You may read the Getting Started document to get started. Regarding the compiler options for performance, please refer to the doc: Intel C++ Compiler User and Reference, there is a section for performance: Optimizing Applications.
For a comparison between gcc and icc on Core I7, in order to include vector optimizations such as sum reduction, I would use comparable options icc -O3 -xSSE4.2 -openmp -ansi-alias -std=c99 -prec-div -prec-sqrt bench.c gcc -O3 -msse4.2 -mtune=barcelona -openmp -std=c99 -ffast-math -fno-cx-limited-range -mno-recip -funroll-loops --param max-unroll-times=4 bench.c and I would test the gcc version with both libgomp and libiomp5. Important options for getting remarks about optimization from the compiler include additional options gcc -ftree-vectorizer-verbose=2 icc -opt-report TBB might be more useful than OpenMP, if using C++, depending on the nature of the application. Intel performance libraries work with both compilers; they come at no extra charge with icc. Both compilers support link-time optimization (lto/ipo) and profile feedback. Evidently, there are enough options that you would have to make a decision on what is relevant for your choice of application and platform, after consulting the docs for both compilers. You can see how most people would see a contradiction between a desire for simplicity and a desire to compare optimization of multiple compilers.