1 /*
2 * Copyright (c) 2002, 2017, Oracle and/or its affiliates. All rights reserved.
3 * Copyright (c) 2012, 2017 SAP SE. All rights reserved.
4 * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
5 *
6 * This code is free software; you can redistribute it and/or modify it
7 * under the terms of the GNU General Public License version 2 only, as
8 * published by the Free Software Foundation.
9 *
10 * This code is distributed in the hope that it will be useful, but WITHOUT
11 * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
12 * FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License
13 * version 2 for more details (a copy is included in the LICENSE file that
14 * accompanied this code).
15 *
16 * You should have received a copy of the GNU General Public License version
17 * 2 along with this work; if not, write to the Free Software Foundation,
18 * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA.
19 *
20 * Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA
21 * or visit www.oracle.com if you need additional information or have any
22 * questions.
23 *
24 */
25
26 #ifndef CPU_PPC_VM_ASSEMBLER_PPC_HPP
27 #define CPU_PPC_VM_ASSEMBLER_PPC_HPP
28
29 #include "asm/register.hpp"
30
31 // Address is an abstraction used to represent a memory location
32 // as used in assembler instructions.
33 // PPC instructions grok either baseReg + indexReg or baseReg + disp.
34 class Address VALUE_OBJ_CLASS_SPEC {
35 private:
36 Register _base; // Base register.
37 Register _index; // Index register.
38 intptr_t _disp; // Displacement.
39
40 public:
41 Address(Register b, Register i, address d = 0)
42 : _base(b), _index(i), _disp((intptr_t)d) {
43 assert(i == noreg || d == 0, "can't have both");
44 }
45
46 Address(Register b, address d = 0)
47 : _base(b), _index(noreg), _disp((intptr_t)d) {}
48
49 Address(Register b, intptr_t d)
50 : _base(b), _index(noreg), _disp(d) {}
51
52 Address(Register b, RegisterOrConstant roc)
53 : _base(b), _index(noreg), _disp(0) {
54 if (roc.is_constant()) _disp = roc.as_constant(); else _index = roc.as_register();
55 }
56
57 Address()
58 : _base(noreg), _index(noreg), _disp(0) {}
59
60 // accessors
61 Register base() const { return _base; }
62 Register index() const { return _index; }
63 int disp() const { return (int)_disp; }
64 bool is_const() const { return _base == noreg && _index == noreg; }
65 };
66
67 class AddressLiteral VALUE_OBJ_CLASS_SPEC {
68 private:
69 address _address;
70 RelocationHolder _rspec;
71
72 RelocationHolder rspec_from_rtype(relocInfo::relocType rtype, address addr) {
73 switch (rtype) {
74 case relocInfo::external_word_type:
75 return external_word_Relocation::spec(addr);
76 case relocInfo::internal_word_type:
77 return internal_word_Relocation::spec(addr);
78 case relocInfo::opt_virtual_call_type:
79 return opt_virtual_call_Relocation::spec();
80 case relocInfo::static_call_type:
81 return static_call_Relocation::spec();
82 case relocInfo::runtime_call_type:
83 return runtime_call_Relocation::spec();
84 case relocInfo::none:
85 return RelocationHolder();
86 default:
87 ShouldNotReachHere();
88 return RelocationHolder();
89 }
90 }
91
92 protected:
93 // creation
94 AddressLiteral() : _address(NULL), _rspec(NULL) {}
95
96 public:
97 AddressLiteral(address addr, RelocationHolder const& rspec)
98 : _address(addr),
99 _rspec(rspec) {}
100
101 AddressLiteral(address addr, relocInfo::relocType rtype = relocInfo::none)
102 : _address((address) addr),
103 _rspec(rspec_from_rtype(rtype, (address) addr)) {}
104
105 AddressLiteral(oop* addr, relocInfo::relocType rtype = relocInfo::none)
106 : _address((address) addr),
107 _rspec(rspec_from_rtype(rtype, (address) addr)) {}
108
109 intptr_t value() const { return (intptr_t) _address; }
110
111 const RelocationHolder& rspec() const { return _rspec; }
112 };
113
114 // Argument is an abstraction used to represent an outgoing
115 // actual argument or an incoming formal parameter, whether
116 // it resides in memory or in a register, in a manner consistent
117 // with the PPC Application Binary Interface, or ABI. This is
118 // often referred to as the native or C calling convention.
119
120 class Argument VALUE_OBJ_CLASS_SPEC {
121 private:
122 int _number; // The number of the argument.
123 public:
124 enum {
125 // Only 8 registers may contain integer parameters.
126 n_register_parameters = 8,
127 // Can have up to 8 floating registers.
128 n_float_register_parameters = 8,
129
130 // PPC C calling conventions.
131 // The first eight arguments are passed in int regs if they are int.
132 n_int_register_parameters_c = 8,
133 // The first thirteen float arguments are passed in float regs.
134 n_float_register_parameters_c = 13,
135 // Only the first 8 parameters are not placed on the stack. Aix disassembly
136 // shows that xlC places all float args after argument 8 on the stack AND
137 // in a register. This is not documented, but we follow this convention, too.
138 n_regs_not_on_stack_c = 8,
139 };
140 // creation
141 Argument(int number) : _number(number) {}
142
143 int number() const { return _number; }
144
145 // Locating register-based arguments:
146 bool is_register() const { return _number < n_register_parameters; }
147
148 Register as_register() const {
149 assert(is_register(), "must be a register argument");
150 return as_Register(number() + R3_ARG1->encoding());
151 }
152 };
153
154 #if !defined(ABI_ELFv2)
155 // A ppc64 function descriptor.
156 struct FunctionDescriptor VALUE_OBJ_CLASS_SPEC {
157 private:
158 address _entry;
159 address _toc;
160 address _env;
161
162 public:
163 inline address entry() const { return _entry; }
164 inline address toc() const { return _toc; }
165 inline address env() const { return _env; }
166
167 inline void set_entry(address entry) { _entry = entry; }
168 inline void set_toc( address toc) { _toc = toc; }
169 inline void set_env( address env) { _env = env; }
170
171 inline static ByteSize entry_offset() { return byte_offset_of(FunctionDescriptor, _entry); }
172 inline static ByteSize toc_offset() { return byte_offset_of(FunctionDescriptor, _toc); }
173 inline static ByteSize env_offset() { return byte_offset_of(FunctionDescriptor, _env); }
174
175 // Friend functions can be called without loading toc and env.
176 enum {
177 friend_toc = 0xcafe,
178 friend_env = 0xc0de
179 };
180
181 inline bool is_friend_function() const {
182 return (toc() == (address) friend_toc) && (env() == (address) friend_env);
183 }
184
185 // Constructor for stack-allocated instances.
186 FunctionDescriptor() {
187 _entry = (address) 0xbad;
188 _toc = (address) 0xbad;
189 _env = (address) 0xbad;
190 }
191 };
192 #endif
193
194
195 // The PPC Assembler: Pure assembler doing NO optimizations on the
196 // instruction level; i.e., what you write is what you get. The
197 // Assembler is generating code into a CodeBuffer.
198
199 class Assembler : public AbstractAssembler {
200 protected:
201 // Displacement routines
202 static int patched_branch(int dest_pos, int inst, int inst_pos);
203 static int branch_destination(int inst, int pos);
204
205 friend class AbstractAssembler;
206
207 // Code patchers need various routines like inv_wdisp()
208 friend class NativeInstruction;
209 friend class NativeGeneralJump;
210 friend class Relocation;
211
212 public:
213
214 enum shifts {
215 XO_21_29_SHIFT = 2,
216 XO_21_30_SHIFT = 1,
217 XO_27_29_SHIFT = 2,
218 XO_30_31_SHIFT = 0,
219 SPR_5_9_SHIFT = 11u, // SPR_5_9 field in bits 11 -- 15
220 SPR_0_4_SHIFT = 16u, // SPR_0_4 field in bits 16 -- 20
221 RS_SHIFT = 21u, // RS field in bits 21 -- 25
222 OPCODE_SHIFT = 26u, // opcode in bits 26 -- 31
223 };
224
225 enum opcdxos_masks {
226 XL_FORM_OPCODE_MASK = (63u << OPCODE_SHIFT) | (1023u << 1),
227 ADDI_OPCODE_MASK = (63u << OPCODE_SHIFT),
228 ADDIS_OPCODE_MASK = (63u << OPCODE_SHIFT),
229 BXX_OPCODE_MASK = (63u << OPCODE_SHIFT),
230 BCXX_OPCODE_MASK = (63u << OPCODE_SHIFT),
231 // trap instructions
232 TDI_OPCODE_MASK = (63u << OPCODE_SHIFT),
233 TWI_OPCODE_MASK = (63u << OPCODE_SHIFT),
234 TD_OPCODE_MASK = (63u << OPCODE_SHIFT) | (1023u << 1),
235 TW_OPCODE_MASK = (63u << OPCODE_SHIFT) | (1023u << 1),
236 LD_OPCODE_MASK = (63u << OPCODE_SHIFT) | (3u << XO_30_31_SHIFT), // DS-FORM
237 STD_OPCODE_MASK = LD_OPCODE_MASK,
238 STDU_OPCODE_MASK = STD_OPCODE_MASK,
239 STDX_OPCODE_MASK = (63u << OPCODE_SHIFT) | (1023u << 1),
240 STDUX_OPCODE_MASK = STDX_OPCODE_MASK,
241 STW_OPCODE_MASK = (63u << OPCODE_SHIFT),
242 STWU_OPCODE_MASK = STW_OPCODE_MASK,
243 STWX_OPCODE_MASK = (63u << OPCODE_SHIFT) | (1023u << 1),
244 STWUX_OPCODE_MASK = STWX_OPCODE_MASK,
245 MTCTR_OPCODE_MASK = ~(31u << RS_SHIFT),
246 ORI_OPCODE_MASK = (63u << OPCODE_SHIFT),
247 ORIS_OPCODE_MASK = (63u << OPCODE_SHIFT),
248 RLDICR_OPCODE_MASK = (63u << OPCODE_SHIFT) | (7u << XO_27_29_SHIFT)
249 };
250
251 enum opcdxos {
252 ADD_OPCODE = (31u << OPCODE_SHIFT | 266u << 1),
253 ADDC_OPCODE = (31u << OPCODE_SHIFT | 10u << 1),
254 ADDI_OPCODE = (14u << OPCODE_SHIFT),
255 ADDIS_OPCODE = (15u << OPCODE_SHIFT),
256 ADDIC__OPCODE = (13u << OPCODE_SHIFT),
257 ADDE_OPCODE = (31u << OPCODE_SHIFT | 138u << 1),
258 ADDME_OPCODE = (31u << OPCODE_SHIFT | 234u << 1),
259 ADDZE_OPCODE = (31u << OPCODE_SHIFT | 202u << 1),
260 SUBF_OPCODE = (31u << OPCODE_SHIFT | 40u << 1),
261 SUBFC_OPCODE = (31u << OPCODE_SHIFT | 8u << 1),
262 SUBFE_OPCODE = (31u << OPCODE_SHIFT | 136u << 1),
263 SUBFIC_OPCODE = (8u << OPCODE_SHIFT),
264 SUBFME_OPCODE = (31u << OPCODE_SHIFT | 232u << 1),
265 SUBFZE_OPCODE = (31u << OPCODE_SHIFT | 200u << 1),
266 DIVW_OPCODE = (31u << OPCODE_SHIFT | 491u << 1),
267 MULLW_OPCODE = (31u << OPCODE_SHIFT | 235u << 1),
268 MULHW_OPCODE = (31u << OPCODE_SHIFT | 75u << 1),
269 MULHWU_OPCODE = (31u << OPCODE_SHIFT | 11u << 1),
270 MULLI_OPCODE = (7u << OPCODE_SHIFT),
271 AND_OPCODE = (31u << OPCODE_SHIFT | 28u << 1),
272 ANDI_OPCODE = (28u << OPCODE_SHIFT),
273 ANDIS_OPCODE = (29u << OPCODE_SHIFT),
274 ANDC_OPCODE = (31u << OPCODE_SHIFT | 60u << 1),
275 ORC_OPCODE = (31u << OPCODE_SHIFT | 412u << 1),
276 OR_OPCODE = (31u << OPCODE_SHIFT | 444u << 1),
277 ORI_OPCODE = (24u << OPCODE_SHIFT),
278 ORIS_OPCODE = (25u << OPCODE_SHIFT),
279 XOR_OPCODE = (31u << OPCODE_SHIFT | 316u << 1),
280 XORI_OPCODE = (26u << OPCODE_SHIFT),
281 XORIS_OPCODE = (27u << OPCODE_SHIFT),
282
283 NEG_OPCODE = (31u << OPCODE_SHIFT | 104u << 1),
284
285 RLWINM_OPCODE = (21u << OPCODE_SHIFT),
286 CLRRWI_OPCODE = RLWINM_OPCODE,
287 CLRLWI_OPCODE = RLWINM_OPCODE,
288
289 RLWIMI_OPCODE = (20u << OPCODE_SHIFT),
290
291 SLW_OPCODE = (31u << OPCODE_SHIFT | 24u << 1),
292 SLWI_OPCODE = RLWINM_OPCODE,
293 SRW_OPCODE = (31u << OPCODE_SHIFT | 536u << 1),
294 SRWI_OPCODE = RLWINM_OPCODE,
295 SRAW_OPCODE = (31u << OPCODE_SHIFT | 792u << 1),
296 SRAWI_OPCODE = (31u << OPCODE_SHIFT | 824u << 1),
297
298 CMP_OPCODE = (31u << OPCODE_SHIFT | 0u << 1),
299 CMPI_OPCODE = (11u << OPCODE_SHIFT),
300 CMPL_OPCODE = (31u << OPCODE_SHIFT | 32u << 1),
301 CMPLI_OPCODE = (10u << OPCODE_SHIFT),
302
303 ISEL_OPCODE = (31u << OPCODE_SHIFT | 15u << 1),
304
305 // Special purpose registers
306 MTSPR_OPCODE = (31u << OPCODE_SHIFT | 467u << 1),
307 MFSPR_OPCODE = (31u << OPCODE_SHIFT | 339u << 1),
308
309 MTXER_OPCODE = (MTSPR_OPCODE | 1 << SPR_0_4_SHIFT),
310 MFXER_OPCODE = (MFSPR_OPCODE | 1 << SPR_0_4_SHIFT),
311
312 MTDSCR_OPCODE = (MTSPR_OPCODE | 3 << SPR_0_4_SHIFT),
313 MFDSCR_OPCODE = (MFSPR_OPCODE | 3 << SPR_0_4_SHIFT),
314
315 MTLR_OPCODE = (MTSPR_OPCODE | 8 << SPR_0_4_SHIFT),
316 MFLR_OPCODE = (MFSPR_OPCODE | 8 << SPR_0_4_SHIFT),
317
318 MTCTR_OPCODE = (MTSPR_OPCODE | 9 << SPR_0_4_SHIFT),
319 MFCTR_OPCODE = (MFSPR_OPCODE | 9 << SPR_0_4_SHIFT),
320
321 // Attention: Higher and lower half are inserted in reversed order.
