diff --git a/llvm/docs/AMDGPUUsage.rst b/llvm/docs/AMDGPUUsage.rst
--- a/llvm/docs/AMDGPUUsage.rst
+++ b/llvm/docs/AMDGPUUsage.rst
@@ -6968,6 +6968,25 @@
   the top 32 bits of the pipeline, so the shader may use the program
   counter's top 32 bits.
 
+.. _pal_call-convention:
+
+Call Convention
+~~~~~~~~~~~~~~~
+
+For graphics use cases, the calling convention is `amdgpu_gfx`.
+
+.. note::
+
+  `amdgpu_gfx` Function calls are currently in development and are
+  subject to major changes.
+
+This calling convention shares most properties with calling non-kernel
+functions (see
+:ref:`amdgpu-amdhsa-function-call-convention-non-kernel-functions`).
+Differences are:
+
+ - Currently there are none, differences will be listed here
+
 Unspecified OS
 --------------
 
diff --git a/llvm/include/llvm/IR/CallingConv.h b/llvm/include/llvm/IR/CallingConv.h
--- a/llvm/include/llvm/IR/CallingConv.h
+++ b/llvm/include/llvm/IR/CallingConv.h
@@ -241,6 +241,9 @@
     /// The remainder matches the regular calling convention.
     WASM_EmscriptenInvoke = 99,
 
+    /// Calling convention used for AMD graphics targets.
+    AMDGPU_Gfx = 100,
+
     /// The highest possible calling convention ID. Must be some 2^k - 1.
     MaxID = 1023
   };
diff --git a/llvm/lib/AsmParser/LLLexer.cpp b/llvm/lib/AsmParser/LLLexer.cpp
--- a/llvm/lib/AsmParser/LLLexer.cpp
+++ b/llvm/lib/AsmParser/LLLexer.cpp
@@ -624,6 +624,7 @@
   KEYWORD(amdgpu_ps);
   KEYWORD(amdgpu_cs);
   KEYWORD(amdgpu_kernel);
+  KEYWORD(amdgpu_gfx);
   KEYWORD(tailcc);
 
   KEYWORD(cc);
diff --git a/llvm/lib/AsmParser/LLParser.cpp b/llvm/lib/AsmParser/LLParser.cpp
--- a/llvm/lib/AsmParser/LLParser.cpp
+++ b/llvm/lib/AsmParser/LLParser.cpp
@@ -2086,6 +2086,9 @@
   case lltok::kw_hhvm_ccc:       CC = CallingConv::HHVM_C; break;
   case lltok::kw_cxx_fast_tlscc: CC = CallingConv::CXX_FAST_TLS; break;
   case lltok::kw_amdgpu_vs:      CC = CallingConv::AMDGPU_VS; break;
+  case lltok::kw_amdgpu_gfx:
+    CC = CallingConv::AMDGPU_Gfx;
+    break;
   case lltok::kw_amdgpu_ls:      CC = CallingConv::AMDGPU_LS; break;
   case lltok::kw_amdgpu_hs:      CC = CallingConv::AMDGPU_HS; break;
   case lltok::kw_amdgpu_es:      CC = CallingConv::AMDGPU_ES; break;
diff --git a/llvm/lib/AsmParser/LLToken.h b/llvm/lib/AsmParser/LLToken.h
--- a/llvm/lib/AsmParser/LLToken.h
+++ b/llvm/lib/AsmParser/LLToken.h
@@ -170,6 +170,7 @@
   kw_amdgpu_ps,
   kw_amdgpu_cs,
   kw_amdgpu_kernel,
+  kw_amdgpu_gfx,
   kw_tailcc,
 
   // Attributes:
diff --git a/llvm/lib/IR/AsmWriter.cpp b/llvm/lib/IR/AsmWriter.cpp
--- a/llvm/lib/IR/AsmWriter.cpp
+++ b/llvm/lib/IR/AsmWriter.cpp
@@ -388,6 +388,9 @@
   case CallingConv::AMDGPU_PS:     Out << "amdgpu_ps"; break;
   case CallingConv::AMDGPU_CS:     Out << "amdgpu_cs"; break;
   case CallingConv::AMDGPU_KERNEL: Out << "amdgpu_kernel"; break;
+  case CallingConv::AMDGPU_Gfx:
+    Out << "amdgpu_gfx";
+    break;
   }
 }
 
diff --git a/llvm/lib/Target/AMDGPU/AMDGPUAsmPrinter.cpp b/llvm/lib/Target/AMDGPU/AMDGPUAsmPrinter.cpp
--- a/llvm/lib/Target/AMDGPU/AMDGPUAsmPrinter.cpp
+++ b/llvm/lib/Target/AMDGPU/AMDGPUAsmPrinter.cpp
@@ -456,7 +456,7 @@
     Info = analyzeResourceUsage(MF);
   }
 
-  if (STM.isAmdPalOS())
+  if (STM.isAmdPalOS() && MFI->isEntryFunction())
     EmitPALMetadata(MF, CurrentProgramInfo);
   else if (!STM.isAmdHsaOS()) {
     EmitProgramInfoSI(MF, CurrentProgramInfo);
diff --git a/llvm/lib/Target/AMDGPU/AMDGPUCallLowering.cpp b/llvm/lib/Target/AMDGPU/AMDGPUCallLowering.cpp
--- a/llvm/lib/Target/AMDGPU/AMDGPUCallLowering.cpp
+++ b/llvm/lib/Target/AMDGPU/AMDGPUCallLowering.cpp
@@ -459,9 +459,9 @@
   assert(!Val == VRegs.empty() && "Return value without a vreg");
 
   CallingConv::ID CC = B.getMF().getFunction().getCallingConv();
-  const bool IsShader = AMDGPU::isShader(CC);
-  const bool IsWaveEnd = (IsShader && MFI->returnsVoid()) ||
-                         AMDGPU::isKernel(CC);
+  const bool IsGraphics = AMDGPU::isGraphics(CC);
+  const bool IsWaveEnd =
+      (IsGraphics && MFI->returnsVoid()) || AMDGPU::isKernel(CC);
   if (IsWaveEnd) {
     B.buildInstr(AMDGPU::S_ENDPGM)
       .addImm(0);
@@ -471,7 +471,7 @@
   auto const &ST = MF.getSubtarget<GCNSubtarget>();
 
   unsigned ReturnOpc =
-      IsShader ? AMDGPU::SI_RETURN_TO_EPILOG : AMDGPU::S_SETPC_B64_return;
+      IsGraphics ? AMDGPU::SI_RETURN_TO_EPILOG : AMDGPU::S_SETPC_B64_return;
 
   auto Ret = B.buildInstrNoInsert(ReturnOpc);
   Register ReturnAddrVReg;
@@ -785,7 +785,7 @@
   if (CC == CallingConv::AMDGPU_KERNEL)
     return lowerFormalArgumentsKernel(B, F, VRegs);
 
-  const bool IsShader = AMDGPU::isShader(CC);
+  const bool IsGraphics = AMDGPU::isGraphics(CC);
   const bool IsEntryFunc = AMDGPU::isEntryFunctionCC(CC);
 
   MachineFunction &MF = B.getMF();
@@ -826,7 +826,7 @@
     const bool InReg = Arg.hasAttribute(Attribute::InReg);
 
     // SGPR arguments to functions not implemented.
-    if (!IsShader && InReg)
+    if (!IsGraphics && InReg)
       return false;
 
     if (Arg.hasAttribute(Attribute::SwiftSelf) ||
@@ -937,7 +937,7 @@
 
   // Start adding system SGPRs.
   if (IsEntryFunc) {
-    TLI.allocateSystemSGPRs(CCInfo, MF, *Info, CC, IsShader);
+    TLI.allocateSystemSGPRs(CCInfo, MF, *Info, CC, IsGraphics);
   } else {
     CCInfo.AllocateReg(Info->getScratchRSrcReg());
     TLI.allocateSpecialInputSGPRs(CCInfo, MF, *TRI, *Info);
diff --git a/llvm/lib/Target/AMDGPU/AMDGPUCallingConv.td b/llvm/lib/Target/AMDGPU/AMDGPUCallingConv.td
--- a/llvm/lib/Target/AMDGPU/AMDGPUCallingConv.td
+++ b/llvm/lib/Target/AMDGPU/AMDGPUCallingConv.td
@@ -16,7 +16,75 @@
   : CCIf<"ArgFlags.isSExt() || ArgFlags.isZExt()", A>;
 
 // Calling convention for SI
-def CC_SI : CallingConv<[
+def CC_SI_Gfx : CallingConv<[
+  // 0-3 are reserved for the stack buffer descriptor
+  // 30-31 are reserved for the return address
+  // 32 is reserved for the stack pointer
+  CCIfInReg<CCIfType<[f32, i32, f16, i16, v2i16, v2f16] , CCAssignToReg<[
+    SGPR4, SGPR5, SGPR6, SGPR7,
+    SGPR8, SGPR9, SGPR10, SGPR11, SGPR12, SGPR13, SGPR14, SGPR15,
+    SGPR16, SGPR17, SGPR18, SGPR19, SGPR20, SGPR21, SGPR22, SGPR23,
+    SGPR24, SGPR25, SGPR26, SGPR27, SGPR28, SGPR29,
+  ]>>>,
+
+  CCIfNotInReg<CCIfType<[f32, i32, f16, i16, v2i16, v2f16] , CCAssignToReg<[
+    VGPR0, VGPR1, VGPR2, VGPR3, VGPR4, VGPR5, VGPR6, VGPR7,
+    VGPR8, VGPR9, VGPR10, VGPR11, VGPR12, VGPR13, VGPR14, VGPR15,
+    VGPR16, VGPR17, VGPR18, VGPR19, VGPR20, VGPR21, VGPR22, VGPR23,
+    VGPR24, VGPR25, VGPR26, VGPR27, VGPR28, VGPR29, VGPR30, VGPR31
+  ]>>>,
+
+  CCIfType<[i32, f32, v2i16, v2f16, i16, f16, i1], CCAssignToStack<4, 4>>,
+  CCIfType<[i64, f64, v2i32, v2f32], CCAssignToStack<8, 4>>,
+  CCIfType<[v3i32, v3f32], CCAssignToStack<12, 4>>,
+  CCIfType<[v4i32, v4f32, v2i64, v2f64], CCAssignToStack<16, 4>>,
+  CCIfType<[v5i32, v5f32], CCAssignToStack<20, 4>>,
+  CCIfType<[v8i32, v8f32], CCAssignToStack<32, 4>>,
+  CCIfType<[v16i32, v16f32], CCAssignToStack<64, 4>>
+]>;
+
+def RetCC_SI_Gfx : CallingConv<[
+  // 0-3 are reserved for the stack buffer descriptor
+  // 32 is reserved for the stack pointer
+  CCIfInReg<CCIfType<[f32, i32, f16, i16, v2i16, v2f16] , CCAssignToReg<[
+    SGPR4, SGPR5, SGPR6, SGPR7,
+    SGPR8, SGPR9, SGPR10, SGPR11, SGPR12, SGPR13, SGPR14, SGPR15,
+    SGPR16, SGPR17, SGPR18, SGPR19, SGPR20, SGPR21, SGPR22, SGPR23,
+    SGPR24, SGPR25, SGPR26, SGPR27, SGPR28, SGPR29, SGPR30, SGPR31,
+    SGPR33, SGPR34, SGPR35, SGPR36, SGPR37, SGPR38, SGPR39,
+    SGPR40, SGPR41, SGPR42, SGPR43
+  ]>>>,
+
+  CCIfNotInReg<CCIfType<[f32, i32, f16, i16, v2i16, v2f16] , CCAssignToReg<[
+    VGPR0, VGPR1, VGPR2, VGPR3, VGPR4, VGPR5, VGPR6, VGPR7,
+    VGPR8, VGPR9, VGPR10, VGPR11, VGPR12, VGPR13, VGPR14, VGPR15,
+    VGPR16, VGPR17, VGPR18, VGPR19, VGPR20, VGPR21, VGPR22, VGPR23,
+    VGPR24, VGPR25, VGPR26, VGPR27, VGPR28, VGPR29, VGPR30, VGPR31,
+    VGPR32, VGPR33, VGPR34, VGPR35, VGPR36, VGPR37, VGPR38, VGPR39,
+    VGPR40, VGPR41, VGPR42, VGPR43, VGPR44, VGPR45, VGPR46, VGPR47,
+    VGPR48, VGPR49, VGPR50, VGPR51, VGPR52, VGPR53, VGPR54, VGPR55,
+    VGPR56, VGPR57, VGPR58, VGPR59, VGPR60, VGPR61, VGPR62, VGPR63,
+    VGPR64, VGPR65, VGPR66, VGPR67, VGPR68, VGPR69, VGPR70, VGPR71,
+    VGPR72, VGPR73, VGPR74, VGPR75, VGPR76, VGPR77, VGPR78, VGPR79,
+    VGPR80, VGPR81, VGPR82, VGPR83, VGPR84, VGPR85, VGPR86, VGPR87,
+    VGPR88, VGPR89, VGPR90, VGPR91, VGPR92, VGPR93, VGPR94, VGPR95,
+    VGPR96, VGPR97, VGPR98, VGPR99, VGPR100, VGPR101, VGPR102, VGPR103,
+    VGPR104, VGPR105, VGPR106, VGPR107, VGPR108, VGPR109, VGPR110, VGPR111,
+    VGPR112, VGPR113, VGPR114, VGPR115, VGPR116, VGPR117, VGPR118, VGPR119,
+    VGPR120, VGPR121, VGPR122, VGPR123, VGPR124, VGPR125, VGPR126, VGPR127,
+    VGPR128, VGPR129, VGPR130, VGPR131, VGPR132, VGPR133, VGPR134, VGPR135
+  ]>>>,
+
+  CCIfType<[i32, f32, v2i16, v2f16, i16, f16, i1], CCAssignToStack<4, 4>>,
+  CCIfType<[i64, f64, v2i32, v2f32], CCAssignToStack<8, 4>>,
+  CCIfType<[v3i32, v3f32], CCAssignToStack<12, 4>>,
+  CCIfType<[v4i32, v4f32, v2i64, v2f64], CCAssignToStack<16, 4>>,
+  CCIfType<[v5i32, v5f32], CCAssignToStack<20, 4>>,
+  CCIfType<[v8i32, v8f32], CCAssignToStack<32, 4>>,
+  CCIfType<[v16i32, v16f32], CCAssignToStack<64, 4>>
+]>;
+
+def CC_SI_SHADER : CallingConv<[
 
   CCIfInReg<CCIfType<[f32, i32, f16, i16, v2i16, v2f16] , CCAssignToReg<[
     SGPR0, SGPR1, SGPR2, SGPR3, SGPR4, SGPR5, SGPR6, SGPR7,
@@ -159,7 +227,7 @@
    CCIf<"static_cast<const GCNSubtarget&>"
          "(State.getMachineFunction().getSubtarget()).getGeneration() >= "
            "AMDGPUSubtarget::SOUTHERN_ISLANDS",
-        CCDelegateTo<CC_SI>>,
+        CCDelegateTo<CC_SI_SHADER>>,
    CCIf<"static_cast<const GCNSubtarget&>"
          "(State.getMachineFunction().getSubtarget()).getGeneration() >= "
            "AMDGPUSubtarget::SOUTHERN_ISLANDS && State.getCallingConv() == CallingConv::C",
diff --git a/llvm/lib/Target/AMDGPU/AMDGPUISelLowering.cpp b/llvm/lib/Target/AMDGPU/AMDGPUISelLowering.cpp
--- a/llvm/lib/Target/AMDGPU/AMDGPUISelLowering.cpp
+++ b/llvm/lib/Target/AMDGPU/AMDGPUISelLowering.cpp
@@ -939,6 +939,8 @@
   case CallingConv::Fast:
   case CallingConv::Cold:
     return CC_AMDGPU_Func;
+  case CallingConv::AMDGPU_Gfx:
+    return CC_SI_Gfx;
   case CallingConv::AMDGPU_KERNEL:
   case CallingConv::SPIR_KERNEL:
   default:
@@ -960,6 +962,8 @@
   case CallingConv::AMDGPU_ES:
   case CallingConv::AMDGPU_LS:
     return RetCC_SI_Shader;
+  case CallingConv::AMDGPU_Gfx:
+    return RetCC_SI_Gfx;
   case CallingConv::C:
   case CallingConv::Fast:
   case CallingConv::Cold:
diff --git a/llvm/lib/Target/AMDGPU/AMDGPUTargetTransformInfo.h b/llvm/lib/Target/AMDGPU/AMDGPUTargetTransformInfo.h
--- a/llvm/lib/Target/AMDGPU/AMDGPUTargetTransformInfo.h
+++ b/llvm/lib/Target/AMDGPU/AMDGPUTargetTransformInfo.h
@@ -76,7 +76,7 @@
   const GCNSubtarget *ST;
   const SITargetLowering *TLI;
   AMDGPUTTIImpl CommonTTI;
-  bool IsGraphicsShader;
+  bool IsGraphics;
   bool HasFP32Denormals;
   bool HasFP64FP16Denormals;
   unsigned MaxVGPRs;
@@ -137,7 +137,7 @@
       : BaseT(TM, F.getParent()->getDataLayout()),
         ST(static_cast<const GCNSubtarget *>(TM->getSubtargetImpl(F))),
         TLI(ST->getTargetLowering()), CommonTTI(TM, F),
-        IsGraphicsShader(AMDGPU::isShader(F.getCallingConv())),
+        IsGraphics(AMDGPU::isGraphics(F.getCallingConv())),
         MaxVGPRs(ST->getMaxNumVGPRs(
             std::max(ST->getWavesPerEU(F).first,
                      ST->getWavesPerEUForWorkGroup(
@@ -217,7 +217,7 @@
   unsigned getFlatAddressSpace() const {
     // Don't bother running InferAddressSpaces pass on graphics shaders which
     // don't use flat addressing.
-    if (IsGraphicsShader)
+    if (IsGraphics)
       return -1;
     return AMDGPUAS::FLAT_ADDRESS;
   }
diff --git a/llvm/lib/Target/AMDGPU/AMDGPUTargetTransformInfo.cpp b/llvm/lib/Target/AMDGPU/AMDGPUTargetTransformInfo.cpp
--- a/llvm/lib/Target/AMDGPU/AMDGPUTargetTransformInfo.cpp
+++ b/llvm/lib/Target/AMDGPU/AMDGPUTargetTransformInfo.cpp
@@ -722,31 +722,6 @@
   }
 }
 
-static bool isArgPassedInSGPR(const Argument *A) {
-  const Function *F = A->getParent();
-
-  // Arguments to compute shaders are never a source of divergence.
-  CallingConv::ID CC = F->getCallingConv();
-  switch (CC) {
-  case CallingConv::AMDGPU_KERNEL:
-  case CallingConv::SPIR_KERNEL:
-    return true;
-  case CallingConv::AMDGPU_VS:
-  case CallingConv::AMDGPU_LS:
-  case CallingConv::AMDGPU_HS:
-  case CallingConv::AMDGPU_ES:
-  case CallingConv::AMDGPU_GS:
-  case CallingConv::AMDGPU_PS:
-  case CallingConv::AMDGPU_CS:
-    // For non-compute shaders, SGPR inputs are marked with either inreg.
-    // Everything else is in VGPRs.
-    return F->getAttributes().hasParamAttribute(A->getArgNo(), Attribute::InReg);
-  default:
-    // TODO: Should calls support inreg for SGPR inputs?
-    return false;
-  }
-}
-
 /// Analyze if the results of inline asm are divergent. If \p Indices is empty,
 /// this is analyzing the collective result of all output registers. Otherwise,
 /// this is only querying a specific result index if this returns multiple
@@ -803,7 +778,7 @@
 /// different across workitems in a wavefront.
 bool GCNTTIImpl::isSourceOfDivergence(const Value *V) const {
   if (const Argument *A = dyn_cast<Argument>(V))
-    return !isArgPassedInSGPR(A);
+    return !AMDGPU::isArgPassedInSGPR(A);
 
   // Loads from the private and flat address spaces are divergent, because
   // threads can execute the load instruction with the same inputs and get
diff --git a/llvm/lib/Target/AMDGPU/SIISelLowering.cpp b/llvm/lib/Target/AMDGPU/SIISelLowering.cpp
--- a/llvm/lib/Target/AMDGPU/SIISelLowering.cpp
+++ b/llvm/lib/Target/AMDGPU/SIISelLowering.cpp
@@ -1780,12 +1780,11 @@
   return CreateLiveInRegister(DAG, RC, Reg->getRegister(), VT);
 }
 
-static void processShaderInputArgs(SmallVectorImpl<ISD::InputArg> &Splits,
-                                   CallingConv::ID CallConv,
-                                   ArrayRef<ISD::InputArg> Ins,
-                                   BitVector &Skipped,
-                                   FunctionType *FType,
-                                   SIMachineFunctionInfo *Info) {
+static void processPSInputArgs(SmallVectorImpl<ISD::InputArg> &Splits,
+                               CallingConv::ID CallConv,
+                               ArrayRef<ISD::InputArg> Ins, BitVector &Skipped,
+                               FunctionType *FType,
+                               SIMachineFunctionInfo *Info) {
   for (unsigned I = 0, E = Ins.size(), PSInputNum = 0; I != E; ++I) {
     const ISD::InputArg *Arg = &Ins[I];
 
@@ -2237,7 +2236,7 @@
   FunctionType *FType = MF.getFunction().getFunctionType();
   SIMachineFunctionInfo *Info = MF.getInfo<SIMachineFunctionInfo>();
 
-  if (Subtarget->isAmdHsaOS() && AMDGPU::isShader(CallConv)) {
+  if (Subtarget->isAmdHsaOS() && AMDGPU::isGraphics(CallConv)) {
     DiagnosticInfoUnsupported NoGraphicsHSA(
         Fn, "unsupported non-compute shaders with HSA", DL.getDebugLoc());
     DAG.getContext()->diagnose(NoGraphicsHSA);
@@ -2250,12 +2249,20 @@
   CCState CCInfo(CallConv, isVarArg, DAG.getMachineFunction(), ArgLocs,
                  *DAG.getContext());
 
-  bool IsShader = AMDGPU::isShader(CallConv);
+  bool IsGraphics = AMDGPU::isGraphics(CallConv);
   bool IsKernel = AMDGPU::isKernel(CallConv);
   bool IsEntryFunc = AMDGPU::isEntryFunctionCC(CallConv);
 
-  if (IsShader) {
-    processShaderInputArgs(Splits, CallConv, Ins, Skipped, FType, Info);
+  if (IsGraphics) {
+    assert(!Info->hasDispatchPtr() && !Info->hasKernargSegmentPtr() &&
+           !Info->hasFlatScratchInit() && !Info->hasWorkGroupIDX() &&
+           !Info->hasWorkGroupIDY() && !Info->hasWorkGroupIDZ() &&
+           !Info->hasWorkGroupInfo() && !Info->hasWorkItemIDX() &&
+           !Info->hasWorkItemIDY() && !Info->hasWorkItemIDZ());
+  }
+
+  if (CallConv == CallingConv::AMDGPU_PS) {
+    processPSInputArgs(Splits, CallConv, Ins, Skipped, FType, Info);
 
     // At least one interpolation mode must be enabled or else the GPU will
     // hang.
@@ -2270,39 +2277,28 @@
     // - At least one of PERSP_* (0xF) or LINEAR_* (0x70) must be enabled.
     // - If POS_W_FLOAT (11) is enabled, at least one of PERSP_* must be
     //   enabled too.
-    if (CallConv == CallingConv::AMDGPU_PS) {
-      if ((Info->getPSInputAddr() & 0x7F) == 0 ||
-           ((Info->getPSInputAddr() & 0xF) == 0 &&
-            Info->isPSInputAllocated(11))) {
-        CCInfo.AllocateReg(AMDGPU::VGPR0);
-        CCInfo.AllocateReg(AMDGPU::VGPR1);
-        Info->markPSInputAllocated(0);
-        Info->markPSInputEnabled(0);
-      }
-      if (Subtarget->isAmdPalOS()) {
-        // For isAmdPalOS, the user does not enable some bits after compilation
-        // based on run-time states; the register values being generated here are
-        // the final ones set in hardware. Therefore we need to apply the
-        // workaround to PSInputAddr and PSInputEnable together.  (The case where
-        // a bit is set in PSInputAddr but not PSInputEnable is where the
-        // frontend set up an input arg for a particular interpolation mode, but
-        // nothing uses that input arg. Really we should have an earlier pass
-        // that removes such an arg.)
-        unsigned PsInputBits = Info->getPSInputAddr() & Info->getPSInputEnable();
-        if ((PsInputBits & 0x7F) == 0 ||
-            ((PsInputBits & 0xF) == 0 &&
-             (PsInputBits >> 11 & 1)))
-          Info->markPSInputEnabled(
-              countTrailingZeros(Info->getPSInputAddr(), ZB_Undefined));
-      }
+    if ((Info->getPSInputAddr() & 0x7F) == 0 ||
+        ((Info->getPSInputAddr() & 0xF) == 0 && Info->isPSInputAllocated(11))) {
+      CCInfo.AllocateReg(AMDGPU::VGPR0);
+      CCInfo.AllocateReg(AMDGPU::VGPR1);
+      Info->markPSInputAllocated(0);
+      Info->markPSInputEnabled(0);
+    }
+    if (Subtarget->isAmdPalOS()) {
+      // For isAmdPalOS, the user does not enable some bits after compilation
+      // based on run-time states; the register values being generated here are
+      // the final ones set in hardware. Therefore we need to apply the
+      // workaround to PSInputAddr and PSInputEnable together.  (The case where
+      // a bit is set in PSInputAddr but not PSInputEnable is where the
+      // frontend set up an input arg for a particular interpolation mode, but
+      // nothing uses that input arg. Really we should have an earlier pass
+      // that removes such an arg.)
+      unsigned PsInputBits = Info->getPSInputAddr() & Info->getPSInputEnable();
+      if ((PsInputBits & 0x7F) == 0 ||
+          ((PsInputBits & 0xF) == 0 && (PsInputBits >> 11 & 1)))
+        Info->markPSInputEnabled(
+            countTrailingZeros(Info->getPSInputAddr(), ZB_Undefined));
     }
-
-    assert(!Info->hasDispatchPtr() &&
-           !Info->hasKernargSegmentPtr() && !Info->hasFlatScratchInit() &&
-           !Info->hasWorkGroupIDX() && !Info->hasWorkGroupIDY() &&
-           !Info->hasWorkGroupIDZ() && !Info->hasWorkGroupInfo() &&
-           !Info->hasWorkItemIDX() && !Info->hasWorkItemIDY() &&
-           !Info->hasWorkItemIDZ());
   } else if (IsKernel) {
     assert(Info->hasWorkGroupIDX() && Info->hasWorkItemIDX());
   } else {
@@ -2449,7 +2445,7 @@
 
   // Start adding system SGPRs.
   if (IsEntryFunc) {
-    allocateSystemSGPRs(CCInfo, MF, *Info, CallConv, IsShader);
+    allocateSystemSGPRs(CCInfo, MF, *Info, CallConv, IsGraphics);
   } else {
     CCInfo.AllocateReg(Info->getScratchRSrcReg());
     allocateSpecialInputSGPRs(CCInfo, MF, *TRI, *Info);
@@ -2932,7 +2928,7 @@
     report_fatal_error("unsupported libcall legalization");
 
   if (!AMDGPUTargetMachine::EnableFixedFunctionABI &&
-      !CLI.CB->getCalledFunction()) {
+      !CLI.CB->getCalledFunction() && CallConv != CallingConv::AMDGPU_Gfx) {
     return lowerUnhandledCall(CLI, InVals,
                               "unsupported indirect call to function ");
   }
@@ -2942,11 +2938,19 @@
                               "unsupported required tail call to function ");
   }
 
-  if (AMDGPU::isShader(MF.getFunction().getCallingConv())) {
-    // Note the issue is with the CC of the calling function, not of the call
+  if (AMDGPU::isShader(CallConv)) {
+    // Note the issue is with the CC of the called function, not of the call
     // itself.
     return lowerUnhandledCall(CLI, InVals,
-                          "unsupported call from graphics shader of function ");
+                              "unsupported call to a shader function ");
+  }
+
+  if (AMDGPU::isShader(MF.getFunction().getCallingConv()) &&
+      CallConv != CallingConv::AMDGPU_Gfx) {
+    // Only allow calls with specific calling conventions.
+    return lowerUnhandledCall(CLI, InVals,
+                              "unsupported calling convention for call from "
+                              "graphics shader of function ");
   }
 
   if (IsTailCall) {
@@ -2977,7 +2981,8 @@
   CCState CCInfo(CallConv, IsVarArg, MF, ArgLocs, *DAG.getContext());
   CCAssignFn *AssignFn = CCAssignFnForCall(CallConv, IsVarArg);
 
-  if (AMDGPUTargetMachine::EnableFixedFunctionABI) {
+  if (AMDGPUTargetMachine::EnableFixedFunctionABI &&
+      CallConv != CallingConv::AMDGPU_Gfx) {
     // With a fixed ABI, allocate fixed registers before user arguments.
     passSpecialInputs(CLI, CCInfo, *Info, RegsToPass, MemOpChains, Chain);
   }
@@ -3111,7 +3116,8 @@
     }
   }
 
-  if (!AMDGPUTargetMachine::EnableFixedFunctionABI) {
+  if (!AMDGPUTargetMachine::EnableFixedFunctionABI &&
+      CallConv != CallingConv::AMDGPU_Gfx) {
     // Copy special input registers after user input arguments.
     passSpecialInputs(CLI, CCInfo, *Info, RegsToPass, MemOpChains, Chain);
   }
diff --git a/llvm/lib/Target/AMDGPU/SIInstrInfo.cpp b/llvm/lib/Target/AMDGPU/SIInstrInfo.cpp
--- a/llvm/lib/Target/AMDGPU/SIInstrInfo.cpp
+++ b/llvm/lib/Target/AMDGPU/SIInstrInfo.cpp
@@ -5087,8 +5087,7 @@
   // scratch memory access. In both cases, the legalization never involves
   // conversion to the addr64 form.
   if (isMIMG(MI) ||
-      (AMDGPU::isShader(MF.getFunction().getCallingConv()) &&
-       (isMUBUF(MI) || isMTBUF(MI)))) {
+      ((ST.isAmdPalOS() || ST.isMesa3DOS()) && (isMUBUF(MI) || isMTBUF(MI)))) {
     MachineOperand *SRsrc = getNamedOperand(MI, AMDGPU::OpName::srsrc);
     if (SRsrc && !RI.isSGPRClass(MRI.getRegClass(SRsrc->getReg())))
       loadSRsrcFromVGPR(*this, MI, *SRsrc, MDT);
diff --git a/llvm/lib/Target/AMDGPU/SILowerSGPRSpills.cpp b/llvm/lib/Target/AMDGPU/SILowerSGPRSpills.cpp
--- a/llvm/lib/Target/AMDGPU/SILowerSGPRSpills.cpp
+++ b/llvm/lib/Target/AMDGPU/SILowerSGPRSpills.cpp
@@ -276,6 +276,7 @@
   FuncInfo->setSGPRSpillVGPRs(LowestAvailableVGPR, FI, Index);
 
   for (MachineBasicBlock &MBB : MF) {
+    assert(LowestAvailableVGPR.isValid() && "Did not find an available VGPR");
     MBB.addLiveIn(LowestAvailableVGPR);
     MBB.sortUniqueLiveIns();
   }
diff --git a/llvm/lib/Target/AMDGPU/SIRegisterInfo.cpp b/llvm/lib/Target/AMDGPU/SIRegisterInfo.cpp
--- a/llvm/lib/Target/AMDGPU/SIRegisterInfo.cpp
+++ b/llvm/lib/Target/AMDGPU/SIRegisterInfo.cpp
@@ -125,6 +125,7 @@
   case CallingConv::C:
   case CallingConv::Fast:
   case CallingConv::Cold:
+  case CallingConv::AMDGPU_Gfx:
     return CSR_AMDGPU_HighRegs_SaveList;
   default: {
     // Dummy to not crash RegisterClassInfo.
@@ -145,6 +146,7 @@
   case CallingConv::C:
   case CallingConv::Fast:
   case CallingConv::Cold:
+  case CallingConv::AMDGPU_Gfx:
     return CSR_AMDGPU_HighRegs_RegMask;
   default:
     return nullptr;
diff --git a/llvm/lib/Target/AMDGPU/Utils/AMDGPUBaseInfo.h b/llvm/lib/Target/AMDGPU/Utils/AMDGPUBaseInfo.h
--- a/llvm/lib/Target/AMDGPU/Utils/AMDGPUBaseInfo.h
+++ b/llvm/lib/Target/AMDGPU/Utils/AMDGPUBaseInfo.h
@@ -567,6 +567,9 @@
 LLVM_READNONE
 bool isShader(CallingConv::ID CC);
 
+LLVM_READNONE
+bool isGraphics(CallingConv::ID CC);
+
 LLVM_READNONE
 bool isCompute(CallingConv::ID CC);
 
diff --git a/llvm/lib/Target/AMDGPU/Utils/AMDGPUBaseInfo.cpp b/llvm/lib/Target/AMDGPU/Utils/AMDGPUBaseInfo.cpp
--- a/llvm/lib/Target/AMDGPU/Utils/AMDGPUBaseInfo.cpp
+++ b/llvm/lib/Target/AMDGPU/Utils/AMDGPUBaseInfo.cpp
@@ -1043,8 +1043,12 @@
   }
 }
 
+bool isGraphics(CallingConv::ID cc) {
+  return isShader(cc) || cc == CallingConv::AMDGPU_Gfx;
+}
+
 bool isCompute(CallingConv::ID cc) {
-  return !isShader(cc) || cc == CallingConv::AMDGPU_CS;
+  return !isGraphics(cc) || cc == CallingConv::AMDGPU_CS;
 }
 
