The Intel® OpenMP* runtime can theoretically scale linearly with an increase in cores. A dual socket quad core platform with 8 cores would allow an 8x speedup, and the Intel® OpenMP* runtime has been shown to scale at close to that level. The more computation done in parallel in a single parallel region by the threads of a software application, the closer it can get to the theoretical limit.
The Intel® OpenMP* runtime is used with our commercial compilers on the Intel® Xeon Phi™ coprocessor, which supports 240 hardware threads, and also on large SMP machines from vendors such as SGI* and Bull*. Testing on machines at this scale is an important part of our release qualification process.
No. The open-source Intel® OpenMP* runtime is compatible with the Intel compilers (icc and ifort on Linux*/OS X*, icl and ifort on Windows) and also with the gnu compilers (gcc and gfortran).
Full instructions that describe how to build the runtime are in the README file and the documentation.