• KNOM Review
• /
• v.23 no.1
• /
• pp.43-50
• /
• 2020

As General Purpose Graphics Processing Unit (GPGPU) recently plays an essential role in high-performance computing, several cloud service providers offer GPU service. Most cluster orchestration platforms in a cloud environment using containers allocate the integer number of GPU to jobs and do not allow a node shared with other jobs. In this case, resource utilization of a GPU node might be low if a job does not intensively require either many cores or large size of memory in GPU. GPU virtualization brings opportunities to realize kernel concurrency and share resources. However, performance may vary depending on characteristics of applications running concurrently and interference among them due to resource contention on a node. This paper proposes GPU container co-execution framework with multiple server creation and execution based on Kubernetes, container orchestration platform for measuring interference which may be occurred by sharing GPU resources. Performance changes according to scheduling policies were investigated by executing several jobs on GPU. The result shows that optimal scheduling is not possible only considering GPU memory and computing resource usage. Interference caused by co-execution among applications is measured using the framework.

• Proceedings of the Korea Information Processing Society Conference
• /
• 2010.11a
• /
• pp.1784-1787
• /
• 2010

General Purpose Graphic Processing Unit (GPGPU) computing is a technique that utilizes the high-performance many-core processors of high-end graphic cards for general-purpose computations such as 3D graphics, video/image processing, computer vision, scientific computing, HPC and many more. GPGPUs offer a vast amount of raw computing power, but programming is extremely challenging because of hardware idiosyncrasies. The open computing language (OpenCL) has been proposed as a vendor-independent GPGPU programming interface. OpenCL is very close to the hardware and thus does little to increase GPGPU programmability. In this paper we present how a set of digital camera image signal processing (ISP) filters can be realized efficiently on GPGPUs using OpenCL. Although we found ISP filters to be memory-bound computations, our GPGPU implementations achieve speedups of up to a factor of 64.8 over their sequential counterparts. On GPGPUs, our proposed optimizations achieved speedups between 145% and 275% over their baseline GPGPU implementations. Our experiments have been conducted on a Geforce GTX 275; because of OpenCL we expect our optimizations to be applicable to other architectures as well.

• Journal of the Korea Institute of Military Science and Technology
• /
• v.14 no.4
• /
• pp.663-668
• /
• 2011

Although multi-frame super resolution algorithm has many merits but it demands too much calculation time. Researches have shown that image processing time can be reduced using a CUDA(Compute unified device architecture) which is one of GPGPU(General purpose computing on graphics processing unit) models. In this paper, we show that the processing time of multi-frame super resolution algorithm can be reduced by employing the CUDA. It was applied not to the whole parts but to the largest time consuming parts of the program. The simulation result shows that using a CUDA can reduce an operation time dramatically. Therefore it can be possible that multi-frame super resolution algorithm is implemented in real time by using libraries of image processing algorithms which are made by a CUDA.

• Proceedings of the Korea Water Resources Association Conference
• /
• 2021.06a
• /
• pp.287-287
• /
• 2021

본 연구에서는 CUDA(Compute Unified Device Architecture) 포트란을 이용하여 확산파 강우 유출모형을 개발하였다. CUDA 포트란은 그래픽 처리 장치(Graphic Processing Unit: GPU)에서 수행하는 병렬 연산 알고리즘을 포트란 언어를 사용하여 작성할 수 있도록 하는 GPU상의 범용계산(General-Purpose Computing on Graphics Processing Units: GPGPU) 기술이다. GPU는 그래픽 처리 작업에 특화된 다수의 산술 논리 장치(Arithmetic Logic Unit: ALU)로 구성되어 있어서 중앙 처리 장치(Central Processing Unit: CPU)보다 한 번에 더 많은 연산 수행이 가능하다. 이에 따라, CUDA 포트란기반 확산파모형은 분포형 강우유출모형의 수치모의 연산시간을 단축시킬 수 있다. 분포형모형의 지배방정식은 확산파모형과 Green-Ampt모형으로 구성되었고, 확산파모형은 유한체적법을 이용하여 이산화 하였다. CUDA 포트란기반 확산파모형의 정확성은 기존 연구된 수리실험 결과 및 CPU기반 강우유출모형과 비교하였으며, 연산소요시간에 대한 효율성은 CPU기반 확산파모형과 비교하였다. 그 결과 CUDA 포트란기반 확산파모형의 결과는 수리실험 결과 및 CPU기반 강우유출모형의 결과와 유사한 결과를 나타냈다. 또한, 연산소요시간은 CPU 기반 확산파모형의 연산소요시간보다 단축되었으며, 본 연구에 사용된 장비를 기준으로 최대 100배 정도 단축되었다.

• Journal of IKEEE
• /
• v.19 no.3
• /
• pp.455-459
• /
• 2015

This paper proposes a dispatch unit of GP-GPU with SIMT architecture to support the acceleration of general-purpose operation as well as graphics processing. If all the information of an operand used instructions issued from the warp scheduler is decoded, an unnecessary operand load occurs, resulting in register loads. To resolve this problem, this paper proposes a method that can reduce the operand load and the load on the resister by decoding only the information of the operand using a pre-decoding method. The operand information from the dispatch unit is passed to the operand selection unit with preventing register bank collisions. Thus the overall performance are improved. In the simulation test, the total clock cycles required by processing 10,000 arbitrary instructions issued from the wrap scheduler using ModelSim SE 10.0b are measured. It shows that the application of the dispatch unit equipped with the pre-decoding function proposed in this paper can make an improvement of about 12% in processing performance compared to the conventional method.

