Speaker: Shruthi Kannappan
Topic: Dynamic MaxFlow in Starplat Cuda Backend
Description: Dynamic MaxFlow in Starplat Cuda Backend: This project involves the design, implementation, and evaluation of a dynamic maximum flow algorithm that efficiently handles batch updates of edge capacities. The algorithm is implemented using StarPlat, a high-level Domain-Specific Language (DSL) for graph algorithms, which generates CUDA code for execution on many-core GPU architectures. The generated code significantly reduces recomputation overhead by reusing results from a previously computed static flow while being more efficient in terms of both time and memory usage when compared to recomputing the maximum flow from scratch after each update.
Time 5 pm to 6 pm (IST)
Presenter: Shruthi Kannappan
About the Presenter: https://www.linkedin.com/in/shruthi-kannappan/