AVR-Chat

On Thu, 2005-09-01 at 09:28 -0500, David Kelly wrote:
> On Thu, Sep 01, 2005 at 09:03:53AM -0500, Thomas Keller wrote:
> >   Now, perhaps itmight be argued that thery could have considered
> > modifying the I/O addressing instructiosn to also access this
> extended I/O space, but if I read the data sheet correctly, this would
> have involved making them take longer to execute (having to read an
> > additional byte to obtain the longer addresses).
> To do so Atmel would have had to modify the instruction set. An opcode
> on the AVR is 16 bits. The "argument" is embedded within those 16 bits
> for SFR codes.

  Yes, that was part of my point.

> Its not just a matter of an additional byte, all instructions are
> multiples of 16 bits. Flash is actually 16 bits wide (page 16 of
> ATmega64), but byte addressable. The program counter counts 16 bit
> words, not bytes. If you look at a disassembly such as generated by
> avr-objdump byte addresses are listed. But look into the opcodes and
> one has to double the value to find the disassembled listed value.

  Yup.  I misspoke, I did know this.  You are quite right.

> The 64 address range for SFRs is more heavily ingrained than you
> suspected.

   I dunno abou tthat, but it is certtainly heavily ingrained, anmd as I
pointed out, I think that ultimately, the Atmel decision on how to
handle it was the correct one.

Tom