1 /*
   2  * Copyright (c) 1997, 2015, Oracle and/or its affiliates. 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 Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA
  20  * or visit www.oracle.com if you need additional information or have any
  21  * questions.
  22  *
  23  */
  24 
  25 #ifndef CPU_X86_VM_MACROASSEMBLER_X86_HPP
  26 #define CPU_X86_VM_MACROASSEMBLER_X86_HPP
  27 
  28 #include "asm/assembler.hpp"
  29 #include "utilities/macros.hpp"
  30 #include "runtime/rtmLocking.hpp"
  31 
  32 // MacroAssembler extends Assembler by frequently used macros.
  33 //
  34 // Instructions for which a 'better' code sequence exists depending
  35 // on arguments should also go in here.
  36 
  37 class MacroAssembler: public Assembler {
  38   friend class LIR_Assembler;
  39   friend class Runtime1;      // as_Address()
  40 
  41  protected:
  42 
  43   Address as_Address(AddressLiteral adr);
  44   Address as_Address(ArrayAddress adr);
  45 
  46   // Support for VM calls
  47   //
  48   // This is the base routine called by the different versions of call_VM_leaf. The interpreter
  49   // may customize this version by overriding it for its purposes (e.g., to save/restore
  50   // additional registers when doing a VM call).
  51 #define COMMA ,
  52 
  53   virtual void call_VM_leaf_base(
  54     address entry_point,               // the entry point
  55     int     number_of_arguments        // the number of arguments to pop after the call
  56   );
  57 
  58   // This is the base routine called by the different versions of call_VM. The interpreter
  59   // may customize this version by overriding it for its purposes (e.g., to save/restore
  60   // additional registers when doing a VM call).
  61   //
  62   // If no java_thread register is specified (noreg) than rdi will be used instead. call_VM_base
  63   // returns the register which contains the thread upon return. If a thread register has been
  64   // specified, the return value will correspond to that register. If no last_java_sp is specified
  65   // (noreg) than rsp will be used instead.
  66   virtual void call_VM_base(           // returns the register containing the thread upon return
  67     Register oop_result,               // where an oop-result ends up if any; use noreg otherwise
  68     Register java_thread,              // the thread if computed before     ; use noreg otherwise
  69     Register last_java_sp,             // to set up last_Java_frame in stubs; use noreg otherwise
  70     address  entry_point,              // the entry point
  71     int      number_of_arguments,      // the number of arguments (w/o thread) to pop after the call
  72     bool     check_exceptions          // whether to check for pending exceptions after return
  73   );
  74 
  75   // These routines should emit JVMTI PopFrame and ForceEarlyReturn handling code.
  76   // The implementation is only non-empty for the InterpreterMacroAssembler,
  77   // as only the interpreter handles PopFrame and ForceEarlyReturn requests.
  78   virtual void check_and_handle_popframe(Register java_thread);
  79   virtual void check_and_handle_earlyret(Register java_thread);
  80 
  81   void call_VM_helper(Register oop_result, address entry_point, int number_of_arguments, bool check_exceptions = true);
  82 
  83   // helpers for FPU flag access
  84   // tmp is a temporary register, if none is available use noreg
  85   void save_rax   (Register tmp);
  86   void restore_rax(Register tmp);
  87 
  88  public:
  89   MacroAssembler(CodeBuffer* code) : Assembler(code) {}
  90 
  91   // Support for NULL-checks
  92   //
  93   // Generates code that causes a NULL OS exception if the content of reg is NULL.
  94   // If the accessed location is M[reg + offset] and the offset is known, provide the
  95   // offset. No explicit code generation is needed if the offset is within a certain
  96   // range (0 <= offset <= page_size).
  97 
  98   void null_check(Register reg, int offset = -1);
  99   static bool needs_explicit_null_check(intptr_t offset);
 100 
 101   // Required platform-specific helpers for Label::patch_instructions.
 102   // They _shadow_ the declarations in AbstractAssembler, which are undefined.
 103   void pd_patch_instruction(address branch, address target) {
 104     unsigned char op = branch[0];
 105     assert(op == 0xE8 /* call */ ||
 106         op == 0xE9 /* jmp */ ||
 107         op == 0xEB /* short jmp */ ||
 108         (op & 0xF0) == 0x70 /* short jcc */ ||
 109         op == 0x0F && (branch[1] & 0xF0) == 0x80 /* jcc */ ||
 110         op == 0xC7 && branch[1] == 0xF8 /* xbegin */,
 111         "Invalid opcode at patch point");
 112 
 113     if (op == 0xEB || (op & 0xF0) == 0x70) {
 114       // short offset operators (jmp and jcc)
 115       char* disp = (char*) &branch[1];
 116       int imm8 = target - (address) &disp[1];
 117       guarantee(this->is8bit(imm8), "Short forward jump exceeds 8-bit offset");
 118       *disp = imm8;
 119     } else {
 120       int* disp = (int*) &branch[(op == 0x0F || op == 0xC7)? 2: 1];
 121       int imm32 = target - (address) &disp[1];
 122       *disp = imm32;
 123     }
 124   }
 125 
 126   // The following 4 methods return the offset of the appropriate move instruction
 127 
 128   // Support for fast byte/short loading with zero extension (depending on particular CPU)
 129   int load_unsigned_byte(Register dst, Address src);
 130   int load_unsigned_short(Register dst, Address src);
 131 
 132   // Support for fast byte/short loading with sign extension (depending on particular CPU)
 133   int load_signed_byte(Register dst, Address src);
 134   int load_signed_short(Register dst, Address src);
 135 
 136   // Support for sign-extension (hi:lo = extend_sign(lo))
 137   void extend_sign(Register hi, Register lo);
 138 
 139   // Load and store values by size and signed-ness
 140   void load_sized_value(Register dst, Address src, size_t size_in_bytes, bool is_signed, Register dst2 = noreg);
 141   void store_sized_value(Address dst, Register src, size_t size_in_bytes, Register src2 = noreg);
 142 
 143   // Support for inc/dec with optimal instruction selection depending on value
 144 
 145   void increment(Register reg, int value = 1) { LP64_ONLY(incrementq(reg, value)) NOT_LP64(incrementl(reg, value)) ; }
 146   void decrement(Register reg, int value = 1) { LP64_ONLY(decrementq(reg, value)) NOT_LP64(decrementl(reg, value)) ; }
 147 
 148   void decrementl(Address dst, int value = 1);
 149   void decrementl(Register reg, int value = 1);
 150 
 151   void decrementq(Register reg, int value = 1);
 152   void decrementq(Address dst, int value = 1);
 153 
 154   void incrementl(Address dst, int value = 1);
 155   void incrementl(Register reg, int value = 1);
 156 
 157   void incrementq(Register reg, int value = 1);
 158   void incrementq(Address dst, int value = 1);
 159 
 160   // Support optimal SSE move instructions.
 161   void movflt(XMMRegister dst, XMMRegister src) {
 162     if (UseXmmRegToRegMoveAll) { movaps(dst, src); return; }
 163     else                       { movss (dst, src); return; }
 164   }
 165   void movflt(XMMRegister dst, Address src) { movss(dst, src); }
 166   void movflt(XMMRegister dst, AddressLiteral src);
 167   void movflt(Address dst, XMMRegister src) { movss(dst, src); }
 168 
 169   void movdbl(XMMRegister dst, XMMRegister src) {
 170     if (UseXmmRegToRegMoveAll) { movapd(dst, src); return; }
 171     else                       { movsd (dst, src); return; }
 172   }
 173 
 174   void movdbl(XMMRegister dst, AddressLiteral src);
 175 
 176   void movdbl(XMMRegister dst, Address src) {
 177     if (UseXmmLoadAndClearUpper) { movsd (dst, src); return; }
 178     else                         { movlpd(dst, src); return; }
 179   }
 180   void movdbl(Address dst, XMMRegister src) { movsd(dst, src); }
 181 
 182   void incrementl(AddressLiteral dst);
 183   void incrementl(ArrayAddress dst);
 184 
 185   void incrementq(AddressLiteral dst);
 186 
 187   // Alignment
 188   void align(int modulus);
 189   void align(int modulus, int target);
 190 
 191   // A 5 byte nop that is safe for patching (see patch_verified_entry)
 192   void fat_nop();
 193 
 194   // Stack frame creation/removal
 195   void enter();
 196   void leave();
 197 
 198   // Support for getting the JavaThread pointer (i.e.; a reference to thread-local information)
 199   // The pointer will be loaded into the thread register.
 200   void get_thread(Register thread);
 201 
 202 
 203   // Support for VM calls
 204   //
 205   // It is imperative that all calls into the VM are handled via the call_VM macros.
 206   // They make sure that the stack linkage is setup correctly. call_VM's correspond
 207   // to ENTRY/ENTRY_X entry points while call_VM_leaf's correspond to LEAF entry points.
