Parallel Units Research Articles

Temporal Parallelization of Inference in Hidden Markov Models

This paper presents algorithms for parallelization of inference in hidden Markov models (HMMs). In particular, we propose parallel backward-forward type of filtering and smoothing algorithm as well as parallel Viterbi-type maximum-a-posteriori (MAP) algorithm. We define associative elements and operators to pose these inference problems as parallel-prefix-sum computations in sum-product and max-product algorithms and parallelize them using parallel-scan algorithms. The advantage of the proposed algorithms is that they are computationally efficient in HMM inference problems with long time horizons. We empirically compare the performance of the proposed methods to classical methods on a highly parallel graphical processing unit (GPU).

SIKE on GPU: Accelerating Supersingular Isogeny-Based Key Encapsulation Mechanism on Graphic Processing Units

Since the introduction of the Supersingular isogeny Diffie–Hellman (SIDH) key exchange protocol by Jao and de Feo in 2011, it and its variation (SIKE) have gained significant attention as a promising candidate for post-quantum cryptography (PQC). Until now, even though several implementations of the state-of-the-art SIKE mechanism were presented on CPUs and embedded MCUs, there was no consideration of implementing SIKE on parallel graphic processing units (GPUs). With the advent of the IoT era, a number of IT devices will communicate with application servers. Thus, developing efficient instance of SIKE on server sides is also important. GPUs have been considered as a promising candidate for a cryptographic accelerator. In this paper, we present an efficient implementation of Supersingular Isogeny Key Encapsulation (SIKE) mechanism on GPUs. Even though SIKE has fascinating advantages of much smaller key and ciphertext sizes compared with other NIST PQC candidates, its computational overhead is extremely high. Until now, a large amount of research has been conducted for enhancing the performance of SIKE with respect to software on typical CPU and embedded MCUs and hardware optimization on ASIC and FPGA. However, generic software optimization utilizing GPUs has not been considered yet. We target the GPU implementation of SIKEp503 security parameters which provides the security level 2 (At least as difficult to break as SHA256). For efficiency, we optimize the underlying field arithmetic, especially field multiplication and reduction over $p503=2^{250}3^{159}-1$ and take full advantage of the properties of GPU architecture including memory hierarchy. The proposed GPU software based on RTX2080Ti provides around 36376.61 KeyGens/s, 25603.72 Encaps/s, and 22211.61 Decaps/s. These are about 140.64, 157.66, and 146.81 times of improvements to the SIKE CPU Software on Intel i9-10900K CPU, respectively. As far as we know, this is the first efficient implementation of SIKE software on GPU side.

Existence of some classes of N(k)-quasi Einstein manifolds

The object of the present paper is to study some classes of N(k)-quasi Einstein manifolds. The existence of such manifolds are proved by giving non-trivial physical and geometrical examples. It is also proved that the characteristic vector field of the manifold is killing as well as parallel unit vector fields under certain curvaturerestrictions.

THE COMPUTER SYSTEM FOR OPTIMAL CONTROL OF NATURAL GAS PUMPING UNITS

Method of the optimal control of the natural gas compression process by a compressor station, which is equipped with gas pumping units with a gas turbine drive, has been developed.
 The parallel scheme of connection of gas pumping units working on the general collector is considered.
 The cost of fuel gas taking into account the power of nitrogen oxide emissions into the atmosphere is chosen as the criterion of optimality. This takes into account the restrictions on technological parameters and the planned task for the transportation of natural gas.
 The formed problem is divided into two subtasks - the choice of the optimal number of parallel units in each of the subgroups and determining the optimal mode of operation of each unit in the group.
 Formalization of optimal control problem of GPA operation takes into account the uncertainty inherent in the process of natural gas compression. For this purpose, the apparatus of fuzzy algebra was used, which made it possible to take into account the inaccuracies that arise both as a result of measuring technological parameters and as a result of processing the obtained data on the technological process.
 The optimal control system is integrated into the upper level (dispatcher) and is an integral part of the overall control system of the GPA and has a hierarchical structure. At the lower level there are measuring devices of technological parameters, which are a source of information for the upper level where the solution of optimization problems takes place.
 The developed methods formed the basis of the algorithmic and software of the computer system for optimal control of a separate unit of the compressor station, which can be integrated into the existing automated system.

