0

As we know, we should build project first with /GENPROFILE and run it for training, and then switch the command to /USEPROFILE to build the optimized bundle.

I have printed the time of each time-costing function and I found that it is more faster under /GENPROFILE stage than /USEPROFILE, actually the prior one is far beyond the performance than the latter one. Why this happened? If the compiler is using the trained pgd file, why the final .exe which using the trained pgd file is slower?

user3126461
  • 152
  • 8
  • Where's your code? – 1201ProgramAlarm May 04 '17 at 04:12
  • A lot of test code cannot be pasted here, but why this happened? Or, in what circumstances this will happen? – user3126461 May 04 '17 at 04:14
  • 1
    It will happen in the case that the profiler does a bad job of collecting information and optimizing the code. There's nothing more specific than that that can be said; it doesn't happen often, obviously. Perhaps you ran the profiling build under unusual conditions or with data that doesn't model real-world usage? If you want an answer to this, you need to create an [mcve]. – Cody Gray - on strike May 06 '17 at 11:48

0 Answers0