diff --git a/llvm/lib/Target/AMDGPU/AMDGPUHSAMetadataStreamer.cpp b/llvm/lib/Target/AMDGPU/AMDGPUHSAMetadataStreamer.cpp
--- a/llvm/lib/Target/AMDGPU/AMDGPUHSAMetadataStreamer.cpp
+++ b/llvm/lib/Target/AMDGPU/AMDGPUHSAMetadataStreamer.cpp
@@ -798,7 +798,8 @@
   if (!HiddenArgNumBytes)
     return;
 
-  auto &DL = Func.getParent()->getDataLayout();
+  const Module *M = Func.getParent();
+  auto &DL = M->getDataLayout();
   auto Int64Ty = Type::getInt64Ty(Func.getContext());
 
   if (HiddenArgNumBytes >= 8)
@@ -814,16 +815,18 @@
   auto Int8PtrTy =
       Type::getInt8PtrTy(Func.getContext(), AMDGPUAS::GLOBAL_ADDRESS);
 
-  // Emit "printf buffer" argument if printf is used, otherwise emit dummy
-  // "none" argument.
+  // Emit "hidden_printf_buffer" argument if printf is used,
+  // "hidden_hostcall_buffer" if hostcalls are used,
+  // otherwise emit dummy "none" argument.  
   if (HiddenArgNumBytes >= 32) {
-    if (Func.getParent()->getNamedMetadata("llvm.printf.fmts"))
+    if (M->getNamedMetadata("llvm.printf.fmts"))
       emitKernelArg(DL, Int8PtrTy, Align(8), "hidden_printf_buffer", Offset,
                     Args);
-    else if (Func.getParent()->getFunction("__ockl_hostcall_internal")) {
+    else if ([&](){ for (auto &F: *M) if(F.hasFnAttribute("amdgpu-hostcalls")) {
+                                      return true; break; } return false;}() ) {
       // The printf runtime binding pass should have ensured that hostcall and
       // printf are not used in the same module.
-      assert(!Func.getParent()->getNamedMetadata("llvm.printf.fmts"));
+      assert(!M->getNamedMetadata("llvm.printf.fmts"));
       emitKernelArg(DL, Int8PtrTy, Align(8), "hidden_hostcall_buffer", Offset,
                     Args);
     } else
diff --git a/llvm/lib/Transforms/Utils/AMDGPUEmitPrintf.cpp b/llvm/lib/Transforms/Utils/AMDGPUEmitPrintf.cpp
--- a/llvm/lib/Transforms/Utils/AMDGPUEmitPrintf.cpp
+++ b/llvm/lib/Transforms/Utils/AMDGPUEmitPrintf.cpp
@@ -63,7 +63,11 @@
   auto Int64Ty = Builder.getInt64Ty();
   auto M = Builder.GetInsertBlock()->getModule();
   auto Fn = M->getOrInsertFunction("__ockl_printf_begin", Int64Ty, Int64Ty);
-  return Builder.CreateCall(Fn, Version);
+  auto Call = Builder.CreateCall(Fn, Version);
+  auto Bb = Call->getParent();
+  assert(Bb && Bb->getParent());
+  Bb->getParent()->addFnAttr("amdgpu-hostcalls");
+  return Call;
 }
 
 static Value *callAppendArgs(IRBuilder<> &Builder, Value *Desc, int NumArgs,
diff --git a/llvm/test/CodeGen/AMDGPU/hsa-metadata-hostcall-present-v3.ll b/llvm/test/CodeGen/AMDGPU/hsa-metadata-hostcall-present-v3.ll
--- a/llvm/test/CodeGen/AMDGPU/hsa-metadata-hostcall-present-v3.ll
+++ b/llvm/test/CodeGen/AMDGPU/hsa-metadata-hostcall-present-v3.ll
@@ -41,7 +41,7 @@
 ; CHECK-NEXT: - 1
 ; CHECK-NEXT: - 0
 
-attributes #0 = { "amdgpu-implicitarg-num-bytes"="48" }
+attributes #0 = { "amdgpu-implicitarg-num-bytes"="48" "amdgpu-hostcalls"}
 
 !1 = !{i32 0}
 !2 = !{!"none"}