1 /* 2 * Optimizations for Tiny Code Generator for QEMU 3 * 4 * Copyright (c) 2010 Samsung Electronics. 5 * Contributed by Kirill Batuzov <batuzovk@ispras.ru> 6 * 7 * Permission is hereby granted, free of charge, to any person obtaining a copy 8 * of this software and associated documentation files (the "Software"), to deal 9 * in the Software without restriction, including without limitation the rights 10 * to use, copy, modify, merge, publish, distribute, sublicense, and/or sell 11 * copies of the Software, and to permit persons to whom the Software is 12 * furnished to do so, subject to the following conditions: 13 * 14 * The above copyright notice and this permission notice shall be included in 15 * all copies or substantial portions of the Software. 16 * 17 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR 18 * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, 19 * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL 20 * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER 21 * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, 22 * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN 23 * THE SOFTWARE. 24 */ 25 26 #include "qemu/osdep.h" 27 #include "qemu/int128.h" 28 #include "qemu/interval-tree.h" 29 #include "tcg/tcg-op-common.h" 30 #include "tcg-internal.h" 31 32 #define CASE_OP_32_64(x) \ 33 glue(glue(case INDEX_op_, x), _i32): \ 34 glue(glue(case INDEX_op_, x), _i64) 35 36 #define CASE_OP_32_64_VEC(x) \ 37 glue(glue(case INDEX_op_, x), _i32): \ 38 glue(glue(case INDEX_op_, x), _i64): \ 39 glue(glue(case INDEX_op_, x), _vec) 40 41 typedef struct MemCopyInfo { 42 IntervalTreeNode itree; 43 QSIMPLEQ_ENTRY (MemCopyInfo) next; 44 TCGTemp *ts; 45 TCGType type; 46 } MemCopyInfo; 47 48 typedef struct TempOptInfo { 49 bool is_const; 50 TCGTemp *prev_copy; 51 TCGTemp *next_copy; 52 QSIMPLEQ_HEAD(, MemCopyInfo) mem_copy; 53 uint64_t val; 54 uint64_t z_mask; /* mask bit is 0 if and only if value bit is 0 */ 55 uint64_t s_mask; /* a left-aligned mask of clrsb(value) bits. */ 56 } TempOptInfo; 57 58 typedef struct OptContext { 59 TCGContext *tcg; 60 TCGOp *prev_mb; 61 TCGTempSet temps_used; 62 63 IntervalTreeRoot mem_copy; 64 QSIMPLEQ_HEAD(, MemCopyInfo) mem_free; 65 66 /* In flight values from optimization. */ 67 uint64_t a_mask; /* mask bit is 0 iff value identical to first input */ 68 uint64_t z_mask; /* mask bit is 0 iff value bit is 0 */ 69 uint64_t s_mask; /* mask of clrsb(value) bits */ 70 TCGType type; 71 } OptContext; 72 73 /* Calculate the smask for a specific value. */ 74 static uint64_t smask_from_value(uint64_t value) 75 { 76 int rep = clrsb64(value); 77 return ~(~0ull >> rep); 78 } 79 80 /* 81 * Calculate the smask for a given set of known-zeros. 82 * If there are lots of zeros on the left, we can consider the remainder 83 * an unsigned field, and thus the corresponding signed field is one bit 84 * larger. 85 */ 86 static uint64_t smask_from_zmask(uint64_t zmask) 87 { 88 /* 89 * Only the 0 bits are significant for zmask, thus the msb itself 90 * must be zero, else we have no sign information. 91 */ 92 int rep = clz64(zmask); 93 if (rep == 0) { 94 return 0; 95 } 96 rep -= 1; 97 return ~(~0ull >> rep); 98 } 99 100 /* 101 * Recreate a properly left-aligned smask after manipulation. 102 * Some bit-shuffling, particularly shifts and rotates, may 103 * retain sign bits on the left, but may scatter disconnected 104 * sign bits on the right. Retain only what remains to the left. 105 */ 106 static uint64_t smask_from_smask(int64_t smask) 107 { 108 /* Only the 1 bits are significant for smask */ 109 return smask_from_zmask(~smask); 110 } 111 112 static inline TempOptInfo *ts_info(TCGTemp *ts) 113 { 114 return ts->state_ptr; 115 } 116 117 static inline TempOptInfo *arg_info(TCGArg arg) 118 { 119 return ts_info(arg_temp(arg)); 120 } 121 122 static inline bool ts_is_const(TCGTemp *ts) 123 { 124 return ts_info(ts)->is_const; 125 } 126 127 static inline bool ts_is_const_val(TCGTemp *ts, uint64_t val) 128 { 129 TempOptInfo *ti = ts_info(ts); 130 return ti->is_const && ti->val == val; 131 } 132 133 static inline bool arg_is_const(TCGArg arg) 134 { 135 return ts_is_const(arg_temp(arg)); 136 } 137 138 static inline bool arg_is_const_val(TCGArg arg, uint64_t val) 139 { 140 return ts_is_const_val(arg_temp(arg), val); 141 } 142 143 static inline bool ts_is_copy(TCGTemp *ts) 144 { 145 return ts_info(ts)->next_copy != ts; 146 } 147 148 static TCGTemp *cmp_better_copy(TCGTemp *a, TCGTemp *b) 149 { 150 return a->kind < b->kind ? b : a; 151 } 152 153 /* Initialize and activate a temporary. */ 154 static void init_ts_info(OptContext *ctx, TCGTemp *ts) 155 { 156 size_t idx = temp_idx(ts); 157 TempOptInfo *ti; 158 159 if (test_bit(idx, ctx->temps_used.l)) { 160 return; 161 } 162 set_bit(idx, ctx->temps_used.l); 163 164 ti = ts->state_ptr; 165 if (ti == NULL) { 166 ti = tcg_malloc(sizeof(TempOptInfo)); 167 ts->state_ptr = ti; 168 } 169 170 ti->next_copy = ts; 171 ti->prev_copy = ts; 172 QSIMPLEQ_INIT(&ti->mem_copy); 173 if (ts->kind == TEMP_CONST) { 174 ti->is_const = true; 175 ti->val = ts->val; 176 ti->z_mask = ts->val; 177 ti->s_mask = smask_from_value(ts->val); 178 } else { 179 ti->is_const = false; 180 ti->z_mask = -1; 181 ti->s_mask = 0; 182 } 183 } 184 185 static MemCopyInfo *mem_copy_first(OptContext *ctx, intptr_t s, intptr_t l) 186 { 187 IntervalTreeNode *r = interval_tree_iter_first(&ctx->mem_copy, s, l); 188 return r ? container_of(r, MemCopyInfo, itree) : NULL; 189 } 190 191 static MemCopyInfo *mem_copy_next(MemCopyInfo *mem, intptr_t s, intptr_t l) 192 { 193 IntervalTreeNode *r = interval_tree_iter_next(&mem->itree, s, l); 194 return r ? container_of(r, MemCopyInfo, itree) : NULL; 195 } 196 197 static void remove_mem_copy(OptContext *ctx, MemCopyInfo *mc) 198 { 199 TCGTemp *ts = mc->ts; 200 TempOptInfo *ti = ts_info(ts); 201 202 interval_tree_remove(&mc->itree, &ctx->mem_copy); 203 QSIMPLEQ_REMOVE(&ti->mem_copy, mc, MemCopyInfo, next); 204 QSIMPLEQ_INSERT_TAIL(&ctx->mem_free, mc, next); 205 } 206 207 static void remove_mem_copy_in(OptContext *ctx, intptr_t s, intptr_t l) 208 { 209 while (true) { 210 MemCopyInfo *mc = mem_copy_first(ctx, s, l); 211 if (!mc) { 212 break; 213 } 214 remove_mem_copy(ctx, mc); 215 } 216 } 217 218 static void remove_mem_copy_all(OptContext *ctx) 219 { 220 remove_mem_copy_in(ctx, 0, -1); 221 tcg_debug_assert(interval_tree_is_empty(&ctx->mem_copy)); 222 } 223 224 static TCGTemp *find_better_copy(TCGTemp *ts) 225 { 226 TCGTemp *i, *ret; 227 228 /* If this is already readonly, we can't do better. */ 229 if (temp_readonly(ts)) { 230 return ts; 231 } 232 233 ret = ts; 234 for (i = ts_info(ts)->next_copy; i != ts; i = ts_info(i)->next_copy) { 235 ret = cmp_better_copy(ret, i); 236 } 237 return ret; 238 } 239 240 static void move_mem_copies(TCGTemp *dst_ts, TCGTemp *src_ts) 241 { 242 TempOptInfo *si = ts_info(src_ts); 243 TempOptInfo *di = ts_info(dst_ts); 244 MemCopyInfo *mc; 245 246 QSIMPLEQ_FOREACH(mc, &si->mem_copy, next) { 247 tcg_debug_assert(mc->ts == src_ts); 248 mc->ts = dst_ts; 249 } 250 QSIMPLEQ_CONCAT(&di->mem_copy, &si->mem_copy); 251 } 252 253 /* Reset TEMP's state, possibly removing the temp for the list of copies. */ 254 static void reset_ts(OptContext *ctx, TCGTemp *ts) 255 { 256 TempOptInfo *ti = ts_info(ts); 257 TCGTemp *pts = ti->prev_copy; 258 TCGTemp *nts = ti->next_copy; 259 TempOptInfo *pi = ts_info(pts); 260 TempOptInfo *ni = ts_info(nts); 261 262 ni->prev_copy = ti->prev_copy; 263 pi->next_copy = ti->next_copy; 264 ti->next_copy = ts; 265 ti->prev_copy = ts; 266 ti->is_const = false; 267 ti->z_mask = -1; 268 ti->s_mask = 0; 269 270 if (!QSIMPLEQ_EMPTY(&ti->mem_copy)) { 271 if (ts == nts) { 272 /* Last temp copy being removed, the mem copies die. */ 273 MemCopyInfo *mc; 274 QSIMPLEQ_FOREACH(mc, &ti->mem_copy, next) { 275 interval_tree_remove(&mc->itree, &ctx->mem_copy); 276 } 277 QSIMPLEQ_CONCAT(&ctx->mem_free, &ti->mem_copy); 278 } else { 279 move_mem_copies(find_better_copy(nts), ts); 280 } 281 } 282 } 283 284 static void reset_temp(OptContext *ctx, TCGArg arg) 285 { 286 reset_ts(ctx, arg_temp(arg)); 287 } 288 289 static void record_mem_copy(OptContext *ctx, TCGType type, 290 TCGTemp *ts, intptr_t start, intptr_t last) 291 { 292 MemCopyInfo *mc; 293 TempOptInfo *ti; 294 295 mc = QSIMPLEQ_FIRST(&ctx->mem_free); 296 if (mc) { 297 QSIMPLEQ_REMOVE_HEAD(&ctx->mem_free, next); 298 } else { 299 mc = tcg_malloc(sizeof(*mc)); 300 } 301 302 memset(mc, 0, sizeof(*mc)); 303 mc->itree.start = start; 304 mc->itree.last = last; 305 mc->type = type; 306 interval_tree_insert(&mc->itree, &ctx->mem_copy); 307 308 ts = find_better_copy(ts); 309 ti = ts_info(ts); 310 mc->ts = ts; 311 QSIMPLEQ_INSERT_TAIL(&ti->mem_copy, mc, next); 312 } 313 314 static bool ts_are_copies(TCGTemp *ts1, TCGTemp *ts2) 315 { 316 TCGTemp *i; 317 318 if (ts1 == ts2) { 319 return true; 320 } 321 322 if (!ts_is_copy(ts1) || !ts_is_copy(ts2)) { 323 return false; 324 } 325 326 for (i = ts_info(ts1)->next_copy; i != ts1; i = ts_info(i)->next_copy) { 327 if (i == ts2) { 328 return true; 329 } 330 } 331 332 return false; 333 } 334 335 static bool args_are_copies(TCGArg arg1, TCGArg arg2) 336 { 337 return ts_are_copies(arg_temp(arg1), arg_temp(arg2)); 338 } 339 340 static TCGTemp *find_mem_copy_for(OptContext *ctx, TCGType type, intptr_t s) 341 { 342 MemCopyInfo *mc; 343 344 for (mc = mem_copy_first(ctx, s, s); mc; mc = mem_copy_next(mc, s, s)) { 345 if (mc->itree.start == s && mc->type == type) { 346 return find_better_copy(mc->ts); 347 } 348 } 349 return NULL; 350 } 351 352 static TCGArg arg_new_constant(OptContext *ctx, uint64_t val) 353 { 354 TCGType type = ctx->type; 355 TCGTemp *ts; 356 357 if (type == TCG_TYPE_I32) { 358 val = (int32_t)val; 359 } 360 361 ts = tcg_constant_internal(type, val); 362 init_ts_info(ctx, ts); 363 364 return temp_arg(ts); 365 } 366 367 static bool tcg_opt_gen_mov(OptContext *ctx, TCGOp *op, TCGArg dst, TCGArg src) 368 { 369 TCGTemp *dst_ts = arg_temp(dst); 370 TCGTemp *src_ts = arg_temp(src); 371 TempOptInfo *di; 372 TempOptInfo *si; 373 TCGOpcode new_op; 374 375 if (ts_are_copies(dst_ts, src_ts)) { 376 tcg_op_remove(ctx->tcg, op); 377 return true; 378 } 379 380 reset_ts(ctx, dst_ts); 381 di = ts_info(dst_ts); 382 si = ts_info(src_ts); 383 384 switch (ctx->type) { 385 case TCG_TYPE_I32: 386 new_op = INDEX_op_mov_i32; 387 break; 388 case TCG_TYPE_I64: 389 new_op = INDEX_op_mov_i64; 390 break; 391 case TCG_TYPE_V64: 392 case TCG_TYPE_V128: 393 case TCG_TYPE_V256: 394 /* TCGOP_VECL and TCGOP_VECE remain unchanged. */ 395 new_op = INDEX_op_mov_vec; 396 break; 397 default: 398 g_assert_not_reached(); 399 } 400 op->opc = new_op; 401 op->args[0] = dst; 402 op->args[1] = src; 403 404 di->z_mask = si->z_mask; 405 di->s_mask = si->s_mask; 406 407 if (src_ts->type == dst_ts->type) { 408 TempOptInfo *ni = ts_info(si->next_copy); 409 410 di->next_copy = si->next_copy; 411 di->prev_copy = src_ts; 412 ni->prev_copy = dst_ts; 413 si->next_copy = dst_ts; 414 di->is_const = si->is_const; 415 di->val = si->val; 416 417 if (!QSIMPLEQ_EMPTY(&si->mem_copy) 418 && cmp_better_copy(src_ts, dst_ts) == dst_ts) { 419 move_mem_copies(dst_ts, src_ts); 420 } 421 } 422 return true; 423 } 424 425 static bool tcg_opt_gen_movi(OptContext *ctx, TCGOp *op, 426 TCGArg dst, uint64_t val) 427 { 428 /* Convert movi to mov with constant temp. */ 429 return tcg_opt_gen_mov(ctx, op, dst, arg_new_constant(ctx, val)); 430 } 431 432 static uint64_t do_constant_folding_2(TCGOpcode op, uint64_t x, uint64_t y) 433 { 434 uint64_t l64, h64; 435 436 switch (op) { 437 CASE_OP_32_64(add): 438 return x + y; 439 440 CASE_OP_32_64(sub): 441 return x - y; 442 443 CASE_OP_32_64(mul): 444 return x * y; 445 446 CASE_OP_32_64_VEC(and): 447 return x & y; 448 449 CASE_OP_32_64_VEC(or): 450 return x | y; 451 452 CASE_OP_32_64_VEC(xor): 453 return x ^ y; 454 455 case INDEX_op_shl_i32: 456 return (uint32_t)x << (y & 31); 457 458 case INDEX_op_shl_i64: 459 return (uint64_t)x << (y & 63); 460 461 case INDEX_op_shr_i32: 462 return (uint32_t)x >> (y & 31); 463 464 case INDEX_op_shr_i64: 465 return (uint64_t)x >> (y & 63); 466 467 case INDEX_op_sar_i32: 468 return (int32_t)x >> (y & 31); 469 470 case INDEX_op_sar_i64: 471 return (int64_t)x >> (y & 63); 472 473 case INDEX_op_rotr_i32: 474 return ror32(x, y & 31); 475 476 case INDEX_op_rotr_i64: 477 return ror64(x, y & 63); 478 479 case INDEX_op_rotl_i32: 480 return rol32(x, y & 31); 481 482 case INDEX_op_rotl_i64: 483 return rol64(x, y & 63); 484 485 CASE_OP_32_64_VEC(not): 486 return ~x; 487 488 CASE_OP_32_64(neg): 489 return -x; 490 491 CASE_OP_32_64_VEC(andc): 492 return x & ~y; 493 494 CASE_OP_32_64_VEC(orc): 495 return x | ~y; 496 497 CASE_OP_32_64_VEC(eqv): 498 return ~(x ^ y); 499 500 CASE_OP_32_64_VEC(nand): 501 return ~(x & y); 502 503 CASE_OP_32_64_VEC(nor): 504 return ~(x | y); 505 506 case INDEX_op_clz_i32: 507 return (uint32_t)x ? clz32(x) : y; 508 509 case INDEX_op_clz_i64: 510 return x ? clz64(x) : y; 511 512 case INDEX_op_ctz_i32: 513 return (uint32_t)x ? ctz32(x) : y; 514 515 case INDEX_op_ctz_i64: 516 return x ? ctz64(x) : y; 517 518 case INDEX_op_ctpop_i32: 519 return ctpop32(x); 520 521 case INDEX_op_ctpop_i64: 522 return ctpop64(x); 523 524 CASE_OP_32_64(ext8s): 525 return (int8_t)x; 526 527 CASE_OP_32_64(ext16s): 528 return (int16_t)x; 529 530 CASE_OP_32_64(ext8u): 531 return (uint8_t)x; 532 533 CASE_OP_32_64(ext16u): 534 return (uint16_t)x; 535 536 CASE_OP_32_64(bswap16): 537 x = bswap16(x); 538 return y & TCG_BSWAP_OS ? (int16_t)x : x; 539 540 CASE_OP_32_64(bswap32): 541 x = bswap32(x); 542 return y & TCG_BSWAP_OS ? (int32_t)x : x; 543 544 case INDEX_op_bswap64_i64: 545 return bswap64(x); 546 547 case INDEX_op_ext_i32_i64: 548 case INDEX_op_ext32s_i64: 549 return (int32_t)x; 550 551 case INDEX_op_extu_i32_i64: 552 case INDEX_op_extrl_i64_i32: 553 case INDEX_op_ext32u_i64: 554 return (uint32_t)x; 555 556 case INDEX_op_extrh_i64_i32: 557 return (uint64_t)x >> 32; 558 559 case INDEX_op_muluh_i32: 560 return ((uint64_t)(uint32_t)x * (uint32_t)y) >> 32; 561 case INDEX_op_mulsh_i32: 562 return ((int64_t)(int32_t)x * (int32_t)y) >> 32; 563 564 case INDEX_op_muluh_i64: 565 mulu64(&l64, &h64, x, y); 566 return h64; 567 case INDEX_op_mulsh_i64: 568 muls64(&l64, &h64, x, y); 569 return h64; 570 571 case INDEX_op_div_i32: 572 /* Avoid crashing on divide by zero, otherwise undefined. */ 573 return (int32_t)x / ((int32_t)y ? : 1); 574 case INDEX_op_divu_i32: 575 return (uint32_t)x / ((uint32_t)y ? : 1); 576 case INDEX_op_div_i64: 577 return (int64_t)x / ((int64_t)y ? : 1); 578 case INDEX_op_divu_i64: 579 return (uint64_t)x / ((uint64_t)y ? : 1); 580 581 case INDEX_op_rem_i32: 582 return (int32_t)x % ((int32_t)y ? : 1); 583 case INDEX_op_remu_i32: 584 return (uint32_t)x % ((uint32_t)y ? : 1); 585 case INDEX_op_rem_i64: 586 return (int64_t)x % ((int64_t)y ? : 1); 587 case INDEX_op_remu_i64: 588 return (uint64_t)x % ((uint64_t)y ? : 1); 589 590 default: 591 g_assert_not_reached(); 592 } 593 } 594 595 static uint64_t do_constant_folding(TCGOpcode op, TCGType type, 596 uint64_t x, uint64_t y) 597 { 598 uint64_t res = do_constant_folding_2(op, x, y); 599 if (type == TCG_TYPE_I32) { 600 res = (int32_t)res; 601 } 602 return res; 603 } 604 605 static bool do_constant_folding_cond_32(uint32_t x, uint32_t y, TCGCond c) 606 { 607 switch (c) { 608 case TCG_COND_EQ: 609 return x == y; 610 case TCG_COND_NE: 611 return x != y; 612 case TCG_COND_LT: 613 return (int32_t)x < (int32_t)y; 614 case TCG_COND_GE: 615 return (int32_t)x >= (int32_t)y; 616 case TCG_COND_LE: 617 return (int32_t)x <= (int32_t)y; 618 case TCG_COND_GT: 619 return (int32_t)x > (int32_t)y; 620 case TCG_COND_LTU: 621 return x < y; 622 case TCG_COND_GEU: 623 return x >= y; 624 case TCG_COND_LEU: 625 return x <= y; 626 case TCG_COND_GTU: 627 return x > y; 628 case TCG_COND_TSTEQ: 629 return (x & y) == 0; 630 case TCG_COND_TSTNE: 631 return (x & y) != 0; 632 case TCG_COND_ALWAYS: 633 case TCG_COND_NEVER: 634 break; 635 } 636 g_assert_not_reached(); 637 } 638 639 static bool do_constant_folding_cond_64(uint64_t x, uint64_t y, TCGCond c) 640 { 641 switch (c) { 642 case TCG_COND_EQ: 643 return x == y; 644 case TCG_COND_NE: 645 return x != y; 646 case TCG_COND_LT: 647 return (int64_t)x < (int64_t)y; 648 case TCG_COND_GE: 649 return (int64_t)x >= (int64_t)y; 650 case TCG_COND_LE: 651 return (int64_t)x <= (int64_t)y; 652 case TCG_COND_GT: 653 return (int64_t)x > (int64_t)y; 654 case TCG_COND_LTU: 655 return x < y; 656 case TCG_COND_GEU: 657 return x >= y; 658 case TCG_COND_LEU: 659 return x <= y; 660 case TCG_COND_GTU: 661 return x > y; 662 case TCG_COND_TSTEQ: 663 return (x & y) == 0; 664 case TCG_COND_TSTNE: 665 return (x & y) != 0; 666 case TCG_COND_ALWAYS: 667 case TCG_COND_NEVER: 668 break; 669 } 670 g_assert_not_reached(); 671 } 672 673 static int do_constant_folding_cond_eq(TCGCond c) 674 { 675 switch (c) { 676 case TCG_COND_GT: 677 case TCG_COND_LTU: 678 case TCG_COND_LT: 679 case TCG_COND_GTU: 680 case TCG_COND_NE: 681 return 0; 682 case TCG_COND_GE: 683 case TCG_COND_GEU: 684 case TCG_COND_LE: 685 case TCG_COND_LEU: 686 case TCG_COND_EQ: 687 return 1; 688 case TCG_COND_TSTEQ: 689 case TCG_COND_TSTNE: 690 return -1; 691 case TCG_COND_ALWAYS: 692 case TCG_COND_NEVER: 693 break; 694 } 695 g_assert_not_reached(); 696 } 697 698 /* 699 * Return -1 if the condition can't be simplified, 700 * and the result of the condition (0 or 1) if it can. 701 */ 702 static int do_constant_folding_cond(TCGType type, TCGArg x, 703 TCGArg y, TCGCond c) 704 { 705 if (arg_is_const(x) && arg_is_const(y)) { 706 uint64_t xv = arg_info(x)->val; 707 uint64_t yv = arg_info(y)->val; 708 709 switch (type) { 710 case TCG_TYPE_I32: 711 return do_constant_folding_cond_32(xv, yv, c); 712 case TCG_TYPE_I64: 713 return do_constant_folding_cond_64(xv, yv, c); 714 default: 715 /* Only scalar comparisons are optimizable */ 716 return -1; 717 } 718 } else if (args_are_copies(x, y)) { 719 return do_constant_folding_cond_eq(c); 720 } else if (arg_is_const_val(y, 0)) { 721 switch (c) { 722 case TCG_COND_LTU: 723 case TCG_COND_TSTNE: 724 return 0; 725 case TCG_COND_GEU: 726 case TCG_COND_TSTEQ: 727 return 1; 728 default: 729 return -1; 730 } 731 } 732 return -1; 733 } 734 735 /** 736 * swap_commutative: 737 * @dest: TCGArg of the destination argument, or NO_DEST. 738 * @p1: first paired argument 739 * @p2: second paired argument 740 * 741 * If *@p1 is a constant and *@p2 is not, swap. 742 * If *@p2 matches @dest, swap. 743 * Return true if a swap was performed. 744 */ 745 746 #define NO_DEST temp_arg(NULL) 747 748 static bool swap_commutative(TCGArg dest, TCGArg *p1, TCGArg *p2) 749 { 750 TCGArg a1 = *p1, a2 = *p2; 751 int sum = 0; 752 sum += arg_is_const(a1); 753 sum -= arg_is_const(a2); 754 755 /* Prefer the constant in second argument, and then the form 756 op a, a, b, which is better handled on non-RISC hosts. */ 757 if (sum > 0 || (sum == 0 && dest == a2)) { 758 *p1 = a2; 759 *p2 = a1; 760 return true; 761 } 762 return false; 763 } 764 765 static bool swap_commutative2(TCGArg *p1, TCGArg *p2) 766 { 767 int sum = 0; 768 sum += arg_is_const(p1[0]); 769 sum += arg_is_const(p1[1]); 770 sum -= arg_is_const(p2[0]); 771 sum -= arg_is_const(p2[1]); 772 if (sum > 0) { 773 TCGArg t; 774 t = p1[0], p1[0] = p2[0], p2[0] = t; 775 t = p1[1], p1[1] = p2[1], p2[1] = t; 776 return true; 777 } 778 return false; 779 } 780 781 /* 782 * Return -1 if the condition can't be simplified, 783 * and the result of the condition (0 or 1) if it can. 784 */ 785 static int do_constant_folding_cond1(OptContext *ctx, TCGArg dest, 786 TCGArg *p1, TCGArg *p2, TCGArg *pcond) 787 { 788 TCGCond cond; 789 bool swap; 790 int r; 791 792 swap = swap_commutative(dest, p1, p2); 793 cond = *pcond; 794 if (swap) { 795 *pcond = cond = tcg_swap_cond(cond); 796 } 797 798 r = do_constant_folding_cond(ctx->type, *p1, *p2, cond); 799 if (r >= 0) { 800 return r; 801 } 802 if (!is_tst_cond(cond)) { 803 return -1; 804 } 805 806 /* 807 * TSTNE x,x -> NE x,0 808 * TSTNE x,-1 -> NE x,0 809 */ 810 if (args_are_copies(*p1, *p2) || arg_is_const_val(*p2, -1)) { 811 *p2 = arg_new_constant(ctx, 0); 812 *pcond = tcg_tst_eqne_cond(cond); 813 return -1; 814 } 815 816 /* TSTNE x,sign -> LT x,0 */ 817 if (arg_is_const_val(*p2, (ctx->type == TCG_TYPE_I32 818 ? INT32_MIN : INT64_MIN))) { 819 *p2 = arg_new_constant(ctx, 0); 820 *pcond = tcg_tst_ltge_cond(cond); 821 } 822 return -1; 823 } 824 825 static int do_constant_folding_cond2(OptContext *ctx, TCGArg *args) 826 { 827 TCGArg al, ah, bl, bh; 828 TCGCond c; 829 bool swap; 830 int r; 831 832 swap = swap_commutative2(args, args + 2); 833 c = args[4]; 834 if (swap) { 835 args[4] = c = tcg_swap_cond(c); 836 } 837 838 al = args[0]; 839 ah = args[1]; 840 bl = args[2]; 841 bh = args[3]; 842 843 if (arg_is_const(bl) && arg_is_const(bh)) { 844 tcg_target_ulong blv = arg_info(bl)->val; 845 tcg_target_ulong bhv = arg_info(bh)->val; 846 uint64_t b = deposit64(blv, 32, 32, bhv); 847 848 if (arg_is_const(al) && arg_is_const(ah)) { 849 tcg_target_ulong alv = arg_info(al)->val; 850 tcg_target_ulong ahv = arg_info(ah)->val; 851 uint64_t a = deposit64(alv, 32, 32, ahv); 852 853 r = do_constant_folding_cond_64(a, b, c); 854 if (r >= 0) { 855 return r; 856 } 857 } 858 859 if (b == 0) { 860 switch (c) { 861 case TCG_COND_LTU: 862 case TCG_COND_TSTNE: 863 return 0; 864 case TCG_COND_GEU: 865 case TCG_COND_TSTEQ: 866 return 1; 867 default: 868 break; 869 } 870 } 871 872 /* TSTNE x,-1 -> NE x,0 */ 873 if (b == -1 && is_tst_cond(c)) { 874 args[3] = args[2] = arg_new_constant(ctx, 0); 875 args[4] = tcg_tst_eqne_cond(c); 876 return -1; 877 } 878 879 /* TSTNE x,sign -> LT x,0 */ 880 if (b == INT64_MIN && is_tst_cond(c)) { 881 /* bl must be 0, so copy that to bh */ 882 args[3] = bl; 883 args[4] = tcg_tst_ltge_cond(c); 884 return -1; 885 } 886 } 887 888 if (args_are_copies(al, bl) && args_are_copies(ah, bh)) { 889 r = do_constant_folding_cond_eq(c); 890 if (r >= 0) { 891 return r; 892 } 893 894 /* TSTNE x,x -> NE x,0 */ 895 if (is_tst_cond(c)) { 896 args[3] = args[2] = arg_new_constant(ctx, 0); 897 args[4] = tcg_tst_eqne_cond(c); 898 return -1; 899 } 900 } 901 return -1; 902 } 903 904 static void init_arguments(OptContext *ctx, TCGOp *op, int nb_args) 905 { 906 for (int i = 0; i < nb_args; i++) { 907 TCGTemp *ts = arg_temp(op->args[i]); 908 init_ts_info(ctx, ts); 909 } 910 } 911 912 static void copy_propagate(OptContext *ctx, TCGOp *op, 913 int nb_oargs, int nb_iargs) 914 { 915 for (int i = nb_oargs; i < nb_oargs + nb_iargs; i++) { 916 TCGTemp *ts = arg_temp(op->args[i]); 917 if (ts_is_copy(ts)) { 918 op->args[i] = temp_arg(find_better_copy(ts)); 919 } 920 } 921 } 922 923 static void finish_folding(OptContext *ctx, TCGOp *op) 924 { 925 const TCGOpDef *def = &tcg_op_defs[op->opc]; 926 int i, nb_oargs; 927 928 /* 929 * We only optimize extended basic blocks. If the opcode ends a BB 930 * and is not a conditional branch, reset all temp data. 931 */ 932 if (def->flags & TCG_OPF_BB_END) { 933 ctx->prev_mb = NULL; 934 if (!(def->flags & TCG_OPF_COND_BRANCH)) { 935 memset(&ctx->temps_used, 0, sizeof(ctx->temps_used)); 936 remove_mem_copy_all(ctx); 937 } 938 return; 939 } 940 941 nb_oargs = def->nb_oargs; 942 for (i = 0; i < nb_oargs; i++) { 943 TCGTemp *ts = arg_temp(op->args[i]); 944 reset_ts(ctx, ts); 945 /* 946 * Save the corresponding known-zero/sign bits mask for the 947 * first output argument (only one supported so far). 948 */ 949 if (i == 0) { 950 ts_info(ts)->z_mask = ctx->z_mask; 951 ts_info(ts)->s_mask = ctx->s_mask; 952 } 953 } 954 } 955 956 /* 957 * The fold_* functions return true when processing is complete, 958 * usually by folding the operation to a constant or to a copy, 959 * and calling tcg_opt_gen_{mov,movi}. They may do other things, 960 * like collect information about the value produced, for use in 961 * optimizing a subsequent operation. 962 * 963 * These first fold_* functions are all helpers, used by other 964 * folders for more specific operations. 965 */ 966 967 static bool fold_const1(OptContext *ctx, TCGOp *op) 968 { 969 if (arg_is_const(op->args[1])) { 970 uint64_t t; 971 972 t = arg_info(op->args[1])->val; 973 t = do_constant_folding(op->opc, ctx->type, t, 0); 974 return tcg_opt_gen_movi(ctx, op, op->args[0], t); 975 } 976 return false; 977 } 978 979 static bool fold_const2(OptContext *ctx, TCGOp *op) 980 { 981 if (arg_is_const(op->args[1]) && arg_is_const(op->args[2])) { 982 uint64_t t1 = arg_info(op->args[1])->val; 983 uint64_t t2 = arg_info(op->args[2])->val; 984 985 t1 = do_constant_folding(op->opc, ctx->type, t1, t2); 986 return tcg_opt_gen_movi(ctx, op, op->args[0], t1); 987 } 988 return false; 989 } 990 991 static bool fold_commutative(OptContext *ctx, TCGOp *op) 992 { 993 swap_commutative(op->args[0], &op->args[1], &op->args[2]); 994 return false; 995 } 996 997 static bool fold_const2_commutative(OptContext *ctx, TCGOp *op) 998 { 999 swap_commutative(op->args[0], &op->args[1], &op->args[2]); 1000 return fold_const2(ctx, op); 1001 } 1002 1003 static bool fold_masks(OptContext *ctx, TCGOp *op) 1004 { 1005 uint64_t a_mask = ctx->a_mask; 1006 uint64_t z_mask = ctx->z_mask; 1007 uint64_t s_mask = ctx->s_mask; 1008 1009 /* 1010 * 32-bit ops generate 32-bit results, which for the purpose of 1011 * simplifying tcg are sign-extended. Certainly that's how we 1012 * represent our constants elsewhere. Note that the bits will 1013 * be reset properly for a 64-bit value when encountering the 1014 * type changing opcodes. 1015 */ 1016 if (ctx->type == TCG_TYPE_I32) { 1017 a_mask = (int32_t)a_mask; 1018 z_mask = (int32_t)z_mask; 1019 s_mask |= MAKE_64BIT_MASK(32, 32); 1020 ctx->z_mask = z_mask; 1021 ctx->s_mask = s_mask; 1022 } 1023 1024 if (z_mask == 0) { 1025 return tcg_opt_gen_movi(ctx, op, op->args[0], 0); 1026 } 1027 if (a_mask == 0) { 1028 return tcg_opt_gen_mov(ctx, op, op->args[0], op->args[1]); 1029 } 1030 return false; 1031 } 1032 1033 /* 1034 * Convert @op to NOT, if NOT is supported by the host. 1035 * Return true f the conversion is successful, which will still 1036 * indicate that the processing is complete. 1037 */ 1038 static bool fold_not(OptContext *ctx, TCGOp *op); 1039 static bool fold_to_not(OptContext *ctx, TCGOp *op, int idx) 1040 { 1041 TCGOpcode not_op; 1042 bool have_not; 1043 1044 switch (ctx->type) { 1045 case TCG_TYPE_I32: 1046 not_op = INDEX_op_not_i32; 1047 have_not = TCG_TARGET_HAS_not_i32; 1048 break; 1049 case TCG_TYPE_I64: 1050 not_op = INDEX_op_not_i64; 1051 have_not = TCG_TARGET_HAS_not_i64; 1052 break; 1053 case TCG_TYPE_V64: 1054 case TCG_TYPE_V128: 1055 case TCG_TYPE_V256: 1056 not_op = INDEX_op_not_vec; 1057 have_not = TCG_TARGET_HAS_not_vec; 1058 break; 1059 default: 1060 g_assert_not_reached(); 1061 } 1062 if (have_not) { 1063 op->opc = not_op; 1064 op->args[1] = op->args[idx]; 1065 return fold_not(ctx, op); 1066 } 1067 return false; 1068 } 1069 1070 /* If the binary operation has first argument @i, fold to @i. */ 1071 static bool fold_ix_to_i(OptContext *ctx, TCGOp *op, uint64_t i) 1072 { 1073 if (arg_is_const_val(op->args[1], i)) { 1074 return tcg_opt_gen_movi(ctx, op, op->args[0], i); 1075 } 1076 return false; 1077 } 1078 1079 /* If the binary operation has first argument @i, fold to NOT. */ 1080 static bool fold_ix_to_not(OptContext *ctx, TCGOp *op, uint64_t i) 1081 { 1082 if (arg_is_const_val(op->args[1], i)) { 1083 return fold_to_not(ctx, op, 2); 1084 } 1085 return false; 1086 } 1087 1088 /* If the binary operation has second argument @i, fold to @i. */ 1089 static bool fold_xi_to_i(OptContext *ctx, TCGOp *op, uint64_t i) 1090 { 1091 if (arg_is_const_val(op->args[2], i)) { 1092 return tcg_opt_gen_movi(ctx, op, op->args[0], i); 1093 } 1094 return false; 1095 } 1096 1097 /* If the binary operation has second argument @i, fold to identity. */ 1098 static bool fold_xi_to_x(OptContext *ctx, TCGOp *op, uint64_t i) 1099 { 1100 if (arg_is_const_val(op->args[2], i)) { 1101 return tcg_opt_gen_mov(ctx, op, op->args[0], op->args[1]); 1102 } 1103 return false; 1104 } 1105 1106 /* If the binary operation has second argument @i, fold to NOT. */ 1107 static bool fold_xi_to_not(OptContext *ctx, TCGOp *op, uint64_t i) 1108 { 1109 if (arg_is_const_val(op->args[2], i)) { 1110 return fold_to_not(ctx, op, 1); 1111 } 1112 return false; 1113 } 1114 1115 /* If the binary operation has both arguments equal, fold to @i. */ 1116 static bool fold_xx_to_i(OptContext *ctx, TCGOp *op, uint64_t i) 1117 { 1118 if (args_are_copies(op->args[1], op->args[2])) { 1119 return tcg_opt_gen_movi(ctx, op, op->args[0], i); 1120 } 1121 return false; 1122 } 1123 1124 /* If the binary operation has both arguments equal, fold to identity. */ 1125 static bool fold_xx_to_x(OptContext *ctx, TCGOp *op) 1126 { 1127 if (args_are_copies(op->args[1], op->args[2])) { 1128 return tcg_opt_gen_mov(ctx, op, op->args[0], op->args[1]); 1129 } 1130 return false; 1131 } 1132 1133 /* 1134 * These outermost fold_<op> functions are sorted alphabetically. 1135 * 1136 * The ordering of the transformations should be: 1137 * 1) those that produce a constant 1138 * 2) those that produce a copy 1139 * 3) those that produce information about the result value. 1140 */ 1141 1142 static bool fold_add(OptContext *ctx, TCGOp *op) 1143 { 1144 if (fold_const2_commutative(ctx, op) || 1145 fold_xi_to_x(ctx, op, 0)) { 1146 return true; 1147 } 1148 return false; 1149 } 1150 1151 /* We cannot as yet do_constant_folding with vectors. */ 1152 static bool fold_add_vec(OptContext *ctx, TCGOp *op) 1153 { 1154 if (fold_commutative(ctx, op) || 1155 fold_xi_to_x(ctx, op, 0)) { 1156 return true; 1157 } 1158 return false; 1159 } 1160 1161 static bool fold_addsub2(OptContext *ctx, TCGOp *op, bool add) 1162 { 1163 bool a_const = arg_is_const(op->args[2]) && arg_is_const(op->args[3]); 1164 bool b_const = arg_is_const(op->args[4]) && arg_is_const(op->args[5]); 1165 1166 if (a_const && b_const) { 1167 uint64_t al = arg_info(op->args[2])->val; 1168 uint64_t ah = arg_info(op->args[3])->val; 1169 uint64_t bl = arg_info(op->args[4])->val; 1170 uint64_t bh = arg_info(op->args[5])->val; 1171 TCGArg rl, rh; 1172 TCGOp *op2; 1173 1174 if (ctx->type == TCG_TYPE_I32) { 1175 uint64_t a = deposit64(al, 32, 32, ah); 1176 uint64_t b = deposit64(bl, 32, 32, bh); 1177 1178 if (add) { 1179 a += b; 1180 } else { 1181 a -= b; 1182 } 1183 1184 al = sextract64(a, 0, 32); 1185 ah = sextract64(a, 32, 32); 1186 } else { 1187 Int128 a = int128_make128(al, ah); 1188 Int128 b = int128_make128(bl, bh); 1189 1190 if (add) { 1191 a = int128_add(a, b); 1192 } else { 1193 a = int128_sub(a, b); 1194 } 1195 1196 al = int128_getlo(a); 1197 ah = int128_gethi(a); 1198 } 1199 1200 rl = op->args[0]; 1201 rh = op->args[1]; 1202 1203 /* The proper opcode is supplied by tcg_opt_gen_mov. */ 1204 op2 = tcg_op_insert_before(ctx->tcg, op, 0, 2); 1205 1206 tcg_opt_gen_movi(ctx, op, rl, al); 1207 tcg_opt_gen_movi(ctx, op2, rh, ah); 1208 return true; 1209 } 1210 1211 /* Fold sub2 r,x,i to add2 r,x,-i */ 1212 if (!add && b_const) { 1213 uint64_t bl = arg_info(op->args[4])->val; 1214 uint64_t bh = arg_info(op->args[5])->val; 1215 1216 /* Negate the two parts without assembling and disassembling. */ 1217 bl = -bl; 1218 bh = ~bh + !bl; 1219 1220 op->opc = (ctx->type == TCG_TYPE_I32 1221 ? INDEX_op_add2_i32 : INDEX_op_add2_i64); 1222 op->args[4] = arg_new_constant(ctx, bl); 1223 op->args[5] = arg_new_constant(ctx, bh); 1224 } 1225 return false; 1226 } 1227 1228 static bool fold_add2(OptContext *ctx, TCGOp *op) 1229 { 1230 /* Note that the high and low parts may be independently swapped. */ 1231 swap_commutative(op->args[0], &op->args[2], &op->args[4]); 1232 swap_commutative(op->args[1], &op->args[3], &op->args[5]); 1233 1234 return fold_addsub2(ctx, op, true); 1235 } 1236 1237 static bool fold_and(OptContext *ctx, TCGOp *op) 1238 { 1239 uint64_t z1, z2; 1240 1241 if (fold_const2_commutative(ctx, op) || 1242 fold_xi_to_i(ctx, op, 0) || 1243 fold_xi_to_x(ctx, op, -1) || 1244 fold_xx_to_x(ctx, op)) { 1245 return true; 1246 } 1247 1248 z1 = arg_info(op->args[1])->z_mask; 1249 z2 = arg_info(op->args[2])->z_mask; 1250 ctx->z_mask = z1 & z2; 1251 1252 /* 1253 * Sign repetitions are perforce all identical, whether they are 1 or 0. 1254 * Bitwise operations preserve the relative quantity of the repetitions. 1255 */ 1256 ctx->s_mask = arg_info(op->args[1])->s_mask 1257 & arg_info(op->args[2])->s_mask; 1258 1259 /* 1260 * Known-zeros does not imply known-ones. Therefore unless 1261 * arg2 is constant, we can't infer affected bits from it. 1262 */ 1263 if (arg_is_const(op->args[2])) { 1264 ctx->a_mask = z1 & ~z2; 1265 } 1266 1267 return fold_masks(ctx, op); 1268 } 1269 1270 static bool fold_andc(OptContext *ctx, TCGOp *op) 1271 { 1272 uint64_t z1; 1273 1274 if (fold_const2(ctx, op) || 1275 fold_xx_to_i(ctx, op, 0) || 1276 fold_xi_to_x(ctx, op, 0) || 1277 fold_ix_to_not(ctx, op, -1)) { 1278 return true; 1279 } 1280 1281 z1 = arg_info(op->args[1])->z_mask; 1282 1283 /* 1284 * Known-zeros does not imply known-ones. Therefore unless 1285 * arg2 is constant, we can't infer anything from it. 1286 */ 1287 if (arg_is_const(op->args[2])) { 1288 uint64_t z2 = ~arg_info(op->args[2])->z_mask; 1289 ctx->a_mask = z1 & ~z2; 1290 z1 &= z2; 1291 } 1292 ctx->z_mask = z1; 1293 1294 ctx->s_mask = arg_info(op->args[1])->s_mask 1295 & arg_info(op->args[2])->s_mask; 1296 return fold_masks(ctx, op); 1297 } 1298 1299 static bool fold_brcond(OptContext *ctx, TCGOp *op) 1300 { 1301 int i = do_constant_folding_cond1(ctx, NO_DEST, &op->args[0], 1302 &op->args[1], &op->args[2]); 1303 if (i == 0) { 1304 tcg_op_remove(ctx->tcg, op); 1305 return true; 1306 } 1307 if (i > 0) { 1308 op->opc = INDEX_op_br; 1309 op->args[0] = op->args[3]; 1310 } 1311 return false; 1312 } 1313 1314 static bool fold_brcond2(OptContext *ctx, TCGOp *op) 1315 { 1316 TCGCond cond; 1317 TCGArg label; 1318 int i, inv = 0; 1319 1320 i = do_constant_folding_cond2(ctx, &op->args[0]); 1321 cond = op->args[4]; 1322 label = op->args[5]; 1323 if (i >= 0) { 1324 goto do_brcond_const; 1325 } 1326 1327 switch (cond) { 1328 case TCG_COND_LT: 1329 case TCG_COND_GE: 1330 /* 1331 * Simplify LT/GE comparisons vs zero to a single compare 1332 * vs the high word of the input. 1333 */ 1334 if (arg_is_const_val(op->args[2], 0) && 1335 arg_is_const_val(op->args[3], 0)) { 1336 goto do_brcond_high; 1337 } 1338 break; 1339 1340 case TCG_COND_NE: 1341 inv = 1; 1342 QEMU_FALLTHROUGH; 1343 case TCG_COND_EQ: 1344 /* 1345 * Simplify EQ/NE comparisons where one of the pairs 1346 * can be simplified. 