JITMathIC.h   [plain text]

/*
 * Copyright (C) 2016-2018 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. 
 */

#pragma once

#if ENABLE(JIT)

#include "ArithProfile.h"
#include "CCallHelpers.h"
#include "JITAddGenerator.h"
#include "JITMathICInlineResult.h"
#include "JITMulGenerator.h"
#include "JITNegGenerator.h"
#include "JITSubGenerator.h"
#include "LinkBuffer.h"
#include "Repatch.h"

namespace JSC {

class LinkBuffer;

struct MathICGenerationState {
    MacroAssembler::Label fastPathStart;
    MacroAssembler::Label fastPathEnd;
    MacroAssembler::Label slowPathStart;
    MacroAssembler::Call slowPathCall;
    MacroAssembler::JumpList slowPathJumps;
    bool shouldSlowPathRepatch;
};

#define ENABLE_MATH_IC_STATS 0

template <typename GeneratorType, bool(*isProfileEmpty)(ArithProfile&)>
class JITMathIC {
    WTF_MAKE_FAST_ALLOCATED;
public:
    JITMathIC(ArithProfile* arithProfile, Instruction* instruction)
        : m_arithProfile(arithProfile)
        , m_instruction(instruction)
    {
    }

    CodeLocationLabel<JSInternalPtrTag> doneLocation() { return m_inlineEnd; }
    CodeLocationCall<JSInternalPtrTag> slowPathCallLocation() { return m_slowPathCallLocation; }
    CodeLocationLabel<JSInternalPtrTag> slowPathStartLocation() { return m_slowPathStartLocation; }

    bool generateInline(CCallHelpers& jit, MathICGenerationState& state, bool shouldEmitProfiling = true)
    {
#if CPU(ARM_TRADITIONAL)
        // FIXME: Remove this workaround once the proper fixes are landed.
        // [ARM] Disable Inline Caching on ARMv7 traditional until proper fix
        // https://bugs.webkit.org/show_bug.cgi?id=159759
        return false;
#endif

        state.fastPathStart = jit.label();
        size_t startSize = jit.m_assembler.buffer().codeSize();

        if (m_arithProfile) {
            if (isProfileEmpty(*m_arithProfile)) {
                // It looks like the MathIC has yet to execute. We don't want to emit code in this
                // case for a couple reasons. First, the operation may never execute, so if we don't emit
                // code, it's a win. Second, if the operation does execute, we can emit better code
                // once we have an idea about the types.
                state.slowPathJumps.append(jit.patchableJump());
                size_t inlineSize = jit.m_assembler.buffer().codeSize() - startSize;
                ASSERT_UNUSED(inlineSize, static_cast<ptrdiff_t>(inlineSize) <= MacroAssembler::patchableJumpSize());
                state.shouldSlowPathRepatch = true;
                state.fastPathEnd = jit.label();
                ASSERT(!m_generateFastPathOnRepatch); // We should have gathered some observed type info about the types before trying to regenerate again.
                m_generateFastPathOnRepatch = true;
                return true;
            }
        }

        JITMathICInlineResult result = m_generator.generateInline(jit, state, m_arithProfile);

        switch (result) {
        case JITMathICInlineResult::GeneratedFastPath: {
            size_t inlineSize = jit.m_assembler.buffer().codeSize() - startSize;
            if (static_cast<ptrdiff_t>(inlineSize) < MacroAssembler::patchableJumpSize()) {
                size_t nopsToEmitInBytes = MacroAssembler::patchableJumpSize() - inlineSize;
                jit.emitNops(nopsToEmitInBytes);
            }
            state.shouldSlowPathRepatch = true;
            state.fastPathEnd = jit.label();
            return true;
        }
        case JITMathICInlineResult::GenerateFullSnippet: {
            MacroAssembler::JumpList endJumpList;
            bool result = m_generator.generateFastPath(jit, endJumpList, state.slowPathJumps, m_arithProfile, shouldEmitProfiling);
            if (result) {
                state.fastPathEnd = jit.label();
                state.shouldSlowPathRepatch = false;
                endJumpList.link(&jit);
                return true;
            }
            return false;
        }
        case JITMathICInlineResult::DontGenerate: {
            return false;
        }
        default:
            ASSERT_NOT_REACHED();
        }

        return false;
    }

    void generateOutOfLine(CodeBlock* codeBlock, FunctionPtr<CFunctionPtrTag> callReplacement)
    {
        auto linkJumpToOutOfLineSnippet = [&] () {
            CCallHelpers jit(codeBlock);
            auto jump = jit.jump();
            // We don't need a nop sled here because nobody should be jumping into the middle of an IC.
            bool needsBranchCompaction = false;
            RELEASE_ASSERT(jit.m_assembler.buffer().codeSize() <= static_cast<size_t>(MacroAssembler::differenceBetweenCodePtr(m_inlineStart, m_inlineEnd)));
            LinkBuffer linkBuffer(jit, m_inlineStart, jit.m_assembler.buffer().codeSize(), JITCompilationMustSucceed, needsBranchCompaction);
            RELEASE_ASSERT(linkBuffer.isValid());
            linkBuffer.link(jump, CodeLocationLabel<JITStubRoutinePtrTag>(m_code.code()));
            FINALIZE_CODE(linkBuffer, NoPtrTag, "JITMathIC: linking constant jump to out of line stub");
        };

        auto replaceCall = [&] () {
#if COMPILER(MSVC)
            ftlThunkAwareRepatchCall(codeBlock, slowPathCallLocation().retagged<JSInternalPtrTag>(), callReplacement);
#else
            ftlThunkAwareRepatchCall(codeBlock, slowPathCallLocation().template retagged<JSInternalPtrTag>(), callReplacement);
#endif
        };

        bool shouldEmitProfiling = !JITCode::isOptimizingJIT(codeBlock->jitType());

        if (m_generateFastPathOnRepatch) {

            CCallHelpers jit(codeBlock);
            MathICGenerationState generationState;
            bool generatedInline = generateInline(jit, generationState, shouldEmitProfiling);

