1 /* 2 * Toshiba TX79-specific instructions translation routines 3 * 4 * Copyright (c) 2018 Fredrik Noring 5 * Copyright (c) 2021 Philippe Mathieu-Daudé 6 * 7 * SPDX-License-Identifier: GPL-2.0-or-later 8 */ 9 10 #include "qemu/osdep.h" 11 #include "tcg/tcg-op.h" 12 #include "tcg/tcg-op-gvec.h" 13 #include "exec/helper-gen.h" 14 #include "translate.h" 15 16 /* Include the auto-generated decoder. */ 17 #include "decode-tx79.c.inc" 18 19 /* 20 * Overview of the TX79-specific instruction set 21 * ============================================= 22 * 23 * The R5900 and the C790 have 128-bit wide GPRs, where the upper 64 bits 24 * are only used by the specific quadword (128-bit) LQ/SQ load/store 25 * instructions and certain multimedia instructions (MMIs). These MMIs 26 * configure the 128-bit data path as two 64-bit, four 32-bit, eight 16-bit 27 * or sixteen 8-bit paths. 28 * 29 * Reference: 30 * 31 * The Toshiba TX System RISC TX79 Core Architecture manual, 32 * https://wiki.qemu.org/File:C790.pdf 33 */ 34 35 bool decode_ext_tx79(DisasContext *ctx, uint32_t insn) 36 { 37 if (TARGET_LONG_BITS == 64 && decode_tx79(ctx, insn)) { 38 return true; 39 } 40 return false; 41 } 42 43 /* 44 * Three-Operand Multiply and Multiply-Add (4 instructions) 45 * -------------------------------------------------------- 46 * MADD [rd,] rs, rt Multiply/Add 47 * MADDU [rd,] rs, rt Multiply/Add Unsigned 48 * MULT [rd,] rs, rt Multiply (3-operand) 49 * MULTU [rd,] rs, rt Multiply Unsigned (3-operand) 50 */ 51 52 /* 53 * Multiply Instructions for Pipeline 1 (10 instructions) 54 * ------------------------------------------------------ 55 * MULT1 [rd,] rs, rt Multiply Pipeline 1 56 * MULTU1 [rd,] rs, rt Multiply Unsigned Pipeline 1 57 * DIV1 rs, rt Divide Pipeline 1 58 * DIVU1 rs, rt Divide Unsigned Pipeline 1 59 * MADD1 [rd,] rs, rt Multiply-Add Pipeline 1 60 * MADDU1 [rd,] rs, rt Multiply-Add Unsigned Pipeline 1 61 * MFHI1 rd Move From HI1 Register 62 * MFLO1 rd Move From LO1 Register 63 * MTHI1 rs Move To HI1 Register 64 * MTLO1 rs Move To LO1 Register 65 */ 66 67 static bool trans_MFHI1(DisasContext *ctx, arg_rtype *a) 68 { 69 gen_store_gpr(cpu_HI[1], a->rd); 70 71 return true; 72 } 73 74 static bool trans_MFLO1(DisasContext *ctx, arg_rtype *a) 75 { 76 gen_store_gpr(cpu_LO[1], a->rd); 77 78 return true; 79 } 80 81 static bool trans_MTHI1(DisasContext *ctx, arg_rtype *a) 82 { 83 gen_load_gpr(cpu_HI[1], a->rs); 84 85 return true; 86 } 87 88 static bool trans_MTLO1(DisasContext *ctx, arg_rtype *a) 89 { 90 gen_load_gpr(cpu_LO[1], a->rs); 91 92 return true; 93 } 94 95 /* 96 * Arithmetic (19 instructions) 97 * ---------------------------- 98 * PADDB rd, rs, rt Parallel Add Byte 99 * PSUBB rd, rs, rt