Mercurial > hg > truffle
annotate src/cpu/x86/vm/assembler_x86.hpp @ 4409:9f8e4aeec1a9
Merge
author | Andreas Woess <andreas.woess@jku.at> |
---|---|
date | Sun, 29 Jan 2012 11:40:04 +0100 |
parents | 6729bbc1fcd6 |
children | 127b3692c168 |
rev | line source |
---|---|
0 | 1 /* |
2100
b1a2afa37ec4
7003271: Hotspot should track cumulative Java heap bytes allocated on a per-thread basis
phh
parents:
2008
diff
changeset
|
2 * Copyright (c) 1997, 2011, Oracle and/or its affiliates. All rights reserved. |
0 | 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 * | |
1552
c18cbe5936b8
6941466: Oracle rebranding changes for Hotspot repositories
trims
parents:
1503
diff
changeset
|
19 * Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA |
c18cbe5936b8
6941466: Oracle rebranding changes for Hotspot repositories
trims
parents:
1503
diff
changeset
|
20 * or visit www.oracle.com if you need additional information or have any |
c18cbe5936b8
6941466: Oracle rebranding changes for Hotspot repositories
trims
parents:
1503
diff
changeset
|
21 * questions. |
0 | 22 * |
23 */ | |
24 | |
1972 | 25 #ifndef CPU_X86_VM_ASSEMBLER_X86_HPP |
26 #define CPU_X86_VM_ASSEMBLER_X86_HPP | |
27 | |
0 | 28 class BiasedLockingCounters; |
29 | |
30 // Contains all the definitions needed for x86 assembly code generation. | |
31 | |
32 // Calling convention | |
33 class Argument VALUE_OBJ_CLASS_SPEC { | |
34 public: | |
35 enum { | |
36 #ifdef _LP64 | |
37 #ifdef _WIN64 | |
38 n_int_register_parameters_c = 4, // rcx, rdx, r8, r9 (c_rarg0, c_rarg1, ...) | |
39 n_float_register_parameters_c = 4, // xmm0 - xmm3 (c_farg0, c_farg1, ... ) | |
40 #else | |
41 n_int_register_parameters_c = 6, // rdi, rsi, rdx, rcx, r8, r9 (c_rarg0, c_rarg1, ...) | |
42 n_float_register_parameters_c = 8, // xmm0 - xmm7 (c_farg0, c_farg1, ... ) | |
43 #endif // _WIN64 | |
44 n_int_register_parameters_j = 6, // j_rarg0, j_rarg1, ... | |
45 n_float_register_parameters_j = 8 // j_farg0, j_farg1, ... | |
46 #else | |
47 n_register_parameters = 0 // 0 registers used to pass arguments | |
48 #endif // _LP64 | |
49 }; | |
50 }; | |
51 | |
52 | |
53 #ifdef _LP64 | |
54 // Symbolically name the register arguments used by the c calling convention. | |
55 // Windows is different from linux/solaris. So much for standards... | |
56 | |
57 #ifdef _WIN64 | |
58 | |
59 REGISTER_DECLARATION(Register, c_rarg0, rcx); | |
60 REGISTER_DECLARATION(Register, c_rarg1, rdx); | |
61 REGISTER_DECLARATION(Register, c_rarg2, r8); | |
62 REGISTER_DECLARATION(Register, c_rarg3, r9); | |
63 | |
304 | 64 REGISTER_DECLARATION(XMMRegister, c_farg0, xmm0); |
65 REGISTER_DECLARATION(XMMRegister, c_farg1, xmm1); | |
66 REGISTER_DECLARATION(XMMRegister, c_farg2, xmm2); | |
67 REGISTER_DECLARATION(XMMRegister, c_farg3, xmm3); | |
0 | 68 |
69 #else | |
70 | |
71 REGISTER_DECLARATION(Register, c_rarg0, rdi); | |
72 REGISTER_DECLARATION(Register, c_rarg1, rsi); | |
73 REGISTER_DECLARATION(Register, c_rarg2, rdx); | |
74 REGISTER_DECLARATION(Register, c_rarg3, rcx); | |
75 REGISTER_DECLARATION(Register, c_rarg4, r8); | |
76 REGISTER_DECLARATION(Register, c_rarg5, r9); | |
77 | |
304 | 78 REGISTER_DECLARATION(XMMRegister, c_farg0, xmm0); |
79 REGISTER_DECLARATION(XMMRegister, c_farg1, xmm1); | |
80 REGISTER_DECLARATION(XMMRegister, c_farg2, xmm2); | |
81 REGISTER_DECLARATION(XMMRegister, c_farg3, xmm3); | |
82 REGISTER_DECLARATION(XMMRegister, c_farg4, xmm4); | |
83 REGISTER_DECLARATION(XMMRegister, c_farg5, xmm5); | |
84 REGISTER_DECLARATION(XMMRegister, c_farg6, xmm6); | |
85 REGISTER_DECLARATION(XMMRegister, c_farg7, xmm7); | |
0 | 86 |
87 #endif // _WIN64 | |
88 | |
89 // Symbolically name the register arguments used by the Java calling convention. | |
90 // We have control over the convention for java so we can do what we please. | |
91 // What pleases us is to offset the java calling convention so that when | |
92 // we call a suitable jni method the arguments are lined up and we don't | |
93 // have to do little shuffling. A suitable jni method is non-static and a | |
94 // small number of arguments (two fewer args on windows) | |
95 // | |
96 // |-------------------------------------------------------| | |
97 // | c_rarg0 c_rarg1 c_rarg2 c_rarg3 c_rarg4 c_rarg5 | | |
98 // |-------------------------------------------------------| | |
99 // | rcx rdx r8 r9 rdi* rsi* | windows (* not a c_rarg) | |
100 // | rdi rsi rdx rcx r8 r9 | solaris/linux | |
101 // |-------------------------------------------------------| | |
102 // | j_rarg5 j_rarg0 j_rarg1 j_rarg2 j_rarg3 j_rarg4 | | |
103 // |-------------------------------------------------------| | |
104 | |
105 REGISTER_DECLARATION(Register, j_rarg0, c_rarg1); | |
106 REGISTER_DECLARATION(Register, j_rarg1, c_rarg2); | |
107 REGISTER_DECLARATION(Register, j_rarg2, c_rarg3); | |
108 // Windows runs out of register args here | |
109 #ifdef _WIN64 | |
110 REGISTER_DECLARATION(Register, j_rarg3, rdi); | |
111 REGISTER_DECLARATION(Register, j_rarg4, rsi); | |
112 #else | |
113 REGISTER_DECLARATION(Register, j_rarg3, c_rarg4); | |
114 REGISTER_DECLARATION(Register, j_rarg4, c_rarg5); | |
115 #endif /* _WIN64 */ | |
116 REGISTER_DECLARATION(Register, j_rarg5, c_rarg0); | |
117 | |
304 | 118 REGISTER_DECLARATION(XMMRegister, j_farg0, xmm0); |
119 REGISTER_DECLARATION(XMMRegister, j_farg1, xmm1); | |
120 REGISTER_DECLARATION(XMMRegister, j_farg2, xmm2); | |
121 REGISTER_DECLARATION(XMMRegister, j_farg3, xmm3); | |
122 REGISTER_DECLARATION(XMMRegister, j_farg4, xmm4); | |
123 REGISTER_DECLARATION(XMMRegister, j_farg5, xmm5); | |
124 REGISTER_DECLARATION(XMMRegister, j_farg6, xmm6); | |
125 REGISTER_DECLARATION(XMMRegister, j_farg7, xmm7); | |
0 | 126 |
127 REGISTER_DECLARATION(Register, rscratch1, r10); // volatile | |
128 REGISTER_DECLARATION(Register, rscratch2, r11); // volatile | |
129 | |
304 | 130 REGISTER_DECLARATION(Register, r12_heapbase, r12); // callee-saved |
0 | 131 REGISTER_DECLARATION(Register, r15_thread, r15); // callee-saved |
132 | |
304 | 133 #else |
134 // rscratch1 will apear in 32bit code that is dead but of course must compile | |
135 // Using noreg ensures if the dead code is incorrectly live and executed it | |
136 // will cause an assertion failure | |
137 #define rscratch1 noreg | |
2002 | 138 #define rscratch2 noreg |
304 | 139 |
0 | 140 #endif // _LP64 |
141 | |
1564 | 142 // JSR 292 fixed register usages: |
143 REGISTER_DECLARATION(Register, rbp_mh_SP_save, rbp); | |
144 | |
0 | 145 // Address is an abstraction used to represent a memory location |
146 // using any of the amd64 addressing modes with one object. | |
147 // | |
148 // Note: A register location is represented via a Register, not | |
149 // via an address for efficiency & simplicity reasons. | |
150 | |
151 class ArrayAddress; | |
152 | |
153 class Address VALUE_OBJ_CLASS_SPEC { | |
154 public: | |
155 enum ScaleFactor { | |
156 no_scale = -1, | |
157 times_1 = 0, | |
158 times_2 = 1, | |
159 times_4 = 2, | |
304 | 160 times_8 = 3, |
161 times_ptr = LP64_ONLY(times_8) NOT_LP64(times_4) | |
0 | 162 }; |
622
56aae7be60d4
6812678: macro assembler needs delayed binding of a few constants (for 6655638)
jrose
parents:
420
diff
changeset
|
163 static ScaleFactor times(int size) { |
56aae7be60d4
6812678: macro assembler needs delayed binding of a few constants (for 6655638)
jrose
parents:
420
diff
changeset
|
164 assert(size >= 1 && size <= 8 && is_power_of_2(size), "bad scale size"); |
56aae7be60d4
6812678: macro assembler needs delayed binding of a few constants (for 6655638)
jrose
parents:
420
diff
changeset
|
165 if (size == 8) return times_8; |
56aae7be60d4
6812678: macro assembler needs delayed binding of a few constants (for 6655638)
jrose
parents:
420
diff
changeset
|
166 if (size == 4) return times_4; |
56aae7be60d4
6812678: macro assembler needs delayed binding of a few constants (for 6655638)
jrose
parents:
420
diff
changeset
|
167 if (size == 2) return times_2; |
56aae7be60d4
6812678: macro assembler needs delayed binding of a few constants (for 6655638)
jrose
parents:
420
diff
changeset
|
168 return times_1; |
56aae7be60d4
6812678: macro assembler needs delayed binding of a few constants (for 6655638)
jrose
parents:
420
diff
changeset
|
169 } |
56aae7be60d4
6812678: macro assembler needs delayed binding of a few constants (for 6655638)
jrose
parents:
420
diff
changeset
|
170 static int scale_size(ScaleFactor scale) { |
56aae7be60d4
6812678: macro assembler needs delayed binding of a few constants (for 6655638)
jrose
parents:
420
diff
changeset
|
171 assert(scale != no_scale, ""); |
56aae7be60d4
6812678: macro assembler needs delayed binding of a few constants (for 6655638)
jrose
parents:
420
diff
changeset
|
172 assert(((1 << (int)times_1) == 1 && |
56aae7be60d4
6812678: macro assembler needs delayed binding of a few constants (for 6655638)
jrose
parents:
420
diff
changeset
|
173 (1 << (int)times_2) == 2 && |
56aae7be60d4
6812678: macro assembler needs delayed binding of a few constants (for 6655638)
jrose
parents:
420
diff
changeset
|
174 (1 << (int)times_4) == 4 && |
56aae7be60d4
6812678: macro assembler needs delayed binding of a few constants (for 6655638)
jrose
parents:
420
diff
changeset
|
175 (1 << (int)times_8) == 8), ""); |
56aae7be60d4
6812678: macro assembler needs delayed binding of a few constants (for 6655638)
jrose
parents:
420
diff
changeset
|
176 return (1 << (int)scale); |
56aae7be60d4
6812678: macro assembler needs delayed binding of a few constants (for 6655638)
jrose
parents:
420
diff
changeset
|
177 } |
0 | 178 |
179 private: | |
180 Register _base; | |
181 Register _index; | |
182 ScaleFactor _scale; | |
183 int _disp; | |
184 RelocationHolder _rspec; | |
185 | |
304 | 186 // Easily misused constructors make them private |
187 // %%% can we make these go away? | |
188 NOT_LP64(Address(address loc, RelocationHolder spec);) | |
189 Address(int disp, address loc, relocInfo::relocType rtype); | |
190 Address(int disp, address loc, RelocationHolder spec); | |
0 | 191 |
192 public: | |
304 | 193 |
194 int disp() { return _disp; } | |
0 | 195 // creation |
196 Address() | |
197 : _base(noreg), | |
198 _index(noreg), | |
199 _scale(no_scale), | |
200 _disp(0) { | |
201 } | |
202 | |
203 // No default displacement otherwise Register can be implicitly | |
204 // converted to 0(Register) which is quite a different animal. | |
205 | |
206 Address(Register base, int disp) | |
207 : _base(base), | |
208 _index(noreg), | |
209 _scale(no_scale), | |
210 _disp(disp) { | |
211 } | |
212 | |
213 Address(Register base, Register index, ScaleFactor scale, int disp = 0) | |
214 : _base (base), | |
215 _index(index), | |
216 _scale(scale), | |
217 _disp (disp) { | |
218 assert(!index->is_valid() == (scale == Address::no_scale), | |
219 "inconsistent address"); | |
220 } | |
221 | |
665
c89f86385056
6814659: separable cleanups and subroutines for 6655638
jrose
parents:
644
diff
changeset
|
222 Address(Register base, RegisterOrConstant index, ScaleFactor scale = times_1, int disp = 0) |
622
56aae7be60d4
6812678: macro assembler needs delayed binding of a few constants (for 6655638)
jrose
parents:
420
diff
changeset
|
223 : _base (base), |
56aae7be60d4
6812678: macro assembler needs delayed binding of a few constants (for 6655638)
jrose
parents:
420
diff
changeset
|
224 _index(index.register_or_noreg()), |
56aae7be60d4
6812678: macro assembler needs delayed binding of a few constants (for 6655638)
jrose
parents:
420
diff
changeset
|
225 _scale(scale), |
56aae7be60d4
6812678: macro assembler needs delayed binding of a few constants (for 6655638)
jrose
parents:
420
diff
changeset
|
226 _disp (disp + (index.constant_or_zero() * scale_size(scale))) { |
56aae7be60d4
6812678: macro assembler needs delayed binding of a few constants (for 6655638)
jrose
parents:
420
diff
changeset
|
227 if (!index.is_register()) scale = Address::no_scale; |
56aae7be60d4
6812678: macro assembler needs delayed binding of a few constants (for 6655638)
jrose
parents:
420
diff
changeset
|
228 assert(!_index->is_valid() == (scale == Address::no_scale), |
56aae7be60d4
6812678: macro assembler needs delayed binding of a few constants (for 6655638)
jrose
parents:
420
diff
changeset
|
229 "inconsistent address"); |
56aae7be60d4
6812678: macro assembler needs delayed binding of a few constants (for 6655638)
jrose
parents:
420
diff
changeset
|
230 } |
56aae7be60d4
6812678: macro assembler needs delayed binding of a few constants (for 6655638)
jrose
parents:
420
diff
changeset
|
231 |
56aae7be60d4
6812678: macro assembler needs delayed binding of a few constants (for 6655638)
jrose
parents:
420
diff
changeset
|
232 Address plus_disp(int disp) const { |
56aae7be60d4
6812678: macro assembler needs delayed binding of a few constants (for 6655638)
jrose
parents:
420
diff
changeset
|
233 Address a = (*this); |
56aae7be60d4
6812678: macro assembler needs delayed binding of a few constants (for 6655638)
jrose
parents:
420
diff
changeset
|
234 a._disp += disp; |
56aae7be60d4
6812678: macro assembler needs delayed binding of a few constants (for 6655638)
jrose
parents:
420
diff
changeset
|
235 return a; |
56aae7be60d4
6812678: macro assembler needs delayed binding of a few constants (for 6655638)
jrose
parents:
420
diff
changeset
|
236 } |
3363
167b70ff3abc
6939861: JVM should handle more conversion operations
never
parents:
3336
diff
changeset
|
237 Address plus_disp(RegisterOrConstant disp, ScaleFactor scale = times_1) const { |
167b70ff3abc
6939861: JVM should handle more conversion operations
never
parents:
3336
diff
changeset
|
238 Address a = (*this); |
167b70ff3abc
6939861: JVM should handle more conversion operations
never
parents:
3336
diff
changeset
|
239 a._disp += disp.constant_or_zero() * scale_size(scale); |
167b70ff3abc
6939861: JVM should handle more conversion operations
never
parents:
3336
diff
changeset
|
240 if (disp.is_register()) { |
167b70ff3abc
6939861: JVM should handle more conversion operations
never
parents:
3336
diff
changeset
|
241 assert(!a.index()->is_valid(), "competing indexes"); |
167b70ff3abc
6939861: JVM should handle more conversion operations
never
parents:
3336
diff
changeset
|
242 a._index = disp.as_register(); |
167b70ff3abc
6939861: JVM should handle more conversion operations
never
parents:
3336
diff
changeset
|
243 a._scale = scale; |
167b70ff3abc
6939861: JVM should handle more conversion operations
never
parents:
3336
diff
changeset
|
244 } |
167b70ff3abc
6939861: JVM should handle more conversion operations
never
parents:
3336
diff
changeset
|
245 return a; |
167b70ff3abc
6939861: JVM should handle more conversion operations
never
parents:
3336
diff
changeset
|
246 } |
167b70ff3abc
6939861: JVM should handle more conversion operations
never
parents:
3336
diff
changeset
|
247 bool is_same_address(Address a) const { |
167b70ff3abc
6939861: JVM should handle more conversion operations
never
parents:
3336
diff
changeset
|
248 // disregard _rspec |
167b70ff3abc
6939861: JVM should handle more conversion operations
never
parents:
3336
diff
changeset
|
249 return _base == a._base && _disp == a._disp && _index == a._index && _scale == a._scale; |
167b70ff3abc
6939861: JVM should handle more conversion operations
never
parents:
3336
diff
changeset
|
250 } |
622
56aae7be60d4
6812678: macro assembler needs delayed binding of a few constants (for 6655638)
jrose
parents:
420
diff
changeset
|
251 |
0 | 252 // The following two overloads are used in connection with the |
253 // ByteSize type (see sizes.hpp). They simplify the use of | |
254 // ByteSize'd arguments in assembly code. Note that their equivalent | |
255 // for the optimized build are the member functions with int disp | |
256 // argument since ByteSize is mapped to an int type in that case. | |
257 // | |
258 // Note: DO NOT introduce similar overloaded functions for WordSize | |
259 // arguments as in the optimized mode, both ByteSize and WordSize | |
260 // are mapped to the same type and thus the compiler cannot make a | |
261 // distinction anymore (=> compiler errors). | |
262 | |
263 #ifdef ASSERT | |
264 Address(Register base, ByteSize disp) | |
265 : _base(base), | |
266 _index(noreg), | |
267 _scale(no_scale), | |
268 _disp(in_bytes(disp)) { | |
269 } | |
270 | |
271 Address(Register base, Register index, ScaleFactor scale, ByteSize disp) | |
272 : _base(base), | |
273 _index(index), | |
274 _scale(scale), | |
275 _disp(in_bytes(disp)) { | |
276 assert(!index->is_valid() == (scale == Address::no_scale), | |
