[CodeGen] __builtin_sqrt should map to the compiler's intrinsic sqrt function

[spatel]

Summary

This patch is intended to answer a question raised in PR27435:
https://bugs.llvm.org/show_bug.cgi?id=27435

Is a programmer using __builtin_sqrt() invoking the compiler's intrinsic definition of sqrt or the mathlib definition of sqrt?

I know there are follow-up bugs that still need to be solved (D28335), but I think we should answer this first. Also if this patch is correct, it is enough to produce the hoped-for result in the PR27435 examples because we'll get this IR at -O2:

$ ./clang -O2 27435.c -S -o - -emit-llvm | grep llvm.sqrt

%2 = call <2 x double> @llvm.sqrt.v2f64(<2 x double> %1)
%2 = call <4 x float> @llvm.sqrt.v4f32(<4 x float> %1)

[spatel]

Oops - I wrongly made llvm-commits a subscriber rather than cfe-commits. Let me know if I should reopen under a new thread to get the patch to hit the right mailing list.

[hfinkel]

In D39160#902918, @spatel wrote:

Oops - I wrongly made llvm-commits a subscriber rather than cfe-commits. Let me know if I should reopen under a new thread to get the patch to hit the right mailing list.

Yes, please reopen.

[spatel]

Reincarnated as D39204.

lib/CodeGen/CGBuiltin.cpp

RValue CodeGenFunction::EmitBuiltinExpr(const FunctionDecl *FD,
   case Builtin::BI__builtin_nearbyintl: {
     return RValue::get(emitUnaryBuiltin(*this, E, Intrinsic::nearbyint));
   }
   case Builtin::BI__builtin_round:
   case Builtin::BI__builtin_roundf:
   case Builtin::BI__builtin_roundl: {
     return RValue::get(emitUnaryBuiltin(*this, E, Intrinsic::round));
   }
+
+  case Builtin::BI__builtin_sqrt:
+  case Builtin::BI__builtin_sqrtf:
+  case Builtin::BI__builtin_sqrtl:
+    return RValue::get(emitUnaryBuiltin(*this, E, Intrinsic::sqrt));
+
   case Builtin::BI__builtin_fmin:
   case Builtin::BI__builtin_fminf:
   case Builtin::BI__builtin_fminl: {
     return RValue::get(emitBinaryBuiltin(*this, E, Intrinsic::minnum));
   }
   case Builtin::BI__builtin_fmax:
   case Builtin::BI__builtin_fmaxf:
   case Builtin::BI__builtin_fmaxl: {

test/CodeGen/builtins.c

void test_float_builtin_ops(float F, double D, long double LD) {
 
   resd = __builtin_floor(D);
   // CHECK: call double @llvm.floor.f64
 
   resld = __builtin_floorl(LD);
   // CHECK: call x86_fp80 @llvm.floor.f80
 
   resf = __builtin_sqrtf(F);
-  // CHECK: call float @sqrtf(
+  // CHECK: call float @llvm.sqrt.f32
 
   resd = __builtin_sqrt(D);
-  // CHECK: call double @sqrt(
+  // CHECK: call double @llvm.sqrt.f64
 
   resld = __builtin_sqrtl(LD);
-  // CHECK: call x86_fp80 @sqrtl(
+  // CHECK: call x86_fp80 @llvm.sqrt.f80
 
   resf = __builtin_truncf(F);
   // CHECK: call float @llvm.trunc.f32
 
   resd = __builtin_trunc(D);
   // CHECK: call double @llvm.trunc.f64
 
   resld = __builtin_truncl(LD);