| commit | d025b72f88c4e2de6be8d29e863f8f9dff27bd5a | [log] [tgz] |
|---|---|---|
| author | bjacob <benoitjacob@google.com> | Mon Nov 21 22:06:50 2022 +0000 |
| committer | GitHub <noreply@github.com> | Mon Nov 21 22:06:50 2022 +0000 |
| tree | 2c3c30437285b83de8e9a829f8ad67fae20420ef | |
| parent | 9408934a471ea20b573a401a1b897b729ed07a3c [diff] |
ASan tests: account for multi-driver Lit tests (#11242) This PR reenables all tests that were disabled under ASan! The LSAN failures that motivated the current exclude list should all be fixed by #11206. Yet, simply dropping the exclude-list doesn't work, because some of the entries in that list are Lit tests that exercise multiple drivers in a way that's encoded as `//RUN` comments within the test source, opaque to the build and test system --- there was no way to tell which of those tests involved Vulkan. It would be nice to only disable LSAN for the Vulkan part of these tests, but this isn't feasible as things are currently structured, with the multiple driver runs conflated withing a single Lit test source. At first I introduced a new tag/label, `multi-driver-including-vulkan`, to label those tests, so that `build_and_test_asan.sh` could disable LSAN accordingly. Then I thought, how does this scale? Will there be a `multi-driver-XYZ` tag for all combinations of (possibly multiple?!) drivers? Then I thought, let's just use the existing `driver=vulkan` tag for this, giving it "multiple choices possible" semantics. Actually, "tags" are meant to work that way; perhaps the only thing about that tag that makes it somewhat surprising that it would work that way is the `=` sign: it might seem strange to have `driver=X` and `driver=Y` tags together on the same rule. Then again maybe we'll get used to it, and maybe precisely having such tags side by side (as enforced by `buildifier`) will make it clear that it does have such semantics.
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.