Wibheda+
Deshya Wijesundera, Alok Prakash, Thilina Perera, Kalindu Herath, Thambipillai Srikanthan · 2018
FPGA-based system-on-chip (SoC) devices for Internet of Things (IoT) applications require hardware-software (HW-SW) partitioning techniques to optimize for performance under stringent area and power constraints. To obtain an optimally partitioned design it is necessary to account for the data communication cost between hardware and software. However, the large design space during partitioning makes it challenging to account for this cost while optimizing for stringent constraints using an exhaustive approach. Hence, we propose Wibheda+, a heuristic based framework for fine-grained data dependency-aware multi-constrained HW-SW partitioning that can be employed to partition designs for FPGA-based SoCs used in IoT. Wibheda+, evaluated on 10 applications from the CHStone benchmark suite, has been shown to find solutions with 99% accuracy within several milliseconds compared to several minutes or hours taken in a state-of-the-art and an exhaustive approach, respectively.