hotspot/src/cpu/sparc/vm/assembler_sparc.hpp
changeset 1 489c9b5090e2
child 360 21d113ecbf6a
equal deleted inserted replaced
0:fd16c54261b3 1:489c9b5090e2
       
     1 /*
       
     2  * Copyright 1997-2007 Sun Microsystems, Inc.  All Rights Reserved.
       
     3  * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
       
     4  *
       
     5  * This code is free software; you can redistribute it and/or modify it
       
     6  * under the terms of the GNU General Public License version 2 only, as
       
     7  * published by the Free Software Foundation.
       
     8  *
       
     9  * This code is distributed in the hope that it will be useful, but WITHOUT
       
    10  * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
       
    11  * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License
       
    12  * version 2 for more details (a copy is included in the LICENSE file that
       
    13  * accompanied this code).
       
    14  *
       
    15  * You should have received a copy of the GNU General Public License version
       
    16  * 2 along with this work; if not, write to the Free Software Foundation,
       
    17  * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA.
       
    18  *
       
    19  * Please contact Sun Microsystems, Inc., 4150 Network Circle, Santa Clara,
       
    20  * CA 95054 USA or visit www.sun.com if you need additional information or
       
    21  * have any questions.
       
    22  *
       
    23  */
       
    24 
       
    25 class BiasedLockingCounters;
       
    26 
       
    27 // <sys/trap.h> promises that the system will not use traps 16-31
       
    28 #define ST_RESERVED_FOR_USER_0 0x10
       
    29 
       
    30 /* Written: David Ungar 4/19/97 */
       
    31 
       
    32 // Contains all the definitions needed for sparc assembly code generation.
       
    33 
       
    34 // Register aliases for parts of the system:
       
    35 
       
    36 // 64 bit values can be kept in g1-g5, o1-o5 and o7 and all 64 bits are safe
       
    37 // across context switches in V8+ ABI.  Of course, there are no 64 bit regs
       
    38 // in V8 ABI. All 64 bits are preserved in V9 ABI for all registers.
       
    39 
       
    40 // g2-g4 are scratch registers called "application globals".  Their
       
    41 // meaning is reserved to the "compilation system"--which means us!
       
    42 // They are are not supposed to be touched by ordinary C code, although
       
    43 // highly-optimized C code might steal them for temps.  They are safe
       
    44 // across thread switches, and the ABI requires that they be safe
       
    45 // across function calls.
       
    46 //
       
    47 // g1 and g3 are touched by more modules.  V8 allows g1 to be clobbered
       
    48 // across func calls, and V8+ also allows g5 to be clobbered across
       
    49 // func calls.  Also, g1 and g5 can get touched while doing shared
       
    50 // library loading.
       
    51 //
       
    52 // We must not touch g7 (it is the thread-self register) and g6 is
       
    53 // reserved for certain tools.  g0, of course, is always zero.
       
    54 //
       
    55 // (Sources:  SunSoft Compilers Group, thread library engineers.)
       
    56 
       
    57 // %%%% The interpreter should be revisited to reduce global scratch regs.
       
    58 
       
    59 // This global always holds the current JavaThread pointer:
       
    60 
       
    61 REGISTER_DECLARATION(Register, G2_thread , G2);
       
    62 
       
    63 // The following globals are part of the Java calling convention:
       
    64 
       
    65 REGISTER_DECLARATION(Register, G5_method             , G5);
       
    66 REGISTER_DECLARATION(Register, G5_megamorphic_method , G5_method);
       
    67 REGISTER_DECLARATION(Register, G5_inline_cache_reg   , G5_method);
       
    68 
       
    69 // The following globals are used for the new C1 & interpreter calling convention:
       
    70 REGISTER_DECLARATION(Register, Gargs        , G4); // pointing to the last argument
       
    71 
       
    72 // This local is used to preserve G2_thread in the interpreter and in stubs:
       
    73 REGISTER_DECLARATION(Register, L7_thread_cache , L7);
       
    74 
       
    75 // These globals are used as scratch registers in the interpreter:
       
    76 
       
    77 REGISTER_DECLARATION(Register, Gframe_size   , G1); // SAME REG as G1_scratch
       
    78 REGISTER_DECLARATION(Register, G1_scratch    , G1); // also SAME
       
    79 REGISTER_DECLARATION(Register, G3_scratch    , G3);
       
    80 REGISTER_DECLARATION(Register, G4_scratch    , G4);
       
    81 
       
    82 // These globals are used as short-lived scratch registers in the compiler:
       
    83 
       
    84 REGISTER_DECLARATION(Register, Gtemp  , G5);
       
    85 
       
    86 // The compiler requires that G5_megamorphic_method is G5_inline_cache_klass,
       
    87 // because a single patchable "set" instruction (NativeMovConstReg,
       
    88 // or NativeMovConstPatching for compiler1) instruction
       
    89 // serves to set up either quantity, depending on whether the compiled
       
    90 // call site is an inline cache or is megamorphic.  See the function
       
    91 // CompiledIC::set_to_megamorphic.
       
    92 //
       
    93 // On the other hand, G5_inline_cache_klass must differ from G5_method,
       
    94 // because both registers are needed for an inline cache that calls
       
    95 // an interpreted method.
       
    96 //
       
    97 // Note that G5_method is only the method-self for the interpreter,
       
    98 // and is logically unrelated to G5_megamorphic_method.
       
    99 //
       
   100 // Invariants on G2_thread (the JavaThread pointer):
       
   101 //  - it should not be used for any other purpose anywhere
       
   102 //  - it must be re-initialized by StubRoutines::call_stub()
       
   103 //  - it must be preserved around every use of call_VM
       
   104 
       
   105 // We can consider using g2/g3/g4 to cache more values than the
       
   106 // JavaThread, such as the card-marking base or perhaps pointers into
       
   107 // Eden.  It's something of a waste to use them as scratch temporaries,
       
   108 // since they are not supposed to be volatile.  (Of course, if we find
       
   109 // that Java doesn't benefit from application globals, then we can just
       
   110 // use them as ordinary temporaries.)
       
   111 //
       
   112 // Since g1 and g5 (and/or g6) are the volatile (caller-save) registers,
       
   113 // it makes sense to use them routinely for procedure linkage,
       
   114 // whenever the On registers are not applicable.  Examples:  G5_method,
       
   115 // G5_inline_cache_klass, and a double handful of miscellaneous compiler
       
   116 // stubs.  This means that compiler stubs, etc., should be kept to a
       
   117 // maximum of two or three G-register arguments.
       
   118 
       
   119 
       
   120 // stub frames
       
   121 
       
   122 REGISTER_DECLARATION(Register, Lentry_args      , L0); // pointer to args passed to callee (interpreter) not stub itself
       
   123 
       
   124 // Interpreter frames
       
   125 
       
   126 #ifdef CC_INTERP
       
   127 REGISTER_DECLARATION(Register, Lstate           , L0); // interpreter state object pointer
       
   128 REGISTER_DECLARATION(Register, L1_scratch       , L1); // scratch
       
   129 REGISTER_DECLARATION(Register, Lmirror          , L1); // mirror (for native methods only)
       
   130 REGISTER_DECLARATION(Register, L2_scratch       , L2);
       
   131 REGISTER_DECLARATION(Register, L3_scratch       , L3);
       
   132 REGISTER_DECLARATION(Register, L4_scratch       , L4);
       
   133 REGISTER_DECLARATION(Register, Lscratch         , L5); // C1 uses
       
   134 REGISTER_DECLARATION(Register, Lscratch2        , L6); // C1 uses
       
   135 REGISTER_DECLARATION(Register, L7_scratch       , L7); // constant pool cache
       
   136 REGISTER_DECLARATION(Register, O5_savedSP       , O5);
       
   137 REGISTER_DECLARATION(Register, I5_savedSP       , I5); // Saved SP before bumping for locals.  This is simply
       
   138                                                        // a copy SP, so in 64-bit it's a biased value.  The bias
       
   139                                                        // is added and removed as needed in the frame code.
       
   140 // Interface to signature handler
       
   141 REGISTER_DECLARATION(Register, Llocals          , L7); // pointer to locals for signature handler
       
   142 REGISTER_DECLARATION(Register, Lmethod          , L6); // methodOop when calling signature handler
       
   143 
       
   144 #else
       
   145 REGISTER_DECLARATION(Register, Lesp             , L0); // expression stack pointer
       
   146 REGISTER_DECLARATION(Register, Lbcp             , L1); // pointer to next bytecode
       
   147 REGISTER_DECLARATION(Register, Lmethod          , L2);
       
   148 REGISTER_DECLARATION(Register, Llocals          , L3);
       
   149 REGISTER_DECLARATION(Register, Largs            , L3); // pointer to locals for signature handler
       
   150                                                        // must match Llocals in asm interpreter
       
   151 REGISTER_DECLARATION(Register, Lmonitors        , L4);
       
   152 REGISTER_DECLARATION(Register, Lbyte_code       , L5);
       
   153 // When calling out from the interpreter we record SP so that we can remove any extra stack
       
   154 // space allocated during adapter transitions. This register is only live from the point
       
   155 // of the call until we return.
       
   156 REGISTER_DECLARATION(Register, Llast_SP         , L5);
       
   157 REGISTER_DECLARATION(Register, Lscratch         , L5);
       
   158 REGISTER_DECLARATION(Register, Lscratch2        , L6);
       
   159 REGISTER_DECLARATION(Register, LcpoolCache      , L6); // constant pool cache
       
   160 
       
   161 REGISTER_DECLARATION(Register, O5_savedSP       , O5);
       
   162 REGISTER_DECLARATION(Register, I5_savedSP       , I5); // Saved SP before bumping for locals.  This is simply
       
   163                                                        // a copy SP, so in 64-bit it's a biased value.  The bias
       
   164                                                        // is added and removed as needed in the frame code.
       
   165 REGISTER_DECLARATION(Register, IdispatchTables  , I4); // Base address of the bytecode dispatch tables
       
   166 REGISTER_DECLARATION(Register, IdispatchAddress , I3); // Register which saves the dispatch address for each bytecode
       
   167 REGISTER_DECLARATION(Register, ImethodDataPtr   , I2); // Pointer to the current method data
       
   168 #endif /* CC_INTERP */
       
   169 
       
   170 // NOTE: Lscratch2 and LcpoolCache point to the same registers in
       
   171 //       the interpreter code. If Lscratch2 needs to be used for some
       
   172 //       purpose than LcpoolCache should be restore after that for
       
   173 //       the interpreter to work right
       
   174 // (These assignments must be compatible with L7_thread_cache; see above.)
       
   175 
       
   176 // Since Lbcp points into the middle of the method object,
       
   177 // it is temporarily converted into a "bcx" during GC.
       
   178 
       
   179 // Exception processing
       
   180 // These registers are passed into exception handlers.
       
   181 // All exception handlers require the exception object being thrown.
       
   182 // In addition, an nmethod's exception handler must be passed
       
   183 // the address of the call site within the nmethod, to allow
       
   184 // proper selection of the applicable catch block.
       
   185 // (Interpreter frames use their own bcp() for this purpose.)
       
   186 //
       
   187 // The Oissuing_pc value is not always needed.  When jumping to a
       
   188 // handler that is known to be interpreted, the Oissuing_pc value can be
       
   189 // omitted.  An actual catch block in compiled code receives (from its
       
   190 // nmethod's exception handler) the thrown exception in the Oexception,
       
   191 // but it doesn't need the Oissuing_pc.
       
   192 //
       
   193 // If an exception handler (either interpreted or compiled)
       
   194 // discovers there is no applicable catch block, it updates
       
   195 // the Oissuing_pc to the continuation PC of its own caller,
       
   196 // pops back to that caller's stack frame, and executes that
       
   197 // caller's exception handler.  Obviously, this process will
       
   198 // iterate until the control stack is popped back to a method
       
   199 // containing an applicable catch block.  A key invariant is
       
   200 // that the Oissuing_pc value is always a value local to
       
   201 // the method whose exception handler is currently executing.
       
   202 //
       
   203 // Note:  The issuing PC value is __not__ a raw return address (I7 value).
       
   204 // It is a "return pc", the address __following__ the call.
       
   205 // Raw return addresses are converted to issuing PCs by frame::pc(),
       
   206 // or by stubs.  Issuing PCs can be used directly with PC range tables.
       
   207 //
       
   208 REGISTER_DECLARATION(Register, Oexception  , O0); // exception being thrown
       
   209 REGISTER_DECLARATION(Register, Oissuing_pc , O1); // where the exception is coming from
       
   210 
       
   211 
       
   212 // These must occur after the declarations above
       
   213 #ifndef DONT_USE_REGISTER_DEFINES
       
   214 
       
   215 #define Gthread             AS_REGISTER(Register, Gthread)
       
   216 #define Gmethod             AS_REGISTER(Register, Gmethod)
       
   217 #define Gmegamorphic_method AS_REGISTER(Register, Gmegamorphic_method)
       
   218 #define Ginline_cache_reg   AS_REGISTER(Register, Ginline_cache_reg)
       
   219 #define Gargs               AS_REGISTER(Register, Gargs)
       
   220 #define Lthread_cache       AS_REGISTER(Register, Lthread_cache)
       
   221 #define Gframe_size         AS_REGISTER(Register, Gframe_size)
       
   222 #define Gtemp               AS_REGISTER(Register, Gtemp)
       
   223 
       
   224 #ifdef CC_INTERP
       
   225 #define Lstate              AS_REGISTER(Register, Lstate)
       
   226 #define Lesp                AS_REGISTER(Register, Lesp)
       
   227 #define L1_scratch          AS_REGISTER(Register, L1_scratch)
       
   228 #define Lmirror             AS_REGISTER(Register, Lmirror)
       
   229 #define L2_scratch          AS_REGISTER(Register, L2_scratch)
       
   230 #define L3_scratch          AS_REGISTER(Register, L3_scratch)
       
   231 #define L4_scratch          AS_REGISTER(Register, L4_scratch)
       
   232 #define Lscratch            AS_REGISTER(Register, Lscratch)
       
   233 #define Lscratch2           AS_REGISTER(Register, Lscratch2)
       
   234 #define L7_scratch          AS_REGISTER(Register, L7_scratch)
       
   235 #define Ostate              AS_REGISTER(Register, Ostate)
       
   236 #else
       
   237 #define Lesp                AS_REGISTER(Register, Lesp)
       
   238 #define Lbcp                AS_REGISTER(Register, Lbcp)
       
   239 #define Lmethod             AS_REGISTER(Register, Lmethod)
       
   240 #define Llocals             AS_REGISTER(Register, Llocals)
       
   241 #define Lmonitors           AS_REGISTER(Register, Lmonitors)
       
   242 #define Lbyte_code          AS_REGISTER(Register, Lbyte_code)
       
   243 #define Lscratch            AS_REGISTER(Register, Lscratch)
       
   244 #define Lscratch2           AS_REGISTER(Register, Lscratch2)
       
   245 #define LcpoolCache         AS_REGISTER(Register, LcpoolCache)
       
   246 #endif /* ! CC_INTERP */
       
   247 
       
   248 #define Lentry_args         AS_REGISTER(Register, Lentry_args)
       
   249 #define I5_savedSP          AS_REGISTER(Register, I5_savedSP)
       
   250 #define O5_savedSP          AS_REGISTER(Register, O5_savedSP)
       
   251 #define IdispatchAddress    AS_REGISTER(Register, IdispatchAddress)
       
   252 #define ImethodDataPtr      AS_REGISTER(Register, ImethodDataPtr)
       
   253 #define IdispatchTables     AS_REGISTER(Register, IdispatchTables)
       
   254 
       
   255 #define Oexception          AS_REGISTER(Register, Oexception)
       
   256 #define Oissuing_pc         AS_REGISTER(Register, Oissuing_pc)
       
   257 
       
   258 
       
   259 #endif
       
   260 
       
   261 // Address is an abstraction used to represent a memory location.
       
   262 //
       
   263 // Note: A register location is represented via a Register, not
       
   264 //       via an address for efficiency & simplicity reasons.
       