• The Journal of the Convergence on Culture Technology
• /
• v.3 no.4
• /
• pp.165-169
• /
• 2017

CNN(Convolutional Nerual Network) is one of the algorithms that show superior performance in image recognition and classification among machine learning algorithms. CNN is simple, but it has a large amount of computation and it takes a lot of time. Consequently, in this paper we performed an parallel processing unit for the convolution layer, pooling layer and the fully connected layer, which consumes a lot of handling time in the process of CNN, through the SIMT(Single Instruction Multiple Thread)'s structure of GPGPU(General-Purpose computing on Graphics Processing Units).And we also expect to improve performance by reducing the number of memory accesses and directly using the output of convolution layer not storing it in pooling layer. In this paper, we use MNIST dataset to verify this experiment and confirm that the proposed CNN structure is 12.38% better than existing structure.

• Proceedings of the Korea Water Resources Association Conference
• /
• 2018.05a
• /
• pp.271-271
• /
• 2018

그래픽 처리 장치(GPU: Graphic Processing Units)는 그래픽 처리에 특화된 수많은 산술논리연산자 (ALU: Arithmetic Logic Unit)와 이에 관련된 인스트럭션Instruction)으로 인해 중앙 처리 장치(CPU: Central Processing Units) 보다 훨씬 빠른 계산 처리를 수행할 수 있다. 최근에는 FORTRAN에 의해 구현된 많은 수치모형들이 현실적인 모델링 방법의 발달로 인해 더 많은 계산량과 계산시간을 필요로 한다. 이 연구에서는 GPU 상의 범용 계산GPGPU : General-Purpose computing on Graphics Processing Units) 기반 운동파 강우유출모형(Kinematic Wave Rainfall-Runoff Model)이 CUDA(Compute Unified Device Architecture) FORTRAN을 사용하여 구현되었다. CUDA FORTRAN 운동파 강우유출모형의 계산 결과는 검증된 CPU 기반 운동파 강우유출모형의 계산 결과와 비교하여 검증되었으며, 잘 일치함을 보여 주었다. CUDA FORTRAN 운동파 강우유출모형은 CPU 기반 모형에 비해 약 20 배 더 빠른 계산 시간을 보였다. 또한 계산 영역이 커짐에 따라 CPU 버전에 비해 CUDA FORTRAN 버전의 계산 효율이 향상되었다.

• The Transactions of The Korean Institute of Electrical Engineers
• /
• v.65 no.4
• /
• pp.666-672
• /
• 2016

Recently the fingerprint is widely used as one of biometrics to improve the security of financial mobile applications, because of its user convenience and high recognition rate. However, in order to apply fingerprint algorithms to finance and security applications, the recognition rate and processing speed of the fingerprint algorithms have to be improved further. In this paper, we propose the parallel fingerprint enhancement algorithm on general-purpose computing on graphics processing unit (GPGPU) using OpenCL. We discuss the analysis of the parallelism in the fingerprint algorithm as well as the exploration of optimization parameters of the parallel fingerprint algorithm to improve the performance. The experimental results showed that the execution of parallel fingerprint enhancement algorithm on GPGPUs was accelerated from 29.4 upto 69.2 times compared with the execution of the original one on the host CPUs.

• Proceedings of the Korean Society of Computer Information Conference
• /
• 2021.07a
• /
• pp.541-542
• /
• 2021

클라우드 컴퓨팅 환경에서는 고성능 컴퓨팅을 지원하기 위해 사용자에게 GPU(Graphic Processing Unit)가 할당된 가상머신을 제공하여 사용자가 고성능 응용을 실행할 수 있도록 지원한다. 일반적인 컴퓨팅 환경에서 한 명의 사용자가 GPU를 독점해서 사용하기 때문에 자원 경쟁으로 인한 문제가 상대적으로 적게 발생하지만 독립적인 여러 사용자가 컴퓨팅 자원을 공유하는 클라우드 환경에서는 자원 경쟁으로 인해 서로 성능 영향을 미치는 문제를 발생시킨다. 본 논문에서는 여러 개의 가상머신이 단일 GPU를 공유하는 RPC(Remote Procedure Call) 기반 GPU 가상화 환경에서 다수의 가상머신이 GPGPU(General Purpose computing on Graphics Processing Units) 작업을 수행할 때 GPU 메모리 입력 경쟁으로 인해 발생하는 커널 함수의 실행 지연 문제를 분석한다.

• Smart Media Journal
• /
• v.4 no.4
• /
• pp.56-63
• /
• 2015

Modern GPU can execute mass parallel computation by exploiting many GPU core. GPGPU architecture, which is one of approaches exploiting outstanding computational resources on GPU, executes general-purpose applications as well as graphics applications, effectively. In this paper, we investigate the impact of memory-efficiency and performance according to number of CTAs(Cooperative Thread Array) on a SM(Streaming Multiprocessors), since the analysis of relation between number of CTA on a SM and them provides inspiration for researchers who study the GPU to improve the performance. Our simulation results show that almost benchmarks increasing the number of CTAs on a SM improve the performance. On the other hand, some benchmarks cannot provide performance improvement. This is because the number of CTAs generated from same kernel is a little or the number of CTAs executed simultaneously is not enough. To precisely classify the analysis of performance according to number of CTA on a SM, we also analyze the relations between performance and memory stall, dram stall due to the interconnect congestion, pipeline stall at the memory stage. We expect that our analysis results help the study to improve the parallelism and memory-efficiency on GPGPU architecture.