Myung Kuk Yoon
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Previous Organizations / Institutes
The organizations / institutes from the scholar's other articles in the past ten years
Research Subjects
Myung Kuk Yoon has a total of 45 co-authors. Here are the top co-authors by number of shared publications:
- Yunho Oh
School of Electrical Engineering, Korea University, Seoul, South Korea
14 shared publications - Gunjae Koo
Korea University, Seoul, Republic of Korea
11 shared publications - Jaebeom Jeon
Korea University, Seoul, Republic of Korea
5 shared publications - Junsu Kim
Korea University, Seoul, Republic of Korea
4 shared publications - Sangun Choi
School of Electrical Engineering, Korea University, Seoul, South Korea
4 shared publications - Won Woo Ro
Yonsei University
4 shared publications
Latest Publications
- Open Access
- 20 April 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 14, 61010-61024
Restructuring the Implicit GEMM Workflow for Complex-Valued Convolution
Complex-valued neural networks (CV-CNNs) are valuable in signal-centric domains because they preserve magnitude–phase structure that can be lost in purely real-valued representations. However, this benefit does not directly translate to efficient GPU inference. In common implementations, complex-valued convolution is typically decomposed into four real-valued terms, each executed via implicit GEMM. This fragmented execution incurs workflow (structural) overheads from repeated tensor preparation, output reconstruction, and fine-grained kernel launches beyond the core GEMM computation, increasing redundant memory movement and kernel-launch frequency and accounting for a substantial fraction of the end-to-end runtime of ComplexConv2d. This paper revisits the implicit GEMM execution structure of ComplexConv2d and proposes threeCV-CNN specialized restructurings that optimize the implicitGEMM workflow. First, the epilogue-fused design integrates recurring output reconstruction postprocessing into the implicit GEMM epilogue to reduce auxiliary kernels and intermediate traffic. Second, the operation-reordered design reorders the four real-valued convolution terms to increase reuse of already materialized real and imaginary input tensors within the implicit GEMM workflow, reducing repeated tensor preparation. Third, the unified design implements an implicit CGEMM kernel that directly consumes real and imaginary components through an integrated load path and performs complex accumulation within a consolidated execution flow, thereby substantially reducing these overheads. We evaluate the proposed designs across four representative CV-CNN backbones (ResNet18, ResNet50, VGG16, and VGG19). On average, the epilogue-fused, operation-reordered, and unified designs reduce the normalized end-to-end runtime by 7.6%, 10.1%, and 27.8% relative to baseline, respectively, with the unified design consistently achieving the largest improvement across the evaluated models.
- 20 April 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 14, 61010-61024
Open AccessRestructuring the Implicit GEMM Workflow for Complex-Valued Convolution
Complex-valued neural networks (CV-CNNs) are valuable in signal-centric domains because they preserve magnitude–phase structure that can be lost in purely real-valued representations. However, this benefit does not directly translate to efficient GPU inference. In common implementations, complex-valued convolution is typically decomposed into four real-valued terms, each executed via implicit GEMM. This fragmented execution incurs workflow (structural) overheads from repeated tensor preparation, output reconstruction, and fine-grained kernel launches beyond the core GEMM computation, increasing redundant memory movement and kernel-launch frequency and accounting for a substantial fraction of the end-to-end runtime of ComplexConv2d. This paper revisits the implicit GEMM execution structure of ComplexConv2d and proposes threeCV-CNN specialized restructurings that optimize the implicitGEMM workflow. First, the epilogue-fused design integrates recurring output reconstruction postprocessing into the implicit GEMM epilogue to reduce auxiliary kernels and intermediate traffic. Second, the operation-reordered design reorders the four real-valued convolution terms to increase reuse of already materialized real and imaginary input tensors within the implicit GEMM workflow, reducing repeated tensor preparation. Third, the unified design implements an implicit CGEMM kernel that directly consumes real and imaginary components through an integrated load path and performs complex accumulation within a consolidated execution flow, thereby substantially reducing these overheads. We evaluate the proposed designs across four representative CV-CNN backbones (ResNet18, ResNet50, VGG16, and VGG19). On average, the epilogue-fused, operation-reordered, and unified designs reduce the normalized end-to-end runtime by 7.6%, 10.1%, and 27.8% relative to baseline, respectively, with the unified design consistently achieving the largest improvement across the evaluated models.
