[Scipy-tickets] [SciPy] #840: Potential subtractive cancellation bug in scipy/interpolate/Rbf.py

SciPy scipy-tickets@scipy....
Fri Jan 9 17:47:20 CST 2009


#840: Potential subtractive cancellation bug in scipy/interpolate/Rbf.py
-------------------------------+--------------------------------------------
 Reporter:  fghorow            |        Owner:  somebody
     Type:  defect             |       Status:  closed  
 Priority:  normal             |    Milestone:  0.7.1   
Component:  scipy.interpolate  |      Version:  devel   
 Severity:  normal             |   Resolution:  wontfix 
 Keywords:  Rbf                |  
-------------------------------+--------------------------------------------
Comment (by fghorow):

 Replying to [comment:3 rkern]:
 > Actually, it turns out that the only time that function gets called,
 there is already a cast to float64. So much for principle.

 Not true. The `_euclidean_norm()` function is called twice. Once at
 "linear system setup" time (where things are indeed cast to float64), and
 once at "solution evaluation" time, where as far as I can tell there is no
 cast.

 I do take your earlier point about calling with the correct precision
 arguments, and actually agree with that philosophy.

-- 
Ticket URL: <http://scipy.org/scipy/scipy/ticket/840#comment:4>
SciPy <http://www.scipy.org/>
SciPy is open-source software for mathematics, science, and engineering.


More information about the Scipy-tickets mailing list