commit | d5fc2c2acc9785104133e8c4cc0e2f649035b473 | [log] [tgz] |
---|---|---|
author | Lei Zhang <antiagainst@google.com> | Thu Nov 11 11:16:41 2021 -0500 |
committer | GitHub <noreply@github.com> | Thu Nov 11 11:16:41 2021 -0500 |
tree | 4294bdccfe32509dae450adff7c318a0c47cfbe2 | |
parent | 954b7497cf1800a2b58347bd773678725d41b87e [diff] |
Use upstream patterns to fold `affine.min` ops in loops (#7564) This commit adds wrapper patterns to call upstream `affine.min` canonicalization utilities and perform folding for them. Wrapping is needed because we tile and distribute at the same time, while the upstream pattern only expects tiling to function. So the pattern "discards" the distribution aspect and feeds the tiling aspect into the upstream pattern to make it work. This makes it possible to replace some of the home brewed ad-hoc patterns. The SPIR-V side is cleaned up for this; but we don't delete all such `affine.min` patterns yet; they are also relied on in other backends and can be removed once those are migrated over.
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.