regop.isa revision 10184:bbfa3152bdea
1// Copyright (c) 2007-2008 The Hewlett-Packard Development Company
2// All rights reserved.
3//
4// The license below extends only to copyright in the software and shall
5// not be construed as granting a license to any other intellectual
6// property including but not limited to intellectual property relating
7// to a hardware implementation of the functionality of the software
8// licensed hereunder.  You may use the software subject to the license
9// terms below provided that you ensure that this notice is replicated
10// unmodified and in its entirety in all distributions of the software,
11// modified or unmodified, in source code or in binary form.
12//
13// Redistribution and use in source and binary forms, with or without
14// modification, are permitted provided that the following conditions are
15// met: redistributions of source code must retain the above copyright
16// notice, this list of conditions and the following disclaimer;
17// redistributions in binary form must reproduce the above copyright
18// notice, this list of conditions and the following disclaimer in the
19// documentation and/or other materials provided with the distribution;
20// neither the name of the copyright holders nor the names of its
21// contributors may be used to endorse or promote products derived from
22// this software without specific prior written permission.
23//
24// THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
25// "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
26// LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
27// A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
28// OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
29// SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
30// LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
31// DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
32// THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
33// (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
34// OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
35//
36// Authors: Gabe Black
37
38//////////////////////////////////////////////////////////////////////////
39//
40// RegOp Microop templates
41//
42//////////////////////////////////////////////////////////////////////////
43
44def template MicroRegOpExecute {{
45        Fault %(class_name)s::execute(%(CPU_exec_context)s *xc,
46                Trace::InstRecord *traceData) const
47        {
48            Fault fault = NoFault;
49
50            DPRINTF(X86, "The data size is %d\n", dataSize);
51            %(op_decl)s;
52            %(op_rd)s;
53
54            IntReg result M5_VAR_USED;
55
56            if(%(cond_check)s)
57            {
58                %(code)s;
59                %(flag_code)s;
60            }
61            else
62            {
63                %(else_code)s;
64            }
65
66            //Write the resulting state to the execution context
67            if(fault == NoFault)
68            {
69                %(op_wb)s;
70            }
71            return fault;
72        }
73}};
74
75def template MicroRegOpImmExecute {{
76        Fault %(class_name)s::execute(%(CPU_exec_context)s *xc,
77                Trace::InstRecord *traceData) const
78        {
79            Fault fault = NoFault;
80
81            %(op_decl)s;
82            %(op_rd)s;
83
84            IntReg result M5_VAR_USED;
85
86            if(%(cond_check)s)
87            {
88                %(code)s;
89                %(flag_code)s;
90            }
91            else
92            {
93                %(else_code)s;
94            }
95
96            //Write the resulting state to the execution context
97            if(fault == NoFault)
98            {
99                %(op_wb)s;
100            }
101            return fault;
102        }
103}};
104
105def template MicroRegOpDeclare {{
106    class %(class_name)s : public %(base_class)s
107    {
108      public:
109        %(class_name)s(ExtMachInst _machInst,
110                const char * instMnem, uint64_t setFlags,
111                InstRegIndex _src1, InstRegIndex _src2, InstRegIndex _dest,
112                uint8_t _dataSize, uint16_t _ext);
113
114        %(BasicExecDeclare)s
115    };
116}};
117
118def template MicroRegOpImmDeclare {{
119
120    class %(class_name)s : public %(base_class)s
121    {
122      public:
123        %(class_name)s(ExtMachInst _machInst,
124                const char * instMnem, uint64_t setFlags,
125                InstRegIndex _src1, uint8_t _imm8, InstRegIndex _dest,
126                uint8_t _dataSize, uint16_t _ext);
127
128        %(BasicExecDeclare)s
129    };
130}};
131
132def template MicroRegOpConstructor {{
133    %(class_name)s::%(class_name)s(
134            ExtMachInst machInst, const char * instMnem, uint64_t setFlags,
135            InstRegIndex _src1, InstRegIndex _src2, InstRegIndex _dest,
136            uint8_t _dataSize, uint16_t _ext) :
137        %(base_class)s(machInst, "%(mnemonic)s", instMnem, setFlags,
138                _src1, _src2, _dest, _dataSize, _ext,
139                %(op_class)s)
140    {
141        %(constructor)s;
142        %(cond_control_flag_init)s;
143    }
144}};
145
146def template MicroRegOpImmConstructor {{
147    %(class_name)s::%(class_name)s(
148            ExtMachInst machInst, const char * instMnem, uint64_t setFlags,
149            InstRegIndex _src1, uint8_t _imm8, InstRegIndex _dest,
150            uint8_t _dataSize, uint16_t _ext) :
151        %(base_class)s(machInst, "%(mnemonic)s", instMnem, setFlags,
152                _src1, _imm8, _dest, _dataSize, _ext,
153                %(op_class)s)
154    {
155        %(constructor)s;
156        %(cond_control_flag_init)s;
157    }
158}};
159
160output header {{
161    void
162    divide(uint64_t dividend, uint64_t divisor,
163            uint64_t &quotient, uint64_t &remainder);
164
165    enum SegmentSelectorCheck {
166      SegNoCheck, SegCSCheck, SegCallGateCheck, SegIntGateCheck,
167      SegSoftIntGateCheck, SegSSCheck, SegIretCheck, SegIntCSCheck,
168      SegTRCheck, SegTSSCheck, SegInGDTCheck, SegLDTCheck
169    };
170
171    enum LongModeDescriptorType {
172        LDT64 = 2,
173        AvailableTSS64 = 9,
174        BusyTSS64 = 0xb,
175        CallGate64 = 0xc,
176        IntGate64 = 0xe,
177        TrapGate64 = 0xf
178    };
179}};
180
181output decoder {{
182    void
183    divide(uint64_t dividend, uint64_t divisor,
184            uint64_t &quotient, uint64_t &remainder)
185    {
186        //Check for divide by zero.
187        assert(divisor != 0);
188        //If the divisor is bigger than the dividend, don't do anything.
189        if (divisor <= dividend) {
190            //Shift the divisor so it's msb lines up with the dividend.
191            int dividendMsb = findMsbSet(dividend);
192            int divisorMsb = findMsbSet(divisor);
193            int shift = dividendMsb - divisorMsb;
194            divisor <<= shift;
195            //Compute what we'll add to the quotient if the divisor isn't
196            //now larger than the dividend.
197            uint64_t quotientBit = 1;
198            quotientBit <<= shift;
199            //If we need to step back a bit (no pun intended) because the
200            //divisor got too to large, do that here. This is the "or two"
201            //part of one or two bit division.
202            if (divisor > dividend) {
203                quotientBit >>= 1;
204                divisor >>= 1;
205            }
206            //Decrement the remainder and increment the quotient.
207            quotient += quotientBit;
208            remainder -= divisor;
209        }
210    }
211}};
212
213let {{
214    # Make these empty strings so that concatenating onto
215    # them will always work.
216    header_output = ""
217    decoder_output = ""
218    exec_output = ""
219
220    immTemplates = (
221            MicroRegOpImmDeclare,
222            MicroRegOpImmConstructor,
223            MicroRegOpImmExecute)
224
225    regTemplates = (
226            MicroRegOpDeclare,
227            MicroRegOpConstructor,
228            MicroRegOpExecute)
229
230    class RegOpMeta(type):
231        def buildCppClasses(self, name, Name, suffix, code, big_code, \
232                flag_code, cond_check, else_code, cond_control_flag_init,
233                op_class):
234
235            # Globals to stick the output in
236            global header_output
237            global decoder_output
238            global exec_output
239
240            # Stick all the code together so it can be searched at once
241            allCode = "|".join((code, flag_code, cond_check, else_code, 
242                                cond_control_flag_init))
243            allBigCode = "|".join((big_code, flag_code, cond_check, else_code,
244                                   cond_control_flag_init))
245
246            # If op2 is used anywhere, make register and immediate versions
247            # of this code.
248            matcher = re.compile(r"(?<!\w)(?P<prefix>s?)op2(?P<typeQual>_[^\W_]+)?")
249            match = matcher.search(allCode + allBigCode)
250            if match:
251                typeQual = ""
252                if match.group("typeQual"):
253                    typeQual = match.group("typeQual")
254                src2_name = "%spsrc2%s" % (match.group("prefix"), typeQual)
255                self.buildCppClasses(name, Name, suffix,
256                        matcher.sub(src2_name, code),
257                        matcher.sub(src2_name, big_code),
258                        matcher.sub(src2_name, flag_code),
259                        matcher.sub(src2_name, cond_check),
260                        matcher.sub(src2_name, else_code),
261                        matcher.sub(src2_name, cond_control_flag_init),
262                        op_class)
263                imm_name = "%simm8" % match.group("prefix")
264                self.buildCppClasses(name + "i", Name, suffix + "Imm",
265                        matcher.sub(imm_name, code),
266                        matcher.sub(imm_name, big_code),
267                        matcher.sub(imm_name, flag_code),
268                        matcher.sub(imm_name, cond_check),
269                        matcher.sub(imm_name, else_code),
270                        matcher.sub(imm_name, cond_control_flag_init),
271                        op_class)
272                return
273
274            # If there's something optional to do with flags, generate
275            # a version without it and fix up this version to use it.
