Graphics Processing Units Research Articles

PyC[formula omitted]Ray: A flexible and GPU-accelerated radiative transfer framework for simulating the cosmic epoch of reionization

Detailed modeling of the evolution of neutral hydrogen in the intergalactic medium during the Epoch of Reionization, 5≤z≤20, is critical in interpreting the cosmological signals from current and upcoming 21-cm experiments such as the Low-Frequency Array (LOFAR) and the Square Kilometre Array (SKA). Numerical radiative transfer codes provide the most physically accurate models of the reionization process. However, they are computationally expensive as they must encompass enormous cosmological volumes while accurately capturing astrophysical processes occurring at small scales (≲Mpc). Here, we present pyC2Ray, an updated version of the massively parallel ray-tracing and chemistry code, C2-Ray, which has been extensively employed in reionization simulations. The most time-consuming part of the code is calculating the hydrogen column density along the path of the ionizing photons. Here, we present the Accelerated Short-characteristics Octahedral ray-tracing (ASORA) method, a ray-tracing algorithm specifically designed to run on graphical processing units (GPUs). We include a modern Python interface, allowing easy and customized use of the code without compromising computational efficiency. We test pyC2Ray on a series of standard ray-tracing tests and a complete cosmological simulation with volume size (349Mpc)3, mesh size of 2503 and approximately 106 sources. Compared to the original code, pyC2Ray achieves the same results with negligible fractional differences, ∼10−5, and a speedup factor of two orders of magnitude. Benchmark analysis shows that ASORA takes a few nanoseconds per source per voxel and scales linearly for an increasing number of sources and voxels within the ray-tracing radii.

Sentiment analysis of social media comments based on multimodal attention fusion network

Social media comments are no longer in a single textual modality, but heterogeneous data in multiple modalities, such as vision, sound, and text, which is why multimodal sentiment analysis strategies has been introduced. However, among the multimodal sentiment analysis domains, a majority of the current multimodal sentiment analysis models employ the Transformer architecture due to its great impact and benefits, thereby leading to an augmentation in resource overhead. In this paper, a multimodal attention fusion (MAF) network model is proposed for sentiment analysis of multimodal data. MAF is mainly composed of the cross attention and residual unit. The Cross Attention Unit is designed to select one core modality out of three modes, while the remaining modes serve as base modalities. The core modality is then combined with the base modality information to facilitate significant interaction between the two modalities, resulting in three sets of two-by-two attention computations. Moreover, a residual unit is employed to integrate the overall information into the attentional information. This approach not only enables modality-to-modality interaction, but also supplements the overall information. In the end, experiments are conducted on two publicly available multimodal sentiment analysis datasets from Carnegie Mellon University(CMU), CMU-MOSEI (abbreviated as MOSEI) and CMU-MOSI (abbreviated as MOSI), to validate that the method achieves high performance while removing the complex structure, and is comparable to the State-Of-The-Art(SOTA) model with high-performance A100 and V100 Graphics Processing Units(GPU) in an ordinary hardware environment.

A High‐Performance Coupled Human And Natural Systems (CHANS) Model for Flood Risk Assessment and Reduction

AbstractIn recent years, flood risk in urban areas has been rapidly increasing due to unsustainable urban development, changes of hydrological processes and frequent occurrence of extreme weather events. Flood risk assessment should realistically take into account the complex interactions between human and natural systems to better inform risk management and improve resilience. In this study, we propose a novel Coupled Human And Natural Systems (CHANS) modeling framework to capture the intricate interactive human behaviors and flooding process at a high spatial resolution. The new CHANS modeling framework integrates a high‐performance hydrodynamic model with an agent‐based model to simulate the complex responses of individual households to the evolving flood conditions, leveraging the computing power of graphics processing units (GPUs) to achieve real‐time simulation. The framework is applied to reproduce the 2015 Desmond flood in the 2,500 km2 Eden Catchment in England, demonstrating its ability to predict interactive flood‐human dynamics and assess flood impact at the household‐level. The study also further explores the effectiveness of different flood risk management strategies, including the provision of early warning and distribution of sandbags, in mitigating flood impact. The new CHANS model potentially provides a useful tool for understanding short‐term human behaviors and their impact on flood risk during a flood event, which is important for the development of effective disaster risk management plans.

