mipssim.cc

Aller à la documentation de ce fichier.

// mipssim.cc -- simulate a MIPS R2/3000 processor
//
//   This code has been adapted from Ousterhout's MIPSSIM package.
//   Byte ordering is little-endian, so we can be compatible with
//   DEC RISC systems.
//
//   DO NOT CHANGE -- part of the machine emulation
//
// Copyright (c) 1992-1996 The Regents of the University of California.
// All rights reserved.  See copyright.h for copyright notice and limitation 
// of liability and disclaimer of warranty provisions.

// Simulation fixes done by Peter E Reissner, class of Winter 1994/95 (York)
// I've not been able to test this extensively.
// Ported to newer version of Nachos at Waterloo by Scott Graham (Mar 99).


#include "copyright.h"

#include "debug.h"
#include "machine.h"
#include "mipssim.h"
#include "main.h"

static void Mult(int a, int b, bool signedArith, int* hiPtr, int* loPtr);

// The following class defines an instruction, represented in both
//      undecoded binary form
//      decoded to identify
//          operation to do
//          registers to act on
//          any immediate operand value

class Instruction {
  public:
    void Decode();      // decode the binary representation of the instruction

    unsigned int value; // binary representation of the instruction

    char opCode;     // Type of instruction.  This is NOT the same as the
                     // opcode field from the instruction: see defs in mips.h
    char rs, rt, rd; // Three registers from instruction.
    int extra;       // Immediate or target or shamt field or offset.
                     // Immediates are sign-extended.
};

//----------------------------------------------------------------------
// Machine::Run
//      Simulate the execution of a user-level program on Nachos.
//      Called by the kernel when the program starts up; never returns.
//
//      This routine is re-entrant, in that it can be called multiple
//      times concurrently -- one for each thread executing user code.
//----------------------------------------------------------------------

void
Machine::Run()
{
    Instruction *instr = new Instruction;  // storage for decoded instruction

    if (debug->IsEnabled('m')) {
        cerr << "Starting program in thread: " << kernel->currentThread->getName();
        cerr << ", at time: " << kernel->stats->totalTicks << "\n";
    }
    kernel->interrupt->setStatus(UserMode);
    for (;;) {
        OneInstruction(instr);
        kernel->interrupt->OneTick();
        if (singleStep && (runUntilTime <= kernel->stats->totalTicks))
          Debugger();
    }
}


//----------------------------------------------------------------------
// TypeToReg
//      Retrieve the register # referred to in an instruction. 
//----------------------------------------------------------------------

static int 
TypeToReg(RegType reg, Instruction *instr)
{
    switch (reg) {
      case RS:
        return instr->rs;
      case RT:
        return instr->rt;
      case RD:
        return instr->rd;
      case EXTRA:
        return instr->extra;
      default:
        return -1;
    }
}

//----------------------------------------------------------------------
// Machine::OneInstruction
//      Execute one instruction from a user-level program
//
//      If there is any kind of exception or interrupt, we invoke the 
//      exception handler, and when it returns, we return to Run(), which
//      will re-invoke us in a loop.  This allows us to
//      re-start the instruction execution from the beginning, in
//      case any of our state has changed.  On a syscall,
//      the OS software must increment the PC so execution begins
//      at the instruction immediately after the syscall. 
//
//      This routine is re-entrant, in that it can be called multiple
//      times concurrently -- one for each thread executing user code.
//      We get re-entrancy by never caching any data -- we always re-start the
//      simulation from scratch each time we are called (or after trapping
//      back to the Nachos kernel on an exception or interrupt), and we always
//      store all data back to the machine registers and memory before
//      leaving.  This allows the Nachos kernel to control our behavior
//      by controlling the contents of memory, the translation table,
//      and the register set.
//----------------------------------------------------------------------

void
Machine::OneInstruction(Instruction *instr)
{
#ifdef SIM_FIX
    int byte;       // described in Kane for LWL,LWR,...
#endif

    int raw;
    int nextLoadReg = 0;        
    int nextLoadValue = 0;      // record delayed load operation, to apply
                                // in the future

    // Fetch instruction 
    if (!ReadMem(registers[PCReg], 4, &raw))
        return;                 // exception occurred
    instr->value = raw;
    instr->Decode();

    if (debug->IsEnabled('m')) {
        struct OpString *str = &opStrings[instr->opCode];
        char buf[80];

