Thu, 12 Jun 2008 13:50:55 -0700
Merge
1 /*
2 * Copyright 2003-2007 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 */
25 class BiasedLockingCounters;
27 // Contains all the definitions needed for amd64 assembly code generation.
29 #ifdef _LP64
30 // Calling convention
31 class Argument VALUE_OBJ_CLASS_SPEC {
32 public:
33 enum {
34 #ifdef _WIN64
35 n_int_register_parameters_c = 4, // rcx, rdx, r8, r9 (c_rarg0, c_rarg1, ...)
36 n_float_register_parameters_c = 4, // xmm0 - xmm3 (c_farg0, c_farg1, ... )
37 #else
38 n_int_register_parameters_c = 6, // rdi, rsi, rdx, rcx, r8, r9 (c_rarg0, c_rarg1, ...)
39 n_float_register_parameters_c = 8, // xmm0 - xmm7 (c_farg0, c_farg1, ... )
40 #endif // _WIN64
41 n_int_register_parameters_j = 6, // j_rarg0, j_rarg1, ...
42 n_float_register_parameters_j = 8 // j_farg0, j_farg1, ...
43 };
44 };
47 // Symbolically name the register arguments used by the c calling convention.
48 // Windows is different from linux/solaris. So much for standards...
50 #ifdef _WIN64
52 REGISTER_DECLARATION(Register, c_rarg0, rcx);
53 REGISTER_DECLARATION(Register, c_rarg1, rdx);
54 REGISTER_DECLARATION(Register, c_rarg2, r8);
55 REGISTER_DECLARATION(Register, c_rarg3, r9);
57 REGISTER_DECLARATION(XMMRegister, c_farg0, xmm0);
58 REGISTER_DECLARATION(XMMRegister, c_farg1, xmm1);
59 REGISTER_DECLARATION(XMMRegister, c_farg2, xmm2);
60 REGISTER_DECLARATION(XMMRegister, c_farg3, xmm3);
62 #else
64 REGISTER_DECLARATION(Register, c_rarg0, rdi);
65 REGISTER_DECLARATION(Register, c_rarg1, rsi);
66 REGISTER_DECLARATION(Register, c_rarg2, rdx);
67 REGISTER_DECLARATION(Register, c_rarg3, rcx);
68 REGISTER_DECLARATION(Register, c_rarg4, r8);
69 REGISTER_DECLARATION(Register, c_rarg5, r9);
71 REGISTER_DECLARATION(XMMRegister, c_farg0, xmm0);
72 REGISTER_DECLARATION(XMMRegister, c_farg1, xmm1);
73 REGISTER_DECLARATION(XMMRegister, c_farg2, xmm2);
74 REGISTER_DECLARATION(XMMRegister, c_farg3, xmm3);
75 REGISTER_DECLARATION(XMMRegister, c_farg4, xmm4);
76 REGISTER_DECLARATION(XMMRegister, c_farg5, xmm5);
77 REGISTER_DECLARATION(XMMRegister, c_farg6, xmm6);
78 REGISTER_DECLARATION(XMMRegister, c_farg7, xmm7);
80 #endif // _WIN64
82 // Symbolically name the register arguments used by the Java calling convention.
83 // We have control over the convention for java so we can do what we please.
84 // What pleases us is to offset the java calling convention so that when
85 // we call a suitable jni method the arguments are lined up and we don't
86 // have to do little shuffling. A suitable jni method is non-static and a
87 // small number of arguments (two fewer args on windows)
88 //
89 // |-------------------------------------------------------|
90 // | c_rarg0 c_rarg1 c_rarg2 c_rarg3 c_rarg4 c_rarg5 |
91 // |-------------------------------------------------------|
92 // | rcx rdx r8 r9 rdi* rsi* | windows (* not a c_rarg)
93 // | rdi rsi rdx rcx r8 r9 | solaris/linux
94 // |-------------------------------------------------------|
95 // | j_rarg5 j_rarg0 j_rarg1 j_rarg2 j_rarg3 j_rarg4 |
96 // |-------------------------------------------------------|
98 REGISTER_DECLARATION(Register, j_rarg0, c_rarg1);
99 REGISTER_DECLARATION(Register, j_rarg1, c_rarg2);
100 REGISTER_DECLARATION(Register, j_rarg2, c_rarg3);
101 // Windows runs out of register args here
102 #ifdef _WIN64
103 REGISTER_DECLARATION(Register, j_rarg3, rdi);
104 REGISTER_DECLARATION(Register, j_rarg4, rsi);
105 #else
106 REGISTER_DECLARATION(Register, j_rarg3, c_rarg4);
107 REGISTER_DECLARATION(Register, j_rarg4, c_rarg5);
108 #endif // _WIN64
109 REGISTER_DECLARATION(Register, j_rarg5, c_rarg0);
111 REGISTER_DECLARATION(XMMRegister, j_farg0, xmm0);
112 REGISTER_DECLARATION(XMMRegister, j_farg1, xmm1);
113 REGISTER_DECLARATION(XMMRegister, j_farg2, xmm2);
114 REGISTER_DECLARATION(XMMRegister, j_farg3, xmm3);
115 REGISTER_DECLARATION(XMMRegister, j_farg4, xmm4);
116 REGISTER_DECLARATION(XMMRegister, j_farg5, xmm5);
117 REGISTER_DECLARATION(XMMRegister, j_farg6, xmm6);
118 REGISTER_DECLARATION(XMMRegister, j_farg7, xmm7);
120 REGISTER_DECLARATION(Register, rscratch1, r10); // volatile
121 REGISTER_DECLARATION(Register, rscratch2, r11); // volatile
123 REGISTER_DECLARATION(Register, r12_heapbase, r12); // callee-saved
124 REGISTER_DECLARATION(Register, r15_thread, r15); // callee-saved
126 #endif // _LP64
128 // Address is an abstraction used to represent a memory location
129 // using any of the amd64 addressing modes with one object.
130 //
131 // Note: A register location is represented via a Register, not
132 // via an address for efficiency & simplicity reasons.
134 class ArrayAddress;
136 class Address VALUE_OBJ_CLASS_SPEC {
137 public:
138 enum ScaleFactor {
139 no_scale = -1,
140 times_1 = 0,
141 times_2 = 1,
142 times_4 = 2,
143 times_8 = 3
144 };
146 private:
147 Register _base;
148 Register _index;
149 ScaleFactor _scale;
150 int _disp;
151 RelocationHolder _rspec;
153 // Easily misused constructors make them private
154 Address(int disp, address loc, relocInfo::relocType rtype);
155 Address(int disp, address loc, RelocationHolder spec);
157 public:
158 // creation
159 Address()
160 : _base(noreg),
161 _index(noreg),
162 _scale(no_scale),
163 _disp(0) {
164 }
166 // No default displacement otherwise Register can be implicitly
167 // converted to 0(Register) which is quite a different animal.
