64b
Subj : Re: pthread library architecture (threading model)
To   : comp.programming.threads
From : David Butenhof
Date : Mon Sep 19 2005 01:48 pm

pankaj-startup wrote:
>> The pthread API doesn't imply any particular threading model. Implementations
>> on other platforms provide 1:1, m:n, and m:1 models.
>> --
> 
> are there any pthread apis to choose one of these models?

POSIX thread implementations may support "1:1", called "System 
Contention Scope", or "M:N", called "Process Contention Scope". "M:1", 
or multiplexing on top of a single "kernel scheduling entity" cannot 
generally be done without special kernel support that would really make 
it more of a "M:N" -- a pure "M:1" implementation would allow 
indefinitely blocking one thread behind another due to problems in 
preempting UNIX synchronous syscalls, and would violate POSIX requirements.

Relatively few implementations support both System and Process 
contention scopes, and there's no API to determine which are supported. 
You can however ask for one or the other explicitly using the 
pthread_attr_setscope() operation on an attributes object that you use 
to create threads. Check your documentation. (Note that since contention 
scope is a scheduling attribute, it may be inherited from the creating 
thread by default; always use pthread_attr_setinheritsched() to 
PTHREAD_EXPLICIT_SCHED.)

-- 
Dave Butenhof, David.Butenhof@hp.com
HP Utility Pricing software, POSIX thread consultant
Manageability Solutions Lab (MSL), Hewlett-Packard Company
110 Spit Brook Road, ZK2/3-Q18, Nashua, NH 03062

.

0