#1

#10

#100

#101

#102

#103

#104

#106

#107

#108

#109

#11

#110

#111

#112

#113

#114

#116

#117

#118

#119

#12

#122

#123

#124

#125

#13

#131

#132

#136

#14

#141

#142

#145

#145

#146

#147

#15

#150

#159

#16

#162

#166

#167

#169

#17

#170

#171

#172

#178

#179

#18

#180

#181

#182

#183

#184

#185

#186

#188

#189

#19

#190

#195

#196

#197

#2

#20

#200

#201

#202

#203

#204

#208

#209

#21

#211

#213

#214

#217

#218

#219

#22

#220

#221

#222

#223

#225

#229

#23

#230

#231

#232

#233

#234

#235

#236

#237

#237

#238

#239

#24

#240

#241

#243

#244

#245

#246

#247

#248

#249

#25

#250

#251

#252

#255

#256

#257

#258

#259

#26

#261

#262

#264

#267

#268

#27

#270

#271

#272

#273

#275

#276

#277

#278

#279

#28

#280

#283

#284

#286

#29

#290

#291

#292

#293

#294

#295

#296

#297

#298

#299

#3

#30

#300

#301

#303

#304

#305

#306

#307

#308

#309

#310

#312

#313

#314

#316

#317

#318

#319

#32

#321

#322

#324

#325

#326

#327

#328

#329

#33

#330

#331

#332

#333

#334

#336

#337

#338

#339

#34

#340

#341

#342

#343

#344

#345

#346

#347

#348

#35

#351

#352

#353

#354

#355

#356

#357

#358

#36

#360

#363

#364

#365

#366

#366

#367

#369

#37

#370

#371

#371

#372

#373

#374

#375

#376

#378

#379

#38

#380

#381

#382

#383

#384

#385

#386

#389

#39

#390

#391

#392

#393

#396

#399

#4

#40

#400

#401

#402

#403

#404

#405

#406

#407

#409

#41

#410

#411

#412

#413

#414

#415

#416

#417

#419

#42

#420

#421

#422

#423

#424

#425

#426

#427

#428

#429

#43

#430

#433

#434

#435

#436

#437

#44

#441

#442

#443

#444

#445

#445

#446

#447

#448

#449

#45

#450

#451

#452

#453

#455

#456

#457

#458

#459

#46

#460

#461

#462

#463

#464

#464

#465

#466

#468

#469

#47

#471

#472

#474

#475

#476

#478

#479

#48

#481

#483

#484

#485

#486

#488

#489

#49

#490

#491

#492

#497

#498

#499

#5

#50

#500

#501

#502

#505

#506

#507

#508

#509

#51

#510

#511

#512

#513

#514

#515

#518

#519

#52

#520

#521

#522

#523

#524

#525

#526

#527

#528

#529

#53

#530

#532

#534

#537

#538

#539

#54

#541

#543

#544

#545

#546

#547

#549

#55

#550

#551

#553

#555

#557

#558

#559

#56

#560

#561

#562

#563

#564

#565

#566

#567

#568

#569

#57

#570

#571

#571

#572

#573

#573

#574

#575

#576

#577

#578

#578

#579

#58

#580

#581

#582

#583

#584

#585

#586

#587

#588

#589

#59

#590

#591

#592

#593

#594

#595

#596

#596

#597

#598

#599

#6

#60

#600

#601

#602

#603

#604

#605

#606

#607

#608

#609

#61

#610

#611

#612

#613

#614

#615

#616

#617

#618

#619

#62

#620

#621

#624

#625

#626

#627

#627

#628

#629

#629

#63

#630

#631

#632

#633

#634

#636

#637

#638

#639

