A work stealing based approach for enabling scalable optimal sequence homology detection
Not Available
SEARCH · Search NASA
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.
Not Available
In this work, a parallel algorithm is described for the coupled-cluster singles and doubles method augmented with a perturbative correction for triple excitations [CCSD(T)] using the resolution-of-the-identity (RI) approximation for two-electron repulsion integrals (ERIs). The algorithm bypasses the storage of four-center ERIs by adopting an integral-direct strategy. The CCSD amplitude equations are given in a compact quasi-linear form by factorizing them in terms of amplitude-dressed three-center intermediates. A hybrid MPI/OpenMP parallelization scheme is employed, which uses the OpenMP-based shared memory model for intranode parallelization and the MPI-based distributed memory model for internode parallelization. Parallel efficiency has been optimized for all terms in the CCSD amplitude equations. Two different algorithms have been implemented for the rate-limiting terms in the CCSD amplitude equations that entail and -scaling computational costs, where N O and N V denote the number of correlated occupied and virtual orbitals, respectively. One of the algorithms assembles the four-center ERIs requiring N V 4 and N O 2 N V 2 -scaling memory costs in a distributed manner on a number of MPI ranks, while the other algorithm completely bypasses the assembling of quartic memory-scaling ERIs and thus largely reduces the memory demand. It is demonstrated that the former memory-expensive algorithm is faster on a few hundred cores, while the latter memory-economic algorithm shows a better strong scaling in the limit of a few thousand cores. The program is shown to exhibit a near-linear scaling, in particular for the compute-intensive triples correction step, on up to 8000 cores. The performance of the program is demonstrated via calculations involving molecules with 24–51 atoms and up to 1624 atomic basis functions. As the first application, the complete basis set (CBS) limit for the interaction energy of the π-stacked uracil dimer from the S66 data set has been investigated. This work reports the first calculation of the interaction energy at the CCSD(T)/aug-cc-pVQZ level without local orbital approximation. The CBS limit for the CCSD correlation contribution to the interaction energy was found to be -8.01 kcal/mol, which agrees very well with the value -7.99 kcal/mol reported by Schmitz, Hättig, and Tew [ Phys. Chem. Chem. Phys. 2014 , 16 , 22167-22178]. The CBS limit for the total interaction energy was estimated to be -9.64 kcal/mol.
In the area of power electronics control, Field Programmable Gate Arrays (FPGAs) have the capability to outperform their Digital Signal Processor (DSP) counterparts due to the FPGA’s ability to implement true parallel processing and therefore facilitate higher switching frequencies, higher control bandwidth, and/or enhanced functionality. National Instruments (NI) has developed two platforms, Compact RIO (cRIO) and Single Board RIO (sbRIO), which combine a real-time processor with an FPGA. The FPGA can be programmed with a subset of the well-known LabVIEW graphical programming language. The use of cRIO and sbRIO for power electronics control has developed over the last few years to include control of three-phase inverters. Most three-phase inverter topologies include three switching legs. The addition of a fourth-leg to natively generate the neutral connection allows the inverter to serve single-phase loads in a microgrid or stand-alone power system and to balance the three-phase voltages in the presence of significant load imbalance. However, the control of a four-leg inverter is much more complex. In particular, instead of standard two-dimensional space vector modulation (SVM), the inverter requires three-dimensional space vector modulation (3D-SVM). The candidate software implements complete control algorithms in LabVIEW FPGA for a three-phase four-leg inverter. The software includes feedback control loops, three-dimensional space vector modulation gate-drive algorithms, advanced alarm handling capabilities, contactor control, power measurements, and debugging and tuning tools. The feedback control loops allow inverter operation in AC voltage control, AC current control, or DC bus voltage control modes based on external mode selection by a user or supervisory controller. The software includes the ability to synchronize its AC output to the grid or other voltage-source before connection. The software also includes provisions to allow inverter operation in parallel with other voltage regulating devices on the AC or DC buses. This flexibility allows the Inverter to operate as a stand-alone voltage source, connected to the grid, or in parallel with other controllable voltage sources as part of a microgrid or remote power system. In addition, as the inverter is expected to operate under severe unbalanced conditions, the software includes algorithms to accurately compute real and reactive power for each phase based on definitions provided in the IEEE Standard 1459: IEEE Standard Definitions for the Measurement of Electric Power Quantities Under Sinusoidal, Nonsinusoidal, Balanced, or Unbalanced Conditions. Finally, the software includes code to output analog signals for debugging and for tuning of control loops. The software fits on the Xilinx Virtex V LX110 FPGA embedded in the NI cRIO-9118 FPGA chassis, and with a 40 MHz base clock, supports a modulation update rate of 40 MHz, user-settable switching frequencies and synchronized control loop update rates of tens of kHz, and reference waveform generation, including Phase Lock Loop (PLL), update rate of 100 kHz.
