Landing: 997a8a9bb85b

← All landings

Project / Subsystem

gcc / gcc

Date

2026-06-05

Author

Jakub Jelinek

Commit

997a8a9bb85b615e9c1ec7fed802e203a6aaa4a3

Source

github

Perf win

No

Breaking

No

View on GitHub →

All attributes

project: gcc
subsystem: gcc
patch_id: —
commit_hash: 997a8a9bb85b615e9c1ec7fed802e203a6aaa4a3
source_type: github
headline: Store-merging pass considers bswap and fixes RISC-V build
tldr: The store-merging optimization in GCC now correctly handles byte-swap operations and resolves a build failure on RISC-V.
author: Jakub Jelinek
outcome: committed
performance_win: false
breaking_change: false
series_id: —
series_parts: []
tags: • optimization
• risc-v
• byte-swap
• internal api
discussion_id_link: —
bugzilla_pr: —
url: https://github.com/gcc-mirror/gcc/commit/997a8a9bb85b615e9c1ec7fed802e203a6aaa4a3
date: 2026-06-05T00:00:00.000Z

This commit enhances the store-merging optimization pass in GCC to account for widen_bswap_or_bitreverse operations, rather than relying on an expander for 32-bit byte swaps. This change resolves a build issue on the RISC-V architecture where the bswapsi2 expander caused a failure due to an incorrect FAIL condition. Now, the store-merging pass considers __builtin_bswap32 as cheap if a bswapdi2 expander is present, leading to more robust and correct code generation, especially for 64-bit targets.