   265 
       
   266 class Address VALUE_OBJ_CLASS_SPEC {
       
   267  private:
       
   268   Register              _base;
       
   269 #ifdef _LP64
       
   270   int                   _hi32;          // bits 63::32
       
   271   int                   _low32;         // bits 31::0
       
   272 #endif
       
   273   int                   _hi;
       
   274   int                   _disp;
       
   275   RelocationHolder      _rspec;
       
   276 
       
   277   RelocationHolder rspec_from_rtype(relocInfo::relocType rt, address a = NULL) {
       
   278     switch (rt) {
       
   279     case relocInfo::external_word_type:
       
   280       return external_word_Relocation::spec(a);
       
   281     case relocInfo::internal_word_type:
       
   282       return internal_word_Relocation::spec(a);
       
   283 #ifdef _LP64
       
   284     case relocInfo::opt_virtual_call_type:
       
   285       return opt_virtual_call_Relocation::spec();
       
   286     case relocInfo::static_call_type:
       
   287       return static_call_Relocation::spec();
       
   288     case relocInfo::runtime_call_type:
       
   289       return runtime_call_Relocation::spec();
       
   290 #endif
       
   291     case relocInfo::none:
       
   292       return RelocationHolder();
       
   293     default:
       
   294       ShouldNotReachHere();
       
   295       return RelocationHolder();
       
   296     }
       
   297   }
       
   298 
       
   299  public:
       
   300   Address(Register b, address a, relocInfo::relocType rt = relocInfo::none)
       
   301     : _rspec(rspec_from_rtype(rt, a))
       
   302   {
       
   303     _base  = b;
       
   304 #ifdef _LP64
       
   305     _hi32  = (intptr_t)a >> 32;    // top 32 bits in 64 bit word
       
   306     _low32 = (intptr_t)a & ~0;     // low 32 bits in 64 bit word
       
   307 #endif
       
   308     _hi    = (intptr_t)a & ~0x3ff; // top    22 bits in low word
       
   309     _disp  = (intptr_t)a &  0x3ff; // bottom 10 bits
       
   310   }
       
   311 
       
   312   Address(Register b, address a, RelocationHolder const& rspec)
       
   313     : _rspec(rspec)
       
   314   {
       
   315     _base  = b;
       
   316 #ifdef _LP64
       
   317     _hi32  = (intptr_t)a >> 32;    // top 32 bits in 64 bit word
       
   318     _low32 = (intptr_t)a & ~0;     // low 32 bits in 64 bit word
       
   319 #endif
       
   320     _hi    = (intptr_t)a & ~0x3ff; // top    22 bits
       
   321     _disp  = (intptr_t)a &  0x3ff; // bottom 10 bits
       
   322   }
       
   323 
       
   324   Address(Register b, intptr_t h, intptr_t d, RelocationHolder const& rspec = RelocationHolder())
       
   325     : _rspec(rspec)
       
   326   {
       
   327     _base  = b;
       
   328 #ifdef _LP64
       
   329 // [RGV] Put in Assert to force me to check usage of this constructor
       
   330      assert( h == 0, "Check usage of this constructor" );
       
   331     _hi32  = h;
       
   332     _low32 = d;
       
   333     _hi    = h;
       
   334     _disp  = d;
       
   335 #else
       
   336     _hi    = h;
       
   337     _disp  = d;
       
   338 #endif
       
   339   }
       
   340 
       
   341   Address()
       
   342     : _rspec(RelocationHolder())
       
   343   {
       
   344     _base  = G0;
       
   345 #ifdef _LP64
       
   346     _hi32  = 0;
       
   347     _low32 = 0;
       
   348 #endif
       
   349     _hi    = 0;
       
   350     _disp  = 0;
       
   351   }
       
   352 
       
   353   // fancier constructors
       
   354 
       
   355   enum addr_type {
       
   356     extra_in_argument,  // in the In registers
       
   357     extra_out_argument  // in the Outs
       
   358   };
       
   359 
       
   360   Address( addr_type, int );
       
   361 
       
   362   // accessors
       
   363 
       
   364   Register               base() const { return _base; }
       
   365 #ifdef _LP64
       
   366   int                   hi32()  const { return _hi32; }
       
   367   int                   low32() const { return _low32; }
       
   368 #endif
       
   369   int                      hi() const { return _hi;  }
       
   370   int                    disp() const { return _disp; }
       
   371 #ifdef _LP64
       
   372   intptr_t              value() const { return ((intptr_t)_hi32 << 32) |
       
   373                                                 (intptr_t)(uint32_t)_low32; }
       
   374 #else
       
   375   int                   value() const { return _hi | _disp; }
       
   376 #endif
       
   377   const relocInfo::relocType  rtype() { return _rspec.type(); }
       
   378   const RelocationHolder&     rspec() { return _rspec; }
       
   379 
       
   380   RelocationHolder      rspec(int offset) const {
       
   381     return offset == 0 ? _rspec : _rspec.plus(offset);
       
   382   }
       
   383 
       
   384   inline bool is_simm13(int offset = 0);  // check disp+offset for overflow
       
   385 
       
   386   Address split_disp() const {            // deal with disp overflow
       
   387     Address a = (*this);
       
   388     int hi_disp = _disp & ~0x3ff;
       
   389     if (hi_disp != 0) {
       
   390       a._disp -= hi_disp;
       
   391       a._hi   += hi_disp;
       
   392     }
       
   393     return a;
       
   394   }
       
   395 
       
   396   Address after_save() const {
       
   397     Address a = (*this);
       
   398     a._base = a._base->after_save();
       
   399     return a;
       
   400   }
       
   401 
       
   402   Address after_restore() const {
       
   403     Address a = (*this);
       
   404     a._base = a._base->after_restore();
       
   405     return a;
       
   406   }
       
   407 
       
   408   friend class Assembler;
       
   409 };
       
   410 
       
   411 
       
   412 inline Address RegisterImpl::address_in_saved_window() const {
       
   413    return (Address(SP, 0, (sp_offset_in_saved_window() * wordSize) + STACK_BIAS));
       
   414 }
       
   415 
       
   416 
       
   417 
       
   418 // Argument is an abstraction used to represent an outgoing
       
   419 // actual argument or an incoming formal parameter, whether
       
   420 // it resides in memory or in a register, in a manner consistent
       
   421 // with the SPARC Application Binary Interface, or ABI.  This is
       
   422 // often referred to as the native or C calling convention.
       
   423 
       
   424 class Argument VALUE_OBJ_CLASS_SPEC {
       
   425  private:
       
   426   int _number;
       
   427   bool _is_in;
       
   428 
       
   429  public:
       
   430 #ifdef _LP64
       
   431   enum {
       
   432     n_register_parameters = 6,          // only 6 registers may contain integer parameters
       
   433     n_float_register_parameters = 16    // Can have up to 16 floating registers
       
   434   };
       
   435 #else
       
   436   enum {
       
   437     n_register_parameters = 6           // only 6 registers may contain integer parameters
       
   438   };
       
   439 #endif
       
   440 
       
   441   // creation
       
   442   Argument(int number, bool is_in) : _number(number), _is_in(is_in) {}
       
   443 
       
   444   int  number() const  { return _number;  }
       
   445   bool is_in()  const  { return _is_in;   }
       
   446   bool is_out() const  { return !is_in(); }
       
   447 
       
   448   Argument successor() const  { return Argument(number() + 1, is_in()); }
       
   449   Argument as_in()     const  { return Argument(number(), true ); }
       
   450   Argument as_out()    const  { return Argument(number(), false); }
       
   451 
       
   452   // locating register-based arguments:
       
   453   bool is_register() const { return _number < n_register_parameters; }
       
   454 
       
   455 #ifdef _LP64
       
   456   // locating Floating Point register-based arguments:
       
   457   bool is_float_register() const { return _number < n_float_register_parameters; }
       
   458 
       
   459   FloatRegister as_float_register() const {
       
   460     assert(is_float_register(), "must be a register argument");
       
   461     return as_FloatRegister(( number() *2 ) + 1);
       
   462   }
       
   463   FloatRegister as_double_register() const {
       
   464     assert(is_float_register(), "must be a register argument");
       
   465     return as_FloatRegister(( number() *2 ));
       
   466   }
       
   467 #endif
       
   468 
       
   469   Register as_register() const {
       
   470     assert(is_register(), "must be a register argument");
       
   471     return is_in() ? as_iRegister(number()) : as_oRegister(number());
       
   472   }
       
   473 
       
   474   // locating memory-based arguments
       
   475   Address as_address() const {
       
   476     assert(!is_register(), "must be a memory argument");
       
   477     return address_in_frame();
       
   478   }
       
   479 
       
   480   // When applied to a register-based argument, give the corresponding address
       
   481   // into the 6-word area "into which callee may store register arguments"
       
   482   // (This is a different place than the corresponding register-save area location.)
       
   483   Address address_in_frame() const {
       
   484     return Address( is_in()   ? Address::extra_in_argument
       
   485                               : Address::extra_out_argument,
       
   486                     _number );
       
   487   }
       
   488 
       
   489   // debugging
       
   490   const char* name() const;
       
   491 
       
   492   friend class Assembler;
       
   493 };
       
   494 
       
   495 
       
   496 // The SPARC Assembler: Pure assembler doing NO optimizations on the instruction
       
   497 // level; i.e., what you write
       
   498 // is what you get. The Assembler is generating code into a CodeBuffer.
       
   499 
       
   500 class Assembler : public AbstractAssembler  {
       
   501  protected:
       
   502 
       
   503   static void print_instruction(int inst);
       
   504   static int  patched_branch(int dest_pos, int inst, int inst_pos);
       
   505   static int  branch_destination(int inst, int pos);
       
   506 
       
   507 
       
   508   friend class AbstractAssembler;
       
   509 
       
   510   // code patchers need various routines like inv_wdisp()
       
   511   friend class NativeInstruction;
       
   512   friend class NativeGeneralJump;
       
   513   friend class Relocation;
       
   514   friend class Label;
       
   515 
       
   516  public:
       
   517   // op carries format info; see page 62 & 267
       
   518 
       
   519   enum ops {
       
   520     call_op   = 1, // fmt 1
       
   521     branch_op = 0, // also sethi (fmt2)
       
   522     arith_op  = 2, // fmt 3, arith & misc
       
   523     ldst_op   = 3  // fmt 3, load/store
       
   524   };
       
   525 
       
   526   enum op2s {
       
   527     bpr_op2   = 3,
       
   528     fb_op2    = 6,
       
   529     fbp_op2   = 5,
       
   530     br_op2    = 2,
       
   531     bp_op2    = 1,
       
   532     cb_op2    = 7, // V8
       
   533     sethi_op2 = 4
       
   534   };
       
   535 
       
   536   enum op3s {
       
   537     // selected op3s
       
   538     add_op3      = 0x00,
       
   539     and_op3      = 0x01,
       
   540     or_op3       = 0x02,
       
   541     xor_op3      = 0x03,
       
   542     sub_op3      = 0x04,
       
   543     andn_op3     = 0x05,
       
   544     orn_op3      = 0x06,
       
   545     xnor_op3     = 0x07,
       
   546     addc_op3     = 0x08,
       
   547     mulx_op3     = 0x09,
       
   548     umul_op3     = 0x0a,
       
   549     smul_op3     = 0x0b,
       
   550     subc_op3     = 0x0c,
       
   551     udivx_op3    = 0x0d,
       
   552     udiv_op3     = 0x0e,
       
   553     sdiv_op3     = 0x0f,
       
   554 
       
   555     addcc_op3    = 0x10,
       
   556     andcc_op3    = 0x11,
       
   557     orcc_op3     = 0x12,
       
   558     xorcc_op3    = 0x13,
       
   559     subcc_op3    = 0x14,
       
   560     andncc_op3   = 0x15,
       
   561     orncc_op3    = 0x16,
       
   562     xnorcc_op3   = 0x17,
       
   563     addccc_op3   = 0x18,
       
   564     umulcc_op3   = 0x1a,
       
   565     smulcc_op3   = 0x1b,
       
   566     subccc_op3   = 0x1c,
       
   567     udivcc_op3   = 0x1e,
       
   568     sdivcc_op3   = 0x1f,
       
   569 
       
   570     taddcc_op3   = 0x20,
       
   571     tsubcc_op3   = 0x21,
       
   572     taddcctv_op3 = 0x22,
       
   573     tsubcctv_op3 = 0x23,
       
   574     mulscc_op3   = 0x24,
       
   575     sll_op3      = 0x25,
       
   576     sllx_op3     = 0x25,
       
   577     srl_op3      = 0x26,
       
   578     srlx_op3     = 0x26,
       
   579     sra_op3      = 0x27,
       
   580     srax_op3     = 0x27,
       
   581     rdreg_op3    = 0x28,
       
   582     membar_op3   = 0x28,
       
   583 
       
   584     flushw_op3   = 0x2b,
       
   585     movcc_op3    = 0x2c,
       
   586     sdivx_op3    = 0x2d,
       
   587     popc_op3     = 0x2e,
       
   588     movr_op3     = 0x2f,
       
   589 
       
   590     sir_op3      = 0x30,
       
   591     wrreg_op3    = 0x30,
       
   592     saved_op3    = 0x31,
       
   593 
       
   594     fpop1_op3    = 0x34,
       
   595     fpop2_op3    = 0x35,
       
   596     impdep1_op3  = 0x36,
       
   597     impdep2_op3  = 0x37,
       
   598     jmpl_op3     = 0x38,
       
   599     rett_op3     = 0x39,
       
   600     trap_op3     = 0x3a,
       
   601     flush_op3    = 0x3b,
       
   602     save_op3     = 0x3c,
       
   603     restore_op3  = 0x3d,
       
   604     done_op3     = 0x3e,
       
   605     retry_op3    = 0x3e,
       
   606 
       
   607     lduw_op3     = 0x00,
       
   608     ldub_op3     = 0x01,
       
   609     lduh_op3     = 0x02,
       
   610     ldd_op3      = 0x03,
       
   611     stw_op3      = 0x04,
       
   612     stb_op3      = 0x05,
       
   613     sth_op3      = 0x06,
       
   614     std_op3      = 0x07,
       
   615     ldsw_op3     = 0x08,
       
   616     ldsb_op3     = 0x09,
       
   617     ldsh_op3     = 0x0a,
       
   618     ldx_op3      = 0x0b,
       
   619 
       
   620     ldstub_op3   = 0x0d,
       
   621     stx_op3      = 0x0e,
       
   622     swap_op3     = 0x0f,
       
   623 
       
   624     lduwa_op3    = 0x10,
       
   625     ldxa_op3     = 0x1b,
       
   626 
       
   627     stwa_op3     = 0x14,
       
   628     stxa_op3     = 0x1e,
       
   629 
       
   630     ldf_op3      = 0x20,
       
   631     ldfsr_op3    = 0x21,
       
   632     ldqf_op3     = 0x22,
       
   633     lddf_op3     = 0x23,
       
   634     stf_op3      = 0x24,
       
   635     stfsr_op3    = 0x25,
       
   636     stqf_op3     = 0x26,
       
   637     stdf_op3     = 0x27,
       
   638 
       
   639     prefetch_op3 = 0x2d,
       
   640 
       
   641 
       
   642     ldc_op3      = 0x30,
       
   643     ldcsr_op3    = 0x31,
       
   644     lddc_op3     = 0x33,
       
   645     stc_op3      = 0x34,
       
   646     stcsr_op3    = 0x35,
       
   647     stdcq_op3    = 0x36,
       
   648     stdc_op3     = 0x37,
       
   649 
       
   650     casa_op3     = 0x3c,
       
   651     casxa_op3    = 0x3e,
       
   652 
       
   653     alt_bit_op3  = 0x10,
       
   654      cc_bit_op3  = 0x10
       
   655   };
       
   656 
       
   657   enum opfs {
       
   658     // selected opfs
       
   659     fmovs_opf   = 0x01,
       
   660     fmovd_opf   = 0x02,
       
   661 
       
   662     fnegs_opf   = 0x05,
       
   663     fnegd_opf   = 0x06,
       
   664 
       
   665     fadds_opf   = 0x41,
       
   666     faddd_opf   = 0x42,
       
   667     fsubs_opf   = 0x45,
       
   668     fsubd_opf   = 0x46,
       
   669 
       
   670     fmuls_opf   = 0x49,
       
   671     fmuld_opf   = 0x4a,
       
   672     fdivs_opf   = 0x4d,
       
   673     fdivd_opf   = 0x4e,
       
   674 
       
   675     fcmps_opf   = 0x51,
       
   676     fcmpd_opf   = 0x52,
       
   677 
       
   678     fstox_opf   = 0x81,
       
   679     fdtox_opf   = 0x82,
       
   680     fxtos_opf   = 0x84,
       
   681     fxtod_opf   = 0x88,
       
   682     fitos_opf   = 0xc4,
       
   683     fdtos_opf   = 0xc6,
       
   684     fitod_opf   = 0xc8,
       
   685     fstod_opf   = 0xc9,
       
   686     fstoi_opf   = 0xd1,
       
   687     fdtoi_opf   = 0xd2
       
   688   };
       
   689 
       
   690   enum RCondition {  rc_z = 1,  rc_lez = 2,  rc_lz = 3, rc_nz = 5, rc_gz = 6, rc_gez = 7  };
       
   691 
       
   692   enum Condition {
       
   693      // for FBfcc & FBPfcc instruction
       
   694     f_never                     = 0,
       
   695     f_notEqual                  = 1,
       
   696     f_notZero                   = 1,
       
   697     f_lessOrGreater             = 2,
       
   698     f_unorderedOrLess           = 3,
       
   699     f_less                      = 4,
       
   700     f_unorderedOrGreater        = 5,
       
   701     f_greater                   = 6,
       
   702     f_unordered                 = 7,
       
   703     f_always                    = 8,
       
   704     f_equal                     = 9,
       
   705     f_zero                      = 9,
       
   706     f_unorderedOrEqual          = 10,
       
   707     f_greaterOrEqual            = 11,
       
   708     f_unorderedOrGreaterOrEqual = 12,
       
   709     f_lessOrEqual               = 13,
       
   710     f_unorderedOrLessOrEqual    = 14,
       
   711     f_ordered                   = 15,
       
   712 
       
   713     // V8 coproc, pp 123 v8 manual
       
   714 
       
   715     cp_always  = 8,
       
   716     cp_never   = 0,
       
   717     cp_3       = 7,
       
   718     cp_2       = 6,
       
   719     cp_2or3    = 5,
       
   720     cp_1       = 4,
       
   721     cp_1or3    = 3,
       
   722     cp_1or2    = 2,
       
   723     cp_1or2or3 = 1,
       
   724     cp_0       = 9,
       
   725     cp_0or3    = 10,
       
   726     cp_0or2    = 11,
       
   727     cp_0or2or3 = 12,
       
   728     cp_0or1    = 13,
       
   729     cp_0or1or3 = 14,
       
   730     cp_0or1or2 = 15,
       
   731 
       
   732 
       
   733     // for integers
       
   734 
       
   735     never                 =  0,
       
   736     equal                 =  1,
       
   737     zero                  =  1,
       
   738     lessEqual             =  2,
       
   739     less                  =  3,
       
   740     lessEqualUnsigned     =  4,
       
   741     lessUnsigned          =  5,
       
   742     carrySet              =  5,
       
   743     negative              =  6,
       
   744     overflowSet           =  7,
       
   745     always                =  8,
       
   746     notEqual              =  9,
       
   747     notZero               =  9,
       
   748     greater               =  10,
       
   749     greaterEqual          =  11,
       
   750     greaterUnsigned       =  12,
       
   751     greaterEqualUnsigned  =  13,
       
   752     carryClear            =  13,
       
   753     positive              =  14,
       
   754     overflowClear         =  15
       
   755   };
       
   756 
       
   757   enum CC {
       
   758     icc  = 0,  xcc  = 2,
       
   759     // ptr_cc is the correct condition code for a pointer or intptr_t:
       
   760     ptr_cc = NOT_LP64(icc) LP64_ONLY(xcc),
       
   761     fcc0 = 0,  fcc1 = 1, fcc2 = 2, fcc3 = 3
       
   762   };
       
   763 
       
   764   enum PrefetchFcn {
       
   765     severalReads = 0,  oneRead = 1,  severalWritesAndPossiblyReads = 2, oneWrite = 3, page = 4
       
   766   };
       
   767 
       
   768  public:
       
   769   // Helper functions for groups of instructions
       
   770 
       
   771   enum Predict { pt = 1, pn = 0 }; // pt = predict taken
       
   772 
       
   773   enum Membar_mask_bits { // page 184, v9
       
   774     StoreStore = 1 << 3,
       
   775     LoadStore  = 1 << 2,
       
   776     StoreLoad  = 1 << 1,
       
   777     LoadLoad   = 1 << 0,
       
   778 
       
   779     Sync       = 1 << 6,
       
   780     MemIssue   = 1 << 5,
       
   781     Lookaside  = 1 << 4
       
   782   };
       
   783 
       
   784   // test if x is within signed immediate range for nbits
       
   785   static bool is_simm(int x, int nbits) { return -( 1 << nbits-1 )  <= x   &&   x  <  ( 1 << nbits-1 ); }
       
   786 
       
   787   // test if -4096 <= x <= 4095
       
   788   static bool is_simm13(int x) { return is_simm(x, 13); }
       
   789 
       
   790   enum ASIs { // page 72, v9
       
   791     ASI_PRIMARY        = 0x80,
       
   792     ASI_PRIMARY_LITTLE = 0x88
       
   793     // add more from book as needed
       
   794   };
       
   795 
       
   796  protected:
       
   797   // helpers
       
   798 
       
   799   // x is supposed to fit in a field "nbits" wide
       
   800   // and be sign-extended. Check the range.
       
