Power Of Graphics Processing Units Research Articles

The impact of stochastic microstructures on the macroscopic fracture properties of brick and mortar composites

This paper examines the effect of non-uniform microstructures on the macroscopic fracture properties of idealized brick and mortar composites, which consist of rigid bricks bonded with elastic–plastic mortar that ruptures at finite strain. A simulation tool that harnesses the parallel processing power of graphics processing units (GPUs) was used to simulate fracture in virtual specimens, whose microstructures were generated by sampling a probability distribution of brick sizes. In the simulations, crack advance is a natural outcome of local ruptures in the cohesive zones bonding the bricks: the macroscopic initiation toughness for small-scale yielding is inferred by correlating the critical load needed to advance a pre-defined crack with an associated far-field energy release rate. Quantitative connections between the statistical parameters defining heterogeneous brick distributions and the statistics of initiation toughness are presented. The nature of crack tip damage and stresses ahead of the crack tip are illustrated as a function of brick size variability. The results offer quantitative insights that can be used to identify microstructural targets for process development, notably specific brick size distributions that still provide macroscopic toughening.

Libcloudph++ 1.0: a single-moment bulk, double-moment bulk, and particle-based warm-rain microphysics library in C++

Abstract. This paper introduces a library of algorithms for representing cloud microphysics in numerical models. The library is written in C++, hence the name libcloudph++. In the current release, the library covers three warm-rain schemes: the single- and double-moment bulk schemes, and the particle-based scheme with Monte Carlo coalescence. The three schemes are intended for modelling frameworks of different dimensionalities and complexities ranging from parcel models to multi-dimensional cloud-resolving (e.g. large-eddy) simulations. A two-dimensional (2-D) prescribed-flow framework is used in the paper to illustrate the library features. The libcloudph++ and all its mandatory dependencies are free and open-source software. The Boost.units library is used for zero-overhead dimensional analysis of the code at compile time. The particle-based scheme is implemented using the Thrust library that allows one to leverage the power of graphics processing units (GPU), retaining the possibility of compiling the unchanged code for execution on single or multiple standard processors (CPUs). The paper includes a complete description of the programming interface (API) of the library and a performance analysis including comparison of GPU and CPU set-ups.

CHOLLA: A NEW MASSIVELY PARALLEL HYDRODYNAMICS CODE FOR ASTROPHYSICAL SIMULATION

We present Cholla (Computational Hydrodynamics On ParaLLel Architectures), a new three-dimensional hydrodynamics code that harnesses the power of graphics processing units (GPUs) to accelerate astrophysical simulations. Cholla models the Euler equations on a static mesh using state-of-the-art techniques, including the unsplit Corner Transport Upwind (CTU) algorithm, a variety of exact and approximate Riemann solvers, and multiple spatial reconstruction techniques including the piecewise parabolic method (PPM). Using GPUs, Cholla evolves the fluid properties of thousands of cells simultaneously and can update over ten million cells per GPU-second while using an exact Riemann solver and PPM reconstruction. Owing to the massively-parallel architecture of GPUs and the design of the Cholla code, astrophysical simulations with physically interesting grid resolutions (> 256^3) can easily be computed on a single device. We use the Message Passing Interface library to extend calculations onto multiple devices and demonstrate nearly ideal scaling beyond 64 GPUs. A suite of test problems highlights the physical accuracy of our modeling and provides a useful comparison to other codes. We then use Cholla to simulate the interaction of a shock wave with a gas cloud in the interstellar medium, showing that the evolution of the cloud is highly dependent on its density structure. We reconcile the computed mixing time of a turbulent cloud with a realistic density distribution destroyed by a strong shock with the existing analytic theory for spherical cloud destruction by describing the system in terms of its median gas density.

GPUDePiCt: A Parallel Implementation of a Clustering Algorithm for Computing Degenerate Primers on Graphics Processing Units.