Sensitivity Analysis of Genetic Algorithms for Optimization of Operating Conditions for Protein Production Plants

This work introduces genetic algorithms, which will be investigated and used to perform sensitivity analysis. Genetic algorithms are part of a collection of stochastic optimization algorithms based on the concepts of biological evolutionary theory. This research deals with the problem of the search for optimal investment cost of multiproduct batch chemical plants found in a chemical engineering process with uncertain demand. The aim of this work is to minimize the investment cost and find out the number and size of parallel equipment units in each stage. For this purpose, it is proposed to solve the problem by using Genetics Algorithms (GAs). This GAs consider an effective mixed continuous discrete coding method with a four point crossover operator, which take into account, the uncertainty on the demand using Gaussian process modeling. Experiments indicated that relatively good results could be obtained using 4-point crossover with an applied rate of 0.7 and mutation rate ranged between 0.01 and 0.09 promised to give best performance. The results (number and size of equipment, investment cost, production time (Hi), CPU time and Idle times in plant) obtained by GAs are much more rapidly than mixed integer linear programming. This methodology can help the decision makers and constitutes a very promising framework for finding a set of “good solutions”.

Effect of Al Concentration on the Microstructural Evolution of Fe-Cr-Al Systems: A Phase-Field Approach

In this study, the microstructural evolution of an Fe-Cr-Al system was simulated in two-dimensional (2D) and three-dimensional (3D) systems using the phase-field method. We investigated the effect of Al concentration on the microstructural evolution of the systems, with a focus on the nucleation and growth of the Cr-rich α′ phase. In addition, we quantitatively analyzed the mechanism of the effect of Al concentration on the microstructural characteristics of the 2D and 3D systems, such as the number of precipitates, average precipitate area (volume), and α′ phase fraction. In particular, we analyzed the effect of Al concentration and the dimensions of the system cell on the formation of the interconnected structure at high Cr concentrations, such as 35 Crat% and 40 Crat%. To enhance the performance of the simulations, we applied a semi-implicit Fourier spectral method for the ternary system and a parallel graphics processing unit computing technique. The results revealed that the initiation of phase separation in the 2D and 3D simulations was enhanced with an increase in the average Al concentration in the system. In addition, with an increase in the average Al concentration in both systems, the α′ phase fraction increased, while the change in the phase fraction decreased.

Breaking the Interaction Wall: A DLPU-Centric Deep Learning Computing System

Due to the broad successes of deep learning, many CPU-centric artificial intelligent computing systems employ specialized devices such as GPUs, FPGAs, and ASICs, which can be named as Deep Learning Processing Units (DLPUs), for processing computation-intensive deep learning tasks. The separation between the scalar control operations mapped on CPUs and the vector computation operations mapped on DLPUs causes the frequent and costly interactions between CPUs and DLPUs, leading to the <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">Interaction Wall</i> . Moreover, the increasing algorithm complexity and DLPU computation speed would further aggravate the interaction wall substantially. To break the interaction wall, we propose a novel DLPU-centric deep learning computing system consisting of an <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">exception-oriented programming (EOP) model</i> and the architectural support of <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">CPULESS DLPU</i> . The EOP model processes scalar control operations of a deep learning task as exception handlers to maximally avoid stalling the crucial and dominated vector computation operations. Together with the CPULESS DLPU which integrates a scalar processing unit (SPU) for scalar control operations and the parallel processing unit (PPU) for vector computation operations into a fused pipeline, the proposed DLPU-centric system can cost-effectively leverage the EOP model to execute the two kinds of operations simultaneously without disturbing each other. Compared with a state-of-the-art commodity CPU-centric system with discrete V100 GPU via PCIe bus, experimental results show that our DLPU-centric system achieves 10.30× better performance and 92.99 percent energy savings, respectively. Moreover, compared with a CPU-centric version of DLPU system where the SPU serves as the host with integrated PPU, the proposed DLPU-centric system still achieves 15.60 percent better performance from avoided interactions.

