The Wolfram Profiler is also useful for parallel Mathematica applications. To do this you first need to launch your parallel application in debug mode.
Now you should load the parallel profiler package.
Now you can execute the parallel code you want to profile using
the function ParallelProfile.
This should produce a profiler output report, but now there is one section
for each subkernel.
As described in the Wolfram Profiler documentation, you can link from the report to the source code that was executed. Note that this only works for code that was loaded from the project.