Search NASA⌕ Search

SEARCH · Search NASA

Results for “Integer Optimization”

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 865 records · Page 48

fPINNs: Fractional Physics-Informed Neural Networks

Physics-informed neural networks (PINNs), introduced, are effective in solving integer-order partial differential equations (PDEs) based on scattered and noisy data. PINNs employ standard feedforward neural networks (NNs) with the PDEs explicitly encoded into the NN using automatic differentiation, while the sum of the mean-squared PDE residuals and the mean-squared error in initial-boundary conditions is minimized with respect to the NN parameters. Here we extend PINNs to fractional PINNs (fPINNs) to solve space-time fractional advection-diffusion equations (fractional ADEs), and we study systematically their convergence, hence explaining both fPINNs and PINNs for the first time. Specifically, we demonstrate their accuracy and effectiveness in solving multidimensional forward and inverse problems with forcing terms whose values are only known at randomly scattered spatio-temporal coordinates (black-box (BB) forcing terms). A novel element of the fPINNs is the hybrid approach that we introduce for constructing the residual in the loss function using both automatic differentiation for the integer-order operators and numerical discretization for the fractional operators. This approach bypasses the difficulties stemming from the fact that automatic differentiation is not applicable to fractional operators because the standard chain rule in integer calculus is not valid in fractional calculus. To discretize the fractional operators, we employ the Grünwald--Letnikov (GL) formula in one-dimensional fractional ADEs and the vector GL formula in conjunction with the directional fractional Laplacian in two- and three-dimensional fractional ADEs. We first consider the one-dimensional fractional Poisson equation and compare the convergence of the fPINNs against the finite difference method (FDM). We present the solution convergence using both the mean L 2 error as well as the standard deviation due to sensitivity to NN parameter initializations. Using different GL formulas we observe first-, second-, and third-order convergence rates for small size training sets but the error saturates for larger training sets. We explain these results by analyzing the four sources of numerical errors due to discretization, sampling, NN approximation, and optimization. The total error decays monotonically (below 10 –5 for a third-order GL formula) but it saturates beyond that point due to the optimization error. We also analyze the relative balance between discretization and sampling errors and observe that the sampling size and the number of discretization points (auxiliary points) should be comparable to achieve the highest accuracy. As we increase the depth of the NN up to certain value, the mean error decreases and the standard deviation increases, whereas the width has essentially no effect unless its value is either too small or too large. We next consider time-dependent fractional ADEs and compare white-box (WB) and BB forcing. We observe that for the WB forcing, our results are similar to the aforementioned cases; however, for the BB forcing fPINNs outperform FDM. Subsequently, we consider multidimensional time-, space-, and space-time-fractional ADEs using the directional fractional Laplacian and we observe relative errors of 10 –3 ~ 10 –4 . Lastly, we solve several inverse problems in one, two, and three dimensions to identify the fractional orders, diffusion coefficients, and transport velocities and obtain accurate results given proper initializations even in the presence of significant noise.

97 MATHEMATICS AND COMPUTING↗

Exploring Integer Sum Reduction using Atomics on Intel CPU

Atomic functions are useful in updating a shared variable by multiple threads, barrier synchronizations, constructing complex data structures, and building high-level frameworks. In this paper, we focus on the evaluation and analysis of integer sum reduction, a common data parallel primitive. We convert the sequential reduction into parallel OpenCL implementations on the CPU. We also develop three micro kernels, which allow us to understand the relationships between the kernel performance and the operations involved in reduction. The results of the micro kernels show that increasing the work-group size linearly can linearly improve the kernel performance. There is a sweet spot in the relationship between the work-group size and barrier synchronization overhead. The performance of the atomics over local memory are not sensitive to the work-group size. The sum reduction kernel with vectorized memory accesses can improve the performance of the baseline kernel for a wide range of work-group sizes. However, the vectorization efficiency shrinks with the growing work-group size. We also find that the vendor’s default OpenCL kernel optimization does not improve the kernel performance. On average, disabling the optimization can reduce the execution time of the kernel with vectorized memory accesses by 15%. We attribute the performance drop to the fact that the default kernel optimizations instantiate a large number of atomics over global memory when implicitly vectorizing the kernel computation.

atomics↗

Efficient and Collision-Free Anticipative Cruise Control in Randomly Mixed Strings

Connected intelligent vehicle following can improve safety and efficiency compared to today's road transport, but real traffic in the near future will not provide an ideal setting for its deployment. Unconnected human-driven vehicles will follow variable behavior patterns that combine with differences in dynamic capability to create heterogeneous scenarios. In this paper, connected automated vehicles employ model predictive control for following in traffic that may include both heavy and passenger vehicles at quasi-random positions. Adding further realism, some quasi-randomly mixed vehicles are completely unconnected and equipped with the reactive intelligent driver model using pseudorandom parameters. A mixed-integer quadratic programming formulation adapts the predictive algorithm to diverse powertrain operating point constraints. The preceding vehicle's control input is estimated from velocity and brake light observations and used to probabilistically generate a preview for the ego vehicle. Here, a terminal constraint designed using particle kinematics prevents collisions due to shortsightedness. Results are simulated at various heavy and predictive vehicle concentrations in the presence of packet loss. Linear fuel economy improvements between 1.4% and 1.9% per 10% increase in predictive vehicle penetration rate are shown.

