Subj : Re: What is the real costs of LOCK on x86 multiprocesor machine?
To   : comp.programming.threads
From : Mirek Fidler
Date : Sun Jul 31 2005 02:51 pm

Joe Seigh wrote:
> Peter Dimov wrote:
> 
>> Mirek Fidler wrote:
>>
>>>> In practice, the actual amortized cost appears to be much lower than
>>>> 50-100x. I see a 4x difference on shared_ptr_timing_test (which
>>>> basically tests reference count increments and decrements in a somewhat
>>>> real-life worst case scenario.)
>>>
>>>
>>> Have you tested on real MP machine? (4x is close to what I have measured
>>> on my SP).
>>
>>
>>
>> No, it was a single P4/2GHz without HT.
>>
> 
> Now that a lot of people have weighed in on this, why does the cost of
> atomic increment/decrement matter?
> 

A lot. It is decisive factor in choosing COW structure over full copy.

Mirek

.