277 "inconsistent address"); | |
278 } | |
622
56aae7be60d4
6812678: macro assembler needs delayed binding of a few constants (for 6655638)
jrose
parents:
420
diff
changeset
|
279 |
665
c89f86385056
6814659: separable cleanups and subroutines for 6655638
jrose
parents:
644
diff
changeset
|
280 Address(Register base, RegisterOrConstant index, ScaleFactor scale, ByteSize disp) |
622
56aae7be60d4
6812678: macro assembler needs delayed binding of a few constants (for 6655638)
jrose
parents:
420
diff
changeset
|
281 : _base (base), |
56aae7be60d4
6812678: macro assembler needs delayed binding of a few constants (for 6655638)
jrose
parents:
420
diff
changeset
|
282 _index(index.register_or_noreg()), |
56aae7be60d4
6812678: macro assembler needs delayed binding of a few constants (for 6655638)
jrose
parents:
420
diff
changeset
|
283 _scale(scale), |
56aae7be60d4
6812678: macro assembler needs delayed binding of a few constants (for 6655638)
jrose
parents:
420
diff
changeset
|
284 _disp (in_bytes(disp) + (index.constant_or_zero() * scale_size(scale))) { |
56aae7be60d4
6812678: macro assembler needs delayed binding of a few constants (for 6655638)
jrose
parents:
420
diff
changeset
|
285 if (!index.is_register()) scale = Address::no_scale; |
56aae7be60d4
6812678: macro assembler needs delayed binding of a few constants (for 6655638)
jrose
parents:
420
diff
changeset
|
286 assert(!_index->is_valid() == (scale == Address::no_scale), |
56aae7be60d4
6812678: macro assembler needs delayed binding of a few constants (for 6655638)
jrose
parents:
420
diff
changeset
|
287 "inconsistent address"); |
56aae7be60d4
6812678: macro assembler needs delayed binding of a few constants (for 6655638)
jrose
parents:
420
diff
changeset
|
288 } |
56aae7be60d4
6812678: macro assembler needs delayed binding of a few constants (for 6655638)
jrose
parents:
420
diff
changeset
|
289 |
0 | 290 #endif // ASSERT |
291 | |
292 // accessors | |
342
37f87013dfd8
6711316: Open source the Garbage-First garbage collector
ysr
parents:
71
diff
changeset
|
293 bool uses(Register reg) const { return _base == reg || _index == reg; } |
37f87013dfd8
6711316: Open source the Garbage-First garbage collector
ysr
parents:
71
diff
changeset
|
294 Register base() const { return _base; } |
37f87013dfd8
6711316: Open source the Garbage-First garbage collector
ysr
parents:
71
diff
changeset
|
295 Register index() const { return _index; } |
37f87013dfd8
6711316: Open source the Garbage-First garbage collector
ysr
parents:
71
diff
changeset
|
296 ScaleFactor scale() const { return _scale; } |
37f87013dfd8
6711316: Open source the Garbage-First garbage collector
ysr
parents:
71
diff
changeset
|
297 int disp() const { return _disp; } |
0 | 298 |
299 // Convert the raw encoding form into the form expected by the constructor for | |
300 // Address. An index of 4 (rsp) corresponds to having no index, so convert | |
301 // that to noreg for the Address constructor. | |
624 | 302 static Address make_raw(int base, int index, int scale, int disp, bool disp_is_oop); |
0 | 303 |
304 static Address make_array(ArrayAddress); | |
305 | |
306 private: | |
307 bool base_needs_rex() const { | |
308 return _base != noreg && _base->encoding() >= 8; | |
309 } | |
310 | |
311 bool index_needs_rex() const { | |
312 return _index != noreg &&_index->encoding() >= 8; | |
313 } | |
314 | |
315 relocInfo::relocType reloc() const { return _rspec.type(); } | |
316 | |
317 friend class Assembler; | |
318 friend class MacroAssembler; | |
319 friend class LIR_Assembler; // base/index/scale/disp | |
320 }; | |
321 | |
322 // | |
323 // AddressLiteral has been split out from Address because operands of this type | |
324 // need to be treated specially on 32bit vs. 64bit platforms. By splitting it out | |
325 // the few instructions that need to deal with address literals are unique and the | |
326 // MacroAssembler does not have to implement every instruction in the Assembler | |
327 // in order to search for address literals that may need special handling depending | |
328 // on the instruction and the platform. As small step on the way to merging i486/amd64 | |
329 // directories. | |
330 // | |
331 class AddressLiteral VALUE_OBJ_CLASS_SPEC { | |
332 friend class ArrayAddress; | |
333 RelocationHolder _rspec; | |
334 // Typically we use AddressLiterals we want to use their rval | |
335 // However in some situations we want the lval (effect address) of the item. | |
336 // We provide a special factory for making those lvals. | |
337 bool _is_lval; | |
338 | |
339 // If the target is far we'll need to load the ea of this to | |
340 // a register to reach it. Otherwise if near we can do rip | |
341 // relative addressing. | |
342 | |
343 address _target; | |
344 | |
345 protected: | |
346 // creation | |
347 AddressLiteral() | |
348 : _is_lval(false), | |
349 _target(NULL) | |
350 {} | |
351 | |
352 public: | |
353 | |
354 | |
355 AddressLiteral(address target, relocInfo::relocType rtype); | |
356 | |
357 AddressLiteral(address target, RelocationHolder const& rspec) | |
358 : _rspec(rspec), | |
359 _is_lval(false), | |
360 _target(target) | |
361 {} | |
362 | |
363 AddressLiteral addr() { | |
364 AddressLiteral ret = *this; | |
365 ret._is_lval = true; | |
366 return ret; | |
367 } | |
368 | |
369 | |
370 private: | |
371 | |
372 address target() { return _target; } | |
373 bool is_lval() { return _is_lval; } | |
374 | |
375 relocInfo::relocType reloc() const { return _rspec.type(); } | |
376 const RelocationHolder& rspec() const { return _rspec; } | |
377 | |
378 friend class Assembler; | |
379 friend class MacroAssembler; | |
380 friend class Address; | |
381 friend class LIR_Assembler; | |
382 }; | |
383 | |
384 // Convience classes | |
385 class RuntimeAddress: public AddressLiteral { | |
386 | |
387 public: | |
388 | |
389 RuntimeAddress(address target) : AddressLiteral(target, relocInfo::runtime_call_type) {} | |
390 | |
391 }; | |
392 | |
393 class OopAddress: public AddressLiteral { | |
394 | |
395 public: | |
396 | |
397 OopAddress(address target) : AddressLiteral(target, relocInfo::oop_type){} | |
398 | |
399 }; | |
400 | |
401 class ExternalAddress: public AddressLiteral { | |
2455
479b4b4b6950
6777083: assert(target != __null,"must not be null")
never
parents:
2415
diff
changeset
|
402 private: |
479b4b4b6950
6777083: assert(target != __null,"must not be null")
never
parents:
2415
diff
changeset
|
403 static relocInfo::relocType reloc_for_target(address target) { |
479b4b4b6950
6777083: assert(target != __null,"must not be null")
never
parents:
2415
diff
changeset
|
404 // Sometimes ExternalAddress is used for values which aren't |
479b4b4b6950
6777083: assert(target != __null,"must not be null")
never
parents:
2415
diff
changeset
|
405 // exactly addresses, like the card table base. |
479b4b4b6950
6777083: assert(target != __null,"must not be null")
never
parents:
2415
diff
changeset
|
406 // external_word_type can't be used for values in the first page |
479b4b4b6950
6777083: assert(target != __null,"must not be null")
never
parents:
2415
diff
changeset
|
407 // so just skip the reloc in that case. |
479b4b4b6950
6777083: assert(target != __null,"must not be null")
never
parents:
2415
diff
changeset
|
408 return external_word_Relocation::can_be_relocated(target) ? relocInfo::external_word_type : relocInfo::none; |
479b4b4b6950
6777083: assert(target != __null,"must not be null")
never
parents:
2415
diff
changeset
|
409 } |
479b4b4b6950
6777083: assert(target != __null,"must not be null")
never
parents:
2415
diff
changeset
|
410 |
479b4b4b6950
6777083: assert(target != __null,"must not be null")
never
parents:
2415
diff
changeset
|
411 public: |
479b4b4b6950
6777083: assert(target != __null,"must not be null")
never
parents:
2415
diff
changeset
|
412 |
479b4b4b6950
6777083: assert(target != __null,"must not be null")
never
parents:
2415
diff
changeset
|
413 ExternalAddress(address target) : AddressLiteral(target, reloc_for_target(target)) {} |
0 | 414 |
415 }; | |
416 | |
417 class InternalAddress: public AddressLiteral { | |
418 | |
419 public: | |
420 | |
421 InternalAddress(address target) : AddressLiteral(target, relocInfo::internal_word_type) {} | |
422 | |
423 }; | |
424 | |
425 // x86 can do array addressing as a single operation since disp can be an absolute | |
426 // address amd64 can't. We create a class that expresses the concept but does extra | |
427 // magic on amd64 to get the final result | |
428 | |
429 class ArrayAddress VALUE_OBJ_CLASS_SPEC { | |
430 private: | |
431 | |
432 AddressLiteral _base; | |
433 Address _index; | |
434 | |
435 public: | |
436 | |
437 ArrayAddress() {}; | |
438 ArrayAddress(AddressLiteral base, Address index): _base(base), _index(index) {}; | |
439 AddressLiteral base() { return _base; } | |
440 Address index() { return _index; } | |
441 | |
442 }; | |
443 | |
304 | 444 const int FPUStateSizeInWords = NOT_LP64(27) LP64_ONLY( 512 / wordSize); |
0 | 445 |
446 // The Intel x86/Amd64 Assembler: Pure assembler doing NO optimizations on the instruction | |
447 // level (e.g. mov rax, 0 is not translated into xor rax, rax!); i.e., what you write | |
448 // is what you get. The Assembler is generating code into a CodeBuffer. | |
449 | |
450 class Assembler : public AbstractAssembler { | |
451 friend class AbstractAssembler; // for the non-virtual hack | |
452 friend class LIR_Assembler; // as_Address() | |
304 | 453 friend class StubGenerator; |
0 | 454 |
455 public: | |
456 enum Condition { // The x86 condition codes used for conditional jumps/moves. | |
457 zero = 0x4, | |
458 notZero = 0x5, | |
459 equal = 0x4, | |
460 notEqual = 0x5, | |
461 less = 0xc, | |
462 lessEqual = 0xe, | |
463 greater = 0xf, | |
464 greaterEqual = 0xd, | |
465 below = 0x2, | |
466 belowEqual = 0x6, | |
467 above = 0x7, | |
468 aboveEqual = 0x3, | |
469 overflow = 0x0, | |
470 noOverflow = 0x1, | |
471 carrySet = 0x2, | |
472 carryClear = 0x3, | |
473 negative = 0x8, | |
474 positive = 0x9, | |
475 parity = 0xa, | |
476 noParity = 0xb | |
477 }; | |
478 | |
479 enum Prefix { | |
480 // segment overrides | |
481 CS_segment = 0x2e, | |
482 SS_segment = 0x36, | |
483 DS_segment = 0x3e, | |
484 ES_segment = 0x26, | |
485 FS_segment = 0x64, | |
486 GS_segment = 0x65, | |
487 | |
488 REX = 0x40, | |
489 | |
490 REX_B = 0x41, | |
491 REX_X = 0x42, | |
492 REX_XB = 0x43, | |
493 REX_R = 0x44, | |
494 REX_RB = 0x45, | |
495 REX_RX = 0x46, | |
496 REX_RXB = 0x47, | |
497 | |
498 REX_W = 0x48, | |
499 | |
500 REX_WB = 0x49, | |
501 REX_WX = 0x4A, | |
502 REX_WXB = 0x4B, | |
503 REX_WR = 0x4C, | |
504 REX_WRB = 0x4D, | |
505 REX_WRX = 0x4E, | |
506 REX_WRXB = 0x4F | |
507 }; | |
508 | |
509 enum WhichOperand { | |
510 // input to locate_operand, and format code for relocations | |
304 | 511 imm_operand = 0, // embedded 32-bit|64-bit immediate operand |
0 | 512 disp32_operand = 1, // embedded 32-bit displacement or address |
513 call32_operand = 2, // embedded 32-bit self-relative displacement | |
304 | 514 #ifndef _LP64 |
0 | 515 _WhichOperand_limit = 3 |
304 | 516 #else |
517 narrow_oop_operand = 3, // embedded 32-bit immediate narrow oop | |
518 _WhichOperand_limit = 4 | |
519 #endif | |
0 | 520 }; |
521 | |
304 | 522 |
523 | |
524 // NOTE: The general philopsophy of the declarations here is that 64bit versions | |
525 // of instructions are freely declared without the need for wrapping them an ifdef. | |
526 // (Some dangerous instructions are ifdef's out of inappropriate jvm's.) | |
527 // In the .cpp file the implementations are wrapped so that they are dropped out | |
528 // of the resulting jvm. This is done mostly to keep the footprint of KERNEL | |
529 // to the size it was prior to merging up the 32bit and 64bit assemblers. | |
530 // | |
531 // This does mean you'll get a linker/runtime error if you use a 64bit only instruction | |
532 // in a 32bit vm. This is somewhat unfortunate but keeps the ifdef noise down. | |
533 | |
534 private: | |
535 | |
536 | |
537 // 64bit prefixes | |
538 int prefix_and_encode(int reg_enc, bool byteinst = false); | |
539 int prefixq_and_encode(int reg_enc); | |
540 | |
541 int prefix_and_encode(int dst_enc, int src_enc, bool byteinst = false); | |
542 int prefixq_and_encode(int dst_enc, int src_enc); | |
543 | |
544 void prefix(Register reg); | |
545 void prefix(Address adr); | |
546 void prefixq(Address adr); | |
547 | |
548 void prefix(Address adr, Register reg, bool byteinst = false); | |
549 void prefixq(Address adr, Register reg); | |
550 | |
551 void prefix(Address adr, XMMRegister reg); | |
552 | |
553 void prefetch_prefix(Address src); | |
554 | |
555 // Helper functions for groups of instructions | |
556 void emit_arith_b(int op1, int op2, Register dst, int imm8); | |
557 | |
558 void emit_arith(int op1, int op2, Register dst, int32_t imm32); | |
559 // only 32bit?? | |
560 void emit_arith(int op1, int op2, Register dst, jobject obj); | |
561 void emit_arith(int op1, int op2, Register dst, Register src); | |
562 | |
563 void emit_operand(Register reg, | |
564 Register base, Register index, Address::ScaleFactor scale, | |
565 int disp, | |
566 RelocationHolder const& rspec, | |
567 int rip_relative_correction = 0); | |
568 | |
569 void emit_operand(Register reg, Address adr, int rip_relative_correction = 0); | |
570 | |
571 // operands that only take the original 32bit registers | |
572 void emit_operand32(Register reg, Address adr); | |
573 | |
574 void emit_operand(XMMRegister reg, | |
575 Register base, Register index, Address::ScaleFactor scale, | |
576 int disp, | |
577 RelocationHolder const& rspec); | |
578 | |
579 void emit_operand(XMMRegister reg, Address adr); | |
580 | |
581 void emit_operand(MMXRegister reg, Address adr); | |
582 | |
583 // workaround gcc (3.2.1-7) bug | |
584 void emit_operand(Address adr, MMXRegister reg); | |
585 | |
586 | |
587 // Immediate-to-memory forms | |
588 void emit_arith_operand(int op1, Register rm, Address adr, int32_t imm32); | |
589 | |
590 void emit_farith(int b1, int b2, int i); | |
591 | |
592 | |
593 protected: | |
594 #ifdef ASSERT | |
595 void check_relocation(RelocationHolder const& rspec, int format); | |
596 #endif | |
597 | |
598 inline void emit_long64(jlong x); | |
599 | |
600 void emit_data(jint data, relocInfo::relocType rtype, int format); | |
601 void emit_data(jint data, RelocationHolder const& rspec, int format); | |
602 void emit_data64(jlong data, relocInfo::relocType rtype, int format = 0); | |
603 void emit_data64(jlong data, RelocationHolder const& rspec, int format = 0); | |
604 | |
605 bool reachable(AddressLiteral adr) NOT_LP64({ return true;}); | |
606 | |
607 // These are all easily abused and hence protected | |
608 | |
609 // 32BIT ONLY SECTION | |
610 #ifndef _LP64 | |
611 // Make these disappear in 64bit mode since they would never be correct | |
612 void cmp_literal32(Register src1, int32_t imm32, RelocationHolder const& rspec); // 32BIT ONLY | |
613 void cmp_literal32(Address src1, int32_t imm32, RelocationHolder const& rspec); // 32BIT ONLY | |
614 | |
642
660978a2a31a
6791178: Specialize for zero as the compressed oop vm heap base
kvn
parents:
624
diff
changeset
|
615 void mov_literal32(Register dst, int32_t imm32, RelocationHolder const& rspec); // 32BIT ONLY |
304 | 616 void mov_literal32(Address dst, int32_t imm32, RelocationHolder const& rspec); // 32BIT ONLY |
617 | |
618 void push_literal32(int32_t imm32, RelocationHolder const& rspec); // 32BIT ONLY | |
619 #else | |
620 // 64BIT ONLY SECTION | |
621 void mov_literal64(Register dst, intptr_t imm64, RelocationHolder const& rspec); // 64BIT ONLY | |
642
660978a2a31a
6791178: Specialize for zero as the compressed oop vm heap base
kvn
parents:
624
diff
changeset
|
622 |
660978a2a31a
6791178: Specialize for zero as the compressed oop vm heap base
kvn
parents:
624
diff
changeset
|
623 void cmp_narrow_oop(Register src1, int32_t imm32, RelocationHolder const& rspec); |
660978a2a31a
6791178: Specialize for zero as the compressed oop vm heap base
kvn
parents:
624
diff
changeset
|
624 void cmp_narrow_oop(Address src1, int32_t imm32, RelocationHolder const& rspec); |
660978a2a31a
6791178: Specialize for zero as the compressed oop vm heap base
kvn
parents:
624
diff
changeset
|
625 |
660978a2a31a
6791178: Specialize for zero as the compressed oop vm heap base
kvn
parents:
624
diff
changeset
|
626 void mov_narrow_oop(Register dst, int32_t imm32, RelocationHolder const& rspec); |
660978a2a31a