33 ADVANCED PROPULSION SYSTEMS↗

Water Security Toolkit User Manual Version 1.2.

The Water Security Toolkit (WST) is a suite of open source software tools that can be used by water utilities to create response strategies to reduce the impact of contamination in a water distribution network . WST includes hydraulic and water quality modeling software , optimizati on methodologies , and visualization tools to identify: (1) sensor locations to detect contamination, (2) locations in the network in which the contamination was introduced, (3) hydrants to remove contaminated water from the distribution system, (4) locations in the network to inject decontamination agents to inactivate, remove, or destroy contaminants, (5) locations in the network to take grab sample s to help identify the source of contamination and (6) valves to close in order to isolate contaminate d areas of the network. This user manual describes the different components of WST , along w ith examples and case studies. License Notice The Water Security Toolkit (WST) v.1.2 Copyright c 2012 Sandia Corporation. Under the terms of Contract DE-AC04-94AL85000, there is a non-exclusive license for use of this work by or on behalf of the U.S. government. This software is distributed under the Revised BSD License (see below). In addition, WST leverages a variety of third-party software packages, which have separate licensing policies: Acro Revised BSD License argparse Python Software Foundation License Boost Boost Software License Coopr Revised BSD License Coverage BSD License Distribute Python Software Foundation License / Zope Public License EPANET Public Domain EPANET-ERD Revised BSD License EPANET-MSX GNU Lesser General Public License (LGPL) v.3 gcovr Revised BSD License GRASP AT&T Commercial License for noncommercial use; includes randomsample and sideconstraints executable files LZMA SDK Public Domain nose GNU Lesser General Public License (LGPL) v.2.1 ordereddict MIT License pip MIT License PLY BSD License PyEPANET Revised BSD License Pyro MIT License PyUtilib Revised BSD License PyYAML MIT License runpy2 Python Software Foundation License setuptools Python Software Foundation License / Zope Public License six MIT License TinyXML zlib License unittest2 BSD License Utilib Revised BSD License virtualenv MIT License Vol Common Public License vpykit Revised BSD License Additionally, some precompiled WST binary distributions might bundle other third-party executables files: Coliny Revised BSD License (part of Acro project) Dakota GNU Lesser General Public License (LGPL) v.2.1 PICO Revised BSD License (part of Acro project) i Revised BSD License Redistribution and use in source and binary forms, with or without modification, are permitted provided that the following conditions are met: * Redistributions of source code must retain the above copyright notice, this list of conditions and the following disclaimer. * Redistributions in binary form must reproduce the above copyright notice, this list of conditions and the following disclaimer in the documentation and/or other materials provided with the distribution. * Neither the name of Sandia National Laboratories nor Sandia Corporation nor the names of its con- tributors may be used to endorse or promote products derived from this software without specific prior written permission. THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IM- PLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL SANDIA CORPORATION BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUD- ING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. ii Acknowledgements This work was supported by the U.S. Environmental Protection Agency through its Office of Research and Development (Interagency Agreement # DW8992192801). The material in this document has been subject to technical and policy review by the U.S. EPA, and approved for publication. The views expressed by individual authors, however, are their own, and do not necessarily reflect those of the U.S. Environmental Protection Agency. Mention of trade names, products, or services does not convey official U.S. EPA approval, endorsement, or recommendation. The Water Security Toolkit is an extension of the Threat Ensemble Vulnerability Assessment-Sensor Place- ment Optimization Tool (TEVA-SPOT), which was also developed with funding from the U.S. Environ- mental Protection Agency through its Office of Research and Development (Interagency Agreement # DW8992192801). The authors acknowledge the following individuals for their contributions to the devel- opment of TEVA-SPOT: Jonathan Berry (Sandia National Laboratories), Erik Boman (Sandia National Laboratories), Lee Ann Riesen (Sandia National Laboratories), James Uber (University of Cincinnati), and Jean-Paul Watson (Sandia National Laboratories). iii Acronyms ATUS American Time-Use Survey BLAS Basic linear algebra sub-routines CFU Colony-forming unit CVAR Conditional value at risk CWS Contamination warning system EA Evolutionary algorithm EDS Event detection system EPA U.S. Environmental Protection Agency EC Extent of Contamination ERD EPANET results database file GLPK GNU Linear Programming Kit GRASP Greedy randomized adaptive sampling process HEX Hexadecimal HTML HyperText markup language INP EPANET input file LP Linear program MC Mass consumed MILP Mixed integer linear program MIP Mixed integer program MSX Multi-species extension for EPANET NFD Number of failed detections NS Number of sensors NZD Non-zero demand PD Population dosed PE Population exposed PK Population killed TAI Threat assessment input file TCE Tailed-conditioned expectation TD Time to detection TEC Timed extent of contamination TEVA Threat ensemble vulnerability assessment TSB Tryptic soy broth TSG Threat scenario generation file TSI Threat simulation input file VAR Value at risk VC Volume consumed WST Water Security Toolkit YML YAML configuration file format for WST iv Symbols Notation Definition Example { , } set brackets { 1,2,3 } means a set containing the values 1,2, and 3. [?] is an element of s [?] S means that s is an element of the set S . [?] for all s = 1 [?] s [?] S means that the statement s = 1 is true for all s in set S . P summation P n i =1 s i means s 1 + s 2 + * * * + s n . \ set minus S \ T means the set that contains all those elements of S that are not in set T . %7C given %7C is used to define conditional probability. P ( s %7C t ) means the prob- ability of s occurring given that t occurs. %7C ... %7C cardinality Cardinality of a set is the number of elements of the set. If set S = { 2,4,6 } , then %7C S %7C = 3. v

