Re: [Bug-gama] questions about Gnu-Gama

[Ales Cepek]

Hi Jacques,

I have traced the bug down deep to a routine for computing values of normal distributions N(0,1) and it seems that the problem stems from the compiler code optimisation. The source of the function comes as far as to good old FORTRAN days, and I'll try to make it fool proof (there is a test on nullity of a value that fails and results in an endless loop).  I am sending the first results of your network which is quite interesting, I haven't analyse it and neither I can say at the moment if the results are reasonable, you are using constructs I have never met (a kind of point pairs, right?). BTW solving your data took less then 2 seconds real time on my laptop, so the size of the task seems really not to be an issue.

Ales

prelim.zip
Description: Zip archive