Implement call lowering without parameters on AIX

[jasonliu]

Summary

This patch implements call lowering for calls without parameters on AIX as initial support.

[aheejin]

llvm/lib/Target/TargetMachine.cpp

181

No TT.isOSBinFormatWasm() anymore?

[jasonliu]

llvm/lib/Target/TargetMachine.cpp

181

Thanks for catching this. I will add it back in.

[sfertile]

llvm/lib/Target/PowerPC/PPCISelLowering.cpp

6612

minor nit: missing period.

6616

We don't need !isTailCall in the condition here.

6624

With this patch AIX is using IsEligibleForTailCallOptimization to determine if a call can be a tail-call. Is this safe because we expect getTargetMachine().Options.GuaranteedTailCallOpt to be false on AIX ? If thats the case then we should assert that here. Or should we have an llvm_unreachable here for unsupported tail-calls for now?

[efriedma]

llvm/lib/Target/PowerPC/PPCISelLowering.cpp

6595

Please use report_fatal_error instead of llvm_unreachable for code that you expect is actually reachable.

[hubert.reinterpretcast]

llvm/lib/Target/PowerPC/PPCCallingConv.td

309

Testing seems to confirm that r13 is a non-volatile as is stated in Table 5 of AIX Version 7.2: Assembler Language Reference under the 32-bit environment.

int main(void) {
  int ret, val = 55;
  __asm__(
      "mr 13,%1\n\t"
      "mr %0,13"
      : "=r"(ret)
      : "r"(val)
      : "r13"
      );
  return ret;
}

[hubert.reinterpretcast]

llvm/lib/Target/PowerPC/PPCFrameLowering.cpp

74

Minor nit: We should group checks of "similar" properties together. So, AIX and Darwin together.

llvm/lib/Target/PowerPC/PPCISelLowering.cpp

5157–5158

Minor: No comma for lists of less than three items.

Suggestion:

SVR4 or AIX ABIs.

5164

The NOP may change to become some other NOP.

5167–5169

Minor nit: Missing space before !isPatchPoint.

llvm/lib/Target/PowerPC/PPCISelLowering.h

1120

Minor nit: Indentation needs adjustment.

llvm/lib/Target/PowerPC/PPCInstrInfo.td

1460

Minor nit: Other blocks align to after the <.

[hubert.reinterpretcast]

llvm/lib/Target/PowerPC/PPCISelLowering.cpp

6599

This should say "reserved". The Darwin version uses pre-reserved; the prefix seems redundant.

6604

s/its varargs/the callee is variadic/;

llvm/test/CodeGen/PowerPC/test_call_aix.ll

12

Question: Where is the code that ensures quadword (16-byte) alignment of the stack?

[jasonliu]

llvm/lib/Target/PowerPC/PPCISelLowering.cpp

6624

getTargetMachine().Options.GuaranteedTailCallOpt could return true if -tailcallopt is passed in from the command line.
I will put a report_fatal_error here to indicate unsupported tail calls for now.

llvm/test/CodeGen/PowerPC/test_call_aix.ll

12

I believe it's done in PPCFrameLowering::determineFrameLayout.

[jasonliu]

Address review comment round 2.

[hubert.reinterpretcast]

One minor comment; otherwise, all of my comments have been addressed.

llvm/lib/Target/PowerPC/PPCISelLowering.cpp

5157–5158

Add "the" before AIX to clarify binding of "64-bit" as applying only to SVR4.

[jasonliu]

Clarify the comments modified for AIX.

