1 /* 2 * Copyright (c) 2002, 2018, Oracle and/or its affiliates. All rights reserved. 3 * Copyright (c) 2012, 2017 SAP SE. All rights reserved. 4 * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER. 5 * 6 * This code is free software; you can redistribute it and/or modify it 7 * under the terms of the GNU General Public License version 2 only, as 8 * published by the Free Software Foundation. 9 * 10 * This code is distributed in the hope that it will be useful, but WITHOUT 11 * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or 12 * FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License 13 * version 2 for more details (a copy is included in the LICENSE file that 14 * accompanied this code). 15 * 16 * You should have received a copy of the GNU General Public License version 17 * 2 along with this work; if not, write to the Free Software Foundation, 18 * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA. 19 * 20 * Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA 21 * or visit www.oracle.com if you need additional information or have any 22 * questions. 23 * 24 */ 25 26 #ifndef CPU_PPC_VM_ASSEMBLER_PPC_HPP 27 #define CPU_PPC_VM_ASSEMBLER_PPC_HPP 28 29 #include "asm/register.hpp" 30 31 // Address is an abstraction used to represent a memory location 32 // as used in assembler instructions. 33 // PPC instructions grok either baseReg + indexReg or baseReg + disp. 34 class Address { 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 { 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 { 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 { 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 CMPRB_OPCODE = (31u << OPCODE_SHIFT | 192u << 1), 303 CMPEQB_OPCODE = (31u << OPCODE_SHIFT | 224u << 1), 304 305 ISEL_OPCODE = (31u << OPCODE_SHIFT | 15u << 1), 306 307 // Special purpose registers 308 MTSPR_OPCODE = (31u << OPCODE_SHIFT | 467u << 1), 309 MFSPR_OPCODE = (31u << OPCODE_SHIFT | 339u << 1), 310 311 MTXER_OPCODE = (MTSPR_OPCODE | 1 << SPR_0_4_SHIFT), 312 MFXER_OPCODE = (MFSPR_OPCODE | 1 << SPR_0_4_SHIFT), 313 314 MTDSCR_OPCODE = (MTSPR_OPCODE | 3 << SPR_0_4_SHIFT), 315 MFDSCR_OPCODE = (MFSPR_OPCODE | 3 << SPR_0_4_SHIFT), 316 317 MTLR_OPCODE = (MTSPR_OPCODE | 8 << SPR_0_4_SHIFT), 318 MFLR_OPCODE = (MFSPR_OPCODE | 8 << SPR_0_4_SHIFT), 319 320 MTCTR_OPCODE = (MTSPR_OPCODE | 9 << SPR_0_4_SHIFT), 321 MFCTR_OPCODE = (MFSPR_OPCODE | 9 << SPR_0_4_SHIFT), 322 323 // Attention: Higher and lower half are inserted in reversed order. 324 MTTFHAR_OPCODE = (MTSPR_OPCODE | 4 << SPR_5_9_SHIFT | 0 << SPR_0_4_SHIFT), 325 MFTFHAR_OPCODE = (MFSPR_OPCODE | 4 << SPR_5_9_SHIFT | 0 << SPR_0_4_SHIFT), 326 MTTFIAR_OPCODE = (MTSPR_OPCODE | 4 << SPR_5_9_SHIFT | 1 << SPR_0_4_SHIFT), 327 MFTFIAR_OPCODE = (MFSPR_OPCODE | 4 << SPR_5_9_SHIFT | 1 << SPR_0_4_SHIFT), 328 MTTEXASR_OPCODE = (MTSPR_OPCODE | 4 << SPR_5_9_SHIFT | 2 << SPR_0_4_SHIFT), 329 MFTEXASR_OPCODE = (MFSPR_OPCODE | 4 << SPR_5_9_SHIFT | 2 << SPR_0_4_SHIFT), 330 MTTEXASRU_OPCODE = (MTSPR_OPCODE | 4 << SPR_5_9_SHIFT | 3 << SPR_0_4_SHIFT), 331 MFTEXASRU_OPCODE = (MFSPR_OPCODE | 4 << SPR_5_9_SHIFT | 3 << SPR_0_4_SHIFT), 332 333 MTVRSAVE_OPCODE = (MTSPR_OPCODE | 8 << SPR_5_9_SHIFT | 0 << SPR_0_4_SHIFT), 334 MFVRSAVE_OPCODE = (MFSPR_OPCODE | 8 << SPR_5_9_SHIFT | 0 << SPR_0_4_SHIFT), 335 336 MFTB_OPCODE = (MFSPR_OPCODE | 8 << SPR_5_9_SHIFT | 12 << SPR_0_4_SHIFT), 337 338 MTCRF_OPCODE = (31u << OPCODE_SHIFT | 144u << 1), 339 MFCR_OPCODE = (31u << OPCODE_SHIFT | 19u << 1), 340 MCRF_OPCODE = (19u << OPCODE_SHIFT | 0u << 1), 341 SETB_OPCODE = (31u << OPCODE_SHIFT | 128u << 1), 342 343 // condition register logic instructions 344 CRAND_OPCODE = (19u << OPCODE_SHIFT | 257u << 1), 345 CRNAND_OPCODE = (19u << OPCODE_SHIFT | 225u << 1), 346 CROR_OPCODE = (19u << OPCODE_SHIFT | 449u << 1), 347 CRXOR_OPCODE = (19u << OPCODE_SHIFT | 193u << 1), 348 CRNOR_OPCODE = (19u << OPCODE_SHIFT | 33u << 1), 349 CREQV_OPCODE = (19u << OPCODE_SHIFT | 289u << 1), 350 CRANDC_OPCODE = (19u << OPCODE_SHIFT | 129u << 1), 351 CRORC_OPCODE = (19u << OPCODE_SHIFT | 417u << 1), 352 353 BCLR_OPCODE = (19u << OPCODE_SHIFT | 16u << 1), 354 BXX_OPCODE = (18u << OPCODE_SHIFT), 355 BCXX_OPCODE = (16u << OPCODE_SHIFT), 356 357 // CTR-related opcodes 358 BCCTR_OPCODE = (19u << OPCODE_SHIFT | 528u << 1), 359 360 LWZ_OPCODE = (32u << OPCODE_SHIFT), 361 LWZX_OPCODE = (31u << OPCODE_SHIFT | 23u << 1), 362 LWZU_OPCODE = (33u << OPCODE_SHIFT), 363 LWBRX_OPCODE = (31u << OPCODE_SHIFT | 534 << 1), 364 365 LHA_OPCODE = (42u << OPCODE_SHIFT), 366 LHAX_OPCODE = (31u << OPCODE_SHIFT | 343u << 1), 367 LHAU_OPCODE = (43u << OPCODE_SHIFT), 368 369 LHZ_OPCODE = (40u << OPCODE_SHIFT), 370 LHZX_OPCODE = (31u << OPCODE_SHIFT | 279u << 1), 371 LHZU_OPCODE = (41u << OPCODE_SHIFT), 372 LHBRX_OPCODE = (31u << OPCODE_SHIFT | 790 << 1), 373 374 LBZ_OPCODE = (34u << OPCODE_SHIFT), 375 LBZX_OPCODE = (31u << OPCODE_SHIFT | 87u << 1), 376 LBZU_OPCODE = (35u << OPCODE_SHIFT), 377 378 STW_OPCODE = (36u << OPCODE_SHIFT), 379 STWX_OPCODE = (31u << OPCODE_SHIFT | 151u << 1), 380 STWU_OPCODE = (37u << OPCODE_SHIFT), 381 STWUX_OPCODE = (31u << OPCODE_SHIFT | 183u << 1), 382 STWBRX_OPCODE = (31u << OPCODE_SHIFT | 662u << 1), 383 384 STH_OPCODE = (44u << OPCODE_SHIFT), 385 STHX_OPCODE = (31u << OPCODE_SHIFT | 407u << 1), 386 STHU_OPCODE = (45u << OPCODE_SHIFT), 387 STHBRX_OPCODE = (31u << OPCODE_SHIFT | 918u << 1), 388 389 STB_OPCODE = (38u << OPCODE_SHIFT), 390 STBX_OPCODE = (31u << OPCODE_SHIFT | 215u << 1), 391 STBU_OPCODE = (39u << OPCODE_SHIFT), 392 393 EXTSB_OPCODE = (31u << OPCODE_SHIFT | 954u << 1), 394 EXTSH_OPCODE = (31u << OPCODE_SHIFT | 922u << 1), 395 EXTSW_OPCODE = (31u << OPCODE_SHIFT | 986u << 1), // X-FORM 396 397 // 32 bit opcode encodings 398 399 LWA_OPCODE = (58u << OPCODE_SHIFT | 2u << XO_30_31_SHIFT), // DS-FORM 400 LWAX_OPCODE = (31u << OPCODE_SHIFT | 341u << XO_21_30_SHIFT), // X-FORM 401 402 CNTLZW_OPCODE = (31u << OPCODE_SHIFT | 26u << XO_21_30_SHIFT), // X-FORM 403 CNTTZW_OPCODE = (31u << OPCODE_SHIFT | 538u << XO_21_30_SHIFT), // X-FORM 404 405 // 64 bit opcode encodings 406 407 LD_OPCODE = (58u << OPCODE_SHIFT | 0u << XO_30_31_SHIFT), // DS-FORM 408 LDU_OPCODE = (58u << OPCODE_SHIFT | 1u << XO_30_31_SHIFT), // DS-FORM 409 LDX_OPCODE = (31u << OPCODE_SHIFT | 21u << XO_21_30_SHIFT), // X-FORM 410 LDBRX_OPCODE = (31u << OPCODE_SHIFT | 532u << 1), // X-FORM 411 412 STD_OPCODE = (62u << OPCODE_SHIFT | 0u << XO_30_31_SHIFT), // DS-FORM 413 STDU_OPCODE = (62u << OPCODE_SHIFT | 1u << XO_30_31_SHIFT), // DS-FORM 414 STDUX_OPCODE = (31u << OPCODE_SHIFT | 181u << 1), // X-FORM 415 STDX_OPCODE = (31u << OPCODE_SHIFT | 149u << XO_21_30_SHIFT), // X-FORM 416 STDBRX_OPCODE = (31u << OPCODE_SHIFT | 660u << 1), // X-FORM 417 418 RLDICR_OPCODE = (30u << OPCODE_SHIFT | 1u << XO_27_29_SHIFT), // MD-FORM 419 RLDICL_OPCODE = (30u << OPCODE_SHIFT | 0u << XO_27_29_SHIFT), // MD-FORM 420 RLDIC_OPCODE = (30u << OPCODE_SHIFT | 2u << XO_27_29_SHIFT), // MD-FORM 421 RLDIMI_OPCODE = (30u << OPCODE_SHIFT | 3u << XO_27_29_SHIFT), // MD-FORM 422 423 SRADI_OPCODE = (31u << OPCODE_SHIFT | 413u << XO_21_29_SHIFT), // XS-FORM 424 425 SLD_OPCODE = (31u << OPCODE_SHIFT | 27u << 1), // X-FORM 426 SRD_OPCODE = (31u << OPCODE_SHIFT | 539u << 1), // X-FORM 427 SRAD_OPCODE = (31u << OPCODE_SHIFT | 794u << 1), // X-FORM 428 429 MULLD_OPCODE = (31u << OPCODE_SHIFT | 233u << 1), // XO-FORM 430 MULHD_OPCODE = (31u << OPCODE_SHIFT | 73u << 1), // XO-FORM 431 MULHDU_OPCODE = (31u << OPCODE_SHIFT | 9u << 1), // XO-FORM 432 DIVD_OPCODE = (31u << OPCODE_SHIFT | 489u << 1), // XO-FORM 433 434 CNTLZD_OPCODE = (31u << OPCODE_SHIFT | 58u << XO_21_30_SHIFT), // X-FORM 435 CNTTZD_OPCODE = (31u << OPCODE_SHIFT | 570u << XO_21_30_SHIFT), // X-FORM 436 NAND_OPCODE = (31u << OPCODE_SHIFT | 476u << XO_21_30_SHIFT), // X-FORM 437 NOR_OPCODE = (31u << OPCODE_SHIFT | 124u << XO_21_30_SHIFT), // X-FORM 438 439 440 // opcodes only used for floating arithmetic 441 FADD_OPCODE = (63u << OPCODE_SHIFT | 21u << 1), 442 FADDS_OPCODE = (59u << OPCODE_SHIFT | 21u << 1), 443 FCMPU_OPCODE = (63u << OPCODE_SHIFT | 00u << 1), 444 FDIV_OPCODE = (63u << OPCODE_SHIFT | 18u << 1), 445 FDIVS_OPCODE = (59u << OPCODE_SHIFT | 18u << 1), 446 FMR_OPCODE = (63u << OPCODE_SHIFT | 72u << 1), 447 // These are special Power6 opcodes, reused for "lfdepx" and "stfdepx" 448 // on Power7. Do not use. 449 // MFFGPR_OPCODE = (31u << OPCODE_SHIFT | 607u << 1), 450 // MFTGPR_OPCODE = (31u << OPCODE_SHIFT | 735u << 1), 451 CMPB_OPCODE = (31u << OPCODE_SHIFT | 508 << 1), 452 POPCNTB_OPCODE = (31u << OPCODE_SHIFT | 122 << 1), 453 POPCNTW_OPCODE = (31u << OPCODE_SHIFT | 378 << 1), 454 POPCNTD_OPCODE = (31u << OPCODE_SHIFT | 506 << 1), 455 FABS_OPCODE = (63u << OPCODE_SHIFT | 264u << 1), 456 FNABS_OPCODE = (63u << OPCODE_SHIFT | 136u << 1), 457 FMUL_OPCODE = (63u << OPCODE_SHIFT | 25u << 1), 458 FMULS_OPCODE = (59u << OPCODE_SHIFT | 25u << 1), 459 FNEG_OPCODE = (63u << OPCODE_SHIFT | 40u << 1), 460 FSUB_OPCODE = (63u << OPCODE_SHIFT | 20u << 1), 461 FSUBS_OPCODE = (59u << OPCODE_SHIFT | 20u << 1), 462 463 // PPC64-internal FPU conversion opcodes 464 FCFID_OPCODE = (63u << OPCODE_SHIFT | 846u << 1), 465 FCFIDS_OPCODE = (59u << OPCODE_SHIFT | 846u << 1), 466 FCTID_OPCODE = (63u << OPCODE_SHIFT | 814u << 1), 467 FCTIDZ_OPCODE = (63u << OPCODE_SHIFT | 815u << 1), 468 FCTIW_OPCODE = (63u << OPCODE_SHIFT | 14u << 1), 469 FCTIWZ_OPCODE = (63u << OPCODE_SHIFT | 15u << 1), 470 FRSP_OPCODE = (63u << OPCODE_SHIFT | 12u << 1), 471 472 // Fused multiply-accumulate instructions. 473 FMADD_OPCODE = (63u << OPCODE_SHIFT | 29u << 1), 474 FMADDS_OPCODE = (59u << OPCODE_SHIFT | 29u << 1), 475 FMSUB_OPCODE = (63u << OPCODE_SHIFT | 28u << 1), 476 FMSUBS_OPCODE = (59u << OPCODE_SHIFT | 28u << 1), 477 FNMADD_OPCODE = (63u << OPCODE_SHIFT | 31u << 1), 478 FNMADDS_OPCODE = (59u << OPCODE_SHIFT | 31u << 1), 479 FNMSUB_OPCODE = (63u << OPCODE_SHIFT | 30u << 1), 480 FNMSUBS_OPCODE = (59u << OPCODE_SHIFT | 30u << 1), 481 482 LFD_OPCODE = (50u << OPCODE_SHIFT | 00u << 1), 483 LFDU_OPCODE = (51u << OPCODE_SHIFT | 00u << 1), 484 LFDX_OPCODE = (31u << OPCODE_SHIFT | 599u << 1), 485 LFS_OPCODE = (48u << OPCODE_SHIFT | 00u << 1), 486 LFSU_OPCODE = (49u << OPCODE_SHIFT | 00u << 1), 487 LFSX_OPCODE = (31u << OPCODE_SHIFT | 535u << 1), 488 489 STFD_OPCODE = (54u << OPCODE_SHIFT | 00u << 1), 490 STFDU_OPCODE = (55u << OPCODE_SHIFT | 00u << 1), 491 STFDX_OPCODE = (31u << OPCODE_SHIFT | 727u << 1), 492 STFS_OPCODE = (52u << OPCODE_SHIFT | 00u << 1), 493 STFSU_OPCODE = (53u << OPCODE_SHIFT | 00u << 1), 494 STFSX_OPCODE = (31u << OPCODE_SHIFT | 663u << 1), 495 496 FSQRT_OPCODE = (63u << OPCODE_SHIFT | 22u << 1), // A-FORM 497 FSQRTS_OPCODE = (59u << OPCODE_SHIFT | 22u << 1), // A-FORM 498 499 // Vector instruction support for >= Power6 500 // Vector Storage Access 501 LVEBX_OPCODE = (31u << OPCODE_SHIFT | 7u << 1), 502 LVEHX_OPCODE = (31u << OPCODE_SHIFT | 39u << 1), 503 LVEWX_OPCODE = (31u << OPCODE_SHIFT | 71u << 1), 504 LVX_OPCODE = (31u << OPCODE_SHIFT | 103u << 1), 505 LVXL_OPCODE = (31u << OPCODE_SHIFT | 359u << 1), 506 STVEBX_OPCODE = (31u << OPCODE_SHIFT | 135u << 1), 507 STVEHX_OPCODE = (31u << OPCODE_SHIFT | 167u << 1), 508 STVEWX_OPCODE = (31u << OPCODE_SHIFT | 199u << 1), 509 STVX_OPCODE = (31u << OPCODE_SHIFT | 231u << 1), 510 STVXL_OPCODE = (31u << OPCODE_SHIFT | 487u << 1), 511 LVSL_OPCODE = (31u << OPCODE_SHIFT | 6u << 1), 512 LVSR_OPCODE = (31u << OPCODE_SHIFT | 38u << 1), 513 514 // Vector-Scalar (VSX) instruction support. 