A 186Mvertices/s 161mW Floating-Point Vertex Processor for Mobile Graphics Systems

Abstract

In this paper, a power-efficient vertex processor with a geometry-specific arithmetic unit, vertex caches, and a vertex texturing unit is presented for mobile graphics environments. The arithmetic unit takes advantages of the geometry operations; a four-threaded and four-issue expanded VLIW datapath with a quad-float vertex texture fetcher. The vertex caches are optimized to acquire higher power efficiency. Moreover, an instruction-level power control method is adopted with an operand sharing and writeback re-allocation methods as well as operand isolations and gated clocks. The proposed vertex processor achieves 186 Mvertices/s of geometry performance which is 1.6 times faster than the previous results which adopt the IEEE754-compliant arithmetic units, and it supports OpenGL ES 2.0 and Vertex Shader Model 3.0. The processor is implemented in a 0.18-mum 1P4M CMOS process.