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

// Copyright N) 2007 MIPS Technologies, Inc.  All Rights Reserved

//  This software is part of the M5 simulator.

//  THIS IS A LEGAL AGREEMENT.  BY DOWNLOADING, USING, COPYING, CREATING
//  DERIVATIVE WORKS, AND/OR DISTRIBUTING THIS SOFTWARE YOU ARE AGREEING
//  TO THESE TERMS AND CONDITIONS.

//  Permission is granted to use, copy, create derivative works and
//  distribute this software and such derivative works for any purpose,
//  so long as (1) the copyright notice above, this grant of permission,
//  and the disclaimer below appear in all copies and derivative works
//  made, (2) the copyright notice above is augmented as appropriate to
//  reflect the addition of any new copyrightable work in a derivative
//  work (e.g., Copyright N) <Publication Year> Copyright Owner), and (3)
//  the name of MIPS Technologies, Inc. ($(B!H(BMIPS$(B!I(B) is not used in any
//  advertising or publicity pertaining to the use or distribution of
//  this software without specific, written prior authorization.

//  THIS SOFTWARE IS PROVIDED $(B!H(BAS IS.$(B!I(B  MIPS MAKES NO WARRANTIES AND
//  DISCLAIMS ALL WARRANTIES, WHETHER EXPRESS, STATUTORY, IMPLIED OR
//  OTHERWISE, INCLUDING BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF
//  MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE, AND
//  NON-INFRINGEMENT OF THIRD PARTY RIGHTS, REGARDING THIS SOFTWARE.
//  IN NO EVENT SHALL MIPS BE LIABLE FOR ANY DAMAGES, INCLUDING DIRECT,
//  INDIRECT, INCIDENTAL, CONSEQUENTIAL, SPECIAL, OR PUNITIVE DAMAGES OF
//  ANY KIND OR NATURE, ARISING OUT OF OR IN CONNECTION WITH THIS AGREEMENT,
//  THIS SOFTWARE AND/OR THE USE OF THIS SOFTWARE, WHETHER SUCH LIABILITY
//  IS ASSERTED ON THE BASIS OF CONTRACT, TORT (INCLUDING NEGLIGENCE OR
//  STRICT LIABILITY), OR OTHERWISE, EVEN IF MIPS HAS BEEN WARNED OF THE
//  POSSIBILITY OF ANY SUCH LOSS OR DAMAGE IN ADVANCE.

//Authors: Korey L. Sewell

////////////////////////////////////////////////////////////////////
//
// Floating Point operate instructions
//

output header {{
        /**
         * Base class for FP operations.
         */
        class FPOp : public MipsStaticInst
        {
                protected:

                /// Constructor
                FPOp(const char *mnem, MachInst _machInst, OpClass __opClass) : MipsStaticInst(mnem, _machInst, __opClass)
                {
                }

            //std::string generateDisassembly(Addr pc, const SymbolTable *symtab) const;

                //needs function to check for fpEnable or not
        };

        class FPCompareOp : public FPOp
        {
          protected:
            FPCompareOp(const char *mnem, MachInst _machInst, OpClass __opClass) : FPOp(mnem, _machInst, __opClass)
                {
                }

            std::string generateDisassembly(Addr pc, const SymbolTable *symtab) const;

        };
}};

output decoder {{
        std::string FPCompareOp::generateDisassembly(Addr pc, const SymbolTable *symtab) const
        {
            std::stringstream ss;

            ccprintf(ss, "%-10s ", mnemonic);

            ccprintf(ss,"%d",CC);

            if(_numSrcRegs > 0) {
                ss << ", ";
                printReg(ss, _srcRegIdx[0]);
            }

            if(_numSrcRegs > 1) {
                ss << ", ";
                printReg(ss, _srcRegIdx[1]);
            }

            return ss.str();
        }
}};

output exec {{
        inline Fault checkFpEnableFault(%(CPU_exec_context)s *xc)
        {
            //@TODO: Implement correct CP0 checks to see if the CP1
            // unit is enable or not
          if (!isCoprocessorEnabled(xc, 1))
             return new CoprocessorUnusableFault(1);

          return NoFault;
        }

        //If any operand is Nan return the appropriate QNaN
        template <class T>
        bool
        fpNanOperands(FPOp *inst, %(CPU_exec_context)s *xc, const T &src_type,
                      Trace::InstRecord *traceData)
        {
            uint64_t mips_nan = 0;
            T src_op = 0;
            int size = sizeof(src_op) * 8;

            for (int i = 0; i < inst->numSrcRegs(); i++) {
                uint64_t src_bits = xc->readFloatRegOperandBits(inst, 0, size);

                if (isNan(&src_bits, size) ) {
                    if (isSnan(&src_bits, size)) {
                        switch (size)
                        {
                          case 32: mips_nan = MIPS32_QNAN; break;
                          case 64: mips_nan = MIPS64_QNAN; break;
                          default: panic("Unsupported Floating Point Size (%d)", size);
                        }
                    } else {
                        mips_nan = src_bits;
                    }

