|
|
 |
|
Studienarbeit / Bachelor thesis
Rigid body dynamics: Performance optimization
Supervision:
Background:
The pe rigid body physics engine is a framework for the physically
correct simulation of rigid bodies of arbitrary shape.
A basic component of rigid body dynamic simulations are calculations with 3-dimensional
vectors and matrices for translations, transformations and rotations. A large part of
the necessary calculations for a single time step is spent with these operations.
In order to optimize the performance of the simulation (e.g. to make the framework
suitable for real-time appliciations) the optimization of those operations is of essential
importance. One possible optimization strategy is the exploitation of the target computer
architecture. Modern CPUs offer special purpose registers and calculations units like e.g.
the MMX and SSE extensions of the x86 processor family. Another possible optimization
approach is the application of assembler programming in performance critical parts of
the calculations.
The goal of this thesis is to exploit the hardware architectures as good as possible if
the calculations are performed on a suitable CPU. The task is to extend the mathematics
module of the pe physics engine by MMX and SSE intrinsics and/or
assembler instructions to increase the performance of all vector and matrix operations.
Tasks:
- Performance optimization of the mathematics module of the pe physics
engine
- Performance evaluation and measurement of the implementation
- Development of demonstration examples
Recommended knowledge:
Status:
Running
|
|
|
|
