Reproduce “Multiprocessor Scheduling of a Multi-mode Dataflow Graph Considering Mode Transition Delay” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Hanwoong Jung, Hyunok Oh, Soonhoi Ha
Researchers from Seoul National University developed a genetic algorithm framework for multiprocessor scheduling of multi-mode dataflow graphs, minimizing processor count while ensuring throughput by explicitly accounting for task migration costs and mode transition delays. This approach consistently required fewer processors and robustly handled increased migration overheads compared to baseline methods.
Key insights
- Allowing task migration between modes, when optimized holistically and with explicit cost consideration, can significantly reduce the number of required processors compared to fixed-mapping strategies.
- A conservative calculation of individual mode throughput requirements, accounting for all mode transition delays, is essential to guarantee overall system throughput.
- Simultaneous scheduling of all modes, rather than independent mode optimization, leads to more robust and resource-efficient solutions, especially as task migration overheads increase.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.