Selecting linear algebra kernel composition using response time prediction

Numerical linear algebra libraries provide many kernels that can be composed to perform complex computations. For a given computation, there is typically a large number of functionally equivalent kernel compositions. Some of these compositions achieve better response times than others for particular data and when executed on a particular computer architecture. Previous research provides methods to enumerate (a subset of) these kernel compositions. In this work, we study the problem of determining the composition that yields the lowest response time. Our approach is based on a response time prediction for each candidate combination. While this prediction could in principle be obtained using analytical and/or empirical performance models, developing accurate such models is known to be challenging. Instead, we define a feature space that captures salient properties of kernel combinations and predict response time using supervised machine learning. We experiment with a standard set of machine learning algorithms and identify an effective algorithm for our kernel composition selection problem. Using this algorithm, our approach widely outperforms the strategy that would consist in always using the simplest kernel composition and is often close to the fastest kernel compositions among those evaluated. We quantify the potential benefit of our approach if it were to be implemented as part of an interactive computational tool. We find that although the potential benefit is substantial, a limiting factor is the kernel composition enumeration overhead. Copyright © 2014 John Wiley & Sons, Ltd.     

On the Easy Use of Scientific Computing Services for Large Scale Linear Algebra and Parallel Decision Making with the P-Grade Portal

Scientific research is becoming increasingly dependent on the large-scale analysis of data using distributed computing infrastructures (Grid, cloud, GPU, etc.). Scientific computing (Petitet et al. 1999) aims at constructing mathematical models and numerical solution techniques for solving problems arising in science and engineering. In this paper, we describe the services of an integrated portal based on the P-Grade (Parallel Grid Run-time and Application Development Environment) portal (http://www.p-grade.hu) that enables the solution of large-scale linear systems of equations using direct solvers, makes easier the use of parallel block iterative algorithm and provides an interface for parallel decision making algorithms. The ultimate goal is to develop a single sign on integrated multi-service environment providing an easy access to different kind of mathematical calculations and algorithms to be performed on hybrid distributed computing infrastructures combining the benefits of large clusters, Grid or cloud, when needed.     

The long and winding road to inhaled TB therapy: not only the bug’s fault

Not all of the issues impacting the success of tuberculosis (TB) treatment arise from pathogen-related disease characteristics. Nowadays, there is an increasing awareness that antibiotic treatment is not the only answer to the TB problem, promoting the search for alternative administration strategies and host-directed therapies. Among all the administration routes, being the lungs the main TB focus, inhalation is conceptually a logical solution to enhance treatment effectiveness and compliance. Nevertheless, research efforts and funding are almost entirely conveyed to conventional approaches. This review will critically evaluate the reasons constraining research in this field, providing some future perspectives. The most recent advances in inhalation approaches for TB will be discussed, either at the preclinical or clinical phase, illustrating the risk of failure and chances of success.     

The effects of asymmetrical cold rolling on kinetics,grain size and texture in IF steels

Asymmetrical rolling is a promising way to achieve high rolling reductions and potentially grain refinement. However in steels the effect of asymmetrical rolling is not thoroughly quantified. In this paper the impact of cold asymmetrical rolling obtained by changing the roll velocity ratio (from 1.1 to 1.45) on the deformation and recrystallization features has been finely investigated by Electron Back-Scattering Diffraction in a Scanning Electron Microscope. In the investigated conditions (with low thickness reduction), it is found that asymmetrical rolling induces grain refinement and more homogeneous gamma fiber texture.     

Physicochemical and pasting properties of maize flour as a function of the interactive effect of natural-fermentation and roasting

The effect of natural-fermentation and roasting on maize physicochemical and functional properties were evaluated. Natural fermented maize seeds were soaked for variable times (0, 24 and 48 h) and directly decorticated or roasted before milling into flour. Generally during the 24 to 48 h natural fermentation, total sugars (80.3–92.0 g/100 g) and proteins contents (4.0–7.6 g/100 g) decreased while they significantly increased with soaking. Soluble sugars content of the unroasted maize significantly decreased with fermentation while that of roasted maize significantly increased. Reverse observations were made on soluble proteins. The antinutrients (phytates and total polyphenols) contents of the grains as well as the functional properties of their flours were observed to have been significantly changed following fermentation and roasting. In particular the least gelation concentration (6 to 18 g/100 mL), an important reverse index of gelating power significantly increased with fermentation and roasting. As consequence the viscosity of the fermented maize flours were systematically significantly lower than that of the unfermented flour. Generally while the effects of duration of fermentation and roasting on the viscosity were not consistent, the 48 h natural fermented and roasted flour was observed to particularly produce flours of much lower viscosity. This highlighted the positive effect of combining fermentation and roasting to improve the quality of weaning flour made from maize.     

