v8/test/unittests/compiler/instruction-sequence-unittest.cc

567 lines
19 KiB
C++
Raw Normal View History

// Copyright 2014 the V8 project authors. All rights reserved.
// Use of this source code is governed by a BSD-style license that can be
// found in the LICENSE file.
#include "src/base/utils/random-number-generator.h"
#include "src/compiler/pipeline.h"
#include "test/unittests/compiler/instruction-sequence-unittest.h"
#include "test/unittests/test-utils.h"
#include "testing/gmock/include/gmock/gmock.h"
namespace v8 {
namespace internal {
namespace compiler {
static const char*
general_register_names_[RegisterConfiguration::kMaxGeneralRegisters];
static const char*
double_register_names_[RegisterConfiguration::kMaxFPRegisters];
static char register_names_[10 * (RegisterConfiguration::kMaxGeneralRegisters +
RegisterConfiguration::kMaxFPRegisters)];
Revert of MIPS: Fix bad RegisterConfiguration usage in InstructionSequence unit tests. (patchset #3 id:40001 of https://codereview.chromium.org/2433093002/ ) Reason for revert: This change rendered InstructionSequenceTest::SetNumRegs ineffectual, thus loosening the tests that were using that API to ensure correct register allocation under intentionally constrained setups. For the problem stated in this CL, a solution needs to continue supporting the intentionally set-up test configuration. Original issue's description: > MIPS: Fix bad RegisterConfiguration usage in InstructionSequence unit tests. > > Test InstructionSequenceTest has been initialized with a testing RegisterConfiguration > instance defined in instruction-sequence-unittest.h, whereas class ExplicitOperand which > is being tested used RegisterConfiguration from instruction.cc. In case these two > instances are different, the tests would fail. The issue is fixed by using the same > instance of RegisterConfiguration both for test code and code under test. > > Additionally, the tests in register-allocator-unittest.cc use hardcoded values > for register and begin failing is the hardcoded register is not available for > allocation. Fix by forcing the use of allocatable registers only. > > TEST=unittests.MoveOptimizerTest.RemovesRedundantExplicit,unittests.RegisterAllocatorTest.SpillPhi > BUG= > > Committed: https://crrev.com/0cf56232209d4c9c669b8426680de18806f6c29a > Cr-Commit-Position: refs/heads/master@{#40862} TBR=dcarney@chromium.org,bmeurer@chromium.org,mstarzinger@chromium.org,vogelheim@chromium.org,titzer@chromium.org,ivica.bogosavljevic@imgtec.com # Not skipping CQ checks because original CL landed more than 1 days ago. BUG= Review-Url: https://codereview.chromium.org/2587593002 Cr-Commit-Position: refs/heads/master@{#41777}
2016-12-16 23:09:34 +00:00
namespace {
static int allocatable_codes[InstructionSequenceTest::kDefaultNRegs] = {
0, 1, 2, 3, 4, 5, 6, 7};
}
static void InitializeRegisterNames() {
char* loc = register_names_;
for (int i = 0; i < RegisterConfiguration::kMaxGeneralRegisters; ++i) {
general_register_names_[i] = loc;
loc += base::OS::SNPrintF(loc, 100, "gp_%d", i);
*loc++ = 0;
}
for (int i = 0; i < RegisterConfiguration::kMaxFPRegisters; ++i) {
double_register_names_[i] = loc;
loc += base::OS::SNPrintF(loc, 100, "fp_%d", i) + 1;
*loc++ = 0;
}
}
InstructionSequenceTest::InstructionSequenceTest()
: sequence_(nullptr),
num_general_registers_(kDefaultNRegs),
num_double_registers_(kDefaultNRegs),
instruction_blocks_(zone()),
current_block_(nullptr),
block_returns_(false) {
InitializeRegisterNames();
