testmasm.cpp   [plain text]

/*
 * Copyright (C) 2017-2018 Apple Inc. All rights reserved.
 *
 * Redistribution and use in source and binary forms, with or without
 * modification, are permitted provided that the following conditions
 * are met:
 * 1. Redistributions of source code must retain the above copyright
 *    notice, this list of conditions and the following disclaimer.
 * 2. Redistributions in binary form must reproduce the above copyright
 *    notice, this list of conditions and the following disclaimer in the
 *    documentation and/or other materials provided with the distribution.
 *
 * THIS SOFTWARE IS PROVIDED BY APPLE INC. ``AS IS'' AND ANY
 * EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
 * PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL APPLE INC. OR
 * CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL,
 * EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO,
 * PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR
 * PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY
 * OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
 * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
 * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. 
 */

#include "config.h"

#include "CCallHelpers.h"
#include "CPU.h"
#include "FPRInfo.h"
#include "GPRInfo.h"
#include "InitializeThreading.h"
#include "LinkBuffer.h"
#include "ProbeContext.h"
#include "StackAlignment.h"
#include <limits>
#include <wtf/Compiler.h>
#include <wtf/DataLog.h>
#include <wtf/Function.h>
#include <wtf/Lock.h>
#include <wtf/NumberOfCores.h>
#include <wtf/Threading.h>

// We don't have a NO_RETURN_DUE_TO_EXIT, nor should we. That's ridiculous.
static bool hiddenTruthBecauseNoReturnIsStupid() { return true; }

static void usage()
{
    dataLog("Usage: testmasm [<filter>]\n");
    if (hiddenTruthBecauseNoReturnIsStupid())
        exit(1);
}

#if ENABLE(JIT)

#if ENABLE(MASM_PROBE)
namespace WTF {

static void printInternal(PrintStream& out, void* value)
{
    out.printf("%p", value);
}

} // namespace WTF
#endif // ENABLE(MASM_PROBE)

namespace JSC {
namespace Probe {

JS_EXPORT_PRIVATE void* probeStateForContext(Probe::Context&);

} // namespace Probe
} // namespace JSC

using namespace JSC;

namespace {

#if ENABLE(MASM_PROBE)
using CPUState = Probe::CPUState;
#endif

Lock crashLock;

typedef WTF::Function<void(CCallHelpers&)> Generator;

template<typename T> T nextID(T id) { return static_cast<T>(id + 1); }

#define TESTWORD64 0x0c0defefebeef000
#define TESTWORD32 0x0beef000

#define testWord32(x) (TESTWORD32 + static_cast<uint32_t>(x))
#define testWord64(x) (TESTWORD64 + static_cast<uint64_t>(x))

#if USE(JSVALUE64)
#define testWord(x) testWord64(x)
#else
#define testWord(x) testWord32(x)
#endif

// Nothing fancy for now; we just use the existing WTF assertion machinery.
#define CHECK_EQ(_actual, _expected) do {                               \
        if ((_actual) == (_expected))                                   \
            break;                                                      \
        crashLock.lock();                                               \
        dataLog("FAILED while testing " #_actual ": expected: ", _expected, ", actual: ", _actual, "\n"); \
        WTFReportAssertionFailure(__FILE__, __LINE__, WTF_PRETTY_FUNCTION, "CHECK_EQ("#_actual ", " #_expected ")"); \
        CRASH();                                                        \
    } while (false)

#if ENABLE(MASM_PROBE)
bool isPC(MacroAssembler::RegisterID id)
{
#if CPU(ARM_THUMB2) || CPU(ARM_TRADITIONAL)
    return id == ARMRegisters::pc;
#else
    UNUSED_PARAM(id);
    return false;
#endif
}

bool isSP(MacroAssembler::RegisterID id)
{
    return id == MacroAssembler::stackPointerRegister;
}

bool isFP(MacroAssembler::RegisterID id)
{
    return id == MacroAssembler::framePointerRegister;
}

bool isSpecialGPR(MacroAssembler::RegisterID id)
{
    if (isPC(id) || isSP(id) || isFP(id))
        return true;
#if CPU(ARM64)
    if (id == ARM64Registers::x18)
        return true;
#elif CPU(MIPS)
    if (id == MIPSRegisters::zero || id == MIPSRegisters::k0 || id == MIPSRegisters::k1)
        return true;
#endif
    return false;
}
#endif // ENABLE(MASM_PROBE)

MacroAssemblerCodeRef<JSEntryPtrTag> compile(Generator&& generate)
{
    CCallHelpers jit;
    generate(jit);
    LinkBuffer linkBuffer(jit, nullptr);
    return FINALIZE_CODE(linkBuffer, JSEntryPtrTag, "testmasm compilation");
}

template<typename T, typename... Arguments>
T invoke(MacroAssemblerCodeRef<JSEntryPtrTag> code, Arguments... arguments)
{
    void* executableAddress = untagCFunctionPtr<JSEntryPtrTag>(code.code().executableAddress());
    T (*function)(Arguments...) = bitwise_cast<T(*)(Arguments...)>(executableAddress);
    return function(arguments...);
}

template<typename T, typename... Arguments>
T compileAndRun(Generator&& generator, Arguments... arguments)
{
    return invoke<T>(compile(WTFMove(generator)), arguments...);
}