#64

#640

#641

#642

#643

#644

#645

#646

#647

#648

#650

#651

#652

#653

#655

#656

#659

#66

#660

#661

#662

#663

#664

#666

#667

#668

#669

#67

#670

#671

#673

#676

#677

#679

#68

#680

#681

#683

#684

#685

#686

#687

#688

#689

#69

#690

#691

#692

#693

#694

#696

#697

#697

#698

#699

#7

#70

#700

#701

#702

#703

#704

#706

#707

#708

#709

#71

#710

#711

#712

#715

#716

#717

#718

#719

#72

#720

#721

#722

#723

#724

#725

#726

#727

#728

#729

#73

#730

#731

#732

#733

#734

#735

#736

#737

#738

#739

#74

#740

#741

#742

#743

#744

#745

#746

#747

#748

#749

#75

#750

#751

#752

#753

#754

#755

#756

#757

#758

#759

#76

#760

#761

#762

#762

#763

#764

#765

#766

#768

#769

#77

#77

#770

#771

#771

#773

#774

#775

#776

#777

#778

#779

#78

#780

#781

#782

#783

#783

#784

#784

#785

#786

#787

#789

#789

#79

#790

#792

#793

#794

#794

#795

#796

#796

#797

#798

#798

#799

#799

#8

#80

#81

#82

#83

#84

#85

#86

#87

#88

#9

#93

#94

#95

#96

#97

#98

rocm-v0.7.0-patch

rocm-v0.7.0-patch-1

v0.1.0

v0.2.0

v0.3.0

v0.4.0

v0.4.1

v0.4.2

v0.4.3

v0.5.0

v0.5.1

v0.5.2

v0.6.0

v0.7.0

v0.8.0

v0.9.0

Mono Color

• 00ca2a2051 ext/ep: lift cached_notify nc cap, strided-warp head fixup (HT 442/399 GB/s) qinghuazhou/expert_parallel_gb200 Qinghua Zhou 2026-05-11 05:32:57 +00:00
• ce1554bd1f ext/ep: fix kRDMASender epilogue tail-write race (unblocks chunk_send>16) Qinghua Zhou 2026-05-11 03:40:55 +00:00
• 01a10e00de ext/ep: HT perf - lower lazy head-feedback threshold to chunk/4 Qinghua Zhou 2026-05-10 19:21:25 +00:00
• e0a1bb2c42 ext/ep: WIP Phase 4 fix NVLS self-overcount + cached_notify NVLS barrier Qinghua Zhou 2026-05-10 07:23:36 +00:00
• f2228b07bb ext/ep: WIP Phase 4 fabric-VA cross-node tail/head counter (bypass NVLS multimem.red) Qinghua Zhou 2026-05-10 06:07:46 +00:00
• bf0a7e788a ext/ep: WIP Phase 4 normalize fence/sync ordering in combine writer Qinghua Zhou 2026-05-10 03:58:20 +00:00
• 28f1d722e1 ext/ep: WIP Phase 4 NVLS HT B2 cooperative-copy + relaxed multimem.red Qinghua Zhou 2026-05-10 01:33:14 +00:00
• 4569c4e751 Phase 11: hybrid NVLink + RDMA LL dispatch (+70% throughput) qinghuazhou/expert_parallel_ll_opt Qinghua Zhou 2026-05-09 23:04:15 +00:00
• 5f219b5cda Phase 10: validate vs nccl-ep — prior 30 GB/s gap was apples-to-oranges Qinghua Zhou 2026-05-09 22:22:07 +00:00
• 591fe8272b ext/ep: WIP Phase 4 NVLS HT B2 third multimem barrier in notify_dispatch Qinghua Zhou 2026-05-09 21:39:46 +00:00
• 46701d4161 ext/ep: WIP Phase 4 NVLS HT B2 fabric-IPC data path Qinghua Zhou 2026-05-09 20:59:34 +00:00
• 04f047fc5a Phase 9: multi-NIC striping (NEGATIVE RESULT) Qinghua Zhou 2026-05-09 20:22:00 +00:00
• 3ab2e43b79 ext/ep: NVLS HT B2 phases 1-3 (notify_dispatch barrier + counter fast path) Qinghua Zhou 2026-05-09 19:25:29 +00:00
