axis-dev88: Connect the PIC upstream IRQs directly to the CPU
[qemu/ar7.git] / target-i386 / translate.c
blobb0f227915ab37035f0f7913dd756ffb738abca9b
1 /*
2 * i386 translation
4 * Copyright (c) 2003 Fabrice Bellard
6 * This library is free software; you can redistribute it and/or
7 * modify it under the terms of the GNU Lesser General Public
8 * License as published by the Free Software Foundation; either
9 * version 2 of the License, or (at your option) any later version.
11 * This library is distributed in the hope that it will be useful,
12 * but WITHOUT ANY WARRANTY; without even the implied warranty of
13 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
14 * Lesser General Public License for more details.
16 * You should have received a copy of the GNU Lesser General Public
17 * License along with this library; if not, see <http://www.gnu.org/licenses/>.
19 #include <stdarg.h>
20 #include <stdlib.h>
21 #include <stdio.h>
22 #include <string.h>
23 #include <inttypes.h>
24 #include <signal.h>
26 #include "qemu/host-utils.h"
27 #include "cpu.h"
28 #include "disas/disas.h"
29 #include "tcg-op.h"
31 #include "helper.h"
32 #define GEN_HELPER 1
33 #include "helper.h"
35 #define PREFIX_REPZ 0x01
36 #define PREFIX_REPNZ 0x02
37 #define PREFIX_LOCK 0x04
38 #define PREFIX_DATA 0x08
39 #define PREFIX_ADR 0x10
40 #define PREFIX_VEX 0x20
42 #ifdef TARGET_X86_64
43 #define CODE64(s) ((s)->code64)
44 #define REX_X(s) ((s)->rex_x)
45 #define REX_B(s) ((s)->rex_b)
46 #else
47 #define CODE64(s) 0
48 #define REX_X(s) 0
49 #define REX_B(s) 0
50 #endif
52 #ifdef TARGET_X86_64
53 # define ctztl ctz64
54 # define clztl clz64
55 #else
56 # define ctztl ctz32
57 # define clztl clz32
58 #endif
60 //#define MACRO_TEST 1
62 /* global register indexes */
63 static TCGv_ptr cpu_env;
64 static TCGv cpu_A0;
65 static TCGv cpu_cc_dst, cpu_cc_src, cpu_cc_src2, cpu_cc_srcT;
66 static TCGv_i32 cpu_cc_op;
67 static TCGv cpu_regs[CPU_NB_REGS];
68 /* local temps */
69 static TCGv cpu_T[2];
70 /* local register indexes (only used inside old micro ops) */
71 static TCGv cpu_tmp0, cpu_tmp4;
72 static TCGv_ptr cpu_ptr0, cpu_ptr1;
73 static TCGv_i32 cpu_tmp2_i32, cpu_tmp3_i32;
74 static TCGv_i64 cpu_tmp1_i64;
76 static uint8_t gen_opc_cc_op[OPC_BUF_SIZE];
78 #include "exec/gen-icount.h"
80 #ifdef TARGET_X86_64
81 static int x86_64_hregs;
82 #endif
84 typedef struct DisasContext {
85 /* current insn context */
86 int override; /* -1 if no override */
87 int prefix;
88 TCGMemOp aflag;
89 TCGMemOp dflag;
90 target_ulong pc; /* pc = eip + cs_base */
91 int is_jmp; /* 1 = means jump (stop translation), 2 means CPU
92 static state change (stop translation) */
93 /* current block context */
94 target_ulong cs_base; /* base of CS segment */
95 int pe; /* protected mode */
96 int code32; /* 32 bit code segment */
97 #ifdef TARGET_X86_64
98 int lma; /* long mode active */
99 int code64; /* 64 bit code segment */
100 int rex_x, rex_b;
101 #endif
102 int vex_l; /* vex vector length */
103 int vex_v; /* vex vvvv register, without 1's compliment. */
104 int ss32; /* 32 bit stack segment */
105 CCOp cc_op; /* current CC operation */
106 bool cc_op_dirty;
107 int addseg; /* non zero if either DS/ES/SS have a non zero base */
108 int f_st; /* currently unused */
109 int vm86; /* vm86 mode */
110 int cpl;
111 int iopl;
112 int tf; /* TF cpu flag */
113 int singlestep_enabled; /* "hardware" single step enabled */
114 int jmp_opt; /* use direct block chaining for direct jumps */
115 int mem_index; /* select memory access functions */
116 uint64_t flags; /* all execution flags */
117 struct TranslationBlock *tb;
118 int popl_esp_hack; /* for correct popl with esp base handling */
119 int rip_offset; /* only used in x86_64, but left for simplicity */
120 int cpuid_features;
121 int cpuid_ext_features;
122 int cpuid_ext2_features;
123 int cpuid_ext3_features;
124 int cpuid_7_0_ebx_features;
125 } DisasContext;
127 static void gen_eob(DisasContext *s);
128 static void gen_jmp(DisasContext *s, target_ulong eip);
129 static void gen_jmp_tb(DisasContext *s, target_ulong eip, int tb_num);
130 static void gen_op(DisasContext *s1, int op, TCGMemOp ot, int d);
132 /* i386 arith/logic operations */
133 enum {
134 OP_ADDL,
135 OP_ORL,
136 OP_ADCL,
137 OP_SBBL,
138 OP_ANDL,
139 OP_SUBL,
140 OP_XORL,
141 OP_CMPL,
144 /* i386 shift ops */
145 enum {
146 OP_ROL,
147 OP_ROR,
148 OP_RCL,
149 OP_RCR,
150 OP_SHL,
151 OP_SHR,
152 OP_SHL1, /* undocumented */
153 OP_SAR = 7,
156 enum {
157 JCC_O,
158 JCC_B,
159 JCC_Z,
160 JCC_BE,
161 JCC_S,
162 JCC_P,
163 JCC_L,
164 JCC_LE,
167 enum {
168 /* I386 int registers */
169 OR_EAX, /* MUST be even numbered */
170 OR_ECX,
171 OR_EDX,
172 OR_EBX,
173 OR_ESP,
174 OR_EBP,
175 OR_ESI,
176 OR_EDI,
178 OR_TMP0 = 16, /* temporary operand register */
179 OR_TMP1,
180 OR_A0, /* temporary register used when doing address evaluation */
183 enum {
184 USES_CC_DST = 1,
185 USES_CC_SRC = 2,
186 USES_CC_SRC2 = 4,
187 USES_CC_SRCT = 8,
190 /* Bit set if the global variable is live after setting CC_OP to X. */
191 static const uint8_t cc_op_live[CC_OP_NB] = {
192 [CC_OP_DYNAMIC] = USES_CC_DST | USES_CC_SRC | USES_CC_SRC2,
193 [CC_OP_EFLAGS] = USES_CC_SRC,
194 [CC_OP_MULB ... CC_OP_MULQ] = USES_CC_DST | USES_CC_SRC,
195 [CC_OP_ADDB ... CC_OP_ADDQ] = USES_CC_DST | USES_CC_SRC,
196 [CC_OP_ADCB ... CC_OP_ADCQ] = USES_CC_DST | USES_CC_SRC | USES_CC_SRC2,
197 [CC_OP_SUBB ... CC_OP_SUBQ] = USES_CC_DST | USES_CC_SRC | USES_CC_SRCT,
198 [CC_OP_SBBB ... CC_OP_SBBQ] = USES_CC_DST | USES_CC_SRC | USES_CC_SRC2,
199 [CC_OP_LOGICB ... CC_OP_LOGICQ] = USES_CC_DST,
200 [CC_OP_INCB ... CC_OP_INCQ] = USES_CC_DST | USES_CC_SRC,
201 [CC_OP_DECB ... CC_OP_DECQ] = USES_CC_DST | USES_CC_SRC,
202 [CC_OP_SHLB ... CC_OP_SHLQ] = USES_CC_DST | USES_CC_SRC,
203 [CC_OP_SARB ... CC_OP_SARQ] = USES_CC_DST | USES_CC_SRC,
204 [CC_OP_BMILGB ... CC_OP_BMILGQ] = USES_CC_DST | USES_CC_SRC,
205 [CC_OP_ADCX] = USES_CC_DST | USES_CC_SRC,
206 [CC_OP_ADOX] = USES_CC_SRC | USES_CC_SRC2,
207 [CC_OP_ADCOX] = USES_CC_DST | USES_CC_SRC | USES_CC_SRC2,
208 [CC_OP_CLR] = 0,
211 static void set_cc_op(DisasContext *s, CCOp op)
213 int dead;
215 if (s->cc_op == op) {
216 return;
219 /* Discard CC computation that will no longer be used. */
220 dead = cc_op_live[s->cc_op] & ~cc_op_live[op];
221 if (dead & USES_CC_DST) {
222 tcg_gen_discard_tl(cpu_cc_dst);
224 if (dead & USES_CC_SRC) {
225 tcg_gen_discard_tl(cpu_cc_src);
227 if (dead & USES_CC_SRC2) {
228 tcg_gen_discard_tl(cpu_cc_src2);
230 if (dead & USES_CC_SRCT) {
231 tcg_gen_discard_tl(cpu_cc_srcT);
234 if (op == CC_OP_DYNAMIC) {
235 /* The DYNAMIC setting is translator only, and should never be
236 stored. Thus we always consider it clean. */
237 s->cc_op_dirty = false;
238 } else {
239 /* Discard any computed CC_OP value (see shifts). */
240 if (s->cc_op == CC_OP_DYNAMIC) {
241 tcg_gen_discard_i32(cpu_cc_op);
243 s->cc_op_dirty = true;
245 s->cc_op = op;
248 static void gen_update_cc_op(DisasContext *s)
250 if (s->cc_op_dirty) {
251 tcg_gen_movi_i32(cpu_cc_op, s->cc_op);
252 s->cc_op_dirty = false;
256 #ifdef TARGET_X86_64
258 #define NB_OP_SIZES 4
260 #else /* !TARGET_X86_64 */
262 #define NB_OP_SIZES 3
264 #endif /* !TARGET_X86_64 */
266 #if defined(HOST_WORDS_BIGENDIAN)
267 #define REG_B_OFFSET (sizeof(target_ulong) - 1)
268 #define REG_H_OFFSET (sizeof(target_ulong) - 2)
269 #define REG_W_OFFSET (sizeof(target_ulong) - 2)
270 #define REG_L_OFFSET (sizeof(target_ulong) - 4)
271 #define REG_LH_OFFSET (sizeof(target_ulong) - 8)
272 #else
273 #define REG_B_OFFSET 0
274 #define REG_H_OFFSET 1
275 #define REG_W_OFFSET 0
276 #define REG_L_OFFSET 0
277 #define REG_LH_OFFSET 4
278 #endif
280 /* In instruction encodings for byte register accesses the
281 * register number usually indicates "low 8 bits of register N";
282 * however there are some special cases where N 4..7 indicates
283 * [AH, CH, DH, BH], ie "bits 15..8 of register N-4". Return
284 * true for this special case, false otherwise.
286 static inline bool byte_reg_is_xH(int reg)
288 if (reg < 4) {
289 return false;
291 #ifdef TARGET_X86_64
292 if (reg >= 8 || x86_64_hregs) {
293 return false;
295 #endif
296 return true;
299 /* Select the size of a push/pop operation. */
300 static inline TCGMemOp mo_pushpop(DisasContext *s, TCGMemOp ot)
302 if (CODE64(s)) {
303 return ot == MO_16 ? MO_16 : MO_64;
304 } else {
305 return ot;
309 /* Select only size 64 else 32. Used for SSE operand sizes. */
310 static inline TCGMemOp mo_64_32(TCGMemOp ot)
312 #ifdef TARGET_X86_64
313 return ot == MO_64 ? MO_64 : MO_32;
314 #else
315 return MO_32;
316 #endif
319 /* Select size 8 if lsb of B is clear, else OT. Used for decoding
320 byte vs word opcodes. */
321 static inline TCGMemOp mo_b_d(int b, TCGMemOp ot)
323 return b & 1 ? ot : MO_8;
326 /* Select size 8 if lsb of B is clear, else OT capped at 32.
327 Used for decoding operand size of port opcodes. */
328 static inline TCGMemOp mo_b_d32(int b, TCGMemOp ot)
330 return b & 1 ? (ot == MO_16 ? MO_16 : MO_32) : MO_8;
333 static void gen_op_mov_reg_v(TCGMemOp ot, int reg, TCGv t0)
335 switch(ot) {
336 case MO_8:
337 if (!byte_reg_is_xH(reg)) {
338 tcg_gen_deposit_tl(cpu_regs[reg], cpu_regs[reg], t0, 0, 8);
339 } else {
340 tcg_gen_deposit_tl(cpu_regs[reg - 4], cpu_regs[reg - 4], t0, 8, 8);
342 break;
343 case MO_16:
344 tcg_gen_deposit_tl(cpu_regs[reg], cpu_regs[reg], t0, 0, 16);
345 break;
346 case MO_32:
347 /* For x86_64, this sets the higher half of register to zero.
348 For i386, this is equivalent to a mov. */
349 tcg_gen_ext32u_tl(cpu_regs[reg], t0);
350 break;
351 #ifdef TARGET_X86_64
352 case MO_64:
353 tcg_gen_mov_tl(cpu_regs[reg], t0);
354 break;
355 #endif
356 default:
357 tcg_abort();
361 static inline void gen_op_mov_v_reg(TCGMemOp ot, TCGv t0, int reg)
363 if (ot == MO_8 && byte_reg_is_xH(reg)) {
364 tcg_gen_shri_tl(t0, cpu_regs[reg - 4], 8);
365 tcg_gen_ext8u_tl(t0, t0);
366 } else {
367 tcg_gen_mov_tl(t0, cpu_regs[reg]);
371 static inline void gen_op_movl_A0_reg(int reg)
373 tcg_gen_mov_tl(cpu_A0, cpu_regs[reg]);
376 static inline void gen_op_addl_A0_im(int32_t val)
378 tcg_gen_addi_tl(cpu_A0, cpu_A0, val);
379 #ifdef TARGET_X86_64
380 tcg_gen_andi_tl(cpu_A0, cpu_A0, 0xffffffff);
381 #endif
384 #ifdef TARGET_X86_64
385 static inline void gen_op_addq_A0_im(int64_t val)
387 tcg_gen_addi_tl(cpu_A0, cpu_A0, val);
389 #endif
391 static void gen_add_A0_im(DisasContext *s, int val)
393 #ifdef TARGET_X86_64
394 if (CODE64(s))
395 gen_op_addq_A0_im(val);
396 else
397 #endif
398 gen_op_addl_A0_im(val);
401 static inline void gen_op_jmp_v(TCGv dest)
403 tcg_gen_st_tl(dest, cpu_env, offsetof(CPUX86State, eip));
406 static inline void gen_op_add_reg_im(TCGMemOp size, int reg, int32_t val)
408 tcg_gen_addi_tl(cpu_tmp0, cpu_regs[reg], val);
409 gen_op_mov_reg_v(size, reg, cpu_tmp0);
412 static inline void gen_op_add_reg_T0(TCGMemOp size, int reg)
414 tcg_gen_add_tl(cpu_tmp0, cpu_regs[reg], cpu_T[0]);
415 gen_op_mov_reg_v(size, reg, cpu_tmp0);
418 static inline void gen_op_addl_A0_reg_sN(int shift, int reg)
420 tcg_gen_mov_tl(cpu_tmp0, cpu_regs[reg]);
421 if (shift != 0)
422 tcg_gen_shli_tl(cpu_tmp0, cpu_tmp0, shift);
423 tcg_gen_add_tl(cpu_A0, cpu_A0, cpu_tmp0);
424 /* For x86_64, this sets the higher half of register to zero.
425 For i386, this is equivalent to a nop. */
426 tcg_gen_ext32u_tl(cpu_A0, cpu_A0);
429 static inline void gen_op_movl_A0_seg(int reg)
431 tcg_gen_ld32u_tl(cpu_A0, cpu_env, offsetof(CPUX86State, segs[reg].base) + REG_L_OFFSET);
434 static inline void gen_op_addl_A0_seg(DisasContext *s, int reg)
436 tcg_gen_ld_tl(cpu_tmp0, cpu_env, offsetof(CPUX86State, segs[reg].base));
437 #ifdef TARGET_X86_64
438 if (CODE64(s)) {
439 tcg_gen_andi_tl(cpu_A0, cpu_A0, 0xffffffff);
440 tcg_gen_add_tl(cpu_A0, cpu_A0, cpu_tmp0);
441 } else {
442 tcg_gen_add_tl(cpu_A0, cpu_A0, cpu_tmp0);
443 tcg_gen_andi_tl(cpu_A0, cpu_A0, 0xffffffff);
445 #else
446 tcg_gen_add_tl(cpu_A0, cpu_A0, cpu_tmp0);
447 #endif
450 #ifdef TARGET_X86_64
451 static inline void gen_op_movq_A0_seg(int reg)
453 tcg_gen_ld_tl(cpu_A0, cpu_env, offsetof(CPUX86State, segs[reg].base));
456 static inline void gen_op_addq_A0_seg(int reg)
458 tcg_gen_ld_tl(cpu_tmp0, cpu_env, offsetof(CPUX86State, segs[reg].base));
459 tcg_gen_add_tl(cpu_A0, cpu_A0, cpu_tmp0);
462 static inline void gen_op_movq_A0_reg(int reg)
464 tcg_gen_mov_tl(cpu_A0, cpu_regs[reg]);
467 static inline void gen_op_addq_A0_reg_sN(int shift, int reg)
469 tcg_gen_mov_tl(cpu_tmp0, cpu_regs[reg]);
470 if (shift != 0)
471 tcg_gen_shli_tl(cpu_tmp0, cpu_tmp0, shift);
472 tcg_gen_add_tl(cpu_A0, cpu_A0, cpu_tmp0);
474 #endif
476 static inline void gen_op_ld_v(DisasContext *s, int idx, TCGv t0, TCGv a0)
478 tcg_gen_qemu_ld_tl(t0, a0, s->mem_index, idx | MO_LE);
481 static inline void gen_op_st_v(DisasContext *s, int idx, TCGv t0, TCGv a0)
483 tcg_gen_qemu_st_tl(t0, a0, s->mem_index, idx | MO_LE);
486 static inline void gen_op_st_rm_T0_A0(DisasContext *s, int idx, int d)
488 if (d == OR_TMP0) {
489 gen_op_st_v(s, idx, cpu_T[0], cpu_A0);
490 } else {
491 gen_op_mov_reg_v(idx, d, cpu_T[0]);
495 static inline void gen_jmp_im(target_ulong pc)
497 tcg_gen_movi_tl(cpu_tmp0, pc);
498 gen_op_jmp_v(cpu_tmp0);
501 static inline void gen_string_movl_A0_ESI(DisasContext *s)
503 int override;
505 override = s->override;
506 switch (s->aflag) {
507 #ifdef TARGET_X86_64
508 case MO_64:
509 if (override >= 0) {
510 gen_op_movq_A0_seg(override);
511 gen_op_addq_A0_reg_sN(0, R_ESI);
512 } else {
513 gen_op_movq_A0_reg(R_ESI);
515 break;
516 #endif
517 case MO_32:
518 /* 32 bit address */
519 if (s->addseg && override < 0)
520 override = R_DS;
521 if (override >= 0) {
522 gen_op_movl_A0_seg(override);
523 gen_op_addl_A0_reg_sN(0, R_ESI);
524 } else {
525 gen_op_movl_A0_reg(R_ESI);
527 break;
528 case MO_16:
529 /* 16 address, always override */
530 if (override < 0)
531 override = R_DS;
532 tcg_gen_ext16u_tl(cpu_A0, cpu_regs[R_ESI]);
533 gen_op_addl_A0_seg(s, override);
534 break;
535 default:
536 tcg_abort();
540 static inline void gen_string_movl_A0_EDI(DisasContext *s)
542 switch (s->aflag) {
543 #ifdef TARGET_X86_64
544 case MO_64:
545 gen_op_movq_A0_reg(R_EDI);
546 break;
547 #endif
548 case MO_32:
549 if (s->addseg) {
550 gen_op_movl_A0_seg(R_ES);
551 gen_op_addl_A0_reg_sN(0, R_EDI);
552 } else {
553 gen_op_movl_A0_reg(R_EDI);
555 break;
556 case MO_16:
557 tcg_gen_ext16u_tl(cpu_A0, cpu_regs[R_EDI]);
558 gen_op_addl_A0_seg(s, R_ES);
559 break;
560 default:
561 tcg_abort();
565 static inline void gen_op_movl_T0_Dshift(TCGMemOp ot)
567 tcg_gen_ld32s_tl(cpu_T[0], cpu_env, offsetof(CPUX86State, df));
568 tcg_gen_shli_tl(cpu_T[0], cpu_T[0], ot);
571 static TCGv gen_ext_tl(TCGv dst, TCGv src, TCGMemOp size, bool sign)
573 switch (size) {
574 case MO_8:
575 if (sign) {
576 tcg_gen_ext8s_tl(dst, src);
577 } else {
578 tcg_gen_ext8u_tl(dst, src);
580 return dst;
581 case MO_16:
582 if (sign) {
583 tcg_gen_ext16s_tl(dst, src);
584 } else {
585 tcg_gen_ext16u_tl(dst, src);
587 return dst;
588 #ifdef TARGET_X86_64
589 case MO_32:
590 if (sign) {
591 tcg_gen_ext32s_tl(dst, src);
592 } else {
593 tcg_gen_ext32u_tl(dst, src);
595 return dst;
596 #endif
597 default:
598 return src;
602 static void gen_extu(TCGMemOp ot, TCGv reg)
604 gen_ext_tl(reg, reg, ot, false);
607 static void gen_exts(TCGMemOp ot, TCGv reg)
609 gen_ext_tl(reg, reg, ot, true);
612 static inline void gen_op_jnz_ecx(TCGMemOp size, int label1)
614 tcg_gen_mov_tl(cpu_tmp0, cpu_regs[R_ECX]);
615 gen_extu(size, cpu_tmp0);
616 tcg_gen_brcondi_tl(TCG_COND_NE, cpu_tmp0, 0, label1);
619 static inline void gen_op_jz_ecx(TCGMemOp size, int label1)
621 tcg_gen_mov_tl(cpu_tmp0, cpu_regs[R_ECX]);
622 gen_extu(size, cpu_tmp0);
623 tcg_gen_brcondi_tl(TCG_COND_EQ, cpu_tmp0, 0, label1);
626 static void gen_helper_in_func(TCGMemOp ot, TCGv v, TCGv_i32 n)
628 switch (ot) {
629 case MO_8:
630 gen_helper_inb(v, n);
631 break;
632 case MO_16:
633 gen_helper_inw(v, n);
634 break;
635 case MO_32:
636 gen_helper_inl(v, n);
637 break;
638 default:
639 tcg_abort();
643 static void gen_helper_out_func(TCGMemOp ot, TCGv_i32 v, TCGv_i32 n)
645 switch (ot) {
646 case MO_8:
647 gen_helper_outb(v, n);
648 break;
649 case MO_16:
650 gen_helper_outw(v, n);
651 break;
652 case MO_32:
653 gen_helper_outl(v, n);
654 break;
655 default:
656 tcg_abort();
660 static void gen_check_io(DisasContext *s, TCGMemOp ot, target_ulong cur_eip,
661 uint32_t svm_flags)
663 int state_saved;
664 target_ulong next_eip;
666 state_saved = 0;
667 if (s->pe && (s->cpl > s->iopl || s->vm86)) {
668 gen_update_cc_op(s);
669 gen_jmp_im(cur_eip);
670 state_saved = 1;
671 tcg_gen_trunc_tl_i32(cpu_tmp2_i32, cpu_T[0]);
672 switch (ot) {
673 case MO_8:
674 gen_helper_check_iob(cpu_env, cpu_tmp2_i32);
675 break;
676 case MO_16:
677 gen_helper_check_iow(cpu_env, cpu_tmp2_i32);
678 break;
679 case MO_32:
680 gen_helper_check_iol(cpu_env, cpu_tmp2_i32);
681 break;
682 default:
683 tcg_abort();
686 if(s->flags & HF_SVMI_MASK) {
687 if (!state_saved) {
688 gen_update_cc_op(s);
689 gen_jmp_im(cur_eip);
691 svm_flags |= (1 << (4 + ot));
692 next_eip = s->pc - s->cs_base;
693 tcg_gen_trunc_tl_i32(cpu_tmp2_i32, cpu_T[0]);
694 gen_helper_svm_check_io(cpu_env, cpu_tmp2_i32,
695 tcg_const_i32(svm_flags),
696 tcg_const_i32(next_eip - cur_eip));
700 static inline void gen_movs(DisasContext *s, TCGMemOp ot)
702 gen_string_movl_A0_ESI(s);
703 gen_op_ld_v(s, ot, cpu_T[0], cpu_A0);
704 gen_string_movl_A0_EDI(s);
705 gen_op_st_v(s, ot, cpu_T[0], cpu_A0);
706 gen_op_movl_T0_Dshift(ot);
707 gen_op_add_reg_T0(s->aflag, R_ESI);
708 gen_op_add_reg_T0(s->aflag, R_EDI);
711 static void gen_op_update1_cc(void)
713 tcg_gen_mov_tl(cpu_cc_dst, cpu_T[0]);
716 static void gen_op_update2_cc(void)
718 tcg_gen_mov_tl(cpu_cc_src, cpu_T[1]);
719 tcg_gen_mov_tl(cpu_cc_dst, cpu_T[0]);
722 static void gen_op_update3_cc(TCGv reg)
724 tcg_gen_mov_tl(cpu_cc_src2, reg);
725 tcg_gen_mov_tl(cpu_cc_src, cpu_T[1]);
726 tcg_gen_mov_tl(cpu_cc_dst, cpu_T[0]);
729 static inline void gen_op_testl_T0_T1_cc(void)
731 tcg_gen_and_tl(cpu_cc_dst, cpu_T[0], cpu_T[1]);
734 static void gen_op_update_neg_cc(void)
736 tcg_gen_mov_tl(cpu_cc_dst, cpu_T[0]);
737 tcg_gen_neg_tl(cpu_cc_src, cpu_T[0]);
738 tcg_gen_movi_tl(cpu_cc_srcT, 0);
741 /* compute all eflags to cc_src */
742 static void gen_compute_eflags(DisasContext *s)
744 TCGv zero, dst, src1, src2;
745 int live, dead;
747 if (s->cc_op == CC_OP_EFLAGS) {
748 return;
750 if (s->cc_op == CC_OP_CLR) {
751 tcg_gen_movi_tl(cpu_cc_src, CC_Z);
752 set_cc_op(s, CC_OP_EFLAGS);
753 return;
756 TCGV_UNUSED(zero);
757 dst = cpu_cc_dst;
758 src1 = cpu_cc_src;
759 src2 = cpu_cc_src2;
761 /* Take care to not read values that are not live. */
762 live = cc_op_live[s->cc_op] & ~USES_CC_SRCT;
763 dead = live ^ (USES_CC_DST | USES_CC_SRC | USES_CC_SRC2);
764 if (dead) {
765 zero = tcg_const_tl(0);
766 if (dead & USES_CC_DST) {
767 dst = zero;
769 if (dead & USES_CC_SRC) {
770 src1 = zero;
772 if (dead & USES_CC_SRC2) {
773 src2 = zero;
777 gen_update_cc_op(s);
778 gen_helper_cc_compute_all(cpu_cc_src, dst, src1, src2, cpu_cc_op);
779 set_cc_op(s, CC_OP_EFLAGS);
781 if (dead) {
782 tcg_temp_free(zero);
786 typedef struct CCPrepare {
787 TCGCond cond;
788 TCGv reg;
789 TCGv reg2;
790 target_ulong imm;
791 target_ulong mask;
792 bool use_reg2;
793 bool no_setcond;
794 } CCPrepare;
796 /* compute eflags.C to reg */
797 static CCPrepare gen_prepare_eflags_c(DisasContext *s, TCGv reg)
799 TCGv t0, t1;
800 int size, shift;
802 switch (s->cc_op) {
803 case CC_OP_SUBB ... CC_OP_SUBQ:
804 /* (DATA_TYPE)CC_SRCT < (DATA_TYPE)CC_SRC */
805 size = s->cc_op - CC_OP_SUBB;
806 t1 = gen_ext_tl(cpu_tmp0, cpu_cc_src, size, false);
807 /* If no temporary was used, be careful not to alias t1 and t0. */
808 t0 = TCGV_EQUAL(t1, cpu_cc_src) ? cpu_tmp0 : reg;
809 tcg_gen_mov_tl(t0, cpu_cc_srcT);
810 gen_extu(size, t0);
811 goto add_sub;
813 case CC_OP_ADDB ... CC_OP_ADDQ:
814 /* (DATA_TYPE)CC_DST < (DATA_TYPE)CC_SRC */
815 size = s->cc_op - CC_OP_ADDB;
816 t1 = gen_ext_tl(cpu_tmp0, cpu_cc_src, size, false);
817 t0 = gen_ext_tl(reg, cpu_cc_dst, size, false);
818 add_sub:
819 return (CCPrepare) { .cond = TCG_COND_LTU, .reg = t0,
820 .reg2 = t1, .mask = -1, .use_reg2 = true };
822 case CC_OP_LOGICB ... CC_OP_LOGICQ:
823 case CC_OP_CLR:
824 return (CCPrepare) { .cond = TCG_COND_NEVER, .mask = -1 };
826 case CC_OP_INCB ... CC_OP_INCQ:
827 case CC_OP_DECB ... CC_OP_DECQ:
828 return (CCPrepare) { .cond = TCG_COND_NE, .reg = cpu_cc_src,
829 .mask = -1, .no_setcond = true };
831 case CC_OP_SHLB ... CC_OP_SHLQ:
832 /* (CC_SRC >> (DATA_BITS - 1)) & 1 */
833 size = s->cc_op - CC_OP_SHLB;
834 shift = (8 << size) - 1;
835 return (CCPrepare) { .cond = TCG_COND_NE, .reg = cpu_cc_src,
836 .mask = (target_ulong)1 << shift };
838 case CC_OP_MULB ... CC_OP_MULQ:
839 return (CCPrepare) { .cond = TCG_COND_NE,
840 .reg = cpu_cc_src, .mask = -1 };
842 case CC_OP_BMILGB ... CC_OP_BMILGQ:
843 size = s->cc_op - CC_OP_BMILGB;
844 t0 = gen_ext_tl(reg, cpu_cc_src, size, false);
845 return (CCPrepare) { .cond = TCG_COND_EQ, .reg = t0, .mask = -1 };
847 case CC_OP_ADCX:
848 case CC_OP_ADCOX:
849 return (CCPrepare) { .cond = TCG_COND_NE, .reg = cpu_cc_dst,
850 .mask = -1, .no_setcond = true };
852 case CC_OP_EFLAGS:
853 case CC_OP_SARB ... CC_OP_SARQ:
854 /* CC_SRC & 1 */
855 return (CCPrepare) { .cond = TCG_COND_NE,
856 .reg = cpu_cc_src, .mask = CC_C };
858 default:
859 /* The need to compute only C from CC_OP_DYNAMIC is important
860 in efficiently implementing e.g. INC at the start of a TB. */
861 gen_update_cc_op(s);
862 gen_helper_cc_compute_c(reg, cpu_cc_dst, cpu_cc_src,
863 cpu_cc_src2, cpu_cc_op);
864 return (CCPrepare) { .cond = TCG_COND_NE, .reg = reg,
865 .mask = -1, .no_setcond = true };
869 /* compute eflags.P to reg */
870 static CCPrepare gen_prepare_eflags_p(DisasContext *s, TCGv reg)
872 gen_compute_eflags(s);
873 return (CCPrepare) { .cond = TCG_COND_NE, .reg = cpu_cc_src,
874 .mask = CC_P };
877 /* compute eflags.S to reg */
878 static CCPrepare gen_prepare_eflags_s(DisasContext *s, TCGv reg)
880 switch (s->cc_op) {
881 case CC_OP_DYNAMIC:
882 gen_compute_eflags(s);
883 /* FALLTHRU */
884 case CC_OP_EFLAGS:
885 case CC_OP_ADCX:
886 case CC_OP_ADOX:
887 case CC_OP_ADCOX:
888 return (CCPrepare) { .cond = TCG_COND_NE, .reg = cpu_cc_src,
889 .mask = CC_S };
890 case CC_OP_CLR:
891 return (CCPrepare) { .cond = TCG_COND_NEVER, .mask = -1 };
892 default:
894 TCGMemOp size = (s->cc_op - CC_OP_ADDB) & 3;
895 TCGv t0 = gen_ext_tl(reg, cpu_cc_dst, size, true);
896 return (CCPrepare) { .cond = TCG_COND_LT, .reg = t0, .mask = -1 };
901 /* compute eflags.O to reg */
902 static CCPrepare gen_prepare_eflags_o(DisasContext *s, TCGv reg)
904 switch (s->cc_op) {
905 case CC_OP_ADOX:
906 case CC_OP_ADCOX:
907 return (CCPrepare) { .cond = TCG_COND_NE, .reg = cpu_cc_src2,
908 .mask = -1, .no_setcond = true };
909 case CC_OP_CLR:
910 return (CCPrepare) { .cond = TCG_COND_NEVER, .mask = -1 };
911 default:
912 gen_compute_eflags(s);
913 return (CCPrepare) { .cond = TCG_COND_NE, .reg = cpu_cc_src,
914 .mask = CC_O };
918 /* compute eflags.Z to reg */
919 static CCPrepare gen_prepare_eflags_z(DisasContext *s, TCGv reg)
921 switch (s->cc_op) {
922 case CC_OP_DYNAMIC:
923 gen_compute_eflags(s);
924 /* FALLTHRU */
925 case CC_OP_EFLAGS:
926 case CC_OP_ADCX:
927 case CC_OP_ADOX:
928 case CC_OP_ADCOX:
929 return (CCPrepare) { .cond = TCG_COND_NE, .reg = cpu_cc_src,
930 .mask = CC_Z };
931 case CC_OP_CLR:
932 return (CCPrepare) { .cond = TCG_COND_ALWAYS, .mask = -1 };
933 default:
935 TCGMemOp size = (s->cc_op - CC_OP_ADDB) & 3;
936 TCGv t0 = gen_ext_tl(reg, cpu_cc_dst, size, false);
937 return (CCPrepare) { .cond = TCG_COND_EQ, .reg = t0, .mask = -1 };
942 /* perform a conditional store into register 'reg' according to jump opcode
943 value 'b'. In the fast case, T0 is guaranted not to be used. */
944 static CCPrepare gen_prepare_cc(DisasContext *s, int b, TCGv reg)
946 int inv, jcc_op, cond;
947 TCGMemOp size;
948 CCPrepare cc;
949 TCGv t0;
951 inv = b & 1;
952 jcc_op = (b >> 1) & 7;
954 switch (s->cc_op) {
955 case CC_OP_SUBB ... CC_OP_SUBQ:
956 /* We optimize relational operators for the cmp/jcc case. */
957 size = s->cc_op - CC_OP_SUBB;
958 switch (jcc_op) {
959 case JCC_BE:
960 tcg_gen_mov_tl(cpu_tmp4, cpu_cc_srcT);
961 gen_extu(size, cpu_tmp4);
962 t0 = gen_ext_tl(cpu_tmp0, cpu_cc_src, size, false);
963 cc = (CCPrepare) { .cond = TCG_COND_LEU, .reg = cpu_tmp4,
964 .reg2 = t0, .mask = -1, .use_reg2 = true };
965 break;
967 case JCC_L:
968 cond = TCG_COND_LT;
969 goto fast_jcc_l;
970 case JCC_LE:
971 cond = TCG_COND_LE;
972 fast_jcc_l:
973 tcg_gen_mov_tl(cpu_tmp4, cpu_cc_srcT);
974 gen_exts(size, cpu_tmp4);
975 t0 = gen_ext_tl(cpu_tmp0, cpu_cc_src, size, true);
976 cc = (CCPrepare) { .cond = cond, .reg = cpu_tmp4,
977 .reg2 = t0, .mask = -1, .use_reg2 = true };
978 break;
980 default:
981 goto slow_jcc;
983 break;
985 default:
986 slow_jcc:
987 /* This actually generates good code for JC, JZ and JS. */
988 switch (jcc_op) {
989 case JCC_O:
990 cc = gen_prepare_eflags_o(s, reg);
991 break;
992 case JCC_B:
993 cc = gen_prepare_eflags_c(s, reg);
994 break;
995 case JCC_Z:
996 cc = gen_prepare_eflags_z(s, reg);
997 break;
998 case JCC_BE:
999 gen_compute_eflags(s);
1000 cc = (CCPrepare) { .cond = TCG_COND_NE, .reg = cpu_cc_src,
1001 .mask = CC_Z | CC_C };
1002 break;
1003 case JCC_S:
1004 cc = gen_prepare_eflags_s(s, reg);
1005 break;
1006 case JCC_P:
1007 cc = gen_prepare_eflags_p(s, reg);
1008 break;
1009 case JCC_L:
1010 gen_compute_eflags(s);
1011 if (TCGV_EQUAL(reg, cpu_cc_src)) {
1012 reg = cpu_tmp0;
1014 tcg_gen_shri_tl(reg, cpu_cc_src, 4); /* CC_O -> CC_S */
1015 tcg_gen_xor_tl(reg, reg, cpu_cc_src);
1016 cc = (CCPrepare) { .cond = TCG_COND_NE, .reg = reg,
1017 .mask = CC_S };
1018 break;
1019 default:
1020 case JCC_LE:
1021 gen_compute_eflags(s);
1022 if (TCGV_EQUAL(reg, cpu_cc_src)) {
1023 reg = cpu_tmp0;
1025 tcg_gen_shri_tl(reg, cpu_cc_src, 4); /* CC_O -> CC_S */
1026 tcg_gen_xor_tl(reg, reg, cpu_cc_src);
1027 cc = (CCPrepare) { .cond = TCG_COND_NE, .reg = reg,
1028 .mask = CC_S | CC_Z };
1029 break;
1031 break;
1034 if (inv) {
1035 cc.cond = tcg_invert_cond(cc.cond);
1037 return cc;
1040 static void gen_setcc1(DisasContext *s, int b, TCGv reg)
1042 CCPrepare cc = gen_prepare_cc(s, b, reg);
1044 if (cc.no_setcond) {
1045 if (cc.cond == TCG_COND_EQ) {
1046 tcg_gen_xori_tl(reg, cc.reg, 1);
1047 } else {
1048 tcg_gen_mov_tl(reg, cc.reg);
1050 return;
1053 if (cc.cond == TCG_COND_NE && !cc.use_reg2 && cc.imm == 0 &&
1054 cc.mask != 0 && (cc.mask & (cc.mask - 1)) == 0) {
1055 tcg_gen_shri_tl(reg, cc.reg, ctztl(cc.mask));
1056 tcg_gen_andi_tl(reg, reg, 1);
1057 return;
1059 if (cc.mask != -1) {
1060 tcg_gen_andi_tl(reg, cc.reg, cc.mask);
1061 cc.reg = reg;
1063 if (cc.use_reg2) {
1064 tcg_gen_setcond_tl(cc.cond, reg, cc.reg, cc.reg2);
1065 } else {
1066 tcg_gen_setcondi_tl(cc.cond, reg, cc.reg, cc.imm);
1070 static inline void gen_compute_eflags_c(DisasContext *s, TCGv reg)
1072 gen_setcc1(s, JCC_B << 1, reg);
1075 /* generate a conditional jump to label 'l1' according to jump opcode
1076 value 'b'. In the fast case, T0 is guaranted not to be used. */
1077 static inline void gen_jcc1_noeob(DisasContext *s, int b, int l1)
1079 CCPrepare cc = gen_prepare_cc(s, b, cpu_T[0]);
1081 if (cc.mask != -1) {
1082 tcg_gen_andi_tl(cpu_T[0], cc.reg, cc.mask);
1083 cc.reg = cpu_T[0];
1085 if (cc.use_reg2) {
1086 tcg_gen_brcond_tl(cc.cond, cc.reg, cc.reg2, l1);
1087 } else {
1088 tcg_gen_brcondi_tl(cc.cond, cc.reg, cc.imm, l1);
1092 /* Generate a conditional jump to label 'l1' according to jump opcode
1093 value 'b'. In the fast case, T0 is guaranted not to be used.