void testSimple()
{
    CHECK_EQ(compileAndRun<int>([] (CCallHelpers& jit) {
        jit.emitFunctionPrologue();
        jit.move(CCallHelpers::TrustedImm32(42), GPRInfo::returnValueGPR);
        jit.emitFunctionEpilogue();
        jit.ret();
    }), 42);
}

void testGetEffectiveAddress(size_t pointer, ptrdiff_t length, int32_t offset, CCallHelpers::Scale scale)
{
    CHECK_EQ(compileAndRun<size_t>([=] (CCallHelpers& jit) {
        jit.emitFunctionPrologue();
        jit.move(CCallHelpers::TrustedImmPtr(bitwise_cast<void*>(pointer)), GPRInfo::regT0);
        jit.move(CCallHelpers::TrustedImmPtr(bitwise_cast<void*>(length)), GPRInfo::regT1);
        jit.getEffectiveAddress(CCallHelpers::BaseIndex(GPRInfo::regT0, GPRInfo::regT1, scale, offset), GPRInfo::returnValueGPR);
        jit.emitFunctionEpilogue();
        jit.ret();
    }), pointer + offset + (static_cast<size_t>(1) << static_cast<int>(scale)) * length);
}

// branchTruncateDoubleToInt32(), when encountering Infinity, -Infinity or a
// Nan, should either yield 0 in dest or fail.
void testBranchTruncateDoubleToInt32(double val, int32_t expected)
{
    const uint64_t valAsUInt = *reinterpret_cast<uint64_t*>(&val);
#if CPU(BIG_ENDIAN)
    const bool isBigEndian = true;
#else
    const bool isBigEndian = false;
#endif
    CHECK_EQ(compileAndRun<int>([&] (CCallHelpers& jit) {
        jit.emitFunctionPrologue();
        jit.subPtr(CCallHelpers::TrustedImm32(stackAlignmentBytes()), MacroAssembler::stackPointerRegister);
        if (isBigEndian) {
            jit.store32(CCallHelpers::TrustedImm32(valAsUInt >> 32),
                MacroAssembler::stackPointerRegister);
            jit.store32(CCallHelpers::TrustedImm32(valAsUInt & 0xffffffff),
                MacroAssembler::Address(MacroAssembler::stackPointerRegister, 4));
        } else {
            jit.store32(CCallHelpers::TrustedImm32(valAsUInt & 0xffffffff),
                MacroAssembler::stackPointerRegister);
            jit.store32(CCallHelpers::TrustedImm32(valAsUInt >> 32),
                MacroAssembler::Address(MacroAssembler::stackPointerRegister, 4));
        }
        jit.loadDouble(MacroAssembler::stackPointerRegister, FPRInfo::fpRegT0);

        MacroAssembler::Jump done;
        done = jit.branchTruncateDoubleToInt32(FPRInfo::fpRegT0, GPRInfo::returnValueGPR, MacroAssembler::BranchIfTruncateSuccessful);

        jit.move(CCallHelpers::TrustedImm32(0), GPRInfo::returnValueGPR);

        done.link(&jit);
        jit.addPtr(CCallHelpers::TrustedImm32(stackAlignmentBytes()), MacroAssembler::stackPointerRegister);
        jit.emitFunctionEpilogue();
        jit.ret();
    }), expected);
}


static Vector<double> doubleOperands()
{
    return Vector<double> {
        0,
        -0,
        1,
        -1,
        42,
        -42,
        std::numeric_limits<double>::max(),
        std::numeric_limits<double>::min(),
        std::numeric_limits<double>::lowest(),
        std::numeric_limits<double>::quiet_NaN(),
        std::numeric_limits<double>::infinity(),
        -std::numeric_limits<double>::infinity(),
    };
}


static Vector<float> floatOperands()
{
    return Vector<float> {
        0,
        -0,
        1,
        -1,
        42,
        -42,
        std::numeric_limits<float>::max(),
        std::numeric_limits<float>::min(),
        std::numeric_limits<float>::lowest(),
        std::numeric_limits<float>::quiet_NaN(),
        std::numeric_limits<float>::infinity(),
        -std::numeric_limits<float>::infinity(),
    };
}


void testCompareDouble(MacroAssembler::DoubleCondition condition)
{
    double arg1 = 0;
    double arg2 = 0;

    auto compareDouble = compile([&, condition] (CCallHelpers& jit) {
        jit.emitFunctionPrologue();

        jit.loadDouble(CCallHelpers::TrustedImmPtr(&arg1), FPRInfo::fpRegT0);
        jit.loadDouble(CCallHelpers::TrustedImmPtr(&arg2), FPRInfo::fpRegT1);
        jit.move(CCallHelpers::TrustedImm32(-1), GPRInfo::returnValueGPR);
        jit.compareDouble(condition, FPRInfo::fpRegT0, FPRInfo::fpRegT1, GPRInfo::returnValueGPR);

        jit.emitFunctionEpilogue();
        jit.ret();
    });

    auto compareDoubleGeneric = compile([&, condition] (CCallHelpers& jit) {
        jit.emitFunctionPrologue();

        jit.loadDouble(CCallHelpers::TrustedImmPtr(&arg1), FPRInfo::fpRegT0);
        jit.loadDouble(CCallHelpers::TrustedImmPtr(&arg2), FPRInfo::fpRegT1);
        jit.move(CCallHelpers::TrustedImm32(1), GPRInfo::returnValueGPR);
        auto jump = jit.branchDouble(condition, FPRInfo::fpRegT0, FPRInfo::fpRegT1);
        jit.move(CCallHelpers::TrustedImm32(0), GPRInfo::returnValueGPR);
        jump.link(&jit);

        jit.emitFunctionEpilogue();
        jit.ret();
    });

    auto operands = doubleOperands();
    for (auto a : operands) {
        for (auto b : operands) {
            arg1 = a;
            arg2 = b;
            CHECK_EQ(invoke<int>(compareDouble), invoke<int>(compareDoubleGeneric));
        }
    }
}

