[llvm] [RISCV][TTI] Properly model odd vector sized LD/ST operations (PR #100436)

[Philip Reames via llvm-commits]

preames wrote:

A couple bits of context here.

The goal here is to get simple cases with -slp-vectorize-non-power-of-2 working.  We're probably got other places in the cost model which need fixed here, but I'm starting with a simple one.

This code could be better integrated with the default costing.  I do want to take a look at that, but it's slightly more involved than I wanted to couple into the -slp-vectorize-non-power-of-2 work.  

https://github.com/llvm/llvm-project/pull/100436