6791178: Specialize for zero as the compressed oop vm heap base
kvn
parents:
624
diff
changeset
|
627 void mov_narrow_oop(Address dst, int32_t imm32, RelocationHolder const& rspec); |
304 | 628 #endif // _LP64 |
629 | |
630 // These are unique in that we are ensured by the caller that the 32bit | |
631 // relative in these instructions will always be able to reach the potentially | |
632 // 64bit address described by entry. Since they can take a 64bit address they | |
633 // don't have the 32 suffix like the other instructions in this class. | |
634 | |
635 void call_literal(address entry, RelocationHolder const& rspec); | |
636 void jmp_literal(address entry, RelocationHolder const& rspec); | |
637 | |
638 // Avoid using directly section | |
639 // Instructions in this section are actually usable by anyone without danger | |
640 // of failure but have performance issues that are addressed my enhanced | |
641 // instructions which will do the proper thing base on the particular cpu. | |
642 // We protect them because we don't trust you... | |
643 | |
644 // Don't use next inc() and dec() methods directly. INC & DEC instructions | |
645 // could cause a partial flag stall since they don't set CF flag. | |
646 // Use MacroAssembler::decrement() & MacroAssembler::increment() methods | |
647 // which call inc() & dec() or add() & sub() in accordance with | |
648 // the product flag UseIncDec value. | |
649 | |
650 void decl(Register dst); | |
651 void decl(Address dst); | |
652 void decq(Register dst); | |
653 void decq(Address dst); | |
654 | |
655 void incl(Register dst); | |
656 void incl(Address dst); | |
657 void incq(Register dst); | |
658 void incq(Address dst); | |
659 | |
660 // New cpus require use of movsd and movss to avoid partial register stall | |
661 // when loading from memory. But for old Opteron use movlpd instead of movsd. | |
662 // The selection is done in MacroAssembler::movdbl() and movflt(). | |
663 | |
664 // Move Scalar Single-Precision Floating-Point Values | |
665 void movss(XMMRegister dst, Address src); | |
666 void movss(XMMRegister dst, XMMRegister src); | |
667 void movss(Address dst, XMMRegister src); | |
668 | |
669 // Move Scalar Double-Precision Floating-Point Values | |
670 void movsd(XMMRegister dst, Address src); | |
671 void movsd(XMMRegister dst, XMMRegister src); | |
672 void movsd(Address dst, XMMRegister src); | |
673 void movlpd(XMMRegister dst, Address src); | |
674 | |
675 // New cpus require use of movaps and movapd to avoid partial register stall | |
676 // when moving between registers. | |
677 void movaps(XMMRegister dst, XMMRegister src); | |
678 void movapd(XMMRegister dst, XMMRegister src); | |
679 | |
680 // End avoid using directly | |
681 | |
682 | |
683 // Instruction prefixes | |
684 void prefix(Prefix p); | |
685 | |
0 | 686 public: |
687 | |
688 // Creation | |
689 Assembler(CodeBuffer* code) : AbstractAssembler(code) {} | |
690 | |
691 // Decoding | |
692 static address locate_operand(address inst, WhichOperand which); | |
693 static address locate_next_instruction(address inst); | |
694 | |
304 | 695 // Utilities |
2404
b40d4fa697bf
6964776: c2 should ensure the polling page is reachable on 64 bit
iveresov
parents:
2320
diff
changeset
|
696 static bool is_polling_page_far() NOT_LP64({ return false;}); |
b40d4fa697bf
6964776: c2 should ensure the polling page is reachable on 64 bit
iveresov
parents:
2320
diff
changeset
|
697 |
304 | 698 // Generic instructions |
699 // Does 32bit or 64bit as needed for the platform. In some sense these | |
700 // belong in macro assembler but there is no need for both varieties to exist | |
701 | |
702 void lea(Register dst, Address src); | |
703 | |
704 void mov(Register dst, Register src); | |
705 | |
706 void pusha(); | |
707 void popa(); | |
708 | |
709 void pushf(); | |
710 void popf(); | |
711 | |
712 void push(int32_t imm32); | |
713 | |
714 void push(Register src); | |
715 | |
716 void pop(Register dst); | |
717 | |
718 // These are dummies to prevent surprise implicit conversions to Register | |
719 void push(void* v); | |
720 void pop(void* v); | |
721 | |
722 // These do register sized moves/scans | |
723 void rep_mov(); | |
724 void rep_set(); | |
725 void repne_scan(); | |
726 #ifdef _LP64 | |
727 void repne_scanl(); | |
728 #endif | |
729 | |
730 // Vanilla instructions in lexical order | |
731 | |
2100
b1a2afa37ec4
7003271: Hotspot should track cumulative Java heap bytes allocated on a per-thread basis
phh
parents:
2008
diff
changeset
|
732 void adcl(Address dst, int32_t imm32); |
b1a2afa37ec4
7003271: Hotspot should track cumulative Java heap bytes allocated on a per-thread basis
phh
parents:
2008
diff
changeset
|
733 void adcl(Address dst, Register src); |
304 | 734 void adcl(Register dst, int32_t imm32); |
0 | 735 void adcl(Register dst, Address src); |
736 void adcl(Register dst, Register src); | |
737 | |
304 | 738 void adcq(Register dst, int32_t imm32); |
739 void adcq(Register dst, Address src); | |
740 void adcq(Register dst, Register src); | |
741 | |
742 void addl(Address dst, int32_t imm32); | |
0 | 743 void addl(Address dst, Register src); |
304 | 744 void addl(Register dst, int32_t imm32); |
0 | 745 void addl(Register dst, Address src); |
746 void addl(Register dst, Register src); | |
747 | |
304 | 748 void addq(Address dst, int32_t imm32); |
749 void addq(Address dst, Register src); | |
750 void addq(Register dst, int32_t imm32); | |
751 void addq(Register dst, Address src); | |
752 void addq(Register dst, Register src); | |
753 | |
0 | 754 void addr_nop_4(); |
755 void addr_nop_5(); | |
756 void addr_nop_7(); | |
757 void addr_nop_8(); | |
758 | |
304 | 759 // Add Scalar Double-Precision Floating-Point Values |
760 void addsd(XMMRegister dst, Address src); | |
761 void addsd(XMMRegister dst, XMMRegister src); | |
762 | |
763 // Add Scalar Single-Precision Floating-Point Values | |
764 void addss(XMMRegister dst, Address src); | |
765 void addss(XMMRegister dst, XMMRegister src); | |
766 | |
767 void andl(Register dst, int32_t imm32); | |
768 void andl(Register dst, Address src); | |
769 void andl(Register dst, Register src); | |
770 | |
3783 | 771 void andq(Address dst, int32_t imm32); |
304 | 772 void andq(Register dst, int32_t imm32); |
773 void andq(Register dst, Address src); | |
774 void andq(Register dst, Register src); | |
775 | |
776 // Bitwise Logical AND of Packed Double-Precision Floating-Point Values | |
777 void andpd(XMMRegister dst, Address src); | |
778 void andpd(XMMRegister dst, XMMRegister src); | |
779 | |
775
93c14e5562c4
6823354: Add intrinsics for {Integer,Long}.{numberOfLeadingZeros,numberOfTrailingZeros}()
twisti
parents:
710
diff
changeset
|
780 void bsfl(Register dst, Register src); |
93c14e5562c4
6823354: Add intrinsics for {Integer,Long}.{numberOfLeadingZeros,numberOfTrailingZeros}()
twisti
parents:
710
diff
changeset
|
781 void bsrl(Register dst, Register src); |
93c14e5562c4
6823354: Add intrinsics for {Integer,Long}.{numberOfLeadingZeros,numberOfTrailingZeros}()
twisti
parents:
710
diff
changeset
|
782 |
93c14e5562c4
6823354: Add intrinsics for {Integer,Long}.{numberOfLeadingZeros,numberOfTrailingZeros}()
twisti
parents:
710
diff
changeset
|
783 #ifdef _LP64 |
93c14e5562c4
6823354: Add intrinsics for {Integer,Long}.{numberOfLeadingZeros,numberOfTrailingZeros}()
twisti
parents:
710
diff
changeset
|
784 void bsfq(Register dst, Register src); |
93c14e5562c4
6823354: Add intrinsics for {Integer,Long}.{numberOfLeadingZeros,numberOfTrailingZeros}()
twisti
parents:
710
diff
changeset
|
785 void bsrq(Register dst, Register src); |
93c14e5562c4
6823354: Add intrinsics for {Integer,Long}.{numberOfLeadingZeros,numberOfTrailingZeros}()
twisti
parents:
710
diff
changeset
|
786 #endif |
93c14e5562c4
6823354: Add intrinsics for {Integer,Long}.{numberOfLeadingZeros,numberOfTrailingZeros}()
twisti
parents:
710
diff
changeset
|
787 |
304 | 788 void bswapl(Register reg); |
789 | |
790 void bswapq(Register reg); | |
791 | |
0 | 792 void call(Label& L, relocInfo::relocType rtype); |
793 void call(Register reg); // push pc; pc <- reg | |
794 void call(Address adr); // push pc; pc <- adr | |
795 | |
304 | 796 void cdql(); |
797 | |
798 void cdqq(); | |
799 | |
800 void cld() { emit_byte(0xfc); } | |
801 | |
802 void clflush(Address adr); | |
803 | |
804 void cmovl(Condition cc, Register dst, Register src); | |
805 void cmovl(Condition cc, Register dst, Address src); | |
806 | |
807 void cmovq(Condition cc, Register dst, Register src); | |
808 void cmovq(Condition cc, Register dst, Address src); | |
809 | |
810 | |
811 void cmpb(Address dst, int imm8); | |
812 | |
813 void cmpl(Address dst, int32_t imm32); | |
814 | |
815 void cmpl(Register dst, int32_t imm32); | |
816 void cmpl(Register dst, Register src); | |
817 void cmpl(Register dst, Address src); | |
818 | |
819 void cmpq(Address dst, int32_t imm32); | |
820 void cmpq(Address dst, Register src); | |
821 | |
822 void cmpq(Register dst, int32_t imm32); | |
823 void cmpq(Register dst, Register src); | |
824 void cmpq(Register dst, Address src); | |
825 | |
826 // these are dummies used to catch attempting to convert NULL to Register | |
827 void cmpl(Register dst, void* junk); // dummy | |
828 void cmpq(Register dst, void* junk); // dummy | |
829 | |
830 void cmpw(Address dst, int imm16); | |
831 | |
832 void cmpxchg8 (Address adr); | |
833 | |
834 void cmpxchgl(Register reg, Address adr); | |
835 | |
836 void cmpxchgq(Register reg, Address adr); | |
837 | |
838 // Ordered Compare Scalar Double-Precision Floating-Point Values and set EFLAGS | |
839 void comisd(XMMRegister dst, Address src); | |
840 | |
841 // Ordered Compare Scalar Single-Precision Floating-Point Values and set EFLAGS | |
842 void comiss(XMMRegister dst, Address src); | |
843 | |
844 // Identify processor type and features | |
845 void cpuid() { | |
846 emit_byte(0x0F); | |
847 emit_byte(0xA2); | |
848 } | |
849 | |
850 // Convert Scalar Double-Precision Floating-Point Value to Scalar Single-Precision Floating-Point Value | |
851 void cvtsd2ss(XMMRegister dst, XMMRegister src); | |
852 | |
853 // Convert Doubleword Integer to Scalar Double-Precision Floating-Point Value | |
854 void cvtsi2sdl(XMMRegister dst, Register src); | |
855 void cvtsi2sdq(XMMRegister dst, Register src); | |
856 | |
857 // Convert Doubleword Integer to Scalar Single-Precision Floating-Point Value | |
858 void cvtsi2ssl(XMMRegister dst, Register src); | |
859 void cvtsi2ssq(XMMRegister dst, Register src); | |
860 | |
861 // Convert Packed Signed Doubleword Integers to Packed Double-Precision Floating-Point Value | |
862 void cvtdq2pd(XMMRegister dst, XMMRegister src); | |
863 | |
864 // Convert Packed Signed Doubleword Integers to Packed Single-Precision Floating-Point Value | |
865 void cvtdq2ps(XMMRegister dst, XMMRegister src); | |
866 | |
867 // Convert Scalar Single-Precision Floating-Point Value to Scalar Double-Precision Floating-Point Value | |
868 void cvtss2sd(XMMRegister dst, XMMRegister src); | |
869 | |
870 // Convert with Truncation Scalar Double-Precision Floating-Point Value to Doubleword Integer | |
871 void cvttsd2sil(Register dst, Address src); | |
872 void cvttsd2sil(Register dst, XMMRegister src); | |
873 void cvttsd2siq(Register dst, XMMRegister src); | |
874 | |
875 // Convert with Truncation Scalar Single-Precision Floating-Point Value to Doubleword Integer | |
876 void cvttss2sil(Register dst, XMMRegister src); | |
877 void cvttss2siq(Register dst, XMMRegister src); | |
878 | |
879 // Divide Scalar Double-Precision Floating-Point Values | |
880 void divsd(XMMRegister dst, Address src); | |
881 void divsd(XMMRegister dst, XMMRegister src); | |
882 | |
883 // Divide Scalar Single-Precision Floating-Point Values | |
884 void divss(XMMRegister dst, Address src); | |
885 void divss(XMMRegister dst, XMMRegister src); | |
886 | |
887 void emms(); | |
888 | |
889 void fabs(); | |
890 | |
891 void fadd(int i); | |
892 | |
893 void fadd_d(Address src); | |
894 void fadd_s(Address src); | |
895 | |
896 // "Alternate" versions of x87 instructions place result down in FPU | |
897 // stack instead of on TOS | |
898 | |
899 void fadda(int i); // "alternate" fadd | |
900 void faddp(int i = 1); | |
901 | |
902 void fchs(); | |
903 | |
904 void fcom(int i); | |
905 | |
906 void fcomp(int i = 1); | |
907 void fcomp_d(Address src); | |
908 void fcomp_s(Address src); | |
909 | |
910 void fcompp(); | |
911 | |
912 void fcos(); | |
913 | |
914 void fdecstp(); | |
915 | |
916 void fdiv(int i); | |
917 void fdiv_d(Address src); | |
918 void fdivr_s(Address src); | |
919 void fdiva(int i); // "alternate" fdiv | |
920 void fdivp(int i = 1); | |
921 | |
922 void fdivr(int i); | |
923 void fdivr_d(Address src); | |
924 void fdiv_s(Address src); | |
925 | |
926 void fdivra(int i); // "alternate" reversed fdiv | |
927 | |
928 void fdivrp(int i = 1); | |
929 | |
930 void ffree(int i = 0); | |
931 | |
932 void fild_d(Address adr); | |
933 void fild_s(Address adr); | |
934 | |
935 void fincstp(); | |
936 | |
937 void finit(); | |
938 | |
939 void fist_s (Address adr); | |
940 void fistp_d(Address adr); | |
941 void fistp_s(Address adr); | |
942 | |
943 void fld1(); | |
944 | |
945 void fld_d(Address adr); | |
946 void fld_s(Address adr); | |
947 void fld_s(int index); | |
948 void fld_x(Address adr); // extended-precision (80-bit) format | |
949 | |
950 void fldcw(Address src); | |
951 | |
952 void fldenv(Address src); | |
953 | |
954 void fldlg2(); | |
955 | |
956 void fldln2(); | |
957 | |
958 void fldz(); | |
959 | |
960 void flog(); | |
961 void flog10(); | |
962 | |
963 void fmul(int i); | |
964 | |
965 void fmul_d(Address src); | |
966 void fmul_s(Address src); | |
967 | |
968 void fmula(int i); // "alternate" fmul | |
969 | |
970 void fmulp(int i = 1); | |
971 | |
972 void fnsave(Address dst); | |
973 | |
974 void fnstcw(Address src); | |
975 | |
976 void fnstsw_ax(); | |
977 | |
978 void fprem(); | |
979 void fprem1(); | |
980 | |
981 void frstor(Address src); | |
982 | |
983 void fsin(); | |
984 | |
985 void fsqrt(); | |
986 | |
987 void fst_d(Address adr); | |
988 void fst_s(Address adr); | |
989 | |
990 void fstp_d(Address adr); | |
991 void fstp_d(int index); | |
992 void fstp_s(Address adr); | |
993 void fstp_x(Address adr); // extended-precision (80-bit) format | |
994 | |
995 void fsub(int i); | |
996 void fsub_d(Address src); | |
997 void fsub_s(Address src); | |
998 | |
999 void fsuba(int i); // "alternate" fsub | |
1000 | |
1001 void fsubp(int i = 1); | |
1002 | |
1003 void fsubr(int i); | |
1004 void fsubr_d(Address src); | |
1005 void fsubr_s(Address src); | |
1006 | |
1007 void fsubra(int i); // "alternate" reversed fsub | |
1008 | |
1009 void fsubrp(int i = 1); | |
1010 | |
1011 void ftan(); | |
1012 | |
1013 void ftst(); | |
1014 | |
1015 void fucomi(int i = 1); | |
1016 void fucomip(int i = 1); | |
1017 | |
1018 void fwait(); | |
1019 | |
1020 void fxch(int i = 1); | |
1021 | |
1022 void fxrstor(Address src); | |
1023 | |
1024 void fxsave(Address dst); | |
1025 | |
1026 void fyl2x(); | |
1027 | |
1028 void hlt(); | |
1029 | |
1030 void idivl(Register src); | |
1920 | 1031 void divl(Register src); // Unsigned division |
304 | 1032 |
1033 void idivq(Register src); | |
1034 | |
1035 void imull(Register dst, Register src); | |
1036 void imull(Register dst, Register src, int value); | |
1037 | |
1038 void imulq(Register dst, Register src); | |
1039 void imulq(Register dst, Register src, int value); | |
1040 | |
0 | 1041 |
1042 // jcc is the generic conditional branch generator to run- | |
1043 // time routines, jcc is used for branches to labels. jcc | |
1044 // takes a branch opcode (cc) and a label (L) and generates | |
1045 // either a backward branch or a forward branch and links it | |
1046 // to the label fixup chain. Usage: | |
1047 // | |
1048 // Label L; // unbound label | |
1049 // jcc(cc, L); // forward branch to unbound label | |
1050 // bind(L); // bind label to the current pc | |
1051 // jcc(cc, L); // backward branch to bound label | |
1052 // bind(L); // illegal: a label may be bound only once | |
1053 // | |
1054 // Note: The same Label can be used for forward and backward branches | |
1055 // but it may be bound only once. | |
1056 | |
3851 | 1057 void jcc(Condition cc, Label& L, bool maybe_short = true); |
0 | 1058 |
1059 // Conditional jump to a 8-bit offset to L. | |
1060 // WARNING: be very careful using this for forward jumps. If the label is | |
