Pipelining computing stages in configurable multicore architectures

Recently, there has been increasing interest on using task-level pipelining to accelerate the overall execution of applications mainly consisting of Producer-Consumer (P/C) tasks. In this PhD work we propose an approach to achieve pipelining execution of P/C pairs of tasks in FPGA-based multicore architectures. The current approach is able to speedup the overall execution of successive, data-dependent tasks, by using multiple cores and specific customization features provided by FPGAs. An important component of our approach is the use of customized inter-stage buffer schemes to communicate data and to synchronize the cores associated to the P/C tasks. To improve the performance, we propose a technique to optimize out-of-order communication between P/C pairs when the consumer requests more than once each data element produced, a behavior present in many applications (e.g., image processing). The current FPGA-based experimental results show the feasibility of our approach in both in-order and out-of-order P/C tasks. Moreover, the results using our approach to task-level pipelining and a multicore architecture reveal noticeable performance improvements for a number of benchmarks over a single core implementation without using task-level pipelining.