Abstract
To effectively utilize an ever increasing number of processors during parallel rendering, hardware and software designers rely on sophisticated load balancing strategies. While dynamic load balancing is a powerful solution, it requires complex work distribution and synchronization mechanisms. Graphics hardware manufacturers have opted to employ static load balancing strategies instead. Specifically, triangle data is distributed to processors based on its overlap with screenspace tiles arranged in a fixed pattern. While the current strategy of using simple patterns for a small number of fast rasterizers achieves formidable performance, it is questionable how this approach will scale as the number of processors increases further. To address this issue, we analyze real-world rendering workloads, derive requirements for effective patterns, and present ten different pattern design strategies based on these requirements. In addition to a theoretical evaluation of these design strategies, we compare the performance of select patterns in a parallel sort-middle software rendering pipeline on an extensive set of triangle data captured from eight recent video games. As a result, we are able to identify a set of patterns that scale well and exhibit significantly improved performance over naïve approaches.
Originalsprache | englisch |
---|---|
Titel | Proceedings of High Performance Graphics, HPG 2017 |
Herausgeber (Verlag) | Association of Computing Machinery |
ISBN (elektronisch) | 9781450351010 |
DOIs | |
Publikationsstatus | Veröffentlicht - 28 Juli 2017 |
Veranstaltung | 9th Conference on High Performance Graphics: HPG 2017 - Los Angeles, USA / Vereinigte Staaten Dauer: 28 Juli 2017 → 30 Juli 2017 |
Konferenz
Konferenz | 9th Conference on High Performance Graphics |
---|---|
Land/Gebiet | USA / Vereinigte Staaten |
Ort | Los Angeles |
Zeitraum | 28/07/17 → 30/07/17 |
ASJC Scopus subject areas
- Computergrafik und computergestütztes Design
- Maschinelles Sehen und Mustererkennung
- Human-computer interaction