Approximate Computing with FIR Accelerators
Type: Semester/Master project
Description:
Some applications not always require 100% accuracy, therefore it is possible to trade-off less power consumption for less accuracy. In particular, multipliers are widely used in computing systems and their power depends on the computed operands. This has been verified in reconfigurable FIR filters where the power largely depend on the configured coefficients. The goal of this project is to investigate how the power consumption changes in reconfigurable system based on the knowledge of the implemented multipliers.
Work distribution: 10% literature review, 30% HDL coding, 30 simulations, 30% analysis of power reports
Areas: approximate computing, FIR filters, multipliers, VLSI design, reconfigurable systems
Supervisor: Andrea Bonetti