[PATCH] D54498: AbstractCallSite -- A unified interface for (in)direct and callback calls

[Siddharth via Phabricator via llvm-commits]

bollu added a comment.

I have a question in the context of optimising sequential programs -- I'm interested in optimising across a trampoline call, since I have a haskell-like lazy language, which lowers CPS into LLVM-IR by constructing a trampoline. Example IR here <https://gist.github.com/bollu/8834682e5970550afcd6a0c991777fe1#file-ackermann-ll-L220> Some salient features are:

- All functions are `musttail`, and communicate data by custom "stacks" which are allocated on the heap.
- LLVM manages to remove some of the trampoline code, but it's still super messy (I can generate a clean version where the trampoline is visible)

Would the `AbstractCallSite` representation allow one to represent the control flow between:

  trampoline -> fn1 -> trampoline -> fn2 -> ...

And could this infrastructure help implement a pass that would resolve these custom stack pushes and pops?

Thanks, and sorry for the somewhat unstructured comment. I'm asking this question as a Haskeller interested in representing laziness within LLVM.


Repository:
  rL LLVM

https://reviews.llvm.org/D54498