Landing: fb21f1686938
Project / Subsystem
gcc / middle-end
Date
2026-09-07
Author
Roger Sayle
Commit
fb21f1686938d6cca13fa7efd9f94830a7a6ef04
Source
github
Perf win
Yes
Breaking
No
All attributes
- project
- gcc
- subsystem
- middle-end
- patch_id
- —
- commit_hash
- fb21f1686938d6cca13fa7efd9f94830a7a6ef04
- source_type
- github
- headline
- Middle-end: Improve bitreverse RTL expansion with AOP
- tldr
- New optab improves bitwise operations by allowing backends to choose optimal instructions.
- author
- Roger Sayle
- outcome
- committed
- performance_win
- true
- breaking_change
- false
- series_id
- —
- series_parts
- []
- tags
-
- • optimization
- • middle-end
- • RISC-V
- • x86
- • AVR
- discussion_id_link
- —
- bugzilla_pr
- —
- date
- 2026-09-07T00:00:00.000Z
This patch enhances the RTL expansion for bitreverse and store_fixed_bitfield operations by introducing a new aop_optab. This aop_optab allows target backends to precisely choose between bitwise OR, XOR, or addition instructions, leading to better code generation on various architectures like x86, RISC-V, and AVR. Previously, the expansion relied on a generic PLUS or IOR that might not be optimal. This change provides finer control, yielding more efficient code, as demonstrated by an improved x86_64 expansion for a struct assignment.