1 /* 2 * Copyright (c) 2002, 2016, Oracle and/or its affiliates. All rights reserved. 3 * Copyright (c) 2012, 2016 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), 379 380 STH_OPCODE = (44u << OPCODE_SHIFT), 381 STHX_OPCODE = (31u << OPCODE_SHIFT | 407u << 1), 382 STHU_OPCODE = (45u << OPCODE_SHIFT), 383 384 STB_OPCODE = (38u << OPCODE_SHIFT), 385 STBX_OPCODE = (31u << OPCODE_SHIFT | 215u << 1), 386 STBU_OPCODE = (39u << OPCODE_SHIFT), 387 388 EXTSB_OPCODE = (31u << OPCODE_SHIFT | 954u << 1), 389 EXTSH_OPCODE = (31u << OPCODE_SHIFT | 922u << 1), 390 EXTSW_OPCODE = (31u << OPCODE_SHIFT | 986u << 1), // X-FORM 391 392 // 32 bit opcode encodings 393 394 LWA_OPCODE = (58u << OPCODE_SHIFT | 2u << XO_30_31_SHIFT), // DS-FORM 395 LWAX_OPCODE = (31u << OPCODE_SHIFT | 341u << XO_21_30_SHIFT), // X-FORM 396 397 CNTLZW_OPCODE = (31u << OPCODE_SHIFT | 26u << XO_21_30_SHIFT), // X-FORM 398 399 // 64 bit opcode encodings 400 401 LD_OPCODE = (58u << OPCODE_SHIFT | 0u << XO_30_31_SHIFT), // DS-FORM 402 LDU_OPCODE = (58u << OPCODE_SHIFT | 1u << XO_30_31_SHIFT), // DS-FORM 403 LDX_OPCODE = (31u << OPCODE_SHIFT | 21u << XO_21_30_SHIFT), // X-FORM 404 405 STD_OPCODE = (62u << OPCODE_SHIFT | 0u << XO_30_31_SHIFT), // DS-FORM 406 STDU_OPCODE = (62u << OPCODE_SHIFT | 1u << XO_30_31_SHIFT), // DS-FORM 407 STDUX_OPCODE = (31u << OPCODE_SHIFT | 181u << 1), // X-FORM 408 STDX_OPCODE = (31u << OPCODE_SHIFT | 149u << XO_21_30_SHIFT), // X-FORM 409 410 RLDICR_OPCODE = (30u << OPCODE_SHIFT | 1u << XO_27_29_SHIFT), // MD-FORM 411 RLDICL_OPCODE = (30u << OPCODE_SHIFT | 0u << XO_27_29_SHIFT), // MD-FORM 412 RLDIC_OPCODE = (30u << OPCODE_SHIFT | 2u << XO_27_29_SHIFT), // MD-FORM 413 RLDIMI_OPCODE = (30u << OPCODE_SHIFT | 3u << XO_27_29_SHIFT), // MD-FORM 414 415 SRADI_OPCODE = (31u << OPCODE_SHIFT | 413u << XO_21_29_SHIFT), // XS-FORM 416 417 SLD_OPCODE = (31u << OPCODE_SHIFT | 27u << 1), // X-FORM 418 SRD_OPCODE = (31u << OPCODE_SHIFT | 539u << 1), // X-FORM 419 SRAD_OPCODE = (31u << OPCODE_SHIFT | 794u << 1), // X-FORM 420 421 MULLD_OPCODE = (31u << OPCODE_SHIFT | 233u << 1), // XO-FORM 422 MULHD_OPCODE = (31u << OPCODE_SHIFT | 73u << 1), // XO-FORM 423 MULHDU_OPCODE = (31u << OPCODE_SHIFT | 9u << 1), // XO-FORM 424 DIVD_OPCODE = (31u << OPCODE_SHIFT | 489u << 1), // XO-FORM 425 426 CNTLZD_OPCODE = (31u << OPCODE_SHIFT | 58u << XO_21_30_SHIFT), // X-FORM 427 NAND_OPCODE = (31u << OPCODE_SHIFT | 476u << XO_21_30_SHIFT), // X-FORM 428 NOR_OPCODE = (31u << OPCODE_SHIFT | 124u << XO_21_30_SHIFT), // X-FORM 429 430 431 // opcodes only used for floating arithmetic 432 FADD_OPCODE = (63u << OPCODE_SHIFT | 21u << 1), 433 FADDS_OPCODE = (59u << OPCODE_SHIFT | 21u << 1), 434 FCMPU_OPCODE = (63u << OPCODE_SHIFT | 00u << 1), 435 FDIV_OPCODE = (63u << OPCODE_SHIFT | 18u << 1), 436 FDIVS_OPCODE = (59u << OPCODE_SHIFT | 18u << 1), 437 FMR_OPCODE = (63u << OPCODE_SHIFT | 72u << 1), 438 // These are special Power6 opcodes, reused for "lfdepx" and "stfdepx" 439 // on Power7. Do not use. 440 // MFFGPR_OPCODE = (31u << OPCODE_SHIFT | 607u << 1), 441 // MFTGPR_OPCODE = (31u << OPCODE_SHIFT | 735u << 1), 442 CMPB_OPCODE = (31u << OPCODE_SHIFT | 508 << 1), 443 POPCNTB_OPCODE = (31u << OPCODE_SHIFT | 122 << 1), 444 POPCNTW_OPCODE = (31u << OPCODE_SHIFT | 378 << 1), 445 POPCNTD_OPCODE = (31u << OPCODE_SHIFT | 506 << 1), 446 FABS_OPCODE = (63u << OPCODE_SHIFT | 264u << 1), 447 FNABS_OPCODE = (63u << OPCODE_SHIFT | 136u << 1), 448 FMUL_OPCODE = (63u << OPCODE_SHIFT | 25u << 1), 449 FMULS_OPCODE = (59u << OPCODE_SHIFT | 25u << 1), 450 FNEG_OPCODE = (63u << OPCODE_SHIFT | 40u << 1), 451 FSUB_OPCODE = (63u << OPCODE_SHIFT | 20u << 1), 452 FSUBS_OPCODE = (59u << OPCODE_SHIFT | 20u << 1), 453 454 // PPC64-internal FPU conversion opcodes 455 FCFID_OPCODE = (63u << OPCODE_SHIFT | 846u << 1), 456 FCFIDS_OPCODE = (59u << OPCODE_SHIFT | 846u << 1), 457 FCTID_OPCODE = (63u << OPCODE_SHIFT | 814u << 1), 458 FCTIDZ_OPCODE = (63u << OPCODE_SHIFT | 815u << 1), 459 FCTIW_OPCODE = (63u << OPCODE_SHIFT | 14u << 1), 460 FCTIWZ_OPCODE = (63u << OPCODE_SHIFT | 15u << 1), 461 FRSP_OPCODE = (63u << OPCODE_SHIFT | 12u << 1), 462 463 // Fused multiply-accumulate instructions. 464 FMADD_OPCODE = (63u << OPCODE_SHIFT | 29u << 1), 465 FMADDS_OPCODE = (59u << OPCODE_SHIFT | 29u << 1), 466 FMSUB_OPCODE = (63u << OPCODE_SHIFT | 28u << 1), 467 FMSUBS_OPCODE = (59u << OPCODE_SHIFT | 28u << 1), 468 FNMADD_OPCODE = (63u << OPCODE_SHIFT | 31u << 1), 469 FNMADDS_OPCODE = (59u << OPCODE_SHIFT | 31u << 1), 470 FNMSUB_OPCODE = (63u << OPCODE_SHIFT | 30u << 1), 471 FNMSUBS_OPCODE = (59u << OPCODE_SHIFT | 30u << 1), 472 473 LFD_OPCODE = (50u << OPCODE_SHIFT | 00u << 1), 474 LFDU_OPCODE = (51u << OPCODE_SHIFT | 00u << 1), 475 LFDX_OPCODE = (31u << OPCODE_SHIFT | 599u << 1), 476 LFS_OPCODE = (48u << OPCODE_SHIFT | 00u << 1), 477 LFSU_OPCODE = (49u << OPCODE_SHIFT | 00u << 1), 478 LFSX_OPCODE = (31u << OPCODE_SHIFT | 535u << 1), 479 480 STFD_OPCODE = (54u << OPCODE_SHIFT | 00u << 1), 481 STFDU_OPCODE = (55u << OPCODE_SHIFT | 00u << 1), 482 STFDX_OPCODE = (31u << OPCODE_SHIFT | 727u << 1), 483 STFS_OPCODE = (52u << OPCODE_SHIFT | 00u << 1), 484 STFSU_OPCODE = (53u << OPCODE_SHIFT | 00u << 1), 485 STFSX_OPCODE = (31u << OPCODE_SHIFT | 663u << 1), 486 487 FSQRT_OPCODE = (63u << OPCODE_SHIFT | 22u << 1), // A-FORM 488 FSQRTS_OPCODE = (59u << OPCODE_SHIFT | 22u << 1), // A-FORM 489 490 // Vector instruction support for >= Power6 491 // Vector Storage Access 492 LVEBX_OPCODE = (31u << OPCODE_SHIFT | 7u << 1), 493 LVEHX_OPCODE = (31u << OPCODE_SHIFT | 39u << 1), 494 LVEWX_OPCODE = (31u << OPCODE_SHIFT | 71u << 1), 495 LVX_OPCODE = (31u << OPCODE_SHIFT | 103u << 1), 496 LVXL_OPCODE = (31u << OPCODE_SHIFT | 359u << 1), 497 STVEBX_OPCODE = (31u << OPCODE_SHIFT | 135u << 1), 498 STVEHX_OPCODE = (31u << OPCODE_SHIFT | 167u << 1), 499 STVEWX_OPCODE = (31u << OPCODE_SHIFT | 199u << 1), 500 STVX_OPCODE = (31u << OPCODE_SHIFT | 231u << 1), 501 STVXL_OPCODE = (31u << OPCODE_SHIFT | 487u << 1), 502 LVSL_OPCODE = (31u << OPCODE_SHIFT | 6u << 1), 503 LVSR_OPCODE = (31u << OPCODE_SHIFT | 38u << 1), 504 505 // Vector-Scalar (VSX) instruction support. 506 LXVD2X_OPCODE = (31u << OPCODE_SHIFT | 844u << 1), 507 STXVD2X_OPCODE = (31u << OPCODE_SHIFT | 972u << 1), 508 MTVSRD_OPCODE = (31u << OPCODE_SHIFT | 179u << 1), 509 MFVSRD_OPCODE = (31u << OPCODE_SHIFT | 51u << 1), 510 511 // Vector Permute and Formatting 512 VPKPX_OPCODE = (4u << OPCODE_SHIFT | 782u ), 513 VPKSHSS_OPCODE = (4u << OPCODE_SHIFT | 398u ), 514 VPKSWSS_OPCODE = (4u << OPCODE_SHIFT | 462u ), 515 VPKSHUS_OPCODE = (4u << OPCODE_SHIFT | 270u ), 516 VPKSWUS_OPCODE = (4u << OPCODE_SHIFT | 334u ), 517 VPKUHUM_OPCODE = (4u << OPCODE_SHIFT | 14u ), 518 VPKUWUM_OPCODE = (4u << OPCODE_SHIFT | 78u ), 519 VPKUHUS_OPCODE = (4u << OPCODE_SHIFT | 142u ), 520 VPKUWUS_OPCODE = (4u << OPCODE_SHIFT | 206u ), 521 VUPKHPX_OPCODE = (4u << OPCODE_SHIFT | 846u ), 522 VUPKHSB_OPCODE = (4u << OPCODE_SHIFT | 526u ), 523 VUPKHSH_OPCODE = (4u << OPCODE_SHIFT | 590u ), 524 VUPKLPX_OPCODE = (4u << OPCODE_SHIFT | 974u ), 525 VUPKLSB_OPCODE = (4u << OPCODE_SHIFT | 654u ), 526 VUPKLSH_OPCODE = (4u << OPCODE_SHIFT | 718u ), 527 528 VMRGHB_OPCODE = (4u << OPCODE_SHIFT | 12u ), 529 VMRGHW_OPCODE = (4u << OPCODE_SHIFT | 140u ), 530 VMRGHH_OPCODE = (4u << OPCODE_SHIFT | 76u ), 531 VMRGLB_OPCODE = (4u << OPCODE_SHIFT | 268u ), 532 VMRGLW_OPCODE = (4u << OPCODE_SHIFT | 396u ), 533 VMRGLH_OPCODE = (4u << OPCODE_SHIFT | 332u ), 534 535 VSPLT_OPCODE = (4u << OPCODE_SHIFT | 524u ), 536 VSPLTH_OPCODE = (4u << OPCODE_SHIFT | 588u ), 537 VSPLTW_OPCODE = (4u << OPCODE_SHIFT | 652u ), 538 VSPLTISB_OPCODE= (4u << OPCODE_SHIFT | 780u ), 539 VSPLTISH_OPCODE= (4u << OPCODE_SHIFT | 844u ), 540 VSPLTISW_OPCODE= (4u << OPCODE_SHIFT | 908u ), 541 542 VPERM_OPCODE = (4u << OPCODE_SHIFT | 43u ), 543 VSEL_OPCODE = (4u << OPCODE_SHIFT | 42u ), 544 545 VSL_OPCODE = (4u << OPCODE_SHIFT | 452u ), 546 VSLDOI_OPCODE = (4u << OPCODE_SHIFT | 44u ), 547 VSLO_OPCODE = (4u << OPCODE_SHIFT | 1036u ), 548 VSR_OPCODE = (4u << OPCODE_SHIFT | 708u ), 549 VSRO_OPCODE = (4u << OPCODE_SHIFT | 1100u ), 550 551 // Vector Integer 552 VADDCUW_OPCODE = (4u << OPCODE_SHIFT | 384u ), 553 VADDSHS_OPCODE = (4u << OPCODE_SHIFT | 832u ), 554 VADDSBS_OPCODE = (4u << OPCODE_SHIFT | 768u ), 555 VADDSWS_OPCODE = (4u << OPCODE_SHIFT | 896u ), 556 VADDUBM_OPCODE = (4u << OPCODE_SHIFT | 0u ), 557 VADDUWM_OPCODE = (4u << OPCODE_SHIFT | 128u ), 558 VADDUHM_OPCODE = (4u << OPCODE_SHIFT | 64u ), 559 VADDUBS_OPCODE = (4u << OPCODE_SHIFT | 512u ), 560 VADDUWS_OPCODE = (4u << OPCODE_SHIFT | 640u ), 561 VADDUHS_OPCODE = (4u << OPCODE_SHIFT | 576u ), 562 VSUBCUW_OPCODE = (4u << OPCODE_SHIFT | 1408u ), 563 VSUBSHS_OPCODE = (4u << OPCODE_SHIFT | 1856u ), 564 VSUBSBS_OPCODE = (4u << OPCODE_SHIFT | 1792u ), 565 VSUBSWS_OPCODE = (4u << OPCODE_SHIFT | 1920u ), 566 VSUBUBM_OPCODE = (4u << OPCODE_SHIFT | 1024u ), 567 VSUBUWM_OPCODE = (4u << OPCODE_SHIFT | 1152u ), 568 VSUBUHM_OPCODE = (4u << OPCODE_SHIFT | 1088u ), 569 VSUBUBS_OPCODE = (4u << OPCODE_SHIFT | 1536u ), 570 VSUBUWS_OPCODE = (4u << OPCODE_SHIFT | 1664u ), 571 VSUBUHS_OPCODE = (4u << OPCODE_SHIFT | 1600u ), 572 573 VMULESB_OPCODE = (4u << OPCODE_SHIFT | 776u ), 574 VMULEUB_OPCODE = (4u << OPCODE_SHIFT | 520u ), 575 VMULESH_OPCODE = (4u << OPCODE_SHIFT | 840u ), 576 VMULEUH_OPCODE = (4u << OPCODE_SHIFT | 584u ), 577 VMULOSB_OPCODE = (4u << OPCODE_SHIFT | 264u ), 578 VMULOUB_OPCODE = (4u << OPCODE_SHIFT | 8u ), 579 VMULOSH_OPCODE = (4u << OPCODE_SHIFT | 328u ), 580 VMULOUH_OPCODE = (4u << OPCODE_SHIFT | 72u ), 581 VMHADDSHS_OPCODE=(4u << OPCODE_SHIFT | 32u ), 582 VMHRADDSHS_OPCODE=(4u << OPCODE_SHIFT | 33u ), 583 VMLADDUHM_OPCODE=(4u << OPCODE_SHIFT | 34u ), 584 VMSUBUHM_OPCODE= (4u << OPCODE_SHIFT | 36u ), 585 VMSUMMBM_OPCODE= (4u << OPCODE_SHIFT | 37u ), 586 VMSUMSHM_OPCODE= (4u << OPCODE_SHIFT | 40u ), 587 VMSUMSHS_OPCODE= (4u << OPCODE_SHIFT | 41u ), 588 VMSUMUHM_OPCODE= (4u << OPCODE_SHIFT | 38u ), 589 VMSUMUHS_OPCODE= (4u << OPCODE_SHIFT | 39u ), 590 591 VSUMSWS_OPCODE = (4u << OPCODE_SHIFT | 1928u ), 592 VSUM2SWS_OPCODE= (4u << OPCODE_SHIFT | 1672u ), 593 VSUM4SBS_OPCODE= (4u << OPCODE_SHIFT | 1800u ), 594 VSUM4UBS_OPCODE= (4u << OPCODE_SHIFT | 1544u ), 595 VSUM4SHS_OPCODE= (4u << OPCODE_SHIFT | 1608u ), 596 597 VAVGSB_OPCODE = (4u << OPCODE_SHIFT | 1282u ), 598 VAVGSW_OPCODE = (4u << OPCODE_SHIFT | 1410u ), 599 VAVGSH_OPCODE = (4u << OPCODE_SHIFT | 1346u ), 600 VAVGUB_OPCODE = (4u << OPCODE_SHIFT | 1026u ), 601 VAVGUW_OPCODE = (4u << OPCODE_SHIFT | 1154u ), 602 VAVGUH_OPCODE = (4u << OPCODE_SHIFT | 1090u ), 603 604 VMAXSB_OPCODE = (4u << OPCODE_SHIFT | 258u ), 605 VMAXSW_OPCODE = (4u << OPCODE_SHIFT | 386u ), 606 VMAXSH_OPCODE = (4u << OPCODE_SHIFT | 322u ), 607 VMAXUB_OPCODE = (4u << OPCODE_SHIFT | 2u ), 608 VMAXUW_OPCODE = (4u << OPCODE_SHIFT | 130u ), 609 VMAXUH_OPCODE = (4u << OPCODE_SHIFT | 66u ), 610 VMINSB_OPCODE = (4u << OPCODE_SHIFT | 770u ), 611 VMINSW_OPCODE = (4u << OPCODE_SHIFT | 898u ), 612 VMINSH_OPCODE = (4u << OPCODE_SHIFT | 834u ), 613 VMINUB_OPCODE = (4u << OPCODE_SHIFT | 514u ), 614 VMINUW_OPCODE = (4u << OPCODE_SHIFT | 642u ), 615 VMINUH_OPCODE = (4u << OPCODE_SHIFT | 578u ), 616 617 VCMPEQUB_OPCODE= (4u << OPCODE_SHIFT | 6u ), 618 VCMPEQUH_OPCODE= (4u << OPCODE_SHIFT | 70u ), 619 VCMPEQUW_OPCODE= (4u << OPCODE_SHIFT | 134u ), 620 VCMPGTSH_OPCODE= (4u << OPCODE_SHIFT | 838u ), 621 VCMPGTSB_OPCODE= (4u << OPCODE_SHIFT | 774u ), 622 VCMPGTSW_OPCODE= (4u << OPCODE_SHIFT | 902u ), 623 VCMPGTUB_OPCODE= (4u << OPCODE_SHIFT | 518u ), 624 VCMPGTUH_OPCODE= (4u << OPCODE_SHIFT | 582u ), 625 VCMPGTUW_OPCODE= (4u << OPCODE_SHIFT | 646u ), 626 627 VAND_OPCODE = (4u << OPCODE_SHIFT | 1028u ), 628 VANDC_OPCODE = (4u << OPCODE_SHIFT | 1092u ), 629 VNOR_OPCODE = (4u << OPCODE_SHIFT | 1284u ), 630 VOR_OPCODE = (4u << OPCODE_SHIFT | 1156u ), 631 VXOR_OPCODE = (4u << OPCODE_SHIFT | 1220u ), 632 VRLD_OPCODE = (4u << OPCODE_SHIFT | 196u ), 633 VRLB_OPCODE = (4u << OPCODE_SHIFT | 4u ), 634 VRLW_OPCODE = (4u << OPCODE_SHIFT | 132u ), 635 VRLH_OPCODE = (4u << OPCODE_SHIFT | 68u ), 636 VSLB_OPCODE = (4u << OPCODE_SHIFT | 260u ), 637 VSKW_OPCODE = (4u << OPCODE_SHIFT | 388u ), 638 VSLH_OPCODE = (4u << OPCODE_SHIFT | 324u ), 639 VSRB_OPCODE = (4u << OPCODE_SHIFT | 516u ), 640 VSRW_OPCODE = (4u << OPCODE_SHIFT | 644u ), 641 VSRH_OPCODE = (4u << OPCODE_SHIFT | 580u ), 642 VSRAB_OPCODE = (4u << OPCODE_SHIFT | 772u ), 643 VSRAW_OPCODE = (4u << OPCODE_SHIFT | 900u ), 644 VSRAH_OPCODE = (4u << OPCODE_SHIFT | 836u ), 645 646 // Vector Floating-Point 647 // not implemented yet 648 649 // Vector Status and Control 650 MTVSCR_OPCODE = (4u << OPCODE_SHIFT | 1604u ), 651 MFVSCR_OPCODE = (4u << OPCODE_SHIFT | 1540u ), 652 653 // AES (introduced with Power 8) 654 VCIPHER_OPCODE = (4u << OPCODE_SHIFT | 1288u), 655 VCIPHERLAST_OPCODE = (4u << OPCODE_SHIFT | 1289u), 656 VNCIPHER_OPCODE = (4u << OPCODE_SHIFT | 1352u), 657 VNCIPHERLAST_OPCODE = (4u << OPCODE_SHIFT | 1353u), 658 VSBOX_OPCODE = (4u << OPCODE_SHIFT | 1480u), 659 660 // SHA (introduced with Power 8) 661 VSHASIGMAD_OPCODE = (4u << OPCODE_SHIFT | 1730u), 662 VSHASIGMAW_OPCODE = (4u << OPCODE_SHIFT | 1666u), 663 664 // Vector Binary Polynomial Multiplication (introduced with Power 8) 665 VPMSUMB_OPCODE = (4u << OPCODE_SHIFT | 1032u), 666 VPMSUMD_OPCODE = (4u << OPCODE_SHIFT | 1224u), 667 VPMSUMH_OPCODE = (4u << OPCODE_SHIFT | 1096u), 668 VPMSUMW_OPCODE = (4u << OPCODE_SHIFT | 1160u), 669 670 // Vector Permute and Xor (introduced with Power 8) 671 VPERMXOR_OPCODE = (4u << OPCODE_SHIFT | 45u), 672 673 // Transactional Memory instructions (introduced with Power 8) 674 TBEGIN_OPCODE = (31u << OPCODE_SHIFT | 654u << 1), 675 TEND_OPCODE = (31u << OPCODE_SHIFT | 686u << 1), 676 TABORT_OPCODE = (31u << OPCODE_SHIFT | 910u << 1), 677 TABORTWC_OPCODE = (31u << OPCODE_SHIFT | 782u << 1), 678 TABORTWCI_OPCODE = (31u << OPCODE_SHIFT | 846u << 1), 679 TABORTDC_OPCODE = (31u << OPCODE_SHIFT | 814u << 1), 680 TABORTDCI_OPCODE = (31u << OPCODE_SHIFT | 878u << 1), 681 TSR_OPCODE = (31u << OPCODE_SHIFT | 750u << 1), 682 TCHECK_OPCODE = (31u << OPCODE_SHIFT | 718u << 1), 683 684 // Icache and dcache related instructions 685 DCBA_OPCODE = (31u << OPCODE_SHIFT | 758u << 1), 686 DCBZ_OPCODE = (31u << OPCODE_SHIFT | 1014u << 1), 687 DCBST_OPCODE = (31u << OPCODE_SHIFT | 54u << 1), 688 DCBF_OPCODE = (31u << OPCODE_SHIFT | 86u << 1), 689 690 DCBT_OPCODE = (31u << OPCODE_SHIFT | 278u << 1), 691 DCBTST_OPCODE = (31u << OPCODE_SHIFT | 246u << 1), 692 ICBI_OPCODE = (31u << OPCODE_SHIFT | 982u << 1), 693 694 // Instruction synchronization 695 ISYNC_OPCODE = (19u << OPCODE_SHIFT | 150u << 1), 696 // Memory barriers 697 SYNC_OPCODE = (31u << OPCODE_SHIFT | 598u << 1), 698 EIEIO_OPCODE = (31u << OPCODE_SHIFT | 854u << 1), 699 700 // Wait instructions for polling. 701 WAIT_OPCODE = (31u << OPCODE_SHIFT | 62u << 1), 702 703 // Trap instructions 704 TDI_OPCODE = (2u << OPCODE_SHIFT), 705 TWI_OPCODE = (3u << OPCODE_SHIFT), 706 TD_OPCODE = (31u << OPCODE_SHIFT | 68u << 1), 707 TW_OPCODE = (31u << OPCODE_SHIFT | 4u << 1), 708 709 // Atomics. 710 LBARX_OPCODE = (31u << OPCODE_SHIFT | 52u << 1), 711 LHARX_OPCODE = (31u << OPCODE_SHIFT | 116u << 1), 712 LWARX_OPCODE = (31u << OPCODE_SHIFT | 20u << 1), 713 LDARX_OPCODE = (31u << OPCODE_SHIFT | 84u << 1), 714 LQARX_OPCODE = (31u << OPCODE_SHIFT | 276u << 1), 715 STBCX_OPCODE = (31u << OPCODE_SHIFT | 694u << 1), 716 STHCX_OPCODE = (31u << OPCODE_SHIFT | 726u << 1), 717 STWCX_OPCODE = (31u << OPCODE_SHIFT | 150u << 1), 718 STDCX_OPCODE = (31u << OPCODE_SHIFT | 214u << 1), 719 STQCX_OPCODE = (31u << OPCODE_SHIFT | 182u << 1) 720 721 }; 722 723 // Trap instructions TO bits 724 enum trap_to_bits { 725 // single bits 726 traptoLessThanSigned = 1 << 4, // 0, left end 727 traptoGreaterThanSigned = 1 << 3, 728 traptoEqual = 1 << 2, 729 traptoLessThanUnsigned = 1 << 1, 730 traptoGreaterThanUnsigned = 1 << 0, // 4, right end 731 732 // compound ones 733 traptoUnconditional = (traptoLessThanSigned | 734 traptoGreaterThanSigned | 735 traptoEqual | 736 traptoLessThanUnsigned | 737 traptoGreaterThanUnsigned) 738 }; 739 740 // Branch hints BH field 741 enum branch_hint_bh { 742 // bclr cases: 743 bhintbhBCLRisReturn = 0, 744 bhintbhBCLRisNotReturnButSame = 1, 745 bhintbhBCLRisNotPredictable = 3, 746 747 // bcctr cases: 748 bhintbhBCCTRisNotReturnButSame = 0, 749 bhintbhBCCTRisNotPredictable = 3 750 }; 751 752 // Branch prediction hints AT field 753 enum branch_hint_at { 754 bhintatNoHint = 0, // at=00 755 bhintatIsNotTaken = 2, // at=10 756 bhintatIsTaken = 3 // at=11 757 }; 758 759 // Branch prediction hints 760 enum branch_hint_concept { 761 // Use the same encoding as branch_hint_at to simply code. 762 bhintNoHint = bhintatNoHint, 763 bhintIsNotTaken = bhintatIsNotTaken, 764 bhintIsTaken = bhintatIsTaken 765 }; 766 767 // Used in BO field of branch instruction. 768 enum branch_condition { 769 bcondCRbiIs0 = 4, // bo=001at 770 bcondCRbiIs1 = 12, // bo=011at 771 bcondAlways = 20 // bo=10100 772 }; 773 774 // Branch condition with combined prediction hints. 775 enum branch_condition_with_hint { 776 bcondCRbiIs0_bhintNoHint = bcondCRbiIs0 | bhintatNoHint, 777 bcondCRbiIs0_bhintIsNotTaken = bcondCRbiIs0 | bhintatIsNotTaken, 778 bcondCRbiIs0_bhintIsTaken = bcondCRbiIs0 | bhintatIsTaken, 779 bcondCRbiIs1_bhintNoHint = bcondCRbiIs1 | bhintatNoHint, 780 bcondCRbiIs1_bhintIsNotTaken = bcondCRbiIs1 | bhintatIsNotTaken, 781 bcondCRbiIs1_bhintIsTaken = bcondCRbiIs1 | bhintatIsTaken, 782 }; 783 784 // Elemental Memory Barriers (>=Power 8) 785 enum Elemental_Membar_mask_bits { 786 StoreStore = 1 << 0, 787 StoreLoad = 1 << 1, 788 LoadStore = 1 << 2, 789 LoadLoad = 1 << 3 790 }; 791 792 // Branch prediction hints. 793 inline static int add_bhint_to_boint(const int bhint, const int boint) { 794 switch (boint) { 795 case bcondCRbiIs0: 796 case bcondCRbiIs1: 797 // branch_hint and branch_hint_at have same encodings 798 assert( (int)bhintNoHint == (int)bhintatNoHint 799 && (int)bhintIsNotTaken == (int)bhintatIsNotTaken 800 && (int)bhintIsTaken == (int)bhintatIsTaken, 801 "wrong encodings"); 802 assert((bhint & 0x03) == bhint, "wrong encodings"); 803 return (boint & ~0x03) | bhint; 804 case bcondAlways: 805 // no branch_hint 806 return boint; 807 default: 808 ShouldNotReachHere(); 809 return 0; 810 } 811 } 812 813 // Extract bcond from boint. 814 inline static int inv_boint_bcond(const int boint) { 815 int r_bcond = boint & ~0x03; 816 assert(r_bcond == bcondCRbiIs0 || 817 r_bcond == bcondCRbiIs1 || 818 r_bcond == bcondAlways, 819 "bad branch condition"); 820 return r_bcond; 821 } 822 823 // Extract bhint from boint. 824 inline static int inv_boint_bhint(const int boint) { 825 int r_bhint = boint & 0x03; 826 assert(r_bhint == bhintatNoHint || 827 r_bhint == bhintatIsNotTaken || 828 r_bhint == bhintatIsTaken, 829 "bad branch hint"); 830 return r_bhint; 831 } 832 833 // Calculate opposite of given bcond. 834 inline static int opposite_bcond(const int bcond) { 835 switch (bcond) { 836 case bcondCRbiIs0: 837 return bcondCRbiIs1; 838 case bcondCRbiIs1: 839 return bcondCRbiIs0; 840 default: 841 ShouldNotReachHere(); 842 return 0; 843 } 844 } 845 846 // Calculate opposite of given bhint. 847 inline static int opposite_bhint(const int bhint) { 848 switch (bhint) { 849 case bhintatNoHint: 850 return bhintatNoHint; 851 case bhintatIsNotTaken: 852 return bhintatIsTaken; 853 case bhintatIsTaken: 854 return bhintatIsNotTaken; 855 default: 856 ShouldNotReachHere(); 857 return 0; 858 } 859 } 860 861 // PPC branch instructions 862 enum ppcops { 863 b_op = 18, 864 bc_op = 16, 865 bcr_op = 19 866 }; 867 868 enum Condition { 869 negative = 0, 870 less = 0, 871 positive = 1, 872 greater = 1, 873 zero = 2, 874 equal = 2, 875 summary_overflow = 3, 876 }; 877 878 public: 879 // Helper functions for groups of instructions 880 881 enum Predict { pt = 1, pn = 0 }; // pt = predict taken 882 883 // Instruction must start at passed address. 884 static int instr_len(unsigned char *instr) { return BytesPerInstWord; } 885 886 // longest instructions 887 static int instr_maxlen() { return BytesPerInstWord; } 888 889 // Test if x is within signed immediate range for nbits. 890 static bool is_simm(int x, unsigned int nbits) { 891 assert(0 < nbits && nbits < 32, "out of bounds"); 892 const int min = -(((int)1) << nbits-1); 893 const int maxplus1 = (((int)1) << nbits-1); 894 return min <= x && x < maxplus1; 895 } 896 897 static bool is_simm(jlong x, unsigned int nbits) { 898 assert(0 < nbits && nbits < 64, "out of bounds"); 899 const jlong min = -(((jlong)1) << nbits-1); 900 const jlong maxplus1 = (((jlong)1) << nbits-1); 901 return min <= x && x < maxplus1; 902 } 903 904 // Test if x is within unsigned immediate range for nbits. 905 static bool is_uimm(int x, unsigned int nbits) { 906 assert(0 < nbits && nbits < 32, "out of bounds"); 907 const unsigned int maxplus1 = (((unsigned int)1) << nbits); 908 return (unsigned int)x < maxplus1; 909 } 910 911 static bool is_uimm(jlong x, unsigned int nbits) { 912 assert(0 < nbits && nbits < 64, "out of bounds"); 913 const julong maxplus1 = (((julong)1) << nbits); 914 return (julong)x < maxplus1; 915 } 916 917 protected: 918 // helpers 919 920 // X is supposed to fit in a field "nbits" wide 921 // and be sign-extended. Check the range. 922 static void assert_signed_range(intptr_t x, int nbits) { 923 assert(nbits == 32 || (-(1 << nbits-1) <= x && x < (1 << nbits-1)), 924 "value out of range"); 925 } 926 927 static void assert_signed_word_disp_range(intptr_t x, int nbits) { 928 assert((x & 3) == 0, "not word aligned"); 929 assert_signed_range(x, nbits + 2); 930 } 931 932 static void assert_unsigned_const(int x, int nbits) { 933 assert(juint(x) < juint(1 << nbits), "unsigned constant out of range"); 934 } 935 936 static int fmask(juint hi_bit, juint lo_bit) { 937 assert(hi_bit >= lo_bit && hi_bit < 32, "bad bits"); 938 return (1 << ( hi_bit-lo_bit + 1 )) - 1; 939 } 940 941 // inverse of u_field 942 static int inv_u_field(int x, int hi_bit, int lo_bit) { 943 juint r = juint(x) >> lo_bit; 944 r &= fmask(hi_bit, lo_bit); 945 return int(r); 946 } 947 948 // signed version: extract from field and sign-extend 949 static int inv_s_field_ppc(int x, int hi_bit, int lo_bit) { 950 x = x << (31-hi_bit); 951 x = x >> (31-hi_bit+lo_bit); 952 return x; 953 } 954 955 static int u_field(int x, int hi_bit, int lo_bit) { 956 assert((x & ~fmask(hi_bit, lo_bit)) == 0, "value out of range"); 957 int r = x << lo_bit; 958 assert(inv_u_field(r, hi_bit, lo_bit) == x, "just checking"); 959 return r; 960 } 961 962 // Same as u_field for signed values 963 static int s_field(int x, int hi_bit, int lo_bit) { 964 int nbits = hi_bit - lo_bit + 1; 965 assert(nbits == 32 || (-(1 << nbits-1) <= x && x < (1 << nbits-1)), 966 "value out of range"); 967 x &= fmask(hi_bit, lo_bit); 968 int r = x << lo_bit; 969 return r; 970 } 971 972 // inv_op for ppc instructions 973 static int inv_op_ppc(int x) { return inv_u_field(x, 31, 26); } 974 975 // Determine target address from li, bd field of branch instruction. 976 static intptr_t inv_li_field(int x) { 977 intptr_t r = inv_s_field_ppc(x, 25, 2); 978 r = (r << 2); 979 return r; 980 } 981 static intptr_t inv_bd_field(int x, intptr_t pos) { 982 intptr_t r = inv_s_field_ppc(x, 15, 2); 983 r = (r << 2) + pos; 984 return r; 985 } 986 987 #define inv_opp_u_field(x, hi_bit, lo_bit) inv_u_field(x, 31-(lo_bit), 31-(hi_bit)) 988 #define inv_opp_s_field(x, hi_bit, lo_bit) inv_s_field_ppc(x, 31-(lo_bit), 31-(hi_bit)) 989 // Extract instruction fields from instruction words. 