Bug 24412 - Add CCX-awareness to qspinlock
Summary: Add CCX-awareness to qspinlock
Status: NEW
Alias: None
Product: ANCK 5.10 Dev
Classification: ANCK
Component: X86 (show other bugs) X86
Version: unspecified
Hardware: x86_64 Linux
: P3-Medium S3-normal
Target Milestone: ---
Assignee: Guanjun
QA Contact:
URL:
Whiteboard:
Keywords:
Depends on:
Blocks:
 
Reported: 2025-09-04 10:33 UTC by kongyingqiao
Modified: 2025-09-05 18:38 UTC (History)
0 users

See Also:


Attachments

Note You need to log in before you can comment on or make changes to this bug.
Description kongyingqiao hygon_group 2025-09-04 10:33:44 UTC
NUMA-aware qspinlock has been integrated into the kernel starting for versions 5.10 and 6.6-dev. This patch is derived from the compact NUMA-aware lock and enhances qspinlock by considering NUMA node locality.
https://gitee.com/anolis/cloud-kernel/pulls/535
https://gitee.com/anolis/cloud-kernel/pulls/4406

On x86 platforms, where each CCX shares the same last-level cache (LLC), we further introduce CCX-aware qspinlock. This allows users to flexibly choose between CCX-awareness and NUMA-awareness depending on the platform characteristics and workload requirements.

In benchmarks on platforms with 8 physical cores per CCX, the CCX-aware qspinlock delivers over 20% performance improvement in MySQL OLTP read_write scenarios running across 6 CCXs.
Comment 1 小龙 admin 2025-09-05 18:38:34 UTC
The PR Link: https://gitee.com/anolis/cloud-kernel/pulls/5742