Neither the C standard nor IEEE-754 specify anything about floating-point operations in non-standard modes like DAZ or FZ. (Nor do they require that support fo

Neither the C standard nor IEEE-754 specify anything about floating-point operations in non-standard modes like DAZ or FZ. (Nor do they require that support for signaling NaNs be implemented). The behavior of -0.0 suffices to block the optimization under strict fp modes, however.