From: Peter Zijlstra <peterz@infradead.org>
To: Jianyong Wu <wujianyong@hygon.cn>
Cc: Ingo Molnar <mingo@redhat.com>,
Juri Lelli <juri.lelli@redhat.com>,
Vincent Guittot <vincent.guittot@linaro.org>,
Chen Yu <yu.c.chen@intel.com>,
Tim Chen <tim.c.chen@linux.intel.com>,
Dietmar Eggemann <dietmar.eggemann@arm.com>,
Steven Rostedt <rostedt@goodmis.org>,
Ben Segall <bsegall@google.com>, Mel Gorman <mgorman@suse.de>,
Valentin Schneider <vschneid@redhat.com>,
K Prateek Nayak <kprateek.nayak@amd.com>,
Shrikanth Hegde <sshegde@linux.ibm.com>,
Phil Auld <pauld@redhat.com>,
Andrew Morton <akpm@linux-foundation.org>,
David Hildenbrand <david@kernel.org>,
linux-kernel@vger.kernel.org, linux-mm@kvack.org,
jianyong.wu@outlook.com, zhongyuan@hygon.cn, huangsj@hygon.cn,
wangfengyu@hygon.cn, yingzhiwei@hygon.cn, justin.he@arm.com
Subject: Re: [RFC PATCH v2 04/23] sched/topology: Introduce a method to calculate the llc distance
Date: Mon, 31 Aug 2026 15:12:44 +0200 [thread overview]
Message-ID: <20260831131244.GJ4120091@noisy.programming.kicks-ass.net> (raw)
In-Reply-To: <20260827122816.756234-5-wujianyong@hygon.cn>
On Thu, Aug 27, 2026 at 08:27:57PM +0800, Jianyong Wu wrote:
> The previous patch addresses NUMA inter-node distance handling. This
> patch shifts focus to the LLC layer and introduces a scheme to compute
> pairwise distances between sibling LLCs inside one NUMA node.
> The calculation only guarantees distinct distance values for all peers
> when viewed from a single reference LLC; it does not enforce globally
> unique distances for every arbitrary LLC pair on the node.
>
> One valid calculation scheme is defined by the formula below:
>
> dist(LLC1, LLC2) = (rank1 + rank2) % k + 1
>
> Constraints: LLC1 and LLC2 must reside on the identical NUMA node.
> rank1 = local index of LLC1 within its node; rank2 follows the same rule.
> k = total count of LLCs available on the node.
>
> The formula never yields zero, so diagonal self-distance entries are
> explicitly hardcoded to 0.
>
> For demonstration, consider a NUMA node with four LLCs (local ranks 0~3):
> LLC0 LLC1 LLC2 LLC3
> LLC0 0 2 3 4
> LLC1 2 0 4 1
> LLC2 3 4 0 2
> LLC3 4 1 2 0
>
> Sorting each matrix row in ascending order generates a unique LLC affinity
> sequence, which drives thread migration policies implemented in follow-up
> patches.
Aah, now I see. I got confused by your sched_cache_node_distance naming,
and thought it was cache-to-cache distance.
next prev parent reply other threads:[~2026-08-31 13:12 UTC|newest]
Thread overview: 72+ messages / expand[flat|nested] mbox.gz Atom feed top
2026-08-27 12:27 [RFC PATCH v2 00/23] sched: Scale cache-aware aggregation at LLC granularity Jianyong Wu
2026-08-27 12:27 ` [RFC PATCH v2 01/23] sched/topology: Add llc_to_node() to translate LLC id to NUMA node Jianyong Wu
2026-08-29 10:31 ` Peter Zijlstra
2026-08-31 9:43 ` Jianyong Wu
2026-08-27 12:27 ` [RFC PATCH v2 02/23] sched/topology: Introduce a NUMA distance matrix with unique distance values Jianyong Wu
2026-08-31 11:50 ` Peter Zijlstra
2026-09-01 6:57 ` Jianyong Wu
2026-09-01 7:13 ` Peter Zijlstra
2026-09-01 7:37 ` Jianyong Wu
2026-09-22 18:38 ` Tim Chen
2026-09-23 3:14 ` Jianyong Wu
2026-09-23 18:29 ` Tim Chen
2026-09-24 5:41 ` Jianyong Wu
2026-09-24 15:46 ` Tim Chen
2026-09-28 9:39 ` Jianyong Wu
2026-10-01 22:04 ` Tim Chen
2026-10-08 7:35 ` Jianyong Wu
2026-10-08 16:35 ` Chen Yu
2026-09-01 8:48 ` Peter Zijlstra
2026-09-02 7:11 ` Jianyong Wu
2026-08-27 12:27 ` [RFC PATCH v2 03/23] sched/topology: Introduce a macro to traverse node Jianyong Wu
2026-08-27 12:27 ` [RFC PATCH v2 04/23] sched/topology: Introduce a method to calculate the llc distance Jianyong Wu
2026-08-31 13:12 ` Peter Zijlstra [this message]
2026-08-27 12:27 ` [RFC PATCH v2 05/23] sched/topology: Introduce a macro to traverse LLC inside node Jianyong Wu
