Failures on [typed_data] Implement IEEE min/max for Float32x4/Float64x2 and add an exhaustive test

There are new test failures on [typed_data] Implement IEEE min/max for Float32x4/Float64x2 and add an exhaustive test.

The tests

lib/typed_data/min_max_edge_cases_test RuntimeError (expected Pass)

are failing on configurations

vm-aot-android-debug-arm_x64
vm-aot-android-release-arm_x64
vm-modaot-mac-debug-arm64

@modulovalue

@modulovalue

Also fails under QEMU (not run on a bot):

./tools/build.py -aarm && ./tools/test.py -aarm --use-qemu lib/typed_data/min_max_edge_cases_test

Thanks, will take a look!

We do not handle flushing in our arm32 simulator NEON parts properly. I'll try to fix the simulator first so qemu and our simulator agree, that is, our simulator trybots should also fail.

I'm not sure yet what effect flushing has on the algebraic properties I assert there. I'll investigate that.

(In any case, I'm very happy to see that those tests help us catch such nasty bugs!)

Navigation Menu

Repository navigation

Failures on [typed_data] Implement IEEE min/max for Float32x4/Float64x2 and add an exhaustive test #64521

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

Failures on [typed_data] Implement IEEE min/max for Float32x4/Float64x2 and add an exhaustive test #64521

Description

Activity

rmacnak-google commented on Oct 7, 2026

rmacnak-google commented on Oct 7, 2026

modulovalue commented on Oct 7, 2026

modulovalue commented on Oct 11, 2026

Metadata

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

Issue actions