Commits

Breadcrumbs

History for

onmain

User selector

All users

Datepicker

All time

Commit History

Commits on May 21, 2023

• NFC: Use fully quailified mlir::Value type in interface declaration (iree-org#13716)

  ezhulenev

  authored
  76e36ff

  Copy full SHA for 76e36ff

Commits on May 20, 2023

• Revert "[CPU] Enable 'iree-llvmcpu-reassociate-fp-reductions' by default" (iree-org#13713)

  Show description for 40a68af
  dcaballe

  authored
  40a68af

  Copy full SHA for 40a68af

• Make opsets really optional in auto pipeline (iree-org#13712)

  Show description for 3c111c3
  jpienaar

  authored
  3c111c3

  Copy full SHA for 3c111c3

• Improve reduction code generation (iree-org#13688)

  Show description for 3ceb80f
  harsh-nod

  authored
  3ceb80f

  Copy full SHA for 3ceb80f

Commits on May 19, 2023

• NFC: Use fully qualified mlir::ValueRange type in interface declaration (iree-org#13695)

  Show description for 7a97d4f
  ezhulenev

  authored
  7a97d4f

  Copy full SHA for 7a97d4f

• CPU ukernels as bitcode (x86-only for now) (iree-org#13460)

  Show description for 29647b3
  

  MaheshRavishankar

  and

  bjacob

  authored
  29647b3

  Copy full SHA for 29647b3

• [CPU] Enable 'iree-llvmcpu-reassociate-fp-reductions' by default (iree-org#13685)

  Show description for 55a2780
  dcaballe

  authored
  55a2780

  Copy full SHA for 55a2780

• Adding util.cast/!util.object and lowering to vm.cast.* ops. (iree-org#13687)

  Show description for 90ed2d0
  benvanik

  authored
  90ed2d0

  Copy full SHA for 90ed2d0

• Add support for two matmuls in layout analysis (iree-org#13664)

  Show description for aa9d4f6
  harsh-nod

  authored
  aa9d4f6

  Copy full SHA for aa9d4f6

• Only apply WMMA unrolling to elementwise ops of sufficient rank (iree-org#13627)

  Show description for 4795765
  qedawkins

  authored
  4795765

  Copy full SHA for 4795765

Commits on May 18, 2023

• [spirv] Use GPUCheckResourceUsage to check shared memory usage (iree-org#13684)

  Show description for 5824221
  antiagainst

  authored
  5824221

  Copy full SHA for 5824221

• NFC: Move LLVMGPU resource usage check pass to Common for reuse (iree-org#13679)

  antiagainst

  authored
  7216615

  Copy full SHA for 7216615

• HLO input support for iree.abi.output and iree.abi.encoding attrs. (iree-org#13678)

  Show description for 74b4df9
  benvanik

  authored
  74b4df9

  Copy full SHA for 74b4df9

• Enable gather vectorization in LLVMGPU backend iree-org#12737 (iree-org#13358)

  pavanimajety

  authored
  f4e77e5

  Copy full SHA for f4e77e5

• Adding iree.abi.encoding for specifying buffer view encoding. (iree-org#13676)

  Show description for dedade3
  benvanik

  authored
  dedade3

  Copy full SHA for dedade3

• NFC: Move common GPU passes under Common/GPU/ directory (iree-org#13672)

  Show description for 5663e58
  antiagainst

  authored
  5663e58

  Copy full SHA for 5663e58

• [CPU] Move 'vector.shape_cast' lowering to the end of LLVMCPUVectorLowering (iree-org#13629)

  Show description for e7d3346
  dcaballe

  authored
  e7d3346

  Copy full SHA for e7d3346

• Re-enable vectorization of gathers (iree-org#13365)

  Show description for 0d9f457
  banach-space

  authored
  0d9f457

  Copy full SHA for 0d9f457

• Adding some basic channel op folding. (iree-org#13654)

  benvanik

  authored
  098764e

  Copy full SHA for 098764e

Commits on May 17, 2023

• [SPIRV] Unroll vector.gather to native vector sizes before lowering (iree-org#13670)

  Show description for 66a286f
  qedawkins

  authored
  66a286f

  Copy full SHA for 66a286f

• Disabling detensoring by default until iree-org#6948 can be completed. (iree-org#13658)

  Show description for 9ba88a2
  benvanik

  authored
  9ba88a2

  Copy full SHA for 9ba88a2

• Remove arith.bitcast from being always legal (iree-org#13667)

  Show description for 114e5cc
  rsuderman

  authored
  114e5cc

  Copy full SHA for 114e5cc

• NFC: Move LLVMGPUDistribute pass to Common/ (iree-org#13663)

  Show description for 1b47584
  antiagainst

  authored
  1b47584

  Copy full SHA for 1b47584

• NFC: Move two LLVMGPU passes to Common/ for sharing with SPIR-V (iree-org#13620)

  Show description for 01c1c12
  antiagainst

  authored
  01c1c12

  Copy full SHA for 01c1c12

• Support mhlo.collective_permute with NCCL (iree-org#13502)

  Show description for 7016b8c
  trevor-m

  authored
  7016b8c

  Copy full SHA for 7016b8c

• e2e matmul test improvements (iree-org#13657)

  Show description for a8daf0e
  bjacob

  authored
  a8daf0e

  Copy full SHA for a8daf0e

• Bucket before deduping executables. (iree-org#13652)

  Show description for 72c8f56
  jpienaar

  authored
  72c8f56

  Copy full SHA for 72c8f56

• Split Passes.h into 6 per target devices (NFC) (iree-org#13591)

  Show description for bb7e776
  grypp

  authored
  bb7e776

  Copy full SHA for bb7e776

Commits on May 16, 2023

• Remove bf16 buffer operations for all backends (iree-org#13607)

  Show description for 6cf092d
  rsuderman

  authored
  6cf092d

  Copy full SHA for 6cf092d

• [spirv] Enable i8 dot product for 2-D normal convolution (iree-org#13116)

  Show description for fc3b764
  antiagainst

  authored
  fc3b764

  Copy full SHA for fc3b764

• [LLVMCPU] Enable shuffle_16x16 for pack codegen. (iree-org#13318)

  Show description for c1b424c
  hanhanW

  authored
  c1b424c

  Copy full SHA for c1b424c

• [spirv] Enable generating vector3 compute ops during vectorization (iree-org#13642)

  Show description for 5b3bbba
  antiagainst

  authored
  5b3bbba

  Copy full SHA for 5b3bbba

• [spirv] Better calculate shared memory usage for cooperative matrix (iree-org#13609)

  Show description for 70f76fc
  antiagainst

  authored
  70f76fc

  Copy full SHA for 70f76fc

• Avoid redundant transfers during usage refinement. (iree-org#13632)

  Show description for 28a085e
  benvanik

  authored
  28a085e

  Copy full SHA for 28a085e

• Cast Address Space in FuncOp and CallOpo for LLVMGPUs (iree-org#13518)

  Show description for 8858543
  grypp

  authored
  8858543

  Copy full SHA for 8858543

Pagination

Previous

Next