Change 64 bit ArtMethod fields to be pointer sized
Changed the 64 bit entrypoint and gc map fields in ArtMethod to be
pointer sized. This saves a large amount of memory on 32 bit systems.
Reduces ArtMethod size by 16 bytes on 32 bit.
Total number of ArtMethod on low memory mako: 169957
Image size: 49203 methods -> 787248 image size reduction.
Zygote space size: 1070 methods -> 17120 size reduction.
App methods: ~120k -> 2 MB savings.
Savings per app on low memory mako: 125K+ per app
(less active apps -> more image methods per app).
Savings depend on how often the shared methods are on dirty pages vs
shared.
TODO in another CL, delete gc map field from ArtMethod since we
should be able to get it from the Oat method header.
Bug: 17643507
Change-Id: Ie9508f05907a9f693882d4d32a564460bf273ee8
(cherry picked from commit e832e64a7e82d7f72aedbd7d798fb929d458ee8f)
diff --git a/compiler/image_writer.cc b/compiler/image_writer.cc
index 6096625..64d2de1 100644
--- a/compiler/image_writer.cc
+++ b/compiler/image_writer.cc
@@ -68,6 +68,7 @@
namespace art {
bool ImageWriter::PrepareImageAddressSpace() {
+ target_ptr_size_ = InstructionSetPointerSize(compiler_driver_.GetInstructionSet());
{
Thread::Current()->TransitionFromSuspendedToRunnable();
PruneNonImageClasses(); // Remove junk
@@ -214,7 +215,14 @@
void ImageWriter::AssignImageOffset(mirror::Object* object) {
DCHECK(object != nullptr);
SetImageOffset(object, image_end_);
- image_end_ += RoundUp(object->SizeOf(), 8); // 64-bit alignment
+ size_t object_size;
+ if (object->IsArtMethod()) {
+ // Methods are sized based on the target pointer size.
+ object_size = mirror::ArtMethod::InstanceSize(target_ptr_size_);
+ } else {
+ object_size = object->SizeOf();
+ }
+ image_end_ += RoundUp(object_size, 8); // 64-bit alignment
DCHECK_LT(image_end_, image_->Size());
}
@@ -754,7 +762,14 @@
size_t offset = image_writer->GetImageOffset(obj);
uint8_t* dst = image_writer->image_->Begin() + offset;
const uint8_t* src = reinterpret_cast<const uint8_t*>(obj);
- size_t n = obj->SizeOf();
+ size_t n;
+ if (obj->IsArtMethod()) {
+ // Size without pointer fields since we don't want to overrun the buffer if target art method
+ // is 32 bits but source is 64 bits.
+ n = mirror::ArtMethod::SizeWithoutPointerFields();
+ } else {
+ n = obj->SizeOf();
+ }
DCHECK_LT(offset + n, image_writer->image_->Size());
memcpy(dst, src, n);
Object* copy = reinterpret_cast<Object*>(dst);
@@ -834,6 +849,10 @@
}
if (orig->IsArtMethod<kVerifyNone>()) {
FixupMethod(orig->AsArtMethod<kVerifyNone>(), down_cast<ArtMethod*>(copy));
+ } else if (orig->IsClass() && orig->AsClass()->IsArtMethodClass()) {
+ // Set the right size for the target.
+ size_t size = mirror::ArtMethod::InstanceSize(target_ptr_size_);
+ down_cast<mirror::Class*>(copy)->SetObjectSizeWithoutChecks(size);
}
}
@@ -892,29 +911,48 @@
void ImageWriter::FixupMethod(ArtMethod* orig, ArtMethod* copy) {
// OatWriter replaces the code_ with an offset value. Here we re-adjust to a pointer relative to
// oat_begin_
+ // For 64 bit targets we need to repack the current runtime pointer sized fields to the right
+ // locations.
+ // Copy all of the fields from the runtime methods to the target methods first since we did a
+ // bytewise copy earlier.
+ copy->SetEntryPointFromPortableCompiledCodePtrSize<kVerifyNone>(
+ orig->GetEntryPointFromPortableCompiledCode(), target_ptr_size_);
+ copy->SetEntryPointFromInterpreterPtrSize<kVerifyNone>(orig->GetEntryPointFromInterpreter(),
+ target_ptr_size_);
+ copy->SetEntryPointFromJniPtrSize<kVerifyNone>(orig->GetEntryPointFromJni(), target_ptr_size_);
+ copy->SetEntryPointFromQuickCompiledCodePtrSize<kVerifyNone>(
+ orig->GetEntryPointFromQuickCompiledCode(), target_ptr_size_);
+ copy->SetNativeGcMapPtrSize<kVerifyNone>(orig->GetNativeGcMap(), target_ptr_size_);
// The resolution method has a special trampoline to call.
