diff --git a/bench/RefCntBench.cpp b/bench/RefCntBench.cpp
index 351513b83..f846b1ac6 100644
--- a/bench/RefCntBench.cpp
+++ b/bench/RefCntBench.cpp
@@ -14,6 +14,54 @@ enum {
     M = 2
 };
 
+class AtomicInc32 : public Benchmark {
+public:
+    AtomicInc32() : fX(0) {}
+
+    virtual bool isSuitableFor(Backend backend) SK_OVERRIDE {
+        return backend == kNonRendering_Backend;
+    }
+
+protected:
+    virtual const char* onGetName() {
+        return "atomic_inc_32";
+    }
+
+    virtual void onDraw(const int loops, SkCanvas*) {
+        for (int i = 0; i < loops; ++i) {
+            sk_atomic_inc(&fX);
+        }
+    }
+
+private:
+    int32_t fX;
+    typedef Benchmark INHERITED;
+};
+
+class AtomicInc64 : public Benchmark {
+public:
+    AtomicInc64() : fX(0) {}
+
+    virtual bool isSuitableFor(Backend backend) SK_OVERRIDE {
+        return backend == kNonRendering_Backend;
+    }
+
+protected:
+    virtual const char* onGetName() {
+        return "atomic_inc_64";
+    }
+
+    virtual void onDraw(const int loops, SkCanvas*) {
+        for (int i = 0; i < loops; ++i) {
+            sk_atomic_inc(&fX);
+        }
+    }
+
+private:
+    int64_t fX;
+    typedef Benchmark INHERITED;
+};
+
 class RefCntBench_Stack : public Benchmark {
 public:
     virtual bool isSuitableFor(Backend backend) SK_OVERRIDE {
@@ -191,6 +239,9 @@ private:
 
 ///////////////////////////////////////////////////////////////////////////////
 
+DEF_BENCH( return new AtomicInc32(); )
+DEF_BENCH( return new AtomicInc64(); )
+
 DEF_BENCH( return new RefCntBench_Stack(); )
 DEF_BENCH( return new RefCntBench_Heap(); )
 DEF_BENCH( return new RefCntBench_New(); )
diff --git a/include/core/SkThread.h b/include/core/SkThread.h
index 4f7f32609..403b288f1 100644
--- a/include/core/SkThread.h
+++ b/include/core/SkThread.h
@@ -16,22 +16,26 @@
  *  No additional memory barrier is required; this must act as a compiler barrier.
  */
 static int32_t sk_atomic_inc(int32_t* addr);
+static int64_t sk_atomic_inc(int64_t* addr);
 
 /** Atomically adds inc to the int referenced by addr and returns the previous value.
  *  No additional memory barrier is required; this must act as a compiler barrier.
  */
 static int32_t sk_atomic_add(int32_t* addr, int32_t inc);
+static int64_t sk_atomic_add(int64_t* addr, int64_t inc);
 
 /** Atomically subtracts one from the int referenced by addr and returns the previous value.
  *  This must act as a release (SL/S) memory barrier and as a compiler barrier.
  */
 static int32_t sk_atomic_dec(int32_t* addr);
+static int64_t sk_atomic_dec(int64_t* addr);
 
 /** Atomic compare and set.
  *  If *addr == before, set *addr to after and return true, otherwise return false.
  *  This must act as a release (SL/S) memory barrier and as a compiler barrier.
  */
 static bool sk_atomic_cas(int32_t* addr, int32_t before, int32_t after);
+static bool sk_atomic_cas(int64_t* addr, int64_t before, int64_t after);
 
 /** If sk_atomic_dec does not act as an acquire (L/SL) barrier,
  *  this must act as an acquire (L/SL) memory barrier and as a compiler barrier.
@@ -49,8 +53,8 @@ static void sk_membar_acquire__after_atomic_conditional_inc();
  *  and returns the previous value.
  *  No additional memory barrier is required; this must act as a compiler barrier.
  */
-static inline int32_t sk_atomic_conditional_inc(int32_t* addr) {
-    int32_t prev;
+template<typename INT_TYPE> static inline INT_TYPE sk_atomic_conditional_inc(INT_TYPE* addr) {
+    INT_TYPE prev;
     do {
         prev = *addr;
         if (0 == prev) {
diff --git a/src/ports/SkAtomics_sync.h b/src/ports/SkAtomics_sync.h
index b0d17527f..8135ae2e2 100644
--- a/src/ports/SkAtomics_sync.h
+++ b/src/ports/SkAtomics_sync.h
@@ -16,14 +16,26 @@ static inline __attribute__((always_inline)) int32_t sk_atomic_inc(int32_t* addr
     return __sync_fetch_and_add(addr, 1);
 }
 
