*BSD News Article 29925


Return to BSD News archive

Path: sserve!newshost.anu.edu.au!harbinger.cc.monash.edu.au!yarrina.connect.com.au!warrane.connect.com.au!kralizec.zeta.org.au!not-for-mail
From: bde@kralizec.zeta.org.au (Bruce Evans)
Newsgroups: comp.os.386bsd.development
Subject: Re: FreeBSD 1.1G - SLIP performance problems
Date: 2 May 1994 15:28:35 +1000
Organization: Kralizec Dialup Unix Sydney - +61-2-837-1183, v.32bis v.42bis
Lines: 28
Message-ID: <2q2323$12g@kralizec.zeta.org.au>
References: <2pqst6$g4s@opine.cs.umass.edu> <Cp15CK.qII@mozo.cc.purdue.edu>
NNTP-Posting-Host: kralizec.zeta.org.au

In article <Cp15CK.qII@mozo.cc.purdue.edu>,
Ben Jackson <bj@kidd.vet.purdue.edu> wrote:
>In article <2pqst6$g4s@opine.cs.umass.edu>,
>Jim Doyle <doyle@cs.umass.edu> wrote:

>I think the "one-way" nature of your problem is the key...
>
>>*  The 386PC (386dx40) has a 16550A on the port used for the SLIP interface.
>                              ^^^^^^
>>   It is connected to the PCRoute box useing a hardwired null-modem
>>   SLIP link. The SLIP link ran at 38400 bps, I did not try 57600.
>
>The 16550A has a FIFO which helps prevent data loss at high transfer
>rates.  As I understand it, though, there is just *one* FIFO, and you
>have to choose which way it buffers (input or output).

No, there is one dedicated to input and one dedicated to output.  Perhaps
the input one is broken?  1.1-BETA finally uses the full capabilities of
the output fifo, but the output direction is the one that works.

>I don't think it would be unreasonable to estimate that the FreeBSD
>box is getting 16x the interrupt traffic in the non-buffered direction.

FreeBSD has efficient serial interrupt handling and would hardly notice
16x the receiver interrupt traffic at a low speed like 38400 (it would
cost about 5% on a 386DX40).
-- 
Bruce Evans  bde@kralizec.zeta.org.au