Date   : Mon, 19 Dec 2011 16:30:40 +0000 (GMT)
From   : dabenavidesd@... (Daniel Alejandro Benavides D.)
Subject: 32016 + 32082

Hi all:
I think you must be talking about the rule of the thumb, if so, they were
pretty sure they will be over-head of the benchmarks:

http://www.bbc.co.uk/news/technology-11082701

If so, they wouldn't need to modify the HW in ARM2 for the ARX in ARM3, but
yeah still that wasn't good enough for the DRAM of the day.

Thanks in advance

> On Sun, 18 Dec 2011 12:08:46 +0100, Patrick Kaell <sparc@...>
> wrote:
> 
> >On 12/17/2011 05:21 AM, David Hunt wrote:
> >>> I've just acquired a 32016 Second Processor,
> and I've noticed that this
> >>> one has a 32082 MMU on board.
> >
> >How did the performance compare to the 68000?
> >
> >According to this Wikipedia article:
> >
> >http://en.wikipedia.org/wiki/Microprocessor#32-bit_designs
> >
> >the NS32016/32 outperformed the MC68000/10.
> >
> >What about the 80186 in the Master 512? The 80186
> contained many 
> >optimizations of the 80286 without the protected mode.
> Since Acorn 
> >considered adopting 80286 before designing the ARM, it
> must have a real 
> >speed advantage to the 6502, which seems not to be the
> case for the 32016?
> 
> You have to be very careful making comparisons like that.
> The hardware itself is one factor, but (lack of)
> efficiency
> in software (including compilers) is another, and is
> almost
> never evaluated in benchmarks.
> 
> Certainly when a company adopts a new processor, knowledge
> on how to use it efficiently is thin on the ground, which
> tends to lead to suboptimal software performance, early
> on.
> 
> It is doubtful, for example, that BASIC for the 32016 was
> created from scratch, and went through some comprehensive
> optimization phases. It is likely that it is a crude or
> even
> automatic translation of the 6502 version. The performance
> suffers accordingly. The 6502 and 32016 are quite
> different
> beasts, and coding in 32016 assembly would probably
> require
> rethinking the coding strategy in many parts of the code.
> 
> >[...]
> 
> 
> John Kortink
> 
> -- 
> 
> Email    : kortink@...
> Homepage : http://www.inter.nl.net/users/J.Kortink
> 
>