Speedup Ratio Research Articles

An improved efficient adaptive method for large-scale multi-explosives explosion simulations

Shock wave caused by a sudden release of high-energy, such as explosion and blast, usually affects a significant range of areas. The utilization of a uniform fine mesh to capture sharp shock wave and to obtain precise results is inefficient in terms of computational resource. This is particularly evident when large-scale fluid field simulations are conducted with significant differences in computational domain size. In this work, a variable-domain-size adaptive mesh enlargement (vAME) method is developed based on the proposed adaptive mesh enlargement (AME) method for modeling multi-explosives explosion problems. The vAME method reduces the division of numerous empty areas or unnecessary computational domains by adaptively suspending enlargement operation in one or two directions, rather than in all directions as in AME method. A series of numerical tests via AME and vAME with varying nonintegral enlargement ratios and different mesh numbers are simulated to verify the efficiency and order of accuracy. An estimate of speedup ratio is analyzed for further efficiency comparison. Several large-scale near-ground explosion experiments with single/multiple explosives are performed to analyze the shock wave superposition formed by the incident wave, reflected wave, and Mach wave. Additionally, the vAME method is employed to validate the accuracy, as well as to investigate the performance of the fluid field and shock wave propagation, considering explosive quantities ranging from 1 to 5 while maintaining a constant total mass. The results show a satisfactory correlation between the overpressure versus time curves for experiments and numerical simulations. The vAME method yields a competitive efficiency, increasing the computational speed to 3.0 and approximately 120,000 times in comparison to AME and the fully fine mesh method, respectively. It indicates that the vAME method reduces the computational cost with minimal impact on the results for such large-scale high-energy release problems with significant differences in computational domain size.

IDAD: An improved tensor train based distributed DDoS attack detection framework and its application in complex networks

With the vigorous development of Internet technology, the scale of systems in the network has increased sharply, which provides a great opportunity for potential attacks, especially the Distributed Denial of Service (DDoS) attack. In this case, detecting DDoS attacks is critical to system security. However, current detection methods exhibit limitations, leading to compromises in accuracy and efficiency. To cope with it, three key strategies are implemented in this paper: (i) Using tensors to model large-scale and heterogeneous data in complex networks; (ii) Proposing a denoising algorithm based on the improved and distributed tensor train (IDTT) decomposition, which optimizes the tensor train(TT) decomposition in terms of parallel computation and low-rank estimation; (iii) Combining (i), (ii) and Light Gradient Boosting Machine (LightGBM) classification model, an efficient DDoS attack detection framework is proposed. Datasets CIC-DDoS2019 and NSL-KDD are used to evaluate the framework, and results demonstrate that accuracy can reach 99.19% while having the characteristics of low storage consumption and well speedup ratio.

Multi-layered solid element for seismic analysis of rubber bearings considering stress continuity based on finite particle method

Rubber bearings are crucial components in seismic isolation systems to mitigate the damaging effects of earthquakes on structures. Rubber bearings generally consist of alternate layers of steel and rubber bonded together. Simplified two-point spring models omit this layered construction and have limitations in adaptability and programmability, while refined multiple solid models lead to substantial computational costs. This study presents an effective and efficient approach to simulate rubber bearings by using multi-layered solid element based on the finite particle method (FPM). The traditional “Zigzag” shape function is extended into the 3D situation to reflect the serrated deformation within the layered construction. Based on the simplified assumption of stress continuity, an iterative approach is presented for the deformation gradient to achieve the force continuity between layers in dynamic nonlinear scenarios. The multi-layered solid models for both natural rubber bearings (NRBs) and lead rubber bearings (LRBs) are proposed based on this element. The errors of the hysteresis curves of rubber bearings are below 3.1 % between the numerical results obtained by using the multi-layered solid model and the experimental results. The computational speedup ratio of the proposed multi-layered solid model is verified to be 19 relative to the refined multiple solid model. The proposed method is applied to a base-isolated frame structure to demonstrate its effectiveness in structural seismic analysis.