            // We no longer want to try to regenerate the fast path.
            m_generateFastPathOnRepatch = false;

            if (generatedInline) {
                auto jumpToDone = jit.jump();

                LinkBuffer linkBuffer(jit, codeBlock, JITCompilationCanFail);
                if (!linkBuffer.didFailToAllocate()) {
                    linkBuffer.link(generationState.slowPathJumps, slowPathStartLocation());
                    linkBuffer.link(jumpToDone, doneLocation());

                    m_code = FINALIZE_CODE_FOR(
                        codeBlock, linkBuffer, JITStubRoutinePtrTag, "JITMathIC: generating out of line fast IC snippet");

                    if (!generationState.shouldSlowPathRepatch) {
                        // We won't need to regenerate, so we can wire the slow path call
                        // to a non repatching variant.
                        replaceCall();
                    }

                    linkJumpToOutOfLineSnippet();

                    return;
                }
            }
            
            // We weren't able to generate an out of line fast path.
            // We just generate the snippet in its full generality.
        }

        // We rewire to the alternate regardless of whether or not we can allocate the out of line path
        // because if we fail allocating the out of line path, we don't want to waste time trying to
        // allocate it in the future.
        replaceCall();

        {
            CCallHelpers jit(codeBlock);

            MacroAssembler::JumpList endJumpList; 
            MacroAssembler::JumpList slowPathJumpList; 

            bool emittedFastPath = m_generator.generateFastPath(jit, endJumpList, slowPathJumpList, m_arithProfile, shouldEmitProfiling);
            if (!emittedFastPath)
                return;
            endJumpList.append(jit.jump());

            LinkBuffer linkBuffer(jit, codeBlock, JITCompilationCanFail);
            if (linkBuffer.didFailToAllocate())
                return;

            linkBuffer.link(endJumpList, doneLocation());
            linkBuffer.link(slowPathJumpList, slowPathStartLocation());

            m_code = FINALIZE_CODE_FOR(
                codeBlock, linkBuffer, JITStubRoutinePtrTag, "JITMathIC: generating out of line IC snippet");
        }

        linkJumpToOutOfLineSnippet();
    }

    void finalizeInlineCode(const MathICGenerationState& state, LinkBuffer& linkBuffer)
    {
        CodeLocationLabel<JSInternalPtrTag> start = linkBuffer.locationOf<JSInternalPtrTag>(state.fastPathStart);
        m_inlineStart = start;

        m_inlineEnd = linkBuffer.locationOf<JSInternalPtrTag>(state.fastPathEnd);
        ASSERT(m_inlineEnd.untaggedExecutableAddress() > m_inlineStart.untaggedExecutableAddress());

        m_slowPathCallLocation = linkBuffer.locationOf<JSInternalPtrTag>(state.slowPathCall);
        m_slowPathStartLocation = linkBuffer.locationOf<JSInternalPtrTag>(state.slowPathStart);
    }

    ArithProfile* arithProfile() const { return m_arithProfile; }
    Instruction* instruction() const { return m_instruction; }

#if ENABLE(MATH_IC_STATS)
    size_t m_generatedCodeSize { 0 };
    size_t codeSize() const
    {
        size_t result = m_generatedCodeSize;
        if (m_code)
            result += m_code.size();
        return result;
    }
#endif

    ArithProfile* m_arithProfile;
    Instruction* m_instruction;
    MacroAssemblerCodeRef<JITStubRoutinePtrTag> m_code;
    CodeLocationLabel<JSInternalPtrTag> m_inlineStart;
    CodeLocationLabel<JSInternalPtrTag> m_inlineEnd;
    CodeLocationLabel<JSInternalPtrTag> m_slowPathCallLocation;
    CodeLocationLabel<JSInternalPtrTag> m_slowPathStartLocation;
    bool m_generateFastPathOnRepatch { false };
    GeneratorType m_generator;
};

inline bool isBinaryProfileEmpty(ArithProfile& arithProfile)
{
    return arithProfile.lhsObservedType().isEmpty() || arithProfile.rhsObservedType().isEmpty();
}
template <typename GeneratorType>
class JITBinaryMathIC : public JITMathIC<GeneratorType, isBinaryProfileEmpty> {
public:
    JITBinaryMathIC(ArithProfile* arithProfile, Instruction* instruction)
        : JITMathIC<GeneratorType, isBinaryProfileEmpty>(arithProfile, instruction)
    {
    }
};

typedef JITBinaryMathIC<JITAddGenerator> JITAddIC;
typedef JITBinaryMathIC<JITMulGenerator> JITMulIC;
typedef JITBinaryMathIC<JITSubGenerator> JITSubIC;


inline bool isUnaryProfileEmpty(ArithProfile& arithProfile)
{
    return arithProfile.lhsObservedType().isEmpty();
}
template <typename GeneratorType>
class JITUnaryMathIC : public JITMathIC<GeneratorType, isUnaryProfileEmpty> {
public:
    JITUnaryMathIC(ArithProfile* arithProfile, Instruction* instruction)
        : JITMathIC<GeneratorType, isUnaryProfileEmpty>(arithProfile, instruction)
    {
    }
};

typedef JITUnaryMathIC<JITNegGenerator> JITNegIC;

} // namespace JSC

#endif // ENABLE(JIT)