ADT: Give ilist<T>::reverse_iterator a handle to the current node

Reverse iterators to doubly-linked lists can be simpler (and cheaper)
than std::reverse_iterator.  Make it so.

In particular, change ilist<T>::reverse_iterator so that it is *never*
invalidated unless the node it references is deleted.  This matches the
guarantees of ilist<T>::iterator.

(Note: MachineBasicBlock::iterator is *not* an ilist iterator, but a
MachineInstrBundleIterator<MachineInstr>.  This commit does not change
MachineBasicBlock::reverse_iterator, but it does update
MachineBasicBlock::reverse_instr_iterator.  See note at end of commit
message for details on bundle iterators.)

Given the list (with the Sentinel showing twice for simplicity):

     [Sentinel] <-> A <-> B <-> [Sentinel]

the following is now true:
 1. begin() represents A.
 2. begin() holds the pointer for A.
 3. end() represents [Sentinel].
 4. end() holds the poitner for [Sentinel].
 5. rbegin() represents B.
 6. rbegin() holds the pointer for B.
 7. rend() represents [Sentinel].
 8. rend() holds the pointer for [Sentinel].

The changes are apple#6 and apple#8.  Here are some properties from the old
scheme (which used std::reverse_iterator):
- rbegin() held the pointer for [Sentinel] and rend() held the pointer
  for A;
- operator*() cost two dereferences instead of one;
- converting from a valid iterator to its valid reverse_iterator
  involved a confusing increment; and
- "RI++->erase()" left RI invalid.  The unintuitive replacement was
  "RI->erase(), RE = end()".

With vector-like data structures these properties are hard to avoid
(since past-the-beginning is not a valid pointer), and don't impose a
real cost (since there's still only one dereference, and all iterators
are invalidated on erase).  But with lists, this was a poor design.

Specifically, the following code (which obviously works with normal
iterators) now works with ilist::reverse_iterator as well:

    for (auto RI = L.rbegin(), RE = L.rend(); RI != RE;)
      fooThatMightRemoveArgFromList(*RI++);

Converting between iterator and reverse_iterator for the same node uses
the getReverse() function.

    reverse_iterator iterator::getReverse();
    iterator reverse_iterator::getReverse();

Why doesn't iterator <=> reverse_iterator conversion use constructors?

In order to catch and update old code, reverse_iterator does not even
have an explicit conversion from iterator.  It wouldn't be safe because
there would be no reasonable way to catch all the bugs from the changed
semantic (see the changes at call sites that are part of this patch).

Old code used this API:

    std::reverse_iterator::reverse_iterator(iterator);
    iterator std::reverse_iterator::base();

Here's how to update from old code to new (that incorporates the
semantic change), assuming I is an ilist<>::iterator and RI is an
ilist<>::reverse_iterator:

            [Old]         ==>          [New]
    reverse_iterator(I)       (--I).getReverse()
    reverse_iterator(I)         ++I.getReverse()
  --reverse_iterator(I)           I.getReverse()
    reverse_iterator(++I)         I.getReverse()
          RI.base()          (--RI).getReverse()
          RI.base()            ++RI.getReverse()
        --RI.base()              RI.getReverse()
      (++RI).base()              RI.getReverse()
  delete &*RI, RE = end()         delete &*RI++
  RI->erase(), RE = end()         RI++->erase()

=======================================
Note: bundle iterators are out of scope
=======================================

MachineBasicBlock::iterator, also known as
MachineInstrBundleIterator<MachineInstr>, is a wrapper to represent
MachineInstr bundles.  The idea is that each operator++ takes you to the
beginning of the next bundle.  Implementing a sane reverse iterator for
this is harder than ilist.  Here are the options:
- Use std::reverse_iterator<MBB::i>.  Store a handle to the beginning of
  the next bundle.  A call to operator*() runs a loop (usually
  operator--() will be called 1 time, for unbundled instructions).
  Increment/decrement just works.  This is the status quo.
- Store a handle to the final node in the bundle.  A call to operator*()
  still runs a loop, but it iterates one time fewer (usually
  operator--() will be called 0 times, for unbundled instructions).
  Increment/decrement just works.
- Make the ilist_sentinel<MachineInstr> *always* store that it's the
  sentinel (instead of just in asserts mode).  Then the bundle iterator
  can sniff the sentinel bit in operator++().

