|
From: | John W. Eaton |
Subject: | [Octave-bug-tracker] [bug #41065] Single precision matrix inverse gives bogus singular matrix warning |
Date: | Wed, 08 Jan 2014 03:11:14 +0000 |
User-agent: | Mozilla/5.0 (X11; Linux x86_64; rv:17.0) Gecko/20130917 Firefox/17.0 Iceweasel/17.0.9 |
Follow-up Comment #5, bug #41065 (project octave): Thinking about this a little more, I'm not sure the fix is the right thing to do. We want to test rcond+1 == 1 but in float precision for floats and double precision for doubles. What we are doing instead is always doing the test as a double. I think that's wrong. For example: octave:2> x = eps ('single') x = 1.1921e-07 octave:3> x = eps ('single') / 2 x = 5.9605e-08 octave:4> x + 1 == 1 ans = 1 octave:5> double (x) + 1 == 1 ans = 0 I also think this is worth applying to stable. It seems bad if Octave can't even invert a simple float matrix without issuing a warning. I checked in the following change on stable and merged it with gui-release and default: http://hg.savannah.gnu.org/hgweb/octave/rev/36057e2411f8 _______________________________________________________ Reply to this item at: <http://savannah.gnu.org/bugs/?41065> _______________________________________________ Message sent via/by Savannah http://savannah.gnu.org/
[Prev in Thread] | Current Thread | [Next in Thread] |