Parallel rendering of volumetric dataset on distributed memory architectures

A parallel solution to the visualization and rendering of volume data is presented. Based on the ray tracing (RT) visualization technique, the system works on a distributed memory parallel architecture. The volumetric dataset is internally represented by the STICKS coding scheme, a data structure which exploits the one-dimensional coherence of the data, and is distributed between the nodes using a slice-partitioning technique. The similarity between the chosen data partition, the communication pattern of the visualization processes and the topology of the physical system architecture is one of the keypoints of the proposal and improves both software design and efficiency. The partitioning technique used and the network interconnection topology reduce the communication overhead and allow an efficient implementation of a static load balancing technique based on the pre-rendering of a low resolution image. Details related to the practical issues involved in the parallelization of volumetric RT are reported.