Abstract
This chapter presents dataflow paradigm in general and cyclic execution graphs, auto-loop offsets, and counters as key points for acceleration and discusses implementation details of iterative rule based algorithms on the dataflow accelerators. Auto-loop offsets create buffers in cyclic execution graphs for streaming results from previous iteration to the next. Counters control input and outputs of the execution graph based on auto-loop offsets. It is shown how part of an algorithm (iterative steps) can be migrated using advanced optimization constructs.
Reference5 articles.
1. AhmadR.KhanumA. (2008). Document topic generation in text mining by using cluster analysis with EROCK. International Journal of Computer Science & Security.
2. Ali, H. I., Akesson, B., & Pinho, L. M. (2015). Generalized extraction of real-time parameters for homogeneous synchronous dataflow graphs. 2015 23rd Euromicro International Conference on Parallel, Distributed, and Network-Based Processing, 701-710.
3. Maxeler Data-Flow in Computational Finance
4. Induction of Fuzzy-Rule-Based Classifiers With Evolutionary Boosting Algorithms
5. Comparing Controlflow and Dataflow for Tensor Calculus: Speed, Power, Complexity, and MTBF