Home | History | Annotate | Download | only in codeflinger
      1 /*
      2  * Copyright (C) 2013 The Android Open Source Project
      3  * All rights reserved.
      4  *
      5  * Redistribution and use in source and binary forms, with or without
      6  * modification, are permitted provided that the following conditions
      7  * are met:
      8  *  * Redistributions of source code must retain the above copyright
      9  *    notice, this list of conditions and the following disclaimer.
     10  *  * Redistributions in binary form must reproduce the above copyright
     11  *    notice, this list of conditions and the following disclaimer in
     12  *    the documentation and/or other materials provided with the
     13  *    distribution.
     14  *
     15  * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
     16  * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
     17  * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS
     18  * FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE
     19  * COPYRIGHT OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT,
     20  * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING,
     21  * BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS
     22  * OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED
     23  * AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
     24  * OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT
     25  * OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
     26  * SUCH DAMAGE.
     27  */
     28 
     29 #define LOG_TAG "ArmToArm64Assembler"
     30 
     31 #include <stdio.h>
     32 #include <stdlib.h>
     33 #include <string.h>
     34 
     35 #include <cutils/properties.h>
     36 #include <log/log.h>
     37 #include <private/pixelflinger/ggl_context.h>
     38 
     39 #include "codeflinger/Arm64Assembler.h"
     40 #include "codeflinger/Arm64Disassembler.h"
     41 #include "codeflinger/CodeCache.h"
     42 
     43 /*
     44 ** --------------------------------------------
     45 ** Support for Arm64 in GGLAssembler JIT
     46 ** --------------------------------------------
     47 **
     48 ** Approach
     49 ** - GGLAssembler and associated files are largely un-changed.
     50 ** - A translator class maps ArmAssemblerInterface calls to
     51 **   generate Arm64 instructions.
     52 **
     53 ** ----------------------
     54 ** ArmToArm64Assembler
     55 ** ----------------------
     56 **
     57 ** - Subclassed from ArmAssemblerInterface
     58 **
     59 ** - Translates each ArmAssemblerInterface call to generate
     60 **   one or more Arm64 instructions  as necessary.
     61 **
     62 ** - Does not implement ArmAssemblerInterface portions unused by GGLAssembler
     63 **   It calls NOT_IMPLEMENTED() for such cases, which in turn logs
     64 **    a fatal message.
     65 **
     66 ** - Uses A64_.. series of functions to generate instruction machine code
     67 **   for Arm64 instructions. These functions also log the instruction
     68 **   to LOG, if ARM64_ASM_DEBUG define is set to 1
     69 **
     70 ** - Dumps machine code and eqvt assembly if "debug.pf.disasm" option is set
     71 **   It uses arm64_disassemble to perform disassembly
     72 **
     73 ** - Uses register 13 (SP in ARM), 15 (PC in ARM), 16, 17 for storing
     74 **   intermediate results. GGLAssembler does not use SP and PC as these
     75 **   registers are marked as reserved. The temporary registers are not
     76 **   saved/restored on stack as these are caller-saved registers in Arm64
     77 **
     78 ** - Uses CSEL instruction to support conditional execution. The result is
     79 **   stored in a temporary register and then copied to the target register
     80 **   if the condition is true.
     81 **
     82 ** - In the case of conditional data transfer instructions, conditional
     83 **   branch is used to skip over instruction, if the condition is false
     84 **
     85 ** - Wherever possible, immediate values are transferred to temporary
     86 **   register prior to processing. This simplifies overall implementation
     87 **   as instructions requiring immediate values are converted to
     88 **   move immediate instructions followed by register-register instruction.
     89 **
     90 ** --------------------------------------------
     91 ** ArmToArm64Assembler unit test bench
     92 ** --------------------------------------------
     93 **
     94 ** - Tests ArmToArm64Assembler interface for all the possible
     95 **   ways in which GGLAssembler uses ArmAssemblerInterface interface.
     96 **
     97 ** - Uses test jacket (written in assembly) to set the registers,
     98 **   condition flags prior to calling generated instruction. It also
     99 **   copies registers and flags at the end of execution. Caller then
    100 **   checks if generated code performed correct operation based on
    101 **   output registers and flags.
    102 **
    103 ** - Broadly contains three type of tests, (i) data operation tests
    104 **   (ii) data transfer tests and (iii) LDM/STM tests.
    105 **
    106 ** ----------------------
    107 ** Arm64 disassembler
    108 ** ----------------------
    109 ** - This disassembler disassembles only those machine codes which can be
    110 **   generated by ArmToArm64Assembler. It has a unit testbench which
    111 **   tests all the instructions supported by the disassembler.
    112 **
    113 ** ------------------------------------------------------------------
    114 ** ARMAssembler/ARMAssemblerInterface/ARMAssemblerProxy changes
    115 ** ------------------------------------------------------------------
    116 **
    117 ** - In existing code, addresses were being handled as 32 bit values at
    118 **   certain places.
    119 **
    120 ** - Added a new set of functions for address load/store/manipulation.
    121 **   These are ADDR_LDR, ADDR_STR, ADDR_ADD, ADDR_SUB and they map to
    122 **   default 32 bit implementations in ARMAssemblerInterface.
    123 **
    124 ** - ArmToArm64Assembler maps these functions to appropriate 64 bit
    125 **   functions.
    126 **
    127 ** ----------------------
    128 ** GGLAssembler changes
    129 ** ----------------------
    130 ** - Since ArmToArm64Assembler can generate 4 Arm64 instructions for
    131 **   each call in worst case, the memory required is set to 4 times
    132 **   ARM memory
    133 **
    134 ** - Address load/store/manipulation were changed to use new functions
    135 **   added in the ARMAssemblerInterface.