Optimizing energy expenditure in agricultural autonomous ground vehicles through a GPU-accelerated particle swarm optimization-artificial neural network framework

The accurate energy consumption prediction in Agricultural Ground Vehicles (AGVs) holds immense potential for optimizing operational efficiency and minimizing environmental impact. However, existing optimization methods for such prediction tasks often suffer from high computational demands, hindering their practical implementation. This paper introduces a ground-breaking approach that overcomes this limitation by leveraging the potent computational power of Graphics Processing Units (GPUs) to accelerate the optimization process dramatically. We propose a novel adaptation of the Particle Swarm Optimization (PSO) algorithm, specifically tailored to the intricate multi-objective challenges of AGV energy prediction. This framework harnesses the strengths of a multi-objective approach, enabling the simultaneous optimization of prediction accuracy and model complexity. To further enhance efficiency, we seamlessly integrate GPU parallelization techniques, significantly expediting both the optimization process and the training of Artificial Neural Networks (ANNs) employed for prediction. Preliminary results demonstrate a remarkable improvement in the accuracy of AGV energy consumption predictions, directly attributed to the synergistic effect of optimizing the ANN architecture and parameters through our proposed PSO framework. This tailored PSO adaptation distinguishes itself by its ability to tackle the complex multi-objective nature of AGV energy prediction with enhanced efficiency and precision. It thus emerges as a compelling and novel solution within the realm of Machine Learning and heuristic methods for agricultural robotics, paving the way for sustainable and optimal AGV operations.

Investigation of the bubble deflection using graphics processing unit accelerated multiple relaxation time lattice Boltzmann method: The effect of bubble size and liquid viscosity

Bubble deflection is important for bubble motion and bubble mass transfer in multiphase reactors. However, the mechanism triggering the instability is controversial. In this study, we first perform a graphics processing unit-accelerated multiple relaxation time lattice Boltzmann method (MRT-LBM) to simulate the motion process of single bubbles (1.8–8.0 mm) freely rising in stationary liquids with different viscosities. Then, we further analyze the influence of the bubble size and liquid viscosity on the characteristics of the time-dependent bubble motion, including the bubble shape, velocity, and deflection, based on the high-precision numerical simulation results. The results show that the formation and shedding of bubble wake vortices are responsible for the oscillation of the bubble shape. The interaction of bubble wake vortices and bubble shape is the primary cause of bubble deflection. Finally, an empirical model, taking into account the bubble acceleration and deflection angle, is developed to characterize the effect of bubble size and liquid viscosity on the drag coefficient of a rising bubble. This study provides insight into the bubble deflection and the bubble dynamics.

Deep Learning Algorithm Analysis of Potato Disease Classification for System on Chip Implementation

Recently, every aspect of human existence has been affected by modern technologies including agriculture. The broad range of crops has witnessed setbacks in different capacities due to climate change among other factors, hence leading to diseases and infections; thereby leading to negatively impacted nutrition. This work uses a deep learning algorithm to investigate the classification of potato diseases as a case study which can be leveraged on by other agricultural products for reference. Potatoes play a vital role in global agriculture, constituting a significant portion of fresh produce consumption. In southern Africa, potatoes hold particular importance, comprising 39% of total fresh produce consumption. However, the industry faces challenges from diseases such as early blight and late blight, resulting in substantial yield losses. Early detection is crucial, prompting the exploration of Convolutional Neural Networks (CNNs) for their disease classification capabilities. CNNs have shown excellent results in plant disease classification based on image data set. This study proposes the potential of aligning existing software-based Central Processing Units (CPUs) and Graphic Processing Units (GPUs) with FPGA-based potato disease classification using CNNs. To this end, five CNN models were trained on the Plant Village dataset for analysis. The models were compared using various performance metrics such as sensitivity, F1-Score, precision and recall. Based on the analysis the most suitable CNN model for FPGA implementation was selected. The chosen model was optimized for size using quantization technique. Subsequently the model was converted to Hardware Description Language (HDL) using Vitis High level synthesis (HLS) tool performance in terms of resources required to implement the model in System on Chip (SoC) was analysed.

Automatically optimized component model computation for power system simulation on GPU

This work provides an approach that automatically optimizes the component computations on graphics processing unit (GPU) devices from different vendors. The approach consists of a two-level optimization, where the first level considers the linear part of the computation for vectorization and applies mixed matrix formats to increase computational throughput further. Then, the second optimization level treats the combination of linear and non-linear parts as a black box and searches for the optimal configuration of parameters such as the degree of vectorization, the combination of matrix formats, and the group (of threads) sizes during parallel execution on GPU. Moreover, we also introduce constraints that reduce the optimization procedure’s execution time. Finally, we select three different types of components that could be representative to computational tasks in power system and perform our optimization approach on these kernels. The computational performance is compared with unoptimized baseline and sparse linear algebra library based implementations, result shows that our optimization leads to better performance and more efficient memory utilization.

