[llvm] [InstCombine][CVP] Fold length-one memset with variable fill (PR #213240)

[Antonio Frighetto via llvm-commits]

https://github.com/antoniofrighetto requested changes to this pull request.

I'm not sure this should need CVP/LVI. For the simple reported case, using computeKnownBits() should suffice, and I believe it should suffice for the dominating-condition case too (https://godbolt.org/z/7o6bbMse9). If CFG changes are needed, the transformation should possibly live in AggressiveInstCombine.

https://github.com/llvm/llvm-project/pull/213240