1094 A translation block must end soon. */
1095 static inline void gen_jcc1(DisasContext *s, int b, int l1)
1097 CCPrepare cc = gen_prepare_cc(s, b, cpu_T[0]);
1099 gen_update_cc_op(s);
1100 if (cc.mask != -1) {
1101 tcg_gen_andi_tl(cpu_T[0], cc.reg, cc.mask);
1102 cc.reg = cpu_T[0];
1104 set_cc_op(s, CC_OP_DYNAMIC);
1105 if (cc.use_reg2) {
1106 tcg_gen_brcond_tl(cc.cond, cc.reg, cc.reg2, l1);
1107 } else {
1108 tcg_gen_brcondi_tl(cc.cond, cc.reg, cc.imm, l1);
1112 /* XXX: does not work with gdbstub "ice" single step - not a
1113 serious problem */
1114 static int gen_jz_ecx_string(DisasContext *s, target_ulong next_eip)
1116 int l1, l2;
1118 l1 = gen_new_label();
1119 l2 = gen_new_label();
1120 gen_op_jnz_ecx(s->aflag, l1);
1121 gen_set_label(l2);
1122 gen_jmp_tb(s, next_eip, 1);
1123 gen_set_label(l1);
1124 return l2;
1127 static inline void gen_stos(DisasContext *s, TCGMemOp ot)
1129 gen_op_mov_v_reg(MO_32, cpu_T[0], R_EAX);
1130 gen_string_movl_A0_EDI(s);
1131 gen_op_st_v(s, ot, cpu_T[0], cpu_A0);
1132 gen_op_movl_T0_Dshift(ot);
1133 gen_op_add_reg_T0(s->aflag, R_EDI);
1136 static inline void gen_lods(DisasContext *s, TCGMemOp ot)
1138 gen_string_movl_A0_ESI(s);
1139 gen_op_ld_v(s, ot, cpu_T[0], cpu_A0);
1140 gen_op_mov_reg_v(ot, R_EAX, cpu_T[0]);
1141 gen_op_movl_T0_Dshift(ot);
1142 gen_op_add_reg_T0(s->aflag, R_ESI);
1145 static inline void gen_scas(DisasContext *s, TCGMemOp ot)
1147 gen_string_movl_A0_EDI(s);
1148 gen_op_ld_v(s, ot, cpu_T[1], cpu_A0);
1149 gen_op(s, OP_CMPL, ot, R_EAX);
1150 gen_op_movl_T0_Dshift(ot);
1151 gen_op_add_reg_T0(s->aflag, R_EDI);
1154 static inline void gen_cmps(DisasContext *s, TCGMemOp ot)
1156 gen_string_movl_A0_EDI(s);
1157 gen_op_ld_v(s, ot, cpu_T[1], cpu_A0);
1158 gen_string_movl_A0_ESI(s);
1159 gen_op(s, OP_CMPL, ot, OR_TMP0);
1160 gen_op_movl_T0_Dshift(ot);
1161 gen_op_add_reg_T0(s->aflag, R_ESI);
1162 gen_op_add_reg_T0(s->aflag, R_EDI);
1165 static inline void gen_ins(DisasContext *s, TCGMemOp ot)
1167 if (use_icount)
1168 gen_io_start();
1169 gen_string_movl_A0_EDI(s);
1170 /* Note: we must do this dummy write first to be restartable in
1171 case of page fault. */
1172 tcg_gen_movi_tl(cpu_T[0], 0);
1173 gen_op_st_v(s, ot, cpu_T[0], cpu_A0);
1174 tcg_gen_trunc_tl_i32(cpu_tmp2_i32, cpu_regs[R_EDX]);
1175 tcg_gen_andi_i32(cpu_tmp2_i32, cpu_tmp2_i32, 0xffff);
1176 gen_helper_in_func(ot, cpu_T[0], cpu_tmp2_i32);
1177 gen_op_st_v(s, ot, cpu_T[0], cpu_A0);
1178 gen_op_movl_T0_Dshift(ot);
1179 gen_op_add_reg_T0(s->aflag, R_EDI);
1180 if (use_icount)
1181 gen_io_end();
1184 static inline void gen_outs(DisasContext *s, TCGMemOp ot)
1186 if (use_icount)
1187 gen_io_start();
1188 gen_string_movl_A0_ESI(s);
1189 gen_op_ld_v(s, ot, cpu_T[0], cpu_A0);
1191 tcg_gen_trunc_tl_i32(cpu_tmp2_i32, cpu_regs[R_EDX]);
1192 tcg_gen_andi_i32(cpu_tmp2_i32, cpu_tmp2_i32, 0xffff);
1193 tcg_gen_trunc_tl_i32(cpu_tmp3_i32, cpu_T[0]);
1194 gen_helper_out_func(ot, cpu_tmp2_i32, cpu_tmp3_i32);
1196 gen_op_movl_T0_Dshift(ot);
1197 gen_op_add_reg_T0(s->aflag, R_ESI);
1198 if (use_icount)
1199 gen_io_end();
1202 /* same method as Valgrind : we generate jumps to current or next
1203 instruction */
1204 #define GEN_REPZ(op) \
1205 static inline void gen_repz_ ## op(DisasContext *s, TCGMemOp ot, \
1206 target_ulong cur_eip, target_ulong next_eip) \
1208 int l2;\
1209 gen_update_cc_op(s); \
1210 l2 = gen_jz_ecx_string(s, next_eip); \
1211 gen_ ## op(s, ot); \
1212 gen_op_add_reg_im(s->aflag, R_ECX, -1); \
1213 /* a loop would cause two single step exceptions if ECX = 1 \
1214 before rep string_insn */ \
1215 if (!s->jmp_opt) \
1216 gen_op_jz_ecx(s->aflag, l2); \
1217 gen_jmp(s, cur_eip); \
1220 #define GEN_REPZ2(op) \
1221 static inline void gen_repz_ ## op(DisasContext *s, TCGMemOp ot, \
1222 target_ulong cur_eip, \
1223 target_ulong next_eip, \
1224 int nz) \
1226 int l2;\
1227 gen_update_cc_op(s); \
1228 l2 = gen_jz_ecx_string(s, next_eip); \
1229 gen_ ## op(s, ot); \
1230 gen_op_add_reg_im(s->aflag, R_ECX, -1); \
1231 gen_update_cc_op(s); \
1232 gen_jcc1(s, (JCC_Z << 1) | (nz ^ 1), l2); \
1233 if (!s->jmp_opt) \
1234 gen_op_jz_ecx(s->aflag, l2); \
1235 gen_jmp(s, cur_eip); \
1238 GEN_REPZ(movs)
1239 GEN_REPZ(stos)
1240 GEN_REPZ(lods)
1241 GEN_REPZ(ins)
1242 GEN_REPZ(outs)
1243 GEN_REPZ2(scas)
1244 GEN_REPZ2(cmps)
1246 static void gen_helper_fp_arith_ST0_FT0(int op)
1248 switch (op) {
1249 case 0:
1250 gen_helper_fadd_ST0_FT0(cpu_env);
1251 break;
1252 case 1:
1253 gen_helper_fmul_ST0_FT0(cpu_env);
1254 break;
1255 case 2:
1256 gen_helper_fcom_ST0_FT0(cpu_env);
1257 break;
1258 case 3:
1259 gen_helper_fcom_ST0_FT0(cpu_env);
1260 break;
1261 case 4:
1262 gen_helper_fsub_ST0_FT0(cpu_env);
1263 break;
1264 case 5:
1265 gen_helper_fsubr_ST0_FT0(cpu_env);
1266 break;
1267 case 6:
1268 gen_helper_fdiv_ST0_FT0(cpu_env);
1269 break;
1270 case 7:
1271 gen_helper_fdivr_ST0_FT0(cpu_env);
1272 break;
1276 /* NOTE the exception in "r" op ordering */
1277 static void gen_helper_fp_arith_STN_ST0(int op, int opreg)
1279 TCGv_i32 tmp = tcg_const_i32(opreg);
1280 switch (op) {
1281 case 0:
1282 gen_helper_fadd_STN_ST0(cpu_env, tmp);
1283 break;
1284 case 1:
1285 gen_helper_fmul_STN_ST0(cpu_env, tmp);
1286 break;
1287 case 4:
1288 gen_helper_fsubr_STN_ST0(cpu_env, tmp);
1289 break;
1290 case 5:
1291 gen_helper_fsub_STN_ST0(cpu_env, tmp);
1292 break;
1293 case 6:
1294 gen_helper_fdivr_STN_ST0(cpu_env, tmp);
1295 break;
1296 case 7:
1297 gen_helper_fdiv_STN_ST0(cpu_env, tmp);
1298 break;
1302 /* if d == OR_TMP0, it means memory operand (address in A0) */
1303 static void gen_op(DisasContext *s1, int op, TCGMemOp ot, int d)
1305 if (d != OR_TMP0) {
1306 gen_op_mov_v_reg(ot, cpu_T[0], d);
1307 } else {
1308 gen_op_ld_v(s1, ot, cpu_T[0], cpu_A0);
1310 switch(op) {
1311 case OP_ADCL:
1312 gen_compute_eflags_c(s1, cpu_tmp4);
1313 tcg_gen_add_tl(cpu_T[0], cpu_T[0], cpu_T[1]);
1314 tcg_gen_add_tl(cpu_T[0], cpu_T[0], cpu_tmp4);
1315 gen_op_st_rm_T0_A0(s1, ot, d);
1316 gen_op_update3_cc(cpu_tmp4);
1317 set_cc_op(s1, CC_OP_ADCB + ot);
1318 break;
1319 case OP_SBBL:
1320 gen_compute_eflags_c(s1, cpu_tmp4);
1321 tcg_gen_sub_tl(cpu_T[0], cpu_T[0], cpu_T[1]);
1322 tcg_gen_sub_tl(cpu_T[0], cpu_T[0], cpu_tmp4);
1323 gen_op_st_rm_T0_A0(s1, ot, d);
1324 gen_op_update3_cc(cpu_tmp4);
1325 set_cc_op(s1, CC_OP_SBBB + ot);
1326 break;
1327 case OP_ADDL:
1328 tcg_gen_add_tl(cpu_T[0], cpu_T[0], cpu_T[1]);
1329 gen_op_st_rm_T0_A0(s1, ot, d);
1330 gen_op_update2_cc();
1331 set_cc_op(s1, CC_OP_ADDB + ot);
1332 break;
1333 case OP_SUBL:
1334 tcg_gen_mov_tl(cpu_cc_srcT, cpu_T[0]);
1335 tcg_gen_sub_tl(cpu_T[0], cpu_T[0], cpu_T[1]);
1336 gen_op_st_rm_T0_A0(s1, ot, d);
1337 gen_op_update2_cc();
1338 set_cc_op(s1, CC_OP_SUBB + ot);
1339 break;
1340 default:
1341 case OP_ANDL:
1342 tcg_gen_and_tl(cpu_T[0], cpu_T[0], cpu_T[1]);
1343 gen_op_st_rm_T0_A0(s1, ot, d);
1344 gen_op_update1_cc();
1345 set_cc_op(s1, CC_OP_LOGICB + ot);
1346 break;
1347 case OP_ORL:
1348 tcg_gen_or_tl(cpu_T[0], cpu_T[0], cpu_T[1]);
1349 gen_op_st_rm_T0_A0(s1, ot, d);
1350 gen_op_update1_cc();
1351 set_cc_op(s1, CC_OP_LOGICB + ot);
1352 break;
1353 case OP_XORL:
1354 tcg_gen_xor_tl(cpu_T[0], cpu_T[0], cpu_T[1]);
1355 gen_op_st_rm_T0_A0(s1, ot, d);
1356 gen_op_update1_cc();
1357 set_cc_op(s1, CC_OP_LOGICB + ot);
1358 break;
1359 case OP_CMPL:
1360 tcg_gen_mov_tl(cpu_cc_src, cpu_T[1]);
1361 tcg_gen_mov_tl(cpu_cc_srcT, cpu_T[0]);
1362 tcg_gen_sub_tl(cpu_cc_dst, cpu_T[0], cpu_T[1]);
1363 set_cc_op(s1, CC_OP_SUBB + ot);
1364 break;
1368 /* if d == OR_TMP0, it means memory operand (address in A0) */
1369 static void gen_inc(DisasContext *s1, TCGMemOp ot, int d, int c)
1371 if (d != OR_TMP0) {
1372 gen_op_mov_v_reg(ot, cpu_T[0], d);
1373 } else {
1374 gen_op_ld_v(s1, ot, cpu_T[0], cpu_A0);
1376 gen_compute_eflags_c(s1, cpu_cc_src);
1377 if (c > 0) {
1378 tcg_gen_addi_tl(cpu_T[0], cpu_T[0], 1);
1379 set_cc_op(s1, CC_OP_INCB + ot);
1380 } else {
1381 tcg_gen_addi_tl(cpu_T[0], cpu_T[0], -1);
1382 set_cc_op(s1, CC_OP_DECB + ot);
1384 gen_op_st_rm_T0_A0(s1, ot, d);
1385 tcg_gen_mov_tl(cpu_cc_dst, cpu_T[0]);
1388 static void gen_shift_flags(DisasContext *s, TCGMemOp ot, TCGv result,
1389 TCGv shm1, TCGv count, bool is_right)
1391 TCGv_i32 z32, s32, oldop;
1392 TCGv z_tl;
1394 /* Store the results into the CC variables. If we know that the
1395 variable must be dead, store unconditionally. Otherwise we'll
1396 need to not disrupt the current contents. */
1397 z_tl = tcg_const_tl(0);
1398 if (cc_op_live[s->cc_op] & USES_CC_DST) {
1399 tcg_gen_movcond_tl(TCG_COND_NE, cpu_cc_dst, count, z_tl,
1400 result, cpu_cc_dst);
1401 } else {
1402 tcg_gen_mov_tl(cpu_cc_dst, result);
1404 if (cc_op_live[s->cc_op] & USES_CC_SRC) {
1405 tcg_gen_movcond_tl(TCG_COND_NE, cpu_cc_src, count, z_tl,
1406 shm1, cpu_cc_src);
1407 } else {
1408 tcg_gen_mov_tl(cpu_cc_src, shm1);
1410 tcg_temp_free(z_tl);
1412 /* Get the two potential CC_OP values into temporaries. */
1413 tcg_gen_movi_i32(cpu_tmp2_i32, (is_right ? CC_OP_SARB : CC_OP_SHLB) + ot);
1414 if (s->cc_op == CC_OP_DYNAMIC) {
1415 oldop = cpu_cc_op;
1416 } else {
1417 tcg_gen_movi_i32(cpu_tmp3_i32, s->cc_op);
1418 oldop = cpu_tmp3_i32;
1421 /* Conditionally store the CC_OP value. */
1422 z32 = tcg_const_i32(0);
1423 s32 = tcg_temp_new_i32();
1424 tcg_gen_trunc_tl_i32(s32, count);
1425 tcg_gen_movcond_i32(TCG_COND_NE, cpu_cc_op, s32, z32, cpu_tmp2_i32, oldop);
1426 tcg_temp_free_i32(z32);
1427 tcg_temp_free_i32(s32);
1429 /* The CC_OP value is no longer predictable. */
1430 set_cc_op(s, CC_OP_DYNAMIC);
1433 static void gen_shift_rm_T1(DisasContext *s, TCGMemOp ot, int op1,
1434 int is_right, int is_arith)
1436 target_ulong mask = (ot == MO_64 ? 0x3f : 0x1f);
1438 /* load */
1439 if (op1 == OR_TMP0) {
1440 gen_op_ld_v(s, ot, cpu_T[0], cpu_A0);
1441 } else {
1442 gen_op_mov_v_reg(ot, cpu_T[0], op1);
1445 tcg_gen_andi_tl(cpu_T[1], cpu_T[1], mask);
1446 tcg_gen_subi_tl(cpu_tmp0, cpu_T[1], 1);
1448 if (is_right) {
1449 if (is_arith) {
1450 gen_exts(ot, cpu_T[0]);
1451 tcg_gen_sar_tl(cpu_tmp0, cpu_T[0], cpu_tmp0);
1452 tcg_gen_sar_tl(cpu_T[0], cpu_T[0], cpu_T[1]);
1453 } else {
1454 gen_extu(ot, cpu_T[0]);
1455 tcg_gen_shr_tl(cpu_tmp0, cpu_T[0], cpu_tmp0);
1456 tcg_gen_shr_tl(cpu_T[0], cpu_T[0], cpu_T[1]);
1458 } else {
1459 tcg_gen_shl_tl(cpu_tmp0, cpu_T[0], cpu_tmp0);
1460 tcg_gen_shl_tl(cpu_T[0], cpu_T[0], cpu_T[1]);
1463 /* store */
1464 gen_op_st_rm_T0_A0(s, ot, op1);
1466 gen_shift_flags(s, ot, cpu_T[0], cpu_tmp0, cpu_T[1], is_right);
1469 static void gen_shift_rm_im(DisasContext *s, TCGMemOp ot, int op1, int op2,
1470 int is_right, int is_arith)
1472 int mask = (ot == MO_64 ? 0x3f : 0x1f);
1474 /* load */
1475 if (op1 == OR_TMP0)
1476 gen_op_ld_v(s, ot, cpu_T[0], cpu_A0);
1477 else
1478 gen_op_mov_v_reg(ot, cpu_T[0], op1);
1480 op2 &= mask;
1481 if (op2 != 0) {
1482 if (is_right) {
1483 if (is_arith) {
1484 gen_exts(ot, cpu_T[0]);
1485 tcg_gen_sari_tl(cpu_tmp4, cpu_T[0], op2 - 1);
1486 tcg_gen_sari_tl(cpu_T[0], cpu_T[0], op2);
1487 } else {
1488 gen_extu(ot, cpu_T[0]);
1489 tcg_gen_shri_tl(cpu_tmp4, cpu_T[0], op2 - 1);
1490 tcg_gen_shri_tl(cpu_T[0], cpu_T[0], op2);
1492 } else {
1493 tcg_gen_shli_tl(cpu_tmp4, cpu_T[0], op2 - 1);
1494 tcg_gen_shli_tl(cpu_T[0], cpu_T[0], op2);
1498 /* store */
1499 gen_op_st_rm_T0_A0(s, ot, op1);
1501 /* update eflags if non zero shift */
1502 if (op2 != 0) {
1503 tcg_gen_mov_tl(cpu_cc_src, cpu_tmp4);
1504 tcg_gen_mov_tl(cpu_cc_dst, cpu_T[0]);
1505 set_cc_op(s, (is_right ? CC_OP_SARB : CC_OP_SHLB) + ot);
1509 static inline void tcg_gen_lshift(TCGv ret, TCGv arg1, target_long arg2)
1511 if (arg2 >= 0)
1512 tcg_gen_shli_tl(ret, arg1, arg2);
1513 else
1514 tcg_gen_shri_tl(ret, arg1, -arg2);
1517 static void gen_rot_rm_T1(DisasContext *s, TCGMemOp ot, int op1, int is_right)
1519 target_ulong mask = (ot == MO_64 ? 0x3f : 0x1f);
1520 TCGv_i32 t0, t1;
1522 /* load */
1523 if (op1 == OR_TMP0) {
1524 gen_op_ld_v(s, ot, cpu_T[0], cpu_A0);
1525 } else {
1526 gen_op_mov_v_reg(ot, cpu_T[0], op1);
1529 tcg_gen_andi_tl(cpu_T[1], cpu_T[1], mask);
1531 switch (ot) {
1532 case MO_8:
1533 /* Replicate the 8-bit input so that a 32-bit rotate works. */
1534 tcg_gen_ext8u_tl(cpu_T[0], cpu_T[0]);
1535 tcg_gen_muli_tl(cpu_T[0], cpu_T[0], 0x01010101);
1536 goto do_long;
1537 case MO_16:
1538 /* Replicate the 16-bit input so that a 32-bit rotate works. */
1539 tcg_gen_deposit_tl(cpu_T[0], cpu_T[0], cpu_T[0], 16, 16);
1540 goto do_long;
1541 do_long:
1542 #ifdef TARGET_X86_64
1543 case MO_32:
1544 tcg_gen_trunc_tl_i32(cpu_tmp2_i32, cpu_T[0]);
1545 tcg_gen_trunc_tl_i32(cpu_tmp3_i32, cpu_T[1]);
1546 if (is_right) {
1547 tcg_gen_rotr_i32(cpu_tmp2_i32, cpu_tmp2_i32, cpu_tmp3_i32);
1548 } else {
1549 tcg_gen_rotl_i32(cpu_tmp2_i32, cpu_tmp2_i32, cpu_tmp3_i32);
1551 tcg_gen_extu_i32_tl(cpu_T[0], cpu_tmp2_i32);
1552 break;
1553 #endif
1554 default:
1555 if (is_right) {
1556 tcg_gen_rotr_tl(cpu_T[0], cpu_T[0], cpu_T[1]);
1557 } else {
1558 tcg_gen_rotl_tl(cpu_T[0], cpu_T[0], cpu_T[1]);
1560 break;
1563 /* store */
1564 gen_op_st_rm_T0_A0(s, ot, op1);
1566 /* We'll need the flags computed into CC_SRC. */
1567 gen_compute_eflags(s);
1569 /* The value that was "rotated out" is now present at the other end
1570 of the word. Compute C into CC_DST and O into CC_SRC2. Note that
1571 since we've computed the flags into CC_SRC, these variables are
1572 currently dead. */
1573 if (is_right) {
1574 tcg_gen_shri_tl(cpu_cc_src2, cpu_T[0], mask - 1);
1575 tcg_gen_shri_tl(cpu_cc_dst, cpu_T[0], mask);
1576 tcg_gen_andi_tl(cpu_cc_dst, cpu_cc_dst, 1);
1577 } else {
1578 tcg_gen_shri_tl(cpu_cc_src2, cpu_T[0], mask);
1579 tcg_gen_andi_tl(cpu_cc_dst, cpu_T[0], 1);
1581 tcg_gen_andi_tl(cpu_cc_src2, cpu_cc_src2, 1);
1582 tcg_gen_xor_tl(cpu_cc_src2, cpu_cc_src2, cpu_cc_dst);
1584 /* Now conditionally store the new CC_OP value. If the shift count
1585 is 0 we keep the CC_OP_EFLAGS setting so that only CC_SRC is live.
1586 Otherwise reuse CC_OP_ADCOX which have the C and O flags split out
1587 exactly as we computed above. */
1588 t0 = tcg_const_i32(0);
1589 t1 = tcg_temp_new_i32();
1590 tcg_gen_trunc_tl_i32(t1, cpu_T[1]);
1591 tcg_gen_movi_i32(cpu_tmp2_i32, CC_OP_ADCOX);
1592 tcg_gen_movi_i32(cpu_tmp3_i32, CC_OP_EFLAGS);
1593 tcg_gen_movcond_i32(TCG_COND_NE, cpu_cc_op, t1, t0,
1594 cpu_tmp2_i32, cpu_tmp3_i32);
1595 tcg_temp_free_i32(t0);
1596 tcg_temp_free_i32(t1);
1598 /* The CC_OP value is no longer predictable. */
1599 set_cc_op(s, CC_OP_DYNAMIC);
1602 static void gen_rot_rm_im(DisasContext *s, TCGMemOp ot, int op1, int op2,
1603 int is_right)
1605 int mask = (ot == MO_64 ? 0x3f : 0x1f);
1606 int shift;
1608 /* load */
1609 if (op1 == OR_TMP0) {
1610 gen_op_ld_v(s, ot, cpu_T[0], cpu_A0);
1611 } else {
1612 gen_op_mov_v_reg(ot, cpu_T[0], op1);
1615 op2 &= mask;
1616 if (op2 != 0) {
1617 switch (ot) {
1618 #ifdef TARGET_X86_64
1619 case MO_32:
1620 tcg_gen_trunc_tl_i32(cpu_tmp2_i32, cpu_T[0]);
1621 if (is_right) {
1622 tcg_gen_rotri_i32(cpu_tmp2_i32, cpu_tmp2_i32, op2);
1623 } else {
1624 tcg_gen_rotli_i32(cpu_tmp2_i32, cpu_tmp2_i32, op2);
1626 tcg_gen_extu_i32_tl(cpu_T[0], cpu_tmp2_i32);
1627 break;
1628 #endif
1629 default:
1630 if (is_right) {
1631 tcg_gen_rotri_tl(cpu_T[0], cpu_T[0], op2);
1632 } else {
1633 tcg_gen_rotli_tl(cpu_T[0], cpu_T[0], op2);
1635 break;
1636 case MO_8:
1637 mask = 7;
1638 goto do_shifts;
1639 case MO_16:
1640 mask = 15;
1641 do_shifts:
1642 shift = op2 & mask;
1643 if (is_right) {
1644 shift = mask + 1 - shift;
1646 gen_extu(ot, cpu_T[0]);
1647 tcg_gen_shli_tl(cpu_tmp0, cpu_T[0], shift);
1648 tcg_gen_shri_tl(cpu_T[0], cpu_T[0], mask + 1 - shift);
1649 tcg_gen_or_tl(cpu_T[0], cpu_T[0], cpu_tmp0);
1650 break;
1654 /* store */
1655 gen_op_st_rm_T0_A0(s, ot, op1);
1657 if (op2 != 0) {
1658 /* Compute the flags into CC_SRC. */
1659 gen_compute_eflags(s);
1661 /* The value that was "rotated out" is now present at the other end
1662 of the word. Compute C into CC_DST and O into CC_SRC2. Note that
1663 since we've computed the flags into CC_SRC, these variables are
1664 currently dead. */
1665 if (is_right) {
1666 tcg_gen_shri_tl(cpu_cc_src2, cpu_T[0], mask - 1);
1667 tcg_gen_shri_tl(cpu_cc_dst, cpu_T[0], mask);
1668 tcg_gen_andi_tl(cpu_cc_dst, cpu_cc_dst, 1);
1669 } else {
1670 tcg_gen_shri_tl(cpu_cc_src2, cpu_T[0], mask);
1671 tcg_gen_andi_tl(cpu_cc_dst, cpu_T[0], 1);
1673 tcg_gen_andi_tl(cpu_cc_src2, cpu_cc_src2, 1);
1674 tcg_gen_xor_tl(cpu_cc_src2, cpu_cc_src2, cpu_cc_dst);
1675 set_cc_op(s, CC_OP_ADCOX);
1679 /* XXX: add faster immediate = 1 case */
1680 static void gen_rotc_rm_T1(DisasContext *s, TCGMemOp ot, int op1,
1681 int is_right)
1683 gen_compute_eflags(s);
1684 assert(s->cc_op == CC_OP_EFLAGS);
1686 /* load */
1687 if (op1 == OR_TMP0)
1688 gen_op_ld_v(s, ot, cpu_T[0], cpu_A0);
1689 else
1690 gen_op_mov_v_reg(ot, cpu_T[0], op1);
1692 if (is_right) {
1693 switch (ot) {
1694 case MO_8:
1695 gen_helper_rcrb(cpu_T[0], cpu_env, cpu_T[0], cpu_T[1]);
1696 break;
1697 case MO_16:
1698 gen_helper_rcrw(cpu_T[0], cpu_env, cpu_T[0], cpu_T[1]);
1699 break;
1700 case MO_32:
1701 gen_helper_rcrl(cpu_T[0], cpu_env, cpu_T[0], cpu_T[1]);
1702 break;
1703 #ifdef TARGET_X86_64
1704 case MO_64:
1705 gen_helper_rcrq(cpu_T[0], cpu_env, cpu_T[0], cpu_T[1]);
1706 break;
1707 #endif
1708 default:
1709 tcg_abort();
1711 } else {
1712 switch (ot) {
1713 case MO_8:
1714 gen_helper_rclb(cpu_T[0], cpu_env, cpu_T[0], cpu_T[1]);
1715 break;
1716 case MO_16:
1717 gen_helper_rclw(cpu_T[0], cpu_env, cpu_T[0], cpu_T[1]);
1718 break;
1719 case MO_32:
1720 gen_helper_rcll(cpu_T[0], cpu_env, cpu_T[0], cpu_T[1]);
1721 break;
1722 #ifdef TARGET_X86_64
1723 case MO_64:
1724 gen_helper_rclq(cpu_T[0], cpu_env, cpu_T[0], cpu_T[1]);
1725 break;
1726 #endif
1727 default:
1728 tcg_abort();
1731 /* store */
1732 gen_op_st_rm_T0_A0(s, ot, op1);
1735 /* XXX: add faster immediate case */
1736 static void gen_shiftd_rm_T1(DisasContext *s, TCGMemOp ot, int op1,
1737 bool is_right, TCGv count_in)
1739 target_ulong mask = (ot == MO_64 ? 63 : 31);
1740 TCGv count;
1742 /* load */
1743 if (op1 == OR_TMP0) {
1744 gen_op_ld_v(s, ot, cpu_T[0], cpu_A0);
1745 } else {
1746 gen_op_mov_v_reg(ot, cpu_T[0], op1);
1749 count = tcg_temp_new();
1750 tcg_gen_andi_tl(count, count_in, mask);
1752 switch (ot) {
1753 case MO_16:
1754 /* Note: we implement the Intel behaviour for shift count > 16.