High-Selectivity Frequency-Selective Rasorber Based on Low-Profile Bandpass Filter

A frequency-selective rasorber (FSR) with a high selectivity transmission passband and two absorption bands is proposed in this letter. The design method is introduced based on the equivalent circuit model (ECM). This structure consists of a lossy layer and a lossless layer. The lossy layer is composed of Jerusalem elements with a parallel resonance unit in center, and load with resistors on the metal branch. The lossless layer is integrated by a multilayer frequency selective surface to realize the high selectivity and wide transmission band. The surface current distributions and surface loss density were illustrated to explain the working principle. The simulated results show that the 1 dB transmission window is obtained from 7.9 to 9.0 GHz. The −10 dB absorption bands are over 3.3–7.1 GHz and 9.4–12.0 GHz. A prototype of the proposed FSR was fabricated and measured to verify our design; the measured results show a good agreement with simulated results.

A comparison of different metaheuristics for the quadratic assignment problem in accelerated systems

Although metaheuristics are used for solving complex and real-world problems, they do not provide an exact solution, but only an approximate solution in feasible time. However, for very large instances, a metaheuristic may take a considerable amount of time. Therefore, we utilize a highly parallel metaheuristic to run on a modern massively parallel graphics processing unit (GPU) to further reduce the execution time. Additionally, because metaheuristics are not problem-specific, it is interesting to determine which metaheuristic is best for each problem type. In this study, using a massive parallel machine such as a GPU, we evaluate the performance of different metaheuristics such as the iterated local search, simulated annealing, genetic algorithm, tabu search, particle swarm optimization, and crow search algorithm with respect to the quadratic assignment problem.

Hygrothermal performance comparison study on bamboo and timber construction in Asia-Pacific bamboo areas

The hygrothermal feasibility is one of the premises for applying bamboo in building industry. This study carries out material property tests on typical bamboo variants and hygrothermal performance simulation on bamboo exterior wall constructions in the Asia-Pacific bamboo areas, and makes comparison with reference timber units in parallel. Results show that, except for bamboo particleboard, the rest bamboo variants have overall lower hygric properties and higher thermal properties, and some indicators of bamboo scrimber and bamboo mat board exceed the property range of the reference timber. The construction groups with bamboo particleboard as interior or (thickened) interlayer boards show overall better hygrothermal performance than those with timber units given the same construction and space size. For example, the annual total cooling demand and cooling peak are reduced by up to 14%, respectively. The combination of bamboo mat board and bamboo particleboard in bamboo groups shows better performance in most indicators than the combination of OSB and Spruce in timber groups. Thanks to the advantages of lower cost and technical requirement, bamboo particleboard and bamboo mat board have the potential to be local climate adaptive building materials and are competitive compared with the corresponding timber products. The multi-level comparison between bamboo and timber units in this study demonstrates the feasibility to ‘substitute timber with bamboo’ in terms of hygrothermal performance, identifies the dominant bamboo variants and offers suggestions on its market competition with timber.

