Re: [PATCH v3] mm/khugepaged: sched to numa node when collapse huge page

[maobibo]

在 2022/5/13 09:19, maobibo 写道:
> 
> 
> 在 2022/5/13 08:36, Andrew Morton 写道:
>> On Thu, 28 Apr 2022 12:34:07 -0400 Peter Xu <peterx@xxxxxxxxxx> wrote:
>>
>>> On Thu, Apr 28, 2022 at 05:17:07PM +0200, David Hildenbrand wrote:
>>>> On 17.03.22 07:50, Bibo Mao wrote:
>>>>> collapse huge page will copy huge page from general small pages,
>>>>> dest node is calculated from most one of source pages, however
>>>>> THP daemon is not scheduled on dest node. The performance may be
>>>>> poor since huge page copying across nodes, also cache is not used
>>>>> for target node. With this patch, khugepaged daemon switches to
>>>>> the same numa node with huge page. It saves copying time and makes
>>>>> use of local cache better.
>>>>>
>>>>> With this patch, specint 2006 base performance is improved with 6%
>>>>> on Loongson 3C5000L platform with 32 cores and 8 numa nodes.
>>>>
>>>> If it helps, that's nice as long as it doesn't hurt other cases.
>>>>
>>
>> Quite a bit of doubtful feedback and we have yet to hear from the
>> author.  I'll drop the patch.
>>
>> Bibo, please resend at a later time if you feel the patch remains
>> desirable.  Please attempt to address the feedback via code changes
>> and/or changelogging.
> Sorry for the late response, the mail is filtered and I did not notice that. The result is not so obvious after bandwidth is improved between nodes, it is only about 1% improvement for specint2006 for 32 core on my box.
> 
> Now I do not see negative effective about this patch unless user wants to keep some cores separated from khugepaged daemon process.

Can we provide an extra parameter to let khugepaged daemon scheduling binded to node or freely? If can, I will provide updated patch.

> 
> 
> regards
> bibo,mao