1755 This means "shrdw C, B, A" shifts A:B:A >> C. Build the B:A
1756 portion by constructing it as a 32-bit value. */
1757 if (is_right) {
1758 tcg_gen_deposit_tl(cpu_tmp0, cpu_T[0], cpu_T[1], 16, 16);
1759 tcg_gen_mov_tl(cpu_T[1], cpu_T[0]);
1760 tcg_gen_mov_tl(cpu_T[0], cpu_tmp0);
1761 } else {
1762 tcg_gen_deposit_tl(cpu_T[1], cpu_T[0], cpu_T[1], 16, 16);
1764 /* FALLTHRU */
1765 #ifdef TARGET_X86_64
1766 case MO_32:
1767 /* Concatenate the two 32-bit values and use a 64-bit shift. */
1768 tcg_gen_subi_tl(cpu_tmp0, count, 1);
1769 if (is_right) {
1770 tcg_gen_concat_tl_i64(cpu_T[0], cpu_T[0], cpu_T[1]);
1771 tcg_gen_shr_i64(cpu_tmp0, cpu_T[0], cpu_tmp0);
1772 tcg_gen_shr_i64(cpu_T[0], cpu_T[0], count);
1773 } else {
1774 tcg_gen_concat_tl_i64(cpu_T[0], cpu_T[1], cpu_T[0]);
1775 tcg_gen_shl_i64(cpu_tmp0, cpu_T[0], cpu_tmp0);
1776 tcg_gen_shl_i64(cpu_T[0], cpu_T[0], count);
1777 tcg_gen_shri_i64(cpu_tmp0, cpu_tmp0, 32);
1778 tcg_gen_shri_i64(cpu_T[0], cpu_T[0], 32);
1780 break;
1781 #endif
1782 default:
1783 tcg_gen_subi_tl(cpu_tmp0, count, 1);
1784 if (is_right) {
1785 tcg_gen_shr_tl(cpu_tmp0, cpu_T[0], cpu_tmp0);
1787 tcg_gen_subfi_tl(cpu_tmp4, mask + 1, count);
1788 tcg_gen_shr_tl(cpu_T[0], cpu_T[0], count);
1789 tcg_gen_shl_tl(cpu_T[1], cpu_T[1], cpu_tmp4);
1790 } else {
1791 tcg_gen_shl_tl(cpu_tmp0, cpu_T[0], cpu_tmp0);
1792 if (ot == MO_16) {
1793 /* Only needed if count > 16, for Intel behaviour. */
1794 tcg_gen_subfi_tl(cpu_tmp4, 33, count);
1795 tcg_gen_shr_tl(cpu_tmp4, cpu_T[1], cpu_tmp4);
1796 tcg_gen_or_tl(cpu_tmp0, cpu_tmp0, cpu_tmp4);
1799 tcg_gen_subfi_tl(cpu_tmp4, mask + 1, count);
1800 tcg_gen_shl_tl(cpu_T[0], cpu_T[0], count);
1801 tcg_gen_shr_tl(cpu_T[1], cpu_T[1], cpu_tmp4);
1803 tcg_gen_movi_tl(cpu_tmp4, 0);
1804 tcg_gen_movcond_tl(TCG_COND_EQ, cpu_T[1], count, cpu_tmp4,
1805 cpu_tmp4, cpu_T[1]);
1806 tcg_gen_or_tl(cpu_T[0], cpu_T[0], cpu_T[1]);
1807 break;
1810 /* store */
1811 gen_op_st_rm_T0_A0(s, ot, op1);
1813 gen_shift_flags(s, ot, cpu_T[0], cpu_tmp0, count, is_right);
1814 tcg_temp_free(count);
1817 static void gen_shift(DisasContext *s1, int op, TCGMemOp ot, int d, int s)
1819 if (s != OR_TMP1)
1820 gen_op_mov_v_reg(ot, cpu_T[1], s);
1821 switch(op) {
1822 case OP_ROL:
1823 gen_rot_rm_T1(s1, ot, d, 0);
1824 break;
1825 case OP_ROR:
1826 gen_rot_rm_T1(s1, ot, d, 1);
1827 break;
1828 case OP_SHL:
1829 case OP_SHL1:
1830 gen_shift_rm_T1(s1, ot, d, 0, 0);
1831 break;
1832 case OP_SHR:
1833 gen_shift_rm_T1(s1, ot, d, 1, 0);
1834 break;
1835 case OP_SAR:
1836 gen_shift_rm_T1(s1, ot, d, 1, 1);
1837 break;
1838 case OP_RCL:
1839 gen_rotc_rm_T1(s1, ot, d, 0);
1840 break;
1841 case OP_RCR:
1842 gen_rotc_rm_T1(s1, ot, d, 1);
1843 break;
1847 static void gen_shifti(DisasContext *s1, int op, TCGMemOp ot, int d, int c)
1849 switch(op) {
1850 case OP_ROL:
1851 gen_rot_rm_im(s1, ot, d, c, 0);
1852 break;
1853 case OP_ROR:
1854 gen_rot_rm_im(s1, ot, d, c, 1);
1855 break;
1856 case OP_SHL:
1857 case OP_SHL1:
1858 gen_shift_rm_im(s1, ot, d, c, 0, 0);
1859 break;
1860 case OP_SHR:
1861 gen_shift_rm_im(s1, ot, d, c, 1, 0);
1862 break;
1863 case OP_SAR:
1864 gen_shift_rm_im(s1, ot, d, c, 1, 1);
1865 break;
1866 default:
1867 /* currently not optimized */
1868 tcg_gen_movi_tl(cpu_T[1], c);
1869 gen_shift(s1, op, ot, d, OR_TMP1);
1870 break;
1874 static void gen_lea_modrm(CPUX86State *env, DisasContext *s, int modrm)
1876 target_long disp;
1877 int havesib;
1878 int base;
1879 int index;
1880 int scale;
1881 int mod, rm, code, override, must_add_seg;
1882 TCGv sum;
1884 override = s->override;
1885 must_add_seg = s->addseg;
1886 if (override >= 0)
1887 must_add_seg = 1;
1888 mod = (modrm >> 6) & 3;
1889 rm = modrm & 7;
1891 switch (s->aflag) {
1892 case MO_64:
1893 case MO_32:
1894 havesib = 0;
1895 base = rm;
1896 index = -1;
1897 scale = 0;
1899 if (base == 4) {
1900 havesib = 1;
1901 code = cpu_ldub_code(env, s->pc++);
1902 scale = (code >> 6) & 3;
1903 index = ((code >> 3) & 7) | REX_X(s);
1904 if (index == 4) {
1905 index = -1; /* no index */
1907 base = (code & 7);
1909 base |= REX_B(s);
1911 switch (mod) {
1912 case 0:
1913 if ((base & 7) == 5) {
1914 base = -1;
1915 disp = (int32_t)cpu_ldl_code(env, s->pc);
1916 s->pc += 4;
1917 if (CODE64(s) && !havesib) {
1918 disp += s->pc + s->rip_offset;
1920 } else {
1921 disp = 0;
1923 break;
1924 case 1:
1925 disp = (int8_t)cpu_ldub_code(env, s->pc++);
1926 break;
1927 default:
1928 case 2:
1929 disp = (int32_t)cpu_ldl_code(env, s->pc);
1930 s->pc += 4;
1931 break;
1934 /* For correct popl handling with esp. */
1935 if (base == R_ESP && s->popl_esp_hack) {
1936 disp += s->popl_esp_hack;
1939 /* Compute the address, with a minimum number of TCG ops. */
1940 TCGV_UNUSED(sum);
1941 if (index >= 0) {
1942 if (scale == 0) {
1943 sum = cpu_regs[index];
1944 } else {
1945 tcg_gen_shli_tl(cpu_A0, cpu_regs[index], scale);
1946 sum = cpu_A0;
1948 if (base >= 0) {
1949 tcg_gen_add_tl(cpu_A0, sum, cpu_regs[base]);
1950 sum = cpu_A0;
1952 } else if (base >= 0) {
1953 sum = cpu_regs[base];
1955 if (TCGV_IS_UNUSED(sum)) {
1956 tcg_gen_movi_tl(cpu_A0, disp);
1957 } else {
1958 tcg_gen_addi_tl(cpu_A0, sum, disp);
1961 if (must_add_seg) {
1962 if (override < 0) {
1963 if (base == R_EBP || base == R_ESP) {
1964 override = R_SS;
1965 } else {
1966 override = R_DS;
1970 tcg_gen_ld_tl(cpu_tmp0, cpu_env,
1971 offsetof(CPUX86State, segs[override].base));
1972 if (CODE64(s)) {
1973 if (s->aflag == MO_32) {
1974 tcg_gen_ext32u_tl(cpu_A0, cpu_A0);
1976 tcg_gen_add_tl(cpu_A0, cpu_A0, cpu_tmp0);
1977 return;
1980 tcg_gen_add_tl(cpu_A0, cpu_A0, cpu_tmp0);
1983 if (s->aflag == MO_32) {
1984 tcg_gen_ext32u_tl(cpu_A0, cpu_A0);
1986 break;
1988 case MO_16:
1989 switch (mod) {
1990 case 0:
1991 if (rm == 6) {
1992 disp = cpu_lduw_code(env, s->pc);
1993 s->pc += 2;
1994 tcg_gen_movi_tl(cpu_A0, disp);
1995 rm = 0; /* avoid SS override */
1996 goto no_rm;
1997 } else {
1998 disp = 0;
2000 break;
2001 case 1:
2002 disp = (int8_t)cpu_ldub_code(env, s->pc++);
2003 break;
2004 default:
2005 case 2:
2006 disp = (int16_t)cpu_lduw_code(env, s->pc);
2007 s->pc += 2;
2008 break;
2011 sum = cpu_A0;
2012 switch (rm) {
2013 case 0:
2014 tcg_gen_add_tl(cpu_A0, cpu_regs[R_EBX], cpu_regs[R_ESI]);
2015 break;
2016 case 1:
2017 tcg_gen_add_tl(cpu_A0, cpu_regs[R_EBX], cpu_regs[R_EDI]);
2018 break;
2019 case 2:
2020 tcg_gen_add_tl(cpu_A0, cpu_regs[R_EBP], cpu_regs[R_ESI]);
2021 break;
2022 case 3:
2023 tcg_gen_add_tl(cpu_A0, cpu_regs[R_EBP], cpu_regs[R_EDI]);
2024 break;
2025 case 4:
2026 sum = cpu_regs[R_ESI];
2027 break;
2028 case 5:
2029 sum = cpu_regs[R_EDI];
2030 break;
2031 case 6:
2032 sum = cpu_regs[R_EBP];
2033 break;
2034 default:
2035 case 7:
2036 sum = cpu_regs[R_EBX];
2037 break;
2039 tcg_gen_addi_tl(cpu_A0, sum, disp);
2040 tcg_gen_ext16u_tl(cpu_A0, cpu_A0);
2041 no_rm:
2042 if (must_add_seg) {
2043 if (override < 0) {
2044 if (rm == 2 || rm == 3 || rm == 6) {
2045 override = R_SS;
2046 } else {
2047 override = R_DS;
2050 gen_op_addl_A0_seg(s, override);
2052 break;
2054 default:
2055 tcg_abort();
2059 static void gen_nop_modrm(CPUX86State *env, DisasContext *s, int modrm)
2061 int mod, rm, base, code;
2063 mod = (modrm >> 6) & 3;
2064 if (mod == 3)
2065 return;
2066 rm = modrm & 7;
2068 switch (s->aflag) {
2069 case MO_64:
2070 case MO_32:
2071 base = rm;
2073 if (base == 4) {
2074 code = cpu_ldub_code(env, s->pc++);
2075 base = (code & 7);
2078 switch (mod) {
2079 case 0:
2080 if (base == 5) {
2081 s->pc += 4;
2083 break;
2084 case 1:
2085 s->pc++;
2086 break;
2087 default:
2088 case 2:
2089 s->pc += 4;
2090 break;
2092 break;
2094 case MO_16:
2095 switch (mod) {
2096 case 0:
2097 if (rm == 6) {
2098 s->pc += 2;
2100 break;
2101 case 1:
2102 s->pc++;
2103 break;
2104 default:
2105 case 2:
2106 s->pc += 2;
2107 break;
2109 break;
2111 default:
2112 tcg_abort();
2116 /* used for LEA and MOV AX, mem */
2117 static void gen_add_A0_ds_seg(DisasContext *s)
2119 int override, must_add_seg;
2120 must_add_seg = s->addseg;
2121 override = R_DS;
2122 if (s->override >= 0) {
2123 override = s->override;
2124 must_add_seg = 1;
2126 if (must_add_seg) {
2127 #ifdef TARGET_X86_64
2128 if (CODE64(s)) {
2129 gen_op_addq_A0_seg(override);
2130 } else
2131 #endif
2133 gen_op_addl_A0_seg(s, override);
2138 /* generate modrm memory load or store of 'reg'. TMP0 is used if reg ==
2139 OR_TMP0 */
2140 static void gen_ldst_modrm(CPUX86State *env, DisasContext *s, int modrm,
2141 TCGMemOp ot, int reg, int is_store)
2143 int mod, rm;
2145 mod = (modrm >> 6) & 3;
2146 rm = (modrm & 7) | REX_B(s);
2147 if (mod == 3) {
2148 if (is_store) {
2149 if (reg != OR_TMP0)
2150 gen_op_mov_v_reg(ot, cpu_T[0], reg);
2151 gen_op_mov_reg_v(ot, rm, cpu_T[0]);
2152 } else {
2153 gen_op_mov_v_reg(ot, cpu_T[0], rm);
2154 if (reg != OR_TMP0)
2155 gen_op_mov_reg_v(ot, reg, cpu_T[0]);
2157 } else {
2158 gen_lea_modrm(env, s, modrm);
2159 if (is_store) {
2160 if (reg != OR_TMP0)
2161 gen_op_mov_v_reg(ot, cpu_T[0], reg);
2162 gen_op_st_v(s, ot, cpu_T[0], cpu_A0);
2163 } else {
2164 gen_op_ld_v(s, ot, cpu_T[0], cpu_A0);
2165 if (reg != OR_TMP0)
2166 gen_op_mov_reg_v(ot, reg, cpu_T[0]);
2171 static inline uint32_t insn_get(CPUX86State *env, DisasContext *s, TCGMemOp ot)
2173 uint32_t ret;
2175 switch (ot) {
2176 case MO_8:
2177 ret = cpu_ldub_code(env, s->pc);
2178 s->pc++;
2179 break;
2180 case MO_16:
2181 ret = cpu_lduw_code(env, s->pc);
2182 s->pc += 2;
2183 break;
2184 case MO_32:
2185 #ifdef TARGET_X86_64
2186 case MO_64:
2187 #endif
2188 ret = cpu_ldl_code(env, s->pc);
2189 s->pc += 4;
2190 break;
2191 default:
2192 tcg_abort();
2194 return ret;
2197 static inline int insn_const_size(TCGMemOp ot)
2199 if (ot <= MO_32) {
2200 return 1 << ot;
2201 } else {
2202 return 4;
2206 static inline void gen_goto_tb(DisasContext *s, int tb_num, target_ulong eip)
2208 TranslationBlock *tb;
2209 target_ulong pc;
2211 pc = s->cs_base + eip;
2212 tb = s->tb;
2213 /* NOTE: we handle the case where the TB spans two pages here */
2214 if ((pc & TARGET_PAGE_MASK) == (tb->pc & TARGET_PAGE_MASK) ||
2215 (pc & TARGET_PAGE_MASK) == ((s->pc - 1) & TARGET_PAGE_MASK)) {
2216 /* jump to same page: we can use a direct jump */
2217 tcg_gen_goto_tb(tb_num);
2218 gen_jmp_im(eip);
2219 tcg_gen_exit_tb((uintptr_t)tb + tb_num);
2220 } else {
2221 /* jump to another page: currently not optimized */
2222 gen_jmp_im(eip);
2223 gen_eob(s);
2227 static inline void gen_jcc(DisasContext *s, int b,
2228 target_ulong val, target_ulong next_eip)
2230 int l1, l2;
2232 if (s->jmp_opt) {
2233 l1 = gen_new_label();
2234 gen_jcc1(s, b, l1);
2236 gen_goto_tb(s, 0, next_eip);
2238 gen_set_label(l1);
2239 gen_goto_tb(s, 1, val);
2240 s->is_jmp = DISAS_TB_JUMP;
2241 } else {
2242 l1 = gen_new_label();
2243 l2 = gen_new_label();
2244 gen_jcc1(s, b, l1);
2246 gen_jmp_im(next_eip);
2247 tcg_gen_br(l2);
2249 gen_set_label(l1);
2250 gen_jmp_im(val);
2251 gen_set_label(l2);
2252 gen_eob(s);
2256 static void gen_cmovcc1(CPUX86State *env, DisasContext *s, TCGMemOp ot, int b,
2257 int modrm, int reg)
2259 CCPrepare cc;
2261 gen_ldst_modrm(env, s, modrm, ot, OR_TMP0, 0);
2263 cc = gen_prepare_cc(s, b, cpu_T[1]);
2264 if (cc.mask != -1) {
2265 TCGv t0 = tcg_temp_new();
2266 tcg_gen_andi_tl(t0, cc.reg, cc.mask);
2267 cc.reg = t0;
2269 if (!cc.use_reg2) {
2270 cc.reg2 = tcg_const_tl(cc.imm);
2273 tcg_gen_movcond_tl(cc.cond, cpu_T[0], cc.reg, cc.reg2,
2274 cpu_T[0], cpu_regs[reg]);
2275 gen_op_mov_reg_v(ot, reg, cpu_T[0]);
2277 if (cc.mask != -1) {
2278 tcg_temp_free(cc.reg);
2280 if (!cc.use_reg2) {
2281 tcg_temp_free(cc.reg2);
2285 static inline void gen_op_movl_T0_seg(int seg_reg)
2287 tcg_gen_ld32u_tl(cpu_T[0], cpu_env,
2288 offsetof(CPUX86State,segs[seg_reg].selector));
2291 static inline void gen_op_movl_seg_T0_vm(int seg_reg)
2293 tcg_gen_andi_tl(cpu_T[0], cpu_T[0], 0xffff);
2294 tcg_gen_st32_tl(cpu_T[0], cpu_env,
2295 offsetof(CPUX86State,segs[seg_reg].selector));
2296 tcg_gen_shli_tl(cpu_T[0], cpu_T[0], 4);
2297 tcg_gen_st_tl(cpu_T[0], cpu_env,
2298 offsetof(CPUX86State,segs[seg_reg].base));
2301 /* move T0 to seg_reg and compute if the CPU state may change. Never
2302 call this function with seg_reg == R_CS */
2303 static void gen_movl_seg_T0(DisasContext *s, int seg_reg, target_ulong cur_eip)
2305 if (s->pe && !s->vm86) {
2306 /* XXX: optimize by finding processor state dynamically */
2307 gen_update_cc_op(s);
2308 gen_jmp_im(cur_eip);
2309 tcg_gen_trunc_tl_i32(cpu_tmp2_i32, cpu_T[0]);
2310 gen_helper_load_seg(cpu_env, tcg_const_i32(seg_reg), cpu_tmp2_i32);
2311 /* abort translation because the addseg value may change or
2312 because ss32 may change. For R_SS, translation must always
2313 stop as a special handling must be done to disable hardware
2314 interrupts for the next instruction */
2315 if (seg_reg == R_SS || (s->code32 && seg_reg < R_FS))
2316 s->is_jmp = DISAS_TB_JUMP;
2317 } else {
2318 gen_op_movl_seg_T0_vm(seg_reg);
2319 if (seg_reg == R_SS)
2320 s->is_jmp = DISAS_TB_JUMP;
2324 static inline int svm_is_rep(int prefixes)
2326 return ((prefixes & (PREFIX_REPZ | PREFIX_REPNZ)) ? 8 : 0);
2329 static inline void
2330 gen_svm_check_intercept_param(DisasContext *s, target_ulong pc_start,
2331 uint32_t type, uint64_t param)
2333 /* no SVM activated; fast case */
2334 if (likely(!(s->flags & HF_SVMI_MASK)))
2335 return;
2336 gen_update_cc_op(s);
2337 gen_jmp_im(pc_start - s->cs_base);
2338 gen_helper_svm_check_intercept_param(cpu_env, tcg_const_i32(type),
2339 tcg_const_i64(param));
2342 static inline void
2343 gen_svm_check_intercept(DisasContext *s, target_ulong pc_start, uint64_t type)
2345 gen_svm_check_intercept_param(s, pc_start, type, 0);
2348 static inline void gen_stack_update(DisasContext *s, int addend)
2350 #ifdef TARGET_X86_64
2351 if (CODE64(s)) {
2352 gen_op_add_reg_im(MO_64, R_ESP, addend);
2353 } else
2354 #endif
2355 if (s->ss32) {
2356 gen_op_add_reg_im(MO_32, R_ESP, addend);
2357 } else {
2358 gen_op_add_reg_im(MO_16, R_ESP, addend);
2362 /* Generate a push. It depends on ss32, addseg and dflag. */
2363 static void gen_push_v(DisasContext *s, TCGv val)
2365 TCGMemOp a_ot, d_ot = mo_pushpop(s, s->dflag);
2366 int size = 1 << d_ot;
2367 TCGv new_esp = cpu_A0;
2369 tcg_gen_subi_tl(cpu_A0, cpu_regs[R_ESP], size);
2371 if (CODE64(s)) {
2372 a_ot = MO_64;
2373 } else if (s->ss32) {
2374 a_ot = MO_32;
2375 if (s->addseg) {
2376 new_esp = cpu_tmp4;
2377 tcg_gen_mov_tl(new_esp, cpu_A0);
2378 gen_op_addl_A0_seg(s, R_SS);
2379 } else {
2380 tcg_gen_ext32u_tl(cpu_A0, cpu_A0);
2382 } else {
2383 a_ot = MO_16;
2384 new_esp = cpu_tmp4;
2385 tcg_gen_ext16u_tl(cpu_A0, cpu_A0);
2386 tcg_gen_mov_tl(new_esp, cpu_A0);
2387 gen_op_addl_A0_seg(s, R_SS);
2390 gen_op_st_v(s, d_ot, val, cpu_A0);
2391 gen_op_mov_reg_v(a_ot, R_ESP, new_esp);
2394 /* two step pop is necessary for precise exceptions */
2395 static TCGMemOp gen_pop_T0(DisasContext *s)
2397 TCGMemOp d_ot = mo_pushpop(s, s->dflag);
2398 TCGv addr = cpu_A0;
2400 if (CODE64(s)) {
2401 addr = cpu_regs[R_ESP];
2402 } else if (!s->ss32) {
2403 tcg_gen_ext16u_tl(cpu_A0, cpu_regs[R_ESP]);
2404 gen_op_addl_A0_seg(s, R_SS);
2405 } else if (s->addseg) {
2406 tcg_gen_mov_tl(cpu_A0, cpu_regs[R_ESP]);
2407 gen_op_addl_A0_seg(s, R_SS);
2408 } else {
2409 tcg_gen_ext32u_tl(cpu_A0, cpu_regs[R_ESP]);
2412 gen_op_ld_v(s, d_ot, cpu_T[0], addr);
2413 return d_ot;
2416 static void gen_pop_update(DisasContext *s, TCGMemOp ot)
2418 gen_stack_update(s, 1 << ot);
2421 static void gen_stack_A0(DisasContext *s)
2423 gen_op_movl_A0_reg(R_ESP);
2424 if (!s->ss32)
2425 tcg_gen_ext16u_tl(cpu_A0, cpu_A0);
2426 tcg_gen_mov_tl(cpu_T[1], cpu_A0);
2427 if (s->addseg)
2428 gen_op_addl_A0_seg(s, R_SS);
2431 /* NOTE: wrap around in 16 bit not fully handled */
2432 static void gen_pusha(DisasContext *s)
2434 int i;
2435 gen_op_movl_A0_reg(R_ESP);
2436 gen_op_addl_A0_im(-8 << s->dflag);
2437 if (!s->ss32)
2438 tcg_gen_ext16u_tl(cpu_A0, cpu_A0);
2439 tcg_gen_mov_tl(cpu_T[1], cpu_A0);
2440 if (s->addseg)
2441 gen_op_addl_A0_seg(s, R_SS);
2442 for(i = 0;i < 8; i++) {
2443 gen_op_mov_v_reg(MO_32, cpu_T[0], 7 - i);
2444 gen_op_st_v(s, s->dflag, cpu_T[0], cpu_A0);
2445 gen_op_addl_A0_im(1 << s->dflag);
2447 gen_op_mov_reg_v(MO_16 + s->ss32, R_ESP, cpu_T[1]);
2450 /* NOTE: wrap around in 16 bit not fully handled */
2451 static void gen_popa(DisasContext *s)
2453 int i;
2454 gen_op_movl_A0_reg(R_ESP);
2455 if (!s->ss32)
2456 tcg_gen_ext16u_tl(cpu_A0, cpu_A0);
2457 tcg_gen_mov_tl(cpu_T[1], cpu_A0);
2458 tcg_gen_addi_tl(cpu_T[1], cpu_T[1], 8 << s->dflag);
2459 if (s->addseg)
2460 gen_op_addl_A0_seg(s, R_SS);
2461 for(i = 0;i < 8; i++) {
2462 /* ESP is not reloaded */
2463 if (i != 3) {
2464 gen_op_ld_v(s, s->dflag, cpu_T[0], cpu_A0);
2465 gen_op_mov_reg_v(s->dflag, 7 - i, cpu_T[0]);
2467 gen_op_addl_A0_im(1 << s->dflag);
2469 gen_op_mov_reg_v(MO_16 + s->ss32, R_ESP, cpu_T[1]);
2472 static void gen_enter(DisasContext *s, int esp_addend, int level)
2474 TCGMemOp ot = mo_pushpop(s, s->dflag);
2475 int opsize = 1 << ot;
2477 level &= 0x1f;
2478 #ifdef TARGET_X86_64
2479 if (CODE64(s)) {
2480 gen_op_movl_A0_reg(R_ESP);
2481 gen_op_addq_A0_im(-opsize);
2482 tcg_gen_mov_tl(cpu_T[1], cpu_A0);
2484 /* push bp */
2485 gen_op_mov_v_reg(MO_32, cpu_T[0], R_EBP);
2486 gen_op_st_v(s, ot, cpu_T[0], cpu_A0);
2487 if (level) {
2488 /* XXX: must save state */
2489 gen_helper_enter64_level(cpu_env, tcg_const_i32(level),
2490 tcg_const_i32((ot == MO_64)),
2491 cpu_T[1]);
2493 gen_op_mov_reg_v(ot, R_EBP, cpu_T[1]);
2494 tcg_gen_addi_tl(cpu_T[1], cpu_T[1], -esp_addend + (-opsize * level));
2495 gen_op_mov_reg_v(MO_64, R_ESP, cpu_T[1]);
2496 } else
2497 #endif
2499 gen_op_movl_A0_reg(R_ESP);
2500 gen_op_addl_A0_im(-opsize);
2501 if (!s->ss32)
2502 tcg_gen_ext16u_tl(cpu_A0, cpu_A0);
2503 tcg_gen_mov_tl(cpu_T[1], cpu_A0);
2504 if (s->addseg)
2505 gen_op_addl_A0_seg(s, R_SS);
2506 /* push bp */
2507 gen_op_mov_v_reg(MO_32, cpu_T[0], R_EBP);
2508 gen_op_st_v(s, ot, cpu_T[0], cpu_A0);
2509 if (level) {
2510 /* XXX: must save state */
2511 gen_helper_enter_level(cpu_env, tcg_const_i32(level),
2512 tcg_const_i32(s->dflag - 1),
2513 cpu_T[1]);
2515 gen_op_mov_reg_v(ot, R_EBP, cpu_T[1]);
2516 tcg_gen_addi_tl(cpu_T[1], cpu_T[1], -esp_addend + (-opsize * level));
2517 gen_op_mov_reg_v(MO_16 + s->ss32, R_ESP, cpu_T[1]);
2521 static void gen_exception(DisasContext *s, int trapno, target_ulong cur_eip)
2523 gen_update_cc_op(s);
2524 gen_jmp_im(cur_eip);
2525 gen_helper_raise_exception(cpu_env, tcg_const_i32(trapno));
2526 s->is_jmp = DISAS_TB_JUMP;
2529 /* an interrupt is different from an exception because of the
2530 privilege checks */
2531 static void gen_interrupt(DisasContext *s, int intno,
2532 target_ulong cur_eip, target_ulong next_eip)
2534 gen_update_cc_op(s);
2535 gen_jmp_im(cur_eip);
2536 gen_helper_raise_interrupt(cpu_env, tcg_const_i32(intno),
2537 tcg_const_i32(next_eip - cur_eip));
2538 s->is_jmp = DISAS_TB_JUMP;
2541 static void gen_debug(DisasContext *s, target_ulong cur_eip)
2543 gen_update_cc_op(s);
2544 gen_jmp_im(cur_eip);
2545 gen_helper_debug(cpu_env);
2546 s->is_jmp = DISAS_TB_JUMP;
2549 /* generate a generic end of block. Trace exception is also generated
2550 if needed */
2551 static void gen_eob(DisasContext *s)
2553 gen_update_cc_op(s);
2554 if (s->tb->flags & HF_INHIBIT_IRQ_MASK) {
2555 gen_helper_reset_inhibit_irq(cpu_env);
2557 if (s->tb->flags & HF_RF_MASK) {
2558 gen_helper_reset_rf(cpu_env);
2560 if (s->singlestep_enabled) {
2561 gen_helper_debug(cpu_env);
2562 } else if (s->tf) {
2563 gen_helper_single_step(cpu_env);
2564 } else {
2565 tcg_gen_exit_tb(0);
2567 s->is_jmp = DISAS_TB_JUMP;
2570 /* generate a jump to eip. No segment change must happen before as a
2571 direct call to the next block may occur */
2572 static void gen_jmp_tb(DisasContext *s, target_ulong eip, int tb_num)
2574 gen_update_cc_op(s);
2575 set_cc_op(s, CC_OP_DYNAMIC);
2576 if (s->jmp_opt) {
2577 gen_goto_tb(s, tb_num, eip);
2578 s->is_jmp = DISAS_TB_JUMP;
2579 } else {
2580 gen_jmp_im(eip);
2581 gen_eob(s);
2585 static void gen_jmp(DisasContext *s, target_ulong eip)
2587 gen_jmp_tb(s, eip, 0);
2590 static inline void gen_ldq_env_A0(DisasContext *s, int offset)
2592 tcg_gen_qemu_ld_i64(cpu_tmp1_i64, cpu_A0, s->mem_index, MO_LEQ);
2593 tcg_gen_st_i64(cpu_tmp1_i64, cpu_env, offset);
2596 static inline void gen_stq_env_A0(DisasContext *s, int offset)
2598 tcg_gen_ld_i64(cpu_tmp1_i64, cpu_env, offset);
2599 tcg_gen_qemu_st_i64(cpu_tmp1_i64, cpu_A0, s->mem_index, MO_LEQ);
2602 static inline void gen_ldo_env_A0(DisasContext *s, int offset)
2604 int mem_index = s->mem_index;
2605 tcg_gen_qemu_ld_i64(cpu_tmp1_i64, cpu_A0, mem_index, MO_LEQ);
2606 tcg_gen_st_i64(cpu_tmp1_i64, cpu_env, offset + offsetof(XMMReg, XMM_Q(0)));
2607 tcg_gen_addi_tl(cpu_tmp0, cpu_A0, 8);
2608 tcg_gen_qemu_ld_i64(cpu_tmp1_i64, cpu_tmp0, mem_index, MO_LEQ);
2609 tcg_gen_st_i64(cpu_tmp1_i64, cpu_env, offset + offsetof(XMMReg, XMM_Q(1)));
2612 static inline void gen_sto_env_A0(DisasContext *s, int offset)
2614 int mem_index = s->mem_index;
2615 tcg_gen_ld_i64(cpu_tmp1_i64, cpu_env, offset + offsetof(XMMReg, XMM_Q(0)));
2616 tcg_gen_qemu_st_i64(cpu_tmp1_i64, cpu_A0, mem_index, MO_LEQ);
2617 tcg_gen_addi_tl(cpu_tmp0, cpu_A0, 8);
2618 tcg_gen_ld_i64(cpu_tmp1_i64, cpu_env, offset + offsetof(XMMReg, XMM_Q(1)));
2619 tcg_gen_qemu_st_i64(cpu_tmp1_i64, cpu_tmp0, mem_index, MO_LEQ);
2622 static inline void gen_op_movo(int d_offset, int s_offset)
2624 tcg_gen_ld_i64(cpu_tmp1_i64, cpu_env, s_offset);
2625 tcg_gen_st_i64(cpu_tmp1_i64, cpu_env, d_offset);
2626 tcg_gen_ld_i64(cpu_tmp1_i64, cpu_env, s_offset + 8);
2627 tcg_gen_st_i64(cpu_tmp1_i64, cpu_env, d_offset + 8);
2630 static inline void gen_op_movq(int d_offset, int s_offset)
2632 tcg_gen_ld_i64(cpu_tmp1_i64, cpu_env, s_offset);
2633 tcg_gen_st_i64(cpu_tmp1_i64, cpu_env, d_offset);
2636 static inline void gen_op_movl(int d_offset, int s_offset)
2638 tcg_gen_ld_i32(cpu_tmp2_i32, cpu_env, s_offset);
2639 tcg_gen_st_i32(cpu_tmp2_i32, cpu_env, d_offset);
2642 static inline void gen_op_movq_env_0(int d_offset)
2644 tcg_gen_movi_i64(cpu_tmp1_i64, 0);
2645 tcg_gen_st_i64(cpu_tmp1_i64, cpu_env, d_offset);
2648 typedef void (*SSEFunc_i_ep)(TCGv_i32 val, TCGv_ptr env, TCGv_ptr reg);
2649 typedef void (*SSEFunc_l_ep)(TCGv_i64 val, TCGv_ptr env, TCGv_ptr reg);
2650 typedef void (*SSEFunc_0_epi)(TCGv_ptr env, TCGv_ptr reg, TCGv_i32 val);
2651 typedef void (*SSEFunc_0_epl)(TCGv_ptr env, TCGv_ptr reg, TCGv_i64 val);
2652 typedef void (*SSEFunc_0_epp)(TCGv_ptr env, TCGv_ptr reg_a, TCGv_ptr reg_b);
2653 typedef void (*SSEFunc_0_eppi)(TCGv_ptr env, TCGv_ptr reg_a, TCGv_ptr reg_b,
2654 TCGv_i32 val);
2655 typedef void (*SSEFunc_0_ppi)(TCGv_ptr reg_a, TCGv_ptr reg_b, TCGv_i32 val);
2656 typedef void (*SSEFunc_0_eppt)(TCGv_ptr env, TCGv_ptr reg_a, TCGv_ptr reg_b,
2657 TCGv val);
2659 #define SSE_SPECIAL ((void *)1)
2660 #define SSE_DUMMY ((void *)2)
2662 #define MMX_OP2(x) { gen_helper_ ## x ## _mmx, gen_helper_ ## x ## _xmm }
2663 #define SSE_FOP(x) { gen_helper_ ## x ## ps, gen_helper_ ## x ## pd, \
2664 gen_helper_ ## x ## ss, gen_helper_ ## x ## sd, }
2666 static const SSEFunc_0_epp sse_op_table1[256][4] = {
2667 /* 3DNow! extensions */
2668 [0x0e] = { SSE_DUMMY }, /* femms */
2669 [0x0f] = { SSE_DUMMY }, /* pf... */
2670 /* pure SSE operations */
2671 [0x10] = { SSE_SPECIAL, SSE_SPECIAL, SSE_SPECIAL, SSE_SPECIAL }, /* movups, movupd, movss, movsd */
2672 [0x11] = { SSE_SPECIAL, SSE_SPECIAL, SSE_SPECIAL, SSE_SPECIAL }, /* movups, movupd, movss, movsd */
2673 [0x12] = { SSE_SPECIAL, SSE_SPECIAL, SSE_SPECIAL, SSE_SPECIAL }, /* movlps, movlpd, movsldup, movddup */
2674 [0x13] = { SSE_SPECIAL, SSE_SPECIAL }, /* movlps, movlpd */
2675 [0x14] = { gen_helper_punpckldq_xmm, gen_helper_punpcklqdq_xmm },
2676 [0x15] = { gen_helper_punpckhdq_xmm, gen_helper_punpckhqdq_xmm },
2677 [0x16] = { SSE_SPECIAL, SSE_SPECIAL, SSE_SPECIAL }, /* movhps, movhpd, movshdup */
2678 [0x17] = { SSE_SPECIAL, SSE_SPECIAL }, /* movhps, movhpd */
2680 [0x28] = { SSE_SPECIAL, SSE_SPECIAL }, /* movaps, movapd */
2681 [0x29] = { SSE_SPECIAL, SSE_SPECIAL }, /* movaps, movapd */
2682 [0x2a] = { SSE_SPECIAL, SSE_SPECIAL, SSE_SPECIAL, SSE_SPECIAL }, /* cvtpi2ps, cvtpi2pd, cvtsi2ss, cvtsi2sd */
2683 [0x2b] = { SSE_SPECIAL, SSE_SPECIAL, SSE_SPECIAL, SSE_SPECIAL }, /* movntps, movntpd, movntss, movntsd */
2684 [0x2c] = { SSE_SPECIAL, SSE_SPECIAL, SSE_SPECIAL, SSE_SPECIAL }, /* cvttps2pi, cvttpd2pi, cvttsd2si, cvttss2si */
2685 [0x2d] = { SSE_SPECIAL, SSE_SPECIAL, SSE_SPECIAL, SSE_SPECIAL }, /* cvtps2pi, cvtpd2pi, cvtsd2si, cvtss2si */
2686 [0x2e] = { gen_helper_ucomiss, gen_helper_ucomisd },
2687 [0x2f] = { gen_helper_comiss, gen_helper_comisd },
2688 [0x50] = { SSE_SPECIAL, SSE_SPECIAL }, /* movmskps, movmskpd */
2689 [0x51] = SSE_FOP(sqrt),
2690 [0x52] = { gen_helper_rsqrtps, NULL, gen_helper_rsqrtss, NULL },
2691 [0x53] = { gen_helper_rcpps, NULL, gen_helper_rcpss, NULL },
2692 [0x54] = { gen_helper_pand_xmm, gen_helper_pand_xmm }, /* andps, andpd */
2693 [0x55] = { gen_helper_pandn_xmm, gen_helper_pandn_xmm }, /* andnps, andnpd */
2694 [0x56] = { gen_helper_por_xmm, gen_helper_por_xmm }, /* orps, orpd */
2695 [0x57] = { gen_helper_pxor_xmm, gen_helper_pxor_xmm }, /* xorps, xorpd */
2696 [0x58] = SSE_FOP(add),
2697 [0x59] = SSE_FOP(mul),
2698 [0x5a] = { gen_helper_cvtps2pd, gen_helper_cvtpd2ps,
2699 gen_helper_cvtss2sd, gen_helper_cvtsd2ss },
2700 [0x5b] = { gen_helper_cvtdq2ps, gen_helper_cvtps2dq, gen_helper_cvttps2dq },
2701 [0x5c] = SSE_FOP(sub),
2702 [0x5d] = SSE_FOP(min),
2703 [0x5e] = SSE_FOP(div),
2704 [0x5f] = SSE_FOP(max),
2706 [0xc2] = SSE_FOP(cmpeq),
2707 [0xc6] = { (SSEFunc_0_epp)gen_helper_shufps,
2708 (SSEFunc_0_epp)gen_helper_shufpd }, /* XXX: casts */
2710 /* SSSE3, SSE4, MOVBE, CRC32, BMI1, BMI2, ADX. */
2711 [0x38] = { SSE_SPECIAL, SSE_SPECIAL, SSE_SPECIAL, SSE_SPECIAL },
2712 [0x3a] = { SSE_SPECIAL, SSE_SPECIAL, SSE_SPECIAL, SSE_SPECIAL },
2714 /* MMX ops and their SSE extensions */
2715 [0x60] = MMX_OP2(punpcklbw),
2716 [0x61] = MMX_OP2(punpcklwd),
2717 [0x62] = MMX_OP2(punpckldq),
2718 [0x63] = MMX_OP2(packsswb),
2719 [0x64] = MMX_OP2(pcmpgtb),
2720 [0x65] = MMX_OP2(pcmpgtw),
2721 [0x66] = MMX_OP2(pcmpgtl),
2722 [0x67] = MMX_OP2(packuswb),
2723 [0x68] = MMX_OP2(punpckhbw),
2724 [0x69] = MMX_OP2(punpckhwd),
2725 [0x6a] = MMX_OP2(punpckhdq),
2726 [0x6b] = MMX_OP2(packssdw),
2727 [0x6c] = { NULL, gen_helper_punpcklqdq_xmm },
2728 [0x6d] = { NULL, gen_helper_punpckhqdq_xmm },
2729 [0x6e] = { SSE_SPECIAL, SSE_SPECIAL }, /* movd mm, ea */
2730 [0x6f] = { SSE_SPECIAL, SSE_SPECIAL, SSE_SPECIAL }, /* movq, movdqa, , movqdu */
2731 [0x70] = { (SSEFunc_0_epp)gen_helper_pshufw_mmx,
2732 (SSEFunc_0_epp)gen_helper_pshufd_xmm,
2733 (SSEFunc_0_epp)gen_helper_pshufhw_xmm,
2734 (SSEFunc_0_epp)gen_helper_pshuflw_xmm }, /* XXX: casts */
2735 [0x71] = { SSE_SPECIAL, SSE_SPECIAL }, /* shiftw */
2736 [0x72] = { SSE_SPECIAL, SSE_SPECIAL }, /* shiftd */
2737 [0x73] = { SSE_SPECIAL, SSE_SPECIAL }, /* shiftq */
2738 [0x74] = MMX_OP2(pcmpeqb),
2739 [0x75] = MMX_OP2(pcmpeqw),
2740 [0x76] = MMX_OP2(pcmpeql),
2741 [0x77] = { SSE_DUMMY }, /* emms */
2742 [0x78] = { NULL, SSE_SPECIAL, NULL, SSE_SPECIAL }, /* extrq_i, insertq_i */
2743 [0x79] = { NULL, gen_helper_extrq_r, NULL, gen_helper_insertq_r },
2744 [0x7c] = { NULL, gen_helper_haddpd, NULL, gen_helper_haddps },
2745 [0x7d] = { NULL, gen_helper_hsubpd, NULL, gen_helper_hsubps },
2746 [0x7e] = { SSE_SPECIAL, SSE_SPECIAL, SSE_SPECIAL }, /* movd, movd, , movq */
2747 [0x7f] = { SSE_SPECIAL, SSE_SPECIAL, SSE_SPECIAL }, /* movq, movdqa, movdqu */
2748 [0xc4] = { SSE_SPECIAL, SSE_SPECIAL }, /* pinsrw */
2749 [0xc5] = { SSE_SPECIAL, SSE_SPECIAL }, /* pextrw */
2750 [0xd0] = { NULL, gen_helper_addsubpd, NULL, gen_helper_addsubps },
2751 [0xd1] = MMX_OP2(psrlw),
2752 [0xd2] = MMX_OP2(psrld),
2753 [0xd3] = MMX_OP2(psrlq),
2754 [0xd4] = MMX_OP2(paddq),
2755 [0xd5] = MMX_OP2(pmullw),
2756 [0xd6] = { NULL, SSE_SPECIAL, SSE_SPECIAL, SSE_SPECIAL },
2757 [0xd7] = { SSE_SPECIAL, SSE_SPECIAL }, /* pmovmskb */
2758 [0xd8] = MMX_OP2(psubusb),
2759 [0xd9] = MMX_OP2(psubusw),
2760 [0xda] = MMX_OP2(pminub),
2761 [0xdb] = MMX_OP2(pand),
2762 [0xdc] = MMX_OP2(paddusb),
2763 [0xdd] = MMX_OP2(paddusw),
2764 [0xde] = MMX_OP2(pmaxub),
2765 [0xdf] = MMX_OP2(pandn),
2766 [0xe0] = MMX_OP2(pavgb),
2767 [0xe1] = MMX_OP2(psraw),
2768 [0xe2] = MMX_OP2(psrad),
2769 [0xe3] = MMX_OP2(pavgw),
2770 [0xe4] = MMX_OP2(pmulhuw),
2771 [0xe5] = MMX_OP2(pmulhw),
2772 [0xe6] = { NULL, gen_helper_cvttpd2dq, gen_helper_cvtdq2pd, gen_helper_cvtpd2dq },
2773 [0xe7] = { SSE_SPECIAL , SSE_SPECIAL }, /* movntq, movntq */
2774 [0xe8] = MMX_OP2(psubsb),
2775 [0xe9] = MMX_OP2(psubsw),
2776 [0xea] = MMX_OP2(pminsw),
2777 [0xeb] = MMX_OP2(por),
2778 [0xec] = MMX_OP2(paddsb),
2779 [0xed] = MMX_OP2(paddsw),
2780 [0xee] = MMX_OP2(pmaxsw),
2781 [0xef] = MMX_OP2(pxor),
2782 [0xf0] = { NULL, NULL, NULL, SSE_SPECIAL }, /* lddqu */
2783 [0xf1] = MMX_OP2(psllw),
2784 [0xf2] = MMX_OP2(pslld),
2785 [0xf3] = MMX_OP2(psllq),
2786 [0xf4] = MMX_OP2(pmuludq),
2787 [0xf5] = MMX_OP2(pmaddwd),
2788 [0xf6] = MMX_OP2(psadbw),
2789 [0xf7] = { (SSEFunc_0_epp)gen_helper_maskmov_mmx,
2790 (SSEFunc_0_epp)gen_helper_maskmov_xmm }, /* XXX: casts */
2791 [0xf8] = MMX_OP2(psubb),
2792 [0xf9] = MMX_OP2(psubw),
2793 [0xfa] = MMX_OP2(psubl),
2794 [0xfb] = MMX_OP2(psubq),
2795 [0xfc] = MMX_OP2(paddb),
2796 [0xfd] = MMX_OP2(paddw),
2797 [0xfe] = MMX_OP2(paddl),
2800 static const SSEFunc_0_epp sse_op_table2[3 * 8][2] = {
2801 [0 + 2] = MMX_OP2(psrlw),
2802 [0 + 4] = MMX_OP2(psraw),
2803 [0 + 6] = MMX_OP2(psllw),
2804 [8 + 2] = MMX_OP2(psrld),
2805 [8 + 4] = MMX_OP2(psrad),
2806 [8 + 6] = MMX_OP2(pslld),
2807 [16 + 2] = MMX_OP2(psrlq),
2808 [16 + 3] = { NULL, gen_helper_psrldq_xmm },
2809 [16 + 6] = MMX_OP2(psllq),
2810 [16 + 7] = { NULL, gen_helper_pslldq_xmm },
