Search NASA⌕ Search

SEARCH · Search NASA

Results for “Parallel application”

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 577 records · Page 32

Simulating the Birth of the Universe on a PetaFlop

In the beginning, when the universe was less than one microsecond old and more than one trillion degrees hot, it transformed from a plasma of quarks and gluons into bound states of quarks we reefer to as protons and neutrons, the fundamental building blocks of nuclear matter that make up most of the visible universe. We believe this happened because the theory of quantum chromodynamics (QCD), which governs the interactions of the strong nuclear force, predicts it should happen when such conditions occur. Recent experiments at the Relativistic Heavy Ion Collider at Brookhaven National Laboratory have provided direct evidence of the existence of this phase transition in collisions between gold nuclei at the highest attainable energies. But calculating the properties of this phase transition has been notoriously difficult and computationally challenging. What we've been able to discern via this theory has come by situating space and time onto a 4D grid of lattice points, on a volume no bigger than the size of a large nucleus. This discrete formulation of QCD known as lattice QCD (LQCD) can be tamed in a way that's numerically amenable to massively parallel machines. Indeed, our LQCD code scaled with perfect speedup all the way to the 131,072 CPU cores on the world's fastest supercomputer, the BlueGene/L (BG/L) system at Lawrence Livermore National Laboratory (LLNL). Efforts are currently under way to calculate this phase transition's properties using tens of teraflops spread across many of the world's fastest computers. However, the lattice discretization used in our calculations of space-time breaks one of the most crucial properties of the underlying theory: the basic symmetry of chiral rotations. Chiral symmetry can't be fully restored on the lattice, but it can be restored and controlled to a high degree by using the method of domain wall fermions (DWF). DWF introduces an extra fifth dimension, in which chiral symmetry is restored with only a small amount of breaking, which decreases as the fifth dimension's size increases. In particular, researchers have estimated that a fifth dimension of roughly 64 lattice points is sufficient to restore symmetry with a systematic error of only a few percent, even around the transition temperature. Studies have also shown that DWF exhibits excellent fidelity in incorporating the transition's basic driving forces. But the computational cost of this calculation is roughly 2 x 64 = 128 times higher than the cost of current methods that don't preserve this symmetry on the lattice, and thus it falls outside the realm of current terascale supercomputing. With petascale supercomputing, however, this restriction is lifted. If we had petascale supercomputing for our calculation, we could use computational techniques similar to those used on BG/L to calculate the thermodynamic properties--or more precisely, the equation of state--of the quark-gluon plasma just as it begins to form protons and neutrons as it did in the very early universe. The basic level of parallelism again comes from the division of the 4D grid into subgrids assigned to nodes on a massively parallel machine, thereby providing perfect speedup. DWF's fifth dimension provides yet another level of parallelism: given that the nodes of a massively parallel petascale supercomputer will likely consist of several CPU cores per chip, it's natural to fully map the fifth dimension along the node chip's CPU cores. On a 16-CPU core chip, for example, we could assign four fifth-dimension 'slices' per CPU core for a total of 64. In this way, the new level of parallelism in the application matches the new level of parallelism in the hardware. The communication patterns along the fifth dimension are nearest-lattice-neighbor only and can be handled via shared memory or any other on-chip data transfer mechanism, thus alleviating off-chip data transfers while exploiting the hardware's native capability.

72 PHYSICS OF ELEMENTARY PARTICLES AND FIELDS↗

Enhancing Application Performance Using Mini-Apps: Comparison of Hybrid Parallel Programming Paradigms

In many fields, real-world applications for High Performance Computing have already been developed. For these applications to stay up-to-date, new parallel strategies must be explored to yield the best performance; however, restructuring or modifying a real-world application may be daunting depending on the size of the code. In this case, a mini-app may be employed to quickly explore such options without modifying the entire code. In this work, several mini-apps have been created to enhance a real-world application performance, namely the VULCAN code for complex flow analysis developed at the NASA Langley Research Center. These mini-apps explore hybrid parallel programming paradigms with Message Passing Interface (MPI) for distributed memory access and either Shared MPI (SMPI) or OpenMP for shared memory accesses. Performance testing shows that MPI+SMPI yields the best execution performance, while requiring the largest number of code changes. A maximum speedup of 23 was measured for MPI+SMPI, but only 11 was measured for MPI+OpenMP.

