Re: problems with gsl_linalg_SV_decomp

[Brian Gough <bjg at network-theory dot co dot uk>]

David Clifford writes:
 > 
 > I compiled the program using a make file with the lines
 > 
 > svd:    svd.c
 >         g++ -g -o svd svd.c -lgsl -lgslcblas
 > 
 > I am running the program on red hat linux release 7.2 (enigma)

Hi, 

The gcc 2.96-98 compiler on redhat 7.2 has known problems with
optimisation. These should show up as "make check" failures on
installation.  If you use another version of gcc it should work ok
(see below).  If you installed from precompiled rpms let us know.

See http://sources.redhat.com/ml/gsl-discuss/2002-q2/msg00108.html
for more information

regards
Brian Gough

bjg|debian> g++ -Wall -g svd.c -lgsl -lgslcblas -lm  
                             # gsl was compiled with gcc-2.7.2.3
bjg|debian> ./a.out 
our matrix is

2.000000 2.000000 0.000000 
0.000000 0.000000 0.000000 
0.000000 0.000000 5.000000 

defining variables for svd...
getting the svd of the matrix...
finished singular value decomposition

U is

0.000000 1.000000 0.000000 
0.000000 0.000000 1.000000 
1.000000 0.000000 0.000000 

V is

0.000000 0.707107 -0.707107 
0.000000 0.707107 0.707107 
1.000000 0.000000 0.000000 

s is

5.000000 2.828427 0.000000