T. A. Funkhouser. Coarse-grained parallelism for hierarchical radiosity using group iterative methods. In H. Rushmeier, editor, SIGGRAPH 96 Conference Proceedings, Annual Conference Series, pages 343--352. ACM SIGGRAPH, Addison Wesley, August 1996. held in New Orleans, Louisiana.  Home/Search   Document Details and Download   Summary   Related Articles   Check

....graphics pipeline [33] which insulates applications from the details of the underlying implementations of a common API. 2 Background and Related Work 2. 1 Cluster Graphics Clusters have long been used for parallelizing traditionally noninteractive graphics tasks such as ray tracing, radiosity [5, 25], and volume rendering [6] Other cluster parallel rendering efforts have largely concentrated on exploiting inter frame parallelism rather than trying to make each individual frame run faster [20] We are interested in enabling fast, interactive rendering on clusters, so these techniques tend to ....

Thomas Funkhouser. Coarse-grained parallelism for hierarchical radiosity using group iterative methods. Proceedings of SIGGRAPH 96, pages 343--352, August 1996.

....speed ups are obtained. iMAGIS is a joint research project of CNRS, INRIA, INPG and UJF. Different authors describe parallel implementations of hierarchical radiosity on clusters of workstations. Funkhouser describes an algorithm where multiple hierarchical radiosity solvers work in parallel [Fun96]. A master process distributes sets of polygons over the set of workstations to determine an approximate and partial radiosity solution. The master then collects and merges solutions. This mechanism is iterated until convergence. This approach is well adapted to very complex scenes which could not ....

.... calculations, the global nature of light propagation (where each object can potentially illuminate many other objects in different areas of the scene) makes it very difficult to organize and monitor data locality, unless a very strong spatial structure is present as in some architectural scenes [Fun96]. Distributed Shared Memory (DSM) systems therefore appear particularly suited to radiosity calculations. Even for the simpler shared memory case, however, a principal difficulty remains the segmentation of the work into tasks that can be efficiently distributed to the processors. A major factor ....

Thomas A. Funkhouser. Coarse-Grained Parallelism for Hierarchical Radiosity Using Group Iterative Methods. In Computer Graphics Proceedings, Annual Conference Series, 1996 (ACM SIGGRAPH '96 Proceedings), pages 343--352, 1996.

....energy group, rather than patch. This energy distributed the patches within group. Grouping patches reduces aliasing number iterations needed to converge a solution. In parallel radiosity implementations, the amount of sent each processors could greatly reduced energy is exchanged between groups [49]. In parallel rendering, both geometric simplification and grouping patches would pre processing algorithms prior rendering. Whenever a remote processor requests data, geometry sent processor could derived from distance in network. When processor receives a ray to intersect with locally stored ....

Funkhouser. Coarse-grained parallelism hierarchical radiosity using group iterative meth- ods. Rushmeier, editor, SIGGRAPH Con- ference Proceedings, Annual Conference Series, pages 343--352. ACM SIGGRAPH, Addison Wes-

....bene ts from the energy of all other patches in the group by multiple re ections. Consequently, it is important to take into account the level of interaction between the patches during group building. In order to improve the quality of groups, we took inspiration from the techniques of Funkhouser [Funkh96a] in which surfaces of the scene are subdivided in groups according to a form factors graph. In our method, groups are build dynamically during the resolution. One group is considered at a time so we have only one group to build from the graph. The form factors are already known, nevertheless ....

Funkhouser, T.A. : Coarse-Grained Parallelism for Hierarchical Radiosity Using Group Iterative Methods, Computer Graphics (SIGGRAPH'96 Proceedings), pp.343-352, 1996.

....texture caching in a similar manner, and extends these ideas to support efficient geometry caching. Algorithms that explicitly take advantage of the dense occlusion present in large architectural models have been used to compute radiosity solutions in scenes that would otherwise be intractable[20, 6]. These algorithms break the computation into nearly independent sub problems, based on sets of mutually interacting objects. Computation is reordered so that only a spatially local part of the data is processed at a time, and computation is scheduled based on which parts of the scene are already ....

Thomas A. Funkhouser. Coarse-grained parallelism for hierarchical radiosity using group iterative methods. In Holly Rushmeier, editor, SIGGRAPH 96 Conference Proceedings, pages 343--352. Addison Wesley, August 1996.

....static and dynamic, using INRIA Towards Efficient Parallel Radiosity for DSM based Parallel Computers . 17 new mapping of the sub environments on processors, or even new placements for virtual interfaces. One solution to be investigated is the use of graph partitioning techniques as proposed in [5]. For both techniques, a trade off between load balancing and data locality would have to be found. Our study focus on the traditional and well know progressive radiosity method, which has been replaced nowadays by hierarchical methods. Theses methods seem to be even more memory consuming than ....

