[RISCV] Support lowering FrameIndex

[asb]

Summary

Introduces the AddrFI "addressing mode", which is necessary simple because it's not possible to write a pattern that directly matches a frameindex.

Ensures callee-saved registers are accessed relative to the stackpointer. This is necessary as callee saved register spills are performed before the frame pointer is set.

[mgrang]

lib/Target/RISCV/RISCVInstrInfo.td

333 ↗

(On Diff #122260)

nit: Period after comment.

[mgrang]

lib/Target/RISCV/RISCVRegisterInfo.cpp

74 ↗

(On Diff #122260)

This will have an "unused variable 'Reg'" error in a non-asserts build. No?

[asb]

Thanks Mandeep, updated the patch. I've moved a bit of logic from the successor patch (D39849) to this one, as I think it makes more sense to use the correct register in eliminateFrameIndex from the start, rather than adding the extra logic in the next patch.

lib/Target/RISCV/RISCVRegisterInfo.cpp

74 ↗

(On Diff #122260)

No, Reg gets used in the code below.

[asb]

Rebase on latest LLVM and remove redundant return in eliminateFrameIndex.

[apazos]

lib/Target/RISCV/RISCVISelDAGToDAG.cpp

94 ↗

(On Diff #122635)

too many ( )

lib/Target/RISCV/RISCVRegisterInfo.cpp

90 ↗

(On Diff #122635)

you can make this the default value and eliminate the else.

122 ↗

(On Diff #122635)

maybe move default to the first check.

[asb]

Updated patch to address review comments.

lib/Target/RISCV/RISCVISelDAGToDAG.cpp

94 ↗

(On Diff #122635)

The extra parens are actually necessary to avoid a compiler warning "warning: using the result of an assignment as a condition without parentheses [-Wparentheses]". This is a fairly ugly idiom and we can avoid the need for it altogether, so I've refactored the code to do so.

[sabuasal]

Hi @asb

Can you point me to a documentation for the addressing modes names you are using (ADDRii, MEMii ..) ? I can't find these on the GitHub page.

[asb]

I should add a clarifying comment. These are codegen-only constructs that as far as I can see are required for frameindex lowering, in order to represent a frameindex+offset pair.

[kparzysz]

Why do you want the special load/store instructions?

lib/Target/RISCV/RISCVRegisterInfo.cpp

71 ↗

(On Diff #122789)

You should implement getFrameIndexReference in TargetFrameLowering (in any case). You could use it here instead of all this logic.

[asb]

Rebased and updated to implement getFrameIndexReference.

@kparzysz honestly I don't really want the special load/store instructions, but was unable to write a pattern that would allow frameindex+offset to be matched by the normal load/store. Perhaps I'm missing something obvious?

[sabuasal]

hi Alex,

this doesn't apply for me on ToT. what version is it re based on?

[asb]

Just tested again with r318797. Applies cleanly, compiles, all tests pass.

[kparzysz]

FrameIndex cannot be matched directly in patterns, but it can be matched by a ComplexPattern, which then can be used in selection patterns.

We have this in Hexagon (the relevant part is the getTargetFrameIndex):

bool HexagonDAGToDAGISel::SelectAddrFI(SDValue &N, SDValue &R) {
  if (N.getOpcode() != ISD::FrameIndex)
    return false;
  auto &HFI = *HST->getFrameLowering();
  MachineFrameInfo &MFI = MF->getFrameInfo();
  int FX = cast<FrameIndexSDNode>(N)->getIndex();
  if (!MFI.isFixedObjectIndex(FX) && HFI.needsAligna(*MF))
    return false;
  R = CurDAG->getTargetFrameIndex(FX, MVT::i32);
  return true;
}

Then in HexagonPatterns.td:

def AddrFI: ComplexPattern<i32, 1, "SelectAddrFI", [frameindex], []>;

And then some Pat:

def: Pat<(VT (Load (add (i32 AddrFI:$fi), ImmPred:$Off))),
         (VT (MI AddrFI:$fi, imm:$Off))>;

[asb]

In D39848#932109, @kparzysz wrote:

FrameIndex cannot be matched directly in patterns, but it can be matched by a ComplexPattern, which then can be used in selection patterns.

We have this in Hexagon (the relevant part is the getTargetFrameIndex):

Thanks Krzysztof. This patch does something similar, in that I define a ComplexPattern in order to match a FrameIndex. The bit I remember having problems with was using the operand produced by the complex pattern with my targets standard GPR+imm load instructions. Of course my complexpattern was trying to get both the base and offset from the frameindex. Sticking to _just_ matching the frameindex in the ComplexPattern like your Hexagon snippets does seems more promising. I'll have another look tomorrow.

[asb]

I've had a good look at the alternative approach, whereby a ComplexPattern is used that _just_ matches the FrameIndex and extra load and store tablegen patterns are written using that. This also requires lowering FrameIndex in RISCVDAGToDAGISel::Select. If going that way, it probably makes sense to move HexagonDAGToDAGISel::isOrEquivalentToAdd to target-independent code rather than copying it in to RISCVDagToDAGISel.

I've managed to achieve identical codegen apart from one regression, where a simple varargs test case starts with:

t0: ch = EntryToken
      t39: i32 = add nuw FrameIndex:i32<-1>, Constant:i32<4>
    t55: ch = store<ST4[%va]> t0, t39, FrameIndex:i32<0>, undef:i32
      t9: i32,ch = CopyFromReg t0, Register:i32 %vreg2
    t11: ch = store<ST4[FixedStack-4](align=8)> t0, t9, FrameIndex:i32<-4>, undef:i32
      t13: i32,ch = CopyFromReg t0, Register:i32 %vreg3
    t15: ch = store<ST4[<unknown>]> t0, t13, FrameIndex:i32<-5>, undef:i32
      t17: i32,ch = CopyFromReg t0, Register:i32 %vreg4
    t19: ch = store<ST4[FixedStack-6](align=16)> t0, t17, FrameIndex:i32<-6>, undef:i32
      t21: i32,ch = CopyFromReg t0, Register:i32 %vreg5
    t23: ch = store<ST4[<unknown>]> t0, t21, FrameIndex:i32<-7>, undef:i32
      t25: i32,ch = CopyFromReg t0, Register:i32 %vreg6
    t27: ch = store<ST4[FixedStack-8](align=8)> t0, t25, FrameIndex:i32<-8>, undef:i32
      t29: i32,ch = CopyFromReg t0, Register:i32 %vreg7
    t31: ch = store<ST4[<unknown>]> t0, t29, FrameIndex:i32<-9>, undef:i32
  t60: ch = TokenFactor t58:1, t55, t11, t15, t19, t23, t27, t31
t44: ch,glue = CopyToReg t60, Register:i32 %X10, t58
    t4: i32,ch = CopyFromReg t0, Register:i32 %vreg1
  t7: ch = store<ST4[<unknown>]> t0, t4, FrameIndex:i32<-3>, undef:i32
t58: i32,ch = load<LD4[%2]> t7, FrameIndex:i32<-1>, undef:i32
t45: ch = RISCVISD::RET_FLAG t44, Register:i32 %X10, t44:1

ends up as the following (two addi when one would suffice):

	%vreg7<def> = COPY %X17; GPR:%vreg7
	%vreg6<def> = COPY %X16; GPR:%vreg6
	%vreg5<def> = COPY %X15; GPR:%vreg5
	%vreg4<def> = COPY %X14; GPR:%vreg4
	%vreg3<def> = COPY %X13; GPR:%vreg3
	%vreg2<def> = COPY %X12; GPR:%vreg2
	%vreg1<def> = COPY %X11; GPR:%vreg1
	SW %vreg1, <fi#-3>, 0; mem:ST4[<unknown>] GPR:%vreg1
	SW %vreg7, <fi#-9>, 0; mem:ST4[<unknown>] GPR:%vreg7
	SW %vreg6, <fi#-8>, 0; mem:ST4[FixedStack-8](align=8) GPR:%vreg6
	SW %vreg5, <fi#-7>, 0; mem:ST4[<unknown>] GPR:%vreg5
	SW %vreg4, <fi#-6>, 0; mem:ST4[FixedStack-6](align=16) GPR:%vreg4
	SW %vreg3, <fi#-5>, 0; mem:ST4[<unknown>] GPR:%vreg3
	SW %vreg2, <fi#-4>, 0; mem:ST4[FixedStack-4](align=8) GPR:%vreg2
	%vreg8<def> = ADDI <fi#-1>, 0; GPR:%vreg8
	%vreg9<def> = ADDI %vreg8<kill>, 4; GPR:%vreg9,%vreg8
	SW %vreg9<kill>, <fi#0>, 0; mem:ST4[%va] GPR:%vreg9
	%vreg10<def> = LW <fi#-1>, 0; mem:LD4[%2] GPR:%vreg10
	%X10<def> = COPY %vreg10; GPR:%vreg10
	PseudoRET %X10<imp-use>

One the one hand, the RISC-V backend is starting by focusing on correct and well tested output, with performance work coming later. This means sub-optimal codegen can be tolerated initially. On the other hand, it seems a shame to regress code generation in this way.

I'll have a look if there's a solution that doesn't result in too much additional complexity.

[asb]

Just to update on my previous comment and to make sure nobody wastes time on this: looking at the issue with a fresh pair of eyes this morning the selection issues were rather straight-forward to resolve. Krzysztof's proposed solution seems cleaner overall and is now working as a drop-in replacement for the RV32I tests. I'll clean up my implementation, test against the full RV32IMAFD+RV64IMAFD patchset and refresh this patch.

