arXiv subjects

Wenzhao Wu

Publications and source records attributed to Wenzhao Wu.

7 recordsLinked to original sources

arXiv2026

Sharing a Fabric with Collective Communication: Two Storage Penalties in Deep Learning Training

Distributed DL training on HPC systems often shares one network fabric between NCCL/RCCL collective communication and parallel-filesystem I/O. Using a real GNN training workload on a Slingshot-11 system, we show that this sharing imposes two distinct costs. The primary cost is heavy-tailed DataLoader stalls: the typical DataLoader wait is just 15 ms at steady state, yet spikes to multiple seconds in 28% of Lustre iterations and 12% of VAST iterations. The secondary cost is traffic-class contention on collective communication: Lustre I/O stalls the all-reduce by up to 145$\times$ in an isolated benchmark. The two costs arise from different mechanisms. I/O stall latency affects any storage path that traverses the shared fabric, whereas all-reduce network contention occurs only when storage and collective communication share the same traffic class. Their common root cause is that storage I/O traverses the shared fabric. This work shows that node-local NVMe staging via DYAD (Our code is publicly available at https://github.com/flux-framework/dyad) eliminates both effects by keeping storage I/O off that path. Across a full training epoch, DYAD achieves a 7.4 times speedup over direct Lustre reads and a 1.06 times speedup over VAST. By the second epoch, once the local cache is fully warmed, DataLoader stalls are eliminated entirely, allowing DYAD to reach a 1.31 times speedup over VAST.

cs.DC↗

arXiv2026

CodegenBench: Can LLMs Write Efficient Code Across Architectures?

While large language models (LLMs) have been extensively evaluated on code generation tasks for general-purpose programming and GPU-accelerated environments (e.g., PyTorch, CUDA), their capabilities in CPU-oriented high-performance computing (HPC) across diverse architectures remain underexplored. To bridge this gap, we introduce CodegenBench, a comprehensive benchmark suite designed to evaluate the generation of efficient parallel code across three distinct hardware platforms: x86_64, Sunway, and Kunpeng. Our benchmark comprises 106 standard Basic Linear Algebra Subprograms (BLAS) routines establishing a fundamental baseline, alongside 20 specialized computational kernels adapted for each of the unique supercomputing architectures (LeetSunway and LeetKunpeng). Our extensive evaluation reveals that while state-of-the-art LLMs can generate optimized code for ubiquitous architectures like x86_64, they exhibit significant performance degradation on domain-specific architectures with limited public documentation and training data, highlighting critical limitations in cross-platform generalization. Furthermore, our analysis of factors influencing code quality such as implementation length and task complexity indicates that current LLMs are most effective for moderately difficult problems requiring concise code snippets. We open-source our dataset and automated evaluation infrastructure to facilitate future research in LLM-driven high-performance code generation. The resources are available at https://anonymous.4open.science/r/CodegenBench-EDE1/ and https://anonymous.4open.science/r/CodegenBenchDataset-2551.

cs.SE↗

arXiv2025

Detecting Non-Optimal Decisions of Embodied Agents via Diversity-Guided Metamorphic Testing

As embodied agents advance toward real-world deployment, ensuring optimal decisions becomes critical for resource-constrained applications. Current evaluation methods focus primarily on functional correctness, overlooking the non-functional optimality of generated plans. This gap can lead to significant performance degradation and resource waste. We identify and formalize the problem of Non-optimal Decisions (NoDs), where agents complete tasks successfully but inefficiently. We present NoD-DGMT, a systematic framework for detecting NoDs in embodied agent task planning via diversity-guided metamorphic testing. Our key insight is that optimal planners should exhibit invariant behavioral properties under specific transformations. We design four novel metamorphic relations capturing fundamental optimality properties: position detour suboptimality, action optimality completeness, condition refinement monotonicity, and scene perturbation invariance. To maximize detection efficiency, we introduce a diversity-guided selection strategy that actively selects test cases exploring different violation categories, avoiding redundant evaluations while ensuring comprehensive diversity coverage. Extensive experiments on the AI2-THOR simulator with four state-of-the-art planning models demonstrate that NoD-DGMT achieves violation detection rates of 31.9% on average, with our diversity-guided filter improving rates by 4.3% and diversity scores by 3.3 on average. NoD-DGMT significantly outperforms six baseline methods, with 16.8% relative improvement over the best baseline, and demonstrates consistent superiority across different model architectures and task complexities.

cs.SE↗

arXiv2024

Validating quantum-supremacy experiments with exact and fast tensor network contraction

