I've thought of doing something like that too... building the whole system using icc. Has anyone done this? I'd be slightly worried that some of the builds written in C++ wouldn't compile.
Brian On Wed, 2 Jul 2003, Michael Gruetzner wrote: > That sounds ok to me. But how can I find out what ebuilds support icc. > I'd like to try some others than povray. Maybe QT or KDE woul be good > but I'm not sure if this will work - probably not :-( . > > Michael > > Brian Budge wrote: > > ICC can produce faster code, but it won't always... It tends to be about > > 5% faster on my own > > path tracer. One thing to note is that there are several compiler options > > for optimization. There are things like global interprocedural opts, even > > interfile opts, and you can also compile using statistics from previous > > runs (to aid branch prediction, etc...). > > > > I would say that most of the time, ICC produces slightly faster code. My > > argument is that GCC can do things from the C++ standard that ICC cannot, > > like partial template specialization. > > > > Brian > > > > On Wed, 2 Jul 2003, Michael Gruetzner wrote: > > > > > >>Harald Arnesen wrote: > >> > >>>Martin LORANG <[EMAIL PROTECTED]> writes: > >>> > >>> > >>> > >>>>What is exactely the benefit of ICC ? > >>> > >>> > >>>It produces _much_ faster code than gcc (even the very latest 3.4-stuff) > >>>for most applications I have tested. > >> > >>I just testet it with povray. The gcc compiled version of povray is > >>faster than the icc version. > >>These are the scenes I tested: > >> > >>/usr/share/povray/scenes/balcony/balcony.pov > >> > >>gcc: 125s > >>icc: 190s > >> > >>/usr/share/povray/scenes/abyss.pov > >> > >>gcc: 23s > >>icc: 23s > >> > >>So I don't think, that icc produces faster code. > >> > >>Michael > >> > >> > > > > > > > > -- > > [EMAIL PROTECTED] mailing list > > > > > > > -- [EMAIL PROTECTED] mailing list