1061 // not bound within an 8-bit offset of this instruction, a run-time error | |
1062 // will occur. | |
1063 void jccb(Condition cc, Label& L); | |
1064 | |
304 | 1065 void jmp(Address entry); // pc <- entry |
1066 | |
1067 // Label operations & relative jumps (PPUM Appendix D) | |
3851 | 1068 void jmp(Label& L, bool maybe_short = true); // unconditional jump to L |
304 | 1069 |
1070 void jmp(Register entry); // pc <- entry | |
1071 | |
1072 // Unconditional 8-bit offset jump to L. | |
1073 // WARNING: be very careful using this for forward jumps. If the label is | |
1074 // not bound within an 8-bit offset of this instruction, a run-time error | |
1075 // will occur. | |
1076 void jmpb(Label& L); | |
1077 | |
1078 void ldmxcsr( Address src ); | |
1079 | |
1080 void leal(Register dst, Address src); | |
1081 | |
1082 void leaq(Register dst, Address src); | |
1083 | |
1084 void lfence() { | |
1085 emit_byte(0x0F); | |
1086 emit_byte(0xAE); | |
1087 emit_byte(0xE8); | |
1088 } | |
1089 | |
1090 void lock(); | |
1091 | |
775
93c14e5562c4
6823354: Add intrinsics for {Integer,Long}.{numberOfLeadingZeros,numberOfTrailingZeros}()
twisti
parents:
710
diff
changeset
|
1092 void lzcntl(Register dst, Register src); |
93c14e5562c4
6823354: Add intrinsics for {Integer,Long}.{numberOfLeadingZeros,numberOfTrailingZeros}()
twisti
parents:
710
diff
changeset
|
1093 |
93c14e5562c4
6823354: Add intrinsics for {Integer,Long}.{numberOfLeadingZeros,numberOfTrailingZeros}()
twisti
parents:
710
diff
changeset
|
1094 #ifdef _LP64 |
93c14e5562c4
6823354: Add intrinsics for {Integer,Long}.{numberOfLeadingZeros,numberOfTrailingZeros}()
twisti
parents:
710
diff
changeset
|
1095 void lzcntq(Register dst, Register src); |
93c14e5562c4
6823354: Add intrinsics for {Integer,Long}.{numberOfLeadingZeros,numberOfTrailingZeros}()
twisti
parents:
710
diff
changeset
|
1096 #endif |
93c14e5562c4
6823354: Add intrinsics for {Integer,Long}.{numberOfLeadingZeros,numberOfTrailingZeros}()
twisti
parents:
710
diff
changeset
|
1097 |
304 | 1098 enum Membar_mask_bits { |
1099 StoreStore = 1 << 3, | |
1100 LoadStore = 1 << 2, | |
1101 StoreLoad = 1 << 1, | |
1102 LoadLoad = 1 << 0 | |
1103 }; | |
1104 | |
671
d0994e5bebce
6822204: volatile fences should prefer lock:addl to actual mfence instructions
never
parents:
665
diff
changeset
|
1105 // Serializes memory and blows flags |
304 | 1106 void membar(Membar_mask_bits order_constraint) { |
671
d0994e5bebce
6822204: volatile fences should prefer lock:addl to actual mfence instructions
never
parents:
665
diff
changeset
|
1107 if (os::is_MP()) { |
d0994e5bebce
6822204: volatile fences should prefer lock:addl to actual mfence instructions
never
parents:
665
diff
changeset
|
1108 // We only have to handle StoreLoad |
d0994e5bebce
6822204: volatile fences should prefer lock:addl to actual mfence instructions
never
parents:
665
diff
changeset
|
1109 if (order_constraint & StoreLoad) { |
d0994e5bebce
6822204: volatile fences should prefer lock:addl to actual mfence instructions
never
parents:
665
diff
changeset
|
1110 // All usable chips support "locked" instructions which suffice |
d0994e5bebce
6822204: volatile fences should prefer lock:addl to actual mfence instructions
never
parents:
665
diff
changeset
|
1111 // as barriers, and are much faster than the alternative of |
d0994e5bebce
6822204: volatile fences should prefer lock:addl to actual mfence instructions
never
parents:
665
diff
changeset
|
1112 // using cpuid instruction. We use here a locked add [esp],0. |
d0994e5bebce
6822204: volatile fences should prefer lock:addl to actual mfence instructions
never
parents:
665
diff
changeset
|
1113 // This is conveniently otherwise a no-op except for blowing |
d0994e5bebce
6822204: volatile fences should prefer lock:addl to actual mfence instructions
never
parents:
665
diff
changeset
|
1114 // flags. |
d0994e5bebce
6822204: volatile fences should prefer lock:addl to actual mfence instructions
never
parents:
665
diff
changeset
|
1115 // Any change to this code may need to revisit other places in |
d0994e5bebce
6822204: volatile fences should prefer lock:addl to actual mfence instructions
never
parents:
665
diff
changeset
|
1116 // the code where this idiom is used, in particular the |
d0994e5bebce
6822204: volatile fences should prefer lock:addl to actual mfence instructions
never
parents:
665
diff
changeset
|
1117 // orderAccess code. |
d0994e5bebce
6822204: volatile fences should prefer lock:addl to actual mfence instructions
never
parents:
665
diff
changeset
|
1118 lock(); |
d0994e5bebce
6822204: volatile fences should prefer lock:addl to actual mfence instructions
never
parents:
665
diff
changeset
|
1119 addl(Address(rsp, 0), 0);// Assert the lock# signal here |
d0994e5bebce
6822204: volatile fences should prefer lock:addl to actual mfence instructions
never
parents:
665
diff
changeset
|
1120 } |
d0994e5bebce
6822204: volatile fences should prefer lock:addl to actual mfence instructions
never
parents:
665
diff
changeset
|
1121 } |
304 | 1122 } |
1123 | |
1124 void mfence(); | |
1125 | |
1126 // Moves | |
1127 | |
1128 void mov64(Register dst, int64_t imm64); | |
1129 | |
1130 void movb(Address dst, Register src); | |
1131 void movb(Address dst, int imm8); | |
1132 void movb(Register dst, Address src); | |
1133 | |
1134 void movdl(XMMRegister dst, Register src); | |
1135 void movdl(Register dst, XMMRegister src); | |
2320
41d4973cf100
6942326: x86 code in string_indexof() could read beyond reserved heap space
kvn
parents:
2262
diff
changeset
|
1136 void movdl(XMMRegister dst, Address src); |
304 | 1137 |
1138 // Move Double Quadword | |
1139 void movdq(XMMRegister dst, Register src); | |
1140 void movdq(Register dst, XMMRegister src); | |
1141 | |
1142 // Move Aligned Double Quadword | |
1143 void movdqa(Address dst, XMMRegister src); | |
1144 void movdqa(XMMRegister dst, Address src); | |
1145 void movdqa(XMMRegister dst, XMMRegister src); | |
1146 | |
405 | 1147 // Move Unaligned Double Quadword |
1148 void movdqu(Address dst, XMMRegister src); | |
1149 void movdqu(XMMRegister dst, Address src); | |
1150 void movdqu(XMMRegister dst, XMMRegister src); | |
1151 | |
304 | 1152 void movl(Register dst, int32_t imm32); |
1153 void movl(Address dst, int32_t imm32); | |
1154 void movl(Register dst, Register src); | |
1155 void movl(Register dst, Address src); | |
1156 void movl(Address dst, Register src); | |
1157 | |
1158 // These dummies prevent using movl from converting a zero (like NULL) into Register | |
1159 // by giving the compiler two choices it can't resolve | |
1160 | |
1161 void movl(Address dst, void* junk); | |
1162 void movl(Register dst, void* junk); | |
1163 | |
1164 #ifdef _LP64 | |
1165 void movq(Register dst, Register src); | |
1166 void movq(Register dst, Address src); | |
2100
b1a2afa37ec4
7003271: Hotspot should track cumulative Java heap bytes allocated on a per-thread basis
phh
parents:
2008
diff
changeset
|
1167 void movq(Address dst, Register src); |
304 | 1168 #endif |
1169 | |
1170 void movq(Address dst, MMXRegister src ); | |
1171 void movq(MMXRegister dst, Address src ); | |
1172 | |
1173 #ifdef _LP64 | |
1174 // These dummies prevent using movq from converting a zero (like NULL) into Register | |
1175 // by giving the compiler two choices it can't resolve | |
1176 | |
1177 void movq(Address dst, void* dummy); | |
1178 void movq(Register dst, void* dummy); | |
1179 #endif | |
1180 | |
1181 // Move Quadword | |
1182 void movq(Address dst, XMMRegister src); | |
1183 void movq(XMMRegister dst, Address src); | |
1184 | |
1185 void movsbl(Register dst, Address src); | |
1186 void movsbl(Register dst, Register src); | |
1187 | |
1188 #ifdef _LP64 | |
624 | 1189 void movsbq(Register dst, Address src); |
1190 void movsbq(Register dst, Register src); | |
1191 | |
304 | 1192 // Move signed 32bit immediate to 64bit extending sign |
2100
b1a2afa37ec4
7003271: Hotspot should track cumulative Java heap bytes allocated on a per-thread basis
phh
parents:
2008
diff
changeset
|
1193 void movslq(Address dst, int32_t imm64); |
304 | 1194 void movslq(Register dst, int32_t imm64); |
1195 | |
1196 void movslq(Register dst, Address src); | |
1197 void movslq(Register dst, Register src); | |
1198 void movslq(Register dst, void* src); // Dummy declaration to cause NULL to be ambiguous | |
1199 #endif | |
1200 | |
1201 void movswl(Register dst, Address src); | |
1202 void movswl(Register dst, Register src); | |
1203 | |
624 | 1204 #ifdef _LP64 |
1205 void movswq(Register dst, Address src); | |
1206 void movswq(Register dst, Register src); | |
1207 #endif | |
1208 | |
304 | 1209 void movw(Address dst, int imm16); |
1210 void movw(Register dst, Address src); | |
1211 void movw(Address dst, Register src); | |
1212 | |
1213 void movzbl(Register dst, Address src); | |
1214 void movzbl(Register dst, Register src); | |
1215 | |
624 | 1216 #ifdef _LP64 |
1217 void movzbq(Register dst, Address src); | |
1218 void movzbq(Register dst, Register src); | |
1219 #endif | |
1220 | |
304 | 1221 void movzwl(Register dst, Address src); |
1222 void movzwl(Register dst, Register src); | |
1223 | |
624 | 1224 #ifdef _LP64 |
1225 void movzwq(Register dst, Address src); | |
1226 void movzwq(Register dst, Register src); | |
1227 #endif | |
1228 | |
304 | 1229 void mull(Address src); |
1230 void mull(Register src); | |
1231 | |
1232 // Multiply Scalar Double-Precision Floating-Point Values | |
1233 void mulsd(XMMRegister dst, Address src); | |
1234 void mulsd(XMMRegister dst, XMMRegister src); | |
1235 | |
1236 // Multiply Scalar Single-Precision Floating-Point Values | |
1237 void mulss(XMMRegister dst, Address src); | |
1238 void mulss(XMMRegister dst, XMMRegister src); | |
1239 | |
1240 void negl(Register dst); | |
1241 | |
1242 #ifdef _LP64 | |
1243 void negq(Register dst); | |
1244 #endif | |
1245 | |
1246 void nop(int i = 1); | |
1247 | |
1248 void notl(Register dst); | |
1249 | |
1250 #ifdef _LP64 | |
1251 void notq(Register dst); | |
1252 #endif | |
1253 | |
1254 void orl(Address dst, int32_t imm32); | |
1255 void orl(Register dst, int32_t imm32); | |
1256 void orl(Register dst, Address src); | |
1257 void orl(Register dst, Register src); | |
1258 | |
1259 void orq(Address dst, int32_t imm32); | |
1260 void orq(Register dst, int32_t imm32); | |
1261 void orq(Register dst, Address src); | |
1262 void orq(Register dst, Register src); | |
1263 | |
681 | 1264 // SSE4.2 string instructions |
1265 void pcmpestri(XMMRegister xmm1, XMMRegister xmm2, int imm8); | |
1266 void pcmpestri(XMMRegister xmm1, Address src, int imm8); | |
1267 | |
1060 | 1268 #ifndef _LP64 // no 32bit push/pop on amd64 |
304 | 1269 void popl(Address dst); |
1060 | 1270 #endif |
304 | 1271 |
1272 #ifdef _LP64 | |
1273 void popq(Address dst); | |
1274 #endif | |
1275 | |
643
c771b7f43bbf
6378821: bitCount() should use POPC on SPARC processors and AMD+10h
twisti
parents:
642
diff
changeset
|
1276 void popcntl(Register dst, Address src); |
c771b7f43bbf
6378821: bitCount() should use POPC on SPARC processors and AMD+10h
twisti
parents:
642
diff
changeset
|
1277 void popcntl(Register dst, Register src); |
c771b7f43bbf
6378821: bitCount() should use POPC on SPARC processors and AMD+10h
twisti
parents:
642
diff
changeset
|
1278 |
c771b7f43bbf
6378821: bitCount() should use POPC on SPARC processors and AMD+10h
twisti
parents:
642
diff
changeset
|
1279 #ifdef _LP64 |
c771b7f43bbf
6378821: bitCount() should use POPC on SPARC processors and AMD+10h
twisti
parents:
642
diff
changeset
|
1280 void popcntq(Register dst, Address src); |
c771b7f43bbf
6378821: bitCount() should use POPC on SPARC processors and AMD+10h
twisti
parents:
642
diff
changeset
|
1281 void popcntq(Register dst, Register src); |
c771b7f43bbf
6378821: bitCount() should use POPC on SPARC processors and AMD+10h
twisti
parents:
642
diff
changeset
|
1282 #endif |
c771b7f43bbf
6378821: bitCount() should use POPC on SPARC processors and AMD+10h
twisti
parents:
642
diff
changeset
|
1283 |
304 | 1284 // Prefetches (SSE, SSE2, 3DNOW only) |
1285 | |
1286 void prefetchnta(Address src); | |
1287 void prefetchr(Address src); | |
1288 void prefetcht0(Address src); | |
1289 void prefetcht1(Address src); | |
1290 void prefetcht2(Address src); | |
1291 void prefetchw(Address src); | |
1292 | |
2262 | 1293 // POR - Bitwise logical OR |
1294 void por(XMMRegister dst, XMMRegister src); | |
1295 | |
304 | 1296 // Shuffle Packed Doublewords |
1297 void pshufd(XMMRegister dst, XMMRegister src, int mode); | |
1298 void pshufd(XMMRegister dst, Address src, int mode); | |
1299 | |
1300 // Shuffle Packed Low Words | |
1301 void pshuflw(XMMRegister dst, XMMRegister src, int mode); | |
1302 void pshuflw(XMMRegister dst, Address src, int mode); | |
1303 | |
2320
41d4973cf100
6942326: x86 code in string_indexof() could read beyond reserved heap space
kvn
parents:
2262
diff
changeset
|
1304 // Shift Right by bits Logical Quadword Immediate |
304 | 1305 void psrlq(XMMRegister dst, int shift); |
1306 | |
2320
41d4973cf100
6942326: x86 code in string_indexof() could read beyond reserved heap space
kvn
parents:
2262
diff
changeset
|
1307 // Shift Right by bytes Logical DoubleQuadword Immediate |
41d4973cf100
6942326: x86 code in string_indexof() could read beyond reserved heap space
kvn
parents:
2262
diff
changeset
|
1308 void psrldq(XMMRegister dst, int shift); |
41d4973cf100
6942326: x86 code in string_indexof() could read beyond reserved heap space
kvn
parents:
2262
diff
changeset
|
1309 |
681 | 1310 // Logical Compare Double Quadword |
1311 void ptest(XMMRegister dst, XMMRegister src); | |
1312 void ptest(XMMRegister dst, Address src); | |
1313 | |
304 | 1314 // Interleave Low Bytes |
1315 void punpcklbw(XMMRegister dst, XMMRegister src); | |
1316 | |
1060 | 1317 #ifndef _LP64 // no 32bit push/pop on amd64 |
304 | 1318 void pushl(Address src); |
1060 | 1319 #endif |
304 | 1320 |
1321 void pushq(Address src); | |
1322 | |
1323 // Xor Packed Byte Integer Values | |
1324 void pxor(XMMRegister dst, Address src); | |
1325 void pxor(XMMRegister dst, XMMRegister src); | |
1326 | |
1327 void rcll(Register dst, int imm8); | |
1328 | |
1329 void rclq(Register dst, int imm8); | |
1330 | |
1331 void ret(int imm16); | |
0 | 1332 |
1333 void sahf(); | |
1334 | |
304 | 1335 void sarl(Register dst, int imm8); |
1336 void sarl(Register dst); | |
1337 | |
1338 void sarq(Register dst, int imm8); | |
1339 void sarq(Register dst); | |
1340 | |
1341 void sbbl(Address dst, int32_t imm32); | |
1342 void sbbl(Register dst, int32_t imm32); | |
1343 void sbbl(Register dst, Address src); | |
1344 void sbbl(Register dst, Register src); | |
1345 | |
1346 void sbbq(Address dst, int32_t imm32); | |
1347 void sbbq(Register dst, int32_t imm32); | |
1348 void sbbq(Register dst, Address src); | |
1349 void sbbq(Register dst, Register src); | |
1350 | |
1351 void setb(Condition cc, Register dst); | |
1352 | |
1353 void shldl(Register dst, Register src); | |
1354 | |
1355 void shll(Register dst, int imm8); | |
1356 void shll(Register dst); | |
1357 | |
1358 void shlq(Register dst, int imm8); | |
1359 void shlq(Register dst); | |
1360 | |
1361 void shrdl(Register dst, Register src); | |
1362 | |
1363 void shrl(Register dst, int imm8); | |
1364 void shrl(Register dst); | |
1365 | |
1366 void shrq(Register dst, int imm8); | |
1367 void shrq(Register dst); | |
1368 | |
1369 void smovl(); // QQQ generic? | |
1370 | |
1371 // Compute Square Root of Scalar Double-Precision Floating-Point Value | |
1372 void sqrtsd(XMMRegister dst, Address src); | |
1373 void sqrtsd(XMMRegister dst, XMMRegister src); | |
1374 | |
2008 | 1375 // Compute Square Root of Scalar Single-Precision Floating-Point Value |
1376 void sqrtss(XMMRegister dst, Address src); | |
1377 void sqrtss(XMMRegister dst, XMMRegister src); | |
1378 | |
304 | 1379 void std() { emit_byte(0xfd); } |
1380 | |
1381 void stmxcsr( Address dst ); | |
1382 | |
1383 void subl(Address dst, int32_t imm32); | |
1384 void subl(Address dst, Register src); | |
1385 void subl(Register dst, int32_t imm32); | |
1386 void subl(Register dst, Address src); | |
1387 void subl(Register dst, Register src); | |
1388 | |
1389 void subq(Address dst, int32_t imm32); | |
1390 void subq(Address dst, Register src); | |
1391 void subq(Register dst, int32_t imm32); | |
1392 void subq(Register dst, Address src); | |
1393 void subq(Register dst, Register src); | |
1394 | |
1395 | |
1396 // Subtract Scalar Double-Precision Floating-Point Values | |
1397 void subsd(XMMRegister dst, Address src); | |
