Date   : Mon, 12 Feb 1996 23:17:34 +0000 (GMT)
From   : James Fidell <james@...>
Subject: Re: The BBC's Logical, Well-Documented and Non-Perverse Keyboard

Tom Seddon wrote:

> As you might have guessed I've got the keyboard working. My problem was 
> that when a CA2 int was being genmerated on the keyboard column scan, 
> because the IER was ddisabling CA2 the bit wasn't being set in the IFR. 
> I've now moved my if{}s around and it works great. I claim the prize for 
> most complete system VIA implementation :-) (since I have input latching 
> as well)

I'd not claim the prize yet, if I were you :-)

> Unfortunately SHIFT doesn't work but I'm going to sort that out (I think 
> the SHIFT key is generating an interrupt or something).

Don't know about that one.

> Now, I have another problem. My virtual BBC is now about 33% slower than 
> a real BBC, on a 486dx2-80. After I've done the obvious optimisations 
> (update VIA clocks every X cycles, that sort of thing), if the speed 
> isn't up to scratch what do people suggest?
> 
> 1. Leave it 'as is'--ie slow?
> 2. Go as far as I can with 'C'?
> 3. Convert it all/bits of it to asm? I think I'd have to convert the 
> whole lot, or nothing; as I'm using djgpp and go32 I'd have to use the 
> GNU assembler and its icky 680x0-style 80x86 syntax. Don't get me wrong, 
> I think the 680x0 family is infinitely better (especially the assembly 
> language), but when I'm in front of a PC I just cannot bring myself to 
> put the destination second!
> 
> I'd be interested to hear your views on this.

1. No.

2. I don't think you should need to use anything else.

3. No.

Start using a profiler if you can.  Think about what's slowest and why.
Think about how you allocate variables and how they're used.  I doubled
the speed on Xbeeb just by careful consideration of what got used most
and how I could optimise that to run more quickly.

Under DOS I'd have thought you'd have more advantages because you
effectively have complete control of the machine.  Xbeeb has reached the
stage where the bottleneck is really the display, but there's little
you can do about the fundamental architecture of X Windows.  When there's
little screen updating being done, it will run two to three times as fast
as a standard Beeb on my DX2-66.

> PS Checking through my old e-mail I noticed that James Fidell's signature 
> states he now works at netcom--so what's happening to the list? In fact, 
> what's happened to James himself?! Are you out there...

As of 1996, I work for Netcom, currently setting up their dial-up IP
service here in the UK.  The list is still at OiT, though I hope to
move it to Netcom at some stage.  My change of job, coupled with a
skiing holiday, working for some scuba diving qualifications and various
other bits and pieces have meant that I've been a little quiet of late.
However, I expect to have Xbeeb v0.3 out before the end of the month.

James.

-- 
 "Yield to temptation --             | Work: james@...          
  it may not pass your way again"    | Play: james@...                 
                                     | http://www.OiT.co.uk/~james/
        - Lazarus Long               |              James Fidell