[llvm] [LV] Implement integer part of VPDerivedIV cost model (PR #198252)

[David Sherwood via llvm-commits]

david-arm wrote:

Hi @ElvinaYakubova, that's strange because it directly contradicts https://github.com/llvm/llvm-project/pull/153039 that said this wasn't worth vectorising. I can only assume that since PR #153039 landed we did vectorise the loop, but applied some transformations after loop vectorisation that made the performance better. I'd rather not revert this PR if possible because I believe that overall the cost model changes are an improvement. I'll see if there is a way to reduce the cost of some of the other operations in the loop.

https://github.com/llvm/llvm-project/pull/198252