   801 
       
   802   static void assert_signed_range(intptr_t x, int nbits) {
       
   803     assert( nbits == 32
       
   804         ||  -(1 << nbits-1) <= x  &&  x < ( 1 << nbits-1),
       
   805       "value out of range");
       
   806   }
       
   807 
       
   808   static void assert_signed_word_disp_range(intptr_t x, int nbits) {
       
   809     assert( (x & 3) == 0, "not word aligned");
       
   810     assert_signed_range(x, nbits + 2);
       
   811   }
       
   812 
       
   813   static void assert_unsigned_const(int x, int nbits) {
       
   814     assert( juint(x)  <  juint(1 << nbits), "unsigned constant out of range");
       
   815   }
       
   816 
       
   817   // fields: note bits numbered from LSB = 0,
       
   818   //  fields known by inclusive bit range
       
   819 
       
   820   static int fmask(juint hi_bit, juint lo_bit) {
       
   821     assert( hi_bit >= lo_bit  &&  0 <= lo_bit  &&  hi_bit < 32, "bad bits");
       
   822     return (1 << ( hi_bit-lo_bit + 1 )) - 1;
       
   823   }
       
   824 
       
   825   // inverse of u_field
       
   826 
       
   827   static int inv_u_field(int x, int hi_bit, int lo_bit) {
       
   828     juint r = juint(x) >> lo_bit;
       
   829     r &= fmask( hi_bit, lo_bit);
       
   830     return int(r);
       
   831   }
       
   832 
       
   833 
       
   834   // signed version: extract from field and sign-extend
       
   835 
       
   836   static int inv_s_field(int x, int hi_bit, int lo_bit) {
       
   837     int sign_shift = 31 - hi_bit;
       
   838     return inv_u_field( ((x << sign_shift) >> sign_shift), hi_bit, lo_bit);
       
   839   }
       
   840 
       
   841   // given a field that ranges from hi_bit to lo_bit (inclusive,
       
   842   // LSB = 0), and an unsigned value for the field,
       
   843   // shift it into the field
       
   844 
       
   845 #ifdef ASSERT
       
   846   static int u_field(int x, int hi_bit, int lo_bit) {
       
   847     assert( ( x & ~fmask(hi_bit, lo_bit))  == 0,
       
   848             "value out of range");
       
   849     int r = x << lo_bit;
       
   850     assert( inv_u_field(r, hi_bit, lo_bit) == x, "just checking");
       
   851     return r;
       
   852   }
       
   853 #else
       
   854   // make sure this is inlined as it will reduce code size significantly
       
   855   #define u_field(x, hi_bit, lo_bit)   ((x) << (lo_bit))
       
   856 #endif
       
   857 
       
   858   static int inv_op(  int x ) { return inv_u_field(x, 31, 30); }
       
   859   static int inv_op2( int x ) { return inv_u_field(x, 24, 22); }
       
   860   static int inv_op3( int x ) { return inv_u_field(x, 24, 19); }
       
   861   static int inv_cond( int x ){ return inv_u_field(x, 28, 25); }
       
   862 
       
   863   static bool inv_immed( int x ) { return (x & Assembler::immed(true)) != 0; }
       
   864 
       
   865   static Register inv_rd(  int x ) { return as_Register(inv_u_field(x, 29, 25)); }
       
   866   static Register inv_rs1( int x ) { return as_Register(inv_u_field(x, 18, 14)); }
       
   867   static Register inv_rs2( int x ) { return as_Register(inv_u_field(x,  4,  0)); }
       
   868 
       
   869   static int op(       int         x)  { return  u_field(x,             31, 30); }
       
   870   static int rd(       Register    r)  { return  u_field(r->encoding(), 29, 25); }
       
   871   static int fcn(      int         x)  { return  u_field(x,             29, 25); }
       
   872   static int op3(      int         x)  { return  u_field(x,             24, 19); }
       
   873   static int rs1(      Register    r)  { return  u_field(r->encoding(), 18, 14); }
       
   874   static int rs2(      Register    r)  { return  u_field(r->encoding(),  4,  0); }
       
   875   static int annul(    bool        a)  { return  u_field(a ? 1 : 0,     29, 29); }
       
   876   static int cond(     int         x)  { return  u_field(x,             28, 25); }
       
   877   static int cond_mov( int         x)  { return  u_field(x,             17, 14); }
       
   878   static int rcond(    RCondition  x)  { return  u_field(x,             12, 10); }
       
   879   static int op2(      int         x)  { return  u_field(x,             24, 22); }
       
   880   static int predict(  bool        p)  { return  u_field(p ? 1 : 0,     19, 19); }
       
   881   static int branchcc( CC       fcca)  { return  u_field(fcca,          21, 20); }
       
   882   static int cmpcc(    CC       fcca)  { return  u_field(fcca,          26, 25); }
       
   883   static int imm_asi(  int         x)  { return  u_field(x,             12,  5); }
       
   884   static int immed(    bool        i)  { return  u_field(i ? 1 : 0,     13, 13); }
       
   885   static int opf_low6( int         w)  { return  u_field(w,             10,  5); }
       
   886   static int opf_low5( int         w)  { return  u_field(w,              9,  5); }
       
   887   static int trapcc(   CC         cc)  { return  u_field(cc,            12, 11); }
       
   888   static int sx(       int         i)  { return  u_field(i,             12, 12); } // shift x=1 means 64-bit
       
   889   static int opf(      int         x)  { return  u_field(x,             13,  5); }
       
   890 
       
   891   static int opf_cc(   CC          c, bool useFloat ) { return u_field((useFloat ? 0 : 4) + c, 13, 11); }
       
   892   static int mov_cc(   CC          c, bool useFloat ) { return u_field(useFloat ? 0 : 1,  18, 18) | u_field(c, 12, 11); }
       
   893 
       
   894   static int fd( FloatRegister r,  FloatRegisterImpl::Width fwa) { return u_field(r->encoding(fwa), 29, 25); };
       
   895   static int fs1(FloatRegister r,  FloatRegisterImpl::Width fwa) { return u_field(r->encoding(fwa), 18, 14); };
       
   896   static int fs2(FloatRegister r,  FloatRegisterImpl::Width fwa) { return u_field(r->encoding(fwa),  4,  0); };
       
   897 
       
   898   // some float instructions use this encoding on the op3 field
       
   899   static int alt_op3(int op, FloatRegisterImpl::Width w) {
       
   900     int r;
       
   901     switch(w) {
       
   902      case FloatRegisterImpl::S: r = op + 0;  break;
       
   903      case FloatRegisterImpl::D: r = op + 3;  break;
       
   904      case FloatRegisterImpl::Q: r = op + 2;  break;
       
   905      default: ShouldNotReachHere(); break;
       
   906     }
       
   907     return op3(r);
       
   908   }
       
   909 
       
   910 
       
   911   // compute inverse of simm
       
   912   static int inv_simm(int x, int nbits) {
       
   913     return (int)(x << (32 - nbits)) >> (32 - nbits);
       
   914   }
       
   915 
       
   916   static int inv_simm13( int x ) { return inv_simm(x, 13); }
       
   917 
       
   918   // signed immediate, in low bits, nbits long
       
   919   static int simm(int x, int nbits) {
       
   920     assert_signed_range(x, nbits);
       
   921     return x  &  (( 1 << nbits ) - 1);
       
   922   }
       
   923 
       
   924   // compute inverse of wdisp16
       
   925   static intptr_t inv_wdisp16(int x, intptr_t pos) {
       
   926     int lo = x & (( 1 << 14 ) - 1);
       
   927     int hi = (x >> 20) & 3;
       
   928     if (hi >= 2) hi |= ~1;
       
   929     return (((hi << 14) | lo) << 2) + pos;
       
   930   }
       
   931 
       
   932   // word offset, 14 bits at LSend, 2 bits at B21, B20
       
   933   static int wdisp16(intptr_t x, intptr_t off) {
       
   934     intptr_t xx = x - off;
       
   935     assert_signed_word_disp_range(xx, 16);
       
   936     int r =  (xx >> 2) & ((1 << 14) - 1)
       
   937            |  (  ( (xx>>(2+14)) & 3 )  <<  20 );
       
   938     assert( inv_wdisp16(r, off) == x,  "inverse is not inverse");
       
   939     return r;
       
   940   }
       
   941 
       
   942 
       
   943   // word displacement in low-order nbits bits
       
   944 
       
   945   static intptr_t inv_wdisp( int x, intptr_t pos, int nbits ) {
       
   946     int pre_sign_extend = x & (( 1 << nbits ) - 1);
       
   947     int r =  pre_sign_extend >= ( 1 << (nbits-1) )
       
   948        ?   pre_sign_extend | ~(( 1 << nbits ) - 1)
       
   949        :   pre_sign_extend;
       
   950     return (r << 2) + pos;
       
   951   }
       
   952 
       
   953   static int wdisp( intptr_t x, intptr_t off, int nbits ) {
       
   954     intptr_t xx = x - off;
       
   955     assert_signed_word_disp_range(xx, nbits);
       
   956     int r =  (xx >> 2) & (( 1 << nbits ) - 1);
       
   957     assert( inv_wdisp( r, off, nbits )  ==  x, "inverse not inverse");
       
   958     return r;
       
   959   }
       
   960 
       
   961 
       
   962   // Extract the top 32 bits in a 64 bit word
       
   963   static int32_t hi32( int64_t x ) {
       
   964     int32_t r = int32_t( (uint64_t)x >> 32 );
       
   965     return r;
       
   966   }
       
   967 
       
   968   // given a sethi instruction, extract the constant, left-justified
       
   969   static int inv_hi22( int x ) {
       
   970     return x << 10;
       
   971   }
       
   972 
       
   973   // create an imm22 field, given a 32-bit left-justified constant
       
   974   static int hi22( int x ) {
       
   975     int r = int( juint(x) >> 10 );
       
   976     assert( (r & ~((1 << 22) - 1))  ==  0, "just checkin'");
       
   977     return r;
       
   978   }
       
   979 
       
   980   // create a low10 __value__ (not a field) for a given a 32-bit constant
       
   981   static int low10( int x ) {
       
   982     return x & ((1 << 10) - 1);
       
   983   }
       
   984 
       
   985   // instruction only in v9
       
   986   static void v9_only() { assert( VM_Version::v9_instructions_work(), "This instruction only works on SPARC V9"); }
       
   987 
       
   988   // instruction only in v8
       
   989   static void v8_only() { assert( VM_Version::v8_instructions_work(), "This instruction only works on SPARC V8"); }
       
   990 
       
   991   // instruction deprecated in v9
       
   992   static void v9_dep()  { } // do nothing for now
       
   993 
       
   994   // some float instructions only exist for single prec. on v8
       
   995   static void v8_s_only(FloatRegisterImpl::Width w)  { if (w != FloatRegisterImpl::S)  v9_only(); }
       
   996 
       
   997   // v8 has no CC field
       
   998   static void v8_no_cc(CC cc)  { if (cc)  v9_only(); }
       
   999 
       
  1000  protected:
       
  1001   // Simple delay-slot scheme:
       
  1002   // In order to check the programmer, the assembler keeps track of deley slots.
       
  1003   // It forbids CTIs in delay slots (conservative, but should be OK).
       
  1004   // Also, when putting an instruction into a delay slot, you must say
       
  1005   // asm->delayed()->add(...), in order to check that you don't omit
       
  1006   // delay-slot instructions.
       
  1007   // To implement this, we use a simple FSA
       
  1008 
       
  1009 #ifdef ASSERT
       
  1010   #define CHECK_DELAY
       
  1011 #endif
       
  1012 #ifdef CHECK_DELAY
       
  1013   enum Delay_state { no_delay, at_delay_slot, filling_delay_slot } delay_state;
       
  1014 #endif
       
  1015 
       
  1016  public:
       
  1017   // Tells assembler next instruction must NOT be in delay slot.
       
  1018   // Use at start of multinstruction macros.
       
  1019   void assert_not_delayed() {
       
  1020     // This is a separate overloading to avoid creation of string constants
       
  1021     // in non-asserted code--with some compilers this pollutes the object code.
       
  1022 #ifdef CHECK_DELAY
       
  1023     assert_not_delayed("next instruction should not be a delay slot");
       
  1024 #endif
       
  1025   }
       
  1026   void assert_not_delayed(const char* msg) {
       
  1027 #ifdef CHECK_DELAY
       
  1028     assert_msg ( delay_state == no_delay, msg);
       
  1029 #endif
       
  1030   }
       
  1031 
       
  1032  protected:
       
  1033   // Delay slot helpers
       
  1034   // cti is called when emitting control-transfer instruction,
       
  1035   // BEFORE doing the emitting.
       
  1036   // Only effective when assertion-checking is enabled.
       
  1037   void cti() {
       
  1038 #ifdef CHECK_DELAY
       
  1039     assert_not_delayed("cti should not be in delay slot");
       
  1040 #endif
       
  1041   }
       
  1042 
       
  1043   // called when emitting cti with a delay slot, AFTER emitting
       
  1044   void has_delay_slot() {
       
  1045 #ifdef CHECK_DELAY
       
  1046     assert_not_delayed("just checking");
       
  1047     delay_state = at_delay_slot;
       
  1048 #endif
       
  1049   }
       
  1050 
       
  1051 public:
       
  1052   // Tells assembler you know that next instruction is delayed
       
  1053   Assembler* delayed() {
       
  1054 #ifdef CHECK_DELAY
       
  1055     assert ( delay_state == at_delay_slot, "delayed instruction is not in delay slot");
       
  1056     delay_state = filling_delay_slot;
       
  1057 #endif
       
  1058     return this;
       
  1059   }
       
  1060 
       
  1061   void flush() {
       
  1062 #ifdef CHECK_DELAY
       
  1063     assert ( delay_state == no_delay, "ending code with a delay slot");
       
  1064 #endif
       
  1065     AbstractAssembler::flush();
       
  1066   }
       
  1067 
       
  1068   inline void emit_long(int);  // shadows AbstractAssembler::emit_long
       
  1069   inline void emit_data(int x) { emit_long(x); }
       
  1070   inline void emit_data(int, RelocationHolder const&);
       
  1071   inline void emit_data(int, relocInfo::relocType rtype);
       
  1072   // helper for above fcns
       
  1073   inline void check_delay();
       
  1074 
       
  1075 
       
  1076  public:
       
  1077   // instructions, refer to page numbers in the SPARC Architecture Manual, V9
       
  1078 
       
  1079   // pp 135 (addc was addx in v8)
       
  1080 
       
  1081   inline void add(    Register s1, Register s2, Register d );
       
  1082   inline void add(    Register s1, int simm13a, Register d, relocInfo::relocType rtype = relocInfo::none);
       
  1083   inline void add(    Register s1, int simm13a, Register d, RelocationHolder const& rspec);
       
  1084   inline void add(    const Address&  a,              Register d, int offset = 0);
       
  1085 
       
  1086   void addcc(  Register s1, Register s2, Register d ) { emit_long( op(arith_op) | rd(d) | op3(add_op3  | cc_bit_op3) | rs1(s1) | rs2(s2) ); }
       
  1087   void addcc(  Register s1, int simm13a, Register d ) { emit_long( op(arith_op) | rd(d) | op3(add_op3  | cc_bit_op3) | rs1(s1) | immed(true) | simm(simm13a, 13) ); }
       
  1088   void addc(   Register s1, Register s2, Register d ) { emit_long( op(arith_op) | rd(d) | op3(addc_op3             ) | rs1(s1) | rs2(s2) ); }
       
  1089   void addc(   Register s1, int simm13a, Register d ) { emit_long( op(arith_op) | rd(d) | op3(addc_op3             ) | rs1(s1) | immed(true) | simm(simm13a, 13) ); }
       
  1090   void addccc( Register s1, Register s2, Register d ) { emit_long( op(arith_op) | rd(d) | op3(addc_op3 | cc_bit_op3) | rs1(s1) | rs2(s2) ); }
       
  1091   void addccc( Register s1, int simm13a, Register d ) { emit_long( op(arith_op) | rd(d) | op3(addc_op3 | cc_bit_op3) | rs1(s1) | immed(true) | simm(simm13a, 13) ); }
       
  1092 
       
  1093   // pp 136
       
  1094 
       
  1095   inline void bpr( RCondition c, bool a, Predict p, Register s1, address d, relocInfo::relocType rt = relocInfo::none );
       
  1096   inline void bpr( RCondition c, bool a, Predict p, Register s1, Label& L);
       
  1097 
       
  1098  protected: // use MacroAssembler::br instead
       
  1099 
       
  1100   // pp 138
       
  1101 
       
  1102   inline void fb( Condition c, bool a, address d, relocInfo::relocType rt = relocInfo::none );
       
  1103   inline void fb( Condition c, bool a, Label& L );
       
  1104 
       
  1105   // pp 141
       
  1106 
       
  1107   inline void fbp( Condition c, bool a, CC cc, Predict p, address d, relocInfo::relocType rt = relocInfo::none );
       
  1108   inline void fbp( Condition c, bool a, CC cc, Predict p, Label& L );
       
  1109 
       
  1110  public:
       
  1111 
       
  1112   // pp 144
       
  1113 
       
  1114   inline void br( Condition c, bool a, address d, relocInfo::relocType rt = relocInfo::none );
       
  1115   inline void br( Condition c, bool a, Label& L );
       
  1116 
       
  1117   // pp 146
       
  1118 
       
  1119   inline void bp( Condition c, bool a, CC cc, Predict p, address d, relocInfo::relocType rt = relocInfo::none );
       
  1120   inline void bp( Condition c, bool a, CC cc, Predict p, Label& L );
       
  1121 
       
  1122   // pp 121 (V8)
       
  1123 
       
  1124   inline void cb( Condition c, bool a, address d, relocInfo::relocType rt = relocInfo::none );
       
  1125   inline void cb( Condition c, bool a, Label& L );
       
  1126 
       
  1127   // pp 149
       
  1128 
       
  1129   inline void call( address d,  relocInfo::relocType rt = relocInfo::runtime_call_type );
       
  1130   inline void call( Label& L,   relocInfo::relocType rt = relocInfo::runtime_call_type );
       
  1131 
       
  1132   // pp 150
       
  1133 
       
  1134   // These instructions compare the contents of s2 with the contents of
       
  1135   // memory at address in s1. If the values are equal, the contents of memory
       
  1136   // at address s1 is swapped with the data in d. If the values are not equal,
       
  1137   // the the contents of memory at s1 is loaded into d, without the swap.
       
