Re: [avr-gcc-list] Experimental WinAVR/avr-gcc update for

[Erik Christiansen]

On Wed, Aug 20, 2003 at 04:58:13PM -0400, Dave Hansen wrote:
> According to an old usenet post apparently from an engineer at IAR, "In 
> fact, the AVR was designed by Atmel with help from IAR, to make the 
> architecture suitable for high level language programming."

In "The AVR Microcontroller and C Compiler Co-Design", found on Atmel's
site IIRC, Dr. Gaute Myklebust of ATMEL Development Center, Trondheim,
Norway, writes in a section titled "Fine tuning AVR":

"By allowing professional compiler developers at IAR Systems in Sweden
to comment on the architecture and instruction set, we were able to make
a microcontroller very well suited for C compiler generated code."

Remembering the scarcity of tools for the TMS9900, back in the
seventies, and its early demise, it seems understandable that the AVR
should be supported by a warranted commercial compiler. (Regardless of
the fact that gcc has unequalled standing amongst multi-platform
developers.)

However, if the AVR development environment is to become as mature as
e.g. PowerPC, then not only will Atmel need to update to ELF, but the
two compilers will need to become more compatible, I think. The C
function calling convention (ABI) is not shared, so object files cannot
be shared. (On PowerPC, this can be done if your compiler produces EABI
compliant code.)

But perhaps that's reaching too far into the future.

Regards,
Erik

These views are mine alone. (As if anyone would desire otherwise.)