Benchmark and profiling data underlying the publication: WaterLily.jl: A differentiable and backend-agnostic Julia solver for incompressible viscous flow around dynamic bodies

This dataset contains the benchmark and profiling results of WaterLily v1.4.0 which have been used to generate Figures 1,2,3 of the linked article. The dataset allows to analyze the computational performance of the solver. The benchmarks data is a JSON file which includes the solver's execution time of 100 time steps for different test cases, backends, and simulation sizes. The profiling data is the NVIDIA NVTX measured time range of the main routines of the solver (.nsys-rep files), and the performance of the main kernels (pressure solver 327, and convection/diffusion 355) measured by NVIDIA NCU (.nsys-ncu files).

本数据集包含WaterLily v1.4.0的基准测试与性能剖析结果，这些数据已用于生成关联文章中的图1、图2与图3。本数据集可用于分析该求解器的计算性能。基准测试数据为JSON格式文件，涵盖了不同测试用例、后端环境以及模拟规模下，求解器执行100个时间步的耗时情况。性能剖析数据包含两部分：一是通过NVIDIA NVTX测量得到的求解器主例程时间范围（存储为.nsys-rep文件）；二是由NVIDIA NCU测得的两类核心内核的性能数据——压力求解器327与对流/扩散355，对应文件格式为.nsys-ncu。