Lawson, Gary↗

Profiling and Improving I/O Performance of a Large-Scale Climate Scientific Application

Exascale computing systems are soon to emerge, which will pose great challenges on the huge gap between computing and I/O performance. Many large-scale scientific applications play an important role in our daily life. The huge amounts of data generated by such applications require highly parallel and efficient I/O management policies. In this paper, we adopt a mission-critical scientific application, GEOS-5, as a case to profile and analyze the communication and I/O issues that are preventing applications from fully utilizing the underlying parallel storage systems. Through in-detail architectural and experimental characterization, we observe that current legacy I/O schemes incur significant network communication overheads and are unable to fully parallelize the data access, thus degrading applications' I/O performance and scalability. To address these inefficiencies, we redesign its I/O framework along with a set of parallel I/O techniques to achieve high scalability and performance. Evaluation results on the NASA discover cluster show that our optimization of GEOS-5 with ADIOS has led to significant performance improvements compared to the original GEOS-5 implementation.

Exascale↗

An Asynchronous Many-Task Implementation of In-Situ Statistical Analysis using Legion.

In this report, we propose a framework for the design and implementation of in-situ analy- ses using an asynchronous many-task (AMT) model, using the Legion programming model together with the MiniAero mini-application as a surrogate for full-scale parallel scientific computing applications. The bulk of this work consists of converting the Learn/Derive/Assess model which we had initially developed for parallel statistical analysis using MPI [PTBM11], from a SPMD to an AMT model. In this goal, we propose an original use of the concept of Legion logical regions as a replacement for the parallel communication schemes used for the only operation of the statistics engines that require explicit communication. We then evaluate this proposed scheme in a shared memory environment, using the Legion port of MiniAero as a proxy for a full-scale scientific application, as a means to provide input data sets of variable size for the in-situ statistical analyses in an AMT context. We demonstrate in particular that the approach has merit, and warrants further investigation, in collaboration with ongoing efforts to improve the overall parallel performance of the Legion system.

Pebay, Philippe Pierre↗

CCM vs. CRM Design Optimization of a Boost-derived Parallel Active Power Decoupler for Microinverter Applications

Single-phase inverter or rectifier systems often make use of an auxiliary active power decoupler (APD) to balance the mismatch between steady DC power and fluctuating AC power. This paper deals with efficiency and size optimization of a parallel boost-type APD circuit for PV microinverter applications. Specifically, design of an eGaNFET-based, 400 W APD circuit, employing planar inductor and operating in either continuous conduction mode (CCM) or critical conduction mode (CRM) is considered. Available design variables including inductance value, inductor core geometry, capacitor voltage, switching frequency, and modulation scheme (CCM vs. CRM) are explored to identify Pareto-optimal configurations, which can achieve low California Energy Commission (CEC) efficiency drop while also reducing the footprint area of the inductor. The theoretical study predicts that the optimal CRM design can achieve 37% reduced inductor size, while operating with similar efficiency drop, compared to the optimal CCM design. Experimental results, obtained using two separate 40 V, 400 W hardware prototypes for CCM and CRM, are presented to verify the analyses.

14 SOLAR ENERGY↗

Position Paper - pFLogger: The Parallel Fortran Logging framework for HPC Applications

In the context of high performance computing (HPC), software investments in support of text-based diagnostics, which monitor a running application, are typically limited compared to those for other types of IO. Examples of such diagnostics include reiteration of configuration parameters, progress indicators, simple metrics (e.g., mass conservation, convergence of solvers, etc.), and timers. To some degree, this difference in priority is justifiable as other forms of output are the primary products of a scientific model and, due to their large data volume, much more likely to be a significant performance concern. In contrast, text-based diagnostic content is generally not shared beyond the individual or group running an application and is most often used to troubleshoot when something goes wrong. We suggest that a more systematic approach enabled by a logging facility (or logger) similar to those routinely used by many communities would provide significant value to complex scientific applications. In the context of high-performance computing, an appropriate logger would provide specialized support for distributed and shared-memory parallelism and have low performance overhead. In this paper, we present our prototype implementation of pFlogger a parallel Fortran-based logging framework, and assess its suitability for use in a complex scientific application.

Fortran↗

POSITION PAPER - pFLogger: The Parallel Fortran Logging Framework for HPC Applications

In the context of high performance computing (HPC), software investments in support of text-based diagnostics, which monitor a running application, are typically limited compared to those for other types of IO. Examples of such diagnostics include reiteration of configuration parameters, progress indicators, simple metrics (e.g., mass conservation, convergence of solvers, etc.), and timers. To some degree, this difference in priority is justifiable as other forms of output are the primary products of a scientific model and, due to their large data volume, much more likely to be a significant performance concern. In contrast, text-based diagnostic content is generally not shared beyond the individual or group running an application and is most often used to troubleshoot when something goes wrong. We suggest that a more systematic approach enabled by a logging facility (or 'logger') similar to those routinely used by many communities would provide significant value to complex scientific applications. In the context of high-performance computing, an appropriate logger would provide specialized support for distributed and shared-memory parallelism and have low performance overhead. In this paper, we present our prototype implementation of pFlogger - a parallel Fortran-based logging framework, and assess its suitability for use in a complex scientific application.

