[libc-commits] [libc] [llvm] Reland "[libc][math] Refactor fmaximum_mag_num family to header-only" (PR #194194)

[Muhammad Bassiouni via libc-commits]

================
@@ -33,25 +33,30 @@ LIBC_INLINE constexpr T abs(T x) {
 namespace internal {
 
 template <typename T>
-LIBC_INLINE cpp::enable_if_t<cpp::is_floating_point_v<T>, T> max(T x, T y) {
+LIBC_INLINE constexpr cpp::enable_if_t<cpp::is_floating_point_v<T>, T>
+constexpr_max(T x, T y) {
   FPBits<T> x_bits(x);
   FPBits<T> y_bits(y);
-
-  // To make sure that fmax(+0, -0) == +0 == fmax(-0, +0), whenever x and y
-  // have different signs and both are not NaNs, we return the number with
-  // positive sign.
   if (x_bits.sign() != y_bits.sign())
----------------
bassiounix wrote:

```suggestion
LIBC_INLINE constexpr cpp::enable_if_t<cpp::is_floating_point_v<T>, T>
constexpr_max(T x, T y) {
  FPBits<T> x_bits(x);
  FPBits<T> y_bits(y);

  // To make sure that fmax(+0, -0) == +0 == fmax(-0, +0), whenever x and y
  // have different signs and both are not NaNs, we return the number with
  // positive sign.
  if (x_bits.sign() != y_bits.sign())
```

https://github.com/llvm/llvm-project/pull/194194