990 public: 991 static int inv_ra_field(int x) { return inv_opp_u_field(x, 15, 11); } 992 static int inv_rb_field(int x) { return inv_opp_u_field(x, 20, 16); } 993 static int inv_rt_field(int x) { return inv_opp_u_field(x, 10, 6); } 994 static int inv_rta_field(int x) { return inv_opp_u_field(x, 15, 11); } 995 static int inv_rs_field(int x) { return inv_opp_u_field(x, 10, 6); } 996 // Ds uses opp_s_field(x, 31, 16), but lowest 2 bits must be 0. 997 // Inv_ds_field uses range (x, 29, 16) but shifts by 2 to ensure that lowest bits are 0. 998 static int inv_ds_field(int x) { return inv_opp_s_field(x, 29, 16) << 2; } 999 static int inv_d1_field(int x) { return inv_opp_s_field(x, 31, 16); } 1000 static int inv_si_field(int x) { return inv_opp_s_field(x, 31, 16); } 1001 static int inv_to_field(int x) { return inv_opp_u_field(x, 10, 6); } 1002 static int inv_lk_field(int x) { return inv_opp_u_field(x, 31, 31); } 1003 static int inv_bo_field(int x) { return inv_opp_u_field(x, 10, 6); } 1004 static int inv_bi_field(int x) { return inv_opp_u_field(x, 15, 11); } 1005 1006 #define opp_u_field(x, hi_bit, lo_bit) u_field(x, 31-(lo_bit), 31-(hi_bit)) 1007 #define opp_s_field(x, hi_bit, lo_bit) s_field(x, 31-(lo_bit), 31-(hi_bit)) 1008 1009 // instruction fields 1010 static int aa( int x) { return opp_u_field(x, 30, 30); } 1011 static int ba( int x) { return opp_u_field(x, 15, 11); } 1012 static int bb( int x) { return opp_u_field(x, 20, 16); } 1013 static int bc( int x) { return opp_u_field(x, 25, 21); } 1014 static int bd( int x) { return opp_s_field(x, 29, 16); } 1015 static int bf( ConditionRegister cr) { return bf(cr->encoding()); } 1016 static int bf( int x) { return opp_u_field(x, 8, 6); } 1017 static int bfa(ConditionRegister cr) { return bfa(cr->encoding()); } 1018 static int bfa( int x) { return opp_u_field(x, 13, 11); } 1019 static int bh( int x) { return opp_u_field(x, 20, 19); } 1020 static int bi( int x) { return opp_u_field(x, 15, 11); } 1021 static int bi0(ConditionRegister cr, Condition c) { return (cr->encoding() << 2) | c; } 1022 static int bo( int x) { return opp_u_field(x, 10, 6); } 1023 static int bt( int x) { return opp_u_field(x, 10, 6); } 1024 static int d1( int x) { return opp_s_field(x, 31, 16); } 1025 static int ds( int x) { assert((x & 0x3) == 0, "unaligned offset"); return opp_s_field(x, 31, 16); } 1026 static int eh( int x) { return opp_u_field(x, 31, 31); } 1027 static int flm( int x) { return opp_u_field(x, 14, 7); } 1028 static int fra( FloatRegister r) { return fra(r->encoding());} 1029 static int frb( FloatRegister r) { return frb(r->encoding());} 1030 static int frc( FloatRegister r) { return frc(r->encoding());} 1031 static int frs( FloatRegister r) { return frs(r->encoding());} 1032 static int frt( FloatRegister r) { return frt(r->encoding());} 1033 static int fra( int x) { return opp_u_field(x, 15, 11); } 1034 static int frb( int x) { return opp_u_field(x, 20, 16); } 1035 static int frc( int x) { return opp_u_field(x, 25, 21); } 1036 static int frs( int x) { return opp_u_field(x, 10, 6); } 1037 static int frt( int x) { return opp_u_field(x, 10, 6); } 1038 static int fxm( int x) { return opp_u_field(x, 19, 12); } 1039 static int l10( int x) { return opp_u_field(x, 10, 10); } 1040 static int l15( int x) { return opp_u_field(x, 15, 15); } 1041 static int l910( int x) { return opp_u_field(x, 10, 9); } 1042 static int e1215( int x) { return opp_u_field(x, 15, 12); } 1043 static int lev( int x) { return opp_u_field(x, 26, 20); } 1044 static int li( int x) { return opp_s_field(x, 29, 6); } 1045 static int lk( int x) { return opp_u_field(x, 31, 31); } 1046 static int mb2125( int x) { return opp_u_field(x, 25, 21); } 1047 static int me2630( int x) { return opp_u_field(x, 30, 26); } 1048 static int mb2126( int x) { return opp_u_field(((x & 0x1f) << 1) | ((x & 0x20) >> 5), 26, 21); } 1049 static int me2126( int x) { return mb2126(x); } 1050 static int nb( int x) { return opp_u_field(x, 20, 16); } 1051 //static int opcd( int x) { return opp_u_field(x, 5, 0); } // is contained in our opcodes 1052 static int oe( int x) { return opp_u_field(x, 21, 21); } 1053 static int ra( Register r) { return ra(r->encoding()); } 1054 static int ra( int x) { return opp_u_field(x, 15, 11); } 1055 static int rb( Register r) { return rb(r->encoding()); } 1056 static int rb( int x) { return opp_u_field(x, 20, 16); } 1057 static int rc( int x) { return opp_u_field(x, 31, 31); } 1058 static int rs( Register r) { return rs(r->encoding()); } 1059 static int rs( int x) { return opp_u_field(x, 10, 6); } 1060 // we don't want to use R0 in memory accesses, because it has value `0' then 1061 static int ra0mem( Register r) { assert(r != R0, "cannot use register R0 in memory access"); return ra(r); } 1062 static int ra0mem( int x) { assert(x != 0, "cannot use register 0 in memory access"); return ra(x); } 1063 1064 // register r is target 1065 static int rt( Register r) { return rs(r); } 1066 static int rt( int x) { return rs(x); } 1067 static int rta( Register r) { return ra(r); } 1068 static int rta0mem( Register r) { rta(r); return ra0mem(r); } 1069 1070 static int sh1620( int x) { return opp_u_field(x, 20, 16); } 1071 static int sh30( int x) { return opp_u_field(x, 30, 30); } 1072 static int sh162030( int x) { return sh1620(x & 0x1f) | sh30((x & 0x20) >> 5); } 1073 static int si( int x) { return opp_s_field(x, 31, 16); } 1074 static int spr( int x) { return opp_u_field(x, 20, 11); } 1075 static int sr( int x) { return opp_u_field(x, 15, 12); } 1076 static int tbr( int x) { return opp_u_field(x, 20, 11); } 1077 static int th( int x) { return opp_u_field(x, 10, 7); } 1078 static int thct( int x) { assert((x&8) == 0, "must be valid cache specification"); return th(x); } 1079 static int thds( int x) { assert((x&8) == 8, "must be valid stream specification"); return th(x); } 1080 static int to( int x) { return opp_u_field(x, 10, 6); } 1081 static int u( int x) { return opp_u_field(x, 19, 16); } 1082 static int ui( int x) { return opp_u_field(x, 31, 16); } 1083 1084 // Support vector instructions for >= Power6. 1085 static int vra( int x) { return opp_u_field(x, 15, 11); } 1086 static int vrb( int x) { return opp_u_field(x, 20, 16); } 1087 static int vrc( int x) { return opp_u_field(x, 25, 21); } 1088 static int vrs( int x) { return opp_u_field(x, 10, 6); } 1089 static int vrt( int x) { return opp_u_field(x, 10, 6); } 1090 1091 static int vra( VectorRegister r) { return vra(r->encoding());} 1092 static int vrb( VectorRegister r) { return vrb(r->encoding());} 1093 static int vrc( VectorRegister r) { return vrc(r->encoding());} 1094 static int vrs( VectorRegister r) { return vrs(r->encoding());} 1095 static int vrt( VectorRegister r) { return vrt(r->encoding());} 1096 1097 // Support Vector-Scalar (VSX) instructions. 1098 static int vsra( int x) { return opp_u_field(x, 15, 11); } 1099 static int vsrb( int x) { return opp_u_field(x, 20, 16); } 1100 static int vsrc( int x) { return opp_u_field(x, 25, 21); } 1101 static int vsrs( int x) { return opp_u_field(x, 10, 6); } 1102 static int vsrt( int x) { return opp_u_field(x, 10, 6); } 1103 1104 static int vsra( VectorSRegister r) { return vsra(r->encoding());} 1105 static int vsrb( VectorSRegister r) { return vsrb(r->encoding());} 1106 static int vsrc( VectorSRegister r) { return vsrc(r->encoding());} 1107 static int vsrs( VectorSRegister r) { return vsrs(r->encoding());} 1108 static int vsrt( VectorSRegister r) { return vsrt(r->encoding());} 1109 1110 static int vsplt_uim( int x) { return opp_u_field(x, 15, 12); } // for vsplt* instructions 1111 static int vsplti_sim(int x) { return opp_u_field(x, 15, 11); } // for vsplti* instructions 1112 static int vsldoi_shb(int x) { return opp_u_field(x, 25, 22); } // for vsldoi instruction 1113 static int vcmp_rc( int x) { return opp_u_field(x, 21, 21); } // for vcmp* instructions 1114 1115 //static int xo1( int x) { return opp_u_field(x, 29, 21); }// is contained in our opcodes 1116 //static int xo2( int x) { return opp_u_field(x, 30, 21); }// is contained in our opcodes 1117 //static int xo3( int x) { return opp_u_field(x, 30, 22); }// is contained in our opcodes 1118 //static int xo4( int x) { return opp_u_field(x, 30, 26); }// is contained in our opcodes 1119 //static int xo5( int x) { return opp_u_field(x, 29, 27); }// is contained in our opcodes 1120 //static int xo6( int x) { return opp_u_field(x, 30, 27); }// is contained in our opcodes 1121 //static int xo7( int x) { return opp_u_field(x, 31, 30); }// is contained in our opcodes 1122 1123 protected: 1124 // Compute relative address for branch. 1125 static intptr_t disp(intptr_t x, intptr_t off) { 1126 int xx = x - off; 1127 xx = xx >> 2; 1128 return xx; 1129 } 1130 1131 public: 1132 // signed immediate, in low bits, nbits long 1133 static int simm(int x, int nbits) { 1134 assert_signed_range(x, nbits); 1135 return x & ((1 << nbits) - 1); 1136 } 1137 1138 // unsigned immediate, in low bits, nbits long 1139 static int uimm(int x, int nbits) { 1140 assert_unsigned_const(x, nbits); 1141 return x & ((1 << nbits) - 1); 1142 } 1143 1144 static void set_imm(int* instr, short s) { 1145 // imm is always in the lower 16 bits of the instruction, 1146 // so this is endian-neutral. Same for the get_imm below. 1147 uint32_t w = *(uint32_t *)instr; 1148 *instr = (int)((w & ~0x0000FFFF) | (s & 0x0000FFFF)); 1149 } 1150 1151 static int get_imm(address a, int instruction_number) { 1152 return (short)((int *)a)[instruction_number]; 1153 } 1154 1155 static inline int hi16_signed( int x) { return (int)(int16_t)(x >> 16); } 1156 static inline int lo16_unsigned(int x) { return x & 0xffff; } 1157 1158 protected: 1159 1160 // Extract the top 32 bits in a 64 bit word. 1161 static int32_t hi32(int64_t x) { 1162 int32_t r = int32_t((uint64_t)x >> 32); 1163 return r; 1164 } 1165 1166 public: 1167 1168 static inline unsigned int align_addr(unsigned int addr, unsigned int a) { 1169 return ((addr + (a - 1)) & ~(a - 1)); 1170 } 1171 1172 static inline bool is_aligned(unsigned int addr, unsigned int a) { 1173 return (0 == addr % a); 1174 } 1175 1176 void flush() { 1177 AbstractAssembler::flush(); 1178 } 1179 1180 inline void emit_int32(int); // shadows AbstractAssembler::emit_int32 1181 inline void emit_data(int); 1182 inline void emit_data(int, RelocationHolder const&); 1183 inline void emit_data(int, relocInfo::relocType rtype); 1184 1185 // Emit an address. 1186 inline address emit_addr(const address addr = NULL); 1187 1188 #if !defined(ABI_ELFv2) 1189 // Emit a function descriptor with the specified entry point, TOC, 1190 // and ENV. If the entry point is NULL, the descriptor will point 1191 // just past the descriptor. 1192 // Use values from friend functions as defaults. 1193 inline address emit_fd(address entry = NULL, 1194 address toc = (address) FunctionDescriptor::friend_toc, 1195 address env = (address) FunctionDescriptor::friend_env); 1196 #endif 1197 1198 ///////////////////////////////////////////////////////////////////////////////////// 1199 // PPC instructions 1200 ///////////////////////////////////////////////////////////////////////////////////// 1201 1202 // Memory instructions use r0 as hard coded 0, e.g. to simulate loading 1203 // immediates. The normal instruction encoders enforce that r0 is not 1204 // passed to them. Use either extended mnemonics encoders or the special ra0 1205 // versions. 1206 1207 // Issue an illegal instruction. 1208 inline void illtrap(); 1209 static inline bool is_illtrap(int x); 1210 1211 // PPC 1, section 3.3.8, Fixed-Point Arithmetic Instructions 1212 inline void addi( Register d, Register a, int si16); 1213 inline void addis(Register d, Register a, int si16); 1214 private: 1215 inline void addi_r0ok( Register d, Register a, int si16); 1216 inline void addis_r0ok(Register d, Register a, int si16); 1217 public: 1218 inline void addic_( Register d, Register a, int si16); 1219 inline void subfic( Register d, Register a, int si16); 1220 inline void add( Register d, Register a, Register b); 1221 inline void add_( Register d, Register a, Register b); 1222 inline void subf( Register d, Register a, Register b); // d = b - a "Sub_from", as in ppc spec. 1223 inline void sub( Register d, Register a, Register b); // d = a - b Swap operands of subf for readability. 