276            if flag_code != "" or cond_check != "true":
277                self.buildCppClasses(name, Name, suffix,
278                        code, big_code, "", "true", else_code, "", op_class)
279                suffix = "Flags" + suffix
280
281            # If psrc1 or psrc2 is used, we need to actually insert code to
282            # compute it.
283            for (big, all) in ((False, allCode), (True, allBigCode)):
284                prefix = ""
285                for (rex, decl) in (
286                        ("(?<!\w)psrc1(?!\w)",
287                         "uint64_t psrc1 = pick(SrcReg1, 0, dataSize);"),
288                        ("(?<!\w)psrc2(?!\w)",
289                         "uint64_t psrc2 = pick(SrcReg2, 1, dataSize);"),
290                        ("(?<!\w)spsrc1(?!\w)",
291                         "int64_t spsrc1 = signedPick(SrcReg1, 0, dataSize);"),
292                        ("(?<!\w)spsrc2(?!\w)",
293                         "int64_t spsrc2 = signedPick(SrcReg2, 1, dataSize);"),
294                        ("(?<!\w)simm8(?!\w)",
295                         "int8_t simm8 = imm8;")):
296                    matcher = re.compile(rex)
297                    if matcher.search(all):
298                        prefix += decl + "\n"
299                if big:
300                    if big_code != "":
301                        big_code = prefix + big_code
302                else:
303                    code = prefix + code
304
305            base = "X86ISA::RegOp"
306
307            # If imm8 shows up in the code, use the immediate templates, if
308            # not, hopefully the register ones will be correct.
309            templates = regTemplates
310            matcher = re.compile("(?<!\w)s?imm8(?!\w)")
311            if matcher.search(allCode):
312                base += "Imm"
313                templates = immTemplates
314
315            # Get everything ready for the substitution
316            iops = [InstObjParams(name, Name + suffix, base,
317                    {"code" : code,
318                     "flag_code" : flag_code,
319                     "cond_check" : cond_check,
320                     "else_code" : else_code,
321                     "cond_control_flag_init" : cond_control_flag_init,
322                     "op_class" : op_class})]
323            if big_code != "":
324                iops += [InstObjParams(name, Name + suffix + "Big", base,
325                         {"code" : big_code,
326                          "flag_code" : flag_code,
327                          "cond_check" : cond_check,
328                          "else_code" : else_code,
329                          "cond_control_flag_init" : cond_control_flag_init,
330                          "op_class" : op_class})]
331
332            # Generate the actual code (finally!)
333            for iop in iops:
334                header_output += templates[0].subst(iop)
335                decoder_output += templates[1].subst(iop)
336                exec_output += templates[2].subst(iop)
337
338
339        def __new__(mcls, Name, bases, dict):
340            abstract = False
341            name = Name.lower()
342            if "abstract" in dict:
343                abstract = dict['abstract']
344                del dict['abstract']
345
346            cls = super(RegOpMeta, mcls).__new__(mcls, Name, bases, dict)
347            if not abstract:
348                cls.className = Name
349                cls.base_mnemonic = name
350                code = cls.code
351                big_code = cls.big_code
352                flag_code = cls.flag_code
353                cond_check = cls.cond_check
354                else_code = cls.else_code
355                cond_control_flag_init = cls.cond_control_flag_init
356                op_class = cls.op_class
357
358                # Set up the C++ classes
359                mcls.buildCppClasses(cls, name, Name, "", code, big_code,
360                        flag_code, cond_check, else_code,
361                        cond_control_flag_init, op_class)
362
363                # Hook into the microassembler dict
364                global microopClasses
365                microopClasses[name] = cls
366
367                allCode = "|".join((code, flag_code, cond_check, else_code,
368                                    cond_control_flag_init))
369
370                # If op2 is used anywhere, make register and immediate versions
371                # of this code.
372                matcher = re.compile(r"op2(?P<typeQual>_[^\W_]+)?")
373                if matcher.search(allCode):
374                    microopClasses[name + 'i'] = cls
375            return cls
376
377
378    class RegOp(X86Microop):
379        __metaclass__ = RegOpMeta
380        # This class itself doesn't act as a microop
381        abstract = True
382
383        # Default template parameter values
384        big_code = ""
385        flag_code = ""
386        cond_check = "true"
387        else_code = ";"
388        cond_control_flag_init = ""
389        op_class = "IntAluOp"
390
391        def __init__(self, dest, src1, op2, flags = None, dataSize = "env.dataSize"):
392            self.dest = dest
393            self.src1 = src1
394            self.op2 = op2
395            self.flags = flags
396            self.dataSize = dataSize
397            if flags is None:
398                self.ext = 0
399            else:
400                if not isinstance(flags, (list, tuple)):
401                    raise Exception, "flags must be a list or tuple of flags"
402                self.ext = " | ".join(flags)
403                self.className += "Flags"
404
405        def getAllocator(self, microFlags):
406            if self.big_code != "":
407                className = self.className
408                if self.mnemonic == self.base_mnemonic + 'i':
409                    className += "Imm"
410                allocString = '''
411                    (%(dataSize)s >= 4) ?
412                        (StaticInstPtr)(new %(class_name)sBig(machInst,
413                            macrocodeBlock, %(flags)s, %(src1)s, %(op2)s,
414                            %(dest)s, %(dataSize)s, %(ext)s)) :
415                        (StaticInstPtr)(new %(class_name)s(machInst,
416                            macrocodeBlock, %(flags)s, %(src1)s, %(op2)s,
417                            %(dest)s, %(dataSize)s, %(ext)s))
418                    '''
419                allocator = allocString % {
420                    "class_name" : className,
421                    "flags" : self.microFlagsText(microFlags),
422                    "src1" : self.src1, "op2" : self.op2,
423                    "dest" : self.dest,
424                    "dataSize" : self.dataSize,
425                    "ext" : self.ext}
426                return allocator
427            else:
428                className = self.className
429                if self.mnemonic == self.base_mnemonic + 'i':
430                    className += "Imm"
431                allocator = '''new %(class_name)s(machInst, macrocodeBlock,
432                        %(flags)s, %(src1)s, %(op2)s, %(dest)s,
433                        %(dataSize)s, %(ext)s)''' % {
434                    "class_name" : className,
435                    "flags" : self.microFlagsText(microFlags),
436                    "src1" : self.src1, "op2" : self.op2,
437                    "dest" : self.dest,
438                    "dataSize" : self.dataSize,
439                    "ext" : self.ext}
440                return allocator
441
442    class LogicRegOp(RegOp):
443        abstract = True
444        flag_code = '''
445            //Don't have genFlags handle the OF or CF bits
446            uint64_t mask = CFBit | ECFBit | OFBit;
447            uint64_t newFlags = genFlags(PredccFlagBits | PreddfBit |
448                                 PredezfBit, ext & ~mask, result, psrc1, op2);
449            PredezfBit = newFlags & EZFBit;
450            PreddfBit = newFlags & DFBit;
451            PredccFlagBits = newFlags & ccFlagMask;
452
453            //If a logic microop wants to set these, it wants to set them to 0.
