Pub Date : 2024-01-28DOI: 10.1109/ICEIC61013.2024.10457116
Dong-Hoe Heo, Tae-Hyeon Kim, Kwang-Ho Lee, Min-Seong Choo
This paper presents a 32-Gb/s full-rate clock and data recovery (CDR) architecture based on a phase interpolator (PI), which incorporates high-speed channel equalization to widen the ppm tolerance or locking range. This work focuses on the receiver-side implementation. Therefore, the feed-forward equalizer (FFE) tap is not utilized, and only the voltage swing level is adjusted at the transmitter side. The channel is modeled using Verilog language with a -10 dB loss at 10 GHz. The overall architecture comprises several components: a continuous-time linear equalizer (CTLE), 1-tap decision feedback equalizer (DFE), 7-bit PI, digital loop filter, and 2x oversampling phase detector. By individually employing the DFE and CTLE, the optimal tap coefficient value for the DFE, which produces the widest eye pattern, and the pole and zero positions of the CTLE are determined. Finally, CTLE and 1-tap DFE ensure optimal vertical 322 mV and timing margin of 25.8 ps. It also relaxes phase modulation to obtain acceptable error of the phase interpolator up to ±15625 ppm.
{"title":"An Analysis of 32-Gb/s and Full-Rate Phase Interpolator based Clock and Data Recovery","authors":"Dong-Hoe Heo, Tae-Hyeon Kim, Kwang-Ho Lee, Min-Seong Choo","doi":"10.1109/ICEIC61013.2024.10457116","DOIUrl":"https://doi.org/10.1109/ICEIC61013.2024.10457116","url":null,"abstract":"This paper presents a 32-Gb/s full-rate clock and data recovery (CDR) architecture based on a phase interpolator (PI), which incorporates high-speed channel equalization to widen the ppm tolerance or locking range. This work focuses on the receiver-side implementation. Therefore, the feed-forward equalizer (FFE) tap is not utilized, and only the voltage swing level is adjusted at the transmitter side. The channel is modeled using Verilog language with a -10 dB loss at 10 GHz. The overall architecture comprises several components: a continuous-time linear equalizer (CTLE), 1-tap decision feedback equalizer (DFE), 7-bit PI, digital loop filter, and 2x oversampling phase detector. By individually employing the DFE and CTLE, the optimal tap coefficient value for the DFE, which produces the widest eye pattern, and the pole and zero positions of the CTLE are determined. Finally, CTLE and 1-tap DFE ensure optimal vertical 322 mV and timing margin of 25.8 ps. It also relaxes phase modulation to obtain acceptable error of the phase interpolator up to ±15625 ppm.","PeriodicalId":518726,"journal":{"name":"2024 International Conference on Electronics, Information, and Communication (ICEIC)","volume":"155 6","pages":"1-4"},"PeriodicalIF":0.0,"publicationDate":"2024-01-28","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"140530084","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2024-01-28DOI: 10.1109/ICEIC61013.2024.10457255
Ga-Young Kim, Su-Hong Eom, Eunghyuk Lee, Jeon-Min Kang
This study proposes a following method based on ultrasonic sensors in consideration of the situation that requires collaboration while performing autonomous driving-based service robot tasks. Since it is impossible to follow the target when it is outside the transmission and reception range of ultrasonic sensors, this study proposes a method of extending the range by changing the angle of the ultrasonic sensor. It was confirmed that the proposed method can extend the transmission and reception range in comparing the transmission and reception range with the basic sensor installation method in which ultrasonic sensors are installed facing the front with each other.
{"title":"A Method of Extending the Transmission and Reception Range of Ultrasonic Sensors for Stable Following in a Narrow Indoor Space","authors":"Ga-Young Kim, Su-Hong Eom, Eunghyuk Lee, Jeon-Min Kang","doi":"10.1109/ICEIC61013.2024.10457255","DOIUrl":"https://doi.org/10.1109/ICEIC61013.2024.10457255","url":null,"abstract":"This study proposes a following method based on ultrasonic sensors in consideration of the situation that requires collaboration while performing autonomous driving-based service robot tasks. Since it is impossible to follow the target when it is outside the transmission and reception range of ultrasonic sensors, this study proposes a method of extending the range by changing the angle of the ultrasonic sensor. It was confirmed that the proposed method can extend the transmission and reception range in comparing the transmission and reception range with the basic sensor installation method in which ultrasonic sensors are installed facing the front with each other.","PeriodicalId":518726,"journal":{"name":"2024 International Conference on Electronics, Information, and Communication (ICEIC)","volume":"151 2","pages":"1-4"},"PeriodicalIF":0.0,"publicationDate":"2024-01-28","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"140530087","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2024-01-28DOI: 10.1109/ICEIC61013.2024.10457155
Donghyeon Lim, Changhoon Yim
The use of meta-learning has been proven efficient to address the limitations of insufficient data for no-reference image quality assessment (NR-IQA). While meta-learning methods have been developed as training process, the works for appropriate network models were not sufficient, which posed limitations on performance improvement. The goal of this work is to design a suitable network model for meta-learning to enhance NR-IQA performance. The proposed method follows the training process of optimization-based meta-learning for each distortion type. The proposed network model learns efficiently distortion-specific features and adapts easily to unknown distortions. Experimental results show that the proposed network model provides superior performance than the previous NR-IQA methods using meta-learning.
