EGPGV19: Eurographics Symposium on Parallel Graphics and Visualization

Permanent URI for this collection

https://diglib.eg.org/handle/10.2312/2632789

Browse

Now showing 1 - 13 of 13

PGV 2019: Frontmatter
(The Eurographics Association, 2019) Childs, Hank; Frey, Steffen; Childs, Hank and Frey, Steffen
Screen Partitioning Load Balancing for Parallel Rendering on a Multi-GPU Multi-Display Workstation
(The Eurographics Association, 2019) Dong, Yangzi; Peng, Chao; Childs, Hank and Frey, Steffen
Commodity workstations with multiple GPUs have been built by engineers and scientists for real-time rendering applications. As a result, a high display resolution can be achieved by connecting each GPU to a display monitor (resulting in a tiled large display). Using a multi-GPU workstation may not always produce a highly interactive rendering rate due to imbalanced rendering workloads among GPUs. In this work, we propose a parallel load balancing algorithm based on a screen partitioning strategy to dynamically balance the amount of vertices and triangles rendered by each GPU. Each GPU renders a screen region whose size may be different from the screen regions of other GPUs, but the amounts of vertices and triangles in those screen regions are balanced. It is possible that a screen region rendered by a GPU has to be displayed by another GPU. We propose a frame exchanging algorithm that allows GPUs to exchange screen regions efficiently. The inter-GPU communication overhead is very small since the data transferred between GPUs are a small amount of image pixels.
Task-based Augmented Reeb Graphs with Dynamic ST-Trees
(The Eurographics Association, 2019) Gueunet, Charles; Fortin, Pierre; Jomier, Julien; Tierny, Julien; Childs, Hank and Frey, Steffen
This paper presents, to the best of our knowledge, the first parallel algorithm for the computation of the augmented Reeb graph of piecewise linear scalar data. Such augmented Reeb graphs have a wide range of applications, including contour seeding and feature based segmentation. Our approach targets shared-memory multi-core workstations. For this, it completely revisits the optimal, but sequential, Reeb graph algorithm, which is capable of handing data in arbitrary dimension and with optimal time complexity. We take advantage of Fibonacci heaps to exploit the ST-Tree data structure through independent local propagations, while maintaining the optimal, linearithmic time complexity of the sequential reference algorithm. These independent propagations can be expressed using OpenMP tasks, hence benefiting in parallel from the dynamic load balancing of the task runtime while enabling us to increase the parallelism degree thanks to a dual sweep. We present performance results on triangulated surfaces and tetrahedral meshes. We provide comparisons to related work and show that our new algorithm results in superior time performance in practice, both in sequential and in parallel. An open-source C++ implementation is provided for reproducibility.
Statistical Analysis of Parallel Data Uploading using OpenGL
(The Eurographics Association, 2019) Wiedemann, Markus; Kranzlmüller, Dieter; Childs, Hank and Frey, Steffen
Modern real-time visualizations of large-scale datasets require constant high frame rates while their datasets might exceed the available graphics memory. This requires sophisticated upload strategies from host memory to the memory of the graphics cards. A possible solution uses outsourcing of all data uploads onto concurrent threads and disconnecting prohibitive data dependencies. OpenGL provides a variety of functions and parameters but not all allow minimal interference on rendering. In this work, we present a thorough and statistically sound analysis of various effects introduced by choosing different input parameters, such as size, partitioning and number of threads for uploading, as well as combinations of buffer usage hints and uploading functions. This approach provides insight into the problem and offers a basis for future optimizations.
Dynamic I/O Budget Reallocation For In Situ Wavelet Compression
(The Eurographics Association, 2019) Marsaglia, Nicole J.; Li, Shaomeng; Belcher, Kristi; Larsen, Matthew; Childs, Hank; Childs, Hank and Frey, Steffen
In situ wavelet compression is a potential solution for enabling post hoc visualization on supercomputers with slow I/O systems. While this in situ compression is typically accomplished by allocating an equal storage budget to each parallel process, we propose an adaptive approach. With our approach, we introduce an assessment step prior to compression, where each process characterizes the variation in its portion of the data, and then dynamically adapts storage budgets to the processes with the most variation. We conducted experiments comparing our adaptive approach with the traditional, non-adaptive approach, on two different simulation codes with concurrencies of 512 cores and mesh resolutions of one billion cells. Our findings show that our adaptive approach yields three orders of magnitude of improvement for one simulation and is not harmful for the other.
