v0.1.3 · tile-ai tilelang · Discussion #265 · GitHub

LeiWang1999
Mar 23, 2025
Maintainer

What's Changed

[Docker] Add libstdcxx-ng-12 to Dockerfiles for CUDA versions by @LeiWang1999 in [Docker] Add libstdcxx-ng-12 to Dockerfiles for CUDA versions #160
Add cpu jit with backend ctypes by @xs-keju in Add cpu jit with backend ctypes #154
[Carver] Multi-Threads Compilation for Fast Auto Tuning by @SiriusNEO in [Carver] Multi-Threads Compilation for Fast Auto Tuning #156
[Refactor] Replace T.If with native Python if statement for mla paged kernel by @LeiWang1999 in [Refactor] Replace T.If with native Python if statement for mla paged kernel #162
[Enhancement] Improve CUDA path detection by @xwhzz in [Enhancement] Improve CUDA path detection #157
[Refactor] Replace T.thread_binding with T.get_thread_binding in examples and test cases by @LeiWang1999 in [Refactor] Replace T.thread_binding with T.get_thread_binding in examples and test cases #163
[Bugfix] Cast bool dtype into int8 in blocksparse examples by @LeiWang1999 in [Bugfix] Cast bool dtype into int8 in blocksparse examples #167
[Example] Implement NSA Decode tilelang exampls by @LeiWang1999 in [Example] Implement NSA Decode tilelang exampls #168
[Release] Bump version to v0.1.2.post1 by @LeiWang1999 in [Release] Bump version to v0.1.2.post1 #166
Use SS-GEMM for PV in mla by @YouJiacheng in Use SS-GEMM for PV in mla #165
[Example] Implement tilelang native sparse attention varlen example by @LeiWang1999 in [Example] Implement tilelang native sparse attention varlen example #170
[Bugfix] Implement boundary check for the buffer shape with dynamic symbolic by @LeiWang1999 in [Bugfix] Implement boundary check for the buffer shape with dynamic symbolic #173
[AutoTune] Enable config-performance trace by @LeiWang1999 in [AutoTune] Enable config-performance trace #174
[Feat] Append Pass Context and TMA lowering configuration option by @LeiWang1999 in [Feat] Append Pass Context and TMA lowering configuration option #175
[Feat] Introduce new caching mechanism for compiled kernels by @LeiWang1999 in [Feat] Introduce new caching mechanism for compiled kernels #176
[Refactor] Enhance GPU Kernel Launch with Environment Thread Creation by @LeiWang1999 in [Refactor] Enhance GPU Kernel Launch with Environment Thread Creation #178
[Bugfix] Improve Thread Variable Handling in Layout Inference by @LeiWang1999 in [Bugfix] Improve Thread Variable Handling in Layout Inference #179
[Examples] Implement NSA Backward kernels by @LeiWang1999 in [Examples] Implement NSA Backward kernels #180
[Enhancement] Optimize CMake build process with dynamic job count calculation by @LeiWang1999 in [Enhancement] Optimize CMake build process with dynamic job count calculation #183
[Bugfix] Add dynamic shape support with out_idx in Cython JIT kernel compilation by @LeiWang1999 in [Bugfix] Add dynamic shape support with out_idx in Cython JIT kernel compilation #185
[Dev][Bugfix] Add RMS Normalization Kernels and Fix Reduce Bug by @chengyupku in [Dev][Bugfix] Add RMS Normalization Kernels and Fix Reduce Bug #188
[Dev] Add the failed nvcc command to the exception message by @penguin-wwy in [Dev] Add the failed nvcc command to the exception message #189
[Bugfix] Fix T.copy for scalar datatypes by @LeiWang1999 in [Bugfix] Fix T.copy for scalar datatypes #190
[Enhancement] Simplify GEMM example with direct kernel compilation by @LeiWang1999 in [Enhancement] Simplify GEMM example with direct kernel compilation #191
[Bugfix] Make quickstart work properly on cu118 by @penguin-wwy in [Bugfix] Make quickstart work properly on cu118 #193
[Language] Support clamp in language by @hyx1999 in [Language] Support clamp in language #192
[Refactor] Add SetMaxNRegCollector to Improve Register Hint Handling in Warp Specialized Rewriter by @chengyupku in [Refactor] Add SetMaxNRegCollector to Improve Register Hint Handling in Warp Specialized Rewriter #194