 208 
 209 
 210   void call_VM(Register oop_result,
 211                address entry_point,
 212                bool check_exceptions = true);
 213   void call_VM(Register oop_result,
 214                address entry_point,
 215                Register arg_1,
 216                bool check_exceptions = true);
 217   void call_VM(Register oop_result,
 218                address entry_point,
 219                Register arg_1, Register arg_2,
 220                bool check_exceptions = true);
 221   void call_VM(Register oop_result,
 222                address entry_point,
 223                Register arg_1, Register arg_2, Register arg_3,
 224                bool check_exceptions = true);
 225 
 226   // Overloadings with last_Java_sp
 227   void call_VM(Register oop_result,
 228                Register last_java_sp,
 229                address entry_point,
 230                int number_of_arguments = 0,
 231                bool check_exceptions = true);
 232   void call_VM(Register oop_result,
 233                Register last_java_sp,
 234                address entry_point,
 235                Register arg_1, bool
 236                check_exceptions = true);
 237   void call_VM(Register oop_result,
 238                Register last_java_sp,
 239                address entry_point,
 240                Register arg_1, Register arg_2,
 241                bool check_exceptions = true);
 242   void call_VM(Register oop_result,
 243                Register last_java_sp,
 244                address entry_point,
 245                Register arg_1, Register arg_2, Register arg_3,
 246                bool check_exceptions = true);
 247 
 248   void get_vm_result  (Register oop_result, Register thread);
 249   void get_vm_result_2(Register metadata_result, Register thread);
 250 
 251   // These always tightly bind to MacroAssembler::call_VM_base
 252   // bypassing the virtual implementation
 253   void super_call_VM(Register oop_result, Register last_java_sp, address entry_point, int number_of_arguments = 0, bool check_exceptions = true);
 254   void super_call_VM(Register oop_result, Register last_java_sp, address entry_point, Register arg_1, bool check_exceptions = true);
 255   void super_call_VM(Register oop_result, Register last_java_sp, address entry_point, Register arg_1, Register arg_2, bool check_exceptions = true);
 256   void super_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);
 257   void super_call_VM(Register oop_result, Register last_java_sp, address entry_point, Register arg_1, Register arg_2, Register arg_3, Register arg_4, bool check_exceptions = true);
 258 
 259   void call_VM_leaf(address entry_point,
 260                     int number_of_arguments = 0);
 261   void call_VM_leaf(address entry_point,
 262                     Register arg_1);
 263   void call_VM_leaf(address entry_point,
 264                     Register arg_1, Register arg_2);
 265   void call_VM_leaf(address entry_point,
 266                     Register arg_1, Register arg_2, Register arg_3);
 267 
 268   // These always tightly bind to MacroAssembler::call_VM_leaf_base
 269   // bypassing the virtual implementation
 270   void super_call_VM_leaf(address entry_point);
 271   void super_call_VM_leaf(address entry_point, Register arg_1);
 272   void super_call_VM_leaf(address entry_point, Register arg_1, Register arg_2);
 273   void super_call_VM_leaf(address entry_point, Register arg_1, Register arg_2, Register arg_3);
 274   void super_call_VM_leaf(address entry_point, Register arg_1, Register arg_2, Register arg_3, Register arg_4);
 275 
 276   // last Java Frame (fills frame anchor)
 277   void set_last_Java_frame(Register thread,
 278                            Register last_java_sp,
 279                            Register last_java_fp,
 280                            address last_java_pc);
 281 
 282   // thread in the default location (r15_thread on 64bit)
 283   void set_last_Java_frame(Register last_java_sp,
 284                            Register last_java_fp,
 285                            address last_java_pc);
 286 
 287   void reset_last_Java_frame(Register thread, bool clear_fp, bool clear_pc);
 288 
 289   // thread in the default location (r15_thread on 64bit)
 290   void reset_last_Java_frame(bool clear_fp, bool clear_pc);
 291 
 292   // Stores
 293   void store_check(Register obj);                // store check for obj - register is destroyed afterwards
 294   void store_check(Register obj, Address dst);   // same as above, dst is exact store location (reg. is destroyed)
 295 
 296 #if INCLUDE_ALL_GCS
 297 
 298   void g1_write_barrier_pre(Register obj,
 299                             Register pre_val,
 300                             Register thread,
 301                             Register tmp,
 302                             bool tosca_live,
 303                             bool expand_call);
 304 
 305   void g1_write_barrier_post(Register store_addr,
 306                              Register new_val,
 307                              Register thread,
 308                              Register tmp,
 309                              Register tmp2);
 310 
 311 #endif // INCLUDE_ALL_GCS
 312 
 313   // C 'boolean' to Java boolean: x == 0 ? 0 : 1
 314   void c2bool(Register x);
 315 
 316   // C++ bool manipulation
 317 
 318   void movbool(Register dst, Address src);
 319   void movbool(Address dst, bool boolconst);
 320   void movbool(Address dst, Register src);
 321   void testbool(Register dst);
 322 
 323   // oop manipulations
 324   void load_klass(Register dst, Register src);
 325   void store_klass(Register dst, Register src);
 326 
 327   void load_heap_oop(Register dst, Address src);
 328   void load_heap_oop_not_null(Register dst, Address src);
 329   void store_heap_oop(Address dst, Register src);
 330   void cmp_heap_oop(Register src1, Address src2, Register tmp = noreg);
 331 
 332   // Used for storing NULL. All other oop constants should be
 333   // stored using routines that take a jobject.
 334   void store_heap_oop_null(Address dst);
 335 
 336   void load_prototype_header(Register dst, Register src);
 337 
 338 #ifdef _LP64
 339   void store_klass_gap(Register dst, Register src);
 340 
 341   // This dummy is to prevent a call to store_heap_oop from
 342   // converting a zero (like NULL) into a Register by giving
 343   // the compiler two choices it can't resolve
 344 
 345   void store_heap_oop(Address dst, void* dummy);
 346 
 347   void encode_heap_oop(Register r);
 348   void decode_heap_oop(Register r);
 349   void encode_heap_oop_not_null(Register r);
 350   void decode_heap_oop_not_null(Register r);
 351   void encode_heap_oop_not_null(Register dst, Register src);
 352   void decode_heap_oop_not_null(Register dst, Register src);
 353 
 354   void set_narrow_oop(Register dst, jobject obj);
 355   void set_narrow_oop(Address dst, jobject obj);
 356   void cmp_narrow_oop(Register dst, jobject obj);
 357   void cmp_narrow_oop(Address dst, jobject obj);
 358 
 359   void encode_klass_not_null(Register r);
 360   void decode_klass_not_null(Register r);
 361   void encode_klass_not_null(Register dst, Register src);
 362   void decode_klass_not_null(Register dst, Register src);
 363   void set_narrow_klass(Register dst, Klass* k);
 364   void set_narrow_klass(Address dst, Klass* k);
 365   void cmp_narrow_klass(Register dst, Klass* k);
 366   void cmp_narrow_klass(Address dst, Klass* k);
 367 
 368   // Returns the byte size of the instructions generated by decode_klass_not_null()
 369   // when compressed klass pointers are being used.
 370   static int instr_size_for_decode_klass_not_null();
 371 
 372   // if heap base register is used - reinit it with the correct value
 373   void reinit_heapbase();
 374 
 375   DEBUG_ONLY(void verify_heapbase(const char* msg);)
 376 
 377 #endif // _LP64
 378 
 379   // Int division/remainder for Java
 380   // (as idivl, but checks for special case as described in JVM spec.)
 381   // returns idivl instruction offset for implicit exception handling
 382   int corrected_idivl(Register reg);
 383 
 384   // Long division/remainder for Java
 385   // (as idivq, but checks for special case as described in JVM spec.)
 386   // returns idivq instruction offset for implicit exception handling
 387   int corrected_idivq(Register reg);
 388 
 389   void int3();
 390 
 391   // Long operation macros for a 32bit cpu
 392   // Long negation for Java
 393   void lneg(Register hi, Register lo);
 394 
 395   // Long multiplication for Java
 396   // (destroys contents of eax, ebx, ecx and edx)
 397   void lmul(int x_rsp_offset, int y_rsp_offset); // rdx:rax = x * y
 398 
 399   // Long shifts for Java
 400   // (semantics as described in JVM spec.)
 401   void lshl(Register hi, Register lo);                               // hi:lo << (rcx & 0x3f)
 402   void lshr(Register hi, Register lo, bool sign_extension = false);  // hi:lo >> (rcx & 0x3f)
 403 
 404   // Long compare for Java
 405   // (semantics as described in JVM spec.)
 406   void lcmp2int(Register x_hi, Register x_lo, Register y_hi, Register y_lo); // x_hi = lcmp(x, y)
 407 
 408 
 409   // misc
 410 
 411   // Sign extension
 412   void sign_extend_short(Register reg);
 413   void sign_extend_byte(Register reg);
 414 
 415   // Division by power of 2, rounding towards 0
 416   void division_with_shift(Register reg, int shift_value);
 417 
 418   // Compares the top-most stack entries on the FPU stack and sets the eflags as follows:
 419   //
 420   // CF (corresponds to C0) if x < y
 421   // PF (corresponds to C2) if unordered
 422   // ZF (corresponds to C3) if x = y
 423   //
 424   // The arguments are in reversed order on the stack (i.e., top of stack is first argument).
 425   // tmp is a temporary register, if none is available use noreg (only matters for non-P6 code)
 426   void fcmp(Register tmp);
 427   // Variant of the above which allows y to be further down the stack
 428   // and which only pops x and y if specified. If pop_right is
 429   // specified then pop_left must also be specified.
 430   void fcmp(Register tmp, int index, bool pop_left, bool pop_right);
 431 
 432   // Floating-point comparison for Java
 433   // Compares the top-most stack entries on the FPU stack and stores the result in dst.
 434   // The arguments are in reversed order on the stack (i.e., top of stack is first argument).
 435   // (semantics as described in JVM spec.)
 436   void fcmp2int(Register dst, bool unordered_is_less);
 437   // Variant of the above which allows y to be further down the stack
 438   // and which only pops x and y if specified. If pop_right is
 439   // specified then pop_left must also be specified.
 440   void fcmp2int(Register dst, bool unordered_is_less, int index, bool pop_left, bool pop_right);
 441 
 442   // Floating-point remainder for Java (ST0 = ST0 fremr ST1, ST1 is empty afterwards)
 443   // tmp is a temporary register, if none is available use noreg
 444   void fremr(Register tmp);
 445 
 446 
 447   // same as fcmp2int, but using SSE2
 448   void cmpss2int(XMMRegister opr1, XMMRegister opr2, Register dst, bool unordered_is_less);
 449   void cmpsd2int(XMMRegister opr1, XMMRegister opr2, Register dst, bool unordered_is_less);
 450 
 451   // Inlined sin/cos generator for Java; must not use CPU instruction
 452   // directly on Intel as it does not have high enough precision
 453   // outside of the range [-pi/4, pi/4]. Extra argument indicate the
 454   // number of FPU stack slots in use; all but the topmost will
 455   // require saving if a slow case is necessary. Assumes argument is
 456   // on FP TOS; result is on FP TOS.  No cpu registers are changed by
 457   // this code.