169 Address(Register base, int disp)
170 : _base(base),
171 _index(noreg),
172 _scale(no_scale),
173 _disp(disp) {
174 }
176 Address(Register base, Register index, ScaleFactor scale, int disp = 0)
177 : _base (base),
178 _index(index),
179 _scale(scale),
180 _disp (disp) {
181 assert(!index->is_valid() == (scale == Address::no_scale),
182 "inconsistent address");
183 }
185 // The following two overloads are used in connection with the
186 // ByteSize type (see sizes.hpp). They simplify the use of
187 // ByteSize'd arguments in assembly code. Note that their equivalent
188 // for the optimized build are the member functions with int disp
189 // argument since ByteSize is mapped to an int type in that case.
190 //
191 // Note: DO NOT introduce similar overloaded functions for WordSize
192 // arguments as in the optimized mode, both ByteSize and WordSize
193 // are mapped to the same type and thus the compiler cannot make a
194 // distinction anymore (=> compiler errors).
196 #ifdef ASSERT
197 Address(Register base, ByteSize disp)
198 : _base(base),
199 _index(noreg),
200 _scale(no_scale),
201 _disp(in_bytes(disp)) {
202 }
204 Address(Register base, Register index, ScaleFactor scale, ByteSize disp)
205 : _base(base),
206 _index(index),
207 _scale(scale),
208 _disp(in_bytes(disp)) {
209 assert(!index->is_valid() == (scale == Address::no_scale),
210 "inconsistent address");
211 }
212 #endif // ASSERT
214 // accessors
215 bool uses(Register reg) const {
216 return _base == reg || _index == reg;
217 }
219 // Convert the raw encoding form into the form expected by the constructor for
220 // Address. An index of 4 (rsp) corresponds to having no index, so convert
221 // that to noreg for the Address constructor.
222 static Address make_raw(int base, int index, int scale, int disp);
224 static Address make_array(ArrayAddress);
225 Register base() const {
226 return _base;
227 }
229 Register index() const {
230 return _index;
231 }
233 int disp() const {
234 return _disp;
235 }
238 private:
239 bool base_needs_rex() const {
240 return _base != noreg && _base->encoding() >= 8;
241 }
243 bool index_needs_rex() const {
244 return _index != noreg &&_index->encoding() >= 8;
245 }
247 relocInfo::relocType reloc() const { return _rspec.type(); }
249 friend class Assembler;
250 friend class MacroAssembler;
251 friend class LIR_Assembler; // base/index/scale/disp
252 };
254 //
255 // AddressLiteral has been split out from Address because operands of this type
256 // need to be treated specially on 32bit vs. 64bit platforms. By splitting it out
257 // the few instructions that need to deal with address literals are unique and the
258 // MacroAssembler does not have to implement every instruction in the Assembler
259 // in order to search for address literals that may need special handling depending
260 // on the instruction and the platform. As small step on the way to merging i486/amd64
261 // directories.
262 //
263 class AddressLiteral VALUE_OBJ_CLASS_SPEC {
264 friend class ArrayAddress;
265 RelocationHolder _rspec;
266 // Typically we use AddressLiterals we want to use their rval
267 // However in some situations we want the lval (effect address) of the item.
268 // We provide a special factory for making those lvals.
269 bool _is_lval;
271 // If the target is far we'll need to load the ea of this to
272 // a register to reach it. Otherwise if near we can do rip
273 // relative addressing.
275 address _target;
277 protected:
278 // creation
279 AddressLiteral()
280 : _is_lval(false),
281 _target(NULL)
282 {}
284 public:
287 AddressLiteral(address target, relocInfo::relocType rtype);
289 AddressLiteral(address target, RelocationHolder const& rspec)
290 : _rspec(rspec),
291 _is_lval(false),
292 _target(target)
293 {}
295 AddressLiteral addr() {
296 AddressLiteral ret = *this;
297 ret._is_lval = true;
298 return ret;
299 }
302 private:
304 address target() { return _target; }
305 bool is_lval() { return _is_lval; }
307 relocInfo::relocType reloc() const { return _rspec.type(); }
308 const RelocationHolder& rspec() const { return _rspec; }
310 friend class Assembler;
311 friend class MacroAssembler;
312 friend class Address;
313 friend class LIR_Assembler;
314 };
316 // Convience classes
317 class RuntimeAddress: public AddressLiteral {
319 public:
321 RuntimeAddress(address target) : AddressLiteral(target, relocInfo::runtime_call_type) {}
323 };
325 class OopAddress: public AddressLiteral {
327 public:
329 OopAddress(address target) : AddressLiteral(target, relocInfo::oop_type){}
331 };
333 class ExternalAddress: public AddressLiteral {
335 public:
337 ExternalAddress(address target) : AddressLiteral(target, relocInfo::external_word_type){}
339 };
341 class InternalAddress: public AddressLiteral {
343 public:
345 InternalAddress(address target) : AddressLiteral(target, relocInfo::internal_word_type) {}
347 };
349 // x86 can do array addressing as a single operation since disp can be an absolute
350 // address but amd64 can't [e.g. array_base(rx, ry:width) ]. We create a class
351 // that expresses the concept but does extra magic on amd64 to get the final result
353 class ArrayAddress VALUE_OBJ_CLASS_SPEC {
354 private:
356 AddressLiteral _base;
357 Address _index;
359 public:
361 ArrayAddress() {};
362 ArrayAddress(AddressLiteral base, Address index): _base(base), _index(index) {};
363 AddressLiteral base() { return _base; }
364 Address index() { return _index; }
366 };
368 // The amd64 Assembler: Pure assembler doing NO optimizations on
369 // the instruction level (e.g. mov rax, 0 is not translated into xor
370 // rax, rax!); i.e., what you write is what you get. The Assembler is
371 // generating code into a CodeBuffer.
373 const int FPUStateSizeInWords = 512 / wordSize;
375 class Assembler : public AbstractAssembler {
376 friend class AbstractAssembler; // for the non-virtual hack
377 friend class StubGenerator;
380 protected:
381 #ifdef ASSERT
382 void check_relocation(RelocationHolder const& rspec, int format);
383 #endif
385 inline void emit_long64(jlong x);
387 void emit_data(jint data, relocInfo::relocType rtype, int format /* = 1 */);
388 void emit_data(jint data, RelocationHolder const& rspec, int format /* = 1 */);
389 void emit_data64(jlong data, relocInfo::relocType rtype, int format = 0);
390 void emit_data64(jlong data, RelocationHolder const& rspec, int format = 0);
392 // Helper functions for groups of instructions
393 void emit_arith_b(int op1, int op2, Register dst, int imm8);
395 void emit_arith(int op1, int op2, Register dst, int imm32);
396 // only x86??
