[CIR][CIRGen][Builtin] add __builtin_tan (#1502)

Closes #1354.

Author: el-ev

clang/include/clang/CIR/Dialect/IR/CIROps.td

@@ -4752,6 +4752,7 @@ def RoundOp : UnaryFPToFPBuiltinOp<"round", "RoundOp">;
 def RoundEvenOp : UnaryFPToFPBuiltinOp<"roundeven", "RoundEvenOp">;
 def SinOp : UnaryFPToFPBuiltinOp<"sin", "SinOp">;
 def SqrtOp : UnaryFPToFPBuiltinOp<"sqrt", "SqrtOp">;
+def TanOp : UnaryFPToFPBuiltinOp<"tan", "TanOp">;
 def TruncOp : UnaryFPToFPBuiltinOp<"trunc", "FTruncOp">;
 
 def AbsOp : CIR_Op<"abs", [Pure, SameOperandsAndResultType]> {

clang/lib/CIR/CodeGen/CIRGenBuiltin.cpp

@@ -896,7 +896,8 @@ RValue CIRGenFunction::emitBuiltinExpr(const GlobalDecl GD, unsigned BuiltinID,
     case Builtin::BI__builtin_tanf16:
     case Builtin::BI__builtin_tanl:
     case Builtin::BI__builtin_tanf128:
-      llvm_unreachable("Builtin::BItan like NYI");
+      assert(!cir::MissingFeatures::fastMathFlags());
+      return emitUnaryMaybeConstrainedFPBuiltin<cir::TanOp>(*this, *E);
 
     case Builtin::BItanh:
     case Builtin::BItanhf:

clang/test/CIR/CodeGen/builtin-floating-point.c

@@ -1159,6 +1159,74 @@ long double call_sqrtl(long double f) {
   // LLVM: }
 }
 
+// tan
+
+float my_tanf(float f) {
+  return __builtin_tanf(f);
+  // CHECK: cir.func @my_tanf
+  // CHECK: {{.+}} = cir.tan {{.+}} : !cir.float
+
+  // LLVM: define dso_local float @my_tanf(float %0)
+  // LLVM:   %{{.+}} = call float @llvm.tan.f32(float %{{.+}})
+  // LLVM: }
+}
+
+double my_tan(double f) {
+  return __builtin_tan(f);
+  // CHECK: cir.func @my_tan
+  // CHECK: {{.+}} = cir.tan {{.+}} : !cir.double
+
+  // LLVM: define dso_local double @my_tan(double %0)
+  // LLVM:   %{{.+}} = call double @llvm.tan.f64(double %{{.+}})
+  // LLVM: }
+}
+
+long double my_tanl(long double f) {
+  return __builtin_tanl(f);
+  // CHECK: cir.func @my_tanl
+  // CHECK: {{.+}} = cir.tan {{.+}} : !cir.long_double<!cir.f80>
+  // AARCH64: {{.+}} = cir.tan {{.+}} : !cir.long_double<!cir.double>
+
+  // LLVM: define dso_local x86_fp80 @my_tanl(x86_fp80 %0)
+  // LLVM:   %{{.+}} = call x86_fp80 @llvm.tan.f80(x86_fp80 %{{.+}})
+  // LLVM: }
+}
+
+float tanf(float);
+double tan(double);
+long double tanl(long double);
+
+float call_tanf(float f) {
+  return tanf(f);
+  // CHECK: cir.func @call_tanf
+  // CHECK: {{.+}} = cir.tan {{.+}} : !cir.float
+
+  // LLVM: define dso_local float @call_tanf(float %0)
+  // LLVM:   %{{.+}} = call float @llvm.tan.f32(float %{{.+}})
+  // LLVM: }
+}
+
+double call_tan(double f) {
+  return tan(f);
+  // CHECK: cir.func @call_tan
+  // CHECK: {{.+}} = cir.tan {{.+}} : !cir.double
+
+  // LLVM: define dso_local double @call_tan(double %0)
+  // LLVM:   %{{.+}} = call double @llvm.tan.f64(double %{{.+}})
+  // LLVM: }
+}
+
+long double call_tanl(long double f) {
+  return tanl(f);
+  // CHECK: cir.func @call_tanl
+  // CHECK: {{.+}} = cir.tan {{.+}} : !cir.long_double<!cir.f80>
+  // AARCH64: {{.+}} = cir.tan {{.+}} : !cir.long_double<!cir.double>
+
+  // LLVM: define dso_local x86_fp80 @call_tanl(x86_fp80 %0)
+  // LLVM:   %{{.+}} = call x86_fp80 @llvm.tan.f80(x86_fp80 %{{.+}})
+  // LLVM: }
+}
+
 // trunc
 
 float my_truncf(float f) {