[sfertile]

LGTM.

llvm/lib/Target/PowerPC/P9InstrResources.td

 // Two Cycle Branch
 def : InstRW<[P9_BR_2C, DISP_1C, DISP_1C],
       (instrs
   (instregex "BCCCTR(L)?(8)?$"),
   (instregex "BCCL(A|R|RL)?$"),
   (instregex "BCCTR(L)?(8)?(n)?$"),
   (instregex "BD(N)?Z(8|A|Am|Ap|m|p)?$"),
   (instregex "BD(N)?ZL(A|Am|Ap|R|R8|RL|RLm|RLp|Rm|Rp|m|p)?$"),
-  (instregex "BL(_TLS)?$"),
+  (instregex "BL(_TLS|_NOP)?$"),
   (instregex "BL8(_TLS|_NOP|_NOP_TLS|_TLS_)?$"),
   (instregex "BLA(8|8_NOP)?$"),
   (instregex "BLR(8|L)?$"),
   (instregex "TAILB(A)?(8)?$"),
   (instregex "TAILBCTR(8)?$"),
   (instregex "gBC(A|Aat|CTR|CTRL|L|LA|LAat|LR|LRL|Lat|at)?$"),
   (instregex "BCLR(L)?(n)?$"),
   (instregex "BCTR(L)?(8)?$"),

llvm/lib/Target/PowerPC/PPCCallingConv.td

 def CSR_SPE : CalleeSavedRegs<(add S14, S15, S16, S17, S18, S19, S20, S21, S22,
                                    S23, S24, S25, S26, S27, S28, S29, S30, S31
                               )>;
 
 def CSR_SVR432_Altivec : CalleeSavedRegs<(add CSR_SVR432, CSR_Altivec)>;
 
 def CSR_SVR432_SPE : CalleeSavedRegs<(add CSR_SVR432_COMM, CSR_SPE)>;
 
+def CSR_AIX32 : CalleeSavedRegs<(add R14, R15, R16, R17, R18, R19, R20,
+                                     R21, R22, R23, R24, R25, R26, R27, R28,
+                                     R29, R30, R31, F14, F15, F16, F17, F18,
+                                     F19, F20, F21, F22, F23, F24, F25, F26,
+                                     F27, F28, F29, F30, F31, CR2, CR3, CR4
+                                )>;
+
 def CSR_Darwin64 : CalleeSavedRegs<(add X13, X14, X15, X16, X17, X18, X19, X20,
                                         X21, X22, X23, X24, X25, X26, X27, X28,
                                         X29, X30, X31, F14, F15, F16, F17, F18,
                                         F19, F20, F21, F22, F23, F24, F25, F26,
                                         F27, F28, F29, F30, F31, CR2, CR3, CR4
                                    )>;
 
 def CSR_Darwin64_Altivec : CalleeSavedRegs<(add CSR_Darwin64, CSR_Altivec)>;
 
 def CSR_SVR464   : CalleeSavedRegs<(add X14, X15, X16, X17, X18, X19, X20,
                                         X21, X22, X23, X24, X25, X26, X27, X28,
                                         X29, X30, X31, F14, F15, F16, F17, F18,
                                         F19, F20, F21, F22, F23, F24, F25, F26,
                                         F27, F28, F29, F30, F31, CR2, CR3, CR4
                                    )>;
 
+def CSR_AIX64 : CalleeSavedRegs<(add X14, X15, X16, X17, X18, X19, X20,
+                                     X21, X22, X23, X24, X25, X26, X27, X28,
+                                     X29, X30, X31, F14, F15, F16, F17, F18,
+                                     F19, F20, F21, F22, F23, F24, F25, F26,
+                                     F27, F28, F29, F30, F31, CR2, CR3, CR4
+                                )>;
+
 // CSRs that are handled by prologue, epilogue.
 def CSR_SRV464_TLS_PE : CalleeSavedRegs<(add)>;
 
 def CSR_SVR464_ViaCopy : CalleeSavedRegs<(add CSR_SVR464)>;
 
 def CSR_SVR464_Altivec : CalleeSavedRegs<(add CSR_SVR464, CSR_Altivec)>;
 
 def CSR_SVR464_Altivec_ViaCopy : CalleeSavedRegs<(add CSR_SVR464_Altivec)>;

llvm/lib/Target/PowerPC/PPCFrameLowering.cpp

static unsigned computeFramePointerSaveOffset(const PPCSubtarget &STI) {
   if (STI.isDarwinABI())
     return STI.isPPC64() ? -8U : -4U;
 
