importing code in IR (GIMPLE and/or RTL)

Luchezar Belev l_belev@yahoo.com
Sat May 15 15:29:00 GMT 2004


--- Robert Dewar wrote:
> > I see. That's not a technical issue but some sort of a legal one.
> > In that case probably a new clause in the gcc license could be added
> > stating explicitely that any software doing so (feeding IR into the
> > compiler at some point) will be considered "derived work" for the
> > purpose of the GPL, and this should eliminate the possibility for
> > unfree front end, right?
> 
> No, this is not so simple, the license can only control copying.
> And furthermore it is not productive to discuss this issue on this list.

Ok, as you say.

I have another question (not related with this one).
Won't it be possible the order of the optimizing passes not to be hardcoded,
but to be controlled by a config file or something.
I mean that while I'm experimenting with how gcc optimizes some specific
cases I come across many cases where the particular order of the passes
is not optimal. If the order is configurable, it would expose a great
opportunity for experimenting and potential for gradually improving
the gcc's optimizing ability. This includes the possibility any particular
pass to be repeated any number of times. So we have an ordered sequence like
(A,B,D,C,A,D,B) where each letter denotes one particular pass.
For example for -O4, -O5 and so on we could have ever longer sequences.

Lucho.



	
		
__________________________________
Do you Yahoo!?
SBC Yahoo! - Internet access at a great low price.
http://promo.yahoo.com/sbc/



More information about the Gcc mailing list