Optimize LiftTrivialIf and re-run it only on block-count change#2460
Open
ThrudPrimrose wants to merge 4 commits into
Open
Optimize LiftTrivialIf and re-run it only on block-count change#2460ThrudPrimrose wants to merge 4 commits into
ThrudPrimrose wants to merge 4 commits into
Conversation
Ports the optimized LiftTrivialIf: range-aware folding of loop boundary guards (iteration-range tautology/contradiction), cartesian-product/unsatisfiable branch pruning for if/elif chains, deterministic block relabeling, and gating the whole-SDFG bookkeeping walks (set_nested_sdfg_parent_references + reset_cfg_list) on an actual removal so the pass is no longer quadratic in block count. should_reapply now fires only on Modifies.States (a ConditionalBlock added or removed, or a loop peel that shifts an enclosing range) instead of the whole CFG: interstate-edge content changes never create or destroy a conditional, so reacting to them only re-scanned the SDFG for no new work.
ThrudPrimrose
marked this pull request as draft
July 23, 2026 14:48
The parent-reference / cfg-list repair after a branch splice no longer walks the whole SDFG once per control-flow block. add_node already fixes every moved block's parent_graph/sdfg, so only the deep-copied NestedSDFG nodes' parent pointers are repaired, scoped to the spliced subtree, and the CFG-list index is refreshed once at the end of the pass (only when something changed). The up-front whole-SDFG relabel is replaced by relabeling just the spliced top-level blocks. The per-condition triviality check (pystr_to_symbolic + evaluate, the pass's real hot spot) is memoized in a module-level lru_cache(maxsize=16384, typed=True) keyed on the condition string. Because it is module-level, not per-instance, it also hits across the FixedPointPipeline's repeated pass re-invocations. On the CloudSC frontend this cuts a warm LiftTrivialIf run from ~640ms to ~36ms; results are unchanged (removal counts identical, all validated).
ThrudPrimrose
marked this pull request as ready for review
July 23, 2026 15:19
2 tasks
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Ports the optimized
LiftTrivialIf(range-aware boundary-guard folding, cartesian-product/unsatisfiable branch pruning, deterministic relabeling, whole-SDFG bookkeeping walks gated on an actual removal so it is no longer quadratic in block count) and narrowsshould_reapplytoModifies.Statesso the pass re-runs only when a conditional block is added or removed, not on every interstate-edge change.