[llvm] [APFloat] Add APFloat support for E8M0 type (PR #107127)

[Matt Arsenault via llvm-commits]

================
@@ -1165,7 +1200,7 @@ IEEEFloat::IEEEFloat(const fltSemantics &ourSemantics, integerPart value) {
 
 IEEEFloat::IEEEFloat(const fltSemantics &ourSemantics) {
   initialize(&ourSemantics);
-  makeZero(false);
+  ourSemantics.hasZero ? makeZero(false) : makeSmallestNormalized(false);
----------------
arsenm wrote:

It's not obvious to me that makeSmallestNormalized is the alternative. What's wrong with makeZero? You'll get a zero exponent, and an empty significand? 

https://github.com/llvm/llvm-project/pull/107127