Search NASA⌕ Search

SEARCH · Search NASA

Results for “Parallel processing”

Search indexed NASA NTRS and DOE OSTI research on propulsion, heat transfer, battery materials and energy systems. Follow report and document links to the original sources.

Quote a phrase for an exact phrase match. Source license links do not imply unrestricted reuse.

At least 973 records · Page 54

Controlled exfoliation of (100) GaAs-based devices by spalling fracture

The importance of exfoliation techniques increases as the semiconductor industry progresses toward thinner devices as a way to reduce material costs and improve performance. Here, the controlled spalling technique is a recently developed substrate removal process that utilizes the physics of fracture to create wafer cleavage parallel to the surface at a precise depth. In this letter, we apply principles of linear elastic fracture mechanics to predict the process conditions needed to exfoliate (100) GaAs of a desired thickness. Spalling can be initiated in a controllable manner, by depositing a stressor film of a residual stress value just below the threshold value to induce a spontaneous spall. Experimental data show process window requirements to controllably spall (100) GaAs. Additionally, experimental spall depth in (100) GaAs compares well to spalling mechanics predictions when the effects of wafer thickness and modulus are considered. To test spalled material quality, III-V single junction photovoltaic devices are lifted off of a (100)-GaAs substrate by spalling methods and electrical characteristics are recorded. No degradation is observed in the spalled device, illustrating the potential of this method to rapidly produce thin, high quality devices.

14 SOLAR ENERGY↗

Enhanced Hot Tensile Ductility of Mg-3Al-1Zn Alloy Thin-Walled Tubes Processed Via a Combined Severe Plastic Deformation

In the current study, combined parallel tubular channel angular pressing (PTCAP) and tube backward extrusion (TBE), as a recently developed severe plastic deformation (SPD) method, were applied at 300 °C on a commercial Mg-3Al-1Zn alloy tubes to achieve an ultrafine grained structure. Then, the microstructure, hardness, tensile properties, and fractography evaluations were done at room temperature on the SPD-processed samples. Also, to study the hot tensile ductility of the SPD-processed samples, tensile testing was performed at an elevated temperature of 400 °C, and then, the fractured surface of the tensile samples was studied. It was observed that a bimodal microstructure, with large gains surrounded by many tiny ones, was created in the sample processed by PTCAP followed by TBE. This microstructure led to reach higher hardness and higher strength at room temperature and also led to reach very high elongation to failure (~ 181%) at 400 °C. Also, the value of elongation to failure for this sample was ~ 14.1% at room temperature. The fractographic SEM images showed the occurrence of predominately ductile fracture in the samples pulled at 400 °C. This was mostly due to the nucleation of microvoids and their subsequent growth and coalescence with each other.

