Alberto José Proença

JaSkel: a Java skeleton-based framework for structured cluster and grid computing

This paper presents JaSkel, a skeleton-based framework to develop parallel and grid applications. The framework provides a set of Java abstract classes as a skeleton catalogue, which implements recurring parallel interaction paradigms. This approach aims to improve code efficiency and portability. It also helps to structure scalable applications through the refinement and composition of skeletons. Evaluation results show that using the provided skeletons do contribute to improve both application development time and execution performance.

Specification of industrial digital controllers with object-oriented Petri nets

The main purpose of this paper is to present an object-oriented PN model (shobi-PN) to specify industrial digital controllers. The shobi-PN model (synchronous, hierarchical, object-oriented and interpreted Petri net), was developed to support the use of hierarchy to model both the control unit and the plant of the systems. A CAD environment, SOFHIA, was developed to model digital controllers, to validate their properties and to simulate their behaviour. SOFHIA has an open architecture, which eases the integration of multiple code generator blocks to allow the implementation of the system in a wide range of technologies (hardwired, microprogrammed, programmed). Modelling of an industrial reactor control system is considered as a case study to illustrate the models applicability and capabilities.

Image-Based Empirical Information Acquisition, Scientific Reliability, and Long-Term Digital Preservation for the Natural Sciences and Cultural Heritage

The tools and standards of best practice adopted by natural science (NS) and cultural heritage (CH) professionals will determine the digital future of NS and CH digital imaging work. This tutorial discusses emerging digital technologies and explores issues influencing widespread adoption of digital practices for NS and CH. The tutorial explores a possible digital future for NS and CH through key concepts; adoption of digital surrogates, empirical (scientific) provenance, perpetual digital conservation, and ‘born archival’ semantic knowledge management. The tutorial discusses multiple image based technologies along with current research including; Reflectance Transformation Imaging (RTI), Photometric Stereo, and new research in the next generation of multi-view RTI. This research involves extending stereo correspondence methods. These technologies permit generation of digital surrogates that can serve as trusted representations of ‘real world’ content. The tutorial explores how empirical provenance contributes to the reliability of digital surrogates, and how perpetual digital conservation can ensure that digital surrogates will be archived and available for future generations. The tutorial investigates the role of semantically based knowledge management strategies and their use in simplifying ease of use by natural science and CH professionals as well as long term preservation activities. The tutorial also investigates these emerging technologies’ potential to democratize digital technology, making digital tools and methods easy to adopt and make NS and CH materials widely available to diverse audiences. The tutorial concludes with handson demonstrations of image-based capture and processing methods and a practical problem solving Q&A with the audience.

SOFHIA: a CAD environment to design digital control systems

Petri Nets (PNs) prove to be an efficient methodology to model discrete-event systems with parallel activities. The main advantages lie on the graphical interface and on the availability of a set of techniques for formal analysis, including the validation and the test of the modelled system. A proposal to modify the normal PN behaviour is presented, which aims a fast specification of synchronous parallel digital systems, including both the data path and the control unit. A CAD environment, SOFHIA, was developed to model digital systems, to validate their properties and to simulate their behaviour. The environment includes the automatic generation of VHDL code to allow simulation and synthesis on existing CAD tools.

Enabling JaSkel skeletons for clusters and computational Grids

JaSkel is a skeleton-based framework to develop efficient concurrent, parallel and Grid applications. It provides a set of Java abstract classes that implement recurring parallel interaction paradigms. The key feature of JaSkel is to use aspect-oriented external tools to address distributed execution, by injecting code to support communication middleware into JaSkel built-in skeleton implementations. This feature, when combined with the ability to develop nested skeletons, can help to tailor JaSkel applications to efficiently run on a grid of clusters systems, by taking advantage of inter/intra-cluster and/or intra-node communications. This paper describes the JaSkel distributed execution tools and how they interplay with the JaSkel framework to transparently run applications on a wide range of computing platforms, from multi-core to computational grids. Results are presented to show the feasibility and scalability of this approach.

Hierarchical mechanisms for high-level modeling and simulation of digital systems

The main purpose of this paper is to present a useful set of hierarchical mechanisms of specification to model and simulate digital systems. The shobi-PN model (Synchronous, Hierarchical, Object-Oriented and Interpreted Petri Net), was developed to support the use of hierarchy to model both the control unit and the data path of the systems. Modeling of a pipelined architecture unit of a microprocessor is considered as a case study to illustrate the capabilities of the hierarchical mechanisms proposed.

ParC++: a simple extension of C++ to parallel systems

The ParC++ system is an object oriented programming and execution environment, based on a programming model that aims the development of portable, efficient and scalable parallel applications. The paper presents the main components in the ParC++ system: a pre-processor-that supports a parallel extension to C++ objects and extracts parallelism from standard C++ code-and a run-time system-that dynamically adapts the grain size of parallel activities, according to the computing and communication system load.

A Generic and Highly Efficient Parallel Variant of Boruvka's Algorithm

This paper presents (i) a parallel, platform independent variant of Boruvkas algorithm, an efficient Minimum Spanning Tree (MST) solver, and (ii) a comprehensive comparison of MST-solver implementations, both on multi-core CPU-chips and GPUs. The core of our variant is an effective and explicit contraction of the graph. Our multi-core CPU implementation scales linearly up to 8 threads, whereas the GPU implementation performs considerably better than the optimal number of threads running on the CPU. We also show that our implementations outperform all other parallel MST-solver implementations in (ii), for a broad set of publicly available road network graphs.

Dynamic grain-size adaptation on object oriented parallel programming. The SCOOPP approach

This paper presents the SCOOPP (SCalable Object Oriented Parallel Programming) approach to support the design and execution of scalable parallel applications. The SCOOPP programming model aims the portability, dynamic scalability and efficiency of parallel applications. The SCOOPP is an hybrid compile and run-time system, which can perform parallelism extraction, supports explicit parallelism and performs dynamic granularity control at run-time. The mechanism that supports dynamic grain-size adaptation is presented and performance evaluated on two parallel systems. The measured results show the feasibility of the proposed dynamic grain-size adaptation and a scalability improvement of parallel applications over static parallel OO environments, which suggests cost benefits to develop scalable parallel applications to run on multiple platforms.

A Run-Time System for Dynamic Grain Packing

The SCOOPP (Scalable Object Oriented Parallel Programming) system is an hybrid compile and run-time system, that extracts parallelism, supports explicit parallelism and dynamically serialises parallel tasks in excess, to dynamically scale applications through a wide range of target platforms. This paper describes the run-time system of the current SCOOPP prototype - the ParC++ - and its mechanism to serialise parallelism. Low level performance results are presented, which indicate that the proposed methodology is effective and provides an high reduction in parallelism overheads. These features can improve the scalability of parallel applications with excessive parallelism.