 458   void trigfunc(char trig, int num_fpu_regs_in_use = 1);
 459 
 460   // branch to L if FPU flag C2 is set/not set
 461   // tmp is a temporary register, if none is available use noreg
 462   void jC2 (Register tmp, Label& L);
 463   void jnC2(Register tmp, Label& L);
 464 
 465   // Pop ST (ffree & fincstp combined)
 466   void fpop();
 467 
 468   // Load float value from 'address'. If UseSSE >= 1, the value is loaded into
 469   // register xmm0. Otherwise, the value is loaded onto the FPU stack.
 470   void load_float(Address src);
 471 
 472   // Store float value to 'address'. If UseSSE >= 1, the value is stored
 473   // from register xmm0. Otherwise, the value is stored from the FPU stack.
 474   void store_float(Address dst);
 475 
 476   // Load double value from 'address'. If UseSSE >= 2, the value is loaded into
 477   // register xmm0. Otherwise, the value is loaded onto the FPU stack.
 478   void load_double(Address src);
 479 
 480   // Store double value to 'address'. If UseSSE >= 2, the value is stored
 481   // from register xmm0. Otherwise, the value is stored from the FPU stack.
 482   void store_double(Address dst);
 483 
 484   // pushes double TOS element of FPU stack on CPU stack; pops from FPU stack
 485   void push_fTOS();
 486 
 487   // pops double TOS element from CPU stack and pushes on FPU stack
 488   void pop_fTOS();
 489 
 490   void empty_FPU_stack();
 491 
 492   void push_IU_state();
 493   void pop_IU_state();
 494 
 495   void push_FPU_state();
 496   void pop_FPU_state();
 497 
 498   void push_CPU_state();
 499   void pop_CPU_state();
 500 
 501   // Round up to a power of two
 502   void round_to(Register reg, int modulus);
 503 
 504   // Callee saved registers handling
 505   void push_callee_saved_registers();
 506   void pop_callee_saved_registers();
 507 
 508   // allocation
 509   void eden_allocate(
 510     Register obj,                      // result: pointer to object after successful allocation
 511     Register var_size_in_bytes,        // object size in bytes if unknown at compile time; invalid otherwise
 512     int      con_size_in_bytes,        // object size in bytes if   known at compile time
 513     Register t1,                       // temp register
 514     Label&   slow_case                 // continuation point if fast allocation fails
 515   );
 516   void tlab_allocate(
 517     Register obj,                      // result: pointer to object after successful allocation
 518     Register var_size_in_bytes,        // object size in bytes if unknown at compile time; invalid otherwise
 519     int      con_size_in_bytes,        // object size in bytes if   known at compile time
 520     Register t1,                       // temp register
 521     Register t2,                       // temp register
 522     Label&   slow_case                 // continuation point if fast allocation fails
 523   );
 524   Register tlab_refill(Label& retry_tlab, Label& try_eden, Label& slow_case); // returns TLS address
 525   void zero_memory(Register address, Register length_in_bytes, int offset_in_bytes, Register temp);
 526 
 527   void incr_allocated_bytes(Register thread,
 528                             Register var_size_in_bytes, int con_size_in_bytes,
 529                             Register t1 = noreg);
 530 
 531   // interface method calling
 532   void lookup_interface_method(Register recv_klass,
 533                                Register intf_klass,
 534                                RegisterOrConstant itable_index,
 535                                Register method_result,
 536                                Register scan_temp,
 537                                Label& no_such_interface);
 538 
 539   // virtual method calling
 540   void lookup_virtual_method(Register recv_klass,
 541                              RegisterOrConstant vtable_index,
 542                              Register method_result);
 543 
 544   // Test sub_klass against super_klass, with fast and slow paths.
 545 
 546   // The fast path produces a tri-state answer: yes / no / maybe-slow.
 547   // One of the three labels can be NULL, meaning take the fall-through.
 548   // If super_check_offset is -1, the value is loaded up from super_klass.
 549   // No registers are killed, except temp_reg.
 550   void check_klass_subtype_fast_path(Register sub_klass,
 551                                      Register super_klass,
 552                                      Register temp_reg,
 553                                      Label* L_success,
 554                                      Label* L_failure,
 555                                      Label* L_slow_path,
 556                 RegisterOrConstant super_check_offset = RegisterOrConstant(-1));
 557 
 558   // The rest of the type check; must be wired to a corresponding fast path.
 559   // It does not repeat the fast path logic, so don't use it standalone.
 560   // The temp_reg and temp2_reg can be noreg, if no temps are available.
 561   // Updates the sub's secondary super cache as necessary.
 562   // If set_cond_codes, condition codes will be Z on success, NZ on failure.
 563   void check_klass_subtype_slow_path(Register sub_klass,
 564                                      Register super_klass,
 565                                      Register temp_reg,
 566                                      Register temp2_reg,
 567                                      Label* L_success,
 568                                      Label* L_failure,
 569                                      bool set_cond_codes = false);
 570 
 571   // Simplified, combined version, good for typical uses.
 572   // Falls through on failure.
 573   void check_klass_subtype(Register sub_klass,
 574                            Register super_klass,
 575                            Register temp_reg,
 576                            Label& L_success);
 577 
 578   // method handles (JSR 292)
 579   Address argument_address(RegisterOrConstant arg_slot, int extra_slot_offset = 0);
 580 
 581   //----
 582   void set_word_if_not_zero(Register reg); // sets reg to 1 if not zero, otherwise 0
 583 
 584   // Debugging
 585 
 586   // only if +VerifyOops
 587   // TODO: Make these macros with file and line like sparc version!
 588   void verify_oop(Register reg, const char* s = "broken oop");
 589   void verify_oop_addr(Address addr, const char * s = "broken oop addr");
 590 
 591   // TODO: verify method and klass metadata (compare against vptr?)
 592   void _verify_method_ptr(Register reg, const char * msg, const char * file, int line) {}
 593   void _verify_klass_ptr(Register reg, const char * msg, const char * file, int line){}
 594 
 595 #define verify_method_ptr(reg) _verify_method_ptr(reg, "broken method " #reg, __FILE__, __LINE__)
 596 #define verify_klass_ptr(reg) _verify_klass_ptr(reg, "broken klass " #reg, __FILE__, __LINE__)
 597 
 598   // only if +VerifyFPU
 599   void verify_FPU(int stack_depth, const char* s = "illegal FPU state");
 600 
 601   // Verify or restore cpu control state after JNI call
 602   void restore_cpu_control_state_after_jni();
 603 
 604   // prints msg, dumps registers and stops execution
 605   void stop(const char* msg);
 606 
 607   // prints msg and continues
 608   void warn(const char* msg);
 609 
 610   // dumps registers and other state
 611   void print_state();
 612 
 613   static void debug32(int rdi, int rsi, int rbp, int rsp, int rbx, int rdx, int rcx, int rax, int eip, char* msg);
 614   static void debug64(char* msg, int64_t pc, int64_t regs[]);
 615   static void print_state32(int rdi, int rsi, int rbp, int rsp, int rbx, int rdx, int rcx, int rax, int eip);
 616   static void print_state64(int64_t pc, int64_t regs[]);
 617 
 618   void os_breakpoint();
 619 
 620   void untested()                                { stop("untested"); }
 621 
 622   void unimplemented(const char* what = "")      { char* b = new char[1024];  jio_snprintf(b, 1024, "unimplemented: %s", what);  stop(b); }
 623 
 624   void should_not_reach_here()                   { stop("should not reach here"); }
 625 
 626   void print_CPU_state();
 627 
 628   // Stack overflow checking
 629   void bang_stack_with_offset(int offset) {
 630     // stack grows down, caller passes positive offset
 631     assert(offset > 0, "must bang with negative offset");
 632     movl(Address(rsp, (-offset)), rax);
 633   }
 634 
 635   // Writes to stack successive pages until offset reached to check for
 636   // stack overflow + shadow pages.  Also, clobbers tmp
 637   void bang_stack_size(Register size, Register tmp);
 638 
 639   // Check for reserved stack access in method being exited (for JIT)
 640   void reserved_stack_check();
 641 
 642   virtual RegisterOrConstant delayed_value_impl(intptr_t* delayed_value_addr,
 643                                                 Register tmp,
 644                                                 int offset);
 645 
 646   // Support for serializing memory accesses between threads
 647   void serialize_memory(Register thread, Register tmp);
 648 
 649   void verify_tlab();
 650 
 651   // Biased locking support
 652   // lock_reg and obj_reg must be loaded up with the appropriate values.
 653   // swap_reg must be rax, and is killed.
 654   // tmp_reg is optional. If it is supplied (i.e., != noreg) it will
 655   // be killed; if not supplied, push/pop will be used internally to
 656   // allocate a temporary (inefficient, avoid if possible).
 657   // Optional slow case is for implementations (interpreter and C1) which branch to
 658   // slow case directly. Leaves condition codes set for C2's Fast_Lock node.
 659   // Returns offset of first potentially-faulting instruction for null
 660   // check info (currently consumed only by C1). If
 661   // swap_reg_contains_mark is true then returns -1 as it is assumed
 662   // the calling code has already passed any potential faults.
