[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [Xen-devel] [PATCHv2 5/5] evtchn: pad struct evtchn to 64 bytes



On 16/06/15 13:57, Julien Grall wrote:
> On 16/06/15 12:59, Jan Beulich wrote:
>>>>> On 16.06.15 at 13:14, <julien.grall@xxxxxxxxxx> wrote:
>>> On 15/06/2015 16:48, David Vrabel wrote:
>>>> diff --git a/xen/include/xen/sched.h b/xen/include/xen/sched.h
>>>> index 44ea92d..a0ff9d2 100644
>>>> --- a/xen/include/xen/sched.h
>>>> +++ b/xen/include/xen/sched.h
>>>> @@ -129,7 +129,7 @@ struct evtchn
>>>>   #endif
>>>>       } ssid;
>>>>   #endif
>>>> -};
>>>> +} __attribute__((aligned(64)));
>>>
>>> Why don't you use __cacheline_aligned?
>>
>> That would double the size on x86, for little or no benefit.
> 
> Well, the cache line size is not necessarily 64 bytes on every
> architecture. In the case of ARM, the cache line depends on the
> processor version.
> 
> __cacheline_aligned is the only way to ensure that the cache line is not
> shared on ARM.
> 
> AFAIU, the goal of this patch is to avoid sharing the cache line. If
> not, the commit message is misleading because it claims that a cache
> line is always 64 bytes...

We want to avoid sharing the cache line where we can do so for no
additional memory cost.

David

_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxx
http://lists.xen.org/xen-devel


 


Rackspace

Lists.xenproject.org is hosted with RackSpace, monitoring our
servers 24x7x365 and backed by RackSpace's Fanatical Support®.