[heap] Concurrency support for heap book-keeping info

src/heap/spaces.h

Index: src/heap/spaces.h
diff --git a/src/heap/spaces.h b/src/heap/spaces.h
index b11f926fde3cf373b1144b0dc67ebc5f859c0a02..943f04d182bea57b839598bee5bd583a8411e8ff 100644
--- a/src/heap/spaces.h
+++ b/src/heap/spaces.h
@@ -6,6 +6,7 @@
 #define V8_HEAP_SPACES_H_
 
 #include "src/allocation.h"
+#include "src/atomic-utils.h"
 #include "src/base/atomicops.h"
 #include "src/base/bits.h"
 #include "src/base/platform/mutex.h"
@@ -560,14 +561,14 @@ class MemoryChunk {
 
   static const size_t kMinHeaderSize =
       kWriteBarrierCounterOffset +
-      kIntptrSize      // intptr_t write_barrier_counter_
-      + kIntSize       // int progress_bar_
-      + kIntSize       // int high_water_mark_
-      + kPointerSize   // base::Mutex* mutex_
-      + kPointerSize   // base::AtomicWord parallel_sweeping_
-      + 5 * kIntSize   // int free-list statistics
-      + kPointerSize   // base::AtomicWord next_chunk_
-      + kPointerSize;  // base::AtomicWord prev_chunk_
+      kIntptrSize         // intptr_t write_barrier_counter_
+      + kIntSize          // int progress_bar_
+      + kPointerSize      // AtomicValue high_water_mark_
+      + kPointerSize      // base::Mutex* mutex_
+      + kPointerSize      // base::AtomicWord parallel_sweeping_
+      + 5 * kPointerSize  // AtomicNumber free-list statistics
+      + kPointerSize      // base::AtomicWord next_chunk_
+      + kPointerSize;     // base::AtomicWord prev_chunk_
 
   // We add some more space to the computed header size to amount for missing
   // alignment requirements in our computation.
@@ -674,21 +675,23 @@ class MemoryChunk {
   bool CommitArea(size_t requested);
 
   // Approximate amount of physical memory committed for this chunk.
-  size_t CommittedPhysicalMemory() { return high_water_mark_; }
+  size_t CommittedPhysicalMemory() { return high_water_mark_.Value(); }
 
   // Should be called when memory chunk is about to be freed.
   void ReleaseAllocatedMemory();
 
   static inline void UpdateHighWaterMark(Address mark) {
-    if (mark == NULL) return;
+    if (mark == nullptr) return;
     // Need to subtract one from the mark because when a chunk is full the
     // top points to the next address after the chunk, which effectively belongs
     // to another chunk. See the comment to Page::FromAllocationTop.
     MemoryChunk* chunk = MemoryChunk::FromAddress(mark - 1);
-    int new_mark = static_cast<int>(mark - chunk->address());
-    if (new_mark > chunk->high_water_mark_) {
-      chunk->high_water_mark_ = new_mark;
-    }
+    intptr_t new_mark = static_cast<intptr_t>(mark - chunk->address());
+    intptr_t old_mark = 0;
+    do {
+      old_mark = chunk->high_water_mark_.Value();
+    } while ((new_mark > old_mark) &&
+             !chunk->high_water_mark_.TrySetValue(old_mark, new_mark));
   }
 
  protected:
@@ -719,17 +722,17 @@ class MemoryChunk {
   int progress_bar_;
   // Assuming the initial allocation on a page is sequential,
   // count highest number of bytes ever allocated on the page.
-  int high_water_mark_;
+  AtomicValue<intptr_t> high_water_mark_;
 
   base::Mutex* mutex_;
   base::AtomicWord parallel_sweeping_;
 
   // PagedSpace free-list statistics.
-  int available_in_small_free_list_;
-  int available_in_medium_free_list_;
-  int available_in_large_free_list_;
-  int available_in_huge_free_list_;
-  int non_available_small_blocks_;
+  AtomicNumber<intptr_t> available_in_small_free_list_;
+  AtomicNumber<intptr_t> available_in_medium_free_list_;
+  AtomicNumber<intptr_t> available_in_large_free_list_;
+  AtomicNumber<intptr_t> available_in_huge_free_list_;
+  AtomicNumber<intptr_t> non_available_small_blocks_;
 
