[RISCV] Handle register spill in branch relaxation

[piggynl]

Summary

In branch relaxation pass, j's with offset over 1MiB will be relaxed to jump pseudo-instructions.

This patch allocates a stack slot for functions with a size greater than 1MiB. If the register scavenger cannot find a scratch register for jump, spill a register to the slot before the jump and restore it after the jump.

.mbb:
        foo
        j       .dest_bb
        bar
        bar
        bar
.dest_bb:
        baz

The above code will be relaxed to the following code.

.mbb:
        foo
        sd      s11, 0(sp)
        jump    .restore_bb, s11
        bar
        bar
        bar
        j       .dest_bb
.restore_bb:
        ld      s11, 0(sp)
.dest_bb:
        baz

Depends on D129999.

[piggynl]

llvm/lib/Target/RISCV/RISCVInstrInfo.cpp

1003–1004

Do we need another spill slot to calculate the offset of the slot in eliminateFrameIndex?

[craig.topper]

llvm/lib/Target/RISCV/RISCVFrameLowering.cpp

1008

ARM places constant pools and jump tables in the middle of functions via the ConstantIslands pass . I don't think we do that on RISC-V. So I think we only need to count instruction size.

[StephenFan]

llvm/lib/Target/RISCV/RISCVFrameLowering.cpp

1109

Add this stack object as scavenge frame index in RS. The reason will be explained in the comment below.

llvm/lib/Target/RISCV/RISCVInstrInfo.cpp

1003–1004

if the offset of BranchRelaxationScratchFrameIndex is out of range of Int<12>, then an extra scratch register will be created. To avoid this extra scratch register, we need to make sure that the BranchRelaxationScratchFrameIndex is close to the stack base register. Fortunately, PEI will float or sink the scavenge frame index to close to the stack base register(See https://reviews.llvm.org/D89239). But only if BranchRelaxationScratchFrameIndex is a scavenging frame index.

[efriedma]

llvm/lib/Target/RISCV/RISCVInstrInfo.cpp

999

You probably want to report_fatal_error if you don't have a frame index, so you get an error in release builds if the heuristic breaks somehow.

[efriedma]

llvm/lib/Target/RISCV/RISCVFrameLowering.cpp

1006

You might want to consider using a more conservative estimate for conditional branches, since they can be relaxed to unconditional branches. (I know we don't do this in the ARM function you borrowed this from, but probably we should.)

[piggynl]

• More conservative estimate for branches in EstimateFunctionSizeInBytes.

• Remove the size of constants and jump tables in EstimateFunctionSizeInBytes.

• Add the new stack frame index to RS in RISCVFrameLowering::processFunctionBeforeFrameFinalized.

• Use report_fatal_error in RISCVInstrInfo::insertIndirectBranch if function size is underestimated.

[piggynl]

Fix code style via clang-format.

[StephenFan]

llvm/lib/Target/RISCV/RISCVFrameLowering.cpp

1113

If it would be better to add a test case that has a large stack and the offset of BranchRelaxationScratchFrameIndex is out of range of Int<12>(if it is not floated or sunk)?

[piggynl]

Add test for a large stack that requires adjusting the offset of BranchRelaxationScratchFrameIndex.

[piggynl]

Fix code style via clang-format.

[piggynl]

Remove unnecessary #include.

[StephenFan]

llvm/lib/Target/RISCV/RISCVFrameLowering.cpp

1108

Actually, scavenging spill slots can be reused. But here we always create a scavenging spill slot even if there is already a scavenging spill slot.

[piggynl]

llvm/lib/Target/RISCV/RISCVFrameLowering.cpp

1108

Since branch relaxation pass is after instruction scheduling, I'm not sure whether the jump could be placed in the middle of a spill from previous pass, and the slot's content could be overwritten if we reuse the slot.

I'm not familiar with instruction scheduling, so please correct me for any mistake I made. Thank you in advance!

[CoelacanthusHex]

There seems to be a problem here: if we perform only some operations on registers without sd after a jump (which can easily happen if only a few short instructions are executed between jumps), we then jump to another place. If we still restore immediately in this case, we are actually wasting resources.

[StephenFan]

llvm/lib/Target/RISCV/RISCVFrameLowering.cpp

1108

As far as I know, llvm has not implemented global instruction scheduling, it only schedules instruction within basic blocks. Since Jump instruction is the terminator of a basic block, it may not be scheduled. Otherwise, the control flow graph would be changed. But I'm not an expert in instruction scheduling. Maybe @HsiangKai can answer this question.

[hiraditya]

llvm/lib/Target/RISCV/RISCVFrameLowering.cpp

1107

Can we make an API like isFarBranch or something like it? Other places can reuse it possibly.

[piggynl]

Reuse scavenging spill slot for scratch register's spill slot in branch relaxation.

[piggynl]

llvm/lib/Target/RISCV/RISCVFrameLowering.cpp

1100

Do we need this IsLargeFunction?

[piggynl]

In D130560#3684918, @CoelacanthusHex wrote:

There seems to be a problem here: if we perform only some operations on registers without sd after a jump (which can easily happen if only a few short instructions are executed between jumps), we then jump to another place. If we still restore immediately in this case, we are actually wasting resources.

Yes. We can construct a loop with a few instructions executed and a long jump at each iteration, making the jump scratch register saved/restored frequently. Maybe we can add a TODO in BranchRelaxation.cpp for future improvements?

llvm/lib/Target/RISCV/RISCVFrameLowering.cpp

1107

There is BranchRelaxation::isBlockInRange() in BranchRelaxation pass, but we cannot use that in PEI since there are some late optimizations between PEI and BranchRelaxation (for example MachineBlockPlacement), making us cannot determine whether a branch is a far branch or not. That's also the reason why we reserve a spill slot for BranchRelaxation pass if we estimate the function has a size over 1MiB (which means it is possible that have a branch over 1MiB).

As for an API to estimate the size of a function, it is hard to achieve in a target-independent way. (See https://reviews.llvm.org/D129975) For example, ARM places constant pools and jump tables in the middle of functions via the ConstantIslands pass while RISCV doesn't, meaning that RISCV functions never contain jumps over constant pools and jump tables. As a result, in RISCV, the size of the function does not need to include the constant pool and jump table, whereas in ARM it does.

[CoelacanthusHex]

In D130560#3686584, @piggynl wrote:

In D130560#3684918, @CoelacanthusHex wrote:

There seems to be a problem here: if we perform only some operations on registers without sd after a jump (which can easily happen if only a few short instructions are executed between jumps), we then jump to another place. If we still restore immediately in this case, we are actually wasting resources.

Yes. We can construct a loop with a few instructions executed and a long jump at each iteration, making the jump scratch register saved/restored frequently. Maybe we can add a TODO in BranchRelaxation.cpp for future improvements?

Yes, I think we should add TODO.
I have some ideas, but the expected effect is not good. If I can improve the algorithm and implement it, I will open a PR.

[piggynl]

Ping :) Is there anything else I can do to improve this?

[StephenFan]

llvm/lib/Target/RISCV/RISCVFrameLowering.cpp

1095–1096

Capitalize i

1100

I think we need.

1100

Is this more readable?

[piggynl]

• Revise according to @StephenFan's comments.

• Fix coding style via clang-tidy.

• Move comments about calculating ScavSlotsNum closer to their codes.

• Move tests relax_spill in branch-relaxation-spill-{32,64}.ll to relax_jal_spill_{32,64}{,_adjust_spill_slot} in branch-relaxation.ll.

[StephenFan]

LGTM

[piggynl]

Thank you! Would you mind taking a look at D129999? This patch depends on it for estimating functions' sizes.

[luismarques]

The tests don't seem to have been properly updated with update_llc_test_checks.py. llvm/test/CodeGen/RISCV/branch-relaxation.ll contains RV64 RUN lines but the corresponding CHECK lines are missing in some functions.

[luismarques]

In D130560#3746379, @luismarques wrote:

The tests don't seem to have been properly updated with update_llc_test_checks.py. llvm/test/CodeGen/RISCV/branch-relaxation.ll contains RV64 RUN lines but the corresponding CHECK lines are missing in some functions.

Looking more closely at this, I guess you tried to only include the CHECK-RV64 and CHECK-RV32 checks when relevant. That's a good instinct but I guess it goes a bit against how we normally use update_llc_test_checks.py. My understanding of the trade-off of using that tool is that the test updates are much easier, even if sometimes the CHECKs aren't as tight as something more tailormade.

[piggynl]

In D130560#3746417, @luismarques wrote:

In D130560#3746379, @luismarques wrote:

The tests don't seem to have been properly updated with update_llc_test_checks.py. llvm/test/CodeGen/RISCV/branch-relaxation.ll contains RV64 RUN lines but the corresponding CHECK lines are missing in some functions.

Looking more closely at this, I guess you tried to only include the CHECK-RV64 and CHECK-RV32 checks when relevant. That's a good instinct but I guess it goes a bit against how we normally use update_llc_test_checks.py. My understanding of the trade-off of using that tool is that the test updates are much easier, even if sometimes the CHECKs aren't as tight as something more tailormade.

You are absolutely right. I've created D132625 to update the tests. Thank you!

llvm/lib/Target/RISCV/RISCVFrameLowering.cpp

 
 static bool hasRVVFrameObject(const MachineFunction &MF) {
   // Originally, the function will scan all the stack objects to check whether
   // if there is any scalable vector object on the stack or not. However, it
   // causes errors in the register allocator. In issue 53016, it returns false
   // before RA because there is no RVV stack objects. After RA, it returns true
   // because there are spilling slots for RVV values during RA. It will not
   // reserve BP during register allocation and generate BP access in the PEI
   // pass due to the inconsistent behavior of the function.
   //
   // The function is changed to use hasVInstructions() as the return value. It
   // is not precise, but it can make the register allocation correct.
   //
   // FIXME: Find a better way to make the decision or revisit the solution in
   // D103622.
   //
   // Refer to https://github.com/llvm/llvm-project/issues/53016.
   return MF.getSubtarget<RISCVSubtarget>().hasVInstructions();
 }
 
+static unsigned estimateFunctionSizeInBytes(const MachineFunction &MF,
+                                            const RISCVInstrInfo &TII) {
+  unsigned FnSize = 0;
+  for (auto &MBB : MF) {
+    for (auto &MI : MBB) {
+      // Far branches over 20-bit offset will be relaxed in branch relaxation
+      // pass. In the worst case, conditional branches will be relaxed into
+      // the following instruction sequence. Unconditional branches are
+      // relaxed in the same way, with the exception that there is no first
+      // branch instruction.
+      //
+      //        foo
+      //        bne     t5, t6, .rev_cond # `TII->getInstSizeInBytes(MI)` bytes
+      //        sd      s11, 0(sp)        # 4 bytes, or 2 bytes in RVC
+      //        jump    .restore, s11     # 8 bytes
+      // .rev_cond
+      //        bar
+      //        j       .dest_bb          # 4 bytes, or 2 bytes in RVC
+      // .restore:
+      //        ld      s11, 0(sp)        # 4 bytes, or 2 bytes in RVC
+      // .dest:
+      //        baz
+      if (MI.isConditionalBranch())
+        FnSize += TII.getInstSizeInBytes(MI);
+      if (MI.isConditionalBranch() || MI.isUnconditionalBranch()) {
+        if (MF.getSubtarget<RISCVSubtarget>().hasStdExtC())
+          FnSize += 2 + 8 + 2 + 2;
+        else
+          FnSize += 4 + 8 + 4 + 4;
+        continue;
+      }
+
+      FnSize += TII.getInstSizeInBytes(MI);
+    }
+  }
+  return FnSize;
+}
+
 void RISCVFrameLowering::processFunctionBeforeFrameFinalized(
     MachineFunction &MF, RegScavenger *RS) const {
   const RISCVRegisterInfo *RegInfo =
       MF.getSubtarget<RISCVSubtarget>().getRegisterInfo();
+  const RISCVInstrInfo *TII = MF.getSubtarget<RISCVSubtarget>().getInstrInfo();
   MachineFrameInfo &MFI = MF.getFrameInfo();
   const TargetRegisterClass *RC = &RISCV::GPRRegClass;
   auto *RVFI = MF.getInfo<RISCVMachineFunctionInfo>();
 
   int64_t RVVStackSize;
   Align RVVStackAlign;
   std::tie(RVVStackSize, RVVStackAlign) = assignRVVStackObjectOffsets(MF);
 
   RVFI->setRVVStackSize(RVVStackSize);
   RVFI->setRVVStackAlign(RVVStackAlign);
 
   if (hasRVVFrameObject(MF)) {
     // Ensure the entire stack is aligned to at least the RVV requirement: some
     // scalable-vector object alignments are not considered by the
     // target-independent code.
     MFI.ensureMaxAlignment(RVVStackAlign);
   }
 
+  unsigned ScavSlotsNum = 0;
+
   // estimateStackSize has been observed to under-estimate the final stack
   // size, so give ourselves wiggle-room by checking for stack size
   // representable an 11-bit signed field rather than 12-bits.
-  // FIXME: It may be possible to craft a function with a small stack that
-  // still needs an emergency spill slot for branch relaxation. This case
-  // would currently be missed.
-  // RVV loads & stores have no capacity to hold the immediate address offsets
-  // so we must always reserve an emergency spill slot if the MachineFunction
-  // contains any RVV spills.
-  unsigned ScavSlotsNum = 0;
   if (!isInt<11>(MFI.estimateStackSize(MF)))
     ScavSlotsNum = 1;
 