2813 static const SSEFunc_0_epi sse_op_table3ai[] = {
2814 gen_helper_cvtsi2ss,
2815 gen_helper_cvtsi2sd
2818 #ifdef TARGET_X86_64
2819 static const SSEFunc_0_epl sse_op_table3aq[] = {
2820 gen_helper_cvtsq2ss,
2821 gen_helper_cvtsq2sd
2823 #endif
2825 static const SSEFunc_i_ep sse_op_table3bi[] = {
2826 gen_helper_cvttss2si,
2827 gen_helper_cvtss2si,
2828 gen_helper_cvttsd2si,
2829 gen_helper_cvtsd2si
2832 #ifdef TARGET_X86_64
2833 static const SSEFunc_l_ep sse_op_table3bq[] = {
2834 gen_helper_cvttss2sq,
2835 gen_helper_cvtss2sq,
2836 gen_helper_cvttsd2sq,
2837 gen_helper_cvtsd2sq
2839 #endif
2841 static const SSEFunc_0_epp sse_op_table4[8][4] = {
2842 SSE_FOP(cmpeq),
2843 SSE_FOP(cmplt),
2844 SSE_FOP(cmple),
2845 SSE_FOP(cmpunord),
2846 SSE_FOP(cmpneq),
2847 SSE_FOP(cmpnlt),
2848 SSE_FOP(cmpnle),
2849 SSE_FOP(cmpord),
2852 static const SSEFunc_0_epp sse_op_table5[256] = {
2853 [0x0c] = gen_helper_pi2fw,
2854 [0x0d] = gen_helper_pi2fd,
2855 [0x1c] = gen_helper_pf2iw,
2856 [0x1d] = gen_helper_pf2id,
2857 [0x8a] = gen_helper_pfnacc,
2858 [0x8e] = gen_helper_pfpnacc,
2859 [0x90] = gen_helper_pfcmpge,
2860 [0x94] = gen_helper_pfmin,
2861 [0x96] = gen_helper_pfrcp,
2862 [0x97] = gen_helper_pfrsqrt,
2863 [0x9a] = gen_helper_pfsub,
2864 [0x9e] = gen_helper_pfadd,
2865 [0xa0] = gen_helper_pfcmpgt,
2866 [0xa4] = gen_helper_pfmax,
2867 [0xa6] = gen_helper_movq, /* pfrcpit1; no need to actually increase precision */
2868 [0xa7] = gen_helper_movq, /* pfrsqit1 */
2869 [0xaa] = gen_helper_pfsubr,
2870 [0xae] = gen_helper_pfacc,
2871 [0xb0] = gen_helper_pfcmpeq,
2872 [0xb4] = gen_helper_pfmul,
2873 [0xb6] = gen_helper_movq, /* pfrcpit2 */
2874 [0xb7] = gen_helper_pmulhrw_mmx,
2875 [0xbb] = gen_helper_pswapd,
2876 [0xbf] = gen_helper_pavgb_mmx /* pavgusb */
2879 struct SSEOpHelper_epp {
2880 SSEFunc_0_epp op[2];
2881 uint32_t ext_mask;
2884 struct SSEOpHelper_eppi {
2885 SSEFunc_0_eppi op[2];
2886 uint32_t ext_mask;
2889 #define SSSE3_OP(x) { MMX_OP2(x), CPUID_EXT_SSSE3 }
2890 #define SSE41_OP(x) { { NULL, gen_helper_ ## x ## _xmm }, CPUID_EXT_SSE41 }
2891 #define SSE42_OP(x) { { NULL, gen_helper_ ## x ## _xmm }, CPUID_EXT_SSE42 }
2892 #define SSE41_SPECIAL { { NULL, SSE_SPECIAL }, CPUID_EXT_SSE41 }
2893 #define PCLMULQDQ_OP(x) { { NULL, gen_helper_ ## x ## _xmm }, \
2894 CPUID_EXT_PCLMULQDQ }
2895 #define AESNI_OP(x) { { NULL, gen_helper_ ## x ## _xmm }, CPUID_EXT_AES }
2897 static const struct SSEOpHelper_epp sse_op_table6[256] = {
2898 [0x00] = SSSE3_OP(pshufb),
2899 [0x01] = SSSE3_OP(phaddw),
2900 [0x02] = SSSE3_OP(phaddd),
2901 [0x03] = SSSE3_OP(phaddsw),
2902 [0x04] = SSSE3_OP(pmaddubsw),
2903 [0x05] = SSSE3_OP(phsubw),
2904 [0x06] = SSSE3_OP(phsubd),
2905 [0x07] = SSSE3_OP(phsubsw),
2906 [0x08] = SSSE3_OP(psignb),
2907 [0x09] = SSSE3_OP(psignw),
2908 [0x0a] = SSSE3_OP(psignd),
2909 [0x0b] = SSSE3_OP(pmulhrsw),
2910 [0x10] = SSE41_OP(pblendvb),
2911 [0x14] = SSE41_OP(blendvps),
2912 [0x15] = SSE41_OP(blendvpd),
2913 [0x17] = SSE41_OP(ptest),
2914 [0x1c] = SSSE3_OP(pabsb),
2915 [0x1d] = SSSE3_OP(pabsw),
2916 [0x1e] = SSSE3_OP(pabsd),
2917 [0x20] = SSE41_OP(pmovsxbw),
2918 [0x21] = SSE41_OP(pmovsxbd),
2919 [0x22] = SSE41_OP(pmovsxbq),
2920 [0x23] = SSE41_OP(pmovsxwd),
2921 [0x24] = SSE41_OP(pmovsxwq),
2922 [0x25] = SSE41_OP(pmovsxdq),
2923 [0x28] = SSE41_OP(pmuldq),
2924 [0x29] = SSE41_OP(pcmpeqq),
2925 [0x2a] = SSE41_SPECIAL, /* movntqda */
2926 [0x2b] = SSE41_OP(packusdw),
2927 [0x30] = SSE41_OP(pmovzxbw),
2928 [0x31] = SSE41_OP(pmovzxbd),
2929 [0x32] = SSE41_OP(pmovzxbq),
2930 [0x33] = SSE41_OP(pmovzxwd),
2931 [0x34] = SSE41_OP(pmovzxwq),
2932 [0x35] = SSE41_OP(pmovzxdq),
2933 [0x37] = SSE42_OP(pcmpgtq),
2934 [0x38] = SSE41_OP(pminsb),
2935 [0x39] = SSE41_OP(pminsd),
2936 [0x3a] = SSE41_OP(pminuw),
2937 [0x3b] = SSE41_OP(pminud),
2938 [0x3c] = SSE41_OP(pmaxsb),
2939 [0x3d] = SSE41_OP(pmaxsd),
2940 [0x3e] = SSE41_OP(pmaxuw),
2941 [0x3f] = SSE41_OP(pmaxud),
2942 [0x40] = SSE41_OP(pmulld),
2943 [0x41] = SSE41_OP(phminposuw),
2944 [0xdb] = AESNI_OP(aesimc),
2945 [0xdc] = AESNI_OP(aesenc),
2946 [0xdd] = AESNI_OP(aesenclast),
2947 [0xde] = AESNI_OP(aesdec),
2948 [0xdf] = AESNI_OP(aesdeclast),
2951 static const struct SSEOpHelper_eppi sse_op_table7[256] = {
2952 [0x08] = SSE41_OP(roundps),
2953 [0x09] = SSE41_OP(roundpd),
2954 [0x0a] = SSE41_OP(roundss),
2955 [0x0b] = SSE41_OP(roundsd),
2956 [0x0c] = SSE41_OP(blendps),
2957 [0x0d] = SSE41_OP(blendpd),
2958 [0x0e] = SSE41_OP(pblendw),
2959 [0x0f] = SSSE3_OP(palignr),
2960 [0x14] = SSE41_SPECIAL, /* pextrb */
2961 [0x15] = SSE41_SPECIAL, /* pextrw */
2962 [0x16] = SSE41_SPECIAL, /* pextrd/pextrq */
2963 [0x17] = SSE41_SPECIAL, /* extractps */
2964 [0x20] = SSE41_SPECIAL, /* pinsrb */
2965 [0x21] = SSE41_SPECIAL, /* insertps */
2966 [0x22] = SSE41_SPECIAL, /* pinsrd/pinsrq */
2967 [0x40] = SSE41_OP(dpps),
2968 [0x41] = SSE41_OP(dppd),
2969 [0x42] = SSE41_OP(mpsadbw),
2970 [0x44] = PCLMULQDQ_OP(pclmulqdq),
2971 [0x60] = SSE42_OP(pcmpestrm),
2972 [0x61] = SSE42_OP(pcmpestri),
2973 [0x62] = SSE42_OP(pcmpistrm),
2974 [0x63] = SSE42_OP(pcmpistri),
2975 [0xdf] = AESNI_OP(aeskeygenassist),
2978 static void gen_sse(CPUX86State *env, DisasContext *s, int b,
2979 target_ulong pc_start, int rex_r)
2981 int b1, op1_offset, op2_offset, is_xmm, val;
2982 int modrm, mod, rm, reg;
2983 SSEFunc_0_epp sse_fn_epp;
2984 SSEFunc_0_eppi sse_fn_eppi;
2985 SSEFunc_0_ppi sse_fn_ppi;
2986 SSEFunc_0_eppt sse_fn_eppt;
2987 TCGMemOp ot;
2989 b &= 0xff;
2990 if (s->prefix & PREFIX_DATA)
2991 b1 = 1;
2992 else if (s->prefix & PREFIX_REPZ)
2993 b1 = 2;
2994 else if (s->prefix & PREFIX_REPNZ)
2995 b1 = 3;
2996 else
2997 b1 = 0;
2998 sse_fn_epp = sse_op_table1[b][b1];
2999 if (!sse_fn_epp) {
3000 goto illegal_op;
3002 if ((b <= 0x5f && b >= 0x10) || b == 0xc6 || b == 0xc2) {
3003 is_xmm = 1;
3004 } else {
3005 if (b1 == 0) {
3006 /* MMX case */
3007 is_xmm = 0;
3008 } else {
3009 is_xmm = 1;
3012 /* simple MMX/SSE operation */
3013 if (s->flags & HF_TS_MASK) {
3014 gen_exception(s, EXCP07_PREX, pc_start - s->cs_base);
3015 return;
3017 if (s->flags & HF_EM_MASK) {
3018 illegal_op:
3019 gen_exception(s, EXCP06_ILLOP, pc_start - s->cs_base);
3020 return;
3022 if (is_xmm && !(s->flags & HF_OSFXSR_MASK))
3023 if ((b != 0x38 && b != 0x3a) || (s->prefix & PREFIX_DATA))
3024 goto illegal_op;
3025 if (b == 0x0e) {
3026 if (!(s->cpuid_ext2_features & CPUID_EXT2_3DNOW))
3027 goto illegal_op;
3028 /* femms */
3029 gen_helper_emms(cpu_env);
3030 return;
3032 if (b == 0x77) {
3033 /* emms */
3034 gen_helper_emms(cpu_env);
3035 return;
3037 /* prepare MMX state (XXX: optimize by storing fptt and fptags in
3038 the static cpu state) */
3039 if (!is_xmm) {
3040 gen_helper_enter_mmx(cpu_env);
3043 modrm = cpu_ldub_code(env, s->pc++);
3044 reg = ((modrm >> 3) & 7);
3045 if (is_xmm)
3046 reg |= rex_r;
3047 mod = (modrm >> 6) & 3;
3048 if (sse_fn_epp == SSE_SPECIAL) {
3049 b |= (b1 << 8);
3050 switch(b) {
3051 case 0x0e7: /* movntq */
3052 if (mod == 3)
3053 goto illegal_op;
3054 gen_lea_modrm(env, s, modrm);
3055 gen_stq_env_A0(s, offsetof(CPUX86State, fpregs[reg].mmx));
3056 break;
3057 case 0x1e7: /* movntdq */
3058 case 0x02b: /* movntps */
3059 case 0x12b: /* movntps */
3060 if (mod == 3)
3061 goto illegal_op;
3062 gen_lea_modrm(env, s, modrm);
3063 gen_sto_env_A0(s, offsetof(CPUX86State, xmm_regs[reg]));
3064 break;
3065 case 0x3f0: /* lddqu */
3066 if (mod == 3)
3067 goto illegal_op;
3068 gen_lea_modrm(env, s, modrm);
3069 gen_ldo_env_A0(s, offsetof(CPUX86State, xmm_regs[reg]));
3070 break;
3071 case 0x22b: /* movntss */
3072 case 0x32b: /* movntsd */
3073 if (mod == 3)
3074 goto illegal_op;
3075 gen_lea_modrm(env, s, modrm);
3076 if (b1 & 1) {
3077 gen_stq_env_A0(s, offsetof(CPUX86State, xmm_regs[reg]));
3078 } else {
3079 tcg_gen_ld32u_tl(cpu_T[0], cpu_env, offsetof(CPUX86State,
3080 xmm_regs[reg].XMM_L(0)));
3081 gen_op_st_v(s, MO_32, cpu_T[0], cpu_A0);
3083 break;
3084 case 0x6e: /* movd mm, ea */
3085 #ifdef TARGET_X86_64
3086 if (s->dflag == MO_64) {
3087 gen_ldst_modrm(env, s, modrm, MO_64, OR_TMP0, 0);
3088 tcg_gen_st_tl(cpu_T[0], cpu_env, offsetof(CPUX86State,fpregs[reg].mmx));
3089 } else
3090 #endif
3092 gen_ldst_modrm(env, s, modrm, MO_32, OR_TMP0, 0);
3093 tcg_gen_addi_ptr(cpu_ptr0, cpu_env,
3094 offsetof(CPUX86State,fpregs[reg].mmx));
3095 tcg_gen_trunc_tl_i32(cpu_tmp2_i32, cpu_T[0]);
3096 gen_helper_movl_mm_T0_mmx(cpu_ptr0, cpu_tmp2_i32);
3098 break;
3099 case 0x16e: /* movd xmm, ea */
3100 #ifdef TARGET_X86_64
3101 if (s->dflag == MO_64) {
3102 gen_ldst_modrm(env, s, modrm, MO_64, OR_TMP0, 0);
3103 tcg_gen_addi_ptr(cpu_ptr0, cpu_env,
3104 offsetof(CPUX86State,xmm_regs[reg]));
3105 gen_helper_movq_mm_T0_xmm(cpu_ptr0, cpu_T[0]);
3106 } else
3107 #endif
3109 gen_ldst_modrm(env, s, modrm, MO_32, OR_TMP0, 0);
3110 tcg_gen_addi_ptr(cpu_ptr0, cpu_env,
3111 offsetof(CPUX86State,xmm_regs[reg]));
3112 tcg_gen_trunc_tl_i32(cpu_tmp2_i32, cpu_T[0]);
3113 gen_helper_movl_mm_T0_xmm(cpu_ptr0, cpu_tmp2_i32);
3115 break;
3116 case 0x6f: /* movq mm, ea */
3117 if (mod != 3) {
3118 gen_lea_modrm(env, s, modrm);
3119 gen_ldq_env_A0(s, offsetof(CPUX86State, fpregs[reg].mmx));
3120 } else {
3121 rm = (modrm & 7);
3122 tcg_gen_ld_i64(cpu_tmp1_i64, cpu_env,
3123 offsetof(CPUX86State,fpregs[rm].mmx));
3124 tcg_gen_st_i64(cpu_tmp1_i64, cpu_env,
3125 offsetof(CPUX86State,fpregs[reg].mmx));
3127 break;
3128 case 0x010: /* movups */
3129 case 0x110: /* movupd */
3130 case 0x028: /* movaps */
3131 case 0x128: /* movapd */
3132 case 0x16f: /* movdqa xmm, ea */
3133 case 0x26f: /* movdqu xmm, ea */
3134 if (mod != 3) {
3135 gen_lea_modrm(env, s, modrm);
3136 gen_ldo_env_A0(s, offsetof(CPUX86State, xmm_regs[reg]));
3137 } else {
3138 rm = (modrm & 7) | REX_B(s);
3139 gen_op_movo(offsetof(CPUX86State,xmm_regs[reg]),
3140 offsetof(CPUX86State,xmm_regs[rm]));
3142 break;
3143 case 0x210: /* movss xmm, ea */
3144 if (mod != 3) {
3145 gen_lea_modrm(env, s, modrm);
3146 gen_op_ld_v(s, MO_32, cpu_T[0], cpu_A0);
3147 tcg_gen_st32_tl(cpu_T[0], cpu_env, offsetof(CPUX86State,xmm_regs[reg].XMM_L(0)));
3148 tcg_gen_movi_tl(cpu_T[0], 0);
3149 tcg_gen_st32_tl(cpu_T[0], cpu_env, offsetof(CPUX86State,xmm_regs[reg].XMM_L(1)));
3150 tcg_gen_st32_tl(cpu_T[0], cpu_env, offsetof(CPUX86State,xmm_regs[reg].XMM_L(2)));
3151 tcg_gen_st32_tl(cpu_T[0], cpu_env, offsetof(CPUX86State,xmm_regs[reg].XMM_L(3)));
3152 } else {
3153 rm = (modrm & 7) | REX_B(s);
3154 gen_op_movl(offsetof(CPUX86State,xmm_regs[reg].XMM_L(0)),
3155 offsetof(CPUX86State,xmm_regs[rm].XMM_L(0)));
3157 break;
3158 case 0x310: /* movsd xmm, ea */
3159 if (mod != 3) {
3160 gen_lea_modrm(env, s, modrm);
3161 gen_ldq_env_A0(s, offsetof(CPUX86State,
3162 xmm_regs[reg].XMM_Q(0)));
3163 tcg_gen_movi_tl(cpu_T[0], 0);
3164 tcg_gen_st32_tl(cpu_T[0], cpu_env, offsetof(CPUX86State,xmm_regs[reg].XMM_L(2)));
3165 tcg_gen_st32_tl(cpu_T[0], cpu_env, offsetof(CPUX86State,xmm_regs[reg].XMM_L(3)));
3166 } else {
3167 rm = (modrm & 7) | REX_B(s);
3168 gen_op_movq(offsetof(CPUX86State,xmm_regs[reg].XMM_Q(0)),
3169 offsetof(CPUX86State,xmm_regs[rm].XMM_Q(0)));
3171 break;
3172 case 0x012: /* movlps */
3173 case 0x112: /* movlpd */
3174 if (mod != 3) {
3175 gen_lea_modrm(env, s, modrm);
3176 gen_ldq_env_A0(s, offsetof(CPUX86State,
3177 xmm_regs[reg].XMM_Q(0)));
3178 } else {
3179 /* movhlps */
3180 rm = (modrm & 7) | REX_B(s);
3181 gen_op_movq(offsetof(CPUX86State,xmm_regs[reg].XMM_Q(0)),
3182 offsetof(CPUX86State,xmm_regs[rm].XMM_Q(1)));
3184 break;
3185 case 0x212: /* movsldup */
3186 if (mod != 3) {
3187 gen_lea_modrm(env, s, modrm);
3188 gen_ldo_env_A0(s, offsetof(CPUX86State, xmm_regs[reg]));
3189 } else {
3190 rm = (modrm & 7) | REX_B(s);
3191 gen_op_movl(offsetof(CPUX86State,xmm_regs[reg].XMM_L(0)),
3192 offsetof(CPUX86State,xmm_regs[rm].XMM_L(0)));
3193 gen_op_movl(offsetof(CPUX86State,xmm_regs[reg].XMM_L(2)),
3194 offsetof(CPUX86State,xmm_regs[rm].XMM_L(2)));
3196 gen_op_movl(offsetof(CPUX86State,xmm_regs[reg].XMM_L(1)),
3197 offsetof(CPUX86State,xmm_regs[reg].XMM_L(0)));
3198 gen_op_movl(offsetof(CPUX86State,xmm_regs[reg].XMM_L(3)),
3199 offsetof(CPUX86State,xmm_regs[reg].XMM_L(2)));
3200 break;
3201 case 0x312: /* movddup */
3202 if (mod != 3) {
3203 gen_lea_modrm(env, s, modrm);
3204 gen_ldq_env_A0(s, offsetof(CPUX86State,
3205 xmm_regs[reg].XMM_Q(0)));
3206 } else {
3207 rm = (modrm & 7) | REX_B(s);
3208 gen_op_movq(offsetof(CPUX86State,xmm_regs[reg].XMM_Q(0)),
3209 offsetof(CPUX86State,xmm_regs[rm].XMM_Q(0)));
3211 gen_op_movq(offsetof(CPUX86State,xmm_regs[reg].XMM_Q(1)),
3212 offsetof(CPUX86State,xmm_regs[reg].XMM_Q(0)));
3213 break;
3214 case 0x016: /* movhps */
3215 case 0x116: /* movhpd */
3216 if (mod != 3) {
3217 gen_lea_modrm(env, s, modrm);
3218 gen_ldq_env_A0(s, offsetof(CPUX86State,
3219 xmm_regs[reg].XMM_Q(1)));
3220 } else {
3221 /* movlhps */
3222 rm = (modrm & 7) | REX_B(s);
3223 gen_op_movq(offsetof(CPUX86State,xmm_regs[reg].XMM_Q(1)),
3224 offsetof(CPUX86State,xmm_regs[rm].XMM_Q(0)));
3226 break;
3227 case 0x216: /* movshdup */
3228 if (mod != 3) {
3229 gen_lea_modrm(env, s, modrm);
3230 gen_ldo_env_A0(s, offsetof(CPUX86State, xmm_regs[reg]));
3231 } else {
3232 rm = (modrm & 7) | REX_B(s);
3233 gen_op_movl(offsetof(CPUX86State,xmm_regs[reg].XMM_L(1)),
3234 offsetof(CPUX86State,xmm_regs[rm].XMM_L(1)));
3235 gen_op_movl(offsetof(CPUX86State,xmm_regs[reg].XMM_L(3)),
3236 offsetof(CPUX86State,xmm_regs[rm].XMM_L(3)));
3238 gen_op_movl(offsetof(CPUX86State,xmm_regs[reg].XMM_L(0)),
3239 offsetof(CPUX86State,xmm_regs[reg].XMM_L(1)));
3240 gen_op_movl(offsetof(CPUX86State,xmm_regs[reg].XMM_L(2)),
3241 offsetof(CPUX86State,xmm_regs[reg].XMM_L(3)));
3242 break;
3243 case 0x178:
3244 case 0x378:
3246 int bit_index, field_length;
3248 if (b1 == 1 && reg != 0)
3249 goto illegal_op;
3250 field_length = cpu_ldub_code(env, s->pc++) & 0x3F;
3251 bit_index = cpu_ldub_code(env, s->pc++) & 0x3F;
3252 tcg_gen_addi_ptr(cpu_ptr0, cpu_env,
3253 offsetof(CPUX86State,xmm_regs[reg]));
3254 if (b1 == 1)
3255 gen_helper_extrq_i(cpu_env, cpu_ptr0,
3256 tcg_const_i32(bit_index),
3257 tcg_const_i32(field_length));
3258 else
3259 gen_helper_insertq_i(cpu_env, cpu_ptr0,
3260 tcg_const_i32(bit_index),
3261 tcg_const_i32(field_length));
3263 break;
3264 case 0x7e: /* movd ea, mm */
3265 #ifdef TARGET_X86_64
3266 if (s->dflag == MO_64) {
3267 tcg_gen_ld_i64(cpu_T[0], cpu_env,
3268 offsetof(CPUX86State,fpregs[reg].mmx));
3269 gen_ldst_modrm(env, s, modrm, MO_64, OR_TMP0, 1);
3270 } else
3271 #endif
3273 tcg_gen_ld32u_tl(cpu_T[0], cpu_env,
3274 offsetof(CPUX86State,fpregs[reg].mmx.MMX_L(0)));
3275 gen_ldst_modrm(env, s, modrm, MO_32, OR_TMP0, 1);
3277 break;
3278 case 0x17e: /* movd ea, xmm */
3279 #ifdef TARGET_X86_64
3280 if (s->dflag == MO_64) {
3281 tcg_gen_ld_i64(cpu_T[0], cpu_env,
3282 offsetof(CPUX86State,xmm_regs[reg].XMM_Q(0)));
3283 gen_ldst_modrm(env, s, modrm, MO_64, OR_TMP0, 1);
3284 } else
3285 #endif
3287 tcg_gen_ld32u_tl(cpu_T[0], cpu_env,
3288 offsetof(CPUX86State,xmm_regs[reg].XMM_L(0)));
3289 gen_ldst_modrm(env, s, modrm, MO_32, OR_TMP0, 1);
3291 break;
3292 case 0x27e: /* movq xmm, ea */
3293 if (mod != 3) {
3294 gen_lea_modrm(env, s, modrm);
3295 gen_ldq_env_A0(s, offsetof(CPUX86State,
3296 xmm_regs[reg].XMM_Q(0)));
3297 } else {
3298 rm = (modrm & 7) | REX_B(s);
3299 gen_op_movq(offsetof(CPUX86State,xmm_regs[reg].XMM_Q(0)),
3300 offsetof(CPUX86State,xmm_regs[rm].XMM_Q(0)));
3302 gen_op_movq_env_0(offsetof(CPUX86State,xmm_regs[reg].XMM_Q(1)));
3303 break;
3304 case 0x7f: /* movq ea, mm */
3305 if (mod != 3) {
3306 gen_lea_modrm(env, s, modrm);
3307 gen_stq_env_A0(s, offsetof(CPUX86State, fpregs[reg].mmx));
3308 } else {
3309 rm = (modrm & 7);
3310 gen_op_movq(offsetof(CPUX86State,fpregs[rm].mmx),
3311 offsetof(CPUX86State,fpregs[reg].mmx));
3313 break;
3314 case 0x011: /* movups */
3315 case 0x111: /* movupd */
3316 case 0x029: /* movaps */
3317 case 0x129: /* movapd */
3318 case 0x17f: /* movdqa ea, xmm */
3319 case 0x27f: /* movdqu ea, xmm */
3320 if (mod != 3) {
3321 gen_lea_modrm(env, s, modrm);
3322 gen_sto_env_A0(s, offsetof(CPUX86State, xmm_regs[reg]));
3323 } else {
3324 rm = (modrm & 7) | REX_B(s);
3325 gen_op_movo(offsetof(CPUX86State,xmm_regs[rm]),
3326 offsetof(CPUX86State,xmm_regs[reg]));
3328 break;
3329 case 0x211: /* movss ea, xmm */
3330 if (mod != 3) {
3331 gen_lea_modrm(env, s, modrm);
3332 tcg_gen_ld32u_tl(cpu_T[0], cpu_env, offsetof(CPUX86State,xmm_regs[reg].XMM_L(0)));
3333 gen_op_st_v(s, MO_32, cpu_T[0], cpu_A0);
3334 } else {
3335 rm = (modrm & 7) | REX_B(s);
3336 gen_op_movl(offsetof(CPUX86State,xmm_regs[rm].XMM_L(0)),
3337 offsetof(CPUX86State,xmm_regs[reg].XMM_L(0)));
3339 break;
3340 case 0x311: /* movsd ea, xmm */
3341 if (mod != 3) {
3342 gen_lea_modrm(env, s, modrm);
3343 gen_stq_env_A0(s, offsetof(CPUX86State,
3344 xmm_regs[reg].XMM_Q(0)));
3345 } else {
3346 rm = (modrm & 7) | REX_B(s);
3347 gen_op_movq(offsetof(CPUX86State,xmm_regs[rm].XMM_Q(0)),
3348 offsetof(CPUX86State,xmm_regs[reg].XMM_Q(0)));
3350 break;
3351 case 0x013: /* movlps */
3352 case 0x113: /* movlpd */
3353 if (mod != 3) {
3354 gen_lea_modrm(env, s, modrm);
3355 gen_stq_env_A0(s, offsetof(CPUX86State,
3356 xmm_regs[reg].XMM_Q(0)));
3357 } else {
3358 goto illegal_op;
3360 break;
3361 case 0x017: /* movhps */
3362 case 0x117: /* movhpd */
3363 if (mod != 3) {
3364 gen_lea_modrm(env, s, modrm);
3365 gen_stq_env_A0(s, offsetof(CPUX86State,
3366 xmm_regs[reg].XMM_Q(1)));
3367 } else {
3368 goto illegal_op;
3370 break;
3371 case 0x71: /* shift mm, im */
3372 case 0x72:
3373 case 0x73:
3374 case 0x171: /* shift xmm, im */
3375 case 0x172:
3376 case 0x173:
3377 if (b1 >= 2) {
3378 goto illegal_op;
3380 val = cpu_ldub_code(env, s->pc++);
3381 if (is_xmm) {
3382 tcg_gen_movi_tl(cpu_T[0], val);
3383 tcg_gen_st32_tl(cpu_T[0], cpu_env, offsetof(CPUX86State,xmm_t0.XMM_L(0)));
3384 tcg_gen_movi_tl(cpu_T[0], 0);
3385 tcg_gen_st32_tl(cpu_T[0], cpu_env, offsetof(CPUX86State,xmm_t0.XMM_L(1)));
3386 op1_offset = offsetof(CPUX86State,xmm_t0);
3387 } else {
3388 tcg_gen_movi_tl(cpu_T[0], val);
3389 tcg_gen_st32_tl(cpu_T[0], cpu_env, offsetof(CPUX86State,mmx_t0.MMX_L(0)));
3390 tcg_gen_movi_tl(cpu_T[0], 0);
3391 tcg_gen_st32_tl(cpu_T[0], cpu_env, offsetof(CPUX86State,mmx_t0.MMX_L(1)));
3392 op1_offset = offsetof(CPUX86State,mmx_t0);
3394 sse_fn_epp = sse_op_table2[((b - 1) & 3) * 8 +
3395 (((modrm >> 3)) & 7)][b1];
3396 if (!sse_fn_epp) {
3397 goto illegal_op;
3399 if (is_xmm) {
3400 rm = (modrm & 7) | REX_B(s);
3401 op2_offset = offsetof(CPUX86State,xmm_regs[rm]);
3402 } else {
3403 rm = (modrm & 7);
3404 op2_offset = offsetof(CPUX86State,fpregs[rm].mmx);
3406 tcg_gen_addi_ptr(cpu_ptr0, cpu_env, op2_offset);
3407 tcg_gen_addi_ptr(cpu_ptr1, cpu_env, op1_offset);
3408 sse_fn_epp(cpu_env, cpu_ptr0, cpu_ptr1);
3409 break;
3410 case 0x050: /* movmskps */
3411 rm = (modrm & 7) | REX_B(s);
3412 tcg_gen_addi_ptr(cpu_ptr0, cpu_env,
3413 offsetof(CPUX86State,xmm_regs[rm]));
3414 gen_helper_movmskps(cpu_tmp2_i32, cpu_env, cpu_ptr0);
3415 tcg_gen_extu_i32_tl(cpu_regs[reg], cpu_tmp2_i32);
3416 break;
3417 case 0x150: /* movmskpd */
3418 rm = (modrm & 7) | REX_B(s);
3419 tcg_gen_addi_ptr(cpu_ptr0, cpu_env,
3420 offsetof(CPUX86State,xmm_regs[rm]));
3421 gen_helper_movmskpd(cpu_tmp2_i32, cpu_env, cpu_ptr0);
3422 tcg_gen_extu_i32_tl(cpu_regs[reg], cpu_tmp2_i32);
3423 break;
3424 case 0x02a: /* cvtpi2ps */
3425 case 0x12a: /* cvtpi2pd */
3426 gen_helper_enter_mmx(cpu_env);
3427 if (mod != 3) {
3428 gen_lea_modrm(env, s, modrm);
3429 op2_offset = offsetof(CPUX86State,mmx_t0);
3430 gen_ldq_env_A0(s, op2_offset);
3431 } else {
3432 rm = (modrm & 7);
3433 op2_offset = offsetof(CPUX86State,fpregs[rm].mmx);
3435 op1_offset = offsetof(CPUX86State,xmm_regs[reg]);
3436 tcg_gen_addi_ptr(cpu_ptr0, cpu_env, op1_offset);
3437 tcg_gen_addi_ptr(cpu_ptr1, cpu_env, op2_offset);
3438 switch(b >> 8) {
3439 case 0x0:
3440 gen_helper_cvtpi2ps(cpu_env, cpu_ptr0, cpu_ptr1);
3441 break;
3442 default:
3443 case 0x1:
3444 gen_helper_cvtpi2pd(cpu_env, cpu_ptr0, cpu_ptr1);
3445 break;
3447 break;
3448 case 0x22a: /* cvtsi2ss */
3449 case 0x32a: /* cvtsi2sd */
3450 ot = mo_64_32(s->dflag);
3451 gen_ldst_modrm(env, s, modrm, ot, OR_TMP0, 0);
3452 op1_offset = offsetof(CPUX86State,xmm_regs[reg]);
3453 tcg_gen_addi_ptr(cpu_ptr0, cpu_env, op1_offset);
3454 if (ot == MO_32) {
3455 SSEFunc_0_epi sse_fn_epi = sse_op_table3ai[(b >> 8) & 1];
3456 tcg_gen_trunc_tl_i32(cpu_tmp2_i32, cpu_T[0]);
3457 sse_fn_epi(cpu_env, cpu_ptr0, cpu_tmp2_i32);
3458 } else {
3459 #ifdef TARGET_X86_64
3460 SSEFunc_0_epl sse_fn_epl = sse_op_table3aq[(b >> 8) & 1];
3461 sse_fn_epl(cpu_env, cpu_ptr0, cpu_T[0]);
3462 #else
3463 goto illegal_op;
3464 #endif
3466 break;
3467 case 0x02c: /* cvttps2pi */
3468 case 0x12c: /* cvttpd2pi */
3469 case 0x02d: /* cvtps2pi */
3470 case 0x12d: /* cvtpd2pi */
3471 gen_helper_enter_mmx(cpu_env);
3472 if (mod != 3) {
3473 gen_lea_modrm(env, s, modrm);
3474 op2_offset = offsetof(CPUX86State,xmm_t0);
3475 gen_ldo_env_A0(s, op2_offset);
3476 } else {
3477 rm = (modrm & 7) | REX_B(s);
3478 op2_offset = offsetof(CPUX86State,xmm_regs[rm]);
3480 op1_offset = offsetof(CPUX86State,fpregs[reg & 7].mmx);
3481 tcg_gen_addi_ptr(cpu_ptr0, cpu_env, op1_offset);
3482 tcg_gen_addi_ptr(cpu_ptr1, cpu_env, op2_offset);
3483 switch(b) {
3484 case 0x02c:
3485 gen_helper_cvttps2pi(cpu_env, cpu_ptr0, cpu_ptr1);
3486 break;
3487 case 0x12c:
3488 gen_helper_cvttpd2pi(cpu_env, cpu_ptr0, cpu_ptr1);
3489 break;
3490 case 0x02d:
3491 gen_helper_cvtps2pi(cpu_env, cpu_ptr0, cpu_ptr1);
3492 break;
3493 case 0x12d:
3494 gen_helper_cvtpd2pi(cpu_env, cpu_ptr0, cpu_ptr1);
3495 break;
3497 break;
3498 case 0x22c: /* cvttss2si */
3499 case 0x32c: /* cvttsd2si */
3500 case 0x22d: /* cvtss2si */
3501 case 0x32d: /* cvtsd2si */
3502 ot = mo_64_32(s->dflag);
3503 if (mod != 3) {
3504 gen_lea_modrm(env, s, modrm);
3505 if ((b >> 8) & 1) {
3506 gen_ldq_env_A0(s, offsetof(CPUX86State, xmm_t0.XMM_Q(0)));
3507 } else {
3508 gen_op_ld_v(s, MO_32, cpu_T[0], cpu_A0);
3509 tcg_gen_st32_tl(cpu_T[0], cpu_env, offsetof(CPUX86State,xmm_t0.XMM_L(0)));
3511 op2_offset = offsetof(CPUX86State,xmm_t0);
3512 } else {
3513 rm = (modrm & 7) | REX_B(s);
3514 op2_offset = offsetof(CPUX86State,xmm_regs[rm]);
3516 tcg_gen_addi_ptr(cpu_ptr0, cpu_env, op2_offset);
3517 if (ot == MO_32) {
3518 SSEFunc_i_ep sse_fn_i_ep =
3519 sse_op_table3bi[((b >> 7) & 2) | (b & 1)];
3520 sse_fn_i_ep(cpu_tmp2_i32, cpu_env, cpu_ptr0);
3521 tcg_gen_extu_i32_tl(cpu_T[0], cpu_tmp2_i32);
3522 } else {
3523 #ifdef TARGET_X86_64
3524 SSEFunc_l_ep sse_fn_l_ep =
3525 sse_op_table3bq[((b >> 7) & 2) | (b & 1)];
3526 sse_fn_l_ep(cpu_T[0], cpu_env, cpu_ptr0);
3527 #else
3528 goto illegal_op;
3529 #endif
3531 gen_op_mov_reg_v(ot, reg, cpu_T[0]);
3532 break;
3533 case 0xc4: /* pinsrw */
3534 case 0x1c4:
3535 s->rip_offset = 1;
3536 gen_ldst_modrm(env, s, modrm, MO_16, OR_TMP0, 0);
3537 val = cpu_ldub_code(env, s->pc++);
3538 if (b1) {
3539 val &= 7;
3540 tcg_gen_st16_tl(cpu_T[0], cpu_env,
3541 offsetof(CPUX86State,xmm_regs[reg].XMM_W(val)));
3542 } else {
3543 val &= 3;
3544 tcg_gen_st16_tl(cpu_T[0], cpu_env,
3545 offsetof(CPUX86State,fpregs[reg].mmx.MMX_W(val)));
3547 break;
3548 case 0xc5: /* pextrw */
3549 case 0x1c5:
3550 if (mod != 3)
3551 goto illegal_op;
3552 ot = mo_64_32(s->dflag);
3553 val = cpu_ldub_code(env, s->pc++);
3554 if (b1) {
3555 val &= 7;
3556 rm = (modrm & 7) | REX_B(s);
3557 tcg_gen_ld16u_tl(cpu_T[0], cpu_env,
3558 offsetof(CPUX86State,xmm_regs[rm].XMM_W(val)));
3559 } else {
3560 val &= 3;
3561 rm = (modrm & 7);
3562 tcg_gen_ld16u_tl(cpu_T[0], cpu_env,
3563 offsetof(CPUX86State,fpregs[rm].mmx.MMX_W(val)));
3565 reg = ((modrm >> 3) & 7) | rex_r;
3566 gen_op_mov_reg_v(ot, reg, cpu_T[0]);
3567 break;
3568 case 0x1d6: /* movq ea, xmm */
3569 if (mod != 3) {
3570 gen_lea_modrm(env, s, modrm);
3571 gen_stq_env_A0(s, offsetof(CPUX86State,
3572 xmm_regs[reg].XMM_Q(0)));
3573 } else {
3574 rm = (modrm & 7) | REX_B(s);
3575 gen_op_movq(offsetof(CPUX86State,xmm_regs[rm].XMM_Q(0)),
3576 offsetof(CPUX86State,xmm_regs[reg].XMM_Q(0)));
3577 gen_op_movq_env_0(offsetof(CPUX86State,xmm_regs[rm].XMM_Q(1)));
3579 break;
3580 case 0x2d6: /* movq2dq */
3581 gen_helper_enter_mmx(cpu_env);
3582 rm = (modrm & 7);
3583 gen_op_movq(offsetof(CPUX86State,xmm_regs[reg].XMM_Q(0)),
3584 offsetof(CPUX86State,fpregs[rm].mmx));
3585 gen_op_movq_env_0(offsetof(CPUX86State,xmm_regs[reg].XMM_Q(1)));
3586 break;
3587 case 0x3d6: /* movdq2q */
3588 gen_helper_enter_mmx(cpu_env);
3589 rm = (modrm & 7) | REX_B(s);
3590 gen_op_movq(offsetof(CPUX86State,fpregs[reg & 7].mmx),
3591 offsetof(CPUX86State,xmm_regs[rm].XMM_Q(0)));
3592 break;
3593 case 0xd7: /* pmovmskb */
3594 case 0x1d7:
3595 if (mod != 3)
3596 goto illegal_op;
3597 if (b1) {
3598 rm = (modrm & 7) | REX_B(s);
3599 tcg_gen_addi_ptr(cpu_ptr0, cpu_env, offsetof(CPUX86State,xmm_regs[rm]));
3600 gen_helper_pmovmskb_xmm(cpu_tmp2_i32, cpu_env, cpu_ptr0);
3601 } else {
3602 rm = (modrm & 7);
3603 tcg_gen_addi_ptr(cpu_ptr0, cpu_env, offsetof(CPUX86State,fpregs[rm].mmx));
3604 gen_helper_pmovmskb_mmx(cpu_tmp2_i32, cpu_env, cpu_ptr0);
3606 reg = ((modrm >> 3) & 7) | rex_r;
3607 tcg_gen_extu_i32_tl(cpu_regs[reg], cpu_tmp2_i32);
3608 break;
3610 case 0x138:
3611 case 0x038:
3612 b = modrm;
3613 if ((b & 0xf0) == 0xf0) {
3614 goto do_0f_38_fx;
3616 modrm = cpu_ldub_code(env, s->pc++);
3617 rm = modrm & 7;
3618 reg = ((modrm >> 3) & 7) | rex_r;
3619 mod = (modrm >> 6) & 3;
3620 if (b1 >= 2) {
3621 goto illegal_op;
3624 sse_fn_epp = sse_op_table6[b].op[b1];
3625 if (!sse_fn_epp) {
3626 goto illegal_op;
3628 if (!(s->cpuid_ext_features & sse_op_table6[b].ext_mask))
3629 goto illegal_op;
3631 if (b1) {
3632 op1_offset = offsetof(CPUX86State,xmm_regs[reg]);
3633 if (mod == 3) {
3634 op2_offset = offsetof(CPUX86State,xmm_regs[rm | REX_B(s)]);
3635 } else {
3636 op2_offset = offsetof(CPUX86State,xmm_t0);
3637 gen_lea_modrm(env, s, modrm);
3638 switch (b) {
3639 case 0x20: case 0x30: /* pmovsxbw, pmovzxbw */
3640 case 0x23: case 0x33: /* pmovsxwd, pmovzxwd */
3641 case 0x25: case 0x35: /* pmovsxdq, pmovzxdq */
3642 gen_ldq_env_A0(s, op2_offset +
3643 offsetof(XMMReg, XMM_Q(0)));
3644 break;
3645 case 0x21: case 0x31: /* pmovsxbd, pmovzxbd */
3646 case 0x24: case 0x34: /* pmovsxwq, pmovzxwq */
3647 tcg_gen_qemu_ld_i32(cpu_tmp2_i32, cpu_A0,
3648 s->mem_index, MO_LEUL);
3649 tcg_gen_st_i32(cpu_tmp2_i32, cpu_env, op2_offset +
3650 offsetof(XMMReg, XMM_L(0)));
3651 break;
3652 case 0x22: case 0x32: /* pmovsxbq, pmovzxbq */
3653 tcg_gen_qemu_ld_tl(cpu_tmp0, cpu_A0,
3654 s->mem_index, MO_LEUW);
3655 tcg_gen_st16_tl(cpu_tmp0, cpu_env, op2_offset +
3656 offsetof(XMMReg, XMM_W(0)));
3657 break;
3658 case 0x2a: /* movntqda */
3659 gen_ldo_env_A0(s, op1_offset);
3660 return;
3661 default:
3662 gen_ldo_env_A0(s, op2_offset);
3665 } else {
3666 op1_offset = offsetof(CPUX86State,fpregs[reg].mmx);
3667 if (mod == 3) {
3668 op2_offset = offsetof(CPUX86State,fpregs[rm].mmx);
3669 } else {
3670 op2_offset = offsetof(CPUX86State,mmx_t0);
3671 gen_lea_modrm(env, s, modrm);
3672 gen_ldq_env_A0(s, op2_offset);
3675 if (sse_fn_epp == SSE_SPECIAL) {
3676 goto illegal_op;
3679 tcg_gen_addi_ptr(cpu_ptr0, cpu_env, op1_offset);
3680 tcg_gen_addi_ptr(cpu_ptr1, cpu_env, op2_offset);
3681 sse_fn_epp(cpu_env, cpu_ptr0, cpu_ptr1);
3683 if (b == 0x17) {
3684 set_cc_op(s, CC_OP_EFLAGS);
3686 break;
3688 case 0x238:
3689 case 0x338:
3690 do_0f_38_fx:
3691 /* Various integer extensions at 0f 38 f[0-f]. */
3692 b = modrm | (b1 << 8);
3693 modrm = cpu_ldub_code(env, s->pc++);
3694 reg = ((modrm >> 3) & 7) | rex_r;
3696 switch (b) {
3697 case 0x3f0: /* crc32 Gd,Eb */
3698 case 0x3f1: /* crc32 Gd,Ey */
3699 do_crc32:
3700 if (!(s->cpuid_ext_features & CPUID_EXT_SSE42)) {
3701 goto illegal_op;
3703 if ((b & 0xff) == 0xf0) {
3704 ot = MO_8;
3705 } else if (s->dflag != MO_64) {
3706 ot = (s->prefix & PREFIX_DATA ? MO_16 : MO_32);
3707 } else {
3708 ot = MO_64;
3711 tcg_gen_trunc_tl_i32(cpu_tmp2_i32, cpu_regs[reg]);
3712 gen_ldst_modrm(env, s, modrm, ot, OR_TMP0, 0);
3713 gen_helper_crc32(cpu_T[0], cpu_tmp2_i32,
3714 cpu_T[0], tcg_const_i32(8 << ot));
3716 ot = mo_64_32(s->dflag);
3717 gen_op_mov_reg_v(ot, reg, cpu_T[0]);
3718 break;
3720 case 0x1f0: /* crc32 or movbe */
3721 case 0x1f1:
3722 /* For these insns, the f3 prefix is supposed to have priority
3723 over the 66 prefix, but that's not what we implement above
3724 setting b1. */
3725 if (s->prefix & PREFIX_REPNZ) {
3726 goto do_crc32;
3728 /* FALLTHRU */
3729 case 0x0f0: /* movbe Gy,My */
3730 case 0x0f1: /* movbe My,Gy */
3731 if (!(s->cpuid_ext_features & CPUID_EXT_MOVBE)) {
3732 goto illegal_op;
3734 if (s->dflag != MO_64) {
3735 ot = (s->prefix & PREFIX_DATA ? MO_16 : MO_32);
3736 } else {
3737 ot = MO_64;
3740 gen_lea_modrm(env, s, modrm);
3741 if ((b & 1) == 0) {
3742 tcg_gen_qemu_ld_tl(cpu_T[0], cpu_A0,
3743 s->mem_index, ot | MO_BE);
3744 gen_op_mov_reg_v(ot, reg, cpu_T[0]);
3745 } else {
3746 tcg_gen_qemu_st_tl(cpu_regs[reg], cpu_A0,
3747 s->mem_index, ot | MO_BE);
3749 break;
3751 case 0x0f2: /* andn Gy, By, Ey */
3752 if (!(s->cpuid_7_0_ebx_features & CPUID_7_0_EBX_BMI1)
3753 || !(s->prefix & PREFIX_VEX)
3754 || s->vex_l != 0) {
3755 goto illegal_op;
3757 ot = mo_64_32(s->dflag);
3758 gen_ldst_modrm(env, s, modrm, ot, OR_TMP0, 0);
3759 tcg_gen_andc_tl(cpu_T[0], cpu_regs[s->vex_v], cpu_T[0]);
3760 gen_op_mov_reg_v(ot, reg, cpu_T[0]);
3761 gen_op_update1_cc();
3762 set_cc_op(s, CC_OP_LOGICB + ot);
3763 break;
3765 case 0x0f7: /* bextr Gy, Ey, By */
3766 if (!(s->cpuid_7_0_ebx_features & CPUID_7_0_EBX_BMI1)
3767 || !(s->prefix & PREFIX_VEX)
3768 || s->vex_l != 0) {
3769 goto illegal_op;
3771 ot = mo_64_32(s->dflag);
3773 TCGv bound, zero;
3775 gen_ldst_modrm(env, s, modrm, ot, OR_TMP0, 0);
3776 /* Extract START, and shift the operand.