454            PredcfofBits = PredcfofBits & ~((CFBit | OFBit) & ext);
455            PredecfBit = PredecfBit & ~(ECFBit & ext);
456        '''
457
458    class FlagRegOp(RegOp):
459        abstract = True
460        flag_code = '''
461            uint64_t newFlags = genFlags(PredccFlagBits | PredcfofBits |
462                                    PreddfBit | PredecfBit | PredezfBit,
463                                    ext, result, psrc1, op2);
464
465            PredcfofBits = newFlags & cfofMask;
466            PredecfBit = newFlags & ECFBit;
467            PredezfBit = newFlags & EZFBit;
468            PreddfBit = newFlags & DFBit;
469            PredccFlagBits = newFlags & ccFlagMask;
470        '''
471
472    class SubRegOp(RegOp):
473        abstract = True
474        flag_code = '''
475            uint64_t newFlags = genFlags(PredccFlagBits | PredcfofBits |
476                                         PreddfBit | PredecfBit | PredezfBit,
477                                         ext, result, psrc1, ~op2, true);
478
479            PredcfofBits = newFlags & cfofMask;
480            PredecfBit = newFlags & ECFBit;
481            PredezfBit = newFlags & EZFBit;
482            PreddfBit = newFlags & DFBit;
483            PredccFlagBits = newFlags & ccFlagMask;
484        '''
485
486    class CondRegOp(RegOp):
487        abstract = True
488        cond_check = "checkCondition(ccFlagBits | cfofBits | dfBit | ecfBit | \
489                                     ezfBit, ext)"
490        cond_control_flag_init = "flags[IsCondControl] = flags[IsControl];"
491
492    class RdRegOp(RegOp):
493        abstract = True
494        def __init__(self, dest, src1=None, dataSize="env.dataSize"):
495            if not src1:
496                src1 = dest
497            super(RdRegOp, self).__init__(dest, src1, \
498                    "InstRegIndex(NUM_INTREGS)", None, dataSize)
499
500    class WrRegOp(RegOp):
501        abstract = True
502        def __init__(self, src1, src2, flags=None, dataSize="env.dataSize"):
503            super(WrRegOp, self).__init__("InstRegIndex(NUM_INTREGS)", \
504                    src1, src2, flags, dataSize)
505
506    class Add(FlagRegOp):
507        code = 'DestReg = merge(DestReg, result = (psrc1 + op2), dataSize);'
508        big_code = 'DestReg = result = (psrc1 + op2) & mask(dataSize * 8);'
509
510    class Or(LogicRegOp):
511        code = 'DestReg = merge(DestReg, result = (psrc1 | op2), dataSize);'
512        big_code = 'DestReg = result = (psrc1 | op2) & mask(dataSize * 8);'
513
514    class Adc(FlagRegOp):
515        code = '''
516            CCFlagBits flags = cfofBits;
517            DestReg = merge(DestReg, result = (psrc1 + op2 + flags.cf), dataSize);
518            '''
519        big_code = '''
520            CCFlagBits flags = cfofBits;
521            DestReg = result = (psrc1 + op2 + flags.cf) & mask(dataSize * 8);
522            '''
523
524    class Sbb(SubRegOp):
525        code = '''
526            CCFlagBits flags = cfofBits;
527            DestReg = merge(DestReg, result = (psrc1 - op2 - flags.cf), dataSize);
528            '''
529        big_code = '''
530            CCFlagBits flags = cfofBits;
531            DestReg = result = (psrc1 - op2 - flags.cf) & mask(dataSize * 8);
532            '''
533
534    class And(LogicRegOp):
535        code = 'DestReg = merge(DestReg, result = (psrc1 & op2), dataSize)'
536        big_code = 'DestReg = result = (psrc1 & op2) & mask(dataSize * 8)'
537
538    class Sub(SubRegOp):
539        code = 'DestReg = merge(DestReg, result = (psrc1 - op2), dataSize)'
540        big_code = 'DestReg = result = (psrc1 - op2) & mask(dataSize * 8)'
541
542    class Xor(LogicRegOp):
543        code = 'DestReg = merge(DestReg, result = (psrc1 ^ op2), dataSize)'
544        big_code = 'DestReg = result = (psrc1 ^ op2) & mask(dataSize * 8)'
545
546    class Mul1s(WrRegOp):
547        op_class = 'IntMultOp'
548
549        code = '''
550            ProdLow = psrc1 * op2;
551            int halfSize = (dataSize * 8) / 2;
552            uint64_t shifter = (ULL(1) << halfSize);
553            uint64_t hiResult;
554            uint64_t psrc1_h = psrc1 / shifter;
555            uint64_t psrc1_l = psrc1 & mask(halfSize);
556            uint64_t psrc2_h = (op2 / shifter) & mask(halfSize);
557            uint64_t psrc2_l = op2 & mask(halfSize);
558            hiResult = ((psrc1_l * psrc2_h + psrc1_h * psrc2_l +
559                        ((psrc1_l * psrc2_l) / shifter)) /shifter) +
560                       psrc1_h * psrc2_h;
561            if (bits(psrc1, dataSize * 8 - 1))
562                hiResult -= op2;
563            if (bits(op2, dataSize * 8 - 1))
564                hiResult -= psrc1;
565            ProdHi = hiResult;
566            '''
567        flag_code = '''
568            if ((-ProdHi & mask(dataSize * 8)) !=
569                    bits(ProdLow, dataSize * 8 - 1)) {
570                PredcfofBits = PredcfofBits | (ext & (CFBit | OFBit));
571                PredecfBit = PredecfBit | (ext & ECFBit);
572            } else {
573                PredcfofBits = PredcfofBits & ~(ext & (CFBit | OFBit));
574                PredecfBit = PredecfBit & ~(ext & ECFBit);
575            }
576        '''
577
578    class Mul1u(WrRegOp):
579        op_class = 'IntMultOp'
580
581        code = '''
582            ProdLow = psrc1 * op2;
583            int halfSize = (dataSize * 8) / 2;
584            uint64_t shifter = (ULL(1) << halfSize);
585            uint64_t psrc1_h = psrc1 / shifter;
586            uint64_t psrc1_l = psrc1 & mask(halfSize);
587            uint64_t psrc2_h = (op2 / shifter) & mask(halfSize);
588            uint64_t psrc2_l = op2 & mask(halfSize);
589            ProdHi = ((psrc1_l * psrc2_h + psrc1_h * psrc2_l +
590                      ((psrc1_l * psrc2_l) / shifter)) / shifter) +
591                     psrc1_h * psrc2_h;
592            '''
593        flag_code = '''
594            if (ProdHi) {
595                PredcfofBits = PredcfofBits | (ext & (CFBit | OFBit));
596                PredecfBit = PredecfBit | (ext & ECFBit);
597            } else {
598                PredcfofBits = PredcfofBits & ~(ext & (CFBit | OFBit));
599                PredecfBit = PredecfBit & ~(ext & ECFBit);
600            }
601        '''
602
603    class Mulel(RdRegOp):
604        code = 'DestReg = merge(SrcReg1, ProdLow, dataSize);'
605        big_code = 'DestReg = ProdLow & mask(dataSize * 8);'
606
607    class Muleh(RdRegOp):
608        def __init__(self, dest, src1=None, flags=None, dataSize="env.dataSize"):
609            if not src1:
610                src1 = dest
611            super(RdRegOp, self).__init__(dest, src1, \
612                    "InstRegIndex(NUM_INTREGS)", flags, dataSize)
613        code = 'DestReg = merge(SrcReg1, ProdHi, dataSize);'
614        big_code = 'DestReg = ProdHi & mask(dataSize * 8);'
615
616    # One or two bit divide
617    class Div1(WrRegOp):
618        op_class = 'IntDivOp'
619
620        code = '''
621            //These are temporaries so that modifying them later won't make
622            //the ISA parser think they're also sources.
623            uint64_t quotient = 0;
624            uint64_t remainder = psrc1;
625            //Similarly, this is a temporary so changing it doesn't make it
626            //a source.
627            uint64_t divisor = op2;
628            //This is a temporary just for consistency and clarity.
629            uint64_t dividend = remainder;
630            //Do the division.
631            if (divisor == 0) {
632                fault = new DivideByZero;
633            } else {
634                divide(dividend, divisor, quotient, remainder);
635                //Record the final results.
636                Remainder = remainder;
637                Quotient = quotient;
638                Divisor = divisor;
639            }
640            '''
641
642    # Step divide
643    class Div2(RegOp):
644        op_class = 'IntDivOp'
645
646        divCode = '''
647            uint64_t dividend = Remainder;
648            uint64_t divisor = Divisor;
649            uint64_t quotient = Quotient;
650            uint64_t remainder = dividend;
651            int remaining = op2;
652            //If we overshot, do nothing. This lets us unrool division loops a
653            //little.
654            if (divisor == 0) {
655                fault = new DivideByZero;
656            } else if (remaining) {
657                if (divisor & (ULL(1) << 63)) {
658                    while (remaining && !(dividend & (ULL(1) << 63))) {
659                        dividend = (dividend << 1) |
660                            bits(SrcReg1, remaining - 1);
661                        quotient <<= 1;
662                        remaining--;
663                    }
664                    if (dividend & (ULL(1) << 63)) {
665                        bool highBit = false;
666                        if (dividend < divisor && remaining) {
667                            highBit = true;
668                            dividend = (dividend << 1) |
669                                bits(SrcReg1, remaining - 1);
670                            quotient <<= 1;
671                            remaining--;
672                        }
673                        if (highBit || divisor <= dividend) {
674                            quotient++;
675                            dividend -= divisor;
676                        }
677                    }
678                    remainder = dividend;
679                } else {
680                    //Shift in bits from the low order portion of the dividend
681                    while (dividend < divisor && remaining) {
682                        dividend = (dividend << 1) |
683                            bits(SrcReg1, remaining - 1);
684                        quotient <<= 1;
685                        remaining--;
686                    }
687                    remainder = dividend;
688                    //Do the division.
689                    divide(dividend, divisor, quotient, remainder);
690                }
691            }
692            //Keep track of how many bits there are still to pull in.
