// -*- mode: c++ -*-

// Copyright (c) 2012-2014 ARM Limited
// All rights reserved
//
// The license below extends only to copyright in the software and shall
// not be construed as granting a license to any other intellectual
// property including but not limited to intellectual property relating
// to a hardware implementation of the functionality of the software
// licensed hereunder.  You may use the software subject to the license
// terms below provided that you ensure that this notice is replicated
// unmodified and in its entirety in all distributions of the software,
// modified or unmodified, in source code or in binary form.
//
// Redistribution and use in source and binary forms, with or without
// modification, are permitted provided that the following conditions are
// met: redistributions of source code must retain the above copyright
// notice, this list of conditions and the following disclaimer;
// 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;
// neither the name of the copyright holders nor the names of its
// contributors may be used to endorse or promote products derived from
// this software without specific prior written permission.
//
// THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
// "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 THE COPYRIGHT
// OWNER 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.
//
// Authors: Mbou Eyole
//          Giacomo Gabrielli

let {{

    header_output = ''
    decoder_output = ''
    exec_output = ''

    zeroSveVecRegUpperPartCode = '''
        TheISA::ISA::zeroSveVecRegUpperPart(%s,
            ArmStaticInst::getCurSveVecLen<uint64_t>(xc->tcBase()));
    '''

    def mkMemAccMicroOp(name):
        global header_output, decoder_output, exec_output
        SPAlignmentCheckCodeNeon = '''
            if (baseIsSP && bits(XURa, 3, 0) &&
                SPAlignmentCheckEnabled(xc->tcBase())) {
                return std::make_shared<SPAlignmentFault>();
            }
        '''
        eaCode = SPAlignmentCheckCodeNeon + '''
            EA = XURa + imm;
        '''
        memDecl = '''
            const int MaxNumBytes = 16;
            union MemUnion {
                uint8_t bytes[MaxNumBytes];
                uint32_t floatRegBits[MaxNumBytes / 4];
            };
        '''

        # Do endian conversion for all the elements
        convCode = '''
            VReg x = {0, 0};

            x.lo = (((XReg) memUnion.floatRegBits[1]) << 32) |
                (XReg) memUnion.floatRegBits[0];
            x.hi = (((XReg) memUnion.floatRegBits[3]) << 32) |
                (XReg) memUnion.floatRegBits[2];

            const unsigned eCount = 16 / (1 << eSize);

            if (isBigEndian64(xc->tcBase())) {
                for (unsigned i = 0; i < eCount; i++) {
                    switch (eSize) {
                      case 0x3:  // 64-bit
                        writeVecElem(&x, (XReg) gtobe(
                            (uint64_t) readVecElem(x, i, eSize)), i, eSize);
                        break;
                      case 0x2:  // 32-bit
                        writeVecElem(&x, (XReg) gtobe(
                            (uint32_t) readVecElem(x, i, eSize)), i, eSize);
                        break;
                      case 0x1:  // 16-bit
                        writeVecElem(&x, (XReg) gtobe(
                            (uint16_t) readVecElem(x, i, eSize)), i, eSize);
                        break;
                      default:  // 8-bit
                        break;  // Nothing to do here
                    }
                }
            } else {
                for (unsigned i = 0; i < eCount; i++) {
                    switch (eSize) {
                      case 0x3:  // 64-bit
                        writeVecElem(&x, (XReg) gtole(
                            (uint64_t) readVecElem(x, i, eSize)), i, eSize);
                        break;
                      case 0x2:  // 32-bit
                        writeVecElem(&x, (XReg) gtole(
                            (uint32_t) readVecElem(x, i, eSize)), i, eSize);
                        break;
                      case 0x1:  // 16-bit
                        writeVecElem(&x, (XReg) gtole(
                            (uint16_t) readVecElem(x, i, eSize)), i, eSize);
                        break;
                      default:  // 8-bit
                        break;  // Nothing to do here
                    }
                }
            }

            memUnion.floatRegBits[0] = (uint32_t) x.lo;
            memUnion.floatRegBits[1] = (uint32_t) (x.lo >> 32);
            memUnion.floatRegBits[2] = (uint32_t) x.hi;
            memUnion.floatRegBits[3] = (uint32_t) (x.hi >> 32);
        '''

        # Offload everything into registers
        regSetCode = ''
        for reg in range(4):
            regSetCode += '''
            AA64FpDestP%(reg)d_uw = gtoh(memUnion.floatRegBits[%(reg)d]);
            ''' % { 'reg' : reg }

