The computational capacity of graphics cards for general-purpose computinghave progressed fast over the last decade. A major reason is computational heavycomputer games, where standard of performance and high quality graphics constantlyrise. Another reason is better suitable technologies for programming thegraphics cards. Combined, the product is high raw performance devices andmeans to access that performance. This thesis investigates some of the currenttechnologies for general-purpose computing on graphics processing units. Technologiesare primarily compared by means of benchmarking performance andsecondarily by factors concerning programming and implementation. The choiceof technology can have a large impact on performance. The benchmark applicationfound the difference in execution time of the fastest technology, CUDA, comparedto the slowest, OpenCL, to be twice a factor of two. The benchmark applicationalso found out that the older technologies, OpenGL and DirectX, are competitivewith CUDA and OpenCL in terms of resulting raw performance.

<p>An OpenGL applications usually renders to a single frame. Multi-view or 3D displays on the other hand, needs more more images representing different viewing directions on the same scene, but modifying a large number of applications would be unsuitable and problematic. However, intercepting and modifying these calls before they reach the GPU would dramatically decrease the amount of work needed to support a large number of applications on a new type of multi-view or 3D display. This thesis describes different ways on intercepting, enqueueing and replaying these calls to support rendering form different view points. Intercepting with both an own implementation of opengl32.dll and an OpenGL driver is discussed, and enqueueing using classes, function pointers and enumeration of functions is tried. The different techniques are discussed quickly with the focus being a working implementation. This resulting in an fully blown OpenGL interceptor with the ability to enqueue and replay a frame multiple times while modifying parameters such as the projection matrix. This implementation uses an own implementation of opengl32.dll that is placed in the application directory to be loaded before the real one. Enqueueing is performed by enumerating all OpenGL calls, pushing this enumeration value and all call data to a list Replaying is done by reading the same list and calling the function pointer associated with the enumeration value with the data in the list.</p>

In this physically-based tornado simulation, the tornado-scale approach techniques are applied to simulate the tornado formation environment. The three-dimensional Navier-Stokes equations for incompressible viscous fluid flows are used to model the tornado dynamics. The boundary conditions applied in this simulation lead to rotating and uplifting flow movement as found in real tornadoes and tornado research literatures. Moreover, a particle system is incorporated with the model equation solutions to model the irregular tornado shapes. Also, together with appropriate boundary conditions, varied particle control schemes produce tornadoes with different shapes. Furthermore, a modified metaball scheme is used to smooth the density distribution. Texture mapping, antialising, animation and volume rendering are applied to produce realistic visual results. The rendering algorithm is implemented in OpenGL.

In this physically-based tornado simulation, the tornado-scale approach techniques are applied to simulate the tornado formation environment. The three-dimensional Navier-Stokes equations for incompressible viscous fluid flows are used to model the tornado dynamics. The boundary conditions applied in this simulation lead to rotating and uplifting flow movement as found in real tornadoes and tornado research literatures. Moreover, a particle system is incorporated with the model equation solutions to model the irregular tornado shapes. Also, together with appropriate boundary conditions, varied particle control schemes produce tornadoes with different shapes. Furthermore, a modified metaball scheme is used to smooth the density distribution. Texture mapping, antialising, animation and volume rendering are applied to produce realistic visual results. The rendering algorithm is implemented in OpenGL.

¡@¡@Embedded systems often use several ways to emulate floating-point computation, due to the limitation of hardware resources and the performance/cost consideration. The first part of this thesis will discuss how systems without hardware support emulate floating point operation, and the performance difference between whether hardware floating-point units exists or not, and finally the performance evaluation between the usage of floating-point and fixed-point. In the mean time, we will also discuss the porting process of Nano-X fixed-point version to the Versatile PB/926EJ-S development board. ¡@¡@Due to the growing market demand and the big improvement of hardware, there are several 3D-display applications beginning to be presented to the public. In order to develop 3D programs, we need a standard API to reduce our development time. OpenGL~ES is a royalty-free, cross-platform API for full-function 2D and 3D graphics on embedded systems developed by The Khronos Group. The second part of this thesis will discuss the implementation of EGL---the interface between windowing system and the OpenGL~ES rendering API---and the GLUT (OpenGL Utility Toolkits)-like library, in order to make OpenGL programmers' life easier.

