Adaptive load balancing for raycasting of non-uniformly bricked volumes

Müller, Christoph; Strengert, Magnus; Ertl, Thomas

doi:10.1016/j.parco.2006.12.002

Cited by 20 publications

(11 citation statements)

References 6 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…After trying different methodologies, including multi-layer compositing [see Beeson et al (2003)], using different dedicated compositing nodes, and even applying different compression algorithms, the direct send methodology was found to be the best from the perspective of the total rendering time and the cost of the hardware needed. The same conclusion was reached by Stuart et al (2010), Muller et al (2006), andMuller et al (2007).…”

Section: Communicationsupporting

confidence: 78%

“…Using a GPU-cluster to perform volume rendering using ray-casting performed in the fragment shader unit and volume bricking was addressed by Muller et al (2006Muller et al ( , 2007 . They investigated the effect of using empty-space skipping, static and dynamic loadbalancing approaches, and uniform and non-uniform bricking on the frame rendering time.…”

Section: Gpu-cluster Volume Renderingmentioning

confidence: 99%

See 1 more Smart Citation

A Distributed GPU-Based Framework for Real-Time 3D Volume Rendering of Large Astronomical Data Cubes

Hassan

Fluke

Barnes

2012

Publ. – Astron. Soc. Aust

View full text Add to dashboard Cite

We present a framework to volume-render three-dimensional data cubes interactively using distributed ray-casting and volume-bricking over a cluster of workstations powered by one or more graphics processing units (GPUs) and a multi-core central processing unit (CPU). The main design target for this framework is to provide an in-core visualization solution able to provide three-dimensional interactive views of terabyte-sized data cubes. We tested the presented framework using a computing cluster comprising 64 nodes with a total of 128 GPUs. The framework proved to be scalable to render a 204 GB data cube with an average of 30 frames per second. Our performance analyses also compare the use of NVIDIA Tesla 1060 and 2050 GPU architectures and the effect of increasing the visualization output resolution on the rendering performance. Although our initial focus, as shown in the examples presented in this work, is volume rendering of spectral data cubes from radio astronomy, we contend that our approach has applicability to other disciplines where close to real-time volume rendering of terabyte-order three-dimensional data sets is a requirement.

show abstract

Section: Communicationsupporting

confidence: 78%

Section: Gpu-cluster Volume Renderingmentioning

confidence: 99%

A Distributed GPU-Based Framework for Real-Time 3D Volume Rendering of Large Astronomical Data Cubes

Hassan

Fluke

Barnes

2012

Publ. – Astron. Soc. Aust

View full text Add to dashboard Cite

show abstract

“…However, in the literature, there are few studies addressing this issue on GPUs. The load imbalance issue of graphic problems was discussed in [10,18], and authors observed that it is of fundamental importance for high performance implementations on GPUs. Several static and dynamic load balancing strategies were evaluated for an octree partitioning problem on GPUs in [6].…”

Section: Related Workmentioning

confidence: 99%

Dynamic load balancing on single- and multi-GPU systems

Chen

Villa

Krishnamoorthy

et al. 2010

2010 IEEE International Symposium on Parallel &Amp; Distributed Processing (IPDPS)

106

View full text Add to dashboard Cite

The computational power provided by many-core graphics processing units (GPUs) has been exploited in many applications. The programming techniques currently employed on these GPUs are not sufficient to address problems exhibiting irregular, and unbalanced workload. The problem is exacerbated when trying to effectively exploit multiple GPUs concurrently, which are commonly available in many modern systems. In this paper, we propose a task-based dynamic load-balancing solution for singleand multi-GPU systems. The solution allows load balancing at a finer granularity than what is supported in current GPU programming APIs, such as NVIDIA's CUDA. We evaluate our approach using both micro-benchmarks and a molecular dynamics application that exhibits significant load imbalance. Experimental results with a single-GPU configuration show that our fine-grained task solution can utilize the hardware more efficiently than the CUDA scheduler for unbalanced workload. On multi-GPU systems, our solution achieves near-linear speedup, load balance, and significant performance improvement over techniques based on standard CUDA APIs.

show abstract

“…These problems may be categorized as static [1,2,3,4,5,6,7,8,9] and dynamic [10,11,12,13,14] types of task allocation. Some of the other related methods have been reported in the literature, such as, Integer Programming [15], Load Balancing [16,17,18,19,20,21,22,23], Divide and Conquer [24], Grid Computing [25] and Branch and Bound [26,27]. Tasks are allocated to various processors of the distributed environment in such a way that overall processing cost of the network should be minimized.…”

Section: Introductionmentioning

confidence: 99%

A Smart Algorithm for Dynamic Task Allocation for Distributed Processing Environment

Govil¹

2011

IJCA

View full text Add to dashboard Cite

A Distributed Processing Environment (DPE) consists of multiple autonomous computers that communicate through a communication media. In DPE a task is divided into many fractions and each of which is to be get processed. The task allocation problem can be explained in terms of number of tasks and number of processors available. In the present method propose a dynamic model for task allocation in DPE. Present method describes the allocation of m tasks in the environment of distributed processing with n processors (m>n) that completes in k number of phases. This method allocates the tasks to the processor to increases the performance of the DPE; and based on the inter task communication cost between executing task and another tasks. Residing cost and reallocation cost in various phases has also taken in consideration. On implemented the suggested algorithm we have obtained the phase wise optimal allocation and overall optimal cost. The run time complexity has been computed and compared with existing approaches. It is found that suggested algorithm is much better as compared to others.

show abstract

Adaptive load balancing for raycasting of non-uniformly bricked volumes

Cited by 20 publications

References 6 publications

A Distributed GPU-Based Framework for Real-Time 3D Volume Rendering of Large Astronomical Data Cubes

A Distributed GPU-Based Framework for Real-Time 3D Volume Rendering of Large Astronomical Data Cubes

Dynamic load balancing on single- and multi-GPU systems

A Smart Algorithm for Dynamic Task Allocation for Distributed Processing Environment

Contact Info

Product

Resources

About