[asb]

Rewrite to avoid the need for the fake *_FI instructions. I also move isOrEquivalentToAdd from HexagonISelDAGToDAG to the SelectionDAGISel base class, so we can make use of it.

[asb]

No patch changes, just refreshing the diff. If we can get this reviewed, it unblocks the following two patches which already have reviews.

[sabuasal]

LGTM.

@kparzysz should probably take a look too

[kparzysz]

lib/Target/RISCV/RISCVInstrInfo.td

119 ↗

(On Diff #125373)

This looks unused. Can it be removed?

[asb]

Rebase the patch and remove MEMii which is no longer needed.

[asb]

@kparzysz, does this look good to you now?

[kparzysz]

Yes, LGTM.

llvm/trunk/include/llvm/CodeGen/SelectionDAGISel.h

void SelectCodeCommon(SDNode *NodeToMatch, const unsigned char *MatcherTable,
                         unsigned TableSize);
 
   /// \brief Return true if complex patterns for this target can mutate the
   /// DAG.
   virtual bool ComplexPatternFuncMutatesDAG() const {
     return false;
   }
 
+  bool isOrEquivalentToAdd(const SDNode *N) const;
+
 private:
 
   // Calls to these functions are generated by tblgen.
   void Select_INLINEASM(SDNode *N);
   void Select_READ_REGISTER(SDNode *N);
   void Select_WRITE_REGISTER(SDNode *N);
   void Select_UNDEF(SDNode *N);
   void CannotYetSelect(SDNode *N);

llvm/trunk/lib/CodeGen/SelectionDAG/SelectionDAGISel.cpp

while (true) {
 
       // End of this scope, pop it and try the next child in the containing
       // scope.
       MatchScopes.pop_back();
     }
   }
 }
 
+bool SelectionDAGISel::isOrEquivalentToAdd(const SDNode *N) const {
+  assert(N->getOpcode() == ISD::OR && "Unexpected opcode");
+  auto *C = dyn_cast<ConstantSDNode>(N->getOperand(1));
+  if (!C)
+    return false;
+
+  // Detect when "or" is used to add an offset to a stack object.
+  if (auto *FN = dyn_cast<FrameIndexSDNode>(N->getOperand(0))) {
+    MachineFrameInfo &MFI = MF->getFrameInfo();
+    unsigned A = MFI.getObjectAlignment(FN->getIndex());
+    assert(isPowerOf2_32(A) && "Unexpected alignment");
+    int32_t Off = C->getSExtValue();
+    // If the alleged offset fits in the zero bits guaranteed by
+    // the alignment, then this or is really an add.
+    return (Off >= 0) && (((A - 1) & Off) == unsigned(Off));
+  }
+  return false;
+}
+
 void SelectionDAGISel::CannotYetSelect(SDNode *N) {
   std::string msg;
   raw_string_ostream Msg(msg);
   Msg << "Cannot select: ";
 
   if (N->getOpcode() != ISD::INTRINSIC_W_CHAIN &&
       N->getOpcode() != ISD::INTRINSIC_WO_CHAIN &&
       N->getOpcode() != ISD::INTRINSIC_VOID) {

llvm/trunk/lib/Target/Hexagon/HexagonISelDAGToDAG.h

SDValue selectUndef(const SDLoc &dl, MVT ResTy) {
     SDNode *U = CurDAG->getMachineNode(TargetOpcode::IMPLICIT_DEF, dl, ResTy);
     return SDValue(U, 0);
   }
 
   void SelectHvxShuffle(SDNode *N);
   void SelectHvxRor(SDNode *N);
 
   bool keepsLowBits(const SDValue &Val, unsigned NumBits, SDValue &Src);
-  bool isOrEquivalentToAdd(const SDNode *N) const;
   bool isAlignedMemNode(const MemSDNode *N) const;
   bool isSmallStackStore(const StoreSDNode *N) const;
   bool isPositiveHalfWord(const SDNode *N) const;
   bool hasOneUse(const SDNode *N) const;
 
   // DAG preprocessing functions.
   void ppSimplifyOrSelect0(std::vector<SDNode*> &&Nodes);
   void ppAddrReorderAddShl(std::vector<SDNode*> &&Nodes);

llvm/trunk/lib/Target/Hexagon/HexagonISelDAGToDAG.cpp

case ISD::XOR: {
     }
   }
   default:
     break;
   }
   return false;
 }
 
-
-bool HexagonDAGToDAGISel::isOrEquivalentToAdd(const SDNode *N) const {
-  assert(N->getOpcode() == ISD::OR);
-  auto *C = dyn_cast<ConstantSDNode>(N->getOperand(1));
-  if (!C)
-    return false;
-
-  // Detect when "or" is used to add an offset to a stack object.
-  if (auto *FN = dyn_cast<FrameIndexSDNode>(N->getOperand(0))) {
-    MachineFrameInfo &MFI = MF->getFrameInfo();
-    unsigned A = MFI.getObjectAlignment(FN->getIndex());
-    assert(isPowerOf2_32(A));
-    int32_t Off = C->getSExtValue();
-    // If the alleged offset fits in the zero bits guaranteed by
-    // the alignment, then this or is really an add.
-    return (Off >= 0) && (((A-1) & Off) == unsigned(Off));
-  }
-  return false;
-}
-
 bool HexagonDAGToDAGISel::isAlignedMemNode(const MemSDNode *N) const {
   return N->getAlignment() >= N->getMemoryVT().getStoreSize();
 }
 
 bool HexagonDAGToDAGISel::isSmallStackStore(const StoreSDNode *N) const {
   unsigned StackSize = MF->getFrameInfo().estimateStackSize(*MF);
   switch (N->getMemoryVT().getStoreSize()) {
     case 1:

llvm/trunk/lib/Target/RISCV/RISCVFrameLowering.h

public:
   explicit RISCVFrameLowering(const RISCVSubtarget &STI)
       : TargetFrameLowering(StackGrowsDown,
                             /*StackAlignment=*/16,
                             /*LocalAreaOffset=*/0) {}
 
   void emitPrologue(MachineFunction &MF, MachineBasicBlock &MBB) const override;
   void emitEpilogue(MachineFunction &MF, MachineBasicBlock &MBB) const override;
 
+  int getFrameIndexReference(const MachineFunction &MF, int FI,
+                             unsigned &FrameReg) const override;
+
   bool hasFP(const MachineFunction &MF) const override;
 
   MachineBasicBlock::iterator
   eliminateCallFramePseudoInstr(MachineFunction &MF, MachineBasicBlock &MBB,
                                 MachineBasicBlock::iterator MI) const override {
     return MBB.erase(MI);
   }
 };
 }
 #endif

llvm/trunk/lib/Target/RISCV/RISCVFrameLowering.cpp

 
 bool RISCVFrameLowering::hasFP(const MachineFunction &MF) const { return true; }
 
 void RISCVFrameLowering::emitPrologue(MachineFunction &MF,
                                       MachineBasicBlock &MBB) const {}
 
 void RISCVFrameLowering::emitEpilogue(MachineFunction &MF,
                                       MachineBasicBlock &MBB) const {}
+
+int RISCVFrameLowering::getFrameIndexReference(const MachineFunction &MF,
+                                               int FI,
+                                               unsigned &FrameReg) const {
+  const MachineFrameInfo &MFI = MF.getFrameInfo();
+  const TargetRegisterInfo *RI = MF.getSubtarget().getRegisterInfo();
+
+  // Callee-saved registers should be referenced relative to the stack
+  // pointer (positive offset), otherwise use the frame pointer (negative
+  // offset).
+  const std::vector<CalleeSavedInfo> &CSI = MFI.getCalleeSavedInfo();
+  int MinCSFI = 0;
+  int MaxCSFI = -1;
+
+  int Offset = MFI.getObjectOffset(FI) - getOffsetOfLocalArea() +
+               MFI.getOffsetAdjustment();
+
+  if (CSI.size()) {
+    MinCSFI = CSI[0].getFrameIdx();
+    MaxCSFI = CSI[CSI.size() - 1].getFrameIdx();
+  }
+
+  FrameReg = RI->getFrameRegister(MF);
+  if (FI >= MinCSFI && FI <= MaxCSFI) {
+    FrameReg = RISCV::X2;
+    Offset += MF.getFrameInfo().getStackSize();
+  }
+  return Offset;
+}

llvm/trunk/lib/Target/RISCV/RISCVISelDAGToDAG.cpp

 //===-- RISCVISelDAGToDAG.cpp - A dag to dag inst selector for RISCV ------===//
 //
 //                     The LLVM Compiler Infrastructure
 //
 // This file is distributed under the University of Illinois Open Source
 // License. See LICENSE.TXT for details.
 //
 //===----------------------------------------------------------------------===//
 //
 // This file defines an instruction selector for the RISCV target.
 //
 //===----------------------------------------------------------------------===//
 
 #include "RISCV.h"
 #include "MCTargetDesc/RISCVMCTargetDesc.h"
 #include "RISCVTargetMachine.h"
+#include "llvm/CodeGen/MachineFrameInfo.h"
 #include "llvm/CodeGen/SelectionDAGISel.h"
 #include "llvm/Support/Debug.h"
 #include "llvm/Support/MathExtras.h"
 #include "llvm/Support/raw_ostream.h"
 using namespace llvm;
 
 #define DEBUG_TYPE "riscv-isel"
 
public:
 
   bool runOnMachineFunction(MachineFunction &MF) override {
     Subtarget = &MF.getSubtarget<RISCVSubtarget>();
     return SelectionDAGISel::runOnMachineFunction(MF);
   }
 
   void Select(SDNode *Node) override;
 
