[llvm] [RISCV][TTI] Improve SiFive7 reduction cost (PR #90951)

[Pengcheng Wang via llvm-commits]

wangpc-pp wrote:

> > I just wonder if we can make cost model bind to scheduling model.
> 
> Considering
> 
>     1. The interpretation of Latency and ReleaseAtCycles in the scheduling model may not align directly with TCK_Latency and TCK_RecipThroughput in the TTI cost model.
> 
>     2. In the scheduling model, the vector length is assumed to be VLMAX, whereas in the TTI cost model, we can derive the element count from fixed-length vector types.
> 
> 
> It might be better to keep them separate for now.

I'm not opposed to add processor-specific cost models, but I think we should well organize the code or it may be hard to mantain as time goes on.

https://github.com/llvm/llvm-project/pull/90951