Speedup Measurements Research Articles

Speed-up Measures and Associated Problems for Large-scale DEM-Models

This contribution discusses the possibilities to increase the efficiency of large Discrete Element Method (DEM) simulations. Simulations were conducted to test particle upscaling, decreasing shear modulus and using GPU instead of CPU for the computation. The conducted simulations modelled a simple ore extraction process from a defined outlet of a stope in a cave mine. The analysis is based on the influence of the approaches on the computation speed-up and the accuracy of the result. It was found that the real shear modulus could be decreased in the simulation by a factor of 103 without interfering strongly with the result, provided that the decreased shear modulus is not smaller than 108 Pa. However, a reduction of 102 was found to bring the highest speed-up to the present application.For the particle upscaling in this contribution without parameter calibration, the upscaling factor should not exceed 3. Higher upscaling factors affect the results significantly. However, the flow dynamics was already differing for an upscaling factor of 2, even though the flow zones were comparable. Using a GPU instead of a CPU is only recommended if the simulation contains a high number of particles. The speed-up for a simulation with almost 250,000 particles was over 8, but the advantage diminished for less particles. Furthermore, differences in the results between CPU and GPU computation could be found, which could be a starting point for future work. In summary, this research significantly aids in the development of more efficient DEM simulations for large-scale applications, such as cave mining.

Efficiently simulating Lagrangian particles in large-scale ocean flows — Data structures and their impact on geophysical applications

Studying oceanography by using Lagrangian simulations has been adopted for a range of scenarios, such as the determining the fate of microplastics in the ocean, simulating the origin locations of microplankton used for palaeoceanographic reconstructions, and for studying the impact of fish aggregation devices on the migration behaviour of tuna. These simulations are complex and represent a considerable runtime effort to obtain trajectory results, which is the prime motivation for enhancing the performance of Lagrangian particle simulators. This paper assesses established performance enhancing techniques from Eulerian simulators in light of computational conditions and demands of Lagrangian simulators. A performance enhancement strategy specifically targeting physics-based Lagrangian particle simulations is outlined to address the performance gaps, and techniques for closing the performance gap are presented and implemented. Realistic experiments are derived from three specific oceanographic application scenarios, and the suggested performance-enhancing techniques are benchmarked in detail, so to allow for a good attribution of speed-up measurements to individual techniques. The impacts and insights of the performance enhancement strategy are further discussed for Lagrangian simulations in other geoscience applications. The experiments show that I/O-enhancing techniques, such as dynamic loading and buffering, lead to considerable speed-up on-par with an idealised parallelisation of the process over 20 nodes. Conversely, while the cache-efficient structure-of-arrays collection yields a visible speed-up, other alternative data structures fail in fulfilling the theoretically-expected performance increase. This insight demonstrates the importance of good data alignment in memory and caches for Lagrangian physics simulations.

Brillouin Frequency Estimation in Distributed Brillouin Optical Fiber Sensors Based on Instantaneous Frequency

Short-time Fourier transform (STFT) algorithm with Lorentzian fitting is popular in distributed Brillouin optical fiber sensors to acquire Brillouin frequency. However, STFT often suffers from the tradeoff between Brillouin frequency measurement resolution and computation complexity. To lower the complexity of Brillouin frequency analysis and speedup measurement, a novel method based on instantaneous frequency (IF) analysis is proposed. The instantaneous Brillouin frequency can be obtained through the instantaneous phase change between two adjacent samples in the time domain. The requirement on the spectral resolution is, therefore, greatly relaxed, significantly reducing computation complexity and lowering measurement time cost. A Brillouin optical time-domain reflectometry (BOTDR) scheme is demonstrated with a spatial resolution of 1.2 m over a 5-km sensing fiber. For an average count of 4096 over measurements, experiments show that the IF algorithm can reduce computation complexity by 122 times at a standard deviation (STD) of 0.5065 MHz for BFS estimation which is comparable to the STD obtained through STFT in conjunction with Lorentzian fitting.

