Google Git
Sign in
opensecura / 3p / openxla / iree / 7f14078bf688bc720095271420cb136df4ca639a / compiler / src
  1. 9731fed Pass to block dynamic dimensions of operands of `iree_linalg_ext.attention`. (#18874) by MaheshRavishankar · 5 months ago
  2. 03c744e [GPU] Support multiple contraction dims in MmaSchedules (#18720) by Max191 · 5 months ago
  3. 0c2c627 [NFC] Update old naming from flow to dispatch creation (#18904) by Ian Wood · 5 months ago
  4. 55c5562 [LLVMGPU][NFC] Create LLVMGPU pass for IGEMM (#18871) by Max191 · 5 months ago
  5. c6b3592 [Dispatch Creation] Bubble up ExtractSliceOp with FillOp when the latter has multiple consumers (#18896) by Nithin Meganathan · 5 months ago
  6. 1aa5825 [LLVMGPU] Combine parallel and reduction padding in LLVMGPUPadAndVectorDistribute (#18771) by Kunwar Grover · 5 months ago
  7. 1fc6e5b Add CDNA3 MFMA BF16 intrinsics. (#18892) by Benoit Jacob · 5 months ago
  8. 3b751a4 [LLVMCPU] Enable tileDispatchUsingForall as default (#18777) by Prashant Kumar · 5 months ago
  9. e96e3c0 [VectorLayout] Fix insertion of new constOp for non dominate issue. (#18894) by Stanley Winata · 5 months ago
  10. aef6e1f [GPU] Bail out in GPUReduceBankConflicts if we have collapse_shape user (#18863) by Nirvedh Meshram · 5 months ago
  11. 4ad834b Support F8E5M2FNUZ MFMA on CDNA3 (#18887) by Benoit Jacob · 5 months ago
  12. 2291b38 Support 8-bit floats in the compiler. (#18886) by Benoit Jacob · 5 months ago
  13. abe3f89 Add conversions for 1x1 conv_2d to matmul (#18736) by Ian Wood · 5 months ago
  14. c3fae2f [LLVMGPU] Use forall workgroup distribution in TileAndFuse pipeline (#18565) by Max191 · 5 months ago
  15. 4d20b82 Emit an error when affinity analysis fails. (#18883) by Ben Vanik · 5 months ago
  16. 9f5610d Preserving `nosideeffects` on func.func -> util.func import. (#18882) by Ben Vanik · 5 months ago
  17. e1469b2 [Codegen] Add pass to decompose pack unpack ops at dispatch boundaries (#18852) by Max191 · 5 months ago
  18. 00104b5 Allow dynamic dimensions during folding of `tensor.expand_shape/collapse_shape` into `flow.dispatch.tensor.load/store`. (#18873) by MaheshRavishankar · 5 months ago
  19. 81c8b25 [Codegen] Allow multiple reduction dimensions in VectorDistribution (#18868) by Kunwar Grover · 5 months ago
  20. b922a70 GPU data tiling: query the target's list of MMA intrinsics. Add FP8 test. (#18862) by Benoit Jacob · 5 months ago
  21. bb71f7d [Attention] Only clamp attention for low precision types (#18848) by Kunwar Grover · 5 months ago
  22. 4cc6671 [CPU] Limit vectorization tile sizes for SVE (#18846) by Cullen Rhodes · 5 months ago
  23. d0269f3 Integrate llvm-project @864902e9b4d8bc6d3f0852d5c475e3dc97dd8335 (#18843) by Max191 · 5 months ago
  24. c08362a GPU target parameters for data tiling. (#18839) by Benoit Jacob · 5 months ago
  25. 114a142 [LLVMGPU] Embed mma_intrinsic in to_layout and infer contraction's intrinsic from it. (#18842) by Stanley Winata · 5 months ago
  26. 66342ab Reland #18804 (#18840) by Maksim Levental · 5 months ago
  27. 556c945 [Codegen] Fix bug in IGEMM pass for non conv contractions (#18838) by Max191 · 5 months ago
  28. df5e5aa Revert "`iree_gpu` Python bindings (`GPUPipelineOptionsAttr`)" (#18833) by Scott Todd · 5 months ago
  29. 5546ae0 Run `compiler/src/iree/compiler/API/generate_exports.py`. (#18832) by Scott Todd · 5 months ago
  30. 012f8a6 [Codegen] Drop TransformStrategies (#18820) by Quinn Dawkins · 5 months ago
  31. fb18c42 `iree_gpu` Python bindings (`GPUPipelineOptionsAttr`) (#18804) by Maksim Levental · 5 months ago
  32. 4b32a1d [TileSwizzle] Make the dump and variable name match. (NFC) (#18821) by Han-Chung Wang · 5 months ago
  33. fecccdc [Flow] Fix FoldSplatReshapeIntoSplat pattern (#18818) by Ian Wood · 5 months ago
