---

Using Shark

Shark is one of the most powerful analysis tools you can use to find performance problems in your code. Shark can generate a time-based profile of your program’s execution, trace function calls, and graph memory allocations over time. You can use Shark to track information for your program or for the entire system. Shark even captures execution information for kernel entities such as drivers and kernel extensions.

Despite the power of Shark, using the tool is very simple. Upon launch, Shark presents you with the window shown in Figure 4-4. Click Start (or use the Option-Escape global shortcut) to start sampling all system processes. Click the button again (or use the same global shortcut) to stop gathering data and display the samples (Figure 4-5).

If you want to limit sampling to only one process, you can use the pop-up menu on the right side of the window to select that process. Also, if you want to do something other than a statistical time profile, choose the appropriate option from the configuration pop-up menu before you click Start.

When you stop sampling, Shark displays the profile window (Figure 4-5) with the gathered data. This is the main window you use to identify potential problems. You can configure this window to display a heavy view, tree view, or both.

The heavy view sorts function calls based on the frequency by which they appeared. This view identifies your program’s hot spots. If you see one of your program’s functions near the top of this view, you should investigate. Functions tend to have higher weights if they are poorly optimized but a more likely scenario is that the function is being called too frequently from some other place. This can indicate an inefficient higher-level algorithm.

The tree view shows the same data organized by calling hierarchy. This view offers a convenient way to understand the context in which a particular function is called. This is the more traditional way to view call stack data and can be used in conjunction with the heavy view to track down hot spots and to see the surrounding context quickly.

The Shark profile window makes it easy to filter out irrelevant code and libraries through a feature called data mining. The Data Mining menu has several options for excluding or flattening symbols and libraries. If you apply these commands, Shark rolls the costs of calling those symbols and libraries into the function that called them. For example, if you know your code makes several calls to Core Foundation and you exclude the Core Foundation library, the time spent in Core Foundation now appears as time spent in your code. If the amount of time spent in your code jumps significantly afterward, you might want to investigate ways to call fewer Core Foundation functions.

Shark also can help you identify performance issues within a given function through the code view. When you double-click a function, Shark displays the source code for that function if it is available (see Figure 4-6). It applies color to lines of code based on the amount of time spent in that code. For each line of source, brighter shades of yellow indicate where more time was spent during the profile.

Shark occasionally offers specific tuning tips and comments in the margin. Clicking the exclamation-point icon displays a tip that you can use to improve your code. The comment column displays a summary of the tip.

For more information about Shark and its features, choose Help > Shark Help to display the Shark User Guide.

---