[llvm] [RISCV] custom scmp(x, 0) and scmp(0, x) lowering for RVV (PR #151753)

[Luke Lau via llvm-commits]

https://github.com/lukel97 commented:

Makes sense to me.

I wish we could write this as a combine on `vselect (setcc 0, n, gt), 1, 0 -> shl n, sew - 1`, but given that the fixed and scalable nodes diverge I think custom lowering scmp is as good as we're going to get. 

https://github.com/llvm/llvm-project/pull/151753