Dataset for PhD Thesis

perf.hdf5 Logfiles and data used for performance plots can be found in the file perf.hdf5. Its structure is as follows: Paths with the structure: /<plot_filename.pdf>/logfiles/<path/to/logfiles.log> contain full dumps of logfiles for all runs participating in the generation of plot_filename.pdf. They contain the raw logfile contents as string. Every binary dumps its input file and configuration to the logfile. Used git branches and commit hashes are printed too. Runs are thus completely reproducible by inspecting the logfile contents. Data from logfiles is parsed using regular expressions and aggregated by taking averages, sums, products or other data aggregation. This data can be found under paths: /<plot_filename.pdf>/data/patterns/<pattern>/<path/to/logfiles.log>/raw for raw parsed data. /<plot_filename.pdf>/data/patterns/<pattern>/<path/to/logfiles.log>/aggregated for aggregated data passed to the plot script for further processing. where <pattern> is the regular expression used to extract information from the logfile and <path/to/logfiles.log> is the unique path/name of the logfile. The contents of that logfile can be found as described above. In some cases the path looks as /<plot_filename.pdf>/data/patterns/<pattern>/glob/<glob>/raw for raw parsed data. /<plot_filename.pdf>/data/patterns/<pattern>/glob/<glob>/aggregated for aggregated data passed to the plot script for further processing. where <glob> matches multiple logfiles that are parsed together. Aggregated data ending up in actual performance plots can be found under paths: /<plot_filename.pdf>/plot/Axis: <axis>/Label: <label>/x_data for values along the x-axis. /<plot_filename.pdf>/plot/Axis: <axis>/Label: <label>/y_data for values along the y-axis. /<plot_filename.pdf>/plot/Axis: <axis>/Label: <label>/x_abserr for absolute errors of the values along the x-axis. /<plot_filename.pdf>/plot/Axis: <axis>/Label: <label>/y_abserr for absolute errors of the values along the y-axis. /<plot_filename.pdf>/plot/Axis: <axis>/Label: <label>/x_relerr for relative errors of the values along the x-axis. /<plot_filename.pdf>/plot/Axis: <axis>/Label: <label>/y_relerr for relative errors of the values along the y-axis. where <axis> is the title or id of the plot axis and <label> is the label of the line on the plot and some of the above paths may be absent if data is zero exactly. Plot scripts can be found in the repository https://gitlab.com/roman.gruber/scripts. E7.hdf5, F7.hdf5 and G7.hdf5 For data related to the variance analysis, the files E7.hdf5, F7.hdf5 and G7.hdf5 contain contraction data of the respective lattices. They all share the same path structure: /<estimator>/<level> where <estimator> is the name of the estimator according to the main text in the document and <level> denotes the multigrid level. Under each of these paths, one can find a 4 dimensional dataset. The dimensions are labelled in this order: config enumerates the gauge configurations. Actual config numbers might not start from zero. timeslice enumerates the timeslices where random wall-sources have been placed. This mostly corresponds to the temporal extent of the lattice. src_per_timeslice enumerates the individual sources on each timeslice. time indicates the temporal lattice extent. This is raw correlator data in terms of complex float64 numbers, no data analysis or average has been performed on them. Plot scripts can be found in the repository https://gitlab.com/roman.gruber/scripts.perf.hdf5 用于生成性能绘图的日志文件与数据可在perf.hdf5文件中获取，其结构如下： 路径格式为： /<绘图文件名.pdf>/logfiles/<日志文件路径.log>，该路径下存储了参与生成该绘图文件的所有运行任务的完整日志转储，内容为原始日志字符串。每个二进制程序都会将输入文件与配置写入日志，同时还会输出所用的Git分支与提交哈希值，因此通过检视日志内容即可完全复现所有运行任务。 通过正则表达式解析日志文件中的数据，并通过求平均、求和、求积或其他聚合方式进行整合后的数据，可在以下路径中找到： /<绘图文件名.pdf>/data/patterns/<正则表达式模式>/<日志文件路径.log>/raw，对应原始解析数据； /<绘图文件名.pdf>/data/patterns/<正则表达式模式>/<日志文件路径.log>/aggregated，对应传入绘图脚本进行后续处理的聚合后数据。 其中<正则表达式模式>为用于从日志文件中提取信息的正则表达式，<日志文件路径.log>为该日志文件的唯一路径与名称，该日志文件的内容可如前文所述获取。 在部分场景中，路径会采用如下形式： /<绘图文件名.pdf>/data/patterns/<正则表达式模式>/glob/<通配符模式>/raw，对应原始解析数据； /<绘图文件名.pdf>/data/patterns/<正则表达式模式>/glob/<通配符模式>/aggregated，对应传入绘图脚本进行后续处理的聚合后数据。 此处<通配符模式>可匹配多个被一同解析的日志文件。 最终用于生成正式性能绘图的聚合数据可在以下路径中找到： /<绘图文件名.pdf>/plot/Axis: <轴标识>/Label: <曲线标签>/x_data，对应沿X轴分布的数值； /<绘图文件名.pdf>/plot/Axis: <轴标识>/Label: <曲线标签>/y_data，对应沿Y轴分布的数值； /<绘图文件名.pdf>/plot/Axis: <轴标识>/Label: <曲线标签>/x_abserr，对应X轴数值的绝对误差； /<绘图文件名.pdf>/plot/Axis: <轴标识>/Label: <曲线标签>/y_abserr，对应Y轴数值的绝对误差； /<绘图文件名.pdf>/plot/Axis: <轴标识>/Label: <曲线标签>/x_relerr，对应X轴数值的相对误差； /<绘图文件名.pdf>/plot/Axis: <轴标识>/Label: <曲线标签>/y_relerr，对应Y轴数值的相对误差。 其中<轴标识>为绘图坐标轴的标题或唯一标识符，<曲线标签>为绘图中曲线的标注。若某类误差数值恰好为0，则对应的路径可能不存在。 绘图脚本可在仓库https://gitlab.com/roman.gruber/scripts中获取。 E7.hdf5、F7.hdf5与G7.hdf5 用于方差分析的相关数据存储于E7.hdf5、F7.hdf5与G7.hdf5文件中，这些文件分别存储了对应格点的收缩数据，三者拥有一致的路径结构： /<估计器名称>/<多重网格层级>，其中<估计器名称>为文档正文中提及的估计器名称，<多重网格层级>表示多重网格的层级。 在每个路径下，均可找到一个四维数据集，各维度的标注顺序如下： config：枚举规范配置，实际配置编号未必从0开始； timeslice：枚举放置随机壁源的时间片，该维度通常对应格点的时间延展范围； src_per_timeslice：枚举每个时间片上的单个源； time：表示格点的时间延展范围。 该数据集为采用64位复数浮点数格式存储的原始关联函数数据，未经过任何数据分析或平均处理。 绘图脚本可在仓库https://gitlab.com/roman.gruber/scripts中获取。