Landing: a3f5541bcf62

Project / Subsystem

gcc / risc-v

Date

2026-09-28

Author

Pan Li

Commit

a3f5541bcf62d1142cea7f7f70afcabf888d947c

Source

github

Perf win

Yes

Breaking

No

All attributes

project
gcc
subsystem
risc-v
patch_id
—
commit_hash
a3f5541bcf62d1142cea7f7f70afcabf888d947c
source_type
github
headline
RISC-V: Allow RVV register overlap for vfw{add,sub,mul}.vf
tldr
RISC-V compiler now permits register overlap for wide floating-point vector instructions.
author
Pan Li
outcome
committed
performance_win
true
breaking_change
false
series_id
—
series_parts
[]
tags
  • • risc-v
  • • vector
  • • optimization
  • • rvv
discussion_id_link
—
bugzilla_pr
—
date
2026-09-28T00:00:00.000Z

The RISC-V GCC backend now allows register group overlap for wide floating-point vector instructions, specifically vfwadd.vf, vfwsub.vf, and vfwmul.vf. This change aligns the handling of these instructions with their vector-vector counterparts (.vv) and leverages the Wvr constraint in the vector.md file, potentially enabling more efficient instruction scheduling.