[llvm] [RISCV] enable VTYPE before whole RVVReg move (PR #117866)

[Craig Topper via llvm-commits]

topperc wrote:

> Do you have any performance data available with this patch? My primary concern with the entire "treat this as a compiler problem" is that we may have some lurking performance problem exposed by the mitigation. I would _strongly_ like to see measurements of this on representative hardware - in particular, hardware like the BP3 which _doesn't_ have the faulting behavior.
> 

For most workloads this probably doesn't show up much. The call would have also invalidated the vector registers so there's nothing to copy unless we also just did a whole register load and immediately copied it's result to another register. I think the case that's more likely to show performance issues is when there are vector arguments/returns. That's going to require intrinsic code.

So what should we measure performance on?

https://github.com/llvm/llvm-project/pull/117866