Spontaneous emission of a quantum emitter near a graphene nanodisk under strong light-matter coupling

We investigate the dynamical evolution of the spontaneous emission of a two-level quantum emitter near a graphene nanodisk. We employ the macroscopic quantum electrodynamics methodology to study the reversible population dynamics of the excited state of the quantum emitter in the non-Markovian limit. Our results indicate that the quantum-emitter--nanodisk interaction enters the strong light-matter coupling regime, as manifested in the excited-state probability where Rabi oscillations and population trapping effects are observed. The level of non-Markovianity is estimated by computing three different well-established measures and relate them to the quantum speedup due to the strong-coupling interaction of the emitter with the graphene nanodisk. Importantly, high values of the non-Markovianity and quantum speedup measures are achieved when the emitter-nanodisk interaction is strong, whereas decreasing coupling strength leads to smaller values.

Collisional growth in a particle-based cloud microphysical model: insights from column model simulations using LCM1D (v1.0)

Abstract. Lagrangian cloud models (LCMs) are considered the future of cloud microphysical modelling. Compared to bulk models, however, LCMs are computationally expensive due to the typically high number of simulation particles (SIPs) necessary to represent microphysical processes such as collisional growth of hydrometeors successfully. In this study, the representation of collisional growth is explored in one-dimensional column simulations, allowing for the explicit consideration of sedimentation, complementing the authors' previous study on zero-dimensional collection in a single grid box. Two variants of the Lagrangian probabilistic all-or-nothing (AON) collection algorithm are tested that mainly differ in the assumed spatial distribution of the droplet ensemble: the first variant assumes the droplet ensemble to be well-mixed in a predefined three-dimensional grid box (WM3D), while the second variant considers the (sub-grid) vertical position of the SIPs, reducing the well-mixed assumption to a two-dimensional, horizontal plane (WM2D). Since the number of calculations in AON depends quadratically on the number of SIPs, an established approach is tested that reduces the number of calculations to a linear dependence (so-called linear sampling). All variants are compared to established Eulerian bin model solutions. Generally, all methods approach the same solutions and agree well if the methods are applied with sufficiently high resolution (foremost is the number of SIPs, and to a lesser extent time step and vertical grid spacing). Converging results were found for fairly large time steps, larger than those typically used in the numerical solution of diffusional growth. The dependence on the vertical grid spacing can be reduced if AON-WM2D is applied. The study also shows that AON-WM3D simulations with linear sampling, a common speed-up measure, converge only slightly slower compared to simulations with a quadratic SIP sampling. Hence, AON with linear sampling is the preferred choice when computation time is a limiting factor. Most importantly, the study highlights that results generally require a smaller number of SIPs per grid box for convergence than previous one-dimensional box simulations indicated. The reason is the ability of sedimenting SIPs to interact with a larger ensemble of particles when they are not restricted to a single grid box. Since sedimentation is considered in most commonly applied three-dimensional models, the results indicate smaller computational requirements for successful simulations, encouraging a wider use of LCMs in the future.

GPU Parallelization of a Hybrid Pseudospectral Geophysical Turbulence Framework Using CUDA

An existing hybrid MPI-OpenMP scheme is augmented with a CUDA-based fine grain parallelization approach for multidimensional distributed Fourier transforms, in a well-characterized pseudospectral fluid turbulence code. Basics of the hybrid scheme are reviewed, and heuristics provided to show a potential benefit of the CUDA implementation. The method draws heavily on the CUDA runtime library to handle memory management and on the cuFFT library for computing local FFTs. The manner in which the interfaces to these libraries are constructed, and ISO bindings utilized to facilitate platform portability, are discussed. CUDA streams are implemented to overlap data transfer with cuFFT computation. Testing with a baseline solver demonstrated significant aggregate speed-up over the hybrid MPI-OpenMP solver by offloading to GPUs on an NVLink-based test system. While the batch streamed approach provided little benefit with NVLink, we saw a performance gain of 30 % when tuned for the optimal number of streams on a PCIe-based system. It was found that strong GPU scaling is nearly ideal, in all cases. Profiling of the CUDA kernels shows that the transform computation achieves 15% of the attainable peak FlOp-rate based on a roofline model for the system. In addition to speed-up measurements for the fiducial solver, we also considered several other solvers with different numbers of transform operations and found that aggregate speed-ups are nearly constant for all solvers.