Klise, Katherine A.↗

LONG-TIMESCALE BEHAVIOR OF THE BLAZHKO EFFECT FROM RECTIFIED KEPLER DATA

In order to benefit from the four year unprecedented precision of the Kepler data, we extracted light curves from the pixel photometric data of the Kepler space telescope for 15 Blazhko RR Lyrae stars. To collect all the flux from a given target as accurately as possible, we defined tailor-made apertures for each star and quarter. In some cases, the aperture finding process yielded sub-optimal results, because some flux have been lost even if the aperture contains all available pixels around the star. This fact stresses the importance of those methods that rely on the whole light curve instead of focusing on the extrema (O – C diagrams and other amplitude independent methods). We carried out detailed Fourier analysis of the light curves and the amplitude independent O – C diagram. We found 12 (80%) multiperiodically modulated stars in our sample. This ratio is much higher than previously found. Resonant coupling between radial modes, a recent theory for explaining the Blazhko effect, allows single, multiperiodic, or even chaotic modulations. Among the stars with two modulations, we found three stars (V355 Lyr, V366 Lyr, and V450 Lyr) where one of the periods dominates in amplitude modulation, but the other period has a larger frequency modulation amplitude. The ratio between the primary and secondary modulation periods is almost always very close to the ratios of small integer numbers. It may indicate the effect of undiscovered resonances. Furthermore, we detected the excitation of the second radial overtone mode f {sub 2} for three stars where this feature was formerly unknown. Our data set comprises the longest continuous, most precise observations of Blazhko RR Lyrae stars ever published. These data are made publicly available and will be unrivaled for years to come.

79 ASTRONOMY AND ASTROPHYSICS↗

Stochastic Microgrid Scheduling With Chance‐Constrained Resilience Consideration

Traditionally, it is assumed that microgrids transition seamlessly from grid‐connected operation to islanded mode in the event of sudden main grid outages. In reality, the islanding process, especially unintentional islanding, is rarely seamless. Instead, it is subject to voltage and frequency fluctuations caused by the instantaneous disconnection of the point of common coupling (PCC) switch, variations in loads and renewable generation output and even the protection tripping of distributed energy resources (DERs). To mitigate these fluctuations and facilitate a smooth islanding process, we propose a stochastic microgrid scheduling model that incorporates chance‐constrained resilience measures. Specifically, the resilience measure is defined as the probability of successful islanding (PSI), that is, the probability that a microgrid can mitigate the generation‐demand imbalance caused by the disconnection of the PCC switch, variations in load and renewable generation and DER tripping. This measure is modelled using chance constraints. Unlike existing reliability and resilience indices, which typically neglect the possibility of microgrid/DER failure under extreme events and assume their survival while primarily focussing on reducing impact duration or magnitude, the proposed PSI‐based framework explicitly addresses microgrid and DER survival during the islanding transition. The formulated nonlinear chance constraints are approximated using a multiinterval approach and equivalently represented as a mixed‐integer linear programming (MILP) formulation. Case study results validate the proposed method, showing that the PSI estimation error is reduced to less than 8%, compared to approximately 28% with existing methods. Various sensitivity analyses on the DER tripping rate and PSI settings were performed to validate the robustness of the proposed method. In particular, the necessity of accounting for DER tripping in the PSI calculation was demonstrated.

