From owner-freebsd-hackers  Sun Mar 14 11:47:24 1999
Delivered-To: freebsd-hackers@freebsd.org
Received: from rah.star-gate.com (rah.star-gate.com [209.249.129.138])
	by hub.freebsd.org (Postfix) with ESMTP id 0864614D25
	for <freebsd-hackers@FreeBSD.ORG>; Sun, 14 Mar 1999 11:47:20 -0800 (PST)
	(envelope-from hasty@rah.star-gate.com)
Received: from rah.star-gate.com (localhost.star-gate.com [127.0.0.1])
          by rah.star-gate.com (8.9.1/8.8.8) with ESMTP id LAA87562;
          Sun, 14 Mar 1999 11:45:21 -0800 (PST)
          (envelope-from hasty@rah.star-gate.com)
Message-Id: <199903141945.LAA87562@rah.star-gate.com>
X-Mailer: exmh version 2.0.2 2/24/98
To: Matthew Dillon <dillon@apollo.backplane.com>
Cc: Wes Peters <wes@softweyr.com>, Cory Kempf <ckempf@enigami.com>,
	Bill Paul <wpaul@skynet.ctr.columbia.edu>,
	freebsd-hackers@FreeBSD.ORG
Subject: Re: Gigabit ethernet -- what am I doing wrong? 
In-reply-to: Your message of "Sun, 14 Mar 1999 11:20:28 PST."
             <199903141920.LAA93395@apollo.backplane.com> 
Mime-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Date: Sun, 14 Mar 1999 11:45:21 -0800
From: Amancio Hasty <hasty@rah.star-gate.com>
Sender: owner-freebsd-hackers@FreeBSD.ORG
Precedence: bulk
X-Loop: FreeBSD.ORG

If the pci device has the concept of a program store like in the case of 
a bt848 chipset it is conceivable for dma or internal operations to do a retry.
It is a different issue if the network chipset designers chose not to have
a programmable dma or process control like in the bt848  .


	Best Regards,
	Amancio


> 
> :>     cannot be done.  So routing a single packet requires the data to flow
> :>     over the PCI bus twice.  The 132 MBytes/sec become 66 MBytes/sec right
> :>     off the bat.
> :
> :I am not sure that I can follow you here . Most PCI cards which are capable of 
> :doing dma to the host system's memory can do card - to - card transfer 
> :;however,
> :the target "card" most be able to use the stored data in the case of a network
> :card it must have memory to receive the pack or a very elaborate protocol 
> :to accept short dma bursts which it can then process.
> :
> :
> :	Amancio
> 
>     You can always do a card-to-card transfer, but since most modern network
>     cards do *NOT* have on-card memory doing a card-to-card transfer typically
>     doesn't work.  For example, if the destination card hits a collision/retry,
>     the source card's FIFO can overflow.  
> 
>     It just doesn't work.
> 
>     Using a card as a DMA destination only works well for cards that map
>     memory, such as a video card. 
> 
> 					-Matt
> 					Matthew Dillon 
> 					<dillon@backplane.com>
> 




To Unsubscribe: send mail to majordomo@FreeBSD.org
with "unsubscribe freebsd-hackers" in the body of the message