commit | f3362af2671acf0956a98f6b5a3a1f34b41c04b2 | [log] [tgz] |
---|---|---|
author | Oleksandr "Alex" Zinenko <zinenko@google.com> | Thu Jun 09 18:04:56 2022 +0200 |
committer | GitHub <noreply@github.com> | Thu Jun 09 09:04:56 2022 -0700 |
tree | 1d201b210a8a928b70fca3953954b7f15cb43bee | |
parent | d97e8bc0d306568ea047f408610779f7450f9230 [diff] |
Use upstream vectorize transform op (#9412) This is a functional change due to the differences in the implementation of the upstream vectorize op and its local equivalent. Specifically, the upstream op can be only targeted at an isolated-from-above payload IR op, as opposed to the local equivalent that could be targeted at any payload IR op or untargeted. Targeting an arbitrary op for vectorization is problematic because vectorization is implemented as a set of patterns, some of which perform vectorization and some of which are enabling or cleaning; targeting an individual payload IR (implemented by outlining the op) hinders the application of the latter class of patterns. This is most visible in the double-tiling.mlir test, where the local approach was not actually vectorizing `tensor.pad` operations despite the corresponding option being set.
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.