From owner-freebsd-current  Sat Nov  7 04:44:48 1998
Return-Path: <owner-freebsd-current@FreeBSD.ORG>
Received: (from majordom@localhost)
          by hub.freebsd.org (8.8.8/8.8.8) id EAA01198
          for freebsd-current-outgoing; Sat, 7 Nov 1998 04:44:48 -0800 (PST)
          (envelope-from owner-freebsd-current@FreeBSD.ORG)
Received: from post.mail.demon.net (post-12.mail.demon.net [194.217.242.41])
          by hub.freebsd.org (8.8.8/8.8.8) with ESMTP id EAA01189
          for <current@freebsd.org>; Sat, 7 Nov 1998 04:44:46 -0800 (PST)
          (envelope-from james@westongold.com)
Received: from [158.152.96.124] (helo=wgp01.wgold.demon.co.uk)
	by post.mail.demon.net with esmtp (Exim 2.05demon1 #1)
	id 0zc7j7-0004tN-00; Sat, 7 Nov 1998 12:44:26 +0000
Received: by WGP01 with Internet Mail Service (5.5.1960.3)
	id <W2PQAPHH>; Sat, 7 Nov 1998 12:57:15 -0000
Message-ID: <32BABEF63EAED111B2C5204C4F4F5020180B@WGP01>
From: James Mansion <james@westongold.com>
To: Daniel Eischen <eischen@vigrid.com>,
        James Mansion
	 <james@westongold.com>, peter@netplex.com.au
Cc: current@FreeBSD.ORG, jb@cimlogic.com.au, lists@tar.com
Subject: RE: Kernel threading (was Re: Thread Scheduler bug)
Date: Sat, 7 Nov 1998 12:57:08 -0000 
MIME-Version: 1.0
X-Mailer: Internet Mail Service (5.5.1960.3)
Content-Type: text/plain
Sender: owner-freebsd-current@FreeBSD.ORG
Precedence: bulk
X-Loop: FreeBSD.ORG

> From: Daniel Eischen [mailto:eischen@vigrid.com]
> > *My* concern is that pthread_self, and access to 
> thread-specific data,
> > should be as fast as possible.  Writing thread-hot libraries without
> > good thread specific data is irksome to say the least.
> 
> My point was that you can't have just one common pointer (address)
> to TSD that is changed on thread schedule as it would limit you
> to being able to execute only one thread per process at a time.
> To take advantage of multiple processors, you'd need at least
> as many TSD pointers as CPUs.  Julian discussed this in a previous
> response.

Sure you can.  But you can't share the same page map between all
the threads (or at least between all the kernel threads that are
executing in the process).

The costs have been discussed.  I didn't say it was going to be
convenient or that it wouldn't make a difference to the cost of
rescheduling a kernel thread onto a new user thread.

Whether these costs are worthwhile, or whether the same effect can
be achieved more effectively, is surely the point of the discussion.

James

To Unsubscribe: send mail to majordomo@FreeBSD.org
with "unsubscribe freebsd-current" in the body of the message