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

[Matt Arsenault via llvm-commits]

================
@@ -1075,6 +1099,11 @@ bool IEEEFloat::isSignificandAllOnesExceptLSB() const {
 }
 
 bool IEEEFloat::isSignificandAllZeros() const {
+  // For the E8M0 format, this is always true since there are no
+  // actual significand bits.
+  if (semantics == &semFloat8E8M0FN)
+    return true;
----------------
arsenm wrote:

I would assume it's possible to rework the code to handle 0 part count instead of an explicit check. I'm also wondering if this is exotic enough to be its own subclass of APFloatBase instead of pretending it's IEEEfloat since it's reached a new level of degeneracy 

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