}
void InstructionSequenceTest::SetNumRegs(int num_general_registers,
int num_double_registers) {
CHECK(!config_);
CHECK(instructions_.empty());
CHECK(instruction_blocks_.empty());
num_general_registers_ = num_general_registers;
num_double_registers_ = num_double_registers;
}
int InstructionSequenceTest::GetNumRegs(MachineRepresentation rep) {
switch (rep) {
case MachineRepresentation::kFloat32:
return config()->num_float_registers();
case MachineRepresentation::kFloat64:
return config()->num_double_registers();
case MachineRepresentation::kSimd128:
return config()->num_simd128_registers();
default:
return config()->num_general_registers();
}
}
int InstructionSequenceTest::GetAllocatableCode(int index,
MachineRepresentation rep) {
switch (rep) {
case MachineRepresentation::kFloat32:
return config()->GetAllocatableFloatCode(index);
case MachineRepresentation::kFloat64:
return config()->GetAllocatableDoubleCode(index);
case MachineRepresentation::kSimd128:
return config()->GetAllocatableSimd128Code(index);
default:
return config()->GetAllocatableGeneralCode(index);
}
}
const RegisterConfiguration* InstructionSequenceTest::config() {
Revert of MIPS: Fix bad RegisterConfiguration usage in InstructionSequence unit tests. (patchset #3 id:40001 of https://codereview.chromium.org/2433093002/ ) Reason for revert: This change rendered InstructionSequenceTest::SetNumRegs ineffectual, thus loosening the tests that were using that API to ensure correct register allocation under intentionally constrained setups. For the problem stated in this CL, a solution needs to continue supporting the intentionally set-up test configuration. Original issue's description: > MIPS: Fix bad RegisterConfiguration usage in InstructionSequence unit tests. > > Test InstructionSequenceTest has been initialized with a testing RegisterConfiguration > instance defined in instruction-sequence-unittest.h, whereas class ExplicitOperand which > is being tested used RegisterConfiguration from instruction.cc. In case these two > instances are different, the tests would fail. The issue is fixed by using the same > instance of RegisterConfiguration both for test code and code under test. > > Additionally, the tests in register-allocator-unittest.cc use hardcoded values > for register and begin failing is the hardcoded register is not available for > allocation. Fix by forcing the use of allocatable registers only. > > TEST=unittests.MoveOptimizerTest.RemovesRedundantExplicit,unittests.RegisterAllocatorTest.SpillPhi > BUG= > > Committed: https://crrev.com/0cf56232209d4c9c669b8426680de18806f6c29a > Cr-Commit-Position: refs/heads/master@{#40862} TBR=dcarney@chromium.org,bmeurer@chromium.org,mstarzinger@chromium.org,vogelheim@chromium.org,titzer@chromium.org,ivica.bogosavljevic@imgtec.com # Not skipping CQ checks because original CL landed more than 1 days ago. BUG= Review-Url: https://codereview.chromium.org/2587593002 Cr-Commit-Position: refs/heads/master@{#41777}
2016-12-16 23:09:34 +00:00
if (!config_) {
config_.reset(new RegisterConfiguration(
num_general_registers_, num_double_registers_, num_general_registers_,
num_double_registers_, allocatable_codes, allocatable_codes,
kSimpleFPAliasing ? RegisterConfiguration::OVERLAP
: RegisterConfiguration::COMBINE,
general_register_names_,
double_register_names_, // float register names
double_register_names_,
double_register_names_)); // SIMD 128 register names
}
return config_.get();
}
InstructionSequence* InstructionSequenceTest::sequence() {