322 MTTFHAR_OPCODE = (MTSPR_OPCODE | 4 << SPR_5_9_SHIFT | 0 << SPR_0_4_SHIFT),
323 MFTFHAR_OPCODE = (MFSPR_OPCODE | 4 << SPR_5_9_SHIFT | 0 << SPR_0_4_SHIFT),
324 MTTFIAR_OPCODE = (MTSPR_OPCODE | 4 << SPR_5_9_SHIFT | 1 << SPR_0_4_SHIFT),
325 MFTFIAR_OPCODE = (MFSPR_OPCODE | 4 << SPR_5_9_SHIFT | 1 << SPR_0_4_SHIFT),
326 MTTEXASR_OPCODE = (MTSPR_OPCODE | 4 << SPR_5_9_SHIFT | 2 << SPR_0_4_SHIFT),
327 MFTEXASR_OPCODE = (MFSPR_OPCODE | 4 << SPR_5_9_SHIFT | 2 << SPR_0_4_SHIFT),
328 MTTEXASRU_OPCODE = (MTSPR_OPCODE | 4 << SPR_5_9_SHIFT | 3 << SPR_0_4_SHIFT),
329 MFTEXASRU_OPCODE = (MFSPR_OPCODE | 4 << SPR_5_9_SHIFT | 3 << SPR_0_4_SHIFT),
330
331 MTVRSAVE_OPCODE = (MTSPR_OPCODE | 8 << SPR_5_9_SHIFT | 0 << SPR_0_4_SHIFT),
332 MFVRSAVE_OPCODE = (MFSPR_OPCODE | 8 << SPR_5_9_SHIFT | 0 << SPR_0_4_SHIFT),
333
334 MFTB_OPCODE = (MFSPR_OPCODE | 8 << SPR_5_9_SHIFT | 12 << SPR_0_4_SHIFT),
335
336 MTCRF_OPCODE = (31u << OPCODE_SHIFT | 144u << 1),
337 MFCR_OPCODE = (31u << OPCODE_SHIFT | 19u << 1),
338 MCRF_OPCODE = (19u << OPCODE_SHIFT | 0u << 1),
339
340 // condition register logic instructions
341 CRAND_OPCODE = (19u << OPCODE_SHIFT | 257u << 1),
342 CRNAND_OPCODE = (19u << OPCODE_SHIFT | 225u << 1),
343 CROR_OPCODE = (19u << OPCODE_SHIFT | 449u << 1),
344 CRXOR_OPCODE = (19u << OPCODE_SHIFT | 193u << 1),
345 CRNOR_OPCODE = (19u << OPCODE_SHIFT | 33u << 1),
346 CREQV_OPCODE = (19u << OPCODE_SHIFT | 289u << 1),
347 CRANDC_OPCODE = (19u << OPCODE_SHIFT | 129u << 1),
348 CRORC_OPCODE = (19u << OPCODE_SHIFT | 417u << 1),
349
350 BCLR_OPCODE = (19u << OPCODE_SHIFT | 16u << 1),
351 BXX_OPCODE = (18u << OPCODE_SHIFT),
352 BCXX_OPCODE = (16u << OPCODE_SHIFT),
353
354 // CTR-related opcodes
355 BCCTR_OPCODE = (19u << OPCODE_SHIFT | 528u << 1),
356
357 LWZ_OPCODE = (32u << OPCODE_SHIFT),
358 LWZX_OPCODE = (31u << OPCODE_SHIFT | 23u << 1),
359 LWZU_OPCODE = (33u << OPCODE_SHIFT),
360 LWBRX_OPCODE = (31u << OPCODE_SHIFT | 534 << 1),
361
362 LHA_OPCODE = (42u << OPCODE_SHIFT),
363 LHAX_OPCODE = (31u << OPCODE_SHIFT | 343u << 1),
364 LHAU_OPCODE = (43u << OPCODE_SHIFT),
365
366 LHZ_OPCODE = (40u << OPCODE_SHIFT),
367 LHZX_OPCODE = (31u << OPCODE_SHIFT | 279u << 1),
368 LHZU_OPCODE = (41u << OPCODE_SHIFT),
369 LHBRX_OPCODE = (31u << OPCODE_SHIFT | 790 << 1),
370
371 LBZ_OPCODE = (34u << OPCODE_SHIFT),
372 LBZX_OPCODE = (31u << OPCODE_SHIFT | 87u << 1),
373 LBZU_OPCODE = (35u << OPCODE_SHIFT),
374
375 STW_OPCODE = (36u << OPCODE_SHIFT),
376 STWX_OPCODE = (31u << OPCODE_SHIFT | 151u << 1),
377 STWU_OPCODE = (37u << OPCODE_SHIFT),
378 STWUX_OPCODE = (31u << OPCODE_SHIFT | 183u << 1),
379 STWBRX_OPCODE = (31u << OPCODE_SHIFT | 662u << 1),
380
381 STH_OPCODE = (44u << OPCODE_SHIFT),
382 STHX_OPCODE = (31u << OPCODE_SHIFT | 407u << 1),
383 STHU_OPCODE = (45u << OPCODE_SHIFT),
384 STHBRX_OPCODE = (31u << OPCODE_SHIFT | 918u << 1),
385
386 STB_OPCODE = (38u << OPCODE_SHIFT),
387 STBX_OPCODE = (31u << OPCODE_SHIFT | 215u << 1),
388 STBU_OPCODE = (39u << OPCODE_SHIFT),
389
390 EXTSB_OPCODE = (31u << OPCODE_SHIFT | 954u << 1),
391 EXTSH_OPCODE = (31u << OPCODE_SHIFT | 922u << 1),
392 EXTSW_OPCODE = (31u << OPCODE_SHIFT | 986u << 1), // X-FORM
393
394 // 32 bit opcode encodings
395
396 LWA_OPCODE = (58u << OPCODE_SHIFT | 2u << XO_30_31_SHIFT), // DS-FORM
397 LWAX_OPCODE = (31u << OPCODE_SHIFT | 341u << XO_21_30_SHIFT), // X-FORM
398
399 CNTLZW_OPCODE = (31u << OPCODE_SHIFT | 26u << XO_21_30_SHIFT), // X-FORM
400
401 // 64 bit opcode encodings
402
403 LD_OPCODE = (58u << OPCODE_SHIFT | 0u << XO_30_31_SHIFT), // DS-FORM
404 LDU_OPCODE = (58u << OPCODE_SHIFT | 1u << XO_30_31_SHIFT), // DS-FORM
405 LDX_OPCODE = (31u << OPCODE_SHIFT | 21u << XO_21_30_SHIFT), // X-FORM
406 LDBRX_OPCODE = (31u << OPCODE_SHIFT | 532u << 1), // X-FORM
407
408 STD_OPCODE = (62u << OPCODE_SHIFT | 0u << XO_30_31_SHIFT), // DS-FORM
409 STDU_OPCODE = (62u << OPCODE_SHIFT | 1u << XO_30_31_SHIFT), // DS-FORM
410 STDUX_OPCODE = (31u << OPCODE_SHIFT | 181u << 1), // X-FORM
411 STDX_OPCODE = (31u << OPCODE_SHIFT | 149u << XO_21_30_SHIFT), // X-FORM
412 STDBRX_OPCODE = (31u << OPCODE_SHIFT | 660u << 1), // X-FORM
413
414 RLDICR_OPCODE = (30u << OPCODE_SHIFT | 1u << XO_27_29_SHIFT), // MD-FORM
415 RLDICL_OPCODE = (30u << OPCODE_SHIFT | 0u << XO_27_29_SHIFT), // MD-FORM
416 RLDIC_OPCODE = (30u << OPCODE_SHIFT | 2u << XO_27_29_SHIFT), // MD-FORM
417 RLDIMI_OPCODE = (30u << OPCODE_SHIFT | 3u << XO_27_29_SHIFT), // MD-FORM
418
419 SRADI_OPCODE = (31u << OPCODE_SHIFT | 413u << XO_21_29_SHIFT), // XS-FORM
420
421 SLD_OPCODE = (31u << OPCODE_SHIFT | 27u << 1), // X-FORM
422 SRD_OPCODE = (31u << OPCODE_SHIFT | 539u << 1), // X-FORM
423 SRAD_OPCODE = (31u << OPCODE_SHIFT | 794u << 1), // X-FORM
424
425 MULLD_OPCODE = (31u << OPCODE_SHIFT | 233u << 1), // XO-FORM
426 MULHD_OPCODE = (31u << OPCODE_SHIFT | 73u << 1), // XO-FORM
427 MULHDU_OPCODE = (31u << OPCODE_SHIFT | 9u << 1), // XO-FORM
428 DIVD_OPCODE = (31u << OPCODE_SHIFT | 489u << 1), // XO-FORM
429
430 CNTLZD_OPCODE = (31u << OPCODE_SHIFT | 58u << XO_21_30_SHIFT), // X-FORM
431 NAND_OPCODE = (31u << OPCODE_SHIFT | 476u << XO_21_30_SHIFT), // X-FORM
432 NOR_OPCODE = (31u << OPCODE_SHIFT | 124u << XO_21_30_SHIFT), // X-FORM
433
434
435 // opcodes only used for floating arithmetic
436 FADD_OPCODE = (63u << OPCODE_SHIFT | 21u << 1),
437 FADDS_OPCODE = (59u << OPCODE_SHIFT | 21u << 1),
438 FCMPU_OPCODE = (63u << OPCODE_SHIFT | 00u << 1),
439 FDIV_OPCODE = (63u << OPCODE_SHIFT | 18u << 1),
440 FDIVS_OPCODE = (59u << OPCODE_SHIFT | 18u << 1),
441 FMR_OPCODE = (63u << OPCODE_SHIFT | 72u << 1),
442 // These are special Power6 opcodes, reused for "lfdepx" and "stfdepx"
443 // on Power7. Do not use.
444 // MFFGPR_OPCODE = (31u << OPCODE_SHIFT | 607u << 1),
445 // MFTGPR_OPCODE = (31u << OPCODE_SHIFT | 735u << 1),
446 CMPB_OPCODE = (31u << OPCODE_SHIFT | 508 << 1),
447 POPCNTB_OPCODE = (31u << OPCODE_SHIFT | 122 << 1),
448 POPCNTW_OPCODE = (31u << OPCODE_SHIFT | 378 << 1),
449 POPCNTD_OPCODE = (31u << OPCODE_SHIFT | 506 << 1),
450 FABS_OPCODE = (63u << OPCODE_SHIFT | 264u << 1),
451 FNABS_OPCODE = (63u << OPCODE_SHIFT | 136u << 1),
452 FMUL_OPCODE = (63u << OPCODE_SHIFT | 25u << 1),
453 FMULS_OPCODE = (59u << OPCODE_SHIFT | 25u << 1),
454 FNEG_OPCODE = (63u << OPCODE_SHIFT | 40u << 1),
455 FSUB_OPCODE = (63u << OPCODE_SHIFT | 20u << 1),
456 FSUBS_OPCODE = (59u << OPCODE_SHIFT | 20u << 1),
457
458 // PPC64-internal FPU conversion opcodes
459 FCFID_OPCODE = (63u << OPCODE_SHIFT | 846u << 1),
460 FCFIDS_OPCODE = (59u << OPCODE_SHIFT | 846u << 1),
461 FCTID_OPCODE = (63u << OPCODE_SHIFT | 814u << 1),
462 FCTIDZ_OPCODE = (63u << OPCODE_SHIFT | 815u << 1),
463 FCTIW_OPCODE = (63u << OPCODE_SHIFT | 14u << 1),
464 FCTIWZ_OPCODE = (63u << OPCODE_SHIFT | 15u << 1),
465 FRSP_OPCODE = (63u << OPCODE_SHIFT | 12u << 1),
466
467 // Fused multiply-accumulate instructions.
468 FMADD_OPCODE = (63u << OPCODE_SHIFT | 29u << 1),
469 FMADDS_OPCODE = (59u << OPCODE_SHIFT | 29u << 1),
470 FMSUB_OPCODE = (63u << OPCODE_SHIFT | 28u << 1),
471 FMSUBS_OPCODE = (59u << OPCODE_SHIFT | 28u << 1),
472 FNMADD_OPCODE = (63u << OPCODE_SHIFT | 31u << 1),
473 FNMADDS_OPCODE = (59u << OPCODE_SHIFT | 31u << 1),
474 FNMSUB_OPCODE = (63u << OPCODE_SHIFT | 30u << 1),
475 FNMSUBS_OPCODE = (59u << OPCODE_SHIFT | 30u << 1),
476
477 LFD_OPCODE = (50u << OPCODE_SHIFT | 00u << 1),
478 LFDU_OPCODE = (51u << OPCODE_SHIFT | 00u << 1),
479 LFDX_OPCODE = (31u << OPCODE_SHIFT | 599u << 1),
480 LFS_OPCODE = (48u << OPCODE_SHIFT | 00u << 1),
481 LFSU_OPCODE = (49u << OPCODE_SHIFT | 00u << 1),
482 LFSX_OPCODE = (31u << OPCODE_SHIFT | 535u << 1),
483
484 STFD_OPCODE = (54u << OPCODE_SHIFT | 00u << 1),
485 STFDU_OPCODE = (55u << OPCODE_SHIFT | 00u << 1),
486 STFDX_OPCODE = (31u << OPCODE_SHIFT | 727u << 1),
487 STFS_OPCODE = (52u << OPCODE_SHIFT | 00u << 1),
488 STFSU_OPCODE = (53u << OPCODE_SHIFT | 00u << 1),
489 STFSX_OPCODE = (31u << OPCODE_SHIFT | 663u << 1),
490
491 FSQRT_OPCODE = (63u << OPCODE_SHIFT | 22u << 1), // A-FORM
492 FSQRTS_OPCODE = (59u << OPCODE_SHIFT | 22u << 1), // A-FORM
493
494 // Vector instruction support for >= Power6
495 // Vector Storage Access
496 LVEBX_OPCODE = (31u << OPCODE_SHIFT | 7u << 1),
497 LVEHX_OPCODE = (31u << OPCODE_SHIFT | 39u << 1),
498 LVEWX_OPCODE = (31u << OPCODE_SHIFT | 71u << 1),
499 LVX_OPCODE = (31u << OPCODE_SHIFT | 103u << 1),
500 LVXL_OPCODE = (31u << OPCODE_SHIFT | 359u << 1),
501 STVEBX_OPCODE = (31u << OPCODE_SHIFT | 135u << 1),
502 STVEHX_OPCODE = (31u << OPCODE_SHIFT | 167u << 1),
503 STVEWX_OPCODE = (31u << OPCODE_SHIFT | 199u << 1),
504 STVX_OPCODE = (31u << OPCODE_SHIFT | 231u << 1),
505 STVXL_OPCODE = (31u << OPCODE_SHIFT | 487u << 1),
506 LVSL_OPCODE = (31u << OPCODE_SHIFT | 6u << 1),
507 LVSR_OPCODE = (31u << OPCODE_SHIFT | 38u << 1),
508
509 // Vector-Scalar (VSX) instruction support.
