| 1 | /* |
| 2 | * This file is part of SIS. |
| 3 | * |
| 4 | * SIS, SPARC instruction simulator V1.8 Copyright (C) 1995 Jiri Gaisler, |
| 5 | * European Space Agency |
| 6 | * |
| 7 | * This program is free software; you can redistribute it and/or modify it under |
| 8 | * the terms of the GNU General Public License as published by the Free |
| 9 | * Software Foundation; either version 2 of the License, or (at your option) |
| 10 | * any later version. |
| 11 | * |
| 12 | * This program is distributed in the hope that it will be useful, but WITHOUT |
| 13 | * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or |
| 14 | * FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for |
| 15 | * more details. |
| 16 | * |
| 17 | * You should have received a copy of the GNU General Public License along with |
| 18 | * this program; if not, write to the Free Software Foundation, Inc., 675 |
| 19 | * Mass Ave, Cambridge, MA 02139, USA. |
| 20 | * |
| 21 | */ |
| 22 | |
| 23 | #include "sis.h" |
| 24 | #include "end.h" |
| 25 | #include <math.h> |
| 26 | #include <stdio.h> |
| 27 | |
| 28 | extern int32 sis_verbose, sparclite; |
| 29 | int ext_irl = 0; |
| 30 | |
| 31 | /* Load/store interlock delay */ |
| 32 | #define FLSTHOLD 1 |
| 33 | |
| 34 | /* Load delay (delete if unwanted - speeds up simulation) */ |
| 35 | #define LOAD_DEL 1 |
| 36 | |
| 37 | #define T_LD 2 |
| 38 | #define T_LDD 3 |
| 39 | #define T_ST 3 |
| 40 | #define T_STD 4 |
| 41 | #define T_LDST 4 |
| 42 | #define T_JMPL 2 |
| 43 | #define T_RETT 2 |
| 44 | |
| 45 | #define FSR_QNE 0x2000 |
| 46 | #define FP_EXE_MODE 0 |
| 47 | #define FP_EXC_PE 1 |
| 48 | #define FP_EXC_MODE 2 |
| 49 | |
| 50 | #define FBA 8 |
| 51 | #define FBN 0 |
| 52 | #define FBNE 1 |
| 53 | #define FBLG 2 |
| 54 | #define FBUL 3 |
| 55 | #define FBL 4 |
| 56 | #define FBUG 5 |
| 57 | #define FBG 6 |
| 58 | #define FBU 7 |
| 59 | #define FBA 8 |
| 60 | #define FBE 9 |
| 61 | #define FBUE 10 |
| 62 | #define FBGE 11 |
| 63 | #define FBUGE 12 |
| 64 | #define FBLE 13 |
| 65 | #define FBULE 14 |
| 66 | #define FBO 15 |
| 67 | |
| 68 | #define FCC_E 0 |
| 69 | #define FCC_L 1 |
| 70 | #define FCC_G 2 |
| 71 | #define FCC_U 3 |
| 72 | |
| 73 | #define PSR_ET 0x20 |
| 74 | #define PSR_EF 0x1000 |
| 75 | #define PSR_PS 0x40 |
| 76 | #define PSR_S 0x80 |
| 77 | #define PSR_N 0x0800000 |
| 78 | #define PSR_Z 0x0400000 |
| 79 | #define PSR_V 0x0200000 |
| 80 | #define PSR_C 0x0100000 |
| 81 | #define PSR_CC 0x0F00000 |
| 82 | #define PSR_CWP 0x7 |
| 83 | #define PSR_PIL 0x0f00 |
| 84 | |
| 85 | #define ICC_N (icc >> 3) |
| 86 | #define ICC_Z (icc >> 2) |
| 87 | #define ICC_V (icc >> 1) |
| 88 | #define ICC_C (icc) |
| 89 | |
| 90 | #define FP_PRES (sregs->fpu_pres) |
| 91 | |
| 92 | #define TRAP_IEXC 1 |
| 93 | #define TRAP_UNIMP 2 |
| 94 | #define TRAP_PRIVI 3 |
| 95 | #define TRAP_FPDIS 4 |
| 96 | #define TRAP_WOFL 5 |
| 97 | #define TRAP_WUFL 6 |
| 98 | #define TRAP_UNALI 7 |
| 99 | #define TRAP_FPEXC 8 |
| 100 | #define TRAP_DEXC 9 |
| 101 | #define TRAP_TAG 10 |
| 102 | #define TRAP_DIV0 0x2a |
| 103 | |
| 104 | #define FSR_TT 0x1C000 |
| 105 | #define FP_IEEE 0x04000 |
| 106 | #define FP_UNIMP 0x0C000 |
| 107 | #define FP_SEQ_ERR 0x10000 |
| 108 | |
| 109 | #define BICC_BN 0 |
| 110 | #define BICC_BE 1 |
| 111 | #define BICC_BLE 2 |
| 112 | #define BICC_BL 3 |
| 113 | #define BICC_BLEU 4 |
| 114 | #define BICC_BCS 5 |
| 115 | #define BICC_NEG 6 |
| 116 | #define BICC_BVS 7 |
| 117 | #define BICC_BA 8 |
| 118 | #define BICC_BNE 9 |
| 119 | #define BICC_BG 10 |
| 120 | #define BICC_BGE 11 |
| 121 | #define BICC_BGU 12 |
| 122 | #define BICC_BCC 13 |
| 123 | #define BICC_POS 14 |
| 124 | #define BICC_BVC 15 |
| 125 | |
| 126 | #define INST_SIMM13 0x1fff |
| 127 | #define INST_RS2 0x1f |
| 128 | #define INST_I 0x2000 |
| 129 | #define ADD 0x00 |
| 130 | #define ADDCC 0x10 |
| 131 | #define ADDX 0x08 |
| 132 | #define ADDXCC 0x18 |
| 133 | #define TADDCC 0x20 |
| 134 | #define TSUBCC 0x21 |
| 135 | #define TADDCCTV 0x22 |
| 136 | #define TSUBCCTV 0x23 |
| 137 | #define IAND 0x01 |
| 138 | #define IANDCC 0x11 |
| 139 | #define IANDN 0x05 |
| 140 | #define IANDNCC 0x15 |
| 141 | #define MULScc 0x24 |
| 142 | #define DIVScc 0x1D |
| 143 | #define SMUL 0x0B |
| 144 | #define SMULCC 0x1B |
| 145 | #define UMUL 0x0A |
| 146 | #define UMULCC 0x1A |
| 147 | #define SDIV 0x0F |
| 148 | #define SDIVCC 0x1F |
| 149 | #define UDIV 0x0E |
| 150 | #define UDIVCC 0x1E |
| 151 | #define IOR 0x02 |
| 152 | #define IORCC 0x12 |
| 153 | #define IORN 0x06 |
| 154 | #define IORNCC 0x16 |
| 155 | #define SLL 0x25 |
| 156 | #define SRA 0x27 |
| 157 | #define SRL 0x26 |
| 158 | #define SUB 0x04 |
| 159 | #define SUBCC 0x14 |
| 160 | #define SUBX 0x0C |
| 161 | #define SUBXCC 0x1C |
| 162 | #define IXNOR 0x07 |
| 163 | #define IXNORCC 0x17 |
| 164 | #define IXOR 0x03 |
| 165 | #define IXORCC 0x13 |
| 166 | #define SETHI 0x04 |
| 167 | #define BICC 0x02 |
| 168 | #define FPBCC 0x06 |
| 169 | #define RDY 0x28 |
| 170 | #define RDPSR 0x29 |
| 171 | #define RDWIM 0x2A |
| 172 | #define RDTBR 0x2B |
| 173 | #define SCAN 0x2C |
| 174 | #define WRY 0x30 |
| 175 | #define WRPSR 0x31 |
| 176 | #define WRWIM 0x32 |
| 177 | #define WRTBR 0x33 |
| 178 | #define JMPL 0x38 |
| 179 | #define RETT 0x39 |
| 180 | #define TICC 0x3A |
| 181 | #define SAVE 0x3C |
| 182 | #define RESTORE 0x3D |
| 183 | #define LDD 0x03 |
| 184 | #define LDDA 0x13 |
| 185 | #define LD 0x00 |
| 186 | #define LDA 0x10 |
| 187 | #define LDF 0x20 |
| 188 | #define LDDF 0x23 |
| 189 | #define LDSTUB 0x0D |
| 190 | #define LDSTUBA 0x1D |
| 191 | #define LDUB 0x01 |
| 192 | #define LDUBA 0x11 |
| 193 | #define LDSB 0x09 |
| 194 | #define LDSBA 0x19 |
| 195 | #define LDUH 0x02 |
| 196 | #define LDUHA 0x12 |
| 197 | #define LDSH 0x0A |
| 198 | #define LDSHA 0x1A |
| 199 | #define LDFSR 0x21 |
| 200 | #define ST 0x04 |
| 201 | #define STA 0x14 |
| 202 | #define STB 0x05 |
| 203 | #define STBA 0x15 |
| 204 | #define STD 0x07 |
| 205 | #define STDA 0x17 |
| 206 | #define STF 0x24 |
| 207 | #define STDFQ 0x26 |
| 208 | #define STDF 0x27 |
| 209 | #define STFSR 0x25 |
| 210 | #define STH 0x06 |
| 211 | #define STHA 0x16 |
| 212 | #define SWAP 0x0F |
| 213 | #define SWAPA 0x1F |
| 214 | #define FLUSH 0x3B |
| 215 | |
| 216 | #define SIGN_BIT 0x80000000 |
| 217 | |
| 218 | /* # of cycles overhead when a trap is taken */ |
| 219 | #define TRAP_C 3 |
| 220 | |
| 221 | /* Forward declarations */ |
| 222 | |
| 223 | static uint32 sub_cc PARAMS ((uint32 psr, int32 operand1, int32 operand2, |
| 224 | int32 result)); |
| 225 | static uint32 add_cc PARAMS ((uint32 psr, int32 operand1, int32 operand2, |
| 226 | int32 result)); |
| 227 | static void log_cc PARAMS ((int32 result, struct pstate *sregs)); |
| 228 | static int fpexec PARAMS ((uint32 op3, uint32 rd, uint32 rs1, uint32 rs2, |
| 229 | struct pstate *sregs)); |
| 230 | static int chk_asi PARAMS ((struct pstate *sregs, uint32 *asi, uint32 op3)); |
| 231 | |
| 232 | |
| 233 | extern struct estate ebase; |
| 234 | extern int32 nfp,ift; |
| 235 | |
| 236 | #ifdef ERRINJ |
| 237 | extern uint32 errtt, errftt; |
| 238 | #endif |
| 239 | |
| 240 | static uint32 |
| 241 | sub_cc(psr, operand1, operand2, result) |
| 242 | uint32 psr; |
| 243 | int32 operand1; |
| 244 | int32 operand2; |
| 245 | int32 result; |
| 246 | { |
| 247 | psr = ((psr & ~PSR_N) | ((result >> 8) & PSR_N)); |
| 248 | if (result) |
| 249 | psr &= ~PSR_Z; |
