확장형 VLSI 리바운드 정렬기의 설계

윤지헌,안병철,Yun, Ji-Heon,Ahn, Byoung-Chul 한국정보처리학회 1995 정보처리논문지 Vol.2 No.3

시간 복잡도가 O(Ν)인 고집적 회로(VLSI)의 병렬 정렬기 설계에 관한 논문이다. 발표된 빠른 VLSI 정렬 알고리즘은 Ν개의 데이타를 정렬하기 위해 O(log Ν)시간 복 잡도를 가지고 있다. 그러나 이러한 알고리즘은 입출력 시간을 고려하지 않고, 복잡 한 네트워크 구조를 가지므로 확장이나 실용화하기 힘들다. 입출력 시간이 포함된 병 렬 정렬 알고리즘들의 칩면적과 시간 복잡도를 분석한 후 가장 효과적인 rebound sort 이론을 확장하여 VLSI로 구현한다. 이 리바운드 정렬기는 파이프라인으로 구성하여 O(Ν)의 시간 복잡도를 가지며 한 개의 칩에 8개의 16비트 레코드를 정렬할 수 있다. 그리고 이 정렬 칩은 확장성을 가지고 있어 수직으로 연결할 경우 8개 이상의 레코드 를 정렬할 수 있다. This paper presents an improved VLSI implementation of a parallel sorter to achieve O(Ν) time complexity. Many fast VLSI sort algorithms have been proposed for sorting N elements in O(log Ν) time. However, most such algorithms proposed have complex network structure without considering data input and output time. They are also very difficult to expand or to use in real applications. After analyzing the chip area and time complexity of several parallel sort algorithms with overlapping data input and output time, the most effective algorithm, the rebound sort algorithm, is implemented in VLSI with some improvements. To achieve O(Ν) time complexity, an improved rebound sorter is able to sort 8 16-bits records on a chip. And it is possible to sort more than 8 records by connecting chips in a chain vertically.

CAD : 확장형 VLSI 리바운드 정렬기의 설계

윤지헌(Yun Ji Heon),안병철(Ahn Byoung Chul) 한국정보처리학회 1995 정보처리학회논문지 Vol.2 No.3

시간 복잡도가 O(N)인 고집적 회로(VLSI)의 병렬 정렬기 설계에 관한 논문이다. 발표된 빠른 VLSI 정렬 알고리즘은 N개의 데이타를 정렬하기 위해 O(log N) 시간 복잡도를 가지고 있다. 그러나 이러한 알고리즘은 입출력 시간을 고려하지 않고, 복잡한 네트워크 구조를 가지므로 확장이나 실용화하기 힘들다. 입출력 시간이 포함된 병렬 정렬 알고리즘들의 칩면적과 시간 복잡도를 분석한 후 가장 효과적인 rebound sort이론을 확장하여 VLSI로 구현한다. 이 리바운드 정렬기는 파이프라인으로 구성하여 O(N)의 시간 복잡도를 가지며 한 개의 칩에 8개의 16비트 레코드를 정렬할 수 있다. 그리고 이 정렬 칩은 확장성을 가지고 있어 수직으로 연결할 경우 8개 이상의 레코드를 정렬할 수 있다. This paper presents an improved VLSI implementation of a parallel sorter to achieve O(N) time complexity. Many fast VLSI sort algorithms have been proposed for sorting N elements in O(log N) time. However, most such algorithms proposed have complex network structure without considering data input and output time. They are also very difficult to expand or to use in real applications. After analyzing the chip area and time complexity of several parallel sort algorithms with overlapping data input and output time, the most effective algorithm, the rebound sort algorithm, is implemented in VLSI with some improvements. To achieve O(N) time complexity, an improved rebound sorter is able to sort 8 16-bits records on a chip. And it is possible to sort more than 8 records by connecting chips in a chain vertically.

디스크 스케줄링 방법과 RAIN level에 따른 디스크 엑세스 성능 분석

윤지헌 영남이공대학 2002 論文集 Vol.31 No.-

For some decades, processing speed of microprocessors has been grown up 40% to 60% in a year, memory capacity has been grown up 4 times every 2 or 3 years, and disk storage capacity has increased tremendously. But, disk access time has been improved only about 7% per year, and this leads to I/O crisis. For higher performance needed to relieve the I/O bottleneck and to provide greater storage capacities for data intensive applications, disk array using inexpensive disks appeared. In this study, the performances of disk access were compared for the methods of disk scheduling FCFS, SSTF, SCAN and CVSCAN, and RAID levels 0, 4, 5. The results were simulated with Berkeley raidSim simulator.