+static inline __attribute__((always_inline)) int64_t sk_atomic_inc(int64_t* addr) {
+    return __sync_fetch_and_add(addr, 1);
+}
+
 static inline __attribute__((always_inline)) int32_t sk_atomic_add(int32_t* addr, int32_t inc) {
     return __sync_fetch_and_add(addr, inc);
 }
 
+static inline __attribute__((always_inline)) int64_t sk_atomic_add(int64_t* addr, int64_t inc) {
+    return __sync_fetch_and_add(addr, inc);
+}
+
 static inline __attribute__((always_inline)) int32_t sk_atomic_dec(int32_t* addr) {
     return __sync_fetch_and_add(addr, -1);
 }
 
+static inline __attribute__((always_inline)) int64_t sk_atomic_dec(int64_t* addr) {
+    return __sync_fetch_and_add(addr, -1);
+}
+
 static inline __attribute__((always_inline)) void sk_membar_acquire__after_atomic_dec() { }
 
 static inline __attribute__((always_inline)) bool sk_atomic_cas(int32_t* addr,
@@ -32,6 +44,12 @@ static inline __attribute__((always_inline)) bool sk_atomic_cas(int32_t* addr,
     return __sync_bool_compare_and_swap(addr, before, after);
 }
 
+static inline __attribute__((always_inline)) bool sk_atomic_cas(int64_t* addr,
+                                                                int64_t before,
+                                                                int64_t after) {
+    return __sync_bool_compare_and_swap(addr, before, after);
+}
+
 static inline __attribute__((always_inline)) void* sk_atomic_cas(void** addr,
                                                                  void* before,
                                                                  void* after) {
diff --git a/src/ports/SkAtomics_win.h b/src/ports/SkAtomics_win.h
index 16923947e..4b344dde7 100644
--- a/src/ports/SkAtomics_win.h
+++ b/src/ports/SkAtomics_win.h
@@ -25,21 +25,39 @@ static inline int32_t sk_atomic_inc(int32_t* addr) {
     return _InterlockedIncrement(reinterpret_cast<long*>(addr)) - 1;
 }
 
+static inline int64_t sk_atomic_inc(int64_t* addr) {
+    // InterlockedIncrement returns the new value, we want to return the old.
+    return InterlockedIncrement64(addr) - 1;
+}
+
 static inline int32_t sk_atomic_add(int32_t* addr, int32_t inc) {
     return _InterlockedExchangeAdd(reinterpret_cast<long*>(addr), static_cast<long>(inc));
 }
 
+static inline int64_t sk_atomic_add(int64_t* addr, int64_t inc) {
+    return InterlockedExchangeAdd64(addr, inc);
+}
+
 static inline int32_t sk_atomic_dec(int32_t* addr) {
     // InterlockedDecrement returns the new value, we want to return the old.
     return _InterlockedDecrement(reinterpret_cast<long*>(addr)) + 1;
 }
 
+static inline int64_t sk_atomic_dec(int64_t* addr) {
+    // InterlockedDecrement returns the new value, we want to return the old.
+    return InterlockedDecrement64(addr) + 1;
+}
+
 static inline void sk_membar_acquire__after_atomic_dec() { }
 
 static inline bool sk_atomic_cas(int32_t* addr, int32_t before, int32_t after) {
     return _InterlockedCompareExchange(reinterpret_cast<long*>(addr), after, before) == before;
 }
 
+static inline bool sk_atomic_cas(int64_t* addr, int64_t before, int64_t after) {
+    return _InterlockedCompareExchange64(addr, after, before) == before;
+}
+
 static inline void* sk_atomic_cas(void** addr, void* before, void* after) {
     return InterlockedCompareExchangePointer(addr, after, before);
 }