In order to distribute the large amounts of data that Phylanx will process across a cluster, our team needs to be able to calculate an optimal distribution and layout for this data which will minimize the execution time. This distribution and layout is known as tiling. Determining an efficient tiling is difficult especially as the execution graphs become increasingly complex. In this seminar, Indranil presents the mathematical theory behind approximations we may use to tile the data provided to Phylanx. In the presentation, he covers the definition of graphs and hypergraphs, reviews constraints satisfaction problems, discusses NP-Hardness, and works an example.
You can find links to the seminar materials below:
Seminar Video: https://youtu.be/8yL9KhnFqxM