Frame-Based Automatic Tracing

Introduction

The GPU PerfStudio server has the ability to automatically collect a trace from a connected application at a user-specified frame. When an application is run with GPU PerfStudio, a frame counter is used to count the number of frames rendered (starting at 1). Once the frame counter reaches the value specified, the frame will be captured. In the example below, the application will run until frame 150 is reached, and frame 150 will be traced. After the frame has been traced, the application rendering speed is reduced, indicating that the client can be connected, and the trace can be loaded for viewing.

Introduction

Setup for automatic tracing is done in the included "GPUPerfServer.cfg" file. Refer to GPU PerfServer options to learn more about the GPUPerfServer.cfg file, and the various settings that it provides.
The first step in enabling automatic tracing involves removing the leading '#' from the following line within GPUPerfServer.cfg, and inserting the desired frame index:

# capture-frame=150

In addition to specifying the frame index to trace, the user must also indicate the type of trace to be collected. In the same "GPUPerfServer.cfg" file, find the following line:

# trace-type=API

Uncomment the leading '#', and insert the type of trace to be collected at the frame specified in the step above. Valid options for trace-type are "API", "GPU", or "Linked". Note that this option is case-insensitive.
Refer to the Using GPU PerfStudio with DirectX 12 section for more detail on the capabilities of each type of trace.

After updating the server configuration file, start your application with GPU PerfServer as normal by using a method mentioned in the Starting your application with GPU PerfStudio section. Once your application is up and running, connect to it using the GPUPerfStudio Client. The Client will recognize that automatic tracing is enabled, and will open the trace with the data collected during the frame.