commit | da06cc91a85b1317a3c5cf6b890386364094fa3a | [log] [tgz] |
---|---|---|
author | MaheshRavishankar <1663364+MaheshRavishankar@users.noreply.github.com> | Mon Mar 07 15:26:09 2022 -0800 |
committer | GitHub <noreply@github.com> | Mon Mar 07 15:26:09 2022 -0800 |
tree | bdb234a3169165f245412c21c60d33f0a9765b3f | |
parent | fc4fbc5a8f4a1edb57df3070759d088bee15023c [diff] |
Set configuration using untiled operations. (#8446) With tile and distributed being moved out of Flow, the backends can now use the untiled operations to decide configuration. Make that change and move the TileAndDistributeToWorkgroups Pass to run after configuration specifications on all backends. This makes the `getUntiledShape` and `getUntiledResultShape` methods unnecessary and are removed. The `LoopTileAndDistributionInfo` is still needed since its load bearing for some `affine.min` canonicalizations on the SPIR-V side. - On the CPU side the configuration setting made heavy use of the existence of tiled loops to decide configuration in an op agnostic way. Adapt the configuration selection to not rely on these tiled loops and use the untiled operations directly. - One of the CUDA tests checking for illegal configuration seems to be off and is being looked at in #8456 . Disabling that test now, but will have to be fixed after the fact in that PR. Co-authored-by: Lei Zhang <antiagainst@google.com> Co-authored-by: Thomas Raoux <thomasraoux@google.com>
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.