The x86-specific floating-point subnormal operand exception is now
supported, and can be tested and trapped on x86 systems.
