[llvm] [AArch64] Fix vectorToScalarBitmask BE (#156312) (PR #156314)

[David Green via llvm-commits]

davemgreen wrote:

> I can see what you mean, I've changed the test as you suggested basically coping and pasting the functions which you pointed out.
> 
> The only issue is that I would like to avoid using `update_test_checks` because the generated checks will test behavior outside the scope of this patch. This makes the test itself more brittle and susceptible to small variations in code generation, causing the test to fail even if the mask is correctly generated and used in the same manner.
> 
> It may even prompt the user to rerun `update_test_checks` without rewriting the mask checks.

I see what you mean, but often they can be intertwined and if the codegen changes the lane masks should be changed to accommodate.

Lets go with this version and we can always add extra lines in the future if needed.

https://github.com/llvm/llvm-project/pull/156314