Xueying Wang
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
Xueying Wang has a total of 44 co-authors. Here are the top co-authors by number of shared publications:
- Guangli Li
Institute of Computing Technology, CAS
10 shared publications - Lei Liu
Chinese Academy of Sciences
6 shared publications - Shigang Li
Beijing University of Posts and Telecommunications
6 shared publications - Xiaobing Feng
SKLP, Institute of Computing Technology, CAS; University of Chinese Academy of Sciences, Beijing, Beijing, China
5 shared publications - Peng Zhao
Huawei 2012 Lab, Beijing, China
4 shared publications - Rongtian Fu
Beijing University of Posts and Telecommunications
4 shared publications
Latest Publications
- 1 January 2027
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Parallel and Distributed Systems
- Vol. 38 (1) , 49-66
PowerSparse: GPU-Accelerated Node Aggregation and Edge Messaging for Natural Graphs
- 1 January 2027
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Parallel and Distributed Systems
- Vol. 38 (1) , 49-66
PowerSparse: GPU-Accelerated Node Aggregation and Edge Messaging for Natural Graphs
- 1 November 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Parallel and Distributed Systems
- Vol. 37 (11) , 2508-2523
UltraGNN: A Sparse-Operator-Aware Framework for Accelerating Graph Neural Networks on Tensor Cores
- 1 November 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Parallel and Distributed Systems
- Vol. 37 (11) , 2508-2523
UltraGNN: A Sparse-Operator-Aware Framework for Accelerating Graph Neural Networks on Tensor Cores
- Open Access
- 27 September 2026
- Published by Association for Computing Machinery (ACM)
- p. 747-757
MX-KMeans: Accelerating K-Means Clustering through Microscaling Quantization
K-Means clustering is a classical unsupervised learning method widely used for its simplicity, efficiency, and broad applicability. In this work, we first analyze the numerical distributions of representative K-Means datasets and identify an opportunity for low-precision acceleration through hardware-native microscaling formats. Based on this observation, we propose MX-KMeans, a high-performance K-Means solution that accelerates clustering with precision-adaptive microscaling quantization. In practice, our MX-KMeans assigns different low-precision formats to data points in a pointwise manner according to their tolerance for quantization error: points well inside a cluster can use extremely low precision, while boundary points are protected with higher precision. The dominant point-to-centroid distance computation is then executed as mixed-precision GEMM, where NVIDIA Blackwell Tensor Cores directly consume microscaled low-precision values and their scaling factors without explicit dequantization overhead. Furthermore, MX-KMeans can be extended beyond standard Lloyd-style K-Means by reformulating pruning-based variants, including Elkan, Hamerly, and Yinyang, as a unified masked distance computation and further lowering it to Tensor-Core-friendly dense GEMM operations. Extensive experiments on an NVIDIA RTX 5090 GPU show that MX-KMeans achieves up to 4.08 × and 75.82 × end-to-end speedups over industry-standard cuML and FAISS, respectively, while preserving the clustering quality of high-precision baselines.
- 27 September 2026
- Published by Association for Computing Machinery (ACM)
- p. 747-757
Open AccessMX-KMeans: Accelerating K-Means Clustering through Microscaling Quantization
K-Means clustering is a classical unsupervised learning method widely used for its simplicity, efficiency, and broad applicability. In this work, we first analyze the numerical distributions of representative K-Means datasets and identify an opportunity for low-precision acceleration through hardware-native microscaling formats. Based on this observation, we propose MX-KMeans, a high-performance K-Means solution that accelerates clustering with precision-adaptive microscaling quantization. In practice, our MX-KMeans assigns different low-precision formats to data points in a pointwise manner according to their tolerance for quantization error: points well inside a cluster can use extremely low precision, while boundary points are protected with higher precision. The dominant point-to-centroid distance computation is then executed as mixed-precision GEMM, where NVIDIA Blackwell Tensor Cores directly consume microscaled low-precision values and their scaling factors without explicit dequantization overhead. Furthermore, MX-KMeans can be extended beyond standard Lloyd-style K-Means by reformulating pruning-based variants, including Elkan, Hamerly, and Yinyang, as a unified masked distance computation and further lowering it to Tensor-Core-friendly dense GEMM operations. Extensive experiments on an NVIDIA RTX 5090 GPU show that MX-KMeans achieves up to 4.08 × and 75.82 × end-to-end speedups over industry-standard cuML and FAISS, respectively, while preserving the clustering quality of high-precision baselines.