+  bool SelectAddrFI(SDValue Addr, SDValue &Base);
+
 // Include the pieces autogenerated from the target description.
 #include "RISCVGenDAGISel.inc"
 };
 }
 
 void RISCVDAGToDAGISel::Select(SDNode *Node) {
   unsigned Opcode = Node->getOpcode();
   MVT XLenVT = Subtarget->getXLenVT();
if (Opcode == ISD::Constant && VT == XLenVT) {
     // to propagate these into other instructions.
     if (ConstNode->isNullValue()) {
       SDValue New = CurDAG->getCopyFromReg(CurDAG->getEntryNode(), SDLoc(Node),
                                            RISCV::X0, XLenVT);
       ReplaceNode(Node, New.getNode());
       return;
     }
   }
+  if (Opcode == ISD::FrameIndex) {
+    SDLoc DL(Node);
+    SDValue Imm = CurDAG->getTargetConstant(0, DL, XLenVT);
+    int FI = dyn_cast<FrameIndexSDNode>(Node)->getIndex();
+    EVT VT = Node->getValueType(0);
+    SDValue TFI = CurDAG->getTargetFrameIndex(FI, VT);
+    ReplaceNode(Node, CurDAG->getMachineNode(RISCV::ADDI, DL, VT, TFI, Imm));
+    return;
+  }
 
   // Select the default instruction.
   SelectCode(Node);
 }
 
+bool RISCVDAGToDAGISel::SelectAddrFI(SDValue Addr, SDValue &Base) {
+  if (auto FIN = dyn_cast<FrameIndexSDNode>(Addr)) {
+    Base = CurDAG->getTargetFrameIndex(FIN->getIndex(), Subtarget->getXLenVT());
+    return true;
+  }
+  return false;
+}
+
 // This pass converts a legalized DAG into a RISCV-specific DAG, ready
 // for instruction scheduling.
 FunctionPass *llvm::createRISCVISelDag(RISCVTargetMachine &TM) {
   return new RISCVDAGToDAGISel(TM);
 }

llvm/trunk/lib/Target/RISCV/RISCVInstrInfo.td

 }
 
 // A parameterized register class alternative to i32imm/i64imm from Target.td.
 def ixlenimm : Operand<XLenVT>;
 
 // Standalone (codegen-only) immleaf patterns.
 def simm32 : ImmLeaf<XLenVT, [{return isInt<32>(Imm);}]>;
 
+// Addressing modes.
+// Necessary because a frameindex can't be matched directly in a pattern.
+def AddrFI : ComplexPattern<iPTR, 1, "SelectAddrFI", [frameindex], []>;
+
 // Extract least significant 12 bits from an immediate value and sign extend
 // them.
 def LO12Sext : SDNodeXForm<imm, [{
   return CurDAG->getTargetConstant(SignExtend64<12>(N->getZExtValue()),
                                    SDLoc(N), N->getValueType(0));
 }]>;
 
 // Extract the most significant 20 bits from an immediate value. Add 1 if bit

 class PatGprGpr<SDPatternOperator OpNode, RVInstR Inst>
     : Pat<(OpNode GPR:$rs1, GPR:$rs2), (Inst GPR:$rs1, GPR:$rs2)>;
 class PatGprSimm12<SDPatternOperator OpNode, RVInstI Inst>
     : Pat<(OpNode GPR:$rs1, simm12:$imm12), (Inst GPR:$rs1, simm12:$imm12)>;
 class PatGprUimmLog2XLen<SDPatternOperator OpNode, RVInstIShift Inst>
     : Pat<(OpNode GPR:$rs1, uimmlog2xlen:$shamt),
           (Inst GPR:$rs1, uimmlog2xlen:$shamt)>;
 
+/// Predicates
+
+def IsOrAdd: PatFrag<(ops node:$A, node:$B), (or node:$A, node:$B), [{
+  return isOrEquivalentToAdd(N);
+}]>;
+
 /// Immediates
 
 def : Pat<(simm12:$imm), (ADDI X0, simm12:$imm)>;
 // TODO: Add a pattern for immediates with all zeroes in the lower 12 bits.
 def : Pat<(simm32:$imm), (ADDI (LUI (HI20 imm:$imm)), (LO12Sext imm:$imm))>;
 
 /// Simple arithmetic operations
 
 def : PatGprGpr<add, ADD>;
 def : PatGprSimm12<add, ADDI>;
 def : PatGprGpr<sub, SUB>;
 def : PatGprGpr<or, OR>;
 def : PatGprSimm12<or, ORI>;
 def : PatGprGpr<and, AND>;
 def : PatGprSimm12<and, ANDI>;
 def : PatGprGpr<xor, XOR>;
 def : PatGprSimm12<xor, XORI>;
 def : PatGprGpr<shl, SLL>;
 def : PatGprUimmLog2XLen<shl, SLLI>;
 def : PatGprGpr<srl, SRL>;
 def : PatGprUimmLog2XLen<srl, SRLI>;
 def : PatGprGpr<sra, SRA>;
 def : PatGprUimmLog2XLen<sra, SRAI>;
 
+/// FrameIndex calculations
+
+def : Pat<(add (i32 AddrFI:$Rs), simm12:$imm12),
+          (ADDI (i32 AddrFI:$Rs), simm12:$imm12)>;
+def : Pat<(IsOrAdd (i32 AddrFI:$Rs), simm12:$imm12),
+          (ADDI (i32 AddrFI:$Rs), simm12:$imm12)>;
+
 /// Setcc
 
 def : PatGprGpr<setlt, SLT>;
 def : PatGprSimm12<setlt, SLTI>;
 def : PatGprGpr<setult, SLTU>;
 def : PatGprSimm12<setult, SLTIU>;
 
 // Define pattern expansions for setcc operations that aren't directly

 let isBarrier = 1, isReturn = 1, isTerminator = 1 in
 def PseudoRET : Pseudo<(outs), (ins), [(RetFlag)]>,
                 PseudoInstExpansion<(JALR X0, X1, 0)>;
 
 /// Loads
 
 multiclass LdPat<PatFrag LoadOp, RVInst Inst> {
   def : Pat<(LoadOp GPR:$rs1), (Inst GPR:$rs1, 0)>;
+  def : Pat<(LoadOp AddrFI:$rs1), (Inst AddrFI:$rs1, 0)>;
   def : Pat<(LoadOp (add GPR:$rs1, simm12:$imm12)),
             (Inst GPR:$rs1, simm12:$imm12)>;
+  def : Pat<(LoadOp (add AddrFI:$rs1, simm12:$imm12)),
+            (Inst AddrFI:$rs1, simm12:$imm12)>;
+  def : Pat<(LoadOp (IsOrAdd AddrFI:$rs1, simm12:$imm12)),
+            (Inst AddrFI:$rs1, simm12:$imm12)>;
 }
 
 defm : LdPat<sextloadi8, LB>;
 defm : LdPat<extloadi8, LB>;
 defm : LdPat<sextloadi16, LH>;
 defm : LdPat<extloadi16, LH>;
 defm : LdPat<load, LW>;
 defm : LdPat<zextloadi8, LBU>;
 defm : LdPat<zextloadi16, LHU>;
 
 /// Stores
 
 multiclass StPat<PatFrag StoreOp, RVInst Inst> {
   def : Pat<(StoreOp GPR:$rs2, GPR:$rs1), (Inst GPR:$rs2, GPR:$rs1, 0)>;
+  def : Pat<(StoreOp GPR:$rs2, AddrFI:$rs1), (Inst GPR:$rs2, AddrFI:$rs1, 0)>;
   def : Pat<(StoreOp GPR:$rs2, (add GPR:$rs1, simm12:$imm12)),
             (Inst GPR:$rs2, GPR:$rs1, simm12:$imm12)>;
+  def : Pat<(StoreOp GPR:$rs2, (add AddrFI:$rs1, simm12:$imm12)),
+            (Inst GPR:$rs2, AddrFI:$rs1, simm12:$imm12)>;
+  def : Pat<(StoreOp GPR:$rs2, (IsOrAdd AddrFI:$rs1, simm12:$imm12)),
+            (Inst GPR:$rs2, AddrFI:$rs1, simm12:$imm12)>;
 }
 
 defm : StPat<truncstorei8, SB>;
 defm : StPat<truncstorei16, SH>;
 defm : StPat<store, SW>;
 
 /// Other pseudo-instructions
 

llvm/trunk/lib/Target/RISCV/RISCVRegisterInfo.cpp

 
 const uint32_t *RISCVRegisterInfo::getNoPreservedMask() const {
   return CSR_NoRegs_RegMask;
 }
 
 void RISCVRegisterInfo::eliminateFrameIndex(MachineBasicBlock::iterator II,
                                             int SPAdj, unsigned FIOperandNum,
                                             RegScavenger *RS) const {
-  // TODO: this implementation is a temporary placeholder which does just
-  // enough to allow other aspects of code generation to be tested
-
   assert(SPAdj == 0 && "Unexpected non-zero SPAdj value");
 
   MachineInstr &MI = *II;
   MachineFunction &MF = *MI.getParent()->getParent();
-  const TargetFrameLowering *TFI = MF.getSubtarget().getFrameLowering();
   DebugLoc DL = MI.getDebugLoc();
 
-  unsigned FrameReg = getFrameRegister(MF);
   int FrameIndex = MI.getOperand(FIOperandNum).getIndex();
-  int Offset = TFI->getFrameIndexReference(MF, FrameIndex, FrameReg);
-  Offset += MI.getOperand(FIOperandNum + 1).getImm();
+  unsigned FrameReg;
+  int Offset =
+      getFrameLowering(MF)->getFrameIndexReference(MF, FrameIndex, FrameReg) +
+      MI.getOperand(FIOperandNum + 1).getImm();
 
