Statistical analysis of simulation generated time series: Systolic vs. semi-systolic correlation on the Connection Machine (Q1195165)

The authors are interested in questions connected with the simulation of time series on parallel computing structures, especially algorithms for evaluation of correlation. At first, detailed semi-systolic and full-systolic algorithms for the Connection Machine are presented. The semi-systolic algorithm makes use of an intrinsic, micro-programmed global-add reduction function which is implemented extremely well on the Connection Machine. On the other hand, the full-systolic algorithm which makes use only of local communication and computation operations turns out to be substantially superior to the semi-systolic scheme whose basic step involves a non-local sum computation that extends over the entire machine. Complexity considerations and the discussion of the real-life implementation results conclude this interesting paper.