- Open Access
- 5 July 2026
- Published by Association for Computing Machinery (ACM)
- p. 50-54
Compiler-ASR: Bridging the IR-to-Assembly Gap for Compiler Optimization via Architecture-Specific Reward
Compiler pass sequence optimization using large language models has shown promise for code size reduction. However, existing methods optimize at the IR level and ignore architecture-specific effects. This creates a gap: sequences that reduce IR instruction counts may fail to improve, or even degrade, the final assembly code. We propose Compiler-ASR, a framework that introduces assembly-aware rewards into reinforcement learning training. The reward function combines IR-level and assembly-level signals, penalizing cases where IR improves but assembly regresses. We further expose assembly metrics during chain-of-thought reasoning, enabling the model to incorporate backend feedback before committing to a pass sequence. On x86-64, our method achieves 4.95% and 9.39% code size reduction over -Oz and -O2 on CompilerGym benchmarks, and 4.91% on SPEC CPU 2017.
- 5 July 2026
- Published by Association for Computing Machinery (ACM)
- p. 50-54
Open AccessCompiler-ASR: Bridging the IR-to-Assembly Gap for Compiler Optimization via Architecture-Specific Reward
Compiler pass sequence optimization using large language models has shown promise for code size reduction. However, existing methods optimize at the IR level and ignore architecture-specific effects. This creates a gap: sequences that reduce IR instruction counts may fail to improve, or even degrade, the final assembly code. We propose Compiler-ASR, a framework that introduces assembly-aware rewards into reinforcement learning training. The reward function combines IR-level and assembly-level signals, penalizing cases where IR improves but assembly regresses. We further expose assembly metrics during chain-of-thought reasoning, enabling the model to incorporate backend feedback before committing to a pass sequence. On x86-64, our method achieves 4.95% and 9.39% code size reduction over -Oz and -O2 on CompilerGym benchmarks, and 4.91% on SPEC CPU 2017.
- Open Access
- 26 June 2026
- Published by Association for Computing Machinery (ACM) in ACM Transactions on Architecture and Code Optimization
- Vol. 23 (2) , 1-25
MoonPoly : Bridging Code Generation and Adaptive Execution via Micro-Kernel Polymerization for Optimizing Dynamic-Shape Tensor Operators
The prevalence of dynamic tensor shapes, driven by applications like language model serving with varying sequence lengths, is a defining characteristic of modern deep neural networks. This dynamism poses a fundamental challenge: reconciling the need for intensive, offline code generation to achieve peak performance with the demand for low-latency, adaptive execution to handle unpredictable runtime tensor shapes. Consequently, mainstream strategies are ineffective. Vendor-provided libraries, while highly optimized for a subset of common shapes, suffer performance degradation on unconventional ones. Static tensor compilers are hamstrung by prohibitive just-in-time compilation overheads for each new shape. While recent dynamic-shape compilers offer an alternative, they rely on predefined shape ranges, making them brittle when inputs fall outside these bounds. To resolve this tension, we present MoonPoly , a dynamic-shape tensor compiler that introduces micro-kernel polymerization . Our approach decouples these conflicting requirements through a two-stage process. In the offline stage, it performs intensive auto-tuning to generate a set of micro-kernels and corresponding performance models. The online stage then performs adaptive execution, rapidly assembling a near-optimal tensor operator on-the-fly, guided by a lightweight cost model. Evaluated on an NVIDIA A100 GPU, MoonPoly achieves an average operator-level speedup of 1.27× over the cuBLAS library across a diverse set of operators and data types, which in turn yields end-to-end inference acceleration for a variety of models, including BERT, the Vision Transformer, and large language models.
- 26 June 2026
- Published by Association for Computing Machinery (ACM) in ACM Transactions on Architecture and Code Optimization
- Vol. 23 (2) , 1-25
Open AccessMoonPoly : Bridging Code Generation and Adaptive Execution via Micro-Kernel Polymerization for Optimizing Dynamic-Shape Tensor Operators
The prevalence of dynamic tensor shapes, driven by applications like language model serving with varying sequence lengths, is a defining characteristic of modern deep neural networks. This dynamism poses a fundamental challenge: reconciling the need for intensive, offline code generation to achieve peak performance with the demand for low-latency, adaptive execution to handle unpredictable runtime tensor shapes. Consequently, mainstream strategies are ineffective. Vendor-provided libraries, while highly optimized for a subset of common shapes, suffer performance degradation on unconventional ones. Static tensor compilers are hamstrung by prohibitive just-in-time compilation overheads for each new shape. While recent dynamic-shape compilers offer an alternative, they rely on predefined shape ranges, making them brittle when inputs fall outside these bounds. To resolve this tension, we present MoonPoly , a dynamic-shape tensor compiler that introduces micro-kernel polymerization . Our approach decouples these conflicting requirements through a two-stage process. In the offline stage, it performs intensive auto-tuning to generate a set of micro-kernels and corresponding performance models. The online stage then performs adaptive execution, rapidly assembling a near-optimal tensor operator on-the-fly, guided by a lightweight cost model. Evaluated on an NVIDIA A100 GPU, MoonPoly achieves an average operator-level speedup of 1.27× over the cuBLAS library across a diverse set of operators and data types, which in turn yields end-to-end inference acceleration for a variety of models, including BERT, the Vision Transformer, and large language models.
- Open Access
- 30 June 2025
- Published by Association for Computing Machinery (ACM) in ACM Transactions on Architecture and Code Optimization
- Vol. 22 (2) , 1-27
OptiFX: Automatic Optimization for Convolutional Neural Networks with Aggressive Operator Fusion on GPUs
Convolutional Neural Networks (CNNs) are fundamental to advancing computer vision technologies. As CNNs become more complex and larger, optimizing model inference remains a critical challenge in both industry and academia. On modern GPU platforms, CNN operators are typically memory-bound, leading to significant performance degradation due to memory wall effects. While recent advancements have utilized operator fusion–merging multiple operators into one–to enhance inference performance, the fusion of multiple region-based operators like convolution is seldom addressed. This article introduces AFusion , a novel operator fusion technique aimed at improving inference performance, and OptiFX, an automatic optimization framework based on this approach. OptiFX employs a cost-based backtracking search to identify optimal sub-graphs for fusion and utilizes template-based code generation to create efficient kernels for these fused sub-graphs. We evaluate OptiFX across seven prominent CNN architectures–GoogLeNet, ResNet, DenseNet, MobileNet, SqueezeNet, NasNet, and UNet–on Nvidia A6000 Ada, RTX 4090, and Jetson AGX Orin platforms. Our results demonstrate that OptiFX significantly outperforms existing methods, achieving average speedups of \(2.91\times\) , \(3.30\times\) , and \(2.09\times\) in accelerating inference performance on these platforms, respectively.- 30 June 2025
- Published by Association for Computing Machinery (ACM) in ACM Transactions on Architecture and Code Optimization
- Vol. 22 (2) , 1-27
Open AccessOptiFX: Automatic Optimization for Convolutional Neural Networks with Aggressive Operator Fusion on GPUs
Convolutional Neural Networks (CNNs) are fundamental to advancing computer vision technologies. As CNNs become more complex and larger, optimizing model inference remains a critical challenge in both industry and academia. On modern GPU platforms, CNN operators are typically memory-bound, leading to significant performance degradation due to memory wall effects. While recent advancements have utilized operator fusion–merging multiple operators into one–to enhance inference performance, the fusion of multiple region-based operators like convolution is seldom addressed. This article introduces AFusion , a novel operator fusion technique aimed at improving inference performance, and OptiFX, an automatic optimization framework based on this approach. OptiFX employs a cost-based backtracking search to identify optimal sub-graphs for fusion and utilizes template-based code generation to create efficient kernels for these fused sub-graphs. We evaluate OptiFX across seven prominent CNN architectures–GoogLeNet, ResNet, DenseNet, MobileNet, SqueezeNet, NasNet, and UNet–on Nvidia A6000 Ada, RTX 4090, and Jetson AGX Orin platforms. Our results demonstrate that OptiFX significantly outperforms existing methods, achieving average speedups of \(2.91\times\) , \(3.30\times\) , and \(2.09\times\) in accelerating inference performance on these platforms, respectively. - Top Cited PapersOpen Access
