| commit | e87ff178a9a7b210ec2c0e6726024adb0b0124e6 | [log] [tgz] |
|---|---|---|
| author | Quinn Dawkins <quinn.dawkins@gmail.com> | Fri Apr 19 11:28:12 2024 -0700 |
| committer | GitHub <noreply@github.com> | Fri Apr 19 14:28:12 2024 -0400 |
| tree | 6cc2df5b638a7dc245841d78152e6aca0670b938 | |
| parent | 9f9798900c8da312b572e119ab7ade1030114baf [diff] |
[LLVMGPU] allow multiple m and n dims in contraction distribution (#16943) This adjusts the layout generation logic to allow distribution of contractions with multiple m and n dimensions by greedily using the subgroup/tile_counts of the mma_schedule with the outer dims. The inner most m/n dimensions are still required to be divisible by the intrinsic shape. (and this only supports a single k dimension). This also decouples the ordering logic of the batch/subgroup distribution from the lane distribution for the intrinsics. Currently it assumes intrinsics can only specify three important sizes, an M, N, and K size. To support distributed batches this would require adding a fourth dim type.
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.