Re: [PATCH 0/4] sched: Various reweight_entity() fixes

[Dietmar Eggemann]

On 16.02.26 11:59, Dietmar Eggemann wrote:
> On 16.02.26 04:14, K Prateek Nayak wrote:
>> Hello Shubhang,
>>
>> On 2/14/2026 12:50 PM, Shubhang Kaushik wrote:
>>> Hi Peter,
>>>
>>> On Fri, 30 Jan 2026, Peter Zijlstra wrote:

[...]

> I saw the same issue on my Arm64 big.Little board (12 CPUs) with the old
> 'queue.git sched/core' (i.e before s/div_s64()/div64_long()) in:
> 
> sched/fair: Increase weight bits for avg_vruntime
> 
> and test:
> 
> for ((i=0; i<20000; i++)) do yes > /dev/null & done
> 
> It's gone with:
> 
>   git.kernel.org/pub/scm/linux/kernel/git/peterz/queue.git sched/core
> 
>   commit bdba3187771c ("sched/fair: Use full weight to __calc_delta()")
> 
> Tested-by: Dietmar Eggemann <dietmar.eggemann@xxxxxxx>
> 
> [...]

No regression spotted on 'hammerdb - mysqld' ('client/LDG - server/SUT'
running on different ARM64 VMs)

AWS VM LDG (c7g.16xlarge) and SUT (m7gd.16xlarge ('maxcpus=16 nr_cpus=16').

Kernel              |  Throughput | P50 latency.      |
aarch64             |  (NOPM)     | (larger is worse) |
--------------------+-------------+-------------------+
  tip sched/core *  |   baseline  |     baseline      |
queue sched/core ** |    +0.8%    |       -1.0%       |
--------------------+-------------+--------------------

*  e34881c84c25 - sched: Re-evaluate scheduling when migrating queued
tasks out of throttled cgroups (2026-02-03 Zicheng Qu)

** bdba3187771c - sched/fair: Use full weight to __calc_delta()
(2026-02-13 Peter Zijlstra)