-  assert(TFI->hasFP(MF) && "eliminateFrameIndex currently requires hasFP");
+  assert(MF.getSubtarget().getFrameLowering()->hasFP(MF) &&
+         "eliminateFrameIndex currently requires hasFP");
 
   // Offsets must be directly encoded in a 12-bit immediate field
   if (!isInt<12>(Offset)) {
     report_fatal_error(
         "Frame offsets outside of the signed 12-bit range not supported");
   }
 
   MI.getOperand(FIOperandNum).ChangeToRegister(FrameReg, false);

llvm/trunk/test/CodeGen/RISCV/blockaddress.ll

 ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
 ; RUN: llc -mtriple=riscv32 -verify-machineinstrs < %s \
 ; RUN:   | FileCheck %s -check-prefix=RV32I
 
 @addr = global i8* null
 
 define void @test_blockaddress() nounwind {
 ; RV32I-LABEL: test_blockaddress:
 ; RV32I:       # %bb.0:
-; RV32I-NEXT:    sw ra, 0(s0)
+; RV32I-NEXT:    sw ra, 0(sp)
 ; RV32I-NEXT:    lui a0, %hi(addr)
 ; RV32I-NEXT:    addi a0, a0, %lo(addr)
 ; RV32I-NEXT:    lui a1, %hi(.Ltmp0)
 ; RV32I-NEXT:    addi a1, a1, %lo(.Ltmp0)
 ; RV32I-NEXT:    sw a1, 0(a0)
 ; RV32I-NEXT:    lw a0, 0(a0)
 ; RV32I-NEXT:    jalr zero, a0, 0
 ; RV32I-NEXT:  .Ltmp0: # Block address taken
 ; RV32I-NEXT:  .LBB0_1: # %block
-; RV32I-NEXT:    lw ra, 0(s0)
+; RV32I-NEXT:    lw ra, 0(sp)
 ; RV32I-NEXT:    jalr zero, ra, 0
   store volatile i8* blockaddress(@test_blockaddress, %block), i8** @addr
   %val = load volatile i8*, i8** @addr
   indirectbr i8* %val, [label %block]
 
 block:
   ret void
 }

llvm/trunk/test/CodeGen/RISCV/bswap-ctlz-cttz-ctpop.ll

 ; RV32I-NEXT:    jalr zero, ra, 0
   %tmp = call i64 @llvm.bswap.i64(i64 %a)
   ret i64 %tmp
 }
 
 define i8 @test_cttz_i8(i8 %a) nounwind {
 ; RV32I-LABEL: test_cttz_i8:
 ; RV32I:       # %bb.0:
-; RV32I-NEXT:    sw ra, 12(s0)
+; RV32I-NEXT:    sw ra, 12(sp)
 ; RV32I-NEXT:    addi a1, a0, 0
 ; RV32I-NEXT:    addi a0, zero, 8
 ; RV32I-NEXT:    andi a2, a1, 255
 ; RV32I-NEXT:    beq a2, zero, .LBB3_2
 ; RV32I-NEXT:    jal zero, .LBB3_1
 ; RV32I-NEXT:  .LBB3_1: # %cond.false
 ; RV32I-NEXT:    addi a0, a1, -1
 ; RV32I-NEXT:    xori a1, a1, -1

 ; RV32I-NEXT:    and a0, a0, a1
 ; RV32I-NEXT:    lui a1, 4112
 ; RV32I-NEXT:    addi a1, a1, 257
 ; RV32I-NEXT:    lui a2, %hi(__mulsi3)
 ; RV32I-NEXT:    addi a2, a2, %lo(__mulsi3)
 ; RV32I-NEXT:    jalr ra, a2, 0
 ; RV32I-NEXT:    srli a0, a0, 24
 ; RV32I-NEXT:  .LBB3_2: # %cond.end
-; RV32I-NEXT:    lw ra, 12(s0)
+; RV32I-NEXT:    lw ra, 12(sp)
 ; RV32I-NEXT:    jalr zero, ra, 0
   %tmp = call i8 @llvm.cttz.i8(i8 %a, i1 false)
   ret i8 %tmp
 }
 
 define i16 @test_cttz_i16(i16 %a) nounwind {
 ; RV32I-LABEL: test_cttz_i16:
 ; RV32I:       # %bb.0:
-; RV32I-NEXT:    sw ra, 12(s0)
+; RV32I-NEXT:    sw ra, 12(sp)
 ; RV32I-NEXT:    addi a1, a0, 0
 ; RV32I-NEXT:    addi a0, zero, 16
 ; RV32I-NEXT:    lui a2, 16
 ; RV32I-NEXT:    addi a2, a2, -1
 ; RV32I-NEXT:    and a2, a1, a2
 ; RV32I-NEXT:    beq a2, zero, .LBB4_2
 ; RV32I-NEXT:    jal zero, .LBB4_1
 ; RV32I-NEXT:  .LBB4_1: # %cond.false

 ; RV32I-NEXT:    and a0, a0, a1
 ; RV32I-NEXT:    lui a1, 4112
 ; RV32I-NEXT:    addi a1, a1, 257
 ; RV32I-NEXT:    lui a2, %hi(__mulsi3)
 ; RV32I-NEXT:    addi a2, a2, %lo(__mulsi3)
 ; RV32I-NEXT:    jalr ra, a2, 0
 ; RV32I-NEXT:    srli a0, a0, 24
 ; RV32I-NEXT:  .LBB4_2: # %cond.end
-; RV32I-NEXT:    lw ra, 12(s0)
+; RV32I-NEXT:    lw ra, 12(sp)
 ; RV32I-NEXT:    jalr zero, ra, 0
   %tmp = call i16 @llvm.cttz.i16(i16 %a, i1 false)
   ret i16 %tmp
 }
 
 define i32 @test_cttz_i32(i32 %a) nounwind {
 ; RV32I-LABEL: test_cttz_i32:
 ; RV32I:       # %bb.0:
-; RV32I-NEXT:    sw ra, 12(s0)
+; RV32I-NEXT:    sw ra, 12(sp)
 ; RV32I-NEXT:    addi a1, a0, 0
 ; RV32I-NEXT:    addi a0, zero, 32
 ; RV32I-NEXT:    beq a1, zero, .LBB5_2
 ; RV32I-NEXT:    jal zero, .LBB5_1
 ; RV32I-NEXT:  .LBB5_1: # %cond.false
 ; RV32I-NEXT:    addi a0, a1, -1
 ; RV32I-NEXT:    xori a1, a1, -1
 ; RV32I-NEXT:    and a0, a1, a0

 ; RV32I-NEXT:    and a0, a0, a1
 ; RV32I-NEXT:    lui a1, 4112
 ; RV32I-NEXT:    addi a1, a1, 257
 ; RV32I-NEXT:    lui a2, %hi(__mulsi3)
 ; RV32I-NEXT:    addi a2, a2, %lo(__mulsi3)
 ; RV32I-NEXT:    jalr ra, a2, 0
 ; RV32I-NEXT:    srli a0, a0, 24
 ; RV32I-NEXT:  .LBB5_2: # %cond.end
-; RV32I-NEXT:    lw ra, 12(s0)
+; RV32I-NEXT:    lw ra, 12(sp)
 ; RV32I-NEXT:    jalr zero, ra, 0
   %tmp = call i32 @llvm.cttz.i32(i32 %a, i1 false)
   ret i32 %tmp
 }
 
 define i32 @test_ctlz_i32(i32 %a) nounwind {
 ; RV32I-LABEL: test_ctlz_i32:
 ; RV32I:       # %bb.0:
-; RV32I-NEXT:    sw ra, 12(s0)
+; RV32I-NEXT:    sw ra, 12(sp)
 ; RV32I-NEXT:    addi a1, a0, 0
 ; RV32I-NEXT:    addi a0, zero, 32
 ; RV32I-NEXT:    beq a1, zero, .LBB6_2
 ; RV32I-NEXT:    jal zero, .LBB6_1
 ; RV32I-NEXT:  .LBB6_1: # %cond.false
 ; RV32I-NEXT:    srli a0, a1, 1
 ; RV32I-NEXT:    or a0, a1, a0
 ; RV32I-NEXT:    srli a1, a0, 2