I initially tried implementing the end() option as part of this commit,
but updating iterator/reverse_iterator conversion call sites was
error-prone.  I have a WIP series of patches that implements the final
option.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@280032 91177308-0d34-0410-b5e6-96231b3b80d8

Author: dexonsmith

include/llvm/ADT/ilist.h

@@ -35,7 +35,7 @@
 namespace llvm {
 
 template<typename NodeTy, typename Traits> class iplist;
-template<typename NodeTy> class ilist_iterator;
+template <typename NodeTy, bool IsReverse> class ilist_iterator;
 
 /// An access class for ilist_node private API.
 ///
@@ -146,12 +146,30 @@ template <class NodeTy> struct ConstCorrectNodeType {
 template <class NodeTy> struct ConstCorrectNodeType<const NodeTy> {
   typedef const ilist_node<NodeTy> type;
 };
+
+template <bool IsReverse = false> struct IteratorHelper {
+  template <class T> static void increment(T *&I) {
+    I = ilist_node_access::getNext(*I);
+  }
+  template <class T> static void decrement(T *&I) {
+    I = ilist_node_access::getPrev(*I);
+  }
+};
+template <> struct IteratorHelper<true> {
+  template <class T> static void increment(T *&I) {
+    IteratorHelper<false>::decrement(I);
+  }
+  template <class T> static void decrement(T *&I) {
+    IteratorHelper<false>::increment(I);
+  }
+};
+
 } // end namespace ilist_detail
 
 //===----------------------------------------------------------------------===//
 // Iterator for intrusive list.
 //
-template <typename NodeTy>
+template <typename NodeTy, bool IsReverse>
 class ilist_iterator
     : public std::iterator<std::bidirectional_iterator_tag, NodeTy, ptrdiff_t> {
 public:
@@ -185,19 +203,30 @@ class ilist_iterator
   // a nonconst iterator...
   template <class node_ty>
   ilist_iterator(
-      const ilist_iterator<node_ty> &RHS,
+      const ilist_iterator<node_ty, IsReverse> &RHS,
       typename std::enable_if<std::is_convertible<node_ty *, NodeTy *>::value,
                               void *>::type = nullptr)
       : NodePtr(RHS.getNodePtr()) {}
 
   // This is templated so that we can allow assigning to a const iterator from
   // a nonconst iterator...
   template <class node_ty>
-  const ilist_iterator &operator=(const ilist_iterator<node_ty> &RHS) {
+  const ilist_iterator &
+  operator=(const ilist_iterator<node_ty, IsReverse> &RHS) {
     NodePtr = RHS.getNodePtr();
     return *this;
   }
 
+  /// Convert from an iterator to its reverse.
+  ///
+  /// TODO: Roll this into the implicit constructor once we're sure that no one
+  /// is relying on the std::reverse_iterator off-by-one semantics.
+  ilist_iterator<NodeTy, !IsReverse> getReverse() const {
+    if (NodePtr)
+      return ilist_iterator<NodeTy, !IsReverse>(*NodePtr);
+    return ilist_iterator<NodeTy, !IsReverse>();
+  }
+
   void reset(pointer NP) { NodePtr = NP; }
 
   // Accessors...
@@ -217,12 +246,11 @@ class ilist_iterator
 
   // Increment and decrement operators...
   ilist_iterator &operator--() {
-    NodePtr = ilist_node_access::getPrev(*NodePtr);
-    assert(NodePtr && "--'d off the beginning of an ilist!");
+    ilist_detail::IteratorHelper<IsReverse>::decrement(NodePtr);
     return *this;
   }
   ilist_iterator &operator++() {
-    NodePtr = ilist_node_access::getNext(*NodePtr);
+    ilist_detail::IteratorHelper<IsReverse>::increment(NodePtr);
     return *this;
   }
   ilist_iterator operator--(int) {
@@ -356,8 +384,8 @@ class iplist : public Traits, ilist_base, ilist_node_access {
   typedef ilist_iterator<const NodeTy> const_iterator;
   typedef size_t size_type;
   typedef ptrdiff_t difference_type;
-  typedef std::reverse_iterator<const_iterator>  const_reverse_iterator;
-  typedef std::reverse_iterator<iterator>  reverse_iterator;
+  typedef ilist_iterator<const NodeTy, true> const_reverse_iterator;
+  typedef ilist_iterator<NodeTy, true> reverse_iterator;
 
   iplist() = default;
   ~iplist() { clear(); }
@@ -369,11 +397,10 @@ class iplist : public Traits, ilist_base, ilist_node_access {
   const_iterator end() const { return const_iterator(Sentinel); }
 
