ShrinkWrap: Ignore load/store until stack address computation

[MatzeB]

Summary

No longer conservatively assume a load/store accesses the stack when we can prove that we did not compute any stack-relative address up to this point in the program.

We do this in a cheap not-quite-a-dataflow-analysis: Assume NoStackAddressUsed when all predecessors of a block already guarantee it. Process blocks in reverse post order to guarantee that except for loop headers we have processed all predecessors of a block before processing the block itself. For loops we accept the conservative answer as they are unlikely to be shrink-wrappable anyway.

[MatzeB]

llvm/lib/CodeGen/ShrinkWrap.cpp

304

Was wondering if its worth removing this complex expression with the patch in place, but it's hard to argue that there could be examples where this helps placing the epilogue, so I left it in for now...

[qcolombet]

Hi @MatzeB

The change looks good to me, but can we flip the logic NoStackAddressUsed => StackAddressUsed?

I find the !Noxxx or Noxxx = false difficult to reason about :).

Cheers,
-Quentin

[thegameg]

LG, thanks! Agreed with @qcolombet on the double negation.

[MatzeB]

address review comments

llvm/lib/CodeGen/ShrinkWrap.cpp

class ShrinkWrap : public MachineFunctionPass {
   using SetOfRegs = SmallSetVector<unsigned, 16>;
 
   /// Registers that need to be saved for the current function.
   mutable SetOfRegs CurrentCSRs;
 
   /// Current MachineFunction.
   MachineFunction *MachineFunc = nullptr;
 
+  /// Is `true` for block numbers where we can guarantee no stack access
+  /// or computation of stack-relative addresses on any CFG path including
+  /// the block itself.
+  BitVector StackAddressUsedBlockInfo;
+
   /// Check if \p MI uses or defines a callee-saved register or
   /// a frame index. If this is the case, this means \p MI must happen
   /// after Save and before Restore.
-  bool useOrDefCSROrFI(const MachineInstr &MI, RegScavenger *RS) const;
+  bool useOrDefCSROrFI(const MachineInstr &MI, RegScavenger *RS,
+                       bool StackAddressUsed) const;
 
   const SetOfRegs &getCurrentCSRs(RegScavenger *RS) const {
     if (CurrentCSRs.empty()) {
       BitVector SavedRegs;
       const TargetFrameLowering *TFI =
           MachineFunc->getSubtarget().getFrameLowering();
 
       TFI->determineCalleeSaves(*MachineFunc, SavedRegs, RS);
class ShrinkWrap : public MachineFunctionPass {
   /// the region that is dominated by Save and post-dominated by Restore
   /// and Save and Restore still match the safe point definition.
   /// Such point may not exist and Save and/or Restore may be null after
   /// this call.
   void updateSaveRestorePoints(MachineBasicBlock &MBB, RegScavenger *RS);
 
   // Try to find safe point based on dominance and block frequency without
   // any change in IR.
-  bool performShrinkWrapping(MachineFunction &MF, RegScavenger *RS);
+  bool performShrinkWrapping(
+      const ReversePostOrderTraversal<MachineBasicBlock *> &RPOT,
+      RegScavenger *RS);
 
   /// This function tries to split the restore point if doing so can shrink the
   /// save point further. \return True if restore point is split.
   bool postShrinkWrapping(bool HasCandidate, MachineFunction &MF,
                           RegScavenger *RS);
 
   /// This function analyzes if the restore point can split to create a new
   /// restore point. This function collects

 INITIALIZE_PASS_BEGIN(ShrinkWrap, DEBUG_TYPE, "Shrink Wrap Pass", false, false)
 INITIALIZE_PASS_DEPENDENCY(MachineBlockFrequencyInfo)
 INITIALIZE_PASS_DEPENDENCY(MachineDominatorTree)
 INITIALIZE_PASS_DEPENDENCY(MachinePostDominatorTree)
 INITIALIZE_PASS_DEPENDENCY(MachineLoopInfo)
 INITIALIZE_PASS_DEPENDENCY(MachineOptimizationRemarkEmitterPass)
 INITIALIZE_PASS_END(ShrinkWrap, DEBUG_TYPE, "Shrink Wrap Pass", false, false)
 
-bool ShrinkWrap::useOrDefCSROrFI(const MachineInstr &MI,
-                                 RegScavenger *RS) const {
+bool ShrinkWrap::useOrDefCSROrFI(const MachineInstr &MI, RegScavenger *RS,
+                                 bool StackAddressUsed) const {
   /// Check if \p Op is known to access an address not on the function's stack .
   /// At the moment, accesses where the underlying object is a global, function
   /// argument, or jump table are considered non-stack accesses. Note that the
   /// caller's stack may get accessed when passing an argument via the stack,
   /// but not the stack of the current function.
   ///
   auto IsKnownNonStackPtr = [](MachineMemOperand *Op) {
     if (Op->getValue()) {
       const Value *UO = getUnderlyingObject(Op->getValue());
       if (!UO)
         return false;
       if (auto *Arg = dyn_cast<Argument>(UO))
         return !Arg->hasPassPointeeByValueCopyAttr();
       return isa<GlobalValue>(UO);
     }
     if (const PseudoSourceValue *PSV = Op->getPseudoValue())
       return PSV->isJumpTable();
     return false;
   };
-  // This prevents premature stack popping when occurs a indirect stack
-  // access.  It is overly aggressive for the moment.
-  // TODO:
-  //       - Further, data dependency and alias analysis can validate
-  //         that load and stores never derive from the stack pointer.
-  if (MI.mayLoadOrStore() &&
+  // Load/store operations may access the stack indirectly when we previously
+  // computed an address to a stack location.
+  if (StackAddressUsed && MI.mayLoadOrStore() &&
       (MI.isCall() || MI.hasUnmodeledSideEffects() || MI.memoperands_empty() ||
        !all_of(MI.memoperands(), IsKnownNonStackPtr)))
     return true;
 
   if (MI.getOpcode() == FrameSetupOpcode ||
       MI.getOpcode() == FrameDestroyOpcode) {
     LLVM_DEBUG(dbgs() << "Frame instruction: " << MI << '\n');
     return true;

 // 2. CleanPreds - preds of CurRestore that arent DirtyPreds
 bool ShrinkWrap::checkIfRestoreSplittable(
     const MachineBasicBlock *CurRestore,
     const DenseSet<const MachineBasicBlock *> &ReachableByDirty,
     SmallVectorImpl<MachineBasicBlock *> &DirtyPreds,
     SmallVectorImpl<MachineBasicBlock *> &CleanPreds,
     const TargetInstrInfo *TII, RegScavenger *RS) {
   for (const MachineInstr &MI : *CurRestore)
-    if (useOrDefCSROrFI(MI, RS))
+    if (useOrDefCSROrFI(MI, RS, /*StackAddressUsed=*/true))
       return false;
 
   for (MachineBasicBlock *PredBB : CurRestore->predecessors()) {
     if (!isAnalyzableBB(*TII, *PredBB))
       return false;
 
     if (ReachableByDirty.count(PredBB))
       DirtyPreds.push_back(PredBB);
bool ShrinkWrap::postShrinkWrapping(bool HasCandidate, MachineFunction &MF,
 
   DenseSet<const MachineBasicBlock *> DirtyBBs;
   for (MachineBasicBlock &MBB : MF) {
     if (MBB.isEHPad()) {
       DirtyBBs.insert(&MBB);
       continue;
     }
     for (const MachineInstr &MI : MBB)
-      if (useOrDefCSROrFI(MI, RS)) {
+      if (useOrDefCSROrFI(MI, RS, /*StackAddressUsed=*/true)) {
         DirtyBBs.insert(&MBB);
         break;
       }
   }
 
   // Find blocks reachable from the use or def of CSRs/FI.
   DenseSet<const MachineBasicBlock *> ReachableByDirty;
   collectBlocksReachableByDirty(DirtyBBs, ReachableByDirty);
else if (MPDT->getNode(&MBB)) // If the block is not in the post dom tree, it
     Restore = MPDT->findNearestCommonDominator(Restore, &MBB);
   else
     Restore = nullptr; // Abort, we can't find a restore point in this case.
 
   // Make sure we would be able to insert the restore code before the
   // terminator.
   if (Restore == &MBB) {
     for (const MachineInstr &Terminator : MBB.terminators()) {
-      if (!useOrDefCSROrFI(Terminator, RS))
+      if (!useOrDefCSROrFI(Terminator, RS, /*StackAddressUsed=*/true))
         continue;
       // One of the terminator needs to happen before the restore point.
       if (MBB.succ_empty()) {
         Restore = nullptr; // Abort, we can't find a restore point in this case.
         break;
       }
       // Look for a restore point that post-dominates all the successors.
       // The immediate post-dominator is what we are looking for.
ORE->emit([&]() {
     return MachineOptimizationRemarkMissed(DEBUG_TYPE, RemarkName, Loc, MBB)
            << RemarkMessage;
   });
 
   LLVM_DEBUG(dbgs() << RemarkMessage << '\n');
   return false;
 }
 
-bool ShrinkWrap::performShrinkWrapping(MachineFunction &MF, RegScavenger *RS) {
-  for (MachineBasicBlock &MBB : MF) {
-    LLVM_DEBUG(dbgs() << "Look into: " << MBB.getNumber() << ' '
-                      << MBB.getName() << '\n');
+bool ShrinkWrap::performShrinkWrapping(
+    const ReversePostOrderTraversal<MachineBasicBlock *> &RPOT,
+    RegScavenger *RS) {
+  for (MachineBasicBlock *MBB : RPOT) {
+    LLVM_DEBUG(dbgs() << "Look into: " << printMBBReference(*MBB) << '\n');
 
-    if (MBB.isEHFuncletEntry())
+    if (MBB->isEHFuncletEntry())
       return giveUpWithRemarks(ORE, "UnsupportedEHFunclets",
                                "EH Funclets are not supported yet.",
-                               MBB.front().getDebugLoc(), &MBB);
+                               MBB->front().getDebugLoc(), MBB);
 
-    if (MBB.isEHPad() || MBB.isInlineAsmBrIndirectTarget()) {
+    if (MBB->isEHPad() || MBB->isInlineAsmBrIndirectTarget()) {
       // Push the prologue and epilogue outside of the region that may throw (or
       // jump out via inlineasm_br), by making sure that all the landing pads
       // are at least at the boundary of the save and restore points.  The
       // problem is that a basic block can jump out from the middle in these
       // cases, which we do not handle.
-      updateSaveRestorePoints(MBB, RS);
+      updateSaveRestorePoints(*MBB, RS);
       if (!ArePointsInteresting()) {
         LLVM_DEBUG(dbgs() << "EHPad/inlineasm_br prevents shrink-wrapping\n");
         return false;
       }
       continue;
     }
 
-    for (const MachineInstr &MI : MBB) {
-      if (!useOrDefCSROrFI(MI, RS))
-        continue;
+    bool StackAddressUsed = false;
+    // Check if we found any stack accesses in the predecessors. We are not
+    // doing a full dataflow analysis here to keep things simple but just
+    // rely on a reverse portorder traversal (RPOT) to guarantee predecessors
+    // are already processed except for loops (and accept the conservative
+    // result for loops).
+    for (const MachineBasicBlock *Pred : MBB->predecessors()) {
+      if (StackAddressUsedBlockInfo.test(Pred->getNumber())) {
+        StackAddressUsed = true;
+        break;
+      }
+    }
+
+    for (const MachineInstr &MI : *MBB) {
+      if (useOrDefCSROrFI(MI, RS, StackAddressUsed)) {
         // Save (resp. restore) point must dominate (resp. post dominate)
         // MI. Look for the proper basic block for those.
-      updateSaveRestorePoints(MBB, RS);
+        updateSaveRestorePoints(*MBB, RS);
         // If we are at a point where we cannot improve the placement of
         // save/restore instructions, just give up.
         if (!ArePointsInteresting()) {
           LLVM_DEBUG(dbgs() << "No Shrink wrap candidate found\n");
           return false;
         }
         // No need to look for other instructions, this basic block
         // will already be part of the handled region.
+        StackAddressUsed = true;
         break;
       }
     }
+    StackAddressUsedBlockInfo[MBB->getNumber()] = StackAddressUsed;
+  }
   if (!ArePointsInteresting()) {
     // If the points are not interesting at this point, then they must be null
     // because it means we did not encounter any frame/CSR related code.
     // Otherwise, we would have returned from the previous loop.
     assert(!Save && !Restore && "We miss a shrink-wrap opportunity?!");
     LLVM_DEBUG(dbgs() << "Nothing to shrink-wrap\n");
     return false;
   }
 
   LLVM_DEBUG(dbgs() << "\n ** Results **\nFrequency of the Entry: " << EntryFreq
                     << '\n');
 
-  const TargetFrameLowering *TFI = MF.getSubtarget().getFrameLowering();
+  const TargetFrameLowering *TFI =
+      MachineFunc->getSubtarget().getFrameLowering();
   do {
     LLVM_DEBUG(dbgs() << "Shrink wrap candidates (#, Name, Freq):\nSave: "
-                      << Save->getNumber() << ' ' << Save->getName() << ' '
+                      << printMBBReference(*Save) << ' '
                       << MBFI->getBlockFreq(Save).getFrequency()
-                      << "\nRestore: " << Restore->getNumber() << ' '
-                      << Restore->getName() << ' '
+                      << "\nRestore: " << printMBBReference(*Restore) << ' '
                       << MBFI->getBlockFreq(Restore).getFrequency() << '\n');
 
     bool IsSaveCheap, TargetCanUseSaveAsPrologue = false;
     if (((IsSaveCheap = EntryFreq >= MBFI->getBlockFreq(Save).getFrequency()) &&
          EntryFreq >= MBFI->getBlockFreq(Restore).getFrequency()) &&
         ((TargetCanUseSaveAsPrologue = TFI->canUseAsPrologue(*Save)) &&
          TFI->canUseAsEpilogue(*Restore)))
       break;
bool ShrinkWrap::runOnMachineFunction(MachineFunction &MF) {
   }
 
   const TargetRegisterInfo *TRI = MF.getSubtarget().getRegisterInfo();
   std::unique_ptr<RegScavenger> RS(
       TRI->requiresRegisterScavenging(MF) ? new RegScavenger() : nullptr);
 
   bool Changed = false;
 
-  bool HasCandidate = performShrinkWrapping(MF, RS.get());
+  StackAddressUsedBlockInfo.resize(MF.getNumBlockIDs(), true);
+  bool HasCandidate = performShrinkWrapping(RPOT, RS.get());
+  StackAddressUsedBlockInfo.clear();
   Changed = postShrinkWrapping(HasCandidate, MF, RS.get());
   if (!HasCandidate && !Changed)
     return false;
   if (!ArePointsInteresting())
     return Changed;
 
   LLVM_DEBUG(dbgs() << "Final shrink wrap candidates:\nSave: "
-                    << Save->getNumber() << ' ' << Save->getName()
-                    << "\nRestore: " << Restore->getNumber() << ' '
-                    << Restore->getName() << '\n');
+                    << printMBBReference(*Save) << ' '
+                    << "\nRestore: " << printMBBReference(*Restore) << '\n');
 
   MachineFrameInfo &MFI = MF.getFrameInfo();
   MFI.setSavePoint(Save);
   MFI.setRestorePoint(Restore);
   ++NumCandidates;
   return Changed;
 }
 

llvm/test/CodeGen/AArch64/branch-relax-cbz.ll

 ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
 ; RUN: llc -mtriple=aarch64-apple-darwin -aarch64-cbz-offset-bits=3 < %s | FileCheck %s
 
 
 define void @split_block_no_fallthrough(i64 %val) #0 {
 ; CHECK-LABEL: split_block_no_fallthrough:
 ; CHECK:       ; %bb.0: ; %bb
-; CHECK-NEXT:    stp x29, x30, [sp, #-16]! ; 16-byte Folded Spill
 ; CHECK-NEXT:    cmn x0, #5
 ; CHECK-NEXT:    b.le LBB0_3
 ; CHECK-NEXT:  ; %bb.1: ; %b3
 ; CHECK-NEXT:    ldr w8, [x8]
 ; CHECK-NEXT:    cbnz w8, LBB0_2
 ; CHECK-NEXT:    b LBB0_4
 ; CHECK-NEXT:  LBB0_2: ; %common.ret
-; CHECK-NEXT:    ldp x29, x30, [sp], #16 ; 16-byte Folded Reload
 ; CHECK-NEXT:    ret
 ; CHECK-NEXT:  LBB0_3: ; %b2
-; CHECK-NEXT:    mov w0, #93
+; CHECK-NEXT:    stp x29, x30, [sp, #-16]! ; 16-byte Folded Spill
+; CHECK-NEXT:    mov w0, #93 ; =0x5d
 ; CHECK-NEXT:    bl _extfunc
-; CHECK-NEXT:    cbnz w0, LBB0_2
-; CHECK-NEXT:  LBB0_4: ; %b7
-; CHECK-NEXT:    mov w0, #13
 ; CHECK-NEXT:    ldp x29, x30, [sp], #16 ; 16-byte Folded Reload
+; CHECK-NEXT:    cbz w0, LBB0_4
+; CHECK-NEXT:    b LBB0_2
+; CHECK-NEXT:  LBB0_4: ; %b7
+; CHECK-NEXT:    mov w0, #13 ; =0xd
 ; CHECK-NEXT:    b _extfunc
 bb:
   %c0 = icmp sgt i64 %val, -5
   br i1 %c0, label %b3, label %b2
 
 b2:
   %v0 = tail call i32 @extfunc(i32 93)
   %c1 = icmp eq i32 %v0, 0

llvm/test/CodeGen/AArch64/optimize-cond-branch.ll

 ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
 ; RUN: llc -verify-machineinstrs -o - %s | FileCheck %s
 target triple = "arm64--"
 
 ; AArch64InstrInfo::optimizeCondBranch() optimizes the
 ; "x = and y, 256; cmp x, 0; br" from an "and; cbnz" to a tbnz instruction.
 ; It forgot to clear the a flag resulting in a MachineVerifier complaint.
 ;
 ; Writing a stable/simple test is tricky since most tbz instructions are already
 ; formed in SelectionDAG, optimizeCondBranch() only triggers if the and
 ; instruction is in a different block than the conditional jump.
 
 define void @func() uwtable {
 ; CHECK-LABEL: func:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    mov w8, #1
+; CHECK-NEXT:    mov w8, #1 // =0x1
 ; CHECK-NEXT:    cbnz w8, .LBB0_3
 ; CHECK-NEXT:  // %bb.1: // %b1
-; CHECK-NEXT:    str x30, [sp, #-16]! // 8-byte Folded Spill
-; CHECK-NEXT:    .cfi_def_cfa_offset 16
-; CHECK-NEXT:    .cfi_offset w30, -16
-; CHECK-NEXT:    .cfi_remember_state
 ; CHECK-NEXT:    cbz wzr, .LBB0_4
 ; CHECK-NEXT:  // %bb.2: // %b3
 ; CHECK-NEXT:    ldr w8, [x8]
 ; CHECK-NEXT:    and w0, w8, #0x100
-; CHECK-NEXT:    ldr x30, [sp], #16 // 8-byte Folded Reload
-; CHECK-NEXT:    .cfi_def_cfa_offset 0
-; CHECK-NEXT:    .cfi_restore w30
 ; CHECK-NEXT:    cbz w0, .LBB0_5
 ; CHECK-NEXT:  .LBB0_3: // %common.ret.sink.split
 ; CHECK-NEXT:    b extfunc
 ; CHECK-NEXT:  .LBB0_4: // %b2
-; CHECK-NEXT:    .cfi_restore_state
+; CHECK-NEXT:    str x30, [sp, #-16]! // 8-byte Folded Spill
+; CHECK-NEXT:    .cfi_def_cfa_offset 16
+; CHECK-NEXT:    .cfi_offset w30, -16
 ; CHECK-NEXT:    bl extfunc
 ; CHECK-NEXT:    ldr x30, [sp], #16 // 8-byte Folded Reload
 ; CHECK-NEXT:    .cfi_def_cfa_offset 0
 ; CHECK-NEXT:    .cfi_restore w30
 ; CHECK-NEXT:    cbnz w0, .LBB0_3
 ; CHECK-NEXT:  .LBB0_5: // %common.ret
 ; CHECK-NEXT:    ret
   %c0 = icmp sgt i64 0, 0

llvm/test/CodeGen/ARM/arm-shrink-wrapping.ll

 ; FIXME: This is flakey passing by finding 'bl' somewhere amongst the debug
 ; info (like labels named 'line_table) not because it's found a bl instruction.
 ;
 ; bl
 define float @debug_info(float %gamma, float %slopeLimit, i1 %or.cond, double %tmp) "frame-pointer"="all" {
 ; ARM-ENABLE-LABEL: debug_info:
 ; ARM-ENABLE:       Lfunc_begin12:
 ; ARM-ENABLE-NEXT:  @ %bb.0: @ %bb
+; ARM-ENABLE-NEXT:    tst r2, #1
+; ARM-ENABLE-NEXT:    beq LBB12_2
+; ARM-ENABLE-NEXT:  @ %bb.1: @ %bb3
 ; ARM-ENABLE-NEXT:    push {r4, r7, lr}
 ; ARM-ENABLE-NEXT:    add r7, sp, #4
 ; ARM-ENABLE-NEXT:    sub r4, sp, #16
 ; ARM-ENABLE-NEXT:    bfc r4, #0, #4
 ; ARM-ENABLE-NEXT:    mov sp, r4
-; ARM-ENABLE-NEXT:    tst r2, #1
-; ARM-ENABLE-NEXT:    vst1.64 {d8, d9}, [r4:128]
-; ARM-ENABLE-NEXT:    beq LBB12_2
-; ARM-ENABLE-NEXT:  @ %bb.1: @ %bb3
 ; ARM-ENABLE-NEXT:    ldr r1, [r7, #8]
+; ARM-ENABLE-NEXT:    mov r2, r3
+; ARM-ENABLE-NEXT:    vst1.64 {d8, d9}, [r4:128]
 ; ARM-ENABLE-NEXT:    vmov s16, r0
 ; ARM-ENABLE-NEXT:    mov r0, r3
-; ARM-ENABLE-NEXT:    mov r2, r3
 ; ARM-ENABLE-NEXT:    vmov d9, r3, r1
 ; ARM-ENABLE-NEXT:    mov r3, r1
 ; ARM-ENABLE-NEXT:    bl _pow
 ; ARM-ENABLE-NEXT:    vmov.f32 s0, #1.000000e+00
+; ARM-ENABLE-NEXT:    mov r4, sp
 ; ARM-ENABLE-NEXT:    vmov.f64 d16, #1.000000e+00
 ; ARM-ENABLE-NEXT:    vadd.f64 d16, d9, d16
 ; ARM-ENABLE-NEXT:    vcmp.f32 s16, s0
 ; ARM-ENABLE-NEXT:    vmrs APSR_nzcv, fpscr
 ; ARM-ENABLE-NEXT:    vmov d17, r0, r1
 ; ARM-ENABLE-NEXT:    vmov.f64 d18, d9
 ; ARM-ENABLE-NEXT:    vadd.f64 d17, d17, d17
 ; ARM-ENABLE-NEXT:    vmovgt.f64 d18, d16
 ; ARM-ENABLE-NEXT:    vcmp.f64 d18, d9
 ; ARM-ENABLE-NEXT:    vmrs APSR_nzcv, fpscr
 ; ARM-ENABLE-NEXT:    vmovne.f64 d9, d17
 ; ARM-ENABLE-NEXT:    vcvt.f32.f64 s0, d9
-; ARM-ENABLE-NEXT:    b LBB12_3
+; ARM-ENABLE-NEXT:    vld1.64 {d8, d9}, [r4:128]
+; ARM-ENABLE-NEXT:    sub sp, r7, #4
+; ARM-ENABLE-NEXT:    pop {r4, r7, lr}
+; ARM-ENABLE-NEXT:    vmov r0, s0
+; ARM-ENABLE-NEXT:    bx lr
 ; ARM-ENABLE-NEXT:  LBB12_2:
 ; ARM-ENABLE-NEXT:    vldr s0, LCPI12_0
-; ARM-ENABLE-NEXT:  LBB12_3: @ %bb13
-; ARM-ENABLE-NEXT:    mov r4, sp
-; ARM-ENABLE-NEXT:    vld1.64 {d8, d9}, [r4:128]
 ; ARM-ENABLE-NEXT:    vmov r0, s0
-; ARM-ENABLE-NEXT:    sub sp, r7, #4
-; ARM-ENABLE-NEXT:    pop {r4, r7, pc}
+; ARM-ENABLE-NEXT:    bx lr
 ; ARM-ENABLE-NEXT:    .p2align 2
-; ARM-ENABLE-NEXT:  @ %bb.4:
+; ARM-ENABLE-NEXT:  @ %bb.3:
 ; ARM-ENABLE-NEXT:    .data_region
 ; ARM-ENABLE-NEXT:  LCPI12_0:
 ; ARM-ENABLE-NEXT:    .long 0x00000000 @ float 0
 ; ARM-ENABLE-NEXT:    .end_data_region
 ; ARM-ENABLE-NEXT:  Lfunc_end12:
 ;
 ; ARM-DISABLE-LABEL: debug_info:
 ; ARM-DISABLE:       Lfunc_begin12:

 ; ARM-DISABLE-NEXT:  LCPI12_0:
 ; ARM-DISABLE-NEXT:    .long 0x00000000 @ float 0
 ; ARM-DISABLE-NEXT:    .end_data_region
 ; ARM-DISABLE-NEXT:  Lfunc_end12:
 ;
 ; THUMB-ENABLE-LABEL: debug_info:
 ; THUMB-ENABLE:       Lfunc_begin12:
 ; THUMB-ENABLE-NEXT:  @ %bb.0: @ %bb
+; THUMB-ENABLE-NEXT:    lsls r1, r2, #31
+; THUMB-ENABLE-NEXT:    beq LBB12_2
+; THUMB-ENABLE-NEXT:  @ %bb.1: @ %bb3
 ; THUMB-ENABLE-NEXT:    push {r4, r7, lr}
 ; THUMB-ENABLE-NEXT:    add r7, sp, #4
 ; THUMB-ENABLE-NEXT:    sub.w r4, sp, #16
 ; THUMB-ENABLE-NEXT:    bfc r4, #0, #4
 ; THUMB-ENABLE-NEXT:    mov sp, r4
-; THUMB-ENABLE-NEXT:    lsls r1, r2, #31
-; THUMB-ENABLE-NEXT:    vst1.64 {d8, d9}, [r4:128]
-; THUMB-ENABLE-NEXT:    beq LBB12_2
-; THUMB-ENABLE-NEXT:  @ %bb.1: @ %bb3
 ; THUMB-ENABLE-NEXT:    ldr r1, [r7, #8]
+; THUMB-ENABLE-NEXT:    mov r2, r3
+; THUMB-ENABLE-NEXT:    vst1.64 {d8, d9}, [r4:128]
 ; THUMB-ENABLE-NEXT:    vmov s16, r0
 ; THUMB-ENABLE-NEXT:    mov r0, r3
-; THUMB-ENABLE-NEXT:    mov r2, r3
 ; THUMB-ENABLE-NEXT:    vmov d9, r3, r1
 ; THUMB-ENABLE-NEXT:    mov r3, r1
 ; THUMB-ENABLE-NEXT:    bl _pow
 ; THUMB-ENABLE-NEXT:    vmov.f32 s0, #1.000000e+00
+; THUMB-ENABLE-NEXT:    mov r4, sp
 ; THUMB-ENABLE-NEXT:    vmov.f64 d16, #1.000000e+00
 ; THUMB-ENABLE-NEXT:    vmov.f64 d18, d9
 ; THUMB-ENABLE-NEXT:    vcmp.f32 s16, s0
 ; THUMB-ENABLE-NEXT:    vadd.f64 d16, d9, d16
 ; THUMB-ENABLE-NEXT:    vmrs APSR_nzcv, fpscr
 ; THUMB-ENABLE-NEXT:    it gt
 ; THUMB-ENABLE-NEXT:    vmovgt.f64 d18, d16
 ; THUMB-ENABLE-NEXT:    vcmp.f64 d18, d9
 ; THUMB-ENABLE-NEXT:    vmov d17, r0, r1
 ; THUMB-ENABLE-NEXT:    vmrs APSR_nzcv, fpscr
 ; THUMB-ENABLE-NEXT:    vadd.f64 d17, d17, d17
 ; THUMB-ENABLE-NEXT:    it ne
 ; THUMB-ENABLE-NEXT:    vmovne.f64 d9, d17
 ; THUMB-ENABLE-NEXT:    vcvt.f32.f64 s0, d9
-; THUMB-ENABLE-NEXT:    b LBB12_3
-; THUMB-ENABLE-NEXT:  LBB12_2:
-; THUMB-ENABLE-NEXT:    vldr s0, LCPI12_0
-; THUMB-ENABLE-NEXT:  LBB12_3: @ %bb13
-; THUMB-ENABLE-NEXT:    mov r4, sp
 ; THUMB-ENABLE-NEXT:    vld1.64 {d8, d9}, [r4:128]
 ; THUMB-ENABLE-NEXT:    subs r4, r7, #4
-; THUMB-ENABLE-NEXT:    vmov r0, s0
 ; THUMB-ENABLE-NEXT:    mov sp, r4
-; THUMB-ENABLE-NEXT:    pop {r4, r7, pc}
+; THUMB-ENABLE-NEXT:    pop.w {r4, r7, lr}
+; THUMB-ENABLE-NEXT:    vmov r0, s0
+; THUMB-ENABLE-NEXT:    bx lr
+; THUMB-ENABLE-NEXT:  LBB12_2:
+; THUMB-ENABLE-NEXT:    vldr s0, LCPI12_0
+; THUMB-ENABLE-NEXT:    vmov r0, s0
+; THUMB-ENABLE-NEXT:    bx lr
 ; THUMB-ENABLE-NEXT:    .p2align 2
-; THUMB-ENABLE-NEXT:  @ %bb.4:
+; THUMB-ENABLE-NEXT:  @ %bb.3:
 ; THUMB-ENABLE-NEXT:    .data_region
 ; THUMB-ENABLE-NEXT:  LCPI12_0:
 ; THUMB-ENABLE-NEXT:    .long 0x00000000 @ float 0
 ; THUMB-ENABLE-NEXT:    .end_data_region
 ; THUMB-ENABLE-NEXT:  Lfunc_end12:
 ;
 ; THUMB-DISABLE-LABEL: debug_info:
 ; THUMB-DISABLE:       Lfunc_begin12:

llvm/test/CodeGen/ARM/machine-cse-cmp.ll

 @foo = external global i32
 @bar = external global [250 x i8], align 1
 
 ; CSE of cmp across BB boundary
 ; rdar://10660865
 define void @f2() nounwind ssp {
 ; CHECK-LABEL: f2:
 ; CHECK:       @ %bb.0: @ %entry
-; CHECK-NEXT:    push {lr}
 ; CHECK-NEXT:    movw r0, :lower16:(L_foo$non_lazy_ptr-(LPC1_0+8))
 ; CHECK-NEXT:    movt r0, :upper16:(L_foo$non_lazy_ptr-(LPC1_0+8))
 ; CHECK-NEXT:  LPC1_0:
 ; CHECK-NEXT:    ldr r0, [pc, r0]
 ; CHECK-NEXT:    ldr r2, [r0]
 ; CHECK-NEXT:    cmp r2, #1
-; CHECK-NEXT:    poplt {lr}
 ; CHECK-NEXT:    bxlt lr
 ; CHECK-NEXT:  LBB1_1: @ %for.body.lr.ph
+; CHECK-NEXT:    push {lr}
 ; CHECK-NEXT:    movw r0, :lower16:(L_bar$non_lazy_ptr-(LPC1_1+8))
 ; CHECK-NEXT:    movle r2, #1
 ; CHECK-NEXT:    movt r0, :upper16:(L_bar$non_lazy_ptr-(LPC1_1+8))
 ; CHECK-NEXT:    mov r1, #0
 ; CHECK-NEXT:  LPC1_1:
 ; CHECK-NEXT:    ldr r0, [pc, r0]
 ; CHECK-NEXT:    bl _memset
 ; CHECK-NEXT:    trap

llvm/test/CodeGen/ARM/machine-sink-multidef.ll

 ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
 ; RUN: llc < %s -mtriple=arm-none-eabi | FileCheck %s
 
 %struct.anon.1.19.23.27.35.49.55.57.59.61.89.95 = type { i32, i32 }
 
 @e = external constant [2 x %struct.anon.1.19.23.27.35.49.55.57.59.61.89.95], align 4
 @f = external global i32, align 4
 
 define arm_aapcscc void @g() {
 ; CHECK-LABEL: g:
 ; CHECK:       @ %bb.0: @ %entry
-; CHECK-NEXT:    .save {r11, lr}
-; CHECK-NEXT:    push {r11, lr}
 ; CHECK-NEXT:    ldr r0, .LCPI0_0
 ; CHECK-NEXT:    mov r2, #0
 ; CHECK-NEXT:    ldr r1, .LCPI0_1
 ; CHECK-NEXT:    cmp r2, #0
 ; CHECK-NEXT:    ldr r0, [r0]
 ; CHECK-NEXT:    ldr r0, [r1, r0, lsl #3]!
 ; CHECK-NEXT:    moveq r0, #0
 ; CHECK-NEXT:    cmp r2, #0
-; CHECK-NEXT:    popne {r11, lr}
 ; CHECK-NEXT:    movne pc, lr
 ; CHECK-NEXT:  .LBB0_1: @ %if.then5
+; CHECK-NEXT:    .save {r11, lr}
+; CHECK-NEXT:    push {r11, lr}
 ; CHECK-NEXT:    ldr r1, [r1, #4]
 ; CHECK-NEXT:    bl k
 ; CHECK-NEXT:    .p2align 2
 ; CHECK-NEXT:  @ %bb.2:
 ; CHECK-NEXT:  .LCPI0_0:
 ; CHECK-NEXT:    .long f
 ; CHECK-NEXT:  .LCPI0_1:
 ; CHECK-NEXT:    .long e

llvm/test/CodeGen/ARM/swifterror.ll

handler:
   ret float 1.0
 }
 
 ; "foo_if" is a function that takes a swifterror parameter, it sets swifterror
 ; under a certain condition.
 define float @foo_if(ptr swifterror %error_ptr_ref, i32 %cc) {
 ; CHECK-APPLE-LABEL: foo_if:
 ; CHECK-APPLE:       @ %bb.0: @ %entry
-; CHECK-APPLE-NEXT:    push {lr}
 ; CHECK-APPLE-NEXT:    cmp r0, #0
-; CHECK-APPLE-NEXT:    beq LBB3_2
-; CHECK-APPLE-NEXT:  @ %bb.1: @ %gen_error
+; CHECK-APPLE-NEXT:    vldreq s0, LCPI3_0
+; CHECK-APPLE-NEXT:    vmoveq r0, s0
+; CHECK-APPLE-NEXT:    bxeq lr
+; CHECK-APPLE-NEXT:  LBB3_1: @ %gen_error
+; CHECK-APPLE-NEXT:    push {lr}
 ; CHECK-APPLE-NEXT:    mov r0, #16
 ; CHECK-APPLE-NEXT:    mov r1, #0
 ; CHECK-APPLE-NEXT:    bl _malloc
 ; CHECK-APPLE-NEXT:    mov r8, r0
 ; CHECK-APPLE-NEXT:    mov r0, #1
 ; CHECK-APPLE-NEXT:    vmov.f32 s0, #1.000000e+00
 ; CHECK-APPLE-NEXT:    strb r0, [r8, #8]
-; CHECK-APPLE-NEXT:    b LBB3_3
-; CHECK-APPLE-NEXT:  LBB3_2:
-; CHECK-APPLE-NEXT:    vldr s0, LCPI3_0
-; CHECK-APPLE-NEXT:  LBB3_3: @ %common.ret
-; CHECK-APPLE-NEXT:    vmov r0, s0
 ; CHECK-APPLE-NEXT:    pop {lr}
+; CHECK-APPLE-NEXT:    vmov r0, s0
 ; CHECK-APPLE-NEXT:    bx lr
 ; CHECK-APPLE-NEXT:    .p2align 2
-; CHECK-APPLE-NEXT:  @ %bb.4:
+; CHECK-APPLE-NEXT:  @ %bb.2:
 ; CHECK-APPLE-NEXT:    .data_region
 ; CHECK-APPLE-NEXT:  LCPI3_0:
 ; CHECK-APPLE-NEXT:    .long 0x00000000 @ float 0
 ; CHECK-APPLE-NEXT:    .end_data_region
 ;
 ; CHECK-O0-LABEL: foo_if:
 ; CHECK-O0:       @ %bb.0: @ %entry
 ; CHECK-O0-NEXT:    push {r7, lr}

 ; CHECK-O0-NEXT:  LBB3_2: @ %normal
 ; CHECK-O0-NEXT:    ldr r8, [sp] @ 4-byte Reload
 ; CHECK-O0-NEXT:    mov r0, #0
 ; CHECK-O0-NEXT:    mov sp, r7
 ; CHECK-O0-NEXT:    pop {r7, pc}
 ;
 ; CHECK-ANDROID-LABEL: foo_if:
 ; CHECK-ANDROID:       @ %bb.0: @ %entry
+; CHECK-ANDROID-NEXT:    cmp r0, #0
+; CHECK-ANDROID-NEXT:    vldreq s0, .LCPI3_0
+; CHECK-ANDROID-NEXT:    vmoveq r0, s0
+; CHECK-ANDROID-NEXT:    bxeq lr
+; CHECK-ANDROID-NEXT:  .LBB3_1: @ %gen_error
 ; CHECK-ANDROID-NEXT:    .save {r11, lr}
 ; CHECK-ANDROID-NEXT:    push {r11, lr}
-; CHECK-ANDROID-NEXT:    cmp r0, #0
-; CHECK-ANDROID-NEXT:    beq .LBB3_2
-; CHECK-ANDROID-NEXT:  @ %bb.1: @ %gen_error
 ; CHECK-ANDROID-NEXT:    mov r0, #16
 ; CHECK-ANDROID-NEXT:    mov r1, #0
 ; CHECK-ANDROID-NEXT:    bl malloc
-; CHECK-ANDROID-NEXT:    vmov.f32 s0, #1.000000e+00
 ; CHECK-ANDROID-NEXT:    mov r8, r0
 ; CHECK-ANDROID-NEXT:    mov r0, #1
+; CHECK-ANDROID-NEXT:    vmov.f32 s0, #1.000000e+00
 ; CHECK-ANDROID-NEXT:    strb r0, [r8, #8]
+; CHECK-ANDROID-NEXT:    pop {r11, lr}
 ; CHECK-ANDROID-NEXT:    vmov r0, s0
-; CHECK-ANDROID-NEXT:    pop {r11, pc}
-; CHECK-ANDROID-NEXT:  .LBB3_2:
-; CHECK-ANDROID-NEXT:    vldr s0, .LCPI3_0
-; CHECK-ANDROID-NEXT:    vmov r0, s0
-; CHECK-ANDROID-NEXT:    pop {r11, pc}
+; CHECK-ANDROID-NEXT:    bx lr
 ; CHECK-ANDROID-NEXT:    .p2align 2
-; CHECK-ANDROID-NEXT:  @ %bb.3:
+; CHECK-ANDROID-NEXT:  @ %bb.2:
 ; CHECK-ANDROID-NEXT:  .LCPI3_0:
 ; CHECK-ANDROID-NEXT:    .long 0x00000000 @ float 0
 
 ; spill to stack
 ; reload from stack
 entry:
   %cond = icmp ne i32 %cc, 0
   br i1 %cond, label %gen_error, label %normal

llvm/test/CodeGen/PowerPC/BreakableToken-reduced.ll

 %"class.std::_Rb_tree.145" = type { %"struct.std::_Rb_tree<clang::tooling::Replacement, clang::tooling::Replacement, std::_Identity<clang::tooling::Replacement>, std::less<clang::tooling::Replacement>, std::allocator<clang::tooling::Replacement> >::_Rb_tree_impl" }
 %"struct.std::_Rb_tree<clang::tooling::Replacement, clang::tooling::Replacement, std::_Identity<clang::tooling::Replacement>, std::less<clang::tooling::Replacement>, std::allocator<clang::tooling::Replacement> >::_Rb_tree_impl" = type { %"struct.std::less.149", %"struct.std::_Rb_tree_node_base", i64 }
 %"struct.std::less.149" = type { i8 }
 
 
 define void @_ZN5clang6format22BreakableStringLiteral11insertBreakEjjSt4pairImjERNS0_17WhitespaceManagerE(ptr nocapture readonly %this, i32 zeroext %LineIndex, i32 zeroext %TailOffset, [2 x i64] %Split.coerce, ptr dereferenceable(1504) %Whitespaces) unnamed_addr #1 align 2 {
 ; CHECK-LABEL: _ZN5clang6format22BreakableStringLiteral11insertBreakEjjSt4pairImjERNS0_17WhitespaceManagerE:
 ; CHECK:       # %bb.0: # %entry
-; CHECK-NEXT:    mflr 0
-; CHECK-NEXT:    .cfi_def_cfa_offset 160
-; CHECK-NEXT:    .cfi_offset lr, 16
-; CHECK-NEXT:    .cfi_offset r28, -32
-; CHECK-NEXT:    .cfi_offset r29, -24
-; CHECK-NEXT:    .cfi_offset r30, -16
-; CHECK-NEXT:    std 28, -32(1) # 8-byte Folded Spill
-; CHECK-NEXT:    std 29, -24(1) # 8-byte Folded Spill
-; CHECK-NEXT:    std 30, -16(1) # 8-byte Folded Spill
-; CHECK-NEXT:    stdu 1, -160(1)
-; CHECK-NEXT:    std 0, 176(1)
-; CHECK-NEXT:    mr 12, 8
 ; CHECK-NEXT:    ld 10, 56(3)
 ; CHECK-NEXT:    lwz 0, 40(3)
+; CHECK-NEXT:    mr 12, 8
 ; CHECK-NEXT:    cmpldi 10, 0
 ; CHECK-NEXT:    beq 0, .LBB0_2
 ; CHECK-NEXT:  # %bb.1: # %if.end.i.i
 ; CHECK-NEXT:    ld 9, 48(3)
 ; CHECK-NEXT:    lbz 4, 0(9)
 ; CHECK-NEXT:    cmpwi 4, 64
 ; CHECK-NEXT:    b .LBB0_3
 ; CHECK-NEXT:  .LBB0_2: # %entry._ZNK4llvm9StringRef10startswithES0_.exit_crit_edge
 ; CHECK-NEXT:    ld 9, 48(3)
 ; CHECK-NEXT:    crxor 2, 2, 2
 ; CHECK-NEXT:  .LBB0_3: # %_ZNK4llvm9StringRef10startswithES0_.exit
+; CHECK-NEXT:    mflr 4
+; CHECK-NEXT:    .cfi_def_cfa_offset 160
+; CHECK-NEXT:    .cfi_offset lr, 16
+; CHECK-NEXT:    .cfi_offset r28, -32
+; CHECK-NEXT:    .cfi_offset r29, -24
+; CHECK-NEXT:    .cfi_offset r30, -16
+; CHECK-NEXT:    std 28, -32(1) # 8-byte Folded Spill
+; CHECK-NEXT:    std 29, -24(1) # 8-byte Folded Spill
+; CHECK-NEXT:    std 30, -16(1) # 8-byte Folded Spill
+; CHECK-NEXT:    stdu 1, -160(1)
+; CHECK-NEXT:    std 4, 176(1)
 ; CHECK-NEXT:    li 8, 0
 ; CHECK-NEXT:    li 11, 1
 ; CHECK-NEXT:    add 5, 6, 5
+; CHECK-NEXT:    iseleq 30, 11, 8
+; CHECK-NEXT:    ld 11, 64(3)
 ; CHECK-NEXT:    lbz 29, 20(3)
 ; CHECK-NEXT:    lwz 28, 16(3)
+; CHECK-NEXT:    add 5, 5, 10
 ; CHECK-NEXT:    ld 4, 8(3)
-; CHECK-NEXT:    iseleq 30, 11, 8
-; CHECK-NEXT:    ld 11, 64(3)
 ; CHECK-NEXT:    ld 8, 72(3)
-; CHECK-NEXT:    add 5, 5, 10
 ; CHECK-NEXT:    sub 3, 0, 30
 ; CHECK-NEXT:    clrldi 5, 5, 32
 ; CHECK-NEXT:    li 0, 1
 ; CHECK-NEXT:    clrldi 6, 7, 32
 ; CHECK-NEXT:    extsw 30, 3
 ; CHECK-NEXT:    mr 3, 12
 ; CHECK-NEXT:    mr 7, 11
-; CHECK-NEXT:    std 28, 112(1)
 ; CHECK-NEXT:    std 0, 104(1)
+; CHECK-NEXT:    std 28, 112(1)
 ; CHECK-NEXT:    std 29, 96(1)
 ; CHECK-NEXT:    std 30, 120(1)
 ; CHECK-NEXT:    bl _ZN5clang6format17WhitespaceManager24replaceWhitespaceInTokenERKNS0_11FormatTokenEjjN4llvm9StringRefES6_bjji
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    addi 1, 1, 160
 ; CHECK-NEXT:    ld 0, 16(1)
 ; CHECK-NEXT:    ld 30, -16(1) # 8-byte Folded Reload
 ; CHECK-NEXT:    ld 29, -24(1) # 8-byte Folded Reload

llvm/test/CodeGen/PowerPC/aix-lower-jump-table.ll

 ; RUN: llc -mtriple powerpc64-ibm-aix-xcoff -function-sections < %s | FileCheck \
 ; RUN: --check-prefix=FUNC-ASM %s
 