 ; RV32I-NEXT:    and a0, a0, a1
 ; RV32I-NEXT:    lui a1, 4112
 ; RV32I-NEXT:    addi a1, a1, 257
 ; RV32I-NEXT:    lui a2, %hi(__mulsi3)
 ; RV32I-NEXT:    addi a2, a2, %lo(__mulsi3)
 ; RV32I-NEXT:    jalr ra, a2, 0
 ; RV32I-NEXT:    srli a0, a0, 24
 ; RV32I-NEXT:  .LBB6_2: # %cond.end
-; RV32I-NEXT:    lw ra, 12(s0)
+; RV32I-NEXT:    lw ra, 12(sp)
 ; RV32I-NEXT:    jalr zero, ra, 0
   %tmp = call i32 @llvm.ctlz.i32(i32 %a, i1 false)
   ret i32 %tmp
 }
 
 define i64 @test_cttz_i64(i64 %a) nounwind {
 ; RV32I-LABEL: test_cttz_i64:
 ; RV32I:       # %bb.0:
-; RV32I-NEXT:    sw ra, 28(s0)
-; RV32I-NEXT:    sw s1, 24(s0)
-; RV32I-NEXT:    sw s2, 20(s0)
-; RV32I-NEXT:    sw s3, 16(s0)
-; RV32I-NEXT:    sw s4, 12(s0)
-; RV32I-NEXT:    sw s5, 8(s0)
-; RV32I-NEXT:    sw s6, 4(s0)
-; RV32I-NEXT:    sw s7, 0(s0)
+; RV32I-NEXT:    sw ra, 28(sp)
+; RV32I-NEXT:    sw s1, 24(sp)
+; RV32I-NEXT:    sw s2, 20(sp)
+; RV32I-NEXT:    sw s3, 16(sp)
+; RV32I-NEXT:    sw s4, 12(sp)
+; RV32I-NEXT:    sw s5, 8(sp)
+; RV32I-NEXT:    sw s6, 4(sp)
+; RV32I-NEXT:    sw s7, 0(sp)
 ; RV32I-NEXT:    addi s1, a1, 0
 ; RV32I-NEXT:    addi s2, a0, 0
 ; RV32I-NEXT:    addi a0, s2, -1
 ; RV32I-NEXT:    xori a1, s2, -1
 ; RV32I-NEXT:    and a0, a1, a0
 ; RV32I-NEXT:    lui a1, 349525
 ; RV32I-NEXT:    addi s4, a1, 1365
 ; RV32I-NEXT:    srli a1, a0, 1

 ; RV32I-NEXT:    jalr ra, s6, 0
 ; RV32I-NEXT:    bne s2, zero, .LBB7_2
 ; RV32I-NEXT:  # %bb.1:
 ; RV32I-NEXT:    srli a0, a0, 24
 ; RV32I-NEXT:    addi s1, a0, 32
 ; RV32I-NEXT:  .LBB7_2:
 ; RV32I-NEXT:    addi a0, s1, 0
 ; RV32I-NEXT:    addi a1, zero, 0
-; RV32I-NEXT:    lw s7, 0(s0)
-; RV32I-NEXT:    lw s6, 4(s0)
-; RV32I-NEXT:    lw s5, 8(s0)
-; RV32I-NEXT:    lw s4, 12(s0)
-; RV32I-NEXT:    lw s3, 16(s0)
-; RV32I-NEXT:    lw s2, 20(s0)
-; RV32I-NEXT:    lw s1, 24(s0)
-; RV32I-NEXT:    lw ra, 28(s0)
+; RV32I-NEXT:    lw s7, 0(sp)
+; RV32I-NEXT:    lw s6, 4(sp)
+; RV32I-NEXT:    lw s5, 8(sp)
+; RV32I-NEXT:    lw s4, 12(sp)
+; RV32I-NEXT:    lw s3, 16(sp)
+; RV32I-NEXT:    lw s2, 20(sp)
+; RV32I-NEXT:    lw s1, 24(sp)
+; RV32I-NEXT:    lw ra, 28(sp)
 ; RV32I-NEXT:    jalr zero, ra, 0
   %tmp = call i64 @llvm.cttz.i64(i64 %a, i1 false)
   ret i64 %tmp
 }
 
 define i8 @test_cttz_i8_zero_undef(i8 %a) nounwind {
 ; RV32I-LABEL: test_cttz_i8_zero_undef:
 ; RV32I:       # %bb.0:
-; RV32I-NEXT:    sw ra, 12(s0)
+; RV32I-NEXT:    sw ra, 12(sp)
 ; RV32I-NEXT:    addi a1, a0, -1
 ; RV32I-NEXT:    xori a0, a0, -1
 ; RV32I-NEXT:    and a0, a0, a1
 ; RV32I-NEXT:    lui a1, 349525
 ; RV32I-NEXT:    addi a1, a1, 1365
 ; RV32I-NEXT:    srli a2, a0, 1
 ; RV32I-NEXT:    and a1, a2, a1
 ; RV32I-NEXT:    sub a0, a0, a1

 ; RV32I-NEXT:    addi a1, a1, -241
 ; RV32I-NEXT:    and a0, a0, a1
 ; RV32I-NEXT:    lui a1, 4112
 ; RV32I-NEXT:    addi a1, a1, 257
 ; RV32I-NEXT:    lui a2, %hi(__mulsi3)
 ; RV32I-NEXT:    addi a2, a2, %lo(__mulsi3)
 ; RV32I-NEXT:    jalr ra, a2, 0
 ; RV32I-NEXT:    srli a0, a0, 24
-; RV32I-NEXT:    lw ra, 12(s0)
+; RV32I-NEXT:    lw ra, 12(sp)
 ; RV32I-NEXT:    jalr zero, ra, 0
   %tmp = call i8 @llvm.cttz.i8(i8 %a, i1 true)
   ret i8 %tmp
 }
 
 define i16 @test_cttz_i16_zero_undef(i16 %a) nounwind {
 ; RV32I-LABEL: test_cttz_i16_zero_undef:
 ; RV32I:       # %bb.0:
-; RV32I-NEXT:    sw ra, 12(s0)
+; RV32I-NEXT:    sw ra, 12(sp)
 ; RV32I-NEXT:    addi a1, a0, -1
 ; RV32I-NEXT:    xori a0, a0, -1
 ; RV32I-NEXT:    and a0, a0, a1
 ; RV32I-NEXT:    lui a1, 349525
 ; RV32I-NEXT:    addi a1, a1, 1365
 ; RV32I-NEXT:    srli a2, a0, 1
 ; RV32I-NEXT:    and a1, a2, a1
 ; RV32I-NEXT:    sub a0, a0, a1

 ; RV32I-NEXT:    addi a1, a1, -241
 ; RV32I-NEXT:    and a0, a0, a1
 ; RV32I-NEXT:    lui a1, 4112
 ; RV32I-NEXT:    addi a1, a1, 257
 ; RV32I-NEXT:    lui a2, %hi(__mulsi3)
 ; RV32I-NEXT:    addi a2, a2, %lo(__mulsi3)
 ; RV32I-NEXT:    jalr ra, a2, 0
 ; RV32I-NEXT:    srli a0, a0, 24
-; RV32I-NEXT:    lw ra, 12(s0)
+; RV32I-NEXT:    lw ra, 12(sp)
 ; RV32I-NEXT:    jalr zero, ra, 0
   %tmp = call i16 @llvm.cttz.i16(i16 %a, i1 true)
   ret i16 %tmp
 }
 
 define i32 @test_cttz_i32_zero_undef(i32 %a) nounwind {
 ; RV32I-LABEL: test_cttz_i32_zero_undef:
 ; RV32I:       # %bb.0:
-; RV32I-NEXT:    sw ra, 12(s0)
+; RV32I-NEXT:    sw ra, 12(sp)
 ; RV32I-NEXT:    addi a1, a0, -1
 ; RV32I-NEXT:    xori a0, a0, -1
 ; RV32I-NEXT:    and a0, a0, a1
 ; RV32I-NEXT:    lui a1, 349525
 ; RV32I-NEXT:    addi a1, a1, 1365
 ; RV32I-NEXT:    srli a2, a0, 1
 ; RV32I-NEXT:    and a1, a2, a1
 ; RV32I-NEXT:    sub a0, a0, a1

 ; RV32I-NEXT:    addi a1, a1, -241
 ; RV32I-NEXT:    and a0, a0, a1
 ; RV32I-NEXT:    lui a1, 4112
 ; RV32I-NEXT:    addi a1, a1, 257
 ; RV32I-NEXT:    lui a2, %hi(__mulsi3)
 ; RV32I-NEXT:    addi a2, a2, %lo(__mulsi3)
 ; RV32I-NEXT:    jalr ra, a2, 0
 ; RV32I-NEXT:    srli a0, a0, 24
-; RV32I-NEXT:    lw ra, 12(s0)
+; RV32I-NEXT:    lw ra, 12(sp)
 ; RV32I-NEXT:    jalr zero, ra, 0
   %tmp = call i32 @llvm.cttz.i32(i32 %a, i1 true)
   ret i32 %tmp
 }
 
 define i64 @test_cttz_i64_zero_undef(i64 %a) nounwind {
 ; RV32I-LABEL: test_cttz_i64_zero_undef:
 ; RV32I:       # %bb.0:
-; RV32I-NEXT:    sw ra, 28(s0)
-; RV32I-NEXT:    sw s1, 24(s0)
-; RV32I-NEXT:    sw s2, 20(s0)
-; RV32I-NEXT:    sw s3, 16(s0)
-; RV32I-NEXT:    sw s4, 12(s0)
-; RV32I-NEXT:    sw s5, 8(s0)
-; RV32I-NEXT:    sw s6, 4(s0)
-; RV32I-NEXT:    sw s7, 0(s0)
+; RV32I-NEXT:    sw ra, 28(sp)
+; RV32I-NEXT:    sw s1, 24(sp)
+; RV32I-NEXT:    sw s2, 20(sp)
+; RV32I-NEXT:    sw s3, 16(sp)
+; RV32I-NEXT:    sw s4, 12(sp)
+; RV32I-NEXT:    sw s5, 8(sp)
+; RV32I-NEXT:    sw s6, 4(sp)
+; RV32I-NEXT:    sw s7, 0(sp)
 ; RV32I-NEXT:    addi s1, a1, 0
 ; RV32I-NEXT:    addi s2, a0, 0
 ; RV32I-NEXT:    addi a0, s2, -1
 ; RV32I-NEXT:    xori a1, s2, -1
 ; RV32I-NEXT:    and a0, a1, a0
 ; RV32I-NEXT:    lui a1, 349525
 ; RV32I-NEXT:    addi s4, a1, 1365
 ; RV32I-NEXT:    srli a1, a0, 1