510 LXVD2X_OPCODE = (31u << OPCODE_SHIFT | 844u << 1),
511 STXVD2X_OPCODE = (31u << OPCODE_SHIFT | 972u << 1),
512 MTVSRD_OPCODE = (31u << OPCODE_SHIFT | 179u << 1),
513 MTVSRWZ_OPCODE = (31u << OPCODE_SHIFT | 243u << 1),
514 MFVSRD_OPCODE = (31u << OPCODE_SHIFT | 51u << 1),
515 MTVSRWA_OPCODE = (31u << OPCODE_SHIFT | 211u << 1),
516 MFVSRWZ_OPCODE = (31u << OPCODE_SHIFT | 115u << 1),
517 XXPERMDI_OPCODE= (60u << OPCODE_SHIFT | 10u << 3),
518 XXMRGHW_OPCODE = (60u << OPCODE_SHIFT | 18u << 3),
519 XXMRGLW_OPCODE = (60u << OPCODE_SHIFT | 50u << 3),
520 XXSPLTW_OPCODE = (60u << OPCODE_SHIFT | 164u << 2),
521 XXLXOR_OPCODE = (60u << OPCODE_SHIFT | 154u << 3),
522 XXLEQV_OPCODE = (60u << OPCODE_SHIFT | 186u << 3),
523
524 // Vector Permute and Formatting
525 VPKPX_OPCODE = (4u << OPCODE_SHIFT | 782u ),
526 VPKSHSS_OPCODE = (4u << OPCODE_SHIFT | 398u ),
527 VPKSWSS_OPCODE = (4u << OPCODE_SHIFT | 462u ),
528 VPKSHUS_OPCODE = (4u << OPCODE_SHIFT | 270u ),
529 VPKSWUS_OPCODE = (4u << OPCODE_SHIFT | 334u ),
530 VPKUHUM_OPCODE = (4u << OPCODE_SHIFT | 14u ),
531 VPKUWUM_OPCODE = (4u << OPCODE_SHIFT | 78u ),
532 VPKUHUS_OPCODE = (4u << OPCODE_SHIFT | 142u ),
533 VPKUWUS_OPCODE = (4u << OPCODE_SHIFT | 206u ),
534 VUPKHPX_OPCODE = (4u << OPCODE_SHIFT | 846u ),
535 VUPKHSB_OPCODE = (4u << OPCODE_SHIFT | 526u ),
536 VUPKHSH_OPCODE = (4u << OPCODE_SHIFT | 590u ),
537 VUPKLPX_OPCODE = (4u << OPCODE_SHIFT | 974u ),
538 VUPKLSB_OPCODE = (4u << OPCODE_SHIFT | 654u ),
539 VUPKLSH_OPCODE = (4u << OPCODE_SHIFT | 718u ),
540
541 VMRGHB_OPCODE = (4u << OPCODE_SHIFT | 12u ),
542 VMRGHW_OPCODE = (4u << OPCODE_SHIFT | 140u ),
543 VMRGHH_OPCODE = (4u << OPCODE_SHIFT | 76u ),
544 VMRGLB_OPCODE = (4u << OPCODE_SHIFT | 268u ),
545 VMRGLW_OPCODE = (4u << OPCODE_SHIFT | 396u ),
546 VMRGLH_OPCODE = (4u << OPCODE_SHIFT | 332u ),
547
548 VSPLT_OPCODE = (4u << OPCODE_SHIFT | 524u ),
549 VSPLTH_OPCODE = (4u << OPCODE_SHIFT | 588u ),
550 VSPLTW_OPCODE = (4u << OPCODE_SHIFT | 652u ),
551 VSPLTISB_OPCODE= (4u << OPCODE_SHIFT | 780u ),
552 VSPLTISH_OPCODE= (4u << OPCODE_SHIFT | 844u ),
553 VSPLTISW_OPCODE= (4u << OPCODE_SHIFT | 908u ),
554
555 VPERM_OPCODE = (4u << OPCODE_SHIFT | 43u ),
556 VSEL_OPCODE = (4u << OPCODE_SHIFT | 42u ),
557
558 VSL_OPCODE = (4u << OPCODE_SHIFT | 452u ),
559 VSLDOI_OPCODE = (4u << OPCODE_SHIFT | 44u ),
560 VSLO_OPCODE = (4u << OPCODE_SHIFT | 1036u ),
561 VSR_OPCODE = (4u << OPCODE_SHIFT | 708u ),
562 VSRO_OPCODE = (4u << OPCODE_SHIFT | 1100u ),
563
564 // Vector Integer
565 VADDCUW_OPCODE = (4u << OPCODE_SHIFT | 384u ),
566 VADDSHS_OPCODE = (4u << OPCODE_SHIFT | 832u ),
567 VADDSBS_OPCODE = (4u << OPCODE_SHIFT | 768u ),
568 VADDSWS_OPCODE = (4u << OPCODE_SHIFT | 896u ),
569 VADDUBM_OPCODE = (4u << OPCODE_SHIFT | 0u ),
570 VADDUWM_OPCODE = (4u << OPCODE_SHIFT | 128u ),
571 VADDUHM_OPCODE = (4u << OPCODE_SHIFT | 64u ),
572 VADDUDM_OPCODE = (4u << OPCODE_SHIFT | 192u ),
573 VADDUBS_OPCODE = (4u << OPCODE_SHIFT | 512u ),
574 VADDUWS_OPCODE = (4u << OPCODE_SHIFT | 640u ),
575 VADDUHS_OPCODE = (4u << OPCODE_SHIFT | 576u ),
576 VSUBCUW_OPCODE = (4u << OPCODE_SHIFT | 1408u ),
577 VSUBSHS_OPCODE = (4u << OPCODE_SHIFT | 1856u ),
578 VSUBSBS_OPCODE = (4u << OPCODE_SHIFT | 1792u ),
579 VSUBSWS_OPCODE = (4u << OPCODE_SHIFT | 1920u ),
580 VSUBUBM_OPCODE = (4u << OPCODE_SHIFT | 1024u ),
581 VSUBUWM_OPCODE = (4u << OPCODE_SHIFT | 1152u ),
582 VSUBUHM_OPCODE = (4u << OPCODE_SHIFT | 1088u ),
583 VSUBUBS_OPCODE = (4u << OPCODE_SHIFT | 1536u ),
584 VSUBUWS_OPCODE = (4u << OPCODE_SHIFT | 1664u ),
585 VSUBUHS_OPCODE = (4u << OPCODE_SHIFT | 1600u ),
586
587 VMULESB_OPCODE = (4u << OPCODE_SHIFT | 776u ),
588 VMULEUB_OPCODE = (4u << OPCODE_SHIFT | 520u ),
589 VMULESH_OPCODE = (4u << OPCODE_SHIFT | 840u ),
590 VMULEUH_OPCODE = (4u << OPCODE_SHIFT | 584u ),
591 VMULOSB_OPCODE = (4u << OPCODE_SHIFT | 264u ),
592 VMULOUB_OPCODE = (4u << OPCODE_SHIFT | 8u ),
593 VMULOSH_OPCODE = (4u << OPCODE_SHIFT | 328u ),
594 VMULOUH_OPCODE = (4u << OPCODE_SHIFT | 72u ),
595 VMHADDSHS_OPCODE=(4u << OPCODE_SHIFT | 32u ),
596 VMHRADDSHS_OPCODE=(4u << OPCODE_SHIFT | 33u ),
597 VMLADDUHM_OPCODE=(4u << OPCODE_SHIFT | 34u ),
598 VMSUBUHM_OPCODE= (4u << OPCODE_SHIFT | 36u ),
599 VMSUMMBM_OPCODE= (4u << OPCODE_SHIFT | 37u ),
600 VMSUMSHM_OPCODE= (4u << OPCODE_SHIFT | 40u ),
601 VMSUMSHS_OPCODE= (4u << OPCODE_SHIFT | 41u ),
602 VMSUMUHM_OPCODE= (4u << OPCODE_SHIFT | 38u ),
603 VMSUMUHS_OPCODE= (4u << OPCODE_SHIFT | 39u ),
604
605 VSUMSWS_OPCODE = (4u << OPCODE_SHIFT | 1928u ),
606 VSUM2SWS_OPCODE= (4u << OPCODE_SHIFT | 1672u ),
607 VSUM4SBS_OPCODE= (4u << OPCODE_SHIFT | 1800u ),
608 VSUM4UBS_OPCODE= (4u << OPCODE_SHIFT | 1544u ),
609 VSUM4SHS_OPCODE= (4u << OPCODE_SHIFT | 1608u ),
610
611 VAVGSB_OPCODE = (4u << OPCODE_SHIFT | 1282u ),
612 VAVGSW_OPCODE = (4u << OPCODE_SHIFT | 1410u ),
613 VAVGSH_OPCODE = (4u << OPCODE_SHIFT | 1346u ),
614 VAVGUB_OPCODE = (4u << OPCODE_SHIFT | 1026u ),
615 VAVGUW_OPCODE = (4u << OPCODE_SHIFT | 1154u ),
616 VAVGUH_OPCODE = (4u << OPCODE_SHIFT | 1090u ),
617
618 VMAXSB_OPCODE = (4u << OPCODE_SHIFT | 258u ),
619 VMAXSW_OPCODE = (4u << OPCODE_SHIFT | 386u ),
620 VMAXSH_OPCODE = (4u << OPCODE_SHIFT | 322u ),
621 VMAXUB_OPCODE = (4u << OPCODE_SHIFT | 2u ),
622 VMAXUW_OPCODE = (4u << OPCODE_SHIFT | 130u ),
623 VMAXUH_OPCODE = (4u << OPCODE_SHIFT | 66u ),
624 VMINSB_OPCODE = (4u << OPCODE_SHIFT | 770u ),
625 VMINSW_OPCODE = (4u << OPCODE_SHIFT | 898u ),
626 VMINSH_OPCODE = (4u << OPCODE_SHIFT | 834u ),
627 VMINUB_OPCODE = (4u << OPCODE_SHIFT | 514u ),
628 VMINUW_OPCODE = (4u << OPCODE_SHIFT | 642u ),
629 VMINUH_OPCODE = (4u << OPCODE_SHIFT | 578u ),
630
631 VCMPEQUB_OPCODE= (4u << OPCODE_SHIFT | 6u ),
632 VCMPEQUH_OPCODE= (4u << OPCODE_SHIFT | 70u ),
633 VCMPEQUW_OPCODE= (4u << OPCODE_SHIFT | 134u ),
634 VCMPGTSH_OPCODE= (4u << OPCODE_SHIFT | 838u ),
635 VCMPGTSB_OPCODE= (4u << OPCODE_SHIFT | 774u ),
636 VCMPGTSW_OPCODE= (4u << OPCODE_SHIFT | 902u ),
637 VCMPGTUB_OPCODE= (4u << OPCODE_SHIFT | 518u ),
638 VCMPGTUH_OPCODE= (4u << OPCODE_SHIFT | 582u ),
639 VCMPGTUW_OPCODE= (4u << OPCODE_SHIFT | 646u ),
640
641 VAND_OPCODE = (4u << OPCODE_SHIFT | 1028u ),
642 VANDC_OPCODE = (4u << OPCODE_SHIFT | 1092u ),
643 VNOR_OPCODE = (4u << OPCODE_SHIFT | 1284u ),
644 VOR_OPCODE = (4u << OPCODE_SHIFT | 1156u ),
645 VXOR_OPCODE = (4u << OPCODE_SHIFT | 1220u ),
646 VRLD_OPCODE = (4u << OPCODE_SHIFT | 196u ),
647 VRLB_OPCODE = (4u << OPCODE_SHIFT | 4u ),
648 VRLW_OPCODE = (4u << OPCODE_SHIFT | 132u ),
649 VRLH_OPCODE = (4u << OPCODE_SHIFT | 68u ),
650 VSLB_OPCODE = (4u << OPCODE_SHIFT | 260u ),
651 VSKW_OPCODE = (4u << OPCODE_SHIFT | 388u ),
652 VSLH_OPCODE = (4u << OPCODE_SHIFT | 324u ),
653 VSRB_OPCODE = (4u << OPCODE_SHIFT | 516u ),
654 VSRW_OPCODE = (4u << OPCODE_SHIFT | 644u ),
655 VSRH_OPCODE = (4u << OPCODE_SHIFT | 580u ),
656 VSRAB_OPCODE = (4u << OPCODE_SHIFT | 772u ),
657 VSRAW_OPCODE = (4u << OPCODE_SHIFT | 900u ),
658 VSRAH_OPCODE = (4u << OPCODE_SHIFT | 836u ),
659
660 // Vector Floating-Point
661 // not implemented yet
662
663 // Vector Status and Control
664 MTVSCR_OPCODE = (4u << OPCODE_SHIFT | 1604u ),
665 MFVSCR_OPCODE = (4u << OPCODE_SHIFT | 1540u ),
666
667 // AES (introduced with Power 8)
668 VCIPHER_OPCODE = (4u << OPCODE_SHIFT | 1288u),
669 VCIPHERLAST_OPCODE = (4u << OPCODE_SHIFT | 1289u),
670 VNCIPHER_OPCODE = (4u << OPCODE_SHIFT | 1352u),
671 VNCIPHERLAST_OPCODE = (4u << OPCODE_SHIFT | 1353u),
672 VSBOX_OPCODE = (4u << OPCODE_SHIFT | 1480u),
673
674 // SHA (introduced with Power 8)
675 VSHASIGMAD_OPCODE = (4u << OPCODE_SHIFT | 1730u),
676 VSHASIGMAW_OPCODE = (4u << OPCODE_SHIFT | 1666u),
677
678 // Vector Binary Polynomial Multiplication (introduced with Power 8)
679 VPMSUMB_OPCODE = (4u << OPCODE_SHIFT | 1032u),
680 VPMSUMD_OPCODE = (4u << OPCODE_SHIFT | 1224u),
681 VPMSUMH_OPCODE = (4u << OPCODE_SHIFT | 1096u),
682 VPMSUMW_OPCODE = (4u << OPCODE_SHIFT | 1160u),
683
684 // Vector Permute and Xor (introduced with Power 8)
685 VPERMXOR_OPCODE = (4u << OPCODE_SHIFT | 45u),
686
687 // Transactional Memory instructions (introduced with Power 8)
688 TBEGIN_OPCODE = (31u << OPCODE_SHIFT | 654u << 1),
689 TEND_OPCODE = (31u << OPCODE_SHIFT | 686u << 1),
690 TABORT_OPCODE = (31u << OPCODE_SHIFT | 910u << 1),
691 TABORTWC_OPCODE = (31u << OPCODE_SHIFT | 782u << 1),
692 TABORTWCI_OPCODE = (31u << OPCODE_SHIFT | 846u << 1),
693 TABORTDC_OPCODE = (31u << OPCODE_SHIFT | 814u << 1),
694 TABORTDCI_OPCODE = (31u << OPCODE_SHIFT | 878u << 1),
695 TSR_OPCODE = (31u << OPCODE_SHIFT | 750u << 1),
696 TCHECK_OPCODE = (31u << OPCODE_SHIFT | 718u << 1),
697
698 // Icache and dcache related instructions
699 DCBA_OPCODE = (31u << OPCODE_SHIFT | 758u << 1),
700 DCBZ_OPCODE = (31u << OPCODE_SHIFT | 1014u << 1),
701 DCBST_OPCODE = (31u << OPCODE_SHIFT | 54u << 1),
702 DCBF_OPCODE = (31u << OPCODE_SHIFT | 86u << 1),
703
704 DCBT_OPCODE = (31u << OPCODE_SHIFT | 278u << 1),
705 DCBTST_OPCODE = (31u << OPCODE_SHIFT | 246u << 1),
706 ICBI_OPCODE = (31u << OPCODE_SHIFT | 982u << 1),
707
708 // Instruction synchronization
709 ISYNC_OPCODE = (19u << OPCODE_SHIFT | 150u << 1),
710 // Memory barriers
711 SYNC_OPCODE = (31u << OPCODE_SHIFT | 598u << 1),
712 EIEIO_OPCODE = (31u << OPCODE_SHIFT | 854u << 1),
713
714 // Wait instructions for polling.
715 WAIT_OPCODE = (31u << OPCODE_SHIFT | 62u << 1),
716
717 // Trap instructions
718 TDI_OPCODE = (2u << OPCODE_SHIFT),
719 TWI_OPCODE = (3u << OPCODE_SHIFT),
720 TD_OPCODE = (31u << OPCODE_SHIFT | 68u << 1),
721 TW_OPCODE = (31u << OPCODE_SHIFT | 4u << 1),
722
723 // Atomics.
724 LBARX_OPCODE = (31u << OPCODE_SHIFT | 52u << 1),
725 LHARX_OPCODE = (31u << OPCODE_SHIFT | 116u << 1),
726 LWARX_OPCODE = (31u << OPCODE_SHIFT | 20u << 1),
727 LDARX_OPCODE = (31u << OPCODE_SHIFT | 84u << 1),
728 LQARX_OPCODE = (31u << OPCODE_SHIFT | 276u << 1),
729 STBCX_OPCODE = (31u << OPCODE_SHIFT | 694u << 1),
730 STHCX_OPCODE = (31u << OPCODE_SHIFT | 726u << 1),
731 STWCX_OPCODE = (31u << OPCODE_SHIFT | 150u << 1),
732 STDCX_OPCODE = (31u << OPCODE_SHIFT | 214u << 1),
733 STQCX_OPCODE = (31u << OPCODE_SHIFT | 182u << 1)
734
735 };
736
737 // Trap instructions TO bits
738 enum trap_to_bits {
739 // single bits
740 traptoLessThanSigned = 1 << 4, // 0, left end
741 traptoGreaterThanSigned = 1 << 3,
742 traptoEqual = 1 << 2,
743 traptoLessThanUnsigned = 1 << 1,
744 traptoGreaterThanUnsigned = 1 << 0, // 4, right end
745
746 // compound ones
747 traptoUnconditional = (traptoLessThanSigned |
748 traptoGreaterThanSigned |
749 traptoEqual |
750 traptoLessThanUnsigned |
751 traptoGreaterThanUnsigned)
752 };
753
754 // Branch hints BH field
755 enum branch_hint_bh {
756 // bclr cases:
757 bhintbhBCLRisReturn = 0,
758 bhintbhBCLRisNotReturnButSame = 1,
759 bhintbhBCLRisNotPredictable = 3,
760
761 // bcctr cases:
762 bhintbhBCCTRisNotReturnButSame = 0,
763 bhintbhBCCTRisNotPredictable = 3
764 };
765
766 // Branch prediction hints AT field
767 enum branch_hint_at {
768 bhintatNoHint = 0, // at=00
769 bhintatIsNotTaken = 2, // at=10
770 bhintatIsTaken = 3 // at=11
771 };
772
773 // Branch prediction hints
774 enum branch_hint_concept {
775 // Use the same encoding as branch_hint_at to simply code.
776 bhintNoHint = bhintatNoHint,
777 bhintIsNotTaken = bhintatIsNotTaken,
778 bhintIsTaken = bhintatIsTaken
779 };
780
781 // Used in BO field of branch instruction.
782 enum branch_condition {
783 bcondCRbiIs0 = 4, // bo=001at
784 bcondCRbiIs1 = 12, // bo=011at
785 bcondAlways = 20 // bo=10100
786 };
787
788 // Branch condition with combined prediction hints.
789 enum branch_condition_with_hint {
790 bcondCRbiIs0_bhintNoHint = bcondCRbiIs0 | bhintatNoHint,
791 bcondCRbiIs0_bhintIsNotTaken = bcondCRbiIs0 | bhintatIsNotTaken,
792 bcondCRbiIs0_bhintIsTaken = bcondCRbiIs0 | bhintatIsTaken,
793 bcondCRbiIs1_bhintNoHint = bcondCRbiIs1 | bhintatNoHint,
794 bcondCRbiIs1_bhintIsNotTaken = bcondCRbiIs1 | bhintatIsNotTaken,
795 bcondCRbiIs1_bhintIsTaken = bcondCRbiIs1 | bhintatIsTaken,
796 };
797
798 // Elemental Memory Barriers (>=Power 8)
799 enum Elemental_Membar_mask_bits {
800 StoreStore = 1 << 0,
801 StoreLoad = 1 << 1,
802 LoadStore = 1 << 2,
803 LoadLoad = 1 << 3
804 };
805
806 // Branch prediction hints.
807 inline static int add_bhint_to_boint(const int bhint, const int boint) {
808 switch (boint) {
809 case bcondCRbiIs0:
810 case bcondCRbiIs1:
811 // branch_hint and branch_hint_at have same encodings
812 assert( (int)bhintNoHint == (int)bhintatNoHint
813 && (int)bhintIsNotTaken == (int)bhintatIsNotTaken
814 && (int)bhintIsTaken == (int)bhintatIsTaken,
815 "wrong encodings");
816 assert((bhint & 0x03) == bhint, "wrong encodings");
817 return (boint & ~0x03) | bhint;
818 case bcondAlways:
819 // no branch_hint
820 return boint;
821 default:
822 ShouldNotReachHere();
823 return 0;
824 }
825 }
826
827 // Extract bcond from boint.
828 inline static int inv_boint_bcond(const int boint) {
829 int r_bcond = boint & ~0x03;
830 assert(r_bcond == bcondCRbiIs0 ||
831 r_bcond == bcondCRbiIs1 ||
832 r_bcond == bcondAlways,
833 "bad branch condition");
834 return r_bcond;
835 }
836
837 // Extract bhint from boint.
838 inline static int inv_boint_bhint(const int boint) {
839 int r_bhint = boint & 0x03;
840 assert(r_bhint == bhintatNoHint ||
841 r_bhint == bhintatIsNotTaken ||
842 r_bhint == bhintatIsTaken,
843 "bad branch hint");
844 return r_bhint;
845 }
846
847 // Calculate opposite of given bcond.
848 inline static int opposite_bcond(const int bcond) {
849 switch (bcond) {
850 case bcondCRbiIs0:
851 return bcondCRbiIs1;
852 case bcondCRbiIs1:
853 return bcondCRbiIs0;
854 default:
855 ShouldNotReachHere();
856 return 0;
857 }
858 }
859
860 // Calculate opposite of given bhint.
