MPEG Reconfigurable Video Coding - Signal Processing Systems - page 296

RVC Abstract Decoder Model

CAL FNL BSDL

Abstract
Decoder
Model

Scheduling
Analysis

SW code HW code

generator

Simulator

SDF,
CSDF,
BDF,
DDF

VHDL
Verilog

Ptolemy
II

Open DF

Moses C ARM

Non-normative tools and simulators for RVC

Fig. 22 Illustration of the RVC-C AL simulation and compiling tools

Fig. 23 Dataflow Models of
Computation

5.2 CAL Analysis

This section presents a taxonomy of Models of Computation (MoCs) (Fig. 23 ) that
can model the different types of behavior that RVC-C AL actors can exhibit. The
actors in the Video Tool Library of the RVC standard can behave statically, cyclo-
statically, quasi-statically, or dynamically. Different MoCs exist in the literature that
are suitable to model these types of behaviors.
Dataflow MoCs are defined as subsets of a general model called Dataflow Process
Networks (DPNs). The taxonomy shown on Fig. 23 reflects the fact that MoCs are
progressively restricted from DPN towards SDF with respect to expressiveness, but
at the same time they become more amenable to analysis. We first study the rules
of DPN, and then present the models that can be used to model static, cyclo-static,
quasi-static, and dynamic RVC-CAL actors.
Dataflow models respect the semantics of DPNs: A dataflow model is a directed
graph whose vertices are actors and edges are unidirectional FIFO channels with
unbounded capacity, connected between ports of actors.