 bool isEntryFunctionCC(CallingConv::ID CC) {
@@ -1439,6 +1443,7 @@
   case CallingConv::AMDGPU_GS:
   case CallingConv::AMDGPU_PS:
   case CallingConv::AMDGPU_CS:
+  case CallingConv::AMDGPU_Gfx:
     // For non-compute shaders, SGPR inputs are marked with either inreg or byval.
     // Everything else is in VGPRs.
     return F->getAttributes().hasParamAttribute(A->getArgNo(), Attribute::InReg) ||
diff --git a/llvm/lib/Target/AMDGPU/Utils/AMDGPUPALMetadata.cpp b/llvm/lib/Target/AMDGPU/Utils/AMDGPUPALMetadata.cpp
--- a/llvm/lib/Target/AMDGPU/Utils/AMDGPUPALMetadata.cpp
+++ b/llvm/lib/Target/AMDGPU/Utils/AMDGPUPALMetadata.cpp
@@ -733,6 +733,8 @@
     return ".hs";
   case CallingConv::AMDGPU_LS:
     return ".ls";
+  case CallingConv::AMDGPU_Gfx:
+    llvm_unreachable("Callable shader has no hardware stage");
   default:
     return ".cs";
   }
diff --git a/llvm/test/Bitcode/compatibility.ll b/llvm/test/Bitcode/compatibility.ll
--- a/llvm/test/Bitcode/compatibility.ll
+++ b/llvm/test/Bitcode/compatibility.ll
@@ -478,6 +478,8 @@
 ; CHECK: declare amdgpu_cs void @f.cc90()
 declare amdgpu_cs void @f.amdgpu_cs()
 ; CHECK: declare amdgpu_cs void @f.amdgpu_cs()
+declare amdgpu_gfx void @f.amdgpu_gfx()
+; CHECK: declare amdgpu_gfx void @f.amdgpu_gfx()
 declare cc91 void @f.cc91()
 ; CHECK: declare amdgpu_kernel void @f.cc91()
 declare amdgpu_kernel void @f.amdgpu_kernel()
diff --git a/llvm/test/CodeGen/AMDGPU/amdpal-callable.ll b/llvm/test/CodeGen/AMDGPU/amdpal-callable.ll
new file mode 100644
--- /dev/null
+++ b/llvm/test/CodeGen/AMDGPU/amdpal-callable.ll
@@ -0,0 +1,10 @@
+; RUN: llc -mtriple=amdgcn--amdpal -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=SI -enable-var-scope %s
+; RUN: llc -mtriple=amdgcn--amdpal -mcpu=tonga -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=VI -enable-var-scope %s
+; RUN: llc -mtriple=amdgcn--amdpal -mcpu=gfx900 -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=GFX9 -enable-var-scope %s
+
+; GCN-LABEL: {{^}}gfx_callable_amdpal:
+; GCN: .amd_amdgpu_pal_metadata {{$}}
+define amdgpu_gfx half @gfx_callable_amdpal(half %arg0) {
+  %add = fadd half %arg0, 1.0
+  ret half %add
+}
diff --git a/llvm/test/CodeGen/AMDGPU/gfx-callable-argument-types.ll b/llvm/test/CodeGen/AMDGPU/gfx-callable-argument-types.ll
new file mode 100644
--- /dev/null
+++ b/llvm/test/CodeGen/AMDGPU/gfx-callable-argument-types.ll
@@ -0,0 +1,6566 @@
+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
+; RUN: llc -mtriple=amdgcn--amdpal -mcpu=gfx900 -verify-machineinstrs < %s | FileCheck -check-prefixes=GCN,GFX9 %s
+; RUN: llc -mtriple=amdgcn--amdpal -mcpu=gfx1010 -verify-machineinstrs < %s | FileCheck -check-prefixes=GCN,GFX10 %s
+
+declare hidden amdgpu_gfx void @external_void_func_i1(i1) #0
+declare hidden amdgpu_gfx void @external_void_func_i1_signext(i1 signext) #0
+declare hidden amdgpu_gfx void @external_void_func_i1_zeroext(i1 zeroext) #0
+
+declare hidden amdgpu_gfx void @external_void_func_i8(i8) #0
+declare hidden amdgpu_gfx void @external_void_func_i8_signext(i8 signext) #0
+declare hidden amdgpu_gfx void @external_void_func_i8_zeroext(i8 zeroext) #0
+
+declare hidden amdgpu_gfx void @external_void_func_i16(i16) #0
+declare hidden amdgpu_gfx void @external_void_func_i16_signext(i16 signext) #0
+declare hidden amdgpu_gfx void @external_void_func_i16_zeroext(i16 zeroext) #0
+
+declare hidden amdgpu_gfx void @external_void_func_i32(i32) #0
+declare hidden amdgpu_gfx void @external_void_func_i64(i64) #0
+declare hidden amdgpu_gfx void @external_void_func_v2i64(<2 x i64>) #0
+declare hidden amdgpu_gfx void @external_void_func_v3i64(<3 x i64>) #0
+declare hidden amdgpu_gfx void @external_void_func_v4i64(<4 x i64>) #0
+
+declare hidden amdgpu_gfx void @external_void_func_f16(half) #0
+declare hidden amdgpu_gfx void @external_void_func_f32(float) #0
+declare hidden amdgpu_gfx void @external_void_func_f64(double) #0
+declare hidden amdgpu_gfx void @external_void_func_v2f32(<2 x float>) #0
+declare hidden amdgpu_gfx void @external_void_func_v2f64(<2 x double>) #0
+declare hidden amdgpu_gfx void @external_void_func_v3f32(<3 x float>) #0
+declare hidden amdgpu_gfx void @external_void_func_v3f64(<3 x double>) #0
+declare hidden amdgpu_gfx void @external_void_func_v5f32(<5 x float>) #0
+
+declare hidden amdgpu_gfx void @external_void_func_v2i16(<2 x i16>) #0
+declare hidden amdgpu_gfx void @external_void_func_v2f16(<2 x half>) #0
+declare hidden amdgpu_gfx void @external_void_func_v3i16(<3 x i16>) #0
+declare hidden amdgpu_gfx void @external_void_func_v3f16(<3 x half>) #0
+declare hidden amdgpu_gfx void @external_void_func_v4i16(<4 x i16>) #0
+declare hidden amdgpu_gfx void @external_void_func_v4f16(<4 x half>) #0
+
+declare hidden amdgpu_gfx void @external_void_func_v2i32(<2 x i32>) #0
+declare hidden amdgpu_gfx void @external_void_func_v3i32(<3 x i32>) #0
+declare hidden amdgpu_gfx void @external_void_func_v3i32_i32(<3 x i32>, i32) #0
+declare hidden amdgpu_gfx void @external_void_func_v4i32(<4 x i32>) #0
+declare hidden amdgpu_gfx void @external_void_func_v5i32(<5 x i32>) #0
+declare hidden amdgpu_gfx void @external_void_func_v8i32(<8 x i32>) #0
+declare hidden amdgpu_gfx void @external_void_func_v16i32(<16 x i32>) #0
+declare hidden amdgpu_gfx void @external_void_func_v32i32(<32 x i32>) #0
+declare hidden amdgpu_gfx void @external_void_func_v32i32_i32(<32 x i32>, i32) #0
+
+declare hidden amdgpu_gfx void @external_void_func_i1_inreg(i1 inreg) #0
+declare hidden amdgpu_gfx void @external_void_func_i8_inreg(i8 inreg) #0
+declare hidden amdgpu_gfx void @external_void_func_i16_inreg(i16 inreg) #0
+declare hidden amdgpu_gfx void @external_void_func_i32_inreg(i32 inreg) #0
+declare hidden amdgpu_gfx void @external_void_func_i64_inreg(i64 inreg) #0
+declare hidden amdgpu_gfx void @external_void_func_v2i64_inreg(<2 x i64> inreg) #0
+declare hidden amdgpu_gfx void @external_void_func_v3i64_inreg(<3 x i64> inreg) #0
+declare hidden amdgpu_gfx void @external_void_func_v4i64_inreg(<4 x i64> inreg) #0
+
+declare hidden amdgpu_gfx void @external_void_func_f16_inreg(half inreg) #0
+declare hidden amdgpu_gfx void @external_void_func_f32_inreg(float inreg) #0
+declare hidden amdgpu_gfx void @external_void_func_f64_inreg(double inreg) #0
+declare hidden amdgpu_gfx void @external_void_func_v2f32_inreg(<2 x float> inreg) #0
+declare hidden amdgpu_gfx void @external_void_func_v2f64_inreg(<2 x double> inreg) #0
+declare hidden amdgpu_gfx void @external_void_func_v3f32_inreg(<3 x float> inreg) #0
+declare hidden amdgpu_gfx void @external_void_func_v3f64_inreg(<3 x double> inreg) #0
+declare hidden amdgpu_gfx void @external_void_func_v5f32_inreg(<5 x float> inreg) #0
+
+declare hidden amdgpu_gfx void @external_void_func_v2i16_inreg(<2 x i16> inreg) #0
+declare hidden amdgpu_gfx void @external_void_func_v2f16_inreg(<2 x half> inreg) #0
+declare hidden amdgpu_gfx void @external_void_func_v3i16_inreg(<3 x i16> inreg) #0
+declare hidden amdgpu_gfx void @external_void_func_v3f16_inreg(<3 x half> inreg) #0
+declare hidden amdgpu_gfx void @external_void_func_v4i16_inreg(<4 x i16> inreg) #0
+declare hidden amdgpu_gfx void @external_void_func_v4f16_inreg(<4 x half> inreg) #0
+
+declare hidden amdgpu_gfx void @external_void_func_v2i32_inreg(<2 x i32> inreg) #0
+declare hidden amdgpu_gfx void @external_void_func_v3i32_inreg(<3 x i32> inreg) #0
+declare hidden amdgpu_gfx void @external_void_func_v3i32_i32_inreg(<3 x i32> inreg, i32 inreg) #0
+declare hidden amdgpu_gfx void @external_void_func_v4i32_inreg(<4 x i32> inreg) #0
+declare hidden amdgpu_gfx void @external_void_func_v5i32_inreg(<5 x i32> inreg) #0
+declare hidden amdgpu_gfx void @external_void_func_v8i32_inreg(<8 x i32> inreg) #0
+declare hidden amdgpu_gfx void @external_void_func_v16i32_inreg(<16 x i32> inreg) #0
+declare hidden amdgpu_gfx void @external_void_func_v32i32_inreg(<32 x i32> inreg) #0
+declare hidden amdgpu_gfx void @external_void_func_v32i32_i32_inreg(<32 x i32> inreg, i32 inreg) #0
+
+; return value and argument
+declare hidden amdgpu_gfx i32 @external_i32_func_i32(i32) #0
+
+; Structs
+declare hidden amdgpu_gfx void @external_void_func_struct_i8_i32({ i8, i32 }) #0
+declare hidden amdgpu_gfx void @external_void_func_byval_struct_i8_i32({ i8, i32 } addrspace(5)* byval) #0
+declare hidden amdgpu_gfx void @external_void_func_sret_struct_i8_i32_byval_struct_i8_i32({ i8, i32 } addrspace(5)* sret, { i8, i32 } addrspace(5)* byval) #0
+
+declare hidden amdgpu_gfx void @external_void_func_v16i8(<16 x i8>) #0
+
+define amdgpu_gfx void @test_call_external_void_func_i1_imm() #0 {
+; GFX9-LABEL: test_call_external_void_func_i1_imm:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    s_or_saveexec_b64 s[4:5], -1
+; GFX9-NEXT:    buffer_store_dword v40, off, s[0:3], s32 ; 4-byte Folded Spill
+; GFX9-NEXT:    s_mov_b64 exec, s[4:5]
+; GFX9-NEXT:    v_writelane_b32 v40, s33, 2
+; GFX9-NEXT:    v_writelane_b32 v40, s30, 0
+; GFX9-NEXT:    s_mov_b32 s33, s32
+; GFX9-NEXT:    s_add_u32 s32, s32, 0x400
+; GFX9-NEXT:    v_mov_b32_e32 v0, 1
+; GFX9-NEXT:    v_writelane_b32 v40, s31, 1
+; GFX9-NEXT:    s_getpc_b64 s[4:5]
+; GFX9-NEXT:    s_add_u32 s4, s4, external_void_func_i1@rel32@lo+4
+; GFX9-NEXT:    s_addc_u32 s5, s5, external_void_func_i1@rel32@hi+12
+; GFX9-NEXT:    buffer_store_byte v0, off, s[0:3], s32
+; GFX9-NEXT:    s_swappc_b64 s[30:31], s[4:5]
+; GFX9-NEXT:    v_readlane_b32 s4, v40, 0
+; GFX9-NEXT:    v_readlane_b32 s5, v40, 1
+; GFX9-NEXT:    s_sub_u32 s32, s32, 0x400
+; GFX9-NEXT:    v_readlane_b32 s33, v40, 2
+; GFX9-NEXT:    s_or_saveexec_b64 s[6:7], -1
+; GFX9-NEXT:    buffer_load_dword v40, off, s[0:3], s32 ; 4-byte Folded Reload
+; GFX9-NEXT:    s_mov_b64 exec, s[6:7]
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-NEXT:    s_setpc_b64 s[4:5]
+;
+; GFX10-LABEL: test_call_external_void_func_i1_imm:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
+; GFX10-NEXT:    s_or_saveexec_b32 s4, -1
+; GFX10-NEXT:    buffer_store_dword v40, off, s[0:3], s32 ; 4-byte Folded Spill
+; GFX10-NEXT:    s_waitcnt_depctr 0xffe3
+; GFX10-NEXT:    s_mov_b32 exec_lo, s4
+; GFX10-NEXT:    v_writelane_b32 v40, s33, 2
+; GFX10-NEXT:    v_mov_b32_e32 v0, 1
+; GFX10-NEXT:    s_mov_b32 s33, s32
+; GFX10-NEXT:    s_add_u32 s32, s32, 0x200
+; GFX10-NEXT:    s_getpc_b64 s[4:5]
+; GFX10-NEXT:    s_add_u32 s4, s4, external_void_func_i1@rel32@lo+4
+; GFX10-NEXT:    s_addc_u32 s5, s5, external_void_func_i1@rel32@hi+12
+; GFX10-NEXT:    v_writelane_b32 v40, s30, 0
+; GFX10-NEXT:    ; implicit-def: $vcc_hi
+; GFX10-NEXT:    v_writelane_b32 v40, s31, 1
+; GFX10-NEXT:    buffer_store_byte v0, off, s[0:3], s32
+; GFX10-NEXT:    s_swappc_b64 s[30:31], s[4:5]
+; GFX10-NEXT:    v_readlane_b32 s4, v40, 0
+; GFX10-NEXT:    v_readlane_b32 s5, v40, 1
+; GFX10-NEXT:    s_sub_u32 s32, s32, 0x200
+; GFX10-NEXT:    v_readlane_b32 s33, v40, 2
+; GFX10-NEXT:    s_or_saveexec_b32 s6, -1
+; GFX10-NEXT:    buffer_load_dword v40, off, s[0:3], s32 ; 4-byte Folded Reload
+; GFX10-NEXT:    s_waitcnt_depctr 0xffe3
+; GFX10-NEXT:    s_mov_b32 exec_lo, s6
+; GFX10-NEXT:    s_waitcnt vmcnt(0)
+; GFX10-NEXT:    s_setpc_b64 s[4:5]
+  call amdgpu_gfx void @external_void_func_i1(i1 true)
+  ret void
+}
+
+define amdgpu_gfx void @test_call_external_void_func_i1_signext(i32) #0 {
+; GFX9-LABEL: test_call_external_void_func_i1_signext:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    s_or_saveexec_b64 s[4:5], -1
+; GFX9-NEXT:    buffer_store_dword v40, off, s[0:3], s32 ; 4-byte Folded Spill
+; GFX9-NEXT:    s_mov_b64 exec, s[4:5]
+; GFX9-NEXT:    v_writelane_b32 v40, s33, 2
+; GFX9-NEXT:    v_writelane_b32 v40, s30, 0
+; GFX9-NEXT:    v_writelane_b32 v40, s31, 1
+; GFX9-NEXT:    global_load_ubyte v0, v[0:1], off
+; GFX9-NEXT:    s_mov_b32 s33, s32
+; GFX9-NEXT:    s_add_u32 s32, s32, 0x400
+; GFX9-NEXT:    s_getpc_b64 s[4:5]
+; GFX9-NEXT:    s_add_u32 s4, s4, external_void_func_i1_signext@rel32@lo+4
+; GFX9-NEXT:    s_addc_u32 s5, s5, external_void_func_i1_signext@rel32@hi+12
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-NEXT:    v_and_b32_e32 v0, 1, v0
+; GFX9-NEXT:    buffer_store_byte v0, off, s[0:3], s32
+; GFX9-NEXT:    s_swappc_b64 s[30:31], s[4:5]
+; GFX9-NEXT:    v_readlane_b32 s4, v40, 0
+; GFX9-NEXT:    v_readlane_b32 s5, v40, 1
+; GFX9-NEXT:    s_sub_u32 s32, s32, 0x400
+; GFX9-NEXT:    v_readlane_b32 s33, v40, 2
+; GFX9-NEXT:    s_or_saveexec_b64 s[6:7], -1
+; GFX9-NEXT:    buffer_load_dword v40, off, s[0:3], s32 ; 4-byte Folded Reload
+; GFX9-NEXT:    s_mov_b64 exec, s[6:7]
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-NEXT:    s_setpc_b64 s[4:5]
+;
+; GFX10-LABEL: test_call_external_void_func_i1_signext:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
+; GFX10-NEXT:    s_or_saveexec_b32 s4, -1
+; GFX10-NEXT:    buffer_store_dword v40, off, s[0:3], s32 ; 4-byte Folded Spill
+; GFX10-NEXT:    s_waitcnt_depctr 0xffe3
+; GFX10-NEXT:    s_mov_b32 exec_lo, s4
+; GFX10-NEXT:    v_writelane_b32 v40, s33, 2
+; GFX10-NEXT:    s_mov_b32 s33, s32
+; GFX10-NEXT:    s_add_u32 s32, s32, 0x200
+; GFX10-NEXT:    s_getpc_b64 s[4:5]
+; GFX10-NEXT:    s_add_u32 s4, s4, external_void_func_i1_signext@rel32@lo+4
+; GFX10-NEXT:    s_addc_u32 s5, s5, external_void_func_i1_signext@rel32@hi+12
+; GFX10-NEXT:    ; implicit-def: $vcc_hi
+; GFX10-NEXT:    v_writelane_b32 v40, s30, 0
+; GFX10-NEXT:    v_writelane_b32 v40, s31, 1
+; GFX10-NEXT:    global_load_ubyte v0, v[0:1], off
+; GFX10-NEXT:    s_waitcnt vmcnt(0)
+; GFX10-NEXT:    v_and_b32_e32 v0, 1, v0
+; GFX10-NEXT:    buffer_store_byte v0, off, s[0:3], s32
+; GFX10-NEXT:    s_swappc_b64 s[30:31], s[4:5]
+; GFX10-NEXT:    v_readlane_b32 s4, v40, 0
+; GFX10-NEXT:    v_readlane_b32 s5, v40, 1
+; GFX10-NEXT:    s_sub_u32 s32, s32, 0x200
+; GFX10-NEXT:    v_readlane_b32 s33, v40, 2
+; GFX10-NEXT:    s_or_saveexec_b32 s6, -1
+; GFX10-NEXT:    buffer_load_dword v40, off, s[0:3], s32 ; 4-byte Folded Reload
+; GFX10-NEXT:    s_waitcnt_depctr 0xffe3
+; GFX10-NEXT:    s_mov_b32 exec_lo, s6
+; GFX10-NEXT:    s_waitcnt vmcnt(0)
+; GFX10-NEXT:    s_setpc_b64 s[4:5]
+  %var = load volatile i1, i1 addrspace(1)* undef
+  call amdgpu_gfx void @external_void_func_i1_signext(i1 %var)
+  ret void
+}
+
+define amdgpu_gfx void @test_call_external_void_func_i1_zeroext(i32) #0 {
+; GFX9-LABEL: test_call_external_void_func_i1_zeroext:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    s_or_saveexec_b64 s[4:5], -1
+; GFX9-NEXT:    buffer_store_dword v40, off, s[0:3], s32 ; 4-byte Folded Spill
+; GFX9-NEXT:    s_mov_b64 exec, s[4:5]
+; GFX9-NEXT:    v_writelane_b32 v40, s33, 2
+; GFX9-NEXT:    v_writelane_b32 v40, s30, 0
+; GFX9-NEXT:    v_writelane_b32 v40, s31, 1
+; GFX9-NEXT:    global_load_ubyte v0, v[0:1], off
+; GFX9-NEXT:    s_mov_b32 s33, s32
+; GFX9-NEXT:    s_add_u32 s32, s32, 0x400
+; GFX9-NEXT:    s_getpc_b64 s[4:5]
+; GFX9-NEXT:    s_add_u32 s4, s4, external_void_func_i1_zeroext@rel32@lo+4
+; GFX9-NEXT:    s_addc_u32 s5, s5, external_void_func_i1_zeroext@rel32@hi+12
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-NEXT:    v_and_b32_e32 v0, 1, v0
+; GFX9-NEXT:    buffer_store_byte v0, off, s[0:3], s32
+; GFX9-NEXT:    s_swappc_b64 s[30:31], s[4:5]
+; GFX9-NEXT:    v_readlane_b32 s4, v40, 0
+; GFX9-NEXT:    v_readlane_b32 s5, v40, 1
+; GFX9-NEXT:    s_sub_u32 s32, s32, 0x400
+; GFX9-NEXT:    v_readlane_b32 s33, v40, 2
+; GFX9-NEXT:    s_or_saveexec_b64 s[6:7], -1
+; GFX9-NEXT:    buffer_load_dword v40, off, s[0:3], s32 ; 4-byte Folded Reload
+; GFX9-NEXT:    s_mov_b64 exec, s[6:7]
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-NEXT:    s_setpc_b64 s[4:5]
+;
+; GFX10-LABEL: test_call_external_void_func_i1_zeroext:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
+; GFX10-NEXT:    s_or_saveexec_b32 s4, -1
+; GFX10-NEXT:    buffer_store_dword v40, off, s[0:3], s32 ; 4-byte Folded Spill
+; GFX10-NEXT:    s_waitcnt_depctr 0xffe3
+; GFX10-NEXT:    s_mov_b32 exec_lo, s4
+; GFX10-NEXT:    v_writelane_b32 v40, s33, 2
+; GFX10-NEXT:    s_mov_b32 s33, s32
+; GFX10-NEXT:    s_add_u32 s32, s32, 0x200
+; GFX10-NEXT:    s_getpc_b64 s[4:5]
+; GFX10-NEXT:    s_add_u32 s4, s4, external_void_func_i1_zeroext@rel32@lo+4
+; GFX10-NEXT:    s_addc_u32 s5, s5, external_void_func_i1_zeroext@rel32@hi+12
+; GFX10-NEXT:    ; implicit-def: $vcc_hi
+; GFX10-NEXT:    v_writelane_b32 v40, s30, 0
+; GFX10-NEXT:    v_writelane_b32 v40, s31, 1
+; GFX10-NEXT:    global_load_ubyte v0, v[0:1], off
+; GFX10-NEXT:    s_waitcnt vmcnt(0)
+; GFX10-NEXT:    v_and_b32_e32 v0, 1, v0
+; GFX10-NEXT:    buffer_store_byte v0, off, s[0:3], s32
+; GFX10-NEXT:    s_swappc_b64 s[30:31], s[4:5]
+; GFX10-NEXT:    v_readlane_b32 s4, v40, 0
+; GFX10-NEXT:    v_readlane_b32 s5, v40, 1
+; GFX10-NEXT:    s_sub_u32 s32, s32, 0x200
+; GFX10-NEXT:    v_readlane_b32 s33, v40, 2
+; GFX10-NEXT:    s_or_saveexec_b32 s6, -1
+; GFX10-NEXT:    buffer_load_dword v40, off, s[0:3], s32 ; 4-byte Folded Reload
+; GFX10-NEXT:    s_waitcnt_depctr 0xffe3
+; GFX10-NEXT:    s_mov_b32 exec_lo, s6
+; GFX10-NEXT:    s_waitcnt vmcnt(0)
+; GFX10-NEXT:    s_setpc_b64 s[4:5]
+  %var = load volatile i1, i1 addrspace(1)* undef
+  call amdgpu_gfx void @external_void_func_i1_zeroext(i1 %var)
+  ret void
+}
+
+define amdgpu_gfx void @test_call_external_void_func_i8_imm(i32) #0 {
+; GFX9-LABEL: test_call_external_void_func_i8_imm:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    s_or_saveexec_b64 s[4:5], -1
+; GFX9-NEXT:    buffer_store_dword v40, off, s[0:3], s32 ; 4-byte Folded Spill
+; GFX9-NEXT:    s_mov_b64 exec, s[4:5]
+; GFX9-NEXT:    v_writelane_b32 v40, s33, 2
+; GFX9-NEXT:    v_writelane_b32 v40, s30, 0
+; GFX9-NEXT:    s_mov_b32 s33, s32
+; GFX9-NEXT:    s_add_u32 s32, s32, 0x400
+; GFX9-NEXT:    v_mov_b32_e32 v0, 0x7b
+; GFX9-NEXT:    s_getpc_b64 s[4:5]
+; GFX9-NEXT:    s_add_u32 s4, s4, external_void_func_i8@rel32@lo+4
+; GFX9-NEXT:    s_addc_u32 s5, s5, external_void_func_i8@rel32@hi+12
+; GFX9-NEXT:    v_writelane_b32 v40, s31, 1
+; GFX9-NEXT:    s_swappc_b64 s[30:31], s[4:5]
+; GFX9-NEXT:    v_readlane_b32 s4, v40, 0
+; GFX9-NEXT:    v_readlane_b32 s5, v40, 1
+; GFX9-NEXT:    s_sub_u32 s32, s32, 0x400
+; GFX9-NEXT:    v_readlane_b32 s33, v40, 2
+; GFX9-NEXT:    s_or_saveexec_b64 s[6:7], -1
+; GFX9-NEXT:    buffer_load_dword v40, off, s[0:3], s32 ; 4-byte Folded Reload
+; GFX9-NEXT:    s_mov_b64 exec, s[6:7]
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-NEXT:    s_setpc_b64 s[4:5]
+;
+; GFX10-LABEL: test_call_external_void_func_i8_imm:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
+; GFX10-NEXT:    s_or_saveexec_b32 s4, -1
+; GFX10-NEXT:    buffer_store_dword v40, off, s[0:3], s32 ; 4-byte Folded Spill
+; GFX10-NEXT:    s_waitcnt_depctr 0xffe3
+; GFX10-NEXT:    s_mov_b32 exec_lo, s4
+; GFX10-NEXT:    v_writelane_b32 v40, s33, 2
+; GFX10-NEXT:    v_mov_b32_e32 v0, 0x7b
+; GFX10-NEXT:    s_mov_b32 s33, s32
+; GFX10-NEXT:    s_add_u32 s32, s32, 0x200
+; GFX10-NEXT:    s_getpc_b64 s[4:5]
+; GFX10-NEXT:    s_add_u32 s4, s4, external_void_func_i8@rel32@lo+4
+; GFX10-NEXT:    s_addc_u32 s5, s5, external_void_func_i8@rel32@hi+12
+; GFX10-NEXT:    v_writelane_b32 v40, s30, 0
+; GFX10-NEXT:    ; implicit-def: $vcc_hi
+; GFX10-NEXT:    v_writelane_b32 v40, s31, 1
+; GFX10-NEXT:    s_swappc_b64 s[30:31], s[4:5]
+; GFX10-NEXT:    v_readlane_b32 s4, v40, 0
+; GFX10-NEXT:    v_readlane_b32 s5, v40, 1
+; GFX10-NEXT:    s_sub_u32 s32, s32, 0x200
+; GFX10-NEXT:    v_readlane_b32 s33, v40, 2
+; GFX10-NEXT:    s_or_saveexec_b32 s6, -1
+; GFX10-NEXT:    buffer_load_dword v40, off, s[0:3], s32 ; 4-byte Folded Reload
+; GFX10-NEXT:    s_waitcnt_depctr 0xffe3
+; GFX10-NEXT:    s_mov_b32 exec_lo, s6
+; GFX10-NEXT:    s_waitcnt vmcnt(0)
+; GFX10-NEXT:    s_setpc_b64 s[4:5]
+  call amdgpu_gfx void @external_void_func_i8(i8 123)
+  ret void
+}
+
+define amdgpu_gfx void @test_call_external_void_func_i8_signext(i32) #0 {
+; GFX9-LABEL: test_call_external_void_func_i8_signext:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    s_or_saveexec_b64 s[4:5], -1
+; GFX9-NEXT:    buffer_store_dword v40, off, s[0:3], s32 ; 4-byte Folded Spill
+; GFX9-NEXT:    s_mov_b64 exec, s[4:5]
+; GFX9-NEXT:    v_writelane_b32 v40, s33, 2
+; GFX9-NEXT:    v_writelane_b32 v40, s30, 0
+; GFX9-NEXT:    v_writelane_b32 v40, s31, 1
+; GFX9-NEXT:    global_load_sbyte v0, v[0:1], off
+; GFX9-NEXT:    s_mov_b32 s33, s32
+; GFX9-NEXT:    s_add_u32 s32, s32, 0x400
+; GFX9-NEXT:    s_getpc_b64 s[4:5]
+; GFX9-NEXT:    s_add_u32 s4, s4, external_void_func_i8_signext@rel32@lo+4
+; GFX9-NEXT:    s_addc_u32 s5, s5, external_void_func_i8_signext@rel32@hi+12
+; GFX9-NEXT:    s_swappc_b64 s[30:31], s[4:5]
+; GFX9-NEXT:    v_readlane_b32 s4, v40, 0
+; GFX9-NEXT:    v_readlane_b32 s5, v40, 1
+; GFX9-NEXT:    s_sub_u32 s32, s32, 0x400
+; GFX9-NEXT:    v_readlane_b32 s33, v40, 2
+; GFX9-NEXT:    s_or_saveexec_b64 s[6:7], -1
+; GFX9-NEXT:    buffer_load_dword v40, off, s[0:3], s32 ; 4-byte Folded Reload
+; GFX9-NEXT:    s_mov_b64 exec, s[6:7]
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-NEXT:    s_setpc_b64 s[4:5]
+;
+; GFX10-LABEL: test_call_external_void_func_i8_signext:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
+; GFX10-NEXT:    s_or_saveexec_b32 s4, -1
+; GFX10-NEXT:    buffer_store_dword v40, off, s[0:3], s32 ; 4-byte Folded Spill
+; GFX10-NEXT:    s_waitcnt_depctr 0xffe3
+; GFX10-NEXT:    s_mov_b32 exec_lo, s4
+; GFX10-NEXT:    v_writelane_b32 v40, s33, 2
+; GFX10-NEXT:    s_mov_b32 s33, s32
+; GFX10-NEXT:    s_add_u32 s32, s32, 0x200
+; GFX10-NEXT:    ; implicit-def: $vcc_hi
+; GFX10-NEXT:    s_getpc_b64 s[4:5]
+; GFX10-NEXT:    s_add_u32 s4, s4, external_void_func_i8_signext@rel32@lo+4
+; GFX10-NEXT:    s_addc_u32 s5, s5, external_void_func_i8_signext@rel32@hi+12
+; GFX10-NEXT:    v_writelane_b32 v40, s30, 0
+; GFX10-NEXT:    v_writelane_b32 v40, s31, 1
+; GFX10-NEXT:    global_load_sbyte v0, v[0:1], off
+; GFX10-NEXT:    s_swappc_b64 s[30:31], s[4:5]
+; GFX10-NEXT:    v_readlane_b32 s4, v40, 0
+; GFX10-NEXT:    v_readlane_b32 s5, v40, 1
+; GFX10-NEXT:    s_sub_u32 s32, s32, 0x200
+; GFX10-NEXT:    v_readlane_b32 s33, v40, 2
+; GFX10-NEXT:    s_or_saveexec_b32 s6, -1
+; GFX10-NEXT:    buffer_load_dword v40, off, s[0:3], s32 ; 4-byte Folded Reload
+; GFX10-NEXT:    s_waitcnt_depctr 0xffe3
+; GFX10-NEXT:    s_mov_b32 exec_lo, s6
+; GFX10-NEXT:    s_waitcnt vmcnt(0)
+; GFX10-NEXT:    s_setpc_b64 s[4:5]
+  %var = load volatile i8, i8 addrspace(1)* undef
+  call amdgpu_gfx void @external_void_func_i8_signext(i8 %var)
+  ret void
+}
+
+define amdgpu_gfx void @test_call_external_void_func_i8_zeroext(i32) #0 {
+; GFX9-LABEL: test_call_external_void_func_i8_zeroext:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    s_or_saveexec_b64 s[4:5], -1
+; GFX9-NEXT:    buffer_store_dword v40, off, s[0:3], s32 ; 4-byte Folded Spill
+; GFX9-NEXT:    s_mov_b64 exec, s[4:5]
+; GFX9-NEXT:    v_writelane_b32 v40, s33, 2
+; GFX9-NEXT:    v_writelane_b32 v40, s30, 0
+; GFX9-NEXT:    v_writelane_b32 v40, s31, 1
+; GFX9-NEXT:    global_load_ubyte v0, v[0:1], off
+; GFX9-NEXT:    s_mov_b32 s33, s32
+; GFX9-NEXT:    s_add_u32 s32, s32, 0x400
+; GFX9-NEXT:    s_getpc_b64 s[4:5]
+; GFX9-NEXT:    s_add_u32 s4, s4, external_void_func_i8_zeroext@rel32@lo+4
+; GFX9-NEXT:    s_addc_u32 s5, s5, external_void_func_i8_zeroext@rel32@hi+12
+; GFX9-NEXT:    s_swappc_b64 s[30:31], s[4:5]
+; GFX9-NEXT:    v_readlane_b32 s4, v40, 0
+; GFX9-NEXT:    v_readlane_b32 s5, v40, 1
+; GFX9-NEXT:    s_sub_u32 s32, s32, 0x400
+; GFX9-NEXT:    v_readlane_b32 s33, v40, 2
+; GFX9-NEXT:    s_or_saveexec_b64 s[6:7], -1
+; GFX9-NEXT:    buffer_load_dword v40, off, s[0:3], s32 ; 4-byte Folded Reload
+; GFX9-NEXT:    s_mov_b64 exec, s[6:7]
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-NEXT:    s_setpc_b64 s[4:5]
+;
+; GFX10-LABEL: test_call_external_void_func_i8_zeroext:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
+; GFX10-NEXT:    s_or_saveexec_b32 s4, -1
+; GFX10-NEXT:    buffer_store_dword v40, off, s[0:3], s32 ; 4-byte Folded Spill
+; GFX10-NEXT:    s_waitcnt_depctr 0xffe3
+; GFX10-NEXT:    s_mov_b32 exec_lo, s4
+; GFX10-NEXT:    v_writelane_b32 v40, s33, 2
+; GFX10-NEXT:    s_mov_b32 s33, s32
+; GFX10-NEXT:    s_add_u32 s32, s32, 0x200
+; GFX10-NEXT:    ; implicit-def: $vcc_hi
+; GFX10-NEXT:    s_getpc_b64 s[4:5]
+; GFX10-NEXT:    s_add_u32 s4, s4, external_void_func_i8_zeroext@rel32@lo+4
+; GFX10-NEXT:    s_addc_u32 s5, s5, external_void_func_i8_zeroext@rel32@hi+12
+; GFX10-NEXT:    v_writelane_b32 v40, s30, 0
+; GFX10-NEXT:    v_writelane_b32 v40, s31, 1
+; GFX10-NEXT:    global_load_ubyte v0, v[0:1], off
+; GFX10-NEXT:    s_swappc_b64 s[30:31], s[4:5]
+; GFX10-NEXT:    v_readlane_b32 s4, v40, 0
+; GFX10-NEXT:    v_readlane_b32 s5, v40, 1
+; GFX10-NEXT:    s_sub_u32 s32, s32, 0x200
+; GFX10-NEXT:    v_readlane_b32 s33, v40, 2
+; GFX10-NEXT:    s_or_saveexec_b32 s6, -1
+; GFX10-NEXT:    buffer_load_dword v40, off, s[0:3], s32 ; 4-byte Folded Reload
+; GFX10-NEXT:    s_waitcnt_depctr 0xffe3
+; GFX10-NEXT:    s_mov_b32 exec_lo, s6
+; GFX10-NEXT:    s_waitcnt vmcnt(0)
+; GFX10-NEXT:    s_setpc_b64 s[4:5]
+  %var = load volatile i8, i8 addrspace(1)* undef
+  call amdgpu_gfx void @external_void_func_i8_zeroext(i8 %var)
+  ret void
+}
+
+define amdgpu_gfx void @test_call_external_void_func_i16_imm() #0 {
+; GFX9-LABEL: test_call_external_void_func_i16_imm:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    s_or_saveexec_b64 s[4:5], -1
+; GFX9-NEXT:    buffer_store_dword v40, off, s[0:3], s32 ; 4-byte Folded Spill
+; GFX9-NEXT:    s_mov_b64 exec, s[4:5]
+; GFX9-NEXT:    v_writelane_b32 v40, s33, 2
+; GFX9-NEXT:    v_writelane_b32 v40, s30, 0
+; GFX9-NEXT:    s_mov_b32 s33, s32
+; GFX9-NEXT:    s_add_u32 s32, s32, 0x400
+; GFX9-NEXT:    v_mov_b32_e32 v0, 0x7b
+; GFX9-NEXT:    s_getpc_b64 s[4:5]
+; GFX9-NEXT:    s_add_u32 s4, s4, external_void_func_i16@rel32@lo+4
+; GFX9-NEXT:    s_addc_u32 s5, s5, external_void_func_i16@rel32@hi+12
+; GFX9-NEXT:    v_writelane_b32 v40, s31, 1
+; GFX9-NEXT:    s_swappc_b64 s[30:31], s[4:5]
+; GFX9-NEXT:    v_readlane_b32 s4, v40, 0
+; GFX9-NEXT:    v_readlane_b32 s5, v40, 1
+; GFX9-NEXT:    s_sub_u32 s32, s32, 0x400
+; GFX9-NEXT:    v_readlane_b32 s33, v40, 2
+; GFX9-NEXT:    s_or_saveexec_b64 s[6:7], -1
+; GFX9-NEXT:    buffer_load_dword v40, off, s[0:3], s32 ; 4-byte Folded Reload
+; GFX9-NEXT:    s_mov_b64 exec, s[6:7]
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-NEXT:    s_setpc_b64 s[4:5]
+;
+; GFX10-LABEL: test_call_external_void_func_i16_imm:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
+; GFX10-NEXT:    s_or_saveexec_b32 s4, -1
+; GFX10-NEXT:    buffer_store_dword v40, off, s[0:3], s32 ; 4-byte Folded Spill
+; GFX10-NEXT:    s_waitcnt_depctr 0xffe3
+; GFX10-NEXT:    s_mov_b32 exec_lo, s4
+; GFX10-NEXT:    v_writelane_b32 v40, s33, 2
+; GFX10-NEXT:    v_mov_b32_e32 v0, 0x7b
+; GFX10-NEXT:    s_mov_b32 s33, s32
+; GFX10-NEXT:    s_add_u32 s32, s32, 0x200
+; GFX10-NEXT:    s_getpc_b64 s[4:5]
+; GFX10-NEXT:    s_add_u32 s4, s4, external_void_func_i16@rel32@lo+4
+; GFX10-NEXT:    s_addc_u32 s5, s5, external_void_func_i16@rel32@hi+12
+; GFX10-NEXT:    v_writelane_b32 v40, s30, 0
+; GFX10-NEXT:    ; implicit-def: $vcc_hi
+; GFX10-NEXT:    v_writelane_b32 v40, s31, 1
+; GFX10-NEXT:    s_swappc_b64 s[30:31], s[4:5]
+; GFX10-NEXT:    v_readlane_b32 s4, v40, 0
+; GFX10-NEXT:    v_readlane_b32 s5, v40, 1
+; GFX10-NEXT:    s_sub_u32 s32, s32, 0x200
+; GFX10-NEXT:    v_readlane_b32 s33, v40, 2
+; GFX10-NEXT:    s_or_saveexec_b32 s6, -1
+; GFX10-NEXT:    buffer_load_dword v40, off, s[0:3], s32 ; 4-byte Folded Reload
+; GFX10-NEXT:    s_waitcnt_depctr 0xffe3
+; GFX10-NEXT:    s_mov_b32 exec_lo, s6
+; GFX10-NEXT:    s_waitcnt vmcnt(0)
+; GFX10-NEXT:    s_setpc_b64 s[4:5]
+  call amdgpu_gfx void @external_void_func_i16(i16 123)
+  ret void
+}
+
+define amdgpu_gfx void @test_call_external_void_func_i16_signext(i32) #0 {
+; GFX9-LABEL: test_call_external_void_func_i16_signext:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    s_or_saveexec_b64 s[4:5], -1
+; GFX9-NEXT:    buffer_store_dword v40, off, s[0:3], s32 ; 4-byte Folded Spill
+; GFX9-NEXT:    s_mov_b64 exec, s[4:5]
+; GFX9-NEXT:    v_writelane_b32 v40, s33, 2
+; GFX9-NEXT:    v_writelane_b32 v40, s30, 0
+; GFX9-NEXT:    v_writelane_b32 v40, s31, 1
+; GFX9-NEXT:    global_load_ushort v0, v[0:1], off
+; GFX9-NEXT:    s_mov_b32 s33, s32
+; GFX9-NEXT:    s_add_u32 s32, s32, 0x400
+; GFX9-NEXT:    s_getpc_b64 s[4:5]
+; GFX9-NEXT:    s_add_u32 s4, s4, external_void_func_i16_signext@rel32@lo+4
+; GFX9-NEXT:    s_addc_u32 s5, s5, external_void_func_i16_signext@rel32@hi+12
+; GFX9-NEXT:    s_swappc_b64 s[30:31], s[4:5]
+; GFX9-NEXT:    v_readlane_b32 s4, v40, 0
+; GFX9-NEXT:    v_readlane_b32 s5, v40, 1
+; GFX9-NEXT:    s_sub_u32 s32, s32, 0x400
+; GFX9-NEXT:    v_readlane_b32 s33, v40, 2
+; GFX9-NEXT:    s_or_saveexec_b64 s[6:7], -1
+; GFX9-NEXT:    buffer_load_dword v40, off, s[0:3], s32 ; 4-byte Folded Reload
+; GFX9-NEXT:    s_mov_b64 exec, s[6:7]
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-NEXT:    s_setpc_b64 s[4:5]
+;
+; GFX10-LABEL: test_call_external_void_func_i16_signext:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
+; GFX10-NEXT:    s_or_saveexec_b32 s4, -1
+; GFX10-NEXT:    buffer_store_dword v40, off, s[0:3], s32 ; 4-byte Folded Spill
+; GFX10-NEXT:    s_waitcnt_depctr 0xffe3
+; GFX10-NEXT:    s_mov_b32 exec_lo, s4
+; GFX10-NEXT:    v_writelane_b32 v40, s33, 2
+; GFX10-NEXT:    s_mov_b32 s33, s32
+; GFX10-NEXT:    s_add_u32 s32, s32, 0x200
+; GFX10-NEXT:    ; implicit-def: $vcc_hi
+; GFX10-NEXT:    s_getpc_b64 s[4:5]
+; GFX10-NEXT:    s_add_u32 s4, s4, external_void_func_i16_signext@rel32@lo+4
+; GFX10-NEXT:    s_addc_u32 s5, s5, external_void_func_i16_signext@rel32@hi+12
+; GFX10-NEXT:    v_writelane_b32 v40, s30, 0
+; GFX10-NEXT:    v_writelane_b32 v40, s31, 1
+; GFX10-NEXT:    global_load_ushort v0, v[0:1], off
+; GFX10-NEXT:    s_swappc_b64 s[30:31], s[4:5]
+; GFX10-NEXT:    v_readlane_b32 s4, v40, 0
+; GFX10-NEXT:    v_readlane_b32 s5, v40, 1
+; GFX10-NEXT:    s_sub_u32 s32, s32, 0x200
+; GFX10-NEXT:    v_readlane_b32 s33, v40, 2
+; GFX10-NEXT:    s_or_saveexec_b32 s6, -1
+; GFX10-NEXT:    buffer_load_dword v40, off, s[0:3], s32 ; 4-byte Folded Reload
+; GFX10-NEXT:    s_waitcnt_depctr 0xffe3
+; GFX10-NEXT:    s_mov_b32 exec_lo, s6
+; GFX10-NEXT:    s_waitcnt vmcnt(0)
+; GFX10-NEXT:    s_setpc_b64 s[4:5]
+  %var = load volatile i16, i16 addrspace(1)* undef
+  call amdgpu_gfx void @external_void_func_i16_signext(i16 %var)
+  ret void
+}
+
+define amdgpu_gfx void @test_call_external_void_func_i16_zeroext(i32) #0 {
+; GFX9-LABEL: test_call_external_void_func_i16_zeroext:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    s_or_saveexec_b64 s[4:5], -1
+; GFX9-NEXT:    buffer_store_dword v40, off, s[0:3], s32 ; 4-byte Folded Spill
+; GFX9-NEXT:    s_mov_b64 exec, s[4:5]
+; GFX9-NEXT:    v_writelane_b32 v40, s33, 2
+; GFX9-NEXT:    v_writelane_b32 v40, s30, 0
+; GFX9-NEXT:    v_writelane_b32 v40, s31, 1
+; GFX9-NEXT:    global_load_ushort v0, v[0:1], off
+; GFX9-NEXT:    s_mov_b32 s33, s32
+; GFX9-NEXT:    s_add_u32 s32, s32, 0x400
+; GFX9-NEXT:    s_getpc_b64 s[4:5]
+; GFX9-NEXT:    s_add_u32 s4, s4, external_void_func_i16_zeroext@rel32@lo+4
+; GFX9-NEXT:    s_addc_u32 s5, s5, external_void_func_i16_zeroext@rel32@hi+12
+; GFX9-NEXT:    s_swappc_b64 s[30:31], s[4:5]
+; GFX9-NEXT:    v_readlane_b32 s4, v40, 0
+; GFX9-NEXT:    v_readlane_b32 s5, v40, 1
+; GFX9-NEXT:    s_sub_u32 s32, s32, 0x400
+; GFX9-NEXT:    v_readlane_b32 s33, v40, 2
+; GFX9-NEXT:    s_or_saveexec_b64 s[6:7], -1
+; GFX9-NEXT:    buffer_load_dword v40, off, s[0:3], s32 ; 4-byte Folded Reload
+; GFX9-NEXT:    s_mov_b64 exec, s[6:7]
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-NEXT:    s_setpc_b64 s[4:5]
+;
+; GFX10-LABEL: test_call_external_void_func_i16_zeroext:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
+; GFX10-NEXT:    s_or_saveexec_b32 s4, -1
+; GFX10-NEXT:    buffer_store_dword v40, off, s[0:3], s32 ; 4-byte Folded Spill
+; GFX10-NEXT:    s_waitcnt_depctr 0xffe3
+; GFX10-NEXT:    s_mov_b32 exec_lo, s4
+; GFX10-NEXT:    v_writelane_b32 v40, s33, 2
+; GFX10-NEXT:    s_mov_b32 s33, s32
+; GFX10-NEXT:    s_add_u32 s32, s32, 0x200
+; GFX10-NEXT:    ; implicit-def: $vcc_hi
+; GFX10-NEXT:    s_getpc_b64 s[4:5]
+; GFX10-NEXT:    s_add_u32 s4, s4, external_void_func_i16_zeroext@rel32@lo+4
+; GFX10-NEXT:    s_addc_u32 s5, s5, external_void_func_i16_zeroext@rel32@hi+12
+; GFX10-NEXT:    v_writelane_b32 v40, s30, 0
+; GFX10-NEXT:    v_writelane_b32 v40, s31, 1
+; GFX10-NEXT:    global_load_ushort v0, v[0:1], off
+; GFX10-NEXT:    s_swappc_b64 s[30:31], s[4:5]
+; GFX10-NEXT:    v_readlane_b32 s4, v40, 0
+; GFX10-NEXT:    v_readlane_b32 s5, v40, 1
+; GFX10-NEXT:    s_sub_u32 s32, s32, 0x200
+; GFX10-NEXT:    v_readlane_b32 s33, v40, 2
+; GFX10-NEXT:    s_or_saveexec_b32 s6, -1
+; GFX10-NEXT:    buffer_load_dword v40, off, s[0:3], s32 ; 4-byte Folded Reload
+; GFX10-NEXT:    s_waitcnt_depctr 0xffe3
+; GFX10-NEXT:    s_mov_b32 exec_lo, s6
+; GFX10-NEXT:    s_waitcnt vmcnt(0)
+; GFX10-NEXT:    s_setpc_b64 s[4:5]
+  %var = load volatile i16, i16 addrspace(1)* undef
+  call amdgpu_gfx void @external_void_func_i16_zeroext(i16 %var)
+  ret void
+}
+
+define amdgpu_gfx void @test_call_external_void_func_i32_imm(i32) #0 {
+; GFX9-LABEL: test_call_external_void_func_i32_imm:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    s_or_saveexec_b64 s[4:5], -1
+; GFX9-NEXT:    buffer_store_dword v40, off, s[0:3], s32 ; 4-byte Folded Spill
+; GFX9-NEXT:    s_mov_b64 exec, s[4:5]
+; GFX9-NEXT:    v_writelane_b32 v40, s33, 2
+; GFX9-NEXT:    v_writelane_b32 v40, s30, 0
+; GFX9-NEXT:    s_mov_b32 s33, s32
+; GFX9-NEXT:    s_add_u32 s32, s32, 0x400
+; GFX9-NEXT:    v_mov_b32_e32 v0, 42
+; GFX9-NEXT:    s_getpc_b64 s[4:5]
+; GFX9-NEXT:    s_add_u32 s4, s4, external_void_func_i32@rel32@lo+4
+; GFX9-NEXT:    s_addc_u32 s5, s5, external_void_func_i32@rel32@hi+12
+; GFX9-NEXT:    v_writelane_b32 v40, s31, 1
+; GFX9-NEXT:    s_swappc_b64 s[30:31], s[4:5]
+; GFX9-NEXT:    v_readlane_b32 s4, v40, 0
+; GFX9-NEXT:    v_readlane_b32 s5, v40, 1
+; GFX9-NEXT:    s_sub_u32 s32, s32, 0x400
+; GFX9-NEXT:    v_readlane_b32 s33, v40, 2
+; GFX9-NEXT:    s_or_saveexec_b64 s[6:7], -1
+; GFX9-NEXT:    buffer_load_dword v40, off, s[0:3], s32 ; 4-byte Folded Reload
+; GFX9-NEXT:    s_mov_b64 exec, s[6:7]
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-NEXT:    s_setpc_b64 s[4:5]
+;
+; GFX10-LABEL: test_call_external_void_func_i32_imm:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
+; GFX10-NEXT:    s_or_saveexec_b32 s4, -1
+; GFX10-NEXT:    buffer_store_dword v40, off, s[0:3], s32 ; 4-byte Folded Spill
+; GFX10-NEXT:    s_waitcnt_depctr 0xffe3
+; GFX10-NEXT:    s_mov_b32 exec_lo, s4
+; GFX10-NEXT:    v_writelane_b32 v40, s33, 2
+; GFX10-NEXT:    v_mov_b32_e32 v0, 42
+; GFX10-NEXT:    s_mov_b32 s33, s32
+; GFX10-NEXT:    s_add_u32 s32, s32, 0x200
+; GFX10-NEXT:    s_getpc_b64 s[4:5]
+; GFX10-NEXT:    s_add_u32 s4, s4, external_void_func_i32@rel32@lo+4
+; GFX10-NEXT:    s_addc_u32 s5, s5, external_void_func_i32@rel32@hi+12
+; GFX10-NEXT:    v_writelane_b32 v40, s30, 0
+; GFX10-NEXT:    ; implicit-def: $vcc_hi
+; GFX10-NEXT:    v_writelane_b32 v40, s31, 1
+; GFX10-NEXT:    s_swappc_b64 s[30:31], s[4:5]
+; GFX10-NEXT:    v_readlane_b32 s4, v40, 0
+; GFX10-NEXT:    v_readlane_b32 s5, v40, 1
+; GFX10-NEXT:    s_sub_u32 s32, s32, 0x200
+; GFX10-NEXT:    v_readlane_b32 s33, v40, 2
+; GFX10-NEXT:    s_or_saveexec_b32 s6, -1
+; GFX10-NEXT:    buffer_load_dword v40, off, s[0:3], s32 ; 4-byte Folded Reload
+; GFX10-NEXT:    s_waitcnt_depctr 0xffe3
+; GFX10-NEXT:    s_mov_b32 exec_lo, s6
+; GFX10-NEXT:    s_waitcnt vmcnt(0)
+; GFX10-NEXT:    s_setpc_b64 s[4:5]
+  call amdgpu_gfx void @external_void_func_i32(i32 42)
+  ret void
+}
+
+define amdgpu_gfx void @test_call_external_void_func_i64_imm() #0 {
+; GFX9-LABEL: test_call_external_void_func_i64_imm:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    s_or_saveexec_b64 s[4:5], -1
+; GFX9-NEXT:    buffer_store_dword v40, off, s[0:3], s32 ; 4-byte Folded Spill
+; GFX9-NEXT:    s_mov_b64 exec, s[4:5]
+; GFX9-NEXT:    v_writelane_b32 v40, s33, 2
+; GFX9-NEXT:    v_writelane_b32 v40, s30, 0
+; GFX9-NEXT:    s_mov_b32 s33, s32
+; GFX9-NEXT:    s_add_u32 s32, s32, 0x400
+; GFX9-NEXT:    v_mov_b32_e32 v0, 0x7b
+; GFX9-NEXT:    v_mov_b32_e32 v1, 0
+; GFX9-NEXT:    s_getpc_b64 s[4:5]
+; GFX9-NEXT:    s_add_u32 s4, s4, external_void_func_i64@rel32@lo+4
+; GFX9-NEXT:    s_addc_u32 s5, s5, external_void_func_i64@rel32@hi+12
+; GFX9-NEXT:    v_writelane_b32 v40, s31, 1
+; GFX9-NEXT:    s_swappc_b64 s[30:31], s[4:5]
+; GFX9-NEXT:    v_readlane_b32 s4, v40, 0
+; GFX9-NEXT:    v_readlane_b32 s5, v40, 1
+; GFX9-NEXT:    s_sub_u32 s32, s32, 0x400
+; GFX9-NEXT:    v_readlane_b32 s33, v40, 2
+; GFX9-NEXT:    s_or_saveexec_b64 s[6:7], -1
+; GFX9-NEXT:    buffer_load_dword v40, off, s[0:3], s32 ; 4-byte Folded Reload
+; GFX9-NEXT:    s_mov_b64 exec, s[6:7]
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-NEXT:    s_setpc_b64 s[4:5]
+;
+; GFX10-LABEL: test_call_external_void_func_i64_imm:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
+; GFX10-NEXT:    s_or_saveexec_b32 s4, -1
+; GFX10-NEXT:    buffer_store_dword v40, off, s[0:3], s32 ; 4-byte Folded Spill
+; GFX10-NEXT:    s_waitcnt_depctr 0xffe3
+; GFX10-NEXT:    s_mov_b32 exec_lo, s4
+; GFX10-NEXT:    v_writelane_b32 v40, s33, 2
+; GFX10-NEXT:    v_mov_b32_e32 v0, 0x7b
+; GFX10-NEXT:    v_mov_b32_e32 v1, 0
+; GFX10-NEXT:    s_mov_b32 s33, s32
+; GFX10-NEXT:    s_add_u32 s32, s32, 0x200
+; GFX10-NEXT:    v_writelane_b32 v40, s30, 0
+; GFX10-NEXT:    s_getpc_b64 s[4:5]
+; GFX10-NEXT:    s_add_u32 s4, s4, external_void_func_i64@rel32@lo+4
+; GFX10-NEXT:    s_addc_u32 s5, s5, external_void_func_i64@rel32@hi+12
+; GFX10-NEXT:    ; implicit-def: $vcc_hi
+; GFX10-NEXT:    v_writelane_b32 v40, s31, 1
+; GFX10-NEXT:    s_swappc_b64 s[30:31], s[4:5]
+; GFX10-NEXT:    v_readlane_b32 s4, v40, 0
+; GFX10-NEXT:    v_readlane_b32 s5, v40, 1
+; GFX10-NEXT:    s_sub_u32 s32, s32, 0x200
+; GFX10-NEXT:    v_readlane_b32 s33, v40, 2
+; GFX10-NEXT:    s_or_saveexec_b32 s6, -1
+; GFX10-NEXT:    buffer_load_dword v40, off, s[0:3], s32 ; 4-byte Folded Reload
+; GFX10-NEXT:    s_waitcnt_depctr 0xffe3
+; GFX10-NEXT:    s_mov_b32 exec_lo, s6
+; GFX10-NEXT:    s_waitcnt vmcnt(0)
+; GFX10-NEXT:    s_setpc_b64 s[4:5]
+  call amdgpu_gfx void @external_void_func_i64(i64 123)
+  ret void
+}
+
+define amdgpu_gfx void @test_call_external_void_func_v2i64() #0 {
+; GFX9-LABEL: test_call_external_void_func_v2i64:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    s_or_saveexec_b64 s[4:5], -1
+; GFX9-NEXT:    buffer_store_dword v40, off, s[0:3], s32 ; 4-byte Folded Spill
+; GFX9-NEXT:    s_mov_b64 exec, s[4:5]
+; GFX9-NEXT:    v_writelane_b32 v40, s33, 2
+; GFX9-NEXT:    v_mov_b32_e32 v0, 0
+; GFX9-NEXT:    v_writelane_b32 v40, s30, 0
+; GFX9-NEXT:    v_mov_b32_e32 v1, 0
+; GFX9-NEXT:    v_writelane_b32 v40, s31, 1
+; GFX9-NEXT:    global_load_dwordx4 v[0:3], v[0:1], off
+; GFX9-NEXT:    s_mov_b32 s33, s32
+; GFX9-NEXT:    s_add_u32 s32, s32, 0x400
+; GFX9-NEXT:    s_getpc_b64 s[4:5]
+; GFX9-NEXT:    s_add_u32 s4, s4, external_void_func_v2i64@rel32@lo+4
+; GFX9-NEXT:    s_addc_u32 s5, s5, external_void_func_v2i64@rel32@hi+12
+; GFX9-NEXT:    s_swappc_b64 s[30:31], s[4:5]
+; GFX9-NEXT:    v_readlane_b32 s4, v40, 0
+; GFX9-NEXT:    v_readlane_b32 s5, v40, 1
+; GFX9-NEXT:    s_sub_u32 s32, s32, 0x400
+; GFX9-NEXT:    v_readlane_b32 s33, v40, 2
+; GFX9-NEXT:    s_or_saveexec_b64 s[6:7], -1
+; GFX9-NEXT:    buffer_load_dword v40, off, s[0:3], s32 ; 4-byte Folded Reload
+; GFX9-NEXT:    s_mov_b64 exec, s[6:7]
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-NEXT:    s_setpc_b64 s[4:5]
+;
+; GFX10-LABEL: test_call_external_void_func_v2i64:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
+; GFX10-NEXT:    s_or_saveexec_b32 s4, -1
+; GFX10-NEXT:    buffer_store_dword v40, off, s[0:3], s32 ; 4-byte Folded Spill
+; GFX10-NEXT:    s_waitcnt_depctr 0xffe3
+; GFX10-NEXT:    s_mov_b32 exec_lo, s4
+; GFX10-NEXT:    v_writelane_b32 v40, s33, 2
+; GFX10-NEXT:    v_mov_b32_e32 v0, 0
+; GFX10-NEXT:    v_mov_b32_e32 v1, 0
+; GFX10-NEXT:    s_mov_b32 s33, s32
+; GFX10-NEXT:    s_add_u32 s32, s32, 0x200
+; GFX10-NEXT:    v_writelane_b32 v40, s30, 0
+; GFX10-NEXT:    ; implicit-def: $vcc_hi
+; GFX10-NEXT:    s_getpc_b64 s[4:5]
+; GFX10-NEXT:    s_add_u32 s4, s4, external_void_func_v2i64@rel32@lo+4
+; GFX10-NEXT:    s_addc_u32 s5, s5, external_void_func_v2i64@rel32@hi+12
+; GFX10-NEXT:    v_writelane_b32 v40, s31, 1
+; GFX10-NEXT:    global_load_dwordx4 v[0:3], v[0:1], off
+; GFX10-NEXT:    s_swappc_b64 s[30:31], s[4:5]
+; GFX10-NEXT:    v_readlane_b32 s4, v40, 0
+; GFX10-NEXT:    v_readlane_b32 s5, v40, 1
+; GFX10-NEXT:    s_sub_u32 s32, s32, 0x200
+; GFX10-NEXT:    v_readlane_b32 s33, v40, 2
+; GFX10-NEXT:    s_or_saveexec_b32 s6, -1
+; GFX10-NEXT:    buffer_load_dword v40, off, s[0:3], s32 ; 4-byte Folded Reload
+; GFX10-NEXT:    s_waitcnt_depctr 0xffe3
+; GFX10-NEXT:    s_mov_b32 exec_lo, s6
+; GFX10-NEXT:    s_waitcnt vmcnt(0)
+; GFX10-NEXT:    s_setpc_b64 s[4:5]
+  %val = load <2 x i64>, <2 x i64> addrspace(1)* null
+  call amdgpu_gfx void @external_void_func_v2i64(<2 x i64> %val)
+  ret void
+}
+
+define amdgpu_gfx void @test_call_external_void_func_v2i64_imm() #0 {
+; GFX9-LABEL: test_call_external_void_func_v2i64_imm:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    s_or_saveexec_b64 s[4:5], -1
+; GFX9-NEXT:    buffer_store_dword v40, off, s[0:3], s32 ; 4-byte Folded Spill
+; GFX9-NEXT:    s_mov_b64 exec, s[4:5]
+; GFX9-NEXT:    v_writelane_b32 v40, s33, 2
+; GFX9-NEXT:    v_writelane_b32 v40, s30, 0
+; GFX9-NEXT:    s_mov_b32 s33, s32
+; GFX9-NEXT:    s_add_u32 s32, s32, 0x400
+; GFX9-NEXT:    v_mov_b32_e32 v0, 1
+; GFX9-NEXT:    v_mov_b32_e32 v1, 2
+; GFX9-NEXT:    v_mov_b32_e32 v2, 3
+; GFX9-NEXT:    v_mov_b32_e32 v3, 4
+; GFX9-NEXT:    s_getpc_b64 s[4:5]
+; GFX9-NEXT:    s_add_u32 s4, s4, external_void_func_v2i64@rel32@lo+4
+; GFX9-NEXT:    s_addc_u32 s5, s5, external_void_func_v2i64@rel32@hi+12
+; GFX9-NEXT:    v_writelane_b32 v40, s31, 1
+; GFX9-NEXT:    s_swappc_b64 s[30:31], s[4:5]
+; GFX9-NEXT:    v_readlane_b32 s4, v40, 0
+; GFX9-NEXT:    v_readlane_b32 s5, v40, 1
+; GFX9-NEXT:    s_sub_u32 s32, s32, 0x400
+; GFX9-NEXT:    v_readlane_b32 s33, v40, 2
+; GFX9-NEXT:    s_or_saveexec_b64 s[6:7], -1
+; GFX9-NEXT:    buffer_load_dword v40, off, s[0:3], s32 ; 4-byte Folded Reload
+; GFX9-NEXT:    s_mov_b64 exec, s[6:7]
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-NEXT:    s_setpc_b64 s[4:5]
+;
+; GFX10-LABEL: test_call_external_void_func_v2i64_imm:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
+; GFX10-NEXT:    s_or_saveexec_b32 s4, -1
+; GFX10-NEXT:    buffer_store_dword v40, off, s[0:3], s32 ; 4-byte Folded Spill
+; GFX10-NEXT:    s_waitcnt_depctr 0xffe3
+; GFX10-NEXT:    s_mov_b32 exec_lo, s4
+; GFX10-NEXT:    v_writelane_b32 v40, s33, 2
+; GFX10-NEXT:    v_mov_b32_e32 v0, 1
+; GFX10-NEXT:    v_mov_b32_e32 v1, 2
+; GFX10-NEXT:    v_mov_b32_e32 v2, 3
+; GFX10-NEXT:    v_mov_b32_e32 v3, 4
+; GFX10-NEXT:    v_writelane_b32 v40, s30, 0
+; GFX10-NEXT:    s_mov_b32 s33, s32
+; GFX10-NEXT:    s_add_u32 s32, s32, 0x200
+; GFX10-NEXT:    s_getpc_b64 s[4:5]
+; GFX10-NEXT:    s_add_u32 s4, s4, external_void_func_v2i64@rel32@lo+4
+; GFX10-NEXT:    s_addc_u32 s5, s5, external_void_func_v2i64@rel32@hi+12
+; GFX10-NEXT:    ; implicit-def: $vcc_hi
+; GFX10-NEXT:    v_writelane_b32 v40, s31, 1
+; GFX10-NEXT:    s_swappc_b64 s[30:31], s[4:5]
+; GFX10-NEXT:    v_readlane_b32 s4, v40, 0
+; GFX10-NEXT:    v_readlane_b32 s5, v40, 1
+; GFX10-NEXT:    s_sub_u32 s32, s32, 0x200
+; GFX10-NEXT:    v_readlane_b32 s33, v40, 2
+; GFX10-NEXT:    s_or_saveexec_b32 s6, -1
+; GFX10-NEXT:    buffer_load_dword v40, off, s[0:3], s32 ; 4-byte Folded Reload
+; GFX10-NEXT:    s_waitcnt_depctr 0xffe3
+; GFX10-NEXT:    s_mov_b32 exec_lo, s6
+; GFX10-NEXT:    s_waitcnt vmcnt(0)
+; GFX10-NEXT:    s_setpc_b64 s[4:5]
+  call amdgpu_gfx void @external_void_func_v2i64(<2 x i64> <i64 8589934593, i64 17179869187>)
+  ret void
+}
+
+define amdgpu_gfx void @test_call_external_void_func_v3i64() #0 {
+; GFX9-LABEL: test_call_external_void_func_v3i64:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    s_or_saveexec_b64 s[4:5], -1
+; GFX9-NEXT:    buffer_store_dword v40, off, s[0:3], s32 ; 4-byte Folded Spill
+; GFX9-NEXT:    s_mov_b64 exec, s[4:5]
+; GFX9-NEXT:    v_writelane_b32 v40, s33, 2
+; GFX9-NEXT:    v_mov_b32_e32 v0, 0
+; GFX9-NEXT:    v_writelane_b32 v40, s30, 0
+; GFX9-NEXT:    v_mov_b32_e32 v1, 0
+; GFX9-NEXT:    v_writelane_b32 v40, s31, 1
+; GFX9-NEXT:    global_load_dwordx4 v[0:3], v[0:1], off
+; GFX9-NEXT:    s_mov_b32 s33, s32
+; GFX9-NEXT:    s_add_u32 s32, s32, 0x400
+; GFX9-NEXT:    v_mov_b32_e32 v4, 1
+; GFX9-NEXT:    v_mov_b32_e32 v5, 2
+; GFX9-NEXT:    s_getpc_b64 s[4:5]
+; GFX9-NEXT:    s_add_u32 s4, s4, external_void_func_v3i64@rel32@lo+4
+; GFX9-NEXT:    s_addc_u32 s5, s5, external_void_func_v3i64@rel32@hi+12
+; GFX9-NEXT:    s_swappc_b64 s[30:31], s[4:5]
+; GFX9-NEXT:    v_readlane_b32 s4, v40, 0
+; GFX9-NEXT:    v_readlane_b32 s5, v40, 1
+; GFX9-NEXT:    s_sub_u32 s32, s32, 0x400
+; GFX9-NEXT:    v_readlane_b32 s33, v40, 2
+; GFX9-NEXT:    s_or_saveexec_b64 s[6:7], -1
+; GFX9-NEXT:    buffer_load_dword v40, off, s[0:3], s32 ; 4-byte Folded Reload
+; GFX9-NEXT:    s_mov_b64 exec, s[6:7]
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-NEXT:    s_setpc_b64 s[4:5]
+;
+; GFX10-LABEL: test_call_external_void_func_v3i64:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
+; GFX10-NEXT:    s_or_saveexec_b32 s4, -1
+; GFX10-NEXT:    buffer_store_dword v40, off, s[0:3], s32 ; 4-byte Folded Spill
+; GFX10-NEXT:    s_waitcnt_depctr 0xffe3
+; GFX10-NEXT:    s_mov_b32 exec_lo, s4
+; GFX10-NEXT:    v_writelane_b32 v40, s33, 2
+; GFX10-NEXT:    v_mov_b32_e32 v0, 0
+; GFX10-NEXT:    v_mov_b32_e32 v1, 0
+; GFX10-NEXT:    v_mov_b32_e32 v4, 1
+; GFX10-NEXT:    v_mov_b32_e32 v5, 2
+; GFX10-NEXT:    v_writelane_b32 v40, s30, 0
+; GFX10-NEXT:    s_mov_b32 s33, s32
+; GFX10-NEXT:    s_add_u32 s32, s32, 0x200
+; GFX10-NEXT:    ; implicit-def: $vcc_hi
+; GFX10-NEXT:    s_getpc_b64 s[4:5]
+; GFX10-NEXT:    s_add_u32 s4, s4, external_void_func_v3i64@rel32@lo+4
+; GFX10-NEXT:    s_addc_u32 s5, s5, external_void_func_v3i64@rel32@hi+12
+; GFX10-NEXT:    v_writelane_b32 v40, s31, 1
+; GFX10-NEXT:    global_load_dwordx4 v[0:3], v[0:1], off
+; GFX10-NEXT:    s_swappc_b64 s[30:31], s[4:5]
+; GFX10-NEXT:    v_readlane_b32 s4, v40, 0
+; GFX10-NEXT:    v_readlane_b32 s5, v40, 1
+; GFX10-NEXT:    s_sub_u32 s32, s32, 0x200
+; GFX10-NEXT:    v_readlane_b32 s33, v40, 2
+; GFX10-NEXT:    s_or_saveexec_b32 s6, -1
+; GFX10-NEXT:    buffer_load_dword v40, off, s[0:3], s32 ; 4-byte Folded Reload
+; GFX10-NEXT:    s_waitcnt_depctr 0xffe3
+; GFX10-NEXT:    s_mov_b32 exec_lo, s6
+; GFX10-NEXT:    s_waitcnt vmcnt(0)
+; GFX10-NEXT:    s_setpc_b64 s[4:5]
+  %load = load <2 x i64>, <2 x i64> addrspace(1)* null
+  %val = shufflevector <2 x i64> %load, <2 x i64> <i64 8589934593, i64 undef>, <3 x i32> <i32 0, i32 1, i32 2>
+
+  call amdgpu_gfx void @external_void_func_v3i64(<3 x i64> %val)
+  ret void
+}
+
+define amdgpu_gfx void @test_call_external_void_func_v4i64() #0 {
+; GFX9-LABEL: test_call_external_void_func_v4i64:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    s_or_saveexec_b64 s[4:5], -1
+; GFX9-NEXT:    buffer_store_dword v40, off, s[0:3], s32 ; 4-byte Folded Spill
+; GFX9-NEXT:    s_mov_b64 exec, s[4:5]
+; GFX9-NEXT:    v_writelane_b32 v40, s33, 2
+; GFX9-NEXT:    v_mov_b32_e32 v0, 0
+; GFX9-NEXT:    v_writelane_b32 v40, s30, 0
+; GFX9-NEXT:    v_mov_b32_e32 v1, 0
+; GFX9-NEXT:    v_writelane_b32 v40, s31, 1
+; GFX9-NEXT:    global_load_dwordx4 v[0:3], v[0:1], off
+; GFX9-NEXT:    s_mov_b32 s33, s32
+; GFX9-NEXT:    s_add_u32 s32, s32, 0x400
+; GFX9-NEXT:    v_mov_b32_e32 v4, 1
+; GFX9-NEXT:    v_mov_b32_e32 v5, 2
+; GFX9-NEXT:    v_mov_b32_e32 v6, 3
+; GFX9-NEXT:    v_mov_b32_e32 v7, 4
+; GFX9-NEXT:    s_getpc_b64 s[4:5]
+; GFX9-NEXT:    s_add_u32 s4, s4, external_void_func_v4i64@rel32@lo+4
+; GFX9-NEXT:    s_addc_u32 s5, s5, external_void_func_v4i64@rel32@hi+12
+; GFX9-NEXT:    s_swappc_b64 s[30:31], s[4:5]
+; GFX9-NEXT:    v_readlane_b32 s4, v40, 0
+; GFX9-NEXT:    v_readlane_b32 s5, v40, 1
+; GFX9-NEXT:    s_sub_u32 s32, s32, 0x400
+; GFX9-NEXT:    v_readlane_b32 s33, v40, 2
+; GFX9-NEXT:    s_or_saveexec_b64 s[6:7], -1
+; GFX9-NEXT:    buffer_load_dword v40, off, s[0:3], s32 ; 4-byte Folded Reload
+; GFX9-NEXT:    s_mov_b64 exec, s[6:7]
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-NEXT:    s_setpc_b64 s[4:5]
+;
+; GFX10-LABEL: test_call_external_void_func_v4i64:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
+; GFX10-NEXT:    s_or_saveexec_b32 s4, -1
+; GFX10-NEXT:    buffer_store_dword v40, off, s[0:3], s32 ; 4-byte Folded Spill
+; GFX10-NEXT:    s_waitcnt_depctr 0xffe3
+; GFX10-NEXT:    s_mov_b32 exec_lo, s4
+; GFX10-NEXT:    v_writelane_b32 v40, s33, 2
+; GFX10-NEXT:    v_mov_b32_e32 v0, 0
+; GFX10-NEXT:    v_mov_b32_e32 v1, 0
+; GFX10-NEXT:    v_mov_b32_e32 v4, 1
+; GFX10-NEXT:    v_mov_b32_e32 v5, 2
+; GFX10-NEXT:    v_writelane_b32 v40, s30, 0
+; GFX10-NEXT:    v_mov_b32_e32 v6, 3
+; GFX10-NEXT:    v_mov_b32_e32 v7, 4
+; GFX10-NEXT:    s_mov_b32 s33, s32
+; GFX10-NEXT:    s_add_u32 s32, s32, 0x200
+; GFX10-NEXT:    v_writelane_b32 v40, s31, 1
+; GFX10-NEXT:    global_load_dwordx4 v[0:3], v[0:1], off
+; GFX10-NEXT:    s_getpc_b64 s[4:5]
+; GFX10-NEXT:    s_add_u32 s4, s4, external_void_func_v4i64@rel32@lo+4
+; GFX10-NEXT:    s_addc_u32 s5, s5, external_void_func_v4i64@rel32@hi+12
+; GFX10-NEXT:    ; implicit-def: $vcc_hi
+; GFX10-NEXT:    s_swappc_b64 s[30:31], s[4:5]
+; GFX10-NEXT:    v_readlane_b32 s4, v40, 0
+; GFX10-NEXT:    v_readlane_b32 s5, v40, 1
+; GFX10-NEXT:    s_sub_u32 s32, s32, 0x200
+; GFX10-NEXT:    v_readlane_b32 s33, v40, 2
+; GFX10-NEXT:    s_or_saveexec_b32 s6, -1
+; GFX10-NEXT:    buffer_load_dword v40, off, s[0:3], s32 ; 4-byte Folded Reload
+; GFX10-NEXT:    s_waitcnt_depctr 0xffe3
+; GFX10-NEXT:    s_mov_b32 exec_lo, s6
+; GFX10-NEXT:    s_waitcnt vmcnt(0)
+; GFX10-NEXT:    s_setpc_b64 s[4:5]
+  %load = load <2 x i64>, <2 x i64> addrspace(1)* null
+  %val = shufflevector <2 x i64> %load, <2 x i64> <i64 8589934593, i64 17179869187>, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+  call amdgpu_gfx void @external_void_func_v4i64(<4 x i64> %val)
+  ret void
+}
+
+define amdgpu_gfx void @test_call_external_void_func_f16_imm() #0 {
+; GFX9-LABEL: test_call_external_void_func_f16_imm:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    s_or_saveexec_b64 s[4:5], -1
+; GFX9-NEXT:    buffer_store_dword v40, off, s[0:3], s32 ; 4-byte Folded Spill
+; GFX9-NEXT:    s_mov_b64 exec, s[4:5]
+; GFX9-NEXT:    v_writelane_b32 v40, s33, 2
+; GFX9-NEXT:    v_writelane_b32 v40, s30, 0
+; GFX9-NEXT:    s_mov_b32 s33, s32
+; GFX9-NEXT:    s_add_u32 s32, s32, 0x400
+; GFX9-NEXT:    v_mov_b32_e32 v0, 0x4400
+; GFX9-NEXT:    s_getpc_b64 s[4:5]
+; GFX9-NEXT:    s_add_u32 s4, s4, external_void_func_f16@rel32@lo+4
+; GFX9-NEXT:    s_addc_u32 s5, s5, external_void_func_f16@rel32@hi+12
+; GFX9-NEXT:    v_writelane_b32 v40, s31, 1
+; GFX9-NEXT:    s_swappc_b64 s[30:31], s[4:5]
+; GFX9-NEXT:    v_readlane_b32 s4, v40, 0
+; GFX9-NEXT:    v_readlane_b32 s5, v40, 1
+; GFX9-NEXT:    s_sub_u32 s32, s32, 0x400
+; GFX9-NEXT:    v_readlane_b32 s33, v40, 2
+; GFX9-NEXT:    s_or_saveexec_b64 s[6:7], -1
+; GFX9-NEXT:    buffer_load_dword v40, off, s[0:3], s32 ; 4-byte Folded Reload
+; GFX9-NEXT:    s_mov_b64 exec, s[6:7]
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-NEXT:    s_setpc_b64 s[4:5]
+;
+; GFX10-LABEL: test_call_external_void_func_f16_imm:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
+; GFX10-NEXT:    s_or_saveexec_b32 s4, -1
+; GFX10-NEXT:    buffer_store_dword v40, off, s[0:3], s32 ; 4-byte Folded Spill
+; GFX10-NEXT:    s_waitcnt_depctr 0xffe3
+; GFX10-NEXT:    s_mov_b32 exec_lo, s4
+; GFX10-NEXT:    v_writelane_b32 v40, s33, 2
+; GFX10-NEXT:    v_mov_b32_e32 v0, 0x4400
+; GFX10-NEXT:    s_mov_b32 s33, s32
+; GFX10-NEXT:    s_add_u32 s32, s32, 0x200
+; GFX10-NEXT:    s_getpc_b64 s[4:5]
+; GFX10-NEXT:    s_add_u32 s4, s4, external_void_func_f16@rel32@lo+4
+; GFX10-NEXT:    s_addc_u32 s5, s5, external_void_func_f16@rel32@hi+12
+; GFX10-NEXT:    v_writelane_b32 v40, s30, 0
+; GFX10-NEXT:    ; implicit-def: $vcc_hi
+; GFX10-NEXT:    v_writelane_b32 v40, s31, 1
+; GFX10-NEXT:    s_swappc_b64 s[30:31], s[4:5]
+; GFX10-NEXT:    v_readlane_b32 s4, v40, 0
+; GFX10-NEXT:    v_readlane_b32 s5, v40, 1
+; GFX10-NEXT:    s_sub_u32 s32, s32, 0x200
+; GFX10-NEXT:    v_readlane_b32 s33, v40, 2
+; GFX10-NEXT:    s_or_saveexec_b32 s6, -1
+; GFX10-NEXT:    buffer_load_dword v40, off, s[0:3], s32 ; 4-byte Folded Reload
+; GFX10-NEXT:    s_waitcnt_depctr 0xffe3
+; GFX10-NEXT:    s_mov_b32 exec_lo, s6
+; GFX10-NEXT:    s_waitcnt vmcnt(0)
+; GFX10-NEXT:    s_setpc_b64 s[4:5]
+  call amdgpu_gfx void @external_void_func_f16(half 4.0)
+  ret void
+}
+
+define amdgpu_gfx void @test_call_external_void_func_f32_imm() #0 {
+; GFX9-LABEL: test_call_external_void_func_f32_imm:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    s_or_saveexec_b64 s[4:5], -1
+; GFX9-NEXT:    buffer_store_dword v40, off, s[0:3], s32 ; 4-byte Folded Spill
+; GFX9-NEXT:    s_mov_b64 exec, s[4:5]
+; GFX9-NEXT:    v_writelane_b32 v40, s33, 2
+; GFX9-NEXT:    v_writelane_b32 v40, s30, 0
+; GFX9-NEXT:    s_mov_b32 s33, s32
+; GFX9-NEXT:    s_add_u32 s32, s32, 0x400
+; GFX9-NEXT:    v_mov_b32_e32 v0, 4.0
+; GFX9-NEXT:    s_getpc_b64 s[4:5]
+; GFX9-NEXT:    s_add_u32 s4, s4, external_void_func_f32@rel32@lo+4
+; GFX9-NEXT:    s_addc_u32 s5, s5, external_void_func_f32@rel32@hi+12
+; GFX9-NEXT:    v_writelane_b32 v40, s31, 1
+; GFX9-NEXT:    s_swappc_b64 s[30:31], s[4:5]
+; GFX9-NEXT:    v_readlane_b32 s4, v40, 0
+; GFX9-NEXT:    v_readlane_b32 s5, v40, 1
+; GFX9-NEXT:    s_sub_u32 s32, s32, 0x400
+; GFX9-NEXT:    v_readlane_b32 s33, v40, 2
+; GFX9-NEXT:    s_or_saveexec_b64 s[6:7], -1
+; GFX9-NEXT:    buffer_load_dword v40, off, s[0:3], s32 ; 4-byte Folded Reload
+; GFX9-NEXT:    s_mov_b64 exec, s[6:7]
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-NEXT:    s_setpc_b64 s[4:5]
+;
+; GFX10-LABEL: test_call_external_void_func_f32_imm:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
+; GFX10-NEXT:    s_or_saveexec_b32 s4, -1
+; GFX10-NEXT:    buffer_store_dword v40, off, s[0:3], s32 ; 4-byte Folded Spill
+; GFX10-NEXT:    s_waitcnt_depctr 0xffe3
+; GFX10-NEXT:    s_mov_b32 exec_lo, s4
+; GFX10-NEXT:    v_writelane_b32 v40, s33, 2
+; GFX10-NEXT:    v_mov_b32_e32 v0, 4.0
+; GFX10-NEXT:    s_mov_b32 s33, s32
+; GFX10-NEXT:    s_add_u32 s32, s32, 0x200
+; GFX10-NEXT:    s_getpc_b64 s[4:5]
+; GFX10-NEXT:    s_add_u32 s4, s4, external_void_func_f32@rel32@lo+4
+; GFX10-NEXT:    s_addc_u32 s5, s5, external_void_func_f32@rel32@hi+12
+; GFX10-NEXT:    v_writelane_b32 v40, s30, 0
+; GFX10-NEXT:    ; implicit-def: $vcc_hi
+; GFX10-NEXT:    v_writelane_b32 v40, s31, 1
+; GFX10-NEXT:    s_swappc_b64 s[30:31], s[4:5]
+; GFX10-NEXT:    v_readlane_b32 s4, v40, 0
+; GFX10-NEXT:    v_readlane_b32 s5, v40, 1
+; GFX10-NEXT:    s_sub_u32 s32, s32, 0x200
+; GFX10-NEXT:    v_readlane_b32 s33, v40, 2
+; GFX10-NEXT:    s_or_saveexec_b32 s6, -1
+; GFX10-NEXT:    buffer_load_dword v40, off, s[0:3], s32 ; 4-byte Folded Reload
+; GFX10-NEXT:    s_waitcnt_depctr 0xffe3
+; GFX10-NEXT:    s_mov_b32 exec_lo, s6
+; GFX10-NEXT:    s_waitcnt vmcnt(0)
+; GFX10-NEXT:    s_setpc_b64 s[4:5]
+  call amdgpu_gfx void @external_void_func_f32(float 4.0)
+  ret void
+}
+
+define amdgpu_gfx void @test_call_external_void_func_v2f32_imm() #0 {
+; GFX9-LABEL: test_call_external_void_func_v2f32_imm:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    s_or_saveexec_b64 s[4:5], -1
