The current GPU port was highly optimized for depleted fuel, full core fission reactor simulation problems. Typically these problems have hundreds of nuclides in the fuel region, which results in a majority of the runtime of the simulation being spent performing macroscopic cross section lookups. While fusion simulation problems can also feature materials with a significant number of isotopes, they also tend to have vastly more complex geometry definitions, leading to a relative increase in ray tracing costs vs. cross section lookup costs.
In this task, we will optimize the ray tracing regions of the code (and any other regions needed for high performance on fusion simulations) so as to ensure optimal performance on both fission and fusion energy simulations.
The current GPU port was highly optimized for depleted fuel, full core fission reactor simulation problems. Typically these problems have hundreds of nuclides in the fuel region, which results in a majority of the runtime of the simulation being spent performing macroscopic cross section lookups. While fusion simulation problems can also feature materials with a significant number of isotopes, they also tend to have vastly more complex geometry definitions, leading to a relative increase in ray tracing costs vs. cross section lookup costs.
In this task, we will optimize the ray tracing regions of the code (and any other regions needed for high performance on fusion simulations) so as to ensure optimal performance on both fission and fusion energy simulations.