Differential Performance Analysis Workflow for Algorithmic Changes
TimeSunday, 14 November 20219:05am - 9:35am CST
DescriptionMost performance analysis tools used in HPC focus on a single configuration of an application. In this work, we instead present a novel performance analysis workflow, supporting the in-depth comparison of varied code versions and running conditions.
There exist different code versions for many applications because they comprise parts that can be implemented in various ways or already exist in third-party libraries, like linear solvers. What is the impact of replacing a solver to the rest of the application? Such measurements are often not supported directly and are cumbersome to handle manually with current performance measurement and analysis tools. We present a workflow that handles the multitude of measurements and data collation. Furthermore, we introduce diagrams suited for a clear presentation of the collected performance data. Our application studies with CalculiX and Jukkr highlight that our workflow allows a detailed performance analysis while still being easy to use.