693            %s
694            //Record the final results
695            Remainder = remainder;
696            Quotient = quotient;
697        '''
698        code = divCode % "DestReg = merge(DestReg, remaining, dataSize);"
699        big_code = divCode % "DestReg = remaining & mask(dataSize * 8);"
700        flag_code = '''
701            if (remaining == 0)
702                PredezfBit = PredezfBit | (ext & EZFBit);
703            else
704                PredezfBit = PredezfBit & ~(ext & EZFBit);
705        '''
706
707    class Divq(RdRegOp):
708        code = 'DestReg = merge(SrcReg1, Quotient, dataSize);'
709        big_code = 'DestReg = Quotient & mask(dataSize * 8);'
710
711    class Divr(RdRegOp):
712        code = 'DestReg = merge(SrcReg1, Remainder, dataSize);'
713        big_code = 'DestReg = Remainder & mask(dataSize * 8);'
714
715    class Mov(CondRegOp):
716        code = 'DestReg = merge(SrcReg1, op2, dataSize)'
717        else_code = 'DestReg = DestReg;'
718
719    # Shift instructions
720
721    class Sll(RegOp):
722        code = '''
723            uint8_t shiftAmt = (op2 & ((dataSize == 8) ? mask(6) : mask(5)));
724            DestReg = merge(DestReg, psrc1 << shiftAmt, dataSize);
725            '''
726        big_code = '''
727            uint8_t shiftAmt = (op2 & ((dataSize == 8) ? mask(6) : mask(5)));
728            DestReg = (psrc1 << shiftAmt) & mask(dataSize * 8);
729            '''
730        flag_code = '''
731            // If the shift amount is zero, no flags should be modified.
732            if (shiftAmt) {
733                //Zero out any flags we might modify. This way we only have to
734                //worry about setting them.
735                PredcfofBits = PredcfofBits & ~(ext & (CFBit | OFBit));
736                PredecfBit = PredecfBit & ~(ext & ECFBit);
737
738                int CFBits = 0;
739                //Figure out if we -would- set the CF bits if requested.
740                if (shiftAmt <= dataSize * 8 &&
741                        bits(SrcReg1, dataSize * 8 - shiftAmt)) {
742                    CFBits = 1;
743                }
744
745                //If some combination of the CF bits need to be set, set them.
746                if ((ext & (CFBit | ECFBit)) && CFBits) {
747                    PredcfofBits = PredcfofBits | (ext & CFBit);
748                    PredecfBit = PredecfBit | (ext & ECFBit);
749                }
750
751                //Figure out what the OF bit should be.
752                if ((ext & OFBit) && (CFBits ^ bits(DestReg, dataSize * 8 - 1)))
753                    PredcfofBits = PredcfofBits | OFBit;
754
755                //Use the regular mechanisms to calculate the other flags.
756                uint64_t newFlags = genFlags(PredccFlagBits | PreddfBit |
757                                PredezfBit, ext & ~(CFBit | ECFBit | OFBit),
758                                DestReg, psrc1, op2);
759
760                PredezfBit = newFlags & EZFBit;
761                PreddfBit = newFlags & DFBit;
762                PredccFlagBits = newFlags & ccFlagMask;
763            }
764        '''
765
766    class Srl(RegOp):
767        # Because what happens to the bits shift -in- on a right shift
768        # is not defined in the C/C++ standard, we have to mask them out
769        # to be sure they're zero.
770        code = '''
771            uint8_t shiftAmt = (op2 & ((dataSize == 8) ? mask(6) : mask(5)));
772            uint64_t logicalMask = mask(dataSize * 8 - shiftAmt);
773            DestReg = merge(DestReg, (psrc1 >> shiftAmt) & logicalMask, dataSize);
774            '''
775        big_code = '''
776            uint8_t shiftAmt = (op2 & ((dataSize == 8) ? mask(6) : mask(5)));
777            uint64_t logicalMask = mask(dataSize * 8 - shiftAmt);
778            DestReg = (psrc1 >> shiftAmt) & logicalMask;
779            '''
780        flag_code = '''
781            // If the shift amount is zero, no flags should be modified.
782            if (shiftAmt) {
783                //Zero out any flags we might modify. This way we only have to
784                //worry about setting them.
785                PredcfofBits = PredcfofBits & ~(ext & (CFBit | OFBit));
786                PredecfBit = PredecfBit & ~(ext & ECFBit);
787
788                //If some combination of the CF bits need to be set, set them.
789                if ((ext & (CFBit | ECFBit)) && 
790                        shiftAmt <= dataSize * 8 &&
791                        bits(SrcReg1, shiftAmt - 1)) {
792                    PredcfofBits = PredcfofBits | (ext & CFBit);
793                    PredecfBit = PredecfBit | (ext & ECFBit);
794                }
795
796                //Figure out what the OF bit should be.
797                if ((ext & OFBit) && bits(SrcReg1, dataSize * 8 - 1))
798                    PredcfofBits = PredcfofBits | OFBit;
799
800                //Use the regular mechanisms to calculate the other flags.
801                uint64_t newFlags = genFlags(PredccFlagBits | PreddfBit |
802                                PredezfBit, ext & ~(CFBit | ECFBit | OFBit),
803                                DestReg, psrc1, op2);
804
805                PredezfBit = newFlags & EZFBit;
806                PreddfBit = newFlags & DFBit;
807                PredccFlagBits = newFlags & ccFlagMask;
808            }
809        '''
810
811    class Sra(RegOp):
812        # Because what happens to the bits shift -in- on a right shift
813        # is not defined in the C/C++ standard, we have to sign extend
814        # them manually to be sure.
815        code = '''
816            uint8_t shiftAmt = (op2 & ((dataSize == 8) ? mask(6) : mask(5)));
817            uint64_t arithMask = (shiftAmt == 0) ? 0 :
818                -bits(psrc1, dataSize * 8 - 1) << (dataSize * 8 - shiftAmt);
819            DestReg = merge(DestReg, (psrc1 >> shiftAmt) | arithMask, dataSize);
820            '''
821        big_code = '''
822            uint8_t shiftAmt = (op2 & ((dataSize == 8) ? mask(6) : mask(5)));
823            uint64_t arithMask = (shiftAmt == 0) ? 0 :
824                -bits(psrc1, dataSize * 8 - 1) << (dataSize * 8 - shiftAmt);
825            DestReg = ((psrc1 >> shiftAmt) | arithMask) & mask(dataSize * 8);
826            '''
827        flag_code = '''
828            // If the shift amount is zero, no flags should be modified.
829            if (shiftAmt) {
830                //Zero out any flags we might modify. This way we only have to
831                //worry about setting them.
832                PredcfofBits = PredcfofBits & ~(ext & (CFBit | OFBit));
833                PredecfBit = PredecfBit & ~(ext & ECFBit);
834
835                //If some combination of the CF bits need to be set, set them.
836                uint8_t effectiveShift =
837                    (shiftAmt <= dataSize * 8) ? shiftAmt : (dataSize * 8);
838                if ((ext & (CFBit | ECFBit)) &&
839                        bits(SrcReg1, effectiveShift - 1)) {
840                    PredcfofBits = PredcfofBits | (ext & CFBit);
841                    PredecfBit = PredecfBit | (ext & ECFBit);
842                }
843
844                //Use the regular mechanisms to calculate the other flags.
845                uint64_t newFlags = genFlags(PredccFlagBits | PreddfBit |
846                                PredezfBit, ext & ~(CFBit | ECFBit | OFBit),
847                                DestReg, psrc1, op2);
848
849                PredezfBit = newFlags & EZFBit;
850                PreddfBit = newFlags & DFBit;
851                PredccFlagBits = newFlags & ccFlagMask;
852            }
853        '''
854
855    class Ror(RegOp):
856        code = '''
857            uint8_t shiftAmt =
858                (op2 & ((dataSize == 8) ? mask(6) : mask(5)));
859            uint8_t realShiftAmt = shiftAmt % (dataSize * 8);
860            if (realShiftAmt) {
861                uint64_t top = psrc1 << (dataSize * 8 - realShiftAmt);
862                uint64_t bottom = bits(psrc1, dataSize * 8, realShiftAmt);
863                DestReg = merge(DestReg, top | bottom, dataSize);
864            } else
865                DestReg = merge(DestReg, DestReg, dataSize);
866            '''
867        flag_code = '''
868            // If the shift amount is zero, no flags should be modified.
869            if (shiftAmt) {
870                //Zero out any flags we might modify. This way we only have to
871                //worry about setting them.
872                PredcfofBits = PredcfofBits & ~(ext & (CFBit | OFBit));
873                PredecfBit = PredecfBit & ~(ext & ECFBit);
874
875                //Find the most and second most significant bits of the result.
876                int msb = bits(DestReg, dataSize * 8 - 1);
877                int smsb = bits(DestReg, dataSize * 8 - 2);
878                //If some combination of the CF bits need to be set, set them.
