GPGPU Research Articles

Development and application of a three-dimensional GPGPU-parallelized FDEM for modelling rock fragmentation by blast

Numerical modelling of rock fragmentation by blast is an important step to the optimum fragmentation. Combined finite-discrete element method (FDEM) has recently been proven to be one of the most promising methods for modelling the fracture and fragmentation of rocks under various loading conditions including rock blasting. However, almost all FDEM modellings of rock blasts are conducted in two-dimension. Correspondingly, this study implements a three-dimensional (3D) hybrid finite-discrete element method (HFDEM) parallelized on the basis of the general-purpose graphic-processing-unit (GPGPU) to model the fracture and fragmentation process of rock by blast. The main components of HFDEM3D include robust fracturing algorithm, efficient contact activation approach and various implementations of gas and rock interactions, which are firstly briefly reviewed here and introduced in detail in authors’ various former publications. The GPGPU-parallelized HFDEM3D is then applied to model the rock fracture and fragmentation process by the detonation of a single borehole and simultaneous and consecutive detonations of multiple boreholes.

Globally Consistent 3D LiDAR Mapping With GPU-Accelerated GICP Matching Cost Factors

This paper presents a real-time 3D LiDAR mapping framework based on global matching cost minimization. The proposed method constructs a factor graph that directly minimizes matching costs between frames over the entire map, unlike pose graph-based approaches that minimize errors in the pose space. For real-time global matching cost minimization, we use a voxel data association-based GICP matching cost factor that is able to fully leverage GPU parallel processing. The combination of the matching cost factor and GPU computation enables constraint of the relative pose between frames with a small overlap and creation of a densely connected factor graph. The mapping process is managed based on a voxel-based overlap metric that can quickly be evaluated on a GPU. We incorporate the proposed method with an external loop detection method in order to help the voxel-based matching cost factors to avoid convergence in a local solution. The experimental result on the KITTI dataset shows that the proposed approach improves the estimation accuracy of long trajectories.

A Modified RSA Algorithm Using Machine Learning and DNA Cryptography Techniques

RSA (Rivest Shamir Adleman) is an asymmetric algorithm used for encryption and decryption of data to ensure security through a communication channel. RSA algorithm is preferred over symmetric algorithms as it employs asymmetric keys for encryption and decryption. In this work, RSA optimization is attempted at three stages: key generation, data encryption, and data decryption. Key generation is implemented using the Gradient descent algorithm and Newton's method. Prime numbers required in key generation, data encryption, and data decryption are generated using GPU computations using OpenCL and CPU as the host providing a heterogeneous platform. The second and third stages namely data encryption and data decryption are implemented using the DNA cryptographic techniques and gray coding. The modified RSA designed as depicted above uses a low-key length of 8 bits and is validated for performance through avalanche effect, encryption simulation time, and decryption simulation time. The modified RSA algorithm is further verified by analyzing the chosen ciphertext attack time mathematically for various key lengths from 8 to 2048. The increase in attack time is around 93% for a key length of 8 bits and 78% for 15 bits for a key length of 2048 it increases by 100%.

Development of software and algorithms of parallel learning of artificial neural networks using CUDA technologies

The object of research is to parallelize the learning process of artificial neural networks to automate the procedure of medical image analysis using the Python programming language, PyTorch framework and Compute Unified Device Architecture (CUDA) technology. The operation of this framework is based on the Define-by-Run model. The analysis of the available cloud technologies for realization of the task and the analysis of algorithms of learning of artificial neural networks is carried out. A modified U-Net architecture from the MedicalTorch library was used. The purpose of its application was the need for a network that can effectively learn with small data sets, as in the field of medicine one of the most problematic places is the availability of large datasets, due to the requirements for data confidentiality of this nature. The resulting information system is able to implement the tasks set before it, contains the most user-friendly interface and all the necessary tools to simplify and automate the process of visualization and analysis of data. The efficiency of neural network learning with the help of the central processor (CPU) and with the help of the graphic processor (GPU) with the use of CUDA technologies is compared. Cloud technology was used in the study. Google Colab and Microsoft Azure were considered among cloud services. Colab was first used to build a prototype. Therefore, the Azure service was used to effectively teach the finished architecture of the artificial neural network. Measurements were performed using cloud technologies in both services. The Adam optimizer was used to learn the model. CPU duration measurements were also measured to assess the acceleration of CUDA technology. An estimate of the acceleration obtained through the use of GPU computing and cloud technologies was implemented. CPU duration measurements were also measured to assess the acceleration of CUDA technology. The model developed during the research showed satisfactory results according to the metrics of Jaccard and Dyce in solving the problem. A key factor in the success of this study was cloud computing services.

