• KIISE Transactions on Computing Practices
• /
• v.22 no.1
• /
• pp.20-25
• /
• 2016

With the advent of the era of digital big data, the Hadoop platform has become widely used in various fields. However, the Hadoop MapReduce framework suffers from problems related to the increase of the name node's main memory and map tasks for the processing of large number of small files. In addition, a method for running C++-based tasks in the MapReduce framework is required in order to conjugate GPUs supporting hardware-based data parallelism in the MapReduce framework. Therefore, in this paper, we present a face detection system that generates a sequence file for images to process big data for images in the Hadoop platform. The system also deals with tasks for GPU-based face detection in the MapReduce framework using Hadoop Pipes. We demonstrate a performance increase of around 6.8-fold as compared to a single CPU process.

• Journal of the Korea Computer Graphics Society
• /
• v.23 no.2
• /
• pp.1-10
• /
• 2017

A variety of filters are applied to improve the quality of noise and low resolution medical images. This is necessary to reduce the radiation dose of the patient and to improve the utilization of the conventional spherical imaging equipment. In the conventional method, it is common to perform filtering using the CPU of the PC. However, it is difficult to produce results in real time by applying various calculations and filters to high-resolution human images using only the CPU performance of a PC used in a hospital. In this paper, we analyze the structure and performance of Intel integrated GPU in CPU and propose a method to perform image filtering using OpenCL parallel processing function. By applying complex filters with high computational complexity to medical images, high quality images can be generated in real time.

• KIPS Transactions on Computer and Communication Systems
• /
• v.8 no.5
• /
• pp.111-118
• /
• 2019

Recent graphic processing units (GPUs) provide high throughput by using powerful hardware resources. However, massive memory accesses cause GPU performance degradation due to cache inefficiency. Therefore, the performance of GPU can be improved by reducing thread parallelism when cache suffers memory contention. In this paper, we propose a dynamic warp scheduler which controls thread parallelism according to degree of cache contention. Usually, the greedy then oldest (GTO) policy for issuing warp shows lower parallelism than loose round robin (LRR) policy. Therefore, the proposed warp scheduler employs the LRR warp scheduling policy when Miss Status Holding Register(MSHR) utilization is low. On the other hand, the GTO policy is employed in order to reduce thread parallelism when MSHRs utilization is high. Our proposed technique shows better performance compared with LRR and GTO policy since it selects efficient scheduling policy dynamically. According to our experimental results, our proposed technique provides IPC improvement by 12.8% and 3.5% over LRR and GTO on average, respectively.

• Journal of KIISE:Computer Systems and Theory
• /
• v.32 no.11_12
• /
• pp.692-704
• /
• 2005

With the improvement of computer hardware, GPUs(Graphics Processor Units) have tremendous memory bandwidth and computation power. This leads GPUs to use in general purpose computation. Especially, GPU implementation of compute-intensive physics based simulations is actively studied. In the solution of differential equations which are base of physics simulations, tridiagonal matrix systems occur repeatedly by finite-difference approximation. From the point of view of physics based simulations, fast solution of tridiagonal matrix system is important research field. We propose a fast GPU implementation for the solution of tridiagonal matrix systems. In this paper, we implement the cyclic reduction(also known as odd-even reduction) algorithm which is a popular choice for vector processors. We obtained a considerable performance improvement for solving tridiagonal matrix systems over Thomas method and conjugate gradient method. Thomas method is well known as a method for solving tridiagonal matrix systems on CPU and conjugate gradient method has shown good results on GPU. We experimented our proposed method by applying it to heat conduction, advection-diffusion, and shallow water simulations. The results of these simulations have shown a remarkable performance of over 35 frame-per-second on the 1024x1024 grid.

• Journal of Broadcast Engineering
• /
• v.19 no.2
• /
• pp.140-147
• /
• 2014

In this paper, we propose a method for real-time depth image refinement. In order to improve the quality of the depth map acquired from Kinect camera, we employ constant memory and texture memory which are suitable for a 2D image processing in the graphics processing unit (GPU). In addition, we applied the joint bilateral filter (JBF) in parallel to accelerate the overall execution. To enhance the quality of the depth image, we applied the JBF hierarchically using the compute unified device architecture (CUDA). Finally, we obtain the refined depth image. Experimental results showed that the proposed real-time depth image refinement algorithm improved the subjective quality of the depth image and the computational time was 260 frames per second.

