• Korean Journal of Cognitive Science
• /
• v.33 no.3
• /
• pp.155-168
• /
• 2022

In this paper, we introduce the SNUDM-G, a diffusion model analysis tool with improved computational speed. Although the diffusion model has been applied to explain various cognitive tasks, its use was limited due to computational difficulties. In particular, SNUDM(Koh et al., 2020), one of the diffusion model analysis tools, has a disadvantage in terms of processing speed because it sequentially generates 20,000 data when approximating the diffusion process. To overcome this limitation, we propose to use graphic processing units(GPU) in the process of approximating the diffusion process with a random walk process. Since 20,000 data can be generated in parallel using the graphic processing units, the estimation speed can be increased compared to generating data through sequential processing. As a result of analyzing the data of Experiment 1 by Ratcliff et al. (2004) and recovering the parameters with SNUDM-G using GPU and SNUDM using CPU, SNUDM-G estimated slightly higher values for certain parameters than SNUDM. However, in term of computational speed, SNUDM-G estimated the parameters much faster than SNUDM. This result shows that a more efficient diffusion model analysis for various cognitive tasks is possible using this tool and further suggests that the processing speed of various cognitive models can be improved by using graphic processing units in the future.

• Journal of Internet Computing and Services
• /
• v.12 no.3
• /
• pp.131-137
• /
• 2011

GPUs were originally designed for graphic processing, and GPGPUs are general-purpose GPUs for numerical computation with high performance and low electric power. In this paper, we implemented the parallel LU factorization program for GPGPUs. In CUDA, which is computational environment for Nvidia GPGPUs, domains are divided into blocks, and multi-threads compute each sub-blocks Simultaneously. In LU factorization program, computation order should be artificially decided due to the data dependence. To resolve the data dependancy, we suggested a parallel LU program for GPGPUs, and also explained parallel reduction algorithm for partial pivoting of LU factorization. We finally present performance analysis to show efficiency of the parallel LU factorization program based on multi-threads on GPGPUs.

• Journal of the Institute of Convergence Signal Processing
• /
• v.16 no.1
• /
• pp.1-8
• /
• 2015

In many image processing tasks, connected component labeling (CCL) is performed to extract regions of interest. CCL was usually done in a sequential fashion when image resolution was relatively low and there are small number of input channels. As image resolution gets higher up to HD or Full HD and as the number of input channels increases, sequential CCL is too time-consuming to be used in real time applications. To cope with this situation, parallel CCL framework was introduced where multiple cores are utilized simultaneously. Several parallel CCL methods have been proposed in the literature. Among them are NSZ label equivalence (NSZ-LE) method[1], modified 8 directional label selection (M8DLS) method[2], and HYBRID1 method[3]. Soh [3] showed that HYBRID1 outperforms NSZ-LE and M8DLS, and argued that HYBRID1 is by far the best. In this paper we propose an improved hybrid parallel CCL algorithm termed as HYBRID2 that hybridizes M8DLS with label backtracking (LB) and show that it runs around 20% faster than HYBRID1 for various kinds of images.

• 한국전산유체공학회:학술대회논문집
• /
• 2009.11a
• /
• pp.1-1
• /
• 2009

In this presentation, I talk about various fluid simulation methods that have been developed for computer graphics special effects since 1996. They are all based on CFD but sacrifice physical reality for visual plausability and time. But as the speed of computer increases rapidly and the capability of GPU (graphics processing unit) improves, methods for more physical realism have been tried. In this talk, I will focus on four aspects of fluid simulation methods for computer graphics: (1) particle level-set methods, (2) particle-based simulation, (3) methods for exact satisfaction of incompressibility constraint, and (4) GPU-based simulation. (1) Particle level-set methods evolve the surface of fluid by means of the zero-level set and a band of massless marker particles on both sides of it. The evolution of the zero-level set captures the surface in an approximate manner and the evolution of marker particles captures the fine details of the surface, and the zero-level set is modified based on the particle positions in each step of evolution. (2) Recently the particle-based Lagrangian approach to fluid simulation gains some popularity, because it automatically respects mass conservation and the difficulty of tracking the surface geometry has been somewhat addressed. (3) Until recently fluid simulation algorithm was dominated by approximate fractional step methods. They split the Navier-Stoke equation into two, so that the first one solves the equation without considering the incompressibility constraint and the second finds the pressure which satisfies the constraint. In this approach, the first step introduces error inevitably, producing numerical diffusion in solution. But recently exact fractional step methods without error have been developed by fluid mechanics scholars), and another method was introduced which satisfies the incompressibility constraint by formulating fluid in terms of vorticity field rather than velocity field (by computer graphics scholars). (4) Finally, I want to mention GPU implementation of fluid simulation, which takes advantage of the fact that discrete fluid equations can be solved in parallel.

• Journal of IKEEE
• /
• v.23 no.4
• /
• pp.1314-1320
• /
• 2019

In this paper, we propose a new method for optimizing bit operations and applying them to DNN(Deep Neural Network) in software environment. As a method for this, we propose a packing function for bitwise optimization and a masking matrix multiplication operation for application to DNN. The packing function converts 32-bit real value to 2-bit quantization value through threshold comparison operation. When this sequence is over, four 32-bit real values are changed to one 8-bit value. The masking matrix multiplication operation consists of a special operation for multiplying the packed weight value with the normal input value. And each operation was then processed in parallel using a GPU accelerator. As a result of this experiment, memory saved about 16 times than 32-bit DNN Model. Nevertheless, the accuracy was within 1%, similar to the 32-bit model.