#if CPU(X86) || CPU(X86_64) || CPU(ARM64)
void testCompareFloat(MacroAssembler::DoubleCondition condition)
{
    float arg1 = 0;
    float arg2 = 0;

    auto compareFloat = compile([&, condition] (CCallHelpers& jit) {
        jit.emitFunctionPrologue();

        jit.loadFloat(CCallHelpers::TrustedImmPtr(&arg1), FPRInfo::fpRegT0);
        jit.loadFloat(CCallHelpers::TrustedImmPtr(&arg2), FPRInfo::fpRegT1);
        jit.move(CCallHelpers::TrustedImm32(-1), GPRInfo::returnValueGPR);
        jit.compareFloat(condition, FPRInfo::fpRegT0, FPRInfo::fpRegT1, GPRInfo::returnValueGPR);

        jit.emitFunctionEpilogue();
        jit.ret();
    });

    auto compareFloatGeneric = compile([&, condition] (CCallHelpers& jit) {
        jit.emitFunctionPrologue();

        jit.loadFloat(CCallHelpers::TrustedImmPtr(&arg1), FPRInfo::fpRegT0);
        jit.loadFloat(CCallHelpers::TrustedImmPtr(&arg2), FPRInfo::fpRegT1);
        jit.move(CCallHelpers::TrustedImm32(1), GPRInfo::returnValueGPR);
        auto jump = jit.branchFloat(condition, FPRInfo::fpRegT0, FPRInfo::fpRegT1);
        jit.move(CCallHelpers::TrustedImm32(0), GPRInfo::returnValueGPR);
        jump.link(&jit);

        jit.emitFunctionEpilogue();
        jit.ret();
    });

    auto operands = floatOperands();
    for (auto a : operands) {
        for (auto b : operands) {
            arg1 = a;
            arg2 = b;
            CHECK_EQ(invoke<int>(compareFloat), invoke<int>(compareFloatGeneric));
        }
    }
}
#endif

#if ENABLE(MASM_PROBE)
void testProbeReadsArgumentRegisters()
{
    bool probeWasCalled = false;
    compileAndRun<void>([&] (CCallHelpers& jit) {
        jit.emitFunctionPrologue();

        jit.pushPair(GPRInfo::argumentGPR0, GPRInfo::argumentGPR1);
        jit.pushPair(GPRInfo::argumentGPR2, GPRInfo::argumentGPR3);

        jit.move(CCallHelpers::TrustedImm32(testWord32(0)), GPRInfo::argumentGPR0);
        jit.convertInt32ToDouble(GPRInfo::argumentGPR0, FPRInfo::fpRegT0);
        jit.move(CCallHelpers::TrustedImm32(testWord32(1)), GPRInfo::argumentGPR0);
        jit.convertInt32ToDouble(GPRInfo::argumentGPR0, FPRInfo::fpRegT1);
#if USE(JSVALUE64)
        jit.move(CCallHelpers::TrustedImm64(testWord(0)), GPRInfo::argumentGPR0);
        jit.move(CCallHelpers::TrustedImm64(testWord(1)), GPRInfo::argumentGPR1);
        jit.move(CCallHelpers::TrustedImm64(testWord(2)), GPRInfo::argumentGPR2);
        jit.move(CCallHelpers::TrustedImm64(testWord(3)), GPRInfo::argumentGPR3);
#else
        jit.move(CCallHelpers::TrustedImm32(testWord(0)), GPRInfo::argumentGPR0);
        jit.move(CCallHelpers::TrustedImm32(testWord(1)), GPRInfo::argumentGPR1);
        jit.move(CCallHelpers::TrustedImm32(testWord(2)), GPRInfo::argumentGPR2);
        jit.move(CCallHelpers::TrustedImm32(testWord(3)), GPRInfo::argumentGPR3);
#endif

        jit.probe([&] (Probe::Context& context) {
            auto& cpu = context.cpu;
            probeWasCalled = true;
            CHECK_EQ(cpu.gpr(GPRInfo::argumentGPR0), testWord(0));
            CHECK_EQ(cpu.gpr(GPRInfo::argumentGPR1), testWord(1));
            CHECK_EQ(cpu.gpr(GPRInfo::argumentGPR2), testWord(2));
            CHECK_EQ(cpu.gpr(GPRInfo::argumentGPR3), testWord(3));

            CHECK_EQ(cpu.fpr(FPRInfo::fpRegT0), testWord32(0));
            CHECK_EQ(cpu.fpr(FPRInfo::fpRegT1), testWord32(1));
        });

        jit.popPair(GPRInfo::argumentGPR2, GPRInfo::argumentGPR3);
        jit.popPair(GPRInfo::argumentGPR0, GPRInfo::argumentGPR1);

        jit.emitFunctionEpilogue();
        jit.ret();
    });
    CHECK_EQ(probeWasCalled, true);
}

void testProbeWritesArgumentRegisters()
{
    // This test relies on testProbeReadsArgumentRegisters() having already validated
    // that we can read from argument registers. We'll use that ability to validate
    // that our writes did take effect.
    unsigned probeCallCount = 0;
    compileAndRun<void>([&] (CCallHelpers& jit) {
        jit.emitFunctionPrologue();

        jit.pushPair(GPRInfo::argumentGPR0, GPRInfo::argumentGPR1);
        jit.pushPair(GPRInfo::argumentGPR2, GPRInfo::argumentGPR3);