+; GFX9-NEXT:    buffer_store_dword v40, off, s[0:3], s32 ; 4-byte Folded Spill
+; GFX9-NEXT:    s_mov_b64 exec, s[4:5]
+; GFX9-NEXT:    v_writelane_b32 v40, s33, 2
+; GFX9-NEXT:    v_writelane_b32 v40, s30, 0
+; GFX9-NEXT:    s_mov_b32 s33, s32
+; GFX9-NEXT:    s_add_u32 s32, s32, 0x400
+; GFX9-NEXT:    v_mov_b32_e32 v0, 1.0
+; GFX9-NEXT:    v_mov_b32_e32 v1, 2.0
+; GFX9-NEXT:    s_getpc_b64 s[4:5]
+; GFX9-NEXT:    s_add_u32 s4, s4, external_void_func_v2f32@rel32@lo+4
+; GFX9-NEXT:    s_addc_u32 s5, s5, external_void_func_v2f32@rel32@hi+12
+; GFX9-NEXT:    v_writelane_b32 v40, s31, 1
+; GFX9-NEXT:    s_swappc_b64 s[30:31], s[4:5]
+; GFX9-NEXT:    v_readlane_b32 s4, v40, 0
+; GFX9-NEXT:    v_readlane_b32 s5, v40, 1
+; GFX9-NEXT:    s_sub_u32 s32, s32, 0x400
+; GFX9-NEXT:    v_readlane_b32 s33, v40, 2
+; GFX9-NEXT:    s_or_saveexec_b64 s[6:7], -1
+; GFX9-NEXT:    buffer_load_dword v40, off, s[0:3], s32 ; 4-byte Folded Reload
+; GFX9-NEXT:    s_mov_b64 exec, s[6:7]
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-NEXT:    s_setpc_b64 s[4:5]
+;
+; GFX10-LABEL: test_call_external_void_func_v2f32_imm:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
+; GFX10-NEXT:    s_or_saveexec_b32 s4, -1
+; GFX10-NEXT:    buffer_store_dword v40, off, s[0:3], s32 ; 4-byte Folded Spill
+; GFX10-NEXT:    s_waitcnt_depctr 0xffe3
+; GFX10-NEXT:    s_mov_b32 exec_lo, s4
+; GFX10-NEXT:    v_writelane_b32 v40, s33, 2
+; GFX10-NEXT:    v_mov_b32_e32 v0, 1.0
+; GFX10-NEXT:    v_mov_b32_e32 v1, 2.0
+; GFX10-NEXT:    s_mov_b32 s33, s32
+; GFX10-NEXT:    s_add_u32 s32, s32, 0x200
+; GFX10-NEXT:    v_writelane_b32 v40, s30, 0
+; GFX10-NEXT:    s_getpc_b64 s[4:5]
+; GFX10-NEXT:    s_add_u32 s4, s4, external_void_func_v2f32@rel32@lo+4
+; GFX10-NEXT:    s_addc_u32 s5, s5, external_void_func_v2f32@rel32@hi+12
+; GFX10-NEXT:    ; implicit-def: $vcc_hi
+; GFX10-NEXT:    v_writelane_b32 v40, s31, 1
+; GFX10-NEXT:    s_swappc_b64 s[30:31], s[4:5]
+; GFX10-NEXT:    v_readlane_b32 s4, v40, 0
+; GFX10-NEXT:    v_readlane_b32 s5, v40, 1
+; GFX10-NEXT:    s_sub_u32 s32, s32, 0x200
+; GFX10-NEXT:    v_readlane_b32 s33, v40, 2
+; GFX10-NEXT:    s_or_saveexec_b32 s6, -1
+; GFX10-NEXT:    buffer_load_dword v40, off, s[0:3], s32 ; 4-byte Folded Reload
+; GFX10-NEXT:    s_waitcnt_depctr 0xffe3
+; GFX10-NEXT:    s_mov_b32 exec_lo, s6
+; GFX10-NEXT:    s_waitcnt vmcnt(0)
+; GFX10-NEXT:    s_setpc_b64 s[4:5]
+  call amdgpu_gfx void @external_void_func_v2f32(<2 x float> <float 1.0, float 2.0>)
+  ret void
+}
+
+define amdgpu_gfx void @test_call_external_void_func_v3f32_imm() #0 {
+; GFX9-LABEL: test_call_external_void_func_v3f32_imm:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    s_or_saveexec_b64 s[4:5], -1
+; GFX9-NEXT:    buffer_store_dword v40, off, s[0:3], s32 ; 4-byte Folded Spill
+; GFX9-NEXT:    s_mov_b64 exec, s[4:5]
+; GFX9-NEXT:    v_writelane_b32 v40, s33, 2
+; GFX9-NEXT:    v_writelane_b32 v40, s30, 0
+; GFX9-NEXT:    s_mov_b32 s33, s32
+; GFX9-NEXT:    s_add_u32 s32, s32, 0x400
+; GFX9-NEXT:    v_mov_b32_e32 v0, 1.0
+; GFX9-NEXT:    v_mov_b32_e32 v1, 2.0
+; GFX9-NEXT:    v_mov_b32_e32 v2, 4.0
+; GFX9-NEXT:    s_getpc_b64 s[4:5]
+; GFX9-NEXT:    s_add_u32 s4, s4, external_void_func_v3f32@rel32@lo+4
+; GFX9-NEXT:    s_addc_u32 s5, s5, external_void_func_v3f32@rel32@hi+12
+; GFX9-NEXT:    v_writelane_b32 v40, s31, 1
+; GFX9-NEXT:    s_swappc_b64 s[30:31], s[4:5]
+; GFX9-NEXT:    v_readlane_b32 s4, v40, 0
+; GFX9-NEXT:    v_readlane_b32 s5, v40, 1
+; GFX9-NEXT:    s_sub_u32 s32, s32, 0x400
+; GFX9-NEXT:    v_readlane_b32 s33, v40, 2
+; GFX9-NEXT:    s_or_saveexec_b64 s[6:7], -1
+; GFX9-NEXT:    buffer_load_dword v40, off, s[0:3], s32 ; 4-byte Folded Reload
+; GFX9-NEXT:    s_mov_b64 exec, s[6:7]
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-NEXT:    s_setpc_b64 s[4:5]
+;
+; GFX10-LABEL: test_call_external_void_func_v3f32_imm:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
+; GFX10-NEXT:    s_or_saveexec_b32 s4, -1
+; GFX10-NEXT:    buffer_store_dword v40, off, s[0:3], s32 ; 4-byte Folded Spill
+; GFX10-NEXT:    s_waitcnt_depctr 0xffe3
+; GFX10-NEXT:    s_mov_b32 exec_lo, s4
+; GFX10-NEXT:    v_writelane_b32 v40, s33, 2
+; GFX10-NEXT:    v_mov_b32_e32 v0, 1.0
+; GFX10-NEXT:    v_mov_b32_e32 v1, 2.0
+; GFX10-NEXT:    v_mov_b32_e32 v2, 4.0
+; GFX10-NEXT:    s_mov_b32 s33, s32
+; GFX10-NEXT:    v_writelane_b32 v40, s30, 0
+; GFX10-NEXT:    s_add_u32 s32, s32, 0x200
+; GFX10-NEXT:    s_getpc_b64 s[4:5]
+; GFX10-NEXT:    s_add_u32 s4, s4, external_void_func_v3f32@rel32@lo+4
+; GFX10-NEXT:    s_addc_u32 s5, s5, external_void_func_v3f32@rel32@hi+12
+; GFX10-NEXT:    ; implicit-def: $vcc_hi
+; GFX10-NEXT:    v_writelane_b32 v40, s31, 1
+; GFX10-NEXT:    s_swappc_b64 s[30:31], s[4:5]
+; GFX10-NEXT:    v_readlane_b32 s4, v40, 0
+; GFX10-NEXT:    v_readlane_b32 s5, v40, 1
+; GFX10-NEXT:    s_sub_u32 s32, s32, 0x200
+; GFX10-NEXT:    v_readlane_b32 s33, v40, 2
+; GFX10-NEXT:    s_or_saveexec_b32 s6, -1
+; GFX10-NEXT:    buffer_load_dword v40, off, s[0:3], s32 ; 4-byte Folded Reload
+; GFX10-NEXT:    s_waitcnt_depctr 0xffe3
+; GFX10-NEXT:    s_mov_b32 exec_lo, s6
+; GFX10-NEXT:    s_waitcnt vmcnt(0)
+; GFX10-NEXT:    s_setpc_b64 s[4:5]
+  call amdgpu_gfx void @external_void_func_v3f32(<3 x float> <float 1.0, float 2.0, float 4.0>)
+  ret void
+}
+
+define amdgpu_gfx void @test_call_external_void_func_v5f32_imm() #0 {
+; GFX9-LABEL: test_call_external_void_func_v5f32_imm:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    s_or_saveexec_b64 s[4:5], -1
+; GFX9-NEXT:    buffer_store_dword v40, off, s[0:3], s32 ; 4-byte Folded Spill
+; GFX9-NEXT:    s_mov_b64 exec, s[4:5]
+; GFX9-NEXT:    v_writelane_b32 v40, s33, 2
+; GFX9-NEXT:    v_writelane_b32 v40, s30, 0
+; GFX9-NEXT:    s_mov_b32 s33, s32
+; GFX9-NEXT:    s_add_u32 s32, s32, 0x400
+; GFX9-NEXT:    v_mov_b32_e32 v0, 1.0
+; GFX9-NEXT:    v_mov_b32_e32 v1, 2.0
+; GFX9-NEXT:    v_mov_b32_e32 v2, 4.0
+; GFX9-NEXT:    v_mov_b32_e32 v3, -1.0
+; GFX9-NEXT:    v_mov_b32_e32 v4, 0.5
+; GFX9-NEXT:    s_getpc_b64 s[4:5]
+; GFX9-NEXT:    s_add_u32 s4, s4, external_void_func_v5f32@rel32@lo+4
+; GFX9-NEXT:    s_addc_u32 s5, s5, external_void_func_v5f32@rel32@hi+12
+; GFX9-NEXT:    v_writelane_b32 v40, s31, 1
+; GFX9-NEXT:    s_swappc_b64 s[30:31], s[4:5]
+; GFX9-NEXT:    v_readlane_b32 s4, v40, 0
+; GFX9-NEXT:    v_readlane_b32 s5, v40, 1
+; GFX9-NEXT:    s_sub_u32 s32, s32, 0x400
+; GFX9-NEXT:    v_readlane_b32 s33, v40, 2
+; GFX9-NEXT:    s_or_saveexec_b64 s[6:7], -1
+; GFX9-NEXT:    buffer_load_dword v40, off, s[0:3], s32 ; 4-byte Folded Reload
+; GFX9-NEXT:    s_mov_b64 exec, s[6:7]
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-NEXT:    s_setpc_b64 s[4:5]
+;
+; GFX10-LABEL: test_call_external_void_func_v5f32_imm:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
+; GFX10-NEXT:    s_or_saveexec_b32 s4, -1
+; GFX10-NEXT:    buffer_store_dword v40, off, s[0:3], s32 ; 4-byte Folded Spill
+; GFX10-NEXT:    s_waitcnt_depctr 0xffe3
+; GFX10-NEXT:    s_mov_b32 exec_lo, s4
+; GFX10-NEXT:    v_writelane_b32 v40, s33, 2
+; GFX10-NEXT:    v_mov_b32_e32 v0, 1.0
+; GFX10-NEXT:    v_mov_b32_e32 v1, 2.0
+; GFX10-NEXT:    v_mov_b32_e32 v2, 4.0
+; GFX10-NEXT:    v_mov_b32_e32 v3, -1.0
+; GFX10-NEXT:    v_writelane_b32 v40, s30, 0
+; GFX10-NEXT:    v_mov_b32_e32 v4, 0.5
+; GFX10-NEXT:    s_mov_b32 s33, s32
+; GFX10-NEXT:    s_add_u32 s32, s32, 0x200
+; GFX10-NEXT:    s_getpc_b64 s[4:5]
+; GFX10-NEXT:    s_add_u32 s4, s4, external_void_func_v5f32@rel32@lo+4
+; GFX10-NEXT:    s_addc_u32 s5, s5, external_void_func_v5f32@rel32@hi+12
+; GFX10-NEXT:    v_writelane_b32 v40, s31, 1
+; GFX10-NEXT:    ; implicit-def: $vcc_hi
+; GFX10-NEXT:    s_swappc_b64 s[30:31], s[4:5]
+; GFX10-NEXT:    v_readlane_b32 s4, v40, 0
+; GFX10-NEXT:    v_readlane_b32 s5, v40, 1
+; GFX10-NEXT:    s_sub_u32 s32, s32, 0x200
+; GFX10-NEXT:    v_readlane_b32 s33, v40, 2
+; GFX10-NEXT:    s_or_saveexec_b32 s6, -1
+; GFX10-NEXT:    buffer_load_dword v40, off, s[0:3], s32 ; 4-byte Folded Reload
+; GFX10-NEXT:    s_waitcnt_depctr 0xffe3
+; GFX10-NEXT:    s_mov_b32 exec_lo, s6
+; GFX10-NEXT:    s_waitcnt vmcnt(0)
+; GFX10-NEXT:    s_setpc_b64 s[4:5]
+  call amdgpu_gfx void @external_void_func_v5f32(<5 x float> <float 1.0, float 2.0, float 4.0, float -1.0, float 0.5>)
+  ret void
+}
+
+define amdgpu_gfx void @test_call_external_void_func_f64_imm() #0 {
+; GFX9-LABEL: test_call_external_void_func_f64_imm:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    s_or_saveexec_b64 s[4:5], -1
+; GFX9-NEXT:    buffer_store_dword v40, off, s[0:3], s32 ; 4-byte Folded Spill
+; GFX9-NEXT:    s_mov_b64 exec, s[4:5]
+; GFX9-NEXT:    v_writelane_b32 v40, s33, 2
+; GFX9-NEXT:    v_writelane_b32 v40, s30, 0
+; GFX9-NEXT:    s_mov_b32 s33, s32
+; GFX9-NEXT:    s_add_u32 s32, s32, 0x400
+; GFX9-NEXT:    v_mov_b32_e32 v0, 0
+; GFX9-NEXT:    v_mov_b32_e32 v1, 0x40100000
+; GFX9-NEXT:    s_getpc_b64 s[4:5]
+; GFX9-NEXT:    s_add_u32 s4, s4, external_void_func_f64@rel32@lo+4
+; GFX9-NEXT:    s_addc_u32 s5, s5, external_void_func_f64@rel32@hi+12
+; GFX9-NEXT:    v_writelane_b32 v40, s31, 1
+; GFX9-NEXT:    s_swappc_b64 s[30:31], s[4:5]
+; GFX9-NEXT:    v_readlane_b32 s4, v40, 0
+; GFX9-NEXT:    v_readlane_b32 s5, v40, 1
+; GFX9-NEXT:    s_sub_u32 s32, s32, 0x400
+; GFX9-NEXT:    v_readlane_b32 s33, v40, 2
+; GFX9-NEXT:    s_or_saveexec_b64 s[6:7], -1
+; GFX9-NEXT:    buffer_load_dword v40, off, s[0:3], s32 ; 4-byte Folded Reload
+; GFX9-NEXT:    s_mov_b64 exec, s[6:7]
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-NEXT:    s_setpc_b64 s[4:5]
+;
+; GFX10-LABEL: test_call_external_void_func_f64_imm:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
+; GFX10-NEXT:    s_or_saveexec_b32 s4, -1
+; GFX10-NEXT:    buffer_store_dword v40, off, s[0:3], s32 ; 4-byte Folded Spill
+; GFX10-NEXT:    s_waitcnt_depctr 0xffe3
+; GFX10-NEXT:    s_mov_b32 exec_lo, s4
+; GFX10-NEXT:    v_writelane_b32 v40, s33, 2
+; GFX10-NEXT:    v_mov_b32_e32 v0, 0
+; GFX10-NEXT:    v_mov_b32_e32 v1, 0x40100000
+; GFX10-NEXT:    s_mov_b32 s33, s32
+; GFX10-NEXT:    s_add_u32 s32, s32, 0x200
+; GFX10-NEXT:    v_writelane_b32 v40, s30, 0
+; GFX10-NEXT:    s_getpc_b64 s[4:5]
+; GFX10-NEXT:    s_add_u32 s4, s4, external_void_func_f64@rel32@lo+4
+; GFX10-NEXT:    s_addc_u32 s5, s5, external_void_func_f64@rel32@hi+12
+; GFX10-NEXT:    ; implicit-def: $vcc_hi
+; GFX10-NEXT:    v_writelane_b32 v40, s31, 1
+; GFX10-NEXT:    s_swappc_b64 s[30:31], s[4:5]
+; GFX10-NEXT:    v_readlane_b32 s4, v40, 0
+; GFX10-NEXT:    v_readlane_b32 s5, v40, 1
+; GFX10-NEXT:    s_sub_u32 s32, s32, 0x200
+; GFX10-NEXT:    v_readlane_b32 s33, v40, 2
+; GFX10-NEXT:    s_or_saveexec_b32 s6, -1
+; GFX10-NEXT:    buffer_load_dword v40, off, s[0:3], s32 ; 4-byte Folded Reload
+; GFX10-NEXT:    s_waitcnt_depctr 0xffe3
+; GFX10-NEXT:    s_mov_b32 exec_lo, s6
+; GFX10-NEXT:    s_waitcnt vmcnt(0)
+; GFX10-NEXT:    s_setpc_b64 s[4:5]
+  call amdgpu_gfx void @external_void_func_f64(double 4.0)
+  ret void
+}
+
+define amdgpu_gfx void @test_call_external_void_func_v2f64_imm() #0 {
+; GFX9-LABEL: test_call_external_void_func_v2f64_imm:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    s_or_saveexec_b64 s[4:5], -1
+; GFX9-NEXT:    buffer_store_dword v40, off, s[0:3], s32 ; 4-byte Folded Spill
+; GFX9-NEXT:    s_mov_b64 exec, s[4:5]
+; GFX9-NEXT:    v_writelane_b32 v40, s33, 2
+; GFX9-NEXT:    v_writelane_b32 v40, s30, 0
+; GFX9-NEXT:    s_mov_b32 s33, s32
+; GFX9-NEXT:    s_add_u32 s32, s32, 0x400
+; GFX9-NEXT:    v_mov_b32_e32 v0, 0
+; GFX9-NEXT:    v_mov_b32_e32 v1, 2.0
+; GFX9-NEXT:    v_mov_b32_e32 v2, 0
+; GFX9-NEXT:    v_mov_b32_e32 v3, 0x40100000
+; GFX9-NEXT:    s_getpc_b64 s[4:5]
+; GFX9-NEXT:    s_add_u32 s4, s4, external_void_func_v2f64@rel32@lo+4
+; GFX9-NEXT:    s_addc_u32 s5, s5, external_void_func_v2f64@rel32@hi+12
+; GFX9-NEXT:    v_writelane_b32 v40, s31, 1
+; GFX9-NEXT:    s_swappc_b64 s[30:31], s[4:5]
+; GFX9-NEXT:    v_readlane_b32 s4, v40, 0
+; GFX9-NEXT:    v_readlane_b32 s5, v40, 1
+; GFX9-NEXT:    s_sub_u32 s32, s32, 0x400
+; GFX9-NEXT:    v_readlane_b32 s33, v40, 2
+; GFX9-NEXT:    s_or_saveexec_b64 s[6:7], -1
+; GFX9-NEXT:    buffer_load_dword v40, off, s[0:3], s32 ; 4-byte Folded Reload
+; GFX9-NEXT:    s_mov_b64 exec, s[6:7]
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-NEXT:    s_setpc_b64 s[4:5]
+;
+; GFX10-LABEL: test_call_external_void_func_v2f64_imm:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
+; GFX10-NEXT:    s_or_saveexec_b32 s4, -1
+; GFX10-NEXT:    buffer_store_dword v40, off, s[0:3], s32 ; 4-byte Folded Spill
+; GFX10-NEXT:    s_waitcnt_depctr 0xffe3
+; GFX10-NEXT:    s_mov_b32 exec_lo, s4
+; GFX10-NEXT:    v_writelane_b32 v40, s33, 2
+; GFX10-NEXT:    v_mov_b32_e32 v0, 0
+; GFX10-NEXT:    v_mov_b32_e32 v1, 2.0
+; GFX10-NEXT:    v_mov_b32_e32 v2, 0
+; GFX10-NEXT:    v_mov_b32_e32 v3, 0x40100000
+; GFX10-NEXT:    v_writelane_b32 v40, s30, 0
+; GFX10-NEXT:    s_mov_b32 s33, s32
+; GFX10-NEXT:    s_add_u32 s32, s32, 0x200
+; GFX10-NEXT:    s_getpc_b64 s[4:5]
+; GFX10-NEXT:    s_add_u32 s4, s4, external_void_func_v2f64@rel32@lo+4
+; GFX10-NEXT:    s_addc_u32 s5, s5, external_void_func_v2f64@rel32@hi+12
+; GFX10-NEXT:    ; implicit-def: $vcc_hi
+; GFX10-NEXT:    v_writelane_b32 v40, s31, 1
+; GFX10-NEXT:    s_swappc_b64 s[30:31], s[4:5]
+; GFX10-NEXT:    v_readlane_b32 s4, v40, 0
+; GFX10-NEXT:    v_readlane_b32 s5, v40, 1
+; GFX10-NEXT:    s_sub_u32 s32, s32, 0x200
+; GFX10-NEXT:    v_readlane_b32 s33, v40, 2
+; GFX10-NEXT:    s_or_saveexec_b32 s6, -1
+; GFX10-NEXT:    buffer_load_dword v40, off, s[0:3], s32 ; 4-byte Folded Reload
+; GFX10-NEXT:    s_waitcnt_depctr 0xffe3
+; GFX10-NEXT:    s_mov_b32 exec_lo, s6
+; GFX10-NEXT:    s_waitcnt vmcnt(0)
+; GFX10-NEXT:    s_setpc_b64 s[4:5]
+  call amdgpu_gfx void @external_void_func_v2f64(<2 x double> <double 2.0, double 4.0>)
+  ret void
+}
+
+define amdgpu_gfx void @test_call_external_void_func_v3f64_imm() #0 {
+; GFX9-LABEL: test_call_external_void_func_v3f64_imm:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    s_or_saveexec_b64 s[4:5], -1
+; GFX9-NEXT:    buffer_store_dword v40, off, s[0:3], s32 ; 4-byte Folded Spill
+; GFX9-NEXT:    s_mov_b64 exec, s[4:5]
+; GFX9-NEXT:    v_writelane_b32 v40, s33, 2
+; GFX9-NEXT:    v_writelane_b32 v40, s30, 0
+; GFX9-NEXT:    s_mov_b32 s33, s32
+; GFX9-NEXT:    s_add_u32 s32, s32, 0x400
+; GFX9-NEXT:    v_mov_b32_e32 v0, 0
+; GFX9-NEXT:    v_mov_b32_e32 v1, 2.0
+; GFX9-NEXT:    v_mov_b32_e32 v2, 0
+; GFX9-NEXT:    v_mov_b32_e32 v3, 0x40100000
+; GFX9-NEXT:    v_mov_b32_e32 v4, 0
+; GFX9-NEXT:    v_mov_b32_e32 v5, 0x40200000
+; GFX9-NEXT:    s_getpc_b64 s[4:5]
+; GFX9-NEXT:    s_add_u32 s4, s4, external_void_func_v3f64@rel32@lo+4
+; GFX9-NEXT:    s_addc_u32 s5, s5, external_void_func_v3f64@rel32@hi+12
+; GFX9-NEXT:    v_writelane_b32 v40, s31, 1
+; GFX9-NEXT:    s_swappc_b64 s[30:31], s[4:5]
+; GFX9-NEXT:    v_readlane_b32 s4, v40, 0
+; GFX9-NEXT:    v_readlane_b32 s5, v40, 1
+; GFX9-NEXT:    s_sub_u32 s32, s32, 0x400
+; GFX9-NEXT:    v_readlane_b32 s33, v40, 2
+; GFX9-NEXT:    s_or_saveexec_b64 s[6:7], -1
+; GFX9-NEXT:    buffer_load_dword v40, off, s[0:3], s32 ; 4-byte Folded Reload
+; GFX9-NEXT:    s_mov_b64 exec, s[6:7]
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-NEXT:    s_setpc_b64 s[4:5]
+;
+; GFX10-LABEL: test_call_external_void_func_v3f64_imm:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
+; GFX10-NEXT:    s_or_saveexec_b32 s4, -1
+; GFX10-NEXT:    buffer_store_dword v40, off, s[0:3], s32 ; 4-byte Folded Spill
+; GFX10-NEXT:    s_waitcnt_depctr 0xffe3
+; GFX10-NEXT:    s_mov_b32 exec_lo, s4
+; GFX10-NEXT:    v_writelane_b32 v40, s33, 2
+; GFX10-NEXT:    v_mov_b32_e32 v0, 0
+; GFX10-NEXT:    v_mov_b32_e32 v1, 2.0
+; GFX10-NEXT:    v_mov_b32_e32 v2, 0
+; GFX10-NEXT:    v_mov_b32_e32 v3, 0x40100000
+; GFX10-NEXT:    v_writelane_b32 v40, s30, 0
+; GFX10-NEXT:    v_mov_b32_e32 v4, 0
+; GFX10-NEXT:    v_mov_b32_e32 v5, 0x40200000
+; GFX10-NEXT:    s_mov_b32 s33, s32
+; GFX10-NEXT:    s_add_u32 s32, s32, 0x200
+; GFX10-NEXT:    s_getpc_b64 s[4:5]
+; GFX10-NEXT:    s_add_u32 s4, s4, external_void_func_v3f64@rel32@lo+4
+; GFX10-NEXT:    s_addc_u32 s5, s5, external_void_func_v3f64@rel32@hi+12
+; GFX10-NEXT:    v_writelane_b32 v40, s31, 1
+; GFX10-NEXT:    ; implicit-def: $vcc_hi
+; GFX10-NEXT:    s_swappc_b64 s[30:31], s[4:5]
+; GFX10-NEXT:    v_readlane_b32 s4, v40, 0
+; GFX10-NEXT:    v_readlane_b32 s5, v40, 1
+; GFX10-NEXT:    s_sub_u32 s32, s32, 0x200
+; GFX10-NEXT:    v_readlane_b32 s33, v40, 2
+; GFX10-NEXT:    s_or_saveexec_b32 s6, -1
+; GFX10-NEXT:    buffer_load_dword v40, off, s[0:3], s32 ; 4-byte Folded Reload
+; GFX10-NEXT:    s_waitcnt_depctr 0xffe3
+; GFX10-NEXT:    s_mov_b32 exec_lo, s6
+; GFX10-NEXT:    s_waitcnt vmcnt(0)
+; GFX10-NEXT:    s_setpc_b64 s[4:5]
+  call amdgpu_gfx void @external_void_func_v3f64(<3 x double> <double 2.0, double 4.0, double 8.0>)
+  ret void
+}
+
+define amdgpu_gfx void @test_call_external_void_func_v2i16() #0 {
+; GFX9-LABEL: test_call_external_void_func_v2i16:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    s_or_saveexec_b64 s[4:5], -1
+; GFX9-NEXT:    buffer_store_dword v40, off, s[0:3], s32 ; 4-byte Folded Spill
+; GFX9-NEXT:    s_mov_b64 exec, s[4:5]
+; GFX9-NEXT:    v_writelane_b32 v40, s33, 2
+; GFX9-NEXT:    v_writelane_b32 v40, s30, 0
+; GFX9-NEXT:    v_writelane_b32 v40, s31, 1
+; GFX9-NEXT:    global_load_dword v0, v[0:1], off
+; GFX9-NEXT:    s_mov_b32 s33, s32
+; GFX9-NEXT:    s_add_u32 s32, s32, 0x400
+; GFX9-NEXT:    s_getpc_b64 s[4:5]
+; GFX9-NEXT:    s_add_u32 s4, s4, external_void_func_v2i16@rel32@lo+4
+; GFX9-NEXT:    s_addc_u32 s5, s5, external_void_func_v2i16@rel32@hi+12
+; GFX9-NEXT:    s_swappc_b64 s[30:31], s[4:5]
+; GFX9-NEXT:    v_readlane_b32 s4, v40, 0
+; GFX9-NEXT:    v_readlane_b32 s5, v40, 1
+; GFX9-NEXT:    s_sub_u32 s32, s32, 0x400
+; GFX9-NEXT:    v_readlane_b32 s33, v40, 2
+; GFX9-NEXT:    s_or_saveexec_b64 s[6:7], -1
+; GFX9-NEXT:    buffer_load_dword v40, off, s[0:3], s32 ; 4-byte Folded Reload
+; GFX9-NEXT:    s_mov_b64 exec, s[6:7]
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-NEXT:    s_setpc_b64 s[4:5]
+;
+; GFX10-LABEL: test_call_external_void_func_v2i16:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
+; GFX10-NEXT:    s_or_saveexec_b32 s4, -1
+; GFX10-NEXT:    buffer_store_dword v40, off, s[0:3], s32 ; 4-byte Folded Spill
+; GFX10-NEXT:    s_waitcnt_depctr 0xffe3
+; GFX10-NEXT:    s_mov_b32 exec_lo, s4
+; GFX10-NEXT:    v_writelane_b32 v40, s33, 2
+; GFX10-NEXT:    s_mov_b32 s33, s32
+; GFX10-NEXT:    s_add_u32 s32, s32, 0x200
+; GFX10-NEXT:    ; implicit-def: $vcc_hi
+; GFX10-NEXT:    s_getpc_b64 s[4:5]
+; GFX10-NEXT:    s_add_u32 s4, s4, external_void_func_v2i16@rel32@lo+4
+; GFX10-NEXT:    s_addc_u32 s5, s5, external_void_func_v2i16@rel32@hi+12
+; GFX10-NEXT:    v_writelane_b32 v40, s30, 0
+; GFX10-NEXT:    v_writelane_b32 v40, s31, 1
+; GFX10-NEXT:    global_load_dword v0, v[0:1], off
+; GFX10-NEXT:    s_swappc_b64 s[30:31], s[4:5]
+; GFX10-NEXT:    v_readlane_b32 s4, v40, 0
+; GFX10-NEXT:    v_readlane_b32 s5, v40, 1
+; GFX10-NEXT:    s_sub_u32 s32, s32, 0x200
+; GFX10-NEXT:    v_readlane_b32 s33, v40, 2
+; GFX10-NEXT:    s_or_saveexec_b32 s6, -1
+; GFX10-NEXT:    buffer_load_dword v40, off, s[0:3], s32 ; 4-byte Folded Reload
+; GFX10-NEXT:    s_waitcnt_depctr 0xffe3
+; GFX10-NEXT:    s_mov_b32 exec_lo, s6
+; GFX10-NEXT:    s_waitcnt vmcnt(0)
+; GFX10-NEXT:    s_setpc_b64 s[4:5]
+  %val = load <2 x i16>, <2 x i16> addrspace(1)* undef
+  call amdgpu_gfx void @external_void_func_v2i16(<2 x i16> %val)
+  ret void
+}
+
+define amdgpu_gfx void @test_call_external_void_func_v3i16() #0 {
+; GFX9-LABEL: test_call_external_void_func_v3i16:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    s_or_saveexec_b64 s[4:5], -1
+; GFX9-NEXT:    buffer_store_dword v40, off, s[0:3], s32 ; 4-byte Folded Spill
+; GFX9-NEXT:    s_mov_b64 exec, s[4:5]
+; GFX9-NEXT:    v_writelane_b32 v40, s33, 2
+; GFX9-NEXT:    v_writelane_b32 v40, s30, 0
+; GFX9-NEXT:    v_writelane_b32 v40, s31, 1
+; GFX9-NEXT:    global_load_dwordx2 v[0:1], v[0:1], off
+; GFX9-NEXT:    s_mov_b32 s33, s32
+; GFX9-NEXT:    s_add_u32 s32, s32, 0x400
+; GFX9-NEXT:    s_getpc_b64 s[4:5]
+; GFX9-NEXT:    s_add_u32 s4, s4, external_void_func_v3i16@rel32@lo+4
+; GFX9-NEXT:    s_addc_u32 s5, s5, external_void_func_v3i16@rel32@hi+12
+; GFX9-NEXT:    s_swappc_b64 s[30:31], s[4:5]
+; GFX9-NEXT:    v_readlane_b32 s4, v40, 0
+; GFX9-NEXT:    v_readlane_b32 s5, v40, 1
+; GFX9-NEXT:    s_sub_u32 s32, s32, 0x400
+; GFX9-NEXT:    v_readlane_b32 s33, v40, 2
+; GFX9-NEXT:    s_or_saveexec_b64 s[6:7], -1
+; GFX9-NEXT:    buffer_load_dword v40, off, s[0:3], s32 ; 4-byte Folded Reload
+; GFX9-NEXT:    s_mov_b64 exec, s[6:7]
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-NEXT:    s_setpc_b64 s[4:5]
+;
+; GFX10-LABEL: test_call_external_void_func_v3i16:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
+; GFX10-NEXT:    s_or_saveexec_b32 s4, -1
+; GFX10-NEXT:    buffer_store_dword v40, off, s[0:3], s32 ; 4-byte Folded Spill
+; GFX10-NEXT:    s_waitcnt_depctr 0xffe3
+; GFX10-NEXT:    s_mov_b32 exec_lo, s4
+; GFX10-NEXT:    v_writelane_b32 v40, s33, 2
+; GFX10-NEXT:    s_mov_b32 s33, s32
+; GFX10-NEXT:    s_add_u32 s32, s32, 0x200
+; GFX10-NEXT:    ; implicit-def: $vcc_hi
+; GFX10-NEXT:    s_getpc_b64 s[4:5]
+; GFX10-NEXT:    s_add_u32 s4, s4, external_void_func_v3i16@rel32@lo+4
+; GFX10-NEXT:    s_addc_u32 s5, s5, external_void_func_v3i16@rel32@hi+12
+; GFX10-NEXT:    v_writelane_b32 v40, s30, 0
+; GFX10-NEXT:    v_writelane_b32 v40, s31, 1
+; GFX10-NEXT:    global_load_dwordx2 v[0:1], v[0:1], off
+; GFX10-NEXT:    s_swappc_b64 s[30:31], s[4:5]
+; GFX10-NEXT:    v_readlane_b32 s4, v40, 0
+; GFX10-NEXT:    v_readlane_b32 s5, v40, 1
+; GFX10-NEXT:    s_sub_u32 s32, s32, 0x200
+; GFX10-NEXT:    v_readlane_b32 s33, v40, 2
+; GFX10-NEXT:    s_or_saveexec_b32 s6, -1
+; GFX10-NEXT:    buffer_load_dword v40, off, s[0:3], s32 ; 4-byte Folded Reload
+; GFX10-NEXT:    s_waitcnt_depctr 0xffe3
+; GFX10-NEXT:    s_mov_b32 exec_lo, s6
+; GFX10-NEXT:    s_waitcnt vmcnt(0)
+; GFX10-NEXT:    s_setpc_b64 s[4:5]
+  %val = load <3 x i16>, <3 x i16> addrspace(1)* undef
+  call amdgpu_gfx void @external_void_func_v3i16(<3 x i16> %val)
+  ret void
+}
+
+define amdgpu_gfx void @test_call_external_void_func_v3f16() #0 {
+; GFX9-LABEL: test_call_external_void_func_v3f16:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    s_or_saveexec_b64 s[4:5], -1
+; GFX9-NEXT:    buffer_store_dword v40, off, s[0:3], s32 ; 4-byte Folded Spill
+; GFX9-NEXT:    s_mov_b64 exec, s[4:5]
+; GFX9-NEXT:    v_writelane_b32 v40, s33, 2
+; GFX9-NEXT:    v_writelane_b32 v40, s30, 0
+; GFX9-NEXT:    v_writelane_b32 v40, s31, 1
+; GFX9-NEXT:    global_load_dwordx2 v[0:1], v[0:1], off
+; GFX9-NEXT:    s_mov_b32 s33, s32
+; GFX9-NEXT:    s_add_u32 s32, s32, 0x400
+; GFX9-NEXT:    s_getpc_b64 s[4:5]
+; GFX9-NEXT:    s_add_u32 s4, s4, external_void_func_v3f16@rel32@lo+4
+; GFX9-NEXT:    s_addc_u32 s5, s5, external_void_func_v3f16@rel32@hi+12
+; GFX9-NEXT:    s_swappc_b64 s[30:31], s[4:5]
+; GFX9-NEXT:    v_readlane_b32 s4, v40, 0
+; GFX9-NEXT:    v_readlane_b32 s5, v40, 1
+; GFX9-NEXT:    s_sub_u32 s32, s32, 0x400
+; GFX9-NEXT:    v_readlane_b32 s33, v40, 2
+; GFX9-NEXT:    s_or_saveexec_b64 s[6:7], -1
+; GFX9-NEXT:    buffer_load_dword v40, off, s[0:3], s32 ; 4-byte Folded Reload
+; GFX9-NEXT:    s_mov_b64 exec, s[6:7]
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-NEXT:    s_setpc_b64 s[4:5]
+;
+; GFX10-LABEL: test_call_external_void_func_v3f16:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
+; GFX10-NEXT:    s_or_saveexec_b32 s4, -1
+; GFX10-NEXT:    buffer_store_dword v40, off, s[0:3], s32 ; 4-byte Folded Spill
+; GFX10-NEXT:    s_waitcnt_depctr 0xffe3
+; GFX10-NEXT:    s_mov_b32 exec_lo, s4
+; GFX10-NEXT:    v_writelane_b32 v40, s33, 2
+; GFX10-NEXT:    s_mov_b32 s33, s32
+; GFX10-NEXT:    s_add_u32 s32, s32, 0x200
+; GFX10-NEXT:    ; implicit-def: $vcc_hi
+; GFX10-NEXT:    s_getpc_b64 s[4:5]
+; GFX10-NEXT:    s_add_u32 s4, s4, external_void_func_v3f16@rel32@lo+4
+; GFX10-NEXT:    s_addc_u32 s5, s5, external_void_func_v3f16@rel32@hi+12
+; GFX10-NEXT:    v_writelane_b32 v40, s30, 0
+; GFX10-NEXT:    v_writelane_b32 v40, s31, 1
+; GFX10-NEXT:    global_load_dwordx2 v[0:1], v[0:1], off
+; GFX10-NEXT:    s_swappc_b64 s[30:31], s[4:5]
+; GFX10-NEXT:    v_readlane_b32 s4, v40, 0
+; GFX10-NEXT:    v_readlane_b32 s5, v40, 1
+; GFX10-NEXT:    s_sub_u32 s32, s32, 0x200
+; GFX10-NEXT:    v_readlane_b32 s33, v40, 2
+; GFX10-NEXT:    s_or_saveexec_b32 s6, -1
+; GFX10-NEXT:    buffer_load_dword v40, off, s[0:3], s32 ; 4-byte Folded Reload
+; GFX10-NEXT:    s_waitcnt_depctr 0xffe3
+; GFX10-NEXT:    s_mov_b32 exec_lo, s6
+; GFX10-NEXT:    s_waitcnt vmcnt(0)
+; GFX10-NEXT:    s_setpc_b64 s[4:5]
+  %val = load <3 x half>, <3 x half> addrspace(1)* undef
+  call amdgpu_gfx void @external_void_func_v3f16(<3 x half> %val)
+  ret void
+}
+
+define amdgpu_gfx void @test_call_external_void_func_v3i16_imm() #0 {
+; GFX9-LABEL: test_call_external_void_func_v3i16_imm:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    s_or_saveexec_b64 s[4:5], -1
+; GFX9-NEXT:    buffer_store_dword v40, off, s[0:3], s32 ; 4-byte Folded Spill
+; GFX9-NEXT:    s_mov_b64 exec, s[4:5]
+; GFX9-NEXT:    v_writelane_b32 v40, s33, 2
+; GFX9-NEXT:    v_writelane_b32 v40, s30, 0
+; GFX9-NEXT:    s_mov_b32 s33, s32
+; GFX9-NEXT:    s_add_u32 s32, s32, 0x400
+; GFX9-NEXT:    v_mov_b32_e32 v0, 0x20001
+; GFX9-NEXT:    v_mov_b32_e32 v1, 3
+; GFX9-NEXT:    s_getpc_b64 s[4:5]
+; GFX9-NEXT:    s_add_u32 s4, s4, external_void_func_v3i16@rel32@lo+4
+; GFX9-NEXT:    s_addc_u32 s5, s5, external_void_func_v3i16@rel32@hi+12
+; GFX9-NEXT:    v_writelane_b32 v40, s31, 1
+; GFX9-NEXT:    s_swappc_b64 s[30:31], s[4:5]
+; GFX9-NEXT:    v_readlane_b32 s4, v40, 0
+; GFX9-NEXT:    v_readlane_b32 s5, v40, 1
+; GFX9-NEXT:    s_sub_u32 s32, s32, 0x400
+; GFX9-NEXT:    v_readlane_b32 s33, v40, 2
+; GFX9-NEXT:    s_or_saveexec_b64 s[6:7], -1
+; GFX9-NEXT:    buffer_load_dword v40, off, s[0:3], s32 ; 4-byte Folded Reload
+; GFX9-NEXT:    s_mov_b64 exec, s[6:7]
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-NEXT:    s_setpc_b64 s[4:5]
+;
+; GFX10-LABEL: test_call_external_void_func_v3i16_imm:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
+; GFX10-NEXT:    s_or_saveexec_b32 s4, -1
+; GFX10-NEXT:    buffer_store_dword v40, off, s[0:3], s32 ; 4-byte Folded Spill
+; GFX10-NEXT:    s_waitcnt_depctr 0xffe3
+; GFX10-NEXT:    s_mov_b32 exec_lo, s4
+; GFX10-NEXT:    v_writelane_b32 v40, s33, 2
+; GFX10-NEXT:    v_mov_b32_e32 v0, 0x20001
+; GFX10-NEXT:    v_mov_b32_e32 v1, 3
+; GFX10-NEXT:    s_mov_b32 s33, s32
+; GFX10-NEXT:    s_add_u32 s32, s32, 0x200
+; GFX10-NEXT:    v_writelane_b32 v40, s30, 0
+; GFX10-NEXT:    s_getpc_b64 s[4:5]
+; GFX10-NEXT:    s_add_u32 s4, s4, external_void_func_v3i16@rel32@lo+4
+; GFX10-NEXT:    s_addc_u32 s5, s5, external_void_func_v3i16@rel32@hi+12
+; GFX10-NEXT:    ; implicit-def: $vcc_hi
+; GFX10-NEXT:    v_writelane_b32 v40, s31, 1
+; GFX10-NEXT:    s_swappc_b64 s[30:31], s[4:5]
+; GFX10-NEXT:    v_readlane_b32 s4, v40, 0
+; GFX10-NEXT:    v_readlane_b32 s5, v40, 1
+; GFX10-NEXT:    s_sub_u32 s32, s32, 0x200
+; GFX10-NEXT:    v_readlane_b32 s33, v40, 2
+; GFX10-NEXT:    s_or_saveexec_b32 s6, -1
+; GFX10-NEXT:    buffer_load_dword v40, off, s[0:3], s32 ; 4-byte Folded Reload
+; GFX10-NEXT:    s_waitcnt_depctr 0xffe3
+; GFX10-NEXT:    s_mov_b32 exec_lo, s6
+; GFX10-NEXT:    s_waitcnt vmcnt(0)
+; GFX10-NEXT:    s_setpc_b64 s[4:5]
+  call amdgpu_gfx void @external_void_func_v3i16(<3 x i16> <i16 1, i16 2, i16 3>)
+  ret void
+}
+
+define amdgpu_gfx void @test_call_external_void_func_v3f16_imm() #0 {
+; GFX9-LABEL: test_call_external_void_func_v3f16_imm:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    s_or_saveexec_b64 s[4:5], -1
+; GFX9-NEXT:    buffer_store_dword v40, off, s[0:3], s32 ; 4-byte Folded Spill
+; GFX9-NEXT:    s_mov_b64 exec, s[4:5]
+; GFX9-NEXT:    v_writelane_b32 v40, s33, 2
+; GFX9-NEXT:    v_writelane_b32 v40, s30, 0
+; GFX9-NEXT:    s_mov_b32 s33, s32
+; GFX9-NEXT:    s_add_u32 s32, s32, 0x400
+; GFX9-NEXT:    v_mov_b32_e32 v0, 0x40003c00
+; GFX9-NEXT:    v_mov_b32_e32 v1, 0x4400
+; GFX9-NEXT:    s_getpc_b64 s[4:5]
+; GFX9-NEXT:    s_add_u32 s4, s4, external_void_func_v3f16@rel32@lo+4
+; GFX9-NEXT:    s_addc_u32 s5, s5, external_void_func_v3f16@rel32@hi+12
+; GFX9-NEXT:    v_writelane_b32 v40, s31, 1
+; GFX9-NEXT:    s_swappc_b64 s[30:31], s[4:5]
+; GFX9-NEXT:    v_readlane_b32 s4, v40, 0
+; GFX9-NEXT:    v_readlane_b32 s5, v40, 1
+; GFX9-NEXT:    s_sub_u32 s32, s32, 0x400
+; GFX9-NEXT:    v_readlane_b32 s33, v40, 2
+; GFX9-NEXT:    s_or_saveexec_b64 s[6:7], -1
+; GFX9-NEXT:    buffer_load_dword v40, off, s[0:3], s32 ; 4-byte Folded Reload
+; GFX9-NEXT:    s_mov_b64 exec, s[6:7]
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-NEXT:    s_setpc_b64 s[4:5]
+;
+; GFX10-LABEL: test_call_external_void_func_v3f16_imm:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
+; GFX10-NEXT:    s_or_saveexec_b32 s4, -1
+; GFX10-NEXT:    buffer_store_dword v40, off, s[0:3], s32 ; 4-byte Folded Spill
+; GFX10-NEXT:    s_waitcnt_depctr 0xffe3
+; GFX10-NEXT:    s_mov_b32 exec_lo, s4
+; GFX10-NEXT:    v_writelane_b32 v40, s33, 2
+; GFX10-NEXT:    v_mov_b32_e32 v0, 0x40003c00
+; GFX10-NEXT:    v_mov_b32_e32 v1, 0x4400
+; GFX10-NEXT:    s_mov_b32 s33, s32
+; GFX10-NEXT:    s_add_u32 s32, s32, 0x200
+; GFX10-NEXT:    v_writelane_b32 v40, s30, 0
+; GFX10-NEXT:    s_getpc_b64 s[4:5]
+; GFX10-NEXT:    s_add_u32 s4, s4, external_void_func_v3f16@rel32@lo+4
+; GFX10-NEXT:    s_addc_u32 s5, s5, external_void_func_v3f16@rel32@hi+12
+; GFX10-NEXT:    ; implicit-def: $vcc_hi
+; GFX10-NEXT:    v_writelane_b32 v40, s31, 1
+; GFX10-NEXT:    s_swappc_b64 s[30:31], s[4:5]
+; GFX10-NEXT:    v_readlane_b32 s4, v40, 0
+; GFX10-NEXT:    v_readlane_b32 s5, v40, 1
+; GFX10-NEXT:    s_sub_u32 s32, s32, 0x200
+; GFX10-NEXT:    v_readlane_b32 s33, v40, 2
+; GFX10-NEXT:    s_or_saveexec_b32 s6, -1
+; GFX10-NEXT:    buffer_load_dword v40, off, s[0:3], s32 ; 4-byte Folded Reload
+; GFX10-NEXT:    s_waitcnt_depctr 0xffe3
+; GFX10-NEXT:    s_mov_b32 exec_lo, s6
+; GFX10-NEXT:    s_waitcnt vmcnt(0)
+; GFX10-NEXT:    s_setpc_b64 s[4:5]
+  call amdgpu_gfx void @external_void_func_v3f16(<3 x half> <half 1.0, half 2.0, half 4.0>)
+  ret void
+}
+
+define amdgpu_gfx void @test_call_external_void_func_v4i16() #0 {
+; GFX9-LABEL: test_call_external_void_func_v4i16:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    s_or_saveexec_b64 s[4:5], -1
+; GFX9-NEXT:    buffer_store_dword v40, off, s[0:3], s32 ; 4-byte Folded Spill
+; GFX9-NEXT:    s_mov_b64 exec, s[4:5]
+; GFX9-NEXT:    v_writelane_b32 v40, s33, 2
+; GFX9-NEXT:    v_writelane_b32 v40, s30, 0
+; GFX9-NEXT:    v_writelane_b32 v40, s31, 1
+; GFX9-NEXT:    global_load_dwordx2 v[0:1], v[0:1], off
+; GFX9-NEXT:    s_mov_b32 s33, s32
+; GFX9-NEXT:    s_add_u32 s32, s32, 0x400
+; GFX9-NEXT:    s_getpc_b64 s[4:5]
+; GFX9-NEXT:    s_add_u32 s4, s4, external_void_func_v4i16@rel32@lo+4
+; GFX9-NEXT:    s_addc_u32 s5, s5, external_void_func_v4i16@rel32@hi+12
+; GFX9-NEXT:    s_swappc_b64 s[30:31], s[4:5]
+; GFX9-NEXT:    v_readlane_b32 s4, v40, 0
+; GFX9-NEXT:    v_readlane_b32 s5, v40, 1
+; GFX9-NEXT:    s_sub_u32 s32, s32, 0x400
+; GFX9-NEXT:    v_readlane_b32 s33, v40, 2
+; GFX9-NEXT:    s_or_saveexec_b64 s[6:7], -1
+; GFX9-NEXT:    buffer_load_dword v40, off, s[0:3], s32 ; 4-byte Folded Reload
+; GFX9-NEXT:    s_mov_b64 exec, s[6:7]
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-NEXT:    s_setpc_b64 s[4:5]
+;
+; GFX10-LABEL: test_call_external_void_func_v4i16:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
+; GFX10-NEXT:    s_or_saveexec_b32 s4, -1
+; GFX10-NEXT:    buffer_store_dword v40, off, s[0:3], s32 ; 4-byte Folded Spill
+; GFX10-NEXT:    s_waitcnt_depctr 0xffe3
+; GFX10-NEXT:    s_mov_b32 exec_lo, s4
+; GFX10-NEXT:    v_writelane_b32 v40, s33, 2
+; GFX10-NEXT:    s_mov_b32 s33, s32
+; GFX10-NEXT:    s_add_u32 s32, s32, 0x200
+; GFX10-NEXT:    ; implicit-def: $vcc_hi
+; GFX10-NEXT:    s_getpc_b64 s[4:5]
+; GFX10-NEXT:    s_add_u32 s4, s4, external_void_func_v4i16@rel32@lo+4
+; GFX10-NEXT:    s_addc_u32 s5, s5, external_void_func_v4i16@rel32@hi+12
+; GFX10-NEXT:    v_writelane_b32 v40, s30, 0
+; GFX10-NEXT:    v_writelane_b32 v40, s31, 1
+; GFX10-NEXT:    global_load_dwordx2 v[0:1], v[0:1], off
+; GFX10-NEXT:    s_swappc_b64 s[30:31], s[4:5]
+; GFX10-NEXT:    v_readlane_b32 s4, v40, 0
+; GFX10-NEXT:    v_readlane_b32 s5, v40, 1
+; GFX10-NEXT:    s_sub_u32 s32, s32, 0x200
+; GFX10-NEXT:    v_readlane_b32 s33, v40, 2
+; GFX10-NEXT:    s_or_saveexec_b32 s6, -1
+; GFX10-NEXT:    buffer_load_dword v40, off, s[0:3], s32 ; 4-byte Folded Reload
+; GFX10-NEXT:    s_waitcnt_depctr 0xffe3
+; GFX10-NEXT:    s_mov_b32 exec_lo, s6
+; GFX10-NEXT:    s_waitcnt vmcnt(0)
+; GFX10-NEXT:    s_setpc_b64 s[4:5]
+  %val = load <4 x i16>, <4 x i16> addrspace(1)* undef
+  call amdgpu_gfx void @external_void_func_v4i16(<4 x i16> %val)
+  ret void
+}
+
+define amdgpu_gfx void @test_call_external_void_func_v4i16_imm() #0 {
+; GFX9-LABEL: test_call_external_void_func_v4i16_imm:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    s_or_saveexec_b64 s[4:5], -1
+; GFX9-NEXT:    buffer_store_dword v40, off, s[0:3], s32 ; 4-byte Folded Spill
+; GFX9-NEXT:    s_mov_b64 exec, s[4:5]
+; GFX9-NEXT:    v_writelane_b32 v40, s33, 2
+; GFX9-NEXT:    v_writelane_b32 v40, s30, 0
+; GFX9-NEXT:    s_mov_b32 s33, s32
+; GFX9-NEXT:    s_add_u32 s32, s32, 0x400
+; GFX9-NEXT:    v_mov_b32_e32 v0, 0x20001
+; GFX9-NEXT:    v_mov_b32_e32 v1, 0x40003
+; GFX9-NEXT:    s_getpc_b64 s[4:5]
+; GFX9-NEXT:    s_add_u32 s4, s4, external_void_func_v4i16@rel32@lo+4
+; GFX9-NEXT:    s_addc_u32 s5, s5, external_void_func_v4i16@rel32@hi+12
+; GFX9-NEXT:    v_writelane_b32 v40, s31, 1
+; GFX9-NEXT:    s_swappc_b64 s[30:31], s[4:5]
+; GFX9-NEXT:    v_readlane_b32 s4, v40, 0
+; GFX9-NEXT:    v_readlane_b32 s5, v40, 1
+; GFX9-NEXT:    s_sub_u32 s32, s32, 0x400
+; GFX9-NEXT:    v_readlane_b32 s33, v40, 2
+; GFX9-NEXT:    s_or_saveexec_b64 s[6:7], -1
+; GFX9-NEXT:    buffer_load_dword v40, off, s[0:3], s32 ; 4-byte Folded Reload
+; GFX9-NEXT:    s_mov_b64 exec, s[6:7]
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-NEXT:    s_setpc_b64 s[4:5]
+;
+; GFX10-LABEL: test_call_external_void_func_v4i16_imm:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
+; GFX10-NEXT:    s_or_saveexec_b32 s4, -1
+; GFX10-NEXT:    buffer_store_dword v40, off, s[0:3], s32 ; 4-byte Folded Spill
+; GFX10-NEXT:    s_waitcnt_depctr 0xffe3
+; GFX10-NEXT:    s_mov_b32 exec_lo, s4
+; GFX10-NEXT:    v_writelane_b32 v40, s33, 2
+; GFX10-NEXT:    v_mov_b32_e32 v0, 0x20001
+; GFX10-NEXT:    v_mov_b32_e32 v1, 0x40003
+; GFX10-NEXT:    s_mov_b32 s33, s32
+; GFX10-NEXT:    s_add_u32 s32, s32, 0x200
+; GFX10-NEXT:    v_writelane_b32 v40, s30, 0
+; GFX10-NEXT:    s_getpc_b64 s[4:5]
+; GFX10-NEXT:    s_add_u32 s4, s4, external_void_func_v4i16@rel32@lo+4
+; GFX10-NEXT:    s_addc_u32 s5, s5, external_void_func_v4i16@rel32@hi+12
+; GFX10-NEXT:    ; implicit-def: $vcc_hi
+; GFX10-NEXT:    v_writelane_b32 v40, s31, 1
+; GFX10-NEXT:    s_swappc_b64 s[30:31], s[4:5]
+; GFX10-NEXT:    v_readlane_b32 s4, v40, 0
+; GFX10-NEXT:    v_readlane_b32 s5, v40, 1
+; GFX10-NEXT:    s_sub_u32 s32, s32, 0x200
+; GFX10-NEXT:    v_readlane_b32 s33, v40, 2
+; GFX10-NEXT:    s_or_saveexec_b32 s6, -1
+; GFX10-NEXT:    buffer_load_dword v40, off, s[0:3], s32 ; 4-byte Folded Reload
+; GFX10-NEXT:    s_waitcnt_depctr 0xffe3
+; GFX10-NEXT:    s_mov_b32 exec_lo, s6
+; GFX10-NEXT:    s_waitcnt vmcnt(0)
+; GFX10-NEXT:    s_setpc_b64 s[4:5]
+  call amdgpu_gfx void @external_void_func_v4i16(<4 x i16> <i16 1, i16 2, i16 3, i16 4>)
+  ret void
+}
+
+define amdgpu_gfx void @test_call_external_void_func_v2f16() #0 {
+; GFX9-LABEL: test_call_external_void_func_v2f16:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    s_or_saveexec_b64 s[4:5], -1
+; GFX9-NEXT:    buffer_store_dword v40, off, s[0:3], s32 ; 4-byte Folded Spill
+; GFX9-NEXT:    s_mov_b64 exec, s[4:5]
+; GFX9-NEXT:    v_writelane_b32 v40, s33, 2
+; GFX9-NEXT:    v_writelane_b32 v40, s30, 0
+; GFX9-NEXT:    v_writelane_b32 v40, s31, 1
+; GFX9-NEXT:    global_load_dword v0, v[0:1], off
+; GFX9-NEXT:    s_mov_b32 s33, s32
+; GFX9-NEXT:    s_add_u32 s32, s32, 0x400
+; GFX9-NEXT:    s_getpc_b64 s[4:5]
+; GFX9-NEXT:    s_add_u32 s4, s4, external_void_func_v2f16@rel32@lo+4
+; GFX9-NEXT:    s_addc_u32 s5, s5, external_void_func_v2f16@rel32@hi+12
+; GFX9-NEXT:    s_swappc_b64 s[30:31], s[4:5]
+; GFX9-NEXT:    v_readlane_b32 s4, v40, 0
+; GFX9-NEXT:    v_readlane_b32 s5, v40, 1
+; GFX9-NEXT:    s_sub_u32 s32, s32, 0x400
+; GFX9-NEXT:    v_readlane_b32 s33, v40, 2
+; GFX9-NEXT:    s_or_saveexec_b64 s[6:7], -1
+; GFX9-NEXT:    buffer_load_dword v40, off, s[0:3], s32 ; 4-byte Folded Reload
+; GFX9-NEXT:    s_mov_b64 exec, s[6:7]
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-NEXT:    s_setpc_b64 s[4:5]
+;
+; GFX10-LABEL: test_call_external_void_func_v2f16:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
+; GFX10-NEXT:    s_or_saveexec_b32 s4, -1
+; GFX10-NEXT:    buffer_store_dword v40, off, s[0:3], s32 ; 4-byte Folded Spill
+; GFX10-NEXT:    s_waitcnt_depctr 0xffe3
+; GFX10-NEXT:    s_mov_b32 exec_lo, s4
+; GFX10-NEXT:    v_writelane_b32 v40, s33, 2
+; GFX10-NEXT:    s_mov_b32 s33, s32
+; GFX10-NEXT:    s_add_u32 s32, s32, 0x200
+; GFX10-NEXT:    ; implicit-def: $vcc_hi
+; GFX10-NEXT:    s_getpc_b64 s[4:5]
+; GFX10-NEXT:    s_add_u32 s4, s4, external_void_func_v2f16@rel32@lo+4
+; GFX10-NEXT:    s_addc_u32 s5, s5, external_void_func_v2f16@rel32@hi+12
+; GFX10-NEXT:    v_writelane_b32 v40, s30, 0
+; GFX10-NEXT:    v_writelane_b32 v40, s31, 1
+; GFX10-NEXT:    global_load_dword v0, v[0:1], off
+; GFX10-NEXT:    s_swappc_b64 s[30:31], s[4:5]
+; GFX10-NEXT:    v_readlane_b32 s4, v40, 0
+; GFX10-NEXT:    v_readlane_b32 s5, v40, 1
+; GFX10-NEXT:    s_sub_u32 s32, s32, 0x200
+; GFX10-NEXT:    v_readlane_b32 s33, v40, 2
+; GFX10-NEXT:    s_or_saveexec_b32 s6, -1
+; GFX10-NEXT:    buffer_load_dword v40, off, s[0:3], s32 ; 4-byte Folded Reload
+; GFX10-NEXT:    s_waitcnt_depctr 0xffe3
+; GFX10-NEXT:    s_mov_b32 exec_lo, s6
+; GFX10-NEXT:    s_waitcnt vmcnt(0)
+; GFX10-NEXT:    s_setpc_b64 s[4:5]
+  %val = load <2 x half>, <2 x half> addrspace(1)* undef
+  call amdgpu_gfx void @external_void_func_v2f16(<2 x half> %val)
+  ret void
+}
+
+define amdgpu_gfx void @test_call_external_void_func_v2i32() #0 {
+; GFX9-LABEL: test_call_external_void_func_v2i32:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    s_or_saveexec_b64 s[4:5], -1
+; GFX9-NEXT:    buffer_store_dword v40, off, s[0:3], s32 ; 4-byte Folded Spill
+; GFX9-NEXT:    s_mov_b64 exec, s[4:5]
+; GFX9-NEXT:    v_writelane_b32 v40, s33, 2
+; GFX9-NEXT:    v_writelane_b32 v40, s30, 0
+; GFX9-NEXT:    v_writelane_b32 v40, s31, 1
+; GFX9-NEXT:    global_load_dwordx2 v[0:1], v[0:1], off
+; GFX9-NEXT:    s_mov_b32 s33, s32
+; GFX9-NEXT:    s_add_u32 s32, s32, 0x400
+; GFX9-NEXT:    s_getpc_b64 s[4:5]
+; GFX9-NEXT:    s_add_u32 s4, s4, external_void_func_v2i32@rel32@lo+4
+; GFX9-NEXT:    s_addc_u32 s5, s5, external_void_func_v2i32@rel32@hi+12
+; GFX9-NEXT:    s_swappc_b64 s[30:31], s[4:5]
+; GFX9-NEXT:    v_readlane_b32 s4, v40, 0
+; GFX9-NEXT:    v_readlane_b32 s5, v40, 1
+; GFX9-NEXT:    s_sub_u32 s32, s32, 0x400
+; GFX9-NEXT:    v_readlane_b32 s33, v40, 2
+; GFX9-NEXT:    s_or_saveexec_b64 s[6:7], -1
+; GFX9-NEXT:    buffer_load_dword v40, off, s[0:3], s32 ; 4-byte Folded Reload
+; GFX9-NEXT:    s_mov_b64 exec, s[6:7]
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-NEXT:    s_setpc_b64 s[4:5]
+;
+; GFX10-LABEL: test_call_external_void_func_v2i32:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
+; GFX10-NEXT:    s_or_saveexec_b32 s4, -1
+; GFX10-NEXT:    buffer_store_dword v40, off, s[0:3], s32 ; 4-byte Folded Spill
+; GFX10-NEXT:    s_waitcnt_depctr 0xffe3
+; GFX10-NEXT:    s_mov_b32 exec_lo, s4
+; GFX10-NEXT:    v_writelane_b32 v40, s33, 2
+; GFX10-NEXT:    s_mov_b32 s33, s32
+; GFX10-NEXT:    s_add_u32 s32, s32, 0x200
+; GFX10-NEXT:    ; implicit-def: $vcc_hi
+; GFX10-NEXT:    s_getpc_b64 s[4:5]
+; GFX10-NEXT:    s_add_u32 s4, s4, external_void_func_v2i32@rel32@lo+4
+; GFX10-NEXT:    s_addc_u32 s5, s5, external_void_func_v2i32@rel32@hi+12
+; GFX10-NEXT:    v_writelane_b32 v40, s30, 0
+; GFX10-NEXT:    v_writelane_b32 v40, s31, 1
+; GFX10-NEXT:    global_load_dwordx2 v[0:1], v[0:1], off
+; GFX10-NEXT:    s_swappc_b64 s[30:31], s[4:5]
+; GFX10-NEXT:    v_readlane_b32 s4, v40, 0
+; GFX10-NEXT:    v_readlane_b32 s5, v40, 1
+; GFX10-NEXT:    s_sub_u32 s32, s32, 0x200
+; GFX10-NEXT:    v_readlane_b32 s33, v40, 2
+; GFX10-NEXT:    s_or_saveexec_b32 s6, -1
+; GFX10-NEXT:    buffer_load_dword v40, off, s[0:3], s32 ; 4-byte Folded Reload
+; GFX10-NEXT:    s_waitcnt_depctr 0xffe3
+; GFX10-NEXT:    s_mov_b32 exec_lo, s6
+; GFX10-NEXT:    s_waitcnt vmcnt(0)
+; GFX10-NEXT:    s_setpc_b64 s[4:5]
+  %val = load <2 x i32>, <2 x i32> addrspace(1)* undef
+  call amdgpu_gfx void @external_void_func_v2i32(<2 x i32> %val)
+  ret void
+}
+
+define amdgpu_gfx void @test_call_external_void_func_v2i32_imm() #0 {
+; GFX9-LABEL: test_call_external_void_func_v2i32_imm:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    s_or_saveexec_b64 s[4:5], -1
+; GFX9-NEXT:    buffer_store_dword v40, off, s[0:3], s32 ; 4-byte Folded Spill
+; GFX9-NEXT:    s_mov_b64 exec, s[4:5]
+; GFX9-NEXT:    v_writelane_b32 v40, s33, 2
+; GFX9-NEXT:    v_writelane_b32 v40, s30, 0
+; GFX9-NEXT:    s_mov_b32 s33, s32
+; GFX9-NEXT:    s_add_u32 s32, s32, 0x400
+; GFX9-NEXT:    v_mov_b32_e32 v0, 1
+; GFX9-NEXT:    v_mov_b32_e32 v1, 2
+; GFX9-NEXT:    s_getpc_b64 s[4:5]
+; GFX9-NEXT:    s_add_u32 s4, s4, external_void_func_v2i32@rel32@lo+4
+; GFX9-NEXT:    s_addc_u32 s5, s5, external_void_func_v2i32@rel32@hi+12
+; GFX9-NEXT:    v_writelane_b32 v40, s31, 1
+; GFX9-NEXT:    s_swappc_b64 s[30:31], s[4:5]
+; GFX9-NEXT:    v_readlane_b32 s4, v40, 0
+; GFX9-NEXT:    v_readlane_b32 s5, v40, 1
+; GFX9-NEXT:    s_sub_u32 s32, s32, 0x400
+; GFX9-NEXT:    v_readlane_b32 s33, v40, 2
+; GFX9-NEXT:    s_or_saveexec_b64 s[6:7], -1
+; GFX9-NEXT:    buffer_load_dword v40, off, s[0:3], s32 ; 4-byte Folded Reload
+; GFX9-NEXT:    s_mov_b64 exec, s[6:7]
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-NEXT:    s_setpc_b64 s[4:5]
+;
+; GFX10-LABEL: test_call_external_void_func_v2i32_imm:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
+; GFX10-NEXT:    s_or_saveexec_b32 s4, -1
+; GFX10-NEXT:    buffer_store_dword v40, off, s[0:3], s32 ; 4-byte Folded Spill
+; GFX10-NEXT:    s_waitcnt_depctr 0xffe3
+; GFX10-NEXT:    s_mov_b32 exec_lo, s4
+; GFX10-NEXT:    v_writelane_b32 v40, s33, 2
+; GFX10-NEXT:    v_mov_b32_e32 v0, 1
+; GFX10-NEXT:    v_mov_b32_e32 v1, 2
+; GFX10-NEXT:    s_mov_b32 s33, s32
+; GFX10-NEXT:    s_add_u32 s32, s32, 0x200
+; GFX10-NEXT:    v_writelane_b32 v40, s30, 0
+; GFX10-NEXT:    s_getpc_b64 s[4:5]
+; GFX10-NEXT:    s_add_u32 s4, s4, external_void_func_v2i32@rel32@lo+4
+; GFX10-NEXT:    s_addc_u32 s5, s5, external_void_func_v2i32@rel32@hi+12
+; GFX10-NEXT:    ; implicit-def: $vcc_hi
+; GFX10-NEXT:    v_writelane_b32 v40, s31, 1
+; GFX10-NEXT:    s_swappc_b64 s[30:31], s[4:5]
+; GFX10-NEXT:    v_readlane_b32 s4, v40, 0
+; GFX10-NEXT:    v_readlane_b32 s5, v40, 1
+; GFX10-NEXT:    s_sub_u32 s32, s32, 0x200
+; GFX10-NEXT:    v_readlane_b32 s33, v40, 2
+; GFX10-NEXT:    s_or_saveexec_b32 s6, -1
+; GFX10-NEXT:    buffer_load_dword v40, off, s[0:3], s32 ; 4-byte Folded Reload
+; GFX10-NEXT:    s_waitcnt_depctr 0xffe3
+; GFX10-NEXT:    s_mov_b32 exec_lo, s6
+; GFX10-NEXT:    s_waitcnt vmcnt(0)
+; GFX10-NEXT:    s_setpc_b64 s[4:5]
+  call amdgpu_gfx void @external_void_func_v2i32(<2 x i32> <i32 1, i32 2>)
+  ret void
+}
+
+define amdgpu_gfx void @test_call_external_void_func_v3i32_imm(i32) #0 {
+; GFX9-LABEL: test_call_external_void_func_v3i32_imm:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    s_or_saveexec_b64 s[4:5], -1
+; GFX9-NEXT:    buffer_store_dword v40, off, s[0:3], s32 ; 4-byte Folded Spill
+; GFX9-NEXT:    s_mov_b64 exec, s[4:5]
+; GFX9-NEXT:    v_writelane_b32 v40, s33, 2
+; GFX9-NEXT:    v_writelane_b32 v40, s30, 0
+; GFX9-NEXT:    s_mov_b32 s33, s32
+; GFX9-NEXT:    s_add_u32 s32, s32, 0x400
+; GFX9-NEXT:    v_mov_b32_e32 v0, 3
+; GFX9-NEXT:    v_mov_b32_e32 v1, 4
+; GFX9-NEXT:    v_mov_b32_e32 v2, 5
+; GFX9-NEXT:    s_getpc_b64 s[4:5]
+; GFX9-NEXT:    s_add_u32 s4, s4, external_void_func_v3i32@rel32@lo+4
+; GFX9-NEXT:    s_addc_u32 s5, s5, external_void_func_v3i32@rel32@hi+12
+; GFX9-NEXT:    v_writelane_b32 v40, s31, 1
+; GFX9-NEXT:    s_swappc_b64 s[30:31], s[4:5]
+; GFX9-NEXT:    v_readlane_b32 s4, v40, 0
+; GFX9-NEXT:    v_readlane_b32 s5, v40, 1
+; GFX9-NEXT:    s_sub_u32 s32, s32, 0x400
+; GFX9-NEXT:    v_readlane_b32 s33, v40, 2
+; GFX9-NEXT:    s_or_saveexec_b64 s[6:7], -1
+; GFX9-NEXT:    buffer_load_dword v40, off, s[0:3], s32 ; 4-byte Folded Reload
+; GFX9-NEXT:    s_mov_b64 exec, s[6:7]
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-NEXT:    s_setpc_b64 s[4:5]
+;
+; GFX10-LABEL: test_call_external_void_func_v3i32_imm:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
+; GFX10-NEXT:    s_or_saveexec_b32 s4, -1
+; GFX10-NEXT:    buffer_store_dword v40, off, s[0:3], s32 ; 4-byte Folded Spill
+; GFX10-NEXT:    s_waitcnt_depctr 0xffe3
+; GFX10-NEXT:    s_mov_b32 exec_lo, s4
+; GFX10-NEXT:    v_writelane_b32 v40, s33, 2
+; GFX10-NEXT:    v_mov_b32_e32 v0, 3
+; GFX10-NEXT:    v_mov_b32_e32 v1, 4
+; GFX10-NEXT:    v_mov_b32_e32 v2, 5
+; GFX10-NEXT:    s_mov_b32 s33, s32
+; GFX10-NEXT:    v_writelane_b32 v40, s30, 0
+; GFX10-NEXT:    s_add_u32 s32, s32, 0x200
+; GFX10-NEXT:    s_getpc_b64 s[4:5]
+; GFX10-NEXT:    s_add_u32 s4, s4, external_void_func_v3i32@rel32@lo+4
+; GFX10-NEXT:    s_addc_u32 s5, s5, external_void_func_v3i32@rel32@hi+12
+; GFX10-NEXT:    ; implicit-def: $vcc_hi
+; GFX10-NEXT:    v_writelane_b32 v40, s31, 1
+; GFX10-NEXT:    s_swappc_b64 s[30:31], s[4:5]
+; GFX10-NEXT:    v_readlane_b32 s4, v40, 0
+; GFX10-NEXT:    v_readlane_b32 s5, v40, 1
+; GFX10-NEXT:    s_sub_u32 s32, s32, 0x200
+; GFX10-NEXT:    v_readlane_b32 s33, v40, 2
+; GFX10-NEXT:    s_or_saveexec_b32 s6, -1
+; GFX10-NEXT:    buffer_load_dword v40, off, s[0:3], s32 ; 4-byte Folded Reload
+; GFX10-NEXT:    s_waitcnt_depctr 0xffe3
+; GFX10-NEXT:    s_mov_b32 exec_lo, s6
+; GFX10-NEXT:    s_waitcnt vmcnt(0)
+; GFX10-NEXT:    s_setpc_b64 s[4:5]
+  call amdgpu_gfx void @external_void_func_v3i32(<3 x i32> <i32 3, i32 4, i32 5>)
+  ret void
+}
+
+define amdgpu_gfx void @test_call_external_void_func_v3i32_i32(i32) #0 {
+; GFX9-LABEL: test_call_external_void_func_v3i32_i32:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    s_or_saveexec_b64 s[4:5], -1
+; GFX9-NEXT:    buffer_store_dword v40, off, s[0:3], s32 ; 4-byte Folded Spill
+; GFX9-NEXT:    s_mov_b64 exec, s[4:5]
+; GFX9-NEXT:    v_writelane_b32 v40, s33, 2
+; GFX9-NEXT:    v_writelane_b32 v40, s30, 0
+; GFX9-NEXT:    s_mov_b32 s33, s32
+; GFX9-NEXT:    s_add_u32 s32, s32, 0x400
+; GFX9-NEXT:    v_mov_b32_e32 v0, 3
+; GFX9-NEXT:    v_mov_b32_e32 v1, 4
+; GFX9-NEXT:    v_mov_b32_e32 v2, 5
+; GFX9-NEXT:    v_mov_b32_e32 v3, 6
+; GFX9-NEXT:    s_getpc_b64 s[4:5]
+; GFX9-NEXT:    s_add_u32 s4, s4, external_void_func_v3i32_i32@rel32@lo+4
+; GFX9-NEXT:    s_addc_u32 s5, s5, external_void_func_v3i32_i32@rel32@hi+12
+; GFX9-NEXT:    v_writelane_b32 v40, s31, 1
+; GFX9-NEXT:    s_swappc_b64 s[30:31], s[4:5]
+; GFX9-NEXT:    v_readlane_b32 s4, v40, 0
+; GFX9-NEXT:    v_readlane_b32 s5, v40, 1
+; GFX9-NEXT:    s_sub_u32 s32, s32, 0x400
+; GFX9-NEXT:    v_readlane_b32 s33, v40, 2
+; GFX9-NEXT:    s_or_saveexec_b64 s[6:7], -1
+; GFX9-NEXT:    buffer_load_dword v40, off, s[0:3], s32 ; 4-byte Folded Reload
+; GFX9-NEXT:    s_mov_b64 exec, s[6:7]
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-NEXT:    s_setpc_b64 s[4:5]
+;
+; GFX10-LABEL: test_call_external_void_func_v3i32_i32:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
+; GFX10-NEXT:    s_or_saveexec_b32 s4, -1
+; GFX10-NEXT:    buffer_store_dword v40, off, s[0:3], s32 ; 4-byte Folded Spill
+; GFX10-NEXT:    s_waitcnt_depctr 0xffe3
+; GFX10-NEXT:    s_mov_b32 exec_lo, s4
+; GFX10-NEXT:    v_writelane_b32 v40, s33, 2
+; GFX10-NEXT:    v_mov_b32_e32 v0, 3
+; GFX10-NEXT:    v_mov_b32_e32 v1, 4
+; GFX10-NEXT:    v_mov_b32_e32 v2, 5
+; GFX10-NEXT:    v_mov_b32_e32 v3, 6
+; GFX10-NEXT:    v_writelane_b32 v40, s30, 0
+; GFX10-NEXT:    s_mov_b32 s33, s32
+; GFX10-NEXT:    s_add_u32 s32, s32, 0x200
+; GFX10-NEXT:    s_getpc_b64 s[4:5]
+; GFX10-NEXT:    s_add_u32 s4, s4, external_void_func_v3i32_i32@rel32@lo+4
+; GFX10-NEXT:    s_addc_u32 s5, s5, external_void_func_v3i32_i32@rel32@hi+12
+; GFX10-NEXT:    ; implicit-def: $vcc_hi
+; GFX10-NEXT:    v_writelane_b32 v40, s31, 1
+; GFX10-NEXT:    s_swappc_b64 s[30:31], s[4:5]
+; GFX10-NEXT:    v_readlane_b32 s4, v40, 0
+; GFX10-NEXT:    v_readlane_b32 s5, v40, 1
+; GFX10-NEXT:    s_sub_u32 s32, s32, 0x200
+; GFX10-NEXT:    v_readlane_b32 s33, v40, 2
+; GFX10-NEXT:    s_or_saveexec_b32 s6, -1
+; GFX10-NEXT:    buffer_load_dword v40, off, s[0:3], s32 ; 4-byte Folded Reload
+; GFX10-NEXT:    s_waitcnt_depctr 0xffe3
+; GFX10-NEXT:    s_mov_b32 exec_lo, s6
+; GFX10-NEXT:    s_waitcnt vmcnt(0)
+; GFX10-NEXT:    s_setpc_b64 s[4:5]
+  call amdgpu_gfx void @external_void_func_v3i32_i32(<3 x i32> <i32 3, i32 4, i32 5>, i32 6)
+  ret void
+}
+
+define amdgpu_gfx void @test_call_external_void_func_v4i32() #0 {
+; GFX9-LABEL: test_call_external_void_func_v4i32:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    s_or_saveexec_b64 s[4:5], -1
+; GFX9-NEXT:    buffer_store_dword v40, off, s[0:3], s32 ; 4-byte Folded Spill
+; GFX9-NEXT:    s_mov_b64 exec, s[4:5]
+; GFX9-NEXT:    v_writelane_b32 v40, s33, 2
+; GFX9-NEXT:    v_writelane_b32 v40, s30, 0
+; GFX9-NEXT:    v_writelane_b32 v40, s31, 1
+; GFX9-NEXT:    global_load_dwordx4 v[0:3], v[0:1], off
+; GFX9-NEXT:    s_mov_b32 s33, s32
+; GFX9-NEXT:    s_add_u32 s32, s32, 0x400
+; GFX9-NEXT:    s_getpc_b64 s[4:5]
+; GFX9-NEXT:    s_add_u32 s4, s4, external_void_func_v4i32@rel32@lo+4
+; GFX9-NEXT:    s_addc_u32 s5, s5, external_void_func_v4i32@rel32@hi+12
+; GFX9-NEXT:    s_swappc_b64 s[30:31], s[4:5]
+; GFX9-NEXT:    v_readlane_b32 s4, v40, 0
+; GFX9-NEXT:    v_readlane_b32 s5, v40, 1
+; GFX9-NEXT:    s_sub_u32 s32, s32, 0x400
+; GFX9-NEXT:    v_readlane_b32 s33, v40, 2
+; GFX9-NEXT:    s_or_saveexec_b64 s[6:7], -1
+; GFX9-NEXT:    buffer_load_dword v40, off, s[0:3], s32 ; 4-byte Folded Reload
+; GFX9-NEXT:    s_mov_b64 exec, s[6:7]
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-NEXT:    s_setpc_b64 s[4:5]
+;
+; GFX10-LABEL: test_call_external_void_func_v4i32:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
+; GFX10-NEXT:    s_or_saveexec_b32 s4, -1
+; GFX10-NEXT:    buffer_store_dword v40, off, s[0:3], s32 ; 4-byte Folded Spill
+; GFX10-NEXT:    s_waitcnt_depctr 0xffe3
+; GFX10-NEXT:    s_mov_b32 exec_lo, s4
+; GFX10-NEXT:    v_writelane_b32 v40, s33, 2
+; GFX10-NEXT:    s_mov_b32 s33, s32
+; GFX10-NEXT:    s_add_u32 s32, s32, 0x200
+; GFX10-NEXT:    ; implicit-def: $vcc_hi
+; GFX10-NEXT:    s_getpc_b64 s[4:5]
+; GFX10-NEXT:    s_add_u32 s4, s4, external_void_func_v4i32@rel32@lo+4
+; GFX10-NEXT:    s_addc_u32 s5, s5, external_void_func_v4i32@rel32@hi+12
+; GFX10-NEXT:    v_writelane_b32 v40, s30, 0
+; GFX10-NEXT:    v_writelane_b32 v40, s31, 1
+; GFX10-NEXT:    global_load_dwordx4 v[0:3], v[0:1], off
+; GFX10-NEXT:    s_swappc_b64 s[30:31], s[4:5]
+; GFX10-NEXT:    v_readlane_b32 s4, v40, 0
+; GFX10-NEXT:    v_readlane_b32 s5, v40, 1
+; GFX10-NEXT:    s_sub_u32 s32, s32, 0x200
+; GFX10-NEXT:    v_readlane_b32 s33, v40, 2
+; GFX10-NEXT:    s_or_saveexec_b32 s6, -1
+; GFX10-NEXT:    buffer_load_dword v40, off, s[0:3], s32 ; 4-byte Folded Reload
+; GFX10-NEXT:    s_waitcnt_depctr 0xffe3
+; GFX10-NEXT:    s_mov_b32 exec_lo, s6
+; GFX10-NEXT:    s_waitcnt vmcnt(0)
+; GFX10-NEXT:    s_setpc_b64 s[4:5]
+  %val = load <4 x i32>, <4 x i32> addrspace(1)* undef
+  call amdgpu_gfx void @external_void_func_v4i32(<4 x i32> %val)
+  ret void
+}
+
+define amdgpu_gfx void @test_call_external_void_func_v4i32_imm() #0 {
+; GFX9-LABEL: test_call_external_void_func_v4i32_imm:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    s_or_saveexec_b64 s[4:5], -1
+; GFX9-NEXT:    buffer_store_dword v40, off, s[0:3], s32 ; 4-byte Folded Spill
+; GFX9-NEXT:    s_mov_b64 exec, s[4:5]
+; GFX9-NEXT:    v_writelane_b32 v40, s33, 2
+; GFX9-NEXT:    v_writelane_b32 v40, s30, 0
+; GFX9-NEXT:    s_mov_b32 s33, s32
+; GFX9-NEXT:    s_add_u32 s32, s32, 0x400
+; GFX9-NEXT:    v_mov_b32_e32 v0, 1
+; GFX9-NEXT:    v_mov_b32_e32 v1, 2
+; GFX9-NEXT:    v_mov_b32_e32 v2, 3
+; GFX9-NEXT:    v_mov_b32_e32 v3, 4
+; GFX9-NEXT:    s_getpc_b64 s[4:5]
+; GFX9-NEXT:    s_add_u32 s4, s4, external_void_func_v4i32@rel32@lo+4
+; GFX9-NEXT:    s_addc_u32 s5, s5, external_void_func_v4i32@rel32@hi+12
+; GFX9-NEXT:    v_writelane_b32 v40, s31, 1
+; GFX9-NEXT:    s_swappc_b64 s[30:31], s[4:5]
+; GFX9-NEXT:    v_readlane_b32 s4, v40, 0
+; GFX9-NEXT:    v_readlane_b32 s5, v40, 1
+; GFX9-NEXT:    s_sub_u32 s32, s32, 0x400
+; GFX9-NEXT:    v_readlane_b32 s33, v40, 2
+; GFX9-NEXT:    s_or_saveexec_b64 s[6:7], -1
+; GFX9-NEXT:    buffer_load_dword v40, off, s[0:3], s32 ; 4-byte Folded Reload
+; GFX9-NEXT:    s_mov_b64 exec, s[6:7]
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-NEXT:    s_setpc_b64 s[4:5]
+;
+; GFX10-LABEL: test_call_external_void_func_v4i32_imm:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
+; GFX10-NEXT:    s_or_saveexec_b32 s4, -1
+; GFX10-NEXT:    buffer_store_dword v40, off, s[0:3], s32 ; 4-byte Folded Spill
+; GFX10-NEXT:    s_waitcnt_depctr 0xffe3
+; GFX10-NEXT:    s_mov_b32 exec_lo, s4
+; GFX10-NEXT:    v_writelane_b32 v40, s33, 2
+; GFX10-NEXT:    v_mov_b32_e32 v0, 1
+; GFX10-NEXT:    v_mov_b32_e32 v1, 2
+; GFX10-NEXT:    v_mov_b32_e32 v2, 3
+; GFX10-NEXT:    v_mov_b32_e32 v3, 4
+; GFX10-NEXT:    v_writelane_b32 v40, s30, 0
+; GFX10-NEXT:    s_mov_b32 s33, s32
+; GFX10-NEXT:    s_add_u32 s32, s32, 0x200
+; GFX10-NEXT:    s_getpc_b64 s[4:5]
+; GFX10-NEXT:    s_add_u32 s4, s4, external_void_func_v4i32@rel32@lo+4
+; GFX10-NEXT:    s_addc_u32 s5, s5, external_void_func_v4i32@rel32@hi+12
+; GFX10-NEXT:    ; implicit-def: $vcc_hi
+; GFX10-NEXT:    v_writelane_b32 v40, s31, 1
+; GFX10-NEXT:    s_swappc_b64 s[30:31], s[4:5]
+; GFX10-NEXT:    v_readlane_b32 s4, v40, 0
+; GFX10-NEXT:    v_readlane_b32 s5, v40, 1
+; GFX10-NEXT:    s_sub_u32 s32, s32, 0x200
+; GFX10-NEXT:    v_readlane_b32 s33, v40, 2
+; GFX10-NEXT:    s_or_saveexec_b32 s6, -1
+; GFX10-NEXT:    buffer_load_dword v40, off, s[0:3], s32 ; 4-byte Folded Reload
+; GFX10-NEXT:    s_waitcnt_depctr 0xffe3
+; GFX10-NEXT:    s_mov_b32 exec_lo, s6
+; GFX10-NEXT:    s_waitcnt vmcnt(0)
+; GFX10-NEXT:    s_setpc_b64 s[4:5]
+  call amdgpu_gfx void @external_void_func_v4i32(<4 x i32> <i32 1, i32 2, i32 3, i32 4>)
+  ret void
+}
+
+define amdgpu_gfx void @test_call_external_void_func_v5i32_imm() #0 {
+; GFX9-LABEL: test_call_external_void_func_v5i32_imm:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    s_or_saveexec_b64 s[4:5], -1
+; GFX9-NEXT:    buffer_store_dword v40, off, s[0:3], s32 ; 4-byte Folded Spill
+; GFX9-NEXT:    s_mov_b64 exec, s[4:5]
+; GFX9-NEXT:    v_writelane_b32 v40, s33, 2
+; GFX9-NEXT:    v_writelane_b32 v40, s30, 0
+; GFX9-NEXT:    s_mov_b32 s33, s32
+; GFX9-NEXT:    s_add_u32 s32, s32, 0x400
+; GFX9-NEXT:    v_mov_b32_e32 v0, 1
+; GFX9-NEXT:    v_mov_b32_e32 v1, 2
+; GFX9-NEXT:    v_mov_b32_e32 v2, 3
+; GFX9-NEXT:    v_mov_b32_e32 v3, 4
+; GFX9-NEXT:    v_mov_b32_e32 v4, 5
+; GFX9-NEXT:    s_getpc_b64 s[4:5]
+; GFX9-NEXT:    s_add_u32 s4, s4, external_void_func_v5i32@rel32@lo+4
+; GFX9-NEXT:    s_addc_u32 s5, s5, external_void_func_v5i32@rel32@hi+12