   // SVR4 ABI: First slot in the general register save area.
   return STI.isPPC64() ? -8U : -4U;
 }
 
 static unsigned computeLinkageSize(const PPCSubtarget &STI) {
-  if (STI.isDarwinABI() || STI.isPPC64())
+  if (STI.isDarwinABI() || STI.isPPC64() || STI.isAIXABI())
     return (STI.isELFv2ABI() ? 4 : 6) * (STI.isPPC64() ? 8 : 4);
 
-  // SVR4 ABI:
+  // 32-bit SVR4 ABI:
   return 8;
 }
 
 static unsigned computeBasePointerSaveOffset(const PPCSubtarget &STI) {
   if (STI.isDarwinABI())
     return STI.isPPC64() ? -16U : -8U;
 
   // SVR4 ABI: First slot in the general register save area.

llvm/lib/Target/PowerPC/PPCISelLowering.h

enum NodeType : unsigned {
       /// The combination of sra[wd]i and addze used to implemented signed
       /// integer division by a power of 2. The first operand is the dividend,
       /// and the second is the constant shift amount (representing the
       /// divisor).
       SRA_ADDZE,
 
       /// CALL - A direct function call.
       /// CALL_NOP is a call with the special NOP which follows 64-bit
-      /// SVR4 calls.
+      /// SVR4 calls and 32-bit/64-bit AIX calls.
       CALL, CALL_NOP,
 
       /// CHAIN,FLAG = MTCTR(VAL, CHAIN[, INFLAG]) - Directly corresponds to a
       /// MTCTR instruction.
       MTCTR,
 
       /// CHAIN,FLAG = BCTRL(CHAIN, INFLAG) - Directly corresponds to a
       /// BCTRL instruction.
SDValue LowerCall_32SVR4(SDValue Chain, SDValue Callee,
                              CallingConv::ID CallConv, bool isVarArg,
                              bool isTailCall, bool isPatchPoint,
                              const SmallVectorImpl<ISD::OutputArg> &Outs,
                              const SmallVectorImpl<SDValue> &OutVals,
                              const SmallVectorImpl<ISD::InputArg> &Ins,
                              const SDLoc &dl, SelectionDAG &DAG,
                              SmallVectorImpl<SDValue> &InVals,
                              ImmutableCallSite CS) const;
+    SDValue LowerCall_AIX(SDValue Chain, SDValue Callee,
+                             CallingConv::ID CallConv, bool isVarArg,
+                             bool isTailCall, bool isPatchPoint,
+                             const SmallVectorImpl<ISD::OutputArg> &Outs,
+                             const SmallVectorImpl<SDValue> &OutVals,
+                             const SmallVectorImpl<ISD::InputArg> &Ins,
+                             const SDLoc &dl, SelectionDAG &DAG,
+                             SmallVectorImpl<SDValue> &InVals,
+                             ImmutableCallSite CS) const;
 
     SDValue lowerEH_SJLJ_SETJMP(SDValue Op, SelectionDAG &DAG) const;
     SDValue lowerEH_SJLJ_LONGJMP(SDValue Op, SelectionDAG &DAG) const;
     SDValue LowerBITCAST(SDValue Op, SelectionDAG &DAG) const;
 
     SDValue DAGCombineExtBoolTrunc(SDNode *N, DAGCombinerInfo &DCI) const;
     SDValue DAGCombineBuildVector(SDNode *N, DAGCombinerInfo &DCI) const;
     SDValue DAGCombineTruncBoolExt(SDNode *N, DAGCombinerInfo &DCI) const;

llvm/lib/Target/PowerPC/PPCISelLowering.cpp

assert(((Callee.getOpcode() == ISD::Register &&
             isa<ConstantSDNode>(Callee)) &&
     "Expecting an global address, external symbol, absolute value or register");
 
