[R-SIG-Mac]timing R and R and R and R

[Stefano Iacus]

Jan, thank you for this test.

What would be nice to do it to integrate the Carbon GUI of R into the 
Darwin Mach-O port.
After R 1.4 will be released I'll be mostly involved in this and I'm not 
sure yet if I'll continue to develop the Carbon version as Apple is 
going to abandon old MacOS as MacOS X is now more than 1 year old. (With 
this I mean that I'll not implement new features for the Carbon version 
but just recompile it for one or two releases after 1.4).
I hope that in the end I can finally attack the quartz device so that 
the darwin version can have its own Aqua-gui and quartz-device.


Probably this list is a good place to put this question:

does it take sense to continue to support old MacOS'es ?

I'm looking forward for comments on this.

Stefano

On Marted́, novembre 13, 2001, at 02:42 , Jan de Leeuw wrote:

> We have two versions of R-1.3.1. The first one is the Carbon version
> of Stefano Iacus. This is a CFM/PEF version, so it can run on OS 9.x
> and on OS X 10.x. We can run this version in OS 9.x in two ways: by
> booting into OS (.x and by running it in the Classic environment in
> OS X.  The second version we use is the Darwin/X11 version,
> which only runs on OS X 10.x, because it is a Mach-O program.
>
> We have a single G4 with two 800MHz CPU's. It runs MacOS
> 9.2.2b7c2 and MacOS X 10.1.2 (5M64).
>
> We have a single test. Define
>
> hilbert<-function(n) 1/(outer(seq(n),seq(n),"+")-1)
> system.time(eigen(hilbert(xxx)))
>
> where xxx is either 500 or 1000.
>
> For xxx equal to 500 we need 21 seconds under OS 9, 20 seconds
> under OS 10, and 14 seconds under Darwin. It requires 21
> seconds under Classic.
>
> For xxx equal to 1000 we need 177 seconds under OS 9, 173
> seconds under OS 10, and 126 seconds under Darwin. Classic
> uses 176 seconds.
>
> Thus for this particular example Darwin is about 40% faster than
> OS 10 Carbon, which may be about 1-2% faster than OS 9 Carbon.
> There can be various reasons, of course. The compilers are
> different. CFM applications are translated by the OS to
> Mach-O applications, which could take some time. It is quite
> possible that SMP works better for Mach-O under Darwin
> (this is a two-processor machine, after all). In any case, the
> speed difference seems significant.
>
> It does not seem to make a difference if we run the Carbon program
> natively in OS X, or natively in OS 9, or in Classic in OS X. That's
> quite remarkable. Of course if we want to run it natively under OS 9
> we have to give R a lot of memory, and both in OS 9 and Classic
> the cooperative MP freezes the system during the computations.
> ===
> Jan de Leeuw; Professor and Chair, UCLA Department of Statistics;
> US mail: 9432 Boelter Hall, Box 951554, Los Angeles, CA 90095-1554
> phone (310)-825-9550;  fax (310)-206-5658;  email: deleeuw@stat.ucla.edu
> homepage: http://www.stat.ucla.edu/~deleeuw
> ========================================================
>           No matter where you go, there you are. --- Buckaroo Banzai
>                    http://www.stat.ucla.edu/~deleeuw/sounds/nomatter.au
> ========================================================
>
> _______________________________________________
> R-SIG-Mac mailing list
> R-SIG-Mac@stat.math.ethz.ch
> http://www.stat.math.ethz.ch/mailman/listinfo/r-sig-mac
>