Overview
MADGRAF is investigating proprietary architectures which enable heterogeneous mobile hosts with varied resources (memory, CPU, power, etc) to render large geometric models with high polygon counts. The architectures explored will typically be designed around scalability and communication and involve some distribution of work over multiple commodity machines based on their resources. In addition to the architectural techniques, we will also look at integrating adaptive graphics pipeline techniques including culling, polygon simplification taking into account required level-of-detail and mesh compression (and encoding) techniques to account for low bandwidth (high error) links. Other techniques being explored includes intelligent caching. Finally prototypes of representative applications of this framework will be developed including visualization, location-aware applications and mobile gaming.
Publications
- Estimating Mobile Memory Requirements and Rendering Time for Remote Execution of the Graphics Pipeline
Kutty Banerjee, Fan Wu, Emmanuel Agu
in Proc. Eurographics 2005 (to appear)
- Remote Execution for 3D Graphics on Mobile Devices,
Kutty Banerjee, Emmanuel Agu,
in proc. IEEE WirelessCom Symposium on Mobile Computing, Maui, Hawaii, 2005 (to appear).
- PowerSpy: Fine-Grained Software Energy Profiling for Mobile Devices
Kutty Banerjee, Emmanuel Agu,
in Proc IEEE WirelessCom Symposium on Mobile Computing, Maui, Hawaii, 2005 (to appear).
- A Middleware Architecture for Mobile 3D Graphics”,
Emmanuel Agu, Kutty Banerjee, Shirish Nilekar, Oleg Rekutin, Diane Kramer
in Proc. 3rd International Workshop on Mobile Distributed Computing (MDC'05), to be held in conjuction with the 25th Int'l Conf. on Distributed Computing Systems (ICDCS'05), Columbus, Ohio (to appear)
Bibliography (limited access)
Links
Current MADGRAF Prototype Implementation
PowerSpy Presentation