• 대한임베디드공학회논문지
• /
• 제11권4호
• /
• pp.209-218
• /
• 2016

This paper proposes a new lifetime aware buffer mapping method of a synchronous dataflow (SDF) graph on a hybrid memory system with DRAM and PRAM. Since the number of write operations on PRAM is limited, the number of written samples on PRAM is minimized to maximize the lifetime of PRAM. We improve the utilization of DRAM by mapping more buffers on DRAM through buffer sharing. The problem is formulated formally and solved by an optimal approach of an answer set programming. In experiment, the buffer mapping method with buffer sharing improves the PRAM lifetime by 63%.

• JSTS:Journal of Semiconductor Technology and Science
• /
• 제17권2호
• /
• pp.199-203
• /
• 2017

Mutually-actuated-nano-electromechanical (MA-NEM) memory switches are proposed for scalability improvement. While conventional NEM memory switches have fixed electrode lines, the proposed MA-NEM memory switches have mutually-actuated cantilever-like electrode lines. Thus, MA-NEM memory switches show smaller deformations of beams in switching. This unique feature of MA-NEM memory switches allows aggressive reduction of the beam length while maintaining nonvolatile property. Also, the scalability of MA-NEM memory switches is confirmed by using finite-element (FE) simulations. MA-NEM memory switches can be promising solutions for reconfigurable logic (RL) circuits.

• 한국정보통신학회논문지
• /
• 제8권2호
• /
• pp.430-439
• /
• 2004

OFDM 방식의 DVB-T 수신기에서 다수 반송파의 변ㆍ복조를 수행하는 8192점/2048점 FFT/IFFT 프로세서 (CFFT8k2k)를 설계하였다. 8192점 FFT와 같이 변환 크기가 큰 경우에는 매우 큰 용량의 메모리가 필요하므로, 메모리 효율적인 설계가 중요하다. 본 논문에서는 R4SDC (Radix-4 Single-path Delay Commutator)와 R4SDF (Radix-4 Single-path Delay Feedback)를 혼합한 Hybrid 구조를 적용함으로써 R4SDC 단일 구조에 비해 약 20%의 메모리를 줄였으며, 2단계 수렴 블록 부동점 스케일링 기법을 적용함으로써 기존의 CBFP 방식에비해 약 24%의 메모리를 감소시켰다. 이와 같은 메모리 효율적인 설계를 통해, 기존 방식의 약 57%의 메모리만으로 구현되었으며, 칩 면적과 전력소모가 크게 감소되었다. CFFT8k2k 코어는 Verilog-HDL로 설계되었으며, 102,000여 개의 게이트, 292k 비트의 RAM, 그리고 39k 비트의 ROM으로 구현되었다. $0.25-{\um}m$ CMOS라이브러리로 합성된 게이트 레벨 netlst와 SDF를 이용한 타이밍 시뮬레이션 결과, 2.5-V 전원전압에서 50-MHz로 안전하게 동작함을 확인하였으며, 8192점 FFT/IFFT 연산에 164-${\mu}\textrm{s}$가 소요되어 DVB-T 사양을 만족하는 것으로 평가되었다. 설계된 CFFT8k2k 코어는 FPGA로 구현하여 정상 동작함을 확인하였으며, 8192점 FFT의 평균 SQNR은 약 60-㏈로 분석되었다.

• 한국산업융합학회 논문집
• /
• 제24권5호
• /
• pp.531-536
• /
• 2021

As artificial intelligence technology advances, it is being applied to various application fields. Artificial intelligence is performing well in the field of image recognition and classification. Chip design specialized in this field is also actively being studied. Artificial intelligence-specific chips are designed to provide optimal performance for the applications. At the design task, memory component optimization is becoming an important issue. In this study, the optimal algorithm for the memory size exploration is presented, and the optimal memory size is becoming as a important factor in providing a proper design that meets the requirements of performance, cost, and power consumption.

• 한국통신학회:학술대회논문집
• /
• 한국통신학회 1987년도 춘계학술발표회 논문집
• /
• pp.31-34
• /
• 1987

A new efficientcoding system which can transmit video conferenceof viedeophone signals at a 64kbps is proposed. In addition to the interframe and CRC (Conditional Repleni shment Coding) system, BTVQ (Binary Tree searched Vector Quantizer)and RLC (Run Length Coding) methods are incorporated. Couble buffer memory is used for simple comtrol of channel symbol transmission and memory underflow And also buffer memory onerfolw is easily controlled by the thresholds of a MAD (Moving Area Betector)

• 한국광학회:학술대회논문집
• /
• 한국광학회 1989년도 제4회 파동 및 레이저 학술발표회 4th Conference on Waves and lasers 논문집 - 한국광학회
• /
• pp.89-94
• /
• 1989

In this paper, we propose a hybrid optical/digital version of the associative memory which improve hardware efficiency and increase convergence rates. Multifocus hololens are used as space-varient optical element for performing inner product and summation function. The real-time input and the stored states of memory matrix is formated using LCTV. One method of adaptively changing the weights of stored vectors during each iteration is implemented electronically. A design for a optical implementation scheme is discussed and the proposed architecture is demonstrated the ability of retrieving with computer simmulation.

