Re: [MiNT] New MiNT distribution is n.. ..pro

[Michael Schwingen <rincewind@discworld.dascon.de>]

On Mon, Feb 21, 2000 at 09:40:58PM +0200, Martin-Éric Racine wrote:
> 
> Given this, I'm sure if Michael and Ali went to Motorola and said
> "build us a 200 MHz 060 with FPU _and_ 030 instructions that were
> removed since the 040 and we'll buy 1000 units now and 1000 more
> in 6 months, after the first batch of Milan 3 is sold," Motorola
> could not refuse.

I guess you would have to add some zeroes to that numbers - I am sure that
Motorola would not even think about a 060 redesign for 10k-100k quantities.

And BTW, I have no problem with the removed instructions on the 060: these
are rare and are handles fine by the emulation code supplied by Motorola.
Using movep in a disk transfer loop, and movep emulation code on a 060, I
still got about 600K/s throughput, so for the rare case of a few movep
instructions in user code, this is definitely no problem.

You have to make a tradeoff between rarely used instructions and complexity
of the CPU. If dropping movep enables the designers to get more speed out of
the pipelines, then that is OK by me - you can't get 100% compatibility at
full speed.

> Btw, one interresting avenue to explore would be a dual-CPU Atari
> using, let's say, twin 060's at 80 MHz.

All the people talking about multiple 68k CPUs should start to think about
what OS they want to use. Multiple CPUs only give a real benefit for the
money if you have multithreading both in the OS and in the applications,
otherwise, it is just wasted money in most cases.

cu
Michael

[1] Getting a 060 produced in a current (0.25u or less) production process
will require a complete redesign, since you can't simply shrink the whole
masks and expect the resulting chip to work.
-- 
Michael Schwingen, Ahornstrasse 36, 52074 Aachen