 define i32 @jump_table(i32 %a) {
 ; 32SMALL-ASM-LABEL: jump_table:
 ; 32SMALL-ASM:       # %bb.0: # %entry
 ; 32SMALL-ASM-NEXT:    addi 3, 3, -1
 ; 32SMALL-ASM-NEXT:    cmplwi 3, 3
-; 32SMALL-ASM-NEXT:    bgt 0, L..BB0_6
+; 32SMALL-ASM-NEXT:    bgt 0, L..BB0_3
 ; 32SMALL-ASM-NEXT:  # %bb.1: # %entry
 ; 32SMALL-ASM-NEXT:    lwz 4, L..C0(2) # %jump-table.0
 ; 32SMALL-ASM-NEXT:    slwi 3, 3, 2
 ; 32SMALL-ASM-NEXT:    lwzx 3, 3, 4
 ; 32SMALL-ASM-NEXT:    add 3, 3, 4
 ; 32SMALL-ASM-NEXT:    mtctr 3
 ; 32SMALL-ASM-NEXT:    bctr
 ; 32SMALL-ASM-NEXT:  L..BB0_2: # %sw.bb
-; 32SMALL-ASM-NEXT:    li 3, 0
 ; 32SMALL-ASM-NEXT:    #APP
 ; 32SMALL-ASM-NEXT:    #NO_APP
+; 32SMALL-ASM-NEXT:  L..BB0_3: # %sw.epilog
+; 32SMALL-ASM-NEXT:    li 3, 0
 ; 32SMALL-ASM-NEXT:    blr
-; 32SMALL-ASM-NEXT:  L..BB0_3: # %sw.bb1
+; 32SMALL-ASM-NEXT:  L..BB0_4: # %sw.bb1
 ; 32SMALL-ASM-NEXT:    li 3, 0
 ; 32SMALL-ASM-NEXT:    #APP
 ; 32SMALL-ASM-NEXT:    #NO_APP
 ; 32SMALL-ASM-NEXT:    blr
-; 32SMALL-ASM-NEXT:  L..BB0_4: # %sw.bb2
+; 32SMALL-ASM-NEXT:  L..BB0_5: # %sw.bb2
 ; 32SMALL-ASM-NEXT:    li 3, 0
 ; 32SMALL-ASM-NEXT:    #APP
 ; 32SMALL-ASM-NEXT:    #NO_APP
 ; 32SMALL-ASM-NEXT:    blr
-; 32SMALL-ASM-NEXT:  L..BB0_5: # %sw.bb3
+; 32SMALL-ASM-NEXT:  L..BB0_6: # %sw.bb3
+; 32SMALL-ASM-NEXT:    li 3, 0
 ; 32SMALL-ASM-NEXT:    #APP
 ; 32SMALL-ASM-NEXT:    #NO_APP
-; 32SMALL-ASM-NEXT:  L..BB0_6: # %sw.epilog
-; 32SMALL-ASM-NEXT:    li 3, 0
 ; 32SMALL-ASM-NEXT:    blr
 ;
 ; 32LARGE-ASM-LABEL: jump_table:
 ; 32LARGE-ASM:       # %bb.0: # %entry
 ; 32LARGE-ASM-NEXT:    addi 3, 3, -1
 ; 32LARGE-ASM-NEXT:    cmplwi 3, 3
-; 32LARGE-ASM-NEXT:    bgt 0, L..BB0_6
+; 32LARGE-ASM-NEXT:    bgt 0, L..BB0_3
 ; 32LARGE-ASM-NEXT:  # %bb.1: # %entry
 ; 32LARGE-ASM-NEXT:    addis 4, L..C0@u(2)
 ; 32LARGE-ASM-NEXT:    slwi 3, 3, 2
 ; 32LARGE-ASM-NEXT:    lwz 4, L..C0@l(4)
 ; 32LARGE-ASM-NEXT:    lwzx 3, 3, 4
 ; 32LARGE-ASM-NEXT:    add 3, 3, 4
 ; 32LARGE-ASM-NEXT:    mtctr 3
 ; 32LARGE-ASM-NEXT:    bctr
 ; 32LARGE-ASM-NEXT:  L..BB0_2: # %sw.bb
-; 32LARGE-ASM-NEXT:    li 3, 0
 ; 32LARGE-ASM-NEXT:    #APP
 ; 32LARGE-ASM-NEXT:    #NO_APP
+; 32LARGE-ASM-NEXT:  L..BB0_3: # %sw.epilog
+; 32LARGE-ASM-NEXT:    li 3, 0
 ; 32LARGE-ASM-NEXT:    blr
-; 32LARGE-ASM-NEXT:  L..BB0_3: # %sw.bb1
+; 32LARGE-ASM-NEXT:  L..BB0_4: # %sw.bb1
 ; 32LARGE-ASM-NEXT:    li 3, 0
 ; 32LARGE-ASM-NEXT:    #APP
 ; 32LARGE-ASM-NEXT:    #NO_APP
 ; 32LARGE-ASM-NEXT:    blr
-; 32LARGE-ASM-NEXT:  L..BB0_4: # %sw.bb2
+; 32LARGE-ASM-NEXT:  L..BB0_5: # %sw.bb2
 ; 32LARGE-ASM-NEXT:    li 3, 0
 ; 32LARGE-ASM-NEXT:    #APP
 ; 32LARGE-ASM-NEXT:    #NO_APP
 ; 32LARGE-ASM-NEXT:    blr
-; 32LARGE-ASM-NEXT:  L..BB0_5: # %sw.bb3
+; 32LARGE-ASM-NEXT:  L..BB0_6: # %sw.bb3
+; 32LARGE-ASM-NEXT:    li 3, 0
 ; 32LARGE-ASM-NEXT:    #APP
 ; 32LARGE-ASM-NEXT:    #NO_APP
-; 32LARGE-ASM-NEXT:  L..BB0_6: # %sw.epilog
-; 32LARGE-ASM-NEXT:    li 3, 0
 ; 32LARGE-ASM-NEXT:    blr
 ;
 ; 64SMALL-ASM-LABEL: jump_table:
 ; 64SMALL-ASM:       # %bb.0: # %entry
 ; 64SMALL-ASM-NEXT:    addi 3, 3, -1
 ; 64SMALL-ASM-NEXT:    cmplwi 3, 3
-; 64SMALL-ASM-NEXT:    bgt 0, L..BB0_6
+; 64SMALL-ASM-NEXT:    bgt 0, L..BB0_3
 ; 64SMALL-ASM-NEXT:  # %bb.1: # %entry
 ; 64SMALL-ASM-NEXT:    ld 4, L..C0(2) # %jump-table.0
 ; 64SMALL-ASM-NEXT:    rldic 3, 3, 2, 30
 ; 64SMALL-ASM-NEXT:    lwax 3, 3, 4
 ; 64SMALL-ASM-NEXT:    add 3, 3, 4
 ; 64SMALL-ASM-NEXT:    mtctr 3
 ; 64SMALL-ASM-NEXT:    bctr
 ; 64SMALL-ASM-NEXT:  L..BB0_2: # %sw.bb
-; 64SMALL-ASM-NEXT:    li 3, 0
 ; 64SMALL-ASM-NEXT:    #APP
 ; 64SMALL-ASM-NEXT:    #NO_APP
+; 64SMALL-ASM-NEXT:  L..BB0_3: # %sw.epilog
+; 64SMALL-ASM-NEXT:    li 3, 0
 ; 64SMALL-ASM-NEXT:    blr
-; 64SMALL-ASM-NEXT:  L..BB0_3: # %sw.bb1
+; 64SMALL-ASM-NEXT:  L..BB0_4: # %sw.bb1
 ; 64SMALL-ASM-NEXT:    li 3, 0
 ; 64SMALL-ASM-NEXT:    #APP
 ; 64SMALL-ASM-NEXT:    #NO_APP
 ; 64SMALL-ASM-NEXT:    blr
-; 64SMALL-ASM-NEXT:  L..BB0_4: # %sw.bb2
+; 64SMALL-ASM-NEXT:  L..BB0_5: # %sw.bb2
 ; 64SMALL-ASM-NEXT:    li 3, 0
 ; 64SMALL-ASM-NEXT:    #APP
 ; 64SMALL-ASM-NEXT:    #NO_APP
 ; 64SMALL-ASM-NEXT:    blr
-; 64SMALL-ASM-NEXT:  L..BB0_5: # %sw.bb3
+; 64SMALL-ASM-NEXT:  L..BB0_6: # %sw.bb3
+; 64SMALL-ASM-NEXT:    li 3, 0
 ; 64SMALL-ASM-NEXT:    #APP
 ; 64SMALL-ASM-NEXT:    #NO_APP
-; 64SMALL-ASM-NEXT:  L..BB0_6: # %sw.epilog
-; 64SMALL-ASM-NEXT:    li 3, 0
 ; 64SMALL-ASM-NEXT:    blr
 ;
 ; 64LARGE-ASM-LABEL: jump_table:
 ; 64LARGE-ASM:       # %bb.0: # %entry
 ; 64LARGE-ASM-NEXT:    addi 3, 3, -1
 ; 64LARGE-ASM-NEXT:    cmplwi 3, 3
-; 64LARGE-ASM-NEXT:    bgt 0, L..BB0_6
+; 64LARGE-ASM-NEXT:    bgt 0, L..BB0_3
 ; 64LARGE-ASM-NEXT:  # %bb.1: # %entry
 ; 64LARGE-ASM-NEXT:    addis 4, L..C0@u(2)
 ; 64LARGE-ASM-NEXT:    rldic 3, 3, 2, 30
 ; 64LARGE-ASM-NEXT:    ld 4, L..C0@l(4)
 ; 64LARGE-ASM-NEXT:    lwax 3, 3, 4
 ; 64LARGE-ASM-NEXT:    add 3, 3, 4
 ; 64LARGE-ASM-NEXT:    mtctr 3
 ; 64LARGE-ASM-NEXT:    bctr
 ; 64LARGE-ASM-NEXT:  L..BB0_2: # %sw.bb
-; 64LARGE-ASM-NEXT:    li 3, 0
 ; 64LARGE-ASM-NEXT:    #APP
 ; 64LARGE-ASM-NEXT:    #NO_APP
+; 64LARGE-ASM-NEXT:  L..BB0_3: # %sw.epilog
+; 64LARGE-ASM-NEXT:    li 3, 0
 ; 64LARGE-ASM-NEXT:    blr
-; 64LARGE-ASM-NEXT:  L..BB0_3: # %sw.bb1
+; 64LARGE-ASM-NEXT:  L..BB0_4: # %sw.bb1
 ; 64LARGE-ASM-NEXT:    li 3, 0
 ; 64LARGE-ASM-NEXT:    #APP
 ; 64LARGE-ASM-NEXT:    #NO_APP
 ; 64LARGE-ASM-NEXT:    blr
-; 64LARGE-ASM-NEXT:  L..BB0_4: # %sw.bb2
+; 64LARGE-ASM-NEXT:  L..BB0_5: # %sw.bb2
 ; 64LARGE-ASM-NEXT:    li 3, 0
 ; 64LARGE-ASM-NEXT:    #APP
 ; 64LARGE-ASM-NEXT:    #NO_APP
 ; 64LARGE-ASM-NEXT:    blr
-; 64LARGE-ASM-NEXT:  L..BB0_5: # %sw.bb3
+; 64LARGE-ASM-NEXT:  L..BB0_6: # %sw.bb3
+; 64LARGE-ASM-NEXT:    li 3, 0
 ; 64LARGE-ASM-NEXT:    #APP
 ; 64LARGE-ASM-NEXT:    #NO_APP
-; 64LARGE-ASM-NEXT:  L..BB0_6: # %sw.epilog
-; 64LARGE-ASM-NEXT:    li 3, 0
 ; 64LARGE-ASM-NEXT:    blr
 entry:
   switch i32 %a, label %sw.epilog [
     i32 1, label %sw.bb
     i32 2, label %sw.bb1
     i32 3, label %sw.bb2
     i32 4, label %sw.bb3
   ]

llvm/test/CodeGen/PowerPC/licm-tocReg.ll

 ;    BCC 44, killed %19, %bb.2.if.then
 ;    B %bb.3.if.end
 
 @ga = external global i32, align 4
 @gb = external global i32, align 4
 define signext i32 @test(ptr nocapture %FP) local_unnamed_addr #0 {
 ; CHECKLX-LABEL: test:
 ; CHECKLX:       # %bb.0: # %entry
-; CHECKLX-NEXT:    mflr 0
-; CHECKLX-NEXT:    stdu 1, -32(1)
-; CHECKLX-NEXT:    std 2, 24(1)
-; CHECKLX-NEXT:    std 0, 48(1)
-; CHECKLX-NEXT:    .cfi_def_cfa_offset 32
-; CHECKLX-NEXT:    .cfi_offset lr, 16
 ; CHECKLX-NEXT:    addis 4, 2, .LC0@toc@ha
 ; CHECKLX-NEXT:    addis 5, 2, .LC1@toc@ha
 ; CHECKLX-NEXT:    mr 12, 3
 ; CHECKLX-NEXT:    ld 4, .LC0@toc@l(4)
 ; CHECKLX-NEXT:    ld 5, .LC1@toc@l(5)
 ; CHECKLX-NEXT:    lwz 6, 0(4)
 ; CHECKLX-NEXT:    lwz 7, 0(5)
 ; CHECKLX-NEXT:    cmpw 6, 7
 ; CHECKLX-NEXT:    lwz 6, 0(4)
 ; CHECKLX-NEXT:    bgt 0, .LBB0_2
 ; CHECKLX-NEXT:    .p2align 5
 ; CHECKLX-NEXT:  .LBB0_1: # %if.end
 ; CHECKLX-NEXT:    #
 ; CHECKLX-NEXT:    addi 3, 6, 1
 ; CHECKLX-NEXT:    stw 3, 0(4)
 ; CHECKLX-NEXT:    lwz 3, 0(4)
 ; CHECKLX-NEXT:    lwz 6, 0(5)
 ; CHECKLX-NEXT:    cmpw 3, 6
 ; CHECKLX-NEXT:    lwz 6, 0(4)
 ; CHECKLX-NEXT:    ble 0, .LBB0_1
 ; CHECKLX-NEXT:  .LBB0_2: # %if.then
+; CHECKLX-NEXT:    mflr 0
+; CHECKLX-NEXT:    stdu 1, -32(1)
+; CHECKLX-NEXT:    std 2, 24(1)
+; CHECKLX-NEXT:    std 0, 48(1)
+; CHECKLX-NEXT:    .cfi_def_cfa_offset 32
+; CHECKLX-NEXT:    .cfi_offset lr, 16
 ; CHECKLX-NEXT:    extsw 3, 6
 ; CHECKLX-NEXT:    mtctr 12
 ; CHECKLX-NEXT:    bctrl
 ; CHECKLX-NEXT:    ld 2, 24(1)
 ; CHECKLX-NEXT:    addi 1, 1, 32
 ; CHECKLX-NEXT:    ld 0, 16(1)
 ; CHECKLX-NEXT:    mtlr 0
 ; CHECKLX-NEXT:    blr
 ;
 ; CHECKAIX-LABEL: test:
 ; CHECKAIX:       # %bb.0: # %entry
-; CHECKAIX-NEXT:    mflr 0
-; CHECKAIX-NEXT:    stdu 1, -112(1)
-; CHECKAIX-NEXT:    std 0, 128(1)
 ; CHECKAIX-NEXT:    ld 5, L..C0(2) # @ga
 ; CHECKAIX-NEXT:    ld 6, L..C1(2) # @gb
 ; CHECKAIX-NEXT:  L..BB0_1: # %if.end
 ; CHECKAIX-NEXT:    #
 ; CHECKAIX-NEXT:    lwz 4, 0(5)
 ; CHECKAIX-NEXT:    lwz 7, 0(6)
 ; CHECKAIX-NEXT:    cmpw 4, 7
 ; CHECKAIX-NEXT:    lwz 4, 0(5)
 ; CHECKAIX-NEXT:    bgt 0, L..BB0_3
 ; CHECKAIX-NEXT:  # %bb.2: # %if.end
 ; CHECKAIX-NEXT:    #
 ; CHECKAIX-NEXT:    addi 4, 4, 1
 ; CHECKAIX-NEXT:    stw 4, 0(5)
 ; CHECKAIX-NEXT:    b L..BB0_1
 ; CHECKAIX-NEXT:  L..BB0_3: # %if.then
+; CHECKAIX-NEXT:    mflr 0
+; CHECKAIX-NEXT:    stdu 1, -112(1)
 ; CHECKAIX-NEXT:    ld 5, 0(3)
+; CHECKAIX-NEXT:    std 0, 128(1)
 ; CHECKAIX-NEXT:    ld 11, 16(3)
 ; CHECKAIX-NEXT:    std 2, 40(1)
 ; CHECKAIX-NEXT:    ld 2, 8(3)
 ; CHECKAIX-NEXT:    extsw 3, 4
 ; CHECKAIX-NEXT:    mtctr 5
 ; CHECKAIX-NEXT:    bctrl
 ; CHECKAIX-NEXT:    ld 2, 40(1)
 ; CHECKAIX-NEXT:    addi 1, 1, 112
 ; CHECKAIX-NEXT:    ld 0, 16(1)
 ; CHECKAIX-NEXT:    mtlr 0
 ; CHECKAIX-NEXT:    blr
 ;
 ; CHECKAIX32-LABEL: test:
 ; CHECKAIX32:       # %bb.0: # %entry
-; CHECKAIX32-NEXT:    mflr 0
-; CHECKAIX32-NEXT:    stwu 1, -64(1)
-; CHECKAIX32-NEXT:    stw 0, 72(1)
 ; CHECKAIX32-NEXT:    lwz 5, L..C0(2) # @ga
 ; CHECKAIX32-NEXT:    lwz 6, L..C1(2) # @gb
 ; CHECKAIX32-NEXT:  L..BB0_1: # %if.end
 ; CHECKAIX32-NEXT:    #
 ; CHECKAIX32-NEXT:    lwz 4, 0(5)
 ; CHECKAIX32-NEXT:    lwz 7, 0(6)
 ; CHECKAIX32-NEXT:    cmpw 4, 7
 ; CHECKAIX32-NEXT:    lwz 4, 0(5)
 ; CHECKAIX32-NEXT:    bgt 0, L..BB0_3
 ; CHECKAIX32-NEXT:  # %bb.2: # %if.end
 ; CHECKAIX32-NEXT:    #
 ; CHECKAIX32-NEXT:    addi 4, 4, 1
 ; CHECKAIX32-NEXT:    stw 4, 0(5)
 ; CHECKAIX32-NEXT:    b L..BB0_1
 ; CHECKAIX32-NEXT:  L..BB0_3: # %if.then
+; CHECKAIX32-NEXT:    mflr 0
+; CHECKAIX32-NEXT:    stwu 1, -64(1)
 ; CHECKAIX32-NEXT:    lwz 5, 0(3)
+; CHECKAIX32-NEXT:    stw 0, 72(1)
 ; CHECKAIX32-NEXT:    stw 2, 20(1)
-; CHECKAIX32-NEXT:    lwz 11, 8(3)
 ; CHECKAIX32-NEXT:    mtctr 5
+; CHECKAIX32-NEXT:    lwz 11, 8(3)
 ; CHECKAIX32-NEXT:    lwz 2, 4(3)
 ; CHECKAIX32-NEXT:    mr 3, 4
 ; CHECKAIX32-NEXT:    bctrl
 ; CHECKAIX32-NEXT:    lwz 2, 20(1)
 ; CHECKAIX32-NEXT:    addi 1, 1, 64
 ; CHECKAIX32-NEXT:    lwz 0, 8(1)
 ; CHECKAIX32-NEXT:    mtlr 0
 ; CHECKAIX32-NEXT:    blr

llvm/test/CodeGen/PowerPC/p10-spill-creq.ll

 %1 = type { ptr, ptr, i32 }
 %2 = type { [200 x i8], [200 x i8], ptr, ptr, ptr, ptr, ptr, ptr, ptr, i64 }
 %3 = type { i64, i32, ptr, ptr, ptr, ptr, ptr, ptr, ptr, ptr, i64, i32, i32 }
 %4 = type { i32, i64, ptr, ptr, i16, ptr, ptr, i64, i64 }
 
 define dso_local double @P10_Spill_CR_EQ(ptr %arg) local_unnamed_addr #0 {
 ; CHECK-LABEL: P10_Spill_CR_EQ:
 ; CHECK:       # %bb.0: # %bb
-; CHECK-NEXT:    mfcr r12
-; CHECK-NEXT:    stw r12, 8(r1)
 ; CHECK-NEXT:    ld r3, 0(r3)
 ; CHECK-NEXT:    ld r4, 0(0)
-; CHECK-NEXT:    ld r5, 56(0)
 ; CHECK-NEXT:    cmpdi r3, 0
+; CHECK-NEXT:    ld r5, 56(0)
 ; CHECK-NEXT:    cmpdi cr1, r4, 0
 ; CHECK-NEXT:    cmpdi cr5, r5, 0
 ; CHECK-NEXT:    cmpldi cr6, r3, 0
 ; CHECK-NEXT:    beq cr6, .LBB0_3
 ; CHECK-NEXT:  # %bb.1: # %bb10
 ; CHECK-NEXT:    lwz r3, 0(r3)
 ; CHECK-NEXT:    bc 12, 4*cr1+eq, .LBB0_4
 ; CHECK-NEXT:  .LBB0_2: # %bb14

 ; CHECK-NEXT:    crnot 4*cr5+un, 4*cr5+eq
 ; CHECK-NEXT:    bc 12, 4*cr5+eq, .LBB0_7
 ; CHECK-NEXT:  # %bb.6: # %bb18
 ; CHECK-NEXT:    lwz r4, 0(r3)
 ; CHECK-NEXT:    b .LBB0_8
 ; CHECK-NEXT:  .LBB0_7:
 ; CHECK-NEXT:    # implicit-def: $r4
 ; CHECK-NEXT:  .LBB0_8: # %bb20
+; CHECK-NEXT:    mfcr r12
 ; CHECK-NEXT:    cmpwi cr2, r3, -1
 ; CHECK-NEXT:    cmpwi cr3, r4, -1
+; CHECK-NEXT:    stw r12, 8(r1)
 ; CHECK-NEXT:    cmpwi cr7, r3, 0
 ; CHECK-NEXT:    cmpwi cr6, r4, 0
-; CHECK-NEXT:    # implicit-def: $x3
 ; CHECK-NEXT:    crand 4*cr5+gt, 4*cr2+gt, 4*cr1+lt
 ; CHECK-NEXT:    crand 4*cr5+lt, 4*cr3+gt, 4*cr5+un
+; CHECK-NEXT:    # implicit-def: $x3
 ; CHECK-NEXT:    bc 4, 4*cr5+gt, .LBB0_10
 ; CHECK-NEXT:  # %bb.9: # %bb34
 ; CHECK-NEXT:    ld r3, 0(r3)
 ; CHECK-NEXT:  .LBB0_10: # %bb36
 ; CHECK-NEXT:    cmpwi cr2, r5, 0
 ; CHECK-NEXT:    # implicit-def: $x4
 ; CHECK-NEXT:    bc 4, 4*cr5+lt, .LBB0_12
 ; CHECK-NEXT:  # %bb.11: # %bb38

llvm/test/CodeGen/PowerPC/sms-cpy-1.ll

 ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
 ; RUN: llc < %s -mtriple=powerpc64le-unknown-linux-gnu -verify-machineinstrs\
 ; RUN:       -mcpu=pwr9 --ppc-enable-pipeliner 2>&1 | FileCheck %s
 
 @.str.28 = external unnamed_addr constant [69 x i8], align 1
 
 define void @print_res() nounwind {
 ; CHECK-LABEL: print_res:
 ; CHECK:       # %bb.0:
-; CHECK-NEXT:    mflr 0
-; CHECK-NEXT:    stdu 1, -128(1)
-; CHECK-NEXT:    std 0, 144(1)
 ; CHECK-NEXT:    lwz 3, 0(3)
+; CHECK-NEXT:    mflr 0
 ; CHECK-NEXT:    addi 3, 3, -1
 ; CHECK-NEXT:    clrldi 4, 3, 32
 ; CHECK-NEXT:    cmplwi 3, 3
 ; CHECK-NEXT:    li 3, 3
 ; CHECK-NEXT:    isellt 3, 4, 3
 ; CHECK-NEXT:    li 4, 1
 ; CHECK-NEXT:    cmpldi 3, 1
 ; CHECK-NEXT:    iselgt 3, 3, 4
 ; CHECK-NEXT:    li 4, 0
-; CHECK-NEXT:    li 5, 0
 ; CHECK-NEXT:    mtctr 3
+; CHECK-NEXT:    stdu 1, -128(1)
+; CHECK-NEXT:    li 5, 0
+; CHECK-NEXT:    std 0, 144(1)
+; CHECK-NEXT:    li 3, 1
 ; CHECK-NEXT:    li 7, -1
 ; CHECK-NEXT:    lbz 5, 0(5)
-; CHECK-NEXT:    li 3, 1
 ; CHECK-NEXT:    bdz .LBB0_6
 ; CHECK-NEXT:  # %bb.1:
 ; CHECK-NEXT:    xori 6, 5, 84
 ; CHECK-NEXT:    clrldi 5, 7, 32
 ; CHECK-NEXT:    addi 3, 3, 1
 ; CHECK-NEXT:    addi 8, 7, -1
 ; CHECK-NEXT:    lbz 5, 0(5)
 ; CHECK-NEXT:    bdz .LBB0_5

llvm/test/CodeGen/RISCV/half-round-conv-sat.ll

; CHECKIZHINXMIN-NEXT:    ret
   %a = call half @llvm.floor.f16(half %x)
   %b = call i32 @llvm.fptosi.sat.i32.f16(half %a)
   ret i32 %b
 }
 
 define i64 @test_floor_si64(half %x) nounwind {
 ; RV32IZFH-LABEL: test_floor_si64:
 ; RV32IZFH:       # %bb.0:
-; RV32IZFH-NEXT:    addi sp, sp, -16
-; RV32IZFH-NEXT:    sw ra, 12(sp) # 4-byte Folded Spill
-; RV32IZFH-NEXT:    sw s0, 8(sp) # 4-byte Folded Spill
-; RV32IZFH-NEXT:    fsw fs0, 4(sp) # 4-byte Folded Spill
 ; RV32IZFH-NEXT:    lui a0, %hi(.LCPI1_0)
 ; RV32IZFH-NEXT:    flh fa5, %lo(.LCPI1_0)(a0)
 ; RV32IZFH-NEXT:    fabs.h fa4, fa0
 ; RV32IZFH-NEXT:    flt.h a0, fa4, fa5
 ; RV32IZFH-NEXT:    beqz a0, .LBB1_2
 ; RV32IZFH-NEXT:  # %bb.1:
 ; RV32IZFH-NEXT:    fcvt.w.h a0, fa0, rdn
 ; RV32IZFH-NEXT:    fcvt.h.w fa5, a0, rdn
 ; RV32IZFH-NEXT:    fsgnj.h fa0, fa5, fa0
 ; RV32IZFH-NEXT:  .LBB1_2:
+; RV32IZFH-NEXT:    addi sp, sp, -16
+; RV32IZFH-NEXT:    sw ra, 12(sp) # 4-byte Folded Spill
+; RV32IZFH-NEXT:    sw s0, 8(sp) # 4-byte Folded Spill
+; RV32IZFH-NEXT:    fsw fs0, 4(sp) # 4-byte Folded Spill
 ; RV32IZFH-NEXT:    fcvt.s.h fs0, fa0
 ; RV32IZFH-NEXT:    lui a0, 913408
 ; RV32IZFH-NEXT:    fmv.w.x fa5, a0
 ; RV32IZFH-NEXT:    fle.s s0, fa5, fs0
 ; RV32IZFH-NEXT:    fmv.s fa0, fs0
 ; RV32IZFH-NEXT:    call __fixsfdi@plt
 ; RV32IZFH-NEXT:    lui a4, 524288
 ; RV32IZFH-NEXT:    lui a2, 524288

 ; RV64IZFH-NEXT:    feq.h a1, fa0, fa0
 ; RV64IZFH-NEXT:    seqz a1, a1
 ; RV64IZFH-NEXT:    addi a1, a1, -1
 ; RV64IZFH-NEXT:    and a0, a1, a0
 ; RV64IZFH-NEXT:    ret
 ;
 ; RV32IZHINX-LABEL: test_floor_si64:
 ; RV32IZHINX:       # %bb.0:
-; RV32IZHINX-NEXT:    addi sp, sp, -16
-; RV32IZHINX-NEXT:    sw ra, 12(sp) # 4-byte Folded Spill
-; RV32IZHINX-NEXT:    sw s0, 8(sp) # 4-byte Folded Spill
-; RV32IZHINX-NEXT:    sw s1, 4(sp) # 4-byte Folded Spill
-; RV32IZHINX-NEXT:    sw s2, 0(sp) # 4-byte Folded Spill
 ; RV32IZHINX-NEXT:    lui a1, %hi(.LCPI1_0)
 ; RV32IZHINX-NEXT:    lh a1, %lo(.LCPI1_0)(a1)
 ; RV32IZHINX-NEXT:    fabs.h a2, a0
 ; RV32IZHINX-NEXT:    flt.h a1, a2, a1
 ; RV32IZHINX-NEXT:    beqz a1, .LBB1_2
 ; RV32IZHINX-NEXT:  # %bb.1:
 ; RV32IZHINX-NEXT:    fcvt.w.h a1, a0, rdn
 ; RV32IZHINX-NEXT:    fcvt.h.w a1, a1, rdn
 ; RV32IZHINX-NEXT:    fsgnj.h a0, a1, a0
 ; RV32IZHINX-NEXT:  .LBB1_2:
+; RV32IZHINX-NEXT:    addi sp, sp, -16
+; RV32IZHINX-NEXT:    sw ra, 12(sp) # 4-byte Folded Spill
+; RV32IZHINX-NEXT:    sw s0, 8(sp) # 4-byte Folded Spill
+; RV32IZHINX-NEXT:    sw s1, 4(sp) # 4-byte Folded Spill
+; RV32IZHINX-NEXT:    sw s2, 0(sp) # 4-byte Folded Spill
 ; RV32IZHINX-NEXT:    fcvt.s.h s0, a0
 ; RV32IZHINX-NEXT:    lui a0, 913408
 ; RV32IZHINX-NEXT:    fle.s s1, a0, s0
 ; RV32IZHINX-NEXT:    neg s2, s1
 ; RV32IZHINX-NEXT:    mv a0, s0
 ; RV32IZHINX-NEXT:    call __fixsfdi@plt
 ; RV32IZHINX-NEXT:    lui a2, %hi(.LCPI1_1)
 ; RV32IZHINX-NEXT:    lw a2, %lo(.LCPI1_1)(a2)
 ; RV32IZHINX-NEXT:    and a0, s2, a0
 ; RV32IZHINX-NEXT:    flt.s a4, a2, s0
 ; RV32IZHINX-NEXT:    neg a2, a4
 ; RV32IZHINX-NEXT:    or a0, a2, a0
 ; RV32IZHINX-NEXT:    feq.s a2, s0, s0
 ; RV32IZHINX-NEXT:    neg a2, a2
 ; RV32IZHINX-NEXT:    lui a5, 524288
 ; RV32IZHINX-NEXT:    lui a3, 524288
 ; RV32IZHINX-NEXT:    beqz s1, .LBB1_4
 ; RV32IZHINX-NEXT:  # %bb.3:
 ; RV32IZHINX-NEXT:    mv a3, a1
 ; RV32IZHINX-NEXT:  .LBB1_4:
 ; RV32IZHINX-NEXT:    and a0, a2, a0
-; RV32IZHINX-NEXT:    beqz a4, .LBB1_6
-; RV32IZHINX-NEXT:  # %bb.5:
-; RV32IZHINX-NEXT:    addi a3, a5, -1
-; RV32IZHINX-NEXT:  .LBB1_6:
-; RV32IZHINX-NEXT:    and a1, a2, a3
 ; RV32IZHINX-NEXT:    lw ra, 12(sp) # 4-byte Folded Reload
 ; RV32IZHINX-NEXT:    lw s0, 8(sp) # 4-byte Folded Reload
 ; RV32IZHINX-NEXT:    lw s1, 4(sp) # 4-byte Folded Reload
 ; RV32IZHINX-NEXT:    lw s2, 0(sp) # 4-byte Folded Reload
 ; RV32IZHINX-NEXT:    addi sp, sp, 16
+; RV32IZHINX-NEXT:    beqz a4, .LBB1_6
+; RV32IZHINX-NEXT:  # %bb.5:
+; RV32IZHINX-NEXT:    addi a3, a5, -1
+; RV32IZHINX-NEXT:  .LBB1_6:
+; RV32IZHINX-NEXT:    and a1, a2, a3
 ; RV32IZHINX-NEXT:    ret
 ;
 ; RV64IZHINX-LABEL: test_floor_si64:
 ; RV64IZHINX:       # %bb.0:
 ; RV64IZHINX-NEXT:    lui a1, %hi(.LCPI1_0)
 ; RV64IZHINX-NEXT:    lh a1, %lo(.LCPI1_0)(a1)
 ; RV64IZHINX-NEXT:    fabs.h a2, a0
 ; RV64IZHINX-NEXT:    flt.h a1, a2, a1
; RV64IZHINXMIN-NEXT:    ret
   %a = call half @llvm.floor.f16(half %x)
   %b = call i32 @llvm.fptoui.sat.i32.f16(half %a)
   ret i32 %b
 }
 
 define i64 @test_floor_ui64(half %x) nounwind {
 ; RV32IZFH-LABEL: test_floor_ui64:
 ; RV32IZFH:       # %bb.0:
-; RV32IZFH-NEXT:    addi sp, sp, -16
-; RV32IZFH-NEXT:    sw ra, 12(sp) # 4-byte Folded Spill
-; RV32IZFH-NEXT:    sw s0, 8(sp) # 4-byte Folded Spill
-; RV32IZFH-NEXT:    fsw fs0, 4(sp) # 4-byte Folded Spill
 ; RV32IZFH-NEXT:    lui a0, %hi(.LCPI3_0)
 ; RV32IZFH-NEXT:    flh fa5, %lo(.LCPI3_0)(a0)
 ; RV32IZFH-NEXT:    fabs.h fa4, fa0
 ; RV32IZFH-NEXT:    flt.h a0, fa4, fa5
 ; RV32IZFH-NEXT:    beqz a0, .LBB3_2
 ; RV32IZFH-NEXT:  # %bb.1:
 ; RV32IZFH-NEXT:    fcvt.w.h a0, fa0, rdn
 ; RV32IZFH-NEXT:    fcvt.h.w fa5, a0, rdn
 ; RV32IZFH-NEXT:    fsgnj.h fa0, fa5, fa0
 ; RV32IZFH-NEXT:  .LBB3_2:
+; RV32IZFH-NEXT:    addi sp, sp, -16
+; RV32IZFH-NEXT:    sw ra, 12(sp) # 4-byte Folded Spill
+; RV32IZFH-NEXT:    sw s0, 8(sp) # 4-byte Folded Spill
+; RV32IZFH-NEXT:    fsw fs0, 4(sp) # 4-byte Folded Spill
 ; RV32IZFH-NEXT:    fcvt.s.h fs0, fa0
 ; RV32IZFH-NEXT:    fmv.w.x fa5, zero
 ; RV32IZFH-NEXT:    fle.s a0, fa5, fs0
 ; RV32IZFH-NEXT:    neg s0, a0
 ; RV32IZFH-NEXT:    fmv.s fa0, fs0
 ; RV32IZFH-NEXT:    call __fixunssfdi@plt
 ; RV32IZFH-NEXT:    lui a2, %hi(.LCPI3_1)
 ; RV32IZFH-NEXT:    flw fa5, %lo(.LCPI3_1)(a2)

 ; RV64IZFH-NEXT:    feq.h a1, fa0, fa0
 ; RV64IZFH-NEXT:    seqz a1, a1
 ; RV64IZFH-NEXT:    addi a1, a1, -1
 ; RV64IZFH-NEXT:    and a0, a1, a0
 ; RV64IZFH-NEXT:    ret
 ;
 ; RV32IZHINX-LABEL: test_floor_ui64:
 ; RV32IZHINX:       # %bb.0:
-; RV32IZHINX-NEXT:    addi sp, sp, -16
-; RV32IZHINX-NEXT:    sw ra, 12(sp) # 4-byte Folded Spill
-; RV32IZHINX-NEXT:    sw s0, 8(sp) # 4-byte Folded Spill
-; RV32IZHINX-NEXT:    sw s1, 4(sp) # 4-byte Folded Spill
 ; RV32IZHINX-NEXT:    lui a1, %hi(.LCPI3_0)
 ; RV32IZHINX-NEXT:    lh a1, %lo(.LCPI3_0)(a1)
 ; RV32IZHINX-NEXT:    fabs.h a2, a0
 ; RV32IZHINX-NEXT:    flt.h a1, a2, a1
 ; RV32IZHINX-NEXT:    beqz a1, .LBB3_2
 ; RV32IZHINX-NEXT:  # %bb.1:
 ; RV32IZHINX-NEXT:    fcvt.w.h a1, a0, rdn
 ; RV32IZHINX-NEXT:    fcvt.h.w a1, a1, rdn
 ; RV32IZHINX-NEXT:    fsgnj.h a0, a1, a0
 ; RV32IZHINX-NEXT:  .LBB3_2:
+; RV32IZHINX-NEXT:    addi sp, sp, -16
+; RV32IZHINX-NEXT:    sw ra, 12(sp) # 4-byte Folded Spill
+; RV32IZHINX-NEXT:    sw s0, 8(sp) # 4-byte Folded Spill
+; RV32IZHINX-NEXT:    sw s1, 4(sp) # 4-byte Folded Spill
 ; RV32IZHINX-NEXT:    fcvt.s.h s0, a0
 ; RV32IZHINX-NEXT:    fle.s a0, zero, s0
 ; RV32IZHINX-NEXT:    neg s1, a0
 ; RV32IZHINX-NEXT:    mv a0, s0
 ; RV32IZHINX-NEXT:    call __fixunssfdi@plt
 ; RV32IZHINX-NEXT:    lui a2, %hi(.LCPI3_1)
 ; RV32IZHINX-NEXT:    lw a2, %lo(.LCPI3_1)(a2)
 ; RV32IZHINX-NEXT:    and a0, s1, a0
; CHECKIZHINXMIN-NEXT:    ret
   %a = call half @llvm.ceil.f16(half %x)
   %b = call i32 @llvm.fptosi.sat.i32.f16(half %a)
   ret i32 %b
 }
 
 define i64 @test_ceil_si64(half %x) nounwind {
 ; RV32IZFH-LABEL: test_ceil_si64:
 ; RV32IZFH:       # %bb.0:
-; RV32IZFH-NEXT:    addi sp, sp, -16
-; RV32IZFH-NEXT:    sw ra, 12(sp) # 4-byte Folded Spill
-; RV32IZFH-NEXT:    sw s0, 8(sp) # 4-byte Folded Spill
-; RV32IZFH-NEXT:    fsw fs0, 4(sp) # 4-byte Folded Spill
 ; RV32IZFH-NEXT:    lui a0, %hi(.LCPI5_0)
 ; RV32IZFH-NEXT:    flh fa5, %lo(.LCPI5_0)(a0)
 ; RV32IZFH-NEXT:    fabs.h fa4, fa0
 ; RV32IZFH-NEXT:    flt.h a0, fa4, fa5
 ; RV32IZFH-NEXT:    beqz a0, .LBB5_2
 ; RV32IZFH-NEXT:  # %bb.1:
 ; RV32IZFH-NEXT:    fcvt.w.h a0, fa0, rup
 ; RV32IZFH-NEXT:    fcvt.h.w fa5, a0, rup
 ; RV32IZFH-NEXT:    fsgnj.h fa0, fa5, fa0
 ; RV32IZFH-NEXT:  .LBB5_2:
+; RV32IZFH-NEXT:    addi sp, sp, -16
+; RV32IZFH-NEXT:    sw ra, 12(sp) # 4-byte Folded Spill
+; RV32IZFH-NEXT:    sw s0, 8(sp) # 4-byte Folded Spill
+; RV32IZFH-NEXT:    fsw fs0, 4(sp) # 4-byte Folded Spill
 ; RV32IZFH-NEXT:    fcvt.s.h fs0, fa0
 ; RV32IZFH-NEXT:    lui a0, 913408
 ; RV32IZFH-NEXT:    fmv.w.x fa5, a0
 ; RV32IZFH-NEXT:    fle.s s0, fa5, fs0
 ; RV32IZFH-NEXT:    fmv.s fa0, fs0
 ; RV32IZFH-NEXT:    call __fixsfdi@plt
 ; RV32IZFH-NEXT:    lui a4, 524288
 ; RV32IZFH-NEXT:    lui a2, 524288

 ; RV64IZFH-NEXT:    feq.h a1, fa0, fa0
 ; RV64IZFH-NEXT:    seqz a1, a1
 ; RV64IZFH-NEXT:    addi a1, a1, -1
 ; RV64IZFH-NEXT:    and a0, a1, a0
 ; RV64IZFH-NEXT:    ret
 ;
 ; RV32IZHINX-LABEL: test_ceil_si64:
 ; RV32IZHINX:       # %bb.0:
-; RV32IZHINX-NEXT:    addi sp, sp, -16
-; RV32IZHINX-NEXT:    sw ra, 12(sp) # 4-byte Folded Spill
-; RV32IZHINX-NEXT:    sw s0, 8(sp) # 4-byte Folded Spill
-; RV32IZHINX-NEXT:    sw s1, 4(sp) # 4-byte Folded Spill
-; RV32IZHINX-NEXT:    sw s2, 0(sp) # 4-byte Folded Spill
 ; RV32IZHINX-NEXT:    lui a1, %hi(.LCPI5_0)
 ; RV32IZHINX-NEXT:    lh a1, %lo(.LCPI5_0)(a1)
 ; RV32IZHINX-NEXT:    fabs.h a2, a0
 ; RV32IZHINX-NEXT:    flt.h a1, a2, a1
 ; RV32IZHINX-NEXT:    beqz a1, .LBB5_2
 ; RV32IZHINX-NEXT:  # %bb.1:
 ; RV32IZHINX-NEXT:    fcvt.w.h a1, a0, rup
 ; RV32IZHINX-NEXT:    fcvt.h.w a1, a1, rup
 ; RV32IZHINX-NEXT:    fsgnj.h a0, a1, a0
 ; RV32IZHINX-NEXT:  .LBB5_2:
+; RV32IZHINX-NEXT:    addi sp, sp, -16
+; RV32IZHINX-NEXT:    sw ra, 12(sp) # 4-byte Folded Spill
+; RV32IZHINX-NEXT:    sw s0, 8(sp) # 4-byte Folded Spill
+; RV32IZHINX-NEXT:    sw s1, 4(sp) # 4-byte Folded Spill
+; RV32IZHINX-NEXT:    sw s2, 0(sp) # 4-byte Folded Spill
 ; RV32IZHINX-NEXT:    fcvt.s.h s0, a0
 ; RV32IZHINX-NEXT:    lui a0, 913408
 ; RV32IZHINX-NEXT:    fle.s s1, a0, s0
 ; RV32IZHINX-NEXT:    neg s2, s1
 ; RV32IZHINX-NEXT:    mv a0, s0
 ; RV32IZHINX-NEXT:    call __fixsfdi@plt
 ; RV32IZHINX-NEXT:    lui a2, %hi(.LCPI5_1)
 ; RV32IZHINX-NEXT:    lw a2, %lo(.LCPI5_1)(a2)
 ; RV32IZHINX-NEXT:    and a0, s2, a0
 ; RV32IZHINX-NEXT:    flt.s a4, a2, s0
 ; RV32IZHINX-NEXT:    neg a2, a4
 ; RV32IZHINX-NEXT:    or a0, a2, a0
 ; RV32IZHINX-NEXT:    feq.s a2, s0, s0
 ; RV32IZHINX-NEXT:    neg a2, a2
 ; RV32IZHINX-NEXT:    lui a5, 524288
 ; RV32IZHINX-NEXT:    lui a3, 524288
 ; RV32IZHINX-NEXT:    beqz s1, .LBB5_4
 ; RV32IZHINX-NEXT:  # %bb.3:
 ; RV32IZHINX-NEXT:    mv a3, a1
 ; RV32IZHINX-NEXT:  .LBB5_4:
 ; RV32IZHINX-NEXT:    and a0, a2, a0
-; RV32IZHINX-NEXT:    beqz a4, .LBB5_6
-; RV32IZHINX-NEXT:  # %bb.5:
-; RV32IZHINX-NEXT:    addi a3, a5, -1
-; RV32IZHINX-NEXT:  .LBB5_6:
-; RV32IZHINX-NEXT:    and a1, a2, a3
 ; RV32IZHINX-NEXT:    lw ra, 12(sp) # 4-byte Folded Reload
 ; RV32IZHINX-NEXT:    lw s0, 8(sp) # 4-byte Folded Reload
 ; RV32IZHINX-NEXT:    lw s1, 4(sp) # 4-byte Folded Reload
 ; RV32IZHINX-NEXT:    lw s2, 0(sp) # 4-byte Folded Reload
 ; RV32IZHINX-NEXT:    addi sp, sp, 16
+; RV32IZHINX-NEXT:    beqz a4, .LBB5_6
+; RV32IZHINX-NEXT:  # %bb.5:
+; RV32IZHINX-NEXT:    addi a3, a5, -1
+; RV32IZHINX-NEXT:  .LBB5_6:
+; RV32IZHINX-NEXT:    and a1, a2, a3
 ; RV32IZHINX-NEXT:    ret
 ;
 ; RV64IZHINX-LABEL: test_ceil_si64:
 ; RV64IZHINX:       # %bb.0:
 ; RV64IZHINX-NEXT:    lui a1, %hi(.LCPI5_0)
 ; RV64IZHINX-NEXT:    lh a1, %lo(.LCPI5_0)(a1)
 ; RV64IZHINX-NEXT:    fabs.h a2, a0
 ; RV64IZHINX-NEXT:    flt.h a1, a2, a1
; RV64IZHINXMIN-NEXT:    ret
   %a = call half @llvm.ceil.f16(half %x)
   %b = call i32 @llvm.fptoui.sat.i32.f16(half %a)
   ret i32 %b
 }
 
 define i64 @test_ceil_ui64(half %x) nounwind {
 ; RV32IZFH-LABEL: test_ceil_ui64:
 ; RV32IZFH:       # %bb.0:
-; RV32IZFH-NEXT:    addi sp, sp, -16
-; RV32IZFH-NEXT:    sw ra, 12(sp) # 4-byte Folded Spill
-; RV32IZFH-NEXT:    sw s0, 8(sp) # 4-byte Folded Spill
-; RV32IZFH-NEXT:    fsw fs0, 4(sp) # 4-byte Folded Spill
 ; RV32IZFH-NEXT:    lui a0, %hi(.LCPI7_0)
 ; RV32IZFH-NEXT:    flh fa5, %lo(.LCPI7_0)(a0)
 ; RV32IZFH-NEXT:    fabs.h fa4, fa0
 ; RV32IZFH-NEXT:    flt.h a0, fa4, fa5
 ; RV32IZFH-NEXT:    beqz a0, .LBB7_2
 ; RV32IZFH-NEXT:  # %bb.1:
 ; RV32IZFH-NEXT:    fcvt.w.h a0, fa0, rup
 ; RV32IZFH-NEXT:    fcvt.h.w fa5, a0, rup
 ; RV32IZFH-NEXT:    fsgnj.h fa0, fa5, fa0
 ; RV32IZFH-NEXT:  .LBB7_2:
+; RV32IZFH-NEXT:    addi sp, sp, -16
+; RV32IZFH-NEXT:    sw ra, 12(sp) # 4-byte Folded Spill
+; RV32IZFH-NEXT:    sw s0, 8(sp) # 4-byte Folded Spill
+; RV32IZFH-NEXT:    fsw fs0, 4(sp) # 4-byte Folded Spill
 ; RV32IZFH-NEXT:    fcvt.s.h fs0, fa0
 ; RV32IZFH-NEXT:    fmv.w.x fa5, zero
 ; RV32IZFH-NEXT:    fle.s a0, fa5, fs0
 ; RV32IZFH-NEXT:    neg s0, a0
 ; RV32IZFH-NEXT:    fmv.s fa0, fs0
 ; RV32IZFH-NEXT:    call __fixunssfdi@plt
 ; RV32IZFH-NEXT:    lui a2, %hi(.LCPI7_1)
 ; RV32IZFH-NEXT:    flw fa5, %lo(.LCPI7_1)(a2)

 ; RV64IZFH-NEXT:    feq.h a1, fa0, fa0
 ; RV64IZFH-NEXT:    seqz a1, a1
 ; RV64IZFH-NEXT:    addi a1, a1, -1
 ; RV64IZFH-NEXT:    and a0, a1, a0
 ; RV64IZFH-NEXT:    ret
 ;
 ; RV32IZHINX-LABEL: test_ceil_ui64:
 ; RV32IZHINX:       # %bb.0:
-; RV32IZHINX-NEXT:    addi sp, sp, -16
-; RV32IZHINX-NEXT:    sw ra, 12(sp) # 4-byte Folded Spill
-; RV32IZHINX-NEXT:    sw s0, 8(sp) # 4-byte Folded Spill
-; RV32IZHINX-NEXT:    sw s1, 4(sp) # 4-byte Folded Spill
 ; RV32IZHINX-NEXT:    lui a1, %hi(.LCPI7_0)
 ; RV32IZHINX-NEXT:    lh a1, %lo(.LCPI7_0)(a1)
 ; RV32IZHINX-NEXT:    fabs.h a2, a0
 ; RV32IZHINX-NEXT:    flt.h a1, a2, a1
 ; RV32IZHINX-NEXT:    beqz a1, .LBB7_2
 ; RV32IZHINX-NEXT:  # %bb.1:
 ; RV32IZHINX-NEXT:    fcvt.w.h a1, a0, rup
 ; RV32IZHINX-NEXT:    fcvt.h.w a1, a1, rup
 ; RV32IZHINX-NEXT:    fsgnj.h a0, a1, a0
 ; RV32IZHINX-NEXT:  .LBB7_2:
+; RV32IZHINX-NEXT:    addi sp, sp, -16
+; RV32IZHINX-NEXT:    sw ra, 12(sp) # 4-byte Folded Spill
+; RV32IZHINX-NEXT:    sw s0, 8(sp) # 4-byte Folded Spill
+; RV32IZHINX-NEXT:    sw s1, 4(sp) # 4-byte Folded Spill
 ; RV32IZHINX-NEXT:    fcvt.s.h s0, a0
 ; RV32IZHINX-NEXT:    fle.s a0, zero, s0
 ; RV32IZHINX-NEXT:    neg s1, a0
 ; RV32IZHINX-NEXT:    mv a0, s0
 ; RV32IZHINX-NEXT:    call __fixunssfdi@plt
 ; RV32IZHINX-NEXT:    lui a2, %hi(.LCPI7_1)
 ; RV32IZHINX-NEXT:    lw a2, %lo(.LCPI7_1)(a2)
 ; RV32IZHINX-NEXT:    and a0, s1, a0
; CHECKIZHINXMIN-NEXT:    ret
   %a = call half @llvm.trunc.f16(half %x)
   %b = call i32 @llvm.fptosi.sat.i32.f16(half %a)
   ret i32 %b
 }
 
 define i64 @test_trunc_si64(half %x) nounwind {
 ; RV32IZFH-LABEL: test_trunc_si64:
 ; RV32IZFH:       # %bb.0:
-; RV32IZFH-NEXT:    addi sp, sp, -16
-; RV32IZFH-NEXT:    sw ra, 12(sp) # 4-byte Folded Spill
-; RV32IZFH-NEXT:    sw s0, 8(sp) # 4-byte Folded Spill
-; RV32IZFH-NEXT:    fsw fs0, 4(sp) # 4-byte Folded Spill
 ; RV32IZFH-NEXT:    lui a0, %hi(.LCPI9_0)
 ; RV32IZFH-NEXT:    flh fa5, %lo(.LCPI9_0)(a0)
 ; RV32IZFH-NEXT:    fabs.h fa4, fa0
 ; RV32IZFH-NEXT:    flt.h a0, fa4, fa5
 ; RV32IZFH-NEXT:    beqz a0, .LBB9_2
 ; RV32IZFH-NEXT:  # %bb.1:
 ; RV32IZFH-NEXT:    fcvt.w.h a0, fa0, rtz
 ; RV32IZFH-NEXT:    fcvt.h.w fa5, a0, rtz
 ; RV32IZFH-NEXT:    fsgnj.h fa0, fa5, fa0
 ; RV32IZFH-NEXT:  .LBB9_2:
+; RV32IZFH-NEXT:    addi sp, sp, -16
+; RV32IZFH-NEXT:    sw ra, 12(sp) # 4-byte Folded Spill
+; RV32IZFH-NEXT:    sw s0, 8(sp) # 4-byte Folded Spill
+; RV32IZFH-NEXT:    fsw fs0, 4(sp) # 4-byte Folded Spill
 ; RV32IZFH-NEXT:    fcvt.s.h fs0, fa0
 ; RV32IZFH-NEXT:    lui a0, 913408
 ; RV32IZFH-NEXT:    fmv.w.x fa5, a0
 ; RV32IZFH-NEXT:    fle.s s0, fa5, fs0
 ; RV32IZFH-NEXT:    fmv.s fa0, fs0
 ; RV32IZFH-NEXT:    call __fixsfdi@plt
 ; RV32IZFH-NEXT:    lui a4, 524288
 ; RV32IZFH-NEXT:    lui a2, 524288

