Reland [NFC-I] Remove hack for fp-classification builtins

The FP-classification builtins (__builtin_isfinite, etc) use variadic
packs in the definition file to mean an overload set.  Because of that,
floats were converted to doubles, which is incorrect. There WAS a patch
to remove the cast after the fact.

THis patch switches these builtins to just be custom type checking,
calls the implicit conversions for the integer members, and makes sure
the correct L->R casts are put into place, then does type checking like
normal.

A future direction (that wouldn't be NFC) would consider making
conversions for the floating point parameter legal.

Note: The initial patch for this missed that certain systems need to
still convert half to float, since they dont' support that type.

1 files changed, 11 insertions, 1 deletions

diff --git a/clang/test/CodeGen/builtin_float.c b/clang/test/CodeGen/builtin_float.c
index b97cf1be348..b257854278c 100644
--- a/clang/test/CodeGen/builtin_float.c
+++ b/clang/test/CodeGen/builtin_float.c
@@ -1,4 +1,5 @@
-// RUN: %clang_cc1 -emit-llvm -o - %s | FileCheck %s
+// RUN: %clang_cc1 -emit-llvm -triple x86_64-windows-pc -o - %s | FileCheck %s --check-prefixes=CHECK,FP16
+// RUN: %clang_cc1 -emit-llvm -triple ppc64-be -o - %s -DNO_FP16 | FileCheck %s --check-prefixes=CHECK,NOFP16
 
 // test to ensure that these builtins don't do the variadic promotion of float->double.
 void test_floats(float f1, float f2) {
@@ -43,6 +44,15 @@ void test_doubles(double d1, double f2) {
   // CHECK-NEXT: zext i1
 }
 
+void test_half(__fp16 *H, __fp16 *H2) {
+  (void)__builtin_isgreater(*H, *H2);
+  // CHECK: fcmp ogt float
+  // CHECK-NEXT: zext i1
+  (void)__builtin_isinf(*H);
+  // NOFP16: fcmp oeq float %{{.+}}, 0x7FF
+  // FP16: fcmp oeq half %{{.+}}, 0xH7C
+}
+
 void test_mixed(double d1, float f2) {
   (void)__builtin_isgreater(d1, f2);
   // CHECK: fpext float {{.*}} to double