Interfacing hardware accelerators to a time-division multiplexing network-on-chip

This paper addresses the integration of stateless hardware accelerators into time-predictable multi-core platforms based on time-division multiplexing networks-on-chip. Stateless hardware accelerators, like floating-point units, are typically attached as co-processors to individual processors in the platform. Our design takes a different approach and connects the hardware accelerators to the network-on-chip in the same way as processor cores. Each processor that uses a hardware accelerator is assigned a virtual channel for sending instructions to the hardware accelerator and a virtual channel for receiving results. This allows a stateless and possibly pipelined hardware accelerator to be shared in an interleaved fashion without any form of reservation, and this opens for interesting area-performance trade-offs. The design is developed with a focus on time-predictability, area-efficiency, and FPGA implementation. The design evaluation is carried out using the open source T-CREST multi-core platform implemented on an Altera Cyclone IV FPGA. The size of the proposed design, including a floating-point accelerator, is about two-thirds of a processor.