397 void emit_arith(int op1, int op2, Register dst, jobject obj);
398 void emit_arith(int op1, int op2, Register dst, Register src);
400 void emit_operand(Register reg,
401 Register base, Register index, Address::ScaleFactor scale,
402 int disp,
403 RelocationHolder const& rspec,
404 int rip_relative_correction = 0);
405 void emit_operand(Register reg, Address adr,
406 int rip_relative_correction = 0);
407 void emit_operand(XMMRegister reg,
408 Register base, Register index, Address::ScaleFactor scale,
409 int disp,
410 RelocationHolder const& rspec,
411 int rip_relative_correction = 0);
412 void emit_operand(XMMRegister reg, Address adr,
413 int rip_relative_correction = 0);
415 // Immediate-to-memory forms
416 void emit_arith_operand(int op1, Register rm, Address adr, int imm32);
418 void emit_farith(int b1, int b2, int i);
420 bool reachable(AddressLiteral adr);
422 // These are all easily abused and hence protected
424 // Make these disappear in 64bit mode since they would never be correct
425 #ifndef _LP64
426 void cmp_literal32(Register src1, int32_t imm32, RelocationHolder const& rspec);
427 void cmp_literal32(Address src1, int32_t imm32, RelocationHolder const& rspec);
429 void mov_literal32(Register dst, int32_t imm32, RelocationHolder const& rspec);
430 void mov_literal32(Address dst, int32_t imm32, RelocationHolder const& rspec);
432 void push_literal32(int32_t imm32, RelocationHolder const& rspec);
433 #endif // _LP64
436 void mov_literal64(Register dst, intptr_t imm64, RelocationHolder const& rspec);
438 // These are unique in that we are ensured by the caller that the 32bit
439 // relative in these instructions will always be able to reach the potentially
440 // 64bit address described by entry. Since they can take a 64bit address they
441 // don't have the 32 suffix like the other instructions in this class.
442 void jmp_literal(address entry, RelocationHolder const& rspec);
443 void call_literal(address entry, RelocationHolder const& rspec);
445 public:
446 enum Condition { // The amd64 condition codes used for conditional jumps/moves.
447 zero = 0x4,
448 notZero = 0x5,
449 equal = 0x4,
450 notEqual = 0x5,
451 less = 0xc,
452 lessEqual = 0xe,
453 greater = 0xf,
454 greaterEqual = 0xd,
455 below = 0x2,
456 belowEqual = 0x6,
457 above = 0x7,
458 aboveEqual = 0x3,
459 overflow = 0x0,
460 noOverflow = 0x1,
461 carrySet = 0x2,
462 carryClear = 0x3,
463 negative = 0x8,
464 positive = 0x9,
465 parity = 0xa,
466 noParity = 0xb
467 };
469 enum Prefix {
470 // segment overrides
471 // XXX remove segment prefixes
472 CS_segment = 0x2e,
473 SS_segment = 0x36,
474 DS_segment = 0x3e,
475 ES_segment = 0x26,
476 FS_segment = 0x64,
477 GS_segment = 0x65,
479 REX = 0x40,
481 REX_B = 0x41,
482 REX_X = 0x42,
483 REX_XB = 0x43,
484 REX_R = 0x44,
485 REX_RB = 0x45,
486 REX_RX = 0x46,
487 REX_RXB = 0x47,
489 REX_W = 0x48,
491 REX_WB = 0x49,
492 REX_WX = 0x4A,
493 REX_WXB = 0x4B,
494 REX_WR = 0x4C,
495 REX_WRB = 0x4D,
496 REX_WRX = 0x4E,
497 REX_WRXB = 0x4F
498 };
500 enum WhichOperand {
501 // input to locate_operand, and format code for relocations
502 imm64_operand = 0, // embedded 64-bit immediate operand
503 disp32_operand = 1, // embedded 32-bit displacement
504 call32_operand = 2, // embedded 32-bit self-relative displacement
505 #ifndef AMD64
506 _WhichOperand_limit = 3
507 #else
508 narrow_oop_operand = 3, // embedded 32-bit immediate narrow oop
509 _WhichOperand_limit = 4
510 #endif
511 };
513 public:
515 // Creation
516 Assembler(CodeBuffer* code)
517 : AbstractAssembler(code) {
518 }
520 // Decoding
521 static address locate_operand(address inst, WhichOperand which);
522 static address locate_next_instruction(address inst);
524 // Utilities
526 static bool is_simm(int64_t x, int nbits) { return -( CONST64(1) << (nbits-1) ) <= x && x < ( CONST64(1) << (nbits-1) ); }
527 static bool is_simm32 (int64_t x) { return x == (int64_t)(int32_t)x; }
530 // Stack
531 void pushaq();
532 void popaq();
534 void pushfq();
535 void popfq();
537 void pushq(int imm32);
539 void pushq(Register src);
540 void pushq(Address src);
542 void popq(Register dst);
543 void popq(Address dst);
545 // Instruction prefixes
546 void prefix(Prefix p);
548 int prefix_and_encode(int reg_enc, bool byteinst = false);
549 int prefixq_and_encode(int reg_enc);
551 int prefix_and_encode(int dst_enc, int src_enc, bool byteinst = false);
552 int prefixq_and_encode(int dst_enc, int src_enc);
554 void prefix(Register reg);
555 void prefix(Address adr);
556 void prefixq(Address adr);
558 void prefix(Address adr, Register reg, bool byteinst = false);
559 void prefixq(Address adr, Register reg);
561 void prefix(Address adr, XMMRegister reg);
563 // Moves
564 void movb(Register dst, Address src);
565 void movb(Address dst, int imm8);
566 void movb(Address dst, Register src);
568 void movw(Address dst, int imm16);
569 void movw(Register dst, Address src);
570 void movw(Address dst, Register src);
572 void movl(Register dst, int imm32);
573 void movl(Register dst, Register src);
574 void movl(Register dst, Address src);
575 void movl(Address dst, int imm32);
576 void movl(Address dst, Register src);
578 void movq(Register dst, Register src);
579 void movq(Register dst, Address src);
580 void movq(Address dst, Register src);
581 // These prevent using movq from converting a zero (like NULL) into Register
582 // by giving the compiler two choices it can't resolve
583 void movq(Address dst, void* dummy);
584 void movq(Register dst, void* dummy);
586 void mov64(Register dst, intptr_t imm64);
587 void mov64(Address dst, intptr_t imm64);
589 void movsbl(Register dst, Address src);
590 void movsbl(Register dst, Register src);
591 void movswl(Register dst, Address src);
592 void movswl(Register dst, Register src);
593 void movslq(Register dst, Address src);
594 void movslq(Register dst, Register src);
596 void movzbl(Register dst, Address src);
597 void movzbl(Register dst, Register src);
598 void movzwl(Register dst, Address src);
599 void movzwl(Register dst, Register src);
601 protected: // Avoid using the next instructions directly.
