“An Extensible, Data-oriented Architecture for High-performance, Many-world Simulation” by Shacklett, Rosenzweig, Sarkar, Xie, Szot, et al. …
Conference:
Type(s):
Title:
- An Extensible, Data-oriented Architecture for High-performance, Many-world Simulation
Session/Category Title: Cloud Rendering: Your GPU Is Somewhere Else
Presenter(s)/Author(s):
- Brennan Shacklett
- Luc Guy Rosenzweig
- Bidipta Sarkar
- Zhiqiang Xie
- Andrew Szot
- Erik Wijmans
- Vladlen Koltun
- Dhruv Batra
- Kayvon Fatahalian
Moderator(s):
Abstract:
Training agents to learn skills in simulated environments is an increasingly important workload at the intersection of AI and graphics. The Madrona framework allows users to build high-performance simulators for novel tasks by leveraging entity-component-system (ECS) interfaces that map task logic onto parallel GPU execution across thousands of environments.
Additional Images:
![©Brennan Shacklett, Luc Guy Rosenzweig, Bidipta Sarkar, Zhiqiang Xie, Andrew Szot, Erik Wijmans, Vladlen Koltun, Dhruv Batra, and Kayvon Fatahalian](https://history.siggraph.org/wp-content/uploads/2024/02/2023-Tech-Papers-Shacklett_A-Programmable-Data-Oriented-Architecture-for-High-Performance-Many-World-Simulation.jpg)