| 250 | else |
| 251 | psr |= PSR_Z; |
| 252 | psr = (psr & ~PSR_V) | ((((operand1 & ~operand2 & ~result) | |
| 253 | (~operand1 & operand2 & result)) >> 10) & PSR_V); |
| 254 | psr = (psr & ~PSR_C) | ((((~operand1 & operand2) | |
| 255 | ((~operand1 | operand2) & result)) >> 11) & PSR_C); |
| 256 | return (psr); |
| 257 | } |
| 258 | |
| 259 | uint32 |
| 260 | add_cc(psr, operand1, operand2, result) |
| 261 | uint32 psr; |
| 262 | int32 operand1; |
| 263 | int32 operand2; |
| 264 | int32 result; |
| 265 | { |
| 266 | psr = ((psr & ~PSR_N) | ((result >> 8) & PSR_N)); |
| 267 | if (result) |
| 268 | psr &= ~PSR_Z; |
| 269 | else |
| 270 | psr |= PSR_Z; |
| 271 | psr = (psr & ~PSR_V) | ((((operand1 & operand2 & ~result) | |
| 272 | (~operand1 & ~operand2 & result)) >> 10) & PSR_V); |
| 273 | psr = (psr & ~PSR_C) | ((((operand1 & operand2) | |
| 274 | ((operand1 | operand2) & ~result)) >> 11) & PSR_C); |
| 275 | return(psr); |
| 276 | } |
| 277 | |
| 278 | static void |
| 279 | log_cc(result, sregs) |
| 280 | int32 result; |
| 281 | struct pstate *sregs; |
| 282 | { |
| 283 | sregs->psr &= ~(PSR_CC); /* Zero CC bits */ |
| 284 | sregs->psr = (sregs->psr | ((result >> 8) & PSR_N)); |
| 285 | if (result == 0) |
| 286 | sregs->psr |= PSR_Z; |
| 287 | } |
| 288 | |
| 289 | /* Add two unsigned 32-bit integers, and calculate the carry out. */ |
| 290 | |
| 291 | static uint32 |
| 292 | add32 (uint32 n1, uint32 n2, int *carry) |
| 293 | { |
| 294 | uint32 result = n1 + n2; |
| 295 | |
| 296 | *carry = result < n1 || result < n1; |
| 297 | return(result); |
| 298 | } |
| 299 | |
| 300 | /* Multiply two 32-bit integers. */ |
| 301 | |
| 302 | static void |
| 303 | mul64 (uint32 n1, uint32 n2, uint32 *result_hi, uint32 *result_lo, int msigned) |
| 304 | { |
| 305 | uint32 lo, mid1, mid2, hi, reg_lo, reg_hi; |
| 306 | int carry; |
| 307 | int sign = 0; |
| 308 | |
| 309 | /* If this is a signed multiply, calculate the sign of the result |
| 310 | and make the operands positive. */ |
| 311 | if (msigned) |
| 312 | { |
| 313 | sign = (n1 ^ n2) & SIGN_BIT; |
| 314 | if (n1 & SIGN_BIT) |
| 315 | n1 = -n1; |
| 316 | if (n2 & SIGN_BIT) |
| 317 | n2 = -n2; |
| 318 | |
| 319 | } |
| 320 | |
| 321 | /* We can split the 32x32 into four 16x16 operations. This ensures |
| 322 | that we do not lose precision on 32bit only hosts: */ |
| 323 | lo = ((n1 & 0xFFFF) * (n2 & 0xFFFF)); |
| 324 | mid1 = ((n1 & 0xFFFF) * ((n2 >> 16) & 0xFFFF)); |
| 325 | mid2 = (((n1 >> 16) & 0xFFFF) * (n2 & 0xFFFF)); |
| 326 | hi = (((n1 >> 16) & 0xFFFF) * ((n2 >> 16) & 0xFFFF)); |
| 327 | |
| 328 | /* We now need to add all of these results together, taking care |
| 329 | to propogate the carries from the additions: */ |
| 330 | reg_lo = add32 (lo, (mid1 << 16), &carry); |
| 331 | reg_hi = carry; |
| 332 | reg_lo = add32 (reg_lo, (mid2 << 16), &carry); |
| 333 | reg_hi += (carry + ((mid1 >> 16) & 0xFFFF) + ((mid2 >> 16) & 0xFFFF) + hi); |
| 334 | |
| 335 | /* Negate result if necessary. */ |
| 336 | if (sign) |
| 337 | { |
| 338 | reg_hi = ~ reg_hi; |
| 339 | reg_lo = - reg_lo; |
| 340 | if (reg_lo == 0) |
| 341 | reg_hi++; |
| 342 | } |
| 343 | |
| 344 | *result_lo = reg_lo; |
| 345 | *result_hi = reg_hi; |
| 346 | } |
| 347 | |
| 348 | |
| 349 | /* Divide a 64-bit integer by a 32-bit integer. We cheat and assume |
| 350 | that the host compiler supports long long operations. */ |
| 351 | |
| 352 | static void |
| 353 | div64 (uint32 n1_hi, uint32 n1_low, uint32 n2, uint32 *result, int msigned) |
| 354 | { |
| 355 | uint64 n1; |
| 356 | |
| 357 | n1 = ((uint64) n1_hi) << 32; |
| 358 | n1 |= ((uint64) n1_low) & 0xffffffff; |
| 359 | |
| 360 | if (msigned) |
| 361 | { |
| 362 | int64 n1_s = (int64) n1; |
| 363 | int32 n2_s = (int32) n2; |
| 364 | n1_s = n1_s / n2_s; |
| 365 | n1 = (uint64) n1_s; |
| 366 | } |
| 367 | else |
| 368 | n1 = n1 / n2; |
| 369 | |
| 370 | *result = (uint32) (n1 & 0xffffffff); |
| 371 | } |
| 372 | |
| 373 | |
| 374 | int |
| 375 | dispatch_instruction(sregs) |
| 376 | struct pstate *sregs; |
| 377 | { |
| 378 | |
| 379 | uint32 cwp, op, op2, op3, asi, rd, cond, rs1, |
| 380 | rs2; |
| 381 | uint32 ldep, icc; |
| 382 | int32 operand1, operand2, *rdd, result, eicc, |
| 383 | new_cwp; |
| 384 | int32 pc, npc, data, address, ws, mexc, fcc; |
| 385 | int32 ddata[2]; |
| 386 | |
| 387 | sregs->ninst++; |
| 388 | cwp = ((sregs->psr & PSR_CWP) << 4); |
| 389 | op = sregs->inst >> 30; |
| 390 | pc = sregs->npc; |
| 391 | npc = sregs->npc + 4; |
| 392 | op3 = rd = rs1 = operand2 = eicc = 0; |
| 393 | rdd = 0; |
| 394 | if (op & 2) { |
| 395 | |
| 396 | op3 = (sregs->inst >> 19) & 0x3f; |
| 397 | rs1 = (sregs->inst >> 14) & 0x1f; |
| 398 | rd = (sregs->inst >> 25) & 0x1f; |
| 399 | |
| 400 | #ifdef LOAD_DEL |
| 401 | |
| 402 | /* Check if load dependecy is possible */ |
| 403 | if (ebase.simtime <= sregs->ildtime) |
| 404 | ldep = (((op3 & 0x38) != 0x28) && ((op3 & 0x3e) != 0x34) && (sregs->ildreg != 0)); |
| 405 | else |
| 406 | ldep = 0; |
| 407 | if (sregs->inst & INST_I) { |
| 408 | if (ldep && (sregs->ildreg == rs1)) |
| 409 | sregs->hold++; |
| 410 | operand2 = sregs->inst; |
| 411 | operand2 = ((operand2 << 19) >> 19); /* sign extend */ |
| 412 | } else { |
| 413 | rs2 = sregs->inst & INST_RS2; |
| 414 | if (rs2 > 7) |
| 415 | operand2 = sregs->r[(cwp + rs2) & 0x7f]; |
| 416 | else |
| 417 | operand2 = sregs->g[rs2]; |
| 418 | if (ldep && ((sregs->ildreg == rs1) || (sregs->ildreg == rs2))) |
| 419 | sregs->hold++; |
| 420 | } |
| 421 | #else |
| 422 | if (sregs->inst & INST_I) { |
| 423 | operand2 = sregs->inst; |
| 424 | operand2 = ((operand2 << 19) >> 19); /* sign extend */ |
| 425 | } else { |
| 426 | rs2 = sregs->inst & INST_RS2; |
| 427 | if (rs2 > 7) |
| 428 | operand2 = sregs->r[(cwp + rs2) & 0x7f]; |
| 429 | else |
| 430 | operand2 = sregs->g[rs2]; |
| 431 | } |
| 432 | #endif |
| 433 | |
| 434 | if (rd > 7) |
| 435 | rdd = &(sregs->r[(cwp + rd) & 0x7f]); |
| 436 | else |
| 437 | rdd = &(sregs->g[rd]); |
| 438 | if (rs1 > 7) |
| 439 | rs1 = sregs->r[(cwp + rs1) & 0x7f]; |
| 440 | else |
| 441 | rs1 = sregs->g[rs1]; |
| 442 | } |
| 443 | switch (op) { |
| 444 | case 0: |
| 445 | op2 = (sregs->inst >> 22) & 0x7; |
| 446 | switch (op2) { |
| 447 | case SETHI: |
| 448 | rd = (sregs->inst >> 25) & 0x1f; |
| 449 | if (rd > 7) |
| 450 | rdd = &(sregs->r[(cwp + rd) & 0x7f]); |
| 451 | else |
| 452 | rdd = &(sregs->g[rd]); |
| 453 | *rdd = sregs->inst << 10; |
| 454 | break; |
| 455 | case BICC: |
| 456 | #ifdef STAT |
| 457 | sregs->nbranch++; |
| 458 | #endif |
| 459 | icc = sregs->psr >> 20; |
| 460 | cond = ((sregs->inst >> 25) & 0x0f); |
| 461 | switch (cond) { |
| 462 | case BICC_BN: |
| 463 | eicc = 0; |
| 464 | break; |
| 465 | case BICC_BE: |
| 466 | eicc = ICC_Z; |
| 467 | break; |
| 468 | case BICC_BLE: |
| 469 | eicc = ICC_Z | (ICC_N ^ ICC_V); |
| 470 | break; |
| 471 | case BICC_BL: |
| 472 | eicc = (ICC_N ^ ICC_V); |
| 473 | break; |
| 474 | case BICC_BLEU: |
| 475 | eicc = ICC_C | ICC_Z; |
| 476 | break; |
| 477 | case BICC_BCS: |
| 478 | eicc = ICC_C; |
| 479 | break; |
| 480 | case BICC_NEG: |
| 481 | eicc = ICC_N; |
| 482 | break; |
| 483 | case BICC_BVS: |
| 484 | eicc = ICC_V; |
| 485 | break; |
| 486 | case BICC_BA: |
| 487 | eicc = 1; |
| 488 | if (sregs->inst & 0x20000000) |
| 489 | sregs->annul = 1; |
| 490 | break; |
| 491 | case BICC_BNE: |
| 492 | eicc = ~(ICC_Z); |
| 493 | break; |
| 494 | case BICC_BG: |