Residential and commercial buildings in China, India, the United States (US), United Kingdom (UK), and Italy consume 39-45% of each nation's primary energy. Building energy models can be used to automatically optimize the return-on-investment for retrofits to improve a building’s energy efficiency. However, with an average of 3,000 building descriptors necessary to accurately simulate a single building, there is a market need to reduce the transaction cost for creating a more robust model for simulating every building in a city and accurately estimate savings prior to capital expenditures.We used two of the world’s fastest supercomputers, assembled unique datasets, and developed innovative algorithms for big data mining to assess different methods to create an accurate building energy model. The team has leveraged a total of eight high performance computing resources and 218 servers to analyze the best methods, metrics, and algorithms for creating an accurate building energy model beyond current industry standards necessary for private-sector financing. The project developed the world’s fastest buildings simulator, has completed over 8 million simulations totaling over 300TB, and mined this data with over 130,000 parallel artificial intelligence algorithms. This was used to quantify accuracy of AI-developed calibration algorithms with results that surpass industry standard guidelines and can identify individual building parameters to between 15% and 32% of their actual value.
Viskores is a scientific visualization library that is the primary deployment of such algorithms to the parallel accelerated processors of modern DOE supercomputers. In this paper, we review the capabilities provided by Viskores and how these capabilities are leveraged by other software in the high-performance computing ecosystem. We discuss the Viskores data representation and pay particular attention to array management. Through this array management we describe how data is adapted between Viskores and other software along with strategies for converting dynamic, polymorphic objects to static representations better suited to GPU processing. We conclude with several examples of Viskores integrating with high-performance software that is used in production today.
Extremely large scale analysis is becoming increasingly important as supercomputers and their simulations move from petascale to exascale. The lack of dedicated hardware acceleration for rendering on today's supercomputing platforms motivates our detailed evaluation of the possibility of interactive rendering on the supercomputer. In order to facilitate our understanding of rendering on the supercomputing platform, we focus on scalability of rendering algorithms and architecture envisioned for exascale datasets. To understand tradeoffs for dealing with extremely large datasets, we compare three different rendering algorithms for large polygonal data: software based ray tracing, software based rasterization and hardware accelerated rasterization. We present a case study of strong and weak scaling of rendering extremely large data on both GPU and CPU based parallel supercomputers using Para View, a parallel visualization tool. Wc use three different data sets: two synthetic and one from a scientific application. At an extreme scale, algorithmic rendering choices make a difference and should be considered while approaching exascale computing, visualization, and analysis. We find software based ray-tracing offers a viable approach for scalable rendering of the projected future massive data sizes.
