On Sat, Sep 7, 2013 at 12:59 PM, Marc Glisse <marc.gli...@inria.fr> wrote:
>> Furthermore, I do think that the compiler should have special nodes >> for both standard placement new and the global operator new functions, > > > That's one way to do it. Since this is the first time I look at those, I > don't really see the advantage compared to the current status, but I trust > you. What would you do with this special-node placement new? placement new really is about "calling" a contractor, and marking the beginning of the lifetime of a new object, hence aiding lifetime-based alias analysis. > Keep it as is > until after vrp so we can use the !=0 property and then expand it to its > first argument? Or expand it early to the equivalent of the library code I > wrote above? > > >> as I explained in previous messages. > > > I couldn't find them, sorry if they contained information that makes this > post irrelevant. http://gcc.gnu.org/ml/gcc-patches/2013-08/msg01276.html http://gcc.gnu.org/ml/gcc-patches/2013-08/msg01291.html -- Gaby
