Add section in profiling section to measure processing times of particular...

Add section in profiling section to measure processing times of particular algorithmic phases or cell operations.

29_profiling.tex

@@ -7,6 +7,41 @@ Intel's VTune Amplifier and Likwid, we equip the engine with a couple of profile
 options and features.
 
 
+\section{Profiling scope}
+
+Profiling can target the whole code or only particular algorithmic phases.
+The profiling target can be selected via a block
+
+\begin{verbatim}
+begin profiling
+  profiling-target = <enum>
+end profiling
+\end{verbatim}
+
+where we can choose between \texttt{whole\_code} and \texttt{neighbour\_merge}, \texttt{predictor}, and \texttt{update}.
+The default is \texttt{whole\_code}.
+The block must be placed within the \texttt{exahype-project} environment.
+Note that it is important to run the code for a certain
+number of \texttt{time-steps} instead of specifying a simulation
+\texttt{end-time} if a profiling target other than \texttt{whole\_code}
+is specified.
+
+Alternatively or additionally, we can measure the processing time of cell-wise operations
+such as (fused) cell update or space-time predictor computation via:
+
+\begin{verbatim}
+begin profiling
+  measure_cell_processing_times = true
+  measure_cell_processing_times_iter = <integer>
+end profiling
+\end{verbatim}
+
+The second parameter specifies the number of measurements to perform when measuring cell processing times. 
+The results are averaged over the number of measurements. The default is one measurement.
+
+Note that these real-time measurements can be performed in release mode.
+No special profiling build target must be specified.
+
 \section{Build targets}
 
 We currently support two build targets for profiling: