Cosmic-Calc is a project that explores parallel programming concepts through the spatial distribution analysis of galaxies using the two-point angular correlation function. This statistic measures how galaxies cluster relative to a random distribution and can serve as evidence of cosmic dark matter.
The computation was developed in three versions to study parallelization trade-offs: pure C as a baseline, OpenMP for multi-core parallel processing, and a GPU-accelerated CUDA version that achieved a 58x performance boost over the sequential baseline.
References
- Cosmological Calculations on the GPU — the paper this project implements; good starting point for the physics and algorithm
- CUDA C Programming Guide — official NVIDIA reference for kernel launches, memory, and thread hierarchy
- CUDA by Example (Sanders & Kandrot) — gentle intro if GPU programming is new
- OpenMP specification and tutorials — for the multi-core CPU parallel version
- Two-point correlation function (overview) — quick primer on what the statistic actually measures