| 495 | eicc = ~(ICC_Z | (ICC_N ^ ICC_V)); |
| 496 | break; |
| 497 | case BICC_BGE: |
| 498 | eicc = ~(ICC_N ^ ICC_V); |
| 499 | break; |
| 500 | case BICC_BGU: |
| 501 | eicc = ~(ICC_C | ICC_Z); |
| 502 | break; |
| 503 | case BICC_BCC: |
| 504 | eicc = ~(ICC_C); |
| 505 | break; |
| 506 | case BICC_POS: |
| 507 | eicc = ~(ICC_N); |
| 508 | break; |
| 509 | case BICC_BVC: |
| 510 | eicc = ~(ICC_V); |
| 511 | break; |
| 512 | } |
| 513 | if (eicc & 1) { |
| 514 | operand1 = sregs->inst; |
| 515 | operand1 = ((operand1 << 10) >> 8); /* sign extend */ |
| 516 | npc = sregs->pc + operand1; |
| 517 | } else { |
| 518 | if (sregs->inst & 0x20000000) |
| 519 | sregs->annul = 1; |
| 520 | } |
| 521 | break; |
| 522 | case FPBCC: |
| 523 | #ifdef STAT |
| 524 | sregs->nbranch++; |
| 525 | #endif |
| 526 | if (!((sregs->psr & PSR_EF) && FP_PRES)) { |
| 527 | sregs->trap = TRAP_FPDIS; |
| 528 | break; |
| 529 | } |
| 530 | if (ebase.simtime < sregs->ftime) { |
| 531 | sregs->ftime = ebase.simtime + sregs->hold; |
| 532 | } |
| 533 | cond = ((sregs->inst >> 25) & 0x0f); |
| 534 | fcc = (sregs->fsr >> 10) & 0x3; |
| 535 | switch (cond) { |
| 536 | case FBN: |
| 537 | eicc = 0; |
| 538 | break; |
| 539 | case FBNE: |
| 540 | eicc = (fcc != FCC_E); |
| 541 | break; |
| 542 | case FBLG: |
| 543 | eicc = (fcc == FCC_L) || (fcc == FCC_G); |
| 544 | break; |
| 545 | case FBUL: |
| 546 | eicc = (fcc == FCC_L) || (fcc == FCC_U); |
| 547 | break; |
| 548 | case FBL: |
| 549 | eicc = (fcc == FCC_L); |
| 550 | break; |
| 551 | case FBUG: |
| 552 | eicc = (fcc == FCC_G) || (fcc == FCC_U); |
| 553 | break; |
| 554 | case FBG: |
| 555 | eicc = (fcc == FCC_G); |
| 556 | break; |
| 557 | case FBU: |
| 558 | eicc = (fcc == FCC_U); |
| 559 | break; |
| 560 | case FBA: |
| 561 | eicc = 1; |
| 562 | if (sregs->inst & 0x20000000) |
| 563 | sregs->annul = 1; |
| 564 | break; |
| 565 | case FBE: |
| 566 | eicc = !(fcc != FCC_E); |
| 567 | break; |
| 568 | case FBUE: |
| 569 | eicc = !((fcc == FCC_L) || (fcc == FCC_G)); |
| 570 | break; |
| 571 | case FBGE: |
| 572 | eicc = !((fcc == FCC_L) || (fcc == FCC_U)); |
| 573 | break; |
| 574 | case FBUGE: |
| 575 | eicc = !(fcc == FCC_L); |
| 576 | break; |
| 577 | case FBLE: |
| 578 | eicc = !((fcc == FCC_G) || (fcc == FCC_U)); |
| 579 | break; |
| 580 | case FBULE: |
| 581 | eicc = !(fcc == FCC_G); |
| 582 | break; |
| 583 | case FBO: |
| 584 | eicc = !(fcc == FCC_U); |
| 585 | break; |
| 586 | } |
| 587 | if (eicc) { |
| 588 | operand1 = sregs->inst; |
| 589 | operand1 = ((operand1 << 10) >> 8); /* sign extend */ |
| 590 | npc = sregs->pc + operand1; |
| 591 | } else { |
| 592 | if (sregs->inst & 0x20000000) |
| 593 | sregs->annul = 1; |
| 594 | } |
| 595 | break; |
| 596 | |
| 597 | default: |
| 598 | sregs->trap = TRAP_UNIMP; |
| 599 | break; |
| 600 | } |
| 601 | break; |
| 602 | case 1: /* CALL */ |
| 603 | #ifdef STAT |
| 604 | sregs->nbranch++; |
| 605 | #endif |
| 606 | sregs->r[(cwp + 15) & 0x7f] = sregs->pc; |
| 607 | npc = sregs->pc + (sregs->inst << 2); |
| 608 | break; |
| 609 | |
| 610 | case 2: |
| 611 | if ((op3 >> 1) == 0x1a) { |
| 612 | if (!((sregs->psr & PSR_EF) && FP_PRES)) { |
| 613 | sregs->trap = TRAP_FPDIS; |
| 614 | } else { |
| 615 | rs1 = (sregs->inst >> 14) & 0x1f; |
| 616 | rs2 = sregs->inst & 0x1f; |
| 617 | sregs->trap = fpexec(op3, rd, rs1, rs2, sregs); |
| 618 | } |
| 619 | } else { |
| 620 | |
| 621 | switch (op3) { |
| 622 | case TICC: |
| 623 | icc = sregs->psr >> 20; |
| 624 | cond = ((sregs->inst >> 25) & 0x0f); |
| 625 | switch (cond) { |
| 626 | case BICC_BN: |
| 627 | eicc = 0; |
| 628 | break; |
| 629 | case BICC_BE: |
| 630 | eicc = ICC_Z; |
| 631 | break; |
| 632 | case BICC_BLE: |
| 633 | eicc = ICC_Z | (ICC_N ^ ICC_V); |
| 634 | break; |
| 635 | case BICC_BL: |
| 636 | eicc = (ICC_N ^ ICC_V); |
| 637 | break; |
| 638 | case BICC_BLEU: |
| 639 | eicc = ICC_C | ICC_Z; |
| 640 | break; |
| 641 | case BICC_BCS: |
| 642 | eicc = ICC_C; |
| 643 | break; |
| 644 | case BICC_NEG: |
| 645 | eicc = ICC_N; |
| 646 | break; |
| 647 | case BICC_BVS: |
| 648 | eicc = ICC_V; |
| 649 | break; |
| 650 | case BICC_BA: |
| 651 | eicc = 1; |
| 652 | break; |
| 653 | case BICC_BNE: |
| 654 | eicc = ~(ICC_Z); |
| 655 | break; |
| 656 | case BICC_BG: |
| 657 | eicc = ~(ICC_Z | (ICC_N ^ ICC_V)); |
| 658 | break; |
| 659 | case BICC_BGE: |
| 660 | eicc = ~(ICC_N ^ ICC_V); |
| 661 | break; |
| 662 | case BICC_BGU: |
| 663 | eicc = ~(ICC_C | ICC_Z); |
| 664 | break; |
| 665 | case BICC_BCC: |
| 666 | eicc = ~(ICC_C); |
| 667 | break; |
| 668 | case BICC_POS: |
| 669 | eicc = ~(ICC_N); |
| 670 | break; |
| 671 | case BICC_BVC: |
| 672 | eicc = ~(ICC_V); |
| 673 | break; |
| 674 | } |
| 675 | if (eicc & 1) { |
| 676 | sregs->trap = (0x80 | ((rs1 + operand2) & 0x7f)); |
| 677 | } |
| 678 | break; |
| 679 | |
| 680 | case MULScc: |
| 681 | operand1 = |
| 682 | (((sregs->psr & PSR_V) ^ ((sregs->psr & PSR_N) >> 2)) |
| 683 | << 10) | (rs1 >> 1); |
| 684 | if ((sregs->y & 1) == 0) |
| 685 | operand2 = 0; |
| 686 | *rdd = operand1 + operand2; |
| 687 | sregs->y = (rs1 << 31) | (sregs->y >> 1); |
| 688 | sregs->psr = add_cc(sregs->psr, operand1, operand2, *rdd); |
| 689 | break; |
| 690 | case DIVScc: |
| 691 | { |
| 692 | int sign; |
| 693 | uint32 result, remainder; |
| 694 | int c0, y31; |
| 695 | |
| 696 | if (!sparclite) { |
| 697 | sregs->trap = TRAP_UNIMP; |
| 698 | break; |
| 699 | } |
| 700 | |
| 701 | sign = ((sregs->psr & PSR_V) != 0) ^ ((sregs->psr & PSR_N) != 0); |
| 702 | |
| 703 | remainder = (sregs->y << 1) | (rs1 >> 31); |
| 704 | |
| 705 | /* If true sign is positive, calculate remainder - divisor. |
| 706 | Otherwise, calculate remainder + divisor. */ |
| 707 | if (sign == 0) |
| 708 | operand2 = ~operand2 + 1; |
| 709 | result = remainder + operand2; |
| 710 | |
| 711 | /* The SPARClite User's Manual is not clear on how |
| 712 | the "carry out" of the above ALU operation is to |
| 713 | be calculated. From trial and error tests |
| 714 | on the the chip itself, it appears that it is |
| 715 | a normal addition carry, and not a subtraction borrow, |
| 716 | even in cases where the divisor is subtracted |
| 717 | from the remainder. FIXME: get the true story |
| 718 | from Fujitsu. */ |
| 719 | c0 = result < (uint32) remainder |
| 720 | || result < (uint32) operand2; |
| 721 | |
| 722 | if (result & 0x80000000) |
| 723 | sregs->psr |= PSR_N; |
| 724 | else |
| 725 | sregs->psr &= ~PSR_N; |
| 726 | |
| 727 | y31 = (sregs->y & 0x80000000) == 0x80000000; |
| 728 | |
| 729 | if (result == 0 && sign == y31) |
| 730 | sregs->psr |= PSR_Z; |
| 731 | else |
| 732 | sregs->psr &= ~PSR_Z; |
| 733 | |
| 734 | sign = (sign && !y31) || (!c0 && (sign || !y31)); |
| 735 | |
| 736 | if (sign ^ (result >> 31)) |
| 737 | sregs->psr |= PSR_V; |
| 738 | else |
| 739 | sregs->psr &= ~PSR_V; |
| 740 | |
| 741 | if (!sign) |
| 742 | sregs->psr |= PSR_C; |
| 743 | else |
| 744 | sregs->psr &= ~PSR_C; |
| 745 | |
| 746 | sregs->y = result; |
| 747 | |
| 748 | if (rd != 0) |
| 749 | *rdd = (rs1 << 1) | !sign; |
| 750 | } |
| 751 | break; |
| 752 | case SMUL: |
| 753 | { |
| 754 | mul64 (rs1, operand2, &sregs->y, rdd, 1); |
| 755 | } |
| 756 | break; |
| 757 | case SMULCC: |
| 758 | { |
| 759 | uint32 result; |
| 760 | |
| 761 | mul64 (rs1, operand2, &sregs->y, &result, 1); |
| 762 | |
| 763 | if (result & 0x80000000) |
| 764 | sregs->psr |= PSR_N; |
| 765 | else |
| 766 | sregs->psr &= ~PSR_N; |
| 767 | |
| 768 | if (result == 0) |