- Open Access
- 1 January 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- p. 1
BiKD: Bidirectional Kernel Decomposition for Large-Scale GCNs on GPU
- 1 January 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- p. 1
Open AccessBiKD: Bidirectional Kernel Decomposition for Large-Scale GCNs on GPU
- Open Access
- 3 November 2025
- Published by Association for Computing Machinery (ACM) in ACM Transactions on Storage
- Vol. 21 (4) , 1-26
TM-Training: An Energy-Efficient Tiered Memory System for Deep Learning Training in NPUs
DRAM accounts for a large fraction of the total cost of ownership of memory systems in deep learning acceleration systems. To achieve sustainable scalability, tiered memory systems with denser technologies become critical. Prior work has proposed various tiered memory systems, but no matter how a system is designed, data movements between tiers consume substantial energy. In particular, as model sizes and memory capacity demands grow, the data movements between memory tiers become more frequent, posing a challenge that tiered memory systems may reduce deployment costs but suffer from low energy efficiency. If a memory system proactively places data into a tier and timely fetches it from the tier, then the excessive data movement between tiers can be mitigated. We find that a system can statically anticipate such behaviors for all pages and localities. With this insight, we propose a new DNN acceleration system called TM-Training using flash memory. TM-Training capitalizes on the repetitive nature of the same computational patterns during execution, enabling the static establishment of optimal data placement for subsequent operations. Moreover, TM-Training employs a new data-splitting scheme to enable precise memory management. Our evaluation demonstrates that TM-Training reduces inter-tier data traffic by 64% and achieves a 55% higher throughput per watt in training than prior work.
- 3 November 2025
- Published by Association for Computing Machinery (ACM) in ACM Transactions on Storage
- Vol. 21 (4) , 1-26
Open AccessTM-Training: An Energy-Efficient Tiered Memory System for Deep Learning Training in NPUs
DRAM accounts for a large fraction of the total cost of ownership of memory systems in deep learning acceleration systems. To achieve sustainable scalability, tiered memory systems with denser technologies become critical. Prior work has proposed various tiered memory systems, but no matter how a system is designed, data movements between tiers consume substantial energy. In particular, as model sizes and memory capacity demands grow, the data movements between memory tiers become more frequent, posing a challenge that tiered memory systems may reduce deployment costs but suffer from low energy efficiency. If a memory system proactively places data into a tier and timely fetches it from the tier, then the excessive data movement between tiers can be mitigated. We find that a system can statically anticipate such behaviors for all pages and localities. With this insight, we propose a new DNN acceleration system called TM-Training using flash memory. TM-Training capitalizes on the repetitive nature of the same computational patterns during execution, enabling the static establishment of optimal data placement for subsequent operations. Moreover, TM-Training employs a new data-splitting scheme to enable precise memory management. Our evaluation demonstrates that TM-Training reduces inter-tier data traffic by 64% and achieves a 55% higher throughput per watt in training than prior work.
