• Proceedings of the Korea Information Processing Society Conference
• /
• 2024.05a
• /
• pp.47-50
• /
• 2024

희소 행렬은 대부분의 요소가 0 인 행렬이다. 이러한 희소 행렬-행렬 곱셈을 수행할 경우 0 인 데이터 또한 곱셈을 수행하니 불필요한 연산이 발생한다. 이러한 문제를 해결하고자 행렬 압축 알고리즘 또는 곱셈의 부분합의 수를 줄이는 연구들이 활발히 진행 중이다. 하지만 현재의 연구들은 주로 단일 행렬 연산에 집중되어 있어 FPGA(Field Programmable Gate Array)와 특정 용도로 사용하는 가속기에서는 리소스를 충분히 활용하지 못해 비효율적이다. 본 연구는 FPGA 의 모든 리소스를 사용하여 다중 희소 행렬 곱셈을 수행하는 아키텍처를 제안한다.

• ETRI Journal
• /
• v.42 no.4
• /
• pp.491-504
• /
• 2020

We present AB9, a neural processor for inference acceleration. AB9 consists of a systolic tensor core (STC) neural network accelerator designed to accelerate artificial intelligence applications by exploiting the data reuse and parallelism characteristics inherent in neural networks while providing fast access to large on-chip memory. Complementing the hardware is an intuitive and user-friendly development environment that includes a simulator and an implementation flow that provides a high degree of programmability with a short development time. Along with a 40-TFLOP STC that includes 32k arithmetic units and over 36 MB of on-chip SRAM, our baseline implementation of AB9 consists of a 1-GHz quad-core setup with other various industry-standard peripheral intellectual properties. The acceleration performance and power efficiency were evaluated using YOLOv2, and the results show that AB9 has superior performance and power efficiency to that of a general-purpose graphics processing unit implementation. AB9 has been taped out in the TSMC 28-nm process with a chip size of 17 × 23 ㎟. Delivery is expected later this year.

• Journal of Broadcast Engineering
• /
• v.26 no.2
• /
• pp.167-174
• /
• 2021

Recently, deep learning technology is widely used in various computer vision applications, such as object recognition, classification, and image generation. In particular, the deep learning-based super-resolution has been gaining significant performance improvement. Fast super-resolution convolutional neural network (FSRCNN) is a well-known model as a deep learning-based super-resolution algorithm that output image is generated by a deconvolutional layer. In this paper, we propose an FPGA-based convolutional neural networks accelerator that considers parallel computing efficiency. In addition, the proposed method proposes Optimal-FSRCNN, which is modified the structure of FSRCNN. The number of multipliers is compressed by 3.47 times compared to FSRCNN. Moreover, PSNR has similar performance to FSRCNN. We developed a real-time image processing technology that implements on FPGA.

• Journal of IKEEE
• /
• v.25 no.4
• /
• pp.762-765
• /
• 2021

In this paper, an automotive ECU (electronic control unit) with Kalman filter accelerator is designed and implemented. RISC-V is exploited as a processor core. Accelerator for Kalman filter matrix operation, CAN (controller area network) controller for in-vehicle network, and LIN (local interconnect network) controller are designed and embedded. Kalman filter operation consists of time update process and measurement update process. Current state variable and its error covariance are estimated in time update process. Final values are corrected from input measurement data and Kalman gain in measurement update process. Usually floating-point multiplication is exploited in software implementation, but fixed-point multiplier considering accuracy analysis is exploited in this paper to reduce hardware area. In 28nm silicon fabrication, its operating frequency, area, and gate counts are 100MHz, 0.37mm², and 760k gates, respectively.

• Journal of IKEEE
• /
• v.26 no.3
• /
• pp.364-372
• /
• 2022

This paper proposes the design and implementation results for human and object classification systems utilizing frequency modulated continuous wave (FMCW) radar sensor. Such a system requires the process of radar sensor signal processing for multi-target detection and the process of deep learning for the classification of human and object. Since deep learning requires such a great amount of computation and data processing, the lightweight process is utmost essential. Therefore, binary neural network (BNN) structure was adopted, operating convolution neural network (CNN) computation in a binary condition. In addition, for the real-time operation, a hardware accelerator was implemented and verified via FPGA platform. Based on performance evaluation and verified results, it is confirmed that the accuracy for multi-target classification of 90.5%, reduced memory usage by 96.87% compared to CNN and the run time of 5ms are achieved.

