| commit | 4f8ee5187e8fb76d7ba40239018b2a6c99a41b62 | [log] [tgz] |
|---|---|---|
| author | Ben Vanik <ben.vanik@gmail.com> | Thu May 16 16:53:55 2024 -0700 |
| committer | GitHub <noreply@github.com> | Thu May 16 16:53:55 2024 -0700 |
| tree | ca2c1e909b1d36d7760e59eed44737fede734abf | |
| parent | dece30eb0c164a4a9914cbc321055901939f191d [diff] |
Moving demotion/promotion passes to input conversion. (#17422) They need to happen before ABI materialization. Really they should happen in frontend dialects where the semantics can be enumerated and verified. This at least removes the duplication we had between the pass flags in StableHLO and the global optimization flags that could conflict. The f64->f32 demotion is still enabled by default because codegen has issues (CPU pulls in libm functions like exp, SPIR-V for all targets has issues with f64 memrefs, etc). At least now there's a single flag and it happens in a place that somewhat works and users can start to disable for testing. StableHLO previously defaulted i64->i32 but all other input types didn't. The default is now i64->i32 disabled by default everywhere. People using StableHLO and wanting i64->i32 demotion will need to opt in to that behavior with `--iree-input-demote-i64-to-i32=true`. Closes #8745.
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.
Community meeting recordings: IREE YouTube channel
IREE is licensed under the terms of the Apache 2.0 License with LLVM Exceptions. See LICENSE for more information.