From owner-freebsd-hackers Sun May 5 13:53:58 1996 Return-Path: owner-hackers Received: (from root@localhost) by freefall.freebsd.org (8.7.3/8.7.3) id NAA09258 for hackers-outgoing; Sun, 5 May 1996 13:53:58 -0700 (PDT) Received: from mail.think.com (Mail1.Think.COM [131.239.33.245]) by freefall.freebsd.org (8.7.3/8.7.3) with SMTP id NAA09252 for ; Sun, 5 May 1996 13:53:55 -0700 (PDT) Received: from Early-Bird-1.Think.COM by mail.think.com; Sun, 5 May 96 16:53:54 -0400 Received: from by Early-Bird.Think.COM; Sun, 5 May 96 16:53:52 EDT Received: (from alk@localhost) by compound.Think.COM (8.7.5/8.7.3) id DAA06080; Sun, 5 May 1996 03:47:45 -0500 (CDT) Date: Sun, 5 May 1996 03:47:45 -0500 (CDT) Message-Id: <199605050847.DAA06080@compound.Think.COM> From: Tony Kimball To: hackers@freebsd.org Subject: co-scheduling Sender: owner-hackers@freebsd.org X-Loop: FreeBSD.org Precedence: bulk I'd really like to co-schedule processes on LAN nodes. To be clear: I want to insure that the processes of a group of processes distributed one-to-one on a set of FBSD boxes in a LAN execute with a maximal degree of synchronous overlap. The reason is that the processes are exchanging data at a high rate, and cannot make progress unless data is exchanged. When operating with vanishing orthogonal load, this is not a problem; however, when the orthogonal load increases, the slowdown is, ermmm, impressive. Can anyone recommend the most expeditious means of implementing such a co-scheduling facility? On a related note, I am curious whether there would be much interest in an HPF compiler for x86 BSD networks, given that it would only be available in binary form, were it made freely available. Oh, and if anyone wants to *buy* one, do please let me know:-) Finally, are there to be recommended any examples of code which sends/recvs ethernet pakets using exclusively user-space code, especially using 100bT devices? //alk