Parallel Hybrid BBO Search Method for Twitter Sentiment Analysis of Large Scale Datasets Using MapReduce

Sentiment analysis is an eminent part of data mining for the investigation of user perception. Twitter is one of the popular social platforms for expressing thoughts in the form of tweets. Nowadays, tweets are widely used for analyzing the sentiments of the users, and utilized for decision making purposes. Though clustering and classification methods are used for the twitter sentiment analysis, meta-heuristic based clustering methods has witnessed better performance due to subjective nature of tweets. However, sequential meta-heuristic based clustering methods are computation intensive for large scale datasets. Therefore, in this paper, a novel MapReduce based K-means biogeography based optimizer(MR-KBBO) is proposed to leverage the strength of biogeography based optimizer with MapReduce model to efficiently cluster the large scale data. The proposed method is validated against four state-of-the-art MapReduce based clustering methods namely; parallel K-means, parallel K-means particle swarm optimization, MapReduce based artificial bee colony optimization, dynamic frequency based parallel k-bat algorithm on four large scale twitter datasets. Further, speedup measure is used to illustrate the computation performance on varying number of nodes. Experimental results demonstrate that the proposed method is efficient in sentiment mining for the large scale twitter datasets.

An efficient genetic algorithm for large-scale transmit power control of dense and robust wireless networks in harsh industrial environments

The industrial wireless local area network (IWLAN) is increasingly dense, due to not only the penetration of wireless applications to shop floors and warehouses, but also the rising need of redundancy for robust wireless coverage. Instead of simply powering on all access points (APs), there is an unavoidable need to dynamically control the transmit power of APs on a large scale, in order to minimize interference and adapt the coverage to the latest shadowing effects of dominant obstacles in an industrial indoor environment. To fulfill this need, this paper formulates a transmit power control (TPC) model that enables both powering on/off APs and transmit power calibration of each AP that is powered on. This TPC model uses an empirical one-slope path loss model considering three-dimensional obstacle shadowing effects, to enable accurate yet simple coverage prediction. An efficient genetic algorithm (GA), named GATPC, is designed to solve this TPC model even on a large scale. To this end, it leverages repair mechanism-based population initialization, crossover and mutation, parallelism as well as dedicated speedup measures. The GATPC was experimentally validated in a small-scale IWLAN that is deployed a real industrial indoor environment. It was further numerically demonstrated and benchmarked on both small- and large-scales, regarding the effectiveness and the scalability of TPC. Moreover, sensitivity analysis was performed to reveal the produced interference and the qualification rate of GATPC in function of varying target coverage percentage as well as number and placement direction of dominant obstacles.

Single-pass incremental force updates for adaptively restrained molecular dynamics.

Adaptively restrained molecular dynamics (ARMD) allows users to perform more integration steps in wall-clock time by switching on and off positional degrees of freedoms. This article presents new, single-pass incremental force updates algorithms to efficiently simulate a system using ARMD. We assessed different algorithms for speedup measurements and implemented them in the LAMMPS MD package. We validated the single-pass incremental force update algorithm on four different benchmarks using diverse pair potentials. The proposed algorithm allows us to perform simulation of a system faster than traditional MD in both NVE and NVT ensembles. Moreover, ARMD using the new single-pass algorithm speeds up the convergence of observables in wall-clock time. © 2017 Wiley Periodicals, Inc.

An efficient genetic algorithm for large-scale planning of dense and robust industrial wireless networks

With the penetration of Internet of things in manufacturing industry, it is an unavoidable issue to maintain robust wireless connections among machines and human workers in harsh industrial environments. However, the existing wireless planning tools focus on office environments, which are less harsh than industrial environments regarding shadowing effects of diverse obstacles. To fill this gap, this paper proposes an over-dimensioning (OD) model, which automates the decision making on deploying a robust industrial wireless local area network (IWLAN). This model creates two full coverage layers while minimizing the deployment cost, and guaranteeing a minimal separation distance between two access points (APs) to prevent APs that cover the same region from being simultaneously shadowed by an obstacle. Moreover, an empirical one-slope path loss model, which considers three-dimensional obstacle shadowing effects, is proposed for simple yet precise coverage calculation. To solve this OD model even at a large size, an efficient genetic algorithm based over-dimensioning (GAOD) algorithm is designed. Genetic operators, parallelism, and speedup measures are tailored to enable large-scale optimization. A greedy heuristic based over-dimensioning (GHOD) algorithm is further proposed, as a state-of-the-art heuristic benchmark algorithm. In small- and large-size OD problems based on industrial data, the GAOD was demonstrated to be 20%–25% more economical than benchmark algorithms for OD in the same environment. The effectiveness of GAOD was further experimentally validated with a real deployment system. Though this paper focuses on an IWLAN, the proposed GAOD can serve as a decision making tool for deploying other types of robust industrial wireless networks in terms of coverage, such as wireless sensor networks and radio-frequency identification (RFID) networks.