602 // New cpus require use of movsd and movss to avoid partial register stall
603 // when loading from memory. But for old Opteron use movlpd instead of movsd.
604 // The selection is done in MacroAssembler::movdbl() and movflt().
605 void movss(XMMRegister dst, XMMRegister src);
606 void movss(XMMRegister dst, Address src);
607 void movss(Address dst, XMMRegister src);
608 void movsd(XMMRegister dst, XMMRegister src);
609 void movsd(Address dst, XMMRegister src);
610 void movsd(XMMRegister dst, Address src);
611 void movlpd(XMMRegister dst, Address src);
612 // New cpus require use of movaps and movapd to avoid partial register stall
613 // when moving between registers.
614 void movapd(XMMRegister dst, XMMRegister src);
615 void movaps(XMMRegister dst, XMMRegister src);
616 public:
618 void movdl(XMMRegister dst, Register src);
619 void movdl(Register dst, XMMRegister src);
620 void movdq(XMMRegister dst, Register src);
621 void movdq(Register dst, XMMRegister src);
623 void cmovl(Condition cc, Register dst, Register src);
624 void cmovl(Condition cc, Register dst, Address src);
625 void cmovq(Condition cc, Register dst, Register src);
626 void cmovq(Condition cc, Register dst, Address src);
628 // Prefetches
629 private:
630 void prefetch_prefix(Address src);
631 public:
632 void prefetcht0(Address src);
633 void prefetcht1(Address src);
634 void prefetcht2(Address src);
635 void prefetchnta(Address src);
636 void prefetchw(Address src);
638 // Arithmetics
639 void adcl(Register dst, int imm32);
640 void adcl(Register dst, Address src);
641 void adcl(Register dst, Register src);
642 void adcq(Register dst, int imm32);
643 void adcq(Register dst, Address src);
644 void adcq(Register dst, Register src);
646 void addl(Address dst, int imm32);
647 void addl(Address dst, Register src);
648 void addl(Register dst, int imm32);
649 void addl(Register dst, Address src);
650 void addl(Register dst, Register src);
651 void addq(Address dst, int imm32);
652 void addq(Address dst, Register src);
653 void addq(Register dst, int imm32);
654 void addq(Register dst, Address src);
655 void addq(Register dst, Register src);
657 void andl(Register dst, int imm32);
658 void andl(Register dst, Address src);
659 void andl(Register dst, Register src);
660 void andq(Register dst, int imm32);
661 void andq(Register dst, Address src);
662 void andq(Register dst, Register src);
664 void cmpb(Address dst, int imm8);
665 void cmpl(Address dst, int imm32);
666 void cmpl(Register dst, int imm32);
667 void cmpl(Register dst, Register src);
668 void cmpl(Register dst, Address src);
669 void cmpq(Address dst, int imm32);
670 void cmpq(Address dst, Register src);
671 void cmpq(Register dst, int imm32);
672 void cmpq(Register dst, Register src);
673 void cmpq(Register dst, Address src);
675 void ucomiss(XMMRegister dst, XMMRegister src);
676 void ucomisd(XMMRegister dst, XMMRegister src);
678 protected:
679 // Don't use next inc() and dec() methods directly. INC & DEC instructions
680 // could cause a partial flag stall since they don't set CF flag.
681 // Use MacroAssembler::decrement() & MacroAssembler::increment() methods
682 // which call inc() & dec() or add() & sub() in accordance with
683 // the product flag UseIncDec value.
685 void decl(Register dst);
686 void decl(Address dst);
687 void decq(Register dst);
688 void decq(Address dst);
690 void incl(Register dst);
691 void incl(Address dst);
692 void incq(Register dst);
693 void incq(Address dst);
695 public:
696 void idivl(Register src);
697 void idivq(Register src);
698 void cdql();
699 void cdqq();
701 void imull(Register dst, Register src);
702 void imull(Register dst, Register src, int value);
703 void imulq(Register dst, Register src);
704 void imulq(Register dst, Register src, int value);
706 void leal(Register dst, Address src);
707 void leaq(Register dst, Address src);
709 void mull(Address src);
710 void mull(Register src);
712 void negl(Register dst);
713 void negq(Register dst);
715 void notl(Register dst);
716 void notq(Register dst);
718 void orl(Address dst, int imm32);
719 void orl(Register dst, int imm32);
720 void orl(Register dst, Address src);
721 void orl(Register dst, Register src);
722 void orq(Address dst, int imm32);
723 void orq(Register dst, int imm32);
724 void orq(Register dst, Address src);
725 void orq(Register dst, Register src);
727 void rcll(Register dst, int imm8);
728 void rclq(Register dst, int imm8);
730 void sarl(Register dst, int imm8);
731 void sarl(Register dst);
732 void sarq(Register dst, int imm8);
733 void sarq(Register dst);
735 void sbbl(Address dst, int imm32);
736 void sbbl(Register dst, int imm32);
737 void sbbl(Register dst, Address src);
738 void sbbl(Register dst, Register src);
739 void sbbq(Address dst, int imm32);
740 void sbbq(Register dst, int imm32);
741 void sbbq(Register dst, Address src);
742 void sbbq(Register dst, Register src);
744 void shll(Register dst, int imm8);
745 void shll(Register dst);
746 void shlq(Register dst, int imm8);
747 void shlq(Register dst);
749 void shrl(Register dst, int imm8);
750 void shrl(Register dst);
751 void shrq(Register dst, int imm8);
752 void shrq(Register dst);
754 void subl(Address dst, int imm32);
755 void subl(Address dst, Register src);
756 void subl(Register dst, int imm32);
757 void subl(Register dst, Address src);
758 void subl(Register dst, Register src);
759 void subq(Address dst, int imm32);
760 void subq(Address dst, Register src);
761 void subq(Register dst, int imm32);
762 void subq(Register dst, Address src);
763 void subq(Register dst, Register src);
765 void testb(Register dst, int imm8);
766 void testl(Register dst, int imm32);
767 void testl(Register dst, Register src);
768 void testq(Register dst, int imm32);
769 void testq(Register dst, Register src);
771 void xaddl(Address dst, Register src);
772 void xaddq(Address dst, Register src);
774 void xorl(Register dst, int imm32);
775 void xorl(Register dst, Address src);
776 void xorl(Register dst, Register src);
777 void xorq(Register dst, int imm32);
778 void xorq(Register dst, Address src);
779 void xorq(Register dst, Register src);
781 // Miscellaneous
782 void bswapl(Register reg);
783 void bswapq(Register reg);
784 void lock();
786 void xchgl(Register reg, Address adr);
787 void xchgl(Register dst, Register src);
788 void xchgq(Register reg, Address adr);
789 void xchgq(Register dst, Register src);
791 void cmpxchgl(Register reg, Address adr);
792 void cmpxchgq(Register reg, Address adr);
794 void nop(int i = 1);
795 void addr_nop_4();
796 void addr_nop_5();
797 void addr_nop_7();
798 void addr_nop_8();
800 void hlt();
801 void ret(int imm16);
802 void smovl();
803 void rep_movl();
804 void rep_movq();
805 void rep_set();
806 void repne_scanl();
807 void repne_scanq();
808 void setb(Condition cc, Register dst);
810 void clflush(Address adr);
812 enum Membar_mask_bits {
813 StoreStore = 1 << 3,
814 LoadStore = 1 << 2,
815 StoreLoad = 1 << 1,
816 LoadLoad = 1 << 0
817 };
819 // Serializes memory.
