v8/test/unittests/interpreter/bytecodes-unittest.cc

365 lines
15 KiB
C++
Raw Normal View History

// Copyright 2015 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 <vector>
#include "src/v8.h"
#include "src/interpreter/bytecode-register.h"
#include "src/interpreter/bytecodes.h"
#include "test/unittests/test-utils.h"
namespace v8 {
namespace internal {
namespace interpreter {
TEST(OperandConversion, Registers) {
int register_count = 128;
int step = register_count / 7;
for (int i = 0; i < register_count; i += step) {
if (i <= kMaxInt8) {
uint32_t operand0 = Register(i).ToOperand();
Register reg0 = Register::FromOperand(operand0);
CHECK_EQ(i, reg0.index());
}
uint32_t operand1 = Register(i).ToOperand();
Register reg1 = Register::FromOperand(operand1);
CHECK_EQ(i, reg1.index());
uint32_t operand2 = Register(i).ToOperand();
Register reg2 = Register::FromOperand(operand2);
CHECK_EQ(i, reg2.index());
}
}
TEST(OperandConversion, Parameters) {
int parameter_counts[] = {7, 13, 99};
size_t count = sizeof(parameter_counts) / sizeof(parameter_counts[0]);
for (size_t p = 0; p < count; p++) {
int parameter_count = parameter_counts[p];
for (int i = 0; i < parameter_count; i++) {
Register r = Register::FromParameterIndex(i, parameter_count);
uint32_t operand_value = r.ToOperand();
Register s = Register::FromOperand(operand_value);
CHECK_EQ(i, s.ToParameterIndex(parameter_count));
}
}
}
TEST(OperandConversion, RegistersParametersNoOverlap) {
int register_count = 128;
int parameter_count = 100;
int32_t register_space_size = base::bits::RoundUpToPowerOfTwo32(
static_cast<uint32_t>(register_count + parameter_count));
uint32_t range = static_cast<uint32_t>(register_space_size);
std::vector<uint8_t> operand_count(range);
for (int i = 0; i < register_count; i += 1) {
Register r = Register(i);
int32_t operand = r.ToOperand();
uint8_t index = static_cast<uint8_t>(operand);
CHECK_LT(index, operand_count.size());
operand_count[index] += 1;
CHECK_EQ(operand_count[index], 1);
}
for (int i = 0; i < parameter_count; i += 1) {
Register r = Register::FromParameterIndex(i, parameter_count);
uint32_t operand = r.ToOperand();
uint8_t index = static_cast<uint8_t>(operand);
CHECK_LT(index, operand_count.size());
operand_count[index] += 1;
CHECK_EQ(operand_count[index], 1);
}
}
TEST(OperandScaling, ScalableAndNonScalable) {
const OperandScale kOperandScales[] = {
#define VALUE(Name, _) OperandScale::k##Name,
OPERAND_SCALE_LIST(VALUE)
#undef VALUE
};
for (OperandScale operand_scale : kOperandScales) {
int scale = static_cast<int>(operand_scale);
CHECK_EQ(Bytecodes::Size(Bytecode::kCallRuntime, operand_scale),
1 + 2 + 2 * scale);
CHECK_EQ(Bytecodes::Size(Bytecode::kCreateObjectLiteral, operand_scale),
1 + 2 * scale + 1);
CHECK_EQ(Bytecodes::Size(Bytecode::kTestIn, operand_scale), 1 + 2 * scale);
}
}
TEST(Bytecodes, RegisterOperands) {
CHECK(Bytecodes::IsRegisterOperandType(OperandType::kReg));
CHECK(Bytecodes::IsRegisterOperandType(OperandType::kRegPair));
CHECK(Bytecodes::IsRegisterInputOperandType(OperandType::kReg));
CHECK(Bytecodes::IsRegisterInputOperandType(OperandType::kRegPair));
CHECK(Bytecodes::IsRegisterInputOperandType(OperandType::kRegList));
CHECK(!Bytecodes::IsRegisterOutputOperandType(OperandType::kReg));
CHECK(!Bytecodes::IsRegisterInputOperandType(OperandType::kRegOut));
