| commit | ebb5b7d4a3c59b8dac9c5a4fe4ba78f0c736dc7f | [log] [tgz] |
|---|---|---|
| author | Scott Todd <scotttodd@google.com> | Tue Nov 14 10:05:27 2023 -0800 |
| committer | GitHub <noreply@github.com> | Tue Nov 14 10:05:27 2023 -0800 |
| tree | c1ac68247f6ddfdffd45496c4b9147dd3201c41b | |
| parent | 643b46710be909cf5af0bb90a755c7f8808c01eb [diff] |
Migrate StableHLO input conversion to a compiler plugin. (#15568) Progress on https://github.com/openxla/iree/issues/15468 StableHLO-specific changes: * Added `input_stablehlo` plugin at `compiler/plugins/input/StableHLO/stablehlo-iree/` * Moved StableHLO input conversion code into the plugin * Note file paths: `stablehlo-iree/Conversion` is used instead of `stablehlo-iree/InputConversion` to save on lengths * Torch and TOSA both have fewer characters and do not have a `InputConversion/Preprocessing/` subfolder so they don't hit the same limit Cleanup now that all input dialect conversion is handled via plugins consistently: * Deleted `init_input_dialects.[h, cc]` and `init_input_passes.[h, cc]` * Removed dialect-specific code paths from `compiler/Pipelines/[Pipelines, Options].cpp` * Generated more `CMakeList.txt` files from `BUILD.bazel` files (input dialect dependencies and source files are organized into plugins with their own top-level filtering) * Added `auto_input_conversion.mlir` tests for each input plugin that use `--compile-to=input` to show how the "auto" input type handles their dialects
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 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!
See our website for more information.
IREE is licensed under the terms of the Apache 2.0 License with LLVM Exceptions. See LICENSE for more information.