[clang] [Clang][OpenCL][AMDGPU] Allow a kernel to call another kernel (PR #115821)

[Aniket Lal via cfe-commits]

================
@@ -6138,6 +6150,17 @@ void CodeGenModule::EmitGlobalFunctionDefinition(GlobalDecl GD,
   CodeGenFunction(*this).GenerateCode(GD, Fn, FI);
 
   setNonAliasAttributes(GD, Fn);
+
+  if (D->hasAttr<OpenCLKernelAttr>()) {
+    if (GD.getKernelReferenceKind() == KernelReferenceKind::Stub) {
+      if (Fn->hasFnAttribute(llvm::Attribute::NoInline))
+        Fn->removeFnAttr(llvm::Attribute::NoInline);
+      if (Fn->hasFnAttribute(llvm::Attribute::InlineHint))
+        Fn->removeFnAttr(llvm::Attribute::InlineHint);
+      Fn->addFnAttr(llvm::Attribute::AlwaysInline);
+    }
+  }
----------------
lalaniket8 wrote:

I see, yes it seems that adding the alwaysinline attribute is not necessary. I have removed it in the latest commit and made required changes in the test files.

https://github.com/llvm/llvm-project/pull/115821