+  // Far branches over 20-bit offset require a spill slot for scratch register.
+  bool IsLargeFunction = !isInt<20>(estimateFunctionSizeInBytes(MF, *TII));
+  if (IsLargeFunction)
+    ScavSlotsNum = std::max(ScavSlotsNum, 1u);
+
+  // RVV loads & stores have no capacity to hold the immediate address offsets
+  // so we must always reserve an emergency spill slot if the MachineFunction
+  // contains any RVV spills.
   ScavSlotsNum = std::max(ScavSlotsNum, getScavSlotsNumForRVV(MF));
-  for (unsigned i = 0; i < ScavSlotsNum; i++) {
-    RS->addScavengingFrameIndex(MFI.CreateStackObject(
-        RegInfo->getSpillSize(*RC), RegInfo->getSpillAlign(*RC), false));
+
+  for (unsigned I = 0; I < ScavSlotsNum; I++) {
+    int FI = MFI.CreateStackObject(RegInfo->getSpillSize(*RC),
+                                   RegInfo->getSpillAlign(*RC), false);
+    RS->addScavengingFrameIndex(FI);
+
+    if (IsLargeFunction && RVFI->getBranchRelaxationScratchFrameIndex() == -1)
+      RVFI->setBranchRelaxationScratchFrameIndex(FI);
   }
 
   if (MFI.getCalleeSavedInfo().empty() || RVFI->useSaveRestoreLibCalls(MF)) {
     RVFI->setCalleeSavedStackSize(0);
     return;
   }
 
   unsigned Size = 0;
   for (const auto &Info : MFI.getCalleeSavedInfo()) {
     int FrameIdx = Info.getFrameIdx();
     if (MFI.getStackID(FrameIdx) != TargetStackID::Default)
       continue;
 
     Size += MFI.getObjectSize(FrameIdx);
   }
   RVFI->setCalleeSavedStackSize(Size);
 }
 

llvm/lib/Target/RISCV/RISCVInstrInfo.cpp

void RISCVInstrInfo::insertIndirectBranch(MachineBasicBlock &MBB,
                                           MachineBasicBlock &DestBB,
                                           MachineBasicBlock &RestoreBB,
                                           const DebugLoc &DL, int64_t BrOffset,
                                           RegScavenger *RS) const {
   assert(RS && "RegScavenger required for long branching");
   assert(MBB.empty() &&
          "new block should be inserted for expanding unconditional branch");
   assert(MBB.pred_size() == 1);
+  assert(RestoreBB.empty() &&
+         "restore block should be inserted for restoring clobbered registers");
 
   MachineFunction *MF = MBB.getParent();
   MachineRegisterInfo &MRI = MF->getRegInfo();
+  RISCVMachineFunctionInfo *RVFI = MF->getInfo<RISCVMachineFunctionInfo>();
+  const TargetRegisterInfo *TRI = MF->getSubtarget().getRegisterInfo();
 
   if (!isInt<32>(BrOffset))
     report_fatal_error(
         "Branch offsets outside of the signed 32-bit range not supported");
 
   // FIXME: A virtual register must be used initially, as the register
   // scavenger won't work with empty blocks (SIInstrInfo::insertIndirectBranch
   // uses the same workaround).
   Register ScratchReg = MRI.createVirtualRegister(&RISCV::GPRRegClass);
   auto II = MBB.end();
-
+  // We may also update the jump target to RestoreBB later.
   MachineInstr &MI = *BuildMI(MBB, II, DL, get(RISCV::PseudoJump))
                           .addReg(ScratchReg, RegState::Define | RegState::Dead)
                           .addMBB(&DestBB, RISCVII::MO_CALL);
 
   RS->enterBasicBlockEnd(MBB);
-  Register Scav = RS->scavengeRegisterBackwards(RISCV::GPRRegClass,
-                                                MI.getIterator(), false, 0);
-  // TODO: The case when there is no scavenged register needs special handling.
-  assert(Scav != RISCV::NoRegister && "No register is scavenged!");
-  MRI.replaceRegWith(ScratchReg, Scav);
+  Register TmpGPR =
+      RS->scavengeRegisterBackwards(RISCV::GPRRegClass, MI.getIterator(),
+                                    /*RestoreAfter=*/false, /*SpAdj=*/0,
+                                    /*AllowSpill=*/false);
+  if (TmpGPR != RISCV::NoRegister)
+    RS->setRegUsed(TmpGPR);
+  else {
+    // The case when there is no scavenged register needs special handling.
+
+    // Pick s11 because it doesn't make a difference.
+    TmpGPR = RISCV::X27;
+
+    int FrameIndex = RVFI->getBranchRelaxationScratchFrameIndex();
+    if (FrameIndex == -1)
+      report_fatal_error("underestimated function size");
+
+    storeRegToStackSlot(MBB, MI, TmpGPR, /*IsKill=*/true, FrameIndex,
+                        &RISCV::GPRRegClass, TRI);
+    TRI->eliminateFrameIndex(std::prev(MI.getIterator()),
+                             /*SpAdj=*/0, /*FIOperandNum=*/1);
+
+    MI.getOperand(1).setMBB(&RestoreBB);
+
+    loadRegFromStackSlot(RestoreBB, RestoreBB.end(), TmpGPR, FrameIndex,
+                         &RISCV::GPRRegClass, TRI);
+    TRI->eliminateFrameIndex(RestoreBB.back(),
+                             /*SpAdj=*/0, /*FIOperandNum=*/1);
+  }
+
+  MRI.replaceRegWith(ScratchReg, TmpGPR);
   MRI.clearVirtRegs();
-  RS->setRegUsed(Scav);
 }
 
 bool RISCVInstrInfo::reverseBranchCondition(
     SmallVectorImpl<MachineOperand> &Cond) const {
   assert((Cond.size() == 3) && "Invalid branch condition!");
   auto CC = static_cast<RISCVCC::CondCode>(Cond[0].getImm());
   Cond[0].setImm(getOppositeBranchCondition(CC));
   return false;

llvm/lib/Target/RISCV/RISCVMachineFunctionInfo.h

 private:
   /// FrameIndex for start of varargs area
   int VarArgsFrameIndex = 0;
   /// Size of the save area used for varargs
   int VarArgsSaveSize = 0;
   /// FrameIndex used for transferring values between 64-bit FPRs and a pair
   /// of 32-bit GPRs via the stack.
   int MoveF64FrameIndex = -1;
+  /// FrameIndex of the spill slot for the scratch register in BranchRelaxation.
+  int BranchRelaxationScratchFrameIndex = -1;
   /// Size of any opaque stack adjustment due to save/restore libcalls.
   unsigned LibCallStackSize = 0;
   /// Size of RVV stack.
   uint64_t RVVStackSize = 0;
   /// Alignment of RVV stack.
   Align RVVStackAlign;
   /// Padding required to keep RVV stack aligned within the main stack.
   uint64_t RVVPadding = 0;
public:
 
   int getMoveF64FrameIndex(MachineFunction &MF) {
     if (MoveF64FrameIndex == -1)
       MoveF64FrameIndex =
           MF.getFrameInfo().CreateStackObject(8, Align(8), false);
     return MoveF64FrameIndex;
   }
 
+  int getBranchRelaxationScratchFrameIndex() const {
+    return BranchRelaxationScratchFrameIndex;
+  }
+  void setBranchRelaxationScratchFrameIndex(int Index) {
+    BranchRelaxationScratchFrameIndex = Index;
+  }
+
   unsigned getLibCallStackSize() const { return LibCallStackSize; }
   void setLibCallStackSize(unsigned Size) { LibCallStackSize = Size; }
 
   bool useSaveRestoreLibCalls(const MachineFunction &MF) const {
     // We cannot use fixed locations for the callee saved spill slots if the
     // function uses a varargs save area, or is an interrupt handler.
     return MF.getSubtarget<RISCVSubtarget>().enableSaveRestore() &&
            VarArgsSaveSize == 0 && !MF.getFrameInfo().hasTailCall() &&

llvm/test/CodeGen/RISCV/branch-relaxation.ll

 ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
 ; RUN: llc -mtriple=riscv32 -verify-machineinstrs -filetype=obj < %s \
 ; RUN:   -o /dev/null 2>&1
 ; RUN: llc -mtriple=riscv32 -relocation-model=pic -verify-machineinstrs \
 ; RUN:   -filetype=obj < %s -o /dev/null 2>&1
-; RUN: llc -mtriple=riscv32 -verify-machineinstrs < %s | FileCheck %s
+; RUN: llc -mtriple=riscv32 -verify-machineinstrs < %s \
+; RUN:   | FileCheck %s --check-prefixes=CHECK,CHECK-RV32
 ; RUN: llc -mtriple=riscv32 -relocation-model=pic -verify-machineinstrs < %s \
-; RUN:   | FileCheck %s
+; RUN:   | FileCheck %s --check-prefixes=CHECK,CHECK-RV32
+; RUN: llc -mtriple=riscv64 -verify-machineinstrs -filetype=obj < %s \
+; RUN:   -o /dev/null 2>&1
+; RUN: llc -mtriple=riscv64 -relocation-model=pic -verify-machineinstrs \
+; RUN:   -filetype=obj < %s -o /dev/null 2>&1
+; RUN: llc -mtriple=riscv64 -verify-machineinstrs < %s \
+; RUN:   | FileCheck %s --check-prefixes=CHECK,CHECK-RV64
+; RUN: llc -mtriple=riscv64 -relocation-model=pic -verify-machineinstrs < %s \
+; RUN:   | FileCheck %s --check-prefixes=CHECK,CHECK-RV64
 
 define void @relax_bcc(i1 %a) nounwind {
 ; CHECK-LABEL: relax_bcc:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    andi a0, a0, 1
 ; CHECK-NEXT:    bnez a0, .LBB0_1
 ; CHECK-NEXT:    j .LBB0_2
 ; CHECK-NEXT:  .LBB0_1: # %iftrue

 
 tail:
   ret void
 }
 
 define i32 @relax_jal(i1 %a) nounwind {
 ; CHECK-LABEL: relax_jal:
 ; CHECK:       # %bb.0:
+; CHECK-NEXT:    addi sp, sp, -16
 ; CHECK-NEXT:    andi a0, a0, 1
 ; CHECK-NEXT:    bnez a0, .LBB1_1
-; CHECK-NEXT:  # %bb.3:
+; CHECK-NEXT:  # %bb.4:
 ; CHECK-NEXT:    jump .LBB1_2, a0
 ; CHECK-NEXT:  .LBB1_1: # %iftrue
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    .zero 1048576
 ; CHECK-NEXT:    #NO_APP
-; CHECK-NEXT:    li a0, 1
-; CHECK-NEXT:    ret
+; CHECK-NEXT:    j .LBB1_3
 ; CHECK-NEXT:  .LBB1_2: # %jmp
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    #NO_APP
+; CHECK-NEXT:  .LBB1_3: # %tail
 ; CHECK-NEXT:    li a0, 1
+; CHECK-NEXT:    addi sp, sp, 16
 ; CHECK-NEXT:    ret
   br i1 %a, label %iftrue, label %jmp
 