        ASSERT(instr->opCode <= MaxOpcode);
        cout << "At PC = " << registers[PCReg];
        sprintf(buf, str->format, TypeToReg(str->args[0], instr),
             TypeToReg(str->args[1], instr), TypeToReg(str->args[2], instr));
        cerr << "\t" << buf << "\n";
    }
    
    // Compute next pc, but don't install in case there's an error or branch.
    int pcAfter = registers[NextPCReg] + 4;
    int sum, diff, tmp, value;
    unsigned int rs, rt, imm;

    // Execute the instruction (cf. Kane's book)
    switch (instr->opCode) {
        
      case OP_ADD:
        sum = registers[instr->rs] + registers[instr->rt];
        if (!((registers[instr->rs] ^ registers[instr->rt]) & SIGN_BIT) &&
            ((registers[instr->rs] ^ sum) & SIGN_BIT)) {
            RaiseException(OverflowException, 0);
            return;
        }
        registers[instr->rd] = sum;
        break;
        
      case OP_ADDI:
        sum = registers[instr->rs] + instr->extra;
        if (!((registers[instr->rs] ^ instr->extra) & SIGN_BIT) &&
            ((instr->extra ^ sum) & SIGN_BIT)) {
            RaiseException(OverflowException, 0);
            return;
        }
        registers[instr->rt] = sum;
        break;
        
      case OP_ADDIU:
        registers[instr->rt] = registers[instr->rs] + instr->extra;
        break;
        
      case OP_ADDU:
        registers[instr->rd] = registers[instr->rs] + registers[instr->rt];
        break;
        
      case OP_AND:
        registers[instr->rd] = registers[instr->rs] & registers[instr->rt];
        break;
        
      case OP_ANDI:
        registers[instr->rt] = registers[instr->rs] & (instr->extra & 0xffff);
        break;
        
      case OP_BEQ:
        if (registers[instr->rs] == registers[instr->rt])
            pcAfter = registers[NextPCReg] + IndexToAddr(instr->extra);
        break;
        
      case OP_BGEZAL:
        registers[R31] = registers[NextPCReg] + 4;
      case OP_BGEZ:
        if (!(registers[instr->rs] & SIGN_BIT))
            pcAfter = registers[NextPCReg] + IndexToAddr(instr->extra);
        break;
        
      case OP_BGTZ:
        if (registers[instr->rs] > 0)
            pcAfter = registers[NextPCReg] + IndexToAddr(instr->extra);
        break;
        
      case OP_BLEZ:
        if (registers[instr->rs] <= 0)
            pcAfter = registers[NextPCReg] + IndexToAddr(instr->extra);
        break;
        
      case OP_BLTZAL:
        registers[R31] = registers[NextPCReg] + 4;
      case OP_BLTZ:
        if (registers[instr->rs] & SIGN_BIT)
            pcAfter = registers[NextPCReg] + IndexToAddr(instr->extra);
        break;
        
      case OP_BNE:
        if (registers[instr->rs] != registers[instr->rt])
            pcAfter = registers[NextPCReg] + IndexToAddr(instr->extra);
        break;
        
      case OP_DIV:
        if (registers[instr->rt] == 0) {
            registers[LoReg] = 0;
            registers[HiReg] = 0;
        } else {
            registers[LoReg] =  registers[instr->rs] / registers[instr->rt];
            registers[HiReg] = registers[instr->rs] % registers[instr->rt];
        }
        break;
        
      case OP_DIVU:       
          rs = (unsigned int) registers[instr->rs];
          rt = (unsigned int) registers[instr->rt];
          if (rt == 0) {
              registers[LoReg] = 0;
              registers[HiReg] = 0;
          } else {
              tmp = rs / rt;
              registers[LoReg] = (int) tmp;
              tmp = rs % rt;
              registers[HiReg] = (int) tmp;
          }
          break;
        
      case OP_JAL:
        registers[R31] = registers[NextPCReg] + 4;
      case OP_J:
        pcAfter = (pcAfter & 0xf0000000) | IndexToAddr(instr->extra);
        break;
        