36 MATERIALS SCIENCE↗

Basic Research Needs for Carbon Capture: Beyond 2020

This report is based on a SC/FE workshop on Carbon Capture: Beyond 2020, held March 4–5, 2010, to assess the basic research needed to address the current technical bottlenecks in carbon capture processes and to identify key research priority directions that will provide the foundations for future carbon capture technologies. The problem of thermodynamically efficient and scalable carbon capture stands as one of the greatest challenges for modern energy researchers. The vast majority of US and global energy use derives from fossil fuels, the combustion of which results in the emission of carbon dioxide into the atmosphere. These anthropogenic emissions are now altering the climate. Although many alternatives to combustion are being considered, the fact is that combustion will remain a principal component of the global energy system for decades to come. Today’s carbon capture technologies are expensive and cumbersome and energy intensive. If scientists could develop practical and cost-effective methods to capture carbon, those methods would at once alter the future of the largest industry in the world and provide a technical solution to one of the most vexing problems facing humanity. The carbon capture problem is a true grand challenge for today’s scientists. Postcombustion CO2 capture requires major new developments in disciplines spanning fundamental theoretical and experimental physical chemistry, materials design and synthesis, and chemical engineering. To start with, the CO2 molecule itself is thermodynamically stable and binding to it requires a distortion of the molecule away from its linear and symmetric arrangement. This binding of the gas molecule cannot be too strong, however; the sheer quantity of CO2 that must be captured ultimately dictates that the capture medium must be recycled over and over. Hence the CO2 once bound, must be released with relatively little energy input. Further, the CO2 must be rapidly and selectively pulled out of a mixture that contains many other gaseous components. The related processes of precombustion capture and oxycombustion pose similar challenges. It is this nexus of high-speed capture with high selectivity and minimal energy loss that makes this a true grand challenge problem, far beyond any of today’s artificial molecular manipulation technologies, and one whose solution will drive the advancement of molecular science to a new level of sophistication. We have only to look to nature, where such chemical separations are performed routinely, to imagine what may be achieved. The hemoglobin molecule transports oxygen in the blood rapidly and selectively and releases it with minimal energy penalty. Despite our improved understanding of how this biological system works, we have yet to engineer a molecular capture system that uses the fundamental cooperativity process that lies at the heart of the functionality of hemoglobin. While such biological examples provide inspiration, we also note that newly developed theoretical and computational capabilities; the synthesis of new molecules, materials, and membranes; and the remarkable advances in characterization techniques enabled by the Department of Energy’s measurement facilities all create a favorable environment for a major new basic research push to solve the carbon capture problem within the next decade. The Department of Energy has established a comprehensive strategy to meet the nation’s needs in the carbon capture arena. This framework has been developed following a series of workshops that have engaged all the critical stakeholder communities. The strategy that has emerged is based upon a tiered approach, with Fossil Energy taking the lead in a series of applied research programs that will test and extend our current systems. ARPA-E (Advanced Research Projects Agency–Energy) is supporting potential breakthroughs based upon innovative proposals to rapidly harness today’s technical capabilities in ways not previously considered. These needs and plans have been well summarized in the report from a recent workshop—Carbon Capture 2020, held in October 5 and 6, 2009—focused on near-term strategies for carbon capture improvements (http://www.netl.doe.gov/publications/ proceedings/09/CC2020/pdfs/Richards_Summary.pdf ). Yet the fact remains that when the carbon capture problem is looked at closely, we see today’s technologies fall far short of making carbon capture an economically viable process. This situation reinforces the need for a parallel, intensive use-inspired basic research effort to address the problem. This was the overwhelming conclusion of a recent workshop—Carbon Capture: Beyond 2020, held March 4 and 5, 2010—and is the subject of the present report. To prepare for the second workshop, an in-depth assessment of current technologies for carbon capture was conducted; the result of this study was a factual document, Technology and Applied R&D Needs for Carbon Capture: Beyond 2020. This document, which was prepared by experts in current carbon capture processes, also summarized the technological gaps or bottlenecks that limit currently available carbon capture technologies. The report considered the separation processes needed for all three CO2 emission reduction strategies—postcombustion, precombustion, and oxycombustion—and assessed three primary separation technologies based on liquid absorption, membranes, and solid adsorption. The workshop “Carbon Capture: Beyond 2020” convened approximately 80 attendees from universities, national laboratories, and industry to assess the basic research needed to address the current technical bottlenecks in carbon capture processes and to identify key research priority directions that will provide the foundations for future carbon capture technologies. The workshop began with a plenary session including speakers who summarized the extent of the carbon capture challenge, the various current approaches, and the limitations of these technologies. Workshop attendees were then given the charge to identify high-priority basic research directions that could provide revolutionary new concepts to form the basis for separation technologies in 2020 and beyond. The participants were divided into three major panels corresponding to different approaches for separating gases to reduce carbon emissions—liquid absorption, solid adsorption, and membrane separations. Two other panels were instructed to attend each of these three technology panels to assess crosscutting issues relevant to characterization and computation. At the end of the workshop, a final plenary session was convened to summarize the most critical research needs identified by the workshop attendees in each of the three major technical panels and from the two cross-cutting panels. The reports of the three technical panels included a set of high level Priority Research Directions meant to serve as inspiration to researchers in multiple disciplines—materials science, chemistry, biology, computational science, engineering, and others—to address the huge scientific challenges facing this nation and the world as we seek technologies for large-scale carbon capture beyond 2020. These Priority Research Directions were clustered around three main areas, all tightly coupled: Understand and control the dynamic atomic-level and molecular-level interactions of the targeted species with the separation media. Discover and design new materials that incorporate designed structures and functionalities tuned for optimum separation properties. Tailor capture/release processes with alternative driving forces, taking advantage of a new generation of materials. In each of the technical panels, the participants identified two major crosscutting research themes. The first was the development of new analytical tools that can characterize materials structure and molecular processes across broad spatial and temporal scales and under realistic conditions that mimic those encountered in actual separation processes. Such tools are needed to examine interfaces and thin films at the atomic and molecular levels, achieving an atomic/molecular-scale understanding of gas–host structures, kinetics, and dynamics, and understanding and control of nanoscale synthesis in multiple dimensions. A second major crosscutting theme was the development of new computational tools for theory, modeling, and simulation of separation processes. Computational techniques can be used to elucidate mechanisms responsible for observed separations, predict new desired features for advanced separations materials, and guide future experiments, thus complementing synthesis and characterization efforts. These two crosscut areas underscored the fact that the challenge for future carbon capture technologies will be met only with multidisciplinary teams of scientists and engineers. In addition, it was noted that success in this fundamental research area must be closely coupled with successful applied research to ensure the continuing assessment and maturation of new technologies as they undergo scale-up and deployment. Carbon capture is a very rich scientific problem, replete with opportunity for basic researchers to advance the frontiers of science as they engage on one of the most important technical challenges of our times. This workshop report outlines an ambitious agenda for addressing the very difficult problem of carbon capture by creating foundational new basic science. This new science will in turn pave the way for many additional advances across a broad range of scientific disciplines and technology sectors.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗

Scalable deep text comprehension for Cancer surveillance on high-performance computing

Background: Deep Learning (DL) has advanced the state-of-the-art capabilities in bioinformatics applications which has resulted in trends of increasingly sophisticated and computationally demanding models trained by larger and larger data sets. This vastly increased computational demand challenges the feasibility of conducting cutting-edge research. One solution is to distribute the vast computational workload across multiple computing cluster nodes with data parallelism algorithms. In this study, we used a High-Performance Computing environment and implemented the Downpour Stochastic Gradient Descent algorithm for data parallelism to train a Convolutional Neural Network (CNN) for the natural language processing task of information extraction from a massive dataset of cancer pathology reports. We evaluated the scalability improvements using data parallelism training and the Titan supercomputer at Oak Ridge Leadership Computing Facility. To evaluate scalability, we used different numbers of worker nodes and performed a set of experiments comparing the effects of different training batch sizes and optimizer functions.Results: We found that Adadelta would consistently converge at a lower validation loss, though requiring over twice as many training epochs as the fastest converging optimizer, RMSProp. The Adam optimizer consistently achieved a close 2nd place minimum validation loss significantly faster; using a batch size of 16 and 32 allowed the network to converge in only 4.5 training epochs.Conclusions: We demonstrated that the networked training process is scalable across multiple compute nodes communicating with message passing interface while achieving higher classification accuracy compared to a traditional machine learning algorithm.

Qiu, John X.↗

Large Scale GPU Accelerated PPMLR-MHD Simulations for Space Weather Forecast

PPMLR-MHD is a new magnetohydrodynamics (MHD) model used to simulate the interactions of the solar wind with the magnetosphere, which has been proved to be the key element of the space weather cause-and-effect chain process from the Sun to Earth. Compared to existing MHD methods, PPMLR-MHD achieves the advantage of high order spatial accuracy and low numerical dissipation. However, the accuracy comes at a cost. On one hand, this method requires more intensive computation. On the other hand, more boundary data is subject to be transferred during the process of simulation. In this work, we present a parallel hybrid solution of the PPMLR-MHD model implemented using the computing capabilities of both CPUs and GPUs. We demonstrate that our optimized implementation alleviates the data transfer overhead by using GPU Direct technology and can scale up to 151 processes and achieve significant performance gains by distributing the workload among the CPUs and GPUs on Titan at Oak Ridge National Laboratory. The performance results show that our implementation is fast enough to carry out highly accurate MHD simulations in real time.

Computer Science↗

Multi-core and many-core shared-memory parallel raycasting volume rendering optimization and tuning

Given the computing industry trend of increasing processing capacity by adding more cores to a chip, the focus of this work is tuning the performance of a staple visualization algorithm, raycasting volume rendering, for shared-memory parallelism on multi-core CPUs and many-core GPUs. Our approach is to vary tunable algorithmic settings, along with known algorithmic optimizations and two different memory layouts, and measure performance in terms of absolute runtime and L2 memory cache misses. Our results indicate there is a wide variation in runtime performance on all platforms, as much as 254% for the tunable parameters we test on multi-core CPUs and 265% on many-core GPUs, and the optimal configurations vary across platforms, often in a non-obvious way. For example, our results indicate the optimal configurations on the GPU occur at a crossover point between those that maintain good cache utilization and those that saturate computational throughput. This result is likely to be extremely difficult to predict with an empirical performance model for this particular algorithm because it has an unstructured memory access pattern that varies locally for individual rays and globally for the selected viewpoint. Our results also show that optimal parameters on modern architectures are markedly different from those in previous studies run on older architectures. In addition, given the dramatic performance variation across platforms for both optimal algorithm settings and performance results, there is a clear benefit for production visualization and analysis codes to adopt a strategy for performance optimization through auto-tuning. These benefits will likely become more pronounced in the future as the number of cores per chip and the cost of moving data through the memory hierarchy both increase.