3777 Shifts larger than operand size get zeros. */
3778 tcg_gen_ext8u_tl(cpu_A0, cpu_regs[s->vex_v]);
3779 tcg_gen_shr_tl(cpu_T[0], cpu_T[0], cpu_A0);
3781 bound = tcg_const_tl(ot == MO_64 ? 63 : 31);
3782 zero = tcg_const_tl(0);
3783 tcg_gen_movcond_tl(TCG_COND_LEU, cpu_T[0], cpu_A0, bound,
3784 cpu_T[0], zero);
3785 tcg_temp_free(zero);
3787 /* Extract the LEN into a mask. Lengths larger than
3788 operand size get all ones. */
3789 tcg_gen_shri_tl(cpu_A0, cpu_regs[s->vex_v], 8);
3790 tcg_gen_ext8u_tl(cpu_A0, cpu_A0);
3791 tcg_gen_movcond_tl(TCG_COND_LEU, cpu_A0, cpu_A0, bound,
3792 cpu_A0, bound);
3793 tcg_temp_free(bound);
3794 tcg_gen_movi_tl(cpu_T[1], 1);
3795 tcg_gen_shl_tl(cpu_T[1], cpu_T[1], cpu_A0);
3796 tcg_gen_subi_tl(cpu_T[1], cpu_T[1], 1);
3797 tcg_gen_and_tl(cpu_T[0], cpu_T[0], cpu_T[1]);
3799 gen_op_mov_reg_v(ot, reg, cpu_T[0]);
3800 gen_op_update1_cc();
3801 set_cc_op(s, CC_OP_LOGICB + ot);
3803 break;
3805 case 0x0f5: /* bzhi Gy, Ey, By */
3806 if (!(s->cpuid_7_0_ebx_features & CPUID_7_0_EBX_BMI2)
3807 || !(s->prefix & PREFIX_VEX)
3808 || s->vex_l != 0) {
3809 goto illegal_op;
3811 ot = mo_64_32(s->dflag);
3812 gen_ldst_modrm(env, s, modrm, ot, OR_TMP0, 0);
3813 tcg_gen_ext8u_tl(cpu_T[1], cpu_regs[s->vex_v]);
3815 TCGv bound = tcg_const_tl(ot == MO_64 ? 63 : 31);
3816 /* Note that since we're using BMILG (in order to get O
3817 cleared) we need to store the inverse into C. */
3818 tcg_gen_setcond_tl(TCG_COND_LT, cpu_cc_src,
3819 cpu_T[1], bound);
3820 tcg_gen_movcond_tl(TCG_COND_GT, cpu_T[1], cpu_T[1],
3821 bound, bound, cpu_T[1]);
3822 tcg_temp_free(bound);
3824 tcg_gen_movi_tl(cpu_A0, -1);
3825 tcg_gen_shl_tl(cpu_A0, cpu_A0, cpu_T[1]);
3826 tcg_gen_andc_tl(cpu_T[0], cpu_T[0], cpu_A0);
3827 gen_op_mov_reg_v(ot, reg, cpu_T[0]);
3828 gen_op_update1_cc();
3829 set_cc_op(s, CC_OP_BMILGB + ot);
3830 break;
3832 case 0x3f6: /* mulx By, Gy, rdx, Ey */
3833 if (!(s->cpuid_7_0_ebx_features & CPUID_7_0_EBX_BMI2)
3834 || !(s->prefix & PREFIX_VEX)
3835 || s->vex_l != 0) {
3836 goto illegal_op;
3838 ot = mo_64_32(s->dflag);
3839 gen_ldst_modrm(env, s, modrm, ot, OR_TMP0, 0);
3840 switch (ot) {
3841 default:
3842 tcg_gen_trunc_tl_i32(cpu_tmp2_i32, cpu_T[0]);
3843 tcg_gen_trunc_tl_i32(cpu_tmp3_i32, cpu_regs[R_EDX]);
3844 tcg_gen_mulu2_i32(cpu_tmp2_i32, cpu_tmp3_i32,
3845 cpu_tmp2_i32, cpu_tmp3_i32);
3846 tcg_gen_extu_i32_tl(cpu_regs[s->vex_v], cpu_tmp2_i32);
3847 tcg_gen_extu_i32_tl(cpu_regs[reg], cpu_tmp3_i32);
3848 break;
3849 #ifdef TARGET_X86_64
3850 case MO_64:
3851 tcg_gen_mulu2_i64(cpu_regs[s->vex_v], cpu_regs[reg],
3852 cpu_T[0], cpu_regs[R_EDX]);
3853 break;
3854 #endif
3856 break;
3858 case 0x3f5: /* pdep Gy, By, Ey */
3859 if (!(s->cpuid_7_0_ebx_features & CPUID_7_0_EBX_BMI2)
3860 || !(s->prefix & PREFIX_VEX)
3861 || s->vex_l != 0) {
3862 goto illegal_op;
3864 ot = mo_64_32(s->dflag);
3865 gen_ldst_modrm(env, s, modrm, ot, OR_TMP0, 0);
3866 /* Note that by zero-extending the mask operand, we
3867 automatically handle zero-extending the result. */
3868 if (ot == MO_64) {
3869 tcg_gen_mov_tl(cpu_T[1], cpu_regs[s->vex_v]);
3870 } else {
3871 tcg_gen_ext32u_tl(cpu_T[1], cpu_regs[s->vex_v]);
3873 gen_helper_pdep(cpu_regs[reg], cpu_T[0], cpu_T[1]);
3874 break;
3876 case 0x2f5: /* pext Gy, By, Ey */
3877 if (!(s->cpuid_7_0_ebx_features & CPUID_7_0_EBX_BMI2)
3878 || !(s->prefix & PREFIX_VEX)
3879 || s->vex_l != 0) {
3880 goto illegal_op;
3882 ot = mo_64_32(s->dflag);
3883 gen_ldst_modrm(env, s, modrm, ot, OR_TMP0, 0);
3884 /* Note that by zero-extending the mask operand, we
3885 automatically handle zero-extending the result. */
3886 if (ot == MO_64) {
3887 tcg_gen_mov_tl(cpu_T[1], cpu_regs[s->vex_v]);
3888 } else {
3889 tcg_gen_ext32u_tl(cpu_T[1], cpu_regs[s->vex_v]);
3891 gen_helper_pext(cpu_regs[reg], cpu_T[0], cpu_T[1]);
3892 break;
3894 case 0x1f6: /* adcx Gy, Ey */
3895 case 0x2f6: /* adox Gy, Ey */
3896 if (!(s->cpuid_7_0_ebx_features & CPUID_7_0_EBX_ADX)) {
3897 goto illegal_op;
3898 } else {
3899 TCGv carry_in, carry_out, zero;
3900 int end_op;
3902 ot = mo_64_32(s->dflag);
3903 gen_ldst_modrm(env, s, modrm, ot, OR_TMP0, 0);
3905 /* Re-use the carry-out from a previous round. */
3906 TCGV_UNUSED(carry_in);
3907 carry_out = (b == 0x1f6 ? cpu_cc_dst : cpu_cc_src2);
3908 switch (s->cc_op) {
3909 case CC_OP_ADCX:
3910 if (b == 0x1f6) {
3911 carry_in = cpu_cc_dst;
3912 end_op = CC_OP_ADCX;
3913 } else {
3914 end_op = CC_OP_ADCOX;
3916 break;
3917 case CC_OP_ADOX:
3918 if (b == 0x1f6) {
3919 end_op = CC_OP_ADCOX;
3920 } else {
3921 carry_in = cpu_cc_src2;
3922 end_op = CC_OP_ADOX;
3924 break;
3925 case CC_OP_ADCOX:
3926 end_op = CC_OP_ADCOX;
3927 carry_in = carry_out;
3928 break;
3929 default:
3930 end_op = (b == 0x1f6 ? CC_OP_ADCX : CC_OP_ADOX);
3931 break;
3933 /* If we can't reuse carry-out, get it out of EFLAGS. */
3934 if (TCGV_IS_UNUSED(carry_in)) {
3935 if (s->cc_op != CC_OP_ADCX && s->cc_op != CC_OP_ADOX) {
3936 gen_compute_eflags(s);
3938 carry_in = cpu_tmp0;
3939 tcg_gen_shri_tl(carry_in, cpu_cc_src,
3940 ctz32(b == 0x1f6 ? CC_C : CC_O));
3941 tcg_gen_andi_tl(carry_in, carry_in, 1);
3944 switch (ot) {
3945 #ifdef TARGET_X86_64
3946 case MO_32:
3947 /* If we know TL is 64-bit, and we want a 32-bit
3948 result, just do everything in 64-bit arithmetic. */
3949 tcg_gen_ext32u_i64(cpu_regs[reg], cpu_regs[reg]);
3950 tcg_gen_ext32u_i64(cpu_T[0], cpu_T[0]);
3951 tcg_gen_add_i64(cpu_T[0], cpu_T[0], cpu_regs[reg]);
3952 tcg_gen_add_i64(cpu_T[0], cpu_T[0], carry_in);
3953 tcg_gen_ext32u_i64(cpu_regs[reg], cpu_T[0]);
3954 tcg_gen_shri_i64(carry_out, cpu_T[0], 32);
3955 break;
3956 #endif
3957 default:
3958 /* Otherwise compute the carry-out in two steps. */
3959 zero = tcg_const_tl(0);
3960 tcg_gen_add2_tl(cpu_T[0], carry_out,
3961 cpu_T[0], zero,
3962 carry_in, zero);
3963 tcg_gen_add2_tl(cpu_regs[reg], carry_out,
3964 cpu_regs[reg], carry_out,
3965 cpu_T[0], zero);
3966 tcg_temp_free(zero);
3967 break;
3969 set_cc_op(s, end_op);
3971 break;
3973 case 0x1f7: /* shlx Gy, Ey, By */
3974 case 0x2f7: /* sarx Gy, Ey, By */
3975 case 0x3f7: /* shrx Gy, Ey, By */
3976 if (!(s->cpuid_7_0_ebx_features & CPUID_7_0_EBX_BMI2)
3977 || !(s->prefix & PREFIX_VEX)
3978 || s->vex_l != 0) {
3979 goto illegal_op;
3981 ot = mo_64_32(s->dflag);
3982 gen_ldst_modrm(env, s, modrm, ot, OR_TMP0, 0);
3983 if (ot == MO_64) {
3984 tcg_gen_andi_tl(cpu_T[1], cpu_regs[s->vex_v], 63);
3985 } else {
3986 tcg_gen_andi_tl(cpu_T[1], cpu_regs[s->vex_v], 31);
3988 if (b == 0x1f7) {
3989 tcg_gen_shl_tl(cpu_T[0], cpu_T[0], cpu_T[1]);
3990 } else if (b == 0x2f7) {
3991 if (ot != MO_64) {
3992 tcg_gen_ext32s_tl(cpu_T[0], cpu_T[0]);
3994 tcg_gen_sar_tl(cpu_T[0], cpu_T[0], cpu_T[1]);
3995 } else {
3996 if (ot != MO_64) {
3997 tcg_gen_ext32u_tl(cpu_T[0], cpu_T[0]);
3999 tcg_gen_shr_tl(cpu_T[0], cpu_T[0], cpu_T[1]);
4001 gen_op_mov_reg_v(ot, reg, cpu_T[0]);
4002 break;
4004 case 0x0f3:
4005 case 0x1f3:
4006 case 0x2f3:
4007 case 0x3f3: /* Group 17 */
4008 if (!(s->cpuid_7_0_ebx_features & CPUID_7_0_EBX_BMI1)
4009 || !(s->prefix & PREFIX_VEX)
4010 || s->vex_l != 0) {
4011 goto illegal_op;
4013 ot = mo_64_32(s->dflag);
4014 gen_ldst_modrm(env, s, modrm, ot, OR_TMP0, 0);
4016 switch (reg & 7) {
4017 case 1: /* blsr By,Ey */
4018 tcg_gen_neg_tl(cpu_T[1], cpu_T[0]);
4019 tcg_gen_and_tl(cpu_T[0], cpu_T[0], cpu_T[1]);
4020 gen_op_mov_reg_v(ot, s->vex_v, cpu_T[0]);
4021 gen_op_update2_cc();
4022 set_cc_op(s, CC_OP_BMILGB + ot);
4023 break;
4025 case 2: /* blsmsk By,Ey */
4026 tcg_gen_mov_tl(cpu_cc_src, cpu_T[0]);
4027 tcg_gen_subi_tl(cpu_T[0], cpu_T[0], 1);
4028 tcg_gen_xor_tl(cpu_T[0], cpu_T[0], cpu_cc_src);
4029 tcg_gen_mov_tl(cpu_cc_dst, cpu_T[0]);
4030 set_cc_op(s, CC_OP_BMILGB + ot);
4031 break;
4033 case 3: /* blsi By, Ey */
4034 tcg_gen_mov_tl(cpu_cc_src, cpu_T[0]);
4035 tcg_gen_subi_tl(cpu_T[0], cpu_T[0], 1);
4036 tcg_gen_and_tl(cpu_T[0], cpu_T[0], cpu_cc_src);
4037 tcg_gen_mov_tl(cpu_cc_dst, cpu_T[0]);
4038 set_cc_op(s, CC_OP_BMILGB + ot);
4039 break;
4041 default:
4042 goto illegal_op;
4044 break;
4046 default:
4047 goto illegal_op;
4049 break;
4051 case 0x03a:
4052 case 0x13a:
4053 b = modrm;
4054 modrm = cpu_ldub_code(env, s->pc++);
4055 rm = modrm & 7;
4056 reg = ((modrm >> 3) & 7) | rex_r;
4057 mod = (modrm >> 6) & 3;
4058 if (b1 >= 2) {
4059 goto illegal_op;
4062 sse_fn_eppi = sse_op_table7[b].op[b1];
4063 if (!sse_fn_eppi) {
4064 goto illegal_op;
4066 if (!(s->cpuid_ext_features & sse_op_table7[b].ext_mask))
4067 goto illegal_op;
4069 if (sse_fn_eppi == SSE_SPECIAL) {
4070 ot = mo_64_32(s->dflag);
4071 rm = (modrm & 7) | REX_B(s);
4072 if (mod != 3)
4073 gen_lea_modrm(env, s, modrm);
4074 reg = ((modrm >> 3) & 7) | rex_r;
4075 val = cpu_ldub_code(env, s->pc++);
4076 switch (b) {
4077 case 0x14: /* pextrb */
4078 tcg_gen_ld8u_tl(cpu_T[0], cpu_env, offsetof(CPUX86State,
4079 xmm_regs[reg].XMM_B(val & 15)));
4080 if (mod == 3) {
4081 gen_op_mov_reg_v(ot, rm, cpu_T[0]);
4082 } else {
4083 tcg_gen_qemu_st_tl(cpu_T[0], cpu_A0,
4084 s->mem_index, MO_UB);
4086 break;
4087 case 0x15: /* pextrw */
4088 tcg_gen_ld16u_tl(cpu_T[0], cpu_env, offsetof(CPUX86State,
4089 xmm_regs[reg].XMM_W(val & 7)));
4090 if (mod == 3) {
4091 gen_op_mov_reg_v(ot, rm, cpu_T[0]);
4092 } else {
4093 tcg_gen_qemu_st_tl(cpu_T[0], cpu_A0,
4094 s->mem_index, MO_LEUW);
4096 break;
4097 case 0x16:
4098 if (ot == MO_32) { /* pextrd */
4099 tcg_gen_ld_i32(cpu_tmp2_i32, cpu_env,
4100 offsetof(CPUX86State,
4101 xmm_regs[reg].XMM_L(val & 3)));
4102 if (mod == 3) {
4103 tcg_gen_extu_i32_tl(cpu_regs[rm], cpu_tmp2_i32);
4104 } else {
4105 tcg_gen_qemu_st_i32(cpu_tmp2_i32, cpu_A0,
4106 s->mem_index, MO_LEUL);
4108 } else { /* pextrq */
4109 #ifdef TARGET_X86_64
4110 tcg_gen_ld_i64(cpu_tmp1_i64, cpu_env,
4111 offsetof(CPUX86State,
4112 xmm_regs[reg].XMM_Q(val & 1)));
4113 if (mod == 3) {
4114 tcg_gen_mov_i64(cpu_regs[rm], cpu_tmp1_i64);
4115 } else {
4116 tcg_gen_qemu_st_i64(cpu_tmp1_i64, cpu_A0,
4117 s->mem_index, MO_LEQ);
4119 #else
4120 goto illegal_op;
4121 #endif
4123 break;
4124 case 0x17: /* extractps */
4125 tcg_gen_ld32u_tl(cpu_T[0], cpu_env, offsetof(CPUX86State,
4126 xmm_regs[reg].XMM_L(val & 3)));
4127 if (mod == 3) {
4128 gen_op_mov_reg_v(ot, rm, cpu_T[0]);
4129 } else {
4130 tcg_gen_qemu_st_tl(cpu_T[0], cpu_A0,
4131 s->mem_index, MO_LEUL);
4133 break;
4134 case 0x20: /* pinsrb */
4135 if (mod == 3) {
4136 gen_op_mov_v_reg(MO_32, cpu_T[0], rm);
4137 } else {
4138 tcg_gen_qemu_ld_tl(cpu_T[0], cpu_A0,
4139 s->mem_index, MO_UB);
4141 tcg_gen_st8_tl(cpu_T[0], cpu_env, offsetof(CPUX86State,
4142 xmm_regs[reg].XMM_B(val & 15)));
4143 break;
4144 case 0x21: /* insertps */
4145 if (mod == 3) {
4146 tcg_gen_ld_i32(cpu_tmp2_i32, cpu_env,
4147 offsetof(CPUX86State,xmm_regs[rm]
4148 .XMM_L((val >> 6) & 3)));
4149 } else {
4150 tcg_gen_qemu_ld_i32(cpu_tmp2_i32, cpu_A0,
4151 s->mem_index, MO_LEUL);
4153 tcg_gen_st_i32(cpu_tmp2_i32, cpu_env,
4154 offsetof(CPUX86State,xmm_regs[reg]
4155 .XMM_L((val >> 4) & 3)));
4156 if ((val >> 0) & 1)
4157 tcg_gen_st_i32(tcg_const_i32(0 /*float32_zero*/),
4158 cpu_env, offsetof(CPUX86State,
4159 xmm_regs[reg].XMM_L(0)));
4160 if ((val >> 1) & 1)
4161 tcg_gen_st_i32(tcg_const_i32(0 /*float32_zero*/),
4162 cpu_env, offsetof(CPUX86State,
4163 xmm_regs[reg].XMM_L(1)));
4164 if ((val >> 2) & 1)
4165 tcg_gen_st_i32(tcg_const_i32(0 /*float32_zero*/),
4166 cpu_env, offsetof(CPUX86State,
4167 xmm_regs[reg].XMM_L(2)));
4168 if ((val >> 3) & 1)
4169 tcg_gen_st_i32(tcg_const_i32(0 /*float32_zero*/),
4170 cpu_env, offsetof(CPUX86State,
4171 xmm_regs[reg].XMM_L(3)));
4172 break;
4173 case 0x22:
4174 if (ot == MO_32) { /* pinsrd */
4175 if (mod == 3) {
4176 tcg_gen_trunc_tl_i32(cpu_tmp2_i32, cpu_regs[rm]);
4177 } else {
4178 tcg_gen_qemu_ld_i32(cpu_tmp2_i32, cpu_A0,
4179 s->mem_index, MO_LEUL);
4181 tcg_gen_st_i32(cpu_tmp2_i32, cpu_env,
4182 offsetof(CPUX86State,
4183 xmm_regs[reg].XMM_L(val & 3)));
4184 } else { /* pinsrq */
4185 #ifdef TARGET_X86_64
4186 if (mod == 3) {
4187 gen_op_mov_v_reg(ot, cpu_tmp1_i64, rm);
4188 } else {
4189 tcg_gen_qemu_ld_i64(cpu_tmp1_i64, cpu_A0,
4190 s->mem_index, MO_LEQ);
4192 tcg_gen_st_i64(cpu_tmp1_i64, cpu_env,
4193 offsetof(CPUX86State,
4194 xmm_regs[reg].XMM_Q(val & 1)));
4195 #else
4196 goto illegal_op;
4197 #endif
4199 break;
4201 return;
4204 if (b1) {
4205 op1_offset = offsetof(CPUX86State,xmm_regs[reg]);
4206 if (mod == 3) {
4207 op2_offset = offsetof(CPUX86State,xmm_regs[rm | REX_B(s)]);
4208 } else {
4209 op2_offset = offsetof(CPUX86State,xmm_t0);
4210 gen_lea_modrm(env, s, modrm);
4211 gen_ldo_env_A0(s, op2_offset);
4213 } else {
4214 op1_offset = offsetof(CPUX86State,fpregs[reg].mmx);
4215 if (mod == 3) {
4216 op2_offset = offsetof(CPUX86State,fpregs[rm].mmx);
4217 } else {
4218 op2_offset = offsetof(CPUX86State,mmx_t0);
4219 gen_lea_modrm(env, s, modrm);
4220 gen_ldq_env_A0(s, op2_offset);
4223 val = cpu_ldub_code(env, s->pc++);
4225 if ((b & 0xfc) == 0x60) { /* pcmpXstrX */
4226 set_cc_op(s, CC_OP_EFLAGS);
4228 if (s->dflag == MO_64) {
4229 /* The helper must use entire 64-bit gp registers */
4230 val |= 1 << 8;
4234 tcg_gen_addi_ptr(cpu_ptr0, cpu_env, op1_offset);
4235 tcg_gen_addi_ptr(cpu_ptr1, cpu_env, op2_offset);
4236 sse_fn_eppi(cpu_env, cpu_ptr0, cpu_ptr1, tcg_const_i32(val));
4237 break;
4239 case 0x33a:
4240 /* Various integer extensions at 0f 3a f[0-f]. */
4241 b = modrm | (b1 << 8);
4242 modrm = cpu_ldub_code(env, s->pc++);
4243 reg = ((modrm >> 3) & 7) | rex_r;
4245 switch (b) {
4246 case 0x3f0: /* rorx Gy,Ey, Ib */
4247 if (!(s->cpuid_7_0_ebx_features & CPUID_7_0_EBX_BMI2)
4248 || !(s->prefix & PREFIX_VEX)
4249 || s->vex_l != 0) {
4250 goto illegal_op;
4252 ot = mo_64_32(s->dflag);
4253 gen_ldst_modrm(env, s, modrm, ot, OR_TMP0, 0);
4254 b = cpu_ldub_code(env, s->pc++);
4255 if (ot == MO_64) {
4256 tcg_gen_rotri_tl(cpu_T[0], cpu_T[0], b & 63);
4257 } else {
4258 tcg_gen_trunc_tl_i32(cpu_tmp2_i32, cpu_T[0]);
4259 tcg_gen_rotri_i32(cpu_tmp2_i32, cpu_tmp2_i32, b & 31);
4260 tcg_gen_extu_i32_tl(cpu_T[0], cpu_tmp2_i32);
4262 gen_op_mov_reg_v(ot, reg, cpu_T[0]);
4263 break;
4265 default:
4266 goto illegal_op;
4268 break;
4270 default:
4271 goto illegal_op;
4273 } else {
4274 /* generic MMX or SSE operation */
4275 switch(b) {
4276 case 0x70: /* pshufx insn */
4277 case 0xc6: /* pshufx insn */
4278 case 0xc2: /* compare insns */
4279 s->rip_offset = 1;
4280 break;
4281 default:
4282 break;
4284 if (is_xmm) {
4285 op1_offset = offsetof(CPUX86State,xmm_regs[reg]);
4286 if (mod != 3) {
4287 gen_lea_modrm(env, s, modrm);
4288 op2_offset = offsetof(CPUX86State,xmm_t0);
4289 if (b1 >= 2 && ((b >= 0x50 && b <= 0x5f && b != 0x5b) ||
4290 b == 0xc2)) {
4291 /* specific case for SSE single instructions */
4292 if (b1 == 2) {
4293 /* 32 bit access */
4294 gen_op_ld_v(s, MO_32, cpu_T[0], cpu_A0);
4295 tcg_gen_st32_tl(cpu_T[0], cpu_env, offsetof(CPUX86State,xmm_t0.XMM_L(0)));
4296 } else {
4297 /* 64 bit access */
4298 gen_ldq_env_A0(s, offsetof(CPUX86State,
4299 xmm_t0.XMM_D(0)));
4301 } else {
4302 gen_ldo_env_A0(s, op2_offset);
4304 } else {
4305 rm = (modrm & 7) | REX_B(s);
4306 op2_offset = offsetof(CPUX86State,xmm_regs[rm]);
4308 } else {
4309 op1_offset = offsetof(CPUX86State,fpregs[reg].mmx);
4310 if (mod != 3) {
4311 gen_lea_modrm(env, s, modrm);
4312 op2_offset = offsetof(CPUX86State,mmx_t0);
4313 gen_ldq_env_A0(s, op2_offset);
4314 } else {
4315 rm = (modrm & 7);
4316 op2_offset = offsetof(CPUX86State,fpregs[rm].mmx);
4319 switch(b) {
4320 case 0x0f: /* 3DNow! data insns */
4321 if (!(s->cpuid_ext2_features & CPUID_EXT2_3DNOW))
4322 goto illegal_op;
4323 val = cpu_ldub_code(env, s->pc++);
4324 sse_fn_epp = sse_op_table5[val];
4325 if (!sse_fn_epp) {
4326 goto illegal_op;
4328 tcg_gen_addi_ptr(cpu_ptr0, cpu_env, op1_offset);
4329 tcg_gen_addi_ptr(cpu_ptr1, cpu_env, op2_offset);
4330 sse_fn_epp(cpu_env, cpu_ptr0, cpu_ptr1);
4331 break;
4332 case 0x70: /* pshufx insn */
4333 case 0xc6: /* pshufx insn */
4334 val = cpu_ldub_code(env, s->pc++);
4335 tcg_gen_addi_ptr(cpu_ptr0, cpu_env, op1_offset);
4336 tcg_gen_addi_ptr(cpu_ptr1, cpu_env, op2_offset);
4337 /* XXX: introduce a new table? */
4338 sse_fn_ppi = (SSEFunc_0_ppi)sse_fn_epp;
4339 sse_fn_ppi(cpu_ptr0, cpu_ptr1, tcg_const_i32(val));
4340 break;
4341 case 0xc2:
4342 /* compare insns */
4343 val = cpu_ldub_code(env, s->pc++);
4344 if (val >= 8)
4345 goto illegal_op;
4346 sse_fn_epp = sse_op_table4[val][b1];
4348 tcg_gen_addi_ptr(cpu_ptr0, cpu_env, op1_offset);
4349 tcg_gen_addi_ptr(cpu_ptr1, cpu_env, op2_offset);
4350 sse_fn_epp(cpu_env, cpu_ptr0, cpu_ptr1);
4351 break;
4352 case 0xf7:
4353 /* maskmov : we must prepare A0 */
4354 if (mod != 3)
4355 goto illegal_op;
4356 tcg_gen_mov_tl(cpu_A0, cpu_regs[R_EDI]);
4357 gen_extu(s->aflag, cpu_A0);
4358 gen_add_A0_ds_seg(s);
4360 tcg_gen_addi_ptr(cpu_ptr0, cpu_env, op1_offset);
4361 tcg_gen_addi_ptr(cpu_ptr1, cpu_env, op2_offset);
4362 /* XXX: introduce a new table? */
4363 sse_fn_eppt = (SSEFunc_0_eppt)sse_fn_epp;
4364 sse_fn_eppt(cpu_env, cpu_ptr0, cpu_ptr1, cpu_A0);
4365 break;
4366 default:
4367 tcg_gen_addi_ptr(cpu_ptr0, cpu_env, op1_offset);
4368 tcg_gen_addi_ptr(cpu_ptr1, cpu_env, op2_offset);
4369 sse_fn_epp(cpu_env, cpu_ptr0, cpu_ptr1);
4370 break;
4372 if (b == 0x2e || b == 0x2f) {
4373 set_cc_op(s, CC_OP_EFLAGS);
4378 /* convert one instruction. s->is_jmp is set if the translation must
4379 be stopped. Return the next pc value */
4380 static target_ulong disas_insn(CPUX86State *env, DisasContext *s,
4381 target_ulong pc_start)
4383 int b, prefixes;
4384 int shift;
4385 TCGMemOp ot, aflag, dflag;
4386 int modrm, reg, rm, mod, op, opreg, val;
4387 target_ulong next_eip, tval;
4388 int rex_w, rex_r;
4390 if (unlikely(qemu_loglevel_mask(CPU_LOG_TB_OP | CPU_LOG_TB_OP_OPT))) {
4391 tcg_gen_debug_insn_start(pc_start);
4393 s->pc = pc_start;
4394 prefixes = 0;
4395 s->override = -1;
4396 rex_w = -1;
4397 rex_r = 0;
4398 #ifdef TARGET_X86_64
4399 s->rex_x = 0;
4400 s->rex_b = 0;
4401 x86_64_hregs = 0;
4402 #endif
4403 s->rip_offset = 0; /* for relative ip address */
4404 s->vex_l = 0;
4405 s->vex_v = 0;
4406 next_byte:
4407 b = cpu_ldub_code(env, s->pc);
4408 s->pc++;
4409 /* Collect prefixes. */
4410 switch (b) {
4411 case 0xf3:
4412 prefixes |= PREFIX_REPZ;
4413 goto next_byte;
4414 case 0xf2:
4415 prefixes |= PREFIX_REPNZ;
4416 goto next_byte;
4417 case 0xf0:
4418 prefixes |= PREFIX_LOCK;
4419 goto next_byte;
4420 case 0x2e:
4421 s->override = R_CS;
4422 goto next_byte;
4423 case 0x36:
4424 s->override = R_SS;
4425 goto next_byte;
4426 case 0x3e:
4427 s->override = R_DS;
4428 goto next_byte;
4429 case 0x26:
4430 s->override = R_ES;
4431 goto next_byte;
4432 case 0x64:
4433 s->override = R_FS;
4434 goto next_byte;
4435 case 0x65:
4436 s->override = R_GS;
4437 goto next_byte;
4438 case 0x66:
4439 prefixes |= PREFIX_DATA;
4440 goto next_byte;
4441 case 0x67:
4442 prefixes |= PREFIX_ADR;
4443 goto next_byte;
4444 #ifdef TARGET_X86_64
4445 case 0x40 ... 0x4f:
4446 if (CODE64(s)) {
4447 /* REX prefix */
4448 rex_w = (b >> 3) & 1;
4449 rex_r = (b & 0x4) << 1;
4450 s->rex_x = (b & 0x2) << 2;
4451 REX_B(s) = (b & 0x1) << 3;
4452 x86_64_hregs = 1; /* select uniform byte register addressing */
4453 goto next_byte;
4455 break;
4456 #endif
4457 case 0xc5: /* 2-byte VEX */
4458 case 0xc4: /* 3-byte VEX */
4459 /* VEX prefixes cannot be used except in 32-bit mode.