879                if ((ext & (CFBit | ECFBit)) && msb) {
880                    PredcfofBits = PredcfofBits | (ext & CFBit);
881                    PredecfBit = PredecfBit | (ext & ECFBit);
882                }
883
884                //Figure out what the OF bit should be.
885                if ((ext & OFBit) && (msb ^ smsb))
886                    PredcfofBits = PredcfofBits | OFBit;
887
888                //Use the regular mechanisms to calculate the other flags.
889                uint64_t newFlags = genFlags(PredccFlagBits | PreddfBit |
890                                PredezfBit, ext & ~(CFBit | ECFBit | OFBit),
891                                DestReg, psrc1, op2);
892
893                PredezfBit = newFlags & EZFBit;
894                PreddfBit = newFlags & DFBit;
895                PredccFlagBits = newFlags & ccFlagMask;
896            }
897        '''
898
899    class Rcr(RegOp):
900        code = '''
901            uint8_t shiftAmt =
902                (op2 & ((dataSize == 8) ? mask(6) : mask(5)));
903            uint8_t realShiftAmt = shiftAmt % (dataSize * 8 + 1);
904            if (realShiftAmt) {
905                CCFlagBits flags = cfofBits;
906                uint64_t top = flags.cf << (dataSize * 8 - realShiftAmt);
907                if (realShiftAmt > 1)
908                    top |= psrc1 << (dataSize * 8 - realShiftAmt + 1);
909                uint64_t bottom = bits(psrc1, dataSize * 8 - 1, realShiftAmt);
910                DestReg = merge(DestReg, top | bottom, dataSize);
911            } else
912                DestReg = merge(DestReg, DestReg, dataSize);
913            '''
914        flag_code = '''
915            // If the shift amount is zero, no flags should be modified.
916            if (shiftAmt) {
917                int origCFBit = (cfofBits & CFBit) ? 1 : 0;
918                //Zero out any flags we might modify. This way we only have to
919                //worry about setting them.
920                PredcfofBits = PredcfofBits & ~(ext & (CFBit | OFBit));
921                PredecfBit = PredecfBit & ~(ext & ECFBit);
922
923                //Figure out what the OF bit should be.
924                if ((ext & OFBit) && (origCFBit ^
925                                      bits(SrcReg1, dataSize * 8 - 1))) {
926                    PredcfofBits = PredcfofBits | OFBit;
927                }
928                //If some combination of the CF bits need to be set, set them.
929                if ((ext & (CFBit | ECFBit)) &&
930                        (realShiftAmt == 0) ? origCFBit :
931                        bits(SrcReg1, realShiftAmt - 1)) {
932                    PredcfofBits = PredcfofBits | (ext & CFBit);
933                    PredecfBit = PredecfBit | (ext & ECFBit);
934                }
935
936                //Use the regular mechanisms to calculate the other flags.
937                uint64_t newFlags = genFlags(PredccFlagBits | PreddfBit |
938                                PredezfBit, ext & ~(CFBit | ECFBit | OFBit),
939                                DestReg, psrc1, op2);
940
941                PredezfBit = newFlags & EZFBit;
942                PreddfBit = newFlags & DFBit;
943                PredccFlagBits = newFlags & ccFlagMask;
944            }
945        '''
946
947    class Rol(RegOp):
948        code = '''
949            uint8_t shiftAmt =
950                (op2 & ((dataSize == 8) ? mask(6) : mask(5)));
951            uint8_t realShiftAmt = shiftAmt % (dataSize * 8);
952            if (realShiftAmt) {
953                uint64_t top = psrc1 << realShiftAmt;
954                uint64_t bottom =
955                    bits(psrc1, dataSize * 8 - 1, dataSize * 8 - realShiftAmt);
956                DestReg = merge(DestReg, top | bottom, dataSize);
957            } else
958                DestReg = merge(DestReg, DestReg, dataSize);
959            '''
960        flag_code = '''
961            // If the shift amount is zero, no flags should be modified.
962            if (shiftAmt) {
963                //Zero out any flags we might modify. This way we only have to
964                //worry about setting them.
965                PredcfofBits = PredcfofBits & ~(ext & (CFBit | OFBit));
966                PredecfBit = PredecfBit & ~(ext & ECFBit);
967
968                //The CF bits, if set, would be set to the lsb of the result.
969                int lsb = DestReg & 0x1;
970                int msb = bits(DestReg, dataSize * 8 - 1);
971                //If some combination of the CF bits need to be set, set them.
972                if ((ext & (CFBit | ECFBit)) && lsb) {
973                    PredcfofBits = PredcfofBits | (ext & CFBit);
974                    PredecfBit = PredecfBit | (ext & ECFBit);
975                }
976
977                //Figure out what the OF bit should be.
978                if ((ext & OFBit) && (msb ^ lsb))
979                    PredcfofBits = PredcfofBits | OFBit;
980
981                //Use the regular mechanisms to calculate the other flags.
982                uint64_t newFlags = genFlags(PredccFlagBits | PreddfBit |
983                                PredezfBit, ext & ~(CFBit | ECFBit | OFBit),
984                                DestReg, psrc1, op2);
985
986                PredezfBit = newFlags & EZFBit;
987                PreddfBit = newFlags & DFBit;
988                PredccFlagBits = newFlags & ccFlagMask;
989            }
990        '''
991
992    class Rcl(RegOp):
993        code = '''
994            uint8_t shiftAmt =
995                (op2 & ((dataSize == 8) ? mask(6) : mask(5)));
996            uint8_t realShiftAmt = shiftAmt % (dataSize * 8 + 1);
997            if (realShiftAmt) {
998                CCFlagBits flags = cfofBits;
999                uint64_t top = psrc1 << realShiftAmt;
1000                uint64_t bottom = flags.cf << (realShiftAmt - 1);
1001                if(shiftAmt > 1)
1002                    bottom |=
1003                        bits(psrc1, dataSize * 8 - 1,
1004                                   dataSize * 8 - realShiftAmt + 1);
1005                DestReg = merge(DestReg, top | bottom, dataSize);
1006            } else
1007                DestReg = merge(DestReg, DestReg, dataSize);
1008            '''
1009        flag_code = '''
1010            // If the shift amount is zero, no flags should be modified.
1011            if (shiftAmt) {
1012                int origCFBit = (cfofBits & CFBit) ? 1 : 0;
1013                //Zero out any flags we might modify. This way we only have to
1014                //worry about setting them.
1015                PredcfofBits = PredcfofBits & ~(ext & (CFBit | OFBit));
1016                PredecfBit = PredecfBit & ~(ext & ECFBit);
1017
1018                int msb = bits(DestReg, dataSize * 8 - 1);
1019                int CFBits = bits(SrcReg1, dataSize * 8 - realShiftAmt);
1020                //If some combination of the CF bits need to be set, set them.
1021                if ((ext & (CFBit | ECFBit)) && 
1022                        (realShiftAmt == 0) ? origCFBit : CFBits) {
1023                    PredcfofBits = PredcfofBits | (ext & CFBit);
1024                    PredecfBit = PredecfBit | (ext & ECFBit);
1025                }
1026
1027                //Figure out what the OF bit should be.
1028                if ((ext & OFBit) && (msb ^ CFBits))
1029                    PredcfofBits = PredcfofBits | OFBit;
1030
1031                //Use the regular mechanisms to calculate the other flags.
1032                uint64_t newFlags = genFlags(PredccFlagBits | PreddfBit |
1033                                PredezfBit, ext & ~(CFBit | ECFBit | OFBit),
1034                                DestReg, psrc1, op2);
1035
1036                PredezfBit = newFlags & EZFBit;
1037                PreddfBit = newFlags & DFBit;
1038                PredccFlagBits = newFlags & ccFlagMask;
1039            }
1040        '''
1041
1042    class Sld(RegOp):
1043        sldCode = '''
1044            uint8_t shiftAmt = (op2 & ((dataSize == 8) ? mask(6) : mask(5)));
1045            uint8_t dataBits = dataSize * 8;
1046            uint8_t realShiftAmt = shiftAmt %% (2 * dataBits);
1047            uint64_t result;
1048            if (realShiftAmt == 0) {
1049                result = psrc1;
1050            } else if (realShiftAmt < dataBits) {
1051                result = (psrc1 << realShiftAmt) |
1052                         (DoubleBits >> (dataBits - realShiftAmt));
1053            } else {
1054                result = (DoubleBits << (realShiftAmt - dataBits)) |
1055                         (psrc1 >> (2 * dataBits - realShiftAmt));
1056            }
1057            %s
1058            '''
1059        code = sldCode % "DestReg = merge(DestReg, result, dataSize);"
1060        big_code = sldCode % "DestReg = result & mask(dataSize * 8);"
1061        flag_code = '''
1062            // If the shift amount is zero, no flags should be modified.
1063            if (shiftAmt) {
1064                //Zero out any flags we might modify. This way we only have to
1065                //worry about setting them.