CPU: Cross-Rack-Aware Pipelining Update for Erasure-Coded Storage

Erasure coding is widely used in distributed storage systems (DSSs) to efficiently achieve fault tolerance. However, when the original data need to be updated, erasure coding must update every encoded block, resulting in long update time and high bandwidth consumption. Exiting solutions are mainly focused on coding schemes to minimize the size of transmitted update information, while ignoring more efficient utilization of bandwidth among update racks. In this article, we propose a parallel Cross-rack Pipelining Update scheme ( <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">CPU</i> ), which divides the update information into small-size units and transmits these units in parallel along with an update pipeline path among multiple racks. The performance of <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">CPU</i> is mainly determined by slice size and update path. More slices bring finer-grained parallel transmissions over cross-rack links, but also introduces more overheads. An update path that traverses all racks with large-bandwidth links provide short update time. We formulate the proposed pipelining update scheme as an optimization problem, based on a new theoretical pipelining update model. We prove the optimization problem is NP-hard and develop a heuristic algorithm to solve it based on the features of practical DSSs and our implementations, including <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">Big chunk</i> and <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">Small overhead</i> . Specifically, we determine the best update path first by solving a max-min problem and then decide the slice size. We further simplify the slice size selection by offline learning a range of interesting ( <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">RoI</i> ), in which all slice sizes provide similar performance. We implement <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">CPU</i> and conduct experiments on Amazon EC2 under a variety of scenarios. The results show that <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">CPU</i> can reduce the average update time by 48.2 percent, compared with the state-of-the-art update schemes.

Visual Simultaneous Localization and Mapping with Applications to Monitoring of Underground Transportation Infrastructure

This paper presents the results of the development, design, and implementation of a visual simultaneous localization and mapping (SLAM) system for autonomous real-time localization with application to underground transportation infrastructure (UTI) such as tunnels. Localization is achieved in the absence of any global positioning system (GPS) or auxiliary system. The indoor localization system is a necessary element of a fully autonomous platform for the detection of cracks and other anomalies on the interior surfaces of tunnels and other UTI. It can be used for tagging of high-resolution sensor data obtained with low-cost prototype data acquisition platforms previously developed. Visual based SLAM has been used as the core element in an architecture employing a commercial off-the-shelf (COTS) ZED stereo camera from Stereolabs. To achieve real-time operation, an NVIDIA Jetson TX2 massively parallel graphics processing unit (GPU) was used as the core computational engine employing two different software libraries. We achieved localization at 5 frames per second (FPS) using ORBSLAM2 open-source software library, and the much lighter, but proprietary, ZED SDK was able to deliver a performance at nearly 60 FPS. To assess the accuracy of the relative localization system, we conducted several tests at 30 FPS and reported on the resulting error variances that were found to be consistently very small. Finally, we conducted several tests in a tunnel in the Los Angeles county area and confirmed the applicability of the method for monitoring UTI.

Correction: Nitric oxide acts as a cotransmitter in a subset of dopaminergic neurons to diversify memory dynamics

Animals employ diverse learning rules and synaptic plasticity dynamics to record temporal and statistical information about the world. However, the molecular mechanisms underlying this diversity are poorly understood. The anatomically defined compartments of the insect mushroom body function as parallel units of associative learning, with different learning rates, memory decay dynamics and flexibility (Aso and Rubin, 2016). Here, we show that nitric oxide (NO) acts as a neurotransmitter in a subset of dopaminergic neurons in Drosophila. NO's effects develop more slowly than those of dopamine and depend on soluble guanylate cyclase in postsynaptic Kenyon cells. NO acts antagonistically to dopamine; it shortens memory retention and facilitates the rapid updating of memories. The interplay of NO and dopamine enables memories stored in local domains along Kenyon cell axons to be specialized for predicting the value of odors based only on recent events. Our results provide key mechanistic insights into how diverse memory dynamics are established in parallel memory systems.

