[llvm] [VPlan] First step towards VPlan cost modeling. (PR #92555)

[Florian Hahn via llvm-commits]

fhahn wrote:

> FYI the new assert `VF.Width == Width && "VPlan cost model and legacy cost model disagreed"` is also firing when compiling Fortran benchmark SPEC ACCEL calculix and cam. These applications are huge, so I will need time to give a reproducer/insructions. I think my tests included your [eea150c](https://github.com/llvm/llvm-project/commit/eea150c84053035163f307b46549a2997a343ce9) fix, but I also need to double check that. I will come back to you next week. [edit: this is when compiling the Fortran application with flang]

@jeanPerier  it would be great if you could share a reproducer once you have one. It doesn't need to be small or reduced if that's easier.


https://github.com/llvm/llvm-project/pull/92555