{"title":"An Effective Meta-Learning Network Model for No-Reference Image Quality Assessment","authors":"Donghyeon Lim, Changhoon Yim","doi":"10.1109/ICEIC61013.2024.10457155","DOIUrl":"https://doi.org/10.1109/ICEIC61013.2024.10457155","url":null,"abstract":"The use of meta-learning has been proven efficient to address the limitations of insufficient data for no-reference image quality assessment (NR-IQA). While meta-learning methods have been developed as training process, the works for appropriate network models were not sufficient, which posed limitations on performance improvement. The goal of this work is to design a suitable network model for meta-learning to enhance NR-IQA performance. The proposed method follows the training process of optimization-based meta-learning for each distortion type. The proposed network model learns efficiently distortion-specific features and adapts easily to unknown distortions. Experimental results show that the proposed network model provides superior performance than the previous NR-IQA methods using meta-learning.","PeriodicalId":518726,"journal":{"name":"2024 International Conference on Electronics, Information, and Communication (ICEIC)","volume":"295 1","pages":"1-4"},"PeriodicalIF":0.0,"publicationDate":"2024-01-28","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"140530046","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2024-01-28DOI: 10.1109/ICEIC61013.2024.10457099
Minseok Kang, Minhyeok Lee, Sangyoun Lee
Semantic segmentation, a fundamental task in computer vision, has evolved significantly with the introduction of deep learning techniques, particularly fully convolutional networks (FCNs). In the context of real-time semantic segmentation, the demand for efficient yet accurate models has grown, particularly for resource-constrained devices. Recent advancements have explored the fusion of global context and local details through bidirectional network structures, exemplified by BiSeNet, STDCNet, and DDRNet. However, issues of pixel inconsistency within the same object classification persist. Attention-based models like SETR and SegFormer have shown promise in mitigating this issue by capturing intricate spatial dependencies. This paper introduces the concept of ‘Mask Disarrange’ and proposes a lightweight attention mechanism suitable for real-time semantic segmentation. The Cross Patch Attention (CPA) and Inter Patch Attention (IPA) methods are presented, addressing fusion and mask disarrange challenges while maintaining computational efficiency. Experimental results on the Cityscapes dataset demonstrate the effectiveness of the proposed Bilateral Patch-Net (BPNet) in achieving superior segmentation performance and increased frames per second (FPS) compared to the state-of-the-art PIDNet. BPNet's contributions lie in its simplicity, efficiency, and applicability to diverse domains, offering potential for broader adoption in computer vision applications.
