Optimizing MPI Communications

 Use the Intel® Trace Analyzer to analyze an MPI application's behavior to improve performance at the inter-process level.

This part of the tutorial uses the sample trace files poisson_sendrecv.single.stf and poisson_icomm.single.stf to demonstrate how to detect and remove serialization in your application.

Step 1: Prepare for analysis

Use the Intel Trace Analyzer Event Timeline chart to zoom in to a single iteration of your application.

Step 2: Detect serialization

Step 3: Remove serialization

Improve your application performance by replacing the problem-causing function.

Step 4: Check the result

Use the Intel Trace Analyzer Comparison chart to compare the serialized application with the revised one.

Step 5: Analyze optimized communications

Analyze the revised application in the Event Timeline to see if the revised code needs further optimization.

Key Terms

Idealized Tracefile
Serialization

For more complete information about compiler optimizations, see our Optimization Notice.