1347 */ 1348 i = do_constant_folding_cond(TCG_TYPE_I32, op->args[0], 1349 op->args[2], cond); 1350 switch (i ^ inv) { 1351 case 0: 1352 goto do_brcond_const; 1353 case 1: 1354 goto do_brcond_high; 1355 } 1356 1357 i = do_constant_folding_cond(TCG_TYPE_I32, op->args[1], 1358 op->args[3], cond); 1359 switch (i ^ inv) { 1360 case 0: 1361 goto do_brcond_const; 1362 case 1: 1363 goto do_brcond_low; 1364 } 1365 break; 1366 1367 case TCG_COND_TSTEQ: 1368 case TCG_COND_TSTNE: 1369 if (arg_is_const_val(op->args[2], 0)) { 1370 goto do_brcond_high; 1371 } 1372 if (arg_is_const_val(op->args[3], 0)) { 1373 goto do_brcond_low; 1374 } 1375 break; 1376 1377 default: 1378 break; 1379 1380 do_brcond_low: 1381 op->opc = INDEX_op_brcond_i32; 1382 op->args[1] = op->args[2]; 1383 op->args[2] = cond; 1384 op->args[3] = label; 1385 return fold_brcond(ctx, op); 1386 1387 do_brcond_high: 1388 op->opc = INDEX_op_brcond_i32; 1389 op->args[0] = op->args[1]; 1390 op->args[1] = op->args[3]; 1391 op->args[2] = cond; 1392 op->args[3] = label; 1393 return fold_brcond(ctx, op); 1394 1395 do_brcond_const: 1396 if (i == 0) { 1397 tcg_op_remove(ctx->tcg, op); 1398 return true; 1399 } 1400 op->opc = INDEX_op_br; 1401 op->args[0] = label; 1402 break; 1403 } 1404 return false; 1405 } 1406 1407 static bool fold_bswap(OptContext *ctx, TCGOp *op) 1408 { 1409 uint64_t z_mask, s_mask, sign; 1410 1411 if (arg_is_const(op->args[1])) { 1412 uint64_t t = arg_info(op->args[1])->val; 1413 1414 t = do_constant_folding(op->opc, ctx->type, t, op->args[2]); 1415 return tcg_opt_gen_movi(ctx, op, op->args[0], t); 1416 } 1417 1418 z_mask = arg_info(op->args[1])->z_mask; 1419 1420 switch (op->opc) { 1421 case INDEX_op_bswap16_i32: 1422 case INDEX_op_bswap16_i64: 1423 z_mask = bswap16(z_mask); 1424 sign = INT16_MIN; 1425 break; 1426 case INDEX_op_bswap32_i32: 1427 case INDEX_op_bswap32_i64: 1428 z_mask = bswap32(z_mask); 1429 sign = INT32_MIN; 1430 break; 1431 case INDEX_op_bswap64_i64: 1432 z_mask = bswap64(z_mask); 1433 sign = INT64_MIN; 1434 break; 1435 default: 1436 g_assert_not_reached(); 1437 } 1438 s_mask = smask_from_zmask(z_mask); 1439 1440 switch (op->args[2] & (TCG_BSWAP_OZ | TCG_BSWAP_OS)) { 1441 case TCG_BSWAP_OZ: 1442 break; 1443 case TCG_BSWAP_OS: 1444 /* If the sign bit may be 1, force all the bits above to 1. */ 1445 if (z_mask & sign) { 1446 z_mask |= sign; 1447 s_mask = sign << 1; 1448 } 1449 break; 1450 default: 1451 /* The high bits are undefined: force all bits above the sign to 1. */ 1452 z_mask |= sign << 1; 1453 s_mask = 0; 1454 break; 1455 } 1456 ctx->z_mask = z_mask; 1457 ctx->s_mask = s_mask; 1458 1459 return fold_masks(ctx, op); 1460 } 1461 1462 static bool fold_call(OptContext *ctx, TCGOp *op) 1463 { 1464 TCGContext *s = ctx->tcg; 1465 int nb_oargs = TCGOP_CALLO(op); 1466 int nb_iargs = TCGOP_CALLI(op); 1467 int flags, i; 1468 1469 init_arguments(ctx, op, nb_oargs + nb_iargs); 1470 copy_propagate(ctx, op, nb_oargs, nb_iargs); 1471 1472 /* If the function reads or writes globals, reset temp data. */ 1473 flags = tcg_call_flags(op); 1474 if (!(flags & (TCG_CALL_NO_READ_GLOBALS | TCG_CALL_NO_WRITE_GLOBALS))) { 1475 int nb_globals = s->nb_globals; 1476 1477 for (i = 0; i < nb_globals; i++) { 1478 if (test_bit(i, ctx->temps_used.l)) { 1479 reset_ts(ctx, &ctx->tcg->temps[i]); 1480 } 1481 } 1482 } 1483 1484 /* If the function has side effects, reset mem data. */ 1485 if (!(flags & TCG_CALL_NO_SIDE_EFFECTS)) { 1486 remove_mem_copy_all(ctx); 1487 } 1488 1489 /* Reset temp data for outputs. */ 1490 for (i = 0; i < nb_oargs; i++) { 1491 reset_temp(ctx, op->args[i]); 1492 } 1493 1494 /* Stop optimizing MB across calls. */ 1495 ctx->prev_mb = NULL; 1496 return true; 1497 } 1498 1499 static bool fold_count_zeros(OptContext *ctx, TCGOp *op) 1500 { 1501 uint64_t z_mask; 1502 1503 if (arg_is_const(op->args[1])) { 1504 uint64_t t = arg_info(op->args[1])->val; 1505 1506 if (t != 0) { 1507 t = do_constant_folding(op->opc, ctx->type, t, 0); 1508 return tcg_opt_gen_movi(ctx, op, op->args[0], t); 1509 } 1510 return tcg_opt_gen_mov(ctx, op, op->args[0], op->args[2]); 1511 } 1512 1513 switch (ctx->type) { 1514 case TCG_TYPE_I32: 1515 z_mask = 31; 1516 break; 1517 case TCG_TYPE_I64: 1518 z_mask = 63; 1519 break; 1520 default: 1521 g_assert_not_reached(); 1522 } 1523 ctx->z_mask = arg_info(op->args[2])->z_mask | z_mask; 1524 ctx->s_mask = smask_from_zmask(ctx->z_mask); 1525 return false; 1526 } 1527 1528 static bool fold_ctpop(OptContext *ctx, TCGOp *op) 1529 { 1530 if (fold_const1(ctx, op)) { 1531 return true; 1532 } 1533 1534 switch (ctx->type) { 1535 case TCG_TYPE_I32: 1536 ctx->z_mask = 32 | 31; 1537 break; 1538 case TCG_TYPE_I64: 1539 ctx->z_mask = 64 | 63; 1540 break; 1541 default: 1542 g_assert_not_reached(); 1543 } 1544 ctx->s_mask = smask_from_zmask(ctx->z_mask); 1545 return false; 1546 } 1547 1548 static bool fold_deposit(OptContext *ctx, TCGOp *op) 1549 { 1550 TCGOpcode and_opc; 1551 1552 if (arg_is_const(op->args[1]) && arg_is_const(op->args[2])) { 1553 uint64_t t1 = arg_info(op->args[1])->val; 1554 uint64_t t2 = arg_info(op->args[2])->val; 1555 1556 t1 = deposit64(t1, op->args[3], op->args[4], t2); 1557 return tcg_opt_gen_movi(ctx, op, op->args[0], t1); 1558 } 1559 1560 switch (ctx->type) { 1561 case TCG_TYPE_I32: 1562 and_opc = INDEX_op_and_i32; 1563 break; 1564 case TCG_TYPE_I64: 1565 and_opc = INDEX_op_and_i64; 1566 break; 1567 default: 1568 g_assert_not_reached(); 1569 } 1570 1571 /* Inserting a value into zero at offset 0. */ 1572 if (arg_is_const_val(op->args[1], 0) && op->args[3] == 0) { 1573 uint64_t mask = MAKE_64BIT_MASK(0, op->args[4]); 1574 1575 op->opc = and_opc; 1576 op->args[1] = op->args[2]; 1577 op->args[2] = arg_new_constant(ctx, mask); 1578 ctx->z_mask = mask & arg_info(op->args[1])->z_mask; 1579 return false; 1580 } 1581 1582 /* Inserting zero into a value. */ 1583 if (arg_is_const_val(op->args[2], 0)) { 1584 uint64_t mask = deposit64(-1, op->args[3], op->args[4], 0); 1585 1586 op->opc = and_opc; 1587 op->args[2] = arg_new_constant(ctx, mask); 1588 ctx->z_mask = mask & arg_info(op->args[1])->z_mask; 1589 return false; 1590 } 1591 1592 ctx->z_mask = deposit64(arg_info(op->args[1])->z_mask, 1593 op->args[3], op->args[4], 1594 arg_info(op->args[2])->z_mask); 1595 return false; 1596 } 1597 1598 static bool fold_divide(OptContext *ctx, TCGOp *op) 1599 { 1600 if (fold_const2(ctx, op) || 1601 fold_xi_to_x(ctx, op, 1)) { 1602 return true; 1603 } 1604 return false; 1605 } 1606 1607 static bool fold_dup(OptContext *ctx, TCGOp *op) 1608 { 1609 if (arg_is_const(op->args[1])) { 1610 uint64_t t = arg_info(op->args[1])->val; 1611 t = dup_const(TCGOP_VECE(op), t); 1612 return tcg_opt_gen_movi(ctx, op, op->args[0], t); 1613 } 1614 return false; 1615 } 1616 1617 static bool fold_dup2(OptContext *ctx, TCGOp *op) 1618 { 1619 if (arg_is_const(op->args[1]) && arg_is_const(op->args[2])) { 1620 uint64_t t = deposit64(arg_info(op->args[1])->val, 32, 32, 1621 arg_info(op->args[2])->val); 1622 return tcg_opt_gen_movi(ctx, op, op->args[0], t); 1623 } 1624 1625 if (args_are_copies(op->args[1], op->args[2])) { 1626 op->opc = INDEX_op_dup_vec; 1627 TCGOP_VECE(op) = MO_32; 1628 } 1629 return false; 1630 } 1631 1632 static bool fold_eqv(OptContext *ctx, TCGOp *op) 1633 { 1634 if (fold_const2_commutative(ctx, op) || 1635 fold_xi_to_x(ctx, op, -1) || 1636 fold_xi_to_not(ctx, op, 0)) { 1637 return true; 1638 } 1639 1640 ctx->s_mask = arg_info(op->args[1])->s_mask 1641 & arg_info(op->args[2])->s_mask; 1642 return false; 1643 } 1644 1645 static bool fold_extract(OptContext *ctx, TCGOp *op) 1646 { 1647 uint64_t z_mask_old, z_mask; 1648 int pos = op->args[2]; 1649 int len = op->args[3]; 1650 1651 if (arg_is_const(op->args[1])) { 1652 uint64_t t; 1653 1654 t = arg_info(op->args[1])->val; 1655 t = extract64(t, pos, len); 1656 return tcg_opt_gen_movi(ctx, op, op->args[0], t); 1657 } 1658 1659 z_mask_old = arg_info(op->args[1])->z_mask; 1660 z_mask = extract64(z_mask_old, pos, len); 1661 if (pos == 0) { 1662 ctx->a_mask = z_mask_old ^ z_mask; 1663 } 1664 ctx->z_mask = z_mask; 1665 ctx->s_mask = smask_from_zmask(z_mask); 1666 1667 return fold_masks(ctx, op); 1668 } 1669 1670 static bool fold_extract2(OptContext *ctx, TCGOp *op) 1671 { 1672 if (arg_is_const(op->args[1]) && arg_is_const(op->args[2])) { 1673 uint64_t v1 = arg_info(op->args[1])->val; 1674 uint64_t v2 = arg_info(op->args[2])->val; 1675 int shr = op->args[3]; 1676 1677 if (op->opc == INDEX_op_extract2_i64) { 1678 v1 >>= shr; 1679 v2 <<= 64 - shr; 1680 } else { 1681 v1 = (uint32_t)v1 >> shr; 1682 v2 = (uint64_t)((int32_t)v2 << (32 - shr)); 1683 } 1684 return tcg_opt_gen_movi(ctx, op, op->args[0], v1 | v2); 1685 } 1686 return false; 1687 } 1688 1689 static bool fold_exts(OptContext *ctx, TCGOp *op) 1690 { 1691 uint64_t s_mask_old, s_mask, z_mask, sign; 1692 bool type_change = false; 1693 1694 if (fold_const1(ctx, op)) { 1695 return true; 1696 } 1697 1698 z_mask = arg_info(op->args[1])->z_mask; 1699 s_mask = arg_info(op->args[1])->s_mask; 1700 s_mask_old = s_mask; 1701 1702 switch (op->opc) { 1703 CASE_OP_32_64(ext8s): 1704 sign = INT8_MIN; 1705 z_mask = (uint8_t)z_mask; 1706 break; 1707 CASE_OP_32_64(ext16s): 1708 sign = INT16_MIN; 1709 z_mask = (uint16_t)z_mask; 1710 break; 1711 case INDEX_op_ext_i32_i64: 1712 type_change = true; 1713 QEMU_FALLTHROUGH; 1714 case INDEX_op_ext32s_i64: 1715 sign = INT32_MIN; 1716 z_mask = (uint32_t)z_mask; 1717 break; 1718 default: 1719 g_assert_not_reached(); 1720 } 1721 1722 if (z_mask & sign) { 1723 z_mask |= sign; 1724 } 1725 s_mask |= sign << 1; 1726 1727 ctx->z_mask = z_mask; 1728 ctx->s_mask = s_mask; 1729 if (!type_change) { 1730 ctx->a_mask = s_mask & ~s_mask_old; 1731 } 1732 1733 return fold_masks(ctx, op); 1734 } 1735 1736 static bool fold_extu(OptContext *ctx, TCGOp *op) 1737 { 1738 uint64_t z_mask_old, z_mask; 1739 bool type_change = false; 1740 1741 if (fold_const1(ctx, op)) { 1742 return true; 1743 } 1744 1745 z_mask_old = z_mask = arg_info(op->args[1])->z_mask; 1746 1747 switch (op->opc) { 1748 CASE_OP_32_64(ext8u): 1749 z_mask = (uint8_t)z_mask; 1750 break; 1751 CASE_OP_32_64(ext16u): 1752 z_mask = (uint16_t)z_mask; 1753 break; 1754 case INDEX_op_extrl_i64_i32: 1755 case INDEX_op_extu_i32_i64: 1756 type_change = true; 1757 QEMU_FALLTHROUGH; 1758 case INDEX_op_ext32u_i64: 1759 z_mask = (uint32_t)z_mask; 1760 break; 1761 case INDEX_op_extrh_i64_i32: 1762 type_change = true; 1763 z_mask >>= 32; 1764 break; 1765 default: 1766 g_assert_not_reached(); 1767 } 1768 1769 ctx->z_mask = z_mask; 1770 ctx->s_mask = smask_from_zmask(z_mask); 1771 if (!type_change) { 1772 ctx->a_mask = z_mask_old ^ z_mask; 1773 } 1774 return fold_masks(ctx, op); 1775 } 1776 1777 static bool fold_mb(OptContext *ctx, TCGOp *op) 1778 { 1779 /* Eliminate duplicate and redundant fence instructions. */ 1780 if (ctx->prev_mb) { 1781 /* 1782 * Merge two barriers of the same type into one, 1783 * or a weaker barrier into a stronger one, 1784 * or two weaker barriers into a stronger one. 1785 * mb X; mb Y => mb X|Y 1786 * mb; strl => mb; st 1787 * ldaq; mb => ld; mb 1788 * ldaq; strl => ld; mb; st 1789 * Other combinations are also merged into a strong 1790 * barrier. This is stricter than specified but for 1791 * the purposes of TCG is better than not optimizing. 