Landing: 8f8968aa70a4

← All landings

Project / Subsystem

gcc / sh

Date

2026-06-29

Author

Oleg Endo

Commit

8f8968aa70a455fa55d267a4d448af094639838c

Source

github

Perf win

Breaking

View on GitHub →

All attributes

project: gcc
subsystem: sh
patch_id: —
commit_hash: 8f8968aa70a455fa55d267a4d448af094639838c
source_type: github
headline: SH: Add movv2sf patterns to support LRA
tldr: New patterns for `movv2sf` and related instructions are added to the SH backend to prevent LRA from entering reload cycles.
author: Oleg Endo
outcome: committed
performance_win: false
breaking_change: false
series_id: —
series_parts: []
tags: • sh
• register allocation
• compiler patterns
discussion_id_link: —
bugzilla_pr: —
url: https://github.com/gcc-mirror/gcc/commit/8f8968aa70a455fa55d267a4d448af094639838c
date: 2026-06-29T00:00:00.000Z

The SH backend introduces new patterns for the movv2sf instruction and its variations to properly handle splits after register allocation. Without these patterns, the Long-Range Addresser (LRA) can get stuck in infinite reload cycles. The primary use case identified for V2SF mode is the FSCA instruction.