On Mar 31, 2006, at 12:05 PM, Wouter van Ooijen wrote:

>> Mip for mip, are the ARM instructions capable of much more?
>
> As always, it depends. If you can away with the RMW problem, a PIC can
> set or reset a pin in a single instruction. This is a bit more
> complicated on an ARM, but it can be done.
>
> An ARM is a register-register architecture. So no direct
> manipulation of bits in memory. Compare:
>
>    # one 12, 14 or 16 bit instruction
>    bsf some_place_in_memory, a_specific_bit
>
> with
>
>    # 4 instructions, 32 bits each
>    ldr r0, =some_place_in_memory
>    ldr r1, [ r0 ]
>    and r0, r0, # ( 1 << a_specific_bit )
>    str r1, [ r0 ]
>
I've been reading the Luminary data sheets (and designing
hypothetical PCBs...)  There is apparently a new feature in the
cortex M3 core called "bit banding" where ARM throws address
space at the bit-addressing "problem" of the architecture.

That means that for an arbitrary bit N in some register address M,
you can access JUST the one bit by using memory address:
    bitbandbase + (M *32) + (N*4)
making the above example something like:
	ldr r0, =(bitbandcalc(some_place_in_memory, a_specific_bit)
	str r_known_one, [r0]

Where at least one of those instructions is only 16 bits...

Sorta neat.  I'm not sure how I feel about "wasting" 32 bytes of
address space for each byte  of "bit-banded" memory, but I guess
a small ARM processors DOES have address space to burn!

BillW

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