A Monte-Carlo test program with vectorized geometry routines
A Monte-Carlo test program with vectorized geometry routines
Utilization of parallel processing computers to perform radiation transport calculations means to perform parallel computing on problems that are of merely consecutive nature, namely the histories of single particles interacting with matter. Parallelism can be achieved with some success orily by con...
Saved in:
Personal Name(s): | Cloth, P. |
---|---|
Filges, D. | |
Contributing Institute: |
Publikationen vor 2000; PRE-2000; Retrocat |
Imprint: |
Jülich
Kernforschungsanlage Jülich, Verlag
1988
|
Physical Description: |
44 p. |
Document Type: |
Report Book |
Research Program: |
Addenda |
Series Title: |
Berichte der Kernforschungsanlage Jülich
2202 |
Link: |
OpenAccess OpenAccess |
Publikationsportal JuSER |
Utilization of parallel processing computers to perform radiation transport calculations means to perform parallel computing on problems that are of merely consecutive nature, namely the histories of single particles interacting with matter. Parallelism can be achieved with some success orily by considering a certain number of particle histories simultaneously, i. e. one must try, to do what nature does, when particles interact independently with matter at the same time. We must find an algorithmthat can treat an ensemble of particles in a similar way. We must, however, be aware of the fact that the today available supercomputers are SIMD machines, i.e. they can only perform the same action on all elements of a vector (Single Instruction - Multiple Data). This makes a big difference compared to nature, and most of the difficulties in vectorization of Monte Carlo programs arise from this fact. |