commit | 835f496118c21b2942b760c141e81816666f55b8 | [log] [tgz] |
---|---|---|
author | Lei Zhang <antiagainst@google.com> | Wed Nov 17 11:06:07 2021 -0500 |
committer | GitHub <noreply@github.com> | Wed Nov 17 11:06:07 2021 -0500 |
tree | c4837127ecf2f798a7bf7ffeeecdf070b14b9f9d | |
parent | b793d9197e34baccf321c420635e28d59bf39e56 [diff] |
Avoid out-of-bound access when setting shape dimension size (#7650) The number of implicit loops in a Linalg op can be greater than the shape rank of the result. It's not always guaranteed that the first batch of implicit loops are used to indexing into the output. For example, for pooling ops, we have the following output indexing map: ``` (d0, d1, d2, d3, d4, d5) -> (d0, d1, d2, d5)> ``` Directly using the loop index to access the output shape will cause out of bound access (at `d5` in the above). This patch guards against such cases by considering implicit loop to output shape dimension mapping.
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.