[DispatchCreation] Fuse packed contractions with epilogues and unpack (#24967)

Packed contractions lose their fused epilogues when they arrive at
dispatch formation as mmt4d or a data-tiled convolution. The result
unpack can also become a separate dispatch.

Add opt-in fusion (`fuse-mmt4d`, `fuse-data-tiled-convolution`) of
layout-preserving elementwise consumers and the result unpack. Notes for
review:

- Epilogues must read the contraction result and write their own result
through identity maps. Epilogues that update the result through their
init operand are not matched.
- Only the unpack layouts that data-tiling materialization produces are
matched. For mmt4d that is the row-major form and the narrow-N
transposed form. For convolutions it is NHWC, NCHW and batchless HWC;
batchless CHW is not matched.
- The result unpack is the last op of the group, so a consumer of a
claimed unpack gets its own dispatch instead of fusing with the unpack.
- An unpack that its root does not claim becomes a root at its own
position in program order, as without the options, so it fuses with its
own consumers as before.
- A group holds pack-like ops (`linalg.pack`, `set_encoding`) or the
result unpack, but not both, whichever joins first. A unit-batch
collapse only joins together with the unpack that consumes it, so a
group never ends in the collapse.
- The matchers hardcode the CPU layouts; a TODO describes moving the
layout decision into its own component that dispatch formation queries.

No pipeline enables the options until #24969.

This is one of five PRs adding an experimental CPU path that
materializes data-tiling layouts before dispatch creation, so that
`linalg.mmt4d` and data-tiled convolutions fuse with their elementwise
epilogues and result unpacks.


Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
9 files changed
tree: cb1f98a28ed8171e0cfdac618e0b82f60aa56b88
  1. .github/
  2. build_tools/
  3. compiler/
  4. docs/
  5. experimental/
  6. integrations/
  7. lib/
  8. llvm-external-projects/
  9. runtime/
  10. samples/
  11. tests/
  12. third_party/
  13. tools/
  14. .bazel_to_cmake.cfg.py
  15. .bazelignore
  16. .bazelrc
  17. .bazelversion
  18. .clang-format
  19. .git-blame-ignore-revs
  20. .gitattributes
  21. .gitignore
  22. .gitmodules
  23. .pre-commit-config.yaml
  24. .yamllint.yml
  25. AUTHORS
  26. BUILD.bazel
  27. CITATION.cff
  28. CMakeLists.txt
  29. configure_bazel.py
  30. CONTRIBUTING.md
  31. GOVERNANCE.md
  32. LICENSE
  33. MAINTAINERS.md
  34. MODULE.bazel
  35. README.md
  36. RELEASING.md
README.md

IREE: Intermediate Representation Execution Environment

IREE (Intermediate Representation Execution Environment, pronounced as “eerie”) is an MLIR-based end-to-end compiler and runtime that lowers Machine Learning (ML) models to a unified IR that scales up to meet the needs of the datacenter and down to satisfy the constraints and special considerations of mobile and edge deployments.

See our website for project details, user guides, and instructions on building from source.

IREE Discord Status pre-commit OpenSSF Best Practices

Project news

Project status

Release status

Releases notes are published on GitHub releases.

PackageRelease status
GitHub release (stable)GitHub Release
GitHub release (nightly)GitHub Release
iree-base-compilerPyPI version
iree-base-runtimePyPI version

For more details on the release process, see https://iree.dev/developers/general/release-management/.

Build status

CI PkgCI

Nightly build status

Operating systemBuild status
LinuxCI - Linux arm64 clang
macOSCI - macOS arm64 clang

For the full list of workflows see https://iree.dev/developers/general/github-actions/.

Communication channels

Related project channels

  • MLIR topic within LLVM Discourse: IREE is enabled by and heavily relies on MLIR. IREE sometimes is referred to in certain MLIR discussions. Useful if you are also interested in MLIR evolution.

Architecture overview

IREE Architecture IREE Architecture

See our website for more information.

Presentations and talks

Community meeting recordings: IREE YouTube channel

DateTitleRecordingSlides
2025-06-10Data-Tiling in IREE: Achieving High Performance Through Compiler Design (AsiaLLVM)recordingslides
2025-05-17Introduction to GPU architecture and IREE's GPU CodeGen Pipelinerecordingslides
2025-02-12The Long Tail of AI: SPIR-V in IREE and MLIR (Vulkanised)recordingslides
2024-10-01Unveiling the Inner Workings of IREE: An MLIR-Based Compiler for Diverse Hardwarerecording
2021-06-09IREE Runtime Design Tech Talkrecordingslides
2020-08-20IREE CodeGen (MLIR Open Design Meeting)recordingslides
2020-03-18Interactive HAL IR Walkthroughrecording
2020-01-31End-to-end MLIR Workflow in IREE (MLIR Open Design Meeting)recordingslides

License

IREE is licensed under the terms of the Apache 2.0 License with LLVM Exceptions. See LICENSE for more information.