chance constrained optimization↗

An interregional optimization approach for time series aggregation in continent-scale electricity system models

Modeling electric power systems with high shares of weather-dependent resources requires tradeoffs between temporal, spatial, and operational resolution. Many studies perform time series aggregation using clustering algorithms to reduce the temporal dimension, but when modeling continent-scale electricity systems that are large enough to contain multiple independent weather systems, this approach requires large numbers of representative periods to minimize errors in regional wind and solar capacity factors. Here, a new optimization-based approach for representative period selection and weighting is introduced that minimizes regional errors in average renewable capacity factors and electricity demand. The method delivers higher regional fidelity with fewer representative periods than alternative clustering methods when applied to wind, solar, and demand profiles for the contiguous United States. When representative periods are selected from multiple weather years, the optimized method reproduces regional averages with lower error than a complete 365-day time series from any single weather year. The method identifies only representative (as opposed to outlying) periods but can be combined with an iterative "stress period" identification approach to guide efficient decision-making considering both average and high-risk weather conditions.

24 POWER TRANSMISSION AND DISTRIBUTION↗

Multiphase complete exchange on a circuit switched hypercube

On a distributed memory parallel computer, the complete exchange (all-to-all personalized) communication pattern requires each of n processors to send a different block of data to each of the remaining n - 1 processors. This pattern is at the heart of many important algorithms, most notably the matrix transpose. For a circuit switched hypercube of dimension d(n = 2(sup d)), two algorithms for achieving complete exchange are known. These are (1) the Standard Exchange approach that employs d transmissions of size 2(sup d-1) blocks each and is useful for small block sizes, and (2) the Optimal Circuit Switched algorithm that employs 2(sup d) - 1 transmissions of 1 block each and is best for large block sizes. A unified multiphase algorithm is described that includes these two algorithms as special cases. The complete exchange on a hypercube of dimension d and block size m is achieved by carrying out k partial exchange on subcubes of dimension d(sub i) Sigma(sup k)(sub i=1) d(sub i) = d and effective block size m(sub i) = m2(sup d-di). When k = d and all d(sub i) = 1, this corresponds to algorithm (1) above. For the case of k = 1 and d(sub i) = d, this becomes the circuit switched algorithm (2). Changing the subcube dimensions d, varies the effective block size and permits a compromise between the data permutation and block transmission overhead of (1) and the startup overhead of (2). For a hypercube of dimension d, the number of possible combinations of subcubes is p(d), the number of partitions of the integer d. This is an exponential but very slowly growing function and it is feasible over these partitions to discover the best combination for a given message size. The approach was analyzed for, and implemented on, the Intel iPSC-860 circuit switched hypercube. Measurements show good agreement with predictions and demonstrate that the multiphase approach can substantially improve performance for block sizes in the 0 to 160 byte range. This range, which corresponds to 0 to 40 floating point numbers per processor, is commonly encountered in practical numeric applications. The multiphase technique is applicable to all circuit-switched hypercubes that use the common e-cube routing strategy.

Bokhari, Shahid H.↗

ORBITAL MOTION OF HR 8799 b, c, d USING HUBBLE SPACE TELESCOPE DATA FROM 1998: CONSTRAINTS ON INCLINATION, ECCENTRICITY, AND STABILITY

HR 8799 is currently the only multiple-planet system that has been detected with direct imaging, with four giant planets of masses 7-10 M{sub Jup} orbiting at large separations (15-68 AU) from this young late A star. Orbital motion provides insight into the stability and possible formation mechanisms of this planetary system. Dynamical studies can also provide constraints on the planets' masses, which help calibrate evolutionary models, yet measuring the orbital motion is a very difficult task because the long-period orbits (50-500 yr) require long time baselines and high-precision astrometry. This paper studies the three planets HR 8799b, c, and d in the archival data set of HR 8799 obtained with the Hubble Space Telescope (HST) NICMOS coronagraph in 1998. The detection of all three planets is made possible by a careful optimization of the Locally Optimized Combination of Images algorithm, and we used a statistical analysis of a large number of reduced images. This work confirms previous astrometry for planet b and presents new detections and astrometry for planets c and d. These HST images provide a ten-year baseline with the discovery images from 2008, and therefore offer a unique opportunity to constrain their orbital motion now. Recent dynamical studies of this system show the existence of a few possible stable solutions involving mean motion resonances (MMRs), where the interaction between c and d plays a major role. We study the compatibility of a few of these stable scenarios (1d:1c, 1d:2c, or 1d:2c:4d) with the new astrometric data from HST. In the hypothesis of a 1d:2c:4b MMR our best orbit fit is close to the stable solution previously identified for a three-planet system and involves low eccentricity for planet d (e{sub d} = 0.10) and moderate inclination of the system (i = 28.0 deg), assuming a coplanar system, circular orbits for b and c, and exact resonance with integer period ratios. Under these assumptions, we can place strong constraints on the inclination of the system (27.3-31.4 deg) and on the eccentricity for d e{sub d} < 0.46. Our results are robust to small departures from exact integer period ratios and consistent with previously published results based on dynamical studies for a three-planet system prior to the discovery of the fourth planet.

