[R] R under Pocket PC

Peter Dalgaard p.dalgaard at biostat.ku.dk
Wed Nov 10 12:03:36 CET 2004


"Stuart Leask" <stuart.leask at nottingham.ac.uk> writes:

> > One concern I recall from looking into this a while back (hopefully
> > not founded) was the issue with floating point handling on the ARMs
> > and similar PDA CPUs.
> >
> > This was relevant to the familiar/intimate linux distro's around 18-24
> > months ago, when I was considering an IPAQ or Zaurus to replace my old
> > but not dead yet palm.
> 
> Thanks to the excellent work of the R developers in keeping all the code
> quite standard, and some specific work by Simon Pickering (Bath University),
> I have had R working (with graphics) on my Sharp Zaurus for some time.
> Sadly, checking today his site is down, but binaries for the Zaurus are
> available.
> 
> There was a problem with NA handling that needed specifically addressing,
> and some missing fonts, but the fpu emulation worked fine. This was using
> X11, keypebble and VNC. It could _just about_ run in the standard RAM -
> although was a lot happier with extra memory. Most impressive, and well done
> Simon (and the R team).

Right. As I recall it, the FPU/NA issue was a case of configure
getting the byte-order wrong and not (as was believed for a while) a
problem with non-IEEE arithmetic.
 
> However, a note of caution - presumably due to the lack of a hardware fpu,
> and perhaps also the relatively slow access speed of SD ram, benchmarks on
> the zaurus ran anything up to (wait for it!) 100 times slower than a pentium
> of the same clock speed. This has sadly left R on my Zaurus largely idle, as
> in practice I found it just too slow to be usable.

Hmm. So you're probably limited to problems that require on the order
of tens of milliseconds FPU time on a normal machine. I suppose that
there are usages where this is feasible.

> I fear this problem could afflict any handheld implementation, until they
> start putting fpu's on these chips.

Yes. Now where are the video gamers when you need them?

> (PS. I have the same problem with a mini-ITX 'silent PC' - many of these
> low-power cpus (eg. Eden) lack hardware FPU. They run many office apps fine,
> but when I benchmarked with a stats program, I found it 50 times slower than
> a pentium of equivalent clock speed! I am advised that even those that do
> have hardware FPU eg. the Nehemiah chip are only 50-60% as fast as a
> pentium-class cpu of similar clock speed)

Hmm, the newer ones (Eden-N, Eden-ESP) claim "full-speed FPU". I've
been toying around with the idea of building a system around the
MythTV stuff to get a harddisk video recording, etc. Could be kind of
fun to run R on it with display to the TV (but could you control R
with the TV remote control?)

-- 
   O__  ---- Peter Dalgaard             Blegdamsvej 3  
  c/ /'_ --- Dept. of Biostatistics     2200 Cph. N   
 (*) \(*) -- University of Copenhagen   Denmark      Ph: (+45) 35327918
~~~~~~~~~~ - (p.dalgaard at biostat.ku.dk)             FAX: (+45) 35327907




More information about the R-help mailing list