    136 **
    137 */
    138 
    139 
    140 #define NOT_IMPLEMENTED()  LOG_FATAL("Arm instruction %s not yet implemented\n", __func__)
    141 
    142 #define ARM64_ASM_DEBUG 0
    143 
    144 #if ARM64_ASM_DEBUG
    145     #define LOG_INSTR(...) ALOGD("\t" __VA_ARGS__)
    146     #define LOG_LABEL(...) ALOGD(__VA_ARGS__)
    147 #else
    148     #define LOG_INSTR(...) ((void)0)
    149     #define LOG_LABEL(...) ((void)0)
    150 #endif
    151 
    152 namespace android {
    153 
    154 static const char* shift_codes[] =
    155 {
    156     "LSL", "LSR", "ASR", "ROR"
    157 };
    158 static const char *cc_codes[] =
    159 {
    160     "EQ", "NE", "CS", "CC", "MI",
    161     "PL", "VS", "VC", "HI", "LS",
    162     "GE", "LT", "GT", "LE", "AL", "NV"
    163 };
    164 
    165 ArmToArm64Assembler::ArmToArm64Assembler(const sp<Assembly>& assembly)
    166     :   ARMAssemblerInterface(),
    167         mAssembly(assembly)
    168 {
    169     mBase = mPC = (uint32_t *)assembly->base();
    170     mDuration = ggl_system_time();
    171     mZeroReg = 13;
    172     mTmpReg1 = 15;
    173     mTmpReg2 = 16;
    174     mTmpReg3 = 17;
    175 }
    176 
    177 ArmToArm64Assembler::ArmToArm64Assembler(void *base)
    178     :   ARMAssemblerInterface(), mAssembly(NULL)
    179 {
    180     mBase = mPC = (uint32_t *)base;
    181     mDuration = ggl_system_time();
    182     // Regs 13, 15, 16, 17 are used as temporary registers
    183     mZeroReg = 13;
    184     mTmpReg1 = 15;
    185     mTmpReg2 = 16;
    186     mTmpReg3 = 17;
    187 }
    188 
    189 ArmToArm64Assembler::~ArmToArm64Assembler()
    190 {
    191 }
    192 
    193 uint32_t* ArmToArm64Assembler::pc() const
    194 {
    195     return mPC;
    196 }
    197 
    198 uint32_t* ArmToArm64Assembler::base() const
    199 {
    200     return mBase;
    201 }
    202 
    203 void ArmToArm64Assembler::reset()
    204 {
    205     if(mAssembly == NULL)
    206         mPC = mBase;
    207     else
    208         mBase = mPC = (uint32_t *)mAssembly->base();
    209     mBranchTargets.clear();
    210     mLabels.clear();
    211     mLabelsInverseMapping.clear();
    212     mComments.clear();
    213 #if ARM64_ASM_DEBUG
    214     ALOGI("RESET\n");
    215 #endif
    216 }
    217 
    218 int ArmToArm64Assembler::getCodegenArch()
    219 {
    220     return CODEGEN_ARCH_ARM64;
    221 }
    222 
    223 // ----------------------------------------------------------------------------
    224 
    225 void ArmToArm64Assembler::disassemble(const char* name)
    226 {
    227     if(name)
    228     {
    229         printf("%s:\n", name);
    230     }
    231     size_t count = pc()-base();
    232     uint32_t* i = base();
    233     while (count--)
    234     {
    235         ssize_t label = mLabelsInverseMapping.indexOfKey(i);
    236         if (label >= 0)
    237         {
    238             printf("%s:\n", mLabelsInverseMapping.valueAt(label));
    239         }
    240         ssize_t comment = mComments.indexOfKey(i);
    241         if (comment >= 0)
    242         {
    243             printf("; %s\n", mComments.valueAt(comment));
    244         }
    245         printf("%p:    %08x    ", i, uint32_t(i[0]));
    246         {
    247             char instr[256];
    248             ::arm64_disassemble(*i, instr);
    249             printf("%s\n", instr);
    250         }
    251         i++;
    252     }
    253 }
    254 
    255 void ArmToArm64Assembler::comment(const char* string)
    256 {
    257     mComments.add(mPC, string);
    258     LOG_INSTR("//%s\n", string);
    259 }
    260 
    261 void ArmToArm64Assembler::label(const char* theLabel)
    262 {
    263     mLabels.add(theLabel, mPC);
    264     mLabelsInverseMapping.add(mPC, theLabel);
    265     LOG_LABEL("%s:\n", theLabel);
    266 }
    267 
    268 void ArmToArm64Assembler::B(int cc, const char* label)
    269 {
    270     mBranchTargets.add(branch_target_t(label, mPC));
    271     LOG_INSTR("B%s %s\n", cc_codes[cc], label );
    272     *mPC++ = (0x54 << 24) | cc;
    273 }
    274 
    275 void ArmToArm64Assembler::BL(int /*cc*/, const char* /*label*/)
    276 {
    277     NOT_IMPLEMENTED(); //Not Required
    278 }
    279 
    280 // ----------------------------------------------------------------------------
    281 //Prolog/Epilog & Generate...
    282 // ----------------------------------------------------------------------------
    283 
    284 void ArmToArm64Assembler::prolog()
    285 {
    286     // write prolog code
    287     mPrologPC = mPC;
    288     *mPC++ = A64_MOVZ_X(mZeroReg,0,0);
    289 }
    290 
    291 void ArmToArm64Assembler::epilog(uint32_t /*touched*/)
    292 {
    293     // write epilog code
    294     static const int XLR = 30;
    295     *mPC++ = A64_RET(XLR);
    296 }
    297 
    298 int ArmToArm64Assembler::generate(const char* name)
    299 {
    300     // fixup all the branches
    301     size_t count = mBranchTargets.size();
    302     while (count--)
    303     {
    304         const branch_target_t& bt = mBranchTargets[count];
    305         uint32_t* target_pc = mLabels.valueFor(bt.label);
    306         LOG_ALWAYS_FATAL_IF(!target_pc,
    307                 "error resolving branch targets, target_pc is null");
    308         int32_t offset = int32_t(target_pc - bt.pc);
    309         *bt.pc |= (offset & 0x7FFFF) << 5;
    310     }
    311 
    312     if(mAssembly != NULL)
    313         mAssembly->resize( int(pc()-base())*4 );
    314 
    315     // the instruction cache is flushed by CodeCache
    316     const int64_t duration = ggl_system_time() - mDuration;
    317     const char * const format = "generated %s (%d ins) at [%p:%p] in %ld ns\n";
    318     ALOGI(format, name, int(pc()-base()), base(), pc(), duration);
    319 
    320 
    321     char value[PROPERTY_VALUE_MAX];
    322     property_get("debug.pf.disasm", value, "0");
    323     if (atoi(value) != 0)
    324     {
    325         printf(format, name, int(pc()-base()), base(), pc(), duration);
    326         disassemble(name);
    327     }
    328     return NO_ERROR;
    329 }
    330 
    331 uint32_t* ArmToArm64Assembler::pcForLabel(const char* label)
    332 {
    333     return mLabels.valueFor(label);
    334 }
    335 
    336 // ----------------------------------------------------------------------------
    337 // Data Processing...