      case OP_JALR:
        registers[instr->rd] = registers[NextPCReg] + 4;
      case OP_JR:
        pcAfter = registers[instr->rs];
        break;
        
      case OP_LB:
      case OP_LBU:
        tmp = registers[instr->rs] + instr->extra;
        if (!ReadMem(tmp, 1, &value))
            return;

        if ((value & 0x80) && (instr->opCode == OP_LB))
            value |= 0xffffff00;
        else
            value &= 0xff;
        nextLoadReg = instr->rt;
        nextLoadValue = value;
        break;
        
      case OP_LH:
      case OP_LHU:        
        tmp = registers[instr->rs] + instr->extra;
        if (tmp & 0x1) {
            RaiseException(AddressErrorException, tmp);
            return;
        }
        if (!ReadMem(tmp, 2, &value))
            return;

        if ((value & 0x8000) && (instr->opCode == OP_LH))
            value |= 0xffff0000;
        else
            value &= 0xffff;
        nextLoadReg = instr->rt;
        nextLoadValue = value;
        break;
        
      case OP_LUI:
        DEBUG(dbgMach, "Executing: LUI r" << instr->rt << ", " << instr->extra);
        registers[instr->rt] = instr->extra << 16;
        break;
        
      case OP_LW:
        tmp = registers[instr->rs] + instr->extra;
        if (tmp & 0x3) {
            RaiseException(AddressErrorException, tmp);
            return;
        }
        if (!ReadMem(tmp, 4, &value))
            return;
        nextLoadReg = instr->rt;
        nextLoadValue = value;
        break;
        
      case OP_LWL:        
        tmp = registers[instr->rs] + instr->extra;

#ifdef SIM_FIX
        // The only difference between this code and the BIG ENDIAN code
        // is that the ReadMem call is guaranteed an aligned access as it
        // should be (Kane's book hides the fact that all memory access
        // are done using aligned loads - what the instruction asks for
        // is a arbitrary) This is the whole purpose of LWL and LWR etc.
        // Then the switch uses  3 - (tmp & 0x3)  instead of (tmp & 0x3)

        byte = tmp & 0x3;
        // DEBUG('P', "Addr 0x%X\n",tmp-byte);

        if (!ReadMem(tmp-byte, 4, &value))
            return;
#else
        // ReadMem assumes all 4 byte requests are aligned on an even 
        // word boundary.  Also, the little endian/big endian swap code would
        // fail (I think) if the other cases are ever exercised.
        ASSERT((tmp & 0x3) == 0);  

        if (!ReadMem(tmp, 4, &value))
            return;
#endif

        if (registers[LoadReg] == instr->rt)
            nextLoadValue = registers[LoadValueReg];
        else
            nextLoadValue = registers[instr->rt];
#ifdef SIM_FIX
        switch (3 - byte) 
#else
        switch (tmp & 0x3)
#endif
          {
          case 0:
            nextLoadValue = value;
            break;
          case 1:
            nextLoadValue = (nextLoadValue & 0xff) | (value << 8);
            break;
          case 2:
            nextLoadValue = (nextLoadValue & 0xffff) | (value << 16);
            break;
          case 3:
            nextLoadValue = (nextLoadValue & 0xffffff) | (value << 24);
            break;
        }
        nextLoadReg = instr->rt;
        break;
        
      case OP_LWR:
        tmp = registers[instr->rs] + instr->extra;

#ifdef SIM_FIX
        // The only difference between this code and the BIG ENDIAN code
        // is that the ReadMem call is guaranteed an aligned access as it
        // should be (Kane's book hides the fact that all memory access
        // are done using aligned loads - what the instruction asks 
        // for is a arbitrary) This is the whole purpose of LWL and LWR etc.
        // Then the switch uses  3 - (tmp & 0x3)  instead of (tmp & 0x3)

        byte = tmp & 0x3;
        // DEBUG('P', "Addr 0x%X\n",tmp-byte);

        if (!ReadMem(tmp-byte, 4, &value))
            return;
#else
        // ReadMem assumes all 4 byte requests are aligned on an even 
        // word boundary.  Also, the little endian/big endian swap code would
        // fail (I think) if the other cases are ever exercised.
        ASSERT((tmp & 0x3) == 0);  

        if (!ReadMem(tmp, 4, &value))
            return;
#endif

        if (registers[LoadReg] == instr->rt)
            nextLoadValue = registers[LoadValueReg];
        else
            nextLoadValue = registers[instr->rt];