• 9d729d795e ext/ep: document Phase 8 - combine TMA/cp.async declined after profiling qinghuazhou 2026-05-09 17:56:54 +00:00
• f1808501e9 ext/ep: document Phase 7 attempts (multi-SGE, CUDA Graph, skew) qinghuazhou 2026-05-09 17:32:16 +00:00
• addb6932b8 ext/ep: tune IBGDA LL grid to (1,32) — +6% dispatch / +5% combine qinghuazhou 2026-05-09 05:31:54 +00:00
• 825fc124a5 address hang issue binyli/mnnvl Binyang Li 2026-05-09 03:16:33 +00:00
• 8f2c4e7d98 ext/ep: route LL internode atomics over NVL72 fabric (Proposal A) Qinghua Zhou 2026-05-08 22:04:35 +00:00
• e5ccac520c ext/ep: env-tunable IBGDA channel count (MSCCLPP_EP_IBGDA_CHANNELS) Qinghua Zhou 2026-05-08 21:56:31 +00:00
• b557fe289a ext/ep: split combine 'reduce' bucket into grid_sync + reduce-arith Qinghua Zhou 2026-05-08 21:50:00 +00:00
• f63bf15378 ext/ep: log per-rank IB device + add MSCCLPP_EP_IB_DEVICE_OVERRIDE Qinghua Zhou 2026-05-08 19:30:17 +00:00
• 0e46d3052a ext/ep: per-warp_group dispatch profile slots + sampled readback Qinghua Zhou 2026-05-08 18:38:08 +00:00
• e208cc326b WIP Binyang Li 2026-05-08 04:30:05 +00:00
• 5516bdbb6b fix Binyang Li 2026-05-08 04:22:50 +00:00
• 654bcfa6ba update Binyang Li 2026-05-08 03:54:32 +00:00
• 9ff7e1c2c3 update Binyang Li 2026-05-08 03:43:34 +00:00
• 113d859d13 fix Binyang Li 2026-05-08 03:00:53 +00:00
• 5d16ac958e EP GB200 (4 GPUs/node) support Qinghua Zhou 2026-05-08 01:42:21 +00:00
• fec40601b8 ext/ep: add opt-in in-kernel timestamp profiling for LL dispatch/combine Qinghua Zhou 2026-05-07 21:35:45 +00:00
• 6547d7756c wip caiorocha/fix_er_tol Caio Rocha 2026-05-07 07:10:47 +00:00
• 04ebba7563 ext/ep: GPU-initiated IBGDA path for low-latency dispatch/combine Qinghua Zhou 2026-05-07 05:14:15 +00:00
• d1b04a3b26 NVLS zero-copy allreduce: support FP16 accumulator for FP8 inputs Binyang Li 2026-05-07 00:38:31 +00:00
• 7d80a33360 Default torch example SYMMETRIC_MEMORY env to 1 Binyang Li 2026-05-06 23:43:37 +00:00
• e8caab7c8e Strip preflight validation blocks from NVLS pipeline allreduce kernels Binyang Li 2026-05-06 23:04:41 +00:00
• 639b80de7b Tie AllreduceAllpairPacket maxBlockNum_ to MAX_IPC_DOMAIN_NRANKS - 1 Binyang Li 2026-05-06 22:31:15 +00:00
• 095cfff11d Revert RSAG nBlocks default to 64 Binyang Li 2026-05-06 22:23:18 +00:00
• bde23ce38e Revert verbose RSAG zero-copy comment; rename NRanksPerNode template param Binyang Li 2026-05-06 22:16:08 +00:00
• f0c6ac081f Fold validateIpcDomainSpansWorld into getIpcDomainNranks Binyang Li 2026-05-06 21:49:48 +00:00
• 307a471888 Shorten verbose comments and use THROW in validateIpcDomainSpansWorld Binyang Li 2026-05-06 21:37:09 +00:00
• 4a0d5b29d5 Simplify torch-integration tuning example Binyang Li 2026-05-06 21:14:36 +00:00
• 905b23d9a8 Drop non-MNNVL multi_node regime from torch-integration example Binyang Li 2026-05-06 19:00:22 +00:00
