RE: [avr-gcc-list] Re: inline control

[Weddington, Eric]

> -----Original Message-----
> From: 
> address@hidden 
> [mailto:address@hidden
> org] On Behalf Of David Brown
> Sent: Monday, February 16, 2009 5:04 PM
> To: address@hidden
> Subject: [avr-gcc-list] Re: inline control
> 
> Weddington, Eric wrote:
> > 
> > 
> >> -----Original Message----- From: 
> >> address@hidden 
> >> [mailto:address@hidden org]
> >> On Behalf Of David Brown Sent: Monday, February 16, 2009 2:32 PM 
> >> To: address@hidden Subject: [avr-gcc-list] Re: inline
> >> control
> >> 
> >> 
> >> Have you tried -frtl-abstract-sequences ?
> >> 
> > 
> > In my tests, that switch did not help one whit, and it's severely
> > broken on several other targets to the point that the gcc folks have
> > removed that switch (or will do so). Too bad, really, because the
> > theory of what it does has merit (IMO).
> 
> I haven't tried it myself, and needn't bother now!  It could 
> easily have 
> been a useful trick for small avr's (ImageCraft's avr 
> compiler does this 
> - they call it their "code compressor").

Sure, it's a known optimization technique. The gcc folks asked if anyone was 
going to maintain that switch, and no one stepped forward to get it fixed on 
the targets that it was broken on (like x86, for one). I think that a properly 
implemented version of that would be very useful for the avr.