515 LXVD2X_OPCODE = (31u << OPCODE_SHIFT | 844u << 1), 516 STXVD2X_OPCODE = (31u << OPCODE_SHIFT | 972u << 1), 517 MTVSRD_OPCODE = (31u << OPCODE_SHIFT | 179u << 1), 518 MTVSRWZ_OPCODE = (31u << OPCODE_SHIFT | 243u << 1), 519 MFVSRD_OPCODE = (31u << OPCODE_SHIFT | 51u << 1), 520 MTVSRWA_OPCODE = (31u << OPCODE_SHIFT | 211u << 1), 521 MFVSRWZ_OPCODE = (31u << OPCODE_SHIFT | 115u << 1), 522 XXPERMDI_OPCODE= (60u << OPCODE_SHIFT | 10u << 3), 523 XXMRGHW_OPCODE = (60u << OPCODE_SHIFT | 18u << 3), 524 XXMRGLW_OPCODE = (60u << OPCODE_SHIFT | 50u << 3), 525 XXSPLTW_OPCODE = (60u << OPCODE_SHIFT | 164u << 2), 526 XXLOR_OPCODE = (60u << OPCODE_SHIFT | 146u << 3), 527 XXLXOR_OPCODE = (60u << OPCODE_SHIFT | 154u << 3), 528 XXLEQV_OPCODE = (60u << OPCODE_SHIFT | 186u << 3), 529 XVDIVSP_OPCODE = (60u << OPCODE_SHIFT | 88u << 3), 530 XVDIVDP_OPCODE = (60u << OPCODE_SHIFT | 120u << 3), 531 XVABSSP_OPCODE = (60u << OPCODE_SHIFT | 409u << 2), 532 XVABSDP_OPCODE = (60u << OPCODE_SHIFT | 473u << 2), 533 XVNEGSP_OPCODE = (60u << OPCODE_SHIFT | 441u << 2), 534 XVNEGDP_OPCODE = (60u << OPCODE_SHIFT | 505u << 2), 535 XVSQRTSP_OPCODE= (60u << OPCODE_SHIFT | 139u << 2), 536 XVSQRTDP_OPCODE= (60u << OPCODE_SHIFT | 203u << 2), 537 XSCVDPSPN_OPCODE=(60u << OPCODE_SHIFT | 267u << 2), 538 XVADDDP_OPCODE = (60u << OPCODE_SHIFT | 96u << 3), 539 XVSUBDP_OPCODE = (60u << OPCODE_SHIFT | 104u << 3), 540 XVMULSP_OPCODE = (60u << OPCODE_SHIFT | 80u << 3), 541 XVMULDP_OPCODE = (60u << OPCODE_SHIFT | 112u << 3), 542 543 // Deliver A Random Number (introduced with POWER9) 544 DARN_OPCODE = (31u << OPCODE_SHIFT | 755u << 1), 545 546 // Vector Permute and Formatting 547 VPKPX_OPCODE = (4u << OPCODE_SHIFT | 782u ), 548 VPKSHSS_OPCODE = (4u << OPCODE_SHIFT | 398u ), 549 VPKSWSS_OPCODE = (4u << OPCODE_SHIFT | 462u ), 550 VPKSHUS_OPCODE = (4u << OPCODE_SHIFT | 270u ), 551 VPKSWUS_OPCODE = (4u << OPCODE_SHIFT | 334u ), 552 VPKUHUM_OPCODE = (4u << OPCODE_SHIFT | 14u ), 553 VPKUWUM_OPCODE = (4u << OPCODE_SHIFT | 78u ), 554 VPKUHUS_OPCODE = (4u << OPCODE_SHIFT | 142u ), 555 VPKUWUS_OPCODE = (4u << OPCODE_SHIFT | 206u ), 556 VUPKHPX_OPCODE = (4u << OPCODE_SHIFT | 846u ), 557 VUPKHSB_OPCODE = (4u << OPCODE_SHIFT | 526u ), 558 VUPKHSH_OPCODE = (4u << OPCODE_SHIFT | 590u ), 559 VUPKLPX_OPCODE = (4u << OPCODE_SHIFT | 974u ), 560 VUPKLSB_OPCODE = (4u << OPCODE_SHIFT | 654u ), 561 VUPKLSH_OPCODE = (4u << OPCODE_SHIFT | 718u ), 562 563 VMRGHB_OPCODE = (4u << OPCODE_SHIFT | 12u ), 564 VMRGHW_OPCODE = (4u << OPCODE_SHIFT | 140u ), 565 VMRGHH_OPCODE = (4u << OPCODE_SHIFT | 76u ), 566 VMRGLB_OPCODE = (4u << OPCODE_SHIFT | 268u ), 567 VMRGLW_OPCODE = (4u << OPCODE_SHIFT | 396u ), 568 VMRGLH_OPCODE = (4u << OPCODE_SHIFT | 332u ), 569 570 VSPLT_OPCODE = (4u << OPCODE_SHIFT | 524u ), 571 VSPLTH_OPCODE = (4u << OPCODE_SHIFT | 588u ), 572 VSPLTW_OPCODE = (4u << OPCODE_SHIFT | 652u ), 573 VSPLTISB_OPCODE= (4u << OPCODE_SHIFT | 780u ), 574 VSPLTISH_OPCODE= (4u << OPCODE_SHIFT | 844u ), 575 VSPLTISW_OPCODE= (4u << OPCODE_SHIFT | 908u ), 576 577 VPERM_OPCODE = (4u << OPCODE_SHIFT | 43u ), 578 VSEL_OPCODE = (4u << OPCODE_SHIFT | 42u ), 579 580 VSL_OPCODE = (4u << OPCODE_SHIFT | 452u ), 581 VSLDOI_OPCODE = (4u << OPCODE_SHIFT | 44u ), 582 VSLO_OPCODE = (4u << OPCODE_SHIFT | 1036u ), 583 VSR_OPCODE = (4u << OPCODE_SHIFT | 708u ), 584 VSRO_OPCODE = (4u << OPCODE_SHIFT | 1100u ), 585 586 // Vector Integer 587 VADDCUW_OPCODE = (4u << OPCODE_SHIFT | 384u ), 588 VADDSHS_OPCODE = (4u << OPCODE_SHIFT | 832u ), 589 VADDSBS_OPCODE = (4u << OPCODE_SHIFT | 768u ), 590 VADDSWS_OPCODE = (4u << OPCODE_SHIFT | 896u ), 591 VADDUBM_OPCODE = (4u << OPCODE_SHIFT | 0u ), 592 VADDUWM_OPCODE = (4u << OPCODE_SHIFT | 128u ), 593 VADDUHM_OPCODE = (4u << OPCODE_SHIFT | 64u ), 594 VADDUDM_OPCODE = (4u << OPCODE_SHIFT | 192u ), 595 VADDUBS_OPCODE = (4u << OPCODE_SHIFT | 512u ), 596 VADDUWS_OPCODE = (4u << OPCODE_SHIFT | 640u ), 597 VADDUHS_OPCODE = (4u << OPCODE_SHIFT | 576u ), 598 VADDFP_OPCODE = (4u << OPCODE_SHIFT | 10u ), 599 VSUBCUW_OPCODE = (4u << OPCODE_SHIFT | 1408u ), 600 VSUBSHS_OPCODE = (4u << OPCODE_SHIFT | 1856u ), 601 VSUBSBS_OPCODE = (4u << OPCODE_SHIFT | 1792u ), 602 VSUBSWS_OPCODE = (4u << OPCODE_SHIFT | 1920u ), 603 VSUBUBM_OPCODE = (4u << OPCODE_SHIFT | 1024u ), 604 VSUBUWM_OPCODE = (4u << OPCODE_SHIFT | 1152u ), 605 VSUBUHM_OPCODE = (4u << OPCODE_SHIFT | 1088u ), 606 VSUBUDM_OPCODE = (4u << OPCODE_SHIFT | 1216u ), 607 VSUBUBS_OPCODE = (4u << OPCODE_SHIFT | 1536u ), 608 VSUBUWS_OPCODE = (4u << OPCODE_SHIFT | 1664u ), 609 VSUBUHS_OPCODE = (4u << OPCODE_SHIFT | 1600u ), 610 VSUBFP_OPCODE = (4u << OPCODE_SHIFT | 74u ), 611 612 VMULESB_OPCODE = (4u << OPCODE_SHIFT | 776u ), 613 VMULEUB_OPCODE = (4u << OPCODE_SHIFT | 520u ), 614 VMULESH_OPCODE = (4u << OPCODE_SHIFT | 840u ), 615 VMULEUH_OPCODE = (4u << OPCODE_SHIFT | 584u ), 616 VMULOSB_OPCODE = (4u << OPCODE_SHIFT | 264u ), 617 VMULOUB_OPCODE = (4u << OPCODE_SHIFT | 8u ), 618 VMULOSH_OPCODE = (4u << OPCODE_SHIFT | 328u ), 619 VMULOSW_OPCODE = (4u << OPCODE_SHIFT | 392u ), 620 VMULOUH_OPCODE = (4u << OPCODE_SHIFT | 72u ), 621 VMULUWM_OPCODE = (4u << OPCODE_SHIFT | 137u ), 622 VMHADDSHS_OPCODE=(4u << OPCODE_SHIFT | 32u ), 623 VMHRADDSHS_OPCODE=(4u << OPCODE_SHIFT | 33u ), 624 VMLADDUHM_OPCODE=(4u << OPCODE_SHIFT | 34u ), 625 VMSUBUHM_OPCODE= (4u << OPCODE_SHIFT | 36u ), 626 VMSUMMBM_OPCODE= (4u << OPCODE_SHIFT | 37u ), 627 VMSUMSHM_OPCODE= (4u << OPCODE_SHIFT | 40u ), 628 VMSUMSHS_OPCODE= (4u << OPCODE_SHIFT | 41u ), 629 VMSUMUHM_OPCODE= (4u << OPCODE_SHIFT | 38u ), 630 VMSUMUHS_OPCODE= (4u << OPCODE_SHIFT | 39u ), 631 VMADDFP_OPCODE = (4u << OPCODE_SHIFT | 46u ), 632 633 VSUMSWS_OPCODE = (4u << OPCODE_SHIFT | 1928u ), 634 VSUM2SWS_OPCODE= (4u << OPCODE_SHIFT | 1672u ), 635 VSUM4SBS_OPCODE= (4u << OPCODE_SHIFT | 1800u ), 636 VSUM4UBS_OPCODE= (4u << OPCODE_SHIFT | 1544u ), 637 VSUM4SHS_OPCODE= (4u << OPCODE_SHIFT | 1608u ), 638 639 VAVGSB_OPCODE = (4u << OPCODE_SHIFT | 1282u ), 640 VAVGSW_OPCODE = (4u << OPCODE_SHIFT | 1410u ), 641 VAVGSH_OPCODE = (4u << OPCODE_SHIFT | 1346u ), 642 VAVGUB_OPCODE = (4u << OPCODE_SHIFT | 1026u ), 643 VAVGUW_OPCODE = (4u << OPCODE_SHIFT | 1154u ), 644 VAVGUH_OPCODE = (4u << OPCODE_SHIFT | 1090u ), 645 646 VMAXSB_OPCODE = (4u << OPCODE_SHIFT | 258u ), 647 VMAXSW_OPCODE = (4u << OPCODE_SHIFT | 386u ), 648 VMAXSH_OPCODE = (4u << OPCODE_SHIFT | 322u ), 649 VMAXUB_OPCODE = (4u << OPCODE_SHIFT | 2u ), 650 VMAXUW_OPCODE = (4u << OPCODE_SHIFT | 130u ), 651 VMAXUH_OPCODE = (4u << OPCODE_SHIFT | 66u ), 652 VMINSB_OPCODE = (4u << OPCODE_SHIFT | 770u ), 653 VMINSW_OPCODE = (4u << OPCODE_SHIFT | 898u ), 654 VMINSH_OPCODE = (4u << OPCODE_SHIFT | 834u ), 655 VMINUB_OPCODE = (4u << OPCODE_SHIFT | 514u ), 656 VMINUW_OPCODE = (4u << OPCODE_SHIFT | 642u ), 657 VMINUH_OPCODE = (4u << OPCODE_SHIFT | 578u ), 658 659 VCMPEQUB_OPCODE= (4u << OPCODE_SHIFT | 6u ), 660 VCMPEQUH_OPCODE= (4u << OPCODE_SHIFT | 70u ), 661 VCMPEQUW_OPCODE= (4u << OPCODE_SHIFT | 134u ), 662 VCMPGTSH_OPCODE= (4u << OPCODE_SHIFT | 838u ), 663 VCMPGTSB_OPCODE= (4u << OPCODE_SHIFT | 774u ), 664 VCMPGTSW_OPCODE= (4u << OPCODE_SHIFT | 902u ), 665 VCMPGTUB_OPCODE= (4u << OPCODE_SHIFT | 518u ), 666 VCMPGTUH_OPCODE= (4u << OPCODE_SHIFT | 582u ), 667 VCMPGTUW_OPCODE= (4u << OPCODE_SHIFT | 646u ), 668 669 VAND_OPCODE = (4u << OPCODE_SHIFT | 1028u ), 670 VANDC_OPCODE = (4u << OPCODE_SHIFT | 1092u ), 671 VNOR_OPCODE = (4u << OPCODE_SHIFT | 1284u ), 672 VOR_OPCODE = (4u << OPCODE_SHIFT | 1156u ), 673 VXOR_OPCODE = (4u << OPCODE_SHIFT | 1220u ), 674 VRLD_OPCODE = (4u << OPCODE_SHIFT | 196u ), 675 VRLB_OPCODE = (4u << OPCODE_SHIFT | 4u ), 676 VRLW_OPCODE = (4u << OPCODE_SHIFT | 132u ), 677 VRLH_OPCODE = (4u << OPCODE_SHIFT | 68u ), 678 VSLB_OPCODE = (4u << OPCODE_SHIFT | 260u ), 679 VSKW_OPCODE = (4u << OPCODE_SHIFT | 388u ), 680 VSLH_OPCODE = (4u << OPCODE_SHIFT | 324u ), 681 VSRB_OPCODE = (4u << OPCODE_SHIFT | 516u ), 682 VSRW_OPCODE = (4u << OPCODE_SHIFT | 644u ), 683 VSRH_OPCODE = (4u << OPCODE_SHIFT | 580u ), 684 VSRAB_OPCODE = (4u << OPCODE_SHIFT | 772u ), 685 VSRAW_OPCODE = (4u << OPCODE_SHIFT | 900u ), 686 VSRAH_OPCODE = (4u << OPCODE_SHIFT | 836u ), 687 VPOPCNTW_OPCODE= (4u << OPCODE_SHIFT | 1923u ), 688 689 // Vector Floating-Point 690 // not implemented yet 691 692 // Vector Status and Control 693 MTVSCR_OPCODE = (4u << OPCODE_SHIFT | 1604u ), 694 MFVSCR_OPCODE = (4u << OPCODE_SHIFT | 1540u ), 695 696 // AES (introduced with Power 8) 697 VCIPHER_OPCODE = (4u << OPCODE_SHIFT | 1288u), 698 VCIPHERLAST_OPCODE = (4u << OPCODE_SHIFT | 1289u), 699 VNCIPHER_OPCODE = (4u << OPCODE_SHIFT | 1352u), 700 VNCIPHERLAST_OPCODE = (4u << OPCODE_SHIFT | 1353u), 701 VSBOX_OPCODE = (4u << OPCODE_SHIFT | 1480u), 702 703 // SHA (introduced with Power 8) 704 VSHASIGMAD_OPCODE = (4u << OPCODE_SHIFT | 1730u), 705 VSHASIGMAW_OPCODE = (4u << OPCODE_SHIFT | 1666u), 706 707 // Vector Binary Polynomial Multiplication (introduced with Power 8) 708 VPMSUMB_OPCODE = (4u << OPCODE_SHIFT | 1032u), 709 VPMSUMD_OPCODE = (4u << OPCODE_SHIFT | 1224u), 710 VPMSUMH_OPCODE = (4u << OPCODE_SHIFT | 1096u), 711 VPMSUMW_OPCODE = (4u << OPCODE_SHIFT | 1160u), 712 713 // Vector Permute and Xor (introduced with Power 8) 714 VPERMXOR_OPCODE = (4u << OPCODE_SHIFT | 45u), 715 716 // Transactional Memory instructions (introduced with Power 8) 717 TBEGIN_OPCODE = (31u << OPCODE_SHIFT | 654u << 1), 718 TEND_OPCODE = (31u << OPCODE_SHIFT | 686u << 1), 719 TABORT_OPCODE = (31u << OPCODE_SHIFT | 910u << 1), 720 TABORTWC_OPCODE = (31u << OPCODE_SHIFT | 782u << 1), 721 TABORTWCI_OPCODE = (31u << OPCODE_SHIFT | 846u << 1), 722 TABORTDC_OPCODE = (31u << OPCODE_SHIFT | 814u << 1), 723 TABORTDCI_OPCODE = (31u << OPCODE_SHIFT | 878u << 1), 724 TSR_OPCODE = (31u << OPCODE_SHIFT | 750u << 1), 725 TCHECK_OPCODE = (31u << OPCODE_SHIFT | 718u << 1), 726 727 // Icache and dcache related instructions 728 DCBA_OPCODE = (31u << OPCODE_SHIFT | 758u << 1), 729 DCBZ_OPCODE = (31u << OPCODE_SHIFT | 1014u << 1), 730 DCBST_OPCODE = (31u << OPCODE_SHIFT | 54u << 1), 731 DCBF_OPCODE = (31u << OPCODE_SHIFT | 86u << 1), 732 733 DCBT_OPCODE = (31u << OPCODE_SHIFT | 278u << 1), 734 DCBTST_OPCODE = (31u << OPCODE_SHIFT | 246u << 1), 735 ICBI_OPCODE = (31u << OPCODE_SHIFT | 982u << 1), 736 737 // Instruction synchronization 738 ISYNC_OPCODE = (19u << OPCODE_SHIFT | 150u << 1), 739 // Memory barriers 740 SYNC_OPCODE = (31u << OPCODE_SHIFT | 598u << 1), 741 EIEIO_OPCODE = (31u << OPCODE_SHIFT | 854u << 1), 742 743 // Wait instructions for polling. 744 WAIT_OPCODE = (31u << OPCODE_SHIFT | 62u << 1), 745 746 // Trap instructions 747 TDI_OPCODE = (2u << OPCODE_SHIFT), 748 TWI_OPCODE = (3u << OPCODE_SHIFT), 749 TD_OPCODE = (31u << OPCODE_SHIFT | 68u << 1), 750 TW_OPCODE = (31u << OPCODE_SHIFT | 4u << 1), 751 752 // Atomics. 753 LBARX_OPCODE = (31u << OPCODE_SHIFT | 52u << 1), 754 LHARX_OPCODE = (31u << OPCODE_SHIFT | 116u << 1), 755 LWARX_OPCODE = (31u << OPCODE_SHIFT | 20u << 1), 756 LDARX_OPCODE = (31u << OPCODE_SHIFT | 84u << 1), 757 LQARX_OPCODE = (31u << OPCODE_SHIFT | 276u << 1), 758 STBCX_OPCODE = (31u << OPCODE_SHIFT | 694u << 1), 759 STHCX_OPCODE = (31u << OPCODE_SHIFT | 726u << 1), 760 STWCX_OPCODE = (31u << OPCODE_SHIFT | 150u << 1), 761 STDCX_OPCODE = (31u << OPCODE_SHIFT | 214u << 1), 762 STQCX_OPCODE = (31u << OPCODE_SHIFT | 182u << 1) 763 764 }; 765 766 // Trap instructions TO bits 767 enum trap_to_bits { 768 // single bits 769 traptoLessThanSigned = 1 << 4, // 0, left end 770 traptoGreaterThanSigned = 1 << 3, 771 traptoEqual = 1 << 2, 772 traptoLessThanUnsigned = 1 << 1, 773 traptoGreaterThanUnsigned = 1 << 0, // 4, right end 774 775 // compound ones 776 traptoUnconditional = (traptoLessThanSigned | 777 traptoGreaterThanSigned | 778 traptoEqual | 779 traptoLessThanUnsigned | 780 traptoGreaterThanUnsigned) 781 }; 782 783 // Branch hints BH field 784 enum branch_hint_bh { 785 // bclr cases: 786 bhintbhBCLRisReturn = 0, 787 bhintbhBCLRisNotReturnButSame = 1, 788 bhintbhBCLRisNotPredictable = 3, 789 790 // bcctr cases: 791 bhintbhBCCTRisNotReturnButSame = 0, 792 bhintbhBCCTRisNotPredictable = 3 793 }; 794 795 // Branch prediction hints AT field 796 enum branch_hint_at { 797 bhintatNoHint = 0, // at=00 798 bhintatIsNotTaken = 2, // at=10 799 bhintatIsTaken = 3 // at=11 800 }; 801 802 // Branch prediction hints 803 enum branch_hint_concept { 804 // Use the same encoding as branch_hint_at to simply code. 805 bhintNoHint = bhintatNoHint, 806 bhintIsNotTaken = bhintatIsNotTaken, 807 bhintIsTaken = bhintatIsTaken 808 }; 809 810 // Used in BO field of branch instruction. 811 enum branch_condition { 812 bcondCRbiIs0 = 4, // bo=001at 813 bcondCRbiIs1 = 12, // bo=011at 814 bcondAlways = 20 // bo=10100 815 }; 816 817 // Branch condition with combined prediction hints. 818 enum branch_condition_with_hint { 819 bcondCRbiIs0_bhintNoHint = bcondCRbiIs0 | bhintatNoHint, 820 bcondCRbiIs0_bhintIsNotTaken = bcondCRbiIs0 | bhintatIsNotTaken, 821 bcondCRbiIs0_bhintIsTaken = bcondCRbiIs0 | bhintatIsTaken, 822 bcondCRbiIs1_bhintNoHint = bcondCRbiIs1 | bhintatNoHint, 823 bcondCRbiIs1_bhintIsNotTaken = bcondCRbiIs1 | bhintatIsNotTaken, 824 bcondCRbiIs1_bhintIsTaken = bcondCRbiIs1 | bhintatIsTaken, 825 }; 826 827 // Elemental Memory Barriers (>=Power 8) 828 enum Elemental_Membar_mask_bits { 829 StoreStore = 1 << 0, 830 StoreLoad = 1 << 1, 831 LoadStore = 1 << 2, 832 LoadLoad = 1 << 3 833 }; 834 835 // Branch prediction hints. 836 inline static int add_bhint_to_boint(const int bhint, const int boint) { 837 switch (boint) { 838 case bcondCRbiIs0: 839 case bcondCRbiIs1: 840 // branch_hint and branch_hint_at have same encodings 841 assert( (int)bhintNoHint == (int)bhintatNoHint 842 && (int)bhintIsNotTaken == (int)bhintatIsNotTaken 843 && (int)bhintIsTaken == (int)bhintatIsTaken, 844 "wrong encodings"); 845 assert((bhint & 0x03) == bhint, "wrong encodings"); 846 return (boint & ~0x03) | bhint; 847 case bcondAlways: 848 // no branch_hint 849 return boint; 850 default: 851 ShouldNotReachHere(); 852 return 0; 853 } 854 } 855 856 // Extract bcond from boint. 857 inline static int inv_boint_bcond(const int boint) { 858 int r_bcond = boint & ~0x03; 859 assert(r_bcond == bcondCRbiIs0 || 860 r_bcond == bcondCRbiIs1 || 861 r_bcond == bcondAlways, 862 "bad branch condition"); 863 return r_bcond; 864 } 865 866 // Extract bhint from boint. 867 inline static int inv_boint_bhint(const int boint) { 868 int r_bhint = boint & 0x03; 869 assert(r_bhint == bhintatNoHint || 870 r_bhint == bhintatIsNotTaken || 871 r_bhint == bhintatIsTaken, 872 "bad branch hint"); 873 return r_bhint; 874 } 875 876 // Calculate opposite of given bcond. 877 inline static int opposite_bcond(const int bcond) { 878 switch (bcond) { 879 case bcondCRbiIs0: 880 return bcondCRbiIs1; 881 case bcondCRbiIs1: 882 return bcondCRbiIs0; 883 default: 884 ShouldNotReachHere(); 885 return 0; 886 } 887 } 888 889 // Calculate opposite of given bhint. 890 inline static int opposite_bhint(const int bhint) { 891 switch (bhint) { 892 case bhintatNoHint: 893 return bhintatNoHint; 894 case bhintatIsNotTaken: 895 return bhintatIsTaken; 896 case bhintatIsTaken: 897 return bhintatIsNotTaken; 898 default: 899 ShouldNotReachHere(); 900 return 0; 901 } 902 } 903 904 // PPC branch instructions 905 enum ppcops { 906 b_op = 18, 907 bc_op = 16, 908 bcr_op = 19 909 }; 910 911 enum Condition { 912 negative = 0, 913 less = 0, 914 positive = 1, 915 greater = 1, 916 zero = 2, 917 equal = 2, 918 summary_overflow = 3, 919 }; 920 921 public: 922 // Helper functions for groups of instructions 923 924 enum Predict { pt = 1, pn = 0 }; // pt = predict taken 925 926 // Instruction must start at passed address. 