| commit | a69124b9a0aac36097cbe9c9a57f39b4190d6e64 | [log] [tgz] |
|---|---|---|
| author | Stella Laurenzo <stellaraccident@gmail.com> | Mon Sep 06 13:14:10 2021 -0700 |
| committer | GitHub <noreply@github.com> | Mon Sep 06 13:14:10 2021 -0700 |
| tree | e83eaeff5ca299dbc5a25453495b7cbd8dd60221 | |
| parent | aa64b7d1d0a99bf5253062c961a3bd072d181205 [diff] |
Health and welfare on the iree_pydm dialect. (#6978) Dialect changes: * Simplify file naming to limit repetition. * Add PythonTypeInterface, implemented by all Python types (currently used for some identity and numeric promotion rules). * Apply previous recommendation and rework var alloc/load/store into a !free_var_ref type, alloc_free_var, load_var, store_var. Cell variables will need something different but this should generalize (i.e. cell vars need to be resolved symbolically, inter-procedurally). * Add a UnionType in order to support type refinement (not yet used, and still needs some refinement). * Forked scf.if into `functional_if` for the specific case where we are emitting conditional Python code of a functional nature (shows up in conditionals and short-circuit evals a lot, but most Python control flow is naturally CFG based). With this change, the pydm dialect is self-complete, not relying on ops from outside of itself. This will help with type inference, etc. * Implemented OpAsmOpInterface::getDefaultDialect on `func` and `functional_if`, making all pydm ops able to be used prefix-free, cleaning up IR a lot. * Made `none` ConstantLike. Added `success` and `failure` ops to produce `ExceptionResults`. * Added `make_tuple` op (not yet used). * Added `promote_numeric` op. * Implemented simple folders for `constant`, `none`, `success`, `as_bool`, `bool_to_pred`, `raise_on_failure`, `select`. * Implemented static numeric promotion with: a folder+canonicalizer on `promote_numeric`, a canonicalizer on `dynamic_binary_promote` which reduces it to primitives for static cases. * Implemented no-op box/unbox canonicalizations. With this, the dialect is in good shape to start building out the optimization and lowering pipelines.
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!
| CI System | Build System | Platform | Architecture | Configuration / Component | Status |
|---|---|---|---|---|---|
| Kokoro | Bazel | Linux | x86-64 |  | |
| Kokoro | CMake & Bazel | Linux | x86-64 (swiftshader) | Integrations |  |
| Kokoro | CMake & Bazel | Linux | x86-64 (turing) | Integrations |  |
| Kokoro | CMake | Linux | x86-64 (swiftshader) |  | |
| Kokoro | CMake | Linux | x86-64 (swiftshader) | asan |  |
| Kokoro | CMake | Linux | x86-64 (turing) |  | |
| Kokoro | CMake | Android | arm64-v8a | Runtime (build only) |  |
| Kokoro | CMake | Bare Metal | risc-v-32 | Runtime |  |
| Kokoro | CMake | Linux | risc-v-64 | Runtime |  |
| Buildkite | CMake | Android | arm64-v8a | Runtime |  |
| BuildKite | CMake | Android | arm64-v8a | Runtime Benchmarks |  |
| BuildKite | CMake | Linux | x86-64 | Tracing + Standalone Runtime |  |
IREE is licensed under the terms of the Apache 2.0 License with LLVM Exceptions. See LICENSE for more information.