[Octave-bug-tracker] [bug #45651] Crash when inverting matrices sized 10x10 or bigger

[anonymous]

URL:
  <http://savannah.gnu.org/bugs/?45651>

                 Summary: Crash when inverting matrices sized 10x10 or bigger
                 Project: GNU Octave
            Submitted by: None
            Submitted on: Wed 29 Jul 2015 06:58:17 PM UTC
                Category: Octave Function
                Severity: 3 - Normal
                Priority: 5 - Normal
              Item Group: Crash
                  Status: None
             Assigned to: None
         Originator Name: 
        Originator Email: address@hidden
             Open/Closed: Open
         Discussion Lock: Any
                 Release: 3.8.2
        Operating System: GNU/Linux

    _______________________________________________________

Details:

octave --version
  GNU Octave, version 3.8.2
  [...]
   Octave was configured for "x86_64-pc-linux-gnu"
 
%N=11;
N=12;
A=zeros(N);
for n=2:N-1
  A(n,n-1)= 1;
  A(n,n  )=-2;
  A(n,n+1)= 1;
end
A=A(2:end-1,2:end-1);
disp(A);
inv(A);

"Segmentation fault"

This only happens for N=12, not N=11 or less (I cut off the first and last row
and column therefore the actual matrix is N-2xN-2 big). I guess it switches to
some other method of calculation.

The same happens if I use backslash:
B=zeros(N);
for n=1:N
  B(n,n)=1
end
AB

"Segmentation fault"

It doesn't happen for inv(B) though... Or in general it doesn't seem to hapen
for diagonal matrices.


Temporary measure: implement a function mat_inv using a Gauss-Jordan algorithm
myself. (very slow!)





    _______________________________________________________

Reply to this item at:

  <http://savannah.gnu.org/bugs/?45651>

_______________________________________________
  Message sent via/by Savannah
  http://savannah.gnu.org/