820 void membar(Membar_mask_bits order_constraint) {
821 // We only have to handle StoreLoad and LoadLoad
822 if (order_constraint & StoreLoad) {
823 // MFENCE subsumes LFENCE
824 mfence();
825 } /* [jk] not needed currently: else if (order_constraint & LoadLoad) {
826 lfence();
827 } */
828 }
830 void lfence() {
831 emit_byte(0x0F);
832 emit_byte(0xAE);
833 emit_byte(0xE8);
834 }
836 void mfence() {
837 emit_byte(0x0F);
838 emit_byte(0xAE);
839 emit_byte(0xF0);
840 }
842 // Identify processor type and features
843 void cpuid() {
844 emit_byte(0x0F);
845 emit_byte(0xA2);
846 }
848 void cld() { emit_byte(0xfc);
849 }
851 void std() { emit_byte(0xfd);
852 }
855 // Calls
857 void call(Label& L, relocInfo::relocType rtype);
858 void call(Register reg);
859 void call(Address adr);
861 // Jumps
863 void jmp(Register reg);
864 void jmp(Address adr);
866 // Label operations & relative jumps (PPUM Appendix D)
867 // unconditional jump to L
868 void jmp(Label& L, relocInfo::relocType rtype = relocInfo::none);
871 // Unconditional 8-bit offset jump to L.
872 // WARNING: be very careful using this for forward jumps. If the label is
873 // not bound within an 8-bit offset of this instruction, a run-time error
874 // will occur.
875 void jmpb(Label& L);
877 // jcc is the generic conditional branch generator to run- time
878 // routines, jcc is used for branches to labels. jcc takes a branch
879 // opcode (cc) and a label (L) and generates either a backward
880 // branch or a forward branch and links it to the label fixup
881 // chain. Usage:
882 //
883 // Label L; // unbound label
884 // jcc(cc, L); // forward branch to unbound label
885 // bind(L); // bind label to the current pc
886 // jcc(cc, L); // backward branch to bound label
887 // bind(L); // illegal: a label may be bound only once
888 //
889 // Note: The same Label can be used for forward and backward branches
890 // but it may be bound only once.
892 void jcc(Condition cc, Label& L,
893 relocInfo::relocType rtype = relocInfo::none);
895 // Conditional jump to a 8-bit offset to L.
896 // WARNING: be very careful using this for forward jumps. If the label is
897 // not bound within an 8-bit offset of this instruction, a run-time error
898 // will occur.
899 void jccb(Condition cc, Label& L);
901 // Floating-point operations
903 void fxsave(Address dst);
904 void fxrstor(Address src);
905 void ldmxcsr(Address src);
906 void stmxcsr(Address dst);
908 void addss(XMMRegister dst, XMMRegister src);
909 void addss(XMMRegister dst, Address src);
910 void subss(XMMRegister dst, XMMRegister src);
911 void subss(XMMRegister dst, Address src);
912 void mulss(XMMRegister dst, XMMRegister src);
913 void mulss(XMMRegister dst, Address src);
914 void divss(XMMRegister dst, XMMRegister src);
915 void divss(XMMRegister dst, Address src);
916 void addsd(XMMRegister dst, XMMRegister src);
917 void addsd(XMMRegister dst, Address src);
918 void subsd(XMMRegister dst, XMMRegister src);
919 void subsd(XMMRegister dst, Address src);
920 void mulsd(XMMRegister dst, XMMRegister src);
921 void mulsd(XMMRegister dst, Address src);
922 void divsd(XMMRegister dst, XMMRegister src);
923 void divsd(XMMRegister dst, Address src);
925 // We only need the double form
926 void sqrtsd(XMMRegister dst, XMMRegister src);
927 void sqrtsd(XMMRegister dst, Address src);
929 void xorps(XMMRegister dst, XMMRegister src);
930 void xorps(XMMRegister dst, Address src);
931 void xorpd(XMMRegister dst, XMMRegister src);
932 void xorpd(XMMRegister dst, Address src);
934 void cvtsi2ssl(XMMRegister dst, Register src);
935 void cvtsi2ssq(XMMRegister dst, Register src);
936 void cvtsi2sdl(XMMRegister dst, Register src);
937 void cvtsi2sdq(XMMRegister dst, Register src);
938 void cvttss2sil(Register dst, XMMRegister src); // truncates
939 void cvttss2siq(Register dst, XMMRegister src); // truncates
940 void cvttsd2sil(Register dst, XMMRegister src); // truncates
941 void cvttsd2siq(Register dst, XMMRegister src); // truncates
942 void cvtss2sd(XMMRegister dst, XMMRegister src);
943 void cvtsd2ss(XMMRegister dst, XMMRegister src);
944 void cvtdq2pd(XMMRegister dst, XMMRegister src);
945 void cvtdq2ps(XMMRegister dst, XMMRegister src);
947 void pxor(XMMRegister dst, Address src); // Xor Packed Byte Integer Values
948 void pxor(XMMRegister dst, XMMRegister src); // Xor Packed Byte Integer Values
950 void movdqa(XMMRegister dst, Address src); // Move Aligned Double Quadword
951 void movdqa(XMMRegister dst, XMMRegister src);
952 void movdqa(Address dst, XMMRegister src);
954 void movq(XMMRegister dst, Address src);
955 void movq(Address dst, XMMRegister src);
957 void pshufd(XMMRegister dst, XMMRegister src, int mode); // Shuffle Packed Doublewords
958 void pshufd(XMMRegister dst, Address src, int mode);
959 void pshuflw(XMMRegister dst, XMMRegister src, int mode); // Shuffle Packed Low Words
960 void pshuflw(XMMRegister dst, Address src, int mode);
962 void psrlq(XMMRegister dst, int shift); // Shift Right Logical Quadword Immediate
964 void punpcklbw(XMMRegister dst, XMMRegister src); // Interleave Low Bytes
965 void punpcklbw(XMMRegister dst, Address src);
966 };
969 // MacroAssembler extends Assembler by frequently used macros.