 jmp:
   call void asm sideeffect "", ""()
   br label %tail
 
 iftrue:
   call void asm sideeffect "", ""()
   br label %space
 
 space:
   call void asm sideeffect ".space 1048576", ""()
   br label %tail
 
 tail:
   ret i32 1
 }
+
+define void @relax_jal_spill_32() {
+; CHECK-RV32-LABEL: relax_jal_spill_32:
+; CHECK-RV32:       # %bb.0:
+; CHECK-RV32-NEXT:    addi sp, sp, -64
+; CHECK-RV32-NEXT:    .cfi_def_cfa_offset 64
+; CHECK-RV32-NEXT:    sw ra, 60(sp) # 4-byte Folded Spill
+; CHECK-RV32-NEXT:    sw s0, 56(sp) # 4-byte Folded Spill
+; CHECK-RV32-NEXT:    sw s1, 52(sp) # 4-byte Folded Spill
+; CHECK-RV32-NEXT:    sw s2, 48(sp) # 4-byte Folded Spill
+; CHECK-RV32-NEXT:    sw s3, 44(sp) # 4-byte Folded Spill
+; CHECK-RV32-NEXT:    sw s4, 40(sp) # 4-byte Folded Spill
+; CHECK-RV32-NEXT:    sw s5, 36(sp) # 4-byte Folded Spill
+; CHECK-RV32-NEXT:    sw s6, 32(sp) # 4-byte Folded Spill
+; CHECK-RV32-NEXT:    sw s7, 28(sp) # 4-byte Folded Spill
+; CHECK-RV32-NEXT:    sw s8, 24(sp) # 4-byte Folded Spill
+; CHECK-RV32-NEXT:    sw s9, 20(sp) # 4-byte Folded Spill
+; CHECK-RV32-NEXT:    sw s10, 16(sp) # 4-byte Folded Spill
+; CHECK-RV32-NEXT:    sw s11, 12(sp) # 4-byte Folded Spill
+; CHECK-RV32-NEXT:    .cfi_offset ra, -4
+; CHECK-RV32-NEXT:    .cfi_offset s0, -8
+; CHECK-RV32-NEXT:    .cfi_offset s1, -12
+; CHECK-RV32-NEXT:    .cfi_offset s2, -16
+; CHECK-RV32-NEXT:    .cfi_offset s3, -20
+; CHECK-RV32-NEXT:    .cfi_offset s4, -24
+; CHECK-RV32-NEXT:    .cfi_offset s5, -28
+; CHECK-RV32-NEXT:    .cfi_offset s6, -32
+; CHECK-RV32-NEXT:    .cfi_offset s7, -36
+; CHECK-RV32-NEXT:    .cfi_offset s8, -40
+; CHECK-RV32-NEXT:    .cfi_offset s9, -44
+; CHECK-RV32-NEXT:    .cfi_offset s10, -48
+; CHECK-RV32-NEXT:    .cfi_offset s11, -52
+; CHECK-RV32-NEXT:    #APP
+; CHECK-RV32-NEXT:    li ra, 1
+; CHECK-RV32-NEXT:    #NO_APP
+; CHECK-RV32-NEXT:    #APP
+; CHECK-RV32-NEXT:    li t0, 5
+; CHECK-RV32-NEXT:    #NO_APP
+; CHECK-RV32-NEXT:    #APP
+; CHECK-RV32-NEXT:    li t1, 6
+; CHECK-RV32-NEXT:    #NO_APP
+; CHECK-RV32-NEXT:    #APP
+; CHECK-RV32-NEXT:    li t2, 7
+; CHECK-RV32-NEXT:    #NO_APP
+; CHECK-RV32-NEXT:    #APP
+; CHECK-RV32-NEXT:    li s0, 8
+; CHECK-RV32-NEXT:    #NO_APP
+; CHECK-RV32-NEXT:    #APP
+; CHECK-RV32-NEXT:    li s1, 9
+; CHECK-RV32-NEXT:    #NO_APP
+; CHECK-RV32-NEXT:    #APP
+; CHECK-RV32-NEXT:    li a0, 10
+; CHECK-RV32-NEXT:    #NO_APP
+; CHECK-RV32-NEXT:    #APP
+; CHECK-RV32-NEXT:    li a1, 11
+; CHECK-RV32-NEXT:    #NO_APP
+; CHECK-RV32-NEXT:    #APP
+; CHECK-RV32-NEXT:    li a2, 12
+; CHECK-RV32-NEXT:    #NO_APP
+; CHECK-RV32-NEXT:    #APP
+; CHECK-RV32-NEXT:    li a3, 13
+; CHECK-RV32-NEXT:    #NO_APP
+; CHECK-RV32-NEXT:    #APP
+; CHECK-RV32-NEXT:    li a4, 14
+; CHECK-RV32-NEXT:    #NO_APP
+; CHECK-RV32-NEXT:    #APP
+; CHECK-RV32-NEXT:    li a5, 15
+; CHECK-RV32-NEXT:    #NO_APP
+; CHECK-RV32-NEXT:    #APP
+; CHECK-RV32-NEXT:    li a6, 16
+; CHECK-RV32-NEXT:    #NO_APP
+; CHECK-RV32-NEXT:    #APP
+; CHECK-RV32-NEXT:    li a7, 17
+; CHECK-RV32-NEXT:    #NO_APP
+; CHECK-RV32-NEXT:    #APP
+; CHECK-RV32-NEXT:    li s2, 18
+; CHECK-RV32-NEXT:    #NO_APP
+; CHECK-RV32-NEXT:    #APP
+; CHECK-RV32-NEXT:    li s3, 19
+; CHECK-RV32-NEXT:    #NO_APP
+; CHECK-RV32-NEXT:    #APP
+; CHECK-RV32-NEXT:    li s4, 20
+; CHECK-RV32-NEXT:    #NO_APP
+; CHECK-RV32-NEXT:    #APP
+; CHECK-RV32-NEXT:    li s5, 21
+; CHECK-RV32-NEXT:    #NO_APP
+; CHECK-RV32-NEXT:    #APP
+; CHECK-RV32-NEXT:    li s6, 22
+; CHECK-RV32-NEXT:    #NO_APP
+; CHECK-RV32-NEXT:    #APP
+; CHECK-RV32-NEXT:    li s7, 23
+; CHECK-RV32-NEXT:    #NO_APP
+; CHECK-RV32-NEXT:    #APP
+; CHECK-RV32-NEXT:    li s8, 24
+; CHECK-RV32-NEXT:    #NO_APP
+; CHECK-RV32-NEXT:    #APP
+; CHECK-RV32-NEXT:    li s9, 25
+; CHECK-RV32-NEXT:    #NO_APP
+; CHECK-RV32-NEXT:    #APP
+; CHECK-RV32-NEXT:    li s10, 26
+; CHECK-RV32-NEXT:    #NO_APP
+; CHECK-RV32-NEXT:    #APP
+; CHECK-RV32-NEXT:    li s11, 27
+; CHECK-RV32-NEXT:    #NO_APP
+; CHECK-RV32-NEXT:    #APP
+; CHECK-RV32-NEXT:    li t3, 28
+; CHECK-RV32-NEXT:    #NO_APP
+; CHECK-RV32-NEXT:    #APP
+; CHECK-RV32-NEXT:    li t4, 29
+; CHECK-RV32-NEXT:    #NO_APP
+; CHECK-RV32-NEXT:    #APP
+; CHECK-RV32-NEXT:    li t5, 30
+; CHECK-RV32-NEXT:    #NO_APP
+; CHECK-RV32-NEXT:    #APP
+; CHECK-RV32-NEXT:    li t6, 31
+; CHECK-RV32-NEXT:    #NO_APP
+; CHECK-RV32-NEXT:    beq t5, t6, .LBB2_1
+; CHECK-RV32-NEXT:  # %bb.3:
+; CHECK-RV32-NEXT:    sw s11, 0(sp)
+; CHECK-RV32-NEXT:    jump .LBB2_4, s11
+; CHECK-RV32-NEXT:  .LBB2_1: # %branch_1
+; CHECK-RV32-NEXT:    #APP
+; CHECK-RV32-NEXT:    .zero 1048576
+; CHECK-RV32-NEXT:    #NO_APP
+; CHECK-RV32-NEXT:    j .LBB2_2
+; CHECK-RV32-NEXT:  .LBB2_4: # %branch_2
+; CHECK-RV32-NEXT:    lw s11, 0(sp)
+; CHECK-RV32-NEXT:  .LBB2_2: # %branch_2
+; CHECK-RV32-NEXT:    #APP
+; CHECK-RV32-NEXT:    # reg use ra
+; CHECK-RV32-NEXT:    #NO_APP
+; CHECK-RV32-NEXT:    #APP
+; CHECK-RV32-NEXT:    # reg use t0
+; CHECK-RV32-NEXT:    #NO_APP
+; CHECK-RV32-NEXT:    #APP
+; CHECK-RV32-NEXT:    # reg use t1
+; CHECK-RV32-NEXT:    #NO_APP
+; CHECK-RV32-NEXT:    #APP
+; CHECK-RV32-NEXT:    # reg use t2
+; CHECK-RV32-NEXT:    #NO_APP
+; CHECK-RV32-NEXT:    #APP
+; CHECK-RV32-NEXT:    # reg use s0
+; CHECK-RV32-NEXT:    #NO_APP
+; CHECK-RV32-NEXT:    #APP
+; CHECK-RV32-NEXT:    # reg use s1
+; CHECK-RV32-NEXT:    #NO_APP
+; CHECK-RV32-NEXT:    #APP
+; CHECK-RV32-NEXT:    # reg use a0
+; CHECK-RV32-NEXT:    #NO_APP
+; CHECK-RV32-NEXT:    #APP
+; CHECK-RV32-NEXT:    # reg use a1
+; CHECK-RV32-NEXT:    #NO_APP
+; CHECK-RV32-NEXT:    #APP
+; CHECK-RV32-NEXT:    # reg use a2
+; CHECK-RV32-NEXT:    #NO_APP
+; CHECK-RV32-NEXT:    #APP
+; CHECK-RV32-NEXT:    # reg use a3
+; CHECK-RV32-NEXT:    #NO_APP
+; CHECK-RV32-NEXT:    #APP
+; CHECK-RV32-NEXT:    # reg use a4
+; CHECK-RV32-NEXT:    #NO_APP
+; CHECK-RV32-NEXT:    #APP
+; CHECK-RV32-NEXT:    # reg use a5
+; CHECK-RV32-NEXT:    #NO_APP
+; CHECK-RV32-NEXT:    #APP
+; CHECK-RV32-NEXT:    # reg use a6
+; CHECK-RV32-NEXT:    #NO_APP
+; CHECK-RV32-NEXT:    #APP
+; CHECK-RV32-NEXT:    # reg use a7
+; CHECK-RV32-NEXT:    #NO_APP
+; CHECK-RV32-NEXT:    #APP
+; CHECK-RV32-NEXT:    # reg use s2
+; CHECK-RV32-NEXT:    #NO_APP
+; CHECK-RV32-NEXT:    #APP
+; CHECK-RV32-NEXT:    # reg use s3
+; CHECK-RV32-NEXT:    #NO_APP
+; CHECK-RV32-NEXT:    #APP
+; CHECK-RV32-NEXT:    # reg use s4
+; CHECK-RV32-NEXT:    #NO_APP
+; CHECK-RV32-NEXT:    #APP
+; CHECK-RV32-NEXT:    # reg use s5
+; CHECK-RV32-NEXT:    #NO_APP
+; CHECK-RV32-NEXT:    #APP
+; CHECK-RV32-NEXT:    # reg use s6
+; CHECK-RV32-NEXT:    #NO_APP
+; CHECK-RV32-NEXT:    #APP
+; CHECK-RV32-NEXT:    # reg use s7
+; CHECK-RV32-NEXT:    #NO_APP
+; CHECK-RV32-NEXT:    #APP
+; CHECK-RV32-NEXT:    # reg use s8
+; CHECK-RV32-NEXT:    #NO_APP
+; CHECK-RV32-NEXT:    #APP
+; CHECK-RV32-NEXT:    # reg use s9
+; CHECK-RV32-NEXT:    #NO_APP
+; CHECK-RV32-NEXT:    #APP
+; CHECK-RV32-NEXT:    # reg use s10
+; CHECK-RV32-NEXT:    #NO_APP
+; CHECK-RV32-NEXT:    #APP
+; CHECK-RV32-NEXT:    # reg use s11
+; CHECK-RV32-NEXT:    #NO_APP
+; CHECK-RV32-NEXT:    #APP
+; CHECK-RV32-NEXT:    # reg use t3
+; CHECK-RV32-NEXT:    #NO_APP
+; CHECK-RV32-NEXT:    #APP
+; CHECK-RV32-NEXT:    # reg use t4
+; CHECK-RV32-NEXT:    #NO_APP
+; CHECK-RV32-NEXT:    #APP
+; CHECK-RV32-NEXT:    # reg use t5
+; CHECK-RV32-NEXT:    #NO_APP
+; CHECK-RV32-NEXT:    #APP
+; CHECK-RV32-NEXT:    # reg use t6
+; CHECK-RV32-NEXT:    #NO_APP
+; CHECK-RV32-NEXT:    lw ra, 60(sp) # 4-byte Folded Reload
+; CHECK-RV32-NEXT:    lw s0, 56(sp) # 4-byte Folded Reload
+; CHECK-RV32-NEXT:    lw s1, 52(sp) # 4-byte Folded Reload
+; CHECK-RV32-NEXT:    lw s2, 48(sp) # 4-byte Folded Reload
+; CHECK-RV32-NEXT:    lw s3, 44(sp) # 4-byte Folded Reload
+; CHECK-RV32-NEXT:    lw s4, 40(sp) # 4-byte Folded Reload
+; CHECK-RV32-NEXT:    lw s5, 36(sp) # 4-byte Folded Reload
+; CHECK-RV32-NEXT:    lw s6, 32(sp) # 4-byte Folded Reload
+; CHECK-RV32-NEXT:    lw s7, 28(sp) # 4-byte Folded Reload
+; CHECK-RV32-NEXT:    lw s8, 24(sp) # 4-byte Folded Reload
+; CHECK-RV32-NEXT:    lw s9, 20(sp) # 4-byte Folded Reload