 663   int biased_locking_enter(Register lock_reg, Register obj_reg,
 664                            Register swap_reg, Register tmp_reg,
 665                            bool swap_reg_contains_mark,
 666                            Label& done, Label* slow_case = NULL,
 667                            BiasedLockingCounters* counters = NULL);
 668   void biased_locking_exit (Register obj_reg, Register temp_reg, Label& done);
 669 #ifdef COMPILER2
 670   // Code used by cmpFastLock and cmpFastUnlock mach instructions in .ad file.
 671   // See full desription in macroAssembler_x86.cpp.
 672   void fast_lock(Register obj, Register box, Register tmp,
 673                  Register scr, Register cx1, Register cx2,
 674                  BiasedLockingCounters* counters,
 675                  RTMLockingCounters* rtm_counters,
 676                  RTMLockingCounters* stack_rtm_counters,
 677                  Metadata* method_data,
 678                  bool use_rtm, bool profile_rtm);
 679   void fast_unlock(Register obj, Register box, Register tmp, bool use_rtm);
 680 #if INCLUDE_RTM_OPT
 681   void rtm_counters_update(Register abort_status, Register rtm_counters);
 682   void branch_on_random_using_rdtsc(Register tmp, Register scr, int count, Label& brLabel);
 683   void rtm_abort_ratio_calculation(Register tmp, Register rtm_counters_reg,
 684                                    RTMLockingCounters* rtm_counters,
 685                                    Metadata* method_data);
 686   void rtm_profiling(Register abort_status_Reg, Register rtm_counters_Reg,
 687                      RTMLockingCounters* rtm_counters, Metadata* method_data, bool profile_rtm);
 688   void rtm_retry_lock_on_abort(Register retry_count, Register abort_status, Label& retryLabel);
 689   void rtm_retry_lock_on_busy(Register retry_count, Register box, Register tmp, Register scr, Label& retryLabel);
 690   void rtm_stack_locking(Register obj, Register tmp, Register scr,
 691                          Register retry_on_abort_count,
 692                          RTMLockingCounters* stack_rtm_counters,
 693                          Metadata* method_data, bool profile_rtm,
 694                          Label& DONE_LABEL, Label& IsInflated);
 695   void rtm_inflated_locking(Register obj, Register box, Register tmp,
 696                             Register scr, Register retry_on_busy_count,
 697                             Register retry_on_abort_count,
 698                             RTMLockingCounters* rtm_counters,
 699                             Metadata* method_data, bool profile_rtm,
 700                             Label& DONE_LABEL);
 701 #endif
 702 #endif
 703 
 704   Condition negate_condition(Condition cond);
 705 
 706   // Instructions that use AddressLiteral operands. These instruction can handle 32bit/64bit
 707   // operands. In general the names are modified to avoid hiding the instruction in Assembler
 708   // so that we don't need to implement all the varieties in the Assembler with trivial wrappers
 709   // here in MacroAssembler. The major exception to this rule is call
 710 
 711   // Arithmetics
 712 
 713 
 714   void addptr(Address dst, int32_t src) { LP64_ONLY(addq(dst, src)) NOT_LP64(addl(dst, src)) ; }
 715   void addptr(Address dst, Register src);
 716 
 717   void addptr(Register dst, Address src) { LP64_ONLY(addq(dst, src)) NOT_LP64(addl(dst, src)); }
 718   void addptr(Register dst, int32_t src);
 719   void addptr(Register dst, Register src);
 720   void addptr(Register dst, RegisterOrConstant src) {
 721     if (src.is_constant()) addptr(dst, (int) src.as_constant());
 722     else                   addptr(dst,       src.as_register());
 723   }
 724 
 725   void andptr(Register dst, int32_t src);
 726   void andptr(Register src1, Register src2) { LP64_ONLY(andq(src1, src2)) NOT_LP64(andl(src1, src2)) ; }
 727 
 728   void cmp8(AddressLiteral src1, int imm);
 729 
 730   // renamed to drag out the casting of address to int32_t/intptr_t
 731   void cmp32(Register src1, int32_t imm);
 732 
 733   void cmp32(AddressLiteral src1, int32_t imm);
 734   // compare reg - mem, or reg - &mem
 735   void cmp32(Register src1, AddressLiteral src2);
 736 
 737   void cmp32(Register src1, Address src2);
 738 
 739 #ifndef _LP64
 740   void cmpklass(Address dst, Metadata* obj);
 741   void cmpklass(Register dst, Metadata* obj);
 742   void cmpoop(Address dst, jobject obj);
 743   void cmpoop(Register dst, jobject obj);
 744 #endif // _LP64
 745 
 746   // NOTE src2 must be the lval. This is NOT an mem-mem compare
 747   void cmpptr(Address src1, AddressLiteral src2);
 748 
 749   void cmpptr(Register src1, AddressLiteral src2);
 750 
 751   void cmpptr(Register src1, Register src2) { LP64_ONLY(cmpq(src1, src2)) NOT_LP64(cmpl(src1, src2)) ; }
 752   void cmpptr(Register src1, Address src2) { LP64_ONLY(cmpq(src1, src2)) NOT_LP64(cmpl(src1, src2)) ; }
 753   // void cmpptr(Address src1, Register src2) { LP64_ONLY(cmpq(src1, src2)) NOT_LP64(cmpl(src1, src2)) ; }
 754 
 755   void cmpptr(Register src1, int32_t src2) { LP64_ONLY(cmpq(src1, src2)) NOT_LP64(cmpl(src1, src2)) ; }
 756   void cmpptr(Address src1, int32_t src2) { LP64_ONLY(cmpq(src1, src2)) NOT_LP64(cmpl(src1, src2)) ; }
 757 
 758   // cmp64 to avoild hiding cmpq
 759   void cmp64(Register src1, AddressLiteral src);
 760 
 761   void cmpxchgptr(Register reg, Address adr);
 762 
 763   void locked_cmpxchgptr(Register reg, AddressLiteral adr);
 764 
 765 
 766   void imulptr(Register dst, Register src) { LP64_ONLY(imulq(dst, src)) NOT_LP64(imull(dst, src)); }
 767   void imulptr(Register dst, Register src, int imm32) { LP64_ONLY(imulq(dst, src, imm32)) NOT_LP64(imull(dst, src, imm32)); }
 768 
 769 
 770   void negptr(Register dst) { LP64_ONLY(negq(dst)) NOT_LP64(negl(dst)); }
 771 
 772   void notptr(Register dst) { LP64_ONLY(notq(dst)) NOT_LP64(notl(dst)); }
 773 
 774   void shlptr(Register dst, int32_t shift);
 775   void shlptr(Register dst) { LP64_ONLY(shlq(dst)) NOT_LP64(shll(dst)); }
 776 
 777   void shrptr(Register dst, int32_t shift);
 778   void shrptr(Register dst) { LP64_ONLY(shrq(dst)) NOT_LP64(shrl(dst)); }
 779 
 780   void sarptr(Register dst) { LP64_ONLY(sarq(dst)) NOT_LP64(sarl(dst)); }
 781   void sarptr(Register dst, int32_t src) { LP64_ONLY(sarq(dst, src)) NOT_LP64(sarl(dst, src)); }
 782 
 783   void subptr(Address dst, int32_t src) { LP64_ONLY(subq(dst, src)) NOT_LP64(subl(dst, src)); }
 784 
 785   void subptr(Register dst, Address src) { LP64_ONLY(subq(dst, src)) NOT_LP64(subl(dst, src)); }
 786   void subptr(Register dst, int32_t src);
 787   // Force generation of a 4 byte immediate value even if it fits into 8bit
 788   void subptr_imm32(Register dst, int32_t src);
 789   void subptr(Register dst, Register src);
 790   void subptr(Register dst, RegisterOrConstant src) {
 791     if (src.is_constant()) subptr(dst, (int) src.as_constant());
 792     else                   subptr(dst,       src.as_register());
 793   }
 794 
 795   void sbbptr(Address dst, int32_t src) { LP64_ONLY(sbbq(dst, src)) NOT_LP64(sbbl(dst, src)); }
 796   void sbbptr(Register dst, int32_t src) { LP64_ONLY(sbbq(dst, src)) NOT_LP64(sbbl(dst, src)); }
 797 
 798   void xchgptr(Register src1, Register src2) { LP64_ONLY(xchgq(src1, src2)) NOT_LP64(xchgl(src1, src2)) ; }
 799   void xchgptr(Register src1, Address src2) { LP64_ONLY(xchgq(src1, src2)) NOT_LP64(xchgl(src1, src2)) ; }
 800 
 801   void xaddptr(Address src1, Register src2) { LP64_ONLY(xaddq(src1, src2)) NOT_LP64(xaddl(src1, src2)) ; }
 802 
 803 
 804 
 805   // Helper functions for statistics gathering.
 806   // Conditionally (atomically, on MPs) increments passed counter address, preserving condition codes.
 807   void cond_inc32(Condition cond, AddressLiteral counter_addr);
 808   // Unconditional atomic increment.
 809   void atomic_incl(Address counter_addr);
 810   void atomic_incl(AddressLiteral counter_addr, Register scr = rscratch1);
 811 #ifdef _LP64
 812   void atomic_incq(Address counter_addr);
 813   void atomic_incq(AddressLiteral counter_addr, Register scr = rscratch1);
 814 #endif
 815   void atomic_incptr(AddressLiteral counter_addr, Register scr = rscratch1) { LP64_ONLY(atomic_incq(counter_addr, scr)) NOT_LP64(atomic_incl(counter_addr, scr)) ; }
 816   void atomic_incptr(Address counter_addr) { LP64_ONLY(atomic_incq(counter_addr)) NOT_LP64(atomic_incl(counter_addr)) ; }
 817 
 818   void lea(Register dst, AddressLiteral adr);
 819   void lea(Address dst, AddressLiteral adr);
 820   void lea(Register dst, Address adr) { Assembler::lea(dst, adr); }
 821 
 822   void leal32(Register dst, Address src) { leal(dst, src); }
 823 
 824   // Import other testl() methods from the parent class or else
 825   // they will be hidden by the following overriding declaration.
