[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [f-cpu] New suggestion about call convention



> You're right. But i thinl this case is a ill one. I hope nobody code like
> that !

It's append and sometime it can be usefull.

> So maybe the 20-80 rule apply. This case could not be optimised but the
> common one could.

The problem is that you couldn't make any different between a source of 
problem and a correct program that you could optimize. So sometime you will 
generate wrong code and a compiler can never generate wrong code because of 
an optimisation.
  Patching gcc with surch an algorithm will never be acceptable, because of 
regression risk ! If we don't find a solution to this problem, we must forget 
this idea !

Cedric
*************************************************************
To unsubscribe, send an e-mail to majordomo@seul.org with
unsubscribe f-cpu       in the body. http://f-cpu.seul.org/