 ; RV64IZFH-NEXT:    feq.h a1, fa0, fa0
 ; RV64IZFH-NEXT:    seqz a1, a1
 ; RV64IZFH-NEXT:    addi a1, a1, -1
 ; RV64IZFH-NEXT:    and a0, a1, a0
 ; RV64IZFH-NEXT:    ret
 ;
 ; RV32IZHINX-LABEL: test_trunc_si64:
 ; RV32IZHINX:       # %bb.0:
-; RV32IZHINX-NEXT:    addi sp, sp, -16
-; RV32IZHINX-NEXT:    sw ra, 12(sp) # 4-byte Folded Spill
-; RV32IZHINX-NEXT:    sw s0, 8(sp) # 4-byte Folded Spill
-; RV32IZHINX-NEXT:    sw s1, 4(sp) # 4-byte Folded Spill
-; RV32IZHINX-NEXT:    sw s2, 0(sp) # 4-byte Folded Spill
 ; RV32IZHINX-NEXT:    lui a1, %hi(.LCPI9_0)
 ; RV32IZHINX-NEXT:    lh a1, %lo(.LCPI9_0)(a1)
 ; RV32IZHINX-NEXT:    fabs.h a2, a0
 ; RV32IZHINX-NEXT:    flt.h a1, a2, a1
 ; RV32IZHINX-NEXT:    beqz a1, .LBB9_2
 ; RV32IZHINX-NEXT:  # %bb.1:
 ; RV32IZHINX-NEXT:    fcvt.w.h a1, a0, rtz
 ; RV32IZHINX-NEXT:    fcvt.h.w a1, a1, rtz
 ; RV32IZHINX-NEXT:    fsgnj.h a0, a1, a0
 ; RV32IZHINX-NEXT:  .LBB9_2:
+; RV32IZHINX-NEXT:    addi sp, sp, -16
+; RV32IZHINX-NEXT:    sw ra, 12(sp) # 4-byte Folded Spill
+; RV32IZHINX-NEXT:    sw s0, 8(sp) # 4-byte Folded Spill
+; RV32IZHINX-NEXT:    sw s1, 4(sp) # 4-byte Folded Spill
+; RV32IZHINX-NEXT:    sw s2, 0(sp) # 4-byte Folded Spill
 ; RV32IZHINX-NEXT:    fcvt.s.h s0, a0
 ; RV32IZHINX-NEXT:    lui a0, 913408
 ; RV32IZHINX-NEXT:    fle.s s1, a0, s0
 ; RV32IZHINX-NEXT:    neg s2, s1
 ; RV32IZHINX-NEXT:    mv a0, s0
 ; RV32IZHINX-NEXT:    call __fixsfdi@plt
 ; RV32IZHINX-NEXT:    lui a2, %hi(.LCPI9_1)
 ; RV32IZHINX-NEXT:    lw a2, %lo(.LCPI9_1)(a2)
 ; RV32IZHINX-NEXT:    and a0, s2, a0
 ; RV32IZHINX-NEXT:    flt.s a4, a2, s0
 ; RV32IZHINX-NEXT:    neg a2, a4
 ; RV32IZHINX-NEXT:    or a0, a2, a0
 ; RV32IZHINX-NEXT:    feq.s a2, s0, s0
 ; RV32IZHINX-NEXT:    neg a2, a2
 ; RV32IZHINX-NEXT:    lui a5, 524288
 ; RV32IZHINX-NEXT:    lui a3, 524288
 ; RV32IZHINX-NEXT:    beqz s1, .LBB9_4
 ; RV32IZHINX-NEXT:  # %bb.3:
 ; RV32IZHINX-NEXT:    mv a3, a1
 ; RV32IZHINX-NEXT:  .LBB9_4:
 ; RV32IZHINX-NEXT:    and a0, a2, a0
-; RV32IZHINX-NEXT:    beqz a4, .LBB9_6
-; RV32IZHINX-NEXT:  # %bb.5:
-; RV32IZHINX-NEXT:    addi a3, a5, -1
-; RV32IZHINX-NEXT:  .LBB9_6:
-; RV32IZHINX-NEXT:    and a1, a2, a3
 ; RV32IZHINX-NEXT:    lw ra, 12(sp) # 4-byte Folded Reload
 ; RV32IZHINX-NEXT:    lw s0, 8(sp) # 4-byte Folded Reload
 ; RV32IZHINX-NEXT:    lw s1, 4(sp) # 4-byte Folded Reload
 ; RV32IZHINX-NEXT:    lw s2, 0(sp) # 4-byte Folded Reload
 ; RV32IZHINX-NEXT:    addi sp, sp, 16
+; RV32IZHINX-NEXT:    beqz a4, .LBB9_6
+; RV32IZHINX-NEXT:  # %bb.5:
+; RV32IZHINX-NEXT:    addi a3, a5, -1
+; RV32IZHINX-NEXT:  .LBB9_6:
+; RV32IZHINX-NEXT:    and a1, a2, a3
 ; RV32IZHINX-NEXT:    ret
 ;
 ; RV64IZHINX-LABEL: test_trunc_si64:
 ; RV64IZHINX:       # %bb.0:
 ; RV64IZHINX-NEXT:    lui a1, %hi(.LCPI9_0)
 ; RV64IZHINX-NEXT:    lh a1, %lo(.LCPI9_0)(a1)
 ; RV64IZHINX-NEXT:    fabs.h a2, a0
 ; RV64IZHINX-NEXT:    flt.h a1, a2, a1
; RV64IZHINXMIN-NEXT:    ret
   %a = call half @llvm.trunc.f16(half %x)
   %b = call i32 @llvm.fptoui.sat.i32.f16(half %a)
   ret i32 %b
 }
 
 define i64 @test_trunc_ui64(half %x) nounwind {
 ; RV32IZFH-LABEL: test_trunc_ui64:
 ; RV32IZFH:       # %bb.0:
-; RV32IZFH-NEXT:    addi sp, sp, -16
-; RV32IZFH-NEXT:    sw ra, 12(sp) # 4-byte Folded Spill
-; RV32IZFH-NEXT:    sw s0, 8(sp) # 4-byte Folded Spill
-; RV32IZFH-NEXT:    fsw fs0, 4(sp) # 4-byte Folded Spill
 ; RV32IZFH-NEXT:    lui a0, %hi(.LCPI11_0)
 ; RV32IZFH-NEXT:    flh fa5, %lo(.LCPI11_0)(a0)
 ; RV32IZFH-NEXT:    fabs.h fa4, fa0
 ; RV32IZFH-NEXT:    flt.h a0, fa4, fa5
 ; RV32IZFH-NEXT:    beqz a0, .LBB11_2
 ; RV32IZFH-NEXT:  # %bb.1:
 ; RV32IZFH-NEXT:    fcvt.w.h a0, fa0, rtz
 ; RV32IZFH-NEXT:    fcvt.h.w fa5, a0, rtz
 ; RV32IZFH-NEXT:    fsgnj.h fa0, fa5, fa0
 ; RV32IZFH-NEXT:  .LBB11_2:
+; RV32IZFH-NEXT:    addi sp, sp, -16
+; RV32IZFH-NEXT:    sw ra, 12(sp) # 4-byte Folded Spill
+; RV32IZFH-NEXT:    sw s0, 8(sp) # 4-byte Folded Spill
+; RV32IZFH-NEXT:    fsw fs0, 4(sp) # 4-byte Folded Spill
 ; RV32IZFH-NEXT:    fcvt.s.h fs0, fa0
 ; RV32IZFH-NEXT:    fmv.w.x fa5, zero
 ; RV32IZFH-NEXT:    fle.s a0, fa5, fs0
 ; RV32IZFH-NEXT:    neg s0, a0
 ; RV32IZFH-NEXT:    fmv.s fa0, fs0
 ; RV32IZFH-NEXT:    call __fixunssfdi@plt
 ; RV32IZFH-NEXT:    lui a2, %hi(.LCPI11_1)
 ; RV32IZFH-NEXT:    flw fa5, %lo(.LCPI11_1)(a2)

 ; RV64IZFH-NEXT:    feq.h a1, fa0, fa0
 ; RV64IZFH-NEXT:    seqz a1, a1
 ; RV64IZFH-NEXT:    addi a1, a1, -1
 ; RV64IZFH-NEXT:    and a0, a1, a0
 ; RV64IZFH-NEXT:    ret
 ;
 ; RV32IZHINX-LABEL: test_trunc_ui64:
 ; RV32IZHINX:       # %bb.0:
-; RV32IZHINX-NEXT:    addi sp, sp, -16
-; RV32IZHINX-NEXT:    sw ra, 12(sp) # 4-byte Folded Spill
-; RV32IZHINX-NEXT:    sw s0, 8(sp) # 4-byte Folded Spill
-; RV32IZHINX-NEXT:    sw s1, 4(sp) # 4-byte Folded Spill
 ; RV32IZHINX-NEXT:    lui a1, %hi(.LCPI11_0)
 ; RV32IZHINX-NEXT:    lh a1, %lo(.LCPI11_0)(a1)
 ; RV32IZHINX-NEXT:    fabs.h a2, a0
 ; RV32IZHINX-NEXT:    flt.h a1, a2, a1
 ; RV32IZHINX-NEXT:    beqz a1, .LBB11_2
 ; RV32IZHINX-NEXT:  # %bb.1:
 ; RV32IZHINX-NEXT:    fcvt.w.h a1, a0, rtz
 ; RV32IZHINX-NEXT:    fcvt.h.w a1, a1, rtz
 ; RV32IZHINX-NEXT:    fsgnj.h a0, a1, a0
 ; RV32IZHINX-NEXT:  .LBB11_2:
+; RV32IZHINX-NEXT:    addi sp, sp, -16
+; RV32IZHINX-NEXT:    sw ra, 12(sp) # 4-byte Folded Spill
+; RV32IZHINX-NEXT:    sw s0, 8(sp) # 4-byte Folded Spill
+; RV32IZHINX-NEXT:    sw s1, 4(sp) # 4-byte Folded Spill
 ; RV32IZHINX-NEXT:    fcvt.s.h s0, a0
 ; RV32IZHINX-NEXT:    fle.s a0, zero, s0
 ; RV32IZHINX-NEXT:    neg s1, a0
 ; RV32IZHINX-NEXT:    mv a0, s0
 ; RV32IZHINX-NEXT:    call __fixunssfdi@plt
 ; RV32IZHINX-NEXT:    lui a2, %hi(.LCPI11_1)
 ; RV32IZHINX-NEXT:    lw a2, %lo(.LCPI11_1)(a2)
 ; RV32IZHINX-NEXT:    and a0, s1, a0
; CHECKIZHINXMIN-NEXT:    ret
   %a = call half @llvm.round.f16(half %x)
   %b = call i32 @llvm.fptosi.sat.i32.f16(half %a)
   ret i32 %b
 }
 
 define i64 @test_round_si64(half %x) nounwind {
 ; RV32IZFH-LABEL: test_round_si64:
 ; RV32IZFH:       # %bb.0:
-; RV32IZFH-NEXT:    addi sp, sp, -16
-; RV32IZFH-NEXT:    sw ra, 12(sp) # 4-byte Folded Spill
-; RV32IZFH-NEXT:    sw s0, 8(sp) # 4-byte Folded Spill
-; RV32IZFH-NEXT:    fsw fs0, 4(sp) # 4-byte Folded Spill
 ; RV32IZFH-NEXT:    lui a0, %hi(.LCPI13_0)
 ; RV32IZFH-NEXT:    flh fa5, %lo(.LCPI13_0)(a0)
 ; RV32IZFH-NEXT:    fabs.h fa4, fa0
 ; RV32IZFH-NEXT:    flt.h a0, fa4, fa5
 ; RV32IZFH-NEXT:    beqz a0, .LBB13_2
 ; RV32IZFH-NEXT:  # %bb.1:
 ; RV32IZFH-NEXT:    fcvt.w.h a0, fa0, rmm
 ; RV32IZFH-NEXT:    fcvt.h.w fa5, a0, rmm
 ; RV32IZFH-NEXT:    fsgnj.h fa0, fa5, fa0
 ; RV32IZFH-NEXT:  .LBB13_2:
+; RV32IZFH-NEXT:    addi sp, sp, -16
+; RV32IZFH-NEXT:    sw ra, 12(sp) # 4-byte Folded Spill
+; RV32IZFH-NEXT:    sw s0, 8(sp) # 4-byte Folded Spill
+; RV32IZFH-NEXT:    fsw fs0, 4(sp) # 4-byte Folded Spill
 ; RV32IZFH-NEXT:    fcvt.s.h fs0, fa0
 ; RV32IZFH-NEXT:    lui a0, 913408
 ; RV32IZFH-NEXT:    fmv.w.x fa5, a0
 ; RV32IZFH-NEXT:    fle.s s0, fa5, fs0
 ; RV32IZFH-NEXT:    fmv.s fa0, fs0
 ; RV32IZFH-NEXT:    call __fixsfdi@plt
 ; RV32IZFH-NEXT:    lui a4, 524288
 ; RV32IZFH-NEXT:    lui a2, 524288

 ; RV64IZFH-NEXT:    feq.h a1, fa0, fa0
 ; RV64IZFH-NEXT:    seqz a1, a1
 ; RV64IZFH-NEXT:    addi a1, a1, -1
 ; RV64IZFH-NEXT:    and a0, a1, a0
 ; RV64IZFH-NEXT:    ret
 ;
 ; RV32IZHINX-LABEL: test_round_si64:
 ; RV32IZHINX:       # %bb.0:
-; RV32IZHINX-NEXT:    addi sp, sp, -16
-; RV32IZHINX-NEXT:    sw ra, 12(sp) # 4-byte Folded Spill
-; RV32IZHINX-NEXT:    sw s0, 8(sp) # 4-byte Folded Spill
-; RV32IZHINX-NEXT:    sw s1, 4(sp) # 4-byte Folded Spill
-; RV32IZHINX-NEXT:    sw s2, 0(sp) # 4-byte Folded Spill
 ; RV32IZHINX-NEXT:    lui a1, %hi(.LCPI13_0)
 ; RV32IZHINX-NEXT:    lh a1, %lo(.LCPI13_0)(a1)
 ; RV32IZHINX-NEXT:    fabs.h a2, a0
 ; RV32IZHINX-NEXT:    flt.h a1, a2, a1
 ; RV32IZHINX-NEXT:    beqz a1, .LBB13_2
 ; RV32IZHINX-NEXT:  # %bb.1:
 ; RV32IZHINX-NEXT:    fcvt.w.h a1, a0, rmm
 ; RV32IZHINX-NEXT:    fcvt.h.w a1, a1, rmm
 ; RV32IZHINX-NEXT:    fsgnj.h a0, a1, a0
 ; RV32IZHINX-NEXT:  .LBB13_2:
+; RV32IZHINX-NEXT:    addi sp, sp, -16
+; RV32IZHINX-NEXT:    sw ra, 12(sp) # 4-byte Folded Spill
+; RV32IZHINX-NEXT:    sw s0, 8(sp) # 4-byte Folded Spill
+; RV32IZHINX-NEXT:    sw s1, 4(sp) # 4-byte Folded Spill
+; RV32IZHINX-NEXT:    sw s2, 0(sp) # 4-byte Folded Spill
 ; RV32IZHINX-NEXT:    fcvt.s.h s0, a0
 ; RV32IZHINX-NEXT:    lui a0, 913408
 ; RV32IZHINX-NEXT:    fle.s s1, a0, s0
 ; RV32IZHINX-NEXT:    neg s2, s1
 ; RV32IZHINX-NEXT:    mv a0, s0
 ; RV32IZHINX-NEXT:    call __fixsfdi@plt
 ; RV32IZHINX-NEXT:    lui a2, %hi(.LCPI13_1)
 ; RV32IZHINX-NEXT:    lw a2, %lo(.LCPI13_1)(a2)
 ; RV32IZHINX-NEXT:    and a0, s2, a0
 ; RV32IZHINX-NEXT:    flt.s a4, a2, s0
 ; RV32IZHINX-NEXT:    neg a2, a4
 ; RV32IZHINX-NEXT:    or a0, a2, a0
 ; RV32IZHINX-NEXT:    feq.s a2, s0, s0
 ; RV32IZHINX-NEXT:    neg a2, a2
 ; RV32IZHINX-NEXT:    lui a5, 524288
 ; RV32IZHINX-NEXT:    lui a3, 524288
 ; RV32IZHINX-NEXT:    beqz s1, .LBB13_4
 ; RV32IZHINX-NEXT:  # %bb.3:
 ; RV32IZHINX-NEXT:    mv a3, a1
 ; RV32IZHINX-NEXT:  .LBB13_4:
 ; RV32IZHINX-NEXT:    and a0, a2, a0
-; RV32IZHINX-NEXT:    beqz a4, .LBB13_6
-; RV32IZHINX-NEXT:  # %bb.5:
-; RV32IZHINX-NEXT:    addi a3, a5, -1
-; RV32IZHINX-NEXT:  .LBB13_6:
-; RV32IZHINX-NEXT:    and a1, a2, a3
 ; RV32IZHINX-NEXT:    lw ra, 12(sp) # 4-byte Folded Reload
 ; RV32IZHINX-NEXT:    lw s0, 8(sp) # 4-byte Folded Reload
 ; RV32IZHINX-NEXT:    lw s1, 4(sp) # 4-byte Folded Reload
 ; RV32IZHINX-NEXT:    lw s2, 0(sp) # 4-byte Folded Reload
 ; RV32IZHINX-NEXT:    addi sp, sp, 16
+; RV32IZHINX-NEXT:    beqz a4, .LBB13_6
+; RV32IZHINX-NEXT:  # %bb.5:
+; RV32IZHINX-NEXT:    addi a3, a5, -1
+; RV32IZHINX-NEXT:  .LBB13_6:
+; RV32IZHINX-NEXT:    and a1, a2, a3
 ; RV32IZHINX-NEXT:    ret
 ;
 ; RV64IZHINX-LABEL: test_round_si64:
 ; RV64IZHINX:       # %bb.0:
 ; RV64IZHINX-NEXT:    lui a1, %hi(.LCPI13_0)
 ; RV64IZHINX-NEXT:    lh a1, %lo(.LCPI13_0)(a1)
 ; RV64IZHINX-NEXT:    fabs.h a2, a0
 ; RV64IZHINX-NEXT:    flt.h a1, a2, a1
; RV64IZHINXMIN-NEXT:    ret
   %a = call half @llvm.round.f16(half %x)
   %b = call i32 @llvm.fptoui.sat.i32.f16(half %a)
   ret i32 %b
 }
 
 define i64 @test_round_ui64(half %x) nounwind {
 ; RV32IZFH-LABEL: test_round_ui64:
 ; RV32IZFH:       # %bb.0:
-; RV32IZFH-NEXT:    addi sp, sp, -16
-; RV32IZFH-NEXT:    sw ra, 12(sp) # 4-byte Folded Spill
-; RV32IZFH-NEXT:    sw s0, 8(sp) # 4-byte Folded Spill
-; RV32IZFH-NEXT:    fsw fs0, 4(sp) # 4-byte Folded Spill
 ; RV32IZFH-NEXT:    lui a0, %hi(.LCPI15_0)
 ; RV32IZFH-NEXT:    flh fa5, %lo(.LCPI15_0)(a0)
 ; RV32IZFH-NEXT:    fabs.h fa4, fa0
 ; RV32IZFH-NEXT:    flt.h a0, fa4, fa5
 ; RV32IZFH-NEXT:    beqz a0, .LBB15_2
 ; RV32IZFH-NEXT:  # %bb.1:
 ; RV32IZFH-NEXT:    fcvt.w.h a0, fa0, rmm
 ; RV32IZFH-NEXT:    fcvt.h.w fa5, a0, rmm
 ; RV32IZFH-NEXT:    fsgnj.h fa0, fa5, fa0
 ; RV32IZFH-NEXT:  .LBB15_2:
+; RV32IZFH-NEXT:    addi sp, sp, -16
+; RV32IZFH-NEXT:    sw ra, 12(sp) # 4-byte Folded Spill
+; RV32IZFH-NEXT:    sw s0, 8(sp) # 4-byte Folded Spill
+; RV32IZFH-NEXT:    fsw fs0, 4(sp) # 4-byte Folded Spill
 ; RV32IZFH-NEXT:    fcvt.s.h fs0, fa0
 ; RV32IZFH-NEXT:    fmv.w.x fa5, zero
 ; RV32IZFH-NEXT:    fle.s a0, fa5, fs0
 ; RV32IZFH-NEXT:    neg s0, a0
 ; RV32IZFH-NEXT:    fmv.s fa0, fs0
 ; RV32IZFH-NEXT:    call __fixunssfdi@plt
 ; RV32IZFH-NEXT:    lui a2, %hi(.LCPI15_1)
 ; RV32IZFH-NEXT:    flw fa5, %lo(.LCPI15_1)(a2)

 ; RV64IZFH-NEXT:    feq.h a1, fa0, fa0
 ; RV64IZFH-NEXT:    seqz a1, a1
 ; RV64IZFH-NEXT:    addi a1, a1, -1
 ; RV64IZFH-NEXT:    and a0, a1, a0
 ; RV64IZFH-NEXT:    ret
 ;
 ; RV32IZHINX-LABEL: test_round_ui64:
 ; RV32IZHINX:       # %bb.0:
-; RV32IZHINX-NEXT:    addi sp, sp, -16
-; RV32IZHINX-NEXT:    sw ra, 12(sp) # 4-byte Folded Spill
-; RV32IZHINX-NEXT:    sw s0, 8(sp) # 4-byte Folded Spill
-; RV32IZHINX-NEXT:    sw s1, 4(sp) # 4-byte Folded Spill
 ; RV32IZHINX-NEXT:    lui a1, %hi(.LCPI15_0)
 ; RV32IZHINX-NEXT:    lh a1, %lo(.LCPI15_0)(a1)
 ; RV32IZHINX-NEXT:    fabs.h a2, a0
 ; RV32IZHINX-NEXT:    flt.h a1, a2, a1
 ; RV32IZHINX-NEXT:    beqz a1, .LBB15_2
 ; RV32IZHINX-NEXT:  # %bb.1:
 ; RV32IZHINX-NEXT:    fcvt.w.h a1, a0, rmm
 ; RV32IZHINX-NEXT:    fcvt.h.w a1, a1, rmm
 ; RV32IZHINX-NEXT:    fsgnj.h a0, a1, a0
 ; RV32IZHINX-NEXT:  .LBB15_2:
+; RV32IZHINX-NEXT:    addi sp, sp, -16
+; RV32IZHINX-NEXT:    sw ra, 12(sp) # 4-byte Folded Spill
+; RV32IZHINX-NEXT:    sw s0, 8(sp) # 4-byte Folded Spill
+; RV32IZHINX-NEXT:    sw s1, 4(sp) # 4-byte Folded Spill
 ; RV32IZHINX-NEXT:    fcvt.s.h s0, a0
 ; RV32IZHINX-NEXT:    fle.s a0, zero, s0
 ; RV32IZHINX-NEXT:    neg s1, a0
 ; RV32IZHINX-NEXT:    mv a0, s0
 ; RV32IZHINX-NEXT:    call __fixunssfdi@plt
 ; RV32IZHINX-NEXT:    lui a2, %hi(.LCPI15_1)
 ; RV32IZHINX-NEXT:    lw a2, %lo(.LCPI15_1)(a2)
 ; RV32IZHINX-NEXT:    and a0, s1, a0
; CHECKIZHINXMIN-NEXT:    ret
   %a = call half @llvm.roundeven.f16(half %x)
   %b = call i32 @llvm.fptosi.sat.i32.f16(half %a)
   ret i32 %b
 }
 
 define i64 @test_roundeven_si64(half %x) nounwind {
 ; RV32IZFH-LABEL: test_roundeven_si64:
 ; RV32IZFH:       # %bb.0:
-; RV32IZFH-NEXT:    addi sp, sp, -16
-; RV32IZFH-NEXT:    sw ra, 12(sp) # 4-byte Folded Spill
-; RV32IZFH-NEXT:    sw s0, 8(sp) # 4-byte Folded Spill
-; RV32IZFH-NEXT:    fsw fs0, 4(sp) # 4-byte Folded Spill
 ; RV32IZFH-NEXT:    lui a0, %hi(.LCPI17_0)
 ; RV32IZFH-NEXT:    flh fa5, %lo(.LCPI17_0)(a0)
 ; RV32IZFH-NEXT:    fabs.h fa4, fa0
 ; RV32IZFH-NEXT:    flt.h a0, fa4, fa5
 ; RV32IZFH-NEXT:    beqz a0, .LBB17_2
 ; RV32IZFH-NEXT:  # %bb.1:
 ; RV32IZFH-NEXT:    fcvt.w.h a0, fa0, rne
 ; RV32IZFH-NEXT:    fcvt.h.w fa5, a0, rne
 ; RV32IZFH-NEXT:    fsgnj.h fa0, fa5, fa0
 ; RV32IZFH-NEXT:  .LBB17_2:
+; RV32IZFH-NEXT:    addi sp, sp, -16
+; RV32IZFH-NEXT:    sw ra, 12(sp) # 4-byte Folded Spill
+; RV32IZFH-NEXT:    sw s0, 8(sp) # 4-byte Folded Spill
+; RV32IZFH-NEXT:    fsw fs0, 4(sp) # 4-byte Folded Spill
 ; RV32IZFH-NEXT:    fcvt.s.h fs0, fa0
 ; RV32IZFH-NEXT:    lui a0, 913408
 ; RV32IZFH-NEXT:    fmv.w.x fa5, a0
 ; RV32IZFH-NEXT:    fle.s s0, fa5, fs0
 ; RV32IZFH-NEXT:    fmv.s fa0, fs0
 ; RV32IZFH-NEXT:    call __fixsfdi@plt
 ; RV32IZFH-NEXT:    lui a4, 524288
 ; RV32IZFH-NEXT:    lui a2, 524288

 ; RV64IZFH-NEXT:    feq.h a1, fa0, fa0
 ; RV64IZFH-NEXT:    seqz a1, a1
 ; RV64IZFH-NEXT:    addi a1, a1, -1
 ; RV64IZFH-NEXT:    and a0, a1, a0
 ; RV64IZFH-NEXT:    ret
 ;
 ; RV32IZHINX-LABEL: test_roundeven_si64:
 ; RV32IZHINX:       # %bb.0:
-; RV32IZHINX-NEXT:    addi sp, sp, -16
-; RV32IZHINX-NEXT:    sw ra, 12(sp) # 4-byte Folded Spill
-; RV32IZHINX-NEXT:    sw s0, 8(sp) # 4-byte Folded Spill
-; RV32IZHINX-NEXT:    sw s1, 4(sp) # 4-byte Folded Spill
-; RV32IZHINX-NEXT:    sw s2, 0(sp) # 4-byte Folded Spill
 ; RV32IZHINX-NEXT:    lui a1, %hi(.LCPI17_0)
 ; RV32IZHINX-NEXT:    lh a1, %lo(.LCPI17_0)(a1)
 ; RV32IZHINX-NEXT:    fabs.h a2, a0
 ; RV32IZHINX-NEXT:    flt.h a1, a2, a1
 ; RV32IZHINX-NEXT:    beqz a1, .LBB17_2
 ; RV32IZHINX-NEXT:  # %bb.1:
 ; RV32IZHINX-NEXT:    fcvt.w.h a1, a0, rne
 ; RV32IZHINX-NEXT:    fcvt.h.w a1, a1, rne
 ; RV32IZHINX-NEXT:    fsgnj.h a0, a1, a0
 ; RV32IZHINX-NEXT:  .LBB17_2:
+; RV32IZHINX-NEXT:    addi sp, sp, -16
+; RV32IZHINX-NEXT:    sw ra, 12(sp) # 4-byte Folded Spill
+; RV32IZHINX-NEXT:    sw s0, 8(sp) # 4-byte Folded Spill
+; RV32IZHINX-NEXT:    sw s1, 4(sp) # 4-byte Folded Spill
+; RV32IZHINX-NEXT:    sw s2, 0(sp) # 4-byte Folded Spill
 ; RV32IZHINX-NEXT:    fcvt.s.h s0, a0
 ; RV32IZHINX-NEXT:    lui a0, 913408
 ; RV32IZHINX-NEXT:    fle.s s1, a0, s0
 ; RV32IZHINX-NEXT:    neg s2, s1
 ; RV32IZHINX-NEXT:    mv a0, s0
 ; RV32IZHINX-NEXT:    call __fixsfdi@plt
 ; RV32IZHINX-NEXT:    lui a2, %hi(.LCPI17_1)
 ; RV32IZHINX-NEXT:    lw a2, %lo(.LCPI17_1)(a2)
 ; RV32IZHINX-NEXT:    and a0, s2, a0
 ; RV32IZHINX-NEXT:    flt.s a4, a2, s0
 ; RV32IZHINX-NEXT:    neg a2, a4
 ; RV32IZHINX-NEXT:    or a0, a2, a0
 ; RV32IZHINX-NEXT:    feq.s a2, s0, s0
 ; RV32IZHINX-NEXT:    neg a2, a2
 ; RV32IZHINX-NEXT:    lui a5, 524288
 ; RV32IZHINX-NEXT:    lui a3, 524288
 ; RV32IZHINX-NEXT:    beqz s1, .LBB17_4
 ; RV32IZHINX-NEXT:  # %bb.3:
 ; RV32IZHINX-NEXT:    mv a3, a1
 ; RV32IZHINX-NEXT:  .LBB17_4:
 ; RV32IZHINX-NEXT:    and a0, a2, a0
-; RV32IZHINX-NEXT:    beqz a4, .LBB17_6
-; RV32IZHINX-NEXT:  # %bb.5:
-; RV32IZHINX-NEXT:    addi a3, a5, -1
-; RV32IZHINX-NEXT:  .LBB17_6:
-; RV32IZHINX-NEXT:    and a1, a2, a3
 ; RV32IZHINX-NEXT:    lw ra, 12(sp) # 4-byte Folded Reload
 ; RV32IZHINX-NEXT:    lw s0, 8(sp) # 4-byte Folded Reload
 ; RV32IZHINX-NEXT:    lw s1, 4(sp) # 4-byte Folded Reload
 ; RV32IZHINX-NEXT:    lw s2, 0(sp) # 4-byte Folded Reload
 ; RV32IZHINX-NEXT:    addi sp, sp, 16
+; RV32IZHINX-NEXT:    beqz a4, .LBB17_6
+; RV32IZHINX-NEXT:  # %bb.5:
+; RV32IZHINX-NEXT:    addi a3, a5, -1
+; RV32IZHINX-NEXT:  .LBB17_6:
+; RV32IZHINX-NEXT:    and a1, a2, a3
 ; RV32IZHINX-NEXT:    ret
 ;
 ; RV64IZHINX-LABEL: test_roundeven_si64:
 ; RV64IZHINX:       # %bb.0:
 ; RV64IZHINX-NEXT:    lui a1, %hi(.LCPI17_0)
 ; RV64IZHINX-NEXT:    lh a1, %lo(.LCPI17_0)(a1)
 ; RV64IZHINX-NEXT:    fabs.h a2, a0
 ; RV64IZHINX-NEXT:    flt.h a1, a2, a1
; RV64IZHINXMIN-NEXT:    ret
   %a = call half @llvm.roundeven.f16(half %x)
   %b = call i32 @llvm.fptoui.sat.i32.f16(half %a)
   ret i32 %b
 }
 
 define i64 @test_roundeven_ui64(half %x) nounwind {
 ; RV32IZFH-LABEL: test_roundeven_ui64:
 ; RV32IZFH:       # %bb.0:
-; RV32IZFH-NEXT:    addi sp, sp, -16
-; RV32IZFH-NEXT:    sw ra, 12(sp) # 4-byte Folded Spill
-; RV32IZFH-NEXT:    sw s0, 8(sp) # 4-byte Folded Spill
-; RV32IZFH-NEXT:    fsw fs0, 4(sp) # 4-byte Folded Spill
 ; RV32IZFH-NEXT:    lui a0, %hi(.LCPI19_0)
 ; RV32IZFH-NEXT:    flh fa5, %lo(.LCPI19_0)(a0)
 ; RV32IZFH-NEXT:    fabs.h fa4, fa0
 ; RV32IZFH-NEXT:    flt.h a0, fa4, fa5
 ; RV32IZFH-NEXT:    beqz a0, .LBB19_2
 ; RV32IZFH-NEXT:  # %bb.1:
 ; RV32IZFH-NEXT:    fcvt.w.h a0, fa0, rne
 ; RV32IZFH-NEXT:    fcvt.h.w fa5, a0, rne
 ; RV32IZFH-NEXT:    fsgnj.h fa0, fa5, fa0
 ; RV32IZFH-NEXT:  .LBB19_2:
+; RV32IZFH-NEXT:    addi sp, sp, -16
+; RV32IZFH-NEXT:    sw ra, 12(sp) # 4-byte Folded Spill
+; RV32IZFH-NEXT:    sw s0, 8(sp) # 4-byte Folded Spill
+; RV32IZFH-NEXT:    fsw fs0, 4(sp) # 4-byte Folded Spill
 ; RV32IZFH-NEXT:    fcvt.s.h fs0, fa0
 ; RV32IZFH-NEXT:    fmv.w.x fa5, zero
 ; RV32IZFH-NEXT:    fle.s a0, fa5, fs0
 ; RV32IZFH-NEXT:    neg s0, a0
 ; RV32IZFH-NEXT:    fmv.s fa0, fs0
 ; RV32IZFH-NEXT:    call __fixunssfdi@plt
 ; RV32IZFH-NEXT:    lui a2, %hi(.LCPI19_1)
 ; RV32IZFH-NEXT:    flw fa5, %lo(.LCPI19_1)(a2)

 ; RV64IZFH-NEXT:    feq.h a1, fa0, fa0
 ; RV64IZFH-NEXT:    seqz a1, a1
 ; RV64IZFH-NEXT:    addi a1, a1, -1
 ; RV64IZFH-NEXT:    and a0, a1, a0
 ; RV64IZFH-NEXT:    ret
 ;
 ; RV32IZHINX-LABEL: test_roundeven_ui64:
 ; RV32IZHINX:       # %bb.0:
-; RV32IZHINX-NEXT:    addi sp, sp, -16
-; RV32IZHINX-NEXT:    sw ra, 12(sp) # 4-byte Folded Spill
-; RV32IZHINX-NEXT:    sw s0, 8(sp) # 4-byte Folded Spill
-; RV32IZHINX-NEXT:    sw s1, 4(sp) # 4-byte Folded Spill
 ; RV32IZHINX-NEXT:    lui a1, %hi(.LCPI19_0)
 ; RV32IZHINX-NEXT:    lh a1, %lo(.LCPI19_0)(a1)
 ; RV32IZHINX-NEXT:    fabs.h a2, a0
 ; RV32IZHINX-NEXT:    flt.h a1, a2, a1
 ; RV32IZHINX-NEXT:    beqz a1, .LBB19_2
 ; RV32IZHINX-NEXT:  # %bb.1:
 ; RV32IZHINX-NEXT:    fcvt.w.h a1, a0, rne
 ; RV32IZHINX-NEXT:    fcvt.h.w a1, a1, rne
 ; RV32IZHINX-NEXT:    fsgnj.h a0, a1, a0
 ; RV32IZHINX-NEXT:  .LBB19_2:
+; RV32IZHINX-NEXT:    addi sp, sp, -16
+; RV32IZHINX-NEXT:    sw ra, 12(sp) # 4-byte Folded Spill
+; RV32IZHINX-NEXT:    sw s0, 8(sp) # 4-byte Folded Spill
+; RV32IZHINX-NEXT:    sw s1, 4(sp) # 4-byte Folded Spill
 ; RV32IZHINX-NEXT:    fcvt.s.h s0, a0
 ; RV32IZHINX-NEXT:    fle.s a0, zero, s0
 ; RV32IZHINX-NEXT:    neg s1, a0
 ; RV32IZHINX-NEXT:    mv a0, s0
 ; RV32IZHINX-NEXT:    call __fixunssfdi@plt
 ; RV32IZHINX-NEXT:    lui a2, %hi(.LCPI19_1)
 ; RV32IZHINX-NEXT:    lw a2, %lo(.LCPI19_1)(a2)
 ; RV32IZHINX-NEXT:    and a0, s1, a0

llvm/test/CodeGen/RISCV/half-round-conv.ll

; CHECKIZHINXMIN-NEXT:    ret
   %a = call half @llvm.floor.f16(half %x)
   %b = fptosi half %a to i32
   ret i32 %b
 }
 
 define i64 @test_floor_si64(half %x) {
 ; RV32IZFH-LABEL: test_floor_si64:
 ; RV32IZFH:       # %bb.0:
-; RV32IZFH-NEXT:    addi sp, sp, -16
-; RV32IZFH-NEXT:    .cfi_def_cfa_offset 16
-; RV32IZFH-NEXT:    sw ra, 12(sp) # 4-byte Folded Spill
-; RV32IZFH-NEXT:    .cfi_offset ra, -4
 ; RV32IZFH-NEXT:    lui a0, %hi(.LCPI3_0)
 ; RV32IZFH-NEXT:    flh fa5, %lo(.LCPI3_0)(a0)
 ; RV32IZFH-NEXT:    fabs.h fa4, fa0
 ; RV32IZFH-NEXT:    flt.h a0, fa4, fa5
 ; RV32IZFH-NEXT:    beqz a0, .LBB3_2
 ; RV32IZFH-NEXT:  # %bb.1:
 ; RV32IZFH-NEXT:    fcvt.w.h a0, fa0, rdn
 ; RV32IZFH-NEXT:    fcvt.h.w fa5, a0, rdn
 ; RV32IZFH-NEXT:    fsgnj.h fa0, fa5, fa0
 ; RV32IZFH-NEXT:  .LBB3_2:
+; RV32IZFH-NEXT:    addi sp, sp, -16
+; RV32IZFH-NEXT:    .cfi_def_cfa_offset 16
+; RV32IZFH-NEXT:    sw ra, 12(sp) # 4-byte Folded Spill
+; RV32IZFH-NEXT:    .cfi_offset ra, -4
 ; RV32IZFH-NEXT:    call __fixhfdi@plt
 ; RV32IZFH-NEXT:    lw ra, 12(sp) # 4-byte Folded Reload
 ; RV32IZFH-NEXT:    addi sp, sp, 16
 ; RV32IZFH-NEXT:    ret
 ;
 ; RV64IZFH-LABEL: test_floor_si64:
 ; RV64IZFH:       # %bb.0:
 ; RV64IZFH-NEXT:    fcvt.l.h a0, fa0, rdn
 ; RV64IZFH-NEXT:    ret
 ;
 ; RV32IZHINX-LABEL: test_floor_si64:
 ; RV32IZHINX:       # %bb.0:
-; RV32IZHINX-NEXT:    addi sp, sp, -16
-; RV32IZHINX-NEXT:    .cfi_def_cfa_offset 16
-; RV32IZHINX-NEXT:    sw ra, 12(sp) # 4-byte Folded Spill
-; RV32IZHINX-NEXT:    .cfi_offset ra, -4
 ; RV32IZHINX-NEXT:    lui a1, %hi(.LCPI3_0)
 ; RV32IZHINX-NEXT:    lh a1, %lo(.LCPI3_0)(a1)
 ; RV32IZHINX-NEXT:    fabs.h a2, a0
 ; RV32IZHINX-NEXT:    flt.h a1, a2, a1
 ; RV32IZHINX-NEXT:    beqz a1, .LBB3_2
 ; RV32IZHINX-NEXT:  # %bb.1:
 ; RV32IZHINX-NEXT:    fcvt.w.h a1, a0, rdn
 ; RV32IZHINX-NEXT:    fcvt.h.w a1, a1, rdn
 ; RV32IZHINX-NEXT:    fsgnj.h a0, a1, a0
 ; RV32IZHINX-NEXT:  .LBB3_2:
+; RV32IZHINX-NEXT:    addi sp, sp, -16
+; RV32IZHINX-NEXT:    .cfi_def_cfa_offset 16
+; RV32IZHINX-NEXT:    sw ra, 12(sp) # 4-byte Folded Spill
+; RV32IZHINX-NEXT:    .cfi_offset ra, -4
 ; RV32IZHINX-NEXT:    call __fixhfdi@plt
 ; RV32IZHINX-NEXT:    lw ra, 12(sp) # 4-byte Folded Reload
 ; RV32IZHINX-NEXT:    addi sp, sp, 16
 ; RV32IZHINX-NEXT:    ret
 ;
 ; RV64IZHINX-LABEL: test_floor_si64:
 ; RV64IZHINX:       # %bb.0:
 ; RV64IZHINX-NEXT:    lui a1, %hi(.LCPI3_0)
; CHECKIZHINXMIN-NEXT:    ret
   %a = call half @llvm.floor.f16(half %x)
   %b = fptoui half %a to i32
   ret i32 %b
 }
 
 define i64 @test_floor_ui64(half %x) {
 ; RV32IZFH-LABEL: test_floor_ui64:
 ; RV32IZFH:       # %bb.0:
-; RV32IZFH-NEXT:    addi sp, sp, -16
-; RV32IZFH-NEXT:    .cfi_def_cfa_offset 16
-; RV32IZFH-NEXT:    sw ra, 12(sp) # 4-byte Folded Spill
-; RV32IZFH-NEXT:    .cfi_offset ra, -4
 ; RV32IZFH-NEXT:    lui a0, %hi(.LCPI7_0)
 ; RV32IZFH-NEXT:    flh fa5, %lo(.LCPI7_0)(a0)
 ; RV32IZFH-NEXT:    fabs.h fa4, fa0
 ; RV32IZFH-NEXT:    flt.h a0, fa4, fa5
 ; RV32IZFH-NEXT:    beqz a0, .LBB7_2
 ; RV32IZFH-NEXT:  # %bb.1:
 ; RV32IZFH-NEXT:    fcvt.w.h a0, fa0, rdn
 ; RV32IZFH-NEXT:    fcvt.h.w fa5, a0, rdn
 ; RV32IZFH-NEXT:    fsgnj.h fa0, fa5, fa0
 ; RV32IZFH-NEXT:  .LBB7_2:
+; RV32IZFH-NEXT:    addi sp, sp, -16
+; RV32IZFH-NEXT:    .cfi_def_cfa_offset 16
+; RV32IZFH-NEXT:    sw ra, 12(sp) # 4-byte Folded Spill
+; RV32IZFH-NEXT:    .cfi_offset ra, -4
 ; RV32IZFH-NEXT:    call __fixunshfdi@plt
 ; RV32IZFH-NEXT:    lw ra, 12(sp) # 4-byte Folded Reload
 ; RV32IZFH-NEXT:    addi sp, sp, 16
 ; RV32IZFH-NEXT:    ret
 ;
 ; RV64IZFH-LABEL: test_floor_ui64:
 ; RV64IZFH:       # %bb.0:
 ; RV64IZFH-NEXT:    fcvt.lu.h a0, fa0, rdn
 ; RV64IZFH-NEXT:    ret
 ;
 ; RV32IZHINX-LABEL: test_floor_ui64:
 ; RV32IZHINX:       # %bb.0:
-; RV32IZHINX-NEXT:    addi sp, sp, -16
-; RV32IZHINX-NEXT:    .cfi_def_cfa_offset 16
-; RV32IZHINX-NEXT:    sw ra, 12(sp) # 4-byte Folded Spill
-; RV32IZHINX-NEXT:    .cfi_offset ra, -4
 ; RV32IZHINX-NEXT:    lui a1, %hi(.LCPI7_0)
 ; RV32IZHINX-NEXT:    lh a1, %lo(.LCPI7_0)(a1)
 ; RV32IZHINX-NEXT:    fabs.h a2, a0
 ; RV32IZHINX-NEXT:    flt.h a1, a2, a1
 ; RV32IZHINX-NEXT:    beqz a1, .LBB7_2
 ; RV32IZHINX-NEXT:  # %bb.1:
 ; RV32IZHINX-NEXT:    fcvt.w.h a1, a0, rdn
 ; RV32IZHINX-NEXT:    fcvt.h.w a1, a1, rdn
 ; RV32IZHINX-NEXT:    fsgnj.h a0, a1, a0
 ; RV32IZHINX-NEXT:  .LBB7_2:
+; RV32IZHINX-NEXT:    addi sp, sp, -16
+; RV32IZHINX-NEXT:    .cfi_def_cfa_offset 16
+; RV32IZHINX-NEXT:    sw ra, 12(sp) # 4-byte Folded Spill
+; RV32IZHINX-NEXT:    .cfi_offset ra, -4
 ; RV32IZHINX-NEXT:    call __fixunshfdi@plt
 ; RV32IZHINX-NEXT:    lw ra, 12(sp) # 4-byte Folded Reload
 ; RV32IZHINX-NEXT:    addi sp, sp, 16
 ; RV32IZHINX-NEXT:    ret
 ;
 ; RV64IZHINX-LABEL: test_floor_ui64:
 ; RV64IZHINX:       # %bb.0:
 ; RV64IZHINX-NEXT:    lui a1, %hi(.LCPI7_0)
; CHECKIZHINXMIN-NEXT:    ret
   %a = call half @llvm.ceil.f16(half %x)
   %b = fptosi half %a to i32
   ret i32 %b
 }
 
 define i64 @test_ceil_si64(half %x) {
 ; RV32IZFH-LABEL: test_ceil_si64:
 ; RV32IZFH:       # %bb.0:
-; RV32IZFH-NEXT:    addi sp, sp, -16
-; RV32IZFH-NEXT:    .cfi_def_cfa_offset 16
-; RV32IZFH-NEXT:    sw ra, 12(sp) # 4-byte Folded Spill
-; RV32IZFH-NEXT:    .cfi_offset ra, -4
 ; RV32IZFH-NEXT:    lui a0, %hi(.LCPI11_0)
 ; RV32IZFH-NEXT:    flh fa5, %lo(.LCPI11_0)(a0)
 ; RV32IZFH-NEXT:    fabs.h fa4, fa0
 ; RV32IZFH-NEXT:    flt.h a0, fa4, fa5
 ; RV32IZFH-NEXT:    beqz a0, .LBB11_2
 ; RV32IZFH-NEXT:  # %bb.1:
 ; RV32IZFH-NEXT:    fcvt.w.h a0, fa0, rup
 ; RV32IZFH-NEXT:    fcvt.h.w fa5, a0, rup
 ; RV32IZFH-NEXT:    fsgnj.h fa0, fa5, fa0
 ; RV32IZFH-NEXT:  .LBB11_2:
+; RV32IZFH-NEXT:    addi sp, sp, -16
+; RV32IZFH-NEXT:    .cfi_def_cfa_offset 16
+; RV32IZFH-NEXT:    sw ra, 12(sp) # 4-byte Folded Spill
+; RV32IZFH-NEXT:    .cfi_offset ra, -4
 ; RV32IZFH-NEXT:    call __fixhfdi@plt
 ; RV32IZFH-NEXT:    lw ra, 12(sp) # 4-byte Folded Reload
 ; RV32IZFH-NEXT:    addi sp, sp, 16
 ; RV32IZFH-NEXT:    ret
 ;
 ; RV64IZFH-LABEL: test_ceil_si64:
 ; RV64IZFH:       # %bb.0:
 ; RV64IZFH-NEXT:    fcvt.l.h a0, fa0, rup
 ; RV64IZFH-NEXT:    ret
 ;
 ; RV32IZHINX-LABEL: test_ceil_si64:
 ; RV32IZHINX:       # %bb.0:
-; RV32IZHINX-NEXT:    addi sp, sp, -16
-; RV32IZHINX-NEXT:    .cfi_def_cfa_offset 16
-; RV32IZHINX-NEXT:    sw ra, 12(sp) # 4-byte Folded Spill
-; RV32IZHINX-NEXT:    .cfi_offset ra, -4
 ; RV32IZHINX-NEXT:    lui a1, %hi(.LCPI11_0)
 ; RV32IZHINX-NEXT:    lh a1, %lo(.LCPI11_0)(a1)
 ; RV32IZHINX-NEXT:    fabs.h a2, a0
 ; RV32IZHINX-NEXT:    flt.h a1, a2, a1
 ; RV32IZHINX-NEXT:    beqz a1, .LBB11_2
 ; RV32IZHINX-NEXT:  # %bb.1:
 ; RV32IZHINX-NEXT:    fcvt.w.h a1, a0, rup
 ; RV32IZHINX-NEXT:    fcvt.h.w a1, a1, rup
 ; RV32IZHINX-NEXT:    fsgnj.h a0, a1, a0
 ; RV32IZHINX-NEXT:  .LBB11_2:
+; RV32IZHINX-NEXT:    addi sp, sp, -16
+; RV32IZHINX-NEXT:    .cfi_def_cfa_offset 16
+; RV32IZHINX-NEXT:    sw ra, 12(sp) # 4-byte Folded Spill
+; RV32IZHINX-NEXT:    .cfi_offset ra, -4
 ; RV32IZHINX-NEXT:    call __fixhfdi@plt
 ; RV32IZHINX-NEXT:    lw ra, 12(sp) # 4-byte Folded Reload
 ; RV32IZHINX-NEXT:    addi sp, sp, 16
 ; RV32IZHINX-NEXT:    ret
 ;
 ; RV64IZHINX-LABEL: test_ceil_si64:
 ; RV64IZHINX:       # %bb.0:
 ; RV64IZHINX-NEXT:    lui a1, %hi(.LCPI11_0)
; CHECKIZHINXMIN-NEXT:    ret
   %a = call half @llvm.ceil.f16(half %x)
   %b = fptoui half %a to i32
   ret i32 %b
 }
 
