[R] seg fault: some more comments

[Lorenz Gygax]

Dear all,

Thanks for the recent comments on my segmentation fault problem when using
loess.

I realized that I should have mentioned two things. Firstly, I am already
working with R --vsize 100 and still get a crash at n= 4000. Secondly I am
using R 0.63.1 with a SuSE Linux 5.2

Obviously one thing to do is to upgrade (neither in Martin Maechlers nor
in Brian Ripleys attempts there is an actual seg fault but only an error).

A further observation that I made is that loess's performence seems to
deteriorate if called several times in succession even if gc () is called
in between. (Which might make sense in the view of what the problem might
be regarding memory allocation?)

What I currently did is just use a random sample of my data of acceptale
size (to loess) to actually calculate loess and quit and restart R if I
needed loess several times in succession. This may not be very pretty but
I got what I needed and running the commands in batch mode with a shell
scripts does make it pretty simple, too. (Just a side comment, why is
.lib.loc different in batch than in interactive mode?) 

Regards, Lorenz


-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-
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
_._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._