This report aims to empirically understand the limits of machine learning when applied to Big Data. We observe that recent innovations in being able to collect, access, organize, integrate, and query massive amounts of data from a wide variety of data sources have brought statistical data mining and machine learning under more scrutiny, evaluation and application for gleaning insights from the data than ever before. Much is expected from algorithms without understanding their limitations at scale while dealing with massive datasets. In that context, we pose and address the following questions How does a machine learning algorithm perform on measures such as accuracy and execution time with increasing sample size and feature dimensionality? Does training with more samples guarantee better accuracy? How many features to compute for a given problem? Do more features guarantee better accuracy? Do efforts to derive and calculate more features and train on larger samples worth the effort? As problems become more complex and traditional binary classification algorithms are replaced with multi-task, multi-class categorization algorithms do parallel learners perform better? What happens to the accuracy of the learning algorithm when trained to categorize multiple classes within the same feature space? Towards finding answers to these questions, we describe the design of an empirical study and present the results. We conclude with the following observations (i) accuracy of the learning algorithm increases with increasing sample size but saturates at a point, beyond which more samples do not contribute to better accuracy/learning, (ii) the richness of the feature space dictates performance - both accuracy and training time, (iii) increased dimensionality often reflected in better performance (higher accuracy in spite of longer training times) but the improvements are not commensurate the efforts for feature computation and training and (iv) accuracy of the learning algorithms drop significantly with multi-class learners training on the same feature matrix and (v) learning algorithms perform well when categories in labeled data are independent (i.e., no relationship or hierarchy exists among categories).
The design and implementation of a new framework for adaptive mesh refinement calculations are described. It is intended primarily for applications in astrophysical fluid dynamics, but its flexible and modular design enables its use for a wide variety of physics. The framework works with both uniform and nonuniform grids in Cartesian and curvilinear coordinate systems. It adopts a dynamic execution model based on a simple design called a "task list" that improves parallel performance by overlapping communication and computation, simplifies the inclusion of a diverse range of physics, and even enables multiphysics models involving different physics in different regions of the calculation. We describe physics modules implemented in this framework for both nonrelativistic and relativistic magnetohydrodynamics (MHD). These modules adopt mature and robust algorithms originally developed for the Athena MHD code and incorporate new extensions: support for curvilinear coordinates, higher-order time integrators, more realistic physics such as a general equation of state, and diffusion terms that can be integrated with super-time-stepping algorithms. The modules show excellent performance and scaling, with well over 80% parallel efficiency on over half a million threads. The source code has been made publicly available.
This paper presents a parallel log-domain particle filtering algorithm combined with gamma spectrum unfolding to perform localization, identification, and evaluation of multiple point sources of various isotopes in an environment with attenuating obstacles. The method uses sets of precomputed attenuation kernels that map the attenuation characteristics of the environment. These kernels are specific to the energy level of a photopeak of interest. The spectral measurements are deconvolved into count measurements of each photopeak. These count measurements are fed into a set of parallel particle filters using attenuation kernels computed for that photopeak’s energy level. The individual regularized particle filters perform all likelihood calculations in the logarithmic domain to mitigate the effects of particle degeneracy. The output of each particle filter is combined to estimate which isotopes are present as well as their positions and strengths. The performance of the algorithm is characterized in a lab-scale environment using a mobile robot equipped with a gamma ray spectrometer in the presence of up to three different radioactive isotopes simultaneously. The sources were localized to within 10 cm, and their strengths were estimated within 10% of their true values. Furthermore, the isotopes were all correctly identified, and no spurious sources were reported.
More accurate than mean-field methods and more scalable than quantum chemical methods, continuum quantum Monte Carlo (QMC) is an invaluable tool for predicting the properties of matter from fundamental principles. Because QMC algorithms offer multiple forms of parallelism, they're ideal candidates for acceleration in the many-core paradigm.
