3d9
Subj : Re: load-after-store reordering for x86 wrt smr...
To   : comp.programming.threads
From : Joe Seigh
Date : Tue Mar 15 2005 07:34 am

On Mon, 14 Mar 2005 20:21:05 -0800, SenderX <xxx@xxx.com> wrote:

>>> Do you think that the store fence is needed in line 8:
>>>
>>>
>> [...]
>>>
>>>
>>> Line 7 would be the first store, line 9 would be the load-after-store
>>> from
>>> another location condition. It seems that smr requires explicit barriers
>>> on
>>> x86 to prevent the algorithm from crashing...
>>>
>>>
>>>
>> It has to be an mfence.  sfence is just store/store.
>
> I made the change, and now as I expected... Iterating a huge linked-list
> using hazard-pointers is more expensive than using a rw-mutex; you need a
> fence for each item! It looks like proxy collectors are the way to go...
>
Fortunately, most architectures support dependent load memory ordering,
otherwise you'd have to use load/load membars for them.

-- 
Joe Seigh

.

0