1 /* 2 * Copyright (c) 2002, 2015, Oracle and/or its affiliates. All rights reserved. 3 * Copyright 2012, 2015 SAP AG. 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 // So far we do not use this as simplification by this class is low 35 // on PPC with its simple addressing mode. Use RegisterOrConstant to 36 // represent an offset. 37 class Address VALUE_OBJ_CLASS_SPEC { 38 }; 39 40 class AddressLiteral VALUE_OBJ_CLASS_SPEC { 41 private: 42 address _address; 43 RelocationHolder _rspec; 44 45 RelocationHolder rspec_from_rtype(relocInfo::relocType rtype, address addr) { 46 switch (rtype) { 47 case relocInfo::external_word_type: 48 return external_word_Relocation::spec(addr); 49 case relocInfo::internal_word_type: 50 return internal_word_Relocation::spec(addr); 51 case relocInfo::opt_virtual_call_type: 52 return opt_virtual_call_Relocation::spec(); 53 case relocInfo::static_call_type: 54 return static_call_Relocation::spec(); 55 case relocInfo::runtime_call_type: 56 return runtime_call_Relocation::spec(); 57 case relocInfo::none: 58 return RelocationHolder(); 59 default: 60 ShouldNotReachHere(); 61 return RelocationHolder(); 62 } 63 } 64 65 protected: 66 // creation 67 AddressLiteral() : _address(NULL), _rspec(NULL) {} 68 69 public: 70 AddressLiteral(address addr, RelocationHolder const& rspec) 71 : _address(addr), 72 _rspec(rspec) {} 73 74 AddressLiteral(address addr, relocInfo::relocType rtype = relocInfo::none) 75 : _address((address) addr), 76 _rspec(rspec_from_rtype(rtype, (address) addr)) {} 77 78 AddressLiteral(oop* addr, relocInfo::relocType rtype = relocInfo::none) 79 : _address((address) addr), 80 _rspec(rspec_from_rtype(rtype, (address) addr)) {} 81 82 intptr_t value() const { return (intptr_t) _address; } 83 84 const RelocationHolder& rspec() const { return _rspec; } 85 }; 86 87 // Argument is an abstraction used to represent an outgoing 88 // actual argument or an incoming formal parameter, whether 89 // it resides in memory or in a register, in a manner consistent 90 // with the PPC Application Binary Interface, or ABI. This is 91 // often referred to as the native or C calling convention. 92 93 class Argument VALUE_OBJ_CLASS_SPEC { 94 private: 95 int _number; // The number of the argument. 96 public: 97 enum { 98 // Only 8 registers may contain integer parameters. 99 n_register_parameters = 8, 100 // Can have up to 8 floating registers. 101 n_float_register_parameters = 8, 102 103 // PPC C calling conventions. 104 // The first eight arguments are passed in int regs if they are int. 105 n_int_register_parameters_c = 8, 106 // The first thirteen float arguments are passed in float regs. 107 n_float_register_parameters_c = 13, 108 // Only the first 8 parameters are not placed on the stack. Aix disassembly 109 // shows that xlC places all float args after argument 8 on the stack AND 110 // in a register. This is not documented, but we follow this convention, too. 111 n_regs_not_on_stack_c = 8, 112 }; 113 // creation 114 Argument(int number) : _number(number) {} 115 116 int number() const { return _number; } 117 118 // Locating register-based arguments: 119 bool is_register() const { return _number < n_register_parameters; } 120 121 Register as_register() const { 122 assert(is_register(), "must be a register argument"); 123 return as_Register(number() + R3_ARG1->encoding()); 124 } 125 }; 126 127 #if !defined(ABI_ELFv2) 128 // A ppc64 function descriptor. 129 struct FunctionDescriptor VALUE_OBJ_CLASS_SPEC { 130 private: 131 address _entry; 132 address _toc; 133 address _env; 134 135 public: 136 inline address entry() const { return _entry; } 137 inline address toc() const { return _toc; } 138 inline address env() const { return _env; } 139 140 inline void set_entry(address entry) { _entry = entry; } 141 inline void set_toc( address toc) { _toc = toc; } 142 inline void set_env( address env) { _env = env; } 143 144 inline static ByteSize entry_offset() { return byte_offset_of(FunctionDescriptor, _entry); } 145 inline static ByteSize toc_offset() { return byte_offset_of(FunctionDescriptor, _toc); } 146 inline static ByteSize env_offset() { return byte_offset_of(FunctionDescriptor, _env); } 147 148 // Friend functions can be called without loading toc and env. 149 enum { 150 friend_toc = 0xcafe, 151 friend_env = 0xc0de 152 }; 153 154 inline bool is_friend_function() const { 155 return (toc() == (address) friend_toc) && (env() == (address) friend_env); 156 } 157 158 // Constructor for stack-allocated instances. 159 FunctionDescriptor() { 160 _entry = (address) 0xbad; 161 _toc = (address) 0xbad; 162 _env = (address) 0xbad; 163 } 164 }; 165 #endif 166 167 class Assembler : public AbstractAssembler { 168 protected: 169 // Displacement routines 170 static void print_instruction(int inst); 171 static int patched_branch(int dest_pos, int inst, int inst_pos); 172 static int branch_destination(int inst, int pos); 173 174 friend class AbstractAssembler; 175 176 // Code patchers need various routines like inv_wdisp() 177 friend class NativeInstruction; 178 friend class NativeGeneralJump; 179 friend class Relocation; 180 181 public: 182 183 enum shifts { 184 XO_21_29_SHIFT = 2, 185 XO_21_30_SHIFT = 1, 186 XO_27_29_SHIFT = 2, 187 XO_30_31_SHIFT = 0, 188 SPR_5_9_SHIFT = 11u, // SPR_5_9 field in bits 11 -- 15 189 SPR_0_4_SHIFT = 16u, // SPR_0_4 field in bits 16 -- 20 190 RS_SHIFT = 21u, // RS field in bits 21 -- 25 191 OPCODE_SHIFT = 26u, // opcode in bits 26 -- 31 192 }; 193 194 enum opcdxos_masks { 195 XL_FORM_OPCODE_MASK = (63u << OPCODE_SHIFT) | (1023u << 1), 196 ADDI_OPCODE_MASK = (63u << OPCODE_SHIFT), 197 ADDIS_OPCODE_MASK = (63u << OPCODE_SHIFT), 198 BXX_OPCODE_MASK = (63u << OPCODE_SHIFT), 199 BCXX_OPCODE_MASK = (63u << OPCODE_SHIFT), 200 // trap instructions 201 TDI_OPCODE_MASK = (63u << OPCODE_SHIFT), 202 TWI_OPCODE_MASK = (63u << OPCODE_SHIFT), 203 TD_OPCODE_MASK = (63u << OPCODE_SHIFT) | (1023u << 1), 204 TW_OPCODE_MASK = (63u << OPCODE_SHIFT) | (1023u << 1), 205 LD_OPCODE_MASK = (63u << OPCODE_SHIFT) | (3u << XO_30_31_SHIFT), // DS-FORM 206 STD_OPCODE_MASK = LD_OPCODE_MASK, 207 STDU_OPCODE_MASK = STD_OPCODE_MASK, 208 STDX_OPCODE_MASK = (63u << OPCODE_SHIFT) | (1023u << 1), 209 STDUX_OPCODE_MASK = STDX_OPCODE_MASK, 210 STW_OPCODE_MASK = (63u << OPCODE_SHIFT), 211 STWU_OPCODE_MASK = STW_OPCODE_MASK, 212 STWX_OPCODE_MASK = (63u << OPCODE_SHIFT) | (1023u << 1), 213 STWUX_OPCODE_MASK = STWX_OPCODE_MASK, 214 MTCTR_OPCODE_MASK = ~(31u << RS_SHIFT), 215 ORI_OPCODE_MASK = (63u << OPCODE_SHIFT), 216 ORIS_OPCODE_MASK = (63u << OPCODE_SHIFT), 217 RLDICR_OPCODE_MASK = (63u << OPCODE_SHIFT) | (7u << XO_27_29_SHIFT) 218 }; 219 220 enum opcdxos { 221 ADD_OPCODE = (31u << OPCODE_SHIFT | 266u << 1), 222 ADDC_OPCODE = (31u << OPCODE_SHIFT | 10u << 1), 223 ADDI_OPCODE = (14u << OPCODE_SHIFT), 224 ADDIS_OPCODE = (15u << OPCODE_SHIFT), 225 ADDIC__OPCODE = (13u << OPCODE_SHIFT), 226 ADDE_OPCODE = (31u << OPCODE_SHIFT | 138u << 1), 227 SUBF_OPCODE = (31u << OPCODE_SHIFT | 40u << 1), 228 SUBFC_OPCODE = (31u << OPCODE_SHIFT | 8u << 1), 229 SUBFE_OPCODE = (31u << OPCODE_SHIFT | 136u << 1), 230 SUBFIC_OPCODE = (8u << OPCODE_SHIFT), 231 SUBFZE_OPCODE = (31u << OPCODE_SHIFT | 200u << 1), 232 DIVW_OPCODE = (31u << OPCODE_SHIFT | 491u << 1), 233 MULLW_OPCODE = (31u << OPCODE_SHIFT | 235u << 1), 234 MULHW_OPCODE = (31u << OPCODE_SHIFT | 75u << 1), 235 MULHWU_OPCODE = (31u << OPCODE_SHIFT | 11u << 1), 236 MULLI_OPCODE = (7u << OPCODE_SHIFT), 237 AND_OPCODE = (31u << OPCODE_SHIFT | 28u << 1), 238 ANDI_OPCODE = (28u << OPCODE_SHIFT), 239 ANDIS_OPCODE = (29u << OPCODE_SHIFT), 240 ANDC_OPCODE = (31u << OPCODE_SHIFT | 60u << 1), 241 ORC_OPCODE = (31u << OPCODE_SHIFT | 412u << 1), 242 OR_OPCODE = (31u << OPCODE_SHIFT | 444u << 1), 243 ORI_OPCODE = (24u << OPCODE_SHIFT), 244 ORIS_OPCODE = (25u << OPCODE_SHIFT), 245 XOR_OPCODE = (31u << OPCODE_SHIFT | 316u << 1), 246 XORI_OPCODE = (26u << OPCODE_SHIFT), 247 XORIS_OPCODE = (27u << OPCODE_SHIFT), 248 249 NEG_OPCODE = (31u << OPCODE_SHIFT | 104u << 1), 250 251 RLWINM_OPCODE = (21u << OPCODE_SHIFT), 252 CLRRWI_OPCODE = RLWINM_OPCODE, 253 CLRLWI_OPCODE = RLWINM_OPCODE, 254 255 RLWIMI_OPCODE = (20u << OPCODE_SHIFT), 256 257 SLW_OPCODE = (31u << OPCODE_SHIFT | 24u << 1), 258 SLWI_OPCODE = RLWINM_OPCODE, 259 SRW_OPCODE = (31u << OPCODE_SHIFT | 536u << 1), 260 SRWI_OPCODE = RLWINM_OPCODE, 261 SRAW_OPCODE = (31u << OPCODE_SHIFT | 792u << 1), 262 SRAWI_OPCODE = (31u << OPCODE_SHIFT | 824u << 1), 263 264 CMP_OPCODE = (31u << OPCODE_SHIFT | 0u << 1), 265 CMPI_OPCODE = (11u << OPCODE_SHIFT), 266 CMPL_OPCODE = (31u << OPCODE_SHIFT | 32u << 1), 267 CMPLI_OPCODE = (10u << OPCODE_SHIFT), 268 269 ISEL_OPCODE = (31u << OPCODE_SHIFT | 15u << 1), 270 271 // Special purpose registers 272 MTSPR_OPCODE = (31u << OPCODE_SHIFT | 467u << 1), 273 MFSPR_OPCODE = (31u << OPCODE_SHIFT | 339u << 1), 274 275 MTXER_OPCODE = (MTSPR_OPCODE | 1 << SPR_0_4_SHIFT), 276 MFXER_OPCODE = (MFSPR_OPCODE | 1 << SPR_0_4_SHIFT), 277 278 MTDSCR_OPCODE = (MTSPR_OPCODE | 3 << SPR_0_4_SHIFT), 279 MFDSCR_OPCODE = (MFSPR_OPCODE | 3 << SPR_0_4_SHIFT), 280 281 MTLR_OPCODE = (MTSPR_OPCODE | 8 << SPR_0_4_SHIFT), 282 MFLR_OPCODE = (MFSPR_OPCODE | 8 << SPR_0_4_SHIFT), 283 284 MTCTR_OPCODE = (MTSPR_OPCODE | 9 << SPR_0_4_SHIFT), 285 MFCTR_OPCODE = (MFSPR_OPCODE | 9 << SPR_0_4_SHIFT), 286 287 // Attention: Higher and lower half are inserted in reversed order. 288 MTTFHAR_OPCODE = (MTSPR_OPCODE | 4 << SPR_5_9_SHIFT | 0 << SPR_0_4_SHIFT), 289 MFTFHAR_OPCODE = (MFSPR_OPCODE | 4 << SPR_5_9_SHIFT | 0 << SPR_0_4_SHIFT), 290 MTTFIAR_OPCODE = (MTSPR_OPCODE | 4 << SPR_5_9_SHIFT | 1 << SPR_0_4_SHIFT), 291 MFTFIAR_OPCODE = (MFSPR_OPCODE | 4 << SPR_5_9_SHIFT | 1 << SPR_0_4_SHIFT), 292 MTTEXASR_OPCODE = (MTSPR_OPCODE | 4 << SPR_5_9_SHIFT | 2 << SPR_0_4_SHIFT), 293 MFTEXASR_OPCODE = (MFSPR_OPCODE | 4 << SPR_5_9_SHIFT | 2 << SPR_0_4_SHIFT), 294 MTTEXASRU_OPCODE = (MTSPR_OPCODE | 4 << SPR_5_9_SHIFT | 3 << SPR_0_4_SHIFT), 295 MFTEXASRU_OPCODE = (MFSPR_OPCODE | 4 << SPR_5_9_SHIFT | 3 << SPR_0_4_SHIFT), 296 297 MTVRSAVE_OPCODE = (MTSPR_OPCODE | 8 << SPR_5_9_SHIFT | 0 << SPR_0_4_SHIFT), 298 MFVRSAVE_OPCODE = (MFSPR_OPCODE | 8 << SPR_5_9_SHIFT | 0 << SPR_0_4_SHIFT), 299 300 MFTB_OPCODE = (MFSPR_OPCODE | 8 << SPR_5_9_SHIFT | 12 << SPR_0_4_SHIFT), 301 302 MTCRF_OPCODE = (31u << OPCODE_SHIFT | 144u << 1), 303 MFCR_OPCODE = (31u << OPCODE_SHIFT | 19u << 1), 304 MCRF_OPCODE = (19u << OPCODE_SHIFT | 0u << 1), 305 306 // condition register logic instructions 307 CRAND_OPCODE = (19u << OPCODE_SHIFT | 257u << 1), 308 CRNAND_OPCODE = (19u << OPCODE_SHIFT | 225u << 1), 309 CROR_OPCODE = (19u << OPCODE_SHIFT | 449u << 1), 310 CRXOR_OPCODE = (19u << OPCODE_SHIFT | 193u << 1), 311 CRNOR_OPCODE = (19u << OPCODE_SHIFT | 33u << 1), 312 CREQV_OPCODE = (19u << OPCODE_SHIFT | 289u << 1), 313 CRANDC_OPCODE = (19u << OPCODE_SHIFT | 129u << 1), 314 CRORC_OPCODE = (19u << OPCODE_SHIFT | 417u << 1), 315 316 BCLR_OPCODE = (19u << OPCODE_SHIFT | 16u << 1), 317 BXX_OPCODE = (18u << OPCODE_SHIFT), 318 BCXX_OPCODE = (16u << OPCODE_SHIFT), 319 320 // CTR-related opcodes 321 BCCTR_OPCODE = (19u << OPCODE_SHIFT | 528u << 1), 322 323 LWZ_OPCODE = (32u << OPCODE_SHIFT), 324 LWZX_OPCODE = (31u << OPCODE_SHIFT | 23u << 1), 325 LWZU_OPCODE = (33u << OPCODE_SHIFT), 326 LWBRX_OPCODE = (31u << OPCODE_SHIFT | 534 << 1), 327 328 LHA_OPCODE = (42u << OPCODE_SHIFT), 329 LHAX_OPCODE = (31u << OPCODE_SHIFT | 343u << 1), 330 LHAU_OPCODE = (43u << OPCODE_SHIFT), 331 332 LHZ_OPCODE = (40u << OPCODE_SHIFT), 333 LHZX_OPCODE = (31u << OPCODE_SHIFT | 279u << 1), 334 LHZU_OPCODE = (41u << OPCODE_SHIFT), 335 LHBRX_OPCODE = (31u << OPCODE_SHIFT | 790 << 1), 336 337 LBZ_OPCODE = (34u << OPCODE_SHIFT), 338 LBZX_OPCODE = (31u << OPCODE_SHIFT | 87u << 1), 339 LBZU_OPCODE = (35u << OPCODE_SHIFT), 340 341 STW_OPCODE = (36u << OPCODE_SHIFT), 342 STWX_OPCODE = (31u << OPCODE_SHIFT | 151u << 1), 343 STWU_OPCODE = (37u << OPCODE_SHIFT), 344 STWUX_OPCODE = (31u << OPCODE_SHIFT | 183u << 1), 345 346 STH_OPCODE = (44u << OPCODE_SHIFT), 347 STHX_OPCODE = (31u << OPCODE_SHIFT | 407u << 1), 348 STHU_OPCODE = (45u << OPCODE_SHIFT), 349 350 STB_OPCODE = (38u << OPCODE_SHIFT), 351 STBX_OPCODE = (31u << OPCODE_SHIFT | 215u << 1), 352 STBU_OPCODE = (39u << OPCODE_SHIFT), 353 354 EXTSB_OPCODE = (31u << OPCODE_SHIFT | 954u << 1), 355 EXTSH_OPCODE = (31u << OPCODE_SHIFT | 922u << 1), 356 EXTSW_OPCODE = (31u << OPCODE_SHIFT | 986u << 1), // X-FORM 357 358 // 32 bit opcode encodings 359 360 LWA_OPCODE = (58u << OPCODE_SHIFT | 2u << XO_30_31_SHIFT), // DS-FORM 361 LWAX_OPCODE = (31u << OPCODE_SHIFT | 341u << XO_21_30_SHIFT), // X-FORM 362 363 CNTLZW_OPCODE = (31u << OPCODE_SHIFT | 26u << XO_21_30_SHIFT), // X-FORM 364 365 // 64 bit opcode encodings 366 367 LD_OPCODE = (58u << OPCODE_SHIFT | 0u << XO_30_31_SHIFT), // DS-FORM 368 LDU_OPCODE = (58u << OPCODE_SHIFT | 1u << XO_30_31_SHIFT), // DS-FORM 369 LDX_OPCODE = (31u << OPCODE_SHIFT | 21u << XO_21_30_SHIFT), // X-FORM 370 371 STD_OPCODE = (62u << OPCODE_SHIFT | 0u << XO_30_31_SHIFT), // DS-FORM 372 STDU_OPCODE = (62u << OPCODE_SHIFT | 1u << XO_30_31_SHIFT), // DS-FORM 373 STDUX_OPCODE = (31u << OPCODE_SHIFT | 181u << 1), // X-FORM 374 STDX_OPCODE = (31u << OPCODE_SHIFT | 149u << XO_21_30_SHIFT), // X-FORM 375 376 RLDICR_OPCODE = (30u << OPCODE_SHIFT | 1u << XO_27_29_SHIFT), // MD-FORM 377 RLDICL_OPCODE = (30u << OPCODE_SHIFT | 0u << XO_27_29_SHIFT), // MD-FORM 378 RLDIC_OPCODE = (30u << OPCODE_SHIFT | 2u << XO_27_29_SHIFT), // MD-FORM 379 RLDIMI_OPCODE = (30u << OPCODE_SHIFT | 3u << XO_27_29_SHIFT), // MD-FORM 380 381 SRADI_OPCODE = (31u << OPCODE_SHIFT | 413u << XO_21_29_SHIFT), // XS-FORM 382 383 SLD_OPCODE = (31u << OPCODE_SHIFT | 27u << 1), // X-FORM 384 SRD_OPCODE = (31u << OPCODE_SHIFT | 539u << 1), // X-FORM 385 SRAD_OPCODE = (31u << OPCODE_SHIFT | 794u << 1), // X-FORM 386 387 MULLD_OPCODE = (31u << OPCODE_SHIFT | 233u << 1), // XO-FORM 388 MULHD_OPCODE = (31u << OPCODE_SHIFT | 73u << 1), // XO-FORM 389 MULHDU_OPCODE = (31u << OPCODE_SHIFT | 9u << 1), // XO-FORM 390 DIVD_OPCODE = (31u << OPCODE_SHIFT | 489u << 1), // XO-FORM 391 392 CNTLZD_OPCODE = (31u << OPCODE_SHIFT | 58u << XO_21_30_SHIFT), // X-FORM 393 NAND_OPCODE = (31u << OPCODE_SHIFT | 476u << XO_21_30_SHIFT), // X-FORM 394 NOR_OPCODE = (31u << OPCODE_SHIFT | 124u << XO_21_30_SHIFT), // X-FORM 395 396 397 // opcodes only used for floating arithmetic 398 FADD_OPCODE = (63u << OPCODE_SHIFT | 21u << 1), 399 FADDS_OPCODE = (59u << OPCODE_SHIFT | 21u << 1), 400 FCMPU_OPCODE = (63u << OPCODE_SHIFT | 00u << 1), 401 FDIV_OPCODE = (63u << OPCODE_SHIFT | 18u << 1), 402 FDIVS_OPCODE = (59u << OPCODE_SHIFT | 18u << 1), 403 FMR_OPCODE = (63u << OPCODE_SHIFT | 72u << 1), 404 // These are special Power6 opcodes, reused for "lfdepx" and "stfdepx" 405 // on Power7. Do not use. 406 // MFFGPR_OPCODE = (31u << OPCODE_SHIFT | 607u << 1), 407 // MFTGPR_OPCODE = (31u << OPCODE_SHIFT | 735u << 1), 408 CMPB_OPCODE = (31u << OPCODE_SHIFT | 508 << 1), 409 POPCNTB_OPCODE = (31u << OPCODE_SHIFT | 122 << 1), 410 POPCNTW_OPCODE = (31u << OPCODE_SHIFT | 378 << 1), 411 POPCNTD_OPCODE = (31u << OPCODE_SHIFT | 506 << 1), 412 FABS_OPCODE = (63u << OPCODE_SHIFT | 264u << 1), 413 FNABS_OPCODE = (63u << OPCODE_SHIFT | 136u << 1), 414 FMUL_OPCODE = (63u << OPCODE_SHIFT | 25u << 1), 415 FMULS_OPCODE = (59u << OPCODE_SHIFT | 25u << 1), 416 FNEG_OPCODE = (63u << OPCODE_SHIFT | 40u << 1), 417 FSUB_OPCODE = (63u << OPCODE_SHIFT | 20u << 1), 418 FSUBS_OPCODE = (59u << OPCODE_SHIFT | 20u << 1), 419 420 // PPC64-internal FPU conversion opcodes 421 FCFID_OPCODE = (63u << OPCODE_SHIFT | 846u << 1), 422 FCFIDS_OPCODE = (59u << OPCODE_SHIFT | 846u << 1), 423 FCTID_OPCODE = (63u << OPCODE_SHIFT | 814u << 1), 424 FCTIDZ_OPCODE = (63u << OPCODE_SHIFT | 815u << 1), 425 FCTIW_OPCODE = (63u << OPCODE_SHIFT | 14u << 1), 426 FCTIWZ_OPCODE = (63u << OPCODE_SHIFT | 15u << 1), 427 FRSP_OPCODE = (63u << OPCODE_SHIFT | 12u << 1), 428 429 // WARNING: using fmadd results in a non-compliant vm. Some floating 430 // point tck tests will fail. 431 FMADD_OPCODE = (59u << OPCODE_SHIFT | 29u << 1), 432 DMADD_OPCODE = (63u << OPCODE_SHIFT | 29u << 1), 433 FMSUB_OPCODE = (59u << OPCODE_SHIFT | 28u << 1), 434 DMSUB_OPCODE = (63u << OPCODE_SHIFT | 28u << 1), 435 FNMADD_OPCODE = (59u << OPCODE_SHIFT | 31u << 1), 436 DNMADD_OPCODE = (63u << OPCODE_SHIFT | 31u << 1), 437 FNMSUB_OPCODE = (59u << OPCODE_SHIFT | 30u << 1), 438 DNMSUB_OPCODE = (63u << OPCODE_SHIFT | 30u << 1), 439 440 LFD_OPCODE = (50u << OPCODE_SHIFT | 00u << 1), 441 LFDU_OPCODE = (51u << OPCODE_SHIFT | 00u << 1), 442 LFDX_OPCODE = (31u << OPCODE_SHIFT | 599u << 1), 443 LFS_OPCODE = (48u << OPCODE_SHIFT | 00u << 1), 444 LFSU_OPCODE = (49u << OPCODE_SHIFT | 00u << 1), 445 LFSX_OPCODE = (31u << OPCODE_SHIFT | 535u << 1), 446 447 STFD_OPCODE = (54u << OPCODE_SHIFT | 00u << 1), 448 STFDU_OPCODE = (55u << OPCODE_SHIFT | 00u << 1), 449 STFDX_OPCODE = (31u << OPCODE_SHIFT | 727u << 1), 450 STFS_OPCODE = (52u << OPCODE_SHIFT | 00u << 1), 451 STFSU_OPCODE = (53u << OPCODE_SHIFT | 00u << 1), 452 STFSX_OPCODE = (31u << OPCODE_SHIFT | 663u << 1), 453 454 FSQRT_OPCODE = (63u << OPCODE_SHIFT | 22u << 1), // A-FORM 455 FSQRTS_OPCODE = (59u << OPCODE_SHIFT | 22u << 1), // A-FORM 456 457 // Vector instruction support for >= Power6 458 // Vector Storage Access 459 LVEBX_OPCODE = (31u << OPCODE_SHIFT | 7u << 1), 460 LVEHX_OPCODE = (31u << OPCODE_SHIFT | 39u << 1), 461 LVEWX_OPCODE = (31u << OPCODE_SHIFT | 71u << 1), 462 LVX_OPCODE = (31u << OPCODE_SHIFT | 103u << 1), 463 LVXL_OPCODE = (31u << OPCODE_SHIFT | 359u << 1), 464 STVEBX_OPCODE = (31u << OPCODE_SHIFT | 135u << 1), 465 STVEHX_OPCODE = (31u << OPCODE_SHIFT | 167u << 1), 466 STVEWX_OPCODE = (31u << OPCODE_SHIFT | 199u << 1), 467 STVX_OPCODE = (31u << OPCODE_SHIFT | 231u << 1), 468 STVXL_OPCODE = (31u << OPCODE_SHIFT | 487u << 1), 469 LVSL_OPCODE = (31u << OPCODE_SHIFT | 6u << 1), 470 LVSR_OPCODE = (31u << OPCODE_SHIFT | 38u << 1), 471 472 // Vector Permute and Formatting 473 VPKPX_OPCODE = (4u << OPCODE_SHIFT | 782u ), 474 VPKSHSS_OPCODE = (4u << OPCODE_SHIFT | 398u ), 475 VPKSWSS_OPCODE = (4u << OPCODE_SHIFT | 462u ), 476 VPKSHUS_OPCODE = (4u << OPCODE_SHIFT | 270u ), 477 VPKSWUS_OPCODE = (4u << OPCODE_SHIFT | 334u ), 478 VPKUHUM_OPCODE = (4u << OPCODE_SHIFT | 14u ), 479 VPKUWUM_OPCODE = (4u << OPCODE_SHIFT | 78u ), 480 VPKUHUS_OPCODE = (4u << OPCODE_SHIFT | 142u ), 481 VPKUWUS_OPCODE = (4u << OPCODE_SHIFT | 206u ), 482 VUPKHPX_OPCODE = (4u << OPCODE_SHIFT | 846u ), 483 VUPKHSB_OPCODE = (4u << OPCODE_SHIFT | 526u ), 484 VUPKHSH_OPCODE = (4u << OPCODE_SHIFT | 590u ), 485 VUPKLPX_OPCODE = (4u << OPCODE_SHIFT | 974u ), 486 VUPKLSB_OPCODE = (4u << OPCODE_SHIFT | 654u ), 487 VUPKLSH_OPCODE = (4u << OPCODE_SHIFT | 718u ), 488 489 VMRGHB_OPCODE = (4u << OPCODE_SHIFT | 12u ), 490 VMRGHW_OPCODE = (4u << OPCODE_SHIFT | 140u ), 491 VMRGHH_OPCODE = (4u << OPCODE_SHIFT | 76u ), 492 VMRGLB_OPCODE = (4u << OPCODE_SHIFT | 268u ), 493 VMRGLW_OPCODE = (4u << OPCODE_SHIFT | 396u ), 494 VMRGLH_OPCODE = (4u << OPCODE_SHIFT | 332u ), 495 496 VSPLT_OPCODE = (4u << OPCODE_SHIFT | 524u ), 497 VSPLTH_OPCODE = (4u << OPCODE_SHIFT | 588u ), 498 VSPLTW_OPCODE = (4u << OPCODE_SHIFT | 652u ), 499 VSPLTISB_OPCODE= (4u << OPCODE_SHIFT | 780u ), 500 VSPLTISH_OPCODE= (4u << OPCODE_SHIFT | 844u ), 501 VSPLTISW_OPCODE= (4u << OPCODE_SHIFT | 908u ), 502 503 VPERM_OPCODE = (4u << OPCODE_SHIFT | 43u ), 504 VSEL_OPCODE = (4u << OPCODE_SHIFT | 42u ), 505 506 VSL_OPCODE = (4u << OPCODE_SHIFT | 452u ), 507 VSLDOI_OPCODE = (4u << OPCODE_SHIFT | 44u ), 508 VSLO_OPCODE = (4u << OPCODE_SHIFT | 1036u ), 509 VSR_OPCODE = (4u << OPCODE_SHIFT | 708u ), 510 VSRO_OPCODE = (4u << OPCODE_SHIFT | 1100u ), 511 512 // Vector Integer 513 VADDCUW_OPCODE = (4u << OPCODE_SHIFT | 384u ), 514 VADDSHS_OPCODE = (4u << OPCODE_SHIFT | 832u ), 515 VADDSBS_OPCODE = (4u << OPCODE_SHIFT | 768u ), 516 VADDSWS_OPCODE = (4u << OPCODE_SHIFT | 896u ), 517 VADDUBM_OPCODE = (4u << OPCODE_SHIFT | 0u ), 518 VADDUWM_OPCODE = (4u << OPCODE_SHIFT | 128u ), 519 VADDUHM_OPCODE = (4u << OPCODE_SHIFT | 64u ), 520 VADDUBS_OPCODE = (4u << OPCODE_SHIFT | 512u ), 521 VADDUWS_OPCODE = (4u << OPCODE_SHIFT | 640u ), 522 VADDUHS_OPCODE = (4u << OPCODE_SHIFT | 576u ), 523 VSUBCUW_OPCODE = (4u << OPCODE_SHIFT | 1408u ), 524 VSUBSHS_OPCODE = (4u << OPCODE_SHIFT | 1856u ), 525 VSUBSBS_OPCODE = (4u << OPCODE_SHIFT | 1792u ), 526 VSUBSWS_OPCODE = (4u << OPCODE_SHIFT | 1920u ), 527 VSUBUBM_OPCODE = (4u << OPCODE_SHIFT | 1024u ), 528 VSUBUWM_OPCODE = (4u << OPCODE_SHIFT | 1152u ), 529 VSUBUHM_OPCODE = (4u << OPCODE_SHIFT | 1088u ), 530 VSUBUBS_OPCODE = (4u << OPCODE_SHIFT | 1536u ), 531 VSUBUWS_OPCODE = (4u << OPCODE_SHIFT | 1664u ), 532 VSUBUHS_OPCODE = (4u << OPCODE_SHIFT | 1600u ), 533 534 VMULESB_OPCODE = (4u << OPCODE_SHIFT | 776u ), 535 VMULEUB_OPCODE = (4u << OPCODE_SHIFT | 520u ), 536 VMULESH_OPCODE = (4u << OPCODE_SHIFT | 840u ), 537 VMULEUH_OPCODE = (4u << OPCODE_SHIFT | 584u ), 538 VMULOSB_OPCODE = (4u << OPCODE_SHIFT | 264u ), 539 VMULOUB_OPCODE = (4u << OPCODE_SHIFT | 8u ), 540 VMULOSH_OPCODE = (4u << OPCODE_SHIFT | 328u ), 541 VMULOUH_OPCODE = (4u << OPCODE_SHIFT | 72u ), 542 VMHADDSHS_OPCODE=(4u << OPCODE_SHIFT | 32u ), 543 VMHRADDSHS_OPCODE=(4u << OPCODE_SHIFT | 33u ), 544 VMLADDUHM_OPCODE=(4u << OPCODE_SHIFT | 34u ), 545 VMSUBUHM_OPCODE= (4u << OPCODE_SHIFT | 36u ), 546 VMSUMMBM_OPCODE= (4u << OPCODE_SHIFT | 37u ), 547 VMSUMSHM_OPCODE= (4u << OPCODE_SHIFT | 40u ), 548 VMSUMSHS_OPCODE= (4u << OPCODE_SHIFT | 41u ), 549 VMSUMUHM_OPCODE= (4u << OPCODE_SHIFT | 38u ), 550 VMSUMUHS_OPCODE= (4u << OPCODE_SHIFT | 39u ), 551 552 VSUMSWS_OPCODE = (4u << OPCODE_SHIFT | 1928u ), 553 VSUM2SWS_OPCODE= (4u << OPCODE_SHIFT | 1672u ), 554 VSUM4SBS_OPCODE= (4u << OPCODE_SHIFT | 1800u ), 555 VSUM4UBS_OPCODE= (4u << OPCODE_SHIFT | 1544u ), 556 VSUM4SHS_OPCODE= (4u << OPCODE_SHIFT | 1608u ), 557 558 VAVGSB_OPCODE = (4u << OPCODE_SHIFT | 1282u ), 559 VAVGSW_OPCODE = (4u << OPCODE_SHIFT | 1410u ), 560 VAVGSH_OPCODE = (4u << OPCODE_SHIFT | 1346u ), 561 VAVGUB_OPCODE = (4u << OPCODE_SHIFT | 1026u ), 562 VAVGUW_OPCODE = (4u << OPCODE_SHIFT | 1154u ), 563 VAVGUH_OPCODE = (4u << OPCODE_SHIFT | 1090u ), 564 565 VMAXSB_OPCODE = (4u << OPCODE_SHIFT | 258u ), 566 VMAXSW_OPCODE = (4u << OPCODE_SHIFT | 386u ), 567 VMAXSH_OPCODE = (4u << OPCODE_SHIFT | 322u ), 568 VMAXUB_OPCODE = (4u << OPCODE_SHIFT | 2u ), 569 VMAXUW_OPCODE = (4u << OPCODE_SHIFT | 130u ), 570 VMAXUH_OPCODE = (4u << OPCODE_SHIFT | 66u ), 571 VMINSB_OPCODE = (4u << OPCODE_SHIFT | 770u ), 572 VMINSW_OPCODE = (4u << OPCODE_SHIFT | 898u ), 573 VMINSH_OPCODE = (4u << OPCODE_SHIFT | 834u ), 574 VMINUB_OPCODE = (4u << OPCODE_SHIFT | 514u ), 575 VMINUW_OPCODE = (4u << OPCODE_SHIFT | 642u ), 576 VMINUH_OPCODE = (4u << OPCODE_SHIFT | 578u ), 577 578 VCMPEQUB_OPCODE= (4u << OPCODE_SHIFT | 6u ), 579 VCMPEQUH_OPCODE= (4u << OPCODE_SHIFT | 70u ), 580 VCMPEQUW_OPCODE= (4u << OPCODE_SHIFT | 134u ), 581 VCMPGTSH_OPCODE= (4u << OPCODE_SHIFT | 838u ), 582 VCMPGTSB_OPCODE= (4u << OPCODE_SHIFT | 774u ), 583 VCMPGTSW_OPCODE= (4u << OPCODE_SHIFT | 902u ), 584 VCMPGTUB_OPCODE= (4u << OPCODE_SHIFT | 518u ), 585 VCMPGTUH_OPCODE= (4u << OPCODE_SHIFT | 582u ), 586 VCMPGTUW_OPCODE= (4u << OPCODE_SHIFT | 646u ), 587 588 VAND_OPCODE = (4u << OPCODE_SHIFT | 1028u ), 589 VANDC_OPCODE = (4u << OPCODE_SHIFT | 1092u ), 590 VNOR_OPCODE = (4u << OPCODE_SHIFT | 1284u ), 591 VOR_OPCODE = (4u << OPCODE_SHIFT | 1156u ), 592 VXOR_OPCODE = (4u << OPCODE_SHIFT | 1220u ), 593 VRLB_OPCODE = (4u << OPCODE_SHIFT | 4u ), 594 VRLW_OPCODE = (4u << OPCODE_SHIFT | 132u ), 595 VRLH_OPCODE = (4u << OPCODE_SHIFT | 68u ), 596 VSLB_OPCODE = (4u << OPCODE_SHIFT | 260u ), 597 VSKW_OPCODE = (4u << OPCODE_SHIFT | 388u ), 598 VSLH_OPCODE = (4u << OPCODE_SHIFT | 324u ), 599 VSRB_OPCODE = (4u << OPCODE_SHIFT | 516u ), 600 VSRW_OPCODE = (4u << OPCODE_SHIFT | 644u ), 601 VSRH_OPCODE = (4u << OPCODE_SHIFT | 580u ), 602 VSRAB_OPCODE = (4u << OPCODE_SHIFT | 772u ), 603 VSRAW_OPCODE = (4u << OPCODE_SHIFT | 900u ), 604 VSRAH_OPCODE = (4u << OPCODE_SHIFT | 836u ), 605 606 // Vector Floating-Point 607 // not implemented yet 608 609 // Vector Status and Control 610 MTVSCR_OPCODE = (4u << OPCODE_SHIFT | 1604u ), 611 MFVSCR_OPCODE = (4u << OPCODE_SHIFT | 1540u ), 612 613 // AES (introduced with Power 8) 614 VCIPHER_OPCODE = (4u << OPCODE_SHIFT | 1288u), 615 