+; GFX9-NEXT:    v_writelane_b32 v40, s31, 1
+; GFX9-NEXT:    s_swappc_b64 s[30:31], s[4:5]
+; GFX9-NEXT:    v_readlane_b32 s4, v40, 0
+; GFX9-NEXT:    v_readlane_b32 s5, v40, 1
+; GFX9-NEXT:    s_sub_u32 s32, s32, 0x400
+; GFX9-NEXT:    v_readlane_b32 s33, v40, 2
+; GFX9-NEXT:    s_or_saveexec_b64 s[6:7], -1
+; GFX9-NEXT:    buffer_load_dword v40, off, s[0:3], s32 ; 4-byte Folded Reload
+; GFX9-NEXT:    s_mov_b64 exec, s[6:7]
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-NEXT:    s_setpc_b64 s[4:5]
+;
+; GFX10-LABEL: test_call_external_void_func_v5i32_imm:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
+; GFX10-NEXT:    s_or_saveexec_b32 s4, -1
+; GFX10-NEXT:    buffer_store_dword v40, off, s[0:3], s32 ; 4-byte Folded Spill
+; GFX10-NEXT:    s_waitcnt_depctr 0xffe3
+; GFX10-NEXT:    s_mov_b32 exec_lo, s4
+; GFX10-NEXT:    v_writelane_b32 v40, s33, 2
+; GFX10-NEXT:    v_mov_b32_e32 v0, 1
+; GFX10-NEXT:    v_mov_b32_e32 v1, 2
+; GFX10-NEXT:    v_mov_b32_e32 v2, 3
+; GFX10-NEXT:    v_mov_b32_e32 v3, 4
+; GFX10-NEXT:    v_writelane_b32 v40, s30, 0
+; GFX10-NEXT:    v_mov_b32_e32 v4, 5
+; GFX10-NEXT:    s_mov_b32 s33, s32
+; GFX10-NEXT:    s_add_u32 s32, s32, 0x200
+; GFX10-NEXT:    s_getpc_b64 s[4:5]
+; GFX10-NEXT:    s_add_u32 s4, s4, external_void_func_v5i32@rel32@lo+4
+; GFX10-NEXT:    s_addc_u32 s5, s5, external_void_func_v5i32@rel32@hi+12
+; GFX10-NEXT:    v_writelane_b32 v40, s31, 1
+; GFX10-NEXT:    ; implicit-def: $vcc_hi
+; GFX10-NEXT:    s_swappc_b64 s[30:31], s[4:5]
+; GFX10-NEXT:    v_readlane_b32 s4, v40, 0
+; GFX10-NEXT:    v_readlane_b32 s5, v40, 1
+; GFX10-NEXT:    s_sub_u32 s32, s32, 0x200
+; GFX10-NEXT:    v_readlane_b32 s33, v40, 2
+; GFX10-NEXT:    s_or_saveexec_b32 s6, -1
+; GFX10-NEXT:    buffer_load_dword v40, off, s[0:3], s32 ; 4-byte Folded Reload
+; GFX10-NEXT:    s_waitcnt_depctr 0xffe3
+; GFX10-NEXT:    s_mov_b32 exec_lo, s6
+; GFX10-NEXT:    s_waitcnt vmcnt(0)
+; GFX10-NEXT:    s_setpc_b64 s[4:5]
+  call amdgpu_gfx void @external_void_func_v5i32(<5 x i32> <i32 1, i32 2, i32 3, i32 4, i32 5>)
+  ret void
+}
+
+define amdgpu_gfx void @test_call_external_void_func_v8i32() #0 {
+; GFX9-LABEL: test_call_external_void_func_v8i32:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    s_or_saveexec_b64 s[4:5], -1
+; GFX9-NEXT:    buffer_store_dword v40, off, s[0:3], s32 ; 4-byte Folded Spill
+; GFX9-NEXT:    s_mov_b64 exec, s[4:5]
+; GFX9-NEXT:    s_load_dwordx2 s[4:5], s[4:5], 0x0
+; GFX9-NEXT:    v_writelane_b32 v40, s33, 2
+; GFX9-NEXT:    v_writelane_b32 v40, s30, 0
+; GFX9-NEXT:    v_writelane_b32 v40, s31, 1
+; GFX9-NEXT:    s_mov_b32 s33, s32
+; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX9-NEXT:    v_mov_b32_e32 v4, s4
+; GFX9-NEXT:    v_mov_b32_e32 v5, s5
+; GFX9-NEXT:    global_load_dwordx4 v[0:3], v[4:5], off
+; GFX9-NEXT:    global_load_dwordx4 v[4:7], v[4:5], off offset:16
+; GFX9-NEXT:    s_add_u32 s32, s32, 0x400
+; GFX9-NEXT:    s_getpc_b64 s[4:5]
+; GFX9-NEXT:    s_add_u32 s4, s4, external_void_func_v8i32@rel32@lo+4
+; GFX9-NEXT:    s_addc_u32 s5, s5, external_void_func_v8i32@rel32@hi+12
+; GFX9-NEXT:    s_swappc_b64 s[30:31], s[4:5]
+; GFX9-NEXT:    v_readlane_b32 s4, v40, 0
+; GFX9-NEXT:    v_readlane_b32 s5, v40, 1
+; GFX9-NEXT:    s_sub_u32 s32, s32, 0x400
+; GFX9-NEXT:    v_readlane_b32 s33, v40, 2
+; GFX9-NEXT:    s_or_saveexec_b64 s[6:7], -1
+; GFX9-NEXT:    buffer_load_dword v40, off, s[0:3], s32 ; 4-byte Folded Reload
+; GFX9-NEXT:    s_mov_b64 exec, s[6:7]
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-NEXT:    s_setpc_b64 s[4:5]
+;
+; GFX10-LABEL: test_call_external_void_func_v8i32:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
+; GFX10-NEXT:    s_or_saveexec_b32 s4, -1
+; GFX10-NEXT:    buffer_store_dword v40, off, s[0:3], s32 ; 4-byte Folded Spill
+; GFX10-NEXT:    s_waitcnt_depctr 0xffe3
+; GFX10-NEXT:    s_mov_b32 exec_lo, s4
+; GFX10-NEXT:    s_load_dwordx2 s[4:5], s[4:5], 0x0
+; GFX10-NEXT:    v_writelane_b32 v40, s33, 2
+; GFX10-NEXT:    s_mov_b32 s33, s32
+; GFX10-NEXT:    s_add_u32 s32, s32, 0x200
+; GFX10-NEXT:    ; implicit-def: $vcc_hi
+; GFX10-NEXT:    v_writelane_b32 v40, s30, 0
+; GFX10-NEXT:    v_writelane_b32 v40, s31, 1
+; GFX10-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-NEXT:    v_mov_b32_e32 v4, s4
+; GFX10-NEXT:    v_mov_b32_e32 v5, s5
+; GFX10-NEXT:    s_getpc_b64 s[4:5]
+; GFX10-NEXT:    s_add_u32 s4, s4, external_void_func_v8i32@rel32@lo+4
+; GFX10-NEXT:    s_addc_u32 s5, s5, external_void_func_v8i32@rel32@hi+12
+; GFX10-NEXT:    s_clause 0x1
+; GFX10-NEXT:    global_load_dwordx4 v[0:3], v[4:5], off
+; GFX10-NEXT:    global_load_dwordx4 v[4:7], v[4:5], off offset:16
+; GFX10-NEXT:    s_swappc_b64 s[30:31], s[4:5]
+; GFX10-NEXT:    v_readlane_b32 s4, v40, 0
+; GFX10-NEXT:    v_readlane_b32 s5, v40, 1
+; GFX10-NEXT:    s_sub_u32 s32, s32, 0x200
+; GFX10-NEXT:    v_readlane_b32 s33, v40, 2
+; GFX10-NEXT:    s_or_saveexec_b32 s6, -1
+; GFX10-NEXT:    buffer_load_dword v40, off, s[0:3], s32 ; 4-byte Folded Reload
+; GFX10-NEXT:    s_waitcnt_depctr 0xffe3
+; GFX10-NEXT:    s_mov_b32 exec_lo, s6
+; GFX10-NEXT:    s_waitcnt vmcnt(0)
+; GFX10-NEXT:    s_setpc_b64 s[4:5]
+  %ptr = load <8 x i32> addrspace(1)*, <8 x i32> addrspace(1)* addrspace(4)* undef
+  %val = load <8 x i32>, <8 x i32> addrspace(1)* %ptr
+  call amdgpu_gfx void @external_void_func_v8i32(<8 x i32> %val)
+  ret void
+}
+
+define amdgpu_gfx void @test_call_external_void_func_v8i32_imm() #0 {
+; GFX9-LABEL: test_call_external_void_func_v8i32_imm:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    s_or_saveexec_b64 s[4:5], -1
+; GFX9-NEXT:    buffer_store_dword v40, off, s[0:3], s32 ; 4-byte Folded Spill
+; GFX9-NEXT:    s_mov_b64 exec, s[4:5]
+; GFX9-NEXT:    v_writelane_b32 v40, s33, 2
+; GFX9-NEXT:    v_writelane_b32 v40, s30, 0
+; GFX9-NEXT:    s_mov_b32 s33, s32
+; GFX9-NEXT:    s_add_u32 s32, s32, 0x400
+; GFX9-NEXT:    v_mov_b32_e32 v0, 1
+; GFX9-NEXT:    v_mov_b32_e32 v1, 2
+; GFX9-NEXT:    v_mov_b32_e32 v2, 3
+; GFX9-NEXT:    v_mov_b32_e32 v3, 4
+; GFX9-NEXT:    v_mov_b32_e32 v4, 5
+; GFX9-NEXT:    v_mov_b32_e32 v5, 6
+; GFX9-NEXT:    v_mov_b32_e32 v6, 7
+; GFX9-NEXT:    v_mov_b32_e32 v7, 8
+; GFX9-NEXT:    s_getpc_b64 s[4:5]
+; GFX9-NEXT:    s_add_u32 s4, s4, external_void_func_v8i32@rel32@lo+4
+; GFX9-NEXT:    s_addc_u32 s5, s5, external_void_func_v8i32@rel32@hi+12
+; GFX9-NEXT:    v_writelane_b32 v40, s31, 1
+; GFX9-NEXT:    s_swappc_b64 s[30:31], s[4:5]
+; GFX9-NEXT:    v_readlane_b32 s4, v40, 0
+; GFX9-NEXT:    v_readlane_b32 s5, v40, 1
+; GFX9-NEXT:    s_sub_u32 s32, s32, 0x400
+; GFX9-NEXT:    v_readlane_b32 s33, v40, 2
+; GFX9-NEXT:    s_or_saveexec_b64 s[6:7], -1
+; GFX9-NEXT:    buffer_load_dword v40, off, s[0:3], s32 ; 4-byte Folded Reload
+; GFX9-NEXT:    s_mov_b64 exec, s[6:7]
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-NEXT:    s_setpc_b64 s[4:5]
+;
+; GFX10-LABEL: test_call_external_void_func_v8i32_imm:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
+; GFX10-NEXT:    s_or_saveexec_b32 s4, -1
+; GFX10-NEXT:    buffer_store_dword v40, off, s[0:3], s32 ; 4-byte Folded Spill
+; GFX10-NEXT:    s_waitcnt_depctr 0xffe3
+; GFX10-NEXT:    s_mov_b32 exec_lo, s4
+; GFX10-NEXT:    v_writelane_b32 v40, s33, 2
+; GFX10-NEXT:    v_mov_b32_e32 v0, 1
+; GFX10-NEXT:    v_mov_b32_e32 v1, 2
+; GFX10-NEXT:    v_mov_b32_e32 v2, 3
+; GFX10-NEXT:    v_mov_b32_e32 v3, 4
+; GFX10-NEXT:    v_writelane_b32 v40, s30, 0
+; GFX10-NEXT:    v_mov_b32_e32 v4, 5
+; GFX10-NEXT:    v_mov_b32_e32 v5, 6
+; GFX10-NEXT:    v_mov_b32_e32 v6, 7
+; GFX10-NEXT:    v_mov_b32_e32 v7, 8
+; GFX10-NEXT:    s_mov_b32 s33, s32
+; GFX10-NEXT:    s_add_u32 s32, s32, 0x200
+; GFX10-NEXT:    s_getpc_b64 s[4:5]
+; GFX10-NEXT:    s_add_u32 s4, s4, external_void_func_v8i32@rel32@lo+4
+; GFX10-NEXT:    s_addc_u32 s5, s5, external_void_func_v8i32@rel32@hi+12
+; GFX10-NEXT:    v_writelane_b32 v40, s31, 1
+; GFX10-NEXT:    ; implicit-def: $vcc_hi
+; GFX10-NEXT:    s_swappc_b64 s[30:31], s[4:5]
+; GFX10-NEXT:    v_readlane_b32 s4, v40, 0
+; GFX10-NEXT:    v_readlane_b32 s5, v40, 1
+; GFX10-NEXT:    s_sub_u32 s32, s32, 0x200
+; GFX10-NEXT:    v_readlane_b32 s33, v40, 2
+; GFX10-NEXT:    s_or_saveexec_b32 s6, -1
+; GFX10-NEXT:    buffer_load_dword v40, off, s[0:3], s32 ; 4-byte Folded Reload
+; GFX10-NEXT:    s_waitcnt_depctr 0xffe3
+; GFX10-NEXT:    s_mov_b32 exec_lo, s6
+; GFX10-NEXT:    s_waitcnt vmcnt(0)
+; GFX10-NEXT:    s_setpc_b64 s[4:5]
+  call amdgpu_gfx void @external_void_func_v8i32(<8 x i32> <i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8>)
+  ret void
+}
+
+define amdgpu_gfx void @test_call_external_void_func_v16i32() #0 {
+; GFX9-LABEL: test_call_external_void_func_v16i32:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    s_or_saveexec_b64 s[4:5], -1
+; GFX9-NEXT:    buffer_store_dword v40, off, s[0:3], s32 ; 4-byte Folded Spill
+; GFX9-NEXT:    s_mov_b64 exec, s[4:5]
+; GFX9-NEXT:    s_load_dwordx2 s[4:5], s[4:5], 0x0
+; GFX9-NEXT:    v_writelane_b32 v40, s33, 2
+; GFX9-NEXT:    v_writelane_b32 v40, s30, 0
+; GFX9-NEXT:    v_writelane_b32 v40, s31, 1
+; GFX9-NEXT:    s_mov_b32 s33, s32
+; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX9-NEXT:    v_mov_b32_e32 v13, s5
+; GFX9-NEXT:    v_mov_b32_e32 v12, s4
+; GFX9-NEXT:    global_load_dwordx4 v[0:3], v[12:13], off
+; GFX9-NEXT:    global_load_dwordx4 v[4:7], v[12:13], off offset:16
+; GFX9-NEXT:    global_load_dwordx4 v[8:11], v[12:13], off offset:32
+; GFX9-NEXT:    global_load_dwordx4 v[12:15], v[12:13], off offset:48
+; GFX9-NEXT:    s_add_u32 s32, s32, 0x400
+; GFX9-NEXT:    s_getpc_b64 s[4:5]
+; GFX9-NEXT:    s_add_u32 s4, s4, external_void_func_v16i32@rel32@lo+4
+; GFX9-NEXT:    s_addc_u32 s5, s5, external_void_func_v16i32@rel32@hi+12
+; GFX9-NEXT:    s_swappc_b64 s[30:31], s[4:5]
+; GFX9-NEXT:    v_readlane_b32 s4, v40, 0
+; GFX9-NEXT:    v_readlane_b32 s5, v40, 1
+; GFX9-NEXT:    s_sub_u32 s32, s32, 0x400
+; GFX9-NEXT:    v_readlane_b32 s33, v40, 2
+; GFX9-NEXT:    s_or_saveexec_b64 s[6:7], -1
+; GFX9-NEXT:    buffer_load_dword v40, off, s[0:3], s32 ; 4-byte Folded Reload
+; GFX9-NEXT:    s_mov_b64 exec, s[6:7]
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-NEXT:    s_setpc_b64 s[4:5]
+;
+; GFX10-LABEL: test_call_external_void_func_v16i32:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
+; GFX10-NEXT:    s_or_saveexec_b32 s4, -1
+; GFX10-NEXT:    buffer_store_dword v40, off, s[0:3], s32 ; 4-byte Folded Spill
+; GFX10-NEXT:    s_waitcnt_depctr 0xffe3
+; GFX10-NEXT:    s_mov_b32 exec_lo, s4
+; GFX10-NEXT:    s_load_dwordx2 s[4:5], s[4:5], 0x0
+; GFX10-NEXT:    v_writelane_b32 v40, s33, 2
+; GFX10-NEXT:    s_mov_b32 s33, s32
+; GFX10-NEXT:    s_add_u32 s32, s32, 0x200
+; GFX10-NEXT:    ; implicit-def: $vcc_hi
+; GFX10-NEXT:    v_writelane_b32 v40, s30, 0
+; GFX10-NEXT:    v_writelane_b32 v40, s31, 1
+; GFX10-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-NEXT:    v_mov_b32_e32 v13, s5
+; GFX10-NEXT:    v_mov_b32_e32 v12, s4
+; GFX10-NEXT:    s_getpc_b64 s[4:5]
+; GFX10-NEXT:    s_add_u32 s4, s4, external_void_func_v16i32@rel32@lo+4
+; GFX10-NEXT:    s_addc_u32 s5, s5, external_void_func_v16i32@rel32@hi+12
+; GFX10-NEXT:    s_clause 0x3
+; GFX10-NEXT:    global_load_dwordx4 v[0:3], v[12:13], off
+; GFX10-NEXT:    global_load_dwordx4 v[4:7], v[12:13], off offset:16
+; GFX10-NEXT:    global_load_dwordx4 v[8:11], v[12:13], off offset:32
+; GFX10-NEXT:    global_load_dwordx4 v[12:15], v[12:13], off offset:48
+; GFX10-NEXT:    s_swappc_b64 s[30:31], s[4:5]
+; GFX10-NEXT:    v_readlane_b32 s4, v40, 0
+; GFX10-NEXT:    v_readlane_b32 s5, v40, 1
+; GFX10-NEXT:    s_sub_u32 s32, s32, 0x200
+; GFX10-NEXT:    v_readlane_b32 s33, v40, 2
+; GFX10-NEXT:    s_or_saveexec_b32 s6, -1
+; GFX10-NEXT:    buffer_load_dword v40, off, s[0:3], s32 ; 4-byte Folded Reload
+; GFX10-NEXT:    s_waitcnt_depctr 0xffe3
+; GFX10-NEXT:    s_mov_b32 exec_lo, s6
+; GFX10-NEXT:    s_waitcnt vmcnt(0)
+; GFX10-NEXT:    s_setpc_b64 s[4:5]
+  %ptr = load <16 x i32> addrspace(1)*, <16 x i32> addrspace(1)* addrspace(4)* undef
+  %val = load <16 x i32>, <16 x i32> addrspace(1)* %ptr
+  call amdgpu_gfx void @external_void_func_v16i32(<16 x i32> %val)
+  ret void
+}
+
+define amdgpu_gfx void @test_call_external_void_func_v32i32() #0 {
+; GFX9-LABEL: test_call_external_void_func_v32i32:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    s_or_saveexec_b64 s[4:5], -1
+; GFX9-NEXT:    buffer_store_dword v40, off, s[0:3], s32 ; 4-byte Folded Spill
+; GFX9-NEXT:    s_mov_b64 exec, s[4:5]
+; GFX9-NEXT:    s_load_dwordx2 s[4:5], s[4:5], 0x0
+; GFX9-NEXT:    v_writelane_b32 v40, s33, 2
+; GFX9-NEXT:    v_writelane_b32 v40, s30, 0
+; GFX9-NEXT:    v_writelane_b32 v40, s31, 1
+; GFX9-NEXT:    s_mov_b32 s33, s32
+; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX9-NEXT:    v_mov_b32_e32 v29, s5
+; GFX9-NEXT:    v_mov_b32_e32 v28, s4
+; GFX9-NEXT:    global_load_dwordx4 v[0:3], v[28:29], off
+; GFX9-NEXT:    global_load_dwordx4 v[4:7], v[28:29], off offset:16
+; GFX9-NEXT:    global_load_dwordx4 v[8:11], v[28:29], off offset:32
+; GFX9-NEXT:    global_load_dwordx4 v[12:15], v[28:29], off offset:48
+; GFX9-NEXT:    global_load_dwordx4 v[16:19], v[28:29], off offset:64
+; GFX9-NEXT:    global_load_dwordx4 v[20:23], v[28:29], off offset:80
+; GFX9-NEXT:    global_load_dwordx4 v[24:27], v[28:29], off offset:96
+; GFX9-NEXT:    global_load_dwordx4 v[28:31], v[28:29], off offset:112
+; GFX9-NEXT:    s_add_u32 s32, s32, 0x400
+; GFX9-NEXT:    s_getpc_b64 s[4:5]
+; GFX9-NEXT:    s_add_u32 s4, s4, external_void_func_v32i32@rel32@lo+4
+; GFX9-NEXT:    s_addc_u32 s5, s5, external_void_func_v32i32@rel32@hi+12
+; GFX9-NEXT:    s_swappc_b64 s[30:31], s[4:5]
+; GFX9-NEXT:    v_readlane_b32 s4, v40, 0
+; GFX9-NEXT:    v_readlane_b32 s5, v40, 1
+; GFX9-NEXT:    s_sub_u32 s32, s32, 0x400
+; GFX9-NEXT:    v_readlane_b32 s33, v40, 2
+; GFX9-NEXT:    s_or_saveexec_b64 s[6:7], -1
+; GFX9-NEXT:    buffer_load_dword v40, off, s[0:3], s32 ; 4-byte Folded Reload
+; GFX9-NEXT:    s_mov_b64 exec, s[6:7]
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-NEXT:    s_setpc_b64 s[4:5]
+;
+; GFX10-LABEL: test_call_external_void_func_v32i32:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
+; GFX10-NEXT:    s_or_saveexec_b32 s4, -1
+; GFX10-NEXT:    buffer_store_dword v40, off, s[0:3], s32 ; 4-byte Folded Spill
+; GFX10-NEXT:    s_waitcnt_depctr 0xffe3
+; GFX10-NEXT:    s_mov_b32 exec_lo, s4
+; GFX10-NEXT:    s_load_dwordx2 s[4:5], s[4:5], 0x0
+; GFX10-NEXT:    v_writelane_b32 v40, s33, 2
+; GFX10-NEXT:    s_mov_b32 s33, s32
+; GFX10-NEXT:    s_add_u32 s32, s32, 0x200
+; GFX10-NEXT:    ; implicit-def: $vcc_hi
+; GFX10-NEXT:    v_writelane_b32 v40, s30, 0
+; GFX10-NEXT:    v_writelane_b32 v40, s31, 1
+; GFX10-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-NEXT:    v_mov_b32_e32 v29, s5
+; GFX10-NEXT:    v_mov_b32_e32 v28, s4
+; GFX10-NEXT:    s_getpc_b64 s[4:5]
+; GFX10-NEXT:    s_add_u32 s4, s4, external_void_func_v32i32@rel32@lo+4
+; GFX10-NEXT:    s_addc_u32 s5, s5, external_void_func_v32i32@rel32@hi+12
+; GFX10-NEXT:    s_clause 0x7
+; GFX10-NEXT:    global_load_dwordx4 v[0:3], v[28:29], off
+; GFX10-NEXT:    global_load_dwordx4 v[4:7], v[28:29], off offset:16
+; GFX10-NEXT:    global_load_dwordx4 v[8:11], v[28:29], off offset:32
+; GFX10-NEXT:    global_load_dwordx4 v[12:15], v[28:29], off offset:48
+; GFX10-NEXT:    global_load_dwordx4 v[16:19], v[28:29], off offset:64
+; GFX10-NEXT:    global_load_dwordx4 v[20:23], v[28:29], off offset:80
+; GFX10-NEXT:    global_load_dwordx4 v[24:27], v[28:29], off offset:96
+; GFX10-NEXT:    global_load_dwordx4 v[28:31], v[28:29], off offset:112
+; GFX10-NEXT:    s_swappc_b64 s[30:31], s[4:5]
+; GFX10-NEXT:    v_readlane_b32 s4, v40, 0
+; GFX10-NEXT:    v_readlane_b32 s5, v40, 1
+; GFX10-NEXT:    s_sub_u32 s32, s32, 0x200
+; GFX10-NEXT:    v_readlane_b32 s33, v40, 2
+; GFX10-NEXT:    s_or_saveexec_b32 s6, -1
+; GFX10-NEXT:    buffer_load_dword v40, off, s[0:3], s32 ; 4-byte Folded Reload
+; GFX10-NEXT:    s_waitcnt_depctr 0xffe3
+; GFX10-NEXT:    s_mov_b32 exec_lo, s6
+; GFX10-NEXT:    s_waitcnt vmcnt(0)
+; GFX10-NEXT:    s_setpc_b64 s[4:5]
+  %ptr = load <32 x i32> addrspace(1)*, <32 x i32> addrspace(1)* addrspace(4)* undef
+  %val = load <32 x i32>, <32 x i32> addrspace(1)* %ptr
+  call amdgpu_gfx void @external_void_func_v32i32(<32 x i32> %val)
+  ret void
+}
+
+define amdgpu_gfx void @test_call_external_void_func_v32i32_i32(i32) #0 {
+; GFX9-LABEL: test_call_external_void_func_v32i32_i32:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    s_or_saveexec_b64 s[4:5], -1
+; GFX9-NEXT:    buffer_store_dword v40, off, s[0:3], s32 ; 4-byte Folded Spill
+; GFX9-NEXT:    s_mov_b64 exec, s[4:5]
+; GFX9-NEXT:    s_load_dwordx2 s[4:5], s[4:5], 0x0
+; GFX9-NEXT:    v_writelane_b32 v40, s33, 2
+; GFX9-NEXT:    v_writelane_b32 v40, s30, 0
+; GFX9-NEXT:    v_writelane_b32 v40, s31, 1
+; GFX9-NEXT:    s_mov_b32 s33, s32
+; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX9-NEXT:    v_mov_b32_e32 v29, s5
+; GFX9-NEXT:    v_mov_b32_e32 v28, s4
+; GFX9-NEXT:    global_load_dwordx4 v[0:3], v[28:29], off
+; GFX9-NEXT:    global_load_dwordx4 v[4:7], v[28:29], off offset:16
+; GFX9-NEXT:    global_load_dwordx4 v[8:11], v[28:29], off offset:32
+; GFX9-NEXT:    global_load_dwordx4 v[12:15], v[28:29], off offset:48
+; GFX9-NEXT:    global_load_dwordx4 v[16:19], v[28:29], off offset:64
+; GFX9-NEXT:    global_load_dwordx4 v[20:23], v[28:29], off offset:80
+; GFX9-NEXT:    global_load_dwordx4 v[24:27], v[28:29], off offset:96
+; GFX9-NEXT:    global_load_dwordx4 v[28:31], v[28:29], off offset:112
+; GFX9-NEXT:    s_add_u32 s32, s32, 0x400
+; GFX9-NEXT:    s_getpc_b64 s[4:5]
+; GFX9-NEXT:    s_add_u32 s4, s4, external_void_func_v32i32_i32@rel32@lo+4
+; GFX9-NEXT:    s_addc_u32 s5, s5, external_void_func_v32i32_i32@rel32@hi+12
+; GFX9-NEXT:    s_waitcnt vmcnt(7)
+; GFX9-NEXT:    global_load_dword v32, v[0:1], off
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-NEXT:    buffer_store_dword v32, off, s[0:3], s32
+; GFX9-NEXT:    s_swappc_b64 s[30:31], s[4:5]
+; GFX9-NEXT:    v_readlane_b32 s4, v40, 0
+; GFX9-NEXT:    v_readlane_b32 s5, v40, 1
+; GFX9-NEXT:    s_sub_u32 s32, s32, 0x400
+; GFX9-NEXT:    v_readlane_b32 s33, v40, 2
+; GFX9-NEXT:    s_or_saveexec_b64 s[6:7], -1
+; GFX9-NEXT:    buffer_load_dword v40, off, s[0:3], s32 ; 4-byte Folded Reload
+; GFX9-NEXT:    s_mov_b64 exec, s[6:7]
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-NEXT:    s_setpc_b64 s[4:5]
+;
+; GFX10-LABEL: test_call_external_void_func_v32i32_i32:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
+; GFX10-NEXT:    s_or_saveexec_b32 s4, -1
+; GFX10-NEXT:    buffer_store_dword v40, off, s[0:3], s32 ; 4-byte Folded Spill
+; GFX10-NEXT:    s_waitcnt_depctr 0xffe3
+; GFX10-NEXT:    s_mov_b32 exec_lo, s4
+; GFX10-NEXT:    s_load_dwordx2 s[4:5], s[4:5], 0x0
+; GFX10-NEXT:    v_writelane_b32 v40, s33, 2
+; GFX10-NEXT:    s_mov_b32 s33, s32
+; GFX10-NEXT:    s_add_u32 s32, s32, 0x200
+; GFX10-NEXT:    ; implicit-def: $vcc_hi
+; GFX10-NEXT:    v_writelane_b32 v40, s30, 0
+; GFX10-NEXT:    v_writelane_b32 v40, s31, 1
+; GFX10-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-NEXT:    v_mov_b32_e32 v29, s5
+; GFX10-NEXT:    v_mov_b32_e32 v28, s4
+; GFX10-NEXT:    s_getpc_b64 s[4:5]
+; GFX10-NEXT:    s_add_u32 s4, s4, external_void_func_v32i32_i32@rel32@lo+4
+; GFX10-NEXT:    s_addc_u32 s5, s5, external_void_func_v32i32_i32@rel32@hi+12
+; GFX10-NEXT:    global_load_dword v32, v[0:1], off
+; GFX10-NEXT:    s_clause 0x7
+; GFX10-NEXT:    global_load_dwordx4 v[0:3], v[28:29], off
+; GFX10-NEXT:    global_load_dwordx4 v[4:7], v[28:29], off offset:16
+; GFX10-NEXT:    global_load_dwordx4 v[8:11], v[28:29], off offset:32
+; GFX10-NEXT:    global_load_dwordx4 v[12:15], v[28:29], off offset:48
+; GFX10-NEXT:    global_load_dwordx4 v[16:19], v[28:29], off offset:64
+; GFX10-NEXT:    global_load_dwordx4 v[20:23], v[28:29], off offset:80
+; GFX10-NEXT:    global_load_dwordx4 v[24:27], v[28:29], off offset:96
+; GFX10-NEXT:    global_load_dwordx4 v[28:31], v[28:29], off offset:112
+; GFX10-NEXT:    s_waitcnt vmcnt(8)
+; GFX10-NEXT:    buffer_store_dword v32, off, s[0:3], s32
+; GFX10-NEXT:    s_swappc_b64 s[30:31], s[4:5]
+; GFX10-NEXT:    v_readlane_b32 s4, v40, 0
+; GFX10-NEXT:    v_readlane_b32 s5, v40, 1
+; GFX10-NEXT:    s_sub_u32 s32, s32, 0x200
+; GFX10-NEXT:    v_readlane_b32 s33, v40, 2
+; GFX10-NEXT:    s_or_saveexec_b32 s6, -1
+; GFX10-NEXT:    buffer_load_dword v40, off, s[0:3], s32 ; 4-byte Folded Reload
+; GFX10-NEXT:    s_waitcnt_depctr 0xffe3
+; GFX10-NEXT:    s_mov_b32 exec_lo, s6
+; GFX10-NEXT:    s_waitcnt vmcnt(0)
+; GFX10-NEXT:    s_setpc_b64 s[4:5]
+  %ptr0 = load <32 x i32> addrspace(1)*, <32 x i32> addrspace(1)* addrspace(4)* undef
+  %val0 = load <32 x i32>, <32 x i32> addrspace(1)* %ptr0
+  %val1 = load i32, i32 addrspace(1)* undef
+  call amdgpu_gfx void @external_void_func_v32i32_i32(<32 x i32> %val0, i32 %val1)
+  ret void
+}
+
+define amdgpu_gfx void @test_call_external_i32_func_i32_imm(i32 addrspace(1)* %out) #0 {
+; GFX9-LABEL: test_call_external_i32_func_i32_imm:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    s_or_saveexec_b64 s[4:5], -1
+; GFX9-NEXT:    buffer_store_dword v42, off, s[0:3], s32 offset:8 ; 4-byte Folded Spill
+; GFX9-NEXT:    s_mov_b64 exec, s[4:5]
+; GFX9-NEXT:    v_writelane_b32 v42, s33, 2
+; GFX9-NEXT:    s_mov_b32 s33, s32
+; GFX9-NEXT:    s_nop 2
+; GFX9-NEXT:    buffer_store_dword v40, off, s[0:3], s33 offset:4 ; 4-byte Folded Spill
+; GFX9-NEXT:    buffer_store_dword v41, off, s[0:3], s33 ; 4-byte Folded Spill
+; GFX9-NEXT:    v_writelane_b32 v42, s30, 0
+; GFX9-NEXT:    s_add_u32 s32, s32, 0x400
+; GFX9-NEXT:    v_mov_b32_e32 v40, v0
+; GFX9-NEXT:    v_mov_b32_e32 v0, 42
+; GFX9-NEXT:    s_getpc_b64 s[4:5]
+; GFX9-NEXT:    s_add_u32 s4, s4, external_i32_func_i32@rel32@lo+4
+; GFX9-NEXT:    s_addc_u32 s5, s5, external_i32_func_i32@rel32@hi+12
+; GFX9-NEXT:    v_writelane_b32 v42, s31, 1
+; GFX9-NEXT:    v_mov_b32_e32 v41, v1
+; GFX9-NEXT:    s_swappc_b64 s[30:31], s[4:5]
+; GFX9-NEXT:    global_store_dword v[40:41], v0, off
+; GFX9-NEXT:    v_readlane_b32 s4, v42, 0
+; GFX9-NEXT:    v_readlane_b32 s5, v42, 1
+; GFX9-NEXT:    buffer_load_dword v41, off, s[0:3], s33 ; 4-byte Folded Reload
+; GFX9-NEXT:    buffer_load_dword v40, off, s[0:3], s33 offset:4 ; 4-byte Folded Reload
+; GFX9-NEXT:    s_sub_u32 s32, s32, 0x400
+; GFX9-NEXT:    v_readlane_b32 s33, v42, 2
+; GFX9-NEXT:    s_or_saveexec_b64 s[6:7], -1
+; GFX9-NEXT:    buffer_load_dword v42, off, s[0:3], s32 offset:8 ; 4-byte Folded Reload
+; GFX9-NEXT:    s_mov_b64 exec, s[6:7]
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-NEXT:    s_setpc_b64 s[4:5]
+;
+; GFX10-LABEL: test_call_external_i32_func_i32_imm:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
+; GFX10-NEXT:    s_or_saveexec_b32 s4, -1
+; GFX10-NEXT:    buffer_store_dword v42, off, s[0:3], s32 offset:8 ; 4-byte Folded Spill
+; GFX10-NEXT:    s_waitcnt_depctr 0xffe3
+; GFX10-NEXT:    s_mov_b32 exec_lo, s4
+; GFX10-NEXT:    v_writelane_b32 v42, s33, 2
+; GFX10-NEXT:    s_mov_b32 s33, s32
+; GFX10-NEXT:    s_add_u32 s32, s32, 0x200
+; GFX10-NEXT:    s_nop 1
+; GFX10-NEXT:    buffer_store_dword v40, off, s[0:3], s33 offset:4 ; 4-byte Folded Spill
+; GFX10-NEXT:    buffer_store_dword v41, off, s[0:3], s33 ; 4-byte Folded Spill
+; GFX10-NEXT:    v_mov_b32_e32 v40, v0
+; GFX10-NEXT:    v_writelane_b32 v42, s30, 0
+; GFX10-NEXT:    v_mov_b32_e32 v0, 42
+; GFX10-NEXT:    s_getpc_b64 s[4:5]
+; GFX10-NEXT:    s_add_u32 s4, s4, external_i32_func_i32@rel32@lo+4
+; GFX10-NEXT:    s_addc_u32 s5, s5, external_i32_func_i32@rel32@hi+12
+; GFX10-NEXT:    v_mov_b32_e32 v41, v1
+; GFX10-NEXT:    ; implicit-def: $vcc_hi
+; GFX10-NEXT:    v_writelane_b32 v42, s31, 1
+; GFX10-NEXT:    s_swappc_b64 s[30:31], s[4:5]
+; GFX10-NEXT:    global_store_dword v[40:41], v0, off
+; GFX10-NEXT:    v_readlane_b32 s4, v42, 0
+; GFX10-NEXT:    v_readlane_b32 s5, v42, 1
+; GFX10-NEXT:    s_clause 0x1
+; GFX10-NEXT:    buffer_load_dword v41, off, s[0:3], s33
+; GFX10-NEXT:    buffer_load_dword v40, off, s[0:3], s33 offset:4
+; GFX10-NEXT:    s_sub_u32 s32, s32, 0x200
+; GFX10-NEXT:    v_readlane_b32 s33, v42, 2
+; GFX10-NEXT:    s_or_saveexec_b32 s6, -1
+; GFX10-NEXT:    buffer_load_dword v42, off, s[0:3], s32 offset:8 ; 4-byte Folded Reload
+; GFX10-NEXT:    s_waitcnt_depctr 0xffe3
+; GFX10-NEXT:    s_mov_b32 exec_lo, s6
+; GFX10-NEXT:    s_waitcnt vmcnt(0)
+; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
+; GFX10-NEXT:    s_setpc_b64 s[4:5]
+  %val = call amdgpu_gfx i32 @external_i32_func_i32(i32 42)
+  store volatile i32 %val, i32 addrspace(1)* %out
+  ret void
+}
+
+define amdgpu_gfx void @test_call_external_void_func_struct_i8_i32() #0 {
+; GFX9-LABEL: test_call_external_void_func_struct_i8_i32:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    s_or_saveexec_b64 s[4:5], -1
+; GFX9-NEXT:    buffer_store_dword v40, off, s[0:3], s32 ; 4-byte Folded Spill
+; GFX9-NEXT:    s_mov_b64 exec, s[4:5]
+; GFX9-NEXT:    s_load_dwordx2 s[4:5], s[4:5], 0x0
+; GFX9-NEXT:    v_writelane_b32 v40, s33, 2
+; GFX9-NEXT:    v_writelane_b32 v40, s30, 0
+; GFX9-NEXT:    v_writelane_b32 v40, s31, 1
+; GFX9-NEXT:    s_mov_b32 s33, s32
+; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX9-NEXT:    v_mov_b32_e32 v2, s4
+; GFX9-NEXT:    v_mov_b32_e32 v3, s5
+; GFX9-NEXT:    global_load_dword v1, v[2:3], off offset:4
+; GFX9-NEXT:    global_load_ubyte v0, v[2:3], off
+; GFX9-NEXT:    s_add_u32 s32, s32, 0x400
+; GFX9-NEXT:    s_getpc_b64 s[4:5]
+; GFX9-NEXT:    s_add_u32 s4, s4, external_void_func_struct_i8_i32@rel32@lo+4
+; GFX9-NEXT:    s_addc_u32 s5, s5, external_void_func_struct_i8_i32@rel32@hi+12
+; GFX9-NEXT:    s_swappc_b64 s[30:31], s[4:5]
+; GFX9-NEXT:    v_readlane_b32 s4, v40, 0
+; GFX9-NEXT:    v_readlane_b32 s5, v40, 1
+; GFX9-NEXT:    s_sub_u32 s32, s32, 0x400
+; GFX9-NEXT:    v_readlane_b32 s33, v40, 2
+; GFX9-NEXT:    s_or_saveexec_b64 s[6:7], -1
+; GFX9-NEXT:    buffer_load_dword v40, off, s[0:3], s32 ; 4-byte Folded Reload
+; GFX9-NEXT:    s_mov_b64 exec, s[6:7]
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-NEXT:    s_setpc_b64 s[4:5]
+;
+; GFX10-LABEL: test_call_external_void_func_struct_i8_i32:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
+; GFX10-NEXT:    s_or_saveexec_b32 s4, -1
+; GFX10-NEXT:    buffer_store_dword v40, off, s[0:3], s32 ; 4-byte Folded Spill
+; GFX10-NEXT:    s_waitcnt_depctr 0xffe3
+; GFX10-NEXT:    s_mov_b32 exec_lo, s4
+; GFX10-NEXT:    s_load_dwordx2 s[4:5], s[4:5], 0x0
+; GFX10-NEXT:    v_writelane_b32 v40, s33, 2
+; GFX10-NEXT:    s_mov_b32 s33, s32
+; GFX10-NEXT:    s_add_u32 s32, s32, 0x200
+; GFX10-NEXT:    ; implicit-def: $vcc_hi
+; GFX10-NEXT:    v_writelane_b32 v40, s30, 0
+; GFX10-NEXT:    v_writelane_b32 v40, s31, 1
+; GFX10-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-NEXT:    v_mov_b32_e32 v1, s4
+; GFX10-NEXT:    v_mov_b32_e32 v2, s5
+; GFX10-NEXT:    s_getpc_b64 s[4:5]
+; GFX10-NEXT:    s_add_u32 s4, s4, external_void_func_struct_i8_i32@rel32@lo+4
+; GFX10-NEXT:    s_addc_u32 s5, s5, external_void_func_struct_i8_i32@rel32@hi+12
+; GFX10-NEXT:    s_clause 0x1
+; GFX10-NEXT:    global_load_ubyte v0, v[1:2], off
+; GFX10-NEXT:    global_load_dword v1, v[1:2], off offset:4
+; GFX10-NEXT:    s_swappc_b64 s[30:31], s[4:5]
+; GFX10-NEXT:    v_readlane_b32 s4, v40, 0
+; GFX10-NEXT:    v_readlane_b32 s5, v40, 1
+; GFX10-NEXT:    s_sub_u32 s32, s32, 0x200
+; GFX10-NEXT:    v_readlane_b32 s33, v40, 2
+; GFX10-NEXT:    s_or_saveexec_b32 s6, -1
+; GFX10-NEXT:    buffer_load_dword v40, off, s[0:3], s32 ; 4-byte Folded Reload
+; GFX10-NEXT:    s_waitcnt_depctr 0xffe3
+; GFX10-NEXT:    s_mov_b32 exec_lo, s6
+; GFX10-NEXT:    s_waitcnt vmcnt(0)
+; GFX10-NEXT:    s_setpc_b64 s[4:5]
+  %ptr0 = load { i8, i32 } addrspace(1)*, { i8, i32 } addrspace(1)* addrspace(4)* undef
+  %val = load { i8, i32 }, { i8, i32 } addrspace(1)* %ptr0
+  call amdgpu_gfx void @external_void_func_struct_i8_i32({ i8, i32 } %val)
+  ret void
+}
+
+define amdgpu_gfx void @test_call_external_void_func_byval_struct_i8_i32() #0 {
+; GFX9-LABEL: test_call_external_void_func_byval_struct_i8_i32:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    s_or_saveexec_b64 s[4:5], -1
+; GFX9-NEXT:    buffer_store_dword v40, off, s[0:3], s32 offset:8 ; 4-byte Folded Spill
+; GFX9-NEXT:    s_mov_b64 exec, s[4:5]
+; GFX9-NEXT:    v_writelane_b32 v40, s33, 2
+; GFX9-NEXT:    v_writelane_b32 v40, s30, 0
+; GFX9-NEXT:    s_mov_b32 s33, s32
+; GFX9-NEXT:    v_mov_b32_e32 v0, 3
+; GFX9-NEXT:    v_writelane_b32 v40, s31, 1
+; GFX9-NEXT:    buffer_store_byte v0, off, s[0:3], s33
+; GFX9-NEXT:    v_mov_b32_e32 v0, 8
+; GFX9-NEXT:    s_add_u32 s32, s32, 0x400
+; GFX9-NEXT:    buffer_store_dword v0, off, s[0:3], s33 offset:4
+; GFX9-NEXT:    v_lshrrev_b32_e64 v0, 6, s33
+; GFX9-NEXT:    s_getpc_b64 s[4:5]
+; GFX9-NEXT:    s_add_u32 s4, s4, external_void_func_byval_struct_i8_i32@rel32@lo+4
+; GFX9-NEXT:    s_addc_u32 s5, s5, external_void_func_byval_struct_i8_i32@rel32@hi+12
+; GFX9-NEXT:    s_swappc_b64 s[30:31], s[4:5]
+; GFX9-NEXT:    v_readlane_b32 s4, v40, 0
+; GFX9-NEXT:    v_readlane_b32 s5, v40, 1
+; GFX9-NEXT:    s_sub_u32 s32, s32, 0x400
+; GFX9-NEXT:    v_readlane_b32 s33, v40, 2
+; GFX9-NEXT:    s_or_saveexec_b64 s[6:7], -1
+; GFX9-NEXT:    buffer_load_dword v40, off, s[0:3], s32 offset:8 ; 4-byte Folded Reload
+; GFX9-NEXT:    s_mov_b64 exec, s[6:7]
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-NEXT:    s_setpc_b64 s[4:5]
+;
+; GFX10-LABEL: test_call_external_void_func_byval_struct_i8_i32:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
+; GFX10-NEXT:    s_or_saveexec_b32 s4, -1
+; GFX10-NEXT:    buffer_store_dword v40, off, s[0:3], s32 offset:8 ; 4-byte Folded Spill
+; GFX10-NEXT:    s_waitcnt_depctr 0xffe3
+; GFX10-NEXT:    s_mov_b32 exec_lo, s4
+; GFX10-NEXT:    v_writelane_b32 v40, s33, 2
+; GFX10-NEXT:    v_mov_b32_e32 v0, 3
+; GFX10-NEXT:    v_mov_b32_e32 v1, 8
+; GFX10-NEXT:    s_mov_b32 s33, s32
+; GFX10-NEXT:    s_add_u32 s32, s32, 0x200
+; GFX10-NEXT:    v_writelane_b32 v40, s30, 0
+; GFX10-NEXT:    ; implicit-def: $vcc_hi
+; GFX10-NEXT:    s_getpc_b64 s[4:5]
+; GFX10-NEXT:    s_add_u32 s4, s4, external_void_func_byval_struct_i8_i32@rel32@lo+4
+; GFX10-NEXT:    s_addc_u32 s5, s5, external_void_func_byval_struct_i8_i32@rel32@hi+12
+; GFX10-NEXT:    v_writelane_b32 v40, s31, 1
+; GFX10-NEXT:    buffer_store_byte v0, off, s[0:3], s33
+; GFX10-NEXT:    buffer_store_dword v1, off, s[0:3], s33 offset:4
+; GFX10-NEXT:    v_lshrrev_b32_e64 v0, 5, s33
+; GFX10-NEXT:    s_swappc_b64 s[30:31], s[4:5]
+; GFX10-NEXT:    v_readlane_b32 s4, v40, 0
+; GFX10-NEXT:    v_readlane_b32 s5, v40, 1
+; GFX10-NEXT:    s_sub_u32 s32, s32, 0x200
+; GFX10-NEXT:    v_readlane_b32 s33, v40, 2
+; GFX10-NEXT:    s_or_saveexec_b32 s6, -1
+; GFX10-NEXT:    buffer_load_dword v40, off, s[0:3], s32 offset:8 ; 4-byte Folded Reload
+; GFX10-NEXT:    s_waitcnt_depctr 0xffe3
+; GFX10-NEXT:    s_mov_b32 exec_lo, s6
+; GFX10-NEXT:    s_waitcnt vmcnt(0)
+; GFX10-NEXT:    s_setpc_b64 s[4:5]
+  %val = alloca { i8, i32 }, align 4, addrspace(5)
+  %gep0 = getelementptr inbounds { i8, i32 }, { i8, i32 } addrspace(5)* %val, i32 0, i32 0
+  %gep1 = getelementptr inbounds { i8, i32 }, { i8, i32 } addrspace(5)* %val, i32 0, i32 1
+  store i8 3, i8 addrspace(5)* %gep0
+  store i32 8, i32 addrspace(5)* %gep1
+  call amdgpu_gfx void @external_void_func_byval_struct_i8_i32({ i8, i32 } addrspace(5)* %val)
+  ret void
+}
+
+define amdgpu_gfx void @test_call_external_void_func_sret_struct_i8_i32_byval_struct_i8_i32(i32) #0 {
+; GFX9-LABEL: test_call_external_void_func_sret_struct_i8_i32_byval_struct_i8_i32:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    s_or_saveexec_b64 s[4:5], -1
+; GFX9-NEXT:    buffer_store_dword v40, off, s[0:3], s32 offset:16 ; 4-byte Folded Spill
+; GFX9-NEXT:    s_mov_b64 exec, s[4:5]
+; GFX9-NEXT:    v_writelane_b32 v40, s33, 2
+; GFX9-NEXT:    v_writelane_b32 v40, s30, 0
+; GFX9-NEXT:    s_mov_b32 s33, s32
+; GFX9-NEXT:    v_mov_b32_e32 v0, 3
+; GFX9-NEXT:    v_writelane_b32 v40, s31, 1
+; GFX9-NEXT:    buffer_store_byte v0, off, s[0:3], s33
+; GFX9-NEXT:    v_mov_b32_e32 v0, 8
+; GFX9-NEXT:    buffer_store_dword v0, off, s[0:3], s33 offset:4
+; GFX9-NEXT:    v_lshrrev_b32_e64 v0, 6, s33
+; GFX9-NEXT:    s_add_u32 s32, s32, 0x800
+; GFX9-NEXT:    v_add_u32_e32 v0, 8, v0
+; GFX9-NEXT:    v_lshrrev_b32_e64 v1, 6, s33
+; GFX9-NEXT:    s_getpc_b64 s[4:5]
+; GFX9-NEXT:    s_add_u32 s4, s4, external_void_func_sret_struct_i8_i32_byval_struct_i8_i32@rel32@lo+4
+; GFX9-NEXT:    s_addc_u32 s5, s5, external_void_func_sret_struct_i8_i32_byval_struct_i8_i32@rel32@hi+12
+; GFX9-NEXT:    s_swappc_b64 s[30:31], s[4:5]
+; GFX9-NEXT:    buffer_load_ubyte v0, off, s[0:3], s33 offset:8
+; GFX9-NEXT:    buffer_load_dword v1, off, s[0:3], s33 offset:12
+; GFX9-NEXT:    s_sub_u32 s32, s32, 0x800
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-NEXT:    global_store_byte v[0:1], v0, off
+; GFX9-NEXT:    global_store_dword v[0:1], v1, off
+; GFX9-NEXT:    v_readlane_b32 s4, v40, 0
+; GFX9-NEXT:    v_readlane_b32 s5, v40, 1
+; GFX9-NEXT:    v_readlane_b32 s33, v40, 2
+; GFX9-NEXT:    s_or_saveexec_b64 s[6:7], -1
+; GFX9-NEXT:    buffer_load_dword v40, off, s[0:3], s32 offset:16 ; 4-byte Folded Reload
+; GFX9-NEXT:    s_mov_b64 exec, s[6:7]
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-NEXT:    s_setpc_b64 s[4:5]
+;
+; GFX10-LABEL: test_call_external_void_func_sret_struct_i8_i32_byval_struct_i8_i32:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
+; GFX10-NEXT:    s_or_saveexec_b32 s4, -1
+; GFX10-NEXT:    buffer_store_dword v40, off, s[0:3], s32 offset:16 ; 4-byte Folded Spill
+; GFX10-NEXT:    s_waitcnt_depctr 0xffe3
+; GFX10-NEXT:    s_mov_b32 exec_lo, s4
+; GFX10-NEXT:    v_writelane_b32 v40, s33, 2
+; GFX10-NEXT:    v_mov_b32_e32 v0, 3
+; GFX10-NEXT:    v_mov_b32_e32 v1, 8
+; GFX10-NEXT:    s_mov_b32 s33, s32
+; GFX10-NEXT:    s_add_u32 s32, s32, 0x400
+; GFX10-NEXT:    v_writelane_b32 v40, s30, 0
+; GFX10-NEXT:    ; implicit-def: $vcc_hi
+; GFX10-NEXT:    s_getpc_b64 s[4:5]
+; GFX10-NEXT:    s_add_u32 s4, s4, external_void_func_sret_struct_i8_i32_byval_struct_i8_i32@rel32@lo+4
+; GFX10-NEXT:    s_addc_u32 s5, s5, external_void_func_sret_struct_i8_i32_byval_struct_i8_i32@rel32@hi+12
+; GFX10-NEXT:    v_writelane_b32 v40, s31, 1
+; GFX10-NEXT:    buffer_store_byte v0, off, s[0:3], s33
+; GFX10-NEXT:    buffer_store_dword v1, off, s[0:3], s33 offset:4
+; GFX10-NEXT:    v_lshrrev_b32_e64 v0, 5, s33
+; GFX10-NEXT:    v_lshrrev_b32_e64 v1, 5, s33
+; GFX10-NEXT:    v_add_nc_u32_e32 v0, 8, v0
+; GFX10-NEXT:    s_swappc_b64 s[30:31], s[4:5]
+; GFX10-NEXT:    s_clause 0x1
+; GFX10-NEXT:    buffer_load_ubyte v0, off, s[0:3], s33 offset:8
+; GFX10-NEXT:    buffer_load_dword v1, off, s[0:3], s33 offset:12
+; GFX10-NEXT:    s_sub_u32 s32, s32, 0x400
+; GFX10-NEXT:    s_waitcnt vmcnt(0)
+; GFX10-NEXT:    global_store_byte v[0:1], v0, off
+; GFX10-NEXT:    global_store_dword v[0:1], v1, off
+; GFX10-NEXT:    v_readlane_b32 s4, v40, 0
+; GFX10-NEXT:    v_readlane_b32 s5, v40, 1
+; GFX10-NEXT:    v_readlane_b32 s33, v40, 2
+; GFX10-NEXT:    s_or_saveexec_b32 s6, -1
+; GFX10-NEXT:    buffer_load_dword v40, off, s[0:3], s32 offset:16 ; 4-byte Folded Reload
+; GFX10-NEXT:    s_waitcnt_depctr 0xffe3
+; GFX10-NEXT:    s_mov_b32 exec_lo, s6
+; GFX10-NEXT:    s_waitcnt vmcnt(0)
+; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
+; GFX10-NEXT:    s_setpc_b64 s[4:5]
+  %in.val = alloca { i8, i32 }, align 4, addrspace(5)
+  %out.val = alloca { i8, i32 }, align 4, addrspace(5)
+  %in.gep0 = getelementptr inbounds { i8, i32 }, { i8, i32 } addrspace(5)* %in.val, i32 0, i32 0
+  %in.gep1 = getelementptr inbounds { i8, i32 }, { i8, i32 } addrspace(5)* %in.val, i32 0, i32 1
+  store i8 3, i8 addrspace(5)* %in.gep0
+  store i32 8, i32 addrspace(5)* %in.gep1
+  call amdgpu_gfx void @external_void_func_sret_struct_i8_i32_byval_struct_i8_i32({ i8, i32 } addrspace(5)* %out.val, { i8, i32 } addrspace(5)* %in.val)
+  %out.gep0 = getelementptr inbounds { i8, i32 }, { i8, i32 } addrspace(5)* %out.val, i32 0, i32 0
+  %out.gep1 = getelementptr inbounds { i8, i32 }, { i8, i32 } addrspace(5)* %out.val, i32 0, i32 1
+  %out.val0 = load i8, i8 addrspace(5)* %out.gep0
+  %out.val1 = load i32, i32 addrspace(5)* %out.gep1
+
+  store volatile i8 %out.val0, i8 addrspace(1)* undef
+  store volatile i32 %out.val1, i32 addrspace(1)* undef
+  ret void
+}
+
+define amdgpu_gfx void @test_call_external_void_func_v16i8() #0 {
+; GFX9-LABEL: test_call_external_void_func_v16i8:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    s_or_saveexec_b64 s[4:5], -1
+; GFX9-NEXT:    buffer_store_dword v40, off, s[0:3], s32 ; 4-byte Folded Spill
+; GFX9-NEXT:    s_mov_b64 exec, s[4:5]
+; GFX9-NEXT:    s_load_dwordx2 s[4:5], s[4:5], 0x0
+; GFX9-NEXT:    v_writelane_b32 v40, s33, 2
+; GFX9-NEXT:    v_writelane_b32 v40, s30, 0
+; GFX9-NEXT:    v_writelane_b32 v40, s31, 1
+; GFX9-NEXT:    s_mov_b32 s33, s32
+; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX9-NEXT:    v_mov_b32_e32 v0, s4
+; GFX9-NEXT:    v_mov_b32_e32 v1, s5
+; GFX9-NEXT:    global_load_dwordx4 v[0:3], v[0:1], off
+; GFX9-NEXT:    s_add_u32 s32, s32, 0x400
+; GFX9-NEXT:    s_getpc_b64 s[4:5]
+; GFX9-NEXT:    s_add_u32 s4, s4, external_void_func_v16i8@rel32@lo+4
+; GFX9-NEXT:    s_addc_u32 s5, s5, external_void_func_v16i8@rel32@hi+12
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-NEXT:    v_lshrrev_b32_e32 v16, 8, v0
+; GFX9-NEXT:    v_lshrrev_b32_e32 v17, 16, v0
+; GFX9-NEXT:    v_lshrrev_b32_e32 v18, 24, v0
+; GFX9-NEXT:    v_lshrrev_b32_e32 v5, 8, v1
+; GFX9-NEXT:    v_lshrrev_b32_e32 v6, 16, v1
+; GFX9-NEXT:    v_lshrrev_b32_e32 v7, 24, v1
+; GFX9-NEXT:    v_mov_b32_e32 v4, v1
+; GFX9-NEXT:    v_lshrrev_b32_e32 v9, 8, v2
+; GFX9-NEXT:    v_lshrrev_b32_e32 v10, 16, v2
+; GFX9-NEXT:    v_lshrrev_b32_e32 v11, 24, v2
+; GFX9-NEXT:    v_mov_b32_e32 v8, v2
+; GFX9-NEXT:    v_lshrrev_b32_e32 v13, 8, v3
+; GFX9-NEXT:    v_lshrrev_b32_e32 v14, 16, v3
+; GFX9-NEXT:    v_lshrrev_b32_e32 v15, 24, v3
+; GFX9-NEXT:    v_mov_b32_e32 v12, v3
+; GFX9-NEXT:    v_mov_b32_e32 v1, v16
+; GFX9-NEXT:    v_mov_b32_e32 v2, v17
+; GFX9-NEXT:    v_mov_b32_e32 v3, v18
+; GFX9-NEXT:    s_swappc_b64 s[30:31], s[4:5]
+; GFX9-NEXT:    v_readlane_b32 s4, v40, 0
+; GFX9-NEXT:    v_readlane_b32 s5, v40, 1
+; GFX9-NEXT:    s_sub_u32 s32, s32, 0x400
+; GFX9-NEXT:    v_readlane_b32 s33, v40, 2
+; GFX9-NEXT:    s_or_saveexec_b64 s[6:7], -1
+; GFX9-NEXT:    buffer_load_dword v40, off, s[0:3], s32 ; 4-byte Folded Reload
+; GFX9-NEXT:    s_mov_b64 exec, s[6:7]
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-NEXT:    s_setpc_b64 s[4:5]
+;
+; GFX10-LABEL: test_call_external_void_func_v16i8:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
+; GFX10-NEXT:    s_or_saveexec_b32 s4, -1
+; GFX10-NEXT:    buffer_store_dword v40, off, s[0:3], s32 ; 4-byte Folded Spill
+; GFX10-NEXT:    s_waitcnt_depctr 0xffe3
+; GFX10-NEXT:    s_mov_b32 exec_lo, s4
+; GFX10-NEXT:    s_load_dwordx2 s[4:5], s[4:5], 0x0
+; GFX10-NEXT:    v_writelane_b32 v40, s33, 2
+; GFX10-NEXT:    s_mov_b32 s33, s32
+; GFX10-NEXT:    s_add_u32 s32, s32, 0x200
+; GFX10-NEXT:    ; implicit-def: $vcc_hi
+; GFX10-NEXT:    v_writelane_b32 v40, s30, 0
+; GFX10-NEXT:    v_writelane_b32 v40, s31, 1
+; GFX10-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-NEXT:    v_mov_b32_e32 v0, s4
+; GFX10-NEXT:    v_mov_b32_e32 v1, s5
+; GFX10-NEXT:    s_getpc_b64 s[4:5]
+; GFX10-NEXT:    s_add_u32 s4, s4, external_void_func_v16i8@rel32@lo+4
+; GFX10-NEXT:    s_addc_u32 s5, s5, external_void_func_v16i8@rel32@hi+12
+; GFX10-NEXT:    global_load_dwordx4 v[0:3], v[0:1], off
+; GFX10-NEXT:    s_waitcnt vmcnt(0)
+; GFX10-NEXT:    v_lshrrev_b32_e32 v16, 8, v0
+; GFX10-NEXT:    v_lshrrev_b32_e32 v17, 16, v0
+; GFX10-NEXT:    v_lshrrev_b32_e32 v18, 24, v0
+; GFX10-NEXT:    v_lshrrev_b32_e32 v5, 8, v1
+; GFX10-NEXT:    v_lshrrev_b32_e32 v6, 16, v1
+; GFX10-NEXT:    v_lshrrev_b32_e32 v7, 24, v1
+; GFX10-NEXT:    v_mov_b32_e32 v4, v1
+; GFX10-NEXT:    v_lshrrev_b32_e32 v9, 8, v2
+; GFX10-NEXT:    v_lshrrev_b32_e32 v10, 16, v2
+; GFX10-NEXT:    v_lshrrev_b32_e32 v11, 24, v2
+; GFX10-NEXT:    v_mov_b32_e32 v8, v2
+; GFX10-NEXT:    v_lshrrev_b32_e32 v13, 8, v3
+; GFX10-NEXT:    v_lshrrev_b32_e32 v14, 16, v3
+; GFX10-NEXT:    v_lshrrev_b32_e32 v15, 24, v3
+; GFX10-NEXT:    v_mov_b32_e32 v12, v3
+; GFX10-NEXT:    v_mov_b32_e32 v1, v16
+; GFX10-NEXT:    v_mov_b32_e32 v2, v17
+; GFX10-NEXT:    v_mov_b32_e32 v3, v18
+; GFX10-NEXT:    s_swappc_b64 s[30:31], s[4:5]
+; GFX10-NEXT:    v_readlane_b32 s4, v40, 0
+; GFX10-NEXT:    v_readlane_b32 s5, v40, 1
+; GFX10-NEXT:    s_sub_u32 s32, s32, 0x200
+; GFX10-NEXT:    v_readlane_b32 s33, v40, 2
+; GFX10-NEXT:    s_or_saveexec_b32 s6, -1
+; GFX10-NEXT:    buffer_load_dword v40, off, s[0:3], s32 ; 4-byte Folded Reload
+; GFX10-NEXT:    s_waitcnt_depctr 0xffe3
+; GFX10-NEXT:    s_mov_b32 exec_lo, s6
+; GFX10-NEXT:    s_waitcnt vmcnt(0)
+; GFX10-NEXT:    s_setpc_b64 s[4:5]
+  %ptr = load <16 x i8> addrspace(1)*, <16 x i8> addrspace(1)* addrspace(4)* undef
+  %val = load <16 x i8>, <16 x i8> addrspace(1)* %ptr
+  call amdgpu_gfx void @external_void_func_v16i8(<16 x i8> %val)
+  ret void
+}
+
+define void @tail_call_byval_align16(<32 x i32> %val, double %tmp) #0 {
+; GFX9-LABEL: tail_call_byval_align16:
+; GFX9:       ; %bb.0: ; %entry
+; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    s_or_saveexec_b64 s[4:5], -1
+; GFX9-NEXT:    buffer_store_dword v40, off, s[0:3], s32 offset:16 ; 4-byte Folded Spill
+; GFX9-NEXT:    s_mov_b64 exec, s[4:5]
+; GFX9-NEXT:    v_writelane_b32 v40, s33, 2
+; GFX9-NEXT:    v_writelane_b32 v40, s30, 0
+; GFX9-NEXT:    s_mov_b32 s33, s32
+; GFX9-NEXT:    v_writelane_b32 v40, s31, 1
+; GFX9-NEXT:    s_nop 0
+; GFX9-NEXT:    buffer_load_dword v32, off, s[0:3], s33 offset:12
+; GFX9-NEXT:    buffer_load_dword v33, off, s[0:3], s33 offset:8
+; GFX9-NEXT:    s_add_u32 s32, s32, 0x800
+; GFX9-NEXT:    s_getpc_b64 s[4:5]
+; GFX9-NEXT:    s_add_u32 s4, s4, byval_align16_f64_arg@rel32@lo+4
+; GFX9-NEXT:    s_addc_u32 s5, s5, byval_align16_f64_arg@rel32@hi+12
+; GFX9-NEXT:    s_waitcnt vmcnt(1)
+; GFX9-NEXT:    buffer_store_dword v32, off, s[0:3], s32 offset:4
+; GFX9-NEXT:    s_waitcnt vmcnt(1)
+; GFX9-NEXT:    buffer_store_dword v33, off, s[0:3], s32
+; GFX9-NEXT:    s_swappc_b64 s[30:31], s[4:5]
+; GFX9-NEXT:    v_readlane_b32 s4, v40, 0
+; GFX9-NEXT:    v_readlane_b32 s5, v40, 1
+; GFX9-NEXT:    s_sub_u32 s32, s32, 0x800
+; GFX9-NEXT:    v_readlane_b32 s33, v40, 2
+; GFX9-NEXT:    s_or_saveexec_b64 s[6:7], -1
+; GFX9-NEXT:    buffer_load_dword v40, off, s[0:3], s32 offset:16 ; 4-byte Folded Reload
+; GFX9-NEXT:    s_mov_b64 exec, s[6:7]
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-NEXT:    s_setpc_b64 s[4:5]
+;
+; GFX10-LABEL: tail_call_byval_align16:
+; GFX10:       ; %bb.0: ; %entry
+; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
+; GFX10-NEXT:    s_or_saveexec_b32 s4, -1
+; GFX10-NEXT:    buffer_store_dword v40, off, s[0:3], s32 offset:16 ; 4-byte Folded Spill
+; GFX10-NEXT:    s_waitcnt_depctr 0xffe3
+; GFX10-NEXT:    s_mov_b32 exec_lo, s4
+; GFX10-NEXT:    v_writelane_b32 v40, s33, 2
+; GFX10-NEXT:    s_mov_b32 s33, s32
+; GFX10-NEXT:    s_add_u32 s32, s32, 0x400
+; GFX10-NEXT:    s_getpc_b64 s[4:5]
+; GFX10-NEXT:    s_add_u32 s4, s4, byval_align16_f64_arg@rel32@lo+4
+; GFX10-NEXT:    s_addc_u32 s5, s5, byval_align16_f64_arg@rel32@hi+12
+; GFX10-NEXT:    ; implicit-def: $vcc_hi
+; GFX10-NEXT:    v_writelane_b32 v40, s30, 0
+; GFX10-NEXT:    v_writelane_b32 v40, s31, 1
+; GFX10-NEXT:    s_clause 0x1
+; GFX10-NEXT:    buffer_load_dword v32, off, s[0:3], s33 offset:12
+; GFX10-NEXT:    buffer_load_dword v33, off, s[0:3], s33 offset:8
+; GFX10-NEXT:    s_waitcnt vmcnt(1)
+; GFX10-NEXT:    buffer_store_dword v32, off, s[0:3], s32 offset:4
+; GFX10-NEXT:    s_waitcnt vmcnt(0)
+; GFX10-NEXT:    buffer_store_dword v33, off, s[0:3], s32
+; GFX10-NEXT:    s_swappc_b64 s[30:31], s[4:5]
+; GFX10-NEXT:    v_readlane_b32 s4, v40, 0
+; GFX10-NEXT:    v_readlane_b32 s5, v40, 1
+; GFX10-NEXT:    s_sub_u32 s32, s32, 0x400
+; GFX10-NEXT:    v_readlane_b32 s33, v40, 2
+; GFX10-NEXT:    s_or_saveexec_b32 s6, -1
+; GFX10-NEXT:    buffer_load_dword v40, off, s[0:3], s32 offset:16 ; 4-byte Folded Reload
+; GFX10-NEXT:    s_waitcnt_depctr 0xffe3
+; GFX10-NEXT:    s_mov_b32 exec_lo, s6
+; GFX10-NEXT:    s_waitcnt vmcnt(0)
+; GFX10-NEXT:    s_setpc_b64 s[4:5]
+entry:
+  %alloca = alloca double, align 8, addrspace(5)
+  tail call amdgpu_gfx void @byval_align16_f64_arg(<32 x i32> %val, double addrspace(5)* byval align 16 %alloca)
+  ret void
+}
+
+; inreg arguments are put in sgprs
+define amdgpu_gfx void @test_call_external_void_func_i1_imm_inreg() #0 {
+; GFX9-LABEL: test_call_external_void_func_i1_imm_inreg:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    s_or_saveexec_b64 s[4:5], -1
+; GFX9-NEXT:    buffer_store_dword v40, off, s[0:3], s32 ; 4-byte Folded Spill
+; GFX9-NEXT:    s_mov_b64 exec, s[4:5]
+; GFX9-NEXT:    v_writelane_b32 v40, s33, 2
+; GFX9-NEXT:    v_writelane_b32 v40, s30, 0
+; GFX9-NEXT:    s_mov_b32 s33, s32
+; GFX9-NEXT:    s_add_u32 s32, s32, 0x400
+; GFX9-NEXT:    v_mov_b32_e32 v0, 1
+; GFX9-NEXT:    v_writelane_b32 v40, s31, 1
+; GFX9-NEXT:    s_getpc_b64 s[4:5]
+; GFX9-NEXT:    s_add_u32 s4, s4, external_void_func_i1_inreg@rel32@lo+4
+; GFX9-NEXT:    s_addc_u32 s5, s5, external_void_func_i1_inreg@rel32@hi+12
+; GFX9-NEXT:    buffer_store_byte v0, off, s[0:3], s32
+; GFX9-NEXT:    s_swappc_b64 s[30:31], s[4:5]
+; GFX9-NEXT:    v_readlane_b32 s4, v40, 0
+; GFX9-NEXT:    v_readlane_b32 s5, v40, 1
+; GFX9-NEXT:    s_sub_u32 s32, s32, 0x400
+; GFX9-NEXT:    v_readlane_b32 s33, v40, 2
+; GFX9-NEXT:    s_or_saveexec_b64 s[6:7], -1
+; GFX9-NEXT:    buffer_load_dword v40, off, s[0:3], s32 ; 4-byte Folded Reload
+; GFX9-NEXT:    s_mov_b64 exec, s[6:7]
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-NEXT:    s_setpc_b64 s[4:5]
+;
+; GFX10-LABEL: test_call_external_void_func_i1_imm_inreg:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
+; GFX10-NEXT:    s_or_saveexec_b32 s4, -1
+; GFX10-NEXT:    buffer_store_dword v40, off, s[0:3], s32 ; 4-byte Folded Spill
+; GFX10-NEXT:    s_waitcnt_depctr 0xffe3
+; GFX10-NEXT:    s_mov_b32 exec_lo, s4
+; GFX10-NEXT:    v_writelane_b32 v40, s33, 2
+; GFX10-NEXT:    v_mov_b32_e32 v0, 1
+; GFX10-NEXT:    s_mov_b32 s33, s32
+; GFX10-NEXT:    s_add_u32 s32, s32, 0x200
+; GFX10-NEXT:    s_getpc_b64 s[4:5]
+; GFX10-NEXT:    s_add_u32 s4, s4, external_void_func_i1_inreg@rel32@lo+4
+; GFX10-NEXT:    s_addc_u32 s5, s5, external_void_func_i1_inreg@rel32@hi+12
+; GFX10-NEXT:    v_writelane_b32 v40, s30, 0
+; GFX10-NEXT:    ; implicit-def: $vcc_hi
+; GFX10-NEXT:    v_writelane_b32 v40, s31, 1
+; GFX10-NEXT:    buffer_store_byte v0, off, s[0:3], s32
+; GFX10-NEXT:    s_swappc_b64 s[30:31], s[4:5]
+; GFX10-NEXT:    v_readlane_b32 s4, v40, 0
+; GFX10-NEXT:    v_readlane_b32 s5, v40, 1
+; GFX10-NEXT:    s_sub_u32 s32, s32, 0x200
+; GFX10-NEXT:    v_readlane_b32 s33, v40, 2
+; GFX10-NEXT:    s_or_saveexec_b32 s6, -1
+; GFX10-NEXT:    buffer_load_dword v40, off, s[0:3], s32 ; 4-byte Folded Reload
+; GFX10-NEXT:    s_waitcnt_depctr 0xffe3
+; GFX10-NEXT:    s_mov_b32 exec_lo, s6
+; GFX10-NEXT:    s_waitcnt vmcnt(0)
+; GFX10-NEXT:    s_setpc_b64 s[4:5]
+  call amdgpu_gfx void @external_void_func_i1_inreg(i1 true)
+  ret void
+}
+
+define amdgpu_gfx void @test_call_external_void_func_i8_imm_inreg(i32) #0 {
+; GFX9-LABEL: test_call_external_void_func_i8_imm_inreg:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    s_or_saveexec_b64 s[4:5], -1
+; GFX9-NEXT:    buffer_store_dword v40, off, s[0:3], s32 ; 4-byte Folded Spill
+; GFX9-NEXT:    s_mov_b64 exec, s[4:5]
+; GFX9-NEXT:    v_writelane_b32 v40, s33, 2
+; GFX9-NEXT:    v_writelane_b32 v40, s30, 0
+; GFX9-NEXT:    s_mov_b32 s33, s32
+; GFX9-NEXT:    s_add_u32 s32, s32, 0x400
+; GFX9-NEXT:    s_movk_i32 s4, 0x7b
+; GFX9-NEXT:    s_getpc_b64 s[6:7]
+; GFX9-NEXT:    s_add_u32 s6, s6, external_void_func_i8_inreg@rel32@lo+4
+; GFX9-NEXT:    s_addc_u32 s7, s7, external_void_func_i8_inreg@rel32@hi+12
+; GFX9-NEXT:    v_writelane_b32 v40, s31, 1
+; GFX9-NEXT:    s_swappc_b64 s[30:31], s[6:7]
+; GFX9-NEXT:    v_readlane_b32 s4, v40, 0
+; GFX9-NEXT:    v_readlane_b32 s5, v40, 1
+; GFX9-NEXT:    s_sub_u32 s32, s32, 0x400
+; GFX9-NEXT:    v_readlane_b32 s33, v40, 2
+; GFX9-NEXT:    s_or_saveexec_b64 s[6:7], -1
+; GFX9-NEXT:    buffer_load_dword v40, off, s[0:3], s32 ; 4-byte Folded Reload
+; GFX9-NEXT:    s_mov_b64 exec, s[6:7]
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-NEXT:    s_setpc_b64 s[4:5]
+;
+; GFX10-LABEL: test_call_external_void_func_i8_imm_inreg:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
+; GFX10-NEXT:    s_or_saveexec_b32 s4, -1
+; GFX10-NEXT:    buffer_store_dword v40, off, s[0:3], s32 ; 4-byte Folded Spill
+; GFX10-NEXT:    s_waitcnt_depctr 0xffe3
+; GFX10-NEXT:    s_mov_b32 exec_lo, s4
+; GFX10-NEXT:    v_writelane_b32 v40, s33, 2
+; GFX10-NEXT:    s_movk_i32 s4, 0x7b
+; GFX10-NEXT:    s_mov_b32 s33, s32
+; GFX10-NEXT:    s_add_u32 s32, s32, 0x200
+; GFX10-NEXT:    s_getpc_b64 s[6:7]
+; GFX10-NEXT:    s_add_u32 s6, s6, external_void_func_i8_inreg@rel32@lo+4
+; GFX10-NEXT:    s_addc_u32 s7, s7, external_void_func_i8_inreg@rel32@hi+12
+; GFX10-NEXT:    v_writelane_b32 v40, s30, 0
+; GFX10-NEXT:    ; implicit-def: $vcc_hi
+; GFX10-NEXT:    v_writelane_b32 v40, s31, 1
+; GFX10-NEXT:    s_swappc_b64 s[30:31], s[6:7]
+; GFX10-NEXT:    v_readlane_b32 s4, v40, 0
+; GFX10-NEXT:    v_readlane_b32 s5, v40, 1
+; GFX10-NEXT:    s_sub_u32 s32, s32, 0x200
+; GFX10-NEXT:    v_readlane_b32 s33, v40, 2
+; GFX10-NEXT:    s_or_saveexec_b32 s6, -1
+; GFX10-NEXT:    buffer_load_dword v40, off, s[0:3], s32 ; 4-byte Folded Reload
+; GFX10-NEXT:    s_waitcnt_depctr 0xffe3
+; GFX10-NEXT:    s_mov_b32 exec_lo, s6
+; GFX10-NEXT:    s_waitcnt vmcnt(0)
+; GFX10-NEXT:    s_setpc_b64 s[4:5]
+  call amdgpu_gfx void @external_void_func_i8_inreg(i8 123)
+  ret void
+}
+
+define amdgpu_gfx void @test_call_external_void_func_i16_imm_inreg() #0 {
+; GFX9-LABEL: test_call_external_void_func_i16_imm_inreg:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    s_or_saveexec_b64 s[4:5], -1
+; GFX9-NEXT:    buffer_store_dword v40, off, s[0:3], s32 ; 4-byte Folded Spill
+; GFX9-NEXT:    s_mov_b64 exec, s[4:5]
+; GFX9-NEXT:    v_writelane_b32 v40, s33, 2
+; GFX9-NEXT:    v_writelane_b32 v40, s30, 0
+; GFX9-NEXT:    s_mov_b32 s33, s32
+; GFX9-NEXT:    s_add_u32 s32, s32, 0x400
+; GFX9-NEXT:    s_movk_i32 s4, 0x7b
+; GFX9-NEXT:    s_getpc_b64 s[6:7]
+; GFX9-NEXT:    s_add_u32 s6, s6, external_void_func_i16_inreg@rel32@lo+4
+; GFX9-NEXT:    s_addc_u32 s7, s7, external_void_func_i16_inreg@rel32@hi+12
+; GFX9-NEXT:    v_writelane_b32 v40, s31, 1
+; GFX9-NEXT:    s_swappc_b64 s[30:31], s[6:7]
+; GFX9-NEXT:    v_readlane_b32 s4, v40, 0
+; GFX9-NEXT:    v_readlane_b32 s5, v40, 1
+; GFX9-NEXT:    s_sub_u32 s32, s32, 0x400
+; GFX9-NEXT:    v_readlane_b32 s33, v40, 2
+; GFX9-NEXT:    s_or_saveexec_b64 s[6:7], -1
+; GFX9-NEXT:    buffer_load_dword v40, off, s[0:3], s32 ; 4-byte Folded Reload
+; GFX9-NEXT:    s_mov_b64 exec, s[6:7]
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-NEXT:    s_setpc_b64 s[4:5]
+;
+; GFX10-LABEL: test_call_external_void_func_i16_imm_inreg:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
+; GFX10-NEXT:    s_or_saveexec_b32 s4, -1
+; GFX10-NEXT:    buffer_store_dword v40, off, s[0:3], s32 ; 4-byte Folded Spill
+; GFX10-NEXT:    s_waitcnt_depctr 0xffe3
+; GFX10-NEXT:    s_mov_b32 exec_lo, s4
+; GFX10-NEXT:    v_writelane_b32 v40, s33, 2
+; GFX10-NEXT:    s_movk_i32 s4, 0x7b
+; GFX10-NEXT:    s_mov_b32 s33, s32
+; GFX10-NEXT:    s_add_u32 s32, s32, 0x200
+; GFX10-NEXT:    s_getpc_b64 s[6:7]
+; GFX10-NEXT:    s_add_u32 s6, s6, external_void_func_i16_inreg@rel32@lo+4
+; GFX10-NEXT:    s_addc_u32 s7, s7, external_void_func_i16_inreg@rel32@hi+12
+; GFX10-NEXT:    v_writelane_b32 v40, s30, 0
+; GFX10-NEXT:    ; implicit-def: $vcc_hi
+; GFX10-NEXT:    v_writelane_b32 v40, s31, 1
+; GFX10-NEXT:    s_swappc_b64 s[30:31], s[6:7]
+; GFX10-NEXT:    v_readlane_b32 s4, v40, 0
+; GFX10-NEXT:    v_readlane_b32 s5, v40, 1
+; GFX10-NEXT:    s_sub_u32 s32, s32, 0x200
+; GFX10-NEXT:    v_readlane_b32 s33, v40, 2
+; GFX10-NEXT:    s_or_saveexec_b32 s6, -1
+; GFX10-NEXT:    buffer_load_dword v40, off, s[0:3], s32 ; 4-byte Folded Reload
+; GFX10-NEXT:    s_waitcnt_depctr 0xffe3
+; GFX10-NEXT:    s_mov_b32 exec_lo, s6
+; GFX10-NEXT:    s_waitcnt vmcnt(0)
+; GFX10-NEXT:    s_setpc_b64 s[4:5]
+  call amdgpu_gfx void @external_void_func_i16_inreg(i16 123)
+  ret void
+}
+
+define amdgpu_gfx void @test_call_external_void_func_i32_imm_inreg(i32) #0 {
+; GFX9-LABEL: test_call_external_void_func_i32_imm_inreg:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    s_or_saveexec_b64 s[4:5], -1
+; GFX9-NEXT:    buffer_store_dword v40, off, s[0:3], s32 ; 4-byte Folded Spill
+; GFX9-NEXT:    s_mov_b64 exec, s[4:5]
+; GFX9-NEXT:    v_writelane_b32 v40, s33, 2
+; GFX9-NEXT:    v_writelane_b32 v40, s30, 0
+; GFX9-NEXT:    s_mov_b32 s33, s32
+; GFX9-NEXT:    s_add_u32 s32, s32, 0x400
+; GFX9-NEXT:    s_mov_b32 s4, 42
+; GFX9-NEXT:    s_getpc_b64 s[6:7]
+; GFX9-NEXT:    s_add_u32 s6, s6, external_void_func_i32_inreg@rel32@lo+4
+; GFX9-NEXT:    s_addc_u32 s7, s7, external_void_func_i32_inreg@rel32@hi+12
+; GFX9-NEXT:    v_writelane_b32 v40, s31, 1
+; GFX9-NEXT:    s_swappc_b64 s[30:31], s[6:7]
+; GFX9-NEXT:    v_readlane_b32 s4, v40, 0
+; GFX9-NEXT:    v_readlane_b32 s5, v40, 1
+; GFX9-NEXT:    s_sub_u32 s32, s32, 0x400
+; GFX9-NEXT:    v_readlane_b32 s33, v40, 2
+; GFX9-NEXT:    s_or_saveexec_b64 s[6:7], -1
+; GFX9-NEXT:    buffer_load_dword v40, off, s[0:3], s32 ; 4-byte Folded Reload
+; GFX9-NEXT:    s_mov_b64 exec, s[6:7]
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-NEXT:    s_setpc_b64 s[4:5]
+;
+; GFX10-LABEL: test_call_external_void_func_i32_imm_inreg:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
+; GFX10-NEXT:    s_or_saveexec_b32 s4, -1
+; GFX10-NEXT:    buffer_store_dword v40, off, s[0:3], s32 ; 4-byte Folded Spill
+; GFX10-NEXT:    s_waitcnt_depctr 0xffe3
+; GFX10-NEXT:    s_mov_b32 exec_lo, s4
+; GFX10-NEXT:    v_writelane_b32 v40, s33, 2
+; GFX10-NEXT:    s_mov_b32 s4, 42
+; GFX10-NEXT:    s_mov_b32 s33, s32
+; GFX10-NEXT:    s_add_u32 s32, s32, 0x200
+; GFX10-NEXT:    s_getpc_b64 s[6:7]
+; GFX10-NEXT:    s_add_u32 s6, s6, external_void_func_i32_inreg@rel32@lo+4
+; GFX10-NEXT:    s_addc_u32 s7, s7, external_void_func_i32_inreg@rel32@hi+12
+; GFX10-NEXT:    v_writelane_b32 v40, s30, 0
+; GFX10-NEXT:    ; implicit-def: $vcc_hi
+; GFX10-NEXT:    v_writelane_b32 v40, s31, 1
+; GFX10-NEXT:    s_swappc_b64 s[30:31], s[6:7]
+; GFX10-NEXT:    v_readlane_b32 s4, v40, 0
+; GFX10-NEXT:    v_readlane_b32 s5, v40, 1
+; GFX10-NEXT:    s_sub_u32 s32, s32, 0x200
+; GFX10-NEXT:    v_readlane_b32 s33, v40, 2
+; GFX10-NEXT:    s_or_saveexec_b32 s6, -1
+; GFX10-NEXT:    buffer_load_dword v40, off, s[0:3], s32 ; 4-byte Folded Reload
+; GFX10-NEXT:    s_waitcnt_depctr 0xffe3
+; GFX10-NEXT:    s_mov_b32 exec_lo, s6
+; GFX10-NEXT:    s_waitcnt vmcnt(0)
+; GFX10-NEXT:    s_setpc_b64 s[4:5]
+  call amdgpu_gfx void @external_void_func_i32_inreg(i32 42)
+  ret void
+}
+
+define amdgpu_gfx void @test_call_external_void_func_i64_imm_inreg() #0 {
+; GFX9-LABEL: test_call_external_void_func_i64_imm_inreg:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    s_or_saveexec_b64 s[4:5], -1
+; GFX9-NEXT:    buffer_store_dword v40, off, s[0:3], s32 ; 4-byte Folded Spill
+; GFX9-NEXT:    s_mov_b64 exec, s[4:5]
+; GFX9-NEXT:    v_writelane_b32 v40, s33, 2
+; GFX9-NEXT:    v_writelane_b32 v40, s30, 0
+; GFX9-NEXT:    s_mov_b32 s33, s32
+; GFX9-NEXT:    s_add_u32 s32, s32, 0x400
+; GFX9-NEXT:    s_movk_i32 s4, 0x7b