        // Pre-initialize with non-expected values.
#if USE(JSVALUE64)
        jit.move(CCallHelpers::TrustedImm64(0), GPRInfo::argumentGPR0);
        jit.move(CCallHelpers::TrustedImm64(0), GPRInfo::argumentGPR1);
        jit.move(CCallHelpers::TrustedImm64(0), GPRInfo::argumentGPR2);
        jit.move(CCallHelpers::TrustedImm64(0), GPRInfo::argumentGPR3);
#else
        jit.move(CCallHelpers::TrustedImm32(0), GPRInfo::argumentGPR0);
        jit.move(CCallHelpers::TrustedImm32(0), GPRInfo::argumentGPR1);
        jit.move(CCallHelpers::TrustedImm32(0), GPRInfo::argumentGPR2);
        jit.move(CCallHelpers::TrustedImm32(0), GPRInfo::argumentGPR3);
#endif
        jit.convertInt32ToDouble(GPRInfo::argumentGPR0, FPRInfo::fpRegT0);
        jit.convertInt32ToDouble(GPRInfo::argumentGPR0, FPRInfo::fpRegT1);

        // Write expected values.
        jit.probe([&] (Probe::Context& context) {
            auto& cpu = context.cpu;
            probeCallCount++;
            cpu.gpr(GPRInfo::argumentGPR0) = testWord(0);
            cpu.gpr(GPRInfo::argumentGPR1) = testWord(1);
            cpu.gpr(GPRInfo::argumentGPR2) = testWord(2);
            cpu.gpr(GPRInfo::argumentGPR3) = testWord(3);
            
            cpu.fpr(FPRInfo::fpRegT0) = bitwise_cast<double>(testWord64(0));
            cpu.fpr(FPRInfo::fpRegT1) = bitwise_cast<double>(testWord64(1));
        });

        // Validate that expected values were written.
        jit.probe([&] (Probe::Context& context) {
            auto& cpu = context.cpu;
            probeCallCount++;
            CHECK_EQ(cpu.gpr(GPRInfo::argumentGPR0), testWord(0));
            CHECK_EQ(cpu.gpr(GPRInfo::argumentGPR1), testWord(1));
            CHECK_EQ(cpu.gpr(GPRInfo::argumentGPR2), testWord(2));
            CHECK_EQ(cpu.gpr(GPRInfo::argumentGPR3), testWord(3));

            CHECK_EQ(cpu.fpr<uint64_t>(FPRInfo::fpRegT0), testWord64(0));
            CHECK_EQ(cpu.fpr<uint64_t>(FPRInfo::fpRegT1), testWord64(1));
        });

        jit.popPair(GPRInfo::argumentGPR2, GPRInfo::argumentGPR3);
        jit.popPair(GPRInfo::argumentGPR0, GPRInfo::argumentGPR1);

        jit.emitFunctionEpilogue();
        jit.ret();
    });
    CHECK_EQ(probeCallCount, 2);
}

static NEVER_INLINE NOT_TAIL_CALLED int testFunctionToTrashGPRs(int a, int b, int c, int d, int e, int f, int g, int h, int i, int j)
{
    if (j > 0)
        return testFunctionToTrashGPRs(a + 1, b + a, c + b, d + 5, e - a, f * 1.5, g ^ a, h - b, i, j - 1);
    return a + 1;
}
static NEVER_INLINE NOT_TAIL_CALLED double testFunctionToTrashFPRs(double a, double b, double c, double d, double e, double f, double g, double h, double i, double j)
{
    if (j > 0)
        return testFunctionToTrashFPRs(a + 1, b + a, c + b, d + 5, e - a, f * 1.5, pow(g, a), h - b, i, j - 1);
    return a + 1;
}

void testProbePreservesGPRS()
{
    // This test relies on testProbeReadsArgumentRegisters() and testProbeWritesArgumentRegisters()
    // having already validated that we can read and write from registers. We'll use these abilities
    // to validate that the probe preserves register values.
    unsigned probeCallCount = 0;
    CPUState originalState;

    compileAndRun<void>([&] (CCallHelpers& jit) {
        jit.emitFunctionPrologue();

        // Write expected values into the registers (except for sp, fp, and pc).
        jit.probe([&] (Probe::Context& context) {
            auto& cpu = context.cpu;
            probeCallCount++;
            for (auto id = CCallHelpers::firstRegister(); id <= CCallHelpers::lastRegister(); id = nextID(id)) {
                originalState.gpr(id) = cpu.gpr(id);
                if (isSpecialGPR(id))
                    continue;
                cpu.gpr(id) = testWord(static_cast<int>(id));
            }
            for (auto id = CCallHelpers::firstFPRegister(); id <= CCallHelpers::lastFPRegister(); id = nextID(id)) {
                originalState.fpr(id) = cpu.fpr(id);
                cpu.fpr(id) = bitwise_cast<double>(testWord64(id));
            }
        });

