Self-Adapting Linear Algebra Algorithms and Software

One of the main obstacles to the efficient solution of scientific problems is the problem of tuning software, both to the available architecture and to the user problem at hand. We describe approaches for obtaining tuned high-performance kernels and for automatically choosing suitable algorithms. Specifically, we describe the generation of dense and sparse Basic Linear Algebra Subprograms (BLAS) kernels, and the selection of linear solver algorithms. However, the ideas presented here extend beyond these areas, which can be considered proof of concept.     

Multiobjective evaluation of advanced manufacturing system technology investments with risk

Advanced manufacturing system technologies are difficult to evaluate using ordinary financial measures owing to the presence of risk and significant intangible factors. In this paper a methodology is proposed that quantifies financial, strategic, and tactical attributes of each alternative. Attribute values are denned for each period of project life. Both project life and attributes are treated as random variables. Attributes are discounted to the current time and the multiple objectives are combined by composite programming. The distribution of the combined objective is estimated. The procedure is demonstrated on an actual case study.     

Wideband direction-of-arrival estimation using frequency-domain frequency-invariant beamformers: an analysis of performance

This paper presents an extension of the wideband direction-of-arrival estimation method using frequency-domain frequency-invariant beamformers (Do-Hong et al., 2003) with an analysis of performance. An asymptotic expression for the root-mean-square error and a resolution threshold of signal-to-noise-ratio for resolving directions of sources are shown. A comparison of analysis and simulation as well as numerical results are also presented.     

An intervalal gorithm for solving systems of linear equations to prespecified accuracy

We describe an interval arithmetic algorithm for solving a special class of simultaneous linear equations. This class includes but is not limited to systemsAx=b whereA andb have integer entries. The algorithm uses fixed point arithmetic, and has two properties which distinguish it from earlier algorithms: given the absolute accuracy ε desired, the algorithm uses only as much precision as needed to achieve it, and the algorithm can adjust its own parameters to minimize computation time.     

When cache blocking of sparse matrix vector multiply works and why

We present new performance models and more compact data structures for cache blocking when applied to sparse matrix-vector multiply (SpM × V). We extend our prior models by relaxing the assumption that the vectors fit in cache and find that the new models are accurate enough to predict optimum block sizes. In addition, we determine criteria that predict when cache blocking improves performance. We conclude with architectural suggestions that would make memory systems execute SpM × V faster.     

Models and Scheduling Algorithms for Mixed Data and Task Parallel Programs

An increasing number of scientific programs exhibit two forms of parallelism, often in a nested fashion. At the outer level, the application comprises coarse-grained task parallelism, with dependencies between tasks reflected by an acyclic graph. At the inner level, each node of the graph is a data-parallel operation on arrays. Designers of languages, compilers, and runtime systems are building mechanisms to support such applications by providing processor groups and array remapping capabilities. In this paper we explore how to supplement these mechanisms with policy. What properties of an application, its data size, and the parallel machine determine the maximum potential gains from using both kinds of parallelism? It turns out that large gains can be expected only for specific task graph structures. For such applications, what are practical and effective ways to allocate processors to the nodes of the task graph? In principle one could solve the NP-complete problem of finding the best possible allocation of arbitrary processor subsets to nodes in the task graph. Instead of this, our analysis and simulations show that a simpleswitchedscheduling paradigm, which alternates between pure task and pure data parallelism, provides nearly optimal performance for the task graphs considered here. Furthermore, our scheme is much simpler to implement, has less overhead than the optimal allocation, and would be attractive even if the optimal allocation was free to compute. To evaluate switching in real applications, we implemented a switching task scheduler in the parallel numerical library ScaLAPACK and used it in a nonsymmetric eigenvalue program. Even for fairly large input sizes, the efficiency improves by factors of 1.5 on the Intel Paragon and 2.5 on the IBM SP-2. The remapping and scheduling overhead is negligible, between 0.5 and 5%.     

Simultaneous detection of DNA from 10 food allergens by ligation-dependent probe amplification

The simultaneous detection of DNA from different allergenic food ingredients by a ligation-dependent probe amplification (LPA) system is described. The approach allows detection of several targets in a one-tube assay. Synthetic oligonucleotides were designed to detect DNA from peanuts, cashews, pecans, pistachios, hazelnuts, sesame seeds, macadamia nuts, almonds, walnuts and brazil nuts. The specificity of the system was tested with DNA from more than 50 plant and animal species. The sensitivity of the method was suitable to detect allergenic ingredients in the low mg kg^?1 range. The limit of detection (LOD) for single allergens in different food matrices was 5 mg kg^?1. The novel analytical strategy represents a useful tool for the surveillance of established legislation on food allergens within the European Union.     

Trace element concentration in the human pineal body activation analysis of cobalt,iron, rubidium,selenium, zinc,antimony and cesium

100 human pineal bodies (56 male and 44 females) were analyzed by means of instrumental neutron activation analysis for trace element concentrations of cobalt, iron, rubidium, selenium, zinc, antimony and cesium.The results indicated that the measured element concentrations are not related to body-surface, age and fresh weight.Moreover, the mean absolute cobalt value from 97 pineal bodies is increased by a factor of 1.43 over the mean absolute concentration value of 257 other areas of the brain. The mean zinc content is found to be 3.7 fold higher than the mean zinc value from 269 other samples of brain.The iron values from various brain areas do not differ from those of the pineal body. Compared to other brain regions pineal selenium is increased and rubidium is decreased.Correlating the different element concentrations to each other, a positive correlation is found for selenium and rubidium, a negative correlation for cobalt and zinc.The present data suggest that the measured trace elements are somehow related to specific roles in the physiology and biochemistry of the pineal body. This is supported by the constancy of element concentration over a wide range of increasing fresh weights of the organ.It is considered that zinc, cobalt and iron are involved as constituents of enzymes in the metabolism of amino acids, peptides and proteins of the pineal body. Moreover the conspicuous high zinc content of this organ may be related to a so far undetected neurotransmitter.     

Detection of Cashew Nut in Foods by a Specific Real-time PCR Method

Appropriate detection methods have to be provided to assure the compliance with the recently established regulatory provisions concerning the labeling of allergens in food. Therefore, a novel real-time polymerase chain reaction (PCR) system for the specific and sensitive detection of cashew nut (Anacardium occidentale) was developed. Specificity was checked against DNA from 56 plant and animal species to avoid cross-reactivity to phylogenetically related and other food-relevant organisms. The absolute limit of detection (LOD) was determined to be 0.5 pg genomic cashew DNA and 10 copies, respectively, and the practical LOD examined exemplarily for pesto Genovese was 2 mg/kg. In addition, analysis of different retail samples was performed to demonstrate the suitability of the new assay for manifold applications.