  1138 
       
  1139   void casa(  Register s1, Register s2, Register d, int ia = -1 ) { v9_only();  emit_long( op(ldst_op) | rd(d) | op3(casa_op3 ) | rs1(s1) | (ia == -1  ? immed(true) : imm_asi(ia)) | rs2(s2)); }
       
  1140   void casxa( Register s1, Register s2, Register d, int ia = -1 ) { v9_only();  emit_long( op(ldst_op) | rd(d) | op3(casxa_op3) | rs1(s1) | (ia == -1  ? immed(true) : imm_asi(ia)) | rs2(s2)); }
       
  1141 
       
  1142   // pp 152
       
  1143 
       
  1144   void udiv(   Register s1, Register s2, Register d ) { emit_long( op(arith_op) | rd(d) | op3(udiv_op3             ) | rs1(s1) | rs2(s2)); }
       
  1145   void udiv(   Register s1, int simm13a, Register d ) { emit_long( op(arith_op) | rd(d) | op3(udiv_op3             ) | rs1(s1) | immed(true) | simm(simm13a, 13) ); }
       
  1146   void sdiv(   Register s1, Register s2, Register d ) { emit_long( op(arith_op) | rd(d) | op3(sdiv_op3             ) | rs1(s1) | rs2(s2)); }
       
  1147   void sdiv(   Register s1, int simm13a, Register d ) { emit_long( op(arith_op) | rd(d) | op3(sdiv_op3             ) | rs1(s1) | immed(true) | simm(simm13a, 13) ); }
       
  1148   void udivcc( Register s1, Register s2, Register d ) { emit_long( op(arith_op) | rd(d) | op3(udiv_op3 | cc_bit_op3) | rs1(s1) | rs2(s2)); }
       
  1149   void udivcc( Register s1, int simm13a, Register d ) { emit_long( op(arith_op) | rd(d) | op3(udiv_op3 | cc_bit_op3) | rs1(s1) | immed(true) | simm(simm13a, 13) ); }
       
  1150   void sdivcc( Register s1, Register s2, Register d ) { emit_long( op(arith_op) | rd(d) | op3(sdiv_op3 | cc_bit_op3) | rs1(s1) | rs2(s2)); }
       
  1151   void sdivcc( Register s1, int simm13a, Register d ) { emit_long( op(arith_op) | rd(d) | op3(sdiv_op3 | cc_bit_op3) | rs1(s1) | immed(true) | simm(simm13a, 13) ); }
       
  1152 
       
  1153   // pp 155
       
  1154 
       
  1155   void done()  { v9_only();  cti();  emit_long( op(arith_op) | fcn(0) | op3(done_op3) ); }
       
  1156   void retry() { v9_only();  cti();  emit_long( op(arith_op) | fcn(1) | op3(retry_op3) ); }
       
  1157 
       
  1158   // pp 156
       
  1159 
       
  1160   void fadd( FloatRegisterImpl::Width w, FloatRegister s1, FloatRegister s2, FloatRegister d ) { emit_long( op(arith_op) | fd(d, w) | op3(fpop1_op3) | fs1(s1, w) | opf(0x40 + w) | fs2(s2, w)); }
       
  1161   void fsub( FloatRegisterImpl::Width w, FloatRegister s1, FloatRegister s2, FloatRegister d ) { emit_long( op(arith_op) | fd(d, w) | op3(fpop1_op3) | fs1(s1, w) | opf(0x44 + w) | fs2(s2, w)); }
       
  1162 
       
  1163   // pp 157
       
  1164 
       
  1165   void fcmp(  FloatRegisterImpl::Width w, CC cc, FloatRegister s1, FloatRegister s2) { v8_no_cc(cc);  emit_long( op(arith_op) | cmpcc(cc) | op3(fpop2_op3) | fs1(s1, w) | opf(0x50 + w) | fs2(s2, w)); }
       
  1166   void fcmpe( FloatRegisterImpl::Width w, CC cc, FloatRegister s1, FloatRegister s2) { v8_no_cc(cc);  emit_long( op(arith_op) | cmpcc(cc) | op3(fpop2_op3) | fs1(s1, w) | opf(0x54 + w) | fs2(s2, w)); }
       
  1167 
       
  1168   // pp 159
       
  1169 
       
  1170   void ftox( FloatRegisterImpl::Width w, FloatRegister s, FloatRegister d ) { v9_only();  emit_long( op(arith_op) | fd(d, w) | op3(fpop1_op3) | opf(0x80 + w) | fs2(s, w)); }
       
  1171   void ftoi( FloatRegisterImpl::Width w, FloatRegister s, FloatRegister d ) {             emit_long( op(arith_op) | fd(d, w) | op3(fpop1_op3) | opf(0xd0 + w) | fs2(s, w)); }
       
  1172 
       
  1173   // pp 160
       
  1174 
       
  1175   void ftof( FloatRegisterImpl::Width sw, FloatRegisterImpl::Width dw, FloatRegister s, FloatRegister d ) { emit_long( op(arith_op) | fd(d, dw) | op3(fpop1_op3) | opf(0xc0 + sw + dw*4) | fs2(s, sw)); }
       
  1176 
       
  1177   // pp 161
       
  1178 
       
  1179   void fxtof( FloatRegisterImpl::Width w, FloatRegister s, FloatRegister d ) { v9_only();  emit_long( op(arith_op) | fd(d, w) | op3(fpop1_op3) | opf(0x80 + w*4) | fs2(s, w)); }
       
  1180   void fitof( FloatRegisterImpl::Width w, FloatRegister s, FloatRegister d ) {             emit_long( op(arith_op) | fd(d, w) | op3(fpop1_op3) | opf(0xc0 + w*4) | fs2(s, w)); }
       
  1181 
       
  1182   // pp 162
       
  1183 
       
  1184   void fmov( FloatRegisterImpl::Width w, FloatRegister s, FloatRegister d ) { v8_s_only(w);  emit_long( op(arith_op) | fd(d, w) | op3(fpop1_op3) | opf(0x00 + w) | fs2(s, w)); }
       
  1185 
       
  1186   void fneg( FloatRegisterImpl::Width w, FloatRegister s, FloatRegister d ) { v8_s_only(w);  emit_long( op(arith_op) | fd(d, w) | op3(fpop1_op3) | opf(0x04 + w) | fs2(s, w)); }
       
  1187 
       
  1188   // page 144 sparc v8 architecture (double prec works on v8 if the source and destination registers are the same). fnegs is the only instruction available
       
  1189   // on v8 to do negation of single, double and quad precision floats.
       
  1190 
       
  1191   void fneg( FloatRegisterImpl::Width w, FloatRegister sd ) { if (VM_Version::v9_instructions_work()) emit_long( op(arith_op) | fd(sd, w) | op3(fpop1_op3) | opf(0x04 + w) | fs2(sd, w)); else emit_long( op(arith_op) | fd(sd, w) | op3(fpop1_op3) |  opf(0x05) | fs2(sd, w)); }
       
  1192 
       
  1193   void fabs( FloatRegisterImpl::Width w, FloatRegister s, FloatRegister d ) { v8_s_only(w);  emit_long( op(arith_op) | fd(d, w) | op3(fpop1_op3) | opf(0x08 + w) | fs2(s, w)); }
       
  1194 
       
  1195   // page 144 sparc v8 architecture (double prec works on v8 if the source and destination registers are the same). fabss is the only instruction available
       
  1196   // on v8 to do abs operation on single/double/quad precision floats.
       
  1197 
       
  1198   void fabs( FloatRegisterImpl::Width w, FloatRegister sd ) { if (VM_Version::v9_instructions_work()) emit_long( op(arith_op) | fd(sd, w) | op3(fpop1_op3) | opf(0x08 + w) | fs2(sd, w)); else emit_long( op(arith_op) | fd(sd, w) | op3(fpop1_op3) | opf(0x09) | fs2(sd, w)); }
       
  1199 
       
  1200   // pp 163
       
  1201 
       
  1202   void fmul( FloatRegisterImpl::Width w,                            FloatRegister s1, FloatRegister s2, FloatRegister d ) { emit_long( op(arith_op) | fd(d, w)  | op3(fpop1_op3) | fs1(s1, w)  | opf(0x48 + w)         | fs2(s2, w)); }
       
  1203   void fmul( FloatRegisterImpl::Width sw, FloatRegisterImpl::Width dw,  FloatRegister s1, FloatRegister s2, FloatRegister d ) { emit_long( op(arith_op) | fd(d, dw) | op3(fpop1_op3) | fs1(s1, sw) | opf(0x60 + sw + dw*4) | fs2(s2, sw)); }
       
  1204   void fdiv( FloatRegisterImpl::Width w,                            FloatRegister s1, FloatRegister s2, FloatRegister d ) { emit_long( op(arith_op) | fd(d, w)  | op3(fpop1_op3) | fs1(s1, w)  | opf(0x4c + w)         | fs2(s2, w)); }
       
  1205 
       
  1206   // pp 164
       
  1207 
       
  1208   void fsqrt( FloatRegisterImpl::Width w, FloatRegister s, FloatRegister d ) { emit_long( op(arith_op) | fd(d, w) | op3(fpop1_op3) | opf(0x28 + w) | fs2(s, w)); }
       
  1209 
       
  1210   // pp 165
       
  1211 
       
  1212   inline void flush( Register s1, Register s2 );
       
  1213   inline void flush( Register s1, int simm13a);
       
  1214 
       
  1215   // pp 167
       
  1216 
       
  1217   void flushw() { v9_only();  emit_long( op(arith_op) | op3(flushw_op3) ); }
       
  1218 
       
  1219   // pp 168
       
  1220 
       
  1221   void illtrap( int const22a) { if (const22a != 0) v9_only();  emit_long( op(branch_op) | u_field(const22a, 21, 0) ); }
       
  1222   // v8 unimp == illtrap(0)
       
  1223 
       
  1224   // pp 169
       
  1225 
       
  1226   void impdep1( int id1, int const19a ) { v9_only();  emit_long( op(arith_op) | fcn(id1) | op3(impdep1_op3) | u_field(const19a, 18, 0)); }
       
  1227   void impdep2( int id1, int const19a ) { v9_only();  emit_long( op(arith_op) | fcn(id1) | op3(impdep2_op3) | u_field(const19a, 18, 0)); }
       
  1228 
       
  1229   // pp 149 (v8)
       
  1230 
       
  1231   void cpop1( int opc, int cr1, int cr2, int crd ) { v8_only();  emit_long( op(arith_op) | fcn(crd) | op3(impdep1_op3) | u_field(cr1, 18, 14) | opf(opc) | u_field(cr2, 4, 0)); }
       
  1232   void cpop2( int opc, int cr1, int cr2, int crd ) { v8_only();  emit_long( op(arith_op) | fcn(crd) | op3(impdep2_op3) | u_field(cr1, 18, 14) | opf(opc) | u_field(cr2, 4, 0)); }
       
  1233 
       
  1234   // pp 170
       
  1235 
       
  1236   void jmpl( Register s1, Register s2, Register d );
       
  1237   void jmpl( Register s1, int simm13a, Register d, RelocationHolder const& rspec = RelocationHolder() );
       
  1238 
       
  1239   inline void jmpl( Address& a, Register d, int offset = 0);
       
  1240 
       
  1241   // 171
       
  1242 
       
  1243   inline void ldf(    FloatRegisterImpl::Width w, Register s1, Register s2, FloatRegister d );
       
  1244   inline void ldf(    FloatRegisterImpl::Width w, Register s1, int simm13a, FloatRegister d );
       
  1245 
       
  1246   inline void ldf(    FloatRegisterImpl::Width w, const Address& a, FloatRegister d, int offset = 0);
       
  1247 
       
  1248 
       
  1249   inline void ldfsr(  Register s1, Register s2 );
       
  1250   inline void ldfsr(  Register s1, int simm13a);
       
  1251   inline void ldxfsr( Register s1, Register s2 );
       
  1252   inline void ldxfsr( Register s1, int simm13a);
       
  1253 
       
  1254   // pp 94 (v8)
       
  1255 
       
  1256   inline void ldc(   Register s1, Register s2, int crd );
       
  1257   inline void ldc(   Register s1, int simm13a, int crd);
       
  1258   inline void lddc(  Register s1, Register s2, int crd );
       
  1259   inline void lddc(  Register s1, int simm13a, int crd);
       
  1260   inline void ldcsr( Register s1, Register s2, int crd );
       
  1261   inline void ldcsr( Register s1, int simm13a, int crd);
       
  1262 
       
  1263 
       
  1264   // 173
       
  1265 
       
  1266   void ldfa(  FloatRegisterImpl::Width w, Register s1, Register s2, int ia, FloatRegister d ) { v9_only();  emit_long( op(ldst_op) | fd(d, w) | alt_op3(ldf_op3 | alt_bit_op3, w) | rs1(s1) | imm_asi(ia) | rs2(s2) ); }
       
  1267   void ldfa(  FloatRegisterImpl::Width w, Register s1, int simm13a,         FloatRegister d ) { v9_only();  emit_long( op(ldst_op) | fd(d, w) | alt_op3(ldf_op3 | alt_bit_op3, w) | rs1(s1) | immed(true) | simm(simm13a, 13) ); }
       
  1268 
       
  1269   // pp 175, lduw is ld on v8
       
  1270 
       
  1271   inline void ldsb(  Register s1, Register s2, Register d );
       
  1272   inline void ldsb(  Register s1, int simm13a, Register d);
       
  1273   inline void ldsh(  Register s1, Register s2, Register d );
       
  1274   inline void ldsh(  Register s1, int simm13a, Register d);
       
  1275   inline void ldsw(  Register s1, Register s2, Register d );
       
  1276   inline void ldsw(  Register s1, int simm13a, Register d);
       
  1277   inline void ldub(  Register s1, Register s2, Register d );
       
  1278   inline void ldub(  Register s1, int simm13a, Register d);
       
  1279   inline void lduh(  Register s1, Register s2, Register d );
       
  1280   inline void lduh(  Register s1, int simm13a, Register d);
       
  1281   inline void lduw(  Register s1, Register s2, Register d );
       
  1282   inline void lduw(  Register s1, int simm13a, Register d);
       
  1283   inline void ldx(   Register s1, Register s2, Register d );
       
  1284   inline void ldx(   Register s1, int simm13a, Register d);
       
  1285   inline void ld(    Register s1, Register s2, Register d );
       
  1286   inline void ld(    Register s1, int simm13a, Register d);
       
  1287   inline void ldd(   Register s1, Register s2, Register d );
       
  1288   inline void ldd(   Register s1, int simm13a, Register d);
       
  1289 
       
  1290   inline void ldsb( const Address& a, Register d, int offset = 0 );
       
  1291   inline void ldsh( const Address& a, Register d, int offset = 0 );
       
  1292   inline void ldsw( const Address& a, Register d, int offset = 0 );
       
  1293   inline void ldub( const Address& a, Register d, int offset = 0 );
       
  1294   inline void lduh( const Address& a, Register d, int offset = 0 );
       
  1295   inline void lduw( const Address& a, Register d, int offset = 0 );
       
  1296   inline void ldx(  const Address& a, Register d, int offset = 0 );
       
  1297   inline void ld(   const Address& a, Register d, int offset = 0 );
       
  1298   inline void ldd(  const Address& a, Register d, int offset = 0 );
       
  1299 
       
  1300   // pp 177
       
  1301 
       
  1302   void ldsba(  Register s1, Register s2, int ia, Register d ) {             emit_long( op(ldst_op) | rd(d) | op3(ldsb_op3 | alt_bit_op3) | rs1(s1) | imm_asi(ia) | rs2(s2) ); }
       
  1303   void ldsba(  Register s1, int simm13a,         Register d ) {             emit_long( op(ldst_op) | rd(d) | op3(ldsb_op3 | alt_bit_op3) | rs1(s1) | immed(true) | simm(simm13a, 13) ); }
       
  1304   void ldsha(  Register s1, Register s2, int ia, Register d ) {             emit_long( op(ldst_op) | rd(d) | op3(ldsh_op3 | alt_bit_op3) | rs1(s1) | imm_asi(ia) | rs2(s2) ); }
       