1792 */ 1793 ctx->prev_mb->args[0] |= op->args[0]; 1794 tcg_op_remove(ctx->tcg, op); 1795 } else { 1796 ctx->prev_mb = op; 1797 } 1798 return true; 1799 } 1800 1801 static bool fold_mov(OptContext *ctx, TCGOp *op) 1802 { 1803 return tcg_opt_gen_mov(ctx, op, op->args[0], op->args[1]); 1804 } 1805 1806 static bool fold_movcond(OptContext *ctx, TCGOp *op) 1807 { 1808 int i; 1809 1810 /* 1811 * Canonicalize the "false" input reg to match the destination reg so 1812 * that the tcg backend can implement a "move if true" operation. 1813 */ 1814 if (swap_commutative(op->args[0], &op->args[4], &op->args[3])) { 1815 op->args[5] = tcg_invert_cond(op->args[5]); 1816 } 1817 1818 i = do_constant_folding_cond1(ctx, NO_DEST, &op->args[1], 1819 &op->args[2], &op->args[5]); 1820 if (i >= 0) { 1821 return tcg_opt_gen_mov(ctx, op, op->args[0], op->args[4 - i]); 1822 } 1823 1824 ctx->z_mask = arg_info(op->args[3])->z_mask 1825 | arg_info(op->args[4])->z_mask; 1826 ctx->s_mask = arg_info(op->args[3])->s_mask 1827 & arg_info(op->args[4])->s_mask; 1828 1829 if (arg_is_const(op->args[3]) && arg_is_const(op->args[4])) { 1830 uint64_t tv = arg_info(op->args[3])->val; 1831 uint64_t fv = arg_info(op->args[4])->val; 1832 TCGOpcode opc, negopc = 0; 1833 TCGCond cond = op->args[5]; 1834 1835 switch (ctx->type) { 1836 case TCG_TYPE_I32: 1837 opc = INDEX_op_setcond_i32; 1838 if (TCG_TARGET_HAS_negsetcond_i32) { 1839 negopc = INDEX_op_negsetcond_i32; 1840 } 1841 tv = (int32_t)tv; 1842 fv = (int32_t)fv; 1843 break; 1844 case TCG_TYPE_I64: 1845 opc = INDEX_op_setcond_i64; 1846 if (TCG_TARGET_HAS_negsetcond_i64) { 1847 negopc = INDEX_op_negsetcond_i64; 1848 } 1849 break; 1850 default: 1851 g_assert_not_reached(); 1852 } 1853 1854 if (tv == 1 && fv == 0) { 1855 op->opc = opc; 1856 op->args[3] = cond; 1857 } else if (fv == 1 && tv == 0) { 1858 op->opc = opc; 1859 op->args[3] = tcg_invert_cond(cond); 1860 } else if (negopc) { 1861 if (tv == -1 && fv == 0) { 1862 op->opc = negopc; 1863 op->args[3] = cond; 1864 } else if (fv == -1 && tv == 0) { 1865 op->opc = negopc; 1866 op->args[3] = tcg_invert_cond(cond); 1867 } 1868 } 1869 } 1870 return false; 1871 } 1872 1873 static bool fold_mul(OptContext *ctx, TCGOp *op) 1874 { 1875 if (fold_const2(ctx, op) || 1876 fold_xi_to_i(ctx, op, 0) || 1877 fold_xi_to_x(ctx, op, 1)) { 1878 return true; 1879 } 1880 return false; 1881 } 1882 1883 static bool fold_mul_highpart(OptContext *ctx, TCGOp *op) 1884 { 1885 if (fold_const2_commutative(ctx, op) || 1886 fold_xi_to_i(ctx, op, 0)) { 1887 return true; 1888 } 1889 return false; 1890 } 1891 1892 static bool fold_multiply2(OptContext *ctx, TCGOp *op) 1893 { 1894 swap_commutative(op->args[0], &op->args[2], &op->args[3]); 1895 1896 if (arg_is_const(op->args[2]) && arg_is_const(op->args[3])) { 1897 uint64_t a = arg_info(op->args[2])->val; 1898 uint64_t b = arg_info(op->args[3])->val; 1899 uint64_t h, l; 1900 TCGArg rl, rh; 1901 TCGOp *op2; 1902 1903 switch (op->opc) { 1904 case INDEX_op_mulu2_i32: 1905 l = (uint64_t)(uint32_t)a * (uint32_t)b; 1906 h = (int32_t)(l >> 32); 1907 l = (int32_t)l; 1908 break; 1909 case INDEX_op_muls2_i32: 1910 l = (int64_t)(int32_t)a * (int32_t)b; 1911 h = l >> 32; 1912 l = (int32_t)l; 1913 break; 1914 case INDEX_op_mulu2_i64: 1915 mulu64(&l, &h, a, b); 1916 break; 1917 case INDEX_op_muls2_i64: 1918 muls64(&l, &h, a, b); 1919 break; 1920 default: 1921 g_assert_not_reached(); 1922 } 1923 1924 rl = op->args[0]; 1925 rh = op->args[1]; 1926 1927 /* The proper opcode is supplied by tcg_opt_gen_mov. */ 1928 op2 = tcg_op_insert_before(ctx->tcg, op, 0, 2); 1929 1930 tcg_opt_gen_movi(ctx, op, rl, l); 1931 tcg_opt_gen_movi(ctx, op2, rh, h); 1932 return true; 1933 } 1934 return false; 1935 } 1936 1937 static bool fold_nand(OptContext *ctx, TCGOp *op) 1938 { 1939 if (fold_const2_commutative(ctx, op) || 1940 fold_xi_to_not(ctx, op, -1)) { 1941 return true; 1942 } 1943 1944 ctx->s_mask = arg_info(op->args[1])->s_mask 1945 & arg_info(op->args[2])->s_mask; 1946 return false; 1947 } 1948 1949 static bool fold_neg(OptContext *ctx, TCGOp *op) 1950 { 1951 uint64_t z_mask; 1952 1953 if (fold_const1(ctx, op)) { 1954 return true; 1955 } 1956 1957 /* Set to 1 all bits to the left of the rightmost. */ 1958 z_mask = arg_info(op->args[1])->z_mask; 1959 ctx->z_mask = -(z_mask & -z_mask); 1960 1961 /* 1962 * Because of fold_sub_to_neg, we want to always return true, 1963 * via finish_folding. 1964 */ 1965 finish_folding(ctx, op); 1966 return true; 1967 } 1968 1969 static bool fold_nor(OptContext *ctx, TCGOp *op) 1970 { 1971 if (fold_const2_commutative(ctx, op) || 1972 fold_xi_to_not(ctx, op, 0)) { 1973 return true; 1974 } 1975 1976 ctx->s_mask = arg_info(op->args[1])->s_mask 1977 & arg_info(op->args[2])->s_mask; 1978 return false; 1979 } 1980 1981 static bool fold_not(OptContext *ctx, TCGOp *op) 1982 { 1983 if (fold_const1(ctx, op)) { 1984 return true; 1985 } 1986 1987 ctx->s_mask = arg_info(op->args[1])->s_mask; 1988 1989 /* Because of fold_to_not, we want to always return true, via finish. */ 1990 finish_folding(ctx, op); 1991 return true; 1992 } 1993 1994 static bool fold_or(OptContext *ctx, TCGOp *op) 1995 { 1996 if (fold_const2_commutative(ctx, op) || 1997 fold_xi_to_x(ctx, op, 0) || 1998 fold_xx_to_x(ctx, op)) { 1999 return true; 2000 } 2001 2002 ctx->z_mask = arg_info(op->args[1])->z_mask 2003 | arg_info(op->args[2])->z_mask; 2004 ctx->s_mask = arg_info(op->args[1])->s_mask 2005 & arg_info(op->args[2])->s_mask; 2006 return fold_masks(ctx, op); 2007 } 2008 2009 static bool fold_orc(OptContext *ctx, TCGOp *op) 2010 { 2011 if (fold_const2(ctx, op) || 2012 fold_xx_to_i(ctx, op, -1) || 2013 fold_xi_to_x(ctx, op, -1) || 2014 fold_ix_to_not(ctx, op, 0)) { 2015 return true; 2016 } 2017 2018 ctx->s_mask = arg_info(op->args[1])->s_mask 2019 & arg_info(op->args[2])->s_mask; 2020 return false; 2021 } 2022 2023 static bool fold_qemu_ld(OptContext *ctx, TCGOp *op) 2024 { 2025 const TCGOpDef *def = &tcg_op_defs[op->opc]; 2026 MemOpIdx oi = op->args[def->nb_oargs + def->nb_iargs]; 2027 MemOp mop = get_memop(oi); 2028 int width = 8 * memop_size(mop); 2029 2030 if (width < 64) { 2031 ctx->s_mask = MAKE_64BIT_MASK(width, 64 - width); 2032 if (!(mop & MO_SIGN)) { 2033 ctx->z_mask = MAKE_64BIT_MASK(0, width); 2034 ctx->s_mask <<= 1; 2035 } 2036 } 2037 2038 /* Opcodes that touch guest memory stop the mb optimization. */ 2039 ctx->prev_mb = NULL; 2040 return false; 2041 } 2042 2043 static bool fold_qemu_st(OptContext *ctx, TCGOp *op) 2044 { 2045 /* Opcodes that touch guest memory stop the mb optimization. */ 2046 ctx->prev_mb = NULL; 2047 return false; 2048 } 2049 2050 static bool fold_remainder(OptContext *ctx, TCGOp *op) 2051 { 2052 if (fold_const2(ctx, op) || 2053 fold_xx_to_i(ctx, op, 0)) { 2054 return true; 2055 } 2056 return false; 2057 } 2058 2059 static void fold_setcond_tst_pow2(OptContext *ctx, TCGOp *op, bool neg) 2060 { 2061 TCGOpcode and_opc, sub_opc, xor_opc, neg_opc, shr_opc, uext_opc, sext_opc; 2062 TCGCond cond = op->args[3]; 2063 TCGArg ret, src1, src2; 2064 TCGOp *op2; 2065 uint64_t val; 2066 int sh; 2067 bool inv; 2068 2069 if (!is_tst_cond(cond) || !arg_is_const(op->args[2])) { 2070 return; 2071 } 2072 2073 src2 = op->args[2]; 2074 val = arg_info(src2)->val; 2075 if (!is_power_of_2(val)) { 2076 return; 2077 } 2078 sh = ctz64(val); 2079 2080 switch (ctx->type) { 2081 case TCG_TYPE_I32: 2082 and_opc = INDEX_op_and_i32; 2083 sub_opc = INDEX_op_sub_i32; 2084 xor_opc = INDEX_op_xor_i32; 2085 shr_opc = INDEX_op_shr_i32; 2086 neg_opc = INDEX_op_neg_i32; 2087 if (TCG_TARGET_extract_i32_valid(sh, 1)) { 2088 uext_opc = TCG_TARGET_HAS_extract_i32 ? INDEX_op_extract_i32 : 0; 2089 sext_opc = TCG_TARGET_HAS_sextract_i32 ? INDEX_op_sextract_i32 : 0; 2090 } 2091 break; 2092 case TCG_TYPE_I64: 2093 and_opc = INDEX_op_and_i64; 2094 sub_opc = INDEX_op_sub_i64; 2095 xor_opc = INDEX_op_xor_i64; 2096 shr_opc = INDEX_op_shr_i64; 2097 neg_opc = INDEX_op_neg_i64; 2098 if (TCG_TARGET_extract_i64_valid(sh, 1)) { 2099 uext_opc = TCG_TARGET_HAS_extract_i64 ? INDEX_op_extract_i64 : 0; 2100 sext_opc = TCG_TARGET_HAS_sextract_i64 ? INDEX_op_sextract_i64 : 0; 2101 } 2102 break; 2103 default: 2104 g_assert_not_reached(); 2105 } 2106 2107 ret = op->args[0]; 2108 src1 = op->args[1]; 2109 inv = cond == TCG_COND_TSTEQ; 2110 2111 if (sh && sext_opc && neg && !inv) { 2112 op->opc = sext_opc; 2113 op->args[1] = src1; 2114 op->args[2] = sh; 2115 op->args[3] = 1; 2116 return; 2117 } else if (sh && uext_opc) { 2118 op->opc = uext_opc; 2119 op->args[1] = src1; 2120 op->args[2] = sh; 2121 op->args[3] = 1; 2122 } else { 2123 if (sh) { 2124 op2 = tcg_op_insert_before(ctx->tcg, op, shr_opc, 3); 2125 op2->args[0] = ret; 2126 op2->args[1] = src1; 2127 op2->args[2] = arg_new_constant(ctx, sh); 2128 src1 = ret; 2129 } 2130 op->opc = and_opc; 2131 op->args[1] = src1; 2132 op->args[2] = arg_new_constant(ctx, 1); 2133 } 2134 2135 if (neg && inv) { 2136 op2 = tcg_op_insert_after(ctx->tcg, op, sub_opc, 3); 2137 op2->args[0] = ret; 2138 op2->args[1] = ret; 2139 op2->args[2] = arg_new_constant(ctx, 1); 2140 } else if (inv) { 2141 op2 = tcg_op_insert_after(ctx->tcg, op, xor_opc, 3); 2142 op2->args[0] = ret; 2143 op2->args[1] = ret; 2144 op2->args[2] = arg_new_constant(ctx, 1); 2145 } else if (neg) { 2146 op2 = tcg_op_insert_after(ctx->tcg, op, neg_opc, 2); 2147 op2->args[0] = ret; 2148 op2->args[1] = ret; 2149 } 2150 } 2151 2152 static bool fold_setcond(OptContext *ctx, TCGOp *op) 2153 { 2154 int i = do_constant_folding_cond1(ctx, op->args[0], &op->args[1], 2155 &op->args[2], &op->args[3]); 2156 if (i >= 0) { 2157 return tcg_opt_gen_movi(ctx, op, op->args[0], i); 2158 } 2159 fold_setcond_tst_pow2(ctx, op, false); 2160 2161 ctx->z_mask = 1; 2162 ctx->s_mask = smask_from_zmask(1); 2163 return false; 2164 } 2165 2166 static bool fold_negsetcond(OptContext *ctx, TCGOp *op) 2167 { 2168 int i = do_constant_folding_cond1(ctx, op->args[0], &op->args[1], 2169 &op->args[2], &op->args[3]); 2170 if (i >= 0) { 2171 return tcg_opt_gen_movi(ctx, op, op->args[0], -i); 2172 } 2173 fold_setcond_tst_pow2(ctx, op, true); 2174 2175 /* Value is {0,-1} so all bits are repetitions of the sign. */ 2176 ctx->s_mask = -1; 2177 return false; 2178 } 2179 2180 static bool fold_setcond2(OptContext *ctx, TCGOp *op) 2181 { 2182 TCGCond cond; 2183 int i, inv = 0; 2184 2185 i = do_constant_folding_cond2(ctx, &op->args[1]); 2186 cond = op->args[5]; 2187 if (i >= 0) { 2188 goto do_setcond_const; 2189 } 2190 2191 switch (cond) { 2192 case TCG_COND_LT: 2193 case TCG_COND_GE: 2194 /* 2195 * Simplify LT/GE comparisons vs zero to a single compare 2196 * vs the high word of the input. 2197 */ 2198 if (arg_is_const_val(op->args[3], 0) && 2199 arg_is_const_val(op->args[4], 0)) { 2200 goto do_setcond_high; 2201 } 2202 break; 2203 2204 case TCG_COND_NE: 2205 inv = 1; 2206 QEMU_FALLTHROUGH; 2207 case TCG_COND_EQ: 2208 /* 2209 * Simplify EQ/NE comparisons where one of the pairs 2210 * can be simplified. 2211 */ 2212 i = do_constant_folding_cond(TCG_TYPE_I32, op->args[1], 2213 op->args[3], cond); 2214 switch (i ^ inv) { 2215 case 0: 2216 goto do_setcond_const; 2217 case 1: 2218 goto do_setcond_high; 2219 } 2220 2221 i = do_constant_folding_cond(TCG_TYPE_I32, op->args[2], 2222 op->args[4], cond); 2223 switch (i ^ inv) { 2224 case 0: 2225 goto do_setcond_const; 2226 case 1: 2227 goto do_setcond_low; 2228 } 2229 break; 2230 2231 case TCG_COND_TSTEQ: 2232 case TCG_COND_TSTNE: 2233 if (arg_is_const_val(op->args[2], 0)) { 2234 goto do_setcond_high; 2235 } 2236 if (arg_is_const_val(op->args[4], 0)) { 2237 goto do_setcond_low; 2238 } 2239 break; 2240 2241 default: 2242 break; 2243 2244 do_setcond_low: 2245 op->args[2] = op->args[3]; 2246 op->args[3] = cond; 2247 op->opc = INDEX_op_setcond_i32; 2248 return fold_setcond(ctx, op); 2249 2250 do_setcond_high: 2251 op->args[1] = op->args[2]; 2252 op->args[2] = op->args[4]; 2253 op->args[3] = cond; 2254 op->opc = INDEX_op_setcond_i32; 2255 return fold_setcond(ctx, op); 2256 } 2257 2258 ctx->z_mask = 1; 2259 ctx->s_mask = smask_from_zmask(1); 2260 return false; 2261 2262 do_setcond_const: 2263 return tcg_opt_gen_movi(ctx, op, op->args[0], i); 2264 } 2265 2266 static bool fold_sextract(OptContext *ctx, TCGOp *op) 2267 { 2268 uint64_t z_mask, s_mask, s_mask_old; 2269 int pos = op->args[2]; 2270 int len = op->args[3]; 2271 2272 if (arg_is_const(op->args[1])) { 2273 uint64_t t; 2274 2275 t = arg_info(op->args[1])->val; 2276 t = sextract64(t, pos, len); 2277 return tcg_opt_gen_movi(ctx, op, op->args[0], t); 2278 } 2279 2280 z_mask = arg_info(op->args[1])->z_mask; 2281 z_mask = sextract64(z_mask, pos, len); 2282 ctx->z_mask = z_mask; 2283 2284 s_mask_old = arg_info(op->args[1])->s_mask; 2285 s_mask = sextract64(s_mask_old, pos, len); 2286 s_mask |= MAKE_64BIT_MASK(len, 64 - len); 2287 ctx->s_mask = s_mask; 2288 2289 if (pos == 0) { 2290 ctx->a_mask = s_mask & ~s_mask_old; 2291 } 2292 2293 return fold_masks(ctx, op); 2294 } 2295 2296 static bool fold_shift(OptContext *ctx, TCGOp *op) 2297 { 2298 uint64_t s_mask, z_mask, sign; 2299 2300 if (fold_const2(ctx, op) || 2301 fold_ix_to_i(ctx, op, 0) || 2302 fold_xi_to_x(ctx, op, 0)) { 2303 return true; 2304 } 2305 2306 s_mask = arg_info(op->args[1])->s_mask; 2307 z_mask = arg_info(op->args[1])->z_mask; 2308 2309 if (arg_is_const(op->args[2])) { 2310 int sh = arg_info(op->args[2])->val; 2311 2312 ctx->z_mask = do_constant_folding(op->opc, ctx->type, z_mask, sh); 2313 2314 s_mask = do_constant_folding(op->opc, ctx->type, s_mask, sh); 2315 ctx->s_mask = smask_from_smask(s_mask); 2316 2317 return fold_masks(ctx, op); 2318 } 2319 2320 switch (op->opc) { 2321 CASE_OP_32_64(sar): 2322 /* 2323 * Arithmetic right shift will not reduce the number of 2324 * input sign repetitions. 2325 */ 2326 ctx->s_mask = s_mask; 2327 break; 2328 CASE_OP_32_64(shr): 2329 /* 2330 * If the sign bit is known zero, then logical right shift 2331 * will not reduced the number of input sign repetitions. 2332 */ 2333 sign = (s_mask & -s_mask) >> 1; 2334 if (!(z_mask & sign)) { 2335 ctx->s_mask = s_mask; 2336 } 2337 break; 2338 default: 2339 break; 2340 } 2341 2342 return false; 2343 } 2344 2345 static bool fold_sub_to_neg(OptContext *ctx, TCGOp *op) 2346 { 2347 TCGOpcode neg_op; 2348 bool have_neg; 2349 2350 if (!arg_is_const(op->args[1]) || arg_info(op->args[1])->val != 0) { 2351 return false; 2352 } 2353 2354 switch (ctx->type) { 2355 case TCG_TYPE_I32: 2356 neg_op = INDEX_op_neg_i32; 2357 have_neg = true; 2358 break; 2359 case TCG_TYPE_I64: 2360 neg_op = INDEX_op_neg_i64; 2361 have_neg = true; 2362 break; 2363 case TCG_TYPE_V64: 2364 case TCG_TYPE_V128: 2365 case TCG_TYPE_V256: 2366 neg_op = INDEX_op_neg_vec; 2367 have_neg = (TCG_TARGET_HAS_neg_vec && 2368 tcg_can_emit_vec_op(neg_op, ctx->type, TCGOP_VECE(op)) > 0); 2369 break; 2370 default: 2371 g_assert_not_reached(); 2372 } 2373 if (have_neg) { 2374 op->opc = neg_op; 2375 op->args[1] = op->args[2]; 2376 return fold_neg(ctx, op); 2377 } 2378 return false; 2379 } 2380 2381 /* We cannot as yet do_constant_folding with vectors. */ 2382 static bool fold_sub_vec(OptContext *ctx, TCGOp *op) 2383 { 2384 if (fold_xx_to_i(ctx, op, 0) || 2385 fold_xi_to_x(ctx, op, 0) || 2386 fold_sub_to_neg(ctx, op)) { 2387 return true; 2388 } 2389 return false; 2390 } 2391 2392 static bool fold_sub(OptContext *ctx, TCGOp *op) 2393 { 2394 if (fold_const2(ctx, op) || fold_sub_vec(ctx, op)) { 2395 return true; 2396 } 2397 2398 /* Fold sub r,x,i to add r,x,-i */ 2399 if (arg_is_const(op->args[2])) { 2400 uint64_t val = arg_info(op->args[2])->val; 2401 2402 op->opc = (ctx->type == TCG_TYPE_I32 2403 ? INDEX_op_add_i32 : INDEX_op_add_i64); 2404 op->args[2] = arg_new_constant(ctx, -val); 2405 } 2406 return false; 2407 } 2408 2409 static bool fold_sub2(OptContext *ctx, TCGOp *op) 2410 { 2411 return fold_addsub2(ctx, op, false); 2412 } 2413 2414 static bool fold_tcg_ld(OptContext *ctx, TCGOp *op) 2415 { 2416 /* We can't do any folding with a load, but we can record bits. */ 2417 switch (op->opc) { 2418 CASE_OP_32_64(ld8s): 2419 ctx->s_mask = MAKE_64BIT_MASK(8, 56); 2420 break; 2421 CASE_OP_32_64(ld8u): 2422 ctx->z_mask = MAKE_64BIT_MASK(0, 8); 2423 ctx->s_mask = MAKE_64BIT_MASK(9, 55); 2424 break; 2425 CASE_OP_32_64(ld16s): 2426 ctx->s_mask = MAKE_64BIT_MASK(16, 48); 2427 break; 2428 CASE_OP_32_64(ld16u): 2429 ctx->z_mask = MAKE_64BIT_MASK(0, 16); 2430 ctx->s_mask = MAKE_64BIT_MASK(17, 47); 2431 break; 2432 case INDEX_op_ld32s_i64: 2433 ctx->s_mask = MAKE_64BIT_MASK(32, 32); 2434 break; 2435 case INDEX_op_ld32u_i64: 2436 ctx->z_mask = MAKE_64BIT_MASK(0, 32); 2437 ctx->s_mask = MAKE_64BIT_MASK(33, 31); 2438 break; 2439 default: 2440 g_assert_not_reached(); 2441 } 2442 return false; 2443 } 2444 2445 static bool fold_tcg_ld_memcopy(OptContext *ctx, TCGOp *op) 2446 { 2447 TCGTemp *dst, *src; 