Achieving accurate and context‐sensitive timing for code optimization

Key computational kernels must run near their peak efficiency for most high‐performance computing (HPC) applications. Getting this level of efficiency has always required extensive tuning of the kernel on a particular platform of interest. The success or failure of an optimization is usually measured by invoking a timer. Understanding how to build reliable and context‐sensitive timers is one of the most neglected areas in HPC, and this results in a host of HPC software that looks good when reported in the papers, but delivers only a fraction of the reported performance when used by actual HPC applications. In this paper, we motivate the importance of timer design and then discuss the techniques and methodologies we have developed in order to accurately time HPC kernel routines for our well‐known empirical tuning framework, ATLAS. Copyright © 2008 John Wiley & Sons, Ltd.     

Exploring the best sequence LSTM modeling architecture for flood prediction

Neural Computing and Applications - Accurate and efficient models for rainfall–runoff (RR) simulations are crucial for flood risk management. Recently, the success of the recurrent neural...     

Morphology-based approaches for detecting stream channels from ALSM data

To map the Earth's surface at remarkable resolution, Airborne Laser Swath Mapping (ALSM) instrument technology and subsequent algorithms have been used over the last several years. Since forested watersheds have commonly been problematic to study with remote sensing techniques, the ability of ALSM technology to densely sample ground elevations beneath forest canopies is especially considerable. Stream network detection from digital elevation models (DEMs) is a key role in modelling spatially distributed hydrological processes. To detect stream channels, we have developed two approaches. The first approach is based on an encoding of mathematical morphological operators. In the second approach, a composition of geodesic top-hat and bot-hat operations of different sizes is used in order to build a morphological profile (P _M) that records the image structural information. The two proposed methods perform well in terms of detection results and classification accuracies. The second approach is more general than the first, but it also requires training and more computation.     

High Dietary Fat Exacerbates Weight Gain and Obesity in Female Liver Fatty Acid Binding Protein Gene-Ablated Mice

Since liver fatty acid binding protein (L-FABP) facilitates uptake/oxidation of long-chain fatty acids in cultured transfected cells and primary hepatocytes, loss of L-FABP was expected to exacerbate weight gain and/or obesity in response to high dietary fat. Male and female wild-type (WT) and L-FABP gene-ablated mice, pair-fed a defined isocaloric control or high fat diet for 12 weeks, consumed equal amounts of food by weight and kcal. Male WT mice gained weight faster than their female WT counterparts regardless of diet. L-FABP gene ablation enhanced weight gain more in female than male mice—an effect exacerbated by high fat diet. Dual emission X-ray absorptiometry revealed high-fat fed male and female WT mice gained mostly fat tissue mass (FTM). L-FABP gene ablation increased FTM in female, but not male, mice—an effect also exacerbated by high fat diet. Concomitantly, L-FABP gene ablation decreased serum β-hydroxybutyrate in male and female mice fed the control diet and, even more so, on the high-fat diet. Thus, L-FABP gene ablation decreased fat oxidation and sensitized all mice to weight gain as whole body FTM and LTM—with the most gain observed in FTM of control vs high-fat fed female L-FABP null mice. Taken together, these results indicate loss of L-FABP exacerbates weight gain and/or obesity in response to high dietary fat.     

Strategies for stationary and portable fuel cell markets

In the future, hydrogen-based stationary and portable fuel cell systems can help supply some or all of the power demanded with additional advantages of higher reliability, lower emissions, independence from the general grid, and cogeneration capability. In order to understand how to prepare the future for this technology, this paper describes a thorough investigation of past alternative stationary and portable power projects in order for an assessment of the opportunities for stationary and portable fuel cell markets, as well as interactions with transportation hydrogen systems. The lessons learned from the programs are used to establish best practices and recommendations for a hydrogen strategy that addresses opportunities for hydrogen in power generation systems, as well as to make recommendations for market transformation within the hydrogen fuel cell industry.     

A Performance Comparison of Continuous and Discontinuous Finite Element Shallow Water Models

We present a comparative study of two finite element shallow water equation (SWE) models: a generalized wave continuity equation based continuous Galerkin (CG) model—an approach used by several existing SWE models—and a recently developed discontinuous Galerkin (DG) model. While DG methods are known to possess a number of favorable properties, such as local mass conservation, one commonly cited disadvantage is the larger number of degrees of freedom associated with the methods, which naturally translates into a greater computational cost compared to CG methods. However, in a series of numerical tests, we demonstrate that the DG SWE model is generally more efficient than the CG model (i) in terms of achieving a specified error level for a given computational cost and (ii) on large-scale parallel machines because of the inherently local structure of the method. Both models are verified on a series of analytic test cases and validated on a field-scale application.     

Fixed three-dimensional holographic images

Three-dimensional holograms were recorded in a cerium-doped, strontium barium niobate (SBN:75) photorefractive crystal. These holograms are shown to not degrade after more than one week of continuous readout and to reconstruct reproductions of the original object with an observable field of view of approximately 35 degrees.     

A Discontinuous Galerkin Method for Three-Dimensional Shallow Water Equations

We describe the application of a local discontinuous Galerkin method to the numerical solution of the three-dimensional shallow water equations. The shallow water equations are used to model surface water flows where the hydrostatic pressure assumption is valid. The authors recently developed a DG\linebreak method for the depth-integrated shallow water equations. The method described here is an extension of these ideas to non-depth-integrated models. The method and its implementation are discussed, followed by numerical examples on several test problems.This revised version was published online in July 2005 with corrected volume and issue numbers.     

Minimizing development and maintenance costs in supporting persistently optimized BLAS

The Basic Linear Algebra Subprograms (BLAS) define one of the most heavily used performance‐critical APIs in scientific computing today. It has long been understood that the most important of these routines, the dense Level 3 BLAS, may be written efficiently given a highly optimized general matrix multiply routine. In this paper, however, we show that an even larger set of operations can be efficiently maintained using a much simpler matrix multiply kernel. Indeed, this is how our own project, ATLAS (which provides one of the most widely used BLAS implementations in use today), supports a large variety of performance‐critical routines. Copyright © 2004 John Wiley & Sons, Ltd.     

Grab and Go Systems: a CPO approach to concurrent web and grid-based computation

A useful property of web–based information systems [1] is the ability to display partial information. For example, a web program can display a sequence of fuzzy images which is extended by the production of improved images as execution of the program proceeds. A sequence of improving approximations to an image can be modelled by the elements of a complete partial order (CPO). CPOs can also be used to model grid-based computations. For example, consider a collection of iterative processes, each with identical functionality, which individually generate a series of improving approximations towards a desired goal. Sharing approximations from time to time among the processes may produce faster convergence than could be achieved by any of the processes separately. Each process in the collection has the potential to exchange partial information with its companions so that all may make use of the best information available.In this paper a non-blocking communication abstraction, based on CPOs, is used to develop a model of iterative web- and grid-based computations. The abstraction is novel in that it may not directly match a send communication in one process with a corresponding receive communication in another; rather, a receive communication is identified with taking the least upper bound of the set of messages available at an input port - this set may be empty, contain exactly one messsage or contain multiple messages. In all cases the receiver does not wait but gathers the best available information and proceeds with its computation. Thus, the abstraction corresponds to a loosely coupled model of distributed computation. The applicability of the model is illustrated by a number of disparate examples of distributed iterative computation.