4460 Otherwise the instruction is LES or LDS. */
4461 if (s->code32 && !s->vm86) {
4462 static const int pp_prefix[4] = {
4463 0, PREFIX_DATA, PREFIX_REPZ, PREFIX_REPNZ
4465 int vex3, vex2 = cpu_ldub_code(env, s->pc);
4467 if (!CODE64(s) && (vex2 & 0xc0) != 0xc0) {
4468 /* 4.1.4.6: In 32-bit mode, bits [7:6] must be 11b,
4469 otherwise the instruction is LES or LDS. */
4470 break;
4472 s->pc++;
4474 /* 4.1.1-4.1.3: No preceding lock, 66, f2, f3, or rex prefixes. */
4475 if (prefixes & (PREFIX_REPZ | PREFIX_REPNZ
4476 | PREFIX_LOCK | PREFIX_DATA)) {
4477 goto illegal_op;
4479 #ifdef TARGET_X86_64
4480 if (x86_64_hregs) {
4481 goto illegal_op;
4483 #endif
4484 rex_r = (~vex2 >> 4) & 8;
4485 if (b == 0xc5) {
4486 vex3 = vex2;
4487 b = cpu_ldub_code(env, s->pc++);
4488 } else {
4489 #ifdef TARGET_X86_64
4490 s->rex_x = (~vex2 >> 3) & 8;
4491 s->rex_b = (~vex2 >> 2) & 8;
4492 #endif
4493 vex3 = cpu_ldub_code(env, s->pc++);
4494 rex_w = (vex3 >> 7) & 1;
4495 switch (vex2 & 0x1f) {
4496 case 0x01: /* Implied 0f leading opcode bytes. */
4497 b = cpu_ldub_code(env, s->pc++) | 0x100;
4498 break;
4499 case 0x02: /* Implied 0f 38 leading opcode bytes. */
4500 b = 0x138;
4501 break;
4502 case 0x03: /* Implied 0f 3a leading opcode bytes. */
4503 b = 0x13a;
4504 break;
4505 default: /* Reserved for future use. */
4506 goto illegal_op;
4509 s->vex_v = (~vex3 >> 3) & 0xf;
4510 s->vex_l = (vex3 >> 2) & 1;
4511 prefixes |= pp_prefix[vex3 & 3] | PREFIX_VEX;
4513 break;
4516 /* Post-process prefixes. */
4517 if (CODE64(s)) {
4518 /* In 64-bit mode, the default data size is 32-bit. Select 64-bit
4519 data with rex_w, and 16-bit data with 0x66; rex_w takes precedence
4520 over 0x66 if both are present. */
4521 dflag = (rex_w > 0 ? MO_64 : prefixes & PREFIX_DATA ? MO_16 : MO_32);
4522 /* In 64-bit mode, 0x67 selects 32-bit addressing. */
4523 aflag = (prefixes & PREFIX_ADR ? MO_32 : MO_64);
4524 } else {
4525 /* In 16/32-bit mode, 0x66 selects the opposite data size. */
4526 if (s->code32 ^ ((prefixes & PREFIX_DATA) != 0)) {
4527 dflag = MO_32;
4528 } else {
4529 dflag = MO_16;
4531 /* In 16/32-bit mode, 0x67 selects the opposite addressing. */
4532 if (s->code32 ^ ((prefixes & PREFIX_ADR) != 0)) {
4533 aflag = MO_32;
4534 } else {
4535 aflag = MO_16;
4539 s->prefix = prefixes;
4540 s->aflag = aflag;
4541 s->dflag = dflag;
4543 /* lock generation */
4544 if (prefixes & PREFIX_LOCK)
4545 gen_helper_lock();
4547 /* now check op code */
4548 reswitch:
4549 switch(b) {
4550 case 0x0f:
4551 /**************************/
4552 /* extended op code */
4553 b = cpu_ldub_code(env, s->pc++) | 0x100;
4554 goto reswitch;
4556 /**************************/
4557 /* arith & logic */
4558 case 0x00 ... 0x05:
4559 case 0x08 ... 0x0d:
4560 case 0x10 ... 0x15:
4561 case 0x18 ... 0x1d:
4562 case 0x20 ... 0x25:
4563 case 0x28 ... 0x2d:
4564 case 0x30 ... 0x35:
4565 case 0x38 ... 0x3d:
4567 int op, f, val;
4568 op = (b >> 3) & 7;
4569 f = (b >> 1) & 3;
4571 ot = mo_b_d(b, dflag);
4573 switch(f) {
4574 case 0: /* OP Ev, Gv */
4575 modrm = cpu_ldub_code(env, s->pc++);
4576 reg = ((modrm >> 3) & 7) | rex_r;
4577 mod = (modrm >> 6) & 3;
4578 rm = (modrm & 7) | REX_B(s);
4579 if (mod != 3) {
4580 gen_lea_modrm(env, s, modrm);
4581 opreg = OR_TMP0;
4582 } else if (op == OP_XORL && rm == reg) {
4583 xor_zero:
4584 /* xor reg, reg optimisation */
4585 set_cc_op(s, CC_OP_CLR);
4586 tcg_gen_movi_tl(cpu_T[0], 0);
4587 gen_op_mov_reg_v(ot, reg, cpu_T[0]);
4588 break;
4589 } else {
4590 opreg = rm;
4592 gen_op_mov_v_reg(ot, cpu_T[1], reg);
4593 gen_op(s, op, ot, opreg);
4594 break;
4595 case 1: /* OP Gv, Ev */
4596 modrm = cpu_ldub_code(env, s->pc++);
4597 mod = (modrm >> 6) & 3;
4598 reg = ((modrm >> 3) & 7) | rex_r;
4599 rm = (modrm & 7) | REX_B(s);
4600 if (mod != 3) {
4601 gen_lea_modrm(env, s, modrm);
4602 gen_op_ld_v(s, ot, cpu_T[1], cpu_A0);
4603 } else if (op == OP_XORL && rm == reg) {
4604 goto xor_zero;
4605 } else {
4606 gen_op_mov_v_reg(ot, cpu_T[1], rm);
4608 gen_op(s, op, ot, reg);
4609 break;
4610 case 2: /* OP A, Iv */
4611 val = insn_get(env, s, ot);
4612 tcg_gen_movi_tl(cpu_T[1], val);
4613 gen_op(s, op, ot, OR_EAX);
4614 break;
4617 break;
4619 case 0x82:
4620 if (CODE64(s))
4621 goto illegal_op;
4622 case 0x80: /* GRP1 */
4623 case 0x81:
4624 case 0x83:
4626 int val;
4628 ot = mo_b_d(b, dflag);
4630 modrm = cpu_ldub_code(env, s->pc++);
4631 mod = (modrm >> 6) & 3;
4632 rm = (modrm & 7) | REX_B(s);
4633 op = (modrm >> 3) & 7;
4635 if (mod != 3) {
4636 if (b == 0x83)
4637 s->rip_offset = 1;
4638 else
4639 s->rip_offset = insn_const_size(ot);
4640 gen_lea_modrm(env, s, modrm);
4641 opreg = OR_TMP0;
4642 } else {
4643 opreg = rm;
4646 switch(b) {
4647 default:
4648 case 0x80:
4649 case 0x81:
4650 case 0x82:
4651 val = insn_get(env, s, ot);
4652 break;
4653 case 0x83:
4654 val = (int8_t)insn_get(env, s, MO_8);
4655 break;
4657 tcg_gen_movi_tl(cpu_T[1], val);
4658 gen_op(s, op, ot, opreg);
4660 break;
4662 /**************************/
4663 /* inc, dec, and other misc arith */
4664 case 0x40 ... 0x47: /* inc Gv */
4665 ot = dflag;
4666 gen_inc(s, ot, OR_EAX + (b & 7), 1);
4667 break;
4668 case 0x48 ... 0x4f: /* dec Gv */
4669 ot = dflag;
4670 gen_inc(s, ot, OR_EAX + (b & 7), -1);
4671 break;
4672 case 0xf6: /* GRP3 */
4673 case 0xf7:
4674 ot = mo_b_d(b, dflag);
4676 modrm = cpu_ldub_code(env, s->pc++);
4677 mod = (modrm >> 6) & 3;
4678 rm = (modrm & 7) | REX_B(s);
4679 op = (modrm >> 3) & 7;
4680 if (mod != 3) {
4681 if (op == 0)
4682 s->rip_offset = insn_const_size(ot);
4683 gen_lea_modrm(env, s, modrm);
4684 gen_op_ld_v(s, ot, cpu_T[0], cpu_A0);
4685 } else {
4686 gen_op_mov_v_reg(ot, cpu_T[0], rm);
4689 switch(op) {
4690 case 0: /* test */
4691 val = insn_get(env, s, ot);
4692 tcg_gen_movi_tl(cpu_T[1], val);
4693 gen_op_testl_T0_T1_cc();
4694 set_cc_op(s, CC_OP_LOGICB + ot);
4695 break;
4696 case 2: /* not */
4697 tcg_gen_not_tl(cpu_T[0], cpu_T[0]);
4698 if (mod != 3) {
4699 gen_op_st_v(s, ot, cpu_T[0], cpu_A0);
4700 } else {
4701 gen_op_mov_reg_v(ot, rm, cpu_T[0]);
4703 break;
4704 case 3: /* neg */
4705 tcg_gen_neg_tl(cpu_T[0], cpu_T[0]);
4706 if (mod != 3) {
4707 gen_op_st_v(s, ot, cpu_T[0], cpu_A0);
4708 } else {
4709 gen_op_mov_reg_v(ot, rm, cpu_T[0]);
4711 gen_op_update_neg_cc();
4712 set_cc_op(s, CC_OP_SUBB + ot);
4713 break;
4714 case 4: /* mul */
4715 switch(ot) {
4716 case MO_8:
4717 gen_op_mov_v_reg(MO_8, cpu_T[1], R_EAX);
4718 tcg_gen_ext8u_tl(cpu_T[0], cpu_T[0]);
4719 tcg_gen_ext8u_tl(cpu_T[1], cpu_T[1]);
4720 /* XXX: use 32 bit mul which could be faster */
4721 tcg_gen_mul_tl(cpu_T[0], cpu_T[0], cpu_T[1]);
4722 gen_op_mov_reg_v(MO_16, R_EAX, cpu_T[0]);
4723 tcg_gen_mov_tl(cpu_cc_dst, cpu_T[0]);
4724 tcg_gen_andi_tl(cpu_cc_src, cpu_T[0], 0xff00);
4725 set_cc_op(s, CC_OP_MULB);
4726 break;
4727 case MO_16:
4728 gen_op_mov_v_reg(MO_16, cpu_T[1], R_EAX);
4729 tcg_gen_ext16u_tl(cpu_T[0], cpu_T[0]);
4730 tcg_gen_ext16u_tl(cpu_T[1], cpu_T[1]);
4731 /* XXX: use 32 bit mul which could be faster */
4732 tcg_gen_mul_tl(cpu_T[0], cpu_T[0], cpu_T[1]);
4733 gen_op_mov_reg_v(MO_16, R_EAX, cpu_T[0]);
4734 tcg_gen_mov_tl(cpu_cc_dst, cpu_T[0]);
4735 tcg_gen_shri_tl(cpu_T[0], cpu_T[0], 16);
4736 gen_op_mov_reg_v(MO_16, R_EDX, cpu_T[0]);
4737 tcg_gen_mov_tl(cpu_cc_src, cpu_T[0]);
4738 set_cc_op(s, CC_OP_MULW);
4739 break;
4740 default:
4741 case MO_32:
4742 tcg_gen_trunc_tl_i32(cpu_tmp2_i32, cpu_T[0]);
4743 tcg_gen_trunc_tl_i32(cpu_tmp3_i32, cpu_regs[R_EAX]);
4744 tcg_gen_mulu2_i32(cpu_tmp2_i32, cpu_tmp3_i32,
4745 cpu_tmp2_i32, cpu_tmp3_i32);
4746 tcg_gen_extu_i32_tl(cpu_regs[R_EAX], cpu_tmp2_i32);
4747 tcg_gen_extu_i32_tl(cpu_regs[R_EDX], cpu_tmp3_i32);
4748 tcg_gen_mov_tl(cpu_cc_dst, cpu_regs[R_EAX]);
4749 tcg_gen_mov_tl(cpu_cc_src, cpu_regs[R_EDX]);
4750 set_cc_op(s, CC_OP_MULL);
4751 break;
4752 #ifdef TARGET_X86_64
4753 case MO_64:
4754 tcg_gen_mulu2_i64(cpu_regs[R_EAX], cpu_regs[R_EDX],
4755 cpu_T[0], cpu_regs[R_EAX]);
4756 tcg_gen_mov_tl(cpu_cc_dst, cpu_regs[R_EAX]);
4757 tcg_gen_mov_tl(cpu_cc_src, cpu_regs[R_EDX]);
4758 set_cc_op(s, CC_OP_MULQ);
4759 break;
4760 #endif
4762 break;
4763 case 5: /* imul */
4764 switch(ot) {
4765 case MO_8:
4766 gen_op_mov_v_reg(MO_8, cpu_T[1], R_EAX);
4767 tcg_gen_ext8s_tl(cpu_T[0], cpu_T[0]);
4768 tcg_gen_ext8s_tl(cpu_T[1], cpu_T[1]);
4769 /* XXX: use 32 bit mul which could be faster */
4770 tcg_gen_mul_tl(cpu_T[0], cpu_T[0], cpu_T[1]);
4771 gen_op_mov_reg_v(MO_16, R_EAX, cpu_T[0]);
4772 tcg_gen_mov_tl(cpu_cc_dst, cpu_T[0]);
4773 tcg_gen_ext8s_tl(cpu_tmp0, cpu_T[0]);
4774 tcg_gen_sub_tl(cpu_cc_src, cpu_T[0], cpu_tmp0);
4775 set_cc_op(s, CC_OP_MULB);
4776 break;
4777 case MO_16:
4778 gen_op_mov_v_reg(MO_16, cpu_T[1], R_EAX);
4779 tcg_gen_ext16s_tl(cpu_T[0], cpu_T[0]);
4780 tcg_gen_ext16s_tl(cpu_T[1], cpu_T[1]);
4781 /* XXX: use 32 bit mul which could be faster */
4782 tcg_gen_mul_tl(cpu_T[0], cpu_T[0], cpu_T[1]);
4783 gen_op_mov_reg_v(MO_16, R_EAX, cpu_T[0]);
4784 tcg_gen_mov_tl(cpu_cc_dst, cpu_T[0]);
4785 tcg_gen_ext16s_tl(cpu_tmp0, cpu_T[0]);
4786 tcg_gen_sub_tl(cpu_cc_src, cpu_T[0], cpu_tmp0);
4787 tcg_gen_shri_tl(cpu_T[0], cpu_T[0], 16);
4788 gen_op_mov_reg_v(MO_16, R_EDX, cpu_T[0]);
4789 set_cc_op(s, CC_OP_MULW);
4790 break;
4791 default:
4792 case MO_32:
4793 tcg_gen_trunc_tl_i32(cpu_tmp2_i32, cpu_T[0]);
4794 tcg_gen_trunc_tl_i32(cpu_tmp3_i32, cpu_regs[R_EAX]);
4795 tcg_gen_muls2_i32(cpu_tmp2_i32, cpu_tmp3_i32,
4796 cpu_tmp2_i32, cpu_tmp3_i32);
4797 tcg_gen_extu_i32_tl(cpu_regs[R_EAX], cpu_tmp2_i32);
4798 tcg_gen_extu_i32_tl(cpu_regs[R_EDX], cpu_tmp3_i32);
4799 tcg_gen_sari_i32(cpu_tmp2_i32, cpu_tmp2_i32, 31);
4800 tcg_gen_mov_tl(cpu_cc_dst, cpu_regs[R_EAX]);
4801 tcg_gen_sub_i32(cpu_tmp2_i32, cpu_tmp2_i32, cpu_tmp3_i32);
4802 tcg_gen_extu_i32_tl(cpu_cc_src, cpu_tmp2_i32);
4803 set_cc_op(s, CC_OP_MULL);
4804 break;
4805 #ifdef TARGET_X86_64
4806 case MO_64:
4807 tcg_gen_muls2_i64(cpu_regs[R_EAX], cpu_regs[R_EDX],
4808 cpu_T[0], cpu_regs[R_EAX]);
4809 tcg_gen_mov_tl(cpu_cc_dst, cpu_regs[R_EAX]);
4810 tcg_gen_sari_tl(cpu_cc_src, cpu_regs[R_EAX], 63);
4811 tcg_gen_sub_tl(cpu_cc_src, cpu_cc_src, cpu_regs[R_EDX]);
4812 set_cc_op(s, CC_OP_MULQ);
4813 break;
4814 #endif
4816 break;
4817 case 6: /* div */
4818 switch(ot) {
4819 case MO_8:
4820 gen_jmp_im(pc_start - s->cs_base);
4821 gen_helper_divb_AL(cpu_env, cpu_T[0]);
4822 break;
4823 case MO_16:
4824 gen_jmp_im(pc_start - s->cs_base);
4825 gen_helper_divw_AX(cpu_env, cpu_T[0]);
4826 break;
4827 default:
4828 case MO_32:
4829 gen_jmp_im(pc_start - s->cs_base);
4830 gen_helper_divl_EAX(cpu_env, cpu_T[0]);
4831 break;
4832 #ifdef TARGET_X86_64
4833 case MO_64:
4834 gen_jmp_im(pc_start - s->cs_base);
4835 gen_helper_divq_EAX(cpu_env, cpu_T[0]);
4836 break;
4837 #endif
4839 break;
4840 case 7: /* idiv */
4841 switch(ot) {
4842 case MO_8:
4843 gen_jmp_im(pc_start - s->cs_base);
4844 gen_helper_idivb_AL(cpu_env, cpu_T[0]);
4845 break;
4846 case MO_16:
4847 gen_jmp_im(pc_start - s->cs_base);
4848 gen_helper_idivw_AX(cpu_env, cpu_T[0]);
4849 break;
4850 default:
4851 case MO_32:
4852 gen_jmp_im(pc_start - s->cs_base);
4853 gen_helper_idivl_EAX(cpu_env, cpu_T[0]);
4854 break;
4855 #ifdef TARGET_X86_64
4856 case MO_64:
4857 gen_jmp_im(pc_start - s->cs_base);
4858 gen_helper_idivq_EAX(cpu_env, cpu_T[0]);
4859 break;
4860 #endif
4862 break;
4863 default:
4864 goto illegal_op;
4866 break;
4868 case 0xfe: /* GRP4 */
4869 case 0xff: /* GRP5 */
4870 ot = mo_b_d(b, dflag);
4872 modrm = cpu_ldub_code(env, s->pc++);
4873 mod = (modrm >> 6) & 3;
4874 rm = (modrm & 7) | REX_B(s);
4875 op = (modrm >> 3) & 7;
4876 if (op >= 2 && b == 0xfe) {
4877 goto illegal_op;
4879 if (CODE64(s)) {
4880 if (op == 2 || op == 4) {
4881 /* operand size for jumps is 64 bit */
4882 ot = MO_64;
4883 } else if (op == 3 || op == 5) {
4884 ot = dflag != MO_16 ? MO_32 + (rex_w == 1) : MO_16;
4885 } else if (op == 6) {
4886 /* default push size is 64 bit */
4887 ot = mo_pushpop(s, dflag);
4890 if (mod != 3) {
4891 gen_lea_modrm(env, s, modrm);
4892 if (op >= 2 && op != 3 && op != 5)
4893 gen_op_ld_v(s, ot, cpu_T[0], cpu_A0);
4894 } else {
4895 gen_op_mov_v_reg(ot, cpu_T[0], rm);
4898 switch(op) {
4899 case 0: /* inc Ev */
4900 if (mod != 3)
4901 opreg = OR_TMP0;
4902 else
4903 opreg = rm;
4904 gen_inc(s, ot, opreg, 1);
4905 break;
4906 case 1: /* dec Ev */
4907 if (mod != 3)
4908 opreg = OR_TMP0;
4909 else
4910 opreg = rm;
4911 gen_inc(s, ot, opreg, -1);
4912 break;
4913 case 2: /* call Ev */
4914 /* XXX: optimize if memory (no 'and' is necessary) */
4915 if (dflag == MO_16) {
4916 tcg_gen_ext16u_tl(cpu_T[0], cpu_T[0]);
4918 next_eip = s->pc - s->cs_base;
4919 tcg_gen_movi_tl(cpu_T[1], next_eip);
4920 gen_push_v(s, cpu_T[1]);
4921 gen_op_jmp_v(cpu_T[0]);
4922 gen_eob(s);
4923 break;
4924 case 3: /* lcall Ev */
4925 gen_op_ld_v(s, ot, cpu_T[1], cpu_A0);
4926 gen_add_A0_im(s, 1 << ot);
4927 gen_op_ld_v(s, MO_16, cpu_T[0], cpu_A0);
4928 do_lcall:
4929 if (s->pe && !s->vm86) {
4930 gen_update_cc_op(s);
4931 gen_jmp_im(pc_start - s->cs_base);
4932 tcg_gen_trunc_tl_i32(cpu_tmp2_i32, cpu_T[0]);
4933 gen_helper_lcall_protected(cpu_env, cpu_tmp2_i32, cpu_T[1],
4934 tcg_const_i32(dflag - 1),
4935 tcg_const_i32(s->pc - pc_start));
4936 } else {
4937 tcg_gen_trunc_tl_i32(cpu_tmp2_i32, cpu_T[0]);
4938 gen_helper_lcall_real(cpu_env, cpu_tmp2_i32, cpu_T[1],
4939 tcg_const_i32(dflag - 1),
4940 tcg_const_i32(s->pc - s->cs_base));
4942 gen_eob(s);
4943 break;
4944 case 4: /* jmp Ev */
4945 if (dflag == MO_16) {
4946 tcg_gen_ext16u_tl(cpu_T[0], cpu_T[0]);
4948 gen_op_jmp_v(cpu_T[0]);
4949 gen_eob(s);
4950 break;
4951 case 5: /* ljmp Ev */
4952 gen_op_ld_v(s, ot, cpu_T[1], cpu_A0);
4953 gen_add_A0_im(s, 1 << ot);
4954 gen_op_ld_v(s, MO_16, cpu_T[0], cpu_A0);
4955 do_ljmp:
4956 if (s->pe && !s->vm86) {
4957 gen_update_cc_op(s);
4958 gen_jmp_im(pc_start - s->cs_base);
4959 tcg_gen_trunc_tl_i32(cpu_tmp2_i32, cpu_T[0]);
4960 gen_helper_ljmp_protected(cpu_env, cpu_tmp2_i32, cpu_T[1],
4961 tcg_const_i32(s->pc - pc_start));
4962 } else {
4963 gen_op_movl_seg_T0_vm(R_CS);
4964 gen_op_jmp_v(cpu_T[1]);
4966 gen_eob(s);
4967 break;
4968 case 6: /* push Ev */
4969 gen_push_v(s, cpu_T[0]);
4970 break;
4971 default:
4972 goto illegal_op;
4974 break;
4976 case 0x84: /* test Ev, Gv */
4977 case 0x85:
4978 ot = mo_b_d(b, dflag);
4980 modrm = cpu_ldub_code(env, s->pc++);
4981 reg = ((modrm >> 3) & 7) | rex_r;
4983 gen_ldst_modrm(env, s, modrm, ot, OR_TMP0, 0);
4984 gen_op_mov_v_reg(ot, cpu_T[1], reg);
4985 gen_op_testl_T0_T1_cc();
4986 set_cc_op(s, CC_OP_LOGICB + ot);
4987 break;
4989 case 0xa8: /* test eAX, Iv */
4990 case 0xa9:
4991 ot = mo_b_d(b, dflag);
4992 val = insn_get(env, s, ot);
4994 gen_op_mov_v_reg(ot, cpu_T[0], OR_EAX);
4995 tcg_gen_movi_tl(cpu_T[1], val);
4996 gen_op_testl_T0_T1_cc();
4997 set_cc_op(s, CC_OP_LOGICB + ot);
4998 break;
5000 case 0x98: /* CWDE/CBW */
5001 switch (dflag) {
5002 #ifdef TARGET_X86_64
5003 case MO_64:
5004 gen_op_mov_v_reg(MO_32, cpu_T[0], R_EAX);
5005 tcg_gen_ext32s_tl(cpu_T[0], cpu_T[0]);
5006 gen_op_mov_reg_v(MO_64, R_EAX, cpu_T[0]);
5007 break;
5008 #endif
5009 case MO_32:
5010 gen_op_mov_v_reg(MO_16, cpu_T[0], R_EAX);
5011 tcg_gen_ext16s_tl(cpu_T[0], cpu_T[0]);
5012 gen_op_mov_reg_v(MO_32, R_EAX, cpu_T[0]);
5013 break;
5014 case MO_16:
5015 gen_op_mov_v_reg(MO_8, cpu_T[0], R_EAX);
5016 tcg_gen_ext8s_tl(cpu_T[0], cpu_T[0]);
5017 gen_op_mov_reg_v(MO_16, R_EAX, cpu_T[0]);
5018 break;
5019 default:
5020 tcg_abort();
5022 break;
5023 case 0x99: /* CDQ/CWD */
5024 switch (dflag) {
5025 #ifdef TARGET_X86_64
5026 case MO_64:
5027 gen_op_mov_v_reg(MO_64, cpu_T[0], R_EAX);
5028 tcg_gen_sari_tl(cpu_T[0], cpu_T[0], 63);
5029 gen_op_mov_reg_v(MO_64, R_EDX, cpu_T[0]);
5030 break;
5031 #endif
5032 case MO_32:
5033 gen_op_mov_v_reg(MO_32, cpu_T[0], R_EAX);
5034 tcg_gen_ext32s_tl(cpu_T[0], cpu_T[0]);
5035 tcg_gen_sari_tl(cpu_T[0], cpu_T[0], 31);
5036 gen_op_mov_reg_v(MO_32, R_EDX, cpu_T[0]);
5037 break;
5038 case MO_16:
5039 gen_op_mov_v_reg(MO_16, cpu_T[0], R_EAX);
5040 tcg_gen_ext16s_tl(cpu_T[0], cpu_T[0]);
5041 tcg_gen_sari_tl(cpu_T[0], cpu_T[0], 15);
5042 gen_op_mov_reg_v(MO_16, R_EDX, cpu_T[0]);
5043 break;
5044 default:
5045 tcg_abort();
5047 break;
5048 case 0x1af: /* imul Gv, Ev */
5049 case 0x69: /* imul Gv, Ev, I */
5050 case 0x6b:
5051 ot = dflag;
5052 modrm = cpu_ldub_code(env, s->pc++);
5053 reg = ((modrm >> 3) & 7) | rex_r;
5054 if (b == 0x69)
5055 s->rip_offset = insn_const_size(ot);
5056 else if (b == 0x6b)
5057 s->rip_offset = 1;
5058 gen_ldst_modrm(env, s, modrm, ot, OR_TMP0, 0);
5059 if (b == 0x69) {
5060 val = insn_get(env, s, ot);
5061 tcg_gen_movi_tl(cpu_T[1], val);
5062 } else if (b == 0x6b) {
5063 val = (int8_t)insn_get(env, s, MO_8);
5064 tcg_gen_movi_tl(cpu_T[1], val);
5065 } else {
5066 gen_op_mov_v_reg(ot, cpu_T[1], reg);
5068 switch (ot) {
5069 #ifdef TARGET_X86_64
5070 case MO_64:
5071 tcg_gen_muls2_i64(cpu_regs[reg], cpu_T[1], cpu_T[0], cpu_T[1]);
5072 tcg_gen_mov_tl(cpu_cc_dst, cpu_regs[reg]);
5073 tcg_gen_sari_tl(cpu_cc_src, cpu_cc_dst, 63);
5074 tcg_gen_sub_tl(cpu_cc_src, cpu_cc_src, cpu_T[1]);
5075 break;
5076 #endif
5077 case MO_32:
5078 tcg_gen_trunc_tl_i32(cpu_tmp2_i32, cpu_T[0]);
5079 tcg_gen_trunc_tl_i32(cpu_tmp3_i32, cpu_T[1]);
5080 tcg_gen_muls2_i32(cpu_tmp2_i32, cpu_tmp3_i32,
5081 cpu_tmp2_i32, cpu_tmp3_i32);
5082 tcg_gen_extu_i32_tl(cpu_regs[reg], cpu_tmp2_i32);
5083 tcg_gen_sari_i32(cpu_tmp2_i32, cpu_tmp2_i32, 31);
5084 tcg_gen_mov_tl(cpu_cc_dst, cpu_regs[reg]);
5085 tcg_gen_sub_i32(cpu_tmp2_i32, cpu_tmp2_i32, cpu_tmp3_i32);
5086 tcg_gen_extu_i32_tl(cpu_cc_src, cpu_tmp2_i32);
5087 break;
5088 default:
5089 tcg_gen_ext16s_tl(cpu_T[0], cpu_T[0]);
5090 tcg_gen_ext16s_tl(cpu_T[1], cpu_T[1]);
5091 /* XXX: use 32 bit mul which could be faster */
5092 tcg_gen_mul_tl(cpu_T[0], cpu_T[0], cpu_T[1]);
5093 tcg_gen_mov_tl(cpu_cc_dst, cpu_T[0]);
5094 tcg_gen_ext16s_tl(cpu_tmp0, cpu_T[0]);
5095 tcg_gen_sub_tl(cpu_cc_src, cpu_T[0], cpu_tmp0);
5096 gen_op_mov_reg_v(ot, reg, cpu_T[0]);
5097 break;
5099 set_cc_op(s, CC_OP_MULB + ot);
5100 break;
5101 case 0x1c0:
5102 case 0x1c1: /* xadd Ev, Gv */
5103 ot = mo_b_d(b, dflag);
5104 modrm = cpu_ldub_code(env, s->pc++);
5105 reg = ((modrm >> 3) & 7) | rex_r;
5106 mod = (modrm >> 6) & 3;
5107 if (mod == 3) {
5108 rm = (modrm & 7) | REX_B(s);
5109 gen_op_mov_v_reg(ot, cpu_T[0], reg);
5110 gen_op_mov_v_reg(ot, cpu_T[1], rm);
5111 tcg_gen_add_tl(cpu_T[0], cpu_T[0], cpu_T[1]);
5112 gen_op_mov_reg_v(ot, reg, cpu_T[1]);
5113 gen_op_mov_reg_v(ot, rm, cpu_T[0]);
5114 } else {
5115 gen_lea_modrm(env, s, modrm);
5116 gen_op_mov_v_reg(ot, cpu_T[0], reg);
5117 gen_op_ld_v(s, ot, cpu_T[1], cpu_A0);
5118 tcg_gen_add_tl(cpu_T[0], cpu_T[0], cpu_T[1]);
5119 gen_op_st_v(s, ot, cpu_T[0], cpu_A0);
5120 gen_op_mov_reg_v(ot, reg, cpu_T[1]);
5122 gen_op_update2_cc();
5123 set_cc_op(s, CC_OP_ADDB + ot);
5124 break;
5125 case 0x1b0:
5126 case 0x1b1: /* cmpxchg Ev, Gv */
5128 int label1, label2;
5129 TCGv t0, t1, t2, a0;
5131 ot = mo_b_d(b, dflag);
5132 modrm = cpu_ldub_code(env, s->pc++);
5133 reg = ((modrm >> 3) & 7) | rex_r;
5134 mod = (modrm >> 6) & 3;
5135 t0 = tcg_temp_local_new();
5136 t1 = tcg_temp_local_new();
5137 t2 = tcg_temp_local_new();
5138 a0 = tcg_temp_local_new();
5139 gen_op_mov_v_reg(ot, t1, reg);
5140 if (mod == 3) {
5141 rm = (modrm & 7) | REX_B(s);
5142 gen_op_mov_v_reg(ot, t0, rm);
5143 } else {
5144 gen_lea_modrm(env, s, modrm);
5145 tcg_gen_mov_tl(a0, cpu_A0);
5146 gen_op_ld_v(s, ot, t0, a0);
5147 rm = 0; /* avoid warning */
5149 label1 = gen_new_label();
5150 tcg_gen_mov_tl(t2, cpu_regs[R_EAX]);
5151 gen_extu(ot, t0);
5152 gen_extu(ot, t2);
5153 tcg_gen_brcond_tl(TCG_COND_EQ, t2, t0, label1);
5154 label2 = gen_new_label();
5155 if (mod == 3) {
5156 gen_op_mov_reg_v(ot, R_EAX, t0);
5157 tcg_gen_br(label2);
5158 gen_set_label(label1);
5159 gen_op_mov_reg_v(ot, rm, t1);
5160 } else {
5161 /* perform no-op store cycle like physical cpu; must be
5162 before changing accumulator to ensure idempotency if
5163 the store faults and the instruction is restarted */
5164 gen_op_st_v(s, ot, t0, a0);
5165 gen_op_mov_reg_v(ot, R_EAX, t0);
5166 tcg_gen_br(label2);
5167 gen_set_label(label1);
5168 gen_op_st_v(s, ot, t1, a0);
5170 gen_set_label(label2);
5171 tcg_gen_mov_tl(cpu_cc_src, t0);
5172 tcg_gen_mov_tl(cpu_cc_srcT, t2);
5173 tcg_gen_sub_tl(cpu_cc_dst, t2, t0);
5174 set_cc_op(s, CC_OP_SUBB + ot);
5175 tcg_temp_free(t0);
5176 tcg_temp_free(t1);
5177 tcg_temp_free(t2);
5178 tcg_temp_free(a0);
5180 break;
5181 case 0x1c7: /* cmpxchg8b */
5182 modrm = cpu_ldub_code(env, s->pc++);
5183 mod = (modrm >> 6) & 3;
5184 if ((mod == 3) || ((modrm & 0x38) != 0x8))
5185 goto illegal_op;
5186 #ifdef TARGET_X86_64
5187 if (dflag == MO_64) {
5188 if (!(s->cpuid_ext_features & CPUID_EXT_CX16))
5189 goto illegal_op;
5190 gen_jmp_im(pc_start - s->cs_base);
5191 gen_update_cc_op(s);
5192 gen_lea_modrm(env, s, modrm);
5193 gen_helper_cmpxchg16b(cpu_env, cpu_A0);
5194 } else
5195 #endif
5197 if (!(s->cpuid_features & CPUID_CX8))
5198 goto illegal_op;
5199 gen_jmp_im(pc_start - s->cs_base);
5200 gen_update_cc_op(s);
5201 gen_lea_modrm(env, s, modrm);
5202 gen_helper_cmpxchg8b(cpu_env, cpu_A0);
5204 set_cc_op(s, CC_OP_EFLAGS);
5205 break;
5207 /**************************/
5208 /* push/pop */
5209 case 0x50 ... 0x57: /* push */
5210 gen_op_mov_v_reg(MO_32, cpu_T[0], (b & 7) | REX_B(s));
5211 gen_push_v(s, cpu_T[0]);
5212 break;
5213 case 0x58 ... 0x5f: /* pop */
5214 ot = gen_pop_T0(s);
5215 /* NOTE: order is important for pop %sp */
5216 gen_pop_update(s, ot);
5217 gen_op_mov_reg_v(ot, (b & 7) | REX_B(s), cpu_T[0]);
5218 break;
5219 case 0x60: /* pusha */
5220 if (CODE64(s))
5221 goto illegal_op;
5222 gen_pusha(s);
5223 break;
5224 case 0x61: /* popa */
5225 if (CODE64(s))
5226 goto illegal_op;
5227 gen_popa(s);
5228 break;
5229 case 0x68: /* push Iv */
5230 case 0x6a:
5231 ot = mo_pushpop(s, dflag);
5232 if (b == 0x68)
5233 val = insn_get(env, s, ot);
5234 else
5235 val = (int8_t)insn_get(env, s, MO_8);
5236 tcg_gen_movi_tl(cpu_T[0], val);
5237 gen_push_v(s, cpu_T[0]);
5238 break;
5239 case 0x8f: /* pop Ev */
5240 modrm = cpu_ldub_code(env, s->pc++);
5241 mod = (modrm >> 6) & 3;
5242 ot = gen_pop_T0(s);
5243 if (mod == 3) {
5244 /* NOTE: order is important for pop %sp */
5245 gen_pop_update(s, ot);
5246 rm = (modrm & 7) | REX_B(s);
5247 gen_op_mov_reg_v(ot, rm, cpu_T[0]);
5248 } else {
5249 /* NOTE: order is important too for MMU exceptions */
5250 s->popl_esp_hack = 1 << ot;
5251 gen_ldst_modrm(env, s, modrm, ot, OR_TMP0, 1);
5252 s->popl_esp_hack = 0;
5253 gen_pop_update(s, ot);
5255 break;
5256 case 0xc8: /* enter */
5258 int level;
5259 val = cpu_lduw_code(env, s->pc);
5260 s->pc += 2;
5261 level = cpu_ldub_code(env, s->pc++);
5262 gen_enter(s, val, level);
5264 break;
5265 case 0xc9: /* leave */
5266 /* XXX: exception not precise (ESP is updated before potential exception) */
5267 if (CODE64(s)) {
5268 gen_op_mov_v_reg(MO_64, cpu_T[0], R_EBP);
5269 gen_op_mov_reg_v(MO_64, R_ESP, cpu_T[0]);
5270 } else if (s->ss32) {
5271 gen_op_mov_v_reg(MO_32, cpu_T[0], R_EBP);
5272 gen_op_mov_reg_v(MO_32, R_ESP, cpu_T[0]);
5273 } else {
5274 gen_op_mov_v_reg(MO_16, cpu_T[0], R_EBP);
5275 gen_op_mov_reg_v(MO_16, R_ESP, cpu_T[0]);
5277 ot = gen_pop_T0(s);
5278 gen_op_mov_reg_v(ot, R_EBP, cpu_T[0]);
5279 gen_pop_update(s, ot);
5280 break;
5281 case 0x06: /* push es */
5282 case 0x0e: /* push cs */
5283 case 0x16: /* push ss */
5284 case 0x1e: /* push ds */
5285 if (CODE64(s))
5286 goto illegal_op;
5287 gen_op_movl_T0_seg(b >> 3);
5288 gen_push_v(s, cpu_T[0]);
5289 break;
5290 case 0x1a0: /* push fs */
5291 case 0x1a8: /* push gs */
5292 gen_op_movl_T0_seg((b >> 3) & 7);
5293 gen_push_v(s, cpu_T[0]);
5294 break;
5295 case 0x07: /* pop es */
5296 case 0x17: /* pop ss */
5297 case 0x1f: /* pop ds */
5298 if (CODE64(s))
5299 goto illegal_op;
5300 reg = b >> 3;
5301 ot = gen_pop_T0(s);
5302 gen_movl_seg_T0(s, reg, pc_start - s->cs_base);
5303 gen_pop_update(s, ot);
5304 if (reg == R_SS) {
5305 /* if reg == SS, inhibit interrupts/trace. */
5306 /* If several instructions disable interrupts, only the
5307 _first_ does it */
5308 if (!(s->tb->flags & HF_INHIBIT_IRQ_MASK))
5309 gen_helper_set_inhibit_irq(cpu_env);
5310 s->tf = 0;
5312 if (s->is_jmp) {
5313 gen_jmp_im(s->pc - s->cs_base);
5314 gen_eob(s);
5316 break;
5317 case 0x1a1: /* pop fs */
5318 case 0x1a9: /* pop gs */
5319 ot = gen_pop_T0(s);
5320 gen_movl_seg_T0(s, (b >> 3) & 7, pc_start - s->cs_base);
5321 gen_pop_update(s, ot);
5322 if (s->is_jmp) {
5323 gen_jmp_im(s->pc - s->cs_base);
5324 gen_eob(s);
5326 break;
5328 /**************************/
5329 /* mov */
5330 case 0x88:
5331 case 0x89: /* mov Gv, Ev */
5332 ot = mo_b_d(b, dflag);
5333 modrm = cpu_ldub_code(env, s->pc++);
5334 reg = ((modrm >> 3) & 7) | rex_r;
5336 /* generate a generic store */
5337 gen_ldst_modrm(env, s, modrm, ot, reg, 1);
5338 break;
5339 case 0xc6:
5340 case 0xc7: /* mov Ev, Iv */
5341 ot = mo_b_d(b, dflag);
5342 modrm = cpu_ldub_code(env, s->pc++);
5343 mod = (modrm >> 6) & 3;
5344 if (mod != 3) {
5345 s->rip_offset = insn_const_size(ot);
5346 gen_lea_modrm(env, s, modrm);
5348 val = insn_get(env, s, ot);
5349 tcg_gen_movi_tl(cpu_T[0], val);
5350 if (mod != 3) {
5351 gen_op_st_v(s, ot, cpu_T[0], cpu_A0);
5352 } else {
5353 gen_op_mov_reg_v(ot, (modrm & 7) | REX_B(s), cpu_T[0]);
5355 break;
5356 case 0x8a:
5357 case 0x8b: /* mov Ev, Gv */
5358 ot = mo_b_d(b, dflag);
5359 modrm = cpu_ldub_code(env, s->pc++);
5360 reg = ((modrm >> 3) & 7) | rex_r;
5362 gen_ldst_modrm(env, s, modrm, ot, OR_TMP0, 0);
5363 gen_op_mov_reg_v(ot, reg, cpu_T[0]);
5364 break;
5365 case 0x8e: /* mov seg, Gv */
5366 modrm = cpu_ldub_code(env, s->pc++);
5367 reg = (modrm >> 3) & 7;
5368 if (reg >= 6 || reg == R_CS)
5369 goto illegal_op;
5370 gen_ldst_modrm(env, s, modrm, MO_16, OR_TMP0, 0);
5371 gen_movl_seg_T0(s, reg, pc_start - s->cs_base);
5372 if (reg == R_SS) {
5373 /* if reg == SS, inhibit interrupts/trace */
5374 /* If several instructions disable interrupts, only the
5375 _first_ does it */
5376 if (!(s->tb->flags & HF_INHIBIT_IRQ_MASK))
5377 gen_helper_set_inhibit_irq(cpu_env);
5378 s->tf = 0;
5380 if (s->is_jmp) {
5381 gen_jmp_im(s->pc - s->cs_base);
5382 gen_eob(s);
5384 break;
5385 case 0x8c: /* mov Gv, seg */
5386 modrm = cpu_ldub_code(env, s->pc++);
5387 reg = (modrm >> 3) & 7;
5388 mod = (modrm >> 6) & 3;
5389 if (reg >= 6)
5390 goto illegal_op;
5391 gen_op_movl_T0_seg(reg);
5392 ot = mod == 3 ? dflag : MO_16;
5393 gen_ldst_modrm(env, s, modrm, ot, OR_TMP0, 1);
5394 break;
5396 case 0x1b6: /* movzbS Gv, Eb */
5397 case 0x1b7: /* movzwS Gv, Eb */
5398 case 0x1be: /* movsbS Gv, Eb */
5399 case 0x1bf: /* movswS Gv, Eb */
5401 TCGMemOp d_ot;
5402 TCGMemOp s_ot;
5404 /* d_ot is the size of destination */
5405 d_ot = dflag;
5406 /* ot is the size of source */
5407 ot = (b & 1) + MO_8;
5408 /* s_ot is the sign+size of source */
5409 s_ot = b & 8 ? MO_SIGN | ot : ot;
5411 modrm = cpu_ldub_code(env, s->pc++);
5412 reg = ((modrm >> 3) & 7) | rex_r;
5413 mod = (modrm >> 6) & 3;
5414 rm = (modrm & 7) | REX_B(s);
5416 if (mod == 3) {
5417 gen_op_mov_v_reg(ot, cpu_T[0], rm);