 define i64 @test_ceil_ui64(half %x) {
 ; RV32IZFH-LABEL: test_ceil_ui64:
 ; RV32IZFH:       # %bb.0:
-; RV32IZFH-NEXT:    addi sp, sp, -16
-; RV32IZFH-NEXT:    .cfi_def_cfa_offset 16
-; RV32IZFH-NEXT:    sw ra, 12(sp) # 4-byte Folded Spill
-; RV32IZFH-NEXT:    .cfi_offset ra, -4
 ; RV32IZFH-NEXT:    lui a0, %hi(.LCPI15_0)
 ; RV32IZFH-NEXT:    flh fa5, %lo(.LCPI15_0)(a0)
 ; RV32IZFH-NEXT:    fabs.h fa4, fa0
 ; RV32IZFH-NEXT:    flt.h a0, fa4, fa5
 ; RV32IZFH-NEXT:    beqz a0, .LBB15_2
 ; RV32IZFH-NEXT:  # %bb.1:
 ; RV32IZFH-NEXT:    fcvt.w.h a0, fa0, rup
 ; RV32IZFH-NEXT:    fcvt.h.w fa5, a0, rup
 ; RV32IZFH-NEXT:    fsgnj.h fa0, fa5, fa0
 ; RV32IZFH-NEXT:  .LBB15_2:
+; RV32IZFH-NEXT:    addi sp, sp, -16
+; RV32IZFH-NEXT:    .cfi_def_cfa_offset 16
+; RV32IZFH-NEXT:    sw ra, 12(sp) # 4-byte Folded Spill
+; RV32IZFH-NEXT:    .cfi_offset ra, -4
 ; RV32IZFH-NEXT:    call __fixunshfdi@plt
 ; RV32IZFH-NEXT:    lw ra, 12(sp) # 4-byte Folded Reload
 ; RV32IZFH-NEXT:    addi sp, sp, 16
 ; RV32IZFH-NEXT:    ret
 ;
 ; RV64IZFH-LABEL: test_ceil_ui64:
 ; RV64IZFH:       # %bb.0:
 ; RV64IZFH-NEXT:    fcvt.lu.h a0, fa0, rup
 ; RV64IZFH-NEXT:    ret
 ;
 ; RV32IZHINX-LABEL: test_ceil_ui64:
 ; RV32IZHINX:       # %bb.0:
-; RV32IZHINX-NEXT:    addi sp, sp, -16
-; RV32IZHINX-NEXT:    .cfi_def_cfa_offset 16
-; RV32IZHINX-NEXT:    sw ra, 12(sp) # 4-byte Folded Spill
-; RV32IZHINX-NEXT:    .cfi_offset ra, -4
 ; RV32IZHINX-NEXT:    lui a1, %hi(.LCPI15_0)
 ; RV32IZHINX-NEXT:    lh a1, %lo(.LCPI15_0)(a1)
 ; RV32IZHINX-NEXT:    fabs.h a2, a0
 ; RV32IZHINX-NEXT:    flt.h a1, a2, a1
 ; RV32IZHINX-NEXT:    beqz a1, .LBB15_2
 ; RV32IZHINX-NEXT:  # %bb.1:
 ; RV32IZHINX-NEXT:    fcvt.w.h a1, a0, rup
 ; RV32IZHINX-NEXT:    fcvt.h.w a1, a1, rup
 ; RV32IZHINX-NEXT:    fsgnj.h a0, a1, a0
 ; RV32IZHINX-NEXT:  .LBB15_2:
+; RV32IZHINX-NEXT:    addi sp, sp, -16
+; RV32IZHINX-NEXT:    .cfi_def_cfa_offset 16
+; RV32IZHINX-NEXT:    sw ra, 12(sp) # 4-byte Folded Spill
+; RV32IZHINX-NEXT:    .cfi_offset ra, -4
 ; RV32IZHINX-NEXT:    call __fixunshfdi@plt
 ; RV32IZHINX-NEXT:    lw ra, 12(sp) # 4-byte Folded Reload
 ; RV32IZHINX-NEXT:    addi sp, sp, 16
 ; RV32IZHINX-NEXT:    ret
 ;
 ; RV64IZHINX-LABEL: test_ceil_ui64:
 ; RV64IZHINX:       # %bb.0:
 ; RV64IZHINX-NEXT:    lui a1, %hi(.LCPI15_0)
; CHECKIZHINXMIN-NEXT:    ret
   %a = call half @llvm.trunc.f16(half %x)
   %b = fptosi half %a to i32
   ret i32 %b
 }
 
 define i64 @test_trunc_si64(half %x) {
 ; RV32IZFH-LABEL: test_trunc_si64:
 ; RV32IZFH:       # %bb.0:
-; RV32IZFH-NEXT:    addi sp, sp, -16
-; RV32IZFH-NEXT:    .cfi_def_cfa_offset 16
-; RV32IZFH-NEXT:    sw ra, 12(sp) # 4-byte Folded Spill
-; RV32IZFH-NEXT:    .cfi_offset ra, -4
 ; RV32IZFH-NEXT:    lui a0, %hi(.LCPI19_0)
 ; RV32IZFH-NEXT:    flh fa5, %lo(.LCPI19_0)(a0)
 ; RV32IZFH-NEXT:    fabs.h fa4, fa0
 ; RV32IZFH-NEXT:    flt.h a0, fa4, fa5
 ; RV32IZFH-NEXT:    beqz a0, .LBB19_2
 ; RV32IZFH-NEXT:  # %bb.1:
 ; RV32IZFH-NEXT:    fcvt.w.h a0, fa0, rtz
 ; RV32IZFH-NEXT:    fcvt.h.w fa5, a0, rtz
 ; RV32IZFH-NEXT:    fsgnj.h fa0, fa5, fa0
 ; RV32IZFH-NEXT:  .LBB19_2:
+; RV32IZFH-NEXT:    addi sp, sp, -16
+; RV32IZFH-NEXT:    .cfi_def_cfa_offset 16
+; RV32IZFH-NEXT:    sw ra, 12(sp) # 4-byte Folded Spill
+; RV32IZFH-NEXT:    .cfi_offset ra, -4
 ; RV32IZFH-NEXT:    call __fixhfdi@plt
 ; RV32IZFH-NEXT:    lw ra, 12(sp) # 4-byte Folded Reload
 ; RV32IZFH-NEXT:    addi sp, sp, 16
 ; RV32IZFH-NEXT:    ret
 ;
 ; RV64IZFH-LABEL: test_trunc_si64:
 ; RV64IZFH:       # %bb.0:
 ; RV64IZFH-NEXT:    fcvt.l.h a0, fa0, rtz
 ; RV64IZFH-NEXT:    ret
 ;
 ; RV32IZHINX-LABEL: test_trunc_si64:
 ; RV32IZHINX:       # %bb.0:
-; RV32IZHINX-NEXT:    addi sp, sp, -16
-; RV32IZHINX-NEXT:    .cfi_def_cfa_offset 16
-; RV32IZHINX-NEXT:    sw ra, 12(sp) # 4-byte Folded Spill
-; RV32IZHINX-NEXT:    .cfi_offset ra, -4
 ; RV32IZHINX-NEXT:    lui a1, %hi(.LCPI19_0)
 ; RV32IZHINX-NEXT:    lh a1, %lo(.LCPI19_0)(a1)
 ; RV32IZHINX-NEXT:    fabs.h a2, a0
 ; RV32IZHINX-NEXT:    flt.h a1, a2, a1
 ; RV32IZHINX-NEXT:    beqz a1, .LBB19_2
 ; RV32IZHINX-NEXT:  # %bb.1:
 ; RV32IZHINX-NEXT:    fcvt.w.h a1, a0, rtz
 ; RV32IZHINX-NEXT:    fcvt.h.w a1, a1, rtz
 ; RV32IZHINX-NEXT:    fsgnj.h a0, a1, a0
 ; RV32IZHINX-NEXT:  .LBB19_2:
+; RV32IZHINX-NEXT:    addi sp, sp, -16
+; RV32IZHINX-NEXT:    .cfi_def_cfa_offset 16
+; RV32IZHINX-NEXT:    sw ra, 12(sp) # 4-byte Folded Spill
+; RV32IZHINX-NEXT:    .cfi_offset ra, -4
 ; RV32IZHINX-NEXT:    call __fixhfdi@plt
 ; RV32IZHINX-NEXT:    lw ra, 12(sp) # 4-byte Folded Reload
 ; RV32IZHINX-NEXT:    addi sp, sp, 16
 ; RV32IZHINX-NEXT:    ret
 ;
 ; RV64IZHINX-LABEL: test_trunc_si64:
 ; RV64IZHINX:       # %bb.0:
 ; RV64IZHINX-NEXT:    lui a1, %hi(.LCPI19_0)
; CHECKIZHINXMIN-NEXT:    ret
   %a = call half @llvm.trunc.f16(half %x)
   %b = fptoui half %a to i32
   ret i32 %b
 }
 
 define i64 @test_trunc_ui64(half %x) {
 ; RV32IZFH-LABEL: test_trunc_ui64:
 ; RV32IZFH:       # %bb.0:
-; RV32IZFH-NEXT:    addi sp, sp, -16
-; RV32IZFH-NEXT:    .cfi_def_cfa_offset 16
-; RV32IZFH-NEXT:    sw ra, 12(sp) # 4-byte Folded Spill
-; RV32IZFH-NEXT:    .cfi_offset ra, -4
 ; RV32IZFH-NEXT:    lui a0, %hi(.LCPI23_0)
 ; RV32IZFH-NEXT:    flh fa5, %lo(.LCPI23_0)(a0)
 ; RV32IZFH-NEXT:    fabs.h fa4, fa0
 ; RV32IZFH-NEXT:    flt.h a0, fa4, fa5
 ; RV32IZFH-NEXT:    beqz a0, .LBB23_2
 ; RV32IZFH-NEXT:  # %bb.1:
 ; RV32IZFH-NEXT:    fcvt.w.h a0, fa0, rtz
 ; RV32IZFH-NEXT:    fcvt.h.w fa5, a0, rtz
 ; RV32IZFH-NEXT:    fsgnj.h fa0, fa5, fa0
 ; RV32IZFH-NEXT:  .LBB23_2:
+; RV32IZFH-NEXT:    addi sp, sp, -16
+; RV32IZFH-NEXT:    .cfi_def_cfa_offset 16
+; RV32IZFH-NEXT:    sw ra, 12(sp) # 4-byte Folded Spill
+; RV32IZFH-NEXT:    .cfi_offset ra, -4
 ; RV32IZFH-NEXT:    call __fixunshfdi@plt
 ; RV32IZFH-NEXT:    lw ra, 12(sp) # 4-byte Folded Reload
 ; RV32IZFH-NEXT:    addi sp, sp, 16
 ; RV32IZFH-NEXT:    ret
 ;
 ; RV64IZFH-LABEL: test_trunc_ui64:
 ; RV64IZFH:       # %bb.0:
 ; RV64IZFH-NEXT:    fcvt.lu.h a0, fa0, rtz
 ; RV64IZFH-NEXT:    ret
 ;
 ; RV32IZHINX-LABEL: test_trunc_ui64:
 ; RV32IZHINX:       # %bb.0:
-; RV32IZHINX-NEXT:    addi sp, sp, -16
-; RV32IZHINX-NEXT:    .cfi_def_cfa_offset 16
-; RV32IZHINX-NEXT:    sw ra, 12(sp) # 4-byte Folded Spill
-; RV32IZHINX-NEXT:    .cfi_offset ra, -4
 ; RV32IZHINX-NEXT:    lui a1, %hi(.LCPI23_0)
 ; RV32IZHINX-NEXT:    lh a1, %lo(.LCPI23_0)(a1)
 ; RV32IZHINX-NEXT:    fabs.h a2, a0
 ; RV32IZHINX-NEXT:    flt.h a1, a2, a1
 ; RV32IZHINX-NEXT:    beqz a1, .LBB23_2
 ; RV32IZHINX-NEXT:  # %bb.1:
 ; RV32IZHINX-NEXT:    fcvt.w.h a1, a0, rtz
 ; RV32IZHINX-NEXT:    fcvt.h.w a1, a1, rtz
 ; RV32IZHINX-NEXT:    fsgnj.h a0, a1, a0
 ; RV32IZHINX-NEXT:  .LBB23_2:
+; RV32IZHINX-NEXT:    addi sp, sp, -16
+; RV32IZHINX-NEXT:    .cfi_def_cfa_offset 16
+; RV32IZHINX-NEXT:    sw ra, 12(sp) # 4-byte Folded Spill
+; RV32IZHINX-NEXT:    .cfi_offset ra, -4
 ; RV32IZHINX-NEXT:    call __fixunshfdi@plt
 ; RV32IZHINX-NEXT:    lw ra, 12(sp) # 4-byte Folded Reload
 ; RV32IZHINX-NEXT:    addi sp, sp, 16
 ; RV32IZHINX-NEXT:    ret
 ;
 ; RV64IZHINX-LABEL: test_trunc_ui64:
 ; RV64IZHINX:       # %bb.0:
 ; RV64IZHINX-NEXT:    lui a1, %hi(.LCPI23_0)
; CHECKIZHINXMIN-NEXT:    ret
   %a = call half @llvm.round.f16(half %x)
   %b = fptosi half %a to i32
   ret i32 %b
 }
 
 define i64 @test_round_si64(half %x) {
 ; RV32IZFH-LABEL: test_round_si64:
 ; RV32IZFH:       # %bb.0:
-; RV32IZFH-NEXT:    addi sp, sp, -16
-; RV32IZFH-NEXT:    .cfi_def_cfa_offset 16
-; RV32IZFH-NEXT:    sw ra, 12(sp) # 4-byte Folded Spill
-; RV32IZFH-NEXT:    .cfi_offset ra, -4
 ; RV32IZFH-NEXT:    lui a0, %hi(.LCPI27_0)
 ; RV32IZFH-NEXT:    flh fa5, %lo(.LCPI27_0)(a0)
 ; RV32IZFH-NEXT:    fabs.h fa4, fa0
 ; RV32IZFH-NEXT:    flt.h a0, fa4, fa5
 ; RV32IZFH-NEXT:    beqz a0, .LBB27_2
 ; RV32IZFH-NEXT:  # %bb.1:
 ; RV32IZFH-NEXT:    fcvt.w.h a0, fa0, rmm
 ; RV32IZFH-NEXT:    fcvt.h.w fa5, a0, rmm
 ; RV32IZFH-NEXT:    fsgnj.h fa0, fa5, fa0
 ; RV32IZFH-NEXT:  .LBB27_2:
+; RV32IZFH-NEXT:    addi sp, sp, -16
+; RV32IZFH-NEXT:    .cfi_def_cfa_offset 16
+; RV32IZFH-NEXT:    sw ra, 12(sp) # 4-byte Folded Spill
+; RV32IZFH-NEXT:    .cfi_offset ra, -4
 ; RV32IZFH-NEXT:    call __fixhfdi@plt
 ; RV32IZFH-NEXT:    lw ra, 12(sp) # 4-byte Folded Reload
 ; RV32IZFH-NEXT:    addi sp, sp, 16
 ; RV32IZFH-NEXT:    ret
 ;
 ; RV64IZFH-LABEL: test_round_si64:
 ; RV64IZFH:       # %bb.0:
 ; RV64IZFH-NEXT:    fcvt.l.h a0, fa0, rmm
 ; RV64IZFH-NEXT:    ret
 ;
 ; RV32IZHINX-LABEL: test_round_si64:
 ; RV32IZHINX:       # %bb.0:
-; RV32IZHINX-NEXT:    addi sp, sp, -16
-; RV32IZHINX-NEXT:    .cfi_def_cfa_offset 16
-; RV32IZHINX-NEXT:    sw ra, 12(sp) # 4-byte Folded Spill
-; RV32IZHINX-NEXT:    .cfi_offset ra, -4
 ; RV32IZHINX-NEXT:    lui a1, %hi(.LCPI27_0)
 ; RV32IZHINX-NEXT:    lh a1, %lo(.LCPI27_0)(a1)
 ; RV32IZHINX-NEXT:    fabs.h a2, a0
 ; RV32IZHINX-NEXT:    flt.h a1, a2, a1
 ; RV32IZHINX-NEXT:    beqz a1, .LBB27_2
 ; RV32IZHINX-NEXT:  # %bb.1:
 ; RV32IZHINX-NEXT:    fcvt.w.h a1, a0, rmm
 ; RV32IZHINX-NEXT:    fcvt.h.w a1, a1, rmm
 ; RV32IZHINX-NEXT:    fsgnj.h a0, a1, a0
 ; RV32IZHINX-NEXT:  .LBB27_2:
+; RV32IZHINX-NEXT:    addi sp, sp, -16
+; RV32IZHINX-NEXT:    .cfi_def_cfa_offset 16
+; RV32IZHINX-NEXT:    sw ra, 12(sp) # 4-byte Folded Spill
+; RV32IZHINX-NEXT:    .cfi_offset ra, -4
 ; RV32IZHINX-NEXT:    call __fixhfdi@plt
 ; RV32IZHINX-NEXT:    lw ra, 12(sp) # 4-byte Folded Reload
 ; RV32IZHINX-NEXT:    addi sp, sp, 16
 ; RV32IZHINX-NEXT:    ret
 ;
 ; RV64IZHINX-LABEL: test_round_si64:
 ; RV64IZHINX:       # %bb.0:
 ; RV64IZHINX-NEXT:    lui a1, %hi(.LCPI27_0)
; CHECKIZHINXMIN-NEXT:    ret
   %a = call half @llvm.round.f16(half %x)
   %b = fptoui half %a to i32
   ret i32 %b
 }
 
 define i64 @test_round_ui64(half %x) {
 ; RV32IZFH-LABEL: test_round_ui64:
 ; RV32IZFH:       # %bb.0:
-; RV32IZFH-NEXT:    addi sp, sp, -16
-; RV32IZFH-NEXT:    .cfi_def_cfa_offset 16
-; RV32IZFH-NEXT:    sw ra, 12(sp) # 4-byte Folded Spill
-; RV32IZFH-NEXT:    .cfi_offset ra, -4
 ; RV32IZFH-NEXT:    lui a0, %hi(.LCPI31_0)
 ; RV32IZFH-NEXT:    flh fa5, %lo(.LCPI31_0)(a0)
 ; RV32IZFH-NEXT:    fabs.h fa4, fa0
 ; RV32IZFH-NEXT:    flt.h a0, fa4, fa5
 ; RV32IZFH-NEXT:    beqz a0, .LBB31_2
 ; RV32IZFH-NEXT:  # %bb.1:
 ; RV32IZFH-NEXT:    fcvt.w.h a0, fa0, rmm
 ; RV32IZFH-NEXT:    fcvt.h.w fa5, a0, rmm
 ; RV32IZFH-NEXT:    fsgnj.h fa0, fa5, fa0
 ; RV32IZFH-NEXT:  .LBB31_2:
+; RV32IZFH-NEXT:    addi sp, sp, -16
+; RV32IZFH-NEXT:    .cfi_def_cfa_offset 16
+; RV32IZFH-NEXT:    sw ra, 12(sp) # 4-byte Folded Spill
+; RV32IZFH-NEXT:    .cfi_offset ra, -4
 ; RV32IZFH-NEXT:    call __fixunshfdi@plt
 ; RV32IZFH-NEXT:    lw ra, 12(sp) # 4-byte Folded Reload
 ; RV32IZFH-NEXT:    addi sp, sp, 16
 ; RV32IZFH-NEXT:    ret
 ;
 ; RV64IZFH-LABEL: test_round_ui64:
 ; RV64IZFH:       # %bb.0:
 ; RV64IZFH-NEXT:    fcvt.lu.h a0, fa0, rmm
 ; RV64IZFH-NEXT:    ret
 ;
 ; RV32IZHINX-LABEL: test_round_ui64:
 ; RV32IZHINX:       # %bb.0:
-; RV32IZHINX-NEXT:    addi sp, sp, -16
-; RV32IZHINX-NEXT:    .cfi_def_cfa_offset 16
-; RV32IZHINX-NEXT:    sw ra, 12(sp) # 4-byte Folded Spill
-; RV32IZHINX-NEXT:    .cfi_offset ra, -4
 ; RV32IZHINX-NEXT:    lui a1, %hi(.LCPI31_0)
 ; RV32IZHINX-NEXT:    lh a1, %lo(.LCPI31_0)(a1)
 ; RV32IZHINX-NEXT:    fabs.h a2, a0
 ; RV32IZHINX-NEXT:    flt.h a1, a2, a1
 ; RV32IZHINX-NEXT:    beqz a1, .LBB31_2
 ; RV32IZHINX-NEXT:  # %bb.1:
 ; RV32IZHINX-NEXT:    fcvt.w.h a1, a0, rmm
 ; RV32IZHINX-NEXT:    fcvt.h.w a1, a1, rmm
 ; RV32IZHINX-NEXT:    fsgnj.h a0, a1, a0
 ; RV32IZHINX-NEXT:  .LBB31_2:
+; RV32IZHINX-NEXT:    addi sp, sp, -16
+; RV32IZHINX-NEXT:    .cfi_def_cfa_offset 16
+; RV32IZHINX-NEXT:    sw ra, 12(sp) # 4-byte Folded Spill
+; RV32IZHINX-NEXT:    .cfi_offset ra, -4
 ; RV32IZHINX-NEXT:    call __fixunshfdi@plt
 ; RV32IZHINX-NEXT:    lw ra, 12(sp) # 4-byte Folded Reload
 ; RV32IZHINX-NEXT:    addi sp, sp, 16
 ; RV32IZHINX-NEXT:    ret
 ;
 ; RV64IZHINX-LABEL: test_round_ui64:
 ; RV64IZHINX:       # %bb.0:
 ; RV64IZHINX-NEXT:    lui a1, %hi(.LCPI31_0)
; CHECKIZHINXMIN-NEXT:    ret
   %a = call half @llvm.roundeven.f16(half %x)
   %b = fptosi half %a to i32
   ret i32 %b
 }
 
 define i64 @test_roundeven_si64(half %x) {
 ; RV32IZFH-LABEL: test_roundeven_si64:
 ; RV32IZFH:       # %bb.0:
-; RV32IZFH-NEXT:    addi sp, sp, -16
-; RV32IZFH-NEXT:    .cfi_def_cfa_offset 16
-; RV32IZFH-NEXT:    sw ra, 12(sp) # 4-byte Folded Spill
-; RV32IZFH-NEXT:    .cfi_offset ra, -4
 ; RV32IZFH-NEXT:    lui a0, %hi(.LCPI35_0)
 ; RV32IZFH-NEXT:    flh fa5, %lo(.LCPI35_0)(a0)
 ; RV32IZFH-NEXT:    fabs.h fa4, fa0
 ; RV32IZFH-NEXT:    flt.h a0, fa4, fa5
 ; RV32IZFH-NEXT:    beqz a0, .LBB35_2
 ; RV32IZFH-NEXT:  # %bb.1:
 ; RV32IZFH-NEXT:    fcvt.w.h a0, fa0, rne
 ; RV32IZFH-NEXT:    fcvt.h.w fa5, a0, rne
 ; RV32IZFH-NEXT:    fsgnj.h fa0, fa5, fa0
 ; RV32IZFH-NEXT:  .LBB35_2:
+; RV32IZFH-NEXT:    addi sp, sp, -16
+; RV32IZFH-NEXT:    .cfi_def_cfa_offset 16
+; RV32IZFH-NEXT:    sw ra, 12(sp) # 4-byte Folded Spill
+; RV32IZFH-NEXT:    .cfi_offset ra, -4
 ; RV32IZFH-NEXT:    call __fixhfdi@plt
 ; RV32IZFH-NEXT:    lw ra, 12(sp) # 4-byte Folded Reload
 ; RV32IZFH-NEXT:    addi sp, sp, 16
 ; RV32IZFH-NEXT:    ret
 ;
 ; RV64IZFH-LABEL: test_roundeven_si64:
 ; RV64IZFH:       # %bb.0:
 ; RV64IZFH-NEXT:    fcvt.l.h a0, fa0, rne
 ; RV64IZFH-NEXT:    ret
 ;
 ; RV32IZHINX-LABEL: test_roundeven_si64:
 ; RV32IZHINX:       # %bb.0:
-; RV32IZHINX-NEXT:    addi sp, sp, -16
-; RV32IZHINX-NEXT:    .cfi_def_cfa_offset 16
-; RV32IZHINX-NEXT:    sw ra, 12(sp) # 4-byte Folded Spill
-; RV32IZHINX-NEXT:    .cfi_offset ra, -4
 ; RV32IZHINX-NEXT:    lui a1, %hi(.LCPI35_0)
 ; RV32IZHINX-NEXT:    lh a1, %lo(.LCPI35_0)(a1)
 ; RV32IZHINX-NEXT:    fabs.h a2, a0
 ; RV32IZHINX-NEXT:    flt.h a1, a2, a1
 ; RV32IZHINX-NEXT:    beqz a1, .LBB35_2
 ; RV32IZHINX-NEXT:  # %bb.1:
 ; RV32IZHINX-NEXT:    fcvt.w.h a1, a0, rne
 ; RV32IZHINX-NEXT:    fcvt.h.w a1, a1, rne
 ; RV32IZHINX-NEXT:    fsgnj.h a0, a1, a0
 ; RV32IZHINX-NEXT:  .LBB35_2:
+; RV32IZHINX-NEXT:    addi sp, sp, -16
+; RV32IZHINX-NEXT:    .cfi_def_cfa_offset 16
+; RV32IZHINX-NEXT:    sw ra, 12(sp) # 4-byte Folded Spill
+; RV32IZHINX-NEXT:    .cfi_offset ra, -4
 ; RV32IZHINX-NEXT:    call __fixhfdi@plt
 ; RV32IZHINX-NEXT:    lw ra, 12(sp) # 4-byte Folded Reload
 ; RV32IZHINX-NEXT:    addi sp, sp, 16
 ; RV32IZHINX-NEXT:    ret
 ;
 ; RV64IZHINX-LABEL: test_roundeven_si64:
 ; RV64IZHINX:       # %bb.0:
 ; RV64IZHINX-NEXT:    lui a1, %hi(.LCPI35_0)
; CHECKIZHINXMIN-NEXT:    ret
   %a = call half @llvm.roundeven.f16(half %x)
   %b = fptoui half %a to i32
   ret i32 %b
 }
 
 define i64 @test_roundeven_ui64(half %x) {
 ; RV32IZFH-LABEL: test_roundeven_ui64:
 ; RV32IZFH:       # %bb.0:
-; RV32IZFH-NEXT:    addi sp, sp, -16
-; RV32IZFH-NEXT:    .cfi_def_cfa_offset 16
-; RV32IZFH-NEXT:    sw ra, 12(sp) # 4-byte Folded Spill
-; RV32IZFH-NEXT:    .cfi_offset ra, -4
 ; RV32IZFH-NEXT:    lui a0, %hi(.LCPI39_0)
 ; RV32IZFH-NEXT:    flh fa5, %lo(.LCPI39_0)(a0)
 ; RV32IZFH-NEXT:    fabs.h fa4, fa0
 ; RV32IZFH-NEXT:    flt.h a0, fa4, fa5
 ; RV32IZFH-NEXT:    beqz a0, .LBB39_2
 ; RV32IZFH-NEXT:  # %bb.1:
 ; RV32IZFH-NEXT:    fcvt.w.h a0, fa0, rne
 ; RV32IZFH-NEXT:    fcvt.h.w fa5, a0, rne
 ; RV32IZFH-NEXT:    fsgnj.h fa0, fa5, fa0
 ; RV32IZFH-NEXT:  .LBB39_2:
+; RV32IZFH-NEXT:    addi sp, sp, -16
+; RV32IZFH-NEXT:    .cfi_def_cfa_offset 16
+; RV32IZFH-NEXT:    sw ra, 12(sp) # 4-byte Folded Spill
+; RV32IZFH-NEXT:    .cfi_offset ra, -4
 ; RV32IZFH-NEXT:    call __fixunshfdi@plt
 ; RV32IZFH-NEXT:    lw ra, 12(sp) # 4-byte Folded Reload
 ; RV32IZFH-NEXT:    addi sp, sp, 16
 ; RV32IZFH-NEXT:    ret
 ;
 ; RV64IZFH-LABEL: test_roundeven_ui64:
 ; RV64IZFH:       # %bb.0:
 ; RV64IZFH-NEXT:    fcvt.lu.h a0, fa0, rne
 ; RV64IZFH-NEXT:    ret
 ;
 ; RV32IZHINX-LABEL: test_roundeven_ui64:
 ; RV32IZHINX:       # %bb.0:
-; RV32IZHINX-NEXT:    addi sp, sp, -16
-; RV32IZHINX-NEXT:    .cfi_def_cfa_offset 16
-; RV32IZHINX-NEXT:    sw ra, 12(sp) # 4-byte Folded Spill
-; RV32IZHINX-NEXT:    .cfi_offset ra, -4
 ; RV32IZHINX-NEXT:    lui a1, %hi(.LCPI39_0)
 ; RV32IZHINX-NEXT:    lh a1, %lo(.LCPI39_0)(a1)
 ; RV32IZHINX-NEXT:    fabs.h a2, a0
 ; RV32IZHINX-NEXT:    flt.h a1, a2, a1
 ; RV32IZHINX-NEXT:    beqz a1, .LBB39_2
 ; RV32IZHINX-NEXT:  # %bb.1:
 ; RV32IZHINX-NEXT:    fcvt.w.h a1, a0, rne
 ; RV32IZHINX-NEXT:    fcvt.h.w a1, a1, rne
 ; RV32IZHINX-NEXT:    fsgnj.h a0, a1, a0
 ; RV32IZHINX-NEXT:  .LBB39_2:
+; RV32IZHINX-NEXT:    addi sp, sp, -16
+; RV32IZHINX-NEXT:    .cfi_def_cfa_offset 16
+; RV32IZHINX-NEXT:    sw ra, 12(sp) # 4-byte Folded Spill
+; RV32IZHINX-NEXT:    .cfi_offset ra, -4
 ; RV32IZHINX-NEXT:    call __fixunshfdi@plt
 ; RV32IZHINX-NEXT:    lw ra, 12(sp) # 4-byte Folded Reload
 ; RV32IZHINX-NEXT:    addi sp, sp, 16
 ; RV32IZHINX-NEXT:    ret
 ;
 ; RV64IZHINX-LABEL: test_roundeven_ui64:
 ; RV64IZHINX:       # %bb.0:
 ; RV64IZHINX-NEXT:    lui a1, %hi(.LCPI39_0)

llvm/test/CodeGen/RISCV/rvv/fixed-vectors-masked-gather.ll

 ; RV64V:       # %bb.0:
 ; RV64V-NEXT:    vsetivli zero, 8, e64, m4, ta, mu
 ; RV64V-NEXT:    vluxei64.v v12, (zero), v8, v0.t
 ; RV64V-NEXT:    vmv.v.v v8, v12
 ; RV64V-NEXT:    ret
 ;
 ; RV32ZVE32F-LABEL: mgather_v8i64:
 ; RV32ZVE32F:       # %bb.0:
-; RV32ZVE32F-NEXT:    addi sp, sp, -16
-; RV32ZVE32F-NEXT:    .cfi_def_cfa_offset 16
-; RV32ZVE32F-NEXT:    sw s0, 12(sp) # 4-byte Folded Spill
-; RV32ZVE32F-NEXT:    sw s1, 8(sp) # 4-byte Folded Spill
-; RV32ZVE32F-NEXT:    .cfi_offset s0, -4
-; RV32ZVE32F-NEXT:    .cfi_offset s1, -8
 ; RV32ZVE32F-NEXT:    vsetivli zero, 0, e8, mf4, ta, ma
 ; RV32ZVE32F-NEXT:    vmv.x.s t0, v0
 ; RV32ZVE32F-NEXT:    andi a2, t0, 1
-; RV32ZVE32F-NEXT:    beqz a2, .LBB47_9
+; RV32ZVE32F-NEXT:    beqz a2, .LBB47_7
 ; RV32ZVE32F-NEXT:  # %bb.1: # %cond.load
 ; RV32ZVE32F-NEXT:    vsetivli zero, 0, e32, m2, ta, ma
 ; RV32ZVE32F-NEXT:    vmv.x.s a3, v8
 ; RV32ZVE32F-NEXT:    lw a2, 4(a3)
 ; RV32ZVE32F-NEXT:    lw a3, 0(a3)
 ; RV32ZVE32F-NEXT:    andi a4, t0, 2
-; RV32ZVE32F-NEXT:    bnez a4, .LBB47_10
+; RV32ZVE32F-NEXT:    bnez a4, .LBB47_8
 ; RV32ZVE32F-NEXT:  .LBB47_2:
 ; RV32ZVE32F-NEXT:    lw a4, 12(a1)
 ; RV32ZVE32F-NEXT:    lw a5, 8(a1)
 ; RV32ZVE32F-NEXT:    andi a6, t0, 4
-; RV32ZVE32F-NEXT:    bnez a6, .LBB47_11
+; RV32ZVE32F-NEXT:    bnez a6, .LBB47_9
 ; RV32ZVE32F-NEXT:  .LBB47_3:
 ; RV32ZVE32F-NEXT:    lw a6, 20(a1)
 ; RV32ZVE32F-NEXT:    lw a7, 16(a1)
 ; RV32ZVE32F-NEXT:    andi t1, t0, 8
-; RV32ZVE32F-NEXT:    bnez t1, .LBB47_12
+; RV32ZVE32F-NEXT:    bnez t1, .LBB47_10
 ; RV32ZVE32F-NEXT:  .LBB47_4:
 ; RV32ZVE32F-NEXT:    lw t1, 28(a1)
 ; RV32ZVE32F-NEXT:    lw t2, 24(a1)
 ; RV32ZVE32F-NEXT:    andi t3, t0, 16
-; RV32ZVE32F-NEXT:    bnez t3, .LBB47_13
+; RV32ZVE32F-NEXT:    bnez t3, .LBB47_11
 ; RV32ZVE32F-NEXT:  .LBB47_5:
 ; RV32ZVE32F-NEXT:    lw t3, 36(a1)
 ; RV32ZVE32F-NEXT:    lw t4, 32(a1)
 ; RV32ZVE32F-NEXT:    andi t5, t0, 32
-; RV32ZVE32F-NEXT:    bnez t5, .LBB47_14
+; RV32ZVE32F-NEXT:    bnez t5, .LBB47_12
 ; RV32ZVE32F-NEXT:  .LBB47_6:
 ; RV32ZVE32F-NEXT:    lw t5, 44(a1)
 ; RV32ZVE32F-NEXT:    lw t6, 40(a1)
-; RV32ZVE32F-NEXT:    andi s0, t0, 64
-; RV32ZVE32F-NEXT:    bnez s0, .LBB47_15
+; RV32ZVE32F-NEXT:    j .LBB47_13
 ; RV32ZVE32F-NEXT:  .LBB47_7:
-; RV32ZVE32F-NEXT:    lw s0, 52(a1)
-; RV32ZVE32F-NEXT:    lw s1, 48(a1)
-; RV32ZVE32F-NEXT:    andi t0, t0, -128
-; RV32ZVE32F-NEXT:    bnez t0, .LBB47_16
-; RV32ZVE32F-NEXT:  .LBB47_8:
-; RV32ZVE32F-NEXT:    lw t0, 60(a1)
-; RV32ZVE32F-NEXT:    lw a1, 56(a1)
-; RV32ZVE32F-NEXT:    j .LBB47_17
-; RV32ZVE32F-NEXT:  .LBB47_9:
 ; RV32ZVE32F-NEXT:    lw a2, 4(a1)
 ; RV32ZVE32F-NEXT:    lw a3, 0(a1)
 ; RV32ZVE32F-NEXT:    andi a4, t0, 2
 ; RV32ZVE32F-NEXT:    beqz a4, .LBB47_2
-; RV32ZVE32F-NEXT:  .LBB47_10: # %cond.load1
+; RV32ZVE32F-NEXT:  .LBB47_8: # %cond.load1
 ; RV32ZVE32F-NEXT:    vsetivli zero, 1, e32, m2, ta, ma
 ; RV32ZVE32F-NEXT:    vslidedown.vi v10, v8, 1
 ; RV32ZVE32F-NEXT:    vmv.x.s a5, v10
 ; RV32ZVE32F-NEXT:    lw a4, 4(a5)
 ; RV32ZVE32F-NEXT:    lw a5, 0(a5)
 ; RV32ZVE32F-NEXT:    andi a6, t0, 4
 ; RV32ZVE32F-NEXT:    beqz a6, .LBB47_3
-; RV32ZVE32F-NEXT:  .LBB47_11: # %cond.load4
+; RV32ZVE32F-NEXT:  .LBB47_9: # %cond.load4
 ; RV32ZVE32F-NEXT:    vsetivli zero, 1, e32, m2, ta, ma
 ; RV32ZVE32F-NEXT:    vslidedown.vi v10, v8, 2
 ; RV32ZVE32F-NEXT:    vmv.x.s a7, v10
 ; RV32ZVE32F-NEXT:    lw a6, 4(a7)
 ; RV32ZVE32F-NEXT:    lw a7, 0(a7)
 ; RV32ZVE32F-NEXT:    andi t1, t0, 8
 ; RV32ZVE32F-NEXT:    beqz t1, .LBB47_4
-; RV32ZVE32F-NEXT:  .LBB47_12: # %cond.load7
+; RV32ZVE32F-NEXT:  .LBB47_10: # %cond.load7
 ; RV32ZVE32F-NEXT:    vsetivli zero, 1, e32, m2, ta, ma
 ; RV32ZVE32F-NEXT:    vslidedown.vi v10, v8, 3
 ; RV32ZVE32F-NEXT:    vmv.x.s t2, v10
 ; RV32ZVE32F-NEXT:    lw t1, 4(t2)
 ; RV32ZVE32F-NEXT:    lw t2, 0(t2)
 ; RV32ZVE32F-NEXT:    andi t3, t0, 16
 ; RV32ZVE32F-NEXT:    beqz t3, .LBB47_5
-; RV32ZVE32F-NEXT:  .LBB47_13: # %cond.load10
+; RV32ZVE32F-NEXT:  .LBB47_11: # %cond.load10
 ; RV32ZVE32F-NEXT:    vsetivli zero, 1, e32, m2, ta, ma
 ; RV32ZVE32F-NEXT:    vslidedown.vi v10, v8, 4
 ; RV32ZVE32F-NEXT:    vmv.x.s t4, v10
 ; RV32ZVE32F-NEXT:    lw t3, 4(t4)
 ; RV32ZVE32F-NEXT:    lw t4, 0(t4)
 ; RV32ZVE32F-NEXT:    andi t5, t0, 32
 ; RV32ZVE32F-NEXT:    beqz t5, .LBB47_6
-; RV32ZVE32F-NEXT:  .LBB47_14: # %cond.load13
+; RV32ZVE32F-NEXT:  .LBB47_12: # %cond.load13
 ; RV32ZVE32F-NEXT:    vsetivli zero, 1, e32, m2, ta, ma
 ; RV32ZVE32F-NEXT:    vslidedown.vi v10, v8, 5
 ; RV32ZVE32F-NEXT:    vmv.x.s t6, v10
 ; RV32ZVE32F-NEXT:    lw t5, 4(t6)
 ; RV32ZVE32F-NEXT:    lw t6, 0(t6)
+; RV32ZVE32F-NEXT:  .LBB47_13: # %else14
+; RV32ZVE32F-NEXT:    addi sp, sp, -16
+; RV32ZVE32F-NEXT:    .cfi_def_cfa_offset 16
+; RV32ZVE32F-NEXT:    sw s0, 12(sp) # 4-byte Folded Spill
+; RV32ZVE32F-NEXT:    sw s1, 8(sp) # 4-byte Folded Spill
+; RV32ZVE32F-NEXT:    .cfi_offset s0, -4
+; RV32ZVE32F-NEXT:    .cfi_offset s1, -8
 ; RV32ZVE32F-NEXT:    andi s0, t0, 64
-; RV32ZVE32F-NEXT:    beqz s0, .LBB47_7
-; RV32ZVE32F-NEXT:  .LBB47_15: # %cond.load16
+; RV32ZVE32F-NEXT:    beqz s0, .LBB47_16
+; RV32ZVE32F-NEXT:  # %bb.14: # %cond.load16
 ; RV32ZVE32F-NEXT:    vsetivli zero, 1, e32, m2, ta, ma
 ; RV32ZVE32F-NEXT:    vslidedown.vi v10, v8, 6
 ; RV32ZVE32F-NEXT:    vmv.x.s s1, v10
 ; RV32ZVE32F-NEXT:    lw s0, 4(s1)
 ; RV32ZVE32F-NEXT:    lw s1, 0(s1)
 ; RV32ZVE32F-NEXT:    andi t0, t0, -128
-; RV32ZVE32F-NEXT:    beqz t0, .LBB47_8
-; RV32ZVE32F-NEXT:  .LBB47_16: # %cond.load19
+; RV32ZVE32F-NEXT:    bnez t0, .LBB47_17
+; RV32ZVE32F-NEXT:  .LBB47_15:
+; RV32ZVE32F-NEXT:    lw t0, 60(a1)
+; RV32ZVE32F-NEXT:    lw a1, 56(a1)
+; RV32ZVE32F-NEXT:    j .LBB47_18
+; RV32ZVE32F-NEXT:  .LBB47_16:
+; RV32ZVE32F-NEXT:    lw s0, 52(a1)
+; RV32ZVE32F-NEXT:    lw s1, 48(a1)
+; RV32ZVE32F-NEXT:    andi t0, t0, -128
+; RV32ZVE32F-NEXT:    beqz t0, .LBB47_15
+; RV32ZVE32F-NEXT:  .LBB47_17: # %cond.load19
 ; RV32ZVE32F-NEXT:    vsetivli zero, 1, e32, m2, ta, ma
 ; RV32ZVE32F-NEXT:    vslidedown.vi v8, v8, 7
 ; RV32ZVE32F-NEXT:    vmv.x.s a1, v8
 ; RV32ZVE32F-NEXT:    lw t0, 4(a1)
 ; RV32ZVE32F-NEXT:    lw a1, 0(a1)
-; RV32ZVE32F-NEXT:  .LBB47_17: # %else20
+; RV32ZVE32F-NEXT:  .LBB47_18: # %else20
 ; RV32ZVE32F-NEXT:    sw a3, 0(a0)
 ; RV32ZVE32F-NEXT:    sw a2, 4(a0)
 ; RV32ZVE32F-NEXT:    sw a5, 8(a0)
 ; RV32ZVE32F-NEXT:    sw a4, 12(a0)
 ; RV32ZVE32F-NEXT:    sw a7, 16(a0)
 ; RV32ZVE32F-NEXT:    sw a6, 20(a0)
 ; RV32ZVE32F-NEXT:    sw t2, 24(a0)
 ; RV32ZVE32F-NEXT:    sw t1, 28(a0)