+; GFX9-NEXT:    s_mov_b32 s5, 0
+; GFX9-NEXT:    s_getpc_b64 s[6:7]
+; GFX9-NEXT:    s_add_u32 s6, s6, external_void_func_i64_inreg@rel32@lo+4
+; GFX9-NEXT:    s_addc_u32 s7, s7, external_void_func_i64_inreg@rel32@hi+12
+; GFX9-NEXT:    v_writelane_b32 v40, s31, 1
+; GFX9-NEXT:    s_swappc_b64 s[30:31], s[6:7]
+; GFX9-NEXT:    v_readlane_b32 s4, v40, 0
+; GFX9-NEXT:    v_readlane_b32 s5, v40, 1
+; GFX9-NEXT:    s_sub_u32 s32, s32, 0x400
+; GFX9-NEXT:    v_readlane_b32 s33, v40, 2
+; GFX9-NEXT:    s_or_saveexec_b64 s[6:7], -1
+; GFX9-NEXT:    buffer_load_dword v40, off, s[0:3], s32 ; 4-byte Folded Reload
+; GFX9-NEXT:    s_mov_b64 exec, s[6:7]
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-NEXT:    s_setpc_b64 s[4:5]
+;
+; GFX10-LABEL: test_call_external_void_func_i64_imm_inreg:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
+; GFX10-NEXT:    s_or_saveexec_b32 s4, -1
+; GFX10-NEXT:    buffer_store_dword v40, off, s[0:3], s32 ; 4-byte Folded Spill
+; GFX10-NEXT:    s_waitcnt_depctr 0xffe3
+; GFX10-NEXT:    s_mov_b32 exec_lo, s4
+; GFX10-NEXT:    v_writelane_b32 v40, s33, 2
+; GFX10-NEXT:    s_movk_i32 s4, 0x7b
+; GFX10-NEXT:    s_mov_b32 s5, 0
+; GFX10-NEXT:    s_mov_b32 s33, s32
+; GFX10-NEXT:    s_add_u32 s32, s32, 0x200
+; GFX10-NEXT:    v_writelane_b32 v40, s30, 0
+; GFX10-NEXT:    s_getpc_b64 s[6:7]
+; GFX10-NEXT:    s_add_u32 s6, s6, external_void_func_i64_inreg@rel32@lo+4
+; GFX10-NEXT:    s_addc_u32 s7, s7, external_void_func_i64_inreg@rel32@hi+12
+; GFX10-NEXT:    ; implicit-def: $vcc_hi
+; GFX10-NEXT:    v_writelane_b32 v40, s31, 1
+; GFX10-NEXT:    s_swappc_b64 s[30:31], s[6:7]
+; GFX10-NEXT:    v_readlane_b32 s4, v40, 0
+; GFX10-NEXT:    v_readlane_b32 s5, v40, 1
+; GFX10-NEXT:    s_sub_u32 s32, s32, 0x200
+; GFX10-NEXT:    v_readlane_b32 s33, v40, 2
+; GFX10-NEXT:    s_or_saveexec_b32 s6, -1
+; GFX10-NEXT:    buffer_load_dword v40, off, s[0:3], s32 ; 4-byte Folded Reload
+; GFX10-NEXT:    s_waitcnt_depctr 0xffe3
+; GFX10-NEXT:    s_mov_b32 exec_lo, s6
+; GFX10-NEXT:    s_waitcnt vmcnt(0)
+; GFX10-NEXT:    s_setpc_b64 s[4:5]
+  call amdgpu_gfx void @external_void_func_i64_inreg(i64 123)
+  ret void
+}
+
+define amdgpu_gfx void @test_call_external_void_func_v2i64_inreg() #0 {
+; GFX9-LABEL: test_call_external_void_func_v2i64_inreg:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    s_or_saveexec_b64 s[4:5], -1
+; GFX9-NEXT:    buffer_store_dword v40, off, s[0:3], s32 ; 4-byte Folded Spill
+; GFX9-NEXT:    s_mov_b64 exec, s[4:5]
+; GFX9-NEXT:    s_mov_b64 s[4:5], 0
+; GFX9-NEXT:    s_load_dwordx4 s[4:7], s[4:5], 0x0
+; GFX9-NEXT:    v_writelane_b32 v40, s33, 2
+; GFX9-NEXT:    v_writelane_b32 v40, s30, 0
+; GFX9-NEXT:    s_mov_b32 s33, s32
+; GFX9-NEXT:    s_add_u32 s32, s32, 0x400
+; GFX9-NEXT:    s_getpc_b64 s[8:9]
+; GFX9-NEXT:    s_add_u32 s8, s8, external_void_func_v2i64_inreg@rel32@lo+4
+; GFX9-NEXT:    s_addc_u32 s9, s9, external_void_func_v2i64_inreg@rel32@hi+12
+; GFX9-NEXT:    v_writelane_b32 v40, s31, 1
+; GFX9-NEXT:    s_swappc_b64 s[30:31], s[8:9]
+; GFX9-NEXT:    v_readlane_b32 s4, v40, 0
+; GFX9-NEXT:    v_readlane_b32 s5, v40, 1
+; GFX9-NEXT:    s_sub_u32 s32, s32, 0x400
+; GFX9-NEXT:    v_readlane_b32 s33, v40, 2
+; GFX9-NEXT:    s_or_saveexec_b64 s[6:7], -1
+; GFX9-NEXT:    buffer_load_dword v40, off, s[0:3], s32 ; 4-byte Folded Reload
+; GFX9-NEXT:    s_mov_b64 exec, s[6:7]
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-NEXT:    s_setpc_b64 s[4:5]
+;
+; GFX10-LABEL: test_call_external_void_func_v2i64_inreg:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
+; GFX10-NEXT:    s_or_saveexec_b32 s4, -1
+; GFX10-NEXT:    buffer_store_dword v40, off, s[0:3], s32 ; 4-byte Folded Spill
+; GFX10-NEXT:    s_waitcnt_depctr 0xffe3
+; GFX10-NEXT:    s_mov_b32 exec_lo, s4
+; GFX10-NEXT:    s_mov_b64 s[4:5], 0
+; GFX10-NEXT:    v_writelane_b32 v40, s33, 2
+; GFX10-NEXT:    s_load_dwordx4 s[4:7], s[4:5], 0x0
+; GFX10-NEXT:    s_mov_b32 s33, s32
+; GFX10-NEXT:    s_add_u32 s32, s32, 0x200
+; GFX10-NEXT:    s_getpc_b64 s[8:9]
+; GFX10-NEXT:    s_add_u32 s8, s8, external_void_func_v2i64_inreg@rel32@lo+4
+; GFX10-NEXT:    s_addc_u32 s9, s9, external_void_func_v2i64_inreg@rel32@hi+12
+; GFX10-NEXT:    v_writelane_b32 v40, s30, 0
+; GFX10-NEXT:    ; implicit-def: $vcc_hi
+; GFX10-NEXT:    v_writelane_b32 v40, s31, 1
+; GFX10-NEXT:    s_swappc_b64 s[30:31], s[8:9]
+; GFX10-NEXT:    v_readlane_b32 s4, v40, 0
+; GFX10-NEXT:    v_readlane_b32 s5, v40, 1
+; GFX10-NEXT:    s_sub_u32 s32, s32, 0x200
+; GFX10-NEXT:    v_readlane_b32 s33, v40, 2
+; GFX10-NEXT:    s_or_saveexec_b32 s6, -1
+; GFX10-NEXT:    buffer_load_dword v40, off, s[0:3], s32 ; 4-byte Folded Reload
+; GFX10-NEXT:    s_waitcnt_depctr 0xffe3
+; GFX10-NEXT:    s_mov_b32 exec_lo, s6
+; GFX10-NEXT:    s_waitcnt vmcnt(0)
+; GFX10-NEXT:    s_setpc_b64 s[4:5]
+  %val = load <2 x i64>, <2 x i64> addrspace(4)* null
+  call amdgpu_gfx void @external_void_func_v2i64_inreg(<2 x i64> %val)
+  ret void
+}
+
+define amdgpu_gfx void @test_call_external_void_func_v2i64_imm_inreg() #0 {
+; GFX9-LABEL: test_call_external_void_func_v2i64_imm_inreg:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    s_or_saveexec_b64 s[4:5], -1
+; GFX9-NEXT:    buffer_store_dword v40, off, s[0:3], s32 ; 4-byte Folded Spill
+; GFX9-NEXT:    s_mov_b64 exec, s[4:5]
+; GFX9-NEXT:    v_writelane_b32 v40, s33, 2
+; GFX9-NEXT:    v_writelane_b32 v40, s30, 0
+; GFX9-NEXT:    s_mov_b32 s33, s32
+; GFX9-NEXT:    s_add_u32 s32, s32, 0x400
+; GFX9-NEXT:    s_mov_b32 s4, 1
+; GFX9-NEXT:    s_mov_b32 s5, 2
+; GFX9-NEXT:    s_mov_b32 s6, 3
+; GFX9-NEXT:    s_mov_b32 s7, 4
+; GFX9-NEXT:    s_getpc_b64 s[8:9]
+; GFX9-NEXT:    s_add_u32 s8, s8, external_void_func_v2i64_inreg@rel32@lo+4
+; GFX9-NEXT:    s_addc_u32 s9, s9, external_void_func_v2i64_inreg@rel32@hi+12
+; GFX9-NEXT:    v_writelane_b32 v40, s31, 1
+; GFX9-NEXT:    s_swappc_b64 s[30:31], s[8:9]
+; GFX9-NEXT:    v_readlane_b32 s4, v40, 0
+; GFX9-NEXT:    v_readlane_b32 s5, v40, 1
+; GFX9-NEXT:    s_sub_u32 s32, s32, 0x400
+; GFX9-NEXT:    v_readlane_b32 s33, v40, 2
+; GFX9-NEXT:    s_or_saveexec_b64 s[6:7], -1
+; GFX9-NEXT:    buffer_load_dword v40, off, s[0:3], s32 ; 4-byte Folded Reload
+; GFX9-NEXT:    s_mov_b64 exec, s[6:7]
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-NEXT:    s_setpc_b64 s[4:5]
+;
+; GFX10-LABEL: test_call_external_void_func_v2i64_imm_inreg:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
+; GFX10-NEXT:    s_or_saveexec_b32 s4, -1
+; GFX10-NEXT:    buffer_store_dword v40, off, s[0:3], s32 ; 4-byte Folded Spill
+; GFX10-NEXT:    s_waitcnt_depctr 0xffe3
+; GFX10-NEXT:    s_mov_b32 exec_lo, s4
+; GFX10-NEXT:    v_writelane_b32 v40, s33, 2
+; GFX10-NEXT:    s_mov_b32 s4, 1
+; GFX10-NEXT:    s_mov_b32 s5, 2
+; GFX10-NEXT:    s_mov_b32 s6, 3
+; GFX10-NEXT:    s_mov_b32 s7, 4
+; GFX10-NEXT:    v_writelane_b32 v40, s30, 0
+; GFX10-NEXT:    s_mov_b32 s33, s32
+; GFX10-NEXT:    s_add_u32 s32, s32, 0x200
+; GFX10-NEXT:    s_getpc_b64 s[8:9]
+; GFX10-NEXT:    s_add_u32 s8, s8, external_void_func_v2i64_inreg@rel32@lo+4
+; GFX10-NEXT:    s_addc_u32 s9, s9, external_void_func_v2i64_inreg@rel32@hi+12
+; GFX10-NEXT:    ; implicit-def: $vcc_hi
+; GFX10-NEXT:    v_writelane_b32 v40, s31, 1
+; GFX10-NEXT:    s_swappc_b64 s[30:31], s[8:9]
+; GFX10-NEXT:    v_readlane_b32 s4, v40, 0
+; GFX10-NEXT:    v_readlane_b32 s5, v40, 1
+; GFX10-NEXT:    s_sub_u32 s32, s32, 0x200
+; GFX10-NEXT:    v_readlane_b32 s33, v40, 2
+; GFX10-NEXT:    s_or_saveexec_b32 s6, -1
+; GFX10-NEXT:    buffer_load_dword v40, off, s[0:3], s32 ; 4-byte Folded Reload
+; GFX10-NEXT:    s_waitcnt_depctr 0xffe3
+; GFX10-NEXT:    s_mov_b32 exec_lo, s6
+; GFX10-NEXT:    s_waitcnt vmcnt(0)
+; GFX10-NEXT:    s_setpc_b64 s[4:5]
+  call amdgpu_gfx void @external_void_func_v2i64_inreg(<2 x i64> <i64 8589934593, i64 17179869187>)
+  ret void
+}
+
+define amdgpu_gfx void @test_call_external_void_func_v3i64_inreg() #0 {
+; GFX9-LABEL: test_call_external_void_func_v3i64_inreg:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    s_or_saveexec_b64 s[4:5], -1
+; GFX9-NEXT:    buffer_store_dword v40, off, s[0:3], s32 ; 4-byte Folded Spill
+; GFX9-NEXT:    s_mov_b64 exec, s[4:5]
+; GFX9-NEXT:    s_mov_b64 s[4:5], 0
+; GFX9-NEXT:    s_load_dwordx4 s[4:7], s[4:5], 0x0
+; GFX9-NEXT:    v_writelane_b32 v40, s33, 2
+; GFX9-NEXT:    v_writelane_b32 v40, s30, 0
+; GFX9-NEXT:    s_mov_b32 s33, s32
+; GFX9-NEXT:    s_add_u32 s32, s32, 0x400
+; GFX9-NEXT:    s_mov_b32 s8, 1
+; GFX9-NEXT:    s_mov_b32 s9, 2
+; GFX9-NEXT:    s_getpc_b64 s[10:11]
+; GFX9-NEXT:    s_add_u32 s10, s10, external_void_func_v3i64_inreg@rel32@lo+4
+; GFX9-NEXT:    s_addc_u32 s11, s11, external_void_func_v3i64_inreg@rel32@hi+12
+; GFX9-NEXT:    v_writelane_b32 v40, s31, 1
+; GFX9-NEXT:    s_swappc_b64 s[30:31], s[10:11]
+; GFX9-NEXT:    v_readlane_b32 s4, v40, 0
+; GFX9-NEXT:    v_readlane_b32 s5, v40, 1
+; GFX9-NEXT:    s_sub_u32 s32, s32, 0x400
+; GFX9-NEXT:    v_readlane_b32 s33, v40, 2
+; GFX9-NEXT:    s_or_saveexec_b64 s[6:7], -1
+; GFX9-NEXT:    buffer_load_dword v40, off, s[0:3], s32 ; 4-byte Folded Reload
+; GFX9-NEXT:    s_mov_b64 exec, s[6:7]
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-NEXT:    s_setpc_b64 s[4:5]
+;
+; GFX10-LABEL: test_call_external_void_func_v3i64_inreg:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
+; GFX10-NEXT:    s_or_saveexec_b32 s4, -1
+; GFX10-NEXT:    buffer_store_dword v40, off, s[0:3], s32 ; 4-byte Folded Spill
+; GFX10-NEXT:    s_waitcnt_depctr 0xffe3
+; GFX10-NEXT:    s_mov_b32 exec_lo, s4
+; GFX10-NEXT:    s_mov_b64 s[4:5], 0
+; GFX10-NEXT:    v_writelane_b32 v40, s33, 2
+; GFX10-NEXT:    s_load_dwordx4 s[4:7], s[4:5], 0x0
+; GFX10-NEXT:    s_mov_b32 s8, 1
+; GFX10-NEXT:    s_mov_b32 s9, 2
+; GFX10-NEXT:    s_mov_b32 s33, s32
+; GFX10-NEXT:    v_writelane_b32 v40, s30, 0
+; GFX10-NEXT:    s_add_u32 s32, s32, 0x200
+; GFX10-NEXT:    s_getpc_b64 s[10:11]
+; GFX10-NEXT:    s_add_u32 s10, s10, external_void_func_v3i64_inreg@rel32@lo+4
+; GFX10-NEXT:    s_addc_u32 s11, s11, external_void_func_v3i64_inreg@rel32@hi+12
+; GFX10-NEXT:    ; implicit-def: $vcc_hi
+; GFX10-NEXT:    v_writelane_b32 v40, s31, 1
+; GFX10-NEXT:    s_swappc_b64 s[30:31], s[10:11]
+; GFX10-NEXT:    v_readlane_b32 s4, v40, 0
+; GFX10-NEXT:    v_readlane_b32 s5, v40, 1
+; GFX10-NEXT:    s_sub_u32 s32, s32, 0x200
+; GFX10-NEXT:    v_readlane_b32 s33, v40, 2
+; GFX10-NEXT:    s_or_saveexec_b32 s6, -1
+; GFX10-NEXT:    buffer_load_dword v40, off, s[0:3], s32 ; 4-byte Folded Reload
+; GFX10-NEXT:    s_waitcnt_depctr 0xffe3
+; GFX10-NEXT:    s_mov_b32 exec_lo, s6
+; GFX10-NEXT:    s_waitcnt vmcnt(0)
+; GFX10-NEXT:    s_setpc_b64 s[4:5]
+  %load = load <2 x i64>, <2 x i64> addrspace(4)* null
+  %val = shufflevector <2 x i64> %load, <2 x i64> <i64 8589934593, i64 undef>, <3 x i32> <i32 0, i32 1, i32 2>
+
+  call amdgpu_gfx void @external_void_func_v3i64_inreg(<3 x i64> %val)
+  ret void
+}
+
+define amdgpu_gfx void @test_call_external_void_func_v4i64_inreg() #0 {
+; GFX9-LABEL: test_call_external_void_func_v4i64_inreg:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    s_or_saveexec_b64 s[4:5], -1
+; GFX9-NEXT:    buffer_store_dword v40, off, s[0:3], s32 ; 4-byte Folded Spill
+; GFX9-NEXT:    s_mov_b64 exec, s[4:5]
+; GFX9-NEXT:    s_mov_b64 s[4:5], 0
+; GFX9-NEXT:    s_load_dwordx4 s[4:7], s[4:5], 0x0
+; GFX9-NEXT:    v_writelane_b32 v40, s33, 2
+; GFX9-NEXT:    v_writelane_b32 v40, s30, 0
+; GFX9-NEXT:    s_mov_b32 s33, s32
+; GFX9-NEXT:    s_add_u32 s32, s32, 0x400
+; GFX9-NEXT:    s_mov_b32 s8, 1
+; GFX9-NEXT:    s_mov_b32 s9, 2
+; GFX9-NEXT:    s_mov_b32 s10, 3
+; GFX9-NEXT:    s_mov_b32 s11, 4
+; GFX9-NEXT:    s_getpc_b64 s[12:13]
+; GFX9-NEXT:    s_add_u32 s12, s12, external_void_func_v4i64_inreg@rel32@lo+4
+; GFX9-NEXT:    s_addc_u32 s13, s13, external_void_func_v4i64_inreg@rel32@hi+12
+; GFX9-NEXT:    v_writelane_b32 v40, s31, 1
+; GFX9-NEXT:    s_swappc_b64 s[30:31], s[12:13]
+; GFX9-NEXT:    v_readlane_b32 s4, v40, 0
+; GFX9-NEXT:    v_readlane_b32 s5, v40, 1
+; GFX9-NEXT:    s_sub_u32 s32, s32, 0x400
+; GFX9-NEXT:    v_readlane_b32 s33, v40, 2
+; GFX9-NEXT:    s_or_saveexec_b64 s[6:7], -1
+; GFX9-NEXT:    buffer_load_dword v40, off, s[0:3], s32 ; 4-byte Folded Reload
+; GFX9-NEXT:    s_mov_b64 exec, s[6:7]
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-NEXT:    s_setpc_b64 s[4:5]
+;
+; GFX10-LABEL: test_call_external_void_func_v4i64_inreg:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
+; GFX10-NEXT:    s_or_saveexec_b32 s4, -1
+; GFX10-NEXT:    buffer_store_dword v40, off, s[0:3], s32 ; 4-byte Folded Spill
+; GFX10-NEXT:    s_waitcnt_depctr 0xffe3
+; GFX10-NEXT:    s_mov_b32 exec_lo, s4
+; GFX10-NEXT:    s_mov_b64 s[4:5], 0
+; GFX10-NEXT:    v_writelane_b32 v40, s33, 2
+; GFX10-NEXT:    s_load_dwordx4 s[4:7], s[4:5], 0x0
+; GFX10-NEXT:    s_mov_b32 s8, 1
+; GFX10-NEXT:    s_mov_b32 s9, 2
+; GFX10-NEXT:    s_mov_b32 s10, 3
+; GFX10-NEXT:    v_writelane_b32 v40, s30, 0
+; GFX10-NEXT:    s_mov_b32 s11, 4
+; GFX10-NEXT:    s_mov_b32 s33, s32
+; GFX10-NEXT:    s_add_u32 s32, s32, 0x200
+; GFX10-NEXT:    s_getpc_b64 s[12:13]
+; GFX10-NEXT:    s_add_u32 s12, s12, external_void_func_v4i64_inreg@rel32@lo+4
+; GFX10-NEXT:    s_addc_u32 s13, s13, external_void_func_v4i64_inreg@rel32@hi+12
+; GFX10-NEXT:    v_writelane_b32 v40, s31, 1
+; GFX10-NEXT:    ; implicit-def: $vcc_hi
+; GFX10-NEXT:    s_swappc_b64 s[30:31], s[12:13]
+; GFX10-NEXT:    v_readlane_b32 s4, v40, 0
+; GFX10-NEXT:    v_readlane_b32 s5, v40, 1
+; GFX10-NEXT:    s_sub_u32 s32, s32, 0x200
+; GFX10-NEXT:    v_readlane_b32 s33, v40, 2
+; GFX10-NEXT:    s_or_saveexec_b32 s6, -1
+; GFX10-NEXT:    buffer_load_dword v40, off, s[0:3], s32 ; 4-byte Folded Reload
+; GFX10-NEXT:    s_waitcnt_depctr 0xffe3
+; GFX10-NEXT:    s_mov_b32 exec_lo, s6
+; GFX10-NEXT:    s_waitcnt vmcnt(0)
+; GFX10-NEXT:    s_setpc_b64 s[4:5]
+  %load = load <2 x i64>, <2 x i64> addrspace(4)* null
+  %val = shufflevector <2 x i64> %load, <2 x i64> <i64 8589934593, i64 17179869187>, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+  call amdgpu_gfx void @external_void_func_v4i64_inreg(<4 x i64> %val)
+  ret void
+}
+
+define amdgpu_gfx void @test_call_external_void_func_f16_imm_inreg() #0 {
+; GFX9-LABEL: test_call_external_void_func_f16_imm_inreg:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    s_or_saveexec_b64 s[4:5], -1
+; GFX9-NEXT:    buffer_store_dword v40, off, s[0:3], s32 ; 4-byte Folded Spill
+; GFX9-NEXT:    s_mov_b64 exec, s[4:5]
+; GFX9-NEXT:    v_writelane_b32 v40, s33, 2
+; GFX9-NEXT:    v_writelane_b32 v40, s30, 0
+; GFX9-NEXT:    s_mov_b32 s33, s32
+; GFX9-NEXT:    s_add_u32 s32, s32, 0x400
+; GFX9-NEXT:    s_movk_i32 s4, 0x4400
+; GFX9-NEXT:    s_getpc_b64 s[6:7]
+; GFX9-NEXT:    s_add_u32 s6, s6, external_void_func_f16_inreg@rel32@lo+4
+; GFX9-NEXT:    s_addc_u32 s7, s7, external_void_func_f16_inreg@rel32@hi+12
+; GFX9-NEXT:    v_writelane_b32 v40, s31, 1
+; GFX9-NEXT:    s_swappc_b64 s[30:31], s[6:7]
+; GFX9-NEXT:    v_readlane_b32 s4, v40, 0
+; GFX9-NEXT:    v_readlane_b32 s5, v40, 1
+; GFX9-NEXT:    s_sub_u32 s32, s32, 0x400
+; GFX9-NEXT:    v_readlane_b32 s33, v40, 2
+; GFX9-NEXT:    s_or_saveexec_b64 s[6:7], -1
+; GFX9-NEXT:    buffer_load_dword v40, off, s[0:3], s32 ; 4-byte Folded Reload
+; GFX9-NEXT:    s_mov_b64 exec, s[6:7]
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-NEXT:    s_setpc_b64 s[4:5]
+;
+; GFX10-LABEL: test_call_external_void_func_f16_imm_inreg:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
+; GFX10-NEXT:    s_or_saveexec_b32 s4, -1
+; GFX10-NEXT:    buffer_store_dword v40, off, s[0:3], s32 ; 4-byte Folded Spill
+; GFX10-NEXT:    s_waitcnt_depctr 0xffe3
+; GFX10-NEXT:    s_mov_b32 exec_lo, s4
+; GFX10-NEXT:    v_writelane_b32 v40, s33, 2
+; GFX10-NEXT:    s_movk_i32 s4, 0x4400
+; GFX10-NEXT:    s_mov_b32 s33, s32
+; GFX10-NEXT:    s_add_u32 s32, s32, 0x200
+; GFX10-NEXT:    s_getpc_b64 s[6:7]
+; GFX10-NEXT:    s_add_u32 s6, s6, external_void_func_f16_inreg@rel32@lo+4
+; GFX10-NEXT:    s_addc_u32 s7, s7, external_void_func_f16_inreg@rel32@hi+12
+; GFX10-NEXT:    v_writelane_b32 v40, s30, 0
+; GFX10-NEXT:    ; implicit-def: $vcc_hi
+; GFX10-NEXT:    v_writelane_b32 v40, s31, 1
+; GFX10-NEXT:    s_swappc_b64 s[30:31], s[6:7]
+; GFX10-NEXT:    v_readlane_b32 s4, v40, 0
+; GFX10-NEXT:    v_readlane_b32 s5, v40, 1
+; GFX10-NEXT:    s_sub_u32 s32, s32, 0x200
+; GFX10-NEXT:    v_readlane_b32 s33, v40, 2
+; GFX10-NEXT:    s_or_saveexec_b32 s6, -1
+; GFX10-NEXT:    buffer_load_dword v40, off, s[0:3], s32 ; 4-byte Folded Reload
+; GFX10-NEXT:    s_waitcnt_depctr 0xffe3
+; GFX10-NEXT:    s_mov_b32 exec_lo, s6
+; GFX10-NEXT:    s_waitcnt vmcnt(0)
+; GFX10-NEXT:    s_setpc_b64 s[4:5]
+  call amdgpu_gfx void @external_void_func_f16_inreg(half 4.0)
+  ret void
+}
+
+define amdgpu_gfx void @test_call_external_void_func_f32_imm_inreg() #0 {
+; GFX9-LABEL: test_call_external_void_func_f32_imm_inreg:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    s_or_saveexec_b64 s[4:5], -1
+; GFX9-NEXT:    buffer_store_dword v40, off, s[0:3], s32 ; 4-byte Folded Spill
+; GFX9-NEXT:    s_mov_b64 exec, s[4:5]
+; GFX9-NEXT:    v_writelane_b32 v40, s33, 2
+; GFX9-NEXT:    v_writelane_b32 v40, s30, 0
+; GFX9-NEXT:    s_mov_b32 s33, s32
+; GFX9-NEXT:    s_add_u32 s32, s32, 0x400
+; GFX9-NEXT:    s_mov_b32 s4, 4.0
+; GFX9-NEXT:    s_getpc_b64 s[6:7]
+; GFX9-NEXT:    s_add_u32 s6, s6, external_void_func_f32_inreg@rel32@lo+4
+; GFX9-NEXT:    s_addc_u32 s7, s7, external_void_func_f32_inreg@rel32@hi+12
+; GFX9-NEXT:    v_writelane_b32 v40, s31, 1
+; GFX9-NEXT:    s_swappc_b64 s[30:31], s[6:7]
+; GFX9-NEXT:    v_readlane_b32 s4, v40, 0
+; GFX9-NEXT:    v_readlane_b32 s5, v40, 1
+; GFX9-NEXT:    s_sub_u32 s32, s32, 0x400
+; GFX9-NEXT:    v_readlane_b32 s33, v40, 2
+; GFX9-NEXT:    s_or_saveexec_b64 s[6:7], -1
+; GFX9-NEXT:    buffer_load_dword v40, off, s[0:3], s32 ; 4-byte Folded Reload
+; GFX9-NEXT:    s_mov_b64 exec, s[6:7]
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-NEXT:    s_setpc_b64 s[4:5]
+;
+; GFX10-LABEL: test_call_external_void_func_f32_imm_inreg:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
+; GFX10-NEXT:    s_or_saveexec_b32 s4, -1
+; GFX10-NEXT:    buffer_store_dword v40, off, s[0:3], s32 ; 4-byte Folded Spill
+; GFX10-NEXT:    s_waitcnt_depctr 0xffe3
+; GFX10-NEXT:    s_mov_b32 exec_lo, s4
+; GFX10-NEXT:    v_writelane_b32 v40, s33, 2
+; GFX10-NEXT:    s_mov_b32 s4, 4.0
+; GFX10-NEXT:    s_mov_b32 s33, s32
+; GFX10-NEXT:    s_add_u32 s32, s32, 0x200
+; GFX10-NEXT:    s_getpc_b64 s[6:7]
+; GFX10-NEXT:    s_add_u32 s6, s6, external_void_func_f32_inreg@rel32@lo+4
+; GFX10-NEXT:    s_addc_u32 s7, s7, external_void_func_f32_inreg@rel32@hi+12
+; GFX10-NEXT:    v_writelane_b32 v40, s30, 0
+; GFX10-NEXT:    ; implicit-def: $vcc_hi
+; GFX10-NEXT:    v_writelane_b32 v40, s31, 1
+; GFX10-NEXT:    s_swappc_b64 s[30:31], s[6:7]
+; GFX10-NEXT:    v_readlane_b32 s4, v40, 0
+; GFX10-NEXT:    v_readlane_b32 s5, v40, 1
+; GFX10-NEXT:    s_sub_u32 s32, s32, 0x200
+; GFX10-NEXT:    v_readlane_b32 s33, v40, 2
+; GFX10-NEXT:    s_or_saveexec_b32 s6, -1
+; GFX10-NEXT:    buffer_load_dword v40, off, s[0:3], s32 ; 4-byte Folded Reload
+; GFX10-NEXT:    s_waitcnt_depctr 0xffe3
+; GFX10-NEXT:    s_mov_b32 exec_lo, s6
+; GFX10-NEXT:    s_waitcnt vmcnt(0)
+; GFX10-NEXT:    s_setpc_b64 s[4:5]
+  call amdgpu_gfx void @external_void_func_f32_inreg(float 4.0)
+  ret void
+}
+
+define amdgpu_gfx void @test_call_external_void_func_v2f32_imm_inreg() #0 {
+; GFX9-LABEL: test_call_external_void_func_v2f32_imm_inreg:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    s_or_saveexec_b64 s[4:5], -1
+; GFX9-NEXT:    buffer_store_dword v40, off, s[0:3], s32 ; 4-byte Folded Spill
+; GFX9-NEXT:    s_mov_b64 exec, s[4:5]
+; GFX9-NEXT:    v_writelane_b32 v40, s33, 2
+; GFX9-NEXT:    v_writelane_b32 v40, s30, 0
+; GFX9-NEXT:    s_mov_b32 s33, s32
+; GFX9-NEXT:    s_add_u32 s32, s32, 0x400
+; GFX9-NEXT:    s_mov_b32 s4, 1.0
+; GFX9-NEXT:    s_mov_b32 s5, 2.0
+; GFX9-NEXT:    s_getpc_b64 s[6:7]
+; GFX9-NEXT:    s_add_u32 s6, s6, external_void_func_v2f32_inreg@rel32@lo+4
+; GFX9-NEXT:    s_addc_u32 s7, s7, external_void_func_v2f32_inreg@rel32@hi+12
+; GFX9-NEXT:    v_writelane_b32 v40, s31, 1
+; GFX9-NEXT:    s_swappc_b64 s[30:31], s[6:7]
+; GFX9-NEXT:    v_readlane_b32 s4, v40, 0
+; GFX9-NEXT:    v_readlane_b32 s5, v40, 1
+; GFX9-NEXT:    s_sub_u32 s32, s32, 0x400
+; GFX9-NEXT:    v_readlane_b32 s33, v40, 2
+; GFX9-NEXT:    s_or_saveexec_b64 s[6:7], -1
+; GFX9-NEXT:    buffer_load_dword v40, off, s[0:3], s32 ; 4-byte Folded Reload
+; GFX9-NEXT:    s_mov_b64 exec, s[6:7]
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-NEXT:    s_setpc_b64 s[4:5]
+;
+; GFX10-LABEL: test_call_external_void_func_v2f32_imm_inreg:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
+; GFX10-NEXT:    s_or_saveexec_b32 s4, -1
+; GFX10-NEXT:    buffer_store_dword v40, off, s[0:3], s32 ; 4-byte Folded Spill
+; GFX10-NEXT:    s_waitcnt_depctr 0xffe3
+; GFX10-NEXT:    s_mov_b32 exec_lo, s4
+; GFX10-NEXT:    v_writelane_b32 v40, s33, 2
+; GFX10-NEXT:    s_mov_b32 s4, 1.0
+; GFX10-NEXT:    s_mov_b32 s5, 2.0
+; GFX10-NEXT:    s_mov_b32 s33, s32
+; GFX10-NEXT:    s_add_u32 s32, s32, 0x200
+; GFX10-NEXT:    v_writelane_b32 v40, s30, 0
+; GFX10-NEXT:    s_getpc_b64 s[6:7]
+; GFX10-NEXT:    s_add_u32 s6, s6, external_void_func_v2f32_inreg@rel32@lo+4
+; GFX10-NEXT:    s_addc_u32 s7, s7, external_void_func_v2f32_inreg@rel32@hi+12
+; GFX10-NEXT:    ; implicit-def: $vcc_hi
+; GFX10-NEXT:    v_writelane_b32 v40, s31, 1
+; GFX10-NEXT:    s_swappc_b64 s[30:31], s[6:7]
+; GFX10-NEXT:    v_readlane_b32 s4, v40, 0
+; GFX10-NEXT:    v_readlane_b32 s5, v40, 1
+; GFX10-NEXT:    s_sub_u32 s32, s32, 0x200
+; GFX10-NEXT:    v_readlane_b32 s33, v40, 2
+; GFX10-NEXT:    s_or_saveexec_b32 s6, -1
+; GFX10-NEXT:    buffer_load_dword v40, off, s[0:3], s32 ; 4-byte Folded Reload
+; GFX10-NEXT:    s_waitcnt_depctr 0xffe3
+; GFX10-NEXT:    s_mov_b32 exec_lo, s6
+; GFX10-NEXT:    s_waitcnt vmcnt(0)
+; GFX10-NEXT:    s_setpc_b64 s[4:5]
+  call amdgpu_gfx void @external_void_func_v2f32_inreg(<2 x float> <float 1.0, float 2.0>)
+  ret void
+}
+
+define amdgpu_gfx void @test_call_external_void_func_v3f32_imm_inreg() #0 {
+; GFX9-LABEL: test_call_external_void_func_v3f32_imm_inreg:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    s_or_saveexec_b64 s[4:5], -1
+; GFX9-NEXT:    buffer_store_dword v40, off, s[0:3], s32 ; 4-byte Folded Spill
+; GFX9-NEXT:    s_mov_b64 exec, s[4:5]
+; GFX9-NEXT:    v_writelane_b32 v40, s33, 2
+; GFX9-NEXT:    v_writelane_b32 v40, s30, 0
+; GFX9-NEXT:    s_mov_b32 s33, s32
+; GFX9-NEXT:    s_add_u32 s32, s32, 0x400
+; GFX9-NEXT:    s_mov_b32 s4, 1.0
+; GFX9-NEXT:    s_mov_b32 s5, 2.0
+; GFX9-NEXT:    s_mov_b32 s6, 4.0
+; GFX9-NEXT:    s_getpc_b64 s[8:9]
+; GFX9-NEXT:    s_add_u32 s8, s8, external_void_func_v3f32_inreg@rel32@lo+4
+; GFX9-NEXT:    s_addc_u32 s9, s9, external_void_func_v3f32_inreg@rel32@hi+12
+; GFX9-NEXT:    v_writelane_b32 v40, s31, 1
+; GFX9-NEXT:    s_swappc_b64 s[30:31], s[8:9]
+; GFX9-NEXT:    v_readlane_b32 s4, v40, 0
+; GFX9-NEXT:    v_readlane_b32 s5, v40, 1
+; GFX9-NEXT:    s_sub_u32 s32, s32, 0x400
+; GFX9-NEXT:    v_readlane_b32 s33, v40, 2
+; GFX9-NEXT:    s_or_saveexec_b64 s[6:7], -1
+; GFX9-NEXT:    buffer_load_dword v40, off, s[0:3], s32 ; 4-byte Folded Reload
+; GFX9-NEXT:    s_mov_b64 exec, s[6:7]
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-NEXT:    s_setpc_b64 s[4:5]
+;
+; GFX10-LABEL: test_call_external_void_func_v3f32_imm_inreg:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
+; GFX10-NEXT:    s_or_saveexec_b32 s4, -1
+; GFX10-NEXT:    buffer_store_dword v40, off, s[0:3], s32 ; 4-byte Folded Spill
+; GFX10-NEXT:    s_waitcnt_depctr 0xffe3
+; GFX10-NEXT:    s_mov_b32 exec_lo, s4
+; GFX10-NEXT:    v_writelane_b32 v40, s33, 2
+; GFX10-NEXT:    s_mov_b32 s4, 1.0
+; GFX10-NEXT:    s_mov_b32 s5, 2.0
+; GFX10-NEXT:    s_mov_b32 s6, 4.0
+; GFX10-NEXT:    s_mov_b32 s33, s32
+; GFX10-NEXT:    v_writelane_b32 v40, s30, 0
+; GFX10-NEXT:    s_add_u32 s32, s32, 0x200
+; GFX10-NEXT:    s_getpc_b64 s[8:9]
+; GFX10-NEXT:    s_add_u32 s8, s8, external_void_func_v3f32_inreg@rel32@lo+4
+; GFX10-NEXT:    s_addc_u32 s9, s9, external_void_func_v3f32_inreg@rel32@hi+12
+; GFX10-NEXT:    ; implicit-def: $vcc_hi
+; GFX10-NEXT:    v_writelane_b32 v40, s31, 1
+; GFX10-NEXT:    s_swappc_b64 s[30:31], s[8:9]
+; GFX10-NEXT:    v_readlane_b32 s4, v40, 0
+; GFX10-NEXT:    v_readlane_b32 s5, v40, 1
+; GFX10-NEXT:    s_sub_u32 s32, s32, 0x200
+; GFX10-NEXT:    v_readlane_b32 s33, v40, 2
+; GFX10-NEXT:    s_or_saveexec_b32 s6, -1
+; GFX10-NEXT:    buffer_load_dword v40, off, s[0:3], s32 ; 4-byte Folded Reload
+; GFX10-NEXT:    s_waitcnt_depctr 0xffe3
+; GFX10-NEXT:    s_mov_b32 exec_lo, s6
+; GFX10-NEXT:    s_waitcnt vmcnt(0)
+; GFX10-NEXT:    s_setpc_b64 s[4:5]
+  call amdgpu_gfx void @external_void_func_v3f32_inreg(<3 x float> <float 1.0, float 2.0, float 4.0>)
+  ret void
+}
+
+define amdgpu_gfx void @test_call_external_void_func_v5f32_imm_inreg() #0 {
+; GFX9-LABEL: test_call_external_void_func_v5f32_imm_inreg:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    s_or_saveexec_b64 s[4:5], -1
+; GFX9-NEXT:    buffer_store_dword v40, off, s[0:3], s32 ; 4-byte Folded Spill
+; GFX9-NEXT:    s_mov_b64 exec, s[4:5]
+; GFX9-NEXT:    v_writelane_b32 v40, s33, 2
+; GFX9-NEXT:    v_writelane_b32 v40, s30, 0
+; GFX9-NEXT:    s_mov_b32 s33, s32
+; GFX9-NEXT:    s_add_u32 s32, s32, 0x400
+; GFX9-NEXT:    s_mov_b32 s4, 1.0
+; GFX9-NEXT:    s_mov_b32 s5, 2.0
+; GFX9-NEXT:    s_mov_b32 s6, 4.0
+; GFX9-NEXT:    s_mov_b32 s7, -1.0
+; GFX9-NEXT:    s_mov_b32 s8, 0.5
+; GFX9-NEXT:    s_getpc_b64 s[10:11]
+; GFX9-NEXT:    s_add_u32 s10, s10, external_void_func_v5f32_inreg@rel32@lo+4
+; GFX9-NEXT:    s_addc_u32 s11, s11, external_void_func_v5f32_inreg@rel32@hi+12
+; GFX9-NEXT:    v_writelane_b32 v40, s31, 1
+; GFX9-NEXT:    s_swappc_b64 s[30:31], s[10:11]
+; GFX9-NEXT:    v_readlane_b32 s4, v40, 0
+; GFX9-NEXT:    v_readlane_b32 s5, v40, 1
+; GFX9-NEXT:    s_sub_u32 s32, s32, 0x400
+; GFX9-NEXT:    v_readlane_b32 s33, v40, 2
+; GFX9-NEXT:    s_or_saveexec_b64 s[6:7], -1
+; GFX9-NEXT:    buffer_load_dword v40, off, s[0:3], s32 ; 4-byte Folded Reload
+; GFX9-NEXT:    s_mov_b64 exec, s[6:7]
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-NEXT:    s_setpc_b64 s[4:5]
+;
+; GFX10-LABEL: test_call_external_void_func_v5f32_imm_inreg:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
+; GFX10-NEXT:    s_or_saveexec_b32 s4, -1
+; GFX10-NEXT:    buffer_store_dword v40, off, s[0:3], s32 ; 4-byte Folded Spill
+; GFX10-NEXT:    s_waitcnt_depctr 0xffe3
+; GFX10-NEXT:    s_mov_b32 exec_lo, s4
+; GFX10-NEXT:    v_writelane_b32 v40, s33, 2
+; GFX10-NEXT:    s_mov_b32 s4, 1.0
+; GFX10-NEXT:    s_mov_b32 s5, 2.0
+; GFX10-NEXT:    s_mov_b32 s6, 4.0
+; GFX10-NEXT:    s_mov_b32 s7, -1.0
+; GFX10-NEXT:    v_writelane_b32 v40, s30, 0
+; GFX10-NEXT:    s_mov_b32 s8, 0.5
+; GFX10-NEXT:    s_mov_b32 s33, s32
+; GFX10-NEXT:    s_add_u32 s32, s32, 0x200
+; GFX10-NEXT:    s_getpc_b64 s[10:11]
+; GFX10-NEXT:    s_add_u32 s10, s10, external_void_func_v5f32_inreg@rel32@lo+4
+; GFX10-NEXT:    s_addc_u32 s11, s11, external_void_func_v5f32_inreg@rel32@hi+12
+; GFX10-NEXT:    v_writelane_b32 v40, s31, 1
+; GFX10-NEXT:    ; implicit-def: $vcc_hi
+; GFX10-NEXT:    s_swappc_b64 s[30:31], s[10:11]
+; GFX10-NEXT:    v_readlane_b32 s4, v40, 0
+; GFX10-NEXT:    v_readlane_b32 s5, v40, 1
+; GFX10-NEXT:    s_sub_u32 s32, s32, 0x200
+; GFX10-NEXT:    v_readlane_b32 s33, v40, 2
+; GFX10-NEXT:    s_or_saveexec_b32 s6, -1
+; GFX10-NEXT:    buffer_load_dword v40, off, s[0:3], s32 ; 4-byte Folded Reload
+; GFX10-NEXT:    s_waitcnt_depctr 0xffe3
+; GFX10-NEXT:    s_mov_b32 exec_lo, s6
+; GFX10-NEXT:    s_waitcnt vmcnt(0)
+; GFX10-NEXT:    s_setpc_b64 s[4:5]
+  call amdgpu_gfx void @external_void_func_v5f32_inreg(<5 x float> <float 1.0, float 2.0, float 4.0, float -1.0, float 0.5>)
+  ret void
+}
+
+define amdgpu_gfx void @test_call_external_void_func_f64_imm_inreg() #0 {
+; GFX9-LABEL: test_call_external_void_func_f64_imm_inreg:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    s_or_saveexec_b64 s[4:5], -1
+; GFX9-NEXT:    buffer_store_dword v40, off, s[0:3], s32 ; 4-byte Folded Spill
+; GFX9-NEXT:    s_mov_b64 exec, s[4:5]
+; GFX9-NEXT:    v_writelane_b32 v40, s33, 2
+; GFX9-NEXT:    v_writelane_b32 v40, s30, 0
+; GFX9-NEXT:    s_mov_b32 s33, s32
+; GFX9-NEXT:    s_add_u32 s32, s32, 0x400
+; GFX9-NEXT:    s_mov_b32 s4, 0
+; GFX9-NEXT:    s_mov_b32 s5, 0x40100000
+; GFX9-NEXT:    s_getpc_b64 s[6:7]
+; GFX9-NEXT:    s_add_u32 s6, s6, external_void_func_f64_inreg@rel32@lo+4
+; GFX9-NEXT:    s_addc_u32 s7, s7, external_void_func_f64_inreg@rel32@hi+12
+; GFX9-NEXT:    v_writelane_b32 v40, s31, 1
+; GFX9-NEXT:    s_swappc_b64 s[30:31], s[6:7]
+; GFX9-NEXT:    v_readlane_b32 s4, v40, 0
+; GFX9-NEXT:    v_readlane_b32 s5, v40, 1
+; GFX9-NEXT:    s_sub_u32 s32, s32, 0x400
+; GFX9-NEXT:    v_readlane_b32 s33, v40, 2
+; GFX9-NEXT:    s_or_saveexec_b64 s[6:7], -1
+; GFX9-NEXT:    buffer_load_dword v40, off, s[0:3], s32 ; 4-byte Folded Reload
+; GFX9-NEXT:    s_mov_b64 exec, s[6:7]
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-NEXT:    s_setpc_b64 s[4:5]
+;
+; GFX10-LABEL: test_call_external_void_func_f64_imm_inreg:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
+; GFX10-NEXT:    s_or_saveexec_b32 s4, -1
+; GFX10-NEXT:    buffer_store_dword v40, off, s[0:3], s32 ; 4-byte Folded Spill
+; GFX10-NEXT:    s_waitcnt_depctr 0xffe3
+; GFX10-NEXT:    s_mov_b32 exec_lo, s4
+; GFX10-NEXT:    v_writelane_b32 v40, s33, 2
+; GFX10-NEXT:    s_mov_b32 s4, 0
+; GFX10-NEXT:    s_mov_b32 s5, 0x40100000
+; GFX10-NEXT:    s_mov_b32 s33, s32
+; GFX10-NEXT:    s_add_u32 s32, s32, 0x200
+; GFX10-NEXT:    v_writelane_b32 v40, s30, 0
+; GFX10-NEXT:    s_getpc_b64 s[6:7]
+; GFX10-NEXT:    s_add_u32 s6, s6, external_void_func_f64_inreg@rel32@lo+4
+; GFX10-NEXT:    s_addc_u32 s7, s7, external_void_func_f64_inreg@rel32@hi+12
+; GFX10-NEXT:    ; implicit-def: $vcc_hi
+; GFX10-NEXT:    v_writelane_b32 v40, s31, 1
+; GFX10-NEXT:    s_swappc_b64 s[30:31], s[6:7]
+; GFX10-NEXT:    v_readlane_b32 s4, v40, 0
+; GFX10-NEXT:    v_readlane_b32 s5, v40, 1
+; GFX10-NEXT:    s_sub_u32 s32, s32, 0x200
+; GFX10-NEXT:    v_readlane_b32 s33, v40, 2
+; GFX10-NEXT:    s_or_saveexec_b32 s6, -1
+; GFX10-NEXT:    buffer_load_dword v40, off, s[0:3], s32 ; 4-byte Folded Reload
+; GFX10-NEXT:    s_waitcnt_depctr 0xffe3
+; GFX10-NEXT:    s_mov_b32 exec_lo, s6
+; GFX10-NEXT:    s_waitcnt vmcnt(0)
+; GFX10-NEXT:    s_setpc_b64 s[4:5]
+  call amdgpu_gfx void @external_void_func_f64_inreg(double 4.0)
+  ret void
+}
+
+define amdgpu_gfx void @test_call_external_void_func_v2f64_imm_inreg() #0 {
+; GFX9-LABEL: test_call_external_void_func_v2f64_imm_inreg:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    s_or_saveexec_b64 s[4:5], -1
+; GFX9-NEXT:    buffer_store_dword v40, off, s[0:3], s32 ; 4-byte Folded Spill
+; GFX9-NEXT:    s_mov_b64 exec, s[4:5]
+; GFX9-NEXT:    v_writelane_b32 v40, s33, 2
+; GFX9-NEXT:    v_writelane_b32 v40, s30, 0
+; GFX9-NEXT:    s_mov_b32 s33, s32
+; GFX9-NEXT:    s_add_u32 s32, s32, 0x400
+; GFX9-NEXT:    s_mov_b32 s4, 0
+; GFX9-NEXT:    s_mov_b32 s5, 2.0
+; GFX9-NEXT:    s_mov_b32 s6, 0
+; GFX9-NEXT:    s_mov_b32 s7, 0x40100000
+; GFX9-NEXT:    s_getpc_b64 s[8:9]
+; GFX9-NEXT:    s_add_u32 s8, s8, external_void_func_v2f64_inreg@rel32@lo+4
+; GFX9-NEXT:    s_addc_u32 s9, s9, external_void_func_v2f64_inreg@rel32@hi+12
+; GFX9-NEXT:    v_writelane_b32 v40, s31, 1
+; GFX9-NEXT:    s_swappc_b64 s[30:31], s[8:9]
+; GFX9-NEXT:    v_readlane_b32 s4, v40, 0
+; GFX9-NEXT:    v_readlane_b32 s5, v40, 1
+; GFX9-NEXT:    s_sub_u32 s32, s32, 0x400
+; GFX9-NEXT:    v_readlane_b32 s33, v40, 2
+; GFX9-NEXT:    s_or_saveexec_b64 s[6:7], -1
+; GFX9-NEXT:    buffer_load_dword v40, off, s[0:3], s32 ; 4-byte Folded Reload
+; GFX9-NEXT:    s_mov_b64 exec, s[6:7]
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-NEXT:    s_setpc_b64 s[4:5]
+;
+; GFX10-LABEL: test_call_external_void_func_v2f64_imm_inreg:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
+; GFX10-NEXT:    s_or_saveexec_b32 s4, -1
+; GFX10-NEXT:    buffer_store_dword v40, off, s[0:3], s32 ; 4-byte Folded Spill
+; GFX10-NEXT:    s_waitcnt_depctr 0xffe3
+; GFX10-NEXT:    s_mov_b32 exec_lo, s4
+; GFX10-NEXT:    v_writelane_b32 v40, s33, 2
+; GFX10-NEXT:    s_mov_b32 s4, 0
+; GFX10-NEXT:    s_mov_b32 s5, 2.0
+; GFX10-NEXT:    s_mov_b32 s6, 0
+; GFX10-NEXT:    s_mov_b32 s7, 0x40100000
+; GFX10-NEXT:    v_writelane_b32 v40, s30, 0
+; GFX10-NEXT:    s_mov_b32 s33, s32
+; GFX10-NEXT:    s_add_u32 s32, s32, 0x200
+; GFX10-NEXT:    s_getpc_b64 s[8:9]
+; GFX10-NEXT:    s_add_u32 s8, s8, external_void_func_v2f64_inreg@rel32@lo+4
+; GFX10-NEXT:    s_addc_u32 s9, s9, external_void_func_v2f64_inreg@rel32@hi+12
+; GFX10-NEXT:    ; implicit-def: $vcc_hi
+; GFX10-NEXT:    v_writelane_b32 v40, s31, 1
+; GFX10-NEXT:    s_swappc_b64 s[30:31], s[8:9]
+; GFX10-NEXT:    v_readlane_b32 s4, v40, 0
+; GFX10-NEXT:    v_readlane_b32 s5, v40, 1
+; GFX10-NEXT:    s_sub_u32 s32, s32, 0x200
+; GFX10-NEXT:    v_readlane_b32 s33, v40, 2
+; GFX10-NEXT:    s_or_saveexec_b32 s6, -1
+; GFX10-NEXT:    buffer_load_dword v40, off, s[0:3], s32 ; 4-byte Folded Reload
+; GFX10-NEXT:    s_waitcnt_depctr 0xffe3
+; GFX10-NEXT:    s_mov_b32 exec_lo, s6
+; GFX10-NEXT:    s_waitcnt vmcnt(0)
+; GFX10-NEXT:    s_setpc_b64 s[4:5]
+  call amdgpu_gfx void @external_void_func_v2f64_inreg(<2 x double> <double 2.0, double 4.0>)
+  ret void
+}
+
+define amdgpu_gfx void @test_call_external_void_func_v3f64_imm_inreg() #0 {
+; GFX9-LABEL: test_call_external_void_func_v3f64_imm_inreg:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    s_or_saveexec_b64 s[4:5], -1
+; GFX9-NEXT:    buffer_store_dword v40, off, s[0:3], s32 ; 4-byte Folded Spill
+; GFX9-NEXT:    s_mov_b64 exec, s[4:5]
+; GFX9-NEXT:    v_writelane_b32 v40, s33, 2
+; GFX9-NEXT:    v_writelane_b32 v40, s30, 0
+; GFX9-NEXT:    s_mov_b32 s33, s32
+; GFX9-NEXT:    s_add_u32 s32, s32, 0x400
+; GFX9-NEXT:    s_mov_b32 s4, 0
+; GFX9-NEXT:    s_mov_b32 s5, 2.0
+; GFX9-NEXT:    s_mov_b32 s6, 0
+; GFX9-NEXT:    s_mov_b32 s7, 0x40100000
+; GFX9-NEXT:    s_mov_b32 s8, 0
+; GFX9-NEXT:    s_mov_b32 s9, 0x40200000
+; GFX9-NEXT:    s_getpc_b64 s[10:11]
+; GFX9-NEXT:    s_add_u32 s10, s10, external_void_func_v3f64_inreg@rel32@lo+4
+; GFX9-NEXT:    s_addc_u32 s11, s11, external_void_func_v3f64_inreg@rel32@hi+12
+; GFX9-NEXT:    v_writelane_b32 v40, s31, 1
+; GFX9-NEXT:    s_swappc_b64 s[30:31], s[10:11]
+; GFX9-NEXT:    v_readlane_b32 s4, v40, 0
+; GFX9-NEXT:    v_readlane_b32 s5, v40, 1
+; GFX9-NEXT:    s_sub_u32 s32, s32, 0x400
+; GFX9-NEXT:    v_readlane_b32 s33, v40, 2
+; GFX9-NEXT:    s_or_saveexec_b64 s[6:7], -1
+; GFX9-NEXT:    buffer_load_dword v40, off, s[0:3], s32 ; 4-byte Folded Reload
+; GFX9-NEXT:    s_mov_b64 exec, s[6:7]
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-NEXT:    s_setpc_b64 s[4:5]
+;
+; GFX10-LABEL: test_call_external_void_func_v3f64_imm_inreg:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
+; GFX10-NEXT:    s_or_saveexec_b32 s4, -1
+; GFX10-NEXT:    buffer_store_dword v40, off, s[0:3], s32 ; 4-byte Folded Spill
+; GFX10-NEXT:    s_waitcnt_depctr 0xffe3
+; GFX10-NEXT:    s_mov_b32 exec_lo, s4
+; GFX10-NEXT:    v_writelane_b32 v40, s33, 2
+; GFX10-NEXT:    s_mov_b32 s4, 0
+; GFX10-NEXT:    s_mov_b32 s5, 2.0
+; GFX10-NEXT:    s_mov_b32 s6, 0
+; GFX10-NEXT:    s_mov_b32 s7, 0x40100000
+; GFX10-NEXT:    v_writelane_b32 v40, s30, 0
+; GFX10-NEXT:    s_mov_b32 s8, 0
+; GFX10-NEXT:    s_mov_b32 s9, 0x40200000
+; GFX10-NEXT:    s_mov_b32 s33, s32
+; GFX10-NEXT:    s_add_u32 s32, s32, 0x200
+; GFX10-NEXT:    s_getpc_b64 s[10:11]
+; GFX10-NEXT:    s_add_u32 s10, s10, external_void_func_v3f64_inreg@rel32@lo+4
+; GFX10-NEXT:    s_addc_u32 s11, s11, external_void_func_v3f64_inreg@rel32@hi+12
+; GFX10-NEXT:    v_writelane_b32 v40, s31, 1
+; GFX10-NEXT:    ; implicit-def: $vcc_hi
+; GFX10-NEXT:    s_swappc_b64 s[30:31], s[10:11]
+; GFX10-NEXT:    v_readlane_b32 s4, v40, 0
+; GFX10-NEXT:    v_readlane_b32 s5, v40, 1
+; GFX10-NEXT:    s_sub_u32 s32, s32, 0x200
+; GFX10-NEXT:    v_readlane_b32 s33, v40, 2
+; GFX10-NEXT:    s_or_saveexec_b32 s6, -1
+; GFX10-NEXT:    buffer_load_dword v40, off, s[0:3], s32 ; 4-byte Folded Reload
+; GFX10-NEXT:    s_waitcnt_depctr 0xffe3
+; GFX10-NEXT:    s_mov_b32 exec_lo, s6
+; GFX10-NEXT:    s_waitcnt vmcnt(0)
+; GFX10-NEXT:    s_setpc_b64 s[4:5]
+  call amdgpu_gfx void @external_void_func_v3f64_inreg(<3 x double> <double 2.0, double 4.0, double 8.0>)
+  ret void
+}
+
+define amdgpu_gfx void @test_call_external_void_func_v2i16_inreg() #0 {
+; GFX9-LABEL: test_call_external_void_func_v2i16_inreg:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    s_or_saveexec_b64 s[4:5], -1
+; GFX9-NEXT:    buffer_store_dword v40, off, s[0:3], s32 ; 4-byte Folded Spill
+; GFX9-NEXT:    s_mov_b64 exec, s[4:5]
+; GFX9-NEXT:    s_load_dword s4, s[4:5], 0x0
+; GFX9-NEXT:    v_writelane_b32 v40, s33, 2
+; GFX9-NEXT:    v_writelane_b32 v40, s30, 0
+; GFX9-NEXT:    s_mov_b32 s33, s32
+; GFX9-NEXT:    s_add_u32 s32, s32, 0x400
+; GFX9-NEXT:    s_getpc_b64 s[6:7]
+; GFX9-NEXT:    s_add_u32 s6, s6, external_void_func_v2i16_inreg@rel32@lo+4
+; GFX9-NEXT:    s_addc_u32 s7, s7, external_void_func_v2i16_inreg@rel32@hi+12
+; GFX9-NEXT:    v_writelane_b32 v40, s31, 1
+; GFX9-NEXT:    s_swappc_b64 s[30:31], s[6:7]
+; GFX9-NEXT:    v_readlane_b32 s4, v40, 0
+; GFX9-NEXT:    v_readlane_b32 s5, v40, 1
+; GFX9-NEXT:    s_sub_u32 s32, s32, 0x400
+; GFX9-NEXT:    v_readlane_b32 s33, v40, 2
+; GFX9-NEXT:    s_or_saveexec_b64 s[6:7], -1
+; GFX9-NEXT:    buffer_load_dword v40, off, s[0:3], s32 ; 4-byte Folded Reload
+; GFX9-NEXT:    s_mov_b64 exec, s[6:7]
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-NEXT:    s_setpc_b64 s[4:5]
+;
+; GFX10-LABEL: test_call_external_void_func_v2i16_inreg:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
+; GFX10-NEXT:    s_or_saveexec_b32 s4, -1
+; GFX10-NEXT:    buffer_store_dword v40, off, s[0:3], s32 ; 4-byte Folded Spill
+; GFX10-NEXT:    s_waitcnt_depctr 0xffe3
+; GFX10-NEXT:    s_mov_b32 exec_lo, s4
+; GFX10-NEXT:    s_load_dword s4, s[4:5], 0x0
+; GFX10-NEXT:    v_writelane_b32 v40, s33, 2
+; GFX10-NEXT:    s_mov_b32 s33, s32
+; GFX10-NEXT:    s_add_u32 s32, s32, 0x200
+; GFX10-NEXT:    s_getpc_b64 s[6:7]
+; GFX10-NEXT:    s_add_u32 s6, s6, external_void_func_v2i16_inreg@rel32@lo+4
+; GFX10-NEXT:    s_addc_u32 s7, s7, external_void_func_v2i16_inreg@rel32@hi+12
+; GFX10-NEXT:    ; implicit-def: $vcc_hi
+; GFX10-NEXT:    v_writelane_b32 v40, s30, 0
+; GFX10-NEXT:    v_writelane_b32 v40, s31, 1
+; GFX10-NEXT:    s_swappc_b64 s[30:31], s[6:7]
+; GFX10-NEXT:    v_readlane_b32 s4, v40, 0
+; GFX10-NEXT:    v_readlane_b32 s5, v40, 1
+; GFX10-NEXT:    s_sub_u32 s32, s32, 0x200
+; GFX10-NEXT:    v_readlane_b32 s33, v40, 2
+; GFX10-NEXT:    s_or_saveexec_b32 s6, -1
+; GFX10-NEXT:    buffer_load_dword v40, off, s[0:3], s32 ; 4-byte Folded Reload
+; GFX10-NEXT:    s_waitcnt_depctr 0xffe3
+; GFX10-NEXT:    s_mov_b32 exec_lo, s6
+; GFX10-NEXT:    s_waitcnt vmcnt(0)
+; GFX10-NEXT:    s_setpc_b64 s[4:5]
+  %val = load <2 x i16>, <2 x i16> addrspace(4)* undef
+  call amdgpu_gfx void @external_void_func_v2i16_inreg(<2 x i16> %val)
+  ret void
+}
+
+define amdgpu_gfx void @test_call_external_void_func_v3i16_inreg() #0 {
+; GFX9-LABEL: test_call_external_void_func_v3i16_inreg:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    s_or_saveexec_b64 s[4:5], -1
+; GFX9-NEXT:    buffer_store_dword v40, off, s[0:3], s32 ; 4-byte Folded Spill
+; GFX9-NEXT:    s_mov_b64 exec, s[4:5]
+; GFX9-NEXT:    s_load_dwordx2 s[4:5], s[4:5], 0x0
+; GFX9-NEXT:    v_writelane_b32 v40, s33, 2
+; GFX9-NEXT:    v_writelane_b32 v40, s30, 0
+; GFX9-NEXT:    s_mov_b32 s33, s32
+; GFX9-NEXT:    s_add_u32 s32, s32, 0x400
+; GFX9-NEXT:    s_getpc_b64 s[6:7]
+; GFX9-NEXT:    s_add_u32 s6, s6, external_void_func_v3i16_inreg@rel32@lo+4
+; GFX9-NEXT:    s_addc_u32 s7, s7, external_void_func_v3i16_inreg@rel32@hi+12
+; GFX9-NEXT:    v_writelane_b32 v40, s31, 1
+; GFX9-NEXT:    s_swappc_b64 s[30:31], s[6:7]
+; GFX9-NEXT:    v_readlane_b32 s4, v40, 0
+; GFX9-NEXT:    v_readlane_b32 s5, v40, 1
+; GFX9-NEXT:    s_sub_u32 s32, s32, 0x400
+; GFX9-NEXT:    v_readlane_b32 s33, v40, 2
+; GFX9-NEXT:    s_or_saveexec_b64 s[6:7], -1
+; GFX9-NEXT:    buffer_load_dword v40, off, s[0:3], s32 ; 4-byte Folded Reload
+; GFX9-NEXT:    s_mov_b64 exec, s[6:7]
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-NEXT:    s_setpc_b64 s[4:5]
+;
+; GFX10-LABEL: test_call_external_void_func_v3i16_inreg:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
+; GFX10-NEXT:    s_or_saveexec_b32 s4, -1
+; GFX10-NEXT:    buffer_store_dword v40, off, s[0:3], s32 ; 4-byte Folded Spill
+; GFX10-NEXT:    s_waitcnt_depctr 0xffe3
+; GFX10-NEXT:    s_mov_b32 exec_lo, s4
+; GFX10-NEXT:    s_load_dwordx2 s[4:5], s[4:5], 0x0
+; GFX10-NEXT:    v_writelane_b32 v40, s33, 2
+; GFX10-NEXT:    s_mov_b32 s33, s32
+; GFX10-NEXT:    s_add_u32 s32, s32, 0x200
+; GFX10-NEXT:    s_getpc_b64 s[6:7]
+; GFX10-NEXT:    s_add_u32 s6, s6, external_void_func_v3i16_inreg@rel32@lo+4
+; GFX10-NEXT:    s_addc_u32 s7, s7, external_void_func_v3i16_inreg@rel32@hi+12
+; GFX10-NEXT:    ; implicit-def: $vcc_hi
+; GFX10-NEXT:    v_writelane_b32 v40, s30, 0
+; GFX10-NEXT:    v_writelane_b32 v40, s31, 1
+; GFX10-NEXT:    s_swappc_b64 s[30:31], s[6:7]
+; GFX10-NEXT:    v_readlane_b32 s4, v40, 0
+; GFX10-NEXT:    v_readlane_b32 s5, v40, 1
+; GFX10-NEXT:    s_sub_u32 s32, s32, 0x200
+; GFX10-NEXT:    v_readlane_b32 s33, v40, 2
+; GFX10-NEXT:    s_or_saveexec_b32 s6, -1
+; GFX10-NEXT:    buffer_load_dword v40, off, s[0:3], s32 ; 4-byte Folded Reload
+; GFX10-NEXT:    s_waitcnt_depctr 0xffe3
+; GFX10-NEXT:    s_mov_b32 exec_lo, s6
+; GFX10-NEXT:    s_waitcnt vmcnt(0)
+; GFX10-NEXT:    s_setpc_b64 s[4:5]
+  %val = load <3 x i16>, <3 x i16> addrspace(4)* undef
+  call amdgpu_gfx void @external_void_func_v3i16_inreg(<3 x i16> %val)
+  ret void
+}
+
+define amdgpu_gfx void @test_call_external_void_func_v3f16_inreg() #0 {
+; GFX9-LABEL: test_call_external_void_func_v3f16_inreg:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    s_or_saveexec_b64 s[4:5], -1
+; GFX9-NEXT:    buffer_store_dword v40, off, s[0:3], s32 ; 4-byte Folded Spill
+; GFX9-NEXT:    s_mov_b64 exec, s[4:5]
+; GFX9-NEXT:    s_load_dwordx2 s[4:5], s[4:5], 0x0
+; GFX9-NEXT:    v_writelane_b32 v40, s33, 2
+; GFX9-NEXT:    v_writelane_b32 v40, s30, 0
+; GFX9-NEXT:    s_mov_b32 s33, s32
+; GFX9-NEXT:    s_add_u32 s32, s32, 0x400
+; GFX9-NEXT:    s_getpc_b64 s[6:7]
+; GFX9-NEXT:    s_add_u32 s6, s6, external_void_func_v3f16_inreg@rel32@lo+4
+; GFX9-NEXT:    s_addc_u32 s7, s7, external_void_func_v3f16_inreg@rel32@hi+12
+; GFX9-NEXT:    v_writelane_b32 v40, s31, 1
+; GFX9-NEXT:    s_swappc_b64 s[30:31], s[6:7]
+; GFX9-NEXT:    v_readlane_b32 s4, v40, 0
+; GFX9-NEXT:    v_readlane_b32 s5, v40, 1
+; GFX9-NEXT:    s_sub_u32 s32, s32, 0x400
+; GFX9-NEXT:    v_readlane_b32 s33, v40, 2
+; GFX9-NEXT:    s_or_saveexec_b64 s[6:7], -1
+; GFX9-NEXT:    buffer_load_dword v40, off, s[0:3], s32 ; 4-byte Folded Reload
+; GFX9-NEXT:    s_mov_b64 exec, s[6:7]
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-NEXT:    s_setpc_b64 s[4:5]
+;
+; GFX10-LABEL: test_call_external_void_func_v3f16_inreg:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
+; GFX10-NEXT:    s_or_saveexec_b32 s4, -1
+; GFX10-NEXT:    buffer_store_dword v40, off, s[0:3], s32 ; 4-byte Folded Spill
+; GFX10-NEXT:    s_waitcnt_depctr 0xffe3
+; GFX10-NEXT:    s_mov_b32 exec_lo, s4
+; GFX10-NEXT:    s_load_dwordx2 s[4:5], s[4:5], 0x0
+; GFX10-NEXT:    v_writelane_b32 v40, s33, 2
+; GFX10-NEXT:    s_mov_b32 s33, s32
+; GFX10-NEXT:    s_add_u32 s32, s32, 0x200
+; GFX10-NEXT:    s_getpc_b64 s[6:7]
+; GFX10-NEXT:    s_add_u32 s6, s6, external_void_func_v3f16_inreg@rel32@lo+4
+; GFX10-NEXT:    s_addc_u32 s7, s7, external_void_func_v3f16_inreg@rel32@hi+12
+; GFX10-NEXT:    ; implicit-def: $vcc_hi
+; GFX10-NEXT:    v_writelane_b32 v40, s30, 0
+; GFX10-NEXT:    v_writelane_b32 v40, s31, 1
+; GFX10-NEXT:    s_swappc_b64 s[30:31], s[6:7]
+; GFX10-NEXT:    v_readlane_b32 s4, v40, 0
+; GFX10-NEXT:    v_readlane_b32 s5, v40, 1
+; GFX10-NEXT:    s_sub_u32 s32, s32, 0x200
+; GFX10-NEXT:    v_readlane_b32 s33, v40, 2
+; GFX10-NEXT:    s_or_saveexec_b32 s6, -1
+; GFX10-NEXT:    buffer_load_dword v40, off, s[0:3], s32 ; 4-byte Folded Reload
+; GFX10-NEXT:    s_waitcnt_depctr 0xffe3
+; GFX10-NEXT:    s_mov_b32 exec_lo, s6
+; GFX10-NEXT:    s_waitcnt vmcnt(0)
+; GFX10-NEXT:    s_setpc_b64 s[4:5]
+  %val = load <3 x half>, <3 x half> addrspace(4)* undef
+  call amdgpu_gfx void @external_void_func_v3f16_inreg(<3 x half> %val)
+  ret void
+}
+
+define amdgpu_gfx void @test_call_external_void_func_v3i16_imm_inreg() #0 {
+; GFX9-LABEL: test_call_external_void_func_v3i16_imm_inreg:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    s_or_saveexec_b64 s[4:5], -1
+; GFX9-NEXT:    buffer_store_dword v40, off, s[0:3], s32 ; 4-byte Folded Spill
+; GFX9-NEXT:    s_mov_b64 exec, s[4:5]
+; GFX9-NEXT:    v_writelane_b32 v40, s33, 2
+; GFX9-NEXT:    v_writelane_b32 v40, s30, 0
+; GFX9-NEXT:    s_mov_b32 s33, s32
+; GFX9-NEXT:    s_add_u32 s32, s32, 0x400
+; GFX9-NEXT:    s_mov_b32 s4, 0x20001
+; GFX9-NEXT:    s_mov_b32 s5, 3
+; GFX9-NEXT:    s_getpc_b64 s[6:7]
+; GFX9-NEXT:    s_add_u32 s6, s6, external_void_func_v3i16_inreg@rel32@lo+4
+; GFX9-NEXT:    s_addc_u32 s7, s7, external_void_func_v3i16_inreg@rel32@hi+12
+; GFX9-NEXT:    v_writelane_b32 v40, s31, 1
+; GFX9-NEXT:    s_swappc_b64 s[30:31], s[6:7]
+; GFX9-NEXT:    v_readlane_b32 s4, v40, 0
+; GFX9-NEXT:    v_readlane_b32 s5, v40, 1
+; GFX9-NEXT:    s_sub_u32 s32, s32, 0x400
+; GFX9-NEXT:    v_readlane_b32 s33, v40, 2
+; GFX9-NEXT:    s_or_saveexec_b64 s[6:7], -1
+; GFX9-NEXT:    buffer_load_dword v40, off, s[0:3], s32 ; 4-byte Folded Reload
+; GFX9-NEXT:    s_mov_b64 exec, s[6:7]
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-NEXT:    s_setpc_b64 s[4:5]
+;
+; GFX10-LABEL: test_call_external_void_func_v3i16_imm_inreg:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
+; GFX10-NEXT:    s_or_saveexec_b32 s4, -1
+; GFX10-NEXT:    buffer_store_dword v40, off, s[0:3], s32 ; 4-byte Folded Spill
+; GFX10-NEXT:    s_waitcnt_depctr 0xffe3
+; GFX10-NEXT:    s_mov_b32 exec_lo, s4
+; GFX10-NEXT:    v_writelane_b32 v40, s33, 2
+; GFX10-NEXT:    s_mov_b32 s4, 0x20001
+; GFX10-NEXT:    s_mov_b32 s5, 3
+; GFX10-NEXT:    s_mov_b32 s33, s32
+; GFX10-NEXT:    s_add_u32 s32, s32, 0x200
+; GFX10-NEXT:    v_writelane_b32 v40, s30, 0
+; GFX10-NEXT:    s_getpc_b64 s[6:7]
+; GFX10-NEXT:    s_add_u32 s6, s6, external_void_func_v3i16_inreg@rel32@lo+4
+; GFX10-NEXT:    s_addc_u32 s7, s7, external_void_func_v3i16_inreg@rel32@hi+12
+; GFX10-NEXT:    ; implicit-def: $vcc_hi
+; GFX10-NEXT:    v_writelane_b32 v40, s31, 1
+; GFX10-NEXT:    s_swappc_b64 s[30:31], s[6:7]
+; GFX10-NEXT:    v_readlane_b32 s4, v40, 0
+; GFX10-NEXT:    v_readlane_b32 s5, v40, 1
+; GFX10-NEXT:    s_sub_u32 s32, s32, 0x200
+; GFX10-NEXT:    v_readlane_b32 s33, v40, 2
+; GFX10-NEXT:    s_or_saveexec_b32 s6, -1
+; GFX10-NEXT:    buffer_load_dword v40, off, s[0:3], s32 ; 4-byte Folded Reload
+; GFX10-NEXT:    s_waitcnt_depctr 0xffe3
+; GFX10-NEXT:    s_mov_b32 exec_lo, s6
+; GFX10-NEXT:    s_waitcnt vmcnt(0)
+; GFX10-NEXT:    s_setpc_b64 s[4:5]
+  call amdgpu_gfx void @external_void_func_v3i16_inreg(<3 x i16> <i16 1, i16 2, i16 3>)
+  ret void
+}
+
+define amdgpu_gfx void @test_call_external_void_func_v3f16_imm_inreg() #0 {
+; GFX9-LABEL: test_call_external_void_func_v3f16_imm_inreg:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    s_or_saveexec_b64 s[4:5], -1
+; GFX9-NEXT:    buffer_store_dword v40, off, s[0:3], s32 ; 4-byte Folded Spill
+; GFX9-NEXT:    s_mov_b64 exec, s[4:5]
+; GFX9-NEXT:    v_writelane_b32 v40, s33, 2
+; GFX9-NEXT:    v_writelane_b32 v40, s30, 0
+; GFX9-NEXT:    s_mov_b32 s33, s32
+; GFX9-NEXT:    s_add_u32 s32, s32, 0x400
+; GFX9-NEXT:    s_mov_b32 s4, 0x40003c00
+; GFX9-NEXT:    s_movk_i32 s5, 0x4400
+; GFX9-NEXT:    s_getpc_b64 s[6:7]
+; GFX9-NEXT:    s_add_u32 s6, s6, external_void_func_v3f16_inreg@rel32@lo+4
+; GFX9-NEXT:    s_addc_u32 s7, s7, external_void_func_v3f16_inreg@rel32@hi+12
+; GFX9-NEXT:    v_writelane_b32 v40, s31, 1
+; GFX9-NEXT:    s_swappc_b64 s[30:31], s[6:7]
+; GFX9-NEXT:    v_readlane_b32 s4, v40, 0
+; GFX9-NEXT:    v_readlane_b32 s5, v40, 1
+; GFX9-NEXT:    s_sub_u32 s32, s32, 0x400
+; GFX9-NEXT:    v_readlane_b32 s33, v40, 2
+; GFX9-NEXT:    s_or_saveexec_b64 s[6:7], -1
+; GFX9-NEXT:    buffer_load_dword v40, off, s[0:3], s32 ; 4-byte Folded Reload
+; GFX9-NEXT:    s_mov_b64 exec, s[6:7]
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-NEXT:    s_setpc_b64 s[4:5]
+;
+; GFX10-LABEL: test_call_external_void_func_v3f16_imm_inreg:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
+; GFX10-NEXT:    s_or_saveexec_b32 s4, -1
+; GFX10-NEXT:    buffer_store_dword v40, off, s[0:3], s32 ; 4-byte Folded Spill
+; GFX10-NEXT:    s_waitcnt_depctr 0xffe3
+; GFX10-NEXT:    s_mov_b32 exec_lo, s4
+; GFX10-NEXT:    v_writelane_b32 v40, s33, 2
+; GFX10-NEXT:    s_mov_b32 s4, 0x40003c00
+; GFX10-NEXT:    s_movk_i32 s5, 0x4400
+; GFX10-NEXT:    s_mov_b32 s33, s32
+; GFX10-NEXT:    s_add_u32 s32, s32, 0x200
+; GFX10-NEXT:    v_writelane_b32 v40, s30, 0
+; GFX10-NEXT:    s_getpc_b64 s[6:7]
+; GFX10-NEXT:    s_add_u32 s6, s6, external_void_func_v3f16_inreg@rel32@lo+4
+; GFX10-NEXT:    s_addc_u32 s7, s7, external_void_func_v3f16_inreg@rel32@hi+12
+; GFX10-NEXT:    ; implicit-def: $vcc_hi
+; GFX10-NEXT:    v_writelane_b32 v40, s31, 1
+; GFX10-NEXT:    s_swappc_b64 s[30:31], s[6:7]
+; GFX10-NEXT:    v_readlane_b32 s4, v40, 0
+; GFX10-NEXT:    v_readlane_b32 s5, v40, 1
+; GFX10-NEXT:    s_sub_u32 s32, s32, 0x200
+; GFX10-NEXT:    v_readlane_b32 s33, v40, 2
+; GFX10-NEXT:    s_or_saveexec_b32 s6, -1
+; GFX10-NEXT:    buffer_load_dword v40, off, s[0:3], s32 ; 4-byte Folded Reload
+; GFX10-NEXT:    s_waitcnt_depctr 0xffe3
+; GFX10-NEXT:    s_mov_b32 exec_lo, s6
+; GFX10-NEXT:    s_waitcnt vmcnt(0)
+; GFX10-NEXT:    s_setpc_b64 s[4:5]
+  call amdgpu_gfx void @external_void_func_v3f16_inreg(<3 x half> <half 1.0, half 2.0, half 4.0>)
+  ret void
+}
+
+define amdgpu_gfx void @test_call_external_void_func_v4i16_inreg() #0 {
+; GFX9-LABEL: test_call_external_void_func_v4i16_inreg:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    s_or_saveexec_b64 s[4:5], -1
+; GFX9-NEXT:    buffer_store_dword v40, off, s[0:3], s32 ; 4-byte Folded Spill
+; GFX9-NEXT:    s_mov_b64 exec, s[4:5]
+; GFX9-NEXT:    s_load_dwordx2 s[4:5], s[4:5], 0x0
+; GFX9-NEXT:    v_writelane_b32 v40, s33, 2
+; GFX9-NEXT:    v_writelane_b32 v40, s30, 0
+; GFX9-NEXT:    s_mov_b32 s33, s32
+; GFX9-NEXT:    s_add_u32 s32, s32, 0x400
+; GFX9-NEXT:    s_getpc_b64 s[6:7]
+; GFX9-NEXT:    s_add_u32 s6, s6, external_void_func_v4i16_inreg@rel32@lo+4
+; GFX9-NEXT:    s_addc_u32 s7, s7, external_void_func_v4i16_inreg@rel32@hi+12
+; GFX9-NEXT:    v_writelane_b32 v40, s31, 1
+; GFX9-NEXT:    s_swappc_b64 s[30:31], s[6:7]
+; GFX9-NEXT:    v_readlane_b32 s4, v40, 0
+; GFX9-NEXT:    v_readlane_b32 s5, v40, 1
+; GFX9-NEXT:    s_sub_u32 s32, s32, 0x400
+; GFX9-NEXT:    v_readlane_b32 s33, v40, 2
+; GFX9-NEXT:    s_or_saveexec_b64 s[6:7], -1
+; GFX9-NEXT:    buffer_load_dword v40, off, s[0:3], s32 ; 4-byte Folded Reload
+; GFX9-NEXT:    s_mov_b64 exec, s[6:7]
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-NEXT:    s_setpc_b64 s[4:5]
+;
+; GFX10-LABEL: test_call_external_void_func_v4i16_inreg:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
+; GFX10-NEXT:    s_or_saveexec_b32 s4, -1
+; GFX10-NEXT:    buffer_store_dword v40, off, s[0:3], s32 ; 4-byte Folded Spill
+; GFX10-NEXT:    s_waitcnt_depctr 0xffe3
+; GFX10-NEXT:    s_mov_b32 exec_lo, s4
+; GFX10-NEXT:    s_load_dwordx2 s[4:5], s[4:5], 0x0
+; GFX10-NEXT:    v_writelane_b32 v40, s33, 2
+; GFX10-NEXT:    s_mov_b32 s33, s32
+; GFX10-NEXT:    s_add_u32 s32, s32, 0x200
+; GFX10-NEXT:    s_getpc_b64 s[6:7]
+; GFX10-NEXT:    s_add_u32 s6, s6, external_void_func_v4i16_inreg@rel32@lo+4
+; GFX10-NEXT:    s_addc_u32 s7, s7, external_void_func_v4i16_inreg@rel32@hi+12
+; GFX10-NEXT:    ; implicit-def: $vcc_hi
+; GFX10-NEXT:    v_writelane_b32 v40, s30, 0
+; GFX10-NEXT:    v_writelane_b32 v40, s31, 1
+; GFX10-NEXT:    s_swappc_b64 s[30:31], s[6:7]
+; GFX10-NEXT:    v_readlane_b32 s4, v40, 0
+; GFX10-NEXT:    v_readlane_b32 s5, v40, 1
+; GFX10-NEXT:    s_sub_u32 s32, s32, 0x200
+; GFX10-NEXT:    v_readlane_b32 s33, v40, 2
+; GFX10-NEXT:    s_or_saveexec_b32 s6, -1
+; GFX10-NEXT:    buffer_load_dword v40, off, s[0:3], s32 ; 4-byte Folded Reload
+; GFX10-NEXT:    s_waitcnt_depctr 0xffe3
+; GFX10-NEXT:    s_mov_b32 exec_lo, s6
+; GFX10-NEXT:    s_waitcnt vmcnt(0)
+; GFX10-NEXT:    s_setpc_b64 s[4:5]
+  %val = load <4 x i16>, <4 x i16> addrspace(4)* undef
+  call amdgpu_gfx void @external_void_func_v4i16_inreg(<4 x i16> %val)
+  ret void
+}
+
+define amdgpu_gfx void @test_call_external_void_func_v4i16_imm_inreg() #0 {
+; GFX9-LABEL: test_call_external_void_func_v4i16_imm_inreg:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    s_or_saveexec_b64 s[4:5], -1
+; GFX9-NEXT:    buffer_store_dword v40, off, s[0:3], s32 ; 4-byte Folded Spill
+; GFX9-NEXT:    s_mov_b64 exec, s[4:5]
+; GFX9-NEXT:    v_writelane_b32 v40, s33, 2
+; GFX9-NEXT:    v_writelane_b32 v40, s30, 0
+; GFX9-NEXT:    s_mov_b32 s33, s32
+; GFX9-NEXT:    s_add_u32 s32, s32, 0x400
+; GFX9-NEXT:    s_mov_b32 s4, 0x20001
+; GFX9-NEXT:    s_mov_b32 s5, 0x40003
+; GFX9-NEXT:    s_getpc_b64 s[6:7]
+; GFX9-NEXT:    s_add_u32 s6, s6, external_void_func_v4i16_inreg@rel32@lo+4
+; GFX9-NEXT:    s_addc_u32 s7, s7, external_void_func_v4i16_inreg@rel32@hi+12
+; GFX9-NEXT:    v_writelane_b32 v40, s31, 1
+; GFX9-NEXT:    s_swappc_b64 s[30:31], s[6:7]
+; GFX9-NEXT:    v_readlane_b32 s4, v40, 0
+; GFX9-NEXT:    v_readlane_b32 s5, v40, 1
+; GFX9-NEXT:    s_sub_u32 s32, s32, 0x400
+; GFX9-NEXT:    v_readlane_b32 s33, v40, 2
+; GFX9-NEXT:    s_or_saveexec_b64 s[6:7], -1
+; GFX9-NEXT:    buffer_load_dword v40, off, s[0:3], s32 ; 4-byte Folded Reload
+; GFX9-NEXT:    s_mov_b64 exec, s[6:7]
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-NEXT:    s_setpc_b64 s[4:5]
+;
+; GFX10-LABEL: test_call_external_void_func_v4i16_imm_inreg:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
+; GFX10-NEXT:    s_or_saveexec_b32 s4, -1
+; GFX10-NEXT:    buffer_store_dword v40, off, s[0:3], s32 ; 4-byte Folded Spill
+; GFX10-NEXT:    s_waitcnt_depctr 0xffe3
+; GFX10-NEXT:    s_mov_b32 exec_lo, s4
+; GFX10-NEXT:    v_writelane_b32 v40, s33, 2
+; GFX10-NEXT:    s_mov_b32 s4, 0x20001
+; GFX10-NEXT:    s_mov_b32 s5, 0x40003
+; GFX10-NEXT:    s_mov_b32 s33, s32
+; GFX10-NEXT:    s_add_u32 s32, s32, 0x200
+; GFX10-NEXT:    v_writelane_b32 v40, s30, 0
+; GFX10-NEXT:    s_getpc_b64 s[6:7]
+; GFX10-NEXT:    s_add_u32 s6, s6, external_void_func_v4i16_inreg@rel32@lo+4
+; GFX10-NEXT:    s_addc_u32 s7, s7, external_void_func_v4i16_inreg@rel32@hi+12
+; GFX10-NEXT:    ; implicit-def: $vcc_hi
+; GFX10-NEXT:    v_writelane_b32 v40, s31, 1
+; GFX10-NEXT:    s_swappc_b64 s[30:31], s[6:7]
+; GFX10-NEXT:    v_readlane_b32 s4, v40, 0
+; GFX10-NEXT:    v_readlane_b32 s5, v40, 1
+; GFX10-NEXT:    s_sub_u32 s32, s32, 0x200
+; GFX10-NEXT:    v_readlane_b32 s33, v40, 2
+; GFX10-NEXT:    s_or_saveexec_b32 s6, -1
+; GFX10-NEXT:    buffer_load_dword v40, off, s[0:3], s32 ; 4-byte Folded Reload
+; GFX10-NEXT:    s_waitcnt_depctr 0xffe3
+; GFX10-NEXT:    s_mov_b32 exec_lo, s6
+; GFX10-NEXT:    s_waitcnt vmcnt(0)
+; GFX10-NEXT:    s_setpc_b64 s[4:5]
+  call amdgpu_gfx void @external_void_func_v4i16_inreg(<4 x i16> <i16 1, i16 2, i16 3, i16 4>)
+  ret void
+}
+
+define amdgpu_gfx void @test_call_external_void_func_v2f16_inreg() #0 {