861 inline static int opposite_bhint(const int bhint) {
862 switch (bhint) {
863 case bhintatNoHint:
864 return bhintatNoHint;
865 case bhintatIsNotTaken:
866 return bhintatIsTaken;
867 case bhintatIsTaken:
868 return bhintatIsNotTaken;
869 default:
870 ShouldNotReachHere();
871 return 0;
872 }
873 }
874
875 // PPC branch instructions
876 enum ppcops {
877 b_op = 18,
878 bc_op = 16,
879 bcr_op = 19
880 };
881
882 enum Condition {
883 negative = 0,
884 less = 0,
885 positive = 1,
886 greater = 1,
887 zero = 2,
888 equal = 2,
889 summary_overflow = 3,
890 };
891
892 public:
893 // Helper functions for groups of instructions
894
895 enum Predict { pt = 1, pn = 0 }; // pt = predict taken
896
897 // Instruction must start at passed address.
898 static int instr_len(unsigned char *instr) { return BytesPerInstWord; }
899
900 // longest instructions
901 static int instr_maxlen() { return BytesPerInstWord; }
902
903 // Test if x is within signed immediate range for nbits.
904 static bool is_simm(int x, unsigned int nbits) {
905 assert(0 < nbits && nbits < 32, "out of bounds");
906 const int min = -(((int)1) << nbits-1);
907 const int maxplus1 = (((int)1) << nbits-1);
908 return min <= x && x < maxplus1;
909 }
910
911 static bool is_simm(jlong x, unsigned int nbits) {
912 assert(0 < nbits && nbits < 64, "out of bounds");
913 const jlong min = -(((jlong)1) << nbits-1);
914 const jlong maxplus1 = (((jlong)1) << nbits-1);
915 return min <= x && x < maxplus1;
916 }
917
918 // Test if x is within unsigned immediate range for nbits.
919 static bool is_uimm(int x, unsigned int nbits) {
920 assert(0 < nbits && nbits < 32, "out of bounds");
921 const unsigned int maxplus1 = (((unsigned int)1) << nbits);
922 return (unsigned int)x < maxplus1;
923 }
924
925 static bool is_uimm(jlong x, unsigned int nbits) {
926 assert(0 < nbits && nbits < 64, "out of bounds");
927 const julong maxplus1 = (((julong)1) << nbits);
928 return (julong)x < maxplus1;
929 }
930
931 protected:
932 // helpers
933
934 // X is supposed to fit in a field "nbits" wide
935 // and be sign-extended. Check the range.
936 static void assert_signed_range(intptr_t x, int nbits) {
937 assert(nbits == 32 || (-(1 << nbits-1) <= x && x < (1 << nbits-1)),
938 "value out of range");
939 }
940
941 static void assert_signed_word_disp_range(intptr_t x, int nbits) {
942 assert((x & 3) == 0, "not word aligned");
943 assert_signed_range(x, nbits + 2);
944 }
945
946 static void assert_unsigned_const(int x, int nbits) {
947 assert(juint(x) < juint(1 << nbits), "unsigned constant out of range");
948 }
949
950 static int fmask(juint hi_bit, juint lo_bit) {
951 assert(hi_bit >= lo_bit && hi_bit < 32, "bad bits");
952 return (1 << ( hi_bit-lo_bit + 1 )) - 1;
953 }
954
955 // inverse of u_field
956 static int inv_u_field(int x, int hi_bit, int lo_bit) {
957 juint r = juint(x) >> lo_bit;
958 r &= fmask(hi_bit, lo_bit);
959 return int(r);
960 }
961
962 // signed version: extract from field and sign-extend
963 static int inv_s_field_ppc(int x, int hi_bit, int lo_bit) {
964 x = x << (31-hi_bit);
965 x = x >> (31-hi_bit+lo_bit);
966 return x;
967 }
968
969 static int u_field(int x, int hi_bit, int lo_bit) {
970 assert((x & ~fmask(hi_bit, lo_bit)) == 0, "value out of range");
971 int r = x << lo_bit;
972 assert(inv_u_field(r, hi_bit, lo_bit) == x, "just checking");
973 return r;
974 }
975
976 // Same as u_field for signed values
977 static int s_field(int x, int hi_bit, int lo_bit) {
978 int nbits = hi_bit - lo_bit + 1;
979 assert(nbits == 32 || (-(1 << nbits-1) <= x && x < (1 << nbits-1)),
980 "value out of range");
981 x &= fmask(hi_bit, lo_bit);
982 int r = x << lo_bit;
983 return r;
984 }
985
986 // inv_op for ppc instructions
987 static int inv_op_ppc(int x) { return inv_u_field(x, 31, 26); }
988
989 // Determine target address from li, bd field of branch instruction.
990 static intptr_t inv_li_field(int x) {
991 intptr_t r = inv_s_field_ppc(x, 25, 2);
992 r = (r << 2);
993 return r;
994 }
995 static intptr_t inv_bd_field(int x, intptr_t pos) {
996 intptr_t r = inv_s_field_ppc(x, 15, 2);
997 r = (r << 2) + pos;
998 return r;
999 }
1000
1001 #define inv_opp_u_field(x, hi_bit, lo_bit) inv_u_field(x, 31-(lo_bit), 31-(hi_bit))
1002 #define inv_opp_s_field(x, hi_bit, lo_bit) inv_s_field_ppc(x, 31-(lo_bit), 31-(hi_bit))
1003 // Extract instruction fields from instruction words.
1004 public:
1005 static int inv_ra_field(int x) { return inv_opp_u_field(x, 15, 11); }
1006 static int inv_rb_field(int x) { return inv_opp_u_field(x, 20, 16); }
1007 static int inv_rt_field(int x) { return inv_opp_u_field(x, 10, 6); }
1008 static int inv_rta_field(int x) { return inv_opp_u_field(x, 15, 11); }
1009 static int inv_rs_field(int x) { return inv_opp_u_field(x, 10, 6); }
1010 // Ds uses opp_s_field(x, 31, 16), but lowest 2 bits must be 0.
1011 // Inv_ds_field uses range (x, 29, 16) but shifts by 2 to ensure that lowest bits are 0.
1012 static int inv_ds_field(int x) { return inv_opp_s_field(x, 29, 16) << 2; }
1013 static int inv_d1_field(int x) { return inv_opp_s_field(x, 31, 16); }
1014 static int inv_si_field(int x) { return inv_opp_s_field(x, 31, 16); }
1015 static int inv_to_field(int x) { return inv_opp_u_field(x, 10, 6); }
1016 static int inv_lk_field(int x) { return inv_opp_u_field(x, 31, 31); }
1017 static int inv_bo_field(int x) { return inv_opp_u_field(x, 10, 6); }
1018 static int inv_bi_field(int x) { return inv_opp_u_field(x, 15, 11); }
1019
1020 #define opp_u_field(x, hi_bit, lo_bit) u_field(x, 31-(lo_bit), 31-(hi_bit))
1021 #define opp_s_field(x, hi_bit, lo_bit) s_field(x, 31-(lo_bit), 31-(hi_bit))
1022
1023 // instruction fields
1024 static int aa( int x) { return opp_u_field(x, 30, 30); }
1025 static int ba( int x) { return opp_u_field(x, 15, 11); }
1026 static int bb( int x) { return opp_u_field(x, 20, 16); }
1027 static int bc( int x) { return opp_u_field(x, 25, 21); }
1028 static int bd( int x) { return opp_s_field(x, 29, 16); }
1029 static int bf( ConditionRegister cr) { return bf(cr->encoding()); }
1030 static int bf( int x) { return opp_u_field(x, 8, 6); }
1031 static int bfa(ConditionRegister cr) { return bfa(cr->encoding()); }
1032 static int bfa( int x) { return opp_u_field(x, 13, 11); }
1033 static int bh( int x) { return opp_u_field(x, 20, 19); }
1034 static int bi( int x) { return opp_u_field(x, 15, 11); }
1035 static int bi0(ConditionRegister cr, Condition c) { return (cr->encoding() << 2) | c; }
1036 static int bo( int x) { return opp_u_field(x, 10, 6); }
1037 static int bt( int x) { return opp_u_field(x, 10, 6); }
1038 static int d1( int x) { return opp_s_field(x, 31, 16); }
1039 static int ds( int x) { assert((x & 0x3) == 0, "unaligned offset"); return opp_s_field(x, 31, 16); }
1040 static int eh( int x) { return opp_u_field(x, 31, 31); }
1041 static int flm( int x) { return opp_u_field(x, 14, 7); }
1042 static int fra( FloatRegister r) { return fra(r->encoding());}
1043 static int frb( FloatRegister r) { return frb(r->encoding());}
1044 static int frc( FloatRegister r) { return frc(r->encoding());}
1045 static int frs( FloatRegister r) { return frs(r->encoding());}
1046 static int frt( FloatRegister r) { return frt(r->encoding());}
1047 static int fra( int x) { return opp_u_field(x, 15, 11); }
1048 static int frb( int x) { return opp_u_field(x, 20, 16); }
1049 static int frc( int x) { return opp_u_field(x, 25, 21); }
1050 static int frs( int x) { return opp_u_field(x, 10, 6); }
1051 static int frt( int x) { return opp_u_field(x, 10, 6); }
1052 static int fxm( int x) { return opp_u_field(x, 19, 12); }
1053 static int l10( int x) { return opp_u_field(x, 10, 10); }
1054 static int l15( int x) { return opp_u_field(x, 15, 15); }
1055 static int l910( int x) { return opp_u_field(x, 10, 9); }
1056 static int e1215( int x) { return opp_u_field(x, 15, 12); }
1057 static int lev( int x) { return opp_u_field(x, 26, 20); }
1058 static int li( int x) { return opp_s_field(x, 29, 6); }
1059 static int lk( int x) { return opp_u_field(x, 31, 31); }
1060 static int mb2125( int x) { return opp_u_field(x, 25, 21); }
1061 static int me2630( int x) { return opp_u_field(x, 30, 26); }
1062 static int mb2126( int x) { return opp_u_field(((x & 0x1f) << 1) | ((x & 0x20) >> 5), 26, 21); }
1063 static int me2126( int x) { return mb2126(x); }
1064 static int nb( int x) { return opp_u_field(x, 20, 16); }
1065 //static int opcd( int x) { return opp_u_field(x, 5, 0); } // is contained in our opcodes
1066 static int oe( int x) { return opp_u_field(x, 21, 21); }
1067 static int ra( Register r) { return ra(r->encoding()); }
1068 static int ra( int x) { return opp_u_field(x, 15, 11); }
1069 static int rb( Register r) { return rb(r->encoding()); }
1070 static int rb( int x) { return opp_u_field(x, 20, 16); }
1071 static int rc( int x) { return opp_u_field(x, 31, 31); }
1072 static int rs( Register r) { return rs(r->encoding()); }
1073 static int rs( int x) { return opp_u_field(x, 10, 6); }
1074 // we don't want to use R0 in memory accesses, because it has value `0' then
1075 static int ra0mem( Register r) { assert(r != R0, "cannot use register R0 in memory access"); return ra(r); }
1076 static int ra0mem( int x) { assert(x != 0, "cannot use register 0 in memory access"); return ra(x); }
1077
1078 // register r is target
1079 static int rt( Register r) { return rs(r); }
1080 static int rt( int x) { return rs(x); }
1081 static int rta( Register r) { return ra(r); }
1082 static int rta0mem( Register r) { rta(r); return ra0mem(r); }
1083
1084 static int sh1620( int x) { return opp_u_field(x, 20, 16); }
1085 static int sh30( int x) { return opp_u_field(x, 30, 30); }
1086 static int sh162030( int x) { return sh1620(x & 0x1f) | sh30((x & 0x20) >> 5); }
1087 static int si( int x) { return opp_s_field(x, 31, 16); }
1088 static int spr( int x) { return opp_u_field(x, 20, 11); }
1089 static int sr( int x) { return opp_u_field(x, 15, 12); }
1090 static int tbr( int x) { return opp_u_field(x, 20, 11); }
1091 static int th( int x) { return opp_u_field(x, 10, 7); }
1092 static int thct( int x) { assert((x&8) == 0, "must be valid cache specification"); return th(x); }
1093 static int thds( int x) { assert((x&8) == 8, "must be valid stream specification"); return th(x); }
1094 static int to( int x) { return opp_u_field(x, 10, 6); }
1095 static int u( int x) { return opp_u_field(x, 19, 16); }
1096 static int ui( int x) { return opp_u_field(x, 31, 16); }
1097
1098 // Support vector instructions for >= Power6.
1099 static int vra( int x) { return opp_u_field(x, 15, 11); }
1100 static int vrb( int x) { return opp_u_field(x, 20, 16); }
1101 static int vrc( int x) { return opp_u_field(x, 25, 21); }
1102 static int vrs( int x) { return opp_u_field(x, 10, 6); }
1103 static int vrt( int x) { return opp_u_field(x, 10, 6); }
1104
1105 static int vra( VectorRegister r) { return vra(r->encoding());}
1106 static int vrb( VectorRegister r) { return vrb(r->encoding());}
1107 static int vrc( VectorRegister r) { return vrc(r->encoding());}
1108 static int vrs( VectorRegister r) { return vrs(r->encoding());}
1109 static int vrt( VectorRegister r) { return vrt(r->encoding());}
1110
1111 // Only used on SHA sigma instructions (VX-form)
1112 static int vst( int x) { return opp_u_field(x, 16, 16); }
1113 static int vsix( int x) { return opp_u_field(x, 20, 17); }
1114
1115 // Support Vector-Scalar (VSX) instructions.
1116 static int vsra( int x) { return opp_u_field(x & 0x1F, 15, 11) | opp_u_field((x & 0x20) >> 5, 29, 29); }
1117 static int vsrb( int x) { return opp_u_field(x & 0x1F, 20, 16) | opp_u_field((x & 0x20) >> 5, 30, 30); }
1118 static int vsrs( int x) { return opp_u_field(x & 0x1F, 10, 6) | opp_u_field((x & 0x20) >> 5, 31, 31); }
1119 static int vsrt( int x) { return vsrs(x); }
1120 static int vsdm( int x) { return opp_u_field(x, 23, 22); }
1121
1122 static int vsra( VectorSRegister r) { return vsra(r->encoding());}
1123 static int vsrb( VectorSRegister r) { return vsrb(r->encoding());}
1124 static int vsrs( VectorSRegister r) { return vsrs(r->encoding());}
1125 static int vsrt( VectorSRegister r) { return vsrt(r->encoding());}
1126
1127 static int vsplt_uim( int x) { return opp_u_field(x, 15, 12); } // for vsplt* instructions
1128 static int vsplti_sim(int x) { return opp_u_field(x, 15, 11); } // for vsplti* instructions
1129 static int vsldoi_shb(int x) { return opp_u_field(x, 25, 22); } // for vsldoi instruction
1130 static int vcmp_rc( int x) { return opp_u_field(x, 21, 21); } // for vcmp* instructions
1131 static int xxsplt_uim(int x) { return opp_u_field(x, 15, 14); } // for xxsplt* instructions
1132
1133 //static int xo1( int x) { return opp_u_field(x, 29, 21); }// is contained in our opcodes
1134 //static int xo2( int x) { return opp_u_field(x, 30, 21); }// is contained in our opcodes
1135 //static int xo3( int x) { return opp_u_field(x, 30, 22); }// is contained in our opcodes
1136 //static int xo4( int x) { return opp_u_field(x, 30, 26); }// is contained in our opcodes
1137 //static int xo5( int x) { return opp_u_field(x, 29, 27); }// is contained in our opcodes
1138 //static int xo6( int x) { return opp_u_field(x, 30, 27); }// is contained in our opcodes
1139 //static int xo7( int x) { return opp_u_field(x, 31, 30); }// is contained in our opcodes
1140
1141 protected:
1142 // Compute relative address for branch.
1143 static intptr_t disp(intptr_t x, intptr_t off) {
1144 int xx = x - off;
1145 xx = xx >> 2;
1146 return xx;
1147 }
1148
1149 public:
1150 // signed immediate, in low bits, nbits long
1151 static int simm(int x, int nbits) {
1152 assert_signed_range(x, nbits);
1153 return x & ((1 << nbits) - 1);
1154 }
1155
1156 // unsigned immediate, in low bits, nbits long
1157 static int uimm(int x, int nbits) {
1158 assert_unsigned_const(x, nbits);
1159 return x & ((1 << nbits) - 1);
1160 }
1161
1162 static void set_imm(int* instr, short s) {
1163 // imm is always in the lower 16 bits of the instruction,
1164 // so this is endian-neutral. Same for the get_imm below.
1165 uint32_t w = *(uint32_t *)instr;
1166 *instr = (int)((w & ~0x0000FFFF) | (s & 0x0000FFFF));
1167 }
1168
1169 static int get_imm(address a, int instruction_number) {
1170 return (short)((int *)a)[instruction_number];
1171 }
1172
1173 static inline int hi16_signed( int x) { return (int)(int16_t)(x >> 16); }
1174 static inline int lo16_unsigned(int x) { return x & 0xffff; }
1175
1176 protected:
1177
1178 // Extract the top 32 bits in a 64 bit word.
