Landing: 6c8c0809015a

Project / Subsystem

gcc / i386

Date

2026-04-26

Author

Roger Sayle

Commit

6c8c0809015a3f2e663d3e1a3e0d2fe2117fb6a7

Source

github

Perf win

No

Breaking

No

All attributes

project
gcc
subsystem
i386
patch_id
commit_hash
6c8c0809015a3f2e663d3e1a3e0d2fe2117fb6a7
source_type
github
headline
i386: Refactor AVX512 comparisons in machine description sse.md.
tldr
GCC's x86 backend now has a cleaner organization of AVX512 vector comparisons, avoiding nonsensical instruction combinations.
author
Roger Sayle
outcome
committed
performance_win
false
breaking_change
false
series_id
series_parts
[]
tags
  • i386
  • x86
  • avx512
  • vectorization
  • code generation
discussion_id_link
bugzilla_pr
date
2026-04-26T00:00:00.000Z

The x86 backend in GCC has been refactored to improve the organization of AVX512 vector comparisons. The previous organization led to the generation of potentially invalid instructions (e.g., integer comparisons specifying floating-point rounding modes). The new structure uses a more logical decomposition of patterns, ensuring that only valid instruction combinations are generated. This improves code quality by using the correct instructions.