[llvm] MachineBlockPlacement: Add tolerance to comparisons (PR #67197)

[Matthias Braun via llvm-commits]

MatzeB wrote:

> I am curios if there is a way to teach BFI to return appropriately rounded values, so that we don't modify downstream algorithms?

Yes in fact I started out with code "rounding" all BFI values and still have it around here. But initially I wanted this more of a drive-by fix rather than challenging base layer assumptions which would probably need more discussion from stakeholders...

Also part of the motivation of the current fix is plainly that https://github.com/llvm/llvm-project/pull/66285 forces me to update hundreds of unit tests and this here seemed like an easier "drive-by" change so I can frontload a number of those updates...

So yeah may be better, but would like some more feedback / support if we want to go in that direction. In the meantime I think this change shouldn't hurt?

https://github.com/llvm/llvm-project/pull/67197