Thomas A. Funkhouser. Coarse-Grained Parallelism for Hierarchical Radiosity Using Group Iterative Methods. In ACM SIGGRAPH '96 Proceedings, pages 343--352, New Orleans, August 1996.

.... the algorithm in many small tasks which required a redistribution and graph balancing technique [1] Funkhouser presented a parallel implementation for large scene databases including special memory management and an elaborate merging phase to correctly combine parts of the common solution [3]. Both approaches achieve impressive speed ups but suffer from their exclusive hardware requirements or complex implementation. We will now present a new distributed HR algorithm, that is simple and straightforward. Therefore it allows an easy integration into existing rendering packages but ....

Thomas A. Funkhouser. Coarse-Grained Parallelism for Hierarchical Radiosity Using Group Iterative Methods. In Computer Graphics Proceedings, Annual Conference Series, 1996 (ACM SIGGRAPH '96 Proceedings), pages 343--352, 1996.

....a polygon and some objects occluded by a large polygon like a wall. This problem has already been addressed in [2, 3, 4, 5] where the authors propose to partition the scene into subsets of polygons (called cells) before performing radiosity computations whether sequentially [1] or in parallel [6]. Another problem is due to the fact that the amount of data needed for radiosity computations (for very complex scenes) is so important that they cannot fit in memory. One solution to this problem is to partition the scene into cells so that only a subset of cells is maintained in memory for ....

....due to proximity and local intervisibility. From the experiments done by Teller, the fourth approach which exploits the result of the BSP partitioning method, seems giving the best results. A parallel algorithm for radiosity computation in large environments has been proposed by Funkhouser [6]. This algorithm makes use of the same datastructures as those proposed by Teller. The scene is organized in groups of clusters which are distributed by the host processor to the slave processors. Each slave processor is in charge of computing the radiosity solution for the group it is responsible ....

[Article contains additional citation context not shown here]

Thomas Funkhouser. Coarse-grained parallelism for hierarchical radiosity using group iterative methods. ACM SIGGRAPH'96 proceedings, pages 343--352, August 1996. RR n3149 16 Daniel Meneveaux, Kadi Bouatouch and Eric Maisel

....In fact, we achieved the same form factor computations either with regions and visibility masks or without them. In addition, we diffuse the energy in the same way but in different steps. But all these methods do not solve convergence detection rigorously (i.e. when to stop the algorithm) In [8], Funkhouser shows a proof of the convergence of parallel group iterative method by comparison to the standard sequential group Jacobi method. He uses separate instantiations of a hierarchical radiosity solver that can gather radiosity for patches in separate groups in parallel. Usually in ....

Thomas A. Funkhouser. Coarse-Grained Parallelism for Hierarchical Radiosity Using Group Iterative Methods. In Computer Graphics (ACM SIGGRAPH '96 Proceedings), pages 343--352, 1996.

....is a vast literature about parallel raytracing. The above proposal of a parallel raytracer is similar to [24] and to the decomposition approach in [20] Most parallel radiosity implementations treat the full matrix or the progressive algorithm. Only little work has been done on multi level methods [3, 7]. This paper abstracts from the specific rendering algorithm and provides a load balancer as a general tool for applications dealing with geometric objects. PQ = a local part of all initial tasks; while (PQ not empty) while (pending( receive(task) PQ task; endwhile perform first task of PQ ....

T. A. Funkhouser. Coarse-Grained Parallelism For Hierarchical Radiosity Using Group Iterative Methods. In SIGGRAPH '96, New Orleans, LA, August 1996.

....speedup the radiosity time since it still takes a few hours to complete all iterations for progressive radiosity and even if a user wants to view intermediate results, it takes minutes to see the obvious improvement from one view to another. There have been several parallel algorithms proposed in [BW90, BP94, PC94, RGG90, F96] but speedups for progressive radiosity with adaptive meshing are still low because the computation associated with patches in a 3D space is unstructured and the cost varies during iterations. Load balancing is the main challenge in achieving high speedups [PC94] The previous research [BP94, ....

....the radiosity algorithm, choosing a coarse grain parallelism would not allow full exploitation of data locality. Therefore, a tradeoff between selecting a grain size and exploiting the data locality has to be decided. Several parallel radiosity algorithms have been proposed in the literature, e.g. [BW90, BP94, PC94, RGG90, F96]. Baum et al. BW90] have obtained interesting results on a shared memory machine with 8 processors supplied with a hardware Z buffer; however, scalability results on a larger of numbers of processors have not been demonstrated. In [RGG90, F96] a master slave approach is proposed. At each ....

[Article contains additional citation context not shown here]

T.A.Funkhouser, "Coarse-Grained Parallelism for Hierarchical Radiosity Using Group Iterative Methods," Computer Graphics Proceedings, Annual Conference Series, 1996, pp.343352.