SW2D-GPU: A two-dimensional shallow water model accelerated by GPGPU

Shallow water models are used for simulating flood and lake hydrodynamics. However, the computational cost of those models is often high and require high performance computing. We present the SW2D-GPU: a two-dimensional shallow water model accelerated by General Purpose Graphics Processing Unit. The model is implemented in parallel using CUDA C/C++. We exemplify the use of the model with two case studies: (i) Flood simulation in an urban area and (ii) water level simulation in a lake catchment. We have included potential evaporation in the formulation which expands its application to water level simulations in lakes and reservoirs. The SW2D-GPU model is approximately 34 times faster than its equivalent sequential version. The model can be run in any computer equipped with a NVIDIA GPU. Integrated simulations of surface waters in lake watershed and simulations of floods caused by dam break are some of the potential applications.

Experimental and numerical studies on failure behaviours of sandstones subject to freeze-thaw cycles

The freeze-thaw induced damage of rock affects the durability and serviceability of geo-structures, especially those constructed in the regions frequently impacted by climatic changes. A series of laboratory tests including P-wave velocity tests, freeze-thaw tests, uniaxial compression strength (UCS) tests and Brazilian tensile strength (BTS) tests are conducted to investigate the physical-mechanical properties and failure behaviours of Tasmanian sandstones subjected to various freeze-thaw cycles. It is observed that the P-wave velocity, BTS and UCS of the sandstone decrease as the number of freeze-thaw cycles increases, in which the decreasing rate from 0 to 20 freeze-thaw cycles was more pronounced than that from 20 to 40 and 40 to 60 freeze-thaw cycles. Moreover, it is found that the main failure mode of the sandstone changes from axial splitting to shearing along a single plane in the UCS tests and from central smooth fractures to a central zigzag fracture in the BTS tests with the number of freeze-thaw cycles increasing. Three-dimensional (3D) numerical modellings are then conducted using a self-developed 3D hybrid finite-discrete element method (HFDEM) parallelized on the basis of general-purpose graphic processing units (GPGPU) to further investigate the failure mechanisms of Tasmanian sandstones subjected to various freeze-thaw cycles in the UCS and BTS tests. The 3D numerical modellings agree very well with the experimental observations that the physical-mechanical parameters of the sandstone degrade with the increasing number of the freeze-thaw cycles. Moreover, the 3D numerical modellings reveal the deterioration and failure mechanisms of sandstones subjected to various freeze-thaw cycles. For the sandstone specimens without subjecting to freeze-thaw cycles, axial splitting is the main failure pattern while tensile and mixed-mode damages are the dominant failure mechanism in the UCS tests. For the sandstone subjecting to various freeze-thaw cycles, the increasing number of freeze-thaw cycles causes the macroscopic cracks to propagate, interact and coalesce in the shear behaviour resulting in the final shear fracture pattern in the UCS test. The 3D numerical modellings of the BTS test show that, although, for both the models with and without subjecting to freezing and thawing cycles, a central fracture is the eventual failure pattern, the failure surface becomes more zigzag as the number of freeze-thaw cycles increases.

Measurement and analysis of GPU-accelerated applications with HPCToolkit

To address the challenge of performance analysis on the US DOE’s forthcoming exascale supercomputers, Rice University has been extending its HPCToolkit performance tools to support measurement and analysis of GPU-accelerated applications. To help developers understand the performance of accelerated applications as a whole, HPCToolkit’s measurement and analysis tools attribute metrics to calling contexts that span both CPUs and GPUs. To measure GPU-accelerated applications efficiently, HPCToolkit employs a novel wait-free data structure to coordinate monitoring and attribution of GPU performance. To help developers understand the performance of complex GPU code generated from high-level programming models, HPCToolkit constructs sophisticated approximations of call path profiles for GPU computations. To support fine-grained analysis and tuning, HPCToolkit uses PC sampling and instrumentation to measure and attribute GPU performance metrics to source lines, loops, and inlined code. To supplement fine-grained measurements, HPCToolkit can measure GPU kernel executions using hardware performance counters. To provide a view of how an execution evolves over time, HPCToolkit can collect, analyze, and visualize call path traces within and across nodes. Finally, on NVIDIA GPUs, HPCToolkit can derive and attribute a collection of useful performance metrics based on measurements using GPU PC samples. We illustrate HPCToolkit’s new capabilities for analyzing GPU-accelerated applications with several codes developed as part of the Exascale Computing Project.

A GPU-Parallel Image Coregistration Algorithm for InSar Processing at the Edge.

