CMake Cookbook

How it works

Most modern processors provide vector instruction sets. Carefully crafted code can exploit these and achieve enhanced performance with respect to non-vectorized code. The Eigen library has been written with vectorization explicitly in mind since linear algebra operations can greatly benefit from it. All we need to do is instruct the compiler to inspect the processor for us and generate the native set of instructions for the current architecture. Different compiler vendors use different flags to achieve this: the GNU compiler implements this by means of the -march=native flag, whereas the Intel compiler uses the -xHost flag. We then use the check_cxx_compiler_flag function offered by the CheckCXXCompilerFlag.cmake module:

check_cxx_compiler_flag("-march=native" ...