|
| 1 || 0 || 0 || "Negative zero", which is considered equal to regular zero
|-
| any || 0 || != 0 || Denormal/subnormal.
|-
| 0 || 0xFF || 0 || Positive Infinity
|}
Knowing this, determining special cases and can be done by done simple bit operations:
<code><pre>
* Exponent=0 and Mantissa=0: The number is 0.0 (if sign is 0) or -0.0 (negative zero if sign is 1). Note that for all intents and purposes, -0 is considered equal to 0.
fn is_zero(f: u32) -> bool { f & 0x7FFF_FFFF == 0 }
* Exponent=0 and Mantissa!=0: The number is a denormal or subnormal. If a number like this is given to an calculating instruction, an Unimplemented Operation is signaled.
fn is_subnormal(f: u32) -> bool { ((f & 0x7F80_0000) == 0) && ((f & 0x7F_FFFF) != 0) }
* Exponent=0xFF and Mantissa==0: The number is INFINITY (if sign is 0) or -INFINITY (if sign is 1).
fn is_nan(f: u32) -> bool { ((f & 0x7F80_0000) == 0x7F80_0000) && ((f & 0x7F_FFFF) != 0) }
* Exponent=0xFF and Mantissa!=0: The number is a NAN (Not a Number), which indicates an incorrect result (this is for example the result of 0.0/0.0 or sqrt(-2). There are two varieties of NAN, which are differented by the most significant bit of the mantissa: msb=1 is qNAN (quiet) and msb=0 is sNAN (signaling). Any sNAN that is given to a calculating instruction will immediately trigger an UnimplementedOperation. qNAN as input will cause the output of the instruction to be qNAN (though with a different payload) and no exception will be signaled.
fn is_quiet_nan(f: u32) -> bool { (f & 0x7FC0_0000) == 0x7FC0_0000) }
* Anything else: This is a regular floating point number.
fn absolute_value(f: u32) -> bool { f & 0x7FFF_FFFF }
fn negate(f: u32) -> bool { f ^ 0x8000_0000 }
</pre></code>
|