Index: lib/Target/AMDGPU/SIInsertWaits.cpp
===================================================================
--- lib/Target/AMDGPU/SIInsertWaits.cpp
+++ lib/Target/AMDGPU/SIInsertWaits.cpp
@@ -558,14 +558,6 @@
 
     // Wait for everything at the end of the MBB
     Changes |= insertWait(MBB, MBB.getFirstTerminator(), LastIssued);
-
-    // Functions returning something shouldn't contain S_ENDPGM, because other
-    // bytecode will be appended after it.
-    if (!ReturnsVoid) {
-      MachineBasicBlock::iterator I = MBB.getFirstTerminator();
-      if (I != MBB.end() && I->getOpcode() == AMDGPU::S_ENDPGM)
-        I->eraseFromParent();
-    }
   }
 
   return Changes;
Index: lib/Target/AMDGPU/SILowerControlFlow.cpp
===================================================================
--- lib/Target/AMDGPU/SILowerControlFlow.cpp
+++ lib/Target/AMDGPU/SILowerControlFlow.cpp
@@ -486,6 +486,7 @@
   for (MachineFunction::iterator BI = MF.begin(), BE = MF.end();
        BI != BE; ++BI) {
 
+    MachineBasicBlock *EmptyMBBAtEnd = NULL;
     MachineBasicBlock &MBB = *BI;
     MachineBasicBlock::iterator I, Next;
     for (I = MBB.begin(); I != MBB.end(); I = Next) {
@@ -562,6 +563,29 @@
         case AMDGPU::SI_INDIRECT_DST_V16:
           IndirectDst(MI);
           break;
+
+        case AMDGPU::S_ENDPGM: {
+          if (MF.getInfo<SIMachineFunctionInfo>()->returnsVoid())
+            break;
+
+          // Graphics shaders returning non-void shouldn't contain S_ENDPGM,
+          // because external bytecode will be appended at the end.
+          if (BI != --MF.end() || I != MBB.getFirstTerminator()) {
+            // S_ENDPGM is not the last instruction. Add an empty block at
+            // the end and jump there.
+            if (!EmptyMBBAtEnd) {
+              EmptyMBBAtEnd = MF.CreateMachineBasicBlock();
+              MF.insert(MF.end(), EmptyMBBAtEnd);
+            }
+
+            MBB.addSuccessor(EmptyMBBAtEnd);
+            BuildMI(*BI, I, MI.getDebugLoc(), TII->get(AMDGPU::S_BRANCH))
+                    .addMBB(EmptyMBBAtEnd);
+          }
+
+          I->eraseFromParent();
+          break;
+        }
       }
     }
   }
Index: test/CodeGen/AMDGPU/ret_jump.ll
===================================================================
--- /dev/null
+++ test/CodeGen/AMDGPU/ret_jump.ll
@@ -0,0 +1,57 @@
+; RUN: llc -march=amdgcn -mcpu=tahiti -verify-machineinstrs < %s | FileCheck -check-prefix=GCN %s
+; RUN: llc -march=amdgcn -mcpu=tonga -verify-machineinstrs < %s | FileCheck -check-prefix=GCN %s
+
+target triple = "amdgcn--"
+
+; GCN-LABEL: {{^}}main:
+; GCN: BB0_3:
+; GCN-NEXT: s_branch [[LASTBB:BB[0-9]*_[0-9]*]]
+; GCN-NEXT: BB0_
+; GCN: [[LASTBB]]
+; GCN-NEXT: .Lfunc_end0:
+; ModuleID = 'bugpoint-reduced-simplified.bc'
+target triple = "amdgcn--"
+
+define <{ i32, i32, i32, i32, i32, i32, i32, i32, i32, float, float, float, float, float, float, float, float, float, float, float, float, float, float }> @main([9 x <16 x i8>] addrspace(2)* byval, [17 x <16 x i8>] addrspace(2)* byval, [17 x <8 x i32>] addrspace(2)* byval, i32 addrspace(2)* byval, float inreg, i32 inreg, <2 x i32>, <2 x i32>, <2 x i32>, <3 x i32>, <2 x i32>, <2 x i32>, <2 x i32>, float, float, float, float, float, i32, i32, float, i32) #0 {
+main_body:
+  %p83 = call float @llvm.SI.fs.interp(i32 1, i32 0, i32 %5, <2 x i32> %7)
+  %p87 = fmul float undef, %p83
+  %p88 = fadd float %p87, undef
+  %p93 = fadd float %p88, undef
+  %p97 = fmul float %p93, undef
+  %p102 = fsub float %p97, undef
+  %p104 = fmul float %p102, undef
+  %p106 = fadd float 0.000000e+00, %p104
+  %p108 = fadd float undef, %p106
+  br i1 undef, label %ENDIF69, label %ELSE
+
+ELSE:                                             ; preds = %main_body
+  %p124 = fmul float %p108, %p108
+  %p125 = fsub float %p124, undef
+  %p126 = fcmp olt float %p125, 0.000000e+00
+  br i1 %p126, label %ENDIF69, label %ELSE41
+
+ELSE41:                                           ; preds = %ELSE
+  unreachable
+
+ENDIF69:                                          ; preds = %ELSE, %main_body
+  ret <{ i32, i32, i32, i32, i32, i32, i32, i32, i32, float, float, float, float, float, float, float, float, float, float, float, float, float, float }> undef
+}
+
+; Function Attrs: nounwind readnone
+declare float @llvm.SI.load.const(<16 x i8>, i32) #1
+
+; Function Attrs: nounwind readnone
+declare float @llvm.SI.fs.interp(i32, i32, i32, <2 x i32>) #1
+
+; Function Attrs: nounwind readnone
+declare float @llvm.fabs.f32(float) #1
+
+; Function Attrs: nounwind readnone
+declare float @llvm.sqrt.f32(float) #1
+
+; Function Attrs: nounwind readnone
+declare float @llvm.floor.f32(float) #1
+
+attributes #0 = { "InitialPSInputAddr"="36983" "ShaderType"="0" }
+attributes #1 = { nounwind readnone }