 826   using Assembler::testl;
 827   void testl(Register dst, AddressLiteral src);
 828 
 829   void orptr(Register dst, Address src) { LP64_ONLY(orq(dst, src)) NOT_LP64(orl(dst, src)); }
 830   void orptr(Register dst, Register src) { LP64_ONLY(orq(dst, src)) NOT_LP64(orl(dst, src)); }
 831   void orptr(Register dst, int32_t src) { LP64_ONLY(orq(dst, src)) NOT_LP64(orl(dst, src)); }
 832   void orptr(Address dst, int32_t imm32) { LP64_ONLY(orq(dst, imm32)) NOT_LP64(orl(dst, imm32)); }
 833 
 834   void testptr(Register src, int32_t imm32) {  LP64_ONLY(testq(src, imm32)) NOT_LP64(testl(src, imm32)); }
 835   void testptr(Register src1, Register src2);
 836 
 837   void xorptr(Register dst, Register src) { LP64_ONLY(xorq(dst, src)) NOT_LP64(xorl(dst, src)); }
 838   void xorptr(Register dst, Address src) { LP64_ONLY(xorq(dst, src)) NOT_LP64(xorl(dst, src)); }
 839 
 840   // Calls
 841 
 842   void call(Label& L, relocInfo::relocType rtype);
 843   void call(Register entry);
 844 
 845   // NOTE: this call tranfers to the effective address of entry NOT
 846   // the address contained by entry. This is because this is more natural
 847   // for jumps/calls.
 848   void call(AddressLiteral entry);
 849 
 850   // Emit the CompiledIC call idiom
 851   void ic_call(address entry, jint method_index = 0);
 852 
 853   // Jumps
 854 
 855   // NOTE: these jumps tranfer to the effective address of dst NOT
 856   // the address contained by dst. This is because this is more natural
 857   // for jumps/calls.
 858   void jump(AddressLiteral dst);
 859   void jump_cc(Condition cc, AddressLiteral dst);
 860 
 861   // 32bit can do a case table jump in one instruction but we no longer allow the base
 862   // to be installed in the Address class. This jump will tranfers to the address
 863   // contained in the location described by entry (not the address of entry)
 864   void jump(ArrayAddress entry);
 865 
 866   // Floating
 867 
 868   void andpd(XMMRegister dst, Address src) { Assembler::andpd(dst, src); }
 869   void andpd(XMMRegister dst, AddressLiteral src);
 870   void andpd(XMMRegister dst, XMMRegister src) { Assembler::andpd(dst, src); }
 871 
 872   void andps(XMMRegister dst, XMMRegister src) { Assembler::andps(dst, src); }
 873   void andps(XMMRegister dst, Address src) { Assembler::andps(dst, src); }
 874   void andps(XMMRegister dst, AddressLiteral src);
 875 
 876   void comiss(XMMRegister dst, XMMRegister src) { Assembler::comiss(dst, src); }
 877   void comiss(XMMRegister dst, Address src) { Assembler::comiss(dst, src); }
 878   void comiss(XMMRegister dst, AddressLiteral src);
 879 
 880   void comisd(XMMRegister dst, XMMRegister src) { Assembler::comisd(dst, src); }
 881   void comisd(XMMRegister dst, Address src) { Assembler::comisd(dst, src); }
 882   void comisd(XMMRegister dst, AddressLiteral src);
 883 
 884   void fadd_s(Address src)        { Assembler::fadd_s(src); }
 885   void fadd_s(AddressLiteral src) { Assembler::fadd_s(as_Address(src)); }
 886 
 887   void fldcw(Address src) { Assembler::fldcw(src); }
 888   void fldcw(AddressLiteral src);
 889 
 890   void fld_s(int index)   { Assembler::fld_s(index); }
 891   void fld_s(Address src) { Assembler::fld_s(src); }
 892   void fld_s(AddressLiteral src);
 893 
 894   void fld_d(Address src) { Assembler::fld_d(src); }
 895   void fld_d(AddressLiteral src);
 896 
 897   void fld_x(Address src) { Assembler::fld_x(src); }
 898   void fld_x(AddressLiteral src);
 899 
 900   void fmul_s(Address src)        { Assembler::fmul_s(src); }
 901   void fmul_s(AddressLiteral src) { Assembler::fmul_s(as_Address(src)); }
 902 
 903   void ldmxcsr(Address src) { Assembler::ldmxcsr(src); }
 904   void ldmxcsr(AddressLiteral src);
 905 
 906   void fast_exp(XMMRegister xmm0, XMMRegister xmm1, XMMRegister xmm2, XMMRegister xmm3,
 907                 XMMRegister xmm4, XMMRegister xmm5, XMMRegister xmm6, XMMRegister xmm7,
 908                 Register rax, Register rcx, Register rdx, Register tmp);
 909 
 910   void fast_log(XMMRegister xmm0, XMMRegister xmm1, XMMRegister xmm2, XMMRegister xmm3,
 911                 XMMRegister xmm4, XMMRegister xmm5, XMMRegister xmm6, XMMRegister xmm7,
 912                 Register rax, Register rcx, Register rdx, Register tmp1 LP64_ONLY(COMMA Register tmp2));
 913 
 914   void fast_pow(XMMRegister xmm0, XMMRegister xmm1, XMMRegister xmm2, XMMRegister xmm3, XMMRegister xmm4,
 915                 XMMRegister xmm5, XMMRegister xmm6, XMMRegister xmm7, Register rax, Register rcx,
 916                 Register rdx NOT_LP64(COMMA  Register tmp) LP64_ONLY(COMMA  Register tmp1)
 917                 LP64_ONLY(COMMA  Register tmp2) LP64_ONLY(COMMA  Register tmp3) LP64_ONLY(COMMA  Register tmp4));
 918 
 919   void fast_sin(XMMRegister xmm0, XMMRegister xmm1, XMMRegister xmm2, XMMRegister xmm3,
 920                 XMMRegister xmm4, XMMRegister xmm5, XMMRegister xmm6, XMMRegister xmm7,
 921                 Register rax, Register rbx LP64_ONLY(COMMA  Register rcx), Register rdx
 922                 LP64_ONLY(COMMA Register tmp1) LP64_ONLY(COMMA Register tmp2)
 923                 LP64_ONLY(COMMA Register tmp3) LP64_ONLY(COMMA Register tmp4));
 924 
 925   void fast_cos(XMMRegister xmm0, XMMRegister xmm1, XMMRegister xmm2, XMMRegister xmm3,
 926                 XMMRegister xmm4, XMMRegister xmm5, XMMRegister xmm6, XMMRegister xmm7,
 927                 Register rax, Register rcx, Register rdx NOT_LP64(COMMA Register tmp)
 928                 LP64_ONLY(COMMA Register r8) LP64_ONLY(COMMA Register r9)
 929                 LP64_ONLY(COMMA Register r10) LP64_ONLY(COMMA Register r11));
 930 
 931 #ifndef _LP64
 932   void libm_sincos_huge(XMMRegister xmm0, XMMRegister xmm1, Register eax, Register ecx,
 933                         Register edx, Register ebx, Register esi, Register edi,
 934                         Register ebp, Register esp);
 935   void libm_reduce_pi04l(Register eax, Register ecx, Register edx, Register ebx,
 936                          Register esi, Register edi, Register ebp, Register esp);
 937 #endif
 938 
 939   void increase_precision();
 940   void restore_precision();
 941 
 942 private:
 943 
 944   // call runtime as a fallback for trig functions and pow/exp.
