Commits

Branch selector

 main

User selector

qedawkins

Datepicker

All time

Commit History

Commits on Apr 30, 2024

• Update sdxl rocm flags to match branch performance (iree-org#17232)

  Show description for c2571c8
  qedawkins

  authored
  c2571c8

  Copy full SHA for c2571c8

Commits on Apr 26, 2024

• [GlobalOptimization] Drop obsolete transpose propagation pattern (iree-org#17203)

  Show description for ae70704
  qedawkins

  authored
  ae70704

  Copy full SHA for ae70704

Commits on Apr 24, 2024

• [GlobalOptimization] Add patterns for bubbling of transposes (iree-org#17035)

  Show description for 3bc7e29
  qedawkins

  authored
  3bc7e29

  Copy full SHA for 3bc7e29

Commits on Apr 21, 2024

• [Codegen] NFC: Cleanup common transform dialect ops (iree-org#17120)

  Show description for fd79fca
  qedawkins

  authored
  fd79fca

  Copy full SHA for fd79fca

• [Codegen] Drop iree.fold_arith_ext_into_contraction for upstream variant (iree-org#17119)

  Show description for d861372
  qedawkins

  authored
  d861372

  Copy full SHA for d861372

Commits on Apr 19, 2024

• [LLVMGPU] allow multiple m and n dims in contraction distribution (iree-org#16943)

  Show description for e87ff17
  qedawkins

  authored
  e87ff17

  Copy full SHA for e87ff17

Commits on Apr 17, 2024

• NFC: Add VectorLayoutInterface method for getting the layout rank (iree-org#17071)

  Show description for c83f9ba
  qedawkins

  authored
  c83f9ba

  Copy full SHA for c83f9ba

Commits on Apr 15, 2024

• [Preprocessing] NFC: Finish migrating passes to use new tablegen (iree-org#17047)

  qedawkins

  authored
  3d626b5

  Copy full SHA for 3d626b5

• [Flow] Switch to new pass generation tablegen definitions (iree-org#17046)

  Show description for 39091a7
  qedawkins

  authored
  39091a7

  Copy full SHA for 39091a7

Commits on Mar 28, 2024

• [LLVMGPU] Allow sending expanded convolutions down mfma pipeline (iree-org#16917)

  Show description for 719a8a9
  qedawkins

  authored
  719a8a9

  Copy full SHA for 719a8a9

• [Codegen] Fix layout analysis for vector.transpose (iree-org#16820) (iree-org#16921)

  Show description for 5acacb7
  

  qedawkins

  and

  Groverkss

  authored
  5acacb7

  Copy full SHA for 5acacb7

Commits on Mar 13, 2024

• [LLVMGPU] Fix fused elementwise broadcasts in mfma pipeline (iree-org#16756)

  Show description for 858cce6
  qedawkins

  authored
  858cce6

  Copy full SHA for 858cce6

Commits on Mar 12, 2024

• [ROCM] Fix build with runtime tracing enabled (iree-org#16737)

  qedawkins

  authored
  a3603c6

  Copy full SHA for a3603c6

Commits on Mar 5, 2024

• [Codegen][GPU] NFC: Move SPIRVCreateFastSlowPath to Common/GPU (iree-org#16669)

  Show description for 7782a41
  qedawkins

  authored
  7782a41

  Copy full SHA for 7782a41

• [Codegen][ROCDL] Replace custom generalization pass with upstream one (iree-org#16662)

  Show description for 7171014
  qedawkins

  authored
  7171014

  Copy full SHA for 7171014

Commits on Mar 3, 2024

• [LLVMGPU] Fix allocation space for thread local allocations (iree-org#16640)

  Show description for 047211b
  qedawkins

  authored
  047211b

  Copy full SHA for 047211b

Commits on Mar 1, 2024

• [Codegen][rocdl] Generalize all contraction ops before folding unit dims (iree-org#16632)

