VTK-m: Accelerating the Visualization Toolkit for Massively Threaded Architectures

Moreland, Kenneth D.; Sewell, Christopher; Usher, Will; Lo, Li-Ta; Meredith, Jeremy; Pugmire, David; Kress, James; Schroots, Hendrik A.; Ma, Kwan‐Liu; Childs, Hank; Larsen, Matthew C.; Chen, Chun‐Ming; Maynard, Robert; Geveci, Berk

doi:10.1109/mcg.2016.48

Cited by 132 publications

(50 citation statements)

References 8 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Thus it cannot produce an augmented merge tree and consequently cannot support merge tree based data segmentation. Carr et al [50] presented a new algorithm available in the VTK-m library [57]. This approach is based on massive, fine-grained (one thread per input vertex), data parallelism and is specially designed for many-core architectures (like GPUs).…”

Section: A Related Workmentioning

confidence: 99%

Task-Based Augmented Contour Trees with Fibonacci Heaps

Gueunet

Fortin

Jomier³

et al. 2019

IEEE Trans. Parallel Distrib. Syst.

View full text Add to dashboard Cite

This paper presents a new algorithm for the fast, shared memory, multi-core computation of augmented contour trees on triangulations. In contrast to most existing parallel algorithms our technique computes augmented trees, enabling the full extent of contour tree based applications including data segmentation. Our approach completely revisits the traditional, sequential contour tree algorithm to re-formulate all the steps of the computation as a set of independent local tasks. This includes a new computation procedure based on Fibonacci heaps for the join and split trees, two intermediate data structures used to compute the contour tree, whose constructions are efficiently carried out concurrently thanks to the dynamic scheduling of task parallelism. We also introduce a new parallel algorithm for the combination of these two trees into the output global contour tree. Overall, this results in superior time performance in practice, both in sequential and in parallel thanks to the OpenMP task runtime. We report performance numbers that compare our approach to reference sequential and multi-threaded implementations for the computation of augmented merge and contour trees. These experiments demonstrate the run-time efficiency of our approach and its scalability on common workstations. We demonstrate the utility of our approach in data segmentation applications.

show abstract

Section: A Related Workmentioning

confidence: 99%

Task-Based Augmented Contour Trees with Fibonacci Heaps

Gueunet

Fortin

Jomier³

et al. 2019

IEEE Trans. Parallel Distrib. Syst.

View full text Add to dashboard Cite

show abstract

“…[9] presented a method for raytracing consisting entirely of data parallel operators such as map, gather, scatter, reduce, and scan, which are optimized for CPU and GPU. VTK-m library [12] employs Larsen's algorithm for the ray-tracer and supports in-situ visualization mode on various multithreaded devices such as CPU, GPU, and MIC. In our previous work [5], PBVR was implemented on GPU, and an order of magnitude speedup was achieved compared with CPU rendering.…”

Section: Related Workmentioning

confidence: 99%

“…Thirdly, the conventional volume rendering generates view-dependent images, and thus, interactive in-situ visualization becomes extremely costly. Existing parallel visualization libraries such as VTK-m [12], VisIt [1], and ParaView [2] support in-situ visualization based on the conventional volume rendering algorithms. However, the above bottlenecks were not resolved yet.…”

Section: Introductionmentioning

confidence: 99%

Performance Evaluation of Runtime Data Exploration Framework based on In-Situ Particle Based Volume Rendering

