MANIAC

BDD Manipulation for Approximate Computing

The project MANIAC aims to optimise algorithms for approximate computing on the basis of binary decision diagrams. Approximate computing (AC) exploits the capability of many applications to tolerate a certain level of inexactness in computations. Because the result does not have to precise, a much better degree of efficiency can be reached. For example, when processing multi-media content, certain degrees of approximation are tolerated, as the human user does not notice minimal differences in hue or pitch. The approximate computing paradigm results in new requirements on the design process of hardware systems, which CPS scientists aim to meet by using binary decision diagrams (BDDs). BDDs are data structures to represent boolean functions, which are already being employed in hardware design – for example for logic synthesis or verification. In the project MANIAC, BDDs will allow an efficient representation of functions which in the long run will be tolerant against many kinds of deviations, and will thereby improve the design of new, powerful technologies.

Duration: Jan 1, 2016 - Dec 31, 2018
Application field: Hardware Systems   Software Systems  
Grant number: DR 287/29-1
Funding:

Publications within the Project

© DFKI GmbH
last updated 08/02/2017
to top