Landing: e77642a6f29d
Project / Subsystem
gcc / lra
Date
2026-09-25
Author
Vladimir N. Makarov
Commit
e77642a6f29d87c48a935ea6da7c26ce3f8747d3
Source
github
Perf win
No
Breaking
No
All attributes
- project
- gcc
- subsystem
- lra
- patch_id
- —
- commit_hash
- e77642a6f29d87c48a935ea6da7c26ce3f8747d3
- source_type
- github
- headline
- LRA: Handle winning MEM operands to avoid asm rejections
- tldr
- Fixes an issue where LRA incorrectly classified memory operands, leading to assembly rejections.
- author
- Vladimir N. Makarov
- outcome
- committed
- performance_win
- false
- breaking_change
- false
- series_id
- —
- series_parts
- []
- tags
-
- • register allocation
- • assembly
- • bugfix
- discussion_id_link
- github:gcc-mirror/gcc#127326
- bugzilla_pr
- —
- date
- 2026-09-25T00:00:00.000Z
This commit addresses a regression in the Live Register Allocation (LRA) pass where a winning operand that was already a memory location (MEM) was not correctly handled. Previously, it would only fall back to memory if OFFMEMOK was set and the operand was a register. This change ensures that MEM operands are correctly classified, preventing LRA from rejecting valid assembly code.