0 | 1398 void subsd(XMMRegister dst, XMMRegister src); |
1399 | |
304 | 1400 // Subtract Scalar Single-Precision Floating-Point Values |
1401 void subss(XMMRegister dst, Address src); | |
1402 void subss(XMMRegister dst, XMMRegister src); | |
1403 | |
1404 void testb(Register dst, int imm8); | |
1405 | |
1406 void testl(Register dst, int32_t imm32); | |
1407 void testl(Register dst, Register src); | |
1408 void testl(Register dst, Address src); | |
1409 | |
1410 void testq(Register dst, int32_t imm32); | |
1411 void testq(Register dst, Register src); | |
1412 | |
1413 | |
1414 // Unordered Compare Scalar Double-Precision Floating-Point Values and set EFLAGS | |
1415 void ucomisd(XMMRegister dst, Address src); | |
0 | 1416 void ucomisd(XMMRegister dst, XMMRegister src); |
1417 | |
304 | 1418 // Unordered Compare Scalar Single-Precision Floating-Point Values and set EFLAGS |
1419 void ucomiss(XMMRegister dst, Address src); | |
1420 void ucomiss(XMMRegister dst, XMMRegister src); | |
1421 | |
1422 void xaddl(Address dst, Register src); | |
1423 | |
1424 void xaddq(Address dst, Register src); | |
1425 | |
1426 void xchgl(Register reg, Address adr); | |
1427 void xchgl(Register dst, Register src); | |
1428 | |
1429 void xchgq(Register reg, Address adr); | |
1430 void xchgq(Register dst, Register src); | |
1431 | |
1432 void xorl(Register dst, int32_t imm32); | |
1433 void xorl(Register dst, Address src); | |
1434 void xorl(Register dst, Register src); | |
1435 | |
1436 void xorq(Register dst, Address src); | |
1437 void xorq(Register dst, Register src); | |
1438 | |
1439 // Bitwise Logical XOR of Packed Double-Precision Floating-Point Values | |
1440 void xorpd(XMMRegister dst, Address src); | |
1441 void xorpd(XMMRegister dst, XMMRegister src); | |
1442 | |
1443 // Bitwise Logical XOR of Packed Single-Precision Floating-Point Values | |
1444 void xorps(XMMRegister dst, Address src); | |
0 | 1445 void xorps(XMMRegister dst, XMMRegister src); |
304 | 1446 |
1447 void set_byte_if_not_zero(Register dst); // sets reg to 1 if not zero, otherwise 0 | |
0 | 1448 }; |
1449 | |
1450 | |
1451 // MacroAssembler extends Assembler by frequently used macros. | |
1452 // | |
1453 // Instructions for which a 'better' code sequence exists depending | |
1454 // on arguments should also go in here. | |
1455 | |
1456 class MacroAssembler: public Assembler { | |
342
37f87013dfd8
6711316: Open source the Garbage-First garbage collector
ysr
parents:
71
diff
changeset
|
1457 friend class LIR_Assembler; |
37f87013dfd8
6711316: Open source the Garbage-First garbage collector
ysr
parents:
71
diff
changeset
|
1458 friend class Runtime1; // as_Address() |
3249
e1162778c1c8
7009266: G1: assert(obj->is_oop_or_null(true )) failed: Error
johnc
parents:
2320
diff
changeset
|
1459 |
0 | 1460 protected: |
1461 | |
1462 Address as_Address(AddressLiteral adr); | |
1463 Address as_Address(ArrayAddress adr); | |
1464 | |
1465 // Support for VM calls | |
1466 // | |
1467 // This is the base routine called by the different versions of call_VM_leaf. The interpreter | |
1468 // may customize this version by overriding it for its purposes (e.g., to save/restore | |
1469 // additional registers when doing a VM call). | |
1470 #ifdef CC_INTERP | |
1471 // c++ interpreter never wants to use interp_masm version of call_VM | |
1472 #define VIRTUAL | |
1473 #else | |
1474 #define VIRTUAL virtual | |
1475 #endif | |
1476 | |
1477 VIRTUAL void call_VM_leaf_base( | |
1478 address entry_point, // the entry point | |
1479 int number_of_arguments // the number of arguments to pop after the call | |
1480 ); | |
1481 | |
1482 // This is the base routine called by the different versions of call_VM. The interpreter | |
1483 // may customize this version by overriding it for its purposes (e.g., to save/restore | |
1484 // additional registers when doing a VM call). | |
1485 // | |
1486 // If no java_thread register is specified (noreg) than rdi will be used instead. call_VM_base | |
1487 // returns the register which contains the thread upon return. If a thread register has been | |
1488 // specified, the return value will correspond to that register. If no last_java_sp is specified | |
1489 // (noreg) than rsp will be used instead. | |
1490 VIRTUAL void call_VM_base( // returns the register containing the thread upon return | |
1491 Register oop_result, // where an oop-result ends up if any; use noreg otherwise | |
1492 Register java_thread, // the thread if computed before ; use noreg otherwise | |
1493 Register last_java_sp, // to set up last_Java_frame in stubs; use noreg otherwise | |
1494 address entry_point, // the entry point | |
1495 int number_of_arguments, // the number of arguments (w/o thread) to pop after the call | |
1496 bool check_exceptions // whether to check for pending exceptions after return | |
1497 ); | |
1498 | |
1499 // These routines should emit JVMTI PopFrame and ForceEarlyReturn handling code. | |
1500 // The implementation is only non-empty for the InterpreterMacroAssembler, | |
1501 // as only the interpreter handles PopFrame and ForceEarlyReturn requests. | |
1502 virtual void check_and_handle_popframe(Register java_thread); | |
1503 virtual void check_and_handle_earlyret(Register java_thread); | |
1504 | |
1505 void call_VM_helper(Register oop_result, address entry_point, int number_of_arguments, bool check_exceptions = true); | |
1506 | |
1507 // helpers for FPU flag access | |
1508 // tmp is a temporary register, if none is available use noreg | |
1509 void save_rax (Register tmp); | |
1510 void restore_rax(Register tmp); | |
1511 | |
1512 public: | |
1513 MacroAssembler(CodeBuffer* code) : Assembler(code) {} | |
1514 | |
1515 // Support for NULL-checks | |
1516 // | |
1517 // Generates code that causes a NULL OS exception if the content of reg is NULL. | |
1518 // If the accessed location is M[reg + offset] and the offset is known, provide the | |
1519 // offset. No explicit code generation is needed if the offset is within a certain | |
1520 // range (0 <= offset <= page_size). | |
1521 | |
1522 void null_check(Register reg, int offset = -1); | |
168
7793bd37a336
6705887: Compressed Oops: generate x64 addressing and implicit null checks with narrow oops
kvn
parents:
71
diff
changeset
|
1523 static bool needs_explicit_null_check(intptr_t offset); |
0 | 1524 |
1525 // Required platform-specific helpers for Label::patch_instructions. | |
1526 // They _shadow_ the declarations in AbstractAssembler, which are undefined. | |
1527 void pd_patch_instruction(address branch, address target); | |
1528 #ifndef PRODUCT | |
1529 static void pd_print_patched_instruction(address branch); | |
1530 #endif | |
1531 | |
1532 // The following 4 methods return the offset of the appropriate move instruction | |
1533 | |
622
56aae7be60d4
6812678: macro assembler needs delayed binding of a few constants (for 6655638)
jrose
parents:
420
diff
changeset
|
1534 // Support for fast byte/short loading with zero extension (depending on particular CPU) |
0 | 1535 int load_unsigned_byte(Register dst, Address src); |
622
56aae7be60d4
6812678: macro assembler needs delayed binding of a few constants (for 6655638)
jrose
parents:
420
diff
changeset
|
1536 int load_unsigned_short(Register dst, Address src); |
56aae7be60d4
6812678: macro assembler needs delayed binding of a few constants (for 6655638)
jrose
parents:
420
diff
changeset
|
1537 |
56aae7be60d4
6812678: macro assembler needs delayed binding of a few constants (for 6655638)
jrose
parents:
420
diff
changeset
|
1538 // Support for fast byte/short loading with sign extension (depending on particular CPU) |
0 | 1539 int load_signed_byte(Register dst, Address src); |
622
56aae7be60d4
6812678: macro assembler needs delayed binding of a few constants (for 6655638)
jrose
parents:
420
diff
changeset
|
1540 int load_signed_short(Register dst, Address src); |
0 | 1541 |
1542 // Support for sign-extension (hi:lo = extend_sign(lo)) | |
1543 void extend_sign(Register hi, Register lo); | |
1544 | |
2258
28bf941f445e
7018378: JSR 292: _bound_int_mh produces wrong result on 64-bit SPARC
twisti
parents:
2100
diff
changeset
|
1545 // Load and store values by size and signed-ness |
28bf941f445e
7018378: JSR 292: _bound_int_mh produces wrong result on 64-bit SPARC
twisti
parents:
2100
diff
changeset
|
1546 void load_sized_value(Register dst, Address src, size_t size_in_bytes, bool is_signed, Register dst2 = noreg); |
28bf941f445e
7018378: JSR 292: _bound_int_mh produces wrong result on 64-bit SPARC
twisti
parents:
2100
diff
changeset
|
1547 void store_sized_value(Address dst, Register src, size_t size_in_bytes, Register src2 = noreg); |
622
56aae7be60d4
6812678: macro assembler needs delayed binding of a few constants (for 6655638)
jrose
parents:
420
diff
changeset
|
1548 |
0 | 1549 // Support for inc/dec with optimal instruction selection depending on value |
304 | 1550 |
1551 void increment(Register reg, int value = 1) { LP64_ONLY(incrementq(reg, value)) NOT_LP64(incrementl(reg, value)) ; } | |
1552 void decrement(Register reg, int value = 1) { LP64_ONLY(decrementq(reg, value)) NOT_LP64(decrementl(reg, value)) ; } | |
1553 | |
1554 void decrementl(Address dst, int value = 1); | |
1555 void decrementl(Register reg, int value = 1); | |
1556 | |
1557 void decrementq(Register reg, int value = 1); | |
1558 void decrementq(Address dst, int value = 1); | |
1559 | |
1560 void incrementl(Address dst, int value = 1); | |
1561 void incrementl(Register reg, int value = 1); | |
1562 | |
1563 void incrementq(Register reg, int value = 1); | |
1564 void incrementq(Address dst, int value = 1); | |
1565 | |
0 | 1566 |
1567 // Support optimal SSE move instructions. | |
1568 void movflt(XMMRegister dst, XMMRegister src) { | |
1569 if (UseXmmRegToRegMoveAll) { movaps(dst, src); return; } | |
1570 else { movss (dst, src); return; } | |
1571 } | |
1572 void movflt(XMMRegister dst, Address src) { movss(dst, src); } | |
1573 void movflt(XMMRegister dst, AddressLiteral src); | |
1574 void movflt(Address dst, XMMRegister src) { movss(dst, src); } | |
1575 | |
1576 void movdbl(XMMRegister dst, XMMRegister src) { | |
1577 if (UseXmmRegToRegMoveAll) { movapd(dst, src); return; } | |
1578 else { movsd (dst, src); return; } | |
1579 } | |
1580 | |
1581 void movdbl(XMMRegister dst, AddressLiteral src); | |
1582 | |
1583 void movdbl(XMMRegister dst, Address src) { | |
1584 if (UseXmmLoadAndClearUpper) { movsd (dst, src); return; } | |
1585 else { movlpd(dst, src); return; } | |
1586 } | |
1587 void movdbl(Address dst, XMMRegister src) { movsd(dst, src); } | |
1588 | |
304 | 1589 void incrementl(AddressLiteral dst); |
1590 void incrementl(ArrayAddress dst); | |
0 | 1591 |
1592 // Alignment | |
1593 void align(int modulus); | |
1594 | |
1595 // Misc | |
1596 void fat_nop(); // 5 byte nop | |
1597 | |
1598 // Stack frame creation/removal | |
1599 void enter(); | |
1600 void leave(); | |
1601 | |
1602 // Support for getting the JavaThread pointer (i.e.; a reference to thread-local information) | |
1603 // The pointer will be loaded into the thread register. | |
1604 void get_thread(Register thread); | |
1605 | |
362 | 1606 |
0 | 1607 // Support for VM calls |
1608 // | |
1609 // It is imperative that all calls into the VM are handled via the call_VM macros. | |
1610 // They make sure that the stack linkage is setup correctly. call_VM's correspond | |
1611 // to ENTRY/ENTRY_X entry points while call_VM_leaf's correspond to LEAF entry points. | |
1612 | |
304 | 1613 |
1614 void call_VM(Register oop_result, | |
1615 address entry_point, | |
1616 bool check_exceptions = true); | |
1617 void call_VM(Register oop_result, | |
1618 address entry_point, | |
1619 Register arg_1, | |
1620 bool check_exceptions = true); | |
1621 void call_VM(Register oop_result, | |
1622 address entry_point, | |
1623 Register arg_1, Register arg_2, | |
1624 bool check_exceptions = true); | |
1625 void call_VM(Register oop_result, | |
1626 address entry_point, | |
1627 Register arg_1, Register arg_2, Register arg_3, | |
1628 bool check_exceptions = true); | |
1629 | |
1630 // Overloadings with last_Java_sp | |
1631 void call_VM(Register oop_result, | |
1632 Register last_java_sp, | |
1633 address entry_point, | |
1634 int number_of_arguments = 0, | |
1635 bool check_exceptions = true); | |
1636 void call_VM(Register oop_result, | |
1637 Register last_java_sp, | |
1638 address entry_point, | |
1639 Register arg_1, bool | |
1640 check_exceptions = true); | |
1641 void call_VM(Register oop_result, | |
1642 Register last_java_sp, | |
1643 address entry_point, | |
1644 Register arg_1, Register arg_2, | |
1645 bool check_exceptions = true); | |
1646 void call_VM(Register oop_result, | |
1647 Register last_java_sp, | |
1648 address entry_point, | |
1649 Register arg_1, Register arg_2, Register arg_3, | |
1650 bool check_exceptions = true); | |
1651 | |
3755
5cf771a79037
7047697: MethodHandle.invokeExact call for wrong method causes VM failure if run with -Xcomp
jrose
parents:
3363
diff
changeset
|
1652 // These always tightly bind to MacroAssembler::call_VM_base |
5cf771a79037
7047697: MethodHandle.invokeExact call for wrong method causes VM failure if run with -Xcomp
jrose
parents:
3363
diff
changeset
|
1653 // bypassing the virtual implementation |
5cf771a79037
7047697: MethodHandle.invokeExact call for wrong method causes VM failure if run with -Xcomp
jrose
parents:
3363
diff
changeset
|
1654 void super_call_VM(Register oop_result, Register last_java_sp, address entry_point, int number_of_arguments = 0, bool check_exceptions = true); |
5cf771a79037
7047697: MethodHandle.invokeExact call for wrong method causes VM failure if run with -Xcomp
jrose
parents:
3363
diff
changeset
|
1655 void super_call_VM(Register oop_result, Register last_java_sp, address entry_point, Register arg_1, bool check_exceptions = true); |
5cf771a79037
7047697: MethodHandle.invokeExact call for wrong method causes VM failure if run with -Xcomp
jrose
parents:
3363
diff
changeset
|
1656 void super_call_VM(Register oop_result, Register last_java_sp, address entry_point, Register arg_1, Register arg_2, bool check_exceptions = true); |
5cf771a79037
7047697: MethodHandle.invokeExact call for wrong method causes VM failure if run with -Xcomp
jrose
parents:
3363
diff
changeset
|
1657 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); |
5cf771a79037
7047697: MethodHandle.invokeExact call for wrong method causes VM failure if run with -Xcomp
jrose
parents:
3363
diff
changeset
|
1658 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); |
5cf771a79037
7047697: MethodHandle.invokeExact call for wrong method causes VM failure if run with -Xcomp
jrose
parents:
3363
diff
changeset
|
1659 |
304 | 1660 void call_VM_leaf(address entry_point, |
1661 int number_of_arguments = 0); | |
1662 void call_VM_leaf(address entry_point, | |
1663 Register arg_1); | |
1664 void call_VM_leaf(address entry_point, | |
1665 Register arg_1, Register arg_2); | |
1666 void call_VM_leaf(address entry_point, | |
1667 Register arg_1, Register arg_2, Register arg_3); | |
0 | 1668 |
3336
2e038ad0c1d0
7009361: JSR 292 Invalid value on stack on solaris-sparc with -Xcomp
never
parents:
3255
diff
changeset
|
1669 // These always tightly bind to MacroAssembler::call_VM_leaf_base |
2e038ad0c1d0
7009361: JSR 292 Invalid value on stack on solaris-sparc with -Xcomp
never
parents:
3255
diff
changeset
|
1670 // bypassing the virtual implementation |
2e038ad0c1d0
7009361: JSR 292 Invalid value on stack on solaris-sparc with -Xcomp
never
parents:
3255
diff
changeset
|
1671 void super_call_VM_leaf(address entry_point); |
2e038ad0c1d0
7009361: JSR 292 Invalid value on stack on solaris-sparc with -Xcomp
never
parents:
3255
diff
changeset
|
1672 void super_call_VM_leaf(address entry_point, Register arg_1); |
2e038ad0c1d0
7009361: JSR 292 Invalid value on stack on solaris-sparc with -Xcomp
never
parents:
3255
diff
changeset
|
1673 void super_call_VM_leaf(address entry_point, Register arg_1, Register arg_2); |
2e038ad0c1d0
7009361: JSR 292 Invalid value on stack on solaris-sparc with -Xcomp
never
parents:
3255
diff
changeset
|
1674 void super_call_VM_leaf(address entry_point, Register arg_1, Register arg_2, Register arg_3); |
2e038ad0c1d0
7009361: JSR 292 Invalid value on stack on solaris-sparc with -Xcomp
never
parents:
3255
diff
changeset
|
1675 void super_call_VM_leaf(address entry_point, Register arg_1, Register arg_2, Register arg_3, Register arg_4); |
2e038ad0c1d0