CHECK(Bytecodes::IsRegisterOutputOperandType(OperandType::kRegOut));
CHECK(Bytecodes::IsRegisterOutputOperandType(OperandType::kRegOutPair));
}
TEST(Bytecodes, DebugBreakExistForEachBytecode) {
static const OperandScale kOperandScale = OperandScale::kSingle;
#define CHECK_DEBUG_BREAK_SIZE(Name, ...) \
if (!Bytecodes::IsDebugBreak(Bytecode::k##Name) && \
!Bytecodes::IsPrefixScalingBytecode(Bytecode::k##Name)) { \
Bytecode debug_bytecode = Bytecodes::GetDebugBreak(Bytecode::k##Name); \
CHECK_EQ(Bytecodes::Size(Bytecode::k##Name, kOperandScale), \
Bytecodes::Size(debug_bytecode, kOperandScale)); \
}
BYTECODE_LIST(CHECK_DEBUG_BREAK_SIZE)
#undef CHECK_DEBUG_BREAK_SIZE
}
TEST(Bytecodes, DebugBreakForPrefixBytecodes) {
CHECK_EQ(Bytecode::kDebugBreakWide,
Bytecodes::GetDebugBreak(Bytecode::kWide));
CHECK_EQ(Bytecode::kDebugBreakExtraWide,
Bytecodes::GetDebugBreak(Bytecode::kExtraWide));
}
TEST(Bytecodes, PrefixMappings) {
Bytecode prefixes[] = {Bytecode::kWide, Bytecode::kExtraWide};
TRACED_FOREACH(Bytecode, prefix, prefixes) {
CHECK_EQ(prefix, Bytecodes::OperandScaleToPrefixBytecode(
Bytecodes::PrefixBytecodeToOperandScale(prefix)));
}
}
[Interpreter] Optimize BytecodeArrayBuilder and BytecodeArrayWriter. This CL optimizes the code in BytecodeArrayBuilder and BytecodeArrayWriter by making the following main changes: - Move operand scale calculation out of BytecodeArrayWriter to the BytecodeNode constructor, where the decision on which operands are scalable can generally be statically decided by the compiler. - Move the maximum register calculation out of BytecodeArrayWriter and into BytecodeRegisterOptimizer (which is the only place outside BytecodeGenerator which updates which registers are used). This avoids the BytecodeArrayWriter needing to know the operand types of a node as it writes it. - Modify EmitBytecodes to use individual push_backs rather than building a buffer and calling insert, since this turns out to be faster. - Initialize BytecodeArrayWriter's bytecode vector by reserving 512 bytes, - Make common functions in Bytecodes constexpr so that they can be statically calculated by the compiler. - Move common functions and constructors in Bytecodes and BytecodeNode to the header so that they can be inlined. - Change large static switch statements in Bytecodes to const array lookups, and move to the header to allow inlining. I also took the opportunity to remove a number of unused helper functions, and rework some others for consistency. This reduces the percentage of time spent in making BytecodeArrays in CodeLoad from ~15% to ~11% according to perf. The CoadLoad score increase by around 2%. BUG=v8:4280 Committed: https://crrev.com/b11a8b4d41bf09d6b3d6cf214fe3fb61faf01a64 Review-Url: https://codereview.chromium.org/2351763002 Cr-Original-Commit-Position: refs/heads/master@{#39599} Cr-Commit-Position: refs/heads/master@{#39637}
2016-09-22 16:34:16 +00:00
TEST(Bytecodes, ScaleForSignedOperand) {
CHECK_EQ(Bytecodes::ScaleForSignedOperand(0), OperandScale::kSingle);
CHECK_EQ(Bytecodes::ScaleForSignedOperand(kMaxInt8), OperandScale::kSingle);
CHECK_EQ(Bytecodes::ScaleForSignedOperand(kMinInt8), OperandScale::kSingle);
CHECK_EQ(Bytecodes::ScaleForSignedOperand(kMaxInt8 + 1),
OperandScale::kDouble);
CHECK_EQ(Bytecodes::ScaleForSignedOperand(kMinInt8 - 1),
