commit | 6206d35dc9c3fb4502814d974e84500226fd68f9 | [log] [tgz] |
---|---|---|
author | Scott Todd <scotttodd@google.com> | Fri May 27 14:39:59 2022 -0700 |
committer | GitHub <noreply@github.com> | Fri May 27 14:39:59 2022 -0700 |
tree | 7644474f90877d816ae2efc6a6c7c431bb2556e3 | |
parent | cb828660f1092d30cdc484535dbff5441aa3000f [diff] |
Add test runner page and associated scripting for the web. (#9238) This adds a test runner webpage and associated scripting for building IREE's runtime tests with Emscripten. <img src="https://user-images.githubusercontent.com/4010439/170773583-c2002dd9-de4f-459a-b1e8-68e9c1eaecd8.png" width="500"> How this works: * Our tests (and benchmarks) generate binary files that when executed run tests then return an exit code * When compiling for the web using Emscripten, each binary target produces a `.js` and a `.wasm` file (and optionally a `.wasm.debug.wasm` and other debugging artifacts) * Running an Emscripten-produced binary on a webpage requires just defining the `Module` object and importing the `.js` --- Up until this point, I've been building a very narrow list of targets with the Emscripten toolchain. Now that I'm building all runtime tests, I had to disable more targets that aren't compatible with Emscripten - namely `hal/local/elf/` and `executable_library_benchmark` (which uses the embedded elf loader). I'd prefer to not sprinkle branches into our CMake files, but that will need some deeper, and IMO overdue :), refactoring to how executable loaders are built and registered. --- Ideas for future improvements: * Test index page that lists all targets, sorted by package, and lets you run individual tests, package test suites, or all tests * `ctest --show-only=json-v1` can give useful metadata for generating such a page * ctest integration ([here is a good reference](https://www.stevenengelhardt.com/2018/04/17/unit-testing-emscripten-library-in-browser-using-cmake-and-nightwatch-js/)) * Threading / event loop / other flag tweaks to get more tests passing and exercising browser features
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.
IREE is licensed under the terms of the Apache 2.0 License with LLVM Exceptions. See LICENSE for more information.