        // Invoke the probe to call a lot of functions and trash register values.
        jit.probe([&] (Probe::Context&) {
            probeCallCount++;
            CHECK_EQ(testFunctionToTrashGPRs(0, 1, 2, 3, 4, 5, 6, 7, 8, 9), 10);
            CHECK_EQ(testFunctionToTrashFPRs(0, 1, 2, 3, 4, 5, 6, 7, 8, 9), 10);
        });

        // Validate that the registers have the expected values.
        jit.probe([&] (Probe::Context& context) {
            auto& cpu = context.cpu;
            probeCallCount++;
            for (auto id = CCallHelpers::firstRegister(); id <= CCallHelpers::lastRegister(); id = nextID(id)) {
                if (isSP(id) || isFP(id)) {
                    CHECK_EQ(cpu.gpr(id), originalState.gpr(id));
                    continue;
                }
                if (isSpecialGPR(id))
                    continue;
                CHECK_EQ(cpu.gpr(id), testWord(id));
            }
            for (auto id = CCallHelpers::firstFPRegister(); id <= CCallHelpers::lastFPRegister(); id = nextID(id))
#if CPU(MIPS)
                if (!(id & 1))
#endif
                CHECK_EQ(cpu.fpr<uint64_t>(id), testWord64(id));
        });

        // Restore the original state.
        jit.probe([&] (Probe::Context& context) {
            auto& cpu = context.cpu;
            probeCallCount++;
            for (auto id = CCallHelpers::firstRegister(); id <= CCallHelpers::lastRegister(); id = nextID(id)) {
                if (isSpecialGPR(id))
                    continue;
                cpu.gpr(id) = originalState.gpr(id);
            }
            for (auto id = CCallHelpers::firstFPRegister(); id <= CCallHelpers::lastFPRegister(); id = nextID(id))
                cpu.fpr(id) = originalState.fpr(id);
        });

        // Validate that the original state was restored.
        jit.probe([&] (Probe::Context& context) {
            auto& cpu = context.cpu;
            probeCallCount++;
            for (auto id = CCallHelpers::firstRegister(); id <= CCallHelpers::lastRegister(); id = nextID(id)) {
                if (isSpecialGPR(id))
                    continue;
                CHECK_EQ(cpu.gpr(id), originalState.gpr(id));
            }
            for (auto id = CCallHelpers::firstFPRegister(); id <= CCallHelpers::lastFPRegister(); id = nextID(id))
#if CPU(MIPS)
                if (!(id & 1))
#endif
                CHECK_EQ(cpu.fpr<uint64_t>(id), originalState.fpr<uint64_t>(id));
        });

        jit.emitFunctionEpilogue();
        jit.ret();
    });
    CHECK_EQ(probeCallCount, 5);
}

void testProbeModifiesStackPointer(WTF::Function<void*(Probe::Context&)> computeModifiedStackPointer)
{
    unsigned probeCallCount = 0;
    CPUState originalState;
    void* originalSP { nullptr };
    void* modifiedSP { nullptr };
#if !(CPU(MIPS))
    uintptr_t modifiedFlags { 0 };
#endif
    
#if CPU(X86) || CPU(X86_64)
    auto flagsSPR = X86Registers::eflags;
    uintptr_t flagsMask = 0xc5;
#elif CPU(ARM_THUMB2) || CPU(ARM_TRADITIONAL)
    auto flagsSPR = ARMRegisters::apsr;
    uintptr_t flagsMask = 0xf8000000;
#elif CPU(ARM64)
    auto flagsSPR = ARM64Registers::nzcv;
    uintptr_t flagsMask = 0xf0000000;
#endif

    compileAndRun<void>([&] (CCallHelpers& jit) {
        jit.emitFunctionPrologue();

        // Preserve original stack pointer and modify the sp, and
        // write expected values into other registers (except for fp, and pc).
        jit.probe([&] (Probe::Context& context) {
            auto& cpu = context.cpu;
            probeCallCount++;
            for (auto id = CCallHelpers::firstRegister(); id <= CCallHelpers::lastRegister(); id = nextID(id)) {
                originalState.gpr(id) = cpu.gpr(id);
                if (isSpecialGPR(id))
                    continue;
                cpu.gpr(id) = testWord(static_cast<int>(id));
            }
            for (auto id = CCallHelpers::firstFPRegister(); id <= CCallHelpers::lastFPRegister(); id = nextID(id)) {
                originalState.fpr(id) = cpu.fpr(id);
                cpu.fpr(id) = bitwise_cast<double>(testWord64(id));
            }

#if !(CPU(MIPS))
            originalState.spr(flagsSPR) = cpu.spr(flagsSPR);
            modifiedFlags = originalState.spr(flagsSPR) ^ flagsMask;
            cpu.spr(flagsSPR) = modifiedFlags;
#endif

            originalSP = cpu.sp();
            modifiedSP = computeModifiedStackPointer(context);
            cpu.sp() = modifiedSP;
        });

        // Validate that the registers have the expected values.
        jit.probe([&] (Probe::Context& context) {
            auto& cpu = context.cpu;
            probeCallCount++;
            for (auto id = CCallHelpers::firstRegister(); id <= CCallHelpers::lastRegister(); id = nextID(id)) {
                if (isFP(id)) {
                    CHECK_EQ(cpu.gpr(id), originalState.gpr(id));
                    continue;
                }
                if (isSpecialGPR(id))
                    continue;
                CHECK_EQ(cpu.gpr(id), testWord(id));
            }
            for (auto id = CCallHelpers::firstFPRegister(); id <= CCallHelpers::lastFPRegister(); id = nextID(id))
#if CPU(MIPS)
                if (!(id & 1))
#endif
                CHECK_EQ(cpu.fpr<uint64_t>(id), testWord64(id));
#if !(CPU(MIPS))
            CHECK_EQ(cpu.spr(flagsSPR) & flagsMask, modifiedFlags & flagsMask);
#endif
            CHECK_EQ(cpu.sp(), modifiedSP);
        });

        // Restore the original state.
        jit.probe([&] (Probe::Context& context) {
            auto& cpu = context.cpu;
            probeCallCount++;
            for (auto id = CCallHelpers::firstRegister(); id <= CCallHelpers::lastRegister(); id = nextID(id)) {
                if (isSpecialGPR(id))
                    continue;
                cpu.gpr(id) = originalState.gpr(id);
            }
            for (auto id = CCallHelpers::firstFPRegister(); id <= CCallHelpers::lastFPRegister(); id = nextID(id))
                cpu.fpr(id) = originalState.fpr(id);
#if !(CPU(MIPS))
            cpu.spr(flagsSPR) = originalState.spr(flagsSPR);
#endif
            cpu.sp() = originalSP;
        });

        // Validate that the original state was restored.
        jit.probe([&] (Probe::Context& context) {
            auto& cpu = context.cpu;
            probeCallCount++;
            for (auto id = CCallHelpers::firstRegister(); id <= CCallHelpers::lastRegister(); id = nextID(id)) {
                if (isSpecialGPR(id))
                    continue;
                CHECK_EQ(cpu.gpr(id), originalState.gpr(id));
            }
            for (auto id = CCallHelpers::firstFPRegister(); id <= CCallHelpers::lastFPRegister(); id = nextID(id))
#if CPU(MIPS)
                if (!(id & 1))
#endif
                CHECK_EQ(cpu.fpr<uint64_t>(id), originalState.fpr<uint64_t>(id));
#if !(CPU(MIPS))
            CHECK_EQ(cpu.spr(flagsSPR) & flagsMask, originalState.spr(flagsSPR) & flagsMask);
#endif
            CHECK_EQ(cpu.sp(), originalSP);
        });

        jit.emitFunctionEpilogue();
        jit.ret();
    });
    CHECK_EQ(probeCallCount, 4);
}

void testProbeModifiesStackPointerToInsideProbeStateOnStack()
{
    size_t increment = sizeof(uintptr_t);
#if CPU(ARM64)
    // The ARM64 probe uses ldp and stp which require 16 byte alignment.
    increment = 2 * sizeof(uintptr_t);
#endif
    for (size_t offset = 0; offset < sizeof(Probe::State); offset += increment) {
        testProbeModifiesStackPointer([=] (Probe::Context& context) -> void* {
            return reinterpret_cast<uint8_t*>(probeStateForContext(context)) + offset;

        });
    }
}

void testProbeModifiesStackPointerToNBytesBelowSP()
{
    size_t increment = sizeof(uintptr_t);
#if CPU(ARM64)
    // The ARM64 probe uses ldp and stp which require 16 byte alignment.
    increment = 2 * sizeof(uintptr_t);
#endif
    for (size_t offset = 0; offset < 1 * KB; offset += increment) {
        testProbeModifiesStackPointer([=] (Probe::Context& context) -> void* {
            return context.cpu.sp<uint8_t*>() - offset;
        });
    }
}

void testProbeModifiesProgramCounter()
{
    // This test relies on testProbeReadsArgumentRegisters() and testProbeWritesArgumentRegisters()
    // having already validated that we can read and write from registers. We'll use these abilities
    // to validate that the probe preserves register values.
    unsigned probeCallCount = 0;
    bool continuationWasReached = false;

    MacroAssemblerCodeRef<JSEntryPtrTag> continuation = compile([&] (CCallHelpers& jit) {
        // Validate that we reached the continuation.
        jit.probe([&] (Probe::Context&) {
            probeCallCount++;
            continuationWasReached = true;
        });

        jit.emitFunctionEpilogue();
        jit.ret();
    });

    compileAndRun<void>([&] (CCallHelpers& jit) {
        jit.emitFunctionPrologue();

        // Write expected values into the registers.
        jit.probe([&] (Probe::Context& context) {
            probeCallCount++;
            context.cpu.pc() = untagCodePtr(continuation.code().executableAddress(), JSEntryPtrTag);
        });

        jit.breakpoint(); // We should never get here.
    });
    CHECK_EQ(probeCallCount, 2);
    CHECK_EQ(continuationWasReached, true);
}

void testProbeModifiesStackValues()
{
    unsigned probeCallCount = 0;
    CPUState originalState;
    void* originalSP { nullptr };
    void* newSP { nullptr };
    uintptr_t modifiedFlags { 0 };
    size_t numberOfExtraEntriesToWrite { 10 }; // ARM64 requires that this be 2 word aligned.

