leavelet/cutlass
1
0
Fork 0
Code Issues Pull Requests Actions 1 Packages Projects Releases Wiki Activity

Commit Graph

  • fde824af21 Update Hopper performance plot for CUTLASS 3.1 + CTK 12.1 (#967) Vijay Thakkar 2023-06-01 11:52:40 -07:00
  • 7dbf423763 Add conversion from ElementBias to ElementCompute (#961) Jack Kosaian 2023-05-26 23:08:36 -04:00
  • 6f47420213 Update README.md Haicheng Wu 2023-05-24 12:40:31 -04:00
  • 4638250469 Update CHANGELOG.md Haicheng Wu 2023-05-24 12:39:42 -04:00
  • 7859fe322a Update PUBLICATIONS.md Haicheng Wu 2023-05-24 12:36:12 -04:00
  • d3e72719b4 Add support for sparse GEMM with row broadcasted bias vector (#951) Aleksandar Samardžić 2023-05-24 16:25:05 +02:00
  • b4ab501767 Adds CUDA path for x86-64 (#957) Ali Hassani 2023-05-24 07:21:25 -07:00
  • f079619f5e More updates for 3.1 (#958) ANIKET SHIVAM 2023-05-24 07:17:16 -07:00
  • 13f413493a Stream-K with broadcast (#892) Ali Hassani 2023-05-22 16:05:06 -07:00
  • 6fbc0d3380 Update layout.md Haicheng Wu 2023-05-17 20:12:58 -04:00
  • b97404837e Adding 128x256 tile for 16b input datatype WGMMA gemm (#950) Manish Gupta 2023-05-17 14:13:23 -07:00
  • e2953d47c5 Update gemm_api.md Haicheng Wu 2023-05-12 15:37:31 -04:00
  • 19c4a4815e replace division with multiplication in GELU (#942) wll 2023-05-12 22:57:18 +08:00
  • fcfbd23e26 Fix host compilation of cute::cast_smem_ptr_to_uint. (#940) Gregory Meyer (gregjm) 2023-05-09 21:06:54 -07:00
  • b250faccd3 Make operator() const-correct and add missing static functions. (#936) Gregory Meyer (gregjm) 2023-05-09 13:33:01 -07:00
  • 24c8b7d8a2 Fix cuTE compilation with clang (#939) Janusz Lisiecki 2023-05-09 15:51:45 +02:00
  • 7c04f95415 Updates for 3.1 (#932) ANIKET SHIVAM 2023-04-29 06:34:27 -07:00
  • 6f8596ce3f Add missing #include directive to get access to cutlass::epilogue::thread::ScaleType. (#925) Gregory Meyer (gregjm) 2023-04-28 17:02:41 -07:00
  • fe2f491dd7 Get SM count with cudaDeviceGetAttribute in KernelHardwareInfo (#927) Adnan Akhundov 2023-04-28 19:23:23 +02:00
  • df02482f1d Add missing schedules argument in SM90 fp16 op generation (#920) Adnan Akhundov 2023-04-26 22:44:49 +02:00
  • 180c5629bf Add missing checks for NVRTC in CuTe (#921) Jakub Szuppe 2023-04-25 18:52:43 +02:00
  • e36912f961 Fix for dangling references in the MHA example (#918) Alexander Zinoviev 2023-04-19 18:35:46 -07:00
  • 9a83bd3381 CUTLASS 3.1 Python interface documentation (#917) Jack Kosaian 2023-04-18 15:11:35 -04:00
  • 54bebe417d Fix some typos in CuTe tutorials (#912) Adnan Akhundov 2023-04-17 22:00:51 +02:00
  • 43cfbe0086 Allow L2 prefect for clang compiler (#914) Guray Ozen 2023-04-15 07:23:22 +02:00
  • 4a68cf748e added support of b2b bmm (#849) Aleksandr Pivovar 2023-04-15 05:20:02 +02:00
  • d572cc1aab CUTLASS 3.1 (#915) ANIKET SHIVAM 2023-04-14 20:19:34 -07:00
  • 9b8166e3f0 fMHA: Add backward pass (#844) dan_the_3rd 2023-04-07 02:44:58 +02:00
  • e2d439ee7e Add tile_n=32 and tile_k=32 kernels in generator.py (#858) Shuai Shao 2023-04-06 07:00:52 -07:00
  • 0435979f59 Remove const from 3.x GemmUniversalAdapter::operator() (#905) Adnan Akhundov 2023-04-04 02:30:51 +02:00
  • 2ba1ef10be Increase max dynamic SMEM size in GemmSoftmax (#903) Adnan Akhundov 2023-04-03 16:01:12 +02:00
