SVE: Fix the mask values returned by Helper

[kunalspathak]

As pointed in #112264 (comment), the mask values returned by helper methods GetMaskDouble() and GetMaskSingle() are incorrect and were getting returned using BitConverter.DoubleToInt64Bits() and BitConverter.SingleToInt32Bits() respectively. This might make a lane active or inactive, depending on the value of bit 0 (on which mask registers operate on). However, in the validation, we just assume that anything that is non-zero is active. Fix it by returning strictly either 0 or 1.

Fixes: #112264

These might also be related certain functional incorrect failures seen in #112377. Note that in the issue, there was also a reference of getting Index was outside the bounds of the array which might be unrelated and is not addressed by this fix.

[dotnet-policy-service]

Tagging subscribers to this area: @JulieLeeMSFT, @jakobbotsch
See info in area-owners.md if you want to be subscribed.

[Copilot]

Copilot reviewed 1 out of 1 changed files in this pull request and generated 1 comment.

Comments suppressed due to low confidence (2)

src/tests/JIT/HardwareIntrinsics/Arm/Shared/Helpers.cs:7747

• Directly returning the modulo result fixes the incorrect mask value, but confirm that TestLibrary.Generator.GetInt32() always produces non-negative values to ensure the returned mask is strictly 0 or 1.

return (float)(TestLibrary.Generator.GetInt32()%(int)2);

src/tests/JIT/HardwareIntrinsics/Arm/Shared/Helpers.cs:7745

• Consider adding unit tests to validate that both getMaskSingle() and getMaskDouble() always return 0 or 1 as intended.

public static float getMaskSingle() { ... }

[kunalspathak]

@dotnet/jit-contrib @dotnet/arm64-contrib

[amanasifkhalid]

Looks like there are some new CI failures to sort through

[jozkee]

/backport to release/9.0-staging

[github-actions]

Started backporting to release/9.0-staging: https://github.com/dotnet/runtime/actions/runs/15539869299