Code Confidence code analysis integration

The following screenshot illustrates the presentation of basic block execution counts derived from the the eCos twothreads example application:

It can be observed that the application entry function (cyg_user_start) is executed once only, while code within the two application-generated threads is executed multiple times within a loop. This data has been generated by enabling code coverage features within the GNU compiler (GCC). Coverage data is accumulated in memory during application execution and retrieved from the embedded system using Code Confidence™ remote code coverage capabilities. The resulting data files may then be processed by the GCov Integration feature (from the Eclipse Linux Tools project) for presentation within the Eclipse workbench. This feature also summarises coverage data across multiple source code files, providing an immediate indication of code which is not being exercised or may require optimisation:

In the above screenshot, code coverage data for both the example application code and the eCos kernel package may be observed.

The Code Confidence Tools also enable the rapid retrieval of statistical profiling data. Such data may be generated using a profiling timer running on the embedded target, or (non-invasively) using a suitable hardware debug adapter. For details of compatible hardware, please contact us.

The resulting data files may be processed by the GProf Integration feature (from the Eclipse Linux Tools project) for presentation within the Eclipse workbench, providing a breakdown of processor time consumed by various parts of an application:

In the above screenshot, it can be observed that the application is spending the majority of time executing the hal_idle_thread_action function. Significant time is also spent in the delay_loop function.

For an illustration of non-invasive profiling using the Code Confidence Tools, please refer to technical note 0004. Back…