GRAPHICS PIPELINE PERFORMANCE OPTIMIZATION

ruralrompΛογισμικό & κατασκευή λογ/κού

2 Δεκ 2013 (πριν από 3 χρόνια και 6 μήνες)

79 εμφανίσεις

GRAPHICS PIPELINE PERFORMANCE OPTIMIZATION


Sumantra Dasgupta


The present research is geared towards optimizing the performance of the graphics
pipeline.
The rendering pipeline in present day Computer Graphics is highly hardware
-
accelerated. Analysis of t
he performance characteristics of the graphics pipeline has
become extremely complicated due to the fact that not only is the performance CPU
dependent but also GPU dependent. The Graphics processing Unit has
actually become
the heart of the hardware
-
accel
erated graphics pipeline.

The present day computer is no
more a single processor unit as far as rendering and graphics are concerned. Improving
performance in such

a multiprocessor environment requires one

to determine bottleneck
s
and systematically elimi
nate

them. This process is repeated until the performance meets a
predefined standard.
The
present
research will endeavor to identify such bottlenecks

in
the GPU portion of the graphics pipeline

a
nd optimize them.

Wloka [1] discusses methodologies to be us
ed by game developers to better utilize the
GPU and offload the CPU in the process.
In t
his way
,

the CPU is free to do other routine
operations and the overall gaming experience improves
. The NVIDIA GPU
Programming Guide
found in [2]
discusses the various
practices to be employed
for
proper

uti
lization of

the GPU.

Since GPU processing is highly parallel, optimization
techniques used for DSPs [3] might be useful for optimizing GPUs. DirectX developers
discuss various issues relating to GPU optimization [4].

The aim of the present research
would be to

build a systematic approach for improving

GPU

performance. It will be
attempted to

build an optimization model for the whole process. The optimization
techniques will then be used to render a realistic nature s
cene which the user can
navigate. This should serve as a qualitative measure of the research.

The breakup of the project will be as follows:

1.

First Update: Identify the bottlenecks in the graphics pipeline arising inside the
GPU.

A thorough literature surve
y will also be done in the process.

2.

Second Update: Investigate into methods for obviating the bottlenecks.
Model the
optimization process (is any statistical model possible??).

3.

Final Report: Apply the optimization techniques to model a navigable realistic
looking nature scene. The realism of the scene and the quality of navigation will
give a qualitative measure of the research.


References:

1.

Wloka, Matthias, 2003 “Batch, Batch, Batch: What Does It Really Mean?”
Presentation at Game Developers Conference 200
3

2.

http://developer.nvidia.com

3.

http://www.ti.com

4.

http://msdn.microsoft.com/directx