lkml.org 
[lkml]   [2018]   [Jan]   [1]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
SubjectRe: [PATCH v5 7/9] arm64: Topology, rename cluster_id
From
Date
Hi,

On 2017/12/18 20:42, Morten Rasmussen wrote:
> On Fri, Dec 15, 2017 at 10:36:35AM -0600, Jeremy Linton wrote:
>> Hi,
>>
>> On 12/13/2017 12:02 PM, Lorenzo Pieralisi wrote:
>>> [+Morten, Dietmar]
>>>
>>> $SUBJECT should be:
>>>
>>> arm64: topology: rename cluster_id
>>
[cut]
>>
>> I was hoping someone else would comment here, but my take at this point is
>> that it doesn't really matter in a functional sense at the moment.
>> Like the chiplet discussion it can be the subject of a future patch along
>> with the patches which tweak the scheduler to understand the split.
>>
>> BTW, given that i'm OoO next week, and the following that are the holidays,
>> I don't intend to repost this for a couple weeks. I don't think there are
>> any issues with this set.
>>
>>>
>>> There is also arch/arm to take into account, again, this patch is
>>> just renaming (as it should have named since the beginning) a
>>> topology level but we should consider everything from a legacy
>>> perspective.
>
> arch/arm has gone for thread/core/socket for the three topology levels
> it supports.
>
> I'm not sure what short term value keeping cluster_id has? Isn't it just
> about where we make the package = cluster assignment? Currently it is in
> the definition of topology_physical_package_id. If we keep cluster_id
> and add package_id, it gets moved into the MPIDR/DT parsing code.
>
> Keeping cluster_id and introducing a topology_cluster_id function could
> help cleaning up some of the users of topology_physical_package_id that
> currently assumes package_id == cluster_id though.

I think we still need the information describing which cores are in one cluster.
Many arm64 chips have the architecture core/cluster/socket. Cores in one cluster may
share a same L2 cache. That information can be used to build the sched_domain. If we put
cores in one cluster in one sched_domain, the performance will be better.(please see
kernel/sched/topology.c:1197, cpu_coregroup_mask() uses 'core_sibling' to build a multi-core
sched_domain)
So I think we still need variable to record which cores are in one sched_domain for future use.

Thanks,
Xiongfeng

>
> Morten
>
> .
>

\
 
 \ /
  Last update: 2018-01-02 03:30    [W:0.101 / U:4.892 seconds]
©2003-2018 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site