Clune, Thomas L.↗

Proceedings of the Peteflops-Systems Operation Working Review (POWR)

This report constitutes the final technical report. Even as Petaflops performance computing is being achieved for a few important applications on the nation's largest massively parallel processing (MPP) systems, the challenges of realizing far greater performance are being investigated by a strong interdisciplinary team of experts from academia, industry, and government. This team is also exploring the extraordinary opportunity such capabilities would provide for critical areas of strategic national importance. Under what has become known informally as the "Petaflops Initiative," key leaders in research across the areas of device technology, parallel systems architecture, applications and algorithms, and systems software have been exploring the implications, requirements, interrelationships, and trade-offs among these research areas through a series of workshops, studies, and projects sponsored by a number of Federal agencies.

Sterling, Thomas↗

Automatic Parallelization Using OpenMP Based on STL Semantics

Automatic parallelization of sequential applications using OpenMP as a target has been attracting significant attention recently because of the popularity of multicore processors and the simplicity of using OpenMP to express parallelism for shared-memory systems. However, most previous research has only focused on C and Fortran applications operating on primitive data types. C++ applications using high level abstractions such as STL containers are largely ignored due to the lack of research compilers that are readily able to recognize high level object-oriented abstractions of STL. In this paper, we use ROSE, a multiple-language source-to-source compiler infrastructure, to build a parallelizer that can recognize such high level semantics and parallelize C++ applications using certain STL containers. The idea of our work is to automatically insert OpenMP constructs using extended conventional dependence analysis and the known domain-specific semantics of high-level abstractions with optional assistance from source code annotations. In addition, the parallelizer is followed by an OpenMP translator to translate the generated OpenMP programs into multi-threaded code targeted to a popular OpenMP runtime library. Our work extends the applicability of automatic parallelization and provides another way to take advantage of multicore processors.

99 GENERAL AND MISCELLANEOUS↗

Parallel compilation - A design and its application to SIMULA 67

A design for a separate compilation facility for the SIMULA 67 programming language is presented. The paper explores the problems with existing separate compilation schemes, and proposes a new scheme that allows top-down, bottom-up, or even parallel development and integration of program modules. An evaluation of the proposal and a discussion of its applicability to other languages are then given.

Schwartz, R. L.↗

Massively Parallel Near-Linear Scalability Algorithms with Application to Unstructured Video Analysis

This paper describes the use of high performance, massively parallel unstructured data analysis techniques to computationally extract human faces from streaming video data. The use of parallel high-throughput algorithms is essential to maximize the rate at which frames can be analyzed so large volumes of data can be evaluated to answer the question “Have we seen this person’s face before?” This paper will highlight the method used to achieve near-linear scalability according to number of processing cores for principle component analysis (PCA) on the Texas Advanced Computing Center (TACC) Ranger supercomputer. A global summation across processing core operation, accomplished via MPI_Reduce on Ranger, causes a slight deviation from linear scaling behavior. This mapping has been utilized in the past to enable a host of data centric analytic techniques such as neural networks, multi-dimensional scaling (MDS), signal processing and other methods to address, with massively parallel computers, problems of scientific interest in Physics, Mathematics, Biology and Chemistry.

99 GENERAL AND MISCELLANEOUS↗

Measurement of PVFS2 performance on InfiniBand

InfiniBand is becoming increasingly popular as a fast interconnect technology between servers and storage. It has far better price/performance ratio compared to both Gigabit Ethernet and 10 Gigabit Ethernet, and hence is being increasingly used for highperformance computing applications. PVFS2, the second generation Parallel Virtual File System (PVFS), is a distributed file system for parallel data access that is being increasingly used in clustered applications. As previous studies have shown, in general, PVFS2 over InfiniBand offers enhanced I/O rates compared to PVFS2 over TCP and Gigabit Ethernet. Apart from the hardware technology, the application programming interface into the file system also makes a difference. To get better parallel performance, the choice of a file system interface is important. Our study is to benchmark and compare the performance of PVFS2 running over InfiniBand using different file system interfaces. IOR is a popular I/O benchmarking tool that supports the POSIX and MPI I/O file system interfaces. In addition to testing these already supported interfaces, we have written a PVFS2 module extension for IOR to support native PVFS2 interfaces into the PVFS2 file system. As we shall see in this study, using native PVFS2 interface offers significant performance benefit compared to other file system interfaces on the PVFS2 file system. Our benchmarking effort also involves studying the effect of a multi-client environment on the I/O performance of different file system interfaces. Based on the benchmarking results we obtain, we determine the most efficient application programming interface for parallel I/O on PVFS2 in a typical multi-client parallel application scenario.