927 static int instr_len(unsigned char *instr) { return BytesPerInstWord; } 928 929 // longest instructions 930 static int instr_maxlen() { return BytesPerInstWord; } 931 932 // Test if x is within signed immediate range for nbits. 933 static bool is_simm(int x, unsigned int nbits) { 934 assert(0 < nbits && nbits < 32, "out of bounds"); 935 const int min = -(((int)1) << nbits-1); 936 const int maxplus1 = (((int)1) << nbits-1); 937 return min <= x && x < maxplus1; 938 } 939 940 static bool is_simm(jlong x, unsigned int nbits) { 941 assert(0 < nbits && nbits < 64, "out of bounds"); 942 const jlong min = -(((jlong)1) << nbits-1); 943 const jlong maxplus1 = (((jlong)1) << nbits-1); 944 return min <= x && x < maxplus1; 945 } 946 947 // Test if x is within unsigned immediate range for nbits. 948 static bool is_uimm(int x, unsigned int nbits) { 949 assert(0 < nbits && nbits < 32, "out of bounds"); 950 const unsigned int maxplus1 = (((unsigned int)1) << nbits); 951 return (unsigned int)x < maxplus1; 952 } 953 954 static bool is_uimm(jlong x, unsigned int nbits) { 955 assert(0 < nbits && nbits < 64, "out of bounds"); 956 const julong maxplus1 = (((julong)1) << nbits); 957 return (julong)x < maxplus1; 958 } 959 960 protected: 961 // helpers 962 963 // X is supposed to fit in a field "nbits" wide 964 // and be sign-extended. Check the range. 965 static void assert_signed_range(intptr_t x, int nbits) { 966 assert(nbits == 32 || (-(1 << nbits-1) <= x && x < (1 << nbits-1)), 967 "value out of range"); 968 } 969 970 static void assert_signed_word_disp_range(intptr_t x, int nbits) { 971 assert((x & 3) == 0, "not word aligned"); 972 assert_signed_range(x, nbits + 2); 973 } 974 975 static void assert_unsigned_const(int x, int nbits) { 976 assert(juint(x) < juint(1 << nbits), "unsigned constant out of range"); 977 } 978 979 static int fmask(juint hi_bit, juint lo_bit) { 980 assert(hi_bit >= lo_bit && hi_bit < 32, "bad bits"); 981 return (1 << ( hi_bit-lo_bit + 1 )) - 1; 982 } 983 984 // inverse of u_field 985 static int inv_u_field(int x, int hi_bit, int lo_bit) { 986 juint r = juint(x) >> lo_bit; 987 r &= fmask(hi_bit, lo_bit); 988 return int(r); 989 } 990 991 // signed version: extract from field and sign-extend 992 static int inv_s_field_ppc(int x, int hi_bit, int lo_bit) { 993 x = x << (31-hi_bit); 994 x = x >> (31-hi_bit+lo_bit); 995 return x; 996 } 997 998 static int u_field(int x, int hi_bit, int lo_bit) { 999 assert((x & ~fmask(hi_bit, lo_bit)) == 0, "value out of range"); 1000 int r = x << lo_bit; 1001 assert(inv_u_field(r, hi_bit, lo_bit) == x, "just checking"); 1002 return r; 1003 } 1004 1005 // Same as u_field for signed values 1006 static int s_field(int x, int hi_bit, int lo_bit) { 1007 int nbits = hi_bit - lo_bit + 1; 1008 assert(nbits == 32 || (-(1 << nbits-1) <= x && x < (1 << nbits-1)), 1009 "value out of range"); 1010 x &= fmask(hi_bit, lo_bit); 1011 int r = x << lo_bit; 1012 return r; 1013 } 1014 1015 // inv_op for ppc instructions 1016 static int inv_op_ppc(int x) { return inv_u_field(x, 31, 26); } 1017 1018 // Determine target address from li, bd field of branch instruction. 1019 static intptr_t inv_li_field(int x) { 1020 intptr_t r = inv_s_field_ppc(x, 25, 2); 1021 r = (r << 2); 1022 return r; 1023 } 1024 static intptr_t inv_bd_field(int x, intptr_t pos) { 1025 intptr_t r = inv_s_field_ppc(x, 15, 2); 1026 r = (r << 2) + pos; 1027 return r; 1028 } 1029 1030 #define inv_opp_u_field(x, hi_bit, lo_bit) inv_u_field(x, 31-(lo_bit), 31-(hi_bit)) 1031 #define inv_opp_s_field(x, hi_bit, lo_bit) inv_s_field_ppc(x, 31-(lo_bit), 31-(hi_bit)) 1032 // Extract instruction fields from instruction words. 1033 public: 1034 static int inv_ra_field(int x) { return inv_opp_u_field(x, 15, 11); } 1035 static int inv_rb_field(int x) { return inv_opp_u_field(x, 20, 16); } 1036 static int inv_rt_field(int x) { return inv_opp_u_field(x, 10, 6); } 1037 static int inv_rta_field(int x) { return inv_opp_u_field(x, 15, 11); } 1038 static int inv_rs_field(int x) { return inv_opp_u_field(x, 10, 6); } 1039 // Ds uses opp_s_field(x, 31, 16), but lowest 2 bits must be 0. 1040 // Inv_ds_field uses range (x, 29, 16) but shifts by 2 to ensure that lowest bits are 0. 1041 static int inv_ds_field(int x) { return inv_opp_s_field(x, 29, 16) << 2; } 1042 static int inv_d1_field(int x) { return inv_opp_s_field(x, 31, 16); } 1043 static int inv_si_field(int x) { return inv_opp_s_field(x, 31, 16); } 1044 static int inv_to_field(int x) { return inv_opp_u_field(x, 10, 6); } 1045 static int inv_lk_field(int x) { return inv_opp_u_field(x, 31, 31); } 1046 static int inv_bo_field(int x) { return inv_opp_u_field(x, 10, 6); } 1047 static int inv_bi_field(int x) { return inv_opp_u_field(x, 15, 11); } 1048 1049 #define opp_u_field(x, hi_bit, lo_bit) u_field(x, 31-(lo_bit), 31-(hi_bit)) 1050 #define opp_s_field(x, hi_bit, lo_bit) s_field(x, 31-(lo_bit), 31-(hi_bit)) 1051 1052 // instruction fields 1053 static int aa( int x) { return opp_u_field(x, 30, 30); } 1054 static int ba( int x) { return opp_u_field(x, 15, 11); } 1055 static int bb( int x) { return opp_u_field(x, 20, 16); } 1056 static int bc( int x) { return opp_u_field(x, 25, 21); } 1057 static int bd( int x) { return opp_s_field(x, 29, 16); } 1058 static int bf( ConditionRegister cr) { return bf(cr->encoding()); } 1059 static int bf( int x) { return opp_u_field(x, 8, 6); } 1060 static int bfa(ConditionRegister cr) { return bfa(cr->encoding()); } 1061 static int bfa( int x) { return opp_u_field(x, 13, 11); } 1062 static int bh( int x) { return opp_u_field(x, 20, 19); } 1063 static int bi( int x) { return opp_u_field(x, 15, 11); } 1064 static int bi0(ConditionRegister cr, Condition c) { return (cr->encoding() << 2) | c; } 1065 static int bo( int x) { return opp_u_field(x, 10, 6); } 1066 static int bt( int x) { return opp_u_field(x, 10, 6); } 1067 static int d1( int x) { return opp_s_field(x, 31, 16); } 1068 static int ds( int x) { assert((x & 0x3) == 0, "unaligned offset"); return opp_s_field(x, 31, 16); } 1069 static int eh( int x) { return opp_u_field(x, 31, 31); } 1070 static int flm( int x) { return opp_u_field(x, 14, 7); } 1071 static int fra( FloatRegister r) { return fra(r->encoding());} 1072 static int frb( FloatRegister r) { return frb(r->encoding());} 1073 static int frc( FloatRegister r) { return frc(r->encoding());} 1074 static int frs( FloatRegister r) { return frs(r->encoding());} 1075 static int frt( FloatRegister r) { return frt(r->encoding());} 1076 static int fra( int x) { return opp_u_field(x, 15, 11); } 1077 static int frb( int x) { return opp_u_field(x, 20, 16); } 1078 static int frc( int x) { return opp_u_field(x, 25, 21); } 1079 static int frs( int x) { return opp_u_field(x, 10, 6); } 1080 static int frt( int x) { return opp_u_field(x, 10, 6); } 1081 static int fxm( int x) { return opp_u_field(x, 19, 12); } 1082 static int l10( int x) { assert(x == 0 || x == 1, "must be 0 or 1"); return opp_u_field(x, 10, 10); } 1083 static int l14( int x) { return opp_u_field(x, 15, 14); } 1084 static int l15( int x) { return opp_u_field(x, 15, 15); } 1085 static int l910( int x) { return opp_u_field(x, 10, 9); } 1086 static int e1215( int x) { return opp_u_field(x, 15, 12); } 1087 static int lev( int x) { return opp_u_field(x, 26, 20); } 1088 static int li( int x) { return opp_s_field(x, 29, 6); } 1089 static int lk( int x) { return opp_u_field(x, 31, 31); } 1090 static int mb2125( int x) { return opp_u_field(x, 25, 21); } 1091 static int me2630( int x) { return opp_u_field(x, 30, 26); } 1092 static int mb2126( int x) { return opp_u_field(((x & 0x1f) << 1) | ((x & 0x20) >> 5), 26, 21); } 1093 static int me2126( int x) { return mb2126(x); } 1094 static int nb( int x) { return opp_u_field(x, 20, 16); } 1095 //static int opcd( int x) { return opp_u_field(x, 5, 0); } // is contained in our opcodes 1096 static int oe( int x) { return opp_u_field(x, 21, 21); } 1097 static int ra( Register r) { return ra(r->encoding()); } 1098 static int ra( int x) { return opp_u_field(x, 15, 11); } 1099 static int rb( Register r) { return rb(r->encoding()); } 1100 static int rb( int x) { return opp_u_field(x, 20, 16); } 1101 static int rc( int x) { return opp_u_field(x, 31, 31); } 1102 static int rs( Register r) { return rs(r->encoding()); } 1103 static int rs( int x) { return opp_u_field(x, 10, 6); } 1104 // we don't want to use R0 in memory accesses, because it has value `0' then 1105 static int ra0mem( Register r) { assert(r != R0, "cannot use register R0 in memory access"); return ra(r); } 1106 static int ra0mem( int x) { assert(x != 0, "cannot use register 0 in memory access"); return ra(x); } 1107 1108 // register r is target 1109 static int rt( Register r) { return rs(r); } 1110 static int rt( int x) { return rs(x); } 1111 static int rta( Register r) { return ra(r); } 1112 static int rta0mem( Register r) { rta(r); return ra0mem(r); } 1113 1114 static int sh1620( int x) { return opp_u_field(x, 20, 16); } 1115 static int sh30( int x) { return opp_u_field(x, 30, 30); } 1116 static int sh162030( int x) { return sh1620(x & 0x1f) | sh30((x & 0x20) >> 5); } 1117 static int si( int x) { return opp_s_field(x, 31, 16); } 1118 static int spr( int x) { return opp_u_field(x, 20, 11); } 1119 static int sr( int x) { return opp_u_field(x, 15, 12); } 1120 static int tbr( int x) { return opp_u_field(x, 20, 11); } 1121 static int th( int x) { return opp_u_field(x, 10, 7); } 1122 static int thct( int x) { assert((x&8) == 0, "must be valid cache specification"); return th(x); } 1123 static int thds( int x) { assert((x&8) == 8, "must be valid stream specification"); return th(x); } 1124 static int to( int x) { return opp_u_field(x, 10, 6); } 1125 static int u( int x) { return opp_u_field(x, 19, 16); } 1126 static int ui( int x) { return opp_u_field(x, 31, 16); } 1127 1128 // Support vector instructions for >= Power6. 1129 static int vra( int x) { return opp_u_field(x, 15, 11); } 1130 static int vrb( int x) { return opp_u_field(x, 20, 16); } 1131 static int vrc( int x) { return opp_u_field(x, 25, 21); } 1132 static int vrs( int x) { return opp_u_field(x, 10, 6); } 1133 static int vrt( int x) { return opp_u_field(x, 10, 6); } 1134 1135 static int vra( VectorRegister r) { return vra(r->encoding());} 1136 static int vrb( VectorRegister r) { return vrb(r->encoding());} 1137 static int vrc( VectorRegister r) { return vrc(r->encoding());} 1138 static int vrs( VectorRegister r) { return vrs(r->encoding());} 1139 static int vrt( VectorRegister r) { return vrt(r->encoding());} 1140 1141 // Only used on SHA sigma instructions (VX-form) 1142 static int vst( int x) { return opp_u_field(x, 16, 16); } 1143 static int vsix( int x) { return opp_u_field(x, 20, 17); } 1144 1145 // Support Vector-Scalar (VSX) instructions. 1146 static int vsra( int x) { return opp_u_field(x & 0x1F, 15, 11) | opp_u_field((x & 0x20) >> 5, 29, 29); } 1147 static int vsrb( int x) { return opp_u_field(x & 0x1F, 20, 16) | opp_u_field((x & 0x20) >> 5, 30, 30); } 1148 static int vsrs( int x) { return opp_u_field(x & 0x1F, 10, 6) | opp_u_field((x & 0x20) >> 5, 31, 31); } 1149 static int vsrt( int x) { return vsrs(x); } 1150 static int vsdm( int x) { return opp_u_field(x, 23, 22); } 1151 1152 static int vsra( VectorSRegister r) { return vsra(r->encoding());} 1153 static int vsrb( VectorSRegister r) { return vsrb(r->encoding());} 1154 static int vsrs( VectorSRegister r) { return vsrs(r->encoding());} 1155 static int vsrt( VectorSRegister r) { return vsrt(r->encoding());} 1156 1157 static int vsplt_uim( int x) { return opp_u_field(x, 15, 12); } // for vsplt* instructions 1158 static int vsplti_sim(int x) { return opp_u_field(x, 15, 11); } // for vsplti* instructions 1159 static int vsldoi_shb(int x) { return opp_u_field(x, 25, 22); } // for vsldoi instruction 1160 static int vcmp_rc( int x) { return opp_u_field(x, 21, 21); } // for vcmp* instructions 1161 static int xxsplt_uim(int x) { return opp_u_field(x, 15, 14); } // for xxsplt* instructions 1162 1163 //static int xo1( int x) { return opp_u_field(x, 29, 21); }// is contained in our opcodes 1164 //static int xo2( int x) { return opp_u_field(x, 30, 21); }// is contained in our opcodes 1165 //static int xo3( int x) { return opp_u_field(x, 30, 22); }// is contained in our opcodes 1166 //static int xo4( int x) { return opp_u_field(x, 30, 26); }// is contained in our opcodes 1167 //static int xo5( int x) { return opp_u_field(x, 29, 27); }// is contained in our opcodes 1168 //static int xo6( int x) { return opp_u_field(x, 30, 27); }// is contained in our opcodes 1169 //static int xo7( int x) { return opp_u_field(x, 31, 30); }// is contained in our opcodes 1170 1171 protected: 1172 // Compute relative address for branch. 