Count references to interference cache entries.

Each InterferenceCache::Cursor instance references a cache entry. A
non-zero reference count guarantees that the entry won't be reused for a
new register.

This makes it possible to have multiple live cursors examining
interference for different physregs.

The total number of live cursors into a cache must be kept below
InterferenceCache::getMaxCursors().

Code generation should be unaffected by this change, and it doesn't seem
to affect the cache replacement strategy either.

llvm-svn: 135121

lib/CodeGen/InterferenceCache.cpp

@@ -14,6 +14,7 @@
 #define DEBUG_TYPE "regalloc"
 #include "InterferenceCache.h"
 #include "llvm/Target/TargetRegisterInfo.h"
+#include "llvm/Support/ErrorHandling.h"
 
 using namespace llvm;
 
@@ -40,9 +41,18 @@ InterferenceCache::Entry *InterferenceCache::get(unsigned PhysReg) {
   E = RoundRobin;
   if (++RoundRobin == CacheEntries)
     RoundRobin = 0;
-  Entries[E].reset(PhysReg, LIUArray, TRI, MF);
-  PhysRegEntries[PhysReg] = E;
-  return &Entries[E];
+  for (unsigned i = 0; i != CacheEntries; ++i) {
+    // Skip entries that are in use.
+    if (Entries[E].hasRefs()) {
+      if (++E == CacheEntries)
+        E = 0;
+      continue;
+    }
+    Entries[E].reset(PhysReg, LIUArray, TRI, MF);
+    PhysRegEntries[PhysReg] = E;
+    return &Entries[E];
+  }
+  llvm_unreachable("Ran out of interference cache entries.");
 }
 
 /// revalidate - LIU contents have changed, update tags.
@@ -59,6 +69,7 @@ void InterferenceCache::Entry::reset(unsigned physReg,
                                      LiveIntervalUnion *LIUArray,
                                      const TargetRegisterInfo *TRI,
                                      const MachineFunction *MF) {
+  assert(!hasRefs() && "Cannot reset cache entry with references");
   // LIU's changed, invalidate cache.
   ++Tag;
   PhysReg = physReg;

lib/CodeGen/InterferenceCache.h

@@ -43,6 +43,9 @@ class InterferenceCache {
     /// change.
     unsigned Tag;
 
+    /// RefCount - The total number of Cursor instances referring to this Entry.
+    unsigned RefCount;
+
     /// MF - The current function.
     MachineFunction *MF;
 
@@ -68,9 +71,10 @@ class InterferenceCache {
     void update(unsigned MBBNum);
 
   public:
-    Entry() : PhysReg(0), Tag(0), Indexes(0) {}
+    Entry() : PhysReg(0), Tag(0), RefCount(0), Indexes(0) {}
 
     void clear(MachineFunction *mf, SlotIndexes *indexes) {
+      assert(!hasRefs() && "Cannot clear cache entry with references");
       PhysReg = 0;
       MF = mf;
       Indexes = indexes;
@@ -78,6 +82,10 @@ class InterferenceCache {
 
     unsigned getPhysReg() const { return PhysReg; }
 
+    void addRef(int Delta) { RefCount += Delta; }
+
+    bool hasRefs() const { return RefCount > 0; }
+
     void revalidate();
 
     /// valid - Return true if this is a valid entry for physReg.
@@ -122,18 +130,47 @@ public:
   void init(MachineFunction*, LiveIntervalUnion*, SlotIndexes*,
             const TargetRegisterInfo *);
 
+  /// getMaxCursors - Return the maximum number of concurrent cursors that can
+  /// be supported.
+  unsigned getMaxCursors() const { return CacheEntries; }
+
   /// Cursor - The primary query interface for the block interference cache.
   class Cursor {
     Entry *CacheEntry;
     BlockInterference *Current;
+
+    void setEntry(Entry *E) {
+      // Update reference counts. Nothing happens when RefCount reaches 0, so
+      // we don't have to check for E == CacheEntry etc.
+      if (CacheEntry)
+        CacheEntry->addRef(-1);
+      CacheEntry = E;
+      if (CacheEntry)
+        CacheEntry->addRef(+1);
+      Current = 0;
+    }
+
   public:
     /// Cursor - Create a dangling cursor.
     Cursor() : CacheEntry(0), Current(0) {}
+    ~Cursor() { setEntry(0); }
+
+    Cursor(const Cursor &O) {
+      setEntry(O.CacheEntry);
+    }
+
+    Cursor &operator=(const Cursor &O) {
+      setEntry(O.CacheEntry);
+      return *this;
+    }
 
     /// setPhysReg - Point this cursor to PhysReg's interference.
     void setPhysReg(InterferenceCache &Cache, unsigned PhysReg) {
-      CacheEntry = Cache.get(PhysReg);
-      Current = 0;
+      // Release reference before getting a new one. That guarantees we can
+      // actually have CacheEntries live cursors.
+      setEntry(0);
+      if (PhysReg)
+        setEntry(Cache.get(PhysReg));
     }
 
     /// moveTo - Move cursor to basic block MBBNum.

lib/CodeGen/RegAllocGreedy.cpp

@@ -854,11 +854,6 @@ void RAGreedy::splitAroundRegion(LiveInterval &VirtReg,
   });
 
   InterferenceCache::Cursor &Intf = Cand.Intf;
-
-  // FIXME: We need cache reference counts to guarantee that Intf hasn't been
-  // clobbered.
-  Intf.setPhysReg(IntfCache, Cand.PhysReg);
-
   LiveRangeEdit LREdit(VirtReg, NewVRegs, this);
   SE->reset(LREdit);
 
@@ -1252,6 +1247,22 @@ unsigned RAGreedy::tryRegionSplit(LiveInterval &VirtReg, AllocationOrder &Order,
 
   Order.rewind();
   while (unsigned PhysReg = Order.next()) {
+    // Discard bad candidates before we run out of interference cache cursors.
+    // This will only affect register classes with a lot of registers (>32).
+    if (NumCands == IntfCache.getMaxCursors()) {
+      unsigned WorstCount = ~0u;
+      unsigned Worst = 0;
+      for (unsigned i = 0; i != NumCands; ++i) {
+        if (i == BestCand)
+          continue;
+        unsigned Count = GlobalCand[i].LiveBundles.count();
+        if (Count < WorstCount)
+          Worst = i, WorstCount = Count;
+      }
+      --NumCands;
+      GlobalCand[Worst] = GlobalCand[NumCands];
+    }
+
     if (GlobalCand.size() <= NumCands)
       GlobalCand.resize(NumCands+1);
     GlobalSplitCandidate &Cand = GlobalCand[NumCands];