VCIPHERLAST_OPCODE = (4u << OPCODE_SHIFT | 1289u), 616 VNCIPHER_OPCODE = (4u << OPCODE_SHIFT | 1352u), 617 VNCIPHERLAST_OPCODE = (4u << OPCODE_SHIFT | 1353u), 618 VSBOX_OPCODE = (4u << OPCODE_SHIFT | 1480u), 619 620 // SHA (introduced with Power 8) 621 VSHASIGMAD_OPCODE = (4u << OPCODE_SHIFT | 1730u), 622 VSHASIGMAW_OPCODE = (4u << OPCODE_SHIFT | 1666u), 623 624 // Vector Binary Polynomial Multiplication (introduced with Power 8) 625 VPMSUMB_OPCODE = (4u << OPCODE_SHIFT | 1032u), 626 VPMSUMD_OPCODE = (4u << OPCODE_SHIFT | 1224u), 627 VPMSUMH_OPCODE = (4u << OPCODE_SHIFT | 1096u), 628 VPMSUMW_OPCODE = (4u << OPCODE_SHIFT | 1160u), 629 630 // Vector Permute and Xor (introduced with Power 8) 631 VPERMXOR_OPCODE = (4u << OPCODE_SHIFT | 45u), 632 633 // Transactional Memory instructions (introduced with Power 8) 634 TBEGIN_OPCODE = (31u << OPCODE_SHIFT | 654u << 1), 635 TEND_OPCODE = (31u << OPCODE_SHIFT | 686u << 1), 636 TABORT_OPCODE = (31u << OPCODE_SHIFT | 910u << 1), 637 TABORTWC_OPCODE = (31u << OPCODE_SHIFT | 782u << 1), 638 TABORTWCI_OPCODE = (31u << OPCODE_SHIFT | 846u << 1), 639 TABORTDC_OPCODE = (31u << OPCODE_SHIFT | 814u << 1), 640 TABORTDCI_OPCODE = (31u << OPCODE_SHIFT | 878u << 1), 641 TSR_OPCODE = (31u << OPCODE_SHIFT | 750u << 1), 642 TCHECK_OPCODE = (31u << OPCODE_SHIFT | 718u << 1), 643 644 // Icache and dcache related instructions 645 DCBA_OPCODE = (31u << OPCODE_SHIFT | 758u << 1), 646 DCBZ_OPCODE = (31u << OPCODE_SHIFT | 1014u << 1), 647 DCBST_OPCODE = (31u << OPCODE_SHIFT | 54u << 1), 648 DCBF_OPCODE = (31u << OPCODE_SHIFT | 86u << 1), 649 650 DCBT_OPCODE = (31u << OPCODE_SHIFT | 278u << 1), 651 DCBTST_OPCODE = (31u << OPCODE_SHIFT | 246u << 1), 652 ICBI_OPCODE = (31u << OPCODE_SHIFT | 982u << 1), 653 654 // Instruction synchronization 655 ISYNC_OPCODE = (19u << OPCODE_SHIFT | 150u << 1), 656 // Memory barriers 657 SYNC_OPCODE = (31u << OPCODE_SHIFT | 598u << 1), 658 EIEIO_OPCODE = (31u << OPCODE_SHIFT | 854u << 1), 659 660 // Trap instructions 661 TDI_OPCODE = (2u << OPCODE_SHIFT), 662 TWI_OPCODE = (3u << OPCODE_SHIFT), 663 TD_OPCODE = (31u << OPCODE_SHIFT | 68u << 1), 664 TW_OPCODE = (31u << OPCODE_SHIFT | 4u << 1), 665 666 // Atomics. 667 LWARX_OPCODE = (31u << OPCODE_SHIFT | 20u << 1), 668 LDARX_OPCODE = (31u << OPCODE_SHIFT | 84u << 1), 669 STWCX_OPCODE = (31u << OPCODE_SHIFT | 150u << 1), 670 STDCX_OPCODE = (31u << OPCODE_SHIFT | 214u << 1) 671 672 }; 673 674 // Trap instructions TO bits 675 enum trap_to_bits { 676 // single bits 677 traptoLessThanSigned = 1 << 4, // 0, left end 678 traptoGreaterThanSigned = 1 << 3, 679 traptoEqual = 1 << 2, 680 traptoLessThanUnsigned = 1 << 1, 681 traptoGreaterThanUnsigned = 1 << 0, // 4, right end 682 683 // compound ones 684 traptoUnconditional = (traptoLessThanSigned | 685 traptoGreaterThanSigned | 686 traptoEqual | 687 traptoLessThanUnsigned | 688 traptoGreaterThanUnsigned) 689 }; 690 691 // Branch hints BH field 692 enum branch_hint_bh { 693 // bclr cases: 694 bhintbhBCLRisReturn = 0, 695 bhintbhBCLRisNotReturnButSame = 1, 696 bhintbhBCLRisNotPredictable = 3, 697 698 // bcctr cases: 699 bhintbhBCCTRisNotReturnButSame = 0, 700 bhintbhBCCTRisNotPredictable = 3 701 }; 702 703 // Branch prediction hints AT field 704 enum branch_hint_at { 705 bhintatNoHint = 0, // at=00 706 bhintatIsNotTaken = 2, // at=10 707 bhintatIsTaken = 3 // at=11 708 }; 709 710 // Branch prediction hints 711 enum branch_hint_concept { 712 // Use the same encoding as branch_hint_at to simply code. 713 bhintNoHint = bhintatNoHint, 714 bhintIsNotTaken = bhintatIsNotTaken, 715 bhintIsTaken = bhintatIsTaken 716 }; 717 718 // Used in BO field of branch instruction. 719 enum branch_condition { 720 bcondCRbiIs0 = 4, // bo=001at 721 bcondCRbiIs1 = 12, // bo=011at 722 bcondAlways = 20 // bo=10100 723 }; 724 725 // Branch condition with combined prediction hints. 726 enum branch_condition_with_hint { 727 bcondCRbiIs0_bhintNoHint = bcondCRbiIs0 | bhintatNoHint, 728 bcondCRbiIs0_bhintIsNotTaken = bcondCRbiIs0 | bhintatIsNotTaken, 729 bcondCRbiIs0_bhintIsTaken = bcondCRbiIs0 | bhintatIsTaken, 730 bcondCRbiIs1_bhintNoHint = bcondCRbiIs1 | bhintatNoHint, 731 bcondCRbiIs1_bhintIsNotTaken = bcondCRbiIs1 | bhintatIsNotTaken, 732 bcondCRbiIs1_bhintIsTaken = bcondCRbiIs1 | bhintatIsTaken, 733 }; 734 735 // Elemental Memory Barriers (>=Power 8) 736 enum Elemental_Membar_mask_bits { 737 StoreStore = 1 << 0, 738 StoreLoad = 1 << 1, 739 LoadStore = 1 << 2, 740 LoadLoad = 1 << 3 741 }; 742 743 // Branch prediction hints. 744 inline static int add_bhint_to_boint(const int bhint, const int boint) { 745 switch (boint) { 746 case bcondCRbiIs0: 747 case bcondCRbiIs1: 748 // branch_hint and branch_hint_at have same encodings 749 assert( (int)bhintNoHint == (int)bhintatNoHint 750 && (int)bhintIsNotTaken == (int)bhintatIsNotTaken 751 && (int)bhintIsTaken == (int)bhintatIsTaken, 752 "wrong encodings"); 753 assert((bhint & 0x03) == bhint, "wrong encodings"); 754 return (boint & ~0x03) | bhint; 755 case bcondAlways: 756 // no branch_hint 757 return boint; 758 default: 759 ShouldNotReachHere(); 760 return 0; 761 } 762 } 763 764 // Extract bcond from boint. 765 inline static int inv_boint_bcond(const int boint) { 766 int r_bcond = boint & ~0x03; 767 assert(r_bcond == bcondCRbiIs0 || 768 r_bcond == bcondCRbiIs1 || 769 r_bcond == bcondAlways, 770 "bad branch condition"); 771 return r_bcond; 772 } 773 774 // Extract bhint from boint. 775 inline static int inv_boint_bhint(const int boint) { 776 int r_bhint = boint & 0x03; 777 assert(r_bhint == bhintatNoHint || 778 r_bhint == bhintatIsNotTaken || 779 r_bhint == bhintatIsTaken, 780 "bad branch hint"); 781 return r_bhint; 782 } 783 784 // Calculate opposite of given bcond. 785 inline static int opposite_bcond(const int bcond) { 786 switch (bcond) { 787 case bcondCRbiIs0: 788 return bcondCRbiIs1; 789 case bcondCRbiIs1: 790 return bcondCRbiIs0; 791 default: 792 ShouldNotReachHere(); 793 return 0; 794 } 795 } 796 797 // Calculate opposite of given bhint. 798 inline static int opposite_bhint(const int bhint) { 799 switch (bhint) { 800 case bhintatNoHint: 801 return bhintatNoHint; 802 case bhintatIsNotTaken: 803 return bhintatIsTaken; 804 case bhintatIsTaken: 805 return bhintatIsNotTaken; 806 default: 807 ShouldNotReachHere(); 808 return 0; 809 } 810 } 811 812 // PPC branch instructions 813 enum ppcops { 814 b_op = 18, 815 bc_op = 16, 816 bcr_op = 19 817 }; 818 819 enum Condition { 820 negative = 0, 821 less = 0, 822 positive = 1, 823 greater = 1, 824 zero = 2, 825 equal = 2, 826 summary_overflow = 3, 827 }; 828 829 public: 830 // Helper functions for groups of instructions 831 832 enum Predict { pt = 1, pn = 0 }; // pt = predict taken 833 834 // instruction must start at passed address 835 static int instr_len(unsigned char *instr) { return BytesPerInstWord; } 836 837 // instruction must be left-justified in argument 838 static int instr_len(unsigned long instr) { return BytesPerInstWord; } 839 840 // longest instructions 841 static int instr_maxlen() { return BytesPerInstWord; } 842 843 // Test if x is within signed immediate range for nbits. 844 static bool is_simm(int x, unsigned int nbits) { 845 assert(0 < nbits && nbits < 32, "out of bounds"); 846 const int min = -( ((int)1) << nbits-1 ); 847 const int maxplus1 = ( ((int)1) << nbits-1 ); 848 return min <= x && x < maxplus1; 849 } 850 851 static bool is_simm(jlong x, unsigned int nbits) { 852 assert(0 < nbits && nbits < 64, "out of bounds"); 853 const jlong min = -( ((jlong)1) << nbits-1 ); 854 const jlong maxplus1 = ( ((jlong)1) << nbits-1 ); 855 return min <= x && x < maxplus1; 856 } 857 858 // Test if x is within unsigned immediate range for nbits 859 static bool is_uimm(int x, unsigned int nbits) { 860 assert(0 < nbits && nbits < 32, "out of bounds"); 861 const int maxplus1 = ( ((int)1) << nbits ); 862 return 0 <= x && x < maxplus1; 863 } 864 865 static bool is_uimm(jlong x, unsigned int nbits) { 866 assert(0 < nbits && nbits < 64, "out of bounds"); 867 const jlong maxplus1 = ( ((jlong)1) << nbits ); 868 return 0 <= x && x < maxplus1; 869 } 870 871 protected: 872 // helpers 873 874 // X is supposed to fit in a field "nbits" wide 875 // and be sign-extended. Check the range. 876 static void assert_signed_range(intptr_t x, int nbits) { 877 assert(nbits == 32 || (-(1 << nbits-1) <= x && x < (1 << nbits-1)), 878 "value out of range"); 879 } 880 881 static void assert_signed_word_disp_range(intptr_t x, int nbits) { 882 assert((x & 3) == 0, "not word aligned"); 883 assert_signed_range(x, nbits + 2); 884 } 885 886 static void assert_unsigned_const(int x, int nbits) { 887 assert(juint(x) < juint(1 << nbits), "unsigned constant out of range"); 888 } 889 890 static int fmask(juint hi_bit, juint lo_bit) { 891 assert(hi_bit >= lo_bit && hi_bit < 32, "bad bits"); 892 return (1 << ( hi_bit-lo_bit + 1 )) - 1; 893 } 894 895 // inverse of u_field 896 static int inv_u_field(int x, int hi_bit, int lo_bit) { 897 juint r = juint(x) >> lo_bit; 898 r &= fmask(hi_bit, lo_bit); 899 return int(r); 900 } 901 902 // signed version: extract from field and sign-extend 903 static int inv_s_field_ppc(int x, int hi_bit, int lo_bit) { 904 x = x << (31-hi_bit); 905 x = x >> (31-hi_bit+lo_bit); 906 return x; 907 } 908 909 static int u_field(int x, int hi_bit, int lo_bit) { 910 assert((x & ~fmask(hi_bit, lo_bit)) == 0, "value out of range"); 911 int r = x << lo_bit; 912 assert(inv_u_field(r, hi_bit, lo_bit) == x, "just checking"); 913 return r; 914 } 915 916 // Same as u_field for signed values 917 static int s_field(int x, int hi_bit, int lo_bit) { 918 int nbits = hi_bit - lo_bit + 1; 919 assert(nbits == 32 || (-(1 << nbits-1) <= x && x < (1 << nbits-1)), 920 "value out of range"); 921 x &= fmask(hi_bit, lo_bit); 922 int r = x << lo_bit; 923 return r; 924 } 925 926 // inv_op for ppc instructions 927 static int inv_op_ppc(int x) { return inv_u_field(x, 31, 26); } 928 929 // Determine target address from li, bd field of branch instruction. 930 static intptr_t inv_li_field(int x) { 931 intptr_t r = inv_s_field_ppc(x, 25, 2); 932 r = (r << 2); 933 return r; 934 } 935 static intptr_t inv_bd_field(int x, intptr_t pos) { 936 intptr_t r = inv_s_field_ppc(x, 15, 2); 937 r = (r << 2) + pos; 938 return r; 939 } 940 941 #define inv_opp_u_field(x, hi_bit, lo_bit) inv_u_field(x, 31-(lo_bit), 31-(hi_bit)) 942 #define inv_opp_s_field(x, hi_bit, lo_bit) inv_s_field_ppc(x, 31-(lo_bit), 31-(hi_bit)) 943 // Extract instruction fields from instruction words. 944 public: 945 static int inv_ra_field(int x) { return inv_opp_u_field(x, 15, 11); } 946 static int inv_rb_field(int x) { return inv_opp_u_field(x, 20, 16); } 947 static int inv_rt_field(int x) { return inv_opp_u_field(x, 10, 6); } 948 static int inv_rta_field(int x) { return inv_opp_u_field(x, 15, 11); } 949 static int inv_rs_field(int x) { return inv_opp_u_field(x, 10, 6); } 950 // Ds uses opp_s_field(x, 31, 16), but lowest 2 bits must be 0. 951 // Inv_ds_field uses range (x, 29, 16) but shifts by 2 to ensure that lowest bits are 0. 