diff --git a/clang/lib/CodeGen/Targets/X86.cpp b/clang/lib/CodeGen/Targets/X86.cpp
index 7ee8a0dc2a0db..c03ba9487a6dc 100644
--- a/clang/lib/CodeGen/Targets/X86.cpp
+++ b/clang/lib/CodeGen/Targets/X86.cpp
@@ -1513,12 +1513,18 @@ static void initFeatureMaps(const ASTContext &Ctx,
 
 static bool checkAVXParamFeature(DiagnosticsEngine &Diag,
                                  SourceLocation CallLoc,
+                                 const FunctionDecl &Callee,
                                  const llvm::StringMap<bool> &CallerMap,
                                  const llvm::StringMap<bool> &CalleeMap,
                                  QualType Ty, StringRef Feature,
                                  bool IsArgument) {
   bool CallerHasFeat = CallerMap.lookup(Feature);
   bool CalleeHasFeat = CalleeMap.lookup(Feature);
+  // No explicit features and the function is internal, be permissive.
+  if (!CallerHasFeat && !CalleeHasFeat &&
+      (!Callee.isExternallyVisible() || Callee.hasAttr<AlwaysInlineAttr>()))
+    return false;
+
   if (!CallerHasFeat && !CalleeHasFeat)
     return Diag.Report(CallLoc, diag::warn_avx_calling_convention)
            << IsArgument << Ty << Feature;
@@ -1534,18 +1540,18 @@ static bool checkAVXParamFeature(DiagnosticsEngine &Diag,
 }
 
 static bool checkAVXParam(DiagnosticsEngine &Diag, ASTContext &Ctx,
-                          SourceLocation CallLoc,
+                          SourceLocation CallLoc, const FunctionDecl &Callee,
                           const llvm::StringMap<bool> &CallerMap,
                           const llvm::StringMap<bool> &CalleeMap, QualType Ty,
                           bool IsArgument) {
   uint64_t Size = Ctx.getTypeSize(Ty);
   if (Size > 256)
-    return checkAVXParamFeature(Diag, CallLoc, CallerMap, CalleeMap, Ty,
+    return checkAVXParamFeature(Diag, CallLoc, Callee, CallerMap, CalleeMap, Ty,
                                 "avx512f", IsArgument);
 
   if (Size > 128)
-    return checkAVXParamFeature(Diag, CallLoc, CallerMap, CalleeMap, Ty, "avx",
-                                IsArgument);
+    return checkAVXParamFeature(Diag, CallLoc, Callee, CallerMap, CalleeMap, Ty,
+                                "avx", IsArgument);
 
   return false;
 }
@@ -1582,8 +1588,8 @@ void X86_64TargetCodeGenInfo::checkFunctionCallABI(CodeGenModule &CGM,
       if (ArgIndex < Callee->getNumParams())
         Ty = Callee->getParamDecl(ArgIndex)->getType();
 
-      if (checkAVXParam(CGM.getDiags(), CGM.getContext(), CallLoc, CallerMap,
-                        CalleeMap, Ty, /*IsArgument*/ true))
+      if (checkAVXParam(CGM.getDiags(), CGM.getContext(), CallLoc, *Callee,
+                        CallerMap, CalleeMap, Ty, /*IsArgument*/ true))
         return;
     }
     ++ArgIndex;
@@ -1594,7 +1600,7 @@ void X86_64TargetCodeGenInfo::checkFunctionCallABI(CodeGenModule &CGM,
   if (Callee->getReturnType()->isVectorType() &&
       CGM.getContext().getTypeSize(Callee->getReturnType()) > 128) {
     initFeatureMaps(CGM.getContext(), CallerMap, Caller, CalleeMap, Callee);
-    checkAVXParam(CGM.getDiags(), CGM.getContext(), CallLoc, CallerMap,
+    checkAVXParam(CGM.getDiags(), CGM.getContext(), CallLoc, *Callee, CallerMap,
                   CalleeMap, Callee->getReturnType(),
                   /*IsArgument*/ false);
   }
diff --git a/clang/test/CodeGen/target-builtin-error-3.c b/clang/test/CodeGen/target-builtin-error-3.c
index 3de76e253d9b1..056dc940f7a93 100644
--- a/clang/test/CodeGen/target-builtin-error-3.c
+++ b/clang/test/CodeGen/target-builtin-error-3.c
@@ -24,6 +24,26 @@ static inline half16 __attribute__((__overloadable__)) convert_half( float16 a )
 }
 void avx_test( uint16_t *destData, float16 argbF)
 {
-  // expected-warning@+1{{AVX vector argument of type 'float16' (vector of 16 'float' values) without 'avx512f' enabled changes the ABI}}
   ((half16U *)destData)[0] = convert_half(argbF);
 }
+
+half16 test( float16 a ) {
+  half16 r;
+  r.lo = convert_half(a.lo);
+  return r;
+}
+void avx_test2( uint16_t *destData, float16 argbF)
+{
+  // expected-warning@+1{{AVX vector argument of type 'float16' (vector of 16 'float' values) without 'avx512f' enabled changes the ABI}}
+  ((half16U *)destData)[0] = test(argbF);
+}
+
+__attribute__((always_inline)) half16 test2( float16 a ) {
+  half16 r;
+  r.lo = convert_half(a.lo);
+  return r;
+}
+void avx_test3( uint16_t *destData, float16 argbF)
+{
+  ((half16U *)destData)[0] = test2(argbF);
+}