3-center and 4-center 2-particle Gaussian AO integrals on modern accelerated processors.

We report an implementation of the McMurchie-Davidson (MD) algorithm for 3-center and 4-center 2-particle integrals over Gaussian atomic orbitals (AOs) with low and high angular momenta l and varying degrees of contraction for graphical processing units (GPUs). This work builds upon our recent implementation of a matrix form of the MD algorithm that is efficient for GPU evaluation of 4-center 2-particle integrals over Gaussian AOs of high angular momenta (l ≥ 4) [A. Asadchev and E. F. Valeev, J. Phys. Chem. A 127, 10889-10895 (2023)]. The use of unconventional data layouts and three variants of the MD algorithm allow for the evaluation of integrals with double precision and sustained performance between 25% and 70% of the theoretical hardware peak. Performance assessment includes integrals over AOs with l ≤ 6 (a higher l is supported). Preliminary implementation of the Hartree-Fock exchange operator is presented and assessed for computations with up to a quadruple-zeta basis and more than 20 000 AOs. The corresponding C++ code is part of the experimental open-source LibintX library available at https://github.com/ValeevGroup/libintx.

GENESIS 2.1: High-Performance Molecular Dynamics Software for Enhanced Sampling and Free-Energy Calculations for Atomistic, Coarse-Grained, and Quantum Mechanics/Molecular Mechanics Models.

GENeralized-Ensemble SImulation System (GENESIS) is a molecular dynamics (MD) software developed to simulate the conformational dynamics of a single biomolecule, as well as molecular interactions in large biomolecular assemblies and between multiple biomolecules in cellular environments. To achieve the latter purpose, the earlier versions of GENESIS emphasized high performance in atomistic MD simulations on massively parallel supercomputers, with or without graphics processing units (GPUs). Here, we implemented multiscale MD simulations that include atomistic, coarse-grained, and hybrid quantum mechanics/molecular mechanics (QM/MM) calculations. They demonstrate high performance and are integrated with enhanced conformational sampling algorithms and free-energy calculations without using external programs except for the QM programs. In this article, we review new functions, molecular models, and other essential features in GENESIS version 2.1 and discuss ongoing developments for future releases.

A novel hydrodynamic-water quality coupling model for high-efficiency and high-resolution simulations of urban NSPs

Nonpoint source pollution (NSP) is one of the main factors leading to black odor in urban rivers. A better understanding and simulation of the formation and transport mechanism of NSP is critical to the development of effective urban rainwater management and pollution control strategies. This work presents a novel hydrodynamic-water quality coupling model to quantitatively simulate the entire physical process of the urban NSP by integrating two-dimensional (2D) NSP build-up, wash-off and transport modules into the source code of the GPU-accelerated surface water and transport (GAST) model. The advantages of this model are that high-resolution terrain is considered, such as key microtopography, such as roads and houses, which will have a diversion effect on the transport of hydrodynamic water quality and affect, and detailed hydrodynamic processes based on physical processes are fully considered. Additionally, a graphical processing unit (GPU)-based acceleration method is introduced to increase the parallel performance of the computing process. Three scales of cases were used to verify the model. First, indoor experimental data from the model roof and measured data from the campus roof under two short-duration rainstorms were used for nonpoint source model verification. The Nash-Sutcliffe efficiency (NSE), root mean square error (RMSE), Bayesian information criterion (BIC) and Akaike information criterion (AIC) were utilized to assess the applicability of the different NSP wash-off equations added to this model. Based on this model, the NSP mechanism processes of residential buildings were examined. The results indicate that this model is suitable for various pollutant wash-off formulas and is capable of high-efficiency and robust predictions of the sources, transmission paths, and spatiotemporal distribution characteristics of urban nonpoint sources. We quantitatively simulated the generation and transport of NSP from a single roof and proposed an NSP traceability method based on detailed hydrodynamic processes. This model can provide strong technical support for the precise treatment of urban NSP and the construction of sponge facilities.