 945   void fp_runtime_fallback(address runtime_entry, int nb_args, int num_fpu_regs_in_use);
 946 
 947   // these are private because users should be doing movflt/movdbl
 948 
 949   void movss(Address dst, XMMRegister src)     { Assembler::movss(dst, src); }
 950   void movss(XMMRegister dst, XMMRegister src) { Assembler::movss(dst, src); }
 951   void movss(XMMRegister dst, Address src)     { Assembler::movss(dst, src); }
 952   void movss(XMMRegister dst, AddressLiteral src);
 953 
 954   void movlpd(XMMRegister dst, Address src)    {Assembler::movlpd(dst, src); }
 955   void movlpd(XMMRegister dst, AddressLiteral src);
 956 
 957 public:
 958 
 959   void addsd(XMMRegister dst, XMMRegister src)    { Assembler::addsd(dst, src); }
 960   void addsd(XMMRegister dst, Address src)        { Assembler::addsd(dst, src); }
 961   void addsd(XMMRegister dst, AddressLiteral src);
 962 
 963   void addss(XMMRegister dst, XMMRegister src)    { Assembler::addss(dst, src); }
 964   void addss(XMMRegister dst, Address src)        { Assembler::addss(dst, src); }
 965   void addss(XMMRegister dst, AddressLiteral src);
 966 
 967   void addpd(XMMRegister dst, XMMRegister src)    { Assembler::addpd(dst, src); }
 968   void addpd(XMMRegister dst, Address src)        { Assembler::addpd(dst, src); }
 969   void addpd(XMMRegister dst, AddressLiteral src);
 970 
 971   void divsd(XMMRegister dst, XMMRegister src)    { Assembler::divsd(dst, src); }
 972   void divsd(XMMRegister dst, Address src)        { Assembler::divsd(dst, src); }
 973   void divsd(XMMRegister dst, AddressLiteral src);
 974 
 975   void divss(XMMRegister dst, XMMRegister src)    { Assembler::divss(dst, src); }
 976   void divss(XMMRegister dst, Address src)        { Assembler::divss(dst, src); }
 977   void divss(XMMRegister dst, AddressLiteral src);
 978 
 979   // Move Unaligned Double Quadword
 980   void movdqu(Address     dst, XMMRegister src);
 981   void movdqu(XMMRegister dst, Address src);
 982   void movdqu(XMMRegister dst, XMMRegister src);
 983   void movdqu(XMMRegister dst, AddressLiteral src);
 984   // AVX Unaligned forms
 985   void vmovdqu(Address     dst, XMMRegister src);
 986   void vmovdqu(XMMRegister dst, Address src);
 987   void vmovdqu(XMMRegister dst, XMMRegister src);
 988   void vmovdqu(XMMRegister dst, AddressLiteral src);
 989 
 990   // Move Aligned Double Quadword
 991   void movdqa(XMMRegister dst, Address src)       { Assembler::movdqa(dst, src); }
 992   void movdqa(XMMRegister dst, XMMRegister src)   { Assembler::movdqa(dst, src); }
 993   void movdqa(XMMRegister dst, AddressLiteral src);
 994 
 995   void movsd(XMMRegister dst, XMMRegister src) { Assembler::movsd(dst, src); }
 996   void movsd(Address dst, XMMRegister src)     { Assembler::movsd(dst, src); }
 997   void movsd(XMMRegister dst, Address src)     { Assembler::movsd(dst, src); }
 998   void movsd(XMMRegister dst, AddressLiteral src);
 999 
1000   void mulpd(XMMRegister dst, XMMRegister src)    { Assembler::mulpd(dst, src); }
1001   void mulpd(XMMRegister dst, Address src)        { Assembler::mulpd(dst, src); }
1002   void mulpd(XMMRegister dst, AddressLiteral src);
1003 
1004   void mulsd(XMMRegister dst, XMMRegister src)    { Assembler::mulsd(dst, src); }
1005   void mulsd(XMMRegister dst, Address src)        { Assembler::mulsd(dst, src); }
1006   void mulsd(XMMRegister dst, AddressLiteral src);
1007 
1008   void mulss(XMMRegister dst, XMMRegister src)    { Assembler::mulss(dst, src); }
1009   void mulss(XMMRegister dst, Address src)        { Assembler::mulss(dst, src); }
1010   void mulss(XMMRegister dst, AddressLiteral src);
1011 
1012   // Carry-Less Multiplication Quadword
1013   void pclmulldq(XMMRegister dst, XMMRegister src) {
1014     // 0x00 - multiply lower 64 bits [0:63]
1015     Assembler::pclmulqdq(dst, src, 0x00);
1016   }
1017   void pclmulhdq(XMMRegister dst, XMMRegister src) {
1018     // 0x11 - multiply upper 64 bits [64:127]
1019     Assembler::pclmulqdq(dst, src, 0x11);
1020   }
1021 
1022   void pcmpeqb(XMMRegister dst, XMMRegister src);
1023   void pcmpeqw(XMMRegister dst, XMMRegister src);
1024 
1025   void pcmpestri(XMMRegister dst, Address src, int imm8);
1026   void pcmpestri(XMMRegister dst, XMMRegister src, int imm8);
1027 
1028   void pmovzxbw(XMMRegister dst, XMMRegister src);
1029   void pmovzxbw(XMMRegister dst, Address src);
1030 
1031   void pmovmskb(Register dst, XMMRegister src);
1032 
1033   void ptest(XMMRegister dst, XMMRegister src);
1034 
1035   void sqrtsd(XMMRegister dst, XMMRegister src)    { Assembler::sqrtsd(dst, src); }
1036   void sqrtsd(XMMRegister dst, Address src)        { Assembler::sqrtsd(dst, src); }
1037   void sqrtsd(XMMRegister dst, AddressLiteral src);
1038 
1039   void sqrtss(XMMRegister dst, XMMRegister src)    { Assembler::sqrtss(dst, src); }
1040   void sqrtss(XMMRegister dst, Address src)        { Assembler::sqrtss(dst, src); }
1041   void sqrtss(XMMRegister dst, AddressLiteral src);
1042 
1043   void subsd(XMMRegister dst, XMMRegister src)    { Assembler::subsd(dst, src); }
1044   void subsd(XMMRegister dst, Address src)        { Assembler::subsd(dst, src); }
1045   void subsd(XMMRegister dst, AddressLiteral src);
1046 
1047   void subss(XMMRegister dst, XMMRegister src)    { Assembler::subss(dst, src); }
1048   void subss(XMMRegister dst, Address src)        { Assembler::subss(dst, src); }
1049   void subss(XMMRegister dst, AddressLiteral src);
1050 
1051   void ucomiss(XMMRegister dst, XMMRegister src) { Assembler::ucomiss(dst, src); }
1052   void ucomiss(XMMRegister dst, Address src)     { Assembler::ucomiss(dst, src); }
1053   void ucomiss(XMMRegister dst, AddressLiteral src);
1054 
1055   void ucomisd(XMMRegister dst, XMMRegister src) { Assembler::ucomisd(dst, src); }
1056   void ucomisd(XMMRegister dst, Address src)     { Assembler::ucomisd(dst, src); }
1057   void ucomisd(XMMRegister dst, AddressLiteral src);
1058 
1059   // Bitwise Logical XOR of Packed Double-Precision Floating-Point Values
1060   void xorpd(XMMRegister dst, XMMRegister src);
1061   void xorpd(XMMRegister dst, Address src)     { Assembler::xorpd(dst, src); }
1062   void xorpd(XMMRegister dst, AddressLiteral src);
1063 
1064   // Bitwise Logical XOR of Packed Single-Precision Floating-Point Values
1065   void xorps(XMMRegister dst, XMMRegister src);
1066   void xorps(XMMRegister dst, Address src)     { Assembler::xorps(dst, src); }
1067   void xorps(XMMRegister dst, AddressLiteral src);
1068 
1069   // Shuffle Bytes
1070   void pshufb(XMMRegister dst, XMMRegister src) { Assembler::pshufb(dst, src); }
1071   void pshufb(XMMRegister dst, Address src)     { Assembler::pshufb(dst, src); }
1072   void pshufb(XMMRegister dst, AddressLiteral src);
1073   // AVX 3-operands instructions
1074 
1075   void vaddsd(XMMRegister dst, XMMRegister nds, XMMRegister src) { Assembler::vaddsd(dst, nds, src); }
1076   void vaddsd(XMMRegister dst, XMMRegister nds, Address src)     { Assembler::vaddsd(dst, nds, src); }
1077   void vaddsd(XMMRegister dst, XMMRegister nds, AddressLiteral src);
1078 
1079   void vaddss(XMMRegister dst, XMMRegister nds, XMMRegister src) { Assembler::vaddss(dst, nds, src); }
1080   void vaddss(XMMRegister dst, XMMRegister nds, Address src)     { Assembler::vaddss(dst, nds, src); }
1081   void vaddss(XMMRegister dst, XMMRegister nds, AddressLiteral src);
1082 
1083   void vabsss(XMMRegister dst, XMMRegister nds, XMMRegister src, AddressLiteral negate_field, int vector_len);
1084   void vabssd(XMMRegister dst, XMMRegister nds, XMMRegister src, AddressLiteral negate_field, int vector_len);
1085 
1086   void vpaddb(XMMRegister dst, XMMRegister nds, XMMRegister src, int vector_len);
1087   void vpaddb(XMMRegister dst, XMMRegister nds, Address src, int vector_len);
1088 
1089   void vpaddw(XMMRegister dst, XMMRegister nds, XMMRegister src, int vector_len);
1090   void vpaddw(XMMRegister dst, XMMRegister nds, Address src, int vector_len);
1091 
1092   void vpbroadcastw(XMMRegister dst, XMMRegister src);
1093 
1094   void vpcmpeqb(XMMRegister dst, XMMRegister nds, XMMRegister src, int vector_len);
1095   void vpcmpeqw(XMMRegister dst, XMMRegister nds, XMMRegister src, int vector_len);
1096 
1097   void vpmovzxbw(XMMRegister dst, Address src, int vector_len);
1098   void vpmovmskb(Register dst, XMMRegister src);
1099 
1100   void vpmullw(XMMRegister dst, XMMRegister nds, XMMRegister src, int vector_len);
1101   void vpmullw(XMMRegister dst, XMMRegister nds, Address src, int vector_len);
1102 
1103   void vpsubb(XMMRegister dst, XMMRegister nds, XMMRegister src, int vector_len);
1104   void vpsubb(XMMRegister dst, XMMRegister nds, Address src, int vector_len);
1105 
1106   void vpsubw(XMMRegister dst, XMMRegister nds, XMMRegister src, int vector_len);
1107   void vpsubw(XMMRegister dst, XMMRegister nds, Address src, int vector_len);
1108 
1109   void vpsraw(XMMRegister dst, XMMRegister nds, XMMRegister shift, int vector_len);
1110   void vpsraw(XMMRegister dst, XMMRegister nds, int shift, int vector_len);
1111 