+; CHECK-RV32-NEXT:    lw s10, 16(sp) # 4-byte Folded Reload
+; CHECK-RV32-NEXT:    lw s11, 12(sp) # 4-byte Folded Reload
+; CHECK-RV32-NEXT:    addi sp, sp, 64
+; CHECK-RV32-NEXT:    ret
+
+  %ra = call i32 asm sideeffect "addi ra, x0, 1", "={ra}"()
+  %t0 = call i32 asm sideeffect "addi t0, x0, 5", "={t0}"()
+  %t1 = call i32 asm sideeffect "addi t1, x0, 6", "={t1}"()
+  %t2 = call i32 asm sideeffect "addi t2, x0, 7", "={t2}"()
+  %s0 = call i32 asm sideeffect "addi s0, x0, 8", "={s0}"()
+  %s1 = call i32 asm sideeffect "addi s1, x0, 9", "={s1}"()
+  %a0 = call i32 asm sideeffect "addi a0, x0, 10", "={a0}"()
+  %a1 = call i32 asm sideeffect "addi a1, x0, 11", "={a1}"()
+  %a2 = call i32 asm sideeffect "addi a2, x0, 12", "={a2}"()
+  %a3 = call i32 asm sideeffect "addi a3, x0, 13", "={a3}"()
+  %a4 = call i32 asm sideeffect "addi a4, x0, 14", "={a4}"()
+  %a5 = call i32 asm sideeffect "addi a5, x0, 15", "={a5}"()
+  %a6 = call i32 asm sideeffect "addi a6, x0, 16", "={a6}"()
+  %a7 = call i32 asm sideeffect "addi a7, x0, 17", "={a7}"()
+  %s2 = call i32 asm sideeffect "addi s2, x0, 18", "={s2}"()
+  %s3 = call i32 asm sideeffect "addi s3, x0, 19", "={s3}"()
+  %s4 = call i32 asm sideeffect "addi s4, x0, 20", "={s4}"()
+  %s5 = call i32 asm sideeffect "addi s5, x0, 21", "={s5}"()
+  %s6 = call i32 asm sideeffect "addi s6, x0, 22", "={s6}"()
+  %s7 = call i32 asm sideeffect "addi s7, x0, 23", "={s7}"()
+  %s8 = call i32 asm sideeffect "addi s8, x0, 24", "={s8}"()
+  %s9 = call i32 asm sideeffect "addi s9, x0, 25", "={s9}"()
+  %s10 = call i32 asm sideeffect "addi s10, x0, 26", "={s10}"()
+  %s11 = call i32 asm sideeffect "addi s11, x0, 27", "={s11}"()
+  %t3 = call i32 asm sideeffect "addi t3, x0, 28", "={t3}"()
+  %t4 = call i32 asm sideeffect "addi t4, x0, 29", "={t4}"()
+  %t5 = call i32 asm sideeffect "addi t5, x0, 30", "={t5}"()
+  %t6 = call i32 asm sideeffect "addi t6, x0, 31", "={t6}"()
+
+  %cmp = icmp eq i32 %t5, %t6
+  br i1 %cmp, label %branch_1, label %branch_2
+
+branch_1:
+  call void asm sideeffect ".space 1048576", ""()
+  br label %branch_2
+
+branch_2:
+  call void asm sideeffect "# reg use $0", "{ra}"(i32 %ra)
+  call void asm sideeffect "# reg use $0", "{t0}"(i32 %t0)
+  call void asm sideeffect "# reg use $0", "{t1}"(i32 %t1)
+  call void asm sideeffect "# reg use $0", "{t2}"(i32 %t2)
+  call void asm sideeffect "# reg use $0", "{s0}"(i32 %s0)
+  call void asm sideeffect "# reg use $0", "{s1}"(i32 %s1)
+  call void asm sideeffect "# reg use $0", "{a0}"(i32 %a0)
+  call void asm sideeffect "# reg use $0", "{a1}"(i32 %a1)
+  call void asm sideeffect "# reg use $0", "{a2}"(i32 %a2)
+  call void asm sideeffect "# reg use $0", "{a3}"(i32 %a3)
+  call void asm sideeffect "# reg use $0", "{a4}"(i32 %a4)
+  call void asm sideeffect "# reg use $0", "{a5}"(i32 %a5)
+  call void asm sideeffect "# reg use $0", "{a6}"(i32 %a6)
+  call void asm sideeffect "# reg use $0", "{a7}"(i32 %a7)
+  call void asm sideeffect "# reg use $0", "{s2}"(i32 %s2)
+  call void asm sideeffect "# reg use $0", "{s3}"(i32 %s3)
+  call void asm sideeffect "# reg use $0", "{s4}"(i32 %s4)
+  call void asm sideeffect "# reg use $0", "{s5}"(i32 %s5)
+  call void asm sideeffect "# reg use $0", "{s6}"(i32 %s6)
+  call void asm sideeffect "# reg use $0", "{s7}"(i32 %s7)
+  call void asm sideeffect "# reg use $0", "{s8}"(i32 %s8)
+  call void asm sideeffect "# reg use $0", "{s9}"(i32 %s9)
+  call void asm sideeffect "# reg use $0", "{s10}"(i32 %s10)
+  call void asm sideeffect "# reg use $0", "{s11}"(i32 %s11)
+  call void asm sideeffect "# reg use $0", "{t3}"(i32 %t3)
+  call void asm sideeffect "# reg use $0", "{t4}"(i32 %t4)
+  call void asm sideeffect "# reg use $0", "{t5}"(i32 %t5)
+  call void asm sideeffect "# reg use $0", "{t6}"(i32 %t6)
+
+  ret void
+}
+
+define void @relax_jal_spill_32_adjust_spill_slot() {
+; CHECK-RV32-LABEL: relax_jal_spill_32_adjust_spill_slot:
+; CHECK-RV32:       # %bb.0:
+; CHECK-RV32-NEXT:    addi sp, sp, -2032
+; CHECK-RV32-NEXT:    .cfi_def_cfa_offset 2032
+; CHECK-RV32-NEXT:    sw ra, 2028(sp) # 4-byte Folded Spill
+; CHECK-RV32-NEXT:    sw s0, 2024(sp) # 4-byte Folded Spill
+; CHECK-RV32-NEXT:    sw s1, 2020(sp) # 4-byte Folded Spill
+; CHECK-RV32-NEXT:    sw s2, 2016(sp) # 4-byte Folded Spill
+; CHECK-RV32-NEXT:    sw s3, 2012(sp) # 4-byte Folded Spill
+; CHECK-RV32-NEXT:    sw s4, 2008(sp) # 4-byte Folded Spill
+; CHECK-RV32-NEXT:    sw s5, 2004(sp) # 4-byte Folded Spill
+; CHECK-RV32-NEXT:    sw s6, 2000(sp) # 4-byte Folded Spill
+; CHECK-RV32-NEXT:    sw s7, 1996(sp) # 4-byte Folded Spill
+; CHECK-RV32-NEXT:    sw s8, 1992(sp) # 4-byte Folded Spill
+; CHECK-RV32-NEXT:    sw s9, 1988(sp) # 4-byte Folded Spill
+; CHECK-RV32-NEXT:    sw s10, 1984(sp) # 4-byte Folded Spill
+; CHECK-RV32-NEXT:    sw s11, 1980(sp) # 4-byte Folded Spill
+; CHECK-RV32-NEXT:    .cfi_offset ra, -4
+; CHECK-RV32-NEXT:    .cfi_offset s0, -8
+; CHECK-RV32-NEXT:    .cfi_offset s1, -12
+; CHECK-RV32-NEXT:    .cfi_offset s2, -16
+; CHECK-RV32-NEXT:    .cfi_offset s3, -20
+; CHECK-RV32-NEXT:    .cfi_offset s4, -24
+; CHECK-RV32-NEXT:    .cfi_offset s5, -28
+; CHECK-RV32-NEXT:    .cfi_offset s6, -32
+; CHECK-RV32-NEXT:    .cfi_offset s7, -36
+; CHECK-RV32-NEXT:    .cfi_offset s8, -40
+; CHECK-RV32-NEXT:    .cfi_offset s9, -44
+; CHECK-RV32-NEXT:    .cfi_offset s10, -48
+; CHECK-RV32-NEXT:    .cfi_offset s11, -52
+; CHECK-RV32-NEXT:    addi s0, sp, 2032
+; CHECK-RV32-NEXT:    .cfi_def_cfa s0, 0
+; CHECK-RV32-NEXT:    lui a0, 2
+; CHECK-RV32-NEXT:    addi a0, a0, -2032
+; CHECK-RV32-NEXT:    sub sp, sp, a0
+; CHECK-RV32-NEXT:    srli a0, sp, 12
+; CHECK-RV32-NEXT:    slli sp, a0, 12
+; CHECK-RV32-NEXT:    #APP
+; CHECK-RV32-NEXT:    li ra, 1
+; CHECK-RV32-NEXT:    #NO_APP
+; CHECK-RV32-NEXT:    #APP
+; CHECK-RV32-NEXT:    li t0, 5
+; CHECK-RV32-NEXT:    #NO_APP
+; CHECK-RV32-NEXT:    #APP
+; CHECK-RV32-NEXT:    li t1, 6
+; CHECK-RV32-NEXT:    #NO_APP
+; CHECK-RV32-NEXT:    #APP
+; CHECK-RV32-NEXT:    li t2, 7
+; CHECK-RV32-NEXT:    #NO_APP
+; CHECK-RV32-NEXT:    #APP
+; CHECK-RV32-NEXT:    li s0, 8
+; CHECK-RV32-NEXT:    #NO_APP
+; CHECK-RV32-NEXT:    #APP
+; CHECK-RV32-NEXT:    li s1, 9
+; CHECK-RV32-NEXT:    #NO_APP
+; CHECK-RV32-NEXT:    #APP
+; CHECK-RV32-NEXT:    li a0, 10
+; CHECK-RV32-NEXT:    #NO_APP
+; CHECK-RV32-NEXT:    #APP
+; CHECK-RV32-NEXT:    li a1, 11
+; CHECK-RV32-NEXT:    #NO_APP
+; CHECK-RV32-NEXT:    #APP
+; CHECK-RV32-NEXT:    li a2, 12
+; CHECK-RV32-NEXT:    #NO_APP
+; CHECK-RV32-NEXT:    #APP
+; CHECK-RV32-NEXT:    li a3, 13
+; CHECK-RV32-NEXT:    #NO_APP
+; CHECK-RV32-NEXT:    #APP
+; CHECK-RV32-NEXT:    li a4, 14
+; CHECK-RV32-NEXT:    #NO_APP
+; CHECK-RV32-NEXT:    #APP
+; CHECK-RV32-NEXT:    li a5, 15
+; CHECK-RV32-NEXT:    #NO_APP
+; CHECK-RV32-NEXT:    #APP
+; CHECK-RV32-NEXT:    li a6, 16
+; CHECK-RV32-NEXT:    #NO_APP
+; CHECK-RV32-NEXT:    #APP
+; CHECK-RV32-NEXT:    li a7, 17
+; CHECK-RV32-NEXT:    #NO_APP
+; CHECK-RV32-NEXT:    #APP
+; CHECK-RV32-NEXT:    li s2, 18
+; CHECK-RV32-NEXT:    #NO_APP
+; CHECK-RV32-NEXT:    #APP
+; CHECK-RV32-NEXT:    li s3, 19
+; CHECK-RV32-NEXT:    #NO_APP
+; CHECK-RV32-NEXT:    #APP
+; CHECK-RV32-NEXT:    li s4, 20
+; CHECK-RV32-NEXT:    #NO_APP
+; CHECK-RV32-NEXT:    #APP
+; CHECK-RV32-NEXT:    li s5, 21
+; CHECK-RV32-NEXT:    #NO_APP
+; CHECK-RV32-NEXT:    #APP
+; CHECK-RV32-NEXT:    li s6, 22
+; CHECK-RV32-NEXT:    #NO_APP
+; CHECK-RV32-NEXT:    #APP
+; CHECK-RV32-NEXT:    li s7, 23
+; CHECK-RV32-NEXT:    #NO_APP
+; CHECK-RV32-NEXT:    #APP
+; CHECK-RV32-NEXT:    li s8, 24
+; CHECK-RV32-NEXT:    #NO_APP
+; CHECK-RV32-NEXT:    #APP
+; CHECK-RV32-NEXT:    li s9, 25
+; CHECK-RV32-NEXT:    #NO_APP
+; CHECK-RV32-NEXT:    #APP
+; CHECK-RV32-NEXT:    li s10, 26
+; CHECK-RV32-NEXT:    #NO_APP
+; CHECK-RV32-NEXT:    #APP
+; CHECK-RV32-NEXT:    li s11, 27
+; CHECK-RV32-NEXT:    #NO_APP
+; CHECK-RV32-NEXT:    #APP
+; CHECK-RV32-NEXT:    li t3, 28
+; CHECK-RV32-NEXT:    #NO_APP
+; CHECK-RV32-NEXT:    #APP
+; CHECK-RV32-NEXT:    li t4, 29
+; CHECK-RV32-NEXT:    #NO_APP
+; CHECK-RV32-NEXT:    #APP