• Journal of the Korea Institute of Information and Communication Engineering
• /
• v.24 no.10
• /
• pp.1331-1340
• /
• 2020

Object detection in maritime means that the captain detects floating objects that has a risk of colliding with the ship using the computer automatically and as accurately as human eyes. In conventional ships, the presence and distance of objects are determined through radar waves. However, it cannot identify the shape and type. In contrast, with the development of AI, cameras help accurately identify obstacles on the sea route with excellent performance in detecting or recognizing objects. The computer must calculate high-volume pixels to analyze digital images. However, the CPU is specialized for sequential processing; the processing speed is very slow, and smooth service support or security is not guaranteed. Accordingly, this study developed maritime object detection software and implemented it with FPGA to accelerate the processing of large-scale computations. Additionally, the system implementation was improved through embedded boards and FPGA interface, achieving 30 times faster performance than the existing algorithm and a three-times faster entire system.

• The KIPS Transactions:PartA
• /
• v.9A no.3
• /
• pp.327-332
• /
• 2002

Conventional proxy web cache, which is generally used to caching server, is a content-copy based system. This method focuses on speeding up the phase delivery not improving the webserver performance. However, if immense clients attempt to connect the webserver simultaneously, the proxy web cache cannot achieve the desired result. In this paper, we propose the web accelerator called the SCALA-AX, whitch improves web server performance by accelerating the delivery contents. The SCALA-AX is built in the Linux-based kernel as a kernel modulo and works in combination with the conventional webserver program. The SCALA-AX speeds up the processing rate of the webserver, because it processes the requests using the kernel thread. The SCALA-AX also applies the well-developed cache algorithm to the processing, and thus it obtains the advantage of the caching server without installing additional hardware. A banchmarking test demonstrates that the SCALA-AX improves webserver performance by up to 500％ for content delivery.

• Journal of Advanced Navigation Technology
• /
• v.26 no.4
• /
• pp.211-218
• /
• 2022

Continuous wave (CW) radar has the advantage of reliability and accuracy compared to other sensors such as camera and lidar. In addition, binarized neural network (BNN) has a characteristic that dramatically reduces memory usage and complexity compared to other deep learning networks. Therefore, this paper proposes binarized neural network based human identification and motion classification system using CW radar. After receiving a signal from CW radar, a spectrogram is generated through a short-time Fourier transform (STFT). Based on this spectrogram, we propose an algorithm that detects whether a person approaches a radar. Also, we designed an optimized BNN model that can support the accuracy of 90.0% for human identification and 98.3% for motion classification. In order to accelerate BNN operation, we designed BNN hardware accelerator on field programmable gate array (FPGA). The accelerator was implemented with 1,030 logics, 836 registers, and 334.904 Kbit block memory, and it was confirmed that the real-time operation was possible with a total calculation time of 6 ms from inference to transferring result.

• Journal of the Institute of Electronics Engineers of Korea SD
• /
• v.45 no.10
• /
• pp.53-58
• /
• 2008

This paper proposes a new set of instructions to improve the packet processing capacity of a network processor. The proposed set of instructions is able to achieve more efficient packet processing by accelerating integration of packet headers. Furthermore, a hardware configuration dedicated to processing overlay instructions was designed to reduce additional hardware cost. For this purpose, the basic architecture for the network processor was designed using LISA and the overlay block was optimized based on the barrel shifter. The block was synthesized to compare the area and the operation delay, and allocated to a C-level macro function using the compiler known function (CKF). The improvement in performance was confirmed by comparing the execution cycle and the execution time of an application program. Experiments were conducted using the processor designer and the compiler designer from Coware. The result of synthesis with the TSMC ($0.25{\mu}m$) from Synopsys indicated a reduction in operation delay by 20.7% and an improvement in performance of 30.8% with the proposed set of instructions for the entire execution cycle.

• Journal of IKEEE
• /
• v.22 no.3
• /
• pp.805-809
• /
• 2018

In this paper, we analyze the change of accuracy when fixed point arithmetic is used instead of floating point arithmetic in binary weight network(BWN). We observed the change of accuracy by varying total bit size and fraction bit size. If the integer part is not changed after fixed point approximation, there is no significant decrease in accuracy compared to the floating-point operation. When overflow occurs in the integer part, the approximation to the maximum or minimum of the fixed point representation minimizes the decrease in accuracy. The results of this paper can be applied to the minimization of memory and hardware resource requirement in the implementation of FPGA-based BWN accelerator.