*BSD News Article 37286


Return to BSD News archive

Xref: sserve comp.os.linux.development:18485 comp.os.linux.misc:28575 comp.os.386bsd.questions:14160 comp.os.386bsd.misc:3908 sci.electronics:83040
Path: sserve!newshost.anu.edu.au!harbinger.cc.monash.edu.au!msuinfo!uwm.edu!news.alpha.net!news.mathworks.com!europa.eng.gtefsd.com!newsxfer.itd.umich.edu!gumby!andrews-cc!gillham
From: gillham@andrews.edu (Andrew Gillham)
Newsgroups: comp.os.linux.development,comp.os.linux.misc,comp.os.386bsd.questions,comp.os.386bsd.misc,sci.electronics
Subject: Re: 16550 detection
Date: 2 Nov 1994 01:33:06 GMT
Organization: Andrews University
Lines: 35
Message-ID: <396q8i$fap@orion.cc.andrews.edu>
References: <CMETZ.94Oct30051603@itchy.inner.net> <TYTSO.94Oct31162517@dcl.mit.edu> <MICHAELV.94Oct31211019@MindBender.HeadCandy.com> <TYTSO.94Nov1182557@dcl.mit.edu>
NNTP-Posting-Host: edmund.cs.andrews.edu

In article <TYTSO.94Nov1182557@dcl.mit.edu> tytso@athena.mit.edu (Theodore Y. Ts'o) writes:

[stuff that almost sounded like Hayes bashing.. :-) ]

>
>						- Ted
>
>P.S.  That's not to say that a larger FIFO buffer wouldn't be helpful to
>avoid problems.  If you have some interrupt latency problems, and a
>super fast modem, and a slow CPU, the larger FIFO is undoubtedly useful.
>But that presupposes other problems in your operating system, either
>software or hardware, to begin with.  On many systems, the 1024 byte
>FIFO is just loafing...

Not to sound stupid, but why does everybody refer to "interrupt latency"
and CPU/MODEM speed and make it sound like they're only talking about
*ONE* serial port?  Seems to me that if you have 8 ports and want to
run them all at 115200, then you damn well better have a larger FIFO!
I don't know what the interrupt latency numbers look like for NetBSD or
Linux, but I don't imagine the serial port performance is very scalable
with plain old NS16550A's?  1 port at 115kbps should be easy, but what
about 2, 4, 8, 16?  I'd take the 1024 byte fifo any day!  Especially
if I could program the board with separate trigger levels for different
ports.  i.e. 10 byte trigger for an interactive terminal, but maybe a
256 byte trigger on a "non-interactive" device.
On a side note, does anybodyng a multi-port serial card want to
comment on performance with 2-4+ serial ports going full blast?

-Andrew

-- 
==========================================================
Andrew Gillham                       gillham@andrews.edu
LAN/WAN/Netware/Unix Analyst         gillham@whirlpool.com
==========================================================