AMD CodeAnalyst can collect many different kinds of performance data in a single measurement session. This is especially true of event-based profiling and Instruction-Based Sampling where a large number of events may be collected in a single performance experiment. The CodeAnalyst view feature organizes performance data into groups of logically related kinds of data in order to aid interpretation and analysis.
Select a view from the drop-down list of views to change to a different view.
The All Data view is displayed by default after data collection. The data is displayed in as many columns as there are distinct kinds of sample data.
The list of available views is determined by the kind of performance data that was collected. A view is only offered if the right kind of performance data is available to produce the view. CodeAnalyst provides several predefined views. See Predefined views for more details.
The choice of performance data and the aggregation/separation of data by CPU (core), process and thread is controlled through View Management. View Management aids interpretation by filtering out unneeded performance data and by breaking out (or summarizing) performance data across CPUs, processes and threads.
The CodeAnalyst GUI does not provide a way to directly create and save new views. Advanced users may chose to create their own views and should see View Configuration for more information. Expertise with XML is required.