Does using a Intel C++ compiler instead of the standard g++ compiler in compiling opencv codes make difference in algorithm times???
In a pdf linked in the case study below, the author claims that using Intel C++ compiler boosts the algorithm time compared to using a gcc compiler. Can anyone explain why is it so?? and what under circumstances does it happen???
Link to the case study : https://software.intel.com/content/ww...
On Page 3 : he shows
I think there are some areas where ICC (Intel compiler) is better at producing optimized code. For instance, for loop unrolling.
But overall, there are not so much difference between ICC and GCC for producing optimized code I think. Especially nowadays when open-source tools like GCC or clang are contributed but large number of engineers over the world (and by Intel engineers).
There are some benchmarks on the net. Quick glance shows that ICC can produce better optimized code.
But in my opinion, if you target performance, you will have to hand-write optimized code. And the difference between compilers should be negligible.