Hybrid Online Autotuning for Parallel Ray Tracing
(The Eurographics Association, 2019) Herveau, Killian; Pfaffe, Philip; Tillmann, Martin Peter; Tichy, Walter F.; Dachsbacher, Carsten; Childs, Hank and Frey, Steffen
Acceleration structures are key to high performance parallel ray tracing. Maximizing performance requires configuring the degrees of freedom (e.g., construction parameters) these data structures expose. Whether a parameter setting is optimal depends on the input (e.g., the scene and view parameters) and hardware. Manual selection is tedious, error prone, and is not portable. To automate the parameter selection task we use a hybrid of model-based prediction and online autotuning. The combination benefits from the best of both worlds: one-shot configuration selection when inputs are known or similar, effective exploration of the configuration space otherwise. Online tuning additionally serves to train the model on real inputs without requiring a-priori training samples. Online autotuning outperforms best-practice configurations recommended by the literature, by up to 11% median. The model predictions achieve 95% of the online autotuning performance while reducing 90% of the autotuner overhead. Hybrid online autotuning thus enables always-on tuning of parallel ray tracing.
An Interpolation Scheme for VDVP Lagrangian Basis Flows
(The Eurographics Association, 2019) Sane, Sudhanshu; Childs, Hank; Bujack, Roxana; Childs, Hank and Frey, Steffen
Using the Eulerian paradigm, accurate flow visualization of 3D time-varying data requires a high temporal resolution resulting in large storage requirements. The Lagrangian paradigm has proven to be a viable in situ-based approach to tackle this large data visualization problem. However, previous methods constrained the generation of Lagrangian basis flows to the special case of fixed duration and fixed placement (FDFP), in part because reconstructing the flow field using these basis flows is trivial. Our research relaxes this constraint, by considering the general case of variable duration and variable placement (VDVP) with the goal of increasing the amount of information per byte stored. That said, reconstructing the flow field using VDVP basis flows is non-trivial; the primary contribution of our work is a method we call VDVP-Interpolation which solves this problem. VDVP-Interpolation reduces error propagation and limits interpolation error while using VDVP Lagrangian basis flows. As a secondary contribution of the work, we generate VDVP basis flows for multiple data sets and demonstrate improved accuracy-storage propositions compared to previous work. In some cases, we demonstrate up to 40-60% more accurate pathline calculation while using 50% less data storage.
Parallel XPBD Simulation of Modified Morse Potential - an Alternative Spring Model
(The Eurographics Association, 2019) Cetinaslan, Ozan; Childs, Hank and Frey, Steffen
In this paper, we introduce a modified Morse potential as an alternative to the existing spring models within a massively parallel extended Position Based Dynamics (XPBD) algorithm. To date, stretching is one of the most popular constraint types of XPBD frameworks due to its simplicity, robustness and efficiency. However, the underneath mathematical expression of stretching constraint does not fully represent a spring model and behaves too stiff over a certain iteration count or damping coefficient. On the other hand, Hookean spring potential behaves softer and viscoelastic within the XPBD algorithm under the same conditions as stretching constraint. Our modified Morse potential addresses this issue by keeping the simulation of deformable models in between Hooke's law and stretching constraint. To demonstrate the benefits of modified Morse potential with higher frame rates, we develop an efficient Independent Edge Grouping algorithm for XPBD method which provides parallel processing on GPU. We compare the simulation results of cloth and volumetric models with stretching constraint, Hookean and St. Venant-Kirchhoff (STVK) spring potentials. We believe that our modified Morse potential is easy to implement and seamlessly fit into the existing XPBD frameworks.
Real-time Particle-based Snow Simulation on the GPU
(The Eurographics Association, 2019) Goswami, Prashant; Markowicz, Christian; Hassan, Ali; Childs, Hank and Frey, Steffen
This paper presents a novel real-time particle-based method for simulating snow on the GPU. Our method captures compression and bonding between snow particles, and incorporates the thermodynamics to model the realistic behavior of snow. The presented technique is computationally inexpensive, and is capable of supporting rendering in addition to physics simulation at high frame rates. The method is completely parallel and is implemented using CUDA. High efficiency and its simplicity makes our method an ideal candidate for integration in existing game SDK frameworks.
Hybrid Remote Visualization in Immersive Virtual Environments with Vistle
(The Eurographics Association, 2019) Aumüller, Martin; Childs, Hank and Frey, Steffen
Because of the spatial separation of high performance compute resources and immersive visualization systems, their combined use requires remote visualization. Remote rendering incurs increased latency from user interaction to display. For immersive virtual environments, this latency is a bigger problem than for desktop visualization. With hybrid remote visualization we enable the exploration of large-scale remote data sets from immersive virtual environments. This is based on three factors: When appropriate, we enable the local rendering of remote objects. We decouple local interaction from remote rendering as far as possible by depth compositing of remote and local images at a rate independent from remote rendering. Finally, we try to hide this latency by reprojecting 2.5D images for changed viewer positions. In this paper we describe the integration of hybrid remote rendering into the data-parallel visualization system Vistle as well its extension to a distributed system. Thereby arbitrary combinations of object-based and image-based remote visualization become possible.
Fast Mesh Validation in Combustion Simulations through In-Situ Visualization
(The Eurographics Association, 2019) Shudler, Sergei; Ferrier, Nicola; Insley, Joseph; Papka, Michael E.; Patel, Saumil; Rizzi, Silvio; Childs, Hank and Frey, Steffen
In-situ visualization and analysis is a powerful concept that aims to give users the ability to process data while it is still resident in memory, thereby vastly reducing the amount of data left for post-hoc analysis. The problem of having too much data for posthoc analysis is exacerbated in large-scale high-performance computing applications such as Nek5000, a massively-parallel CFD (Computational Fluid Dynamics) code used primarily for thermal hydraulics problems. Specifically, one problem users of Nek5000 often face is validating the mesh, that is identifying the exact location of problematic mesh elements within the whole mesh. Employing the standard post-hoc approach to address this problem is both time consuming and requires vast storage space. In this paper, we demonstrate how in-situ visualization, produced with SENSEI, a generic in-situ platform, helps users quickly validate the mesh. We also provide a bridge between Nek5000 and SENSEI that enables users to use any existing and future analysis routines in SENSEI. The approach is evaluated on a number of realistic datasets.
Efficient Point Merging Using Data Parallel Techniques
(The Eurographics Association, 2019) Yenpure, Abhishek; Childs, Hank; Moreland, Kenneth; Childs, Hank and Frey, Steffen
We study the problem of merging three-dimensional points that are nearby or coincident. We introduce a fast, efficient approach that uses data parallel techniques for execution in various shared-memory environments. Our technique incorporates a heuristic for efficiently clustering spatially close points together, which is one reason our method performs well against other methods. We then compare our approach against methods of a widely-used scientific visualization library accompanied by a performance study that shows our approach works well with different kinds of parallel hardware (many-core CPUs and NVIDIA GPUs) and data sets of various sizes.
Scalable Parallel Flow Visualization Using 3D Line Integral Convolution for Large Scale Unstructured Simulation Data
(The Eurographics Association, 2019) Liao, Yangguang; Matsui, Hiroaki; Kreylos, Oliver; Kellogg, Louise; Childs, Hank and Frey, Steffen
To address the need of highly efficient and scalable parallel flow visualization methods, we developed a flow visualization system for large unstructured simulation data using parallel 3D line integral convolution (LIC). The main consideration for a parallel LIC implementation is a trade-off between the additional memory cost of replicating cells at sub-domain boundaries, or the communication cost of exchanging those data among computation nodes. To improve scalability, we introduce a load-balancing scheme that partitions datasets based on estimated LIC computation time. We also introduce a data-driven sub-domain extension scheme that determines which external cells at sub-domain boundary need to be added based on current boundary cells, which reduces memory overhead because the same visual quality can be achieved with a significantly smaller number of replicated external cells. We evaluate our visualization method by first comparing its parallel scalability to traditional integral field lines methods. Next, we compare our cost-driven domain decomposition method to existing methods to verify that ours leads to more balanced computation and improved scalability. Finally, we compare our data-driven sub-domain expansion method to traditional layer-based expansion methods in terms of memory overhead and visual quality. We conclude that our parallel 3D LIC method is an efficient and scalable approach to visualization of large and complex 3D vector fields.

Browse

Browsing EGPGV19: Eurographics Symposium on Parallel Graphics and Visualization by Issue Date

Results Per Page

Sort Options