Han Zhao
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
Han Zhao has a total of 49 co-authors. Here are the top co-authors by number of shared publications:
- Minyi Guo
Shanghai Jiao Tong University, Shanghai, China
11 shared publications - Quan Chen
Shanghai Jiao Tong University
10 shared publications - Weihao Cui
Shanghai Jiao Tong University, Shanghai, China
9 shared publications - Jingwen Leng
Shanghai Jiao Tong University, Shanghai, China
4 shared publications - Shulai Zhang
Shanghai Jiao Tong University, Shanghai, China
4 shared publications - Chen Chen
Shanghai Jiao Tong University, Shanghai, China
3 shared publications
Latest Publications
- Open Access
- 26 April 2026
- Published by Association for Computing Machinery (ACM)
- p. 89-107
Arena: Efficiently Training Large Models via Dynamic Scheduling and Adaptive Parallelism Co-Design
- 26 April 2026
- Published by Association for Computing Machinery (ACM)
- p. 89-107
Open AccessArena: Efficiently Training Large Models via Dynamic Scheduling and Adaptive Parallelism Co-Design
- Open Access
- 22 March 2026
- Published by Association for Computing Machinery (ACM)
- p. 2030-2047
Towards High-Goodput LLM Serving with Prefill-decode Multiplexing
- 22 March 2026
- Published by Association for Computing Machinery (ACM)
- p. 2030-2047
Open AccessTowards High-Goodput LLM Serving with Prefill-decode Multiplexing
- Open Access
- 30 June 2025
- Published by Association for Computing Machinery (ACM) in ACM Transactions on Architecture and Code Optimization
- Vol. 22 (2) , 1-26
ARACHNE: Optimizing Distributed Parallel Applications with Reduced Inter-Process Communication
In high-performance computing (HPC), parallelization is essential for improving computational efficiency as data and computation scales exceed single-node capacity. Existing methods, such as the polyhedral model used in Pluto -Distmem, focus on loop and array optimizations within shared memory but struggle with high communication overheads and inflexibility in distributed environments. These methods often fail to effectively partition computation and manage data across nodes, leading to suboptimal performance. This paper presents Arachne , an innovative system designed to address these shortcomings by generating distributed parallel code with minimized communication overhead. The system introduces a dynamic programming algorithm to optimally distribute computational tasks across multiple processes, ensuring minimal communication costs. It also incorporates user-friendly compiler directives, allowing programmers to influence code generation easily and accommodate a broader range of parallelization scenarios without needing in-depth knowledge of parallel architectures. Arachne significantly reduces the learning curve and need for extensive code modifications, making parallel programming more accessible and efficient. Evaluation of various HPC benchmarks demonstrates that Arachne outperforms existing methods by reducing communication overhead, lowering memory requirements, and supporting more complex parallel logic, thus enhancing the overall scalability and efficiency of HPC applications.
- 30 June 2025
- Published by Association for Computing Machinery (ACM) in ACM Transactions on Architecture and Code Optimization
- Vol. 22 (2) , 1-26
Open AccessARACHNE: Optimizing Distributed Parallel Applications with Reduced Inter-Process Communication
In high-performance computing (HPC), parallelization is essential for improving computational efficiency as data and computation scales exceed single-node capacity. Existing methods, such as the polyhedral model used in Pluto -Distmem, focus on loop and array optimizations within shared memory but struggle with high communication overheads and inflexibility in distributed environments. These methods often fail to effectively partition computation and manage data across nodes, leading to suboptimal performance. This paper presents Arachne , an innovative system designed to address these shortcomings by generating distributed parallel code with minimized communication overhead. The system introduces a dynamic programming algorithm to optimally distribute computational tasks across multiple processes, ensuring minimal communication costs. It also incorporates user-friendly compiler directives, allowing programmers to influence code generation easily and accommodate a broader range of parallelization scenarios without needing in-depth knowledge of parallel architectures. Arachne significantly reduces the learning curve and need for extensive code modifications, making parallel programming more accessible and efficient. Evaluation of various HPC benchmarks demonstrates that Arachne outperforms existing methods by reducing communication overhead, lowering memory requirements, and supporting more complex parallel logic, thus enhancing the overall scalability and efficiency of HPC applications.
