Exploring Parallelism of a BRDF algorithm using CUDA
Hyuck Yi, Sunho Baek, JunSeong Kim · 2022
While parallel hardware has become common, most typical engineers and scientists tend to follow the traditional single-core processing approach causing major drawbacks in their developments. In this study, to fully utilize the computing power we present practical parallelization approaches for typical engineers. We implement a BRDF estimation algorithm pursuing parallelism at various levels using CUDA. Experiments with a set of real environmental data show that even a simple parallelization can drastically improve performance: the speedup is between 4.93 and 64.10 depending on the approach in parallelization and the problem size. Little efforts in parallel programming can bring efficient computing.