Use intrinsics for RWSpinLock when on MSVC.

[folly.git] / folly / RWSpinLock.h

diff --git a/folly/RWSpinLock.h b/folly/RWSpinLock.h
index 8077720775c7f9528579402230eafc1db20c8962..359343f80c1536d6c310d8f21cb6a10fc246aef4 100644 (file)
--- a/folly/RWSpinLock.h
+++ b/folly/RWSpinLock.h
@@ -1,5 +1,5 @@
 /*
- * Copyright 2013 Facebook, Inc.
+ * Copyright 2015 Facebook, Inc.
  *
  * Licensed under the Apache License, Version 2.0 (the "License");
  * you may not use this file except in compliance with the License.
@@ -46,6 +46,18 @@
  *    RWTicketSpinLock<64> only allows up to 2^16 - 1 concurrent
  *    readers and writers.
  *
+ *    RWTicketSpinLock<..., true> (kFavorWriter = true, that is, strict
+ *    writer priority) is NOT reentrant, even for lock_shared().
+ *
+ *    The lock will not grant any new shared (read) accesses while a thread
+ *    attempting to acquire the lock in write mode is blocked. (That is,
+ *    if the lock is held in shared mode by N threads, and a thread attempts
+ *    to acquire it in write mode, no one else can acquire it in shared mode
+ *    until these N threads release the lock and then the blocked thread
+ *    acquires and releases the exclusive lock.) This also applies for
+ *    attempts to reacquire the lock in shared mode by threads that already
+ *    hold it in shared mode, making the lock non-reentrant.
+ *
  *    RWSpinLock handles 2^30 - 1 concurrent readers.
  *
  * @author Xin Liu <xliux@fb.com>
@@ -106,13 +118,24 @@ pthread_rwlock_t Read        728698     24us       101ns     7.28ms     194us
 
 */
 
-#if defined(__GNUC__) && !defined(__clang__) && \
-  (defined(__i386) || defined(__x86_64__) || \
+#include <folly/Portability.h>
+
+#if defined(__GNUC__) && \
+  (defined(__i386) || FOLLY_X64 || \
    defined(ARCH_K8))
-#define RW_SPINLOCK_USE_X86_INTRINSIC_
-#include <x86intrin.h>
+# define RW_SPINLOCK_USE_X86_INTRINSIC_
+# include <x86intrin.h>
+#elif defined(_MSC_VER) && defined(FOLLY_X64)
+# define RW_SPINLOCK_USE_X86_INTRINSIC_
 #else
-#undef RW_SPINLOCK_USE_X86_INTRINSIC_
+# undef RW_SPINLOCK_USE_X86_INTRINSIC_
+#endif
+
+// iOS doesn't define _mm_cvtsi64_si128 and friends
+#if (FOLLY_SSE >= 2) && !TARGET_OS_IPHONE
+#define RW_SPINLOCK_USE_SSE_INSTRUCTIONS_
+#else
+#undef RW_SPINLOCK_USE_SSE_INSTRUCTIONS_
 #endif
 
 #include <atomic>
@@ -123,7 +146,7 @@ pthread_rwlock_t Read        728698     24us       101ns     7.28ms     194us
 #include <sched.h>
 #include <glog/logging.h>
 
-#include "folly/Likely.h"
+#include <folly/Likely.h>
 
 namespace folly {
 
@@ -272,7 +295,7 @@ class RWSpinLock : boost::noncopyable {
       lock_->lock_shared();
     }
 
-    ReadHolder(ReadHolder&& other) : lock_(other.lock_) {
+    ReadHolder(ReadHolder&& other) noexcept : lock_(other.lock_) {
       other.lock_ = nullptr;
     }
 
@@ -331,7 +354,7 @@ class RWSpinLock : boost::noncopyable {
       if (lock_) lock_->unlock_and_lock_upgrade();
     }
 
-    UpgradedHolder(UpgradedHolder&& other) : lock_(other.lock_) {
+    UpgradedHolder(UpgradedHolder&& other) noexcept : lock_(other.lock_) {
       other.lock_ = nullptr;
     }
 
@@ -381,7 +404,7 @@ class RWSpinLock : boost::noncopyable {
       if (lock_) lock_->unlock_upgrade_and_lock();
     }
 
-    WriteHolder(WriteHolder&& other) : lock_(other.lock_) {
+    WriteHolder(WriteHolder&& other) noexcept : lock_(other.lock_) {
       other.lock_ = nullptr;
     }
 
@@ -440,7 +463,7 @@ struct RWTicketIntTrait<64> {
   typedef uint32_t HalfInt;
   typedef uint16_t QuarterInt;
 
-#ifdef __SSE2__
+#ifdef RW_SPINLOCK_USE_SSE_INSTRUCTIONS_
   static __m128i make128(const uint16_t v[4]) {
     return _mm_set_epi16(0, 0, 0, 0, v[3], v[2], v[1], v[0]);
   }
@@ -462,7 +485,7 @@ struct RWTicketIntTrait<32> {
   typedef uint16_t HalfInt;
   typedef uint8_t QuarterInt;
 
-#ifdef __SSE2__
+#ifdef RW_SPINLOCK_USE_SSE_INSTRUCTIONS_
   static __m128i make128(const uint8_t v[4]) {
     return _mm_set_epi8(0, 0, 0, 0, 0, 0, 0, 0,
         0, 0, 0, 0, v[3], v[2], v[1], v[0]);
@@ -566,7 +589,7 @@ class RWTicketSpinLockT : boost::noncopyable {
     int count = 0;
     QuarterInt val = __sync_fetch_and_add(&ticket.users, 1);
     while (val != load_acquire(&ticket.write)) {
-      asm volatile("pause");
+      asm_volatile_pause();
       if (UNLIKELY(++count > 1000)) sched_yield();
     }
   }
@@ -596,7 +619,7 @@ class RWTicketSpinLockT : boost::noncopyable {
     t.whole = load_acquire(&ticket.whole);
     FullInt old = t.whole;
 
-#ifdef __SSE2__
+#ifdef RW_SPINLOCK_USE_SSE_INSTRUCTIONS_
     // SSE2 can reduce the lock and unlock overhead by 10%
     static const QuarterInt kDeltaBuf[4] = { 1, 1, 0, 0 };   // write/read/user
     static const __m128i kDelta = IntTraitType::make128(kDeltaBuf);
@@ -615,7 +638,7 @@ class RWTicketSpinLockT : boost::noncopyable {
     // need to let threads that already have a shared lock complete
     int count = 0;
     while (!LIKELY(try_lock_shared())) {
-      asm volatile("pause");
+      asm_volatile_pause();
       if (UNLIKELY((++count & 1023) == 0)) sched_yield();
     }
   }
@@ -624,7 +647,7 @@ class RWTicketSpinLockT : boost::noncopyable {
     RWTicket t, old;
     old.whole = t.whole = load_acquire(&ticket.whole);
     old.users = old.read;
-#ifdef  __SSE2__
+#ifdef RW_SPINLOCK_USE_SSE_INSTRUCTIONS_
     // SSE2 may reduce the total lock and unlock overhead by 10%
     static const QuarterInt kDeltaBuf[4] = { 0, 1, 1, 0 };   // write/read/user
     static const __m128i kDelta = IntTraitType::make128(kDeltaBuf);