src/hotspot/share/gc/shared/threadLocalAllocBuffer.hpp

@@ -35,21 +35,33 @@
 // ThreadLocalAllocBuffer: a descriptor for thread-local storage used by
 // the threads for allocation.
 //            It is thread-private at any time, but maybe multiplexed over
 //            time across multiple threads. The park()/unpark() pair is
 //            used to make it available for such multiplexing.
+//
+//            Heap sampling is performed via the end/actual_end fields.
+//            actual_end contains the real end of the tlab allocation,
+//            whereas end can be set to an arbitrary spot in the tlab to
+//            trip the return and sample the allocation.
+//            slow_path_end is used to track if a fast tlab refill occured
+//            between slowpath calls.
 class ThreadLocalAllocBuffer: public CHeapObj<mtThread> {
   friend class VMStructs;
   friend class JVMCIVMStructs;
 private:
   HeapWord* _start;                              // address of TLAB
   HeapWord* _top;                                // address after last allocation
   HeapWord* _pf_top;                             // allocation prefetch watermark
-  HeapWord* _end;                                // allocation end (excluding alignment_reserve)
+  HeapWord* _end;                                // allocation end (can be the sampling end point or
+                                                 //                 the actual TLAB end, excluding alignment_reserve)
+  HeapWord* _actual_end;                         // allocation actual_end (actual TLAB end, excluding alignment_reserve)
+  HeapWord* _slow_path_end;                      // remember the end in case a fast refill occurs.
+
   size_t    _desired_size;                       // desired size   (including alignment_reserve)
   size_t    _refill_waste_limit;                 // hold onto tlab if free() is larger than this
   size_t    _allocated_before_last_gc;           // total bytes allocated up until the last gc
+  size_t    _bytes_until_sample;                 // bytes until sample.
 
   static size_t   _max_size;                          // maximum size of any TLAB
   static int      _reserve_for_allocation_prefetch;   // Reserve at the end of the TLAB
   static unsigned _target_refills;                    // expected number of refills between GCs
 

@@ -64,21 +76,27 @@
   void accumulate_statistics();
   void initialize_statistics();
 
   void set_start(HeapWord* start)                { _start = start; }
   void set_end(HeapWord* end)                    { _end = end; }
+  void set_actual_end(HeapWord* actual_end)      { _actual_end = actual_end; }
+  void set_slow_path_end(HeapWord* slow_path_end)    { _slow_path_end = slow_path_end; }
   void set_top(HeapWord* top)                    { _top = top; }
   void set_pf_top(HeapWord* pf_top)              { _pf_top = pf_top; }
   void set_desired_size(size_t desired_size)     { _desired_size = desired_size; }
   void set_refill_waste_limit(size_t waste)      { _refill_waste_limit = waste;  }
+  void set_bytes_until_sample(size_t bytes)      { _bytes_until_sample = bytes;  }
 
   size_t initial_refill_waste_limit()            { return desired_size() / TLABRefillWasteFraction; }
 
   static int    target_refills()                 { return _target_refills; }
   size_t initial_desired_size();
 
-  size_t remaining() const                       { return end() == NULL ? 0 : pointer_delta(hard_end(), top()); }
+  size_t remaining()                             { return end() == NULL ? 0 : pointer_delta(hard_end(), top()); }
+
+  // Obtain the actual end of the TLAB.
+  void set_sample_end();
 
   // Make parsable and release it.
   void reset();
 
   // Resize based on amount of allocation, etc.

@@ -111,13 +129,13 @@
   static size_t min_size()                       { return align_object_size(MinTLABSize / HeapWordSize) + alignment_reserve(); }
   static size_t max_size()                       { assert(_max_size != 0, "max_size not set up"); return _max_size; }
   static size_t max_size_in_bytes()              { return max_size() * BytesPerWord; }
   static void set_max_size(size_t max_size)      { _max_size = max_size; }
 
+  HeapWord* hard_end();
   HeapWord* start() const                        { return _start; }
   HeapWord* end() const                          { return _end; }
-  HeapWord* hard_end() const                     { return _end + alignment_reserve(); }
   HeapWord* top() const                          { return _top; }
   HeapWord* pf_top() const                       { return _pf_top; }
   size_t desired_size() const                    { return _desired_size; }
   size_t used() const                            { return pointer_delta(top(), start()); }
   size_t used_bytes() const                      { return pointer_delta(top(), start(), 1); }

@@ -160,10 +178,15 @@
   static void resize_all_tlabs();
 
   void fill(HeapWord* start, HeapWord* top, size_t new_size);
   void initialize();
 
+  void pick_next_sample(size_t diff = 0);
+  void set_back_actual_end();
+  void handle_sample(Thread* thread, HeapWord* result, size_t size);
+  bool should_sample() { return _bytes_until_sample == 0; }
+
   static size_t refill_waste_limit_increment()   { return TLABWasteIncrement; }
 
   // Code generation support
   static ByteSize start_offset()                 { return byte_offset_of(ThreadLocalAllocBuffer, _start); }
   static ByteSize end_offset()                   { return byte_offset_of(ThreadLocalAllocBuffer, _end  ); }