Index: lib/Target/AMDGPU/AMDGPUAnnotateKernelFeatures.cpp
===================================================================
--- lib/Target/AMDGPU/AMDGPUAnnotateKernelFeatures.cpp
+++ lib/Target/AMDGPU/AMDGPUAnnotateKernelFeatures.cpp
@@ -235,6 +235,26 @@
           if (!CS.isInlineAsm())
             HaveCall = true;
           continue;
+        }      
+       
+        CallingConv::ID CC = F.getCallingConv();
+        if (ST.isAmdHsaOS()) {
+          if (CC == CallingConv::AMDGPU_KERNEL) {
+            //Check for uniform workgroup size attribute
+            if (F.hasFnAttribute("uniform-work-group-size")) {
+              if (Callee->hasFnAttribute("uniform-work-group-size")) {
+                Attribute attr = Callee->getFnAttribute("uniform-work-group-size");  
+                if (attr.getValueAsString().equals("false"))
+                  continue;
+                else if (attr.getValueAsString().equals("true"))
+                  Callee->addFnAttr(F.getFnAttribute("uniform-work-group-size"));
+              }
+              else {
+                //Copy the kernel attribute to the function 
+                Callee->addFnAttr(F.getFnAttribute("uniform-work-group-size"));
+              }
+            }
+          }
         }
 
         Intrinsic::ID IID = Callee->getIntrinsicID();
Index: lib/Target/AMDGPU/AMDGPUTargetMachine.cpp
===================================================================
--- lib/Target/AMDGPU/AMDGPUTargetMachine.cpp
+++ lib/Target/AMDGPU/AMDGPUTargetMachine.cpp
@@ -667,6 +667,10 @@
 }
 
 void AMDGPUPassConfig::addCodeGenPrepare() {
+
+  if (TM->getTargetTriple().getArch() == Triple::amdgcn )
+    addPass(createAMDGPUAnnotateKernelFeaturesPass());
+
   if (TM->getTargetTriple().getArch() == Triple::amdgcn &&
       EnableLowerKernelArguments)
     addPass(createAMDGPULowerKernelArgumentsPass());
@@ -749,7 +753,7 @@
 
   // FIXME: We need to run a pass to propagate the attributes when calls are
   // supported.
-  addPass(createAMDGPUAnnotateKernelFeaturesPass());
+  //addPass(createAMDGPUAnnotateKernelFeaturesPass());
 
   // Merge divergent exit nodes. StructurizeCFG won't recognize the multi-exit
   // regions formed by them.