Catálogo de publicaciones - libros

The analysis of long running and distributed applications poses a great challenge to software developers. PARADIS is a novel tool that helps the programmer with accomplishing this task. It reconstructs the corresponding event graph from events collected during a program run and provides techniques to address the problems arising from large traces. It offers several modules for specific examinations like the analysis of applications which process transactions and due to its modular architecture it allows an easy extension of the functionality. We show the usefulness on the basis of a real-life application and discuss future enhancements.

Parallel/Distributed programming is a complex task that requires a high degree of expertise to fulfill the expectations of high performance computation. On the one hand, application developers must tackle new programming paradigms, languages, libraries. On the other hand they must consider all the issues concerning application performance. On this context the Master/Worker paradigm appears as one of the most commonly used because it is quite easy to understand and there is a wide range of applications that match this paradigm. However, to reach high performance indeces it is necessary to tune the data distribution or the number of Workers considering the particular features of each run or even the actual behavior that can change dynamically during the execution. Dynamic tuning becomes a necessary and promising approach to reach the desired indeces. In this paper, we show the usage of a dynamic tuning environment that allows for adapting the data distribution applying Factoring algorithm on Master/Worker applications. The results show that such approach improves the execution time significantly when the application modifies its behavior during its execution.

With the increasing complexity of today’s systems, detailed performance analysis is more important than ever. We have developed DynTG, a tool for interactive, dynamic instrumentation. It uses performance module plugins to reconfigure the data acquisition and provides a source browser that allows users to insert any probe functionality provided by the modules dynamically into the target application. Any instrumentation can be added both before and during the application’s execution and the acquired data is presented in realtime within the source viewer. This enables users to monitor their applications’ progress and interactively control and adapt the instrumentation based on their observations.

Analyzing the performance of networks and messaging layers is important for diagnosing anomalous performance in parallel applications. However, general-purpose benchmarks rarely provide sufficient insight into any particular application’s behavior. What is needed is a facility for rapidly developing customized network performance tests that mimic an application’s use of the network but allow for easier experimentation to help determine performance bottlenecks.

In this paper, we contrast four approaches to developing customized network performance tests: straight C, C with a helper library, Python with a helper library, and a domain-specific language. We show that while a special-purpose library can result in significant improvements in functionality without sacrificing language familiarity, the key to facilitating rapid development of network performances tests is to use a domain-specific language designed expressly for that purpose.

While Java is getting an increasingly widely used programming language, Java Virtual Machine has become an important platform for networking and distributed computing. Due to the increasing complexity of programs, the demand for monitoring tool support (debuggers, performance analyzers etc.) for efficient computing is growing as well. A special, open interface J-OMIS, which provides an efficient support for monitoring distributed Java programs, is derived from the On-line Monitoring Interface Specification (OMIS) that enables to use multiple monitoring tools simultaneously. OMIS has been developed with interoperability in mind but not all structural and logical conflicts have been solved. Addressing the missing aspects of interoperability support within J-OMIS is intended to increase the simplicity of developing the monitoring tools which will synergetically support each other. In the paper we present the concept of JINEXT, an extension to OMIS, which is aimed to provide interoperability between monitoring tools.

The visualization tool is presented which aims at helping the programmer to find program transformations to improve temporal data locality. We present a number of locality metrics that capture the necessary information. Based on a cluster analysis of basic block vectors, the tool gives strong hints on which program transformations are needed. The visualizer allowed us to find the necessary transformations for three SPEC2000 programs in just a few minutes. After performing these transformations, the programs run 3 times faster on average on a number of different platforms.

Programmers usually rely on cache performance data to optimize their applications towards high runtime cache hit ratio. In this paper, we introduce a software toolset CacheIn, which uses simulation and monitoring to collect comprehensive cache performance data. CacheIn consists of a cache simulator for modeling the cache activities, a cache monitor for gathering different kind of information, and a multilayer software infrastructure for processing the raw monitoring data towards statistical, high-level representations, like histograms and summarized numbers. CacheIn exhibits both the details of traditional software mechanisms and the feasibility of performance counters. Based on a code instrumentor, we have verified CacheIn using standard benchmarks. Initial experimental results show its full functionality in terms of providing accurate, comprehensive, and corse-grained performance data.

Cache performance is regarded as a critical issue that significantly influences the overall performance of a computer system. Hence, optimization with respect to cache behavior forms an important topic in both research and commercial areas. This work aims at helping the user in the task of cache behavior analysis with a comprehensive visualization tool. The resulted cache visualizer is capable of presenting the various aspects of cache accesses, and shows the performance data in a high-level, user-readable way. More specially, it directly leads the user to the data structures where excessive cache misses occur. It then further depicts the reason of theses misses. In addition, it helps the user to decide important parameters concerning some optimization techniques, e.g. padding stride and size of cache blocks. Examples of initial optimization based on the visualizer have proven the feasibility of this visualization tool.

Wrinkles are important visual features on the skin of an animated character. However, realistic animation of wrinkles is not easy, especially for the hand, as it gathers many types of different wrinkles. In this paper, we present a computationally efficient method for the generation of hand wrinkles, both for the palm and the fingers. This method is developed by exploiting the inherent properties relating to wrinkles, which are volume preservation of the underlying flesh and incompressibility of the skin. Although this method is for the hand model, it is also applicable to other parts of the body where the skin is deformed by underlying skeleton joints.

This paper presents a muscle-driven wrinkle model for simulating dynamic wrinkles that appear during facial expressions, which is geometric in nature. Wrinkles are generated on an anatomy-based face model that incorporates a structure of skin, muscles and skull for physically-based animation. Corresponding to two types of facial muscles, a geometric model is developed to govern how the wrinkle amplitude evolves locally upon skin deformation, taking into account the properties of real wrinkles. During facial animation, wrinkles are generated in the local regions influenced by muscle contraction, simulating resistance to compression of tissues. Using our method, realistic wrinkles can be dynamically rendered on the adaptively refined face mesh at interactive rates.