#ifdef SIM_FIX
        switch (3 - byte) 
#else
        switch (tmp & 0x3)
#endif
          {
          case 0:
            nextLoadValue = (nextLoadValue & 0xffffff00) |
                ((value >> 24) & 0xff);
            break;
          case 1:
            nextLoadValue = (nextLoadValue & 0xffff0000) |
                ((value >> 16) & 0xffff);
            break;
          case 2:
            nextLoadValue = (nextLoadValue & 0xff000000)
                | ((value >> 8) & 0xffffff);
            break;
          case 3:
            nextLoadValue = value;
            break;
        }
        nextLoadReg = instr->rt;
        break;
        
      case OP_MFHI:
        registers[instr->rd] = registers[HiReg];
        break;
        
      case OP_MFLO:
        registers[instr->rd] = registers[LoReg];
        break;
        
      case OP_MTHI:
        registers[HiReg] = registers[instr->rs];
        break;
        
      case OP_MTLO:
        registers[LoReg] = registers[instr->rs];
        break;
        
      case OP_MULT:
        Mult(registers[instr->rs], registers[instr->rt], TRUE,
             &registers[HiReg], &registers[LoReg]);
        break;
        
      case OP_MULTU:
        Mult(registers[instr->rs], registers[instr->rt], FALSE,
             &registers[HiReg], &registers[LoReg]);
        break;
        
      case OP_NOR:
        registers[instr->rd] = ~(registers[instr->rs] | registers[instr->rt]);
        break;
        
      case OP_OR:
        registers[instr->rd] = registers[instr->rs] | registers[instr->rt];
        break;
        
      case OP_ORI:
        registers[instr->rt] = registers[instr->rs] | (instr->extra & 0xffff);
        break;
        
      case OP_SB:
        if (!WriteMem((unsigned) 
                (registers[instr->rs] + instr->extra), 1, registers[instr->rt]))
            return;
        break;
        
      case OP_SH:
        if (!WriteMem((unsigned) 
                (registers[instr->rs] + instr->extra), 2, registers[instr->rt]))
            return;
        break;
        
      case OP_SLL:
        registers[instr->rd] = registers[instr->rt] << instr->extra;
        break;
        
      case OP_SLLV:
        registers[instr->rd] = registers[instr->rt] <<
            (registers[instr->rs] & 0x1f);
        break;
        
      case OP_SLT:
        if (registers[instr->rs] < registers[instr->rt])
            registers[instr->rd] = 1;
        else
            registers[instr->rd] = 0;
        break;
        
      case OP_SLTI:
        if (registers[instr->rs] < instr->extra)
            registers[instr->rt] = 1;
        else
            registers[instr->rt] = 0;
        break;
        
      case OP_SLTIU:      
        rs = registers[instr->rs];
        imm = instr->extra;
        if (rs < imm)
            registers[instr->rt] = 1;
        else
            registers[instr->rt] = 0;
        break;
        
      case OP_SLTU:       
        rs = registers[instr->rs];
        rt = registers[instr->rt];
        if (rs < rt)
            registers[instr->rd] = 1;
        else
            registers[instr->rd] = 0;
        break;
        
      case OP_SRA:
        registers[instr->rd] = registers[instr->rt] >> instr->extra;
        break;
        
      case OP_SRAV:
        registers[instr->rd] = registers[instr->rt] >>
            (registers[instr->rs] & 0x1f);
        break;
        
      case OP_SRL:
        tmp = registers[instr->rt];
        tmp >>= instr->extra;
        registers[instr->rd] = tmp;
        break;
        
      case OP_SRLV:
        tmp = registers[instr->rt];
        tmp >>= (registers[instr->rs] & 0x1f);
        registers[instr->rd] = tmp;
        break;
        
      case OP_SUB:        
        diff = registers[instr->rs] - registers[instr->rt];
        if (((registers[instr->rs] ^ registers[instr->rt]) & SIGN_BIT) &&
            ((registers[instr->rs] ^ diff) & SIGN_BIT)) {
            RaiseException(OverflowException, 0);
            return;
        }
        registers[instr->rd] = diff;
        break;
        
      case OP_SUBU:
        registers[instr->rd] = registers[instr->rs] - registers[instr->rt];
        break;
        
      case OP_SW:
        if (!WriteMem((unsigned) 
                (registers[instr->rs] + instr->extra), 4, registers[instr->rt]))
            return;
        break;
        
      case OP_SWL:        
        tmp = registers[instr->rs] + instr->extra;

