Re: [boost] [threadpool] new version 16: pining of processors(processor affinity)

Subject: Re: [boost] [threadpool] new version 16: pining of processors(processor affinity)
From: Oliver Kowalke (k-oli_at_[hidden])
Date: 2008-11-17 02:04:04

> k-oli_at_[hidden] wrote:
> 
> > Yes - internaly a work-stealing algorithm is used and thatswhy pining
> threads 
> > to processors improves the performance.
> 
> Couldn't it also be inefficient for some patterns?

You are not frced to used the pining. If you specify poolsize in the cosntructor no pining will happen.

Boost.Threadpool uses internaly a work-stealing algorithm. Each worker-thread has its own local queue (containing tasks) which will be filled by sub-tasks (please see the docu). The idea is that memory associated with the work most recently pushed into the queue will still be hot in the cache (queue == Processor affinity is optimizing cache performance (even time-sensitive applications). 

Oliver
-- 
Der GMX SmartSurfer hilft bis zu 70% Ihrer Onlinekosten zu sparen! 
Ideal für Modem und ISDN: http://www.gmx.net/de/go/smartsurfer