Implement checkcast for optimizing.
- Ended up not using HTypeCheck because of how
instanceof and checkcast end up having different logic
for code generation.
- Fix a x86_64 assembler bug triggered by now enabling
more methods to be compiled. Difficult to test today
without b/18117217.
Change-Id: I3022e7ae03befb1d10bea9637ad21fadc430abe0
diff --git a/compiler/optimizing/code_generator_x86_64.cc b/compiler/optimizing/code_generator_x86_64.cc
index e09b6ca..bd1e4f4 100644
--- a/compiler/optimizing/code_generator_x86_64.cc
+++ b/compiler/optimizing/code_generator_x86_64.cc
@@ -284,13 +284,19 @@
class TypeCheckSlowPathX86_64 : public SlowPathCodeX86_64 {
public:
- TypeCheckSlowPathX86_64(HTypeCheck* instruction, Location object_class)
+ TypeCheckSlowPathX86_64(HInstruction* instruction,
+ Location class_to_check,
+ Location object_class,
+ uint32_t dex_pc)
: instruction_(instruction),
- object_class_(object_class) {}
+ class_to_check_(class_to_check),
+ object_class_(object_class),
+ dex_pc_(dex_pc) {}
virtual void EmitNativeCode(CodeGenerator* codegen) OVERRIDE {
LocationSummary* locations = instruction_->GetLocations();
- DCHECK(!locations->GetLiveRegisters()->ContainsCoreRegister(locations->Out().reg()));
+ DCHECK(instruction_->IsCheckCast()
+ || !locations->GetLiveRegisters()->ContainsCoreRegister(locations->Out().reg()));
CodeGeneratorX86_64* x64_codegen = down_cast<CodeGeneratorX86_64*>(codegen);
__ Bind(GetEntryLabel());
@@ -299,7 +305,7 @@
// We're moving two locations to locations that could overlap, so we need a parallel
// move resolver.
InvokeRuntimeCallingConvention calling_convention;
- MoveOperands move1(locations->InAt(1),
+ MoveOperands move1(class_to_check_,
Location::RegisterLocation(calling_convention.GetRegisterAt(0)),
nullptr);
MoveOperands move2(object_class_,
@@ -310,18 +316,29 @@
parallel_move.AddMove(&move2);
x64_codegen->GetMoveResolver()->EmitNativeCode(¶llel_move);
- __ gs()->call(
- Address::Absolute(QUICK_ENTRYPOINT_OFFSET(kX86_64WordSize, pInstanceofNonTrivial), true));
- codegen->RecordPcInfo(instruction_, instruction_->GetDexPc());
- x64_codegen->Move(locations->Out(), Location::RegisterLocation(RAX));
+ if (instruction_->IsInstanceOf()) {
+ __ gs()->call(
+ Address::Absolute(QUICK_ENTRYPOINT_OFFSET(kX86_64WordSize, pInstanceofNonTrivial), true));
+ } else {
+ DCHECK(instruction_->IsCheckCast());
+ __ gs()->call(
+ Address::Absolute(QUICK_ENTRYPOINT_OFFSET(kX86_64WordSize, pCheckCast), true));
+ }
+ codegen->RecordPcInfo(instruction_, dex_pc_);
+
+ if (instruction_->IsInstanceOf()) {
+ x64_codegen->Move(locations->Out(), Location::RegisterLocation(RAX));
+ }
codegen->RestoreLiveRegisters(locations);
__ jmp(GetExitLabel());
}
private:
- HTypeCheck* const instruction_;
+ HInstruction* const instruction_;
+ const Location class_to_check_;
const Location object_class_;
+ const uint32_t dex_pc_;
DISALLOW_COPY_AND_ASSIGN(TypeCheckSlowPathX86_64);
};
@@ -2743,7 +2760,7 @@
codegen_->RecordPcInfo(instruction, instruction->GetDexPc());
}
-void LocationsBuilderX86_64::VisitTypeCheck(HTypeCheck* instruction) {
+void LocationsBuilderX86_64::VisitInstanceOf(HInstanceOf* instruction) {
LocationSummary::CallKind call_kind = instruction->IsClassFinal()
? LocationSummary::kNoCall
: LocationSummary::kCallOnSlowPath;
@@ -2753,7 +2770,7 @@
locations->SetOut(Location::RequiresRegister());
}
-void InstructionCodeGeneratorX86_64::VisitTypeCheck(HTypeCheck* instruction) {
+void InstructionCodeGeneratorX86_64::VisitInstanceOf(HInstanceOf* instruction) {
LocationSummary* locations = instruction->GetLocations();
CpuRegister obj = locations->InAt(0).As<CpuRegister>();
Location cls = locations->InAt(1);
@@ -2783,7 +2800,7 @@
// If the classes are not equal, we go into a slow path.
DCHECK(locations->OnlyCallsOnSlowPath());
slow_path = new (GetGraph()->GetArena()) TypeCheckSlowPathX86_64(
- instruction, Location::RegisterLocation(out.AsRegister()));
+ instruction, locations->InAt(1), locations->Out(), instruction->GetDexPc());
codegen_->AddSlowPath(slow_path);
__ j(kNotEqual, slow_path->GetEntryLabel());
__ movl(out, Immediate(1));
@@ -2797,5 +2814,39 @@
__ Bind(&done);
}
+void LocationsBuilderX86_64::VisitCheckCast(HCheckCast* instruction) {
+ LocationSummary* locations = new (GetGraph()->GetArena()) LocationSummary(
+ instruction, LocationSummary::kCallOnSlowPath);
+ locations->SetInAt(0, Location::RequiresRegister());
+ locations->SetInAt(1, Location::Any());
+ locations->AddTemp(Location::RequiresRegister());
+}
+
+void InstructionCodeGeneratorX86_64::VisitCheckCast(HCheckCast* instruction) {
+ LocationSummary* locations = instruction->GetLocations();
+ CpuRegister obj = locations->InAt(0).As<CpuRegister>();
+ Location cls = locations->InAt(1);
+ CpuRegister temp = locations->GetTemp(0).As<CpuRegister>();
+ uint32_t class_offset = mirror::Object::ClassOffset().Int32Value();
+ SlowPathCodeX86_64* slow_path = new (GetGraph()->GetArena()) TypeCheckSlowPathX86_64(
+ instruction, locations->InAt(1), locations->GetTemp(0), instruction->GetDexPc());
+ codegen_->AddSlowPath(slow_path);
+
+ // TODO: avoid this check if we know obj is not null.
+ __ testl(obj, obj);
+ __ j(kEqual, slow_path->GetExitLabel());
+ // Compare the class of `obj` with `cls`.
+ __ movl(temp, Address(obj, class_offset));
+ if (cls.IsRegister()) {
+ __ cmpl(temp, cls.As<CpuRegister>());
+ } else {
+ DCHECK(cls.IsStackSlot()) << cls;
+ __ cmpl(temp, Address(CpuRegister(RSP), cls.GetStackIndex()));
+ }
+ // Classes must be equal for the checkcast to succeed.
+ __ j(kNotEqual, slow_path->GetEntryLabel());
+ __ Bind(slow_path->GetExitLabel());
+}
+
} // namespace x86_64
} // namespace art