| 769 | sregs->psr |= PSR_Z; |
| 770 | else |
| 771 | sregs->psr &= ~PSR_Z; |
| 772 | |
| 773 | *rdd = result; |
| 774 | } |
| 775 | break; |
| 776 | case UMUL: |
| 777 | { |
| 778 | mul64 (rs1, operand2, &sregs->y, rdd, 0); |
| 779 | } |
| 780 | break; |
| 781 | case UMULCC: |
| 782 | { |
| 783 | uint32 result; |
| 784 | |
| 785 | mul64 (rs1, operand2, &sregs->y, &result, 0); |
| 786 | |
| 787 | if (result & 0x80000000) |
| 788 | sregs->psr |= PSR_N; |
| 789 | else |
| 790 | sregs->psr &= ~PSR_N; |
| 791 | |
| 792 | if (result == 0) |
| 793 | sregs->psr |= PSR_Z; |
| 794 | else |
| 795 | sregs->psr &= ~PSR_Z; |
| 796 | |
| 797 | *rdd = result; |
| 798 | } |
| 799 | break; |
| 800 | case SDIV: |
| 801 | { |
| 802 | if (sparclite) { |
| 803 | sregs->trap = TRAP_UNIMP; |
| 804 | break; |
| 805 | } |
| 806 | |
| 807 | if (operand2 == 0) { |
| 808 | sregs->trap = TRAP_DIV0; |
| 809 | break; |
| 810 | } |
| 811 | |
| 812 | div64 (sregs->y, rs1, operand2, rdd, 1); |
| 813 | } |
| 814 | break; |
| 815 | case SDIVCC: |
| 816 | { |
| 817 | uint32 result; |
| 818 | |
| 819 | if (sparclite) { |
| 820 | sregs->trap = TRAP_UNIMP; |
| 821 | break; |
| 822 | } |
| 823 | |
| 824 | if (operand2 == 0) { |
| 825 | sregs->trap = TRAP_DIV0; |
| 826 | break; |
| 827 | } |
| 828 | |
| 829 | div64 (sregs->y, rs1, operand2, &result, 1); |
| 830 | |
| 831 | if (result & 0x80000000) |
| 832 | sregs->psr |= PSR_N; |
| 833 | else |
| 834 | sregs->psr &= ~PSR_N; |
| 835 | |
| 836 | if (result == 0) |
| 837 | sregs->psr |= PSR_Z; |
| 838 | else |
| 839 | sregs->psr &= ~PSR_Z; |
| 840 | |
| 841 | /* FIXME: should set overflow flag correctly. */ |
| 842 | sregs->psr &= ~(PSR_C | PSR_V); |
| 843 | |
| 844 | *rdd = result; |
| 845 | } |
| 846 | break; |
| 847 | case UDIV: |
| 848 | { |
| 849 | if (sparclite) { |
| 850 | sregs->trap = TRAP_UNIMP; |
| 851 | break; |
| 852 | } |
| 853 | |
| 854 | if (operand2 == 0) { |
| 855 | sregs->trap = TRAP_DIV0; |
| 856 | break; |
| 857 | } |
| 858 | |
| 859 | div64 (sregs->y, rs1, operand2, rdd, 0); |
| 860 | } |
| 861 | break; |
| 862 | case UDIVCC: |
| 863 | { |
| 864 | uint32 result; |
| 865 | |
| 866 | if (sparclite) { |
| 867 | sregs->trap = TRAP_UNIMP; |
| 868 | break; |
| 869 | } |
| 870 | |
| 871 | if (operand2 == 0) { |
| 872 | sregs->trap = TRAP_DIV0; |
| 873 | break; |
| 874 | } |
| 875 | |
| 876 | div64 (sregs->y, rs1, operand2, &result, 0); |
| 877 | |
| 878 | if (result & 0x80000000) |
| 879 | sregs->psr |= PSR_N; |
| 880 | else |
| 881 | sregs->psr &= ~PSR_N; |
| 882 | |
| 883 | if (result == 0) |
| 884 | sregs->psr |= PSR_Z; |
| 885 | else |
| 886 | sregs->psr &= ~PSR_Z; |
| 887 | |
| 888 | /* FIXME: should set overflow flag correctly. */ |
| 889 | sregs->psr &= ~(PSR_C | PSR_V); |
| 890 | |
| 891 | *rdd = result; |
| 892 | } |
| 893 | break; |
| 894 | case IXNOR: |
| 895 | *rdd = rs1 ^ ~operand2; |
| 896 | break; |
| 897 | case IXNORCC: |
| 898 | *rdd = rs1 ^ ~operand2; |
| 899 | log_cc(*rdd, sregs); |
| 900 | break; |
| 901 | case IXOR: |
| 902 | *rdd = rs1 ^ operand2; |
| 903 | break; |
| 904 | case IXORCC: |
| 905 | *rdd = rs1 ^ operand2; |
| 906 | log_cc(*rdd, sregs); |
| 907 | break; |
| 908 | case IOR: |
| 909 | *rdd = rs1 | operand2; |
| 910 | break; |
| 911 | case IORCC: |
| 912 | *rdd = rs1 | operand2; |
| 913 | log_cc(*rdd, sregs); |
| 914 | break; |
| 915 | case IORN: |
| 916 | *rdd = rs1 | ~operand2; |
| 917 | break; |
| 918 | case IORNCC: |
| 919 | *rdd = rs1 | ~operand2; |
| 920 | log_cc(*rdd, sregs); |
| 921 | break; |
| 922 | case IANDNCC: |
| 923 | *rdd = rs1 & ~operand2; |
| 924 | log_cc(*rdd, sregs); |
| 925 | break; |
| 926 | case IANDN: |
| 927 | *rdd = rs1 & ~operand2; |
| 928 | break; |
| 929 | case IAND: |
| 930 | *rdd = rs1 & operand2; |
| 931 | break; |
| 932 | case IANDCC: |
| 933 | *rdd = rs1 & operand2; |
| 934 | log_cc(*rdd, sregs); |
| 935 | break; |
| 936 | case SUB: |
| 937 | *rdd = rs1 - operand2; |
| 938 | break; |
| 939 | case SUBCC: |
| 940 | *rdd = rs1 - operand2; |
| 941 | sregs->psr = sub_cc(sregs->psr, rs1, operand2, *rdd); |
| 942 | break; |
| 943 | case SUBX: |
| 944 | *rdd = rs1 - operand2 - ((sregs->psr >> 20) & 1); |
| 945 | break; |
| 946 | case SUBXCC: |
| 947 | *rdd = rs1 - operand2 - ((sregs->psr >> 20) & 1); |
| 948 | sregs->psr = sub_cc(sregs->psr, rs1, operand2, *rdd); |
| 949 | break; |
| 950 | case ADD: |
| 951 | *rdd = rs1 + operand2; |
| 952 | break; |
| 953 | case ADDCC: |
| 954 | *rdd = rs1 + operand2; |
| 955 | sregs->psr = add_cc(sregs->psr, rs1, operand2, *rdd); |
| 956 | break; |
| 957 | case ADDX: |
| 958 | *rdd = rs1 + operand2 + ((sregs->psr >> 20) & 1); |
| 959 | break; |
| 960 | case ADDXCC: |
| 961 | *rdd = rs1 + operand2 + ((sregs->psr >> 20) & 1); |
| 962 | sregs->psr = add_cc(sregs->psr, rs1, operand2, *rdd); |
| 963 | break; |
| 964 | case TADDCC: |
| 965 | *rdd = rs1 + operand2; |
| 966 | sregs->psr = add_cc(sregs->psr, rs1, operand2, *rdd); |
| 967 | if ((rs1 | operand2) & 0x3) |
| 968 | sregs->psr |= PSR_V; |
| 969 | break; |
| 970 | case TSUBCC: |
| 971 | *rdd = rs1 - operand2; |
| 972 | sregs->psr = sub_cc (sregs->psr, rs1, operand2, *rdd); |
| 973 | if ((rs1 | operand2) & 0x3) |
| 974 | sregs->psr |= PSR_V; |
| 975 | break; |
| 976 | case TADDCCTV: |
| 977 | *rdd = rs1 + operand2; |
| 978 | result = add_cc(0, rs1, operand2, *rdd); |
| 979 | if ((rs1 | operand2) & 0x3) |
| 980 | result |= PSR_V; |
| 981 | if (result & PSR_V) { |
| 982 | sregs->trap = TRAP_TAG; |
| 983 | } else { |
| 984 | sregs->psr = (sregs->psr & ~PSR_CC) | result; |
| 985 | } |
| 986 | break; |
| 987 | case TSUBCCTV: |
| 988 | *rdd = rs1 - operand2; |
| 989 | result = add_cc (0, rs1, operand2, *rdd); |
| 990 | if ((rs1 | operand2) & 0x3) |
| 991 | result |= PSR_V; |
| 992 | if (result & PSR_V) |
| 993 | { |
| 994 | sregs->trap = TRAP_TAG; |
| 995 | } |
| 996 | else |
| 997 | { |
| 998 | sregs->psr = (sregs->psr & ~PSR_CC) | result; |
| 999 | } |
| 1000 | break; |
| 1001 | case SLL: |
| 1002 | *rdd = rs1 << (operand2 & 0x1f); |
| 1003 | break; |
| 1004 | case SRL: |
| 1005 | *rdd = rs1 >> (operand2 & 0x1f); |
| 1006 | break; |
| 1007 | case SRA: |
| 1008 | *rdd = ((int) rs1) >> (operand2 & 0x1f); |
| 1009 | break; |
| 1010 | case FLUSH: |
| 1011 | if (ift) sregs->trap = TRAP_UNIMP; |
| 1012 | break; |
| 1013 | case SAVE: |
| 1014 | new_cwp = ((sregs->psr & PSR_CWP) - 1) & PSR_CWP; |
| 1015 | if (sregs->wim & (1 << new_cwp)) { |
| 1016 | sregs->trap = TRAP_WOFL; |
| 1017 | break; |
| 1018 | } |
| 1019 | if (rd > 7) |
| 1020 | rdd = &(sregs->r[((new_cwp << 4) + rd) & 0x7f]); |
| 1021 | *rdd = rs1 + operand2; |
| 1022 | sregs->psr = (sregs->psr & ~PSR_CWP) | new_cwp; |
| 1023 | break; |
| 1024 | case RESTORE: |
| 1025 | |
| 1026 | new_cwp = ((sregs->psr & PSR_CWP) + 1) & PSR_CWP; |
| 1027 | if (sregs->wim & (1 << new_cwp)) { |
| 1028 | sregs->trap = TRAP_WUFL; |
| 1029 | break; |
| 1030 | } |
| 1031 | if (rd > 7) |
| 1032 | rdd = &(sregs->r[((new_cwp << 4) + rd) & 0x7f]); |
| 1033 | *rdd = rs1 + operand2; |
| 1034 | sregs->psr = (sregs->psr & ~PSR_CWP) | new_cwp; |
| 1035 | break; |
| 1036 | case RDPSR: |
| 1037 | if (!(sregs->psr & PSR_S)) { |
| 1038 | sregs->trap = TRAP_PRIVI; |
| 1039 | break; |
| 1040 | } |
| 1041 | *rdd = sregs->psr; |
| 1042 | break; |
| 1043 | case RDY: |
| 1044 | if (!sparclite) |
| 1045 | *rdd = sregs->y; |
| 1046 | else { |
| 1047 | int rs1_is_asr = (sregs->inst >> 14) & 0x1f; |
| 1048 | if ( 0 == rs1_is_asr ) |