79 ASTRONOMY AND ASTROPHYSICS↗

OPERATIONAL TOPOLOGY ESTIMATION OF THREE PHASE UNBALANCED POWER DISTRIBUTION SYSTEMS WITH OUTAGES

Distribution grid, in general, is constructionally loopy but a radially operated network. Opening and closing of suitable tie-switches allows transition from one radial topology to another. Tie-switches ensure reliability and provide restoration paths when faults (outages) occur. The estimation of the system's current operational topology is a precursor to implementing any optimal control actions (during normal operation) or restorative actions (during outage condition). The existing literature in topology detection suffers with several limitations including: (1) computationally intensive nonlinear formulations that are not scalable for large distribution systems, (2) methods requiring exhaustive search for the correct topology by enumerating the complete topological search space, and (3) approaches requiring large volumes of time-series measurements and are not applicable for estimating continuously changing operational topology. Furthermore, the existing literature on distribution grid topology estimation cannot be extended to estimate the operational topology during outage conditions. For operational topology estimation during outages, our objective is to estimate both energized radial feeder and the outaged sections. Starting with a radial topology, these methods identify outaged feeder sections and hence, are inapplicable when the current radial operational topology is not known. In this work, we present a mathematical programming approach to estimate the operational topology of a three-phase unbalanced power distribution system under both outage and normal operating conditions. The aforementioned limitations of existing literature were countered by modeling the problem as a mixed-integer linear programming (MILP) problem which minimizes the errors in available measurements while satisfying the power flow and structural constraints of the system. Specically, a minimum weighted least absolute value estimator is proposed that uses the instantaneous line flow measurements, historical/ forecasted load data, and ping measurements from a limited set of smart meters and solves an MILP to identify the operational topology based on a linearized three-phase unbalanced power flow model. The proposed approach is tested on IEEE 123-bus feeder and 1069 bus multi-feeder test system and the numerical results prove the applicability of the proposed method to large three phase unbalanced grids with high errors in measurements. It is observed that the approach performs well even at high percentages of measurement errors.

GANDLURU NAGA, ANANDINI↗

Bi-objective approaches for home healthcare medical team planning and scheduling problem

In this paper, we study the medical team planning and scheduling in home healthcare within a weekly horizon, including the planning of nursing visits and the daily traveling route scheduling. Two objectives are considered: the first one is to minimize the total operation cost of the healthcare agency and the second one is to maximize the patient satisfaction. The problem is formulated as a mixed integer program, to seek for a trade-off between two objectives, with the characteristics of the patient requirements including nursing types, nursing frequency and service length considered. Medical team types, available service days and overwork penalty are also respected in the model. Then an ϵ-constraint method is adopted to obtain exact non-dominated solutions. To address large-scale problem instances, three heuristic approaches are developed to generate approximate Pareto fronts in a reasonable time. The set of non-dominated solutions are valuable for decision-making. Computational experiments are conducted and the results demonstrate the efficiency of the approaches.

97 MATHEMATICS AND COMPUTING↗

Some practical universal noiseless coding techniques, part 3, module PSl14,K+

The algorithmic definitions, performance characterizations, and application notes for a high-performance adaptive noiseless coding module are provided. Subsets of these algorithms are currently under development in custom very large scale integration (VLSI) at three NASA centers. The generality of coding algorithms recently reported is extended. The module incorporates a powerful adaptive noiseless coder for Standard Data Sources (i.e., sources whose symbols can be represented by uncorrelated non-negative integers, where smaller integers are more likely than the larger ones). Coders can be specified to provide performance close to the data entropy over any desired dynamic range (of entropy) above 0.75 bit/sample. This is accomplished by adaptively choosing the best of many efficient variable-length coding options to use on each short block of data (e.g., 16 samples) All code options used for entropies above 1.5 bits/sample are 'Huffman Equivalent', but they require no table lookups to implement. The coding can be performed directly on data that have been preprocessed to exhibit the characteristics of a standard source. Alternatively, a built-in predictive preprocessor can be used where applicable. This built-in preprocessor includes the familiar 1-D predictor followed by a function that maps the prediction error sequences into the desired standard form. Additionally, an external prediction can be substituted if desired. A broad range of issues dealing with the interface between the coding module and the data systems it might serve are further addressed. These issues include: multidimensional prediction, archival access, sensor noise, rate control, code rate improvements outside the module, and the optimality of certain internal code options.

