EGPGV15: Eurographics Symposium on Parallel Graphics and Visualization

Permanent URI for this collection

https://diglib.eg.org/handle/10.2312/12465

Browse

Browsing EGPGV15: Eurographics Symposium on Parallel Graphics and Visualization by Subject "Distributed/network graphics"

Now showing 1 - 2 of 2
  • Thumbnail Image
    Item
    Large-Scale Parallel Visualization of Particle-Based Simulations using Point Sprites and Level-Of-Detail
    (The Eurographics Association, 2015) Rizzi, Silvio; Hereld, Mark; Insley, Joseph; Papka, Michael E.; Uram, Thomas; Vishwanath, Venkatram; C. Dachsbacher and P. Navrátil
    Recent large-scale particle-based simulations are generating vast amounts of data posing a challenge to visualization algorithms. One possibility for addressing this challenge is to map particles into a regular grid for volume rendering, which carries the disadvantages of inefficient use of memory and undesired losses of dynamic range. As an alternative, we propose a method to efficiently visualize these massive particle datasets using point rendering techniques with neither loss of dynamic range nor memory overheads. In addition, a hierarchical reorganization of the data is desired to deliver meaningful visual representations of a large number of particles in a limited number of pixels, preserving point locality and also helping achieve interactive frame rates. In this paper, we present a framework for parallel rendering of large-scale particle data sets combining point sprites and z-ordering. The latter is used to create a multi level representation of the data which helps improving frame rates. Performance and scalability are evaluated on a GPU-based visualization cluster, scaling up to 128 GPUs. Results using particle datasets of up to 32 billion particles are shown.
  • Thumbnail Image
    Item
    TOD-Tree: Task-Overlapped Direct send Tree Image Compositing for Hybrid MPI Parallelism
    (The Eurographics Association, 2015) Grosset, A. V. Pascal; Prasad, Manasa; Christensen, Cameron; Knoll, Aaron; Hansen, Charles; C. Dachsbacher and P. Navrátil
    Modern supercomputers have very powerful multi-core CPUs. The programming model on these supercomputer is switching from pure MPI to MPI for inter-node communication, and shared memory and threads for intra-node communication. Consequently the bottleneck in most systems is no longer computation but communication between nodes. In this paper, we present a new compositing algorithm for hybrid MPI parallelism that focuses on communication avoidance and overlapping communication with computation at the expense of evenly balancing the workload. The algorithm has three stages: a direct send stage where nodes are arranged in groups and exchange regions of an image, followed by a tree compositing stage and a gather stage. We compare our algorithm with radix-k and binary-swap from the IceT library in a hybrid OpenMP/MPI setting, show strong scaling results and explain how we generally achieve better performance than these two algorithms.