[llvm] [TTI] Handle experimental.vp.reverse in BasicTTIImpl (PR #147868)

[Luke Lau via llvm-commits]

lukel97 wrote:

> Why the cost of scalable instructions are invalid. I think these are valid operations.

Yeah they're valid operations, but part of the reason we return an invalid cost is because we inadvertently drop the values when costing so we get the scalarized cost, see https://github.com/llvm/llvm-project/pull/147677 which should fix part of it. I think the type based costing will still need to be fixed after this

https://github.com/llvm/llvm-project/pull/147868