Rice, Robert F.↗

HykSort: a new variant of hypercube quicksort on distributed memory architectures. In: ICS '13 Proceedings of the 27th international ACM conference on International conference on supercomputing

In this paper, we present HykSort, an optimized comparison sort for distributed memory architectures that attains more than 2× improvement over bitonic sort and samplesort. The algorithm is based on the hypercube quicksort, but instead of a binary recursion, we perform a k-way recursion in which the pivots are selected accurately with an iterative parallel select algorithm. The single-node sort is performed using a vectorized and multithreaded merge sort. The advantages of HykSort are lower communication costs, better load balancing, and avoidance of O(p)-collective communication primitives. We also present a staged communication samplesort, which is more robust than the original samplesort for large core counts. We conduct an experimental study in which we compare hypercube sort, bitonic sort, the original samplesort, the staged samplesort, and HykSort. We report weak and strong scaling results and study the effect of the grain size. It turns out that no single algorithm performs best and a hybridization strategy is necessary. As a highlight of our study, on our largest experiment on 262,144 AMD cores of the CRAY XK7 "Titan" platform at the Oak Ridge National Laboratory we sorted 8 trillion 32-bit integer keys in 37 seconds achieving 0.9TB/s effective throughput.

Sundar, Hari↗

Rapid Crossing of a Resonance During Tune Modulation by Synchrotron Oscillations

It had become a standard practice to constrain the particle’s tune footprint while designing the ring lattice so that the particle tunes fit between harmful resonances that limit ring dynamic aperture (DA). This approach, known as “tune confinement”, puts tight limits on the magnitude of the tune shifts with amplitude and with momentum. The latter requires labor-intensive optimization of the off-energy dynamic aperture and control over the corresponding tune footprint for the large energy deviations and is a prerequisite to achieving reasonable lifetime if it is dominated by the Touschek scattering. As nonlinearities of the modern ring lattices are much enhanced as compared with the previous generation of synchrotron light sources, it is becoming more and more difficult to keep the off-momentum tune footprint confined. While one could reach reasonable on-energy DA suitable for injection into the machine, the off-energy DA optimization is much more complex and needs to satisfy more constraints so to reach large energy acceptance and secure decent lifetime at the high ring current. In certain cases when the lattice solution with the confined off-energy tune footprint cannot be found, one may ask whether crossing of a resonance stopband from off-momentum particle will lead to loss of this particle. Intuitively, if the particle crosses the stopband quickly, one may expect that the betatron oscillation amplitude will not increase substantially thereby keeping the particle within the machine acceptance. At the same time the tight tolerances of the modern lattice elements as designed and produced afford much narrower resonance stopbands being compared with the machines built decades ago. Under these circumstances the question above may be asked as what are the conditions for operating a ring with a large chromatic tune footprint, which crosses resonances, while not observing particle losses. In this paper we try to answer the questions above by studying the model of a particle with large initial momentum deviation (as, for example, driven by the Touschek scattering event), which crosses stopband of a half-integer resonance while executing synchrotron oscillations in the highly chromatic ring lattice.

36 MATERIALS SCIENCE↗

Data Locality Enhancement of Dynamic Simulations for Exascale Computing (Final Report)

