• Journal of KIISE
• /
• v.42 no.12
• /
• pp.1467-1473
• /
• 2015

In this paper, we implemented a parallel random number generation program on GPU's, which are known for high performance computing, using LCG (Linear Congruential Generator). Random numbers are important in all fields requiring the use of randomness, and LCG is one of the most widely used methods for the generation of pseudo-random numbers. We explained the parallel program using the NVIDIA CUDA model and MPI(Message Passing Interface) and showed uniform distribution and performance results. We also used a Monte Carlo algorithm to calculate pi(${\pi}$) comparing the parallel random number generator with cuRAND, which is a CUDA library function, and showed that our program is much more efficient. Finally we compared performance results using multi-GPU's with those of ideal speedups.

• Journal of the Korea Computer Graphics Society
• /
• v.21 no.1
• /
• pp.13-21
• /
• 2015

We present a robust GPU algorithm constructing a sweep volume boundary for a triangular mesh model. Sweeping geometric entities of a triangular mesh object is first approximated to a set of triangles, the envelope of which becomes the outer boundary of the sweep volume. We find the envelope by computing the arrangement of the triangle set and extracting its outmost boundary. To ensure robustness of the algorithm, we adopt random perturbation of sweep vertices and the interval arithmetic using multi-level precisions. The algorithm is implemented to perform most computation on GPU, and as a result it runs two orders of magnitude faster than other algorithms.

• Journal of Broadcast Engineering
• /
• v.19 no.5
• /
• pp.616-626
• /
• 2014

In this paper, we introduce a real-time stereo-to-multiview conversion system using FPGA and GPU. The system is based on two different devices so that it consists of two major blocks. The first block is a disparity estimation block that is implemented on FPGA. In this block, each disparity map of stereoscopic video is estimated by DP(dynamic programming)-based stereo matching. And then the estimated disparity maps are refined by post-processing. The refined disparity map is transferred to the GPU device through USB 3.0 and PCI-express interfaces. Stereoscopic video is also transferred to the GPU device. These data are used to render arbitrary number of virtual views in next block. In the second block, disparity-based view interpolation is performed to generate virtual multi-view video. As a final step, all generated views have to be re-arranged into a single image at full resolution for presenting on the target autostereoscopic 3D display. All these steps of the second block are performed in parallel on the GPU device.

• Journal of the Korea Society of Computer and Information
• /
• v.18 no.8
• /
• pp.1-8
• /
• 2013

This paper proposes an effective parallel implementation of a physical modeling synthesis of guitar on the GPU environment. We used appropriate filter coefficients and adjusted the length of delay line for each open string to generate 44,100 six-polyphonic guitar sounds (E2, A2, D3, G4, B3, E4) by using physical modeling synthesis. In addition, we analyzed the physical modeling synthesis algorithm and observed that we can exploit parallelism inherent in the length of delay line. Thus, we assigned CUDA cores as many as the length of delay line and effectively implemented the physical modeling synthesis using GPU to achieve the highest performance. Experimental results indicated that synthetic guitar sounds using GPU were very similar to the original sounds when we compared their spectra. In addition, GPU achieved 68x and 3x better performance than high-performance TI DSP and CPU, respectively. Furthermore, this paper implemented and evaluated the performance of multi-GPU systems for the physical modeling algorithm.

• Journal of Institute of Control, Robotics and Systems
• /
• v.19 no.12
• /
• pp.1061-1066
• /
• 2013

In the steel industry, the detection of tiny defects including its 3D characteristics on steel surfaces is very important from the point of view of quality control. A multi-spectral photometric stereo method is an attractive scheme because the shape of the defect can be obtained based on the images which are acquired at the same time by using a multi-channel camera. Moreover, the calculation time required for this scheme can be greatly reduced for real-time application with the aid of a GPU (Graphic Processing Unit). Although a more reliable shape reconstruction of defects can be possible when the numbers of available images are increased, it is not an easy task to construct a camera system which has more than 3 channels in the visible light range. In this paper, a new 6-channel camera system, which can distinguish the vertical/horizontal linearly polarized lights of RGB light sources, was developed by adopting two 3-CCD cameras and two polarized lenses based on the fact that the polarized light is preserved on the steel surface. The photometric stereo scheme with 6 images was accelerated by using a GPU, and the performance of the proposed system was validated through experiments.

