Computation Enhancement using Reconfigurable Computing

In light of the industry’s constant need for better computer performance, this project aims to choose and evaluate an approach for facing this issue. The targeted category of computers is single board computers (e.g. Raspberry Pi). The approach utilized for enhancing performance is the use of reconfigurable computing as to execute computationally expensive calculations on a runtime custom-tailored hardware. The objective of this project is the test of the potential this approach has for increasing computers performance through comparing a software implementation of an algorithm with an FPGA assisted implementation of the same algorithm.
The platforms chosen for this project are the Rapsberry Pi and the Parallella P1602 board with its Zynq SoC for the software implementation and the FPGA assisted implementation in that order. The chosen algorithm is Fourier Fast Transform due to its part in many DSP applications and its suitability for the project objective. While the software solution worked successfully resulting in an asymptotic cost of O(N log N); the reconfigurable computing solution couldn’t be completed due to time constraints and lack of experience of the student. Future work should complete the experiment and add a multicore implementation of the same algorithm to add yet another class to the comparison.