This data set augments the initial study of workload-specific code coverage by a hit count analysis, i.e., counting how many times a specific line of code/statement was executed. The experimental setup is identical to the one in the original study, utilizing the LLVM intermediate language for code analysis.

Data description

We provide coverage data for four (of the nine original) subject systems; for each subject system, we provide three CSV-files containing:

• sample.csv -- A list of configurations assessed in this study (binary configuration options).
• coverage.csv -- A list that maps lines of code (statement level; identified by source code file and line number) of the specific version (see the original paper) and utilized workloads to the respective coverage frequency.

• measurements.csv -- The original performance measurements corresponding to the original study, mapping confiuguration ids to scalar measurements.