[R] performance gap between R 1.7.1 and 1.8.0

Martin Maechler maechler at stat.math.ethz.ch
Sat Nov 29 12:58:38 CET 2003

>>>>> "AndyL" == Liaw, Andy <andy_liaw at merck.com>
>>>>>     on Fri, 28 Nov 2003 22:19:27 -0500 writes:

    AndyL> Dear R-help, A colleague of mine was running some
    AndyL> code on two of our boxes, and noticed a rather large
    AndyL> difference in running time.  We've so far isolated
    AndyL> the problem to the difference between R 1.7.1 and
    AndyL> 1.8.0, but not more than that.  The exact same code
    AndyL> took 933.5 seconds in 1.7.1, and 3594.4 seconds in
    AndyL> 1.8.1, on the same box.

I understand your concern.  I do believe that it must be some
peculiarity of your (or the non-standard packages you used

Some wild guessing: 
Looking at your profiling table, and seeing  read.dcf()  makes
me think about library() -- which does call read.dcf() only when
called on a package that's not yet attached --- and the
bioconductor extra package management -- which for me makes only
the first time (!) attachment of bioconductor packages
relatively slow.  
I still have no idea how this could influence bootstrapping,
unless you detach() and attach() packages many times.
In any case, it might depend on the exact list (and order!?) of
packages present in search() when calling your code.

A very first step of diagnosis might be to activate
  options(verbose = TRUE)

A step further might be to patch read.dcf such that it prints
       if(getOption("verbose")) { <<print what I am trying to read>> }


More information about the R-help mailing list