One-dimensional spatial patterning along mitotic chromosomes: A mechanical basis for macroscopic morphogenesis

Spatial patterns are ubiquitous in both physical and biological systems. We have recently discovered that mitotic chromosomes sequentially acquire two interesting morphological patterns along their structural axes [L. Chu et al., Mol. Cell, 10.1016/j.molcel.2020.07.002 (2020)]. First, axes of closely conjoined sister chromosomes acquire regular undulations comprising nearly planar arrays of sequential half-helices of similar size and alternating handedness, accompanied by periodic kinks. This pattern, which persists through all later stages, provides a case of the geometric form known as a "perversion." Next, as sister chromosomes become distinct parallel units, their individual axes become linked by bridges, which are themselves miniature axes. These bridges are dramatically evenly spaced. Together, these effects comprise a unique instance of spatial patterning in a subcellular biological system. We present evidence that axis undulations and bridge arrays arise by a single continuous mechanically promoted progression, driven by stress within the chromosome axes. We further suggest that, after sister individualization, this same stress also promotes chromosome compaction by rendering the axes susceptible to the requisite molecular remodeling. Thus, by this scenario, the continuous presence of mechanical stress within the chromosome axes could potentially underlie the entire morphogenetic chromosomal program. Direct analogies with meiotic chromosomes suggest that the same effects could underlie interactions between homologous chromosomes as required for gametogenesis. Possible mechanical bases for generation of axis stress and resultant deformations are discussed. Together, these findings provide a perspective on the macroscopic changes of organized chromosomes.

Coordination framework of cadmium(II), harvested from dithiolate-imidazole binary ligand systems: Crystal structure, Hirshfeld surface analysis, antibacterial, and DNA cleavage potential

The novel one-dimensional coordination polymer with binuclear building node {[Cd2(i-mnt)2(Im)2]DMF2}n (1) harvested from 1,1-dicyanoethylene-2,2-dithiolate (i-mnt2−) and imidazole (Im) has been systemically designed, synthesized and structurally characterized by elemental analysis, FT-IR and 1H NMR spectroscopy. Single crystal X-ray diffraction analysis revealed that in 1, each Cd(II) is in penta-coordinated environment having slightly distorted square pyramidal geometry (τ5 = 0.24) and have two bridging thiolate ends of i-mnt−2- along with one DMF solvent molecule. The flexible i-mnt2− ligand acts as a tridentate connector by ligating with Cd(II) centers in parallel dimeric units forming a 1D polymeric chain. Hirshfeld surface analyses and two dimensional (2D) finger print plots were employed to decode all types of non-covalent contacts in the crystal frameworks, allowing a systematic evaluation of the crystal structure. From HS analysis, it was observed that N⋯H/H⋯N type interactions (22.1%) have the major contribution compared to overall non-covalent interacrtions. The share of H⋯O/O⋯H interactions of the total surface is only 3.7%, their role is important in the direction and organization of crystal packaging. The in-vitro antimicrobial study of 1 against Gram-positive (Staphylococcus aureus and Bacillus subtilis) and Gram-negative (E. coli and P. aeruginosa) human pathogenic bacterial strains show promising antimicrobial activities. Furthermore, DNA cleavage study (E. coli DNA) of 1 using gel electrophoresis method revealed promising DNA cleaver property similar to H2O2 due to the efficient coordination of Cd(II) ion.

Person Attribute Recognition by Sequence Contextual Relation Learning

Person attribute recognition aims to identify the attribute labels from the pedestrian images. Extracting contextual relation from the images and attributes, including the spatial-semantic relations, the spatial context and the semantic correlation, is beneficial to enhance the discrimination of the features for recognizing the attributes. Thus, this work proposes a sequence contextual relation learning (SCRL) method to capture these relations. It first embeds the images and attributes into sequences in two branches. Then SCRL flexibly learns the contextual relation from the sequences with the parallel attention model structure, which integrates the inter-attention and intra-attention models. The inter-attention module is utilized to extract the spatial-semantic relations, while the intra-attention is designed to gain the spatial context and the semantic correlation. Both attention modules are comprised of several parallel attention units and each unit can obtain the pairwise relations in one subspace. Therefore, they obtain the relations in multiple subspaces, which can improve the comprehensiveness of the relation learning. Additionally, for the sake of better extraction of spatial-semantic relations, this paper employs connectionist temporal classification (CTC) loss which is capable of driving the network to enforce monotonic alignment between the image and attribute. It can also accelerate the convergence of the network by the algorithm in it. Extensive experiments on five public datasets, i.e., Market-1501 attribute, Duke attribute, PETA, RAP and PA-100K datasets, demonstrate the effectiveness of the proposed method.