OperandScale::kDouble);
CHECK_EQ(Bytecodes::ScaleForSignedOperand(kMaxInt16), OperandScale::kDouble);
CHECK_EQ(Bytecodes::ScaleForSignedOperand(kMinInt16), OperandScale::kDouble);
CHECK_EQ(Bytecodes::ScaleForSignedOperand(kMaxInt16 + 1),
OperandScale::kQuadruple);
CHECK_EQ(Bytecodes::ScaleForSignedOperand(kMinInt16 - 1),
OperandScale::kQuadruple);
CHECK_EQ(Bytecodes::ScaleForSignedOperand(kMaxInt), OperandScale::kQuadruple);
CHECK_EQ(Bytecodes::ScaleForSignedOperand(kMinInt), OperandScale::kQuadruple);
[Interpreter] Optimize BytecodeArrayBuilder and BytecodeArrayWriter. This CL optimizes the code in BytecodeArrayBuilder and BytecodeArrayWriter by making the following main changes: - Move operand scale calculation out of BytecodeArrayWriter to the BytecodeNode constructor, where the decision on which operands are scalable can generally be statically decided by the compiler. - Move the maximum register calculation out of BytecodeArrayWriter and into BytecodeRegisterOptimizer (which is the only place outside BytecodeGenerator which updates which registers are used). This avoids the BytecodeArrayWriter needing to know the operand types of a node as it writes it. - Modify EmitBytecodes to use individual push_backs rather than building a buffer and calling insert, since this turns out to be faster. - Initialize BytecodeArrayWriter's bytecode vector by reserving 512 bytes, - Make common functions in Bytecodes constexpr so that they can be statically calculated by the compiler. - Move common functions and constructors in Bytecodes and BytecodeNode to the header so that they can be inlined. - Change large static switch statements in Bytecodes to const array lookups, and move to the header to allow inlining. I also took the opportunity to remove a number of unused helper functions, and rework some others for consistency. This reduces the percentage of time spent in making BytecodeArrays in CodeLoad from ~15% to ~11% according to perf. The CoadLoad score increase by around 2%. BUG=v8:4280 Committed: https://crrev.com/b11a8b4d41bf09d6b3d6cf214fe3fb61faf01a64 Review-Url: https://codereview.chromium.org/2351763002 Cr-Original-Commit-Position: refs/heads/master@{#39599} Cr-Commit-Position: refs/heads/master@{#39637}
2016-09-22 16:34:16 +00:00
}
TEST(Bytecodes, ScaleForUnsignedOperands) {
// int overloads
CHECK_EQ(Bytecodes::ScaleForUnsignedOperand(0), OperandScale::kSingle);
CHECK_EQ(Bytecodes::ScaleForUnsignedOperand(kMaxUInt8),
OperandScale::kSingle);
CHECK_EQ(Bytecodes::ScaleForUnsignedOperand(kMaxUInt8 + 1),
OperandScale::kDouble);
CHECK_EQ(Bytecodes::ScaleForUnsignedOperand(kMaxUInt16),
OperandScale::kDouble);
CHECK_EQ(Bytecodes::ScaleForUnsignedOperand(kMaxUInt16 + 1),
OperandScale::kQuadruple);
[Interpreter] Optimize BytecodeArrayBuilder and BytecodeArrayWriter. This CL optimizes the code in BytecodeArrayBuilder and BytecodeArrayWriter by making the following main changes: - Move operand scale calculation out of BytecodeArrayWriter to the BytecodeNode constructor, where the decision on which operands are scalable can generally be statically decided by the compiler. - Move the maximum register calculation out of BytecodeArrayWriter and into BytecodeRegisterOptimizer (which is the only place outside BytecodeGenerator which updates which registers are used). This