Re: [MRMC] Bug in Fox-Glynn algorithm implementation

Dear Elvio Amparore,

Thank you for the bug report. We'll correct it in our SVN branch.

Could you perhaps forward us your version of the Computing Poisson Probabilities paper?

We have a version in which the parentheses are actually missing.

Viet Yen

On Dec 14, 2009, at 17:05 , Elvio Amparore wrote:

Dear MRMC developers,

I believe I've found a bug in mrmc. When using transient analysis, MRMC outputs:

   >>>>ERROR: Fox-Glynn: lambda >= 25, underflow. The results  

are UNRELIABLE.

which, I think, is the result of a wrong formula in foxglynn.c, line 199:

   k_prime = k_rtp * sqrt_2 + 3.0 / 2.0 * sqrt_lambda;

This line should be:

   k_prime = k_rtp * sqrt_2 + 3.0 / (2.0 * sqrt_lambda);

as written in Corollary 3 of the "Computing Poisson Probabilities" paper. Without this line, mrmc always thinks that the right Poisson tail will underflow, writing the warning message.

Amparore Elvio _______________________________________________ MRMC-users mailing list MRMC-users@lists.rwth-aachen.de https://mailman.rwth-aachen.de/mailman/listinfo/mrmc-users

Viet Yen -- http://moves.rwth-aachen.de/%7Enguyen