Optimisation of reaction rate calculations in ALEPH

Objectives
1. An optimal implementation of the calculation of the reaction rate convolution integral.
• An optimal data structure for the cross sections.
• Optimal way of calculating the reaction rate integral using point-wise data where both
functions have different grid points. Is it worth creating a “shared” grid to be able to use the
trapezoid rule? Should this common grid be stored or regenerateed each time?
• Analyse the potential of parallellisation taking into account I/O limitations (each process
reading from the same file(s)).
2. A strategy to “coarsen” the cross section and neutron flux grid up-front to reduce computation
time and memory/storage usage.
• Can we, in general or based on a typical application, coarsen both the cross section grid and the
energy grid for neutron spectrum binning without loosing accuracy?
3. Implement the on-the-fly Doppler broadening
• Implement the basic OTFDB approach by calculating the integral above.
• Analyze if and how calculations can be cached. Analyze the trade-off between CPU needs, RAM
use, I/O burden and storage needs.
We expect proof-of-concept implementations that we can integrate in our ALEPH2 code. We will
provide the student with the necessary data sets and typical use case examples