 ; RV64V-NEXT:    vsext.vf8 v16, v8
 ; RV64V-NEXT:    vsll.vi v8, v16, 3
 ; RV64V-NEXT:    vluxei64.v v12, (a0), v8, v0.t
 ; RV64V-NEXT:    vmv.v.v v8, v12
 ; RV64V-NEXT:    ret
 ;
 ; RV32ZVE32F-LABEL: mgather_baseidx_v8i8_v8i64:
 ; RV32ZVE32F:       # %bb.0:
-; RV32ZVE32F-NEXT:    addi sp, sp, -16
-; RV32ZVE32F-NEXT:    .cfi_def_cfa_offset 16
-; RV32ZVE32F-NEXT:    sw s0, 12(sp) # 4-byte Folded Spill
-; RV32ZVE32F-NEXT:    sw s1, 8(sp) # 4-byte Folded Spill
-; RV32ZVE32F-NEXT:    .cfi_offset s0, -4
-; RV32ZVE32F-NEXT:    .cfi_offset s1, -8
 ; RV32ZVE32F-NEXT:    vsetivli zero, 8, e32, m2, ta, ma
 ; RV32ZVE32F-NEXT:    vsext.vf4 v10, v8
 ; RV32ZVE32F-NEXT:    vsll.vi v8, v10, 3
 ; RV32ZVE32F-NEXT:    vadd.vx v8, v8, a1
 ; RV32ZVE32F-NEXT:    vsetivli zero, 0, e8, mf4, ta, ma
 ; RV32ZVE32F-NEXT:    vmv.x.s t0, v0
 ; RV32ZVE32F-NEXT:    andi a1, t0, 1
-; RV32ZVE32F-NEXT:    beqz a1, .LBB48_9
+; RV32ZVE32F-NEXT:    beqz a1, .LBB48_7
 ; RV32ZVE32F-NEXT:  # %bb.1: # %cond.load
 ; RV32ZVE32F-NEXT:    vsetivli zero, 0, e32, m2, ta, ma
 ; RV32ZVE32F-NEXT:    vmv.x.s a3, v8
 ; RV32ZVE32F-NEXT:    lw a1, 4(a3)
 ; RV32ZVE32F-NEXT:    lw a3, 0(a3)
 ; RV32ZVE32F-NEXT:    andi a4, t0, 2
-; RV32ZVE32F-NEXT:    bnez a4, .LBB48_10
+; RV32ZVE32F-NEXT:    bnez a4, .LBB48_8
 ; RV32ZVE32F-NEXT:  .LBB48_2:
 ; RV32ZVE32F-NEXT:    lw a4, 12(a2)
 ; RV32ZVE32F-NEXT:    lw a5, 8(a2)
 ; RV32ZVE32F-NEXT:    andi a6, t0, 4
-; RV32ZVE32F-NEXT:    bnez a6, .LBB48_11
+; RV32ZVE32F-NEXT:    bnez a6, .LBB48_9
 ; RV32ZVE32F-NEXT:  .LBB48_3:
 ; RV32ZVE32F-NEXT:    lw a6, 20(a2)
 ; RV32ZVE32F-NEXT:    lw a7, 16(a2)
 ; RV32ZVE32F-NEXT:    andi t1, t0, 8
-; RV32ZVE32F-NEXT:    bnez t1, .LBB48_12
+; RV32ZVE32F-NEXT:    bnez t1, .LBB48_10
 ; RV32ZVE32F-NEXT:  .LBB48_4:
 ; RV32ZVE32F-NEXT:    lw t1, 28(a2)
 ; RV32ZVE32F-NEXT:    lw t2, 24(a2)
 ; RV32ZVE32F-NEXT:    andi t3, t0, 16
-; RV32ZVE32F-NEXT:    bnez t3, .LBB48_13
+; RV32ZVE32F-NEXT:    bnez t3, .LBB48_11
 ; RV32ZVE32F-NEXT:  .LBB48_5:
 ; RV32ZVE32F-NEXT:    lw t3, 36(a2)
 ; RV32ZVE32F-NEXT:    lw t4, 32(a2)
 ; RV32ZVE32F-NEXT:    andi t5, t0, 32
-; RV32ZVE32F-NEXT:    bnez t5, .LBB48_14
+; RV32ZVE32F-NEXT:    bnez t5, .LBB48_12
 ; RV32ZVE32F-NEXT:  .LBB48_6:
 ; RV32ZVE32F-NEXT:    lw t5, 44(a2)
 ; RV32ZVE32F-NEXT:    lw t6, 40(a2)
-; RV32ZVE32F-NEXT:    andi s0, t0, 64
-; RV32ZVE32F-NEXT:    bnez s0, .LBB48_15
+; RV32ZVE32F-NEXT:    j .LBB48_13
 ; RV32ZVE32F-NEXT:  .LBB48_7:
-; RV32ZVE32F-NEXT:    lw s0, 52(a2)
-; RV32ZVE32F-NEXT:    lw s1, 48(a2)
-; RV32ZVE32F-NEXT:    andi t0, t0, -128
-; RV32ZVE32F-NEXT:    bnez t0, .LBB48_16
-; RV32ZVE32F-NEXT:  .LBB48_8:
-; RV32ZVE32F-NEXT:    lw t0, 60(a2)
-; RV32ZVE32F-NEXT:    lw a2, 56(a2)
-; RV32ZVE32F-NEXT:    j .LBB48_17
-; RV32ZVE32F-NEXT:  .LBB48_9:
 ; RV32ZVE32F-NEXT:    lw a1, 4(a2)
 ; RV32ZVE32F-NEXT:    lw a3, 0(a2)
 ; RV32ZVE32F-NEXT:    andi a4, t0, 2
 ; RV32ZVE32F-NEXT:    beqz a4, .LBB48_2
-; RV32ZVE32F-NEXT:  .LBB48_10: # %cond.load1
+; RV32ZVE32F-NEXT:  .LBB48_8: # %cond.load1
 ; RV32ZVE32F-NEXT:    vsetivli zero, 1, e32, m2, ta, ma
 ; RV32ZVE32F-NEXT:    vslidedown.vi v10, v8, 1
 ; RV32ZVE32F-NEXT:    vmv.x.s a5, v10
 ; RV32ZVE32F-NEXT:    lw a4, 4(a5)
 ; RV32ZVE32F-NEXT:    lw a5, 0(a5)
 ; RV32ZVE32F-NEXT:    andi a6, t0, 4
 ; RV32ZVE32F-NEXT:    beqz a6, .LBB48_3
-; RV32ZVE32F-NEXT:  .LBB48_11: # %cond.load4
+; RV32ZVE32F-NEXT:  .LBB48_9: # %cond.load4
 ; RV32ZVE32F-NEXT:    vsetivli zero, 1, e32, m2, ta, ma
 ; RV32ZVE32F-NEXT:    vslidedown.vi v10, v8, 2
 ; RV32ZVE32F-NEXT:    vmv.x.s a7, v10
 ; RV32ZVE32F-NEXT:    lw a6, 4(a7)
 ; RV32ZVE32F-NEXT:    lw a7, 0(a7)
 ; RV32ZVE32F-NEXT:    andi t1, t0, 8
 ; RV32ZVE32F-NEXT:    beqz t1, .LBB48_4
-; RV32ZVE32F-NEXT:  .LBB48_12: # %cond.load7
+; RV32ZVE32F-NEXT:  .LBB48_10: # %cond.load7
 ; RV32ZVE32F-NEXT:    vsetivli zero, 1, e32, m2, ta, ma
 ; RV32ZVE32F-NEXT:    vslidedown.vi v10, v8, 3
 ; RV32ZVE32F-NEXT:    vmv.x.s t2, v10
 ; RV32ZVE32F-NEXT:    lw t1, 4(t2)
 ; RV32ZVE32F-NEXT:    lw t2, 0(t2)
 ; RV32ZVE32F-NEXT:    andi t3, t0, 16
 ; RV32ZVE32F-NEXT:    beqz t3, .LBB48_5
-; RV32ZVE32F-NEXT:  .LBB48_13: # %cond.load10
+; RV32ZVE32F-NEXT:  .LBB48_11: # %cond.load10
 ; RV32ZVE32F-NEXT:    vsetivli zero, 1, e32, m2, ta, ma
 ; RV32ZVE32F-NEXT:    vslidedown.vi v10, v8, 4
 ; RV32ZVE32F-NEXT:    vmv.x.s t4, v10
 ; RV32ZVE32F-NEXT:    lw t3, 4(t4)
 ; RV32ZVE32F-NEXT:    lw t4, 0(t4)
 ; RV32ZVE32F-NEXT:    andi t5, t0, 32
 ; RV32ZVE32F-NEXT:    beqz t5, .LBB48_6
-; RV32ZVE32F-NEXT:  .LBB48_14: # %cond.load13
+; RV32ZVE32F-NEXT:  .LBB48_12: # %cond.load13
 ; RV32ZVE32F-NEXT:    vsetivli zero, 1, e32, m2, ta, ma
 ; RV32ZVE32F-NEXT:    vslidedown.vi v10, v8, 5
 ; RV32ZVE32F-NEXT:    vmv.x.s t6, v10
 ; RV32ZVE32F-NEXT:    lw t5, 4(t6)
 ; RV32ZVE32F-NEXT:    lw t6, 0(t6)
+; RV32ZVE32F-NEXT:  .LBB48_13: # %else14
+; RV32ZVE32F-NEXT:    addi sp, sp, -16
+; RV32ZVE32F-NEXT:    .cfi_def_cfa_offset 16
+; RV32ZVE32F-NEXT:    sw s0, 12(sp) # 4-byte Folded Spill
+; RV32ZVE32F-NEXT:    sw s1, 8(sp) # 4-byte Folded Spill
+; RV32ZVE32F-NEXT:    .cfi_offset s0, -4
+; RV32ZVE32F-NEXT:    .cfi_offset s1, -8
 ; RV32ZVE32F-NEXT:    andi s0, t0, 64
-; RV32ZVE32F-NEXT:    beqz s0, .LBB48_7
-; RV32ZVE32F-NEXT:  .LBB48_15: # %cond.load16
+; RV32ZVE32F-NEXT:    beqz s0, .LBB48_16
+; RV32ZVE32F-NEXT:  # %bb.14: # %cond.load16
 ; RV32ZVE32F-NEXT:    vsetivli zero, 1, e32, m2, ta, ma
 ; RV32ZVE32F-NEXT:    vslidedown.vi v10, v8, 6
 ; RV32ZVE32F-NEXT:    vmv.x.s s1, v10
 ; RV32ZVE32F-NEXT:    lw s0, 4(s1)
 ; RV32ZVE32F-NEXT:    lw s1, 0(s1)
 ; RV32ZVE32F-NEXT:    andi t0, t0, -128
-; RV32ZVE32F-NEXT:    beqz t0, .LBB48_8
-; RV32ZVE32F-NEXT:  .LBB48_16: # %cond.load19
+; RV32ZVE32F-NEXT:    bnez t0, .LBB48_17
+; RV32ZVE32F-NEXT:  .LBB48_15:
+; RV32ZVE32F-NEXT:    lw t0, 60(a2)
+; RV32ZVE32F-NEXT:    lw a2, 56(a2)
+; RV32ZVE32F-NEXT:    j .LBB48_18
+; RV32ZVE32F-NEXT:  .LBB48_16:
+; RV32ZVE32F-NEXT:    lw s0, 52(a2)
+; RV32ZVE32F-NEXT:    lw s1, 48(a2)
+; RV32ZVE32F-NEXT:    andi t0, t0, -128
+; RV32ZVE32F-NEXT:    beqz t0, .LBB48_15
+; RV32ZVE32F-NEXT:  .LBB48_17: # %cond.load19
 ; RV32ZVE32F-NEXT:    vsetivli zero, 1, e32, m2, ta, ma
 ; RV32ZVE32F-NEXT:    vslidedown.vi v8, v8, 7
 ; RV32ZVE32F-NEXT:    vmv.x.s a2, v8
 ; RV32ZVE32F-NEXT:    lw t0, 4(a2)
 ; RV32ZVE32F-NEXT:    lw a2, 0(a2)
-; RV32ZVE32F-NEXT:  .LBB48_17: # %else20
+; RV32ZVE32F-NEXT:  .LBB48_18: # %else20
 ; RV32ZVE32F-NEXT:    sw a3, 0(a0)
 ; RV32ZVE32F-NEXT:    sw a1, 4(a0)
 ; RV32ZVE32F-NEXT:    sw a5, 8(a0)
 ; RV32ZVE32F-NEXT:    sw a4, 12(a0)
 ; RV32ZVE32F-NEXT:    sw a7, 16(a0)
 ; RV32ZVE32F-NEXT:    sw a6, 20(a0)
 ; RV32ZVE32F-NEXT:    sw t2, 24(a0)
 ; RV32ZVE32F-NEXT:    sw t1, 28(a0)

 ; RV64V-NEXT:    vsext.vf8 v16, v8
 ; RV64V-NEXT:    vsll.vi v8, v16, 3
 ; RV64V-NEXT:    vluxei64.v v12, (a0), v8, v0.t
 ; RV64V-NEXT:    vmv.v.v v8, v12
 ; RV64V-NEXT:    ret
 ;
 ; RV32ZVE32F-LABEL: mgather_baseidx_sext_v8i8_v8i64:
 ; RV32ZVE32F:       # %bb.0:
-; RV32ZVE32F-NEXT:    addi sp, sp, -16
-; RV32ZVE32F-NEXT:    .cfi_def_cfa_offset 16
-; RV32ZVE32F-NEXT:    sw s0, 12(sp) # 4-byte Folded Spill
-; RV32ZVE32F-NEXT:    sw s1, 8(sp) # 4-byte Folded Spill
-; RV32ZVE32F-NEXT:    .cfi_offset s0, -4
-; RV32ZVE32F-NEXT:    .cfi_offset s1, -8
 ; RV32ZVE32F-NEXT:    vsetivli zero, 8, e32, m2, ta, ma
 ; RV32ZVE32F-NEXT:    vsext.vf4 v10, v8
 ; RV32ZVE32F-NEXT:    vsll.vi v8, v10, 3
 ; RV32ZVE32F-NEXT:    vadd.vx v8, v8, a1
 ; RV32ZVE32F-NEXT:    vsetivli zero, 0, e8, mf4, ta, ma
 ; RV32ZVE32F-NEXT:    vmv.x.s t0, v0
 ; RV32ZVE32F-NEXT:    andi a1, t0, 1
-; RV32ZVE32F-NEXT:    beqz a1, .LBB49_9
+; RV32ZVE32F-NEXT:    beqz a1, .LBB49_7
 ; RV32ZVE32F-NEXT:  # %bb.1: # %cond.load
 ; RV32ZVE32F-NEXT:    vsetivli zero, 0, e32, m2, ta, ma
 ; RV32ZVE32F-NEXT:    vmv.x.s a3, v8
 ; RV32ZVE32F-NEXT:    lw a1, 4(a3)
 ; RV32ZVE32F-NEXT:    lw a3, 0(a3)
 ; RV32ZVE32F-NEXT:    andi a4, t0, 2
-; RV32ZVE32F-NEXT:    bnez a4, .LBB49_10
+; RV32ZVE32F-NEXT:    bnez a4, .LBB49_8
 ; RV32ZVE32F-NEXT:  .LBB49_2:
 ; RV32ZVE32F-NEXT:    lw a4, 12(a2)
 ; RV32ZVE32F-NEXT:    lw a5, 8(a2)
 ; RV32ZVE32F-NEXT:    andi a6, t0, 4
-; RV32ZVE32F-NEXT:    bnez a6, .LBB49_11
+; RV32ZVE32F-NEXT:    bnez a6, .LBB49_9
 ; RV32ZVE32F-NEXT:  .LBB49_3:
 ; RV32ZVE32F-NEXT:    lw a6, 20(a2)
 ; RV32ZVE32F-NEXT:    lw a7, 16(a2)
 ; RV32ZVE32F-NEXT:    andi t1, t0, 8
-; RV32ZVE32F-NEXT:    bnez t1, .LBB49_12
+; RV32ZVE32F-NEXT:    bnez t1, .LBB49_10
 ; RV32ZVE32F-NEXT:  .LBB49_4:
 ; RV32ZVE32F-NEXT:    lw t1, 28(a2)
 ; RV32ZVE32F-NEXT:    lw t2, 24(a2)
 ; RV32ZVE32F-NEXT:    andi t3, t0, 16
-; RV32ZVE32F-NEXT:    bnez t3, .LBB49_13
+; RV32ZVE32F-NEXT:    bnez t3, .LBB49_11
 ; RV32ZVE32F-NEXT:  .LBB49_5:
 ; RV32ZVE32F-NEXT:    lw t3, 36(a2)
 ; RV32ZVE32F-NEXT:    lw t4, 32(a2)
 ; RV32ZVE32F-NEXT:    andi t5, t0, 32
-; RV32ZVE32F-NEXT:    bnez t5, .LBB49_14
+; RV32ZVE32F-NEXT:    bnez t5, .LBB49_12
 ; RV32ZVE32F-NEXT:  .LBB49_6:
 ; RV32ZVE32F-NEXT:    lw t5, 44(a2)
 ; RV32ZVE32F-NEXT:    lw t6, 40(a2)
-; RV32ZVE32F-NEXT:    andi s0, t0, 64
-; RV32ZVE32F-NEXT:    bnez s0, .LBB49_15
+; RV32ZVE32F-NEXT:    j .LBB49_13
 ; RV32ZVE32F-NEXT:  .LBB49_7:
-; RV32ZVE32F-NEXT:    lw s0, 52(a2)
-; RV32ZVE32F-NEXT:    lw s1, 48(a2)
-; RV32ZVE32F-NEXT:    andi t0, t0, -128
-; RV32ZVE32F-NEXT:    bnez t0, .LBB49_16
-; RV32ZVE32F-NEXT:  .LBB49_8:
-; RV32ZVE32F-NEXT:    lw t0, 60(a2)
-; RV32ZVE32F-NEXT:    lw a2, 56(a2)
-; RV32ZVE32F-NEXT:    j .LBB49_17
-; RV32ZVE32F-NEXT:  .LBB49_9:
 ; RV32ZVE32F-NEXT:    lw a1, 4(a2)
 ; RV32ZVE32F-NEXT:    lw a3, 0(a2)
 ; RV32ZVE32F-NEXT:    andi a4, t0, 2
 ; RV32ZVE32F-NEXT:    beqz a4, .LBB49_2
-; RV32ZVE32F-NEXT:  .LBB49_10: # %cond.load1
+; RV32ZVE32F-NEXT:  .LBB49_8: # %cond.load1
 ; RV32ZVE32F-NEXT:    vsetivli zero, 1, e32, m2, ta, ma
 ; RV32ZVE32F-NEXT:    vslidedown.vi v10, v8, 1
 ; RV32ZVE32F-NEXT:    vmv.x.s a5, v10
 ; RV32ZVE32F-NEXT:    lw a4, 4(a5)
 ; RV32ZVE32F-NEXT:    lw a5, 0(a5)
 ; RV32ZVE32F-NEXT:    andi a6, t0, 4
 ; RV32ZVE32F-NEXT:    beqz a6, .LBB49_3
-; RV32ZVE32F-NEXT:  .LBB49_11: # %cond.load4
+; RV32ZVE32F-NEXT:  .LBB49_9: # %cond.load4
 ; RV32ZVE32F-NEXT:    vsetivli zero, 1, e32, m2, ta, ma
 ; RV32ZVE32F-NEXT:    vslidedown.vi v10, v8, 2
 ; RV32ZVE32F-NEXT:    vmv.x.s a7, v10
 ; RV32ZVE32F-NEXT:    lw a6, 4(a7)
 ; RV32ZVE32F-NEXT:    lw a7, 0(a7)
 ; RV32ZVE32F-NEXT:    andi t1, t0, 8
 ; RV32ZVE32F-NEXT:    beqz t1, .LBB49_4
-; RV32ZVE32F-NEXT:  .LBB49_12: # %cond.load7
+; RV32ZVE32F-NEXT:  .LBB49_10: # %cond.load7
 ; RV32ZVE32F-NEXT:    vsetivli zero, 1, e32, m2, ta, ma
 ; RV32ZVE32F-NEXT:    vslidedown.vi v10, v8, 3
 ; RV32ZVE32F-NEXT:    vmv.x.s t2, v10
 ; RV32ZVE32F-NEXT:    lw t1, 4(t2)
 ; RV32ZVE32F-NEXT:    lw t2, 0(t2)
 ; RV32ZVE32F-NEXT:    andi t3, t0, 16
 ; RV32ZVE32F-NEXT:    beqz t3, .LBB49_5
-; RV32ZVE32F-NEXT:  .LBB49_13: # %cond.load10
+; RV32ZVE32F-NEXT:  .LBB49_11: # %cond.load10
 ; RV32ZVE32F-NEXT:    vsetivli zero, 1, e32, m2, ta, ma
 ; RV32ZVE32F-NEXT:    vslidedown.vi v10, v8, 4
 ; RV32ZVE32F-NEXT:    vmv.x.s t4, v10
 ; RV32ZVE32F-NEXT:    lw t3, 4(t4)
 ; RV32ZVE32F-NEXT:    lw t4, 0(t4)
 ; RV32ZVE32F-NEXT:    andi t5, t0, 32
 ; RV32ZVE32F-NEXT:    beqz t5, .LBB49_6
-; RV32ZVE32F-NEXT:  .LBB49_14: # %cond.load13
+; RV32ZVE32F-NEXT:  .LBB49_12: # %cond.load13
 ; RV32ZVE32F-NEXT:    vsetivli zero, 1, e32, m2, ta, ma
 ; RV32ZVE32F-NEXT:    vslidedown.vi v10, v8, 5
 ; RV32ZVE32F-NEXT:    vmv.x.s t6, v10
 ; RV32ZVE32F-NEXT:    lw t5, 4(t6)
 ; RV32ZVE32F-NEXT:    lw t6, 0(t6)
+; RV32ZVE32F-NEXT:  .LBB49_13: # %else14
+; RV32ZVE32F-NEXT:    addi sp, sp, -16
+; RV32ZVE32F-NEXT:    .cfi_def_cfa_offset 16
+; RV32ZVE32F-NEXT:    sw s0, 12(sp) # 4-byte Folded Spill
+; RV32ZVE32F-NEXT:    sw s1, 8(sp) # 4-byte Folded Spill
+; RV32ZVE32F-NEXT:    .cfi_offset s0, -4
+; RV32ZVE32F-NEXT:    .cfi_offset s1, -8
 ; RV32ZVE32F-NEXT:    andi s0, t0, 64
-; RV32ZVE32F-NEXT:    beqz s0, .LBB49_7
-; RV32ZVE32F-NEXT:  .LBB49_15: # %cond.load16
+; RV32ZVE32F-NEXT:    beqz s0, .LBB49_16
+; RV32ZVE32F-NEXT:  # %bb.14: # %cond.load16
 ; RV32ZVE32F-NEXT:    vsetivli zero, 1, e32, m2, ta, ma
 ; RV32ZVE32F-NEXT:    vslidedown.vi v10, v8, 6
 ; RV32ZVE32F-NEXT:    vmv.x.s s1, v10
 ; RV32ZVE32F-NEXT:    lw s0, 4(s1)
 ; RV32ZVE32F-NEXT:    lw s1, 0(s1)
 ; RV32ZVE32F-NEXT:    andi t0, t0, -128
-; RV32ZVE32F-NEXT:    beqz t0, .LBB49_8
-; RV32ZVE32F-NEXT:  .LBB49_16: # %cond.load19
+; RV32ZVE32F-NEXT:    bnez t0, .LBB49_17
+; RV32ZVE32F-NEXT:  .LBB49_15:
+; RV32ZVE32F-NEXT:    lw t0, 60(a2)
+; RV32ZVE32F-NEXT:    lw a2, 56(a2)
+; RV32ZVE32F-NEXT:    j .LBB49_18
+; RV32ZVE32F-NEXT:  .LBB49_16:
+; RV32ZVE32F-NEXT:    lw s0, 52(a2)
+; RV32ZVE32F-NEXT:    lw s1, 48(a2)
+; RV32ZVE32F-NEXT:    andi t0, t0, -128
+; RV32ZVE32F-NEXT:    beqz t0, .LBB49_15
+; RV32ZVE32F-NEXT:  .LBB49_17: # %cond.load19
 ; RV32ZVE32F-NEXT:    vsetivli zero, 1, e32, m2, ta, ma
 ; RV32ZVE32F-NEXT:    vslidedown.vi v8, v8, 7
 ; RV32ZVE32F-NEXT:    vmv.x.s a2, v8
 ; RV32ZVE32F-NEXT:    lw t0, 4(a2)
 ; RV32ZVE32F-NEXT:    lw a2, 0(a2)
-; RV32ZVE32F-NEXT:  .LBB49_17: # %else20
+; RV32ZVE32F-NEXT:  .LBB49_18: # %else20
 ; RV32ZVE32F-NEXT:    sw a3, 0(a0)
 ; RV32ZVE32F-NEXT:    sw a1, 4(a0)
 ; RV32ZVE32F-NEXT:    sw a5, 8(a0)
 ; RV32ZVE32F-NEXT:    sw a4, 12(a0)
 ; RV32ZVE32F-NEXT:    sw a7, 16(a0)
 ; RV32ZVE32F-NEXT:    sw a6, 20(a0)
 ; RV32ZVE32F-NEXT:    sw t2, 24(a0)
 ; RV32ZVE32F-NEXT:    sw t1, 28(a0)

 ; RV64V-NEXT:    vzext.vf8 v16, v8
 ; RV64V-NEXT:    vsll.vi v8, v16, 3
 ; RV64V-NEXT:    vluxei64.v v12, (a0), v8, v0.t
 ; RV64V-NEXT:    vmv.v.v v8, v12
 ; RV64V-NEXT:    ret
 ;
 ; RV32ZVE32F-LABEL: mgather_baseidx_zext_v8i8_v8i64:
 ; RV32ZVE32F:       # %bb.0:
-; RV32ZVE32F-NEXT:    addi sp, sp, -16
-; RV32ZVE32F-NEXT:    .cfi_def_cfa_offset 16
-; RV32ZVE32F-NEXT:    sw s0, 12(sp) # 4-byte Folded Spill
-; RV32ZVE32F-NEXT:    sw s1, 8(sp) # 4-byte Folded Spill
-; RV32ZVE32F-NEXT:    .cfi_offset s0, -4
-; RV32ZVE32F-NEXT:    .cfi_offset s1, -8
 ; RV32ZVE32F-NEXT:    vsetivli zero, 8, e32, m2, ta, ma
 ; RV32ZVE32F-NEXT:    vzext.vf4 v10, v8
 ; RV32ZVE32F-NEXT:    vsll.vi v8, v10, 3
 ; RV32ZVE32F-NEXT:    vadd.vx v8, v8, a1
 ; RV32ZVE32F-NEXT:    vsetivli zero, 0, e8, mf4, ta, ma
 ; RV32ZVE32F-NEXT:    vmv.x.s t0, v0
 ; RV32ZVE32F-NEXT:    andi a1, t0, 1
-; RV32ZVE32F-NEXT:    beqz a1, .LBB50_9
+; RV32ZVE32F-NEXT:    beqz a1, .LBB50_7
 ; RV32ZVE32F-NEXT:  # %bb.1: # %cond.load
 ; RV32ZVE32F-NEXT:    vsetivli zero, 0, e32, m2, ta, ma
 ; RV32ZVE32F-NEXT:    vmv.x.s a3, v8
 ; RV32ZVE32F-NEXT:    lw a1, 4(a3)
 ; RV32ZVE32F-NEXT:    lw a3, 0(a3)
 ; RV32ZVE32F-NEXT:    andi a4, t0, 2
-; RV32ZVE32F-NEXT:    bnez a4, .LBB50_10
+; RV32ZVE32F-NEXT:    bnez a4, .LBB50_8
 ; RV32ZVE32F-NEXT:  .LBB50_2:
 ; RV32ZVE32F-NEXT:    lw a4, 12(a2)
 ; RV32ZVE32F-NEXT:    lw a5, 8(a2)
 ; RV32ZVE32F-NEXT:    andi a6, t0, 4
-; RV32ZVE32F-NEXT:    bnez a6, .LBB50_11
+; RV32ZVE32F-NEXT:    bnez a6, .LBB50_9
 ; RV32ZVE32F-NEXT:  .LBB50_3:
 ; RV32ZVE32F-NEXT:    lw a6, 20(a2)
 ; RV32ZVE32F-NEXT:    lw a7, 16(a2)
 ; RV32ZVE32F-NEXT:    andi t1, t0, 8
-; RV32ZVE32F-NEXT:    bnez t1, .LBB50_12
+; RV32ZVE32F-NEXT:    bnez t1, .LBB50_10
 ; RV32ZVE32F-NEXT:  .LBB50_4:
 ; RV32ZVE32F-NEXT:    lw t1, 28(a2)
 ; RV32ZVE32F-NEXT:    lw t2, 24(a2)
 ; RV32ZVE32F-NEXT:    andi t3, t0, 16
-; RV32ZVE32F-NEXT:    bnez t3, .LBB50_13
+; RV32ZVE32F-NEXT:    bnez t3, .LBB50_11
 ; RV32ZVE32F-NEXT:  .LBB50_5:
 ; RV32ZVE32F-NEXT:    lw t3, 36(a2)
 ; RV32ZVE32F-NEXT:    lw t4, 32(a2)
 ; RV32ZVE32F-NEXT:    andi t5, t0, 32
-; RV32ZVE32F-NEXT:    bnez t5, .LBB50_14
+; RV32ZVE32F-NEXT:    bnez t5, .LBB50_12
 ; RV32ZVE32F-NEXT:  .LBB50_6:
 ; RV32ZVE32F-NEXT:    lw t5, 44(a2)
 ; RV32ZVE32F-NEXT:    lw t6, 40(a2)
-; RV32ZVE32F-NEXT:    andi s0, t0, 64
-; RV32ZVE32F-NEXT:    bnez s0, .LBB50_15
+; RV32ZVE32F-NEXT:    j .LBB50_13
 ; RV32ZVE32F-NEXT:  .LBB50_7:
-; RV32ZVE32F-NEXT:    lw s0, 52(a2)
-; RV32ZVE32F-NEXT:    lw s1, 48(a2)
-; RV32ZVE32F-NEXT:    andi t0, t0, -128
-; RV32ZVE32F-NEXT:    bnez t0, .LBB50_16
-; RV32ZVE32F-NEXT:  .LBB50_8:
-; RV32ZVE32F-NEXT:    lw t0, 60(a2)
-; RV32ZVE32F-NEXT:    lw a2, 56(a2)
-; RV32ZVE32F-NEXT:    j .LBB50_17
-; RV32ZVE32F-NEXT:  .LBB50_9:
 ; RV32ZVE32F-NEXT:    lw a1, 4(a2)
 ; RV32ZVE32F-NEXT:    lw a3, 0(a2)
 ; RV32ZVE32F-NEXT:    andi a4, t0, 2
 ; RV32ZVE32F-NEXT:    beqz a4, .LBB50_2
-; RV32ZVE32F-NEXT:  .LBB50_10: # %cond.load1
+; RV32ZVE32F-NEXT:  .LBB50_8: # %cond.load1
 ; RV32ZVE32F-NEXT:    vsetivli zero, 1, e32, m2, ta, ma
 ; RV32ZVE32F-NEXT:    vslidedown.vi v10, v8, 1
 ; RV32ZVE32F-NEXT:    vmv.x.s a5, v10
 ; RV32ZVE32F-NEXT:    lw a4, 4(a5)
 ; RV32ZVE32F-NEXT:    lw a5, 0(a5)
 ; RV32ZVE32F-NEXT:    andi a6, t0, 4
 ; RV32ZVE32F-NEXT:    beqz a6, .LBB50_3
-; RV32ZVE32F-NEXT:  .LBB50_11: # %cond.load4
+; RV32ZVE32F-NEXT:  .LBB50_9: # %cond.load4
 ; RV32ZVE32F-NEXT:    vsetivli zero, 1, e32, m2, ta, ma
 ; RV32ZVE32F-NEXT:    vslidedown.vi v10, v8, 2
 ; RV32ZVE32F-NEXT:    vmv.x.s a7, v10
 ; RV32ZVE32F-NEXT:    lw a6, 4(a7)
 ; RV32ZVE32F-NEXT:    lw a7, 0(a7)
 ; RV32ZVE32F-NEXT:    andi t1, t0, 8
 ; RV32ZVE32F-NEXT:    beqz t1, .LBB50_4
-; RV32ZVE32F-NEXT:  .LBB50_12: # %cond.load7
+; RV32ZVE32F-NEXT:  .LBB50_10: # %cond.load7
 ; RV32ZVE32F-NEXT:    vsetivli zero, 1, e32, m2, ta, ma
 ; RV32ZVE32F-NEXT:    vslidedown.vi v10, v8, 3
 ; RV32ZVE32F-NEXT:    vmv.x.s t2, v10
 ; RV32ZVE32F-NEXT:    lw t1, 4(t2)
 ; RV32ZVE32F-NEXT:    lw t2, 0(t2)
 ; RV32ZVE32F-NEXT:    andi t3, t0, 16
 ; RV32ZVE32F-NEXT:    beqz t3, .LBB50_5
-; RV32ZVE32F-NEXT:  .LBB50_13: # %cond.load10
+; RV32ZVE32F-NEXT:  .LBB50_11: # %cond.load10
 ; RV32ZVE32F-NEXT:    vsetivli zero, 1, e32, m2, ta, ma
 ; RV32ZVE32F-NEXT:    vslidedown.vi v10, v8, 4
 ; RV32ZVE32F-NEXT:    vmv.x.s t4, v10
 ; RV32ZVE32F-NEXT:    lw t3, 4(t4)
 ; RV32ZVE32F-NEXT:    lw t4, 0(t4)
 ; RV32ZVE32F-NEXT:    andi t5, t0, 32
 ; RV32ZVE32F-NEXT:    beqz t5, .LBB50_6
-; RV32ZVE32F-NEXT:  .LBB50_14: # %cond.load13
+; RV32ZVE32F-NEXT:  .LBB50_12: # %cond.load13
 ; RV32ZVE32F-NEXT:    vsetivli zero, 1, e32, m2, ta, ma
 ; RV32ZVE32F-NEXT:    vslidedown.vi v10, v8, 5
 ; RV32ZVE32F-NEXT:    vmv.x.s t6, v10
 ; RV32ZVE32F-NEXT:    lw t5, 4(t6)
 ; RV32ZVE32F-NEXT:    lw t6, 0(t6)
+; RV32ZVE32F-NEXT:  .LBB50_13: # %else14
+; RV32ZVE32F-NEXT:    addi sp, sp, -16
+; RV32ZVE32F-NEXT:    .cfi_def_cfa_offset 16
+; RV32ZVE32F-NEXT:    sw s0, 12(sp) # 4-byte Folded Spill
+; RV32ZVE32F-NEXT:    sw s1, 8(sp) # 4-byte Folded Spill
+; RV32ZVE32F-NEXT:    .cfi_offset s0, -4
+; RV32ZVE32F-NEXT:    .cfi_offset s1, -8
 ; RV32ZVE32F-NEXT:    andi s0, t0, 64
-; RV32ZVE32F-NEXT:    beqz s0, .LBB50_7
-; RV32ZVE32F-NEXT:  .LBB50_15: # %cond.load16
+; RV32ZVE32F-NEXT:    beqz s0, .LBB50_16
+; RV32ZVE32F-NEXT:  # %bb.14: # %cond.load16
 ; RV32ZVE32F-NEXT:    vsetivli zero, 1, e32, m2, ta, ma
 ; RV32ZVE32F-NEXT:    vslidedown.vi v10, v8, 6
 ; RV32ZVE32F-NEXT:    vmv.x.s s1, v10
 ; RV32ZVE32F-NEXT:    lw s0, 4(s1)
 ; RV32ZVE32F-NEXT:    lw s1, 0(s1)
 ; RV32ZVE32F-NEXT:    andi t0, t0, -128
-; RV32ZVE32F-NEXT:    beqz t0, .LBB50_8
-; RV32ZVE32F-NEXT:  .LBB50_16: # %cond.load19
+; RV32ZVE32F-NEXT:    bnez t0, .LBB50_17
+; RV32ZVE32F-NEXT:  .LBB50_15:
+; RV32ZVE32F-NEXT:    lw t0, 60(a2)
+; RV32ZVE32F-NEXT:    lw a2, 56(a2)
+; RV32ZVE32F-NEXT:    j .LBB50_18
+; RV32ZVE32F-NEXT:  .LBB50_16:
+; RV32ZVE32F-NEXT:    lw s0, 52(a2)
+; RV32ZVE32F-NEXT:    lw s1, 48(a2)
+; RV32ZVE32F-NEXT:    andi t0, t0, -128
+; RV32ZVE32F-NEXT:    beqz t0, .LBB50_15
+; RV32ZVE32F-NEXT:  .LBB50_17: # %cond.load19
 ; RV32ZVE32F-NEXT:    vsetivli zero, 1, e32, m2, ta, ma
 ; RV32ZVE32F-NEXT:    vslidedown.vi v8, v8, 7
 ; RV32ZVE32F-NEXT:    vmv.x.s a2, v8
 ; RV32ZVE32F-NEXT:    lw t0, 4(a2)
 ; RV32ZVE32F-NEXT:    lw a2, 0(a2)
-; RV32ZVE32F-NEXT:  .LBB50_17: # %else20
+; RV32ZVE32F-NEXT:  .LBB50_18: # %else20
 ; RV32ZVE32F-NEXT:    sw a3, 0(a0)
 ; RV32ZVE32F-NEXT:    sw a1, 4(a0)
 ; RV32ZVE32F-NEXT:    sw a5, 8(a0)
 ; RV32ZVE32F-NEXT:    sw a4, 12(a0)
 ; RV32ZVE32F-NEXT:    sw a7, 16(a0)
 ; RV32ZVE32F-NEXT:    sw a6, 20(a0)
 ; RV32ZVE32F-NEXT:    sw t2, 24(a0)
 ; RV32ZVE32F-NEXT:    sw t1, 28(a0)

 ; RV64V-NEXT:    vsext.vf4 v16, v8
 ; RV64V-NEXT:    vsll.vi v8, v16, 3
 ; RV64V-NEXT:    vluxei64.v v12, (a0), v8, v0.t
 ; RV64V-NEXT:    vmv.v.v v8, v12
 ; RV64V-NEXT:    ret
 ;
 ; RV32ZVE32F-LABEL: mgather_baseidx_v8i16_v8i64:
 ; RV32ZVE32F:       # %bb.0:
-; RV32ZVE32F-NEXT:    addi sp, sp, -16
-; RV32ZVE32F-NEXT:    .cfi_def_cfa_offset 16
-; RV32ZVE32F-NEXT:    sw s0, 12(sp) # 4-byte Folded Spill
-; RV32ZVE32F-NEXT:    sw s1, 8(sp) # 4-byte Folded Spill
-; RV32ZVE32F-NEXT:    .cfi_offset s0, -4
-; RV32ZVE32F-NEXT:    .cfi_offset s1, -8
 ; RV32ZVE32F-NEXT:    vsetivli zero, 8, e32, m2, ta, ma
 ; RV32ZVE32F-NEXT:    vsext.vf2 v10, v8
 ; RV32ZVE32F-NEXT:    vsll.vi v8, v10, 3
 ; RV32ZVE32F-NEXT:    vadd.vx v8, v8, a1
 ; RV32ZVE32F-NEXT:    vsetivli zero, 0, e8, mf4, ta, ma
 ; RV32ZVE32F-NEXT:    vmv.x.s t0, v0
 ; RV32ZVE32F-NEXT:    andi a1, t0, 1
-; RV32ZVE32F-NEXT:    beqz a1, .LBB51_9
+; RV32ZVE32F-NEXT:    beqz a1, .LBB51_7
 ; RV32ZVE32F-NEXT:  # %bb.1: # %cond.load
 ; RV32ZVE32F-NEXT:    vsetivli zero, 0, e32, m2, ta, ma
 ; RV32ZVE32F-NEXT:    vmv.x.s a3, v8
 ; RV32ZVE32F-NEXT:    lw a1, 4(a3)
 ; RV32ZVE32F-NEXT:    lw a3, 0(a3)
 ; RV32ZVE32F-NEXT:    andi a4, t0, 2
-; RV32ZVE32F-NEXT:    bnez a4, .LBB51_10
+; RV32ZVE32F-NEXT:    bnez a4, .LBB51_8
 ; RV32ZVE32F-NEXT:  .LBB51_2:
 ; RV32ZVE32F-NEXT:    lw a4, 12(a2)
 ; RV32ZVE32F-NEXT:    lw a5, 8(a2)
 ; RV32ZVE32F-NEXT:    andi a6, t0, 4
-; RV32ZVE32F-NEXT:    bnez a6, .LBB51_11
+; RV32ZVE32F-NEXT:    bnez a6, .LBB51_9
 ; RV32ZVE32F-NEXT:  .LBB51_3:
 ; RV32ZVE32F-NEXT:    lw a6, 20(a2)
 ; RV32ZVE32F-NEXT:    lw a7, 16(a2)
 ; RV32ZVE32F-NEXT:    andi t1, t0, 8
-; RV32ZVE32F-NEXT:    bnez t1, .LBB51_12
+; RV32ZVE32F-NEXT:    bnez t1, .LBB51_10
 ; RV32ZVE32F-NEXT:  .LBB51_4:
 ; RV32ZVE32F-NEXT:    lw t1, 28(a2)
 ; RV32ZVE32F-NEXT:    lw t2, 24(a2)
 ; RV32ZVE32F-NEXT:    andi t3, t0, 16
-; RV32ZVE32F-NEXT:    bnez t3, .LBB51_13
+; RV32ZVE32F-NEXT:    bnez t3, .LBB51_11
 ; RV32ZVE32F-NEXT:  .LBB51_5:
 ; RV32ZVE32F-NEXT:    lw t3, 36(a2)
 ; RV32ZVE32F-NEXT:    lw t4, 32(a2)
 ; RV32ZVE32F-NEXT:    andi t5, t0, 32
-; RV32ZVE32F-NEXT:    bnez t5, .LBB51_14
+; RV32ZVE32F-NEXT:    bnez t5, .LBB51_12
 ; RV32ZVE32F-NEXT:  .LBB51_6:
 ; RV32ZVE32F-NEXT:    lw t5, 44(a2)
 ; RV32ZVE32F-NEXT:    lw t6, 40(a2)
-; RV32ZVE32F-NEXT:    andi s0, t0, 64
-; RV32ZVE32F-NEXT:    bnez s0, .LBB51_15
+; RV32ZVE32F-NEXT:    j .LBB51_13
 ; RV32ZVE32F-NEXT:  .LBB51_7:
-; RV32ZVE32F-NEXT:    lw s0, 52(a2)
-; RV32ZVE32F-NEXT:    lw s1, 48(a2)
-; RV32ZVE32F-NEXT:    andi t0, t0, -128
-; RV32ZVE32F-NEXT:    bnez t0, .LBB51_16
-; RV32ZVE32F-NEXT:  .LBB51_8:
-; RV32ZVE32F-NEXT:    lw t0, 60(a2)
-; RV32ZVE32F-NEXT:    lw a2, 56(a2)
-; RV32ZVE32F-NEXT:    j .LBB51_17
-; RV32ZVE32F-NEXT:  .LBB51_9:
 ; RV32ZVE32F-NEXT:    lw a1, 4(a2)
 ; RV32ZVE32F-NEXT:    lw a3, 0(a2)
 ; RV32ZVE32F-NEXT:    andi a4, t0, 2
 ; RV32ZVE32F-NEXT:    beqz a4, .LBB51_2
-; RV32ZVE32F-NEXT:  .LBB51_10: # %cond.load1
+; RV32ZVE32F-NEXT:  .LBB51_8: # %cond.load1
 ; RV32ZVE32F-NEXT:    vsetivli zero, 1, e32, m2, ta, ma
 ; RV32ZVE32F-NEXT:    vslidedown.vi v10, v8, 1
 ; RV32ZVE32F-NEXT:    vmv.x.s a5, v10
 ; RV32ZVE32F-NEXT:    lw a4, 4(a5)
 ; RV32ZVE32F-NEXT:    lw a5, 0(a5)
 ; RV32ZVE32F-NEXT:    andi a6, t0, 4
 ; RV32ZVE32F-NEXT:    beqz a6, .LBB51_3
-; RV32ZVE32F-NEXT:  .LBB51_11: # %cond.load4
+; RV32ZVE32F-NEXT:  .LBB51_9: # %cond.load4
 ; RV32ZVE32F-NEXT:    vsetivli zero, 1, e32, m2, ta, ma
 ; RV32ZVE32F-NEXT:    vslidedown.vi v10, v8, 2
 ; RV32ZVE32F-NEXT:    vmv.x.s a7, v10
 ; RV32ZVE32F-NEXT:    lw a6, 4(a7)
 ; RV32ZVE32F-NEXT:    lw a7, 0(a7)
 ; RV32ZVE32F-NEXT:    andi t1, t0, 8
 ; RV32ZVE32F-NEXT:    beqz t1, .LBB51_4
-; RV32ZVE32F-NEXT:  .LBB51_12: # %cond.load7
+; RV32ZVE32F-NEXT:  .LBB51_10: # %cond.load7
 ; RV32ZVE32F-NEXT:    vsetivli zero, 1, e32, m2, ta, ma
 ; RV32ZVE32F-NEXT:    vslidedown.vi v10, v8, 3
 ; RV32ZVE32F-NEXT:    vmv.x.s t2, v10
 ; RV32ZVE32F-NEXT:    lw t1, 4(t2)
 ; RV32ZVE32F-NEXT:    lw t2, 0(t2)
 ; RV32ZVE32F-NEXT:    andi t3, t0, 16
 ; RV32ZVE32F-NEXT:    beqz t3, .LBB51_5
-; RV32ZVE32F-NEXT:  .LBB51_13: # %cond.load10
+; RV32ZVE32F-NEXT:  .LBB51_11: # %cond.load10
 ; RV32ZVE32F-NEXT:    vsetivli zero, 1, e32, m2, ta, ma
 ; RV32ZVE32F-NEXT:    vslidedown.vi v10, v8, 4
 ; RV32ZVE32F-NEXT:    vmv.x.s t4, v10
 ; RV32ZVE32F-NEXT:    lw t3, 4(t4)
 ; RV32ZVE32F-NEXT:    lw t4, 0(t4)
 ; RV32ZVE32F-NEXT:    andi t5, t0, 32
 ; RV32ZVE32F-NEXT:    beqz t5, .LBB51_6
-; RV32ZVE32F-NEXT:  .LBB51_14: # %cond.load13
+; RV32ZVE32F-NEXT:  .LBB51_12: # %cond.load13
 ; RV32ZVE32F-NEXT:    vsetivli zero, 1, e32, m2, ta, ma
 ; RV32ZVE32F-NEXT:    vslidedown.vi v10, v8, 5
 ; RV32ZVE32F-NEXT:    vmv.x.s t6, v10
 ; RV32ZVE32F-NEXT:    lw t5, 4(t6)
 ; RV32ZVE32F-NEXT:    lw t6, 0(t6)
+; RV32ZVE32F-NEXT:  .LBB51_13: # %else14
+; RV32ZVE32F-NEXT:    addi sp, sp, -16
+; RV32ZVE32F-NEXT:    .cfi_def_cfa_offset 16
+; RV32ZVE32F-NEXT:    sw s0, 12(sp) # 4-byte Folded Spill
+; RV32ZVE32F-NEXT:    sw s1, 8(sp) # 4-byte Folded Spill
+; RV32ZVE32F-NEXT:    .cfi_offset s0, -4
+; RV32ZVE32F-NEXT:    .cfi_offset s1, -8
 ; RV32ZVE32F-NEXT:    andi s0, t0, 64
-; RV32ZVE32F-NEXT:    beqz s0, .LBB51_7
-; RV32ZVE32F-NEXT:  .LBB51_15: # %cond.load16
+; RV32ZVE32F-NEXT:    beqz s0, .LBB51_16
+; RV32ZVE32F-NEXT:  # %bb.14: # %cond.load16
 ; RV32ZVE32F-NEXT:    vsetivli zero, 1, e32, m2, ta, ma
 ; RV32ZVE32F-NEXT:    vslidedown.vi v10, v8, 6
 ; RV32ZVE32F-NEXT:    vmv.x.s s1, v10
 ; RV32ZVE32F-NEXT:    lw s0, 4(s1)
 ; RV32ZVE32F-NEXT:    lw s1, 0(s1)
 ; RV32ZVE32F-NEXT:    andi t0, t0, -128
-; RV32ZVE32F-NEXT:    beqz t0, .LBB51_8
-; RV32ZVE32F-NEXT:  .LBB51_16: # %cond.load19
+; RV32ZVE32F-NEXT:    bnez t0, .LBB51_17
+; RV32ZVE32F-NEXT:  .LBB51_15:
+; RV32ZVE32F-NEXT:    lw t0, 60(a2)
+; RV32ZVE32F-NEXT:    lw a2, 56(a2)
+; RV32ZVE32F-NEXT:    j .LBB51_18
+; RV32ZVE32F-NEXT:  .LBB51_16:
+; RV32ZVE32F-NEXT:    lw s0, 52(a2)
+; RV32ZVE32F-NEXT:    lw s1, 48(a2)
+; RV32ZVE32F-NEXT:    andi t0, t0, -128
+; RV32ZVE32F-NEXT:    beqz t0, .LBB51_15
+; RV32ZVE32F-NEXT:  .LBB51_17: # %cond.load19
 ; RV32ZVE32F-NEXT:    vsetivli zero, 1, e32, m2, ta, ma
 ; RV32ZVE32F-NEXT:    vslidedown.vi v8, v8, 7
 ; RV32ZVE32F-NEXT:    vmv.x.s a2, v8
 ; RV32ZVE32F-NEXT:    lw t0, 4(a2)
 ; RV32ZVE32F-NEXT:    lw a2, 0(a2)
-; RV32ZVE32F-NEXT:  .LBB51_17: # %else20
+; RV32ZVE32F-NEXT:  .LBB51_18: # %else20
 ; RV32ZVE32F-NEXT:    sw a3, 0(a0)
 ; RV32ZVE32F-NEXT:    sw a1, 4(a0)
 ; RV32ZVE32F-NEXT:    sw a5, 8(a0)
 ; RV32ZVE32F-NEXT:    sw a4, 12(a0)
 ; RV32ZVE32F-NEXT:    sw a7, 16(a0)
 ; RV32ZVE32F-NEXT:    sw a6, 20(a0)
 ; RV32ZVE32F-NEXT:    sw t2, 24(a0)
 ; RV32ZVE32F-NEXT:    sw t1, 28(a0)

