Re: [boost] [threadpool] parallel_sort example

Subject: Re: [boost] [threadpool] parallel_sort example
From: vicente.botet (vicente.botet_at_[hidden])
Date: 2009-03-02 15:39:13

----- Original Message ----- 
From: <k-oli_at_[hidden]>
To: <boost_at_[hidden]>
Sent: Monday, March 02, 2009 7:20 PM
Subject: Re: [boost] [threadpool] parallel_sort example

> 
> Am Monday 02 March 2009 00:06:25 schrieb Edouard A.:
>> I also think that having a default pool is a must. What you can do is have
>> a boost::tp:get_default_pool() that would be using Meiers idiom for
>> singleton's (i.e. having the function returning a static local variable).
> 
> How many worker-threads should be created inside the default threadpool?
> Which scheduling strategy should by applied?
> 
> The pool could use FIFO ordering of the queued tasks but the amount of 
> worker-threads depend on the threadpool usage.
> THe pool could create as many worker-threads as CPUs are online an bind each 
> thread to one specific CPU/Core - CPU intensive apps could benefit form this 
> scenario. Applications with lot of IO require usally more worker-threads than 
> CPUs.
> 
> Hmmm

Hi Olivier, 

You are right the number of CPUs, depend on the ussage. The fact that the default pool should be configurable do not implies that we cannot have one. The configuration can be done either by compilations flags, or the user can change the features of the default pool at runtime;  I don't know there are a lot of possibilities. What is clear for me is that the user needs this default. 

Best,
Vicente