[Feature] Add TMA Store Synchronization Support by @chengyupku in [Feature] Add TMA Store Synchronization Support #195
Update expired example code. by @66RING in Update expired example code. #196
[CMake] Add CUDA Major Version Detection for Conditional Compilation by @chengyupku in [CMake] Add CUDA Major Version Detection for Conditional Compilation #197
[Feature] Support Async Pipeline inference within if scope by @LeiWang1999 in [Feature] Support Async Pipeline inference within if scope #198
[Dev] Add new example for FlashAttention with pipelined execution by @chengyupku in [Dev] Add new example for FlashAttention with pipelined execution #200
[Enhancement] Enhancing the handling of conditional statements in the pipeline by @LeiWang1999 in [Enhancement] Enhancing the handling of conditional statements in the pipeline #201
[Feature] Upgrade cutlass version and support fp8 T.gemm by @zqh-wz in [Feature] Upgrade cutlass version and support fp8 T.gemm #202
[Docker] Update Dockerfiles to specify exact version of libstdcxx-ng by @LeiWang1999 in [Docker] Update Dockerfiles to specify exact version of libstdcxx-ng #203
[Dev] Add GQA backward example by @chengyupku in [Dev] Add GQA backward example #205
[LICENSE] Typo fix in LICENSE by @LeiWang1999 in [LICENSE] Typo fix in LICENSE #208
[Enhancement] Allow mma fallback when wgmma is not supported by @LeiWang1999 in [Enhancement] Allow mma fallback when wgmma is not supported #206
[Examples] Expand tuning configurations for FlashAttention example by @chenghuaWang in [Examples] Expand tuning configurations for FlashAttention example #204
[Enhancement] Avoid tvm ffi handling when out_idx is specified by @LeiWang1999 in [Enhancement] Avoid tvm ffi handling when out_idx is specified #209
[Fix] Fix K // block_K to T.ceildiv(K,block_K) and add tests by @hyx1999 in [Fix] Fix K // block_K to T.ceildiv(K,block_K) and add tests #210
[Dev] Implement IfStmtBinding and MergeIfStmt transformations by @chengyupku in [Dev] Implement IfStmtBinding and MergeIfStmt transformations #211
[Language] Introduce T.reshape and T.view by @LeiWang1999 in [Language] Introduce T.reshape and T.view #212
[Enhancement] Improve device handling in Cython kernel adapter by @LeiWang1999 in [Enhancement] Improve device handling in Cython kernel adapter #220
[Enhancement] Update format script to support force compare with upstream by @LeiWang1999 in [Enhancement] Update format script to support force compare with upstream #221
[Refactor] Introduce KernelParam integration across modules by @LeiWang1999 in [Refactor] Introduce KernelParam integration across modules #223
[Bugfix] Fix mismatch of shared memory layout and mma atom on Hopper by @zqh-wz in [Bugfix] Fix mismatch of shared memory layout and mma atom on Hopper #224
[Refactor] Update kernel compilation and profiling in examples by @chengyupku in [Refactor] Update kernel compilation and profiling in examples #225
[Examples] Add fp8 gemm 2xAcc and deepgemm example by @cherichy in [Examples] Add fp8 gemm 2xAcc and deepgemm example #217
[Doc] Add instructions for installing nightly version by @xwhzz in [Doc] Add instructions for installing nightly version #226
[Bugfix] Disable force inline for ldmatrix by @LeiWang1999 in [Bugfix] Disable force inline for ldmatrix #227
[Bugfix] Support duplicate tma desc declaration by @LeiWang1999 in [Bugfix] Support duplicate tma desc declaration #228
[Refactor] Rename clamp functions and enhance dtype handling in tests by @LeiWang1999 in [Refactor] Rename clamp functions and enhance dtype handling in tests #232
[Enhancement] Simplify kernel source extraction in JIT adapters by @LeiWang1999 in [Enhancement] Simplify kernel source extraction in JIT adapters #230
[Feature] Add reduce_max corresponding tests by @LeiWang1999 in [Feature] Add reduce_max corresponding tests #236
[BugFix] Fix bug of missing MBarrierExpectTX by @chengyupku in [BugFix] Fix bug of missing MBarrierExpectTX #241
[Refactor] Refactor for Better Layout Conflict Handling by @LeiWang1999 in [Refactor] Refactor for Better Layout Conflict Handling #240