avoids the BytecodeArrayWriter needing to know the operand types of a node as it writes it. - Modify EmitBytecodes to use individual push_backs rather than building a buffer and calling insert, since this turns out to be faster. - Initialize BytecodeArrayWriter's bytecode vector by reserving 512 bytes, - Make common functions in Bytecodes constexpr so that they can be statically calculated by the compiler. - Move common functions and constructors in Bytecodes and BytecodeNode to the header so that they can be inlined. - Change large static switch statements in Bytecodes to const array lookups, and move to the header to allow inlining. I also took the opportunity to remove a number of unused helper functions, and rework some others for consistency. This reduces the percentage of time spent in making BytecodeArrays in CodeLoad from ~15% to ~11% according to perf. The CoadLoad score increase by around 2%. BUG=v8:4280 Committed: https://crrev.com/b11a8b4d41bf09d6b3d6cf214fe3fb61faf01a64 Review-Url: https://codereview.chromium.org/2351763002 Cr-Original-Commit-Position: refs/heads/master@{#39599} Cr-Commit-Position: refs/heads/master@{#39637}
2016-09-22 16:34:16 +00:00
// size_t overloads
CHECK_EQ(Bytecodes::ScaleForUnsignedOperand(static_cast<size_t>(0)),
OperandScale::kSingle);
CHECK_EQ(Bytecodes::ScaleForUnsignedOperand(static_cast<size_t>(kMaxUInt8)),
OperandScale::kSingle);
[Interpreter] Optimize BytecodeArrayBuilder and BytecodeArrayWriter. This CL optimizes the code in BytecodeArrayBuilder and BytecodeArrayWriter by making the following main changes: - Move operand scale calculation out of BytecodeArrayWriter to the BytecodeNode constructor, where the decision on which operands are scalable can generally be statically decided by the compiler. - Move the maximum register calculation out of BytecodeArrayWriter and into BytecodeRegisterOptimizer (which is the only place outside BytecodeGenerator which updates which registers are used). This avoids the BytecodeArrayWriter needing to know the operand types of a node as it writes it. - Modify EmitBytecodes to use individual push_backs rather than building a buffer and calling insert, since this turns out to be faster. - Initialize BytecodeArrayWriter's bytecode vector by reserving 512 bytes, - Make common functions in Bytecodes constexpr so that they can be statically calculated by the compiler. - Move common functions and constructors in Bytecodes and BytecodeNode to the header so that they can be inlined. - Change large static switch statements in Bytecodes to const array lookups, and move to the header to allow inlining. I also took the opportunity to remove a number of unused helper functions, and rework some others for consistency. This reduces the percentage of time spent in making BytecodeArrays in CodeLoad from ~15% to ~11% according to perf. The CoadLoad score increase by around 2%. BUG=v8:4280 Committed: https://crrev.com/b11a8b4d41bf09d6b3d6cf214fe3fb61faf01a64 Review-Url: https://codereview.chromium.org/2351763002 Cr-Original-Commit-Position: refs/heads/master@{#39599} Cr-Commit-Position: refs/heads/master@{#39637}
2016-09-22 16:34:16 +00:00
CHECK(Bytecodes::ScaleForUnsignedOperand(
static_cast<size_t>(kMaxUInt8 + 1)) == OperandScale::kDouble);
CHECK_EQ(Bytecodes::ScaleForUnsignedOperand(static_cast<size_t>(kMaxUInt16)),