952 static int inv_ds_field(int x) { return inv_opp_s_field(x, 29, 16) << 2; } 953 static int inv_d1_field(int x) { return inv_opp_s_field(x, 31, 16); } 954 static int inv_si_field(int x) { return inv_opp_s_field(x, 31, 16); } 955 static int inv_to_field(int x) { return inv_opp_u_field(x, 10, 6); } 956 static int inv_lk_field(int x) { return inv_opp_u_field(x, 31, 31); } 957 static int inv_bo_field(int x) { return inv_opp_u_field(x, 10, 6); } 958 static int inv_bi_field(int x) { return inv_opp_u_field(x, 15, 11); } 959 960 #define opp_u_field(x, hi_bit, lo_bit) u_field(x, 31-(lo_bit), 31-(hi_bit)) 961 #define opp_s_field(x, hi_bit, lo_bit) s_field(x, 31-(lo_bit), 31-(hi_bit)) 962 963 // instruction fields 964 static int aa( int x) { return opp_u_field(x, 30, 30); } 965 static int ba( int x) { return opp_u_field(x, 15, 11); } 966 static int bb( int x) { return opp_u_field(x, 20, 16); } 967 static int bc( int x) { return opp_u_field(x, 25, 21); } 968 static int bd( int x) { return opp_s_field(x, 29, 16); } 969 static int bf( ConditionRegister cr) { return bf(cr->encoding()); } 970 static int bf( int x) { return opp_u_field(x, 8, 6); } 971 static int bfa(ConditionRegister cr) { return bfa(cr->encoding()); } 972 static int bfa( int x) { return opp_u_field(x, 13, 11); } 973 static int bh( int x) { return opp_u_field(x, 20, 19); } 974 static int bi( int x) { return opp_u_field(x, 15, 11); } 975 static int bi0(ConditionRegister cr, Condition c) { return (cr->encoding() << 2) | c; } 976 static int bo( int x) { return opp_u_field(x, 10, 6); } 977 static int bt( int x) { return opp_u_field(x, 10, 6); } 978 static int d1( int x) { return opp_s_field(x, 31, 16); } 979 static int ds( int x) { assert((x & 0x3) == 0, "unaligned offset"); return opp_s_field(x, 31, 16); } 980 static int eh( int x) { return opp_u_field(x, 31, 31); } 981 static int flm( int x) { return opp_u_field(x, 14, 7); } 982 static int fra( FloatRegister r) { return fra(r->encoding());} 983 static int frb( FloatRegister r) { return frb(r->encoding());} 984 static int frc( FloatRegister r) { return frc(r->encoding());} 985 static int frs( FloatRegister r) { return frs(r->encoding());} 986 static int frt( FloatRegister r) { return frt(r->encoding());} 987 static int fra( int x) { return opp_u_field(x, 15, 11); } 988 static int frb( int x) { return opp_u_field(x, 20, 16); } 989 static int frc( int x) { return opp_u_field(x, 25, 21); } 990 static int frs( int x) { return opp_u_field(x, 10, 6); } 991 static int frt( int x) { return opp_u_field(x, 10, 6); } 992 static int fxm( int x) { return opp_u_field(x, 19, 12); } 993 static int l10( int x) { return opp_u_field(x, 10, 10); } 994 static int l15( int x) { return opp_u_field(x, 15, 15); } 995 static int l910( int x) { return opp_u_field(x, 10, 9); } 996 static int e1215( int x) { return opp_u_field(x, 15, 12); } 997 static int lev( int x) { return opp_u_field(x, 26, 20); } 998 static int li( int x) { return opp_s_field(x, 29, 6); } 999 static int lk( int x) { return opp_u_field(x, 31, 31); } 1000 static int mb2125( int x) { return opp_u_field(x, 25, 21); } 1001 static int me2630( int x) { return opp_u_field(x, 30, 26); } 1002 static int mb2126( int x) { return opp_u_field(((x & 0x1f) << 1) | ((x & 0x20) >> 5), 26, 21); } 1003 static int me2126( int x) { return mb2126(x); } 1004 static int nb( int x) { return opp_u_field(x, 20, 16); } 1005 //static int opcd( int x) { return opp_u_field(x, 5, 0); } // is contained in our opcodes 1006 static int oe( int x) { return opp_u_field(x, 21, 21); } 1007 static int ra( Register r) { return ra(r->encoding()); } 1008 static int ra( int x) { return opp_u_field(x, 15, 11); } 1009 static int rb( Register r) { return rb(r->encoding()); } 1010 static int rb( int x) { return opp_u_field(x, 20, 16); } 1011 static int rc( int x) { return opp_u_field(x, 31, 31); } 1012 static int rs( Register r) { return rs(r->encoding()); } 1013 static int rs( int x) { return opp_u_field(x, 10, 6); } 1014 // we don't want to use R0 in memory accesses, because it has value `0' then 1015 static int ra0mem( Register r) { assert(r != R0, "cannot use register R0 in memory access"); return ra(r); } 1016 static int ra0mem( int x) { assert(x != 0, "cannot use register 0 in memory access"); return ra(x); } 1017 1018 // register r is target 1019 static int rt( Register r) { return rs(r); } 1020 static int rt( int x) { return rs(x); } 1021 static int rta( Register r) { return ra(r); } 1022 static int rta0mem( Register r) { rta(r); return ra0mem(r); } 1023 1024 static int sh1620( int x) { return opp_u_field(x, 20, 16); } 1025 static int sh30( int x) { return opp_u_field(x, 30, 30); } 1026 static int sh162030( int x) { return sh1620(x & 0x1f) | sh30((x & 0x20) >> 5); } 1027 static int si( int x) { return opp_s_field(x, 31, 16); } 1028 static int spr( int x) { return opp_u_field(x, 20, 11); } 1029 static int sr( int x) { return opp_u_field(x, 15, 12); } 1030 static int tbr( int x) { return opp_u_field(x, 20, 11); } 1031 static int th( int x) { return opp_u_field(x, 10, 7); } 1032 static int thct( int x) { assert((x&8) == 0, "must be valid cache specification"); return th(x); } 1033 static int thds( int x) { assert((x&8) == 8, "must be valid stream specification"); return th(x); } 1034 static int to( int x) { return opp_u_field(x, 10, 6); } 1035 static int u( int x) { return opp_u_field(x, 19, 16); } 1036 static int ui( int x) { return opp_u_field(x, 31, 16); } 1037 1038 // Support vector instructions for >= Power6. 1039 static int vra( int x) { return opp_u_field(x, 15, 11); } 1040 static int vrb( int x) { return opp_u_field(x, 20, 16); } 1041 static int vrc( int x) { return opp_u_field(x, 25, 21); } 1042 static int vrs( int x) { return opp_u_field(x, 10, 6); } 1043 static int vrt( int x) { return opp_u_field(x, 10, 6); } 1044 1045 static int vra( VectorRegister r) { return vra(r->encoding());} 1046 static int vrb( VectorRegister r) { return vrb(r->encoding());} 1047 static int vrc( VectorRegister r) { return vrc(r->encoding());} 1048 static int vrs( VectorRegister r) { return vrs(r->encoding());} 1049 static int vrt( VectorRegister r) { return vrt(r->encoding());} 1050 1051 static int vsplt_uim( int x) { return opp_u_field(x, 15, 12); } // for vsplt* instructions 1052 static int vsplti_sim(int x) { return opp_u_field(x, 15, 11); } // for vsplti* instructions 1053 static int vsldoi_shb(int x) { return opp_u_field(x, 25, 22); } // for vsldoi instruction 1054 static int vcmp_rc( int x) { return opp_u_field(x, 21, 21); } // for vcmp* instructions 1055 1056 //static int xo1( int x) { return opp_u_field(x, 29, 21); }// is contained in our opcodes 1057 //static int xo2( int x) { return opp_u_field(x, 30, 21); }// is contained in our opcodes 1058 //static int xo3( int x) { return opp_u_field(x, 30, 22); }// is contained in our opcodes 1059 //static int xo4( int x) { return opp_u_field(x, 30, 26); }// is contained in our opcodes 1060 //static int xo5( int x) { return opp_u_field(x, 29, 27); }// is contained in our opcodes 1061 //static int xo6( int x) { return opp_u_field(x, 30, 27); }// is contained in our opcodes 1062 //static int xo7( int x) { return opp_u_field(x, 31, 30); }// is contained in our opcodes 1063 1064 protected: 1065 // Compute relative address for branch. 1066 static intptr_t disp(intptr_t x, intptr_t off) { 1067 int xx = x - off; 1068 xx = xx >> 2; 1069 return xx; 1070 } 1071 1072 public: 1073 // signed immediate, in low bits, nbits long 1074 static int simm(int x, int nbits) { 1075 assert_signed_range(x, nbits); 1076 return x & ((1 << nbits) - 1); 1077 } 1078 1079 // unsigned immediate, in low bits, nbits long 1080 static int uimm(int x, int nbits) { 1081 assert_unsigned_const(x, nbits); 1082 return x & ((1 << nbits) - 1); 1083 } 1084 1085 static void set_imm(int* instr, short s) { 1086 // imm is always in the lower 16 bits of the instruction, 1087 // so this is endian-neutral. Same for the get_imm below. 1088 uint32_t w = *(uint32_t *)instr; 1089 *instr = (int)((w & ~0x0000FFFF) | (s & 0x0000FFFF)); 1090 } 1091 1092 static int get_imm(address a, int instruction_number) { 1093 return (short)((int *)a)[instruction_number]; 1094 } 1095 1096 static inline int hi16_signed( int x) { return (int)(int16_t)(x >> 16); } 1097 static inline int lo16_unsigned(int x) { return x & 0xffff; } 1098 1099 protected: 1100 1101 // Extract the top 32 bits in a 64 bit word. 1102 static int32_t hi32(int64_t x) { 1103 int32_t r = int32_t((uint64_t)x >> 32); 1104 return r; 1105 } 1106 1107 public: 1108 1109 static inline unsigned int align_addr(unsigned int addr, unsigned int a) { 1110 return ((addr + (a - 1)) & ~(a - 1)); 1111 } 1112 1113 static inline bool is_aligned(unsigned int addr, unsigned int a) { 1114 return (0 == addr % a); 1115 } 1116 1117 void flush() { 1118 AbstractAssembler::flush(); 1119 } 1120 1121 inline void emit_int32(int); // shadows AbstractAssembler::emit_int32 1122 inline void emit_data(int); 1123 inline void emit_data(int, RelocationHolder const&); 1124 inline void emit_data(int, relocInfo::relocType rtype); 1125 1126 // Emit an address. 1127 inline address emit_addr(const address addr = NULL); 1128 1129 #if !defined(ABI_ELFv2) 1130 // Emit a function descriptor with the specified entry point, TOC, 1131 // and ENV. If the entry point is NULL, the descriptor will point 1132 // just past the descriptor. 1133 // Use values from friend functions as defaults. 1134 inline address emit_fd(address entry = NULL, 1135 address toc = (address) FunctionDescriptor::friend_toc, 1136 address env = (address) FunctionDescriptor::friend_env); 1137 #endif 1138 1139 ///////////////////////////////////////////////////////////////////////////////////// 1140 // PPC instructions 1141 ///////////////////////////////////////////////////////////////////////////////////// 1142 1143 // Memory instructions use r0 as hard coded 0, e.g. to simulate loading 1144 // immediates. The normal instruction encoders enforce that r0 is not 1145 // passed to them. Use either extended mnemonics encoders or the special ra0 1146 // versions. 1147 1148 // Issue an illegal instruction. 1149 inline void illtrap(); 1150 static inline bool is_illtrap(int x); 1151 1152 // PPC 1, section 3.3.8, Fixed-Point Arithmetic Instructions 1153 inline void addi( Register d, Register a, int si16); 1154 inline void addis(Register d, Register a, int si16); 1155 private: 1156 inline void addi_r0ok( Register d, Register a, int si16); 1157 inline void addis_r0ok(Register d, Register a, int si16); 1158 public: 1159 inline void addic_( Register d, Register a, int si16); 1160 inline void subfic( Register d, Register a, int si16); 1161 inline void add( Register d, Register a, Register b); 1162 inline void add_( Register d, Register a, Register b); 1163 inline void subf( Register d, Register a, Register b); // d = b - a "Sub_from", as in ppc spec. 1164 inline void sub( Register d, Register a, Register b); // d = a - b Swap operands of subf for readability. 1165 inline void subf_( Register d, Register a, Register b); 1166 inline void addc( Register d, Register a, Register b); 1167 inline void addc_( Register d, Register a, Register b); 1168 inline void subfc( Register d, Register a, Register b); 1169 inline void subfc_( Register d, Register a, Register b); 1170 inline void adde( Register d, Register a, Register b); 1171 inline void adde_( Register d, Register a, Register b); 1172 inline void subfe( Register d, Register a, Register b); 1173 inline void subfe_( Register d, Register a, Register b); 1174 inline void neg( Register d, Register a); 1175 inline void neg_( Register d, Register a); 1176 inline void mulli( Register d, Register a, int si16); 1177 inline void mulld( Register d, Register a, Register b); 1178 inline void mulld_( Register d, Register a, Register b); 1179 inline void mullw( Register d, Register a, Register b); 1180 inline void mullw_( Register d, Register a, Register b); 1181 inline void mulhw( Register d, Register a, Register b); 1182 inline void mulhw_( Register d, Register a, Register b); 1183 inline void mulhd( Register d, Register a, Register b); 1184 inline void mulhd_( Register d, Register a, Register b); 1185 inline void mulhdu( Register d, Register a, Register b); 1186 inline void mulhdu_(Register d, Register a, Register b); 1187 inline void divd( Register d, Register a, Register b); 1188 inline void divd_( Register d, Register a, Register b); 1189 inline void divw( Register d, Register a, Register b); 1190 inline void divw_( Register d, Register a, Register b); 1191 1192 // extended mnemonics 1193 inline void li( Register d, int si16); 1194 inline void lis( Register d, int si16); 1195 inline void addir(Register d, int si16, Register a); 1196 1197 static bool is_addi(int x) { 1198 return ADDI_OPCODE == (x & ADDI_OPCODE_MASK); 1199 } 1200 static bool is_addis(int x) { 1201 return ADDIS_OPCODE == (x & ADDIS_OPCODE_MASK); 1202 } 1203 static bool is_bxx(int x) { 1204 return BXX_OPCODE == (x & BXX_OPCODE_MASK); 1205 } 1206 static bool is_b(int x) { 1207 return BXX_OPCODE == (x & BXX_OPCODE_MASK) && inv_lk_field(x) == 0; 1208 } 1209 static bool is_bl(int x) { 1210 return BXX_OPCODE == (x & BXX_OPCODE_MASK) && inv_lk_field(x) == 1; 1211 } 1212 static bool is_bcxx(int x) { 1213 return BCXX_OPCODE == (x & BCXX_OPCODE_MASK); 1214 } 1215 static bool