GPU accelerated optical light propagation in CORSIKA 8

Optical photons, created from fluorescence or Cherenkov emission in atmospheric cascades induced
through high energetic cosmic rays are of major interest for several experiments. Experiments
like CTA require a significant amount of computing time and funds for the simulation with
CORSIKA 7.
Since individual photons don’t interact they can be simulated without any order. The traditional
sequential approach is simple but leads to reduced utilization of modern hardware infrastructure.
The calculations done on each photon have low complexity, compared to the other aspects of the
simulation. This, as well as the fact that, besides information about the photon itself, nearly no
additional data is needed, favors a data-parallel approach in which several photons are propagated
concurrently. The new CORSIKA 8 framework enables the implementation and verification of
these methods.
With the use of dedicated high parallel acceleration hardware like GPUs the possible benefits with
this data-parallel approach are even higher.