1179 static int32_t hi32(int64_t x) {
1180 int32_t r = int32_t((uint64_t)x >> 32);
1181 return r;
1182 }
1183
1184 public:
1185
1186 static inline unsigned int align_addr(unsigned int addr, unsigned int a) {
1187 return ((addr + (a - 1)) & ~(a - 1));
1188 }
1189
1190 static inline bool is_aligned(unsigned int addr, unsigned int a) {
1191 return (0 == addr % a);
1192 }
1193
1194 void flush() {
1195 AbstractAssembler::flush();
1196 }
1197
1198 inline void emit_int32(int); // shadows AbstractAssembler::emit_int32
1199 inline void emit_data(int);
1200 inline void emit_data(int, RelocationHolder const&);
1201 inline void emit_data(int, relocInfo::relocType rtype);
1202
1203 // Emit an address.
1204 inline address emit_addr(const address addr = NULL);
1205
1206 #if !defined(ABI_ELFv2)
1207 // Emit a function descriptor with the specified entry point, TOC,
1208 // and ENV. If the entry point is NULL, the descriptor will point
1209 // just past the descriptor.
1210 // Use values from friend functions as defaults.
1211 inline address emit_fd(address entry = NULL,
1212 address toc = (address) FunctionDescriptor::friend_toc,
1213 address env = (address) FunctionDescriptor::friend_env);
1214 #endif
1215
1216 /////////////////////////////////////////////////////////////////////////////////////
1217 // PPC instructions
1218 /////////////////////////////////////////////////////////////////////////////////////
1219
1220 // Memory instructions use r0 as hard coded 0, e.g. to simulate loading
1221 // immediates. The normal instruction encoders enforce that r0 is not
1222 // passed to them. Use either extended mnemonics encoders or the special ra0
1223 // versions.
1224
1225 // Issue an illegal instruction.
1226 inline void illtrap();
1227 static inline bool is_illtrap(int x);
1228
1229 // PPC 1, section 3.3.8, Fixed-Point Arithmetic Instructions
1230 inline void addi( Register d, Register a, int si16);
1231 inline void addis(Register d, Register a, int si16);
1232 private:
1233 inline void addi_r0ok( Register d, Register a, int si16);
1234 inline void addis_r0ok(Register d, Register a, int si16);
1235 public:
1236 inline void addic_( Register d, Register a, int si16);
1237 inline void subfic( Register d, Register a, int si16);
1238 inline void add( Register d, Register a, Register b);
1239 inline void add_( Register d, Register a, Register b);
1240 inline void subf( Register d, Register a, Register b); // d = b - a "Sub_from", as in ppc spec.
1241 inline void sub( Register d, Register a, Register b); // d = a - b Swap operands of subf for readability.
1242 inline void subf_( Register d, Register a, Register b);
1243 inline void addc( Register d, Register a, Register b);
1244 inline void addc_( Register d, Register a, Register b);
1245 inline void subfc( Register d, Register a, Register b);
1246 inline void subfc_( Register d, Register a, Register b);
1247 inline void adde( Register d, Register a, Register b);
1248 inline void adde_( Register d, Register a, Register b);
1249 inline void subfe( Register d, Register a, Register b);
1250 inline void subfe_( Register d, Register a, Register b);
1251 inline void addme( Register d, Register a);
1252 inline void addme_( Register d, Register a);
1253 inline void subfme( Register d, Register a);
1254 inline void subfme_(Register d, Register a);
1255 inline void addze( Register d, Register a);
1256 inline void addze_( Register d, Register a);
1257 inline void subfze( Register d, Register a);
1258 inline void subfze_(Register d, Register a);
1259 inline void neg( Register d, Register a);
1260 inline void neg_( Register d, Register a);
1261 inline void mulli( Register d, Register a, int si16);
1262 inline void mulld( Register d, Register a, Register b);
1263 inline void mulld_( Register d, Register a, Register b);
1264 inline void mullw( Register d, Register a, Register b);
1265 inline void mullw_( Register d, Register a, Register b);
1266 inline void mulhw( Register d, Register a, Register b);
1267 inline void mulhw_( Register d, Register a, Register b);
1268 inline void mulhwu( Register d, Register a, Register b);
1269 inline void mulhwu_(Register d, Register a, Register b);
1270 inline void mulhd( Register d, Register a, Register b);
1271 inline void mulhd_( Register d, Register a, Register b);
1272 inline void mulhdu( Register d, Register a, Register b);
1273 inline void mulhdu_(Register d, Register a, Register b);
1274 inline void divd( Register d, Register a, Register b);
1275 inline void divd_( Register d, Register a, Register b);
1276 inline void divw( Register d, Register a, Register b);
1277 inline void divw_( Register d, Register a, Register b);
1278
1279 // Fixed-Point Arithmetic Instructions with Overflow detection
1280 inline void addo( Register d, Register a, Register b);
1281 inline void addo_( Register d, Register a, Register b);
1282 inline void subfo( Register d, Register a, Register b);
1283 inline void subfo_( Register d, Register a, Register b);
1284 inline void addco( Register d, Register a, Register b);
1285 inline void addco_( Register d, Register a, Register b);
1286 inline void subfco( Register d, Register a, Register b);
1287 inline void subfco_( Register d, Register a, Register b);
1288 inline void addeo( Register d, Register a, Register b);
1289 inline void addeo_( Register d, Register a, Register b);
1290 inline void subfeo( Register d, Register a, Register b);
1291 inline void subfeo_( Register d, Register a, Register b);
1292 inline void addmeo( Register d, Register a);
1293 inline void addmeo_( Register d, Register a);
1294 inline void subfmeo( Register d, Register a);
1295 inline void subfmeo_(Register d, Register a);
1296 inline void addzeo( Register d, Register a);
1297 inline void addzeo_( Register d, Register a);
1298 inline void subfzeo( Register d, Register a);
1299 inline void subfzeo_(Register d, Register a);
1300 inline void nego( Register d, Register a);
1301 inline void nego_( Register d, Register a);
1302 inline void mulldo( Register d, Register a, Register b);
1303 inline void mulldo_( Register d, Register a, Register b);
1304 inline void mullwo( Register d, Register a, Register b);
1305 inline void mullwo_( Register d, Register a, Register b);
1306 inline void divdo( Register d, Register a, Register b);
1307 inline void divdo_( Register d, Register a, Register b);
1308 inline void divwo( Register d, Register a, Register b);
1309 inline void divwo_( Register d, Register a, Register b);
1310
1311 // extended mnemonics
1312 inline void li( Register d, int si16);
1313 inline void lis( Register d, int si16);
1314 inline void addir(Register d, int si16, Register a);
1315 inline void subi( Register d, Register a, int si16);
1316
1317 static bool is_addi(int x) {
1318 return ADDI_OPCODE == (x & ADDI_OPCODE_MASK);
1319 }
1320 static bool is_addis(int x) {
1321 return ADDIS_OPCODE == (x & ADDIS_OPCODE_MASK);
1322 }
1323 static bool is_bxx(int x) {
1324 return BXX_OPCODE == (x & BXX_OPCODE_MASK);
1325 }
1326 static bool is_b(int x) {
1327 return BXX_OPCODE == (x & BXX_OPCODE_MASK) && inv_lk_field(x) == 0;
1328 }
1329 static bool is_bl(int x) {
1330 return BXX_OPCODE == (x & BXX_OPCODE_MASK) && inv_lk_field(x) == 1;
1331 }
1332 static bool is_bcxx(int x) {
1333 return BCXX_OPCODE == (x & BCXX_OPCODE_MASK);
1334 }
1335 static bool is_bxx_or_bcxx(int x) {
1336 return is_bxx(x) || is_bcxx(x);
1337 }
1338 static bool is_bctrl(int x) {
1339 return x == 0x4e800421;
1340 }
1341 static bool is_bctr(int x) {
1342 return x == 0x4e800420;
1343 }
1344 static bool is_bclr(int x) {
1345 return BCLR_OPCODE == (x & XL_FORM_OPCODE_MASK);
1346 }
1347 static bool is_li(int x) {
1348 return is_addi(x) && inv_ra_field(x)==0;
1349 }
1350 static bool is_lis(int x) {
1351 return is_addis(x) && inv_ra_field(x)==0;
1352 }
1353 static bool is_mtctr(int x) {
1354 return MTCTR_OPCODE == (x & MTCTR_OPCODE_MASK);
1355 }
1356 static bool is_ld(int x) {
1357 return LD_OPCODE == (x & LD_OPCODE_MASK);
1358 }
1359 static bool is_std(int x) {
1360 return STD_OPCODE == (x & STD_OPCODE_MASK);
1361 }
1362 static bool is_stdu(int x) {
1363 return STDU_OPCODE == (x & STDU_OPCODE_MASK);
1364 }
1365 static bool is_stdx(int x) {
1366 return STDX_OPCODE == (x & STDX_OPCODE_MASK);
1367 }
1368 static bool is_stdux(int x) {
1369 return STDUX_OPCODE == (x & STDUX_OPCODE_MASK);
1370 }
1371 static bool is_stwx(int x) {
1372 return STWX_OPCODE == (x & STWX_OPCODE_MASK);
1373 }
1374 static bool is_stwux(int x) {
1375 return STWUX_OPCODE == (x & STWUX_OPCODE_MASK);
1376 }
1377 static bool is_stw(int x) {
1378 return STW_OPCODE == (x & STW_OPCODE_MASK);
1379 }
1380 static bool is_stwu(int x) {
1381 return STWU_OPCODE == (x & STWU_OPCODE_MASK);
1382 }
1383 static bool is_ori(int x) {
1384 return ORI_OPCODE == (x & ORI_OPCODE_MASK);
1385 };
1386 static bool is_oris(int x) {
1387 return ORIS_OPCODE == (x & ORIS_OPCODE_MASK);
1388 };
1389 static bool is_rldicr(int x) {
1390 return (RLDICR_OPCODE == (x & RLDICR_OPCODE_MASK));
1391 };
1392 static bool is_nop(int x) {
1393 return x == 0x60000000;
1394 }
1395 // endgroup opcode for Power6
1396 static bool is_endgroup(int x) {
1397 return is_ori(x) && inv_ra_field(x) == 1 && inv_rs_field(x) == 1 && inv_d1_field(x) == 0;
1398 }
1399
1400
1401 private:
1402 // PPC 1, section 3.3.9, Fixed-Point Compare Instructions
1403 inline void cmpi( ConditionRegister bf, int l, Register a, int si16);
1404 inline void cmp( ConditionRegister bf, int l, Register a, Register b);
1405 inline void cmpli(ConditionRegister bf, int l, Register a, int ui16);
1406 inline void cmpl( ConditionRegister bf, int l, Register a, Register b);
1407
1408 public:
1409 // extended mnemonics of Compare Instructions
1410 inline void cmpwi( ConditionRegister crx, Register a, int si16);
1411 inline void cmpdi( ConditionRegister crx, Register a, int si16);
1412 inline void cmpw( ConditionRegister crx, Register a, Register b);
1413 inline void cmpd( ConditionRegister crx, Register a, Register b);
1414 inline void cmplwi(ConditionRegister crx, Register a, int ui16);
1415 inline void cmpldi(ConditionRegister crx, Register a, int ui16);
1416 inline void cmplw( ConditionRegister crx, Register a, Register b);
1417 inline void cmpld( ConditionRegister crx, Register a, Register b);
1418
1419 inline void isel( Register d, Register a, Register b, int bc);
1420 // Convenient version which takes: Condition register, Condition code and invert flag. Omit b to keep old value.
1421 inline void isel( Register d, ConditionRegister cr, Condition cc, bool inv, Register a, Register b = noreg);
1422 // Set d = 0 if (cr.cc) equals 1, otherwise b.
1423 inline void isel_0( Register d, ConditionRegister cr, Condition cc, Register b = noreg);
1424
1425 // PPC 1, section 3.3.11, Fixed-Point Logical Instructions
1426 void andi( Register a, Register s, long ui16); // optimized version
1427 inline void andi_( Register a, Register s, int ui16);
1428 inline void andis_( Register a, Register s, int ui16);
1429 inline void ori( Register a, Register s, int ui16);
1430 inline void oris( Register a, Register s, int ui16);
1431 inline void xori( Register a, Register s, int ui16);
1432 inline void xoris( Register a, Register s, int ui16);
1433 inline void andr( Register a, Register s, Register b); // suffixed by 'r' as 'and' is C++ keyword
1434 inline void and_( Register a, Register s, Register b);
1435 // Turn or0(rx,rx,rx) into a nop and avoid that we accidently emit a
1436 // SMT-priority change instruction (see SMT instructions below).
1437 inline void or_unchecked(Register a, Register s, Register b);
1438 inline void orr( Register a, Register s, Register b); // suffixed by 'r' as 'or' is C++ keyword
1439 inline void or_( Register a, Register s, Register b);
1440 inline void xorr( Register a, Register s, Register b); // suffixed by 'r' as 'xor' is C++ keyword
1441 inline void xor_( Register a, Register s, Register b);
1442 inline void nand( Register a, Register s, Register b);
1443 inline void nand_( Register a, Register s, Register b);
1444 inline void nor( Register a, Register s, Register b);
1445 inline void nor_( Register a, Register s, Register b);
1446 inline void andc( Register a, Register s, Register b);
1447 inline void andc_( Register a, Register s, Register b);
1448 inline void orc( Register a, Register s, Register b);
1449 inline void orc_( Register a, Register s, Register b);
1450 inline void extsb( Register a, Register s);
1451 inline void extsb_( Register a, Register s);
1452 inline void extsh( Register a, Register s);
1453 inline void extsh_( Register a, Register s);
1454 inline void extsw( Register a, Register s);
1455 inline void extsw_( Register a, Register s);
1456
1457 // extended mnemonics
1458 inline void nop();
1459 // NOP for FP and BR units (different versions to allow them to be in one group)
1460 inline void fpnop0();
1461 inline void fpnop1();
1462 inline void brnop0();
1463 inline void brnop1();
1464 inline void brnop2();
1465
1466 inline void mr( Register d, Register s);
1467 inline void ori_opt( Register d, int ui16);
1468 inline void oris_opt(Register d, int ui16);
1469
1470 // endgroup opcode for Power6
1471 inline void endgroup();
1472
1473 // count instructions
1474 inline void cntlzw( Register a, Register s);
1475 inline void cntlzw_( Register a, Register s);
1476 inline void cntlzd( Register a, Register s);
1477 inline void cntlzd_( Register a, Register s);
1478
1479 // PPC 1, section 3.3.12, Fixed-Point Rotate and Shift Instructions
1480 inline void sld( Register a, Register s, Register b);
1481 inline void sld_( Register a, Register s, Register b);
1482 inline void slw( Register a, Register s, Register b);
1483 inline void slw_( Register a, Register s, Register b);
1484 inline void srd( Register a, Register s, Register b);
1485 inline void srd_( Register a, Register s, Register b);
1486 inline void srw( Register a, Register s, Register b);
1487 inline void srw_( Register a, Register s, Register b);
1488 inline void srad( Register a, Register s, Register b);
1489 inline void srad_( Register a, Register s, Register b);
1490 inline void sraw( Register a, Register s, Register b);
1491 inline void sraw_( Register a, Register s, Register b);
1492 inline void sradi( Register a, Register s, int sh6);
1493 inline void sradi_( Register a, Register s, int sh6);
1494 inline void srawi( Register a, Register s, int sh5);
1495 inline void srawi_( Register a, Register s, int sh5);
1496
1497 // extended mnemonics for Shift Instructions
1498 inline void sldi( Register a, Register s, int sh6);
1499 inline void sldi_( Register a, Register s, int sh6);
1500 inline void slwi( Register a, Register s, int sh5);
1501 inline void slwi_( Register a, Register s, int sh5);
1502 inline void srdi( Register a, Register s, int sh6);
1503 inline void srdi_( Register a, Register s, int sh6);
1504 inline void srwi( Register a, Register s, int sh5);
1505 inline void srwi_( Register a, Register s, int sh5);
1506
1507 inline void clrrdi( Register a, Register s, int ui6);
1508 inline void clrrdi_( Register a, Register s, int ui6);
1509 inline void clrldi( Register a, Register s, int ui6);
1510 inline void clrldi_( Register a, Register s, int ui6);
1511 inline void clrlsldi(Register a, Register s, int clrl6, int shl6);
1512 inline void clrlsldi_(Register a, Register s, int clrl6, int shl6);
1513 inline void extrdi( Register a, Register s, int n, int b);
1514 // testbit with condition register
1515 inline void testbitdi(ConditionRegister cr, Register a, Register s, int ui6);
1516
1517 // rotate instructions
1518 inline void rotldi( Register a, Register s, int n);
1519 inline void rotrdi( Register a, Register s, int n);
1520 inline void rotlwi( Register a, Register s, int n);
1521 inline void rotrwi( Register a, Register s, int n);
1522
1523 // Rotate Instructions
1524 inline void rldic( Register a, Register s, int sh6, int mb6);
1525 inline void rldic_( Register a, Register s, int sh6, int mb6);
1526 inline void rldicr( Register a, Register s, int sh6, int mb6);
1527 inline void rldicr_( Register a, Register s, int sh6, int mb6);
1528 inline void rldicl( Register a, Register s, int sh6, int mb6);
1529 inline void rldicl_( Register a, Register s, int sh6, int mb6);
1530 inline void rlwinm( Register a, Register s, int sh5, int mb5, int me5);
1531 inline void rlwinm_( Register a, Register s, int sh5, int mb5, int me5);
1532 inline void rldimi( Register a, Register s, int sh6, int mb6);
1533 inline void rldimi_( Register a, Register s, int sh6, int mb6);
1534 inline void rlwimi( Register a, Register s, int sh5, int mb5, int me5);
1535 inline void insrdi( Register a, Register s, int n, int b);
1536 inline void insrwi( Register a, Register s, int n, int b);
1537
1538 // PPC 1, section 3.3.2 Fixed-Point Load Instructions
1539 // 4 bytes
1540 inline void lwzx( Register d, Register s1, Register s2);
1541 inline void lwz( Register d, int si16, Register s1);
1542 inline void lwzu( Register d, int si16, Register s1);
1543
1544 // 4 bytes
1545 inline void lwax( Register d, Register s1, Register s2);
1546 inline void lwa( Register d, int si16, Register s1);
1547
1548 // 4 bytes reversed
1549 inline void lwbrx( Register d, Register s1, Register s2);
1550
1551 // 2 bytes
1552 inline void lhzx( Register d, Register s1, Register s2);
1553 inline void lhz( Register d, int si16, Register s1);
1554 inline void lhzu( Register d, int si16, Register s1);
1555
1556 // 2 bytes reversed
1557 inline void lhbrx( Register d, Register s1, Register s2);
1558
1559 // 2 bytes
1560 inline void lhax( Register d, Register s1, Register s2);
1561 inline void lha( Register d, int si16, Register s1);
1562 inline void lhau( Register d, int si16, Register s1);
1563
1564 // 1 byte
1565 inline void lbzx( Register d, Register s1, Register s2);
1566 inline void lbz( Register d, int si16, Register s1);
1567 inline void lbzu( Register d, int si16, Register s1);
1568
1569 // 8 bytes
1570 inline void ldx( Register d, Register s1, Register s2);
1571 inline void ld( Register d, int si16, Register s1);
1572 inline void ldu( Register d, int si16, Register s1);
1573
1574 // 8 bytes reversed
1575 inline void ldbrx( Register d, Register s1, Register s2);
1576
1577 // For convenience. Load pointer into d from b+s1.
