University of Twente Student Theses


Comparing selfh-healing techniques in approximate MAC accelerators

Raben, K.T. (2019) Comparing selfh-healing techniques in approximate MAC accelerators.

[img] PDF
Abstract:Approximate computing is the technique of trading in accuracy for efficiency in calculations. This efficiency can come in many forms such as reduction in used hardware area or reduced power consumption. To even further these methods, circuits can be designed that introduce self-healing, where some portion of the reduced accuracy can get internally compensated by letting different errors cancel each other out. One of these circuits is a Multiply Accumulate (MAC) circuit, where a multiplication is made and then added to the total result. This work focuses on comparing two different self-healing techniques that can reduce the cost of the multiplier in the MAC, one where a balancing multiplier is added and one where the multiplier is internally balanced. This work provides results on which technique is more effective in what situations executed on an FPGA. A model is made that evaluates different MAC configurations with parameters found by performing FPGA hardware simulations. From these evaluations the conclusion is drawn that a internal self-healing technique, where an approximate MAC is built with a multiplier that is internally balanced, is more effective.
Item Type:Essay (Bachelor)
Faculty:EEMCS: Electrical Engineering, Mathematics and Computer Science
Subject:53 electrotechnology, 54 computer science
Programme:Electrical Engineering BSc (56953)
Link to this item:
Export this item as:BibTeX
HTML Citation
Reference Manager


Repository Staff Only: item control page