| 1049 | *rdd = sregs->y; |
| 1050 | else if ( 17 == rs1_is_asr ) |
| 1051 | *rdd = sregs->asr17; |
| 1052 | else { |
| 1053 | sregs->trap = TRAP_UNIMP; |
| 1054 | break; |
| 1055 | } |
| 1056 | } |
| 1057 | break; |
| 1058 | case RDWIM: |
| 1059 | if (!(sregs->psr & PSR_S)) { |
| 1060 | sregs->trap = TRAP_PRIVI; |
| 1061 | break; |
| 1062 | } |
| 1063 | *rdd = sregs->wim; |
| 1064 | break; |
| 1065 | case RDTBR: |
| 1066 | if (!(sregs->psr & PSR_S)) { |
| 1067 | sregs->trap = TRAP_PRIVI; |
| 1068 | break; |
| 1069 | } |
| 1070 | *rdd = sregs->tbr; |
| 1071 | break; |
| 1072 | case WRPSR: |
| 1073 | if ((sregs->psr & 0x1f) > 7) { |
| 1074 | sregs->trap = TRAP_UNIMP; |
| 1075 | break; |
| 1076 | } |
| 1077 | if (!(sregs->psr & PSR_S)) { |
| 1078 | sregs->trap = TRAP_PRIVI; |
| 1079 | break; |
| 1080 | } |
| 1081 | sregs->psr = (rs1 ^ operand2) & 0x00f03fff; |
| 1082 | break; |
| 1083 | case WRWIM: |
| 1084 | if (!(sregs->psr & PSR_S)) { |
| 1085 | sregs->trap = TRAP_PRIVI; |
| 1086 | break; |
| 1087 | } |
| 1088 | sregs->wim = (rs1 ^ operand2) & 0x0ff; |
| 1089 | break; |
| 1090 | case WRTBR: |
| 1091 | if (!(sregs->psr & PSR_S)) { |
| 1092 | sregs->trap = TRAP_PRIVI; |
| 1093 | break; |
| 1094 | } |
| 1095 | sregs->tbr = (sregs->tbr & 0x00000ff0) | |
| 1096 | ((rs1 ^ operand2) & 0xfffff000); |
| 1097 | break; |
| 1098 | case WRY: |
| 1099 | if (!sparclite) |
| 1100 | sregs->y = (rs1 ^ operand2); |
| 1101 | else { |
| 1102 | if ( 0 == rd ) |
| 1103 | sregs->y = (rs1 ^ operand2); |
| 1104 | else if ( 17 == rd ) |
| 1105 | sregs->asr17 = (rs1 ^ operand2); |
| 1106 | else { |
| 1107 | sregs->trap = TRAP_UNIMP; |
| 1108 | break; |
| 1109 | } |
| 1110 | } |
| 1111 | break; |
| 1112 | case JMPL: |
| 1113 | |
| 1114 | #ifdef STAT |
| 1115 | sregs->nbranch++; |
| 1116 | #endif |
| 1117 | sregs->icnt = T_JMPL; /* JMPL takes two cycles */ |
| 1118 | if (rs1 & 0x3) { |
| 1119 | sregs->trap = TRAP_UNALI; |
| 1120 | break; |
| 1121 | } |
| 1122 | *rdd = sregs->pc; |
| 1123 | npc = rs1 + operand2; |
| 1124 | break; |
| 1125 | case RETT: |
| 1126 | address = rs1 + operand2; |
| 1127 | new_cwp = ((sregs->psr & PSR_CWP) + 1) & PSR_CWP; |
| 1128 | sregs->icnt = T_RETT; /* RETT takes two cycles */ |
| 1129 | if (sregs->psr & PSR_ET) { |
| 1130 | sregs->trap = TRAP_UNIMP; |
| 1131 | break; |
| 1132 | } |
| 1133 | if (!(sregs->psr & PSR_S)) { |
| 1134 | sregs->trap = TRAP_PRIVI; |
| 1135 | break; |
| 1136 | } |
| 1137 | if (sregs->wim & (1 << new_cwp)) { |
| 1138 | sregs->trap = TRAP_WUFL; |
| 1139 | break; |
| 1140 | } |
| 1141 | if (address & 0x3) { |
| 1142 | sregs->trap = TRAP_UNALI; |
| 1143 | break; |
| 1144 | } |
| 1145 | sregs->psr = (sregs->psr & ~PSR_CWP) | new_cwp | PSR_ET; |
| 1146 | sregs->psr = |
| 1147 | (sregs->psr & ~PSR_S) | ((sregs->psr & PSR_PS) << 1); |
| 1148 | npc = address; |
| 1149 | break; |
| 1150 | |
| 1151 | case SCAN: |
| 1152 | { |
| 1153 | uint32 result, mask; |
| 1154 | int i; |
| 1155 | |
| 1156 | if (!sparclite) { |
| 1157 | sregs->trap = TRAP_UNIMP; |
| 1158 | break; |
| 1159 | } |
| 1160 | mask = (operand2 & 0x80000000) | (operand2 >> 1); |
| 1161 | result = rs1 ^ mask; |
| 1162 | |
| 1163 | for (i = 0; i < 32; i++) { |
| 1164 | if (result & 0x80000000) |
| 1165 | break; |
| 1166 | result <<= 1; |
| 1167 | } |
| 1168 | |
| 1169 | *rdd = i == 32 ? 63 : i; |
| 1170 | } |
| 1171 | break; |
| 1172 | |
| 1173 | default: |
| 1174 | sregs->trap = TRAP_UNIMP; |
| 1175 | break; |
| 1176 | } |
| 1177 | } |
| 1178 | break; |
| 1179 | case 3: /* Load/store instructions */ |
| 1180 | |
| 1181 | address = rs1 + operand2; |
| 1182 | |
| 1183 | if (sregs->psr & PSR_S) |
| 1184 | asi = 11; |
| 1185 | else |
| 1186 | asi = 10; |
| 1187 | |
| 1188 | if (op3 & 4) { |
| 1189 | sregs->icnt = T_ST; /* Set store instruction count */ |
| 1190 | #ifdef STAT |
| 1191 | sregs->nstore++; |
| 1192 | #endif |
| 1193 | } else { |
| 1194 | sregs->icnt = T_LD; /* Set load instruction count */ |
| 1195 | #ifdef STAT |
| 1196 | sregs->nload++; |
| 1197 | #endif |
| 1198 | } |
| 1199 | |
| 1200 | /* Decode load/store instructions */ |
| 1201 | |
| 1202 | switch (op3) { |
| 1203 | case LDDA: |
| 1204 | if (!chk_asi(sregs, &asi, op3)) break; |
| 1205 | case LDD: |
| 1206 | if (address & 0x7) { |
| 1207 | sregs->trap = TRAP_UNALI; |
| 1208 | break; |
| 1209 | } |
| 1210 | if (rd & 1) { |
| 1211 | rd &= 0x1e; |
| 1212 | if (rd > 7) |
| 1213 | rdd = &(sregs->r[(cwp + rd) & 0x7f]); |
| 1214 | else |
| 1215 | rdd = &(sregs->g[rd]); |
| 1216 | } |
| 1217 | mexc = memory_read(asi, address, ddata, 3, &ws); |
| 1218 | sregs->hold += ws * 2; |
| 1219 | sregs->icnt = T_LDD; |
| 1220 | if (mexc) { |
| 1221 | sregs->trap = TRAP_DEXC; |
| 1222 | } else { |
| 1223 | rdd[0] = ddata[0]; |
| 1224 | rdd[1] = ddata[1]; |
| 1225 | #ifdef STAT |
| 1226 | sregs->nload++; /* Double load counts twice */ |
| 1227 | #endif |
| 1228 | } |
| 1229 | break; |
| 1230 | |
| 1231 | case LDA: |
| 1232 | if (!chk_asi(sregs, &asi, op3)) break; |
| 1233 | case LD: |
| 1234 | if (address & 0x3) { |
| 1235 | sregs->trap = TRAP_UNALI; |
| 1236 | break; |
| 1237 | } |
| 1238 | mexc = memory_read(asi, address, &data, 2, &ws); |
| 1239 | sregs->hold += ws; |
| 1240 | if (mexc) { |
| 1241 | sregs->trap = TRAP_DEXC; |
| 1242 | } else { |
| 1243 | *rdd = data; |
| 1244 | } |
| 1245 | break; |
| 1246 | case LDSTUBA: |
| 1247 | if (!chk_asi(sregs, &asi, op3)) break; |
| 1248 | case LDSTUB: |
| 1249 | mexc = memory_read(asi, address, &data, 0, &ws); |
| 1250 | sregs->hold += ws; |
| 1251 | sregs->icnt = T_LDST; |
| 1252 | if (mexc) { |
| 1253 | sregs->trap = TRAP_DEXC; |
| 1254 | break; |
| 1255 | } |
| 1256 | *rdd = data; |
| 1257 | data = 0x0ff; |
| 1258 | mexc = memory_write(asi, address, &data, 0, &ws); |
| 1259 | sregs->hold += ws; |
| 1260 | if (mexc) { |
| 1261 | sregs->trap = TRAP_DEXC; |
| 1262 | } |
| 1263 | #ifdef STAT |
| 1264 | sregs->nload++; |
| 1265 | #endif |
| 1266 | break; |
| 1267 | case LDSBA: |
| 1268 | case LDUBA: |
| 1269 | if (!chk_asi(sregs, &asi, op3)) break; |
| 1270 | case LDSB: |
| 1271 | case LDUB: |
| 1272 | mexc = memory_read(asi, address, &data, 0, &ws); |
| 1273 | sregs->hold += ws; |
| 1274 | if (mexc) { |
| 1275 | sregs->trap = TRAP_DEXC; |
| 1276 | break; |
| 1277 | } |
| 1278 | if ((op3 == LDSB) && (data & 0x80)) |
| 1279 | data |= 0xffffff00; |
| 1280 | *rdd = data; |
| 1281 | break; |
| 1282 | case LDSHA: |
| 1283 | case LDUHA: |
| 1284 | if (!chk_asi(sregs, &asi, op3)) break; |
| 1285 | case LDSH: |
| 1286 | case LDUH: |
| 1287 | if (address & 0x1) { |
| 1288 | sregs->trap = TRAP_UNALI; |
| 1289 | break; |
| 1290 | } |
| 1291 | mexc = memory_read(asi, address, &data, 1, &ws); |
| 1292 | sregs->hold += ws; |
| 1293 | if (mexc) { |
| 1294 | sregs->trap = TRAP_DEXC; |
| 1295 | break; |
| 1296 | } |
| 1297 | if ((op3 == LDSH) && (data & 0x8000)) |
| 1298 | data |= 0xffff0000; |
| 1299 | *rdd = data; |
| 1300 | break; |
| 1301 | case LDF: |
| 1302 | if (!((sregs->psr & PSR_EF) && FP_PRES)) { |
| 1303 | sregs->trap = TRAP_FPDIS; |
| 1304 | break; |
| 1305 | } |
| 1306 | if (address & 0x3) { |
| 1307 | sregs->trap = TRAP_UNALI; |
| 1308 | break; |
| 1309 | } |
| 1310 | if (ebase.simtime < sregs->ftime) { |
| 1311 | if ((sregs->frd == rd) || (sregs->frs1 == rd) || |
| 1312 | (sregs->frs2 == rd)) |
| 1313 | sregs->fhold += (sregs->ftime - ebase.simtime); |
| 1314 | } |
| 1315 | mexc = memory_read(asi, address, &data, 2, &ws); |
| 1316 | sregs->hold += ws; |
| 1317 | sregs->flrd = rd; |
