• Journal of the Institute of Electronics Engineers of Korea SD
• /
• v.48 no.10
• /
• pp.82-90
• /
• 2011

Because of the development of Smart phone devices, the demands of high performance FPU(Floating-point Unit) becomes increasing. Therefore, we propose the high-speed single-/double-precision FPU design that includes an elementary add/sub unit and improved multiplier and compare and convert units. The most commonly used add/sub unit is optimized by the parallel rounding unit. The matrix operation is used in complex calculation something like a graphic calculation. We designed the Multiply-Add Fused(MAF) instead of multiplier to calculate the matrix more quickly. The branch instruction that is decided by the compare operation is very frequently used in various programs. We bypassed the result of the compare operation before all the pipeline processes ended to decrease the total execution time. And we included additional convert operations that are added in IEEE754-2008 standard. To verify our RTL designs, we chose four hundred thousand test vectors by weighted random method and simulated each unit. The FPU that was synthesized by Samsung's 45-nm low-power process satisfied the 600-MHz operation frequency. And we confirm a reduction in area by comparing the improved FPU with the existing FPU.

• Journal of the Institute of Electronics Engineers of Korea SD
• /
• v.44 no.10
• /
• pp.49-54
• /
• 2007

As digital contents based on 3D graphics are increased, the requirement for low power 3D graphic hardware for mobile devices is increased. We design a transformation engine for mobile 3D graphic processor. We propose a simplified transformation engine for mobile 3D graphic processor. The area of the transformation engine is reduced by merging a mapping transformation unit into a projective transformation unit and by replacing a clipping unit with a selection unit. It consists of a viewing transformation unit a projective transformation unit a divide by w nit, and a selection unit. It can process 32 bit floating point format of the IEEE-754 standard or a reduced 24 bit floating point format. It has a pipelined architecture so that a vertex is processed every 4 cycles except for the initial latency. The RTL code is verified using an FPGA.

• Journal of Broadcast Engineering
• /
• v.13 no.3
• /
• pp.415-418
• /
• 2008

Some embedded microprocessors do not have an FPU(Floating Point Unit) due to a circuit complexity and power consumption. The performance speed of MPEG-4 AAC decoder on this hardware environment would be slower than corresponding speed for playing back of the decoded results. Therefore, irritating and high-pitched noises are interleaved in the original the audio data. So, in order to play MPEG-4 AAC file on such PDA, a new algorithm that transforms floating-point arithmetic to one with integers, is needed. We have developed a transformation algorithm from floating-point operation to integer operation and implemented the PDA's AAC Player. We also show the efficiency of our proposed method with the experimental results.

• Proceedings of the Korean Institute of Information and Commucation Sciences Conference
• /
• 2010.05a
• /
• pp.320-323
• /
• 2010

This paper proposes efficient log and exponent calculation methods using a dual phase instruction set without additional ALU unit for a mobile enviroment. Using the Dual Phase Instruction set, it extracts exponent and mantissa from expression of floating point and calculates 24bit single precision floating point of log approximation using the Taylor series expansion algorithm. And with dual phase instruction set, it reduces instruction excution cycles. The proposed Dual Phase architecture reduces the performance degradation and maintain smaller size.

• Journal of the Korea Institute of Information and Communication Engineering
• /
• v.19 no.10
• /
• pp.2341-2349
• /
• 2015

The commonly used Goldschmidt's floating-point divider algorithm performs two multiplications in one iteration. In this paper, a tentative error corrected K'th Goldschmidt's floating-point number divider algorithm which performs K times multiplications in one iteration is proposed. Since the number of multiplications performed by the proposed algorithm is dependent on the input values, the average number of multiplications per an operation in single precision and double precision divider is derived from many reciprocal tables with varying sizes. In addition, an error correction algorithm, which consists of one multiplication and a decision, to get exact result in divider is proposed. Since the proposed algorithm only performs the multiplications until the error gets smaller than a given value, it can be used to improve the performance of a divider unit. Also, it can be used to construct optimized approximate reciprocal tables.

