Accelerating floating-point 3D graphics for vector microprocessors

摘要: We present performance results and discuss optimization techniques for floating-point 3D graphics primitives on the ARM vector (VFP) coprocessor. Much of work in is performed by a surprisingly small number primitives. Handheld devices cannot always afford power cost dedicated processors, so there considerable interest performing these handheld's own processor. In this paper, we examine pipeline, break out important primitives, then show how to vectorize optimize VFP.