 ; RV64V-NEXT:    vsext.vf4 v16, v8
 ; RV64V-NEXT:    vsll.vi v8, v16, 3
 ; RV64V-NEXT:    vluxei64.v v12, (a0), v8, v0.t
 ; RV64V-NEXT:    vmv.v.v v8, v12
 ; RV64V-NEXT:    ret
 ;
 ; RV32ZVE32F-LABEL: mgather_baseidx_sext_v8i16_v8i64:
 ; RV32ZVE32F:       # %bb.0:
-; RV32ZVE32F-NEXT:    addi sp, sp, -16
-; RV32ZVE32F-NEXT:    .cfi_def_cfa_offset 16
-; RV32ZVE32F-NEXT:    sw s0, 12(sp) # 4-byte Folded Spill
-; RV32ZVE32F-NEXT:    sw s1, 8(sp) # 4-byte Folded Spill
-; RV32ZVE32F-NEXT:    .cfi_offset s0, -4
-; RV32ZVE32F-NEXT:    .cfi_offset s1, -8
 ; RV32ZVE32F-NEXT:    vsetivli zero, 8, e32, m2, ta, ma
 ; RV32ZVE32F-NEXT:    vsext.vf2 v10, v8
 ; RV32ZVE32F-NEXT:    vsll.vi v8, v10, 3
 ; RV32ZVE32F-NEXT:    vadd.vx v8, v8, a1
 ; RV32ZVE32F-NEXT:    vsetivli zero, 0, e8, mf4, ta, ma
 ; RV32ZVE32F-NEXT:    vmv.x.s t0, v0
 ; RV32ZVE32F-NEXT:    andi a1, t0, 1
-; RV32ZVE32F-NEXT:    beqz a1, .LBB52_9
+; RV32ZVE32F-NEXT:    beqz a1, .LBB52_7
 ; RV32ZVE32F-NEXT:  # %bb.1: # %cond.load
 ; RV32ZVE32F-NEXT:    vsetivli zero, 0, e32, m2, ta, ma
 ; RV32ZVE32F-NEXT:    vmv.x.s a3, v8
 ; RV32ZVE32F-NEXT:    lw a1, 4(a3)
 ; RV32ZVE32F-NEXT:    lw a3, 0(a3)
 ; RV32ZVE32F-NEXT:    andi a4, t0, 2
-; RV32ZVE32F-NEXT:    bnez a4, .LBB52_10
+; RV32ZVE32F-NEXT:    bnez a4, .LBB52_8
 ; RV32ZVE32F-NEXT:  .LBB52_2:
 ; RV32ZVE32F-NEXT:    lw a4, 12(a2)
 ; RV32ZVE32F-NEXT:    lw a5, 8(a2)
 ; RV32ZVE32F-NEXT:    andi a6, t0, 4
-; RV32ZVE32F-NEXT:    bnez a6, .LBB52_11
+; RV32ZVE32F-NEXT:    bnez a6, .LBB52_9
 ; RV32ZVE32F-NEXT:  .LBB52_3:
 ; RV32ZVE32F-NEXT:    lw a6, 20(a2)
 ; RV32ZVE32F-NEXT:    lw a7, 16(a2)
 ; RV32ZVE32F-NEXT:    andi t1, t0, 8
-; RV32ZVE32F-NEXT:    bnez t1, .LBB52_12
+; RV32ZVE32F-NEXT:    bnez t1, .LBB52_10
 ; RV32ZVE32F-NEXT:  .LBB52_4:
 ; RV32ZVE32F-NEXT:    lw t1, 28(a2)
 ; RV32ZVE32F-NEXT:    lw t2, 24(a2)
 ; RV32ZVE32F-NEXT:    andi t3, t0, 16
-; RV32ZVE32F-NEXT:    bnez t3, .LBB52_13
+; RV32ZVE32F-NEXT:    bnez t3, .LBB52_11
 ; RV32ZVE32F-NEXT:  .LBB52_5:
 ; RV32ZVE32F-NEXT:    lw t3, 36(a2)
 ; RV32ZVE32F-NEXT:    lw t4, 32(a2)
 ; RV32ZVE32F-NEXT:    andi t5, t0, 32
-; RV32ZVE32F-NEXT:    bnez t5, .LBB52_14
+; RV32ZVE32F-NEXT:    bnez t5, .LBB52_12
 ; RV32ZVE32F-NEXT:  .LBB52_6:
 ; RV32ZVE32F-NEXT:    lw t5, 44(a2)
 ; RV32ZVE32F-NEXT:    lw t6, 40(a2)
-; RV32ZVE32F-NEXT:    andi s0, t0, 64
-; RV32ZVE32F-NEXT:    bnez s0, .LBB52_15
+; RV32ZVE32F-NEXT:    j .LBB52_13
 ; RV32ZVE32F-NEXT:  .LBB52_7:
-; RV32ZVE32F-NEXT:    lw s0, 52(a2)
-; RV32ZVE32F-NEXT:    lw s1, 48(a2)
-; RV32ZVE32F-NEXT:    andi t0, t0, -128
-; RV32ZVE32F-NEXT:    bnez t0, .LBB52_16
-; RV32ZVE32F-NEXT:  .LBB52_8:
-; RV32ZVE32F-NEXT:    lw t0, 60(a2)
-; RV32ZVE32F-NEXT:    lw a2, 56(a2)
-; RV32ZVE32F-NEXT:    j .LBB52_17
-; RV32ZVE32F-NEXT:  .LBB52_9:
 ; RV32ZVE32F-NEXT:    lw a1, 4(a2)
 ; RV32ZVE32F-NEXT:    lw a3, 0(a2)
 ; RV32ZVE32F-NEXT:    andi a4, t0, 2
 ; RV32ZVE32F-NEXT:    beqz a4, .LBB52_2
-; RV32ZVE32F-NEXT:  .LBB52_10: # %cond.load1
+; RV32ZVE32F-NEXT:  .LBB52_8: # %cond.load1
 ; RV32ZVE32F-NEXT:    vsetivli zero, 1, e32, m2, ta, ma
 ; RV32ZVE32F-NEXT:    vslidedown.vi v10, v8, 1
 ; RV32ZVE32F-NEXT:    vmv.x.s a5, v10
 ; RV32ZVE32F-NEXT:    lw a4, 4(a5)
 ; RV32ZVE32F-NEXT:    lw a5, 0(a5)
 ; RV32ZVE32F-NEXT:    andi a6, t0, 4
 ; RV32ZVE32F-NEXT:    beqz a6, .LBB52_3
-; RV32ZVE32F-NEXT:  .LBB52_11: # %cond.load4
+; RV32ZVE32F-NEXT:  .LBB52_9: # %cond.load4
 ; RV32ZVE32F-NEXT:    vsetivli zero, 1, e32, m2, ta, ma
 ; RV32ZVE32F-NEXT:    vslidedown.vi v10, v8, 2
 ; RV32ZVE32F-NEXT:    vmv.x.s a7, v10
 ; RV32ZVE32F-NEXT:    lw a6, 4(a7)
 ; RV32ZVE32F-NEXT:    lw a7, 0(a7)
 ; RV32ZVE32F-NEXT:    andi t1, t0, 8
 ; RV32ZVE32F-NEXT:    beqz t1, .LBB52_4
-; RV32ZVE32F-NEXT:  .LBB52_12: # %cond.load7
+; RV32ZVE32F-NEXT:  .LBB52_10: # %cond.load7
 ; RV32ZVE32F-NEXT:    vsetivli zero, 1, e32, m2, ta, ma
 ; RV32ZVE32F-NEXT:    vslidedown.vi v10, v8, 3
 ; RV32ZVE32F-NEXT:    vmv.x.s t2, v10
 ; RV32ZVE32F-NEXT:    lw t1, 4(t2)
 ; RV32ZVE32F-NEXT:    lw t2, 0(t2)
 ; RV32ZVE32F-NEXT:    andi t3, t0, 16
 ; RV32ZVE32F-NEXT:    beqz t3, .LBB52_5
-; RV32ZVE32F-NEXT:  .LBB52_13: # %cond.load10
+; RV32ZVE32F-NEXT:  .LBB52_11: # %cond.load10
 ; RV32ZVE32F-NEXT:    vsetivli zero, 1, e32, m2, ta, ma
 ; RV32ZVE32F-NEXT:    vslidedown.vi v10, v8, 4
 ; RV32ZVE32F-NEXT:    vmv.x.s t4, v10
 ; RV32ZVE32F-NEXT:    lw t3, 4(t4)
 ; RV32ZVE32F-NEXT:    lw t4, 0(t4)
 ; RV32ZVE32F-NEXT:    andi t5, t0, 32
 ; RV32ZVE32F-NEXT:    beqz t5, .LBB52_6
-; RV32ZVE32F-NEXT:  .LBB52_14: # %cond.load13
+; RV32ZVE32F-NEXT:  .LBB52_12: # %cond.load13
 ; RV32ZVE32F-NEXT:    vsetivli zero, 1, e32, m2, ta, ma
 ; RV32ZVE32F-NEXT:    vslidedown.vi v10, v8, 5
 ; RV32ZVE32F-NEXT:    vmv.x.s t6, v10
 ; RV32ZVE32F-NEXT:    lw t5, 4(t6)
 ; RV32ZVE32F-NEXT:    lw t6, 0(t6)
+; RV32ZVE32F-NEXT:  .LBB52_13: # %else14
+; RV32ZVE32F-NEXT:    addi sp, sp, -16
+; RV32ZVE32F-NEXT:    .cfi_def_cfa_offset 16
+; RV32ZVE32F-NEXT:    sw s0, 12(sp) # 4-byte Folded Spill
+; RV32ZVE32F-NEXT:    sw s1, 8(sp) # 4-byte Folded Spill
+; RV32ZVE32F-NEXT:    .cfi_offset s0, -4
+; RV32ZVE32F-NEXT:    .cfi_offset s1, -8
 ; RV32ZVE32F-NEXT:    andi s0, t0, 64
-; RV32ZVE32F-NEXT:    beqz s0, .LBB52_7
-; RV32ZVE32F-NEXT:  .LBB52_15: # %cond.load16
+; RV32ZVE32F-NEXT:    beqz s0, .LBB52_16
+; RV32ZVE32F-NEXT:  # %bb.14: # %cond.load16
 ; RV32ZVE32F-NEXT:    vsetivli zero, 1, e32, m2, ta, ma
 ; RV32ZVE32F-NEXT:    vslidedown.vi v10, v8, 6
 ; RV32ZVE32F-NEXT:    vmv.x.s s1, v10
 ; RV32ZVE32F-NEXT:    lw s0, 4(s1)
 ; RV32ZVE32F-NEXT:    lw s1, 0(s1)
 ; RV32ZVE32F-NEXT:    andi t0, t0, -128
-; RV32ZVE32F-NEXT:    beqz t0, .LBB52_8
-; RV32ZVE32F-NEXT:  .LBB52_16: # %cond.load19
+; RV32ZVE32F-NEXT:    bnez t0, .LBB52_17
+; RV32ZVE32F-NEXT:  .LBB52_15:
+; RV32ZVE32F-NEXT:    lw t0, 60(a2)
+; RV32ZVE32F-NEXT:    lw a2, 56(a2)
+; RV32ZVE32F-NEXT:    j .LBB52_18
+; RV32ZVE32F-NEXT:  .LBB52_16:
+; RV32ZVE32F-NEXT:    lw s0, 52(a2)
+; RV32ZVE32F-NEXT:    lw s1, 48(a2)
+; RV32ZVE32F-NEXT:    andi t0, t0, -128
+; RV32ZVE32F-NEXT:    beqz t0, .LBB52_15
+; RV32ZVE32F-NEXT:  .LBB52_17: # %cond.load19
 ; RV32ZVE32F-NEXT:    vsetivli zero, 1, e32, m2, ta, ma
 ; RV32ZVE32F-NEXT:    vslidedown.vi v8, v8, 7
 ; RV32ZVE32F-NEXT:    vmv.x.s a2, v8
 ; RV32ZVE32F-NEXT:    lw t0, 4(a2)
 ; RV32ZVE32F-NEXT:    lw a2, 0(a2)
-; RV32ZVE32F-NEXT:  .LBB52_17: # %else20
+; RV32ZVE32F-NEXT:  .LBB52_18: # %else20
 ; RV32ZVE32F-NEXT:    sw a3, 0(a0)
 ; RV32ZVE32F-NEXT:    sw a1, 4(a0)
 ; RV32ZVE32F-NEXT:    sw a5, 8(a0)
 ; RV32ZVE32F-NEXT:    sw a4, 12(a0)
 ; RV32ZVE32F-NEXT:    sw a7, 16(a0)
 ; RV32ZVE32F-NEXT:    sw a6, 20(a0)
 ; RV32ZVE32F-NEXT:    sw t2, 24(a0)
 ; RV32ZVE32F-NEXT:    sw t1, 28(a0)

 ; RV64V-NEXT:    vzext.vf4 v16, v8
 ; RV64V-NEXT:    vsll.vi v8, v16, 3
 ; RV64V-NEXT:    vluxei64.v v12, (a0), v8, v0.t
 ; RV64V-NEXT:    vmv.v.v v8, v12
 ; RV64V-NEXT:    ret
 ;
 ; RV32ZVE32F-LABEL: mgather_baseidx_zext_v8i16_v8i64:
 ; RV32ZVE32F:       # %bb.0:
-; RV32ZVE32F-NEXT:    addi sp, sp, -16
-; RV32ZVE32F-NEXT:    .cfi_def_cfa_offset 16
-; RV32ZVE32F-NEXT:    sw s0, 12(sp) # 4-byte Folded Spill
-; RV32ZVE32F-NEXT:    sw s1, 8(sp) # 4-byte Folded Spill
-; RV32ZVE32F-NEXT:    .cfi_offset s0, -4
-; RV32ZVE32F-NEXT:    .cfi_offset s1, -8
 ; RV32ZVE32F-NEXT:    vsetivli zero, 8, e32, m2, ta, ma
 ; RV32ZVE32F-NEXT:    vzext.vf2 v10, v8
 ; RV32ZVE32F-NEXT:    vsll.vi v8, v10, 3
 ; RV32ZVE32F-NEXT:    vadd.vx v8, v8, a1
 ; RV32ZVE32F-NEXT:    vsetivli zero, 0, e8, mf4, ta, ma
 ; RV32ZVE32F-NEXT:    vmv.x.s t0, v0
 ; RV32ZVE32F-NEXT:    andi a1, t0, 1
-; RV32ZVE32F-NEXT:    beqz a1, .LBB53_9
+; RV32ZVE32F-NEXT:    beqz a1, .LBB53_7
 ; RV32ZVE32F-NEXT:  # %bb.1: # %cond.load
 ; RV32ZVE32F-NEXT:    vsetivli zero, 0, e32, m2, ta, ma
 ; RV32ZVE32F-NEXT:    vmv.x.s a3, v8
 ; RV32ZVE32F-NEXT:    lw a1, 4(a3)
 ; RV32ZVE32F-NEXT:    lw a3, 0(a3)
 ; RV32ZVE32F-NEXT:    andi a4, t0, 2
-; RV32ZVE32F-NEXT:    bnez a4, .LBB53_10
+; RV32ZVE32F-NEXT:    bnez a4, .LBB53_8
 ; RV32ZVE32F-NEXT:  .LBB53_2:
 ; RV32ZVE32F-NEXT:    lw a4, 12(a2)
 ; RV32ZVE32F-NEXT:    lw a5, 8(a2)
 ; RV32ZVE32F-NEXT:    andi a6, t0, 4
-; RV32ZVE32F-NEXT:    bnez a6, .LBB53_11
+; RV32ZVE32F-NEXT:    bnez a6, .LBB53_9
 ; RV32ZVE32F-NEXT:  .LBB53_3:
 ; RV32ZVE32F-NEXT:    lw a6, 20(a2)
 ; RV32ZVE32F-NEXT:    lw a7, 16(a2)
 ; RV32ZVE32F-NEXT:    andi t1, t0, 8
-; RV32ZVE32F-NEXT:    bnez t1, .LBB53_12
+; RV32ZVE32F-NEXT:    bnez t1, .LBB53_10
 ; RV32ZVE32F-NEXT:  .LBB53_4:
 ; RV32ZVE32F-NEXT:    lw t1, 28(a2)
 ; RV32ZVE32F-NEXT:    lw t2, 24(a2)
 ; RV32ZVE32F-NEXT:    andi t3, t0, 16
-; RV32ZVE32F-NEXT:    bnez t3, .LBB53_13
+; RV32ZVE32F-NEXT:    bnez t3, .LBB53_11
 ; RV32ZVE32F-NEXT:  .LBB53_5:
 ; RV32ZVE32F-NEXT:    lw t3, 36(a2)
 ; RV32ZVE32F-NEXT:    lw t4, 32(a2)
 ; RV32ZVE32F-NEXT:    andi t5, t0, 32
-; RV32ZVE32F-NEXT:    bnez t5, .LBB53_14
+; RV32ZVE32F-NEXT:    bnez t5, .LBB53_12
 ; RV32ZVE32F-NEXT:  .LBB53_6:
 ; RV32ZVE32F-NEXT:    lw t5, 44(a2)
 ; RV32ZVE32F-NEXT:    lw t6, 40(a2)
-; RV32ZVE32F-NEXT:    andi s0, t0, 64
-; RV32ZVE32F-NEXT:    bnez s0, .LBB53_15
+; RV32ZVE32F-NEXT:    j .LBB53_13
 ; RV32ZVE32F-NEXT:  .LBB53_7:
-; RV32ZVE32F-NEXT:    lw s0, 52(a2)
-; RV32ZVE32F-NEXT:    lw s1, 48(a2)
-; RV32ZVE32F-NEXT:    andi t0, t0, -128
-; RV32ZVE32F-NEXT:    bnez t0, .LBB53_16
-; RV32ZVE32F-NEXT:  .LBB53_8:
-; RV32ZVE32F-NEXT:    lw t0, 60(a2)
-; RV32ZVE32F-NEXT:    lw a2, 56(a2)
-; RV32ZVE32F-NEXT:    j .LBB53_17
-; RV32ZVE32F-NEXT:  .LBB53_9:
 ; RV32ZVE32F-NEXT:    lw a1, 4(a2)
 ; RV32ZVE32F-NEXT:    lw a3, 0(a2)
 ; RV32ZVE32F-NEXT:    andi a4, t0, 2
 ; RV32ZVE32F-NEXT:    beqz a4, .LBB53_2
-; RV32ZVE32F-NEXT:  .LBB53_10: # %cond.load1
+; RV32ZVE32F-NEXT:  .LBB53_8: # %cond.load1
 ; RV32ZVE32F-NEXT:    vsetivli zero, 1, e32, m2, ta, ma
 ; RV32ZVE32F-NEXT:    vslidedown.vi v10, v8, 1
 ; RV32ZVE32F-NEXT:    vmv.x.s a5, v10
 ; RV32ZVE32F-NEXT:    lw a4, 4(a5)
 ; RV32ZVE32F-NEXT:    lw a5, 0(a5)
 ; RV32ZVE32F-NEXT:    andi a6, t0, 4
 ; RV32ZVE32F-NEXT:    beqz a6, .LBB53_3
-; RV32ZVE32F-NEXT:  .LBB53_11: # %cond.load4
+; RV32ZVE32F-NEXT:  .LBB53_9: # %cond.load4
 ; RV32ZVE32F-NEXT:    vsetivli zero, 1, e32, m2, ta, ma
 ; RV32ZVE32F-NEXT:    vslidedown.vi v10, v8, 2
 ; RV32ZVE32F-NEXT:    vmv.x.s a7, v10
 ; RV32ZVE32F-NEXT:    lw a6, 4(a7)
 ; RV32ZVE32F-NEXT:    lw a7, 0(a7)
 ; RV32ZVE32F-NEXT:    andi t1, t0, 8
 ; RV32ZVE32F-NEXT:    beqz t1, .LBB53_4
-; RV32ZVE32F-NEXT:  .LBB53_12: # %cond.load7
+; RV32ZVE32F-NEXT:  .LBB53_10: # %cond.load7
 ; RV32ZVE32F-NEXT:    vsetivli zero, 1, e32, m2, ta, ma
 ; RV32ZVE32F-NEXT:    vslidedown.vi v10, v8, 3
 ; RV32ZVE32F-NEXT:    vmv.x.s t2, v10
 ; RV32ZVE32F-NEXT:    lw t1, 4(t2)
 ; RV32ZVE32F-NEXT:    lw t2, 0(t2)
 ; RV32ZVE32F-NEXT:    andi t3, t0, 16
 ; RV32ZVE32F-NEXT:    beqz t3, .LBB53_5
-; RV32ZVE32F-NEXT:  .LBB53_13: # %cond.load10
+; RV32ZVE32F-NEXT:  .LBB53_11: # %cond.load10
 ; RV32ZVE32F-NEXT:    vsetivli zero, 1, e32, m2, ta, ma
 ; RV32ZVE32F-NEXT:    vslidedown.vi v10, v8, 4
 ; RV32ZVE32F-NEXT:    vmv.x.s t4, v10
 ; RV32ZVE32F-NEXT:    lw t3, 4(t4)
 ; RV32ZVE32F-NEXT:    lw t4, 0(t4)
 ; RV32ZVE32F-NEXT:    andi t5, t0, 32
 ; RV32ZVE32F-NEXT:    beqz t5, .LBB53_6
-; RV32ZVE32F-NEXT:  .LBB53_14: # %cond.load13
+; RV32ZVE32F-NEXT:  .LBB53_12: # %cond.load13
 ; RV32ZVE32F-NEXT:    vsetivli zero, 1, e32, m2, ta, ma
 ; RV32ZVE32F-NEXT:    vslidedown.vi v10, v8, 5
 ; RV32ZVE32F-NEXT:    vmv.x.s t6, v10
 ; RV32ZVE32F-NEXT:    lw t5, 4(t6)
 ; RV32ZVE32F-NEXT:    lw t6, 0(t6)
+; RV32ZVE32F-NEXT:  .LBB53_13: # %else14
+; RV32ZVE32F-NEXT:    addi sp, sp, -16
+; RV32ZVE32F-NEXT:    .cfi_def_cfa_offset 16
+; RV32ZVE32F-NEXT:    sw s0, 12(sp) # 4-byte Folded Spill
+; RV32ZVE32F-NEXT:    sw s1, 8(sp) # 4-byte Folded Spill
+; RV32ZVE32F-NEXT:    .cfi_offset s0, -4
+; RV32ZVE32F-NEXT:    .cfi_offset s1, -8
 ; RV32ZVE32F-NEXT:    andi s0, t0, 64
-; RV32ZVE32F-NEXT:    beqz s0, .LBB53_7
-; RV32ZVE32F-NEXT:  .LBB53_15: # %cond.load16
+; RV32ZVE32F-NEXT:    beqz s0, .LBB53_16
+; RV32ZVE32F-NEXT:  # %bb.14: # %cond.load16
 ; RV32ZVE32F-NEXT:    vsetivli zero, 1, e32, m2, ta, ma
 ; RV32ZVE32F-NEXT:    vslidedown.vi v10, v8, 6
 ; RV32ZVE32F-NEXT:    vmv.x.s s1, v10
 ; RV32ZVE32F-NEXT:    lw s0, 4(s1)
 ; RV32ZVE32F-NEXT:    lw s1, 0(s1)
 ; RV32ZVE32F-NEXT:    andi t0, t0, -128
-; RV32ZVE32F-NEXT:    beqz t0, .LBB53_8
-; RV32ZVE32F-NEXT:  .LBB53_16: # %cond.load19
+; RV32ZVE32F-NEXT:    bnez t0, .LBB53_17
+; RV32ZVE32F-NEXT:  .LBB53_15:
+; RV32ZVE32F-NEXT:    lw t0, 60(a2)
+; RV32ZVE32F-NEXT:    lw a2, 56(a2)
+; RV32ZVE32F-NEXT:    j .LBB53_18
+; RV32ZVE32F-NEXT:  .LBB53_16:
+; RV32ZVE32F-NEXT:    lw s0, 52(a2)
+; RV32ZVE32F-NEXT:    lw s1, 48(a2)
+; RV32ZVE32F-NEXT:    andi t0, t0, -128
+; RV32ZVE32F-NEXT:    beqz t0, .LBB53_15
+; RV32ZVE32F-NEXT:  .LBB53_17: # %cond.load19
 ; RV32ZVE32F-NEXT:    vsetivli zero, 1, e32, m2, ta, ma
 ; RV32ZVE32F-NEXT:    vslidedown.vi v8, v8, 7
 ; RV32ZVE32F-NEXT:    vmv.x.s a2, v8
 ; RV32ZVE32F-NEXT:    lw t0, 4(a2)
 ; RV32ZVE32F-NEXT:    lw a2, 0(a2)
-; RV32ZVE32F-NEXT:  .LBB53_17: # %else20
+; RV32ZVE32F-NEXT:  .LBB53_18: # %else20
 ; RV32ZVE32F-NEXT:    sw a3, 0(a0)
 ; RV32ZVE32F-NEXT:    sw a1, 4(a0)
 ; RV32ZVE32F-NEXT:    sw a5, 8(a0)
 ; RV32ZVE32F-NEXT:    sw a4, 12(a0)
 ; RV32ZVE32F-NEXT:    sw a7, 16(a0)
 ; RV32ZVE32F-NEXT:    sw a6, 20(a0)
 ; RV32ZVE32F-NEXT:    sw t2, 24(a0)
 ; RV32ZVE32F-NEXT:    sw t1, 28(a0)

 ; RV64V-NEXT:    vsext.vf2 v16, v8
 ; RV64V-NEXT:    vsll.vi v8, v16, 3
 ; RV64V-NEXT:    vluxei64.v v12, (a0), v8, v0.t
 ; RV64V-NEXT:    vmv.v.v v8, v12
 ; RV64V-NEXT:    ret
 ;
 ; RV32ZVE32F-LABEL: mgather_baseidx_v8i32_v8i64:
 ; RV32ZVE32F:       # %bb.0:
-; RV32ZVE32F-NEXT:    addi sp, sp, -16
-; RV32ZVE32F-NEXT:    .cfi_def_cfa_offset 16
-; RV32ZVE32F-NEXT:    sw s0, 12(sp) # 4-byte Folded Spill
-; RV32ZVE32F-NEXT:    sw s1, 8(sp) # 4-byte Folded Spill
-; RV32ZVE32F-NEXT:    .cfi_offset s0, -4
-; RV32ZVE32F-NEXT:    .cfi_offset s1, -8
 ; RV32ZVE32F-NEXT:    vsetivli zero, 8, e32, m2, ta, ma
 ; RV32ZVE32F-NEXT:    vsll.vi v8, v8, 3
 ; RV32ZVE32F-NEXT:    vadd.vx v8, v8, a1
 ; RV32ZVE32F-NEXT:    vsetivli zero, 0, e8, mf4, ta, ma
 ; RV32ZVE32F-NEXT:    vmv.x.s t0, v0
 ; RV32ZVE32F-NEXT:    andi a1, t0, 1
-; RV32ZVE32F-NEXT:    beqz a1, .LBB54_9
+; RV32ZVE32F-NEXT:    beqz a1, .LBB54_7
 ; RV32ZVE32F-NEXT:  # %bb.1: # %cond.load
 ; RV32ZVE32F-NEXT:    vsetivli zero, 0, e32, m2, ta, ma
 ; RV32ZVE32F-NEXT:    vmv.x.s a3, v8
 ; RV32ZVE32F-NEXT:    lw a1, 4(a3)
 ; RV32ZVE32F-NEXT:    lw a3, 0(a3)
 ; RV32ZVE32F-NEXT:    andi a4, t0, 2
-; RV32ZVE32F-NEXT:    bnez a4, .LBB54_10
+; RV32ZVE32F-NEXT:    bnez a4, .LBB54_8
 ; RV32ZVE32F-NEXT:  .LBB54_2:
 ; RV32ZVE32F-NEXT:    lw a4, 12(a2)
 ; RV32ZVE32F-NEXT:    lw a5, 8(a2)
 ; RV32ZVE32F-NEXT:    andi a6, t0, 4
-; RV32ZVE32F-NEXT:    bnez a6, .LBB54_11
+; RV32ZVE32F-NEXT:    bnez a6, .LBB54_9
 ; RV32ZVE32F-NEXT:  .LBB54_3:
 ; RV32ZVE32F-NEXT:    lw a6, 20(a2)
 ; RV32ZVE32F-NEXT:    lw a7, 16(a2)
 ; RV32ZVE32F-NEXT:    andi t1, t0, 8
-; RV32ZVE32F-NEXT:    bnez t1, .LBB54_12
+; RV32ZVE32F-NEXT:    bnez t1, .LBB54_10
 ; RV32ZVE32F-NEXT:  .LBB54_4:
 ; RV32ZVE32F-NEXT:    lw t1, 28(a2)
 ; RV32ZVE32F-NEXT:    lw t2, 24(a2)
 ; RV32ZVE32F-NEXT:    andi t3, t0, 16
-; RV32ZVE32F-NEXT:    bnez t3, .LBB54_13
+; RV32ZVE32F-NEXT:    bnez t3, .LBB54_11
 ; RV32ZVE32F-NEXT:  .LBB54_5:
 ; RV32ZVE32F-NEXT:    lw t3, 36(a2)
 ; RV32ZVE32F-NEXT:    lw t4, 32(a2)
 ; RV32ZVE32F-NEXT:    andi t5, t0, 32
-; RV32ZVE32F-NEXT:    bnez t5, .LBB54_14
+; RV32ZVE32F-NEXT:    bnez t5, .LBB54_12
 ; RV32ZVE32F-NEXT:  .LBB54_6:
 ; RV32ZVE32F-NEXT:    lw t5, 44(a2)
 ; RV32ZVE32F-NEXT:    lw t6, 40(a2)
-; RV32ZVE32F-NEXT:    andi s0, t0, 64
-; RV32ZVE32F-NEXT:    bnez s0, .LBB54_15
+; RV32ZVE32F-NEXT:    j .LBB54_13
 ; RV32ZVE32F-NEXT:  .LBB54_7:
-; RV32ZVE32F-NEXT:    lw s0, 52(a2)
-; RV32ZVE32F-NEXT:    lw s1, 48(a2)
-; RV32ZVE32F-NEXT:    andi t0, t0, -128
-; RV32ZVE32F-NEXT:    bnez t0, .LBB54_16
-; RV32ZVE32F-NEXT:  .LBB54_8:
-; RV32ZVE32F-NEXT:    lw t0, 60(a2)
-; RV32ZVE32F-NEXT:    lw a2, 56(a2)
-; RV32ZVE32F-NEXT:    j .LBB54_17
-; RV32ZVE32F-NEXT:  .LBB54_9:
 ; RV32ZVE32F-NEXT:    lw a1, 4(a2)
 ; RV32ZVE32F-NEXT:    lw a3, 0(a2)
 ; RV32ZVE32F-NEXT:    andi a4, t0, 2
 ; RV32ZVE32F-NEXT:    beqz a4, .LBB54_2
-; RV32ZVE32F-NEXT:  .LBB54_10: # %cond.load1
+; RV32ZVE32F-NEXT:  .LBB54_8: # %cond.load1
 ; RV32ZVE32F-NEXT:    vsetivli zero, 1, e32, m2, ta, ma
 ; RV32ZVE32F-NEXT:    vslidedown.vi v10, v8, 1
 ; RV32ZVE32F-NEXT:    vmv.x.s a5, v10
 ; RV32ZVE32F-NEXT:    lw a4, 4(a5)
 ; RV32ZVE32F-NEXT:    lw a5, 0(a5)
 ; RV32ZVE32F-NEXT:    andi a6, t0, 4
 ; RV32ZVE32F-NEXT:    beqz a6, .LBB54_3
-; RV32ZVE32F-NEXT:  .LBB54_11: # %cond.load4
+; RV32ZVE32F-NEXT:  .LBB54_9: # %cond.load4
 ; RV32ZVE32F-NEXT:    vsetivli zero, 1, e32, m2, ta, ma
 ; RV32ZVE32F-NEXT:    vslidedown.vi v10, v8, 2
 ; RV32ZVE32F-NEXT:    vmv.x.s a7, v10
 ; RV32ZVE32F-NEXT:    lw a6, 4(a7)
 ; RV32ZVE32F-NEXT:    lw a7, 0(a7)
 ; RV32ZVE32F-NEXT:    andi t1, t0, 8
 ; RV32ZVE32F-NEXT:    beqz t1, .LBB54_4
-; RV32ZVE32F-NEXT:  .LBB54_12: # %cond.load7
+; RV32ZVE32F-NEXT:  .LBB54_10: # %cond.load7
 ; RV32ZVE32F-NEXT:    vsetivli zero, 1, e32, m2, ta, ma
 ; RV32ZVE32F-NEXT:    vslidedown.vi v10, v8, 3
 ; RV32ZVE32F-NEXT:    vmv.x.s t2, v10
 ; RV32ZVE32F-NEXT:    lw t1, 4(t2)
 ; RV32ZVE32F-NEXT:    lw t2, 0(t2)
 ; RV32ZVE32F-NEXT:    andi t3, t0, 16
 ; RV32ZVE32F-NEXT:    beqz t3, .LBB54_5
-; RV32ZVE32F-NEXT:  .LBB54_13: # %cond.load10
+; RV32ZVE32F-NEXT:  .LBB54_11: # %cond.load10
 ; RV32ZVE32F-NEXT:    vsetivli zero, 1, e32, m2, ta, ma
 ; RV32ZVE32F-NEXT:    vslidedown.vi v10, v8, 4
 ; RV32ZVE32F-NEXT:    vmv.x.s t4, v10
 ; RV32ZVE32F-NEXT:    lw t3, 4(t4)
 ; RV32ZVE32F-NEXT:    lw t4, 0(t4)
 ; RV32ZVE32F-NEXT:    andi t5, t0, 32
 ; RV32ZVE32F-NEXT:    beqz t5, .LBB54_6
-; RV32ZVE32F-NEXT:  .LBB54_14: # %cond.load13
+; RV32ZVE32F-NEXT:  .LBB54_12: # %cond.load13
 ; RV32ZVE32F-NEXT:    vsetivli zero, 1, e32, m2, ta, ma
 ; RV32ZVE32F-NEXT:    vslidedown.vi v10, v8, 5
 ; RV32ZVE32F-NEXT:    vmv.x.s t6, v10
 ; RV32ZVE32F-NEXT:    lw t5, 4(t6)
 ; RV32ZVE32F-NEXT:    lw t6, 0(t6)
+; RV32ZVE32F-NEXT:  .LBB54_13: # %else14
+; RV32ZVE32F-NEXT:    addi sp, sp, -16
+; RV32ZVE32F-NEXT:    .cfi_def_cfa_offset 16
+; RV32ZVE32F-NEXT:    sw s0, 12(sp) # 4-byte Folded Spill
+; RV32ZVE32F-NEXT:    sw s1, 8(sp) # 4-byte Folded Spill
+; RV32ZVE32F-NEXT:    .cfi_offset s0, -4
+; RV32ZVE32F-NEXT:    .cfi_offset s1, -8
 ; RV32ZVE32F-NEXT:    andi s0, t0, 64
-; RV32ZVE32F-NEXT:    beqz s0, .LBB54_7
-; RV32ZVE32F-NEXT:  .LBB54_15: # %cond.load16
+; RV32ZVE32F-NEXT:    beqz s0, .LBB54_16
+; RV32ZVE32F-NEXT:  # %bb.14: # %cond.load16
 ; RV32ZVE32F-NEXT:    vsetivli zero, 1, e32, m2, ta, ma
 ; RV32ZVE32F-NEXT:    vslidedown.vi v10, v8, 6
 ; RV32ZVE32F-NEXT:    vmv.x.s s1, v10
 ; RV32ZVE32F-NEXT:    lw s0, 4(s1)
 ; RV32ZVE32F-NEXT:    lw s1, 0(s1)
 ; RV32ZVE32F-NEXT:    andi t0, t0, -128
-; RV32ZVE32F-NEXT:    beqz t0, .LBB54_8
-; RV32ZVE32F-NEXT:  .LBB54_16: # %cond.load19
+; RV32ZVE32F-NEXT:    bnez t0, .LBB54_17
+; RV32ZVE32F-NEXT:  .LBB54_15:
+; RV32ZVE32F-NEXT:    lw t0, 60(a2)
+; RV32ZVE32F-NEXT:    lw a2, 56(a2)
+; RV32ZVE32F-NEXT:    j .LBB54_18
+; RV32ZVE32F-NEXT:  .LBB54_16:
+; RV32ZVE32F-NEXT:    lw s0, 52(a2)
+; RV32ZVE32F-NEXT:    lw s1, 48(a2)
+; RV32ZVE32F-NEXT:    andi t0, t0, -128
+; RV32ZVE32F-NEXT:    beqz t0, .LBB54_15
+; RV32ZVE32F-NEXT:  .LBB54_17: # %cond.load19
 ; RV32ZVE32F-NEXT:    vsetivli zero, 1, e32, m2, ta, ma
 ; RV32ZVE32F-NEXT:    vslidedown.vi v8, v8, 7
 ; RV32ZVE32F-NEXT:    vmv.x.s a2, v8
 ; RV32ZVE32F-NEXT:    lw t0, 4(a2)
 ; RV32ZVE32F-NEXT:    lw a2, 0(a2)
-; RV32ZVE32F-NEXT:  .LBB54_17: # %else20
+; RV32ZVE32F-NEXT:  .LBB54_18: # %else20
 ; RV32ZVE32F-NEXT:    sw a3, 0(a0)
 ; RV32ZVE32F-NEXT:    sw a1, 4(a0)
 ; RV32ZVE32F-NEXT:    sw a5, 8(a0)
 ; RV32ZVE32F-NEXT:    sw a4, 12(a0)
 ; RV32ZVE32F-NEXT:    sw a7, 16(a0)
 ; RV32ZVE32F-NEXT:    sw a6, 20(a0)
 ; RV32ZVE32F-NEXT:    sw t2, 24(a0)
 ; RV32ZVE32F-NEXT:    sw t1, 28(a0)

 ; RV64V-NEXT:    vsext.vf2 v16, v8
 ; RV64V-NEXT:    vsll.vi v8, v16, 3
 ; RV64V-NEXT:    vluxei64.v v12, (a0), v8, v0.t
 ; RV64V-NEXT:    vmv.v.v v8, v12
 ; RV64V-NEXT:    ret
 ;
 ; RV32ZVE32F-LABEL: mgather_baseidx_sext_v8i32_v8i64:
 ; RV32ZVE32F:       # %bb.0:
-; RV32ZVE32F-NEXT:    addi sp, sp, -16
-; RV32ZVE32F-NEXT:    .cfi_def_cfa_offset 16
-; RV32ZVE32F-NEXT:    sw s0, 12(sp) # 4-byte Folded Spill
-; RV32ZVE32F-NEXT:    sw s1, 8(sp) # 4-byte Folded Spill
-; RV32ZVE32F-NEXT:    .cfi_offset s0, -4
-; RV32ZVE32F-NEXT:    .cfi_offset s1, -8
 ; RV32ZVE32F-NEXT:    vsetivli zero, 8, e32, m2, ta, ma
 ; RV32ZVE32F-NEXT:    vsll.vi v8, v8, 3
 ; RV32ZVE32F-NEXT:    vadd.vx v8, v8, a1
 ; RV32ZVE32F-NEXT:    vsetivli zero, 0, e8, mf4, ta, ma
 ; RV32ZVE32F-NEXT:    vmv.x.s t0, v0
 ; RV32ZVE32F-NEXT:    andi a1, t0, 1
-; RV32ZVE32F-NEXT:    beqz a1, .LBB55_9
+; RV32ZVE32F-NEXT:    beqz a1, .LBB55_7
 ; RV32ZVE32F-NEXT:  # %bb.1: # %cond.load
 ; RV32ZVE32F-NEXT:    vsetivli zero, 0, e32, m2, ta, ma
 ; RV32ZVE32F-NEXT:    vmv.x.s a3, v8
 ; RV32ZVE32F-NEXT:    lw a1, 4(a3)
 ; RV32ZVE32F-NEXT:    lw a3, 0(a3)
 ; RV32ZVE32F-NEXT:    andi a4, t0, 2
-; RV32ZVE32F-NEXT:    bnez a4, .LBB55_10
+; RV32ZVE32F-NEXT:    bnez a4, .LBB55_8
 ; RV32ZVE32F-NEXT:  .LBB55_2:
 ; RV32ZVE32F-NEXT:    lw a4, 12(a2)
 ; RV32ZVE32F-NEXT:    lw a5, 8(a2)
 ; RV32ZVE32F-NEXT:    andi a6, t0, 4
-; RV32ZVE32F-NEXT:    bnez a6, .LBB55_11
+; RV32ZVE32F-NEXT:    bnez a6, .LBB55_9
 ; RV32ZVE32F-NEXT:  .LBB55_3:
 ; RV32ZVE32F-NEXT:    lw a6, 20(a2)
 ; RV32ZVE32F-NEXT:    lw a7, 16(a2)
 ; RV32ZVE32F-NEXT:    andi t1, t0, 8
-; RV32ZVE32F-NEXT:    bnez t1, .LBB55_12
+; RV32ZVE32F-NEXT:    bnez t1, .LBB55_10
 ; RV32ZVE32F-NEXT:  .LBB55_4:
 ; RV32ZVE32F-NEXT:    lw t1, 28(a2)
 ; RV32ZVE32F-NEXT:    lw t2, 24(a2)
 ; RV32ZVE32F-NEXT:    andi t3, t0, 16
-; RV32ZVE32F-NEXT:    bnez t3, .LBB55_13
+; RV32ZVE32F-NEXT:    bnez t3, .LBB55_11
 ; RV32ZVE32F-NEXT:  .LBB55_5:
 ; RV32ZVE32F-NEXT:    lw t3, 36(a2)
 ; RV32ZVE32F-NEXT:    lw t4, 32(a2)
 ; RV32ZVE32F-NEXT:    andi t5, t0, 32
-; RV32ZVE32F-NEXT:    bnez t5, .LBB55_14
+; RV32ZVE32F-NEXT:    bnez t5, .LBB55_12
 ; RV32ZVE32F-NEXT:  .LBB55_6:
 ; RV32ZVE32F-NEXT:    lw t5, 44(a2)
 ; RV32ZVE32F-NEXT:    lw t6, 40(a2)
-; RV32ZVE32F-NEXT:    andi s0, t0, 64
-; RV32ZVE32F-NEXT:    bnez s0, .LBB55_15
+; RV32ZVE32F-NEXT:    j .LBB55_13
 ; RV32ZVE32F-NEXT:  .LBB55_7:
-; RV32ZVE32F-NEXT:    lw s0, 52(a2)
-; RV32ZVE32F-NEXT:    lw s1, 48(a2)
-; RV32ZVE32F-NEXT:    andi t0, t0, -128
-; RV32ZVE32F-NEXT:    bnez t0, .LBB55_16
-; RV32ZVE32F-NEXT:  .LBB55_8:
-; RV32ZVE32F-NEXT:    lw t0, 60(a2)
-; RV32ZVE32F-NEXT:    lw a2, 56(a2)
-; RV32ZVE32F-NEXT:    j .LBB55_17
-; RV32ZVE32F-NEXT:  .LBB55_9:
 ; RV32ZVE32F-NEXT:    lw a1, 4(a2)
 ; RV32ZVE32F-NEXT:    lw a3, 0(a2)
 ; RV32ZVE32F-NEXT:    andi a4, t0, 2
 ; RV32ZVE32F-NEXT:    beqz a4, .LBB55_2
-; RV32ZVE32F-NEXT:  .LBB55_10: # %cond.load1
+; RV32ZVE32F-NEXT:  .LBB55_8: # %cond.load1
 ; RV32ZVE32F-NEXT:    vsetivli zero, 1, e32, m2, ta, ma
 ; RV32ZVE32F-NEXT:    vslidedown.vi v10, v8, 1
 ; RV32ZVE32F-NEXT:    vmv.x.s a5, v10
 ; RV32ZVE32F-NEXT:    lw a4, 4(a5)
 ; RV32ZVE32F-NEXT:    lw a5, 0(a5)
 ; RV32ZVE32F-NEXT:    andi a6, t0, 4
 ; RV32ZVE32F-NEXT:    beqz a6, .LBB55_3
-; RV32ZVE32F-NEXT:  .LBB55_11: # %cond.load4
+; RV32ZVE32F-NEXT:  .LBB55_9: # %cond.load4
 ; RV32ZVE32F-NEXT:    vsetivli zero, 1, e32, m2, ta, ma
 ; RV32ZVE32F-NEXT:    vslidedown.vi v10, v8, 2
 ; RV32ZVE32F-NEXT:    vmv.x.s a7, v10
 ; RV32ZVE32F-NEXT:    lw a6, 4(a7)
 ; RV32ZVE32F-NEXT:    lw a7, 0(a7)
 ; RV32ZVE32F-NEXT:    andi t1, t0, 8
 ; RV32ZVE32F-NEXT:    beqz t1, .LBB55_4
-; RV32ZVE32F-NEXT:  .LBB55_12: # %cond.load7
+; RV32ZVE32F-NEXT:  .LBB55_10: # %cond.load7
 ; RV32ZVE32F-NEXT:    vsetivli zero, 1, e32, m2, ta, ma
 ; RV32ZVE32F-NEXT:    vslidedown.vi v10, v8, 3
 ; RV32ZVE32F-NEXT:    vmv.x.s t2, v10
 ; RV32ZVE32F-NEXT:    lw t1, 4(t2)
 ; RV32ZVE32F-NEXT:    lw t2, 0(t2)
 ; RV32ZVE32F-NEXT:    andi t3, t0, 16
 ; RV32ZVE32F-NEXT:    beqz t3, .LBB55_5
-; RV32ZVE32F-NEXT:  .LBB55_13: # %cond.load10
+; RV32ZVE32F-NEXT:  .LBB55_11: # %cond.load10
 ; RV32ZVE32F-NEXT:    vsetivli zero, 1, e32, m2, ta, ma
 ; RV32ZVE32F-NEXT:    vslidedown.vi v10, v8, 4
 ; RV32ZVE32F-NEXT:    vmv.x.s t4, v10
 ; RV32ZVE32F-NEXT:    lw t3, 4(t4)
 ; RV32ZVE32F-NEXT:    lw t4, 0(t4)
 ; RV32ZVE32F-NEXT:    andi t5, t0, 32
 ; RV32ZVE32F-NEXT:    beqz t5, .LBB55_6
-; RV32ZVE32F-NEXT:  .LBB55_14: # %cond.load13
+; RV32ZVE32F-NEXT:  .LBB55_12: # %cond.load13
 ; RV32ZVE32F-NEXT:    vsetivli zero, 1, e32, m2, ta, ma
 ; RV32ZVE32F-NEXT:    vslidedown.vi v10, v8, 5
 ; RV32ZVE32F-NEXT:    vmv.x.s t6, v10
 ; RV32ZVE32F-NEXT:    lw t5, 4(t6)
 ; RV32ZVE32F-NEXT:    lw t6, 0(t6)
+; RV32ZVE32F-NEXT:  .LBB55_13: # %else14
+; RV32ZVE32F-NEXT:    addi sp, sp, -16
+; RV32ZVE32F-NEXT:    .cfi_def_cfa_offset 16
+; RV32ZVE32F-NEXT:    sw s0, 12(sp) # 4-byte Folded Spill
+; RV32ZVE32F-NEXT:    sw s1, 8(sp) # 4-byte Folded Spill
+; RV32ZVE32F-NEXT:    .cfi_offset s0, -4
+; RV32ZVE32F-NEXT:    .cfi_offset s1, -8
 ; RV32ZVE32F-NEXT:    andi s0, t0, 64
-; RV32ZVE32F-NEXT:    beqz s0, .LBB55_7
-; RV32ZVE32F-NEXT:  .LBB55_15: # %cond.load16
+; RV32ZVE32F-NEXT:    beqz s0, .LBB55_16
+; RV32ZVE32F-NEXT:  # %bb.14: # %cond.load16
 ; RV32ZVE32F-NEXT:    vsetivli zero, 1, e32, m2, ta, ma
 ; RV32ZVE32F-NEXT:    vslidedown.vi v10, v8, 6
 ; RV32ZVE32F-NEXT:    vmv.x.s s1, v10
 ; RV32ZVE32F-NEXT:    lw s0, 4(s1)
 ; RV32ZVE32F-NEXT:    lw s1, 0(s1)
 ; RV32ZVE32F-NEXT:    andi t0, t0, -128
-; RV32ZVE32F-NEXT:    beqz t0, .LBB55_8
-; RV32ZVE32F-NEXT:  .LBB55_16: # %cond.load19
+; RV32ZVE32F-NEXT:    bnez t0, .LBB55_17
+; RV32ZVE32F-NEXT:  .LBB55_15:
+; RV32ZVE32F-NEXT:    lw t0, 60(a2)
+; RV32ZVE32F-NEXT:    lw a2, 56(a2)
+; RV32ZVE32F-NEXT:    j .LBB55_18
+; RV32ZVE32F-NEXT:  .LBB55_16:
+; RV32ZVE32F-NEXT:    lw s0, 52(a2)
+; RV32ZVE32F-NEXT:    lw s1, 48(a2)
+; RV32ZVE32F-NEXT:    andi t0, t0, -128
+; RV32ZVE32F-NEXT:    beqz t0, .LBB55_15
+; RV32ZVE32F-NEXT:  .LBB55_17: # %cond.load19
 ; RV32ZVE32F-NEXT:    vsetivli zero, 1, e32, m2, ta, ma
 ; RV32ZVE32F-NEXT:    vslidedown.vi v8, v8, 7
 ; RV32ZVE32F-NEXT:    vmv.x.s a2, v8
 ; RV32ZVE32F-NEXT:    lw t0, 4(a2)
 ; RV32ZVE32F-NEXT:    lw a2, 0(a2)
-; RV32ZVE32F-NEXT:  .LBB55_17: # %else20
+; RV32ZVE32F-NEXT:  .LBB55_18: # %else20
 ; RV32ZVE32F-NEXT:    sw a3, 0(a0)
 ; RV32ZVE32F-NEXT:    sw a1, 4(a0)
 ; RV32ZVE32F-NEXT:    sw a5, 8(a0)
 ; RV32ZVE32F-NEXT:    sw a4, 12(a0)
 ; RV32ZVE32F-NEXT:    sw a7, 16(a0)
 ; RV32ZVE32F-NEXT:    sw a6, 20(a0)
 ; RV32ZVE32F-NEXT:    sw t2, 24(a0)
 ; RV32ZVE32F-NEXT:    sw t1, 28(a0)