This article describes the development of a flexible grazing-incidence small-angle X-ray scattering (GISAXS) simulation code in the framework of the distorted wave Born approximation that effectively utilizes the parallel processing power provided by graphics processors and multicore processors. The code, entitled High-Performance GISAXS , computes the GISAXS image for any given superposition of user-defined custom shapes or morphologies in a material and for various grazing-incidence angles and sample orientations. These capabilities permit treatment of a wide range of possible sample structures, including multilayered polymer films and nanoparticles on top of or embedded in a substrate or polymer film layers. In cases where the material displays regions of significant refractive index contrast, an algorithm has been implemented to perform a slicing of the sample and compute the averaged refractive index profile to be used as the reference geometry of the unperturbed system. A number of case studies are presented, which demonstrate good agreement with the experimental data for a variety of polymer and hybrid polymer/nanoparticle composite materials. The parallelized simulation code is well suited for addressing the analysis efforts required by the increasing amounts of GISAXS data being produced by high-speed detectors and ultrafast light sources.
Parallel Kinetic Monte Carlo (KMC) is a potent tool to simulate stochastic particle systems efficiently. However, despite literature on quantifying domain decomposition errors of the particle system for this class of algorithms in the short and in the long time regime, no study yet explores and quantifies the loss of time-reversibility in Parallel KMC. Inspired by concepts from non-equilibrium statistical mechanics, we propose the entropy production per unit time, or entropy production rate, given in terms of an observable and a corresponding estimator, as a metric that quantifies the loss of reversibility. Typically, this is a quantity that cannot be computed explicitly for Parallel KMC, which is why we develop a posteriori estimators that have good scaling properties with respect to the size of the system. Through these estimators, we can connect the different parameters of the scheme, such as the communication time step of the parallelization, the choice of the domain decomposition, and the computational schedule, with its performance in controlling the loss of reversibility. From this point of view, the entropy production rate can be seen both as an information criterion to compare the reversibility of different parallel schemes and as a tool to diagnose reversibility issues with a particular scheme. As a demonstration, we use Sandia Lab's SPPARKS software to compare different parallelization schemes and different domain (lattice) decompositions.
Abstract not provided
The MCNP6.1 Monte Carlo code, released in 2013, offers many features not available in previous versions, but runs 20-30% slower for most problems and 2-5x slower for some problems. The MCNP 2020 initiative was established in mid-2013 to address code performance and other issues, with 3 basic components: (1) Improve MCNP performance, with the specific goal of achieving a 2x performance gain within 2 years; (2) Upgrade the core MCNP software and algorithms in an evolutionary fashion, with the goal of achieving more agile and sustainable code; (3) Prepare for future exascale systems that will include massive parallelism and limited memory per core, with the goal of providing flexible and adaptable coding for possible exascale systems. This talk reviews the initial performance improvements to MCNP6.1 that have been incorporated into the 2014 update, MCNP6.1.1. The performance improvements to date have included both classic code optimizations and algorithm improvements. Testing on a variety of problems has demonstrated that the performance improvements were effective, yielding speedups by factors of 1.2x - 4x compared to MCNP6.1, depending on the type of problem. For criticality problems, MCNP6.1.1 runs 1.5x - 1.7x faster than MCNP6.1. Much more work is planned for improving MCNP6 performance, structure, and algorithms.
Metagenome assembly is the process of transforming a set of short, overlapping, and potentially erroneous DNA segments from environmental samples into the accurate representation of the underlying microbiomes's genomes. State-of-the-art tools require big shared memory machines and cannot handle contemporary metagenome datasets that exceed Terabytes in size. In this paper, we introduce the MetaHipMer pipeline, a high-quality and high-performance metagenome assembler that employs an iterative de Bruijn graph approach. MetaHipMer leverages a specialized scaffolding algorithm that produces long scaffolds and accommodates the idiosyncrasies of metagenomes. MetaHipMer is end-to-end parallelized using the Unified Parallel C language and therefore can run seamlessly on shared and distributed-memory systems. Experimental results show that MetaHipMer matches or outperforms the state-of-the-art tools in terms of accuracy. Ultimately, MetaHipMer scales efficiently to large concurrencies and is able to assemble previously intractable grand challenge metagenomes. We demonstrate the unprecedented capability of MetaHipMer by computing the first full assembly of the Twitchell Wetlands dataset, consisting of 7.5 billion reads - size 2.6 TBytes.
