RESEARCH OF NESTED PARALLEL PIPELINES ON PARALLEL GRAPHICS RENDERING SYSTEM

Abstract

Existing parallel graphics rendering systems only support single level parallel rendering pipeline. This paper presents a novel high performance parallel graphics rendering architecture on PC-Cluster supporting tiled display wall. It employs a hybrid sort-first and sort-last architecture based on a new rendering and scheduling structure, called dynamic rendering team (DRT as abbreviation), which is composed of multiple PCs instead of single PC to act as a rendering node. Each DRT responds for a certain projector area in the tiled display wall and all DRTs form a outer level parallel rendering pipeline natively. Inside separate DRT there is an optimized parallel Rendering-Composing-Display (R-C-D) pipeline reconstructing the serial workflow to parallel one. The Optimized parallel R-C-D pipeline along with the outer parallel rendering pipelines among DRTs forms a special nested parallel pipeline architecture, which promotes the whole rendering performance of our system greatly. Experiments show that parallel rendering system with the proposed architecture and nested parallel rendering pipelines, called Parallel-SG, can render over 13M triangles at an average speed of 12 frames per second without any accelerating technologies on a tile-display wall of 5*3 projectors array.