97 MATHEMATICS AND COMPUTING↗

Multi-core and Many-core Shared-memory Parallel Raycasting Volume Rendering Optimization and Tuning

Given the computing industry trend of increasing processing capacity by adding more cores to a chip, the focus of this work is tuning the performance of a staple visualization algorithm, raycasting volume rendering, for shared-memory parallelism on multi-core CPUs and many-core GPUs. Our approach is to vary tunable algorithmic settings, along with known algorithmic optimizations and two different memory layouts, and measure performance in terms of absolute runtime and L2 memory cache misses. Our results indicate there is a wide variation in runtime performance on all platforms, as much as 254% for the tunable parameters we test on multi-core CPUs and 265% on many-core GPUs, and the optimal configurations vary across platforms, often in a non-obvious way. For example, our results indicate the optimal configurations on the GPU occur at a crossover point between those that maintain good cache utilization and those that saturate computational throughput. This result is likely to be extremely difficult to predict with an empirical performance model for this particular algorithm because it has an unstructured memory access pattern that varies locally for individual rays and globally for the selected viewpoint. Our results also show that optimal parameters on modern architectures are markedly different from those in previous studies run on older architectures. And, given the dramatic performance variation across platforms for both optimal algorithm settings and performance results, there is a clear benefit for production visualization and analysis codes to adopt a strategy for performance optimization through auto-tuning. These benefits will likely become more pronounced in the future as the number of cores per chip and the cost of moving data through the memory hierarchy both increase.

97 MATHEMATICS AND COMPUTING↗

Flexible and Modular Simultaneous Modeling of Flow and Reactive Transport in Rivers and Hyporheic Zones

Investigations of coupled multiphysics processes in rivers and hyporheic zones have extensively used numerical models. Most existing models use a sequential, one-way coupling between the surface and subsurface domains. Such one-way coupling potentially introduces error. To overcome this, a fully coupled model, hyporheicFoam, was developed using the open-source computational platform OpenFOAM. It captures the coupled flow and multicomponent reactive transport processes within both surface and subsurface domains and across their interface. The coupling between two domains is implemented by mapping conservative flux boundary conditions at the interface through an iterative algorithm. Reactive transport is enabled by specifying a reaction network. To start, we have implemented reaction kinetics following the double Monod-type model with inhibition. The model capability is illustrated through modeling of both conservative and reactive hyporheic flow and transport through dune bedforms. With the novel coupled model, it is now possible to quantify reactions wherein the reactants and products are constantly exchanging between domains and have feedbacks. hyporheicFoam can simulate large, three-dimensional cases owing to the computational flexibility and power offered by the code structure and parallel design of OpenFOAM.

58 GEOSCIENCES↗

Tunnel magnetoresistance in epitaxially grown magnetic tunnel junctions using Heusler alloy electrode and MgO barrier

Epitaxially grown magnetic tunnel junctions (MTJs) with a stacking structure of Co{sub 2}MnSi/MgO/CoFe were fabricated. Their tunnel magnetoresistance (TMR) effects were investigated. The TMR ratio and tunnelling conductance characteristics of MTJs were considerably different between those with an MgO barrier prepared using sputtering (SP-MTJ) and those prepared using EB evaporation (EB-MTJ). The EB-MTJ exhibited a very large TMR ratio of 217% at room temperature and 753% at 2 K. The bias voltage dependence of the tunnelling conductance in the parallel magnetic configuration for the EB-MTJ suggests that the observed large TMR ratio at RT results from the coherent tunnelling process through the crystalline MgO barrier. The tunnelling conductance in the anti-parallel magnetic configuration suggests that the large temperature dependence of the TMR ratio results from the inelastic spin-flip tunnelling process.