• ETRI Journal
• /
• v.10 no.3
• /
• pp.36-48
• /
• 1988

본 논문에서는 부동소숫점연산 프로세서들의 최근 동향을 설명하면서 부동소숫점 연산기의 중요성을 강조하고, 한국전자통신연구소 프로세서구조연구실에서 개발하고 있는 HARP(High-performance Architecture for RISC type Processor)의 개발전략에 따른 부동소숫점 연산기(Floating-Point Unit : FPU)의 구조를 정의한다. 또한 HARP FPU의 설계구현을 마이크로 구조측면에서 설명한다. HARP의 CPU와 동일 칩상에 구현될 HARP FPU는 고유의 구조를 가지며 모든 부동소숫점 연산은 IEEE-754 표준을 따른다. HARP FPU는 고속의 부동소숫점 연산 유니트이며, HARP의 IPU(Integer Processing Unit)와는 독립적으로 동작되도록 설계되어서 HARP CPU의 전체적인 파이프라인 기능에 가능한 한 페날티를 주지 않도록 동작된다.

• Proceedings of the KIEE Conference
• /
• 2001.11c
• /
• pp.165-168
• /
• 2001

This paper presents a design of a divide unit supporting IEEE-754 floating point standard single-precision with 32-bit word length. Its functions have been verified with ALTERA MAX PLUS II tool. For a high-speed division operation, the radix-4 non-restoring algorithm has been applied and CLA(carry-look -ahead) adders has been used in order to improve the area efficiency and the speed of performance for the fraction division part. The prevention of the speed decrement of operations due to clocking has been achieved by taking advantage of combinational logic. A quotient select block which is very complicated and significant in the high-radix part was designed by using P-D plot in order to select the fast and accurate quotient. Also, we designed all division steps with Gate-level which visualize the operations and delay time.

• Journal of the Korean Institute of Telematics and Electronics C
• /
• v.34C no.8
• /
• pp.71-78
• /
• 1997

An LPC cepstrum processor for speech recognition is implemented on CMOS array process. The designed processor contains a 24-bit floating-point MAC unit to perform the correlation quickly, which occupies the majority of operations used in the algorithm, and has 22 register files to store temporary variables. For the purpose of fast operations, the floating-point MAC consists of a 3-stage pipeline and the new post-normalization shceme is proposed and applied to it. Experimental result shows that it takes approximately 266.mu.s to process 200 samples/frame at 15 MHz clock rate. This processor runs at the maximum rate of 16.6 MHz and the number of gates are 27,760.

• Electronics and Telecommunications Trends
• /
• v.37 no.1
• /
• pp.53-62
• /
• 2022

With increasing size of transformer-based neural networks, a light-weight algorithm and efficient AI accelerator has been developed to train these huge networks in practical design time. In this article, we present a survey of state-of-the-art research on the low-precision computational algorithms especially for floating-point formats and their hardware accelerator. We describe the trends by focusing on the work of two leading research groups-IBM and Seoul National University-which have deep knowledge in both AI algorithm and hardware architecture. For the low-precision algorithm, we summarize two efficient floating-point formats (hybrid FP8 and radix-4 FP4) with accuracy-preserving algorithms for training on the main research stream. Moreover, we describe the AI processor architecture supporting the low-bit mixed precision computing unit including the integer engine.

• The Journal of Korean Institute of Communications and Information Sciences
• /
• v.26 no.5B
• /
• pp.685-691
• /
• 2001

본 논문은 4단계 파이프 라인과 VLIW (Very Long Instruction Word) 구조를 갖는 FLOVA라는 DSP 프로세서의 테스트용이 설계 기법을 다룬다. Full-scan design, BIST(Built-In-Self-Test), IEEE 1149.1의 기법들이 플립플롭과 floaing point unit, 내장된 메모리, I/O cell 등에 각각 적용되었다. 이러한 기법들은 테스트 용이도의 관점에서 FLOVA의 구조에 적절하게 적용되었다. 본 논문에서는 이와 같이 FLOVA에 적용된 테스트 용이 설계의 특징들을 중심으로 상세하게 기술한다.