• The Journal of the Institute of Internet, Broadcasting and Communication
• /
• v.19 no.3
• /
• pp.165-171
• /
• 2019

The study of the computer architecture has shown that major improvements in price-to-energy performance stems from domain-specific hardware development. This paper analyzes the tensor processing unit (TPU) ASIC which can accelerate the reasoning of the artificial neural network (NN). The core device of the TPU is a MAC matrix multiplier capable of high-speed operation and software-managed on-chip memory. The execution model of the TPU can meet the reaction time requirements of the artificial neural network better than the existing CPU and the GPU execution models, with the small area and the low power consumption even though it has many MAC and large memory. Utilizing the TPU for the tensor flow benchmark framework, it can achieve higher performance and better power efficiency than the CPU or CPU. In this paper, we analyze TPU, simulate the Python modeled OpenTPU, and synthesize the matrix multiplication unit, which is the key hardware.

• Transactions of the Korean Society of Mechanical Engineers B
• /
• v.40 no.9
• /
• pp.597-604
• /
• 2016

A parallel algorithm of bi-conjugate gradient method was developed based on CUDA for parallel computation of the incompressible Navier-Stokes equations. The governing equations were discretized using splitting P2P1 finite element method. Asymmetric stenotic flow problem was solved to validate the proposed algorithm, and then the parallel performance of the GPU was examined by measuring the elapsed times. Further, the GPU performance for sparse matrix-vector multiplication was also investigated with a matrix of fluid-structure interaction problem. A kernel was generated to simultaneously compute the inner product of each row of sparse matrix and a vector. In addition, the kernel was optimized to improve the performance by using both parallel reduction and memory coalescing. In the kernel construction, the effect of warp on the parallel performance of the present CUDA was also examined. The present GPU computation was more than 7 times faster than the single CPU by double precision.

• Journal of KIISE
• /
• v.41 no.10
• /
• pp.792-798
• /
• 2014

In spatial databases, R-tree is one of the most widely used indexing structures and many variants have been proposed for its performance improvement. Among these variants, Hilbert R-tree is a representative method using Hilbert curve to process large amounts of data without high cost split techniques to construct the R-tree. This Hilbert R-tree, however, is hardly applicable to large-scale applications in practice mainly due to high pre-processing costs and slow bulk-load time. To overcome the limitations of Hilbert R-tree, we propose a novel approach for parallelizing Hilbert mapping and thus accelerating bulk-loading of Hilbert R-tree on GPU memory. Hilbert R-tree based on GPU improves bulk-loading performance by applying the inversed-cell method and exploiting parallelism for packing the R-tree structure. Our experimental results show that the proposed scheme is up to 45 times faster compared to the traditional CPU-based bulk-loading schemes.

• Journal of Broadcast Engineering
• /
• v.13 no.6
• /
• pp.859-874
• /
• 2008

In this paper, we develop a fast view synthesis method that generates multiple intermediate views in real-time for the 3D display system when the camera geometry and depth map of reference views are given in advance. The proposed method achieves faster view synthesis than previous approaches in GPU by processing in parallel the entire computations required for the view synthesis. Specifically, we use $CUDA^{TM}$ (by NVIDIA) to control GPU device. For increasing the processing speed, we adapted all the processes for the view synthesis to single instruction multiple data (SIMD) structure that is a main feature of CUDA, maximized the use of the high-speed memories on GPU device, and optimized the implementation. As a result, we could synthesize 9 intermediate view images with the size of 720 by 480 pixels within 0.128 second.

• Journal of the Korea Society of Computer and Information
• /
• v.27 no.3
• /
• pp.63-69
• /
• 2022

In this paper, we propose a new GPU-based framework for quickly calculating adaptive and continuous SDF(Signed distance fields), and examine cases related to rendering/collision processing using them. The quadtree constructed from the triangle mesh is transferred to the GPU memory, and the Euclidean distance to the triangle is processed in parallel for each thread by using it to find the shortest continuous distance without discontinuity in the adaptive grid space. In this process, it is shown through experiments that the cut-off view of the adaptive distance field, the distance value inquiry at a specific location, real-time raytracing, and collision handling can be performed quickly and efficiently. Using the proposed method, the adaptive sign distance field can be calculated quickly in about 1 second even on a high polygon mesh, so it is a method that can be fully utilized not only for rigid bodies but also for deformable bodies. It shows the stability of the algorithm through various experimental results whether it can accurately sample and represent distance values in various models.