+; CHECK-RV32-NEXT:    li t5, 30
+; CHECK-RV32-NEXT:    #NO_APP
+; CHECK-RV32-NEXT:    #APP
+; CHECK-RV32-NEXT:    li t6, 31
+; CHECK-RV32-NEXT:    #NO_APP
+; CHECK-RV32-NEXT:    beq t5, t6, .LBB3_1
+; CHECK-RV32-NEXT:  # %bb.3:
+; CHECK-RV32-NEXT:    sw s11, 0(sp)
+; CHECK-RV32-NEXT:    jump .LBB3_4, s11
+; CHECK-RV32-NEXT:  .LBB3_1: # %branch_1
+; CHECK-RV32-NEXT:    #APP
+; CHECK-RV32-NEXT:    .zero 1048576
+; CHECK-RV32-NEXT:    #NO_APP
+; CHECK-RV32-NEXT:    j .LBB3_2
+; CHECK-RV32-NEXT:  .LBB3_4: # %branch_2
+; CHECK-RV32-NEXT:    lw s11, 0(sp)
+; CHECK-RV32-NEXT:  .LBB3_2: # %branch_2
+; CHECK-RV32-NEXT:    #APP
+; CHECK-RV32-NEXT:    # reg use ra
+; CHECK-RV32-NEXT:    #NO_APP
+; CHECK-RV32-NEXT:    #APP
+; CHECK-RV32-NEXT:    # reg use t0
+; CHECK-RV32-NEXT:    #NO_APP
+; CHECK-RV32-NEXT:    #APP
+; CHECK-RV32-NEXT:    # reg use t1
+; CHECK-RV32-NEXT:    #NO_APP
+; CHECK-RV32-NEXT:    #APP
+; CHECK-RV32-NEXT:    # reg use t2
+; CHECK-RV32-NEXT:    #NO_APP
+; CHECK-RV32-NEXT:    #APP
+; CHECK-RV32-NEXT:    # reg use s0
+; CHECK-RV32-NEXT:    #NO_APP
+; CHECK-RV32-NEXT:    #APP
+; CHECK-RV32-NEXT:    # reg use s1
+; CHECK-RV32-NEXT:    #NO_APP
+; CHECK-RV32-NEXT:    #APP
+; CHECK-RV32-NEXT:    # reg use a0
+; CHECK-RV32-NEXT:    #NO_APP
+; CHECK-RV32-NEXT:    #APP
+; CHECK-RV32-NEXT:    # reg use a1
+; CHECK-RV32-NEXT:    #NO_APP
+; CHECK-RV32-NEXT:    #APP
+; CHECK-RV32-NEXT:    # reg use a2
+; CHECK-RV32-NEXT:    #NO_APP
+; CHECK-RV32-NEXT:    #APP
+; CHECK-RV32-NEXT:    # reg use a3
+; CHECK-RV32-NEXT:    #NO_APP
+; CHECK-RV32-NEXT:    #APP
+; CHECK-RV32-NEXT:    # reg use a4
+; CHECK-RV32-NEXT:    #NO_APP
+; CHECK-RV32-NEXT:    #APP
+; CHECK-RV32-NEXT:    # reg use a5
+; CHECK-RV32-NEXT:    #NO_APP
+; CHECK-RV32-NEXT:    #APP
+; CHECK-RV32-NEXT:    # reg use a6
+; CHECK-RV32-NEXT:    #NO_APP
+; CHECK-RV32-NEXT:    #APP
+; CHECK-RV32-NEXT:    # reg use a7
+; CHECK-RV32-NEXT:    #NO_APP
+; CHECK-RV32-NEXT:    #APP
+; CHECK-RV32-NEXT:    # reg use s2
+; CHECK-RV32-NEXT:    #NO_APP
+; CHECK-RV32-NEXT:    #APP
+; CHECK-RV32-NEXT:    # reg use s3
+; CHECK-RV32-NEXT:    #NO_APP
+; CHECK-RV32-NEXT:    #APP
+; CHECK-RV32-NEXT:    # reg use s4
+; CHECK-RV32-NEXT:    #NO_APP
+; CHECK-RV32-NEXT:    #APP
+; CHECK-RV32-NEXT:    # reg use s5
+; CHECK-RV32-NEXT:    #NO_APP
+; CHECK-RV32-NEXT:    #APP
+; CHECK-RV32-NEXT:    # reg use s6
+; CHECK-RV32-NEXT:    #NO_APP
+; CHECK-RV32-NEXT:    #APP
+; CHECK-RV32-NEXT:    # reg use s7
+; CHECK-RV32-NEXT:    #NO_APP
+; CHECK-RV32-NEXT:    #APP
+; CHECK-RV32-NEXT:    # reg use s8
+; CHECK-RV32-NEXT:    #NO_APP
+; CHECK-RV32-NEXT:    #APP
+; CHECK-RV32-NEXT:    # reg use s9
+; CHECK-RV32-NEXT:    #NO_APP
+; CHECK-RV32-NEXT:    #APP
+; CHECK-RV32-NEXT:    # reg use s10
+; CHECK-RV32-NEXT:    #NO_APP
+; CHECK-RV32-NEXT:    #APP
+; CHECK-RV32-NEXT:    # reg use s11
+; CHECK-RV32-NEXT:    #NO_APP
+; CHECK-RV32-NEXT:    #APP
+; CHECK-RV32-NEXT:    # reg use t3
+; CHECK-RV32-NEXT:    #NO_APP
+; CHECK-RV32-NEXT:    #APP
+; CHECK-RV32-NEXT:    # reg use t4
+; CHECK-RV32-NEXT:    #NO_APP
+; CHECK-RV32-NEXT:    #APP
+; CHECK-RV32-NEXT:    # reg use t5
+; CHECK-RV32-NEXT:    #NO_APP
+; CHECK-RV32-NEXT:    #APP
+; CHECK-RV32-NEXT:    # reg use t6
+; CHECK-RV32-NEXT:    #NO_APP
+; CHECK-RV32-NEXT:    lui a0, 2
+; CHECK-RV32-NEXT:    sub sp, s0, a0
+; CHECK-RV32-NEXT:    lui a0, 2
+; CHECK-RV32-NEXT:    addi a0, a0, -2032
+; CHECK-RV32-NEXT:    add sp, sp, a0
+; CHECK-RV32-NEXT:    lw ra, 2028(sp) # 4-byte Folded Reload
+; CHECK-RV32-NEXT:    lw s0, 2024(sp) # 4-byte Folded Reload
+; CHECK-RV32-NEXT:    lw s1, 2020(sp) # 4-byte Folded Reload
+; CHECK-RV32-NEXT:    lw s2, 2016(sp) # 4-byte Folded Reload
+; CHECK-RV32-NEXT:    lw s3, 2012(sp) # 4-byte Folded Reload
+; CHECK-RV32-NEXT:    lw s4, 2008(sp) # 4-byte Folded Reload
+; CHECK-RV32-NEXT:    lw s5, 2004(sp) # 4-byte Folded Reload
+; CHECK-RV32-NEXT:    lw s6, 2000(sp) # 4-byte Folded Reload
+; CHECK-RV32-NEXT:    lw s7, 1996(sp) # 4-byte Folded Reload
+; CHECK-RV32-NEXT:    lw s8, 1992(sp) # 4-byte Folded Reload
+; CHECK-RV32-NEXT:    lw s9, 1988(sp) # 4-byte Folded Reload
+; CHECK-RV32-NEXT:    lw s10, 1984(sp) # 4-byte Folded Reload
+; CHECK-RV32-NEXT:    lw s11, 1980(sp) # 4-byte Folded Reload
+; CHECK-RV32-NEXT:    addi sp, sp, 2032
+; CHECK-RV32-NEXT:    ret
+
+  ; If the stack is large and the offset of BranchRelaxationScratchFrameIndex
+  ; is out the range of 12-bit signed integer, check whether the spill slot is
+  ; adjusted to close to the stack base register.
+  %stack_obj = alloca i32, align 4096
+
+  %ra = call i32 asm sideeffect "addi ra, x0, 1", "={ra}"()
+  %t0 = call i32 asm sideeffect "addi t0, x0, 5", "={t0}"()
+  %t1 = call i32 asm sideeffect "addi t1, x0, 6", "={t1}"()
+  %t2 = call i32 asm sideeffect "addi t2, x0, 7", "={t2}"()
+  %s0 = call i32 asm sideeffect "addi s0, x0, 8", "={s0}"()
+  %s1 = call i32 asm sideeffect "addi s1, x0, 9", "={s1}"()
+  %a0 = call i32 asm sideeffect "addi a0, x0, 10", "={a0}"()
+  %a1 = call i32 asm sideeffect "addi a1, x0, 11", "={a1}"()
+  %a2 = call i32 asm sideeffect "addi a2, x0, 12", "={a2}"()
+  %a3 = call i32 asm sideeffect "addi a3, x0, 13", "={a3}"()
+  %a4 = call i32 asm sideeffect "addi a4, x0, 14", "={a4}"()
+  %a5 = call i32 asm sideeffect "addi a5, x0, 15", "={a5}"()
+  %a6 = call i32 asm sideeffect "addi a6, x0, 16", "={a6}"()
+  %a7 = call i32 asm sideeffect "addi a7, x0, 17", "={a7}"()
+  %s2 = call i32 asm sideeffect "addi s2, x0, 18", "={s2}"()
+  %s3 = call i32 asm sideeffect "addi s3, x0, 19", "={s3}"()
+  %s4 = call i32 asm sideeffect "addi s4, x0, 20", "={s4}"()
+  %s5 = call i32 asm sideeffect "addi s5, x0, 21", "={s5}"()
+  %s6 = call i32 asm sideeffect "addi s6, x0, 22", "={s6}"()
+  %s7 = call i32 asm sideeffect "addi s7, x0, 23", "={s7}"()
+  %s8 = call i32 asm sideeffect "addi s8, x0, 24", "={s8}"()
+  %s9 = call i32 asm sideeffect "addi s9, x0, 25", "={s9}"()
+  %s10 = call i32 asm sideeffect "addi s10, x0, 26", "={s10}"()
+  %s11 = call i32 asm sideeffect "addi s11, x0, 27", "={s11}"()
+  %t3 = call i32 asm sideeffect "addi t3, x0, 28", "={t3}"()
+  %t4 = call i32 asm sideeffect "addi t4, x0, 29", "={t4}"()
+  %t5 = call i32 asm sideeffect "addi t5, x0, 30", "={t5}"()
+  %t6 = call i32 asm sideeffect "addi t6, x0, 31", "={t6}"()
+
+  %cmp = icmp eq i32 %t5, %t6
+  br i1 %cmp, label %branch_1, label %branch_2
+
+branch_1:
+  call void asm sideeffect ".space 1048576", ""()
+  br label %branch_2
+
+branch_2:
+  call void asm sideeffect "# reg use $0", "{ra}"(i32 %ra)
+  call void asm sideeffect "# reg use $0", "{t0}"(i32 %t0)
+  call void asm sideeffect "# reg use $0", "{t1}"(i32 %t1)
+  call void asm sideeffect "# reg use $0", "{t2}"(i32 %t2)
+  call void asm sideeffect "# reg use $0", "{s0}"(i32 %s0)
+  call void asm sideeffect "# reg use $0", "{s1}"(i32 %s1)
+  call void asm sideeffect "# reg use $0", "{a0}"(i32 %a0)
+  call void asm sideeffect "# reg use $0", "{a1}"(i32 %a1)
+  call void asm sideeffect "# reg use $0", "{a2}"(i32 %a2)
+  call void asm sideeffect "# reg use $0", "{a3}"(i32 %a3)
+  call void asm sideeffect "# reg use $0", "{a4}"(i32 %a4)
+  call void asm sideeffect "# reg use $0", "{a5}"(i32 %a5)
+  call void asm sideeffect "# reg use $0", "{a6}"(i32 %a6)
+  call void asm sideeffect "# reg use $0", "{a7}"(i32 %a7)
+  call void asm sideeffect "# reg use $0", "{s2}"(i32 %s2)
+  call void asm sideeffect "# reg use $0", "{s3}"(i32 %s3)
+  call void asm sideeffect "# reg use $0", "{s4}"(i32 %s4)
+  call void asm sideeffect "# reg use $0", "{s5}"(i32 %s5)
+  call void asm sideeffect "# reg use $0", "{s6}"(i32 %s6)
+  call void asm sideeffect "# reg use $0", "{s7}"(i32 %s7)
+  call void asm sideeffect "# reg use $0", "{s8}"(i32 %s8)
+  call void asm sideeffect "# reg use $0", "{s9}"(i32 %s9)
+  call void asm sideeffect "# reg use $0", "{s10}"(i32 %s10)
+  call void asm sideeffect "# reg use $0", "{s11}"(i32 %s11)
+  call void asm sideeffect "# reg use $0", "{t3}"(i32 %t3)
+  call void asm sideeffect "# reg use $0", "{t4}"(i32 %t4)