The quantum supremacy experiment, such as Google Sycamore [Nature \textbf{574}, 505 (2019)], poses great challenge for classical verification due to the exponentially-increasing compute cost. Using a new-generation Sunway supercomputer within $8.5$ days, we provide a direct verification by computing three million exact amplitudes for the experimentally generated bitstrings, obtaining an XEB fidelity of $0.191\%$ (the estimated value is $0.224\%$). The leap of simulation capability is built on a multiple-amplitude tensor network contraction algorithm which systematically exploits the ``classical advantage" (the inherent ``store-and-compute" operation mode of von Neumann machines) of current supercomputers, and a fused tensor network contraction algorithm which drastically increases the compute efficiency on heterogeneous architectures. Our method has a far-reaching impact in solving quantum many-body problems, statistical problems as well as combinatorial optimization problems.

quant-ph↗

arXiv2021

Redefining the Quantum Supremacy Baseline With a New Generation Sunway Supercomputer

A major milestone in the era of noisy intermediate scale quantum computers is \textit{quantum supremacy} [Nature \textbf{574}, 505 (2019)] claimed on the Sycamore quantum processor of $53$ qubits, which can perform a random circuit sampling task within $200$ seconds while the same task is estimated to require a runtime of $10,000$ years on Summit. This record has been renewed with two recent experiments on the Zuchongzhi $2.0$ ($56$ qubits) and Zuchongzhi $2.1$ ($60$ qubits) quantum processors. On the other front of quantum supremacy comparison, there has also been continuous improvements on both the classical simulation algorithm as well as the underlying hardware. And a fair justification of the computational advantages for those quantum supremacy experiments would require to practically simulate the same problems on current top supercomputers, which is still in lack. Here we report the full-scale simulations of these problems on new generation Sunway supercomputer, based on a customized tensor network contraction algorithm. Our benchmark shows that the most challenging sampling task performed on Sycamore can be accomplished within $1$ week, thus collapsing the quantum supremacy claim of Sycamore. Additionally, we show that the XEB fidelities of the \textit{quantum supremacy circuits} with up to $14$ cycles can be verified in minutes, which also provides strong consistency check for quantum supremacy experiments. Our results redefine quantum supremacy baseline using the new generation Sunway supercomputer.

quant-ph↗

arXiv2021

Closing the "Quantum Supremacy" Gap: Achieving Real-Time Simulation of a Random Quantum Circuit Using a New Sunway Supercomputer

We develop a high-performance tensor-based simulator for random quantum circuits(RQCs) on the new Sunway supercomputer. Our major innovations include: (1) a near-optimal slicing scheme, and a path-optimization strategy that considers both complexity and compute density; (2) a three-level parallelization scheme that scales to about 42 million cores; (3) a fused permutation and multiplication design that improves the compute efficiency for a wide range of tensor contraction scenarios; and (4) a mixed-precision scheme to further improve the performance. Our simulator effectively expands the scope of simulatable RQCs to include the 10*10(qubits)*(1+40+1)(depth) circuit, with a sustained performance of 1.2 Eflops (single-precision), or 4.4 Eflops (mixed-precision)as a new milestone for classical simulation of quantum circuits; and reduces the simulation sampling time of Google Sycamore to 304 seconds, from the previously claimed 10,000 years.

quant-ph↗

arXiv2020

Cross-regional oil palm tree counting and detection via multi-level attention domain adaptation network

Providing an accurate evaluation of palm tree plantation in a large region can bring meaningful impacts in both economic and ecological aspects. However, the enormous spatial scale and the variety of geological features across regions has made it a grand challenge with limited solutions based on manual human monitoring efforts. Although deep learning based algorithms have demonstrated potential in forming an automated approach in recent years, the labelling efforts needed for covering different features in different regions largely constrain its effectiveness in large-scale problems. In this paper, we propose a novel domain adaptive oil palm tree detection method, i.e., a Multi-level Attention Domain Adaptation Network (MADAN) to reap cross-regional oil palm tree counting and detection. MADAN consists of 4 procedures: First, we adopted a batch-instance normalization network (BIN) based feature extractor for improving the generalization ability of the model, integrating batch normalization and instance normalization. Second, we embedded a multi-level attention mechanism (MLA) into our architecture for enhancing the transferability, including a feature level attention and an entropy level attention. Then we designed a minimum entropy regularization (MER) to increase the confidence of the classifier predictions through assigning the entropy level attention value to the entropy penalty. Finally, we employed a sliding window-based prediction and an IOU based post-processing approach to attain the final detection results. We conducted comprehensive ablation experiments using three different satellite images of large-scale oil palm plantation area with six transfer tasks. MADAN improves the detection accuracy by 14.98% in terms of average F1-score compared with the Baseline method (without DA), and performs 3.55%-14.49% better than existing domain adaptation methods.

cs.CV↗