Application Timeline Trace page

Profile applications that use

Select the API to trace. When one of the OpenCL / HSA radio buttons is clicked, the project setting pages will display the options relevant to the selected API.

Enable navigation to source code (high overhead)

When checked, the Profiler generates a symbol information file from an application's debugging information (the .pdb file on Windows), containing one entry for each called OpenCL™ API. This symbol information file lets you navigate from an item in the API Trace in the Application Timeline Trace Session panel to the source location of the API call.

Write trace data in intervals during program execution (ms)

When selected, the Profiler periodically writes all queued trace data to disk during program execution. The interval (in milliseconds) at which to write trace data is specified using the value following the checkbox. When checked, in addition to writing data periodically, the Profiler also writes all queued trace data when the clReleaseContext OpenCL™ API is called. However, if an application does not call clReleaseContext, or if it calls any OpenCL™ APIs after the final clReleaseContext call, then it is possible that not all trace data is written to the disk. When unchecked, all trace data is written to disk when the application terminates. On Linux, this is the default (and only supported) mode for writing trace data. Thus on Linux, the UI lets you specify the interval but does not let you enable or disable writing the data in intervals.

Maximum number of APIs to trace

This controls how many APIs are traced over an application's lifetime. The default number of APIs to trace is 1 million. Limiting the number of APIs traced helps to prevent running out of memory while profiling. After the limit is reached, no additional APIs is traced, and the trace results do not include any additional information. Because of this, any information provided in the GPU Profiler Summary Pages might not be correct, as a complete trace is required to provide a fully-accurate application summary.

Always show API error codes

When checked, the Profiler reports the return codes for all OpenCL™ API calls. Some OpenCL™ API functions return an error code through a passed-in parameter. If the host application passes in NULL for that parameter, then the OpenCL™ runtime does not report an error code. The Profiler substitutes a non-null parameter in this case, and the API Trace can show the return code.

Collapse consecutive identical clGetEventInfo calls

Some OpenCL™ applications wait for certain Enqueue API calls to complete by continuously checking the status of the event returned by the Enqueue API. These applications do this by calling clGetEventInfo within a loop until the event status reaches a certain state (typically CL_COMPLETE). For these applications, the timeline and API trace can contain thousands of clGetEventInfo calls, making it difficult to easily analyze the timeline and trace data. To make analysis easier, the Profiler can collapse consecutive clGetEventInfo calls that have the same parameters and return values into a single entry in the timeline and API trace.

Generate occupancy information for each OpenCL kernel profiled

When checked, the Profiler generates kernel occupancy data for each OpenCL™ kernel dispatched to a GPU device.