is_bxx_or_bcxx(int x) { 1216 return is_bxx(x) || is_bcxx(x); 1217 } 1218 static bool is_bctrl(int x) { 1219 return x == 0x4e800421; 1220 } 1221 static bool is_bctr(int x) { 1222 return x == 0x4e800420; 1223 } 1224 static bool is_bclr(int x) { 1225 return BCLR_OPCODE == (x & XL_FORM_OPCODE_MASK); 1226 } 1227 static bool is_li(int x) { 1228 return is_addi(x) && inv_ra_field(x)==0; 1229 } 1230 static bool is_lis(int x) { 1231 return is_addis(x) && inv_ra_field(x)==0; 1232 } 1233 static bool is_mtctr(int x) { 1234 return MTCTR_OPCODE == (x & MTCTR_OPCODE_MASK); 1235 } 1236 static bool is_ld(int x) { 1237 return LD_OPCODE == (x & LD_OPCODE_MASK); 1238 } 1239 static bool is_std(int x) { 1240 return STD_OPCODE == (x & STD_OPCODE_MASK); 1241 } 1242 static bool is_stdu(int x) { 1243 return STDU_OPCODE == (x & STDU_OPCODE_MASK); 1244 } 1245 static bool is_stdx(int x) { 1246 return STDX_OPCODE == (x & STDX_OPCODE_MASK); 1247 } 1248 static bool is_stdux(int x) { 1249 return STDUX_OPCODE == (x & STDUX_OPCODE_MASK); 1250 } 1251 static bool is_stwx(int x) { 1252 return STWX_OPCODE == (x & STWX_OPCODE_MASK); 1253 } 1254 static bool is_stwux(int x) { 1255 return STWUX_OPCODE == (x & STWUX_OPCODE_MASK); 1256 } 1257 static bool is_stw(int x) { 1258 return STW_OPCODE == (x & STW_OPCODE_MASK); 1259 } 1260 static bool is_stwu(int x) { 1261 return STWU_OPCODE == (x & STWU_OPCODE_MASK); 1262 } 1263 static bool is_ori(int x) { 1264 return ORI_OPCODE == (x & ORI_OPCODE_MASK); 1265 }; 1266 static bool is_oris(int x) { 1267 return ORIS_OPCODE == (x & ORIS_OPCODE_MASK); 1268 }; 1269 static bool is_rldicr(int x) { 1270 return (RLDICR_OPCODE == (x & RLDICR_OPCODE_MASK)); 1271 }; 1272 static bool is_nop(int x) { 1273 return x == 0x60000000; 1274 } 1275 // endgroup opcode for Power6 1276 static bool is_endgroup(int x) { 1277 return is_ori(x) && inv_ra_field(x) == 1 && inv_rs_field(x) == 1 && inv_d1_field(x) == 0; 1278 } 1279 1280 1281 private: 1282 // PPC 1, section 3.3.9, Fixed-Point Compare Instructions 1283 inline void cmpi( ConditionRegister bf, int l, Register a, int si16); 1284 inline void cmp( ConditionRegister bf, int l, Register a, Register b); 1285 inline void cmpli(ConditionRegister bf, int l, Register a, int ui16); 1286 inline void cmpl( ConditionRegister bf, int l, Register a, Register b); 1287 1288 public: 1289 // extended mnemonics of Compare Instructions 1290 inline void cmpwi( ConditionRegister crx, Register a, int si16); 1291 inline void cmpdi( ConditionRegister crx, Register a, int si16); 1292 inline void cmpw( ConditionRegister crx, Register a, Register b); 1293 inline void cmpd( ConditionRegister crx, Register a, Register b); 1294 inline void cmplwi(ConditionRegister crx, Register a, int ui16); 1295 inline void cmpldi(ConditionRegister crx, Register a, int ui16); 1296 inline void cmplw( ConditionRegister crx, Register a, Register b); 1297 inline void cmpld( ConditionRegister crx, Register a, Register b); 1298 1299 inline void isel( Register d, Register a, Register b, int bc); 1300 // Convenient version which takes: Condition register, Condition code and invert flag. Omit b to keep old value. 1301 inline void isel( Register d, ConditionRegister cr, Condition cc, bool inv, Register a, Register b = noreg); 1302 // Set d = 0 if (cr.cc) equals 1, otherwise b. 1303 inline void isel_0( Register d, ConditionRegister cr, Condition cc, Register b = noreg); 1304 1305 // PPC 1, section 3.3.11, Fixed-Point Logical Instructions 1306 void andi( Register a, Register s, int ui16); // optimized version 1307 inline void andi_( Register a, Register s, int ui16); 1308 inline void andis_( Register a, Register s, int ui16); 1309 inline void ori( Register a, Register s, int ui16); 1310 inline void oris( Register a, Register s, int ui16); 1311 inline void xori( Register a, Register s, int ui16); 1312 inline void xoris( Register a, Register s, int ui16); 1313 inline void andr( Register a, Register s, Register b); // suffixed by 'r' as 'and' is C++ keyword 1314 inline void and_( Register a, Register s, Register b); 1315 // Turn or0(rx,rx,rx) into a nop and avoid that we accidently emit a 1316 // SMT-priority change instruction (see SMT instructions below). 1317 inline void or_unchecked(Register a, Register s, Register b); 1318 inline void orr( Register a, Register s, Register b); // suffixed by 'r' as 'or' is C++ keyword 1319 inline void or_( Register a, Register s, Register b); 1320 inline void xorr( Register a, Register s, Register b); // suffixed by 'r' as 'xor' is C++ keyword 1321 inline void xor_( Register a, Register s, Register b); 1322 inline void nand( Register a, Register s, Register b); 1323 inline void nand_( Register a, Register s, Register b); 1324 inline void nor( Register a, Register s, Register b); 1325 inline void nor_( Register a, Register s, Register b); 1326 inline void andc( Register a, Register s, Register b); 1327 inline void andc_( Register a, Register s, Register b); 1328 inline void orc( Register a, Register s, Register b); 1329 inline void orc_( Register a, Register s, Register b); 1330 inline void extsb( Register a, Register s); 1331 inline void extsh( Register a, Register s); 1332 inline void extsw( Register a, Register s); 1333 1334 // extended mnemonics 1335 inline void nop(); 1336 // NOP for FP and BR units (different versions to allow them to be in one group) 1337 inline void fpnop0(); 1338 inline void fpnop1(); 1339 inline void brnop0(); 1340 inline void brnop1(); 1341 inline void brnop2(); 1342 1343 inline void mr( Register d, Register s); 1344 inline void ori_opt( Register d, int ui16); 1345 inline void oris_opt(Register d, int ui16); 1346 1347 // endgroup opcode for Power6 1348 inline void endgroup(); 1349 1350 // count instructions 1351 inline void cntlzw( Register a, Register s); 1352 inline void cntlzw_( Register a, Register s); 1353 inline void cntlzd( Register a, Register s); 1354 inline void cntlzd_( Register a, Register s); 1355 1356 // PPC 1, section 3.3.12, Fixed-Point Rotate and Shift Instructions 1357 inline void sld( Register a, Register s, Register b); 1358 inline void sld_( Register a, Register s, Register b); 1359 inline void slw( Register a, Register s, Register b); 1360 inline void slw_( Register a, Register s, Register b); 1361 inline void srd( Register a, Register s, Register b); 1362 inline void srd_( Register a, Register s, Register b); 1363 inline void srw( Register a, Register s, Register b); 1364 inline void srw_( Register a, Register s, Register b); 1365 inline void srad( Register a, Register s, Register b); 1366 inline void srad_( Register a, Register s, Register b); 1367 inline void sraw( Register a, Register s, Register b); 1368 inline void sraw_( Register a, Register s, Register b); 1369 inline void sradi( Register a, Register s, int sh6); 1370 inline void sradi_( Register a, Register s, int sh6); 1371 inline void srawi( Register a, Register s, int sh5); 1372 inline void srawi_( Register a, Register s, int sh5); 1373 1374 // extended mnemonics for Shift Instructions 1375 inline void sldi( Register a, Register s, int sh6); 1376 inline void sldi_( Register a, Register s, int sh6); 1377 inline void slwi( Register a, Register s, int sh5); 1378 inline void slwi_( Register a, Register s, int sh5); 1379 inline void srdi( Register a, Register s, int sh6); 1380 inline void srdi_( Register a, Register s, int sh6); 1381 inline void srwi( Register a, Register s, int sh5); 1382 inline void srwi_( Register a, Register s, int sh5); 1383 1384 inline void clrrdi( Register a, Register s, int ui6); 1385 inline void clrrdi_( Register a, Register s, int ui6); 1386 inline void clrldi( Register a, Register s, int ui6); 1387 inline void clrldi_( Register a, Register s, int ui6); 1388 inline void clrlsldi(Register a, Register s, int clrl6, int shl6); 1389 inline void clrlsldi_(Register a, Register s, int clrl6, int shl6); 1390 inline void extrdi( Register a, Register s, int n, int b); 1391 // testbit with condition register 1392 inline void testbitdi(ConditionRegister cr, Register a, Register s, int ui6); 1393 1394 // rotate instructions 1395 inline void rotldi( Register a, Register s, int n); 1396 inline void rotrdi( Register a, Register s, int n); 1397 inline void rotlwi( Register a, Register s, int n); 1398 inline void rotrwi( Register a, Register s, int n); 1399 1400 // Rotate Instructions 1401 inline void rldic( Register a, Register s, int sh6, int mb6); 1402 inline void rldic_( Register a, Register s, int sh6, int mb6); 1403 inline void rldicr( Register a, Register s, int sh6, int mb6); 1404 inline void rldicr_( Register a, Register s, int sh6, int mb6); 1405 inline void rldicl( Register a, Register s, int sh6, int mb6); 1406 inline void rldicl_( Register a, Register s, int sh6, int mb6); 1407 inline void rlwinm( Register a, Register s, int sh5, int mb5, int me5); 1408 inline void rlwinm_( Register a, Register s, int sh5, int mb5, int me5); 1409 inline void rldimi( Register a, Register s, int sh6, int mb6); 1410 inline void rldimi_( Register a, Register s, int sh6, int mb6); 1411 inline void rlwimi( Register a, Register s, int sh5, int mb5, int me5); 1412 inline void insrdi( Register a, Register s, int n, int b); 1413 inline void insrwi( Register a, Register s, int n, int b); 1414 1415 // PPC 1, section 3.3.2 Fixed-Point Load Instructions 1416 // 4 bytes 1417 inline void lwzx( Register d, Register s1, Register s2); 1418 inline void lwz( Register d, int si16, Register s1); 1419 inline void lwzu( Register d, int si16, Register s1); 1420 1421 // 4 bytes 1422 inline void lwax( Register d, Register s1, Register s2); 1423 inline void lwa( Register d, int si16, Register s1); 1424 1425 // 4 bytes reversed 1426 inline void lwbrx( Register d, Register s1, Register s2); 1427 1428 // 2 bytes 1429 inline void lhzx( Register d, Register s1, Register s2); 1430 inline void lhz( Register d, int si16, Register s1); 1431 inline void lhzu( Register d, int si16, Register s1); 1432 1433 // 2 bytes reversed 1434 inline void lhbrx( Register d, Register s1, Register s2); 1435 1436 // 2 bytes 1437 inline void lhax( Register d, Register s1, Register s2); 1438 inline void lha( Register d, int si16, Register s1); 1439 inline void lhau( Register d, int si16, Register s1); 1440 1441 // 1 byte 1442 inline void lbzx( Register d, Register s1, Register s2); 1443 inline void lbz( Register d, int si16, Register s1); 1444 inline void lbzu( Register d, int si16, Register s1); 1445 1446 // 8 bytes 1447 inline void ldx( Register d, Register s1, Register s2); 1448 inline void ld( Register d, int si16, Register s1); 1449 inline void ldu( Register d, int si16, Register s1); 1450 1451 // PPC 1, section 3.3.3 Fixed-Point Store Instructions 1452 inline void stwx( Register d, Register s1, Register s2); 1453 inline void stw( Register d, int si16, Register s1); 1454 inline void stwu( Register d, int si16, Register s1); 1455 1456 inline void sthx( Register d, Register s1, Register s2); 1457 inline void sth( Register d, int si16, Register s1); 1458 inline void sthu( Register d, int si16, Register s1); 1459 1460 inline void stbx( Register d, Register s1, Register s2); 1461 inline void stb( Register d, int si16, Register s1); 1462 inline void stbu( Register d, int si16, Register s1); 1463 1464 inline void stdx( Register d, Register s1, Register s2); 1465 inline void std( Register d, int si16, Register s1); 1466 inline void stdu( Register d, int si16, Register s1); 1467 inline void stdux(Register s, Register a, Register b); 1468 1469 // PPC 1, section 3.3.13 Move To/From System Register Instructions 1470 inline void mtlr( Register s1); 1471 inline void mflr( Register d); 1472 inline void mtctr(Register s1); 1473 inline void mfctr(Register d); 1474 inline void mtcrf(int fxm, Register s); 1475 inline void mfcr( Register d); 1476 inline void mcrf( ConditionRegister crd, ConditionRegister cra); 1477 inline void mtcr( Register s); 1478 1479 // Special purpose registers 1480 // Exception Register 1481 inline void mtxer(Register s1); 1482 inline void mfxer(Register d); 1483 // Vector Register Save Register 1484 inline void mtvrsave(Register s1); 1485 inline void mfvrsave(Register d); 1486 // Timebase 1487 inline void mftb(Register d); 1488 // Introduced with Power 8: 1489 // Data Stream Control Register 1490 inline void mtdscr(Register s1); 1491 inline void mfdscr(Register d ); 1492 // Transactional Memory Registers 1493 inline void mftfhar(Register d); 1494 inline void mftfiar(Register d); 1495 inline void mftexasr(Register d); 1496 inline void mftexasru(Register d); 1497 1498 // TEXASR bit description 1499 enum transaction_failure_reason { 1500 // Upper half (TEXASRU): 1501 tm_failure_persistent = 7, // The failure is likely to recur on each execution. 