Google Git
Sign in
opensecura / 3p / openxla / iree / d5c4ef12db0d24204a208c45a881b852a3e56a42
commitd5c4ef12db0d24204a208c45a881b852a3e56a42[log] [tgz]
authorMax191 <44243577+Max191@users.noreply.github.com>Mon Sep 09 13:06:20 2024 -0700
committerGitHub <noreply@github.com>Mon Sep 09 16:06:20 2024 -0400
tree5d117a2795cfbd4e698d1ac0599947033bb88ee0
parent69ca7dfa042141602f0ff748c456c31be86e3db6 [diff]
[LLVMGPU] Improve mfma tile sizes for TileAndFuse pipeline (#18459)

This attempts to select better tile sizes on the TileAndFuse pipeline
based on the number of elements loaded along the K dimension tile. A new
field is added to `GPUMMAHeuristicSeeds` called
`bestKElementCountPerSubgroup`, which represents the best number of
elements to have along the K dimension of a subgroup tile. This can be
more useful than the existing `bestKTileCountPerSubgroup`, since it
takes into account the size of the intrinsic being selected. Having
control over the specific number of elements loaded helps give better
control over overall VGPR utilization and load queue saturation.

The tile size selection logic in `setMatmulLoweringConfig` has been
tuned using this new field. The current tile sizes improve overall
performance on SDXL for the TileAndFuse pipeline over the old tile size
selection logic.

---------

Signed-off-by: Max Dawkins <max.dawkins@gmail.com>
4 files changed
tree: 5d117a2795cfbd4e698d1ac0599947033bb88ee0
  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. .dockerignore
  20. .git-blame-ignore-revs
  21. .gitattributes
  22. .gitignore
  23. .gitmodules
  24. .pre-commit-config.yaml
  25. .yamllint.yml
  26. AUTHORS
  27. BUILD.bazel
  28. CITATION.cff
  29. CMakeLists.txt
  30. configure_bazel.py
  31. CONTRIBUTING.md
  32. LICENSE
  33. MAINTAINERS.md
  34. README.md
  35. RELEASING.md
  36. WORKSPACE
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 Status

IREE is still in its early phase. We have settled down on the overarching infrastructure and are actively improving various software components as well as project logistics. It is still quite far from ready for everyday use and is made available without any support at the moment. With that said, we welcome any kind of feedback on any communication channels

Release status

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

Build status

CI PkgCI

Host platformBuild status
LinuxCI - Linux x64 clang
CI - Linux arm64 clang
macOSCI - macOS x64 clang
WindowsCI - Windows x64 MSVC

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

  • 2021-06-09: IREE Runtime Design Tech Talk (recording and slides)
  • 2020-08-20: IREE CodeGen: MLIR Open Design Meeting Presentation (recording and slides)
  • 2020-03-18: Interactive HAL IR Walkthrough (recording)
  • 2020-01-31: End-to-end MLIR Workflow in IREE: MLIR Open Design Meeting Presentation (recording and slides)

License

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