Program Challenge (was Re: z80 timing... 6502 timing)

From: Richard Erlacher <edick_at_idcomm.com>
Date: Mon Apr 19 14:27:55 1999

This is a valid viewpoint, though I think, ultimately, the question to be
answered pivots around which processor was potentially the most efficient of
all its resources, including time. However, just the raw speed got a lot of
discussion. In 1983, the 4MHz 6502 was "old hat" and the 8MHz Z-80H was
readily available. However, AFAIK the peripherals for the Z-80H were not,
and, in fact, I didn't ever see them. Somebody said they were out there at
some point, but I've never seen them offered for sale.

You could, of course, postulate that a given processor could be run at a
given rate, whether it's true or not, but if you want a comparison of REAL
parts running REAL code in a REAL environment, then you need something more
than if you just look at what various vendors sold. Every design has
compromises made, and, back in the early '80's, cost of memory was a factor,
as was the assembly cost. In general, I saw lots of boards for lots of
different processors designed to make the most efficient use of memory.

Once we have a good idea of what you're trying to measure, you can make
adjustments to the system variables. If you want to compute how fast a
processor runs code with a given-speed-rated memory, you've got to design
the most efficient memory interface you can and then adjust the processor
clock for that. If you merely want to use a rate which trivially provides a
baud rate generator with a convenient harmonic of the baud rate you intend
to use, you can, I suppose design for that.

I made the statement that it would take, at most, a couple of days to build
a system which would allow straightforward programs to be run and thereby
satisfy the requirement for a test system. I didn't really intend that
people build a unique system just for this test. What I figured was that
folks could look around the basement and see what they could find that would
meet their needs. An old Apple-II would work if you compensate for its
obtuse timing. Since they all seemed to have a serial console, almost any
S-100 system would work.

It's pretty hard to imagine how a limitation like your suggestion would
apply. Newer processors addressed weaknesses in the older ones. One of
those was ease of programming. In some cases, e.g. the 6809, the processor
was designed with a regular instruction set and lots of addressing modes so
as to make generating code easy. It didn't necessarily make it faster. I
don't know how elegant such code will ultimately turn out to be.

My best guess would be that off-the-shelf systems will be what folks use to
program for this "challenge" if anyone does it. There were so few tools for
homebrew or single-board 6502 systems that I doubt anyone with other than an
Apple or an OSI will be interested. OTOH, it will be someone running CP/M
or the like on a Z-80 who enters on the Z-80 side.

That was why I thought a good simulator would be the best solution.

Dick

-----Original Message-----
From: Bill Yakowenko <yakowenk_at_cs.unc.edu>
To: Discussion re-collecting of classic computers
<classiccmp_at_u.washington.edu>
Date: Sunday, April 18, 1999 5:21 PM
Subject: Re: Program Challenge (was Re: z80 timing... 6502 timing)


>You'd normally expect that the winner of any such contest would be
>the most recent processor, wouldn't you? So, if the cut-off date
>was 1982, any processor that was released in late '82 should probably
>beat any that was already available in 1979. Of course, if the 1979
>processor had a much faster version available in 1983 (like a higher
>clock rate), that faster version wouldn't be legal with the 1982 cut-
>off, because that wasn't around in 1982.
>
>So, I suggest that one way of judging cleverness of software hacks
>is that they let an older processor beat a newer one.
>
>Judging that way, we don't even need a cut-off date. But anybody
>coding for a 1999 processor won't have any way to win.
>
>Of course, then we need clear evidence when each processor became
>available at each clock speed, and we can argue about when the chip
>was *really* available (as opposed to being orderable, or just having
>the spec available).
>
> Bill.
>
>
Received on Mon Apr 19 1999 - 14:27:55 BST