[Octave-bug-tracker] [bug #40683] Unequal results when using element-wise and broadcast power operations

[jose]

Follow-up Comment #2, bug #40683 (project octave):

An easier way to reproduce it


a=100+rand;l=1e-10;d=[a-l;a;a+l];
for pow=1:15
  T(:,pow)=diag(diag(d)^pow)-d.^pow;
endfor
T


In octave 3.8.2 (in Kubuntu), the last columns of T are


Columns 12 through 15:

  -4.0265e+08  -3.4360e+10  -4.3980e+12  -2.8147e+14
   2.6844e+08   5.1540e+10   4.3980e+12   5.6295e+14
  -1.3422e+08  -1.7180e+10  -2.1990e+12  -1.4074e+14



    _______________________________________________________

Reply to this item at:

  <http://savannah.gnu.org/bugs/?40683>

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