2026-08-27 12:27 ` [RFC PATCH v2 06/23] sched/topology: Add sd_node for the NODE sched domain Jianyong Wu
2026-08-27 12:28 ` [RFC PATCH v2 07/23] sched/cache: Prioritize preferred NUMA node selection over LLC selection Jianyong Wu
2026-08-31 13:16 ` Peter Zijlstra
2026-09-01 7:44 ` Jianyong Wu
2026-08-31 13:22 ` Peter Zijlstra
2026-09-01 8:05 ` Jianyong Wu
2026-08-27 12:28 ` [RFC PATCH v2 08/23] sched/topology: Introduce a per-CPU tasks NUMA preferred counter Jianyong Wu
2026-08-31 13:23 ` Peter Zijlstra
2026-09-01 8:14 ` Jianyong Wu
2026-08-31 13:24 ` Peter Zijlstra
2026-09-01 8:31 ` Jianyong Wu
2026-09-01 10:21 ` Peter Zijlstra
2026-09-01 13:02 ` Jianyong Wu
2026-08-27 12:28 ` [RFC PATCH v2 09/23] sched/cache: Account percpu sd task NUMA preference Jianyong Wu
2026-09-01 7:54 ` Peter Zijlstra
2026-09-01 8:41 ` Jianyong Wu
2026-08-27 12:28 ` [RFC PATCH v2 10/23] sched/topology: Add per-sd scratch for the load balance affinity score Jianyong Wu
2026-09-01 8:02 ` Peter Zijlstra
2026-09-01 11:55 ` Jianyong Wu
2026-08-27 12:28 ` [RFC PATCH v2 11/23] sched/cache: Introduce helpers for task migration decisions Jianyong Wu
2026-09-01 9:08 ` Peter Zijlstra
2026-09-02 5:08 ` Jianyong Wu
2026-09-01 11:32 ` Peter Zijlstra
2026-09-02 5:46 ` Jianyong Wu
2026-09-02 21:11 ` Tim Chen
2026-09-03 2:04 ` Jianyong Wu
2026-08-27 12:28 ` [RFC PATCH v2 12/23] sched/cache: Introduce rq affinity gain calculation Jianyong Wu
2026-09-01 9:58 ` Peter Zijlstra
2026-09-01 12:23 ` Jianyong Wu
2026-09-01 10:16 ` Peter Zijlstra
2026-09-01 12:34 ` Jianyong Wu
2026-08-27 12:28 ` [RFC PATCH v2 13/23] sched/cache: Pick optimal src rq/group using affinity promotion metric Jianyong Wu
2026-08-27 12:28 ` [RFC PATCH v2 14/23] sched/cache: Drop prefer_sibling restriction for llc_balance Jianyong Wu
2026-09-01 10:29 ` Peter Zijlstra
2026-09-01 13:25 ` Jianyong Wu
2026-08-28 1:58 ` [RFC PATCH v2 15/23] sched/cache: Judge migration eligibility in LLC granularity Jianyong Wu
2026-08-28 2:04 ` [RFC PATCH v2 16/23] sched/cache: Allow un-throttled active balance to spread out of a full LLC Jianyong Wu
2026-08-28 2:07 ` [RFC PATCH v2 17/23] sched/fair: Fine-granularity NUMA balancing Jianyong Wu
2026-09-01 12:47 ` Peter Zijlstra
2026-09-02 6:43 ` Jianyong Wu
2026-08-28 2:09 ` [RFC PATCH v2 18/23] sched/cache: Scan all prefer nodes in thread group Jianyong Wu
2026-08-28 2:10 ` [RFC PATCH v2 19/23] sched/cache: Remove preferred LLC/node check no longer needed Jianyong Wu
2026-08-28 2:11 ` [RFC PATCH v2 20/23] sched/cache: Estimate utilization of the whole thread group Jianyong Wu
2026-09-01 14:44 ` Peter Zijlstra
2026-09-08 7:43 ` Jianyong Wu
2026-08-28 2:13 ` [RFC PATCH v2 21/23] sched/cache: Spread workloads within an estimated LLC range Jianyong Wu
2026-08-28 2:14 ` [RFC PATCH v2 22/23] sched/cache: Walk the preferred node from the preferred LLC Jianyong Wu
2026-08-28 2:15 ` [RFC PATCH v2 23/23] sched/debug: Print task preferred LLC for scheduler debugging Jianyong Wu
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=20260831131244.GJ4120091@noisy.programming.kicks-ass.net \
--to=peterz@infradead.org \
--cc=akpm@linux-foundation.org \
--cc=bsegall@google.com \
--cc=david@kernel.org \
--cc=dietmar.eggemann@arm.com \
--cc=huangsj@hygon.cn \
--cc=jianyong.wu@outlook.com \
--cc=juri.lelli@redhat.com \
--cc=justin.he@arm.com \
--cc=kprateek.nayak@amd.com \
--cc=linux-kernel@vger.kernel.org \
--cc=linux-mm@kvack.org \
--cc=mgorman@suse.de \
--cc=mingo@redhat.com \
--cc=pauld@redhat.com \
--cc=rostedt@goodmis.org \
--cc=sshegde@linux.ibm.com \
--cc=tim.c.chen@linux.intel.com \
--cc=vincent.guittot@linaro.org \
--cc=vschneid@redhat.com \
--cc=wangfengyu@hygon.cn \
--cc=wujianyong@hygon.cn \
--cc=yingzhiwei@hygon.cn \
--cc=yu.c.chen@intel.com \
--cc=zhongyuan@hygon.cn \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.