     DAG.getMachineFunction().getFrameInfo().setHasTailCall();
     return DAG.getNode(PPCISD::TC_RETURN, dl, MVT::Other, Ops);
   }
 
   // Add a NOP immediately after the branch instruction when using the 64-bit
-  // SVR4 ABI. At link time, if caller and callee are in a different module and
+  // SVR4 ABI, or AIX ABI.
+  // At link time, if caller and callee are in a different module and
   // thus have a different TOC, the call will be replaced with a call to a stub
   // function which saves the current TOC, loads the TOC of the callee and
   // branches to the callee. The NOP will be replaced with a load instruction
   // which restores the TOC of the caller from the TOC save slot of the current
   // stack frame. If caller and callee belong to the same module (and have the
   // same TOC), the NOP will remain unchanged.
 
   MachineFunction &MF = DAG.getMachineFunction();
-  if (!isTailCall && Subtarget.isSVR4ABI()&& Subtarget.isPPC64() &&
-      !isPatchPoint) {
+  if (!isTailCall &&!isPatchPoint &&
+      ((Subtarget.isSVR4ABI() && Subtarget.isPPC64()) ||
+       Subtarget.isAIXABI())) {
     if (CallOpc == PPCISD::BCTRL) {
+      if (Subtarget.isAIXABI())
+        llvm_unreachable("Indirect call on AIX is not implemented.");
+
       // This is a call through a function pointer.
       // Restore the caller TOC from the save area into R2.
       // See PrepareCall() for more information about calls through function
       // pointers in the 64-bit SVR4 ABI.
       // We are using a target-specific load with r2 hard coded, because the
       // result of a target-independent load would never go directly into r2,
       // since r2 is a reserved register (which prevents the register allocator
       // from allocating it), resulting in an additional register being
PPCTargetLowering::LowerCall(TargetLowering::CallLoweringInfo &CLI,
 
   // When long calls (i.e. indirect calls) are always used, calls are always
   // made via function pointer. If we have a function name, first translate it
   // into a pointer.
   if (Subtarget.useLongCalls() && isa<GlobalAddressSDNode>(Callee) &&
       !isTailCall)
     Callee = LowerGlobalAddress(Callee, DAG);
 
-  if (Subtarget.isSVR4ABI()) {
-    if (Subtarget.isPPC64())
-      return LowerCall_64SVR4(Chain, Callee, CallConv, isVarArg,
-                              isTailCall, isPatchPoint, Outs, OutVals, Ins,
-                              dl, DAG, InVals, CS);
-    else
-      return LowerCall_32SVR4(Chain, Callee, CallConv, isVarArg,
-                              isTailCall, isPatchPoint, Outs, OutVals, Ins,
-                              dl, DAG, InVals, CS);
-  }
+  if (Subtarget.isSVR4ABI() && Subtarget.isPPC64())
+
+  if (Subtarget.isSVR4ABI())
+
+  if (Subtarget.isAIXABI())
+    return LowerCall_AIX(Chain, Callee, CallConv, isVarArg,
+                         isTailCall, isPatchPoint, Outs, OutVals, Ins,
+                         dl, DAG, InVals, CS);
 
   return LowerCall_Darwin(Chain, Callee, CallConv, isVarArg,
                           isTailCall, isPatchPoint, Outs, OutVals, Ins,
                           dl, DAG, InVals, CS);
 }
 