2448 intptr_t ofs; 2449 TCGType type; 2450 2451 if (op->args[1] != tcgv_ptr_arg(tcg_env)) { 2452 return false; 2453 } 2454 2455 type = ctx->type; 2456 ofs = op->args[2]; 2457 dst = arg_temp(op->args[0]); 2458 src = find_mem_copy_for(ctx, type, ofs); 2459 if (src && src->base_type == type) { 2460 return tcg_opt_gen_mov(ctx, op, temp_arg(dst), temp_arg(src)); 2461 } 2462 2463 reset_ts(ctx, dst); 2464 record_mem_copy(ctx, type, dst, ofs, ofs + tcg_type_size(type) - 1); 2465 return true; 2466 } 2467 2468 static bool fold_tcg_st(OptContext *ctx, TCGOp *op) 2469 { 2470 intptr_t ofs = op->args[2]; 2471 intptr_t lm1; 2472 2473 if (op->args[1] != tcgv_ptr_arg(tcg_env)) { 2474 remove_mem_copy_all(ctx); 2475 return false; 2476 } 2477 2478 switch (op->opc) { 2479 CASE_OP_32_64(st8): 2480 lm1 = 0; 2481 break; 2482 CASE_OP_32_64(st16): 2483 lm1 = 1; 2484 break; 2485 case INDEX_op_st32_i64: 2486 case INDEX_op_st_i32: 2487 lm1 = 3; 2488 break; 2489 case INDEX_op_st_i64: 2490 lm1 = 7; 2491 break; 2492 case INDEX_op_st_vec: 2493 lm1 = tcg_type_size(ctx->type) - 1; 2494 break; 2495 default: 2496 g_assert_not_reached(); 2497 } 2498 remove_mem_copy_in(ctx, ofs, ofs + lm1); 2499 return false; 2500 } 2501 2502 static bool fold_tcg_st_memcopy(OptContext *ctx, TCGOp *op) 2503 { 2504 TCGTemp *src; 2505 intptr_t ofs, last; 2506 TCGType type; 2507 2508 if (op->args[1] != tcgv_ptr_arg(tcg_env)) { 2509 fold_tcg_st(ctx, op); 2510 return false; 2511 } 2512 2513 src = arg_temp(op->args[0]); 2514 ofs = op->args[2]; 2515 type = ctx->type; 2516 2517 /* 2518 * Eliminate duplicate stores of a constant. 2519 * This happens frequently when the target ISA zero-extends. 2520 */ 2521 if (ts_is_const(src)) { 2522 TCGTemp *prev = find_mem_copy_for(ctx, type, ofs); 2523 if (src == prev) { 2524 tcg_op_remove(ctx->tcg, op); 2525 return true; 2526 } 2527 } 2528 2529 last = ofs + tcg_type_size(type) - 1; 2530 remove_mem_copy_in(ctx, ofs, last); 2531 record_mem_copy(ctx, type, src, ofs, last); 2532 return false; 2533 } 2534 2535 static bool fold_xor(OptContext *ctx, TCGOp *op) 2536 { 2537 if (fold_const2_commutative(ctx, op) || 2538 fold_xx_to_i(ctx, op, 0) || 2539 fold_xi_to_x(ctx, op, 0) || 2540 fold_xi_to_not(ctx, op, -1)) { 2541 return true; 2542 } 2543 2544 ctx->z_mask = arg_info(op->args[1])->z_mask 2545 | arg_info(op->args[2])->z_mask; 2546 ctx->s_mask = arg_info(op->args[1])->s_mask 2547 & arg_info(op->args[2])->s_mask; 2548 return fold_masks(ctx, op); 2549 } 2550 2551 /* Propagate constants and copies, fold constant expressions. */ 2552 void tcg_optimize(TCGContext *s) 2553 { 2554 int nb_temps, i; 2555 TCGOp *op, *op_next; 2556 OptContext ctx = { .tcg = s }; 2557 2558 QSIMPLEQ_INIT(&ctx.mem_free); 2559 2560 /* Array VALS has an element for each temp. 2561 If this temp holds a constant then its value is kept in VALS' element. 2562 If this temp is a copy of other ones then the other copies are 2563 available through the doubly linked circular list. */ 2564 2565 nb_temps = s->nb_temps; 2566 for (i = 0; i < nb_temps; ++i) { 2567 s->temps[i].state_ptr = NULL; 2568 } 2569 2570 QTAILQ_FOREACH_SAFE(op, &s->ops, link, op_next) { 2571 TCGOpcode opc = op->opc; 2572 const TCGOpDef *def; 2573 bool done = false; 2574 2575 /* Calls are special. */ 2576 if (opc == INDEX_op_call) { 2577 fold_call(&ctx, op); 2578 continue; 2579 } 2580 2581 def = &tcg_op_defs[opc]; 2582 init_arguments(&ctx, op, def->nb_oargs + def->nb_iargs); 2583 copy_propagate(&ctx, op, def->nb_oargs, def->nb_iargs); 2584 2585 /* Pre-compute the type of the operation. */ 2586 if (def->flags & TCG_OPF_VECTOR) { 2587 ctx.type = TCG_TYPE_V64 + TCGOP_VECL(op); 2588 } else if (def->flags & TCG_OPF_64BIT) { 2589 ctx.type = TCG_TYPE_I64; 2590 } else { 2591 ctx.type = TCG_TYPE_I32; 2592 } 2593 2594 /* Assume all bits affected, no bits known zero, no sign reps. */ 2595 ctx.a_mask = -1; 2596 ctx.z_mask = -1; 2597 ctx.s_mask = 0; 2598 2599 /* 2600 * Process each opcode. 2601 * Sorted alphabetically by opcode as much as possible. 2602 */ 2603 switch (opc) { 2604 CASE_OP_32_64(add): 2605 done = fold_add(&ctx, op); 2606 break; 2607 case INDEX_op_add_vec: 2608 done = fold_add_vec(&ctx, op); 2609 break; 2610 CASE_OP_32_64(add2): 2611 done = fold_add2(&ctx, op); 2612 break; 2613 CASE_OP_32_64_VEC(and): 2614 done = fold_and(&ctx, op); 2615 break; 2616 CASE_OP_32_64_VEC(andc): 2617 done = fold_andc(&ctx, op); 2618 break; 2619 CASE_OP_32_64(brcond): 2620 done = fold_brcond(&ctx, op); 2621 break; 2622 case INDEX_op_brcond2_i32: 2623 done = fold_brcond2(&ctx, op); 2624 break; 2625 CASE_OP_32_64(bswap16): 2626 CASE_OP_32_64(bswap32): 2627 case INDEX_op_bswap64_i64: 2628 done = fold_bswap(&ctx, op); 2629 break; 2630 CASE_OP_32_64(clz): 2631 CASE_OP_32_64(ctz): 2632 done = fold_count_zeros(&ctx, op); 2633 break; 2634 CASE_OP_32_64(ctpop): 2635 done = fold_ctpop(&ctx, op); 2636 break; 2637 CASE_OP_32_64(deposit): 2638 done = fold_deposit(&ctx, op); 2639 break; 2640 CASE_OP_32_64(div): 2641 CASE_OP_32_64(divu): 2642 done = fold_divide(&ctx, op); 2643 break; 2644 case INDEX_op_dup_vec: 2645 done = fold_dup(&ctx, op); 2646 break; 2647 case INDEX_op_dup2_vec: 2648 done = fold_dup2(&ctx, op); 2649 break; 2650 CASE_OP_32_64_VEC(eqv): 2651 done = fold_eqv(&ctx, op); 2652 break; 2653 CASE_OP_32_64(extract): 2654 done = fold_extract(&ctx, op); 2655 break; 2656 CASE_OP_32_64(extract2): 2657 done = fold_extract2(&ctx, op); 2658 break; 2659 CASE_OP_32_64(ext8s): 2660 CASE_OP_32_64(ext16s): 2661 case INDEX_op_ext32s_i64: 2662 case INDEX_op_ext_i32_i64: 2663 done = fold_exts(&ctx, op); 2664 break; 2665 CASE_OP_32_64(ext8u): 2666 CASE_OP_32_64(ext16u): 2667 case INDEX_op_ext32u_i64: 2668 case INDEX_op_extu_i32_i64: 2669 case INDEX_op_extrl_i64_i32: 2670 case INDEX_op_extrh_i64_i32: 2671 done = fold_extu(&ctx, op); 2672 break; 2673 CASE_OP_32_64(ld8s): 2674 CASE_OP_32_64(ld8u): 2675 CASE_OP_32_64(ld16s): 2676 CASE_OP_32_64(ld16u): 2677 case INDEX_op_ld32s_i64: 2678 case INDEX_op_ld32u_i64: 2679 done = fold_tcg_ld(&ctx, op); 2680 break; 2681 case INDEX_op_ld_i32: 2682 case INDEX_op_ld_i64: 2683 case INDEX_op_ld_vec: 2684 done = fold_tcg_ld_memcopy(&ctx, op); 2685 break; 2686 CASE_OP_32_64(st8): 2687 CASE_OP_32_64(st16): 2688 case INDEX_op_st32_i64: 2689 done = fold_tcg_st(&ctx, op); 2690 break; 2691 case INDEX_op_st_i32: 2692 case INDEX_op_st_i64: 2693 case INDEX_op_st_vec: 2694 done = fold_tcg_st_memcopy(&ctx, op); 2695 break; 2696 case INDEX_op_mb: 2697 done = fold_mb(&ctx, op); 2698 break; 2699 CASE_OP_32_64_VEC(mov): 2700 done = fold_mov(&ctx, op); 2701 break; 2702 CASE_OP_32_64(movcond): 2703 done = fold_movcond(&ctx, op); 2704 break; 2705 CASE_OP_32_64(mul): 2706 done = fold_mul(&ctx, op); 2707 break; 2708 CASE_OP_32_64(mulsh): 2709 CASE_OP_32_64(muluh): 2710 done = fold_mul_highpart(&ctx, op); 2711 break; 2712 CASE_OP_32_64(muls2): 2713 CASE_OP_32_64(mulu2): 2714 done = fold_multiply2(&ctx, op); 2715 break; 2716 CASE_OP_32_64_VEC(nand): 2717 done = fold_nand(&ctx, op); 2718 break; 2719 CASE_OP_32_64(neg): 2720 done = fold_neg(&ctx, op); 2721 break; 2722 CASE_OP_32_64_VEC(nor): 2723 done = fold_nor(&ctx, op); 2724 break; 2725 CASE_OP_32_64_VEC(not): 2726 done = fold_not(&ctx, op); 2727 break; 2728 CASE_OP_32_64_VEC(or): 2729 done = fold_or(&ctx, op); 2730 break; 2731 CASE_OP_32_64_VEC(orc): 2732 done = fold_orc(&ctx, op); 2733 break; 2734 case INDEX_op_qemu_ld_a32_i32: 2735 case INDEX_op_qemu_ld_a64_i32: 2736 case INDEX_op_qemu_ld_a32_i64: 2737 case INDEX_op_qemu_ld_a64_i64: 2738 case INDEX_op_qemu_ld_a32_i128: 2739 case INDEX_op_qemu_ld_a64_i128: 2740 done = fold_qemu_ld(&ctx, op); 2741 break; 2742 case INDEX_op_qemu_st8_a32_i32: 2743 case INDEX_op_qemu_st8_a64_i32: 2744 case INDEX_op_qemu_st_a32_i32: 2745 case INDEX_op_qemu_st_a64_i32: 2746 case INDEX_op_qemu_st_a32_i64: 2747 case INDEX_op_qemu_st_a64_i64: 2748 case INDEX_op_qemu_st_a32_i128: 2749 case INDEX_op_qemu_st_a64_i128: 2750 done = fold_qemu_st(&ctx, op); 2751 break; 2752 CASE_OP_32_64(rem): 2753 CASE_OP_32_64(remu): 2754 done = fold_remainder(&ctx, op); 2755 break; 2756 CASE_OP_32_64(rotl): 2757 CASE_OP_32_64(rotr): 2758 CASE_OP_32_64(sar): 2759 CASE_OP_32_64(shl): 2760 CASE_OP_32_64(shr): 2761 done = fold_shift(&ctx, op); 2762 break; 2763 CASE_OP_32_64(setcond): 2764 done = fold_setcond(&ctx, op); 2765 break; 2766 CASE_OP_32_64(negsetcond): 2767 done = fold_negsetcond(&ctx, op); 2768 break; 2769 case INDEX_op_setcond2_i32: 2770 done = fold_setcond2(&ctx, op); 2771 break; 2772 CASE_OP_32_64(sextract): 2773 done = fold_sextract(&ctx, op); 2774 break; 2775 CASE_OP_32_64(sub): 2776 done = fold_sub(&ctx, op); 2777 break; 2778 case INDEX_op_sub_vec: 2779 done = fold_sub_vec(&ctx, op); 2780 break; 2781 CASE_OP_32_64(sub2): 2782 done = fold_sub2(&ctx, op); 2783 break; 2784 CASE_OP_32_64_VEC(xor): 2785 done = fold_xor(&ctx, op); 2786 break; 2787 default: 2788 break; 2789 } 2790 2791 if (!done) { 2792 finish_folding(&ctx, op); 2793 } 2794 } 2795 } 2796