Werner Benger , Markus Haider , Josef Stoeckl , Biagio Cosenza , Marcel Ritter , Dominik Steinhauser , Harald Hoeller Chapter in Astrophysics. InTech - Open Access Publisher year 2011, pages
Numerical simulations of astrophysical theories are an important complement to observational astronomy.
With the increasing computational powers available the demands on accuracy and capabilities of the simulations
grow steadily, leading to growing amounts of complex data sets which require three-dimensional visualization
for proper analysis. Modern graphics hardware provides the suitable rendering technology, however, many
visually impressive approaches used in computer games, for instance, do not transport over to application
upon scientific data sets since priorities are different.
Specific software for scientific visualization of astrophysical data does not necessarily need to provide
photo-realistic rendering, but should rather be able to comprehensively display all features of a data set.
False-color imaging has long been an established approach in observational astronomy,
with three-dimensional time-dependent data sets the optional parameter space is even larger.
As similar requirements occur in other scientific domains beyond astrophysics, data visualization
benefits extraordinarily well from interdisciplinary approaches,
for instance describing curved space in general relativity is similar to diffusion tensor fields
from magneto-resonance imaging in medical visualization.
Visualization of numerical astrophysics is also challenging because of the different data types involved together. For instance, typical visualization data set may involve both particle-based and volumetric data, thus requiring complex techniques to handle transparency.
Yet, interdisciplinary approaches face various hurdles, starting with incompatibilities
of the data format and data model. Common use is to provide specific
solutions to specific problems. Unifying approaches to cover various needs under one
denominator are rare, but essential in the longer term.
In this article we will review a systematic approach for modeling data sets
across scientific disciplines and for visualizing them, demonstrating this approach
upon a set of state-of-the art astrophysical simulations.
inproceedings
Visual Data Mining Using the Point Distribution Tensor
Marcel Ritter , Werner Benger , Biagio Cosenza , Keera Pullman , Hans Moritsch , Wolfgang Leimer IARIS Workshop on Computer Vision and Computer Graphics, VisGra 2012, Saint Gilles, Reunion Island, France year 2012, pages
Keywords: scientific visualization, data mining, metric tensor, point cloud, OpenCL
Abstract
We explore a novel algorithm to analyze arbitrary
distributions of 3D-points. Using a direct tensor field visualization technique allows to easily identify regions of linear, planar
or isotropic structure. This approach is very suitable for visual
data mining and exemplified upon geoscience applications.
It allows to distinguish, for example, power lines and flat
terrains in LIDAR scans. We furthermore present the work
on the optimization of the computationally intensive algorithm
using OpenCL and potentially utilizing the Insieme optimizing
compiler framework.
inproceedings
Distributed Load Balancing for Parallel Agent-Based Simulations
Biagio Cosenza , Gennaro Cordasco , Rosario De Chiara , Vittorio Scarano Proceedings of the 19th International Euromicro Conference on Parallel, Distributed and Network-based Processing, PDP 2011, Ayia Napa, Cyprus, 9-11 February 2011 year 2011, pages 62-69
We focus on agent-based simulations where a large number of agents move in the space, obeying to some
simple rules. Since such kind of simulations are computational intensive, it is challenging, for such a contest, to let the number
of agents to grow and to increase the quality of the simulation.
A fascinating way to answer to this need is by exploiting parallel architectures.
In this paper, we present a novel distributed load balancing schema for a parallel implementation of such simulations. The
purpose of such schema is to achieve an high scalability. Our approach to load balancing is designed to be lightweight and
totally distributed: the calculations for the balancing take place at each computational step, and in?uences the successive step.
To the best of our knowledge, our approach is the first distributed load balancing schema in this context.