Implementation and optimization of SpMV algorithm based on SW26010P many-core processor and stored in BCSR format

The irregular distribution of non-zero elements of large-scale sparse matrix leads to low data access efficiency caused by the unique architecture of the Sunway many-core processor, which brings great challenges to the efficient implementation of sparse matrix–vector multiplication (SpMV) computing by SW26010P many-core processor. To address this problem, a study of SpMV optimization strategies is carried out based on the SW26010P many-core processor. Firstly, we design a memorized data storage transformation strategy to transform the matrix in CSR storage format into BCSR (Block Compressed Sparse Row) storage. Secondly, the dynamic task scheduling method is introduced to the algorithm to realize the load balance between slave cores. Thirdly, the LDM memory is refined and designed, and the slave core dual cache strategy is optimized to further improve the performance. Finally, we selected a large number of representative sparse matrices from the Matrix Market for testing. The results show that the scheme has obviously speedup the processing procedure of sparse matrices with various sizes and sizes, and the master–slave speedup ratio can reach up to 38 times. The optimization method used in this paper has implications for other complex applications of the SW26010P many-core processor.

BagFormer: Better cross-modal retrieval via bag-wise interaction

In the field of cross-modal retrieval, single encoder models tend to perform better than dual encoder models, but they suffer from high latency and low throughput. In this paper, we propose a dual encoder model called BagFormer that utilizes bag-wise late interaction mechanism to improve re-rank performance without sacrificing latency and throughput. BagFormer achieves this by employing a bagging layer, which facilitates the transformation of text to an appropriate granularity. This not only mitigates the issue of modal granularity mismatch but also enables the integration of entity knowledge into the model. Our experiments have shown that BagFormerViT-B outperforms the traditional dual-encoder model CLIPViT-B by 7.97% in zero-shot settings. Under fine-tuned conditions, BagFormerViT-B demonstrates an even more significant improvement of 17.98% over CLIPViT-B. Moreover, BagFormer not only matches the performance of cutting-edge single-encoder models in cross-modal retrieval tasks but also provides efficient inference processes characterized by lower latency and higher throughput. Compared to single-encoder models, BagFormer can achieve a speedup ratio of 38.14 when re-ranking individual candidates. Code and models are available at github.com/howard-hou/BagFormer.

Development of a Spectrum-Based Scheme for Simulating Fine-Grained Sediment Transport in Estuaries

Fine-grained cohesive sediments in estuaries play a critical role in sediment transport and biogeochemical cycles in estuaries. Due to the convergence of marine saltwater and freshwater runoff, combined with periodic tidal cycles, fine-grained sediments exhibit intricate flocculation processes that are challenging to simulate. A size-resolved flocculation module using a bin-based scheme aids in modeling these processes but is hindered by high computational costs. In this study, we develop a new spectrum-based scheme based on the spectral shape of floc size distribution from the original bin-based scheme to expedite modeling execution. This new scheme is implemented in the Stony Brook Parallel Ocean Model (sbPOM) and applied to simulate fine-grained sediment transport in the Hudson River estuary. The effectiveness of this spectrum-based scheme is assessed by comparing its simulations with observations and results from the original bin-based scheme. The findings indicate that the new scheme can simulate the evolution of suspended sediment concentration well at a specific point by comparisons with in-situ observations. Specifically, the results of the 50 paired experiments show an average percentage difference of 1.86% and an average speedup ratio of 4.51 times compared to the original bin-based scheme. In summary, the new spectrum-based scheme offers significant acceleration benefits for the size-resolved flocculation module and has the potential for widespread application in simulating fine-grained sediments in estuaries.

Chance constrained optimal power-water flow: An iterative algorithm assisted by deep neural networks

