Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Path: utzoo!watmath!clyde!caip!nike!ucbcad!ucbvax!jade!eris!mwm
From: mwm@eris.berkeley.edu
Newsgroups: net.micro.amiga
Subject: MC68010 upgrade - how to, and what you get.
Message-ID: <947@jade.BERKELEY.EDU>
Date: Tue, 8-Jul-86 03:24:01 EDT
Article-I.D.: jade.947
Posted: Tue Jul  8 03:24:01 1986
Date-Received: Wed, 9-Jul-86 01:45:29 EDT
References: <902@jade.BERKELEY.EDU> <555@3comvax.UUCP> <2128@ihlpg.UUCP> <305@hp-sdd.UUCP>
Sender: usenet@jade.BERKELEY.EDU
Reply-To: mwm@eris.UUCP ()
Organization: Missionaria Phonibalonica
Lines: 121

In article <305@hp-sdd.UUCP> adp@hp-sdd.UUCP (Tony Parkhurst) writes:
>Would someone please post the info to do the MC68010 upgrade.

Sure.

The upgrade is little more than swapping processors. Exact
instructions can be found on Fish Disk #18. I've included the "How to
Obtain Disks" section of the Fish disk list at the end.

The MC68010 has the following differences when compared with the
MC68000:

	1) Better microcode, so some instructions run faster.
	2) An internal pipeline, and smarts to recognize that loops
		that end in DBcc .-4 will fit into it, and run them out
		of the pipeline.
	3) Better support for virtual memory/machine things, including
	4) move <ea>,SR is now a privileged instruction.

Since I don't write much assembler, and haven't waded around inside
the DOS, I have no idea how much (if any) performance enhancement I'll
get from #2. Hopefully the Lattice "memcopy" routine uses one of
these. #1 buys a little extra performance. #3 is just nice, especially
if you want to play around with doing your own OS. #4 means you need
some code loaded to catch the trap it causes and simulate the MOVE.
All the needed code is on the Fish disk.

I've run three benchmarks on both machines: a simple Mandelbrot
generation, and the dhrystone benchmarks (both with and without
registers). The timings are:


}i			MC68000		MC68010		% performance increase
Mandelbrot (time)	4:27		4:17		3.75
Dhrystone (no regs)	446		454		1.79
Dhrystone (regs)	450		467		3.78

cost (dollars)		1295		1332		2.86

The Mandelbrot is from Fish Disk #5, and is a g on the default set. I
timed the generation with a stopwatch. The Dhrystone was provided by
Matt Dillon, and is supposedly measures the performance of the machine
on an "average" set of integer instructions. The number is the number
of "drystones/second" the machine can do. It is measured in wall-clock
seconds. The boot disk ran my default CLI startup-sequence, and then
had the clock program turned off.

This is one set of timings. The last digit of the %'s is probably
garbage, but provided for completeness. I'd be interested in seeing
other sets of timings, most especially with the MANX C compiler.

Conclusions:

Now that I've made the change, I'm happy about it. I got more increase
in performance than in cost, and have had no problems at all. However,
it's not enough to be obvious when using the system. If i'd known in
advance, I'd NOT have done it; the ulcers involved in me opening the
box aren't worth the minor performance gain.

In other words, I wouldn't recomend this upgrade AT ALL, unless 1)
You're comfortable twiddling hardware, and want to be different, or 2)
you really want to play with OS's. Possibly those wanting Amiga 2000
compatability may be interested in it. Of course, you'll have to make
this decision yourself; I've tried to provide all the relevant data.

Final note: I bought my 68010 from a small parts house, and got the
LAST one they had. It's not clear when or if they'll get another, so
PLEASE don't send me mail asking where to get them or where I got mine
- it won't do you any good, and I won't answer it since you've been
warned. Try calling the places that advertize in the back of Byte.

	<mike

From the list of disks available from Fred Fish, as posted by him.

HOW TO OBTAIN DISKS
-------------------

First, check with your local dealers and user groups.  Many already have
the earlier disks.  Since these disks can be copied freely, and widespread
distribution is encouraged, they propagate out to central distribution
points fairly quickly.

If you just can't wait, or can't find copies locally, I am willing to
make these disks available for the cost of media, mailing materials, 
postage, and miscellaneous expenses (like wear and tear on my drives).
My goal is to get as much software as possible into the hands of people
that can use and enhance it, and make the AMIGA the success it deserves
to be.

Each disk contains all source necessary to recreate the executables
provided.  All programs are currently compiled with the latest Lattice C
and/or Manx C compiler available at the time of release.  In a very few
cases (noted in the description) the code will not compile or run for some
reason, but was considered interesting enough to include anyway.  Almost
all executables are known to run on the latest kickstart/workbench
combination available at time of release.

Disks are typically 85 to 95 percent full.  


HOW TO ORDER
------------
To order, send a list of the disks you want, and $10 per disk, to:

	Fred Fish
	345 Scottsdale Road
	Pleasant Hill, Ca. 94523

	(415) 685-7295    (Sorry, I can only return calls collect.)
	ucbvax!unisoft!fnf	well!fnf	lll-lcc!unisoft!fnf

Time and other jobs permitting, all disks will be mailed via first class
mail within 5 days of receipt of order.  (Tips may help speed the process :-)

Feel free to order more the the current number of disks available.  Excess
funds will be placed "in escrow" (refundable at any time) and drawn against
for automatic mailings of future disks as they become available.  I hope
to add at least two to four disks per month to the library.  Given that I
have a database of about 300Mb of freely distributable software to
draw upon, that should be a fairly easy goal to accomplish.