Proposal: Add CMake option to enable floating point optimizations
I would like to propose to add a compiler-independent CMake option to enable non-standard compliant floating point optimizations (i. e. the -ffast-math
flag of GCC and the corresponding flags on other compilers).
This option could be called CMAKE_ENABLE_FAST_MATH
or similar.
When I write numerical code, I usually compile with these optimizations enabled, but it is quite tedious to sort the flags of different compilers out, and I think controlling this via a CMake option would be suitable. I also think other developers of numerical C/C++/Fortran software would make use of this.
If this is deemed useful, I might look into implementing this myself.
Thoughts?