1578 inline void ld_ptr(Register d, int b, Register s1);
1579 DEBUG_ONLY(inline void ld_ptr(Register d, ByteSize b, Register s1);)
1580
1581 // PPC 1, section 3.3.3 Fixed-Point Store Instructions
1582 inline void stwx( Register d, Register s1, Register s2);
1583 inline void stw( Register d, int si16, Register s1);
1584 inline void stwu( Register d, int si16, Register s1);
1585 inline void stwbrx( Register d, Register s1, Register s2);
1586
1587 inline void sthx( Register d, Register s1, Register s2);
1588 inline void sth( Register d, int si16, Register s1);
1589 inline void sthu( Register d, int si16, Register s1);
1590 inline void sthbrx( Register d, Register s1, Register s2);
1591
1592 inline void stbx( Register d, Register s1, Register s2);
1593 inline void stb( Register d, int si16, Register s1);
1594 inline void stbu( Register d, int si16, Register s1);
1595
1596 inline void stdx( Register d, Register s1, Register s2);
1597 inline void std( Register d, int si16, Register s1);
1598 inline void stdu( Register d, int si16, Register s1);
1599 inline void stdux(Register s, Register a, Register b);
1600 inline void stdbrx( Register d, Register s1, Register s2);
1601
1602 inline void st_ptr(Register d, int si16, Register s1);
1603 DEBUG_ONLY(inline void st_ptr(Register d, ByteSize b, Register s1);)
1604
1605 // PPC 1, section 3.3.13 Move To/From System Register Instructions
1606 inline void mtlr( Register s1);
1607 inline void mflr( Register d);
1608 inline void mtctr(Register s1);
1609 inline void mfctr(Register d);
1610 inline void mtcrf(int fxm, Register s);
1611 inline void mfcr( Register d);
1612 inline void mcrf( ConditionRegister crd, ConditionRegister cra);
1613 inline void mtcr( Register s);
1614
1615 // Special purpose registers
1616 // Exception Register
1617 inline void mtxer(Register s1);
1618 inline void mfxer(Register d);
1619 // Vector Register Save Register
1620 inline void mtvrsave(Register s1);
1621 inline void mfvrsave(Register d);
1622 // Timebase
1623 inline void mftb(Register d);
1624 // Introduced with Power 8:
1625 // Data Stream Control Register
1626 inline void mtdscr(Register s1);
1627 inline void mfdscr(Register d );
1628 // Transactional Memory Registers
1629 inline void mftfhar(Register d);
1630 inline void mftfiar(Register d);
1631 inline void mftexasr(Register d);
1632 inline void mftexasru(Register d);
1633
1634 // TEXASR bit description
1635 enum transaction_failure_reason {
1636 // Upper half (TEXASRU):
1637 tm_failure_persistent = 7, // The failure is likely to recur on each execution.
1638 tm_disallowed = 8, // The instruction is not permitted.
1639 tm_nesting_of = 9, // The maximum transaction level was exceeded.
1640 tm_footprint_of = 10, // The tracking limit for transactional storage accesses was exceeded.
1641 tm_self_induced_cf = 11, // A self-induced conflict occurred in Suspended state.
1642 tm_non_trans_cf = 12, // A conflict occurred with a non-transactional access by another processor.
1643 tm_trans_cf = 13, // A conflict occurred with another transaction.
1644 tm_translation_cf = 14, // A conflict occurred with a TLB invalidation.
1645 tm_inst_fetch_cf = 16, // An instruction fetch was performed from a block that was previously written transactionally.
1646 tm_tabort = 31, // Termination was caused by the execution of an abort instruction.
1647 // Lower half:
1648 tm_suspended = 32, // Failure was recorded in Suspended state.
1649 tm_failure_summary = 36, // Failure has been detected and recorded.
1650 tm_tfiar_exact = 37, // Value in the TFIAR is exact.
1651 tm_rot = 38, // Rollback-only transaction.
1652 };
1653
1654 // PPC 1, section 2.4.1 Branch Instructions
1655 inline void b( address a, relocInfo::relocType rt = relocInfo::none);
1656 inline void b( Label& L);
1657 inline void bl( address a, relocInfo::relocType rt = relocInfo::none);
1658 inline void bl( Label& L);
1659 inline void bc( int boint, int biint, address a, relocInfo::relocType rt = relocInfo::none);
1660 inline void bc( int boint, int biint, Label& L);
1661 inline void bcl(int boint, int biint, address a, relocInfo::relocType rt = relocInfo::none);
1662 inline void bcl(int boint, int biint, Label& L);
1663
1664 inline void bclr( int boint, int biint, int bhint, relocInfo::relocType rt = relocInfo::none);
1665 inline void bclrl( int boint, int biint, int bhint, relocInfo::relocType rt = relocInfo::none);
1666 inline void bcctr( int boint, int biint, int bhint = bhintbhBCCTRisNotReturnButSame,
1667 relocInfo::relocType rt = relocInfo::none);
1668 inline void bcctrl(int boint, int biint, int bhint = bhintbhBCLRisReturn,
1669 relocInfo::relocType rt = relocInfo::none);
1670
1671 // helper function for b, bcxx
1672 inline bool is_within_range_of_b(address a, address pc);
1673 inline bool is_within_range_of_bcxx(address a, address pc);
1674
1675 // get the destination of a bxx branch (b, bl, ba, bla)
1676 static inline address bxx_destination(address baddr);
1677 static inline address bxx_destination(int instr, address pc);
1678 static inline intptr_t bxx_destination_offset(int instr, intptr_t bxx_pos);
1679
1680 // extended mnemonics for branch instructions
1681 inline void blt(ConditionRegister crx, Label& L);
1682 inline void bgt(ConditionRegister crx, Label& L);
1683 inline void beq(ConditionRegister crx, Label& L);
1684 inline void bso(ConditionRegister crx, Label& L);
1685 inline void bge(ConditionRegister crx, Label& L);
1686 inline void ble(ConditionRegister crx, Label& L);
1687 inline void bne(ConditionRegister crx, Label& L);
1688 inline void bns(ConditionRegister crx, Label& L);
1689
1690 // Branch instructions with static prediction hints.
1691 inline void blt_predict_taken( ConditionRegister crx, Label& L);
1692 inline void bgt_predict_taken( ConditionRegister crx, Label& L);
1693 inline void beq_predict_taken( ConditionRegister crx, Label& L);
1694 inline void bso_predict_taken( ConditionRegister crx, Label& L);
1695 inline void bge_predict_taken( ConditionRegister crx, Label& L);
1696 inline void ble_predict_taken( ConditionRegister crx, Label& L);
1697 inline void bne_predict_taken( ConditionRegister crx, Label& L);
1698 inline void bns_predict_taken( ConditionRegister crx, Label& L);
1699 inline void blt_predict_not_taken(ConditionRegister crx, Label& L);
1700 inline void bgt_predict_not_taken(ConditionRegister crx, Label& L);
1701 inline void beq_predict_not_taken(ConditionRegister crx, Label& L);
1702 inline void bso_predict_not_taken(ConditionRegister crx, Label& L);
1703 inline void bge_predict_not_taken(ConditionRegister crx, Label& L);
1704 inline void ble_predict_not_taken(ConditionRegister crx, Label& L);
1705 inline void bne_predict_not_taken(ConditionRegister crx, Label& L);
1706 inline void bns_predict_not_taken(ConditionRegister crx, Label& L);
1707
1708 // for use in conjunction with testbitdi:
1709 inline void btrue( ConditionRegister crx, Label& L);
1710 inline void bfalse(ConditionRegister crx, Label& L);
1711
1712 inline void bltl(ConditionRegister crx, Label& L);
1713 inline void bgtl(ConditionRegister crx, Label& L);
1714 inline void beql(ConditionRegister crx, Label& L);
1715 inline void bsol(ConditionRegister crx, Label& L);
1716 inline void bgel(ConditionRegister crx, Label& L);
1717 inline void blel(ConditionRegister crx, Label& L);
1718 inline void bnel(ConditionRegister crx, Label& L);
1719 inline void bnsl(ConditionRegister crx, Label& L);
1720
1721 // extended mnemonics for Branch Instructions via LR
1722 // We use `blr' for returns.
1723 inline void blr(relocInfo::relocType rt = relocInfo::none);
1724
1725 // extended mnemonics for Branch Instructions with CTR
1726 // bdnz means `decrement CTR and jump to L if CTR is not zero'
1727 inline void bdnz(Label& L);
1728 // Decrement and branch if result is zero.
1729 inline void bdz(Label& L);
1730 // we use `bctr[l]' for jumps/calls in function descriptor glue
1731 // code, e.g. calls to runtime functions
1732 inline void bctr( relocInfo::relocType rt = relocInfo::none);
1733 inline void bctrl(relocInfo::relocType rt = relocInfo::none);
1734 // conditional jumps/branches via CTR
1735 inline void beqctr( ConditionRegister crx, relocInfo::relocType rt = relocInfo::none);
1736 inline void beqctrl(ConditionRegister crx, relocInfo::relocType rt = relocInfo::none);
1737 inline void bnectr( ConditionRegister crx, relocInfo::relocType rt = relocInfo::none);
1738 inline void bnectrl(ConditionRegister crx, relocInfo::relocType rt = relocInfo::none);
1739
1740 // condition register logic instructions
1741 // NOTE: There's a preferred form: d and s2 should point into the same condition register.
1742 inline void crand( int d, int s1, int s2);
1743 inline void crnand(int d, int s1, int s2);
1744 inline void cror( int d, int s1, int s2);
1745 inline void crxor( int d, int s1, int s2);
1746 inline void crnor( int d, int s1, int s2);
1747 inline void creqv( int d, int s1, int s2);
1748 inline void crandc(int d, int s1, int s2);
1749 inline void crorc( int d, int s1, int s2);
1750
1751 // More convenient version.
1752 int condition_register_bit(ConditionRegister cr, Condition c) {
1753 return 4 * (int)(intptr_t)cr + c;
1754 }
1755 void crand( ConditionRegister crdst, Condition cdst, ConditionRegister crsrc, Condition csrc);
1756 void crnand(ConditionRegister crdst, Condition cdst, ConditionRegister crsrc, Condition csrc);
1757 void cror( ConditionRegister crdst, Condition cdst, ConditionRegister crsrc, Condition csrc);
1758 void crxor( ConditionRegister crdst, Condition cdst, ConditionRegister crsrc, Condition csrc);
1759 void crnor( ConditionRegister crdst, Condition cdst, ConditionRegister crsrc, Condition csrc);
1760 void creqv( ConditionRegister crdst, Condition cdst, ConditionRegister crsrc, Condition csrc);
1761 void crandc(ConditionRegister crdst, Condition cdst, ConditionRegister crsrc, Condition csrc);
1762 void crorc( ConditionRegister crdst, Condition cdst, ConditionRegister crsrc, Condition csrc);
1763
1764 // icache and dcache related instructions
1765 inline void icbi( Register s1, Register s2);
1766 //inline void dcba(Register s1, Register s2); // Instruction for embedded processor only.
1767 inline void dcbz( Register s1, Register s2);
1768 inline void dcbst( Register s1, Register s2);
1769 inline void dcbf( Register s1, Register s2);
1770
1771 enum ct_cache_specification {
1772 ct_primary_cache = 0,
1773 ct_secondary_cache = 2
1774 };
1775 // dcache read hint
1776 inline void dcbt( Register s1, Register s2);
1777 inline void dcbtct( Register s1, Register s2, int ct);
1778 inline void dcbtds( Register s1, Register s2, int ds);
1779 // dcache write hint
1780 inline void dcbtst( Register s1, Register s2);
1781 inline void dcbtstct(Register s1, Register s2, int ct);
1782
1783 // machine barrier instructions:
1784 //
1785 // - sync two-way memory barrier, aka fence
1786 // - lwsync orders Store|Store,
1787 // Load|Store,
1788 // Load|Load,
1789 // but not Store|Load
1790 // - eieio orders memory accesses for device memory (only)
1791 // - isync invalidates speculatively executed instructions
1792 // From the Power ISA 2.06 documentation:
1793 // "[...] an isync instruction prevents the execution of
1794 // instructions following the isync until instructions
1795 // preceding the isync have completed, [...]"
1796 // From IBM's AIX assembler reference:
1797 // "The isync [...] instructions causes the processor to
1798 // refetch any instructions that might have been fetched
1799 // prior to the isync instruction. The instruction isync
1800 // causes the processor to wait for all previous instructions
1801 // to complete. Then any instructions already fetched are
1802 // discarded and instruction processing continues in the
1803 // environment established by the previous instructions."
1804 //
1805 // semantic barrier instructions:
1806 // (as defined in orderAccess.hpp)
1807 //
1808 // - release orders Store|Store, (maps to lwsync)
1809 // Load|Store
1810 // - acquire orders Load|Store, (maps to lwsync)
1811 // Load|Load
1812 // - fence orders Store|Store, (maps to sync)
1813 // Load|Store,
1814 // Load|Load,
1815 // Store|Load
1816 //
1817 private:
1818 inline void sync(int l);
1819 public:
1820 inline void sync();
1821 inline void lwsync();
1822 inline void ptesync();
1823 inline void eieio();
1824 inline void isync();
1825 inline void elemental_membar(int e); // Elemental Memory Barriers (>=Power 8)
1826
1827 // Wait instructions for polling. Attention: May result in SIGILL.
1828 inline void wait();
1829 inline void waitrsv(); // >=Power7
1830
1831 // atomics
1832 inline void lbarx_unchecked(Register d, Register a, Register b, int eh1 = 0); // >=Power 8
1833 inline void lharx_unchecked(Register d, Register a, Register b, int eh1 = 0); // >=Power 8
1834 inline void lwarx_unchecked(Register d, Register a, Register b, int eh1 = 0);
1835 inline void ldarx_unchecked(Register d, Register a, Register b, int eh1 = 0);
1836 inline void lqarx_unchecked(Register d, Register a, Register b, int eh1 = 0); // >=Power 8
1837 inline bool lxarx_hint_exclusive_access();
1838 inline void lbarx( Register d, Register a, Register b, bool hint_exclusive_access = false);
1839 inline void lharx( Register d, Register a, Register b, bool hint_exclusive_access = false);
1840 inline void lwarx( Register d, Register a, Register b, bool hint_exclusive_access = false);
1841 inline void ldarx( Register d, Register a, Register b, bool hint_exclusive_access = false);
1842 inline void lqarx( Register d, Register a, Register b, bool hint_exclusive_access = false);
1843 inline void stbcx_( Register s, Register a, Register b);
1844 inline void sthcx_( Register s, Register a, Register b);
1845 inline void stwcx_( Register s, Register a, Register b);
1846 inline void stdcx_( Register s, Register a, Register b);
1847 inline void stqcx_( Register s, Register a, Register b);
1848
1849 // Instructions for adjusting thread priority for simultaneous
1850 // multithreading (SMT) on Power5.
1851 private:
1852 inline void smt_prio_very_low();
1853 inline void smt_prio_medium_high();
1854 inline void smt_prio_high();
1855
1856 public:
1857 inline void smt_prio_low();
1858 inline void smt_prio_medium_low();
1859 inline void smt_prio_medium();
1860 // >= Power7
1861 inline void smt_yield();
1862 inline void smt_mdoio();
1863 inline void smt_mdoom();
1864 // >= Power8
1865 inline void smt_miso();
1866
1867 // trap instructions
1868 inline void twi_0(Register a); // for load with acquire semantics use load+twi_0+isync (trap can't occur)
1869 // NOT FOR DIRECT USE!!
