Exploiting the Reconfigurability of ρ-VEX Processor for Real-Time Robotic Applications

Abstract

Autonomous mobile robots generally have limited computational power on-board, and they have to perform their tasks in real-time in order to interact with their surroundings effectively. Therefore, there is a need to utilize the available computational capabilities efficiently. The ρ-VEX is a run-time reconfigurable VLIW processor. This unique processor allows separation of its issue lanes to form independently operating processing cores. Switching between these configurations during run-time allows optimizing the computing resources for the task(s) it is performing. In this project FreeRTOS is ported to the ρ-VEX processor and a control layer is developed. FreeRTOS manages the applications based on given real time parameters. The control layer decides the number of active cores (hardware contexts) and issue width of each core to best match the processing requirements of the applications. In this way, FreeRTOS and the control layer together can reconfigure the number of active cores at run-time. This is a very unique feature of this thesis project and can not be found in any other multicore implementation of FreeRTOS. The control layer along with FreeRTOS provides the user a facility to run applications under real-time constraints and with the best possible efficiency. In order to evaluate the performance, the overhead of the FreeRTOS is quantified and a performance comparison is made between several configurations of this system. Moreover, impact of reconfigurability of the ρ-VEX on the schedulability of real-time applications is measured and it is concluded that (indeed) reconfigurability of ρ-VEX improves the schedulability of the real-time applications