When you select this option, the compiler generates smaller and faster sequences of instructions for floating-point operations. This can have a significant effect on the performance of programs that contain floating-point expressions, either separately or in loops.
However, when programs are compiled with the IMPRECISE option, floating-point exceptions might not be reported at the precise location where they occur. (This is especially true when the OPTIMIZE option is in effect.) In addition, floating-point operations can produce results that are not precisely IEEE conforming.