Re: [avr-gcc-list] Avr-gcc Produces Incorrect Code with -Os

[Tristan Gingold]

On May 23, 2008, at 9:49 AM, David Brown wrote:
> These "atomic access" builtins are very much specific to the  
> Itanium - they are basically wrappers around some assembly  
> primitives, for use in operating system fundamentals such as locks  
> and synchronisation primitives.

Not really.  These primitives are based on an atomic compare and  
exchange operation which is pretty common
(at least available on x86).

>   For other targets, people normally just use a couple of lines of  
> inline assembly - for the Itanium, assembly language is so hideous  
> that it's worth having these builtin wrappers.

Not at all.  These builtins were defined by intel to ease portability.

> While I think it is worth having a set of consistent atomic access  
> builtins that work across gcc ports, I'm not sure the Itanium's  
> functions are the best model.

These operations are useful for SMP - which is not what AVR users  
care about!


Tristan.