Kawamura

Noda

Idomura³

2017

JSFI

View full text Add to dashboard Cite

We examine the performance of the in-situ data exploration framework based on the in-situ Particle Based Volume Rendering (In-Situ PBVR) on the latest many-core platform. In-Situ PBVR converts extreme scale volume data into small rendering primitive particle data via parallel MonteCarlo sampling without costly visibility ordering. This feature avoids severe bottlenecks such as limited memory size per node and significant performance gap between computation and inter-node communication. In addition, remote in-situ data exploration is enabled by asynchronous file-based control sequences, which transfer the small particle data to client PCs, generate view-independent volume rendering images on client PCs, and change visualization parameters at runtime. In-Situ PBVR shows excellent strong scaling with low memory usage up to ∼ 100k cores on the Oakforest-PACS, which consists of 8,208 Intel Xeon Phi7250 (Knights Landing) processors. This performance is compatible with the remote in-situ data exploration capability.Keywords: in-situ visualization, volume rendering, runtime steering, strong scaling, performance evaluation. IntroductionRecent advances in HPC technology enabled extreme scale simulations at several tens of Peta-FLOPS, while the performance gap between computation and data I/O is enhanced. Because of this severe I/O bottleneck, data handling procedures such as "data output from simulations to storage" and "data input from storage to visualization applications" are becoming very costly, and conventional post processing visualization strategies often fail. In-situ visualization is one of promising solutions to this issue. In this approach, the I/O bottleneck is avoided by combining simulations and visualization applications to visualize simulation data at runtime on the same computing nodes. This requires extreme scale parallel visualization with high scalability at the same level as the simulations.Computational fluid dynamics (CFD) applications are widely used in various science and engineering fields, and are expected to be one of major applications on future exa-scale systems. Although variety of scientific visualization methods have been developed for CFD applications, visualization methods applicable to massively parallel processing of extreme scale volume data are still limited. Therefore, volume rendering methods for in-situ approaches should be carefully selected from the viewpoint of massively parallel processing and flexible visual analytics. Volume rendering is one of scalable visualization methods, which are suitable especially for CFD applications. In addition, in Ref. [4], it was shown that the volume rendering makes visual analytics flexible by extending the definition of transfer functions (TFs) into multi-dimension. Multi-dimensional transfer functions (MDTFs) generate not only three dimensional (3D) volume rendering, but also iso-surfaces, slice planes, image cropping, and image composition.Another important requirement is interactivity of in-situ visualization. In the conventional in-situ ...

show abstract

“…The need for portable physics codes that can run on GPUs and multi‐core CPUs has led to the development of libraries such as Raja and Kokkos, which also provide data‐parallel primitives such as for each, scan, and reduce. Several papers have examined the performance cost of using portable data‐parallel primitives by comparing algorithms, such as isosurfacing and ray tracing, implemented using PISTON, Thrust, or VTK‐m against native reference implementations for the GPU or multi‐core CPU . Due to the similarities between the Marching Cubes isosurface algorithm and interface reconstruction, the portability and performance trade‐offs demonstrated for that algorithm are particularly relevant to this work.…”

Section: Introductionmentioning

confidence: 99%

“…Several papers have examined the performance cost of using portable data-parallel primitives by comparing algorithms, such as isosurfacing and ray tracing, implemented using PISTON, Thrust, or VTK-m against native reference implementations for the GPU or multi-core CPU. 4,7,8 Due to the similarities between the Marching Cubes isosurface algorithm and interface reconstruction, the portability and performance trade-offs demonstrated for that algorithm are particularly relevant to this work.…”

mentioning

confidence: 99%

Portable data‐parallel surface reconstruction on a uniform rectilinear grid

François

Sewell

et al. 2017

Numerical Methods in Fluids

Self Cite

View full text Add to dashboard Cite

Summary With the increasing heterogeneity and on‐node parallelism of high‐performance computing hardware, a major challenge is to develop portable and efficient algorithms and software. In this work, we present our implementation of a portable code to perform surface reconstruction using NVIDIA's Thrust library. Surface reconstruction is a technique commonly used in volume tracking methods for simulations of multimaterial flow with interfaces. We have designed a 3D mesh data structure that is easily mapped to the 1D vectors used by Thrust and at the same time is simple to use and uses familiar data structure terminology (such as cells, faces, vertices, and edges). With this new data structure in place, we have implemented a piecewise linear interface reconstruction algorithm in 3 dimensions that effectively exploits the symmetry present in a uniform rectilinear computational cell. Finally, we report performance results, which show that a single implementation of these algorithms can be compiled to multiple backends (specifically, multi‐core CPUs, NVIDIA GPUs, and Intel Xeon Phi processors), making efficient use of the available parallelism on each. We also compare performance of our implementation to a legacy FORTRAN implementation in Message Passing Interface (MPI) and show performance parity on single and multi‐core CPU and achieved good parallel speed‐ups on GPU. Our research demonstrates the advantage of performance portability of the underlying data‐parallel programming model.

show abstract

VTK-m: Accelerating the Visualization Toolkit for Massively Threaded Architectures

Cited by 132 publications

References 8 publications

Task-Based Augmented Contour Trees with Fibonacci Heaps

Task-Based Augmented Contour Trees with Fibonacci Heaps

Performance Evaluation of Runtime Data Exploration Framework based on In-Situ Particle Based Volume Rendering

Portable data‐parallel surface reconstruction on a uniform rectilinear grid

Contact Info

Product

Resources

About