                    xc->setFloatRegOperandBits(inst, 0, mips_nan, size);
                    if (traceData) { traceData->setData(mips_nan); }
                    return true;
                }
            }
            return false;
        }

        template <class T>
        bool
        fpInvalidOp(FPOp *inst, %(CPU_exec_context)s *cpu, const T dest_val,
                    Trace::InstRecord *traceData)
        {
            uint64_t mips_nan = 0;
            T src_op = dest_val;
            int size = sizeof(src_op) * 8;

            if (isNan(&src_op, size)) {
                switch (size)
                {
                  case 32: mips_nan = MIPS32_QNAN; break;
                  case 64: mips_nan = MIPS64_QNAN; break;
                  default: panic("Unsupported Floating Point Size (%d)", size);
                }

                //Set value to QNAN
                cpu->setFloatRegOperandBits(inst, 0, mips_nan, size);

                //Read FCSR from FloatRegFile
                uint32_t fcsr_bits = cpu->tcBase()->readFloatRegBits(FCSR);

                uint32_t new_fcsr = genInvalidVector(fcsr_bits);

                //Write FCSR from FloatRegFile
                cpu->tcBase()->setFloatRegBits(FCSR, new_fcsr);

                if (traceData) { traceData->setData(mips_nan); }
                return true;
            }

            return false;
        }

        void
        fpResetCauseBits(%(CPU_exec_context)s *cpu)
        {
            //Read FCSR from FloatRegFile
            uint32_t fcsr = cpu->tcBase()->readFloatRegBits(FCSR);

            // TODO: Use utility function here
            fcsr = bits(fcsr, 31, 18) << 18 | bits(fcsr, 11, 0);

            //Write FCSR from FloatRegFile
            cpu->tcBase()->setFloatRegBits(FCSR, fcsr);
        }
}};

def template FloatingPointExecute {{
        Fault %(class_name)s::execute(%(CPU_exec_context)s *xc, Trace::InstRecord *traceData) const
        {
                Fault fault = NoFault;

                %(fp_enable_check)s;


                //When is the right time to reset cause bits?
                //start of every instruction or every cycle?
#if FULL_SYSTEM
                fpResetCauseBits(xc);
#endif
                %(op_decl)s;
                %(op_rd)s;

                //Check if any FP operand is a NaN value
                if (!fpNanOperands((FPOp*)this, xc, Fd, traceData)) {
                    %(code)s;

                    //Change this code for Full-System/Sycall Emulation
                    //separation
                    //----
                    //Should Full System-Mode throw a fault here?
                    //----
                    //Check for IEEE 754 FP Exceptions
                    //fault = fpNanOperands((FPOp*)this, xc, Fd, traceData);
                    if (
#if FULL_SYSTEM
                        !fpInvalidOp((FPOp*)this, xc, Fd, traceData) &&
#endif
                        fault == NoFault)
                    {
                        %(op_wb)s;
                    }
                }

                return fault;
        }
}};

// Primary format for float point operate instructions:
def format FloatOp(code, *flags) {{
        iop = InstObjParams(name, Name, 'FPOp', code, flags)
        header_output = BasicDeclare.subst(iop)
        decoder_output = BasicConstructor.subst(iop)
        decode_block = BasicDecode.subst(iop)
        exec_output = FloatingPointExecute.subst(iop)
}};

def format FloatCompareOp(cond_code, *flags) {{
    import sys

    code = 'bool cond;\n'
    if '.sf' in cond_code or 'SinglePrecision' in flags:
        if 'QnanException' in flags:
            code += 'if (isQnan(&Fs.sf, 32) || isQnan(&Ft.sf, 32)) {\n'
            code += '\tFCSR = genInvalidVector(FCSR);\n'
            code += '\treturn NoFault;'
            code += '}\n else '
        code += 'if (isNan(&Fs.sf, 32) || isNan(&Ft.sf, 32)) {\n'
    elif '.df' in cond_code or 'DoublePrecision' in flags:
        if 'QnanException' in flags:
            code += 'if (isQnan(&Fs.df, 64) || isQnan(&Ft.df, 64)) {\n'
            code += '\tFCSR = genInvalidVector(FCSR);\n'
            code += '\treturn NoFault;'
            code += '}\n else '
        code += 'if (isNan(&Fs.df, 64) || isNan(&Ft.df, 64)) {\n'
    else:
       sys.exit('Decoder Failed: Can\'t Determine Operand Type\n')

    if 'UnorderedTrue' in flags:
       code += 'cond = 1;\n'
    elif 'UnorderedFalse' in flags:
       code += 'cond = 0;\n'
    else:
       sys.exit('Decoder Failed: Float Compare Instruction Needs A Unordered Flag\n')

    code += '} else {\n'
    code +=  cond_code + '}'
    code += 'FCSR = genCCVector(FCSR, CC, cond);\n'

