[Transforms] Apply fixes from performance-for-range-copy (NFC)

Author: kazutakahirata

llvm/lib/Transforms/IPO/MemProfContextDisambiguation.cpp

@@ -758,7 +758,7 @@ void CallsiteContextGraph<DerivedCCG, FuncTy, CallTy>::
   auto UpdateCallers = [&](ContextNode *Node,
                            DenseSet<const ContextEdge *> &Visited,
                            auto &&UpdateCallers) -> void {
-    for (auto Edge : Node->CallerEdges) {
+    for (const auto &Edge : Node->CallerEdges) {
       auto Inserted = Visited.insert(Edge.get());
       if (!Inserted.second)
         continue;
@@ -1119,7 +1119,7 @@ void CallsiteContextGraph<DerivedCCG, FuncTy, CallTy>::updateStackNodes() {
       // not fully matching stack contexts. To do this, subtract any context ids
       // found in caller nodes of the last node found above.
       if (Ids.back() != getLastStackId(Call)) {
-        for (auto PE : LastNode->CallerEdges) {
+        for (const auto &PE : LastNode->CallerEdges) {
           set_subtract(StackSequenceContextIds, PE->getContextIds());
           if (StackSequenceContextIds.empty())
             break;

llvm/lib/Transforms/IPO/PartialInlining.cpp

@@ -981,7 +981,7 @@ PartialInlinerImpl::FunctionCloner::FunctionCloner(
 
   // Go through all Outline Candidate Regions and update all BasicBlock
   // information.
-  for (FunctionOutliningMultiRegionInfo::OutlineRegionInfo RegionInfo :
+  for (const FunctionOutliningMultiRegionInfo::OutlineRegionInfo &RegionInfo :
        OI->ORI) {
     SmallVector<BasicBlock *, 8> Region;
     for (BasicBlock *BB : RegionInfo.Region)

llvm/lib/Transforms/Instrumentation/PGOInstrumentation.cpp

@@ -1338,7 +1338,7 @@ static uint64_t computeStackId(const memprof::Frame &Frame) {
 static void addCallStack(CallStackTrie &AllocTrie,
                          const AllocationInfo *AllocInfo) {
   SmallVector<uint64_t> StackIds;
-  for (auto StackFrame : AllocInfo->CallStack)
+  for (const auto &StackFrame : AllocInfo->CallStack)
     StackIds.push_back(computeStackId(StackFrame));
   auto AllocType = getAllocType(AllocInfo->Info.getMaxAccessCount(),
                                 AllocInfo->Info.getMinSize(),

llvm/lib/Transforms/Scalar/DFAJumpThreading.cpp

@@ -582,7 +582,7 @@ struct AllSwitchPaths {
         continue;
 
       PathsType SuccPaths = paths(Succ, Visited, PathDepth + 1);
-      for (PathType Path : SuccPaths) {
+      for (const PathType &Path : SuccPaths) {
         PathType NewPath(Path);
         NewPath.push_front(BB);
         Res.push_back(NewPath);
@@ -935,7 +935,7 @@ struct TransformDFA {
     SSAUpdaterBulk SSAUpdate;
     SmallVector<Use *, 16> UsesToRename;
 
-    for (auto KV : NewDefs) {
+    for (const auto &KV : NewDefs) {
       Instruction *I = KV.first;
       BasicBlock *BB = I->getParent();
       std::vector<Instruction *> Cloned = KV.second;

llvm/lib/Transforms/Scalar/SimpleLoopUnswitch.cpp

@@ -2762,7 +2762,7 @@ static int CalculateUnswitchCostMultiplier(
   // Count amount of clones that all the candidates might cause during
   // unswitching. Branch/guard counts as 1, switch counts as log2 of its cases.
   int UnswitchedClones = 0;
-  for (auto Candidate : UnswitchCandidates) {
+  for (const auto &Candidate : UnswitchCandidates) {
     const Instruction *CI = Candidate.TI;
     const BasicBlock *CondBlock = CI->getParent();
     bool SkipExitingSuccessors = DT.dominates(CondBlock, Latch);

llvm/lib/Transforms/Utils/Local.cpp

@@ -1948,7 +1948,7 @@ Value *getSalvageOpsForGEP(GetElementPtrInst *GEP, const DataLayout &DL,
     Opcodes.insert(Opcodes.begin(), {dwarf::DW_OP_LLVM_arg, 0});
     CurrentLocOps = 1;
   }
-  for (auto Offset : VariableOffsets) {
+  for (const auto &Offset : VariableOffsets) {
     AdditionalValues.push_back(Offset.first);
     assert(Offset.second.isStrictlyPositive() &&
            "Expected strictly positive multiplier for offset.");

llvm/lib/Transforms/Utils/LoopVersioning.cpp

@@ -210,7 +210,7 @@ void LoopVersioning::prepareNoAliasMetadata() {
   // Finally, transform the above to actually map to scope list which is what
   // the metadata uses.
 
-  for (auto Pair : GroupToNonAliasingScopes)
+  for (const auto &Pair : GroupToNonAliasingScopes)
     GroupToNonAliasingScopeList[Pair.first] = MDNode::get(Context, Pair.second);
 }
 

llvm/lib/Transforms/Utils/UnifyLoopExits.cpp

@@ -113,7 +113,7 @@ static void restoreSSA(const DominatorTree &DT, const Loop *L,
     }
   }
 
-  for (auto II : ExternalUsers) {
+  for (const auto &II : ExternalUsers) {
     // For each Def used outside the loop, create NewPhi in
     // LoopExitBlock. NewPhi receives Def only along exiting blocks that
     // dominate it, while the remaining values are undefined since those paths

llvm/lib/Transforms/Vectorize/LoopVectorize.cpp

@@ -8492,7 +8492,7 @@ VPWidenCallRecipe *VPRecipeBuilder::tryToWidenCall(CallInst *CI,
       VFShape Shape = VFShape::get(*CI, VariantVF, /*HasGlobalPred=*/true);
       unsigned MaskPos = 0;
 
-      for (VFInfo Info : VFDatabase::getMappings(*CI))
+      for (const VFInfo &Info : VFDatabase::getMappings(*CI))
         if (Info.Shape == Shape) {
           assert(Info.isMasked() && "Vector function info shape mismatch");
           MaskPos = Info.getParamIndexForOptionalMask().value();

llvm/lib/Transforms/Vectorize/VectorCombine.cpp

@@ -1565,7 +1565,7 @@ bool VectorCombine::foldSelectShuffle(Instruction &I, bool FromReduction) {
   // Calculate our ReconstructMasks from the OrigReconstructMasks and the
   // modified order of the input shuffles.
   SmallVector<SmallVector<int>> ReconstructMasks;
-  for (auto Mask : OrigReconstructMasks) {
+  for (const auto &Mask : OrigReconstructMasks) {
     SmallVector<int> ReconstructMask;
     for (int M : Mask) {
       auto FindIndex = [](const SmallVector<std::pair<int, int>> &V, int M) {