36 MATERIALS SCIENCE↗

Robust and efficient overset grid assembly for partitioned unstructured meshes

This paper presents a method to perform efficient and automated Overset Grid Assembly (OGA) on a system of overlapping unstructured meshes in a parallel computing environment where all meshes are partitioned into multiple mesh-blocks and processed on multiple cores. The main task of the overset grid assembler is to identify, in parallel, among all points in the overlapping mesh system, at which points the flow solution should be computed (field points), interpolated (receptor points), or ignored (hole points). Point containment search or donor search, an algorithm to efficiently determine the cell that contains a given point, is the core procedure necessary for accomplishing this task. Donor search is particularly challenging for partitioned unstructured meshes because of the complex irregular boundaries that are often created during partitioning. Another challenge arises because of the large variation in the type of mesh-block overlap and the resulting large load imbalance on multiple processors. Desirable traits for the grid assembly method are efficiency (requiring only a small fraction of the solver time), robustness (correct identification of all point types), and full automation (no user input required other than the mesh system). Additionally, the method should be scalable, which is an important challenge due to the inherent load imbalance. This paper describes a fully-automated grid assembly method, which can use two different donor search algorithms. One is based on the use of auxiliary grids and Exact Inverse Maps (EIM), and the other is based on the use of Alternating Digital Trees (ADT). The EIM method is demonstrated to be more efficient than the ADT method, while retaining robustness. An adaptive load re-balance algorithm is also designed and implemented, which considerably improves the scalability of the method.

97 MATHEMATICS AND COMPUTING↗

Transport in the plateau regime in a tokamak pedestal

In a tokamak H-mode, a strong E Multiplication-Sign B flow shear is generated during the L-H transition. Turbulence in a pedestal is suppressed significantly by this E Multiplication-Sign B flow shear. In this case, neoclassical transport may become important. The neoclassical fluxes are calculated in the plateau regime with the parallel plasma flow using their kinetic definitions. In an axisymmetric tokamak, the neoclassical particles fluxes can be decomposed into the banana-plateau flux and the Pfirsch-Schlueter flux. The banana-plateau particle flux is driven by the parallel viscous force and the Pfirsch-Schlueter flux by the poloidal variation of the friction force. The combined quantity of the radial electric field and the parallel flow is determined by the flux surface averaged parallel momentum balance equation rather than requiring the ambipolarity of the total particle fluxes. In this process, the Pfirsch-Schlueter flux does not appear in the flux surface averaged parallel momentum equation. Only the banana-plateau flux is used to determine the parallel flow in the form of the flux surface averaged parallel viscosity. The heat flux, obtained using the solution of the parallel momentum balance equation, decreases exponentially in the presence of sonic M{sub p} without any enhancement over that in the standard neoclassical theory. Here, M{sub p} is a combination of the poloidal E Multiplication-Sign B flow and the parallel mass flow. The neoclassical bootstrap current in the plateau regime is presented. It indicates that the neoclassical bootstrap current also is related only to the banana-plateau fluxes. Finally, transport fluxes are calculated when M{sub p} is large enough to make the parallel electron viscosity comparable with the parallel ion viscosity. It is found that the bootstrap current has a finite value regardless of the magnitude of M{sub p}.

70 PLASMA PHYSICS AND FUSION TECHNOLOGY↗

Kira: Processing Astronomy Imagery Using Big Data Technology

Scientific analyses commonly compose multiple single-process programs into a dataflow. An end-to-end dataflow of single-process programs is known as a many-task application. Typically, HPC tools are used to parallelize these analyses. In this work, we investigate an alternate approach that uses Apache Spark-a modern platform for data intensive computing-to parallelize many-task applications. We implement Kira, a flexible and distributed astronomy image processing toolkit, and its Source Extractor (Kira SE) application. Using Kira SE as a case study, we examine the programming flexibility, dataflow richness, scheduling capacity and performance of Apache Spark running on the Amazon EC2 cloud. By exploiting data locality, Kira SE achieves a 4.1× speedup over an equivalent C program when analyzing a 1TB dataset using 512 cores on the Amazon EC2 cloud. Furthermore, Kira SE on the Amazon EC2 cloud achieves a 1.8× speedup over the C program on the NERSC Edison supercomputer. A 128-core Amazon EC2 cloud deployment of Kira SE using Spark Streaming can achieve a second-scale latency with a sustained throughput of 800 MB/s. Our experience with Kira demonstrates that data intensive computing platforms like Apache Spark are a performant alternative for many-task scientific applications.