970 //
971 // Instructions for which a 'better' code sequence exists depending
972 // on arguments should also go in here.
974 class MacroAssembler : public Assembler {
975 friend class LIR_Assembler;
976 protected:
978 Address as_Address(AddressLiteral adr);
979 Address as_Address(ArrayAddress adr);
981 // Support for VM calls
982 //
983 // This is the base routine called by the different versions of
984 // call_VM_leaf. The interpreter may customize this version by
985 // overriding it for its purposes (e.g., to save/restore additional
986 // registers when doing a VM call).
988 virtual void call_VM_leaf_base(
989 address entry_point, // the entry point
990 int number_of_arguments // the number of arguments to
991 // pop after the call
992 );
994 // This is the base routine called by the different versions of
995 // call_VM. The interpreter may customize this version by overriding
996 // it for its purposes (e.g., to save/restore additional registers
997 // when doing a VM call).
998 //
999 // If no java_thread register is specified (noreg) than rdi will be
1000 // used instead. call_VM_base returns the register which contains
1001 // the thread upon return. If a thread register has been specified,
1002 // the return value will correspond to that register. If no
1003 // last_java_sp is specified (noreg) than rsp will be used instead.
1004 virtual void call_VM_base( // returns the register
1005 // containing the thread upon
1006 // return
1007 Register oop_result, // where an oop-result ends up
1008 // if any; use noreg otherwise
1009 Register java_thread, // the thread if computed
1010 // before ; use noreg otherwise
1011 Register last_java_sp, // to set up last_Java_frame in
1012 // stubs; use noreg otherwise
1013 address entry_point, // the entry point
1014 int number_of_arguments, // the number of arguments (w/o
1015 // thread) to pop after the
1016 // call
1017 bool check_exceptions // whether to check for pending
1018 // exceptions after return
1019 );
1021 // This routines should emit JVMTI PopFrame handling and ForceEarlyReturn code.
1022 // The implementation is only non-empty for the InterpreterMacroAssembler,
1023 // as only the interpreter handles PopFrame and ForceEarlyReturn requests.
1024 virtual void check_and_handle_popframe(Register java_thread);
1025 virtual void check_and_handle_earlyret(Register java_thread);
1027 void call_VM_helper(Register oop_result,
1028 address entry_point,
1029 int number_of_arguments,
1030 bool check_exceptions = true);
1032 public:
1033 MacroAssembler(CodeBuffer* code) : Assembler(code) {}
1035 // Support for NULL-checks
1036 //
1037 // Generates code that causes a NULL OS exception if the content of
1038 // reg is NULL. If the accessed location is M[reg + offset] and the
1039 // offset is known, provide the offset. No explicit code generation
1040 // is needed if the offset is within a certain range (0 <= offset <=
1041 // page_size).
1042 void null_check(Register reg, int offset = -1);
1043 static bool needs_explicit_null_check(intptr_t offset);
1045 // Required platform-specific helpers for Label::patch_instructions.
1046 // They _shadow_ the declarations in AbstractAssembler, which are undefined.
1047 void pd_patch_instruction(address branch, address target);
1048 #ifndef PRODUCT
1049 static void pd_print_patched_instruction(address branch);
1050 #endif
1053 // The following 4 methods return the offset of the appropriate move
1054 // instruction. Note: these are 32 bit instructions
1056 // Support for fast byte/word loading with zero extension (depending
1057 // on particular CPU)
1058 int load_unsigned_byte(Register dst, Address src);
1059 int load_unsigned_word(Register dst, Address src);
1061 // Support for fast byte/word loading with sign extension (depending
1062 // on particular CPU)
1063 int load_signed_byte(Register dst, Address src);
1064 int load_signed_word(Register dst, Address src);
1066 // Support for inc/dec with optimal instruction selection depending
1067 // on value
1068 void incrementl(Register reg, int value = 1);
1069 void decrementl(Register reg, int value = 1);
1070 void incrementq(Register reg, int value = 1);
1071 void decrementq(Register reg, int value = 1);
1073 void incrementl(Address dst, int value = 1);
1074 void decrementl(Address dst, int value = 1);
1075 void incrementq(Address dst, int value = 1);
1076 void decrementq(Address dst, int value = 1);
1078 // Support optimal SSE move instructions.
1079 void movflt(XMMRegister dst, XMMRegister src) {
1080 if (UseXmmRegToRegMoveAll) { movaps(dst, src); return; }
1081 else { movss (dst, src); return; }
1082 }
1084 void movflt(XMMRegister dst, Address src) { movss(dst, src); }
1086 void movflt(XMMRegister dst, AddressLiteral src);
1088 void movflt(Address dst, XMMRegister src) { movss(dst, src); }
1090 void movdbl(XMMRegister dst, XMMRegister src) {
1091 if (UseXmmRegToRegMoveAll) { movapd(dst, src); return; }
1092 else { movsd (dst, src); return; }
1093 }
1095 void movdbl(XMMRegister dst, AddressLiteral src);
1097 void movdbl(XMMRegister dst, Address src) {
1098 if (UseXmmLoadAndClearUpper) { movsd (dst, src); return; }
1099 else { movlpd(dst, src); return; }
1100 }
1102 void movdbl(Address dst, XMMRegister src) { movsd(dst, src); }
1104 void incrementl(AddressLiteral dst);
1105 void incrementl(ArrayAddress dst);
1107 // Alignment
1108 void align(int modulus);
1110 // Misc
1111 void fat_nop(); // 5 byte nop
1114 // C++ bool manipulation
1116 void movbool(Register dst, Address src);
1117 void movbool(Address dst, bool boolconst);
1118 void movbool(Address dst, Register src);
1119 void testbool(Register dst);
1121 // oop manipulations
1122 void load_klass(Register dst, Register src);
1123 void store_klass(Register dst, Register src);
1124 void store_klass_gap(Register dst, Register src);
1126 void load_prototype_header(Register dst, Register src);
1128 void load_heap_oop(Register dst, Address src);
1129 void store_heap_oop(Address dst, Register src);
1130 void encode_heap_oop(Register r);
1131 void decode_heap_oop(Register r);
1132 void encode_heap_oop_not_null(Register r);
1133 void decode_heap_oop_not_null(Register r);
1134 void encode_heap_oop_not_null(Register dst, Register src);
1135 void decode_heap_oop_not_null(Register dst, Register src);
1137 void set_narrow_oop(Register dst, jobject obj);
1139 // Stack frame creation/removal
1140 void enter();
1141 void leave();
1143 // Support for getting the JavaThread pointer (i.e.; a reference to
1144 // thread-local information) The pointer will be loaded into the
1145 // thread register.