Embedded multicore computing and applications

Embedded multicore computing and applications

Ray‐tracing domain decomposition methods for real‐time simulation on multi‐core and multi‐processor systems

SummaryThis paper describes the use of domain decomposition methods for accelerating wave physics simulation. Numerical wave‐based methods provide more accurate simulation than geometrical methods, but at a higher computation cost as well. In the context of virtual reality, the quality of the results is estimated according to human perception, what makes geometrical methods an interesting approach for achieving real‐time physically‐based rendering. Here, we investigate a geometrical method based on both beams and rays tracing, which we enhance by two levels of parallel processing. Techniques from domain decomposition methods are coupled with classical parallel computing on both shared and distributed memory. Both optic and acoustic renderings are experimented to evaluate the acceleration impact of the domain decomposition scheme. Speedup measurements clearly show the efficiency of using domain decomposition methods for real‐time simulation of wave physics. Copyright © 2015 John Wiley & Sons, Ltd.

Asynchronous and implicitly parallel evolutionary computation models

This paper presents the design and the application of asynchronous models of parallel evolutionary algorithms. An overview of the existing parallel evolutionary algorithm (PEA) models and available implementations is given. We present new PEA models in the form of asynchronous algorithms and implicit parallelization, as well as experimental data on their efficiency. The paper also discusses the definition of speedup in PEAs and proposes an appropriate speedup measurement procedure. The described parallel EA algorithms are tested on problems with varying degrees of computational complexity. The results show good efficiency of asynchronous and implicit models compared to existing parallel algorithms.

Speeding up Martins’ algorithm for multiple objective shortest path problems

The latest transportation systems require the best routes in a large network with respect to multiple objectives simultaneously to be calculated in a very short time. The label setting algorithm of Martins efficiently finds this set of Pareto optimal paths, but sometimes tends to be slow, especially for large networks such as transportation networks. In this article we investigate a number of speedup measures, resulting in new algorithms. It is shown that the calculation time to find the Pareto optimal set can be reduced considerably. Moreover, it is mathematically proven that these algorithms still produce the Pareto optimal set of paths.

Runtime Reconfiguration of Multiprocessors Based on Compile-Time Analysis

In multiprocessors, performance improvement is typically achieved by exploring parallelism with fixed granularities, such as instruction-level, task-level, or data-level parallelism. We introduce a new reconfiguration mechanism that facilitates variations in these granularities in order to optimize resource utilization in addition to performance improvements. Our reconfigurable multiprocessor QuadroCore combines the advantages of reconfigurability and parallel processing. In this article, a unified hardware-software approach for the design of our QuadroCore is presented. This design flow is enabled via compiler-driven reconfiguration which matches application-specific characteristics to a fixed set of architectural variations. A special reconfiguration mechanism has been developed that alters the architecture within a single clock cycle. The QuadroCore has been implemented on Xilinx XC2V6000 for functional validation and on UMC’s 90nm standard cell technology for performance estimation. A diverse set of applications have been mapped onto the reconfigurable multiprocessor to meet orthogonal performance characteristics in terms of time and power. Speedup measurements show a 2--11 times performance increase in comparison to a single processor. Additionally, the reconfiguration scheme has been applied to save power in data-parallel applications. Gate-level simulations have been performed to measure the power-performance trade-offs for two computationally complex applications. The power reports confirm that introducing this scheme of reconfiguration results in power savings in the range of 15--24%.

