> Seems quite a bad story to me at the moment.
> I assume that octave (or matlab) which are also based on BLAS +
> LAPACK internally can eigen-decompose the matrix correctly?
> Could anyone check this (without much effort) for us?
Not quite what you asked for, but the following works....
library(mgcv)
M <- as.matrix(read.table("thematrix",header=T))
n <- nrow(M)
d <- array(0,n)
um<-.C("mgcv_symeig",as.double(M),as.double(d),as.integer(n),PACKAGE="mgcv")
um[[2]] ## the eigenvalues
... it's directly calling `dsyevd' from lapack as supplied with R. i.e.
it's using lapack + blas from R. This is on R-devel suse linux 9.3
pentium xeon. eigen(M) hangs on the same platform.
However `dsyevd' isn't the routine called by `eigen'... it uses `dsyevr'
instead (the freeze occurs when this calls DSTEGR). Apparently the
difference is that `dsyevr' uses a `Relatively Robust Representation'
[perhaps `relatively robust' is not roust enough in this case :-)]
Simon
>- Simon Wood, Mathematical Sciences, University of Bath, Bath BA2 7AY
>- +44 (0)1225 386603 www.maths.bath.ac.uk/~sw283/
______________________________________________
R-help@stat.math.ethz.ch mailing list
https://stat.ethz.ch/mailman/listinfo/r-help
PLEASE do read the posting guide! http://www.R-project.org/posting-guide.html