In order to make multiple copies of a target sequence in the laboratory, the technique of Polymerase Chain Reaction (PCR) requires the design of "primers", which are short fragments of nucleotides complementary to the flanking regions of the target sequence. If the same primer is to amplify multiple closely related target sequences, then it is necessary to make the primers "degenerate", which would allow it to hybridize to target sequences with a limited amount of variability that may have been caused by mutations. However, the PCR technique can only allow a limited amount of degeneracy, and therefore the design of degenerate primers requires the identification of reasonably well-conserved regions in the input sequences. We take an existing algorithm for designing degenerate primers that is based on clustering and parallelize it in a web-accessible software package GPUDePiCt, using a shared memory model and the computing power of Graphics Processing Units (GPUs). We test our implementation on large sets of aligned sequences from the human genome and show a multi-fold speedup for clustering using our hybrid GPU/CPU implementation over a pure CPU approach for these sequences, which consist of more than 7,500 nucleotides. We also demonstrate that this speedup is consistent over larger numbers and longer lengths of aligned sequences.

How much past to see the future: a computational study in calibrating urban cellular automata

The objective of this computational study was to investigate to which extent the availability and the way of use of historical maps may affect the quality of the calibration process of cellular automata (CA) urban models. The numerical experiments are based on a constrained CA applied to a case study. Since the model depends on a large number of parameters, we optimize the CA using cooperative coevolutionary particle swarms, which is an approach known for its ability to operate effectively in search spaces with a high number of dimensions. To cope with the relevant computational cost related to the high number of CA simulations required by our study, we use a parallelized CA model that takes advantage of the computing power of graphics processing units. The study has shown that the accuracy of simulations can be significantly influenced by both the number and position in time of the historical maps involved in the calibration.

CMFD and GPU acceleration on method of characteristics for hexagonal cores

Coarse Mesh Finite Difference (CMFD) has been widely adopted as an effective way to accelerate the source iteration of transport calculation. However in a core with hexagonal assemblies there are non-hexagonal meshes around the edges of assemblies, causing a problem for CMFD if the CMFD equations are still to be solved via tri-diagonal matrix inversion by simply scanning the whole core meshes in different directions. To solve this problem, we propose an unequal mesh CMFD formulation that combines the non-hexagonal cells on the boundary of neighboring assemblies into non-regular hexagonal cells. We also investigated the alternative hardware acceleration of using graphics processing units (GPU) with graphics card in a personal computer. The tool CUDA is employed, which is a parallel computing platform and programming model invented by the company NVIDIA for harnessing the power of GPU. To investigate and implement these two acceleration methods, a 2-D hexagonal core transport code using the method of characteristics (MOC) is developed. A hexagonal mini-core benchmark problem is established to confirm the accuracy of the MOC code and to assess the effectiveness of CMFD and GPU parallel acceleration. For this benchmark problem, the CMFD acceleration increases the speed 16 times while the GPU acceleration speeds it up 25 times. When used simultaneously, they provide a speed gain of 292 times.

The challenges of genome-wide interaction studies: lessons to learn from the analysis of HDL blood levels.

Genome-wide association studies (GWAS) have revealed 74 single nucleotide polymorphisms (SNPs) associated with high-density lipoprotein cholesterol (HDL) blood levels. This study is, to our knowledge, the first genome-wide interaction study (GWIS) to identify SNP×SNP interactions associated with HDL levels. We performed a GWIS in the Rotterdam Study (RS) cohort I (RS-I) using the GLIDE tool which leverages the massively parallel computing power of Graphics Processing Units (GPUs) to perform linear regression on all genome-wide pairs of SNPs. By performing a meta-analysis together with Rotterdam Study cohorts II and III (RS-II and RS-III), we were able to filter 181 interaction terms with a p-value<1 · 10−8 that replicated in the two independent cohorts. We were not able to replicate any of these interaction term in the AGES, ARIC, CHS, ERF, FHS and NFBC-66 cohorts (N total = 30,011) when adjusting for multiple testing. Our GWIS resulted in the consistent finding of a possible interaction between rs774801 in ARMC8 (ENSG00000114098) and rs12442098 in SPATA8 (ENSG00000185594) being associated with HDL levels. However, p-values do not reach the preset Bonferroni correction of the p-values. Our study suggest that even for highly genetically determined traits such as HDL the sample sizes needed to detect SNP×SNP interactions are large and the 2-step filtering approaches do not yield a solution. Here we present our analysis plan and our reservations concerning GWIS.

