| commit | 0260947c60b374bff718545290ad693f3111646b | [log] [tgz] |
|---|---|---|
| author | MaheshRavishankar <1663364+MaheshRavishankar@users.noreply.github.com> | Wed May 15 09:51:12 2024 -0700 |
| committer | GitHub <noreply@github.com> | Wed May 15 09:51:12 2024 -0700 |
| tree | 7ef31820dea08bd24b975e2d5be4a68ebd560438 | |
| parent | bf0fbf0419b755e59c3798e9ecf1f3e8c909be99 [diff] |
[GlobalOpt] Simplify the logic used to pick the groups. (#17405) The existing logic uses dominance information and tries to find contractions that can be fused horizontally. The logic though ended up creating a large set of possbily fusable contraction ops, which then fail subsequent checks and result in fewer actual fusions. This change makes the initial set more constrained, but better formed to pass subsequent checks, resulting in more fusion candidates. This reduces number of dispatches in SDXL by 30 and improves execution time from 66 ms -> 64 ms on UNet.
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.
Community meeting recordings: IREE YouTube channel
IREE is licensed under the terms of the Apache 2.0 License with LLVM Exceptions. See LICENSE for more information.