Parallelizing the ZSWEEP Algorithm for Distributed-Shared Memory Architectures

Farias, Ricardo; Silva, Cláudio T.

doi:10.1007/978-3-7091-6756-4_12

Cited by 7 publications

(12 citation statements)

References 9 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Nieh and Levoy [7] developed a parallel volume rendering system. Their algorithm, however, was tested on a distributed-shared memory architecture, so did the PZSweep algorithm proposed by Farias et al [8]. Farias et al [9] soon enhanced the PZSweep routine [8] to configure it on a PC cluster, and strived to find a feasible load balancing strategy applicable for the new architecture.…”

Section: Related Workmentioning

confidence: 99%

“…Their algorithm, however, was tested on a distributed-shared memory architecture, so did the PZSweep algorithm proposed by Farias et al [8]. Farias et al [9] soon enhanced the PZSweep routine [8] to configure it on a PC cluster, and strived to find a feasible load balancing strategy applicable for the new architecture. Chen et al [10] presented a hybrid parallel rendering algorithm on SMP clusters, which make it easier to achieve efficient load balancing.…”

Section: Related Workmentioning

confidence: 99%

See 1 more Smart Citation

A Scalable Parallel Software Volume Rendering Algorithm for Large-Scale Unstructured Data

Wangc

Zheng

2007

Computational Science – ICCS 2007

View full text Add to dashboard Cite

Abstract. In this paper, we develop a highly accurate parallel software scanned cell projection algorithm (PSSCPA) which is applicable of any classification system. This algorithm could handle both convex and non-convex meshes, and provide maximum flexibilities in applicable types of cells. Compared with previous algorithms using 3D commodity graphics hardware, it introduces no the volume decomposition and rendering artifacts in the resulting images. Finally, high resolution images generated by the algorithm are provided, and the scalability of the algorithm is demonstrated on a PC Cluster with modest parallel resources.

show abstract

Section: Related Workmentioning

confidence: 99%

Section: Related Workmentioning

confidence: 99%

A Scalable Parallel Software Volume Rendering Algorithm for Large-Scale Unstructured Data

Wangc

Zheng

2007

Computational Science – ICCS 2007

View full text Add to dashboard Cite

show abstract

“…Farias, Mitchell and Silva [4] proposed a ZSWEEP algorithm to perform cell projection for the rendering of unstructured data sets. Later Farias and Silva reviewed the ZSWEEP algorithm, and proposed an image-based task partitioning scheme for the algorithm [5], making it suitable for distributed shared-memory machines. In this new scheme the screen was divided into several tiles, that could be dynamically sent to several processors, allowing for parallel processing.…”

Section: Related Workmentioning

confidence: 99%

GPU-Based Tiled Ray Casting Using Depth Peeling

Bernardon

Pagot

Comba

et al. 2006

Journal of Graphics Tools

Self Cite

View full text Add to dashboard Cite

Abstract:In this paper we propose several significant improvements to the hardware ray casting algorithm for unstructured meshes proposed byWeiler et al [14]. In their work, ray casting computation is entirely performed in the GPU by advancing intersections against the mesh while evaluating the volume rendering integral. Our contributions can be divided into three categories. First, we propose an alternate representation for mesh data in 2D textures that is more compact and efficient, compared to the 3D textures used in the original work. Second, we use a tile-based subdivision of the screen that allows computation to proceed only at places where it is required, thus reducing fragment processing in the GPU. Finally, we do not introduce imaginary cells that fill space caused by non-convexities of the mesh. Instead, we use a depth-peeling approach that captures when rays re-enter the mesh, which is much more general and does not require a convexification algorithm.We report results on an ATI 9700 Pro, the same hardware used by Weiler et al in their work. Due to the use of the 2D textures and the tiling, our technique is actually much faster than their work, while at the same time being more general, since it can render true non-convex meshes, as compared to their work, which is limited to convex (or convexified ) ones. On the Blunt Fin, our code renders between 400 Ktet/sec to 1.3 Mtet/sec. GPU-based Tiled AbstractIn this paper we propose several significant improvements to the hardware ray casting algorithm for unstructured meshes proposed by Weiler et al [14]. In their work, ray casting computation is entirely performed in the GPU by advancing intersections against the mesh while evaluating the volume rendering integral. Our contributions can be divided into three categories. First, we propose an alternate representation for mesh data in 2D textures that is more compact and efficient, compared to the 3D textures used in the original work. Second, we use a tile-based subdivision of the screen that allows computation to proceed only at places where it is required, thus reducing fragment processing in the GPU. Finally, we do not introduce imaginary cells that fill space caused by non-convexities of the mesh. Instead, we use a depth-peeling approach that captures when rays re-enter the mesh, which is much more general and does not require a convexification algorithm.We report results on an ATI 9700 Pro, the same hardware used by Weiler et al in their work. Due to the use of the 2D textures and the tiling, our technique is actually much faster than their work, while at the same time being more general, since it can render true non-convex meshes, as compared to their work, which is limited to convex (or convexified) ones. On the Blunt Fin, our code renders between 400 Ktet/sec to 1.3 Mtet/sec.

show abstract

“…We believe the limited image size is the main cause of the problem for the Delta dataset. We had similar poor results for our shared-memory version of ZSweep when the resolution was low (see [18]). More interestingly, and somewhat surprisingly, we see that reading data from local disks resulted in no better or sometimes even worse performance than reading data from the remote disk served by a single machine.…”

Section: Dataset Informationmentioning

confidence: 59%

“…Farias et al [16] developed ZSweep, also based on sweeping the data with a plane in the © direction. Very recently, Farias and Silva parallelized the ZSweep algorithm for distributed shared-memory architectures [18], and also developed an out-of-core version of ZSweep [17]. In [18] a tiling idea and an octree partition scheme were proposed; in [17] meta-cells and a similar tiling idea were employed.…”

Section: Volume Rendering For Unstructured Gridsmentioning

confidence: 99%

A unified infrastructure for parallel out-of-core isosurface extraction and volume rendering of unstructured grids

Chiang¹,

Farias²,

Silva³

et al.

Proceedings IEEE 2001 Symposium on Parallel and Large-Data Visualization and Graphics (Cat. No.01EX520)

View full text Add to dashboard Cite

In this paper, we present a unified infrastructure for parallel out-ofcore isosurface extraction and volume rendering of large unstructured grids on distributed-memory parallel machines. We parallelize the out-of-core isosurface extraction algorithm of [9] and the out-of-core ZSweep technique [17] for direct volume rendering, using the meta-cell technique as a unified underlying building block.Our one-time preprocessing first partitions the dataset into metacells that are stored in disk. From the meta-cells, we build a BBIO tree in disk, which can be used to speed up isosurface extraction, and a bounding-box file in disk, which is used for direct volume rendering. At run-time, we use a simple self-scheduling scheme [39] to achieve load balancing among the processors.We perform several experiments on a sixteen-node cluster of PCs connected by a gigabit ethernet, using datasets as large as 6.6 million cells. For the larger datasets, we have found that both our isosurface extraction and direct volume rendering approaches are perfectly scalable up to sixteen nodes.

show abstract

Parallelizing the ZSWEEP Algorithm for Distributed-Shared Memory Architectures

Cited by 7 publications

References 9 publications

A Scalable Parallel Software Volume Rendering Algorithm for Large-Scale Unstructured Data

A Scalable Parallel Software Volume Rendering Algorithm for Large-Scale Unstructured Data

GPU-Based Tiled Ray Casting Using Depth Peeling

A unified infrastructure for parallel out-of-core isosurface extraction and volume rendering of unstructured grids

Contact Info

Product

Resources

About