  34. 05bbcf1 Enable arithmetic optimizations as part of the stream simplification pipeline. (#18806) by Stella Laurenzo · 6 months ago
  35. 929a7da Fixes a range inference overflow with util.align. (#18808) by Stella Laurenzo · 6 months ago
  36. 8da6ba2 Add util.assume.int folder. (#18805) by Stella Laurenzo · 6 months ago
  37. 1500641 Various tweaks to numeric optimizations found while looking at programs. (#18765) by Stella Laurenzo · 6 months ago
  38. e2a352c Integrates LLVM @ 36d936a2d057ddbd7822614edf01e39a0c21d654 (#18801) by Vivian · 6 months ago
  39. a488d38 Add region to `linalg_ext.attention` (#18728) by Rob Suderman · 6 months ago
  40. 206b60c [DispatchCreation] Extend multi-use producer fusion (#18551) by Ian Wood · 6 months ago
  41. f9fa934 Integrates LLVM @ a758bcdbd92efb64a3482eb95d2769d74e33f5bb (#18783) by Vivian · 6 months ago
  42. 2945399 Revert `tensor.cast` to flow reshape conversion (#18772) by Ian Wood · 6 months ago
  43. 9f93073 [CMake] Don't update compile definitions for imported targets for MSCV (#18766) by Maksim Levental · 6 months ago
  44. 3ccd4f1 [Codegen] Replace LICM with a version that checks trip count (#18679) by Quinn Dawkins · 6 months ago
  45. afe18d2 [Codegen][GPU] Add tiling cleanup pattern to fuse pad without zero gaurd (#18748) by Kunwar Grover · 6 months ago
  46. 7622770 Integrate LLVM @ 7900daaa7ba57b5f9729bbbdb54f4e0599a45cd7 (#18773) by Vivian · 6 months ago
  47. ab88871 Add a default lowering config setting for custom_op. (#18737) by MaheshRavishankar · 6 months ago
  48. 6d8f0b2 [LLVMGPU] Add configuration tests for IGEMM, fix NCHW case (#18734) by Max191 · 6 months ago
  49. 2eddaea [IGEMM] Generate matmuls with expanded H and W dims (#18735) by Max191 · 6 months ago
  50. c3d74d0 [Flow] Fold flow reshape with mismatching dyn dims (#18680) by Ian Wood · 6 months ago
  51. 3e52344 [LLVMCPU] Enable tileDispatchUsingForall for multiTilingExpert (#18730) by Prashant Kumar · 6 months ago
  52. 1e6bbb8 [LinalgExt] Remove default implementation for getStaticLoopRanges (#18745) by Kunwar Grover · 6 months ago
  53. d7378bb [Codegen][GPU] Add pass for fallback distribution (#18726) by Quinn Dawkins · 6 months ago
  54. 1e155cc Use integer range and divisibility analysis to propagate int assumptions into dispatch executables. (#18755) by Stella Laurenzo · 6 months ago
  55. 0c69c9e Adding `iree_vm_context_fork` to fork a context. (#18751) by Ben Vanik · 6 months ago
  56. 25e5760 [Codegen] Remove unused arguements from attention op (#18743) by Kunwar Grover · 6 months ago
  57. 726112d [Codegen] Remove memref optimizations from OptimizeTensorInsertExtractSlices (#18732) by Kunwar Grover · 6 months ago
  58. 660f388 Enable analysis based integer optimizations. (#18756) by Stella Laurenzo · 6 months ago
  59. 90a2e1c [NFC][GPU] remove some dead code in setTileAndFuseLoweringConfig (#18754) by Nirvedh Meshram · 6 months ago
  60. ea52c36 Adding `IREE_STATUS_INCOMPATIBLE`. (#18752) by Ben Vanik · 6 months ago
  61. dd3f2a3 [Codegen] Remove old attention transformations (#18740) by Kunwar Grover · 6 months ago
  62. 5f3f863 Add `AggregatedOpInterface` to `iree_linalg_ext.custom_op` (#18700) by MaheshRavishankar · 6 months ago
  63. 598a60e [Im2col] Allow multiple batch, M, and K dimensions on im2col result (#18593) by Max191 · 6 months ago
  64. ce4f098 [Codegen] Use constant int matchers for interface reshape foldings (#18733) by Max191 · 6 months ago
  65. 01b9521 Add `TilingInterface` implementation of `iree_linalg_ext.custom_op`. (#18676) by MaheshRavishankar · 6 months ago
  66. 2963600 Integrate LLVM at f59b151f by Alan Li · 6 months ago
  67. 5270093 Add an integer divisibility analysis. (#18727) by Stella Laurenzo · 6 months ago