The optimal power-water flow (OPWF) is a fundamental tool for operating integrated power and water networks. To address uncertainties in renewable energy and load demands, researchers have used chance-constrained models to investigate OPWF under uncertainty. However, existing work has limitations, such as oversimplified network representations, intractable problem formulations, and high computational burdens. This paper presents a chance-constrained OPWF formulation that incorporates detailed network constraints and stochastic uncertainties from wind speeds and electricity/water demands. By leveraging the concept of quantiles of stochastic variables and convexification techniques, the model is recast into a deterministic mixed-integer second-order cone programming (MISOCP) problem with uncertain margins. To efficiently solve this problem, we propose a novel iterative algorithm assisted by deep neural networks (DNNs). The algorithm integrates two DNNs: one designed to solve power and water flow equations, accelerating the probabilistic evaluation of uncertain margins; and the other to predict pipeline flows and refine variable ranges, facilitating a warm-start approach to the MISOCP model. Case studies conducted on the IEEE 123-node feeder coupled with a 67-node water distribution network validate the two DNNs and the overall iterative algorithm. The findings demonstrate that the proposed DNN-assisted approach effectively balances system security and economy. Compared to existing algorithms, the relative error of the optimization solution is less than 0.03%, with a speedup ratio exceeding 200.

Cross-Scale Modeling of Shallow Water Flows in Coastal Areas with an Improved Local Time-Stepping Method

A shallow water equations-based model with an improved local time-stepping (LTS) scheme is developed for modeling coastal hydrodynamics across multiple scales, from large areas to detailed local regions. To enhance the stability of the shallow water model for long-duration simulations and at larger LTS gradings, a prediction-correction method using a single-layer interface that couples coarse and fine time discretizations is adopted. The proposed scheme improves computational efficiency with an acceptable additional computational burden and ensures accurate conservation of time truncation errors in a discrete sense. The model performance is verified with respect to conservation and computational efficiency through two idealized tests: the spreading of a drop of shallow water and a tidal flat/channel system. The results of both tests demonstrate that the improved LTS scheme maintains precision as the LTS grading increases, preserves conservation properties, and significantly improves computational efficiency with a speedup ratio of up to 2.615. Furthermore, we applied the LTS scheme to simulate tides at grid scales of 40,000 m to 200 m for a portion of the Northwest Pacific. The proposed model shows promise for modeling cross-scale hydrodynamics in complex coastal and ocean engineering problems.

Deflected wind field over a two-dimensional steep ridge subjected to yawed inflow

In this study, large-eddy simulations (LES) are performed to investigate the turbulent wake flow of a two-dimensional ridge with cosine-squared cross-section for representative wind direction. The power law index of the mean velocity profile of the approaching turbulent boundary layer is 0.13. The variations of separation bubble, mean and fluctuating velocities, wind deflection angles, speed-up ratio and power spectra density with inflow angle are systematically investigated. One significant horizontal feature of the flow field is that the yawed inflow deflects further when it encounters the two-dimensional ridge. The yaw angles exhibit strong non-linear relationships with the inflow angle, especially at locations influenced by the separation bubble. The length of the separation bubble projected to the plane perpendicular to the ridge gradually decreases with an increasing inflow angle. Meanwhile, the longitudinal mean velocity above the ridge and the vertical mean velocity near the windward side decrease with a higher inflow angle. A lateral mean velocity emerges and strengthens as the inflow angle increases. In addition, peak frequency in the power spectrum of longitudinal velocity rises with increasing inflow angle, accompanied by higher energy concentrations near the peak frequency.

GPU-accelerated approach for 2D fracture analysis of structures combining finite particle method and cohesive zone model

The fracture analysis of structures typically involves strong discontinuity and nonlinear behaviors, and it requires fine meshing and small time steps, making it time-consuming. This study proposes an approach accelerated by graphics processing units (GPU) for two-dimensional (2D) fracture analysis of structures combining finite particle method (FPM) and cohesive zone model (CZM). Specifically, the equations of motion of particles are presented, and the internal forces of planar triangular elements are derived based on FPM. Subsequently, the internal forces of four-particle cohesive element are derived to describe the CZM, and the linear elastic stage, softening stage, and failure stage are depicted according to the traction-separation criteria. An explicit GPU-based FPM analysis strategy for structural fracture analysis is then developed, and the parallel solvers for particles, triangular elements, and cohesive elements are implemented. A quasi-static fracture analysis of a concrete beam and a dynamic fracture analysis of the Kalthoff problem is conducted to verify the effectiveness of the proposed approach. The obtained crack propagation path, load vs. displacement curves, and crack propagation speed are in good agreement with the experimental results and the results in the literature. The efficiency of the proposed approach is also investigated. The achieved maximum speedup ratio of the GPU-accelerated approach to the central processing unit (CPU)-based approach reaches around 24, and the achieved speedup ratio relative to the commercial finite element software Abaqus/Explicit reaches around 11.5, demonstrating the computational efficiency of the proposed approach.

