Compiler changes for bitstring based type checks.
We guard the use of this feature with a compile-time flag,
set to true in this CL.
Boot image size for aosp_taimen-userdebug in AOSP master:
- before:
arm boot*.oat: 63604740
arm64 boot*.oat: 74237864
- after:
arm boot*.oat: 63531172 (-72KiB, -0.1%)
arm64 boot*.oat: 74135008 (-100KiB, -0.1%)
The new TypeCheckBenchmark yields the following changes
using the little cores of taimen fixed at 1.4016GHz:
32-bit 64-bit
timeCheckCastLevel1ToLevel1 11.48->15.80 11.47->15.78
timeCheckCastLevel2ToLevel1 15.08->15.79 15.08->15.79
timeCheckCastLevel3ToLevel1 19.01->15.82 17.94->15.81
timeCheckCastLevel9ToLevel1 42.55->15.79 42.63->15.81
timeCheckCastLevel9ToLevel2 39.70->14.36 39.70->14.35
timeInstanceOfLevel1ToLevel1 13.74->17.93 13.76->17.95
timeInstanceOfLevel2ToLevel1 17.02->17.95 16.99->17.93
timeInstanceOfLevel3ToLevel1 24.03->17.95 24.45->17.95
timeInstanceOfLevel9ToLevel1 47.13->17.95 47.14->18.00
timeInstanceOfLevel9ToLevel2 44.19->16.52 44.27->16.51
This suggests that the bitstring typecheck should not be
used for exact type checks which would be equivalent to the
"Level1ToLevel1" benchmark. Whether the implementation is
a beneficial replacement for the kClassHierarchyCheck and
kAbstractClassCheck on average depends on how many levels
from the target class (or Object for a negative result) is
a typical object's class.
Test: m test-art-host-gtest
Test: testrunner.py --host --optimizing --jit
Test: testrunner.py --host -t 670-bitstring-type-check
Test: Pixel 2 XL boots.
Test: testrunner.py --target --optimizing --jit
Test: testrunner.py --target -t 670-bitstring-type-check
Bug: 64692057
Bug: 71853552
Bug: 26687569
Change-Id: I538d7e036b5a8ae2cc3fe77662a5903d74854562
diff --git a/compiler/optimizing/code_generator_x86.cc b/compiler/optimizing/code_generator_x86.cc
index cbe9e0a..e85f900 100644
--- a/compiler/optimizing/code_generator_x86.cc
+++ b/compiler/optimizing/code_generator_x86.cc
@@ -6234,6 +6234,27 @@
// No need for memory fence, thanks to the X86 memory model.
}
+void InstructionCodeGeneratorX86::GenerateBitstringTypeCheckCompare(HTypeCheckInstruction* check,
+ Register temp) {
+ uint32_t path_to_root = check->GetBitstringPathToRoot();
+ uint32_t mask = check->GetBitstringMask();
+ DCHECK(IsPowerOfTwo(mask + 1));
+ size_t mask_bits = WhichPowerOf2(mask + 1);
+
+ if ((false) && mask_bits == 16u) {
+ // FIXME: cmpw() erroneously emits the constant as 32 bits instead of 16 bits. b/71853552
+ // Compare the bitstring in memory.
+ __ cmpw(Address(temp, mirror::Class::StatusOffset()), Immediate(path_to_root));
+ } else {
+ // /* uint32_t */ temp = temp->status_
+ __ movl(temp, Address(temp, mirror::Class::StatusOffset()));
+ // Compare the bitstring bits using SUB.
+ __ subl(temp, Immediate(path_to_root));
+ // Shift out bits that do not contribute to the comparison.
+ __ shll(temp, Immediate(32u - mask_bits));
+ }
+}
+
HLoadString::LoadKind CodeGeneratorX86::GetSupportedLoadStringKind(
HLoadString::LoadKind desired_string_load_kind) {
switch (desired_string_load_kind) {
@@ -6426,6 +6447,8 @@
case TypeCheckKind::kInterfaceCheck:
call_kind = LocationSummary::kCallOnSlowPath;
break;
+ case TypeCheckKind::kBitstringCheck:
+ break;
}
LocationSummary* locations =
@@ -6434,7 +6457,13 @@
locations->SetCustomSlowPathCallerSaves(RegisterSet::Empty()); // No caller-save registers.
}
locations->SetInAt(0, Location::RequiresRegister());
- locations->SetInAt(1, Location::Any());
+ if (type_check_kind == TypeCheckKind::kBitstringCheck) {
+ locations->SetInAt(1, Location::ConstantLocation(instruction->InputAt(1)->AsConstant()));
+ locations->SetInAt(2, Location::ConstantLocation(instruction->InputAt(2)->AsConstant()));
+ locations->SetInAt(3, Location::ConstantLocation(instruction->InputAt(3)->AsConstant()));
+ } else {
+ locations->SetInAt(1, Location::Any());
+ }
// Note that TypeCheckSlowPathX86 uses this "out" register too.
locations->SetOut(Location::RequiresRegister());
// When read barriers are enabled, we need a temporary register for some cases.
@@ -6655,6 +6684,21 @@
}
break;
}
+
+ case TypeCheckKind::kBitstringCheck: {
+ // /* HeapReference<Class> */ temp = obj->klass_
+ GenerateReferenceLoadTwoRegisters(instruction,
+ out_loc,
+ obj_loc,
+ class_offset,
+ kWithoutReadBarrier);
+
+ GenerateBitstringTypeCheckCompare(instruction, out);
+ __ j(kNotEqual, &zero);
+ __ movl(out, Immediate(1));
+ __ jmp(&done);
+ break;
+ }
}
if (zero.IsLinked()) {
@@ -6681,6 +6725,10 @@
// Require a register for the interface check since there is a loop that compares the class to
// a memory address.
locations->SetInAt(1, Location::RequiresRegister());
+ } else if (type_check_kind == TypeCheckKind::kBitstringCheck) {
+ locations->SetInAt(1, Location::ConstantLocation(instruction->InputAt(1)->AsConstant()));
+ locations->SetInAt(2, Location::ConstantLocation(instruction->InputAt(2)->AsConstant()));
+ locations->SetInAt(3, Location::ConstantLocation(instruction->InputAt(3)->AsConstant()));
} else {
locations->SetInAt(1, Location::Any());
}
@@ -6900,6 +6948,19 @@
__ MaybeUnpoisonHeapReference(cls.AsRegister<Register>());
break;
}
+
+ case TypeCheckKind::kBitstringCheck: {
+ // /* HeapReference<Class> */ temp = obj->klass_
+ GenerateReferenceLoadTwoRegisters(instruction,
+ temp_loc,
+ obj_loc,
+ class_offset,
+ kWithoutReadBarrier);
+
+ GenerateBitstringTypeCheckCompare(instruction, temp);
+ __ j(kNotEqual, type_check_slow_path->GetEntryLabel());
+ break;
+ }
}
__ Bind(&done);