Tutorial: Finding Hotspots - C++ Sample Code (Windows* OS)

Tutorial

Contents

Compare with Previous Result

You optimized your code to apply a loop interchange mechanism. To understand whether you got rid of the memory bound issue and what kind of optimization you got per function, re-run the Microarchitecture Exploration analysis on the optimized code and compare results:

Analyze Results After Optimization

Run the Microarchitecture Exploration analysis on the modified code.
VTune
Profiler
 automatically opens the
Microarchitecture Exploration
 viewpoint.
The optimized result shows an improvement in the µPipe diagram with about
48%
 of pipeline slots retiring compared to only
11%
 before optimization.

Compare Results Before and After Optimization

  1. Close the new result file.
  2. Select the result in the
    Project Navigator
    , right-click, and choose
    Compare Results
     from the context menu.
    The
    Compare Results
     window opens.
  3. Specify the analysis results you want to compare and click the
    Compare
     button.
    The
    Summary
     window opens, providing a high-level picture of performance improvements in the following format:
    <
    result 1 value
    > - <
    result 2 value
    >
    .

Identify the Performance Gain

In the
Summary
 window, you see that the Elapsed Time shows
50.686
 seconds of optimization for the whole application execution and an improvement from
11%
 of instructions retired to
48.3%
 of instructions retired. The Back-End Bound metric improved by
48.8%
.
Switch to the
Bottom-up
 window to compare the two results and see the differences per metrics side by side.

Product and Performance Information

1

Performance varies by use, configuration and other factors. Learn more at www.Intel.com/PerformanceIndex.