Migrate TOSA input conversion to a compiler plugin. (#15495)

Progress on https://github.com/openxla/iree/issues/15468

Using the plugin API lets us avoid plumbing `#ifdef
IREE_HAVE_TOSA_INPUT` across the source tree. The `IREE_INPUT_TOSA`
CMake define now just chooses whether to include the plugin or not.

Future PRs will migrate other components like the StableHLO input
conversion. I chose to start with the TOSA conversion because it has
fewer files.

## Build system details

This is _mostly_ a code move. The Bazel and CMake configurations were a
little tricky though. Note that some CMake files here were authored
manually, not via bazel_to_cmake.

The Torch input conversion has files located at
`compiler/plugins/input/Torch/torch-iree/**`, with includes rooted on
`torch-iree` (e.g. `#include "torch-iree/InputConversion/Passes.h"`).
However, Bazel wants include paths to match the (unambiguous)
`WORKSPACE`-relative paths (e.g. `#include
"compiler/plugins/input/Torch/torch-iree/InputConversion/Passes.h"`).
Solutions for working around that in Bazel are [documented
here](https://bazel.build/tutorials/cpp-use-cases#add-include-paths).

Similarly, the Torch input conversion roots its CMake targets at
`torch-iree::InputConversion`. In cases where we use relative targets
that isn't an issue, but bazel_to_cmake does not understand that for
absolute target labels.

We have a few options:

1. Use WORKSPACE-relative paths and generate CMake files from Bazel
files
2. (this PR) Use plugin-relative paths and manually author a few CMake
files
3. Use plugin-relative paths and teach bazel_to_cmake about include
copts
4. Use something like `//compiler/src:defs` to propagate include paths
(I tried this a few ways already)
5. Use relative include paths (`./...`)

## Future work

* This new plugin depends on `iree::compiler::Dialect::Flow::Transforms`
for `IREE::Flow::createStripSignednessPass()`. I think this is the only
use of that pass, so it could be moved in to the plugin.
* Mixed input dialect conversion is worth trying now
35 files changed
tree: 35e06fa8e95c3b5a86ef04e62c54316ffff45e64
  1. .devcontainer/
  2. .github/
  3. build_tools/
  4. compiler/
  5. docs/
  6. experimental/
  7. integrations/
  8. lib/
  9. llvm-external-projects/
  10. runtime/
  11. samples/
  12. tests/
  13. third_party/
  14. tools/
  15. .bazel_to_cmake.cfg.py
  16. .bazelignore
  17. .bazelrc
  18. .bazelversion
  19. .clang-format
  20. .dockerignore
  21. .git-blame-ignore-revs
  22. .gitignore
  23. .gitmodules
  24. .yamllint.yml
  25. AUTHORS
  26. BUILD.bazel
  27. CITATION.cff
  28. CMakeLists.txt
  29. configure_bazel.py
  30. CONTRIBUTING.md
  31. LICENSE
  32. README.md
  33. 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.

CI Status

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!

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

License

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