1224 inline void subf_( Register d, Register a, Register b); 1225 inline void addc( Register d, Register a, Register b); 1226 inline void addc_( Register d, Register a, Register b); 1227 inline void subfc( Register d, Register a, Register b); 1228 inline void subfc_( Register d, Register a, Register b); 1229 inline void adde( Register d, Register a, Register b); 1230 inline void adde_( Register d, Register a, Register b); 1231 inline void subfe( Register d, Register a, Register b); 1232 inline void subfe_( Register d, Register a, Register b); 1233 inline void addme( Register d, Register a); 1234 inline void addme_( Register d, Register a); 1235 inline void subfme( Register d, Register a); 1236 inline void subfme_(Register d, Register a); 1237 inline void addze( Register d, Register a); 1238 inline void addze_( Register d, Register a); 1239 inline void subfze( Register d, Register a); 1240 inline void subfze_(Register d, Register a); 1241 inline void neg( Register d, Register a); 1242 inline void neg_( Register d, Register a); 1243 inline void mulli( Register d, Register a, int si16); 1244 inline void mulld( Register d, Register a, Register b); 1245 inline void mulld_( Register d, Register a, Register b); 1246 inline void mullw( Register d, Register a, Register b); 1247 inline void mullw_( Register d, Register a, Register b); 1248 inline void mulhw( Register d, Register a, Register b); 1249 inline void mulhw_( Register d, Register a, Register b); 1250 inline void mulhwu( Register d, Register a, Register b); 1251 inline void mulhwu_(Register d, Register a, Register b); 1252 inline void mulhd( Register d, Register a, Register b); 1253 inline void mulhd_( Register d, Register a, Register b); 1254 inline void mulhdu( Register d, Register a, Register b); 1255 inline void mulhdu_(Register d, Register a, Register b); 1256 inline void divd( Register d, Register a, Register b); 1257 inline void divd_( Register d, Register a, Register b); 1258 inline void divw( Register d, Register a, Register b); 1259 inline void divw_( Register d, Register a, Register b); 1260 1261 // Fixed-Point Arithmetic Instructions with Overflow detection 1262 inline void addo( Register d, Register a, Register b); 1263 inline void addo_( Register d, Register a, Register b); 1264 inline void subfo( Register d, Register a, Register b); 1265 inline void subfo_( Register d, Register a, Register b); 1266 inline void addco( Register d, Register a, Register b); 1267 inline void addco_( Register d, Register a, Register b); 1268 inline void subfco( Register d, Register a, Register b); 1269 inline void subfco_( Register d, Register a, Register b); 1270 inline void addeo( Register d, Register a, Register b); 1271 inline void addeo_( Register d, Register a, Register b); 1272 inline void subfeo( Register d, Register a, Register b); 1273 inline void subfeo_( Register d, Register a, Register b); 1274 inline void addmeo( Register d, Register a); 1275 inline void addmeo_( Register d, Register a); 1276 inline void subfmeo( Register d, Register a); 1277 inline void subfmeo_(Register d, Register a); 1278 inline void addzeo( Register d, Register a); 1279 inline void addzeo_( Register d, Register a); 1280 inline void subfzeo( Register d, Register a); 1281 inline void subfzeo_(Register d, Register a); 1282 inline void nego( Register d, Register a); 1283 inline void nego_( Register d, Register a); 1284 inline void mulldo( Register d, Register a, Register b); 1285 inline void mulldo_( Register d, Register a, Register b); 1286 inline void mullwo( Register d, Register a, Register b); 1287 inline void mullwo_( Register d, Register a, Register b); 1288 inline void divdo( Register d, Register a, Register b); 1289 inline void divdo_( Register d, Register a, Register b); 1290 inline void divwo( Register d, Register a, Register b); 1291 inline void divwo_( Register d, Register a, Register b); 1292 1293 // extended mnemonics 1294 inline void li( Register d, int si16); 1295 inline void lis( Register d, int si16); 1296 inline void addir(Register d, int si16, Register a); 1297 1298 static bool is_addi(int x) { 1299 return ADDI_OPCODE == (x & ADDI_OPCODE_MASK); 1300 } 1301 static bool is_addis(int x) { 1302 return ADDIS_OPCODE == (x & ADDIS_OPCODE_MASK); 1303 } 1304 static bool is_bxx(int x) { 1305 return BXX_OPCODE == (x & BXX_OPCODE_MASK); 1306 } 1307 static bool is_b(int x) { 1308 return BXX_OPCODE == (x & BXX_OPCODE_MASK) && inv_lk_field(x) == 0; 1309 } 1310 static bool is_bl(int x) { 1311 return BXX_OPCODE == (x & BXX_OPCODE_MASK) && inv_lk_field(x) == 1; 1312 } 1313 static bool is_bcxx(int x) { 1314 return BCXX_OPCODE == (x & BCXX_OPCODE_MASK); 1315 } 1316 static bool is_bxx_or_bcxx(int x) { 1317 return is_bxx(x) || is_bcxx(x); 1318 } 1319 static bool is_bctrl(int x) { 1320 return x == 0x4e800421; 1321 } 1322 static bool is_bctr(int x) { 1323 return x == 0x4e800420; 1324 } 1325 static bool is_bclr(int x) { 1326 return BCLR_OPCODE == (x & XL_FORM_OPCODE_MASK); 1327 } 1328 static bool is_li(int x) { 1329 return is_addi(x) && inv_ra_field(x)==0; 1330 } 1331 static bool is_lis(int x) { 1332 return is_addis(x) && inv_ra_field(x)==0; 1333 } 1334 static bool is_mtctr(int x) { 1335 return MTCTR_OPCODE == (x & MTCTR_OPCODE_MASK); 1336 } 1337 static bool is_ld(int x) { 1338 return LD_OPCODE == (x & LD_OPCODE_MASK); 1339 } 1340 static bool is_std(int x) { 1341 return STD_OPCODE == (x & STD_OPCODE_MASK); 1342 } 1343 static bool is_stdu(int x) { 1344 return STDU_OPCODE == (x & STDU_OPCODE_MASK); 1345 } 1346 static bool is_stdx(int x) { 1347 return STDX_OPCODE == (x & STDX_OPCODE_MASK); 1348 } 1349 static bool is_stdux(int x) { 1350 return STDUX_OPCODE == (x & STDUX_OPCODE_MASK); 1351 } 1352 static bool is_stwx(int x) { 1353 return STWX_OPCODE == (x & STWX_OPCODE_MASK); 1354 } 1355 static bool is_stwux(int x) { 1356 return STWUX_OPCODE == (x & STWUX_OPCODE_MASK); 1357 } 1358 static bool is_stw(int x) { 1359 return STW_OPCODE == (x & STW_OPCODE_MASK); 1360 } 1361 static bool is_stwu(int x) { 1362 return STWU_OPCODE == (x & STWU_OPCODE_MASK); 1363 } 1364 static bool is_ori(int x) { 1365 return ORI_OPCODE == (x & ORI_OPCODE_MASK); 1366 }; 1367 static bool is_oris(int x) { 1368 return ORIS_OPCODE == (x & ORIS_OPCODE_MASK); 1369 }; 1370 static bool is_rldicr(int x) { 1371 return (RLDICR_OPCODE == (x & RLDICR_OPCODE_MASK)); 1372 }; 1373 static bool is_nop(int x) { 1374 return x == 0x60000000; 1375 } 1376 // endgroup opcode for Power6 1377 static bool is_endgroup(int x) { 1378 return is_ori(x) && inv_ra_field(x) == 1 && inv_rs_field(x) == 1 && inv_d1_field(x) == 0; 1379 } 1380 1381 1382 private: 1383 // PPC 1, section 3.3.9, Fixed-Point Compare Instructions 1384 inline void cmpi( ConditionRegister bf, int l, Register a, int si16); 1385 inline void cmp( ConditionRegister bf, int l, Register a, Register b); 1386 inline void cmpli(ConditionRegister bf, int l, Register a, int ui16); 1387 inline void cmpl( ConditionRegister bf, int l, Register a, Register b); 1388 1389 public: 1390 // extended mnemonics of Compare Instructions 1391 inline void cmpwi( ConditionRegister crx, Register a, int si16); 1392 inline void cmpdi( ConditionRegister crx, Register a, int si16); 1393 inline void cmpw( ConditionRegister crx, Register a, Register b); 1394 inline void cmpd( ConditionRegister crx, Register a, Register b); 1395 inline void cmplwi(ConditionRegister crx, Register a, int ui16); 1396 inline void cmpldi(ConditionRegister crx, Register a, int ui16); 1397 inline void cmplw( ConditionRegister crx, Register a, Register b); 1398 inline void cmpld( ConditionRegister crx, Register a, Register b); 1399 1400 inline void isel( Register d, Register a, Register b, int bc); 1401 // Convenient version which takes: Condition register, Condition code and invert flag. Omit b to keep old value. 1402 inline void isel( Register d, ConditionRegister cr, Condition cc, bool inv, Register a, Register b = noreg); 1403 // Set d = 0 if (cr.cc) equals 1, otherwise b. 1404 inline void isel_0( Register d, ConditionRegister cr, Condition cc, Register b = noreg); 1405 1406 // PPC 1, section 3.3.11, Fixed-Point Logical Instructions 1407 void andi( Register a, Register s, long ui16); // optimized version 1408 inline void andi_( Register a, Register s, int ui16); 1409 inline void andis_( Register a, Register s, int ui16); 1410 inline void ori( Register a, Register s, int ui16); 1411 inline void oris( Register a, Register s, int ui16); 1412 inline void xori( Register a, Register s, int ui16); 1413 inline void xoris( Register a, Register s, int ui16); 1414 inline void andr( Register a, Register s, Register b); // suffixed by 'r' as 'and' is C++ keyword 1415 inline void and_( Register a, Register s, Register b); 1416 // Turn or0(rx,rx,rx) into a nop and avoid that we accidently emit a 1417 // SMT-priority change instruction (see SMT instructions below). 1418 inline void or_unchecked(Register a, Register s, Register b); 1419 inline void orr( Register a, Register s, Register b); // suffixed by 'r' as 'or' is C++ keyword 1420 inline void or_( Register a, Register s, Register b); 1421 inline void xorr( Register a, Register s, Register b); // suffixed by 'r' as 'xor' is C++ keyword 1422 inline void xor_( Register a, Register s, Register b); 1423 inline void nand( Register a, Register s, Register b); 1424 inline void nand_( Register a, Register s, Register b); 1425 inline void nor( Register a, Register s, Register b); 1426 inline void nor_( Register a, Register s, Register b); 1427 inline void andc( Register a, Register s, Register b); 1428 inline void andc_( Register a, Register s, Register b); 1429 inline void orc( Register a, Register s, Register b); 1430 inline void orc_( Register a, Register s, Register b); 1431 inline void extsb( Register a, Register s); 1432 inline void extsb_( Register a, Register s); 1433 inline void extsh( Register a, Register s); 1434 inline void extsh_( Register a, Register s); 1435 inline void extsw( Register a, Register s); 1436 inline void extsw_( Register a, Register s); 1437 1438 // extended mnemonics 1439 inline void nop(); 1440 // NOP for FP and BR units (different versions to allow them to be in one group) 1441 inline void fpnop0(); 1442 inline void fpnop1(); 1443 inline void brnop0(); 1444 inline void brnop1(); 1445 inline void brnop2(); 1446 1447 inline void mr( Register d, Register s); 1448 inline void ori_opt( Register d, int ui16); 1449 inline void oris_opt(Register d, int ui16); 1450 1451 // endgroup opcode for Power6 1452 inline void endgroup(); 1453 1454 // count instructions 1455 inline void cntlzw( Register a, Register s); 1456 inline void cntlzw_( Register a, Register s); 1457 inline void cntlzd( Register a, Register s); 1458 inline void cntlzd_( Register a, Register s); 1459 1460 // PPC 1, section 3.3.12, Fixed-Point Rotate and Shift Instructions 1461 inline void sld( Register a, Register s, Register b); 1462 inline void sld_( Register a, Register s, Register b); 1463 inline void slw( Register a, Register s, Register b); 1464 inline void slw_( Register a, Register s, Register b); 1465 inline void srd( Register a, Register s, Register b); 1466 inline void srd_( Register a, Register s, Register b); 1467 inline void srw( Register a, Register s, Register b); 1468 inline void srw_( Register a, Register s, Register b); 1469 inline void srad( Register a, Register s, Register b); 1470 inline void srad_( Register