Parallel implementation of hyper-dimensional dynamical particle system on CUDA

Two distinct approaches to parallel implementation are presented, one aiming at a generalized usage of the fast on-chip resources, the other entirely relying on the GPU's on-board global memory. Despite unambiguous mutual differences in their performance, both parallel implementations deliver major speedup compared to the single-thread CPU solution as well as a better scaling of execution time when increasing both the number of particles and dimensions.