5418 switch (s_ot) {
5419 case MO_UB:
5420 tcg_gen_ext8u_tl(cpu_T[0], cpu_T[0]);
5421 break;
5422 case MO_SB:
5423 tcg_gen_ext8s_tl(cpu_T[0], cpu_T[0]);
5424 break;
5425 case MO_UW:
5426 tcg_gen_ext16u_tl(cpu_T[0], cpu_T[0]);
5427 break;
5428 default:
5429 case MO_SW:
5430 tcg_gen_ext16s_tl(cpu_T[0], cpu_T[0]);
5431 break;
5433 gen_op_mov_reg_v(d_ot, reg, cpu_T[0]);
5434 } else {
5435 gen_lea_modrm(env, s, modrm);
5436 gen_op_ld_v(s, s_ot, cpu_T[0], cpu_A0);
5437 gen_op_mov_reg_v(d_ot, reg, cpu_T[0]);
5440 break;
5442 case 0x8d: /* lea */
5443 ot = dflag;
5444 modrm = cpu_ldub_code(env, s->pc++);
5445 mod = (modrm >> 6) & 3;
5446 if (mod == 3)
5447 goto illegal_op;
5448 reg = ((modrm >> 3) & 7) | rex_r;
5449 /* we must ensure that no segment is added */
5450 s->override = -1;
5451 val = s->addseg;
5452 s->addseg = 0;
5453 gen_lea_modrm(env, s, modrm);
5454 s->addseg = val;
5455 gen_op_mov_reg_v(ot, reg, cpu_A0);
5456 break;
5458 case 0xa0: /* mov EAX, Ov */
5459 case 0xa1:
5460 case 0xa2: /* mov Ov, EAX */
5461 case 0xa3:
5463 target_ulong offset_addr;
5465 ot = mo_b_d(b, dflag);
5466 switch (s->aflag) {
5467 #ifdef TARGET_X86_64
5468 case MO_64:
5469 offset_addr = cpu_ldq_code(env, s->pc);
5470 s->pc += 8;
5471 break;
5472 #endif
5473 default:
5474 offset_addr = insn_get(env, s, s->aflag);
5475 break;
5477 tcg_gen_movi_tl(cpu_A0, offset_addr);
5478 gen_add_A0_ds_seg(s);
5479 if ((b & 2) == 0) {
5480 gen_op_ld_v(s, ot, cpu_T[0], cpu_A0);
5481 gen_op_mov_reg_v(ot, R_EAX, cpu_T[0]);
5482 } else {
5483 gen_op_mov_v_reg(ot, cpu_T[0], R_EAX);
5484 gen_op_st_v(s, ot, cpu_T[0], cpu_A0);
5487 break;
5488 case 0xd7: /* xlat */
5489 tcg_gen_mov_tl(cpu_A0, cpu_regs[R_EBX]);
5490 tcg_gen_ext8u_tl(cpu_T[0], cpu_regs[R_EAX]);
5491 tcg_gen_add_tl(cpu_A0, cpu_A0, cpu_T[0]);
5492 gen_extu(s->aflag, cpu_A0);
5493 gen_add_A0_ds_seg(s);
5494 gen_op_ld_v(s, MO_8, cpu_T[0], cpu_A0);
5495 gen_op_mov_reg_v(MO_8, R_EAX, cpu_T[0]);
5496 break;
5497 case 0xb0 ... 0xb7: /* mov R, Ib */
5498 val = insn_get(env, s, MO_8);
5499 tcg_gen_movi_tl(cpu_T[0], val);
5500 gen_op_mov_reg_v(MO_8, (b & 7) | REX_B(s), cpu_T[0]);
5501 break;
5502 case 0xb8 ... 0xbf: /* mov R, Iv */
5503 #ifdef TARGET_X86_64
5504 if (dflag == MO_64) {
5505 uint64_t tmp;
5506 /* 64 bit case */
5507 tmp = cpu_ldq_code(env, s->pc);
5508 s->pc += 8;
5509 reg = (b & 7) | REX_B(s);
5510 tcg_gen_movi_tl(cpu_T[0], tmp);
5511 gen_op_mov_reg_v(MO_64, reg, cpu_T[0]);
5512 } else
5513 #endif
5515 ot = dflag;
5516 val = insn_get(env, s, ot);
5517 reg = (b & 7) | REX_B(s);
5518 tcg_gen_movi_tl(cpu_T[0], val);
5519 gen_op_mov_reg_v(ot, reg, cpu_T[0]);
5521 break;
5523 case 0x91 ... 0x97: /* xchg R, EAX */
5524 do_xchg_reg_eax:
5525 ot = dflag;
5526 reg = (b & 7) | REX_B(s);
5527 rm = R_EAX;
5528 goto do_xchg_reg;
5529 case 0x86:
5530 case 0x87: /* xchg Ev, Gv */
5531 ot = mo_b_d(b, dflag);
5532 modrm = cpu_ldub_code(env, s->pc++);
5533 reg = ((modrm >> 3) & 7) | rex_r;
5534 mod = (modrm >> 6) & 3;
5535 if (mod == 3) {
5536 rm = (modrm & 7) | REX_B(s);
5537 do_xchg_reg:
5538 gen_op_mov_v_reg(ot, cpu_T[0], reg);
5539 gen_op_mov_v_reg(ot, cpu_T[1], rm);
5540 gen_op_mov_reg_v(ot, rm, cpu_T[0]);
5541 gen_op_mov_reg_v(ot, reg, cpu_T[1]);
5542 } else {
5543 gen_lea_modrm(env, s, modrm);
5544 gen_op_mov_v_reg(ot, cpu_T[0], reg);
5545 /* for xchg, lock is implicit */
5546 if (!(prefixes & PREFIX_LOCK))
5547 gen_helper_lock();
5548 gen_op_ld_v(s, ot, cpu_T[1], cpu_A0);
5549 gen_op_st_v(s, ot, cpu_T[0], cpu_A0);
5550 if (!(prefixes & PREFIX_LOCK))
5551 gen_helper_unlock();
5552 gen_op_mov_reg_v(ot, reg, cpu_T[1]);
5554 break;
5555 case 0xc4: /* les Gv */
5556 /* In CODE64 this is VEX3; see above. */
5557 op = R_ES;
5558 goto do_lxx;
5559 case 0xc5: /* lds Gv */
5560 /* In CODE64 this is VEX2; see above. */
5561 op = R_DS;
5562 goto do_lxx;
5563 case 0x1b2: /* lss Gv */
5564 op = R_SS;
5565 goto do_lxx;
5566 case 0x1b4: /* lfs Gv */
5567 op = R_FS;
5568 goto do_lxx;
5569 case 0x1b5: /* lgs Gv */
5570 op = R_GS;
5571 do_lxx:
5572 ot = dflag != MO_16 ? MO_32 : MO_16;
5573 modrm = cpu_ldub_code(env, s->pc++);
5574 reg = ((modrm >> 3) & 7) | rex_r;
5575 mod = (modrm >> 6) & 3;
5576 if (mod == 3)
5577 goto illegal_op;
5578 gen_lea_modrm(env, s, modrm);
5579 gen_op_ld_v(s, ot, cpu_T[1], cpu_A0);
5580 gen_add_A0_im(s, 1 << ot);
5581 /* load the segment first to handle exceptions properly */
5582 gen_op_ld_v(s, MO_16, cpu_T[0], cpu_A0);
5583 gen_movl_seg_T0(s, op, pc_start - s->cs_base);
5584 /* then put the data */
5585 gen_op_mov_reg_v(ot, reg, cpu_T[1]);
5586 if (s->is_jmp) {
5587 gen_jmp_im(s->pc - s->cs_base);
5588 gen_eob(s);
5590 break;
5592 /************************/
5593 /* shifts */
5594 case 0xc0:
5595 case 0xc1:
5596 /* shift Ev,Ib */
5597 shift = 2;
5598 grp2:
5600 ot = mo_b_d(b, dflag);
5601 modrm = cpu_ldub_code(env, s->pc++);
5602 mod = (modrm >> 6) & 3;
5603 op = (modrm >> 3) & 7;
5605 if (mod != 3) {
5606 if (shift == 2) {
5607 s->rip_offset = 1;
5609 gen_lea_modrm(env, s, modrm);
5610 opreg = OR_TMP0;
5611 } else {
5612 opreg = (modrm & 7) | REX_B(s);
5615 /* simpler op */
5616 if (shift == 0) {
5617 gen_shift(s, op, ot, opreg, OR_ECX);
5618 } else {
5619 if (shift == 2) {
5620 shift = cpu_ldub_code(env, s->pc++);
5622 gen_shifti(s, op, ot, opreg, shift);
5625 break;
5626 case 0xd0:
5627 case 0xd1:
5628 /* shift Ev,1 */
5629 shift = 1;
5630 goto grp2;
5631 case 0xd2:
5632 case 0xd3:
5633 /* shift Ev,cl */
5634 shift = 0;
5635 goto grp2;
5637 case 0x1a4: /* shld imm */
5638 op = 0;
5639 shift = 1;
5640 goto do_shiftd;
5641 case 0x1a5: /* shld cl */
5642 op = 0;
5643 shift = 0;
5644 goto do_shiftd;
5645 case 0x1ac: /* shrd imm */
5646 op = 1;
5647 shift = 1;
5648 goto do_shiftd;
5649 case 0x1ad: /* shrd cl */
5650 op = 1;
5651 shift = 0;
5652 do_shiftd:
5653 ot = dflag;
5654 modrm = cpu_ldub_code(env, s->pc++);
5655 mod = (modrm >> 6) & 3;
5656 rm = (modrm & 7) | REX_B(s);
5657 reg = ((modrm >> 3) & 7) | rex_r;
5658 if (mod != 3) {
5659 gen_lea_modrm(env, s, modrm);
5660 opreg = OR_TMP0;
5661 } else {
5662 opreg = rm;
5664 gen_op_mov_v_reg(ot, cpu_T[1], reg);
5666 if (shift) {
5667 TCGv imm = tcg_const_tl(cpu_ldub_code(env, s->pc++));
5668 gen_shiftd_rm_T1(s, ot, opreg, op, imm);
5669 tcg_temp_free(imm);
5670 } else {
5671 gen_shiftd_rm_T1(s, ot, opreg, op, cpu_regs[R_ECX]);
5673 break;
5675 /************************/
5676 /* floats */
5677 case 0xd8 ... 0xdf:
5678 if (s->flags & (HF_EM_MASK | HF_TS_MASK)) {
5679 /* if CR0.EM or CR0.TS are set, generate an FPU exception */
5680 /* XXX: what to do if illegal op ? */
5681 gen_exception(s, EXCP07_PREX, pc_start - s->cs_base);
5682 break;
5684 modrm = cpu_ldub_code(env, s->pc++);
5685 mod = (modrm >> 6) & 3;
5686 rm = modrm & 7;
5687 op = ((b & 7) << 3) | ((modrm >> 3) & 7);
5688 if (mod != 3) {
5689 /* memory op */
5690 gen_lea_modrm(env, s, modrm);
5691 switch(op) {
5692 case 0x00 ... 0x07: /* fxxxs */
5693 case 0x10 ... 0x17: /* fixxxl */
5694 case 0x20 ... 0x27: /* fxxxl */
5695 case 0x30 ... 0x37: /* fixxx */
5697 int op1;
5698 op1 = op & 7;
5700 switch(op >> 4) {
5701 case 0:
5702 tcg_gen_qemu_ld_i32(cpu_tmp2_i32, cpu_A0,
5703 s->mem_index, MO_LEUL);
5704 gen_helper_flds_FT0(cpu_env, cpu_tmp2_i32);
5705 break;
5706 case 1:
5707 tcg_gen_qemu_ld_i32(cpu_tmp2_i32, cpu_A0,
5708 s->mem_index, MO_LEUL);
5709 gen_helper_fildl_FT0(cpu_env, cpu_tmp2_i32);
5710 break;
5711 case 2:
5712 tcg_gen_qemu_ld_i64(cpu_tmp1_i64, cpu_A0,
5713 s->mem_index, MO_LEQ);
5714 gen_helper_fldl_FT0(cpu_env, cpu_tmp1_i64);
5715 break;
5716 case 3:
5717 default:
5718 tcg_gen_qemu_ld_i32(cpu_tmp2_i32, cpu_A0,
5719 s->mem_index, MO_LESW);
5720 gen_helper_fildl_FT0(cpu_env, cpu_tmp2_i32);
5721 break;
5724 gen_helper_fp_arith_ST0_FT0(op1);
5725 if (op1 == 3) {
5726 /* fcomp needs pop */
5727 gen_helper_fpop(cpu_env);
5730 break;
5731 case 0x08: /* flds */
5732 case 0x0a: /* fsts */
5733 case 0x0b: /* fstps */
5734 case 0x18 ... 0x1b: /* fildl, fisttpl, fistl, fistpl */
5735 case 0x28 ... 0x2b: /* fldl, fisttpll, fstl, fstpl */
5736 case 0x38 ... 0x3b: /* filds, fisttps, fists, fistps */
5737 switch(op & 7) {
5738 case 0:
5739 switch(op >> 4) {
5740 case 0:
5741 tcg_gen_qemu_ld_i32(cpu_tmp2_i32, cpu_A0,
5742 s->mem_index, MO_LEUL);
5743 gen_helper_flds_ST0(cpu_env, cpu_tmp2_i32);
5744 break;
5745 case 1:
5746 tcg_gen_qemu_ld_i32(cpu_tmp2_i32, cpu_A0,
5747 s->mem_index, MO_LEUL);
5748 gen_helper_fildl_ST0(cpu_env, cpu_tmp2_i32);
5749 break;
5750 case 2:
5751 tcg_gen_qemu_ld_i64(cpu_tmp1_i64, cpu_A0,
5752 s->mem_index, MO_LEQ);
5753 gen_helper_fldl_ST0(cpu_env, cpu_tmp1_i64);
5754 break;
5755 case 3:
5756 default:
5757 tcg_gen_qemu_ld_i32(cpu_tmp2_i32, cpu_A0,
5758 s->mem_index, MO_LESW);
5759 gen_helper_fildl_ST0(cpu_env, cpu_tmp2_i32);
5760 break;
5762 break;
5763 case 1:
5764 /* XXX: the corresponding CPUID bit must be tested ! */
5765 switch(op >> 4) {
5766 case 1:
5767 gen_helper_fisttl_ST0(cpu_tmp2_i32, cpu_env);
5768 tcg_gen_qemu_st_i32(cpu_tmp2_i32, cpu_A0,
5769 s->mem_index, MO_LEUL);
5770 break;
5771 case 2:
5772 gen_helper_fisttll_ST0(cpu_tmp1_i64, cpu_env);
5773 tcg_gen_qemu_st_i64(cpu_tmp1_i64, cpu_A0,
5774 s->mem_index, MO_LEQ);
5775 break;
5776 case 3:
5777 default:
5778 gen_helper_fistt_ST0(cpu_tmp2_i32, cpu_env);
5779 tcg_gen_qemu_st_i32(cpu_tmp2_i32, cpu_A0,
5780 s->mem_index, MO_LEUW);
5781 break;
5783 gen_helper_fpop(cpu_env);
5784 break;
5785 default:
5786 switch(op >> 4) {
5787 case 0:
5788 gen_helper_fsts_ST0(cpu_tmp2_i32, cpu_env);
5789 tcg_gen_qemu_st_i32(cpu_tmp2_i32, cpu_A0,
5790 s->mem_index, MO_LEUL);
5791 break;
5792 case 1:
5793 gen_helper_fistl_ST0(cpu_tmp2_i32, cpu_env);
5794 tcg_gen_qemu_st_i32(cpu_tmp2_i32, cpu_A0,
5795 s->mem_index, MO_LEUL);
5796 break;
5797 case 2:
5798 gen_helper_fstl_ST0(cpu_tmp1_i64, cpu_env);
5799 tcg_gen_qemu_st_i64(cpu_tmp1_i64, cpu_A0,
5800 s->mem_index, MO_LEQ);
5801 break;
5802 case 3:
5803 default:
5804 gen_helper_fist_ST0(cpu_tmp2_i32, cpu_env);
5805 tcg_gen_qemu_st_i32(cpu_tmp2_i32, cpu_A0,
5806 s->mem_index, MO_LEUW);
5807 break;
5809 if ((op & 7) == 3)
5810 gen_helper_fpop(cpu_env);
5811 break;
5813 break;
5814 case 0x0c: /* fldenv mem */
5815 gen_update_cc_op(s);
5816 gen_jmp_im(pc_start - s->cs_base);
5817 gen_helper_fldenv(cpu_env, cpu_A0, tcg_const_i32(dflag - 1));
5818 break;
5819 case 0x0d: /* fldcw mem */
5820 tcg_gen_qemu_ld_i32(cpu_tmp2_i32, cpu_A0,
5821 s->mem_index, MO_LEUW);
5822 gen_helper_fldcw(cpu_env, cpu_tmp2_i32);
5823 break;
5824 case 0x0e: /* fnstenv mem */
5825 gen_update_cc_op(s);
5826 gen_jmp_im(pc_start - s->cs_base);
5827 gen_helper_fstenv(cpu_env, cpu_A0, tcg_const_i32(dflag - 1));
5828 break;
5829 case 0x0f: /* fnstcw mem */
5830 gen_helper_fnstcw(cpu_tmp2_i32, cpu_env);
5831 tcg_gen_qemu_st_i32(cpu_tmp2_i32, cpu_A0,
5832 s->mem_index, MO_LEUW);
5833 break;
5834 case 0x1d: /* fldt mem */
5835 gen_update_cc_op(s);
5836 gen_jmp_im(pc_start - s->cs_base);
5837 gen_helper_fldt_ST0(cpu_env, cpu_A0);
5838 break;
5839 case 0x1f: /* fstpt mem */
5840 gen_update_cc_op(s);
5841 gen_jmp_im(pc_start - s->cs_base);
5842 gen_helper_fstt_ST0(cpu_env, cpu_A0);
5843 gen_helper_fpop(cpu_env);
5844 break;
5845 case 0x2c: /* frstor mem */
5846 gen_update_cc_op(s);
5847 gen_jmp_im(pc_start - s->cs_base);
5848 gen_helper_frstor(cpu_env, cpu_A0, tcg_const_i32(dflag - 1));
5849 break;
5850 case 0x2e: /* fnsave mem */
5851 gen_update_cc_op(s);
5852 gen_jmp_im(pc_start - s->cs_base);
5853 gen_helper_fsave(cpu_env, cpu_A0, tcg_const_i32(dflag - 1));
5854 break;
5855 case 0x2f: /* fnstsw mem */
5856 gen_helper_fnstsw(cpu_tmp2_i32, cpu_env);
5857 tcg_gen_qemu_st_i32(cpu_tmp2_i32, cpu_A0,
5858 s->mem_index, MO_LEUW);
5859 break;
5860 case 0x3c: /* fbld */
5861 gen_update_cc_op(s);
5862 gen_jmp_im(pc_start - s->cs_base);
5863 gen_helper_fbld_ST0(cpu_env, cpu_A0);
5864 break;
5865 case 0x3e: /* fbstp */
5866 gen_update_cc_op(s);
5867 gen_jmp_im(pc_start - s->cs_base);
5868 gen_helper_fbst_ST0(cpu_env, cpu_A0);
5869 gen_helper_fpop(cpu_env);
5870 break;
5871 case 0x3d: /* fildll */
5872 tcg_gen_qemu_ld_i64(cpu_tmp1_i64, cpu_A0, s->mem_index, MO_LEQ);
5873 gen_helper_fildll_ST0(cpu_env, cpu_tmp1_i64);
5874 break;
5875 case 0x3f: /* fistpll */
5876 gen_helper_fistll_ST0(cpu_tmp1_i64, cpu_env);
5877 tcg_gen_qemu_st_i64(cpu_tmp1_i64, cpu_A0, s->mem_index, MO_LEQ);
5878 gen_helper_fpop(cpu_env);
5879 break;
5880 default:
5881 goto illegal_op;
5883 } else {
5884 /* register float ops */
5885 opreg = rm;
5887 switch(op) {
5888 case 0x08: /* fld sti */
5889 gen_helper_fpush(cpu_env);
5890 gen_helper_fmov_ST0_STN(cpu_env,
5891 tcg_const_i32((opreg + 1) & 7));
5892 break;
5893 case 0x09: /* fxchg sti */
5894 case 0x29: /* fxchg4 sti, undocumented op */
5895 case 0x39: /* fxchg7 sti, undocumented op */
5896 gen_helper_fxchg_ST0_STN(cpu_env, tcg_const_i32(opreg));
5897 break;
5898 case 0x0a: /* grp d9/2 */
5899 switch(rm) {
5900 case 0: /* fnop */
5901 /* check exceptions (FreeBSD FPU probe) */
5902 gen_update_cc_op(s);
5903 gen_jmp_im(pc_start - s->cs_base);
5904 gen_helper_fwait(cpu_env);
5905 break;
5906 default:
5907 goto illegal_op;
5909 break;
5910 case 0x0c: /* grp d9/4 */
5911 switch(rm) {
5912 case 0: /* fchs */
5913 gen_helper_fchs_ST0(cpu_env);
5914 break;
5915 case 1: /* fabs */
5916 gen_helper_fabs_ST0(cpu_env);
5917 break;
5918 case 4: /* ftst */
5919 gen_helper_fldz_FT0(cpu_env);
5920 gen_helper_fcom_ST0_FT0(cpu_env);
5921 break;
5922 case 5: /* fxam */
5923 gen_helper_fxam_ST0(cpu_env);
5924 break;
5925 default:
5926 goto illegal_op;
5928 break;
5929 case 0x0d: /* grp d9/5 */
5931 switch(rm) {
5932 case 0:
5933 gen_helper_fpush(cpu_env);
5934 gen_helper_fld1_ST0(cpu_env);
5935 break;
5936 case 1:
5937 gen_helper_fpush(cpu_env);
5938 gen_helper_fldl2t_ST0(cpu_env);
5939 break;
5940 case 2:
5941 gen_helper_fpush(cpu_env);
5942 gen_helper_fldl2e_ST0(cpu_env);
5943 break;
5944 case 3:
5945 gen_helper_fpush(cpu_env);
5946 gen_helper_fldpi_ST0(cpu_env);
5947 break;
5948 case 4:
5949 gen_helper_fpush(cpu_env);
5950 gen_helper_fldlg2_ST0(cpu_env);
5951 break;
5952 case 5:
5953 gen_helper_fpush(cpu_env);
5954 gen_helper_fldln2_ST0(cpu_env);
5955 break;
5956 case 6:
5957 gen_helper_fpush(cpu_env);
5958 gen_helper_fldz_ST0(cpu_env);
5959 break;
5960 default:
5961 goto illegal_op;
5964 break;
5965 case 0x0e: /* grp d9/6 */
5966 switch(rm) {
5967 case 0: /* f2xm1 */
5968 gen_helper_f2xm1(cpu_env);
5969 break;
5970 case 1: /* fyl2x */
5971 gen_helper_fyl2x(cpu_env);
5972 break;
5973 case 2: /* fptan */
5974 gen_helper_fptan(cpu_env);
5975 break;
5976 case 3: /* fpatan */
5977 gen_helper_fpatan(cpu_env);
5978 break;
5979 case 4: /* fxtract */
5980 gen_helper_fxtract(cpu_env);
5981 break;
5982 case 5: /* fprem1 */
5983 gen_helper_fprem1(cpu_env);
5984 break;
5985 case 6: /* fdecstp */
5986 gen_helper_fdecstp(cpu_env);
5987 break;
5988 default:
5989 case 7: /* fincstp */
5990 gen_helper_fincstp(cpu_env);
5991 break;
5993 break;
5994 case 0x0f: /* grp d9/7 */
5995 switch(rm) {
5996 case 0: /* fprem */
5997 gen_helper_fprem(cpu_env);
5998 break;
5999 case 1: /* fyl2xp1 */
6000 gen_helper_fyl2xp1(cpu_env);
6001 break;
6002 case 2: /* fsqrt */
6003 gen_helper_fsqrt(cpu_env);
6004 break;
6005 case 3: /* fsincos */
6006 gen_helper_fsincos(cpu_env);
6007 break;
6008 case 5: /* fscale */
6009 gen_helper_fscale(cpu_env);
6010 break;
6011 case 4: /* frndint */
6012 gen_helper_frndint(cpu_env);
6013 break;
6014 case 6: /* fsin */
6015 gen_helper_fsin(cpu_env);
6016 break;
6017 default:
6018 case 7: /* fcos */
6019 gen_helper_fcos(cpu_env);
6020 break;
6022 break;
6023 case 0x00: case 0x01: case 0x04 ... 0x07: /* fxxx st, sti */
6024 case 0x20: case 0x21: case 0x24 ... 0x27: /* fxxx sti, st */
6025 case 0x30: case 0x31: case 0x34 ... 0x37: /* fxxxp sti, st */
6027 int op1;
6029 op1 = op & 7;
6030 if (op >= 0x20) {
6031 gen_helper_fp_arith_STN_ST0(op1, opreg);
6032 if (op >= 0x30)
6033 gen_helper_fpop(cpu_env);
6034 } else {
6035 gen_helper_fmov_FT0_STN(cpu_env, tcg_const_i32(opreg));
6036 gen_helper_fp_arith_ST0_FT0(op1);
6039 break;
6040 case 0x02: /* fcom */
6041 case 0x22: /* fcom2, undocumented op */
6042 gen_helper_fmov_FT0_STN(cpu_env, tcg_const_i32(opreg));
6043 gen_helper_fcom_ST0_FT0(cpu_env);
6044 break;
6045 case 0x03: /* fcomp */
6046 case 0x23: /* fcomp3, undocumented op */
6047 case 0x32: /* fcomp5, undocumented op */
6048 gen_helper_fmov_FT0_STN(cpu_env, tcg_const_i32(opreg));
6049 gen_helper_fcom_ST0_FT0(cpu_env);
6050 gen_helper_fpop(cpu_env);
6051 break;
6052 case 0x15: /* da/5 */
6053 switch(rm) {
6054 case 1: /* fucompp */
6055 gen_helper_fmov_FT0_STN(cpu_env, tcg_const_i32(1));
6056 gen_helper_fucom_ST0_FT0(cpu_env);
6057 gen_helper_fpop(cpu_env);
6058 gen_helper_fpop(cpu_env);
6059 break;
6060 default:
6061 goto illegal_op;
6063 break;
6064 case 0x1c:
6065 switch(rm) {
6066 case 0: /* feni (287 only, just do nop here) */
6067 break;
6068 case 1: /* fdisi (287 only, just do nop here) */
6069 break;
6070 case 2: /* fclex */
6071 gen_helper_fclex(cpu_env);
6072 break;
6073 case 3: /* fninit */
6074 gen_helper_fninit(cpu_env);
6075 break;
6076 case 4: /* fsetpm (287 only, just do nop here) */
6077 break;
6078 default:
6079 goto illegal_op;
6081 break;
6082 case 0x1d: /* fucomi */
6083 if (!(s->cpuid_features & CPUID_CMOV)) {
6084 goto illegal_op;
6086 gen_update_cc_op(s);
6087 gen_helper_fmov_FT0_STN(cpu_env, tcg_const_i32(opreg));
6088 gen_helper_fucomi_ST0_FT0(cpu_env);
6089 set_cc_op(s, CC_OP_EFLAGS);
6090 break;
6091 case 0x1e: /* fcomi */
6092 if (!(s->cpuid_features & CPUID_CMOV)) {
6093 goto illegal_op;
6095 gen_update_cc_op(s);
6096 gen_helper_fmov_FT0_STN(cpu_env, tcg_const_i32(opreg));
6097 gen_helper_fcomi_ST0_FT0(cpu_env);
6098 set_cc_op(s, CC_OP_EFLAGS);
6099 break;
6100 case 0x28: /* ffree sti */
6101 gen_helper_ffree_STN(cpu_env, tcg_const_i32(opreg));
6102 break;
6103 case 0x2a: /* fst sti */
6104 gen_helper_fmov_STN_ST0(cpu_env, tcg_const_i32(opreg));
6105 break;
6106 case 0x2b: /* fstp sti */
6107 case 0x0b: /* fstp1 sti, undocumented op */
6108 case 0x3a: /* fstp8 sti, undocumented op */
6109 case 0x3b: /* fstp9 sti, undocumented op */
6110 gen_helper_fmov_STN_ST0(cpu_env, tcg_const_i32(opreg));
6111 gen_helper_fpop(cpu_env);
6112 break;
6113 case 0x2c: /* fucom st(i) */
6114 gen_helper_fmov_FT0_STN(cpu_env, tcg_const_i32(opreg));
6115 gen_helper_fucom_ST0_FT0(cpu_env);
6116 break;
6117 case 0x2d: /* fucomp st(i) */
6118 gen_helper_fmov_FT0_STN(cpu_env, tcg_const_i32(opreg));
6119 gen_helper_fucom_ST0_FT0(cpu_env);
6120 gen_helper_fpop(cpu_env);
6121 break;
6122 case 0x33: /* de/3 */
6123 switch(rm) {
6124 case 1: /* fcompp */
6125 gen_helper_fmov_FT0_STN(cpu_env, tcg_const_i32(1));
6126 gen_helper_fcom_ST0_FT0(cpu_env);
6127 gen_helper_fpop(cpu_env);
6128 gen_helper_fpop(cpu_env);
6129 break;
6130 default:
6131 goto illegal_op;
6133 break;
6134 case 0x38: /* ffreep sti, undocumented op */
6135 gen_helper_ffree_STN(cpu_env, tcg_const_i32(opreg));
6136 gen_helper_fpop(cpu_env);
6137 break;
6138 case 0x3c: /* df/4 */
6139 switch(rm) {
6140 case 0:
6141 gen_helper_fnstsw(cpu_tmp2_i32, cpu_env);
6142 tcg_gen_extu_i32_tl(cpu_T[0], cpu_tmp2_i32);
6143 gen_op_mov_reg_v(MO_16, R_EAX, cpu_T[0]);
6144 break;
6145 default:
6146 goto illegal_op;
6148 break;
6149 case 0x3d: /* fucomip */
6150 if (!(s->cpuid_features & CPUID_CMOV)) {
6151 goto illegal_op;
6153 gen_update_cc_op(s);
6154 gen_helper_fmov_FT0_STN(cpu_env, tcg_const_i32(opreg));
6155 gen_helper_fucomi_ST0_FT0(cpu_env);
6156 gen_helper_fpop(cpu_env);
6157 set_cc_op(s, CC_OP_EFLAGS);
6158 break;
6159 case 0x3e: /* fcomip */
6160 if (!(s->cpuid_features & CPUID_CMOV)) {
6161 goto illegal_op;
6163 gen_update_cc_op(s);
6164 gen_helper_fmov_FT0_STN(cpu_env, tcg_const_i32(opreg));
6165 gen_helper_fcomi_ST0_FT0(cpu_env);
6166 gen_helper_fpop(cpu_env);
6167 set_cc_op(s, CC_OP_EFLAGS);
6168 break;
6169 case 0x10 ... 0x13: /* fcmovxx */
6170 case 0x18 ... 0x1b:
6172 int op1, l1;
6173 static const uint8_t fcmov_cc[8] = {
6174 (JCC_B << 1),
6175 (JCC_Z << 1),
6176 (JCC_BE << 1),
6177 (JCC_P << 1),
6180 if (!(s->cpuid_features & CPUID_CMOV)) {
6181 goto illegal_op;
6183 op1 = fcmov_cc[op & 3] | (((op >> 3) & 1) ^ 1);
6184 l1 = gen_new_label();
6185 gen_jcc1_noeob(s, op1, l1);
6186 gen_helper_fmov_ST0_STN(cpu_env, tcg_const_i32(opreg));
6187 gen_set_label(l1);
6189 break;
6190 default:
6191 goto illegal_op;
6194 break;
6195 /************************/
6196 /* string ops */
6198 case 0xa4: /* movsS */
6199 case 0xa5:
6200 ot = mo_b_d(b, dflag);
6201 if (prefixes & (PREFIX_REPZ | PREFIX_REPNZ)) {
6202 gen_repz_movs(s, ot, pc_start - s->cs_base, s->pc - s->cs_base);
6203 } else {
6204 gen_movs(s, ot);
6206 break;
6208 case 0xaa: /* stosS */
6209 case 0xab:
6210 ot = mo_b_d(b, dflag);
6211 if (prefixes & (PREFIX_REPZ | PREFIX_REPNZ)) {
6212 gen_repz_stos(s, ot, pc_start - s->cs_base, s->pc - s->cs_base);
6213 } else {
6214 gen_stos(s, ot);
6216 break;
6217 case 0xac: /* lodsS */
6218 case 0xad:
6219 ot = mo_b_d(b, dflag);
6220 if (prefixes & (PREFIX_REPZ | PREFIX_REPNZ)) {
6221 gen_repz_lods(s, ot, pc_start - s->cs_base, s->pc - s->cs_base);
6222 } else {
6223 gen_lods(s, ot);
6225 break;
6226 case 0xae: /* scasS */
6227 case 0xaf:
6228 ot = mo_b_d(b, dflag);
6229 if (prefixes & PREFIX_REPNZ) {
6230 gen_repz_scas(s, ot, pc_start - s->cs_base, s->pc - s->cs_base, 1);
6231 } else if (prefixes & PREFIX_REPZ) {
6232 gen_repz_scas(s, ot, pc_start - s->cs_base, s->pc - s->cs_base, 0);
6233 } else {
6234 gen_scas(s, ot);
6236 break;
6238 case 0xa6: /* cmpsS */
6239 case 0xa7:
6240 ot = mo_b_d(b, dflag);
6241 if (prefixes & PREFIX_REPNZ) {
6242 gen_repz_cmps(s, ot, pc_start - s->cs_base, s->pc - s->cs_base, 1);
6243 } else if (prefixes & PREFIX_REPZ) {
6244 gen_repz_cmps(s, ot, pc_start - s->cs_base, s->pc - s->cs_base, 0);
6245 } else {
6246 gen_cmps(s, ot);
6248 break;
6249 case 0x6c: /* insS */
6250 case 0x6d:
6251 ot = mo_b_d32(b, dflag);
6252 tcg_gen_ext16u_tl(cpu_T[0], cpu_regs[R_EDX]);
6253 gen_check_io(s, ot, pc_start - s->cs_base,
6254 SVM_IOIO_TYPE_MASK | svm_is_rep(prefixes) | 4);
6255 if (prefixes & (PREFIX_REPZ | PREFIX_REPNZ)) {
6256 gen_repz_ins(s, ot, pc_start - s->cs_base, s->pc - s->cs_base);
6257 } else {
6258 gen_ins(s, ot);
6259 if (use_icount) {
6260 gen_jmp(s, s->pc - s->cs_base);
6263 break;
6264 case 0x6e: /* outsS */
6265 case 0x6f:
6266 ot = mo_b_d32(b, dflag);
6267 tcg_gen_ext16u_tl(cpu_T[0], cpu_regs[R_EDX]);
6268 gen_check_io(s, ot, pc_start - s->cs_base,
6269 svm_is_rep(prefixes) | 4);
6270 if (prefixes & (PREFIX_REPZ | PREFIX_REPNZ)) {
6271 gen_repz_outs(s, ot, pc_start - s->cs_base, s->pc - s->cs_base);
6272 } else {
6273 gen_outs(s, ot);
6274 if (use_icount) {
6275 gen_jmp(s, s->pc - s->cs_base);
6278 break;
6280 /************************/
6281 /* port I/O */
6283 case 0xe4:
6284 case 0xe5:
6285 ot = mo_b_d32(b, dflag);
6286 val = cpu_ldub_code(env, s->pc++);
6287 gen_check_io(s, ot, pc_start - s->cs_base,
6288 SVM_IOIO_TYPE_MASK | svm_is_rep(prefixes));
6289 if (use_icount)
6290 gen_io_start();
6291 tcg_gen_movi_i32(cpu_tmp2_i32, val);
6292 gen_helper_in_func(ot, cpu_T[1], cpu_tmp2_i32);
6293 gen_op_mov_reg_v(ot, R_EAX, cpu_T[1]);
6294 if (use_icount) {
6295 gen_io_end();
6296 gen_jmp(s, s->pc - s->cs_base);
6298 break;
6299 case 0xe6:
6300 case 0xe7:
6301 ot = mo_b_d32(b, dflag);
6302 val = cpu_ldub_code(env, s->pc++);
6303 gen_check_io(s, ot, pc_start - s->cs_base,
6304 svm_is_rep(prefixes));
6305 gen_op_mov_v_reg(ot, cpu_T[1], R_EAX);
6307 if (use_icount)
6308 gen_io_start();
6309 tcg_gen_movi_i32(cpu_tmp2_i32, val);
6310 tcg_gen_trunc_tl_i32(cpu_tmp3_i32, cpu_T[1]);
6311 gen_helper_out_func(ot, cpu_tmp2_i32, cpu_tmp3_i32);
6312 if (use_icount) {
6313 gen_io_end();
6314 gen_jmp(s, s->pc - s->cs_base);
6316 break;
6317 case 0xec:
6318 case 0xed:
6319 ot = mo_b_d32(b, dflag);
6320 tcg_gen_ext16u_tl(cpu_T[0], cpu_regs[R_EDX]);
6321 gen_check_io(s, ot, pc_start - s->cs_base,
6322 SVM_IOIO_TYPE_MASK | svm_is_rep(prefixes));
6323 if (use_icount)
6324 gen_io_start();
6325 tcg_gen_trunc_tl_i32(cpu_tmp2_i32, cpu_T[0]);
6326 gen_helper_in_func(ot, cpu_T[1], cpu_tmp2_i32);
6327 gen_op_mov_reg_v(ot, R_EAX, cpu_T[1]);
6328 if (use_icount) {
6329 gen_io_end();
6330 gen_jmp(s, s->pc - s->cs_base);
6332 break;
6333 case 0xee:
6334 case 0xef:
6335 ot = mo_b_d32(b, dflag);
6336 tcg_gen_ext16u_tl(cpu_T[0], cpu_regs[R_EDX]);
6337 gen_check_io(s, ot, pc_start - s->cs_base,
6338 svm_is_rep(prefixes));
6339 gen_op_mov_v_reg(ot, cpu_T[1], R_EAX);
6341 if (use_icount)
6342 gen_io_start();
6343 tcg_gen_trunc_tl_i32(cpu_tmp2_i32, cpu_T[0]);
6344 tcg_gen_trunc_tl_i32(cpu_tmp3_i32, cpu_T[1]);
6345 gen_helper_out_func(ot, cpu_tmp2_i32, cpu_tmp3_i32);
6346 if (use_icount) {
6347 gen_io_end();
6348 gen_jmp(s, s->pc - s->cs_base);
6350 break;
6352 /************************/
6353 /* control */
6354 case 0xc2: /* ret im */
6355 val = cpu_ldsw_code(env, s->pc);
6356 s->pc += 2;
6357 ot = gen_pop_T0(s);
6358 gen_stack_update(s, val + (1 << ot));
6359 /* Note that gen_pop_T0 uses a zero-extending load. */
6360 gen_op_jmp_v(cpu_T[0]);
6361 gen_eob(s);
6362 break;
6363 case 0xc3: /* ret */
6364 ot = gen_pop_T0(s);
6365 gen_pop_update(s, ot);
6366 /* Note that gen_pop_T0 uses a zero-extending load. */
6367 gen_op_jmp_v(cpu_T[0]);
6368 gen_eob(s);
6369 break;
6370 case 0xca: /* lret im */
6371 val = cpu_ldsw_code(env, s->pc);
6372 s->pc += 2;
6373 do_lret:
6374 if (s->pe && !s->vm86) {
6375 gen_update_cc_op(s);
6376 gen_jmp_im(pc_start - s->cs_base);
6377 gen_helper_lret_protected(cpu_env, tcg_const_i32(dflag - 1),
6378 tcg_const_i32(val));
6379 } else {
6380 gen_stack_A0(s);
6381 /* pop offset */
6382 gen_op_ld_v(s, dflag, cpu_T[0], cpu_A0);
6383 /* NOTE: keeping EIP updated is not a problem in case of
6384 exception */
6385 gen_op_jmp_v(cpu_T[0]);
6386 /* pop selector */
6387 gen_op_addl_A0_im(1 << dflag);
6388 gen_op_ld_v(s, dflag, cpu_T[0], cpu_A0);
6389 gen_op_movl_seg_T0_vm(R_CS);
6390 /* add stack offset */
6391 gen_stack_update(s, val + (2 << dflag));
6393 gen_eob(s);
6394 break;
6395 case 0xcb: /* lret */
6396 val = 0;
6397 goto do_lret;
6398 case 0xcf: /* iret */
6399 gen_svm_check_intercept(s, pc_start, SVM_EXIT_IRET);
6400 if (!s->pe) {
6401 /* real mode */
6402 gen_helper_iret_real(cpu_env, tcg_const_i32(dflag - 1));
6403 set_cc_op(s, CC_OP_EFLAGS);
6404 } else if (s->vm86) {
6405 if (s->iopl != 3) {
6406 gen_exception(s, EXCP0D_GPF, pc_start - s->cs_base);
6407 } else {
6408 gen_helper_iret_real(cpu_env, tcg_const_i32(dflag - 1));
6409 set_cc_op(s, CC_OP_EFLAGS);
6411 } else {
6412 gen_update_cc_op(s);
6413 gen_jmp_im(pc_start - s->cs_base);
6414 gen_helper_iret_protected(cpu_env, tcg_const_i32(dflag - 1),
6415 tcg_const_i32(s->pc - s->cs_base));
6416 set_cc_op(s, CC_OP_EFLAGS);
6418 gen_eob(s);
6419 break;
6420 case 0xe8: /* call im */
6422 if (dflag != MO_16) {
6423 tval = (int32_t)insn_get(env, s, MO_32);
6424 } else {
6425 tval = (int16_t)insn_get(env, s, MO_16);
6427 next_eip = s->pc - s->cs_base;
6428 tval += next_eip;
6429 if (dflag == MO_16) {
6430 tval &= 0xffff;
6431 } else if (!CODE64(s)) {
6432 tval &= 0xffffffff;
