Using Visualization Tools to Understand Concurrency
IEEE Software
Interactive locality optimization on NUMA architectures
Proceedings of the 2003 ACM symposium on Software visualization
Runtime Visualisation of Object Oriented Software
VISSOFT '02 Proceedings of the 1st International Workshop on Visualizing Software for Understanding and Analysis
Advanced Global Illumination
Realistic Ray Tracing
Physically Based Rendering: From Theory to Implementation
Physically Based Rendering: From Theory to Implementation
Graph visualization for the analysis of the structure and dynamics of extreme-scale supercomputers
Information Visualization - Special issue: Software visualization
Adding parallelism to visual data flow programs
SoftVis '05 Proceedings of the 2005 ACM symposium on Software visualization
A visual approach to investigating shared and global memory behavior of CUDA kernels
EuroVis '13 Proceedings of the 15th Eurographics Conference on Visualization
| Hi-index | 0.02 |
We present a visualization system to assist designers of scheduling-based multi-threaded out-of-core algorithms. Our system facilitates the understanding and improving of the algorithm through a stack of visual widgets that effectively correlate the out-of-core system state with scheduling decisions. The stack presents an increasing refinement in the scope of both time and abstraction level; at the top of the stack, the evolution of a derived efficiency measure is shown for the scope of the entire out-of-core system execution and at the bottom the details of a single scheduling decision are displayed. The stack provides much more than a temporal zoom-effect as each widget presents a different view of the scheduling decision data, presenting distinct aspects of the out-of-core system state as well as correlating them with the neighboring widgets in the stack. This approach allows designers to to better understand and more effectively react to problems in scheduling or algorithm design. As a case study we consider a global illumination renderer and show how visualization of the scheduling behavior has led to key improvements of the renderer's performance.