Text Selection Issue For Parallel Corpus

It is known that the basis of any corpus is its units. Typically, texts of different genres are selected as the corpus unit to ensure the representativeness of the corpus. Therefore, when creating any language corpus, first of all, the principles of selection of texts that are part of it should be defined. Parallel corpus units consist of texts that have been translated one or more times from the original. Which topic and genre text to choose for the parallel corpus is determined by the purpose of the compiler?

Computing Performance Enhancement of VLIW Architecture Using Instruction Level Parallelism

Our paper proposes the new method of processor architecture called as VLIW for enhancing the performance of the architecture. VLIW is the complexity architecture because the enormous number of registers, slices, flip flops, counters, operand, ALUs, and MUXs used. The VLIW has the fife stages of pipelines for executing the architecture are (1) fetching the 128-bit instruction memory, (2) decode stage or it is also called as the operands reading stage because the total number of operands are implemented in this stage, (3) execution stage, here the operations with the parallel executions units which has the four operations, (4) memory stage is used for loading or for storing the data from/to the memory and (5) write back stage in this stage the outputs of all the stage is collected and write back into the register file for storing the output values. The whole process of implementation is implemented in the FPGA of the family of Spartan-6 XC6SLX-3CSG324 device. In this proposed architecture the performance of the architecture is increased by reducing the time taken to execute the CPU of Xst completion of the architecture

Cooperative hybrid evolutionary algorithm for large scale multi-stage multi-product batch plants scheduling problem

As an important part of batch chemical industry scheduling problems, the multi-stage multi-product batch plant scheduling problem (MMSP) has been widely studied for decades. This problem is characterized by multiple stages with non-identical parallel units and operate based on customer orders. In this paper, we focus on the large scale MMSP and treat the minimization of make-span as the objective function. An efficient cooperative hybrid evolutionary algorithm is proposed based on the framework of cooperative co-evolution. First, a novel two-line encoding scheme is developed to represent the unit assignment and sequencing for orders respectively. Second, modified estimation of distribution algorithm (EDA) and differential evolutionary (DE) operations are proposed according to the feature of MMSP. EDA operation with a novel population-based incremental learning strategy is applied to handle the unit assignment variables. And novel DE operation based on a novel encoding method is adopted to deal with sequence variables. Then, two selection strategies are applied to preserve optimal and sub-optimal solutions for the proposed algorithm. The critical path based local search algorithm is adopted to further improve the efficiency of local optimization. The proposed algorithm has been tested by several instances with different sizes and characteristics. The numerical results and comparisons show that the proposed work is very competitive in solving large scale MMSP.

Review of smoothed particle hydrodynamics: towards converged Lagrangian flow modelling.

This paper presents a review of the progress of smoothed particle hydrodynamics (SPH) towards high-order converged simulations. As a mesh-free Lagrangian method suitable for complex flows with interfaces and multiple phases, SPH has developed considerably in the past decade. While original applications were in astrophysics, early engineering applications showed the versatility and robustness of the method without emphasis on accuracy and convergence. The early method was of weakly compressible form resulting in noisy pressures due to spurious pressure waves. This was effectively removed in the incompressible (divergence-free) form which followed; since then the weakly compressible form has been advanced, reducing pressure noise. Now numerical convergence studies are standard. While the method is computationally demanding on conventional processors, it is well suited to parallel processing on massively parallel computing and graphics processing units. Applications are diverse and encompass wave-structure interaction, geophysical flows due to landslides, nuclear sludge flows, welding, gearbox flows and many others. In the state of the art, convergence is typically between the first- and second-order theoretical limits. Recent advances are improving convergence to fourth order (and higher) and these will also be outlined. This can be necessary to resolve multi-scale aspects of turbulent flow.