LuisaGroup / benchmark

Performance benchmark scripts for LuisaRender, Mitsuba 3, and PBRT-v4

Benchmark

Python scripts to automatically benchmark the performance of LuisaRender, Mitsuba3, and PBRT-v4.

Usage

• Please edit run_config.py before running the benchmarks. The dictionary target_settings specifies the renderers and the corresponding settings. You will have to fill in the renderer_settings["<renderer>"]["exe"]["path"] fields for the scripts to successfully find the executables for the renderers.

  Please note that for PBRT-v4, the rr_depth option is ineffective, since the parameter is hard coded in the renderer's source. You will have to modify the source code and recompile it to change rr_depth.

• After configuration, please execute the run.py script with Python 3 to start the benchmarks. The script generates modified scene description files under <renderer>/scenes and creates the outputs folder to hold the results. For example, the renderings are stored in outputs/pictures and timings in outputs/results.csv.