Developing a parallel Monte Carlo transport algorithm using large-grain data flow (Q1105357)

Here is a report on the use of LGDF - a method employing a graphic model for describing parallel processing of data, somewhat similar to Petri nets (without explicit use of tokens) -, and on the experiences gained by converting an existing sequential Fortran package with the help of LGDF into a program for a parallel computer, and running it, and improving it in a critical section; the package (GAMTEB for VAX) being a simulating computation of transport of photons in cylinders of carbon. LGDF and the requirements of the simulation (based on Monte Carlo method) are sketched; they are more fully described in listed earlier publications and reports, which include also a description of the used parallel computer Denelcor HEP. Results, especially observed execution times for the sequential and for the converted parallel version, are given to support the feasability of LGDF in such cases.