Landing: f44527143df2

← All landings

Project / Subsystem

gcc / testsuite/aarch64

Date

2026-07-26

Author

Kyrylo Tkachov

Commit

f44527143df24430447d68e4b0e461685827f0f6

Source

github

Perf win

Breaking

View on GitHub →

All attributes

project: gcc
subsystem: testsuite/aarch64
patch_id: —
commit_hash: f44527143df24430447d68e4b0e461685827f0f6
source_type: github
headline: aarch64 testsuite: use a relative tolerance in vect-vaddv.c
tldr: Corrects floating-point comparison in an aarch64 vector test using relative tolerance.
author: Kyrylo Tkachov
outcome: committed
performance_win: false
breaking_change: false
series_id: —
series_parts: []
tags: • testsuite
• aarch64
• simd
• floating-point
discussion_id_link: —
bugzilla_pr: —
url: https://github.com/gcc-mirror/gcc/commit/f44527143df24430447d68e4b0e461685827f0f6
date: 2026-07-26T00:00:00.000Z

An aarch64 vectorization test (vect-vaddv.c) now uses a relative tolerance for floating-point comparisons instead of an absolute one. This addresses issues where differing associativity in vector summation led to minor discrepancies exceeding the previous absolute tolerance, causing test failures.