{"title":"Real-time Semantic Segmentation with Bilateral Patch Attention","authors":"Minseok Kang, Minhyeok Lee, Sangyoun Lee","doi":"10.1109/ICEIC61013.2024.10457099","DOIUrl":"https://doi.org/10.1109/ICEIC61013.2024.10457099","url":null,"abstract":"Semantic segmentation, a fundamental task in computer vision, has evolved significantly with the introduction of deep learning techniques, particularly fully convolutional networks (FCNs). In the context of real-time semantic segmentation, the demand for efficient yet accurate models has grown, particularly for resource-constrained devices. Recent advancements have explored the fusion of global context and local details through bidirectional network structures, exemplified by BiSeNet, STDCNet, and DDRNet. However, issues of pixel inconsistency within the same object classification persist. Attention-based models like SETR and SegFormer have shown promise in mitigating this issue by capturing intricate spatial dependencies. This paper introduces the concept of ‘Mask Disarrange’ and proposes a lightweight attention mechanism suitable for real-time semantic segmentation. The Cross Patch Attention (CPA) and Inter Patch Attention (IPA) methods are presented, addressing fusion and mask disarrange challenges while maintaining computational efficiency. Experimental results on the Cityscapes dataset demonstrate the effectiveness of the proposed Bilateral Patch-Net (BPNet) in achieving superior segmentation performance and increased frames per second (FPS) compared to the state-of-the-art PIDNet. BPNet's contributions lie in its simplicity, efficiency, and applicability to diverse domains, offering potential for broader adoption in computer vision applications.","PeriodicalId":518726,"journal":{"name":"2024 International Conference on Electronics, Information, and Communication (ICEIC)","volume":"230 4","pages":"1-4"},"PeriodicalIF":0.0,"publicationDate":"2024-01-28","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"140530060","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2024-01-28DOI: 10.1109/ICEIC61013.2024.10457113
Pornnapa Panyadee, P. Champrasert
The flood early warning system can help mitigate the resulting damages by predicting future events. This is achieved through the utilization of data obtained from telemetry stations to predict the values of water levels in the future. Flood hazard maps are considered a tool for representing the potential flood events that occur in an area. This paper proposed a framework to apply the spatial and temporal data to generate flood hazard mapping using the integration of interpolation telemetry station data and a temporal prediction model. The framework consists of two components: 1) the temporal prediction model is applied to water level prediction on hourly and daily scales, and 2) the interpolation of spatial data to generate a flood hazard map. The evaluation results show that the hourly and daily temporal prediction models can predict the water level with an average of MAPE using 500 iterations are 3.17% and 4.88% of training, and 3.48% and 4.72% of testing. Then, the flood hazard map is generated. The accuracy is 70.90% and F1-score is 81.50% compared to the observation flood event.
{"title":"Spatial-Temporal Flood Hazard Mapping Using Integration of Telemetry Data and Prediction Model","authors":"Pornnapa Panyadee, P. Champrasert","doi":"10.1109/ICEIC61013.2024.10457113","DOIUrl":"https://doi.org/10.1109/ICEIC61013.2024.10457113","url":null,"abstract":"The flood early warning system can help mitigate the resulting damages by predicting future events. This is achieved through the utilization of data obtained from telemetry stations to predict the values of water levels in the future. Flood hazard maps are considered a tool for representing the potential flood events that occur in an area. This paper proposed a framework to apply the spatial and temporal data to generate flood hazard mapping using the integration of interpolation telemetry station data and a temporal prediction model. The framework consists of two components: 1) the temporal prediction model is applied to water level prediction on hourly and daily scales, and 2) the interpolation of spatial data to generate a flood hazard map. The evaluation results show that the hourly and daily temporal prediction models can predict the water level with an average of MAPE using 500 iterations are 3.17% and 4.88% of training, and 3.48% and 4.72% of testing. Then, the flood hazard map is generated. The accuracy is 70.90% and F1-score is 81.50% compared to the observation flood event.","PeriodicalId":518726,"journal":{"name":"2024 International Conference on Electronics, Information, and Communication (ICEIC)","volume":"188 1","pages":"1-4"},"PeriodicalIF":0.0,"publicationDate":"2024-01-28","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"140530248","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2024-01-28DOI: 10.1109/ICEIC61013.2024.10457089
Zihui Wang, Xueqin Jiang, Jinming Yu, Miaowen Wen, Jun Li, Han Hai
Dual-mode generalized spatial modulation (DM-GSM) enhances spectral efficiency in GSM systems using two modes across transmit antennas. However, interference between antennas poses a challenge for signal detection. For this, a deep learning detector, the dual-mode deep neural network (DM-DNN), is proposed. The DM-DNN enables simultaneous detection of the antenna mode and modulation symbol through its network structure and label generation. A loss function is proposed to train the DM-DNN, approximating optimal bit error rate (BER) performance. Simulation results demonstrate that the DM-DNN achieves BER performance close to the maximum likelihood (ML) detector while significantly reducing complexity.
