Using the GPU Profiler

CodeXL

PreviousNext
CodeXL User Guide
Help > Using CodeXL > GPU Profiler > Using the GPU Profiler
Using the GPU Profiler

The GPU Profiler provides two modes:

Application Timeline Trace

This mode provides a high-level overview of an OpenCL™ or HSA application. It displays:
1. An API Trace, showing all OpenCL™ of HSA APIs called by the application.
2. A timeline showing the call sequence and duration of all OpenCL™ or HSA APIs called by the host as well as data transfers (OpenCL™ only) and kernels executing on a device.
3. A set of Summary Pages, providing a set of statistics for the application, as well as the results of detailed analysis of the application.
For more information, see:
Application Timeline Trace Session
Application Timeline Trace Summary Pages

Performance Counters

This mode collects performance counters from the AMD GPU or APU for each kernel dispatched to the device by either and OpenCL™, HSA or DirectCompute application.
It also displays statistics from the shader compiler for each kernel dispatched.
The performance counters and statistics can be used to discover bottlenecks in a particular kernel.
This mode also can also display the kernel source code, the generated IL code, and the compiled ISA code for an OpenCL™, DirectCompute kernel dispatched to a GPU.
For more information, see:
GPU Profiler Performance Counters Session
GPU Profiler Code Viewer.

For OpenCL™ programs, both profiling modes can also generate Kernel Occupancy information for each kernel dispatched to a GPU. For HSA applications Kernel Occupancy information is only available in Performance Counter mode. For more information, see GPU Profiler Kernel Occupancy Viewer and GPU Profiler Kernel Occupancy

To use the GPU Profiler:

1.      Create a new project, or open an existing project.

2.      Switch to Profile Mode in CodeXL.
You can switch to Profile Mode using the
CodeXL Toolbar or the CodeXL menu.

3.      In Profile Mode, use the menu to select one of the above two modes.

4.      Start the profile session using the "Start CodeXL Profiling" toolbar or menu item. Profiling results are gathered while the application is running.

Once the application terminates, a new session is added to the CodeXL Explorer. The results of the profile also are displayed by CodeXL.

The following links provide more information on the features available in the GPU Profiler.

Application Timeline Trace Session

GPU Profiler Performance Counters Session

GPU Profiler Summary Pages

GPU Profiler Code Viewer

GPU Profiler Kernel Occupancy Viewer

GPU Profiler Kernel Occupancy

GPU Profile Project Settings

Description of Output Files

Description of Configuration Files

Using the Command-Line Interface

AMDTActivityLogger Library