1870 protected:
1871 inline void tdi_unchecked(int tobits, Register a, int si16);
1872 inline void twi_unchecked(int tobits, Register a, int si16);
1873 inline void tdi( int tobits, Register a, int si16); // asserts UseSIGTRAP
1874 inline void twi( int tobits, Register a, int si16); // asserts UseSIGTRAP
1875 inline void td( int tobits, Register a, Register b); // asserts UseSIGTRAP
1876 inline void tw( int tobits, Register a, Register b); // asserts UseSIGTRAP
1877
1878 static bool is_tdi(int x, int tobits, int ra, int si16) {
1879 return (TDI_OPCODE == (x & TDI_OPCODE_MASK))
1880 && (tobits == inv_to_field(x))
1881 && (ra == -1/*any reg*/ || ra == inv_ra_field(x))
1882 && (si16 == inv_si_field(x));
1883 }
1884
1885 static bool is_twi(int x, int tobits, int ra, int si16) {
1886 return (TWI_OPCODE == (x & TWI_OPCODE_MASK))
1887 && (tobits == inv_to_field(x))
1888 && (ra == -1/*any reg*/ || ra == inv_ra_field(x))
1889 && (si16 == inv_si_field(x));
1890 }
1891
1892 static bool is_twi(int x, int tobits, int ra) {
1893 return (TWI_OPCODE == (x & TWI_OPCODE_MASK))
1894 && (tobits == inv_to_field(x))
1895 && (ra == -1/*any reg*/ || ra == inv_ra_field(x));
1896 }
1897
1898 static bool is_td(int x, int tobits, int ra, int rb) {
1899 return (TD_OPCODE == (x & TD_OPCODE_MASK))
1900 && (tobits == inv_to_field(x))
1901 && (ra == -1/*any reg*/ || ra == inv_ra_field(x))
1902 && (rb == -1/*any reg*/ || rb == inv_rb_field(x));
1903 }
1904
1905 static bool is_tw(int x, int tobits, int ra, int rb) {
1906 return (TW_OPCODE == (x & TW_OPCODE_MASK))
1907 && (tobits == inv_to_field(x))
1908 && (ra == -1/*any reg*/ || ra == inv_ra_field(x))
1909 && (rb == -1/*any reg*/ || rb == inv_rb_field(x));
1910 }
1911
1912 public:
1913 // PPC floating point instructions
1914 // PPC 1, section 4.6.2 Floating-Point Load Instructions
1915 inline void lfs( FloatRegister d, int si16, Register a);
1916 inline void lfsu( FloatRegister d, int si16, Register a);
1917 inline void lfsx( FloatRegister d, Register a, Register b);
1918 inline void lfd( FloatRegister d, int si16, Register a);
1919 inline void lfdu( FloatRegister d, int si16, Register a);
1920 inline void lfdx( FloatRegister d, Register a, Register b);
1921
1922 // PPC 1, section 4.6.3 Floating-Point Store Instructions
1923 inline void stfs( FloatRegister s, int si16, Register a);
1924 inline void stfsu( FloatRegister s, int si16, Register a);
1925 inline void stfsx( FloatRegister s, Register a, Register b);
1926 inline void stfd( FloatRegister s, int si16, Register a);
1927 inline void stfdu( FloatRegister s, int si16, Register a);
1928 inline void stfdx( FloatRegister s, Register a, Register b);
1929
1930 // PPC 1, section 4.6.4 Floating-Point Move Instructions
1931 inline void fmr( FloatRegister d, FloatRegister b);
1932 inline void fmr_( FloatRegister d, FloatRegister b);
1933
1934 // inline void mffgpr( FloatRegister d, Register b);
1935 // inline void mftgpr( Register d, FloatRegister b);
1936 inline void cmpb( Register a, Register s, Register b);
1937 inline void popcntb(Register a, Register s);
1938 inline void popcntw(Register a, Register s);
1939 inline void popcntd(Register a, Register s);
1940
1941 inline void fneg( FloatRegister d, FloatRegister b);
1942 inline void fneg_( FloatRegister d, FloatRegister b);
1943 inline void fabs( FloatRegister d, FloatRegister b);
1944 inline void fabs_( FloatRegister d, FloatRegister b);
1945 inline void fnabs( FloatRegister d, FloatRegister b);
1946 inline void fnabs_(FloatRegister d, FloatRegister b);
1947
1948 // PPC 1, section 4.6.5.1 Floating-Point Elementary Arithmetic Instructions
1949 inline void fadd( FloatRegister d, FloatRegister a, FloatRegister b);
1950 inline void fadd_( FloatRegister d, FloatRegister a, FloatRegister b);
1951 inline void fadds( FloatRegister d, FloatRegister a, FloatRegister b);
1952 inline void fadds_(FloatRegister d, FloatRegister a, FloatRegister b);
1953 inline void fsub( FloatRegister d, FloatRegister a, FloatRegister b);
1954 inline void fsub_( FloatRegister d, FloatRegister a, FloatRegister b);
1955 inline void fsubs( FloatRegister d, FloatRegister a, FloatRegister b);
1956 inline void fsubs_(FloatRegister d, FloatRegister a, FloatRegister b);
1957 inline void fmul( FloatRegister d, FloatRegister a, FloatRegister c);
1958 inline void fmul_( FloatRegister d, FloatRegister a, FloatRegister c);
1959 inline void fmuls( FloatRegister d, FloatRegister a, FloatRegister c);
1960 inline void fmuls_(FloatRegister d, FloatRegister a, FloatRegister c);
1961 inline void fdiv( FloatRegister d, FloatRegister a, FloatRegister b);
1962 inline void fdiv_( FloatRegister d, FloatRegister a, FloatRegister b);
1963 inline void fdivs( FloatRegister d, FloatRegister a, FloatRegister b);
1964 inline void fdivs_(FloatRegister d, FloatRegister a, FloatRegister b);
1965
1966 // Fused multiply-accumulate instructions.
1967 // WARNING: Use only when rounding between the 2 parts is not desired.
1968 // Some floating point tck tests will fail if used incorrectly.
1969 inline void fmadd( FloatRegister d, FloatRegister a, FloatRegister c, FloatRegister b);
1970 inline void fmadd_( FloatRegister d, FloatRegister a, FloatRegister c, FloatRegister b);
1971 inline void fmadds( FloatRegister d, FloatRegister a, FloatRegister c, FloatRegister b);
1972 inline void fmadds_( FloatRegister d, FloatRegister a, FloatRegister c, FloatRegister b);
1973 inline void fmsub( FloatRegister d, FloatRegister a, FloatRegister c, FloatRegister b);
1974 inline void fmsub_( FloatRegister d, FloatRegister a, FloatRegister c, FloatRegister b);
1975 inline void fmsubs( FloatRegister d, FloatRegister a, FloatRegister c, FloatRegister b);
1976 inline void fmsubs_( FloatRegister d, FloatRegister a, FloatRegister c, FloatRegister b);
1977 inline void fnmadd( FloatRegister d, FloatRegister a, FloatRegister c, FloatRegister b);
1978 inline void fnmadd_( FloatRegister d, FloatRegister a, FloatRegister c, FloatRegister b);
1979 inline void fnmadds( FloatRegister d, FloatRegister a, FloatRegister c, FloatRegister b);
1980 inline void fnmadds_(FloatRegister d, FloatRegister a, FloatRegister c, FloatRegister b);
1981 inline void fnmsub( FloatRegister d, FloatRegister a, FloatRegister c, FloatRegister b);
1982 inline void fnmsub_( FloatRegister d, FloatRegister a, FloatRegister c, FloatRegister b);
1983 inline void fnmsubs( FloatRegister d, FloatRegister a, FloatRegister c, FloatRegister b);
1984 inline void fnmsubs_(FloatRegister d, FloatRegister a, FloatRegister c, FloatRegister b);
1985
1986 // PPC 1, section 4.6.6 Floating-Point Rounding and Conversion Instructions
1987 inline void frsp( FloatRegister d, FloatRegister b);
1988 inline void fctid( FloatRegister d, FloatRegister b);
1989 inline void fctidz(FloatRegister d, FloatRegister b);
1990 inline void fctiw( FloatRegister d, FloatRegister b);
1991 inline void fctiwz(FloatRegister d, FloatRegister b);
1992 inline void fcfid( FloatRegister d, FloatRegister b);
1993 inline void fcfids(FloatRegister d, FloatRegister b);
1994
1995 // PPC 1, section 4.6.7 Floating-Point Compare Instructions
1996 inline void fcmpu( ConditionRegister crx, FloatRegister a, FloatRegister b);
1997
1998 inline void fsqrt( FloatRegister d, FloatRegister b);
1999 inline void fsqrts(FloatRegister d, FloatRegister b);
2000
2001 // Vector instructions for >= Power6.
2002 inline void lvebx( VectorRegister d, Register s1, Register s2);
2003 inline void lvehx( VectorRegister d, Register s1, Register s2);
2004 inline void lvewx( VectorRegister d, Register s1, Register s2);
2005 inline void lvx( VectorRegister d, Register s1, Register s2);
2006 inline void lvxl( VectorRegister d, Register s1, Register s2);
2007 inline void stvebx( VectorRegister d, Register s1, Register s2);
2008 inline void stvehx( VectorRegister d, Register s1, Register s2);
2009 inline void stvewx( VectorRegister d, Register s1, Register s2);
2010 inline void stvx( VectorRegister d, Register s1, Register s2);
2011 inline void stvxl( VectorRegister d, Register s1, Register s2);
2012 inline void lvsl( VectorRegister d, Register s1, Register s2);
2013 inline void lvsr( VectorRegister d, Register s1, Register s2);
2014 inline void vpkpx( VectorRegister d, VectorRegister a, VectorRegister b);
2015 inline void vpkshss( VectorRegister d, VectorRegister a, VectorRegister b);
2016 inline void vpkswss( VectorRegister d, VectorRegister a, VectorRegister b);
2017 inline void vpkshus( VectorRegister d, VectorRegister a, VectorRegister b);
2018 inline void vpkswus( VectorRegister d, VectorRegister a, VectorRegister b);
2019 inline void vpkuhum( VectorRegister d, VectorRegister a, VectorRegister b);
2020 inline void vpkuwum( VectorRegister d, VectorRegister a, VectorRegister b);
2021 inline void vpkuhus( VectorRegister d, VectorRegister a, VectorRegister b);
2022 inline void vpkuwus( VectorRegister d, VectorRegister a, VectorRegister b);
2023 inline void vupkhpx( VectorRegister d, VectorRegister b);
2024 inline void vupkhsb( VectorRegister d, VectorRegister b);
2025 inline void vupkhsh( VectorRegister d, VectorRegister b);
2026 inline void vupklpx( VectorRegister d, VectorRegister b);
2027 inline void vupklsb( VectorRegister d, VectorRegister b);
2028 inline void vupklsh( VectorRegister d, VectorRegister b);
2029 inline void vmrghb( VectorRegister d, VectorRegister a, VectorRegister b);
2030 inline void vmrghw( VectorRegister d, VectorRegister a, VectorRegister b);
2031 inline void vmrghh( VectorRegister d, VectorRegister a, VectorRegister b);
2032 inline void vmrglb( VectorRegister d, VectorRegister a, VectorRegister b);
2033 inline void vmrglw( VectorRegister d, VectorRegister a, VectorRegister b);
2034 inline void vmrglh( VectorRegister d, VectorRegister a, VectorRegister b);
2035 inline void vsplt( VectorRegister d, int ui4, VectorRegister b);
2036 inline void vsplth( VectorRegister d, int ui3, VectorRegister b);
2037 inline void vspltw( VectorRegister d, int ui2, VectorRegister b);
2038 inline void vspltisb( VectorRegister d, int si5);
2039 inline void vspltish( VectorRegister d, int si5);
2040 inline void vspltisw( VectorRegister d, int si5);
2041 inline void vperm( VectorRegister d, VectorRegister a, VectorRegister b, VectorRegister c);
2042 inline void vsel( VectorRegister d, VectorRegister a, VectorRegister b, VectorRegister c);
2043 inline void vsl( VectorRegister d, VectorRegister a, VectorRegister b);
2044 inline void vsldoi( VectorRegister d, VectorRegister a, VectorRegister b, int ui4);
2045 inline void vslo( VectorRegister d, VectorRegister a, VectorRegister b);
2046 inline void vsr( VectorRegister d, VectorRegister a, VectorRegister b);
2047 inline void vsro( VectorRegister d, VectorRegister a, VectorRegister b);
2048 inline void vaddcuw( VectorRegister d, VectorRegister a, VectorRegister b);
2049 inline void vaddshs( VectorRegister d, VectorRegister a, VectorRegister b);
2050 inline void vaddsbs( VectorRegister d, VectorRegister a, VectorRegister b);
2051 inline void vaddsws( VectorRegister d, VectorRegister a, VectorRegister b);
2052 inline void vaddubm( VectorRegister d, VectorRegister a, VectorRegister b);
2053 inline void vadduwm( VectorRegister d, VectorRegister a, VectorRegister b);
2054 inline void vadduhm( VectorRegister d, VectorRegister a, VectorRegister b);
2055 inline void vaddudm( VectorRegister d, VectorRegister a, VectorRegister b);
2056 inline void vaddubs( VectorRegister d, VectorRegister a, VectorRegister b);
2057 inline void vadduws( VectorRegister d, VectorRegister a, VectorRegister b);
2058 inline void vadduhs( VectorRegister d, VectorRegister a, VectorRegister b);
2059 inline void vsubcuw( VectorRegister d, VectorRegister a, VectorRegister b);
2060 inline void vsubshs( VectorRegister d, VectorRegister a, VectorRegister b);
2061 inline void vsubsbs( VectorRegister d, VectorRegister a, VectorRegister b);
2062 inline void vsubsws( VectorRegister d, VectorRegister a, VectorRegister b);
2063 inline void vsububm( VectorRegister d, VectorRegister a, VectorRegister b);
2064 inline void vsubuwm( VectorRegister d, VectorRegister a, VectorRegister b);
2065 inline void vsubuhm( VectorRegister d, VectorRegister a, VectorRegister b);
2066 inline void vsububs( VectorRegister d, VectorRegister a, VectorRegister b);
2067 inline void vsubuws( VectorRegister d, VectorRegister a, VectorRegister b);
2068 inline void vsubuhs( VectorRegister d, VectorRegister a, VectorRegister b);
2069 inline void vmulesb( VectorRegister d, VectorRegister a, VectorRegister b);
2070 inline void vmuleub( VectorRegister d, VectorRegister a, VectorRegister b);
2071 inline void vmulesh( VectorRegister d, VectorRegister a, VectorRegister b);
2072 inline void vmuleuh( VectorRegister d, VectorRegister a, VectorRegister b);
2073 inline void vmulosb( VectorRegister d, VectorRegister a, VectorRegister b);
2074 inline void vmuloub( VectorRegister d, VectorRegister a, VectorRegister b);
2075 inline void vmulosh( VectorRegister d, VectorRegister a, VectorRegister b);
2076 inline void vmulouh( VectorRegister d, VectorRegister a, VectorRegister b);
2077 inline void vmhaddshs(VectorRegister d, VectorRegister a, VectorRegister b, VectorRegister c);
2078 inline void vmhraddshs(VectorRegister d,VectorRegister a, VectorRegister b, VectorRegister c);
2079 inline void vmladduhm(VectorRegister d, VectorRegister a, VectorRegister b, VectorRegister c);
2080 inline void vmsubuhm( VectorRegister d, VectorRegister a, VectorRegister b, VectorRegister c);
2081 inline void vmsummbm( VectorRegister d, VectorRegister a, VectorRegister b, VectorRegister c);
2082 inline void vmsumshm( VectorRegister d, VectorRegister a, VectorRegister b, VectorRegister c);
2083 inline void vmsumshs( VectorRegister d, VectorRegister a, VectorRegister b, VectorRegister c);
2084 inline void vmsumuhm( VectorRegister d, VectorRegister a, VectorRegister b, VectorRegister c);
2085 inline void vmsumuhs( VectorRegister d, VectorRegister a, VectorRegister b, VectorRegister c);
2086 inline void vsumsws( VectorRegister d, VectorRegister a, VectorRegister b);
2087 inline void vsum2sws( VectorRegister d, VectorRegister a, VectorRegister b);
2088 inline void vsum4sbs( VectorRegister d, VectorRegister a, VectorRegister b);
2089 inline void vsum4ubs( VectorRegister d, VectorRegister a, VectorRegister b);
2090 inline void vsum4shs( VectorRegister d, VectorRegister a, VectorRegister b);
2091 inline void vavgsb( VectorRegister d, VectorRegister a, VectorRegister b);
2092 inline void vavgsw( VectorRegister d, VectorRegister a, VectorRegister b);
2093 inline void vavgsh( VectorRegister d, VectorRegister a, VectorRegister b);
2094 inline void vavgub( VectorRegister d, VectorRegister a, VectorRegister b);
2095 inline void vavguw( VectorRegister d, VectorRegister a, VectorRegister b);
2096 inline void vavguh( VectorRegister d, VectorRegister a, VectorRegister b);
2097 inline void vmaxsb( VectorRegister d, VectorRegister a, VectorRegister b);
2098 inline void vmaxsw( VectorRegister d, VectorRegister a, VectorRegister b);
2099 inline void vmaxsh( VectorRegister d, VectorRegister a, VectorRegister b);
2100 inline void vmaxub( VectorRegister d, VectorRegister a, VectorRegister b);
2101 inline void vmaxuw( VectorRegister d, VectorRegister a, VectorRegister b);
2102 inline void vmaxuh( VectorRegister d, VectorRegister a, VectorRegister b);
2103 inline void vminsb( VectorRegister d, VectorRegister a, VectorRegister b);
2104 inline void vminsw( VectorRegister d, VectorRegister a, VectorRegister b);
2105 inline void vminsh( VectorRegister d, VectorRegister a, VectorRegister b);
2106 inline void vminub( VectorRegister d, VectorRegister a, VectorRegister b);
2107 inline void vminuw( VectorRegister d, VectorRegister a, VectorRegister b);
2108 inline void vminuh( VectorRegister d, VectorRegister a, VectorRegister b);
2109 inline void vcmpequb( VectorRegister d, VectorRegister a, VectorRegister b);
2110 inline void vcmpequh( VectorRegister d, VectorRegister a, VectorRegister b);
2111 inline void vcmpequw( VectorRegister d, VectorRegister a, VectorRegister b);
2112 inline void vcmpgtsh( VectorRegister d, VectorRegister a, VectorRegister b);
2113 inline void vcmpgtsb( VectorRegister d, VectorRegister a, VectorRegister b);
2114 inline void vcmpgtsw( VectorRegister d, VectorRegister a, VectorRegister b);
2115 inline void vcmpgtub( VectorRegister d, VectorRegister a, VectorRegister b);
2116 inline void vcmpgtuh( VectorRegister d, VectorRegister a, VectorRegister b);
2117 inline void vcmpgtuw( VectorRegister d, VectorRegister a, VectorRegister b);
2118 inline void vcmpequb_(VectorRegister d, VectorRegister a, VectorRegister b);
2119 inline void vcmpequh_(VectorRegister d, VectorRegister a, VectorRegister b);
2120 inline void vcmpequw_(VectorRegister d, VectorRegister a, VectorRegister b);
2121 inline void vcmpgtsh_(VectorRegister d, VectorRegister a, VectorRegister b);
2122 inline void vcmpgtsb_(VectorRegister d, VectorRegister a, VectorRegister b);
2123 inline void vcmpgtsw_(VectorRegister d, VectorRegister a, VectorRegister b);
2124 inline void vcmpgtub_(VectorRegister d, VectorRegister a, VectorRegister b);
2125 inline void vcmpgtuh_(VectorRegister d, VectorRegister a, VectorRegister b);
2126 inline void vcmpgtuw_(VectorRegister d, VectorRegister a, VectorRegister b);
2127 inline void vand( VectorRegister d, VectorRegister a, VectorRegister b);
2128 inline void vandc( VectorRegister d, VectorRegister a, VectorRegister b);
2129 inline void vnor( VectorRegister d, VectorRegister a, VectorRegister b);
2130 inline void vor( VectorRegister d, VectorRegister a, VectorRegister b);
2131 inline void vmr( VectorRegister d, VectorRegister a);
2132 inline void vxor( VectorRegister d, VectorRegister a, VectorRegister b);
2133 inline void vrld( VectorRegister d, VectorRegister a, VectorRegister b);
2134 inline void vrlb( VectorRegister d, VectorRegister a, VectorRegister b);
2135 inline void vrlw( VectorRegister d, VectorRegister a, VectorRegister b);
2136 inline void vrlh( VectorRegister d, VectorRegister a, VectorRegister b);
2137 inline void vslb( VectorRegister d, VectorRegister a, VectorRegister b);
2138 inline void vskw( VectorRegister d, VectorRegister a, VectorRegister b);
2139 inline void vslh( VectorRegister d, VectorRegister a, VectorRegister b);
2140 inline void vsrb( VectorRegister d, VectorRegister a, VectorRegister b);
2141 inline void vsrw( VectorRegister d, VectorRegister a, VectorRegister b);
2142 inline void vsrh( VectorRegister d, VectorRegister a, VectorRegister b);
2143 inline void vsrab( VectorRegister d, VectorRegister a, VectorRegister b);
2144 inline void vsraw( VectorRegister d, VectorRegister a, VectorRegister b);
2145 inline void vsrah( VectorRegister d, VectorRegister a, VectorRegister b);
2146 // Vector Floating-Point not implemented yet
2147 inline void mtvscr( VectorRegister b);
2148 inline void mfvscr( VectorRegister d);
2149
2150 // Vector-Scalar (VSX) instructions.