1112   void vpsrlw(XMMRegister dst, XMMRegister nds, XMMRegister shift, int vector_len);
1113   void vpsrlw(XMMRegister dst, XMMRegister nds, int shift, int vector_len);
1114 
1115   void vpsllw(XMMRegister dst, XMMRegister nds, XMMRegister shift, int vector_len);
1116   void vpsllw(XMMRegister dst, XMMRegister nds, int shift, int vector_len);
1117 
1118   void vptest(XMMRegister dst, XMMRegister src);
1119 
1120   void punpcklbw(XMMRegister dst, XMMRegister src);
1121   void punpcklbw(XMMRegister dst, Address src) { Assembler::punpcklbw(dst, src); }
1122 
1123   void pshuflw(XMMRegister dst, XMMRegister src, int mode);
1124   void pshuflw(XMMRegister dst, Address src, int mode) { Assembler::pshuflw(dst, src, mode); }
1125 
1126   void vandpd(XMMRegister dst, XMMRegister nds, XMMRegister src, int vector_len) { Assembler::vandpd(dst, nds, src, vector_len); }
1127   void vandpd(XMMRegister dst, XMMRegister nds, Address src, int vector_len)     { Assembler::vandpd(dst, nds, src, vector_len); }
1128   void vandpd(XMMRegister dst, XMMRegister nds, AddressLiteral src, int vector_len);
1129 
1130   void vandps(XMMRegister dst, XMMRegister nds, XMMRegister src, int vector_len) { Assembler::vandps(dst, nds, src, vector_len); }
1131   void vandps(XMMRegister dst, XMMRegister nds, Address src, int vector_len)     { Assembler::vandps(dst, nds, src, vector_len); }
1132   void vandps(XMMRegister dst, XMMRegister nds, AddressLiteral src, int vector_len);
1133 
1134   void vdivsd(XMMRegister dst, XMMRegister nds, XMMRegister src) { Assembler::vdivsd(dst, nds, src); }
1135   void vdivsd(XMMRegister dst, XMMRegister nds, Address src)     { Assembler::vdivsd(dst, nds, src); }
1136   void vdivsd(XMMRegister dst, XMMRegister nds, AddressLiteral src);
1137 
1138   void vdivss(XMMRegister dst, XMMRegister nds, XMMRegister src) { Assembler::vdivss(dst, nds, src); }
1139   void vdivss(XMMRegister dst, XMMRegister nds, Address src)     { Assembler::vdivss(dst, nds, src); }
1140   void vdivss(XMMRegister dst, XMMRegister nds, AddressLiteral src);
1141 
1142   void vmulsd(XMMRegister dst, XMMRegister nds, XMMRegister src) { Assembler::vmulsd(dst, nds, src); }
1143   void vmulsd(XMMRegister dst, XMMRegister nds, Address src)     { Assembler::vmulsd(dst, nds, src); }
1144   void vmulsd(XMMRegister dst, XMMRegister nds, AddressLiteral src);
1145 
1146   void vmulss(XMMRegister dst, XMMRegister nds, XMMRegister src) { Assembler::vmulss(dst, nds, src); }
1147   void vmulss(XMMRegister dst, XMMRegister nds, Address src)     { Assembler::vmulss(dst, nds, src); }
1148   void vmulss(XMMRegister dst, XMMRegister nds, AddressLiteral src);
1149 
1150   void vsubsd(XMMRegister dst, XMMRegister nds, XMMRegister src) { Assembler::vsubsd(dst, nds, src); }
1151   void vsubsd(XMMRegister dst, XMMRegister nds, Address src)     { Assembler::vsubsd(dst, nds, src); }
1152   void vsubsd(XMMRegister dst, XMMRegister nds, AddressLiteral src);
1153 
1154   void vsubss(XMMRegister dst, XMMRegister nds, XMMRegister src) { Assembler::vsubss(dst, nds, src); }
1155   void vsubss(XMMRegister dst, XMMRegister nds, Address src)     { Assembler::vsubss(dst, nds, src); }
1156   void vsubss(XMMRegister dst, XMMRegister nds, AddressLiteral src);
1157 
1158   void vnegatess(XMMRegister dst, XMMRegister nds, AddressLiteral src);
1159   void vnegatesd(XMMRegister dst, XMMRegister nds, AddressLiteral src);
1160 
1161   // AVX Vector instructions
1162 
1163   void vxorpd(XMMRegister dst, XMMRegister nds, XMMRegister src, int vector_len) { Assembler::vxorpd(dst, nds, src, vector_len); }
1164   void vxorpd(XMMRegister dst, XMMRegister nds, Address src, int vector_len) { Assembler::vxorpd(dst, nds, src, vector_len); }
1165   void vxorpd(XMMRegister dst, XMMRegister nds, AddressLiteral src, int vector_len);
1166 
1167   void vxorps(XMMRegister dst, XMMRegister nds, XMMRegister src, int vector_len) { Assembler::vxorps(dst, nds, src, vector_len); }
1168   void vxorps(XMMRegister dst, XMMRegister nds, Address src, int vector_len) { Assembler::vxorps(dst, nds, src, vector_len); }
1169   void vxorps(XMMRegister dst, XMMRegister nds, AddressLiteral src, int vector_len);
1170 
1171   void vpxor(XMMRegister dst, XMMRegister nds, XMMRegister src, int vector_len) {
1172     if (UseAVX > 1 || (vector_len < 1)) // vpxor 256 bit is available only in AVX2
1173       Assembler::vpxor(dst, nds, src, vector_len);
1174     else
1175       Assembler::vxorpd(dst, nds, src, vector_len);
1176   }
1177   void vpxor(XMMRegister dst, XMMRegister nds, Address src, int vector_len) {
1178     if (UseAVX > 1 || (vector_len < 1)) // vpxor 256 bit is available only in AVX2
1179       Assembler::vpxor(dst, nds, src, vector_len);
1180     else
1181       Assembler::vxorpd(dst, nds, src, vector_len);
1182   }
1183 
1184   // Simple version for AVX2 256bit vectors
1185   void vpxor(XMMRegister dst, XMMRegister src) { Assembler::vpxor(dst, dst, src, true); }
1186   void vpxor(XMMRegister dst, Address src) { Assembler::vpxor(dst, dst, src, true); }
1187 
1188   // Move packed integer values from low 128 bit to hign 128 bit in 256 bit vector.
1189   void vinserti128h(XMMRegister dst, XMMRegister nds, XMMRegister src) {
1190     if (UseAVX > 1) // vinserti128h is available only in AVX2
1191       Assembler::vinserti128h(dst, nds, src);
1192     else
1193       Assembler::vinsertf128h(dst, nds, src);
1194   }
1195 
1196   // Carry-Less Multiplication Quadword
1197   void vpclmulldq(XMMRegister dst, XMMRegister nds, XMMRegister src) {
1198     // 0x00 - multiply lower 64 bits [0:63]
1199     Assembler::vpclmulqdq(dst, nds, src, 0x00);
1200   }
1201   void vpclmulhdq(XMMRegister dst, XMMRegister nds, XMMRegister src) {
1202     // 0x11 - multiply upper 64 bits [64:127]
1203     Assembler::vpclmulqdq(dst, nds, src, 0x11);
1204   }
1205 
1206   // Data
1207 
1208   void cmov32( Condition cc, Register dst, Address  src);
1209   void cmov32( Condition cc, Register dst, Register src);
1210 
1211   void cmov(   Condition cc, Register dst, Register src) { cmovptr(cc, dst, src); }
1212 
1213   void cmovptr(Condition cc, Register dst, Address  src) { LP64_ONLY(cmovq(cc, dst, src)) NOT_LP64(cmov32(cc, dst, src)); }
1214   void cmovptr(Condition cc, Register dst, Register src) { LP64_ONLY(cmovq(cc, dst, src)) NOT_LP64(cmov32(cc, dst, src)); }
1215 
1216   void movoop(Register dst, jobject obj);
1217   void movoop(Address dst, jobject obj);
1218 
1219   void mov_metadata(Register dst, Metadata* obj);
1220   void mov_metadata(Address dst, Metadata* obj);
1221 
1222   void movptr(ArrayAddress dst, Register src);
1223   // can this do an lea?
1224   void movptr(Register dst, ArrayAddress src);
1225 
1226   void movptr(Register dst, Address src);
1227 
1228 #ifdef _LP64
1229   void movptr(Register dst, AddressLiteral src, Register scratch=rscratch1);
1230 #else
1231   void movptr(Register dst, AddressLiteral src, Register scratch=noreg); // Scratch reg is ignored in 32-bit
1232 #endif
1233 
1234   void movptr(Register dst, intptr_t src);
1235   void movptr(Register dst, Register src);
1236   void movptr(Address dst, intptr_t src);
1237 
1238   void movptr(Address dst, Register src);
1239 
1240   void movptr(Register dst, RegisterOrConstant src) {
1241     if (src.is_constant()) movptr(dst, src.as_constant());
1242     else                   movptr(dst, src.as_register());
1243   }
1244 
1245 #ifdef _LP64
1246   // Generally the next two are only used for moving NULL
1247   // Although there are situations in initializing the mark word where
1248   // they could be used. They are dangerous.
1249 
1250   // They only exist on LP64 so that int32_t and intptr_t are not the same
1251   // and we have ambiguous declarations.
1252 
1253   void movptr(Address dst, int32_t imm32);
1254   void movptr(Register dst, int32_t imm32);
1255 #endif // _LP64
1256 
1257   // to avoid hiding movl
1258   void mov32(AddressLiteral dst, Register src);
1259   void mov32(Register dst, AddressLiteral src);
1260 
1261   // to avoid hiding movb
1262   void movbyte(ArrayAddress dst, int src);
1263 
1264   // Import other mov() methods from the parent class or else
1265   // they will be hidden by the following overriding declaration.
1266   using Assembler::movdl;
1267   using Assembler::movq;
1268   void movdl(XMMRegister dst, AddressLiteral src);
1269   void movq(XMMRegister dst, AddressLiteral src);
1270 
1271   // Can push value or effective address
1272   void pushptr(AddressLiteral src);
1273 
1274   void pushptr(Address src) { LP64_ONLY(pushq(src)) NOT_LP64(pushl(src)); }
1275   void popptr(Address src) { LP64_ONLY(popq(src)) NOT_LP64(popl(src)); }
1276 
1277   void pushoop(jobject obj);
1278   void pushklass(Metadata* obj);
1279 
1280   // sign extend as need a l to ptr sized element
1281   void movl2ptr(Register dst, Address src) { LP64_ONLY(movslq(dst, src)) NOT_LP64(movl(dst, src)); }
1282   void movl2ptr(Register dst, Register src) { LP64_ONLY(movslq(dst, src)) NOT_LP64(if (dst != src) movl(dst, src)); }
1283 
1284   // C2 compiled method's prolog code.