 ; RV32I-NEXT:    jalr ra, s6, 0
 ; RV32I-NEXT:    bne s2, zero, .LBB11_2
 ; RV32I-NEXT:  # %bb.1:
 ; RV32I-NEXT:    srli a0, a0, 24
 ; RV32I-NEXT:    addi s1, a0, 32
 ; RV32I-NEXT:  .LBB11_2:
 ; RV32I-NEXT:    addi a0, s1, 0
 ; RV32I-NEXT:    addi a1, zero, 0
-; RV32I-NEXT:    lw s7, 0(s0)
-; RV32I-NEXT:    lw s6, 4(s0)
-; RV32I-NEXT:    lw s5, 8(s0)
-; RV32I-NEXT:    lw s4, 12(s0)
-; RV32I-NEXT:    lw s3, 16(s0)
-; RV32I-NEXT:    lw s2, 20(s0)
-; RV32I-NEXT:    lw s1, 24(s0)
-; RV32I-NEXT:    lw ra, 28(s0)
+; RV32I-NEXT:    lw s7, 0(sp)
+; RV32I-NEXT:    lw s6, 4(sp)
+; RV32I-NEXT:    lw s5, 8(sp)
+; RV32I-NEXT:    lw s4, 12(sp)
+; RV32I-NEXT:    lw s3, 16(sp)
+; RV32I-NEXT:    lw s2, 20(sp)
+; RV32I-NEXT:    lw s1, 24(sp)
+; RV32I-NEXT:    lw ra, 28(sp)
 ; RV32I-NEXT:    jalr zero, ra, 0
   %tmp = call i64 @llvm.cttz.i64(i64 %a, i1 true)
   ret i64 %tmp
 }
 
 define i32 @test_ctpop_i32(i32 %a) nounwind {
 ; RV32I-LABEL: test_ctpop_i32:
 ; RV32I:       # %bb.0:
-; RV32I-NEXT:    sw ra, 12(s0)
+; RV32I-NEXT:    sw ra, 12(sp)
 ; RV32I-NEXT:    lui a1, 349525
 ; RV32I-NEXT:    addi a1, a1, 1365
 ; RV32I-NEXT:    srli a2, a0, 1
 ; RV32I-NEXT:    and a1, a2, a1
 ; RV32I-NEXT:    sub a0, a0, a1
 ; RV32I-NEXT:    lui a1, 209715
 ; RV32I-NEXT:    addi a1, a1, 819
 ; RV32I-NEXT:    and a2, a0, a1
 ; RV32I-NEXT:    srli a0, a0, 2
 ; RV32I-NEXT:    and a0, a0, a1
 ; RV32I-NEXT:    add a0, a2, a0
 ; RV32I-NEXT:    srli a1, a0, 4
 ; RV32I-NEXT:    add a0, a0, a1
 ; RV32I-NEXT:    lui a1, 61681
 ; RV32I-NEXT:    addi a1, a1, -241
 ; RV32I-NEXT:    and a0, a0, a1
 ; RV32I-NEXT:    lui a1, 4112
 ; RV32I-NEXT:    addi a1, a1, 257
 ; RV32I-NEXT:    lui a2, %hi(__mulsi3)
 ; RV32I-NEXT:    addi a2, a2, %lo(__mulsi3)
 ; RV32I-NEXT:    jalr ra, a2, 0
 ; RV32I-NEXT:    srli a0, a0, 24
-; RV32I-NEXT:    lw ra, 12(s0)
+; RV32I-NEXT:    lw ra, 12(sp)
 ; RV32I-NEXT:    jalr zero, ra, 0
   %1 = call i32 @llvm.ctpop.i32(i32 %a)
   ret i32 %1
 }

llvm/trunk/test/CodeGen/RISCV/calls.ll

 ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
 ; RUN: llc -mtriple=riscv32 -verify-machineinstrs < %s \
 ; RUN:   | FileCheck -check-prefix=RV32I %s
 
 declare i32 @external_function(i32)
 
 define i32 @test_call_external(i32 %a) nounwind {
 ; RV32I-LABEL: test_call_external:
 ; RV32I:       # %bb.0:
-; RV32I-NEXT:    sw ra, 12(s0)
+; RV32I-NEXT:    sw ra, 12(sp)
 ; RV32I-NEXT:    lui a1, %hi(external_function)
 ; RV32I-NEXT:    addi a1, a1, %lo(external_function)
 ; RV32I-NEXT:    jalr ra, a1, 0
-; RV32I-NEXT:    lw ra, 12(s0)
+; RV32I-NEXT:    lw ra, 12(sp)
 ; RV32I-NEXT:    jalr zero, ra, 0
   %1 = call i32 @external_function(i32 %a)
   ret i32 %1
 }
 
 define i32 @defined_function(i32 %a) nounwind {
 ; RV32I-LABEL: defined_function:
 ; RV32I:       # %bb.0:
 ; RV32I-NEXT:    addi a0, a0, 1
 ; RV32I-NEXT:    jalr zero, ra, 0
   %1 = add i32 %a, 1
   ret i32 %1
 }
 
 define i32 @test_call_defined(i32 %a) nounwind {
 ; RV32I-LABEL: test_call_defined:
 ; RV32I:       # %bb.0:
-; RV32I-NEXT:    sw ra, 12(s0)
+; RV32I-NEXT:    sw ra, 12(sp)
 ; RV32I-NEXT:    lui a1, %hi(defined_function)
 ; RV32I-NEXT:    addi a1, a1, %lo(defined_function)
 ; RV32I-NEXT:    jalr ra, a1, 0
-; RV32I-NEXT:    lw ra, 12(s0)
+; RV32I-NEXT:    lw ra, 12(sp)
 ; RV32I-NEXT:    jalr zero, ra, 0
   %1 = call i32 @defined_function(i32 %a) nounwind
   ret i32 %1
 }
 
 define i32 @test_call_indirect(i32 (i32)* %a, i32 %b) nounwind {
 ; RV32I-LABEL: test_call_indirect:
 ; RV32I:       # %bb.0:
-; RV32I-NEXT:    sw ra, 12(s0)
+; RV32I-NEXT:    sw ra, 12(sp)
 ; RV32I-NEXT:    addi a2, a0, 0
 ; RV32I-NEXT:    addi a0, a1, 0
 ; RV32I-NEXT:    jalr ra, a2, 0
-; RV32I-NEXT:    lw ra, 12(s0)
+; RV32I-NEXT:    lw ra, 12(sp)
 ; RV32I-NEXT:    jalr zero, ra, 0
   %1 = call i32 %a(i32 %b)
   ret i32 %1
 }
 
 ; Ensure that calls to fastcc functions aren't rejected. Such calls may be
 ; introduced when compiling with optimisation.
 
 define fastcc i32 @fastcc_function(i32 %a, i32 %b) nounwind {
 ; RV32I-LABEL: fastcc_function:
 ; RV32I:       # %bb.0:
 ; RV32I-NEXT:    add a0, a0, a1
 ; RV32I-NEXT:    jalr zero, ra, 0
  %1 = add i32 %a, %b
  ret i32 %1
 }
 
 define i32 @test_call_fastcc(i32 %a, i32 %b) nounwind {
 ; RV32I-LABEL: test_call_fastcc:
 ; RV32I:       # %bb.0:
-; RV32I-NEXT:    sw ra, 12(s0)
-; RV32I-NEXT:    sw s1, 8(s0)
+; RV32I-NEXT:    sw ra, 12(sp)
+; RV32I-NEXT:    sw s1, 8(sp)
 ; RV32I-NEXT:    addi s1, a0, 0
 ; RV32I-NEXT:    lui a0, %hi(fastcc_function)
 ; RV32I-NEXT:    addi a2, a0, %lo(fastcc_function)
 ; RV32I-NEXT:    addi a0, s1, 0
 ; RV32I-NEXT:    jalr ra, a2, 0
 ; RV32I-NEXT:    addi a0, s1, 0
-; RV32I-NEXT:    lw s1, 8(s0)
-; RV32I-NEXT:    lw ra, 12(s0)
+; RV32I-NEXT:    lw s1, 8(sp)
+; RV32I-NEXT:    lw ra, 12(sp)
 ; RV32I-NEXT:    jalr zero, ra, 0
   %1 = call fastcc i32 @fastcc_function(i32 %a, i32 %b)
   ret i32 %a
 }

llvm/trunk/test/CodeGen/RISCV/div.ll

 ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
 ; RUN: llc -mtriple=riscv32 -verify-machineinstrs < %s \
 ; RUN:   | FileCheck %s -check-prefix=RV32I
 
 define i32 @udiv(i32 %a, i32 %b) {
 ; RV32I-LABEL: udiv:
 ; RV32I:       # %bb.0:
-; RV32I-NEXT:    sw ra, 12(s0)
+; RV32I-NEXT:    sw ra, 12(sp)
 ; RV32I-NEXT:    lui a2, %hi(__udivsi3)
 ; RV32I-NEXT:    addi a2, a2, %lo(__udivsi3)
 ; RV32I-NEXT:    jalr ra, a2, 0
-; RV32I-NEXT:    lw ra, 12(s0)
+; RV32I-NEXT:    lw ra, 12(sp)
 ; RV32I-NEXT:    jalr zero, ra, 0
   %1 = udiv i32 %a, %b
   ret i32 %1
 }
 