Parallel Subtract Byte 100 * PADDH rd, rs, rt Parallel Add Halfword 101 * PSUBH rd, rs, rt Parallel Subtract Halfword 102 * PADDW rd, rs, rt Parallel Add Word 103 * PSUBW rd, rs, rt Parallel Subtract Word 104 * PADSBH rd, rs, rt Parallel Add/Subtract Halfword 105 * PADDSB rd, rs, rt Parallel Add with Signed Saturation Byte 106 * PSUBSB rd, rs, rt Parallel Subtract with Signed Saturation Byte 107 * PADDSH rd, rs, rt Parallel Add with Signed Saturation Halfword 108 * PSUBSH rd, rs, rt Parallel Subtract with Signed Saturation Halfword 109 * PADDSW rd, rs, rt Parallel Add with Signed Saturation Word 110 * PSUBSW rd, rs, rt Parallel Subtract with Signed Saturation Word 111 * PADDUB rd, rs, rt Parallel Add with Unsigned saturation Byte 112 * PSUBUB rd, rs, rt Parallel Subtract with Unsigned saturation Byte 113 * PADDUH rd, rs, rt Parallel Add with Unsigned saturation Halfword 114 * PSUBUH rd, rs, rt Parallel Subtract with Unsigned saturation Halfword 115 * PADDUW rd, rs, rt Parallel Add with Unsigned saturation Word 116 * PSUBUW rd, rs, rt Parallel Subtract with Unsigned saturation Word 117 */ 118 119 static bool trans_parallel_arith(DisasContext *ctx, arg_rtype *a, 120 void (*gen_logic_i64)(TCGv_i64, TCGv_i64, TCGv_i64)) 121 { 122 TCGv_i64 ax, bx; 123 124 if (a->rd == 0) { 125 /* nop */ 126 return true; 127 } 128 129 ax = tcg_temp_new_i64(); 130 bx = tcg_temp_new_i64(); 131 132 /* Lower half */ 133 gen_load_gpr(ax, a->rs); 134 gen_load_gpr(bx, a->rt); 135 gen_logic_i64(cpu_gpr[a->rd], ax, bx); 136 137 /* Upper half */ 138 gen_load_gpr_hi(ax, a->rs); 139 gen_load_gpr_hi(bx, a->rt); 140 gen_logic_i64(cpu_gpr_hi[a->rd], ax, bx); 141 142 tcg_temp_free(bx); 143 tcg_temp_free(ax); 144 145 return true; 146 } 147 148 /* Parallel Subtract Byte */ 149 static bool trans_PSUBB(DisasContext *ctx, arg_rtype *a) 150 { 151 return trans_parallel_arith(ctx, a, tcg_gen_vec_sub8_i64); 152 } 153 154 /* Parallel Subtract Halfword */ 155 static bool trans_PSUBH(DisasContext *ctx, arg_rtype *a) 156 { 157 return trans_parallel_arith(ctx, a, tcg_gen_vec_sub16_i64); 158 } 159 160 /* Parallel Subtract Word */ 161 static bool trans_PSUBW(DisasContext *ctx, arg_rtype *a) 162 { 163 return trans_parallel_arith(ctx, a, tcg_gen_vec_sub32_i64); 164 } 165 166 /* 167 * Min/Max (4 instructions) 168 * ------------------------ 169 * PMAXH rd, rs, rt Parallel Maximum Halfword 170 * PMINH rd, rs, rt Parallel Minimum Halfword 171 * PMAXW rd, rs, rt Parallel Maximum Word 172 * PMINW rd, rs, rt Parallel Minimum Word 173 */ 174 175 /* 176 * Absolute (2 instructions) 177 * ------------------------- 178 * PABSH rd, rt Parallel Absolute Halfword 179 * PABSW rd, rt Parallel Absolute Word 180 */ 181 182 /* 183 * Logical (4 