I built my program using " -g -O2 " and ran valgrind + cachegrind. I am not sure how to interpret the conclusion. Here is the result:
http://daviddoria.com/Uploads/callgrind.CacheMisses
My "whole program" is the InpaintingAlgorithm function, which is 98.4% of the "main" one. So far, so good. Now, looking at InpaintingAlgorithm calls, 92.9% of InpaintingAlgorithm is LinearSearchKNNProperty::operator() . This is my “inner cycle”, and again I expect a huge amount of time to be spent here.
Now that's where I got confused. Looking at the calls of LinearSearchKNNProperty::operator() , is there really nothing there? The largest function is only 7.64%, and the rest is 0.25%. I do not understand how the sum of all calls adds only about 8%. Where are the remaining 92%? (Presumably, the material that I would be looking for it to accelerate!)
If anyone can point me to my error reading these results, I would appreciate it!
source share