14.9.5 hardfloat

The hardfloat modifier is used to indicate that the compiler must use a calling convention that passes certain floating point parameters in VFP registers. (Used on ARM only)