instructions) 184 * ------------------------ 185 * PAND rd, rs, rt Parallel AND 186 * POR rd, rs, rt Parallel OR 187 * PXOR rd, rs, rt Parallel XOR 188 * PNOR rd, rs, rt Parallel NOR 189 */ 190 191 /* Parallel And */ 192 static bool trans_PAND(DisasContext *ctx, arg_rtype *a) 193 { 194 return trans_parallel_arith(ctx, a, tcg_gen_and_i64); 195 } 196 197 /* Parallel Or */ 198 static bool trans_POR(DisasContext *ctx, arg_rtype *a) 199 { 200 return trans_parallel_arith(ctx, a, tcg_gen_or_i64); 201 } 202 203 /* Parallel Exclusive Or */ 204 static bool trans_PXOR(DisasContext *ctx, arg_rtype *a) 205 { 206 return trans_parallel_arith(ctx, a, tcg_gen_xor_i64); 207 } 208 209 /* Parallel Not Or */ 210 static bool trans_PNOR(DisasContext *ctx, arg_rtype *a) 211 { 212 return trans_parallel_arith(ctx, a, tcg_gen_nor_i64); 213 } 214 215 /* 216 * Shift (9 instructions) 217 * ---------------------- 218 * PSLLH rd, rt, sa Parallel Shift Left Logical Halfword 219 * PSRLH rd, rt, sa Parallel Shift Right Logical Halfword 220 * PSRAH rd, rt, sa Parallel Shift Right Arithmetic Halfword 221 * PSLLW rd, rt, sa Parallel Shift Left Logical Word 222 * PSRLW rd, rt, sa Parallel Shift Right Logical Word 223 * PSRAW rd, rt, sa Parallel Shift Right Arithmetic Word 224 * PSLLVW rd, rt, rs Parallel Shift Left Logical Variable Word 225 * PSRLVW rd, rt, rs Parallel Shift Right Logical Variable Word 226 * PSRAVW rd, rt, rs Parallel Shift Right Arithmetic Variable Word 227 */ 228 229 /* 230 * Compare (6 instructions) 231 * ------------------------ 232 * PCGTB rd, rs, rt Parallel Compare for Greater Than Byte 233 * PCEQB rd, rs, rt Parallel Compare for Equal Byte 234 * PCGTH rd, rs, rt Parallel Compare for Greater Than Halfword 235 * PCEQH rd, rs, rt Parallel Compare for Equal Halfword 236 * PCGTW rd, rs, rt Parallel Compare for Greater Than Word 237 * PCEQW rd, rs, rt Parallel Compare for Equal Word 238 */ 239 240 static bool trans_parallel_compare(DisasContext *ctx, arg_rtype *a, 241 TCGCond cond, unsigned wlen) 242 { 243 TCGv_i64 c0, c1, ax, bx, t0, t1, t2; 244 245 if (a->rd == 0) { 246 /* nop */ 247 return true; 248 } 249 250 c0 = tcg_const_tl(0); 251 c1 = tcg_const_tl(0xffffffff); 252 ax = tcg_temp_new_i64(); 253 bx = tcg_temp_new_i64(); 254 t0 = tcg_temp_new_i64(); 255 t1 = tcg_temp_new_i64(); 256 t2 = tcg_temp_new_i64(); 257 258 /* Lower half */ 259 gen_load_gpr(ax, a->rs); 260 gen_load_gpr(bx, a->rt); 261 for (int i = 0; i < (64 / wlen); i++) { 262 tcg_gen_sextract_i64(t0, ax, wlen * i, wlen); 263 tcg_gen_sextract_i64(t1, bx, wlen * i, wlen); 264 tcg_gen_movcond_i64(cond, t2, t1, t0, c1, c0); 265 tcg_gen_deposit_i64(cpu_gpr[a->rd], cpu_gpr[a->rd], t2, wlen * i, wlen); 266 } 267 /* Upper half */ 268 