... | ... | @@ -16,6 +16,11 @@ I move stuff here from the different sections that is on our radar at the moment |
|
|
- Many ideas there, but it is much more complicated than it seems. Monotonicity issues etc.
|
|
|
- !5583 might be a faintly related first step at improvement here
|
|
|
- I'm thinking that we need to properly track data dependencies, e.g., the unstable set idea
|
|
|
- Arity analysis vs. Termination analysis: What are the differences between `exprOkForSpec` and `exprIsCheap`?
|
|
|
- Divergence is cheap, but not speculatable
|
|
|
- Cheap means "Are we OK with evaluating it multiple times or not at all when normally we'd evaluate it exactly once?", e.g., may we defer and unshare the work?
|
|
|
- Speculatable means "Are we OK with evaluating it (once) when normally we might not evaluate it at all?"
|
|
|
- Why can't they share more code? It feels pretty similar; E.g., HNF things are both speculatable (ignoring the annoyance with eval'd unfoldings of case binders) and cheap
|
|
|
|
|
|
# Simplifier and occurrence analysis
|
|
|
|
... | ... | @@ -173,4 +178,4 @@ I move stuff here from the different sections that is on our radar at the moment |
|
|
- Why? What's needed? A formal Specification? Which part? Static or dynamic semantics?
|
|
|
- Also how much? Whole pattern language or just enough of a fragment to explain or patterns?
|
|
|
- I see there is https://gitlab.haskell.org/rae/haskell as a starting point, but it seems to focus entirely on static semantics. But probably the document to complete?
|
|
|
- We talked about it; it's a matter of pushing the proposal forward rather than investing actual elbow grease into an impl. |
|
|
- We talked about it; it's a matter of pushing the proposal forward rather than investing actual elbow grease into an impl. |
|
|
\ No newline at end of file |