7009361: JSR 292 Invalid value on stack on solaris-sparc with -Xcomp
never
parents:
3255
diff
changeset
|
1676 |
0 | 1677 // last Java Frame (fills frame anchor) |
304 | 1678 void set_last_Java_frame(Register thread, |
1679 Register last_java_sp, | |
1680 Register last_java_fp, | |
1681 address last_java_pc); | |
1682 | |
1683 // thread in the default location (r15_thread on 64bit) | |
1684 void set_last_Java_frame(Register last_java_sp, | |
1685 Register last_java_fp, | |
1686 address last_java_pc); | |
1687 | |
0 | 1688 void reset_last_Java_frame(Register thread, bool clear_fp, bool clear_pc); |
1689 | |
304 | 1690 // thread in the default location (r15_thread on 64bit) |
1691 void reset_last_Java_frame(bool clear_fp, bool clear_pc); | |
1692 | |
0 | 1693 // Stores |
1694 void store_check(Register obj); // store check for obj - register is destroyed afterwards | |
1695 void store_check(Register obj, Address dst); // same as above, dst is exact store location (reg. is destroyed) | |
1696 | |
3249
e1162778c1c8
7009266: G1: assert(obj->is_oop_or_null(true )) failed: Error
johnc
parents:
2320
diff
changeset
|
1697 #ifndef SERIALGC |
e1162778c1c8
7009266: G1: assert(obj->is_oop_or_null(true )) failed: Error
johnc
parents:
2320
diff
changeset
|
1698 |
362 | 1699 void g1_write_barrier_pre(Register obj, |
3249
e1162778c1c8
7009266: G1: assert(obj->is_oop_or_null(true )) failed: Error
johnc
parents:
2320
diff
changeset
|
1700 Register pre_val, |
362 | 1701 Register thread, |
1702 Register tmp, | |
3249
e1162778c1c8
7009266: G1: assert(obj->is_oop_or_null(true )) failed: Error
johnc
parents:
2320
diff
changeset
|
1703 bool tosca_live, |
e1162778c1c8
7009266: G1: assert(obj->is_oop_or_null(true )) failed: Error
johnc
parents:
2320
diff
changeset
|
1704 bool expand_call); |
e1162778c1c8
7009266: G1: assert(obj->is_oop_or_null(true )) failed: Error
johnc
parents:
2320
diff
changeset
|
1705 |
362 | 1706 void g1_write_barrier_post(Register store_addr, |
1707 Register new_val, | |
1708 Register thread, | |
1709 Register tmp, | |
1710 Register tmp2); | |
342
37f87013dfd8
6711316: Open source the Garbage-First garbage collector
ysr
parents:
71
diff
changeset
|
1711 |
3249
e1162778c1c8
7009266: G1: assert(obj->is_oop_or_null(true )) failed: Error
johnc
parents:
2320
diff
changeset
|
1712 #endif // SERIALGC |
342
37f87013dfd8
6711316: Open source the Garbage-First garbage collector
ysr
parents:
71
diff
changeset
|
1713 |
0 | 1714 // split store_check(Register obj) to enhance instruction interleaving |
1715 void store_check_part_1(Register obj); | |
1716 void store_check_part_2(Register obj); | |
1717 | |
1718 // C 'boolean' to Java boolean: x == 0 ? 0 : 1 | |
1719 void c2bool(Register x); | |
1720 | |
1721 // C++ bool manipulation | |
1722 | |
1723 void movbool(Register dst, Address src); | |
1724 void movbool(Address dst, bool boolconst); | |
1725 void movbool(Address dst, Register src); | |
1726 void testbool(Register dst); | |
1727 | |
304 | 1728 // oop manipulations |
1729 void load_klass(Register dst, Register src); | |
1730 void store_klass(Register dst, Register src); | |
1731 | |
1846 | 1732 void load_heap_oop(Register dst, Address src); |
2464
d86923d96dca
7034967: C1: assert(false) failed: error (assembler_sparc.cpp:2043)
iveresov
parents:
2455
diff
changeset
|
1733 void load_heap_oop_not_null(Register dst, Address src); |
1846 | 1734 void store_heap_oop(Address dst, Register src); |
1735 | |
1736 // Used for storing NULL. All other oop constants should be | |
1737 // stored using routines that take a jobject. | |
1738 void store_heap_oop_null(Address dst); | |
1739 | |
304 | 1740 void load_prototype_header(Register dst, Register src); |
1741 | |
1742 #ifdef _LP64 | |
1743 void store_klass_gap(Register dst, Register src); | |
1744 | |
1047
beb8f45ee9f0
6889740: G1: OpenDS fails with "unhandled exception in compiled code"
johnc
parents:
986
diff
changeset
|
1745 // This dummy is to prevent a call to store_heap_oop from |
beb8f45ee9f0
6889740: G1: OpenDS fails with "unhandled exception in compiled code"
johnc
parents:
986
diff
changeset
|
1746 // converting a zero (like NULL) into a Register by giving |
beb8f45ee9f0
6889740: G1: OpenDS fails with "unhandled exception in compiled code"
johnc
parents:
986
diff
changeset
|
1747 // the compiler two choices it can't resolve |
beb8f45ee9f0
6889740: G1: OpenDS fails with "unhandled exception in compiled code"
johnc
parents:
986
diff
changeset
|
1748 |
beb8f45ee9f0
6889740: G1: OpenDS fails with "unhandled exception in compiled code"
johnc
parents:
986
diff
changeset
|
1749 void store_heap_oop(Address dst, void* dummy); |
beb8f45ee9f0
6889740: G1: OpenDS fails with "unhandled exception in compiled code"
johnc
parents:
986
diff
changeset
|
1750 |
304 | 1751 void encode_heap_oop(Register r); |
1752 void decode_heap_oop(Register r); | |
1753 void encode_heap_oop_not_null(Register r); | |
1754 void decode_heap_oop_not_null(Register r); | |
1755 void encode_heap_oop_not_null(Register dst, Register src); | |
1756 void decode_heap_oop_not_null(Register dst, Register src); | |
1757 | |
1758 void set_narrow_oop(Register dst, jobject obj); | |
642
660978a2a31a
6791178: Specialize for zero as the compressed oop vm heap base
kvn
parents:
624
diff
changeset
|
1759 void set_narrow_oop(Address dst, jobject obj); |
660978a2a31a
6791178: Specialize for zero as the compressed oop vm heap base
kvn
parents:
624
diff
changeset
|
1760 void cmp_narrow_oop(Register dst, jobject obj); |
660978a2a31a
6791178: Specialize for zero as the compressed oop vm heap base
kvn
parents:
624
diff
changeset
|
1761 void cmp_narrow_oop(Address dst, jobject obj); |
304 | 1762 |
1763 // if heap base register is used - reinit it with the correct value | |
1764 void reinit_heapbase(); | |
1684
66c5dadb4d61
6973308: Missing zero length check before repne scas in check_klass_subtype_slow_path()
kvn
parents:
1579
diff
changeset
|
1765 |
66c5dadb4d61
6973308: Missing zero length check before repne scas in check_klass_subtype_slow_path()
kvn
parents:
1579
diff
changeset
|
1766 DEBUG_ONLY(void verify_heapbase(const char* msg);) |
66c5dadb4d61
6973308: Missing zero length check before repne scas in check_klass_subtype_slow_path()
kvn
parents:
1579
diff
changeset
|
1767 |
304 | 1768 #endif // _LP64 |
1769 | |
1770 // Int division/remainder for Java | |
0 | 1771 // (as idivl, but checks for special case as described in JVM spec.) |
1772 // returns idivl instruction offset for implicit exception handling | |
1773 int corrected_idivl(Register reg); | |
1774 | |
304 | 1775 // Long division/remainder for Java |
1776 // (as idivq, but checks for special case as described in JVM spec.) | |
1777 // returns idivq instruction offset for implicit exception handling | |
1778 int corrected_idivq(Register reg); | |
1779 | |
0 | 1780 void int3(); |
1781 | |
304 | 1782 // Long operation macros for a 32bit cpu |
0 | 1783 // Long negation for Java |
1784 void lneg(Register hi, Register lo); | |
1785 | |
1786 // Long multiplication for Java | |
304 | 1787 // (destroys contents of eax, ebx, ecx and edx) |
0 | 1788 void lmul(int x_rsp_offset, int y_rsp_offset); // rdx:rax = x * y |
1789 | |
1790 // Long shifts for Java | |
1791 // (semantics as described in JVM spec.) | |
1792 void lshl(Register hi, Register lo); // hi:lo << (rcx & 0x3f) | |
1793 void lshr(Register hi, Register lo, bool sign_extension = false); // hi:lo >> (rcx & 0x3f) | |
1794 | |
1795 // Long compare for Java | |
1796 // (semantics as described in JVM spec.) | |
1797 void lcmp2int(Register x_hi, Register x_lo, Register y_hi, Register y_lo); // x_hi = lcmp(x, y) | |
1798 | |
304 | 1799 |
1800 // misc | |
1801 | |
1802 // Sign extension | |
1803 void sign_extend_short(Register reg); | |
1804 void sign_extend_byte(Register reg); | |
1805 | |
1806 // Division by power of 2, rounding towards 0 | |
1807 void division_with_shift(Register reg, int shift_value); | |
1808 | |
0 | 1809 // Compares the top-most stack entries on the FPU stack and sets the eflags as follows: |
1810 // | |
1811 // CF (corresponds to C0) if x < y | |
1812 // PF (corresponds to C2) if unordered | |
1813 // ZF (corresponds to C3) if x = y | |
1814 // | |
1815 // The arguments are in reversed order on the stack (i.e., top of stack is first argument). | |
1816 // tmp is a temporary register, if none is available use noreg (only matters for non-P6 code) | |
1817 void fcmp(Register tmp); | |
1818 // Variant of the above which allows y to be further down the stack | |
1819 // and which only pops x and y if specified. If pop_right is | |
1820 // specified then pop_left must also be specified. | |
1821 void fcmp(Register tmp, int index, bool pop_left, bool pop_right); | |
1822 | |
1823 // Floating-point comparison for Java | |
1824 // Compares the top-most stack entries on the FPU stack and stores the result in dst. | |
1825 // The arguments are in reversed order on the stack (i.e., top of stack is first argument). | |
1826 // (semantics as described in JVM spec.) | |
1827 void fcmp2int(Register dst, bool unordered_is_less); | |
1828 // Variant of the above which allows y to be further down the stack | |
1829 // and which only pops x and y if specified. If pop_right is | |
1830 // specified then pop_left must also be specified. | |
1831 void fcmp2int(Register dst, bool unordered_is_less, int index, bool pop_left, bool pop_right); | |
1832 | |
1833 // Floating-point remainder for Java (ST0 = ST0 fremr ST1, ST1 is empty afterwards) | |
1834 // tmp is a temporary register, if none is available use noreg | |
1835 void fremr(Register tmp); | |
1836 | |
1837 | |
1838 // same as fcmp2int, but using SSE2 | |
1839 void cmpss2int(XMMRegister opr1, XMMRegister opr2, Register dst, bool unordered_is_less); | |
1840 void cmpsd2int(XMMRegister opr1, XMMRegister opr2, Register dst, bool unordered_is_less); | |
1841 | |
1842 // Inlined sin/cos generator for Java; must not use CPU instruction | |
1843 // directly on Intel as it does not have high enough precision | |
1844 // outside of the range [-pi/4, pi/4]. Extra argument indicate the | |
1845 // number of FPU stack slots in use; all but the topmost will | |
1846 // require saving if a slow case is necessary. Assumes argument is | |
1847 // on FP TOS; result is on FP TOS. No cpu registers are changed by | |
1848 // this code. | |
1849 void trigfunc(char trig, int num_fpu_regs_in_use = 1); | |
1850 | |
1851 // branch to L if FPU flag C2 is set/not set | |
1852 // tmp is a temporary register, if none is available use noreg | |
1853 void jC2 (Register tmp, Label& L); | |
1854 void jnC2(Register tmp, Label& L); | |
1855 | |
1856 // Pop ST (ffree & fincstp combined) | |
1857 void fpop(); | |
1858 | |
1859 // pushes double TOS element of FPU stack on CPU stack; pops from FPU stack | |
1860 void push_fTOS(); | |
1861 | |
1862 // pops double TOS element from CPU stack and pushes on FPU stack | |
1863 void pop_fTOS(); | |
1864 | |
1865 void empty_FPU_stack(); | |
1866 | |
1867 void push_IU_state(); | |
1868 void pop_IU_state(); | |
1869 | |
1870 void push_FPU_state(); | |
1871 void pop_FPU_state(); | |
1872 | |
1873 void push_CPU_state(); | |
1874 void pop_CPU_state(); | |
1875 | |
1876 // Round up to a power of two | |
1877 void round_to(Register reg, int modulus); | |
1878 | |
1879 // Callee saved registers handling | |
1880 void push_callee_saved_registers(); | |
1881 void pop_callee_saved_registers(); | |
1882 | |
1883 // allocation | |
1884 void eden_allocate( | |
1885 Register obj, // result: pointer to object after successful allocation | |
1886 Register var_size_in_bytes, // object size in bytes if unknown at compile time; invalid otherwise | |
1887 int con_size_in_bytes, // object size in bytes if known at compile time | |
1888 Register t1, // temp register | |
1889 Label& slow_case // continuation point if fast allocation fails | |
1890 ); | |
1891 void tlab_allocate( | |
1892 Register obj, // result: pointer to object after successful allocation | |
1893 Register var_size_in_bytes, // object size in bytes if unknown at compile time; invalid otherwise | |
1894 int con_size_in_bytes, // object size in bytes if known at compile time | |
1895 Register t1, // temp register | |
1896 Register t2, // temp register | |
1897 Label& slow_case // continuation point if fast allocation fails | |
1898 ); | |
2100
b1a2afa37ec4
7003271: Hotspot should track cumulative Java heap bytes allocated on a per-thread basis
phh
parents:
2008
diff
changeset
|
1899 Register tlab_refill(Label& retry_tlab, Label& try_eden, Label& slow_case); // returns TLS address |
b1a2afa37ec4
7003271: Hotspot should track cumulative Java heap bytes allocated on a per-thread basis
phh
parents:
2008
diff
changeset
|
1900 void incr_allocated_bytes(Register thread, |
b1a2afa37ec4
7003271: Hotspot should track cumulative Java heap bytes allocated on a per-thread basis
phh
parents:
2008
diff
changeset
|
1901 Register var_size_in_bytes, int con_size_in_bytes, |
b1a2afa37ec4
7003271: Hotspot should track cumulative Java heap bytes allocated on a per-thread basis
phh
parents:
2008
diff
changeset
|
1902 Register t1 = noreg); |
0 | 1903 |
623
9adddb8c0fc8
6812831: factor duplicated assembly code for megamorphic invokeinterface (for 6655638)
jrose
parents:
622
diff
changeset
|
1904 // interface method calling |
9adddb8c0fc8
6812831: factor duplicated assembly code for megamorphic invokeinterface (for 6655638)
jrose
parents:
622
diff
changeset
|
1905 void lookup_interface_method(Register recv_klass, |
9adddb8c0fc8
6812831: factor duplicated assembly code for megamorphic invokeinterface (for 6655638)
jrose
parents:
622
diff
changeset
|
1906 Register intf_klass, |
665
c89f86385056
6814659: separable cleanups and subroutines for 6655638
jrose
parents:
644
diff
changeset
|
1907 RegisterOrConstant itable_index, |
623
9adddb8c0fc8
6812831: factor duplicated assembly code for megamorphic invokeinterface (for 6655638)
jrose
parents:
622
diff
changeset
|
1908 Register method_result, |
9adddb8c0fc8
6812831: factor duplicated assembly code for megamorphic invokeinterface (for 6655638)
jrose
parents:
622
diff
changeset
|
1909 Register scan_temp, |
9adddb8c0fc8
6812831: factor duplicated assembly code for megamorphic invokeinterface (for 6655638)
jrose
parents:
622
diff
changeset
|
1910 Label& no_such_interface); |
9adddb8c0fc8
6812831: factor duplicated assembly code for megamorphic invokeinterface (for 6655638)
jrose
parents:
622
diff
changeset
|
1911 |
644
c517646eef23
6813212: factor duplicated assembly code for general subclass check (for 6655638)
jrose
parents:
643
diff
changeset
|
1912 // Test sub_klass against super_klass, with fast and slow paths. |
c517646eef23
6813212: factor duplicated assembly code for general subclass check (for 6655638)
jrose
parents:
643
diff
changeset
|
1913 |
c517646eef23
6813212: factor duplicated assembly code for general subclass check (for 6655638)
jrose
parents:
643
diff
changeset
|
1914 // The fast path produces a tri-state answer: yes / no / maybe-slow. |
c517646eef23
6813212: factor duplicated assembly code for general subclass check (for 6655638)
jrose
parents:
643
diff
changeset
|
1915 // One of the three labels can be NULL, meaning take the fall-through. |
c517646eef23
6813212: factor duplicated assembly code for general subclass check (for 6655638)
jrose
parents:
643
diff
changeset
|
1916 // If super_check_offset is -1, the value is loaded up from super_klass. |
c517646eef23
6813212: factor duplicated assembly code for general subclass check (for 6655638)
jrose
parents:
643
diff
changeset
|
1917 // No registers are killed, except temp_reg. |
c517646eef23
6813212: factor duplicated assembly code for general subclass check (for 6655638)
jrose
parents:
643
diff
changeset
|
1918 void check_klass_subtype_fast_path(Register sub_klass, |
c517646eef23
6813212: factor duplicated assembly code for general subclass check (for 6655638)
jrose
parents:
643
diff
changeset
|
1919 Register super_klass, |
c517646eef23
6813212: factor duplicated assembly code for general subclass check (for 6655638)
jrose
parents:
643
diff
changeset
|
1920 Register temp_reg, |
c517646eef23
6813212: factor duplicated assembly code for general subclass check (for 6655638)
jrose
parents:
643
diff
changeset
|
1921 Label* L_success, |
c517646eef23
6813212: factor duplicated assembly code for general subclass check (for 6655638)
jrose
parents:
643
diff
changeset
|
1922 Label* L_failure, |
c517646eef23
6813212: factor duplicated assembly code for general subclass check (for 6655638)
