Font Size: a A A

GPU Efficiency Based Graphic Engine Investigation And Implementation

Posted on:2014-04-30Degree:MasterType:Thesis
Country:ChinaCandidate:S Q ChenFull Text:PDF
GTID:2268330422454315Subject:Computer application technology
Abstract/Summary:PDF Full Text Request
With the rapid development of internet and mobile devices, the applicationsof3D digital mediaarebecomingmore and more popular in recent years.Rather thanthe beautiful and fantastic rendering images as before, people now care more andmore about the background technology for supporting digital media.3D digitalapplications are now really the hot spot in market. There is also lots of softwaredevelopers are now joining the tide of such abusiness.While developing excellent3Dapplications is usually not easy, one of the main reasons is the complexity of thesource code, and the performance is also a bottleneck.To resolve this problem, weneed a bridge-like component to handle the request between applications and systemresources more efficiently. Actually there have already been a lot of such componentsin the market, which are called as Graphic Engine. While sometimeseven with thegraphic engine, people still face to many kinds of problems, especially performanceissue, one of the mainreasons is that theenginesthemselves are not efficient enough, inanother world, they have to be improved significantly, especially more GPUresources should be used to speed up the algorithms.To solve the problems above,this thesis introduces how to implement a highefficient graphic engine. The efficiency includes: 1. In mesh component, this thesis introduces a new kind of mesh named asBatchMesh which will reduce GPU draw calls by merging meshes having thesame materials.2. In the texture component, we introduce a new concept named asTextureSetto merge textures for materials whose parameters are the same. This will alsohelp to reduce GPU draw calls3. In renderer module, the thesis shows you new algorithm to acceleraterendering-ObjectSorting. This sorting is composed of two aspects: y materialand the object distance tot camera position, it’s also for the purposeofreducing unnecessary GPU draw calls.4. Anew way of selection algorithm is introduced-GPU based selection, thisalgorithmis more efficient than traditional selection system.Lastly, experiments showthat graphic engine ismuch more efficientby using theabove optimizations.
Keywords/Search Tags:Graphic Engine, GPU Optimization, Batch, Texture Set, Sorting, GPU Selection
PDF Full Text Request
Related items