+  call void asm sideeffect "# reg use $0", "{t5}"(i32 %t5)
+  call void asm sideeffect "# reg use $0", "{t6}"(i32 %t6)
+
+  ret void
+}
+
+define void @relax_jal_spill_64() {
+; CHECK-RV64-LABEL: relax_jal_spill_64:
+; CHECK-RV64:       # %bb.0:
+; CHECK-RV64-NEXT:    addi sp, sp, -112
+; CHECK-RV64-NEXT:    .cfi_def_cfa_offset 112
+; CHECK-RV64-NEXT:    sd ra, 104(sp) # 8-byte Folded Spill
+; CHECK-RV64-NEXT:    sd s0, 96(sp) # 8-byte Folded Spill
+; CHECK-RV64-NEXT:    sd s1, 88(sp) # 8-byte Folded Spill
+; CHECK-RV64-NEXT:    sd s2, 80(sp) # 8-byte Folded Spill
+; CHECK-RV64-NEXT:    sd s3, 72(sp) # 8-byte Folded Spill
+; CHECK-RV64-NEXT:    sd s4, 64(sp) # 8-byte Folded Spill
+; CHECK-RV64-NEXT:    sd s5, 56(sp) # 8-byte Folded Spill
+; CHECK-RV64-NEXT:    sd s6, 48(sp) # 8-byte Folded Spill
+; CHECK-RV64-NEXT:    sd s7, 40(sp) # 8-byte Folded Spill
+; CHECK-RV64-NEXT:    sd s8, 32(sp) # 8-byte Folded Spill
+; CHECK-RV64-NEXT:    sd s9, 24(sp) # 8-byte Folded Spill
+; CHECK-RV64-NEXT:    sd s10, 16(sp) # 8-byte Folded Spill
+; CHECK-RV64-NEXT:    sd s11, 8(sp) # 8-byte Folded Spill
+; CHECK-RV64-NEXT:    .cfi_offset ra, -8
+; CHECK-RV64-NEXT:    .cfi_offset s0, -16
+; CHECK-RV64-NEXT:    .cfi_offset s1, -24
+; CHECK-RV64-NEXT:    .cfi_offset s2, -32
+; CHECK-RV64-NEXT:    .cfi_offset s3, -40
+; CHECK-RV64-NEXT:    .cfi_offset s4, -48
+; CHECK-RV64-NEXT:    .cfi_offset s5, -56
+; CHECK-RV64-NEXT:    .cfi_offset s6, -64
+; CHECK-RV64-NEXT:    .cfi_offset s7, -72
+; CHECK-RV64-NEXT:    .cfi_offset s8, -80
+; CHECK-RV64-NEXT:    .cfi_offset s9, -88
+; CHECK-RV64-NEXT:    .cfi_offset s10, -96
+; CHECK-RV64-NEXT:    .cfi_offset s11, -104
+; CHECK-RV64-NEXT:    #APP
+; CHECK-RV64-NEXT:    li ra, 1
+; CHECK-RV64-NEXT:    #NO_APP
+; CHECK-RV64-NEXT:    #APP
+; CHECK-RV64-NEXT:    li t0, 5
+; CHECK-RV64-NEXT:    #NO_APP
+; CHECK-RV64-NEXT:    #APP
+; CHECK-RV64-NEXT:    li t1, 6
+; CHECK-RV64-NEXT:    #NO_APP
+; CHECK-RV64-NEXT:    #APP
+; CHECK-RV64-NEXT:    li t2, 7
+; CHECK-RV64-NEXT:    #NO_APP
+; CHECK-RV64-NEXT:    #APP
+; CHECK-RV64-NEXT:    li s0, 8
+; CHECK-RV64-NEXT:    #NO_APP
+; CHECK-RV64-NEXT:    #APP
+; CHECK-RV64-NEXT:    li s1, 9
+; CHECK-RV64-NEXT:    #NO_APP
+; CHECK-RV64-NEXT:    #APP
+; CHECK-RV64-NEXT:    li a0, 10
+; CHECK-RV64-NEXT:    #NO_APP
+; CHECK-RV64-NEXT:    #APP
+; CHECK-RV64-NEXT:    li a1, 11
+; CHECK-RV64-NEXT:    #NO_APP
+; CHECK-RV64-NEXT:    #APP
+; CHECK-RV64-NEXT:    li a2, 12
+; CHECK-RV64-NEXT:    #NO_APP
+; CHECK-RV64-NEXT:    #APP
+; CHECK-RV64-NEXT:    li a3, 13
+; CHECK-RV64-NEXT:    #NO_APP
+; CHECK-RV64-NEXT:    #APP
+; CHECK-RV64-NEXT:    li a4, 14
+; CHECK-RV64-NEXT:    #NO_APP
+; CHECK-RV64-NEXT:    #APP
+; CHECK-RV64-NEXT:    li a5, 15
+; CHECK-RV64-NEXT:    #NO_APP
+; CHECK-RV64-NEXT:    #APP
+; CHECK-RV64-NEXT:    li a6, 16
+; CHECK-RV64-NEXT:    #NO_APP
+; CHECK-RV64-NEXT:    #APP
+; CHECK-RV64-NEXT:    li a7, 17
+; CHECK-RV64-NEXT:    #NO_APP
+; CHECK-RV64-NEXT:    #APP
+; CHECK-RV64-NEXT:    li s2, 18
+; CHECK-RV64-NEXT:    #NO_APP
+; CHECK-RV64-NEXT:    #APP
+; CHECK-RV64-NEXT:    li s3, 19
+; CHECK-RV64-NEXT:    #NO_APP
+; CHECK-RV64-NEXT:    #APP
+; CHECK-RV64-NEXT:    li s4, 20
+; CHECK-RV64-NEXT:    #NO_APP
+; CHECK-RV64-NEXT:    #APP
+; CHECK-RV64-NEXT:    li s5, 21
+; CHECK-RV64-NEXT:    #NO_APP
+; CHECK-RV64-NEXT:    #APP
+; CHECK-RV64-NEXT:    li s6, 22
+; CHECK-RV64-NEXT:    #NO_APP
+; CHECK-RV64-NEXT:    #APP
+; CHECK-RV64-NEXT:    li s7, 23
+; CHECK-RV64-NEXT:    #NO_APP
+; CHECK-RV64-NEXT:    #APP
+; CHECK-RV64-NEXT:    li s8, 24
+; CHECK-RV64-NEXT:    #NO_APP
+; CHECK-RV64-NEXT:    #APP
+; CHECK-RV64-NEXT:    li s9, 25
+; CHECK-RV64-NEXT:    #NO_APP
+; CHECK-RV64-NEXT:    #APP
+; CHECK-RV64-NEXT:    li s10, 26
+; CHECK-RV64-NEXT:    #NO_APP
+; CHECK-RV64-NEXT:    #APP
+; CHECK-RV64-NEXT:    li s11, 27
+; CHECK-RV64-NEXT:    #NO_APP
+; CHECK-RV64-NEXT:    #APP
+; CHECK-RV64-NEXT:    li t3, 28
+; CHECK-RV64-NEXT:    #NO_APP
+; CHECK-RV64-NEXT:    #APP
+; CHECK-RV64-NEXT:    li t4, 29
+; CHECK-RV64-NEXT:    #NO_APP
+; CHECK-RV64-NEXT:    #APP
+; CHECK-RV64-NEXT:    li t5, 30
+; CHECK-RV64-NEXT:    #NO_APP
+; CHECK-RV64-NEXT:    #APP
+; CHECK-RV64-NEXT:    li t6, 31
+; CHECK-RV64-NEXT:    #NO_APP
+; CHECK-RV64-NEXT:    beq t5, t6, .LBB4_1
+; CHECK-RV64-NEXT:  # %bb.3:
+; CHECK-RV64-NEXT:    sd s11, 0(sp)
+; CHECK-RV64-NEXT:    jump .LBB4_4, s11
+; CHECK-RV64-NEXT:  .LBB4_1: # %branch_1
+; CHECK-RV64-NEXT:    #APP
+; CHECK-RV64-NEXT:    .zero 1048576
+; CHECK-RV64-NEXT:    #NO_APP
+; CHECK-RV64-NEXT:    j .LBB4_2
+; CHECK-RV64-NEXT:  .LBB4_4: # %branch_2
+; CHECK-RV64-NEXT:    ld s11, 0(sp)
+; CHECK-RV64-NEXT:  .LBB4_2: # %branch_2
+; CHECK-RV64-NEXT:    #APP
+; CHECK-RV64-NEXT:    # reg use ra
+; CHECK-RV64-NEXT:    #NO_APP
+; CHECK-RV64-NEXT:    #APP
+; CHECK-RV64-NEXT:    # reg use t0
+; CHECK-RV64-NEXT:    #NO_APP
+; CHECK-RV64-NEXT:    #APP
+; CHECK-RV64-NEXT:    # reg use t1
+; CHECK-RV64-NEXT:    #NO_APP
+; CHECK-RV64-NEXT:    #APP
+; CHECK-RV64-NEXT:    # reg use t2
+; CHECK-RV64-NEXT:    #NO_APP
+; CHECK-RV64-NEXT:    #APP
+; CHECK-RV64-NEXT:    # reg use s0
+; CHECK-RV64-NEXT:    #NO_APP
+; CHECK-RV64-NEXT:    #APP
+; CHECK-RV64-NEXT:    # reg use s1
+; CHECK-RV64-NEXT:    #NO_APP
+; CHECK-RV64-NEXT:    #APP
+; CHECK-RV64-NEXT:    # reg use a0
+; CHECK-RV64-NEXT:    #NO_APP
+; CHECK-RV64-NEXT:    #APP
+; CHECK-RV64-NEXT:    # reg use a1
+; CHECK-RV64-NEXT:    #NO_APP
+; CHECK-RV64-NEXT:    #APP
+; CHECK-RV64-NEXT:    # reg use a2
+; CHECK-RV64-NEXT:    #NO_APP
+; CHECK-RV64-NEXT:    #APP
+; CHECK-RV64-NEXT:    # reg use a3
+; CHECK-RV64-NEXT:    #NO_APP
+; CHECK-RV64-NEXT:    #APP
+; CHECK-RV64-NEXT:    # reg use a4
+; CHECK-RV64-NEXT:    #NO_APP
+; CHECK-RV64-NEXT:    #APP
+; CHECK-RV64-NEXT:    # reg use a5
+; CHECK-RV64-NEXT:    #NO_APP
+; CHECK-RV64-NEXT:    #APP
+; CHECK-RV64-NEXT:    # reg use a6
+; CHECK-RV64-NEXT:    #NO_APP
+; CHECK-RV64-NEXT:    #APP
+; CHECK-RV64-NEXT:    # reg use a7
+; CHECK-RV64-NEXT:    #NO_APP
+; CHECK-RV64-NEXT:    #APP
+; CHECK-RV64-NEXT:    # reg use s2
+; CHECK-RV64-NEXT:    #NO_APP
+; CHECK-RV64-NEXT:    #APP
+; CHECK-RV64-NEXT:    # reg use s3
+; CHECK-RV64-NEXT:    #NO_APP
+; CHECK-RV64-NEXT:    #APP
+; CHECK-RV64-NEXT:    # reg use s4
+; CHECK-RV64-NEXT:    #NO_APP
+; CHECK-RV64-NEXT:    #APP
+; CHECK-RV64-NEXT:    # reg use s5
+; CHECK-RV64-NEXT:    #NO_APP
+; CHECK-RV64-NEXT:    #APP
+; CHECK-RV64-NEXT:    # reg use s6
+; CHECK-RV64-NEXT:    #NO_APP
+; CHECK-RV64-NEXT:    #APP
+; CHECK-RV64-NEXT:    # reg use s7
+; CHECK-RV64-NEXT:    #NO_APP
+; CHECK-RV64-NEXT:    #APP
+; CHECK-RV64-NEXT:    # reg use s8
+; CHECK-RV64-NEXT:    #NO_APP
+; CHECK-RV64-NEXT:    #APP
+; CHECK-RV64-NEXT:    # reg use s9
+; CHECK-RV64-NEXT:    #NO_APP
+; CHECK-RV64-NEXT:    #APP
+; CHECK-RV64-NEXT:    # reg use s10
+; CHECK-RV64-NEXT:    #NO_APP
+; CHECK-RV64-NEXT:    #APP
+; CHECK-RV64-NEXT:    # reg use s11
+; CHECK-RV64-NEXT:    #NO_APP
+; CHECK-RV64-NEXT:    #APP
+; CHECK-RV64-NEXT:    # reg use t3
+; CHECK-RV64-NEXT:    #NO_APP
+; CHECK-RV64-NEXT:    #APP
+; CHECK-RV64-NEXT:    # reg use t4
+; CHECK-RV64-NEXT:    #NO_APP
+; CHECK-RV64-NEXT:    #APP
+; CHECK-RV64-NEXT:    # reg use t5
+; CHECK-RV64-NEXT:    #NO_APP
+; CHECK-RV64-NEXT:    #APP
+; CHECK-RV64-NEXT:    # reg use t6
+; CHECK-RV64-NEXT:    #NO_APP
+; CHECK-RV64-NEXT:    ld ra, 104(sp) # 8-byte Folded Reload
+; CHECK-RV64-NEXT:    ld s0, 96(sp) # 8-byte Folded Reload
+; CHECK-RV64-NEXT:    ld s1, 88(sp) # 8-byte Folded Reload
+; CHECK-RV64-NEXT:    ld s2, 80(sp) # 8-byte Folded Reload
+; CHECK-RV64-NEXT:    ld s3, 72(sp) # 8-byte Folded Reload
+; CHECK-RV64-NEXT:    ld s4, 64(sp) # 8-byte Folded Reload
+; CHECK-RV64-NEXT:    ld s5, 56(sp) # 8-byte Folded Reload
+; CHECK-RV64-NEXT:    ld s6, 48(sp) # 8-byte Folded Reload
+; CHECK-RV64-NEXT:    ld s7, 40(sp) # 8-byte Folded Reload
+; CHECK-RV64-NEXT:    ld s8, 32(sp) # 8-byte Folded Reload
+; CHECK-RV64-NEXT:    ld s9, 24(sp) # 8-byte Folded Reload
+; CHECK-RV64-NEXT:    ld s10, 16(sp) # 8-byte Folded Reload
+; CHECK-RV64-NEXT:    ld s11, 8(sp) # 8-byte Folded Reload
