This allows targets to make more decisions about reserved registers
after isel. For example, now it should be certain there are calls or
stack objects in the frame or not, which could have been introduced by
So if I'm reading this right this just basically just delays TLI->finalizeLowering to the end of ExpandISelPseudos/FinalizeISel. Is there motivation to this other than it makes more logical sense to freeze registers as close to regalloc as possible?
Could you update the summary to explain what this is about?