Adding freelist statistics to blink gc dump provider.

Source/platform/heap/Heap.h

Index: Source/platform/heap/Heap.h
diff --git a/Source/platform/heap/Heap.h b/Source/platform/heap/Heap.h
index 10a873988cf15c42ed1635cae3b7560d05d27621..e075683c5afbc41c2b66cf64ceb5e8cf5c97e859 100644
--- a/Source/platform/heap/Heap.h
+++ b/Source/platform/heap/Heap.h
@@ -656,6 +656,7 @@ public:
     // All FreeListEntries in the given bucket, n, have size >= 2^n.
     static int bucketIndexForSize(size_t);
 
+    void takeSnapshot(const String& dumpBaseName);
 #if ENABLE(GC_PROFILING)
     struct PerBucketFreeListStats {
         size_t entryCount;
@@ -693,6 +694,7 @@ public:
 #if ENABLE(ASSERT) || ENABLE(GC_PROFILING)
     BasePage* findPageFromAddress(Address);
 #endif
+    virtual void takeFreelistSnapshot(String dumpBaseName) { };
 #if ENABLE(GC_PROFILING)
     void snapshot(TracedValue*, ThreadState::SnapshotInfo*);
     virtual void snapshotFreeList(TracedValue&) { };
@@ -752,6 +754,7 @@ public:
     virtual bool isConsistentForGC() override;
     bool pagesToBeSweptContains(Address);
 #endif
+    void takeFreelistSnapshot(String dumpBaseName) override;
 #if ENABLE(GC_PROFILING)
     void snapshotFreeList(TracedValue&) override;
 #endif
@@ -785,11 +788,9 @@ private:
     // The size of promptly freed objects in the heap.
     size_t m_promptlyFreedSize;
 
-#if ENABLE(GC_PROFILING)
     size_t m_cumulativeAllocationSize;
     size_t m_allocationCount;
     size_t m_inlineAllocationCount;
-#endif
 };
 
 class LargeObjectHeap final : public BaseHeap {
@@ -1310,15 +1311,11 @@ void HeapObjectHeader::markDead()
 
 inline Address NormalPageHeap::allocateObject(size_t allocationSize, size_t gcInfoIndex)
 {
-#if ENABLE(GC_PROFILING)

[haraken, 2015/06/23 05:38:58]

This function is performance-sensitive and we don't want to add extra overhead.
Shall we drop this change at least from this change? i.e., we want to keep these
variables only in GC_PROFILING.

[ssid, 2015/06/23 06:03:39]

This means that these values will never be added to trace. Should I have
GC_PROFILING macro in the dumper also, so that these will be added when the flag
is enabled? primiano@ WDYT?

[haraken, 2015/06/23 06:24:24]

Yeah, it's a bit unfortunate we can't have the values in the trace, but I'd
suggest revisiting the issue later. We need to study performance to add the
values. IMO, these values are not that important to understand Oilpan's
workload.

GC_PROFILING builds are super heavy, so it's not an option to enable it on trunk
:-/

[ssid, 2015/06/23 07:39:56]

Okay, removing it.

     m_cumulativeAllocationSize += allocationSize;
     ++m_allocationCount;
-#endif
 
     if (LIKELY(allocationSize <= m_remainingAllocationSize)) {
-#if ENABLE(GC_PROFILING)
         ++m_inlineAllocationCount;
-#endif
         Address headerAddress = m_currentAllocationPoint;
         m_currentAllocationPoint += allocationSize;
         m_remainingAllocationSize -= allocationSize;