The Styria Flyover - LOD management for huge textured terrain model

Michael Kofler, Michael Gervautz, Michael Gruber
In IEEE Proceedings of Computer Graphics International '98, Hannover, Germany, June 1998.

Paper available (in PDF format).

Abstract

With the Styria Flyover you can virtually fly over Styria, a province of Austria. The countryside is characterized by a mixture of rather flat regions in the south east and mountains up to 3000 m in the north west. The digital model covers Styria and parts of its adjoining provinces (30.000 km2); it is based on a DTM consisting of more than 23 million triangles and on 50 Mbyte texture data (satellite images). The framework combines an R-tree data structure with efficient LoD (Level of Details) management (automatic mesh refinement), progressive rendering and dynamic performance adjustment. On a SGI O² it delivers more than 8 frames per second. Our approach demonstrates that efficient algorithms make a fast 3D GIS user interface possible without spending a fortune for hardware. The paper focuses on techniques and data structures to maximize the performance of visualization.

Notes

R-trees and resolution levels

Concept of perspective query at multiple LoD.

Multi-level Texture Caching for 3D Graphics Hardware

Michael Cox¹, Narendra Bhandari² and Michael Shantz²
¹ MRJ/NASA Ames Research Center and Intel Microcomputer Research Labs
and ² Intel Microcomputer Research Labs
Proc. of 25th International Symposium on Computer Architecture, pp.86-97
IEEE Computer Society, Barcelona, Spain, June 27 - July 1, 1998.

Paper available (in PDF format).

Notes

Push architecture :

Large fast memories are co-located with the acceleration hardware so that when pixels from the texture (texels) are required, they can be retrieved with low latency and high bandwidth.

Advantage:

Since local memory is dedicated to graphics, the memory subsystem can be designed to provide exactly the bandwith required.

Disadvantages:

Demand for textures and resolutions keep increasing and demands are usually higher than what manufacturers can provide.
If more texture is required than what texture memory can accomodate, one has to write a segment manager (software) dedicated to the specific machine/architecture.
Usually requires to download entire texture maps to the accelerator (inneficient in most cases).

Pull architecture :

Textures are stored in system memory and the accelerator pulls texels from system memory to accelerator on-chip[ memory as required. There is no local external memory for texture.

Advantages:

Places texture in system memory, relaxing capacity constraints and thereby curbing the growth of local memory associated with the accelerator.
Frees the programmer from resolving a bin-packing problem of texture memory management.
Only required (blocks of) texels are downloaded by the accelerator.

Disadvantage:

Forces system memory to be designed to support bandwidth required by texture mapping.

Multi-Level Texture Tiling

L3: identification number/label for any given texture (map).

L2: integer identifying a tile or block of texels.

L1: integer, associated to a parent block L2, denoting a sub-tile (or sub-block) of texels.

Page created & maintained by Frederic Leymarie, 1998.
Comments, suggestions, etc., mail to: leymarie@lems.brown.edu