  1305   void ldsha(  Register s1, int simm13a,         Register d ) {             emit_long( op(ldst_op) | rd(d) | op3(ldsh_op3 | alt_bit_op3) | rs1(s1) | immed(true) | simm(simm13a, 13) ); }
       
  1306   void ldswa(  Register s1, Register s2, int ia, Register d ) { v9_only();  emit_long( op(ldst_op) | rd(d) | op3(ldsw_op3 | alt_bit_op3) | rs1(s1) | imm_asi(ia) | rs2(s2) ); }
       
  1307   void ldswa(  Register s1, int simm13a,         Register d ) { v9_only();  emit_long( op(ldst_op) | rd(d) | op3(ldsw_op3 | alt_bit_op3) | rs1(s1) | immed(true) | simm(simm13a, 13) ); }
       
  1308   void lduba(  Register s1, Register s2, int ia, Register d ) {             emit_long( op(ldst_op) | rd(d) | op3(ldub_op3 | alt_bit_op3) | rs1(s1) | imm_asi(ia) | rs2(s2) ); }
       
  1309   void lduba(  Register s1, int simm13a,         Register d ) {             emit_long( op(ldst_op) | rd(d) | op3(ldub_op3 | alt_bit_op3) | rs1(s1) | immed(true) | simm(simm13a, 13) ); }
       
  1310   void lduha(  Register s1, Register s2, int ia, Register d ) {             emit_long( op(ldst_op) | rd(d) | op3(lduh_op3 | alt_bit_op3) | rs1(s1) | imm_asi(ia) | rs2(s2) ); }
       
  1311   void lduha(  Register s1, int simm13a,         Register d ) {             emit_long( op(ldst_op) | rd(d) | op3(lduh_op3 | alt_bit_op3) | rs1(s1) | immed(true) | simm(simm13a, 13) ); }
       
  1312   void lduwa(  Register s1, Register s2, int ia, Register d ) {             emit_long( op(ldst_op) | rd(d) | op3(lduw_op3 | alt_bit_op3) | rs1(s1) | imm_asi(ia) | rs2(s2) ); }
       
  1313   void lduwa(  Register s1, int simm13a,         Register d ) {             emit_long( op(ldst_op) | rd(d) | op3(lduw_op3 | alt_bit_op3) | rs1(s1) | immed(true) | simm(simm13a, 13) ); }
       
  1314   void ldxa(   Register s1, Register s2, int ia, Register d ) { v9_only();  emit_long( op(ldst_op) | rd(d) | op3(ldx_op3  | alt_bit_op3) | rs1(s1) | imm_asi(ia) | rs2(s2) ); }
       
  1315   void ldxa(   Register s1, int simm13a,         Register d ) { v9_only();  emit_long( op(ldst_op) | rd(d) | op3(ldx_op3  | alt_bit_op3) | rs1(s1) | immed(true) | simm(simm13a, 13) ); }
       
  1316   void ldda(   Register s1, Register s2, int ia, Register d ) { v9_dep();   emit_long( op(ldst_op) | rd(d) | op3(ldd_op3  | alt_bit_op3) | rs1(s1) | imm_asi(ia) | rs2(s2) ); }
       
  1317   void ldda(   Register s1, int simm13a,         Register d ) { v9_dep();   emit_long( op(ldst_op) | rd(d) | op3(ldd_op3  | alt_bit_op3) | rs1(s1) | immed(true) | simm(simm13a, 13) ); }
       
  1318 
       
  1319   // pp 179
       
  1320 
       
  1321   inline void ldstub(  Register s1, Register s2, Register d );
       
  1322   inline void ldstub(  Register s1, int simm13a, Register d);
       
  1323 
       
  1324   // pp 180
       
  1325 
       
  1326   void ldstuba( Register s1, Register s2, int ia, Register d ) { emit_long( op(ldst_op) | rd(d) | op3(ldstub_op3 | alt_bit_op3) | rs1(s1) | imm_asi(ia) | rs2(s2) ); }
       
  1327   void ldstuba( Register s1, int simm13a,         Register d ) { emit_long( op(ldst_op) | rd(d) | op3(ldstub_op3 | alt_bit_op3) | rs1(s1) | immed(true) | simm(simm13a, 13) ); }
       
  1328 
       
  1329   // pp 181
       
  1330 
       
  1331   void and3(     Register s1, Register s2, Register d ) { emit_long( op(arith_op) | rd(d) | op3(and_op3               ) | rs1(s1) | rs2(s2) ); }
       
  1332   void and3(     Register s1, int simm13a, Register d ) { emit_long( op(arith_op) | rd(d) | op3(and_op3               ) | rs1(s1) | immed(true) | simm(simm13a, 13) ); }
       
  1333   void andcc(   Register s1, Register s2, Register d ) { emit_long( op(arith_op) | rd(d) | op3(and_op3  | cc_bit_op3) | rs1(s1) | rs2(s2) ); }
       
  1334   void andcc(   Register s1, int simm13a, Register d ) { emit_long( op(arith_op) | rd(d) | op3(and_op3  | cc_bit_op3) | rs1(s1) | immed(true) | simm(simm13a, 13) ); }
       
  1335   void andn(    Register s1, Register s2, Register d ) { emit_long( op(arith_op) | rd(d) | op3(andn_op3             ) | rs1(s1) | rs2(s2) ); }
       
  1336   void andn(    Register s1, int simm13a, Register d ) { emit_long( op(arith_op) | rd(d) | op3(andn_op3             ) | rs1(s1) | immed(true) | simm(simm13a, 13) ); }
       
  1337   void andncc(  Register s1, Register s2, Register d ) { emit_long( op(arith_op) | rd(d) | op3(andn_op3 | cc_bit_op3) | rs1(s1) | rs2(s2) ); }
       
  1338   void andncc(  Register s1, int simm13a, Register d ) { emit_long( op(arith_op) | rd(d) | op3(andn_op3 | cc_bit_op3) | rs1(s1) | immed(true) | simm(simm13a, 13) ); }
       
  1339   void or3(      Register s1, Register s2, Register d ) { emit_long( op(arith_op) | rd(d) | op3(or_op3               ) | rs1(s1) | rs2(s2) ); }
       
  1340   void or3(      Register s1, int simm13a, Register d ) { emit_long( op(arith_op) | rd(d) | op3(or_op3               ) | rs1(s1) | immed(true) | simm(simm13a, 13) ); }
       
  1341   void orcc(    Register s1, Register s2, Register d ) { emit_long( op(arith_op) | rd(d) | op3(or_op3   | cc_bit_op3) | rs1(s1) | rs2(s2) ); }
       
  1342   void orcc(    Register s1, int simm13a, Register d ) { emit_long( op(arith_op) | rd(d) | op3(or_op3   | cc_bit_op3) | rs1(s1) | immed(true) | simm(simm13a, 13) ); }
       
  1343   void orn(     Register s1, Register s2, Register d ) { emit_long( op(arith_op) | rd(d) | op3(orn_op3) | rs1(s1) | rs2(s2) ); }
       
  1344   void orn(     Register s1, int simm13a, Register d ) { emit_long( op(arith_op) | rd(d) | op3(orn_op3) | rs1(s1) | immed(true) | simm(simm13a, 13) ); }
       
  1345   void orncc(   Register s1, Register s2, Register d ) { emit_long( op(arith_op) | rd(d) | op3(orn_op3  | cc_bit_op3) | rs1(s1) | rs2(s2) ); }
       
  1346   void orncc(   Register s1, int simm13a, Register d ) { emit_long( op(arith_op) | rd(d) | op3(orn_op3  | cc_bit_op3) | rs1(s1) | immed(true) | simm(simm13a, 13) ); }
       
  1347   void xor3(     Register s1, Register s2, Register d ) { emit_long( op(arith_op) | rd(d) | op3(xor_op3              ) | rs1(s1) | rs2(s2) ); }
       
  1348   void xor3(     Register s1, int simm13a, Register d ) { emit_long( op(arith_op) | rd(d) | op3(xor_op3              ) | rs1(s1) | immed(true) | simm(simm13a, 13) ); }
       
  1349   void xorcc(   Register s1, Register s2, Register d ) { emit_long( op(arith_op) | rd(d) | op3(xor_op3  | cc_bit_op3) | rs1(s1) | rs2(s2) ); }
       
  1350   void xorcc(   Register s1, int simm13a, Register d ) { emit_long( op(arith_op) | rd(d) | op3(xor_op3  | cc_bit_op3) | rs1(s1) | immed(true) | simm(simm13a, 13) ); }
       
  1351   void xnor(    Register s1, Register s2, Register d ) { emit_long( op(arith_op) | rd(d) | op3(xnor_op3             ) | rs1(s1) | rs2(s2) ); }
       
  1352   void xnor(    Register s1, int simm13a, Register d ) { emit_long( op(arith_op) | rd(d) | op3(xnor_op3             ) | rs1(s1) | immed(true) | simm(simm13a, 13) ); }
       
  1353   void xnorcc(  Register s1, Register s2, Register d ) { emit_long( op(arith_op) | rd(d) | op3(xnor_op3 | cc_bit_op3) | rs1(s1) | rs2(s2) ); }
       
  1354   void xnorcc(  Register s1, int simm13a, Register d ) { emit_long( op(arith_op) | rd(d) | op3(xnor_op3 | cc_bit_op3) | rs1(s1) | immed(true) | simm(simm13a, 13) ); }
       
  1355 
       
  1356   // pp 183
       
  1357 
       
  1358   void membar( Membar_mask_bits const7a ) { v9_only(); emit_long( op(arith_op) | op3(membar_op3) | rs1(O7) | immed(true) | u_field( int(const7a), 6, 0)); }
       
  1359 
       
  1360   // pp 185
       
  1361 
       
  1362   void fmov( FloatRegisterImpl::Width w, Condition c,  bool floatCC, CC cca, FloatRegister s2, FloatRegister d ) { v9_only();  emit_long( op(arith_op) | fd(d, w) | op3(fpop2_op3) | cond_mov(c) | opf_cc(cca, floatCC) | opf_low6(w) | fs2(s2, w)); }
       
  1363 
       
  1364   // pp 189
       
  1365 
       
  1366   void fmov( FloatRegisterImpl::Width w, RCondition c, Register s1,  FloatRegister s2, FloatRegister d ) { v9_only();  emit_long( op(arith_op) | fd(d, w) | op3(fpop2_op3) | rs1(s1) | rcond(c) | opf_low5(4 + w) | fs2(s2, w)); }
       
  1367 
       
  1368   // pp 191
       
  1369 
       
  1370   void movcc( Condition c, bool floatCC, CC cca, Register s2, Register d ) { v9_only();  emit_long( op(arith_op) | rd(d) | op3(movcc_op3) | mov_cc(cca, floatCC) | cond_mov(c) | rs2(s2) ); }
       
  1371   void movcc( Condition c, bool floatCC, CC cca, int simm11a, Register d ) { v9_only();  emit_long( op(arith_op) | rd(d) | op3(movcc_op3) | mov_cc(cca, floatCC) | cond_mov(c) | immed(true) | simm(simm11a, 11) ); }
       
  1372 
       
  1373   // pp 195
       
  1374 
       
  1375   void movr( RCondition c, Register s1, Register s2,  Register d ) { v9_only();  emit_long( op(arith_op) | rd(d) | op3(movr_op3) | rs1(s1) | rcond(c) | rs2(s2) ); }
       
  1376   void movr( RCondition c, Register s1, int simm10a,  Register d ) { v9_only();  emit_long( op(arith_op) | rd(d) | op3(movr_op3) | rs1(s1) | rcond(c) | immed(true) | simm(simm10a, 10) ); }
       
  1377 
       
  1378   // pp 196
       
  1379 
       
  1380   void mulx(  Register s1, Register s2, Register d ) { v9_only(); emit_long( op(arith_op) | rd(d) | op3(mulx_op3 ) | rs1(s1) | rs2(s2) ); }
       
  1381   void mulx(  Register s1, int simm13a, Register d ) { v9_only(); emit_long( op(arith_op) | rd(d) | op3(mulx_op3 ) | rs1(s1) | immed(true) | simm(simm13a, 13) ); }
       
  1382   void sdivx( Register s1, Register s2, Register d ) { v9_only(); emit_long( op(arith_op) | rd(d) | op3(sdivx_op3) | rs1(s1) | rs2(s2) ); }
       
  1383   void sdivx( Register s1, int simm13a, Register d ) { v9_only(); emit_long( op(arith_op) | rd(d) | op3(sdivx_op3) | rs1(s1) | immed(true) | simm(simm13a, 13) ); }
       
  1384   void udivx( Register s1, Register s2, Register d ) { v9_only(); emit_long( op(arith_op) | rd(d) | op3(udivx_op3) | rs1(s1) | rs2(s2) ); }
       
  1385   void udivx( Register s1, int simm13a, Register d ) { v9_only(); emit_long( op(arith_op) | rd(d) | op3(udivx_op3) | rs1(s1) | immed(true) | simm(simm13a, 13) ); }
       
  1386 
       
  1387   // pp 197
       
  1388 
       
  1389   void umul(   Register s1, Register s2, Register d ) { emit_long( op(arith_op) | rd(d) | op3(umul_op3             ) | rs1(s1) | rs2(s2) ); }
       
  1390   void umul(   Register s1, int simm13a, Register d ) { emit_long( op(arith_op) | rd(d) | op3(umul_op3             ) | rs1(s1) | immed(true) | simm(simm13a, 13) ); }
       
  1391   void smul(   Register s1, Register s2, Register d ) { emit_long( op(arith_op) | rd(d) | op3(smul_op3             ) | rs1(s1) | rs2(s2) ); }
       
  1392   void smul(   Register s1, int simm13a, Register d ) { emit_long( op(arith_op) | rd(d) | op3(smul_op3             ) | rs1(s1) | immed(true) | simm(simm13a, 13) ); }
       
  1393   void umulcc( Register s1, Register s2, Register d ) { emit_long( op(arith_op) | rd(d) | op3(umul_op3 | cc_bit_op3) | rs1(s1) | rs2(s2) ); }
       
  1394   void umulcc( Register s1, int simm13a, Register d ) { emit_long( op(arith_op) | rd(d) | op3(umul_op3 | cc_bit_op3) | rs1(s1) | immed(true) | simm(simm13a, 13) ); }
       
  1395   void smulcc( Register s1, Register s2, Register d ) { emit_long( op(arith_op) | rd(d) | op3(smul_op3 | cc_bit_op3) | rs1(s1) | rs2(s2) ); }
       
  1396   void smulcc( Register s1, int simm13a, Register d ) { emit_long( op(arith_op) | rd(d) | op3(smul_op3 | cc_bit_op3) | rs1(s1) | immed(true) | simm(simm13a, 13) ); }
       
  1397 
       
  1398   // pp 199
       
  1399 
       
  1400   void mulscc(   Register s1, Register s2, Register d ) { v9_dep();  emit_long( op(arith_op) | rd(d) | op3(mulscc_op3) | rs1(s1) | rs2(s2) ); }
       
  1401   void mulscc(   Register s1, int simm13a, Register d ) { v9_dep();  emit_long( op(arith_op) | rd(d) | op3(mulscc_op3) | rs1(s1) | immed(true) | simm(simm13a, 13) ); }
       
  1402 
       
  1403   // pp 201
       
  1404 
       
  1405   void nop() { emit_long( op(branch_op) | op2(sethi_op2) ); }
       
  1406 
       
  1407 
       
  1408   // pp 202
       
  1409 
       
  1410   void popc( Register s,  Register d) { v9_only();  emit_long( op(arith_op) | rd(d) | op3(popc_op3) | rs2(s)); }
       
  1411   void popc( int simm13a, Register d) { v9_only();  emit_long( op(arith_op) | rd(d) | op3(popc_op3) | immed(true) | simm(simm13a, 13)); }
       
  1412 
       
  1413   // pp 203
       
  1414 
       
  1415   void prefetch(   Register s1, Register s2,         PrefetchFcn f);
       
  1416   void prefetch(   Register s1, int simm13a,         PrefetchFcn f);
       
  1417   void prefetcha(  Register s1, Register s2, int ia, PrefetchFcn f ) { v9_only();  emit_long( op(ldst_op) | fcn(f) | op3(prefetch_op3 | alt_bit_op3) | rs1(s1) | imm_asi(ia) | rs2(s2) ); }
       
  1418   void prefetcha(  Register s1, int simm13a,         PrefetchFcn f ) { v9_only();  emit_long( op(ldst_op) | fcn(f) | op3(prefetch_op3 | alt_bit_op3) | rs1(s1) | immed(true) | simm(simm13a, 13) ); }
       
  1419 
       
  1420   inline void prefetch(const Address& a, PrefetchFcn F, int offset = 0);
       
  1421 
       
  1422   // pp 208
       
  1423 
       
  1424   // not implementing read privileged register
       
  1425 
       
  1426   inline void rdy(    Register d) { v9_dep();  emit_long( op(arith_op) | rd(d) | op3(rdreg_op3) | u_field(0, 18, 14)); }
       
  1427   inline void rdccr(  Register d) { v9_only(); emit_long( op(arith_op) | rd(d) | op3(rdreg_op3) | u_field(2, 18, 14)); }
       
  1428   inline void rdasi(  Register d) { v9_only(); emit_long( op(arith_op) | rd(d) | op3(rdreg_op3) | u_field(3, 18, 14)); }
       
  1429   inline void rdtick( Register d) { v9_only(); emit_long( op(arith_op) | rd(d) | op3(rdreg_op3) | u_field(4, 18, 14)); } // Spoon!
       
