Parallel cache - preliminary

include/ports/SkAtomics_atomic.h

 /*
  * Copyright 2015 Google Inc.
  *
  * Use of this source code is governed by a BSD-style license that can be
  * found in the LICENSE file.
  */
 
 #ifndef SkAtomics_atomic_DEFINED
 #define SkAtomics_atomic_DEFINED
 
@@ skipping 14 matching lines @@
     __atomic_store_n(ptr, val, mo);
 }
 
 template <typename T>
 T sk_atomic_fetch_add(T* ptr, T val, sk_memory_order mo) {
     // All values of mo are valid.
     return __atomic_fetch_add(ptr, val, mo);
 }
 
 template <typename T>
+T sk_atomic_fetch_sub(T* ptr, T val, sk_memory_order mo) {
+    // All values of mo are valid.
+    return __atomic_fetch_sub(ptr, val, mo);
+}
+
+template <typename T>
 bool sk_atomic_compare_exchange(T* ptr, T* expected, T desired,
                                 sk_memory_order success,
                                 sk_memory_order failure) {
     // All values of success are valid.
     SkASSERT(failure == sk_memory_order_relaxed ||
              failure == sk_memory_order_seq_cst ||
              failure == sk_memory_order_acquire ||
              failure == sk_memory_order_consume);
     SkASSERT(failure <= success);
     return __atomic_compare_exchange_n(ptr, expected, desired, false/*weak?*/, success, failure);
 }
 
 template <typename T>
 T sk_atomic_exchange(T* ptr, T val, sk_memory_order mo) {
     // All values of mo are valid.
     return __atomic_exchange_n(ptr, val, mo);
 }
 
 #endif//SkAtomics_atomic_DEFINED