• Smart Media Journal
• /
• v.3 no.1
• /
• pp.33-38
• /
• 2014

Even though the performance of microprocessor is improved continuously, the performance improvement of computing system becomes hard to increase, in order to some drawbacks including increased power consumption. To solve the problem, general-purpose computing on graphics processing units(GPGPUs), which execute general-purpose applications by using specialized parallel-processing device representing graphics processing units(GPUs), have been focused. However, the characteristics of applications related with graphics is substantially different from the characteristics of general-purpose applications. Therefore, GPUs cannot exploit the outstanding computational resources sufficiently due to various constraints, when they execute general-purpose applications. When designing GPUs for GPGPU, memory system is important to effectively exploit the GPUs since typically general-purpose applications requires more memory accesses than graphics applications. Especially, external memory access requiring long latency impose a big overhead on the performance of GPUs. Therefore, the GPU performance must be improved if hierarchical memory architecture which can reduce the number of external memory access is applied. For this reason, we will investigate the analysis of GPU performance according to hierarchical cache architectures in executing various benchmarks.

• Journal of Korea Society of Digital Industry and Information Management
• /
• v.9 no.4
• /
• pp.59-67
• /
• 2013

Compared to the conventional Hardware-oriented base stations, Software Defined Radio (SDR)-based base station provides various advantages especially in flexibility and expandability. It enables the multimode capability required in 4th-generation (4G) environment which aims at a convergence network of various kinds of communication standards. However, since a single base station processes all data required in various multiple waveforms, the SDR base station faces a problem of data processing speed. In this paper, we propose a new concept of SDR base station system which adopts a parallel processing technology of clustering environment. We implemented a WiMAX system with SDR concept which adopts the Message Passing Interface (MPI) technology which enables the speed-up operations. In order to maximize the efficiency of parallel processing in signal processing, we analyze how the algorithm at each of modules is related to data to be processed. Through the implemented system, we show a drastic improvement in operation time due to parallel processing using the proposed MPI technology. In addition, we demonstrate a feasibility of SDR system for 4G or even beyond-4G as well.

• Journal of KIISE:Computing Practices and Letters
• /
• v.14 no.1
• /
• pp.1-12
• /
• 2008

In this paper, we present a real-time algorithm for recognizing the vehicle color from the indoor and outdoor vehicle images based on GPU (Graphics Processing Unit) acceleration. In the preprocessing step, we construct feature victors from the sample vehicle images with different colors. Then, we combine the feature vectors for each color and store them as a reference texture that would be used in the GPU. Given an input vehicle image, the CPU constructs its feature Hector, and then the GPU compares it with the sample feature vectors in the reference texture. The similarities between the input feature vector and the sample feature vectors for each color are measured, and then the result is transferred to the CPU to recognize the vehicle color. The output colors are categorized into seven colors that include three achromatic colors: black, silver, and white and four chromatic colors: red, yellow, blue, and green. We construct feature vectors by using the histograms which consist of hue-saturation pairs and hue-intensity pairs. The weight factor is given to the saturation values. Our algorithm shows 94.67% of successful color recognition rate, by using a large number of sample images captured in various environments, by generating feature vectors that distinguish different colors, and by utilizing an appropriate likelihood function. We also accelerate the speed of color recognition by utilizing the parallel computation functionality in the GPU. In the experiments, we constructed a reference texture from 7,168 sample images, where 1,024 images were used for each color. The average time for generating a feature vector is 0.509ms for the $150{\times}113$ resolution image. After the feature vector is constructed, the execution time for GPU-based color recognition is 2.316ms in average, and this is 5.47 times faster than the case when the algorithm is executed in the CPU. Our experiments were limited to the vehicle images only, but our algorithm can be extended to the input images of the general objects.

• Journal of the Korea Computer Graphics Society
• /
• v.15 no.1
• /
• pp.17-23
• /
• 2009

We present a simple new idea to improve the quality of exemplar based texture synthesis using multiple rotated input exemplars. Our algorithm successfully obtain rotational synthesis feature variations and manages to reduce the artifacts in the results, especially patch seams due to the structure of the exemplars provided which have been inappropriate for previous neighborhood matching synthesis algorithms. Our algorithm is parallel in nature, thus it is possible to implement our algorithm using GPU or multi-core CPU to accelerate synthesis process.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.10 no.7
• /
• pp.3231-3244
• /
• 2016

As processor design has been transiting from homogeneous multicore processor to heterogeneous multicore processor, traditional Amdahl's law cannot meet the new challenges for asymmetric multicore system. In order to further investigate the impact factors related to the Overhead of Data Preparation (ODP) for Asymmetric multicore systems, we evaluate an asymmetric multicore system built with CPU-GPU by measuring the overheads of memory transfer, computing kernel, cache missing and synchronization. This paper demonstrates that decreasing the overhead of data preparation is a promising approach to improve the whole performance of heterogeneous system.