GPU-accelerated parallel image reconstruction strategies for magnetic particle imaging

Objective. Image reconstruction is a fundamental step in magnetic particle imaging (MPI). One of the main challenges is the fact that the reconstructions are computationally intensive and time-consuming, so choosing an algorithm presents a compromise between accuracy and execution time, which depends on the application. This work proposes a method that provides both fast and accurate image reconstructions. Approach. Image reconstruction algorithms were implemented to be executed in parallel in graphics processing units (GPUs) using the CUDA framework. The calculation of the model-based MPI calibration matrix was also implemented in GPU to allow both fast and flexible reconstructions. Main results. The parallel algorithms were able to accelerate the reconstructions by up to about 6,100 times in comparison to the serial Kaczmarz algorithm executed in the CPU, allowing for real-time applications. Reconstructions using the OpenMPIData dataset validated the proposed algorithms and demonstrated that they are able to provide both fast and accurate reconstructions. The calculation of the calibration matrix was accelerated by up to about 37 times. Significance. The parallel algorithms proposed in this work can provide single-frame MPI reconstructions in real time, with frame rates greater than 100 frames per second. The parallel calculation of the calibration matrix can be combined with the parallel reconstruction to deliver images in less time than the serial Kaczmarz reconstruction, potentially eliminating the need of storing the calibration matrix in the main memory, and providing the flexibility of redefining scanning and reconstruction parameters during execution.

Frequency-domain multiplexing of SNSPDs with tunable superconducting resonators

This work culminates in a demonstration of an alternative frequency-domain multiplexing (FDM) scheme for superconducting nanowire single-photon detectors (SNSPDs) using the kinetic inductance parametric up-converter (KPUP) made out of NbTiN. There are multiple multiplexing architectures for SNSPDs that are already in use, but FDM could prove superior in applications where the operational bias currents are very low, especially for mid-infrared and far-infrared SNSPDs. Previous FDM schemes integrated the SNSPD within the resonator, while, in this work, we use an external resonator, which gives more flexibility to optimize the SNSPD architecture. The KPUP is a DC-biased superconducting λ/2 resonator that is sensitive to current perturbations. When coupled to an SNSPD, the KPUP can be used to read out current pulses on a few μA scale. The KPUP is made out of NbTiN, which has a large kinetic inductance and a high operating temperature. Meanwhile, the SNSPD is made from WSi, which is a popular material for broadband SNSPDs. A software-defined radio platform and a graphics processing unit are used to read out the KPUP and SNSPD array. Frequency-domain multiplexed SNSPDs have applications in astronomy, remote sensing, exoplanet science, dark matter detection, and quantum sensing.

Mother-leaf-method accelerated parallel-GPU AMR phase-field simulations of dendrite growth

Parallel computing with multiple graphics processing units (GPUs) for adaptive mesh refinement (AMR) is a powerful method that improves the computational efficiency of phase-field (PF) simulations of dendrite growth. Octree-type block-structured AMR is often used in parallel computation because of its good parallelism; however, the process of filling the buffer region with physical data from adjacent blocks constitutes a major overhead cost that reduces computational performance. In this study, the mother-leaf (ML) method, which can reduce this overhead, was applied to accelerate parallel-GPU AMR PF simulations of dendrite growth. To evaluate the computational performance of the ML method, single- and parallel-GPU AMR PF simulations of dendrite growth were performed. The results show that single- and parallel-GPU AMR computations with the ML method accelerate the PF simulations of dendrite growth approximately 1.2–1.6 times and 1.2–1.4 times, respectively, compared to the conventional method without the ML method. In large-scale dendrite-growth PF simulations, such accelerations are significant for investigating various dendrite growth phenomena.

Analysis of major cannabinoids using Raman microscopy, density functional theory, chemometrics and a novel artificial intelligence approach