The development of modern processors exhibits two trends that complicate the optimizations of modern software. The first is the increasing sensitivity of processors' throughput to irregularities in computation. With more processors produced through a massive integration of simple cores, future systems will increasingly favor regular data-level parallel computations, but deviate from the needs of applications with complex patterns. Some evidences are already shown on Graphic Processing Units (GPU): Irregular data accesses (e.g., indirect references A[D[i]]) and conditional branches are limiting many GPU applications' performance at a level an order of magnitude lower than the peak of GPU. The second hardware trend is the growing gap between memory bandwidth and the aggregate speed—that is, the sum of all cores' computing power—of a Chip Multiprocessor (CMP). Despite the capped growth of the peak CPU speed, the aggregate speed of a CMP keeps increasing as more cores get into a single chip. It is expected that by 2018, node concurrency in an exascale system will increase by hundreds of times, whereas, memory bandwidth will expand by only 10 to 20 times. Consequently, data movement and storage is expected to consume more than 70% of the total system power. Bridging this gap is difficult; the complexities of modern CMP memory hierarchy make it even harder: Data cache becomes shared among computing units, and the sharing is often non-uniform—whether two computing units share a cache depends on their proximity and the level of the cache. On the recent IBM Power7 architecture, for instance, four hardware contexts (or SMT threads) in a core share the entire memory hierarchy, all cores in one chip share an on-chip L3 cache, and cores across chips share L3 and main memory through off-chip connections. These two trends complicate the translation of computing power into performance, especially for a program with either intensive data accesses or complex patterns in data accesses or control flow paths. Unfortunately, both attributes present and will persist in a class of important applications. For instance, many scientific simulations deal with a large volume of data. And meanwhile, as most real-world processes are non-uniform and evolving (e.g., the evolution of a galaxy or the process of a drug injection), both the computations and data accesses of these programs tend to be irregular and dynamically changing. Currently, the lack of support to these applications on modern CMP severely limits their performance. On GPU, as our recent study shows and other studies echo, performance enhancement of a factor of integers is possible when memory accesses or control flows are streamlined for a set of GPU applications. On multicore CPU, our studies show that traditional locality enhancement, for being oblivious to the new features of multicore memory hierarchy, may even cause large slowdown to data-intensive dynamic applications. The severity of the issues is expected to worsen as the two hardware trends continue. Some recent studies try to match software with the trends, but in a limited scope or manner. For irregularities on GPU, most studies focus on irregularities analyzable through static analysis (e.g., data accesses in regular loops). Dynamic irregularities are harder to address because the needed analysis and transformations typically have to happen at run time. Some other research resorts to hardware extensions, an actual adoption of which is unclear for the entailed space cost and complexity. For data locality, recent years have seen some exploitations of the new memory hierarchy on multicore for performance, but most of them are on process or thread scheduling, rather than program transformations. Our recent study reveals that program-level transformations may magnify the scheduling benefits by a factor of seven, concluding that program-level transformation should play a central role for data locality enhancement on modern CMP. But research in this direction has been sparse, and most have focused on data layout or cache performance modeling, rather than program transformations to match with the new memory hierarchy features. Overall, it is still an open question how to bridge the gap between dynamic computations and the two prominent properties of modern processors. The goal of this project is to develop a set of techniques and software tools to enhance the matching between memory accesses in dynamic simulations and the prominent features of modern and future CMP, alleviating the memory performance issues for petascale and exascale computing. This report summarizes the discoveries and products produced throughout this project. It includes free launch, a new software approach to overcoming the shortcomings of both methods; coherence-free multiview, an approach that allows multiple views of a single data object to co-exist on GPU memory during a GPU kernel execution; algorithmic optimizations to data analytics problems, especially those that involve lots of distance calculations. Supporting dynamic parallelism is important for GPU to benefit a broad range of applications. There are currently two fundamental ways for programs to exploit dynamic parallelism on GPU: a software-based approach with software-managed worklists, and a hardware-based approach through dynamic subkernel launches. Neither is satisfactory. The former is complicated to program and is often subject to some load imbalance; the latter suffers large runtime overhead. In this work, we propose free launch, a new software approach to overcoming the shortcomings of both methods. It allows programmers to use subkernel launches to express dynamic parallelism. It employs a novel compiler-based code transformation named subkernel launch removal to replace the subkernel launches with the reuse of parent threads. Coupled with an adaptive task assignment mechanism, the transformation reassigns the tasks in the subkernels to the parent threads with a good load balance. The technique requires no hardware extensions, immediately deployable on existing GPUs. It keeps the programming convenience of the subkernel launch-based approach while avoiding its large runtime overhead. Meanwhile, its superior load balancing makes it outperform manual worklist-based techniques by 3X on average. The work was published at the 48th Annual IEEE/ACM International Symposium on Microarchitecture (Micro'2015). A Graphic Processing Unit (GPU) system is typically equipped with many types of memory (e.g., global, constant, texture, shared, cache). Data placement determines what data are placed on which type of memory, essential for GPU memory performance. Prior optimizations of data placement always require a single view of a data object on memory, which limits the optimization effectiveness. In this work, we propose coherence-free multiview, an approach that allows multiple views of a single data object to co-exist on GPU memory during a GPU kernel execution. We demonstrate that under certain conditions, the multiple views can remain incoherent while facilitating enhanced data placement. We present a theorem and some compiler support to ensure the soundness of the usage of coherence-free multiview. We further develop reference-discerning data placement, a new way to enhance data placements on GPU. It enables more flexible data placements by using coherence-free multiview to leverage the slack in coherence requirement of some GPU programs. Experiments on three types of GPU systems show that, with less than 200KB space cost, the new data placement technique can provide a 1.6X average (up to 4.27X) speedup. In addition, we have examined algorithmic optimizations to data analytics problems, especially those that involve lots of distance calculations. Computing distances among data points is an essential part of many important algorithms in data analytics, graph analysis, and other domains. In each of these domains, developers have spent significant manual effort optimizing algorithms, often through novel applications of the triangle equality, in order to minimize the number of distance computations in the algorithms. In this work, we observe that many algorithms across these domains can be generalized as an instance of a generic distance-related abstraction. Based on this abstraction, we derive seven principles for correctly applying the triangular inequality to optimize distance-related algorithms. Guided by the findings, we develop Triangular OPtimizer (TOP), the first software framework that is able to automatically produce optimized algorithms that either matches or outperforms manually designed algorithms for solving distance-related problems. TOP achieves up to 237x speedups and 2.5X on average. The work has been published at the 32nd International Conference on Machine Learning (ICML'15) and the 41st International Conference on Very Large Data Bases (VLDB'15).