+; CHECK-RV64-NEXT:    addi sp, sp, 112
+; CHECK-RV64-NEXT:    ret
+
+  %ra = call i64 asm sideeffect "addi ra, x0, 1", "={ra}"()
+  %t0 = call i64 asm sideeffect "addi t0, x0, 5", "={t0}"()
+  %t1 = call i64 asm sideeffect "addi t1, x0, 6", "={t1}"()
+  %t2 = call i64 asm sideeffect "addi t2, x0, 7", "={t2}"()
+  %s0 = call i64 asm sideeffect "addi s0, x0, 8", "={s0}"()
+  %s1 = call i64 asm sideeffect "addi s1, x0, 9", "={s1}"()
+  %a0 = call i64 asm sideeffect "addi a0, x0, 10", "={a0}"()
+  %a1 = call i64 asm sideeffect "addi a1, x0, 11", "={a1}"()
+  %a2 = call i64 asm sideeffect "addi a2, x0, 12", "={a2}"()
+  %a3 = call i64 asm sideeffect "addi a3, x0, 13", "={a3}"()
+  %a4 = call i64 asm sideeffect "addi a4, x0, 14", "={a4}"()
+  %a5 = call i64 asm sideeffect "addi a5, x0, 15", "={a5}"()
+  %a6 = call i64 asm sideeffect "addi a6, x0, 16", "={a6}"()
+  %a7 = call i64 asm sideeffect "addi a7, x0, 17", "={a7}"()
+  %s2 = call i64 asm sideeffect "addi s2, x0, 18", "={s2}"()
+  %s3 = call i64 asm sideeffect "addi s3, x0, 19", "={s3}"()
+  %s4 = call i64 asm sideeffect "addi s4, x0, 20", "={s4}"()
+  %s5 = call i64 asm sideeffect "addi s5, x0, 21", "={s5}"()
+  %s6 = call i64 asm sideeffect "addi s6, x0, 22", "={s6}"()
+  %s7 = call i64 asm sideeffect "addi s7, x0, 23", "={s7}"()
+  %s8 = call i64 asm sideeffect "addi s8, x0, 24", "={s8}"()
+  %s9 = call i64 asm sideeffect "addi s9, x0, 25", "={s9}"()
+  %s10 = call i64 asm sideeffect "addi s10, x0, 26", "={s10}"()
+  %s11 = call i64 asm sideeffect "addi s11, x0, 27", "={s11}"()
+  %t3 = call i64 asm sideeffect "addi t3, x0, 28", "={t3}"()
+  %t4 = call i64 asm sideeffect "addi t4, x0, 29", "={t4}"()
+  %t5 = call i64 asm sideeffect "addi t5, x0, 30", "={t5}"()
+  %t6 = call i64 asm sideeffect "addi t6, x0, 31", "={t6}"()
+
+  %cmp = icmp eq i64 %t5, %t6
+  br i1 %cmp, label %branch_1, label %branch_2
+
+branch_1:
+  call void asm sideeffect ".space 1048576", ""()
+  br label %branch_2
+
+branch_2:
+  call void asm sideeffect "# reg use $0", "{ra}"(i64 %ra)
+  call void asm sideeffect "# reg use $0", "{t0}"(i64 %t0)
+  call void asm sideeffect "# reg use $0", "{t1}"(i64 %t1)
+  call void asm sideeffect "# reg use $0", "{t2}"(i64 %t2)
+  call void asm sideeffect "# reg use $0", "{s0}"(i64 %s0)
+  call void asm sideeffect "# reg use $0", "{s1}"(i64 %s1)
+  call void asm sideeffect "# reg use $0", "{a0}"(i64 %a0)
+  call void asm sideeffect "# reg use $0", "{a1}"(i64 %a1)
+  call void asm sideeffect "# reg use $0", "{a2}"(i64 %a2)
+  call void asm sideeffect "# reg use $0", "{a3}"(i64 %a3)
+  call void asm sideeffect "# reg use $0", "{a4}"(i64 %a4)
+  call void asm sideeffect "# reg use $0", "{a5}"(i64 %a5)
+  call void asm sideeffect "# reg use $0", "{a6}"(i64 %a6)
+  call void asm sideeffect "# reg use $0", "{a7}"(i64 %a7)
+  call void asm sideeffect "# reg use $0", "{s2}"(i64 %s2)
+  call void asm sideeffect "# reg use $0", "{s3}"(i64 %s3)
+  call void asm sideeffect "# reg use $0", "{s4}"(i64 %s4)
+  call void asm sideeffect "# reg use $0", "{s5}"(i64 %s5)
+  call void asm sideeffect "# reg use $0", "{s6}"(i64 %s6)
+  call void asm sideeffect "# reg use $0", "{s7}"(i64 %s7)
+  call void asm sideeffect "# reg use $0", "{s8}"(i64 %s8)
+  call void asm sideeffect "# reg use $0", "{s9}"(i64 %s9)
+  call void asm sideeffect "# reg use $0", "{s10}"(i64 %s10)
+  call void asm sideeffect "# reg use $0", "{s11}"(i64 %s11)
+  call void asm sideeffect "# reg use $0", "{t3}"(i64 %t3)
+  call void asm sideeffect "# reg use $0", "{t4}"(i64 %t4)
+  call void asm sideeffect "# reg use $0", "{t5}"(i64 %t5)
+  call void asm sideeffect "# reg use $0", "{t6}"(i64 %t6)
+
+  ret void
+}
+
+define void @relax_jal_spill_64_adjust_spill_slot() {
+; CHECK-RV64-LABEL: relax_jal_spill_64_adjust_spill_slot:
+; CHECK-RV64:       # %bb.0:
+; CHECK-RV64-NEXT:    addi sp, sp, -2032
+; CHECK-RV64-NEXT:    .cfi_def_cfa_offset 2032
+; CHECK-RV64-NEXT:    sd ra, 2024(sp) # 8-byte Folded Spill
+; CHECK-RV64-NEXT:    sd s0, 2016(sp) # 8-byte Folded Spill
+; CHECK-RV64-NEXT:    sd s1, 2008(sp) # 8-byte Folded Spill
+; CHECK-RV64-NEXT:    sd s2, 2000(sp) # 8-byte Folded Spill
+; CHECK-RV64-NEXT:    sd s3, 1992(sp) # 8-byte Folded Spill
+; CHECK-RV64-NEXT:    sd s4, 1984(sp) # 8-byte Folded Spill
+; CHECK-RV64-NEXT:    sd s5, 1976(sp) # 8-byte Folded Spill
+; CHECK-RV64-NEXT:    sd s6, 1968(sp) # 8-byte Folded Spill
+; CHECK-RV64-NEXT:    sd s7, 1960(sp) # 8-byte Folded Spill
+; CHECK-RV64-NEXT:    sd s8, 1952(sp) # 8-byte Folded Spill
+; CHECK-RV64-NEXT:    sd s9, 1944(sp) # 8-byte Folded Spill
+; CHECK-RV64-NEXT:    sd s10, 1936(sp) # 8-byte Folded Spill
+; CHECK-RV64-NEXT:    sd s11, 1928(sp) # 8-byte Folded Spill
+; CHECK-RV64-NEXT:    .cfi_offset ra, -8
+; CHECK-RV64-NEXT:    .cfi_offset s0, -16
+; CHECK-RV64-NEXT:    .cfi_offset s1, -24
+; CHECK-RV64-NEXT:    .cfi_offset s2, -32
+; CHECK-RV64-NEXT:    .cfi_offset s3, -40
+; CHECK-RV64-NEXT:    .cfi_offset s4, -48
+; CHECK-RV64-NEXT:    .cfi_offset s5, -56
+; CHECK-RV64-NEXT:    .cfi_offset s6, -64
+; CHECK-RV64-NEXT:    .cfi_offset s7, -72
+; CHECK-RV64-NEXT:    .cfi_offset s8, -80
+; CHECK-RV64-NEXT:    .cfi_offset s9, -88
+; CHECK-RV64-NEXT:    .cfi_offset s10, -96
+; CHECK-RV64-NEXT:    .cfi_offset s11, -104
+; CHECK-RV64-NEXT:    addi s0, sp, 2032
+; CHECK-RV64-NEXT:    .cfi_def_cfa s0, 0
+; CHECK-RV64-NEXT:    lui a0, 2
+; CHECK-RV64-NEXT:    addiw a0, a0, -2032
+; CHECK-RV64-NEXT:    sub sp, sp, a0
+; CHECK-RV64-NEXT:    srli a0, sp, 12
+; CHECK-RV64-NEXT:    slli sp, a0, 12
+; CHECK-RV64-NEXT:    #APP
+; CHECK-RV64-NEXT:    li ra, 1
+; CHECK-RV64-NEXT:    #NO_APP
+; CHECK-RV64-NEXT:    #APP
+; CHECK-RV64-NEXT:    li t0, 5
+; CHECK-RV64-NEXT:    #NO_APP
+; CHECK-RV64-NEXT:    #APP
+; CHECK-RV64-NEXT:    li t1, 6
+; CHECK-RV64-NEXT:    #NO_APP
+; CHECK-RV64-NEXT:    #APP
+; CHECK-RV64-NEXT:    li t2, 7
+; CHECK-RV64-NEXT:    #NO_APP
+; CHECK-RV64-NEXT:    #APP
+; CHECK-RV64-NEXT:    li s0, 8
+; CHECK-RV64-NEXT:    #NO_APP
+; CHECK-RV64-NEXT:    #APP
+; CHECK-RV64-NEXT:    li s1, 9
+; CHECK-RV64-NEXT:    #NO_APP
+; CHECK-RV64-NEXT:    #APP
+; CHECK-RV64-NEXT:    li a0, 10
+; CHECK-RV64-NEXT:    #NO_APP
+; CHECK-RV64-NEXT:    #APP
+; CHECK-RV64-NEXT:    li a1, 11
+; CHECK-RV64-NEXT:    #NO_APP
+; CHECK-RV64-NEXT:    #APP
+; CHECK-RV64-NEXT:    li a2, 12
+; CHECK-RV64-NEXT:    #NO_APP
+; CHECK-RV64-NEXT:    #APP
+; CHECK-RV64-NEXT:    li a3, 13
+; CHECK-RV64-NEXT:    #NO_APP
+; CHECK-RV64-NEXT:    #APP
+; CHECK-RV64-NEXT:    li a4, 14
+; CHECK-RV64-NEXT:    #NO_APP
+; CHECK-RV64-NEXT:    #APP
+; CHECK-RV64-NEXT:    li a5, 15
+; CHECK-RV64-NEXT:    #NO_APP
+; CHECK-RV64-NEXT:    #APP
+; CHECK-RV64-NEXT:    li a6, 16
+; CHECK-RV64-NEXT:    #NO_APP
+; CHECK-RV64-NEXT:    #APP
+; CHECK-RV64-NEXT:    li a7, 17
+; CHECK-RV64-NEXT:    #NO_APP
+; CHECK-RV64-NEXT:    #APP
+; CHECK-RV64-NEXT:    li s2, 18
+; CHECK-RV64-NEXT:    #NO_APP
+; CHECK-RV64-NEXT:    #APP
+; CHECK-RV64-NEXT:    li s3, 19
+; CHECK-RV64-NEXT:    #NO_APP
+; CHECK-RV64-NEXT:    #APP
+; CHECK-RV64-NEXT:    li s4, 20
+; CHECK-RV64-NEXT:    #NO_APP
+; CHECK-RV64-NEXT:    #APP
+; CHECK-RV64-NEXT:    li s5, 21
+; CHECK-RV64-NEXT:    #NO_APP
+; CHECK-RV64-NEXT:    #APP
+; CHECK-RV64-NEXT:    li s6, 22
+; CHECK-RV64-NEXT:    #NO_APP
+; CHECK-RV64-NEXT:    #APP
+; CHECK-RV64-NEXT:    li s7, 23
+; CHECK-RV64-NEXT:    #NO_APP
+; CHECK-RV64-NEXT:    #APP
+; CHECK-RV64-NEXT:    li s8, 24
+; CHECK-RV64-NEXT:    #NO_APP
+; CHECK-RV64-NEXT:    #APP
+; CHECK-RV64-NEXT:    li s9, 25
+; CHECK-RV64-NEXT:    #NO_APP
+; CHECK-RV64-NEXT:    #APP
+; CHECK-RV64-NEXT:    li s10, 26
+; CHECK-RV64-NEXT:    #NO_APP
+; CHECK-RV64-NEXT:    #APP
+; CHECK-RV64-NEXT:    li s11, 27
+; CHECK-RV64-NEXT:    #NO_APP
+; CHECK-RV64-NEXT:    #APP
+; CHECK-RV64-NEXT:    li t3, 28
+; CHECK-RV64-NEXT:    #NO_APP
+; CHECK-RV64-NEXT:    #APP
+; CHECK-RV64-NEXT:    li t4, 29
+; CHECK-RV64-NEXT:    #NO_APP
+; CHECK-RV64-NEXT:    #APP
+; CHECK-RV64-NEXT:    li t5, 30
+; CHECK-RV64-NEXT:    #NO_APP
+; CHECK-RV64-NEXT:    #APP