+; GFX9-LABEL: test_call_external_void_func_v2f16_inreg:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    s_or_saveexec_b64 s[4:5], -1
+; GFX9-NEXT:    buffer_store_dword v40, off, s[0:3], s32 ; 4-byte Folded Spill
+; GFX9-NEXT:    s_mov_b64 exec, s[4:5]
+; GFX9-NEXT:    s_load_dword s4, s[4:5], 0x0
+; GFX9-NEXT:    v_writelane_b32 v40, s33, 2
+; GFX9-NEXT:    v_writelane_b32 v40, s30, 0
+; GFX9-NEXT:    s_mov_b32 s33, s32
+; GFX9-NEXT:    s_add_u32 s32, s32, 0x400
+; GFX9-NEXT:    s_getpc_b64 s[6:7]
+; GFX9-NEXT:    s_add_u32 s6, s6, external_void_func_v2f16_inreg@rel32@lo+4
+; GFX9-NEXT:    s_addc_u32 s7, s7, external_void_func_v2f16_inreg@rel32@hi+12
+; GFX9-NEXT:    v_writelane_b32 v40, s31, 1
+; GFX9-NEXT:    s_swappc_b64 s[30:31], s[6:7]
+; GFX9-NEXT:    v_readlane_b32 s4, v40, 0
+; GFX9-NEXT:    v_readlane_b32 s5, v40, 1
+; GFX9-NEXT:    s_sub_u32 s32, s32, 0x400
+; GFX9-NEXT:    v_readlane_b32 s33, v40, 2
+; GFX9-NEXT:    s_or_saveexec_b64 s[6:7], -1
+; GFX9-NEXT:    buffer_load_dword v40, off, s[0:3], s32 ; 4-byte Folded Reload
+; GFX9-NEXT:    s_mov_b64 exec, s[6:7]
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-NEXT:    s_setpc_b64 s[4:5]
+;
+; GFX10-LABEL: test_call_external_void_func_v2f16_inreg:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
+; GFX10-NEXT:    s_or_saveexec_b32 s4, -1
+; GFX10-NEXT:    buffer_store_dword v40, off, s[0:3], s32 ; 4-byte Folded Spill
+; GFX10-NEXT:    s_waitcnt_depctr 0xffe3
+; GFX10-NEXT:    s_mov_b32 exec_lo, s4
+; GFX10-NEXT:    s_load_dword s4, s[4:5], 0x0
+; GFX10-NEXT:    v_writelane_b32 v40, s33, 2
+; GFX10-NEXT:    s_mov_b32 s33, s32
+; GFX10-NEXT:    s_add_u32 s32, s32, 0x200
+; GFX10-NEXT:    s_getpc_b64 s[6:7]
+; GFX10-NEXT:    s_add_u32 s6, s6, external_void_func_v2f16_inreg@rel32@lo+4
+; GFX10-NEXT:    s_addc_u32 s7, s7, external_void_func_v2f16_inreg@rel32@hi+12
+; GFX10-NEXT:    ; implicit-def: $vcc_hi
+; GFX10-NEXT:    v_writelane_b32 v40, s30, 0
+; GFX10-NEXT:    v_writelane_b32 v40, s31, 1
+; GFX10-NEXT:    s_swappc_b64 s[30:31], s[6:7]
+; GFX10-NEXT:    v_readlane_b32 s4, v40, 0
+; GFX10-NEXT:    v_readlane_b32 s5, v40, 1
+; GFX10-NEXT:    s_sub_u32 s32, s32, 0x200
+; GFX10-NEXT:    v_readlane_b32 s33, v40, 2
+; GFX10-NEXT:    s_or_saveexec_b32 s6, -1
+; GFX10-NEXT:    buffer_load_dword v40, off, s[0:3], s32 ; 4-byte Folded Reload
+; GFX10-NEXT:    s_waitcnt_depctr 0xffe3
+; GFX10-NEXT:    s_mov_b32 exec_lo, s6
+; GFX10-NEXT:    s_waitcnt vmcnt(0)
+; GFX10-NEXT:    s_setpc_b64 s[4:5]
+  %val = load <2 x half>, <2 x half> addrspace(4)* undef
+  call amdgpu_gfx void @external_void_func_v2f16_inreg(<2 x half> %val)
+  ret void
+}
+
+define amdgpu_gfx void @test_call_external_void_func_v2i32_inreg() #0 {
+; GFX9-LABEL: test_call_external_void_func_v2i32_inreg:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    s_or_saveexec_b64 s[4:5], -1
+; GFX9-NEXT:    buffer_store_dword v40, off, s[0:3], s32 ; 4-byte Folded Spill
+; GFX9-NEXT:    s_mov_b64 exec, s[4:5]
+; GFX9-NEXT:    s_load_dwordx2 s[4:5], s[4:5], 0x0
+; GFX9-NEXT:    v_writelane_b32 v40, s33, 2
+; GFX9-NEXT:    v_writelane_b32 v40, s30, 0
+; GFX9-NEXT:    s_mov_b32 s33, s32
+; GFX9-NEXT:    s_add_u32 s32, s32, 0x400
+; GFX9-NEXT:    s_getpc_b64 s[6:7]
+; GFX9-NEXT:    s_add_u32 s6, s6, external_void_func_v2i32_inreg@rel32@lo+4
+; GFX9-NEXT:    s_addc_u32 s7, s7, external_void_func_v2i32_inreg@rel32@hi+12
+; GFX9-NEXT:    v_writelane_b32 v40, s31, 1
+; GFX9-NEXT:    s_swappc_b64 s[30:31], s[6:7]
+; GFX9-NEXT:    v_readlane_b32 s4, v40, 0
+; GFX9-NEXT:    v_readlane_b32 s5, v40, 1
+; GFX9-NEXT:    s_sub_u32 s32, s32, 0x400
+; GFX9-NEXT:    v_readlane_b32 s33, v40, 2
+; GFX9-NEXT:    s_or_saveexec_b64 s[6:7], -1
+; GFX9-NEXT:    buffer_load_dword v40, off, s[0:3], s32 ; 4-byte Folded Reload
+; GFX9-NEXT:    s_mov_b64 exec, s[6:7]
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-NEXT:    s_setpc_b64 s[4:5]
+;
+; GFX10-LABEL: test_call_external_void_func_v2i32_inreg:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
+; GFX10-NEXT:    s_or_saveexec_b32 s4, -1
+; GFX10-NEXT:    buffer_store_dword v40, off, s[0:3], s32 ; 4-byte Folded Spill
+; GFX10-NEXT:    s_waitcnt_depctr 0xffe3
+; GFX10-NEXT:    s_mov_b32 exec_lo, s4
+; GFX10-NEXT:    s_load_dwordx2 s[4:5], s[4:5], 0x0
+; GFX10-NEXT:    v_writelane_b32 v40, s33, 2
+; GFX10-NEXT:    s_mov_b32 s33, s32
+; GFX10-NEXT:    s_add_u32 s32, s32, 0x200
+; GFX10-NEXT:    s_getpc_b64 s[6:7]
+; GFX10-NEXT:    s_add_u32 s6, s6, external_void_func_v2i32_inreg@rel32@lo+4
+; GFX10-NEXT:    s_addc_u32 s7, s7, external_void_func_v2i32_inreg@rel32@hi+12
+; GFX10-NEXT:    ; implicit-def: $vcc_hi
+; GFX10-NEXT:    v_writelane_b32 v40, s30, 0
+; GFX10-NEXT:    v_writelane_b32 v40, s31, 1
+; GFX10-NEXT:    s_swappc_b64 s[30:31], s[6:7]
+; GFX10-NEXT:    v_readlane_b32 s4, v40, 0
+; GFX10-NEXT:    v_readlane_b32 s5, v40, 1
+; GFX10-NEXT:    s_sub_u32 s32, s32, 0x200
+; GFX10-NEXT:    v_readlane_b32 s33, v40, 2
+; GFX10-NEXT:    s_or_saveexec_b32 s6, -1
+; GFX10-NEXT:    buffer_load_dword v40, off, s[0:3], s32 ; 4-byte Folded Reload
+; GFX10-NEXT:    s_waitcnt_depctr 0xffe3
+; GFX10-NEXT:    s_mov_b32 exec_lo, s6
+; GFX10-NEXT:    s_waitcnt vmcnt(0)
+; GFX10-NEXT:    s_setpc_b64 s[4:5]
+  %val = load <2 x i32>, <2 x i32> addrspace(4)* undef
+  call amdgpu_gfx void @external_void_func_v2i32_inreg(<2 x i32> %val)
+  ret void
+}
+
+define amdgpu_gfx void @test_call_external_void_func_v2i32_imm_inreg() #0 {
+; GFX9-LABEL: test_call_external_void_func_v2i32_imm_inreg:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    s_or_saveexec_b64 s[4:5], -1
+; GFX9-NEXT:    buffer_store_dword v40, off, s[0:3], s32 ; 4-byte Folded Spill
+; GFX9-NEXT:    s_mov_b64 exec, s[4:5]
+; GFX9-NEXT:    v_writelane_b32 v40, s33, 2
+; GFX9-NEXT:    v_writelane_b32 v40, s30, 0
+; GFX9-NEXT:    s_mov_b32 s33, s32
+; GFX9-NEXT:    s_add_u32 s32, s32, 0x400
+; GFX9-NEXT:    s_mov_b32 s4, 1
+; GFX9-NEXT:    s_mov_b32 s5, 2
+; GFX9-NEXT:    s_getpc_b64 s[6:7]
+; GFX9-NEXT:    s_add_u32 s6, s6, external_void_func_v2i32_inreg@rel32@lo+4
+; GFX9-NEXT:    s_addc_u32 s7, s7, external_void_func_v2i32_inreg@rel32@hi+12
+; GFX9-NEXT:    v_writelane_b32 v40, s31, 1
+; GFX9-NEXT:    s_swappc_b64 s[30:31], s[6:7]
+; GFX9-NEXT:    v_readlane_b32 s4, v40, 0
+; GFX9-NEXT:    v_readlane_b32 s5, v40, 1
+; GFX9-NEXT:    s_sub_u32 s32, s32, 0x400
+; GFX9-NEXT:    v_readlane_b32 s33, v40, 2
+; GFX9-NEXT:    s_or_saveexec_b64 s[6:7], -1
+; GFX9-NEXT:    buffer_load_dword v40, off, s[0:3], s32 ; 4-byte Folded Reload
+; GFX9-NEXT:    s_mov_b64 exec, s[6:7]
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-NEXT:    s_setpc_b64 s[4:5]
+;
+; GFX10-LABEL: test_call_external_void_func_v2i32_imm_inreg:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
+; GFX10-NEXT:    s_or_saveexec_b32 s4, -1
+; GFX10-NEXT:    buffer_store_dword v40, off, s[0:3], s32 ; 4-byte Folded Spill
+; GFX10-NEXT:    s_waitcnt_depctr 0xffe3
+; GFX10-NEXT:    s_mov_b32 exec_lo, s4
+; GFX10-NEXT:    v_writelane_b32 v40, s33, 2
+; GFX10-NEXT:    s_mov_b32 s4, 1
+; GFX10-NEXT:    s_mov_b32 s5, 2
+; GFX10-NEXT:    s_mov_b32 s33, s32
+; GFX10-NEXT:    s_add_u32 s32, s32, 0x200
+; GFX10-NEXT:    v_writelane_b32 v40, s30, 0
+; GFX10-NEXT:    s_getpc_b64 s[6:7]
+; GFX10-NEXT:    s_add_u32 s6, s6, external_void_func_v2i32_inreg@rel32@lo+4
+; GFX10-NEXT:    s_addc_u32 s7, s7, external_void_func_v2i32_inreg@rel32@hi+12
+; GFX10-NEXT:    ; implicit-def: $vcc_hi
+; GFX10-NEXT:    v_writelane_b32 v40, s31, 1
+; GFX10-NEXT:    s_swappc_b64 s[30:31], s[6:7]
+; GFX10-NEXT:    v_readlane_b32 s4, v40, 0
+; GFX10-NEXT:    v_readlane_b32 s5, v40, 1
+; GFX10-NEXT:    s_sub_u32 s32, s32, 0x200
+; GFX10-NEXT:    v_readlane_b32 s33, v40, 2
+; GFX10-NEXT:    s_or_saveexec_b32 s6, -1
+; GFX10-NEXT:    buffer_load_dword v40, off, s[0:3], s32 ; 4-byte Folded Reload
+; GFX10-NEXT:    s_waitcnt_depctr 0xffe3
+; GFX10-NEXT:    s_mov_b32 exec_lo, s6
+; GFX10-NEXT:    s_waitcnt vmcnt(0)
+; GFX10-NEXT:    s_setpc_b64 s[4:5]
+  call amdgpu_gfx void @external_void_func_v2i32_inreg(<2 x i32> <i32 1, i32 2>)
+  ret void
+}
+
+define amdgpu_gfx void @test_call_external_void_func_v3i32_imm_inreg(i32) #0 {
+; GFX9-LABEL: test_call_external_void_func_v3i32_imm_inreg:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    s_or_saveexec_b64 s[4:5], -1
+; GFX9-NEXT:    buffer_store_dword v40, off, s[0:3], s32 ; 4-byte Folded Spill
+; GFX9-NEXT:    s_mov_b64 exec, s[4:5]
+; GFX9-NEXT:    v_writelane_b32 v40, s33, 2
+; GFX9-NEXT:    v_writelane_b32 v40, s30, 0
+; GFX9-NEXT:    s_mov_b32 s33, s32
+; GFX9-NEXT:    s_add_u32 s32, s32, 0x400
+; GFX9-NEXT:    s_mov_b32 s4, 3
+; GFX9-NEXT:    s_mov_b32 s5, 4
+; GFX9-NEXT:    s_mov_b32 s6, 5
+; GFX9-NEXT:    s_getpc_b64 s[8:9]
+; GFX9-NEXT:    s_add_u32 s8, s8, external_void_func_v3i32_inreg@rel32@lo+4
+; GFX9-NEXT:    s_addc_u32 s9, s9, external_void_func_v3i32_inreg@rel32@hi+12
+; GFX9-NEXT:    v_writelane_b32 v40, s31, 1
+; GFX9-NEXT:    s_swappc_b64 s[30:31], s[8:9]
+; GFX9-NEXT:    v_readlane_b32 s4, v40, 0
+; GFX9-NEXT:    v_readlane_b32 s5, v40, 1
+; GFX9-NEXT:    s_sub_u32 s32, s32, 0x400
+; GFX9-NEXT:    v_readlane_b32 s33, v40, 2
+; GFX9-NEXT:    s_or_saveexec_b64 s[6:7], -1
+; GFX9-NEXT:    buffer_load_dword v40, off, s[0:3], s32 ; 4-byte Folded Reload
+; GFX9-NEXT:    s_mov_b64 exec, s[6:7]
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-NEXT:    s_setpc_b64 s[4:5]
+;
+; GFX10-LABEL: test_call_external_void_func_v3i32_imm_inreg:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
+; GFX10-NEXT:    s_or_saveexec_b32 s4, -1
+; GFX10-NEXT:    buffer_store_dword v40, off, s[0:3], s32 ; 4-byte Folded Spill
+; GFX10-NEXT:    s_waitcnt_depctr 0xffe3
+; GFX10-NEXT:    s_mov_b32 exec_lo, s4
+; GFX10-NEXT:    v_writelane_b32 v40, s33, 2
+; GFX10-NEXT:    s_mov_b32 s4, 3
+; GFX10-NEXT:    s_mov_b32 s5, 4
+; GFX10-NEXT:    s_mov_b32 s6, 5
+; GFX10-NEXT:    s_mov_b32 s33, s32
+; GFX10-NEXT:    v_writelane_b32 v40, s30, 0
+; GFX10-NEXT:    s_add_u32 s32, s32, 0x200
+; GFX10-NEXT:    s_getpc_b64 s[8:9]
+; GFX10-NEXT:    s_add_u32 s8, s8, external_void_func_v3i32_inreg@rel32@lo+4
+; GFX10-NEXT:    s_addc_u32 s9, s9, external_void_func_v3i32_inreg@rel32@hi+12
+; GFX10-NEXT:    ; implicit-def: $vcc_hi
+; GFX10-NEXT:    v_writelane_b32 v40, s31, 1
+; GFX10-NEXT:    s_swappc_b64 s[30:31], s[8:9]
+; GFX10-NEXT:    v_readlane_b32 s4, v40, 0
+; GFX10-NEXT:    v_readlane_b32 s5, v40, 1
+; GFX10-NEXT:    s_sub_u32 s32, s32, 0x200
+; GFX10-NEXT:    v_readlane_b32 s33, v40, 2
+; GFX10-NEXT:    s_or_saveexec_b32 s6, -1
+; GFX10-NEXT:    buffer_load_dword v40, off, s[0:3], s32 ; 4-byte Folded Reload
+; GFX10-NEXT:    s_waitcnt_depctr 0xffe3
+; GFX10-NEXT:    s_mov_b32 exec_lo, s6
+; GFX10-NEXT:    s_waitcnt vmcnt(0)
+; GFX10-NEXT:    s_setpc_b64 s[4:5]
+  call amdgpu_gfx void @external_void_func_v3i32_inreg(<3 x i32> <i32 3, i32 4, i32 5>)
+  ret void
+}
+
+define amdgpu_gfx void @test_call_external_void_func_v3i32_i32_inreg(i32) #0 {
+; GFX9-LABEL: test_call_external_void_func_v3i32_i32_inreg:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    s_or_saveexec_b64 s[4:5], -1
+; GFX9-NEXT:    buffer_store_dword v40, off, s[0:3], s32 ; 4-byte Folded Spill
+; GFX9-NEXT:    s_mov_b64 exec, s[4:5]
+; GFX9-NEXT:    v_writelane_b32 v40, s33, 2
+; GFX9-NEXT:    v_writelane_b32 v40, s30, 0
+; GFX9-NEXT:    s_mov_b32 s33, s32
+; GFX9-NEXT:    s_add_u32 s32, s32, 0x400
+; GFX9-NEXT:    s_mov_b32 s4, 3
+; GFX9-NEXT:    s_mov_b32 s5, 4
+; GFX9-NEXT:    s_mov_b32 s6, 5
+; GFX9-NEXT:    s_mov_b32 s7, 6
+; GFX9-NEXT:    s_getpc_b64 s[8:9]
+; GFX9-NEXT:    s_add_u32 s8, s8, external_void_func_v3i32_i32_inreg@rel32@lo+4
+; GFX9-NEXT:    s_addc_u32 s9, s9, external_void_func_v3i32_i32_inreg@rel32@hi+12
+; GFX9-NEXT:    v_writelane_b32 v40, s31, 1
+; GFX9-NEXT:    s_swappc_b64 s[30:31], s[8:9]
+; GFX9-NEXT:    v_readlane_b32 s4, v40, 0
+; GFX9-NEXT:    v_readlane_b32 s5, v40, 1
+; GFX9-NEXT:    s_sub_u32 s32, s32, 0x400
+; GFX9-NEXT:    v_readlane_b32 s33, v40, 2
+; GFX9-NEXT:    s_or_saveexec_b64 s[6:7], -1
+; GFX9-NEXT:    buffer_load_dword v40, off, s[0:3], s32 ; 4-byte Folded Reload
+; GFX9-NEXT:    s_mov_b64 exec, s[6:7]
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-NEXT:    s_setpc_b64 s[4:5]
+;
+; GFX10-LABEL: test_call_external_void_func_v3i32_i32_inreg:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
+; GFX10-NEXT:    s_or_saveexec_b32 s4, -1
+; GFX10-NEXT:    buffer_store_dword v40, off, s[0:3], s32 ; 4-byte Folded Spill
+; GFX10-NEXT:    s_waitcnt_depctr 0xffe3
+; GFX10-NEXT:    s_mov_b32 exec_lo, s4
+; GFX10-NEXT:    v_writelane_b32 v40, s33, 2
+; GFX10-NEXT:    s_mov_b32 s4, 3
+; GFX10-NEXT:    s_mov_b32 s5, 4
+; GFX10-NEXT:    s_mov_b32 s6, 5
+; GFX10-NEXT:    s_mov_b32 s7, 6
+; GFX10-NEXT:    v_writelane_b32 v40, s30, 0
+; GFX10-NEXT:    s_mov_b32 s33, s32
+; GFX10-NEXT:    s_add_u32 s32, s32, 0x200
+; GFX10-NEXT:    s_getpc_b64 s[8:9]
+; GFX10-NEXT:    s_add_u32 s8, s8, external_void_func_v3i32_i32_inreg@rel32@lo+4
+; GFX10-NEXT:    s_addc_u32 s9, s9, external_void_func_v3i32_i32_inreg@rel32@hi+12
+; GFX10-NEXT:    ; implicit-def: $vcc_hi
+; GFX10-NEXT:    v_writelane_b32 v40, s31, 1
+; GFX10-NEXT:    s_swappc_b64 s[30:31], s[8:9]
+; GFX10-NEXT:    v_readlane_b32 s4, v40, 0
+; GFX10-NEXT:    v_readlane_b32 s5, v40, 1
+; GFX10-NEXT:    s_sub_u32 s32, s32, 0x200
+; GFX10-NEXT:    v_readlane_b32 s33, v40, 2
+; GFX10-NEXT:    s_or_saveexec_b32 s6, -1
+; GFX10-NEXT:    buffer_load_dword v40, off, s[0:3], s32 ; 4-byte Folded Reload
+; GFX10-NEXT:    s_waitcnt_depctr 0xffe3
+; GFX10-NEXT:    s_mov_b32 exec_lo, s6
+; GFX10-NEXT:    s_waitcnt vmcnt(0)
+; GFX10-NEXT:    s_setpc_b64 s[4:5]
+  call amdgpu_gfx void @external_void_func_v3i32_i32_inreg(<3 x i32> <i32 3, i32 4, i32 5>, i32 6)
+  ret void
+}
+
+define amdgpu_gfx void @test_call_external_void_func_v4i32_inreg() #0 {
+; GFX9-LABEL: test_call_external_void_func_v4i32_inreg:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    s_or_saveexec_b64 s[4:5], -1
+; GFX9-NEXT:    buffer_store_dword v40, off, s[0:3], s32 ; 4-byte Folded Spill
+; GFX9-NEXT:    s_mov_b64 exec, s[4:5]
+; GFX9-NEXT:    s_load_dwordx4 s[4:7], s[4:5], 0x0
+; GFX9-NEXT:    v_writelane_b32 v40, s33, 2
+; GFX9-NEXT:    v_writelane_b32 v40, s30, 0
+; GFX9-NEXT:    s_mov_b32 s33, s32
+; GFX9-NEXT:    s_add_u32 s32, s32, 0x400
+; GFX9-NEXT:    s_getpc_b64 s[8:9]
+; GFX9-NEXT:    s_add_u32 s8, s8, external_void_func_v4i32_inreg@rel32@lo+4
+; GFX9-NEXT:    s_addc_u32 s9, s9, external_void_func_v4i32_inreg@rel32@hi+12
+; GFX9-NEXT:    v_writelane_b32 v40, s31, 1
+; GFX9-NEXT:    s_swappc_b64 s[30:31], s[8:9]
+; GFX9-NEXT:    v_readlane_b32 s4, v40, 0
+; GFX9-NEXT:    v_readlane_b32 s5, v40, 1
+; GFX9-NEXT:    s_sub_u32 s32, s32, 0x400
+; GFX9-NEXT:    v_readlane_b32 s33, v40, 2
+; GFX9-NEXT:    s_or_saveexec_b64 s[6:7], -1
+; GFX9-NEXT:    buffer_load_dword v40, off, s[0:3], s32 ; 4-byte Folded Reload
+; GFX9-NEXT:    s_mov_b64 exec, s[6:7]
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-NEXT:    s_setpc_b64 s[4:5]
+;
+; GFX10-LABEL: test_call_external_void_func_v4i32_inreg:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
+; GFX10-NEXT:    s_or_saveexec_b32 s4, -1
+; GFX10-NEXT:    buffer_store_dword v40, off, s[0:3], s32 ; 4-byte Folded Spill
+; GFX10-NEXT:    s_waitcnt_depctr 0xffe3
+; GFX10-NEXT:    s_mov_b32 exec_lo, s4
+; GFX10-NEXT:    s_load_dwordx4 s[4:7], s[4:5], 0x0
+; GFX10-NEXT:    v_writelane_b32 v40, s33, 2
+; GFX10-NEXT:    s_mov_b32 s33, s32
+; GFX10-NEXT:    s_add_u32 s32, s32, 0x200
+; GFX10-NEXT:    s_getpc_b64 s[8:9]
+; GFX10-NEXT:    s_add_u32 s8, s8, external_void_func_v4i32_inreg@rel32@lo+4
+; GFX10-NEXT:    s_addc_u32 s9, s9, external_void_func_v4i32_inreg@rel32@hi+12
+; GFX10-NEXT:    ; implicit-def: $vcc_hi
+; GFX10-NEXT:    v_writelane_b32 v40, s30, 0
+; GFX10-NEXT:    v_writelane_b32 v40, s31, 1
+; GFX10-NEXT:    s_swappc_b64 s[30:31], s[8:9]
+; GFX10-NEXT:    v_readlane_b32 s4, v40, 0
+; GFX10-NEXT:    v_readlane_b32 s5, v40, 1
+; GFX10-NEXT:    s_sub_u32 s32, s32, 0x200
+; GFX10-NEXT:    v_readlane_b32 s33, v40, 2
+; GFX10-NEXT:    s_or_saveexec_b32 s6, -1
+; GFX10-NEXT:    buffer_load_dword v40, off, s[0:3], s32 ; 4-byte Folded Reload
+; GFX10-NEXT:    s_waitcnt_depctr 0xffe3
+; GFX10-NEXT:    s_mov_b32 exec_lo, s6
+; GFX10-NEXT:    s_waitcnt vmcnt(0)
+; GFX10-NEXT:    s_setpc_b64 s[4:5]
+  %val = load <4 x i32>, <4 x i32> addrspace(4)* undef
+  call amdgpu_gfx void @external_void_func_v4i32_inreg(<4 x i32> %val)
+  ret void
+}
+
+define amdgpu_gfx void @test_call_external_void_func_v4i32_imm_inreg() #0 {
+; GFX9-LABEL: test_call_external_void_func_v4i32_imm_inreg:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    s_or_saveexec_b64 s[4:5], -1
+; GFX9-NEXT:    buffer_store_dword v40, off, s[0:3], s32 ; 4-byte Folded Spill
+; GFX9-NEXT:    s_mov_b64 exec, s[4:5]
+; GFX9-NEXT:    v_writelane_b32 v40, s33, 2
+; GFX9-NEXT:    v_writelane_b32 v40, s30, 0
+; GFX9-NEXT:    s_mov_b32 s33, s32
+; GFX9-NEXT:    s_add_u32 s32, s32, 0x400
+; GFX9-NEXT:    s_mov_b32 s4, 1
+; GFX9-NEXT:    s_mov_b32 s5, 2
+; GFX9-NEXT:    s_mov_b32 s6, 3
+; GFX9-NEXT:    s_mov_b32 s7, 4
+; GFX9-NEXT:    s_getpc_b64 s[8:9]
+; GFX9-NEXT:    s_add_u32 s8, s8, external_void_func_v4i32_inreg@rel32@lo+4
+; GFX9-NEXT:    s_addc_u32 s9, s9, external_void_func_v4i32_inreg@rel32@hi+12
+; GFX9-NEXT:    v_writelane_b32 v40, s31, 1
+; GFX9-NEXT:    s_swappc_b64 s[30:31], s[8:9]
+; GFX9-NEXT:    v_readlane_b32 s4, v40, 0
+; GFX9-NEXT:    v_readlane_b32 s5, v40, 1
+; GFX9-NEXT:    s_sub_u32 s32, s32, 0x400
+; GFX9-NEXT:    v_readlane_b32 s33, v40, 2
+; GFX9-NEXT:    s_or_saveexec_b64 s[6:7], -1
+; GFX9-NEXT:    buffer_load_dword v40, off, s[0:3], s32 ; 4-byte Folded Reload
+; GFX9-NEXT:    s_mov_b64 exec, s[6:7]
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-NEXT:    s_setpc_b64 s[4:5]
+;
+; GFX10-LABEL: test_call_external_void_func_v4i32_imm_inreg:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
+; GFX10-NEXT:    s_or_saveexec_b32 s4, -1
+; GFX10-NEXT:    buffer_store_dword v40, off, s[0:3], s32 ; 4-byte Folded Spill
+; GFX10-NEXT:    s_waitcnt_depctr 0xffe3
+; GFX10-NEXT:    s_mov_b32 exec_lo, s4
+; GFX10-NEXT:    v_writelane_b32 v40, s33, 2
+; GFX10-NEXT:    s_mov_b32 s4, 1
+; GFX10-NEXT:    s_mov_b32 s5, 2
+; GFX10-NEXT:    s_mov_b32 s6, 3
+; GFX10-NEXT:    s_mov_b32 s7, 4
+; GFX10-NEXT:    v_writelane_b32 v40, s30, 0
+; GFX10-NEXT:    s_mov_b32 s33, s32
+; GFX10-NEXT:    s_add_u32 s32, s32, 0x200
+; GFX10-NEXT:    s_getpc_b64 s[8:9]
+; GFX10-NEXT:    s_add_u32 s8, s8, external_void_func_v4i32_inreg@rel32@lo+4
+; GFX10-NEXT:    s_addc_u32 s9, s9, external_void_func_v4i32_inreg@rel32@hi+12
+; GFX10-NEXT:    ; implicit-def: $vcc_hi
+; GFX10-NEXT:    v_writelane_b32 v40, s31, 1
+; GFX10-NEXT:    s_swappc_b64 s[30:31], s[8:9]
+; GFX10-NEXT:    v_readlane_b32 s4, v40, 0
+; GFX10-NEXT:    v_readlane_b32 s5, v40, 1
+; GFX10-NEXT:    s_sub_u32 s32, s32, 0x200
+; GFX10-NEXT:    v_readlane_b32 s33, v40, 2
+; GFX10-NEXT:    s_or_saveexec_b32 s6, -1
+; GFX10-NEXT:    buffer_load_dword v40, off, s[0:3], s32 ; 4-byte Folded Reload
+; GFX10-NEXT:    s_waitcnt_depctr 0xffe3
+; GFX10-NEXT:    s_mov_b32 exec_lo, s6
+; GFX10-NEXT:    s_waitcnt vmcnt(0)
+; GFX10-NEXT:    s_setpc_b64 s[4:5]
+  call amdgpu_gfx void @external_void_func_v4i32_inreg(<4 x i32> <i32 1, i32 2, i32 3, i32 4>)
+  ret void
+}
+
+define amdgpu_gfx void @test_call_external_void_func_v5i32_imm_inreg() #0 {
+; GFX9-LABEL: test_call_external_void_func_v5i32_imm_inreg:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    s_or_saveexec_b64 s[4:5], -1
+; GFX9-NEXT:    buffer_store_dword v40, off, s[0:3], s32 ; 4-byte Folded Spill
+; GFX9-NEXT:    s_mov_b64 exec, s[4:5]
+; GFX9-NEXT:    v_writelane_b32 v40, s33, 2
+; GFX9-NEXT:    v_writelane_b32 v40, s30, 0
+; GFX9-NEXT:    s_mov_b32 s33, s32
+; GFX9-NEXT:    s_add_u32 s32, s32, 0x400
+; GFX9-NEXT:    s_mov_b32 s4, 1
+; GFX9-NEXT:    s_mov_b32 s5, 2
+; GFX9-NEXT:    s_mov_b32 s6, 3
+; GFX9-NEXT:    s_mov_b32 s7, 4
+; GFX9-NEXT:    s_mov_b32 s8, 5
+; GFX9-NEXT:    s_getpc_b64 s[10:11]
+; GFX9-NEXT:    s_add_u32 s10, s10, external_void_func_v5i32_inreg@rel32@lo+4
+; GFX9-NEXT:    s_addc_u32 s11, s11, external_void_func_v5i32_inreg@rel32@hi+12
+; GFX9-NEXT:    v_writelane_b32 v40, s31, 1
+; GFX9-NEXT:    s_swappc_b64 s[30:31], s[10:11]
+; GFX9-NEXT:    v_readlane_b32 s4, v40, 0
+; GFX9-NEXT:    v_readlane_b32 s5, v40, 1
+; GFX9-NEXT:    s_sub_u32 s32, s32, 0x400
+; GFX9-NEXT:    v_readlane_b32 s33, v40, 2
+; GFX9-NEXT:    s_or_saveexec_b64 s[6:7], -1
+; GFX9-NEXT:    buffer_load_dword v40, off, s[0:3], s32 ; 4-byte Folded Reload
+; GFX9-NEXT:    s_mov_b64 exec, s[6:7]
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-NEXT:    s_setpc_b64 s[4:5]
+;
+; GFX10-LABEL: test_call_external_void_func_v5i32_imm_inreg:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
+; GFX10-NEXT:    s_or_saveexec_b32 s4, -1
+; GFX10-NEXT:    buffer_store_dword v40, off, s[0:3], s32 ; 4-byte Folded Spill
+; GFX10-NEXT:    s_waitcnt_depctr 0xffe3
+; GFX10-NEXT:    s_mov_b32 exec_lo, s4
+; GFX10-NEXT:    v_writelane_b32 v40, s33, 2
+; GFX10-NEXT:    s_mov_b32 s4, 1
+; GFX10-NEXT:    s_mov_b32 s5, 2
+; GFX10-NEXT:    s_mov_b32 s6, 3
+; GFX10-NEXT:    s_mov_b32 s7, 4
+; GFX10-NEXT:    v_writelane_b32 v40, s30, 0
+; GFX10-NEXT:    s_mov_b32 s8, 5
+; GFX10-NEXT:    s_mov_b32 s33, s32
+; GFX10-NEXT:    s_add_u32 s32, s32, 0x200
+; GFX10-NEXT:    s_getpc_b64 s[10:11]
+; GFX10-NEXT:    s_add_u32 s10, s10, external_void_func_v5i32_inreg@rel32@lo+4
+; GFX10-NEXT:    s_addc_u32 s11, s11, external_void_func_v5i32_inreg@rel32@hi+12
+; GFX10-NEXT:    v_writelane_b32 v40, s31, 1
+; GFX10-NEXT:    ; implicit-def: $vcc_hi
+; GFX10-NEXT:    s_swappc_b64 s[30:31], s[10:11]
+; GFX10-NEXT:    v_readlane_b32 s4, v40, 0
+; GFX10-NEXT:    v_readlane_b32 s5, v40, 1
+; GFX10-NEXT:    s_sub_u32 s32, s32, 0x200
+; GFX10-NEXT:    v_readlane_b32 s33, v40, 2
+; GFX10-NEXT:    s_or_saveexec_b32 s6, -1
+; GFX10-NEXT:    buffer_load_dword v40, off, s[0:3], s32 ; 4-byte Folded Reload
+; GFX10-NEXT:    s_waitcnt_depctr 0xffe3
+; GFX10-NEXT:    s_mov_b32 exec_lo, s6
+; GFX10-NEXT:    s_waitcnt vmcnt(0)
+; GFX10-NEXT:    s_setpc_b64 s[4:5]
+  call amdgpu_gfx void @external_void_func_v5i32_inreg(<5 x i32> <i32 1, i32 2, i32 3, i32 4, i32 5>)
+  ret void
+}
+
+define amdgpu_gfx void @test_call_external_void_func_v8i32_inreg() #0 {
+; GFX9-LABEL: test_call_external_void_func_v8i32_inreg:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    s_or_saveexec_b64 s[4:5], -1
+; GFX9-NEXT:    buffer_store_dword v40, off, s[0:3], s32 ; 4-byte Folded Spill
+; GFX9-NEXT:    s_mov_b64 exec, s[4:5]
+; GFX9-NEXT:    s_load_dwordx2 s[4:5], s[4:5], 0x0
+; GFX9-NEXT:    v_writelane_b32 v40, s33, 2
+; GFX9-NEXT:    v_writelane_b32 v40, s30, 0
+; GFX9-NEXT:    s_mov_b32 s33, s32
+; GFX9-NEXT:    s_add_u32 s32, s32, 0x400
+; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX9-NEXT:    s_load_dwordx8 s[4:11], s[4:5], 0x0
+; GFX9-NEXT:    s_getpc_b64 s[12:13]
+; GFX9-NEXT:    s_add_u32 s12, s12, external_void_func_v8i32_inreg@rel32@lo+4
+; GFX9-NEXT:    s_addc_u32 s13, s13, external_void_func_v8i32_inreg@rel32@hi+12
+; GFX9-NEXT:    v_writelane_b32 v40, s31, 1
+; GFX9-NEXT:    s_swappc_b64 s[30:31], s[12:13]
+; GFX9-NEXT:    v_readlane_b32 s4, v40, 0
+; GFX9-NEXT:    v_readlane_b32 s5, v40, 1
+; GFX9-NEXT:    s_sub_u32 s32, s32, 0x400
+; GFX9-NEXT:    v_readlane_b32 s33, v40, 2
+; GFX9-NEXT:    s_or_saveexec_b64 s[6:7], -1
+; GFX9-NEXT:    buffer_load_dword v40, off, s[0:3], s32 ; 4-byte Folded Reload
+; GFX9-NEXT:    s_mov_b64 exec, s[6:7]
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-NEXT:    s_setpc_b64 s[4:5]
+;
+; GFX10-LABEL: test_call_external_void_func_v8i32_inreg:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
+; GFX10-NEXT:    s_or_saveexec_b32 s4, -1
+; GFX10-NEXT:    buffer_store_dword v40, off, s[0:3], s32 ; 4-byte Folded Spill
+; GFX10-NEXT:    s_waitcnt_depctr 0xffe3
+; GFX10-NEXT:    s_mov_b32 exec_lo, s4
+; GFX10-NEXT:    s_load_dwordx2 s[4:5], s[4:5], 0x0
+; GFX10-NEXT:    v_writelane_b32 v40, s33, 2
+; GFX10-NEXT:    s_mov_b32 s33, s32
+; GFX10-NEXT:    s_add_u32 s32, s32, 0x200
+; GFX10-NEXT:    ; implicit-def: $vcc_hi
+; GFX10-NEXT:    s_getpc_b64 s[12:13]
+; GFX10-NEXT:    s_add_u32 s12, s12, external_void_func_v8i32_inreg@rel32@lo+4
+; GFX10-NEXT:    s_addc_u32 s13, s13, external_void_func_v8i32_inreg@rel32@hi+12
+; GFX10-NEXT:    v_writelane_b32 v40, s30, 0
+; GFX10-NEXT:    v_writelane_b32 v40, s31, 1
+; GFX10-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-NEXT:    s_load_dwordx8 s[4:11], s[4:5], 0x0
+; GFX10-NEXT:    s_swappc_b64 s[30:31], s[12:13]
+; GFX10-NEXT:    v_readlane_b32 s4, v40, 0
+; GFX10-NEXT:    v_readlane_b32 s5, v40, 1
+; GFX10-NEXT:    s_sub_u32 s32, s32, 0x200
+; GFX10-NEXT:    v_readlane_b32 s33, v40, 2
+; GFX10-NEXT:    s_or_saveexec_b32 s6, -1
+; GFX10-NEXT:    buffer_load_dword v40, off, s[0:3], s32 ; 4-byte Folded Reload
+; GFX10-NEXT:    s_waitcnt_depctr 0xffe3
+; GFX10-NEXT:    s_mov_b32 exec_lo, s6
+; GFX10-NEXT:    s_waitcnt vmcnt(0)
+; GFX10-NEXT:    s_setpc_b64 s[4:5]
+  %ptr = load <8 x i32> addrspace(4)*, <8 x i32> addrspace(4)* addrspace(4)* undef
+  %val = load <8 x i32>, <8 x i32> addrspace(4)* %ptr
+  call amdgpu_gfx void @external_void_func_v8i32_inreg(<8 x i32> %val)
+  ret void
+}
+
+define amdgpu_gfx void @test_call_external_void_func_v8i32_imm_inreg() #0 {
+; GFX9-LABEL: test_call_external_void_func_v8i32_imm_inreg:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    s_or_saveexec_b64 s[4:5], -1
+; GFX9-NEXT:    buffer_store_dword v40, off, s[0:3], s32 ; 4-byte Folded Spill
+; GFX9-NEXT:    s_mov_b64 exec, s[4:5]
+; GFX9-NEXT:    v_writelane_b32 v40, s33, 2
+; GFX9-NEXT:    v_writelane_b32 v40, s30, 0
+; GFX9-NEXT:    s_mov_b32 s33, s32
+; GFX9-NEXT:    s_add_u32 s32, s32, 0x400
+; GFX9-NEXT:    s_mov_b32 s4, 1
+; GFX9-NEXT:    s_mov_b32 s5, 2
+; GFX9-NEXT:    s_mov_b32 s6, 3
+; GFX9-NEXT:    s_mov_b32 s7, 4
+; GFX9-NEXT:    s_mov_b32 s8, 5
+; GFX9-NEXT:    s_mov_b32 s9, 6
+; GFX9-NEXT:    s_mov_b32 s10, 7
+; GFX9-NEXT:    s_mov_b32 s11, 8
+; GFX9-NEXT:    s_getpc_b64 s[12:13]
+; GFX9-NEXT:    s_add_u32 s12, s12, external_void_func_v8i32_inreg@rel32@lo+4
+; GFX9-NEXT:    s_addc_u32 s13, s13, external_void_func_v8i32_inreg@rel32@hi+12
+; GFX9-NEXT:    v_writelane_b32 v40, s31, 1
+; GFX9-NEXT:    s_swappc_b64 s[30:31], s[12:13]
+; GFX9-NEXT:    v_readlane_b32 s4, v40, 0
+; GFX9-NEXT:    v_readlane_b32 s5, v40, 1
+; GFX9-NEXT:    s_sub_u32 s32, s32, 0x400
+; GFX9-NEXT:    v_readlane_b32 s33, v40, 2
+; GFX9-NEXT:    s_or_saveexec_b64 s[6:7], -1
+; GFX9-NEXT:    buffer_load_dword v40, off, s[0:3], s32 ; 4-byte Folded Reload
+; GFX9-NEXT:    s_mov_b64 exec, s[6:7]
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-NEXT:    s_setpc_b64 s[4:5]
+;
+; GFX10-LABEL: test_call_external_void_func_v8i32_imm_inreg:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
+; GFX10-NEXT:    s_or_saveexec_b32 s4, -1
+; GFX10-NEXT:    buffer_store_dword v40, off, s[0:3], s32 ; 4-byte Folded Spill
+; GFX10-NEXT:    s_waitcnt_depctr 0xffe3
+; GFX10-NEXT:    s_mov_b32 exec_lo, s4
+; GFX10-NEXT:    v_writelane_b32 v40, s33, 2
+; GFX10-NEXT:    s_mov_b32 s4, 1
+; GFX10-NEXT:    s_mov_b32 s5, 2
+; GFX10-NEXT:    s_mov_b32 s6, 3
+; GFX10-NEXT:    s_mov_b32 s7, 4
+; GFX10-NEXT:    v_writelane_b32 v40, s30, 0
+; GFX10-NEXT:    s_mov_b32 s8, 5
+; GFX10-NEXT:    s_mov_b32 s9, 6
+; GFX10-NEXT:    s_mov_b32 s10, 7
+; GFX10-NEXT:    s_mov_b32 s11, 8
+; GFX10-NEXT:    s_mov_b32 s33, s32
+; GFX10-NEXT:    s_add_u32 s32, s32, 0x200
+; GFX10-NEXT:    s_getpc_b64 s[12:13]
+; GFX10-NEXT:    s_add_u32 s12, s12, external_void_func_v8i32_inreg@rel32@lo+4
+; GFX10-NEXT:    s_addc_u32 s13, s13, external_void_func_v8i32_inreg@rel32@hi+12
+; GFX10-NEXT:    v_writelane_b32 v40, s31, 1
+; GFX10-NEXT:    ; implicit-def: $vcc_hi
+; GFX10-NEXT:    s_swappc_b64 s[30:31], s[12:13]
+; GFX10-NEXT:    v_readlane_b32 s4, v40, 0
+; GFX10-NEXT:    v_readlane_b32 s5, v40, 1
+; GFX10-NEXT:    s_sub_u32 s32, s32, 0x200
+; GFX10-NEXT:    v_readlane_b32 s33, v40, 2
+; GFX10-NEXT:    s_or_saveexec_b32 s6, -1
+; GFX10-NEXT:    buffer_load_dword v40, off, s[0:3], s32 ; 4-byte Folded Reload
+; GFX10-NEXT:    s_waitcnt_depctr 0xffe3
+; GFX10-NEXT:    s_mov_b32 exec_lo, s6
+; GFX10-NEXT:    s_waitcnt vmcnt(0)
+; GFX10-NEXT:    s_setpc_b64 s[4:5]
+  call amdgpu_gfx void @external_void_func_v8i32_inreg(<8 x i32> <i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8>)
+  ret void
+}
+
+define amdgpu_gfx void @test_call_external_void_func_v16i32_inreg() #0 {
+; GFX9-LABEL: test_call_external_void_func_v16i32_inreg:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    s_or_saveexec_b64 s[4:5], -1
+; GFX9-NEXT:    buffer_store_dword v40, off, s[0:3], s32 ; 4-byte Folded Spill
+; GFX9-NEXT:    s_mov_b64 exec, s[4:5]
+; GFX9-NEXT:    s_load_dwordx2 s[4:5], s[4:5], 0x0
+; GFX9-NEXT:    v_writelane_b32 v40, s33, 2
+; GFX9-NEXT:    v_writelane_b32 v40, s30, 0
+; GFX9-NEXT:    s_mov_b32 s33, s32
+; GFX9-NEXT:    s_add_u32 s32, s32, 0x400
+; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX9-NEXT:    s_load_dwordx16 s[4:19], s[4:5], 0x0
+; GFX9-NEXT:    s_getpc_b64 s[20:21]
+; GFX9-NEXT:    s_add_u32 s20, s20, external_void_func_v16i32_inreg@rel32@lo+4
+; GFX9-NEXT:    s_addc_u32 s21, s21, external_void_func_v16i32_inreg@rel32@hi+12
+; GFX9-NEXT:    v_writelane_b32 v40, s31, 1
+; GFX9-NEXT:    s_swappc_b64 s[30:31], s[20:21]
+; GFX9-NEXT:    v_readlane_b32 s4, v40, 0
+; GFX9-NEXT:    v_readlane_b32 s5, v40, 1
+; GFX9-NEXT:    s_sub_u32 s32, s32, 0x400
+; GFX9-NEXT:    v_readlane_b32 s33, v40, 2
+; GFX9-NEXT:    s_or_saveexec_b64 s[6:7], -1
+; GFX9-NEXT:    buffer_load_dword v40, off, s[0:3], s32 ; 4-byte Folded Reload
+; GFX9-NEXT:    s_mov_b64 exec, s[6:7]
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-NEXT:    s_setpc_b64 s[4:5]
+;
+; GFX10-LABEL: test_call_external_void_func_v16i32_inreg:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
+; GFX10-NEXT:    s_or_saveexec_b32 s4, -1
+; GFX10-NEXT:    buffer_store_dword v40, off, s[0:3], s32 ; 4-byte Folded Spill
+; GFX10-NEXT:    s_waitcnt_depctr 0xffe3
+; GFX10-NEXT:    s_mov_b32 exec_lo, s4
+; GFX10-NEXT:    s_load_dwordx2 s[4:5], s[4:5], 0x0
+; GFX10-NEXT:    v_writelane_b32 v40, s33, 2
+; GFX10-NEXT:    s_mov_b32 s33, s32
+; GFX10-NEXT:    s_add_u32 s32, s32, 0x200
+; GFX10-NEXT:    ; implicit-def: $vcc_hi
+; GFX10-NEXT:    s_getpc_b64 s[20:21]
+; GFX10-NEXT:    s_add_u32 s20, s20, external_void_func_v16i32_inreg@rel32@lo+4
+; GFX10-NEXT:    s_addc_u32 s21, s21, external_void_func_v16i32_inreg@rel32@hi+12
+; GFX10-NEXT:    v_writelane_b32 v40, s30, 0
+; GFX10-NEXT:    v_writelane_b32 v40, s31, 1
+; GFX10-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-NEXT:    s_load_dwordx16 s[4:19], s[4:5], 0x0
+; GFX10-NEXT:    s_swappc_b64 s[30:31], s[20:21]
+; GFX10-NEXT:    v_readlane_b32 s4, v40, 0
+; GFX10-NEXT:    v_readlane_b32 s5, v40, 1
+; GFX10-NEXT:    s_sub_u32 s32, s32, 0x200
+; GFX10-NEXT:    v_readlane_b32 s33, v40, 2
+; GFX10-NEXT:    s_or_saveexec_b32 s6, -1
+; GFX10-NEXT:    buffer_load_dword v40, off, s[0:3], s32 ; 4-byte Folded Reload
+; GFX10-NEXT:    s_waitcnt_depctr 0xffe3
+; GFX10-NEXT:    s_mov_b32 exec_lo, s6
+; GFX10-NEXT:    s_waitcnt vmcnt(0)
+; GFX10-NEXT:    s_setpc_b64 s[4:5]
+  %ptr = load <16 x i32> addrspace(4)*, <16 x i32> addrspace(4)* addrspace(4)* undef
+  %val = load <16 x i32>, <16 x i32> addrspace(4)* %ptr
+  call amdgpu_gfx void @external_void_func_v16i32_inreg(<16 x i32> %val)
+  ret void
+}
+
+define amdgpu_gfx void @test_call_external_void_func_v32i32_inreg() #0 {
+; GFX9-LABEL: test_call_external_void_func_v32i32_inreg:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    s_or_saveexec_b64 s[4:5], -1
+; GFX9-NEXT:    buffer_store_dword v40, off, s[0:3], s32 ; 4-byte Folded Spill
+; GFX9-NEXT:    s_mov_b64 exec, s[4:5]
+; GFX9-NEXT:    v_writelane_b32 v40, s33, 18
+; GFX9-NEXT:    v_writelane_b32 v40, s36, 0
+; GFX9-NEXT:    v_writelane_b32 v40, s37, 1
+; GFX9-NEXT:    v_writelane_b32 v40, s38, 2
+; GFX9-NEXT:    v_writelane_b32 v40, s39, 3
+; GFX9-NEXT:    v_writelane_b32 v40, s40, 4
+; GFX9-NEXT:    v_writelane_b32 v40, s41, 5
+; GFX9-NEXT:    v_writelane_b32 v40, s42, 6
+; GFX9-NEXT:    v_writelane_b32 v40, s43, 7
+; GFX9-NEXT:    v_writelane_b32 v40, s44, 8
+; GFX9-NEXT:    v_writelane_b32 v40, s45, 9
+; GFX9-NEXT:    v_writelane_b32 v40, s46, 10
+; GFX9-NEXT:    v_writelane_b32 v40, s47, 11
+; GFX9-NEXT:    v_writelane_b32 v40, s48, 12
+; GFX9-NEXT:    s_load_dwordx2 s[20:21], s[4:5], 0x0
+; GFX9-NEXT:    v_writelane_b32 v40, s49, 13
+; GFX9-NEXT:    v_writelane_b32 v40, s50, 14
+; GFX9-NEXT:    v_writelane_b32 v40, s51, 15
+; GFX9-NEXT:    v_writelane_b32 v40, s30, 16
+; GFX9-NEXT:    v_writelane_b32 v40, s31, 17
+; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX9-NEXT:    s_load_dwordx16 s[4:19], s[20:21], 0x0
+; GFX9-NEXT:    s_load_dwordx16 s[36:51], s[20:21], 0x40
+; GFX9-NEXT:    s_mov_b32 s33, s32
+; GFX9-NEXT:    s_add_u32 s32, s32, 0x400
+; GFX9-NEXT:    s_getpc_b64 s[30:31]
+; GFX9-NEXT:    s_add_u32 s30, s30, external_void_func_v32i32_inreg@rel32@lo+4
+; GFX9-NEXT:    s_addc_u32 s31, s31, external_void_func_v32i32_inreg@rel32@hi+12
+; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX9-NEXT:    v_mov_b32_e32 v0, s46
+; GFX9-NEXT:    v_mov_b32_e32 v1, s47
+; GFX9-NEXT:    v_mov_b32_e32 v2, s48
+; GFX9-NEXT:    v_mov_b32_e32 v3, s49
+; GFX9-NEXT:    buffer_store_dword v0, off, s[0:3], s32
+; GFX9-NEXT:    buffer_store_dword v1, off, s[0:3], s32 offset:4
+; GFX9-NEXT:    buffer_store_dword v2, off, s[0:3], s32 offset:8
+; GFX9-NEXT:    buffer_store_dword v3, off, s[0:3], s32 offset:12
+; GFX9-NEXT:    v_mov_b32_e32 v0, s50
+; GFX9-NEXT:    buffer_store_dword v0, off, s[0:3], s32 offset:16
+; GFX9-NEXT:    v_mov_b32_e32 v0, s51
+; GFX9-NEXT:    s_mov_b32 s20, s36
+; GFX9-NEXT:    s_mov_b32 s21, s37
+; GFX9-NEXT:    s_mov_b32 s22, s38
+; GFX9-NEXT:    s_mov_b32 s23, s39
+; GFX9-NEXT:    s_mov_b32 s24, s40
+; GFX9-NEXT:    s_mov_b32 s25, s41
+; GFX9-NEXT:    s_mov_b32 s26, s42
+; GFX9-NEXT:    s_mov_b32 s27, s43
+; GFX9-NEXT:    s_mov_b32 s28, s44
+; GFX9-NEXT:    s_mov_b32 s29, s45
+; GFX9-NEXT:    buffer_store_dword v0, off, s[0:3], s32 offset:20
+; GFX9-NEXT:    s_swappc_b64 s[30:31], s[30:31]
+; GFX9-NEXT:    v_readlane_b32 s4, v40, 16
+; GFX9-NEXT:    v_readlane_b32 s5, v40, 17
+; GFX9-NEXT:    v_readlane_b32 s51, v40, 15
+; GFX9-NEXT:    v_readlane_b32 s50, v40, 14
+; GFX9-NEXT:    v_readlane_b32 s49, v40, 13
+; GFX9-NEXT:    v_readlane_b32 s48, v40, 12
+; GFX9-NEXT:    v_readlane_b32 s47, v40, 11
+; GFX9-NEXT:    v_readlane_b32 s46, v40, 10
+; GFX9-NEXT:    v_readlane_b32 s45, v40, 9
+; GFX9-NEXT:    v_readlane_b32 s44, v40, 8
+; GFX9-NEXT:    v_readlane_b32 s43, v40, 7
+; GFX9-NEXT:    v_readlane_b32 s42, v40, 6
+; GFX9-NEXT:    v_readlane_b32 s41, v40, 5
+; GFX9-NEXT:    v_readlane_b32 s40, v40, 4
+; GFX9-NEXT:    v_readlane_b32 s39, v40, 3
+; GFX9-NEXT:    v_readlane_b32 s38, v40, 2
+; GFX9-NEXT:    v_readlane_b32 s37, v40, 1
+; GFX9-NEXT:    v_readlane_b32 s36, v40, 0
+; GFX9-NEXT:    s_sub_u32 s32, s32, 0x400
+; GFX9-NEXT:    v_readlane_b32 s33, v40, 18
+; GFX9-NEXT:    s_or_saveexec_b64 s[6:7], -1
+; GFX9-NEXT:    buffer_load_dword v40, off, s[0:3], s32 ; 4-byte Folded Reload
+; GFX9-NEXT:    s_mov_b64 exec, s[6:7]
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-NEXT:    s_setpc_b64 s[4:5]
+;
+; GFX10-LABEL: test_call_external_void_func_v32i32_inreg:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
+; GFX10-NEXT:    s_or_saveexec_b32 s4, -1
+; GFX10-NEXT:    buffer_store_dword v40, off, s[0:3], s32 ; 4-byte Folded Spill
+; GFX10-NEXT:    s_waitcnt_depctr 0xffe3
+; GFX10-NEXT:    s_mov_b32 exec_lo, s4
+; GFX10-NEXT:    v_writelane_b32 v40, s33, 18
+; GFX10-NEXT:    s_load_dwordx2 s[4:5], s[4:5], 0x0
+; GFX10-NEXT:    s_mov_b32 s33, s32
+; GFX10-NEXT:    s_add_u32 s32, s32, 0x200
+; GFX10-NEXT:    ; implicit-def: $vcc_hi
+; GFX10-NEXT:    v_writelane_b32 v40, s36, 0
+; GFX10-NEXT:    v_writelane_b32 v40, s37, 1
+; GFX10-NEXT:    v_writelane_b32 v40, s38, 2
+; GFX10-NEXT:    v_writelane_b32 v40, s39, 3
+; GFX10-NEXT:    v_writelane_b32 v40, s40, 4
+; GFX10-NEXT:    v_writelane_b32 v40, s41, 5
+; GFX10-NEXT:    v_writelane_b32 v40, s42, 6
+; GFX10-NEXT:    v_writelane_b32 v40, s43, 7
+; GFX10-NEXT:    v_writelane_b32 v40, s44, 8
+; GFX10-NEXT:    v_writelane_b32 v40, s45, 9
+; GFX10-NEXT:    v_writelane_b32 v40, s46, 10
+; GFX10-NEXT:    v_writelane_b32 v40, s47, 11
+; GFX10-NEXT:    v_writelane_b32 v40, s48, 12
+; GFX10-NEXT:    v_writelane_b32 v40, s49, 13
+; GFX10-NEXT:    v_writelane_b32 v40, s50, 14
+; GFX10-NEXT:    v_writelane_b32 v40, s51, 15
+; GFX10-NEXT:    v_writelane_b32 v40, s30, 16
+; GFX10-NEXT:    v_writelane_b32 v40, s31, 17
+; GFX10-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-NEXT:    s_clause 0x1
+; GFX10-NEXT:    s_load_dwordx16 s[36:51], s[4:5], 0x40
+; GFX10-NEXT:    s_load_dwordx16 s[4:19], s[4:5], 0x0
+; GFX10-NEXT:    s_getpc_b64 s[30:31]
+; GFX10-NEXT:    s_add_u32 s30, s30, external_void_func_v32i32_inreg@rel32@lo+4
+; GFX10-NEXT:    s_addc_u32 s31, s31, external_void_func_v32i32_inreg@rel32@hi+12
+; GFX10-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-NEXT:    v_mov_b32_e32 v0, s46
+; GFX10-NEXT:    v_mov_b32_e32 v1, s47
+; GFX10-NEXT:    v_mov_b32_e32 v2, s48
+; GFX10-NEXT:    v_mov_b32_e32 v3, s49
+; GFX10-NEXT:    s_mov_b32 s20, s36
+; GFX10-NEXT:    s_mov_b32 s21, s37
+; GFX10-NEXT:    s_mov_b32 s22, s38
+; GFX10-NEXT:    s_mov_b32 s23, s39
+; GFX10-NEXT:    s_mov_b32 s24, s40
+; GFX10-NEXT:    s_mov_b32 s25, s41
+; GFX10-NEXT:    s_mov_b32 s26, s42
+; GFX10-NEXT:    s_mov_b32 s27, s43
+; GFX10-NEXT:    s_mov_b32 s28, s44
+; GFX10-NEXT:    s_mov_b32 s29, s45
+; GFX10-NEXT:    v_mov_b32_e32 v4, s50
+; GFX10-NEXT:    v_mov_b32_e32 v5, s51
+; GFX10-NEXT:    buffer_store_dword v0, off, s[0:3], s32
+; GFX10-NEXT:    buffer_store_dword v1, off, s[0:3], s32 offset:4
+; GFX10-NEXT:    buffer_store_dword v2, off, s[0:3], s32 offset:8
+; GFX10-NEXT:    buffer_store_dword v3, off, s[0:3], s32 offset:12
+; GFX10-NEXT:    buffer_store_dword v4, off, s[0:3], s32 offset:16
+; GFX10-NEXT:    buffer_store_dword v5, off, s[0:3], s32 offset:20
+; GFX10-NEXT:    s_swappc_b64 s[30:31], s[30:31]
+; GFX10-NEXT:    v_readlane_b32 s4, v40, 16
+; GFX10-NEXT:    v_readlane_b32 s5, v40, 17
+; GFX10-NEXT:    v_readlane_b32 s51, v40, 15
+; GFX10-NEXT:    v_readlane_b32 s50, v40, 14
+; GFX10-NEXT:    v_readlane_b32 s49, v40, 13
+; GFX10-NEXT:    v_readlane_b32 s48, v40, 12
+; GFX10-NEXT:    v_readlane_b32 s47, v40, 11
+; GFX10-NEXT:    v_readlane_b32 s46, v40, 10
+; GFX10-NEXT:    v_readlane_b32 s45, v40, 9
+; GFX10-NEXT:    v_readlane_b32 s44, v40, 8
+; GFX10-NEXT:    v_readlane_b32 s43, v40, 7
+; GFX10-NEXT:    v_readlane_b32 s42, v40, 6
+; GFX10-NEXT:    v_readlane_b32 s41, v40, 5
+; GFX10-NEXT:    v_readlane_b32 s40, v40, 4
+; GFX10-NEXT:    v_readlane_b32 s39, v40, 3
+; GFX10-NEXT:    v_readlane_b32 s38, v40, 2
+; GFX10-NEXT:    v_readlane_b32 s37, v40, 1
+; GFX10-NEXT:    v_readlane_b32 s36, v40, 0
+; GFX10-NEXT:    s_sub_u32 s32, s32, 0x200
+; GFX10-NEXT:    v_readlane_b32 s33, v40, 18
+; GFX10-NEXT:    s_or_saveexec_b32 s6, -1
+; GFX10-NEXT:    buffer_load_dword v40, off, s[0:3], s32 ; 4-byte Folded Reload
+; GFX10-NEXT:    s_waitcnt_depctr 0xffe3
+; GFX10-NEXT:    s_mov_b32 exec_lo, s6
+; GFX10-NEXT:    s_waitcnt vmcnt(0)
+; GFX10-NEXT:    s_setpc_b64 s[4:5]
+  %ptr = load <32 x i32> addrspace(4)*, <32 x i32> addrspace(4)* addrspace(4)* undef
+  %val = load <32 x i32>, <32 x i32> addrspace(4)* %ptr
+  call amdgpu_gfx void @external_void_func_v32i32_inreg(<32 x i32> %val)
+  ret void
+}
+
+define amdgpu_gfx void @test_call_external_void_func_v32i32_i32_inreg(i32) #0 {
+; GFX9-LABEL: test_call_external_void_func_v32i32_i32_inreg:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    s_or_saveexec_b64 s[4:5], -1
+; GFX9-NEXT:    buffer_store_dword v40, off, s[0:3], s32 ; 4-byte Folded Spill
+; GFX9-NEXT:    s_mov_b64 exec, s[4:5]
+; GFX9-NEXT:    v_writelane_b32 v40, s33, 18
+; GFX9-NEXT:    v_writelane_b32 v40, s36, 0
+; GFX9-NEXT:    v_writelane_b32 v40, s37, 1
+; GFX9-NEXT:    v_writelane_b32 v40, s38, 2
+; GFX9-NEXT:    v_writelane_b32 v40, s39, 3
+; GFX9-NEXT:    v_writelane_b32 v40, s40, 4
+; GFX9-NEXT:    v_writelane_b32 v40, s41, 5
+; GFX9-NEXT:    v_writelane_b32 v40, s42, 6
+; GFX9-NEXT:    v_writelane_b32 v40, s43, 7
+; GFX9-NEXT:    v_writelane_b32 v40, s44, 8
+; GFX9-NEXT:    v_writelane_b32 v40, s45, 9
+; GFX9-NEXT:    v_writelane_b32 v40, s46, 10
+; GFX9-NEXT:    v_writelane_b32 v40, s47, 11
+; GFX9-NEXT:    v_writelane_b32 v40, s48, 12
+; GFX9-NEXT:    v_writelane_b32 v40, s49, 13
+; GFX9-NEXT:    v_writelane_b32 v40, s50, 14
+; GFX9-NEXT:    v_writelane_b32 v40, s51, 15
+; GFX9-NEXT:    v_writelane_b32 v40, s30, 16
+; GFX9-NEXT:    v_writelane_b32 v40, s31, 17
+; GFX9-NEXT:    s_load_dwordx2 s[20:21], s[4:5], 0x0
+; GFX9-NEXT:    s_load_dword s22, s[4:5], 0x0
+; GFX9-NEXT:    s_mov_b32 s33, s32
+; GFX9-NEXT:    s_add_u32 s32, s32, 0x400
+; GFX9-NEXT:    s_getpc_b64 s[30:31]
+; GFX9-NEXT:    s_add_u32 s30, s30, external_void_func_v32i32_i32_inreg@rel32@lo+4
+; GFX9-NEXT:    s_addc_u32 s31, s31, external_void_func_v32i32_i32_inreg@rel32@hi+12
+; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX9-NEXT:    s_load_dwordx16 s[4:19], s[20:21], 0x0
+; GFX9-NEXT:    s_load_dwordx16 s[36:51], s[20:21], 0x40
+; GFX9-NEXT:    v_mov_b32_e32 v0, s22
+; GFX9-NEXT:    buffer_store_dword v0, off, s[0:3], s32 offset:24
+; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX9-NEXT:    v_mov_b32_e32 v0, s46
+; GFX9-NEXT:    v_mov_b32_e32 v1, s47
+; GFX9-NEXT:    buffer_store_dword v0, off, s[0:3], s32
+; GFX9-NEXT:    buffer_store_dword v1, off, s[0:3], s32 offset:4
+; GFX9-NEXT:    v_mov_b32_e32 v0, s48
+; GFX9-NEXT:    buffer_store_dword v0, off, s[0:3], s32 offset:8
+; GFX9-NEXT:    v_mov_b32_e32 v0, s49
+; GFX9-NEXT:    buffer_store_dword v0, off, s[0:3], s32 offset:12
+; GFX9-NEXT:    v_mov_b32_e32 v0, s50
+; GFX9-NEXT:    buffer_store_dword v0, off, s[0:3], s32 offset:16
+; GFX9-NEXT:    v_mov_b32_e32 v0, s51
+; GFX9-NEXT:    s_mov_b32 s20, s36
+; GFX9-NEXT:    s_mov_b32 s21, s37
+; GFX9-NEXT:    s_mov_b32 s22, s38
+; GFX9-NEXT:    s_mov_b32 s23, s39
+; GFX9-NEXT:    s_mov_b32 s24, s40
+; GFX9-NEXT:    s_mov_b32 s25, s41
+; GFX9-NEXT:    s_mov_b32 s26, s42
+; GFX9-NEXT:    s_mov_b32 s27, s43
+; GFX9-NEXT:    s_mov_b32 s28, s44
+; GFX9-NEXT:    s_mov_b32 s29, s45
+; GFX9-NEXT:    buffer_store_dword v0, off, s[0:3], s32 offset:20
+; GFX9-NEXT:    s_swappc_b64 s[30:31], s[30:31]
+; GFX9-NEXT:    v_readlane_b32 s4, v40, 16
+; GFX9-NEXT:    v_readlane_b32 s5, v40, 17
+; GFX9-NEXT:    v_readlane_b32 s51, v40, 15
+; GFX9-NEXT:    v_readlane_b32 s50, v40, 14
+; GFX9-NEXT:    v_readlane_b32 s49, v40, 13
+; GFX9-NEXT:    v_readlane_b32 s48, v40, 12
+; GFX9-NEXT:    v_readlane_b32 s47, v40, 11
+; GFX9-NEXT:    v_readlane_b32 s46, v40, 10
+; GFX9-NEXT:    v_readlane_b32 s45, v40, 9
+; GFX9-NEXT:    v_readlane_b32 s44, v40, 8
+; GFX9-NEXT:    v_readlane_b32 s43, v40, 7
+; GFX9-NEXT:    v_readlane_b32 s42, v40, 6
+; GFX9-NEXT:    v_readlane_b32 s41, v40, 5
+; GFX9-NEXT:    v_readlane_b32 s40, v40, 4
+; GFX9-NEXT:    v_readlane_b32 s39, v40, 3
+; GFX9-NEXT:    v_readlane_b32 s38, v40, 2
+; GFX9-NEXT:    v_readlane_b32 s37, v40, 1
+; GFX9-NEXT:    v_readlane_b32 s36, v40, 0
+; GFX9-NEXT:    s_sub_u32 s32, s32, 0x400
+; GFX9-NEXT:    v_readlane_b32 s33, v40, 18
+; GFX9-NEXT:    s_or_saveexec_b64 s[6:7], -1
+; GFX9-NEXT:    buffer_load_dword v40, off, s[0:3], s32 ; 4-byte Folded Reload
+; GFX9-NEXT:    s_mov_b64 exec, s[6:7]
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-NEXT:    s_setpc_b64 s[4:5]
+;
+; GFX10-LABEL: test_call_external_void_func_v32i32_i32_inreg:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
+; GFX10-NEXT:    s_or_saveexec_b32 s4, -1
+; GFX10-NEXT:    buffer_store_dword v40, off, s[0:3], s32 ; 4-byte Folded Spill
+; GFX10-NEXT:    s_waitcnt_depctr 0xffe3
+; GFX10-NEXT:    s_mov_b32 exec_lo, s4
+; GFX10-NEXT:    v_writelane_b32 v40, s33, 18
+; GFX10-NEXT:    s_load_dwordx2 s[4:5], s[4:5], 0x0
+; GFX10-NEXT:    s_mov_b32 s33, s32
+; GFX10-NEXT:    s_add_u32 s32, s32, 0x200
+; GFX10-NEXT:    ; implicit-def: $vcc_hi
+; GFX10-NEXT:    v_writelane_b32 v40, s36, 0
+; GFX10-NEXT:    v_writelane_b32 v40, s37, 1
+; GFX10-NEXT:    v_writelane_b32 v40, s38, 2
+; GFX10-NEXT:    v_writelane_b32 v40, s39, 3
+; GFX10-NEXT:    v_writelane_b32 v40, s40, 4
+; GFX10-NEXT:    v_writelane_b32 v40, s41, 5
+; GFX10-NEXT:    v_writelane_b32 v40, s42, 6
+; GFX10-NEXT:    v_writelane_b32 v40, s43, 7
+; GFX10-NEXT:    v_writelane_b32 v40, s44, 8
+; GFX10-NEXT:    v_writelane_b32 v40, s45, 9
+; GFX10-NEXT:    v_writelane_b32 v40, s46, 10
+; GFX10-NEXT:    v_writelane_b32 v40, s47, 11
+; GFX10-NEXT:    v_writelane_b32 v40, s48, 12
+; GFX10-NEXT:    v_writelane_b32 v40, s49, 13
+; GFX10-NEXT:    v_writelane_b32 v40, s50, 14
+; GFX10-NEXT:    v_writelane_b32 v40, s51, 15
+; GFX10-NEXT:    v_writelane_b32 v40, s30, 16
+; GFX10-NEXT:    v_writelane_b32 v40, s31, 17
+; GFX10-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-NEXT:    s_clause 0x2
+; GFX10-NEXT:    s_load_dword s20, s[4:5], 0x0
+; GFX10-NEXT:    s_load_dwordx16 s[36:51], s[4:5], 0x40
+; GFX10-NEXT:    s_load_dwordx16 s[4:19], s[4:5], 0x0
+; GFX10-NEXT:    s_getpc_b64 s[30:31]
+; GFX10-NEXT:    s_add_u32 s30, s30, external_void_func_v32i32_i32_inreg@rel32@lo+4
+; GFX10-NEXT:    s_addc_u32 s31, s31, external_void_func_v32i32_i32_inreg@rel32@hi+12
+; GFX10-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-NEXT:    v_mov_b32_e32 v0, s20
+; GFX10-NEXT:    v_mov_b32_e32 v1, s47
+; GFX10-NEXT:    v_mov_b32_e32 v2, s48
+; GFX10-NEXT:    v_mov_b32_e32 v3, s49
+; GFX10-NEXT:    s_mov_b32 s20, s36
+; GFX10-NEXT:    buffer_store_dword v0, off, s[0:3], s32 offset:24
+; GFX10-NEXT:    v_mov_b32_e32 v0, s46
+; GFX10-NEXT:    s_mov_b32 s21, s37
+; GFX10-NEXT:    s_mov_b32 s22, s38
+; GFX10-NEXT:    s_mov_b32 s23, s39
+; GFX10-NEXT:    s_mov_b32 s24, s40
+; GFX10-NEXT:    s_mov_b32 s25, s41
+; GFX10-NEXT:    s_mov_b32 s26, s42
+; GFX10-NEXT:    s_mov_b32 s27, s43
+; GFX10-NEXT:    s_mov_b32 s28, s44
+; GFX10-NEXT:    s_mov_b32 s29, s45
+; GFX10-NEXT:    v_mov_b32_e32 v4, s50
+; GFX10-NEXT:    v_mov_b32_e32 v5, s51
+; GFX10-NEXT:    buffer_store_dword v0, off, s[0:3], s32
+; GFX10-NEXT:    buffer_store_dword v1, off, s[0:3], s32 offset:4
+; GFX10-NEXT:    buffer_store_dword v2, off, s[0:3], s32 offset:8
+; GFX10-NEXT:    buffer_store_dword v3, off, s[0:3], s32 offset:12
+; GFX10-NEXT:    buffer_store_dword v4, off, s[0:3], s32 offset:16
+; GFX10-NEXT:    buffer_store_dword v5, off, s[0:3], s32 offset:20
+; GFX10-NEXT:    s_swappc_b64 s[30:31], s[30:31]
+; GFX10-NEXT:    v_readlane_b32 s4, v40, 16
+; GFX10-NEXT:    v_readlane_b32 s5, v40, 17
+; GFX10-NEXT:    v_readlane_b32 s51, v40, 15
+; GFX10-NEXT:    v_readlane_b32 s50, v40, 14
+; GFX10-NEXT:    v_readlane_b32 s49, v40, 13
+; GFX10-NEXT:    v_readlane_b32 s48, v40, 12
+; GFX10-NEXT:    v_readlane_b32 s47, v40, 11
+; GFX10-NEXT:    v_readlane_b32 s46, v40, 10
+; GFX10-NEXT:    v_readlane_b32 s45, v40, 9
+; GFX10-NEXT:    v_readlane_b32 s44, v40, 8
+; GFX10-NEXT:    v_readlane_b32 s43, v40, 7
+; GFX10-NEXT:    v_readlane_b32 s42, v40, 6
+; GFX10-NEXT:    v_readlane_b32 s41, v40, 5
+; GFX10-NEXT:    v_readlane_b32 s40, v40, 4
+; GFX10-NEXT:    v_readlane_b32 s39, v40, 3
+; GFX10-NEXT:    v_readlane_b32 s38, v40, 2
+; GFX10-NEXT:    v_readlane_b32 s37, v40, 1
+; GFX10-NEXT:    v_readlane_b32 s36, v40, 0
+; GFX10-NEXT:    s_sub_u32 s32, s32, 0x200
+; GFX10-NEXT:    v_readlane_b32 s33, v40, 18
+; GFX10-NEXT:    s_or_saveexec_b32 s6, -1
+; GFX10-NEXT:    buffer_load_dword v40, off, s[0:3], s32 ; 4-byte Folded Reload
+; GFX10-NEXT:    s_waitcnt_depctr 0xffe3
+; GFX10-NEXT:    s_mov_b32 exec_lo, s6
+; GFX10-NEXT:    s_waitcnt vmcnt(0)
+; GFX10-NEXT:    s_setpc_b64 s[4:5]
+  %ptr0 = load <32 x i32> addrspace(4)*, <32 x i32> addrspace(4)* addrspace(4)* undef
+  %val0 = load <32 x i32>, <32 x i32> addrspace(4)* %ptr0
+  %val1 = load i32, i32 addrspace(4)* undef
+  call amdgpu_gfx void @external_void_func_v32i32_i32_inreg(<32 x i32> %val0, i32 %val1)
+  ret void
+}
+
+define amdgpu_gfx void @stack_passed_arg_alignment_v32i32_f64(<32 x i32> %val, double %tmp) #0 {
+; GFX9-LABEL: stack_passed_arg_alignment_v32i32_f64:
+; GFX9:       ; %bb.0: ; %entry
+; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    s_or_saveexec_b64 s[4:5], -1
+; GFX9-NEXT:    buffer_store_dword v40, off, s[0:3], s32 offset:8 ; 4-byte Folded Spill
+; GFX9-NEXT:    s_mov_b64 exec, s[4:5]
+; GFX9-NEXT:    v_writelane_b32 v40, s33, 2
+; GFX9-NEXT:    v_writelane_b32 v40, s30, 0
+; GFX9-NEXT:    s_mov_b32 s33, s32
+; GFX9-NEXT:    v_writelane_b32 v40, s31, 1
+; GFX9-NEXT:    s_nop 0
+; GFX9-NEXT:    buffer_load_dword v32, off, s[0:3], s33 offset:4
+; GFX9-NEXT:    buffer_load_dword v33, off, s[0:3], s33
+; GFX9-NEXT:    s_add_u32 s32, s32, 0x400
+; GFX9-NEXT:    s_getpc_b64 s[4:5]
+; GFX9-NEXT:    s_add_u32 s4, s4, stack_passed_f64_arg@rel32@lo+4
+; GFX9-NEXT:    s_addc_u32 s5, s5, stack_passed_f64_arg@rel32@hi+12
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-NEXT:    buffer_store_dword v33, off, s[0:3], s32
+; GFX9-NEXT:    buffer_store_dword v32, off, s[0:3], s32 offset:4
+; GFX9-NEXT:    s_swappc_b64 s[30:31], s[4:5]
+; GFX9-NEXT:    v_readlane_b32 s4, v40, 0
+; GFX9-NEXT:    v_readlane_b32 s5, v40, 1
+; GFX9-NEXT:    s_sub_u32 s32, s32, 0x400
+; GFX9-NEXT:    v_readlane_b32 s33, v40, 2
+; GFX9-NEXT:    s_or_saveexec_b64 s[6:7], -1
+; GFX9-NEXT:    buffer_load_dword v40, off, s[0:3], s32 offset:8 ; 4-byte Folded Reload
+; GFX9-NEXT:    s_mov_b64 exec, s[6:7]
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-NEXT:    s_setpc_b64 s[4:5]
+;
+; GFX10-LABEL: stack_passed_arg_alignment_v32i32_f64:
+; GFX10:       ; %bb.0: ; %entry
+; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
+; GFX10-NEXT:    s_or_saveexec_b32 s4, -1
+; GFX10-NEXT:    buffer_store_dword v40, off, s[0:3], s32 offset:8 ; 4-byte Folded Spill
+; GFX10-NEXT:    s_waitcnt_depctr 0xffe3
+; GFX10-NEXT:    s_mov_b32 exec_lo, s4
+; GFX10-NEXT:    v_writelane_b32 v40, s33, 2
+; GFX10-NEXT:    s_mov_b32 s33, s32
+; GFX10-NEXT:    s_add_u32 s32, s32, 0x200
+; GFX10-NEXT:    s_getpc_b64 s[4:5]
+; GFX10-NEXT:    s_add_u32 s4, s4, stack_passed_f64_arg@rel32@lo+4
+; GFX10-NEXT:    s_addc_u32 s5, s5, stack_passed_f64_arg@rel32@hi+12
+; GFX10-NEXT:    ; implicit-def: $vcc_hi
+; GFX10-NEXT:    v_writelane_b32 v40, s30, 0
+; GFX10-NEXT:    v_writelane_b32 v40, s31, 1
+; GFX10-NEXT:    s_clause 0x1
+; GFX10-NEXT:    buffer_load_dword v32, off, s[0:3], s33
+; GFX10-NEXT:    buffer_load_dword v33, off, s[0:3], s33 offset:4
+; GFX10-NEXT:    s_waitcnt vmcnt(1)
+; GFX10-NEXT:    buffer_store_dword v32, off, s[0:3], s32
+; GFX10-NEXT:    s_waitcnt vmcnt(0)
+; GFX10-NEXT:    buffer_store_dword v33, off, s[0:3], s32 offset:4
+; GFX10-NEXT:    s_swappc_b64 s[30:31], s[4:5]
+; GFX10-NEXT:    v_readlane_b32 s4, v40, 0
+; GFX10-NEXT:    v_readlane_b32 s5, v40, 1
+; GFX10-NEXT:    s_sub_u32 s32, s32, 0x200
+; GFX10-NEXT:    v_readlane_b32 s33, v40, 2
+; GFX10-NEXT:    s_or_saveexec_b32 s6, -1
+; GFX10-NEXT:    buffer_load_dword v40, off, s[0:3], s32 offset:8 ; 4-byte Folded Reload
+; GFX10-NEXT:    s_waitcnt_depctr 0xffe3
+; GFX10-NEXT:    s_mov_b32 exec_lo, s6
+; GFX10-NEXT:    s_waitcnt vmcnt(0)
+; GFX10-NEXT:    s_setpc_b64 s[4:5]
+entry:
+  call amdgpu_gfx void @stack_passed_f64_arg(<32 x i32> %val, double %tmp)
+  ret void
+}
+
+define amdgpu_gfx void @stack_12xv3i32() #0 {
+; GFX9-LABEL: stack_12xv3i32:
+; GFX9:       ; %bb.0: ; %entry
+; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    s_or_saveexec_b64 s[4:5], -1
+; GFX9-NEXT:    buffer_store_dword v40, off, s[0:3], s32 ; 4-byte Folded Spill
+; GFX9-NEXT:    s_mov_b64 exec, s[4:5]
+; GFX9-NEXT:    v_writelane_b32 v40, s33, 2
+; GFX9-NEXT:    v_writelane_b32 v40, s30, 0
+; GFX9-NEXT:    s_mov_b32 s33, s32
+; GFX9-NEXT:    s_add_u32 s32, s32, 0x400
+; GFX9-NEXT:    v_mov_b32_e32 v0, 12
+; GFX9-NEXT:    v_writelane_b32 v40, s31, 1
+; GFX9-NEXT:    buffer_store_dword v0, off, s[0:3], s32
+; GFX9-NEXT:    v_mov_b32_e32 v0, 13
+; GFX9-NEXT:    buffer_store_dword v0, off, s[0:3], s32 offset:4
+; GFX9-NEXT:    v_mov_b32_e32 v0, 14
+; GFX9-NEXT:    buffer_store_dword v0, off, s[0:3], s32 offset:8
+; GFX9-NEXT:    v_mov_b32_e32 v0, 15
+; GFX9-NEXT:    buffer_store_dword v0, off, s[0:3], s32 offset:12
+; GFX9-NEXT:    v_mov_b32_e32 v0, 0
+; GFX9-NEXT:    v_mov_b32_e32 v1, 0
+; GFX9-NEXT:    v_mov_b32_e32 v2, 0
+; GFX9-NEXT:    v_mov_b32_e32 v3, 1
+; GFX9-NEXT:    v_mov_b32_e32 v4, 1
+; GFX9-NEXT:    v_mov_b32_e32 v5, 1
+; GFX9-NEXT:    v_mov_b32_e32 v6, 2
+; GFX9-NEXT:    v_mov_b32_e32 v7, 2
+; GFX9-NEXT:    v_mov_b32_e32 v8, 2
+; GFX9-NEXT:    v_mov_b32_e32 v9, 3
+; GFX9-NEXT:    v_mov_b32_e32 v10, 3
+; GFX9-NEXT:    v_mov_b32_e32 v11, 3
+; GFX9-NEXT:    v_mov_b32_e32 v12, 4
+; GFX9-NEXT:    v_mov_b32_e32 v13, 4
+; GFX9-NEXT:    v_mov_b32_e32 v14, 4
+; GFX9-NEXT:    v_mov_b32_e32 v15, 5
+; GFX9-NEXT:    v_mov_b32_e32 v16, 5
+; GFX9-NEXT:    v_mov_b32_e32 v17, 5
+; GFX9-NEXT:    v_mov_b32_e32 v18, 6
+; GFX9-NEXT:    v_mov_b32_e32 v19, 6
+; GFX9-NEXT:    v_mov_b32_e32 v20, 6
+; GFX9-NEXT:    v_mov_b32_e32 v21, 7
+; GFX9-NEXT:    v_mov_b32_e32 v22, 7
+; GFX9-NEXT:    v_mov_b32_e32 v23, 7
+; GFX9-NEXT:    v_mov_b32_e32 v24, 8
+; GFX9-NEXT:    v_mov_b32_e32 v25, 8
+; GFX9-NEXT:    v_mov_b32_e32 v26, 8
+; GFX9-NEXT:    v_mov_b32_e32 v27, 9
+; GFX9-NEXT:    v_mov_b32_e32 v28, 9
+; GFX9-NEXT:    v_mov_b32_e32 v29, 9
+; GFX9-NEXT:    v_mov_b32_e32 v30, 10
+; GFX9-NEXT:    v_mov_b32_e32 v31, 11
+; GFX9-NEXT:    s_getpc_b64 s[4:5]
+; GFX9-NEXT:    s_add_u32 s4, s4, external_void_func_12xv3i32@rel32@lo+4
+; GFX9-NEXT:    s_addc_u32 s5, s5, external_void_func_12xv3i32@rel32@hi+12
+; GFX9-NEXT:    s_swappc_b64 s[30:31], s[4:5]
+; GFX9-NEXT:    v_readlane_b32 s4, v40, 0
+; GFX9-NEXT:    v_readlane_b32 s5, v40, 1
+; GFX9-NEXT:    s_sub_u32 s32, s32, 0x400
+; GFX9-NEXT:    v_readlane_b32 s33, v40, 2
+; GFX9-NEXT:    s_or_saveexec_b64 s[6:7], -1
+; GFX9-NEXT:    buffer_load_dword v40, off, s[0:3], s32 ; 4-byte Folded Reload
+; GFX9-NEXT:    s_mov_b64 exec, s[6:7]
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-NEXT:    s_setpc_b64 s[4:5]
+;
+; GFX10-LABEL: stack_12xv3i32:
+; GFX10:       ; %bb.0: ; %entry
+; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
+; GFX10-NEXT:    s_or_saveexec_b32 s4, -1
+; GFX10-NEXT:    buffer_store_dword v40, off, s[0:3], s32 ; 4-byte Folded Spill
+; GFX10-NEXT:    s_waitcnt_depctr 0xffe3
+; GFX10-NEXT:    s_mov_b32 exec_lo, s4
+; GFX10-NEXT:    v_writelane_b32 v40, s33, 2
+; GFX10-NEXT:    v_mov_b32_e32 v0, 12
+; GFX10-NEXT:    v_mov_b32_e32 v1, 13
+; GFX10-NEXT:    v_mov_b32_e32 v2, 14
+; GFX10-NEXT:    s_mov_b32 s33, s32
+; GFX10-NEXT:    v_writelane_b32 v40, s30, 0
+; GFX10-NEXT:    s_add_u32 s32, s32, 0x200
+; GFX10-NEXT:    v_mov_b32_e32 v3, 15
+; GFX10-NEXT:    v_mov_b32_e32 v4, 1
+; GFX10-NEXT:    v_mov_b32_e32 v5, 1
+; GFX10-NEXT:    v_writelane_b32 v40, s31, 1
+; GFX10-NEXT:    buffer_store_dword v0, off, s[0:3], s32
+; GFX10-NEXT:    buffer_store_dword v1, off, s[0:3], s32 offset:4
+; GFX10-NEXT:    buffer_store_dword v2, off, s[0:3], s32 offset:8
+; GFX10-NEXT:    buffer_store_dword v3, off, s[0:3], s32 offset:12
+; GFX10-NEXT:    v_mov_b32_e32 v0, 0
+; GFX10-NEXT:    v_mov_b32_e32 v1, 0
+; GFX10-NEXT:    v_mov_b32_e32 v2, 0
+; GFX10-NEXT:    v_mov_b32_e32 v3, 1
+; GFX10-NEXT:    v_mov_b32_e32 v6, 2
+; GFX10-NEXT:    v_mov_b32_e32 v7, 2
+; GFX10-NEXT:    v_mov_b32_e32 v8, 2
+; GFX10-NEXT:    v_mov_b32_e32 v9, 3
+; GFX10-NEXT:    v_mov_b32_e32 v10, 3
+; GFX10-NEXT:    v_mov_b32_e32 v11, 3