| 1318 | sregs->ltime = ebase.simtime + sregs->icnt + FLSTHOLD + |
| 1319 | sregs->hold + sregs->fhold; |
| 1320 | if (mexc) { |
| 1321 | sregs->trap = TRAP_DEXC; |
| 1322 | } else { |
| 1323 | sregs->fs[rd] = *((float32 *) & data); |
| 1324 | } |
| 1325 | break; |
| 1326 | case LDDF: |
| 1327 | if (!((sregs->psr & PSR_EF) && FP_PRES)) { |
| 1328 | sregs->trap = TRAP_FPDIS; |
| 1329 | break; |
| 1330 | } |
| 1331 | if (address & 0x7) { |
| 1332 | sregs->trap = TRAP_UNALI; |
| 1333 | break; |
| 1334 | } |
| 1335 | if (ebase.simtime < sregs->ftime) { |
| 1336 | if (((sregs->frd >> 1) == (rd >> 1)) || |
| 1337 | ((sregs->frs1 >> 1) == (rd >> 1)) || |
| 1338 | ((sregs->frs2 >> 1) == (rd >> 1))) |
| 1339 | sregs->fhold += (sregs->ftime - ebase.simtime); |
| 1340 | } |
| 1341 | mexc = memory_read(asi, address, ddata, 3, &ws); |
| 1342 | sregs->hold += ws * 2; |
| 1343 | sregs->icnt = T_LDD; |
| 1344 | if (mexc) { |
| 1345 | sregs->trap = TRAP_DEXC; |
| 1346 | } else { |
| 1347 | rd &= 0x1E; |
| 1348 | sregs->flrd = rd; |
| 1349 | sregs->fs[rd] = *((float32 *) & ddata[0]); |
| 1350 | #ifdef STAT |
| 1351 | sregs->nload++; /* Double load counts twice */ |
| 1352 | #endif |
| 1353 | sregs->fs[rd + 1] = *((float32 *) & ddata[1]); |
| 1354 | sregs->ltime = ebase.simtime + sregs->icnt + FLSTHOLD + |
| 1355 | sregs->hold + sregs->fhold; |
| 1356 | } |
| 1357 | break; |
| 1358 | case LDFSR: |
| 1359 | if (ebase.simtime < sregs->ftime) { |
| 1360 | sregs->fhold += (sregs->ftime - ebase.simtime); |
| 1361 | } |
| 1362 | if (!((sregs->psr & PSR_EF) && FP_PRES)) { |
| 1363 | sregs->trap = TRAP_FPDIS; |
| 1364 | break; |
| 1365 | } |
| 1366 | if (address & 0x3) { |
| 1367 | sregs->trap = TRAP_UNALI; |
| 1368 | break; |
| 1369 | } |
| 1370 | mexc = memory_read(asi, address, &data, 2, &ws); |
| 1371 | sregs->hold += ws; |
| 1372 | if (mexc) { |
| 1373 | sregs->trap = TRAP_DEXC; |
| 1374 | } else { |
| 1375 | sregs->fsr = |
| 1376 | (sregs->fsr & 0x7FF000) | (data & ~0x7FF000); |
| 1377 | set_fsr(sregs->fsr); |
| 1378 | } |
| 1379 | break; |
| 1380 | case STFSR: |
| 1381 | if (!((sregs->psr & PSR_EF) && FP_PRES)) { |
| 1382 | sregs->trap = TRAP_FPDIS; |
| 1383 | break; |
| 1384 | } |
| 1385 | if (address & 0x3) { |
| 1386 | sregs->trap = TRAP_UNALI; |
| 1387 | break; |
| 1388 | } |
| 1389 | if (ebase.simtime < sregs->ftime) { |
| 1390 | sregs->fhold += (sregs->ftime - ebase.simtime); |
| 1391 | } |
| 1392 | mexc = memory_write(asi, address, &sregs->fsr, 2, &ws); |
| 1393 | sregs->hold += ws; |
| 1394 | if (mexc) { |
| 1395 | sregs->trap = TRAP_DEXC; |
| 1396 | } |
| 1397 | break; |
| 1398 | |
| 1399 | case STA: |
| 1400 | if (!chk_asi(sregs, &asi, op3)) break; |
| 1401 | case ST: |
| 1402 | if (address & 0x3) { |
| 1403 | sregs->trap = TRAP_UNALI; |
| 1404 | break; |
| 1405 | } |
| 1406 | mexc = memory_write(asi, address, rdd, 2, &ws); |
| 1407 | sregs->hold += ws; |
| 1408 | if (mexc) { |
| 1409 | sregs->trap = TRAP_DEXC; |
| 1410 | } |
| 1411 | break; |
| 1412 | case STBA: |
| 1413 | if (!chk_asi(sregs, &asi, op3)) break; |
| 1414 | case STB: |
| 1415 | mexc = memory_write(asi, address, rdd, 0, &ws); |
| 1416 | sregs->hold += ws; |
| 1417 | if (mexc) { |
| 1418 | sregs->trap = TRAP_DEXC; |
| 1419 | } |
| 1420 | break; |
| 1421 | case STDA: |
| 1422 | if (!chk_asi(sregs, &asi, op3)) break; |
| 1423 | case STD: |
| 1424 | if (address & 0x7) { |
| 1425 | sregs->trap = TRAP_UNALI; |
| 1426 | break; |
| 1427 | } |
| 1428 | if (rd & 1) { |
| 1429 | rd &= 0x1e; |
| 1430 | if (rd > 7) |
| 1431 | rdd = &(sregs->r[(cwp + rd) & 0x7f]); |
| 1432 | else |
| 1433 | rdd = &(sregs->g[rd]); |
| 1434 | } |
| 1435 | mexc = memory_write(asi, address, rdd, 3, &ws); |
| 1436 | sregs->hold += ws; |
| 1437 | sregs->icnt = T_STD; |
| 1438 | #ifdef STAT |
| 1439 | sregs->nstore++; /* Double store counts twice */ |
| 1440 | #endif |
| 1441 | if (mexc) { |
| 1442 | sregs->trap = TRAP_DEXC; |
| 1443 | break; |
| 1444 | } |
| 1445 | break; |
| 1446 | case STDFQ: |
| 1447 | if ((sregs->psr & 0x1f) > 7) { |
| 1448 | sregs->trap = TRAP_UNIMP; |
| 1449 | break; |
| 1450 | } |
| 1451 | if (!((sregs->psr & PSR_EF) && FP_PRES)) { |
| 1452 | sregs->trap = TRAP_FPDIS; |
| 1453 | break; |
| 1454 | } |
| 1455 | if (address & 0x7) { |
| 1456 | sregs->trap = TRAP_UNALI; |
| 1457 | break; |
| 1458 | } |
| 1459 | if (!(sregs->fsr & FSR_QNE)) { |
| 1460 | sregs->fsr = (sregs->fsr & ~FSR_TT) | FP_SEQ_ERR; |
| 1461 | break; |
| 1462 | } |
| 1463 | rdd = &(sregs->fpq[0]); |
| 1464 | mexc = memory_write(asi, address, rdd, 3, &ws); |
| 1465 | sregs->hold += ws; |
| 1466 | sregs->icnt = T_STD; |
| 1467 | #ifdef STAT |
| 1468 | sregs->nstore++; /* Double store counts twice */ |
| 1469 | #endif |
| 1470 | if (mexc) { |
| 1471 | sregs->trap = TRAP_DEXC; |
| 1472 | break; |
| 1473 | } else { |
| 1474 | sregs->fsr &= ~FSR_QNE; |
| 1475 | sregs->fpstate = FP_EXE_MODE; |
| 1476 | } |
| 1477 | break; |
| 1478 | case STHA: |
| 1479 | if (!chk_asi(sregs, &asi, op3)) break; |
| 1480 | case STH: |
| 1481 | if (address & 0x1) { |
| 1482 | sregs->trap = TRAP_UNALI; |
| 1483 | break; |
| 1484 | } |
| 1485 | mexc = memory_write(asi, address, rdd, 1, &ws); |
| 1486 | sregs->hold += ws; |
| 1487 | if (mexc) { |
| 1488 | sregs->trap = TRAP_DEXC; |
| 1489 | } |
| 1490 | break; |
| 1491 | case STF: |
| 1492 | if (!((sregs->psr & PSR_EF) && FP_PRES)) { |
| 1493 | sregs->trap = TRAP_FPDIS; |
| 1494 | break; |
| 1495 | } |
| 1496 | if (address & 0x3) { |
| 1497 | sregs->trap = TRAP_UNALI; |
| 1498 | break; |
| 1499 | } |
| 1500 | if (ebase.simtime < sregs->ftime) { |
| 1501 | if (sregs->frd == rd) |
| 1502 | sregs->fhold += (sregs->ftime - ebase.simtime); |
| 1503 | } |
| 1504 | mexc = memory_write(asi, address, &sregs->fsi[rd], 2, &ws); |
| 1505 | sregs->hold += ws; |
| 1506 | if (mexc) { |
| 1507 | sregs->trap = TRAP_DEXC; |
| 1508 | } |
| 1509 | break; |
| 1510 | case STDF: |
| 1511 | if (!((sregs->psr & PSR_EF) && FP_PRES)) { |
| 1512 | sregs->trap = TRAP_FPDIS; |
| 1513 | break; |
| 1514 | } |
| 1515 | if (address & 0x7) { |
| 1516 | sregs->trap = TRAP_UNALI; |
| 1517 | break; |
| 1518 | } |
| 1519 | rd &= 0x1E; |
| 1520 | if (ebase.simtime < sregs->ftime) { |
| 1521 | if ((sregs->frd == rd) || (sregs->frd + 1 == rd)) |
| 1522 | sregs->fhold += (sregs->ftime - ebase.simtime); |
| 1523 | } |
| 1524 | mexc = memory_write(asi, address, &sregs->fsi[rd], 3, &ws); |
| 1525 | sregs->hold += ws; |
| 1526 | sregs->icnt = T_STD; |
| 1527 | #ifdef STAT |
| 1528 | sregs->nstore++; /* Double store counts twice */ |
| 1529 | #endif |
| 1530 | if (mexc) { |
| 1531 | sregs->trap = TRAP_DEXC; |
| 1532 | } |
| 1533 | break; |
| 1534 | case SWAPA: |
| 1535 | if (!chk_asi(sregs, &asi, op3)) break; |
| 1536 | case SWAP: |
| 1537 | if (address & 0x3) { |
| 1538 | sregs->trap = TRAP_UNALI; |
| 1539 | break; |
| 1540 | } |
| 1541 | mexc = memory_read(asi, address, &data, 2, &ws); |
| 1542 | sregs->hold += ws; |
| 1543 | if (mexc) { |
| 1544 | sregs->trap = TRAP_DEXC; |
| 1545 | break; |
| 1546 | } |
| 1547 | mexc = memory_write(asi, address, rdd, 2, &ws); |
| 1548 | sregs->hold += ws; |
| 1549 | sregs->icnt = T_LDST; |
| 1550 | if (mexc) { |
| 1551 | sregs->trap = TRAP_DEXC; |
| 1552 | break; |
| 1553 | } else |
| 1554 | *rdd = data; |
| 1555 | #ifdef STAT |
| 1556 | sregs->nload++; |
| 1557 | #endif |
| 1558 | break; |
| 1559 | |
| 1560 | |
| 1561 | default: |
| 1562 | sregs->trap = TRAP_UNIMP; |
| 1563 | break; |