a, Register s, Register b); 1471 inline void sraw( Register a, Register s, Register b); 1472 inline void sraw_( Register a, Register s, Register b); 1473 inline void sradi( Register a, Register s, int sh6); 1474 inline void sradi_( Register a, Register s, int sh6); 1475 inline void srawi( Register a, Register s, int sh5); 1476 inline void srawi_( Register a, Register s, int sh5); 1477 1478 // extended mnemonics for Shift Instructions 1479 inline void sldi( Register a, Register s, int sh6); 1480 inline void sldi_( Register a, Register s, int sh6); 1481 inline void slwi( Register a, Register s, int sh5); 1482 inline void slwi_( Register a, Register s, int sh5); 1483 inline void srdi( Register a, Register s, int sh6); 1484 inline void srdi_( Register a, Register s, int sh6); 1485 inline void srwi( Register a, Register s, int sh5); 1486 inline void srwi_( Register a, Register s, int sh5); 1487 1488 inline void clrrdi( Register a, Register s, int ui6); 1489 inline void clrrdi_( Register a, Register s, int ui6); 1490 inline void clrldi( Register a, Register s, int ui6); 1491 inline void clrldi_( Register a, Register s, int ui6); 1492 inline void clrlsldi(Register a, Register s, int clrl6, int shl6); 1493 inline void clrlsldi_(Register a, Register s, int clrl6, int shl6); 1494 inline void extrdi( Register a, Register s, int n, int b); 1495 // testbit with condition register 1496 inline void testbitdi(ConditionRegister cr, Register a, Register s, int ui6); 1497 1498 // rotate instructions 1499 inline void rotldi( Register a, Register s, int n); 1500 inline void rotrdi( Register a, Register s, int n); 1501 inline void rotlwi( Register a, Register s, int n); 1502 inline void rotrwi( Register a, Register s, int n); 1503 1504 // Rotate Instructions 1505 inline void rldic( Register a, Register s, int sh6, int mb6); 1506 inline void rldic_( Register a, Register s, int sh6, int mb6); 1507 inline void rldicr( Register a, Register s, int sh6, int mb6); 1508 inline void rldicr_( Register a, Register s, int sh6, int mb6); 1509 inline void rldicl( Register a, Register s, int sh6, int mb6); 1510 inline void rldicl_( Register a, Register s, int sh6, int mb6); 1511 inline void rlwinm( Register a, Register s, int sh5, int mb5, int me5); 1512 inline void rlwinm_( Register a, Register s, int sh5, int mb5, int me5); 1513 inline void rldimi( Register a, Register s, int sh6, int mb6); 1514 inline void rldimi_( Register a, Register s, int sh6, int mb6); 1515 inline void rlwimi( Register a, Register s, int sh5, int mb5, int me5); 1516 inline void insrdi( Register a, Register s, int n, int b); 1517 inline void insrwi( Register a, Register s, int n, int b); 1518 1519 // PPC 1, section 3.3.2 Fixed-Point Load Instructions 1520 // 4 bytes 1521 inline void lwzx( Register d, Register s1, Register s2); 1522 inline void lwz( Register d, int si16, Register s1); 1523 inline void lwzu( Register d, int si16, Register s1); 1524 1525 // 4 bytes 1526 inline void lwax( Register d, Register s1, Register s2); 1527 inline void lwa( Register d, int si16, Register s1); 1528 1529 // 4 bytes reversed 1530 inline void lwbrx( Register d, Register s1, Register s2); 1531 1532 // 2 bytes 1533 inline void lhzx( Register d, Register s1, Register s2); 1534 inline void lhz( Register d, int si16, Register s1); 1535 inline void lhzu( Register d, int si16, Register s1); 1536 1537 // 2 bytes reversed 1538 inline void lhbrx( Register d, Register s1, Register s2); 1539 1540 // 2 bytes 1541 inline void lhax( Register d, Register s1, Register s2); 1542 inline void lha( Register d, int si16, Register s1); 1543 inline void lhau( Register d, int si16, Register s1); 1544 1545 // 1 byte 1546 inline void lbzx( Register d, Register s1, Register s2); 1547 inline void lbz( Register d, int si16, Register s1); 1548 inline void lbzu( Register d, int si16, Register s1); 1549 1550 // 8 bytes 1551 inline void ldx( Register d, Register s1, Register s2); 1552 inline void ld( Register d, int si16, Register s1); 1553 inline void ldu( Register d, int si16, Register s1); 1554 1555 // For convenience. Load pointer into d from b+s1. 1556 inline void ld_ptr(Register d, int b, Register s1); 1557 DEBUG_ONLY(inline void ld_ptr(Register d, ByteSize b, Register s1);) 1558 1559 // PPC 1, section 3.3.3 Fixed-Point Store Instructions 1560 inline void stwx( Register d, Register s1, Register s2); 1561 inline void stw( Register d, int si16, Register s1); 1562 inline void stwu( Register d, int si16, Register s1); 1563 1564 inline void sthx( Register d, Register s1, Register s2); 1565 inline void sth( Register d, int si16, Register s1); 1566 inline void sthu( Register d, int si16, Register s1); 1567 1568 inline void stbx( Register d, Register s1, Register s2); 1569 inline void stb( Register d, int si16, Register s1); 1570 inline void stbu( Register d, int si16, Register s1); 1571 1572 inline void stdx( Register d, Register s1, Register s2); 1573 inline void std( Register d, int si16, Register s1); 1574 inline void stdu( Register d, int si16, Register s1); 1575 inline void stdux(Register s, Register a, Register b); 1576 1577 inline void st_ptr(Register d, int si16, Register s1); 1578 DEBUG_ONLY(inline void st_ptr(Register d, ByteSize b, Register s1);) 1579 1580 // PPC 1, section 3.3.13 Move To/From System Register Instructions 1581 inline void mtlr( Register s1); 1582 inline void mflr( Register d); 1583 inline void mtctr(Register s1); 1584 inline void mfctr(Register d); 1585 inline void mtcrf(int fxm, Register s); 1586 inline void mfcr( Register d); 1587 inline void mcrf( ConditionRegister crd, ConditionRegister cra); 1588 inline void mtcr( Register s); 1589 1590 // Special purpose registers 1591 // Exception Register 1592 inline void mtxer(Register s1); 1593 inline void mfxer(Register d); 1594 // Vector Register Save Register 1595 inline void mtvrsave(Register s1); 1596 inline void mfvrsave(Register d); 1597 // Timebase 1598 inline void mftb(Register d); 1599 // Introduced with Power 8: 1600 // Data Stream Control Register 1601 inline void mtdscr(Register s1); 1602 inline void mfdscr(Register d ); 1603 // Transactional Memory Registers 1604 inline void mftfhar(Register d); 1605 inline void mftfiar(Register d); 1606 inline void mftexasr(Register d); 1607 inline void mftexasru(Register d); 1608 1609 // TEXASR bit description 1610 enum transaction_failure_reason { 1611 // Upper half (TEXASRU): 1612 tm_failure_persistent = 7, // The failure is likely to recur on each execution. 1613 tm_disallowed = 8, // The instruction is not permitted. 1614 tm_nesting_of = 9, // The maximum transaction level was exceeded. 1615 tm_footprint_of = 10, // The tracking limit for transactional storage accesses was exceeded. 1616 tm_self_induced_cf = 11, // A self-induced conflict occurred in Suspended state. 1617 tm_non_trans_cf = 12, // A conflict occurred with a non-transactional access by another processor. 1618 tm_trans_cf = 13, // A conflict occurred with another transaction. 1619 tm_translation_cf = 14, // A conflict occurred with a TLB invalidation. 1620 tm_inst_fetch_cf = 16, // An instruction fetch was performed from a block that was previously written transactionally. 1621 tm_tabort = 31, // Termination was caused by the execution of an abort instruction. 1622 // Lower half: 1623 tm_suspended = 32, // Failure was recorded in Suspended state. 1624 tm_failure_summary = 36, // Failure has been detected and recorded. 1625 tm_tfiar_exact = 37, // Value in the TFIAR is exact. 1626 tm_rot = 38, // Rollback-only transaction. 1627 }; 1628 1629 // PPC 1, section 2.4.1 Branch Instructions 1630 inline void b( address a, relocInfo::relocType rt = relocInfo::none); 1631 inline void b( Label& L); 1632 inline void bl( address a, relocInfo::relocType rt = relocInfo::none); 1633 inline void bl( Label& L); 1634 inline void bc( int boint, int biint, address a, relocInfo::relocType rt = relocInfo::none); 1635 inline void bc( int boint, int biint, Label& L); 1636 inline void bcl(int boint, int biint, address a, relocInfo::relocType rt = relocInfo::none); 1637 inline void bcl(int boint, int biint, Label& L); 1638 1639 inline void bclr( int boint, int biint, int bhint, relocInfo::relocType rt = relocInfo::none); 1640 inline void bclrl( int boint, int biint, int bhint, relocInfo::relocType rt = relocInfo::none); 1641 inline void bcctr( int boint, int biint, int bhint = bhintbhBCCTRisNotReturnButSame, 1642 relocInfo::relocType rt = relocInfo::none); 1643 inline void bcctrl(int boint, int biint, int bhint = bhintbhBCLRisReturn, 1644 relocInfo::relocType rt = relocInfo::none); 1645 1646 // helper function for b, bcxx 1647 inline bool is_within_range_of_b(address a, address pc); 1648 inline bool is_within_range_of_bcxx(address a, address pc); 1649 1650 // get the destination of a bxx branch (b, bl, ba, bla) 1651 static inline address bxx_destination(address baddr); 1652 static inline address bxx_destination(int instr, address pc); 1653 static inline intptr_t bxx_destination_offset(int instr, intptr_t bxx_pos); 1654 1655 // extended mnemonics for branch instructions 1656 inline void blt(ConditionRegister crx, Label& L); 1657 inline void bgt(ConditionRegister crx, Label& L); 1658 inline void beq(ConditionRegister crx, Label& L); 1659 inline void bso(ConditionRegister crx, Label& L); 1660 inline void bge(ConditionRegister crx, Label& L); 1661 inline void ble(ConditionRegister crx, Label& L); 1662 inline void bne(ConditionRegister crx, Label& L); 1663 inline void bns(ConditionRegister crx, Label& L); 1664 1665 // Branch instructions with static prediction hints. 1666 inline void blt_predict_taken( ConditionRegister crx, Label& L); 1667 inline void bgt_predict_taken( ConditionRegister crx, Label& L); 1668 inline void beq_predict_taken( ConditionRegister crx, Label& L); 1669 inline void bso_predict_taken( ConditionRegister crx, Label& L); 1670 inline void bge_predict_taken( ConditionRegister crx, Label& L); 1671 inline void ble_predict_taken( ConditionRegister crx, Label& L); 1672 inline void bne_predict_taken( ConditionRegister crx, Label& L); 1673 inline void bns_predict_taken( ConditionRegister crx, Label& L); 1674 inline void blt_predict_not_taken(ConditionRegister crx, Label& L); 1675 inline void bgt_predict_not_taken(ConditionRegister crx, Label& L); 1676 inline void beq_predict_not_taken(ConditionRegister crx, Label& L); 1677 inline void bso_predict_not_taken(ConditionRegister crx, Label& L); 1678 inline void bge_predict_not_taken(ConditionRegister crx, Label& L); 1679 inline void ble_predict_not_taken(ConditionRegister crx, Label& L); 1680 inline void bne_predict_not_taken(ConditionRegister crx, Label& L); 1681 inline void bns_predict_not_taken(ConditionRegister crx, Label& L); 1682 1683 // for use in conjunction with testbitdi: 1684 inline void btrue( ConditionRegister crx, Label& L); 1685 inline void bfalse(ConditionRegister crx, Label& L); 1686 1687 inline void bltl(ConditionRegister crx, Label& L); 1688 inline void bgtl(ConditionRegister crx, Label& L); 1689 inline void beql(ConditionRegister crx, Label& L); 1690 inline void bsol(ConditionRegister crx, Label& L); 1691 inline void bgel(ConditionRegister crx, Label& L); 1692 inline void blel(ConditionRegister crx, Label& L); 1693 inline void bnel(ConditionRegister crx, Label& L); 1694 inline void bnsl(ConditionRegister crx, Label& L); 1695 1696 // extended mnemonics for Branch Instructions via LR 1697 // We use `blr' for returns. 