+; GFX10-NEXT:    v_mov_b32_e32 v12, 4
+; GFX10-NEXT:    v_mov_b32_e32 v13, 4
+; GFX10-NEXT:    v_mov_b32_e32 v14, 4
+; GFX10-NEXT:    v_mov_b32_e32 v15, 5
+; GFX10-NEXT:    v_mov_b32_e32 v16, 5
+; GFX10-NEXT:    v_mov_b32_e32 v17, 5
+; GFX10-NEXT:    v_mov_b32_e32 v18, 6
+; GFX10-NEXT:    v_mov_b32_e32 v19, 6
+; GFX10-NEXT:    v_mov_b32_e32 v20, 6
+; GFX10-NEXT:    v_mov_b32_e32 v21, 7
+; GFX10-NEXT:    v_mov_b32_e32 v22, 7
+; GFX10-NEXT:    v_mov_b32_e32 v23, 7
+; GFX10-NEXT:    v_mov_b32_e32 v24, 8
+; GFX10-NEXT:    v_mov_b32_e32 v25, 8
+; GFX10-NEXT:    v_mov_b32_e32 v26, 8
+; GFX10-NEXT:    v_mov_b32_e32 v27, 9
+; GFX10-NEXT:    v_mov_b32_e32 v28, 9
+; GFX10-NEXT:    v_mov_b32_e32 v29, 9
+; GFX10-NEXT:    v_mov_b32_e32 v30, 10
+; GFX10-NEXT:    v_mov_b32_e32 v31, 11
+; GFX10-NEXT:    s_getpc_b64 s[4:5]
+; GFX10-NEXT:    s_add_u32 s4, s4, external_void_func_12xv3i32@rel32@lo+4
+; GFX10-NEXT:    s_addc_u32 s5, s5, external_void_func_12xv3i32@rel32@hi+12
+; GFX10-NEXT:    ; implicit-def: $vcc_hi
+; GFX10-NEXT:    s_swappc_b64 s[30:31], s[4:5]
+; GFX10-NEXT:    v_readlane_b32 s4, v40, 0
+; GFX10-NEXT:    v_readlane_b32 s5, v40, 1
+; GFX10-NEXT:    s_sub_u32 s32, s32, 0x200
+; GFX10-NEXT:    v_readlane_b32 s33, v40, 2
+; GFX10-NEXT:    s_or_saveexec_b32 s6, -1
+; GFX10-NEXT:    buffer_load_dword v40, off, s[0:3], s32 ; 4-byte Folded Reload
+; GFX10-NEXT:    s_waitcnt_depctr 0xffe3
+; GFX10-NEXT:    s_mov_b32 exec_lo, s6
+; GFX10-NEXT:    s_waitcnt vmcnt(0)
+; GFX10-NEXT:    s_setpc_b64 s[4:5]
+entry:
+  call amdgpu_gfx void @external_void_func_12xv3i32(
+      <3 x i32><i32 0, i32 0, i32 0>,
+      <3 x i32><i32 1, i32 1, i32 1>,
+      <3 x i32><i32 2, i32 2, i32 2>,
+      <3 x i32><i32 3, i32 3, i32 3>,
+      <3 x i32><i32 4, i32 4, i32 4>,
+      <3 x i32><i32 5, i32 5, i32 5>,
+      <3 x i32><i32 6, i32 6, i32 6>,
+      <3 x i32><i32 7, i32 7, i32 7>,
+      <3 x i32><i32 8, i32 8, i32 8>,
+      <3 x i32><i32 9, i32 9, i32 9>,
+      <3 x i32><i32 10, i32 11, i32 12>,
+      <3 x i32><i32 13, i32 14, i32 15>)
+  ret void
+}
+
+define amdgpu_gfx void @stack_8xv5i32() #0 {
+; GFX9-LABEL: stack_8xv5i32:
+; GFX9:       ; %bb.0: ; %entry
+; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    s_or_saveexec_b64 s[4:5], -1
+; GFX9-NEXT:    buffer_store_dword v40, off, s[0:3], s32 ; 4-byte Folded Spill
+; GFX9-NEXT:    s_mov_b64 exec, s[4:5]
+; GFX9-NEXT:    v_writelane_b32 v40, s33, 2
+; GFX9-NEXT:    v_writelane_b32 v40, s30, 0
+; GFX9-NEXT:    s_mov_b32 s33, s32
+; GFX9-NEXT:    s_add_u32 s32, s32, 0x400
+; GFX9-NEXT:    v_mov_b32_e32 v0, 8
+; GFX9-NEXT:    v_writelane_b32 v40, s31, 1
+; GFX9-NEXT:    buffer_store_dword v0, off, s[0:3], s32
+; GFX9-NEXT:    v_mov_b32_e32 v0, 9
+; GFX9-NEXT:    buffer_store_dword v0, off, s[0:3], s32 offset:4
+; GFX9-NEXT:    v_mov_b32_e32 v0, 10
+; GFX9-NEXT:    buffer_store_dword v0, off, s[0:3], s32 offset:8
+; GFX9-NEXT:    v_mov_b32_e32 v0, 11
+; GFX9-NEXT:    buffer_store_dword v0, off, s[0:3], s32 offset:12
+; GFX9-NEXT:    v_mov_b32_e32 v0, 12
+; GFX9-NEXT:    buffer_store_dword v0, off, s[0:3], s32 offset:16
+; GFX9-NEXT:    v_mov_b32_e32 v0, 13
+; GFX9-NEXT:    buffer_store_dword v0, off, s[0:3], s32 offset:20
+; GFX9-NEXT:    v_mov_b32_e32 v0, 14
+; GFX9-NEXT:    buffer_store_dword v0, off, s[0:3], s32 offset:24
+; GFX9-NEXT:    v_mov_b32_e32 v0, 15
+; GFX9-NEXT:    buffer_store_dword v0, off, s[0:3], s32 offset:28
+; GFX9-NEXT:    v_mov_b32_e32 v0, 0
+; GFX9-NEXT:    v_mov_b32_e32 v1, 0
+; GFX9-NEXT:    v_mov_b32_e32 v2, 0
+; GFX9-NEXT:    v_mov_b32_e32 v3, 0
+; GFX9-NEXT:    v_mov_b32_e32 v4, 0
+; GFX9-NEXT:    v_mov_b32_e32 v5, 1
+; GFX9-NEXT:    v_mov_b32_e32 v6, 1
+; GFX9-NEXT:    v_mov_b32_e32 v7, 1
+; GFX9-NEXT:    v_mov_b32_e32 v8, 1
+; GFX9-NEXT:    v_mov_b32_e32 v9, 1
+; GFX9-NEXT:    v_mov_b32_e32 v10, 2
+; GFX9-NEXT:    v_mov_b32_e32 v11, 2
+; GFX9-NEXT:    v_mov_b32_e32 v12, 2
+; GFX9-NEXT:    v_mov_b32_e32 v13, 2
+; GFX9-NEXT:    v_mov_b32_e32 v14, 2
+; GFX9-NEXT:    v_mov_b32_e32 v15, 3
+; GFX9-NEXT:    v_mov_b32_e32 v16, 3
+; GFX9-NEXT:    v_mov_b32_e32 v17, 3
+; GFX9-NEXT:    v_mov_b32_e32 v18, 3
+; GFX9-NEXT:    v_mov_b32_e32 v19, 3
+; GFX9-NEXT:    v_mov_b32_e32 v20, 4
+; GFX9-NEXT:    v_mov_b32_e32 v21, 4
+; GFX9-NEXT:    v_mov_b32_e32 v22, 4
+; GFX9-NEXT:    v_mov_b32_e32 v23, 4
+; GFX9-NEXT:    v_mov_b32_e32 v24, 4
+; GFX9-NEXT:    v_mov_b32_e32 v25, 5
+; GFX9-NEXT:    v_mov_b32_e32 v26, 5
+; GFX9-NEXT:    v_mov_b32_e32 v27, 5
+; GFX9-NEXT:    v_mov_b32_e32 v28, 5
+; GFX9-NEXT:    v_mov_b32_e32 v29, 5
+; GFX9-NEXT:    v_mov_b32_e32 v30, 6
+; GFX9-NEXT:    v_mov_b32_e32 v31, 7
+; GFX9-NEXT:    s_getpc_b64 s[4:5]
+; GFX9-NEXT:    s_add_u32 s4, s4, external_void_func_8xv5i32@rel32@lo+4
+; GFX9-NEXT:    s_addc_u32 s5, s5, external_void_func_8xv5i32@rel32@hi+12
+; GFX9-NEXT:    s_swappc_b64 s[30:31], s[4:5]
+; GFX9-NEXT:    v_readlane_b32 s4, v40, 0
+; GFX9-NEXT:    v_readlane_b32 s5, v40, 1
+; GFX9-NEXT:    s_sub_u32 s32, s32, 0x400
+; GFX9-NEXT:    v_readlane_b32 s33, v40, 2
+; GFX9-NEXT:    s_or_saveexec_b64 s[6:7], -1
+; GFX9-NEXT:    buffer_load_dword v40, off, s[0:3], s32 ; 4-byte Folded Reload
+; GFX9-NEXT:    s_mov_b64 exec, s[6:7]
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-NEXT:    s_setpc_b64 s[4:5]
+;
+; GFX10-LABEL: stack_8xv5i32:
+; GFX10:       ; %bb.0: ; %entry
+; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
+; GFX10-NEXT:    s_or_saveexec_b32 s4, -1
+; GFX10-NEXT:    buffer_store_dword v40, off, s[0:3], s32 ; 4-byte Folded Spill
+; GFX10-NEXT:    s_waitcnt_depctr 0xffe3
+; GFX10-NEXT:    s_mov_b32 exec_lo, s4
+; GFX10-NEXT:    v_writelane_b32 v40, s33, 2
+; GFX10-NEXT:    v_mov_b32_e32 v0, 8
+; GFX10-NEXT:    v_mov_b32_e32 v1, 9
+; GFX10-NEXT:    v_mov_b32_e32 v2, 10
+; GFX10-NEXT:    s_mov_b32 s33, s32
+; GFX10-NEXT:    v_writelane_b32 v40, s30, 0
+; GFX10-NEXT:    s_add_u32 s32, s32, 0x200
+; GFX10-NEXT:    v_mov_b32_e32 v3, 14
+; GFX10-NEXT:    v_mov_b32_e32 v4, 15
+; GFX10-NEXT:    v_mov_b32_e32 v5, 1
+; GFX10-NEXT:    v_writelane_b32 v40, s31, 1
+; GFX10-NEXT:    buffer_store_dword v0, off, s[0:3], s32
+; GFX10-NEXT:    buffer_store_dword v1, off, s[0:3], s32 offset:4
+; GFX10-NEXT:    buffer_store_dword v2, off, s[0:3], s32 offset:8
+; GFX10-NEXT:    v_mov_b32_e32 v0, 11
+; GFX10-NEXT:    v_mov_b32_e32 v1, 12
+; GFX10-NEXT:    v_mov_b32_e32 v2, 13
+; GFX10-NEXT:    v_mov_b32_e32 v6, 1
+; GFX10-NEXT:    v_mov_b32_e32 v7, 1
+; GFX10-NEXT:    buffer_store_dword v0, off, s[0:3], s32 offset:12
+; GFX10-NEXT:    buffer_store_dword v1, off, s[0:3], s32 offset:16
+; GFX10-NEXT:    buffer_store_dword v2, off, s[0:3], s32 offset:20
+; GFX10-NEXT:    buffer_store_dword v3, off, s[0:3], s32 offset:24
+; GFX10-NEXT:    buffer_store_dword v4, off, s[0:3], s32 offset:28
+; GFX10-NEXT:    v_mov_b32_e32 v0, 0
+; GFX10-NEXT:    v_mov_b32_e32 v1, 0
+; GFX10-NEXT:    v_mov_b32_e32 v2, 0
+; GFX10-NEXT:    v_mov_b32_e32 v3, 0
+; GFX10-NEXT:    v_mov_b32_e32 v4, 0
+; GFX10-NEXT:    v_mov_b32_e32 v8, 1
+; GFX10-NEXT:    v_mov_b32_e32 v9, 1
+; GFX10-NEXT:    v_mov_b32_e32 v10, 2
+; GFX10-NEXT:    v_mov_b32_e32 v11, 2
+; GFX10-NEXT:    v_mov_b32_e32 v12, 2
+; GFX10-NEXT:    v_mov_b32_e32 v13, 2
+; GFX10-NEXT:    v_mov_b32_e32 v14, 2
+; GFX10-NEXT:    v_mov_b32_e32 v15, 3
+; GFX10-NEXT:    v_mov_b32_e32 v16, 3
+; GFX10-NEXT:    v_mov_b32_e32 v17, 3
+; GFX10-NEXT:    v_mov_b32_e32 v18, 3
+; GFX10-NEXT:    v_mov_b32_e32 v19, 3
+; GFX10-NEXT:    v_mov_b32_e32 v20, 4
+; GFX10-NEXT:    v_mov_b32_e32 v21, 4
+; GFX10-NEXT:    v_mov_b32_e32 v22, 4
+; GFX10-NEXT:    v_mov_b32_e32 v23, 4
+; GFX10-NEXT:    v_mov_b32_e32 v24, 4
+; GFX10-NEXT:    v_mov_b32_e32 v25, 5
+; GFX10-NEXT:    v_mov_b32_e32 v26, 5
+; GFX10-NEXT:    v_mov_b32_e32 v27, 5
+; GFX10-NEXT:    v_mov_b32_e32 v28, 5
+; GFX10-NEXT:    v_mov_b32_e32 v29, 5
+; GFX10-NEXT:    v_mov_b32_e32 v30, 6
+; GFX10-NEXT:    v_mov_b32_e32 v31, 7
+; GFX10-NEXT:    s_getpc_b64 s[4:5]
+; GFX10-NEXT:    s_add_u32 s4, s4, external_void_func_8xv5i32@rel32@lo+4
+; GFX10-NEXT:    s_addc_u32 s5, s5, external_void_func_8xv5i32@rel32@hi+12
+; GFX10-NEXT:    ; implicit-def: $vcc_hi
+; GFX10-NEXT:    s_swappc_b64 s[30:31], s[4:5]
+; GFX10-NEXT:    v_readlane_b32 s4, v40, 0
+; GFX10-NEXT:    v_readlane_b32 s5, v40, 1
+; GFX10-NEXT:    s_sub_u32 s32, s32, 0x200
+; GFX10-NEXT:    v_readlane_b32 s33, v40, 2
+; GFX10-NEXT:    s_or_saveexec_b32 s6, -1
+; GFX10-NEXT:    buffer_load_dword v40, off, s[0:3], s32 ; 4-byte Folded Reload
+; GFX10-NEXT:    s_waitcnt_depctr 0xffe3
+; GFX10-NEXT:    s_mov_b32 exec_lo, s6
+; GFX10-NEXT:    s_waitcnt vmcnt(0)
+; GFX10-NEXT:    s_setpc_b64 s[4:5]
+entry:
+  call amdgpu_gfx void @external_void_func_8xv5i32(
+      <5 x i32><i32 0, i32 0, i32 0, i32 0, i32 0>,
+      <5 x i32><i32 1, i32 1, i32 1, i32 1, i32 1>,
+      <5 x i32><i32 2, i32 2, i32 2, i32 2, i32 2>,
+      <5 x i32><i32 3, i32 3, i32 3, i32 3, i32 3>,
+      <5 x i32><i32 4, i32 4, i32 4, i32 4, i32 4>,
+      <5 x i32><i32 5, i32 5, i32 5, i32 5, i32 5>,
+      <5 x i32><i32 6, i32 7, i32 8, i32 9, i32 10>,
+      <5 x i32><i32 11, i32 12, i32 13, i32 14, i32 15>)
+  ret void
+}
+
+define amdgpu_gfx void @stack_8xv5f32() #0 {
+; GFX9-LABEL: stack_8xv5f32:
+; GFX9:       ; %bb.0: ; %entry
+; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    s_or_saveexec_b64 s[4:5], -1
+; GFX9-NEXT:    buffer_store_dword v40, off, s[0:3], s32 ; 4-byte Folded Spill
+; GFX9-NEXT:    s_mov_b64 exec, s[4:5]
+; GFX9-NEXT:    v_writelane_b32 v40, s33, 2
+; GFX9-NEXT:    v_writelane_b32 v40, s30, 0
+; GFX9-NEXT:    s_mov_b32 s33, s32
+; GFX9-NEXT:    s_add_u32 s32, s32, 0x400
+; GFX9-NEXT:    v_mov_b32_e32 v0, 0x41000000
+; GFX9-NEXT:    v_writelane_b32 v40, s31, 1
+; GFX9-NEXT:    buffer_store_dword v0, off, s[0:3], s32
+; GFX9-NEXT:    v_mov_b32_e32 v0, 0x41100000
+; GFX9-NEXT:    buffer_store_dword v0, off, s[0:3], s32 offset:4
+; GFX9-NEXT:    v_mov_b32_e32 v0, 0x41200000
+; GFX9-NEXT:    buffer_store_dword v0, off, s[0:3], s32 offset:8
+; GFX9-NEXT:    v_mov_b32_e32 v0, 0x41300000
+; GFX9-NEXT:    buffer_store_dword v0, off, s[0:3], s32 offset:12
+; GFX9-NEXT:    v_mov_b32_e32 v0, 0x41400000
+; GFX9-NEXT:    buffer_store_dword v0, off, s[0:3], s32 offset:16
+; GFX9-NEXT:    v_mov_b32_e32 v0, 0x41500000
+; GFX9-NEXT:    buffer_store_dword v0, off, s[0:3], s32 offset:20
+; GFX9-NEXT:    v_mov_b32_e32 v0, 0x41600000
+; GFX9-NEXT:    buffer_store_dword v0, off, s[0:3], s32 offset:24
+; GFX9-NEXT:    v_mov_b32_e32 v0, 0x41700000
+; GFX9-NEXT:    buffer_store_dword v0, off, s[0:3], s32 offset:28
+; GFX9-NEXT:    v_mov_b32_e32 v0, 0
+; GFX9-NEXT:    v_mov_b32_e32 v1, 0
+; GFX9-NEXT:    v_mov_b32_e32 v2, 0
+; GFX9-NEXT:    v_mov_b32_e32 v3, 0
+; GFX9-NEXT:    v_mov_b32_e32 v4, 0
+; GFX9-NEXT:    v_mov_b32_e32 v5, 1.0
+; GFX9-NEXT:    v_mov_b32_e32 v6, 1.0
+; GFX9-NEXT:    v_mov_b32_e32 v7, 1.0
+; GFX9-NEXT:    v_mov_b32_e32 v8, 1.0
+; GFX9-NEXT:    v_mov_b32_e32 v9, 1.0
+; GFX9-NEXT:    v_mov_b32_e32 v10, 2.0
+; GFX9-NEXT:    v_mov_b32_e32 v11, 2.0
+; GFX9-NEXT:    v_mov_b32_e32 v12, 2.0
+; GFX9-NEXT:    v_mov_b32_e32 v13, 2.0
+; GFX9-NEXT:    v_mov_b32_e32 v14, 2.0
+; GFX9-NEXT:    v_mov_b32_e32 v15, 0x40400000
+; GFX9-NEXT:    v_mov_b32_e32 v16, 0x40400000
+; GFX9-NEXT:    v_mov_b32_e32 v17, 0x40400000
+; GFX9-NEXT:    v_mov_b32_e32 v18, 0x40400000
+; GFX9-NEXT:    v_mov_b32_e32 v19, 0x40400000
+; GFX9-NEXT:    v_mov_b32_e32 v20, 4.0
+; GFX9-NEXT:    v_mov_b32_e32 v21, 4.0
+; GFX9-NEXT:    v_mov_b32_e32 v22, 4.0
+; GFX9-NEXT:    v_mov_b32_e32 v23, 4.0
+; GFX9-NEXT:    v_mov_b32_e32 v24, 4.0
+; GFX9-NEXT:    v_mov_b32_e32 v25, 0x40a00000
+; GFX9-NEXT:    v_mov_b32_e32 v26, 0x40a00000
+; GFX9-NEXT:    v_mov_b32_e32 v27, 0x40a00000
+; GFX9-NEXT:    v_mov_b32_e32 v28, 0x40a00000
+; GFX9-NEXT:    v_mov_b32_e32 v29, 0x40a00000
+; GFX9-NEXT:    v_mov_b32_e32 v30, 0x40c00000
+; GFX9-NEXT:    v_mov_b32_e32 v31, 0x40e00000
+; GFX9-NEXT:    s_getpc_b64 s[4:5]
+; GFX9-NEXT:    s_add_u32 s4, s4, external_void_func_8xv5f32@rel32@lo+4
+; GFX9-NEXT:    s_addc_u32 s5, s5, external_void_func_8xv5f32@rel32@hi+12
+; GFX9-NEXT:    s_swappc_b64 s[30:31], s[4:5]
+; GFX9-NEXT:    v_readlane_b32 s4, v40, 0
+; GFX9-NEXT:    v_readlane_b32 s5, v40, 1
+; GFX9-NEXT:    s_sub_u32 s32, s32, 0x400
+; GFX9-NEXT:    v_readlane_b32 s33, v40, 2
+; GFX9-NEXT:    s_or_saveexec_b64 s[6:7], -1
+; GFX9-NEXT:    buffer_load_dword v40, off, s[0:3], s32 ; 4-byte Folded Reload
+; GFX9-NEXT:    s_mov_b64 exec, s[6:7]
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-NEXT:    s_setpc_b64 s[4:5]
+;
+; GFX10-LABEL: stack_8xv5f32:
+; GFX10:       ; %bb.0: ; %entry
+; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
+; GFX10-NEXT:    s_or_saveexec_b32 s4, -1
+; GFX10-NEXT:    buffer_store_dword v40, off, s[0:3], s32 ; 4-byte Folded Spill
+; GFX10-NEXT:    s_waitcnt_depctr 0xffe3
+; GFX10-NEXT:    s_mov_b32 exec_lo, s4
+; GFX10-NEXT:    v_writelane_b32 v40, s33, 2
+; GFX10-NEXT:    v_mov_b32_e32 v0, 0x41000000
+; GFX10-NEXT:    v_mov_b32_e32 v1, 0x41100000
+; GFX10-NEXT:    v_mov_b32_e32 v2, 0x41200000
+; GFX10-NEXT:    s_mov_b32 s33, s32
+; GFX10-NEXT:    v_writelane_b32 v40, s30, 0
+; GFX10-NEXT:    s_add_u32 s32, s32, 0x200
+; GFX10-NEXT:    v_mov_b32_e32 v3, 0x41600000
+; GFX10-NEXT:    v_mov_b32_e32 v4, 0x41700000
+; GFX10-NEXT:    v_mov_b32_e32 v5, 1.0
+; GFX10-NEXT:    v_writelane_b32 v40, s31, 1
+; GFX10-NEXT:    buffer_store_dword v0, off, s[0:3], s32
+; GFX10-NEXT:    buffer_store_dword v1, off, s[0:3], s32 offset:4
+; GFX10-NEXT:    buffer_store_dword v2, off, s[0:3], s32 offset:8
+; GFX10-NEXT:    v_mov_b32_e32 v0, 0x41300000
+; GFX10-NEXT:    v_mov_b32_e32 v1, 0x41400000
+; GFX10-NEXT:    v_mov_b32_e32 v2, 0x41500000
+; GFX10-NEXT:    v_mov_b32_e32 v6, 1.0
+; GFX10-NEXT:    v_mov_b32_e32 v7, 1.0
+; GFX10-NEXT:    buffer_store_dword v0, off, s[0:3], s32 offset:12
+; GFX10-NEXT:    buffer_store_dword v1, off, s[0:3], s32 offset:16
+; GFX10-NEXT:    buffer_store_dword v2, off, s[0:3], s32 offset:20
+; GFX10-NEXT:    buffer_store_dword v3, off, s[0:3], s32 offset:24
+; GFX10-NEXT:    buffer_store_dword v4, off, s[0:3], s32 offset:28
+; GFX10-NEXT:    v_mov_b32_e32 v0, 0
+; GFX10-NEXT:    v_mov_b32_e32 v1, 0
+; GFX10-NEXT:    v_mov_b32_e32 v2, 0
+; GFX10-NEXT:    v_mov_b32_e32 v3, 0
+; GFX10-NEXT:    v_mov_b32_e32 v4, 0
+; GFX10-NEXT:    v_mov_b32_e32 v8, 1.0
+; GFX10-NEXT:    v_mov_b32_e32 v9, 1.0
+; GFX10-NEXT:    v_mov_b32_e32 v10, 2.0
+; GFX10-NEXT:    v_mov_b32_e32 v11, 2.0
+; GFX10-NEXT:    v_mov_b32_e32 v12, 2.0
+; GFX10-NEXT:    v_mov_b32_e32 v13, 2.0
+; GFX10-NEXT:    v_mov_b32_e32 v14, 2.0
+; GFX10-NEXT:    v_mov_b32_e32 v15, 0x40400000
+; GFX10-NEXT:    v_mov_b32_e32 v16, 0x40400000
+; GFX10-NEXT:    v_mov_b32_e32 v17, 0x40400000
+; GFX10-NEXT:    v_mov_b32_e32 v18, 0x40400000
+; GFX10-NEXT:    v_mov_b32_e32 v19, 0x40400000
+; GFX10-NEXT:    v_mov_b32_e32 v20, 4.0
+; GFX10-NEXT:    v_mov_b32_e32 v21, 4.0
+; GFX10-NEXT:    v_mov_b32_e32 v22, 4.0
+; GFX10-NEXT:    v_mov_b32_e32 v23, 4.0
+; GFX10-NEXT:    v_mov_b32_e32 v24, 4.0
+; GFX10-NEXT:    v_mov_b32_e32 v25, 0x40a00000
+; GFX10-NEXT:    v_mov_b32_e32 v26, 0x40a00000
+; GFX10-NEXT:    v_mov_b32_e32 v27, 0x40a00000
+; GFX10-NEXT:    v_mov_b32_e32 v28, 0x40a00000
+; GFX10-NEXT:    v_mov_b32_e32 v29, 0x40a00000
+; GFX10-NEXT:    v_mov_b32_e32 v30, 0x40c00000
+; GFX10-NEXT:    v_mov_b32_e32 v31, 0x40e00000
+; GFX10-NEXT:    s_getpc_b64 s[4:5]
+; GFX10-NEXT:    s_add_u32 s4, s4, external_void_func_8xv5f32@rel32@lo+4
+; GFX10-NEXT:    s_addc_u32 s5, s5, external_void_func_8xv5f32@rel32@hi+12
+; GFX10-NEXT:    ; implicit-def: $vcc_hi
+; GFX10-NEXT:    s_swappc_b64 s[30:31], s[4:5]
+; GFX10-NEXT:    v_readlane_b32 s4, v40, 0
+; GFX10-NEXT:    v_readlane_b32 s5, v40, 1
+; GFX10-NEXT:    s_sub_u32 s32, s32, 0x200
+; GFX10-NEXT:    v_readlane_b32 s33, v40, 2
+; GFX10-NEXT:    s_or_saveexec_b32 s6, -1
+; GFX10-NEXT:    buffer_load_dword v40, off, s[0:3], s32 ; 4-byte Folded Reload
+; GFX10-NEXT:    s_waitcnt_depctr 0xffe3
+; GFX10-NEXT:    s_mov_b32 exec_lo, s6
+; GFX10-NEXT:    s_waitcnt vmcnt(0)
+; GFX10-NEXT:    s_setpc_b64 s[4:5]
+entry:
+  call amdgpu_gfx void @external_void_func_8xv5f32(
+      <5 x float><float 0.0, float 0.0, float 0.0, float 0.0, float 0.0>,
+      <5 x float><float 1.0, float 1.0, float 1.0, float 1.0, float 1.0>,
+      <5 x float><float 2.0, float 2.0, float 2.0, float 2.0, float 2.0>,
+      <5 x float><float 3.0, float 3.0, float 3.0, float 3.0, float 3.0>,
+      <5 x float><float 4.0, float 4.0, float 4.0, float 4.0, float 4.0>,
+      <5 x float><float 5.0, float 5.0, float 5.0, float 5.0, float 5.0>,
+      <5 x float><float 6.0, float 7.0, float 8.0, float 9.0, float 10.0>,
+      <5 x float><float 11.0, float 12.0, float 13.0, float 14.0, float 15.0>)
+  ret void
+}
+
+declare hidden amdgpu_gfx void @byval_align16_f64_arg(<32 x i32>, double addrspace(5)* byval align 16) #0
+declare hidden amdgpu_gfx void @stack_passed_f64_arg(<32 x i32>, double) #0
+declare hidden amdgpu_gfx void @external_void_func_12xv3i32(<3 x i32>, <3 x i32>, <3 x i32>, <3 x i32>,
+    <3 x i32>, <3 x i32>, <3 x i32>, <3 x i32>, <3 x i32>, <3 x i32>, <3 x i32>, <3 x i32>) #0
+declare hidden amdgpu_gfx void @external_void_func_8xv5i32(<5 x i32>, <5 x i32>, <5 x i32>, <5 x i32>,
+    <5 x i32>, <5 x i32>, <5 x i32>, <5 x i32>) #0
+declare hidden amdgpu_gfx void @external_void_func_12xv3f32(<3 x float>, <3 x float>, <3 x float>, <3 x float>,
+    <3 x float>, <3 x float>, <3 x float>, <3 x float>, <3 x float>, <3 x float>, <3 x float>, <3 x float>) #0
+declare hidden amdgpu_gfx void @external_void_func_8xv5f32(<5 x float>, <5 x float>, <5 x float>, <5 x float>,
+    <5 x float>, <5 x float>, <5 x float>, <5 x float>) #0
+attributes #0 = { nounwind }
+attributes #1 = { nounwind noinline }
diff --git a/llvm/test/CodeGen/AMDGPU/gfx-callable-preserved-registers.ll b/llvm/test/CodeGen/AMDGPU/gfx-callable-preserved-registers.ll
new file mode 100644
--- /dev/null
+++ b/llvm/test/CodeGen/AMDGPU/gfx-callable-preserved-registers.ll
@@ -0,0 +1,835 @@
+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
+; RUN: llc -mtriple=amdgcn--amdpal -mcpu=gfx900 -enable-ipra=0 -verify-machineinstrs < %s | FileCheck -check-prefixes=GCN,GFX9 %s
+; RUN: llc -mtriple=amdgcn--amdpal -mcpu=gfx1010 -enable-ipra=0 -verify-machineinstrs < %s | FileCheck -check-prefixes=GCN,GFX10 %s
+
+declare hidden amdgpu_gfx void @external_void_func_void() #0
+
+define amdgpu_gfx void @test_call_external_void_func_void_clobber_s30_s31_call_external_void_func_void() #0 {
+; GFX9-LABEL: test_call_external_void_func_void_clobber_s30_s31_call_external_void_func_void:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    s_or_saveexec_b64 s[4:5], -1
+; GFX9-NEXT:    buffer_store_dword v40, off, s[0:3], s32 ; 4-byte Folded Spill
+; GFX9-NEXT:    s_mov_b64 exec, s[4:5]
+; GFX9-NEXT:    v_writelane_b32 v40, s33, 4
+; GFX9-NEXT:    v_writelane_b32 v40, s34, 0
+; GFX9-NEXT:    v_writelane_b32 v40, s35, 1
+; GFX9-NEXT:    v_writelane_b32 v40, s30, 2
+; GFX9-NEXT:    s_mov_b32 s33, s32
+; GFX9-NEXT:    s_add_u32 s32, s32, 0x400
+; GFX9-NEXT:    s_getpc_b64 s[34:35]
+; GFX9-NEXT:    s_add_u32 s34, s34, external_void_func_void@rel32@lo+4
+; GFX9-NEXT:    s_addc_u32 s35, s35, external_void_func_void@rel32@hi+12
+; GFX9-NEXT:    v_writelane_b32 v40, s31, 3
+; GFX9-NEXT:    s_swappc_b64 s[30:31], s[34:35]
+; GFX9-NEXT:    ;;#ASMSTART
+; GFX9-NEXT:    ;;#ASMEND
+; GFX9-NEXT:    s_swappc_b64 s[30:31], s[34:35]
+; GFX9-NEXT:    v_readlane_b32 s4, v40, 2
+; GFX9-NEXT:    v_readlane_b32 s5, v40, 3
+; GFX9-NEXT:    v_readlane_b32 s35, v40, 1
+; GFX9-NEXT:    v_readlane_b32 s34, v40, 0
+; GFX9-NEXT:    s_sub_u32 s32, s32, 0x400
+; GFX9-NEXT:    v_readlane_b32 s33, v40, 4
+; GFX9-NEXT:    s_or_saveexec_b64 s[6:7], -1
+; GFX9-NEXT:    buffer_load_dword v40, off, s[0:3], s32 ; 4-byte Folded Reload
+; GFX9-NEXT:    s_mov_b64 exec, s[6:7]
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-NEXT:    s_setpc_b64 s[4:5]
+;
+; GFX10-LABEL: test_call_external_void_func_void_clobber_s30_s31_call_external_void_func_void:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
+; GFX10-NEXT:    s_or_saveexec_b32 s4, -1
+; GFX10-NEXT:    buffer_store_dword v40, off, s[0:3], s32 ; 4-byte Folded Spill
+; GFX10-NEXT:    s_waitcnt_depctr 0xffe3
+; GFX10-NEXT:    s_mov_b32 exec_lo, s4
+; GFX10-NEXT:    v_writelane_b32 v40, s33, 4
+; GFX10-NEXT:    s_mov_b32 s33, s32
+; GFX10-NEXT:    s_add_u32 s32, s32, 0x200
+; GFX10-NEXT:    ; implicit-def: $vcc_hi
+; GFX10-NEXT:    v_writelane_b32 v40, s34, 0
+; GFX10-NEXT:    v_writelane_b32 v40, s35, 1
+; GFX10-NEXT:    s_getpc_b64 s[34:35]
+; GFX10-NEXT:    s_add_u32 s34, s34, external_void_func_void@rel32@lo+4
+; GFX10-NEXT:    s_addc_u32 s35, s35, external_void_func_void@rel32@hi+12
+; GFX10-NEXT:    v_writelane_b32 v40, s30, 2
+; GFX10-NEXT:    v_writelane_b32 v40, s31, 3
+; GFX10-NEXT:    s_swappc_b64 s[30:31], s[34:35]
+; GFX10-NEXT:    ;;#ASMSTART
+; GFX10-NEXT:    ;;#ASMEND
+; GFX10-NEXT:    s_swappc_b64 s[30:31], s[34:35]
+; GFX10-NEXT:    v_readlane_b32 s4, v40, 2
+; GFX10-NEXT:    v_readlane_b32 s5, v40, 3
+; GFX10-NEXT:    v_readlane_b32 s35, v40, 1
+; GFX10-NEXT:    v_readlane_b32 s34, v40, 0
+; GFX10-NEXT:    s_sub_u32 s32, s32, 0x200
+; GFX10-NEXT:    v_readlane_b32 s33, v40, 4
+; GFX10-NEXT:    s_or_saveexec_b32 s6, -1
+; GFX10-NEXT:    buffer_load_dword v40, off, s[0:3], s32 ; 4-byte Folded Reload
+; GFX10-NEXT:    s_waitcnt_depctr 0xffe3
+; GFX10-NEXT:    s_mov_b32 exec_lo, s6
+; GFX10-NEXT:    s_waitcnt vmcnt(0)
+; GFX10-NEXT:    s_setpc_b64 s[4:5]
+  call amdgpu_gfx void @external_void_func_void()
+  call void asm sideeffect "", ""() #0
+  call amdgpu_gfx void @external_void_func_void()
+  ret void
+}
+
+define amdgpu_gfx void @void_func_void_clobber_s30_s31() #1 {
+; GFX9-LABEL: void_func_void_clobber_s30_s31:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    s_mov_b64 s[4:5], s[30:31]
+; GFX9-NEXT:    ;;#ASMSTART
+; GFX9-NEXT:    ; clobber
+; GFX9-NEXT:    ;;#ASMEND
+; GFX9-NEXT:    s_setpc_b64 s[4:5]
+;
+; GFX10-LABEL: void_func_void_clobber_s30_s31:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
+; GFX10-NEXT:    s_mov_b64 s[4:5], s[30:31]
+; GFX10-NEXT:    ; implicit-def: $vcc_hi
+; GFX10-NEXT:    ;;#ASMSTART
+; GFX10-NEXT:    ; clobber
+; GFX10-NEXT:    ;;#ASMEND
+; GFX10-NEXT:    s_setpc_b64 s[4:5]
+  call void asm sideeffect "; clobber", "~{s[30:31]}"() #0
+  ret void
+}
+
+define amdgpu_gfx void @test_call_void_func_void_mayclobber_s31(i32 addrspace(1)* %out) #0 {
+; GFX9-LABEL: test_call_void_func_void_mayclobber_s31:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    s_or_saveexec_b64 s[4:5], -1
+; GFX9-NEXT:    buffer_store_dword v40, off, s[0:3], s32 ; 4-byte Folded Spill
+; GFX9-NEXT:    s_mov_b64 exec, s[4:5]
+; GFX9-NEXT:    v_writelane_b32 v40, s33, 3
+; GFX9-NEXT:    v_writelane_b32 v40, s34, 0
+; GFX9-NEXT:    v_writelane_b32 v40, s30, 1
+; GFX9-NEXT:    s_mov_b32 s33, s32
+; GFX9-NEXT:    s_add_u32 s32, s32, 0x400
+; GFX9-NEXT:    v_writelane_b32 v40, s31, 2
+; GFX9-NEXT:    ;;#ASMSTART
+; GFX9-NEXT:    ; def s31
+; GFX9-NEXT:    ;;#ASMEND
+; GFX9-NEXT:    s_getpc_b64 s[4:5]
+; GFX9-NEXT:    s_add_u32 s4, s4, external_void_func_void@rel32@lo+4
+; GFX9-NEXT:    s_addc_u32 s5, s5, external_void_func_void@rel32@hi+12
+; GFX9-NEXT:    s_mov_b32 s34, s31
+; GFX9-NEXT:    s_swappc_b64 s[30:31], s[4:5]
+; GFX9-NEXT:    s_mov_b32 s31, s34
+; GFX9-NEXT:    ;;#ASMSTART
+; GFX9-NEXT:    ; use s31
+; GFX9-NEXT:    ;;#ASMEND
+; GFX9-NEXT:    v_readlane_b32 s4, v40, 1
+; GFX9-NEXT:    v_readlane_b32 s5, v40, 2
+; GFX9-NEXT:    v_readlane_b32 s34, v40, 0
+; GFX9-NEXT:    s_sub_u32 s32, s32, 0x400
+; GFX9-NEXT:    v_readlane_b32 s33, v40, 3
+; GFX9-NEXT:    s_or_saveexec_b64 s[6:7], -1
+; GFX9-NEXT:    buffer_load_dword v40, off, s[0:3], s32 ; 4-byte Folded Reload
+; GFX9-NEXT:    s_mov_b64 exec, s[6:7]
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-NEXT:    s_setpc_b64 s[4:5]
+;
+; GFX10-LABEL: test_call_void_func_void_mayclobber_s31:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
+; GFX10-NEXT:    s_or_saveexec_b32 s4, -1
+; GFX10-NEXT:    buffer_store_dword v40, off, s[0:3], s32 ; 4-byte Folded Spill
+; GFX10-NEXT:    s_waitcnt_depctr 0xffe3
+; GFX10-NEXT:    s_mov_b32 exec_lo, s4
+; GFX10-NEXT:    v_writelane_b32 v40, s33, 3
+; GFX10-NEXT:    s_mov_b32 s33, s32
+; GFX10-NEXT:    s_add_u32 s32, s32, 0x200
+; GFX10-NEXT:    s_getpc_b64 s[4:5]
+; GFX10-NEXT:    s_add_u32 s4, s4, external_void_func_void@rel32@lo+4
+; GFX10-NEXT:    s_addc_u32 s5, s5, external_void_func_void@rel32@hi+12
+; GFX10-NEXT:    ; implicit-def: $vcc_hi
+; GFX10-NEXT:    v_writelane_b32 v40, s34, 0
+; GFX10-NEXT:    v_writelane_b32 v40, s30, 1
+; GFX10-NEXT:    v_writelane_b32 v40, s31, 2
+; GFX10-NEXT:    ;;#ASMSTART
+; GFX10-NEXT:    ; def s31
+; GFX10-NEXT:    ;;#ASMEND
+; GFX10-NEXT:    s_mov_b32 s34, s31
+; GFX10-NEXT:    s_swappc_b64 s[30:31], s[4:5]
+; GFX10-NEXT:    s_mov_b32 s31, s34
+; GFX10-NEXT:    ;;#ASMSTART
+; GFX10-NEXT:    ; use s31
+; GFX10-NEXT:    ;;#ASMEND
+; GFX10-NEXT:    v_readlane_b32 s4, v40, 1
+; GFX10-NEXT:    v_readlane_b32 s5, v40, 2
+; GFX10-NEXT:    v_readlane_b32 s34, v40, 0
+; GFX10-NEXT:    s_sub_u32 s32, s32, 0x200
+; GFX10-NEXT:    v_readlane_b32 s33, v40, 3
+; GFX10-NEXT:    s_or_saveexec_b32 s6, -1
+; GFX10-NEXT:    buffer_load_dword v40, off, s[0:3], s32 ; 4-byte Folded Reload
+; GFX10-NEXT:    s_waitcnt_depctr 0xffe3
+; GFX10-NEXT:    s_mov_b32 exec_lo, s6
+; GFX10-NEXT:    s_waitcnt vmcnt(0)
+; GFX10-NEXT:    s_setpc_b64 s[4:5]
+  %s31 = call i32 asm sideeffect "; def $0", "={s31}"()
+  call amdgpu_gfx void @external_void_func_void()
+  call void asm sideeffect "; use $0", "{s31}"(i32 %s31)
+  ret void
+}
+
+define amdgpu_gfx void @test_call_void_func_void_mayclobber_v31(i32 addrspace(1)* %out) #0 {
+; GFX9-LABEL: test_call_void_func_void_mayclobber_v31:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    s_or_saveexec_b64 s[4:5], -1
+; GFX9-NEXT:    buffer_store_dword v41, off, s[0:3], s32 offset:4 ; 4-byte Folded Spill
+; GFX9-NEXT:    s_mov_b64 exec, s[4:5]
+; GFX9-NEXT:    v_writelane_b32 v41, s33, 2
+; GFX9-NEXT:    s_mov_b32 s33, s32
+; GFX9-NEXT:    s_add_u32 s32, s32, 0x400
+; GFX9-NEXT:    s_getpc_b64 s[4:5]
+; GFX9-NEXT:    s_add_u32 s4, s4, external_void_func_void@rel32@lo+4
+; GFX9-NEXT:    s_addc_u32 s5, s5, external_void_func_void@rel32@hi+12
+; GFX9-NEXT:    buffer_store_dword v40, off, s[0:3], s33 ; 4-byte Folded Spill
+; GFX9-NEXT:    v_writelane_b32 v41, s30, 0
+; GFX9-NEXT:    v_writelane_b32 v41, s31, 1
+; GFX9-NEXT:    ;;#ASMSTART
+; GFX9-NEXT:    ; def v31
+; GFX9-NEXT:    ;;#ASMEND
+; GFX9-NEXT:    v_mov_b32_e32 v40, v31
+; GFX9-NEXT:    s_swappc_b64 s[30:31], s[4:5]
+; GFX9-NEXT:    v_mov_b32_e32 v31, v40
+; GFX9-NEXT:    ;;#ASMSTART
+; GFX9-NEXT:    ; use v31
+; GFX9-NEXT:    ;;#ASMEND
+; GFX9-NEXT:    v_readlane_b32 s4, v41, 0
+; GFX9-NEXT:    v_readlane_b32 s5, v41, 1
+; GFX9-NEXT:    buffer_load_dword v40, off, s[0:3], s33 ; 4-byte Folded Reload
+; GFX9-NEXT:    s_sub_u32 s32, s32, 0x400
+; GFX9-NEXT:    v_readlane_b32 s33, v41, 2
+; GFX9-NEXT:    s_or_saveexec_b64 s[6:7], -1
+; GFX9-NEXT:    buffer_load_dword v41, off, s[0:3], s32 offset:4 ; 4-byte Folded Reload
+; GFX9-NEXT:    s_mov_b64 exec, s[6:7]
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-NEXT:    s_setpc_b64 s[4:5]
+;
+; GFX10-LABEL: test_call_void_func_void_mayclobber_v31:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
+; GFX10-NEXT:    s_or_saveexec_b32 s4, -1
+; GFX10-NEXT:    buffer_store_dword v41, off, s[0:3], s32 offset:4 ; 4-byte Folded Spill
+; GFX10-NEXT:    s_waitcnt_depctr 0xffe3
+; GFX10-NEXT:    s_mov_b32 exec_lo, s4
+; GFX10-NEXT:    v_writelane_b32 v41, s33, 2
+; GFX10-NEXT:    s_mov_b32 s33, s32
+; GFX10-NEXT:    s_add_u32 s32, s32, 0x200
+; GFX10-NEXT:    s_getpc_b64 s[4:5]
+; GFX10-NEXT:    s_add_u32 s4, s4, external_void_func_void@rel32@lo+4
+; GFX10-NEXT:    s_addc_u32 s5, s5, external_void_func_void@rel32@hi+12
+; GFX10-NEXT:    buffer_store_dword v40, off, s[0:3], s33 ; 4-byte Folded Spill
+; GFX10-NEXT:    v_writelane_b32 v41, s30, 0
+; GFX10-NEXT:    ; implicit-def: $vcc_hi
+; GFX10-NEXT:    v_writelane_b32 v41, s31, 1
+; GFX10-NEXT:    ;;#ASMSTART
+; GFX10-NEXT:    ; def v31
+; GFX10-NEXT:    ;;#ASMEND
+; GFX10-NEXT:    v_mov_b32_e32 v40, v31
+; GFX10-NEXT:    s_swappc_b64 s[30:31], s[4:5]
+; GFX10-NEXT:    v_mov_b32_e32 v31, v40
+; GFX10-NEXT:    ;;#ASMSTART
+; GFX10-NEXT:    ; use v31
+; GFX10-NEXT:    ;;#ASMEND
+; GFX10-NEXT:    v_readlane_b32 s4, v41, 0
+; GFX10-NEXT:    v_readlane_b32 s5, v41, 1
+; GFX10-NEXT:    buffer_load_dword v40, off, s[0:3], s33 ; 4-byte Folded Reload
+; GFX10-NEXT:    s_sub_u32 s32, s32, 0x200
+; GFX10-NEXT:    v_readlane_b32 s33, v41, 2
+; GFX10-NEXT:    s_or_saveexec_b32 s6, -1
+; GFX10-NEXT:    buffer_load_dword v41, off, s[0:3], s32 offset:4 ; 4-byte Folded Reload
+; GFX10-NEXT:    s_waitcnt_depctr 0xffe3
+; GFX10-NEXT:    s_mov_b32 exec_lo, s6
+; GFX10-NEXT:    s_waitcnt vmcnt(0)
+; GFX10-NEXT:    s_setpc_b64 s[4:5]
+  %v31 = call i32 asm sideeffect "; def $0", "={v31}"()
+  call amdgpu_gfx void @external_void_func_void()
+  call void asm sideeffect "; use $0", "{v31}"(i32 %v31)
+  ret void
+}
+
+
+define amdgpu_gfx void @test_call_void_func_void_preserves_s33(i32 addrspace(1)* %out) #0 {
+; GFX9-LABEL: test_call_void_func_void_preserves_s33:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    s_or_saveexec_b64 s[4:5], -1
+; GFX9-NEXT:    buffer_store_dword v40, off, s[0:3], s32 ; 4-byte Folded Spill
+; GFX9-NEXT:    s_mov_b64 exec, s[4:5]
+; GFX9-NEXT:    v_writelane_b32 v40, s33, 3
+; GFX9-NEXT:    s_mov_b32 s33, s32
+; GFX9-NEXT:    v_writelane_b32 v40, s33, 0
+; GFX9-NEXT:    v_writelane_b32 v40, s30, 1
+; GFX9-NEXT:    s_add_u32 s32, s32, 0x400
+; GFX9-NEXT:    v_writelane_b32 v40, s31, 2
+; GFX9-NEXT:    s_getpc_b64 s[4:5]
+; GFX9-NEXT:    s_add_u32 s4, s4, external_void_func_void@rel32@lo+4
+; GFX9-NEXT:    s_addc_u32 s5, s5, external_void_func_void@rel32@hi+12
+; GFX9-NEXT:    ;;#ASMSTART
+; GFX9-NEXT:    ; def s33
+; GFX9-NEXT:    ;;#ASMEND
+; GFX9-NEXT:    s_swappc_b64 s[30:31], s[4:5]
+; GFX9-NEXT:    ;;#ASMSTART
+; GFX9-NEXT:    ; use s33
+; GFX9-NEXT:    ;;#ASMEND
+; GFX9-NEXT:    v_readlane_b32 s4, v40, 1
+; GFX9-NEXT:    v_readlane_b32 s5, v40, 2
+; GFX9-NEXT:    v_readlane_b32 s33, v40, 0
+; GFX9-NEXT:    s_sub_u32 s32, s32, 0x400
+; GFX9-NEXT:    v_readlane_b32 s33, v40, 3
+; GFX9-NEXT:    s_or_saveexec_b64 s[6:7], -1
+; GFX9-NEXT:    buffer_load_dword v40, off, s[0:3], s32 ; 4-byte Folded Reload
+; GFX9-NEXT:    s_mov_b64 exec, s[6:7]
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-NEXT:    s_setpc_b64 s[4:5]
+;
+; GFX10-LABEL: test_call_void_func_void_preserves_s33:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
+; GFX10-NEXT:    s_or_saveexec_b32 s4, -1
+; GFX10-NEXT:    buffer_store_dword v40, off, s[0:3], s32 ; 4-byte Folded Spill
+; GFX10-NEXT:    s_waitcnt_depctr 0xffe3
+; GFX10-NEXT:    s_mov_b32 exec_lo, s4
+; GFX10-NEXT:    v_writelane_b32 v40, s33, 3
+; GFX10-NEXT:    s_mov_b32 s33, s32
+; GFX10-NEXT:    s_add_u32 s32, s32, 0x200
+; GFX10-NEXT:    s_getpc_b64 s[4:5]
+; GFX10-NEXT:    s_add_u32 s4, s4, external_void_func_void@rel32@lo+4
+; GFX10-NEXT:    s_addc_u32 s5, s5, external_void_func_void@rel32@hi+12
+; GFX10-NEXT:    ; implicit-def: $vcc_hi
+; GFX10-NEXT:    v_writelane_b32 v40, s33, 0
+; GFX10-NEXT:    v_writelane_b32 v40, s30, 1
+; GFX10-NEXT:    v_writelane_b32 v40, s31, 2
+; GFX10-NEXT:    ;;#ASMSTART
+; GFX10-NEXT:    ; def s33
+; GFX10-NEXT:    ;;#ASMEND
+; GFX10-NEXT:    s_swappc_b64 s[30:31], s[4:5]
+; GFX10-NEXT:    ;;#ASMSTART
+; GFX10-NEXT:    ; use s33
+; GFX10-NEXT:    ;;#ASMEND
+; GFX10-NEXT:    v_readlane_b32 s4, v40, 1
+; GFX10-NEXT:    v_readlane_b32 s5, v40, 2
+; GFX10-NEXT:    v_readlane_b32 s33, v40, 0
+; GFX10-NEXT:    s_sub_u32 s32, s32, 0x200
+; GFX10-NEXT:    v_readlane_b32 s33, v40, 3
+; GFX10-NEXT:    s_or_saveexec_b32 s6, -1
+; GFX10-NEXT:    buffer_load_dword v40, off, s[0:3], s32 ; 4-byte Folded Reload
+; GFX10-NEXT:    s_waitcnt_depctr 0xffe3
+; GFX10-NEXT:    s_mov_b32 exec_lo, s6
+; GFX10-NEXT:    s_waitcnt vmcnt(0)
+; GFX10-NEXT:    s_setpc_b64 s[4:5]
+  %s33 = call i32 asm sideeffect "; def $0", "={s33}"()
+  call amdgpu_gfx void @external_void_func_void()
+  call void asm sideeffect "; use $0", "{s33}"(i32 %s33)
+  ret void
+}
+
+define amdgpu_gfx void @test_call_void_func_void_preserves_s34(i32 addrspace(1)* %out) #0 {
+; GFX9-LABEL: test_call_void_func_void_preserves_s34:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    s_or_saveexec_b64 s[4:5], -1
+; GFX9-NEXT:    buffer_store_dword v40, off, s[0:3], s32 ; 4-byte Folded Spill
+; GFX9-NEXT:    s_mov_b64 exec, s[4:5]
+; GFX9-NEXT:    v_writelane_b32 v40, s33, 3
+; GFX9-NEXT:    v_writelane_b32 v40, s34, 0
+; GFX9-NEXT:    v_writelane_b32 v40, s30, 1
+; GFX9-NEXT:    s_mov_b32 s33, s32
+; GFX9-NEXT:    s_add_u32 s32, s32, 0x400
+; GFX9-NEXT:    v_writelane_b32 v40, s31, 2
+; GFX9-NEXT:    s_getpc_b64 s[4:5]
+; GFX9-NEXT:    s_add_u32 s4, s4, external_void_func_void@rel32@lo+4
+; GFX9-NEXT:    s_addc_u32 s5, s5, external_void_func_void@rel32@hi+12
+; GFX9-NEXT:    ;;#ASMSTART
+; GFX9-NEXT:    ; def s34
+; GFX9-NEXT:    ;;#ASMEND
+; GFX9-NEXT:    s_swappc_b64 s[30:31], s[4:5]
+; GFX9-NEXT:    ;;#ASMSTART
+; GFX9-NEXT:    ; use s34
+; GFX9-NEXT:    ;;#ASMEND
+; GFX9-NEXT:    v_readlane_b32 s4, v40, 1
+; GFX9-NEXT:    v_readlane_b32 s5, v40, 2
+; GFX9-NEXT:    v_readlane_b32 s34, v40, 0
+; GFX9-NEXT:    s_sub_u32 s32, s32, 0x400
+; GFX9-NEXT:    v_readlane_b32 s33, v40, 3
+; GFX9-NEXT:    s_or_saveexec_b64 s[6:7], -1
+; GFX9-NEXT:    buffer_load_dword v40, off, s[0:3], s32 ; 4-byte Folded Reload
+; GFX9-NEXT:    s_mov_b64 exec, s[6:7]
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-NEXT:    s_setpc_b64 s[4:5]
+;
+; GFX10-LABEL: test_call_void_func_void_preserves_s34:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
+; GFX10-NEXT:    s_or_saveexec_b32 s4, -1
+; GFX10-NEXT:    buffer_store_dword v40, off, s[0:3], s32 ; 4-byte Folded Spill
+; GFX10-NEXT:    s_waitcnt_depctr 0xffe3
+; GFX10-NEXT:    s_mov_b32 exec_lo, s4
+; GFX10-NEXT:    v_writelane_b32 v40, s33, 3
+; GFX10-NEXT:    s_mov_b32 s33, s32
+; GFX10-NEXT:    s_add_u32 s32, s32, 0x200
+; GFX10-NEXT:    s_getpc_b64 s[4:5]
+; GFX10-NEXT:    s_add_u32 s4, s4, external_void_func_void@rel32@lo+4
+; GFX10-NEXT:    s_addc_u32 s5, s5, external_void_func_void@rel32@hi+12
+; GFX10-NEXT:    ; implicit-def: $vcc_hi
+; GFX10-NEXT:    v_writelane_b32 v40, s34, 0
+; GFX10-NEXT:    v_writelane_b32 v40, s30, 1
+; GFX10-NEXT:    v_writelane_b32 v40, s31, 2
+; GFX10-NEXT:    ;;#ASMSTART
+; GFX10-NEXT:    ; def s34
+; GFX10-NEXT:    ;;#ASMEND
+; GFX10-NEXT:    s_swappc_b64 s[30:31], s[4:5]
+; GFX10-NEXT:    ;;#ASMSTART
+; GFX10-NEXT:    ; use s34
+; GFX10-NEXT:    ;;#ASMEND
+; GFX10-NEXT:    v_readlane_b32 s4, v40, 1
+; GFX10-NEXT:    v_readlane_b32 s5, v40, 2
+; GFX10-NEXT:    v_readlane_b32 s34, v40, 0
+; GFX10-NEXT:    s_sub_u32 s32, s32, 0x200
+; GFX10-NEXT:    v_readlane_b32 s33, v40, 3
+; GFX10-NEXT:    s_or_saveexec_b32 s6, -1
+; GFX10-NEXT:    buffer_load_dword v40, off, s[0:3], s32 ; 4-byte Folded Reload
+; GFX10-NEXT:    s_waitcnt_depctr 0xffe3
+; GFX10-NEXT:    s_mov_b32 exec_lo, s6
+; GFX10-NEXT:    s_waitcnt vmcnt(0)
+; GFX10-NEXT:    s_setpc_b64 s[4:5]
+  %s34 = call i32 asm sideeffect "; def $0", "={s34}"()
+  call amdgpu_gfx void @external_void_func_void()
+  call void asm sideeffect "; use $0", "{s34}"(i32 %s34)
+  ret void
+}
+
+define amdgpu_gfx void @test_call_void_func_void_preserves_v40(i32 addrspace(1)* %out) #0 {
+; GFX9-LABEL: test_call_void_func_void_preserves_v40:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    s_or_saveexec_b64 s[4:5], -1
+; GFX9-NEXT:    buffer_store_dword v41, off, s[0:3], s32 offset:4 ; 4-byte Folded Spill
+; GFX9-NEXT:    s_mov_b64 exec, s[4:5]
+; GFX9-NEXT:    v_writelane_b32 v41, s33, 2
+; GFX9-NEXT:    s_mov_b32 s33, s32
+; GFX9-NEXT:    s_add_u32 s32, s32, 0x400
+; GFX9-NEXT:    s_getpc_b64 s[4:5]
+; GFX9-NEXT:    s_add_u32 s4, s4, external_void_func_void@rel32@lo+4
+; GFX9-NEXT:    s_addc_u32 s5, s5, external_void_func_void@rel32@hi+12
+; GFX9-NEXT:    buffer_store_dword v40, off, s[0:3], s33 ; 4-byte Folded Spill
+; GFX9-NEXT:    v_writelane_b32 v41, s30, 0
+; GFX9-NEXT:    v_writelane_b32 v41, s31, 1
+; GFX9-NEXT:    ;;#ASMSTART
+; GFX9-NEXT:    ; def v40
+; GFX9-NEXT:    ;;#ASMEND
+; GFX9-NEXT:    s_swappc_b64 s[30:31], s[4:5]
+; GFX9-NEXT:    ;;#ASMSTART
+; GFX9-NEXT:    ; use v40
+; GFX9-NEXT:    ;;#ASMEND
+; GFX9-NEXT:    v_readlane_b32 s4, v41, 0
+; GFX9-NEXT:    v_readlane_b32 s5, v41, 1
+; GFX9-NEXT:    buffer_load_dword v40, off, s[0:3], s33 ; 4-byte Folded Reload
+; GFX9-NEXT:    s_sub_u32 s32, s32, 0x400
+; GFX9-NEXT:    v_readlane_b32 s33, v41, 2
+; GFX9-NEXT:    s_or_saveexec_b64 s[6:7], -1
+; GFX9-NEXT:    buffer_load_dword v41, off, s[0:3], s32 offset:4 ; 4-byte Folded Reload
+; GFX9-NEXT:    s_mov_b64 exec, s[6:7]
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-NEXT:    s_setpc_b64 s[4:5]
+;
+; GFX10-LABEL: test_call_void_func_void_preserves_v40:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
+; GFX10-NEXT:    s_or_saveexec_b32 s4, -1
+; GFX10-NEXT:    buffer_store_dword v41, off, s[0:3], s32 offset:4 ; 4-byte Folded Spill
+; GFX10-NEXT:    s_waitcnt_depctr 0xffe3
+; GFX10-NEXT:    s_mov_b32 exec_lo, s4
+; GFX10-NEXT:    v_writelane_b32 v41, s33, 2
+; GFX10-NEXT:    s_mov_b32 s33, s32
+; GFX10-NEXT:    s_add_u32 s32, s32, 0x200
+; GFX10-NEXT:    s_getpc_b64 s[4:5]
+; GFX10-NEXT:    s_add_u32 s4, s4, external_void_func_void@rel32@lo+4
+; GFX10-NEXT:    s_addc_u32 s5, s5, external_void_func_void@rel32@hi+12
+; GFX10-NEXT:    buffer_store_dword v40, off, s[0:3], s33 ; 4-byte Folded Spill
+; GFX10-NEXT:    v_writelane_b32 v41, s30, 0
+; GFX10-NEXT:    ; implicit-def: $vcc_hi
+; GFX10-NEXT:    v_writelane_b32 v41, s31, 1
+; GFX10-NEXT:    ;;#ASMSTART
+; GFX10-NEXT:    ; def v40
+; GFX10-NEXT:    ;;#ASMEND
+; GFX10-NEXT:    s_swappc_b64 s[30:31], s[4:5]
+; GFX10-NEXT:    ;;#ASMSTART
+; GFX10-NEXT:    ; use v40
+; GFX10-NEXT:    ;;#ASMEND
+; GFX10-NEXT:    v_readlane_b32 s4, v41, 0
+; GFX10-NEXT:    v_readlane_b32 s5, v41, 1
+; GFX10-NEXT:    buffer_load_dword v40, off, s[0:3], s33 ; 4-byte Folded Reload
+; GFX10-NEXT:    s_sub_u32 s32, s32, 0x200
+; GFX10-NEXT:    v_readlane_b32 s33, v41, 2
+; GFX10-NEXT:    s_or_saveexec_b32 s6, -1
+; GFX10-NEXT:    buffer_load_dword v41, off, s[0:3], s32 offset:4 ; 4-byte Folded Reload
+; GFX10-NEXT:    s_waitcnt_depctr 0xffe3
+; GFX10-NEXT:    s_mov_b32 exec_lo, s6
+; GFX10-NEXT:    s_waitcnt vmcnt(0)
+; GFX10-NEXT:    s_setpc_b64 s[4:5]
+  %v40 = call i32 asm sideeffect "; def $0", "={v40}"()
+  call amdgpu_gfx void @external_void_func_void()
+  call void asm sideeffect "; use $0", "{v40}"(i32 %v40)
+  ret void
+}
+
+define hidden void @void_func_void_clobber_s33() #1 {
+; GFX9-LABEL: void_func_void_clobber_s33:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    v_writelane_b32 v0, s33, 0
+; GFX9-NEXT:    ;;#ASMSTART
+; GFX9-NEXT:    ; clobber
+; GFX9-NEXT:    ;;#ASMEND
+; GFX9-NEXT:    v_readlane_b32 s33, v0, 0
+; GFX9-NEXT:    s_setpc_b64 s[30:31]
+;
+; GFX10-LABEL: void_func_void_clobber_s33:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
+; GFX10-NEXT:    v_writelane_b32 v0, s33, 0
+; GFX10-NEXT:    ;;#ASMSTART
+; GFX10-NEXT:    ; clobber
+; GFX10-NEXT:    ;;#ASMEND
+; GFX10-NEXT:    ; implicit-def: $vcc_hi
+; GFX10-NEXT:    v_readlane_b32 s33, v0, 0
+; GFX10-NEXT:    s_setpc_b64 s[30:31]
+  call void asm sideeffect "; clobber", "~{s33}"() #0
+  ret void
+}
+
+define hidden void @void_func_void_clobber_s34() #1 {
+; GFX9-LABEL: void_func_void_clobber_s34:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    v_writelane_b32 v0, s34, 0
+; GFX9-NEXT:    ;;#ASMSTART
+; GFX9-NEXT:    ; clobber
+; GFX9-NEXT:    ;;#ASMEND
+; GFX9-NEXT:    v_readlane_b32 s34, v0, 0
+; GFX9-NEXT:    s_setpc_b64 s[30:31]
+;
+; GFX10-LABEL: void_func_void_clobber_s34:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
+; GFX10-NEXT:    v_writelane_b32 v0, s34, 0
+; GFX10-NEXT:    ;;#ASMSTART
+; GFX10-NEXT:    ; clobber
+; GFX10-NEXT:    ;;#ASMEND
+; GFX10-NEXT:    ; implicit-def: $vcc_hi
+; GFX10-NEXT:    v_readlane_b32 s34, v0, 0
+; GFX10-NEXT:    s_setpc_b64 s[30:31]
+  call void asm sideeffect "; clobber", "~{s34}"() #0
+  ret void
+}
+
+define amdgpu_gfx void @test_call_void_func_void_clobber_s33() #0 {
+; GFX9-LABEL: test_call_void_func_void_clobber_s33:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    s_or_saveexec_b64 s[4:5], -1
+; GFX9-NEXT:    buffer_store_dword v40, off, s[0:3], s32 ; 4-byte Folded Spill
+; GFX9-NEXT:    s_mov_b64 exec, s[4:5]
+; GFX9-NEXT:    v_writelane_b32 v40, s33, 2
+; GFX9-NEXT:    v_writelane_b32 v40, s30, 0
+; GFX9-NEXT:    s_mov_b32 s33, s32
+; GFX9-NEXT:    s_add_u32 s32, s32, 0x400
+; GFX9-NEXT:    s_getpc_b64 s[4:5]
+; GFX9-NEXT:    s_add_u32 s4, s4, void_func_void_clobber_s33@rel32@lo+4
+; GFX9-NEXT:    s_addc_u32 s5, s5, void_func_void_clobber_s33@rel32@hi+12
+; GFX9-NEXT:    v_writelane_b32 v40, s31, 1
+; GFX9-NEXT:    s_swappc_b64 s[30:31], s[4:5]
+; GFX9-NEXT:    v_readlane_b32 s4, v40, 0
+; GFX9-NEXT:    v_readlane_b32 s5, v40, 1
+; GFX9-NEXT:    s_sub_u32 s32, s32, 0x400
+; GFX9-NEXT:    v_readlane_b32 s33, v40, 2
+; GFX9-NEXT:    s_or_saveexec_b64 s[6:7], -1
+; GFX9-NEXT:    buffer_load_dword v40, off, s[0:3], s32 ; 4-byte Folded Reload
+; GFX9-NEXT:    s_mov_b64 exec, s[6:7]
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-NEXT:    s_setpc_b64 s[4:5]
+;
+; GFX10-LABEL: test_call_void_func_void_clobber_s33:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
+; GFX10-NEXT:    s_or_saveexec_b32 s4, -1
+; GFX10-NEXT:    buffer_store_dword v40, off, s[0:3], s32 ; 4-byte Folded Spill
+; GFX10-NEXT:    s_waitcnt_depctr 0xffe3
+; GFX10-NEXT:    s_mov_b32 exec_lo, s4
+; GFX10-NEXT:    v_writelane_b32 v40, s33, 2
+; GFX10-NEXT:    s_mov_b32 s33, s32
+; GFX10-NEXT:    s_add_u32 s32, s32, 0x200
+; GFX10-NEXT:    s_getpc_b64 s[4:5]
+; GFX10-NEXT:    s_add_u32 s4, s4, void_func_void_clobber_s33@rel32@lo+4
+; GFX10-NEXT:    s_addc_u32 s5, s5, void_func_void_clobber_s33@rel32@hi+12
+; GFX10-NEXT:    ; implicit-def: $vcc_hi
+; GFX10-NEXT:    v_writelane_b32 v40, s30, 0
+; GFX10-NEXT:    v_writelane_b32 v40, s31, 1
+; GFX10-NEXT:    s_swappc_b64 s[30:31], s[4:5]
+; GFX10-NEXT:    v_readlane_b32 s4, v40, 0
+; GFX10-NEXT:    v_readlane_b32 s5, v40, 1
+; GFX10-NEXT:    s_sub_u32 s32, s32, 0x200
+; GFX10-NEXT:    v_readlane_b32 s33, v40, 2
+; GFX10-NEXT:    s_or_saveexec_b32 s6, -1
+; GFX10-NEXT:    buffer_load_dword v40, off, s[0:3], s32 ; 4-byte Folded Reload
+; GFX10-NEXT:    s_waitcnt_depctr 0xffe3
+; GFX10-NEXT:    s_mov_b32 exec_lo, s6
+; GFX10-NEXT:    s_waitcnt vmcnt(0)
+; GFX10-NEXT:    s_setpc_b64 s[4:5]
+  call amdgpu_gfx void @void_func_void_clobber_s33()
+  ret void
+}
+
+define amdgpu_gfx void @test_call_void_func_void_clobber_s34() #0 {
+; GFX9-LABEL: test_call_void_func_void_clobber_s34:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    s_or_saveexec_b64 s[4:5], -1
+; GFX9-NEXT:    buffer_store_dword v40, off, s[0:3], s32 ; 4-byte Folded Spill
+; GFX9-NEXT:    s_mov_b64 exec, s[4:5]
+; GFX9-NEXT:    v_writelane_b32 v40, s33, 2
+; GFX9-NEXT:    v_writelane_b32 v40, s30, 0
+; GFX9-NEXT:    s_mov_b32 s33, s32
+; GFX9-NEXT:    s_add_u32 s32, s32, 0x400
+; GFX9-NEXT:    s_getpc_b64 s[4:5]
+; GFX9-NEXT:    s_add_u32 s4, s4, void_func_void_clobber_s34@rel32@lo+4
+; GFX9-NEXT:    s_addc_u32 s5, s5, void_func_void_clobber_s34@rel32@hi+12
+; GFX9-NEXT:    v_writelane_b32 v40, s31, 1
+; GFX9-NEXT:    s_swappc_b64 s[30:31], s[4:5]
+; GFX9-NEXT:    v_readlane_b32 s4, v40, 0
+; GFX9-NEXT:    v_readlane_b32 s5, v40, 1
+; GFX9-NEXT:    s_sub_u32 s32, s32, 0x400
+; GFX9-NEXT:    v_readlane_b32 s33, v40, 2
+; GFX9-NEXT:    s_or_saveexec_b64 s[6:7], -1
+; GFX9-NEXT:    buffer_load_dword v40, off, s[0:3], s32 ; 4-byte Folded Reload
+; GFX9-NEXT:    s_mov_b64 exec, s[6:7]
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-NEXT:    s_setpc_b64 s[4:5]
+;
+; GFX10-LABEL: test_call_void_func_void_clobber_s34:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
+; GFX10-NEXT:    s_or_saveexec_b32 s4, -1
+; GFX10-NEXT:    buffer_store_dword v40, off, s[0:3], s32 ; 4-byte Folded Spill
+; GFX10-NEXT:    s_waitcnt_depctr 0xffe3
+; GFX10-NEXT:    s_mov_b32 exec_lo, s4
+; GFX10-NEXT:    v_writelane_b32 v40, s33, 2
+; GFX10-NEXT:    s_mov_b32 s33, s32
+; GFX10-NEXT:    s_add_u32 s32, s32, 0x200
+; GFX10-NEXT:    s_getpc_b64 s[4:5]
+; GFX10-NEXT:    s_add_u32 s4, s4, void_func_void_clobber_s34@rel32@lo+4
+; GFX10-NEXT:    s_addc_u32 s5, s5, void_func_void_clobber_s34@rel32@hi+12
+; GFX10-NEXT:    ; implicit-def: $vcc_hi
+; GFX10-NEXT:    v_writelane_b32 v40, s30, 0
+; GFX10-NEXT:    v_writelane_b32 v40, s31, 1
+; GFX10-NEXT:    s_swappc_b64 s[30:31], s[4:5]
+; GFX10-NEXT:    v_readlane_b32 s4, v40, 0
+; GFX10-NEXT:    v_readlane_b32 s5, v40, 1
+; GFX10-NEXT:    s_sub_u32 s32, s32, 0x200
+; GFX10-NEXT:    v_readlane_b32 s33, v40, 2
+; GFX10-NEXT:    s_or_saveexec_b32 s6, -1
+; GFX10-NEXT:    buffer_load_dword v40, off, s[0:3], s32 ; 4-byte Folded Reload
+; GFX10-NEXT:    s_waitcnt_depctr 0xffe3
+; GFX10-NEXT:    s_mov_b32 exec_lo, s6
+; GFX10-NEXT:    s_waitcnt vmcnt(0)
+; GFX10-NEXT:    s_setpc_b64 s[4:5]
+  call amdgpu_gfx void @void_func_void_clobber_s34()
+  ret void
+}
+
+define amdgpu_gfx void @callee_saved_sgpr_kernel() #1 {
+; GFX9-LABEL: callee_saved_sgpr_kernel:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    s_or_saveexec_b64 s[4:5], -1
+; GFX9-NEXT:    buffer_store_dword v40, off, s[0:3], s32 ; 4-byte Folded Spill
+; GFX9-NEXT:    s_mov_b64 exec, s[4:5]
+; GFX9-NEXT:    v_writelane_b32 v40, s33, 3
+; GFX9-NEXT:    v_writelane_b32 v40, s40, 0
+; GFX9-NEXT:    v_writelane_b32 v40, s30, 1
+; GFX9-NEXT:    s_mov_b32 s33, s32
+; GFX9-NEXT:    s_add_u32 s32, s32, 0x400
+; GFX9-NEXT:    v_writelane_b32 v40, s31, 2
+; GFX9-NEXT:    s_getpc_b64 s[4:5]
+; GFX9-NEXT:    s_add_u32 s4, s4, external_void_func_void@rel32@lo+4
+; GFX9-NEXT:    s_addc_u32 s5, s5, external_void_func_void@rel32@hi+12
+; GFX9-NEXT:    ;;#ASMSTART
+; GFX9-NEXT:    ; def s40
+; GFX9-NEXT:    ;;#ASMEND
+; GFX9-NEXT:    s_swappc_b64 s[30:31], s[4:5]
+; GFX9-NEXT:    ;;#ASMSTART
+; GFX9-NEXT:    ; use s40
+; GFX9-NEXT:    ;;#ASMEND
+; GFX9-NEXT:    v_readlane_b32 s4, v40, 1
+; GFX9-NEXT:    v_readlane_b32 s5, v40, 2
+; GFX9-NEXT:    v_readlane_b32 s40, v40, 0
+; GFX9-NEXT:    s_sub_u32 s32, s32, 0x400
+; GFX9-NEXT:    v_readlane_b32 s33, v40, 3
+; GFX9-NEXT:    s_or_saveexec_b64 s[6:7], -1
+; GFX9-NEXT:    buffer_load_dword v40, off, s[0:3], s32 ; 4-byte Folded Reload
+; GFX9-NEXT:    s_mov_b64 exec, s[6:7]
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-NEXT:    s_setpc_b64 s[4:5]
+;
+; GFX10-LABEL: callee_saved_sgpr_kernel:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
+; GFX10-NEXT:    s_or_saveexec_b32 s4, -1
+; GFX10-NEXT:    buffer_store_dword v40, off, s[0:3], s32 ; 4-byte Folded Spill
+; GFX10-NEXT:    s_waitcnt_depctr 0xffe3
+; GFX10-NEXT:    s_mov_b32 exec_lo, s4
+; GFX10-NEXT:    v_writelane_b32 v40, s33, 3
+; GFX10-NEXT:    s_mov_b32 s33, s32
+; GFX10-NEXT:    s_add_u32 s32, s32, 0x200
+; GFX10-NEXT:    s_getpc_b64 s[4:5]
+; GFX10-NEXT:    s_add_u32 s4, s4, external_void_func_void@rel32@lo+4
+; GFX10-NEXT:    s_addc_u32 s5, s5, external_void_func_void@rel32@hi+12
+; GFX10-NEXT:    ; implicit-def: $vcc_hi
+; GFX10-NEXT:    v_writelane_b32 v40, s40, 0
+; GFX10-NEXT:    v_writelane_b32 v40, s30, 1
+; GFX10-NEXT:    v_writelane_b32 v40, s31, 2
+; GFX10-NEXT:    ;;#ASMSTART
+; GFX10-NEXT:    ; def s40
+; GFX10-NEXT:    ;;#ASMEND
+; GFX10-NEXT:    s_swappc_b64 s[30:31], s[4:5]
+; GFX10-NEXT:    ;;#ASMSTART
+; GFX10-NEXT:    ; use s40
+; GFX10-NEXT:    ;;#ASMEND
+; GFX10-NEXT:    v_readlane_b32 s4, v40, 1
+; GFX10-NEXT:    v_readlane_b32 s5, v40, 2
+; GFX10-NEXT:    v_readlane_b32 s40, v40, 0
+; GFX10-NEXT:    s_sub_u32 s32, s32, 0x200
+; GFX10-NEXT:    v_readlane_b32 s33, v40, 3
+; GFX10-NEXT:    s_or_saveexec_b32 s6, -1
+; GFX10-NEXT:    buffer_load_dword v40, off, s[0:3], s32 ; 4-byte Folded Reload
+; GFX10-NEXT:    s_waitcnt_depctr 0xffe3
+; GFX10-NEXT:    s_mov_b32 exec_lo, s6
+; GFX10-NEXT:    s_waitcnt vmcnt(0)
+; GFX10-NEXT:    s_setpc_b64 s[4:5]
+  %s40 = call i32 asm sideeffect "; def s40", "={s40}"() #0
+  call amdgpu_gfx void @external_void_func_void()
+  call void asm sideeffect "; use $0", "s"(i32 %s40) #0
+  ret void
+}
+
+define amdgpu_gfx void @callee_saved_sgpr_vgpr_kernel() #1 {
+; GFX9-LABEL: callee_saved_sgpr_vgpr_kernel:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    s_or_saveexec_b64 s[4:5], -1
+; GFX9-NEXT:    buffer_store_dword v41, off, s[0:3], s32 offset:4 ; 4-byte Folded Spill
+; GFX9-NEXT:    s_mov_b64 exec, s[4:5]
+; GFX9-NEXT:    v_writelane_b32 v41, s33, 3
+; GFX9-NEXT:    s_mov_b32 s33, s32
+; GFX9-NEXT:    s_add_u32 s32, s32, 0x400
+; GFX9-NEXT:    s_getpc_b64 s[4:5]
+; GFX9-NEXT:    s_add_u32 s4, s4, external_void_func_void@rel32@lo+4
+; GFX9-NEXT:    s_addc_u32 s5, s5, external_void_func_void@rel32@hi+12
+; GFX9-NEXT:    buffer_store_dword v40, off, s[0:3], s33 ; 4-byte Folded Spill
+; GFX9-NEXT:    v_writelane_b32 v41, s40, 0
+; GFX9-NEXT:    v_writelane_b32 v41, s30, 1
+; GFX9-NEXT:    v_writelane_b32 v41, s31, 2
+; GFX9-NEXT:    ;;#ASMSTART
+; GFX9-NEXT:    ; def s40
+; GFX9-NEXT:    ;;#ASMEND
+; GFX9-NEXT:    ;;#ASMSTART
+; GFX9-NEXT:    ; def v32
+; GFX9-NEXT:    ;;#ASMEND
+; GFX9-NEXT:    v_mov_b32_e32 v40, v32
+; GFX9-NEXT:    s_swappc_b64 s[30:31], s[4:5]
+; GFX9-NEXT:    ;;#ASMSTART
+; GFX9-NEXT:    ; use s40
+; GFX9-NEXT:    ;;#ASMEND
+; GFX9-NEXT:    ;;#ASMSTART
+; GFX9-NEXT:    ; use v40
+; GFX9-NEXT:    ;;#ASMEND
+; GFX9-NEXT:    v_readlane_b32 s4, v41, 1
+; GFX9-NEXT:    v_readlane_b32 s5, v41, 2
+; GFX9-NEXT:    v_readlane_b32 s40, v41, 0
+; GFX9-NEXT:    buffer_load_dword v40, off, s[0:3], s33 ; 4-byte Folded Reload
+; GFX9-NEXT:    s_sub_u32 s32, s32, 0x400
+; GFX9-NEXT:    v_readlane_b32 s33, v41, 3
+; GFX9-NEXT:    s_or_saveexec_b64 s[6:7], -1
+; GFX9-NEXT:    buffer_load_dword v41, off, s[0:3], s32 offset:4 ; 4-byte Folded Reload
+; GFX9-NEXT:    s_mov_b64 exec, s[6:7]
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-NEXT:    s_setpc_b64 s[4:5]
+;
+; GFX10-LABEL: callee_saved_sgpr_vgpr_kernel:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
+; GFX10-NEXT:    s_or_saveexec_b32 s4, -1
+; GFX10-NEXT:    buffer_store_dword v41, off, s[0:3], s32 offset:4 ; 4-byte Folded Spill
+; GFX10-NEXT:    s_waitcnt_depctr 0xffe3
+; GFX10-NEXT:    s_mov_b32 exec_lo, s4
+; GFX10-NEXT:    v_writelane_b32 v41, s33, 3
+; GFX10-NEXT:    s_mov_b32 s33, s32
+; GFX10-NEXT:    s_add_u32 s32, s32, 0x200
+; GFX10-NEXT:    s_getpc_b64 s[4:5]
+; GFX10-NEXT:    s_add_u32 s4, s4, external_void_func_void@rel32@lo+4
+; GFX10-NEXT:    s_addc_u32 s5, s5, external_void_func_void@rel32@hi+12
+; GFX10-NEXT:    buffer_store_dword v40, off, s[0:3], s33 ; 4-byte Folded Spill
+; GFX10-NEXT:    v_writelane_b32 v41, s40, 0
+; GFX10-NEXT:    ; implicit-def: $vcc_hi
+; GFX10-NEXT:    v_writelane_b32 v41, s30, 1
+; GFX10-NEXT:    v_writelane_b32 v41, s31, 2
+; GFX10-NEXT:    ;;#ASMSTART
+; GFX10-NEXT:    ; def s40
+; GFX10-NEXT:    ;;#ASMEND
+; GFX10-NEXT:    ;;#ASMSTART
+; GFX10-NEXT:    ; def v32
+; GFX10-NEXT:    ;;#ASMEND
+; GFX10-NEXT:    v_mov_b32_e32 v40, v32
+; GFX10-NEXT:    s_swappc_b64 s[30:31], s[4:5]
+; GFX10-NEXT:    ;;#ASMSTART
+; GFX10-NEXT:    ; use s40
+; GFX10-NEXT:    ;;#ASMEND
+; GFX10-NEXT:    ;;#ASMSTART
+; GFX10-NEXT:    ; use v40
+; GFX10-NEXT:    ;;#ASMEND
+; GFX10-NEXT:    v_readlane_b32 s4, v41, 1
+; GFX10-NEXT:    v_readlane_b32 s5, v41, 2
+; GFX10-NEXT:    v_readlane_b32 s40, v41, 0
+; GFX10-NEXT:    buffer_load_dword v40, off, s[0:3], s33 ; 4-byte Folded Reload
+; GFX10-NEXT:    s_sub_u32 s32, s32, 0x200
+; GFX10-NEXT:    v_readlane_b32 s33, v41, 3
+; GFX10-NEXT:    s_or_saveexec_b32 s6, -1
+; GFX10-NEXT:    buffer_load_dword v41, off, s[0:3], s32 offset:4 ; 4-byte Folded Reload
+; GFX10-NEXT:    s_waitcnt_depctr 0xffe3
+; GFX10-NEXT:    s_mov_b32 exec_lo, s6
+; GFX10-NEXT:    s_waitcnt vmcnt(0)
+; GFX10-NEXT:    s_setpc_b64 s[4:5]
+  %s40 = call i32 asm sideeffect "; def s40", "={s40}"() #0
+  %v32 = call i32 asm sideeffect "; def v32", "={v32}"() #0
+  call amdgpu_gfx void @external_void_func_void()
+  call void asm sideeffect "; use $0", "s"(i32 %s40) #0
+  call void asm sideeffect "; use $0", "v"(i32 %v32) #0
+  ret void
+}
+
+attributes #0 = { nounwind }
+attributes #1 = { nounwind noinline }
diff --git a/llvm/test/CodeGen/AMDGPU/indirect-call.ll b/llvm/test/CodeGen/AMDGPU/indirect-call.ll
--- a/llvm/test/CodeGen/AMDGPU/indirect-call.ll
+++ b/llvm/test/CodeGen/AMDGPU/indirect-call.ll
@@ -469,3 +469,54 @@
   %b = add i32 %a, 1
   ret i32 %b
 }
+
+define void @test_indirect_call_vgpr_ptr_inreg_arg(void(i32)* %fptr) {
+; GCN-LABEL: test_indirect_call_vgpr_ptr_inreg_arg:
+; GCN:       ; %bb.0:
+; GCN-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GCN-NEXT:    s_or_saveexec_b64 s[4:5], -1
+; GCN-NEXT:    buffer_store_dword v42, off, s[0:3], s32 offset:8 ; 4-byte Folded Spill
+; GCN-NEXT:    s_mov_b64 exec, s[4:5]
+; GCN-NEXT:    v_writelane_b32 v42, s33, 6
+; GCN-NEXT:    s_mov_b32 s33, s32
+; GCN-NEXT:    s_add_u32 s32, s32, 0x400
+; GCN-NEXT:    buffer_store_dword v40, off, s[0:3], s33 offset:4 ; 4-byte Folded Spill
+; GCN-NEXT:    buffer_store_dword v41, off, s[0:3], s33 ; 4-byte Folded Spill
+; GCN-NEXT:    v_writelane_b32 v42, s34, 0
+; GCN-NEXT:    v_writelane_b32 v42, s35, 1
+; GCN-NEXT:    v_writelane_b32 v42, s36, 2
+; GCN-NEXT:    v_writelane_b32 v42, s37, 3
+; GCN-NEXT:    v_writelane_b32 v42, s30, 4
+; GCN-NEXT:    v_writelane_b32 v42, s31, 5
+; GCN-NEXT:    v_mov_b32_e32 v41, v1
+; GCN-NEXT:    v_mov_b32_e32 v40, v0
+; GCN-NEXT:    s_mov_b64 s[34:35], exec
+; GCN-NEXT:  BB5_1: ; =>This Inner Loop Header: Depth=1
+; GCN-NEXT:    v_readfirstlane_b32 s6, v40
+; GCN-NEXT:    v_readfirstlane_b32 s7, v41
+; GCN-NEXT:    v_cmp_eq_u64_e32 vcc, s[6:7], v[40:41]
+; GCN-NEXT:    s_and_saveexec_b64 s[36:37], vcc
+; GCN-NEXT:    s_movk_i32 s4, 0x7b
+; GCN-NEXT:    s_swappc_b64 s[30:31], s[6:7]
+; GCN-NEXT:    s_xor_b64 exec, exec, s[36:37]
+; GCN-NEXT:    s_cbranch_execnz BB5_1
+; GCN-NEXT:  ; %bb.2:
+; GCN-NEXT:    s_mov_b64 exec, s[34:35]
+; GCN-NEXT:    v_readlane_b32 s4, v42, 4
+; GCN-NEXT:    v_readlane_b32 s5, v42, 5
+; GCN-NEXT:    v_readlane_b32 s37, v42, 3
+; GCN-NEXT:    v_readlane_b32 s36, v42, 2
+; GCN-NEXT:    v_readlane_b32 s35, v42, 1
+; GCN-NEXT:    v_readlane_b32 s34, v42, 0
+; GCN-NEXT:    buffer_load_dword v41, off, s[0:3], s33 ; 4-byte Folded Reload
+; GCN-NEXT:    buffer_load_dword v40, off, s[0:3], s33 offset:4 ; 4-byte Folded Reload
+; GCN-NEXT:    s_sub_u32 s32, s32, 0x400
+; GCN-NEXT:    v_readlane_b32 s33, v42, 6
+; GCN-NEXT:    s_or_saveexec_b64 s[6:7], -1
+; GCN-NEXT:    buffer_load_dword v42, off, s[0:3], s32 offset:8 ; 4-byte Folded Reload
+; GCN-NEXT:    s_mov_b64 exec, s[6:7]
+; GCN-NEXT:    s_waitcnt vmcnt(0)
+; GCN-NEXT:    s_setpc_b64 s[4:5]
+  call amdgpu_gfx void %fptr(i32 inreg 123)
+  ret void
+}
diff --git a/llvm/test/CodeGen/AMDGPU/unsupported-calls.ll b/llvm/test/CodeGen/AMDGPU/unsupported-calls.ll
--- a/llvm/test/CodeGen/AMDGPU/unsupported-calls.ll
+++ b/llvm/test/CodeGen/AMDGPU/unsupported-calls.ll
@@ -1,4 +1,5 @@
-; RUN: not llc -march=amdgcn -mtriple=amdgcn-- -tailcallopt < %s 2>&1 | FileCheck -check-prefix=GCN %s
+; RUN: not llc -march=amdgcn -mtriple=amdgcn-mesa-mesa3d -tailcallopt < %s 2>&1 | FileCheck -check-prefixes=GCN,MESA %s
+; RUN: not llc -march=amdgcn -mtriple=amdgcn--amdpal -tailcallopt < %s 2>&1 | FileCheck -check-prefixes=GCN,PAL %s
 ; RUN: not llc -march=r600 -mtriple=r600-- -mcpu=cypress -tailcallopt < %s 2>&1 | FileCheck -check-prefix=R600 %s
 