   // reverse iterator creation methods.
-  reverse_iterator rbegin()            { return reverse_iterator(end()); }
-  const_reverse_iterator rbegin() const{ return const_reverse_iterator(end()); }
-  reverse_iterator rend()              { return reverse_iterator(begin()); }
-  const_reverse_iterator rend() const { return const_reverse_iterator(begin());}
-
+  reverse_iterator rbegin()            { return ++reverse_iterator(Sentinel); }
+  const_reverse_iterator rbegin() const{ return ++const_reverse_iterator(Sentinel); }
+  reverse_iterator rend()              { return reverse_iterator(Sentinel); }
+  const_reverse_iterator rend() const { return const_reverse_iterator(Sentinel); }
 
   // Miscellaneous inspection routines.
   size_type max_size() const { return size_type(-1); }

include/llvm/ADT/ilist_node.h

@@ -51,15 +51,16 @@ class ilist_node_base {
 };
 
 struct ilist_node_access;
-template <typename NodeTy> class ilist_iterator;
+template <typename NodeTy, bool IsReverse = false> class ilist_iterator;
 template <typename NodeTy> class ilist_sentinel;
 
 /// Templated wrapper class.
 template <typename NodeTy> class ilist_node : ilist_node_base {
   friend class ilist_base;
   friend struct ilist_node_access;
   friend struct ilist_traits<NodeTy>;
-  friend class ilist_iterator<NodeTy>;
+  friend class ilist_iterator<NodeTy, false>;
+  friend class ilist_iterator<NodeTy, true>;
   friend class ilist_sentinel<NodeTy>;
 
 protected:

include/llvm/CodeGen/MachineBasicBlock.h

@@ -150,9 +150,8 @@ class MachineBasicBlock
 
   typedef Instructions::iterator                                 instr_iterator;
   typedef Instructions::const_iterator                     const_instr_iterator;
-  typedef std::reverse_iterator<instr_iterator>          reverse_instr_iterator;
-  typedef
-  std::reverse_iterator<const_instr_iterator>      const_reverse_instr_iterator;
+  typedef Instructions::reverse_iterator reverse_instr_iterator;
+  typedef Instructions::const_reverse_iterator const_reverse_instr_iterator;
 
   typedef MachineInstrBundleIterator<MachineInstr> iterator;
   typedef MachineInstrBundleIterator<const MachineInstr> const_iterator;
@@ -193,10 +192,14 @@ class MachineBasicBlock
   const_iterator          begin() const { return instr_begin();  }
   iterator                end  ()       { return instr_end();    }
   const_iterator          end  () const { return instr_end();    }
-  reverse_iterator       rbegin()       { return instr_rbegin(); }
-  const_reverse_iterator rbegin() const { return instr_rbegin(); }
-  reverse_iterator       rend  ()       { return instr_rend();   }
-  const_reverse_iterator rend  () const { return instr_rend();   }
+  reverse_iterator rbegin() { return reverse_iterator(end()); }
+  const_reverse_iterator rbegin() const {
+    return const_reverse_iterator(end());
+  }
+  reverse_iterator rend() { return reverse_iterator(begin()); }
+  const_reverse_iterator rend() const {
+    return const_reverse_iterator(begin());
+  }
 
   /// Support for MachineInstr::getNextNode().
   static Instructions MachineBasicBlock::*getSublistAccess(MachineInstr *) {

include/llvm/CodeGen/MachineFunction.h

@@ -445,8 +445,8 @@ class MachineFunction {
   // Provide accessors for the MachineBasicBlock list...
   typedef BasicBlockListType::iterator iterator;
   typedef BasicBlockListType::const_iterator const_iterator;
-  typedef std::reverse_iterator<const_iterator> const_reverse_iterator;
-  typedef std::reverse_iterator<iterator>             reverse_iterator;
+  typedef BasicBlockListType::const_reverse_iterator const_reverse_iterator;
+  typedef BasicBlockListType::reverse_iterator reverse_iterator;
 