1698 inline void blr(relocInfo::relocType rt = relocInfo::none); 1699 1700 // extended mnemonics for Branch Instructions with CTR 1701 // bdnz means `decrement CTR and jump to L if CTR is not zero' 1702 inline void bdnz(Label& L); 1703 // Decrement and branch if result is zero. 1704 inline void bdz(Label& L); 1705 // we use `bctr[l]' for jumps/calls in function descriptor glue 1706 // code, e.g. calls to runtime functions 1707 inline void bctr( relocInfo::relocType rt = relocInfo::none); 1708 inline void bctrl(relocInfo::relocType rt = relocInfo::none); 1709 // conditional jumps/branches via CTR 1710 inline void beqctr( ConditionRegister crx, relocInfo::relocType rt = relocInfo::none); 1711 inline void beqctrl(ConditionRegister crx, relocInfo::relocType rt = relocInfo::none); 1712 inline void bnectr( ConditionRegister crx, relocInfo::relocType rt = relocInfo::none); 1713 inline void bnectrl(ConditionRegister crx, relocInfo::relocType rt = relocInfo::none); 1714 1715 // condition register logic instructions 1716 // NOTE: There's a preferred form: d and s2 should point into the same condition register. 1717 inline void crand( int d, int s1, int s2); 1718 inline void crnand(int d, int s1, int s2); 1719 inline void cror( int d, int s1, int s2); 1720 inline void crxor( int d, int s1, int s2); 1721 inline void crnor( int d, int s1, int s2); 1722 inline void creqv( int d, int s1, int s2); 1723 inline void crandc(int d, int s1, int s2); 1724 inline void crorc( int d, int s1, int s2); 1725 1726 // More convenient version. 1727 int condition_register_bit(ConditionRegister cr, Condition c) { 1728 return 4 * (int)(intptr_t)cr + c; 1729 } 1730 void crand( ConditionRegister crdst, Condition cdst, ConditionRegister crsrc, Condition csrc); 1731 void crnand(ConditionRegister crdst, Condition cdst, ConditionRegister crsrc, Condition csrc); 1732 void cror( ConditionRegister crdst, Condition cdst, ConditionRegister crsrc, Condition csrc); 1733 void crxor( ConditionRegister crdst, Condition cdst, ConditionRegister crsrc, Condition csrc); 1734 void crnor( ConditionRegister crdst, Condition cdst, ConditionRegister crsrc, Condition csrc); 1735 void creqv( ConditionRegister crdst, Condition cdst, ConditionRegister crsrc, Condition csrc); 1736 void crandc(ConditionRegister crdst, Condition cdst, ConditionRegister crsrc, Condition csrc); 1737 void crorc( ConditionRegister crdst, Condition cdst, ConditionRegister crsrc, Condition csrc); 1738 1739 // icache and dcache related instructions 1740 inline void icbi( Register s1, Register s2); 1741 //inline void dcba(Register s1, Register s2); // Instruction for embedded processor only. 1742 inline void dcbz( Register s1, Register s2); 1743 inline void dcbst( Register s1, Register s2); 1744 inline void dcbf( Register s1, Register s2); 1745 1746 enum ct_cache_specification { 1747 ct_primary_cache = 0, 1748 ct_secondary_cache = 2 1749 }; 1750 // dcache read hint 1751 inline void dcbt( Register s1, Register s2); 1752 inline void dcbtct( Register s1, Register s2, int ct); 1753 inline void dcbtds( Register s1, Register s2, int ds); 1754 // dcache write hint 1755 inline void dcbtst( Register s1, Register s2); 1756 inline void dcbtstct(Register s1, Register s2, int ct); 1757 1758 // machine barrier instructions: 1759 // 1760 // - sync two-way memory barrier, aka fence 1761 // - lwsync orders Store|Store, 1762 // Load|Store, 1763 // Load|Load, 1764 // but not Store|Load 1765 // - eieio orders memory accesses for device memory (only) 1766 // - isync invalidates speculatively executed instructions 1767 // From the Power ISA 2.06 documentation: 1768 // "[...] an isync instruction prevents the execution of 1769 // instructions following the isync until instructions 1770 // preceding the isync have completed, [...]" 1771 // From IBM's AIX assembler reference: 1772 // "The isync [...] instructions causes the processor to 1773 // refetch any instructions that might have been fetched 1774 // prior to the isync instruction. The instruction isync 1775 // causes the processor to wait for all previous instructions 1776 // to complete. Then any instructions already fetched are 1777 // discarded and instruction processing continues in the 1778 // environment established by the previous instructions." 1779 // 1780 // semantic barrier instructions: 1781 // (as defined in orderAccess.hpp) 1782 // 1783 // - release orders Store|Store, (maps to lwsync) 1784 // Load|Store 1785 // - acquire orders Load|Store, (maps to lwsync) 1786 // Load|Load 1787 // - fence orders Store|Store, (maps to sync) 1788 // Load|Store, 1789 // Load|Load, 1790 // Store|Load 1791 // 1792 private: 1793 inline void sync(int l); 1794 public: 1795 inline void sync(); 1796 inline void lwsync(); 1797 inline void ptesync(); 1798 inline void eieio(); 1799 inline void isync(); 1800 inline void elemental_membar(int e); // Elemental Memory Barriers (>=Power 8) 1801 1802 // Wait instructions for polling. Attention: May result in SIGILL. 1803 inline void wait(); 1804 inline void waitrsv(); // >=Power7 1805 1806 // atomics 1807 inline void lbarx_unchecked(Register d, Register a, Register b, int eh1 = 0); // >=Power 8 1808 inline void lharx_unchecked(Register d, Register a, Register b, int eh1 = 0); // >=Power 8 1809 inline void lwarx_unchecked(Register d, Register a, Register b, int eh1 = 0); 1810 inline void ldarx_unchecked(Register d, Register a, Register b, int eh1 = 0); 1811 inline void lqarx_unchecked(Register d, Register a, Register b, int eh1 = 0); // >=Power 8 1812 inline bool lxarx_hint_exclusive_access(); 1813 inline void lbarx( Register d, Register a, Register b, bool hint_exclusive_access = false); 1814 inline void lharx( Register d, Register a, Register b, bool hint_exclusive_access = false); 1815 inline void lwarx( Register d, Register a, Register b, bool hint_exclusive_access = false); 1816 inline void ldarx( Register d, Register a, Register b, bool hint_exclusive_access = false); 1817 inline void lqarx( Register d, Register a, Register b, bool hint_exclusive_access = false); 1818 inline void stbcx_( Register s, Register a, Register b); 1819 inline void sthcx_( Register s, Register a, Register b); 1820 inline void stwcx_( Register s, Register a, Register b); 1821 inline void stdcx_( Register s, Register a, Register b); 1822 inline void stqcx_( Register s, Register a, Register b); 1823 1824 // Instructions for adjusting thread priority for simultaneous 1825 // multithreading (SMT) on Power5. 1826 private: 1827 inline void smt_prio_very_low(); 1828 inline void smt_prio_medium_high(); 1829 inline void smt_prio_high(); 1830 1831 public: 1832 inline void smt_prio_low(); 1833 inline void smt_prio_medium_low(); 1834 inline void smt_prio_medium(); 1835 // >= Power7 1836 inline void smt_yield(); 1837 inline void smt_mdoio(); 1838 inline void smt_mdoom(); 1839 // >= Power8 1840 inline void smt_miso(); 1841 1842 // trap instructions 1843 inline void twi_0(Register a); // for load with acquire semantics use load+twi_0+isync (trap can't occur) 1844 // NOT FOR DIRECT USE!! 1845 protected: 1846 inline void tdi_unchecked(int tobits, Register a, int si16); 1847 inline void twi_unchecked(int tobits, Register a, int si16); 1848 inline void tdi( int tobits, Register a, int si16); // asserts UseSIGTRAP 1849 inline void twi( int tobits, Register a, int si16); // asserts UseSIGTRAP 1850 inline void td( int tobits, Register a, Register b); // asserts UseSIGTRAP 1851 inline void tw( int tobits, Register a, Register b); // asserts UseSIGTRAP 1852 1853 static bool is_tdi(int x, int tobits, int ra, int si16) { 1854 return (TDI_OPCODE == (x & TDI_OPCODE_MASK)) 1855 && (tobits == inv_to_field(x)) 1856 && (ra == -1/*any reg*/ || ra == inv_ra_field(x)) 1857 && (si16 == inv_si_field(x)); 1858 } 1859 1860 static bool is_twi(int x, int tobits, int ra, int si16) { 1861 return (TWI_OPCODE == (x & TWI_OPCODE_MASK)) 1862 && (tobits == inv_to_field(x)) 1863 && (ra == -1/*any reg*/ || ra == inv_ra_field(x)) 1864 && (si16 == inv_si_field(x)); 1865 } 1866 1867 static bool is_twi(int x, int tobits, int ra) { 1868 return (TWI_OPCODE == (x & TWI_OPCODE_MASK)) 1869 && (tobits == inv_to_field(x)) 1870 && (ra == -1/*any reg*/ || ra == inv_ra_field(x)); 1871 } 1872 1873 static bool is_td(int x, int tobits, int ra, int rb) { 1874 return (TD_OPCODE == (x & TD_OPCODE_MASK)) 1875 && (tobits == inv_to_field(x)) 1876 && (ra == -1/*any reg*/ || ra == inv_ra_field(x)) 1877 && (rb == -1/*any reg*/ || rb == inv_rb_field(x)); 1878 } 1879 1880 static bool is_tw(int x, int tobits, int ra, int rb) { 1881 return (TW_OPCODE == (x & TW_OPCODE_MASK)) 1882 && (tobits == inv_to_field(x)) 1883 && (ra == -1/*any reg*/ || ra == inv_ra_field(x)) 1884 && (rb == -1/*any reg*/ || rb == inv_rb_field(x)); 1885 } 1886 1887 public: 1888 // PPC floating point instructions 1889 // PPC 1, section 4.6.2 Floating-Point Load Instructions 1890 inline void lfs( FloatRegister d, int si16, Register a); 1891 inline void lfsu( FloatRegister d, int si16, Register a); 1892 inline void lfsx( FloatRegister d, Register a, Register b); 1893 inline void lfd( FloatRegister d, int si16, Register a); 1894 inline void lfdu( FloatRegister d, int si16, Register a); 1895 inline void lfdx( FloatRegister d, Register a, Register b); 1896 1897 // PPC 1, section 4.6.3 Floating-Point Store Instructions 1898 inline void stfs( FloatRegister s, int si16, Register a); 1899 inline void stfsu( FloatRegister s, int si16, Register a); 1900 inline void stfsx( FloatRegister s, Register a, Register b); 1901 inline void stfd( FloatRegister s, int si16, Register a); 1902 inline void stfdu( FloatRegister s, int si16, Register a); 1903 inline void stfdx( FloatRegister s, Register a, Register b); 1904 1905 // PPC 1, section 4.6.4 Floating-Point Move Instructions 1906 inline void fmr( FloatRegister d, FloatRegister b); 1907 inline void fmr_( FloatRegister d, FloatRegister b); 1908 1909 // inline void mffgpr( FloatRegister d, Register b); 1910 // inline void mftgpr( Register d, FloatRegister b); 1911 inline void cmpb( Register a, Register s, Register b); 1912 inline void popcntb(Register a, Register s); 1913 inline void popcntw(Register a, Register s); 1914 inline void popcntd(Register a, Register s); 1915 1916 inline void fneg( FloatRegister d, FloatRegister b); 1917 inline void fneg_( FloatRegister d, FloatRegister b); 1918 inline void fabs( FloatRegister d, FloatRegister b); 1919 inline void fabs_( FloatRegister d, FloatRegister b); 1920 inline void fnabs( FloatRegister d, FloatRegister b); 1921 inline void fnabs_(FloatRegister d, FloatRegister b); 1922 1923 // PPC 1, section 4.6.5.1 Floating-Point Elementary Arithmetic Instructions 1924 inline void fadd( FloatRegister d, FloatRegister a, FloatRegister b); 1925 inline void fadd_( FloatRegister d, FloatRegister a, FloatRegister b); 1926 inline void fadds( FloatRegister d, FloatRegister a, FloatRegister b); 1927 inline void fadds_(FloatRegister d, FloatRegister a, FloatRegister b); 1928 inline void fsub( FloatRegister d, FloatRegister a, FloatRegister b); 1929 inline void fsub_( FloatRegister d, FloatRegister a, FloatRegister b); 1930 inline void fsubs( FloatRegister d, FloatRegister a, FloatRegister b); 1931 inline void fsubs_(FloatRegister d, FloatRegister a, FloatRegister b); 1932 inline void fmul( FloatRegister d, FloatRegister a, FloatRegister c); 1933 inline void fmul_( FloatRegister d, FloatRegister a, FloatRegister c); 1934 inline void fmuls( FloatRegister d, FloatRegister a, FloatRegister c); 1935 inline void fmuls_(FloatRegister d, FloatRegister a, FloatRegister c); 1936 inline void fdiv( FloatRegister d, FloatRegister a, FloatRegister b); 1937 inline void fdiv_( FloatRegister d, FloatRegister a, FloatRegister b); 1938 inline void fdivs( FloatRegister d, FloatRegister a, FloatRegister b); 1939 inline void fdivs_(FloatRegister d, FloatRegister a, FloatRegister b); 1940 1941 // Fused multiply-accumulate instructions. 1942 // WARNING: Use only when rounding between the 2 parts is not desired. 1943 // Some floating point tck tests will fail if used incorrectly. 