Image Coregistration for InSAR processing is a time-consuming procedure that is usually processed in batch mode. With the availability of low-energy GPU accelerators, processing at the edge is now a promising perspective. Starting from the individuation of the most computationally intensive kernels from existing algorithms, we decomposed the cross-correlation problem from a multilevel point of view, intending to design and implement an efficient GPU-parallel algorithm for multiple settings, including the edge computing one. We analyzed the accuracy and performance of the proposed algorithm—also considering power efficiency—and its applicability to the identified settings. Results show that a significant speedup of InSAR processing is possible by exploiting GPU computing in different scenarios with no loss of accuracy, also enabling onboard processing using SoC hardware.

How to Transfer Strided Data into Accelerator

Data transfer from a host central processing unit (CPU) into an accelerator is a performance bottleneck for applications accelerated by accelerators (such as general purpose digital signal processing (GPDSP), many integrated core (MIC), and general purpose graphics processing unit (GPGPU)). It is complicated and inefficient to transfer non-contiguous data with special respect to strided data. In this work, we present three approaches to transfer strided data for different scenarios: Redundant copy (RC), selective copy (SC), and transfer after transformed (TaT). We propose a space and time efficient method named TaT, in which strided data are transformed on the CPU first and then transferred into the accelerator. We simulated regions-of-interest (ROI) coding and validate proposed techniques. TaT was superior to RC on space efficiency and close to SC on saving space, but better than SC on time waste respectively.

Fiber Clustering Acceleration With a Modified Kmeans++ Algorithm Using Data Parallelism

Fiber clustering methods are typically used in brain research to study the organization of white matter bundles from large diffusion MRI tractography datasets. These methods enable exploratory bundle inspection using visualization and other methods that require identifying brain white matter structures in individuals or a population. Some applications, such as real-time visualization and inter-subject clustering, need fast and high-quality intra-subject clustering algorithms. This work proposes a parallel algorithm using a General Purpose Graphics Processing Unit (GPGPU) for fiber clustering based on the FFClust algorithm. The proposed GPGPU implementation exploits data parallelism using both multicore and GPU fine-grained parallelism present in commodity architectures, including current laptops and desktop computers. Our approach implements all FFClust steps in parallel, improving execution times in all of them. In addition, our parallel approach includes a parallel Kmeans++ algorithm implementation and defines a new variant of Kmeans++ to reduce the impact of choosing outliers as initial centroids. The results show that our approach provides clustering quality results very similar to FFClust, and it requires an execution time of 3.5 s for processing about a million fibers, achieving a speedup of 11.5 times compared to FFClust.

"Accelerating a Geostatistical Approach to Groundwater Pollution Source Identification with GPU Computing"

"Accelerating a Geostatistical Approach to Groundwater Pollution Source Identification with GPU Computing"

GPU accelerated parallel reliability-guided digital volume correlation with automatic seed selection based on 3D SIFT

Digital volume correlation (DVC) is a powerful and widely used technique for measuring the internal 3D deformation field of a wide range of materials. One of the most popular DVC algorithms is the reliability-guided DVC (RG-DVC) which is good at dealing with large continuous deformation. However, RG-DVC requires a manually specified seed from which computation starts, and suffers from the efficiency due to a huge amount of computation and data dependency. This paper proposes a GPU accelerated parallel reliability-guided DVC algorithm (CuSIFT-RGDVC) on CUDA, which leverages 3D scale-invariant feature transform (3D SIFT) to assist seed selection to realize fully automation and improves performance utilizing GPU computing. In CuSIFT-RGDVC, reliability-guided displacement tracking (RGDT) is rewritten using sorted array-based batch processing mechanism which is a globally sequential locally parallel model, and multi-granularity parallelism is adopted to maximize GPU utilization. The empirical result shows that the proposed CuSIFT-RGDVC provides up to 29.1x speedup compared with our multi-threaded implementation and achieves the same level of computation speed as the state-of-the-art path-independent DVC without sacrificing accuracy.