GPU-assisted malware

Malware writers constantly seek new methods to increase the infection lifetime of their malicious software. To that end, techniques such as code unpacking and polymorphism have become the norm for hindering automated or manual malware analysis and evading virus scanners. In this paper, we demonstrate how malware can take advantage of the ubiquitous and powerful graphics processing unit (GPU) to increase its robustness against analysis and detection. We present the design and implementation of brute-force unpacking and runtime polymorphism, two code armoring techniques based on the general-purpose computing capabilities of modern graphics processors. By running part of the malicious code on a different processor architecture with ample computational power, these techniques pose significant challenges to existing malware detection and analysis systems, which are tailored to the analysis of CPU code. We also discuss how upcoming GPU features can be used to build even more robust and evasive malware, as well as directions for potential defenses against GPU-assisted malware.

Power Modeling for GPU Architectures Using McPAT

Graphics Processing Units (GPUs) are very popular for both graphics and general-purpose applications. Since GPUs operate many processing units and manage multiple levels of memory hierarchy, they consume a significant amount of power. Although several power models for CPUs are available, the power consumption of GPUs has not been studied much yet. In this article we develop a new power model for GPUs by utilizing McPAT, a CPU power tool. We generate initial power model data from McPAT with a detailed GPU configuration, and then adjust the models by comparing them with empirical data. We use the NVIDIA's Fermi architecture for building the power model, and our model estimates the GPU power consumption with an average error of 7.7% and 12.8% for the microbenchmarks and Merge benchmarks, respectively.

WE-G-18A-04: 3D Dictionary Learning Based Statistical Iterative Reconstruction for Low-Dose Cone Beam CT Imaging

Purpose: To develop a 3D dictionary learning based statistical reconstruction algorithm on graphic processing units (GPU), to improve the quality of low-dose cone beam CT (CBCT) imaging with high efficiency. Methods: A 3D dictionary containing 256 small volumes (atoms) of 3×3×3 voxels was trained from a high quality volume image. During reconstruction, we utilized a Cholesky decomposition based orthogonal matching pursuit algorithm to find a sparse representation on this dictionary basis of each patch in the reconstructed image, in order to regularize the image quality. To accelerate the time-consuming sparse coding in the 3D case, we implemented our algorithm in a parallel fashion by taking advantage of the tremendous computational power of GPU. Evaluations are performed based on a head-neck patient case. FDK reconstruction with full dataset of 364 projections is used as the reference. We compared the proposed 3D dictionary learning based method with a tight frame (TF) based one using a subset data of 121 projections. The image qualities under different resolutions in z-direction, with or without statistical weighting are also studied. Results: Compared to the TF-based CBCT reconstruction, our experiments indicated that 3D dictionary learning based CBCT reconstruction is able to recover finer structures, to remove more streaking artifacts, and is less susceptible to blocky artifacts. It is also observed that statistical reconstruction approach is sensitive to inconsistency between the forward and backward projection operations in parallel computing. Using high a spatial resolution along z direction helps improving the algorithm robustness. Conclusion: 3D dictionary learning based CBCT reconstruction algorithm is able to sense the structural information while suppressing noise, and hence to achieve high quality reconstruction. The GPU realization of the whole algorithm offers a significant efficiency enhancement, making this algorithm more feasible for potential clinical application. A high zresolution is preferred to stabilize statistical iterative reconstruction. This work was supported in part by NIH(1R01CA154747-01), NSFC((No. 61172163), Research Fund for the Doctoral Program of Higher Education of China (No. 20110201110011), China Scholarship Council.

