From owner-freebsd-hardware Sun Jul 30 15:22:51 1995 Return-Path: hardware-owner Received: (from majordom@localhost) by freefall.cdrom.com (8.6.11/8.6.6) id PAA26388 for hardware-outgoing; Sun, 30 Jul 1995 15:22:51 -0700 Received: from godzilla.zeta.org.au (godzilla.zeta.org.au [203.2.228.34]) by freefall.cdrom.com (8.6.11/8.6.6) with ESMTP id PAA26382 for ; Sun, 30 Jul 1995 15:22:44 -0700 Received: (from bde@localhost) by godzilla.zeta.org.au (8.6.9/8.6.9) id IAA18753; Mon, 31 Jul 1995 08:18:58 +1000 Date: Mon, 31 Jul 1995 08:18:58 +1000 From: Bruce Evans Message-Id: <199507302218.IAA18753@godzilla.zeta.org.au> To: freebsd-hardware@freebsd.org, knarf@nasim.cube.net Subject: Re: multi IO boards Sender: hardware-owner@freebsd.org Precedence: bulk >> There's a number of 8-port AST compatibles that work quite well; you can >> get them with 16550's already onboard. (Mandatory for FreeBSD). >Mandatory for FreeBSD? Why? 16550's aren't mandatory for FreeBSD. They are mandatory for Windows and for certain Unux systems with a high interrupt latency. For DOS they are mandatory unless the serial application does something about all the interrupt handlers that turn off interrupts for too long, starting with the keyboard interrupt handler. This is probably too hard to do if DOS is run in V86 mode. >I currently have eight serial ports in my machine. One four port card >with four interrupts and four 16550A (two in one chips) and one AST/4 card >with two 16550A and two 16450. Works fine with five modems and a 115k2 >nullmodem connection right now. It won't work at full speed unless all the ports are never active in both directions at the same time. One port at 115200 bps bidirectional takes about 30% of a 486DX2/66 for interrupt overhead and overhead up to the read()/write() level. 8 * 30% is much larger 100%. With 16550s the overhead is about 12% per port so you can expect to have at least 4% available for doing real work while 8 ports are active :-). Bruce