On 5/20/06, Spehro Pefhany <speff@interlog.com> wrote:

> Not instruction width, code density. How many K bytes does an entire
> application take coded for core A vs. core M? That determines the area
> of silicon required to store the program for a given technology.

The Arm Thumb and Thumb 2 are ARM's solution for higher code
density. In fact, the low end Cortex M3 MCU does not support
ARM instruction set, it only supports Thumb 2 instruction set.

http://www.arm.com/products/CPUs/archi-thumb2.html
http://www.arm.com/products/CPUs/ARM_Cortex-M3.html

Still the core does not really occupy the bulk of the MCU. The
memory and peripherals often occupy more space. Often
the peripherals decide the chip selection. For example, we
decided to use Silicon Labs C8051F MCU for an optic sensor
since it has a fast ADC and no PIC16F/18F have the
ADC performance we need.

http://www.atmel.com/dyn/resources/prod_documents/doc3323.pdf
is a nice white paper from Atmel. (But it is a pain that copy and
paste is not allowed!)

In the conclusion, the author tells us the real challenge:
"The migration from 8-bit to 32-bit microcontrollers is not just a question
of device cost. The integration level reached on 8-bit microcontrollers
during the last years must be matched."

For example, Atmel's ADC implementation in their AVR sucks since
it is very slow. Atmel's ARMs continue to use the slow ADC implementation.
This makes them useless for quite some applications even with a fast
and nice core.

Regards,
Xiaofan

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