| 1564 | } |
| 1565 | |
| 1566 | #ifdef LOAD_DEL |
| 1567 | |
| 1568 | if (!(op3 & 4)) { |
| 1569 | sregs->ildtime = ebase.simtime + sregs->hold + sregs->icnt; |
| 1570 | sregs->ildreg = rd; |
| 1571 | if ((op3 | 0x10) == 0x13) |
| 1572 | sregs->ildreg |= 1; /* Double load, odd register loaded |
| 1573 | * last */ |
| 1574 | } |
| 1575 | #endif |
| 1576 | break; |
| 1577 | |
| 1578 | default: |
| 1579 | sregs->trap = TRAP_UNIMP; |
| 1580 | break; |
| 1581 | } |
| 1582 | sregs->g[0] = 0; |
| 1583 | if (!sregs->trap) { |
| 1584 | sregs->pc = pc; |
| 1585 | sregs->npc = npc; |
| 1586 | } |
| 1587 | return (0); |
| 1588 | } |
| 1589 | |
| 1590 | #define T_FABSs 2 |
| 1591 | #define T_FADDs 4 |
| 1592 | #define T_FADDd 4 |
| 1593 | #define T_FCMPs 4 |
| 1594 | #define T_FCMPd 4 |
| 1595 | #define T_FDIVs 20 |
| 1596 | #define T_FDIVd 35 |
| 1597 | #define T_FMOVs 2 |
| 1598 | #define T_FMULs 5 |
| 1599 | #define T_FMULd 9 |
| 1600 | #define T_FNEGs 2 |
| 1601 | #define T_FSQRTs 37 |
| 1602 | #define T_FSQRTd 65 |
| 1603 | #define T_FSUBs 4 |
| 1604 | #define T_FSUBd 4 |
| 1605 | #define T_FdTOi 7 |
| 1606 | #define T_FdTOs 3 |
| 1607 | #define T_FiTOs 6 |
| 1608 | #define T_FiTOd 6 |
| 1609 | #define T_FsTOi 6 |
| 1610 | #define T_FsTOd 2 |
| 1611 | |
| 1612 | #define FABSs 0x09 |
| 1613 | #define FADDs 0x41 |
| 1614 | #define FADDd 0x42 |
| 1615 | #define FCMPs 0x51 |
| 1616 | #define FCMPd 0x52 |
| 1617 | #define FCMPEs 0x55 |
| 1618 | #define FCMPEd 0x56 |
| 1619 | #define FDIVs 0x4D |
| 1620 | #define FDIVd 0x4E |
| 1621 | #define FMOVs 0x01 |
| 1622 | #define FMULs 0x49 |
| 1623 | #define FMULd 0x4A |
| 1624 | #define FNEGs 0x05 |
| 1625 | #define FSQRTs 0x29 |
| 1626 | #define FSQRTd 0x2A |
| 1627 | #define FSUBs 0x45 |
| 1628 | #define FSUBd 0x46 |
| 1629 | #define FdTOi 0xD2 |
| 1630 | #define FdTOs 0xC6 |
| 1631 | #define FiTOs 0xC4 |
| 1632 | #define FiTOd 0xC8 |
| 1633 | #define FsTOi 0xD1 |
| 1634 | #define FsTOd 0xC9 |
| 1635 | |
| 1636 | |
| 1637 | static int |
| 1638 | fpexec(op3, rd, rs1, rs2, sregs) |
| 1639 | uint32 op3, rd, rs1, rs2; |
| 1640 | struct pstate *sregs; |
| 1641 | { |
| 1642 | uint32 opf, tem, accex; |
| 1643 | int32 fcc; |
| 1644 | uint32 ldadj; |
| 1645 | |
| 1646 | if (sregs->fpstate == FP_EXC_MODE) { |
| 1647 | sregs->fsr = (sregs->fsr & ~FSR_TT) | FP_SEQ_ERR; |
| 1648 | sregs->fpstate = FP_EXC_PE; |
| 1649 | return (0); |
| 1650 | } |
| 1651 | if (sregs->fpstate == FP_EXC_PE) { |
| 1652 | sregs->fpstate = FP_EXC_MODE; |
| 1653 | return (TRAP_FPEXC); |
| 1654 | } |
| 1655 | opf = (sregs->inst >> 5) & 0x1ff; |
| 1656 | |
| 1657 | /* |
| 1658 | * Check if we already have an FPop in the pipe. If so, halt until it is |
| 1659 | * finished by incrementing fhold with the remaining execution time |
| 1660 | */ |
| 1661 | |
| 1662 | if (ebase.simtime < sregs->ftime) { |
| 1663 | sregs->fhold = (sregs->ftime - ebase.simtime); |
| 1664 | } else { |
| 1665 | sregs->fhold = 0; |
| 1666 | |
| 1667 | /* Check load dependencies. */ |
| 1668 | |
| 1669 | if (ebase.simtime < sregs->ltime) { |
| 1670 | |
| 1671 | /* Don't check rs1 if single operand instructions */ |
| 1672 | |
| 1673 | if (((opf >> 6) == 0) || ((opf >> 6) == 3)) |
| 1674 | rs1 = 32; |
| 1675 | |
| 1676 | /* Adjust for double floats */ |
| 1677 | |
| 1678 | ldadj = opf & 1; |
| 1679 | if (!(((sregs->flrd - rs1) >> ldadj) && ((sregs->flrd - rs2) >> ldadj))) |
| 1680 | sregs->fhold++; |
| 1681 | } |
| 1682 | } |
| 1683 | |
| 1684 | sregs->finst++; |
| 1685 | |
| 1686 | sregs->frs1 = rs1; /* Store src and dst for dependecy check */ |
| 1687 | sregs->frs2 = rs2; |
| 1688 | sregs->frd = rd; |
| 1689 | |
| 1690 | sregs->ftime = ebase.simtime + sregs->hold + sregs->fhold; |
| 1691 | |
| 1692 | /* SPARC is big-endian - swap double floats if host is little-endian */ |
| 1693 | /* This is ugly - I know ... */ |
| 1694 | |
| 1695 | /* FIXME: should use (CURRENT_HOST_BYTE_ORDER == CURRENT_TARGET_BYTE_ORDER) |
| 1696 | but what about machines where float values are different endianness |
| 1697 | from integer values? */ |
| 1698 | |
| 1699 | #ifdef HOST_LITTLE_ENDIAN_FLOAT |
| 1700 | rs1 &= 0x1f; |
| 1701 | switch (opf) { |
| 1702 | case FADDd: |
| 1703 | case FDIVd: |
| 1704 | case FMULd: |
| 1705 | case FSQRTd: |
| 1706 | case FSUBd: |
| 1707 | case FCMPd: |
| 1708 | case FCMPEd: |
| 1709 | case FdTOi: |
| 1710 | case FdTOs: |
| 1711 | sregs->fdp[rs1 | 1] = sregs->fs[rs1 & ~1]; |
| 1712 | sregs->fdp[rs1 & ~1] = sregs->fs[rs1 | 1]; |
| 1713 | sregs->fdp[rs2 | 1] = sregs->fs[rs2 & ~1]; |
| 1714 | sregs->fdp[rs2 & ~1] = sregs->fs[rs2 | 1]; |
| 1715 | default: |
| 1716 | } |
| 1717 | #endif |
| 1718 | |
| 1719 | clear_accex(); |
| 1720 | |
| 1721 | switch (opf) { |
| 1722 | case FABSs: |
| 1723 | sregs->fs[rd] = fabs(sregs->fs[rs2]); |
| 1724 | sregs->ftime += T_FABSs; |
| 1725 | sregs->frs1 = 32; /* rs1 ignored */ |
| 1726 | break; |
| 1727 | case FADDs: |
| 1728 | sregs->fs[rd] = sregs->fs[rs1] + sregs->fs[rs2]; |
| 1729 | sregs->ftime += T_FADDs; |
| 1730 | break; |
| 1731 | case FADDd: |
| 1732 | sregs->fd[rd >> 1] = sregs->fd[rs1 >> 1] + sregs->fd[rs2 >> 1]; |
| 1733 | sregs->ftime += T_FADDd; |
| 1734 | break; |
| 1735 | case FCMPs: |
| 1736 | case FCMPEs: |
| 1737 | if (sregs->fs[rs1] == sregs->fs[rs2]) |
| 1738 | fcc = 3; |
| 1739 | else if (sregs->fs[rs1] < sregs->fs[rs2]) |
| 1740 | fcc = 2; |
| 1741 | else if (sregs->fs[rs1] > sregs->fs[rs2]) |
| 1742 | fcc = 1; |
| 1743 | else |
| 1744 | fcc = 0; |
| 1745 | sregs->fsr |= 0x0C00; |
| 1746 | sregs->fsr &= ~(fcc << 10); |
| 1747 | sregs->ftime += T_FCMPs; |
| 1748 | sregs->frd = 32; /* rd ignored */ |
| 1749 | if ((fcc == 0) && (opf == FCMPEs)) { |
| 1750 | sregs->fpstate = FP_EXC_PE; |
| 1751 | sregs->fsr = (sregs->fsr & ~0x1C000) | (1 << 14); |
| 1752 | } |
| 1753 | break; |
| 1754 | case FCMPd: |
| 1755 | case FCMPEd: |
| 1756 | if (sregs->fd[rs1 >> 1] == sregs->fd[rs2 >> 1]) |
| 1757 | fcc = 3; |
| 1758 | else if (sregs->fd[rs1 >> 1] < sregs->fd[rs2 >> 1]) |
| 1759 | fcc = 2; |
| 1760 | else if (sregs->fd[rs1 >> 1] > sregs->fd[rs2 >> 1]) |
| 1761 | fcc = 1; |
| 1762 | else |
| 1763 | fcc = 0; |
| 1764 | sregs->fsr |= 0x0C00; |
| 1765 | sregs->fsr &= ~(fcc << 10); |
| 1766 | sregs->ftime += T_FCMPd; |
| 1767 | sregs->frd = 32; /* rd ignored */ |
| 1768 | if ((fcc == 0) && (opf == FCMPEd)) { |
| 1769 | sregs->fpstate = FP_EXC_PE; |
| 1770 | sregs->fsr = (sregs->fsr & ~FSR_TT) | FP_IEEE; |
| 1771 | } |
| 1772 | break; |
| 1773 | case FDIVs: |
| 1774 | sregs->fs[rd] = sregs->fs[rs1] / sregs->fs[rs2]; |
| 1775 | sregs->ftime += T_FDIVs; |
| 1776 | break; |
| 1777 | case FDIVd: |
| 1778 | sregs->fd[rd >> 1] = sregs->fd[rs1 >> 1] / sregs->fd[rs2 >> 1]; |
| 1779 | sregs->ftime += T_FDIVd; |
| 1780 | break; |
| 1781 | case FMOVs: |
| 1782 | sregs->fs[rd] = sregs->fs[rs2]; |
| 1783 | sregs->ftime += T_FMOVs; |
| 1784 | sregs->frs1 = 32; /* rs1 ignored */ |
| 1785 | break; |
| 1786 | case FMULs: |
| 1787 | sregs->fs[rd] = sregs->fs[rs1] * sregs->fs[rs2]; |
| 1788 | sregs->ftime += T_FMULs; |
| 1789 | break; |
| 1790 | case FMULd: |
| 1791 | sregs->fd[rd >> 1] = sregs->fd[rs1 >> 1] * sregs->fd[rs2 >> 1]; |
| 1792 | sregs->ftime += T_FMULd; |
| 1793 | break; |
| 1794 | case FNEGs: |
| 1795 | sregs->fs[rd] = -sregs->fs[rs2]; |
| 1796 | sregs->ftime += T_FNEGs; |
| 1797 | sregs->frs1 = 32; /* rs1 ignored */ |
| 1798 | break; |
| 1799 | case FSQRTs: |
| 1800 | if (sregs->fs[rs2] < 0.0) { |
| 1801 | sregs->fpstate = FP_EXC_PE; |
