Use checked_sub for inverse approximation

src/float/inv.rs

@@ -7,10 +7,14 @@ use super::F32;
 impl F32 {
     /// Fast approximation of `1/x`.
     pub fn inv(self) -> Self {
-        // Check if the value is too large for the approximation, NaN (e.g. 0x7fC0_0000) or infinity.
-        let bits = self.0.to_bits();
-        if bits >= 0x7f00_0000 {
-            return if self.0.is_infinite() {
+        // Perform the bit manipulation for the approximation
+        // The constant 0x7f00_0000 corresponds to the bit pattern for 1.0 in IEEE 754 format.
+        // Subtracting the bits of the original number from this constant effectively inverts the exponent,
+        // resulting in an approximation of the reciprocal.
+        match 0x7f00_0000_u32.checked_sub(self.0.to_bits()) {
+            Some(result) => Self(f32::from_bits(result)),
+            // Check if the value is too large for the approximation, NaN (e.g. 0x7fC0_0000) or infinity.
+            None => if self.0.is_infinite() {
                 // 1/∞ = 0; by definition.
                 if self.0.is_sign_positive() {
                     Self(0.0)
@@ -24,12 +28,6 @@ impl F32 {
                 Self(0.0)
             }
         }
-
-        // Perform the bit manipulation for the approximation
-        // The constant 0x7f00_0000 corresponds to the bit pattern for 1.0 in IEEE 754 format.
-        // Subtracting the bits of the original number from this constant effectively inverts the exponent,
-        // resulting in an approximation of the reciprocal.
-        Self(f32::from_bits(0x7f00_0000 - bits))
     }
 }