A novel GPGPU-parallelized contact detection algorithm for combined finite-discrete element method

This paper presents a novel algorithm for contact detection for the three-dimensional combined finite-discrete element method (3D FDEM). The contact detection consists of two phases, neighbor search and fine search. Both phases are fully parallelized with Compute Unified Device Architecture (CUDA). In contrast to the non-binary search (NBS) algorithm, the proposed neighbor search algorithm is not sensitive to the element size distribution. To avoid the algorithm being performed at every step, an efficient GPGPU (general purpose graphic processing unit) -parallelized velocity reduction algorithm is employed. For the fine search, the concepts of “positive faces” and “positive edges” are proposed, and the process is considerably simplified and runs faster than the 3D separation axis algorithm. Numerical tests are performed to validate the efficiency and effectiveness of the proposed algorithms. The result shows that for both phases, the computation time is linearly proportional to the number of potential contact pairs regardless of the element size distribution. Furthermore, by using NVIDIA Telsa V100, the resultant overall speed-up ratios of the proposed contact detection algorithms relative to the original Y3D version for uniform and non-uniform element size distributions can reach up to 1982.6 and 13,894.7, respectively. For both quasi-static and dynamic problems, the simulated fracture patterns are in good agreement with the results generated by the NBS algorithm. The proposed methodology can be also employed for 2D FDEM and discrete element method (DEM).

Breast Cancer Segmentation Methods: Current Status and Future Potentials.

Early breast cancer detection is one of the most important issues that need to be addressed worldwide as it can help increase the survival rate of patients. Mammograms have been used to detect breast cancer in the early stages; if detected in the early stages, it can drastically reduce treatment costs. The detection of tumours in the breast depends on segmentation techniques. Segmentation plays a significant role in image analysis and includes detection, feature extraction, classification, and treatment. Segmentation helps physicians quantify the volume of tissue in the breast for treatment planning. In this work, we have grouped segmentation methods into three groups: classical segmentation that includes region-, threshold-, and edge-based segmentation; machine learning segmentation; and supervised and unsupervised and deep learning segmentation. The findings of our study revealed that region-based segmentation is frequently used for classical methods, and the most frequently used techniques are region growing. Further, a median filter is a robust tool for removing noise. Moreover, the MIAS database is frequently used in classical segmentation methods. Meanwhile, in machine learning segmentation, unsupervised machine learning methods are more frequently used, and U-Net is frequently used for mammogram image segmentation because it does not require many annotated images compared with other deep learning models. Furthermore, reviewed papers revealed that it is possible to train a deep learning model without performing any preprocessing or postprocessing and also showed that the U-Net model is frequently used for mammogram segmentation. The U-Net model is frequently used because it does not require many annotated images and also because of the presence of high-performance GPU computing, which makes it easy to train networks with more layers. Additionally, we identified mammograms and utilised widely used databases, wherein 3 and 28 are public and private databases, respectively.

Verification and validation of the high-performance Lorentz-orbit code for use in stellarators and tokamaks (LOCUST)

A novel high-performance computing algorithm, developed in response to the next generation of computational challenges associated with burning plasma regimes in ITER-scale tokamak devices, has been tested and is described herein. The Lorentz-orbit code for use in stellarators and tokamaks (LOCUST) is designed for computationally scalable modelling of fast-ion dynamics, in the presence of detailed first wall geometries and fine 3D magnetic field structures. It achieves this through multiple levels of single instruction, multiple thread parallelism and by leveraging general-purpose graphics processing units. This enables LOCUST to rapidly track the full-orbit trajectories of kinetic Monte Carlo markers to deliver high-resolution fast-ion distribution functions and plasma-facing component power loads. LOCUST has been tested against the prominent NUBEAM and ASCOT fast-ion codes. All codes were compared for collisional plasmas in both high and low-aspect ratio toroidal geometries, with full-orbit and guiding-centre tracking. LOCUST produces statistically consistent results in line with acceptable theoretical and Monte Carlo uncertainties. Synthetic fast-ion D-α diagnostics produced by LOCUST are also compared to experiment using FIDASIM and show good agreement.

Modeling granular material dynamics and its two-way coupling with moving solid bodies using a continuum representation and the SPH method

We outline a continuum approach for treating discrete granular flows that holds across multiple scales: from experiments that focus on centimeter-size control volumes, to tests that involve landslides and large buildings. The time evolution of the continuum used to capture the granular dynamics is resolved in space via the smoothed particle hydrodynamics (SPH) method. The interaction between the granular material and immersed rigid bodies is posed and solved as a “fluid”–solid interaction (FSI) problem using boundary conditions enforcing (BCE) SPH particles rigidly attached onto the boundary of the body interacting with the granular material. A new penetration-based particle shifting technique (PPST) is proposed to enforce the particle regularity and thus a stable simulation. Several numerical experiments (angle of repose, ball drop, and cone penetration) are carried out to validate the accuracy of the proposed methodology. The approach is subsequently demonstrated in conjunction with a 3D landslide simulation and a plowing operation. The approach discussed has been implemented and can be used in an open source simulation platform publicly available on GitHub. The implementation leverages GPU computing.