1146 void get_thread(Register thread);
1148 void int3();
1150 // Support for VM calls
1151 //
1152 // It is imperative that all calls into the VM are handled via the
1153 // call_VM macros. They make sure that the stack linkage is setup
1154 // correctly. call_VM's correspond to ENTRY/ENTRY_X entry points
1155 // while call_VM_leaf's correspond to LEAF entry points.
1156 void call_VM(Register oop_result,
1157 address entry_point,
1158 bool check_exceptions = true);
1159 void call_VM(Register oop_result,
1160 address entry_point,
1161 Register arg_1,
1162 bool check_exceptions = true);
1163 void call_VM(Register oop_result,
1164 address entry_point,
1165 Register arg_1, Register arg_2,
1166 bool check_exceptions = true);
1167 void call_VM(Register oop_result,
1168 address entry_point,
1169 Register arg_1, Register arg_2, Register arg_3,
1170 bool check_exceptions = true);
1172 // Overloadings with last_Java_sp
1173 void call_VM(Register oop_result,
1174 Register last_java_sp,
1175 address entry_point,
1176 int number_of_arguments = 0,
1177 bool check_exceptions = true);
1178 void call_VM(Register oop_result,
1179 Register last_java_sp,
1180 address entry_point,
1181 Register arg_1, bool
1182 check_exceptions = true);
1183 void call_VM(Register oop_result,
1184 Register last_java_sp,
1185 address entry_point,
1186 Register arg_1, Register arg_2,
1187 bool check_exceptions = true);
1188 void call_VM(Register oop_result,
1189 Register last_java_sp,
1190 address entry_point,
1191 Register arg_1, Register arg_2, Register arg_3,
1192 bool check_exceptions = true);
1194 void call_VM_leaf(address entry_point,
1195 int number_of_arguments = 0);
1196 void call_VM_leaf(address entry_point,
1197 Register arg_1);
1198 void call_VM_leaf(address entry_point,
1199 Register arg_1, Register arg_2);
1200 void call_VM_leaf(address entry_point,
1201 Register arg_1, Register arg_2, Register arg_3);
1203 // last Java Frame (fills frame anchor)
1204 void set_last_Java_frame(Register last_java_sp,
1205 Register last_java_fp,
1206 address last_java_pc);
1207 void reset_last_Java_frame(bool clear_fp, bool clear_pc);
1209 // Stores
1210 void store_check(Register obj); // store check for
1211 // obj - register is
1212 // destroyed
1213 // afterwards
1214 void store_check(Register obj, Address dst); // same as above, dst
1215 // is exact store
1216 // location (reg. is
1217 // destroyed)
1219 void g1_write_barrier_pre(Register obj, Register tmp, Register tmp2, bool tosca_live );
1220 void g1_write_barrier_post(Register store_addr, Register new_val, Register tmp, Register tmp2);
1222 // split store_check(Register obj) to enhance instruction interleaving
1223 void store_check_part_1(Register obj);
1224 void store_check_part_2(Register obj);
1226 // C 'boolean' to Java boolean: x == 0 ? 0 : 1
1227 void c2bool(Register x);
1229 // Int division/reminder for Java
1230 // (as idivl, but checks for special case as described in JVM spec.)
1231 // returns idivl instruction offset for implicit exception handling
1232 int corrected_idivl(Register reg);
1233 // Long division/reminder for Java
1234 // (as idivq, but checks for special case as described in JVM spec.)
1235 // returns idivq instruction offset for implicit exception handling
1236 int corrected_idivq(Register reg);
1238 // Push and pop integer/fpu/cpu state
1239 void push_IU_state();
1240 void pop_IU_state();
1242 void push_FPU_state();
1243 void pop_FPU_state();
1245 void push_CPU_state();
1246 void pop_CPU_state();
1248 // Sign extension
1249 void sign_extend_short(Register reg);
1250 void sign_extend_byte(Register reg);
1252 // Division by power of 2, rounding towards 0
1253 void division_with_shift(Register reg, int shift_value);
1255 // Round up to a power of two
1256 void round_to_l(Register reg, int modulus);
1257 void round_to_q(Register reg, int modulus);
1259 // allocation
1260 void eden_allocate(
1261 Register obj, // result: pointer to object after
1262 // successful allocation
1263 Register var_size_in_bytes, // object size in bytes if unknown at
1264 // compile time; invalid otherwise
1265 int con_size_in_bytes, // object size in bytes if known at
1266 // compile time
1267 Register t1, // temp register
1268 Label& slow_case // continuation point if fast
1269 // allocation fails
1270 );
1271 void tlab_allocate(
1272 Register obj, // result: pointer to object after
1273 // successful allocation
1274 Register var_size_in_bytes, // object size in bytes if unknown at
1275 // compile time; invalid otherwise
1276 int con_size_in_bytes, // object size in bytes if known at
1277 // compile time
1278 Register t1, // temp register
1279 Register t2, // temp register
1280 Label& slow_case // continuation point if fast
1281 // allocation fails
1282 );
1283 void tlab_refill(Label& retry_tlab, Label& try_eden, Label& slow_case);
1285 //----
1287 // Debugging
1289 // only if +VerifyOops
1290 void verify_oop(Register reg, const char* s = "broken oop");
1291 void verify_oop_addr(Address addr, const char * s = "broken oop addr");
1293 // if heap base register is used - reinit it with the correct value
1294 void reinit_heapbase();
1296 // only if +VerifyFPU
1297 void verify_FPU(int stack_depth, const char* s = "illegal FPU state") {}
1299 // prints msg, dumps registers and stops execution
1300 void stop(const char* msg);
1302 // prints message and continues
1303 void warn(const char* msg);
1305 static void debug(char* msg, int64_t pc, int64_t regs[]);
1307 void os_breakpoint();
1309 void untested()
1310 {
1311 stop("untested");
1312 }
1314 void unimplemented(const char* what = "")
1315 {
1316 char* b = new char[1024];
1317 sprintf(b, "unimplemented: %s", what);
1318 stop(b);
1319 }
1321 void should_not_reach_here()
1322 {
1323 stop("should not reach here");
1324 }
1326 // Stack overflow checking
1327 void bang_stack_with_offset(int offset)
1328 {
1329 // stack grows down, caller passes positive offset
1330 assert(offset > 0, "must bang with negative offset");
1331 movl(Address(rsp, (-offset)), rax);
1332 }
1334 // Writes to stack successive pages until offset reached to check for
1335 // stack overflow + shadow pages. Also, clobbers tmp
1336 void bang_stack_size(Register offset, Register tmp);
1338 // Support for serializing memory accesses between threads.
