Go backward to Performance Tools Go up to Top Go forward to Performance Tools

Performance Analysis

1. Data collection.
   • Process by which data about program performance are obtained from an executing program.
   • Profiles: time spent in different program parts.
   • Counters: event frequencies or cumulative times.
   • Event traces: occurences of various events.
2. Data transformation.
   • Bring data into form to yield insights.
   • Reduce data volume.
   • Mean values, higher-order statistics, extract profile/counter data from traces.
3. Data visualization.
   • Performance data are multidimensional.
   • Execution times communication costs, ...
   • Different processor numbers, problem sizes.
   • Explore data by bringing them into visual form.