Speaker
Description
The FLUKA general purpose radiation transport Monte Carlo code being developed and maintained by CERN has adopted modern software development standards including a formal quality assurance process. This includes the FLUKAVAL testing framework that takes into account the specific needs of testing a Monte Carlo radiation transport simulation code.
FLUKAVAL is a Python-based command line application that enables semi-automatic batch submission, processing, validation and reporting of test cases. Any FLUKA input and any dedicated tests having access to the FLUKA code at the model level can be adapted and integrated in a few steps into the test framework. FLUKAVAL makes extensive use of the git version control system to store simulation and reference data sets on the CERN GitLab instance. In its current version, FLUKAVAL is optimised for the submission of a large number of test cases to the CERN Batch Service (LXBATCH), which provides distributed computing power with over 220'000 CPU cores.
FLUKA output data are stored in the standardised JSON format, which is compatible with the Geant Validation Portal, to simplify future direct comparisons of identical test cases with Geant4.
The results are processed, validated and compiled in a comprehensive evaluated report, which allows the direct comparison of results from the FLUKA version under validation with results obtained with previous FLUKA versions, other radiation transport codes and/or experimental data. A quantitative analysis is performed in addition to a qualitative comparison whenever possible.
We will demonstrate on real examples how reference benchmark data and output from simulation calculations are analysed and compiled, and how this can be extended to a large amount of test cases. The focus will be on the capabilities of the FLUKAVAL framework with respect to cross-section data for nuclear reactions.