Landing: cd962857ea3d

← All landings

Project / Subsystem

gcc / tree-optimization

Date

2026-08-18

Author

Richard Biener

Commit

cd962857ea3d8db2e7ed08f98df4152e7739e8cd

Source

github

Perf win

No

Breaking

No

View on GitHub →

All attributes

project: gcc
subsystem: tree-optimization
patch_id: —
commit_hash: cd962857ea3d8db2e7ed08f98df4152e7739e8cd
source_type: github
headline: Avoid setting SLP_TREE_REPRESENTATIVE for vector permute operations
tldr: Vector SLP optimization now correctly handles permute nodes by not setting SLP_TREE_REPRESENTATIVE.
author: Richard Biener
outcome: committed
performance_win: false
breaking_change: false
series_id: —
series_parts: []
tags: • tree-optimization
• vectorization
• refactor
discussion_id_link: —
bugzilla_pr: —
url: https://github.com/gcc-mirror/gcc/commit/cd962857ea3d8db2e7ed08f98df4152e7739e8cd
date: 2026-08-18T00:00:00.000Z

This commit refactors GCC’s SLP (Straight-Line-Progression) vectorization pass to consistently avoid setting the SLP_TREE_REPRESENTATIVE flag for VEC_PERM_EXPR nodes. Several functions within tree-vect-slp.cc and tree-vect-slp-patterns.cc have been adjusted to either not set this flag or to correctly handle cases where it is not set for permute operations. This change ensures that the SLP analysis and optimization of vector permutations are handled more robustly, preventing potential issues during CSE (Common Subexpression Elimination) and analysis.