[llvm] [AMDGPU] Recognise bitmask operations as srcmods (PR #149110)

[Chris Jackson via llvm-commits]

chrisjbris wrote:

@LU-JOHN 

> Can you add some scalar test cases? Particularly for the 64-bit cases. The SALU versions do not support source modifiers, so they should not fold if it's a scalar candidate

Yes, I've added some tests to integer-select-src-modifiers.ll.

https://github.com/llvm/llvm-project/pull/149110