As described in Enabling Profiling in Your Application, the
Xilinx Runtime (XRT) collects profiling data on
host applications and kernels when specific options are enabled in the xrt.ini file, such as opencl_trace, xrt_native_api, and
device_trace. XRT captures profiling data for the
host application as it makes calls to the runtime either through OpenCL or XRT API calls. You can also add user calls to
your host application to capture additional profiling information, as explained in Custom Profiling of the Host Application. To capture details of the kernel
operations, you must implement kernels in the .xclbin using the --profile Options as
explained in the next section.
After the application finishes running, the Profile Summary report is saved as .csv files in the directory where the compiled host code is executed. The Profile Summary provides annotated details regarding the overall application performance. All data generated during the execution of the application is grouped into categories. The Profile Summary lets you examine the kernel execution and data transfer statistics.
An example of the Profile Summary report is shown below.
