proposed smp_rendezvous change
Andriy Gapon
avg at FreeBSD.org
Tue May 17 20:16:26 UTC 2011
on 17/05/2011 18:51 John Baldwin said the following:
> On Tuesday, May 17, 2011 10:34:41 am Andriy Gapon wrote:
>> on 17/05/2011 16:58 John Baldwin said the following:
>>> No, it doesn't quite work that way. It wouldn't work on Alpha for example.
>>>
>>> All load_acq is a load with a memory barrier to order other loads after it.
>>> It is still free to load stale data. Only a read-modify-write operation
>>> would actually block until it could access an up-to-date value.
>>
>> Hmm, ok.
>> How about atomic_add_acq_int(&smp_rv_waiters[0], 0) ? :-)
>> Or an equivalent MI action that doesn't actually change smp_rv_waiters[0] value,
>> if there could be any.
>> Maybe explicit atomic_cmpset_acq_int(&smp_rv_waiters[0], 0, 0) ?
>>
>> You see at what I am getting?
>
> Yeah, either of those would work. At this point just leaving the
> atomic_add_int() as-is would be the smallest diff. :)
Yes, I agree about the smallest diff.
But if we are going to remove that first spin-loop, then we would effectively be
wasting one volatile variable. Only a tiny waste, but it could be confusing.
Whatever we decide, this was very educational for me.
--
Andriy Gapon
More information about the freebsd-current
mailing list