| commit | c4c58a15b5bbc99616de2441420b83795dc20c91 | [log] [tgz] |
|---|---|---|
| author | MaheshRavishankar <1663364+MaheshRavishankar@users.noreply.github.com> | Wed Aug 04 08:58:26 2021 -0700 |
| committer | GitHub <noreply@github.com> | Wed Aug 04 08:58:26 2021 -0700 |
| tree | ef6c0c5d3d404c54dc514ea3926c7ec72cf34d1f | |
| parent | f0d243f09d678f33a7150392048e4ec9026efd2a [diff] |
Make the CUDA codegen pipeline consistent with the CPU codegen pipeline. (#6615) This PR finishes the work of "unifying" the CPU backend and the CUDA backend by making LLVMGPULowerExecutableTarget similar to the analogous pass on the CPU side (LLVMCPULowerExecutableTarget). Some changes to the utility methods to set/get attributes that control the lowering. Now the utility methods allow you to override existing attributes and upto individual backends to either override the default or not. Remove the use of llvmgpu.workgroup.size attribute that shadows the workgroup_size attribute on the hal.executable.entry_point.op. Use the data from there where needed. Make initGPULaunchConfiguration handled all compute ops (and not just Linalg ops). This allows for tile and distribute of scatter operations as well. The CUDA side handles only static shapes. To handle dynamic shapes the ConvertToNVVM pass needs to handle shape dialect conversion to LLVM as well as support the ABI for dynamic shapes. Here the support for handling shape dialect is added by using the patterns that are used on the CPU side. For now, the scatter test is split into static and dynamic tests, with dynamic tests not run on CUDA backend.
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.