  • 0964bdb64c update gemm and conv2d cmdline --help output (#878) Adnios 2023-04-01 23:38:13 +08:00
  • ecbd24566c Enable shared memory intrinsics and ldmatrix PTX on Clang. (#754) Gregory Meyer (gregjm) 2023-03-31 18:42:24 -07:00
  • 660a05f581 fix split_k_mode and add reduction kernel for f16 input/accum/output (#896) Manish Gupta 2023-03-30 12:31:08 -07:00
  • bc36122c3f [layout] Fix AffineRank2ColumnMajor::packed() (#879) Feng Shijie 2023-03-29 23:59:48 +08:00
  • 15d9d31f1f CUTLASS 3.0 Hopper GEMMs are GETTs in disguise (#897) Vijay Thakkar 2023-03-29 07:42:40 -07:00
  • 1eef5c3cf1 add guards for __CUDA_ARCH__ >= 530 (#891) ptrblck 2023-03-28 14:47:10 -07:00
  • 87070b6d51 add a CUTLASS publication (#893) Yujia Zhai 2023-03-28 14:06:57 -07:00
  • 77549ae6c8 Update PUBLICATIONS.md Haicheng Wu 2023-03-25 21:17:05 -04:00
  • 42290f5d1c Fix for dangling pointers (#885) Alexander Zinoviev 2023-03-24 22:15:14 -07:00
  • 209faf7b94 remove spurious comma (#871) Vijay Thakkar 2023-03-20 14:25:27 -07:00
  • 6116706c96 Set batch_strides on Params::update (#883) Jack Kosaian 2023-03-20 17:07:47 -04:00
  • 2670b973dd Fix sign-compare warning in reorder_array (#869) Nikita Shulga 2023-03-20 14:07:24 -07:00
  • af332d4aa9 Add missing comma in cutlass/arch/mma_sm90.h (#862) Vijay Thakkar 2023-03-14 09:04:28 -07:00
  • 86cae03cea expose StoreT parameter for potential speed (#838) Edward Rees 2023-03-10 17:58:17 +00:00
  • 29801e348a Hide streams and typinfo from nvrtc (#853) Stepan Tezyunichev 2023-03-10 07:24:47 +03:00
  • 7e370c9637 Fix typos 2 (#842) Alexander Pivovarov 2023-03-09 20:22:56 -08:00
  • c4f6b8c6bc Updates for 3.0 (#857) ANIKET SHIVAM 2023-03-09 12:27:40 -08:00
  • a68e2f95f0 Reduce versbosity in manifest.py (#845) Yinghai Lu 2023-03-07 08:53:01 -08:00
  • a31b43b3f3 Re-enable aarch64 support lost in 277bd6e537 (#846) psaab 2023-03-02 08:17:21 -08:00
  • f396cdd15c ex24[gemm_grouped]: Allow to change layout/dtype (#841) dan_the_3rd 2023-03-01 13:13:51 +01:00
  • 92ebbf1dc4 Fix typos (#839) Alexander Pivovarov 2023-02-27 08:17:58 -08:00
  • 65688c2a87 streamk fix (#836) Haicheng Wu 2023-02-23 16:35:08 -05:00
  • f303889ed9 fMHA: Sync FW with xFormers (#828) dan_the_3rd 2023-02-23 05:25:31 +01:00
  • 9cdbe33570 Add fixed_channel and few_channel mode to int8 in generator (#829) Shuai Shao 2023-02-21 18:15:39 -08:00
  • 95f673ecf7 Update base_grouped.h (#832) Yuxin Wu 2023-02-21 11:48:30 -08:00
  • 91b8de8d32 streamk fix (#830) Haicheng Wu 2023-02-20 11:03:16 -05:00
  • d8359c804b Changes to iterators to support s8 gemm with f16 outputs (#812) Sujan Kumar Gonugondla 2023-02-16 18:37:51 -05:00
  • 34bed24af3 Update helper.h Haicheng Wu 2023-02-16 16:50:04 -05:00
  • a101ac283f Fix some typos (#791) ZZK 2023-02-17 04:56:55 +08:00
  • 9fb38ac048 fix alignmentC=8 for imma N=128 (#822) Haicheng Wu 2023-02-15 12:06:00 -05:00
  • 8f5c242426 Update dual_gemm_common.h Haicheng Wu 2023-02-13 15:35:33 -05:00
  • 3c995c7606 Extend DualGemm: support batched mode + decouple B0/B1 layouts (#790) Adnan Akhundov 2023-02-13 21:27:13 +01:00
  • ce8597dc14 Fix type bug in conv2d/gemm with broadcast (#796) Shuai Shao 2023-02-09 17:53:25 -08:00
  • 2e10404d26 xFormer updates to fMHA FW (#773) dan_the_3rd 2023-02-09 05:00:10 +01:00