97 MATHEMATICS AND COMPUTING↗

An Artificial Retina Processor for Track Reconstruction at the LHC Crossing Rate

The goal of the INFN-RETINA R&D project is to develop and implement a computational methodology that allows to reconstruct events with a large number (> 100) of charged-particle tracks in pixel and silicon strip detectors at 40 MHz, thus matching the requirements for processing LHC events at the full bunch-crossing frequency. Our approach relies on a parallel pattern-recognition algorithm, dubbed artificial retina, inspired by the early stages of image processing by the brain. In order to demonstrate that a track-processing system based on this algorithm is feasible, we built a sizable prototype of a tracking processor tuned to 3 000 patterns, based on already existing readout boards equipped with Altera Stratix III FPGAs. The detailed geometry and charged-particle activity of a large tracking detector currently in operation are used to assess its performances. We report on the test results with such a prototype.

46 INSTRUMENTATION RELATED TO NUCLEAR SCIENCE AND ↗

Mind the Gap: Building Simulation in the Architectural Design Studio

Building modelling and simulation approaches are increasingly being utilized in architectural design studios to guide and inform the design process and offer evidence-based feedback on proposed building performance. The development of intuitive and simplified simulation interfaces has greatly contributed to achieving this integration. One aspect that is often overlooked is the workflow that governs and regulates integrated design, which can have significant impacts on final design outcomes. Currently there are numerous software packages available for building performance simulations. This makes it challenging to select an appropriate tool that provides accurate results yet allows a designer to make informed architectural decisions with a designer-friendly interface. Furthermore, workflows to incorporate simulations into the design process proved to highly impact student’s project design integration. Yet, it is not clear what type of workflows are successful to achieve this goal, under what conditions, and/or for which building and site typologies. This paper addresses these issues by first reviewing three different workflows for integrating building performance simulation processes and highlighting their strengths and weaknesses. Second, a comparative case study approach was employed to test three of the most common workflows in three different integrated design architectural studios at the senior and vertical studio levels as well as in two courses that run parallel and complementary to the design studios. The workflows, processes, and the resultant student projects were further analyzed based on criteria for better integrated design and architectural excellence as outlined in the American Institute of Architects Committee on the Environment (AIA COTE) Top 10 Student’s Competition. Third, to situate the pedagogical case studies’ results within a larger context, a survey of the AIA COTE Top 10 student competition award recipients over the last five years was conducted. The results are summarized in a pedagogical framework that outlines best strategies of the type of workflows, software, design process used, methods to achieve desired interaction between design process and analytical feedback, and metrics for educators to evaluate the success of this integration and their learning outcomes in the design studio. The goal is to help bridge the gap between the building design and simulation within the design studio’s creative process for more integrated design outcomes.

Elzeyadi, Ihab↗

HyperSpace: Distributed Bayesian Hyperparameter Optimization

As machine learning models continue to increase in complexity, so does the potential number of free model parameters commonly known as hyperparameters. While there has been considerable progress toward finding optimal configurations of these hyperparameters, many optimization procedures are treated as black boxes. We believe optimization methods should not only return a set of optimized hyperparameters, but also give insight into the effects of model hyperparameter settings. To this end, we present HyperSpace, a parallel implementation of Bayesian sequential model-based optimization. HyperSpace leverages high performance computing (HPC) resources to better understand unknown, potentially non-convex hyperparameter search spaces. We show that it is possible to learn the dependencies between model hyperparameters through the optimization process. By partitioning large search spaces and running many optimization procedures in parallel, we also show that it is possible to discover families of good hyperparameter settings over a variety of models including unsupervised clustering, regression, and classification tasks.

Computer Science↗

PETSc/TAO Users Manual (Rev. 3.19)