• 9aeeaf0f12 Simplify torch-integration tuning example for MPI-only multi-node testing Binyang Li 2026-05-06 18:51:29 +00:00
• e87c66a85d ext/ep: apply clang-format and black to fix CI lint failures qinghuazhou/expert_parallel Qinghua Zhou 2026-05-06 04:12:20 +00:00
• 01032fa167 core: TODO notes on CUDA-IPC atomicAdd context/flush caveats Qinghua Zhou 2026-05-06 03:44:10 +00:00
• 23e8ce6dbe ext/ep: add pragma once to event.hpp and update validation docs Qinghua Zhou 2026-05-06 03:24:34 +00:00
• c641487c55 ext/ep: fix SWITCH_* macros and add missing standard headers Qinghua Zhou 2026-05-06 03:18:39 +00:00
• 3b96b5ab6e disable flashinfer version rjsouza/sglang-tests empyreus 2026-05-06 03:12:16 +00:00
• b2880652ce ext/ep: remove unused mscclpp_ep CMake target Qinghua Zhou 2026-05-06 03:09:04 +00:00
• 075a43ade7 ext/ep: remove outdated single-rank smoke test Qinghua Zhou 2026-05-06 03:09:04 +00:00
• 5178155be8 ext/ep: add MIT license headers to EP sources and tests Qinghua Zhou 2026-05-06 02:42:49 +00:00
• 89f17dab5b Potential fix for pull request finding Qinghua Zhou 2026-05-05 19:28:37 -07:00
• 89cb62d047 Potential fix for pull request finding Qinghua Zhou 2026-05-05 19:24:28 -07:00
• 1ca7b65db7 host file empyreus 2026-05-05 22:33:37 +00:00
• 783c73b5d9 try to resolve single empyreus 2026-05-05 21:27:30 +00:00
• 1b9f335ddd wip caiorocha/4_nodes_allreduce Caio Rocha 2026-05-05 21:19:23 +00:00
• 822fbb2351 Adding necessary macros for enabling mrc support (#797) main Mahdieh Ghazi 2026-05-05 17:17:41 -04:00
• 47494ea75a check sglang versions empyreus 2026-05-05 21:10:25 +00:00
• 1fe41e05f5 sglang empyreus 2026-05-05 20:17:05 +00:00
• 22a0953a20 fix vmss name empyreus 2026-05-05 20:02:49 +00:00
• f61407ed91 try to install everything not just python empyreus 2026-05-05 18:47:23 +00:00
• 77cb3675b5 fix eof empyreus 2026-05-05 17:57:52 +00:00
• 6296803d87 Make NVLS non-zero-copy allreduce algorithms MNNVL-ready Binyang Li 2026-05-05 04:41:14 +00:00
• 987f80025a Merge remote-tracking branch 'origin/main' into binyli/mnnvl Binyang Li 2026-05-04 23:53:25 +00:00
• 7ca634321a trying without python empyreus 2026-05-04 22:57:37 +00:00
• 3e655e5e02 remove cd empyreus 2026-05-04 22:20:17 +00:00
• 528603856c Merge branch 'main' into caiorocha/4_nodes_allreduce Binyang Li 2026-05-04 15:12:10 -07:00
• 9ec26fa4d1 Reset GPU tokens before reuse (#795) Binyang Li 2026-05-04 15:11:47 -07:00
• 1f3f81b519 make fixes empyreus 2026-05-04 21:20:35 +00:00
• 3f86480097 add container name to deploy empyreus 2026-05-04 20:42:56 +00:00
• 812d43d406 return failed result for new test empyreus 2026-05-04 20:40:12 +00:00
• f6637cc458 attempt to print nvidia-smi for cuda drivers empyreus 2026-05-04 20:11:21 +00:00
• 21197f7c0a change directory empyreus 2026-05-04 20:02:34 +00:00