  • 5ff5209ed5 Add acc2smem in epilogue/threadblock/epilogue.h (#806) Jack Kosaian 2023-02-06 22:04:16 -05:00
  • 5921043981 Re-enable all alignments for int accumulators (#807) Jack Kosaian 2023-02-06 22:01:15 -05:00
  • add4ba622f Fix 8.4 + CUDA 11.4 build (#789) Mark Hoemmen 2023-01-27 07:18:59 -07:00
  • 277bd6e537 CUTLASS 3.0.0 (#786) Vijay Thakkar 2023-01-23 17:55:28 -08:00
  • 66d9cddc83 New updates for 2.11 (#775) ANIKET SHIVAM 2023-01-20 13:32:57 -08:00
  • d49bef88f9 Enable aarch64 support (#779) psaab 2023-01-20 12:51:58 -08:00
  • 8b42e751c6 streamk paper link (#765) Haicheng Wu 2023-01-10 22:10:43 -05:00
  • eb7f99d3dd @hwu36 Adding the individual arXiv link for Stream-K paper. (#764) Muhammad Osama 2023-01-10 17:39:06 -08:00
  • 764b840d6f streamk example and performance tuning (#760) Haicheng Wu 2023-01-10 16:10:02 -05:00
  • a1046d49c1 Adds missing semicolon (#759) Ali Hassani 2023-01-09 18:50:46 -08:00
  • 1cd994b4cf Update PUBLICATIONS.md Haicheng Wu 2023-01-08 00:42:19 -05:00
  • 7bdba07310 Add definitions for tag structs. (#752) Gregory Meyer (gregjm) 2023-01-06 06:46:52 -08:00
  • c54ede3a9e Add const overloads for iterator functions. (#753) Gregory Meyer (gregjm) 2023-01-06 06:46:34 -08:00
  • ff6e733fe1 restore the old epilogue for everything except streamk (#749) Haicheng Wu 2023-01-04 11:02:55 -05:00
  • 5989b7e1d7 Update PUBLICATIONS.md Haicheng Wu 2023-01-04 09:18:38 -05:00
  • 1e64f153b3 improve streamk load balance (#743) Haicheng Wu 2022-12-25 13:56:33 -05:00
  • 78b30d3191 Update README.md Matthew Nicely 2022-12-21 11:58:19 -05:00
  • 59de82688b Update README.md Matthew Nicely 2022-12-21 11:57:55 -05:00
  • b85865d1ad Add missing #include directives (#741) Gregory Meyer (gregjm) 2022-12-21 08:40:20 -08:00
  • 3f2bb17722 minor chagnes (#730) Haicheng Wu 2022-12-10 14:44:53 -05:00
  • 38193d76e3 Updates for stream-k (#728) ANIKET SHIVAM 2022-12-08 20:48:10 -08:00
  • 1d7772f218 Add missing #include directive (#727) Gregory Meyer (gregjm) 2022-12-08 15:58:31 -08:00
  • df81d847d7 Make Python interface work for non-SM80 targets (#726) Jack Kosaian 2022-12-07 21:53:33 -05:00
  • d6117ca362 Relax stream K gemm alignment constraints (#717) Mike Iovine 2022-12-07 11:17:49 -05:00
  • 9c0518608e Fix typos in conv problem sizes (#720) Ali Hassani 2022-12-05 12:54:58 -08:00
  • 9f1f37aa21 misc (#719) Haicheng Wu 2022-12-05 12:07:20 -05:00
  • 84213b0b8e fix: make arch.h self contained (#714) Wenzhuo Liu 2022-12-02 08:25:48 +08:00
  • 8567b87d65 Update quickstart.md (#704) tpoisonooo 2022-11-30 10:43:03 +08:00
  • c975e2ccbb releaase 2.11 (#703) Aditya Atluri 2022-11-19 06:02:15 -08:00
  • 3c90f6aea6 add #pragma once for header file in example 42 (#698) Wenzhuo Liu 2022-11-16 11:50:24 +08:00
  • 06eb90cc0d Fix identity sigmoid activation (#659) seventh 2022-11-10 03:42:23 +08:00
  • 168ea8b0e1 ensure singleton::get thread safe construct instance (#658) seventh 2022-11-09 10:44:32 +08:00
  • 012c62c748 bug fixes and enharcement to gemm reductionK fusion (#682) Haicheng Wu 2022-11-03 11:07:50 -04:00
  • cc85b64cf6 fix typo (#677) FZC 2022-11-02 02:07:33 +08:00
  • 1b4e24470a Example 43 - DualGemm (#670) dan_the_3rd 2022-10-26 20:04:42 +02:00