[llvm] [SimplifyLibCalls] Recognize and simplify f[min/max]imumnum (PR #170699)

[Yingwei Zheng via llvm-commits]

================
@@ -2543,6 +2543,30 @@ Value *LibCallSimplifier::optimizeFMinFMax(CallInst *CI, IRBuilderBase &B) {
                                                 CI->getArgOperand(1), FMF));
 }
 
+Value *LibCallSimplifier::optimizeFMinimumnumFMaximumnum(CallInst *CI,
+                                                         IRBuilderBase &B) {
+  Module *M = CI->getModule();
+
+  // If we can shrink the call to a float function rather than a double
+  // function, do that first.
+  Function *Callee = CI->getCalledFunction();
+  StringRef Name = Callee->getName();
+  if ((Name == "fminimum_num" || Name == "fmaximum_num") &&
+      hasFloatVersion(M, Name))
+    if (Value *Ret = optimizeBinaryDoubleFP(CI, B, TLI))
+      return Ret;
+
+  // The new fminimum_num/fmaximum_num functions, unlike fmin/fmax, *are*
+  // sensitive to the sigh of zero, so we don't change the fast-math flags like
----------------
dtcxzyw wrote:

```suggestion
  // sensitive to the sign of zero, so we don't change the fast-math flags like
```

https://github.com/llvm/llvm-project/pull/170699