- Open Access
- 17 June 2025
- Published by Association for Computing Machinery (ACM) in ACM Transactions on Architecture and Code Optimization
EDAS: Enabling Fast Data Loading for GPU Serverless Computing
Integrating GPUs into serverless computing platforms is crucial for improving efficiency. Many GPU functions, such as DNN inferences and scientific services, benefit from GPU usage, which requires only tens to hundreds of milliseconds for pure computation. Under these circumstances, fast data loading is imperative for function performance. However, existing GPU serverless systems face significant data stall issues, leading to extremely low GPU efficiency. Faced with the above problems, we observe opportunities to optimize data loading, such as data preloading and deduplicated data loading. However, these optimizations are impossible in existing GPU serverless systems due to the lack of insights into data information, such as data sizes and read-write attributes of function inputs. To address this, we propose a novel GPU serverless system, EDAS. EDAS first enhances user request specifications, allowing users to annotate data retrieved by GPU functions from the database with additional attributes. Based on this, EDAS takes over data loading from GPU functions and proposes two innovative data loading management schemes: a parallelized data loading scheme and a multi-stage resource exit scheme. Our experimental results show that EDAS reduces function duration by 16.2 × and improves system throughput by 1.91 × compared to the state-of-the-art serverless platform.
- 17 June 2025
- Published by Association for Computing Machinery (ACM) in ACM Transactions on Architecture and Code Optimization
Open AccessEDAS: Enabling Fast Data Loading for GPU Serverless Computing
Integrating GPUs into serverless computing platforms is crucial for improving efficiency. Many GPU functions, such as DNN inferences and scientific services, benefit from GPU usage, which requires only tens to hundreds of milliseconds for pure computation. Under these circumstances, fast data loading is imperative for function performance. However, existing GPU serverless systems face significant data stall issues, leading to extremely low GPU efficiency. Faced with the above problems, we observe opportunities to optimize data loading, such as data preloading and deduplicated data loading. However, these optimizations are impossible in existing GPU serverless systems due to the lack of insights into data information, such as data sizes and read-write attributes of function inputs. To address this, we propose a novel GPU serverless system, EDAS. EDAS first enhances user request specifications, allowing users to annotate data retrieved by GPU functions from the database with additional attributes. Based on this, EDAS takes over data loading from GPU functions and proposes two innovative data loading management schemes: a parallelized data loading scheme and a multi-stage resource exit scheme. Our experimental results show that EDAS reduces function duration by 16.2 × and improves system throughput by 1.91 × compared to the state-of-the-art serverless platform.
- Top Cited PapersOpen Access
- 30 March 2025
- Published by Association for Computing Machinery (ACM)
- p. 573-588
Improving GPU Sharing Performance through Adaptive Bubbleless Spatial-Temporal Sharing
- 30 March 2025
- Published by Association for Computing Machinery (ACM)
- p. 573-588
Top Cited PapersOpen AccessImproving GPU Sharing Performance through Adaptive Bubbleless Spatial-Temporal Sharing
- Open Access
- 21 March 2025
- Published by Association for Computing Machinery (ACM) in ACM Transactions on Architecture and Code Optimization
- Vol. 22 (1) , 1-24
Taming Flexible Job Packing in Deep Learning Training Clusters
Job packing is an effective technique to harvest the idle resources allocated to the deep learning (DL) training jobs but not fully utilized, especially when clusters may experience low utilization, and users may overestimate their resource needs. However, existing job packing techniques tend to be conservative due to the mismatch in scope and granularity between job packing and cluster scheduling. In particular, tapping the potential of job packing in the training cluster requires a local and fine-grained coordination mechanism. To this end, we propose a novel job-packing middleware named Gimbal , which operates between the cluster scheduler and the hardware resources. As middleware, Gimbal must not only facilitate coordination among the packed jobs but also support various scheduling objectives of different schedulers. Gimbal achieves dual functionality by introducing a set of worker calibration primitives designed to calibrate workers’ execution status in a fine-grained manner. The primitives obscure the complexity of the underlying job and resource management mechanisms, thus offering the generality and extensibility for crafting coordination policies tailored to various scheduling objectives. We implement Gimbal on a real-world GPU cluster and evaluate it with a set of representative DL training jobs. The results show that Gimbal improves different scheduling objectives up to 1.32× compared with the state-of-the-art job packing techniques.
