• Proceedings of the IEEK Conference
• /
• 2006.06a
• /
• pp.573-574
• /
• 2006

A low-power power-aware scalable pipelined Booth recoded multiply & Accumulate unit (PA-MAC) detects the input operands for their dynamic range and accordingly implements a 16-bit, 8-bit or 4-bit multiplication and accumulation operation. The multiplication mode is determined by the dynamic - range detection unit. For the computations, although an area of the proposed PA-MAC is lager than a non-scalable MAC respectively, the proposed PA-MAC proves to be globally more power efficient than a non-scalable MAC.

• Proceedings of the IEEK Conference
• /
• 2000.11b
• /
• pp.349-352
• /
• 2000

In this paper, a SIMD 64bit MAC (Multiply -Accumulate) unit is designed. It is composed of two 32bit MAC unit which supports SIMD 16bit operations. As a result, It can process two 32bit MAC operations or four 16bit operations in one cycle. Proposed MAC unit is described in Verilog HDL. After functional verification is performed, MAC unit is synthesized and optimized with 0.35$\mu\textrm{m}$ standard cell library. The synthesis result shows that this MAC unit can operate at 80㎒ of clock frequency in 85$^{\circ}C$, 3.0V, worst case process and 125㎒ of clock frequency at 25$^{\circ}C$, 3.3V, typical case process. It achieves 320Mops of performance, and is suitable for embedded DSP processors.

• The Journal of Korean Institute of Communications and Information Sciences
• /
• v.29 no.10C
• /
• pp.1460-1468
• /
• 2004

In this paper, we developed a low-area and low-cost SIMD MAC/MAS(Single Instruction Multiple Data Multiply and ACcumulate/Multiply And Subtract) for multimedia that is used much in real life. We compared the result of this research with a previously developed more large and high performance SIMD MAC/MAS. This paper is consist of 5 parts, which are an introduction, the contents of designing SIMD MAC/MAS hardware, a special qualities for previous works, the result of synthesis and conclusion. The design result reduced by size 32% of whole hardware than 64 bit SIMD MAC/MAS block of designed for high performance. This improved ISA (Instruction Set Architecture) to be suitable to embedded DSP(Digital Signal Processor), and shortened bit range of 64-bit data to 32-bit and implement more optimally.

• Journal of IKEEE
• /
• v.28 no.3
• /
• pp.375-381
• /
• 2024

In this paper, we propose a lightweight object classification system using a LiDAR sensor for delivery service robots. The 3D point cloud data is encoded into a 2D pseudo image using a Pillar Feature Network (PFN), and then passed through a lightweight classification network designed based on Depthwise Separable Convolutional Neural Networks (DS-CNN). The implementation results show that the designed classification network has 9.08K parameters and 3.49M Multiply-Accumulate (MAC) operations, while supporting a classification accuracy of 94.94%.

• Journal of the Institute of Electronics Engineers of Korea SD
• /
• v.38 no.12
• /
• pp.44-55
• /
• 2001

MAC(Multiply and ACcumulate) is the core operation of multimedia data processing. Because MAC units implemented on traditional DSP units or embedded processors have latency of three cycles and cannot operate on multiple data simultaneously, then, performances are seriously limited. Many high end general purpose microprocessors have SIMD MAC unit as a functional unit. But these high end MAC units must support pipeline structure for various operation modes and high clock frequency, which makes control logic complex and increases chip area. In this paper, a 64bit SIMD MAC unit for embedded processors is designed. It is implemented to have a latency of one clock cycle to remove pipeline control logics and a minimal area overhead for SIMD support is added to existing Booth multipliers.

• Journal of the Institute of Electronics Engineers of Korea SD
• /
• v.39 no.2
• /
• pp.72-81
• /
• 2002

A new high-performance DSP architecture is proposed, which behaves as a coprocessor of a 32bit microcontroller. Because the proposed DSP architecture is a dual MAC(Multiply and Accumulate) DSP architecture, it can process efficiently a number of SOP(sum of product) operations used in many DSP applications. In order to efficiently perform other operations such as pure additions without any restriction, a MAC is composed of a multiplier and a ALU placed in parallel. In addition, it is a 3-way superscalar architecture, which can issue 3 instructions at a time. The benchmark results with 3 thor dual MAC DSPs show that the proposed DSP has the best performance. Futhermore, it is proven that the proposed DSP is more efficient in memory usage, although the performance is comparable in some algorithms such as Viterbi decoding and FFT butterfly.

• The Journal of the Acoustical Society of Korea
• /
• v.38 no.1
• /
• pp.9-22
• /
• 2019

In this paper, we propose a method that can be applied to the efficient implementation of multi-channel active noise controller. Since the normalized MFxLMS (Modified Filtered-x Least Mean Square) algorithm for the multi-channel active noise control requires a large amount of computation, the difficulty has lied in implementing the algorithm using a low-cost MCU (Microcontoller Unit). We implement the multi-channel active noise controller efficiently by optimizing the software based on the features of the MCU. By maximizing the usage of single-cycle MAC (Multiply- Accumulate) operations and minimizing move operations of the delay memory, we can achieve more than 3 times the performance in the aspect of computational optimization, and by parellel processing using the auxillary processor included in the MCU, we can also obtain more than 4 times the performance. In addition, the usage of additional parts can be minimized by maximizing the usage of the peripherals embedded in the MCU.

• Journal of information and communication convergence engineering
• /
• v.7 no.4
• /
• pp.516-520
• /
• 2009

RSA crypto-processors equipped with more than 1024 bits of key space handle the entire key stream in units of blocks. The RSA processor which will be the target design in this paper defines the length of the basic word as 128 bits, and uses an 256-bits register as the accumulator. For efficient execution of 128-bit multiplication, 32b*32b multiplier was designed and adopted and the results are stored in 8 separate 128-bit registers according to the status flag. In this paper, an efficient method to execute 128-bit MAC (multiplication and accumulation) operation is proposed. The suggested method pre-analyzed the all possible cases so that the MAC unit can remove unnecessary calculations to speed up the execution. The proposed architecture prototype of the MAC unit was automatically synthesized, and successfully operated at 20MHz, which will be the operation frequency in the RSA processor.

• IEMEK Journal of Embedded Systems and Applications
• /
• v.4 no.1
• /
• pp.1-8
• /
• 2009

This paper proposes a 24-bit floating point multiply and accumulate(MAC) unit that can be used in geometry transformation process in 3D graphics. The MAC unit is composed of floating point multiplier and floating point accumulator. When separate multiplier and accumulator are used, matrix calculation, used in the transformation process, can't use continuous accumulation values. In the proposed MAC unit the accumulator can get continuous input from the multiplier and the calculation time is reduced. The MAC unit uses about 4,300 gates and can be operated at 150 MHz frequency.

• ETRI Journal
• /
• v.24 no.5
• /
• pp.391-397
• /
• 2002

This paper presents digital signal processor (DSP) instructions and their data processing unit (DPU) architecture for high-speed fast Fourier transforms (FFTs) in orthogonal frequency division multiplexing (OFDM) systems. The proposed instructions jointly perform new operation flows that are more efficient than the operation flow of the multiply and accumulate (MAC) instruction on which existing DSP chips heavily depend. We further propose a DPU architecture that fully supports the instructions and show that the architecture is two times faster than existing DSP chips for FFTs. We simulated the proposed model with a Verilog HDL, performed a logic synthesis using the 0.35 ${\mu}m$ standard cell library, and then verified the functions thoroughly.