1066                PredcfofBits = PredcfofBits & ~(ext & (CFBit | OFBit));
1067                PredecfBit = PredecfBit & ~(ext & ECFBit);
1068                int CFBits = 0;
1069
1070                //Figure out if we -would- set the CF bits if requested.
1071                if ((realShiftAmt == 0 &&
1072                        bits(DoubleBits, 0)) ||
1073                    (realShiftAmt <= dataBits &&
1074                     bits(SrcReg1, dataBits - realShiftAmt)) ||
1075                    (realShiftAmt > dataBits &&
1076                     bits(DoubleBits, 2 * dataBits - realShiftAmt))) {
1077                    CFBits = 1;
1078                }
1079
1080                //If some combination of the CF bits need to be set, set them.
1081                if ((ext & (CFBit | ECFBit)) && CFBits) {
1082                    PredcfofBits = PredcfofBits | (ext & CFBit);
1083                    PredecfBit = PredecfBit | (ext & ECFBit);
1084                }
1085
1086                //Figure out what the OF bit should be.
1087                if ((ext & OFBit) && (bits(SrcReg1, dataBits - 1) ^
1088                                      bits(result, dataBits - 1)))
1089                    PredcfofBits = PredcfofBits | OFBit;
1090
1091                //Use the regular mechanisms to calculate the other flags.
1092                uint64_t newFlags = genFlags(PredccFlagBits | PreddfBit |
1093                                PredezfBit, ext & ~(CFBit | ECFBit | OFBit),
1094                                DestReg, psrc1, op2);
1095
1096                PredezfBit = newFlags & EZFBit;
1097                PreddfBit = newFlags & DFBit;
1098                PredccFlagBits = newFlags & ccFlagMask;
1099            }
1100        '''
1101
1102    class Srd(RegOp):
1103        srdCode = '''
1104            uint8_t shiftAmt = (op2 & ((dataSize == 8) ? mask(6) : mask(5)));
1105            uint8_t dataBits = dataSize * 8;
1106            uint8_t realShiftAmt = shiftAmt %% (2 * dataBits);
1107            uint64_t result;
1108            if (realShiftAmt == 0) {
1109                result = psrc1;
1110            } else if (realShiftAmt < dataBits) {
1111                // Because what happens to the bits shift -in- on a right
1112                // shift is not defined in the C/C++ standard, we have to
1113                // mask them out to be sure they're zero.
1114                uint64_t logicalMask = mask(dataBits - realShiftAmt);
1115                result = ((psrc1 >> realShiftAmt) & logicalMask) |
1116                         (DoubleBits << (dataBits - realShiftAmt));
1117            } else {
1118                uint64_t logicalMask = mask(2 * dataBits - realShiftAmt);
1119                result = ((DoubleBits >> (realShiftAmt - dataBits)) &
1120                          logicalMask) |
1121                         (psrc1 << (2 * dataBits - realShiftAmt));
1122            }
1123            %s
1124            '''
1125        code = srdCode % "DestReg = merge(DestReg, result, dataSize);"
1126        big_code = srdCode % "DestReg = result & mask(dataSize * 8);"
1127        flag_code = '''
1128            // If the shift amount is zero, no flags should be modified.
1129            if (shiftAmt) {
1130                //Zero out any flags we might modify. This way we only have to
1131                //worry about setting them.
1132                PredcfofBits = PredcfofBits & ~(ext & (CFBit | OFBit));
1133                PredecfBit = PredecfBit & ~(ext & ECFBit);
1134                int CFBits = 0;
1135
1136                //If some combination of the CF bits need to be set, set them.
1137                if ((realShiftAmt == 0 &&
1138                            bits(DoubleBits, dataBits - 1)) ||
1139                        (realShiftAmt <= dataBits &&
1140                         bits(SrcReg1, realShiftAmt - 1)) ||
1141                        (realShiftAmt > dataBits &&
1142                         bits(DoubleBits, realShiftAmt - dataBits - 1))) {
1143                    CFBits = 1;
1144                }
1145
1146                //If some combination of the CF bits need to be set, set them.
1147                if ((ext & (CFBit | ECFBit)) && CFBits) {
1148                    PredcfofBits = PredcfofBits | (ext & CFBit);
1149                    PredecfBit = PredecfBit | (ext & ECFBit);
1150                }
1151
1152                //Figure out what the OF bit should be.
1153                if ((ext & OFBit) && (bits(SrcReg1, dataBits - 1) ^
1154                                      bits(result, dataBits - 1)))
1155                    PredcfofBits = PredcfofBits | OFBit;
1156
1157                //Use the regular mechanisms to calculate the other flags.
1158                uint64_t newFlags = genFlags(PredccFlagBits | PreddfBit |
1159                                PredezfBit, ext & ~(CFBit | ECFBit | OFBit),
1160                                DestReg, psrc1, op2);
1161
1162                PredezfBit = newFlags & EZFBit;
1163                PreddfBit = newFlags & DFBit;
1164                PredccFlagBits = newFlags & ccFlagMask;
1165            }
1166        '''
1167
1168    class Mdb(WrRegOp):
1169        code = 'DoubleBits = psrc1 ^ op2;'
1170
1171    class Wrip(WrRegOp, CondRegOp):
1172        code = 'NRIP = psrc1 + sop2 + CSBase;'
1173        else_code = "NRIP = NRIP;"
1174
1175    class Wruflags(WrRegOp):
1176        code = '''
1177            uint64_t newFlags = psrc1 ^ op2;
1178            cfofBits = newFlags & cfofMask;
1179            ecfBit = newFlags & ECFBit;
1180            ezfBit = newFlags & EZFBit;
1181            dfBit = newFlags & DFBit;
1182            ccFlagBits = newFlags & ccFlagMask;
1183        '''
1184
1185    class Wrflags(WrRegOp):
1186        code = '''
1187            MiscReg newFlags = psrc1 ^ op2;
1188            MiscReg userFlagMask = 0xDD5;
1189
1190            // Get only the user flags
1191            ccFlagBits = newFlags & ccFlagMask;
1192            dfBit = newFlags & DFBit;
1193            cfofBits = newFlags & cfofMask;
1194            ecfBit = 0;
1195            ezfBit = 0;
1196
1197            // Get everything else
1198            nccFlagBits = newFlags & ~userFlagMask;
1199        '''
1200
1201    class Rdip(RdRegOp):
1202        code = 'DestReg = NRIP - CSBase;'
1203
1204    class Ruflags(RdRegOp):
1205        code = 'DestReg = ccFlagBits | cfofBits | dfBit | ecfBit | ezfBit;'
1206
1207    class Rflags(RdRegOp):
1208        code = '''
1209            DestReg = ccFlagBits | cfofBits | dfBit |
1210                      ecfBit | ezfBit | nccFlagBits;
1211            '''
1212
1213    class Ruflag(RegOp):
1214        code = '''
1215            int flag = bits(ccFlagBits | cfofBits | dfBit |
1216                            ecfBit | ezfBit, imm8);
1217            DestReg = merge(DestReg, flag, dataSize);
1218            ezfBit = (flag == 0) ? EZFBit : 0;
1219            '''
1220
1221        big_code = '''
1222            int flag = bits(ccFlagBits | cfofBits | dfBit |
1223                            ecfBit | ezfBit, imm8);
1224            DestReg = flag & mask(dataSize * 8);
1225            ezfBit = (flag == 0) ? EZFBit : 0;
1226            '''
1227
1228        def __init__(self, dest, imm, flags=None, \
1229                dataSize="env.dataSize"):
1230            super(Ruflag, self).__init__(dest, \
1231                    "InstRegIndex(NUM_INTREGS)", imm, flags, dataSize)
1232
1233    class Rflag(RegOp):
1234        code = '''
1235            MiscReg flagMask = 0x3F7FDD5;
1236            MiscReg flags = (nccFlagBits | ccFlagBits | cfofBits | dfBit |
1237                             ecfBit | ezfBit) & flagMask;
1238
1239            int flag = bits(flags, imm8);
1240            DestReg = merge(DestReg, flag, dataSize);
1241            ezfBit = (flag == 0) ? EZFBit : 0;
1242            '''
1243
1244        big_code = '''
1245            MiscReg flagMask = 0x3F7FDD5;
1246            MiscReg flags = (nccFlagBits | ccFlagBits | cfofBits | dfBit |
1247                             ecfBit | ezfBit) & flagMask;
1248
1249            int flag = bits(flags, imm8);
1250            DestReg = flag & mask(dataSize * 8);
1251            ezfBit = (flag == 0) ? EZFBit : 0;
1252            '''
1253
1254        def __init__(self, dest, imm, flags=None, \
1255                dataSize="env.dataSize"):
1256            super(Rflag, self).__init__(dest, \
1257                    "InstRegIndex(NUM_INTREGS)", imm, flags, dataSize)
1258
1259    class Sext(RegOp):
1260        code = '''
1261            IntReg val = psrc1;
1262            // Mask the bit position so that it wraps.