  Show description for 62da59b
  qedawkins

  authored
  62da59b

  Copy full SHA for 62da59b

• [GlobalOpt] Propagate transposes through unary elementwise ops (iree-org#16623)

  Show description for af65386
  qedawkins

  authored
  af65386

  Copy full SHA for af65386

• [Codegen][GPU] Add support for transpose distribution with nested layouts (iree-org#16630)

  Show description for 71f87af
  qedawkins

  authored
  71f87af

  Copy full SHA for 71f87af

Commits on Feb 29, 2024

• [Flow] Improve flow-break/trace match failure debug output (iree-org#16625)

  Show description for 03bc749
  qedawkins

  authored
  03bc749

  Copy full SHA for 03bc749

• [Flow] Avoid fusion of dequantization-like ops with producers (iree-org#16610)

  Show description for 8237d9a
  qedawkins

  authored
  8237d9a

  Copy full SHA for 8237d9a

Commits on Feb 28, 2024

• [Codegen][GPU] Fix multi-dim warp reduction (iree-org#16602)

  Show description for 021b41c
  qedawkins

  authored
  021b41c

  Copy full SHA for 021b41c

• Integrate llvm/llvm-project@80cff273906b (iree-org#16597)

  Show description for da1e547
  

  qedawkins

  and

  ScottTodd

  authored
  da1e547

  Copy full SHA for da1e547

Commits on Feb 27, 2024

• [Codegen][ROCDL] Extend mfma pipeline to support a few more matmul variants (iree-org#16582)

  Show description for 6b995b9
  qedawkins

  authored
  6b995b9

  Copy full SHA for 6b995b9

• [Codegen][ROCDL] Add support for nhwc convolution with mfma (iree-org#16579)

  Show description for db677a8
  qedawkins

  authored
  db677a8

  Copy full SHA for db677a8

Commits on Feb 26, 2024

• [Codegen][GPU] Add pass to generalize named convolution ops (iree-org#16575)

  Show description for baeffa7
  qedawkins

  authored
  baeffa7

  Copy full SHA for baeffa7

• [ROCM] Use translation info to store waves-per-eu (iree-org#16573)

  Show description for c730000
  qedawkins

  authored
  c730000

  Copy full SHA for c730000

• Integrate llvm/llvm-project@c2042c3bc823 (iree-org#16567)

  Show description for 7e1e7b0
  qedawkins

  authored
  7e1e7b0

  Copy full SHA for 7e1e7b0

Commits on Feb 23, 2024

• NFC: Unify registration of Util interfaces on external ops (iree-org#16553)

  Show description for 4a613b6
  qedawkins

  authored
  4a613b6

  Copy full SHA for 4a613b6

• [Codegen] Add missing barrier to GPUVectorAlloc (iree-org#16551)

  Show description for 11089e8
  qedawkins

  authored
  11089e8

  Copy full SHA for 11089e8

• [Codegen][GPU] Add support for distributing broadcasts with nested layouts (iree-org#16532)

  qedawkins

  authored
  fb1151a

  Copy full SHA for fb1151a

Commits on Feb 22, 2024

• [VectorExt] Add support for projecting nested layouts (iree-org#16528)

  Show description for d4b6d74
  qedawkins

  authored
  d4b6d74

  Copy full SHA for d4b6d74

• [rocdl] Use a full slice of vector values to avoid anchors (iree-org#16534)

  Show description for 4182c40
  qedawkins

  authored
  4182c40

  Copy full SHA for 4182c40

Commits on Feb 21, 2024

• [VectorExt] Add custom parser/printer to elide identity orderings (iree-org#16522)

  Show description for 91a6bf7
  qedawkins

  authored
  91a6bf7

  Copy full SHA for 91a6bf7

• [Preprocessing] Add pre-configured pass pipeline for conv transpose (iree-org#16520)

  Show description for 566d6c2
  qedawkins

  authored
  566d6c2

  Copy full SHA for 566d6c2

Pagination

Previous

Next