Date   : Sun, 20 Jun 2004 22:45:43 +0200 (BST)
From   : Johan Heuseveldt <johan@...>
Subject: Re: PC/Econet

Hi Jules,

On Sun 20 Jun, Jules Richardson wrote:
> On Sun, 2004-06-20 at 10:45, Philip Blundell wrote:
> > On Sun, 2004-06-20 at 10:40, Jules Richardson wrote:
> > > I was actually wondering if a PC parallel port had enough speed to just
> > > drive one of the Master-type Econet modules. Raw-speed-wise I think it
> > > should have, but timing could be a problem due to the nature of the
> > > polled-IO approach to driving a parallel port. Another back-burner
> > > project I'd like to fiddle with sometime though...
> > 
> > Yeah, it'd be interesting to try.  Data throughput probably wouldn't be
> > much of a problem, but again I suspect latency might be an issue.

If you mean interupt latency, yes, I think that's the problem.

> Yep. If it really were that easy, why would Acorn invest time and money
> in developing an entire ISA card? I believe PC parallel ports are still
> hung up conforming to outdated ISA hardware, so even on a modern machine
> there's a huge bottleneck in trying to drive a parallel port at high
> speed.
> 
> Of course maybe a 'modern' ISA design with just a 68B54 and a few buffer
> / latch ICs thrown onto a prototyping board could do the job too. 
> 
> I'm not sure where the hang-up requiring Acorn to include a 6502 on the
> ISA Econet board was - whether it was a speed issue with ISA bus itself

There was another brand of ISA PC Econet card with a processor.
This time it was the Z80 processor; SJ Research probably.

> (and therefore still a problem with modern ISA bus systems which still
> have to run at the original speed), or whether it was a speed issue with
> the PC CPUs of the time (which wouldn't be a problem with modern
> hardware)

AFAIUI it is interupt latency:
Servicing the interupt on an Acorn Beeb is via NMI, which is quite
fast. On a PC, the interupt can't be serviced that fast.
I think the 6502 NMI response is quite unique in view of response, and is
faster than 680x stacking more registers = needing more clock cycles.



The actual data throughput with ISA isn't an issue I think.

greetings,
Johan

-- 
Johan Heuseveldt <johan@...              >
  aka  waarland

  The best place is a Riscy place
 
Cloning is the sincerest form of flattery.