Date: Tue, 05 Dec 2006 12:29:14 +0100 From: "Joshua M" <katsuo_harada_evil_does@hotmail.com> To: julian@elischer.org Cc: freebsd-threads@freebsd.org Subject: Re: Threading arch quetions Message-ID: <BAY128-F94639FEBAE6BC000BBF8293DE0@phx.gbl> In-Reply-To: <457499BD.9080000@elischer.org>
next in thread | previous in thread | raw e-mail | index | archive | help
>From: Julian Elischer <julian@elischer.org> >To: Joshua M <katsuo_harada_evil_does@hotmail.com> >CC: freebsd-threads@freebsd.org >Subject: Re: Threading arch quetions >Date: Mon, 04 Dec 2006 13:57:17 -0800 >We adopted a scheme which would allow us to implement both M:N threads >and 1:1 threads with compatible libraries that do it each way. >The hope was that this would allow people to experiment with these. >and for us to be able to select the best approach. Thank you for your response. Can you please name those libs ? >There are examples where M:N outperforms 1:1 but they are the >minority, so we will be switching the default library to >1:1 threads What is the default library ? > >Note that the kernel support for threads is the same for both models, >where processes have sub-entities (kernel schedulable entities) called >threads, as opposed to in Linux where each thread is a separate process. This is the point i dont understand at all. As i understood KSE is an N to M approach and it is implemented in FBSD. So how 1:1 is supposed to coexist if finally everythig is converted to N to M ? does it mean that basically FBSD if it wants 1 to 1 always creates 1 KSEG per thread and thus create a *simulation* of 1 to 1 approach ? Thank you again. _________________________________________________________________ Windows Live Spaces : créez votre Space à votre image ! http://www.windowslivespaces.fr/
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?BAY128-F94639FEBAE6BC000BBF8293DE0>