jrose
parents:
643
diff
changeset
|
1923 Label* L_slow_path, |
665
c89f86385056
6814659: separable cleanups and subroutines for 6655638
jrose
parents:
644
diff
changeset
|
1924 RegisterOrConstant super_check_offset = RegisterOrConstant(-1)); |
644
c517646eef23
6813212: factor duplicated assembly code for general subclass check (for 6655638)
jrose
parents:
643
diff
changeset
|
1925 |
c517646eef23
6813212: factor duplicated assembly code for general subclass check (for 6655638)
jrose
parents:
643
diff
changeset
|
1926 // The rest of the type check; must be wired to a corresponding fast path. |
c517646eef23
6813212: factor duplicated assembly code for general subclass check (for 6655638)
jrose
parents:
643
diff
changeset
|
1927 // It does not repeat the fast path logic, so don't use it standalone. |
c517646eef23
6813212: factor duplicated assembly code for general subclass check (for 6655638)
jrose
parents:
643
diff
changeset
|
1928 // The temp_reg and temp2_reg can be noreg, if no temps are available. |
c517646eef23
6813212: factor duplicated assembly code for general subclass check (for 6655638)
jrose
parents:
643
diff
changeset
|
1929 // Updates the sub's secondary super cache as necessary. |
c517646eef23
6813212: factor duplicated assembly code for general subclass check (for 6655638)
jrose
parents:
643
diff
changeset
|
1930 // If set_cond_codes, condition codes will be Z on success, NZ on failure. |
c517646eef23
6813212: factor duplicated assembly code for general subclass check (for 6655638)
jrose
parents:
643
diff
changeset
|
1931 void check_klass_subtype_slow_path(Register sub_klass, |
c517646eef23
6813212: factor duplicated assembly code for general subclass check (for 6655638)
jrose
parents:
643
diff
changeset
|
1932 Register super_klass, |
c517646eef23
6813212: factor duplicated assembly code for general subclass check (for 6655638)
jrose
parents:
643
diff
changeset
|
1933 Register temp_reg, |
c517646eef23
6813212: factor duplicated assembly code for general subclass check (for 6655638)
jrose
parents:
643
diff
changeset
|
1934 Register temp2_reg, |
c517646eef23
6813212: factor duplicated assembly code for general subclass check (for 6655638)
jrose
parents:
643
diff
changeset
|
1935 Label* L_success, |
c517646eef23
6813212: factor duplicated assembly code for general subclass check (for 6655638)
jrose
parents:
643
diff
changeset
|
1936 Label* L_failure, |
c517646eef23
6813212: factor duplicated assembly code for general subclass check (for 6655638)
jrose
parents:
643
diff
changeset
|
1937 bool set_cond_codes = false); |
c517646eef23
6813212: factor duplicated assembly code for general subclass check (for 6655638)
jrose
parents:
643
diff
changeset
|
1938 |
c517646eef23
6813212: factor duplicated assembly code for general subclass check (for 6655638)
jrose
parents:
643
diff
changeset
|
1939 // Simplified, combined version, good for typical uses. |
c517646eef23
6813212: factor duplicated assembly code for general subclass check (for 6655638)
jrose
parents:
643
diff
changeset
|
1940 // Falls through on failure. |
c517646eef23
6813212: factor duplicated assembly code for general subclass check (for 6655638)
jrose
parents:
643
diff
changeset
|
1941 void check_klass_subtype(Register sub_klass, |
c517646eef23
6813212: factor duplicated assembly code for general subclass check (for 6655638)
jrose
parents:
643
diff
changeset
|
1942 Register super_klass, |
c517646eef23
6813212: factor duplicated assembly code for general subclass check (for 6655638)
jrose
parents:
643
diff
changeset
|
1943 Register temp_reg, |
c517646eef23
6813212: factor duplicated assembly code for general subclass check (for 6655638)
jrose
parents:
643
diff
changeset
|
1944 Label& L_success); |
c517646eef23
6813212: factor duplicated assembly code for general subclass check (for 6655638)
jrose
parents:
643
diff
changeset
|
1945 |
710 | 1946 // method handles (JSR 292) |
1947 void check_method_handle_type(Register mtype_reg, Register mh_reg, | |
1948 Register temp_reg, | |
1949 Label& wrong_method_type); | |
1950 void load_method_handle_vmslots(Register vmslots_reg, Register mh_reg, | |
1951 Register temp_reg); | |
1952 void jump_to_method_handle_entry(Register mh_reg, Register temp_reg); | |
1953 Address argument_address(RegisterOrConstant arg_slot, int extra_slot_offset = 0); | |
1954 | |
1955 | |
0 | 1956 //---- |
1957 void set_word_if_not_zero(Register reg); // sets reg to 1 if not zero, otherwise 0 | |
1958 | |
1959 // Debugging | |
304 | 1960 |
1961 // only if +VerifyOops | |
1962 void verify_oop(Register reg, const char* s = "broken oop"); | |
0 | 1963 void verify_oop_addr(Address addr, const char * s = "broken oop addr"); |
1964 | |
304 | 1965 // only if +VerifyFPU |
1966 void verify_FPU(int stack_depth, const char* s = "illegal FPU state"); | |
1967 | |
1968 // prints msg, dumps registers and stops execution | |
1969 void stop(const char* msg); | |
1970 | |
1971 // prints msg and continues | |
1972 void warn(const char* msg); | |
1973 | |
1974 static void debug32(int rdi, int rsi, int rbp, int rsp, int rbx, int rdx, int rcx, int rax, int eip, char* msg); | |
1975 static void debug64(char* msg, int64_t pc, int64_t regs[]); | |
1976 | |
0 | 1977 void os_breakpoint(); |
304 | 1978 |
0 | 1979 void untested() { stop("untested"); } |
304 | 1980 |
1846 | 1981 void unimplemented(const char* what = "") { char* b = new char[1024]; jio_snprintf(b, 1024, "unimplemented: %s", what); stop(b); } |
304 | 1982 |
0 | 1983 void should_not_reach_here() { stop("should not reach here"); } |
304 | 1984 |
0 | 1985 void print_CPU_state(); |
1986 | |
1987 // Stack overflow checking | |
1988 void bang_stack_with_offset(int offset) { | |
1989 // stack grows down, caller passes positive offset | |
1990 assert(offset > 0, "must bang with negative offset"); | |
1991 movl(Address(rsp, (-offset)), rax); | |
1992 } | |
1993 | |
1994 // Writes to stack successive pages until offset reached to check for | |
1995 // stack overflow + shadow pages. Also, clobbers tmp | |
1996 void bang_stack_size(Register size, Register tmp); | |
1997 | |
665
c89f86385056
6814659: separable cleanups and subroutines for 6655638
jrose
parents:
644
diff
changeset
|
1998 virtual RegisterOrConstant delayed_value_impl(intptr_t* delayed_value_addr, |
c89f86385056
6814659: separable cleanups and subroutines for 6655638
jrose
parents:
644
diff
changeset
|
1999 Register tmp, |
c89f86385056
6814659: separable cleanups and subroutines for 6655638
jrose
parents:
644
diff
changeset
|
2000 int offset); |
622
56aae7be60d4
6812678: macro assembler needs delayed binding of a few constants (for 6655638)
jrose
parents:
420
diff
changeset
|
2001 |
0 | 2002 // Support for serializing memory accesses between threads |
2003 void serialize_memory(Register thread, Register tmp); | |
2004 | |
2005 void verify_tlab(); | |
2006 | |
2007 // Biased locking support | |
2008 // lock_reg and obj_reg must be loaded up with the appropriate values. | |
2009 // swap_reg must be rax, and is killed. | |
2010 // tmp_reg is optional. If it is supplied (i.e., != noreg) it will | |
2011 // be killed; if not supplied, push/pop will be used internally to | |
2012 // allocate a temporary (inefficient, avoid if possible). | |
2013 // Optional slow case is for implementations (interpreter and C1) which branch to | |
2014 // slow case directly. Leaves condition codes set for C2's Fast_Lock node. | |
2015 // Returns offset of first potentially-faulting instruction for null | |
2016 // check info (currently consumed only by C1). If | |
2017 // swap_reg_contains_mark is true then returns -1 as it is assumed | |
2018 // the calling code has already passed any potential faults. | |
420
a1980da045cc
6462850: generate biased locking code in C2 ideal graph
kvn
parents:
405
diff
changeset
|
2019 int biased_locking_enter(Register lock_reg, Register obj_reg, |
a1980da045cc
6462850: generate biased locking code in C2 ideal graph
kvn
parents:
405
diff
changeset
|
2020 Register swap_reg, Register tmp_reg, |
0 | 2021 bool swap_reg_contains_mark, |
2022 Label& done, Label* slow_case = NULL, | |
2023 BiasedLockingCounters* counters = NULL); | |
2024 void biased_locking_exit (Register obj_reg, Register temp_reg, Label& done); | |
2025 | |
2026 | |
2027 Condition negate_condition(Condition cond); | |
2028 | |
2029 // Instructions that use AddressLiteral operands. These instruction can handle 32bit/64bit | |
2030 // operands. In general the names are modified to avoid hiding the instruction in Assembler | |
2031 // so that we don't need to implement all the varieties in the Assembler with trivial wrappers | |
2032 // here in MacroAssembler. The major exception to this rule is call | |
2033 | |
2034 // Arithmetics | |
2035 | |
304 | 2036 |
2037 void addptr(Address dst, int32_t src) { LP64_ONLY(addq(dst, src)) NOT_LP64(addl(dst, src)) ; } | |
2038 void addptr(Address dst, Register src); | |
2039 | |
2040 void addptr(Register dst, Address src) { LP64_ONLY(addq(dst, src)) NOT_LP64(addl(dst, src)); } | |
2041 void addptr(Register dst, int32_t src); | |
2042 void addptr(Register dst, Register src); | |
3363
167b70ff3abc
6939861: JVM should handle more conversion operations
never
parents:
3336
diff
changeset
|
2043 void addptr(Register dst, RegisterOrConstant src) { |
167b70ff3abc
6939861: JVM should handle more conversion operations
never
parents:
3336
diff
changeset
|
2044 if (src.is_constant()) addptr(dst, (int) src.as_constant()); |
167b70ff3abc
6939861: JVM should handle more conversion operations
never
parents:
3336
diff
changeset
|
2045 else addptr(dst, src.as_register()); |
167b70ff3abc
6939861: JVM should handle more conversion operations
never
parents:
3336
diff
changeset
|
2046 } |
304 | 2047 |
2048 void andptr(Register dst, int32_t src); | |
2049 void andptr(Register src1, Register src2) { LP64_ONLY(andq(src1, src2)) NOT_LP64(andl(src1, src2)) ; } | |
2050 | |
2051 void cmp8(AddressLiteral src1, int imm); | |
2052 | |
2053 // renamed to drag out the casting of address to int32_t/intptr_t | |
0 | 2054 void cmp32(Register src1, int32_t imm); |
2055 | |
2056 void cmp32(AddressLiteral src1, int32_t imm); | |
2057 // compare reg - mem, or reg - &mem | |
2058 void cmp32(Register src1, AddressLiteral src2); | |
2059 | |
2060 void cmp32(Register src1, Address src2); | |
2061 | |
304 | 2062 #ifndef _LP64 |
2063 void cmpoop(Address dst, jobject obj); | |
2064 void cmpoop(Register dst, jobject obj); | |
2065 #endif // _LP64 | |
2066 | |
0 | 2067 // NOTE src2 must be the lval. This is NOT an mem-mem compare |
2068 void cmpptr(Address src1, AddressLiteral src2); | |
2069 | |
2070 void cmpptr(Register src1, AddressLiteral src2); | |
2071 | |
304 | 2072 void cmpptr(Register src1, Register src2) { LP64_ONLY(cmpq(src1, src2)) NOT_LP64(cmpl(src1, src2)) ; } |
2073 void cmpptr(Register src1, Address src2) { LP64_ONLY(cmpq(src1, src2)) NOT_LP64(cmpl(src1, src2)) ; } | |
2074 // void cmpptr(Address src1, Register src2) { LP64_ONLY(cmpq(src1, src2)) NOT_LP64(cmpl(src1, src2)) ; } | |
2075 | |
2076 void cmpptr(Register src1, int32_t src2) { LP64_ONLY(cmpq(src1, src2)) NOT_LP64(cmpl(src1, src2)) ; } | |
2077 void cmpptr(Address src1, int32_t src2) { LP64_ONLY(cmpq(src1, src2)) NOT_LP64(cmpl(src1, src2)) ; } | |
2078 | |
2079 // cmp64 to avoild hiding cmpq | |
2080 void cmp64(Register src1, AddressLiteral src); | |
2081 | |
2082 void cmpxchgptr(Register reg, Address adr); | |
2083 | |
2084 void locked_cmpxchgptr(Register reg, AddressLiteral adr); | |
2085 | |
2086 | |
2087 void imulptr(Register dst, Register src) { LP64_ONLY(imulq(dst, src)) NOT_LP64(imull(dst, src)); } | |
2088 | |
2089 | |
2090 void negptr(Register dst) { LP64_ONLY(negq(dst)) NOT_LP64(negl(dst)); } | |
2091 | |
2092 void notptr(Register dst) { LP64_ONLY(notq(dst)) NOT_LP64(notl(dst)); } | |
2093 | |
2094 void shlptr(Register dst, int32_t shift); | |
2095 void shlptr(Register dst) { LP64_ONLY(shlq(dst)) NOT_LP64(shll(dst)); } | |
2096 | |
2097 void shrptr(Register dst, int32_t shift); | |
2098 void shrptr(Register dst) { LP64_ONLY(shrq(dst)) NOT_LP64(shrl(dst)); } | |
2099 | |
2100 void sarptr(Register dst) { LP64_ONLY(sarq(dst)) NOT_LP64(sarl(dst)); } | |
2101 void sarptr(Register dst, int32_t src) { LP64_ONLY(sarq(dst, src)) NOT_LP64(sarl(dst, src)); } | |
2102 | |
2103 void subptr(Address dst, int32_t src) { LP64_ONLY(subq(dst, src)) NOT_LP64(subl(dst, src)); } | |
2104 | |
2105 void subptr(Register dst, Address src) { LP64_ONLY(subq(dst, src)) NOT_LP64(subl(dst, src)); } | |
2106 void subptr(Register dst, int32_t src); | |
2107 void subptr(Register dst, Register src); | |
3363
167b70ff3abc
6939861: JVM should handle more conversion operations
never
parents:
3336
diff
changeset
|
2108 void subptr(Register dst, RegisterOrConstant src) { |
167b70ff3abc
6939861: JVM should handle more conversion operations
never
parents:
3336
diff
changeset
|
2109 if (src.is_constant()) subptr(dst, (int) src.as_constant()); |
167b70ff3abc
6939861: JVM should handle more conversion operations
never
parents:
3336
diff
changeset
|
2110 else subptr(dst, src.as_register()); |
167b70ff3abc
6939861: JVM should handle more conversion operations
never
parents:
3336
diff
changeset
|
2111 } |
304 | 2112 |
2113 void sbbptr(Address dst, int32_t src) { LP64_ONLY(sbbq(dst, src)) NOT_LP64(sbbl(dst, src)); } | |
2114 void sbbptr(Register dst, int32_t src) { LP64_ONLY(sbbq(dst, src)) NOT_LP64(sbbl(dst, src)); } | |
2115 | |
2116 void xchgptr(Register src1, Register src2) { LP64_ONLY(xchgq(src1, src2)) NOT_LP64(xchgl(src1, src2)) ; } | |
2117 void xchgptr(Register src1, Address src2) { LP64_ONLY(xchgq(src1, src2)) NOT_LP64(xchgl(src1, src2)) ; } | |
2118 | |
2119 void xaddptr(Address src1, Register src2) { LP64_ONLY(xaddq(src1, src2)) NOT_LP64(xaddl(src1, src2)) ; } | |
2120 | |
2121 | |
0 | 2122 |
2123 // Helper functions for statistics gathering. | |
2124 // Conditionally (atomically, on MPs) increments passed counter address, preserving condition codes. | |
2125 void cond_inc32(Condition cond, AddressLiteral counter_addr); | |
2126 // Unconditional atomic increment. | |
2127 void atomic_incl(AddressLiteral counter_addr); | |
2128 | |
2129 void lea(Register dst, AddressLiteral adr); | |
2130 void lea(Address dst, AddressLiteral adr); | |
304 | 2131 void lea(Register dst, Address adr) { Assembler::lea(dst, adr); } |
2132 | |
2133 void leal32(Register dst, Address src) { leal(dst, src); } | |
2134 | |
2404
b40d4fa697bf
6964776: c2 should ensure the polling page is reachable on 64 bit
iveresov
parents:
2320
diff
changeset
|
2135 // Import other testl() methods from the parent class or else |
b40d4fa697bf
6964776: c2 should ensure the polling page is reachable on 64 bit
iveresov
parents:
2320
diff
changeset
|
2136 // they will be hidden by the following overriding declaration. |
b40d4fa697bf
6964776: c2 should ensure the polling page is reachable on 64 bit
iveresov
parents:
2320
diff
changeset
|
2137 using Assembler::testl; |
b40d4fa697bf
6964776: c2 should ensure the polling page is reachable on 64 bit
iveresov
parents:
2320
diff
changeset
|
2138 void testl(Register dst, AddressLiteral src); |
304 | 2139 |
2140 void orptr(Register dst, Address src) { LP64_ONLY(orq(dst, src)) NOT_LP64(orl(dst, src)); } | |
2141 void orptr(Register dst, Register src) { LP64_ONLY(orq(dst, src)) NOT_LP64(orl(dst, src)); } | |
2142 void orptr(Register dst, int32_t src) { LP64_ONLY(orq(dst, src)) NOT_LP64(orl(dst, src)); } | |
2143 | |
2144 void testptr(Register src, int32_t imm32) { LP64_ONLY(testq(src, imm32)) NOT_LP64(testl(src, imm32)); } | |
2145 void testptr(Register src1, Register src2); | |
2146 | |
2147 void xorptr(Register dst, Register src) { LP64_ONLY(xorq(dst, src)) NOT_LP64(xorl(dst, src)); } | |
2148 void xorptr(Register dst, Address src) { LP64_ONLY(xorq(dst, src)) NOT_LP64(xorl(dst, src)); } | |
0 | 2149 |
2150 // Calls | |
2151 | |
2152 void call(Label& L, relocInfo::relocType rtype); | |
2153 void call(Register entry); | |
2154 | |
2155 // NOTE: this call tranfers to the effective address of entry NOT | |
2156 // the address contained by entry. This is because this is more natural | |
2157 // for jumps/calls. | |
2158 void call(AddressLiteral entry); | |
2159 | |
2160 // Jumps | |
2161 | |
2162 // NOTE: these jumps tranfer to the effective address of dst NOT | |
2163 // the address contained by dst. This is because this is more natural | |
2164 // for jumps/calls. | |
2165 void jump(AddressLiteral dst); | |
