Re: [avr-llvm-devel] Inline assembly. Mostly just a stub.

[Borja F. <bor...@gm...>]

Haha that's fine Eric, indeed. I don't have very much experience with
inline asm, but if there are ways to make it better than GCC we will add
them, you probably know improvements since you worked for the c lib. One
important thing for this specific topic is that we will able to easily
compile avr-libc since it has a lot of inline asm present.

On the other side for example one thing that avr-llvm is not going to
support is the PROGMEM horrible hack, we support prog data through the
named address space extension which is the correct way of doing it. Ok
we're breaking gcc compat because users will have to adopt the new way but
adding support for this would be a huge step back. So yes it's all a matter
of balence.
Nice to see that atmel spaces is gaining popularity :)

Tonight i will try to make our code compat again with LLVM ToT.



2013/6/13 Weddington, Eric <Eri...@at...>

> > -----Original Message-----
> > From: Borja Ferrer [mailto:bor...@gm...]
> > Sent: Thursday, June 13, 2013 9:48 AM
> > To: Stepan Dyatkovskiy
> > Cc: avr-llvm
> > Subject: Re: [avr-llvm-devel] Inline assembly. Mostly just a stub.
> >
> > Heh Eric, that's ok, for inline asm there's no room for optimizations
> > but for other compiler features sure :)
> >
>
> I guess what I'm trying to say is this:
> - We don't need to be a slave to what GCC does, if you think there is a
> better way to do something
> - However, I can tell you that GCC compatibility will also be a fast road
> to adoption.
>
> It's a balance between the two. I would like to see a fast adoption of AVR
> LLVM. But I would also like to see AVR LLVM be better than what can be done
> with AVR GCC too, even if takes time to do that. ;-)
>
> If the easy path right now is to implement inline assembly like what GCC
> does, then let me just get out of the way. :-)
>
> Eric
>