        # Pull everything in from registers
        regGetCode = ''
        for reg in range(4):
            regGetCode += '''
            memUnion.floatRegBits[%(reg)d] = htog(AA64FpDestP%(reg)d_uw);
            ''' % { 'reg' : reg }

        loadMemAccCode = convCode + regSetCode
        storeMemAccCode = regGetCode + convCode

        loadIop = InstObjParams(name + 'ld',
                'MicroNeonLoad64',
                'MicroNeonMemOp',
            {   'mem_decl' : memDecl,
                'memacc_code' : loadMemAccCode,
                'ea_code' : simd64EnabledCheckCode + eaCode,
            },
            [ 'IsMicroop', 'IsMemRef', 'IsLoad' ])
        loadIop.snippets["memacc_code"] += zeroSveVecRegUpperPartCode % \
            "AA64FpDest"
        storeIop = InstObjParams(name + 'st',
                'MicroNeonStore64',
                'MicroNeonMemOp',
            {   'mem_decl' : memDecl,
                'memacc_code' : storeMemAccCode,
                'ea_code' : simd64EnabledCheckCode + eaCode,
            },
            [ 'IsMicroop', 'IsMemRef', 'IsStore' ])

        exec_output += NeonLoadExecute64.subst(loadIop) + \
            NeonLoadInitiateAcc64.subst(loadIop) + \
            NeonLoadCompleteAcc64.subst(loadIop) + \
            NeonStoreExecute64.subst(storeIop) + \
            NeonStoreInitiateAcc64.subst(storeIop) + \
            NeonStoreCompleteAcc64.subst(storeIop)
        header_output += MicroNeonMemDeclare64.subst(loadIop) + \
            MicroNeonMemDeclare64.subst(storeIop)

    def mkMarshalMicroOp(name, Name, numRegs=4):
        global header_output, decoder_output, exec_output

        getInputCodeOp1L = ''
        for v in range(numRegs):
            for p in range(4):
                getInputCodeOp1L += '''
            writeVecElem(&input[%(v)d], (XReg) AA64FpOp1P%(p)dV%(v)d_uw,
                         %(p)d, 0x2);
            ''' % { 'v' : v, 'p' : p }

        getInputCodeOp1S = ''
        for v in range(numRegs):
            for p in range(4):
                getInputCodeOp1S += '''
            writeVecElem(&input[%(v)d], (XReg) AA64FpOp1P%(p)dV%(v)dS_uw,
                         %(p)d, 0x2);
            ''' % { 'v' : v, 'p' : p }

        if name == 'deint_neon_uop':

            eCode = '''
                // input data from scratch area
                VReg input[4] = { {0, 0}, {0, 0}, {0, 0}, {0, 0} };
                VReg output[2];  // output data to arch. SIMD regs
                VReg temp;
                temp.lo = 0;
                temp.hi = 0;
            '''
            for p in range(4):
                eCode += '''
                writeVecElem(&temp, (XReg) AA64FpDestP%(p)dV1L_uw, %(p)d, 0x2);
                ''' % { 'p' : p }
            eCode += getInputCodeOp1L

            # Note that numRegs is not always the same as numStructElems; in
            # particular, for LD1/ST1, numStructElems is 1 but numRegs can be
            # 1, 2, 3 or 4

            eCode += '''
                output[0].lo = 0;
                output[0].hi = 0;
                output[1].lo = 0;
                output[1].hi = 0;

                int eCount = dataSize / (8 << eSize);
                int eSizeBytes = 1 << eSize;  // element size in bytes
                int numBytes = step * dataSize / 4;
                int totNumBytes = numRegs * dataSize / 8;

                int structElemNo, pos, a, b;
                XReg data;

                for (int r = 0; r < 2; ++r) {
                    for (int i = 0; i < eCount; ++i) {
                        if (numBytes < totNumBytes) {
                            structElemNo = r + (step * 2);
                            if (numStructElems == 1) {
                                pos = (eSizeBytes * i) +
                                    (eCount * structElemNo * eSizeBytes);
                            } else {
                                pos = (numStructElems * eSizeBytes * i) +
                                    (structElemNo * eSizeBytes);
                            }
                            a = pos / 16;
                            b = (pos % 16) / eSizeBytes;
                            data = (XReg) readVecElem(input[a], (XReg) b,
                                                      eSize);
                            writeVecElem(&output[r], data, i, eSize);
                            numBytes += eSizeBytes;
                        }
                    }
                }
            '''
            for p in range(4):
                eCode += '''
                AA64FpDestP%(p)dV0L_uw = (uint32_t) readVecElem(output[0],
                    %(p)d, 0x2);
                ''' % { 'p' : p }
            eCode += '''
                if ((numRegs % 2 == 0) || (numRegs == 3 && step == 0)) {
            '''
            for p in range(4):
                eCode += '''
                    AA64FpDestP%(p)dV1L_uw = (uint32_t) readVecElem(
                        output[1], %(p)d, 0x2);
                ''' % { 'p' : p }
            eCode += '''
                } else {
            '''
            for p in range(4):
                eCode += '''
                    AA64FpDestP%(p)dV1L_uw = (uint32_t) readVecElem(temp,
                        %(p)d, 0x2);
                ''' % { 'p' : p }
            eCode += '''
                }
            '''