if (sequence_ == nullptr) {
sequence_ = new (zone())
InstructionSequence(isolate(), zone(), &instruction_blocks_);
sequence_->SetRegisterConfigurationForTesting(
InstructionSequenceTest::config());
}
return sequence_;
}
void InstructionSequenceTest::StartLoop(int loop_blocks) {
CHECK_NULL(current_block_);
if (!loop_blocks_.empty()) {
CHECK(!loop_blocks_.back().loop_header_.IsValid());
}
LoopData loop_data = {Rpo::Invalid(), loop_blocks};
loop_blocks_.push_back(loop_data);
}
void InstructionSequenceTest::EndLoop() {
CHECK_NULL(current_block_);
CHECK(!loop_blocks_.empty());
CHECK_EQ(0, loop_blocks_.back().expected_blocks_);
loop_blocks_.pop_back();
}
void InstructionSequenceTest::StartBlock(bool deferred) {
block_returns_ = false;
NewBlock(deferred);
}
Instruction* InstructionSequenceTest::EndBlock(BlockCompletion completion) {
Instruction* result = nullptr;
if (block_returns_) {
CHECK(completion.type_ == kBlockEnd || completion.type_ == kFallThrough);
completion.type_ = kBlockEnd;
}
switch (completion.type_) {
case kBlockEnd:
break;
case kFallThrough:
result = EmitJump();
break;
case kJump:
CHECK(!block_returns_);
result = EmitJump();
break;
case kBranch:
CHECK(!block_returns_);
result = EmitBranch(completion.op_);
break;
}
completions_.push_back(completion);
CHECK_NOT_NULL(current_block_);
sequence()->EndBlock(current_block_->rpo_number());
current_block_ = nullptr;
return result;
}
InstructionSequenceTest::TestOperand InstructionSequenceTest::Imm(int32_t imm) {
return TestOperand(kImmediate, imm);
}
InstructionSequenceTest::VReg InstructionSequenceTest::Define(
TestOperand output_op) {
VReg vreg = NewReg(output_op);
InstructionOperand outputs[1]{ConvertOutputOp(vreg, output_op)};
Emit(kArchNop, 1, outputs);
return vreg;
}
Instruction* InstructionSequenceTest::Return(TestOperand input_op_0) {
block_returns_ = true;
InstructionOperand inputs[1]{ConvertInputOp(input_op_0)};
return Emit(kArchRet, 0, nullptr, 1, inputs);
}
PhiInstruction* InstructionSequenceTest::Phi(VReg incoming_vreg_0,
VReg incoming_vreg_1,
VReg incoming_vreg_2,
VReg incoming_vreg_3) {
VReg inputs[] = {incoming_vreg_0, incoming_vreg_1, incoming_vreg_2,
incoming_vreg_3};
size_t input_count = 0;
for (; input_count < arraysize(inputs); ++input_count) {
if (inputs[input_count].value_ == kNoValue) break;
}
CHECK_LT(0, input_count);
auto phi = new (zone()) PhiInstruction(zone(), NewReg().value_, input_count);
for (size_t i = 0; i < input_count; ++i) {
SetInput(phi, i, inputs[i]);
}
current_block_->AddPhi(phi);
return phi;
}
PhiInstruction* InstructionSequenceTest::Phi(VReg incoming_vreg_0,
size_t input_count) {
auto phi = new (zone()) PhiInstruction(zone(), NewReg().value_, input_count);
SetInput(phi, 0, incoming_vreg_0);
current_block_->AddPhi(phi);
return phi;
}
void InstructionSequenceTest::SetInput(PhiInstruction* phi, size_t input,
VReg vreg) {
CHECK_NE(kNoValue, vreg.value_);
phi->SetInput(input, vreg.value_);
}
InstructionSequenceTest::VReg InstructionSequenceTest::DefineConstant(
int32_t imm) {
VReg vreg = NewReg();
sequence()->AddConstant(vreg.value_, Constant(imm));
InstructionOperand outputs[1]{ConstantOperand(vreg.value_)};
Emit(kArchNop, 1, outputs);
return vreg;
}
Instruction* InstructionSequenceTest::EmitNop() { return Emit(kArchNop); }
static size_t CountInputs(size_t size,
InstructionSequenceTest::TestOperand* inputs) {
size_t i = 0;
for (; i < size; ++i) {
if (inputs[i].type_ == InstructionSequenceTest::kInvalid) break;
}
return i;
