Sequential Scheduling of Dataflow Graphs for Memory Peak Minimization

Explore a 17-minute conference talk from LCTES 2023 that delves into the optimization of memory peak in sequential scheduling of dataflow graphs. Learn about novel task graph transformations and an optimized branch and bound algorithm designed to minimize memory usage in a broader class of task graphs. Discover how this approach extends to Synchronous DataFlow (SDF) graphs and a new suboptimal method for reducing problem size in large graphs. Gain insights into the researchers' evaluation of their approach on classic benchmarks, demonstrating consistent outperformance of state-of-the-art methods. Understand the implications for computing systems with fixed shared memory constraints and the potential for improved efficiency in various applications modeled with task or SDF graphs.

[LCTES'23] Sequential Scheduling of Dataflow Graphs for Memory Peak Minimization