1173 static intptr_t disp(intptr_t x, intptr_t off) { 1174 int xx = x - off; 1175 xx = xx >> 2; 1176 return xx; 1177 } 1178 1179 public: 1180 // signed immediate, in low bits, nbits long 1181 static int simm(int x, int nbits) { 1182 assert_signed_range(x, nbits); 1183 return x & ((1 << nbits) - 1); 1184 } 1185 1186 // unsigned immediate, in low bits, nbits long 1187 static int uimm(int x, int nbits) { 1188 assert_unsigned_const(x, nbits); 1189 return x & ((1 << nbits) - 1); 1190 } 1191 1192 static void set_imm(int* instr, short s) { 1193 // imm is always in the lower 16 bits of the instruction, 1194 // so this is endian-neutral. Same for the get_imm below. 1195 uint32_t w = *(uint32_t *)instr; 1196 *instr = (int)((w & ~0x0000FFFF) | (s & 0x0000FFFF)); 1197 } 1198 1199 static int get_imm(address a, int instruction_number) { 1200 return (short)((int *)a)[instruction_number]; 1201 } 1202 1203 static inline int hi16_signed( int x) { return (int)(int16_t)(x >> 16); } 1204 static inline int lo16_unsigned(int x) { return x & 0xffff; } 1205 1206 protected: 1207 1208 // Extract the top 32 bits in a 64 bit word. 1209 static int32_t hi32(int64_t x) { 1210 int32_t r = int32_t((uint64_t)x >> 32); 1211 return r; 1212 } 1213 1214 public: 1215 1216 static inline unsigned int align_addr(unsigned int addr, unsigned int a) { 1217 return ((addr + (a - 1)) & ~(a - 1)); 1218 } 1219 1220 static inline bool is_aligned(unsigned int addr, unsigned int a) { 1221 return (0 == addr % a); 1222 } 1223 1224 void flush() { 1225 AbstractAssembler::flush(); 1226 } 1227 1228 inline void emit_int32(int); // shadows AbstractAssembler::emit_int32 1229 inline void emit_data(int); 1230 inline void emit_data(int, RelocationHolder const&); 1231 inline void emit_data(int, relocInfo::relocType rtype); 1232 1233 // Emit an address. 1234 inline address emit_addr(const address addr = NULL); 1235 1236 #if !defined(ABI_ELFv2) 1237 // Emit a function descriptor with the specified entry point, TOC, 1238 // and ENV. If the entry point is NULL, the descriptor will point 1239 // just past the descriptor. 1240 // Use values from friend functions as defaults. 1241 inline address emit_fd(address entry = NULL, 1242 address toc = (address) FunctionDescriptor::friend_toc, 1243 address env = (address) FunctionDescriptor::friend_env); 1244 #endif 1245 1246 ///////////////////////////////////////////////////////////////////////////////////// 1247 // PPC instructions 1248 ///////////////////////////////////////////////////////////////////////////////////// 1249 1250 // Memory instructions use r0 as hard coded 0, e.g. to simulate loading 1251 // immediates. The normal instruction encoders enforce that r0 is not 1252 // passed to them. Use either extended mnemonics encoders or the special ra0 1253 // versions. 1254 1255 // Issue an illegal instruction. 1256 inline void illtrap(); 1257 static inline bool is_illtrap(int x); 1258 1259 // PPC 1, section 3.3.8, Fixed-Point Arithmetic Instructions 1260 inline void addi( Register d, Register a, int si16); 1261 inline void addis(Register d, Register a, int si16); 1262 private: 1263 inline void addi_r0ok( Register d, Register a, int si16); 1264 inline void addis_r0ok(Register d, Register a, int si16); 1265 public: 1266 inline void addic_( Register d, Register a, int si16); 1267 inline void subfic( Register d, Register a, int si16); 1268 inline void add( Register d, Register a, Register b); 1269 inline void add_( Register d, Register a, Register b); 1270 inline void subf( Register d, Register a, Register b); // d = b - a "Sub_from", as in ppc spec. 1271 inline void sub( Register d, Register a, Register b); // d = a - b Swap operands of subf for readability. 1272 inline void subf_( Register d, Register a, Register b); 1273 inline void addc( Register d, Register a, Register b); 1274 inline void addc_( Register d, Register a, Register b); 1275 inline void subfc( Register d, Register a, Register b); 1276 inline void subfc_( Register d, Register a, Register b); 1277 inline void adde( Register d, Register a, Register b); 1278 inline void adde_( Register d, Register a, Register b); 1279 inline void subfe( Register d, Register a, Register b); 1280 inline void subfe_( Register d, Register a, Register b); 1281 inline void addme( Register d, Register a); 1282 inline void addme_( Register d, Register a); 1283 inline void subfme( Register d, Register a); 1284 inline void subfme_(Register d, Register a); 1285 inline void addze( Register d, Register a); 1286 inline void addze_( Register d, Register a); 1287 inline void subfze( Register d, Register a); 1288 inline void subfze_(Register d, Register a); 1289 inline void neg( Register d, Register a); 1290 inline void neg_( Register d, Register a); 1291 inline void mulli( Register d, Register a, int si16); 1292 inline void mulld( Register d, Register a, Register b); 1293 inline void mulld_( Register d, Register a, Register b); 1294 inline void mullw( Register d, Register a, Register b); 1295 inline void mullw_( Register d, Register a, Register b); 1296 inline void mulhw( Register d, Register a, Register b); 1297 inline void mulhw_( Register d, Register a, Register b); 1298 inline void mulhwu( Register d, Register a, Register b); 1299 inline void mulhwu_(Register d, Register a, Register b); 1300 inline void mulhd( Register d, Register a, Register b); 1301 inline void mulhd_( Register d, Register a, Register b); 1302 inline void mulhdu( Register d, Register a, Register b); 1303 inline void mulhdu_(Register d, Register a, Register b); 1304 inline void divd( Register d, Register a, Register b); 1305 inline void divd_( Register d, Register a, Register b); 1306 inline void divw( Register d, Register a, Register b); 1307 inline void divw_( Register d, Register a, Register b); 1308 1309 // Fixed-Point Arithmetic Instructions with Overflow detection 1310 inline void addo( Register d, Register a, Register b); 1311 inline void addo_( Register d, Register a, Register b); 1312 inline void subfo( Register d, Register a, Register b); 1313 inline void subfo_( Register d, Register a, Register b); 1314 inline void addco( Register d, Register a, Register b); 1315 inline void addco_( Register d, Register a, Register b); 1316 inline void subfco( Register d, Register a, Register b); 1317 inline void subfco_( Register d, Register a, Register b); 1318 inline void addeo( Register d, Register a, Register b); 1319 inline void addeo_( Register d, Register a, Register b); 1320 inline void subfeo( Register d, Register a, Register b); 1321 inline void subfeo_( Register d, Register a, Register b); 1322 inline void addmeo( Register d, Register a); 1323 inline void addmeo_( Register d, Register a); 1324 inline void subfmeo( Register d, Register a); 1325 inline void subfmeo_(Register d, Register a); 1326 inline void addzeo( Register d, Register a); 1327 inline void addzeo_( Register d, Register a); 1328 inline void subfzeo( Register d, Register a); 1329 inline void subfzeo_(Register d, Register a); 1330 inline void nego( Register d, Register a); 1331 inline void nego_( Register d, Register a); 1332 inline void mulldo( Register d, Register a, Register b); 1333 inline void mulldo_( Register d, Register a, Register b); 1334 inline void mullwo( Register d, Register a, Register b); 1335 inline void mullwo_( Register d, Register a, Register b); 1336 inline void divdo( Register d, Register a, Register b); 1337 inline void divdo_( Register d, Register a, Register b); 1338 inline void divwo( Register d, Register a, Register b); 1339 inline void divwo_( Register d, Register a, Register b); 1340 1341 // extended mnemonics 1342 inline void li( Register d, int si16); 1343 inline void lis( Register d, int si16); 1344 inline void addir(Register d, int si16, Register a); 1345 inline void subi( Register d, Register a, int si16); 1346 1347 static bool is_addi(int x) { 1348 return ADDI_OPCODE == (x & ADDI_OPCODE_MASK); 1349 } 1350 static bool is_addis(int x) { 1351 return ADDIS_OPCODE == (x & ADDIS_OPCODE_MASK); 1352 } 1353 static bool is_bxx(int x) { 1354 return BXX_OPCODE == (x & BXX_OPCODE_MASK); 1355 } 1356 static bool is_b(int x) { 1357 return BXX_OPCODE == (x & BXX_OPCODE_MASK) && inv_lk_field(x) == 0; 1358 } 1359 static bool is_bl(int x) { 1360 return BXX_OPCODE == (x & BXX_OPCODE_MASK) && inv_lk_field(x) == 1; 1361 } 1362 static bool is_bcxx(int x) { 1363 return BCXX_OPCODE == (x & BCXX_OPCODE_MASK); 1364 } 1365 static bool is_bxx_or_bcxx(int x) { 1366 return is_bxx(x) || is_bcxx(x); 1367 } 1368 static bool is_bctrl(int x) { 1369 return x == 0x4e800421; 1370 } 1371 static bool is_bctr(int x) { 1372 return x == 0x4e800420; 1373 } 1374 static bool is_bclr(int x) { 1375 return BCLR_OPCODE == (x & XL_FORM_OPCODE_MASK); 1376 } 1377 static bool is_li(int x) { 1378 return is_addi(x) && inv_ra_field(x)==0; 1379 } 1380 static bool is_lis(int x) { 1381 return is_addis(x) && inv_ra_field(x)==0; 1382 } 1383 static bool is_mtctr(int x) { 1384 return MTCTR_OPCODE == (x & MTCTR_OPCODE_MASK); 1385 } 1386 static bool is_ld(int x) { 1387 return LD_OPCODE == (x & LD_OPCODE_MASK); 1388 } 1389 static bool is_std(int x) { 1390 return STD_OPCODE == (x & STD_OPCODE_MASK); 1391 } 1392 static bool is_stdu(int x) { 1393 return STDU_OPCODE == (x & STDU_OPCODE_MASK); 1394 } 1395 static bool is_stdx(int x) { 1396 return STDX_OPCODE == (x & STDX_OPCODE_MASK); 1397 } 1398 static bool is_stdux(int x) { 1399 return STDUX_OPCODE == (x & STDUX_OPCODE_MASK); 1400 } 1401 static bool is_stwx(int x) { 1402 return STWX_OPCODE == (x & STWX_OPCODE_MASK); 1403 } 1404 static bool is_stwux(int x) { 1405 return STWUX_OPCODE == (x & STWUX_OPCODE_MASK); 1406 } 1407 static bool is_stw(int x) { 1408 return STW_OPCODE == (x & STW_OPCODE_MASK); 1409 } 1410 static bool is_stwu(int x) { 1411 return STWU_OPCODE == (x & STWU_OPCODE_MASK); 1412 } 1413 static bool is_ori(int x) { 1414 return ORI_OPCODE == (x & ORI_OPCODE_MASK); 1415 }; 1416 static bool is_oris(int x) { 1417 return ORIS_OPCODE == (x & ORIS_OPCODE_MASK); 1418 }; 1419 static bool is_rldicr(int x) { 1420 return (RLDICR_OPCODE == (x & RLDICR_OPCODE_MASK)); 1421 }; 1422 static bool is_nop(int x) { 1423 return x == 0x60000000; 1424 } 1425 // endgroup opcode for Power6 1426 static bool is_endgroup(int x) { 1427 return is_ori(x) && inv_ra_field(x) == 1 && inv_rs_field(x) == 1 && inv_d1_field(x) == 0; 1428 } 1429 1430 1431 private: 1432 // PPC 1, section 3.3.9, Fixed-Point Compare Instructions 1433 inline void cmpi( ConditionRegister bf, int l, Register a, int si16); 1434 inline void cmp( ConditionRegister bf, int l, Register a, Register b); 1435 inline void cmpli(ConditionRegister bf, int l, Register a, int ui16); 1436 inline void cmpl( ConditionRegister bf, int l, Register a, Register b); 1437 1438 public: 1439 // extended mnemonics of Compare Instructions 1440 inline void cmpwi( ConditionRegister crx, Register a, int si16); 1441 inline void cmpdi( ConditionRegister crx, Register a, int si16); 1442 inline void cmpw( ConditionRegister crx, Register a, Register b); 1443 inline void cmpd( ConditionRegister crx, Register a, Register b); 1444 inline void cmplwi(ConditionRegister crx, Register a, int ui16); 1445 inline void cmpldi(ConditionRegister crx, Register a, int ui16); 1446 inline void cmplw( ConditionRegister crx, Register a, Register b); 1447 inline void cmpld( ConditionRegister crx, Register a, Register b); 1448 1449 // >= Power9 1450 inline void cmprb( ConditionRegister bf, int l, Register a, Register b); 1451 inline void cmpeqb(ConditionRegister bf, Register a, Register b); 1452 1453 inline void isel( Register d, Register a, Register b, int bc); 1454 // Convenient version which takes: Condition register, Condition code and invert flag. Omit b to keep old value. 1455 inline void isel( Register d, ConditionRegister cr, Condition cc, bool inv, Register a, Register b = noreg); 1456 // Set d = 0 if (cr.cc) equals 1, otherwise b. 1457 inline void isel_0( Register d, ConditionRegister cr, Condition cc, Register b = noreg); 1458 1459 // PPC 1, section 3.3.11, Fixed-Point Logical Instructions 1460 void andi( Register a, Register s, long ui16); // optimized version 1461 inline void andi_( Register a, Register s, int ui16); 1462 inline void andis_( Register a, Register s, int ui16); 1463 inline void ori( Register a, Register s, int ui16); 1464 inline void oris( Register a, Register s, int ui16); 1465 inline void xori( Register a, Register s, int ui16); 1466 inline void xoris( Register a, Register s, int ui16); 1467 inline void andr( Register a, Register s, Register b); // suffixed by 'r' as 'and' is C++ keyword 1468 inline void and_( Register a, Register s, Register b); 1469 // Turn or0(rx,rx,rx) into a nop and avoid that we accidently emit a 1470 // SMT-priority change instruction (see SMT instructions below). 