}
Instruction* InstructionSequenceTest::EmitI(size_t input_size,
TestOperand* inputs) {
InstructionOperand* mapped_inputs = ConvertInputs(input_size, inputs);
return Emit(kArchNop, 0, nullptr, input_size, mapped_inputs);
}
Instruction* InstructionSequenceTest::EmitI(TestOperand input_op_0,
TestOperand input_op_1,
TestOperand input_op_2,
TestOperand input_op_3) {
TestOperand inputs[] = {input_op_0, input_op_1, input_op_2, input_op_3};
return EmitI(CountInputs(arraysize(inputs), inputs), inputs);
}
InstructionSequenceTest::VReg InstructionSequenceTest::EmitOI(
TestOperand output_op, size_t input_size, TestOperand* inputs) {
VReg output_vreg = NewReg(output_op);
InstructionOperand outputs[1]{ConvertOutputOp(output_vreg, output_op)};
InstructionOperand* mapped_inputs = ConvertInputs(input_size, inputs);
Emit(kArchNop, 1, outputs, input_size, mapped_inputs);
return output_vreg;
}
InstructionSequenceTest::VReg InstructionSequenceTest::EmitOI(
TestOperand output_op, TestOperand input_op_0, TestOperand input_op_1,
TestOperand input_op_2, TestOperand input_op_3) {
TestOperand inputs[] = {input_op_0, input_op_1, input_op_2, input_op_3};
return EmitOI(output_op, CountInputs(arraysize(inputs), inputs), inputs);
}
InstructionSequenceTest::VRegPair InstructionSequenceTest::EmitOOI(
TestOperand output_op_0, TestOperand output_op_1, size_t input_size,
TestOperand* inputs) {
VRegPair output_vregs =
std::make_pair(NewReg(output_op_0), NewReg(output_op_1));
InstructionOperand outputs[2]{
ConvertOutputOp(output_vregs.first, output_op_0),
ConvertOutputOp(output_vregs.second, output_op_1)};
InstructionOperand* mapped_inputs = ConvertInputs(input_size, inputs);
Emit(kArchNop, 2, outputs, input_size, mapped_inputs);
return output_vregs;
}
InstructionSequenceTest::VRegPair InstructionSequenceTest::EmitOOI(
TestOperand output_op_0, TestOperand output_op_1, TestOperand input_op_0,
TestOperand input_op_1, TestOperand input_op_2, TestOperand input_op_3) {
TestOperand inputs[] = {input_op_0, input_op_1, input_op_2, input_op_3};
return EmitOOI(output_op_0, output_op_1,
CountInputs(arraysize(inputs), inputs), inputs);
}
InstructionSequenceTest::VReg InstructionSequenceTest::EmitCall(
TestOperand output_op, size_t input_size, TestOperand* inputs) {
VReg output_vreg = NewReg(output_op);
InstructionOperand outputs[1]{ConvertOutputOp(output_vreg, output_op)};
CHECK(UnallocatedOperand::cast(outputs[0]).HasFixedPolicy());
InstructionOperand* mapped_inputs = ConvertInputs(input_size, inputs);
Emit(kArchCallCodeObject, 1, outputs, input_size, mapped_inputs, 0, nullptr,
true);
return output_vreg;
}
InstructionSequenceTest::VReg InstructionSequenceTest::EmitCall(
TestOperand output_op, TestOperand input_op_0, TestOperand input_op_1,
TestOperand input_op_2, TestOperand input_op_3) {
TestOperand inputs[] = {input_op_0, input_op_1, input_op_2, input_op_3};
return EmitCall(output_op, CountInputs(arraysize(inputs), inputs), inputs);
}
Instruction* InstructionSequenceTest::EmitBranch(TestOperand input_op) {
InstructionOperand inputs[4]{ConvertInputOp(input_op), ConvertInputOp(Imm()),
ConvertInputOp(Imm()), ConvertInputOp(Imm())};
InstructionCode opcode = kArchJmp | FlagsModeField::encode(kFlags_branch) |
FlagsConditionField::encode(kEqual);
auto instruction = NewInstruction(opcode, 0, nullptr, 4, inputs);
return AddInstruction(instruction);
}
Instruction* InstructionSequenceTest::EmitFallThrough() {
auto instruction = NewInstruction(kArchNop, 0, nullptr);
