| commit | acddd0c2542f79ea681cf5cea8364eccfdcfe137 | [log] [tgz] |
|---|---|---|
| author | Ben Vanik <ben.vanik@gmail.com> | Mon Oct 16 14:59:05 2023 -0700 |
| committer | GitHub <noreply@github.com> | Mon Oct 16 17:59:05 2023 -0400 |
| tree | af8ea1739631ee800a66859b74a281fd1bd4274e | |
| parent | d7f97bac19721a133545ba4bc237857fb639ca2e [diff] |
Adding hal.dispatch.extern op. (#15193) This is an inline dispatch op on tensors that can be inserted in source programs to fully specify an externally derived dispatch region. Improvements were made for external executables in flow/stream/hal. There's a decent amount of cruft on these paths that'll need to be cleaned up; particularly MaterializeInterfaces has long been in need of a reworking (likely decomposing with the assistance of a new op or two) and it's not great that flow now has handling of a HAL pseudo-op. The stream convert HAL->stream which handles hal.tensor.import/export would be a better place but executable deduplication happens in flow today and we need inlined dispatches like this to deduplicate. I toyed with also having a DeduplicateExecutables in stream but decided against it for today.
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.