  68. 1b719b3 [Codegen][GPU] Do not promote fill matmul operands (#18722) by Quinn Dawkins · 6 months ago
  69. 0e16a89 [Codegen][GPU] Disable consumer fusion for multi use cases (#18723) by Quinn Dawkins · 6 months ago
  70. 0f28d44 Use upstream dataflow tooling to build an arithmetic opt pass. (#18702) by Stella Laurenzo · 6 months ago
  71. 0889d13 Add the support to yield multiple results (#18717) by Prashant Kumar · 6 months ago
  72. 7fb28e0 [Codegen][GPU] Fix delinearized index order in forall resolution (#18724) by Quinn Dawkins · 6 months ago
  73. 4636257 [Encoding][NFC] Moving Encoding attr/enum to Encoding[Types|Attrs].* (#18711) by Han-Chung Wang · 6 months ago
  74. e8ff07e [Codegen][GPU] Stop using redundant transfer hoisting pattern on memrefs (#18721) by Quinn Dawkins · 6 months ago
  75. 6c095ed Revert "Revert "[CPU] Enable tileDispatchUsingForall for mmt4d and co… (#18718) by Prashant Kumar · 6 months ago
  76. e8ab5f8 Integrate LLVM at 634c57d7 by Alan Li · 6 months ago
  77. f6a9b6b [Codegen][GPU] Enable destination fusion for unit trip loops (#18674) by Quinn Dawkins · 6 months ago
  78. ad68964 [LLVMGPU] Pad to intrinsic shape in LLVMGPUPadAndVectorDistribute pipeline (#18632) by Kunwar Grover · 6 months ago
  79. 6001f9c Fix distribution logic when number of parallel loops is greater than 3 (#18714) by MaheshRavishankar · 6 months ago
  80. 5b0680d [Codegen] Add option to disable copy vectorization (#18673) by Quinn Dawkins · 6 months ago
  81. cc3b28f [Codegen][GPU] Improve loop fusion pattern verification (#18671) by Quinn Dawkins · 6 months ago
  82. 62f8d7e Revert "[CPU] Enable tileDispatchUsingForall for mmt4d and convolution pipelines. " (#18707) by Han-Chung Wang · 6 months ago
  83. 65158ac Rework util.assume.* ops to util.assume.int and base on attributes. (#18703) by Stella Laurenzo · 6 months ago
  84. caacf6c Honor narrow dims in GPUMaterializeEncoding (#18622) by Benoit Jacob · 6 months ago
  85. bb5f2f5 [CPU] Enable tileDispatchUsingForall for mmt4d and convolution pipelines. (#18618) by Prashant Kumar · 6 months ago
  86. 126e334 [CPU][NFC] Collapsing methods that query a config is enabled. (#18686) by Han-Chung Wang · 6 months ago
  87. b89ba05 [CPU] Switching to linalg::LinalgOp for MaterializeContractionOp pattern (#18690) by Han-Chung Wang · 6 months ago
  88. 3801a5d [GPU][DT] Add support for materializing gemm with batch dimensions. (#18693) by Han-Chung Wang · 6 months ago
  89. 58dac26 [Encoding] Introduce getElementTypesArray helper in EncodingAttr. (#18692) by Han-Chung Wang · 6 months ago
  90. f7a02a0 [DT] Use createOrFold to simplify the logic of creating tensor.cast op. (#18691) by Han-Chung Wang · 6 months ago
  91. 067ba0e [GPU][NFC] Updates comments/style/TODO/debug_message for GPU data-tiling (#18688) by Han-Chung Wang · 6 months ago
  92. 79e979f Legalize newly added compiler hints ops. (#18689) by Stella Laurenzo · 6 months ago
  93. a9f5628 Limit transposition of narrow-N to only CPUMaterializeEncoding (#18684) by Benoit Jacob · 6 months ago
  94. 747081a [Codegen][GPU] Force linalg ops to always use the same vector size (#18654) by Quinn Dawkins · 6 months ago
  95. 0e8a573 Adding vm.cast.f32.si64 and vm.cast.f32.ui64 ops (#18642) by Nirvedh Meshram · 6 months ago
  96. 0bbc432 [VectorDistribution]Add distribution pattern and test mlir file for vector.gather (#18660) by Bangtian Liu · 6 months ago
  97. 88153eb [Codegen][GPU] Fuse into destinations for parallel tiling (#18666) by Quinn Dawkins · 6 months ago
  98. 24ee841 [Codegen][GPU] Fix alloc creation for dynamic outputs in loop fusion (#18665) by Quinn Dawkins · 6 months ago
  99. 206c1f2 [Codegen] Allow vectorizing linalg.copy ops on memrefs (#18672) by Quinn Dawkins · 6 months ago
  100. 903ab0a Integrate LLVM at 9fa55ec3 (#18670) by Benoit Jacob · 6 months ago