A three-dimensional pseudo-potential multiphase model with super-large density ratio and adjustable surface tension

In this paper, an improved three-dimensional pseudo-potential multiphase flow model is proposed. A high-precision difference scheme is employed to improve the computational accuracy of the interaction forces to achieve super-large density ratio. The present model is verified to obtain two-phase density ratio up to hundreds of thousands for all selected equations of state, and the spurious currents can be suppressed to a relatively low level. A modification pressure tensor is introduced to implement a wide range of surface tension adjustments independently of the density ratio. The model is applied to simulate droplet impacts on a liquid film as well as droplet impacts on a dry surface. Numerical results show that the model still has good numerical stability in simulating complex fluid problems with very large density ratio, adjustable surface tension, high Reynolds number, or containing the wettable surface. In addition, to improve the computational efficiency, an efficient parallel algorithm based on graphics processing unit (GPU) is designed for the present model. A maximum speedup ratio of 687 times is obtained compared to the corresponding CPU-based serial algorithm, which can significantly accelerate the numerical simulation study.

Reduced-order methods for neutron transport kinetics problem based on proper orthogonal decomposition and dynamic mode decomposition

This work proposes two reduced-order methods to address the high computing resource consumption of neutron transport kinetics problems. The first method is based on proper orthogonal decomposition (POD), which integrates the POD basis with the governing equation. The second method is based on dynamic mode decomposition (DMD), which constructs an approximate linear relationship to predict the time-dependent physical quantities. Numerical results demonstrate that both two reduced-order methods can achieve high accuracy and efficient prediction. For the selected cases, POD can obtain a speed-up ratio of 105 – 429, while DMD can obtain a speed-up ratio of 2300 – 12500. For step transient situations, POD and DMD predict power errors do not exceed 0.42% and 0.05%. For ramp transient situations, POD and DMD predict power errors do not exceed 0.69% and 1.9%. This work can provide some useful suggestions for applications and further development in neutron transport kinetics reduced-order model.

Judgment criteria for significant wind speed-up regions in natural complex terrain

Flashover incidents due to wind-induced motion of overhead power line components are a not uncommon phenomenon during very strong winds, e.g. typhoons, and an important contributing factor is the speed-up effect caused by small-scale topographic features. In this study, wind speed-up ratios and significant speed-up regions in natural complex terrain are comprehensively analyzed from three perspectives: analysis of landform characteristics of flashover locations, Computational Fluid Dynamics (CFD) simulation, and wind tunnel experiments. Through the analysis of flashover locations, judgment criteria based on the elevation coefficient of variation are proposed to attempt to identify significant speed-up regions in natural complex terrain. Wind tunnel experiments and CFD simulations are conducted to analyze the wind speed-up ratios in 12 wind directions for both a small isolated mountainous area and a large continuous mountainous area. The significant speed-up regions are determined by consolidating the speed-up ratios from all 12 wind directions. The reliability of the judgment criteria is then validated. The results confirm that significant speed-up ratios are predominantly concentrated at the edges of mountainous areas, particularly at peaks and high ridges around areas with larger elevation coefficient of variation. These regions largely coincide with the locations where flashover incidents mainly occurred and the results from wind tunnel experiments and CFD simulations. In summary, the clearly identified significant speed-up regions include (1) small isolated mountainous areas; (2) peaks and high ridges around areas with larger elevation coefficient of variation within large continuous mountainous areas, and the higher crests within these regions. Based on these findings, the study suggests that wind speed-up ratios in design standards should be amplified in the vicinity of these regions to enhance safety measures.