1471 inline void or_unchecked(Register a, Register s, Register b); 1472 inline void orr( Register a, Register s, Register b); // suffixed by 'r' as 'or' is C++ keyword 1473 inline void or_( Register a, Register s, Register b); 1474 inline void xorr( Register a, Register s, Register b); // suffixed by 'r' as 'xor' is C++ keyword 1475 inline void xor_( Register a, Register s, Register b); 1476 inline void nand( Register a, Register s, Register b); 1477 inline void nand_( Register a, Register s, Register b); 1478 inline void nor( Register a, Register s, Register b); 1479 inline void nor_( Register a, Register s, Register b); 1480 inline void andc( Register a, Register s, Register b); 1481 inline void andc_( Register a, Register s, Register b); 1482 inline void orc( Register a, Register s, Register b); 1483 inline void orc_( Register a, Register s, Register b); 1484 inline void extsb( Register a, Register s); 1485 inline void extsb_( Register a, Register s); 1486 inline void extsh( Register a, Register s); 1487 inline void extsh_( Register a, Register s); 1488 inline void extsw( Register a, Register s); 1489 inline void extsw_( Register a, Register s); 1490 1491 // extended mnemonics 1492 inline void nop(); 1493 // NOP for FP and BR units (different versions to allow them to be in one group) 1494 inline void fpnop0(); 1495 inline void fpnop1(); 1496 inline void brnop0(); 1497 inline void brnop1(); 1498 inline void brnop2(); 1499 1500 inline void mr( Register d, Register s); 1501 inline void ori_opt( Register d, int ui16); 1502 inline void oris_opt(Register d, int ui16); 1503 1504 // endgroup opcode for Power6 1505 inline void endgroup(); 1506 1507 // count instructions 1508 inline void cntlzw( Register a, Register s); 1509 inline void cntlzw_( Register a, Register s); 1510 inline void cntlzd( Register a, Register s); 1511 inline void cntlzd_( Register a, Register s); 1512 inline void cnttzw( Register a, Register s); 1513 inline void cnttzw_( Register a, Register s); 1514 inline void cnttzd( Register a, Register s); 1515 inline void cnttzd_( Register a, Register s); 1516 1517 // PPC 1, section 3.3.12, Fixed-Point Rotate and Shift Instructions 1518 inline void sld( Register a, Register s, Register b); 1519 inline void sld_( Register a, Register s, Register b); 1520 inline void slw( Register a, Register s, Register b); 1521 inline void slw_( Register a, Register s, Register b); 1522 inline void srd( Register a, Register s, Register b); 1523 inline void srd_( Register a, Register s, Register b); 1524 inline void srw( Register a, Register s, Register b); 1525 inline void srw_( Register a, Register s, Register b); 1526 inline void srad( Register a, Register s, Register b); 1527 inline void srad_( Register a, Register s, Register b); 1528 inline void sraw( Register a, Register s, Register b); 1529 inline void sraw_( Register a, Register s, Register b); 1530 inline void sradi( Register a, Register s, int sh6); 1531 inline void sradi_( Register a, Register s, int sh6); 1532 inline void srawi( Register a, Register s, int sh5); 1533 inline void srawi_( Register a, Register s, int sh5); 1534 1535 // extended mnemonics for Shift Instructions 1536 inline void sldi( Register a, Register s, int sh6); 1537 inline void sldi_( Register a, Register s, int sh6); 1538 inline void slwi( Register a, Register s, int sh5); 1539 inline void slwi_( Register a, Register s, int sh5); 1540 inline void srdi( Register a, Register s, int sh6); 1541 inline void srdi_( Register a, Register s, int sh6); 1542 inline void srwi( Register a, Register s, int sh5); 1543 inline void srwi_( Register a, Register s, int sh5); 1544 1545 inline void clrrdi( Register a, Register s, int ui6); 1546 inline void clrrdi_( Register a, Register s, int ui6); 1547 inline void clrldi( Register a, Register s, int ui6); 1548 inline void clrldi_( Register a, Register s, int ui6); 1549 inline void clrlsldi(Register a, Register s, int clrl6, int shl6); 1550 inline void clrlsldi_(Register a, Register s, int clrl6, int shl6); 1551 inline void extrdi( Register a, Register s, int n, int b); 1552 // testbit with condition register 1553 inline void testbitdi(ConditionRegister cr, Register a, Register s, int ui6); 1554 1555 // rotate instructions 1556 inline void rotldi( Register a, Register s, int n); 1557 inline void rotrdi( Register a, Register s, int n); 1558 inline void rotlwi( Register a, Register s, int n); 1559 inline void rotrwi( Register a, Register s, int n); 1560 1561 // Rotate Instructions 1562 inline void rldic( Register a, Register s, int sh6, int mb6); 1563 inline void rldic_( Register a, Register s, int sh6, int mb6); 1564 inline void rldicr( Register a, Register s, int sh6, int mb6); 1565 inline void rldicr_( Register a, Register s, int sh6, int mb6); 1566 inline void rldicl( Register a, Register s, int sh6, int mb6); 1567 inline void rldicl_( Register a, Register s, int sh6, int mb6); 1568 inline void rlwinm( Register a, Register s, int sh5, int mb5, int me5); 1569 inline void rlwinm_( Register a, Register s, int sh5, int mb5, int me5); 1570 inline void rldimi( Register a, Register s, int sh6, int mb6); 1571 inline void rldimi_( Register a, Register s, int sh6, int mb6); 1572 inline void rlwimi( Register a, Register s, int sh5, int mb5, int me5); 1573 inline void insrdi( Register a, Register s, int n, int b); 1574 inline void insrwi( Register a, Register s, int n, int b); 1575 1576 // PPC 1, section 3.3.2 Fixed-Point Load Instructions 1577 // 4 bytes 1578 inline void lwzx( Register d, Register s1, Register s2); 1579 inline void lwz( Register d, int si16, Register s1); 1580 inline void lwzu( Register d, int si16, Register s1); 1581 1582 // 4 bytes 1583 inline void lwax( Register d, Register s1, Register s2); 1584 inline void lwa( Register d, int si16, Register s1); 1585 1586 // 4 bytes reversed 1587 inline void lwbrx( Register d, Register s1, Register s2); 1588 1589 // 2 bytes 1590 inline void lhzx( Register d, Register s1, Register s2); 1591 inline void lhz( Register d, int si16, Register s1); 1592 inline void lhzu( Register d, int si16, Register s1); 1593 1594 // 2 bytes reversed 1595 inline void lhbrx( Register d, Register s1, Register s2); 1596 1597 // 2 bytes 1598 inline void lhax( Register d, Register s1, Register s2); 1599 inline void lha( Register d, int si16, Register s1); 1600 inline void lhau( Register d, int si16, Register s1); 1601 1602 // 1 byte 1603 inline void lbzx( Register d, Register s1, Register s2); 1604 inline void lbz( Register d, int si16, Register s1); 1605 inline void lbzu( Register d, int si16, Register s1); 1606 1607 // 8 bytes 1608 inline void ldx( Register d, Register s1, Register s2); 1609 inline void ld( Register d, int si16, Register s1); 1610 inline void ldu( Register d, int si16, Register s1); 1611 1612 // 8 bytes reversed 1613 inline void ldbrx( Register d, Register s1, Register s2); 1614 1615 // For convenience. Load pointer into d from b+s1. 1616 inline void ld_ptr(Register d, int b, Register s1); 1617 DEBUG_ONLY(inline void ld_ptr(Register d, ByteSize b, Register s1);) 1618 1619 // PPC 1, section 3.3.3 Fixed-Point Store Instructions 1620 inline void stwx( Register d, Register s1, Register s2); 1621 inline void stw( Register d, int si16, Register s1); 1622 inline void stwu( Register d, int si16, Register s1); 1623 inline void stwbrx( Register d, Register s1, Register s2); 1624 1625 inline void sthx( Register d, Register s1, Register s2); 1626 inline void sth( Register d, int si16, Register s1); 1627 inline void sthu( Register d, int si16, Register s1); 1628 inline void sthbrx( Register d, Register s1, Register s2); 1629 1630 inline void stbx( Register d, Register s1, Register s2); 1631 inline void stb( Register d, int si16, Register s1); 1632 inline void stbu( Register d, int si16, Register s1); 1633 1634 inline void stdx( Register d, Register s1, Register s2); 1635 inline void std( Register d, int si16, Register s1); 1636 inline void stdu( Register d, int si16, Register s1); 1637 inline void stdux(Register s, Register a, Register b); 1638 inline void stdbrx( Register d, Register s1, Register s2); 1639 1640 inline void st_ptr(Register d, int si16, Register s1); 1641 DEBUG_ONLY(inline void st_ptr(Register d, ByteSize b, Register s1);) 1642 1643 // PPC 1, section 3.3.13 Move To/From System Register Instructions 1644 inline void mtlr( Register s1); 1645 inline void mflr( Register d); 1646 inline void mtctr(Register s1); 1647 inline void mfctr(Register d); 1648 inline void mtcrf(int fxm, Register s); 1649 inline void mfcr( Register d); 1650 inline void mcrf( ConditionRegister crd, ConditionRegister cra); 1651 inline void mtcr( Register s); 1652 // >= Power9 1653 inline void setb( Register d, ConditionRegister cra); 1654 1655 // Special purpose registers 1656 // Exception Register 1657 inline void mtxer(Register s1); 1658 inline void mfxer(Register d); 1659 // Vector Register Save Register 1660 inline void mtvrsave(Register s1); 1661 inline void mfvrsave(Register d); 1662 // Timebase 1663 inline void mftb(Register d); 1664 // Introduced with Power 8: 1665 // Data Stream Control Register 1666 inline void mtdscr(Register s1); 1667 inline void mfdscr(Register d ); 1668 // Transactional Memory Registers 1669 inline void mftfhar(Register d); 1670 inline void mftfiar(Register d); 1671 inline void mftexasr(Register d); 1672 inline void mftexasru(Register d); 1673 1674 // TEXASR bit description 1675 enum transaction_failure_reason { 1676 // Upper half (TEXASRU): 1677 tm_failure_code = 0, // The Failure Code is copied from tabort or treclaim operand. 1678 tm_failure_persistent = 7, // The failure is likely to recur on each execution. 1679 tm_disallowed = 8, // The instruction is not permitted. 1680 tm_nesting_of = 9, // The maximum transaction level was exceeded. 1681 tm_footprint_of = 10, // The tracking limit for transactional storage accesses was exceeded. 1682 tm_self_induced_cf = 11, // A self-induced conflict occurred in Suspended state. 1683 tm_non_trans_cf = 12, // A conflict occurred with a non-transactional access by another processor. 1684 tm_trans_cf = 13, // A conflict occurred with another transaction. 1685 tm_translation_cf = 14, // A conflict occurred with a TLB invalidation. 1686 tm_inst_fetch_cf = 16, // An instruction fetch was performed from a block that was previously written transactionally. 1687 tm_tabort = 31, // Termination was caused by the execution of an abort instruction. 1688 // Lower half: 1689 tm_suspended = 32, // Failure was recorded in Suspended state. 1690 tm_failure_summary = 36, // Failure has been detected and recorded. 1691 tm_tfiar_exact = 37, // Value in the TFIAR is exact. 1692 tm_rot = 38, // Rollback-only transaction. 1693 tm_transaction_level = 52, // Transaction level (nesting depth + 1). 1694 }; 1695 1696 // PPC 1, section 2.4.1 Branch Instructions 1697 inline void b( address a, relocInfo::relocType rt = relocInfo::none); 1698 inline void b( Label& L); 1699 inline void bl( address a, relocInfo::relocType rt = relocInfo::none); 1700 inline void bl( Label& L); 1701 inline void bc( int boint, int biint, address a, relocInfo::relocType rt = relocInfo::none); 1702 inline void bc( int boint, int biint, Label& L); 1703 inline void bcl(int boint, int biint, address a, relocInfo::relocType rt = relocInfo::none); 1704 inline void bcl(int boint, int biint, Label& L); 1705 1706 inline void bclr( int boint, int biint, int bhint, relocInfo::relocType rt = relocInfo::none); 1707 inline void bclrl( int boint, int biint, int bhint, relocInfo::relocType rt = relocInfo::none); 1708 inline void bcctr( int boint, int biint, int bhint = bhintbhBCCTRisNotReturnButSame, 1709 relocInfo::relocType rt = relocInfo::none); 1710 inline void bcctrl(int boint, int biint, int bhint = bhintbhBCLRisReturn, 1711 relocInfo::relocType rt = relocInfo::none); 1712 1713 // helper function for b, bcxx 1714 inline bool is_within_range_of_b(address a, address pc); 1715 inline bool is_within_range_of_bcxx(address a, address pc); 1716 1717 // get the destination of a bxx branch (b, bl, ba, bla) 1718 static inline address bxx_destination(address baddr); 1719 static inline address bxx_destination(int instr, address pc); 1720 static inline intptr_t bxx_destination_offset(int instr, intptr_t bxx_pos); 1721 1722 // extended mnemonics for branch instructions 1723 inline void blt(ConditionRegister crx, Label& L); 1724 inline void bgt(ConditionRegister crx, Label& L); 1725 inline void beq(ConditionRegister crx, Label& L); 1726 inline void bso(ConditionRegister crx, Label& L); 1727 inline void bge(ConditionRegister crx, Label& L); 1728 inline void ble(ConditionRegister crx, Label& L); 1729 inline void bne(ConditionRegister crx, Label& L); 1730 inline void bns(ConditionRegister crx, Label& L); 1731 1732 // Branch instructions with static prediction hints. 1733 inline void blt_predict_taken( ConditionRegister crx, Label& L); 1734 inline void bgt_predict_taken( ConditionRegister crx, Label& L); 1735 inline void beq_predict_taken( ConditionRegister crx, Label& L); 1736 inline void bso_predict_taken( ConditionRegister crx, Label& L); 1737 inline void bge_predict_taken( ConditionRegister crx, Label& L); 1738 inline void ble_predict_taken( ConditionRegister crx, Label& L); 1739 inline void bne_predict_taken( ConditionRegister crx, Label& L); 1740 inline void bns_predict_taken( ConditionRegister crx, Label& L); 1741 inline void blt_predict_not_taken(ConditionRegister crx, Label& L); 1742 inline void bgt_predict_not_taken(ConditionRegister crx, Label& L); 1743 inline void beq_predict_not_taken(ConditionRegister crx, Label& L); 1744 inline void bso_predict_not_taken(ConditionRegister crx, Label& L); 1745 inline void bge_predict_not_taken(ConditionRegister crx, Label& L); 1746 inline void ble_predict_not_taken(ConditionRegister crx, Label& L); 1747 inline void bne_predict_not_taken(ConditionRegister crx, Label& L); 1748 inline void bns_predict_not_taken(ConditionRegister crx, Label& L); 1749 1750 // for use in conjunction with testbitdi: 1751 inline void btrue( ConditionRegister crx, Label& L); 1752 inline void bfalse(ConditionRegister crx, Label& L); 1753 1754 inline void bltl(ConditionRegister crx, Label& L); 1755 inline void bgtl(ConditionRegister crx, Label& L); 1756 inline void beql(ConditionRegister crx, Label& L); 1757 inline void bsol(ConditionRegister crx, Label& L); 1758 inline void bgel(ConditionRegister crx, Label& L); 1759 inline void blel(ConditionRegister crx, Label& L); 1760 inline void bnel(ConditionRegister crx, Label& L); 1761 inline void bnsl(ConditionRegister crx, Label& L); 1762 1763 // extended mnemonics for Branch Instructions via LR 1764 // We use `blr' for returns. 