{"title":"Efficient Dual-Mode Generalized Spatial Modulation Detection with Enhanced DNN Architecture","authors":"Zihui Wang, Xueqin Jiang, Jinming Yu, Miaowen Wen, Jun Li, Han Hai","doi":"10.1109/ICEIC61013.2024.10457089","DOIUrl":"https://doi.org/10.1109/ICEIC61013.2024.10457089","url":null,"abstract":"Dual-mode generalized spatial modulation (DM-GSM) enhances spectral efficiency in GSM systems using two modes across transmit antennas. However, interference between antennas poses a challenge for signal detection. For this, a deep learning detector, the dual-mode deep neural network (DM-DNN), is proposed. The DM-DNN enables simultaneous detection of the antenna mode and modulation symbol through its network structure and label generation. A loss function is proposed to train the DM-DNN, approximating optimal bit error rate (BER) performance. Simulation results demonstrate that the DM-DNN achieves BER performance close to the maximum likelihood (ML) detector while significantly reducing complexity.","PeriodicalId":518726,"journal":{"name":"2024 International Conference on Electronics, Information, and Communication (ICEIC)","volume":"12 1","pages":"1-4"},"PeriodicalIF":0.0,"publicationDate":"2024-01-28","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"140530207","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
GPS uses ECCs to see if an error occurs when the data sent from the satellite reaches the user. Each message structure uses ECCs such as Hamming Code, CRC, BCH Code, and LDPC Code. If the satellite contains all of the encoders, it has a negative impact to the area and power consumption. Therefore, in this paper, we propose a CRC-BCH unified encoder for GPS, which is efficient in terms of space and power consumption. Since both the CRC and BCH encoders use shift registers, the design was made using this part. To replace the existing encoder, the CRC-BCH encoder must have the same output. To validate this, we used individual CRC and BCH encoders and confirmed that the generated output was identical to the output of the proposed encoder. The proposed CRC-BCH unified encoder was synthesized at an operating frequency of 400 MHz using the CMOS 28nm process. The synthesis results showed that it used 16.67% less area and consumed 19.68% less power than the existing encoder. Therefore, the proposed CRC-BCH unified encoder offers advantages in terms of satellite weight and energy efficiency.
{"title":"Efficient CRC-BCH Unified Encoder for Global Positioning System","authors":"Yongtaek Hwang, Jiwoo Hwang, Yuseok Lee, Hoyoung Yoo","doi":"10.1109/ICEIC61013.2024.10457230","DOIUrl":"https://doi.org/10.1109/ICEIC61013.2024.10457230","url":null,"abstract":"GPS uses ECCs to see if an error occurs when the data sent from the satellite reaches the user. Each message structure uses ECCs such as Hamming Code, CRC, BCH Code, and LDPC Code. If the satellite contains all of the encoders, it has a negative impact to the area and power consumption. Therefore, in this paper, we propose a CRC-BCH unified encoder for GPS, which is efficient in terms of space and power consumption. Since both the CRC and BCH encoders use shift registers, the design was made using this part. To replace the existing encoder, the CRC-BCH encoder must have the same output. To validate this, we used individual CRC and BCH encoders and confirmed that the generated output was identical to the output of the proposed encoder. The proposed CRC-BCH unified encoder was synthesized at an operating frequency of 400 MHz using the CMOS 28nm process. The synthesis results showed that it used 16.67% less area and consumed 19.68% less power than the existing encoder. Therefore, the proposed CRC-BCH unified encoder offers advantages in terms of satellite weight and energy efficiency.","PeriodicalId":518726,"journal":{"name":"2024 International Conference on Electronics, Information, and Communication (ICEIC)","volume":"322 5","pages":"1-3"},"PeriodicalIF":0.0,"publicationDate":"2024-01-28","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"140530211","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2024-01-28DOI: 10.1109/ICEIC61013.2024.10457206
Eunsu Jang, Su-Hong Eom, D. Kim, Eunghyuk Lee
As the number of people who need electric wheelchairs increases around the world, there is a lot of demand, but there are many people who cannot use them because they are difficult to operate. Accordingly, the introduction of autonomous driving technology is being studied for wheelchairs that do not require separate manipulation. Currently, such autonomous driving is based on Map, but it is difficult to have Map in all environments. Thus, it is necessary to develop driving assistance technology in a Mapless environment. This study focuses on the position estimation system by fusion of UWB/Encoder/IMU for the development of driving assistance systems in the Mapless environment of wheelchairs.
{"title":"A Study on the UWB/Encoder/IMU Sensor Fusion Position Estimation System for the Development of Driving Assistance Technology in Autonomous Driving Wheelchairs","authors":"Eunsu Jang, Su-Hong Eom, D. Kim, Eunghyuk Lee","doi":"10.1109/ICEIC61013.2024.10457206","DOIUrl":"https://doi.org/10.1109/ICEIC61013.2024.10457206","url":null,"abstract":"As the number of people who need electric wheelchairs increases around the world, there is a lot of demand, but there are many people who cannot use them because they are difficult to operate. Accordingly, the introduction of autonomous driving technology is being studied for wheelchairs that do not require separate manipulation. Currently, such autonomous driving is based on Map, but it is difficult to have Map in all environments. Thus, it is necessary to develop driving assistance technology in a Mapless environment. This study focuses on the position estimation system by fusion of UWB/Encoder/IMU for the development of driving assistance systems in the Mapless environment of wheelchairs.","PeriodicalId":518726,"journal":{"name":"2024 International Conference on Electronics, Information, and Communication (ICEIC)","volume":"408 2","pages":"1-4"},"PeriodicalIF":0.0,"publicationDate":"2024-01-28","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"140530016","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2024-01-28DOI: 10.1109/ICEIC61013.2024.10457111
Youngdeok Hwang, Janghwan Lee, Jiwoong Park, Jieun Lim, Jungwook Choi
Large Language Models (LLMs) have shown remarkable success in various natural language processing tasks. However, their extensive parameter count leads to significant memory and computational demands. To tackle these challenges, there is growing interest in employing post-training quantization (PTQ) with reduced-precision floating-point (FP) operations. Yet, the optimal FP configuration remains a topic of debate. Existing studies often overlook a thorough analysis of the diverse data distributions found in LLMs and the crucial design choice, denormal. In this paper, we conduct a comprehensive examination of the various data distributions within LLMs and the significance of denormal representation, presenting a mixed-format floating-point framework. Our proposed framework allows for sub-8-bit inference with minimal performance degradation in language modeling and reasoning tasks across a broad spectrum of LLMs.
