-
Notifications
You must be signed in to change notification settings - Fork 4.8k
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
System.Numerics.Tensors.Tests.SingleGenericTensorPrimitives.SpanScalarDestination_SpecialValues fails under jitstress on win-x86 #102335
Comments
Tagging subscribers to this area: @JulieLeeMSFT, @jakobbotsch |
Same for |
It's interesting that either of those are failing for jitstress runs. JitStress shouldn't impact whether or not something is imported as intrinsic by the JIT and so the actual IR at importation should be the same here. I'm not going to have time to look at this until after next week (due to Microsoft Build being next week), but I wonder if this is exposing some long-existing issue between |
Failed in: runtime-coreclr libraries-jitstress 20240522.1 Failed tests:
Error message:
Stack trace:
|
Failed in: runtime-coreclr libraries-jitstress 20240523.1 Failed tests:
Error message:
Stack trace:
|
Failed in: runtime-coreclr libraries-jitstress 20240526.1 Failed tests:
Error message:
Stack trace:
|
Failed in: runtime-coreclr libraries-jitstress 20240527.1 Failed tests:
Error message:
Stack trace:
|
Failed in: runtime-coreclr libraries-jitstress 20240528.1 Failed tests:
Error message:
Stack trace:
|
Failed in: runtime-coreclr libraries-jitstress 20240602.1 Failed tests:
Error message:
Stack trace:
|
Failed in: runtime-coreclr libraries-jitstress 20240603.1 Failed tests:
Error message:
Stack trace:
|
@AndyAyersMS this looks to be DPGO related. The repro involves:
and stops reproing if we also set:
Nothing immediately jumped out from the JitDisasm as to why there'd be a difference here. There's a couple more spills and of course the calls to |
Ok, let me take a look, hopefully today... |
Failed in: runtime-coreclr libraries-jitstress 20240605.1 Failed tests:
Error message:
Stack trace:
|
Failed in: runtime-coreclr libraries-jitstress 20240606.1 Failed tests:
Error message:
Stack trace:
|
Failed in: runtime-coreclr libraries-jitstress 20240610.1 Failed tests:
Error message:
Stack trace:
|
Failed in: runtime-coreclr libraries-jitstress 20240611.1 Failed tests:
Error message:
Stack trace:
|
Failed in: runtime-coreclr libraries-jitstress 20240612.1 Failed tests:
Error message:
Stack trace:
|
Failed in: runtime-coreclr libraries-jitstress 20240613.1 Failed tests:
Error message:
Stack trace:
|
Failed in: runtime-coreclr libraries-jitstress 20240616.1 Failed tests:
Error message:
Stack trace:
|
Failed in: runtime-coreclr libraries-jitstress 20240617.1 Failed tests:
Error message:
Stack trace:
|
Any luck, @AndyAyersMS? |
Ooops, forgot about this one. I will look at it today for sure. |
has bad codegen when stressed via
Haven't drilled in further yet. |
Failed in: runtime-coreclr libraries-jitstress 20240618.1 Failed tests:
Error message:
Stack trace:
|
Guessing the bug is that we have the wrong type on a permute node
which becomes
But I suspect |
In `gtNewSimdSumNode` we need to type permute nodes as `simd16`, not the sum's type. If we split the tree at the permute we spill to the wrong typed temp. Also provide the ability to halt stress tree splitting after some number of splits. Fixes dotnet#102335.
Fixing that seems to do the trick. |
Thanks! |
Example pipeline rune: https://dev.azure.com/dnceng-public/public/_build/results?buildId=677384&view=ms.vss-test-web.build-test-results-tab
Example console log: https://helixre107v0xdcypoyl9e7f.blob.core.windows.net/dotnet-runtime-refs-pull-102261-merge-6809f596adec438f97/System.Numerics.Tensors.Tests/1/console.2b07caea.log?helixlogtype=result
cc @tannergooding
There are also failures in
NonGenericSingleTensorPrimitivesTests.CosineSimilarity_AllLengths
in this pipeline run for win-x64.The text was updated successfully, but these errors were encountered: