Landing: aac02c966618
Project / Subsystem
gcc / gcc/optimization
Date
2026-07-09
Author
Konstantinos Eleftheriou
Commit
aac02c9666184d19334a0527ec444b5e55fbd113
Source
github
Perf win
Yes
Breaking
No
All attributes
- project
- gcc
- subsystem
- gcc/optimization
- patch_id
- —
- commit_hash
- aac02c9666184d19334a0527ec444b5e55fbd113
- source_type
- github
- headline
- widening_mul: Lower long-multiply chains to inline longhand
- tldr
- Improves long-multiply optimization by lowering wide intermediate products to N-bit operations when target support is lacking.
- author
- Konstantinos Eleftheriou
- outcome
- committed
- performance_win
- true
- breaking_change
- false
- series_id
- —
- series_parts
- []
- tags
-
- • gcc
- • optimization
- • widening_mul
- • long-multiply
- discussion_id_link
- —
- bugzilla_pr
- —
- date
- 2026-07-09T00:00:00.000Z
This commit enhances the widening_mul optimization pass by introducing a mechanism to lower intermediate wide multiply products back to N-bit operations. When the target architecture cannot efficiently handle the full 2N-bit result of a multiply, this pass now breaks down the calculation into smaller, N-bit partial products that can be handled, improving code generation for such cases. It also refines how operands wider than N bits are split and how results are materialized in GIMPLE.