- 21 March 2025
- Published by Association for Computing Machinery (ACM) in ACM Transactions on Architecture and Code Optimization
- Vol. 22 (1) , 1-24
Open AccessTaming Flexible Job Packing in Deep Learning Training Clusters
Job packing is an effective technique to harvest the idle resources allocated to the deep learning (DL) training jobs but not fully utilized, especially when clusters may experience low utilization, and users may overestimate their resource needs. However, existing job packing techniques tend to be conservative due to the mismatch in scope and granularity between job packing and cluster scheduling. In particular, tapping the potential of job packing in the training cluster requires a local and fine-grained coordination mechanism. To this end, we propose a novel job-packing middleware named Gimbal , which operates between the cluster scheduler and the hardware resources. As middleware, Gimbal must not only facilitate coordination among the packed jobs but also support various scheduling objectives of different schedulers. Gimbal achieves dual functionality by introducing a set of worker calibration primitives designed to calibrate workers’ execution status in a fine-grained manner. The primitives obscure the complexity of the underlying job and resource management mechanisms, thus offering the generality and extensibility for crafting coordination policies tailored to various scheduling objectives. We implement Gimbal on a real-world GPU cluster and evaluate it with a set of representative DL training jobs. The results show that Gimbal improves different scheduling objectives up to 1.32× compared with the state-of-the-art job packing techniques.
- 22 November 2024
- Published by Springer Nature in Frontiers of Computer Science
- Vol. 19 (5) , 1-3
FLAPS: fluctuation-aware power auction strategy for reducing the power overload probability
- 22 November 2024
- Published by Springer Nature in Frontiers of Computer Science
- Vol. 19 (5) , 1-3
FLAPS: fluctuation-aware power auction strategy for reducing the power overload probability
- Open Access
- 20 November 2024
- Published by Association for Computing Machinery (ACM) in ACM Transactions on Architecture and Code Optimization
- Vol. 21 (4) , 1-25
Potamoi: Accelerating Neural Rendering via a Unified Streaming Architecture
Neural Radiance Field (NeRF) has emerged as a promising alternative for photorealistic rendering. Despite recent algorithmic advancements, achieving real-time performance on today’s resource-constrained devices remains challenging. In this article, we identify the primary bottlenecks in current NeRF algorithms and introduce a unified algorithm-architecture co-design, Potamoi , designed to accommodate various NeRF algorithms. Specifically, we introduce a runtime system featuring a plug-and-play algorithm, SpaRW , which significantly reduces the per-frame computational workload and alleviates compute inefficiencies. Furthermore, our unified streaming pipeline coupled with customized hardware support effectively tames both SRAM and DRAM inefficiencies by minimizing repetitive DRAM access and completely eliminating SRAM bank conflicts. When evaluated against a baseline utilizing a dedicated DNN accelerator, our framework demonstrates a speedup and energy reduction of 53.1× and 67.7×, respectively, all while maintaining high visual quality with less than a 1.0 dB reduction in peak signal-to-noise ratio.
- 20 November 2024
- Published by Association for Computing Machinery (ACM) in ACM Transactions on Architecture and Code Optimization
- Vol. 21 (4) , 1-25
Open AccessPotamoi: Accelerating Neural Rendering via a Unified Streaming Architecture
Neural Radiance Field (NeRF) has emerged as a promising alternative for photorealistic rendering. Despite recent algorithmic advancements, achieving real-time performance on today’s resource-constrained devices remains challenging. In this article, we identify the primary bottlenecks in current NeRF algorithms and introduce a unified algorithm-architecture co-design, Potamoi , designed to accommodate various NeRF algorithms. Specifically, we introduce a runtime system featuring a plug-and-play algorithm, SpaRW , which significantly reduces the per-frame computational workload and alleviates compute inefficiencies. Furthermore, our unified streaming pipeline coupled with customized hardware support effectively tames both SRAM and DRAM inefficiencies by minimizing repetitive DRAM access and completely eliminating SRAM bank conflicts. When evaluated against a baseline utilizing a dedicated DNN accelerator, our framework demonstrates a speedup and energy reduction of 53.1× and 67.7×, respectively, all while maintaining high visual quality with less than a 1.0 dB reduction in peak signal-to-noise ratio.
- 10 October 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Computers
- Vol. 74 (2) , 386-400
Adaptive Kernel Fusion for Improving the GPU Utilization While Ensuring QoS
The prosperity of machine learning applications has promoted the rapid development of GPU architecture. It continues to integrate more CUDA Cores, larger L2 cache and memory bandwidth within SM. Moreover, the GPU integrates Tensor Core dedicated to matrix multiplication. Although studies have shown that task co-location could effectively improve system throughput, existing works only focus on resource scheduling at the SM level and cannot improve resource utilization within the SM. In this paper, we propose Aker, a static kernel fusion and scheduling approach to improve resource utilization inside the SM while ensuring the QoS (Quality-of-Service) of co-located tasks. Aker consists of a static kernel fuser, a duration predictor for fused kernels, an adaptive fused kernel selector, and an enhanced QoS-aware kernel manager. The kernel fuser enables the static and flexible fusion for a kernel pair. The kernel pair could be Tensor Core kernel and CUDA Core kernel, or computing-prefer CUDA Core kernel and memory-prefer CUDA Core kernel. After the kernel fuser provides multiple fused kernel versions for a kernel pair, the duration predictor precisely predicts the duration of the fused kernels and the adaptive fused kernel selector locates the optimal fused kernel version. Finally, the kernel manager invokes the fused kernel or the original kernel based on the QoS headroom of latency-critical tasks to improve the system throughput. Our experimental results show that Aker improves the throughput of best-effort applications compared with state-of-the-art solutions by 50.1% on average, while ensuring the QoS of latency-critical tasks.
