> Let's not start throwing up our respective hands in disgust! Nothing's been
> attempted yet. In fact, nothing's been suggested yet except a couple of
> things which at first inspection didn't seem like they'd work. Now, Hans
> Franke suggested something like a KIM-1. There's no reason one couldn't
> code for something LIKE a KIM-1, even the guys working the Z-80 side, but
> it's not convenient programming a 6532 or whatever those ROM-I/O-Timer
> things were, or even a 6522 for simple I/O if that's what's needed. I
> suggested a published algorithm which solves a published problem or
> something close to that.
Since I'm mentioned, I'll take this to jump in and come up with a more
detailed version of a), the sort thing. To be more specific, I'll may
stay to the KIM-1 as an example.
Task: Read a bunch of punch cards and sort them via columne 1-8
and output the result to a 'card puncher' (Don't you remember the
last time when your just finished source programm droped just in
the moment before feeding the punch card reader :)
I assume that there are 200 Cards to be sorted, so the address space
won't be challenged :)
The input is done by reading a paralell port (6522/32 or PIO or ...)
and issue a strobe for each byte. Input is started thru a signal
issued from the testant. Also the output is done via parallel,
waiting for a strobe signal. Output end has to be signaled from
the testant.
The only allowed sort algorythm is bubble - we don't want to compare
basic algos, but rather CPUs performing the same task.
Time is taken at 4 points:
a) start of input (a Ready signal goes active)
b) end of input (Last byte of last card read)
c) begin of output (First output strobe)
d) end of output (the Ready signal goes inactive)
The external Hardware includes a feeding system (emulating
the card reader and puncher) and some devices to take the
timing (could be the same system - but if s.o. has timers
available, we could use them). The feeder should be fast
enough so support even the fastest system - so this must
be a fast 'modern' thing - like an 286-20 for example :)
The feeder also generates _ONE_ set of random cards to be
used with all contestants.
This is a task involving I/O and 'huge' computing, is 'real world'
orientated, big enough to give an optical visible result and
it can be performed easy on simple systems (KIM etc.) or on
more complex - If you want to use a S100 system, just do it.
Now if we take the time needed for each little programm to run,
and normalie it to the clock frequency, we get a number to show
the clock ratio needed to fulfill the same job within the same
time. Now we just have to compare this ratio with the ratios
available at a given time - example: If we get a ratio of 3:5
between 6502 and Z80, and the given ratio of available CPUs
was at the time in question 1:2, the Z80 is clearly faster on
this competive task.
_Please_ tell me if I'm completely wrong, since I belive this
is what we are talkink about.
And Dick, as far as I might concerne right now, if you stay to standard
components, availabe at the given time, you could build your own device
> There haven't been many suggestions made yet, so
> it's inappropriate to choose. If one wants the hardware, it should be the
> SAME hardware throughout the exercise, though. That's why I was suggesting
> a simulator. All that's really needed is a run to see if it actually will
> execute and end up with the desired result when code is submitted to the
> hardware. A simulator would be adequate so long as it was trusted to give
> honest timing results. That way, nobody would have to risk burning his
> fingers.
Shure, but you still have to do the same calculations _and_
you have to trust the simulation (as said).
Gruss
H.
P.S.: in fact, this could be a VCF 3.0 challenge - hey, Salam,
don't you have a nice dumpster born 1st price to offer ?
--
Stimm gegen SPAM: http://www.politik-digital.de/spam/de/
Vote against SPAM: http://www.politik-digital.de/spam/en/
Votez contre le SPAM: http://www.politik-digital.de/spam/fr/
Ich denke, also bin ich, also gut
HRK
Received on Sat Apr 17 1999 - 19:40:40 BST