cmdscale in package mva (PR#1027)

On Tue, 17 Jul 2001, Kurt Hornik wrote:

laurent  writes:

Full_Name: Laurent Gautier
Version: 1.3.0-patched
OS: IRIX 6.5
Submission from: (NULL) (130.225.67.199)

Hello,

The function La.eigen, called by cmdscale in the package mva behaves
an unexplicable way (for me).

The following lines show what happened.  I tried the very same on
linux, and it worked fine.

a <- matrix(c(1,2,3,2),3,3)
a

     [,1] [,2] [,3]
[1,]    1    2    3
[2,]    2    1    2
[3,]    3    2    1

library(mva)
cmdscale(a)

Error in La.eigen(Tmat, symmetric = TRUE) :
	On entry to DSYEV  parameter number 8 had an illegal value

I can confirm this on Naras's IRIX machine (in fact, was just about to
report this myself ...)

Can you put a breakpoint at line 72 of Lapack.c and see what value of lwork
is returned?  Alternatively, it may fail on the first dsyev call, in which
case I would guess that the wrong LAPACK routine is being found (and that's
a plausible explanation).

I get 102 with the built-in BLAS: the answer should be at least max(1, 3*n-1)

What BLAS is in use?  The intention is that the workspace is chosen large
enough to allow the BLAS to work efficiently: it might just be that
the BLAS is reporting erroneously.

-- 
Brian D. Ripley,                  ripley@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-devel 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-devel-request@stat.math.ethz.ch
_._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._

Duncan Temple Lang writes:

For what it is worth, I saw this problem a few weeks ago but didn't
have time to investigate it in depth and managed to find a work
around.  It arose for me on Solaris, specifically when I was compiling
with the native C compiler cc and f77.  Switching to gcc and f77
removed the problem.  (There is a possibility that I have this
backwards, but I think my memory is correct on this point.)