....divided into clusters, each assigned to a different processor. The communication pattern thus reflects the energy casting from one cluster to another, and seems suitable for implementation in a message passing parallel system. Another approach which also uses the clustering method is described in [6]. In this work the visibility is calculated offline. The program uses the master slave paradigm, so that the master assigns the clusters to the slaves for computation. Due to the synchrony of the communication that is used, the master slave paradigm limits the utilization that is obtained to 65 ....

T.A. Funkhouser. Coarse-Grained Parallelism for Hierarchical Radiosity using Group Iterative Methods. In ACM SIGGRAPH, pages 343--352, 1996.

....different levels of precision. An initial polygonal model which fits in main memory before the simulation may require several times more memory for the global illumination solution, making such a solution infeasible. Despite important efforts to reduce memory used by global illumination solutions [18, 5, 17], no overall framework has been proposed which allows the control of memory usage for both the element hierarchy and the links, and which maintains the advantage of the representation of global light transfer. The new approach we present here first develops a novel framework, based on the ....

....the treatment of large databases possible in the general case. In particular the largest model treated by a radiosity algorithm to date (to the authors knowledge) is in the order of 258,000 input polygons using hierarchical radiosity with clustering, coupled with a parallel and distributed system [5]. A completely different approach involves the use of stochastic methods. One recent example is [2] which also includes a thorough bibliography. Due to the fundamental differences compared to finite element approaches, we do not consider this direction further. The storage of links for ....

T. A. Funkhouser. Coarse-grained parallelism for hierarchical radiosity using group iterative methods. In Proc. of SIGGRAPH 96, pages 343--352, August 1996.

....parallel rendering algorithms. Rather, most prior cluster based polygon rendering systems have utilized only inter frame parallelism [13, 25] rendering separate frames of an image sequence on separate computers in parallel. Other parallel rendering algorithms, such as ray tracing, radiosity [11, 26], and volumetric rendering [12] have been implemented with PC clusters. However, they generally have not achieved fast, interactive frame rates (i.e. thirty frames per second) We are not aware of any prior system that has achieved scalable speedups via intra frame data parallelism while ....

Thomas A. Funkhouser. Coarse-grained parallelism for hierarchical radiosity using group iterative methods. In Computer Graphics (SIGGRAPH 96), 1996.

....rather than on intra frame parallelism, which is required to achieve scalable speedups in a real time system utilizing many processors. Networks of workstations have been used successfully for parallel graphics algorithms with coarse granularity, such as volume rendering [19, 16] radiosity [30, 14], and batch rendering of image sequences [17, 27] We are not aware of any prior system that has achieved scalable intra frame polygon rendering speedups via dynamic screen space partitions using a cluster of PCs. 3 Overview of Approach In multi projector systems, the most common approach to ....

T. Funkhouser, Coarse-Grained Parallelism for Hierarchical Radiosity Using Group Iterative Methods, Computer Graphics (SIGGRAPH 96), August 1996.

....rather than on intra frame parallelism, which is required to achieve scalable speedups in a real time system utilizing many processors. Networks of workstations have been used successfully for parallel graphics algorithms with coarse granularity, such as volume rendering [23, 16] radiosity [32, 13], and batch rendering of image sequences [17, 30] We are not aware of any prior system that has achieved scalable polygon rendering speedups via intra frame data parallelism across a network of workstations. 3 Problem Statement In our investigation, we aim to construct a sort first system ....

T. Funkhouser, Coarse-Grained Parallelism for Hierarchical Radiosity Using Group Iterative Methods, Computer Graphics (SIGGRAPH 96), August 1996.

No context found.

T. A. Funkhouser. Coarse-grained parallelism for hierarchical radiosity using group iterative methods. In H. Rushmeier, editor, SIGGRAPH 96 Conference Proceedings, Annual Conference Series, pages 343--352. ACM SIGGRAPH, Addison Wesley, August 1996. held in New Orleans, Louisiana.

No context found.

Funkhouser, T.: 1996, `Coarse-grained Parallelism for Hierarchical Radiosity using Group Iterative Methods'. In: ACM Computer Graphics (Proc. of SIGGRAPH'96). pp. 343--352.

No context found.

T. A. Funkhouser. Coarse-grained parallelism for hierarchical radiosity using group iterative methods. Computer Graphics, 30(4), 1996.

No context found.

T. A. Funkhouser. Coarse-grained parallelism for hierarchical radiosity using group iterative methods. In H. Rushmeier, editor, SIGGRAPH 96 Conference Proceedings, Annual Conference Series, pages 343--352. ACM SIGGRAPH, Addison Wesley, aug 1996. held in New Orleans, Louisiana, 04-09 August 1996.

Online articles have much greater impact   More about CiteSeer.IST   Add search form to your site   Submit documents   Feedback  

CiteSeer.IST - Copyright Penn State and NEC