[llvm] [APFloat] Add APFloat::format (PR #99088)

[Matt Arsenault via llvm-commits]

================
@@ -5217,6 +5249,50 @@ bool DoubleAPFloat::getExactInverse(APFloat *inv) const {
   return Ret;
 }
 
+bool DoubleAPFloat::decomposeToIntegerAndPowerOf2(APInt &I, int &exp) const {
+  APInt a, b;
+  int aExp, bExp;
+  bool aOK = Floats[0].decomposeToIntegerAndPowerOf2(a, aExp);
+  bool bOK = Floats[1].decomposeToIntegerAndPowerOf2(b, bExp);
+
+  if (aOK && bOK) {
+    if (aExp > bExp) {
+      int deltaExp = aExp - bExp;
+      // shift a left to reduce its exponent to be the same as b
----------------
arsenm wrote:

```suggestion
      // Shift a left to reduce its exponent to be the same as b
```

https://github.com/llvm/llvm-project/pull/99088