In article <20020716230156187+0200@news.povray.org>,
 WhiteGandalf <zus### [at] liberoit> wrote:

> In <chr### [at] netplexaussieorg> Christopher 
> James Huff  wrote:
> > You forgot to take the compiler into account...the version of gcc in 
> > the  development tools is nowhere near as good at optimization as 
> > Metrowerks  CodeWarrior (which is used for the official version). The 
> > next update of  the development tools will use a more up to date 
> > version of gcc which is  apparently much better at optimization of PPC 
> > code, I don't know how it  compares to CodeWarrior. If you want to 
> > compare 3.1 and MegaPOV, compile  MegaPOV with gcc as well.
> 
> This exaclty was the point. I used gcc and complained about it!

You said "pov from command line should be advantaged.", which seemed to 
say you were comparing gcc-compiled POV 3.1 against the GUI MegaPOV, 
which is compiled with CodeWarrior. What exactly were you doing?


> Well, that's not completely true, after all without preview povray isn't 
> doing much.

Uh, what? You aren't supposed to have the preview on for benchmarks 
anyway. Of course a preview will eat up CPU cycles...but without preview 
it is still a GUI program.


> And I must add, CodeWarrior, at least the older versions, isn't such a 
> speedy compiler. Under os classic where MPW was available it was 
> consistently outperformed as T. Fr?hlich will remind. He sent me an 
> unofficial MPW compile and it was quite faster.

True, but it has gotten much better over time...I'm not sure which is 
faster now, but they are close, and both are much better than gcc 2.95.2.


> strictly speaking "cycles" doesn't make much sense in a unix environment. 
> It is not a single program that decides to acquire or release the CPU 
> but the kernel.

I am aware of that, but POV-Ray Mac works in the Carbon environment, and 
is written to be compatible with pre-OS X versions of Mac OS. This 
probably makes it less efficient (having to constantly stop rendering 
and check for events to stay responsive) and might prevent it from 
getting all the available time. The command-line version shouldn't have 
this problem...at least my command line programs have no trouble pinning 
the processor use at 100%.

-- 
Christopher James Huff <chr### [at] maccom>
POV-Ray TAG e-mail: chr### [at] tagpovrayorg
TAG web site: http://tag.povray.org/

Post a reply to this message