Fast Propagation of Hamming and Signal Distances for Register-Transfer Level Datapaths

Reimer, Axel and Kosmann, Lars and Lorenz, Daniel and Nebel, Wolfgang
Hamming and signal distance are common model variables for characterising register-transfer level components for power. In order to use power estimation models based on this characterisation it is necessary to know exact input-stimuli for a circuit in order to propagate it through the datapath and calculate the Hamming and signal distancesat each node.In this work a new propagation approach for Hamming and signal distances is presented. It is based on precharacterised components for Hamming and signal distances. Using this approach it is not necessary to know the exact input data of a circuit, only statistical information of the input-stream is needed. The errors of the estimated Hamming and signal distance properties are in the range of 5% to 14% The result of an estimation is available nearly instantaneously since look-up tables are used for implementation.
09 / 2012
Methodik zum Entwurf von energiesparenden, verifizierten Systemen