OperandScale::kDouble);
[Interpreter] Optimize BytecodeArrayBuilder and BytecodeArrayWriter. This CL optimizes the code in BytecodeArrayBuilder and BytecodeArrayWriter by making the following main changes: - Move operand scale calculation out of BytecodeArrayWriter to the BytecodeNode constructor, where the decision on which operands are scalable can generally be statically decided by the compiler. - Move the maximum register calculation out of BytecodeArrayWriter and into BytecodeRegisterOptimizer (which is the only place outside BytecodeGenerator which updates which registers are used). This avoids the BytecodeArrayWriter needing to know the operand types of a node as it writes it. - Modify EmitBytecodes to use individual push_backs rather than building a buffer and calling insert, since this turns out to be faster. - Initialize BytecodeArrayWriter's bytecode vector by reserving 512 bytes, - Make common functions in Bytecodes constexpr so that they can be statically calculated by the compiler. - Move common functions and constructors in Bytecodes and BytecodeNode to the header so that they can be inlined. - Change large static switch statements in Bytecodes to const array lookups, and move to the header to allow inlining. I also took the opportunity to remove a number of unused helper functions, and rework some others for consistency. This reduces the percentage of time spent in making BytecodeArrays in CodeLoad from ~15% to ~11% according to perf. The CoadLoad score increase by around 2%. BUG=v8:4280 Committed: https://crrev.com/b11a8b4d41bf09d6b3d6cf214fe3fb61faf01a64 Review-Url: https://codereview.chromium.org/2351763002 Cr-Original-Commit-Position: refs/heads/master@{#39599} Cr-Commit-Position: refs/heads/master@{#39637}
2016-09-22 16:34:16 +00:00
CHECK(Bytecodes::ScaleForUnsignedOperand(
static_cast<size_t>(kMaxUInt16 + 1)) == OperandScale::kQuadruple);
CHECK_EQ(Bytecodes::ScaleForUnsignedOperand(static_cast<size_t>(kMaxUInt32)),
OperandScale::kQuadruple);
}
TEST(Bytecodes, SizesForUnsignedOperands) {
// int overloads
CHECK_EQ(Bytecodes::SizeForUnsignedOperand(0), OperandSize::kByte);
CHECK_EQ(Bytecodes::SizeForUnsignedOperand(kMaxUInt8), OperandSize::kByte);
CHECK_EQ(Bytecodes::SizeForUnsignedOperand(kMaxUInt8 + 1),
OperandSize::kShort);
CHECK_EQ(Bytecodes::SizeForUnsignedOperand(kMaxUInt16), OperandSize::kShort);
CHECK_EQ(Bytecodes::SizeForUnsignedOperand(kMaxUInt16 + 1),
OperandSize::kQuad);
// size_t overloads
CHECK_EQ(Bytecodes::SizeForUnsignedOperand(static_cast<size_t>(0)),
OperandSize::kByte);
CHECK_EQ(Bytecodes::SizeForUnsignedOperand(static_cast<size_t>(kMaxUInt8)),
OperandSize::kByte);
CHECK_EQ(
Bytecodes::SizeForUnsignedOperand(static_cast<size_t>(kMaxUInt8 + 1)),
OperandSize::kShort);
CHECK_EQ(Bytecodes::SizeForUnsignedOperand(static_cast<size_t>(kMaxUInt16)),
OperandSize::kShort);
CHECK(Bytecodes::SizeForUnsignedOperand(
static_cast<size_t>(kMaxUInt16 + 1)) == OperandSize::kQuad);
CHECK_EQ(Bytecodes::SizeForUnsignedOperand(static_cast<size_t>(kMaxUInt32)),
OperandSize::kQuad);
}
// Helper macros to generate a check for if a bytecode is in a macro list of
// bytecodes. We can use these to exhaustively test a check over all bytecodes,
// both those that should pass and those that should fail the check.
