Jmchenn/AutoThread-dataset-runtime-log

DES多线程轨迹数据集（AutoThread数据集）旨在加速离散事件模拟（DES）系统，以支持仿真在环强化学习（SiL-RL）研究。该数据集记录了在不同应用负载、硬件平台和工作线程数量下的DES环境多线程执行轨迹，支持线程级行为分析、性能建模和自适应线程调优研究。数据集按四个维度组织：应用 × 硬件平台 × 负载参数 × 工作线程数，包含两类数据：PCS/和UAV/目录下的静态线程数扫描数据，以及Dynamic/目录下的动态负载比较实验轨迹。应用案例包括基于DES范式的PCS（模拟无线网络中多类通信任务）和UAV（基于真实飞行轨迹模拟多无人机自主导航和避障）。数据集在AMD EPYC 9734和Intel Xeon Gold 6338硬件平台上收集，包含超过10,000个多线程执行样本，每个样本由三个跟踪文件组成：硬件层跟踪（pidstat_out和perf_out）和应用层跟踪（simulation_out），涵盖34个特征变量。The DES Multithreaded Trajectory Dataset (AutoThread Dataset) targets research on accelerating Discrete Event Simulation (DES) systems for Simulation-in-the-Loop Reinforcement Learning (SiL-RL). It records multithreaded execution traces of DES environments under varying application workloads, hardware platforms, and worker-thread counts, supporting research in thread-level behavior analysis, performance modeling, and adaptive thread tuning. The dataset is organized along four dimensions—application × hardware platform × workload parameters × worker-thread count—and contains two categories of data: full static thread-count sweeps in PCS/ and UAV/ directories, and comparative experiment traces from dynamic workloads in Dynamic/. Application cases include PCS, a classic DES benchmark simulating multi-class communication tasks in wireless networks, and UAV, a case study simulating multi-UAV autonomous navigation based on real flight trajectories. Traces were collected on AMD EPYC 9734 and Intel Xeon Gold 6338 platforms, comprising over 10,000 multithreaded execution samples, each consisting of three trace files: hardware-layer traces (pidstat_out and perf_out) and application-layer traces (simulation_out), covering 34 feature variables.