algorand
Inline subroutines
Summary
It would be a good optimization to identify subroutines that can be inlined directly into the caller code instead of being a legitimate subroutine. A common use case for subroutines is to increase readability, not necessarily because they are called frequently.
Scope
If a subroutine is only called from one place, it will always be more efficient to inline that subroutine into the caller code.
If a subroutine is called from multiple places, it might be more efficient to inline it if it's body is sufficiently small, but this is a more complex case to analyze, and likely less important. I only include it here for completeness--this issue should probably not address it.
One question: when I reviewed all the previous tickets, I think the reason for introducing this ticket is for reducing scratch slot usage in subroutine pre version 8.
After introduction of frame pointer, this seems no longer a pressing problem in functionality, for no dependency of scratch slot in subroutine is strictly required any more.
What do yall think?