With a rise in the prominence of cannabis usage, due to its widespread availability and varying legal status, there has been an increased emphasis on the differentiation of cannabinoids present within cannabis using various analytical techniques. The present study aimed to exploit the capability of Raman microscopy to collect high-quality spectra of seven cannabinoid analytical standards, followed by their classification using linear discriminant analysis (LDA) and a novel transfer learning approach. Additionally, the experimental Raman spectra of delta-9-tetrahydrocannabinol (Δ9-THC), cannabidiol (CBD), and cannabichromene (CBC) were compared to simulated spectra from density functional theory calculations (DFT) to connect the spectral features to the underlying vibrational motions. A microscopical approach enabled the determination of the optimal sampling areas to collect Raman spectra for the nonacidic and acidic cannabinoids. An initial visualization of the data using principal component analysis (PCA) confirmed the spectral differences observable by visual comparisons of the spectra of the cannabinoid standards. The application of LDA implemented with a 5-fold cross-validation with 10 repeats, resulted in a classification accuracy of 99.83 %. For the transfer learning approach, the artificial intelligence (AI) model training was conducted in less than 10 min in a graphical processing unit (GPU) environment. All seven cannabinoids were successfully classified into respective classes based on scalograms transformed from Raman spectra, with 100 % classification accuracy. The average prediction probability for correct classification was 99.31 %. The classification outcome provided by the AI model included both prediction labels and probability, which provided a comprehensive evaluation of the samples.

Hashinator: a portable hybrid hashmap designed for heterogeneous high performance computing

Scientific computing has become increasingly parallel and heterogeneous with the proliferation of graphics processing unit (GPU) use in data centers, allowing for thousands of simultaneous calculations accessing high-bandwidth memory. Adoption of these resources may require re-design of scientific software. Hashmaps are a widely used data structure linking unsorted unique keys with values for fast data retrieval and storage. Several parallel libraries exist for performing hashmap operations utilizing GPU hardware, but none have yet supported GPUs and CPUs interchangeably. We introduce Hashinator, a novel portable hashmap designed to operate efficiently on both CPUs and GPUs using CUDA or HIP/ROCm Unified Memory, offering host access methods, in-kernel access methods, and efficient GPU offloading capability on both NVIDIA and AMD hardware. Hashinator utilizes open addressing with Fibonacci hashing and power-of-two capacity. By comparing against existing implementations, we showcase the excellent performance and flexibility of Hashinator, making it easier to port scientific codes that rely heavily on the use of hashmaps to heterogeneous architectures.

TORO Indexer: a PyTorch-based indexing algorithm for kilohertz serial crystallography.

Serial crystallography (SX) involves combining observations from a very large number of diffraction patterns coming from crystals in random orientations. To compile a complete data set, these patterns must be indexed (i.e. their orientation determined), integrated and merged. Introduced here is TORO (Torch-powered robust optimization) Indexer, a robust and adaptable indexing algorithm developed using the PyTorch framework. TORO is capable of operating on graphics processing units (GPUs), central processing units (CPUs) and other hardware accelerators supported by PyTorch, ensuring compatibility with a wide variety of computational setups. In tests, TORO outpaces existing solutions, indexing thousands of frames per second when running on GPUs, which positions it as an attractive candidate to produce real-time indexing and user feedback. The algorithm streamlines some of the ideas introduced by previous indexers like DIALS real-space grid search [Gildea, Waterman, Parkhurst, Axford, Sutton, Stuart, Sauter, Evans & Winter (2014). Acta Cryst. D70, 2652-2666] and XGandalf [Gevorkov, Yefanov, Barty, White, Mariani, Brehm, Tolstikova, Grigat & Chapman (2019). Acta Cryst. A75, 694-704] and refines them using faster and principled robust optimization techniques which result in a concise code base consisting of less than 500 lines. On the basis of evaluations across four proteins, TORO consistently matches, and in certain instances outperforms, established algorithms such as XGandalf and MOSFLM [Powell (1999). Acta Cryst. D55, 1690-1695], occasionally amplifying the quality of the consolidated data while achieving superior indexing speed. The inherent modularity of TORO and the versatility of PyTorch code bases facilitate its deployment into a wide array of architectures, software platforms and bespoke applications, highlighting its prospective significance in SX.

Efficient and Fast Algorithm for Attribute Reduction of Large Dimensional Data Using Rough Set Theory on Graphics Processing Unit

Efficient and Fast Algorithm for Attribute Reduction of Large Dimensional Data Using Rough Set Theory on Graphics Processing Unit

Subspace graph networks for real-time granular flow simulation with applications to machine-terrain interactions