1263            int bitPos = op2 & (dataSize * 8 - 1);
1264            int sign_bit = bits(val, bitPos, bitPos);
1265            uint64_t maskVal = mask(bitPos+1);
1266            val = sign_bit ? (val | ~maskVal) : (val & maskVal);
1267            DestReg = merge(DestReg, val, dataSize);
1268            '''
1269
1270        big_code = '''
1271            IntReg val = psrc1;
1272            // Mask the bit position so that it wraps.
1273            int bitPos = op2 & (dataSize * 8 - 1);
1274            int sign_bit = bits(val, bitPos, bitPos);
1275            uint64_t maskVal = mask(bitPos+1);
1276            val = sign_bit ? (val | ~maskVal) : (val & maskVal);
1277            DestReg = val & mask(dataSize * 8);
1278            '''
1279
1280        flag_code = '''
1281            if (!sign_bit) {
1282                PredccFlagBits = PredccFlagBits & ~(ext & (ZFBit));
1283                PredcfofBits = PredcfofBits & ~(ext & (CFBit));
1284                PredecfBit = PredecfBit & ~(ext & ECFBit);
1285                PredezfBit = PredezfBit & ~(ext & EZFBit);
1286            } else {
1287                PredccFlagBits = PredccFlagBits | (ext & (ZFBit));
1288                PredcfofBits = PredcfofBits | (ext & (CFBit));
1289                PredecfBit = PredecfBit | (ext & ECFBit);
1290                PredezfBit = PredezfBit | (ext & EZFBit);
1291            }
1292            '''
1293
1294    class Zext(RegOp):
1295        code = 'DestReg = merge(DestReg, bits(psrc1, op2, 0), dataSize);'
1296        big_code = 'DestReg = bits(psrc1, op2, 0) & mask(dataSize * 8);'
1297
1298    class Rddr(RegOp):
1299        def __init__(self, dest, src1, flags=None, dataSize="env.dataSize"):
1300            super(Rddr, self).__init__(dest, \
1301                    src1, "InstRegIndex(NUM_INTREGS)", flags, dataSize)
1302        rdrCode = '''
1303            CR4 cr4 = CR4Op;
1304            DR7 dr7 = DR7Op;
1305            if ((cr4.de == 1 && (src1 == 4 || src1 == 5)) || src1 >= 8) {
1306                fault = new InvalidOpcode();
1307            } else if (dr7.gd) {
1308                fault = new DebugException();
1309            } else {
1310                %s
1311            }
1312        '''
1313        code = rdrCode % "DestReg = merge(DestReg, DebugSrc1, dataSize);"
1314        big_code = rdrCode % "DestReg = DebugSrc1 & mask(dataSize * 8);"
1315
1316    class Wrdr(RegOp):
1317        def __init__(self, dest, src1, flags=None, dataSize="env.dataSize"):
1318            super(Wrdr, self).__init__(dest, \
1319                    src1, "InstRegIndex(NUM_INTREGS)", flags, dataSize)
1320        code = '''
1321            CR4 cr4 = CR4Op;
1322            DR7 dr7 = DR7Op;
1323            if ((cr4.de == 1 && (dest == 4 || dest == 5)) || dest >= 8) {
1324                fault = new InvalidOpcode();
1325            } else if ((dest == 6 || dest == 7) && bits(psrc1, 63, 32) &&
1326                    machInst.mode.mode == LongMode) {
1327                fault = new GeneralProtection(0);
1328            } else if (dr7.gd) {
1329                fault = new DebugException();
1330            } else {
1331                DebugDest = psrc1;
1332            }
1333        '''
1334
1335    class Rdcr(RegOp):
1336        def __init__(self, dest, src1, flags=None, dataSize="env.dataSize"):
1337            super(Rdcr, self).__init__(dest, \
1338                    src1, "InstRegIndex(NUM_INTREGS)", flags, dataSize)
1339        rdcrCode = '''
1340            if (src1 == 1 || (src1 > 4 && src1 < 8) || (src1 > 8)) {
1341                fault = new InvalidOpcode();
1342            } else {
1343                %s
1344            }
1345        '''
1346        code = rdcrCode % "DestReg = merge(DestReg, ControlSrc1, dataSize);"
1347        big_code = rdcrCode % "DestReg = ControlSrc1 & mask(dataSize * 8);"
1348
1349    class Wrcr(RegOp):
1350        def __init__(self, dest, src1, flags=None, dataSize="env.dataSize"):
1351            super(Wrcr, self).__init__(dest, \
1352                    src1, "InstRegIndex(NUM_INTREGS)", flags, dataSize)
1353        code = '''
1354            if (dest == 1 || (dest > 4 && dest < 8) || (dest > 8)) {
1355                fault = new InvalidOpcode();
1356            } else {
1357                // There are *s in the line below so it doesn't confuse the
1358                // parser. They may be unnecessary.
1359                //Mis*cReg old*Val = pick(Cont*rolDest, 0, dat*aSize);
1360                MiscReg newVal = psrc1;
1361
1362                // Check for any modifications that would cause a fault.
1363                switch(dest) {
1364                  case 0:
1365                    {
1366                        Efer efer = EferOp;
1367                        CR0 cr0 = newVal;
1368                        CR4 oldCr4 = CR4Op;
1369                        if (bits(newVal, 63, 32) ||
1370                                (!cr0.pe && cr0.pg) ||
1371                                (!cr0.cd && cr0.nw) ||
1372                                (cr0.pg && efer.lme && !oldCr4.pae))
1373                            fault = new GeneralProtection(0);
1374                    }
1375                    break;
1376                  case 2:
1377                    break;
1378                  case 3:
1379                    break;
1380                  case 4:
1381                    {
1382                        CR4 cr4 = newVal;
1383                        // PAE can't be disabled in long mode.
1384                        if (bits(newVal, 63, 11) ||
1385                                (machInst.mode.mode == LongMode && !cr4.pae))
1386                            fault = new GeneralProtection(0);
1387                    }
1388                    break;
1389                  case 8:
1390                    {
1391                        if (bits(newVal, 63, 4))
1392                            fault = new GeneralProtection(0);
1393                    }
1394                  default:
1395                    fault = new GenericISA::M5PanicFault(
1396                            "Unrecognized control register %d.\\n", dest);
1397                }
1398                ControlDest = newVal;
1399            }
1400            '''
1401
1402    # Microops for manipulating segmentation registers
1403    class SegOp(CondRegOp):
1404        abstract = True
1405        def __init__(self, dest, src1, flags=None, dataSize="env.dataSize"):
1406            super(SegOp, self).__init__(dest, \
1407                    src1, "InstRegIndex(NUM_INTREGS)", flags, dataSize)
1408
1409    class Wrbase(SegOp):
1410        code = '''
1411            SegBaseDest = psrc1;
1412        '''
1413
1414    class Wrlimit(SegOp):
1415        code = '''
1416            SegLimitDest = psrc1;
1417        '''
1418
1419    class Wrsel(SegOp):
1420        code = '''
1421            SegSelDest = psrc1;
1422        '''
1423
1424    class WrAttr(SegOp):
1425        code = '''
1426            SegAttrDest = psrc1;
1427        '''
1428
1429    class Rdbase(SegOp):
1430        code = 'DestReg = merge(DestReg, SegBaseSrc1, dataSize);'
1431        big_code = 'DestReg = SegBaseSrc1 & mask(dataSize * 8);'
1432
1433    class Rdlimit(SegOp):
1434        code = 'DestReg = merge(DestReg, SegLimitSrc1, dataSize);'
1435        big_code = 'DestReg = SegLimitSrc1 & mask(dataSize * 8);'
1436
1437    class RdAttr(SegOp):
1438        code = 'DestReg = merge(DestReg, SegAttrSrc1, dataSize);'
1439        big_code = 'DestReg = SegAttrSrc1 & mask(dataSize * 8);'
1440
1441    class Rdsel(SegOp):
1442        code = 'DestReg = merge(DestReg, SegSelSrc1, dataSize);'
1443        big_code = 'DestReg = SegSelSrc1 & mask(dataSize * 8);'
1444
1445    class Rdval(RegOp):
1446        def __init__(self, dest, src1, flags=None, dataSize="env.dataSize"):
1447            super(Rdval, self).__init__(dest, src1, \
1448                    "InstRegIndex(NUM_INTREGS)", flags, dataSize)
1449        code = '''
1450            DestReg = MiscRegSrc1;
1451        '''
1452
1453    class Wrval(RegOp):
1454        def __init__(self, dest, src1, flags=None, dataSize="env.dataSize"):
1455            super(Wrval, self).__init__(dest, src1, \
1456                    "InstRegIndex(NUM_INTREGS)", flags, dataSize)
1457        code = '''
1458            MiscRegDest = SrcReg1;
1459        '''
1460
1461    class Chks(RegOp):
1462        def __init__(self, dest, src1, src2=0,
1463                flags=None, dataSize="env.dataSize"):
1464            super(Chks, self).__init__(dest,
1465                    src1, src2, flags, dataSize)
1466        code = '''
1467            // The selector is in source 1 and can be at most 16 bits.
