[R] resetting the interpreter
Prof Brian Ripley
ripley at stats.ox.ac.uk
Fri Apr 23 10:01:09 CEST 2004
On Fri, 23 Apr 2004, Richard A. O'Keefe wrote:
> Gabor Grothendieck <ggrothendieck at myway.com> wrote:
> If you haven't tried 1.9.0 yet it does start up faster than 1.8.1.
>
> For what it's worth, on my SunBlade 100,
> R 1.7.1 started in 4 seconds of CPU time, 25 seconds of wall-clock time
> R 1.9.0 starts in 3 seconds of CPU time, 20 seconds of wall-clock time.
>
> I measured the time by doing
> % time R
> > quit("no")
Better to do
% time R --vanilla < /dev/null
That does seem a rather slow machine: our Sun server manages it in 1.0.
> It's really good to see *any* improvement in startup time, and I must
> thank the R developers for even *trying* to do anything about it. There's
> some way to go yet; octave starts in under a second.
It is documented in the NEWS file:
o There has been a concerted effort to speed up the startup of
an R session: it now takes about 2/3rds of the time of 1.8.1.
and BTW without those efforts it would have been 1.5x the time of 1.8.1.
If you want a much greater increase, remove `methods' from the list of
default packages as it is responsible for about half the current startup
time. And if you want a really fast startup with minimal facilities, use
no default packages.
On my office desktop (RH8.0, Athlon MP2600):
time R --vanilla < /dev/null
0.873u 0.035s 0:00.91 98.9% 0+0k 0+0io 1098pf+0w
setenv R_DEFAULT_PACKAGES "utils,graphics,stats"
time R --vanilla < /dev/null
0.505u 0.041s 0:00.55 98.1% 0+0k 0+0io 1090pf+0w
setenv R_DEFAULT_PACKAGES NULL
time R --vanilla < /dev/null
0.099u 0.017s 0:00.11 90.9% 0+0k 0+0io 1033pf+0w
There are planned improvements (hopefully for 2.0.0 next October) which
will load objects on first use and so about halve the second of these
times.
--
Brian D. Ripley, ripley at stats.ox.ac.uk
Professor of Applied Statistics, http://www.stats.ox.ac.uk/~ripley/
University of Oxford, Tel: +44 1865 272861 (self)
1 South Parks Road, +44 1865 272866 (PA)
Oxford OX1 3TG, UK Fax: +44 1865 272595
More information about the R-help
mailing list