commit fed3e2fd4919b58f8e4f8cbc317ee101f3b9af49
author Hiroshi Yamauchi <yamauchi@google.com> Tue Oct 20 11:11:56 2015 -0700
committer Hiroshi Yamauchi <yamauchi@google.com> Wed Oct 21 16:02:43 2015 -0700
tree 284c4eeec50336a4424a873d1e934481375b2823
parent 114873103db3d4d6e0da42ca02bad1ea8826443b

Relax some CASes for the CC collector.

That is, removing some unnecessary memory fences.

We can use the relaxed CAS for the mark bitmap and reference field/GC
root updates because only the atomicity of the updated word matters
there.

We can use the release CAS for the read barrier bits in the lock word
because it needs to make sure the reference field updates are visible
when the object changes black from gray (the field update stores won't
be reordered after the CAS.)

The CC collector's Ritz EAAC GC time decreases from 34.7s to
29.1s (-16%) on N5.

Bug: 12687968

Change-Id: If082d5911a25fac695df66263a8f55ce8149b199

runtime/mirror/object.h

diff --git a/runtime/mirror/object.h b/runtime/mirror/object.h
index f75b8ae..022f31d 100644
--- a/runtime/mirror/object.h
+++ b/runtime/mirror/object.h
@@ -135,6 +135,8 @@
       SHARED_REQUIRES(Locks::mutator_lock_);
   bool CasLockWordWeakRelaxed(LockWord old_val, LockWord new_val)
       SHARED_REQUIRES(Locks::mutator_lock_);
+  bool CasLockWordWeakRelease(LockWord old_val, LockWord new_val)
+      SHARED_REQUIRES(Locks::mutator_lock_);
   uint32_t GetLockOwnerThreadId();
 
   mirror::Object* MonitorEnter(Thread* self)
@@ -276,7 +278,6 @@
                                                                    Object* old_value,
                                                                    Object* new_value)
       SHARED_REQUIRES(Locks::mutator_lock_);
-
   template<bool kTransactionActive, bool kCheckTransaction = true,
       VerifyObjectFlags kVerifyFlags = kDefaultVerifyFlags>
   bool CasFieldStrongSequentiallyConsistentObject(MemberOffset field_offset, Object* old_value,
@@ -288,6 +289,18 @@
                                                                      Object* old_value,
                                                                      Object* new_value)
       SHARED_REQUIRES(Locks::mutator_lock_);
+  template<bool kTransactionActive, bool kCheckTransaction = true,
+      VerifyObjectFlags kVerifyFlags = kDefaultVerifyFlags>
+  bool CasFieldWeakRelaxedObjectWithoutWriteBarrier(MemberOffset field_offset,
+                                                    Object* old_value,
+                                                    Object* new_value)
+      SHARED_REQUIRES(Locks::mutator_lock_);
+  template<bool kTransactionActive, bool kCheckTransaction = true,
+      VerifyObjectFlags kVerifyFlags = kDefaultVerifyFlags>
+  bool CasFieldStrongRelaxedObjectWithoutWriteBarrier(MemberOffset field_offset,
+                                                      Object* old_value,
+                                                      Object* new_value)
+      SHARED_REQUIRES(Locks::mutator_lock_);
 
   template<VerifyObjectFlags kVerifyFlags = kDefaultVerifyFlags>
   HeapReference<Object>* GetFieldObjectReferenceAddr(MemberOffset field_offset);
@@ -396,6 +409,12 @@
 
   template<bool kTransactionActive, bool kCheckTransaction = true,
       VerifyObjectFlags kVerifyFlags = kDefaultVerifyFlags>
+  bool CasFieldWeakRelease32(MemberOffset field_offset, int32_t old_value,
+                             int32_t new_value) ALWAYS_INLINE
+      SHARED_REQUIRES(Locks::mutator_lock_);
+
+  template<bool kTransactionActive, bool kCheckTransaction = true,
+      VerifyObjectFlags kVerifyFlags = kDefaultVerifyFlags>
   bool CasFieldStrongSequentiallyConsistent32(MemberOffset field_offset, int32_t old_value,
                                               int32_t new_value) ALWAYS_INLINE
       SHARED_REQUIRES(Locks::mutator_lock_);