1468            SegSelector selector = DestReg;
1469            SegDescriptor desc = SrcReg1;
1470            HandyM5Reg m5reg = M5Reg;
1471
1472            switch (imm8)
1473            {
1474              case SegNoCheck:
1475                break;
1476              case SegCSCheck:
1477                // Make sure it's the right type
1478                if (desc.s == 0 || desc.type.codeOrData != 1) {
1479                    fault = new GeneralProtection(0);
1480                } else if (m5reg.cpl != desc.dpl) {
1481                    fault = new GeneralProtection(0);
1482                }
1483                break;
1484              case SegCallGateCheck:
1485                fault = new GenericISA::M5PanicFault("CS checks for far "
1486                        "calls/jumps through call gates not implemented.\\n");
1487                break;
1488              case SegSoftIntGateCheck:
1489                // Check permissions.
1490                if (desc.dpl < m5reg.cpl) {
1491                    fault = new GeneralProtection(selector);
1492                    break;
1493                }
1494                // Fall through on purpose
1495              case SegIntGateCheck:
1496                // Make sure the gate's the right type.
1497                if ((m5reg.mode == LongMode && (desc.type & 0xe) != 0xe) ||
1498                        ((desc.type & 0x6) != 0x6)) {
1499                    fault = new GeneralProtection(0);
1500                }
1501                break;
1502              case SegSSCheck:
1503                if (selector.si || selector.ti) {
1504                    if (!desc.p) {
1505                        fault = new StackFault(selector);
1506                    } else if (!(desc.s == 1 && desc.type.codeOrData == 0 &&
1507                                desc.type.w) ||
1508                            (desc.dpl != m5reg.cpl) ||
1509                            (selector.rpl != m5reg.cpl)) {
1510                        fault = new GeneralProtection(selector);
1511                    }
1512                } else if (m5reg.submode != SixtyFourBitMode ||
1513                        m5reg.cpl == 3) {
1514                    fault = new GeneralProtection(selector);
1515                }
1516                break;
1517              case SegIretCheck:
1518                {
1519                    if ((!selector.si && !selector.ti) ||
1520                            (selector.rpl < m5reg.cpl) ||
1521                            !(desc.s == 1 && desc.type.codeOrData == 1) ||
1522                            (!desc.type.c && desc.dpl != selector.rpl) ||
1523                            (desc.type.c && desc.dpl > selector.rpl)) {
1524                        fault = new GeneralProtection(selector);
1525                    } else if (!desc.p) {
1526                        fault = new SegmentNotPresent(selector);
1527                    }
1528                    break;
1529                }
1530              case SegIntCSCheck:
1531                if (m5reg.mode == LongMode) {
1532                    if (desc.l != 1 || desc.d != 0) {
1533                        fault = new GeneralProtection(selector);
1534                    }
1535                } else {
1536                    fault = new GenericISA::M5PanicFault("Interrupt CS "
1537                            "checks not implemented in legacy mode.\\n");
1538                }
1539                break;
1540              case SegTRCheck:
1541                if (!selector.si || selector.ti) {
1542                    fault = new GeneralProtection(selector);
1543                }
1544                break;
1545              case SegTSSCheck:
1546                if (!desc.p) {
1547                    fault = new SegmentNotPresent(selector);
1548                } else if (!(desc.type == 0x9 ||
1549                        (desc.type == 1 &&
1550                         m5reg.mode != LongMode))) {
1551                    fault = new GeneralProtection(selector);
1552                }
1553                break;
1554              case SegInGDTCheck:
1555                if (selector.ti) {
1556                    fault = new GeneralProtection(selector);
1557                }
1558                break;
1559              case SegLDTCheck:
1560                if (!desc.p) {
1561                    fault = new SegmentNotPresent(selector);
1562                } else if (desc.type != 0x2) {
1563                    fault = new GeneralProtection(selector);
1564                }
1565                break;
1566              default:
1567                fault = new GenericISA::M5PanicFault(
1568                        "Undefined segment check type.\\n");
1569            }
1570        '''
1571        flag_code = '''
1572            // Check for a NULL selector and set ZF,EZF appropriately.
1573            PredccFlagBits = PredccFlagBits & ~(ext & ZFBit);
1574            PredezfBit = PredezfBit & ~(ext & EZFBit);
1575
1576            if (!selector.si && !selector.ti) {
1577                PredccFlagBits = PredccFlagBits | (ext & ZFBit);
1578                PredezfBit = PredezfBit | (ext & EZFBit);
1579            }
1580        '''
1581
1582    class Wrdh(RegOp):
1583        code = '''
1584            SegDescriptor desc = SrcReg1;
1585
1586            uint64_t target = bits(SrcReg2, 31, 0) << 32;
1587            switch(desc.type) {
1588              case LDT64:
1589              case AvailableTSS64:
1590              case BusyTSS64:
1591                replaceBits(target, 23, 0, desc.baseLow);
1592                replaceBits(target, 31, 24, desc.baseHigh);
1593                break;
1594              case CallGate64:
1595              case IntGate64:
1596              case TrapGate64:
1597                replaceBits(target, 15, 0, bits(desc, 15, 0));
1598                replaceBits(target, 31, 16, bits(desc, 63, 48));
1599                break;
1600              default:
1601                fault = new GenericISA::M5PanicFault(
1602                        "Wrdh used with wrong descriptor type!\\n");
1603            }
1604            DestReg = target;
1605        '''
1606
1607    class Wrtsc(WrRegOp):
1608        code = '''
1609            TscOp = psrc1;
1610        '''
1611
1612    class Rdtsc(RdRegOp):
1613        code = '''
1614            DestReg = TscOp;
1615        '''
1616
1617    class Rdm5reg(RdRegOp):
1618        code = '''
1619            DestReg = M5Reg;
1620        '''
1621
1622    class Wrdl(RegOp):
1623        code = '''
1624            SegDescriptor desc = SrcReg1;
1625            SegSelector selector = SrcReg2;
1626            // This while loop is so we can use break statements in the code
1627            // below to skip the rest of this section without a bunch of
1628            // nesting.
1629            while (true) {
1630                if (selector.si || selector.ti) {
1631                    if (!desc.p) {
1632                        fault = new GenericISA::M5PanicFault(
1633                                "Segment not present.\\n");
1634                        break;
1635                    }
1636                    SegAttr attr = 0;
1637                    attr.dpl = desc.dpl;
1638                    attr.unusable = 0;
1639                    attr.defaultSize = desc.d;
1640                    attr.longMode = desc.l;
1641                    attr.avl = desc.avl;
1642                    attr.granularity = desc.g;
1643                    attr.present = desc.p;
1644                    attr.system = desc.s;
1645                    attr.type = desc.type;
1646                    if (!desc.s) {
1647                        // The expand down bit happens to be set for gates.
1648                        if (desc.type.e) {
1649                            fault = new GenericISA::M5PanicFault(
1650                                    "Gate descriptor encountered.\\n");
1651                            break;
1652                        }
1653                        attr.readable = 1;
1654                        attr.writable = 1;
1655                        attr.expandDown = 0;
1656                    } else {
1657                        if (desc.type.codeOrData) {
1658                            attr.expandDown = 0;
1659                            attr.readable = desc.type.r;
1660                            attr.writable = 0;
1661                        } else {
1662                            attr.expandDown = desc.type.e;
1663                            attr.readable = 1;
1664                            attr.writable = desc.type.w;
1665                        }
1666                    }
1667                    Addr base = desc.baseLow | (desc.baseHigh << 24);
1668                    Addr limit = desc.limitLow | (desc.limitHigh << 16);
1669                    if (desc.g)
1670                        limit = (limit << 12) | mask(12);
1671                    SegBaseDest = base;
1672                    SegLimitDest = limit;
1673                    SegAttrDest = attr;
1674                } else {
1675                    SegBaseDest = SegBaseDest;
1676                    SegLimitDest = SegLimitDest;
1677                    SegAttrDest = SegAttrDest;
1678                }
1679                break;
1680            }
1681        '''
1682
1683    class Wrxftw(WrRegOp):
1684        def __init__(self, src1, **kwargs):
1685            super(Wrxftw, self).__init__(src1, "InstRegIndex(NUM_INTREGS)", \
1686                                         **kwargs)
1687
1688        code = '''
1689            FTW = X86ISA::convX87XTagsToTags(SrcReg1);
1690        '''
1691
1692    class Rdxftw(RdRegOp):
1693        code = '''
1694            DestReg = X86ISA::convX87TagsToXTags(FTW);
1695        '''
1696}};
1697