#if CPU(X86) || CPU(X86_64)
    MacroAssembler::SPRegisterID flagsSPR = X86Registers::eflags;
    uintptr_t flagsMask = 0xc5;
#elif CPU(ARM_THUMB2) || CPU(ARM_TRADITIONAL)
    MacroAssembler::SPRegisterID flagsSPR = ARMRegisters::apsr;
    uintptr_t flagsMask = 0xf8000000;
#elif CPU(ARM64)
    MacroAssembler::SPRegisterID flagsSPR = ARM64Registers::nzcv;
    uintptr_t flagsMask = 0xf0000000;
#endif

    compileAndRun<void>([&] (CCallHelpers& jit) {
        jit.emitFunctionPrologue();

        // Write expected values into the registers.
        jit.probe([&] (Probe::Context& context) {
            auto& cpu = context.cpu;
            auto& stack = context.stack();
            probeCallCount++;

            // Preserve the original CPU state.
            for (auto id = CCallHelpers::firstRegister(); id <= CCallHelpers::lastRegister(); id = nextID(id)) {
                originalState.gpr(id) = cpu.gpr(id);
                if (isSpecialGPR(id))
                    continue;
                cpu.gpr(id) = testWord(static_cast<int>(id));
            }
            for (auto id = CCallHelpers::firstFPRegister(); id <= CCallHelpers::lastFPRegister(); id = nextID(id)) {
                originalState.fpr(id) = cpu.fpr(id);
                cpu.fpr(id) = bitwise_cast<double>(testWord64(id));
            }
#if !(CPU(MIPS))
            originalState.spr(flagsSPR) = cpu.spr(flagsSPR);
            modifiedFlags = originalState.spr(flagsSPR) ^ flagsMask;
            cpu.spr(flagsSPR) = modifiedFlags;
#endif

            // Ensure that we'll be writing over the regions of the stack where the Probe::State is.
            originalSP = cpu.sp();
            newSP = reinterpret_cast<uintptr_t*>(probeStateForContext(context)) - numberOfExtraEntriesToWrite;
            cpu.sp() = newSP;

            // Fill the stack with values.
            uintptr_t* p = reinterpret_cast<uintptr_t*>(newSP);
            int count = 0;
            stack.set<double>(p++, 1.234567);
            if (is32Bit())
                p++; // On 32-bit targets, a double takes up 2 uintptr_t.
            while (p < reinterpret_cast<uintptr_t*>(originalSP))
                stack.set<uintptr_t>(p++, testWord(count++));
        });

        // Validate that the registers and stack have the expected values.
        jit.probe([&] (Probe::Context& context) {
            auto& cpu = context.cpu;
            auto& stack = context.stack();
            probeCallCount++;

            // Validate the register values.
            for (auto id = CCallHelpers::firstRegister(); id <= CCallHelpers::lastRegister(); id = nextID(id)) {
                if (isFP(id)) {
                    CHECK_EQ(cpu.gpr(id), originalState.gpr(id));
                    continue;
                }
                if (isSpecialGPR(id))
                    continue;
                CHECK_EQ(cpu.gpr(id), testWord(id));
            }
            for (auto id = CCallHelpers::firstFPRegister(); id <= CCallHelpers::lastFPRegister(); id = nextID(id))
#if CPU(MIPS)
                if (!(id & 1))
#endif
                CHECK_EQ(cpu.fpr<uint64_t>(id), testWord64(id));
#if !(CPU(MIPS))
            CHECK_EQ(cpu.spr(flagsSPR) & flagsMask, modifiedFlags & flagsMask);
#endif
            CHECK_EQ(cpu.sp(), newSP);

            // Validate the stack values.
            uintptr_t* p = reinterpret_cast<uintptr_t*>(newSP);
            int count = 0;
            CHECK_EQ(stack.get<double>(p++), 1.234567);
            if (is32Bit())
                p++; // On 32-bit targets, a double takes up 2 uintptr_t.
            while (p < reinterpret_cast<uintptr_t*>(originalSP))
                CHECK_EQ(stack.get<uintptr_t>(p++), testWord(count++));
        });

        // Restore the original state.
        jit.probe([&] (Probe::Context& context) {
            auto& cpu = context.cpu;
            probeCallCount++;
            for (auto id = CCallHelpers::firstRegister(); id <= CCallHelpers::lastRegister(); id = nextID(id)) {
                if (isSpecialGPR(id))
                    continue;
                cpu.gpr(id) = originalState.gpr(id);
            }
            for (auto id = CCallHelpers::firstFPRegister(); id <= CCallHelpers::lastFPRegister(); id = nextID(id))
                cpu.fpr(id) = originalState.fpr(id);
#if !(CPU(MIPS))
            cpu.spr(flagsSPR) = originalState.spr(flagsSPR);
#endif
            cpu.sp() = originalSP;
        });

        jit.emitFunctionEpilogue();
        jit.ret();
    });

    CHECK_EQ(probeCallCount, 3);
}
#endif // ENABLE(MASM_PROBE)

#define RUN(test) do {                          \
        if (!shouldRun(#test))                  \
            break;                              \
        numberOfTests++;                        \
        tasks.append(                           \
            createSharedTask<void()>(           \
                [&] () {                        \
                    dataLog(#test "...\n");     \
                    test;                       \
                    dataLog(#test ": OK!\n");   \
                }));                            \
    } while (false);

void run(const char* filter)
{
    JSC::initializeThreading();
    unsigned numberOfTests = 0;

    Deque<RefPtr<SharedTask<void()>>> tasks;

    auto shouldRun = [&] (const char* testName) -> bool {
#if OS(UNIX)
        return !filter || !!strcasestr(testName, filter);
#else
        return !filter || !!strstr(testName, filter);
#endif
    };

