| commit | be62a3c854e17a418ceb1178d00a30090b601df1 | [log] [tgz] |
|---|---|---|
| author | bjacob <benoitjacob@google.com> | Mon Apr 03 12:04:41 2023 -0400 |
| committer | GitHub <noreply@github.com> | Mon Apr 03 16:04:41 2023 +0000 |
| tree | 368da939f11e0d0389469bb195eae66c3c88d6e7 | |
| parent | e768c17f8c6ee10c59be9d88051d9d4fa61d4e58 [diff] |
e2e matmul benchmark as standalone C calling pack, mmt4d, unpack ukernels (#12848) (This is take-2 of #12818). This is equivalent to compiling a `linalg.matmul` with data-tiling and ukernels enabled and benchmarking the resulting module, but as a standalone C program directly calling the ukernels, it's trivially optimal in all sorts of way that the full-blown iree-compile-based flavor isn't at the moment. It can be useful to quickly evaluate potential performance if all sorts of issues were fixed, and with profiling it can help answer questions of how much packing/unpacking overhead is inherent in the data-tiling approach.
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.