• dfdc9f701e update pool empyreus 2026-05-04 18:29:54 +00:00
• eaa611f220 split multi node test empyreus 2026-05-04 18:10:42 +00:00
• de244e528b update sglang bench empyreus 2026-05-04 18:04:30 +00:00
• 97a4b1aa69 remove duplicate stop empyreus 2026-05-04 17:23:01 +00:00
• cb430b35d4 clean up deploy empyreus 2026-05-04 17:21:56 +00:00
• a8b959946a Inital new test empyreus 2026-05-04 17:18:02 +00:00
• e091f65143 Merge branch 'main' into rjsouza/sglang-tests empyreus 2026-05-04 17:06:18 +00:00
• b30752a94f review: fix docstring, trailing comma, import placement, and filename mismatch copilot-swe-agent[bot] 2026-05-03 22:51:44 +00:00
• 2a34a7ed11 Merge branch 'main' into caiorocha/4_nodes_allreduce Binyang Li 2026-05-03 15:45:22 -07:00
• 9a36884369 Rename gpuMemset wrapper and zero TokenPool slots in deleter Binyang Li 2026-05-02 03:32:18 +00:00
• 7bc5e0406b Reset GPU tokens before reuse Binyang Li 2026-05-02 03:19:31 +00:00
• 1c29817566 Revert AllreduceRsAgZeroCopy non-symmetric ctx key tag back to ++tag Binyang Li 2026-05-01 23:40:11 +00:00
• 2efda4d819 Restore compile-time templated NRanksPerNode for rsag_zero_copy Binyang Li 2026-05-01 23:09:22 +00:00
• 2a2fca8a58 Rename collective ctx/kernel param nRanksPerNode to ipcDomainNranks Binyang Li 2026-05-01 19:06:07 +00:00
• 45a651b2c8 Decouple IPC-domain hint from bootstrap nRanksPerNode Binyang Li 2026-05-01 18:27:17 +00:00
• fdf7d579dc ext/ep: optional preallocated outputs for low_latency_dispatch Qinghua Zhou 2026-04-30 18:45:44 +00:00
• 2529774868 tests/ep: intranode send-side counts unique (token, dst_node) to match NCCL-EP Qinghua Zhou 2026-04-29 23:31:47 +00:00
• 6ad82e8bbe tests/ep: disable NCCL HeartbeatMonitor to silence mpirun shutdown noise Qinghua Zhou 2026-04-29 20:44:37 +00:00
• e752dbaf97 tests/ep: add NCCL-EP six-metric BW breakdown (send/recv x total/nvl/rdma) Qinghua Zhou 2026-04-29 20:44:10 +00:00
• f2feb120b8 ext/ep: refresh README to reflect current LL, proxy sharding, and bench harness Qinghua Zhou 2026-04-29 18:26:36 +00:00
• 9213587ffe ep tests: report dispatch/combine min, avg, max time and use avg for BW Qinghua Zhou 2026-04-29 16:50:33 +00:00
• afbdcd6a3d ep tests: clean shutdown to silence TCPStore/HeartbeatMonitor noise Qinghua Zhou 2026-04-29 05:16:22 +00:00
• 2c52937b26 Fix FP8 ROCm build/test issues and dtype naming (#792) Binyang Li 2026-04-28 15:02:22 -07:00
• 533f329971 Tune no-sym MNNVL with RSAG zero-copy Binyang Li 2026-04-28 16:23:23 +00:00
• 3bc00cb7f0 Enable NVLS zero-copy without symmetric memory flag Binyang Li 2026-04-28 08:24:49 +00:00
• 865c2bc795 Optimize MNNVL allreduce without symmetric memory Binyang Li 2026-04-28 07:55:52 +00:00
• dded5e0e39 Improve MNNVL allreduce tuning performance Binyang Li 2026-04-28 06:41:17 +00:00
• 893a08e69c Enable MNNVL allreduce tuning Binyang Li 2026-04-28 05:38:59 +00:00