    338 // ----------------------------------------------------------------------------
    339 void ArmToArm64Assembler::dataProcessingCommon(int opcode,
    340         int s, int Rd, int Rn, uint32_t Op2)
    341 {
    342     if(opcode != opSUB && s == 1)
    343     {
    344         NOT_IMPLEMENTED(); //Not required
    345         return;
    346     }
    347 
    348     if(opcode != opSUB && opcode != opADD && opcode != opAND &&
    349        opcode != opORR && opcode != opMVN)
    350     {
    351         NOT_IMPLEMENTED(); //Not required
    352         return;
    353     }
    354 
    355     if(Op2 == OPERAND_REG_IMM && mAddrMode.reg_imm_shift > 31)
    356         {
    357         NOT_IMPLEMENTED();
    358         return;
    359     }
    360 
    361     //Store immediate in temporary register and convert
    362     //immediate operation into register operation
    363     if(Op2 == OPERAND_IMM)
    364     {
    365         int imm = mAddrMode.immediate;
    366         *mPC++ = A64_MOVZ_W(mTmpReg2, imm & 0x0000FFFF, 0);
    367         *mPC++ = A64_MOVK_W(mTmpReg2, (imm >> 16) & 0x0000FFFF, 16);
    368         Op2 = mTmpReg2;
    369     }
    370 
    371 
    372     {
    373         uint32_t shift;
    374         uint32_t amount;
    375         uint32_t Rm;
    376 
    377         if(Op2 == OPERAND_REG_IMM)
    378         {
    379             shift   = mAddrMode.reg_imm_type;
    380             amount  = mAddrMode.reg_imm_shift;
    381             Rm      = mAddrMode.reg_imm_Rm;
    382         }
    383         else if(Op2 < OPERAND_REG)
    384         {
    385             shift   = 0;
    386             amount  = 0;
    387             Rm      = Op2;
    388         }
    389         else
    390         {
    391             NOT_IMPLEMENTED(); //Not required
    392             return;
    393         }
    394 
    395         switch(opcode)
    396         {
    397             case opADD: *mPC++ = A64_ADD_W(Rd, Rn, Rm, shift, amount); break;
    398             case opAND: *mPC++ = A64_AND_W(Rd, Rn, Rm, shift, amount); break;
    399             case opORR: *mPC++ = A64_ORR_W(Rd, Rn, Rm, shift, amount); break;
    400             case opMVN: *mPC++ = A64_ORN_W(Rd, Rn, Rm, shift, amount); break;
    401             case opSUB: *mPC++ = A64_SUB_W(Rd, Rn, Rm, shift, amount, s);break;
    402         };
    403 
    404     }
    405 }
    406 
    407 void ArmToArm64Assembler::dataProcessing(int opcode, int cc,
    408         int s, int Rd, int Rn, uint32_t Op2)
    409 {
    410     uint32_t Wd;
    411 
    412     if(cc != AL)
    413         Wd = mTmpReg1;
    414     else
    415         Wd = Rd;
    416 
    417     if(opcode == opADD || opcode == opAND || opcode == opORR ||opcode == opSUB)
    418     {
    419         dataProcessingCommon(opcode, s, Wd, Rn, Op2);
    420     }
    421     else if(opcode == opCMP)
    422     {
    423         dataProcessingCommon(opSUB, 1, mTmpReg3, Rn, Op2);
    424     }
    425     else if(opcode == opRSB)
    426     {
    427         dataProcessingCommon(opSUB, s, Wd, Rn, Op2);
    428         dataProcessingCommon(opSUB, s, Wd, mZeroReg, Wd);
    429     }
    430     else if(opcode == opMOV)
    431     {
    432         dataProcessingCommon(opORR, 0, Wd, mZeroReg, Op2);
    433         if(s == 1)
    434         {
    435             dataProcessingCommon(opSUB, 1, mTmpReg3, Wd, mZeroReg);
    436         }
    437     }
    438     else if(opcode == opMVN)
    439     {
    440         dataProcessingCommon(opMVN, s, Wd, mZeroReg, Op2);
    441     }
    442     else if(opcode == opBIC)
    443     {
    444         dataProcessingCommon(opMVN, s, mTmpReg3, mZeroReg, Op2);
    445         dataProcessingCommon(opAND, s, Wd, Rn, mTmpReg3);
    446     }
    447     else
    448     {
    449         NOT_IMPLEMENTED();
    450         return;
    451     }
    452 
    453     if(cc != AL)
    454     {
    455         *mPC++ = A64_CSEL_W(Rd, mTmpReg1, Rd, cc);
    456     }
    457 }
    458 // ----------------------------------------------------------------------------
    459 // Address Processing...
    460 // ----------------------------------------------------------------------------
    461 
    462 void ArmToArm64Assembler::ADDR_ADD(int cc,
    463         int s, int Rd, int Rn, uint32_t Op2)
    464 {
    465     if(cc != AL){ NOT_IMPLEMENTED(); return;} //Not required
    466     if(s  != 0) { NOT_IMPLEMENTED(); return;} //Not required
    467 
    468 
    469     if(Op2 == OPERAND_REG_IMM && mAddrMode.reg_imm_type == LSL)
    470     {
    471         int Rm = mAddrMode.reg_imm_Rm;
    472         int amount = mAddrMode.reg_imm_shift;
    473         *mPC++ = A64_ADD_X_Wm_SXTW(Rd, Rn, Rm, amount);
    474     }
    475     else if(Op2 < OPERAND_REG)
    476     {
    477         int Rm = Op2;
    478         int amount = 0;
    479         *mPC++ = A64_ADD_X_Wm_SXTW(Rd, Rn, Rm, amount);
    480     }
    481     else if(Op2 == OPERAND_IMM)
    482     {
    483         int imm = mAddrMode.immediate;
    484         *mPC++ = A64_MOVZ_W(mTmpReg1, imm & 0x0000FFFF, 0);
    485         *mPC++ = A64_MOVK_W(mTmpReg1, (imm >> 16) & 0x0000FFFF, 16);
    486 
    487         int Rm = mTmpReg1;
    488         int amount = 0;
    489         *mPC++ = A64_ADD_X_Wm_SXTW(Rd, Rn, Rm, amount);
    490     }
    491     else
    492     {
    493         NOT_IMPLEMENTED(); //Not required
    494     }
    495 }
    496 
    497 void ArmToArm64Assembler::ADDR_SUB(int cc,
    498         int s, int Rd, int Rn, uint32_t Op2)
    499 {
    500     if(cc != AL){ NOT_IMPLEMENTED(); return;} //Not required
    501     if(s  != 0) { NOT_IMPLEMENTED(); return;} //Not required
    502 
    503     if(Op2 == OPERAND_REG_IMM && mAddrMode.reg_imm_type == LSR)
    504     {
    505         *mPC++ = A64_ADD_W(mTmpReg1, mZeroReg, mAddrMode.reg_imm_Rm,
    506                            LSR, mAddrMode.reg_imm_shift);
    507         *mPC++ = A64_SUB_X_Wm_SXTW(Rd, Rn, mTmpReg1, 0);
    508     }
    509     else
    510     {
    511         NOT_IMPLEMENTED(); //Not required
    512     }
    513 }
    514 
    515 // ----------------------------------------------------------------------------
    516 // multiply...