+; CHECK-RV64-NEXT:    li t6, 31
+; CHECK-RV64-NEXT:    #NO_APP
+; CHECK-RV64-NEXT:    beq t5, t6, .LBB5_1
+; CHECK-RV64-NEXT:  # %bb.3:
+; CHECK-RV64-NEXT:    sd s11, 0(sp)
+; CHECK-RV64-NEXT:    jump .LBB5_4, s11
+; CHECK-RV64-NEXT:  .LBB5_1: # %branch_1
+; CHECK-RV64-NEXT:    #APP
+; CHECK-RV64-NEXT:    .zero 1048576
+; CHECK-RV64-NEXT:    #NO_APP
+; CHECK-RV64-NEXT:    j .LBB5_2
+; CHECK-RV64-NEXT:  .LBB5_4: # %branch_2
+; CHECK-RV64-NEXT:    ld s11, 0(sp)
+; CHECK-RV64-NEXT:  .LBB5_2: # %branch_2
+; CHECK-RV64-NEXT:    #APP
+; CHECK-RV64-NEXT:    # reg use ra
+; CHECK-RV64-NEXT:    #NO_APP
+; CHECK-RV64-NEXT:    #APP
+; CHECK-RV64-NEXT:    # reg use t0
+; CHECK-RV64-NEXT:    #NO_APP
+; CHECK-RV64-NEXT:    #APP
+; CHECK-RV64-NEXT:    # reg use t1
+; CHECK-RV64-NEXT:    #NO_APP
+; CHECK-RV64-NEXT:    #APP
+; CHECK-RV64-NEXT:    # reg use t2
+; CHECK-RV64-NEXT:    #NO_APP
+; CHECK-RV64-NEXT:    #APP
+; CHECK-RV64-NEXT:    # reg use s0
+; CHECK-RV64-NEXT:    #NO_APP
+; CHECK-RV64-NEXT:    #APP
+; CHECK-RV64-NEXT:    # reg use s1
+; CHECK-RV64-NEXT:    #NO_APP
+; CHECK-RV64-NEXT:    #APP
+; CHECK-RV64-NEXT:    # reg use a0
+; CHECK-RV64-NEXT:    #NO_APP
+; CHECK-RV64-NEXT:    #APP
+; CHECK-RV64-NEXT:    # reg use a1
+; CHECK-RV64-NEXT:    #NO_APP
+; CHECK-RV64-NEXT:    #APP
+; CHECK-RV64-NEXT:    # reg use a2
+; CHECK-RV64-NEXT:    #NO_APP
+; CHECK-RV64-NEXT:    #APP
+; CHECK-RV64-NEXT:    # reg use a3
+; CHECK-RV64-NEXT:    #NO_APP
+; CHECK-RV64-NEXT:    #APP
+; CHECK-RV64-NEXT:    # reg use a4
+; CHECK-RV64-NEXT:    #NO_APP
+; CHECK-RV64-NEXT:    #APP
+; CHECK-RV64-NEXT:    # reg use a5
+; CHECK-RV64-NEXT:    #NO_APP
+; CHECK-RV64-NEXT:    #APP
+; CHECK-RV64-NEXT:    # reg use a6
+; CHECK-RV64-NEXT:    #NO_APP
+; CHECK-RV64-NEXT:    #APP
+; CHECK-RV64-NEXT:    # reg use a7
+; CHECK-RV64-NEXT:    #NO_APP
+; CHECK-RV64-NEXT:    #APP
+; CHECK-RV64-NEXT:    # reg use s2
+; CHECK-RV64-NEXT:    #NO_APP
+; CHECK-RV64-NEXT:    #APP
+; CHECK-RV64-NEXT:    # reg use s3
+; CHECK-RV64-NEXT:    #NO_APP
+; CHECK-RV64-NEXT:    #APP
+; CHECK-RV64-NEXT:    # reg use s4
+; CHECK-RV64-NEXT:    #NO_APP
+; CHECK-RV64-NEXT:    #APP
+; CHECK-RV64-NEXT:    # reg use s5
+; CHECK-RV64-NEXT:    #NO_APP
+; CHECK-RV64-NEXT:    #APP
+; CHECK-RV64-NEXT:    # reg use s6
+; CHECK-RV64-NEXT:    #NO_APP
+; CHECK-RV64-NEXT:    #APP
+; CHECK-RV64-NEXT:    # reg use s7
+; CHECK-RV64-NEXT:    #NO_APP
+; CHECK-RV64-NEXT:    #APP
+; CHECK-RV64-NEXT:    # reg use s8
+; CHECK-RV64-NEXT:    #NO_APP
+; CHECK-RV64-NEXT:    #APP
+; CHECK-RV64-NEXT:    # reg use s9
+; CHECK-RV64-NEXT:    #NO_APP
+; CHECK-RV64-NEXT:    #APP
+; CHECK-RV64-NEXT:    # reg use s10
+; CHECK-RV64-NEXT:    #NO_APP
+; CHECK-RV64-NEXT:    #APP
+; CHECK-RV64-NEXT:    # reg use s11
+; CHECK-RV64-NEXT:    #NO_APP
+; CHECK-RV64-NEXT:    #APP
+; CHECK-RV64-NEXT:    # reg use t3
+; CHECK-RV64-NEXT:    #NO_APP
+; CHECK-RV64-NEXT:    #APP
+; CHECK-RV64-NEXT:    # reg use t4
+; CHECK-RV64-NEXT:    #NO_APP
+; CHECK-RV64-NEXT:    #APP
+; CHECK-RV64-NEXT:    # reg use t5
+; CHECK-RV64-NEXT:    #NO_APP
+; CHECK-RV64-NEXT:    #APP
+; CHECK-RV64-NEXT:    # reg use t6
+; CHECK-RV64-NEXT:    #NO_APP
+; CHECK-RV64-NEXT:    lui a0, 2
+; CHECK-RV64-NEXT:    sub sp, s0, a0
+; CHECK-RV64-NEXT:    lui a0, 2
+; CHECK-RV64-NEXT:    addiw a0, a0, -2032
+; CHECK-RV64-NEXT:    add sp, sp, a0
+; CHECK-RV64-NEXT:    ld ra, 2024(sp) # 8-byte Folded Reload
+; CHECK-RV64-NEXT:    ld s0, 2016(sp) # 8-byte Folded Reload
+; CHECK-RV64-NEXT:    ld s1, 2008(sp) # 8-byte Folded Reload
+; CHECK-RV64-NEXT:    ld s2, 2000(sp) # 8-byte Folded Reload
+; CHECK-RV64-NEXT:    ld s3, 1992(sp) # 8-byte Folded Reload
+; CHECK-RV64-NEXT:    ld s4, 1984(sp) # 8-byte Folded Reload
+; CHECK-RV64-NEXT:    ld s5, 1976(sp) # 8-byte Folded Reload
+; CHECK-RV64-NEXT:    ld s6, 1968(sp) # 8-byte Folded Reload
+; CHECK-RV64-NEXT:    ld s7, 1960(sp) # 8-byte Folded Reload
+; CHECK-RV64-NEXT:    ld s8, 1952(sp) # 8-byte Folded Reload
+; CHECK-RV64-NEXT:    ld s9, 1944(sp) # 8-byte Folded Reload
+; CHECK-RV64-NEXT:    ld s10, 1936(sp) # 8-byte Folded Reload
+; CHECK-RV64-NEXT:    ld s11, 1928(sp) # 8-byte Folded Reload
+; CHECK-RV64-NEXT:    addi sp, sp, 2032
+; CHECK-RV64-NEXT:    ret
+
+  ; If the stack is large and the offset of BranchRelaxationScratchFrameIndex
+  ; is out the range of 12-bit signed integer, check whether the spill slot is
+  ; adjusted to close to the stack base register.
+  %stack_obj = alloca i64, align 4096
+
+  %ra = call i64 asm sideeffect "addi ra, x0, 1", "={ra}"()
+  %t0 = call i64 asm sideeffect "addi t0, x0, 5", "={t0}"()
+  %t1 = call i64 asm sideeffect "addi t1, x0, 6", "={t1}"()
+  %t2 = call i64 asm sideeffect "addi t2, x0, 7", "={t2}"()
+  %s0 = call i64 asm sideeffect "addi s0, x0, 8", "={s0}"()
+  %s1 = call i64 asm sideeffect "addi s1, x0, 9", "={s1}"()
+  %a0 = call i64 asm sideeffect "addi a0, x0, 10", "={a0}"()
+  %a1 = call i64 asm sideeffect "addi a1, x0, 11", "={a1}"()
+  %a2 = call i64 asm sideeffect "addi a2, x0, 12", "={a2}"()
+  %a3 = call i64 asm sideeffect "addi a3, x0, 13", "={a3}"()
+  %a4 = call i64 asm sideeffect "addi a4, x0, 14", "={a4}"()
+  %a5 = call i64 asm sideeffect "addi a5, x0, 15", "={a5}"()
+  %a6 = call i64 asm sideeffect "addi a6, x0, 16", "={a6}"()
+  %a7 = call i64 asm sideeffect "addi a7, x0, 17", "={a7}"()
+  %s2 = call i64 asm sideeffect "addi s2, x0, 18", "={s2}"()
+  %s3 = call i64 asm sideeffect "addi s3, x0, 19", "={s3}"()
+  %s4 = call i64 asm sideeffect "addi s4, x0, 20", "={s4}"()
+  %s5 = call i64 asm sideeffect "addi s5, x0, 21", "={s5}"()
+  %s6 = call i64 asm sideeffect "addi s6, x0, 22", "={s6}"()
+  %s7 = call i64 asm sideeffect "addi s7, x0, 23", "={s7}"()
+  %s8 = call i64 asm sideeffect "addi s8, x0, 24", "={s8}"()
+  %s9 = call i64 asm sideeffect "addi s9, x0, 25", "={s9}"()
+  %s10 = call i64 asm sideeffect "addi s10, x0, 26", "={s10}"()
+  %s11 = call i64 asm sideeffect "addi s11, x0, 27", "={s11}"()
+  %t3 = call i64 asm sideeffect "addi t3, x0, 28", "={t3}"()
+  %t4 = call i64 asm sideeffect "addi t4, x0, 29", "={t4}"()
+  %t5 = call i64 asm sideeffect "addi t5, x0, 30", "={t5}"()
+  %t6 = call i64 asm sideeffect "addi t6, x0, 31", "={t6}"()
+
+  %cmp = icmp eq i64 %t5, %t6
+  br i1 %cmp, label %branch_1, label %branch_2
+
+branch_1:
+  call void asm sideeffect ".space 1048576", ""()
+  br label %branch_2
+
+branch_2:
+  call void asm sideeffect "# reg use $0", "{ra}"(i64 %ra)
+  call void asm sideeffect "# reg use $0", "{t0}"(i64 %t0)
+  call void asm sideeffect "# reg use $0", "{t1}"(i64 %t1)
+  call void asm sideeffect "# reg use $0", "{t2}"(i64 %t2)
+  call void asm sideeffect "# reg use $0", "{s0}"(i64 %s0)
+  call void asm sideeffect "# reg use $0", "{s1}"(i64 %s1)
+  call void asm sideeffect "# reg use $0", "{a0}"(i64 %a0)
+  call void asm sideeffect "# reg use $0", "{a1}"(i64 %a1)
+  call void asm sideeffect "# reg use $0", "{a2}"(i64 %a2)
+  call void asm sideeffect "# reg use $0", "{a3}"(i64 %a3)
+  call void asm sideeffect "# reg use $0", "{a4}"(i64 %a4)
+  call void asm sideeffect "# reg use $0", "{a5}"(i64 %a5)
+  call void asm sideeffect "# reg use $0", "{a6}"(i64 %a6)
+  call void asm sideeffect "# reg use $0", "{a7}"(i64 %a7)
+  call void asm sideeffect "# reg use $0", "{s2}"(i64 %s2)
+  call void asm sideeffect "# reg use $0", "{s3}"(i64 %s3)
+  call void asm sideeffect "# reg use $0", "{s4}"(i64 %s4)
+  call void asm sideeffect "# reg use $0", "{s5}"(i64 %s5)
+  call void asm sideeffect "# reg use $0", "{s6}"(i64 %s6)
+  call void asm sideeffect "# reg use $0", "{s7}"(i64 %s7)
+  call void asm sideeffect "# reg use $0", "{s8}"(i64 %s8)
+  call void asm sideeffect "# reg use $0", "{s9}"(i64 %s9)
+  call void asm sideeffect "# reg use $0", "{s10}"(i64 %s10)
+  call void asm sideeffect "# reg use $0", "{s11}"(i64 %s11)
+  call void asm sideeffect "# reg use $0", "{t3}"(i64 %t3)
+  call void asm sideeffect "# reg use $0", "{t4}"(i64 %t4)
+  call void asm sideeffect "# reg use $0", "{t5}"(i64 %t5)
+  call void asm sideeffect "# reg use $0", "{t6}"(i64 %t6)
+
+  ret void
+}