[Blackmax] BlackMax bug

[Ian Preston]

Hi guys,

I thought I'd tell you I found a pretty serious bug in the blackmax code
whilst converting it to Java. The decision whether to emit another particle
via hawking radiation or not compares two mass values which are double
precision floating point numbers. However most of the time for the last
decay these are identical up the accuracy of a float. However after the
subtraction spurious digits are obtained and thus the choice to decay is
essentially random. The way to handle this is not to test if Ma > Mb, but to
test if Ma-Mb > epsilon. Where epsilon is the smallest digit of precision
representable with a double precision float (2-53 off the top of my head). I
suspect there are other places where branching on the comparison of two
floats could cause problems as well.

cheers,
Ian

-- 
JPC Lead Developer
Oxford University
http://jpc.sourceforge.net/
http://www-nereus.physics.ox.ac.uk/
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://www.hepforge.org/lists-archive/blackmax/attachments/20100908/33e6436b/attachment.htm