#define OR_IS_BYTECODE(Name, ...) || bytecode == Bytecode::k##Name
#define IN_BYTECODE_LIST(BYTECODE, LIST) \
([](Bytecode bytecode) { return false LIST(OR_IS_BYTECODE); }(BYTECODE))
TEST(Bytecodes, IsJump) {
#define TEST_BYTECODE(Name, ...) \
if (IN_BYTECODE_LIST(Bytecode::k##Name, JUMP_BYTECODE_LIST)) { \
EXPECT_TRUE(Bytecodes::IsJump(Bytecode::k##Name)); \
} else { \
EXPECT_FALSE(Bytecodes::IsJump(Bytecode::k##Name)); \
}
BYTECODE_LIST(TEST_BYTECODE)
#undef TEST_BYTECODE
}
TEST(Bytecodes, IsForwardJump) {
#define TEST_BYTECODE(Name, ...) \
if (IN_BYTECODE_LIST(Bytecode::k##Name, JUMP_FORWARD_BYTECODE_LIST)) { \
EXPECT_TRUE(Bytecodes::IsForwardJump(Bytecode::k##Name)); \
} else { \
EXPECT_FALSE(Bytecodes::IsForwardJump(Bytecode::k##Name)); \
}
BYTECODE_LIST(TEST_BYTECODE)
#undef TEST_BYTECODE
}
TEST(Bytecodes, IsConditionalJump) {
#define TEST_BYTECODE(Name, ...) \
if (IN_BYTECODE_LIST(Bytecode::k##Name, JUMP_CONDITIONAL_BYTECODE_LIST)) { \
EXPECT_TRUE(Bytecodes::IsConditionalJump(Bytecode::k##Name)); \
} else { \
EXPECT_FALSE(Bytecodes::IsConditionalJump(Bytecode::k##Name)); \
}
BYTECODE_LIST(TEST_BYTECODE)
#undef TEST_BYTECODE
}
TEST(Bytecodes, IsUnconditionalJump) {
#define TEST_BYTECODE(Name, ...) \
if (IN_BYTECODE_LIST(Bytecode::k##Name, JUMP_UNCONDITIONAL_BYTECODE_LIST)) { \
EXPECT_TRUE(Bytecodes::IsUnconditionalJump(Bytecode::k##Name)); \
} else { \
EXPECT_FALSE(Bytecodes::IsUnconditionalJump(Bytecode::k##Name)); \
}
BYTECODE_LIST(TEST_BYTECODE)
#undef TEST_BYTECODE
}
TEST(Bytecodes, IsJumpImmediate) {
#define TEST_BYTECODE(Name, ...) \
if (IN_BYTECODE_LIST(Bytecode::k##Name, JUMP_IMMEDIATE_BYTECODE_LIST)) { \
EXPECT_TRUE(Bytecodes::IsJumpImmediate(Bytecode::k##Name)); \
} else { \
EXPECT_FALSE(Bytecodes::IsJumpImmediate(Bytecode::k##Name)); \
}
BYTECODE_LIST(TEST_BYTECODE)
#undef TEST_BYTECODE
}
TEST(Bytecodes, IsJumpConstant) {
#define TEST_BYTECODE(Name, ...) \
if (IN_BYTECODE_LIST(Bytecode::k##Name, JUMP_CONSTANT_BYTECODE_LIST)) { \
EXPECT_TRUE(Bytecodes::IsJumpConstant(Bytecode::k##Name)); \
} else { \
EXPECT_FALSE(Bytecodes::IsJumpConstant(Bytecode::k##Name)); \
}
BYTECODE_LIST(TEST_BYTECODE)
#undef TEST_BYTECODE
}
TEST(Bytecodes, IsConditionalJumpImmediate) {
#define TEST_BYTECODE(Name, ...) \
if (IN_BYTECODE_LIST(Bytecode::k##Name, JUMP_CONDITIONAL_BYTECODE_LIST) && \
IN_BYTECODE_LIST(Bytecode::k##Name, JUMP_IMMEDIATE_BYTECODE_LIST)) { \
EXPECT_TRUE(Bytecodes::IsConditionalJumpImmediate(Bytecode::k##Name)); \
} else { \
