[llvm] [AArch64] Prefer `SADDLV` over `ADDV` for vector mask reductions (PR #183522)

[David Green via llvm-commits]

https://github.com/davemgreen commented:

Hi - Apart from the changes cause by SelectArithExtendedRegister, do you have a case where this is better? I'm not saying it will be worse, I'm just not sure any of your tests show an improvement. Could you add a test where it is worth transforming?

https://github.com/llvm/llvm-project/pull/183522