Landing: d8a26b20e474
Project / Subsystem
gcc / gcc/x86
Date
2026-08-11
Author
H.J. Lu
Commit
d8a26b20e474d15e1842321be0a7163e52e68442
Source
github
Perf win
No
Breaking
No
All attributes
- project
- gcc
- subsystem
- gcc/x86
- patch_id
- —
- commit_hash
- d8a26b20e474d15e1842321be0a7163e52e68442
- source_type
- github
- headline
- Enable 128-bit vector smaxmin/umaxmin with SSE4.1
- tldr
- x86 vector operations smaxmin and umaxmin are now enabled for 128-bit vectors using SSE4.1, requiring int128.
- author
- H.J. Lu
- outcome
- committed
- performance_win
- false
- breaking_change
- false
- series_id
- —
- series_parts
- []
- tags
-
- • x86
- • sse4.1
- • vectorization
- • optimization
- discussion_id_link
- —
- bugzilla_pr
- —
- date
- 2026-08-11T00:00:00.000Z
This change enables 128-bit vector smaxmin and umaxmin operations for the x86 architecture using SSE4.1 instructions. These operations do not require AVX512VL or AVX512DQ extensions. The relevant test cases, vec-narrow-1.C and vec-narrow-minmax-2.C, are updated to compile with ‘-msse4’ and specify the int128 requirement for x86. Previously, these 128-bit operations were not available for ia32 even with the ‘-mmmx’ flag.