42 ENGINEERING↗

Security-Constrained Unit Commitment for Electricity Market: Modeling, Solution Methods, and Future Challenges

This paper summarizes the technical activities of the IEEE Task Force on Solving Large Scale Optimization Problems in Electricity Market and Power System Applications. This Task Force was established by the IEEE Technology and Innovation Subcommittee to first review the state-of-the-art of the security-constrained unit commitment (SCUC) business model, its mathematical formulation, and solution techniques in solving electricity market clearing problems. The Task Force then investigated the emerging challenges of future market clearing problems and presented efforts in building benchmark mathematical and business models.

24 POWER TRANSMISSION AND DISTRIBUTION↗

Probabilistic N - k failure-identification for power systems

This work considers a probabilistic generalization of the N - k failure-identification problem in power transmission networks, where the probability of failure of each component in the network is known a priori and the goal of the problem is to find a set of k components that maximizes disruption to the system loads weighted by the probability of simultaneous failure of the k components. The resulting problem is formulated as a bilevel mixed-integer nonlinear program. Convex relaxations, linear approximations, and heuristics are developed to obtain feasible solutions that are close to the optimum. A general cutting-plane algorithm is proposed to solve the convex relaxation and linear approximations of the N - k problem. Extensive numerical results corroborate the effectiveness of the proposed algorithms on small-, medium-, and large-scale test instances; the test instances include the IEEE 14-bus system, the IEEE single-area and three-area RTS96 systems, the IEEE 118-bus system, the WECC 240-bus test system, the 1354-bus PEGASE system, and the 2383-bus Polish winter-peak test system.

24 POWER TRANSMISSION AND DISTRIBUTION↗

Optimizing ensemble NV − spin properties of fluorescent diamond microparticles by systematic low pressure high temperature annealing

Low pressure high temperature annealing is a means for driving nitrogen and defect diffusion in diamond to reduce internal lattice damage without the need for technically complicated high-pressure cells. Herein, we perform a systematic time (5, 15, and 30 min) and temperature (1200 °C–1800 °C) study of effects of low-pressure high temperature annealing on photoluminescence, spin concentrations, and spin relaxation properties of NV centers in ca. 3 μm synthetic type 1b diamond particles. Annealing in the temperature range of ca. 1400 °C–1700 °C for even 5 min leads to a higher optically detected magnetic resonance contrast as compared to standard annealing at 900 °C for 2 h. Particles annealed at 1700 °C for 5 min exhibit a contrast close to about 13% as compared to about 9% for those annealed at 900 °C for 2 h. A reduction in the zero-field splitting strain parameter from E ≈ 4.5 MHz to ≈ 2.5 MHz and spectral linewidth from Δν ≈ 7 MHz to ≈ 4 MHz are observed even after 5 min annealing at 1700 °C. Improvements in these spectral parameters resulted in a roughly 2-fold reduction in the noise level of temperature monitoring experiment utilizing an ensemble of NV centers in the particles. Annealing in the temperature range of 1600 °C for 15 or 30 min or 1700 °C for 5 min resulted in NV T 1 relaxation times approaching ca. 5 ms typically observed for bulk diamond. Quantitative electron paramagnetic resonance (EPR) allowed for estimations of thermal activation energies of paramagnetic center annihilation. Monitoring the primary defect concentration (P1 and other defects with half integer spins) and utilizing second order kinetic modeling, an activation energy of 3.63 ± 0.28 eV was estimated. Alternatively, using the NV half field EPR signal and first order kinetic modeling, a similar activation energy 3.89 ± 0.29 eV was estimated.

NV centers↗