Fix impNonConstFallback

[EgorBo]

Fixes #91335

I don't think it needs to be ported to 8.0 because non-constant values for imm-like arguments are not a production scenario (and is expected to cause compiler warrnings).

[ghost]

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

Issue Details

---

Fixes #91335

I don't think it needs to be ported to 8.0 because non-constant values for imm-like arguments are not a production scenario (and is expected to cause compiler warrnings).

Author:	EgorBo
Assignees:	-
Labels:	area-CodeGen-coreclr
Milestone:	-

[tannergooding]

I don't think it needs to be ported to 8.0 because non-constant values for imm-like arguments are not a production scenario (and is expected to cause compiler warrnings).

These can happen frequently for cases like the evaluating things in the debugger.

I think more notably there won't be any problems here because the underlying instruction expects 16-byte alignment and that's what the constant will be (and more concretely, the instructions support any alignment technically). The assert is only because it's expecting homogenous arguments and so checking for 32-byte alignment since we're expecting all constants are properly aligned (because that's good for perf).

Since there's no functional correctness (nor even a bad CQ) issue here, there's no need to backport as it will never surface to the end user.