Reorder and align LRUCacheShard data members to reduce false sharing

cache/lru_cache.cc

@@ -24,7 +24,7 @@
 
 namespace rocksdb {
 
-LRUHandleTable::LRUHandleTable() : length_(0), elems_(0), list_(nullptr) {
+LRUHandleTable::LRUHandleTable() : list_(nullptr), length_(0), elems_(0) {
   Resize();
 }
 
@@ -102,7 +102,7 @@ void LRUHandleTable::Resize() {
 }
 
 LRUCacheShard::LRUCacheShard()
-    : usage_(0), lru_usage_(0), high_pri_pool_usage_(0) {
+    : high_pri_pool_usage_(0), usage_(0), lru_usage_(0) {
   // Make empty circular linked list
   lru_.next = &lru_;
   lru_.prev = &lru_;

cache/lru_cache.h

@@ -150,13 +150,13 @@ class LRUHandleTable {
 
   // The table consists of an array of buckets where each bucket is
   // a linked list of cache entries that hash into the bucket.
+  LRUHandle** list_;
   uint32_t length_;
   uint32_t elems_;
-  LRUHandle** list_;
 };
 
 // A single shard of sharded cache.
-class LRUCacheShard : public CacheShard {
+class ALIGN_AS(CACHE_LINE_SIZE) LRUCacheShard : public CacheShard {
  public:
   LRUCacheShard();
   virtual ~LRUCacheShard();
@@ -221,12 +221,6 @@ class LRUCacheShard : public CacheShard {
   // Initialized before use.
   size_t capacity_;
 
-  // Memory size for entries residing in the cache
-  size_t usage_;
-
-  // Memory size for entries residing only in the LRU list
-  size_t lru_usage_;
-
   // Memory size for entries in high-pri pool.
   size_t high_pri_pool_usage_;
 
@@ -240,11 +234,6 @@ class LRUCacheShard : public CacheShard {
   // Remember the value to avoid recomputing each time.
   double high_pri_pool_capacity_;
 
-  // mutex_ protects the following state.
-  // We don't count mutex_ as the cache's internal state so semantically we
-  // don't mind mutex_ invoking the non-const actions.
-  mutable port::Mutex mutex_;
-
   // Dummy head of LRU list.
   // lru.prev is newest entry, lru.next is oldest entry.
   // LRU contains items which can be evicted, ie reference only by cache
@@ -253,7 +242,29 @@ class LRUCacheShard : public CacheShard {
   // Pointer to head of low-pri pool in LRU list.
   LRUHandle* lru_low_pri_;
 
+  // ------------^^^^^^^^^^^^^-----------
+  // Not frequently modified data members
+  // ------------------------------------
+  //
+  // We separate data members that are updated frequently from the ones that
+  // are not frequently updated so that they don't share the same cache line
+  // which will lead into false cache sharing
+  //
+  // ------------------------------------
+  // Frequently modified data members
+  // ------------vvvvvvvvvvvvv-----------
   LRUHandleTable table_;
+
+  // Memory size for entries residing in the cache
+  size_t usage_;
+
+  // Memory size for entries residing only in the LRU list
+  size_t lru_usage_;
+
+  // mutex_ protects the following state.
+  // We don't count mutex_ as the cache's internal state so semantically we
+  // don't mind mutex_ invoking the non-const actions.
+  mutable port::Mutex mutex_;
 };
 
 class LRUCache : public ShardedCache {

port/port_posix.h

@@ -189,6 +189,8 @@ extern void InitOnce(OnceType* once, void (*initializer)());
 #define CACHE_LINE_SIZE 64U
 #endif
 
+#define ALIGN_AS(n) alignas(n)
+
 #define PREFETCH(addr, rw, locality) __builtin_prefetch(addr, rw, locality)
 
 extern void Crash(const std::string& srcfile, int srcline);

port/win/port_win.h

@@ -241,6 +241,8 @@ extern void InitOnce(OnceType* once, void (*initializer)());
 #define CACHE_LINE_SIZE 64U
 #endif
 
+#define ALIGN_AS(n)
+
 static inline void AsmVolatilePause() {
 #if defined(_M_IX86) || defined(_M_X64)
   YieldProcessor();