diff options
Diffstat (limited to '')
| -rw-r--r-- | compiler/InterpreterLoopsJoinCtxs.ml | 28 |
1 files changed, 27 insertions, 1 deletions
diff --git a/compiler/InterpreterLoopsJoinCtxs.ml b/compiler/InterpreterLoopsJoinCtxs.ml index 4b5cfb82..930bffac 100644 --- a/compiler/InterpreterLoopsJoinCtxs.ml +++ b/compiler/InterpreterLoopsJoinCtxs.ml @@ -292,7 +292,33 @@ let reduce_ctx_with_markers (merge_funs : merge_duplicates_funcs option) (* Reduce_ctx can only be called in a context with no markers *) let reduce_ctx = reduce_ctx_with_markers None -(* TODO Adapt and comment *) +(* Collapse an environment + + This is the second part of a join, where we attempt to simplify and remove all projection markers. + This function is called after reducing the environments, and attempting to simplify all the pairs + of borrows and loans. + + We traverse all abstractions, and merge abstractions when they contain the same element, + but with dual markers (i.e., PLeft and PRight). + + For instance, if we have the abstractions + + abs@0 { | MB l0 _ |, ML l1 } + abs@1 { ︙MB l0 _ ︙, ML l2 } + + we will merge abs@0 and abs@1 into a new abstraction abs@2, removing the marker for duplicated elements, + and taking the join of the remaining elements + + abs@2 { MB l0 _, ML l1, ML l2 } + + Rem.: Doing this might introduce new pairs of borrow/loans to be merged in different abstractions: in the example above, + this could occur if there was another abstraction in the context containing ML l0, which would need to be simplified through + a further reduce. + It is unclear whether this can happen in practice. If so, a solution would be to preprocess the environments when doing + a join: while not in the current formalism, it is sound to split an element with no markers into a duplicated pair of the + same element with left and right markers. Doing this before reduce would allow to reduce all possible pairs of borrow/loans, + before finally collapsing and removing all markers. +*) let collapse_ctx_markers (span : Meta.span) (loop_id : LoopId.id) (merge_funs : merge_duplicates_funcs) (old_ids : ids_sets) (ctx0 : eval_ctx) : eval_ctx = |