1765 inline void blr(relocInfo::relocType rt = relocInfo::none); 1766 1767 // extended mnemonics for Branch Instructions with CTR 1768 // bdnz means `decrement CTR and jump to L if CTR is not zero' 1769 inline void bdnz(Label& L); 1770 // Decrement and branch if result is zero. 1771 inline void bdz(Label& L); 1772 // we use `bctr[l]' for jumps/calls in function descriptor glue 1773 // code, e.g. calls to runtime functions 1774 inline void bctr( relocInfo::relocType rt = relocInfo::none); 1775 inline void bctrl(relocInfo::relocType rt = relocInfo::none); 1776 // conditional jumps/branches via CTR 1777 inline void beqctr( ConditionRegister crx, relocInfo::relocType rt = relocInfo::none); 1778 inline void beqctrl(ConditionRegister crx, relocInfo::relocType rt = relocInfo::none); 1779 inline void bnectr( ConditionRegister crx, relocInfo::relocType rt = relocInfo::none); 1780 inline void bnectrl(ConditionRegister crx, relocInfo::relocType rt = relocInfo::none); 1781 1782 // condition register logic instructions 1783 // NOTE: There's a preferred form: d and s2 should point into the same condition register. 1784 inline void crand( int d, int s1, int s2); 1785 inline void crnand(int d, int s1, int s2); 1786 inline void cror( int d, int s1, int s2); 1787 inline void crxor( int d, int s1, int s2); 1788 inline void crnor( int d, int s1, int s2); 1789 inline void creqv( int d, int s1, int s2); 1790 inline void crandc(int d, int s1, int s2); 1791 inline void crorc( int d, int s1, int s2); 1792 1793 // More convenient version. 1794 int condition_register_bit(ConditionRegister cr, Condition c) { 1795 return 4 * (int)(intptr_t)cr + c; 1796 } 1797 void crand( ConditionRegister crdst, Condition cdst, ConditionRegister crsrc, Condition csrc); 1798 void crnand(ConditionRegister crdst, Condition cdst, ConditionRegister crsrc, Condition csrc); 1799 void cror( ConditionRegister crdst, Condition cdst, ConditionRegister crsrc, Condition csrc); 1800 void crxor( ConditionRegister crdst, Condition cdst, ConditionRegister crsrc, Condition csrc); 1801 void crnor( ConditionRegister crdst, Condition cdst, ConditionRegister crsrc, Condition csrc); 1802 void creqv( ConditionRegister crdst, Condition cdst, ConditionRegister crsrc, Condition csrc); 1803 void crandc(ConditionRegister crdst, Condition cdst, ConditionRegister crsrc, Condition csrc); 1804 void crorc( ConditionRegister crdst, Condition cdst, ConditionRegister crsrc, Condition csrc); 1805 1806 // icache and dcache related instructions 1807 inline void icbi( Register s1, Register s2); 1808 //inline void dcba(Register s1, Register s2); // Instruction for embedded processor only. 1809 inline void dcbz( Register s1, Register s2); 1810 inline void dcbst( Register s1, Register s2); 1811 inline void dcbf( Register s1, Register s2); 1812 1813 enum ct_cache_specification { 1814 ct_primary_cache = 0, 1815 ct_secondary_cache = 2 1816 }; 1817 // dcache read hint 1818 inline void dcbt( Register s1, Register s2); 1819 inline void dcbtct( Register s1, Register s2, int ct); 1820 inline void dcbtds( Register s1, Register s2, int ds); 1821 // dcache write hint 1822 inline void dcbtst( Register s1, Register s2); 1823 inline void dcbtstct(Register s1, Register s2, int ct); 1824 1825 // machine barrier instructions: 1826 // 1827 // - sync two-way memory barrier, aka fence 1828 // - lwsync orders Store|Store, 1829 // Load|Store, 1830 // Load|Load, 1831 // but not Store|Load 1832 // - eieio orders memory accesses for device memory (only) 1833 // - isync invalidates speculatively executed instructions 1834 // From the Power ISA 2.06 documentation: 1835 // "[...] an isync instruction prevents the execution of 1836 // instructions following the isync until instructions 1837 // preceding the isync have completed, [...]" 1838 // From IBM's AIX assembler reference: 1839 // "The isync [...] instructions causes the processor to 1840 // refetch any instructions that might have been fetched 1841 // prior to the isync instruction. The instruction isync 1842 // causes the processor to wait for all previous instructions 1843 // to complete. Then any instructions already fetched are 1844 // discarded and instruction processing continues in the 1845 // environment established by the previous instructions." 1846 // 1847 // semantic barrier instructions: 1848 // (as defined in orderAccess.hpp) 1849 // 1850 // - release orders Store|Store, (maps to lwsync) 1851 // Load|Store 1852 // - acquire orders Load|Store, (maps to lwsync) 1853 // Load|Load 1854 // - fence orders Store|Store, (maps to sync) 1855 // Load|Store, 1856 // Load|Load, 1857 // Store|Load 1858 // 1859 private: 1860 inline void sync(int l); 1861 public: 1862 inline void sync(); 1863 inline void lwsync(); 1864 inline void ptesync(); 1865 inline void eieio(); 1866 inline void isync(); 1867 inline void elemental_membar(int e); // Elemental Memory Barriers (>=Power 8) 1868 1869 // Wait instructions for polling. Attention: May result in SIGILL. 1870 inline void wait(); 1871 inline void waitrsv(); // >=Power7 1872 1873 // atomics 1874 inline void lbarx_unchecked(Register d, Register a, Register b, int eh1 = 0); // >=Power 8 1875 inline void lharx_unchecked(Register d, Register a, Register b, int eh1 = 0); // >=Power 8 1876 inline void lwarx_unchecked(Register d, Register a, Register b, int eh1 = 0); 1877 inline void ldarx_unchecked(Register d, Register a, Register b, int eh1 = 0); 1878 inline void lqarx_unchecked(Register d, Register a, Register b, int eh1 = 0); // >=Power 8 1879 inline bool lxarx_hint_exclusive_access(); 1880 inline void lbarx( Register d, Register a, Register b, bool hint_exclusive_access = false); 1881 inline void lharx( Register d, Register a, Register b, bool hint_exclusive_access = false); 1882 inline void lwarx( Register d, Register a, Register b, bool hint_exclusive_access = false); 1883 inline void ldarx( Register d, Register a, Register b, bool hint_exclusive_access = false); 1884 inline void lqarx( Register d, Register a, Register b, bool hint_exclusive_access = false); 1885 inline void stbcx_( Register s, Register a, Register b); 1886 inline void sthcx_( Register s, Register a, Register b); 1887 inline void stwcx_( Register s, Register a, Register b); 1888 inline void stdcx_( Register s, Register a, Register b); 1889 inline void stqcx_( Register s, Register a, Register b); 1890 1891 // Instructions for adjusting thread priority for simultaneous 1892 // multithreading (SMT) on Power5. 1893 private: 1894 inline void smt_prio_very_low(); 1895 inline void smt_prio_medium_high(); 1896 inline void smt_prio_high(); 1897 1898 public: 1899 inline void smt_prio_low(); 1900 inline void smt_prio_medium_low(); 1901 inline void smt_prio_medium(); 1902 // >= Power7 1903 inline void smt_yield(); 1904 inline void smt_mdoio(); 1905 inline void smt_mdoom(); 1906 // >= Power8 1907 inline void smt_miso(); 1908 1909 // trap instructions 1910 inline void twi_0(Register a); // for load with acquire semantics use load+twi_0+isync (trap can't occur) 1911 // NOT FOR DIRECT USE!! 1912 protected: 1913 inline void tdi_unchecked(int tobits, Register a, int si16); 1914 inline void twi_unchecked(int tobits, Register a, int si16); 1915 inline void tdi( int tobits, Register a, int si16); // asserts UseSIGTRAP 1916 inline void twi( int tobits, Register a, int si16); // asserts UseSIGTRAP 1917 inline void td( int tobits, Register a, Register b); // asserts UseSIGTRAP 1918 inline void tw( int tobits, Register a, Register b); // asserts UseSIGTRAP 1919 1920 static bool is_tdi(int x, int tobits, int ra, int si16) { 1921 return (TDI_OPCODE == (x & TDI_OPCODE_MASK)) 1922 && (tobits == inv_to_field(x)) 1923 && (ra == -1/*any reg*/ || ra == inv_ra_field(x)) 1924 && (si16 == inv_si_field(x)); 1925 } 1926 1927 static bool is_twi(int x, int tobits, int ra, int si16) { 1928 return (TWI_OPCODE == (x & TWI_OPCODE_MASK)) 1929 && (tobits == inv_to_field(x)) 1930 && (ra == -1/*any reg*/ || ra == inv_ra_field(x)) 1931 && (si16 == inv_si_field(x)); 1932 } 1933 1934 static bool is_twi(int x, int tobits, int ra) { 1935 return (TWI_OPCODE == (x & TWI_OPCODE_MASK)) 1936 && (tobits == inv_to_field(x)) 1937 && (ra == -1/*any reg*/ || ra == inv_ra_field(x)); 1938 } 1939 1940 static bool is_td(int x, int tobits, int ra, int rb) { 1941 return (TD_OPCODE == (x & TD_OPCODE_MASK)) 1942 && (tobits == inv_to_field(x)) 1943 && (ra == -1/*any reg*/ || ra == inv_ra_field(x)) 1944 && (rb == -1/*any reg*/ || rb == inv_rb_field(x)); 1945 } 1946 1947 static bool is_tw(int x, int tobits, int ra, int rb) { 1948 return (TW_OPCODE == (x & TW_OPCODE_MASK)) 1949 && (tobits == inv_to_field(x)) 1950 && (ra == -1/*any reg*/ || ra == inv_ra_field(x)) 1951 && (rb == -1/*any reg*/ || rb == inv_rb_field(x)); 1952 } 1953 1954 public: 1955 // PPC floating point instructions 1956 // PPC 1, section 4.6.2 Floating-Point Load Instructions 1957 inline void lfs( FloatRegister d, int si16, Register a); 1958 inline void lfsu( FloatRegister d, int si16, Register a); 1959 inline void lfsx( FloatRegister d, Register a, Register b); 1960 inline void lfd( FloatRegister d, int si16, Register a); 1961 inline void lfdu( FloatRegister d, int si16, Register a); 1962 inline void lfdx( FloatRegister d, Register a, Register b); 1963 1964 // PPC 1, section 4.6.3 Floating-Point Store Instructions 1965 inline void stfs( FloatRegister s, int si16, Register a); 1966 inline void stfsu( FloatRegister s, int si16, Register a); 1967 inline void stfsx( FloatRegister s, Register a, Register b); 1968 inline void stfd( FloatRegister s, int si16, Register a); 1969 inline void stfdu( FloatRegister s, int si16, Register a); 1970 inline void stfdx( FloatRegister s, Register a, Register b); 1971 1972 // PPC 1, section 4.6.4 Floating-Point Move Instructions 1973 inline void fmr( FloatRegister d, FloatRegister b); 1974 inline void fmr_( FloatRegister d, FloatRegister b); 1975 1976 // inline void mffgpr( FloatRegister d, Register b); 1977 // inline void mftgpr( Register d, FloatRegister b); 1978 inline void cmpb( Register a, Register s, Register b); 1979 inline void popcntb(Register a, Register s); 1980 inline void popcntw(Register a, Register s); 1981 inline void popcntd(Register a, Register s); 1982 1983 inline void fneg( FloatRegister d, FloatRegister b); 1984 inline void fneg_( FloatRegister d, FloatRegister b); 1985 inline void fabs( FloatRegister d, FloatRegister b); 1986 inline void fabs_( FloatRegister d, FloatRegister b); 1987 inline void fnabs( FloatRegister d, FloatRegister b); 1988 inline void fnabs_(FloatRegister d, FloatRegister b); 1989 1990 // PPC 1, section 4.6.5.1 Floating-Point Elementary Arithmetic Instructions 1991 inline void fadd( FloatRegister d, FloatRegister a, FloatRegister b); 1992 inline void fadd_( FloatRegister d, FloatRegister a, FloatRegister b); 1993 inline void fadds( FloatRegister d, FloatRegister a, FloatRegister b); 1994 inline void fadds_(FloatRegister d, FloatRegister a, FloatRegister b); 1995 inline void fsub( FloatRegister d, FloatRegister a, FloatRegister b); 1996 inline void fsub_( FloatRegister d, FloatRegister a, FloatRegister b); 1997 inline void fsubs( FloatRegister d, FloatRegister a, FloatRegister b); 1998 inline void fsubs_(FloatRegister d, FloatRegister a, FloatRegister b); 1999 inline void fmul( FloatRegister d, FloatRegister a, FloatRegister c); 2000 inline void fmul_( FloatRegister d, FloatRegister a, FloatRegister c); 2001 inline void fmuls( FloatRegister d, FloatRegister a, FloatRegister c); 2002 inline void fmuls_(FloatRegister d, FloatRegister a, FloatRegister c); 2003 inline void fdiv( FloatRegister d, FloatRegister a, FloatRegister b); 2004 inline void fdiv_( FloatRegister d, FloatRegister a, FloatRegister b); 2005 inline void fdivs( FloatRegister d, FloatRegister a, FloatRegister b); 2006 inline void fdivs_(FloatRegister d, FloatRegister a, FloatRegister b); 2007 2008 // Fused multiply-accumulate instructions. 2009 // WARNING: Use only when rounding between the 2 parts is not desired. 2010 // Some floating point tck tests will fail if used incorrectly. 