HLS enabled partially reconfigurable module implementation

Abstract

Making full use of the capabilities of the FPGA as an accelerator is difficult for non hardware experts, especially if partial reconfiguration is to be employed. One of the issues that arise is to physically implement modules into bounding boxes of minimum size for improving fragmentation cost and reconfiguration time. In this paper we present a method which automates the modules designing step, fulfilling module resource requirements and architectural FPGA constraints. We present a case study that shows how our automatic module implementation flow can be used to generate run-time reconfigurable bitstreams that are suited for stitching together processing pipelines directly from a Maxeler MaxJ HLS specification. This takes into consideration design alternatives, fragmentation, and routing failure mitigation strategies.