| 1802 | sregs->fsr = (sregs->fsr & ~FSR_TT) | FP_IEEE; |
| 1803 | sregs->fsr = (sregs->fsr & 0x1f) | 0x10; |
| 1804 | break; |
| 1805 | } |
| 1806 | sregs->fs[rd] = sqrt(sregs->fs[rs2]); |
| 1807 | sregs->ftime += T_FSQRTs; |
| 1808 | sregs->frs1 = 32; /* rs1 ignored */ |
| 1809 | break; |
| 1810 | case FSQRTd: |
| 1811 | if (sregs->fd[rs2 >> 1] < 0.0) { |
| 1812 | sregs->fpstate = FP_EXC_PE; |
| 1813 | sregs->fsr = (sregs->fsr & ~FSR_TT) | FP_IEEE; |
| 1814 | sregs->fsr = (sregs->fsr & 0x1f) | 0x10; |
| 1815 | break; |
| 1816 | } |
| 1817 | sregs->fd[rd >> 1] = sqrt(sregs->fd[rs2 >> 1]); |
| 1818 | sregs->ftime += T_FSQRTd; |
| 1819 | sregs->frs1 = 32; /* rs1 ignored */ |
| 1820 | break; |
| 1821 | case FSUBs: |
| 1822 | sregs->fs[rd] = sregs->fs[rs1] - sregs->fs[rs2]; |
| 1823 | sregs->ftime += T_FSUBs; |
| 1824 | break; |
| 1825 | case FSUBd: |
| 1826 | sregs->fd[rd >> 1] = sregs->fd[rs1 >> 1] - sregs->fd[rs2 >> 1]; |
| 1827 | sregs->ftime += T_FSUBd; |
| 1828 | break; |
| 1829 | case FdTOi: |
| 1830 | sregs->fsi[rd] = (int) sregs->fd[rs2 >> 1]; |
| 1831 | sregs->ftime += T_FdTOi; |
| 1832 | sregs->frs1 = 32; /* rs1 ignored */ |
| 1833 | break; |
| 1834 | case FdTOs: |
| 1835 | sregs->fs[rd] = (float32) sregs->fd[rs2 >> 1]; |
| 1836 | sregs->ftime += T_FdTOs; |
| 1837 | sregs->frs1 = 32; /* rs1 ignored */ |
| 1838 | break; |
| 1839 | case FiTOs: |
| 1840 | sregs->fs[rd] = (float32) sregs->fsi[rs2]; |
| 1841 | sregs->ftime += T_FiTOs; |
| 1842 | sregs->frs1 = 32; /* rs1 ignored */ |
| 1843 | break; |
| 1844 | case FiTOd: |
| 1845 | sregs->fd[rd >> 1] = (float64) sregs->fsi[rs2]; |
| 1846 | sregs->ftime += T_FiTOd; |
| 1847 | sregs->frs1 = 32; /* rs1 ignored */ |
| 1848 | break; |
| 1849 | case FsTOi: |
| 1850 | sregs->fsi[rd] = (int) sregs->fs[rs2]; |
| 1851 | sregs->ftime += T_FsTOi; |
| 1852 | sregs->frs1 = 32; /* rs1 ignored */ |
| 1853 | break; |
| 1854 | case FsTOd: |
| 1855 | sregs->fd[rd >> 1] = sregs->fs[rs2]; |
| 1856 | sregs->ftime += T_FsTOd; |
| 1857 | sregs->frs1 = 32; /* rs1 ignored */ |
| 1858 | break; |
| 1859 | |
| 1860 | default: |
| 1861 | sregs->fsr = (sregs->fsr & ~FSR_TT) | FP_UNIMP; |
| 1862 | sregs->fpstate = FP_EXC_PE; |
| 1863 | } |
| 1864 | |
| 1865 | #ifdef ERRINJ |
| 1866 | if (errftt) { |
| 1867 | sregs->fsr = (sregs->fsr & ~FSR_TT) | (errftt << 14); |
| 1868 | sregs->fpstate = FP_EXC_PE; |
| 1869 | if (sis_verbose) printf("Inserted fpu error %X\n",errftt); |
| 1870 | errftt = 0; |
| 1871 | } |
| 1872 | #endif |
| 1873 | |
| 1874 | accex = get_accex(); |
| 1875 | |
| 1876 | #ifdef HOST_LITTLE_ENDIAN_FLOAT |
| 1877 | switch (opf) { |
| 1878 | case FADDd: |
| 1879 | case FDIVd: |
| 1880 | case FMULd: |
| 1881 | case FSQRTd: |
| 1882 | case FSUBd: |
| 1883 | case FiTOd: |
| 1884 | case FsTOd: |
| 1885 | sregs->fs[rd & ~1] = sregs->fdp[rd | 1]; |
| 1886 | sregs->fs[rd | 1] = sregs->fdp[rd & ~1]; |
| 1887 | default: |
| 1888 | } |
| 1889 | #endif |
| 1890 | if (sregs->fpstate == FP_EXC_PE) { |
| 1891 | sregs->fpq[0] = sregs->pc; |
| 1892 | sregs->fpq[1] = sregs->inst; |
| 1893 | sregs->fsr |= FSR_QNE; |
| 1894 | } else { |
| 1895 | tem = (sregs->fsr >> 23) & 0x1f; |
| 1896 | if (tem & accex) { |
| 1897 | sregs->fpstate = FP_EXC_PE; |
| 1898 | sregs->fsr = (sregs->fsr & ~FSR_TT) | FP_IEEE; |
| 1899 | sregs->fsr = ((sregs->fsr & ~0x1f) | accex); |
| 1900 | } else { |
| 1901 | sregs->fsr = ((((sregs->fsr >> 5) | accex) << 5) | accex); |
| 1902 | } |
| 1903 | if (sregs->fpstate == FP_EXC_PE) { |
| 1904 | sregs->fpq[0] = sregs->pc; |
| 1905 | sregs->fpq[1] = sregs->inst; |
| 1906 | sregs->fsr |= FSR_QNE; |
| 1907 | } |
| 1908 | } |
| 1909 | clear_accex(); |
| 1910 | |
| 1911 | return (0); |
| 1912 | |
| 1913 | |
| 1914 | } |
| 1915 | |
| 1916 | static int |
| 1917 | chk_asi(sregs, asi, op3) |
| 1918 | struct pstate *sregs; |
| 1919 | uint32 *asi, op3; |
| 1920 | |
| 1921 | { |
| 1922 | if (!(sregs->psr & PSR_S)) { |
| 1923 | sregs->trap = TRAP_PRIVI; |
| 1924 | return (0); |
| 1925 | } else if (sregs->inst & INST_I) { |
| 1926 | sregs->trap = TRAP_UNIMP; |
| 1927 | return (0); |
| 1928 | } else |
| 1929 | *asi = (sregs->inst >> 5) & 0x0ff; |
| 1930 | return(1); |
| 1931 | } |
| 1932 | |
| 1933 | int |
| 1934 | execute_trap(sregs) |
| 1935 | struct pstate *sregs; |
| 1936 | { |
| 1937 | int32 cwp; |
| 1938 | |
| 1939 | if (sregs->trap == 256) { |
| 1940 | sregs->pc = 0; |
| 1941 | sregs->npc = 4; |
| 1942 | sregs->trap = 0; |
| 1943 | } else if (sregs->trap == 257) { |
| 1944 | return (ERROR); |
| 1945 | } else { |
| 1946 | |
| 1947 | if ((sregs->psr & PSR_ET) == 0) |
| 1948 | return (ERROR); |
| 1949 | |
| 1950 | sregs->tbr = (sregs->tbr & 0xfffff000) | (sregs->trap << 4); |
| 1951 | sregs->trap = 0; |
| 1952 | sregs->psr &= ~PSR_ET; |
| 1953 | sregs->psr |= ((sregs->psr & PSR_S) >> 1); |
| 1954 | sregs->annul = 0; |
| 1955 | sregs->psr = (((sregs->psr & PSR_CWP) - 1) & 0x7) | (sregs->psr & ~PSR_CWP); |
| 1956 | cwp = ((sregs->psr & PSR_CWP) << 4); |
| 1957 | sregs->r[(cwp + 17) & 0x7f] = sregs->pc; |
| 1958 | sregs->r[(cwp + 18) & 0x7f] = sregs->npc; |
| 1959 | sregs->psr |= PSR_S; |
| 1960 | sregs->pc = sregs->tbr; |
| 1961 | sregs->npc = sregs->tbr + 4; |
| 1962 | |
| 1963 | if ( 0 != (1 & sregs->asr17) ) { |
| 1964 | /* single vector trapping! */ |
| 1965 | sregs->pc = sregs->tbr & 0xfffff000; |
| 1966 | sregs->npc = sregs->pc + 4; |
| 1967 | } |
| 1968 | |
| 1969 | /* Increase simulator time */ |
| 1970 | sregs->icnt = TRAP_C; |
| 1971 | |
| 1972 | } |
| 1973 | |
| 1974 | |
| 1975 | return (0); |
| 1976 | |
| 1977 | } |
| 1978 | |
| 1979 | extern struct irqcell irqarr[16]; |
| 1980 | |
| 1981 | int |
| 1982 | check_interrupts(sregs) |
| 1983 | struct pstate *sregs; |
| 1984 | { |
| 1985 | #ifdef ERRINJ |
| 1986 | if (errtt) { |
| 1987 | sregs->trap = errtt; |
| 1988 | if (sis_verbose) printf("Inserted error trap 0x%02X\n",errtt); |
| 1989 | errtt = 0; |
| 1990 | } |
| 1991 | #endif |
| 1992 | |
| 1993 | if ((ext_irl) && (sregs->psr & PSR_ET) && |
| 1994 | ((ext_irl == 15) || (ext_irl > (int) ((sregs->psr & PSR_PIL) >> 8)))) { |
| 1995 | if (sregs->trap == 0) { |
| 1996 | sregs->trap = 16 + ext_irl; |
| 1997 | irqarr[ext_irl & 0x0f].callback(irqarr[ext_irl & 0x0f].arg); |
| 1998 | return(1); |
| 1999 | } |
| 2000 | } |
| 2001 | return(0); |
| 2002 | } |
| 2003 | |
| 2004 | void |
| 2005 | init_regs(sregs) |
| 2006 | struct pstate *sregs; |
| 2007 | { |
| 2008 | sregs->pc = 0; |
| 2009 | sregs->npc = 4; |
| 2010 | sregs->trap = 0; |
| 2011 | sregs->psr &= 0x00f03fdf; |
| 2012 | sregs->psr |= 0x080; /* Set supervisor bit */ |
| 2013 | sregs->breakpoint = 0; |
| 2014 | sregs->annul = 0; |
| 2015 | sregs->fpstate = FP_EXE_MODE; |
| 2016 | sregs->fpqn = 0; |
| 2017 | sregs->ftime = 0; |
| 2018 | sregs->ltime = 0; |
| 2019 | sregs->err_mode = 0; |
| 2020 | ext_irl = 0; |
| 2021 | sregs->g[0] = 0; |
| 2022 | #ifdef HOST_LITTLE_ENDIAN_FLOAT |
| 2023 | sregs->fdp = (float32 *) sregs->fd; |
| 2024 | sregs->fsi = (int32 *) sregs->fs; |
| 2025 | #else |
| 2026 | sregs->fs = (float32 *) sregs->fd; |
| 2027 | sregs->fsi = (int32 *) sregs->fd; |
| 2028 | #endif |
| 2029 | sregs->fsr = 0; |
| 2030 | sregs->fpu_pres = !nfp; |
| 2031 | set_fsr(sregs->fsr); |
| 2032 | sregs->bphit = 0; |
| 2033 | sregs->ildreg = 0; |
| 2034 | sregs->ildtime = 0; |
| 2035 | |
| 2036 | sregs->y = 0; |
| 2037 | sregs->asr17 = 0; |
| 2038 | |
| 2039 | sregs->rett_err = 0; |
| 2040 | sregs->jmpltime = 0; |
| 2041 | } |