Landing: 4288735a9d03

Project / Subsystem

gcc / gcc/vectorizer

Date

2026-08-05

Author

Richard Biener

Commit

4288735a9d034b47ded8a38d0d82e9b3e536c534

Source

github

Perf win

No

Breaking

No

All attributes

project
gcc
subsystem
gcc/vectorizer
patch_id
—
commit_hash
4288735a9d034b47ded8a38d0d82e9b3e536c534
source_type
github
headline
Cleanup vectorizable_conversion related code.
tldr
Vectorizer simplifies handling of scalar operands in vectorizable conversions.
author
Richard Biener
outcome
committed
performance_win
false
breaking_change
false
series_id
—
series_parts
[]
tags
  • • vectorizer
  • • optimization
discussion_id_link
—
bugzilla_pr
—
date
2026-08-05T00:00:00.000Z

This commit refactors parts of the GCC vectorizer, specifically the vectorizable_conversion function. It elides requirements for scalar operands op0 and op1 when they are fed by SLP permutations, which cannot be computed in that context. A new overload for vect_is_simple_use is introduced to ease this transition, and an unused scalar operand argument is removed from supportable_indirect_convert_operation.