2151 inline void lxvd2x( VectorSRegister d, Register a);
2152 inline void lxvd2x( VectorSRegister d, Register a, Register b);
2153 inline void stxvd2x( VectorSRegister d, Register a);
2154 inline void stxvd2x( VectorSRegister d, Register a, Register b);
2155 inline void mtvrwz( VectorRegister d, Register a);
2156 inline void mfvrwz( Register a, VectorRegister d);
2157 inline void mtvrd( VectorRegister d, Register a);
2158 inline void mfvrd( Register a, VectorRegister d);
2159 inline void xxpermdi( VectorSRegister d, VectorSRegister a, VectorSRegister b, int dm);
2160 inline void xxmrghw( VectorSRegister d, VectorSRegister a, VectorSRegister b);
2161 inline void xxmrglw( VectorSRegister d, VectorSRegister a, VectorSRegister b);
2162 inline void mtvsrd( VectorSRegister d, Register a);
2163 inline void mtvsrwz( VectorSRegister d, Register a);
2164 inline void xxspltw( VectorSRegister d, VectorSRegister b, int ui2);
2165 inline void xxlxor( VectorSRegister d, VectorSRegister a, VectorSRegister b);
2166 inline void xxleqv( VectorSRegister d, VectorSRegister a, VectorSRegister b);
2167
2168 // VSX Extended Mnemonics
2169 inline void xxspltd( VectorSRegister d, VectorSRegister a, int x);
2170 inline void xxmrghd( VectorSRegister d, VectorSRegister a, VectorSRegister b);
2171 inline void xxmrgld( VectorSRegister d, VectorSRegister a, VectorSRegister b);
2172 inline void xxswapd( VectorSRegister d, VectorSRegister a);
2173
2174 // Vector-Scalar (VSX) instructions.
2175 inline void mtfprd( FloatRegister d, Register a);
2176 inline void mtfprwa( FloatRegister d, Register a);
2177 inline void mffprd( Register a, FloatRegister d);
2178
2179 // AES (introduced with Power 8)
2180 inline void vcipher( VectorRegister d, VectorRegister a, VectorRegister b);
2181 inline void vcipherlast( VectorRegister d, VectorRegister a, VectorRegister b);
2182 inline void vncipher( VectorRegister d, VectorRegister a, VectorRegister b);
2183 inline void vncipherlast(VectorRegister d, VectorRegister a, VectorRegister b);
2184 inline void vsbox( VectorRegister d, VectorRegister a);
2185
2186 // SHA (introduced with Power 8)
2187 inline void vshasigmad(VectorRegister d, VectorRegister a, bool st, int six);
2188 inline void vshasigmaw(VectorRegister d, VectorRegister a, bool st, int six);
2189
2190 // Vector Binary Polynomial Multiplication (introduced with Power 8)
2191 inline void vpmsumb( VectorRegister d, VectorRegister a, VectorRegister b);
2192 inline void vpmsumd( VectorRegister d, VectorRegister a, VectorRegister b);
2193 inline void vpmsumh( VectorRegister d, VectorRegister a, VectorRegister b);
2194 inline void vpmsumw( VectorRegister d, VectorRegister a, VectorRegister b);
2195
2196 // Vector Permute and Xor (introduced with Power 8)
2197 inline void vpermxor( VectorRegister d, VectorRegister a, VectorRegister b, VectorRegister c);
2198
2199 // Transactional Memory instructions (introduced with Power 8)
2200 inline void tbegin_(); // R=0
2201 inline void tbeginrot_(); // R=1 Rollback-Only Transaction
2202 inline void tend_(); // A=0
2203 inline void tendall_(); // A=1
2204 inline void tabort_();
2205 inline void tabort_(Register a);
2206 inline void tabortwc_(int t, Register a, Register b);
2207 inline void tabortwci_(int t, Register a, int si);
2208 inline void tabortdc_(int t, Register a, Register b);
2209 inline void tabortdci_(int t, Register a, int si);
2210 inline void tsuspend_(); // tsr with L=0
2211 inline void tresume_(); // tsr with L=1
2212 inline void tcheck(int f);
2213
2214 static bool is_tbegin(int x) {
2215 return TBEGIN_OPCODE == (x & (0x3f << OPCODE_SHIFT | 0x3ff << 1));
2216 }
2217
2218 // The following encoders use r0 as second operand. These instructions
2219 // read r0 as '0'.
2220 inline void lwzx( Register d, Register s2);
2221 inline void lwz( Register d, int si16);
2222 inline void lwax( Register d, Register s2);
2223 inline void lwa( Register d, int si16);
2224 inline void lwbrx(Register d, Register s2);
2225 inline void lhzx( Register d, Register s2);
2226 inline void lhz( Register d, int si16);
2227 inline void lhax( Register d, Register s2);
2228 inline void lha( Register d, int si16);
2229 inline void lhbrx(Register d, Register s2);
2230 inline void lbzx( Register d, Register s2);
2231 inline void lbz( Register d, int si16);
2232 inline void ldx( Register d, Register s2);
2233 inline void ld( Register d, int si16);
2234 inline void ldbrx(Register d, Register s2);
2235 inline void stwx( Register d, Register s2);
2236 inline void stw( Register d, int si16);
2237 inline void stwbrx( Register d, Register s2);
2238 inline void sthx( Register d, Register s2);
2239 inline void sth( Register d, int si16);
2240 inline void sthbrx( Register d, Register s2);
2241 inline void stbx( Register d, Register s2);
2242 inline void stb( Register d, int si16);
2243 inline void stdx( Register d, Register s2);
2244 inline void std( Register d, int si16);
2245 inline void stdbrx( Register d, Register s2);
2246
2247 // PPC 2, section 3.2.1 Instruction Cache Instructions
2248 inline void icbi( Register s2);
2249 // PPC 2, section 3.2.2 Data Cache Instructions
2250 //inlinevoid dcba( Register s2); // Instruction for embedded processor only.
2251 inline void dcbz( Register s2);
2252 inline void dcbst( Register s2);
2253 inline void dcbf( Register s2);
2254 // dcache read hint
2255 inline void dcbt( Register s2);
2256 inline void dcbtct( Register s2, int ct);
2257 inline void dcbtds( Register s2, int ds);
2258 // dcache write hint
2259 inline void dcbtst( Register s2);
2260 inline void dcbtstct(Register s2, int ct);
2261
2262 // Atomics: use ra0mem to disallow R0 as base.
2263 inline void lbarx_unchecked(Register d, Register b, int eh1);
2264 inline void lharx_unchecked(Register d, Register b, int eh1);
2265 inline void lwarx_unchecked(Register d, Register b, int eh1);
2266 inline void ldarx_unchecked(Register d, Register b, int eh1);
2267 inline void lqarx_unchecked(Register d, Register b, int eh1);
2268 inline void lbarx( Register d, Register b, bool hint_exclusive_access);
2269 inline void lharx( Register d, Register b, bool hint_exclusive_access);
2270 inline void lwarx( Register d, Register b, bool hint_exclusive_access);
2271 inline void ldarx( Register d, Register b, bool hint_exclusive_access);
2272 inline void lqarx( Register d, Register b, bool hint_exclusive_access);
2273 inline void stbcx_(Register s, Register b);
2274 inline void sthcx_(Register s, Register b);
2275 inline void stwcx_(Register s, Register b);
2276 inline void stdcx_(Register s, Register b);
2277 inline void stqcx_(Register s, Register b);
2278 inline void lfs( FloatRegister d, int si16);
2279 inline void lfsx( FloatRegister d, Register b);
2280 inline void lfd( FloatRegister d, int si16);
2281 inline void lfdx( FloatRegister d, Register b);
2282 inline void stfs( FloatRegister s, int si16);
2283 inline void stfsx( FloatRegister s, Register b);
2284 inline void stfd( FloatRegister s, int si16);
2285 inline void stfdx( FloatRegister s, Register b);
2286 inline void lvebx( VectorRegister d, Register s2);
2287 inline void lvehx( VectorRegister d, Register s2);
2288 inline void lvewx( VectorRegister d, Register s2);
2289 inline void lvx( VectorRegister d, Register s2);
2290 inline void lvxl( VectorRegister d, Register s2);
2291 inline void stvebx(VectorRegister d, Register s2);
2292 inline void stvehx(VectorRegister d, Register s2);
2293 inline void stvewx(VectorRegister d, Register s2);
2294 inline void stvx( VectorRegister d, Register s2);
2295 inline void stvxl( VectorRegister d, Register s2);
2296 inline void lvsl( VectorRegister d, Register s2);
2297 inline void lvsr( VectorRegister d, Register s2);
2298
2299 // Endianess specific concatenation of 2 loaded vectors.
2300 inline void load_perm(VectorRegister perm, Register addr);
2301 inline void vec_perm(VectorRegister first_dest, VectorRegister second, VectorRegister perm);
2302
2303 // RegisterOrConstant versions.
2304 // These emitters choose between the versions using two registers and
2305 // those with register and immediate, depending on the content of roc.
2306 // If the constant is not encodable as immediate, instructions to
2307 // load the constant are emitted beforehand. Store instructions need a
2308 // tmp reg if the constant is not encodable as immediate.
2309 // Size unpredictable.
2310 void ld( Register d, RegisterOrConstant roc, Register s1 = noreg);
2311 void lwa( Register d, RegisterOrConstant roc, Register s1 = noreg);
2312 void lwz( Register d, RegisterOrConstant roc, Register s1 = noreg);
2313 void lha( Register d, RegisterOrConstant roc, Register s1 = noreg);
2314 void lhz( Register d, RegisterOrConstant roc, Register s1 = noreg);
2315 void lbz( Register d, RegisterOrConstant roc, Register s1 = noreg);
2316 void std( Register d, RegisterOrConstant roc, Register s1 = noreg, Register tmp = noreg);
2317 void stw( Register d, RegisterOrConstant roc, Register s1 = noreg, Register tmp = noreg);
2318 void sth( Register d, RegisterOrConstant roc, Register s1 = noreg, Register tmp = noreg);
2319 void stb( Register d, RegisterOrConstant roc, Register s1 = noreg, Register tmp = noreg);
2320 void add( Register d, RegisterOrConstant roc, Register s1);
2321 void subf(Register d, RegisterOrConstant roc, Register s1);
2322 void cmpd(ConditionRegister d, RegisterOrConstant roc, Register s1);
2323 // Load pointer d from s1+roc.
2324 void ld_ptr(Register d, RegisterOrConstant roc, Register s1 = noreg) { ld(d, roc, s1); }
2325
2326 // Emit several instructions to load a 64 bit constant. This issues a fixed
2327 // instruction pattern so that the constant can be patched later on.
2328 enum {
2329 load_const_size = 5 * BytesPerInstWord
2330 };
2331 void load_const(Register d, long a, Register tmp = noreg);
2332 inline void load_const(Register d, void* a, Register tmp = noreg);
2333 inline void load_const(Register d, Label& L, Register tmp = noreg);
2334 inline void load_const(Register d, AddressLiteral& a, Register tmp = noreg);
2335 inline void load_const32(Register d, int i); // load signed int (patchable)
2336
2337 // Load a 64 bit constant, optimized, not identifyable.
2338 // Tmp can be used to increase ILP. Set return_simm16_rest = true to get a
2339 // 16 bit immediate offset. This is useful if the offset can be encoded in
2340 // a succeeding instruction.
2341 int load_const_optimized(Register d, long a, Register tmp = noreg, bool return_simm16_rest = false);
2342 inline int load_const_optimized(Register d, void* a, Register tmp = noreg, bool return_simm16_rest = false) {
2343 return load_const_optimized(d, (long)(unsigned long)a, tmp, return_simm16_rest);
2344 }
2345
2346 // If return_simm16_rest, the return value needs to get added afterwards.
2347 int add_const_optimized(Register d, Register s, long x, Register tmp = R0, bool return_simm16_rest = false);
2348 inline int add_const_optimized(Register d, Register s, void* a, Register tmp = R0, bool return_simm16_rest = false) {
2349 return add_const_optimized(d, s, (long)(unsigned long)a, tmp, return_simm16_rest);
2350 }
2351
2352 // If return_simm16_rest, the return value needs to get added afterwards.
2353 inline int sub_const_optimized(Register d, Register s, long x, Register tmp = R0, bool return_simm16_rest = false) {
2354 return add_const_optimized(d, s, -x, tmp, return_simm16_rest);
2355 }
2356 inline int sub_const_optimized(Register d, Register s, void* a, Register tmp = R0, bool return_simm16_rest = false) {
2357 return sub_const_optimized(d, s, (long)(unsigned long)a, tmp, return_simm16_rest);
2358 }
2359
2360 // Creation
2361 Assembler(CodeBuffer* code) : AbstractAssembler(code) {
2362 #ifdef CHECK_DELAY
2363 delay_state = no_delay;
2364 #endif
2365 }
2366
2367 // Testing
2368 #ifndef PRODUCT
2369 void test_asm();
2370 #endif
2371 };
2372
2373
2374 #endif // CPU_PPC_VM_ASSEMBLER_PPC_HPP