• 한국차세대컴퓨팅학회논문지
• /
• 제13권2호
• /
• pp.49-61
• /
• 2017

트랜잭셔널 메모리는 트랜잭션 처리를 위한 병렬 프로그래밍 패러다임을 크게 바꾸었으며, 하드웨어 및 소프트웨어 방식에 따라 STM, HTM, HyTM으로 구분된다. 그러나, 기존 연구들은 모든 워크로드에 대해 획일적인 재시도 정책을 제공하는 문제점이 존재한다. 이러한 문제점을 해결하기 위해, 본 논문에서는 멀티코어 환경에서 샘플링 기반 유연한 재시도 정책을 이용한 하이브리드 트랜잭셔널 메모리 기법을 제안한다. 첫째, 제안하는 기법은 트랜잭션의 특성을 파악하여 HTM 혹은 STM을 선택하여 수행하거나, 블룸필터를 이용하여 동시에 HTM과 STM을 수행한다. 둘째, 제안하는 기법은 각 워크로드 내의 트랜잭션의 특성을 반영한 유연한 HTM 재시도 정책을 제공한다. 마지막으로 STAMP를 이용한 성능평가를 통해, 제안하는 기법이 기존 연구에 비해 10~20%의 성능 향상이 있음을 보인다.

• 한국컴퓨터정보학회논문지
• /
• 제24권7호
• /
• pp.1-8
• /
• 2019

Recently, an SSD with hybrid buffer memories is actively researching to reduce the overall latency in server computing systems. However, existing hybrid buffer policies caused many swapping operations in pages because it did not consider the overall latency such as read/write operations of flash chips in the SSD. This paper proposes the clock with hybrid buffer memories (CLOCK-HBM) for a new hybrid buffer policy in the SSD with server computing systems. The CLOCK-HBM constructs new policies based on unique characteristics in both DRAM buffer and NVMs buffer for reducing the number of swapping operations in the SSD. In experimental results, the CLOCK-HBM reduced the number of swapping operations in the SSD by 43.5% on average, compared with LRU, CLOCK, and CLOCK-DNV.

• 대한기계학회논문집A
• /
• 제23권11호
• /
• pp.1904-1911
• /
• 1999

The most effective material in the shape memory alloy(SMA) is the TiNi alloy, because its shape recovery characteristics are very excellent. We molded the composite material with shape memory function. The fiber of it is $Ti_{50}-Ni_{50}$ shape memory alloy and matrix of it is epoxy resin(Araldite B41, Hardner HT903. Ciba Geigy), its adhesive and optical sensitivity are very excellent. It was assured that the composite material could be used as model material of photoelastic experiment for intelligent materials or structures. In this research, the composite material with shape memory function is used as model material of photoelastic experiment. Photoelastic experimental hybrid method is developed in this research, it is assured that it is useful on the obtaining stress intensity factor and the separation of stress components from only isochromatic data. The measuring method of stress intensity factor of intelligent material by photoelastic experiment is introduced. In the mode I state, we can know that stress intensity factors are decreased more than 50% of stress intensity factor of room temperature when temperature of fiber is greater than 4$0^{\circ}C$, prestrain greater than 5% and fiber volume ratio greater than 0.42% and that stress intensity factors are decreased by 100% when fiber volume ratio is greater than 0.84%, prestrain greater than 5% and temperature greater than 60 $^{\circ}C$.

• 디지털콘텐츠학회 논문지
• /
• 제18권1호
• /
• pp.93-99
• /
• 2017

Long Short Term Memory (LSTM) Recurrent Neural Network (RNN)를 이용한 hybrid 방법은 음성 인식률을 크게 향상시켰다. Hybrid 방법에 기반한 음향모델을 학습하기 위해서는 Gaussian Mixture Model (GMM)-Hidden Markov Model (HMM)로부터 forced align된 HMM state sequence가 필요하다. 그러나, GMM-HMM을 학습하기 위해서 많은 연산 시간이 요구되고 있다. 본 논문에서는 학습 속도를 향상하기 위해, LSTM RNN 기반 한국어 음성인식을 위한 end-to-end 방법을 제안한다. 이를 구현하기 위해, Connectionist Temporal Classification (CTC) 알고리즘을 제안한다. 제안하는 방법은 기존의 방법과 비슷한 인식률을 보였지만, 학습 속도는 1.27 배 더 빨라진 성능을 보였다.