Repository navigation
[wasm][CoreCLR] Run Wasm HardwareIntrinsics tests, gated on IsSupported - #134771
Merged
Merged
Conversation
…yToRun Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>
|
Azure Pipelines: Successfully started running 6 pipeline(s). 10 pipeline(s) were filtered out due to trigger conditions. There may be pipelines that require an authorized user to comment /azp run to run. |
Contributor
|
Tagging subscribers to this area: @dotnet/area-system-runtime-intrinsics |
jkotas
reviewed
Sep 28, 2026
…msbuild opt-in Remove EnableWasmHWIntrinsicsTests and follow the model used by other platforms: each test returns early when PackedSimd.IsSupported is false, so the tests run in regular browser lanes (interpreted, no-op) as well as in the ReadyToRun hardware-intrinsics-wasm pipeline. Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>
lewing
changed the title
Member
|
|
…o-helix The template left helixType empty when readyToRun was true, so SendHelixJob failed with a missing Type. Mirror run-test-job.yml. Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>
The pipeline built no tests: in browser CI build.proj only builds test group 1 (groups 2..n are skipped to avoid OOM, #114123), and the Wasm HWI projects fall in priority-1 group 3. -tree:JIT/HardwareIntrinsics/Wasm also excluded the merged runner projects that wrap them. Build the two merged runners with -test:, and only apply the browser group restriction to full-tree builds, not -test/-dir/-tree subsets. Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>
This was referenced Sep 30, 2026
steveisok
approved these changes
Sep 30, 2026
lewing
added a commit
that referenced
this pull request
Oct 1, 2026
…loop (#134960) **Why:** Follow-up to #134771, where @jkotas asked for this. Today the only browser runtime-test leg that runs everything under R2R is the browser_wasm `R2R_CG2` leg in `runtime.yml`, and that leg is innerloop (priority 0 only). `runtime-coreclr outerloop` (`coreclr/ci.yml`) runs priority-1 tests on browser only in the checked interpreter leg, which uses R2R CoreLib with interpreted tests. browser_wasm is not in its `R2R_CG2` platform list. As a result, no pipeline runs priority-1 tests fully under ReadyToRun on browser. **Change:** Add a checked browser_wasm `R2R_CG2` outerloop leg to `eng/pipelines/coreclr/ci.yml`. It is a separate platform-matrix entry rather than an addition to the shared R2R platform list, because browser needs its own `extraBuildArgs` (`-os browser -p:HostConfiguration=Release`), the same as the existing browser interpreter leg. It reuses the browser build artifacts that `ci.yml` already produces. **Quarantines and skips:** Each failure below is specific to R2R and limited to wasm ReadyToRun. Some came from crashed or failed work items. Others appeared only as failed test results inside work items that passed, because merged runners record test failures in their results rather than in their exit code. | Failure | Issue | Mechanism | |---|---|---| | `b51875`, `b77707`: an exception thrown from a `finally` reached by leave is reported as unhandled (`Regression_8`, `Regression_PdbOnly_r_2`) | #134975 | `ActiveIssue` with `IsWasmReadyToRun` | | `self_override5`: `!ShouldCallPrestub()` assert (`Loader`) | #134949 | IL `ActiveIssue` with `IsWasmReadyToRun` | | `AboveStackLimit`: the R2R image exceeds the 1000-parameter wasm limit and fails to load (`Regressions`) | #132855 | `ActiveIssue` with `IsWasmReadyToRun`. The generated wrapper skips the call, so the image is never loaded. | | crossgen2 aborts or hangs in `Lowering::LowerStoreLocCommon` on `Vector128_1_*`, `Vector128_*` (`HardwareIntrinsics_General_r`/`_ro`) | #134976 | Runner-level `R2RIncompatible` on wasm CoreCLR, since more than one assembly is affected. Interpreter coverage is unchanged. | | `Samples/gc`, `arrres_il_r`: an object outlives the expected `GC.Collect`, so finalization and resurrection checks fail (`GC-scenarios1`, `Methodical_r1`) | By design, same class as #134803 | Permanent