    iop = InstObjParams(name, Name, 'FPCompareOp', code)
    header_output = BasicDeclare.subst(iop)
    decoder_output = BasicConstructor.subst(iop)
    decode_block = BasicDecode.subst(iop)
    exec_output = BasicExecute.subst(iop)
}};

def format FloatConvertOp(code, *flags) {{
    import sys

    #Determine Source Type
    convert = 'fpConvert('
    if '.sf' in code:
        code = 'float ' + code + '\n'
        convert += 'SINGLE_TO_'
    elif '.df' in code:
        code = 'double ' + code + '\n'
        convert += 'DOUBLE_TO_'
    elif '.uw' in code:
        code = 'uint32_t ' + code + '\n'
        convert += 'WORD_TO_'
    elif '.ud' in code:
        code = 'uint64_t ' + code + '\n'
        convert += 'LONG_TO_'
    else:
        sys.exit("Error Determining Source Type for Conversion")

    #Determine Destination Type
    if 'ToSingle' in flags:
        code += 'Fd.uw = ' + convert + 'SINGLE, '
    elif 'ToDouble' in flags:
        code += 'Fd.ud = ' + convert + 'DOUBLE, '
    elif 'ToWord' in flags:
        code += 'Fd.uw = ' + convert + 'WORD, '
    elif 'ToLong' in flags:
        code += 'Fd.ud = ' + convert + 'LONG, '
    else:
        sys.exit("Error Determining Destination Type for Conversion")

    #Figure out how to round value
    if 'Ceil' in flags:
        code += 'ceil(val)); '
    elif 'Floor' in flags:
        code += 'floor(val)); '
    elif 'Round' in flags:
        code += 'roundFP(val, 0)); '
    elif 'Trunc' in flags:
        code += 'truncFP(val));'
    else:
        code += 'val); '

    iop = InstObjParams(name, Name, 'FPOp', code)
    header_output = BasicDeclare.subst(iop)
    decoder_output = BasicConstructor.subst(iop)
    decode_block = BasicDecode.subst(iop)
    exec_output = BasicExecute.subst(iop)
}};

def format FloatAccOp(code, *flags) {{
        iop = InstObjParams(name, Name, 'FPOp', code, flags)
        header_output = BasicDeclare.subst(iop)
        decoder_output = BasicConstructor.subst(iop)
        decode_block = BasicDecode.subst(iop)
        exec_output = BasicExecute.subst(iop)
}};

// Primary format for float64 operate instructions:
def format Float64Op(code, *flags) {{
        iop = InstObjParams(name, Name, 'MipsStaticInst', code, flags)
        header_output = BasicDeclare.subst(iop)
        decoder_output = BasicConstructor.subst(iop)
        decode_block = BasicDecode.subst(iop)
        exec_output = BasicExecute.subst(iop)
}};

def format FloatPSCompareOp(cond_code1, cond_code2, *flags) {{
    import sys

    code = 'bool cond1, cond2;\n'
    code += 'bool code_block1, code_block2;\n'
    code += 'code_block1 = code_block2 = true;\n'

    if 'QnanException' in flags:
        code += 'if (isQnan(&Fs1.sf, 32) || isQnan(&Ft1.sf, 32)) {\n'
        code += '\tFCSR = genInvalidVector(FCSR);\n'
        code += 'code_block1 = false;'
        code += '}\n'
        code += 'if (isQnan(&Fs2.sf, 32) || isQnan(&Ft2.sf, 32)) {\n'
        code += '\tFCSR = genInvalidVector(FCSR);\n'
        code += 'code_block2 = false;'
        code += '}\n'

    code += 'if (code_block1) {'
    code += '\tif (isNan(&Fs1.sf, 32) || isNan(&Ft1.sf, 32)) {\n'
    if 'UnorderedTrue' in flags:
       code += 'cond1 = 1;\n'
    elif 'UnorderedFalse' in flags:
       code += 'cond1 = 0;\n'
    else:
       sys.exit('Decoder Failed: Float Compare Instruction Needs A Unordered Flag\n')
    code += '} else {\n'
    code +=  cond_code1
    code += 'FCSR = genCCVector(FCSR, CC, cond1);}\n}\n'

    code += 'if (code_block2) {'
    code += '\tif (isNan(&Fs2.sf, 32) || isNan(&Ft2.sf, 32)) {\n'
    if 'UnorderedTrue' in flags:
       code += 'cond2 = 1;\n'
    elif 'UnorderedFalse' in flags:
       code += 'cond2 = 0;\n'
    else:
       sys.exit('Decoder Failed: Float Compare Instruction Needs A Unordered Flag\n')
    code += '} else {\n'
    code +=  cond_code2
    code += 'FCSR = genCCVector(FCSR, CC, cond2);}\n}'

    iop = InstObjParams(name, Name, 'FPCompareOp', code)
    header_output = BasicDeclare.subst(iop)
    decoder_output = BasicConstructor.subst(iop)
    decode_block = BasicDecode.subst(iop)
    exec_output = BasicExecute.subst(iop)
}};