Coupled-cluster methods provide highly accurate models of molecular structure by explicit numerical calculation of tensors representing the correlation between electrons. These calculations are dominated by a sequence of tensor contractions, motivating the development of numerical libraries for such operations. While based on matrix-matrix multiplication, these libraries are specialized to exploit symmetries in the molecular structure and in electronic interactions, and thus reduce the size of the tensor representation and the complexity of contractions. The resulting algorithms are irregular and their parallelization has been previously achieved via the use of dynamic scheduling or specialized data decompositions. We introduce our efforts to extend the Libtensor framework to work in the distributed memory environment in a scalable and energy efficient manner. We achieve up to 240 speedup compared with the best optimized shared memory implementation. We attain scalability to hundreds of thousands of compute cores on three distributed-memory architectures, (Cray XC30&XC40, BlueGene/Q), and on a heterogeneous GPU-CPU system (Cray XK7). As the bottlenecks shift from being compute-bound DGEMM's to communication-bound collectives as the size of the molecular system scales, we adopt two radically different parallelization approaches for handling load-imbalance. Nevertheless, we preserve a uni ed interface to both programming models to maintain the productivity of computational quantum chemists.
The idea of attacking the phase problem by crowdsourcing is introduced. Using an interactive, multi-player, web-based system, participants work simultaneously to select phase sets that correspond to better electron-density maps in order to solve low-resolution phasing problems. The human mind innately excels at some complex tasks that are difficult to solve using computers alone. For complex problems amenable to parallelization, strategies can be developed to exploit human intelligence in a collective form: such approaches are sometimes referred to as ‘crowdsourcing’. Here, a first attempt at a crowdsourced approach for low-resolution ab initio phasing in macromolecular crystallography is proposed. A collaborative online game named CrowdPhase was designed, which relies on a human-powered genetic algorithm, where players control the selection mechanism during the evolutionary process. The algorithm starts from a population of ‘individuals’, each with a random genetic makeup, in this case a map prepared from a random set of phases, and tries to cause the population to evolve towards individuals with better phases based on Darwinian survival of the fittest. Players apply their pattern-recognition capabilities to evaluate the electron-density maps generated from these sets of phases and to select the fittest individuals. A user-friendly interface, a training stage and a competitive scoring system foster a network of well trained players who can guide the genetic algorithm towards better solutions from generation to generation via gameplay. CrowdPhase was applied to two synthetic low-resolution phasing puzzles and it was shown that players could successfully obtain phase sets in the 30° phase error range and corresponding molecular envelopes showing agreement with the low-resolution models. The successful preliminary studies suggest that with further development the crowdsourcing approach could fill a gap in current crystallographic methods by making it possible to extract meaningful information in cases where limited resolution might otherwise prevent initial phasing.
In this paper, we develop new techniques for solving the large, coupled linear systems that arise from fully implicit Runge–Kutta methods. This method makes use of the iterative preconditioned GMRES algorithm for solving the linear systems, which has seen success for fluid flow problems and discontinuous Galerkin discretizations. By transforming the resulting linear system of equations, one can obtain a method which is much less computationally expensive than the untransformed formulation, and which compares competitively with other time-integration schemes, such as diagonally implicit Runge–Kutta (DIRK) methods. We develop and test several ILU-based preconditioners effective for these large systems. We additionally employ a parallel-in-time strategy to compute the Runge–Kutta stages simultaneously. Numerical experiments are performed on the Navier–Stokes equations using Euler vortex and 2D and 3D NACA airfoil test cases in serial and in parallel settings. In conclusion, the fully implicit Radau IIA Runge–Kutta methods compare favorably with equal-order DIRK methods in terms of accuracy, number of GMRES iterations, number of matrix–vector multiplications, and wall-clock time, for a wide range of time steps.