- Open Access
- 20 June 2025
- Published by Association for Computing Machinery (ACM)
- p. 153-165
Avant-Garde: Empowering GPUs with Scaled Numeric Formats
- 20 June 2025
- Published by Association for Computing Machinery (ACM)
- p. 153-165
Open AccessAvant-Garde: Empowering GPUs with Scaled Numeric Formats
- 17 June 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Computer Architecture Letters
- Vol. 24 (2) , 213-216
MOST: Memory Oversubscription-Aware Scheduling for Tensor Migration on GPU Unified Storage
Deep Neural Network (DNN) training demands large memory capacities that exceed the limits of current GPU onboard memory. Expanding GPU memory with SSDs is a cost-effective approach. However, the low bandwidth of SSDs introduces severe performance bottlenecks in data management, particularly for Unified Virtual Memory (UVM)-based systems. The default on-demand migration mechanism in UVM causes frequent page faults and stalls, exacerbated by memory oversubscription and eviction processes along the critical path. To address these challenges, this paper proposes Memory Oversubscription-aware Scheduling for Tensor Migration (MOST), a software framework designed to improve data migration in UVM environments. MOST profiles memory access behavior and quantifies the impact of memory oversubscription stalls and schedules tensor migrations to minimize overall training time. With the profiling results, MOST executes newly designed pre-eviction and prefetching instructions within DNN kernel code. MOST effectively selects and migrates tensors that can mitigate memory oversubscription stalls, thus reducing training time. Our evaluation shows that MOST achieves an average speedup of 22.9% and 12.8% over state-of-the-art techniques, DeepUM and G10, respectively.
- 17 June 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Computer Architecture Letters
- Vol. 24 (2) , 213-216
MOST: Memory Oversubscription-Aware Scheduling for Tensor Migration on GPU Unified Storage
Deep Neural Network (DNN) training demands large memory capacities that exceed the limits of current GPU onboard memory. Expanding GPU memory with SSDs is a cost-effective approach. However, the low bandwidth of SSDs introduces severe performance bottlenecks in data management, particularly for Unified Virtual Memory (UVM)-based systems. The default on-demand migration mechanism in UVM causes frequent page faults and stalls, exacerbated by memory oversubscription and eviction processes along the critical path. To address these challenges, this paper proposes Memory Oversubscription-aware Scheduling for Tensor Migration (MOST), a software framework designed to improve data migration in UVM environments. MOST profiles memory access behavior and quantifies the impact of memory oversubscription stalls and schedules tensor migrations to minimize overall training time. With the profiling results, MOST executes newly designed pre-eviction and prefetching instructions within DNN kernel code. MOST effectively selects and migrates tensors that can mitigate memory oversubscription stalls, thus reducing training time. Our evaluation shows that MOST achieves an average speedup of 22.9% and 12.8% over state-of-the-art techniques, DeepUM and G10, respectively.
- Open Access
- 13 June 2025
- Published by Association for Computing Machinery (ACM)
- p. 85-96
SSFFT: Energy-Efficient Selective Scaling for Fast Fourier Transform in Embedded GPUs
- 13 June 2025
- Published by Association for Computing Machinery (ACM)
- p. 85-96
Open AccessSSFFT: Energy-Efficient Selective Scaling for Fast Fourier Transform in Embedded GPUs
- 4 March 2025
- Published by Elsevier in Journal of Systems Architecture
- Vol. 162
Beyond VABlock: Improving Transformer workloads through aggressive prefetching
- 4 March 2025
- Published by Elsevier in Journal of Systems Architecture
- Vol. 162
Beyond VABlock: Improving Transformer workloads through aggressive prefetching
- 14 November 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Embedded Systems Letters
- Vol. 17 (3) , 180-183
TLP Balancer: Predictive Thread Allocation for Multitenant Inference in Embedded GPUs
This paper introduces a novel software technique to optimize thread allocation for merged and fused kernels in multi-tenant inference systems on embedded Graphics Processing Units (GPUs). Embedded systems equipped with GPUs face challenges in managing diverse deep learning workloads while adhering to Quality-of-Service (QoS) standards, primarily due to limited hardware resources and the varied nature of deep learning models. Prior work has relied on static thread allocation strategies, often leading to suboptimal hardware utilization. To address these challenges, we propose a new software technique called TLP Balancer. TLP Balancer automatically identifies the best-performing number of threads based on performance modeling. This approach significantly enhances hardware utilization and ensures QoS compliance, outperforming traditional fixed-thread allocation methods. Our evaluation shows that TLP Balancer improves throughput by 40% compared to the state-of-the-art automated kernel merge and fusion techniques.