Parallel processing in structural reliability

The present contribution addresses the parallelization of advanced simulation methods for structural reliability analysis, which have recently been developed for large-scale structures with a high number of uncertain parameters. In particular, the Line Sampling method and the Subset Simulation method are considered. The proposed parallel algorithms exploit the parallelism associated with the possibility to simultaneously perform independent FE analyses. For the Line Sampling method a parallelization scheme is proposed both for the actual sampling process, and for the statistical gradient estimation method used to identify the so-called important direction of the Line Sampling scheme. Two parallelization strategies are investigated for the Subset Simulation method: the first one consists in the embarrassingly parallel advancement of distinct Markov chains; in this case the speedup is bounded by the number of chains advanced simultaneously. The second parallel Subset Simulation algorithm utilizes the concept of speculative computing. Speedup measurements in context with the FE model of a multistory building (24,000 DOFs) show the reduction of the wall-clock time to a very viable amount (<10 minutes for Line Sampling and <TEX>${\approx}$</TEX> 1 hour for Subset Simulation). The measurements, conducted on clusters of multi-core nodes, also indicate a strong sensitivity of the parallel performance to the load level of the nodes, in terms of the number of simultaneously used cores. This performance degradation is related to memory bottlenecks during the modal analysis required during each FE analysis.

Grid-enabled evolutionary algorithm application in the mechanical optimization problems

The application of evolutionary algorithms to the optimization problems is presented in the paper. The Alchemi grid framework is used during computations. The fitness function is evaluated by using finite-element method. The numerical examples and speedup measurements for a different number of processors are shown.

Fiducial line based skew estimation

Skew estimation for textual document images is a well-researched topic and numerals of methods have been reported in the literature. One of the major challenges is the presence of interfering non-textual objects of various types and quantities in the document images. Many existing methods require proper separation of the textual objects which are well aligned from the non-textual objects which are mostly nonaligned. Some comparative evaluation work on the existing methods chooses only the text zones of the test image database. Therefore, the object filtering or zoning stage is crucial to the skew detection stage. However, it is difficult if not impossible to design general-purpose filters that are able to discriminate noises from textual components. This paper presents a robust, general-purpose skew estimation method that does not need any filtering or zoning preprocessing. In fact, this method does apply filtering, but not on the input components at the beginning of the detection process, rather on the output spectrum at the end of the detection process. Therefore, the problem of finding a textual component filter has been transformed into finding a convolution filter on the output accumulator array. This method consists of three steps: (1) the calculation of the slopes of the virtual lines that pass through the centroids of all the unique pairs of the connected components in an image, and quantizes the arctangents of the slopes into a 1-D accumulator array that covers the range from - 90 ∘ to + 90 ∘ ; (2) a special convolution on the resultant histogram, after which there remain only the prominent peaks that possibly correspond to the skew angles of the image; (3) the verification of the detection result. Its computational complexity and detection precision are uncoupled, unlike those projection-profile-based or Hough-transform-based methods whose speeds drop when higher precision is in demand. Speedup measures on the baseline implementation are also presented. The University of Washington English Document Image Database I (UWDB-I) contains a large number of scanned document images with significant amount of non-textual objects. Therefore, it is a good image database for evaluating the proposed method.

Advanced eager scheduling for Java‐based adaptive parallel computing

AbstractJavelin 3 is a software system for developing large‐scale, fault‐tolerant, adaptively parallel applications. When all or part of their application can be cast as a master–worker or branch‐and‐bound computation, Javelin 3 frees application developers from concerns about inter‐processor communication and fault tolerance among networked hosts, allowing them to focus on the underlying application. The paper describes a fault‐tolerant task scheduler and its performance analysis. The task scheduler integrates work stealing with an advanced form of eager scheduling. It enables dynamic task decomposition, which improves host load‐balancing in the presence of tasks whose non‐uniform computational load is evident only at execution time. Speedup measurements are presented of actual performance on up to 1000 hosts. We analyze the expected performance degradation due to unresponsive hosts, and measure actual performance degradation due to unresponsive hosts. Copyright © 2005 John Wiley &amp; Sons, Ltd.

Iterative solution of elliptic collocation systems on a cognitive parallel computer

The performance of parallel algorithms implementing the block SOR iterative method, used for the solution of linear systems arising from the discretization of elliptic BVPs by the finite element collocation method, is investigated in this paper. By making use of a recently introduced block tridiagonal partitioning of the collocation matrix, which yields faster rates of convergence, and its associated modified red-black ordering of unknowns and equations, we succeeded to increase the scalability of the problem. The study of the communicational and computational needs of the problem on a virtual parallel machine made possible the mapping of the resulted SOR-algorithm on a fixed size distributed memory parallel system, utilizing a simple pipeline architecture among its processors, in an optimum way. The proposed parallel schemes are being realized on a Parsytec cognitive computer via an SPMD programming model. Speedup measurements are used to reveal the efficiency of our implementation.