Using the pipeline framework, OpenGL implementations can be classified by the extent to which their OpenGL pipeline is implemented using specialized graphics hardware.
In graphics hardware architecture, a computer's general purpose
processor that drives the graphics hardware is called the host.
Deciding what stages in the graphics pipeline should be performed on
the host or in specialized graphics hardware is an important question
to answer when designing or evaluating a computer's graphics
subsystem. Often, tasks performed on the host are said to be
``implemented in software'' meaning no specialized hardware is used to
perform the stages. Tasks performed by specialized graphics hardware
are said to be ``implemented in hardware'' because dedicated hardware
is used to accelerate the tasks.
Table 1: Various graphics subsystems described using Akeley's
taxonomy for classifying graphics architectures.
Kurt Akeley, a coauthor of the OpenGL specification
[9], has described a useful taxonomy and associated
nomenclature for classifying graphics subsystems based on how the
general graphics pipeline stages are partitioned between software
implementation on the host and dedicated graphics hardware
[1]. The letters GTXRD abbreviate the general pipeline
stages. A hyphen is used to indicate what stages are (largely)
implemented in hardware. For example, a PC graphics board like the
standard VGA adapter would be an example of a GTXR-D architecture
since only the display stage is implemented in hardware. Table
1 describes various graphics subsystems using Akeley's
taxonomy. In practice, the taxonomy is a coarse one because there
is variability in the extent to which specific functionality is
implemented in hardware.