Graphics Reference
In-Depth Information
Chapter 16
Survey of Real-Time
Now that you've seen the core ideas that let us use computer graphics to make
pictures, we're going to describe to you the variety of approaches that have been
developed to encapsulate this knowledge. Such approaches have the benefit of
isolating the programmer from the details of the graphics hardware, which helps
with maintainability of programs. They also let application developers concentrate
on things specific to their application domain, rather than the way that images are
presented to the user. The variety of approaches available is due, in part, to the
pattern of hardware development, which is where we'll begin our survey.
As impressive as the rate of improvement in commodity CPUs has been dur-
ing the past four decades, the evolution of graphics hardware has been even
more remarkable. Hardware-based graphics acceleration—removing the burden
of executing the 3D pipeline from the primary processor by offloading it onto
a peripheral—was first commercialized for vector displays in the late 1960s
and for raster displays in the 1980s. Rendering pipelines moved from soft-
ware to raster graphics hardware via the development of geometry- and pixel-
processing chips that were integrated into graphics workstations built by high-end
real-time 3D graphics vendors such as SGI and Evans & Sutherland, and mid-
level-performance raster graphics was provided by workstation vendors including
Apollo and Sun Microsystems. These devices were expensive, affordable only
to academic and corporate institutions. The maturation of this technology into a
true commodity, available on personal computers, took place in the mid-1990s
in the form of graphics cards featuring cheap but powerful GPUs (described in
Chapter 38).
Since each brand/model of GPU has its own native instruction set and inter-
face, a standard API providing hardware independence is essential. The two dom-
inant APIs providing this important abstraction layer are Microsoft's proprietary
451