 define i32 @udiv_constant(i32 %a) {
 ; RV32I-LABEL: udiv_constant:
 ; RV32I:       # %bb.0:
-; RV32I-NEXT:    sw ra, 12(s0)
+; RV32I-NEXT:    sw ra, 12(sp)
 ; RV32I-NEXT:    lui a1, %hi(__udivsi3)
 ; RV32I-NEXT:    addi a2, a1, %lo(__udivsi3)
 ; RV32I-NEXT:    addi a1, zero, 5
 ; RV32I-NEXT:    jalr ra, a2, 0
-; RV32I-NEXT:    lw ra, 12(s0)
+; RV32I-NEXT:    lw ra, 12(sp)
 ; RV32I-NEXT:    jalr zero, ra, 0
   %1 = udiv i32 %a, 5
   ret i32 %1
 }
 
 define i32 @udiv_pow2(i32 %a) {
 ; RV32I-LABEL: udiv_pow2:
 ; RV32I:       # %bb.0:
 ; RV32I-NEXT:    srli a0, a0, 3
 ; RV32I-NEXT:    jalr zero, ra, 0
   %1 = udiv i32 %a, 8
   ret i32 %1
 }
 
 define i64 @udiv64(i64 %a, i64 %b) {
 ; RV32I-LABEL: udiv64:
 ; RV32I:       # %bb.0:
-; RV32I-NEXT:    sw ra, 12(s0)
+; RV32I-NEXT:    sw ra, 12(sp)
 ; RV32I-NEXT:    lui a4, %hi(__udivdi3)
 ; RV32I-NEXT:    addi a4, a4, %lo(__udivdi3)
 ; RV32I-NEXT:    jalr ra, a4, 0
-; RV32I-NEXT:    lw ra, 12(s0)
+; RV32I-NEXT:    lw ra, 12(sp)
 ; RV32I-NEXT:    jalr zero, ra, 0
   %1 = udiv i64 %a, %b
   ret i64 %1
 }
 
 define i64 @udiv64_constant(i64 %a) {
 ; RV32I-LABEL: udiv64_constant:
 ; RV32I:       # %bb.0:
-; RV32I-NEXT:    sw ra, 12(s0)
+; RV32I-NEXT:    sw ra, 12(sp)
 ; RV32I-NEXT:    lui a2, %hi(__udivdi3)
 ; RV32I-NEXT:    addi a4, a2, %lo(__udivdi3)
 ; RV32I-NEXT:    addi a2, zero, 5
 ; RV32I-NEXT:    addi a3, zero, 0
 ; RV32I-NEXT:    jalr ra, a4, 0
-; RV32I-NEXT:    lw ra, 12(s0)
+; RV32I-NEXT:    lw ra, 12(sp)
 ; RV32I-NEXT:    jalr zero, ra, 0
   %1 = udiv i64 %a, 5
   ret i64 %1
 }
 
 define i32 @sdiv(i32 %a, i32 %b) {
 ; RV32I-LABEL: sdiv:
 ; RV32I:       # %bb.0:
-; RV32I-NEXT:    sw ra, 12(s0)
+; RV32I-NEXT:    sw ra, 12(sp)
 ; RV32I-NEXT:    lui a2, %hi(__divsi3)
 ; RV32I-NEXT:    addi a2, a2, %lo(__divsi3)
 ; RV32I-NEXT:    jalr ra, a2, 0
-; RV32I-NEXT:    lw ra, 12(s0)
+; RV32I-NEXT:    lw ra, 12(sp)
 ; RV32I-NEXT:    jalr zero, ra, 0
   %1 = sdiv i32 %a, %b
   ret i32 %1
 }
 
 define i32 @sdiv_constant(i32 %a) {
 ; RV32I-LABEL: sdiv_constant:
 ; RV32I:       # %bb.0:
-; RV32I-NEXT:    sw ra, 12(s0)
+; RV32I-NEXT:    sw ra, 12(sp)
 ; RV32I-NEXT:    lui a1, %hi(__divsi3)
 ; RV32I-NEXT:    addi a2, a1, %lo(__divsi3)
 ; RV32I-NEXT:    addi a1, zero, 5
 ; RV32I-NEXT:    jalr ra, a2, 0
-; RV32I-NEXT:    lw ra, 12(s0)
+; RV32I-NEXT:    lw ra, 12(sp)
 ; RV32I-NEXT:    jalr zero, ra, 0
   %1 = sdiv i32 %a, 5
   ret i32 %1
 }
 
 define i32 @sdiv_pow2(i32 %a) {
 ; RV32I-LABEL: sdiv_pow2:
 ; RV32I:       # %bb.0:
 ; RV32I-NEXT:    srai a1, a0, 31
 ; RV32I-NEXT:    srli a1, a1, 29
 ; RV32I-NEXT:    add a0, a0, a1
 ; RV32I-NEXT:    srai a0, a0, 3
 ; RV32I-NEXT:    jalr zero, ra, 0
   %1 = sdiv i32 %a, 8
   ret i32 %1
 }
 
 define i64 @sdiv64(i64 %a, i64 %b) {
 ; RV32I-LABEL: sdiv64:
 ; RV32I:       # %bb.0:
-; RV32I-NEXT:    sw ra, 12(s0)
+; RV32I-NEXT:    sw ra, 12(sp)
 ; RV32I-NEXT:    lui a4, %hi(__divdi3)
 ; RV32I-NEXT:    addi a4, a4, %lo(__divdi3)
 ; RV32I-NEXT:    jalr ra, a4, 0
-; RV32I-NEXT:    lw ra, 12(s0)
+; RV32I-NEXT:    lw ra, 12(sp)
 ; RV32I-NEXT:    jalr zero, ra, 0
   %1 = sdiv i64 %a, %b
   ret i64 %1
 }
 
 define i64 @sdiv64_constant(i64 %a) {
 ; RV32I-LABEL: sdiv64_constant:
 ; RV32I:       # %bb.0:
-; RV32I-NEXT:    sw ra, 12(s0)
+; RV32I-NEXT:    sw ra, 12(sp)
 ; RV32I-NEXT:    lui a2, %hi(__divdi3)
 ; RV32I-NEXT:    addi a4, a2, %lo(__divdi3)
 ; RV32I-NEXT:    addi a2, zero, 5
 ; RV32I-NEXT:    addi a3, zero, 0
 ; RV32I-NEXT:    jalr ra, a4, 0
-; RV32I-NEXT:    lw ra, 12(s0)
+; RV32I-NEXT:    lw ra, 12(sp)
 ; RV32I-NEXT:    jalr zero, ra, 0
   %1 = sdiv i64 %a, 5
   ret i64 %1
 }

llvm/trunk/test/CodeGen/RISCV/frame.ll

+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
+; RUN: llc -mtriple=riscv32 -verify-machineinstrs < %s \
+; RUN:   | FileCheck -check-prefix=RV32I %s
+
+%struct.key_t = type { i32, [16 x i8] }
+
+; FIXME: prologue and epilogue insertion must be implemented to complete this
+; test
+
+define i32 @test() nounwind {
+; RV32I-LABEL: test:
+; RV32I:       # %bb.0:
+; RV32I-NEXT:    sw ra, 28(sp)
+; RV32I-NEXT:    sw zero, -8(s0)
+; RV32I-NEXT:    sw zero, -12(s0)
+; RV32I-NEXT:    sw zero, -16(s0)
+; RV32I-NEXT:    sw zero, -20(s0)
+; RV32I-NEXT:    sw zero, -24(s0)
+; RV32I-NEXT:    lui a0, %hi(test1)
+; RV32I-NEXT:    addi a1, a0, %lo(test1)
+; RV32I-NEXT:    addi a0, s0, -20
+; RV32I-NEXT:    jalr ra, a1, 0
+; RV32I-NEXT:    addi a0, zero, 0
+; RV32I-NEXT:    lw ra, 28(sp)
+; RV32I-NEXT:    jalr zero, ra, 0
+  %key = alloca %struct.key_t, align 4
+  %1 = bitcast %struct.key_t* %key to i8*
+  call void @llvm.memset.p0i8.i64(i8* %1, i8 0, i64 20, i32 4, i1 false)
+  %2 = getelementptr inbounds %struct.key_t, %struct.key_t* %key, i64 0, i32 1, i64 0
+  call void @test1(i8* %2) #3
+  ret i32 0
+}
+
+declare void @llvm.memset.p0i8.i64(i8* nocapture, i8, i64, i32, i1)
+
+declare void @test1(i8*)

llvm/trunk/test/CodeGen/RISCV/indirectbr.ll

 ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
 ; RUN: llc -mtriple=riscv32 -verify-machineinstrs < %s \
 ; RUN:   | FileCheck %s -check-prefix=RV32I
 
 define i32 @indirectbr(i8* %target) nounwind {
 ; RV32I-LABEL: indirectbr:
 ; RV32I:       # %bb.0:
-; RV32I-NEXT:    sw ra, 0(s0)
+; RV32I-NEXT:    sw ra, 0(sp)
 ; RV32I-NEXT:    jalr zero, a0, 0
 ; RV32I-NEXT:  .LBB0_1: # %ret
 ; RV32I-NEXT:    addi a0, zero, 0
-; RV32I-NEXT:    lw ra, 0(s0)
+; RV32I-NEXT:    lw ra, 0(sp)
 ; RV32I-NEXT:    jalr zero, ra, 0
   indirectbr i8* %target, [label %test_label]
 test_label:
   br label %ret
 ret:
   ret i32 0
 }
 