1944 inline void fmadd( FloatRegister d, FloatRegister a, FloatRegister c, FloatRegister b); 1945 inline void fmadd_( FloatRegister d, FloatRegister a, FloatRegister c, FloatRegister b); 1946 inline void fmadds( FloatRegister d, FloatRegister a, FloatRegister c, FloatRegister b); 1947 inline void fmadds_( FloatRegister d, FloatRegister a, FloatRegister c, FloatRegister b); 1948 inline void fmsub( FloatRegister d, FloatRegister a, FloatRegister c, FloatRegister b); 1949 inline void fmsub_( FloatRegister d, FloatRegister a, FloatRegister c, FloatRegister b); 1950 inline void fmsubs( FloatRegister d, FloatRegister a, FloatRegister c, FloatRegister b); 1951 inline void fmsubs_( FloatRegister d, FloatRegister a, FloatRegister c, FloatRegister b); 1952 inline void fnmadd( FloatRegister d, FloatRegister a, FloatRegister c, FloatRegister b); 1953 inline void fnmadd_( FloatRegister d, FloatRegister a, FloatRegister c, FloatRegister b); 1954 inline void fnmadds( FloatRegister d, FloatRegister a, FloatRegister c, FloatRegister b); 1955 inline void fnmadds_(FloatRegister d, FloatRegister a, FloatRegister c, FloatRegister b); 1956 inline void fnmsub( FloatRegister d, FloatRegister a, FloatRegister c, FloatRegister b); 1957 inline void fnmsub_( FloatRegister d, FloatRegister a, FloatRegister c, FloatRegister b); 1958 inline void fnmsubs( FloatRegister d, FloatRegister a, FloatRegister c, FloatRegister b); 1959 inline void fnmsubs_(FloatRegister d, FloatRegister a, FloatRegister c, FloatRegister b); 1960 1961 // PPC 1, section 4.6.6 Floating-Point Rounding and Conversion Instructions 1962 inline void frsp( FloatRegister d, FloatRegister b); 1963 inline void fctid( FloatRegister d, FloatRegister b); 1964 inline void fctidz(FloatRegister d, FloatRegister b); 1965 inline void fctiw( FloatRegister d, FloatRegister b); 1966 inline void fctiwz(FloatRegister d, FloatRegister b); 1967 inline void fcfid( FloatRegister d, FloatRegister b); 1968 inline void fcfids(FloatRegister d, FloatRegister b); 1969 1970 // PPC 1, section 4.6.7 Floating-Point Compare Instructions 1971 inline void fcmpu( ConditionRegister crx, FloatRegister a, FloatRegister b); 1972 1973 inline void fsqrt( FloatRegister d, FloatRegister b); 1974 inline void fsqrts(FloatRegister d, FloatRegister b); 1975 1976 // Vector instructions for >= Power6. 1977 inline void lvebx( VectorRegister d, Register s1, Register s2); 1978 inline void lvehx( VectorRegister d, Register s1, Register s2); 1979 inline void lvewx( VectorRegister d, Register s1, Register s2); 1980 inline void lvx( VectorRegister d, Register s1, Register s2); 1981 inline void lvxl( VectorRegister d, Register s1, Register s2); 1982 inline void stvebx( VectorRegister d, Register s1, Register s2); 1983 inline void stvehx( VectorRegister d, Register s1, Register s2); 1984 inline void stvewx( VectorRegister d, Register s1, Register s2); 1985 inline void stvx( VectorRegister d, Register s1, Register s2); 1986 inline void stvxl( VectorRegister d, Register s1, Register s2); 1987 inline void lvsl( VectorRegister d, Register s1, Register s2); 1988 inline void lvsr( VectorRegister d, Register s1, Register s2); 1989 inline void vpkpx( VectorRegister d, VectorRegister a, VectorRegister b); 1990 inline void vpkshss( VectorRegister d, VectorRegister a, VectorRegister b); 1991 inline void vpkswss( VectorRegister d, VectorRegister a, VectorRegister b); 1992 inline void vpkshus( VectorRegister d, VectorRegister a, VectorRegister b); 1993 inline void vpkswus( VectorRegister d, VectorRegister a, VectorRegister b); 1994 inline void vpkuhum( VectorRegister d, VectorRegister a, VectorRegister b); 1995 inline void vpkuwum( VectorRegister d, VectorRegister a, VectorRegister b); 1996 inline void vpkuhus( VectorRegister d, VectorRegister a, VectorRegister b); 1997 inline void vpkuwus( VectorRegister d, VectorRegister a, VectorRegister b); 1998 inline void vupkhpx( VectorRegister d, VectorRegister b); 1999 inline void vupkhsb( VectorRegister d, VectorRegister b); 2000 inline void vupkhsh( VectorRegister d, VectorRegister b); 2001 inline void vupklpx( VectorRegister d, VectorRegister b); 2002 inline void vupklsb( VectorRegister d, VectorRegister b); 2003 inline void vupklsh( VectorRegister d, VectorRegister b); 2004 inline void vmrghb( VectorRegister d, VectorRegister a, VectorRegister b); 2005 inline void vmrghw( VectorRegister d, VectorRegister a, VectorRegister b); 2006 inline void vmrghh( VectorRegister d, VectorRegister a, VectorRegister b); 2007 inline void vmrglb( VectorRegister d, VectorRegister a, VectorRegister b); 2008 inline void vmrglw( VectorRegister d, VectorRegister a, VectorRegister b); 2009 inline void vmrglh( VectorRegister d, VectorRegister a, VectorRegister b); 2010 inline void vsplt( VectorRegister d, int ui4, VectorRegister b); 2011 inline void vsplth( VectorRegister d, int ui3, VectorRegister b); 2012 inline void vspltw( VectorRegister d, int ui2, VectorRegister b); 2013 inline void vspltisb( VectorRegister d, int si5); 2014 inline void vspltish( VectorRegister d, int si5); 2015 inline void vspltisw( VectorRegister d, int si5); 2016 inline void vperm( VectorRegister d, VectorRegister a, VectorRegister b, VectorRegister c); 2017 inline void vsel( VectorRegister d, VectorRegister a, VectorRegister b, VectorRegister c); 2018 inline void vsl( VectorRegister d, VectorRegister a, VectorRegister b); 2019 inline void vsldoi( VectorRegister d, VectorRegister a, VectorRegister b, int si4); 2020 inline void vslo( VectorRegister d, VectorRegister a, VectorRegister b); 2021 inline void vsr( VectorRegister d, VectorRegister a, VectorRegister b); 2022 inline void vsro( VectorRegister d, VectorRegister a, VectorRegister b); 2023 inline void vaddcuw( VectorRegister d, VectorRegister a, VectorRegister b); 2024 inline void vaddshs( VectorRegister d, VectorRegister a, VectorRegister b); 2025 inline void vaddsbs( VectorRegister d, VectorRegister a, VectorRegister b); 2026 inline void vaddsws( VectorRegister d, VectorRegister a, VectorRegister b); 2027 inline void vaddubm( VectorRegister d, VectorRegister a, VectorRegister b); 2028 inline void vadduwm( VectorRegister d, VectorRegister a, VectorRegister b); 2029 inline void vadduhm( VectorRegister d, VectorRegister a, VectorRegister b); 2030 inline void vaddubs( VectorRegister d, VectorRegister a, VectorRegister b); 2031 inline void vadduws( VectorRegister d, VectorRegister a, VectorRegister b); 2032 inline void vadduhs( VectorRegister d, VectorRegister a, VectorRegister b); 2033 inline void vsubcuw( VectorRegister d, VectorRegister a, VectorRegister b); 2034 inline void vsubshs( VectorRegister d, VectorRegister a, VectorRegister b); 2035 inline void vsubsbs( VectorRegister d, VectorRegister a, VectorRegister b); 2036 inline void vsubsws( VectorRegister d, VectorRegister a, VectorRegister b); 2037 inline void vsububm( VectorRegister d, VectorRegister a, VectorRegister b); 2038 inline void vsubuwm( VectorRegister d, VectorRegister a, VectorRegister b); 2039 inline void vsubuhm( VectorRegister d, VectorRegister a, VectorRegister b); 2040 inline void vsububs( VectorRegister d, VectorRegister a, VectorRegister b); 2041 inline void vsubuws( VectorRegister d, VectorRegister a, VectorRegister b); 2042 inline void vsubuhs( VectorRegister d, VectorRegister a, VectorRegister b); 2043 inline void vmulesb( VectorRegister d, VectorRegister a, VectorRegister b); 2044 inline void vmuleub( VectorRegister d, VectorRegister a, VectorRegister b); 2045 inline void vmulesh( VectorRegister d, VectorRegister a, VectorRegister b); 2046 inline void vmuleuh( VectorRegister d, VectorRegister a, VectorRegister b); 2047 inline void vmulosb( VectorRegister d, VectorRegister a, VectorRegister b); 2048 inline void vmuloub( VectorRegister d, VectorRegister a, VectorRegister b); 2049 inline void vmulosh( VectorRegister d, VectorRegister a, VectorRegister b); 2050 inline void vmulouh( VectorRegister d, VectorRegister a, VectorRegister b); 2051 inline void vmhaddshs(VectorRegister d, VectorRegister a, VectorRegister b, VectorRegister c); 2052 inline void vmhraddshs(VectorRegister d,VectorRegister a, VectorRegister b, VectorRegister c); 2053 inline void vmladduhm(VectorRegister d, VectorRegister a, VectorRegister b, VectorRegister c); 2054 inline void vmsubuhm( VectorRegister d, VectorRegister a, VectorRegister b, VectorRegister c); 2055 inline void vmsummbm( VectorRegister d, VectorRegister a, VectorRegister b, VectorRegister c); 2056 inline void vmsumshm( VectorRegister d, VectorRegister a, VectorRegister b, VectorRegister c); 2057 inline void vmsumshs( VectorRegister d, VectorRegister a, VectorRegister b, VectorRegister c); 2058 inline void vmsumuhm( VectorRegister d, VectorRegister a, VectorRegister b, VectorRegister c); 2059 inline void vmsumuhs( VectorRegister d, VectorRegister a, VectorRegister b, VectorRegister c); 2060 inline void vsumsws( VectorRegister d, VectorRegister a, VectorRegister b); 2061 inline void vsum2sws( VectorRegister d, VectorRegister a, VectorRegister b); 2062 inline void vsum4sbs( VectorRegister d, VectorRegister a, VectorRegister b); 2063 inline void vsum4ubs( VectorRegister d, VectorRegister a, VectorRegister b); 2064 inline void vsum4shs( VectorRegister d, VectorRegister a, VectorRegister b); 2065 inline void vavgsb( VectorRegister d, VectorRegister a, VectorRegister b); 2066 inline void vavgsw( VectorRegister d, VectorRegister a, VectorRegister b); 2067 inline void vavgsh( VectorRegister d, VectorRegister a, VectorRegister b); 2068 inline void vavgub( VectorRegister d, VectorRegister a, VectorRegister b); 2069 inline void vavguw( VectorRegister d, VectorRegister a, VectorRegister b); 2070 inline void vavguh( VectorRegister d, VectorRegister a, VectorRegister b); 2071 inline void vmaxsb( VectorRegister d, VectorRegister a, VectorRegister b); 2072 inline void vmaxsw( VectorRegister d, VectorRegister a, VectorRegister b); 2073 inline void vmaxsh( VectorRegister d, VectorRegister a, VectorRegister b); 2074 inline void vmaxub( VectorRegister d, VectorRegister a, VectorRegister b); 2075 inline void vmaxuw( VectorRegister d, VectorRegister a, VectorRegister b); 2076 inline void vmaxuh( VectorRegister d, VectorRegister a, VectorRegister b); 2077 inline void vminsb( VectorRegister d, VectorRegister a, VectorRegister b); 2078 inline void vminsw( VectorRegister d, VectorRegister a, VectorRegister b); 2079 inline void vminsh( VectorRegister d, VectorRegister a, VectorRegister b); 2080 inline void vminub( VectorRegister d, VectorRegister a, VectorRegister b); 2081 inline void vminuw( VectorRegister d, VectorRegister a, VectorRegister b); 2082 inline void vminuh( VectorRegister d, VectorRegister a, VectorRegister b); 2083 inline void vcmpequb( VectorRegister d, VectorRegister a, VectorRegister b); 2084 inline void vcmpequh( VectorRegister d, VectorRegister a, VectorRegister b); 2085 inline void vcmpequw( VectorRegister d, VectorRegister a, VectorRegister b); 2086 inline void vcmpgtsh( VectorRegister d, VectorRegister a, VectorRegister b); 2087 inline void vcmpgtsb( VectorRegister d, VectorRegister a, VectorRegister b); 2088 inline void vcmpgtsw( VectorRegister d, VectorRegister a, VectorRegister b); 2089 inline void vcmpgtub( VectorRegister d, VectorRegister a, VectorRegister b); 2090 inline void vcmpgtuh( VectorRegister d, VectorRegister a, VectorRegister b); 2091 inline void vcmpgtuw( VectorRegister d, VectorRegister a, VectorRegister b); 2092 inline void vcmpequb_(VectorRegister d, VectorRegister a, VectorRegister b); 2093 inline void vcmpequh_(VectorRegister d, VectorRegister a, VectorRegister b); 2094 inline void vcmpequw_(VectorRegister d, VectorRegister a, VectorRegister b); 2095 inline void vcmpgtsh_(VectorRegister d, VectorRegister a, VectorRegister b); 2096 inline void vcmpgtsb_(VectorRegister d, VectorRegister a, VectorRegister b); 2097 inline void vcmpgtsw_(VectorRegister d, VectorRegister a, VectorRegister b); 2098 inline void vcmpgtub_(VectorRegister