We present both the design and the implementation that allowed us to perform a number of experiments, with up-to
1,000,000 agents. Tests show that, in spite of the fact that the load balancing algorithm is local, the workload distribution is
balanced while the communication overhead is negligible.
inproceedings
Load Balancing in Mesh-like Computations using Prediction Binary Trees
Biagio Cosenza , Gennaro Cordasco , Rosario De Chiara , Ugo Erra , Vittorio Scarano 7th International Symposium on Parallel and Distributed Computing (ISPDC) year 2008, pages 139-146
We present a load-balancing technique that exploits the temporal coherence, among successive computation phases, in mesh-like computations to be mapped on a cluster of processors. Our method partitions the computation in balanced tasks and distributes them to independent processors through the Prediction Binary Tree (PBT). At each new phase, current PBT is updated by using previous phase computing time (for each task) as (next phase) cost estimate. The PBT is designed so that it balances the load across the tasks as well as reduce dependency among processors for higher performances. Reducing dependency is obtained by using rectangular tiles of the mesh, of almost-square shape (i.e. one dimension is at most twice the other). By reducing dependency, one can reduce inter-processors communication or exploit local dependencies among tasks (such as data locality). Our strategy has been assessed on a significant problem, Parallel Ray Tracing. Our implementation shows a good scalability, and improves over coherence-oblivious implementations. We report different measurements showing that granularity of tasks is a key point for the performances of our decomposition/mapping strategy.
inproceedings
On Estimating the Effectiveness of Temporal and Spatial Coherence in Parallel Ray Tracing
Biagio Cosenza , Gennaro Cordasco , Rosario De Chiara , Ugo Erra , Vittorio Scarano Eurographics Italian Chapter Conference 2008, Salerno, Italy year 2008, pages 97-104
In this paper we estimate the effectiveness of exploiting coherence in Parallel Ray Tracing. We present a load-balancing technique which divides the original rendering problem in balanced subtasks and distribute them to independent processors through a Prediction Binary Tree (PBT). Furthermore the PBT allows to exploit temporal coherence among successive image frames. At each new frame, it updates the current PBT using a cost function which uses the previous rendering time as cost estimate. We also provide two heuristics which take advantage of data-locality. We assess the effectiveness of the proposed solution by running two experiments. The first one aims to investigate the accurancy of predictions made using the PBT. Results show that such predictions are quite accurate even considering a heavily unbalanced scene and a fast moving camera. The second experiment evaluates the two locality-aware heuristics showing a modest improvement.
inproceedings
A Survey on Exploiting Grids for Ray Tracing
Biagio Cosenza Eurographics Italian Chapter Conference 2008, Salerno, Italy year 2008, pages 89-96
Keywords: computer graphics, ray tracing, grid, graphics data structures
Abstract
Grid is one of the first data structure introduced at the very beginning of computer graphics. Grids are used in several applications of computer graphics, especially in rendering algorithms. Lately, in ray tracing dynamic scenes, grid has received attention for its appealing linear time building time. In this paper, we aim to survey several aspects behind the use of grids in ray tracing. In particular we investigate grid traversal algorithms, building techniques and several approaches for hierarchical grids.
article
Experiences with Mesh-like computations using Prediction Binary
Trees
Gennaro Cordasco , Biagio Cosenza , Rosario De Chiara , Ugo Erra , Vittorio Scarano Scalable Computing: Practice and Experience, Scientific international
journal for parallel and distributed computing (SCPE) year 2009, pages 173--187
In this paper we aim at exploiting the temporal coherence among successive
phases of a computation, in order to implement a load-balancing technique
in mesh-like computations to be mapped on a cluster of processors.
A key concept on which the load balancing schema is built on the
use of a Predictor component that is in charge of providing an estimation
of the unbalancing between successive phases.
By using this information, our method partitions the computation in
balanced tasks through the Prediction Binary Tree (PBT).
At each new phase, current PBT is updated by using previous phase
computing time (for each task) as (next phase) cost estimate. The
PBT is designed so that it balances the load across the tasks as
well as reduces {\em dependency} among processors for higher performances.
Reducing dependency is obtained by using rectangular tiles of the
mesh, of almost-square shape (i.e. one dimension is at most twice
the other). By reducing dependency, one can reduce inter-processors
communication or exploit local dependencies among tasks (such as
data locality). Furthermore, we also provide two heuristics which
take advantage of data-locality.
Our strategy has been assessed on a significant problem, Parallel
Ray Tracing. Our implementation shows a good scalability, and improves
performance in both cheaper commodity cluster and high performance
clusters with low latency networks.
