about the Tulip & Intel PRO 1000

[Pramod John]

Robert,

    What Tulip based cards do you know of that are readily available ? Any
multiport cards ?

thx,

Pramod

----- Original Message -----
From: "Robert Morris" <rtm at amsterdam.lcs.mit.edu>
To: "Giorgio Calarco" <gcalarco at deis.unibo.it>
Cc: <click at amsterdam.lcs.mit.edu>
Sent: Monday, April 14, 2003 3:30 AM
Subject: Re: about the Tulip & Intel PRO 1000


> Giorgio,
>
> Yes, by DMA we mean direct memory access. The Tulip does transfer
> packets to/from memory without direct CPU intervention.
>
> The Intel Pro 1000 also uses DMA, and is generally similar to the
> Tulip. However, the CPU must use one PIO instruction per batch of
> packets to notify the Pro 1000 that there is new work for it to do.
>
> Robert
>
> > From: "Giorgio Calarco" <gcalarco at deis.unibo.it>
> > To: <click at amsterdam.lcs.mit.edu>
> > Subject: about the Tulip & Intel PRO 1000
> > Date: Mon, 14 Apr 2003 11:52:09 +0200
> > Organization: DEIS
> >
> > Hi,
> >
> > I 'm reading your paper "programming language optimizations
> > for modular router configurations" and there are some details
> > which I cannot understand, probably due to my poor knowledge
> > of how the hardware platform works...
> > I try to explain what I think I have understood, plese correct
> > me if I mistake... I refer to paragraph 8.4 - PCI limitations.
> >
> > You wrote that packets can be dropped on the receving
> > tulip card because the internal buffer is full or because the NIC
> > is not able to fetch a ready DMA descriptor etc etc.
> > My first question is: with DMA you mean the "direct memory access" ?
> > If so, are you transferring the packets from the NIC to memory
> > making the tulip card to be the PCI bus master and
> > programming it to transfer the packets without
> > involving the CPU ? If not, how the transferral works ?
> >
> > If what I've thought is true, I have a second question...
> > about the Intel PRO 1000.
> > I'm reading that this can be slower than the tulip card because the
> > packets are transferred via PIO. Is there a reason for this ?
> > I mean, the Intel PRO1000 is unable to become the PCI
> > bus master and the only way to move the packets around
> > needs the CPU to be involved (being it the bus master)?
> > Or is this choice due to other reasons ?
> >
> >
> > Thanks in advance for your answers
> >
> > giorgio
> >
> >
> >
> >
> > Ing. Giorgio Calarco
> > DEIS - Università di Bologna
> > Viale Risorgimento, 2 40136 Bologna - Italy
> > Tel: 051 2093776 Fax: 051 2093053
> > E-mail: gcalarco at deis.unibo.it
> > PGP: http://www.deis.unibo.it/GCalarco/Giorgio_Calarco.asc
> >
> >
>