d, VectorRegister a, VectorRegister b); 2099 inline void vcmpgtuh_(VectorRegister d, VectorRegister a, VectorRegister b); 2100 inline void vcmpgtuw_(VectorRegister d, VectorRegister a, VectorRegister b); 2101 inline void vand( VectorRegister d, VectorRegister a, VectorRegister b); 2102 inline void vandc( VectorRegister d, VectorRegister a, VectorRegister b); 2103 inline void vnor( VectorRegister d, VectorRegister a, VectorRegister b); 2104 inline void vor( VectorRegister d, VectorRegister a, VectorRegister b); 2105 inline void vxor( VectorRegister d, VectorRegister a, VectorRegister b); 2106 inline void vrld( VectorRegister d, VectorRegister a, VectorRegister b); 2107 inline void vrlb( VectorRegister d, VectorRegister a, VectorRegister b); 2108 inline void vrlw( VectorRegister d, VectorRegister a, VectorRegister b); 2109 inline void vrlh( VectorRegister d, VectorRegister a, VectorRegister b); 2110 inline void vslb( VectorRegister d, VectorRegister a, VectorRegister b); 2111 inline void vskw( VectorRegister d, VectorRegister a, VectorRegister b); 2112 inline void vslh( VectorRegister d, VectorRegister a, VectorRegister b); 2113 inline void vsrb( VectorRegister d, VectorRegister a, VectorRegister b); 2114 inline void vsrw( VectorRegister d, VectorRegister a, VectorRegister b); 2115 inline void vsrh( VectorRegister d, VectorRegister a, VectorRegister b); 2116 inline void vsrab( VectorRegister d, VectorRegister a, VectorRegister b); 2117 inline void vsraw( VectorRegister d, VectorRegister a, VectorRegister b); 2118 inline void vsrah( VectorRegister d, VectorRegister a, VectorRegister b); 2119 // Vector Floating-Point not implemented yet 2120 inline void mtvscr( VectorRegister b); 2121 inline void mfvscr( VectorRegister d); 2122 2123 // Vector-Scalar (VSX) instructions. 2124 inline void lxvd2x( VectorSRegister d, Register a); 2125 inline void lxvd2x( VectorSRegister d, Register a, Register b); 2126 inline void stxvd2x( VectorSRegister d, Register a); 2127 inline void stxvd2x( VectorSRegister d, Register a, Register b); 2128 inline void mtvrd( VectorRegister d, Register a); 2129 inline void mfvrd( Register a, VectorRegister d); 2130 2131 // AES (introduced with Power 8) 2132 inline void vcipher( VectorRegister d, VectorRegister a, VectorRegister b); 2133 inline void vcipherlast( VectorRegister d, VectorRegister a, VectorRegister b); 2134 inline void vncipher( VectorRegister d, VectorRegister a, VectorRegister b); 2135 inline void vncipherlast(VectorRegister d, VectorRegister a, VectorRegister b); 2136 inline void vsbox( VectorRegister d, VectorRegister a); 2137 2138 // SHA (introduced with Power 8) 2139 // Not yet implemented. 2140 2141 // Vector Binary Polynomial Multiplication (introduced with Power 8) 2142 inline void vpmsumb( VectorRegister d, VectorRegister a, VectorRegister b); 2143 inline void vpmsumd( VectorRegister d, VectorRegister a, VectorRegister b); 2144 inline void vpmsumh( VectorRegister d, VectorRegister a, VectorRegister b); 2145 inline void vpmsumw( VectorRegister d, VectorRegister a, VectorRegister b); 2146 2147 // Vector Permute and Xor (introduced with Power 8) 2148 inline void vpermxor( VectorRegister d, VectorRegister a, VectorRegister b, VectorRegister c); 2149 2150 // Transactional Memory instructions (introduced with Power 8) 2151 inline void tbegin_(); // R=0 2152 inline void tbeginrot_(); // R=1 Rollback-Only Transaction 2153 inline void tend_(); // A=0 2154 inline void tendall_(); // A=1 2155 inline void tabort_(); 2156 inline void tabort_(Register a); 2157 inline void tabortwc_(int t, Register a, Register b); 2158 inline void tabortwci_(int t, Register a, int si); 2159 inline void tabortdc_(int t, Register a, Register b); 2160 inline void tabortdci_(int t, Register a, int si); 2161 inline void tsuspend_(); // tsr with L=0 2162 inline void tresume_(); // tsr with L=1 2163 inline void tcheck(int f); 2164 2165 static bool is_tbegin(int x) { 2166 return TBEGIN_OPCODE == (x & (0x3f << OPCODE_SHIFT | 0x3ff << 1)); 2167 } 2168 2169 // The following encoders use r0 as second operand. These instructions 2170 // read r0 as '0'. 2171 inline void lwzx( Register d, Register s2); 2172 inline void lwz( Register d, int si16); 2173 inline void lwax( Register d, Register s2); 2174 inline void lwa( Register d, int si16); 2175 inline void lwbrx(Register d, Register s2); 2176 inline void lhzx( Register d, Register s2); 2177 inline void lhz( Register d, int si16); 2178 inline void lhax( Register d, Register s2); 2179 inline void lha( Register d, int si16); 2180 inline void lhbrx(Register d, Register s2); 2181 inline void lbzx( Register d, Register s2); 2182 inline void lbz( Register d, int si16); 2183 inline void ldx( Register d, Register s2); 2184 inline void ld( Register d, int si16); 2185 inline void stwx( Register d, Register s2); 2186 inline void stw( Register d, int si16); 2187 inline void sthx( Register d, Register s2); 2188 inline void sth( Register d, int si16); 2189 inline void stbx( Register d, Register s2); 2190 inline void stb( Register d, int si16); 2191 inline void stdx( Register d, Register s2); 2192 inline void std( Register d, int si16); 2193 2194 // PPC 2, section 3.2.1 Instruction Cache Instructions 2195 inline void icbi( Register s2); 2196 // PPC 2, section 3.2.2 Data Cache Instructions 2197 //inlinevoid dcba( Register s2); // Instruction for embedded processor only. 2198 inline void dcbz( Register s2); 2199 inline void dcbst( Register s2); 2200 inline void dcbf( Register s2); 2201 // dcache read hint 2202 inline void dcbt( Register s2); 2203 inline void dcbtct( Register s2, int ct); 2204 inline void dcbtds( Register s2, int ds); 2205 // dcache write hint 2206 inline void dcbtst( Register s2); 2207 inline void dcbtstct(Register s2, int ct); 2208 2209 // Atomics: use ra0mem to disallow R0 as base. 2210 inline void lbarx_unchecked(Register d, Register b, int eh1); 2211 inline void lharx_unchecked(Register d, Register b, int eh1); 2212 inline void lwarx_unchecked(Register d, Register b, int eh1); 2213 inline void ldarx_unchecked(Register d, Register b, int eh1); 2214 inline void lqarx_unchecked(Register d, Register b, int eh1); 2215 inline void lbarx( Register d, Register b, bool hint_exclusive_access); 2216 inline void lharx( Register d, Register b, bool hint_exclusive_access); 2217 inline void lwarx( Register d, Register b, bool hint_exclusive_access); 2218 inline void ldarx( Register d, Register b, bool hint_exclusive_access); 2219 inline void lqarx( Register d, Register b, bool hint_exclusive_access); 2220 inline void stbcx_(Register s, Register b); 2221 inline void sthcx_(Register s, Register b); 2222 inline void stwcx_(Register s, Register b); 2223 inline void stdcx_(Register s, Register b); 2224 inline void stqcx_(Register s, Register b); 2225 inline void lfs( FloatRegister d, int si16); 2226 inline void lfsx( FloatRegister d, Register b); 2227 inline void lfd( FloatRegister d, int si16); 2228 inline void lfdx( FloatRegister d, Register b); 2229 inline void stfs( FloatRegister s, int si16); 2230 inline void stfsx( FloatRegister s, Register b); 2231 inline void stfd( FloatRegister s, int si16); 2232 inline void stfdx( FloatRegister s, Register b); 2233 inline void lvebx( VectorRegister d, Register s2); 2234 inline void lvehx( VectorRegister d, Register s2); 2235 inline void lvewx( VectorRegister d, Register s2); 2236 inline void lvx( VectorRegister d, Register s2); 2237 inline void lvxl( VectorRegister d, Register s2); 2238 inline void stvebx(VectorRegister d, Register s2); 2239 inline void stvehx(VectorRegister d, Register s2); 2240 inline void stvewx(VectorRegister d, Register s2); 2241 inline void stvx( VectorRegister d, Register s2); 2242 inline void stvxl( VectorRegister d, Register s2); 2243 inline void lvsl( VectorRegister d, Register s2); 2244 inline void lvsr( VectorRegister d, Register s2); 2245 2246 // RegisterOrConstant versions. 2247 // These emitters choose between the versions using two registers and 2248 // those with register and immediate, depending on the content of roc. 2249 // If the constant is not encodable as immediate, instructions to 2250 // load the constant are emitted beforehand. Store instructions need a 2251 // tmp reg if the constant is not encodable as immediate. 2252 // Size unpredictable. 2253 void ld( Register d, RegisterOrConstant roc, Register s1 = noreg); 2254 void lwa( Register d, RegisterOrConstant roc, Register s1 = noreg); 2255 void lwz( Register d, RegisterOrConstant roc, Register s1 = noreg); 2256 void lha( Register d, RegisterOrConstant roc, Register s1 = noreg); 2257 void lhz( Register d, RegisterOrConstant roc, Register s1 = noreg); 2258 void lbz( Register d, RegisterOrConstant roc, Register s1 = noreg); 2259 void std( Register d, RegisterOrConstant roc, Register s1 = noreg, Register tmp = noreg); 2260 void stw( Register d, RegisterOrConstant roc, Register s1 = noreg, Register tmp = noreg); 2261 void sth( Register d, RegisterOrConstant roc, Register s1 = noreg, Register tmp = noreg); 2262 void stb( Register d, RegisterOrConstant roc, Register s1 = noreg, Register tmp = noreg); 2263 void add( Register d, RegisterOrConstant roc, Register s1); 2264 void subf(Register d, RegisterOrConstant roc, Register s1); 2265 void cmpd(ConditionRegister d, RegisterOrConstant roc, Register s1); 2266 // Load pointer d from s1+roc. 2267 void ld_ptr(Register d, RegisterOrConstant roc, Register s1 = noreg) { ld(d, roc, s1); } 2268 2269 // Emit several instructions to load a 64 bit constant. This issues a fixed 2270 // instruction pattern so that the constant can be patched later on. 2271 enum { 2272 load_const_size = 5 * BytesPerInstWord 2273 }; 2274 void load_const(Register d, long a, Register tmp = noreg); 2275 inline void load_const(Register d, void* a, Register tmp = noreg); 2276 inline void load_const(Register d, Label& L, Register tmp = noreg); 2277 inline void load_const(Register d, AddressLiteral& a, Register tmp = noreg); 2278 inline void load_const32(Register d, int i); // load signed int (patchable) 2279 2280 // Load a 64 bit constant, optimized, not identifyable. 2281 // Tmp can be used to increase ILP. Set return_simm16_rest = true to get a 2282 // 16 bit immediate offset. This is useful if the offset can be encoded in 2283 // a succeeding instruction. 2284 int load_const_optimized(Register d, long a, Register tmp = noreg, bool return_simm16_rest = false); 2285 inline int load_const_optimized(Register d, void* a, Register tmp = noreg, bool return_simm16_rest = false) { 2286 return load_const_optimized(d, (long)(unsigned long)a, tmp, return_simm16_rest); 2287 } 2288 2289 // If return_simm16_rest, the return value needs to get added afterwards. 2290 int add_const_optimized(Register d, Register s, long x, Register tmp = R0, bool return_simm16_rest = false); 2291 inline int add_const_optimized(Register d, Register s, void* a, Register tmp = R0, bool return_simm16_rest = false) { 2292 return add_const_optimized(d, s, (long)(unsigned long)a, tmp, return_simm16_rest); 2293 } 2294 2295 // If return_simm16_rest, the return value needs to get added afterwards. 2296 inline int sub_const_optimized(Register d, Register s, long x, Register tmp = R0, bool return_simm16_rest = false) { 2297 return add_const_optimized(d, s, -x, tmp, return_simm16_rest); 2298 } 2299 inline int sub_const_optimized(Register d, Register s, void* a, Register tmp = R0, bool return_simm16_rest = false) { 2300 return sub_const_optimized(d, s, (long)(unsigned long)a, tmp, return_simm16_rest); 2301 } 2302 2303 // Creation 2304 Assembler(CodeBuffer* code) : AbstractAssembler(code) { 2305 #ifdef CHECK_DELAY 2306 delay_state = no_delay; 2307 #endif 2308 } 2309 2310 // Testing 2311 #ifndef PRODUCT 2312 void test_asm(); 2313 #endif 2314 }; 2315 2316 2317 #endif // CPU_PPC_VM_ASSEMBLER_PPC_HPP