----- Original Message ----- From: "Oliver Kowalke" <k-oli@gmx.de> To: <boost@lists.boost.org> Sent: Wednesday, November 05, 2008 9:30 AM Subject: Re: [boost] [threadpool] mixture of recursive invokation and fibers?
Could you tell me more about when we need more that one pool (piplined architectures for instance)?
Pipelined architectures could be used in network-applications/services (for instance web-services). The idea is to parallelize the receiving, parsing, processing of network-messages in different stages. Each stage could have its own threadpool (the stages maybe also use only one pool). The pattern comes from the pipeline architecture of modern RISC processors. A better description can be found here (chapter 6): http://www.cs.uwaterloo.ca/~brecht/papers/getpaper.php?file=eurosys-2007.pdf
What is the advantage to have one pool for each stage respect one pool for all the stages?
Even if this is the case, this do not forbid to have enablers/disablers taking the pool as parameter.
yes
If at the end we think that it would be interesting to have a default thread pool we can have namespace the_thread_pool which defines enablers/disablers respect to this default thread pool.
maybe we can hijack this_thread - a static threadpool with some default configuration (channel type etc.)?
Great, Vicente