Thanks to libraries like CUDA, modern GPUs offer the possibility of offloading a large amount of calculations to thousands of cores in a SIMD fashion. This is particularly useful when working with simulators and a large number of independent runs of the same model are necessary to investigate its emergent properties. I started developing GPU-powered simulators in 2011 for my Master’s degree thesis, and I keep doing that nowadays. Here is a partial list of our projects:
- cupSODA: coarse-grained adaptive deterministic biochemical simulation
- pySB/cupSODA: cupSODA was integrated into the pySB rule-based modeling framework
- cuTauLeaping: coarse-grained adaptive stochastic biochemical simulation
- ginSODA: coarse-grained adaptive deterministic simulation of arbitrary ODE-based models
- FiCoS: fine- and coarse-grained adaptive deterministic biochemical simulation
- SDEGnO: GPU-accelerated stochastic simulation of cosmic rays propagation within the heliosphere
- cuProCell: GPU-powered stochastic simulation of cell proliferation
- HERESY: accelerated simulation of reaction systems