1502 tm_disallowed = 8, // The instruction is not permitted. 1503 tm_nesting_of = 9, // The maximum transaction level was exceeded. 1504 tm_footprint_of = 10, // The tracking limit for transactional storage accesses was exceeded. 1505 tm_self_induced_cf = 11, // A self-induced conflict occurred in Suspended state. 1506 tm_non_trans_cf = 12, // A conflict occurred with a non-transactional access by another processor. 1507 tm_trans_cf = 13, // A conflict occurred with another transaction. 1508 tm_translation_cf = 14, // A conflict occurred with a TLB invalidation. 1509 tm_inst_fetch_cf = 16, // An instruction fetch was performed from a block that was previously written transactionally. 1510 tm_tabort = 31, // Termination was caused by the execution of an abort instruction. 1511 // Lower half: 1512 tm_suspended = 32, // Failure was recorded in Suspended state. 1513 tm_failure_summary = 36, // Failure has been detected and recorded. 1514 tm_tfiar_exact = 37, // Value in the TFIAR is exact. 1515 tm_rot = 38, // Rollback-only transaction. 1516 }; 1517 1518 // PPC 1, section 2.4.1 Branch Instructions 1519 inline void b( address a, relocInfo::relocType rt = relocInfo::none); 1520 inline void b( Label& L); 1521 inline void bl( address a, relocInfo::relocType rt = relocInfo::none); 1522 inline void bl( Label& L); 1523 inline void bc( int boint, int biint, address a, relocInfo::relocType rt = relocInfo::none); 1524 inline void bc( int boint, int biint, Label& L); 1525 inline void bcl(int boint, int biint, address a, relocInfo::relocType rt = relocInfo::none); 1526 inline void bcl(int boint, int biint, Label& L); 1527 1528 inline void bclr( int boint, int biint, int bhint, relocInfo::relocType rt = relocInfo::none); 1529 inline void bclrl( int boint, int biint, int bhint, relocInfo::relocType rt = relocInfo::none); 1530 inline void bcctr( int boint, int biint, int bhint = bhintbhBCCTRisNotReturnButSame, 1531 relocInfo::relocType rt = relocInfo::none); 1532 inline void bcctrl(int boint, int biint, int bhint = bhintbhBCLRisReturn, 1533 relocInfo::relocType rt = relocInfo::none); 1534 1535 // helper function for b, bcxx 1536 inline bool is_within_range_of_b(address a, address pc); 1537 inline bool is_within_range_of_bcxx(address a, address pc); 1538 1539 // get the destination of a bxx branch (b, bl, ba, bla) 1540 static inline address bxx_destination(address baddr); 1541 static inline address bxx_destination(int instr, address pc); 1542 static inline intptr_t bxx_destination_offset(int instr, intptr_t bxx_pos); 1543 1544 // extended mnemonics for branch instructions 1545 inline void blt(ConditionRegister crx, Label& L); 1546 inline void bgt(ConditionRegister crx, Label& L); 1547 inline void beq(ConditionRegister crx, Label& L); 1548 inline void bso(ConditionRegister crx, Label& L); 1549 inline void bge(ConditionRegister crx, Label& L); 1550 inline void ble(ConditionRegister crx, Label& L); 1551 inline void bne(ConditionRegister crx, Label& L); 1552 inline void bns(ConditionRegister crx, Label& L); 1553 1554 // Branch instructions with static prediction hints. 1555 inline void blt_predict_taken( ConditionRegister crx, Label& L); 1556 inline void bgt_predict_taken( ConditionRegister crx, Label& L); 1557 inline void beq_predict_taken( ConditionRegister crx, Label& L); 1558 inline void bso_predict_taken( ConditionRegister crx, Label& L); 1559 inline void bge_predict_taken( ConditionRegister crx, Label& L); 1560 inline void ble_predict_taken( ConditionRegister crx, Label& L); 1561 inline void bne_predict_taken( ConditionRegister crx, Label& L); 1562 inline void bns_predict_taken( ConditionRegister crx, Label& L); 1563 inline void blt_predict_not_taken(ConditionRegister crx, Label& L); 1564 inline void bgt_predict_not_taken(ConditionRegister crx, Label& L); 1565 inline void beq_predict_not_taken(ConditionRegister crx, Label& L); 1566 inline void bso_predict_not_taken(ConditionRegister crx, Label& L); 1567 inline void bge_predict_not_taken(ConditionRegister crx, Label& L); 1568 inline void ble_predict_not_taken(ConditionRegister crx, Label& L); 1569 inline void bne_predict_not_taken(ConditionRegister crx, Label& L); 1570 inline void bns_predict_not_taken(ConditionRegister crx, Label& L); 1571 1572 // for use in conjunction with testbitdi: 1573 inline void btrue( ConditionRegister crx, Label& L); 1574 inline void bfalse(ConditionRegister crx, Label& L); 1575 1576 inline void bltl(ConditionRegister crx, Label& L); 1577 inline void bgtl(ConditionRegister crx, Label& L); 1578 inline void beql(ConditionRegister crx, Label& L); 1579 inline void bsol(ConditionRegister crx, Label& L); 1580 inline void bgel(ConditionRegister crx, Label& L); 1581 inline void blel(ConditionRegister crx, Label& L); 1582 inline void bnel(ConditionRegister crx, Label& L); 1583 inline void bnsl(ConditionRegister crx, Label& L); 1584 1585 // extended mnemonics for Branch Instructions via LR 1586 // We use `blr' for returns. 1587 inline void blr(relocInfo::relocType rt = relocInfo::none); 1588 1589 // extended mnemonics for Branch Instructions with CTR 1590 // bdnz means `decrement CTR and jump to L if CTR is not zero' 1591 inline void bdnz(Label& L); 1592 // Decrement and branch if result is zero. 1593 inline void bdz(Label& L); 1594 // we use `bctr[l]' for jumps/calls in function descriptor glue 1595 // code, e.g. calls to runtime functions 1596 inline void bctr( relocInfo::relocType rt = relocInfo::none); 1597 inline void bctrl(relocInfo::relocType rt = relocInfo::none); 1598 // conditional jumps/branches via CTR 1599 inline void beqctr( ConditionRegister crx, relocInfo::relocType rt = relocInfo::none); 1600 inline void beqctrl(ConditionRegister crx, relocInfo::relocType rt = relocInfo::none); 1601 inline void bnectr( ConditionRegister crx, relocInfo::relocType rt = relocInfo::none); 1602 inline void bnectrl(ConditionRegister crx, relocInfo::relocType rt = relocInfo::none); 1603 1604 // condition register logic instructions 1605 // NOTE: There's a preferred form: d and s2 should point into the same condition register. 1606 inline void crand( int d, int s1, int s2); 1607 inline void crnand(int d, int s1, int s2); 1608 inline void cror( int d, int s1, int s2); 1609 inline void crxor( int d, int s1, int s2); 1610 inline void crnor( int d, int s1, int s2); 1611 inline void creqv( int d, int s1, int s2); 1612 inline void crandc(int d, int s1, int s2); 1613 inline void crorc( int d, int s1, int s2); 1614 1615 // More convenient version. 1616 int condition_register_bit(ConditionRegister cr, Condition c) { 1617 return 4 * (int)(intptr_t)cr + c; 1618 } 1619 void crand( ConditionRegister crdst, Condition cdst, ConditionRegister crsrc, Condition csrc); 1620 void crnand(ConditionRegister crdst, Condition cdst, ConditionRegister crsrc, Condition csrc); 1621 void cror( ConditionRegister crdst, Condition cdst, ConditionRegister crsrc, Condition csrc); 1622 void crxor( ConditionRegister crdst, Condition cdst, ConditionRegister crsrc, Condition csrc); 1623 void crnor( ConditionRegister crdst, Condition cdst, ConditionRegister crsrc, Condition csrc); 1624 void creqv( ConditionRegister crdst, Condition cdst, ConditionRegister crsrc, Condition csrc); 1625 void crandc(ConditionRegister crdst, Condition cdst, ConditionRegister crsrc, Condition csrc); 1626 void crorc( ConditionRegister crdst, Condition cdst, ConditionRegister crsrc, Condition csrc); 1627 1628 // icache and dcache related instructions 1629 inline void icbi( Register s1, Register s2); 1630 //inline void dcba(Register s1, Register s2); // Instruction for embedded processor only. 1631 inline void dcbz( Register s1, Register s2); 1632 inline void dcbst( Register s1, Register s2); 1633 inline void dcbf( Register s1, Register s2); 1634 1635 enum ct_cache_specification { 1636 ct_primary_cache = 0, 1637 ct_secondary_cache = 2 1638 }; 1639 // dcache read hint 1640 inline void dcbt( Register s1, Register s2); 1641 inline void dcbtct( Register s1, Register s2, int ct); 1642 inline void dcbtds( Register s1, Register s2, int ds); 1643 // dcache write hint 1644 inline void dcbtst( Register s1, Register s2); 1645 inline void dcbtstct(Register s1, Register s2, int ct); 1646 1647 // machine barrier instructions: 1648 // 1649 // - sync two-way memory barrier, aka fence 1650 // - lwsync orders Store|Store, 1651 // Load|Store, 1652 // Load|Load, 1653 // but not Store|Load 1654 // - eieio orders memory accesses for device memory (only) 1655 // - isync invalidates speculatively executed instructions 1656 // From the Power ISA 2.06 documentation: 1657 // "[...] an isync instruction prevents the execution of 1658 // instructions following the isync until instructions 1659 // preceding the isync have completed, [...]" 1660 // From IBM's AIX assembler reference: 1661 // "The isync [...] instructions causes the processor to 1662 // refetch any instructions that might have been fetched 1663 // prior to the isync instruction. The instruction isync 1664 // causes the processor to wait for all previous instructions 1665 // to complete. Then any instructions already fetched are 1666 // discarded and instruction processing continues in the 1667 // environment established by the previous instructions." 1668 // 1669 // semantic barrier instructions: 1670 // (as defined in orderAccess.hpp) 1671 // 1672 // - release orders Store|Store, (maps to lwsync) 1673 // Load|Store 1674 // - acquire orders Load|Store, (maps to lwsync) 1675 // Load|Load 1676 // - fence orders Store|Store, (maps to sync) 1677 // Load|Store, 1678 // Load|Load, 1679 // Store|Load 1680 // 1681 private: 1682 inline void sync(int l); 1683 public: 1684 inline void sync(); 1685 inline void lwsync(); 1686 inline void ptesync(); 1687 inline void eieio(); 1688 inline void isync(); 1689 inline void elemental_membar(int e); // Elemental Memory Barriers (>=Power 8) 1690 1691 // atomics 1692 inline void lwarx_unchecked(Register d, Register a, Register b, int eh1 = 0); 1693 inline void ldarx_unchecked(Register d, Register a, Register b, int eh1 = 0); 1694 inline bool lxarx_hint_exclusive_access(); 1695 inline void lwarx( Register d, Register a, Register b, bool hint_exclusive_access = false); 1696 inline void ldarx( Register d, Register a, Register b, bool hint_exclusive_access = false); 1697 inline void stwcx_( Register s, Register a, Register b); 1698 inline void stdcx_( Register s, Register a, Register b); 1699 1700 // Instructions for adjusting thread priority for simultaneous 1701 // multithreading (SMT) on Power5. 