1285   void verified_entry(int framesize, int stack_bang_size, bool fp_mode_24b);
1286 
1287   // clear memory of size 'cnt' qwords, starting at 'base'.
1288   void clear_mem(Register base, Register cnt, Register rtmp);
1289 
1290 #ifdef COMPILER2
1291   void string_indexof_char(Register str1, Register cnt1, Register ch, Register result,
1292                            XMMRegister vec1, XMMRegister vec2, XMMRegister vec3, Register tmp);
1293 
1294   // IndexOf strings.
1295   // Small strings are loaded through stack if they cross page boundary.
1296   void string_indexof(Register str1, Register str2,
1297                       Register cnt1, Register cnt2,
1298                       int int_cnt2,  Register result,
1299                       XMMRegister vec, Register tmp,
1300                       int ae);
1301 
1302   // IndexOf for constant substrings with size >= 8 elements
1303   // which don't need to be loaded through stack.
1304   void string_indexofC8(Register str1, Register str2,
1305                       Register cnt1, Register cnt2,
1306                       int int_cnt2,  Register result,
1307                       XMMRegister vec, Register tmp,
1308                       int ae);
1309 
1310     // Smallest code: we don't need to load through stack,
1311     // check string tail.
1312 
1313   // helper function for string_compare
1314   void load_next_elements(Register elem1, Register elem2, Register str1, Register str2,
1315                           Address::ScaleFactor scale, Address::ScaleFactor scale1,
1316                           Address::ScaleFactor scale2, Register index, int ae);
1317   // Compare strings.
1318   void string_compare(Register str1, Register str2,
1319                       Register cnt1, Register cnt2, Register result,
1320                       XMMRegister vec1, int ae);
1321 
1322   // Search for Non-ASCII character (Negative byte value) in a byte array,
1323   // return true if it has any and false otherwise.
1324   void has_negatives(Register ary1, Register len,
1325                      Register result, Register tmp1,
1326                      XMMRegister vec1, XMMRegister vec2);
1327 
1328   // Compare char[] or byte[] arrays.
1329   void arrays_equals(bool is_array_equ, Register ary1, Register ary2,
1330                      Register limit, Register result, Register chr,
1331                      XMMRegister vec1, XMMRegister vec2, bool is_char);
1332 
1333 #endif
1334 
1335   // Fill primitive arrays
1336   void generate_fill(BasicType t, bool aligned,
1337                      Register to, Register value, Register count,
1338                      Register rtmp, XMMRegister xtmp);
1339 
1340   void encode_iso_array(Register src, Register dst, Register len,
1341                         XMMRegister tmp1, XMMRegister tmp2, XMMRegister tmp3,
1342                         XMMRegister tmp4, Register tmp5, Register result);
1343 
1344 #ifdef _LP64
1345   void add2_with_carry(Register dest_hi, Register dest_lo, Register src1, Register src2);
1346   void multiply_64_x_64_loop(Register x, Register xstart, Register x_xstart,
1347                              Register y, Register y_idx, Register z,
1348                              Register carry, Register product,
1349                              Register idx, Register kdx);
1350   void multiply_add_128_x_128(Register x_xstart, Register y, Register z,
1351                               Register yz_idx, Register idx,
1352                               Register carry, Register product, int offset);
1353   void multiply_128_x_128_bmi2_loop(Register y, Register z,
1354                                     Register carry, Register carry2,
1355                                     Register idx, Register jdx,
1356                                     Register yz_idx1, Register yz_idx2,
1357                                     Register tmp, Register tmp3, Register tmp4);
1358   void multiply_128_x_128_loop(Register x_xstart, Register y, Register z,
1359                                Register yz_idx, Register idx, Register jdx,
1360                                Register carry, Register product,
1361                                Register carry2);
1362   void multiply_to_len(Register x, Register xlen, Register y, Register ylen, Register z, Register zlen,
1363                        Register tmp1, Register tmp2, Register tmp3, Register tmp4, Register tmp5);
1364   void square_rshift(Register x, Register len, Register z, Register tmp1, Register tmp3,
1365                      Register tmp4, Register tmp5, Register rdxReg, Register raxReg);
1366   void multiply_add_64_bmi2(Register sum, Register op1, Register op2, Register carry,
1367                             Register tmp2);
1368   void multiply_add_64(Register sum, Register op1, Register op2, Register carry,
1369                        Register rdxReg, Register raxReg);
1370   void add_one_64(Register z, Register zlen, Register carry, Register tmp1);
1371   void lshift_by_1(Register x, Register len, Register z, Register zlen, Register tmp1, Register tmp2,
1372                        Register tmp3, Register tmp4);
1373   void square_to_len(Register x, Register len, Register z, Register zlen, Register tmp1, Register tmp2,
1374                      Register tmp3, Register tmp4, Register tmp5, Register rdxReg, Register raxReg);
1375 
1376   void mul_add_128_x_32_loop(Register out, Register in, Register offset, Register len, Register tmp1,
1377                Register tmp2, Register tmp3, Register tmp4, Register tmp5, Register rdxReg,
1378                Register raxReg);
1379   void mul_add(Register out, Register in, Register offset, Register len, Register k, Register tmp1,
1380                Register tmp2, Register tmp3, Register tmp4, Register tmp5, Register rdxReg,
1381                Register raxReg);
1382   void vectorized_mismatch(Register obja, Register objb, Register length, Register log2_array_indxscale,
1383                            Register result, Register tmp1, Register tmp2,
1384                            XMMRegister vec1, XMMRegister vec2, XMMRegister vec3);
1385 #endif
1386 
1387   // CRC32 code for java.util.zip.CRC32::updateBytes() intrinsic.
1388   void update_byte_crc32(Register crc, Register val, Register table);
1389   void kernel_crc32(Register crc, Register buf, Register len, Register table, Register tmp);
1390   // CRC32C code for java.util.zip.CRC32C::updateBytes() intrinsic
1391   // Note on a naming convention:
1392   // Prefix w = register only used on a Westmere+ architecture
1393   // Prefix n = register only used on a Nehalem architecture
1394 #ifdef _LP64
1395   void crc32c_ipl_alg4(Register in_out, uint32_t n,
1396                        Register tmp1, Register tmp2, Register tmp3);
1397 #else
1398   void crc32c_ipl_alg4(Register in_out, uint32_t n,
1399                        Register tmp1, Register tmp2, Register tmp3,
1400                        XMMRegister xtmp1, XMMRegister xtmp2);
1401 #endif
1402   void crc32c_pclmulqdq(XMMRegister w_xtmp1,
1403                         Register in_out,
1404                         uint32_t const_or_pre_comp_const_index, bool is_pclmulqdq_supported,
1405                         XMMRegister w_xtmp2,
1406                         Register tmp1,
1407                         Register n_tmp2, Register n_tmp3);
1408   void crc32c_rec_alt2(uint32_t const_or_pre_comp_const_index_u1, uint32_t const_or_pre_comp_const_index_u2, bool is_pclmulqdq_supported, Register in_out, Register in1, Register in2,
1409                        XMMRegister w_xtmp1, XMMRegister w_xtmp2, XMMRegister w_xtmp3,
1410                        Register tmp1, Register tmp2,
1411                        Register n_tmp3);
1412   void crc32c_proc_chunk(uint32_t size, uint32_t const_or_pre_comp_const_index_u1, uint32_t const_or_pre_comp_const_index_u2, bool is_pclmulqdq_supported,
1413                          Register in_out1, Register in_out2, Register in_out3,
1414                          Register tmp1, Register tmp2, Register tmp3,
1415                          XMMRegister w_xtmp1, XMMRegister w_xtmp2, XMMRegister w_xtmp3,
1416                          Register tmp4, Register tmp5,
1417                          Register n_tmp6);
1418   void crc32c_ipl_alg2_alt2(Register in_out, Register in1, Register in2,
1419                             Register tmp1, Register tmp2, Register tmp3,
1420                             Register tmp4, Register tmp5, Register tmp6,
1421                             XMMRegister w_xtmp1, XMMRegister w_xtmp2, XMMRegister w_xtmp3,
1422                             bool is_pclmulqdq_supported);
1423   // Fold 128-bit data chunk
1424   void fold_128bit_crc32(XMMRegister xcrc, XMMRegister xK, XMMRegister xtmp, Register buf, int offset);
1425   void fold_128bit_crc32(XMMRegister xcrc, XMMRegister xK, XMMRegister xtmp, XMMRegister xbuf);
1426   // Fold 8-bit data
1427   void fold_8bit_crc32(Register crc, Register table, Register tmp);
1428   void fold_8bit_crc32(XMMRegister crc, Register table, XMMRegister xtmp, Register tmp);
1429 
1430   // Compress char[] array to byte[].
1431   void char_array_compress(Register src, Register dst, Register len,
1432                            XMMRegister tmp1, XMMRegister tmp2, XMMRegister tmp3,
1433                            XMMRegister tmp4, Register tmp5, Register result);
1434 
1435   // Inflate byte[] array to char[].
1436   void byte_array_inflate(Register src, Register dst, Register len,
1437                           XMMRegister tmp1, Register tmp2);
1438 
1439 };
1440 
1441 /**
1442  * class SkipIfEqual:
1443  *
1444  * Instantiating this class will result in assembly code being output that will
1445  * jump around any code emitted between the creation of the instance and it's
1446  * automatic destruction at the end of a scope block, depending on the value of
1447  * the flag passed to the constructor, which will be checked at run-time.
1448  */
1449 class SkipIfEqual {
1450  private:
1451   MacroAssembler* _masm;
1452   Label _label;
1453 
1454  public:
1455    SkipIfEqual(MacroAssembler*, const bool* flag_addr, bool value);
1456    ~SkipIfEqual();
1457 };
1458 
1459 #endif // CPU_X86_VM_MACROASSEMBLER_X86_HPP