    517 // ----------------------------------------------------------------------------
    518 void ArmToArm64Assembler::MLA(int cc, int s,int Rd, int Rm, int Rs, int Rn)
    519 {
    520     if(cc != AL){ NOT_IMPLEMENTED(); return;} //Not required
    521 
    522     *mPC++ = A64_MADD_W(Rd, Rm, Rs, Rn);
    523     if(s == 1)
    524         dataProcessingCommon(opSUB, 1, mTmpReg1, Rd, mZeroReg);
    525 }
    526 void ArmToArm64Assembler::MUL(int cc, int s, int Rd, int Rm, int Rs)
    527 {
    528     if(cc != AL){ NOT_IMPLEMENTED(); return;} //Not required
    529     if(s  != 0) { NOT_IMPLEMENTED(); return;} //Not required
    530     *mPC++ = A64_MADD_W(Rd, Rm, Rs, mZeroReg);
    531 }
    532 void ArmToArm64Assembler::UMULL(int /*cc*/, int /*s*/,
    533         int /*RdLo*/, int /*RdHi*/, int /*Rm*/, int /*Rs*/)
    534 {
    535     NOT_IMPLEMENTED(); //Not required
    536 }
    537 void ArmToArm64Assembler::UMUAL(int /*cc*/, int /*s*/,
    538         int /*RdLo*/, int /*RdHi*/, int /*Rm*/, int /*Rs*/)
    539 {
    540     NOT_IMPLEMENTED(); //Not required
    541 }
    542 void ArmToArm64Assembler::SMULL(int /*cc*/, int /*s*/,
    543         int /*RdLo*/, int /*RdHi*/, int /*Rm*/, int /*Rs*/)
    544 {
    545     NOT_IMPLEMENTED(); //Not required
    546 }
    547 void ArmToArm64Assembler::SMUAL(int /*cc*/, int /*s*/,
    548         int /*RdLo*/, int /*RdHi*/, int /*Rm*/, int /*Rs*/)
    549 {
    550     NOT_IMPLEMENTED(); //Not required
    551 }
    552 
    553 // ----------------------------------------------------------------------------
    554 // branches relative to PC...
    555 // ----------------------------------------------------------------------------
    556 void ArmToArm64Assembler::B(int /*cc*/, uint32_t* /*pc*/){
    557     NOT_IMPLEMENTED(); //Not required
    558 }
    559 
    560 void ArmToArm64Assembler::BL(int /*cc*/, uint32_t* /*pc*/){
    561     NOT_IMPLEMENTED(); //Not required
    562 }
    563 
    564 void ArmToArm64Assembler::BX(int /*cc*/, int /*Rn*/){
    565     NOT_IMPLEMENTED(); //Not required
    566 }
    567 
    568 // ----------------------------------------------------------------------------
    569 // data transfer...
    570 // ----------------------------------------------------------------------------
    571 enum dataTransferOp
    572 {
    573     opLDR,opLDRB,opLDRH,opSTR,opSTRB,opSTRH
    574 };
    575 
    576 void ArmToArm64Assembler::dataTransfer(int op, int cc,
    577                             int Rd, int Rn, uint32_t op_type, uint32_t size)
    578 {
    579     const int XSP = 31;
    580     if(Rn == SP)
    581         Rn = XSP;
    582 
    583     if(op_type == OPERAND_IMM)
    584     {
    585         int addrReg;
    586         int imm = mAddrMode.immediate;
    587         if(imm >= 0 && imm < (1<<12))
    588             *mPC++ = A64_ADD_IMM_X(mTmpReg1, mZeroReg, imm, 0);
    589         else if(imm < 0 && -imm < (1<<12))
    590             *mPC++ = A64_SUB_IMM_X(mTmpReg1, mZeroReg, -imm, 0);
    591         else
    592         {
    593             NOT_IMPLEMENTED();
    594             return;
    595         }
    596 
    597         addrReg = Rn;
    598         if(mAddrMode.preindex == true || mAddrMode.postindex == true)
    599         {
    600             *mPC++ = A64_ADD_X(mTmpReg2, addrReg, mTmpReg1);
    601             if(mAddrMode.preindex == true)
    602                 addrReg = mTmpReg2;
    603         }
    604 
    605         if(cc != AL)
    606             *mPC++ = A64_B_COND(cc^1, 8);
    607 
    608         *mPC++ = A64_LDRSTR_Wm_SXTW_0(op, size, Rd, addrReg, mZeroReg);
    609 
    610         if(mAddrMode.writeback == true)
    611             *mPC++ = A64_CSEL_X(Rn, mTmpReg2, Rn, cc);
    612     }
    613     else if(op_type == OPERAND_REG_OFFSET)
    614     {
    615         if(cc != AL)
    616             *mPC++ = A64_B_COND(cc^1, 8);
    617         *mPC++ = A64_LDRSTR_Wm_SXTW_0(op, size, Rd, Rn, mAddrMode.reg_offset);
    618 
    619     }
    620     else if(op_type > OPERAND_UNSUPPORTED)
    621     {
    622         if(cc != AL)
    623             *mPC++ = A64_B_COND(cc^1, 8);
    624         *mPC++ = A64_LDRSTR_Wm_SXTW_0(op, size, Rd, Rn, mZeroReg);
    625     }
    626     else
    627     {
    628         NOT_IMPLEMENTED(); // Not required
    629     }
    630     return;
    631 
    632 }
    633 void ArmToArm64Assembler::ADDR_LDR(int cc, int Rd, int Rn, uint32_t op_type)
    634 {
    635     return dataTransfer(opLDR, cc, Rd, Rn, op_type, 64);
    636 }
    637 void ArmToArm64Assembler::ADDR_STR(int cc, int Rd, int Rn, uint32_t op_type)
    638 {
    639     return dataTransfer(opSTR, cc, Rd, Rn, op_type, 64);
    640 }
    641 void ArmToArm64Assembler::LDR(int cc, int Rd, int Rn, uint32_t op_type)
    642 {
    643     return dataTransfer(opLDR, cc, Rd, Rn, op_type);
    644 }
    645 void ArmToArm64Assembler::LDRB(int cc, int Rd, int Rn, uint32_t op_type)
    646 {
    647     return dataTransfer(opLDRB, cc, Rd, Rn, op_type);
    648 }
    649 void ArmToArm64Assembler::STR(int cc, int Rd, int Rn, uint32_t op_type)
    650 {
    651     return dataTransfer(opSTR, cc, Rd, Rn, op_type);
    652 }
    653 
    654 void ArmToArm64Assembler::STRB(int cc, int Rd, int Rn, uint32_t op_type)
    655 {
    656     return dataTransfer(opSTRB, cc, Rd, Rn, op_type);
    657 }
    658 
    659 void ArmToArm64Assembler::LDRH(int cc, int Rd, int Rn, uint32_t op_type)
    660 {
    661     return dataTransfer(opLDRH, cc, Rd, Rn, op_type);
    662 }
    663 void ArmToArm64Assembler::LDRSB(int /*cc*/, int /*Rd*/, int /*Rn*/, uint32_t /*offset*/)
    664 {
    665     NOT_IMPLEMENTED(); //Not required
    666 }
    667 void ArmToArm64Assembler::LDRSH(int /*cc*/, int /*Rd*/, int /*Rn*/, uint32_t /*offset*/)
    668 {
    669     NOT_IMPLEMENTED(); //Not required
    670 }
    671 
    672 void ArmToArm64Assembler::STRH(int cc, int Rd, int Rn, uint32_t op_type)
    673 {
    674     return dataTransfer(opSTRH, cc, Rd, Rn, op_type);
    675 }
    676 
    677 // ----------------------------------------------------------------------------
    678 // block data transfer...
