| commit | af998b2c4733bc9f14dcd9415d6e31ea9964ed4b | [log] [tgz] |
|---|---|---|
| author | Scott Todd <scotttodd@google.com> | Thu Aug 05 12:59:22 2021 -0700 |
| committer | GitHub <noreply@github.com> | Thu Aug 05 12:59:22 2021 -0700 |
| tree | c845cea02a416437da8d8a9febc35afdce7dae1c | |
| parent | c846ef9c123acc77c3e59e4131c2c4dd61cbf1b6 [diff] |
Re-land moving tensor->flow passes. (#6651) Reverts https://github.com/google/iree/pull/6648, rolling-forward https://github.com/google/iree/pull/6586. First commit is the pure revert, second is a set of proposed fixes for the performance regression. The single pass (with a true/false option) is now split further: * `createConvertTensorOpsPass` * `createConvertLinalgTensorOpsPass(true)` * `createConvertLinalgTensorOpsPass(false)` --- This gets the original change closer to an NFC (but it is still not an NFC). Before this change: | stage | patterns | | --- | --- | | `buildCommonInputConversionPassPipeline` | `tensor::CastOp`, `tensor::FromElementsOp` | | Flow, before DispatchRegionFormation | `tensor::InsertSliceOp`, `tensor::ExtractSliceOp`, <br> `linalg::TensorCollapseShapeOp`, `linalg::TensorExpandShapeOp` | | Flow, after DispatchRegionFormation | `linalg::FillOp` | After this change: | stage | patterns | | --- | --- | | `buildCommonInputConversionPassPipeline` | (deleted) | | Flow, before DispatchRegionFormation | `tensor::CastOp`, `tensor::FromElementsOp`, <br>`tensor::InsertSliceOp`, `tensor::ExtractSliceOp`, <br>`linalg::TensorCollapseShapeOp`, `linalg::TensorExpandShapeOp` | | Flow, after DispatchRegionFormation | `linalg::FillOp` | Passes on `linalg` ops are the same, but two `tensor` ops are now converted after linalg fusion, rather than at input conversion time. The first version of this change had those `linalg` ops converted during the "common input conversion pipeline".
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 (Mako) |  |
| 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.