  1430   inline void rdpc(   Register d) { v9_only(); emit_long( op(arith_op) | rd(d) | op3(rdreg_op3) | u_field(5, 18, 14)); }
       
  1431   inline void rdfprs( Register d) { v9_only(); emit_long( op(arith_op) | rd(d) | op3(rdreg_op3) | u_field(6, 18, 14)); }
       
  1432 
       
  1433   // pp 213
       
  1434 
       
  1435   inline void rett( Register s1, Register s2);
       
  1436   inline void rett( Register s1, int simm13a, relocInfo::relocType rt = relocInfo::none);
       
  1437 
       
  1438   // pp 214
       
  1439 
       
  1440   void save(    Register s1, Register s2, Register d ) { emit_long( op(arith_op) | rd(d) | op3(save_op3) | rs1(s1) | rs2(s2) ); }
       
  1441   void save(    Register s1, int simm13a, Register d ) { emit_long( op(arith_op) | rd(d) | op3(save_op3) | rs1(s1) | immed(true) | simm(simm13a, 13) ); }
       
  1442 
       
  1443   void restore( Register s1 = G0,  Register s2 = G0, Register d = G0 ) { emit_long( op(arith_op) | rd(d) | op3(restore_op3) | rs1(s1) | rs2(s2) ); }
       
  1444   void restore( Register s1,       int simm13a,      Register d      ) { emit_long( op(arith_op) | rd(d) | op3(restore_op3) | rs1(s1) | immed(true) | simm(simm13a, 13) ); }
       
  1445 
       
  1446   // pp 216
       
  1447 
       
  1448   void saved()    { v9_only();  emit_long( op(arith_op) | fcn(0) | op3(saved_op3)); }
       
  1449   void restored() { v9_only();  emit_long( op(arith_op) | fcn(1) | op3(saved_op3)); }
       
  1450 
       
  1451   // pp 217
       
  1452 
       
  1453   inline void sethi( int imm22a, Register d, RelocationHolder const& rspec = RelocationHolder() );
       
  1454   // pp 218
       
  1455 
       
  1456   void sll(  Register s1, Register s2, Register d ) { emit_long( op(arith_op) | rd(d) | op3(sll_op3) | rs1(s1) | sx(0) | rs2(s2) ); }
       
  1457   void sll(  Register s1, int imm5a,   Register d ) { emit_long( op(arith_op) | rd(d) | op3(sll_op3) | rs1(s1) | sx(0) | immed(true) | u_field(imm5a, 4, 0) ); }
       
  1458   void srl(  Register s1, Register s2, Register d ) { emit_long( op(arith_op) | rd(d) | op3(srl_op3) | rs1(s1) | sx(0) | rs2(s2) ); }
       
  1459   void srl(  Register s1, int imm5a,   Register d ) { emit_long( op(arith_op) | rd(d) | op3(srl_op3) | rs1(s1) | sx(0) | immed(true) | u_field(imm5a, 4, 0) ); }
       
  1460   void sra(  Register s1, Register s2, Register d ) { emit_long( op(arith_op) | rd(d) | op3(sra_op3) | rs1(s1) | sx(0) | rs2(s2) ); }
       
  1461   void sra(  Register s1, int imm5a,   Register d ) { emit_long( op(arith_op) | rd(d) | op3(sra_op3) | rs1(s1) | sx(0) | immed(true) | u_field(imm5a, 4, 0) ); }
       
  1462 
       
  1463   void sllx( Register s1, Register s2, Register d ) { v9_only();  emit_long( op(arith_op) | rd(d) | op3(sll_op3) | rs1(s1) | sx(1) | rs2(s2) ); }
       
  1464   void sllx( Register s1, int imm6a,   Register d ) { v9_only();  emit_long( op(arith_op) | rd(d) | op3(sll_op3) | rs1(s1) | sx(1) | immed(true) | u_field(imm6a, 5, 0) ); }
       
  1465   void srlx( Register s1, Register s2, Register d ) { v9_only();  emit_long( op(arith_op) | rd(d) | op3(srl_op3) | rs1(s1) | sx(1) | rs2(s2) ); }
       
  1466   void srlx( Register s1, int imm6a,   Register d ) { v9_only();  emit_long( op(arith_op) | rd(d) | op3(srl_op3) | rs1(s1) | sx(1) | immed(true) | u_field(imm6a, 5, 0) ); }
       
  1467   void srax( Register s1, Register s2, Register d ) { v9_only();  emit_long( op(arith_op) | rd(d) | op3(sra_op3) | rs1(s1) | sx(1) | rs2(s2) ); }
       
  1468   void srax( Register s1, int imm6a,   Register d ) { v9_only();  emit_long( op(arith_op) | rd(d) | op3(sra_op3) | rs1(s1) | sx(1) | immed(true) | u_field(imm6a, 5, 0) ); }
       
  1469 
       
  1470   // pp 220
       
  1471 
       
  1472   void sir( int simm13a ) { emit_long( op(arith_op) | fcn(15) | op3(sir_op3) | immed(true) | simm(simm13a, 13)); }
       
  1473 
       
  1474   // pp 221
       
  1475 
       
  1476   void stbar() { emit_long( op(arith_op) | op3(membar_op3) | u_field(15, 18, 14)); }
       
  1477 
       
  1478   // pp 222
       
  1479 
       
  1480   inline void stf(    FloatRegisterImpl::Width w, FloatRegister d, Register s1, Register s2 );
       
  1481   inline void stf(    FloatRegisterImpl::Width w, FloatRegister d, Register s1, int simm13a);
       
  1482   inline void stf(    FloatRegisterImpl::Width w, FloatRegister d, const Address& a, int offset = 0);
       
  1483 
       
  1484   inline void stfsr(  Register s1, Register s2 );
       
  1485   inline void stfsr(  Register s1, int simm13a);
       
  1486   inline void stxfsr( Register s1, Register s2 );
       
  1487   inline void stxfsr( Register s1, int simm13a);
       
  1488 
       
  1489   //  pp 224
       
  1490 
       
  1491   void stfa(  FloatRegisterImpl::Width w, FloatRegister d, Register s1, Register s2, int ia ) { v9_only();  emit_long( op(ldst_op) | fd(d, w) | alt_op3(stf_op3 | alt_bit_op3, w) | rs1(s1) | imm_asi(ia) | rs2(s2) ); }
       
  1492   void stfa(  FloatRegisterImpl::Width w, FloatRegister d, Register s1, int simm13a         ) { v9_only();  emit_long( op(ldst_op) | fd(d, w) | alt_op3(stf_op3 | alt_bit_op3, w) | rs1(s1) | immed(true) | simm(simm13a, 13) ); }
       
  1493 
       
  1494   // p 226
       
  1495 
       
  1496   inline void stb(  Register d, Register s1, Register s2 );
       
  1497   inline void stb(  Register d, Register s1, int simm13a);
       
  1498   inline void sth(  Register d, Register s1, Register s2 );
       
  1499   inline void sth(  Register d, Register s1, int simm13a);
       
  1500   inline void stw(  Register d, Register s1, Register s2 );
       
  1501   inline void stw(  Register d, Register s1, int simm13a);
       
  1502   inline void st(   Register d, Register s1, Register s2 );
       
  1503   inline void st(   Register d, Register s1, int simm13a);
       
  1504   inline void stx(  Register d, Register s1, Register s2 );
       
  1505   inline void stx(  Register d, Register s1, int simm13a);
       
  1506   inline void std(  Register d, Register s1, Register s2 );
       
  1507   inline void std(  Register d, Register s1, int simm13a);
       
  1508 
       
  1509   inline void stb(  Register d, const Address& a, int offset = 0 );
       
  1510   inline void sth(  Register d, const Address& a, int offset = 0 );
       
  1511   inline void stw(  Register d, const Address& a, int offset = 0 );
       
  1512   inline void stx(  Register d, const Address& a, int offset = 0 );
       
  1513   inline void st(   Register d, const Address& a, int offset = 0 );
       
  1514   inline void std(  Register d, const Address& a, int offset = 0 );
       
  1515 
       
  1516   // pp 177
       
  1517 
       
  1518   void stba(  Register d, Register s1, Register s2, int ia ) {             emit_long( op(ldst_op) | rd(d) | op3(stb_op3 | alt_bit_op3) | rs1(s1) | imm_asi(ia) | rs2(s2) ); }
       
  1519   void stba(  Register d, Register s1, int simm13a         ) {             emit_long( op(ldst_op) | rd(d) | op3(stb_op3 | alt_bit_op3) | rs1(s1) | immed(true) | simm(simm13a, 13) ); }
       
  1520   void stha(  Register d, Register s1, Register s2, int ia ) {             emit_long( op(ldst_op) | rd(d) | op3(sth_op3 | alt_bit_op3) | rs1(s1) | imm_asi(ia) | rs2(s2) ); }
       
  1521   void stha(  Register d, Register s1, int simm13a         ) {             emit_long( op(ldst_op) | rd(d) | op3(sth_op3 | alt_bit_op3) | rs1(s1) | immed(true) | simm(simm13a, 13) ); }
       
  1522   void stwa(  Register d, Register s1, Register s2, int ia ) {             emit_long( op(ldst_op) | rd(d) | op3(stw_op3 | alt_bit_op3) | rs1(s1) | imm_asi(ia) | rs2(s2) ); }
       
  1523   void stwa(  Register d, Register s1, int simm13a         ) {             emit_long( op(ldst_op) | rd(d) | op3(stw_op3 | alt_bit_op3) | rs1(s1) | immed(true) | simm(simm13a, 13) ); }
       
  1524   void stxa(  Register d, Register s1, Register s2, int ia ) { v9_only();  emit_long( op(ldst_op) | rd(d) | op3(stx_op3 | alt_bit_op3) | rs1(s1) | imm_asi(ia) | rs2(s2) ); }
       
  1525   void stxa(  Register d, Register s1, int simm13a         ) { v9_only();  emit_long( op(ldst_op) | rd(d) | op3(stx_op3 | alt_bit_op3) | rs1(s1) | immed(true) | simm(simm13a, 13) ); }
       
  1526   void stda(  Register d, Register s1, Register s2, int ia ) {             emit_long( op(ldst_op) | rd(d) | op3(std_op3 | alt_bit_op3) | rs1(s1) | imm_asi(ia) | rs2(s2) ); }
       
  1527   void stda(  Register d, Register s1, int simm13a         ) {             emit_long( op(ldst_op) | rd(d) | op3(std_op3 | alt_bit_op3) | rs1(s1) | immed(true) | simm(simm13a, 13) ); }
       
  1528 
       
  1529   // pp 97 (v8)
       
  1530 
       
  1531   inline void stc(   int crd, Register s1, Register s2 );
       
  1532   inline void stc(   int crd, Register s1, int simm13a);
       
  1533   inline void stdc(  int crd, Register s1, Register s2 );
       
  1534   inline void stdc(  int crd, Register s1, int simm13a);
       
  1535   inline void stcsr( int crd, Register s1, Register s2 );
       
  1536   inline void stcsr( int crd, Register s1, int simm13a);
       
  1537   inline void stdcq( int crd, Register s1, Register s2 );
       
  1538   inline void stdcq( int crd, Register s1, int simm13a);
       
  1539 
       
  1540   // pp 230
       
  1541 
       
  1542   void sub(    Register s1, Register s2, Register d ) { emit_long( op(arith_op) | rd(d) | op3(sub_op3              ) | rs1(s1) | rs2(s2) ); }
       
  1543   void sub(    Register s1, int simm13a, Register d ) { emit_long( op(arith_op) | rd(d) | op3(sub_op3              ) | rs1(s1) | immed(true) | simm(simm13a, 13) ); }
       
  1544   void subcc(  Register s1, Register s2, Register d ) { emit_long( op(arith_op) | rd(d) | op3(sub_op3 | cc_bit_op3 ) | rs1(s1) | rs2(s2) ); }
       
  1545   void subcc(  Register s1, int simm13a, Register d ) { emit_long( op(arith_op) | rd(d) | op3(sub_op3 | cc_bit_op3 ) | rs1(s1) | immed(true) | simm(simm13a, 13) ); }
       
  1546   void subc(   Register s1, Register s2, Register d ) { emit_long( op(arith_op) | rd(d) | op3(subc_op3             ) | rs1(s1) | rs2(s2) ); }
       
  1547   void subc(   Register s1, int simm13a, Register d ) { emit_long( op(arith_op) | rd(d) | op3(subc_op3             ) | rs1(s1) | immed(true) | simm(simm13a, 13) ); }
       
  1548   void subccc( Register s1, Register s2, Register d ) { emit_long( op(arith_op) | rd(d) | op3(subc_op3 | cc_bit_op3) | rs1(s1) | rs2(s2) ); }
       
  1549   void subccc( Register s1, int simm13a, Register d ) { emit_long( op(arith_op) | rd(d) | op3(subc_op3 | cc_bit_op3) | rs1(s1) | immed(true) | simm(simm13a, 13) ); }
       
  1550 
       
  1551   // pp 231
       
  1552 
       
  1553   inline void swap( Register s1, Register s2, Register d );
       
  1554   inline void swap( Register s1, int simm13a, Register d);
       
  1555   inline void swap( Address& a,               Register d, int offset = 0 );
       
  1556 
       
  1557   // pp 232
       
  1558 
       
  1559   void swapa(   Register s1, Register s2, int ia, Register d ) { v9_dep();  emit_long( op(ldst_op) | rd(d) | op3(swap_op3 | alt_bit_op3) | rs1(s1) | imm_asi(ia) | rs2(s2) ); }
       
  1560   void swapa(   Register s1, int simm13a,         Register d ) { v9_dep();  emit_long( op(ldst_op) | rd(d) | op3(swap_op3 | alt_bit_op3) | rs1(s1) | immed(true) | simm(simm13a, 13) ); }
       
  1561 
       
  1562   // pp 234, note op in book is wrong, see pp 268
       
  1563 
       
  1564   void taddcc(    Register s1, Register s2, Register d ) {            emit_long( op(arith_op) | rd(d) | op3(taddcc_op3  ) | rs1(s1) | rs2(s2) ); }
       
  1565   void taddcc(    Register s1, int simm13a, Register d ) {            emit_long( op(arith_op) | rd(d) | op3(taddcc_op3  ) | rs1(s1) | immed(true) | simm(simm13a, 13) ); }
       
  1566   void taddcctv(  Register s1, Register s2, Register d ) { v9_dep();  emit_long( op(arith_op) | rd(d) | op3(taddcctv_op3) | rs1(s1) | rs2(s2) ); }
       
  1567   void taddcctv(  Register s1, int simm13a, Register d ) { v9_dep();  emit_long( op(arith_op) | rd(d) | op3(taddcctv_op3) | rs1(s1) | immed(true) | simm(simm13a, 13) ); }
       
  1568 
       
  1569   // pp 235
       
  1570 
       
  1571   void tsubcc(    Register s1, Register s2, Register d ) { emit_long( op(arith_op) | rd(d) | op3(tsubcc_op3  ) | rs1(s1) | rs2(s2) ); }
       
  1572   void tsubcc(    Register s1, int simm13a, Register d ) { emit_long( op(arith_op) | rd(d) | op3(tsubcc_op3  ) | rs1(s1) | immed(true) | simm(simm13a, 13) ); }
       
  1573   void tsubcctv(  Register s1, Register s2, Register d ) { emit_long( op(arith_op) | rd(d) | op3(tsubcctv_op3) | rs1(s1) | rs2(s2) ); }
       
  1574   void tsubcctv(  Register s1, int simm13a, Register d ) { emit_long( op(arith_op) | rd(d) | op3(tsubcctv_op3) | rs1(s1) | immed(true) | simm(simm13a, 13) ); }
       
  1575 
       
  1576   // pp 237
       
  1577 
       
  1578   void trap( Condition c, CC cc, Register s1, Register s2 ) { v8_no_cc(cc);  emit_long( op(arith_op) | cond(c) | op3(trap_op3) | rs1(s1) | trapcc(cc) | rs2(s2)); }
       
  1579   void trap( Condition c, CC cc, Register s1, int trapa   ) { v8_no_cc(cc);  emit_long( op(arith_op) | cond(c) | op3(trap_op3) | rs1(s1) | trapcc(cc) | immed(true) | u_field(trapa, 6, 0)); }
       
  1580   // simple uncond. trap
       
  1581   void trap( int trapa ) { trap( always, icc, G0, trapa ); }
       
  1582 
       
  1583   // pp 239 omit write priv register for now
       
  1584 
       
  1585   inline void wry(    Register d) { v9_dep();  emit_long( op(arith_op) | rs1(d) | op3(wrreg_op3) | u_field(0, 29, 25)); }
       
  1586   inline void wrccr(Register s) { v9_only(); emit_long( op(arith_op) | rs1(s) | op3(wrreg_op3) | u_field(2, 29, 25)); }
       
  1587   inline void wrccr(Register s, int simm13a) { v9_only(); emit_long( op(arith_op) |
       
  1588                                                                            rs1(s) |
       
  1589                                                                            op3(wrreg_op3) |
       
  1590                                                                            u_field(2, 29, 25) |
       
  1591                                                                            u_field(1, 13, 13) |
       
  1592                                                                            simm(simm13a, 13)); }
       
  1593   inline void wrasi(  Register d) { v9_only(); emit_long( op(arith_op) | rs1(d) | op3(wrreg_op3) | u_field(3, 29, 25)); }
       
  1594   inline void wrfprs( Register d) { v9_only(); emit_long( op(arith_op) | rs1(d) | op3(wrreg_op3) | u_field(6, 29, 25)); }
       
  1595 
       
  1596 
       
  1597   // Creation
       
  1598   Assembler(CodeBuffer* code) : AbstractAssembler(code) {
       
  1599 #ifdef CHECK_DELAY
       
  1600     delay_state = no_delay;
       
  1601 #endif
       
  1602   }
       
  1603 
       
  1604   // Testing
       
  1605 #ifndef PRODUCT
       
  1606   void test_v9();
       
  1607   void test_v8_onlys();
       
  1608 #endif
       
  1609 };
       
  1610 
       
  1611 
       
  1612 class RegistersForDebugging : public StackObj {
       
  1613  public:
       
  1614   intptr_t i[8], l[8], o[8], g[8];
       
  1615   float    f[32];
       
  1616   double   d[32];
       
  1617 
       
  1618   void print(outputStream* s);
       
  1619 
       
  1620   static int i_offset(int j) { return offset_of(RegistersForDebugging, i[j]); }
       
  1621   static int l_offset(int j) { return offset_of(RegistersForDebugging, l[j]); }
       
  1622   static int o_offset(int j) { return offset_of(RegistersForDebugging, o[j]); }
       
  1623   static int g_offset(int j) { return offset_of(RegistersForDebugging, g[j]); }
       
  1624   static int f_offset(int j) { return offset_of(RegistersForDebugging, f[j]); }
       
  1625   static int d_offset(int j) { return offset_of(RegistersForDebugging, d[j / 2]); }
       
  1626 
       
  1627   // gen asm code to save regs
       
  1628   static void save_registers(MacroAssembler* a);
       
  1629 
       
  1630   // restore global registers in case C code disturbed them
       
  1631   static void restore_registers(MacroAssembler* a, Register r);
       
  1632 };
       
  1633 
       
  1634 
       
  1635 // MacroAssembler extends Assembler by a few frequently used macros.
       
