Evaluation of High-Level Synthesis Techniques for Memory and Datapath Tradeoffs in FPGA Based SoC Architectures

Abstract

While High-Level Synthesis (HLS) has been a research topic for more than 30 years, modern HLS tools have made possible the development of production ready circuits, taking advantage of improvements in areas like Field Programmable Gate Arrays (FPGAs) and multicore System-on-Chip (SoC) architectures. The efficient use of these improvements however require technology aware coding and constraint specification, through specific tool directives. In this paper, a modern HLS environment is used to investigate different architecture alternatives, for the design of hardware accelerators, taking into account tradeoffs between memory and datapath component utilization, in modern, SoC aware FPGA devices. Experimental results show that correct architectural option selections can lead to an almost 120X speedup and better scaling performance while the input size is increasing, with practical negligible resource utilization overheads in medium and large scale devices.

Keywords

These keywords were added by machine and not by the authors. This process is experimental and the keywords may be updated as the learning algorithm improves.

This work was partially supported by “TEAChER: TEach AdvanCEd Reconfigurable architectures and tools” project funded by DAAD (2014) and CIDCIP and MENELAOS projects funded by the Ministry of Development under the National Strategic Reference Framework NSRF 2007-2013, action “Creation of innovation clusters” “A GREEK PRODUCT, A SINGLE MARKET: THE PLANET”.