gen_load_gpr_hi(ax, a->rs); 269 gen_load_gpr_hi(bx, a->rt); 270 for (int i = 0; i < (64 / wlen); i++) { 271 tcg_gen_sextract_i64(t0, ax, wlen * i, wlen); 272 tcg_gen_sextract_i64(t1, bx, wlen * i, wlen); 273 tcg_gen_movcond_i64(cond, t2, t1, t0, c1, c0); 274 tcg_gen_deposit_i64(cpu_gpr_hi[a->rd], cpu_gpr_hi[a->rd], t2, wlen * i, wlen); 275 } 276 277 tcg_temp_free(t2); 278 tcg_temp_free(t1); 279 tcg_temp_free(t0); 280 tcg_temp_free(bx); 281 tcg_temp_free(ax); 282 tcg_temp_free(c1); 283 tcg_temp_free(c0); 284 285 return true; 286 } 287 288 /* Parallel Compare for Equal Byte */ 289 static bool trans_PCEQB(DisasContext *ctx, arg_rtype *a) 290 { 291 return trans_parallel_compare(ctx, a, TCG_COND_EQ, 8); 292 } 293 294 /* Parallel Compare for Equal Halfword */ 295 static bool trans_PCEQH(DisasContext *ctx, arg_rtype *a) 296 { 297 return trans_parallel_compare(ctx, a, TCG_COND_EQ, 16); 298 } 299 300 /* Parallel Compare for Equal Word */ 301 static bool trans_PCEQW(DisasContext *ctx, arg_rtype *a) 302 { 303 return trans_parallel_compare(ctx, a, TCG_COND_EQ, 32); 304 } 305 306 /* 307 * LZC (1 instruction) 308 * ------------------- 309 * PLZCW rd, rs Parallel Leading Zero or One Count Word 310 */ 311 312 /* 313 * Quadword Load and Store (2 instructions) 314 * ---------------------------------------- 315 * LQ rt, offset(base) Load Quadword 316 * SQ rt, offset(base) Store Quadword 317 */ 318 319 /* 320 * Multiply and Divide (19 instructions) 321 * ------------------------------------- 322 * PMULTW rd, rs, rt Parallel Multiply Word 323 * PMULTUW rd, rs, rt Parallel Multiply Unsigned Word 324 * PDIVW rs, rt Parallel Divide Word 325 * PDIVUW rs, rt Parallel Divide Unsigned Word 326 * PMADDW rd, rs, rt Parallel Multiply-Add Word 327 * PMADDUW rd, rs, rt Parallel Multiply-Add Unsigned Word 328 * PMSUBW rd, rs, rt Parallel Multiply-Subtract Word 329 * PMULTH rd, rs, rt Parallel Multiply Halfword 330 * PMADDH rd, rs, rt Parallel Multiply-Add Halfword 331 * PMSUBH rd, rs, rt Parallel Multiply-Subtract Halfword 332 * PHMADH rd, rs, rt Parallel Horizontal Multiply-Add Halfword 333 * PHMSBH rd, rs, rt Parallel Horizontal Multiply-Subtract Halfword 334 * PDIVBW rs, rt Parallel Divide Broadcast Word 335 * PMFHI rd Parallel Move From HI Register 336 * PMFLO rd Parallel Move From LO Register 337 * PMTHI rs Parallel Move To HI Register 338 * PMTLO rs Parallel Move To LO Register 339 * PMFHL rd Parallel Move From HI/LO Register 340 * PMTHL rs Parallel Move To HI/LO Register 341 */ 342 343 /* 344 * Pack/Extend (11 instructions) 345 * ----------------------------- 346 * PPAC5 rd, rt Parallel Pack to 5 bits 347 * PPACB rd, rs, rt Parallel Pack to Byte 348 * PPACH rd, rs, rt Parallel Pack to Halfword 349 * PPACW rd, rs, rt Parallel Pack to Word 350 * PEXT5 rd, rt Parallel Extend Upper from 5 bits 351 * PEXTUB rd, rs, rt Parallel Extend Upper from Byte 352 * PEXTLB rd, rs, rt Parallel Extend Lower from Byte 353 * PEXTUH rd, rs, rt Parallel Extend Upper from Halfword 354 * PEXTLH rd, rs, rt Parallel Extend Lower from Halfword 355 * PEXTUW rd, rs, rt Parallel Extend Upper from Word 356 * PEXTLW rd, rs, rt Parallel Extend Lower from Word 357 */ 358 359 static void gen_pextw(TCGv_i64 dl, TCGv_i64 dh, TCGv_i64 a, TCGv_i64 b) 360 { 361 tcg_gen_deposit_i64(dl, b, a, 32, 32); 362 tcg_gen_shri_i64(b, b, 32); 363 tcg_gen_deposit_i64(dh, a, b, 0, 32); 364 } 365 366 static bool trans_PEXTLx(DisasContext *ctx, arg_rtype *a, unsigned wlen) 367 { 368 TCGv_i64 ax, bx; 369 370 if (a->rd == 0) { 371 /* nop */ 372 return true; 373 } 374 375 ax = tcg_temp_new_i64(); 376 bx = tcg_temp_new_i64(); 377 378 gen_load_gpr(ax, a->rs); 379 gen_load_gpr(bx, a->rt); 380 381 /* Lower half */ 382 for (int i = 0; i < 64 / (2 * wlen); i++) { 383 tcg_gen_deposit_i64(cpu_gpr[a->rd], 384 cpu_gpr[a->rd], bx, 2 * wlen * i, wlen); 385 tcg_gen_deposit_i64(cpu_gpr[a->rd], 386 cpu_gpr[a->rd], ax, 2 * wlen * i + wlen, wlen); 387 tcg_gen_shri_i64(bx, bx, wlen); 388 tcg_gen_shri_i64(ax, ax, wlen); 389 } 390 /* Upper half */ 391 for (int i = 0; i < 64 / (2 * wlen); i++) { 392 tcg_gen_deposit_i64(cpu_gpr_hi[a->rd], 393 cpu_gpr_hi[a->rd], bx, 2 * wlen * i, wlen); 394 tcg_gen_deposit_i64(cpu_gpr_hi[a->rd], 395 cpu_gpr_hi[a->rd], ax, 2 * wlen * i + wlen, wlen); 396 tcg_gen_shri_i64(bx, bx, wlen); 397 tcg_gen_shri_i64(ax, ax, wlen); 398 } 399 400 tcg_temp_free(bx); 401 tcg_temp_free(ax); 402 403 return true; 404 } 405 406 /* Parallel Extend Lower from Byte */ 407 static bool trans_PEXTLB(DisasContext *ctx, arg_rtype *a) 408 { 409 return trans_PEXTLx(ctx, a, 8); 410 } 411 412 /* Parallel Extend Lower from Halfword */ 413 static bool trans_PEXTLH(DisasContext *ctx, arg_rtype *a) 414 { 415 return trans_PEXTLx(ctx, a, 16); 416 } 417 418 /* Parallel Extend Lower from Word */ 419 static bool trans_PEXTLW(DisasContext *ctx, arg_rtype *a) 420 { 421 TCGv_i64 ax, bx; 422 423 if (a->rd == 0) { 424 /* nop */ 425 return true; 426 } 427 428 ax = tcg_temp_new_i64(); 429 bx = tcg_temp_new_i64(); 430 431 gen_load_gpr(ax, a->rs); 432 gen_load_gpr(bx, a->rt); 433 gen_pextw(cpu_gpr[a->rd], cpu_gpr_hi[a->rd], ax, bx); 434 435 tcg_temp_free(bx); 436 tcg_temp_free(ax); 437 438 return true; 439 } 440 441 /* Parallel Extend Upper from Word */ 442 static bool trans_PEXTUW(DisasContext *ctx, arg_rtype *a) 443 { 444 TCGv_i64 ax, bx; 445 446 if (a->rd == 0) { 447 /* nop */ 448 return true; 449 } 450 451 ax = tcg_temp_new_i64(); 452 bx = tcg_temp_new_i64(); 