Lagrangian particle-based simulation of fluid-solid coupling on graphics processing units

Lagrangian particle method has been widely used in computer physics and graphics; however, numerically solving the partial differential physical equation on a great number of particles is a computationally complex task. In this paper, a unified particle method on graphics processing units is proposed to simulate fluid-solid interaction with large density ratio interactively. Motivated by microscopic molecular dynamics, we consider the solid object as a particular fluid limited to solid motions; therefore, fluid-solid interaction as well as solid-solid interaction could be solved directly using multiphase weakly compressible smoothed particle hydrodynamics solvers. And then, we present a momentum-conserving particle collision handling scheme to prevent fluid penetrating into solid objects. In the simulation, a measure of particle densities is used to handle density discontinuities at fluid-solid interfaces, and consequently, new formulations for density-weighted inter-particle pressure and viscous forces are derived. Moreover, to realistically simulate various small-scale interaction phenomena such as water droplets flowing on solids' surfaces, a surface tension model that uses density-weighted color gradient and can obtain a stable and accurate surface curvature is employed to capture the interfacial fluid-solid tensions. Because all of the computation is carried out on graphics processing unit and no CPU processing is needed, the proposed algorithm can exploit the massive computational power of graphics processing unit for interactive simulation with a higher particle resolution. The experiment results show that our method can simulate realistic fluid-solid couplings at interactive frame-rates even for up to 126 k particles. Copyright © 2014 John Wiley & Sons, Ltd.

3D dictionary learning based iterative cone beam CT reconstruction

Purpose: This work is to develop a 3D dictionary learning based cone beam CT (CBCT) reconstruction algorithm on graphic processing units (GPU) to improve the quality of sparse-view CBCT reconstruction with high efficiency. Methods : A 3D dictionary containing 256 small volumes (atoms) of 3 × 3 × 3 was trained from a large number of blocks extracted from a high quality volume image. On the basis, we utilized cholesky decomposition based orthogonal matching pursuit algorithm to find the sparse representation of each block. To accelerate the time-consuming sparse coding in the 3D case, we implemented the sparse coding in a parallel fashion by taking advantage of the tremendous computational power of GPU. Conjugate gradient least square algorithm was adopted to minimize the data fidelity term. Evaluations are performed based on a head-neck patient case. FDK reconstruction with full dataset of 364 projections is used as the reference. We compared the proposed 3D dictionary learning based method with tight frame (TF) by performing reconstructions on a subset data of 121 projections. Results: Compared to TF based CBCT reconstruction that shows good overall performance, our experiments indicated that 3D dictionary learning based CBCT reconstruction is able to recover finer structures, remove more streaking artifacts and also induce less blocky artifacts. Conclusion : 3D dictionary learning based CBCT reconstruction algorithm is able to sense the structural information while suppress the noise, and hence to achieve high quality reconstruction under the case of sparse view. The GPU realization of the whole algorithm offers a significant efficiency enhancement, making this algorithm more feasible for potential clinical application. ------------------------------- Cite this article as: Bai T, Yan H, Shi F, Jia X, Lou Y, Xu Q, Jiang S, Mou X. 3D dictionary learning based iterative cone beam CT reconstruction. Int J Cancer Ther Oncol 2014; 2(2):020240. DOI: 10.14319/ijcto.0202.40

Large-Scale Pairwise Alignments on GPU Clusters: Exploring the Implementation Space

Several problems in computational biology require the all-against-all pairwise comparisons of tens of thousands of individual biological sequences. Each such comparison can be performed with the well-known Needleman-Wunsch alignment algorithm. However, with the rapid growth of biological databases, performing all possible comparisons with this algorithm in serial becomes extremely time-consuming. The massive computational power of graphics processing units (GPUs) makes them an appealing choice for accelerating these computations. As such, CPU-GPU clusters can enable all-against-all comparisons on large datasets. In this work, we present four GPU implementations for large-scale pairwise sequence alignment: TiledDScan-mNW, DScan-mNW, RScan-mNW and LazyRScan-mNW. The proposed GPU kernels exhibit different parallelization patterns: we discuss how each parallelization strategy affects the memory accesses and the utilization of the underlying GPU hardware. We evaluate our implementations on a variety of low- and high-end GPUs with different compute capabilities. Our results show that all the proposed solutions outperform the existing open-source implementation from the Rodinia Benchmark Suite, and LazyRScan-mNW is the preferred solution for applications that require performing the trace-back operation only on a subset of the considered sequence pairs (for example, the pairs whose alignment score exceeds a predefined threshold). Finally, we discuss the integration of the proposed GPU kernels into a hybrid MPI-CUDA framework for deployment on CPU-GPU clusters. In particular, our proposed distributed design targets both homogeneous and heterogeneous clusters with nodes that differ amongst themselves in their hardware configuration.