- 28 February 2025
- Published by Association for Computing Machinery (ACM)
- p. 312-325
FlashSparse: Minimizing Computation Redundancy for Fast Sparse Matrix Multiplications on Tensor Cores
Sparse Matrix-matrix Multiplication (SpMM) and Sampled Dense-dense Matrix Multiplication (SDDMM) are important sparse operators in scientific computing and deep learning. Tensor Core Units (TCUs) enhance modern accelerators with superior computing power, which is promising to boost the performance of matrix operators to a higher level. However, due to the irregularity of unstructured sparse data, it is difficult to deliver practical speedups on TCUs. To this end, we propose FlashSparse, a novel approach to bridge the gap between sparse workloads and the TCU architecture. Specifically, FlashSparse minimizes the sparse granularity for SpMM and SDDMM on TCUs through a novel swap-and-transpose matrix multiplication strategy. Benefiting from the minimum sparse granularity, the computation redundancy is remarkably reduced while the computing power of TCUs is fully utilized. Besides, FlashSparse is equipped with a memory-efficient thread mapping strategy for coalesced data access and a sparse matrix storage format to save memory footprint. Extensive experimental results on H100 and RTX 4090 GPUs show that FlashSparse sets a new state-of-the-art for sparse matrix multiplications (geometric mean 5.5x speedup over DTC-SpMM and 3.22x speedup over RoDe).
- 28 February 2025
- Published by Association for Computing Machinery (ACM)
- p. 312-325
Top Cited PapersOpen AccessFlashSparse: Minimizing Computation Redundancy for Fast Sparse Matrix Multiplications on Tensor Cores
Sparse Matrix-matrix Multiplication (SpMM) and Sampled Dense-dense Matrix Multiplication (SDDMM) are important sparse operators in scientific computing and deep learning. Tensor Core Units (TCUs) enhance modern accelerators with superior computing power, which is promising to boost the performance of matrix operators to a higher level. However, due to the irregularity of unstructured sparse data, it is difficult to deliver practical speedups on TCUs. To this end, we propose FlashSparse, a novel approach to bridge the gap between sparse workloads and the TCU architecture. Specifically, FlashSparse minimizes the sparse granularity for SpMM and SDDMM on TCUs through a novel swap-and-transpose matrix multiplication strategy. Benefiting from the minimum sparse granularity, the computation redundancy is remarkably reduced while the computing power of TCUs is fully utilized. Besides, FlashSparse is equipped with a memory-efficient thread mapping strategy for coalesced data access and a sparse matrix storage format to save memory footprint. Extensive experimental results on H100 and RTX 4090 GPUs show that FlashSparse sets a new state-of-the-art for sparse matrix multiplications (geometric mean 5.5x speedup over DTC-SpMM and 3.22x speedup over RoDe).
- Open Access
- 19 January 2024
- Published by Association for Computing Machinery (ACM) in ACM Transactions on Architecture and Code Optimization
- Vol. 21 (1) , 1-26
Fast Convolution Meets Low Precision: Exploring Efficient Quantized Winograd Convolution on Modern CPUs
Low-precision computation has emerged as one of the most effective techniques for accelerating convolutional neural networks and has garnered widespread support on modern hardware. Despite its effectiveness in accelerating convolutional neural networks, low-precision computation has not been commonly applied to fast convolutions, such as the Winograd algorithm, due to numerical issues. In this article, we propose an effective quantized Winograd convolution, named LoWino, which employs an in-side quantization method in the Winograd domain to reduce the precision loss caused by transformations. Meanwhile, we present an efficient implementation that integrates well-designed optimization techniques, allowing us to fully exploit the capabilities of low-precision computation on modern CPUs. We evaluate LoWino on two Intel Xeon Scalable Processor platforms with representative convolutional layers and neural network models. The experimental results demonstrate that our approach can achieve an average of 1.84× and 1.91× operator speedups over state-of-the-art implementations in the vendor library while preserving accuracy loss at a reasonable level.