    679 // ----------------------------------------------------------------------------
    680 void ArmToArm64Assembler::LDM(int cc, int dir,
    681         int Rn, int W, uint32_t reg_list)
    682 {
    683     const int XSP = 31;
    684     if(cc != AL || dir != IA || W == 0 || Rn != SP)
    685     {
    686         NOT_IMPLEMENTED();
    687         return;
    688     }
    689 
    690     for(int i = 0; i < 32; ++i)
    691     {
    692         if((reg_list & (1 << i)))
    693         {
    694             int reg = i;
    695             int size = 16;
    696             *mPC++ = A64_LDR_IMM_PostIndex(reg, XSP, size);
    697         }
    698     }
    699 }
    700 
    701 void ArmToArm64Assembler::STM(int cc, int dir,
    702         int Rn, int W, uint32_t reg_list)
    703 {
    704     const int XSP = 31;
    705     if(cc != AL || dir != DB || W == 0 || Rn != SP)
    706     {
    707         NOT_IMPLEMENTED();
    708         return;
    709     }
    710 
    711     for(int i = 31; i >= 0; --i)
    712     {
    713         if((reg_list & (1 << i)))
    714         {
    715             int size = -16;
    716             int reg  = i;
    717             *mPC++ = A64_STR_IMM_PreIndex(reg, XSP, size);
    718         }
    719     }
    720 }
    721 
    722 // ----------------------------------------------------------------------------
    723 // special...
    724 // ----------------------------------------------------------------------------
    725 void ArmToArm64Assembler::SWP(int /*cc*/, int /*Rn*/, int /*Rd*/, int /*Rm*/)
    726 {
    727     NOT_IMPLEMENTED(); //Not required
    728 }
    729 void ArmToArm64Assembler::SWPB(int /*cc*/, int /*Rn*/, int /*Rd*/, int /*Rm*/)
    730 {
    731     NOT_IMPLEMENTED(); //Not required
    732 }
    733 void ArmToArm64Assembler::SWI(int /*cc*/, uint32_t /*comment*/)
    734 {
    735     NOT_IMPLEMENTED(); //Not required
    736 }
    737 
    738 // ----------------------------------------------------------------------------
    739 // DSP instructions...
    740 // ----------------------------------------------------------------------------
    741 void ArmToArm64Assembler::PLD(int /*Rn*/, uint32_t /*offset*/) {
    742     NOT_IMPLEMENTED(); //Not required
    743 }
    744 
    745 void ArmToArm64Assembler::CLZ(int /*cc*/, int /*Rd*/, int /*Rm*/)
    746 {
    747     NOT_IMPLEMENTED(); //Not required
    748 }
    749 
    750 void ArmToArm64Assembler::QADD(int /*cc*/, int /*Rd*/, int /*Rm*/, int /*Rn*/)
    751 {
    752     NOT_IMPLEMENTED(); //Not required
    753 }
    754 
    755 void ArmToArm64Assembler::QDADD(int /*cc*/, int /*Rd*/, int /*Rm*/, int /*Rn*/)
    756 {
    757     NOT_IMPLEMENTED(); //Not required
    758 }
    759 
    760 void ArmToArm64Assembler::QSUB(int /*cc*/, int /*Rd*/, int /*Rm*/, int /*Rn*/)
    761 {
    762     NOT_IMPLEMENTED(); //Not required
    763 }
    764 
    765 void ArmToArm64Assembler::QDSUB(int /*cc*/, int /*Rd*/, int /*Rm*/, int /*Rn*/)
    766 {
    767     NOT_IMPLEMENTED(); //Not required
    768 }
    769 
    770 // ----------------------------------------------------------------------------
    771 // 16 x 16 multiplication
    772 // ----------------------------------------------------------------------------
    773 void ArmToArm64Assembler::SMUL(int cc, int xy,
    774                 int Rd, int Rm, int Rs)
    775 {
    776     if(cc != AL){ NOT_IMPLEMENTED(); return;} //Not required
    777 
    778     if (xy & xyTB)
    779         *mPC++ = A64_SBFM_W(mTmpReg1, Rm, 16, 31);
    780     else
    781         *mPC++ = A64_SBFM_W(mTmpReg1, Rm, 0, 15);
    782 
    783     if (xy & xyBT)
    784         *mPC++ = A64_SBFM_W(mTmpReg2, Rs, 16, 31);
    785     else
    786         *mPC++ = A64_SBFM_W(mTmpReg2, Rs, 0, 15);
    787 
    788     *mPC++ = A64_MADD_W(Rd,mTmpReg1,mTmpReg2, mZeroReg);
    789 }
    790 // ----------------------------------------------------------------------------
    791 // 32 x 16 multiplication
    792 // ----------------------------------------------------------------------------
    793 void ArmToArm64Assembler::SMULW(int cc, int y, int Rd, int Rm, int Rs)
    794 {
    795     if(cc != AL){ NOT_IMPLEMENTED(); return;} //Not required
    796 
    797     if (y & yT)
    798         *mPC++ = A64_SBFM_W(mTmpReg1, Rs, 16, 31);
    799     else
    800         *mPC++ = A64_SBFM_W(mTmpReg1, Rs, 0, 15);
    801 
    802     *mPC++ = A64_SBFM_W(mTmpReg2, Rm, 0, 31);
    803     *mPC++ = A64_SMADDL(mTmpReg3,mTmpReg1,mTmpReg2, mZeroReg);
    804     *mPC++ = A64_UBFM_X(Rd,mTmpReg3, 16, 47);
    805 }
    806 // ----------------------------------------------------------------------------
    807 // 16 x 16 multiplication and accumulate
    808 // ----------------------------------------------------------------------------
    809 void ArmToArm64Assembler::SMLA(int cc, int xy, int Rd, int Rm, int Rs, int Rn)
    810 {
    811     if(cc != AL){ NOT_IMPLEMENTED(); return;} //Not required
    812     if(xy != xyBB) { NOT_IMPLEMENTED(); return;} //Not required
    813 
    814     *mPC++ = A64_SBFM_W(mTmpReg1, Rm, 0, 15);
    815     *mPC++ = A64_SBFM_W(mTmpReg2, Rs, 0, 15);
    816     *mPC++ = A64_MADD_W(Rd, mTmpReg1, mTmpReg2, Rn);
    817 }
    818 
    819 void ArmToArm64Assembler::SMLAL(int /*cc*/, int /*xy*/,
    820                 int /*RdHi*/, int /*RdLo*/, int /*Rs*/, int /*Rm*/)
    821 {
    822     NOT_IMPLEMENTED(); //Not required
    823     return;
    824 }
    825 
    826 void ArmToArm64Assembler::SMLAW(int /*cc*/, int /*y*/,
    827                 int /*Rd*/, int /*Rm*/, int /*Rs*/, int /*Rn*/)
    828 {
    829     NOT_IMPLEMENTED(); //Not required
    830     return;
    831 }
    832 
    833 // ----------------------------------------------------------------------------
    834 // Byte/half word extract and extend
    835 // ----------------------------------------------------------------------------
    836 void ArmToArm64Assembler::UXTB16(int cc, int Rd, int Rm, int rotate)
    837 {
    838     if(cc != AL){ NOT_IMPLEMENTED(); return;} //Not required
    839 
    840     *mPC++ = A64_EXTR_W(mTmpReg1, Rm, Rm, rotate * 8);
    841 
    842     uint32_t imm = 0x00FF00FF;
    843     *mPC++ = A64_MOVZ_W(mTmpReg2, imm & 0xFFFF, 0);
    844     *mPC++ = A64_MOVK_W(mTmpReg2, (imm >> 16) & 0x0000FFFF, 16);
    845     *mPC++ = A64_AND_W(Rd,mTmpReg1, mTmpReg2);
    846 }
    847 
    848 // ----------------------------------------------------------------------------
    849 // Bit manipulation
    850 // ----------------------------------------------------------------------------
    851 void ArmToArm64Assembler::UBFX(int cc, int Rd, int Rn, int lsb, int width)
    852 {
    853     if(cc != AL){ NOT_IMPLEMENTED(); return;} //Not required
    854     *mPC++ = A64_UBFM_W(Rd, Rn, lsb, lsb + width - 1);
    855 }
    856 // ----------------------------------------------------------------------------
    857 // Shifters...