BROCCOLI: Software for fast fMRI analysis on many-core CPUs and GPUs

Analysis of functional magnetic resonance imaging (fMRI) data is becoming ever more computationally demanding as temporal and spatial resolutions improve, and large, publicly available data sets proliferate. Moreover, methodological improvements in the neuroimaging pipeline, such as non-linear spatial normalization, non-parametric permutation tests and Bayesian Markov Chain Monte Carlo approaches, can dramatically increase the computational burden. Despite these challenges, there do not yet exist any fMRI software packages which leverage inexpensive and powerful graphics processing units (GPUs) to perform these analyses. Here, we therefore present BROCCOLI, a free software package written in OpenCL (Open Computing Language) that can be used for parallel analysis of fMRI data on a large variety of hardware configurations. BROCCOLI has, for example, been tested with an Intel CPU, an Nvidia GPU, and an AMD GPU. These tests show that parallel processing of fMRI data can lead to significantly faster analysis pipelines. This speedup can be achieved on relatively standard hardware, but further, dramatic speed improvements require only a modest investment in GPU hardware. BROCCOLI (running on a GPU) can perform non-linear spatial normalization to a 1 mm3 brain template in 4–6 s, and run a second level permutation test with 10,000 permutations in about a minute. These non-parametric tests are generally more robust than their parametric counterparts, and can also enable more sophisticated analyses by estimating complicated null distributions. Additionally, BROCCOLI includes support for Bayesian first-level fMRI analysis using a Gibbs sampler. The new software is freely available under GNU GPL3 and can be downloaded from github (https://github.com/wanderine/BROCCOLI/).

Efficient and Large-Scale Dissipative Particle Dynamics Simulations on GPU

In this paper, we introduce a complete implementation for efficient and large-scale dissipative particle dynamics (DPD) simulation on Graphics Processing Unit (GPU). The implementation is designed and optimized according to the nature of DPD simulation technique and also takes fully advantage of the computational power of GPUs. From studies of benchmarks, we show that the GPU-based implementation can predict the results correctly and provide nearly 60 times speedup over LAMMPS on a single Central Processing Unit (CPU) core. By using a novel divide-and-conquer (D&C) algorithm to reduce the memory requirement in simulation, our implementation has the capability to perform large-scale DPD simulations with some ten millions of particles on a single current GPU. Furthermore, the thermal fluctuation analysis of a superior large-scale lamellar system (11,059,200 particles) is presented as an important example of practical application of our implementation, and a scaling law at large wavelengths that is inaccessible to small simulation system, is observed. As a result, our GPU-based DPD implementation is very promising for studying various interesting phenomena, which often take place on the mesoscopic length and time scales and are not easily addressed by a conventional CPU-based implementation.

High‐speed parallel implementations of the rainbow method based on perfect tables in a heterogeneous system

SummaryThe computing power of graphics processing units (GPU) has increased rapidly, and there has been extensive research on general‐purpose computing on GPU (GPGPU) for cryptographic algorithms such as RSA, Elliptic Curve Cryptosystem (ECC), NTRU, and Advanced Encryption Standard. With the rise of GPGPU, commodity computers have become complex heterogeneous GPU+CPU systems. This new architecture poses new challenges and opportunities in high‐performance computing. In this paper, we present high‐speed parallel implementations of the rainbow method based on perfect tables, which is known as the most efficient time‐memory trade‐off, in the heterogeneous GPU+CPU system. We give a complete analysis of the effect of multiple checkpoints on reducing the cost of false alarms and take advantage of it for load balancing between GPU and CPU. For GTX460, our implementation is about 1.86 and 3.25 times faster than other GPU‐accelerated implementations, RainbowCrack and Cryptohaze, respectively, and for GTX580, 1.53 and 2.40 times faster. Copyright © 2014 John Wiley &amp; Sons, Ltd.