We report different measurements showing that tasks granularity is
a key point for the performances of our decomposition/mapping strategy.
article
SambVca: A Web Application for the Calculation of the Buried Volume of N-Heterocyclic Carbene Ligands
Albert Poater , Biagio Cosenza , Andrea Correa , Simona Giudice , Francesco Ragone , Vittorio Scarano , Luigi Cavallo European Journal of Inorganic Chemistry year 2009, pages 1759-1766
Keywords: chemistry, inorganic chemistry, web application, buried volume
Abstract
We present a free web application for the calculation of the buried volume V-Bur) of NHC ligands. The web application provides a graphic and user-friendly interface to the SambVca program, developed for the calculation of % V-Bur values not only of NHC ligands but also of other classic organometallic ligands such as, for example, phosphanes and cyclopentadienyl-based ligands. To provide a reliable procedure for the calculation of % V-Bur values we tested our approach in the interpretation of the binding energies of NHC ligands in CpRu(NHC)Cl complexes in terms of steric and electronic parameters.
techreport
Synergy Effects of Hybrid CPU-GPU Architectures for Interactive Parallel Ray Tracing
Biagio Cosenza CINECA, ISBN 978-88-86037-23-5 year 2009, pages 61
Rendering algorithms range from o?-line, computationally costly, physicallybased, realistic techniques, to fast, approximate and plausible interactive
ones. Ray tracing is a popular rendering technique aiming for high realism,
and is a basis for global illumination algorithms. Interactive ray tracing has
seen enormous progress in recent years. Highly optimized packet-based ray
tracing implementations allow the computation of millions of ray-triangle intersection per second, and fully exploit modern multi-core CPUs, or GPUs.
However, complex scenes and lighting, and high-quality renderings with antialiasing are still not feasible at interactive speed, and only possible when
using compute clusters. In these scenarios, good load balancing is crucial in
order to exploit the computational power, and not to su?er from communication overhead and synchronization barriers. We introduce a Parallel Ray
Tracing architecture that exploits several parallelization techniques in order
to reach interactive performance. First, it exploits SIMD vector instructions
and more coherent memory accesses by tracing more rays at once (packet
of rays). Second, we take advantage of the multi core availability by using
multiple threads and scheduling packets of rays to di?erent threads. Finally,
we a?ord the distributed memory parallelization by using several techniques.
In particular, we introduce a method, which uses a cheap GPU rendering
technique to compute a cost map: An estimation of the per-pixel cost when
rendering the image using ray tracing. Using this information, we improve
load balancing, task scheduling, and work stealing strategies.
techreport
Evaluation of Adaptive Subdivision Schemas for Parallel Ray Tracing
Biagio Cosenza CINECA, ISBN 978-88-86037-22-8 year 2008, pages 206-216
Ray tracing algorithm is widely used for rendering images aiming at an high realism. Considerable efforts have been spent in order to investigate new ways to overcome the high computational demands of Ray Tracing. Speeding up ray tracing for interactive use with parallel architectures has received a big impulse during last years.
Despite several techniques are employed in order to amortize communication costs and manage load balancing, they still represents a bottleneck to the scalability.
The common approach in designing a Parallel Ray Tracer is screen-based, where each tile, a rectangular area of the screen space, represents a task sent to the worker node while the scene is replicated.
We propose a screen-based synchronous implementation where load balancing is afforded during the subdivision phase instead of during the assignment phase.
Our approach uses a predictive heuristic able to estimate the computating time need by a task. This estimate is then used in order to obtained a more balanced subdivision. A Prediction Binary Tree (PBT) is in charge of direct the subdivision of the whole job in balanced task.
The PBT stores the current subdivision of the whole image in a full binary tree, and it is updated by using a preditive heuristic.
This approach provides an adaptive subdivision of the whole problem in balanced task and is compatible with well-known load balancing strategies such as work stealing and work sharing, assuring higher scalability and performance.
Our parallel implementation is tailored for distributed memory architectures such as a cluster of workstations. Moreover, our work can be easily extended to shared memory architectures.
misc
Load Balancing Techniques for Parallel Ray Tracing
Biagio Cosenza Poster at HPC-Europa++ TAM-Workshop 2008, presented on 15-17/12/08 at HLRS Supercomputing Center, Universität Stuttgart year 2008, pages