 SDValue PPCTargetLowering::LowerCall_32SVR4(
     SDValue Chain, SDValue Callee, CallingConv::ID CallConv, bool isVarArg,
PrepareTailCall(DAG, InFlag, Chain, dl, SPDiff, NumBytes, LROp, FPOp,
                     TailCallArguments);
 
   return FinishCall(CallConv, dl, isTailCall, isVarArg, isPatchPoint,
                     /* unused except on PPC64 ELFv1 */ false, DAG,
                     RegsToPass, InFlag, Chain, CallSeqStart, Callee, SPDiff,
                     NumBytes, Ins, InVals, CS);
 }
 
+
+SDValue PPCTargetLowering::LowerCall_AIX(
+    SDValue Chain, SDValue Callee, CallingConv::ID CallConv, bool isVarArg,
+    bool isTailCall, bool isPatchPoint,
+    const SmallVectorImpl<ISD::OutputArg> &Outs,
+    const SmallVectorImpl<SDValue> &OutVals,
+    const SmallVectorImpl<ISD::InputArg> &Ins, const SDLoc &dl,
+    SelectionDAG &DAG, SmallVectorImpl<SDValue> &InVals,
+    ImmutableCallSite CS) const {
+
+  assert((CallConv == CallingConv::C || CallConv == CallingConv::Fast) &&
+         "Unimplemented calling convention!");
+  if (isVarArg || isTailCall || isPatchPoint)
+    llvm_unreachable("This call type is unimplemented on AIX.");
+
+  EVT PtrVT = getPointerTy(DAG.getDataLayout());
+  bool isPPC64 = PtrVT == MVT::i64;
+  unsigned PtrByteSize = isPPC64 ? 8 : 4;
+  unsigned NumOps = Outs.size();
+
+  if (NumOps != 0)
+    llvm_unreachable("Call lowering with parameters is not implemented "
+                     "on AIX yet.");
+
+  // Count how many bytes are to be pushed on the stack, including the linkage
+  // area, parameter list area.
+  // On XCOFF, we start with 24/48, which is preserved space for 
+  // [SP][CR][LR][2 x reserved][TOC].
+  unsigned LinkageSize = Subtarget.getFrameLowering()->getLinkageSize();
+
+  // The prolog code of the callee may store up to 8 GPR argument registers to
+  // the stack, allowing va_start to index over them in memory if its varargs.
+  // Because we cannot tell if this is needed on the caller side, we have to
+  // conservatively assume that it is needed.  As such, make sure we have at
+  // least enough stack space for the caller to store the 8 GPRs.
+  unsigned NumBytes = LinkageSize + 8 * PtrByteSize;
+
+  // Adjust the stack pointer for the new arguments...
+  // These operations are automatically eliminated by the prolog/epilog
+  // inserter pass
+  Chain = DAG.getCALLSEQ_START(Chain, NumBytes, 0, dl);
+  SDValue CallSeqStart = Chain;
+
+  if (!isTailCall &&
+      !isFunctionGlobalAddress(Callee) &&
+      !isa<ExternalSymbolSDNode>(Callee))
+    llvm_unreachable("Handling of indirect call is unimplemented!");
+  
+  SmallVector<std::pair<unsigned, SDValue>, 8> RegsToPass;
+  SDValue InFlag;
+  
+  // TODO: when tail call is supported, use CalculateTailCallSPDiff() 
+  // to calculate by how many bytes the stack has to be adjusted in case
+  // of tail call optimization.
+  int SPDiff = 0;
+
+  return FinishCall(CallConv, dl, isTailCall, isVarArg, isPatchPoint,
+                    /* unused except on PPC64 ELFv1 */ false, DAG,
+                    RegsToPass, InFlag, Chain, CallSeqStart, Callee, SPDiff,
+                    NumBytes, Ins, InVals, CS);
+}
+
 bool
 PPCTargetLowering::CanLowerReturn(CallingConv::ID CallConv,
                                   MachineFunction &MF, bool isVarArg,
                                   const SmallVectorImpl<ISD::OutputArg> &Outs,
                                   LLVMContext &Context) const {
   SmallVector<CCValAssign, 16> RVLocs;
   CCState CCInfo(CallConv, isVarArg, MF, RVLocs, Context);
   return CCInfo.CheckReturn(

