Landing: a9c22e6d049b
Project / Subsystem
gcc / tree-optimization
Date
2026-08-18
Author
Richard Biener
Commit
a9c22e6d049bf10e44b6a76384a1eb00991c3eb8
Source
github
Perf win
Yes
Breaking
No
All attributes
- project
- gcc
- subsystem
- tree-optimization
- patch_id
- —
- commit_hash
- a9c22e6d049bf10e44b6a76384a1eb00991c3eb8
- source_type
- github
- headline
- Back-to-back SLP instance scheduling fix [PR126926]
- tldr
- Fixes SLP vectorization scheduling when instances interact, preventing incorrect transformations.
- author
- Richard Biener
- outcome
- committed
- performance_win
- true
- breaking_change
- false
- series_id
- —
- series_parts
- []
- tags
-
- • tree-optimization
- • vectorization
- • slp
- discussion_id_link
- —
- bugzilla_pr
- —
- date
- 2026-08-18T00:00:00.000Z
This commit addresses a bug in the SLP (Shortest Path) vectorization scheduling where interacting SLP instances caused incorrect transformations. When one SLP instance transformed a vector constructor that was also a component of a second SLP instance, the scheduling process was disrupted. The fix ensures that only the right-hand side of the vector constructing statement is replaced, preserving the integrity of the SLP scheduling for both instances and resolving a PR for the tree-optimization component.