In search of the lost polychromy of English medieval alabaster panels in the Southwest of France

Polychrome alabaster carvings, representing religious scenes, constitute England's major contribution to European mediaeval sculpture. During the period between 1350 and 1550, large number of alabaster carvings were produced in the Midlands. Despite their massive destruction during the English Reformation (ca.1534), more than 2500 of them still survive, either legally exported or sold clandestinely to the Continent. With more than 100 panels still existing, the Bordeaux region holds one of the largest concentrations of these reliefs. Most of them have unfortunately lost their medieval polychromy, but 20 can still provide valuable information. They are currently studied in our multidisciplinary research program “ALBATRES” (Polychromie, pigments, perception: les albâtres anglais de la fin du Moyen Âge conservés sur le territoire aquitain. Labex LaScArBx project [2018-2020]). Focused on the panels' polychromy, the project links art history, archaeometry, optical and 3D engineering, and experimental archeology. The first aim is to obtain complete information on the materials and techniques used by medieval painters. The second consists in studying the functions of polychromy and its perception, by determining and interpreting the selection criteria and aesthetic uses of colors and gilding. The polychrome remains of the Aquitaine corpus are studied by means of visual examination, macrophotographic documentation, and noninvasive portable spectroscopic analyses (reflectance spectroscopy by fiber optics in visible and infrared ranges, fluorimetry, etc.). They reveal, among others, the presence of cinnabar, red lead, red ochre, copper green pigments, yellow ochre, gold, lead white, and others. The results allow the production of model samples to help understand the medieval color organization before proposing the coloring of a real copy of one of the studied panels (Virgin's Assumption, Musée d'Aquitaine, Bordeaux). The shape and the appearance of the color rendering could be improved thanks to a special device built at the IOGS to develop a shader that approximates as closely as possible the appearance of the panels. Finally, all information will be integrated on 3D model panels for better appreciation of the carvings and the artistic taste of medieval spectators.     

Magnetic spheres in a superconductor

We consider a cubic lattice of small magnetic spheres imbedded in a type II superconductor. We compute the field distribution for one sphere and the interaction energy when the moments are aligned. It is pointed out that the dipolar field due to the nearest neighbors, in opposition to the normal situation, is nonzero in the superconducting state and induces then an antiferromagnetic ordering of the moments.On leave of absence from Service de Physique des Solides, Faculté des Sciences d'Orsay, 91-Orsay, France. Partially supported by NSF and NATO.     

Fine (PM2.5) and coarse (PM2.5-10) particulate matter on a heavily trafficked London highway: sources and processes

A large dataset for PM2.5 and PMcoarse (PM2.5-10) concentrations monitored near a busy London highway (Marylebone Road) has been analyzed to define the factors that lead to high concentrations. The following have been highlighted as major influencing parameters: wind speed, prevailing wind direction (because of its role on the microscale dispersion within the street), the daily cycle of the atmospheric boundary layer (stable during the night/ convective and mixed during the day), and traffic density. The mainly diesel heavy-duty vehicles are the main source of fine particulate matter at Marylebone Road. In particular, lorries (trucks) dominate PM10 exhaust emissions which are mainly in the fine (<2.5 microm) size range. A strong correlation with PMcoarse suggests that the heavy-duty traffic is largely responsible for this component also. Substantial local increments in PM2.5 and PMcoarse due to traffic have been estimated and a large part of the increment in PMcoarse concentrations is inferred to arise from resuspended road dust emissions since the contribution of abrasion processes estimated from emission factors is modest. Despite the strong influence of traffic on PM concentrations measured at Marylebone Road the analysis of factors leading to the highest 5% of hourly concentrations of PM10 at Marylebone Road reveals that almost half of these events were due to building works. The other events occurred when all or most of the key factors occurred simultaneously (heavy traffic, poor dispersion, etc.). Some episodes of high PM2.5 concentrations were associated with long-range transport in which the regional PM2.5 constituted most of the local concentrations.