{"title":"Searching Optimal Floating-Point Format for Sub-8-Bit Large Language Model Inference","authors":"Youngdeok Hwang, Janghwan Lee, Jiwoong Park, Jieun Lim, Jungwook Choi","doi":"10.1109/ICEIC61013.2024.10457111","DOIUrl":"https://doi.org/10.1109/ICEIC61013.2024.10457111","url":null,"abstract":"Large Language Models (LLMs) have shown remarkable success in various natural language processing tasks. However, their extensive parameter count leads to significant memory and computational demands. To tackle these challenges, there is growing interest in employing post-training quantization (PTQ) with reduced-precision floating-point (FP) operations. Yet, the optimal FP configuration remains a topic of debate. Existing studies often overlook a thorough analysis of the diverse data distributions found in LLMs and the crucial design choice, denormal. In this paper, we conduct a comprehensive examination of the various data distributions within LLMs and the significance of denormal representation, presenting a mixed-format floating-point framework. Our proposed framework allows for sub-8-bit inference with minimal performance degradation in language modeling and reasoning tasks across a broad spectrum of LLMs.","PeriodicalId":518726,"journal":{"name":"2024 International Conference on Electronics, Information, and Communication (ICEIC)","volume":"40 1","pages":"1-4"},"PeriodicalIF":0.0,"publicationDate":"2024-01-28","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"140530261","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2024-01-28DOI: 10.1109/ICEIC61013.2024.10457163
Inseong Hwang, Jihoon Jang, Hyun Kim
The emulation or layout in the study of processing-in-memory (PIM) is a highly time-consuming process. Especially, the processing-using-memory (PUM), a subset of PIM, is much more complex due to the positioning of the processing unit in the high-density data array. Because of this reason, it is important to efficiently verify PIM hardware using simulation to activate the PIM study. To this end, we modify the DRAMsim3, a memory simulator, to implement a PUM system, and propose a PIM operation compiler in the Zsim, a CPU simulator. The PIM operation compiler performs the role of tracing instructions from various precision deep neural network (DNN) workloads and generating PIM operation commands. Finally, we propose an architecture-level PUM simulation framework that can simulate the PUM system with DNN workloads based on the PIM command generated by the compiler.
{"title":"An Architecture-Level Framework for Enabling Processing-Using-Memory Simulations in Deep Neural Networks","authors":"Inseong Hwang, Jihoon Jang, Hyun Kim","doi":"10.1109/ICEIC61013.2024.10457163","DOIUrl":"https://doi.org/10.1109/ICEIC61013.2024.10457163","url":null,"abstract":"The emulation or layout in the study of processing-in-memory (PIM) is a highly time-consuming process. Especially, the processing-using-memory (PUM), a subset of PIM, is much more complex due to the positioning of the processing unit in the high-density data array. Because of this reason, it is important to efficiently verify PIM hardware using simulation to activate the PIM study. To this end, we modify the DRAMsim3, a memory simulator, to implement a PUM system, and propose a PIM operation compiler in the Zsim, a CPU simulator. The PIM operation compiler performs the role of tracing instructions from various precision deep neural network (DNN) workloads and generating PIM operation commands. Finally, we propose an architecture-level PUM simulation framework that can simulate the PUM system with DNN workloads based on the PIM command generated by the compiler.","PeriodicalId":518726,"journal":{"name":"2024 International Conference on Electronics, Information, and Communication (ICEIC)","volume":"364 6","pages":"1-3"},"PeriodicalIF":0.0,"publicationDate":"2024-01-28","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"140530459","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
京公网安备 11010802042870号
京ICP备2023020795号-1
扫码关注我们
求助内容：
应助结果提醒方式：