  1636 //
       
  1637 // Most of the standard SPARC synthetic ops are defined here.
       
  1638 // Instructions for which a 'better' code sequence exists depending
       
  1639 // on arguments should also go in here.
       
  1640 
       
  1641 #define JMP2(r1, r2) jmp(r1, r2, __FILE__, __LINE__)
       
  1642 #define JMP(r1, off) jmp(r1, off, __FILE__, __LINE__)
       
  1643 #define JUMP(a, off)     jump(a, off, __FILE__, __LINE__)
       
  1644 #define JUMPL(a, d, off) jumpl(a, d, off, __FILE__, __LINE__)
       
  1645 
       
  1646 
       
  1647 class MacroAssembler: public Assembler {
       
  1648  protected:
       
  1649   // Support for VM calls
       
  1650   // This is the base routine called by the different versions of call_VM_leaf. The interpreter
       
  1651   // may customize this version by overriding it for its purposes (e.g., to save/restore
       
  1652   // additional registers when doing a VM call).
       
  1653 #ifdef CC_INTERP
       
  1654   #define VIRTUAL
       
  1655 #else
       
  1656   #define VIRTUAL virtual
       
  1657 #endif
       
  1658 
       
  1659   VIRTUAL void call_VM_leaf_base(Register thread_cache, address entry_point, int number_of_arguments);
       
  1660 
       
  1661   //
       
  1662   // It is imperative that all calls into the VM are handled via the call_VM macros.
       
  1663   // They make sure that the stack linkage is setup correctly. call_VM's correspond
       
  1664   // to ENTRY/ENTRY_X entry points while call_VM_leaf's correspond to LEAF entry points.
       
  1665   //
       
  1666   // This is the base routine called by the different versions of call_VM. The interpreter
       
  1667   // may customize this version by overriding it for its purposes (e.g., to save/restore
       
  1668   // additional registers when doing a VM call).
       
  1669   //
       
  1670   // A non-volatile java_thread_cache register should be specified so
       
  1671   // that the G2_thread value can be preserved across the call.
       
  1672   // (If java_thread_cache is noreg, then a slow get_thread call
       
  1673   // will re-initialize the G2_thread.) call_VM_base returns the register that contains the
       
  1674   // thread.
       
  1675   //
       
  1676   // If no last_java_sp is specified (noreg) than SP will be used instead.
       
  1677 
       
  1678   virtual void call_VM_base(
       
  1679     Register        oop_result,             // where an oop-result ends up if any; use noreg otherwise
       
  1680     Register        java_thread_cache,      // the thread if computed before     ; use noreg otherwise
       
  1681     Register        last_java_sp,           // to set up last_Java_frame in stubs; use noreg otherwise
       
  1682     address         entry_point,            // the entry point
       
  1683     int             number_of_arguments,    // the number of arguments (w/o thread) to pop after call
       
  1684     bool            check_exception=true    // flag which indicates if exception should be checked
       
  1685   );
       
  1686 
       
  1687   // This routine should emit JVMTI PopFrame and ForceEarlyReturn handling code.
       
  1688   // The implementation is only non-empty for the InterpreterMacroAssembler,
       
  1689   // as only the interpreter handles and ForceEarlyReturn PopFrame requests.
       
  1690   virtual void check_and_handle_popframe(Register scratch_reg);
       
  1691   virtual void check_and_handle_earlyret(Register scratch_reg);
       
  1692 
       
  1693  public:
       
  1694   MacroAssembler(CodeBuffer* code) : Assembler(code) {}
       
  1695 
       
  1696   // Support for NULL-checks
       
  1697   //
       
  1698   // Generates code that causes a NULL OS exception if the content of reg is NULL.
       
  1699   // If the accessed location is M[reg + offset] and the offset is known, provide the
       
  1700   // offset.  No explicit code generation is needed if the offset is within a certain
       
  1701   // range (0 <= offset <= page_size).
       
  1702   //
       
  1703   // %%%%%% Currently not done for SPARC
       
  1704 
       
  1705   void null_check(Register reg, int offset = -1);
       
  1706   static bool needs_explicit_null_check(intptr_t offset);
       
  1707 
       
  1708   // support for delayed instructions
       
  1709   MacroAssembler* delayed() { Assembler::delayed();  return this; }
       
  1710 
       
  1711   // branches that use right instruction for v8 vs. v9
       
  1712   inline void br( Condition c, bool a, Predict p, address d, relocInfo::relocType rt = relocInfo::none );
       
  1713   inline void br( Condition c, bool a, Predict p, Label& L );
       
  1714   inline void fb( Condition c, bool a, Predict p, address d, relocInfo::relocType rt = relocInfo::none );
       
  1715   inline void fb( Condition c, bool a, Predict p, Label& L );
       
  1716 
       
  1717   // compares register with zero and branches (V9 and V8 instructions)
       
  1718   void br_zero( Condition c, bool a, Predict p, Register s1, Label& L);
       
  1719   // Compares a pointer register with zero and branches on (not)null.
       
  1720   // Does a test & branch on 32-bit systems and a register-branch on 64-bit.
       
  1721   void br_null   ( Register s1, bool a, Predict p, Label& L );
       
  1722   void br_notnull( Register s1, bool a, Predict p, Label& L );
       
  1723 
       
  1724   inline void bp( Condition c, bool a, CC cc, Predict p, address d, relocInfo::relocType rt = relocInfo::none );
       
  1725   inline void bp( Condition c, bool a, CC cc, Predict p, Label& L );
       
  1726 
       
  1727   // Branch that tests xcc in LP64 and icc in !LP64
       
  1728   inline void brx( Condition c, bool a, Predict p, address d, relocInfo::relocType rt = relocInfo::none );
       
  1729   inline void brx( Condition c, bool a, Predict p, Label& L );
       
  1730 
       
  1731   // unconditional short branch
       
  1732   inline void ba( bool a, Label& L );
       
  1733 
       
  1734   // Branch that tests fp condition codes
       
  1735   inline void fbp( Condition c, bool a, CC cc, Predict p, address d, relocInfo::relocType rt = relocInfo::none );
       
  1736   inline void fbp( Condition c, bool a, CC cc, Predict p, Label& L );
       
  1737 
       
  1738   // get PC the best way
       
  1739   inline int get_pc( Register d );
       
  1740 
       
  1741   // Sparc shorthands(pp 85, V8 manual, pp 289 V9 manual)
       
  1742   inline void cmp(  Register s1, Register s2 ) { subcc( s1, s2, G0 ); }
       
  1743   inline void cmp(  Register s1, int simm13a ) { subcc( s1, simm13a, G0 ); }
       
  1744 
       
  1745   inline void jmp( Register s1, Register s2 );
       
  1746   inline void jmp( Register s1, int simm13a, RelocationHolder const& rspec = RelocationHolder() );
       
  1747 
       
  1748   inline void call( address d,  relocInfo::relocType rt = relocInfo::runtime_call_type );
       
  1749   inline void call( Label& L,   relocInfo::relocType rt = relocInfo::runtime_call_type );
       
  1750   inline void callr( Register s1, Register s2 );
       
  1751   inline void callr( Register s1, int simm13a, RelocationHolder const& rspec = RelocationHolder() );
       
  1752 
       
  1753   // Emits nothing on V8
       
  1754   inline void iprefetch( address d, relocInfo::relocType rt = relocInfo::none );
       
  1755   inline void iprefetch( Label& L);
       
  1756 
       
  1757   inline void tst( Register s ) { orcc( G0, s, G0 ); }
       
  1758 
       
  1759 #ifdef PRODUCT
       
  1760   inline void ret(  bool trace = TraceJumps )   { if (trace) {
       
  1761                                                     mov(I7, O7); // traceable register
       
  1762                                                     JMP(O7, 2 * BytesPerInstWord);
       
  1763                                                   } else {
       
  1764                                                     jmpl( I7, 2 * BytesPerInstWord, G0 );
       
  1765                                                   }
       
  1766                                                 }
       
  1767 
       
  1768   inline void retl( bool trace = TraceJumps )  { if (trace) JMP(O7, 2 * BytesPerInstWord);
       
  1769                                                  else jmpl( O7, 2 * BytesPerInstWord, G0 ); }
       
  1770 #else
       
  1771   void ret(  bool trace = TraceJumps );
       
  1772   void retl( bool trace = TraceJumps );
       
  1773 #endif /* PRODUCT */
       
  1774 
       
  1775   // Required platform-specific helpers for Label::patch_instructions.
       
  1776   // They _shadow_ the declarations in AbstractAssembler, which are undefined.
       
  1777   void pd_patch_instruction(address branch, address target);
       
  1778 #ifndef PRODUCT
       
  1779   static void pd_print_patched_instruction(address branch);
       
  1780 #endif
       
  1781 
       
  1782   // sethi Macro handles optimizations and relocations
       
  1783   void sethi( Address& a, bool ForceRelocatable = false );
       
  1784   void sethi( intptr_t imm22a, Register d, bool ForceRelocatable = false, RelocationHolder const& rspec = RelocationHolder());
       
  1785 
       
  1786   // compute the size of a sethi/set
       
  1787   static int  size_of_sethi( address a, bool worst_case = false );
       
  1788   static int  worst_case_size_of_set();
       
  1789 
       
  1790   // set may be either setsw or setuw (high 32 bits may be zero or sign)
       
  1791   void set(    intptr_t value, Register d, RelocationHolder const& rspec = RelocationHolder() );
       
  1792   void setsw(  int value, Register d, RelocationHolder const& rspec = RelocationHolder() );
       
  1793   void set64(  jlong value, Register d, Register tmp);
       
  1794 
       
  1795   // sign-extend 32 to 64
       
  1796   inline void signx( Register s, Register d ) { sra( s, G0, d); }
       
  1797   inline void signx( Register d )             { sra( d, G0, d); }
       
  1798 
       
  1799   inline void not1( Register s, Register d ) { xnor( s, G0, d ); }
       
  1800   inline void not1( Register d )             { xnor( d, G0, d ); }
       
  1801 
       
  1802   inline void neg( Register s, Register d ) { sub( G0, s, d ); }
       
  1803   inline void neg( Register d )             { sub( G0, d, d ); }
       
  1804 
       
  1805   inline void cas(  Register s1, Register s2, Register d) { casa( s1, s2, d, ASI_PRIMARY); }
       
  1806   inline void casx( Register s1, Register s2, Register d) { casxa(s1, s2, d, ASI_PRIMARY); }
       
  1807   // Functions for isolating 64 bit atomic swaps for LP64
       
  1808   // cas_ptr will perform cas for 32 bit VM's and casx for 64 bit VM's
       
  1809   inline void cas_ptr(  Register s1, Register s2, Register d) {
       
  1810 #ifdef _LP64
       
  1811     casx( s1, s2, d );
       
  1812 #else
       
  1813     cas( s1, s2, d );
       
  1814 #endif
       
  1815   }
       
  1816 
       
  1817   // Functions for isolating 64 bit shifts for LP64
       
  1818   inline void sll_ptr( Register s1, Register s2, Register d );
       
  1819   inline void sll_ptr( Register s1, int imm6a,   Register d );
       
  1820   inline void srl_ptr( Register s1, Register s2, Register d );
       
  1821   inline void srl_ptr( Register s1, int imm6a,   Register d );
       
  1822 
       
  1823   // little-endian
       
  1824   inline void casl(  Register s1, Register s2, Register d) { casa( s1, s2, d, ASI_PRIMARY_LITTLE); }
       
  1825   inline void casxl( Register s1, Register s2, Register d) { casxa(s1, s2, d, ASI_PRIMARY_LITTLE); }
       
  1826 
       
  1827   inline void inc(   Register d,  int const13 = 1 ) { add(   d, const13, d); }
       
  1828   inline void inccc( Register d,  int const13 = 1 ) { addcc( d, const13, d); }
       
  1829 
       
  1830   inline void dec(   Register d,  int const13 = 1 ) { sub(   d, const13, d); }
       
  1831   inline void deccc( Register d,  int const13 = 1 ) { subcc( d, const13, d); }
       
  1832 
       
  1833   inline void btst( Register s1,  Register s2 ) { andcc( s1, s2, G0 ); }
       
  1834   inline void btst( int simm13a,  Register s )  { andcc( s,  simm13a, G0 ); }
       
  1835 
       
  1836   inline void bset( Register s1,  Register s2 ) { or3( s1, s2, s2 ); }
       
  1837   inline void bset( int simm13a,  Register s )  { or3( s,  simm13a, s ); }
       
  1838 
       
  1839   inline void bclr( Register s1,  Register s2 ) { andn( s1, s2, s2 ); }
       
  1840   inline void bclr( int simm13a,  Register s )  { andn( s,  simm13a, s ); }
       
  1841 
       
  1842   inline void btog( Register s1,  Register s2 ) { xor3( s1, s2, s2 ); }
       
  1843   inline void btog( int simm13a,  Register s )  { xor3( s,  simm13a, s ); }
       
  1844 
       
  1845   inline void clr( Register d ) { or3( G0, G0, d ); }
       
  1846 
       
  1847   inline void clrb( Register s1, Register s2);
       
  1848   inline void clrh( Register s1, Register s2);
       
  1849   inline void clr(  Register s1, Register s2);
       
  1850   inline void clrx( Register s1, Register s2);
       
  1851 
       
  1852   inline void clrb( Register s1, int simm13a);
       
  1853   inline void clrh( Register s1, int simm13a);
       
  1854   inline void clr(  Register s1, int simm13a);
       
  1855   inline void clrx( Register s1, int simm13a);
       
  1856 
       
  1857   // copy & clear upper word
       
  1858   inline void clruw( Register s, Register d ) { srl( s, G0, d); }
       
  1859   // clear upper word
       
  1860   inline void clruwu( Register d ) { srl( d, G0, d); }
       
  1861 
       
  1862   // membar psuedo instruction.  takes into account target memory model.
       
  1863   inline void membar( Assembler::Membar_mask_bits const7a );
       
  1864 
       
  1865   // returns if membar generates anything.
       
  1866   inline bool membar_has_effect( Assembler::Membar_mask_bits const7a );
       
  1867 
       
  1868   // mov pseudo instructions
       
  1869   inline void mov( Register s,  Register d) {
       
  1870     if ( s != d )    or3( G0, s, d);
       
  1871     else             assert_not_delayed();  // Put something useful in the delay slot!
       
  1872   }
       
  1873 
       
  1874   inline void mov_or_nop( Register s,  Register d) {
       
  1875     if ( s != d )    or3( G0, s, d);
       
  1876     else             nop();
       
  1877   }
       
  1878 
       
  1879   inline void mov( int simm13a, Register d) { or3( G0, simm13a, d); }
       
  1880 
       
  1881   // address pseudos: make these names unlike instruction names to avoid confusion
       
  1882   inline void split_disp(    Address& a, Register temp );
       
  1883   inline intptr_t load_pc_address( Register reg, int bytes_to_skip );
       
  1884   inline void load_address(  Address& a, int offset = 0 );
       
  1885   inline void load_contents( Address& a, Register d, int offset = 0 );
       
  1886   inline void load_ptr_contents( Address& a, Register d, int offset = 0 );
       
  1887   inline void store_contents( Register s, Address& a, int offset = 0 );
       
  1888   inline void store_ptr_contents( Register s, Address& a, int offset = 0 );
       
  1889   inline void jumpl_to( Address& a, Register d, int offset = 0 );
       
  1890   inline void jump_to(  Address& a,             int offset = 0 );
       
  1891 
       
  1892   // ring buffer traceable jumps
       
  1893 
       
  1894   void jmp2( Register r1, Register r2, const char* file, int line );
       
  1895   void jmp ( Register r1, int offset,  const char* file, int line );
       
  1896 
       
  1897   void jumpl( Address& a, Register d, int offset, const char* file, int line );
       
  1898   void jump ( Address& a,             int offset, const char* file, int line );
       
  1899 
       
  1900 
       
  1901   // argument pseudos:
       
  1902 
       
  1903   inline void load_argument( Argument& a, Register  d );
       
  1904   inline void store_argument( Register s, Argument& a );
       
  1905   inline void store_ptr_argument( Register s, Argument& a );
       
  1906   inline void store_float_argument( FloatRegister s, Argument& a );
       
  1907   inline void store_double_argument( FloatRegister s, Argument& a );
       
  1908   inline void store_long_argument( Register s, Argument& a );
       
  1909 
       
  1910   // handy macros:
       
  1911 
       
  1912   inline void round_to( Register r, int modulus ) {
       
  1913     assert_not_delayed();
       
  1914     inc( r, modulus - 1 );
       
  1915     and3( r, -modulus, r );
       
  1916   }
       
  1917 
       
  1918   // --------------------------------------------------
       
  1919 
       
  1920   // Functions for isolating 64 bit loads for LP64
       
  1921   // ld_ptr will perform ld for 32 bit VM's and ldx for 64 bit VM's
       
  1922   // st_ptr will perform st for 32 bit VM's and stx for 64 bit VM's
       
  1923   inline void ld_ptr(   Register s1, Register s2, Register d );
       
  1924   inline void ld_ptr(   Register s1, int simm13a, Register d);
       
  1925   inline void ld_ptr(  const Address& a, Register d, int offset = 0 );
       
  1926   inline void st_ptr(  Register d, Register s1, Register s2 );
       
  1927   inline void st_ptr(  Register d, Register s1, int simm13a);
       
  1928   inline void st_ptr(  Register d, const Address& a, int offset = 0 );
       
  1929 
       
  1930   // ld_long will perform ld for 32 bit VM's and ldx for 64 bit VM's
       
  1931   // st_long will perform st for 32 bit VM's and stx for 64 bit VM's
       
  1932   inline void ld_long( Register s1, Register s2, Register d );
       
  1933   inline void ld_long( Register s1, int simm13a, Register d );
       
  1934   inline void ld_long( const Address& a, Register d, int offset = 0 );
       
  1935   inline void st_long( Register d, Register s1, Register s2 );
       
  1936   inline void st_long( Register d, Register s1, int simm13a );
       
  1937   inline void st_long( Register d, const Address& a, int offset = 0 );
       
  1938 
       
  1939   // --------------------------------------------------
       
  1940 
       
  1941  public:
       
  1942   // traps as per trap.h (SPARC ABI?)
       