Runtime* runtime = Runtime::Current();
if (UNLIKELY(orig == runtime->GetResolutionMethod())) {
- copy->SetEntryPointFromPortableCompiledCode<kVerifyNone>(GetOatAddress(portable_resolution_trampoline_offset_));
- copy->SetEntryPointFromQuickCompiledCode<kVerifyNone>(GetOatAddress(quick_resolution_trampoline_offset_));
+ copy->SetEntryPointFromPortableCompiledCodePtrSize<kVerifyNone>(
+ GetOatAddress(portable_resolution_trampoline_offset_), target_ptr_size_);
+ copy->SetEntryPointFromQuickCompiledCodePtrSize<kVerifyNone>(
+ GetOatAddress(quick_resolution_trampoline_offset_), target_ptr_size_);
} else if (UNLIKELY(orig == runtime->GetImtConflictMethod() ||
orig == runtime->GetImtUnimplementedMethod())) {
- copy->SetEntryPointFromPortableCompiledCode<kVerifyNone>(GetOatAddress(portable_imt_conflict_trampoline_offset_));
- copy->SetEntryPointFromQuickCompiledCode<kVerifyNone>(GetOatAddress(quick_imt_conflict_trampoline_offset_));
+ copy->SetEntryPointFromPortableCompiledCodePtrSize<kVerifyNone>(
+ GetOatAddress(portable_imt_conflict_trampoline_offset_), target_ptr_size_);
+ copy->SetEntryPointFromQuickCompiledCodePtrSize<kVerifyNone>(
+ GetOatAddress(quick_imt_conflict_trampoline_offset_), target_ptr_size_);
} else {
// We assume all methods have code. If they don't currently then we set them to the use the
// resolution trampoline. Abstract methods never have code and so we need to make sure their
// use results in an AbstractMethodError. We use the interpreter to achieve this.
if (UNLIKELY(orig->IsAbstract())) {
- copy->SetEntryPointFromPortableCompiledCode<kVerifyNone>(GetOatAddress(portable_to_interpreter_bridge_offset_));
- copy->SetEntryPointFromQuickCompiledCode<kVerifyNone>(GetOatAddress(quick_to_interpreter_bridge_offset_));
- copy->SetEntryPointFromInterpreter<kVerifyNone>(reinterpret_cast<EntryPointFromInterpreter*>
- (const_cast<uint8_t*>(GetOatAddress(interpreter_to_interpreter_bridge_offset_))));
+ copy->SetEntryPointFromPortableCompiledCodePtrSize<kVerifyNone>(
+ GetOatAddress(portable_to_interpreter_bridge_offset_), target_ptr_size_);
+ copy->SetEntryPointFromQuickCompiledCodePtrSize<kVerifyNone>(
+ GetOatAddress(quick_to_interpreter_bridge_offset_), target_ptr_size_);
+ copy->SetEntryPointFromInterpreterPtrSize<kVerifyNone>(
+ reinterpret_cast<EntryPointFromInterpreter*>(const_cast<uint8_t*>(
+ GetOatAddress(interpreter_to_interpreter_bridge_offset_))), target_ptr_size_);
} else {
bool quick_is_interpreted;
const uint8_t* quick_code = GetQuickCode(orig, &quick_is_interpreted);
- copy->SetEntryPointFromQuickCompiledCode<kVerifyNone>(quick_code);
+ copy->SetEntryPointFromQuickCompiledCodePtrSize<kVerifyNone>(quick_code, target_ptr_size_);
// Portable entrypoint:
const uint8_t* portable_code = GetOatAddress(orig->GetPortableOatCodeOffset());
@@ -937,18 +975,19 @@
// initialization.
portable_code = GetOatAddress(portable_resolution_trampoline_offset_);
}
- copy->SetEntryPointFromPortableCompiledCode<kVerifyNone>(portable_code);
-
+ copy->SetEntryPointFromPortableCompiledCodePtrSize<kVerifyNone>(
+ portable_code, target_ptr_size_);
// JNI entrypoint:
if (orig->IsNative()) {
// The native method's pointer is set to a stub to lookup via dlsym.
// Note this is not the code_ pointer, that is handled above.
- copy->SetNativeMethod<kVerifyNone>(GetOatAddress(jni_dlsym_lookup_offset_));
+ copy->SetEntryPointFromJniPtrSize<kVerifyNone>(GetOatAddress(jni_dlsym_lookup_offset_),
+ target_ptr_size_);
} else {
// Normal (non-abstract non-native) methods have various tables to relocate.
uint32_t native_gc_map_offset = orig->GetOatNativeGcMapOffset();
const uint8_t* native_gc_map = GetOatAddress(native_gc_map_offset);
- copy->SetNativeGcMap<kVerifyNone>(reinterpret_cast<const uint8_t*>(native_gc_map));
+ copy->SetNativeGcMapPtrSize<kVerifyNone>(native_gc_map, target_ptr_size_);
}
// Interpreter entrypoint:
@@ -956,9 +995,11 @@
uint32_t interpreter_code = (quick_is_interpreted && portable_is_interpreted)
? interpreter_to_interpreter_bridge_offset_
: interpreter_to_compiled_code_bridge_offset_;
- copy->SetEntryPointFromInterpreter<kVerifyNone>(
+ EntryPointFromInterpreter* interpreter_entrypoint =
reinterpret_cast<EntryPointFromInterpreter*>(
- const_cast<uint8_t*>(GetOatAddress(interpreter_code))));
+ const_cast<uint8_t*>(GetOatAddress(interpreter_code)));
+ copy->SetEntryPointFromInterpreterPtrSize<kVerifyNone>(
+ interpreter_entrypoint, target_ptr_size_);
}
}
}