 declare i32 @external_function(i32) nounwind
@@ -68,13 +69,20 @@
   ret void
 }
 
-; GCN: :0:0: in function test_call_from_shader i32 (): unsupported call from graphics shader of function defined_function
-; R600: in function test_call{{.*}}: unsupported call to function defined_function
-define amdgpu_ps i32 @test_call_from_shader() {
+; GCN: :0:0: in function test_c_call_from_shader i32 (): unsupported calling convention for call from graphics shader of function defined_function
+; R600: in function test_c_call{{.*}}: unsupported call to function defined_function
+define amdgpu_ps i32 @test_c_call_from_shader() {
   %call = call i32 @defined_function(i32 0)
   ret i32 %call
 }
 
+; GCN-NOT: in function test_gfx_call{{.*}}unsupported
+; R600: in function test_gfx_call{{.*}}: unsupported call to function defined_function
+define amdgpu_ps i32 @test_gfx_call_from_shader() {
+  %call = call amdgpu_gfx i32 @defined_function(i32 0)
+  ret i32 %call
+}
+
 ; FIXME: Bad error message
 ; GCN: error: <unknown>:0:0: in function test_call_absolute void (): unsupported indirect call to function <unknown>
 ; R600: error: <unknown>:0:0: in function test_call_absolute void (): unsupported call to function <unknown>