return AddInstruction(instruction);
}
Instruction* InstructionSequenceTest::EmitJump() {
InstructionOperand inputs[1]{ConvertInputOp(Imm())};
auto instruction = NewInstruction(kArchJmp, 0, nullptr, 1, inputs);
return AddInstruction(instruction);
}
Instruction* InstructionSequenceTest::NewInstruction(
InstructionCode code, size_t outputs_size, InstructionOperand* outputs,
size_t inputs_size, InstructionOperand* inputs, size_t temps_size,
InstructionOperand* temps) {
CHECK(current_block_);
return Instruction::New(zone(), code, outputs_size, outputs, inputs_size,
inputs, temps_size, temps);
}
InstructionOperand InstructionSequenceTest::Unallocated(
TestOperand op, UnallocatedOperand::ExtendedPolicy policy) {
return UnallocatedOperand(policy, op.vreg_.value_);
}
InstructionOperand InstructionSequenceTest::Unallocated(
TestOperand op, UnallocatedOperand::ExtendedPolicy policy,
UnallocatedOperand::Lifetime lifetime) {
return UnallocatedOperand(policy, lifetime, op.vreg_.value_);
}
InstructionOperand InstructionSequenceTest::Unallocated(
TestOperand op, UnallocatedOperand::ExtendedPolicy policy, int index) {
return UnallocatedOperand(policy, index, op.vreg_.value_);
}
InstructionOperand InstructionSequenceTest::Unallocated(
TestOperand op, UnallocatedOperand::BasicPolicy policy, int index) {
return UnallocatedOperand(policy, index, op.vreg_.value_);
}
InstructionOperand* InstructionSequenceTest::ConvertInputs(
size_t input_size, TestOperand* inputs) {
InstructionOperand* mapped_inputs =
zone()->NewArray<InstructionOperand>(static_cast<int>(input_size));
for (size_t i = 0; i < input_size; ++i) {
mapped_inputs[i] = ConvertInputOp(inputs[i]);
}
return mapped_inputs;
}
InstructionOperand InstructionSequenceTest::ConvertInputOp(TestOperand op) {
if (op.type_ == kImmediate) {
CHECK_EQ(op.vreg_.value_, kNoValue);
return ImmediateOperand(ImmediateOperand::INLINE, op.value_);
}
CHECK_NE(op.vreg_.value_, kNoValue);
switch (op.type_) {
case kNone:
return Unallocated(op, UnallocatedOperand::NONE,
UnallocatedOperand::USED_AT_START);
case kUnique:
return Unallocated(op, UnallocatedOperand::NONE);
case kUniqueRegister:
return Unallocated(op, UnallocatedOperand::MUST_HAVE_REGISTER);
case kRegister:
return Unallocated(op, UnallocatedOperand::MUST_HAVE_REGISTER,
UnallocatedOperand::USED_AT_START);
case kSlot:
return Unallocated(op, UnallocatedOperand::MUST_HAVE_SLOT,
UnallocatedOperand::USED_AT_START);
case kFixedRegister: {
MachineRepresentation rep = GetCanonicalRep(op);
CHECK(0 <= op.value_ && op.value_ < GetNumRegs(rep));
if (DoesRegisterAllocation()) {
auto extended_policy = IsFloatingPoint(rep)
? UnallocatedOperand::FIXED_FP_REGISTER
: UnallocatedOperand::FIXED_REGISTER;
return Unallocated(op, extended_policy, op.value_);
} else {
return AllocatedOperand(LocationOperand::REGISTER, rep, op.value_);
}
}
case kFixedSlot:
if (DoesRegisterAllocation()) {
return Unallocated(op, UnallocatedOperand::FIXED_SLOT, op.value_);
} else {
return AllocatedOperand(LocationOperand::STACK_SLOT,
GetCanonicalRep(op), op.value_);
}
default:
break;
}
CHECK(false);
return InstructionOperand();
}
InstructionOperand InstructionSequenceTest::ConvertOutputOp(VReg vreg,
TestOperand op) {
CHECK_EQ(op.vreg_.value_, kNoValue);
op.vreg_ = vreg;
switch (op.type_) {
case kSameAsFirst:
return Unallocated(op, UnallocatedOperand::SAME_AS_FIRST_INPUT);
case kRegister:
return Unallocated(op, UnallocatedOperand::MUST_HAVE_REGISTER);
case kFixedSlot:
if (DoesRegisterAllocation()) {
