Merge "Replace memory barriers to better reflect Java needs."
diff --git a/runtime/arch/x86_64/quick_entrypoints_x86_64.S b/runtime/arch/x86_64/quick_entrypoints_x86_64.S
index 7f7226c..885fbfd 100644
--- a/runtime/arch/x86_64/quick_entrypoints_x86_64.S
+++ b/runtime/arch/x86_64/quick_entrypoints_x86_64.S
@@ -18,7 +18,7 @@
MACRO0(SETUP_FP_CALLEE_SAVE_FRAME)
// Create space for ART FP callee-saved registers
- subq LITERAL(4 * 8), %rsp
+ subq MACRO_LITERAL(4 * 8), %rsp
CFI_ADJUST_CFA_OFFSET(4 * 8)
movq %xmm12, 0(%rsp)
movq %xmm13, 8(%rsp)
@@ -32,7 +32,7 @@
movq 8(%rsp), %xmm13
movq 16(%rsp), %xmm14
movq 24(%rsp), %xmm15
- addq LITERAL(4 * 8), %rsp
+ addq MACRO_LITERAL(4 * 8), %rsp
CFI_ADJUST_CFA_OFFSET(- 4 * 8)
END_MACRO
diff --git a/runtime/gc/heap-inl.h b/runtime/gc/heap-inl.h
index 58ba61b..419af30 100644
--- a/runtime/gc/heap-inl.h
+++ b/runtime/gc/heap-inl.h
@@ -41,11 +41,11 @@
const PreFenceVisitor& pre_fence_visitor) {
if (kIsDebugBuild) {
CheckPreconditionsForAllocObject(klass, byte_count);
+ // Since allocation can cause a GC which will need to SuspendAll, make sure all allocations are
+ // done in the runnable state where suspension is expected.
+ CHECK_EQ(self->GetState(), kRunnable);
+ self->AssertThreadSuspensionIsAllowable();
}
- // Since allocation can cause a GC which will need to SuspendAll, make sure all allocations are
- // done in the runnable state where suspension is expected.
- DCHECK_EQ(self->GetState(), kRunnable);
- self->AssertThreadSuspensionIsAllowable();
// Need to check that we arent the large object allocator since the large object allocation code
// path this function. If we didn't check we would have an infinite loop.
if (kCheckLargeObject && UNLIKELY(ShouldAllocLargeObject(klass, byte_count))) {
@@ -54,49 +54,70 @@
}
mirror::Object* obj;
AllocationTimer alloc_timer(this, &obj);
- size_t bytes_allocated, usable_size;
- obj = TryToAllocate<kInstrumented, false>(self, allocator, byte_count, &bytes_allocated,
- &usable_size);
- if (UNLIKELY(obj == nullptr)) {
- bool is_current_allocator = allocator == GetCurrentAllocator();
- obj = AllocateInternalWithGc(self, allocator, byte_count, &bytes_allocated, &usable_size,
- &klass);
- if (obj == nullptr) {
- bool after_is_current_allocator = allocator == GetCurrentAllocator();
- if (is_current_allocator && !after_is_current_allocator) {
- // If the allocator changed, we need to restart the allocation.
- return AllocObject<kInstrumented>(self, klass, byte_count, pre_fence_visitor);
+ size_t bytes_allocated;
+ size_t usable_size;
+ size_t new_num_bytes_allocated = 0;
+ if (allocator == kAllocatorTypeTLAB) {
+ byte_count = RoundUp(byte_count, space::BumpPointerSpace::kAlignment);
+ }
+ // If we have a thread local allocation we don't need to update bytes allocated.
+ if (allocator == kAllocatorTypeTLAB && byte_count <= self->TlabSize()) {
+ obj = self->AllocTlab(byte_count);
+ obj->SetClass(klass);
+ if (kUseBakerOrBrooksReadBarrier) {
+ if (kUseBrooksReadBarrier) {
+ obj->SetReadBarrierPointer(obj);
}
- return nullptr;
+ obj->AssertReadBarrierPointer();
}
- }
- DCHECK_GT(bytes_allocated, 0u);
- DCHECK_GT(usable_size, 0u);
- obj->SetClass(klass);
- if (kUseBakerOrBrooksReadBarrier) {
- if (kUseBrooksReadBarrier) {
- obj->SetReadBarrierPointer(obj);
+ bytes_allocated = byte_count;
+ pre_fence_visitor(obj, bytes_allocated);
+ QuasiAtomic::ThreadFenceForConstructor();
+ } else {
+ obj = TryToAllocate<kInstrumented, false>(self, allocator, byte_count, &bytes_allocated,
+ &usable_size);
+ if (UNLIKELY(obj == nullptr)) {
+ bool is_current_allocator = allocator == GetCurrentAllocator();
+ obj = AllocateInternalWithGc(self, allocator, byte_count, &bytes_allocated, &usable_size,
+ &klass);
+ if (obj == nullptr) {
+ bool after_is_current_allocator = allocator == GetCurrentAllocator();
+ if (is_current_allocator && !after_is_current_allocator) {
+ // If the allocator changed, we need to restart the allocation.