97 MATHEMATICS AND COMPUTING↗

Towards a theory of economic value for modeling and simulation: incremental cost of parallel simulation (wip)

A useful perspective on the economic value of simulation in engineering comes from considering the role simulation plays in constructing a system. We propose a conceptual model that combines an embryonic concept of the value of an information bit with cost estimation tools used in software engineering. The model provides an economic justification for the common conception that models should be built to address specific questions rather than to comprehensively represent a referent system. Extending these ideas, we offer a model for the incremental value of parallel execution to explain why simulation applications that take advantage of sophisticated parallel discrete event simulation tools form a relatively narrow niche despite the ubiquity of parallel computing hardware. The proposed model enables us to bring value of information criteria into model development and to make predictions about the future of parallel simulation applications.

Nutaro, James J↗

Parallel processing and expert systems

Whether it be monitoring the thermal subsystem of Space Station Freedom, or controlling the navigation of the autonomous rover on Mars, NASA missions in the 1990s cannot enjoy an increased level of autonomy without the efficient implementation of expert systems. Merely increasing the computational speed of uniprocessors may not be able to guarantee that real-time demands are met for larger systems. Speedup via parallel processing must be pursued alongside the optimization of sequential implementations. Prototypes of parallel expert systems have been built at universities and industrial laboratories in the U.S. and Japan. The state-of-the-art research in progress related to parallel execution of expert systems is surveyed. The survey discusses multiprocessors for expert systems, parallel languages for symbolic computations, and mapping expert systems to multiprocessors. Results to date indicate that the parallelism achieved for these systems is small. The main reasons are (1) the body of knowledge applicable in any given situation and the amount of computation executed by each rule firing are small, (2) dividing the problem solving process into relatively independent partitions is difficult, and (3) implementation decisions that enable expert systems to be incrementally refined hamper compile-time optimization. In order to obtain greater speedups, data parallelism and application parallelism must be exploited.

Lau, Sonie↗

Automated Generation of Message-Passing Programs: An Evaluation of CAPTools using NAS Benchmarks

Scientists at NASA Ames Research Center have been developing computational aeroscience applications on highly parallel architectures over the past ten years. During the same time period, a steady transition of hardware and system software also occurred, forcing us to expand great efforts into migrating and receding our applications. As applications and machine architectures continue to become increasingly complex, the cost and time required for this process will become prohibitive. Various attempts to exploit software tools to assist and automate the parallelization process have not produced favorable results. In this paper, we evaluate an interactive parallelization tool, CAPTools, for parallelizing serial versions of the NAB Parallel Benchmarks. Finally, we compare the performance of the resulting CAPTools generated code to the hand-coded benchmarks on the Origin 2000 and IBM SP2. Based on these results, a discussion on the feasibility of automated parallelization of aerospace applications is presented along with suggestions for future work.

Hribar, Michelle R.↗

Understanding scale-dependent soft-error behavior of scientific applications

Analyzing application fault behavior on large-scale systems is time-consuming and resource-demanding. Currently, researchers need to perform fault injection campaigns at full scale to understand the effects of soft errors on applications and whether these faults result in silent data corruption. Both time and resource requirements greatly limit the scope of the resilience studies that can be currently performed. In this work, we propose a methodology to model application fault behavior at large scale based on a reduced set of experiments performed at small scale. We employ machine learning techniques to accurately model application fault behavior using a set of experiments that can be executed in parallel at small scale. Our methodology drastically reduces the set and the scale of the fault injection experiments to be performed and provides a validated methodology to study application fault behavior at large scale. We show that our methodology can accurately model application fault behavior at large scale by using only small scale experiments. In some cases, we can model the fault behavior of a parallel application running on 4,096 cores with about 90% accuracy based on experiments on a single core.

Kestor, Gokcen G.↗

PyPele Rewritten To Use MPI

A computer program known as PyPele, originally written as a Pythonlanguage extension module of a C++ language program, has been rewritten in pure Python language. The original version of PyPele dispatches and coordinates parallel-processing tasks on cluster computers and provides a conceptual framework for spacecraft-mission- design and -analysis software tools to run in an embarrassingly parallel mode. The original version of PyPele uses SSH (Secure Shell a set of standards and an associated network protocol for establishing a secure channel between a local and a remote computer) to coordinate parallel processing. Instead of SSH, the present Python version of PyPele uses Message Passing Interface (MPI) [an unofficial de-facto standard language-independent application programming interface for message- passing on a parallel computer] while keeping the same user interface. The use of MPI instead of SSH and the preservation of the original PyPele user interface make it possible for parallel application programs written previously for the original version of PyPele to run on MPI-based cluster computers. As a result, engineers using the previously written application programs can take advantage of embarrassing parallelism without need to rewrite those programs.

Hockney, George↗