   // next_chunk_ holds a pointer of type MemoryChunk
   base::AtomicWord next_chunk_;
@@ -828,21 +831,22 @@ class Page : public MemoryChunk {
   void ResetFreeListStatistics();
 
   int LiveBytesFromFreeList() {
-    return area_size() - non_available_small_blocks_ -
-           available_in_small_free_list_ - available_in_medium_free_list_ -
-           available_in_large_free_list_ - available_in_huge_free_list_;
+    return static_cast<int>(
+        area_size() - non_available_small_blocks() -
+        available_in_small_free_list() - available_in_medium_free_list() -
+        available_in_large_free_list() - available_in_huge_free_list());
   }
 
-#define FRAGMENTATION_STATS_ACCESSORS(type, name) \
-  type name() { return name##_; }                 \
-  void set_##name(type name) { name##_ = name; }  \
-  void add_##name(type name) { name##_ += name; }
+#define FRAGMENTATION_STATS_ACCESSORS(type, name)        \
+  type name() { return name##_.Value(); }                \
+  void set_##name(type name) { name##_.SetValue(name); } \
+  void add_##name(type name) { name##_.Increment(name); }
 
-  FRAGMENTATION_STATS_ACCESSORS(int, non_available_small_blocks)
-  FRAGMENTATION_STATS_ACCESSORS(int, available_in_small_free_list)
-  FRAGMENTATION_STATS_ACCESSORS(int, available_in_medium_free_list)
-  FRAGMENTATION_STATS_ACCESSORS(int, available_in_large_free_list)
-  FRAGMENTATION_STATS_ACCESSORS(int, available_in_huge_free_list)
+  FRAGMENTATION_STATS_ACCESSORS(intptr_t, non_available_small_blocks)
+  FRAGMENTATION_STATS_ACCESSORS(intptr_t, available_in_small_free_list)
+  FRAGMENTATION_STATS_ACCESSORS(intptr_t, available_in_medium_free_list)
+  FRAGMENTATION_STATS_ACCESSORS(intptr_t, available_in_large_free_list)
+  FRAGMENTATION_STATS_ACCESSORS(intptr_t, available_in_huge_free_list)
 
 #undef FRAGMENTATION_STATS_ACCESSORS
 
@@ -1150,9 +1154,9 @@ class MemoryAllocator {
 
   // Returns an indication of whether a pointer is in a space that has
   // been allocated by this MemoryAllocator.
-  V8_INLINE bool IsOutsideAllocatedSpace(const void* address) const {
-    return address < lowest_ever_allocated_ ||
-           address >= highest_ever_allocated_;
+  V8_INLINE bool IsOutsideAllocatedSpace(const void* address) {
+    return address < lowest_ever_allocated_.Value() ||
+           address >= highest_ever_allocated_.Value();
   }
 
 #ifdef DEBUG
@@ -1246,8 +1250,8 @@ class MemoryAllocator {
   // conservative, i.e. not all addrsses in 'allocated' space are allocated
   // to our heap. The range is [lowest, highest[, inclusive on the low end
   // and exclusive on the high end.
-  void* lowest_ever_allocated_;
-  void* highest_ever_allocated_;
+  AtomicValue<void*> lowest_ever_allocated_;
+  AtomicValue<void*> highest_ever_allocated_;
 
   struct MemoryAllocationCallbackRegistration {
     MemoryAllocationCallbackRegistration(MemoryAllocationCallback callback,
@@ -1270,8 +1274,13 @@ class MemoryAllocator {
                                PagedSpace* owner);
 
   void UpdateAllocatedSpaceLimits(void* low, void* high) {
-    lowest_ever_allocated_ = Min(lowest_ever_allocated_, low);
-    highest_ever_allocated_ = Max(highest_ever_allocated_, high);
+    void* ptr = nullptr;
+    do {
+      ptr = lowest_ever_allocated_.Value();

[Hannes Payer (out of office), 2015/09/15 09:55:36]

Can this be a function on AtomicValue?

[Michael Lippautz, 2015/09/15 11:54:53]

As discussed offline: While lamdas are supported, std::function is not. Doing
the dance between lambda and function ptr seems like an overkill for this one
case. Furthermore I would like to avoid having SetValueIfLess() on AtomicValue
as it's very specific to this use here.

I added a comment on what the code does though, as this is in principal
concurrency code that should be kept isolate.

+    } while ((low < ptr) && !lowest_ever_allocated_.TrySetValue(ptr, low));
+    do {
+      ptr = highest_ever_allocated_.Value();
+    } while ((high > ptr) && !highest_ever_allocated_.TrySetValue(ptr, high));
   }
 
   DISALLOW_IMPLICIT_CONSTRUCTORS(MemoryAllocator);