[Boost-users] Re: [Signals] Observation about performance

From: Keith MacDonald (boost_at_[hidden])
Date: 2005-05-10 09:25:00

Doug,

I omitted to state that I linked with Boost.Signals from the current CVS 
head.  Is that different from 1.33.0?

Keith MacDonald

"Doug Gregor" <dgregor_at_[hidden]> wrote in message 
news:9d494fe2c9d82d8a37c8c9a18f1b553e_at_cs.indiana.edu...
>
> On May 10, 2005, at 7:11 AM, Keith MacDonald wrote:
>
>> I have an application that implements the Observer pattern, based on the 
>> GOF
>> code and was interested in using Boost.Signals in a new application. 
>> Before
>> committing myself, I decided to check the relative performance when
>> signalling two observers.  Both tests were performed on Windows XP, using
>> MSVC 8 beta 2, with the /Ox optimisation.  Here are the times for 
>> 6,000,000
>> signals, measured with the Win32 GetTickCount() API on a 3GHz P4 with 1GB
>> RAM:
>>
>>    GOF Observer:   109
>>    Boost.Signals: 6203
>>
>> Clearly, you have to trade convenience for performance with this library.
>
> The performance of Signals has been drastically improved in 1.33.0 thanks 
> to a patch from Robert Zeh. It may still not be as quick as a specific 
> implementation of the Observer pattern, but it's much more reasonable now.
>
> Doug