 ; RV64V-NEXT:    vzext.vf2 v16, v8
 ; RV64V-NEXT:    vsll.vi v8, v16, 3
 ; RV64V-NEXT:    vluxei64.v v12, (a0), v8, v0.t
 ; RV64V-NEXT:    vmv.v.v v8, v12
 ; RV64V-NEXT:    ret
 ;
 ; RV32ZVE32F-LABEL: mgather_baseidx_zext_v8i32_v8i64:
 ; RV32ZVE32F:       # %bb.0:
-; RV32ZVE32F-NEXT:    addi sp, sp, -16
-; RV32ZVE32F-NEXT:    .cfi_def_cfa_offset 16
-; RV32ZVE32F-NEXT:    sw s0, 12(sp) # 4-byte Folded Spill
-; RV32ZVE32F-NEXT:    sw s1, 8(sp) # 4-byte Folded Spill
-; RV32ZVE32F-NEXT:    .cfi_offset s0, -4
-; RV32ZVE32F-NEXT:    .cfi_offset s1, -8
 ; RV32ZVE32F-NEXT:    vsetivli zero, 8, e32, m2, ta, ma
 ; RV32ZVE32F-NEXT:    vsll.vi v8, v8, 3
 ; RV32ZVE32F-NEXT:    vadd.vx v8, v8, a1
 ; RV32ZVE32F-NEXT:    vsetivli zero, 0, e8, mf4, ta, ma
 ; RV32ZVE32F-NEXT:    vmv.x.s t0, v0
 ; RV32ZVE32F-NEXT:    andi a1, t0, 1
-; RV32ZVE32F-NEXT:    beqz a1, .LBB56_9
+; RV32ZVE32F-NEXT:    beqz a1, .LBB56_7
 ; RV32ZVE32F-NEXT:  # %bb.1: # %cond.load
 ; RV32ZVE32F-NEXT:    vsetivli zero, 0, e32, m2, ta, ma
 ; RV32ZVE32F-NEXT:    vmv.x.s a3, v8
 ; RV32ZVE32F-NEXT:    lw a1, 4(a3)
 ; RV32ZVE32F-NEXT:    lw a3, 0(a3)
 ; RV32ZVE32F-NEXT:    andi a4, t0, 2
-; RV32ZVE32F-NEXT:    bnez a4, .LBB56_10
+; RV32ZVE32F-NEXT:    bnez a4, .LBB56_8
 ; RV32ZVE32F-NEXT:  .LBB56_2:
 ; RV32ZVE32F-NEXT:    lw a4, 12(a2)
 ; RV32ZVE32F-NEXT:    lw a5, 8(a2)
 ; RV32ZVE32F-NEXT:    andi a6, t0, 4
-; RV32ZVE32F-NEXT:    bnez a6, .LBB56_11
+; RV32ZVE32F-NEXT:    bnez a6, .LBB56_9
 ; RV32ZVE32F-NEXT:  .LBB56_3:
 ; RV32ZVE32F-NEXT:    lw a6, 20(a2)
 ; RV32ZVE32F-NEXT:    lw a7, 16(a2)
 ; RV32ZVE32F-NEXT:    andi t1, t0, 8
-; RV32ZVE32F-NEXT:    bnez t1, .LBB56_12
+; RV32ZVE32F-NEXT:    bnez t1, .LBB56_10
 ; RV32ZVE32F-NEXT:  .LBB56_4:
 ; RV32ZVE32F-NEXT:    lw t1, 28(a2)
 ; RV32ZVE32F-NEXT:    lw t2, 24(a2)
 ; RV32ZVE32F-NEXT:    andi t3, t0, 16
-; RV32ZVE32F-NEXT:    bnez t3, .LBB56_13
+; RV32ZVE32F-NEXT:    bnez t3, .LBB56_11
 ; RV32ZVE32F-NEXT:  .LBB56_5:
 ; RV32ZVE32F-NEXT:    lw t3, 36(a2)
 ; RV32ZVE32F-NEXT:    lw t4, 32(a2)
 ; RV32ZVE32F-NEXT:    andi t5, t0, 32
-; RV32ZVE32F-NEXT:    bnez t5, .LBB56_14
+; RV32ZVE32F-NEXT:    bnez t5, .LBB56_12
 ; RV32ZVE32F-NEXT:  .LBB56_6:
 ; RV32ZVE32F-NEXT:    lw t5, 44(a2)
 ; RV32ZVE32F-NEXT:    lw t6, 40(a2)
-; RV32ZVE32F-NEXT:    andi s0, t0, 64
-; RV32ZVE32F-NEXT:    bnez s0, .LBB56_15
+; RV32ZVE32F-NEXT:    j .LBB56_13
 ; RV32ZVE32F-NEXT:  .LBB56_7:
-; RV32ZVE32F-NEXT:    lw s0, 52(a2)
-; RV32ZVE32F-NEXT:    lw s1, 48(a2)
-; RV32ZVE32F-NEXT:    andi t0, t0, -128
-; RV32ZVE32F-NEXT:    bnez t0, .LBB56_16
-; RV32ZVE32F-NEXT:  .LBB56_8:
-; RV32ZVE32F-NEXT:    lw t0, 60(a2)
-; RV32ZVE32F-NEXT:    lw a2, 56(a2)
-; RV32ZVE32F-NEXT:    j .LBB56_17
-; RV32ZVE32F-NEXT:  .LBB56_9:
 ; RV32ZVE32F-NEXT:    lw a1, 4(a2)
 ; RV32ZVE32F-NEXT:    lw a3, 0(a2)
 ; RV32ZVE32F-NEXT:    andi a4, t0, 2
 ; RV32ZVE32F-NEXT:    beqz a4, .LBB56_2
-; RV32ZVE32F-NEXT:  .LBB56_10: # %cond.load1
+; RV32ZVE32F-NEXT:  .LBB56_8: # %cond.load1
 ; RV32ZVE32F-NEXT:    vsetivli zero, 1, e32, m2, ta, ma
 ; RV32ZVE32F-NEXT:    vslidedown.vi v10, v8, 1
 ; RV32ZVE32F-NEXT:    vmv.x.s a5, v10
 ; RV32ZVE32F-NEXT:    lw a4, 4(a5)
 ; RV32ZVE32F-NEXT:    lw a5, 0(a5)
 ; RV32ZVE32F-NEXT:    andi a6, t0, 4
 ; RV32ZVE32F-NEXT:    beqz a6, .LBB56_3
-; RV32ZVE32F-NEXT:  .LBB56_11: # %cond.load4
+; RV32ZVE32F-NEXT:  .LBB56_9: # %cond.load4
 ; RV32ZVE32F-NEXT:    vsetivli zero, 1, e32, m2, ta, ma
 ; RV32ZVE32F-NEXT:    vslidedown.vi v10, v8, 2
 ; RV32ZVE32F-NEXT:    vmv.x.s a7, v10
 ; RV32ZVE32F-NEXT:    lw a6, 4(a7)
 ; RV32ZVE32F-NEXT:    lw a7, 0(a7)
 ; RV32ZVE32F-NEXT:    andi t1, t0, 8
 ; RV32ZVE32F-NEXT:    beqz t1, .LBB56_4
-; RV32ZVE32F-NEXT:  .LBB56_12: # %cond.load7
+; RV32ZVE32F-NEXT:  .LBB56_10: # %cond.load7
 ; RV32ZVE32F-NEXT:    vsetivli zero, 1, e32, m2, ta, ma
 ; RV32ZVE32F-NEXT:    vslidedown.vi v10, v8, 3
 ; RV32ZVE32F-NEXT:    vmv.x.s t2, v10
 ; RV32ZVE32F-NEXT:    lw t1, 4(t2)
 ; RV32ZVE32F-NEXT:    lw t2, 0(t2)
 ; RV32ZVE32F-NEXT:    andi t3, t0, 16
 ; RV32ZVE32F-NEXT:    beqz t3, .LBB56_5
-; RV32ZVE32F-NEXT:  .LBB56_13: # %cond.load10
+; RV32ZVE32F-NEXT:  .LBB56_11: # %cond.load10
 ; RV32ZVE32F-NEXT:    vsetivli zero, 1, e32, m2, ta, ma
 ; RV32ZVE32F-NEXT:    vslidedown.vi v10, v8, 4
 ; RV32ZVE32F-NEXT:    vmv.x.s t4, v10
 ; RV32ZVE32F-NEXT:    lw t3, 4(t4)
 ; RV32ZVE32F-NEXT:    lw t4, 0(t4)
 ; RV32ZVE32F-NEXT:    andi t5, t0, 32
 ; RV32ZVE32F-NEXT:    beqz t5, .LBB56_6
-; RV32ZVE32F-NEXT:  .LBB56_14: # %cond.load13
+; RV32ZVE32F-NEXT:  .LBB56_12: # %cond.load13
 ; RV32ZVE32F-NEXT:    vsetivli zero, 1, e32, m2, ta, ma
 ; RV32ZVE32F-NEXT:    vslidedown.vi v10, v8, 5
 ; RV32ZVE32F-NEXT:    vmv.x.s t6, v10
 ; RV32ZVE32F-NEXT:    lw t5, 4(t6)
 ; RV32ZVE32F-NEXT:    lw t6, 0(t6)
+; RV32ZVE32F-NEXT:  .LBB56_13: # %else14
+; RV32ZVE32F-NEXT:    addi sp, sp, -16
+; RV32ZVE32F-NEXT:    .cfi_def_cfa_offset 16
+; RV32ZVE32F-NEXT:    sw s0, 12(sp) # 4-byte Folded Spill
+; RV32ZVE32F-NEXT:    sw s1, 8(sp) # 4-byte Folded Spill
+; RV32ZVE32F-NEXT:    .cfi_offset s0, -4
+; RV32ZVE32F-NEXT:    .cfi_offset s1, -8
 ; RV32ZVE32F-NEXT:    andi s0, t0, 64
-; RV32ZVE32F-NEXT:    beqz s0, .LBB56_7
-; RV32ZVE32F-NEXT:  .LBB56_15: # %cond.load16
+; RV32ZVE32F-NEXT:    beqz s0, .LBB56_16
+; RV32ZVE32F-NEXT:  # %bb.14: # %cond.load16
 ; RV32ZVE32F-NEXT:    vsetivli zero, 1, e32, m2, ta, ma
 ; RV32ZVE32F-NEXT:    vslidedown.vi v10, v8, 6
 ; RV32ZVE32F-NEXT:    vmv.x.s s1, v10
 ; RV32ZVE32F-NEXT:    lw s0, 4(s1)
 ; RV32ZVE32F-NEXT:    lw s1, 0(s1)
 ; RV32ZVE32F-NEXT:    andi t0, t0, -128
-; RV32ZVE32F-NEXT:    beqz t0, .LBB56_8
-; RV32ZVE32F-NEXT:  .LBB56_16: # %cond.load19
+; RV32ZVE32F-NEXT:    bnez t0, .LBB56_17
+; RV32ZVE32F-NEXT:  .LBB56_15:
+; RV32ZVE32F-NEXT:    lw t0, 60(a2)
+; RV32ZVE32F-NEXT:    lw a2, 56(a2)
+; RV32ZVE32F-NEXT:    j .LBB56_18
+; RV32ZVE32F-NEXT:  .LBB56_16:
+; RV32ZVE32F-NEXT:    lw s0, 52(a2)
+; RV32ZVE32F-NEXT:    lw s1, 48(a2)
+; RV32ZVE32F-NEXT:    andi t0, t0, -128
+; RV32ZVE32F-NEXT:    beqz t0, .LBB56_15
+; RV32ZVE32F-NEXT:  .LBB56_17: # %cond.load19
 ; RV32ZVE32F-NEXT:    vsetivli zero, 1, e32, m2, ta, ma
 ; RV32ZVE32F-NEXT:    vslidedown.vi v8, v8, 7
 ; RV32ZVE32F-NEXT:    vmv.x.s a2, v8
 ; RV32ZVE32F-NEXT:    lw t0, 4(a2)
 ; RV32ZVE32F-NEXT:    lw a2, 0(a2)
-; RV32ZVE32F-NEXT:  .LBB56_17: # %else20
+; RV32ZVE32F-NEXT:  .LBB56_18: # %else20
 ; RV32ZVE32F-NEXT:    sw a3, 0(a0)
 ; RV32ZVE32F-NEXT:    sw a1, 4(a0)
 ; RV32ZVE32F-NEXT:    sw a5, 8(a0)
 ; RV32ZVE32F-NEXT:    sw a4, 12(a0)
 ; RV32ZVE32F-NEXT:    sw a7, 16(a0)
 ; RV32ZVE32F-NEXT:    sw a6, 20(a0)
 ; RV32ZVE32F-NEXT:    sw t2, 24(a0)
 ; RV32ZVE32F-NEXT:    sw t1, 28(a0)

 ; RV64V-NEXT:    vsetivli zero, 8, e64, m4, ta, mu
 ; RV64V-NEXT:    vsll.vi v8, v8, 3
 ; RV64V-NEXT:    vluxei64.v v12, (a0), v8, v0.t
 ; RV64V-NEXT:    vmv.v.v v8, v12
 ; RV64V-NEXT:    ret
 ;
 ; RV32ZVE32F-LABEL: mgather_baseidx_v8i64:
 ; RV32ZVE32F:       # %bb.0:
-; RV32ZVE32F-NEXT:    addi sp, sp, -16
-; RV32ZVE32F-NEXT:    .cfi_def_cfa_offset 16
-; RV32ZVE32F-NEXT:    sw s0, 12(sp) # 4-byte Folded Spill
-; RV32ZVE32F-NEXT:    sw s1, 8(sp) # 4-byte Folded Spill
-; RV32ZVE32F-NEXT:    .cfi_offset s0, -4
-; RV32ZVE32F-NEXT:    .cfi_offset s1, -8
 ; RV32ZVE32F-NEXT:    lw a4, 56(a2)
 ; RV32ZVE32F-NEXT:    lw a5, 48(a2)
 ; RV32ZVE32F-NEXT:    lw a6, 40(a2)
 ; RV32ZVE32F-NEXT:    lw a7, 32(a2)
 ; RV32ZVE32F-NEXT:    lw t0, 24(a2)
 ; RV32ZVE32F-NEXT:    lw t1, 0(a2)
 ; RV32ZVE32F-NEXT:    lw t2, 8(a2)
 ; RV32ZVE32F-NEXT:    lw a2, 16(a2)
 ; RV32ZVE32F-NEXT:    vsetivli zero, 8, e32, m2, ta, ma
 ; RV32ZVE32F-NEXT:    vslide1down.vx v8, v8, t1
 ; RV32ZVE32F-NEXT:    vslide1down.vx v8, v8, t2
 ; RV32ZVE32F-NEXT:    vslide1down.vx v8, v8, a2
 ; RV32ZVE32F-NEXT:    vslide1down.vx v8, v8, t0
 ; RV32ZVE32F-NEXT:    vslide1down.vx v8, v8, a7
 ; RV32ZVE32F-NEXT:    vslide1down.vx v8, v8, a6
 ; RV32ZVE32F-NEXT:    vslide1down.vx v8, v8, a5
 ; RV32ZVE32F-NEXT:    vslide1down.vx v8, v8, a4
 ; RV32ZVE32F-NEXT:    vsll.vi v8, v8, 3
 ; RV32ZVE32F-NEXT:    vadd.vx v8, v8, a1
 ; RV32ZVE32F-NEXT:    vsetivli zero, 0, e8, mf4, ta, ma
 ; RV32ZVE32F-NEXT:    vmv.x.s t0, v0
 ; RV32ZVE32F-NEXT:    andi a1, t0, 1
-; RV32ZVE32F-NEXT:    beqz a1, .LBB57_9
+; RV32ZVE32F-NEXT:    beqz a1, .LBB57_7
 ; RV32ZVE32F-NEXT:  # %bb.1: # %cond.load
 ; RV32ZVE32F-NEXT:    vsetivli zero, 0, e32, m2, ta, ma
 ; RV32ZVE32F-NEXT:    vmv.x.s a2, v8
 ; RV32ZVE32F-NEXT:    lw a1, 4(a2)
 ; RV32ZVE32F-NEXT:    lw a2, 0(a2)
 ; RV32ZVE32F-NEXT:    andi a4, t0, 2
-; RV32ZVE32F-NEXT:    bnez a4, .LBB57_10
+; RV32ZVE32F-NEXT:    bnez a4, .LBB57_8
 ; RV32ZVE32F-NEXT:  .LBB57_2:
 ; RV32ZVE32F-NEXT:    lw a4, 12(a3)
 ; RV32ZVE32F-NEXT:    lw a5, 8(a3)
 ; RV32ZVE32F-NEXT:    andi a6, t0, 4
-; RV32ZVE32F-NEXT:    bnez a6, .LBB57_11
+; RV32ZVE32F-NEXT:    bnez a6, .LBB57_9
 ; RV32ZVE32F-NEXT:  .LBB57_3:
 ; RV32ZVE32F-NEXT:    lw a6, 20(a3)
 ; RV32ZVE32F-NEXT:    lw a7, 16(a3)
 ; RV32ZVE32F-NEXT:    andi t1, t0, 8
-; RV32ZVE32F-NEXT:    bnez t1, .LBB57_12
+; RV32ZVE32F-NEXT:    bnez t1, .LBB57_10
 ; RV32ZVE32F-NEXT:  .LBB57_4:
 ; RV32ZVE32F-NEXT:    lw t1, 28(a3)
 ; RV32ZVE32F-NEXT:    lw t2, 24(a3)
 ; RV32ZVE32F-NEXT:    andi t3, t0, 16
-; RV32ZVE32F-NEXT:    bnez t3, .LBB57_13
+; RV32ZVE32F-NEXT:    bnez t3, .LBB57_11
 ; RV32ZVE32F-NEXT:  .LBB57_5:
 ; RV32ZVE32F-NEXT:    lw t3, 36(a3)
 ; RV32ZVE32F-NEXT:    lw t4, 32(a3)
 ; RV32ZVE32F-NEXT:    andi t5, t0, 32
-; RV32ZVE32F-NEXT:    bnez t5, .LBB57_14
+; RV32ZVE32F-NEXT:    bnez t5, .LBB57_12
 ; RV32ZVE32F-NEXT:  .LBB57_6:
 ; RV32ZVE32F-NEXT:    lw t5, 44(a3)
 ; RV32ZVE32F-NEXT:    lw t6, 40(a3)
-; RV32ZVE32F-NEXT:    andi s0, t0, 64
-; RV32ZVE32F-NEXT:    bnez s0, .LBB57_15
+; RV32ZVE32F-NEXT:    j .LBB57_13
 ; RV32ZVE32F-NEXT:  .LBB57_7:
-; RV32ZVE32F-NEXT:    lw s0, 52(a3)
-; RV32ZVE32F-NEXT:    lw s1, 48(a3)
-; RV32ZVE32F-NEXT:    andi t0, t0, -128
-; RV32ZVE32F-NEXT:    bnez t0, .LBB57_16
-; RV32ZVE32F-NEXT:  .LBB57_8:
-; RV32ZVE32F-NEXT:    lw t0, 60(a3)
-; RV32ZVE32F-NEXT:    lw a3, 56(a3)
-; RV32ZVE32F-NEXT:    j .LBB57_17
-; RV32ZVE32F-NEXT:  .LBB57_9:
 ; RV32ZVE32F-NEXT:    lw a1, 4(a3)
 ; RV32ZVE32F-NEXT:    lw a2, 0(a3)
 ; RV32ZVE32F-NEXT:    andi a4, t0, 2
 ; RV32ZVE32F-NEXT:    beqz a4, .LBB57_2
-; RV32ZVE32F-NEXT:  .LBB57_10: # %cond.load1
+; RV32ZVE32F-NEXT:  .LBB57_8: # %cond.load1
 ; RV32ZVE32F-NEXT:    vsetivli zero, 1, e32, m2, ta, ma
 ; RV32ZVE32F-NEXT:    vslidedown.vi v10, v8, 1
 ; RV32ZVE32F-NEXT:    vmv.x.s a5, v10
 ; RV32ZVE32F-NEXT:    lw a4, 4(a5)
 ; RV32ZVE32F-NEXT:    lw a5, 0(a5)
 ; RV32ZVE32F-NEXT:    andi a6, t0, 4
 ; RV32ZVE32F-NEXT:    beqz a6, .LBB57_3
-; RV32ZVE32F-NEXT:  .LBB57_11: # %cond.load4
+; RV32ZVE32F-NEXT:  .LBB57_9: # %cond.load4
 ; RV32ZVE32F-NEXT:    vsetivli zero, 1, e32, m2, ta, ma
 ; RV32ZVE32F-NEXT:    vslidedown.vi v10, v8, 2
 ; RV32ZVE32F-NEXT:    vmv.x.s a7, v10
 ; RV32ZVE32F-NEXT:    lw a6, 4(a7)
 ; RV32ZVE32F-NEXT:    lw a7, 0(a7)
 ; RV32ZVE32F-NEXT:    andi t1, t0, 8
 ; RV32ZVE32F-NEXT:    beqz t1, .LBB57_4
-; RV32ZVE32F-NEXT:  .LBB57_12: # %cond.load7
+; RV32ZVE32F-NEXT:  .LBB57_10: # %cond.load7
 ; RV32ZVE32F-NEXT:    vsetivli zero, 1, e32, m2, ta, ma
 ; RV32ZVE32F-NEXT:    vslidedown.vi v10, v8, 3
 ; RV32ZVE32F-NEXT:    vmv.x.s t2, v10
 ; RV32ZVE32F-NEXT:    lw t1, 4(t2)
 ; RV32ZVE32F-NEXT:    lw t2, 0(t2)
 ; RV32ZVE32F-NEXT:    andi t3, t0, 16
 ; RV32ZVE32F-NEXT:    beqz t3, .LBB57_5
-; RV32ZVE32F-NEXT:  .LBB57_13: # %cond.load10
+; RV32ZVE32F-NEXT:  .LBB57_11: # %cond.load10
 ; RV32ZVE32F-NEXT:    vsetivli zero, 1, e32, m2, ta, ma
 ; RV32ZVE32F-NEXT:    vslidedown.vi v10, v8, 4
 ; RV32ZVE32F-NEXT:    vmv.x.s t4, v10
 ; RV32ZVE32F-NEXT:    lw t3, 4(t4)
 ; RV32ZVE32F-NEXT:    lw t4, 0(t4)
 ; RV32ZVE32F-NEXT:    andi t5, t0, 32
 ; RV32ZVE32F-NEXT:    beqz t5, .LBB57_6
-; RV32ZVE32F-NEXT:  .LBB57_14: # %cond.load13
+; RV32ZVE32F-NEXT:  .LBB57_12: # %cond.load13
 ; RV32ZVE32F-NEXT:    vsetivli zero, 1, e32, m2, ta, ma
 ; RV32ZVE32F-NEXT:    vslidedown.vi v10, v8, 5
 ; RV32ZVE32F-NEXT:    vmv.x.s t6, v10
 ; RV32ZVE32F-NEXT:    lw t5, 4(t6)
 ; RV32ZVE32F-NEXT:    lw t6, 0(t6)
+; RV32ZVE32F-NEXT:  .LBB57_13: # %else14
+; RV32ZVE32F-NEXT:    addi sp, sp, -16
+; RV32ZVE32F-NEXT:    .cfi_def_cfa_offset 16
+; RV32ZVE32F-NEXT:    sw s0, 12(sp) # 4-byte Folded Spill
+; RV32ZVE32F-NEXT:    sw s1, 8(sp) # 4-byte Folded Spill
+; RV32ZVE32F-NEXT:    .cfi_offset s0, -4
+; RV32ZVE32F-NEXT:    .cfi_offset s1, -8
 ; RV32ZVE32F-NEXT:    andi s0, t0, 64
-; RV32ZVE32F-NEXT:    beqz s0, .LBB57_7
-; RV32ZVE32F-NEXT:  .LBB57_15: # %cond.load16
+; RV32ZVE32F-NEXT:    beqz s0, .LBB57_16
+; RV32ZVE32F-NEXT:  # %bb.14: # %cond.load16
 ; RV32ZVE32F-NEXT:    vsetivli zero, 1, e32, m2, ta, ma
 ; RV32ZVE32F-NEXT:    vslidedown.vi v10, v8, 6
 ; RV32ZVE32F-NEXT:    vmv.x.s s1, v10
 ; RV32ZVE32F-NEXT:    lw s0, 4(s1)
 ; RV32ZVE32F-NEXT:    lw s1, 0(s1)
 ; RV32ZVE32F-NEXT:    andi t0, t0, -128
-; RV32ZVE32F-NEXT:    beqz t0, .LBB57_8
-; RV32ZVE32F-NEXT:  .LBB57_16: # %cond.load19
+; RV32ZVE32F-NEXT:    bnez t0, .LBB57_17
+; RV32ZVE32F-NEXT:  .LBB57_15:
+; RV32ZVE32F-NEXT:    lw t0, 60(a3)
+; RV32ZVE32F-NEXT:    lw a3, 56(a3)
+; RV32ZVE32F-NEXT:    j .LBB57_18
+; RV32ZVE32F-NEXT:  .LBB57_16:
+; RV32ZVE32F-NEXT:    lw s0, 52(a3)
+; RV32ZVE32F-NEXT:    lw s1, 48(a3)
+; RV32ZVE32F-NEXT:    andi t0, t0, -128
+; RV32ZVE32F-NEXT:    beqz t0, .LBB57_15
+; RV32ZVE32F-NEXT:  .LBB57_17: # %cond.load19
 ; RV32ZVE32F-NEXT:    vsetivli zero, 1, e32, m2, ta, ma
 ; RV32ZVE32F-NEXT:    vslidedown.vi v8, v8, 7
 ; RV32ZVE32F-NEXT:    vmv.x.s a3, v8
 ; RV32ZVE32F-NEXT:    lw t0, 4(a3)
 ; RV32ZVE32F-NEXT:    lw a3, 0(a3)
-; RV32ZVE32F-NEXT:  .LBB57_17: # %else20
+; RV32ZVE32F-NEXT:  .LBB57_18: # %else20
 ; RV32ZVE32F-NEXT:    sw a2, 0(a0)
 ; RV32ZVE32F-NEXT:    sw a1, 4(a0)
 ; RV32ZVE32F-NEXT:    sw a5, 8(a0)
 ; RV32ZVE32F-NEXT:    sw a4, 12(a0)
 ; RV32ZVE32F-NEXT:    sw a7, 16(a0)
 ; RV32ZVE32F-NEXT:    sw a6, 20(a0)
 ; RV32ZVE32F-NEXT:    sw t2, 24(a0)
 ; RV32ZVE32F-NEXT:    sw t1, 28(a0)

llvm/test/CodeGen/Thumb2/LowOverheadLoops/fast-fp-loops.ll

for.body:                                         ; preds = %for.body, %for.body.preheader.new
   %niter.nsub.3 = add i32 %niter, -4
   %niter.ncmp.3 = icmp eq i32 %niter.nsub.3, 0
   br i1 %niter.ncmp.3, label %for.cond.cleanup.loopexit.unr-lcssa, label %for.body
 }
 
 define arm_aapcs_vfpcc float @fast_float_mac(ptr nocapture readonly %b, ptr nocapture readonly %c, i32 %N) {
 ; CHECK-LABEL: fast_float_mac:
 ; CHECK:       @ %bb.0: @ %entry
+; CHECK-NEXT:    cmp r2, #0
+; CHECK-NEXT:    itt eq
+; CHECK-NEXT:    vldreq s0, .LCPI1_0
+; CHECK-NEXT:    bxeq lr
+; CHECK-NEXT:  .LBB1_1: @ %vector.ph
 ; CHECK-NEXT:    push {r7, lr}
-; CHECK-NEXT:    cbz r2, .LBB1_4
-; CHECK-NEXT:  @ %bb.1: @ %vector.ph
 ; CHECK-NEXT:    adds r3, r2, #3
 ; CHECK-NEXT:    mov.w r12, #1
 ; CHECK-NEXT:    bic r3, r3, #3
 ; CHECK-NEXT:    vmov.i32 q0, #0x0
 ; CHECK-NEXT:    subs r3, #4
 ; CHECK-NEXT:    add.w r3, r12, r3, lsr #2
 ; CHECK-NEXT:    dls lr, r3
 ; CHECK-NEXT:  .LBB1_2: @ %vector.body

 ; CHECK-NEXT:  @ %bb.3: @ %middle.block
 ; CHECK-NEXT:    vpsel q0, q0, q1
 ; CHECK-NEXT:    vmov.f32 s4, s2
 ; CHECK-NEXT:    vmov.f32 s5, s3
 ; CHECK-NEXT:    vadd.f32 q0, q0, q1
 ; CHECK-NEXT:    vmov r0, s1
 ; CHECK-NEXT:    vadd.f32 q0, q0, r0
 ; CHECK-NEXT:    pop {r7, pc}
-; CHECK-NEXT:  .LBB1_4:
-; CHECK-NEXT:    vldr s0, .LCPI1_0
-; CHECK-NEXT:    pop {r7, pc}
 ; CHECK-NEXT:    .p2align 2
-; CHECK-NEXT:  @ %bb.5:
+; CHECK-NEXT:  @ %bb.4:
 ; CHECK-NEXT:  .LCPI1_0:
 ; CHECK-NEXT:    .long 0x00000000 @ float 0
 entry:
   %cmp8 = icmp eq i32 %N, 0
   br i1 %cmp8, label %for.cond.cleanup, label %vector.ph
 
 vector.ph:                                        ; preds = %entry
   %n.rnd.up = add i32 %N, 3

 for.cond.cleanup:                                 ; preds = %middle.block, %entry
   %a.0.lcssa = phi float [ 0.000000e+00, %entry ], [ %i9, %middle.block ]
   ret float %a.0.lcssa
 }
 
 define arm_aapcs_vfpcc float @fast_float_half_mac(ptr nocapture readonly %b, ptr nocapture readonly %c, i32 %N) {
 ; CHECK-LABEL: fast_float_half_mac:
 ; CHECK:       @ %bb.0: @ %entry
+; CHECK-NEXT:    cmp r2, #0
+; CHECK-NEXT:    itt eq
+; CHECK-NEXT:    vldreq s0, .LCPI2_0
+; CHECK-NEXT:    bxeq lr
+; CHECK-NEXT:  .LBB2_1: @ %vector.ph
 ; CHECK-NEXT:    push {r4, r5, r7, lr}
 ; CHECK-NEXT:    vpush {d8, d9, d10, d11, d12, d13}
 ; CHECK-NEXT:    sub sp, #8
-; CHECK-NEXT:    cmp r2, #0
-; CHECK-NEXT:    beq.w .LBB2_20
-; CHECK-NEXT:  @ %bb.1: @ %vector.ph
 ; CHECK-NEXT:    adds r3, r2, #3
 ; CHECK-NEXT:    vmov.i32 q5, #0x0
 ; CHECK-NEXT:    bic r3, r3, #3
 ; CHECK-NEXT:    sub.w r12, r3, #4
 ; CHECK-NEXT:    movs r3, #1
 ; CHECK-NEXT:    add.w lr, r3, r12, lsr #2
 ; CHECK-NEXT:    sub.w r12, r2, #1
 ; CHECK-NEXT:    adr r2, .LCPI2_1

 ; CHECK-NEXT:    vdup.32 q0, r12
 ; CHECK-NEXT:    vcmp.u32 cs, q0, q4
 ; CHECK-NEXT:    vpsel q0, q5, q3
 ; CHECK-NEXT:    vmov.f32 s4, s2
 ; CHECK-NEXT:    vmov.f32 s5, s3
 ; CHECK-NEXT:    vadd.f32 q0, q0, q1
 ; CHECK-NEXT:    vmov r0, s1
 ; CHECK-NEXT:    vadd.f32 q0, q0, r0
-; CHECK-NEXT:    b .LBB2_21
-; CHECK-NEXT:  .LBB2_20:
-; CHECK-NEXT:    vldr s0, .LCPI2_0
-; CHECK-NEXT:  .LBB2_21: @ %for.cond.cleanup
 ; CHECK-NEXT:    add sp, #8
 ; CHECK-NEXT:    vpop {d8, d9, d10, d11, d12, d13}
 ; CHECK-NEXT:    pop {r4, r5, r7, pc}
 ; CHECK-NEXT:    .p2align 4
-; CHECK-NEXT:  @ %bb.22:
+; CHECK-NEXT:  @ %bb.20:
 ; CHECK-NEXT:  .LCPI2_1:
 ; CHECK-NEXT:    .long 0 @ 0x0
 ; CHECK-NEXT:    .long 1 @ 0x1
 ; CHECK-NEXT:    .long 2 @ 0x2
 ; CHECK-NEXT:    .long 3 @ 0x3
 ; CHECK-NEXT:  .LCPI2_0:
 ; CHECK-NEXT:    .long 0x00000000 @ float 0
 entry:

llvm/test/CodeGen/X86/MachineSink-eflags.ll

 ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
 ; RUN: llc < %s | FileCheck %s
 target datalayout = "e-p:64:64:64-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:64:64-f32:32:32-f64:64:64-v64:64:64-v128:128:128-a0:0:64-s0:64:64-f80:128:128-n8:16:32:64"
 target triple = "x86_64-pc-linux"
 
 
 %0 = type <{ i64, i64, %1, %1, [21 x %2] }>
 %1 = type <{ i64, i64, i64 }>
 %2 = type <{ i32, i32, ptr addrspace(2) }>
 %3 = type { ptr, ptr, ptr, ptr, i32 }
 %4 = type <{ ptr, ptr, i32, i32, [4 x i64], [4 x i64], [4 x i64], [4 x i64], [4 x i64] }>
 %5 = type <{ ptr, ptr, ptr }>
 
 define void @foo(ptr nocapture %_stubArgs) nounwind {
 ; CHECK-LABEL: foo:
 ; CHECK:       # %bb.0: # %entry
-; CHECK-NEXT:    subq $152, %rsp
 ; CHECK-NEXT:    movq 48(%rdi), %rax
 ; CHECK-NEXT:    movl 64(%rdi), %ecx
 ; CHECK-NEXT:    movl $200, %esi
 ; CHECK-NEXT:    addl 68(%rdi), %esi
 ; CHECK-NEXT:    imull $46, %ecx, %edx
 ; CHECK-NEXT:    addq %rsi, %rdx
 ; CHECK-NEXT:    shlq $4, %rdx
 ; CHECK-NEXT:    imull $47, %ecx, %ecx
 ; CHECK-NEXT:    addq %rsi, %rcx
 ; CHECK-NEXT:    shlq $4, %rcx
 ; CHECK-NEXT:    cmpl $0, (%rdi)
 ; CHECK-NEXT:    jne .LBB0_1
 ; CHECK-NEXT:  # %bb.2: # %entry
 ; CHECK-NEXT:    xorps %xmm0, %xmm0
 ; CHECK-NEXT:    jmp .LBB0_3
 ; CHECK-NEXT:  .LBB0_1:
 ; CHECK-NEXT:    movaps (%rax,%rdx), %xmm0
 ; CHECK-NEXT:  .LBB0_3: # %entry
+; CHECK-NEXT:    leaq -{{[0-9]+}}(%rsp), %rsp
 ; CHECK-NEXT:    movaps (%rax,%rcx), %xmm1
 ; CHECK-NEXT:    movaps %xmm0, -{{[0-9]+}}(%rsp)
 ; CHECK-NEXT:    jne .LBB0_5
 ; CHECK-NEXT:  # %bb.4: # %entry
 ; CHECK-NEXT:    xorps %xmm1, %xmm1
 ; CHECK-NEXT:  .LBB0_5: # %entry
 ; CHECK-NEXT:    movaps %xmm1, -{{[0-9]+}}(%rsp)
 ; CHECK-NEXT:    addq $152, %rsp

llvm/test/CodeGen/X86/callbr-asm-label-addr.ll

 ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 2
 ; RUN: llc < %s -mtriple=x86_64-unknown-linux-gnu | FileCheck %s
 
 define i32 @test1(i32 %x) {
 ; CHECK-LABEL: test1:
 ; CHECK:       # %bb.0: # %entry
-; CHECK-NEXT:    pushq %rax
-; CHECK-NEXT:    .cfi_def_cfa_offset 16
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    .quad .Ltmp0
 ; CHECK-NEXT:    .quad .LBB0_1
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:  .LBB0_1: # Block address taken
 ; CHECK-NEXT:    # %bar
 ; CHECK-NEXT:    # Label of block must be emitted
+; CHECK-NEXT:    pushq %rax
+; CHECK-NEXT:    .cfi_def_cfa_offset 16
 ; CHECK-NEXT:    callq foo@PLT
+; CHECK-NEXT:    addq $8, %rsp
+; CHECK-NEXT:    .cfi_def_cfa_offset 8
 ; CHECK-NEXT:  .Ltmp0: # Block address taken
 ; CHECK-NEXT:  # %bb.2: # %baz
 ; CHECK-NEXT:    movl %eax, %edi
-; CHECK-NEXT:    popq %rax
-; CHECK-NEXT:    .cfi_def_cfa_offset 8
 ; CHECK-NEXT:    jmp mux@PLT # TAILCALL
 entry:
   callbr void asm sideeffect ".quad ${0:l}\0A\09.quad ${1:l}", "i,!i,~{dirflag},~{fpsr},~{flags}"(ptr blockaddress(@test1, %baz))
           to label %asm.fallthrough [label %bar]
 
 asm.fallthrough:
   br label %bar
 

llvm/test/CodeGen/X86/dag-update-nodetomatch.ll

entry:
   store i16 %conv52, ptr @b, align 2
   ret void
 }
 
 ; Test for PR31536
 define void @_Z2x6v() local_unnamed_addr {
 ; CHECK-LABEL: _Z2x6v:
 ; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    movq x1@GOTPCREL(%rip), %rax
+; CHECK-NEXT:    movl (%rax), %esi
+; CHECK-NEXT:    andl $511, %esi # imm = 0x1FF
+; CHECK-NEXT:    leaq 1(%rsi), %rax
+; CHECK-NEXT:    movq x4@GOTPCREL(%rip), %rcx
+; CHECK-NEXT:    movl %eax, (%rcx)
+; CHECK-NEXT:    movq x3@GOTPCREL(%rip), %rcx
+; CHECK-NEXT:    movl (%rcx), %edx
+; CHECK-NEXT:    testl %edx, %edx
+; CHECK-NEXT:    je .LBB1_18
+; CHECK-NEXT:  # %bb.1: # %for.cond1thread-pre-split.lr.ph
 ; CHECK-NEXT:    pushq %rbp
 ; CHECK-NEXT:    .cfi_def_cfa_offset 16
 ; CHECK-NEXT:    pushq %r15
 ; CHECK-NEXT:    .cfi_def_cfa_offset 24
 ; CHECK-NEXT:    pushq %r14
 ; CHECK-NEXT:    .cfi_def_cfa_offset 32
 ; CHECK-NEXT:    pushq %r13
 ; CHECK-NEXT:    .cfi_def_cfa_offset 40
 ; CHECK-NEXT:    pushq %r12
 ; CHECK-NEXT:    .cfi_def_cfa_offset 48
 ; CHECK-NEXT:    pushq %rbx
 ; CHECK-NEXT:    .cfi_def_cfa_offset 56
 ; CHECK-NEXT:    .cfi_offset %rbx, -56
 ; CHECK-NEXT:    .cfi_offset %r12, -48
 ; CHECK-NEXT:    .cfi_offset %r13, -40
 ; CHECK-NEXT:    .cfi_offset %r14, -32
 ; CHECK-NEXT:    .cfi_offset %r15, -24
 ; CHECK-NEXT:    .cfi_offset %rbp, -16
-; CHECK-NEXT:    movq x1@GOTPCREL(%rip), %rax
-; CHECK-NEXT:    movl (%rax), %esi
-; CHECK-NEXT:    andl $511, %esi # imm = 0x1FF
-; CHECK-NEXT:    leaq 1(%rsi), %rax
-; CHECK-NEXT:    movq x4@GOTPCREL(%rip), %rcx
-; CHECK-NEXT:    movl %eax, (%rcx)
-; CHECK-NEXT:    movq x3@GOTPCREL(%rip), %rcx
-; CHECK-NEXT:    movl (%rcx), %edx
-; CHECK-NEXT:    testl %edx, %edx
-; CHECK-NEXT:    je .LBB1_18
-; CHECK-NEXT:  # %bb.1: # %for.cond1thread-pre-split.lr.ph
 ; CHECK-NEXT:    movq x5@GOTPCREL(%rip), %rcx
 ; CHECK-NEXT:    movq (%rcx), %rdi
 ; CHECK-NEXT:    movl %edx, %ecx
 ; CHECK-NEXT:    notl %ecx
 ; CHECK-NEXT:    leaq 8(,%rcx,8), %rcx
 ; CHECK-NEXT:    imulq %rax, %rcx
 ; CHECK-NEXT:    addq %rdi, %rcx
 ; CHECK-NEXT:    movq %rcx, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill

 ; CHECK-NEXT:    jl .LBB1_14
 ; CHECK-NEXT:    jmp .LBB1_15
 ; CHECK-NEXT:  .LBB1_17: # %for.cond.for.end5_crit_edge
 ; CHECK-NEXT:    movq x5@GOTPCREL(%rip), %rax
 ; CHECK-NEXT:    movq {{[-0-9]+}}(%r{{[sb]}}p), %rcx # 8-byte Reload
 ; CHECK-NEXT:    movq %rcx, (%rax)
 ; CHECK-NEXT:    movq x3@GOTPCREL(%rip), %rax
 ; CHECK-NEXT:    movl $0, (%rax)
-; CHECK-NEXT:  .LBB1_18: # %for.end5
 ; CHECK-NEXT:    popq %rbx
 ; CHECK-NEXT:    .cfi_def_cfa_offset 48
 ; CHECK-NEXT:    popq %r12
 ; CHECK-NEXT:    .cfi_def_cfa_offset 40
 ; CHECK-NEXT:    popq %r13
 ; CHECK-NEXT:    .cfi_def_cfa_offset 32
 ; CHECK-NEXT:    popq %r14
 ; CHECK-NEXT:    .cfi_def_cfa_offset 24
 ; CHECK-NEXT:    popq %r15
 ; CHECK-NEXT:    .cfi_def_cfa_offset 16
 ; CHECK-NEXT:    popq %rbp
 ; CHECK-NEXT:    .cfi_def_cfa_offset 8
+; CHECK-NEXT:    .cfi_restore %rbx
+; CHECK-NEXT:    .cfi_restore %r12
+; CHECK-NEXT:    .cfi_restore %r13
+; CHECK-NEXT:    .cfi_restore %r14
+; CHECK-NEXT:    .cfi_restore %r15
+; CHECK-NEXT:    .cfi_restore %rbp
+; CHECK-NEXT:  .LBB1_18: # %for.end5
 ; CHECK-NEXT:    retq
 entry:
   %0 = load i32, ptr @x1, align 4
   %and = and i32 %0, 511
   %add = add nuw nsw i32 %and, 1
   store i32 %add, ptr @x4, align 4
   %.pr = load i32, ptr @x3, align 4
   %tobool8 = icmp eq i32 %.pr, 0

llvm/test/CodeGen/X86/fold-pcmpeqd-2.ll

; There should be no pcmpeqd instructions, everybody should the constant pool.
 
   %struct.__ImageExecInfo = type <{ <4 x i32>, <4 x float>, <2 x i64>, ptr, ptr, ptr, i32, i32, i32, i32, i32 }>
   %struct._cl_image_format_t = type <{ i32, i32, i32 }>
   %struct._image2d_t = type <{ ptr, %struct._cl_image_format_t, i32, i32, i32, i32, i32, i32 }>
 
 define void @program_1(ptr %dest, ptr %t0, <4 x float> %p0, <4 x float> %p1, <4 x float> %p4, <4 x float> %p5, <4 x float> %p6) nounwind {
 ; X32-LABEL: program_1:
 ; X32:       ## %bb.0: ## %entry
-; X32-NEXT:    pushl %esi
-; X32-NEXT:    subl $88, %esp
 ; X32-NEXT:    cmpl $0, 0
 ; X32-NEXT:    jle LBB0_2
 ; X32-NEXT:  ## %bb.1: ## %forcond
 ; X32-NEXT:    cmpl $0, 0
 ; X32-NEXT:    jg LBB0_3
 ; X32-NEXT:  LBB0_2: ## %ifthen
-; X32-NEXT:    addl $88, %esp
-; X32-NEXT:    popl %esi
 ; X32-NEXT:    retl
 ; X32-NEXT:  LBB0_3: ## %forbody
+; X32-NEXT:    pushl %esi
+; X32-NEXT:    subl $88, %esp
 ; X32-NEXT:    movaps {{.*#+}} xmm1 = [1.28E+2,1.28E+2,1.28E+2,1.28E+2]
 ; X32-NEXT:    minps {{\.?LCPI[0-9]+_[0-9]+}}, %xmm1
 ; X32-NEXT:    cvttps2dq %xmm1, %xmm0
 ; X32-NEXT:    cvtdq2ps %xmm0, %xmm0
 ; X32-NEXT:    subps %xmm0, %xmm1
 ; X32-NEXT:    movaps %xmm1, {{[-0-9]+}}(%e{{[sb]}}p) ## 16-byte Spill
 ; X32-NEXT:    movaps {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 ## 16-byte Reload
 ; X32-NEXT:    mulps {{\.?LCPI[0-9]+_[0-9]+}}, %xmm0

 ; X32-NEXT:    movaps {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 ## 16-byte Reload
 ; X32-NEXT:    movaps {{[-0-9]+}}(%e{{[sb]}}p), %xmm1 ## 16-byte Reload
 ; X32-NEXT:    movaps {{[-0-9]+}}(%e{{[sb]}}p), %xmm2 ## 16-byte Reload
 ; X32-NEXT:    calll *%esi
 ; X32-NEXT:    ud2
 ;
 ; X64-LABEL: program_1:
 ; X64:       ## %bb.0: ## %entry
-; X64-NEXT:    pushq %rbx
-; X64-NEXT:    subq $64, %rsp
 ; X64-NEXT:    cmpl $0, 0
 ; X64-NEXT:    jle LBB0_2
 ; X64-NEXT:  ## %bb.1: ## %forcond
 ; X64-NEXT:    cmpl $0, 0
 ; X64-NEXT:    jg LBB0_3
 ; X64-NEXT:  LBB0_2: ## %ifthen
-; X64-NEXT:    addq $64, %rsp
-; X64-NEXT:    popq %rbx
 ; X64-NEXT:    retq
 ; X64-NEXT:  LBB0_3: ## %forbody
+; X64-NEXT:    pushq %rbx
+; X64-NEXT:    subq $64, %rsp
 ; X64-NEXT:    xorps %xmm0, %xmm0
 ; X64-NEXT:    movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) ## 16-byte Spill
 ; X64-NEXT:    movaps {{.*#+}} xmm1 = [1.28E+2,1.28E+2,1.28E+2,1.28E+2]
 ; X64-NEXT:    minps {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm1
 ; X64-NEXT:    cvttps2dq %xmm1, %xmm0
 ; X64-NEXT:    cvtdq2ps %xmm0, %xmm0
 ; X64-NEXT:    subps %xmm0, %xmm1
 ; X64-NEXT:    movaps %xmm1, (%rsp) ## 16-byte Spill

llvm/test/CodeGen/X86/fp128-select.ll

 ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
 ; RUN: llc < %s -O2 -mtriple=x86_64-linux-android -mattr=+sse \
 ; RUN:     -enable-legalize-types-checking | FileCheck %s --check-prefix=SSE
 ; RUN: llc < %s -O2 -mtriple=x86_64-linux-gnu -mattr=+sse \
 ; RUN:     -enable-legalize-types-checking | FileCheck %s --check-prefix=SSE
 ; RUN: llc < %s -O2 -mtriple=x86_64-linux-android -mattr=-sse \
 ; RUN:     -enable-legalize-types-checking | FileCheck %s --check-prefix=NOSSE
 ; RUN: llc < %s -O2 -mtriple=x86_64-linux-gnu -mattr=-sse \
 ; RUN:     -enable-legalize-types-checking | FileCheck %s --check-prefix=NOSSE
 
 define void @test_select(ptr %p, ptr %q, i1 zeroext %c) {
 ; SSE-LABEL: test_select:
 ; SSE:       # %bb.0:
 ; SSE-NEXT:    testl %edx, %edx
 ; SSE-NEXT:    jne .LBB0_1
-; SSE-NEXT:  # %bb.2:
+; SSE-NEXT:  # %bb.3:
 ; SSE-NEXT:    movaps {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
 ; SSE-NEXT:    movaps %xmm0, (%rsi)
 ; SSE-NEXT:    retq
 ; SSE-NEXT:  .LBB0_1:
 ; SSE-NEXT:    movups (%rdi), %xmm0
 ; SSE-NEXT:    movaps %xmm0, (%rsi)
 ; SSE-NEXT:    retq
 ;

llvm/test/CodeGen/X86/i386-shrink-wrapping.ll

 
 ; Check that we are clobbering the flags when they are live-in of the
 ; prologue block and the prologue needs to adjust the stack.
 ; PR25607.
 
 define i32 @eflagsLiveInPrologue() #0 {
 ; ENABLE-LABEL: eflagsLiveInPrologue:
 ; ENABLE:       ## %bb.0: ## %entry
-; ENABLE-NEXT:    pushl %esi
-; ENABLE-NEXT:    subl $8, %esp
 ; ENABLE-NEXT:    movl L_a$non_lazy_ptr, %eax
 ; ENABLE-NEXT:    cmpl $0, (%eax)
 ; ENABLE-NEXT:    je LBB0_2
 ; ENABLE-NEXT:  ## %bb.1: ## %if.then
 ; ENABLE-NEXT:    movb $1, _d
 ; ENABLE-NEXT:  LBB0_2: ## %for.cond.preheader
 ; ENABLE-NEXT:    movl L_b$non_lazy_ptr, %eax
 ; ENABLE-NEXT:    movl (%eax), %eax
 ; ENABLE-NEXT:    testl %eax, %eax
 ; ENABLE-NEXT:    je LBB0_4
 ; ENABLE-NEXT:    .p2align 4, 0x90
 ; ENABLE-NEXT:  LBB0_3: ## %for.body
 ; ENABLE-NEXT:    ## =>This Inner Loop Header: Depth=1
 ; ENABLE-NEXT:    jmp LBB0_3
 ; ENABLE-NEXT:  LBB0_4: ## %for.end
+; ENABLE-NEXT:    pushl %esi
+; ENABLE-NEXT:    subl $8, %esp
 ; ENABLE-NEXT:    xorl %edx, %edx
 ; ENABLE-NEXT:    cmpb $0, _d
 ; ENABLE-NEXT:    movl $6, %ecx
 ; ENABLE-NEXT:    cmovnel %edx, %ecx
 ; ENABLE-NEXT:    movl L_e$non_lazy_ptr, %edx
 ; ENABLE-NEXT:    movb %cl, (%edx)
 ; ENABLE-NEXT:    leal 1(%ecx), %esi
 ; ENABLE-NEXT:    cltd

llvm/test/CodeGen/X86/inline-asm-flag-output.ll

 ; X32-NEXT:    jbe .LBB28_2
 ; X32-NEXT:  # %bb.1: # %then
 ; X32-NEXT:    calll bar@PLT
 ; X32-NEXT:  .LBB28_2: # %exit
 ; X32-NEXT:    retl
 ;
 ; X64-LABEL: test_cca_branch:
 ; X64:       # %bb.0: # %entry
-; X64-NEXT:    pushq %rax
 ; X64-NEXT:    #APP
 ; X64-NEXT:    cmp %rdi,(%rsi)
 ; X64-NEXT:    #NO_APP
 ; X64-NEXT:    jbe .LBB28_2
 ; X64-NEXT:  # %bb.1: # %then
+; X64-NEXT:    pushq %rax
 ; X64-NEXT:    callq bar@PLT
+; X64-NEXT:    addq $8, %rsp
 ; X64-NEXT:  .LBB28_2: # %exit
-; X64-NEXT:    popq %rax
 ; X64-NEXT:    retq
 entry:
   %cc = tail call i8 asm "cmp $2,$1", "={@cca},=*m,r,~{cc},~{dirflag},~{fpsr},~{flags}"(ptr elementtype(i64) %addr, i64 %nr) nounwind
   %tobool = icmp ne i8 %cc, 0
   br i1 %tobool, label %then, label %exit
 
 then:
   call void @bar()