    858 // ----------------------------------------------------------------------------
    859 int ArmToArm64Assembler::buildImmediate(
    860         uint32_t immediate, uint32_t& rot, uint32_t& imm)
    861 {
    862     rot = 0;
    863     imm = immediate;
    864     return 0; // Always true
    865 }
    866 
    867 
    868 bool ArmToArm64Assembler::isValidImmediate(uint32_t immediate)
    869 {
    870     uint32_t rot, imm;
    871     return buildImmediate(immediate, rot, imm) == 0;
    872 }
    873 
    874 uint32_t ArmToArm64Assembler::imm(uint32_t immediate)
    875 {
    876     mAddrMode.immediate = immediate;
    877     mAddrMode.writeback = false;
    878     mAddrMode.preindex  = false;
    879     mAddrMode.postindex = false;
    880     return OPERAND_IMM;
    881 
    882 }
    883 
    884 uint32_t ArmToArm64Assembler::reg_imm(int Rm, int type, uint32_t shift)
    885 {
    886     mAddrMode.reg_imm_Rm = Rm;
    887     mAddrMode.reg_imm_type = type;
    888     mAddrMode.reg_imm_shift = shift;
    889     return OPERAND_REG_IMM;
    890 }
    891 
    892 uint32_t ArmToArm64Assembler::reg_rrx(int /*Rm*/)
    893 {
    894     NOT_IMPLEMENTED();
    895     return OPERAND_UNSUPPORTED;
    896 }
    897 
    898 uint32_t ArmToArm64Assembler::reg_reg(int /*Rm*/, int /*type*/, int /*Rs*/)
    899 {
    900     NOT_IMPLEMENTED(); //Not required
    901     return OPERAND_UNSUPPORTED;
    902 }
    903 // ----------------------------------------------------------------------------
    904 // Addressing modes...
    905 // ----------------------------------------------------------------------------
    906 uint32_t ArmToArm64Assembler::immed12_pre(int32_t immed12, int W)
    907 {
    908     mAddrMode.immediate = immed12;
    909     mAddrMode.writeback = W;
    910     mAddrMode.preindex  = true;
    911     mAddrMode.postindex = false;
    912     return OPERAND_IMM;
    913 }
    914 
    915 uint32_t ArmToArm64Assembler::immed12_post(int32_t immed12)
    916 {
    917     mAddrMode.immediate = immed12;
    918     mAddrMode.writeback = true;
    919     mAddrMode.preindex  = false;
    920     mAddrMode.postindex = true;
    921     return OPERAND_IMM;
    922 }
    923 
    924 uint32_t ArmToArm64Assembler::reg_scale_pre(int Rm, int type,
    925         uint32_t shift, int W)
    926 {
    927     if(type != 0 || shift != 0 || W != 0)
    928     {
    929         NOT_IMPLEMENTED(); //Not required
    930         return OPERAND_UNSUPPORTED;
    931     }
    932     else
    933     {
    934         mAddrMode.reg_offset = Rm;
    935         return OPERAND_REG_OFFSET;
    936     }
    937 }
    938 
    939 uint32_t ArmToArm64Assembler::reg_scale_post(int /*Rm*/, int /*type*/, uint32_t /*shift*/)
    940 {
    941     NOT_IMPLEMENTED(); //Not required
    942     return OPERAND_UNSUPPORTED;
    943 }
    944 
    945 uint32_t ArmToArm64Assembler::immed8_pre(int32_t immed8, int W)
    946 {
    947     mAddrMode.immediate = immed8;
    948     mAddrMode.writeback = W;
    949     mAddrMode.preindex  = true;
    950     mAddrMode.postindex = false;
    951     return OPERAND_IMM;
    952 }
    953 
    954 uint32_t ArmToArm64Assembler::immed8_post(int32_t immed8)
    955 {
    956     mAddrMode.immediate = immed8;
    957     mAddrMode.writeback = true;
    958     mAddrMode.preindex  = false;
    959     mAddrMode.postindex = true;
    960     return OPERAND_IMM;
    961 }
    962 
    963 uint32_t ArmToArm64Assembler::reg_pre(int Rm, int W)
    964 {
    965     if(W != 0)
    966     {
    967         NOT_IMPLEMENTED(); //Not required
    968         return OPERAND_UNSUPPORTED;
    969     }
    970     else
    971     {
    972         mAddrMode.reg_offset = Rm;
    973         return OPERAND_REG_OFFSET;
    974     }
    975 }
    976 
    977 uint32_t ArmToArm64Assembler::reg_post(int /*Rm*/)
    978 {
    979     NOT_IMPLEMENTED(); //Not required
    980     return OPERAND_UNSUPPORTED;
    981 }
    982 
    983 // ----------------------------------------------------------------------------
    984 // A64 instructions
    985 // ----------------------------------------------------------------------------
    986 
    987 static const char * dataTransferOpName[] =
    988 {
    989     "LDR","LDRB","LDRH","STR","STRB","STRH"
    990 };
    991 
    992 static const uint32_t dataTransferOpCode [] =
    993 {
    994     ((0xB8u << 24) | (0x3 << 21) | (0x6 << 13) | (0x0 << 12) |(0x1 << 11)),
    995     ((0x38u << 24) | (0x3 << 21) | (0x6 << 13) | (0x1 << 12) |(0x1 << 11)),
    996     ((0x78u << 24) | (0x3 << 21) | (0x6 << 13) | (0x0 << 12) |(0x1 << 11)),
    997     ((0xB8u << 24) | (0x1 << 21) | (0x6 << 13) | (0x0 << 12) |(0x1 << 11)),
    998     ((0x38u << 24) | (0x1 << 21) | (0x6 << 13) | (0x1 << 12) |(0x1 << 11)),
    999     ((0x78u << 24) | (0x1 << 21) | (0x6 << 13) | (0x0 << 12) |(0x1 << 11))
   1000 };
   1001 uint32_t ArmToArm64Assembler::A64_LDRSTR_Wm_SXTW_0(uint32_t op,
   1002                             uint32_t size, uint32_t Rt,
   1003                             uint32_t Rn, uint32_t Rm)
   1004 {
   1005     if(size == 32)
   1006     {
   1007         LOG_INSTR("%s W%d, [X%d, W%d, SXTW #0]\n",
   1008                    dataTransferOpName[op], Rt, Rn, Rm);
   1009         return(dataTransferOpCode[op] | (Rm << 16) | (Rn << 5) | Rt);
   1010     }
   1011     else
   1012     {
   1013         LOG_INSTR("%s X%d, [X%d, W%d, SXTW #0]\n",
   1014                   dataTransferOpName[op], Rt, Rn, Rm);
   1015         return(dataTransferOpCode[op] | (0x1<<30) | (Rm<<16) | (Rn<<5)|Rt);
   1016     }
   1017 }
   1018 
   1019 uint32_t ArmToArm64Assembler::A64_STR_IMM_PreIndex(uint32_t Rt,
   1020                             uint32_t Rn, int32_t simm)
   1021 {
   1022     if(Rn == 31)
   1023         LOG_INSTR("STR W%d, [SP, #%d]!\n", Rt, simm);
   1024     else
   1025         LOG_INSTR("STR W%d, [X%d, #%d]!\n", Rt, Rn, simm);
   1026 
   1027     uint32_t imm9 = (unsigned)(simm) & 0x01FF;
   1028     return (0xB8 << 24) | (imm9 << 12) | (0x3 << 10) | (Rn << 5) | Rt;
   1029 }
   1030 
   1031 uint32_t ArmToArm64Assembler::A64_LDR_IMM_PostIndex(uint32_t Rt,
   1032                             uint32_t Rn, int32_t simm)
   1033 {
   1034     if(Rn == 31)
   1035         LOG_INSTR("LDR W%d, [SP], #%d\n",Rt,simm);
   1036     else
   1037         LOG_INSTR("LDR W%d, [X%d], #%d\n",Rt, Rn, simm);
   1038 
   1039     uint32_t imm9 = (unsigned)(simm) & 0x01FF;
   1040     return (0xB8 << 24) | (0x1 << 22) |
   1041              (imm9 << 12) | (0x1 << 10) | (Rn << 5) | Rt;
   1042 
   1043 }
   1044 uint32_t ArmToArm64Assembler::A64_ADD_X_Wm_SXTW(uint32_t Rd,
   1045                                uint32_t Rn,
   1046                                uint32_t Rm,
   1047                                uint32_t amount)
   1048 {
   1049     LOG_INSTR("ADD X%d, X%d, W%d, SXTW #%d\n", Rd, Rn, Rm, amount);
   1050     return ((0x8B << 24) | (0x1 << 21) |(Rm << 16) |
   1051               (0x6 << 13) | (amount << 10) | (Rn << 5) | Rd);
   1052 
   1053 }
   1054 
   1055 uint32_t ArmToArm64Assembler::A64_SUB_X_Wm_SXTW(uint32_t Rd,
   1056                                uint32_t Rn,
   1057                                uint32_t Rm,
   1058                                uint32_t amount)
   1059 {
   1060     LOG_INSTR("SUB X%d, X%d, W%d, SXTW #%d\n", Rd, Rn, Rm, amount);
   1061     return ((0xCB << 24) | (0x1 << 21) |(Rm << 16) |
   1062             (0x6 << 13) | (amount << 10) | (Rn << 5) | Rd);
   1063 
   1064 }
   1065 
   1066 uint32_t ArmToArm64Assembler::A64_B_COND(uint32_t cc, uint32_t offset)
   1067 {
   1068     LOG_INSTR("B.%s #.+%d\n", cc_codes[cc], offset);
   1069     return (0x54 << 24) | ((offset/4) << 5) | (cc);
   1070 
   1071 }
   1072 uint32_t ArmToArm64Assembler::A64_ADD_X(uint32_t Rd, uint32_t Rn,
   1073                                           uint32_t Rm, uint32_t shift,
   1074                                           uint32_t amount)
   1075 {
   1076     LOG_INSTR("ADD X%d, X%d, X%d, %s #%d\n",
   1077                Rd, Rn, Rm, shift_codes[shift], amount);
   1078     return ((0x8B << 24) | (shift << 22) | ( Rm << 16) |
   1079             (amount << 10) |(Rn << 5) | Rd);
   1080 }
   1081 uint32_t ArmToArm64Assembler::A64_ADD_IMM_X(uint32_t Rd, uint32_t Rn,
   1082                                           uint32_t imm, uint32_t shift)
   1083 {
   1084     LOG_INSTR("ADD X%d, X%d, #%d, LSL #%d\n", Rd, Rn, imm, shift);
   1085     return (0x91 << 24) | ((shift/12) << 22) | (imm << 10) | (Rn << 5) | Rd;
   1086 }
   1087 
   1088 uint32_t ArmToArm64Assembler::A64_SUB_IMM_X(uint32_t Rd, uint32_t Rn,
   1089                                           uint32_t imm, uint32_t shift)
   1090 {
   1091     LOG_INSTR("SUB X%d, X%d, #%d, LSL #%d\n", Rd, Rn, imm, shift);
   1092     return (0xD1 << 24) | ((shift/12) << 22) | (imm << 10) | (Rn << 5) | Rd;
   1093 }
   1094 
   1095 uint32_t ArmToArm64Assembler::A64_ADD_W(uint32_t Rd, uint32_t Rn,
   1096                                           uint32_t Rm, uint32_t shift,
   1097                                           uint32_t amount)
   1098 {
   1099     LOG_INSTR("ADD W%d, W%d, W%d, %s #%d\n",
   1100                Rd, Rn, Rm, shift_codes[shift], amount);
   1101     return ((0x0B << 24) | (shift << 22) | ( Rm << 16) |
   1102             (amount << 10) |(Rn << 5) | Rd);
   1103 }
   1104 
   1105 uint32_t ArmToArm64Assembler::A64_SUB_W(uint32_t Rd, uint32_t Rn,
   1106                                           uint32_t Rm, uint32_t shift,
   1107                                           uint32_t amount,
   1108                                           uint32_t setflag)
   1109 {
   1110     if(setflag == 0)
   1111     {
   1112         LOG_INSTR("SUB W%d, W%d, W%d, %s #%d\n",
   1113                Rd, Rn, Rm, shift_codes[shift], amount);
   1114         return ((0x4B << 24) | (shift << 22) | ( Rm << 16) |
   1115                 (amount << 10) |(Rn << 5) | Rd);
   1116     }
   1117     else
   1118     {
   1119         LOG_INSTR("SUBS W%d, W%d, W%d, %s #%d\n",
   1120                    Rd, Rn, Rm, shift_codes[shift], amount);
   1121         return ((0x6B << 24) | (shift << 22) | ( Rm << 16) |
   1122                 (amount << 10) |(Rn << 5) | Rd);
   1123     }
   1124 }
   1125 
   1126 uint32_t ArmToArm64Assembler::A64_AND_W(uint32_t Rd, uint32_t Rn,
   1127                                           uint32_t Rm, uint32_t shift,