#ifdef SIM_FIX
        // The only difference between this code and the BIG ENDIAN code
        // is that the ReadMem call is guaranteed an aligned access as it
        // should be (Kane's book hides the fact that all memory access
        // are done using aligned loads - what the instruction asks for
        // is a arbitrary) This is the whole purpose of LWL and LWR etc.

        byte = tmp & 0x3;
        // DEBUG('P', "Addr 0x%X\n",tmp-byte);
        if (!ReadMem(tmp-byte, 4, &value))
            return;

        // DEBUG('P', "Value 0x%X\n",value);
#else

        // The little endian/big endian swap code would
        // fail (I think) if the other cases are ever exercised.
        ASSERT((tmp & 0x3) == 0);  

        if (!ReadMem((tmp & ~0x3), 4, &value))
            return;
#endif

#ifdef SIM_FIX
        switch( 3 - byte )
#else
          switch (tmp & 0x3) 
#endif // SIM_FIX
            {
          case 0:
            value = registers[instr->rt];
            break;
          case 1:
            value = (value & 0xff000000) | ((registers[instr->rt] >> 8) &
                                            0xffffff);
            break;
          case 2:
            value = (value & 0xffff0000) | ((registers[instr->rt] >> 16) &
                                            0xffff);
            break;
          case 3:
            value = (value & 0xffffff00) | ((registers[instr->rt] >> 24) &
                                            0xff);
            break;
        }
#ifndef SIM_FIX
        if (!WriteMem((tmp & ~0x3), 4, value))
            return;
#else
        // DEBUG('P', "Value 0x%X\n",value);

        if (!WriteMem((tmp - byte), 4, value))
            return;
#endif // SIM_FIX
        break;
        
      case OP_SWR:        
        tmp = registers[instr->rs] + instr->extra;

#ifndef SIM_FIX
        // The little endian/big endian swap code would
        // fail (I think) if the other cases are ever exercised.
        ASSERT((tmp & 0x3) == 0);  

        if (!ReadMem((tmp & ~0x3), 4, &value))
            return;
#else
        // The only difference between this code and the BIG ENDIAN code
        // is that the ReadMem call is guaranteed an aligned access as 
        // it should be (Kane's book hides the fact that all memory 
        // access are done using aligned loads - what the instruction 
        // asks for is a arbitrary) This is the whole purpose of LWL 
        // and LWR etc.

        byte = tmp & 0x3;
        // DEBUG('P', "Addr 0x%X\n",tmp-byte);

        if (!ReadMem(tmp-byte, 4, &value))
            return;
        // DEBUG('P', "Value 0x%X\n",value);
#endif // SIM_FIX

#ifndef SIM_FIX
        switch (tmp & 0x3) 
#else
          switch( 3 - byte ) 
#endif // SIM_FIX
            {
            case 0:
            value = (value & 0xffffff) | (registers[instr->rt] << 24);
            break;
          case 1:
            value = (value & 0xffff) | (registers[instr->rt] << 16);
            break;
          case 2:
            value = (value & 0xff) | (registers[instr->rt] << 8);
            break;
          case 3:
            value = registers[instr->rt];
            break;
        }

#ifndef SIM_FIX
        if (!WriteMem((tmp & ~0x3), 4, value))
            return;
#else
        // DEBUG('P', "Value 0x%X\n",value);

        if (!WriteMem((tmp - byte), 4, value))
            return;
#endif // SIM_FIX


        break;
        
      case OP_SYSCALL:
        RaiseException(SyscallException, 0);
        return; 
        
      case OP_XOR:
        registers[instr->rd] = registers[instr->rs] ^ registers[instr->rt];
        break;
        
      case OP_XORI:
        registers[instr->rt] = registers[instr->rs] ^ (instr->extra & 0xffff);
        break;
        
      case OP_RES:
      case OP_UNIMP:
        RaiseException(IllegalInstrException, 0);
        return;
        
      default:
        ASSERT(FALSE);
    }
    
    // Now we have successfully executed the instruction.
    
