From owner-freebsd-hardware  Wed Jul  3 13:51:23 1996
Return-Path: owner-hardware
Received: (from root@localhost)
          by freefall.freebsd.org (8.7.5/8.7.3) id NAA24465
          for hardware-outgoing; Wed, 3 Jul 1996 13:51:23 -0700 (PDT)
Received: from godzilla.zeta.org.au (godzilla.zeta.org.au [203.2.228.19])
          by freefall.freebsd.org (8.7.5/8.7.3) with SMTP id NAA24458
          for <hardware@freebsd.org>; Wed, 3 Jul 1996 13:51:20 -0700 (PDT)
Received: (from bde@localhost) by godzilla.zeta.org.au (8.6.12/8.6.9) id GAA15967; Thu, 4 Jul 1996 06:47:36 +1000
Date: Thu, 4 Jul 1996 06:47:36 +1000
From: Bruce Evans <bde@zeta.org.au>
Message-Id: <199607032047.GAA15967@godzilla.zeta.org.au>
To: msmith@atrad.adelaide.edu.au, stesin@elvisti.kiev.ua
Subject: Re: muliport boards - building a PPP dialup server
Cc: hardware@freebsd.org
Sender: owner-hardware@freebsd.org
X-Loop: FreeBSD.org
Precedence: bulk

># > 	1. tty-level driver overhead, that's what I'm watching
># > 	   now with 16 FIFOed ports in 486dx4/100 -- interrupt
># > 	   load is tiny, and most of CPU is eaten by system.
># > 	   "Smart" card with it's own CPU (I treat it like
># > 	   an I/O co-processor) should minimize this factor. (?)
># 
># No.  None of the 'smart' cards currently around perform any tty processing.

>	Won't tty-level overhead decrease when the data flow through a port
>	will be going in bigger chunks?

Only if 'smart' actually means 'fast' :-).  16550 output already has a
chunk size of 16 bytes.  Since 8-bit ISA I/O is so slow, this size is
large enough to get most of the benefits of chunking.  16550 input is
about twice as slow (on >= 486's) as 16550 output because the status
register has to be read for every byte of input.  The input chunk size
is large enough for the UART interface to become the bottleneck.

># > 	2. ISA bus itself, no matter what will I put into it.
># > 	   ...  Yes, probably I'm
># > 	   wrong and ISA can deal i.e. with 32x115200; but how
># > 	   reliable will this configuration be?
># 
># This is a significant issue.  32x115200 is ~350K/sec. presuming 100%
># efficiency.  IIRC, Bruce quoted more like 50%, so the ISA bus would be

You forgot output.  32x115200bps is ~700K/sec.  16 ports is already
350K/sec.  This will take about 80% of the CPU for ISA I/O alone
(assuming 16550s and 2 I/O's per byte).

Bruce