LUAEMA: A Loop Unrolling Approach Extending Memory Accessing for Vector Very-Long-Instruction-Word Digital Signal Processor with Multiple Register Files

Loop unrolling can provide more instruction-level parallelism opportunities for code and enables a greater range of instruction pipeline scheduling. In high-performance very-long-instruction-word (VLIW) digital signal processors (DSPs), there are special registers to address. To further improve the instruction-level parallelism of code for such DSPs by making full use of these registers, in this paper, we propose a more effective loop unrolling approach through extending memory accessing (LUAEMA). In this approach, the final unrolling factor is computed by a model in which every register kind and every memory accessing operation are considered. For basic digital signal processing algorithms, the unrolling factor under the LUAEMA is larger than that under the conventional loop unrolling approach. We also provide the opportunity to reduce the number of instructions in a loop during the code transformation of loop unrolling. The experimental results show that the loop unrolling approach proposed in this paper can achieve an average speedup ratio ranging from 1.14 to 1.81 compared with the conventional loop unrolling approach. For some algorithms, the peak speedup ratio is up to 2.11.

A 3D virtual geographic environment for flood representation towards risk communication

Risk communication seeks to develop a shared understanding of disaster among stakeholders, thereby amplifying public awareness and empowering them to respond more effectively to emergencies. However, existing studies have overemphasized specialized numerical modelling, making the professional output challenging to understand and use by non-research stakeholders. In this context, this article proposes a 3D virtual geographic environment for flood representation towards risk communication, which integrates flood modelling, parallel computation, and 3D representation in a pipeline. Finally, a section of the Rhine River in Bonn, Germany, is selected for experiment analysis. The experimental results show that the proposed approach is capable of flood modelling and 3D representation within a few hours, the parallel speedup ratio reached 6.45. The intuitive flood scene with 3D city models is beneficial for promoting flood risk communication and is particularly helpful for participants without direct experience of floods to understand its spatiotemporal process. It also can be embedded in the Geospatial Infrastructure Management Ecosystem (GeoIME) cloud application for intelligent flood systems.

A universal parallel simulation framework for energy pipeline networks on high-performance computers

Energy distribution networks represent crucial infrastructures for modern society, and various simulation tools have been widely used by energy suppliers to manage these intricate networks. However, simulation calculations include a large number of fluid control equations, and computational overhead limits the performance of simulation software. This paper proposes a universal parallel simulation framework for energy pipeline networks that takes advantages of data parallelism and computational independence between network elements. A non-pipe model of an energy supply network is optimized, and the input and output of the network model in the proposed framework are modified, which can reduce the development burden during the numerical computations of the pipeline network and weaken the computational correlation between different simulated components. In addition, independent computations can be performed concurrently through periodic data exchange procedures between component instances, improving the parallelism and efficiency of simulation computations. Further, a parallel water pipelines network simulation computing paradigm based on a heterogeneous computer hardware architecture is used to evaluate the proposed framework’s performance. A series of tests are conducted to verify the accuracy of the proposed framework, and simulation errors of less than 5% are achieved. The results of multi-threaded simulation experiments have demonstrated the feasibility of the proposed framework in a parallel computing approach. Moreover, an Advanced Micro Devices (AMD) Deep Computing Unit (DCU)-parallel program is implemented into a water supply network simulation system; the computational efficiency of this system is compared with that of its serial counterpart. The experimental results show that the proposed framework is appropriate for high-performance computer architectures, and the 18x speed-up ratio demonstrates that the parallel program based on the proposed universal framework outperforms the serial program. That provides the basis for the application of pipe network simulation on high-performance computers.

Load-Balanced Parallel Implementation on GPUs for Multi-Scalar Multiplication Algorithm

Multi-scalar multiplication (MSM) is an important building block in most of elliptic-curve-based zero-knowledge proof systems, such as Groth16 and PLONK. Recently, Lu et al. proposed cuZK, a new parallel MSM algorithm on GPUs. In this paper, we revisit this scheme and present a new GPU-based implementation to further improve the performance of MSM algorithm. First, we propose a novel method for mapping scalars into Pippenger’s bucket indices, largely reducing the number of buckets compared to the original Pippenger algorithm. Second, in the case that memory is sufficient, we develop a new efficient algorithm based on homogeneous coordinates in the bucket accumulation phase. Moreover, our accumulation phase is load-balanced, which means the parallel speedup ratio is almost linear growth as the number of device threads increases. Finally, we also propose a parallel layered reduction algorithm for the bucket aggregation phase, whose time complexity remains at the logarithmic level of the number of buckets. The implementation results over the BLS12-381 curve on the V100 graphics card show that our proposed algorithm achieves up to 1.998x, 1.821x and 1.818x speedup compared to cuZK at scales of 221, 222, and 223, respectively.

