From owner-freebsd-hackers  Mon Jan 29 00:09:48 1996
Return-Path: owner-hackers
Received: (from root@localhost)
          by freefall.freebsd.org (8.7.3/8.7.3) id AAA12390
          for hackers-outgoing; Mon, 29 Jan 1996 00:09:48 -0800 (PST)
Received: from labinfo.iet.unipi.it (labinfo.iet.unipi.it [131.114.9.5])
          by freefall.freebsd.org (8.7.3/8.7.3) with SMTP id AAA12374
          for <hackers@FreeBSD.org>; Mon, 29 Jan 1996 00:09:10 -0800 (PST)
Received: from localhost (luigi@localhost) by labinfo.iet.unipi.it (8.6.5/8.6.5) id JAA04923; Mon, 29 Jan 1996 09:03:15 +0100
From: Luigi Rizzo <luigi@labinfo.iet.unipi.it>
Message-Id: <199601290803.JAA04923@labinfo.iet.unipi.it>
Subject: Re: [FCC Warning!] The Dangerous effect of Direct TV !!
To: hasty@rah.star-gate.com (Amancio Hasty Jr.)
Date: Mon, 29 Jan 1996 09:03:15 +0100 (MET)
Cc: terry@lambert.org, jkh@time.cdrom.com, james@miller.cs.uwm.edu,
        dufault@hda.com, hackers@FreeBSD.org, multimedia@rah.star-gate.com
In-Reply-To: <199601290155.RAA00586@rah.star-gate.com> from "Amancio Hasty Jr." at Jan 28, 96 05:55:13 pm
X-Mailer: ELM [version 2.4 PL23]
Content-Type: text
Sender: owner-hackers@FreeBSD.org
Precedence: bulk

> As I am typing this I am watching TV on my FreeBSD box 8)
> Whats different is that Matrox Meteor is dumping video directly
> to my video graphic card very little CPU intervention. The 
> reason for the very little intervention is that the program
> "tv" is capturing a single frame at a time.

There is no point in doing this, I believe.

>  > How do you deal with bus-to-bus DMA cache effects?  It seems likely
>  > that this type of usage is one that wasn't considered by the
>  > motherboard manufacturers... considering how many of them have blown
>  > simple bus-to-memory DMA cache update/invalidate.
> 
> I still can't answer your question however the program seems to
> be behaving quite well. 

Consider a few things:

* why should the VRAM be cacheable in the first place, given that the
  video chipset is doing all sorts of things on it.
* often both the grabber and the video board are on the same PCI bus.
* every location is written by the grabber only once per frame.
* one frame does not fit into 256K of cache

--> we should not bother too much about the problem mentioned by Terry.

	Luigi