            iop = InstObjParams(name, Name, 'MicroNeonMixOp64',
                                { 'code' : eCode, 'op_class' : 'No_OpClass' },
                                ['IsMicroop'])
            header_output += MicroNeonMixDeclare64.subst(iop)
            exec_output += MicroNeonMixExecute64.subst(iop)

        elif name == 'int_neon_uop':

            eCode = '''
                // input data from arch. SIMD regs
                VReg input[4] = { {0, 0}, {0, 0}, {0, 0}, {0, 0} };
                VReg output[2];  // output data to scratch area
            '''

            eCode += getInputCodeOp1S

            # Note that numRegs is not always the same as numStructElems; in
            # particular, for LD1/ST1, numStructElems is 1 but numRegs can be
            # 1, 2, 3 or 4

            eCode += '''
                int eCount = dataSize / (8 << eSize);
                int eSizeBytes = 1 << eSize;
                int totNumBytes = numRegs * dataSize / 8;
                int numOutputElems = 128 / (8 << eSize);
                int stepOffset = step * 32;

                for (int i = 0; i < 2; ++i) {
                    output[i].lo = 0;
                    output[i].hi = 0;
                }

                int r = 0, k = 0, i, j;
                XReg data;

                for (int pos = stepOffset; pos < 32 + stepOffset;
                        pos += eSizeBytes) {
                    if (pos < totNumBytes) {
                        if (numStructElems == 1) {
                            i = (pos / eSizeBytes) % eCount;
                            j = pos / (eCount * eSizeBytes);
                        } else {
                            i = pos / (numStructElems * eSizeBytes);
                            j = (pos % (numStructElems * eSizeBytes)) /
                                eSizeBytes;
                        }
                        data = (XReg) readVecElem(input[j], (XReg) i, eSize);
                        writeVecElem(&output[r], data, k, eSize);
                        k++;
                        if (k == numOutputElems){
                            k = 0;
                            ++r;
                        }
                    }
                }
                '''
            for v in range(2):
                for p in range(4):
                    eCode += '''
                AA64FpDestP%(p)dV%(v)d_uw = (uint32_t) readVecElem(
                    output[%(v)d], %(p)d, 0x2);
                ''' % { 'v': v, 'p': p}

            iop = InstObjParams(name, Name, 'MicroNeonMixOp64',
                                { 'code' : eCode, 'op_class' : 'No_OpClass' },
                                ['IsMicroop'])
            header_output += MicroNeonMixDeclare64.subst(iop)
            exec_output += MicroNeonMixExecute64.subst(iop)

        elif name == 'unpack_neon_uop':

            eCode = '''
                //input data from scratch area
                VReg input[4] = { {0, 0}, {0, 0}, {0, 0}, {0, 0} };
                //output data to arch. SIMD regs
                VReg output[2] = { {0, 0}, {0, 0} };
            '''

            eCode += getInputCodeOp1L

            # Fill output regs with register data initially.  Note that
            # elements in output register outside indexed lanes are left
            # untouched
            for v in range(2):
                for p in range(4):
                    eCode += '''
                writeVecElem(&output[%(v)d], (XReg) AA64FpDestP%(p)dV%(v)dL_uw,
                             %(p)d, 0x2);
                ''' % { 'v': v, 'p': p}
            eCode += '''
                int eCount = dataSize / (8 << eSize);
                int eCount128 = 128 / (8 << eSize);
                int eSizeBytes = 1 << eSize;
                int totNumBytes = numStructElems * eSizeBytes;
                int numInputElems = eCount128;
                int stepOffset = step * 2 * eSizeBytes;
                int stepLimit = 2 * eSizeBytes;

                int r = 0, i, j;
                XReg data;

                for (int pos = stepOffset; pos < stepLimit + stepOffset;
                        pos += eSizeBytes) {
                    if (pos < totNumBytes) {
                        r = pos / eSizeBytes;
                        j = r / numInputElems;
                        i = r % numInputElems;
                        data = (XReg) readVecElem(input[j], (XReg) i, eSize);