Fragment-Based Deep Learning for Simultaneous Prediction of Polarizabilities and NMR Shieldings of Macromolecules and Their Aggregates.

Simultaneous prediction of the molecular response properties, such as polarizability and the NMR shielding constant, at a low computational cost is an unresolved issue. We propose to combine a linear-scaling generalized energy-based fragmentation (GEBF) method and deep learning (DL) with both molecular and atomic information-theoretic approach (ITA) quantities as effective descriptors. In GEBF, the total molecular polarizability can be assembled as a linear combination of the corresponding quantities calculated from a set of small embedded subsystems in GEBF. In the new GEBF-DL(ITA) protocol, one can predict subsystem polarizabilities based on the corresponding molecular wave function (thus electron density and ITA quantities) and DL model rather than calculate them from the computationally intensive coupled-perturbed Hartree-Fock or Kohn-Sham equations and finally obtain the total molecular polarizability via a linear combination equation. As a proof-of-concept application, we predict the molecular polarizabilities of large proteins and protein aggregates. GEBF-DL(ITA) is shown to be as accurate enough as GEBF, with mean absolute percentage error <1%. For the largest protein aggregate (>4000 atoms), GEBF-DL(ITA) gains a speedup ratio of 3 compared with GEBF. It is anticipated that when more advanced electronic structure methods are used, this advantage will be more appealing. Moreover, one can also predict the NMR chemical shieldings of proteins with reasonably good accuracy. Overall, the cost-efficient GEBF-DL(ITA) protocol should be a robust theoretical tool for simultaneously predicting polarizabilities and NMR shieldings of large systems.

Efficient Top-k Frequent Itemset Mining on Massive Data

Top-k frequent itemset mining (top-k FIM) plays an important role in many practical applications. It reports the k itemsets with the highest supports. Rather than the subtle minimum support threshold specified in FIM, top-k FIM only needs the more understandable parameter of the result number. The existing algorithms require at least two passes of scan on the table, and incur high execution cost on massive data. This paper develops a prefix-partitioning-based PTF algorithm to mine top-k frequent itemsets efficiently, where each prefix-based partition keeps the transactions sharing the same prefix item. PTF can skip most of the partitions directly which cannot generate any top-k frequent itemsets. Vertical mining is developed to process the partitions of vertical representation with the high-support-first principle, and only a small fraction of the items are involved in the processing of the partitions. Two improvements are proposed to reduce execution cost further. Hybrid vertical storage mode maintains the prefix-based partitions adaptively and the candidate pruning reduces the number of the explored candidates. The extensive experimental results show that, on massive data, PTF can achieve up to 1348.53 times speedup ratio and involve up to 355.31 times less I/O cost compared with the state-of-the-art algorithms.

Parallel Accelerating Number Theoretic Transform for Bootstrapping on a Graphics Processing Unit

The bootstrapping procedure has become the main bottleneck affecting the efficiency of all known fully homomorphic encryption (FHE) schemes. The state-of-the-art scheme for efficient bootstrapping, which is called fully homomorphic encryption over the torus (TFHE), accelerates polynomial multiplication by leveraging number theoretic transform (NTT) and implementing NTT in parallel on a GPU. Unfortunately, almost none of the recent advancements in NTT take full advantage of a GPU, leading to the need for more time. With this in mind, in this work, a novel faster number theoretic transform based on a GPU is proposed, in which matrix multiplication is used to implement a decomposed small-point NTT. When implementing matrix multiplication, we introduce a merging preprocessing method to merge multiple inputs of the small-point NTT, aiming to effectively minimize the count of modulo operations. Subsequently, when the merged result is multiplied by rotation factors, we use logical left shift rather than arithmetic multiplication to improve the computational efficiency. Our scheme can easily be used to realize a 1024-point NTT and the results of the experiments show that the speedup ratio of our method over the butterfly algorithm is about 2.49.