2166 void jump_cc(Condition cc, AddressLiteral dst); | |
2167 | |
2168 // 32bit can do a case table jump in one instruction but we no longer allow the base | |
2169 // to be installed in the Address class. This jump will tranfers to the address | |
2170 // contained in the location described by entry (not the address of entry) | |
2171 void jump(ArrayAddress entry); | |
2172 | |
2173 // Floating | |
2174 | |
2175 void andpd(XMMRegister dst, Address src) { Assembler::andpd(dst, src); } | |
2176 void andpd(XMMRegister dst, AddressLiteral src); | |
2177 | |
2178 void comiss(XMMRegister dst, Address src) { Assembler::comiss(dst, src); } | |
2179 void comiss(XMMRegister dst, AddressLiteral src); | |
2180 | |
2181 void comisd(XMMRegister dst, Address src) { Assembler::comisd(dst, src); } | |
2182 void comisd(XMMRegister dst, AddressLiteral src); | |
2183 | |
2008 | 2184 void fadd_s(Address src) { Assembler::fadd_s(src); } |
2185 void fadd_s(AddressLiteral src) { Assembler::fadd_s(as_Address(src)); } | |
2186 | |
0 | 2187 void fldcw(Address src) { Assembler::fldcw(src); } |
2188 void fldcw(AddressLiteral src); | |
2189 | |
2190 void fld_s(int index) { Assembler::fld_s(index); } | |
2191 void fld_s(Address src) { Assembler::fld_s(src); } | |
2192 void fld_s(AddressLiteral src); | |
2193 | |
2194 void fld_d(Address src) { Assembler::fld_d(src); } | |
2195 void fld_d(AddressLiteral src); | |
2196 | |
2197 void fld_x(Address src) { Assembler::fld_x(src); } | |
2198 void fld_x(AddressLiteral src); | |
2199 | |
2008 | 2200 void fmul_s(Address src) { Assembler::fmul_s(src); } |
2201 void fmul_s(AddressLiteral src) { Assembler::fmul_s(as_Address(src)); } | |
2202 | |
0 | 2203 void ldmxcsr(Address src) { Assembler::ldmxcsr(src); } |
2204 void ldmxcsr(AddressLiteral src); | |
2205 | |
304 | 2206 private: |
2207 // these are private because users should be doing movflt/movdbl | |
2208 | |
0 | 2209 void movss(Address dst, XMMRegister src) { Assembler::movss(dst, src); } |
2210 void movss(XMMRegister dst, XMMRegister src) { Assembler::movss(dst, src); } | |
2211 void movss(XMMRegister dst, Address src) { Assembler::movss(dst, src); } | |
2212 void movss(XMMRegister dst, AddressLiteral src); | |
2213 | |
304 | 2214 void movlpd(XMMRegister dst, Address src) {Assembler::movlpd(dst, src); } |
2215 void movlpd(XMMRegister dst, AddressLiteral src); | |
2216 | |
2217 public: | |
2218 | |
2008 | 2219 void addsd(XMMRegister dst, XMMRegister src) { Assembler::addsd(dst, src); } |
2220 void addsd(XMMRegister dst, Address src) { Assembler::addsd(dst, src); } | |
2221 void addsd(XMMRegister dst, AddressLiteral src) { Assembler::addsd(dst, as_Address(src)); } | |
2222 | |
2223 void addss(XMMRegister dst, XMMRegister src) { Assembler::addss(dst, src); } | |
2224 void addss(XMMRegister dst, Address src) { Assembler::addss(dst, src); } | |
2225 void addss(XMMRegister dst, AddressLiteral src) { Assembler::addss(dst, as_Address(src)); } | |
2226 | |
2227 void divsd(XMMRegister dst, XMMRegister src) { Assembler::divsd(dst, src); } | |
2228 void divsd(XMMRegister dst, Address src) { Assembler::divsd(dst, src); } | |
2229 void divsd(XMMRegister dst, AddressLiteral src) { Assembler::divsd(dst, as_Address(src)); } | |
2230 | |
2231 void divss(XMMRegister dst, XMMRegister src) { Assembler::divss(dst, src); } | |
2232 void divss(XMMRegister dst, Address src) { Assembler::divss(dst, src); } | |
2233 void divss(XMMRegister dst, AddressLiteral src) { Assembler::divss(dst, as_Address(src)); } | |
2234 | |
2100
b1a2afa37ec4
7003271: Hotspot should track cumulative Java heap bytes allocated on a per-thread basis
phh
parents:
2008
diff
changeset
|
2235 void movsd(XMMRegister dst, XMMRegister src) { Assembler::movsd(dst, src); } |
b1a2afa37ec4
7003271: Hotspot should track cumulative Java heap bytes allocated on a per-thread basis
phh
parents:
2008
diff
changeset
|
2236 void movsd(Address dst, XMMRegister src) { Assembler::movsd(dst, src); } |
b1a2afa37ec4
7003271: Hotspot should track cumulative Java heap bytes allocated on a per-thread basis
phh
parents:
2008
diff
changeset
|
2237 void movsd(XMMRegister dst, Address src) { Assembler::movsd(dst, src); } |
2008 | 2238 void movsd(XMMRegister dst, AddressLiteral src) { Assembler::movsd(dst, as_Address(src)); } |
2239 | |
2240 void mulsd(XMMRegister dst, XMMRegister src) { Assembler::mulsd(dst, src); } | |
2241 void mulsd(XMMRegister dst, Address src) { Assembler::mulsd(dst, src); } | |
2242 void mulsd(XMMRegister dst, AddressLiteral src) { Assembler::mulsd(dst, as_Address(src)); } | |
2243 | |
2244 void mulss(XMMRegister dst, XMMRegister src) { Assembler::mulss(dst, src); } | |
2245 void mulss(XMMRegister dst, Address src) { Assembler::mulss(dst, src); } | |
2246 void mulss(XMMRegister dst, AddressLiteral src) { Assembler::mulss(dst, as_Address(src)); } | |
2247 | |
2248 void sqrtsd(XMMRegister dst, XMMRegister src) { Assembler::sqrtsd(dst, src); } | |
2249 void sqrtsd(XMMRegister dst, Address src) { Assembler::sqrtsd(dst, src); } | |
2250 void sqrtsd(XMMRegister dst, AddressLiteral src) { Assembler::sqrtsd(dst, as_Address(src)); } | |
2251 | |
2252 void sqrtss(XMMRegister dst, XMMRegister src) { Assembler::sqrtss(dst, src); } | |
2253 void sqrtss(XMMRegister dst, Address src) { Assembler::sqrtss(dst, src); } | |
2254 void sqrtss(XMMRegister dst, AddressLiteral src) { Assembler::sqrtss(dst, as_Address(src)); } | |
2255 | |
2256 void subsd(XMMRegister dst, XMMRegister src) { Assembler::subsd(dst, src); } | |
2257 void subsd(XMMRegister dst, Address src) { Assembler::subsd(dst, src); } | |
2258 void subsd(XMMRegister dst, AddressLiteral src) { Assembler::subsd(dst, as_Address(src)); } | |
2259 | |
2260 void subss(XMMRegister dst, XMMRegister src) { Assembler::subss(dst, src); } | |
2261 void subss(XMMRegister dst, Address src) { Assembler::subss(dst, src); } | |
2262 void subss(XMMRegister dst, AddressLiteral src) { Assembler::subss(dst, as_Address(src)); } | |
0 | 2263 |
2264 void ucomiss(XMMRegister dst, XMMRegister src) { Assembler::ucomiss(dst, src); } | |
2265 void ucomiss(XMMRegister dst, Address src) { Assembler::ucomiss(dst, src); } | |
2266 void ucomiss(XMMRegister dst, AddressLiteral src); | |
2267 | |
2268 void ucomisd(XMMRegister dst, XMMRegister src) { Assembler::ucomisd(dst, src); } | |
2269 void ucomisd(XMMRegister dst, Address src) { Assembler::ucomisd(dst, src); } | |
2270 void ucomisd(XMMRegister dst, AddressLiteral src); | |
2271 | |
2272 // Bitwise Logical XOR of Packed Double-Precision Floating-Point Values | |
2273 void xorpd(XMMRegister dst, XMMRegister src) { Assembler::xorpd(dst, src); } | |
2274 void xorpd(XMMRegister dst, Address src) { Assembler::xorpd(dst, src); } | |
2275 void xorpd(XMMRegister dst, AddressLiteral src); | |
2276 | |
2277 // Bitwise Logical XOR of Packed Single-Precision Floating-Point Values | |
2278 void xorps(XMMRegister dst, XMMRegister src) { Assembler::xorps(dst, src); } | |
2279 void xorps(XMMRegister dst, Address src) { Assembler::xorps(dst, src); } | |
2280 void xorps(XMMRegister dst, AddressLiteral src); | |
2281 | |
2282 // Data | |
2283 | |
2415
09f96c3ff1ad
7032388: guarantee(VM_Version::supports_cmov()) failed: illegal instruction on i586 after 6919934
twisti
parents:
2404
diff
changeset
|
2284 void cmov32( Condition cc, Register dst, Address src); |
09f96c3ff1ad
7032388: guarantee(VM_Version::supports_cmov()) failed: illegal instruction on i586 after 6919934
twisti
parents:
2404
diff
changeset
|
2285 void cmov32( Condition cc, Register dst, Register src); |
09f96c3ff1ad
7032388: guarantee(VM_Version::supports_cmov()) failed: illegal instruction on i586 after 6919934
twisti
parents:
2404
diff
changeset
|
2286 |
09f96c3ff1ad
7032388: guarantee(VM_Version::supports_cmov()) failed: illegal instruction on i586 after 6919934
twisti
parents:
2404
diff
changeset
|
2287 void cmov( Condition cc, Register dst, Register src) { cmovptr(cc, dst, src); } |
09f96c3ff1ad
7032388: guarantee(VM_Version::supports_cmov()) failed: illegal instruction on i586 after 6919934
twisti
parents:
2404
diff
changeset
|
2288 |
09f96c3ff1ad
7032388: guarantee(VM_Version::supports_cmov()) failed: illegal instruction on i586 after 6919934
twisti
parents:
2404
diff
changeset
|
2289 void cmovptr(Condition cc, Register dst, Address src) { LP64_ONLY(cmovq(cc, dst, src)) NOT_LP64(cmov32(cc, dst, src)); } |
09f96c3ff1ad
7032388: guarantee(VM_Version::supports_cmov()) failed: illegal instruction on i586 after 6919934
twisti
parents:
2404
diff
changeset
|
2290 void cmovptr(Condition cc, Register dst, Register src) { LP64_ONLY(cmovq(cc, dst, src)) NOT_LP64(cmov32(cc, dst, src)); } |
304 | 2291 |
0 | 2292 void movoop(Register dst, jobject obj); |
2293 void movoop(Address dst, jobject obj); | |
2294 | |
2295 void movptr(ArrayAddress dst, Register src); | |
2296 // can this do an lea? | |
2297 void movptr(Register dst, ArrayAddress src); | |
2298 | |
304 | 2299 void movptr(Register dst, Address src); |
2300 | |
0 | 2301 void movptr(Register dst, AddressLiteral src); |
2302 | |
304 | 2303 void movptr(Register dst, intptr_t src); |
2304 void movptr(Register dst, Register src); | |
2305 void movptr(Address dst, intptr_t src); | |
2306 | |
2307 void movptr(Address dst, Register src); | |
2308 | |
3363
167b70ff3abc
6939861: JVM should handle more conversion operations
never
parents:
3336
diff
changeset
|
2309 void movptr(Register dst, RegisterOrConstant src) { |
167b70ff3abc
6939861: JVM should handle more conversion operations
never
parents:
3336
diff
changeset
|
2310 if (src.is_constant()) movptr(dst, src.as_constant()); |
167b70ff3abc
6939861: JVM should handle more conversion operations
never
parents:
3336
diff
changeset
|
2311 else movptr(dst, src.as_register()); |
167b70ff3abc
6939861: JVM should handle more conversion operations
never
parents:
3336
diff
changeset
|
2312 } |
167b70ff3abc
6939861: JVM should handle more conversion operations
never
parents:
3336
diff
changeset
|
2313 |
304 | 2314 #ifdef _LP64 |
2315 // Generally the next two are only used for moving NULL | |
2316 // Although there are situations in initializing the mark word where | |
2317 // they could be used. They are dangerous. | |
2318 | |
2319 // They only exist on LP64 so that int32_t and intptr_t are not the same | |
2320 // and we have ambiguous declarations. | |
2321 | |
2322 void movptr(Address dst, int32_t imm32); | |
2323 void movptr(Register dst, int32_t imm32); | |
2324 #endif // _LP64 | |
2325 | |
0 | 2326 // to avoid hiding movl |
2327 void mov32(AddressLiteral dst, Register src); | |
2328 void mov32(Register dst, AddressLiteral src); | |
304 | 2329 |
0 | 2330 // to avoid hiding movb |
2331 void movbyte(ArrayAddress dst, int src); | |
2332 | |
2333 // Can push value or effective address | |
2334 void pushptr(AddressLiteral src); | |
2335 | |
304 | 2336 void pushptr(Address src) { LP64_ONLY(pushq(src)) NOT_LP64(pushl(src)); } |
2337 void popptr(Address src) { LP64_ONLY(popq(src)) NOT_LP64(popl(src)); } | |
2338 | |
2339 void pushoop(jobject obj); | |
2340 | |
2341 // sign extend as need a l to ptr sized element | |
2342 void movl2ptr(Register dst, Address src) { LP64_ONLY(movslq(dst, src)) NOT_LP64(movl(dst, src)); } | |
2343 void movl2ptr(Register dst, Register src) { LP64_ONLY(movslq(dst, src)) NOT_LP64(if (dst != src) movl(dst, src)); } | |
2344 | |
986
62001a362ce9
6827605: new String intrinsics may prevent EA scalar replacement
kvn
parents:
775
diff
changeset
|
2345 // IndexOf strings. |
2320
41d4973cf100
6942326: x86 code in string_indexof() could read beyond reserved heap space
kvn
parents:
2262
diff
changeset
|
2346 // Small strings are loaded through stack if they cross page boundary. |
986
62001a362ce9
6827605: new String intrinsics may prevent EA scalar replacement
kvn
parents:
775
diff
changeset
|
2347 void string_indexof(Register str1, Register str2, |
2320
41d4973cf100
6942326: x86 code in string_indexof() could read beyond reserved heap space
kvn
parents:
2262
diff
changeset
|
2348 Register cnt1, Register cnt2, |
41d4973cf100
6942326: x86 code in string_indexof() could read beyond reserved heap space
kvn
parents:
2262
diff
changeset
|
2349 int int_cnt2, Register result, |
986
62001a362ce9
6827605: new String intrinsics may prevent EA scalar replacement
kvn
parents:
775
diff
changeset
|
2350 XMMRegister vec, Register tmp); |
62001a362ce9
6827605: new String intrinsics may prevent EA scalar replacement
kvn
parents:
775
diff
changeset
|
2351 |
2320
41d4973cf100
6942326: x86 code in string_indexof() could read beyond reserved heap space
kvn
parents:
2262
diff
changeset
|
2352 // IndexOf for constant substrings with size >= 8 elements |
41d4973cf100
6942326: x86 code in string_indexof() could read beyond reserved heap space
kvn
parents:
2262
diff
changeset
|
2353 // which don't need to be loaded through stack. |
41d4973cf100
6942326: x86 code in string_indexof() could read beyond reserved heap space
kvn
parents:
2262
diff
changeset
|
2354 void string_indexofC8(Register str1, Register str2, |
41d4973cf100
6942326: x86 code in string_indexof() could read beyond reserved heap space
kvn
parents:
2262
diff
changeset
|
2355 Register cnt1, Register cnt2, |
41d4973cf100
6942326: x86 code in string_indexof() could read beyond reserved heap space
kvn
parents:
2262
diff
changeset
|
2356 int int_cnt2, Register result, |
41d4973cf100
6942326: x86 code in string_indexof() could read beyond reserved heap space
kvn
parents:
2262
diff
changeset
|
2357 XMMRegister vec, Register tmp); |
41d4973cf100
6942326: x86 code in string_indexof() could read beyond reserved heap space
kvn
parents:
2262
diff
changeset
|
2358 |
41d4973cf100
6942326: x86 code in string_indexof() could read beyond reserved heap space
kvn
parents:
2262
diff
changeset
|
2359 // Smallest code: we don't need to load through stack, |
41d4973cf100
6942326: x86 code in string_indexof() could read beyond reserved heap space
kvn
parents:
2262
diff
changeset
|
2360 // check string tail. |
41d4973cf100
6942326: x86 code in string_indexof() could read beyond reserved heap space
kvn
parents:
2262
diff
changeset
|
2361 |
986
62001a362ce9
6827605: new String intrinsics may prevent EA scalar replacement
kvn
parents:
775
diff
changeset
|
2362 // Compare strings. |
62001a362ce9
6827605: new String intrinsics may prevent EA scalar replacement
kvn
parents:
775
diff
changeset
|
2363 void string_compare(Register str1, Register str2, |
62001a362ce9
6827605: new String intrinsics may prevent EA scalar replacement
kvn
parents:
775
diff
changeset
|
2364 Register cnt1, Register cnt2, Register result, |
2262 | 2365 XMMRegister vec1); |
986
62001a362ce9
6827605: new String intrinsics may prevent EA scalar replacement
kvn
parents:
775
diff
changeset
|
2366 |
62001a362ce9
6827605: new String intrinsics may prevent EA scalar replacement
kvn
parents:
775
diff
changeset
|
2367 // Compare char[] arrays. |
62001a362ce9
6827605: new String intrinsics may prevent EA scalar replacement
kvn
parents:
775
diff
changeset
|
2368 void char_arrays_equals(bool is_array_equ, Register ary1, Register ary2, |
62001a362ce9
6827605: new String intrinsics may prevent EA scalar replacement
kvn
parents:
775
diff
changeset
|
2369 Register limit, Register result, Register chr, |
62001a362ce9
6827605: new String intrinsics may prevent EA scalar replacement
kvn
parents:
775
diff
changeset
|
2370 XMMRegister vec1, XMMRegister vec2); |
304 | 2371 |
1763 | 2372 // Fill primitive arrays |
2373 void generate_fill(BasicType t, bool aligned, | |
2374 Register to, Register value, Register count, | |
2375 Register rtmp, XMMRegister xtmp); | |
2376 | |
0 | 2377 #undef VIRTUAL |
2378 | |
2379 }; | |
2380 | |
2381 /** | |
2382 * class SkipIfEqual: | |
2383 * | |
2384 * Instantiating this class will result in assembly code being output that will | |
2385 * jump around any code emitted between the creation of the instance and it's | |
2386 * automatic destruction at the end of a scope block, depending on the value of | |
2387 * the flag passed to the constructor, which will be checked at run-time. | |
2388 */ | |
2389 class SkipIfEqual { | |
2390 private: | |
2391 MacroAssembler* _masm; | |
2392 Label _label; | |
2393 | |
2394 public: | |
2395 SkipIfEqual(MacroAssembler*, const bool* flag_addr, bool value); | |
2396 ~SkipIfEqual(); | |
2397 }; | |
2398 | |
2399 #ifdef ASSERT | |
2400 inline bool AbstractAssembler::pd_check_instruction_mark() { return true; } | |
2401 #endif | |
1972 | 2402 |
2403 #endif // CPU_X86_VM_ASSEMBLER_X86_HPP |