   1128                                           uint32_t amount)
   1129 {
   1130     LOG_INSTR("AND W%d, W%d, W%d, %s #%d\n",
   1131                Rd, Rn, Rm, shift_codes[shift], amount);
   1132     return ((0x0A << 24) | (shift << 22) | ( Rm << 16) |
   1133             (amount << 10) |(Rn << 5) | Rd);
   1134 }
   1135 
   1136 uint32_t ArmToArm64Assembler::A64_ORR_W(uint32_t Rd, uint32_t Rn,
   1137                                           uint32_t Rm, uint32_t shift,
   1138                                           uint32_t amount)
   1139 {
   1140     LOG_INSTR("ORR W%d, W%d, W%d, %s #%d\n",
   1141                Rd, Rn, Rm, shift_codes[shift], amount);
   1142     return ((0x2A << 24) | (shift << 22) | ( Rm << 16) |
   1143             (amount << 10) |(Rn << 5) | Rd);
   1144 }
   1145 
   1146 uint32_t ArmToArm64Assembler::A64_ORN_W(uint32_t Rd, uint32_t Rn,
   1147                                           uint32_t Rm, uint32_t shift,
   1148                                           uint32_t amount)
   1149 {
   1150     LOG_INSTR("ORN W%d, W%d, W%d, %s #%d\n",
   1151                Rd, Rn, Rm, shift_codes[shift], amount);
   1152     return ((0x2A << 24) | (shift << 22) | (0x1 << 21) | ( Rm << 16) |
   1153             (amount << 10) |(Rn << 5) | Rd);
   1154 }
   1155 
   1156 uint32_t ArmToArm64Assembler::A64_CSEL_X(uint32_t Rd, uint32_t Rn,
   1157                                            uint32_t Rm, uint32_t cond)
   1158 {
   1159     LOG_INSTR("CSEL X%d, X%d, X%d, %s\n", Rd, Rn, Rm, cc_codes[cond]);
   1160     return ((0x9A << 24)|(0x1 << 23)|(Rm << 16) |(cond << 12)| (Rn << 5) | Rd);
   1161 }
   1162 
   1163 uint32_t ArmToArm64Assembler::A64_CSEL_W(uint32_t Rd, uint32_t Rn,
   1164                                            uint32_t Rm, uint32_t cond)
   1165 {
   1166     LOG_INSTR("CSEL W%d, W%d, W%d, %s\n", Rd, Rn, Rm, cc_codes[cond]);
   1167     return ((0x1A << 24)|(0x1 << 23)|(Rm << 16) |(cond << 12)| (Rn << 5) | Rd);
   1168 }
   1169 
   1170 uint32_t ArmToArm64Assembler::A64_RET(uint32_t Rn)
   1171 {
   1172     LOG_INSTR("RET X%d\n", Rn);
   1173     return ((0xD6 << 24) | (0x1 << 22) | (0x1F << 16) | (Rn << 5));
   1174 }
   1175 
   1176 uint32_t ArmToArm64Assembler::A64_MOVZ_X(uint32_t Rd, uint32_t imm,
   1177                                          uint32_t shift)
   1178 {
   1179     LOG_INSTR("MOVZ X%d, #0x%x, LSL #%d\n", Rd, imm, shift);
   1180     return(0xD2 << 24) | (0x1 << 23) | ((shift/16) << 21) |  (imm << 5) | Rd;
   1181 }
   1182 
   1183 uint32_t ArmToArm64Assembler::A64_MOVK_W(uint32_t Rd, uint32_t imm,
   1184                                          uint32_t shift)
   1185 {
   1186     LOG_INSTR("MOVK W%d, #0x%x, LSL #%d\n", Rd, imm, shift);
   1187     return (0x72 << 24) | (0x1 << 23) | ((shift/16) << 21) | (imm << 5) | Rd;
   1188 }
   1189 
   1190 uint32_t ArmToArm64Assembler::A64_MOVZ_W(uint32_t Rd, uint32_t imm,
   1191                                          uint32_t shift)
   1192 {
   1193     LOG_INSTR("MOVZ W%d, #0x%x, LSL #%d\n", Rd, imm, shift);
   1194     return(0x52 << 24) | (0x1 << 23) | ((shift/16) << 21) |  (imm << 5) | Rd;
   1195 }
   1196 
   1197 uint32_t ArmToArm64Assembler::A64_SMADDL(uint32_t Rd, uint32_t Rn,
   1198                                            uint32_t Rm, uint32_t Ra)
   1199 {
   1200     LOG_INSTR("SMADDL X%d, W%d, W%d, X%d\n",Rd, Rn, Rm, Ra);
   1201     return ((0x9B << 24) | (0x1 << 21) | (Rm << 16)|(Ra << 10)|(Rn << 5) | Rd);
   1202 }
   1203 
   1204 uint32_t ArmToArm64Assembler::A64_MADD_W(uint32_t Rd, uint32_t Rn,
   1205                                            uint32_t Rm, uint32_t Ra)
   1206 {
   1207     LOG_INSTR("MADD W%d, W%d, W%d, W%d\n",Rd, Rn, Rm, Ra);
   1208     return ((0x1B << 24) | (Rm << 16) | (Ra << 10) |(Rn << 5) | Rd);
   1209 }
   1210 
   1211 uint32_t ArmToArm64Assembler::A64_SBFM_W(uint32_t Rd, uint32_t Rn,
   1212                                            uint32_t immr, uint32_t imms)
   1213 {
   1214     LOG_INSTR("SBFM W%d, W%d, #%d, #%d\n", Rd, Rn, immr, imms);
   1215     return ((0x13 << 24) | (immr << 16) | (imms << 10) | (Rn << 5) | Rd);
   1216 
   1217 }
   1218 uint32_t ArmToArm64Assembler::A64_UBFM_W(uint32_t Rd, uint32_t Rn,
   1219                                            uint32_t immr, uint32_t imms)
   1220 {
   1221     LOG_INSTR("UBFM W%d, W%d, #%d, #%d\n", Rd, Rn, immr, imms);
   1222     return ((0x53 << 24) | (immr << 16) | (imms << 10) | (Rn << 5) | Rd);
   1223 
   1224 }
   1225 uint32_t ArmToArm64Assembler::A64_UBFM_X(uint32_t Rd, uint32_t Rn,
   1226                                            uint32_t immr, uint32_t imms)
   1227 {
   1228     LOG_INSTR("UBFM X%d, X%d, #%d, #%d\n", Rd, Rn, immr, imms);
   1229     return ((0xD3 << 24) | (0x1 << 22) |
   1230             (immr << 16) | (imms << 10) | (Rn << 5) | Rd);
   1231 
   1232 }
   1233 uint32_t ArmToArm64Assembler::A64_EXTR_W(uint32_t Rd, uint32_t Rn,
   1234                                            uint32_t Rm, uint32_t lsb)
   1235 {
   1236     LOG_INSTR("EXTR W%d, W%d, W%d, #%d\n", Rd, Rn, Rm, lsb);
   1237     return (0x13 << 24)|(0x1 << 23) | (Rm << 16) | (lsb << 10)|(Rn << 5) | Rd;
   1238 }
   1239 
   1240 }; // namespace android
   1241 
   1242