   /// Support for MachineBasicBlock::getNextNode().
   static BasicBlockListType MachineFunction::*

include/llvm/IR/SymbolTableListTraits.h

@@ -30,10 +30,6 @@
 namespace llvm {
 class ValueSymbolTable;
 
-template <typename NodeTy> class ilist_iterator;
-template <typename NodeTy, typename Traits> class iplist;
-template <typename Ty> struct ilist_traits;
-
 /// Template metafunction to get the parent type for a symbol table list.
 ///
 /// Implementations create a typedef called \c type so that we only need a
@@ -66,6 +62,7 @@ template <typename NodeTy> class SymbolTableList;
 template <typename ValueSubClass>
 class SymbolTableListTraits : public ilist_node_traits<ValueSubClass> {
   typedef SymbolTableList<ValueSubClass> ListTy;
+  typedef ilist_iterator<ValueSubClass, false> iterator;
   typedef
       typename SymbolTableListParentType<ValueSubClass>::type ItemParentClass;
 
@@ -94,10 +91,9 @@ class SymbolTableListTraits : public ilist_node_traits<ValueSubClass> {
 public:
   void addNodeToList(ValueSubClass *V);
   void removeNodeFromList(ValueSubClass *V);
-  void transferNodesFromList(SymbolTableListTraits &L2,
-                             ilist_iterator<ValueSubClass> first,
-                             ilist_iterator<ValueSubClass> last);
-//private:
+  void transferNodesFromList(SymbolTableListTraits &L2, iterator first,
+                             iterator last);
+  // private:
   template<typename TPtr>
   void setSymTabObject(TPtr *, TPtr);
   static ValueSymbolTable *toPtr(ValueSymbolTable *P) { return P; }

lib/CodeGen/MachinePipeliner.cpp

@@ -2880,8 +2880,7 @@ void SwingSchedulerDAG::removeDeadInstructions(MachineBasicBlock *KernelBB,
         used = false;
       }
       if (!used) {
-        MI->eraseFromParent();
-        ME = (*MBB)->instr_rend();
+        MI++->eraseFromParent();
         continue;
       }
       ++MI;

lib/Target/Lanai/LanaiDelaySlotFiller.cpp

@@ -105,7 +105,7 @@ bool Filler::runOnMachineBasicBlock(MachineBasicBlock &MBB) {
         // RET is generated as part of epilogue generation and hence we know
         // what the two instructions preceding it are and that it is safe to
         // insert RET above them.
-        MachineBasicBlock::reverse_instr_iterator RI(I);
+        MachineBasicBlock::reverse_instr_iterator RI = ++I.getReverse();
         assert(RI->getOpcode() == Lanai::LDW_RI && RI->getOperand(0).isReg() &&
                RI->getOperand(0).getReg() == Lanai::FP &&
                RI->getOperand(1).isReg() &&
@@ -117,8 +117,7 @@ bool Filler::runOnMachineBasicBlock(MachineBasicBlock &MBB) {
                RI->getOperand(0).getReg() == Lanai::SP &&
                RI->getOperand(1).isReg() &&
                RI->getOperand(1).getReg() == Lanai::FP);
-        ++RI;
-        MachineBasicBlock::instr_iterator FI(RI.base());
+        MachineBasicBlock::instr_iterator FI = RI.getReverse();
         MBB.splice(std::next(I), &MBB, FI, I);
         FilledSlots += 2;
       } else {
@@ -154,14 +153,14 @@ bool Filler::findDelayInstr(MachineBasicBlock &MBB,
   bool SawLoad = false;
   bool SawStore = false;
 
-  for (MachineBasicBlock::reverse_instr_iterator I(Slot); I != MBB.instr_rend();
-       ++I) {
+  for (MachineBasicBlock::reverse_instr_iterator I = ++Slot.getReverse();
+       I != MBB.instr_rend(); ++I) {
     // skip debug value
     if (I->isDebugValue())
       continue;
 