效率的인 한글 符號에 관한 硏究(Ⅱ)

尹志憲 영남이공대학 1984 論文集 Vol.13 No.-

Because of special characteristics of Korean character, the files of Korean data, that had not processed any other conversion, couldn't sort out for dictionary order of Korean. In the prior report, the codes that have two kinds of vowel which one has not the last consonant, the other has the last consonant in Korean character, were offered. Using this offered code in the prior report, it is able to sort out for dictionary order of Korean without "gather-writing" edition. In this paper, the algorithm for input conversion and "gather-writing" edition using offered code is processed.

PC에서 數値計算의 誤差에 關한 考察

尹志憲 영남이공대학 1988 論文集 Vol.17 No.-

This paper introduces errors generated by hardware architecture in personal computer and by software used for problem solving. The errors in accuracy never occur in calculation for fixed-point number, but for floating-point number, occur several kinds of errors like truncation, number system conversion and roundoff error. And the errors for floating-point number are differed in according to software used. In this paper, the errors in GW-BASIC and IBM S/36 FORTRAN were compared. From these results, it is observed that the accuracy in GW-I3ASIC is better than that of IBM S/36 because of the method of number representation and the method of roundoff process in the result of calculation.

CMOS-Ic를 使用한 3値 Exclusive-OR回路의 構成

尹志憲 영남이공대학 1983 論文集 Vol.12 No.-

This paper gives a definition of the ternary Exclusive-OR logic function from the concept of binary Exclusive-OR, that is, if two inputs are equal, the output is "0" state, and the other states are ORed. The logic expression of ternary Exclusive-OR is simplified using the standard form of the special sum of products. Figure 7 shows that the output waveform coincides with the function of ternary Exclusive-OR logic. This circuit was composed of CMOS-ICs only, and if it is fabricated in integrated circuit, it is useful for other ternary logic circuits.

PDP-11 FORTRAN Ⅳ에서의 INTEGER*4 定數 表現方法에 關하여

尹志憲 영남이공대학 1980 論文集 Vol.9 No.-

There are occasions when it is necessary to have the computer represent constants and/or variables in the double integer INTEGER*4. Generally, INTEGER*4 in FORTRAN Ⅳ languages is permitted on almost all computer systems with the exception of POP-11 series. In POP-Ⅱ FORTRAN If, an INTEGERs4 declaration allocates two words, which are limited to the values under ±2^(15); furthermore, INTEGER*4 constants over ±2^(15) are not permitted for input or output. This paper presents the method to convert the input string of INTEGER*4 constants into internal representation and its internal representation into output string by means of the shift-control. This shift-control method allows more than two words to be used to present integer constants.

'79학년도 본대학 신입생에 대한 학업성적 분석

김정두,윤지헌 영남이공대학 1979 論文集 Vol.8 No.-

This study is one of the continuous work following the preceding analysis in 1977 which has suggested some data in the better improvement of evaluation methods on the students' achievement. Throughout the study, the academic records of 1,699 YJCT freshmen have been used as the reference materials. The main results obtained by this study have revealed the following: First, differences of standard in marking the test scores among the individual course subjects are still existing. Second, the achievement scores of the general selection students for the college admission are higher than the special selection students in the subjects of general education course such as Korean, Ethics, History, English and Mathematics, but contrary lower in the major subjects. Third, the scores for college admission including both Preliminary College Entrance Examination and High School Achievement are not so highly correlated with the college achievement scores as expected. Fourth, no distinct differences have been found according to the years passed after the high school graduation. The college academic achievement scores of the technical high school graduates, however, seemed to be influenced by the reputation of their parent schools. With the above results, it is suggested that a professional study to develop the better selection system of entrance examination, to improve the methods for instruction, evaluation and organization of the course subjects, and to render more effective student guidance should be continued.

마이크로프로세서 시스템에 의한 한국어 모음의 실시간 인식에 관한 고찰

최성운,윤지헌 영남이공대학 1988 論文集 Vol.17 No.-

This paper introduces on development of a highly dependable real-time recognition systemfof Korean-vowels. In this paper, Korean monophthongs, /아, 어, 오, 우, 으, 이/. and diphthongs. /야, 여, 요, 유/. are selected and an APPLE computer system is used for the recognition experiment. The real-time recognition and reduction of information using microprocessor system is generally difficult. but was solved through the extraction of the fornuurt frequency of a vowel using digital filter. It was observed that, for a specific speaker, the rate of recognition for the monophthongs is upto 98% and for the dipthongs upto 92%, for an unspecific speaker, the rate of recognition for the monophthongs is upto 84% and for the diphthongs upto 68% in 0.8 seconds.

構造化 FORTRAN을 위한 Preprocessor의 開發에 關한 硏究 (Ⅰ)

金明昊,尹志憲 영남이공대학 1984 論文集 Vol.13 No.-

This paper describes the design and implementation of the preprocessor which translates from the extented FORTRAN code for the structured programming to standard FORTRAN code. The basic logic structures required in structured coding in FORTRAN and its emulations are suggested. Finally the detailed algoritbms for implementing the preprocessor are discussed.