The xrt.run_summary file can be loaded into Vitis Analyzer and an example of the results is shown in the picture. The application run summary is presented under the System Compilation section of the Profile Summary tab. It contains the following three tables,
- Host Iterations: This table shows a summary of the send and the receive threads iterations, and the number of compute() calls, issued by the application layer and useful runtime statistics. In this example, 13 compute jobs were run with a total runtime of 3.552 milliseconds.
- Host Transfer Details: This table captures the number of data transfers between the application code (running on a host) and the device. In this example, data is transferred using the HBM banks 0 and 2. Several average runtime statistics on the data transfers are shown.
- CU-pipeline usage: This table captures the activity in the compute unit pipelines. Each row represents a CU in the hardware and average run time statics are shown, In this example, the 13 compute jobs were distributed with 8 on CU0 and 5 on CU1.
Note: These columns have tooltips in Vitis Analyzer. If the mouse hovers over any column header in the table, a pop-up will show the explanation of the column. For example, hovering over "Average Time (ms)" will show "= Total time (in milliseconds) / Number of calls."
Figure 1. Profile Summary