    // Do any delayed load operation
    DelayedLoad(nextLoadReg, nextLoadValue);
    
    // Advance program counters.
    registers[PrevPCReg] = registers[PCReg];    // for debugging, in case we
                                                // are jumping into lala-land
    registers[PCReg] = registers[NextPCReg];
    registers[NextPCReg] = pcAfter;
}

//----------------------------------------------------------------------
// Machine::DelayedLoad
//      Simulate effects of a delayed load.
//
//      NOTE -- RaiseException/CheckInterrupts must also call DelayedLoad,
//      since any delayed load must get applied before we trap to the kernel.
//----------------------------------------------------------------------

void
Machine::DelayedLoad(int nextReg, int nextValue)
{
    registers[registers[LoadReg]] = registers[LoadValueReg];
    registers[LoadReg] = nextReg;
    registers[LoadValueReg] = nextValue;
    registers[0] = 0;   // and always make sure R0 stays zero.
}

//----------------------------------------------------------------------
// Instruction::Decode
//      Decode a MIPS instruction 
//----------------------------------------------------------------------

void
Instruction::Decode()
{
    OpInfo *opPtr;
    
    rs = (value >> 21) & 0x1f;
    rt = (value >> 16) & 0x1f;
    rd = (value >> 11) & 0x1f;
    opPtr = &opTable[(value >> 26) & 0x3f];
    opCode = opPtr->opCode;
    if (opPtr->format == IFMT) {
        extra = value & 0xffff;
        if (extra & 0x8000) {
           extra |= 0xffff0000;
        }
    } else if (opPtr->format == RFMT) {
        extra = (value >> 6) & 0x1f;
    } else {
        extra = value & 0x3ffffff;
    }
    if (opCode == SPECIAL) {
        opCode = specialTable[value & 0x3f];
    } else if (opCode == BCOND) {
        int i = value & 0x1f0000;

        if (i == 0) {
            opCode = OP_BLTZ;
        } else if (i == 0x10000) {
            opCode = OP_BGEZ;
        } else if (i == 0x100000) {
            opCode = OP_BLTZAL;
        } else if (i == 0x110000) {
            opCode = OP_BGEZAL;
        } else {
            opCode = OP_UNIMP;
        }
    }
}

//----------------------------------------------------------------------
// Mult
//      Simulate R2000 multiplication.
//      The words at *hiPtr and *loPtr are overwritten with the
//      double-length result of the multiplication.
//----------------------------------------------------------------------

static void
Mult(int a, int b, bool signedArith, int* hiPtr, int* loPtr)
{
    if ((a == 0) || (b == 0)) {
        *hiPtr = *loPtr = 0;
        return;
    }

    // Compute the sign of the result, then make everything positive
    // so unsigned computation can be done in the main loop.
    bool negative = FALSE;
    if (signedArith) {
        if (a < 0) {
            negative = !negative;
            a = -a;
        }
        if (b < 0) {
            negative = !negative;
            b = -b;
        }
    }

    // Compute the result in unsigned arithmetic (check a's bits one at
    // a time, and add in a shifted value of b).
    unsigned int bLo = b;
    unsigned int bHi = 0;
    unsigned int lo = 0;
    unsigned int hi = 0;
    for (int i = 0; i < 32; i++) {
        if (a & 1) {
            lo += bLo;
            if (lo < bLo)  // Carry out of the low bits?
                hi += 1;
            hi += bHi;
            if ((a & 0xfffffffe) == 0)
                break;
        }
        bHi <<= 1;
        if (bLo & 0x80000000)
            bHi |= 1;
        
        bLo <<= 1;
        a >>= 1;
    }

    // If the result is supposed to be negative, compute the two's
    // complement of the double-word result.
    if (negative) {
        hi = ~hi;
        lo = ~lo;
        lo++;
        if (lo == 0)
            hi++;
    }
    
    *hiPtr = (int) hi;
    *loPtr = (int) lo;
}

---