[Refactor] Align torch_assert_close tensor comparison with torch.testing.assert_close by @xwhzz in [Refactor] Align torch_assert_close tensor comparison with torch.testing.assert_close #239
[Dev] Implement FlashAttention3 Backward by @chengyupku in [Dev] Implement FlashAttention3 Backward #244
[BugFix] Fix bug of mismatching dtype in testing by @xwhzz in [BugFix] Fix bug of mismatching dtype in testing #245
[Enhancement] Add zero initialization option to GEMM operations by @chengyupku in [Enhancement] Add zero initialization option to GEMM operations #246
[Enhancement][CUDA] Avoid C7508 for CUDA backend via assigning default value to minBlocksPerMultiprocesor by @cherichy in [Enhancement][CUDA] Avoid C7508 for CUDA backend via assigning default value to minBlocksPerMultiprocesor #248
[Feature] Add database storage for JITKernel cache with Cython and Ctypes adapters by @Alex4210987 in [Feature] Add database storage for JITKernel cache with Cython and Ctypes adapters #213
[Examples] Implement elementwise add kernel by @chenghuaWang in [Examples] Implement elementwise add kernel #219
[Refactor] Phaseout LLVM Dependency by Making it Optional by @LeiWang1999 in [Refactor] Phaseout LLVM Dependency by Making it Optional #247
[Readme] Update Bib Citation Section by @LeiWang1999 in [Readme] Update Bib Citation Section #249
[Enhancement] Support float variable as arguments by @LeiWang1999 in [Enhancement] Support float variable as arguments #250
add autotune to example_gemm.py by @yyttt6 in add autotune to example_gemm.py #252
[Language] Introduce T.alloc_var to define a variable like int var; by @LeiWang1999 in [Language] Introduce T.alloc_var to define a variable like int var; #255
[Example] Implement Kernel Example cumsum by @LeiWang1999 in [Example] Implement Kernel Example cumsum #258
[Refactor] Refactor CUDA post-processing callback registration in TileLang by @LeiWang1999 in [Refactor] Refactor CUDA post-processing callback registration in TileLang #259
[Refactor] Move compilation outside critical section by @YouJiacheng in [Refactor] Move compilation outside critical section #260
[CI] Use auditwheel to generate manylinux wheels by @oraluben in [CI] Use auditwheel to generate manylinux wheels #251
[Bugfix] Fix Benchmark/Example Code for Autotuning by @SiriusNEO in [Bugfix] Fix Benchmark/Example Code for Autotuning #254
[Language] Enhance alias to support blockwise memory load by @LeiWang1999 in [Language] Enhance alias to support blockwise memory load #261
[Bugfix] Fix auto tuning tma handling by @LeiWang1999 in [Bugfix] Fix auto tuning tma handling #263
[Release] Bump version to 0.1.3 by @LeiWang1999 in [Release] Bump version to 0.1.3 #264

New Contributors

@xs-keju made their first contribution in Add cpu jit with backend ctypes #154
@YouJiacheng made their first contribution in Use SS-GEMM for PV in mla #165
@penguin-wwy made their first contribution in [Dev] Add the failed nvcc command to the exception message #189
@hyx1999 made their first contribution in [Language] Support clamp in language #192
@66RING made their first contribution in Update expired example code. #196
@zqh-wz made their first contribution in [Feature] Upgrade cutlass version and support fp8 T.gemm #202
@chenghuaWang made their first contribution in [Examples] Expand tuning configurations for FlashAttention example #204
@cherichy made their first contribution in [Examples] Add fp8 gemm 2xAcc and deepgemm example #217
@Alex4210987 made their first contribution in [Feature] Add database storage for JITKernel cache with Cython and Ctypes adapters #213
@yyttt6 made their first contribution in add autotune to example_gemm.py #252
@oraluben made their first contribution in [CI] Use auditwheel to generate manylinux wheels #251

Full Changelog: v0.1.2...v0.1.3

This discussion was created from the release v0.1.3.

Replies: 0 comments

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment