Abstract
The deployment of efficient measurement solutions to assist network management tasks without interfering with normal network operation assumes a prominent role in today's high-speed networks attending to the huge amounts of traffic involved. From a myriad of proposals for traffic measurement, sampling techniques are particularly relevant contributing effectively for this purpose as only a subset of the overall traffic volume is handled for processing, preserving ideally the correct estimation of network statistical behavior. In this context, this paper proposes MuST - a multiadaptive sampling technique based on linear prediction, aiming at reducing significantly the measurement overhead and still assuring that traffic samples reflect the statistical characteristics of the global network traffic under analysis. Conversely to current sampling techniques, MuST is a multi and self-adaptive technique as both the sample size and interval between samples are self-adjustable parameters according to the ongoing network activity and the accuracy of prediction achieved. The tests carried out demonstrate that the proposed sampling technique is able to achieve accurate network estimations with reduced overhead, using throughput as reference parameter. The evaluation results, obtained resorting to real traffic traces representing wired and wireless aggregated traffic scenarios and actual network services, prove that the simplicity, flexibility and self-adaptability of the proposed technique can be successfully explored to improve network measurements efficiency over distinct traffic conditions. For optimization purposes, this paper also includes a study of the impact of varying the order of prediction, i.e., of considering different degrees of past memory in the self-adaptive estimation mechanism. The significance of the obtained results is demonstrated through statistical benchmarking.

Abstract

Abstract

Abstract
Clusters that combine heterogeneous compute device architectures, coupled with novel programming models, have created a true alternative to traditional (homogeneous) cluster computing, allowing to leverage the performance of parallel applications. In this paper we introduce clOpenCL, a platform that supports the simple deployment and efficient running of OpenCL-based parallel applications that may span several cluster nodes, expanding the original single-node OpenCL model. clOpenCL is deployed through user level services, thus allowing OpenCL applications from different users to share the same cluster nodes and their compute devices. Data exchanges between distributed clOpenCL components rely on Open-MX, a high-performance communication library. We also present extensive experimental data and key conditions that must be addressed when exploiting clOpenCL with real applications.

Abstract
High fidelity interactive rendering is of major importance for footwear designers, since it allows experimenting with virtual prototypes of new products, rather than producing expensive physical mock-ups. This requires capturing the appearance of complex materials using image based approaches, such as the Bidirectional Texture Function (BTF), to allow subsequent interactive visualization, while still maintaining the capability to edit the materials' appearance. However, interactive global illumination rendering of compressed editable BTFs with ordinary computing resources remains to be demonstrated. In this paper we demonstrate interactive global illumination by using a GPU ray tracing engine and the Sparse Parametric Mixture Model representation of BTFs, which is particularly well suited for BTF editing. We propose a rendering pipeline and data layout which allow for interactive frame rates and provide a scalability analysis with respect to the scene's complexity. We also include soft shadows from area light sources and approximate global illumination with ambient occlusion by resorting to progressive refinement, which quickly converges to a high quality image while maintaining interactive frame rates by limiting the number of rays shot per frame. Acceptable performance is also demonstrated under dynamic settings, including camera movements, changing lighting conditions and dynamic geometry.

Abstract
The Monte Carlo method has proved to be very powerful to cope with global illumination problems but it remains costly in terms of sampling operations. In various applications, previous work has shown that Bayesian Monte Carlo can significantly outperform importance sampling Monte Carlo thanks to a more effective use of the prior knowledge and of the information brought by the samples set. These good results have been confirmed in the context of global illumination but strictly limited to the perfect diffuse case. Our main goal in this paper is to propose a more general Bayesian Monte Carlo solution that allows dealing with nondiffuse BRDFs thanks to a spherical Gaussian-based framework. We also propose a fast hyperparameters determination method that avoids learning the hyperparameters for each BRDF. These contributions represent two major steps toward generalizing Bayesian Monte Carlo for global illumination rendering. We show that we achieve substantial quality improvements over importance sampling at comparable computational cost.