+ return AllocObject<kInstrumented>(self, klass, byte_count, pre_fence_visitor);
+ }
+ return nullptr;
+ }
}
- obj->AssertReadBarrierPointer();
+ DCHECK_GT(bytes_allocated, 0u);
+ DCHECK_GT(usable_size, 0u);
+ obj->SetClass(klass);
+ if (kUseBakerOrBrooksReadBarrier) {
+ if (kUseBrooksReadBarrier) {
+ obj->SetReadBarrierPointer(obj);
+ }
+ obj->AssertReadBarrierPointer();
+ }
+ if (collector::SemiSpace::kUseRememberedSet && UNLIKELY(allocator == kAllocatorTypeNonMoving)) {
+ // (Note this if statement will be constant folded away for the
+ // fast-path quick entry points.) Because SetClass() has no write
+ // barrier, if a non-moving space allocation, we need a write
+ // barrier as the class pointer may point to the bump pointer
+ // space (where the class pointer is an "old-to-young" reference,
+ // though rare) under the GSS collector with the remembered set
+ // enabled. We don't need this for kAllocatorTypeRosAlloc/DlMalloc
+ // cases because we don't directly allocate into the main alloc
+ // space (besides promotions) under the SS/GSS collector.
+ WriteBarrierField(obj, mirror::Object::ClassOffset(), klass);
+ }
+ pre_fence_visitor(obj, usable_size);
+ if (kIsDebugBuild && Runtime::Current()->IsStarted()) {
+ CHECK_LE(obj->SizeOf(), usable_size);
+ }
+ new_num_bytes_allocated =
+ static_cast<size_t>(num_bytes_allocated_.FetchAndAddSequentiallyConsistent(bytes_allocated))
+ + bytes_allocated;
}
- if (collector::SemiSpace::kUseRememberedSet && UNLIKELY(allocator == kAllocatorTypeNonMoving)) {
- // (Note this if statement will be constant folded away for the
- // fast-path quick entry points.) Because SetClass() has no write
- // barrier, if a non-moving space allocation, we need a write
- // barrier as the class pointer may point to the bump pointer
- // space (where the class pointer is an "old-to-young" reference,
- // though rare) under the GSS collector with the remembered set
- // enabled. We don't need this for kAllocatorTypeRosAlloc/DlMalloc
- // cases because we don't directly allocate into the main alloc
- // space (besides promotions) under the SS/GSS collector.
- WriteBarrierField(obj, mirror::Object::ClassOffset(), klass);
- }
- pre_fence_visitor(obj, usable_size);
- if (kIsDebugBuild && Runtime::Current()->IsStarted()) {
- CHECK_LE(obj->SizeOf(), usable_size);
- }
- const size_t new_num_bytes_allocated =
- static_cast<size_t>(num_bytes_allocated_.FetchAndAddSequentiallyConsistent(bytes_allocated)) + bytes_allocated;
// TODO: Deprecate.
if (kInstrumented) {
if (Runtime::Current()->HasStatsEnabled()) {
@@ -158,7 +179,8 @@
inline mirror::Object* Heap::TryToAllocate(Thread* self, AllocatorType allocator_type,
size_t alloc_size, size_t* bytes_allocated,
size_t* usable_size) {
- if (UNLIKELY(IsOutOfMemoryOnAllocation<kGrow>(allocator_type, alloc_size))) {
+ if (allocator_type != kAllocatorTypeTLAB &&
+ UNLIKELY(IsOutOfMemoryOnAllocation<kGrow>(allocator_type, alloc_size))) {
return nullptr;
}
mirror::Object* ret;
@@ -206,18 +228,24 @@
break;
}
case kAllocatorTypeTLAB: {
- alloc_size = RoundUp(alloc_size, space::BumpPointerSpace::kAlignment);
+ DCHECK_ALIGNED(alloc_size, space::BumpPointerSpace::kAlignment);
if (UNLIKELY(self->TlabSize() < alloc_size)) {
- // Try allocating a new thread local buffer, if the allocaiton fails the space must be
- // full so return nullptr.
- if (!bump_pointer_space_->AllocNewTlab(self, alloc_size + kDefaultTLABSize)) {
+ const size_t new_tlab_size = alloc_size + kDefaultTLABSize;
+ if (UNLIKELY(IsOutOfMemoryOnAllocation<kGrow>(allocator_type, new_tlab_size))) {
return nullptr;
}
+ // Try allocating a new thread local buffer, if the allocaiton fails the space must be
+ // full so return nullptr.
+ if (!bump_pointer_space_->AllocNewTlab(self, new_tlab_size)) {
+ return nullptr;
+ }
+ *bytes_allocated = new_tlab_size;
+ } else {
+ *bytes_allocated = 0;
}
// The allocation can't fail.
ret = self->AllocTlab(alloc_size);
DCHECK(ret != nullptr);
- *bytes_allocated = alloc_size;
*usable_size = alloc_size;
break;
}
diff --git a/runtime/thread.h b/runtime/thread.h
index 4312741..1b335c8 100644
--- a/runtime/thread.h
+++ b/runtime/thread.h
@@ -781,7 +781,7 @@
void RevokeThreadLocalAllocationStack();
size_t GetThreadLocalBytesAllocated() const {
- return tlsPtr_.thread_local_pos - tlsPtr_.thread_local_start;
+ return tlsPtr_.thread_local_end - tlsPtr_.thread_local_start;
}
size_t GetThreadLocalObjectsAllocated() const {