This manual describes the use of the Portable, Extensible Toolkit for Scientific Computation (PETSc) and the Toolkit for Advanced Optimization (TAO) for the numerical solution of partial differential equations and related problems on high-performance computers. PETSc/TAO is a suite of data structures and routines that provide the building blocks for the implementation of large-scale application codes on parallel (and serial) computers. PETSc uses the MPI standard for all distributed memory communication. PETSc/TAO includes a large suite of parallel linear solvers, nonlinear solvers, time integrators, and opti mization that may be used in application codes written in Fortran, C, C++, and Python (via petsc4py; see Getting Started). PETSc provides many of the mechanisms needed within parallel application codes, such as parallel matrix and vector assembly routines. The library is organized hierarchically, enabling users to employ the level of abstraction that is most appropriate for a particular problem. By using techniques of object-oriented programming, PETSc provides enormous flexibility for users. PETSc is a sophisticated set of software tools; as such, for some users it initially has a much steeper learning curve than packages such as MATLAB or a simple subroutine library. In particular, for individuals without some computer science background, experience programming in C, C++, python, or Fortran and experience using a debugger such as gdb or lldb, it may require a significant amount of time to take full advantage of the features that enable efficient software use. However, the power of the PETSc design and the algorithms it incorporates may make the efficient implementation of many application codes simpler than “rolling them” yourself. For many tasks a package such as MATLAB is often the best tool; PETSc is not intended for the classes of problems for which effective MATLAB code can be written. There are several packages, built on PETSc, that may satisfy your needs without requiring directly using PETSc. We recommend reviewing these packages functionality before starting to code directly with PETSc. PETSc can be used to provide a “MPI parallel linear solver” in an otherwise sequential, or OpenMP parallel code. This approach cannot provide extremely large improvements in the application time by utilizing large numbers of MPI processes but can still improve the performance. Certainly all parts of a previously sequential code need not be parallelized but the matrix generation portion must be parallelized to expect true scalability to large numbers of MPI processes. See PCMPI for details on how to utilize the PETSc MPI linear solver server. Since PETSc is under continued development, small changes in usage and calling sequences of routines will occur. PETSc has been supported for twenty-five years; see mailing list information on our website for information on contacting support.

97 MATHEMATICS AND COMPUTING↗

International Perspective on the Application of Non-Destructive Assay Technology Platforms for Sentencing and Disposal of Radioactive Waste - 12113

Over the past decade, major technology improvements have been introduced in the field of Non-Destructive Assay (NDA) for the management and disposal of radioactive waste in compliance with an evolving regulatory structure. For example in the United States, various NDA technologies have been successfully developed to meet the stringent characterization requirements of the Department of Energy. The use of this instrumentation, combined with the compliant operational processes and expertise levels that have emerged in parallel, have enabled over 75,000 m{sup 3} (or in excess of 145,000 containers) of contact and remote handled transuranic (TRU) waste to be sentenced to date to the Waste Isolation Pilot Plant from 10 different consignor sites. Many of these techniques have applicability that transcends national borders and can be used for common characterization challenges in waste sentencing and disposal on an international basis. Applicable waste streams could include LLW, ILW, TRU and HLW. There are specific design aspects of assay equipment that must be tailored to meet the applicable regulatory requirements for detection and quantification of a set of nuclides of interest to a prescribed limit of detection and measurement uncertainty. Each host nation will have specific challenges in the form of matrix types and processes, availability of historical information, needs for portable versus fixed instruments and the requirement to measure all containers versus assay of a representative sample. Furthermore, the practice of load management (combining smaller packages into a larger package designed to meet the overall waste acceptance criteria for the bulk container) may not have universal acceptability. An evaluation has been performed on a sample of the most successful technologies that have recently emerged to understand their applicability in other countries. Two types of instrumentation 'suite' are considered for measurements on drums and larger boxes / crates: (i) High Efficiency Neutron Coincidence counting combined with gamma isotopic analysis (e.g. the SuperHENC), and (ii) Portable NDA using neutron slab counters and far field high resolution gamma spectroscopy. Target countries (outside the United States) for these applications are: China, Japan, UK, Sweden and Canada. Specific challenges in each country are addressed and the applicability of the technologies as 'universal platforms' is considered. (authors)

12 MANAGEMENT OF RADIOACTIVE AND NON-RADIOACTIVE W↗

Transparent runtime parallelization of the R scripting language

Scripting languages such as R and Matlab are widely used in scientific data processing. As the data volume and the complexity of analysis tasks both grow, sequential data processing using these tools often becomes the bottleneck in scientific workflows. We describe pR, a runtime framework for automatic and transparent parallelization of the popular R language used in statistical computing. Recognizing scripting languages interpreted nature and data analysis codes use pattern, we propose several novel techniques: (1) applying parallelizing compiler technology to runtime, whole-program dependence analysis of scripting languages, (2) incremental code analysis assisted with evaluation results, and (3) runtime parallelization of file accesses. Our framework does not require any modification to either the source code or the underlying R implementation. Experimental results demonstrate that pR can exploit both task and data parallelism transparently and overall has better performance as well as scalability compared to an existing parallel R package that requires code modification.

99 GENERAL AND MISCELLANEOUS↗