Despite its proven usefulness for vector field visualization, LIC requires large amounts of processing power and memory. Parallel processing is an effective approach for computational intensive problems, however in the case of LIC it can suffer from workload imbalance depending on the distribution and orientation of the vectors elements in a given vector field. Specifically, elements of the boundary and singularity regions have the potential to cause workload imbalance, therefore affecting the parallel processing performance. In this paper, we propose and evaluate a coarse-grid pre-processing based load balancing method for minimizing this undesirable performance loss. By utilizing the same LIC algorithm in both pre-processing and final LIC computation stages, we can avoid redundant computations and minimize the impact of the pre-processing stage on the total LIC computation time. Since this method does not influence the original LIC algorithm, it is expected to work along with other enhancements proposed for LIC. We verified the effectiveness of the proposed method for maximizing the performance of parallel 3D LIC.
View full abstract