Static Scheduling for Synthesis of DSP Algorithms on Various Models

Abstract:

Given a behavioral description of a DSP algorithm represented by a
data-flow graph, we show how to obtain a rate-optimal static schedule
with the minimum unfolding factor under two models, integral grid
model and fractional grid model, and two kinds of implementations for
each model, pipelined implementation and non-pipelined implementation.
We present a simple and unified approach to deal with the four
possible combinations. A unified polynomial-time scheduling algorithm
is presented, which works on the original data-flow graphs without
really unfolding. The values of the minimum rate-optimal unfolding
factors and the general properties for all the four combinations are
proved.