Introduction to the Intel® GPA Frame Analyzer
The Intel® GPA Frame Analyzer is an interactive tool that enables you to profile application workloads within a single frame and to experiment with individual ergs (that is, any DirectX* call that potentially renders one or more pixels to the frame buffer) and various settings for the entire rendering pipeline, thereby allowing you to perform “what-if” experiments within the Intel GPA Frame Analyzer without modifying your original application code.
The Intel GPA Frame Analyzer works with a capture file that contains all the information used to render your original frame. This file is created by the Intel® GPA System Analyzer HUD or the Intel® GPA System Analyzer.
The Intel GPA Frame Analyzer provides a graphical user interface for conducting your experiments and hypothesis testing.
The Intel GPA Frame Analyzer is a client that connects to the Intel® GPA Monitor.
Once a single rendering frame from the application is captured, use the Intel GPA Frame Analyzer to play back the scene and gather performance metrics for that frame. If you adjust rendering states, the Intel GPA Frame Analyzer will re-render the frame with the new rendering state, enabling you to see the visual effect of the change and the difference in the rendering time for the change.
With the Intel® GPA Frame Analyzer you can:
- Load a frame capture file
- Analyze and experiment with the data to find performance bottlenecks:
-
-
Discover how much time the graphics application spends within each pass
-
Determine the most expensive ergs
-
Run experiments
-
Modify DirectX* states
-
Modify the shader code to see whether it is possible to improve render time
-
Determine whether texture bandwidth is a performance bottleneck
-
Minimize overdraw
-
See Also
Introduction to the Intel® GPA HUD
Intel® GPA Monitor
Loading a Frame Capture File
Discovering How Much Time the Graphics Application Spends within Each Pass
Discovering the Most Expensive Ergs
Running Experiments
Modifying DirectX* States
Modifying the Shader Code
Determining whether Texture Bandwidth is a Performance Bottleneck
Minimizing Overdraw