Hydrodynamic force analysis of magnetite medium inside dense medium cyclone using multiphase GPU parallelized ASM model

In order to tackle the high compute cost of the multiphase simulations for dense medium cyclones, the development of an open source high fidelity LES based algebraic slip mixture multiphase flow solver is presented. The solver is parallelized on a general purpose GPU using a full load model, which provides around 13 times speed up as compared to a CPU implementation thereby, making design explorations using the current solver more feasible and less time consuming. The traditional ASM model was improved upon using more realistic slurry viscosity model and polydispersed hindered correlation for dense medium operation. The performance of the developed solver for dense medium cyclone operation was validated with the literature based gamma ray tomography (GRT) experimental results. The solver was then used to study the implication of variation in the key aspects of DMC operation such as pressure head, underflow diameter and feed particle concentration. The medium segregation is explained based on a novel non-dimensional force analysis of the drag, shear lift and turbulent dispersion. Finally, the response of the DMC in terms of hydrodynamic force analysis and density differential with the variation in design and operating variables has discussed. Based on the comprehensive force analysis of the medium a justification is provided as to why a lower pressure head is preferred for DMCs used in mineral processing.

Symbolic and Numeric Kernel Division for Graphics Processing Unit-Based Finite Element Analysis Assembly of Regular Meshes With Modified Sparse Storage Formats

Abstract This paper presents an efficient strategy to perform the assembly stage of finite element analysis (FEA) on general purpose graphics processing units (GPUs). This strategy involves dividing the assembly task using symbolic and numeric kernels, and thereby reducing the complexity of the standard single-kernel assembly approach. Two sparse storage formats based on the proposed strategy are also developed by modifying the existing sparse storage formats with the intention of removing the degrees-of-freedom-based redundancies in the global matrix. The inherent problem of race condition is resolved through the implementation of coloring and atomics. The proposed strategy is compared with the state-of-the-art GPU-based and central processing unit (CPU)-based assembly techniques. These comparisons reveal a significant number of benefits in terms of reducing storage space requirements and execution time and increasing performance (GFLOPS). Moreover, using the proposed strategy, it is found that the coloring method is more effective compared to the atomics-based method for the existing as well as the modified storage formats.

Efficient parallelization of SPH algorithm on modern multi-core CPUs and massively parallel GPUs

Smoothed Particle Hydrodynamics (SPH) is fast emerging as a practically useful computational simulation tool for a wide variety of engineering problems. SPH is also gaining popularity as the back bone for fast and realistic animations in graphics and video games. The Lagrangian and mesh-free nature of the method facilitates fast and accurate simulation of material deformation, interface capture, etc. Typically, particle-based methods would necessitate particle search and locate algorithms to be implemented efficiently, as continuous creation of neighbor particle lists is a computationally expensive step. Hence, it is advantageous to implement SPH, on modern multi-core platforms with the help of High-Performance Computing (HPC) tools. In this work, the computational performance of an SPH algorithm is assessed on multi-core Central Processing Unit (CPU) as well as massively parallel General Purpose Graphical Processing Units (GP-GPU). Parallelizing SPH faces several challenges such as, scalability of the neighbor search process, force calculations, minimizing thread divergence, achieving coalesced memory access patterns, balancing workload, ensuring optimum use of computational resources, etc. While addressing some of these challenges, detailed analysis of performance metrics such as speedup, global load efficiency, global store efficiency, warp execution efficiency, occupancy, etc. is evaluated. The OpenMP and Compute Unified Device Architecture[Formula: see text] parallel programming models have been used for parallel computing on Intel Xeon[Formula: see text] E5-[Formula: see text] multi-core CPU and NVIDIA Quadro M[Formula: see text] and NVIDIA Tesla p[Formula: see text] massively parallel GPU architectures. Standard benchmark problems from the Computational Fluid Dynamics (CFD) literature are chosen for the validation. The key concern of how to identify a suitable architecture for mesh-less methods which essentially require heavy workload of neighbor search and evaluation of local force fields from neighbor interactions is addressed.

High-Performance Interactive Scientific Visualization With Datoviz via the Vulkan Low-Level GPU API

We reported initial work towards a new fast and scalable scientific visualization technology that leverages the Vulkan API to achieve unprecedented performance through GPUs. This technology is implemented in a C/C++ library called Datoviz that offers an intermediate-level API for scientific visualization libraries and software. Datoviz provides a unified graphics stack for 2-D, 3-D, graphical user interfaces, and natively supports efficient interactions between rendering and general-purpose GPU computing. A major direction of development is to investigate the integration of Datoviz as a low-level backend of a future version of VisPy, a popular Python scientific plotting library