On Mar 31, 2006, at 3:30 PM, Xiaofan Chen wrote:

> Cortex-M3 is different from ARM7TDMI from Philips ... It does
> not support 32-bit ARM instructions. It only support thumb-2
> instruction sets.

Do thumb-2 instructions instruction still execute in single cycles?
I couldn't find anything about the actual instruction set on the
luminarmicro web site (I'm not sure you CAN describe the thumb2
IS except in terms of the full ARM instruction set; an interesting
state of affairs!  Some instructions will be missing entirely.)
An 8kbyte micro where each instruction is 4 bytes is pretty
inefficient...


>> 20MHz, 8k flash, 2k ram.  Not exactly what one expects from an ARM.
>
> Why not? It is what you get in a 16F.

Because ARMs have traditionally been high-clockrate, high-memory
CPUs.  More memory usages is a typical penalty for wider architectures,
(every time you call a subroutine, you push a 4byte PC, and every
register you save is 4 bytes...)  2kbytes of 32bit is only 512
32 variables, which is pretty minimal.  But the clockrate is the
big surprise.  This is the same speed as most 8 and 16 bit micros.
I would have hoped for at least 40MHz, and 32k/8k of memory (an
LPC2103 in a smaller package, with the cortex core.)  (Mind you,
this is in the absence of any particular application, so the "hope
for" is pretty vague...


> What does 32 bits get you?

Well, faster math if your numbers are bigger, for sure.  Easier
addressing of large memory arrays.  A relaxation (for a while)
of peripheral addressing.  32bit CPUs tend to come with enhanced
management capabilities as well (generically, this like exception
traps, singlestep capability, breakpoints and watchpoints, etc.)

> Mip for mip, are the ARM instructions capable of much more?
> Similar?

Wouter summed it up pretty well.  In a single instruction, the
ARM can do math on two 32bit numbers; that's much more than a
PIC.  OTOH, it cannot set the state of an ouput pin in a single
instruction.  Are your applications math intensive or pin-twiddle
intensive?

> if a pic was running at 4 MIPs and one of these ARM devices was
> running at 20 MIPs, is it 5 times faster? Or are the ARM mips
> more powerful?

Insufficient data :-)

>
> but $1 is the price for 1k probably.

Mouser's prices for the luminary micros start at $2.18 for 1.
That's pretty good for someone advertising a $1 (in quantity)
micro.  OTOH, the development kit is $775!  (ouch.)  I'd like
to predict a cheap development kit an a CCI (or other) design
contest, but I don't know if a fabless startup like luminary
can fund something like that...

Has any microchip vendor every done a development contest for
tools?  That'd be interesting.  (with the ubiquity of open
source "heritage" for many tool chains, it could be "interesting"
in all sorts of BAD ways.  Sigh.  If someone spends a couple
weeks adding some neat bells and whistles to someone else's
open source tool that's been around forever, who gets the prize?
(actually, that's a "problem" I'd like to see get addressed.))

BillW
-- 
http://www.piclist.com PIC/SX FAQ & list archive
View/change your membership options at
http://mailman.mit.edu/mailman/listinfo/piclist