From: Terje Slettebø (tslettebo_at_[hidden])
Date: 2002-11-10 11:57:16
>From: "Gennaro Prota" <gennaro_prota_at_[hidden]>
> On Sat, 09 Nov 2002 21:00:00 -0500, David Abrahams
> <dave_at_[hidden]> wrote:
>
> >The RVO doesn't fall into the category of back-end or
> >intermediate-language optimizations that compilers usually let you
> >turn on and off. It has to happen in the front-end, since it often
> >implies observable differences in the code's behavior.
>
> Well, I'm not a compiler expert, but several compilers have switches
> that actually control the front-end. So there could be one for
> NRVO/URVO too. Or is there any other reason which prevents that?
Since, as Dave said, it may change the observable behaviour, it's perhaps
not such a good idea to have an optimisation switch that changes the
observable behaviour. That could be confusing.
Regards,
Terje