return Unallocated(op, UnallocatedOperand::FIXED_SLOT, op.value_);
} else {
return AllocatedOperand(LocationOperand::STACK_SLOT,
GetCanonicalRep(op), op.value_);
}
case kFixedRegister: {
MachineRepresentation rep = GetCanonicalRep(op);
CHECK(0 <= op.value_ && op.value_ < GetNumRegs(rep));
if (DoesRegisterAllocation()) {
auto extended_policy = IsFloatingPoint(rep)
? UnallocatedOperand::FIXED_FP_REGISTER
: UnallocatedOperand::FIXED_REGISTER;
return Unallocated(op, extended_policy, op.value_);
} else {
return AllocatedOperand(LocationOperand::REGISTER, rep, op.value_);
}
}
default:
break;
}
CHECK(false);
return InstructionOperand();
}
InstructionBlock* InstructionSequenceTest::NewBlock(bool deferred) {
CHECK_NULL(current_block_);
Rpo rpo = Rpo::FromInt(static_cast<int>(instruction_blocks_.size()));
Rpo loop_header = Rpo::Invalid();
Rpo loop_end = Rpo::Invalid();
if (!loop_blocks_.empty()) {
auto& loop_data = loop_blocks_.back();
// This is a loop header.
if (!loop_data.loop_header_.IsValid()) {
loop_end = Rpo::FromInt(rpo.ToInt() + loop_data.expected_blocks_);
loop_data.expected_blocks_--;
loop_data.loop_header_ = rpo;
} else {
// This is a loop body.
CHECK_NE(0, loop_data.expected_blocks_);
// TODO(dcarney): handle nested loops.
loop_data.expected_blocks_--;
loop_header = loop_data.loop_header_;
}
}
// Construct instruction block.
auto instruction_block = new (zone())
InstructionBlock(zone(), rpo, loop_header, loop_end, deferred, false);
instruction_blocks_.push_back(instruction_block);
current_block_ = instruction_block;
sequence()->StartBlock(rpo);
return instruction_block;
}
void InstructionSequenceTest::WireBlocks() {
CHECK(!current_block());
CHECK(instruction_blocks_.size() == completions_.size());
CHECK(loop_blocks_.empty());
// Wire in end block to look like a scheduler produced cfg.
auto end_block = NewBlock();
current_block_ = nullptr;
sequence()->EndBlock(end_block->rpo_number());
size_t offset = 0;
for (const auto& completion : completions_) {
switch (completion.type_) {
case kBlockEnd: {
auto block = instruction_blocks_[offset];
block->successors().push_back(end_block->rpo_number());
end_block->predecessors().push_back(block->rpo_number());
break;
}
case kFallThrough: // Fallthrough.
case kJump:
WireBlock(offset, completion.offset_0_);
break;
case kBranch:
WireBlock(offset, completion.offset_0_);
WireBlock(offset, completion.offset_1_);
break;
}
++offset;
}
}
void InstructionSequenceTest::WireBlock(size_t block_offset, int jump_offset) {
size_t target_block_offset = block_offset + static_cast<size_t>(jump_offset);
CHECK(block_offset < instruction_blocks_.size());
CHECK(target_block_offset < instruction_blocks_.size());
auto block = instruction_blocks_[block_offset];
auto target = instruction_blocks_[target_block_offset];
block->successors().push_back(target->rpo_number());
target->predecessors().push_back(block->rpo_number());
}
Instruction* InstructionSequenceTest::Emit(
InstructionCode code, size_t outputs_size, InstructionOperand* outputs,
size_t inputs_size, InstructionOperand* inputs, size_t temps_size,
InstructionOperand* temps, bool is_call) {
auto instruction = NewInstruction(code, outputs_size, outputs, inputs_size,
inputs, temps_size, temps);
if (is_call) instruction->MarkAsCall();
return AddInstruction(instruction);
}
Instruction* InstructionSequenceTest::AddInstruction(Instruction* instruction) {
sequence()->AddInstruction(instruction);
return instruction;
}
} // namespace compiler
} // namespace internal
} // namespace v8