Granular flows interacting with rigid bodies are key to machine-terrain interactions in robotics and related fields, and still pose several open problems. Continuum methods and deep learning, separately and in combination, show promise. This research advances machine learning methods for modeling rigid body-driven granular flows, for terrestrial industrial machine and planetary rover (where gravity is an important factor) applications. The original contribution of this research is the application of a subspace machine learning simulation approach for these engineering problems. To generate training datasets, a high-fidelity continuum method, the material point method (MPM) is utilized. Principal component analysis (PCA) is used to reduce the dimensionality of data. It is shown that the first few (8 in our datasets) principal components of our high-dimensional data keep almost the entire variance in data. A graph network simulator (GNS) is trained to learn the underlying subspace dynamics. The learned GNS is then able to predict particle positions and interaction forces with good accuracy, averaging 6×10−5 and 3×10−6 mean squared position error and 7% and 17% mean percentage force error, for excavation and wheel data, respectively. More importantly, PCA significantly enhances the time and memory efficiency of GNS in both training and rollout. This enables GNS to be trained using a single desktop graphics processing unit (GPU) with moderate video random-access memory (VRAM). This also makes the GNS real-time on large-scale 3D physics configurations (700x faster than our continuum method), which is the first demonstration of such significantly accelerated high-fidelity granular flow engineering simulations.

Multiple-relaxation-time lattice boltzmann simulation of natural convection of ethylene Glycol -Al[formula omitted]O[formula omitted] power-law Non-newtonian nanofluid in an open enclosure with adiabatic fins

The heat transfer by natural convection of a nanofluid, which is ethylene glycol−Al2O3 has been analyzed in an open cavity numerically using the multiple-relaxation-time - lattice Boltzmann method by the graphics processing unit high-performance parallel computing. The right side of the cavity is open, and different boundary conditions have been applied to all the walls. Besides, one adiabatic fin has been installed on each side of the enclosure’s top and bottom sides. Here, the Prandtl number is fixed at 16.6, and the Rayleigh number changes from 104−106 with the nanoparticle volume fraction from 0%−5% has been used for numerical simulations. Besides, in this work, the power-law index is an important parameter as well, and 0.7, 0.8, 1, 1.2, and 1.4 are the values of this parameter. Results are presented concerning both the average and local Nusselt numbers in the form of streamlines, isotherms, temperature distributions, velocity distributions, heat transfer rate, and entropy production. It is observed when increases, average Nusselt number increases 607.94%, and for this reason, the overall heat transfer rate rises because of buoyancy force. In addition, the average Nusselt number falls by 83.28% when the power-law index rises; as a result, the total heat transfer rate falls because fluid viscosity increases with the power-law index. It is also observed that for shear-thickening fluids, the temperature gradient is higher. On the contrary, the temperature started decreasing with the increase of the power-law index. Additionally, the local Nusselt number value rises as power-law index falls. Moreover, the heat transfer rate increases by 7.08% when volume fraction increases. The intensity of buoyancy force reduces with the increase of volume fraction. Besides, the overall entropy generation rises when the Rayleigh number and the volume fraction increase, but it decreases when the power-law index increases. So, when the Rayleigh number is 106, the power-law index is 0.7, and the volume fraction is 0.00 then the entropy generation is the highest. This current research has many applications for example heat exchangers, electronic cooling equipment, solar heating systems, aerospace applications, medical devices, and entropy generation-related systems.

Matrix‐based implementation and GPU acceleration of hybrid FEM and peridynamic model for hydro‐mechanical coupled problems

AbstractThe hybrid finite element‐peridynamic (FEM‐PD) models have been evidenced for their exceptional ability to address hydro‐mechanical coupled problems involving cracks. Nevertheless, the non‐local characteristics of the PD equations and the required inversion operations when solving fluid equations result in prohibitively high computational costs. In this paper, a fast explicit solution scheme for FEM‐PD models based on matrix operation is introduced, where the graphics processing units (GPUs) are used to accelerate the computation. An in‐house software is developed in MATLAB in both CPU and GPU versions. We first solve a problem related to pore pressure distribution in a single crack, demonstrating the accuracy of the proposed method by a comparison of FEM‐PD solutions with those of PD‐only models and analytical solutions. Subsequently, several examples are solved, including a one‐dimensional dynamic consolidation problem and the fluid‐driven hydraulic fracture propagation problems in both 2D and 3D cases, to comprehensively validate the effectiveness of the proposed methods in simulating deformation and fracture in saturated porous media under the influence of hydro‐mechanical coupling. In the presented numerical results, some typical strong dynamic phenomena, such as stepwise crack advancement, crack branching, and pressure oscillations, are observed. In addition, we compare the wall times of all the cases executed on both the GPU and CPU, highlighting the substantial acceleration performance of the GPU, particularly when tackling problems with a significant computational workload.