 ; X32-NEXT:    jb .LBB29_2
 ; X32-NEXT:  # %bb.1: # %then
 ; X32-NEXT:    calll bar@PLT
 ; X32-NEXT:  .LBB29_2: # %exit
 ; X32-NEXT:    retl
 ;
 ; X64-LABEL: test_ccae_branch:
 ; X64:       # %bb.0: # %entry
-; X64-NEXT:    pushq %rax
 ; X64-NEXT:    #APP
 ; X64-NEXT:    cmp %rdi,(%rsi)
 ; X64-NEXT:    #NO_APP
 ; X64-NEXT:    jb .LBB29_2
 ; X64-NEXT:  # %bb.1: # %then
+; X64-NEXT:    pushq %rax
 ; X64-NEXT:    callq bar@PLT
+; X64-NEXT:    addq $8, %rsp
 ; X64-NEXT:  .LBB29_2: # %exit
-; X64-NEXT:    popq %rax
 ; X64-NEXT:    retq
 entry:
   %cc = tail call i8 asm "cmp $2,$1", "={@ccae},=*m,r,~{cc},~{dirflag},~{fpsr},~{flags}"(ptr elementtype(i64) %addr, i64 %nr) nounwind
   %tobool = icmp ne i8 %cc, 0
   br i1 %tobool, label %then, label %exit
 
 then:
   call void @bar()

 ; X32-NEXT:    jae .LBB30_2
 ; X32-NEXT:  # %bb.1: # %then
 ; X32-NEXT:    calll bar@PLT
 ; X32-NEXT:  .LBB30_2: # %exit
 ; X32-NEXT:    retl
 ;
 ; X64-LABEL: test_ccb_branch:
 ; X64:       # %bb.0: # %entry
-; X64-NEXT:    pushq %rax
 ; X64-NEXT:    #APP
 ; X64-NEXT:    cmp %rdi,(%rsi)
 ; X64-NEXT:    #NO_APP
 ; X64-NEXT:    jae .LBB30_2
 ; X64-NEXT:  # %bb.1: # %then
+; X64-NEXT:    pushq %rax
 ; X64-NEXT:    callq bar@PLT
+; X64-NEXT:    addq $8, %rsp
 ; X64-NEXT:  .LBB30_2: # %exit
-; X64-NEXT:    popq %rax
 ; X64-NEXT:    retq
 entry:
   %cc = tail call i8 asm "cmp $2,$1", "={@ccb},=*m,r,~{cc},~{dirflag},~{fpsr},~{flags}"(ptr elementtype(i64) %addr, i64 %nr) nounwind
   %tobool = icmp ne i8 %cc, 0
   br i1 %tobool, label %then, label %exit
 
 then:
   call void @bar()

 ; X32-NEXT:    ja .LBB31_2
 ; X32-NEXT:  # %bb.1: # %then
 ; X32-NEXT:    calll bar@PLT
 ; X32-NEXT:  .LBB31_2: # %exit
 ; X32-NEXT:    retl
 ;
 ; X64-LABEL: test_ccbe_branch:
 ; X64:       # %bb.0: # %entry
-; X64-NEXT:    pushq %rax
 ; X64-NEXT:    #APP
 ; X64-NEXT:    cmp %rdi,(%rsi)
 ; X64-NEXT:    #NO_APP
 ; X64-NEXT:    ja .LBB31_2
 ; X64-NEXT:  # %bb.1: # %then
+; X64-NEXT:    pushq %rax
 ; X64-NEXT:    callq bar@PLT
+; X64-NEXT:    addq $8, %rsp
 ; X64-NEXT:  .LBB31_2: # %exit
-; X64-NEXT:    popq %rax
 ; X64-NEXT:    retq
 entry:
   %cc = tail call i8 asm "cmp $2,$1", "={@ccbe},=*m,r,~{cc},~{dirflag},~{fpsr},~{flags}"(ptr elementtype(i64) %addr, i64 %nr) nounwind
   %tobool = icmp ne i8 %cc, 0
   br i1 %tobool, label %then, label %exit
 
 then:
   call void @bar()

 ; X32-NEXT:    jae .LBB32_2
 ; X32-NEXT:  # %bb.1: # %then
 ; X32-NEXT:    calll bar@PLT
 ; X32-NEXT:  .LBB32_2: # %exit
 ; X32-NEXT:    retl
 ;
 ; X64-LABEL: test_ccc_branch:
 ; X64:       # %bb.0: # %entry
-; X64-NEXT:    pushq %rax
 ; X64-NEXT:    #APP
 ; X64-NEXT:    cmp %rdi,(%rsi)
 ; X64-NEXT:    #NO_APP
 ; X64-NEXT:    jae .LBB32_2
 ; X64-NEXT:  # %bb.1: # %then
+; X64-NEXT:    pushq %rax
 ; X64-NEXT:    callq bar@PLT
+; X64-NEXT:    addq $8, %rsp
 ; X64-NEXT:  .LBB32_2: # %exit
-; X64-NEXT:    popq %rax
 ; X64-NEXT:    retq
 entry:
   %cc = tail call i8 asm "cmp $2,$1", "={@ccc},=*m,r,~{cc},~{dirflag},~{fpsr},~{flags}"(ptr elementtype(i64) %addr, i64 %nr) nounwind
   %tobool = icmp ne i8 %cc, 0
   br i1 %tobool, label %then, label %exit
 
 then:
   call void @bar()

 ; X32-NEXT:    jne .LBB33_2
 ; X32-NEXT:  # %bb.1: # %then
 ; X32-NEXT:    calll bar@PLT
 ; X32-NEXT:  .LBB33_2: # %exit
 ; X32-NEXT:    retl
 ;
 ; X64-LABEL: test_cce_branch:
 ; X64:       # %bb.0: # %entry
-; X64-NEXT:    pushq %rax
 ; X64-NEXT:    #APP
 ; X64-NEXT:    cmp %rdi,(%rsi)
 ; X64-NEXT:    #NO_APP
 ; X64-NEXT:    jne .LBB33_2
 ; X64-NEXT:  # %bb.1: # %then
+; X64-NEXT:    pushq %rax
 ; X64-NEXT:    callq bar@PLT
+; X64-NEXT:    addq $8, %rsp
 ; X64-NEXT:  .LBB33_2: # %exit
-; X64-NEXT:    popq %rax
 ; X64-NEXT:    retq
 entry:
   %cc = tail call i8 asm "cmp $2,$1", "={@cce},=*m,r,~{cc},~{dirflag},~{fpsr},~{flags}"(ptr elementtype(i64) %addr, i64 %nr) nounwind
   %tobool = icmp ne i8 %cc, 0
   br i1 %tobool, label %then, label %exit
 
 then:
   call void @bar()

 ; X32-NEXT:    jne .LBB34_2
 ; X32-NEXT:  # %bb.1: # %then
 ; X32-NEXT:    calll bar@PLT
 ; X32-NEXT:  .LBB34_2: # %exit
 ; X32-NEXT:    retl
 ;
 ; X64-LABEL: test_ccz_branch:
 ; X64:       # %bb.0: # %entry
-; X64-NEXT:    pushq %rax
 ; X64-NEXT:    #APP
 ; X64-NEXT:    cmp %rdi,(%rsi)
 ; X64-NEXT:    #NO_APP
 ; X64-NEXT:    jne .LBB34_2
 ; X64-NEXT:  # %bb.1: # %then
+; X64-NEXT:    pushq %rax
 ; X64-NEXT:    callq bar@PLT
+; X64-NEXT:    addq $8, %rsp
 ; X64-NEXT:  .LBB34_2: # %exit
-; X64-NEXT:    popq %rax
 ; X64-NEXT:    retq
 entry:
   %cc = tail call i8 asm "cmp $2,$1", "={@ccz},=*m,r,~{cc},~{dirflag},~{fpsr},~{flags}"(ptr elementtype(i64) %addr, i64 %nr) nounwind
   %tobool = icmp ne i8 %cc, 0
   br i1 %tobool, label %then, label %exit
 
 then:
   call void @bar()

 ; X32-NEXT:    jle .LBB35_2
 ; X32-NEXT:  # %bb.1: # %then
 ; X32-NEXT:    calll bar@PLT
 ; X32-NEXT:  .LBB35_2: # %exit
 ; X32-NEXT:    retl
 ;
 ; X64-LABEL: test_ccg_branch:
 ; X64:       # %bb.0: # %entry
-; X64-NEXT:    pushq %rax
 ; X64-NEXT:    #APP
 ; X64-NEXT:    cmp %rdi,(%rsi)
 ; X64-NEXT:    #NO_APP
 ; X64-NEXT:    jle .LBB35_2
 ; X64-NEXT:  # %bb.1: # %then
+; X64-NEXT:    pushq %rax
 ; X64-NEXT:    callq bar@PLT
+; X64-NEXT:    addq $8, %rsp
 ; X64-NEXT:  .LBB35_2: # %exit
-; X64-NEXT:    popq %rax
 ; X64-NEXT:    retq
 entry:
   %cc = tail call i8 asm "cmp $2,$1", "={@ccg},=*m,r,~{cc},~{dirflag},~{fpsr},~{flags}"(ptr elementtype(i64) %addr, i64 %nr) nounwind
   %tobool = icmp ne i8 %cc, 0
   br i1 %tobool, label %then, label %exit
 
 then:
   call void @bar()

 ; X32-NEXT:    jl .LBB36_2
 ; X32-NEXT:  # %bb.1: # %then
 ; X32-NEXT:    calll bar@PLT
 ; X32-NEXT:  .LBB36_2: # %exit
 ; X32-NEXT:    retl
 ;
 ; X64-LABEL: test_ccge_branch:
 ; X64:       # %bb.0: # %entry
-; X64-NEXT:    pushq %rax
 ; X64-NEXT:    #APP
 ; X64-NEXT:    cmp %rdi,(%rsi)
 ; X64-NEXT:    #NO_APP
 ; X64-NEXT:    jl .LBB36_2
 ; X64-NEXT:  # %bb.1: # %then
+; X64-NEXT:    pushq %rax
 ; X64-NEXT:    callq bar@PLT
+; X64-NEXT:    addq $8, %rsp
 ; X64-NEXT:  .LBB36_2: # %exit
-; X64-NEXT:    popq %rax
 ; X64-NEXT:    retq
 entry:
   %cc = tail call i8 asm "cmp $2,$1", "={@ccge},=*m,r,~{cc},~{dirflag},~{fpsr},~{flags}"(ptr elementtype(i64) %addr, i64 %nr) nounwind
   %tobool = icmp ne i8 %cc, 0
   br i1 %tobool, label %then, label %exit
 
 then:
   call void @bar()

 ; X32-NEXT:    jge .LBB37_2
 ; X32-NEXT:  # %bb.1: # %then
 ; X32-NEXT:    calll bar@PLT
 ; X32-NEXT:  .LBB37_2: # %exit
 ; X32-NEXT:    retl
 ;
 ; X64-LABEL: test_ccl_branch:
 ; X64:       # %bb.0: # %entry
-; X64-NEXT:    pushq %rax
 ; X64-NEXT:    #APP
 ; X64-NEXT:    cmp %rdi,(%rsi)
 ; X64-NEXT:    #NO_APP
 ; X64-NEXT:    jge .LBB37_2
 ; X64-NEXT:  # %bb.1: # %then
+; X64-NEXT:    pushq %rax
 ; X64-NEXT:    callq bar@PLT
+; X64-NEXT:    addq $8, %rsp
 ; X64-NEXT:  .LBB37_2: # %exit
-; X64-NEXT:    popq %rax
 ; X64-NEXT:    retq
 entry:
   %cc = tail call i8 asm "cmp $2,$1", "={@ccl},=*m,r,~{cc},~{dirflag},~{fpsr},~{flags}"(ptr elementtype(i64) %addr, i64 %nr) nounwind
   %tobool = icmp ne i8 %cc, 0
   br i1 %tobool, label %then, label %exit
 
 then:
   call void @bar()

 ; X32-NEXT:    jg .LBB38_2
 ; X32-NEXT:  # %bb.1: # %then
 ; X32-NEXT:    calll bar@PLT
 ; X32-NEXT:  .LBB38_2: # %exit
 ; X32-NEXT:    retl
 ;
 ; X64-LABEL: test_ccle_branch:
 ; X64:       # %bb.0: # %entry
-; X64-NEXT:    pushq %rax
 ; X64-NEXT:    #APP
 ; X64-NEXT:    cmp %rdi,(%rsi)
 ; X64-NEXT:    #NO_APP
 ; X64-NEXT:    jg .LBB38_2
 ; X64-NEXT:  # %bb.1: # %then
+; X64-NEXT:    pushq %rax
 ; X64-NEXT:    callq bar@PLT
+; X64-NEXT:    addq $8, %rsp
 ; X64-NEXT:  .LBB38_2: # %exit
-; X64-NEXT:    popq %rax
 ; X64-NEXT:    retq
 entry:
   %cc = tail call i8 asm "cmp $2,$1", "={@ccle},=*m,r,~{cc},~{dirflag},~{fpsr},~{flags}"(ptr elementtype(i64) %addr, i64 %nr) nounwind
   %tobool = icmp ne i8 %cc, 0
   br i1 %tobool, label %then, label %exit
 
 then:
   call void @bar()

 ; X32-NEXT:    ja .LBB39_2
 ; X32-NEXT:  # %bb.1: # %then
 ; X32-NEXT:    calll bar@PLT
 ; X32-NEXT:  .LBB39_2: # %exit
 ; X32-NEXT:    retl
 ;
 ; X64-LABEL: test_ccna_branch:
 ; X64:       # %bb.0: # %entry
-; X64-NEXT:    pushq %rax
 ; X64-NEXT:    #APP
 ; X64-NEXT:    cmp %rdi,(%rsi)
 ; X64-NEXT:    #NO_APP
 ; X64-NEXT:    ja .LBB39_2
 ; X64-NEXT:  # %bb.1: # %then
+; X64-NEXT:    pushq %rax
 ; X64-NEXT:    callq bar@PLT
+; X64-NEXT:    addq $8, %rsp
 ; X64-NEXT:  .LBB39_2: # %exit
-; X64-NEXT:    popq %rax
 ; X64-NEXT:    retq
 entry:
   %cc = tail call i8 asm "cmp $2,$1", "={@ccna},=*m,r,~{cc},~{dirflag},~{fpsr},~{flags}"(ptr elementtype(i64) %addr, i64 %nr) nounwind
   %tobool = icmp ne i8 %cc, 0
   br i1 %tobool, label %then, label %exit
 
 then:
   call void @bar()

 ; X32-NEXT:    jae .LBB40_2
 ; X32-NEXT:  # %bb.1: # %then
 ; X32-NEXT:    calll bar@PLT
 ; X32-NEXT:  .LBB40_2: # %exit
 ; X32-NEXT:    retl
 ;
 ; X64-LABEL: test_ccnae_branch:
 ; X64:       # %bb.0: # %entry
-; X64-NEXT:    pushq %rax
 ; X64-NEXT:    #APP
 ; X64-NEXT:    cmp %rdi,(%rsi)
 ; X64-NEXT:    #NO_APP
 ; X64-NEXT:    jae .LBB40_2
 ; X64-NEXT:  # %bb.1: # %then
+; X64-NEXT:    pushq %rax
 ; X64-NEXT:    callq bar@PLT
+; X64-NEXT:    addq $8, %rsp
 ; X64-NEXT:  .LBB40_2: # %exit
-; X64-NEXT:    popq %rax
 ; X64-NEXT:    retq
 entry:
   %cc = tail call i8 asm "cmp $2,$1", "={@ccnae},=*m,r,~{cc},~{dirflag},~{fpsr},~{flags}"(ptr elementtype(i64) %addr, i64 %nr) nounwind
   %tobool = icmp ne i8 %cc, 0
   br i1 %tobool, label %then, label %exit
 
 then:
   call void @bar()

 ; X32-NEXT:    jb .LBB41_2
 ; X32-NEXT:  # %bb.1: # %then
 ; X32-NEXT:    calll bar@PLT
 ; X32-NEXT:  .LBB41_2: # %exit
 ; X32-NEXT:    retl
 ;
 ; X64-LABEL: test_ccnb_branch:
 ; X64:       # %bb.0: # %entry
-; X64-NEXT:    pushq %rax
 ; X64-NEXT:    #APP
 ; X64-NEXT:    cmp %rdi,(%rsi)
 ; X64-NEXT:    #NO_APP
 ; X64-NEXT:    jb .LBB41_2
 ; X64-NEXT:  # %bb.1: # %then
+; X64-NEXT:    pushq %rax
 ; X64-NEXT:    callq bar@PLT
+; X64-NEXT:    addq $8, %rsp
 ; X64-NEXT:  .LBB41_2: # %exit
-; X64-NEXT:    popq %rax
 ; X64-NEXT:    retq
 entry:
   %cc = tail call i8 asm "cmp $2,$1", "={@ccnb},=*m,r,~{cc},~{dirflag},~{fpsr},~{flags}"(ptr elementtype(i64) %addr, i64 %nr) nounwind
   %tobool = icmp ne i8 %cc, 0
   br i1 %tobool, label %then, label %exit
 
 then:
   call void @bar()

 ; X32-NEXT:    jbe .LBB42_2
 ; X32-NEXT:  # %bb.1: # %then
 ; X32-NEXT:    calll bar@PLT
 ; X32-NEXT:  .LBB42_2: # %exit
 ; X32-NEXT:    retl
 ;
 ; X64-LABEL: test_ccnbe_branch:
 ; X64:       # %bb.0: # %entry
-; X64-NEXT:    pushq %rax
 ; X64-NEXT:    #APP
 ; X64-NEXT:    cmp %rdi,(%rsi)
 ; X64-NEXT:    #NO_APP
 ; X64-NEXT:    jbe .LBB42_2
 ; X64-NEXT:  # %bb.1: # %then
+; X64-NEXT:    pushq %rax
 ; X64-NEXT:    callq bar@PLT
+; X64-NEXT:    addq $8, %rsp
 ; X64-NEXT:  .LBB42_2: # %exit
-; X64-NEXT:    popq %rax
 ; X64-NEXT:    retq
 entry:
   %cc = tail call i8 asm "cmp $2,$1", "={@ccnbe},=*m,r,~{cc},~{dirflag},~{fpsr},~{flags}"(ptr elementtype(i64) %addr, i64 %nr) nounwind
   %tobool = icmp ne i8 %cc, 0
   br i1 %tobool, label %then, label %exit
 
 then:
   call void @bar()

 ; X32-NEXT:    jb .LBB43_2
 ; X32-NEXT:  # %bb.1: # %then
 ; X32-NEXT:    calll bar@PLT
 ; X32-NEXT:  .LBB43_2: # %exit
 ; X32-NEXT:    retl
 ;
 ; X64-LABEL: test_ccnc_branch:
 ; X64:       # %bb.0: # %entry
-; X64-NEXT:    pushq %rax
 ; X64-NEXT:    #APP
 ; X64-NEXT:    cmp %rdi,(%rsi)
 ; X64-NEXT:    #NO_APP
 ; X64-NEXT:    jb .LBB43_2
 ; X64-NEXT:  # %bb.1: # %then
+; X64-NEXT:    pushq %rax
 ; X64-NEXT:    callq bar@PLT
+; X64-NEXT:    addq $8, %rsp
 ; X64-NEXT:  .LBB43_2: # %exit
-; X64-NEXT:    popq %rax
 ; X64-NEXT:    retq
 entry:
   %cc = tail call i8 asm "cmp $2,$1", "={@ccnc},=*m,r,~{cc},~{dirflag},~{fpsr},~{flags}"(ptr elementtype(i64) %addr, i64 %nr) nounwind
   %tobool = icmp ne i8 %cc, 0
   br i1 %tobool, label %then, label %exit
 
 then:
   call void @bar()

 ; X32-NEXT:    je .LBB44_2
 ; X32-NEXT:  # %bb.1: # %then
 ; X32-NEXT:    calll bar@PLT
 ; X32-NEXT:  .LBB44_2: # %exit
 ; X32-NEXT:    retl
 ;
 ; X64-LABEL: test_ccne_branch:
 ; X64:       # %bb.0: # %entry
-; X64-NEXT:    pushq %rax
 ; X64-NEXT:    #APP
 ; X64-NEXT:    cmp %rdi,(%rsi)
 ; X64-NEXT:    #NO_APP
 ; X64-NEXT:    je .LBB44_2
 ; X64-NEXT:  # %bb.1: # %then
+; X64-NEXT:    pushq %rax
 ; X64-NEXT:    callq bar@PLT
+; X64-NEXT:    addq $8, %rsp
 ; X64-NEXT:  .LBB44_2: # %exit
-; X64-NEXT:    popq %rax
 ; X64-NEXT:    retq
 entry:
   %cc = tail call i8 asm "cmp $2,$1", "={@ccne},=*m,r,~{cc},~{dirflag},~{fpsr},~{flags}"(ptr elementtype(i64) %addr, i64 %nr) nounwind
   %tobool = icmp ne i8 %cc, 0
   br i1 %tobool, label %then, label %exit
 
 then:
   call void @bar()

 ; X32-NEXT:    je .LBB45_2
 ; X32-NEXT:  # %bb.1: # %then
 ; X32-NEXT:    calll bar@PLT
 ; X32-NEXT:  .LBB45_2: # %exit
 ; X32-NEXT:    retl
 ;
 ; X64-LABEL: test_ccnz_branch:
 ; X64:       # %bb.0: # %entry
-; X64-NEXT:    pushq %rax
 ; X64-NEXT:    #APP
 ; X64-NEXT:    cmp %rdi,(%rsi)
 ; X64-NEXT:    #NO_APP
 ; X64-NEXT:    je .LBB45_2
 ; X64-NEXT:  # %bb.1: # %then
+; X64-NEXT:    pushq %rax
 ; X64-NEXT:    callq bar@PLT
+; X64-NEXT:    addq $8, %rsp
 ; X64-NEXT:  .LBB45_2: # %exit
-; X64-NEXT:    popq %rax
 ; X64-NEXT:    retq
 entry:
   %cc = tail call i8 asm "cmp $2,$1", "={@ccnz},=*m,r,~{cc},~{dirflag},~{fpsr},~{flags}"(ptr elementtype(i64) %addr, i64 %nr) nounwind
   %tobool = icmp ne i8 %cc, 0
   br i1 %tobool, label %then, label %exit
 
 then:
   call void @bar()

 ; X32-NEXT:    jg .LBB46_2
 ; X32-NEXT:  # %bb.1: # %then
 ; X32-NEXT:    calll bar@PLT
 ; X32-NEXT:  .LBB46_2: # %exit
 ; X32-NEXT:    retl
 ;
 ; X64-LABEL: test_ccng_branch:
 ; X64:       # %bb.0: # %entry
-; X64-NEXT:    pushq %rax
 ; X64-NEXT:    #APP
 ; X64-NEXT:    cmp %rdi,(%rsi)
 ; X64-NEXT:    #NO_APP
 ; X64-NEXT:    jg .LBB46_2
 ; X64-NEXT:  # %bb.1: # %then
+; X64-NEXT:    pushq %rax
 ; X64-NEXT:    callq bar@PLT
+; X64-NEXT:    addq $8, %rsp
 ; X64-NEXT:  .LBB46_2: # %exit
-; X64-NEXT:    popq %rax
 ; X64-NEXT:    retq
 entry:
   %cc = tail call i8 asm "cmp $2,$1", "={@ccng},=*m,r,~{cc},~{dirflag},~{fpsr},~{flags}"(ptr elementtype(i64) %addr, i64 %nr) nounwind
   %tobool = icmp ne i8 %cc, 0
   br i1 %tobool, label %then, label %exit
 
 then:
   call void @bar()

 ; X32-NEXT:    jge .LBB47_2
 ; X32-NEXT:  # %bb.1: # %then
 ; X32-NEXT:    calll bar@PLT
 ; X32-NEXT:  .LBB47_2: # %exit
 ; X32-NEXT:    retl
 ;
 ; X64-LABEL: test_ccnge_branch:
 ; X64:       # %bb.0: # %entry
-; X64-NEXT:    pushq %rax
 ; X64-NEXT:    #APP
 ; X64-NEXT:    cmp %rdi,(%rsi)
 ; X64-NEXT:    #NO_APP
 ; X64-NEXT:    jge .LBB47_2
 ; X64-NEXT:  # %bb.1: # %then
+; X64-NEXT:    pushq %rax
 ; X64-NEXT:    callq bar@PLT
+; X64-NEXT:    addq $8, %rsp
 ; X64-NEXT:  .LBB47_2: # %exit
-; X64-NEXT:    popq %rax
 ; X64-NEXT:    retq
 entry:
   %cc = tail call i8 asm "cmp $2,$1", "={@ccnge},=*m,r,~{cc},~{dirflag},~{fpsr},~{flags}"(ptr elementtype(i64) %addr, i64 %nr) nounwind
   %tobool = icmp ne i8 %cc, 0
   br i1 %tobool, label %then, label %exit
 
 then:
   call void @bar()

 ; X32-NEXT:    jl .LBB48_2
 ; X32-NEXT:  # %bb.1: # %then
 ; X32-NEXT:    calll bar@PLT
 ; X32-NEXT:  .LBB48_2: # %exit
 ; X32-NEXT:    retl
 ;
 ; X64-LABEL: test_ccnl_branch:
 ; X64:       # %bb.0: # %entry
-; X64-NEXT:    pushq %rax
 ; X64-NEXT:    #APP
 ; X64-NEXT:    cmp %rdi,(%rsi)
 ; X64-NEXT:    #NO_APP
 ; X64-NEXT:    jl .LBB48_2
 ; X64-NEXT:  # %bb.1: # %then
+; X64-NEXT:    pushq %rax
 ; X64-NEXT:    callq bar@PLT
+; X64-NEXT:    addq $8, %rsp
 ; X64-NEXT:  .LBB48_2: # %exit
-; X64-NEXT:    popq %rax
 ; X64-NEXT:    retq
 entry:
   %cc = tail call i8 asm "cmp $2,$1", "={@ccnl},=*m,r,~{cc},~{dirflag},~{fpsr},~{flags}"(ptr elementtype(i64) %addr, i64 %nr) nounwind
   %tobool = icmp ne i8 %cc, 0
   br i1 %tobool, label %then, label %exit
 
 then:
   call void @bar()

 ; X32-NEXT:    jle .LBB49_2
 ; X32-NEXT:  # %bb.1: # %then
 ; X32-NEXT:    calll bar@PLT
 ; X32-NEXT:  .LBB49_2: # %exit
 ; X32-NEXT:    retl
 ;
 ; X64-LABEL: test_ccnle_branch:
 ; X64:       # %bb.0: # %entry
-; X64-NEXT:    pushq %rax
 ; X64-NEXT:    #APP
 ; X64-NEXT:    cmp %rdi,(%rsi)
 ; X64-NEXT:    #NO_APP
 ; X64-NEXT:    jle .LBB49_2
 ; X64-NEXT:  # %bb.1: # %then
+; X64-NEXT:    pushq %rax
 ; X64-NEXT:    callq bar@PLT
+; X64-NEXT:    addq $8, %rsp
 ; X64-NEXT:  .LBB49_2: # %exit
-; X64-NEXT:    popq %rax
 ; X64-NEXT:    retq
 entry:
   %cc = tail call i8 asm "cmp $2,$1", "={@ccnle},=*m,r,~{cc},~{dirflag},~{fpsr},~{flags}"(ptr elementtype(i64) %addr, i64 %nr) nounwind
   %tobool = icmp ne i8 %cc, 0
   br i1 %tobool, label %then, label %exit
 
 then:
   call void @bar()

 ; X32-NEXT:    jo .LBB50_2
 ; X32-NEXT:  # %bb.1: # %then
 ; X32-NEXT:    calll bar@PLT
 ; X32-NEXT:  .LBB50_2: # %exit
 ; X32-NEXT:    retl
 ;
 ; X64-LABEL: test_ccno_branch:
 ; X64:       # %bb.0: # %entry
-; X64-NEXT:    pushq %rax
 ; X64-NEXT:    #APP
 ; X64-NEXT:    cmp %rdi,(%rsi)
 ; X64-NEXT:    #NO_APP
 ; X64-NEXT:    jo .LBB50_2
 ; X64-NEXT:  # %bb.1: # %then
+; X64-NEXT:    pushq %rax
 ; X64-NEXT:    callq bar@PLT
+; X64-NEXT:    addq $8, %rsp
 ; X64-NEXT:  .LBB50_2: # %exit
-; X64-NEXT:    popq %rax
 ; X64-NEXT:    retq
 entry:
   %cc = tail call i8 asm "cmp $2,$1", "={@ccno},=*m,r,~{cc},~{dirflag},~{fpsr},~{flags}"(ptr elementtype(i64) %addr, i64 %nr) nounwind
   %tobool = icmp ne i8 %cc, 0
   br i1 %tobool, label %then, label %exit
 
 then:
   call void @bar()

 ; X32-NEXT:    jp .LBB51_2
 ; X32-NEXT:  # %bb.1: # %then
 ; X32-NEXT:    calll bar@PLT
 ; X32-NEXT:  .LBB51_2: # %exit
 ; X32-NEXT:    retl
 ;
 ; X64-LABEL: test_ccnp_branch:
 ; X64:       # %bb.0: # %entry
-; X64-NEXT:    pushq %rax
 ; X64-NEXT:    #APP
 ; X64-NEXT:    cmp %rdi,(%rsi)
 ; X64-NEXT:    #NO_APP
 ; X64-NEXT:    jp .LBB51_2
 ; X64-NEXT:  # %bb.1: # %then
+; X64-NEXT:    pushq %rax
 ; X64-NEXT:    callq bar@PLT
+; X64-NEXT:    addq $8, %rsp
 ; X64-NEXT:  .LBB51_2: # %exit
-; X64-NEXT:    popq %rax
 ; X64-NEXT:    retq
 entry:
   %cc = tail call i8 asm "cmp $2,$1", "={@ccnp},=*m,r,~{cc},~{dirflag},~{fpsr},~{flags}"(ptr elementtype(i64) %addr, i64 %nr) nounwind
   %tobool = icmp ne i8 %cc, 0
   br i1 %tobool, label %then, label %exit
 
 then:
   call void @bar()

 ; X32-NEXT:    js .LBB52_2
 ; X32-NEXT:  # %bb.1: # %then
 ; X32-NEXT:    calll bar@PLT
 ; X32-NEXT:  .LBB52_2: # %exit
 ; X32-NEXT:    retl
 ;
 ; X64-LABEL: test_ccns_branch:
 ; X64:       # %bb.0: # %entry
-; X64-NEXT:    pushq %rax
 ; X64-NEXT:    #APP
 ; X64-NEXT:    cmp %rdi,(%rsi)
 ; X64-NEXT:    #NO_APP
 ; X64-NEXT:    js .LBB52_2
 ; X64-NEXT:  # %bb.1: # %then
+; X64-NEXT:    pushq %rax
 ; X64-NEXT:    callq bar@PLT
+; X64-NEXT:    addq $8, %rsp
 ; X64-NEXT:  .LBB52_2: # %exit
-; X64-NEXT:    popq %rax
 ; X64-NEXT:    retq
 entry:
   %cc = tail call i8 asm "cmp $2,$1", "={@ccns},=*m,r,~{cc},~{dirflag},~{fpsr},~{flags}"(ptr elementtype(i64) %addr, i64 %nr) nounwind
   %tobool = icmp ne i8 %cc, 0
   br i1 %tobool, label %then, label %exit
 
 then:
   call void @bar()

 ; X32-NEXT:    jno .LBB53_2
 ; X32-NEXT:  # %bb.1: # %then
 ; X32-NEXT:    calll bar@PLT
 ; X32-NEXT:  .LBB53_2: # %exit
 ; X32-NEXT:    retl
 ;
 ; X64-LABEL: test_cco_branch:
 ; X64:       # %bb.0: # %entry
-; X64-NEXT:    pushq %rax
 ; X64-NEXT:    #APP
 ; X64-NEXT:    cmp %rdi,(%rsi)
 ; X64-NEXT:    #NO_APP
 ; X64-NEXT:    jno .LBB53_2
 ; X64-NEXT:  # %bb.1: # %then
+; X64-NEXT:    pushq %rax
 ; X64-NEXT:    callq bar@PLT
+; X64-NEXT:    addq $8, %rsp
 ; X64-NEXT:  .LBB53_2: # %exit
-; X64-NEXT:    popq %rax
 ; X64-NEXT:    retq
 entry:
   %cc = tail call i8 asm "cmp $2,$1", "={@cco},=*m,r,~{cc},~{dirflag},~{fpsr},~{flags}"(ptr elementtype(i64) %addr, i64 %nr) nounwind
   %tobool = icmp ne i8 %cc, 0
   br i1 %tobool, label %then, label %exit
 
 then:
   call void @bar()

 ; X32-NEXT:    jnp .LBB54_2
 ; X32-NEXT:  # %bb.1: # %then
 ; X32-NEXT:    calll bar@PLT
 ; X32-NEXT:  .LBB54_2: # %exit
 ; X32-NEXT:    retl
 ;
 ; X64-LABEL: test_ccp_branch:
 ; X64:       # %bb.0: # %entry
-; X64-NEXT:    pushq %rax
 ; X64-NEXT:    #APP
 ; X64-NEXT:    cmp %rdi,(%rsi)
 ; X64-NEXT:    #NO_APP
 ; X64-NEXT:    jnp .LBB54_2
 ; X64-NEXT:  # %bb.1: # %then
+; X64-NEXT:    pushq %rax
 ; X64-NEXT:    callq bar@PLT
+; X64-NEXT:    addq $8, %rsp
 ; X64-NEXT:  .LBB54_2: # %exit
-; X64-NEXT:    popq %rax
 ; X64-NEXT:    retq
 entry:
   %cc = tail call i8 asm "cmp $2,$1", "={@ccp},=*m,r,~{cc},~{dirflag},~{fpsr},~{flags}"(ptr elementtype(i64) %addr, i64 %nr) nounwind
   %tobool = icmp ne i8 %cc, 0
   br i1 %tobool, label %then, label %exit
 
 then:
   call void @bar()

 ; X32-NEXT:    jns .LBB55_2
 ; X32-NEXT:  # %bb.1: # %then
 ; X32-NEXT:    calll bar@PLT
 ; X32-NEXT:  .LBB55_2: # %exit
 ; X32-NEXT:    retl
 ;
 ; X64-LABEL: test_ccs_branch:
 ; X64:       # %bb.0: # %entry
-; X64-NEXT:    pushq %rax
 ; X64-NEXT:    #APP
 ; X64-NEXT:    cmp %rdi,(%rsi)
 ; X64-NEXT:    #NO_APP
 ; X64-NEXT:    jns .LBB55_2
 ; X64-NEXT:  # %bb.1: # %then
+; X64-NEXT:    pushq %rax
 ; X64-NEXT:    callq bar@PLT
+; X64-NEXT:    addq $8, %rsp
 ; X64-NEXT:  .LBB55_2: # %exit
-; X64-NEXT:    popq %rax
 ; X64-NEXT:    retq
 entry:
   %cc = tail call i8 asm "cmp $2,$1", "={@ccs},=*m,r,~{cc},~{dirflag},~{fpsr},~{flags}"(ptr elementtype(i64) %addr, i64 %nr) nounwind
   %tobool = icmp ne i8 %cc, 0
   br i1 %tobool, label %then, label %exit
 
 then:
   call void @bar()
   br label %exit
 
 exit:
   ret void
 }

llvm/test/CodeGen/X86/pr56103.ll

 ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
 ; RUN: llc < %s -mtriple=x86_64-linux-generic | FileCheck %s
 
 @e = global i16 0, align 2
 @a = global i32 0, align 4
 @c = global i32 0, align 4
 @b = global i64 0, align 8
 
 ; Check the test instruction won't be optimizated by peephole opt.
 
 define dso_local i32 @main() nounwind {
 ; CHECK-LABEL: main:
 ; CHECK:       # %bb.0: # %entry
-; CHECK-NEXT:    pushq %rax
 ; CHECK-NEXT:    movq e@GOTPCREL(%rip), %rax
 ; CHECK-NEXT:    movw $1, (%rax)
 ; CHECK-NEXT:    movq b@GOTPCREL(%rip), %rax
 ; CHECK-NEXT:    movq $1, (%rax)
 ; CHECK-NEXT:    movq a@GOTPCREL(%rip), %rax
 ; CHECK-NEXT:    movl (%rax), %ecx
 ; CHECK-NEXT:    movl $-2, %eax
 ; CHECK-NEXT:    sarl %cl, %eax
 ; CHECK-NEXT:    movq c@GOTPCREL(%rip), %rdx
 ; CHECK-NEXT:    movl (%rdx), %edx
 ; CHECK-NEXT:    decl %edx
 ; CHECK-NEXT:    movzwl %ax, %eax
 ; CHECK-NEXT:    decl %eax
 ; CHECK-NEXT:    xorl %edx, %eax
 ; CHECK-NEXT:    notl %ecx
 ; CHECK-NEXT:    andl %eax, %ecx
 ; CHECK-NEXT:    testq %rcx, %rcx
 ; CHECK-NEXT:    jle .LBB0_2
 ; CHECK-NEXT:  # %bb.1: # %if.end
 ; CHECK-NEXT:    xorl %eax, %eax
-; CHECK-NEXT:    popq %rcx
 ; CHECK-NEXT:    retq
 ; CHECK-NEXT:  .LBB0_2: # %if.then
+; CHECK-NEXT:    pushq %rax
 ; CHECK-NEXT:    callq abort@PLT
 entry:
   store i16 1, ptr @e, align 2
   store i64 1, ptr @b, align 8
   %0 = load i32, ptr @a, align 4
   %shr = ashr i32 -2, %0
   %1 = load i32, ptr @c, align 4
   %sub = add i32 %1, -1

llvm/test/CodeGen/X86/test-shrink-bug.ll

 ; CHECK-X86-NEXT:  ## %bb.2: ## %no
 ; CHECK-X86-NEXT:    calll _bar
 ; CHECK-X86-NEXT:  LBB1_3: ## %yes
 ; CHECK-X86-NEXT:    addl $12, %esp
 ; CHECK-X86-NEXT:    retl
 ;
 ; CHECK-X64-LABEL: fail:
 ; CHECK-X64:       # %bb.0:
-; CHECK-X64-NEXT:    pushq %rax
-; CHECK-X64-NEXT:    .cfi_def_cfa_offset 16
 ; CHECK-X64-NEXT:    testl $263, %edi # imm = 0x107
 ; CHECK-X64-NEXT:    je .LBB1_3
 ; CHECK-X64-NEXT:  # %bb.1:
 ; CHECK-X64-NEXT:    pcmpeqb {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
 ; CHECK-X64-NEXT:    pslldq {{.*#+}} xmm0 = zero,zero,zero,zero,zero,zero,zero,xmm0[0,1,2,3,4,5,6,7,8]
 ; CHECK-X64-NEXT:    pextrw $4, %xmm0, %eax
 ; CHECK-X64-NEXT:    testb $1, %al
 ; CHECK-X64-NEXT:    jne .LBB1_3
 ; CHECK-X64-NEXT:  # %bb.2: # %no
+; CHECK-X64-NEXT:    pushq %rax
+; CHECK-X64-NEXT:    .cfi_def_cfa_offset 16
 ; CHECK-X64-NEXT:    callq bar@PLT
-; CHECK-X64-NEXT:  .LBB1_3: # %yes
 ; CHECK-X64-NEXT:    popq %rax
 ; CHECK-X64-NEXT:    .cfi_def_cfa_offset 8
+; CHECK-X64-NEXT:  .LBB1_3: # %yes
 ; CHECK-X64-NEXT:    retq
   %1 = icmp eq <2 x i8> %b, <i8 40, i8 123>
   %2 = extractelement <2 x i1> %1, i32 1
   %3 = and i16 %a, 263
   %4 = icmp eq i16 %3, 0
   %merge = or i1 %4, %2
   br i1 %merge, label %yes, label %no
 

llvm/test/CodeGen/X86/xchg-nofold.ll

 ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
 ; RUN: llc -mtriple=x86_64-linux-gnu < %s | FileCheck %s
 
 %"struct.std::atomic" = type { %"struct.std::atomic_bool" }
 %"struct.std::atomic_bool" = type { %"struct.std::__atomic_base" }
 %"struct.std::__atomic_base" = type { i8 }
 
 ; CHECK-LABEL: _Z3fooRSt6atomicIbEb
 define zeroext i1 @_Z3fooRSt6atomicIbEb(ptr nocapture dereferenceable(1) %a, i1 returned zeroext %b) nounwind {
 ; CHECK-LABEL: _Z3fooRSt6atomicIbEb:
 ; CHECK:       # %bb.0: # %entry
-; CHECK-NEXT:    pushq %rax
 ; CHECK-NEXT:    movl %esi, %eax
 ; CHECK-NEXT:    movq %rdi, %rcx
 ; CHECK-NEXT:    shrq $3, %rcx
 ; CHECK-NEXT:    movzbl 2147450880(%rcx), %ecx
 ; CHECK-NEXT:    testb %cl, %cl
 ; CHECK-NEXT:    je .LBB0_3
 ; CHECK-NEXT:  # %bb.1:
 ; CHECK-NEXT:    movl %edi, %edx
 ; CHECK-NEXT:    andb $7, %dl
 ; CHECK-NEXT:    cmpb %cl, %dl
 ; CHECK-NEXT:    jge .LBB0_2
 ; CHECK-NEXT:  .LBB0_3:
 ; CHECK-NEXT:    movl %eax, %ecx
 ; CHECK-NEXT:    xchgb %cl, (%rdi)
 ; CHECK-NEXT:    # kill: def $al killed $al killed $eax
-; CHECK-NEXT:    popq %rcx
 ; CHECK-NEXT:    retq
 ; CHECK-NEXT:  .LBB0_2:
+; CHECK-NEXT:    pushq %rax
 ; CHECK-NEXT:    callq __asan_report_store1@PLT
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    #NO_APP
 entry:
   %frombool.i.i = zext i1 %b to i8
   %0 = ptrtoint ptr %a to i64
   %1 = lshr i64 %0, 3
   %2 = add i64 %1, 2147450880

llvm/test/Transforms/CodeGenPrepare/AArch64/large-offset-gep.ll

 ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
 ; RUN: llc -mtriple=aarch64-linux-gnu -verify-machineinstrs -o - %s | FileCheck %s
 
 %struct_type = type { [10000 x i32], i32, i32 }
 
 define void @test1(ptr %s, i32 %n) {
 ; CHECK-LABEL: test1:
 ; CHECK:       // %bb.0: // %entry
 ; CHECK-NEXT:    ldr x9, [x0]
-; CHECK-NEXT:    mov w10, #40000
+; CHECK-NEXT:    mov w10, #40000 // =0x9c40
 ; CHECK-NEXT:    mov w8, wzr
 ; CHECK-NEXT:    add x9, x9, x10
 ; CHECK-NEXT:    cmp w8, w1
 ; CHECK-NEXT:    b.ge .LBB0_2
 ; CHECK-NEXT:  .LBB0_1: // %while_body
 ; CHECK-NEXT:    // =>This Inner Loop Header: Depth=1
 ; CHECK-NEXT:    add w10, w8, #1
 ; CHECK-NEXT:    stp w10, w8, [x9]
while_end:
   ret void
 }
 
 define void @test2(ptr %struct, i32 %n) {
 ; CHECK-LABEL: test2:
 ; CHECK:       // %bb.0: // %entry
 ; CHECK-NEXT:    cbz x0, .LBB1_3
 ; CHECK-NEXT:  // %bb.1: // %while_cond.preheader
-; CHECK-NEXT:    mov w9, #40000
+; CHECK-NEXT:    mov w9, #40000 // =0x9c40
 ; CHECK-NEXT:    mov w8, wzr
 ; CHECK-NEXT:    add x9, x0, x9
 ; CHECK-NEXT:    cmp w8, w1
 ; CHECK-NEXT:    b.ge .LBB1_3
 ; CHECK-NEXT:  .LBB1_2: // %while_body
 ; CHECK-NEXT:    // =>This Inner Loop Header: Depth=1
 ; CHECK-NEXT:    add w10, w8, #1
 ; CHECK-NEXT:    stp w10, w8, [x9]

 
 define void @test3(ptr %s1, ptr %s2, i1 %cond, i32 %n) {
 ; CHECK-LABEL: test3:
 ; CHECK:       // %bb.0: // %entry
 ; CHECK-NEXT:    tst w2, #0x1
 ; CHECK-NEXT:    csel x9, x1, x0, ne
 ; CHECK-NEXT:    cbz x9, .LBB2_3
 ; CHECK-NEXT:  // %bb.1: // %while_cond.preheader
-; CHECK-NEXT:    mov w10, #40000
+; CHECK-NEXT:    mov w10, #40000 // =0x9c40
 ; CHECK-NEXT:    mov w8, wzr
 ; CHECK-NEXT:    add x9, x9, x10
 ; CHECK-NEXT:    cmp w8, w3
 ; CHECK-NEXT:    b.ge .LBB2_3
 ; CHECK-NEXT:  .LBB2_2: // %while_body
 ; CHECK-NEXT:    // =>This Inner Loop Header: Depth=1
 ; CHECK-NEXT:    add w10, w8, #1
 ; CHECK-NEXT:    stp w10, w8, [x9]

 ; CHECK-NEXT:    stp x20, x19, [sp, #16] // 16-byte Folded Spill
 ; CHECK-NEXT:    .cfi_offset w19, -8
 ; CHECK-NEXT:    .cfi_offset w20, -16
 ; CHECK-NEXT:    .cfi_offset w21, -24
 ; CHECK-NEXT:    .cfi_offset w30, -32
 ; CHECK-NEXT:    .cfi_remember_state
 ; CHECK-NEXT:    mov w19, w0
 ; CHECK-NEXT:    mov w20, wzr
-; CHECK-NEXT:    mov w21, #40000
+; CHECK-NEXT:    mov w21, #40000 // =0x9c40
 ; CHECK-NEXT:  .LBB3_1: // %while_cond
 ; CHECK-NEXT:    // =>This Inner Loop Header: Depth=1
 ; CHECK-NEXT:  .Ltmp0:
 ; CHECK-NEXT:    bl foo
 ; CHECK-NEXT:  .Ltmp1:
 ; CHECK-NEXT:  // %bb.2: // %while_cond_x.split
 ; CHECK-NEXT:    // in Loop: Header=BB3_1 Depth=1
 ; CHECK-NEXT:    add x8, x0, x21
while_end:
   ret void
 }
 
 declare ptr @llvm.strip.invariant.group.p0(ptr)
 
 define void @test_invariant_group(i32 %arg, i1 %c) {
 ; CHECK-LABEL: test_invariant_group:
 ; CHECK:       // %bb.0: // %bb
-; CHECK-NEXT:    tbz w1, #0, .LBB5_3
+; CHECK-NEXT:    tbz w1, #0, .LBB5_4
 ; CHECK-NEXT:  // %bb.1: // %bb6
-; CHECK-NEXT:    cbz w0, .LBB5_4
+; CHECK-NEXT:    cbz w0, .LBB5_3
 ; CHECK-NEXT:  .LBB5_2: // %bb1
 ; CHECK-NEXT:    // =>This Inner Loop Header: Depth=1
 ; CHECK-NEXT:    tbnz w1, #0, .LBB5_2
-; CHECK-NEXT:  .LBB5_3: // %bb5
+; CHECK-NEXT:    b .LBB5_4
+; CHECK-NEXT:  .LBB5_3: // %bb2
+; CHECK-NEXT:    tbz w1, #0, .LBB5_5
+; CHECK-NEXT:  .LBB5_4: // %bb5
 ; CHECK-NEXT:    ret
-; CHECK-NEXT:  .LBB5_4: // %bb2
-; CHECK-NEXT:    tbnz w1, #0, .LBB5_3
-; CHECK-NEXT:  // %bb.5: // %bb4
-; CHECK-NEXT:    mov w8, #1
+; CHECK-NEXT:  .LBB5_5: // %bb4
+; CHECK-NEXT:    mov w8, #1 // =0x1
 ; CHECK-NEXT:    str x8, [x8]
 ; CHECK-NEXT:    ret
 bb:
   br i1 %c, label %bb6, label %bb5
 
 bb1:                                              ; preds = %bb6, %bb1
   br i1 %c, label %bb1, label %bb5