 define i32 @indirectbr_with_offset(i8* %a) nounwind {
 ; RV32I-LABEL: indirectbr_with_offset:
 ; RV32I:       # %bb.0:
-; RV32I-NEXT:    sw ra, 0(s0)
+; RV32I-NEXT:    sw ra, 0(sp)
 ; RV32I-NEXT:    jalr zero, a0, 1380
 ; RV32I-NEXT:  .LBB1_1: # %ret
 ; RV32I-NEXT:    addi a0, zero, 0
-; RV32I-NEXT:    lw ra, 0(s0)
+; RV32I-NEXT:    lw ra, 0(sp)
 ; RV32I-NEXT:    jalr zero, ra, 0
   %target = getelementptr inbounds i8, i8* %a, i32 1380
   indirectbr i8* %target, [label %test_label]
 test_label:
   br label %ret
 ret:
   ret i32 0
 }

llvm/trunk/test/CodeGen/RISCV/mul.ll

 ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
 ; RUN: llc -mtriple=riscv32 -verify-machineinstrs < %s \
 ; RUN:   | FileCheck %s -check-prefix=RV32I
 
 define i32 @square(i32 %a) {
 ; RV32I-LABEL: square:
 ; RV32I:       # %bb.0:
-; RV32I-NEXT:    sw ra, 12(s0)
+; RV32I-NEXT:    sw ra, 12(sp)
 ; RV32I-NEXT:    lui a1, %hi(__mulsi3)
 ; RV32I-NEXT:    addi a2, a1, %lo(__mulsi3)
 ; RV32I-NEXT:    addi a1, a0, 0
 ; RV32I-NEXT:    jalr ra, a2, 0
-; RV32I-NEXT:    lw ra, 12(s0)
+; RV32I-NEXT:    lw ra, 12(sp)
 ; RV32I-NEXT:    jalr zero, ra, 0
   %1 = mul i32 %a, %a
   ret i32 %1
 }
 
 define i32 @mul(i32 %a, i32 %b) {
 ; RV32I-LABEL: mul:
 ; RV32I:       # %bb.0:
-; RV32I-NEXT:    sw ra, 12(s0)
+; RV32I-NEXT:    sw ra, 12(sp)
 ; RV32I-NEXT:    lui a2, %hi(__mulsi3)
 ; RV32I-NEXT:    addi a2, a2, %lo(__mulsi3)
 ; RV32I-NEXT:    jalr ra, a2, 0
-; RV32I-NEXT:    lw ra, 12(s0)
+; RV32I-NEXT:    lw ra, 12(sp)
 ; RV32I-NEXT:    jalr zero, ra, 0
   %1 = mul i32 %a, %b
   ret i32 %1
 }
 
 define i32 @mul_constant(i32 %a) {
 ; RV32I-LABEL: mul_constant:
 ; RV32I:       # %bb.0:
-; RV32I-NEXT:    sw ra, 12(s0)
+; RV32I-NEXT:    sw ra, 12(sp)
 ; RV32I-NEXT:    lui a1, %hi(__mulsi3)
 ; RV32I-NEXT:    addi a2, a1, %lo(__mulsi3)
 ; RV32I-NEXT:    addi a1, zero, 5
 ; RV32I-NEXT:    jalr ra, a2, 0
-; RV32I-NEXT:    lw ra, 12(s0)
+; RV32I-NEXT:    lw ra, 12(sp)
 ; RV32I-NEXT:    jalr zero, ra, 0
   %1 = mul i32 %a, 5
   ret i32 %1
 }
 
 define i32 @mul_pow2(i32 %a) {
 ; RV32I-LABEL: mul_pow2:
 ; RV32I:       # %bb.0:
 ; RV32I-NEXT:    slli a0, a0, 3
 ; RV32I-NEXT:    jalr zero, ra, 0
   %1 = mul i32 %a, 8
   ret i32 %1
 }
 
 define i64 @mul64(i64 %a, i64 %b) {
 ; RV32I-LABEL: mul64:
 ; RV32I:       # %bb.0:
-; RV32I-NEXT:    sw ra, 12(s0)
+; RV32I-NEXT:    sw ra, 12(sp)
 ; RV32I-NEXT:    lui a4, %hi(__muldi3)
 ; RV32I-NEXT:    addi a4, a4, %lo(__muldi3)
 ; RV32I-NEXT:    jalr ra, a4, 0
-; RV32I-NEXT:    lw ra, 12(s0)
+; RV32I-NEXT:    lw ra, 12(sp)
 ; RV32I-NEXT:    jalr zero, ra, 0
   %1 = mul i64 %a, %b
   ret i64 %1
 }
 
 define i64 @mul64_constant(i64 %a) {
 ; RV32I-LABEL: mul64_constant:
 ; RV32I:       # %bb.0:
-; RV32I-NEXT:    sw ra, 12(s0)
+; RV32I-NEXT:    sw ra, 12(sp)
 ; RV32I-NEXT:    lui a2, %hi(__muldi3)
 ; RV32I-NEXT:    addi a4, a2, %lo(__muldi3)
 ; RV32I-NEXT:    addi a2, zero, 5
 ; RV32I-NEXT:    addi a3, zero, 0
 ; RV32I-NEXT:    jalr ra, a4, 0
-; RV32I-NEXT:    lw ra, 12(s0)
+; RV32I-NEXT:    lw ra, 12(sp)
 ; RV32I-NEXT:    jalr zero, ra, 0
   %1 = mul i64 %a, 5
   ret i64 %1
 }

llvm/trunk/test/CodeGen/RISCV/rem.ll

 ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
 ; RUN: llc -mtriple=riscv32 -verify-machineinstrs < %s \
 ; RUN:   | FileCheck %s -check-prefix=RV32I
 
 define i32 @urem(i32 %a, i32 %b) nounwind {
 ; RV32I-LABEL: urem:
 ; RV32I:       # %bb.0:
-; RV32I-NEXT:    sw ra, 12(s0)
+; RV32I-NEXT:    sw ra, 12(sp)
 ; RV32I-NEXT:    lui a2, %hi(__umodsi3)
 ; RV32I-NEXT:    addi a2, a2, %lo(__umodsi3)
 ; RV32I-NEXT:    jalr ra, a2, 0
-; RV32I-NEXT:    lw ra, 12(s0)
+; RV32I-NEXT:    lw ra, 12(sp)
 ; RV32I-NEXT:    jalr zero, ra, 0
   %1 = urem i32 %a, %b
   ret i32 %1
 }
 
 define i32 @srem(i32 %a, i32 %b) nounwind {
 ; RV32I-LABEL: srem:
 ; RV32I:       # %bb.0:
-; RV32I-NEXT:    sw ra, 12(s0)
+; RV32I-NEXT:    sw ra, 12(sp)
 ; RV32I-NEXT:    lui a2, %hi(__modsi3)
 ; RV32I-NEXT:    addi a2, a2, %lo(__modsi3)
 ; RV32I-NEXT:    jalr ra, a2, 0
-; RV32I-NEXT:    lw ra, 12(s0)
+; RV32I-NEXT:    lw ra, 12(sp)
 ; RV32I-NEXT:    jalr zero, ra, 0
   %1 = srem i32 %a, %b
   ret i32 %1
 }

llvm/trunk/test/CodeGen/RISCV/shifts.ll

 ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
 ; RUN: llc -mtriple=riscv32 -verify-machineinstrs < %s \
 ; RUN:   | FileCheck %s -check-prefix=RV32I
 
 ; Basic shift support is tested as part of ALU.ll. This file ensures that
 ; shifts which may not be supported natively are lowered properly.
 
 define i64 @lshr64(i64 %a, i64 %b) nounwind {
 ; RV32I-LABEL: lshr64:
 ; RV32I:       # %bb.0:
-; RV32I-NEXT:    sw ra, 12(s0)
+; RV32I-NEXT:    sw ra, 12(sp)
 ; RV32I-NEXT:    lui a3, %hi(__lshrdi3)
 ; RV32I-NEXT:    addi a3, a3, %lo(__lshrdi3)
 ; RV32I-NEXT:    jalr ra, a3, 0
-; RV32I-NEXT:    lw ra, 12(s0)
+; RV32I-NEXT:    lw ra, 12(sp)
 ; RV32I-NEXT:    jalr zero, ra, 0
   %1 = lshr i64 %a, %b
   ret i64 %1
 }
 
 define i64 @ashr64(i64 %a, i64 %b) nounwind {
 ; RV32I-LABEL: ashr64:
 ; RV32I:       # %bb.0:
-; RV32I-NEXT:    sw ra, 12(s0)
+; RV32I-NEXT:    sw ra, 12(sp)
 ; RV32I-NEXT:    lui a3, %hi(__ashrdi3)
 ; RV32I-NEXT:    addi a3, a3, %lo(__ashrdi3)
 ; RV32I-NEXT:    jalr ra, a3, 0
-; RV32I-NEXT:    lw ra, 12(s0)
+; RV32I-NEXT:    lw ra, 12(sp)
 ; RV32I-NEXT:    jalr zero, ra, 0
   %1 = ashr i64 %a, %b
   ret i64 %1
 }
 
 define i64 @shl64(i64 %a, i64 %b) nounwind {
 ; RV32I-LABEL: shl64:
 ; RV32I:       # %bb.0:
-; RV32I-NEXT:    sw ra, 12(s0)
+; RV32I-NEXT:    sw ra, 12(sp)
 ; RV32I-NEXT:    lui a3, %hi(__ashldi3)
 ; RV32I-NEXT:    addi a3, a3, %lo(__ashldi3)
 ; RV32I-NEXT:    jalr ra, a3, 0
-; RV32I-NEXT:    lw ra, 12(s0)
+; RV32I-NEXT:    lw ra, 12(sp)
 ; RV32I-NEXT:    jalr zero, ra, 0
   %1 = shl i64 %a, %b
   ret i64 %1
 }