    RUN(testSimple());
    RUN(testGetEffectiveAddress(0xff00, 42, 8, CCallHelpers::TimesEight));
    RUN(testGetEffectiveAddress(0xff00, -200, -300, CCallHelpers::TimesEight));
    RUN(testBranchTruncateDoubleToInt32(0, 0));
    RUN(testBranchTruncateDoubleToInt32(42, 42));
    RUN(testBranchTruncateDoubleToInt32(42.7, 42));
    RUN(testBranchTruncateDoubleToInt32(-1234, -1234));
    RUN(testBranchTruncateDoubleToInt32(-1234.56, -1234));
    RUN(testBranchTruncateDoubleToInt32(std::numeric_limits<double>::infinity(), 0));
    RUN(testBranchTruncateDoubleToInt32(-std::numeric_limits<double>::infinity(), 0));
    RUN(testBranchTruncateDoubleToInt32(std::numeric_limits<double>::quiet_NaN(), 0));
    RUN(testBranchTruncateDoubleToInt32(std::numeric_limits<double>::signaling_NaN(), 0));
    RUN(testBranchTruncateDoubleToInt32(std::numeric_limits<double>::max(), 0));
    RUN(testBranchTruncateDoubleToInt32(-std::numeric_limits<double>::max(), 0));
    // We run this last one to make sure that we don't use flags that were not
    // reset to check a conversion result
    RUN(testBranchTruncateDoubleToInt32(123, 123));

    RUN(testCompareDouble(MacroAssembler::DoubleEqual));
    RUN(testCompareDouble(MacroAssembler::DoubleNotEqual));
    RUN(testCompareDouble(MacroAssembler::DoubleGreaterThan));
    RUN(testCompareDouble(MacroAssembler::DoubleGreaterThanOrEqual));
    RUN(testCompareDouble(MacroAssembler::DoubleLessThan));
    RUN(testCompareDouble(MacroAssembler::DoubleLessThanOrEqual));
    RUN(testCompareDouble(MacroAssembler::DoubleEqualOrUnordered));
    RUN(testCompareDouble(MacroAssembler::DoubleNotEqualOrUnordered));
    RUN(testCompareDouble(MacroAssembler::DoubleGreaterThanOrUnordered));
    RUN(testCompareDouble(MacroAssembler::DoubleGreaterThanOrEqualOrUnordered));
    RUN(testCompareDouble(MacroAssembler::DoubleLessThanOrUnordered));
    RUN(testCompareDouble(MacroAssembler::DoubleLessThanOrEqualOrUnordered));

#if CPU(X86) || CPU(X86_64) || CPU(ARM64)
    RUN(testCompareFloat(MacroAssembler::DoubleEqual));
    RUN(testCompareFloat(MacroAssembler::DoubleNotEqual));
    RUN(testCompareFloat(MacroAssembler::DoubleGreaterThan));
    RUN(testCompareFloat(MacroAssembler::DoubleGreaterThanOrEqual));
    RUN(testCompareFloat(MacroAssembler::DoubleLessThan));
    RUN(testCompareFloat(MacroAssembler::DoubleLessThanOrEqual));
    RUN(testCompareFloat(MacroAssembler::DoubleEqualOrUnordered));
    RUN(testCompareFloat(MacroAssembler::DoubleNotEqualOrUnordered));
    RUN(testCompareFloat(MacroAssembler::DoubleGreaterThanOrUnordered));
    RUN(testCompareFloat(MacroAssembler::DoubleGreaterThanOrEqualOrUnordered));
    RUN(testCompareFloat(MacroAssembler::DoubleLessThanOrUnordered));
    RUN(testCompareFloat(MacroAssembler::DoubleLessThanOrEqualOrUnordered));
#endif

#if ENABLE(MASM_PROBE)
    RUN(testProbeReadsArgumentRegisters());
    RUN(testProbeWritesArgumentRegisters());
    RUN(testProbePreservesGPRS());
    RUN(testProbeModifiesStackPointerToInsideProbeStateOnStack());
    RUN(testProbeModifiesStackPointerToNBytesBelowSP());
    RUN(testProbeModifiesProgramCounter());
    RUN(testProbeModifiesStackValues());
#endif // ENABLE(MASM_PROBE)

    if (tasks.isEmpty())
        usage();

    Lock lock;

    Vector<Ref<Thread>> threads;
    for (unsigned i = filter ? 1 : WTF::numberOfProcessorCores(); i--;) {
        threads.append(
            Thread::create(
                "testmasm thread",
                [&] () {
                    for (;;) {
                        RefPtr<SharedTask<void()>> task;
                        {
                            LockHolder locker(lock);
                            if (tasks.isEmpty())
                                return;
                            task = tasks.takeFirst();
                        }

                        task->run();
                    }
                }));
    }

    for (auto& thread : threads)
        thread->waitForCompletion();
    crashLock.lock();
    dataLog("Completed ", numberOfTests, " tests\n");
}

} // anonymous namespace

#else // not ENABLE(JIT)

static void run(const char*)
{
    dataLog("JIT is not enabled.\n");
}

#endif // ENABLE(JIT)

int main(int argc, char** argv)
{
    const char* filter = nullptr;
    switch (argc) {
    case 1:
        break;
    case 2:
        filter = argv[1];
        break;
    default:
        usage();
        break;
    }

    run(filter);
    return 0;
}

#if OS(WINDOWS)
extern "C" __declspec(dllexport) int WINAPI dllLauncherEntryPoint(int argc, const char* argv[])
{
    return main(argc, const_cast<char**>(argv));
}
#endif