Re: Problems in "Modernizing" a //gs

[bjblackbear@ibm.net (B.J. Major)]

To:          Scott G.
From:        Lewin A.R.W. Edwards, sysadm@zws.com
Re:		        Response to your comments

(please send all further comments/discrepancies to Lewin directly at
sysadm@zws.com from this point onward--thanks.)

>I see he programs for OS/2 and thus for the Intel architecture.  I'm
>surprised he would feel this way considering the Pentium has evolved
>from the 8-bit Intel 8080 and we don't think of the Pentium as
>"basically an 8-bit CPU at heart."  Let me see what's wrong with his

But the Pentium *is* basically an 8-bit CPU at heart, it simply has 16 
and 32-bit extensions. Nobody except Intel regards the 8086 as being a 
real 16-bit CPU, and the Pentium situation is closely analogous. Proof of 
the point lies in the fact that the Pentium has 8-bit opcodes.

>>>Hmm. I hate to sound like a wet blanket, but if the 65832 is
>>>pin-compatible with the 65816 then it is still basically an 8-bit CPU
>
>Here's the problem.  Who said the next generation 65816 would be pin
>compatible?  It would only have to be instruction compatible!  A

The guy who wrote the original message [to B.J.] said it was pin
compatible! I can only work with the information I'm given, you know. It
wasn't an 
assumption, it was information received.

>>>horsepower. The 68000, clock for clock, is substantially more
>>>powerfulthan the 65C816 used in the IIgs.
>
>but the 65C816 is much more efficient at handling memory.  It can
>access memory in a single cycle whereas the 68000 needed 4 cycles.  In

Right, and how many major programs do you know that have nothing but :

      move 1000000,d0
loop: move (a0)+,(a1)+
      dbra d0,loop

as the entire meat of the program? That's the only situation where your 
observation will have a real-world effect. I stand by what I said.

>Design Center also claims it has some RISC like features.

Yippee. Go to <http://www.microchip.com/> and look at the specs of some 
of their RISC microcontrollers. They are pipelined Harvard-architecture 
microcontrollers with 12-bit, 24-bit (etc) instruction words and 8-bit, 
16-bit, 32-bit data file busses. The one I'm using at the moment is 
clocked at up to 20MHz. Doesn't make it a blisteringly fast CPU just 
"because it has some RISC like features".

>Similarly, I clocked a Macintosh LC '020 running at 16MHz against a
>GS running at 9MHz and found that the GS came was consistently faster
>at tasks such as scrolling through a word processing document, sizing

This is a lovely meaningless benchmark. The GS's operating system is 
radically different from MacOS (it might look similar in some respects, 
but it's very different). You can't compare the two. There are a bunch of 
other factors too - color depth at which the display was running, size of 
the display area, wait states imposed on accessing video memory, and what 
custom hardware has to share the same buses.

Why don't you do the same little test on your Intel machine running 
MS-Word for DOS (one of the graphical versions)? I bet you the results 
are reversed.

>>>I daresay that a reasonably powerful PowerPC machine running an
>>>emulator will be able to emulate this "mystery 65832" at three or
>>>four times the instruction rate speed, and at least an order of
>>>magnitude higher thruput than the actual hardware.
>
>Real world results disprove this conjecture.  The "Bernie II" emulator
>written by master programmer Henrik Gudat, running on a kick-butt
>PowerMac, can't match my real 13.75MHz GS in running GS software with

Come ON, be realistic. Bernie II is emulating the entire hardware of the 
IIgs; graphics, audio, disk I/O. Ask Henrik to give you his 65816 engine 
bare and run it flat-out on the same machine. I guarantee you it will be 
orders of magnitude faster.

>Apple could have and should have done this.  The GS outsold the entire
>Mac line in its first year.  If Apple had done the right thing and

I see. These couple of lines really tell me a lot.

>>>It would also be necessary to hack (at best... rewrite at worst) the
>>>IIgs's firmware, because there are software timing loops in the ROM
>>>for things like floppy drive step rate, all of which would be thrown
>>>out of whack by a faster CPU.
>
>All this was done when the Mac Plus evolved into modern Macs.  If the

Yes, yes, it can be done - the point is, it needs to be done, which adds 
cost to the project, which is what I was saying.

>Apple is one of the few companies that deliberately held back the
>development of one of its computers (the GS) for fear it would

Few? IBM does it as a matter of routine. Microsoft does it with software. 
DEC does it too. It doesn't always make sense to have two products 
fighting for the same niche. Basically what you have said in this message 
is that Apple shouldn't have made the Mac, they should have made a Mac 
and called it a IIgs "plus" or something.

>B.J., your friend is part of team OS/2!  That's not "Win" but it is
>"tel"  OS/2 runs on Intel machines.  Like I said, there is no such
>thing as a Wintel machine.  :)

And you've never heard of OS/2 for PowerPC? I'm not surprised, but it 
exists. Believe me there are few things in the world I despise more than 
the iAPx86 series of CPUs.



=== From the Apple PowerBook of ======================
Lewin A.R.W. Edwards [Team OS/2] <http://www.zws.com/>
Programmer + Hardware Eng     Certified Apple Engineer
Need a programmer?    <http://www.zws.com/resume.html>
Tel  +61-412809805                   Fax +61-392281335
"No non-trivial programming task is atomic" - LARWE 97
=== (C) Copyright 1997 by Lewin A.R.W. Edwards =======





--bj    Automatic reply will not work; please reply to: bjbear@ibm.net
Proud owner/user of Apple Computers since 1984; NewsHopper, the offline newsreader for the Macintosh!
Barbie and Ken doll fan since 1961