| commit | db62c355db54a57edc424915d0d75ff9ba3d7c73 | [log] [tgz] |
|---|---|---|
| author | MaheshRavishankar <1663364+MaheshRavishankar@users.noreply.github.com> | Thu Feb 18 21:20:36 2021 -0800 |
| committer | GitHub <noreply@github.com> | Thu Feb 18 21:20:36 2021 -0800 |
| tree | 96cf4000a3c57325c9f51bb6e9840b843b387925 | |
| parent | 7cd207d194ba33903f5c6818944d2ae9716b9a6f [diff] |
Add compilation path for lowering linalg on tensors op without using tile and distribute. (#4840) Not all operations in a dispatch region can be lowered through tile and distribute. Operations like linalg.tensor_reshape cannot be tiled. These needs a separate lowering mechanism. In this PR, such a mechanism is added where such operations are put into the dispatch region as is. Its upto the individual backends to handle these appropriately. On the GPU side trivially parallel operations are distribute across all invocations. On the CPU side such operations are executed sequentially for now. This could be modified to execute in parallel too, but the sequential execution is consistent with existing codegen path. The PR also adds support for lowering linalg.tensor_reshape operation by making LinalgBufferize handle this operation (further clean up related to Issue #4759 with removal of dead code) Instead of relying on ad-hoc pattern matching of the code within the dispatch region, and explicit flag is used to distinguish between the Linalg on tensors path and legacy path. On CPU the LinalgTileAndDistribute pass is split to have a pass MaterializeCPULaunchConfigurationPass. The former now is only for the legacy path, and the latter for the linalg on tensors path. On GPU the ConvertToGPUPass is used to distribute untiled operations in dispatch region to invocations.
IREE (Intermediate Representation Execution Environment, pronounced as “eerie”) is an MLIR-based end-to-end compiler that lowers ML models to a unified IR optimized for real-time mobile/edge inference against heterogeneous hardware accelerators. IREE also provides flexible deployment solutions for the compiled ML models.
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!
Python packages are published on the releases page. See the colab/ directory for examples.
IREE can be built from source using both Bazel and CMake on Windows and Linux. We also have experimental macOS support.
Please see the Getting Started pages on IREE's documentation hub to configure, compile, and run IREE in your favorite development environment!
IREE hosts all its documentation and project status dashboards on GitHub Pages. We are still building up the website; please feel free to create issues for the documentation you'd like to see!
We also have some public talks that explain IREE's concepts and architecture:
IREE adopts a holistic approach towards ML model compilation: the IR produced contains both the scheduling logic, required to communicate data dependencies to low-level parallel pipelined hardware/API like Vulkan, and the execution logic, encoding dense computation on the hardware in the form of hardware/API-specific binaries like SPIR-V.
The architecture of IREE is best illustrated by the following picture:
Being compilation-based means IREE does not have a traditional runtime that dispatches “ops” to their fat kernel implementations. What IREE provides is a toolbox for different deployment scenarios. It scales from running generated code on a particular API (such as emitting C code calling external DSP kernels), to a HAL (Hardware Abstraction Layer) that allows the same generated code to target multiple APIs (like Vulkan and Direct3D 12), to a full VM allowing runtime model loading for flexible deployment options and heterogeneous execution.
IREE aims to
IREE is in the early stages of development and not yet ready for broad adoption. Check out the long-term design roadmap to get a sense of where we're headed.
We plan on a quarterly basis using OKRs. Review our latest objectives to get a sense of what we're up to in the near term.
We use GitHub Projects to track progress on IREE components and specific efforts. We use GitHub Milestones to track the work associated with plans for each quarter.
| CI System | Build System | Platform | Architecture | Component | Status |
|---|---|---|---|---|---|
| Kokoro | Bazel | Linux | x86 | Core |  |
| Kokoro | CMake & Bazel | Linux | x86-swiftshader | Integrations |  |
| Kokoro | CMake & Bazel | Linux | x86-turing | Integrations |  |
| Kokoro | CMake | Linux | x86-swiftshader | Core + Bindings |  |
| Kokoro | CMake | Linux | x86-turing | Core + Bindings |  |
| Kokoro | CMake | Android | arm64-v8a | Runtime (build only) |  |
| BuildKite | CMake | Android | arm64-v8a | Runtime |  |
IREE is licensed under the terms of the Apache license. See LICENSE for more information.