DICE, a Swedish game company, wanted a more realistic water simulation. Currently, most large scale water simulations used in games are based upon ocean simulation technology. These techniques falter when used in other scenarios, such as coastlines. In order to produce a more realistic simulation, a new one was created based upon the water simulation technique "Wavefront Parameter Interpolation". This technique involves a rather extensive preprocess that enables ocean simulations to have interactions with the terrain. This paper is about optimizing the current implementation of DICE's water simulation. The goal is to achieve better runtime GPU performance. After implementing various optimizations, a speedup of roughly 4-6x was achieved. Performance was evaluated on the PlayStation 4 gaming console. / DICE, ett svenskt spelföretag, ville ha en mer realistisk vattensimulering. Det flesta storskalna vattensimuleringar som används i spel idag är baserade på havsvattensimuleringstekniker. Dessa tekniker fungerar inte lika bra i andra scenarier, som t.ex. kustlinjer. För att kunna få en mer realistisk simulation, skapades en ny simulation baserad på vattensimuleringstekniken Wavefront Parameter Interpolation. Denna simuleringsteknik involverar en lång förprocess som ger havsvattensimuleringar möjligheten att interagera med terräng. Denna uppsats handlar om att optimera den nuvarande implementationen av DICEs vattensimulering. Målet är att få bättre grafikprestanda under körtid. Efter att olika optimiseringar hade implementerats, blev programmet 4-6x gånger snabbare. Prestandan utvärderades på spelkonsolen PlayStation 4.

This thesis provides an analysis of the application for evolutionary scheduling of collective communications. It proposes possible ways to accelerate the application using general purpose computing on graphics processing units (GPU). This work offers a theoretical overview of systems on a chip, collective communications scheduling and more detailed description of evolutionary algorithms. Further, the work provides a description of the GPU architecture and its memory hierarchy using the OpenCL memory model. Based on the profiling, the work defines a concept for parallel execution of the fitness function. Furthermore, an estimation of the possible level of acceleration is presented. The process of implementation is described with a closer insight into the optimization process. Another important point consists in comparison of the original CPU-based solution and the massively parallel GPU version. As the final point, the thesis proposes distribution of the computation among different devices supported by OpenCL standard. In the conclusion are discussed further advantages, constraints and possibilities of acceleration using distribution on heterogenous computing systems.