llvm/lib/Target/PowerPC/PPCInstrInfo.td

let isCodeGenOnly = 1 in {
                         "b${cond:cc}la${cond:pm} ${cond:reg}, $dst">;
 
       def BCL  : BForm_4<16, 12, 0, 1, (outs),
                          (ins crbitrc:$bi, condbrtarget:$dst),
                          "bcl 12, $bi, $dst">;
       def BCLn : BForm_4<16, 4, 0, 1, (outs),
                          (ins crbitrc:$bi, condbrtarget:$dst),
                          "bcl 4, $bi, $dst">;
+      def BL_NOP  : IForm_and_DForm_4_zero<18, 0, 1, 24,
+                              (outs), (ins calltarget:$func),
+                              "bl $func\n\tnop", IIC_BrB, []>;
     }
   }
   let Uses = [CTR, RM] in {
     def BCTRL : XLForm_2_ext<19, 528, 20, 0, 1, (outs), (ins),
                              "bctrl", IIC_BrB, [(PPCbctrl)]>,
                 Requires<[In32BitMode]>;
 
     let isCodeGenOnly = 1 in {

 
 // RLWNM
 def : Pat<(and (rotl i32:$in, i32:$sh), maskimm32:$imm),
           (RLWNM $in, $sh, (MB maskimm32:$imm), (ME maskimm32:$imm))>;
 
 // Calls
 def : Pat<(PPCcall (i32 tglobaladdr:$dst)),
           (BL tglobaladdr:$dst)>;
+def : Pat<(PPCcall_nop (i32 tglobaladdr:$dst)),
+          (BL_NOP tglobaladdr:$dst)>;
+
 def : Pat<(PPCcall (i32 texternalsym:$dst)),
           (BL texternalsym:$dst)>;
 
 def : Pat<(PPCtc_return (i32 tglobaladdr:$dst),  imm:$imm),
           (TCRETURNdi tglobaladdr:$dst, imm:$imm)>;
 
 def : Pat<(PPCtc_return (i32 texternalsym:$dst), imm:$imm),
           (TCRETURNdi texternalsym:$dst, imm:$imm)>;

llvm/lib/Target/PowerPC/PPCRegisterInfo.cpp

if (CC == CallingConv::AnyReg) {
     return CSR_64_AllRegs_RegMask;
   }
 
   if (Subtarget.isDarwinABI())
     return TM.isPPC64() ? (Subtarget.hasAltivec() ? CSR_Darwin64_Altivec_RegMask
                                                   : CSR_Darwin64_RegMask)
                         : (Subtarget.hasAltivec() ? CSR_Darwin32_Altivec_RegMask
                                                   : CSR_Darwin32_RegMask);
+  if (Subtarget.isAIXABI()) {
+    assert(!Subtarget.hasAltivec() && "Altivec is not implemented on AIX yet.");
+    return TM.isPPC64() ? CSR_AIX64_RegMask : CSR_AIX32_RegMask;
+  }
 
   if (CC == CallingConv::Cold) {
     return TM.isPPC64() ? (Subtarget.hasAltivec() ? CSR_SVR64_ColdCC_Altivec_RegMask
                                                   : CSR_SVR64_ColdCC_RegMask)
                         : (Subtarget.hasAltivec() ? CSR_SVR32_ColdCC_Altivec_RegMask
                                                   : CSR_SVR32_ColdCC_RegMask);
   }
 

llvm/lib/Target/PowerPC/PPCSubtarget.h

public:
   /// isBGQ - True if this is a BG/Q platform.
   bool isBGQ() const { return TargetTriple.getVendor() == Triple::BGQ; }
 
   bool isTargetELF() const { return TargetTriple.isOSBinFormatELF(); }
   bool isTargetMachO() const { return TargetTriple.isOSBinFormatMachO(); }
   bool isTargetLinux() const { return TargetTriple.isOSLinux(); }
 
   bool isDarwinABI() const { return isTargetMachO() || isDarwin(); }
-  bool isSVR4ABI() const { return !isDarwinABI(); }
+  bool isAIXABI() const { return TargetTriple.isOSAIX(); }
+  bool isSVR4ABI() const { return !isDarwinABI() && !isAIXABI(); }
   bool isELFv2ABI() const;
 
