Yangjie Zhou

Large language models show promise for code generation, yet their effectiveness in producing high-performance GPU kernels remains insufficiently understood. Recent agentic workflows improve over direct prompting through specialized generation roles and execution feedback, but our characterization of a representative workflow reveals failures in optimization-strategy selection, hardware-specific instruction realization, and shape-dependent parameter selection. We present AlphaCu, a transformation-driven synthesis framework for LLM-based GPU kernel generation. AlphaCu treats kernel generation as structured synthesis over explicit optimization decisions rather than end-to-end code emission. Specifically, AlphaCu represents these decisions as scoped transformation passes maintained across agent-coordinated synthesis stages, enabling compatibility checking and failure attribution. AlphaCu binds selected hardware-sensitive operations to pre-validated microkernels and uses target-device measurements to revise transformation choices and tune runtime configurations. We evaluate AlphaCu on 104 test cases covering single and fused operators across four GPU architectures: NVIDIA H100, A100, V100, and AMD MI210. Under comparable generation budgets, AlphaCu achieves 100% functional correctness on our benchmark suite and delivers an average speedup of 5.64× over prior LLM workflows, while remaining competitive with vendor libraries and tensor compilers.

Graph neural networks (GNNs) are powerful tools for exploring and learning from graph structures and features. As such, achieving high-performance execution for GNNs becomes crucially important. Prior works have proposed to explore the sparsity (i.e., low density) in the input graph to accelerate GNNs, which uses the full-graph-level or block-level sparsity format. We show that they fail to balance the sparsity benefit and kernel execution efficiency. In this paper, we propose a novel system, referred to as AdaptGear, that addresses the challenge of optimizing GNNs performance by leveraging kernels tailored to the density characteristics at the subgraph level. Meanwhile, we also propose a method that dynamically chooses the optimal set of kernels for a given input graph. Our evaluation shows that AdaptGear can achieve a significant performance improvement, up to 6.49× (1.87× on average), over the state-of-the-art works on two mainstream NVIDIA GPUs across various datasets.

With the ever-increasing computational demand of DNN training workloads, distributed training has been widely adopted. A combination of data, model and pipeline parallelism strategy, called hybrid parallelism distributed training, is imported to tackle the problem of deploying large-scale models. However, how to evaluate the hybrid strategy and the utilization of each device remains a challenge since existing works either profile on a real large-scale cluster with high time and money costs or only analyze a specific type of parallelism without considering the hybrid parallelism. In this work, we proposed DistSim, an event-based performance model to accurately analyze each device's computation and communication activities with low profiling costs. DistDim breaks down the model into events according to the given distributed strategy, which can be profiled on two nodes. Then DistSim leverages the hierarchy of different parallel strategies to generate the computation and communication event-flow from layer level to model level and finally the activity timeline of each device participating in training. Experiment shows that DistSim can reach <4% errors when predicting distributing training batch time and <5% errors when predicting a single device's activity time in various hybrid strategy settings. We also provide a use-case of DistSim, automatically evaluate and search the best distributed training strategy, and find a hybrid strategy with at most 7.37× throughput improvement.