skip for wasm R2R only. Untracked and call-spilled stack GC refs are reported as always live (`clr-abi.md`, "GC References at Call Sites"). `gc` uses a `ConditionalFact`. `arrres_il_r` uses `R2RIncompatible`, because its hand-written IL entry point has no generated wrapper. | | `test115667`, `test132982`, `ResolvedFromDifferentContext`, `Runtime_64883`, `Runtime_66624`: they load their own assembly file, but `Assembly.Location` is empty for assemblies loaded from an R2R wasm image | — | Early return when `Location` is empty, following `ByRefLocals` and #134076 | Merged children are compiled by the all-targets `AnyOS` managed test build, so a wasm-conditioned project property on a merged child does not exclude it. The quarantines therefore use method-level attributes, properties on a runner, or properties on a process-isolated test. Runner and process-isolated test scripts are generated for browser, so those properties do apply. **Expectation:** This leg is expected to find more everything-R2R issues on browser. Failures should be triaged and tracked (Known Build Errors / `ActiveIssue`), not hidden. **Validation:** - [Build 1618757](https://dev.azure.com/dnceng-public/public/_build/results?buildId=1618757), two quarantine commits before the current head: - All 87 browser R2R work items passed, but 4 tests failed inside them: `Runtime_64883`, `Runtime_66624`, `arrres_il_r`, `Samples/gc`. The newer commits address all 4. - The browser interpreter leg had no failures. - Local browser-wasm Checked run of the affected runners and process-isolated tests, with and without `RunCrossGen2=1`, and with `DOTNET_TieredCompilation=0` as in CI: - Under R2R, the quarantined tests are skipped or return early. - In the interpreter, all of them run and pass. - `runtime-coreclr outerloop` has no PR trigger. To validate the change in CI, run `/azp run runtime-coreclr outerloop`. > [!NOTE] > This PR was generated with GitHub Copilot assistance. --------- Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
3 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Problem: the
src/tests/JIT/HardwareIntrinsics/Wasmtests never execute anywhere. They areCLRTestTargetUnsupportedunlessEnableWasmHWIntrinsicsTests=true, and the only pipeline that sets it (hardware-intrinsics-wasm, added in #130975) was build-only (sendToHelix: false). @jkotas pointed this out on #134762: build 1614067 ran no tests (no Helix jobs, no test runs). It turns out it didn't build them either. In browser CI,src/tests/build.projbuilds only test group 1 (#114123), and these priority-1 projects are in group 3.Change:
EnableWasmHWIntrinsicsTestsopt-in and use the model other platforms follow: each PackedSimd test returns early whenPackedSimd.IsSupportedis false. The tests now run in the regular browser lanes, i.e. interpreted test code calling ReadyToRun CoreLib, as well as fully under ReadyToRun.hardware-intrinsics-wasm: build the two merged runners (HardwareIntrinsics_Wasm_r/_ro.csproj) and send them to Helix withreadyToRun: true(crossgen2 at Helix time, same as the browser_wasmR2R_CG2runtime-test lane).src/tests/build.proj: the browser-CI "group 1 only" OOM workaround now applies only to full-tree builds, not to-test/-dir/-treesubsets. Verified locally: a full-tree browser CI build is unchanged, and the subset build now includes the group containing the Wasm HWI runners.build-runtime-tests-and-send-to-helix.yml: sethelixTypefor ReadyToRun sends, mirroringrun-test-job.yml. It was previously empty, soSendHelixJobfailed.PackedSimdIsSupportedno longer assertsIsSupportedis true. It still checksIsSupportedfor consistency with the reflection call andVector128.IsHardwareAccelerated.Evidence it should work: the libraries copy of PackedSimdTests passes 45/45 on browser CoreCLR trimmed R2R locally. The src/tests copy (66 tests) has never been built or executed in CI, so this PR's runs are the first real runs. Any failures should be triaged rather than hidden.
Note
This PR was generated with GitHub Copilot assistance.