- 19 January 2024
- Published by Association for Computing Machinery (ACM) in ACM Transactions on Architecture and Code Optimization
- Vol. 21 (1) , 1-26
Open AccessFast Convolution Meets Low Precision: Exploring Efficient Quantized Winograd Convolution on Modern CPUs
Low-precision computation has emerged as one of the most effective techniques for accelerating convolutional neural networks and has garnered widespread support on modern hardware. Despite its effectiveness in accelerating convolutional neural networks, low-precision computation has not been commonly applied to fast convolutions, such as the Winograd algorithm, due to numerical issues. In this article, we propose an effective quantized Winograd convolution, named LoWino, which employs an in-side quantization method in the Winograd domain to reduce the precision loss caused by transformations. Meanwhile, we present an efficient implementation that integrates well-designed optimization techniques, allowing us to fully exploit the capabilities of low-precision computation on modern CPUs. We evaluate LoWino on two Intel Xeon Scalable Processor platforms with representative convolutional layers and neural network models. The experimental results demonstrate that our approach can achieve an average of 1.84× and 1.91× operator speedups over state-of-the-art implementations in the vendor library while preserving accuracy loss at a reasonable level.
- 25 August 2023
- Published by Elsevier in Journal of Systems Architecture
- Vol. 143
CoAxNN: Optimizing on-device deep learning with conditional approximate neural networks
- 25 August 2023
- Published by Elsevier in Journal of Systems Architecture
- Vol. 143
CoAxNN: Optimizing on-device deep learning with conditional approximate neural networks
- Open Access
- 16 September 2022
- Published by Association for Computing Machinery (ACM) in ACM Transactions on Architecture and Code Optimization
- Vol. 19 (4) , 1-26
An Application-oblivious Memory Scheduling System for DNN Accelerators
Deep Neural Networks (DNNs) tend to go deeper and wider, which poses a significant challenge to the training of DNNs, due to the limited memory capacity of DNN accelerators. Existing solutions for memory-efficient DNN training are densely coupled with the application features of DNN workloads, e.g., layer structures or computational graphs of DNNs are necessary for these solutions. This would result in weak versatility for DNNs with sophisticated layer structures or complicated computation graphs. These schemes usually need to be re-implemented or re-adapted due to the new layer structures or the unusual operators in the computational graphs introduced by these DNNs. In this article, we review the memory pressure issues of DNN training from the perspective of runtime systems and model the memory access behaviors of DNN workloads. We identify the iterative, regularity , and extremalization properties of memory access patterns for DNN workloads. Based on these observations, we propose AppObMem, an application-oblivious memory scheduling system. AppObMem automatically traces the memory behaviors of DNN workloads and schedules the memory swapping to reduce the memory pressure of the device accelerators without the perception of high-level information of layer structures or computation graphs. Evaluations on a variety of DNN models show that, AppObMem obtains 40–60% memory savings with acceptable performance loss. AppObMem is also competitive with other open sourced SOTA schemes.
- 16 September 2022
- Published by Association for Computing Machinery (ACM) in ACM Transactions on Architecture and Code Optimization
- Vol. 19 (4) , 1-26
Open AccessAn Application-oblivious Memory Scheduling System for DNN Accelerators
Deep Neural Networks (DNNs) tend to go deeper and wider, which poses a significant challenge to the training of DNNs, due to the limited memory capacity of DNN accelerators. Existing solutions for memory-efficient DNN training are densely coupled with the application features of DNN workloads, e.g., layer structures or computational graphs of DNNs are necessary for these solutions. This would result in weak versatility for DNNs with sophisticated layer structures or complicated computation graphs. These schemes usually need to be re-implemented or re-adapted due to the new layer structures or the unusual operators in the computational graphs introduced by these DNNs. In this article, we review the memory pressure issues of DNN training from the perspective of runtime systems and model the memory access behaviors of DNN workloads. We identify the iterative, regularity , and extremalization properties of memory access patterns for DNN workloads. Based on these observations, we propose AppObMem, an application-oblivious memory scheduling system. AppObMem automatically traces the memory behaviors of DNN workloads and schedules the memory swapping to reduce the memory pressure of the device accelerators without the perception of high-level information of layer structures or computation graphs. Evaluations on a variety of DNN models show that, AppObMem obtains 40–60% memory savings with acceptable performance loss. AppObMem is also competitive with other open sourced SOTA schemes.