Collisional and collisionless expansion of Yukawa balls

The expansion of Yukawa balls is studied by means of molecular dynamics simulations of collisionless and collisional situations. High computation speed was achieved by using the parallel computing power of graphics processing units. When the radius of the Yukawa ball is large compared to the shielding length, the expansion process starts with the blow-off of the outermost layer. A rarefactive wave subsequently propagates radially inward at the speed of longitudinal phonons. This mechanism is fundamentally different from Coulomb explosions, which employ a self-similar expansion of the entire system. In the collisionless limit, the outer layers carry away most of the available energy. The simulations are compared with analytical estimates. In the collisional case, the expansion process can be described by a nonlinear diffusion equation that is a special case of the porous medium equation.

GPU accelerated segmentation and centerline extraction of tubular structures from medical images

To create a fast and generic method with sufficient quality for extracting tubular structures such as blood vessels and airways from different modalities (CT, MR and US) and organs (brain, lungs and liver) by utilizing the computational power of graphic processing units (GPUs). A cropping algorithm is used to remove unnecessary data from the datasets on the GPU. A model-based tube detection filter combined with a new parallel centerline extraction algorithm and a parallelized region growing segmentation algorithm is used to extract the tubular structures completely on the GPU. Accuracy of the proposed GPU method and centerline algorithm is compared with the ridge traversal and skeletonization/thinning methods using synthetic vascular datasets. The implementation is tested on several datasets from three different modalities: airways from CT, blood vessels from MR, and 3D Doppler Ultrasound. The results show that the method is able to extract airways and vessels in 3-5 s on a modern GPU and is less sensitive to noise than other centerline extraction methods. Tubular structures such as blood vessels and airways can be extracted from various organs imaged by different modalities in a matter of seconds, even for large datasets.

Dynamic load balancing on heterogeneous multi-GPU systems

Actual HPC systems are composed by multicore processors and powerful graphics processing units. Adapting existing code and libraries to these new systems is a fundamental problem due to the important increment on programming difficulties. The heterogeneity, both at architectural and programming levels at the same time, raises the programmability wall. The performance of the code is affected by the large interdependence between the code and the parallel architecture. We have developed a dynamic load balancing library that allows parallel code to be adapted to a wide variety of heterogeneous systems. The overhead introduced by our system is minimal and the cost to the programmer negligible. This system has been successfully applied to solve load imbalance problems appearing in homogeneous and heterogeneous multiGPU platforms. We consider the Dynamic Programming technique as case of study to validate our proposals using different heterogeneous scenarios in multiGPU systems.

Accelerating a hydrological uncertainty ensemble model using graphics processing units (GPUs)

The practical application of hydrological uncertainty models that are designed to generate multiple ensembles can be severely restricted by the available computer processing power and thus, the time taken to generate the results. CPU clusters can help in this regard, but are often costly to use continuously and maintain, causing scientists to look elsewhere for speed improvements. The use of powerful graphics processing units (GPUs) for application acceleration has become a recent trend, owing to their low cost per FLOP, and their highly parallel and throughput-oriented architecture, which makes them ideal for many scientific applications. However, programming these devices efficiently is non-trivial, seemingly making their use impractical for many researchers. In this study, we investigate whether redesigning the CPU code of an adapted Pitman rainfall-runoff uncertainty model is necessary to obtain a satisfactory speedup on GPU devices. A twelvefold speedup over a multithreaded CPU implementation was achieved by using a modern GPU with minimal changes to the model code. This success leads us to believe that redesigning code for the GPU is not always necessary to obtain a worthwhile speedup.