     // Convert to forward iterator.
-    MachineBasicBlock::instr_iterator FI(std::next(I).base());
+    MachineBasicBlock::instr_iterator FI = I.getReverse();
 
     if (I->hasUnmodeledSideEffects() || I->isInlineAsm() || I->isLabel() ||
         FI == LastFiller || I->isPseudo())

lib/Target/X86/X86FixupSetCC.cpp

@@ -99,7 +99,8 @@ bool X86FixupSetCCPass::isSetCCr(unsigned Opcode) {
 MachineInstr *
 X86FixupSetCCPass::findFlagsImpDef(MachineBasicBlock *MBB,
                                    MachineBasicBlock::reverse_iterator MI) {
-  auto MBBStart = MBB->instr_rend();
+  // FIXME: Should this be instr_rend(), and MI be reverse_instr_iterator?
+  auto MBBStart = MBB->rend();
   for (int i = 0; (i < SearchBound) && (MI != MBBStart); ++i, ++MI)
     for (auto &Op : MI->implicit_operands())
       if ((Op.getReg() == X86::EFLAGS) && (Op.isDef()))

lib/Transforms/Scalar/LoopRerollPass.cpp

@@ -1412,13 +1412,12 @@ bool LoopReroll::DAGRootTracker::validate(ReductionTracker &Reductions) {
 void LoopReroll::DAGRootTracker::replace(const SCEV *IterCount) {
   BasicBlock *Header = L->getHeader();
   // Remove instructions associated with non-base iterations.
-  for (BasicBlock::reverse_iterator J = Header->rbegin();
-       J != Header->rend();) {
+  for (BasicBlock::reverse_iterator J = Header->rbegin(), JE = Header->rend();
+       J != JE;) {
     unsigned I = Uses[&*J].find_first();
     if (I > 0 && I < IL_All) {
-      Instruction *D = &*J;
-      DEBUG(dbgs() << "LRR: removing: " << *D << "\n");
-      D->eraseFromParent();
+      DEBUG(dbgs() << "LRR: removing: " << *J << "\n");
+      J++->eraseFromParent();
       continue;
     }
 

lib/Transforms/Scalar/RewriteStatepointsForGC.cpp

@@ -2578,8 +2578,8 @@ static void findLiveSetAtInst(Instruction *Inst, GCPtrLivenessData &Data,
   // call result is not live (normal), nor are it's arguments
   // (unless they're used again later).  This adjustment is
   // specifically what we need to relocate
-  BasicBlock::reverse_iterator rend(Inst->getIterator());
-  computeLiveInValues(BB->rbegin(), rend, LiveOut);
+  computeLiveInValues(BB->rbegin(), ++Inst->getIterator().getReverse(),
+                      LiveOut);
   LiveOut.remove(Inst);
   Out.insert(LiveOut.begin(), LiveOut.end());
 }

lib/Transforms/Vectorize/SLPVectorizer.cpp

@@ -1868,9 +1868,9 @@ int BoUpSLP::getSpillCost() {
       );
 
     // Now find the sequence of instructions between PrevInst and Inst.
-    BasicBlock::reverse_iterator InstIt(Inst->getIterator()),
-        PrevInstIt(PrevInst->getIterator());
-    --PrevInstIt;
+    BasicBlock::reverse_iterator InstIt = ++Inst->getIterator().getReverse(),
+                                 PrevInstIt =
+                                     PrevInst->getIterator().getReverse();
     while (InstIt != PrevInstIt) {
       if (PrevInstIt == PrevInst->getParent()->rend()) {
         PrevInstIt = Inst->getParent()->rbegin();
@@ -3036,9 +3036,10 @@ bool BoUpSLP::BlockScheduling::extendSchedulingRegion(Value *V) {
   }
   // Search up and down at the same time, because we don't know if the new
   // instruction is above or below the existing scheduling region.
-  BasicBlock::reverse_iterator UpIter(ScheduleStart->getIterator());
+  BasicBlock::reverse_iterator UpIter =
+      ++ScheduleStart->getIterator().getReverse();
   BasicBlock::reverse_iterator UpperEnd = BB->rend();
-  BasicBlock::iterator DownIter(ScheduleEnd);
+  BasicBlock::iterator DownIter = ScheduleEnd->getIterator();
   BasicBlock::iterator LowerEnd = BB->end();
   for (;;) {
     if (++ScheduleRegionSize > ScheduleRegionSizeLimit) {

unittests/ADT/CMakeLists.txt

@@ -19,6 +19,7 @@ set(ADTSources
   HashingTest.cpp
   ilistTestTemp.cpp
   IListBaseTest.cpp
+  IListIteratorTest.cpp
   IListNodeBaseTest.cpp
   IListSentinelTest.cpp
   ImmutableMapTest.cpp