[R] variations in how long commands take

[Prof Brian Ripley]

On Tue, 17 Jul 2001, David Kane  <David Kane wrote:

> We are performming a series of S+ 6.0 versus R 1.3 comparisons. (If anyone has
> already written scripts for this purpose, we would be eager for a copy.) The
> purpose is to see if there are any "gotchas" -- places were S+ works and R
> doesn't -- that would prevent us from going all the way over to R. (We have
> lots of legacy code, so the conversion costs are non-trivial.) As part of this
> project, we have some simple matrix code that takes a similar amount of time in
> S+ and in R (with R a little faster) the *first* time that we run it. However,
> repeated invocations of this code in R produce slower and slower times (often
> dramatically so). Here is a simple example (with a new R session):


You mean system.time, nor proc.time!  proc.time is the cumulative usage
of the session and can only increase.  Here's my results:

> m <- matrix(1, 400, 40000)
> system.time(x0 <- apply(m, 2, sum))
[1]  6.98  1.03 17.40  0.00  0.00
> system.time(x0 <- apply(m, 2, sum))
[1] 7.02 0.44 7.55 0.00 0.00
> system.time(x0 <- apply(m, 2, sum))
[1] 6.88 0.48 7.36 0.00 0.00


>
> > version
>          _
> platform sparc-sun-solaris2.6
> arch     sparc
> os       solaris2.6
> system   sparc, solaris2.6
> status
> major    1
> minor    3.0
> year     2001
> month    06
> day      22
> language R
>
> > m <- matrix(1, 400, 40000)
> > proc.time(x0 <- apply(m, 2, sum))
> [1] 21.88  3.16 40.13  0.00  0.00
>
> So far, so good. This is more than twice as fast as the same thing in S+
> 6.0. But when I re-issue this command, I get:
>
> > proc.time(x0 <- apply(m, 2, sum))
> [1] 37.09  4.65 63.13  0.00  0.00
> > proc.time(x0 <- apply(m, 2, sum))
> [1] 52.34  5.44 85.29  0.00  0.00
>
> So, my first question is: Why do the times keep going up? I first assumed that
> there were some problems with overwriting x0 again and again, but I then remove
> everything from the workspace and start "fresh":
>
> > remove(list = ls())
> > ls()
> character(0)
> > m <- matrix(1, 400, 40000)
> > proc.time(x0 <- apply(m, 2, sum))
> [1]  72.50   6.13 137.08   0.00   0.00
>
> Second question is: Why does this continue even with an empty workspace? I do
> not see this behavior in S+.
>
> I have read ?Memory and played around with things like gc() and mem.limits(). I
> am aware that R and S+ are very different in their memory usage. But I suspect
> that there is something fundamental about R that I am missing . . .
>
> Any suggestions or pointers would be much appreciated.
>
> Thanks,
>
> Dave Kane
> -.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-
> r-help mailing list -- Read http://www.ci.tuwien.ac.at/~hornik/R/R-FAQ.html
> Send "info", "help", or "[un]subscribe"
> (in the "body", not the subject !)  To: r-help-request at stat.math.ethz.ch
> _._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._
>

-- 
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 272860 (secr)
Oxford OX1 3TG, UK                Fax:  +44 1865 272595

-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-
r-help mailing list -- Read http://www.ci.tuwien.ac.at/~hornik/R/R-FAQ.html
Send "info", "help", or "[un]subscribe"
(in the "body", not the subject !)  To: r-help-request at stat.math.ethz.ch
_._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._