| commit | 6f2c98f4fa482f6172cb5129fd45135592772b9e | [log] [tgz] |
|---|---|---|
| author | Jakub Kuderski <jakub@nod-labs.com> | Fri Dec 15 16:11:47 2023 -0500 |
| committer | GitHub <noreply@github.com> | Fri Dec 15 21:11:47 2023 +0000 |
| tree | 44ba29628ef23e0090d42c1a29a1a00494ccc6f9 | |
| parent | 416e4b4ea42f86927947be45ae83388cf581a10f [diff] |
[LLVMGPU] Add multi-row vector reduction configuration for ROCm (#15941) This is to speed up matvec. The new configuration is constrained to a simple subset of matvecs and only applied on ROCm targets. I haven't evaluated this on CUDA, while adding it to the SPIR-V pipeline requires more work to lower down to vector types legal in SPIR-V. In the future revisions, I plan to relax the heuristic and apply it to other matvec-like generics, e.g., matvec fused with dequantization.
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.