- 14 November 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Embedded Systems Letters
- Vol. 17 (3) , 180-183
TLP Balancer: Predictive Thread Allocation for Multitenant Inference in Embedded GPUs
This paper introduces a novel software technique to optimize thread allocation for merged and fused kernels in multi-tenant inference systems on embedded Graphics Processing Units (GPUs). Embedded systems equipped with GPUs face challenges in managing diverse deep learning workloads while adhering to Quality-of-Service (QoS) standards, primarily due to limited hardware resources and the varied nature of deep learning models. Prior work has relied on static thread allocation strategies, often leading to suboptimal hardware utilization. To address these challenges, we propose a new software technique called TLP Balancer. TLP Balancer automatically identifies the best-performing number of threads based on performance modeling. This approach significantly enhances hardware utilization and ensures QoS compliance, outperforming traditional fixed-thread allocation methods. Our evaluation shows that TLP Balancer improves throughput by 40% compared to the state-of-the-art automated kernel merge and fusion techniques.
- Open Access
- 12 August 2024
- Published by Association for Computing Machinery (ACM)
- p. 1012-1021
VitBit: Enhancing Embedded GPU Performance for AI Workloads through Register Operand Packing
The rapid advancement of Artificial Intelligence (AI) necessitates significant enhancements in the energy efficiency of Graphics Processing Units (GPUs) for Deep Neural Network (DNN) workloads. Such a challenge is particularly critical for embedded GPUs, which operate within stringent power constraints. Traditional GPU architectures, designed to support a limited set of numeric formats, face challenges in meeting the diverse requirements of modern AI applications. These applications demand support for various numeric formats to optimize computational speed and efficiency. This paper proposes VitBit, a novel software technique designed to overcome these limitations by enabling efficient processing of arbitrary integer format values, especially those 8 bits or fewer, which are increasingly prevalent in AI workloads. VitBit introduces two key innovations: the packing of arbitrary integer formats for parallel computation and the simultaneous execution of Tensor cores, INT and FP (Integer and Floating-Point) CUDA cores. This approach leverages the architectural features of modern GPUs, such as those based on NVIDIA Ampere architecture, which allows concurrent operation of FP32 and INT32 cores at full throughput. Our evaluation of VitBit on NVIDIA Jetson AGX Orin demonstrates substantial improvements in arithmetic density and peak throughput, achieving up to a 22% reduction in execution time for benchmark AI workloads without compromising inference accuracy. VitBit effectively bridges the gap between current hardware capabilities and the computational demands of AI, offering a scalable and cost-effective method for enhancing GPU performance in AI applications.
- 12 August 2024
- Published by Association for Computing Machinery (ACM)
- p. 1012-1021
Open AccessVitBit: Enhancing Embedded GPU Performance for AI Workloads through Register Operand Packing
The rapid advancement of Artificial Intelligence (AI) necessitates significant enhancements in the energy efficiency of Graphics Processing Units (GPUs) for Deep Neural Network (DNN) workloads. Such a challenge is particularly critical for embedded GPUs, which operate within stringent power constraints. Traditional GPU architectures, designed to support a limited set of numeric formats, face challenges in meeting the diverse requirements of modern AI applications. These applications demand support for various numeric formats to optimize computational speed and efficiency. This paper proposes VitBit, a novel software technique designed to overcome these limitations by enabling efficient processing of arbitrary integer format values, especially those 8 bits or fewer, which are increasingly prevalent in AI workloads. VitBit introduces two key innovations: the packing of arbitrary integer formats for parallel computation and the simultaneous execution of Tensor cores, INT and FP (Integer and Floating-Point) CUDA cores. This approach leverages the architectural features of modern GPUs, such as those based on NVIDIA Ampere architecture, which allows concurrent operation of FP32 and INT32 cores at full throughput. Our evaluation of VitBit on NVIDIA Jetson AGX Orin demonstrates substantial improvements in arithmetic density and peak throughput, achieving up to a 22% reduction in execution time for benchmark AI workloads without compromising inference accuracy. VitBit effectively bridges the gap between current hardware capabilities and the computational demands of AI, offering a scalable and cost-effective method for enhancing GPU performance in AI applications.