EXPECT_FALSE(Bytecodes::IsConditionalJumpImmediate(Bytecode::k##Name)); \
}
BYTECODE_LIST(TEST_BYTECODE)
#undef TEST_BYTECODE
}
TEST(Bytecodes, IsConditionalJumpConstant) {
#define TEST_BYTECODE(Name, ...) \
if (IN_BYTECODE_LIST(Bytecode::k##Name, JUMP_CONDITIONAL_BYTECODE_LIST) && \
IN_BYTECODE_LIST(Bytecode::k##Name, JUMP_CONSTANT_BYTECODE_LIST)) { \
EXPECT_TRUE(Bytecodes::IsConditionalJumpConstant(Bytecode::k##Name)); \
} else { \
EXPECT_FALSE(Bytecodes::IsConditionalJumpConstant(Bytecode::k##Name)); \
}
BYTECODE_LIST(TEST_BYTECODE)
#undef TEST_BYTECODE
}
TEST(Bytecodes, IsJumpIfToBoolean) {
#define TEST_BYTECODE(Name, ...) \
if (IN_BYTECODE_LIST(Bytecode::k##Name, JUMP_TO_BOOLEAN_BYTECODE_LIST)) { \
EXPECT_TRUE(Bytecodes::IsJumpIfToBoolean(Bytecode::k##Name)); \
} else { \
EXPECT_FALSE(Bytecodes::IsJumpIfToBoolean(Bytecode::k##Name)); \
}
BYTECODE_LIST(TEST_BYTECODE)
#undef TEST_BYTECODE
}
#undef OR_IS_BYTECODE
#undef IN_BYTECODE_LIST
TEST(OperandScale, PrefixesRequired) {
CHECK(!Bytecodes::OperandScaleRequiresPrefixBytecode(OperandScale::kSingle));
CHECK(Bytecodes::OperandScaleRequiresPrefixBytecode(OperandScale::kDouble));
CHECK(
Bytecodes::OperandScaleRequiresPrefixBytecode(OperandScale::kQuadruple));
CHECK_EQ(Bytecodes::OperandScaleToPrefixBytecode(OperandScale::kDouble),
Bytecode::kWide);
CHECK_EQ(Bytecodes::OperandScaleToPrefixBytecode(OperandScale::kQuadruple),
Bytecode::kExtraWide);
}
TEST(AccumulatorUse, LogicalOperators) {
CHECK_EQ(AccumulatorUse::kNone | AccumulatorUse::kRead,
AccumulatorUse::kRead);
CHECK_EQ(AccumulatorUse::kRead | AccumulatorUse::kWrite,
AccumulatorUse::kReadWrite);
CHECK_EQ(AccumulatorUse::kRead & AccumulatorUse::kReadWrite,
AccumulatorUse::kRead);
CHECK_EQ(AccumulatorUse::kRead & AccumulatorUse::kWrite,
AccumulatorUse::kNone);
}
TEST(AccumulatorUse, SampleBytecodes) {
CHECK(Bytecodes::ReadsAccumulator(Bytecode::kStar));
CHECK(!Bytecodes::WritesAccumulator(Bytecode::kStar));
CHECK_EQ(Bytecodes::GetAccumulatorUse(Bytecode::kStar),
AccumulatorUse::kRead);
CHECK(!Bytecodes::ReadsAccumulator(Bytecode::kLdar));
CHECK(Bytecodes::WritesAccumulator(Bytecode::kLdar));
CHECK_EQ(Bytecodes::GetAccumulatorUse(Bytecode::kLdar),
AccumulatorUse::kWrite);
CHECK(Bytecodes::ReadsAccumulator(Bytecode::kAdd));
CHECK(Bytecodes::WritesAccumulator(Bytecode::kAdd));
CHECK_EQ(Bytecodes::GetAccumulatorUse(Bytecode::kAdd),
AccumulatorUse::kReadWrite);
}
} // namespace interpreter
} // namespace internal
} // namespace v8