Institutional Repository [SANDBOX]
Technical University of Crete
EN  |  EL

Search

Browse

My Space

Comparing C and SystemC based HLS methods for reconfigurable systems design

Georgopoulos Konstantinos, Malakonakis Pavlos, Tampouratzis Nikolaos, Nikitakis Antonios, Chrysos Grigorios, Dollas Apostolos, Pnevmatikatos Dionysios, Papaefstathiou Ioannis

Full record


URI: http://purl.tuc.gr/dl/dias/17C966D7-0032-4083-91B9-B1F6CCDDD66A
Year 2018
Type of Item Conference Full Paper
License
Details
Bibliographic Citation Κ. Georgopoulos, P. Malakonakis, N. Tampouratzis, A. Nikitakis, G. Chrysos, A. Dollas, D. Pnevmatikatos and I. Papaefstathiou, "Comparing C and SystemC based HLS methods for reconfigurable systems design," in 14th International Symposium on Applied Reconfigurable Computing, 2018, pp. 459-470. doi: 10.1007/978-3-319-78890-6_37 https://doi.org/10.1007/978-3-319-78890-6_37
Appears in Collections

Summary

This paper provides an extensive analysis of the key characteristics, efficiency and overall user-friendliness that stem from the use of two different input methods for the design of High-Level Synthesis (HLS) reconfigurable systems, i.e. C-based and SystemC-based. Each input language has been used within the context of a separate HLS tool that is especially suitable for the particular input method chosen. The study has been based on the use of key fundamental computational and data processing algorithms, while the outlined observations have been drawn by traversing the full HLS flows. The algorithms address both memory- and compute-intensive modules, which are the main cores for numerous modern applications. In this way, detailed observations are made not only with respect to the performance of each approach individually but also against each other. Hence, this paper provides information on an extensive list of issues of major interest to modern reconfigurable systems design engineers, such as design cycle definition, time for HLS flow completion, implementable features, parallelisation, input model complexity and more importantly design effort/time. Moreover, this work presents detailed results on implementation issues as well as implementation guidelines concerning the presented schemes; these guidelines can certainly be used as a reference for any designer implementing such classes of algorithms.

Services

Statistics