Landing: 001b3a711d88

← All landings

Project / Subsystem

gcc / risc-v

Date

2026-05-05

Author

Pan Li

Commit

001b3a711d88850d046a59ca8aa38e845cb6747c

Source

github

Perf win

Yes

Breaking

View on GitHub →

All attributes

project: gcc
subsystem: risc-v
patch_id: —
commit_hash: 001b3a711d88850d046a59ca8aa38e845cb6747c
source_type: github
headline: RISC-V: Combine vec_duplicate + vmsgt.vv to vmsgt.vx
tldr: The compiler now combines `vec_duplicate` with `vmsgt.vv` into `vmsgt.vx` on RISC-V when the cost of GR2VR is zero.
author: Pan Li
outcome: committed
performance_win: true
breaking_change: false
series_id: —
series_parts: []
tags: • risc-v
• vectorization
• optimization
discussion_id_link: —
bugzilla_pr: —
url: https://github.com/gcc-mirror/gcc/commit/001b3a711d88850d046a59ca8aa38e845cb6747c
date: 2026-05-05T00:00:00.000Z

The RISC-V backend now combines vec_duplicate followed by vmsgt.vv instructions into a single vmsgt.vx instruction when the cost of transferring data from a general-purpose register to a vector register (GR2VR) is zero. This optimization reduces the number of instructions and can improve performance. The change modifies the predicate matching and RTL generation to enable this combination.