1702 private: 1703 inline void smt_prio_very_low(); 1704 inline void smt_prio_medium_high(); 1705 inline void smt_prio_high(); 1706 1707 public: 1708 inline void smt_prio_low(); 1709 inline void smt_prio_medium_low(); 1710 inline void smt_prio_medium(); 1711 // >= Power7 1712 inline void smt_yield(); 1713 inline void smt_mdoio(); 1714 inline void smt_mdoom(); 1715 1716 // trap instructions 1717 inline void twi_0(Register a); // for load with acquire semantics use load+twi_0+isync (trap can't occur) 1718 // NOT FOR DIRECT USE!! 1719 protected: 1720 inline void tdi_unchecked(int tobits, Register a, int si16); 1721 inline void twi_unchecked(int tobits, Register a, int si16); 1722 inline void tdi( int tobits, Register a, int si16); // asserts UseSIGTRAP 1723 inline void twi( int tobits, Register a, int si16); // asserts UseSIGTRAP 1724 inline void td( int tobits, Register a, Register b); // asserts UseSIGTRAP 1725 inline void tw( int tobits, Register a, Register b); // asserts UseSIGTRAP 1726 1727 static bool is_tdi(int x, int tobits, int ra, int si16) { 1728 return (TDI_OPCODE == (x & TDI_OPCODE_MASK)) 1729 && (tobits == inv_to_field(x)) 1730 && (ra == -1/*any reg*/ || ra == inv_ra_field(x)) 1731 && (si16 == inv_si_field(x)); 1732 } 1733 1734 static bool is_twi(int x, int tobits, int ra, int si16) { 1735 return (TWI_OPCODE == (x & TWI_OPCODE_MASK)) 1736 && (tobits == inv_to_field(x)) 1737 && (ra == -1/*any reg*/ || ra == inv_ra_field(x)) 1738 && (si16 == inv_si_field(x)); 1739 } 1740 1741 static bool is_twi(int x, int tobits, int ra) { 1742 return (TWI_OPCODE == (x & TWI_OPCODE_MASK)) 1743 && (tobits == inv_to_field(x)) 1744 && (ra == -1/*any reg*/ || ra == inv_ra_field(x)); 1745 } 1746 1747 static bool is_td(int x, int tobits, int ra, int rb) { 1748 return (TD_OPCODE == (x & TD_OPCODE_MASK)) 1749 && (tobits == inv_to_field(x)) 1750 && (ra == -1/*any reg*/ || ra == inv_ra_field(x)) 1751 && (rb == -1/*any reg*/ || rb == inv_rb_field(x)); 1752 } 1753 1754 static bool is_tw(int x, int tobits, int ra, int rb) { 1755 return (TW_OPCODE == (x & TW_OPCODE_MASK)) 1756 && (tobits == inv_to_field(x)) 1757 && (ra == -1/*any reg*/ || ra == inv_ra_field(x)) 1758 && (rb == -1/*any reg*/ || rb == inv_rb_field(x)); 1759 } 1760 1761 public: 1762 // PPC floating point instructions 1763 // PPC 1, section 4.6.2 Floating-Point Load Instructions 1764 inline void lfs( FloatRegister d, int si16, Register a); 1765 inline void lfsu( FloatRegister d, int si16, Register a); 1766 inline void lfsx( FloatRegister d, Register a, Register b); 1767 inline void lfd( FloatRegister d, int si16, Register a); 1768 inline void lfdu( FloatRegister d, int si16, Register a); 1769 inline void lfdx( FloatRegister d, Register a, Register b); 1770 1771 // PPC 1, section 4.6.3 Floating-Point Store Instructions 1772 inline void stfs( FloatRegister s, int si16, Register a); 1773 inline void stfsu( FloatRegister s, int si16, Register a); 1774 inline void stfsx( FloatRegister s, Register a, Register b); 1775 inline void stfd( FloatRegister s, int si16, Register a); 1776 inline void stfdu( FloatRegister s, int si16, Register a); 1777 inline void stfdx( FloatRegister s, Register a, Register b); 1778 1779 // PPC 1, section 4.6.4 Floating-Point Move Instructions 1780 inline void fmr( FloatRegister d, FloatRegister b); 1781 inline void fmr_( FloatRegister d, FloatRegister b); 1782 1783 // inline void mffgpr( FloatRegister d, Register b); 1784 // inline void mftgpr( Register d, FloatRegister b); 1785 inline void cmpb( Register a, Register s, Register b); 1786 inline void popcntb(Register a, Register s); 1787 inline void popcntw(Register a, Register s); 1788 inline void popcntd(Register a, Register s); 1789 1790 inline void fneg( FloatRegister d, FloatRegister b); 1791 inline void fneg_( FloatRegister d, FloatRegister b); 1792 inline void fabs( FloatRegister d, FloatRegister b); 1793 inline void fabs_( FloatRegister d, FloatRegister b); 1794 inline void fnabs( FloatRegister d, FloatRegister b); 1795 inline void fnabs_(FloatRegister d, FloatRegister b); 1796 1797 // PPC 1, section 4.6.5.1 Floating-Point Elementary Arithmetic Instructions 1798 inline void fadd( FloatRegister d, FloatRegister a, FloatRegister b); 1799 inline void fadd_( FloatRegister d, FloatRegister a, FloatRegister b); 1800 inline void fadds( FloatRegister d, FloatRegister a, FloatRegister b); 1801 inline void fadds_(FloatRegister d, FloatRegister a, FloatRegister b); 1802 inline void fsub( FloatRegister d, FloatRegister a, FloatRegister b); 1803 inline void fsub_( FloatRegister d, FloatRegister a, FloatRegister b); 1804 inline void fsubs( FloatRegister d, FloatRegister a, FloatRegister b); 1805 inline void fsubs_(FloatRegister d, FloatRegister a, FloatRegister b); 1806 inline void fmul( FloatRegister d, FloatRegister a, FloatRegister c); 1807 inline void fmul_( FloatRegister d, FloatRegister a, FloatRegister c); 1808 inline void fmuls( FloatRegister d, FloatRegister a, FloatRegister c); 1809 inline void fmuls_(FloatRegister d, FloatRegister a, FloatRegister c); 1810 inline void fdiv( FloatRegister d, FloatRegister a, FloatRegister b); 1811 inline void fdiv_( FloatRegister d, FloatRegister a, FloatRegister b); 1812 inline void fdivs( FloatRegister d, FloatRegister a, FloatRegister b); 1813 inline void fdivs_(FloatRegister d, FloatRegister a, FloatRegister b); 1814 1815 // PPC 1, section 4.6.6 Floating-Point Rounding and Conversion Instructions 1816 inline void frsp( FloatRegister d, FloatRegister b); 1817 inline void fctid( FloatRegister d, FloatRegister b); 1818 inline void fctidz(FloatRegister d, FloatRegister b); 1819 inline void fctiw( FloatRegister d, FloatRegister b); 1820 inline void fctiwz(FloatRegister d, FloatRegister b); 1821 inline void fcfid( FloatRegister d, FloatRegister b); 1822 inline void fcfids(FloatRegister d, FloatRegister b); 1823 1824 // PPC 1, section 4.6.7 Floating-Point Compare Instructions 1825 inline void fcmpu( ConditionRegister crx, FloatRegister a, FloatRegister b); 1826 1827 inline void fsqrt( FloatRegister d, FloatRegister b); 1828 inline void fsqrts(FloatRegister d, FloatRegister b); 1829 1830 // Vector instructions for >= Power6. 1831 inline void lvebx( VectorRegister d, Register s1, Register s2); 1832 inline void lvehx( VectorRegister d, Register s1, Register s2); 1833 inline void lvewx( VectorRegister d, Register s1, Register s2); 1834 inline void lvx( VectorRegister d, Register s1, Register s2); 1835 inline void lvxl( VectorRegister d, Register s1, Register s2); 1836 inline void stvebx( VectorRegister d, Register s1, Register s2); 1837 inline void stvehx( VectorRegister d, Register s1, Register s2); 1838 inline void stvewx( VectorRegister d, Register s1, Register s2); 1839 inline void stvx( VectorRegister d, Register s1, Register s2); 1840 inline void stvxl( VectorRegister d, Register s1, Register s2); 1841 inline void lvsl( VectorRegister d, Register s1, Register s2); 1842 inline void lvsr( VectorRegister d, Register s1, Register s2); 1843 inline void vpkpx( VectorRegister d, VectorRegister a, VectorRegister b); 1844 inline void vpkshss( VectorRegister d, VectorRegister a, VectorRegister b); 1845 inline void vpkswss( VectorRegister d, VectorRegister a, VectorRegister b); 1846 inline void vpkshus( VectorRegister d, VectorRegister a, VectorRegister b); 1847 inline void vpkswus( VectorRegister d, VectorRegister a, VectorRegister b); 1848 inline void vpkuhum( VectorRegister d, VectorRegister a, VectorRegister b); 1849 inline void vpkuwum( VectorRegister d, VectorRegister a, VectorRegister b); 1850 inline void vpkuhus( VectorRegister d, VectorRegister a, VectorRegister b); 1851 inline void vpkuwus( VectorRegister d, VectorRegister a, VectorRegister b); 1852 inline void vupkhpx( VectorRegister d, VectorRegister b); 1853 inline void vupkhsb( VectorRegister d, VectorRegister b); 1854 inline void vupkhsh( VectorRegister d, VectorRegister b); 1855 inline void vupklpx( VectorRegister d, VectorRegister b); 1856 inline void vupklsb( VectorRegister d, VectorRegister b); 1857 inline void vupklsh( VectorRegister d, VectorRegister b); 1858 inline void vmrghb( VectorRegister d, VectorRegister a, VectorRegister b); 1859 inline void vmrghw( VectorRegister d, VectorRegister a, VectorRegister b); 1860 inline void vmrghh( VectorRegister d, VectorRegister a, VectorRegister b); 1861 inline void vmrglb( VectorRegister d, VectorRegister a, VectorRegister b); 1862 inline void vmrglw( VectorRegister d, VectorRegister a, VectorRegister b); 1863 inline void vmrglh( VectorRegister d, VectorRegister a, VectorRegister b); 1864 inline void vsplt( VectorRegister d, int ui4, VectorRegister b); 1865 inline void vsplth( VectorRegister d, int ui3, VectorRegister b); 1866 inline void vspltw( VectorRegister d, int ui2, VectorRegister b); 1867 inline void vspltisb( VectorRegister d, int si5); 1868 inline void vspltish( VectorRegister d, int si5); 1869 inline void vspltisw( VectorRegister d, int si5); 1870 inline void vperm( VectorRegister d, VectorRegister a, VectorRegister b, VectorRegister c); 1871 inline void vsel( VectorRegister d, VectorRegister a, VectorRegister b, VectorRegister c); 1872 inline void vsl( VectorRegister d, VectorRegister a, VectorRegister b); 1873 inline void vsldoi( VectorRegister d, VectorRegister a, VectorRegister b, int si4); 1874 inline void vslo( VectorRegister d, VectorRegister a, VectorRegister b); 1875 inline void vsr( VectorRegister d, VectorRegister a, VectorRegister b); 1876 inline void vsro( VectorRegister d, VectorRegister a, VectorRegister b); 1877 inline void vaddcuw( VectorRegister d, VectorRegister a, VectorRegister b); 1878 inline void vaddshs( VectorRegister d, VectorRegister a, VectorRegister b); 1879 inline void vaddsbs( VectorRegister d, VectorRegister a, VectorRegister b); 1880 inline void vaddsws( VectorRegister d, VectorRegister a, VectorRegister b); 1881 inline void vaddubm( VectorRegister d, VectorRegister a, VectorRegister b); 1882 inline void vadduwm( VectorRegister d, VectorRegister a, VectorRegister b); 1883 inline void vadduhm( VectorRegister d, VectorRegister a, VectorRegister b); 1884 inline void vaddubs( VectorRegister d, VectorRegister a, VectorRegister b); 1885 inline void vadduws( VectorRegister d, VectorRegister a, VectorRegister b); 1886 inline void vadduhs( VectorRegister d, VectorRegister a, VectorRegister b); 1887 inline void vsubcuw( VectorRegister d, VectorRegister a, VectorRegister b); 1888 inline void vsubshs( VectorRegister d, VectorRegister a, VectorRegister b); 1889 inline void vsubsbs( VectorRegister d, VectorRegister a, VectorRegister b); 1890 inline void vsubsws( VectorRegister d, VectorRegister a, VectorRegister b); 1891 inline void vsububm( VectorRegister d, VectorRegister a, VectorRegister b); 1892 inline void vsubuwm( VectorRegister d, VectorRegister a, VectorRegister b); 1893 inline void vsubuhm( VectorRegister d, VectorRegister a, VectorRegister b); 1894 inline void vsububs( VectorRegister d, VectorRegister a, VectorRegister b); 1895 inline void vsubuws( VectorRegister d, VectorRegister a, VectorRegister b); 1896 inline void vsubuhs( VectorRegister d, VectorRegister a, VectorRegister b); 1897 inline void vmulesb( VectorRegister d, VectorRegister a, VectorRegister b); 1898 inline void vmuleub( VectorRegister d, VectorRegister a, VectorRegister b); 1899 inline void vmulesh( VectorRegister d, VectorRegister a, VectorRegister b); 1900 inline void vmuleuh( VectorRegister d, VectorRegister a, VectorRegister b); 1901 inline void vmulosb( VectorRegister d, VectorRegister a, VectorRegister b); 1902 inline void vmuloub( VectorRegister d, VectorRegister a, VectorRegister b); 1903 inline void vmulosh( VectorRegister d, VectorRegister a, VectorRegister b); 1904 inline void vmulouh( VectorRegister d, VectorRegister a, VectorRegister b); 1905 inline void vmhaddshs(VectorRegister d, VectorRegister a, VectorRegister b, VectorRegister c); 1906 inline void vmhraddshs(VectorRegister d,VectorRegister a, VectorRegister b, VectorRegister c); 1907 inline void vmladduhm(VectorRegister d, VectorRegister a, VectorRegister b, VectorRegister c); 1908 inline void vmsubuhm( VectorRegister d, VectorRegister a, VectorRegister b, VectorRegister c); 1909 inline void vmsummbm( VectorRegister d, VectorRegister a, VectorRegister b, VectorRegister c); 1910 inline void vmsumshm( VectorRegister d, VectorRegister a, VectorRegister b, VectorRegister c); 1911 inline void vmsumshs( VectorRegister d, VectorRegister a, VectorRegister b, VectorRegister c); 1912 inline void vmsumuhm( VectorRegister d, VectorRegister a, VectorRegister b, VectorRegister c); 1913 inline void vmsumuhs( VectorRegister d, VectorRegister a, VectorRegister b, VectorRegister c); 1914 inline void vsumsws( VectorRegister d, VectorRegister a, VectorRegister b); 1915 inline void vsum2sws( VectorRegister d, VectorRegister a, VectorRegister b); 1916 inline void vsum4sbs( VectorRegister d, VectorRegister a, VectorRegister b); 1917 inline void vsum4ubs( VectorRegister d, VectorRegister a, VectorRegister b); 1918 inline void vsum4shs( VectorRegister d, VectorRegister a, VectorRegister b); 1919 inline void vavgsb( VectorRegister d, VectorRegister a, VectorRegister b); 1920 inline void vavgsw( VectorRegister d, VectorRegister a, VectorRegister b); 1921 