Go backward to Summary of Data Collection Go up to Top Go forward to Traces

Profile and Counts

Data Transformation and Visualization

• Typical information:
  • Time spent in each procedure on each processor.
  • Number of times each procedure is called.
  • Number of messages generated.
• Data reduction techniques to reduce number of dimensions:
  • Scalar data: single numbers (total computation time, total message number, mean message size, ...).
  • 1D data: histograms.
  • 2D data: color and a 2D-matrix.

See Foster, Plates 7 and 8.