2011 inline void fmadd( FloatRegister d, FloatRegister a, FloatRegister c, FloatRegister b); 2012 inline void fmadd_( FloatRegister d, FloatRegister a, FloatRegister c, FloatRegister b); 2013 inline void fmadds( FloatRegister d, FloatRegister a, FloatRegister c, FloatRegister b); 2014 inline void fmadds_( FloatRegister d, FloatRegister a, FloatRegister c, FloatRegister b); 2015 inline void fmsub( FloatRegister d, FloatRegister a, FloatRegister c, FloatRegister b); 2016 inline void fmsub_( FloatRegister d, FloatRegister a, FloatRegister c, FloatRegister b); 2017 inline void fmsubs( FloatRegister d, FloatRegister a, FloatRegister c, FloatRegister b); 2018 inline void fmsubs_( FloatRegister d, FloatRegister a, FloatRegister c, FloatRegister b); 2019 inline void fnmadd( FloatRegister d, FloatRegister a, FloatRegister c, FloatRegister b); 2020 inline void fnmadd_( FloatRegister d, FloatRegister a, FloatRegister c, FloatRegister b); 2021 inline void fnmadds( FloatRegister d, FloatRegister a, FloatRegister c, FloatRegister b); 2022 inline void fnmadds_(FloatRegister d, FloatRegister a, FloatRegister c, FloatRegister b); 2023 inline void fnmsub( FloatRegister d, FloatRegister a, FloatRegister c, FloatRegister b); 2024 inline void fnmsub_( FloatRegister d, FloatRegister a, FloatRegister c, FloatRegister b); 2025 inline void fnmsubs( FloatRegister d, FloatRegister a, FloatRegister c, FloatRegister b); 2026 inline void fnmsubs_(FloatRegister d, FloatRegister a, FloatRegister c, FloatRegister b); 2027 2028 // PPC 1, section 4.6.6 Floating-Point Rounding and Conversion Instructions 2029 inline void frsp( FloatRegister d, FloatRegister b); 2030 inline void fctid( FloatRegister d, FloatRegister b); 2031 inline void fctidz(FloatRegister d, FloatRegister b); 2032 inline void fctiw( FloatRegister d, FloatRegister b); 2033 inline void fctiwz(FloatRegister d, FloatRegister b); 2034 inline void fcfid( FloatRegister d, FloatRegister b); 2035 inline void fcfids(FloatRegister d, FloatRegister b); 2036 2037 // PPC 1, section 4.6.7 Floating-Point Compare Instructions 2038 inline void fcmpu( ConditionRegister crx, FloatRegister a, FloatRegister b); 2039 2040 inline void fsqrt( FloatRegister d, FloatRegister b); 2041 inline void fsqrts(FloatRegister d, FloatRegister b); 2042 2043 // Vector instructions for >= Power6. 2044 inline void lvebx( VectorRegister d, Register s1, Register s2); 2045 inline void lvehx( VectorRegister d, Register s1, Register s2); 2046 inline void lvewx( VectorRegister d, Register s1, Register s2); 2047 inline void lvx( VectorRegister d, Register s1, Register s2); 2048 inline void lvxl( VectorRegister d, Register s1, Register s2); 2049 inline void stvebx( VectorRegister d, Register s1, Register s2); 2050 inline void stvehx( VectorRegister d, Register s1, Register s2); 2051 inline void stvewx( VectorRegister d, Register s1, Register s2); 2052 inline void stvx( VectorRegister d, Register s1, Register s2); 2053 inline void stvxl( VectorRegister d, Register s1, Register s2); 2054 inline void lvsl( VectorRegister d, Register s1, Register s2); 2055 inline void lvsr( VectorRegister d, Register s1, Register s2); 2056 inline void vpkpx( VectorRegister d, VectorRegister a, VectorRegister b); 2057 inline void vpkshss( VectorRegister d, VectorRegister a, VectorRegister b); 2058 inline void vpkswss( VectorRegister d, VectorRegister a, VectorRegister b); 2059 inline void vpkshus( VectorRegister d, VectorRegister a, VectorRegister b); 2060 inline void vpkswus( VectorRegister d, VectorRegister a, VectorRegister b); 2061 inline void vpkuhum( VectorRegister d, VectorRegister a, VectorRegister b); 2062 inline void vpkuwum( VectorRegister d, VectorRegister a, VectorRegister b); 2063 inline void vpkuhus( VectorRegister d, VectorRegister a, VectorRegister b); 2064 inline void vpkuwus( VectorRegister d, VectorRegister a, VectorRegister b); 2065 inline void vupkhpx( VectorRegister d, VectorRegister b); 2066 inline void vupkhsb( VectorRegister d, VectorRegister b); 2067 inline void vupkhsh( VectorRegister d, VectorRegister b); 2068 inline void vupklpx( VectorRegister d, VectorRegister b); 2069 inline void vupklsb( VectorRegister d, VectorRegister b); 2070 inline void vupklsh( VectorRegister d, VectorRegister b); 2071 inline void vmrghb( VectorRegister d, VectorRegister a, VectorRegister b); 2072 inline void vmrghw( VectorRegister d, VectorRegister a, VectorRegister b); 2073 inline void vmrghh( VectorRegister d, VectorRegister a, VectorRegister b); 2074 inline void vmrglb( VectorRegister d, VectorRegister a, VectorRegister b); 2075 inline void vmrglw( VectorRegister d, VectorRegister a, VectorRegister b); 2076 inline void vmrglh( VectorRegister d, VectorRegister a, VectorRegister b); 2077 inline void vsplt( VectorRegister d, int ui4, VectorRegister b); 2078 inline void vsplth( VectorRegister d, int ui3, VectorRegister b); 2079 inline void vspltw( VectorRegister d, int ui2, VectorRegister b); 2080 inline void vspltisb( VectorRegister d, int si5); 2081 inline void vspltish( VectorRegister d, int si5); 2082 inline void vspltisw( VectorRegister d, int si5); 2083 inline void vperm( VectorRegister d, VectorRegister a, VectorRegister b, VectorRegister c); 2084 inline void vsel( VectorRegister d, VectorRegister a, VectorRegister b, VectorRegister c); 2085 inline void vsl( VectorRegister d, VectorRegister a, VectorRegister b); 2086 inline void vsldoi( VectorRegister d, VectorRegister a, VectorRegister b, int ui4); 2087 inline void vslo( VectorRegister d, VectorRegister a, VectorRegister b); 2088 inline void vsr( VectorRegister d, VectorRegister a, VectorRegister b); 2089 inline void vsro( VectorRegister d, VectorRegister a, VectorRegister b); 2090 inline void vaddcuw( VectorRegister d, VectorRegister a, VectorRegister b); 2091 inline void vaddshs( VectorRegister d, VectorRegister a, VectorRegister b); 2092 inline void vaddsbs( VectorRegister d, VectorRegister a, VectorRegister b); 2093 inline void vaddsws( VectorRegister d, VectorRegister a, VectorRegister b); 2094 inline void vaddubm( VectorRegister d, VectorRegister a, VectorRegister b); 2095 inline void vadduwm( VectorRegister d, VectorRegister a, VectorRegister b); 2096 inline void vadduhm( VectorRegister d, VectorRegister a, VectorRegister b); 2097 inline void vaddudm( VectorRegister d, VectorRegister a, VectorRegister b); 2098 inline void vaddubs( VectorRegister d, VectorRegister a, VectorRegister b); 2099 inline void vadduws( VectorRegister d, VectorRegister a, VectorRegister b); 2100 inline void vadduhs( VectorRegister d, VectorRegister a, VectorRegister b); 2101 inline void vaddfp( VectorRegister d, VectorRegister a, VectorRegister b); 2102 inline void vsubcuw( VectorRegister d, VectorRegister a, VectorRegister b); 2103 inline void vsubshs( VectorRegister d, VectorRegister a, VectorRegister b); 2104 inline void vsubsbs( VectorRegister d, VectorRegister a, VectorRegister b); 2105 inline void vsubsws( VectorRegister d, VectorRegister a, VectorRegister b); 2106 inline void vsububm( VectorRegister d, VectorRegister a, VectorRegister b); 2107 inline void vsubuwm( VectorRegister d, VectorRegister a, VectorRegister b); 2108 inline void vsubuhm( VectorRegister d, VectorRegister a, VectorRegister b); 2109 inline void vsubudm( VectorRegister d, VectorRegister a, VectorRegister b); 2110 inline void vsububs( VectorRegister d, VectorRegister a, VectorRegister b); 2111 inline void vsubuws( VectorRegister d, VectorRegister a, VectorRegister b); 2112 inline void vsubuhs( VectorRegister d, VectorRegister a, VectorRegister b); 2113 inline void vsubfp( VectorRegister d, VectorRegister a, VectorRegister b); 2114 inline void vmulesb( VectorRegister d, VectorRegister a, VectorRegister b); 2115 inline void vmuleub( VectorRegister d, VectorRegister a, VectorRegister b); 2116 inline void vmulesh( VectorRegister d, VectorRegister a, VectorRegister b); 2117 inline void vmuleuh( VectorRegister d, VectorRegister a, VectorRegister b); 2118 inline void vmulosb( VectorRegister d, VectorRegister a, VectorRegister b); 2119 inline void vmuloub( VectorRegister d, VectorRegister a, VectorRegister b); 2120 inline void vmulosh( VectorRegister d, VectorRegister a, VectorRegister b); 2121 inline void vmulosw( VectorRegister d, VectorRegister a, VectorRegister b); 2122 inline void vmulouh( VectorRegister d, VectorRegister a, VectorRegister b); 2123 inline void vmuluwm( VectorRegister d, VectorRegister a, VectorRegister b); 2124 inline void vmhaddshs(VectorRegister d, VectorRegister a, VectorRegister b, VectorRegister c); 2125 inline void vmhraddshs(VectorRegister d,VectorRegister a, VectorRegister b, VectorRegister c); 2126 inline void vmladduhm(VectorRegister d, VectorRegister a, VectorRegister b, VectorRegister c); 2127 inline void vmsubuhm( VectorRegister d, VectorRegister a, VectorRegister b, VectorRegister c); 2128 inline void vmsummbm( VectorRegister d, VectorRegister a, VectorRegister b, VectorRegister c); 2129 inline void vmsumshm( VectorRegister d, VectorRegister a, VectorRegister b, VectorRegister c); 2130 inline void vmsumshs( VectorRegister d, VectorRegister a, VectorRegister b, VectorRegister c); 2131 inline void vmsumuhm( VectorRegister d, VectorRegister a, VectorRegister b, VectorRegister c); 2132 inline void vmsumuhs( VectorRegister d, VectorRegister a, VectorRegister b, VectorRegister c); 2133 inline void vmaddfp( VectorRegister d, VectorRegister a, VectorRegister b, VectorRegister c); 2134 inline void vsumsws( VectorRegister d, VectorRegister a, VectorRegister b); 2135 inline void vsum2sws( VectorRegister d, VectorRegister a, VectorRegister b); 2136 inline void vsum4sbs( VectorRegister d, VectorRegister a, VectorRegister b); 2137 inline void vsum4ubs( VectorRegister d, VectorRegister a, VectorRegister b); 2138 inline void vsum4shs( VectorRegister d, VectorRegister a, VectorRegister b); 2139 inline void vavgsb( VectorRegister d, VectorRegister a, VectorRegister b); 2140 inline void vavgsw( VectorRegister d, VectorRegister a, VectorRegister b); 2141 inline void vavgsh( VectorRegister d, VectorRegister a, VectorRegister b); 2142 inline void vavgub( VectorRegister d, VectorRegister a, VectorRegister b); 2143 inline void vavguw( VectorRegister d, VectorRegister a, VectorRegister b); 2144 inline void vavguh( VectorRegister d, VectorRegister a, VectorRegister b); 2145 inline void vmaxsb( VectorRegister d, VectorRegister a, VectorRegister b); 2146 inline void vmaxsw( VectorRegister d, VectorRegister a, VectorRegister b); 2147 inline void vmaxsh( VectorRegister d, VectorRegister a, VectorRegister b); 2148 inline void vmaxub( VectorRegister d, VectorRegister a, VectorRegister b); 2149 inline void vmaxuw( VectorRegister d, VectorRegister a, VectorRegister b); 2150 inline void vmaxuh( VectorRegister d, VectorRegister a, VectorRegister b); 2151 inline void vminsb( VectorRegister d, VectorRegister a, VectorRegister b); 2152 inline void vminsw( VectorRegister d, VectorRegister a, VectorRegister b); 2153 inline void vminsh( VectorRegister d, VectorRegister a, VectorRegister b); 2154 inline void vminub( VectorRegister d, VectorRegister a, VectorRegister b); 2155 inline void vminuw( VectorRegister d, VectorRegister a, VectorRegister b); 2156 inline void vminuh( VectorRegister d, VectorRegister a, VectorRegister b); 2157 inline void vcmpequb( VectorRegister d, VectorRegister a, VectorRegister b); 2158 inline void vcmpequh( VectorRegister d, VectorRegister a, VectorRegister b); 2159 inline void vcmpequw( VectorRegister d, VectorRegister a, VectorRegister b); 2160 inline void vcmpgtsh( VectorRegister d, VectorRegister a, VectorRegister b); 2161 inline void vcmpgtsb( VectorRegister d, VectorRegister a, VectorRegister b); 2162 inline void vcmpgtsw( VectorRegister d, VectorRegister a, VectorRegister b); 2163 inline void vcmpgtub( VectorRegister d, VectorRegister a, VectorRegister b); 2164 inline void vcmpgtuh( VectorRegister d, VectorRegister a, VectorRegister b); 2165 inline void vcmpgtuw( VectorRegister d, VectorRegister a, VectorRegister b); 2166 inline void vcmpequb_(VectorRegister d, VectorRegister a, VectorRegister b); 2167 inline void vcmpequh_(VectorRegister d, VectorRegister a, VectorRegister b); 2168 inline void vcmpequw_(VectorRegister d, VectorRegister a, VectorRegister b); 2169 inline void vcmpgtsh_(VectorRegister d, VectorRegister a, VectorRegister b); 2170 inline void vcmpgtsb_(VectorRegister d, VectorRegister a, VectorRegister b); 2171 inline void vcmpgtsw_(VectorRegister d, VectorRegister a, VectorRegister b); 2172 inline void vcmpgtub_(VectorRegister d, VectorRegister a, VectorRegister b); 2173 inline void vcmpgtuh_(VectorRegister d, VectorRegister a, VectorRegister b); 2174 inline void vcmpgtuw_(VectorRegister d, VectorRegister a, VectorRegister b); 2175 inline void vand( VectorRegister d, VectorRegister a, VectorRegister b); 2176 inline void vandc( VectorRegister d, VectorRegister a, VectorRegister b); 2177 inline void vnor( VectorRegister d, VectorRegister a, VectorRegister b); 2178 inline void vor( VectorRegister d, VectorRegister a, VectorRegister b); 2179 inline void vmr( VectorRegister d, VectorRegister a); 2180 inline void vxor( VectorRegister d, VectorRegister a, VectorRegister b); 2181 inline void vrld( VectorRegister d, VectorRegister a, VectorRegister b); 2182 inline void vrlb( VectorRegister d, VectorRegister a, VectorRegister b); 2183 inline void vrlw( VectorRegister d, VectorRegister a, VectorRegister b); 2184 inline void vrlh( VectorRegister d, VectorRegister a, VectorRegister b); 2185 inline void vslb( VectorRegister d, VectorRegister a, VectorRegister b); 2186 inline void vskw( VectorRegister d, VectorRegister a, VectorRegister b); 2187 inline void vslh( VectorRegister d, VectorRegister a, VectorRegister b); 2188 inline void vsrb( VectorRegister d, VectorRegister a, VectorRegister b); 2189 inline void vsrw( VectorRegister d, VectorRegister a, VectorRegister b); 2190 inline void vsrh( VectorRegister d, VectorRegister a, VectorRegister b); 2191 inline void vsrab( VectorRegister d, VectorRegister a, VectorRegister b); 2192 inline void vsraw( VectorRegister d, VectorRegister a, VectorRegister b); 2193 inline void vsrah( VectorRegister d, VectorRegister a, VectorRegister b); 2194 inline void vpopcntw( VectorRegister d, VectorRegister b); 2195 // Vector Floating-Point not implemented yet 2196 inline void mtvscr( VectorRegister b); 2197 inline void mfvscr( VectorRegister d); 2198 2199 // Vector-Scalar (VSX) instructions. 