Graphics Reference
In-Depth Information
Computing
environment
Disturbance, d
Plant
(Rendering Process)
Input, u
Output, y
FIGURE 5.1
(See colour insert.) Rendering process from system perspective.
Computing
environment
Performance
objective, r
Error, e
Disturbance, d
u
Output, y
Plant
(Rendering Process)
Controller
FIGURE 5.2
(See colour insert.) Closed-loop feedback control system.
FigureĀ 5.2 depicts a feedback control system in which this controller is adopted to
regulate input to the plant to achieve a certain performance objective.
In this configuration, the rendering process produces a series of images, each in
a certain amount of time (frame time) and this plant output is compared with a pre-
defined performance metric through a data feedback channel for every cycle of a ren-
dered frame. To make the comparison of the output of the plant and the performance
objective useful, these two data streams must be expressed in the same unit of mea-
surement. Typically, the time taken to render one frame of image or its mathematical
inverse (frame rate) is the measurement unit. The error between the two quantities
is passed to the controller that subsequently generates a control action for the plant.
One interesting point is that the performance objective may be predefined by a
user or dynamically set by a more elaborate system that measures quality of service
(QoS) in the computing environment. Furthermore, this closed-loop feedback con-
trol system provides corrective action even when disturbance from the computing
environment occurs.
5.2.1
c ontRol s ystem a RchitectuRes foR R eal -t ime R endeRing
A prudent and imperative step in control system design is understanding the plant
characteristics to be controlled. Real world systems and processes seldom display
linear characteristics over their operating ranges because the physical nature of
materials used creates non-linearity in integrated systems.
Real-time rendering is also complex because of the numerous inputs and configu-
ration settings. A plant with varying dynamics would justify the use of an adaptive
controller to meet system performance objectives. In FigureĀ 5.3, a QoS component
 
Search WWH ::




Custom Search