1339 void serialize_memory(Register thread, Register tmp);
1341 void verify_tlab();
1343 // Biased locking support
1344 // lock_reg and obj_reg must be loaded up with the appropriate values.
1345 // swap_reg must be rax and is killed.
1346 // tmp_reg must be supplied and is killed.
1347 // If swap_reg_contains_mark is true then the code assumes that the
1348 // mark word of the object has already been loaded into swap_reg.
1349 // Optional slow case is for implementations (interpreter and C1) which branch to
1350 // slow case directly. Leaves condition codes set for C2's Fast_Lock node.
1351 // Returns offset of first potentially-faulting instruction for null
1352 // check info (currently consumed only by C1). If
1353 // swap_reg_contains_mark is true then returns -1 as it is assumed
1354 // the calling code has already passed any potential faults.
1355 int biased_locking_enter(Register lock_reg, Register obj_reg, Register swap_reg, Register tmp_reg,
1356 bool swap_reg_contains_mark,
1357 Label& done, Label* slow_case = NULL,
1358 BiasedLockingCounters* counters = NULL);
1359 void biased_locking_exit (Register obj_reg, Register temp_reg, Label& done);
1361 Condition negate_condition(Condition cond);
1363 // Instructions that use AddressLiteral operands. These instruction can handle 32bit/64bit
1364 // operands. In general the names are modified to avoid hiding the instruction in Assembler
1365 // so that we don't need to implement all the varieties in the Assembler with trivial wrappers
1366 // here in MacroAssembler. The major exception to this rule is call
1368 // Arithmetics
1370 void cmp8(AddressLiteral src1, int8_t imm32);
1372 void cmp32(AddressLiteral src1, int32_t src2);
1373 // compare reg - mem, or reg - &mem
1374 void cmp32(Register src1, AddressLiteral src2);
1376 void cmp32(Register src1, Address src2);
1378 #ifndef _LP64
1379 void cmpoop(Address dst, jobject obj);
1380 void cmpoop(Register dst, jobject obj);
1381 #endif // _LP64
1383 // NOTE src2 must be the lval. This is NOT an mem-mem compare
1384 void cmpptr(Address src1, AddressLiteral src2);
1386 void cmpptr(Register src1, AddressLiteral src);
1388 // will be cmpreg(?)
1389 void cmp64(Register src1, AddressLiteral src);
1391 void cmpxchgptr(Register reg, Address adr);
1392 void cmpxchgptr(Register reg, AddressLiteral adr);
1394 // Helper functions for statistics gathering.
1395 // Conditionally (atomically, on MPs) increments passed counter address, preserving condition codes.
1396 void cond_inc32(Condition cond, AddressLiteral counter_addr);
1397 // Unconditional atomic increment.
1398 void atomic_incl(AddressLiteral counter_addr);
1401 void lea(Register dst, AddressLiteral src);
1402 void lea(Register dst, Address src);
1405 // Calls
1406 void call(Label& L, relocInfo::relocType rtype);
1407 void call(Register entry);
1408 void call(AddressLiteral entry);
1410 // Jumps
1412 // 32bit can do a case table jump in one instruction but we no longer allow the base
1413 // to be installed in the Address class
1414 void jump(ArrayAddress entry);
1416 void jump(AddressLiteral entry);
1417 void jump_cc(Condition cc, AddressLiteral dst);
1419 // Floating
1421 void ldmxcsr(Address src) { Assembler::ldmxcsr(src); }
1422 void ldmxcsr(AddressLiteral src);
1424 private:
1425 // these are private because users should be doing movflt/movdbl
1427 void movss(XMMRegister dst, XMMRegister src) { Assembler::movss(dst, src); }
1428 void movss(Address dst, XMMRegister src) { Assembler::movss(dst, src); }
1429 void movss(XMMRegister dst, Address src) { Assembler::movss(dst, src); }
1430 void movss(XMMRegister dst, AddressLiteral src);
1432 void movlpd(XMMRegister dst, Address src) {Assembler::movlpd(dst, src); }
1433 void movlpd(XMMRegister dst, AddressLiteral src);
1435 public:
1438 void xorpd(XMMRegister dst, XMMRegister src) {Assembler::xorpd(dst, src); }
1439 void xorpd(XMMRegister dst, Address src) {Assembler::xorpd(dst, src); }
1440 void xorpd(XMMRegister dst, AddressLiteral src);
1442 void xorps(XMMRegister dst, XMMRegister src) {Assembler::xorps(dst, src); }
1443 void xorps(XMMRegister dst, Address src) {Assembler::xorps(dst, src); }
1444 void xorps(XMMRegister dst, AddressLiteral src);
1447 // Data
1449 void movoop(Register dst, jobject obj);
1450 void movoop(Address dst, jobject obj);
1452 void movptr(ArrayAddress dst, Register src);
1453 void movptr(Register dst, AddressLiteral src);
1455 void movptr(Register dst, intptr_t src);
1456 void movptr(Address dst, intptr_t src);
1458 void movptr(Register dst, ArrayAddress src);
1460 // to avoid hiding movl
1461 void mov32(AddressLiteral dst, Register src);
1462 void mov32(Register dst, AddressLiteral src);
1464 void pushoop(jobject obj);
1466 // Can push value or effective address
1467 void pushptr(AddressLiteral src);
1469 };
1471 /**
1472 * class SkipIfEqual:
1473 *
1474 * Instantiating this class will result in assembly code being output that will
1475 * jump around any code emitted between the creation of the instance and it's
1476 * automatic destruction at the end of a scope block, depending on the value of
1477 * the flag passed to the constructor, which will be checked at run-time.
1478 */
1479 class SkipIfEqual {
1480 private:
1481 MacroAssembler* _masm;
1482 Label _label;
1484 public:
1485 SkipIfEqual(MacroAssembler*, const bool* flag_addr, bool value);
1486 ~SkipIfEqual();
1487 };
1490 #ifdef ASSERT
1491 inline bool AbstractAssembler::pd_check_instruction_mark() { return true; }
1492 #endif