commit | 2e49e5eaabefe940e895cc63c47ca488b241a310 | [log] [tgz] |
---|---|---|
author | Han-Chung Wang <hanchung@google.com> | Fri Sep 24 06:32:17 2021 +0800 |
committer | GitHub <noreply@github.com> | Fri Sep 24 06:32:17 2021 +0800 |
tree | 8103a3602f86ef1e2e2f578f0d162861aa9b8db3 | |
parent | 1547ac28f48c6f455adf726828ce548e20a3f3d5 [diff] |
Update semantics of linalg_ext.reverse. (#7155) 1. Allow the op takes multiple reverse dimensions. 2. Update the loop bound of reverse dim with the whole size. Re (2), it is required because of tiling interface. The current tiling interface expects the implmenetation returned a tiled op. However, two tiled ops are required in the previous definition. E.g., say that we have `M` tiles in total. [T_1], [T_2], ... , [T_M] The result of `T_1` will be stored to the position of `T_M`, and vise versa. If we iterate only half of the tensor, we need to create reversed `T_1` and reversed `T_M`, then store them to corresponding offsets. In this context, two tiled op are created, but only one can be returned. The updated semantic also meets the current lowering, instead of emitting the affine maps into the op itself. This PR also adds e2e tests for the op. It is a step toward https://github.com/google/iree/issues/5045
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.