- 10 October 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Computers
- Vol. 74 (2) , 386-400
Adaptive Kernel Fusion for Improving the GPU Utilization While Ensuring QoS
The prosperity of machine learning applications has promoted the rapid development of GPU architecture. It continues to integrate more CUDA Cores, larger L2 cache and memory bandwidth within SM. Moreover, the GPU integrates Tensor Core dedicated to matrix multiplication. Although studies have shown that task co-location could effectively improve system throughput, existing works only focus on resource scheduling at the SM level and cannot improve resource utilization within the SM. In this paper, we propose Aker, a static kernel fusion and scheduling approach to improve resource utilization inside the SM while ensuring the QoS (Quality-of-Service) of co-located tasks. Aker consists of a static kernel fuser, a duration predictor for fused kernels, an adaptive fused kernel selector, and an enhanced QoS-aware kernel manager. The kernel fuser enables the static and flexible fusion for a kernel pair. The kernel pair could be Tensor Core kernel and CUDA Core kernel, or computing-prefer CUDA Core kernel and memory-prefer CUDA Core kernel. After the kernel fuser provides multiple fused kernel versions for a kernel pair, the duration predictor precisely predicts the duration of the fused kernels and the adaptive fused kernel selector locates the optimal fused kernel version. Finally, the kernel manager invokes the fused kernel or the original kernel based on the QoS headroom of latency-critical tasks to improve the system throughput. Our experimental results show that Aker improves the throughput of best-effort applications compared with state-of-the-art solutions by 50.1% on average, while ensuring the QoS of latency-critical tasks.
- Open Access
Potamoi: Accelerating Neural Rendering via a Unified Streaming Architecture
Neural Radiance Field (NeRF) has emerged as a promising alternative for photorealistic rendering. Despite recent algorithmic advancements, achieving real-time performance on today's resource-constrained devices remains challenging. In this paper, we identify the primary bottlenecks in current NeRF algorithms and introduce a unified algorithm-architecture co-design, Potamoi, designed to accommodate various NeRF algorithms. Specifically, we introduce a runtime system featuring a plug-and-play algorithm, SpaRW, which significantly reduces the per-frame computational workload and alleviates compute inefficiencies. Furthermore, our unified streaming pipeline coupled with customized hardware support effectively tames both SRAM and DRAM inefficiencies by minimizing repetitive DRAM access and completely eliminating SRAM bank conflicts. When evaluated against a baseline utilizing a dedicated DNN accelerator, our framework demonstrates a speed-up and energy reduction of 53.1$\times$ and 67.7$\times$, respectively, all while maintaining high visual quality with less than a 1.0 dB reduction in peak signal-to-noise ratio.
Open AccessPotamoi: Accelerating Neural Rendering via a Unified Streaming Architecture
Neural Radiance Field (NeRF) has emerged as a promising alternative for photorealistic rendering. Despite recent algorithmic advancements, achieving real-time performance on today's resource-constrained devices remains challenging. In this paper, we identify the primary bottlenecks in current NeRF algorithms and introduce a unified algorithm-architecture co-design, Potamoi, designed to accommodate various NeRF algorithms. Specifically, we introduce a runtime system featuring a plug-and-play algorithm, SpaRW, which significantly reduces the per-frame computational workload and alleviates compute inefficiencies. Furthermore, our unified streaming pipeline coupled with customized hardware support effectively tames both SRAM and DRAM inefficiencies by minimizing repetitive DRAM access and completely eliminating SRAM bank conflicts. When evaluated against a baseline utilizing a dedicated DNN accelerator, our framework demonstrates a speed-up and energy reduction of 53.1$\times$ and 67.7$\times$, respectively, all while maintaining high visual quality with less than a 1.0 dB reduction in peak signal-to-noise ratio.