• ETRI Journal
• /
• v.44 no.1
• /
• pp.62-72
• /
• 2022

In this study, we propose a multi-GPU-based 8KVR stitching system that operates in real time on both local and cloud machine environments. The proposed system first obtains multiple 4 K video inputs, decodes them, and generates a stitched 8KVR video stream in real time. The generated 8KVR video stream can be downloaded and rendered omnidirectionally in player apps on smartphones, tablets, and head-mounted displays. To speed up processing, we adopt group-of-pictures-based distributed decoding/encoding and buffering with the NV12 format, along with multi-GPU-based parallel processing. Furthermore, we develop several algorithms such as equirectangular projection-based color correction, real-time CG overlay, and object motion-based seam estimation and correction, to improve the stitching quality. From experiments in both local and cloud machine environments, we confirm the feasibility of the proposed 8KVR stitching system with stitching speed of up to 83.7 fps for six-channel and 62.7 fps for eight-channel inputs. In addition, in an 8KVR live streaming test on the 5G MEC/cloud, the proposed system achieves stable performances with 8 K@30 fps in both indoor and outdoor environments, even during motion.

• The Journal of Korean Institute of Communications and Information Sciences
• /
• v.38B no.3
• /
• pp.163-171
• /
• 2013

Recently, the requirement of processing High Definition (HD) video or 3D application on low, mobile devices has been expanded and content data has been increased as well. It is becoming a major issue in Cloud computing where a Virtual Desktop Infrastructure (VDI) Service needs efficient data processing ability to provide Quality of Experience (QoE) in Cloud computing. In this paper, we propose three kind of Thick-Thin VDI Service which can share and delegate VDI service based on Thick Client using CPU and GPU. Furthermore, we propose and discuss the VDI Service Optimization Method in mixed CPU and GPU Heterogeneous Environment using CPU Parallel Processing OpenMP and GPU Parallel Processing CUDA.

• The Journal of the Institute of Internet, Broadcasting and Communication
• /
• v.24 no.3
• /
• pp.63-67
• /
• 2024

Recently, artificial intelligence (AI) workloads encompassing various industries such as smart logistics, FinTech, and entertainment are being executed on the cloud. In this paper, we address the scheduling issues of various AI workloads on a multi-tenant cloud system composed of heterogeneous GPU clusters. Traditional scheduling decreases GPU utilization in such environments, degrading system performance significantly. To resolve these issues, we present a new scheduling approach utilizing genetic algorithm-based optimization techniques, implemented within a process-based event simulation framework. Trace driven simulations with diverse AI workload traces collected from Alibaba's MLaaS cluster demonstrate that the proposed scheduling improves GPU utilization compared to conventional scheduling significantly.

• Journal of Korea Multimedia Society
• /
• v.12 no.2
• /
• pp.209-218
• /
• 2009

3D backprojection is a kind of reconstruction algorithm to generate volume data consisting of tomographic images, which provides spatial information of the original 3D data from hundreds of 2D projections. The computational time of backprojection increases in proportion to the size of volume data and the number of projection images since the value of every voxel in volume data is calculated by considering corresponding pixels from hundreds of projections. For the reduction of computational time, fast GPU based 3D backprojection methods have been studied recently and the performance of them has been improved significantly. This paper presents two multiple GPU based methods to maximize the parallelism of GPU and compares the efficiencies of two methods by considering both the number of projections and the size of volume data. The first method is to generate partial volume data independently for all projections after allocating a half size of volume data on each GPU. The second method is to acquire the entire volume data by merging the incomplete volume data of each GPU on CPU. The in-complete volume data is generated using the half size of projections after allocating the full size of volume data on each GPU. In experimental results, the first method performed better than the second method when the entire volume data can be allocated on GPU. Otherwise, the second method was efficient than the first one.

• The Bulletin of The Korean Astronomical Society
• /
• v.40 no.2
• /
• pp.44.1-44.1
• /
• 2015

We set up a project to make spectrometers for single dish observations of the Korean VLBI Network (KVN), a new future multi-beam receiver of the ASTE (Atacama Submillimeter Telescope Experiment), and the total power (TP) antennas of the Atacama Large Millimeter/submillimeter Array (ALMA). Traditionally, spectrometers based on ASIC (Application-Specific Integrated circuit) and FPGA (Field-Programmable Gate Array) have been used in radio astronomy. It is, however, that a Graphics Processing Unit (GPU) technology is now viable for spectrometers due to the rapid improvement of its performance. A high-resolution spectrometer should have the following functions: poly-phase filter, data-bit conversion, fast Fourier transform, and complex multiplication. We wrote a program based on CUDA (Compute Unified Device Architecture) for a GPU spectrometer. We measured its performance using two GPU cards, Titan X and K40m, from NVIDIA. A non-optimized GPU code can process a data stream of around 2 GHz bandwidth, which is enough for the KVN spectrometer and promising for the ASTE and ALMA TP spectrometers.