   /// Originally, this function return hasISEL(). Now we always enable it,
   /// but may expand the ISEL instruction later.
   bool enableEarlyIfConversion() const override { return true; }
 
   // Scheduling customization.
   bool enableMachineScheduler() const override;

llvm/lib/Target/TargetMachine.cpp

if (GV && !GV->hasDefaultVisibility())
     return true;
 
   if (TT.isOSBinFormatMachO()) {
     if (RM == Reloc::Static)
       return true;
     return GV && GV->isStrongDefinitionForLinker();
   }
 
-  assert(TT.isOSBinFormatELF() || TT.isOSBinFormatWasm());
+  // Due to the AIX linkage model, any global with default visibility is 
+  // considered non-local.
+  if (TT.isOSBinFormatXCOFF())
+    return false;
+
+  assert(TT.isOSBinFormatELF());
   assert(RM != Reloc::DynamicNoPIC);
 
   bool IsExecutable =
       RM == Reloc::Static || M.getPIELevel() != PIELevel::Default;
   if (IsExecutable) {
     // If the symbol is defined, it cannot be preempted.
     if (GV && !GV->isDeclarationForLinker())
       return true;

llvm/test/CodeGen/PowerPC/test_call_aix.ll

+; RUN: llc -mtriple powerpc-ibm-aix-xcoff -stop-after=machine-cp < %s | \ 
+; RUN: FileCheck --check-prefix=32BIT %s
+
+; RUN: llc -mtriple powerpc64-ibm-aix-xcoff -stop-after=machine-cp < %s | \
+; RUN: FileCheck --check-prefix=64BIT %s
+
+declare void @foo(...)
+
+define void @test_call() {
+entry:
+; 32BIT: ADJCALLSTACKDOWN 56, 0, implicit-def dead $r1, implicit $r1
+; 32BIT: BL_NOP @foo, csr_aix32, implicit-def dead $lr, implicit $rm, implicit-def $r1
+; 32BIT: ADJCALLSTACKUP 56, 0, implicit-def dead $r1, implicit $r1
+
+; 64BIT: ADJCALLSTACKDOWN 112, 0, implicit-def dead $r1, implicit $r1
+; 64BIT: BL8_NOP @foo, csr_aix64, implicit-def dead $lr8, implicit $rm, implicit-def $r1
+; 64BIT: ADJCALLSTACKUP 112, 0, implicit-def dead $r1, implicit $r1
+
+  call void bitcast (void (...)* @foo to void ()*)()
+  ret void
+}
+
+define hidden void @foo_local() {
+entry:
+  ret void
+}
+
+define void @test_local_call() {
+entry:
+; 32BIT: ADJCALLSTACKDOWN 56, 0, implicit-def dead $r1, implicit $r1
+; 32BIT: BL @foo_local, csr_aix32, implicit-def dead $lr, implicit $rm, implicit-def $r1
+; 32BIT: ADJCALLSTACKUP 56, 0, implicit-def dead $r1, implicit $r1
+
+; 64BIT: ADJCALLSTACKDOWN 112, 0, implicit-def dead $r1, implicit $r1
+; 64BIT: BL8 @foo_local, csr_aix64, implicit-def dead $lr8, implicit $rm, implicit-def $r1
+; 64BIT: ADJCALLSTACKUP 112, 0, implicit-def dead $r1, implicit $r1
+
+  call void @foo_local()
+  ret void
+}
+