2200 inline void lxvd2x( VectorSRegister d, Register a); 2201 inline void lxvd2x( VectorSRegister d, Register a, Register b); 2202 inline void stxvd2x( VectorSRegister d, Register a); 2203 inline void stxvd2x( VectorSRegister d, Register a, Register b); 2204 inline void mtvrwz( VectorRegister d, Register a); 2205 inline void mfvrwz( Register a, VectorRegister d); 2206 inline void mtvrd( VectorRegister d, Register a); 2207 inline void mfvrd( Register a, VectorRegister d); 2208 inline void xxpermdi( VectorSRegister d, VectorSRegister a, VectorSRegister b, int dm); 2209 inline void xxmrghw( VectorSRegister d, VectorSRegister a, VectorSRegister b); 2210 inline void xxmrglw( VectorSRegister d, VectorSRegister a, VectorSRegister b); 2211 inline void mtvsrd( VectorSRegister d, Register a); 2212 inline void mtvsrwz( VectorSRegister d, Register a); 2213 inline void xxspltw( VectorSRegister d, VectorSRegister b, int ui2); 2214 inline void xxlor( VectorSRegister d, VectorSRegister a, VectorSRegister b); 2215 inline void xxlxor( VectorSRegister d, VectorSRegister a, VectorSRegister b); 2216 inline void xxleqv( VectorSRegister d, VectorSRegister a, VectorSRegister b); 2217 inline void xvdivsp( VectorSRegister d, VectorSRegister a, VectorSRegister b); 2218 inline void xvdivdp( VectorSRegister d, VectorSRegister a, VectorSRegister b); 2219 inline void xvabssp( VectorSRegister d, VectorSRegister b); 2220 inline void xvabsdp( VectorSRegister d, VectorSRegister b); 2221 inline void xvnegsp( VectorSRegister d, VectorSRegister b); 2222 inline void xvnegdp( VectorSRegister d, VectorSRegister b); 2223 inline void xvsqrtsp( VectorSRegister d, VectorSRegister b); 2224 inline void xvsqrtdp( VectorSRegister d, VectorSRegister b); 2225 inline void xscvdpspn(VectorSRegister d, VectorSRegister b); 2226 inline void xvadddp( VectorSRegister d, VectorSRegister a, VectorSRegister b); 2227 inline void xvsubdp( VectorSRegister d, VectorSRegister a, VectorSRegister b); 2228 inline void xvmulsp( VectorSRegister d, VectorSRegister a, VectorSRegister b); 2229 inline void xvmuldp( VectorSRegister d, VectorSRegister a, VectorSRegister b); 2230 2231 // VSX Extended Mnemonics 2232 inline void xxspltd( VectorSRegister d, VectorSRegister a, int x); 2233 inline void xxmrghd( VectorSRegister d, VectorSRegister a, VectorSRegister b); 2234 inline void xxmrgld( VectorSRegister d, VectorSRegister a, VectorSRegister b); 2235 inline void xxswapd( VectorSRegister d, VectorSRegister a); 2236 2237 // Vector-Scalar (VSX) instructions. 2238 inline void mtfprd( FloatRegister d, Register a); 2239 inline void mtfprwa( FloatRegister d, Register a); 2240 inline void mffprd( Register a, FloatRegister d); 2241 2242 // Deliver A Random Number (introduced with POWER9) 2243 inline void darn( Register d, int l = 1 /*L=CRN*/); 2244 2245 // AES (introduced with Power 8) 2246 inline void vcipher( VectorRegister d, VectorRegister a, VectorRegister b); 2247 inline void vcipherlast( VectorRegister d, VectorRegister a, VectorRegister b); 2248 inline void vncipher( VectorRegister d, VectorRegister a, VectorRegister b); 2249 inline void vncipherlast(VectorRegister d, VectorRegister a, VectorRegister b); 2250 inline void vsbox( VectorRegister d, VectorRegister a); 2251 2252 // SHA (introduced with Power 8) 2253 inline void vshasigmad(VectorRegister d, VectorRegister a, bool st, int six); 2254 inline void vshasigmaw(VectorRegister d, VectorRegister a, bool st, int six); 2255 2256 // Vector Binary Polynomial Multiplication (introduced with Power 8) 2257 inline void vpmsumb( VectorRegister d, VectorRegister a, VectorRegister b); 2258 inline void vpmsumd( VectorRegister d, VectorRegister a, VectorRegister b); 2259 inline void vpmsumh( VectorRegister d, VectorRegister a, VectorRegister b); 2260 inline void vpmsumw( VectorRegister d, VectorRegister a, VectorRegister b); 2261 2262 // Vector Permute and Xor (introduced with Power 8) 2263 inline void vpermxor( VectorRegister d, VectorRegister a, VectorRegister b, VectorRegister c); 2264 2265 // Transactional Memory instructions (introduced with Power 8) 2266 inline void tbegin_(); // R=0 2267 inline void tbeginrot_(); // R=1 Rollback-Only Transaction 2268 inline void tend_(); // A=0 2269 inline void tendall_(); // A=1 2270 inline void tabort_(); 2271 inline void tabort_(Register a); 2272 inline void tabortwc_(int t, Register a, Register b); 2273 inline void tabortwci_(int t, Register a, int si); 2274 inline void tabortdc_(int t, Register a, Register b); 2275 inline void tabortdci_(int t, Register a, int si); 2276 inline void tsuspend_(); // tsr with L=0 2277 inline void tresume_(); // tsr with L=1 2278 inline void tcheck(int f); 2279 2280 static bool is_tbegin(int x) { 2281 return TBEGIN_OPCODE == (x & (0x3f << OPCODE_SHIFT | 0x3ff << 1)); 2282 } 2283 2284 // The following encoders use r0 as second operand. These instructions 2285 // read r0 as '0'. 2286 inline void lwzx( Register d, Register s2); 2287 inline void lwz( Register d, int si16); 2288 inline void lwax( Register d, Register s2); 2289 inline void lwa( Register d, int si16); 2290 inline void lwbrx(Register d, Register s2); 2291 inline void lhzx( Register d, Register s2); 2292 inline void lhz( Register d, int si16); 2293 inline void lhax( Register d, Register s2); 2294 inline void lha( Register d, int si16); 2295 inline void lhbrx(Register d, Register s2); 2296 inline void lbzx( Register d, Register s2); 2297 inline void lbz( Register d, int si16); 2298 inline void ldx( Register d, Register s2); 2299 inline void ld( Register d, int si16); 2300 inline void ldbrx(Register d, Register s2); 2301 inline void stwx( Register d, Register s2); 2302 inline void stw( Register d, int si16); 2303 inline void stwbrx( Register d, Register s2); 2304 inline void sthx( Register d, Register s2); 2305 inline void sth( Register d, int si16); 2306 inline void sthbrx( Register d, Register s2); 2307 inline void stbx( Register d, Register s2); 2308 inline void stb( Register d, int si16); 2309 inline void stdx( Register d, Register s2); 2310 inline void std( Register d, int si16); 2311 inline void stdbrx( Register d, Register s2); 2312 2313 // PPC 2, section 3.2.1 Instruction Cache Instructions 2314 inline void icbi( Register s2); 2315 // PPC 2, section 3.2.2 Data Cache Instructions 2316 //inlinevoid dcba( Register s2); // Instruction for embedded processor only. 2317 inline void dcbz( Register s2); 2318 inline void dcbst( Register s2); 2319 inline void dcbf( Register s2); 2320 // dcache read hint 2321 inline void dcbt( Register s2); 2322 inline void dcbtct( Register s2, int ct); 2323 inline void dcbtds( Register s2, int ds); 2324 // dcache write hint 2325 inline void dcbtst( Register s2); 2326 inline void dcbtstct(Register s2, int ct); 2327 2328 // Atomics: use ra0mem to disallow R0 as base. 2329 inline void lbarx_unchecked(Register d, Register b, int eh1); 2330 inline void lharx_unchecked(Register d, Register b, int eh1); 2331 inline void lwarx_unchecked(Register d, Register b, int eh1); 2332 inline void ldarx_unchecked(Register d, Register b, int eh1); 2333 inline void lqarx_unchecked(Register d, Register b, int eh1); 2334 inline void lbarx( Register d, Register b, bool hint_exclusive_access); 2335 inline void lharx( Register d, Register b, bool hint_exclusive_access); 2336 inline void lwarx( Register d, Register b, bool hint_exclusive_access); 2337 inline void ldarx( Register d, Register b, bool hint_exclusive_access); 2338 inline void lqarx( Register d, Register b, bool hint_exclusive_access); 2339 inline void stbcx_(Register s, Register b); 2340 inline void sthcx_(Register s, Register b); 2341 inline void stwcx_(Register s, Register b); 2342 inline void stdcx_(Register s, Register b); 2343 inline void stqcx_(Register s, Register b); 2344 inline void lfs( FloatRegister d, int si16); 2345 inline void lfsx( FloatRegister d, Register b); 2346 inline void lfd( FloatRegister d, int si16); 2347 inline void lfdx( FloatRegister d, Register b); 2348 inline void stfs( FloatRegister s, int si16); 2349 inline void stfsx( FloatRegister s, Register b); 2350 inline void stfd( FloatRegister s, int si16); 2351 inline void stfdx( FloatRegister s, Register b); 2352 inline void lvebx( VectorRegister d, Register s2); 2353 inline void lvehx( VectorRegister d, Register s2); 2354 inline void lvewx( VectorRegister d, Register s2); 2355 inline void lvx( VectorRegister d, Register s2); 2356 inline void lvxl( VectorRegister d, Register s2); 2357 inline void stvebx(VectorRegister d, Register s2); 2358 inline void stvehx(VectorRegister d, Register s2); 2359 inline void stvewx(VectorRegister d, Register s2); 2360 inline void stvx( VectorRegister d, Register s2); 2361 inline void stvxl( VectorRegister d, Register s2); 2362 inline void lvsl( VectorRegister d, Register s2); 2363 inline void lvsr( VectorRegister d, Register s2); 2364 2365 // Endianess specific concatenation of 2 loaded vectors. 2366 inline void load_perm(VectorRegister perm, Register addr); 2367 inline void vec_perm(VectorRegister first_dest, VectorRegister second, VectorRegister perm); 2368 inline void vec_perm(VectorRegister dest, VectorRegister first, VectorRegister second, VectorRegister perm); 2369 2370 // RegisterOrConstant versions. 2371 // These emitters choose between the versions using two registers and 2372 // those with register and immediate, depending on the content of roc. 2373 // If the constant is not encodable as immediate, instructions to 2374 // load the constant are emitted beforehand. Store instructions need a 2375 // tmp reg if the constant is not encodable as immediate. 2376 // Size unpredictable. 2377 void ld( Register d, RegisterOrConstant roc, Register s1 = noreg); 2378 void lwa( Register d, RegisterOrConstant roc, Register s1 = noreg); 2379 void lwz( Register d, RegisterOrConstant roc, Register s1 = noreg); 2380 void lha( Register d, RegisterOrConstant roc, Register s1 = noreg); 2381 void lhz( Register d, RegisterOrConstant roc, Register s1 = noreg); 2382 void lbz( Register d, RegisterOrConstant roc, Register s1 = noreg); 2383 void std( Register d, RegisterOrConstant roc, Register s1 = noreg, Register tmp = noreg); 2384 void stw( Register d, RegisterOrConstant roc, Register s1 = noreg, Register tmp = noreg); 2385 void sth( Register d, RegisterOrConstant roc, Register s1 = noreg, Register tmp = noreg); 2386 void stb( Register d, RegisterOrConstant roc, Register s1 = noreg, Register tmp = noreg); 2387 void add( Register d, RegisterOrConstant roc, Register s1); 2388 void subf(Register d, RegisterOrConstant roc, Register s1); 2389 void cmpd(ConditionRegister d, RegisterOrConstant roc, Register s1); 2390 // Load pointer d from s1+roc. 2391 void ld_ptr(Register d, RegisterOrConstant roc, Register s1 = noreg) { ld(d, roc, s1); } 2392 2393 // Emit several instructions to load a 64 bit constant. This issues a fixed 2394 // instruction pattern so that the constant can be patched later on. 2395 enum { 2396 load_const_size = 5 * BytesPerInstWord 2397 }; 2398 void load_const(Register d, long a, Register tmp = noreg); 2399 inline void load_const(Register d, void* a, Register tmp = noreg); 2400 inline void load_const(Register d, Label& L, Register tmp = noreg); 2401 inline void load_const(Register d, AddressLiteral& a, Register tmp = noreg); 2402 inline void load_const32(Register d, int i); // load signed int (patchable) 2403 2404 // Load a 64 bit constant, optimized, not identifyable. 2405 // Tmp can be used to increase ILP. Set return_simm16_rest = true to get a 2406 // 16 bit immediate offset. This is useful if the offset can be encoded in 2407 // a succeeding instruction. 2408 int load_const_optimized(Register d, long a, Register tmp = noreg, bool return_simm16_rest = false); 2409 inline int load_const_optimized(Register d, void* a, Register tmp = noreg, bool return_simm16_rest = false) { 2410 return load_const_optimized(d, (long)(unsigned long)a, tmp, return_simm16_rest); 2411 } 2412 2413 // If return_simm16_rest, the return value needs to get added afterwards. 2414 int add_const_optimized(Register d, Register s, long x, Register tmp = R0, bool return_simm16_rest = false); 2415 inline int add_const_optimized(Register d, Register s, void* a, Register tmp = R0, bool return_simm16_rest = false) { 2416 return add_const_optimized(d, s, (long)(unsigned long)a, tmp, return_simm16_rest); 2417 } 2418 2419 // If return_simm16_rest, the return value needs to get added afterwards. 2420 inline int sub_const_optimized(Register d, Register s, long x, Register tmp = R0, bool return_simm16_rest = false) { 2421 return add_const_optimized(d, s, -x, tmp, return_simm16_rest); 2422 } 2423 inline int sub_const_optimized(Register d, Register s, void* a, Register tmp = R0, bool return_simm16_rest = false) { 2424 return sub_const_optimized(d, s, (long)(unsigned long)a, tmp, return_simm16_rest); 2425 } 2426 2427 // Creation 2428 Assembler(CodeBuffer* code) : AbstractAssembler(code) { 2429 #ifdef CHECK_DELAY 2430 delay_state = no_delay; 2431 #endif 2432 } 2433 2434 // Testing 2435 #ifndef PRODUCT 2436 void test_asm(); 2437 #endif 2438 }; 2439 2440 2441 #endif // CPU_PPC_VM_ASSEMBLER_PPC_HPP