  1943 
       
  1944   void breakpoint_trap();
       
  1945   void breakpoint_trap(Condition c, CC cc = icc);
       
  1946   void flush_windows_trap();
       
  1947   void clean_windows_trap();
       
  1948   void get_psr_trap();
       
  1949   void set_psr_trap();
       
  1950 
       
  1951   // V8/V9 flush_windows
       
  1952   void flush_windows();
       
  1953 
       
  1954   // Support for serializing memory accesses between threads
       
  1955   void serialize_memory(Register thread, Register tmp1, Register tmp2);
       
  1956 
       
  1957   // Stack frame creation/removal
       
  1958   void enter();
       
  1959   void leave();
       
  1960 
       
  1961   // V8/V9 integer multiply
       
  1962   void mult(Register s1, Register s2, Register d);
       
  1963   void mult(Register s1, int simm13a, Register d);
       
  1964 
       
  1965   // V8/V9 read and write of condition codes.
       
  1966   void read_ccr(Register d);
       
  1967   void write_ccr(Register s);
       
  1968 
       
  1969   // Manipulation of C++ bools
       
  1970   // These are idioms to flag the need for care with accessing bools but on
       
  1971   // this platform we assume byte size
       
  1972 
       
  1973   inline void stbool( Register d, const Address& a, int offset = 0 ) { stb(d, a, offset); }
       
  1974   inline void ldbool( const Address& a, Register d, int offset = 0 ) { ldsb( a, d, offset ); }
       
  1975   inline void tstbool( Register s ) { tst(s); }
       
  1976   inline void movbool( bool boolconst, Register d) { mov( (int) boolconst, d); }
       
  1977 
       
  1978   // Support for managing the JavaThread pointer (i.e.; the reference to
       
  1979   // thread-local information).
       
  1980   void get_thread();                                // load G2_thread
       
  1981   void verify_thread();                             // verify G2_thread contents
       
  1982   void save_thread   (const Register threache); // save to cache
       
  1983   void restore_thread(const Register thread_cache); // restore from cache
       
  1984 
       
  1985   // Support for last Java frame (but use call_VM instead where possible)
       
  1986   void set_last_Java_frame(Register last_java_sp, Register last_Java_pc);
       
  1987   void reset_last_Java_frame(void);
       
  1988 
       
  1989   // Call into the VM.
       
  1990   // Passes the thread pointer (in O0) as a prepended argument.
       
  1991   // Makes sure oop return values are visible to the GC.
       
  1992   void call_VM(Register oop_result, address entry_point, int number_of_arguments = 0, bool check_exceptions = true);
       
  1993   void call_VM(Register oop_result, address entry_point, Register arg_1, bool check_exceptions = true);
       
  1994   void call_VM(Register oop_result, address entry_point, Register arg_1, Register arg_2, bool check_exceptions = true);
       
  1995   void call_VM(Register oop_result, address entry_point, Register arg_1, Register arg_2, Register arg_3, bool check_exceptions = true);
       
  1996 
       
  1997   // these overloadings are not presently used on SPARC:
       
  1998   void call_VM(Register oop_result, Register last_java_sp, address entry_point, int number_of_arguments = 0, bool check_exceptions = true);
       
  1999   void call_VM(Register oop_result, Register last_java_sp, address entry_point, Register arg_1, bool check_exceptions = true);
       
  2000   void call_VM(Register oop_result, Register last_java_sp, address entry_point, Register arg_1, Register arg_2, bool check_exceptions = true);
       
  2001   void call_VM(Register oop_result, Register last_java_sp, address entry_point, Register arg_1, Register arg_2, Register arg_3, bool check_exceptions = true);
       
  2002 
       
  2003   void call_VM_leaf(Register thread_cache, address entry_point, int number_of_arguments = 0);
       
  2004   void call_VM_leaf(Register thread_cache, address entry_point, Register arg_1);
       
  2005   void call_VM_leaf(Register thread_cache, address entry_point, Register arg_1, Register arg_2);
       
  2006   void call_VM_leaf(Register thread_cache, address entry_point, Register arg_1, Register arg_2, Register arg_3);
       
  2007 
       
  2008   void get_vm_result  (Register oop_result);
       
  2009   void get_vm_result_2(Register oop_result);
       
  2010 
       
  2011   // vm result is currently getting hijacked to for oop preservation
       
  2012   void set_vm_result(Register oop_result);
       
  2013 
       
  2014   // if call_VM_base was called with check_exceptions=false, then call
       
  2015   // check_and_forward_exception to handle exceptions when it is safe
       
  2016   void check_and_forward_exception(Register scratch_reg);
       
  2017 
       
  2018  private:
       
  2019   // For V8
       
  2020   void read_ccr_trap(Register ccr_save);
       
  2021   void write_ccr_trap(Register ccr_save1, Register scratch1, Register scratch2);
       
  2022 
       
  2023 #ifdef ASSERT
       
  2024   // For V8 debugging.  Uses V8 instruction sequence and checks
       
  2025   // result with V9 insturctions rdccr and wrccr.
       
  2026   // Uses Gscatch and Gscatch2
       
  2027   void read_ccr_v8_assert(Register ccr_save);
       
  2028   void write_ccr_v8_assert(Register ccr_save);
       
  2029 #endif // ASSERT
       
  2030 
       
  2031  public:
       
  2032   // Stores
       
  2033   void store_check(Register tmp, Register obj);                // store check for obj - register is destroyed afterwards
       
  2034   void store_check(Register tmp, Register obj, Register offset); // store check for obj - register is destroyed afterwards
       
  2035 
       
  2036   // pushes double TOS element of FPU stack on CPU stack; pops from FPU stack
       
  2037   void push_fTOS();
       
  2038 
       
  2039   // pops double TOS element from CPU stack and pushes on FPU stack
       
  2040   void pop_fTOS();
       
  2041 
       
  2042   void empty_FPU_stack();
       
  2043 
       
  2044   void push_IU_state();
       
  2045   void pop_IU_state();
       
  2046 
       
  2047   void push_FPU_state();
       
  2048   void pop_FPU_state();
       
  2049 
       
  2050   void push_CPU_state();
       
  2051   void pop_CPU_state();
       
  2052 
       
  2053   // Debugging
       
  2054   void _verify_oop(Register reg, const char * msg, const char * file, int line);
       
  2055   void _verify_oop_addr(Address addr, const char * msg, const char * file, int line);
       
  2056 
       
  2057 #define verify_oop(reg) _verify_oop(reg, "broken oop " #reg, __FILE__, __LINE__)
       
  2058 #define verify_oop_addr(addr) _verify_oop_addr(addr, "broken oop addr ", __FILE__, __LINE__)
       
  2059 
       
  2060         // only if +VerifyOops
       
  2061   void verify_FPU(int stack_depth, const char* s = "illegal FPU state");
       
  2062         // only if +VerifyFPU
       
  2063   void stop(const char* msg);                          // prints msg, dumps registers and stops execution
       
  2064   void warn(const char* msg);                          // prints msg, but don't stop
       
  2065   void untested(const char* what = "");
       
  2066   void unimplemented(const char* what = "")              { char* b = new char[1024];  sprintf(b, "unimplemented: %s", what);  stop(b); }
       
  2067   void should_not_reach_here()                   { stop("should not reach here"); }
       
  2068   void print_CPU_state();
       
  2069 
       
  2070   // oops in code
       
  2071   Address allocate_oop_address( jobject obj, Register d ); // allocate_index
       
  2072   Address constant_oop_address( jobject obj, Register d ); // find_index
       
  2073   inline void set_oop         ( jobject obj, Register d ); // uses allocate_oop_address
       
  2074   inline void set_oop_constant( jobject obj, Register d ); // uses constant_oop_address
       
  2075   inline void set_oop         ( Address obj_addr );        // same as load_address
       
  2076 
       
  2077   // nop padding
       
  2078   void align(int modulus);
       
  2079 
       
  2080   // declare a safepoint
       
  2081   void safepoint();
       
  2082 
       
  2083   // factor out part of stop into subroutine to save space
       
  2084   void stop_subroutine();
       
  2085   // factor out part of verify_oop into subroutine to save space
       
  2086   void verify_oop_subroutine();
       
  2087 
       
  2088   // side-door communication with signalHandler in os_solaris.cpp
       
  2089   static address _verify_oop_implicit_branch[3];
       
  2090 
       
  2091 #ifndef PRODUCT
       
  2092   static void test();
       
  2093 #endif
       
  2094 
       
  2095   // convert an incoming arglist to varargs format; put the pointer in d
       
  2096   void set_varargs( Argument a, Register d );
       
  2097 
       
  2098   int total_frame_size_in_bytes(int extraWords);
       
  2099 
       
  2100   // used when extraWords known statically
       
  2101   void save_frame(int extraWords);
       
  2102   void save_frame_c1(int size_in_bytes);
       
  2103   // make a frame, and simultaneously pass up one or two register value
       
  2104   // into the new register window
       
  2105   void save_frame_and_mov(int extraWords, Register s1, Register d1, Register s2 = Register(), Register d2 = Register());
       
  2106 
       
  2107   // give no. (outgoing) params, calc # of words will need on frame
       
  2108   void calc_mem_param_words(Register Rparam_words, Register Rresult);
       
  2109 
       
  2110   // used to calculate frame size dynamically
       
  2111   // result is in bytes and must be negated for save inst
       
  2112   void calc_frame_size(Register extraWords, Register resultReg);
       
  2113 
       
  2114   // calc and also save
       
  2115   void calc_frame_size_and_save(Register extraWords, Register resultReg);
       
  2116 
       
  2117   static void debug(char* msg, RegistersForDebugging* outWindow);
       
  2118 
       
  2119   // implementations of bytecodes used by both interpreter and compiler
       
  2120 
       
  2121   void lcmp( Register Ra_hi, Register Ra_low,
       
  2122              Register Rb_hi, Register Rb_low,
       
  2123              Register Rresult);
       
  2124 
       
  2125   void lneg( Register Rhi, Register Rlow );
       
  2126 
       
  2127   void lshl(  Register Rin_high,  Register Rin_low,  Register Rcount,
       
  2128               Register Rout_high, Register Rout_low, Register Rtemp );
       
  2129 
       
  2130   void lshr(  Register Rin_high,  Register Rin_low,  Register Rcount,
       
  2131               Register Rout_high, Register Rout_low, Register Rtemp );
       
  2132 
       
  2133   void lushr( Register Rin_high,  Register Rin_low,  Register Rcount,
       
  2134               Register Rout_high, Register Rout_low, Register Rtemp );
       
  2135 
       
  2136 #ifdef _LP64
       
  2137   void lcmp( Register Ra, Register Rb, Register Rresult);
       
  2138 #endif
       
  2139 
       
  2140   void float_cmp( bool is_float, int unordered_result,
       
  2141                   FloatRegister Fa, FloatRegister Fb,
       
  2142                   Register Rresult);
       
  2143 
       
  2144   void fneg( FloatRegisterImpl::Width w, FloatRegister s, FloatRegister d);
       
  2145   void fneg( FloatRegisterImpl::Width w, FloatRegister sd ) { Assembler::fneg(w, sd); }
       
  2146   void fmov( FloatRegisterImpl::Width w, FloatRegister s, FloatRegister d);
       
  2147   void fabs( FloatRegisterImpl::Width w, FloatRegister s, FloatRegister d);
       
  2148 
       
  2149   void save_all_globals_into_locals();
       
  2150   void restore_globals_from_locals();
       
  2151 
       
  2152   void casx_under_lock(Register top_ptr_reg, Register top_reg, Register ptr_reg,
       
  2153     address lock_addr=0, bool use_call_vm=false);
       
  2154   void cas_under_lock(Register top_ptr_reg, Register top_reg, Register ptr_reg,
       
  2155     address lock_addr=0, bool use_call_vm=false);
       
  2156   void casn (Register addr_reg, Register cmp_reg, Register set_reg) ;
       
  2157 
       
  2158   // These set the icc condition code to equal if the lock succeeded
       
  2159   // and notEqual if it failed and requires a slow case
       
  2160   void compiler_lock_object(Register Roop, Register Rmark, Register Rbox, Register Rscratch,
       
  2161                               BiasedLockingCounters* counters = NULL);
       
  2162   void compiler_unlock_object(Register Roop, Register Rmark, Register Rbox, Register Rscratch);
       
  2163 
       
  2164   // Biased locking support
       
  2165   // Upon entry, lock_reg must point to the lock record on the stack,
       
  2166   // obj_reg must contain the target object, and mark_reg must contain
       
  2167   // the target object's header.
       
  2168   // Destroys mark_reg if an attempt is made to bias an anonymously
       
  2169   // biased lock. In this case a failure will go either to the slow
       
  2170   // case or fall through with the notEqual condition code set with
       
  2171   // the expectation that the slow case in the runtime will be called.
       
  2172   // In the fall-through case where the CAS-based lock is done,
       
  2173   // mark_reg is not destroyed.
       
  2174   void biased_locking_enter(Register obj_reg, Register mark_reg, Register temp_reg,
       
  2175                             Label& done, Label* slow_case = NULL,
       
  2176                             BiasedLockingCounters* counters = NULL);
       
  2177   // Upon entry, the base register of mark_addr must contain the oop.
       
  2178   // Destroys temp_reg.
       
  2179 
       
  2180   // If allow_delay_slot_filling is set to true, the next instruction
       
  2181   // emitted after this one will go in an annulled delay slot if the
       
  2182   // biased locking exit case failed.
       
  2183   void biased_locking_exit(Address mark_addr, Register temp_reg, Label& done, bool allow_delay_slot_filling = false);
       
  2184 
       
  2185   // allocation
       
  2186   void eden_allocate(
       
  2187     Register obj,                      // result: pointer to object after successful allocation
       
  2188     Register var_size_in_bytes,        // object size in bytes if unknown at compile time; invalid otherwise
       
  2189     int      con_size_in_bytes,        // object size in bytes if   known at compile time
       
  2190     Register t1,                       // temp register
       
  2191     Register t2,                       // temp register
       
  2192     Label&   slow_case                 // continuation point if fast allocation fails
       
  2193   );
       
  2194   void tlab_allocate(
       
  2195     Register obj,                      // result: pointer to object after successful allocation
       
  2196     Register var_size_in_bytes,        // object size in bytes if unknown at compile time; invalid otherwise
       
  2197     int      con_size_in_bytes,        // object size in bytes if   known at compile time
       
  2198     Register t1,                       // temp register
       
  2199     Label&   slow_case                 // continuation point if fast allocation fails
       
  2200   );
       
  2201   void tlab_refill(Label& retry_tlab, Label& try_eden, Label& slow_case);
       
  2202 
       
  2203   // Stack overflow checking
       
  2204 
       
  2205   // Note: this clobbers G3_scratch
       
  2206   void bang_stack_with_offset(int offset) {
       
  2207     // stack grows down, caller passes positive offset
       
  2208     assert(offset > 0, "must bang with negative offset");
       
  2209     set((-offset)+STACK_BIAS, G3_scratch);
       
  2210     st(G0, SP, G3_scratch);
       
  2211   }
       
  2212 
       
  2213   // Writes to stack successive pages until offset reached to check for
       
  2214   // stack overflow + shadow pages.  Clobbers tsp and scratch registers.
       
  2215   void bang_stack_size(Register Rsize, Register Rtsp, Register Rscratch);
       
  2216 
       
  2217   void verify_tlab();
       
  2218 
       
  2219   Condition negate_condition(Condition cond);
       
  2220 
       
  2221   // Helper functions for statistics gathering.
       
  2222   // Conditionally (non-atomically) increments passed counter address, preserving condition codes.
       
  2223   void cond_inc(Condition cond, address counter_addr, Register Rtemp1, Register Rtemp2);
       
  2224   // Unconditional increment.
       
  2225   void inc_counter(address counter_addr, Register Rtemp1, Register Rtemp2);
       
  2226 
       
  2227 #undef VIRTUAL
       
  2228 
       
  2229 };
       
  2230 
       
  2231 /**
       
  2232  * class SkipIfEqual:
       
  2233  *
       
  2234  * Instantiating this class will result in assembly code being output that will
       
  2235  * jump around any code emitted between the creation of the instance and it's
       
  2236  * automatic destruction at the end of a scope block, depending on the value of
       
  2237  * the flag passed to the constructor, which will be checked at run-time.
       
  2238  */
       
  2239 class SkipIfEqual : public StackObj {
       
  2240  private:
       
  2241   MacroAssembler* _masm;
       
  2242   Label _label;
       
  2243 
       
  2244  public:
       
  2245    // 'temp' is a temp register that this object can use (and trash)
       
  2246    SkipIfEqual(MacroAssembler*, Register temp,
       
  2247                const bool* flag_addr, Assembler::Condition condition);
       
  2248    ~SkipIfEqual();
       
  2249 };
       
  2250 
       
  2251 #ifdef ASSERT
       
  2252 // On RISC, there's no benefit to verifying instruction boundaries.
       
  2253 inline bool AbstractAssembler::pd_check_instruction_mark() { return false; }
       
  2254 #endif