OpenGL ES 2.0 is the newest 3D graphics technology for hand-held devices established by Khronos. Users need a shading language compiler and a graphics card which is supportive for OpenGL ES 2.0 to develop their application on OpenGL ES 2.0. Without a graphcis processing unit and a corresponding compiler, one can not develop a 3D graphics application based on OpenGL ES 2.0. In order to solve these problems, we proposed an emulator for OpenGL ES 2.0 based on C-language compiler. The proposed emulator applies C-language compiler and CPU to fulfill the specification of OpenGL ES 2.0. With the proposed emulator, application developers can develop a 3D graphics application for OpenGL ES 2.0 without a specific hardware and a corresponding compiler and hardware designers also can compare and debug when designing their own graphics processing unit.

In computer graphics technique, the 3D graphic pipeline flow has two processing modules: Geometry module and Rendering module. The geometry module supports vertex coordinate transformation, vertex lighting computation, backface-culling, pre-clipping, and clipping functions. Clipping module clips the outside part of objects by view volume boundaries. Adding clipping module into geometry module will make 3D graphics pipeline flow more efficiency. Due to the sequential parsing nature of clipping, it causes the challenges to implement clipping function in hardware design. This paper implements a dual-path clipping engine placed after the Vertex Shader in geometry module and supports OpenGL-ES 2.0 specification. With the clipping engine, it reduces the unnecessary operations in 3D graphics pipeline flow and makes the performance efficient. The pipelined and shared hardware design is proposed to improve the area cost and throughput of the interpolation operation in clipping engine. The two vertices in/out clipping method is proposed in this paper. Users have more different choices of clipping algorithms for hardware implementation with respect to the performance and hardware limitation.

<p>Vizz3D is a 3D graphics code analysis tool, developed at Växjö University that optionally can use Java3D or OpenGL. However, initially Java3D was the only programming interface used. No other version was considered. Therefore the applications structure was built with the Java3D way of thought in mind. But code visualization with 3D graphics can be a demanding task for the computers processor and its graphics hardware and Java3D is known to be somewhat inefficient. So an OpenGL version was introduced.</p><p>This thesis reflects on the work restructuring the application’s code to fit both versions within Vizz3D in a structured and object-oriented way. The thesis shows the efforts to be taken to make an existing ever evolving tool easily extendible to other API’s. Additional aspects of OpenGL specific implementations are discussed throughout the thesis.</p>

Die vorliegende Arbeit beschreibt die Spezifikation und den Entwurf eines Systems der Virtuellen Realitaet. Das Verhalten der Objekte in diesem System basiert dabei auf ausgewählten physikalischen Gesetzmaessigkeiten. Es wurde eine kurze Analyse der am Lehrstuhl verfuegbaren Soft- und Hardware durchgefuehrt, um einen Ausgangspunkt fuer die weiteren Betrachtungen zu gewinnen. Bei Spezifikation und Entwurf des Basissystems wurde Wert auf ein strukturiertes Vorgehen, einen modularen Aufbau und eine gute Erweiterbarkeit für spaetere Arbeiten gelegt. Der praktische Teil der Arbeit realisiert die Implementation eines Experimentalsystems auf einem IBM-kompatiblen Computer unter dem Betriebssystem WindowsNT und der Entwicklungs- umgebung VisualC++ 4.0.

Fluid simulations are computationally intensive and therefore time consuming and expensive. In the field of visual effects, it is imperative that artists be able to efficiently move through iterations of the simulation to quickly converge on the desired result. One common fluid simulation technique is the Smoothed Particle Hydrodynamics (SPH) method. This method is highly parellelizable. I have implemented a method to integrate a Graphics Processor Unit (GPU) accelerated SPH capability into the 3D software package Houdini. This helps increase the speed with which artists are able to move through these iterations. This approach is extendable to allow future accelerations of the algorithm with new SPH techniques. Emphasis is placed on the infrastructure design so it can also serve as a guideline for both GPU programming and integrating custom code with Houdini.