Скачать или смотреть Comparing sparse matrix distributions

This video presents two methods for distributing a sparse matrix over the processors of a parallel computer, the fine-grain method and the medium-grain method. These methods and the original Mondriaan method are compared by running numerical experiments on a diverse test set of sparse matrices from the SuiteSparse collection. The medium-grain method has the lowest resulting communication volume and it is also the fastest of the three methods.

In the video, I furthermore explain the usefulness of the geometric mean for comparing widely differing test instances, and the expressive power of performance profiles, which capture the behaviour of several different methods for a large number of problem instances in a single plot.

This video corresponds to Section 4.6 of the book Parallel Scientific Computation: A Structured Approach Using BSP, Second Edition, by Rob H. Bisseling, Oxford University Press, 2020.

An expanded set of slides, solutions to the homework questions, and software accompanying the book can all be found on my personal book page:
https://webspace.science.uu.nl/~bisse...