P1394 Mail Archive: Re: P1394> Rough SBP-2 & IP1394 core component ROM sizes

P1394 Mail Archive: Re: P1394> Rough SBP-2 & IP1394 core component ROM sizes

Re: P1394> Rough SBP-2 & IP1394 core component ROM sizes

Greg Shue (gregs@sdd.hp.com)
Wed, 11 Feb 1998 08:59:39 -0800 (PST)

Some chipsets offer HW assist for portions of the SBP-2 target
protocol. Others offer HW support for DMA to map device memory
to Serial Bus memory, which significantly improves SBP-2 intiator
performance (e.g. OHCI). This does _not_ mean that the chips
are protocol-specific.

I definitely agree that initiator and target functionality are
vastly different. In my mind, the represent opposite ends of a
client-server model. I think it would be extremely difficult to
write a single application that could, in a single instance (or
through a single interface) be both a controller and a controlled
function. (Note, I did not call this a Module, a Node, or a
device. :-) I could very easily see a protocol stack for a node
having both target and initiator support. Any server-type
services would use the target support. Any client-type
applications would use the initiator support. Because there is
so little that is shared, I see the size of the stacks as being
(roughly) the sum of the individual solutions, no matter how many
applications or services my node has.

Am I missing something here?

Greg Shue

>
>
>
> I didn't know chip sets offered "explicit" support for SBP-2. OHCI
> maybe, but I hadn't seen the chip set feature in the brochure for
> "chip-level SBP-2 support".
>
> Also, since 1394 is a peer-to-peer interface, anyone developing a
> peer-to-peer device (like the majority of the Japanese electronics
> industry), would have to essentially duplicate the functionality of
> SBP-2 initiator and target, whether they actually use SBP-2 to do this,
> or some other (potentially proprietary) method for implementing
> peer-to-peer functionality.
>
> Randy
>
>
>
> -----Original Message-----
> From: Stephen Holmstead [SMTP:stephen@hpb16977.boi.hp.com]
> Sent: Wednesday, February 11, 1998 8:30 AM
> To: 'Turner, Randy'; 'Brian Batchelder'; p1394@pwg.org
> Subject: RE: P1394> Rough SBP-2 & IP1394 core component
> ROM sizes
>
> My experiences in the past is that combining target AND
> initiator
> functions into a single device is VERY difficult. The
> complexity of the
> state tables and ambiguities that can arise make it very
> difficult.
> Also, I don't believe any of the chip manufacturers support
> SBP-2 for
> both target and initiator. In my opinion, it's NOT just a
> matter of
> adding the 15K for the target and 15K for the initiator. The
> target and
> initiator have vastly different functions. Designing a model
> that did
> both (well) would be a challenge. It was problems like this
> that fueled
> many of the AEN (asynchronous event notification) debates.
>
> Stephen
>
> > -----Original Message-----
> > From: Turner, Randy [SMTP:rturner@sharplabs.com]
> > Sent: Monday, February 09, 1998 1:03 PM
> > To: 'Brian Batchelder'; p1394@pwg.org
> > Subject: RE: P1394> Rough SBP-2 & IP1394 core component
> ROM sizes
> >
> >
> > I am very interested in the combined target/initiator
> footprint as
> > well.
> >
> > Thx
> >
> > Randy
> >
> >
> > -----Original Message-----
> > From: Brian Batchelder [SMTP:brianb@vcd.hp.com]
> > Sent: Monday, February 09, 1998 9:37 AM
> > To: p1394@pwg.org
> > Subject: Re: P1394> Rough SBP-2 & IP1394 core
> component
> > ROM sizes
> >
> > At 02:04 PM 2/6/98 -0800, Greg Shue wrote:
> >
> > [snip]
> >
> > >I approached them about sharing some rough estimates of
> code
> > size
> > >for the IP1394 glue, SBP-2 target and initiator, and
> 1394 bus
> > >manager modules to get a feel for how "heavy" the
> protocol
> > >implementations really are. They graciously shared the
> flowing
> > >estimates with us, and certainly deserve our thanks.
> Since the
> > >technology is still being developed, these
> implementations are
> > >certainly going to be tuned and the numbers will change
> a
> > little
> > >bit.
> > >
> > > Component Rough ROM size (KBytes)
> > > --------- -----------------------
> > > SBP-2 initiator core 15 +/- 5
> > > SBP-2 target core 15 +/- 5 (incl. Mgmt+Fetch
> agent,
> > no exec agent)
> >
> > Any idea how big a combined SBP-2 target/initiator would
> be?
> > 15-30K? ;-)
> >
> > [snip]
> >
> > Brian
> >
> >
> ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
> > ~
> > Brian Batchelder | Hewlett-Packard |
> > mailto:brianb@vcd.hp.com
> > Connectivity Futurist | 1115 SE 164th Ave. | Phone:
> (360)
> > 212-4107
> > DeskJet Printers | Vancouver, WA 98684 | Fax:
> (360)
> > 212-4227
>