453 454 gen_load_gpr_hi(ax, a->rs); 455 gen_load_gpr_hi(bx, a->rt); 456 gen_pextw(cpu_gpr[a->rd], cpu_gpr_hi[a->rd], ax, bx); 457 458 tcg_temp_free(bx); 459 tcg_temp_free(ax); 460 461 return true; 462 } 463 464 /* 465 * Others (16 instructions) 466 * ------------------------ 467 * PCPYH rd, rt Parallel Copy Halfword 468 * PCPYLD rd, rs, rt Parallel Copy Lower Doubleword 469 * PCPYUD rd, rs, rt Parallel Copy Upper Doubleword 470 * PREVH rd, rt Parallel Reverse Halfword 471 * PINTH rd, rs, rt Parallel Interleave Halfword 472 * PINTEH rd, rs, rt Parallel Interleave Even Halfword 473 * PEXEH rd, rt Parallel Exchange Even Halfword 474 * PEXCH rd, rt Parallel Exchange Center Halfword 475 * PEXEW rd, rt Parallel Exchange Even Word 476 * PEXCW rd, rt Parallel Exchange Center Word 477 * QFSRV rd, rs, rt Quadword Funnel Shift Right Variable 478 * MFSA rd Move from Shift Amount Register 479 * MTSA rs Move to Shift Amount Register 480 * MTSAB rs, immediate Move Byte Count to Shift Amount Register 481 * MTSAH rs, immediate Move Halfword Count to Shift Amount Register 482 * PROT3W rd, rt Parallel Rotate 3 Words 483 */ 484 485 /* Parallel Copy Halfword */ 486 static bool trans_PCPYH(DisasContext *s, arg_rtype *a) 487 { 488 if (a->rd == 0) { 489 /* nop */ 490 return true; 491 } 492 493 if (a->rt == 0) { 494 tcg_gen_movi_i64(cpu_gpr[a->rd], 0); 495 tcg_gen_movi_i64(cpu_gpr_hi[a->rd], 0); 496 return true; 497 } 498 499 tcg_gen_deposit_i64(cpu_gpr[a->rd], cpu_gpr[a->rt], cpu_gpr[a->rt], 16, 16); 500 tcg_gen_deposit_i64(cpu_gpr[a->rd], cpu_gpr[a->rd], cpu_gpr[a->rd], 32, 32); 501 tcg_gen_deposit_i64(cpu_gpr_hi[a->rd], cpu_gpr_hi[a->rt], cpu_gpr_hi[a->rt], 16, 16); 502 tcg_gen_deposit_i64(cpu_gpr_hi[a->rd], cpu_gpr_hi[a->rd], cpu_gpr_hi[a->rd], 32, 32); 503 504 return true; 505 } 506 507 /* Parallel Copy Lower Doubleword */ 508 static bool trans_PCPYLD(DisasContext *s, arg_rtype *a) 509 { 510 if (a->rd == 0) { 511 /* nop */ 512 return true; 513 } 514 515 if (a->rs == 0) { 516 tcg_gen_movi_i64(cpu_gpr_hi[a->rd], 0); 517 } else { 518 tcg_gen_mov_i64(cpu_gpr_hi[a->rd], cpu_gpr[a->rs]); 519 } 520 521 if (a->rt == 0) { 522 tcg_gen_movi_i64(cpu_gpr[a->rd], 0); 523 } else if (a->rd != a->rt) { 524 tcg_gen_mov_i64(cpu_gpr[a->rd], cpu_gpr[a->rt]); 525 } 526 527 return true; 528 } 529 530 /* Parallel Copy Upper Doubleword */ 531 static bool trans_PCPYUD(DisasContext *s, arg_rtype *a) 532 { 533 if (a->rd == 0) { 534 /* nop */ 535 return true; 536 } 537 538 gen_load_gpr_hi(cpu_gpr[a->rd], a->rs); 539 540 if (a->rt == 0) { 541 tcg_gen_movi_i64(cpu_gpr_hi[a->rd], 0); 542 } else if (a->rd != a->rt) { 543 tcg_gen_mov_i64(cpu_gpr_hi[a->rd], cpu_gpr_hi[a->rt]); 544 } 545 546 return true; 547 } 548