                        if (replicate) {
                            for (int i = 0; i < eCount128; ++i) {
                                if (i < eCount) {
                                    writeVecElem(&output[r % 2], data, i,
                                                 eSize);
                                } else {  // zero extend if necessary
                                    writeVecElem(&output[r % 2], (XReg) 0, i,
                                                 eSize);
                                }
                            }
                        } else {
                            writeVecElem(&output[r % 2], data, lane, eSize);
                        }
                    }
                }
            '''
            for v in range(2):
                for p in range(4):
                    eCode += '''
                AA64FpDestP%(p)dV%(v)dL_uw = (uint32_t) readVecElem(
                    output[%(v)d], %(p)d, 0x2);
                ''' % { 'v' : v, 'p' : p }

            iop = InstObjParams(name, Name, 'MicroNeonMixLaneOp64',
                                { 'code' : eCode }, ['IsMicroop'])
            header_output += MicroNeonMixLaneDeclare64.subst(iop)
            exec_output += MicroNeonMixExecute64.subst(iop)

        elif name == 'pack_neon_uop':

            eCode = '''
                // input data from arch. SIMD regs
                VReg input[4] = { {0, 0}, {0, 0}, {0, 0}, {0, 0} };
                VReg output[2];  // output data to scratch area
            '''

            eCode += getInputCodeOp1S

            eCode += '''
                int eSizeBytes = 1 << eSize;
                int numOutputElems = 128 / (8 << eSize);
                int totNumBytes = numStructElems * eSizeBytes;
                int stepOffset = step * 32;
                int stepLimit = 32;

                int r = 0, i, j;
                XReg data;

                for (int i = 0; i < 2; ++i) {
                    output[i].lo = 0;
                    output[i].hi = 0;
                }

                for (int pos = stepOffset; pos < stepLimit + stepOffset;
                        pos += eSizeBytes) {
                    if (pos < totNumBytes) {
                        r = pos / 16;
                        j = pos / eSizeBytes;
                        i = (pos / eSizeBytes) %  numOutputElems;
                        data = (XReg) readVecElem(input[j], lane, eSize);
                        writeVecElem(&output[r % 2], data, i, eSize);
                    }
                }
            '''

            for v in range(2):
                for p in range(4):
                    eCode += '''
                AA64FpDestP%(p)dV%(v)d_uw = (uint32_t) readVecElem(
                    output[%(v)d], %(p)d, 0x2);
                ''' % { 'v' : v, 'p' : p }

            iop = InstObjParams(name, Name, 'MicroNeonMixLaneOp64',
                                { 'code' : eCode }, ['IsMicroop'])
            header_output += MicroNeonMixLaneDeclare64.subst(iop)
            exec_output += MicroNeonMixExecute64.subst(iop)

    # Generate instructions
    mkMemAccMicroOp('mem_neon_uop')
    mkMarshalMicroOp('deint_neon_uop', 'MicroDeintNeon64_1Reg', numRegs=1)
    mkMarshalMicroOp('deint_neon_uop', 'MicroDeintNeon64_2Reg', numRegs=2)
    mkMarshalMicroOp('deint_neon_uop', 'MicroDeintNeon64_3Reg', numRegs=3)
    mkMarshalMicroOp('deint_neon_uop', 'MicroDeintNeon64_4Reg', numRegs=4)
    mkMarshalMicroOp('int_neon_uop', 'MicroIntNeon64_1Reg', numRegs=1)
    mkMarshalMicroOp('int_neon_uop', 'MicroIntNeon64_2Reg', numRegs=2)
    mkMarshalMicroOp('int_neon_uop', 'MicroIntNeon64_3Reg', numRegs=3)
    mkMarshalMicroOp('int_neon_uop', 'MicroIntNeon64_4Reg', numRegs=4)
    mkMarshalMicroOp('unpack_neon_uop', 'MicroUnpackNeon64')
    mkMarshalMicroOp('pack_neon_uop', 'MicroPackNeon64')

}};

let {{

    iop = InstObjParams('vldmult64', 'VldMult64', 'VldMultOp64', '', [])
    header_output += VMemMultDeclare64.subst(iop)
    decoder_output += VMemMultConstructor64.subst(iop)

    iop = InstObjParams('vstmult64', 'VstMult64', 'VstMultOp64', '', [])
    header_output += VMemMultDeclare64.subst(iop)
    decoder_output += VMemMultConstructor64.subst(iop)

    iop = InstObjParams('vldsingle64', 'VldSingle64', 'VldSingleOp64', '', [])
    header_output += VMemSingleDeclare64.subst(iop)
    decoder_output += VMemSingleConstructor64.subst(iop)

    iop = InstObjParams('vstsingle64', 'VstSingle64', 'VstSingleOp64', '', [])
    header_output += VMemSingleDeclare64.subst(iop)
    decoder_output += VMemSingleConstructor64.subst(iop)

}};