← Back to projects
thor: a massively-parallel GPU-accelerated MCRT code

thor: a massively-parallel GPU-accelerated MCRT code

Emission and absorption line features are important diagnostics for the physics underlying extragalactic astronomy. The interpretation of observed signatures involves comparing against forward modeled spectra from galaxy formation simulations as well as more simplified geometries, while including the complex scattering radiative transfer (RT) of resonant emission lines. We introduce thor, a modern C++ radiative transfer code focused initially on resonant emission lines. thor is a high-performance, distributed memory MPI-parallel, multi-target code, running on CPUs, GPUs and other accelerators, yielding large ~10-50x speed-ups compared to previous CPU-only codes.

Benchmark comparison

Benchmark on a common Lyman-alpha radiative transfer “Neufeld solution” (T=10⁴K, τ₀=10⁶) toy problem for different resonant emission MCRT codes. Compiled with Clang 18. CPU: Intel 8360Y with 36 cores at 2.40GHz, GPU: Nvidia A100 40GB. Tested for latest code versions in July 2025.

Weak scaling test

Lyman-alpha radiative transfer weak scaling test for number of volume elements for a cosmological galaxy formation simulation for up to ~200 billion volume elements, memory distributed with up to 15k CPU cores and 200 GPUs.

To be published. Pre-print available.

Official website: thor-rt.org