| commit | 34449df33fbcc37d680abf37fdcfeae875ac04a1 | [log] [tgz] |
|---|---|---|
| author | Ben Vanik <ben.vanik@gmail.com> | Wed May 01 11:19:25 2024 -0700 |
| committer | GitHub <noreply@github.com> | Wed May 01 11:19:25 2024 -0700 |
| tree | e027e289c1c07639cc8f35bd8f6caa26af310e40 | |
| parent | e15968ff884d3f2078d2817e4ae793a0c2b33449 [diff] |
[NFC] Refactoring MeshToFlow pass out from patterns. (#17245) This moves the pass out to the Transforms/ directory and separates the pass logic from the patterns for consistency with all the other conversions. This will also help us move around where the pass is run and possibly merge it with the main ConvertToFlow pass by separating lookup from channel materialization as we do with device globals. Such refactorings will (likely) be required for proper multi-device scheduling but this initial refactoring is just moving code around and leaves the pass running where it does (during the input pipeline). We could probably put the ConvertMeshToFlow pass in InputConversion/Common/ today based on where it's invoked, but I feel like input may be the wrong layer for it if we expect any global optimization phase work to happen on the mesh dialect and we may want to move it later. It depends on whether we treat the flow ops as the optimization mechanism for any partitioning work we do during global optimization.
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.