HERESY: Mesh
Oriented 3D Graphics Architecture
Although triangle meshes are used pervasively in 3D graphics applications and there exist highly efficient mesh representations, almost all existing 3D graphics processors are based on the assumption that individual triangles are processed completely independently of one another. Consequently, none of these graphics processors are able to exploit triangle mesh's vertex/edge sharing property. In this project, we develop a mesh-oriented 3D graphics architecture called Heresy, which treats meshes as first-class objects, and significantly reduces the computation and communication costs of rendering triangle meshes by reusing intermediate computation results and eliminating data redundancy. The central architectural feature of Heresy is a highly efficient triangle mesh representation based on breadth-first mesh traversal, which is applied throughout the entire 3D graphics pipeline, from geometric transformation, clipping, to rasterization and thus enables aggressive exploitation of vertex/edge sharing to minimize both transformation/shading and rasterization cost.
3D Graphics
Workload Characterization
Although PC-class 3D graphics hardware has made significant strides in the last several years, the underlying architectural design principles are still generally considered as a black art. The quantitative approach prevalent in mainstream computer architecture design is rarely applied, at least as far as publicly available research literature is concerned. One main reason for this deficiency is the absence of a detailed workload characterization of 3D applications. The goal of this project is to collect the characteristics of dynamic and static 3D rendering workloads, perform detailed architectural-level analysis, and establish a 3D workload archive accessible over the Internet.
SUNDER: A NOW Based Parallel 3D Graphics Engine
Sunder is a parallel graphics engine based on a network of high-performance Pentium machines connected by a Gbit/sec system area network. Each of the Pentium machines is equipped with multiple low-cost 3D graphics cards. Sunder adopts the sort-first parallel rendering strategy to avoid the expensive communication cost associated with image-space compositing. Sunder also features the following architectural innovations: predictive mesh traversal, multi-resolution transformation computation based on MMX instruction sets, and adaptive image-space rendering using dynamic texture map. The first Sunder prototype will be a 4-node PentiumII-400MHz cluster with Myrinet interconnect and Diamond Monster 3D-II graphics cards.
Parallel Mesa Library for Large-Scale Message-Passing Systems
We have implemented a 3D graphics library that can be readily linked with parallel applications to provide run-time visualization on large-scale message-passing parallel machines, such as Intel Paragon. The prototype implementation is currently fully operational, and is based on a public-domain OpenGL implementation Mesa, and a sort-last parallelization strategy. Through a detailed performance analysis, we show that the scalability of the current prototype is close to the theoretical limit for the given hardware architecture. We also developed a unified framework to describe parallel compositing algorithms and show that two popular parallel compositing algorithms, binary swapping and parallel pipeline compositing, are just two extreme instances of this framework. Such a framework is important because it allows users to tailor the compositing algorithm according to the computation/communication characteristics of specific parallel machines by tuning the parameters appropriately. The current parallel Mesa library prototype implements such a parameterizable family of compositing algorithms.
Heresy: A Virtual Image-Space
3D Rasterization Architecture
Tzi-cker
Chiueh
12th ACM
SIGGRAPH/Eurographics Graphics Hardware Workshop, August 1997
Characterization of
Static 3D Graphics Workloads
(postscript
paper, postscript
talk slides)
Tzi-cker Chiueh and Wei-jen Lin
12th ACM SIGGRAPH/Eurographics Graphics Hardware Workshop, August 1997
A Breadth-First
Approach to Efficient Mesh Traversal
(postscipt
paper)
Tulika
Mitra and Tzi-cker Chiueh
13th ACM SIGGRAPH/Eurographics Graphics Hardware Workshop, August 1998
Implementation
and Evaluation for the Parallel Mesa Library
(postscript
paper)
Tulika Mitra and Tzi-cker Chiueh
IEEE International Conference on Parallel and Distributed Systems, December
1998
Dynamic 3D Graphics
Workload Characterization and the Architectural Implications
(postscript final version) ( pdf final version ) ( Extended Technical Report )
Tulika Mitra
and Tzi-cker Chiueh
32nd ACM/IEEE Annual International Symposium on Microarchitecture, November 1999
Mesh-Oriented
3D Graphics Architecture
(postscript
paper)
Tzi-cker Chiueh and
Tulika Mitra
ECSL Technical
Report (ECSL-TR-62)
PMesa(Parallel
Mesa Library for Large-Scale Message-Passing Systems)
Heresy:
Efficient Geometry Compression Software for Triangle Mesh
Download
the software from ECSL
archive
This page is maintained by Tulika Mitra (mitra@cs.sunysb.edu)