inline void vavgsh( VectorRegister d, VectorRegister a, VectorRegister b); 1922 inline void vavgub( VectorRegister d, VectorRegister a, VectorRegister b); 1923 inline void vavguw( VectorRegister d, VectorRegister a, VectorRegister b); 1924 inline void vavguh( VectorRegister d, VectorRegister a, VectorRegister b); 1925 inline void vmaxsb( VectorRegister d, VectorRegister a, VectorRegister b); 1926 inline void vmaxsw( VectorRegister d, VectorRegister a, VectorRegister b); 1927 inline void vmaxsh( VectorRegister d, VectorRegister a, VectorRegister b); 1928 inline void vmaxub( VectorRegister d, VectorRegister a, VectorRegister b); 1929 inline void vmaxuw( VectorRegister d, VectorRegister a, VectorRegister b); 1930 inline void vmaxuh( VectorRegister d, VectorRegister a, VectorRegister b); 1931 inline void vminsb( VectorRegister d, VectorRegister a, VectorRegister b); 1932 inline void vminsw( VectorRegister d, VectorRegister a, VectorRegister b); 1933 inline void vminsh( VectorRegister d, VectorRegister a, VectorRegister b); 1934 inline void vminub( VectorRegister d, VectorRegister a, VectorRegister b); 1935 inline void vminuw( VectorRegister d, VectorRegister a, VectorRegister b); 1936 inline void vminuh( VectorRegister d, VectorRegister a, VectorRegister b); 1937 inline void vcmpequb( VectorRegister d, VectorRegister a, VectorRegister b); 1938 inline void vcmpequh( VectorRegister d, VectorRegister a, VectorRegister b); 1939 inline void vcmpequw( VectorRegister d, VectorRegister a, VectorRegister b); 1940 inline void vcmpgtsh( VectorRegister d, VectorRegister a, VectorRegister b); 1941 inline void vcmpgtsb( VectorRegister d, VectorRegister a, VectorRegister b); 1942 inline void vcmpgtsw( VectorRegister d, VectorRegister a, VectorRegister b); 1943 inline void vcmpgtub( VectorRegister d, VectorRegister a, VectorRegister b); 1944 inline void vcmpgtuh( VectorRegister d, VectorRegister a, VectorRegister b); 1945 inline void vcmpgtuw( VectorRegister d, VectorRegister a, VectorRegister b); 1946 inline void vcmpequb_(VectorRegister d, VectorRegister a, VectorRegister b); 1947 inline void vcmpequh_(VectorRegister d, VectorRegister a, VectorRegister b); 1948 inline void vcmpequw_(VectorRegister d, VectorRegister a, VectorRegister b); 1949 inline void vcmpgtsh_(VectorRegister d, VectorRegister a, VectorRegister b); 1950 inline void vcmpgtsb_(VectorRegister d, VectorRegister a, VectorRegister b); 1951 inline void vcmpgtsw_(VectorRegister d, VectorRegister a, VectorRegister b); 1952 inline void vcmpgtub_(VectorRegister d, VectorRegister a, VectorRegister b); 1953 inline void vcmpgtuh_(VectorRegister d, VectorRegister a, VectorRegister b); 1954 inline void vcmpgtuw_(VectorRegister d, VectorRegister a, VectorRegister b); 1955 inline void vand( VectorRegister d, VectorRegister a, VectorRegister b); 1956 inline void vandc( VectorRegister d, VectorRegister a, VectorRegister b); 1957 inline void vnor( VectorRegister d, VectorRegister a, VectorRegister b); 1958 inline void vor( VectorRegister d, VectorRegister a, VectorRegister b); 1959 inline void vxor( VectorRegister d, VectorRegister a, VectorRegister b); 1960 inline void vrlb( VectorRegister d, VectorRegister a, VectorRegister b); 1961 inline void vrlw( VectorRegister d, VectorRegister a, VectorRegister b); 1962 inline void vrlh( VectorRegister d, VectorRegister a, VectorRegister b); 1963 inline void vslb( VectorRegister d, VectorRegister a, VectorRegister b); 1964 inline void vskw( VectorRegister d, VectorRegister a, VectorRegister b); 1965 inline void vslh( VectorRegister d, VectorRegister a, VectorRegister b); 1966 inline void vsrb( VectorRegister d, VectorRegister a, VectorRegister b); 1967 inline void vsrw( VectorRegister d, VectorRegister a, VectorRegister b); 1968 inline void vsrh( VectorRegister d, VectorRegister a, VectorRegister b); 1969 inline void vsrab( VectorRegister d, VectorRegister a, VectorRegister b); 1970 inline void vsraw( VectorRegister d, VectorRegister a, VectorRegister b); 1971 inline void vsrah( VectorRegister d, VectorRegister a, VectorRegister b); 1972 // Vector Floating-Point not implemented yet 1973 inline void mtvscr( VectorRegister b); 1974 inline void mfvscr( VectorRegister d); 1975 1976 // AES (introduced with Power 8) 1977 inline void vcipher( VectorRegister d, VectorRegister a, VectorRegister b); 1978 inline void vcipherlast( VectorRegister d, VectorRegister a, VectorRegister b); 1979 inline void vncipher( VectorRegister d, VectorRegister a, VectorRegister b); 1980 inline void vncipherlast(VectorRegister d, VectorRegister a, VectorRegister b); 1981 inline void vsbox( VectorRegister d, VectorRegister a); 1982 1983 // SHA (introduced with Power 8) 1984 // Not yet implemented. 1985 1986 // Vector Binary Polynomial Multiplication (introduced with Power 8) 1987 inline void vpmsumb( VectorRegister d, VectorRegister a, VectorRegister b); 1988 inline void vpmsumd( VectorRegister d, VectorRegister a, VectorRegister b); 1989 inline void vpmsumh( VectorRegister d, VectorRegister a, VectorRegister b); 1990 inline void vpmsumw( VectorRegister d, VectorRegister a, VectorRegister b); 1991 1992 // Vector Permute and Xor (introduced with Power 8) 1993 inline void vpermxor( VectorRegister d, VectorRegister a, VectorRegister b, VectorRegister c); 1994 1995 // Transactional Memory instructions (introduced with Power 8) 1996 inline void tbegin_(); // R=0 1997 inline void tbeginrot_(); // R=1 Rollback-Only Transaction 1998 inline void tend_(); // A=0 1999 inline void tendall_(); // A=1 2000 inline void tabort_(); 2001 inline void tabort_(Register a); 2002 inline void tabortwc_(int t, Register a, Register b); 2003 inline void tabortwci_(int t, Register a, int si); 2004 inline void tabortdc_(int t, Register a, Register b); 2005 inline void tabortdci_(int t, Register a, int si); 2006 inline void tsuspend_(); // tsr with L=0 2007 inline void tresume_(); // tsr with L=1 2008 inline void tcheck(int f); 2009 2010 static bool is_tbegin(int x) { 2011 return TBEGIN_OPCODE == (x & (0x3f << OPCODE_SHIFT | 0x3ff << 1)); 2012 } 2013 2014 // The following encoders use r0 as second operand. These instructions 2015 // read r0 as '0'. 2016 inline void lwzx( Register d, Register s2); 2017 inline void lwz( Register d, int si16); 2018 inline void lwax( Register d, Register s2); 2019 inline void lwa( Register d, int si16); 2020 inline void lwbrx(Register d, Register s2); 2021 inline void lhzx( Register d, Register s2); 2022 inline void lhz( Register d, int si16); 2023 inline void lhax( Register d, Register s2); 2024 inline void lha( Register d, int si16); 2025 inline void lhbrx(Register d, Register s2); 2026 inline void lbzx( Register d, Register s2); 2027 inline void lbz( Register d, int si16); 2028 inline void ldx( Register d, Register s2); 2029 inline void ld( Register d, int si16); 2030 inline void stwx( Register d, Register s2); 2031 inline void stw( Register d, int si16); 2032 inline void sthx( Register d, Register s2); 2033 inline void sth( Register d, int si16); 2034 inline void stbx( Register d, Register s2); 2035 inline void stb( Register d, int si16); 2036 inline void stdx( Register d, Register s2); 2037 inline void std( Register d, int si16); 2038 2039 // PPC 2, section 3.2.1 Instruction Cache Instructions 2040 inline void icbi( Register s2); 2041 // PPC 2, section 3.2.2 Data Cache Instructions 2042 //inlinevoid dcba( Register s2); // Instruction for embedded processor only. 2043 inline void dcbz( Register s2); 2044 inline void dcbst( Register s2); 2045 inline void dcbf( Register s2); 2046 // dcache read hint 2047 inline void dcbt( Register s2); 2048 inline void dcbtct( Register s2, int ct); 2049 inline void dcbtds( Register s2, int ds); 2050 // dcache write hint 2051 inline void dcbtst( Register s2); 2052 inline void dcbtstct(Register s2, int ct); 2053 2054 // Atomics: use ra0mem to disallow R0 as base. 2055 inline void lwarx_unchecked(Register d, Register b, int eh1); 2056 inline void ldarx_unchecked(Register d, Register b, int eh1); 2057 inline void lwarx( Register d, Register b, bool hint_exclusive_access); 2058 inline void ldarx( Register d, Register b, bool hint_exclusive_access); 2059 inline void stwcx_(Register s, Register b); 2060 inline void stdcx_(Register s, Register b); 2061 inline void lfs( FloatRegister d, int si16); 2062 inline void lfsx( FloatRegister d, Register b); 2063 inline void lfd( FloatRegister d, int si16); 2064 inline void lfdx( FloatRegister d, Register b); 2065 inline void stfs( FloatRegister s, int si16); 2066 inline void stfsx( FloatRegister s, Register b); 2067 inline void stfd( FloatRegister s, int si16); 2068 inline void stfdx( FloatRegister s, Register b); 2069 inline void lvebx( VectorRegister d, Register s2); 2070 inline void lvehx( VectorRegister d, Register s2); 2071 inline void lvewx( VectorRegister d, Register s2); 2072 inline void lvx( VectorRegister d, Register s2); 2073 inline void lvxl( VectorRegister d, Register s2); 2074 inline void stvebx(VectorRegister d, Register s2); 2075 inline void stvehx(VectorRegister d, Register s2); 2076 inline void stvewx(VectorRegister d, Register s2); 2077 inline void stvx( VectorRegister d, Register s2); 2078 inline void stvxl( VectorRegister d, Register s2); 2079 inline void lvsl( VectorRegister d, Register s2); 2080 inline void lvsr( VectorRegister d, Register s2); 2081 2082 // RegisterOrConstant versions. 2083 // These emitters choose between the versions using two registers and 2084 // those with register and immediate, depending on the content of roc. 2085 // If the constant is not encodable as immediate, instructions to 2086 // load the constant are emitted beforehand. Store instructions need a 2087 // tmp reg if the constant is not encodable as immediate. 2088 // Size unpredictable. 2089 void ld( Register d, RegisterOrConstant roc, Register s1 = noreg); 2090 void lwa( Register d, RegisterOrConstant roc, Register s1 = noreg); 2091 void lwz( Register d, RegisterOrConstant roc, Register s1 = noreg); 2092 void lha( Register d, RegisterOrConstant roc, Register s1 = noreg); 2093 void lhz( Register d, RegisterOrConstant roc, Register s1 = noreg); 2094 void lbz( Register d, RegisterOrConstant roc, Register s1 = noreg); 2095 void std( Register d, RegisterOrConstant roc, Register s1 = noreg, Register tmp = noreg); 2096 void stw( Register d, RegisterOrConstant roc, Register s1 = noreg, Register tmp = noreg); 2097 void sth( Register d, RegisterOrConstant roc, Register s1 = noreg, Register tmp = noreg); 2098 void stb( Register d, RegisterOrConstant roc, Register s1 = noreg, Register tmp = noreg); 2099 void add( Register d, RegisterOrConstant roc, Register s1); 2100 void subf(Register d, RegisterOrConstant roc, Register s1); 2101 void cmpd(ConditionRegister d, RegisterOrConstant roc, Register s1); 2102 2103 2104 // Emit several instructions to load a 64 bit constant. This issues a fixed 2105 // instruction pattern so that the constant can be patched later on. 2106 enum { 2107 load_const_size = 5 * BytesPerInstWord 2108 }; 2109 void load_const(Register d, long a, Register tmp = noreg); 2110 inline void load_const(Register d, void* a, Register tmp = noreg); 2111 inline void load_const(Register d, Label& L, Register tmp = noreg); 2112 inline void load_const(Register d, AddressLiteral& a, Register tmp = noreg); 2113 2114 // Load a 64 bit constant, optimized, not identifyable. 2115 // Tmp can be used to increase ILP. Set return_simm16_rest = true to get a 2116 // 16 bit immediate offset. This is useful if the offset can be encoded in 2117 // a succeeding instruction. 2118 int load_const_optimized(Register d, long a, Register tmp = noreg, bool return_simm16_rest = false); 2119 inline int load_const_optimized(Register d, void* a, Register tmp = noreg, bool return_simm16_rest = false) { 2120 return load_const_optimized(d, (long)(unsigned long)a, tmp, return_simm16_rest); 2121 } 2122 2123 // Creation 2124 Assembler(CodeBuffer* code) : AbstractAssembler(code) { 2125 #ifdef CHECK_DELAY 2126 delay_state = no_delay; 2127 #endif 2128 } 2129 2130 // Testing 2131 #ifndef PRODUCT 2132 void test_asm(); 2133 #endif 2134 }; 2135 2136 2137 #endif // CPU_PPC_VM_ASSEMBLER_PPC_HPP