From owner-freebsd-current  Wed Aug 23 11:04:51 1995
Return-Path: current-owner
Received: (from majordom@localhost)
          by freefall.FreeBSD.org (8.6.11/8.6.6) id LAA15456
          for current-outgoing; Wed, 23 Aug 1995 11:04:51 -0700
Received: from gndrsh.aac.dev.com (gndrsh.aac.dev.com [198.145.92.241])
          by freefall.FreeBSD.org (8.6.11/8.6.6) with ESMTP id LAA15449
          for <current@freebsd.org>; Wed, 23 Aug 1995 11:04:46 -0700
Received: (from rgrimes@localhost) by gndrsh.aac.dev.com (8.6.11/8.6.9) id LAA09805; Wed, 23 Aug 1995 11:03:20 -0700
From: "Rodney W. Grimes" <rgrimes@gndrsh.aac.dev.com>
Message-Id: <199508231803.LAA09805@gndrsh.aac.dev.com>
Subject: Re: Matrox Meteor Video Capture Card Driver
To: james@miller.cs.uwm.edu (Jim Lowe)
Date: Wed, 23 Aug 1995 11:03:19 -0700 (PDT)
Cc: current@freebsd.org, se@ZPR.Uni-Koeln.DE, tinguely@plains.nodak.edu
In-Reply-To: <199508231537.KAA26926@miller.cs.uwm.edu> from "Jim Lowe" at Aug 23, 95 10:37:28 am
X-Mailer: ELM [version 2.4 PL24]
MIME-Version: 1.0
Content-Type: text/plain; charset=US-ASCII
Content-Transfer-Encoding: 7bit
Content-Length: 2063      
Sender: current-owner@freebsd.org
Precedence: bulk

> 
> > From: "Rodney W. Grimes" <rgrimes@gndrsh.aac.dev.com>
> > 
> > The NCR runs it's firmware from the host memory, try cranking the PCI
> > latency timer down for the NCR card and see if that helps.  (This is
> > in the PCI bios setup screen).  Depending on motherboards (I happen
> > to know what ``Triton'' board Jim just tried this in) there maybe
> > 1 setting for all slots, or each slot may have it's own setting.
> > 
> > Either way, crank it down to about 40 (default should be 80).
> > 
> > From: se@zpr.uni-koeln.de (Stefan Esser)
> > 
> > Did you try lower values of the "PCI Latency 
> > Timer" of the slot the Matrox has been put in ?
> > 
> > (This timer determines the number of bus clocks
> > a PCI card may claim ownerchip of the PCI bus,
> > if another card wants to start a transfer.
> > The NCR executes its instruction stream from 
> > host RAM, and it may get into trouble if it 
> > can only get another instruction every 10us ...)
> > 
> > A latency timer setting of some 0x20 for the 
> > matrox should allow the NCR to access the bus
> > once per micro second. Guess this is a good
> > value ... 
> > 
> > (The latency timer doesn't force the PCI device 
> > to give up bus ownership if there is no request
> > from another device and thus doesn't slow down
> > the Matrox unless it makes the NCR starve ...)
> > 
> 
> I tried lowering the PCI latency timer from 80 to 40, then to 20, then 1,
> then 0.  It helped me get a few frames at 640x480 before the machine
> hung.  I also tried disabling the cpu to pci burst mode.
> 
> I have been running the meteor in a machine with and adaptec controller
> for 5 days and it never hung.  I am still not sure why the machine
> hangs with the ncr controller.

I am affraid to get a real picture of just what is going on here is
going to require the use of a PCI bus analyzer :-(.  I don't happen
to have one of them, double :-(, :-(.


-- 
Rod Grimes                                      rgrimes@gndrsh.aac.dev.com
Accurate Automation Company                 Reliable computers for FreeBSD