The Intel Trace Analyzer (ITA) timeline
In your ParallelPi directory, you should see a file named parallelPi.stf. This trace file was created automatically by Intel Trace Collector (ITC) as parallelPi executed. Once again, no modification within the parallelPi.c file was required.
After executing the parallelPi code, enter
Exercise: Do this!
[agopu@bc81 agopu]$ cd ~/MPI_Tutorial/ParallelPi [agopu@bc81 ParallelPi]$ traceanalyzer parallelPi.stf &
to start ITA. You will first see the initial timeline of the parallelPi program for four processes. Follow the instructions explained previously in the Initial ITA timeline and whole trace and Detailed ITA timeline sections to open up the detailed ITA timeline; you should see something similar to what is shown below:
In parallelPi the only communication requirement comes at the conclusion of the program. Each processor completes its computation, then waits for processor 0 to gather the results, as shown by the blue arrows. After the MPI_Reduce function call completes, processor 0 normalizes and prints the results.
Do not be alarmed if your timeline shows one or more of the processors starting far ahead of one of more of the others. The parallelPi program has a much shorter run time than roundRobin. This is why there is a few milli-second startup difference between processor 0 when compared to processors 1,2 and 3. ITA exagerates this difference in its timeline because of the short total timespan. For longer, more realistic programs this startup difference is inconsequential.