6434 tcg_gen_movi_tl(cpu_T[0], next_eip);
6435 gen_push_v(s, cpu_T[0]);
6436 gen_jmp(s, tval);
6438 break;
6439 case 0x9a: /* lcall im */
6441 unsigned int selector, offset;
6443 if (CODE64(s))
6444 goto illegal_op;
6445 ot = dflag;
6446 offset = insn_get(env, s, ot);
6447 selector = insn_get(env, s, MO_16);
6449 tcg_gen_movi_tl(cpu_T[0], selector);
6450 tcg_gen_movi_tl(cpu_T[1], offset);
6452 goto do_lcall;
6453 case 0xe9: /* jmp im */
6454 if (dflag != MO_16) {
6455 tval = (int32_t)insn_get(env, s, MO_32);
6456 } else {
6457 tval = (int16_t)insn_get(env, s, MO_16);
6459 tval += s->pc - s->cs_base;
6460 if (dflag == MO_16) {
6461 tval &= 0xffff;
6462 } else if (!CODE64(s)) {
6463 tval &= 0xffffffff;
6465 gen_jmp(s, tval);
6466 break;
6467 case 0xea: /* ljmp im */
6469 unsigned int selector, offset;
6471 if (CODE64(s))
6472 goto illegal_op;
6473 ot = dflag;
6474 offset = insn_get(env, s, ot);
6475 selector = insn_get(env, s, MO_16);
6477 tcg_gen_movi_tl(cpu_T[0], selector);
6478 tcg_gen_movi_tl(cpu_T[1], offset);
6480 goto do_ljmp;
6481 case 0xeb: /* jmp Jb */
6482 tval = (int8_t)insn_get(env, s, MO_8);
6483 tval += s->pc - s->cs_base;
6484 if (dflag == MO_16) {
6485 tval &= 0xffff;
6487 gen_jmp(s, tval);
6488 break;
6489 case 0x70 ... 0x7f: /* jcc Jb */
6490 tval = (int8_t)insn_get(env, s, MO_8);
6491 goto do_jcc;
6492 case 0x180 ... 0x18f: /* jcc Jv */
6493 if (dflag != MO_16) {
6494 tval = (int32_t)insn_get(env, s, MO_32);
6495 } else {
6496 tval = (int16_t)insn_get(env, s, MO_16);
6498 do_jcc:
6499 next_eip = s->pc - s->cs_base;
6500 tval += next_eip;
6501 if (dflag == MO_16) {
6502 tval &= 0xffff;
6504 gen_jcc(s, b, tval, next_eip);
6505 break;
6507 case 0x190 ... 0x19f: /* setcc Gv */
6508 modrm = cpu_ldub_code(env, s->pc++);
6509 gen_setcc1(s, b, cpu_T[0]);
6510 gen_ldst_modrm(env, s, modrm, MO_8, OR_TMP0, 1);
6511 break;
6512 case 0x140 ... 0x14f: /* cmov Gv, Ev */
6513 if (!(s->cpuid_features & CPUID_CMOV)) {
6514 goto illegal_op;
6516 ot = dflag;
6517 modrm = cpu_ldub_code(env, s->pc++);
6518 reg = ((modrm >> 3) & 7) | rex_r;
6519 gen_cmovcc1(env, s, ot, b, modrm, reg);
6520 break;
6522 /************************/
6523 /* flags */
6524 case 0x9c: /* pushf */
6525 gen_svm_check_intercept(s, pc_start, SVM_EXIT_PUSHF);
6526 if (s->vm86 && s->iopl != 3) {
6527 gen_exception(s, EXCP0D_GPF, pc_start - s->cs_base);
6528 } else {
6529 gen_update_cc_op(s);
6530 gen_helper_read_eflags(cpu_T[0], cpu_env);
6531 gen_push_v(s, cpu_T[0]);
6533 break;
6534 case 0x9d: /* popf */
6535 gen_svm_check_intercept(s, pc_start, SVM_EXIT_POPF);
6536 if (s->vm86 && s->iopl != 3) {
6537 gen_exception(s, EXCP0D_GPF, pc_start - s->cs_base);
6538 } else {
6539 ot = gen_pop_T0(s);
6540 if (s->cpl == 0) {
6541 if (dflag != MO_16) {
6542 gen_helper_write_eflags(cpu_env, cpu_T[0],
6543 tcg_const_i32((TF_MASK | AC_MASK |
6544 ID_MASK | NT_MASK |
6545 IF_MASK |
6546 IOPL_MASK)));
6547 } else {
6548 gen_helper_write_eflags(cpu_env, cpu_T[0],
6549 tcg_const_i32((TF_MASK | AC_MASK |
6550 ID_MASK | NT_MASK |
6551 IF_MASK | IOPL_MASK)
6552 & 0xffff));
6554 } else {
6555 if (s->cpl <= s->iopl) {
6556 if (dflag != MO_16) {
6557 gen_helper_write_eflags(cpu_env, cpu_T[0],
6558 tcg_const_i32((TF_MASK |
6559 AC_MASK |
6560 ID_MASK |
6561 NT_MASK |
6562 IF_MASK)));
6563 } else {
6564 gen_helper_write_eflags(cpu_env, cpu_T[0],
6565 tcg_const_i32((TF_MASK |
6566 AC_MASK |
6567 ID_MASK |
6568 NT_MASK |
6569 IF_MASK)
6570 & 0xffff));
6572 } else {
6573 if (dflag != MO_16) {
6574 gen_helper_write_eflags(cpu_env, cpu_T[0],
6575 tcg_const_i32((TF_MASK | AC_MASK |
6576 ID_MASK | NT_MASK)));
6577 } else {
6578 gen_helper_write_eflags(cpu_env, cpu_T[0],
6579 tcg_const_i32((TF_MASK | AC_MASK |
6580 ID_MASK | NT_MASK)
6581 & 0xffff));
6585 gen_pop_update(s, ot);
6586 set_cc_op(s, CC_OP_EFLAGS);
6587 /* abort translation because TF/AC flag may change */
6588 gen_jmp_im(s->pc - s->cs_base);
6589 gen_eob(s);
6591 break;
6592 case 0x9e: /* sahf */
6593 if (CODE64(s) && !(s->cpuid_ext3_features & CPUID_EXT3_LAHF_LM))
6594 goto illegal_op;
6595 gen_op_mov_v_reg(MO_8, cpu_T[0], R_AH);
6596 gen_compute_eflags(s);
6597 tcg_gen_andi_tl(cpu_cc_src, cpu_cc_src, CC_O);
6598 tcg_gen_andi_tl(cpu_T[0], cpu_T[0], CC_S | CC_Z | CC_A | CC_P | CC_C);
6599 tcg_gen_or_tl(cpu_cc_src, cpu_cc_src, cpu_T[0]);
6600 break;
6601 case 0x9f: /* lahf */
6602 if (CODE64(s) && !(s->cpuid_ext3_features & CPUID_EXT3_LAHF_LM))
6603 goto illegal_op;
6604 gen_compute_eflags(s);
6605 /* Note: gen_compute_eflags() only gives the condition codes */
6606 tcg_gen_ori_tl(cpu_T[0], cpu_cc_src, 0x02);
6607 gen_op_mov_reg_v(MO_8, R_AH, cpu_T[0]);
6608 break;
6609 case 0xf5: /* cmc */
6610 gen_compute_eflags(s);
6611 tcg_gen_xori_tl(cpu_cc_src, cpu_cc_src, CC_C);
6612 break;
6613 case 0xf8: /* clc */
6614 gen_compute_eflags(s);
6615 tcg_gen_andi_tl(cpu_cc_src, cpu_cc_src, ~CC_C);
6616 break;
6617 case 0xf9: /* stc */
6618 gen_compute_eflags(s);
6619 tcg_gen_ori_tl(cpu_cc_src, cpu_cc_src, CC_C);
6620 break;
6621 case 0xfc: /* cld */
6622 tcg_gen_movi_i32(cpu_tmp2_i32, 1);
6623 tcg_gen_st_i32(cpu_tmp2_i32, cpu_env, offsetof(CPUX86State, df));
6624 break;
6625 case 0xfd: /* std */
6626 tcg_gen_movi_i32(cpu_tmp2_i32, -1);
6627 tcg_gen_st_i32(cpu_tmp2_i32, cpu_env, offsetof(CPUX86State, df));
6628 break;
6630 /************************/
6631 /* bit operations */
6632 case 0x1ba: /* bt/bts/btr/btc Gv, im */
6633 ot = dflag;
6634 modrm = cpu_ldub_code(env, s->pc++);
6635 op = (modrm >> 3) & 7;
6636 mod = (modrm >> 6) & 3;
6637 rm = (modrm & 7) | REX_B(s);
6638 if (mod != 3) {
6639 s->rip_offset = 1;
6640 gen_lea_modrm(env, s, modrm);
6641 gen_op_ld_v(s, ot, cpu_T[0], cpu_A0);
6642 } else {
6643 gen_op_mov_v_reg(ot, cpu_T[0], rm);
6645 /* load shift */
6646 val = cpu_ldub_code(env, s->pc++);
6647 tcg_gen_movi_tl(cpu_T[1], val);
6648 if (op < 4)
6649 goto illegal_op;
6650 op -= 4;
6651 goto bt_op;
6652 case 0x1a3: /* bt Gv, Ev */
6653 op = 0;
6654 goto do_btx;
6655 case 0x1ab: /* bts */
6656 op = 1;
6657 goto do_btx;
6658 case 0x1b3: /* btr */
6659 op = 2;
6660 goto do_btx;
6661 case 0x1bb: /* btc */
6662 op = 3;
6663 do_btx:
6664 ot = dflag;
6665 modrm = cpu_ldub_code(env, s->pc++);
6666 reg = ((modrm >> 3) & 7) | rex_r;
6667 mod = (modrm >> 6) & 3;
6668 rm = (modrm & 7) | REX_B(s);
6669 gen_op_mov_v_reg(MO_32, cpu_T[1], reg);
6670 if (mod != 3) {
6671 gen_lea_modrm(env, s, modrm);
6672 /* specific case: we need to add a displacement */
6673 gen_exts(ot, cpu_T[1]);
6674 tcg_gen_sari_tl(cpu_tmp0, cpu_T[1], 3 + ot);
6675 tcg_gen_shli_tl(cpu_tmp0, cpu_tmp0, ot);
6676 tcg_gen_add_tl(cpu_A0, cpu_A0, cpu_tmp0);
6677 gen_op_ld_v(s, ot, cpu_T[0], cpu_A0);
6678 } else {
6679 gen_op_mov_v_reg(ot, cpu_T[0], rm);
6681 bt_op:
6682 tcg_gen_andi_tl(cpu_T[1], cpu_T[1], (1 << (3 + ot)) - 1);
6683 switch(op) {
6684 case 0:
6685 tcg_gen_shr_tl(cpu_cc_src, cpu_T[0], cpu_T[1]);
6686 tcg_gen_movi_tl(cpu_cc_dst, 0);
6687 break;
6688 case 1:
6689 tcg_gen_shr_tl(cpu_tmp4, cpu_T[0], cpu_T[1]);
6690 tcg_gen_movi_tl(cpu_tmp0, 1);
6691 tcg_gen_shl_tl(cpu_tmp0, cpu_tmp0, cpu_T[1]);
6692 tcg_gen_or_tl(cpu_T[0], cpu_T[0], cpu_tmp0);
6693 break;
6694 case 2:
6695 tcg_gen_shr_tl(cpu_tmp4, cpu_T[0], cpu_T[1]);
6696 tcg_gen_movi_tl(cpu_tmp0, 1);
6697 tcg_gen_shl_tl(cpu_tmp0, cpu_tmp0, cpu_T[1]);
6698 tcg_gen_not_tl(cpu_tmp0, cpu_tmp0);
6699 tcg_gen_and_tl(cpu_T[0], cpu_T[0], cpu_tmp0);
6700 break;
6701 default:
6702 case 3:
6703 tcg_gen_shr_tl(cpu_tmp4, cpu_T[0], cpu_T[1]);
6704 tcg_gen_movi_tl(cpu_tmp0, 1);
6705 tcg_gen_shl_tl(cpu_tmp0, cpu_tmp0, cpu_T[1]);
6706 tcg_gen_xor_tl(cpu_T[0], cpu_T[0], cpu_tmp0);
6707 break;
6709 set_cc_op(s, CC_OP_SARB + ot);
6710 if (op != 0) {
6711 if (mod != 3) {
6712 gen_op_st_v(s, ot, cpu_T[0], cpu_A0);
6713 } else {
6714 gen_op_mov_reg_v(ot, rm, cpu_T[0]);
6716 tcg_gen_mov_tl(cpu_cc_src, cpu_tmp4);
6717 tcg_gen_movi_tl(cpu_cc_dst, 0);
6719 break;
6720 case 0x1bc: /* bsf / tzcnt */
6721 case 0x1bd: /* bsr / lzcnt */
6722 ot = dflag;
6723 modrm = cpu_ldub_code(env, s->pc++);
6724 reg = ((modrm >> 3) & 7) | rex_r;
6725 gen_ldst_modrm(env, s, modrm, ot, OR_TMP0, 0);
6726 gen_extu(ot, cpu_T[0]);
6728 /* Note that lzcnt and tzcnt are in different extensions. */
6729 if ((prefixes & PREFIX_REPZ)
6730 && (b & 1
6731 ? s->cpuid_ext3_features & CPUID_EXT3_ABM
6732 : s->cpuid_7_0_ebx_features & CPUID_7_0_EBX_BMI1)) {
6733 int size = 8 << ot;
6734 tcg_gen_mov_tl(cpu_cc_src, cpu_T[0]);
6735 if (b & 1) {
6736 /* For lzcnt, reduce the target_ulong result by the
6737 number of zeros that we expect to find at the top. */
6738 gen_helper_clz(cpu_T[0], cpu_T[0]);
6739 tcg_gen_subi_tl(cpu_T[0], cpu_T[0], TARGET_LONG_BITS - size);
6740 } else {
6741 /* For tzcnt, a zero input must return the operand size:
6742 force all bits outside the operand size to 1. */
6743 target_ulong mask = (target_ulong)-2 << (size - 1);
6744 tcg_gen_ori_tl(cpu_T[0], cpu_T[0], mask);
6745 gen_helper_ctz(cpu_T[0], cpu_T[0]);
6747 /* For lzcnt/tzcnt, C and Z bits are defined and are
6748 related to the result. */
6749 gen_op_update1_cc();
6750 set_cc_op(s, CC_OP_BMILGB + ot);
6751 } else {
6752 /* For bsr/bsf, only the Z bit is defined and it is related
6753 to the input and not the result. */
6754 tcg_gen_mov_tl(cpu_cc_dst, cpu_T[0]);
6755 set_cc_op(s, CC_OP_LOGICB + ot);
6756 if (b & 1) {
6757 /* For bsr, return the bit index of the first 1 bit,
6758 not the count of leading zeros. */
6759 gen_helper_clz(cpu_T[0], cpu_T[0]);
6760 tcg_gen_xori_tl(cpu_T[0], cpu_T[0], TARGET_LONG_BITS - 1);
6761 } else {
6762 gen_helper_ctz(cpu_T[0], cpu_T[0]);
6764 /* ??? The manual says that the output is undefined when the
6765 input is zero, but real hardware leaves it unchanged, and
6766 real programs appear to depend on that. */
6767 tcg_gen_movi_tl(cpu_tmp0, 0);
6768 tcg_gen_movcond_tl(TCG_COND_EQ, cpu_T[0], cpu_cc_dst, cpu_tmp0,
6769 cpu_regs[reg], cpu_T[0]);
6771 gen_op_mov_reg_v(ot, reg, cpu_T[0]);
6772 break;
6773 /************************/
6774 /* bcd */
6775 case 0x27: /* daa */
6776 if (CODE64(s))
6777 goto illegal_op;
6778 gen_update_cc_op(s);
6779 gen_helper_daa(cpu_env);
6780 set_cc_op(s, CC_OP_EFLAGS);
6781 break;
6782 case 0x2f: /* das */
6783 if (CODE64(s))
6784 goto illegal_op;
6785 gen_update_cc_op(s);
6786 gen_helper_das(cpu_env);
6787 set_cc_op(s, CC_OP_EFLAGS);
6788 break;
6789 case 0x37: /* aaa */
6790 if (CODE64(s))
6791 goto illegal_op;
6792 gen_update_cc_op(s);
6793 gen_helper_aaa(cpu_env);
6794 set_cc_op(s, CC_OP_EFLAGS);
6795 break;
6796 case 0x3f: /* aas */
6797 if (CODE64(s))
6798 goto illegal_op;
6799 gen_update_cc_op(s);
6800 gen_helper_aas(cpu_env);
6801 set_cc_op(s, CC_OP_EFLAGS);
6802 break;
6803 case 0xd4: /* aam */
6804 if (CODE64(s))
6805 goto illegal_op;
6806 val = cpu_ldub_code(env, s->pc++);
6807 if (val == 0) {
6808 gen_exception(s, EXCP00_DIVZ, pc_start - s->cs_base);
6809 } else {
6810 gen_helper_aam(cpu_env, tcg_const_i32(val));
6811 set_cc_op(s, CC_OP_LOGICB);
6813 break;
6814 case 0xd5: /* aad */
6815 if (CODE64(s))
6816 goto illegal_op;
6817 val = cpu_ldub_code(env, s->pc++);
6818 gen_helper_aad(cpu_env, tcg_const_i32(val));
6819 set_cc_op(s, CC_OP_LOGICB);
6820 break;
6821 /************************/
6822 /* misc */
6823 case 0x90: /* nop */
6824 /* XXX: correct lock test for all insn */
6825 if (prefixes & PREFIX_LOCK) {
6826 goto illegal_op;
6828 /* If REX_B is set, then this is xchg eax, r8d, not a nop. */
6829 if (REX_B(s)) {
6830 goto do_xchg_reg_eax;
6832 if (prefixes & PREFIX_REPZ) {
6833 gen_update_cc_op(s);
6834 gen_jmp_im(pc_start - s->cs_base);
6835 gen_helper_pause(cpu_env, tcg_const_i32(s->pc - pc_start));
6836 s->is_jmp = DISAS_TB_JUMP;
6838 break;
6839 case 0x9b: /* fwait */
6840 if ((s->flags & (HF_MP_MASK | HF_TS_MASK)) ==
6841 (HF_MP_MASK | HF_TS_MASK)) {
6842 gen_exception(s, EXCP07_PREX, pc_start - s->cs_base);
6843 } else {
6844 gen_update_cc_op(s);
6845 gen_jmp_im(pc_start - s->cs_base);
6846 gen_helper_fwait(cpu_env);
6848 break;
6849 case 0xcc: /* int3 */
6850 gen_interrupt(s, EXCP03_INT3, pc_start - s->cs_base, s->pc - s->cs_base);
6851 break;
6852 case 0xcd: /* int N */
6853 val = cpu_ldub_code(env, s->pc++);
6854 if (s->vm86 && s->iopl != 3) {
6855 gen_exception(s, EXCP0D_GPF, pc_start - s->cs_base);
6856 } else {
6857 gen_interrupt(s, val, pc_start - s->cs_base, s->pc - s->cs_base);
6859 break;
6860 case 0xce: /* into */
6861 if (CODE64(s))
6862 goto illegal_op;
6863 gen_update_cc_op(s);
6864 gen_jmp_im(pc_start - s->cs_base);
6865 gen_helper_into(cpu_env, tcg_const_i32(s->pc - pc_start));
6866 break;
6867 #ifdef WANT_ICEBP
6868 case 0xf1: /* icebp (undocumented, exits to external debugger) */
6869 gen_svm_check_intercept(s, pc_start, SVM_EXIT_ICEBP);
6870 #if 1
6871 gen_debug(s, pc_start - s->cs_base);
6872 #else
6873 /* start debug */
6874 tb_flush(env);
6875 qemu_set_log(CPU_LOG_INT | CPU_LOG_TB_IN_ASM);
6876 #endif
6877 break;
6878 #endif
6879 case 0xfa: /* cli */
6880 if (!s->vm86) {
6881 if (s->cpl <= s->iopl) {
6882 gen_helper_cli(cpu_env);
6883 } else {
6884 gen_exception(s, EXCP0D_GPF, pc_start - s->cs_base);
6886 } else {
6887 if (s->iopl == 3) {
6888 gen_helper_cli(cpu_env);
6889 } else {
6890 gen_exception(s, EXCP0D_GPF, pc_start - s->cs_base);
6893 break;
6894 case 0xfb: /* sti */
6895 if (!s->vm86) {
6896 if (s->cpl <= s->iopl) {
6897 gen_sti:
6898 gen_helper_sti(cpu_env);
6899 /* interruptions are enabled only the first insn after sti */
6900 /* If several instructions disable interrupts, only the
6901 _first_ does it */
6902 if (!(s->tb->flags & HF_INHIBIT_IRQ_MASK))
6903 gen_helper_set_inhibit_irq(cpu_env);
6904 /* give a chance to handle pending irqs */
6905 gen_jmp_im(s->pc - s->cs_base);
6906 gen_eob(s);
6907 } else {
6908 gen_exception(s, EXCP0D_GPF, pc_start - s->cs_base);
6910 } else {
6911 if (s->iopl == 3) {
6912 goto gen_sti;
6913 } else {
6914 gen_exception(s, EXCP0D_GPF, pc_start - s->cs_base);
6917 break;
6918 case 0x62: /* bound */
6919 if (CODE64(s))
6920 goto illegal_op;
6921 ot = dflag;
6922 modrm = cpu_ldub_code(env, s->pc++);
6923 reg = (modrm >> 3) & 7;
6924 mod = (modrm >> 6) & 3;
6925 if (mod == 3)
6926 goto illegal_op;
6927 gen_op_mov_v_reg(ot, cpu_T[0], reg);
6928 gen_lea_modrm(env, s, modrm);
6929 gen_jmp_im(pc_start - s->cs_base);
6930 tcg_gen_trunc_tl_i32(cpu_tmp2_i32, cpu_T[0]);
6931 if (ot == MO_16) {
6932 gen_helper_boundw(cpu_env, cpu_A0, cpu_tmp2_i32);
6933 } else {
6934 gen_helper_boundl(cpu_env, cpu_A0, cpu_tmp2_i32);
6936 break;
6937 case 0x1c8 ... 0x1cf: /* bswap reg */
6938 reg = (b & 7) | REX_B(s);
6939 #ifdef TARGET_X86_64
6940 if (dflag == MO_64) {
6941 gen_op_mov_v_reg(MO_64, cpu_T[0], reg);
6942 tcg_gen_bswap64_i64(cpu_T[0], cpu_T[0]);
6943 gen_op_mov_reg_v(MO_64, reg, cpu_T[0]);
6944 } else
6945 #endif
6947 gen_op_mov_v_reg(MO_32, cpu_T[0], reg);
6948 tcg_gen_ext32u_tl(cpu_T[0], cpu_T[0]);
6949 tcg_gen_bswap32_tl(cpu_T[0], cpu_T[0]);
6950 gen_op_mov_reg_v(MO_32, reg, cpu_T[0]);
6952 break;
6953 case 0xd6: /* salc */
6954 if (CODE64(s))
6955 goto illegal_op;
6956 gen_compute_eflags_c(s, cpu_T[0]);
6957 tcg_gen_neg_tl(cpu_T[0], cpu_T[0]);
6958 gen_op_mov_reg_v(MO_8, R_EAX, cpu_T[0]);
6959 break;
6960 case 0xe0: /* loopnz */
6961 case 0xe1: /* loopz */
6962 case 0xe2: /* loop */
6963 case 0xe3: /* jecxz */
6965 int l1, l2, l3;
6967 tval = (int8_t)insn_get(env, s, MO_8);
6968 next_eip = s->pc - s->cs_base;
6969 tval += next_eip;
6970 if (dflag == MO_16) {
6971 tval &= 0xffff;
6974 l1 = gen_new_label();
6975 l2 = gen_new_label();
6976 l3 = gen_new_label();
6977 b &= 3;
6978 switch(b) {
6979 case 0: /* loopnz */
6980 case 1: /* loopz */
6981 gen_op_add_reg_im(s->aflag, R_ECX, -1);
6982 gen_op_jz_ecx(s->aflag, l3);
6983 gen_jcc1(s, (JCC_Z << 1) | (b ^ 1), l1);
6984 break;
6985 case 2: /* loop */
6986 gen_op_add_reg_im(s->aflag, R_ECX, -1);
6987 gen_op_jnz_ecx(s->aflag, l1);
6988 break;
6989 default:
6990 case 3: /* jcxz */
6991 gen_op_jz_ecx(s->aflag, l1);
6992 break;
6995 gen_set_label(l3);
6996 gen_jmp_im(next_eip);
6997 tcg_gen_br(l2);
6999 gen_set_label(l1);
7000 gen_jmp_im(tval);
7001 gen_set_label(l2);
7002 gen_eob(s);
7004 break;
7005 case 0x130: /* wrmsr */
7006 case 0x132: /* rdmsr */
7007 if (s->cpl != 0) {
7008 gen_exception(s, EXCP0D_GPF, pc_start - s->cs_base);
7009 } else {
7010 gen_update_cc_op(s);
7011 gen_jmp_im(pc_start - s->cs_base);
7012 if (b & 2) {
7013 gen_helper_rdmsr(cpu_env);
7014 } else {
7015 gen_helper_wrmsr(cpu_env);
7018 break;
7019 case 0x131: /* rdtsc */
7020 gen_update_cc_op(s);
7021 gen_jmp_im(pc_start - s->cs_base);
7022 if (use_icount)
7023 gen_io_start();
7024 gen_helper_rdtsc(cpu_env);
7025 if (use_icount) {
7026 gen_io_end();
7027 gen_jmp(s, s->pc - s->cs_base);
7029 break;
7030 case 0x133: /* rdpmc */
7031 gen_update_cc_op(s);
7032 gen_jmp_im(pc_start - s->cs_base);
7033 gen_helper_rdpmc(cpu_env);
7034 break;
7035 case 0x134: /* sysenter */
7036 /* For Intel SYSENTER is valid on 64-bit */
7037 if (CODE64(s) && env->cpuid_vendor1 != CPUID_VENDOR_INTEL_1)
7038 goto illegal_op;
7039 if (!s->pe) {
7040 gen_exception(s, EXCP0D_GPF, pc_start - s->cs_base);
7041 } else {
7042 gen_update_cc_op(s);
7043 gen_jmp_im(pc_start - s->cs_base);
7044 gen_helper_sysenter(cpu_env);
7045 gen_eob(s);
7047 break;
7048 case 0x135: /* sysexit */
7049 /* For Intel SYSEXIT is valid on 64-bit */
7050 if (CODE64(s) && env->cpuid_vendor1 != CPUID_VENDOR_INTEL_1)
7051 goto illegal_op;
7052 if (!s->pe) {
7053 gen_exception(s, EXCP0D_GPF, pc_start - s->cs_base);
7054 } else {
7055 gen_update_cc_op(s);
7056 gen_jmp_im(pc_start - s->cs_base);
7057 gen_helper_sysexit(cpu_env, tcg_const_i32(dflag - 1));
7058 gen_eob(s);
7060 break;
7061 #ifdef TARGET_X86_64
7062 case 0x105: /* syscall */
7063 /* XXX: is it usable in real mode ? */
7064 gen_update_cc_op(s);
7065 gen_jmp_im(pc_start - s->cs_base);
7066 gen_helper_syscall(cpu_env, tcg_const_i32(s->pc - pc_start));
7067 gen_eob(s);
7068 break;
7069 case 0x107: /* sysret */
7070 if (!s->pe) {
7071 gen_exception(s, EXCP0D_GPF, pc_start - s->cs_base);
7072 } else {
7073 gen_update_cc_op(s);
7074 gen_jmp_im(pc_start - s->cs_base);
7075 gen_helper_sysret(cpu_env, tcg_const_i32(dflag - 1));
7076 /* condition codes are modified only in long mode */
7077 if (s->lma) {
7078 set_cc_op(s, CC_OP_EFLAGS);
7080 gen_eob(s);
7082 break;
7083 #endif
7084 case 0x1a2: /* cpuid */
7085 gen_update_cc_op(s);
7086 gen_jmp_im(pc_start - s->cs_base);
7087 gen_helper_cpuid(cpu_env);
7088 break;
7089 case 0xf4: /* hlt */
7090 if (s->cpl != 0) {
7091 gen_exception(s, EXCP0D_GPF, pc_start - s->cs_base);
7092 } else {
7093 gen_update_cc_op(s);
7094 gen_jmp_im(pc_start - s->cs_base);
7095 gen_helper_hlt(cpu_env, tcg_const_i32(s->pc - pc_start));
7096 s->is_jmp = DISAS_TB_JUMP;
7098 break;
7099 case 0x100:
7100 modrm = cpu_ldub_code(env, s->pc++);
7101 mod = (modrm >> 6) & 3;
7102 op = (modrm >> 3) & 7;
7103 switch(op) {
7104 case 0: /* sldt */
7105 if (!s->pe || s->vm86)
7106 goto illegal_op;
7107 gen_svm_check_intercept(s, pc_start, SVM_EXIT_LDTR_READ);
7108 tcg_gen_ld32u_tl(cpu_T[0], cpu_env, offsetof(CPUX86State,ldt.selector));
7109 ot = mod == 3 ? dflag : MO_16;
7110 gen_ldst_modrm(env, s, modrm, ot, OR_TMP0, 1);
7111 break;
7112 case 2: /* lldt */
7113 if (!s->pe || s->vm86)
7114 goto illegal_op;
7115 if (s->cpl != 0) {
7116 gen_exception(s, EXCP0D_GPF, pc_start - s->cs_base);
7117 } else {
7118 gen_svm_check_intercept(s, pc_start, SVM_EXIT_LDTR_WRITE);
7119 gen_ldst_modrm(env, s, modrm, MO_16, OR_TMP0, 0);
7120 gen_jmp_im(pc_start - s->cs_base);
7121 tcg_gen_trunc_tl_i32(cpu_tmp2_i32, cpu_T[0]);
7122 gen_helper_lldt(cpu_env, cpu_tmp2_i32);
7124 break;
7125 case 1: /* str */
7126 if (!s->pe || s->vm86)
7127 goto illegal_op;
7128 gen_svm_check_intercept(s, pc_start, SVM_EXIT_TR_READ);
7129 tcg_gen_ld32u_tl(cpu_T[0], cpu_env, offsetof(CPUX86State,tr.selector));
7130 ot = mod == 3 ? dflag : MO_16;
7131 gen_ldst_modrm(env, s, modrm, ot, OR_TMP0, 1);
7132 break;
7133 case 3: /* ltr */
7134 if (!s->pe || s->vm86)
7135 goto illegal_op;
7136 if (s->cpl != 0) {
7137 gen_exception(s, EXCP0D_GPF, pc_start - s->cs_base);
7138 } else {
7139 gen_svm_check_intercept(s, pc_start, SVM_EXIT_TR_WRITE);
7140 gen_ldst_modrm(env, s, modrm, MO_16, OR_TMP0, 0);
7141 gen_jmp_im(pc_start - s->cs_base);
7142 tcg_gen_trunc_tl_i32(cpu_tmp2_i32, cpu_T[0]);
7143 gen_helper_ltr(cpu_env, cpu_tmp2_i32);
7145 break;
7146 case 4: /* verr */
7147 case 5: /* verw */
7148 if (!s->pe || s->vm86)
7149 goto illegal_op;
7150 gen_ldst_modrm(env, s, modrm, MO_16, OR_TMP0, 0);
7151 gen_update_cc_op(s);
7152 if (op == 4) {
7153 gen_helper_verr(cpu_env, cpu_T[0]);
7154 } else {
7155 gen_helper_verw(cpu_env, cpu_T[0]);
7157 set_cc_op(s, CC_OP_EFLAGS);
7158 break;
7159 default:
7160 goto illegal_op;
7162 break;
7163 case 0x101:
7164 modrm = cpu_ldub_code(env, s->pc++);
7165 mod = (modrm >> 6) & 3;
7166 op = (modrm >> 3) & 7;
7167 rm = modrm & 7;
7168 switch(op) {
7169 case 0: /* sgdt */
7170 if (mod == 3)
7171 goto illegal_op;
7172 gen_svm_check_intercept(s, pc_start, SVM_EXIT_GDTR_READ);
7173 gen_lea_modrm(env, s, modrm);
7174 tcg_gen_ld32u_tl(cpu_T[0], cpu_env, offsetof(CPUX86State, gdt.limit));
7175 gen_op_st_v(s, MO_16, cpu_T[0], cpu_A0);
7176 gen_add_A0_im(s, 2);
7177 tcg_gen_ld_tl(cpu_T[0], cpu_env, offsetof(CPUX86State, gdt.base));
7178 if (dflag == MO_16) {
7179 tcg_gen_andi_tl(cpu_T[0], cpu_T[0], 0xffffff);
7181 gen_op_st_v(s, CODE64(s) + MO_32, cpu_T[0], cpu_A0);
7182 break;
7183 case 1:
7184 if (mod == 3) {
7185 switch (rm) {
7186 case 0: /* monitor */
7187 if (!(s->cpuid_ext_features & CPUID_EXT_MONITOR) ||
7188 s->cpl != 0)
7189 goto illegal_op;
7190 gen_update_cc_op(s);
7191 gen_jmp_im(pc_start - s->cs_base);
7192 tcg_gen_mov_tl(cpu_A0, cpu_regs[R_EAX]);
7193 gen_extu(s->aflag, cpu_A0);
7194 gen_add_A0_ds_seg(s);
7195 gen_helper_monitor(cpu_env, cpu_A0);
7196 break;
7197 case 1: /* mwait */
7198 if (!(s->cpuid_ext_features & CPUID_EXT_MONITOR) ||
7199 s->cpl != 0)
7200 goto illegal_op;
7201 gen_update_cc_op(s);
7202 gen_jmp_im(pc_start - s->cs_base);
7203 gen_helper_mwait(cpu_env, tcg_const_i32(s->pc - pc_start));
7204 gen_eob(s);
7205 break;
7206 case 2: /* clac */
7207 if (!(s->cpuid_7_0_ebx_features & CPUID_7_0_EBX_SMAP) ||
7208 s->cpl != 0) {
7209 goto illegal_op;
7211 gen_helper_clac(cpu_env);
7212 gen_jmp_im(s->pc - s->cs_base);
7213 gen_eob(s);
7214 break;
7215 case 3: /* stac */
7216 if (!(s->cpuid_7_0_ebx_features & CPUID_7_0_EBX_SMAP) ||
7217 s->cpl != 0) {
7218 goto illegal_op;
7220 gen_helper_stac(cpu_env);
7221 gen_jmp_im(s->pc - s->cs_base);
7222 gen_eob(s);
7223 break;
7224 default:
7225 goto illegal_op;
7227 } else { /* sidt */
7228 gen_svm_check_intercept(s, pc_start, SVM_EXIT_IDTR_READ);
7229 gen_lea_modrm(env, s, modrm);
7230 tcg_gen_ld32u_tl(cpu_T[0], cpu_env, offsetof(CPUX86State, idt.limit));
7231 gen_op_st_v(s, MO_16, cpu_T[0], cpu_A0);
7232 gen_add_A0_im(s, 2);
7233 tcg_gen_ld_tl(cpu_T[0], cpu_env, offsetof(CPUX86State, idt.base));
7234 if (dflag == MO_16) {
7235 tcg_gen_andi_tl(cpu_T[0], cpu_T[0], 0xffffff);
7237 gen_op_st_v(s, CODE64(s) + MO_32, cpu_T[0], cpu_A0);
7239 break;
7240 case 2: /* lgdt */
7241 case 3: /* lidt */
7242 if (mod == 3) {
7243 gen_update_cc_op(s);
7244 gen_jmp_im(pc_start - s->cs_base);
7245 switch(rm) {
7246 case 0: /* VMRUN */
7247 if (!(s->flags & HF_SVME_MASK) || !s->pe)
7248 goto illegal_op;
7249 if (s->cpl != 0) {
7250 gen_exception(s, EXCP0D_GPF, pc_start - s->cs_base);
7251 break;
7252 } else {
7253 gen_helper_vmrun(cpu_env, tcg_const_i32(s->aflag - 1),
7254 tcg_const_i32(s->pc - pc_start));
7255 tcg_gen_exit_tb(0);
7256 s->is_jmp = DISAS_TB_JUMP;
7258 break;
7259 case 1: /* VMMCALL */
7260 if (!(s->flags & HF_SVME_MASK))
7261 goto illegal_op;
7262 gen_helper_vmmcall(cpu_env);
7263 break;
7264 case 2: /* VMLOAD */
7265 if (!(s->flags & HF_SVME_MASK) || !s->pe)
7266 goto illegal_op;
7267 if (s->cpl != 0) {
7268 gen_exception(s, EXCP0D_GPF, pc_start - s->cs_base);
7269 break;
7270 } else {
7271 gen_helper_vmload(cpu_env, tcg_const_i32(s->aflag - 1));
7273 break;
7274 case 3: /* VMSAVE */
7275 if (!(s->flags & HF_SVME_MASK) || !s->pe)
7276 goto illegal_op;
7277 if (s->cpl != 0) {
7278 gen_exception(s, EXCP0D_GPF, pc_start - s->cs_base);
7279 break;
7280 } else {
7281 gen_helper_vmsave(cpu_env, tcg_const_i32(s->aflag - 1));
7283 break;
7284 case 4: /* STGI */
7285 if ((!(s->flags & HF_SVME_MASK) &&
7286 !(s->cpuid_ext3_features & CPUID_EXT3_SKINIT)) ||
7287 !s->pe)
7288 goto illegal_op;
7289 if (s->cpl != 0) {
7290 gen_exception(s, EXCP0D_GPF, pc_start - s->cs_base);
7291 break;
7292 } else {
7293 gen_helper_stgi(cpu_env);
7295 break;
7296 case 5: /* CLGI */
7297 if (!(s->flags & HF_SVME_MASK) || !s->pe)
7298 goto illegal_op;
7299 if (s->cpl != 0) {
7300 gen_exception(s, EXCP0D_GPF, pc_start - s->cs_base);
7301 break;
7302 } else {
7303 gen_helper_clgi(cpu_env);
7305 break;
7306 case 6: /* SKINIT */
7307 if ((!(s->flags & HF_SVME_MASK) &&
7308 !(s->cpuid_ext3_features & CPUID_EXT3_SKINIT)) ||
7309 !s->pe)
7310 goto illegal_op;
7311 gen_helper_skinit(cpu_env);
7312 break;
7313 case 7: /* INVLPGA */
7314 if (!(s->flags & HF_SVME_MASK) || !s->pe)
7315 goto illegal_op;
7316 if (s->cpl != 0) {
7317 gen_exception(s, EXCP0D_GPF, pc_start - s->cs_base);
7318 break;
7319 } else {
7320 gen_helper_invlpga(cpu_env,
7321 tcg_const_i32(s->aflag - 1));
7323 break;
7324 default:
7325 goto illegal_op;
7327 } else if (s->cpl != 0) {
7328 gen_exception(s, EXCP0D_GPF, pc_start - s->cs_base);
7329 } else {
7330 gen_svm_check_intercept(s, pc_start,
7331 op==2 ? SVM_EXIT_GDTR_WRITE : SVM_EXIT_IDTR_WRITE);
7332 gen_lea_modrm(env, s, modrm);
7333 gen_op_ld_v(s, MO_16, cpu_T[1], cpu_A0);
7334 gen_add_A0_im(s, 2);
7335 gen_op_ld_v(s, CODE64(s) + MO_32, cpu_T[0], cpu_A0);
7336 if (dflag == MO_16) {
7337 tcg_gen_andi_tl(cpu_T[0], cpu_T[0], 0xffffff);
7339 if (op == 2) {
7340 tcg_gen_st_tl(cpu_T[0], cpu_env, offsetof(CPUX86State,gdt.base));
7341 tcg_gen_st32_tl(cpu_T[1], cpu_env, offsetof(CPUX86State,gdt.limit));
7342 } else {
7343 tcg_gen_st_tl(cpu_T[0], cpu_env, offsetof(CPUX86State,idt.base));
7344 tcg_gen_st32_tl(cpu_T[1], cpu_env, offsetof(CPUX86State,idt.limit));
7347 break;
7348 case 4: /* smsw */
7349 gen_svm_check_intercept(s, pc_start, SVM_EXIT_READ_CR0);
7350 #if defined TARGET_X86_64 && defined HOST_WORDS_BIGENDIAN
7351 tcg_gen_ld32u_tl(cpu_T[0], cpu_env, offsetof(CPUX86State,cr[0]) + 4);
7352 #else
7353 tcg_gen_ld32u_tl(cpu_T[0], cpu_env, offsetof(CPUX86State,cr[0]));
7354 #endif
7355 gen_ldst_modrm(env, s, modrm, MO_16, OR_TMP0, 1);
7356 break;
7357 case 6: /* lmsw */
7358 if (s->cpl != 0) {
7359 gen_exception(s, EXCP0D_GPF, pc_start - s->cs_base);
7360 } else {
7361 gen_svm_check_intercept(s, pc_start, SVM_EXIT_WRITE_CR0);
7362 gen_ldst_modrm(env, s, modrm, MO_16, OR_TMP0, 0);
7363 gen_helper_lmsw(cpu_env, cpu_T[0]);
7364 gen_jmp_im(s->pc - s->cs_base);
7365 gen_eob(s);
7367 break;
7368 case 7:
7369 if (mod != 3) { /* invlpg */
7370 if (s->cpl != 0) {
7371 gen_exception(s, EXCP0D_GPF, pc_start - s->cs_base);
7372 } else {
7373 gen_update_cc_op(s);
7374 gen_jmp_im(pc_start - s->cs_base);
7375 gen_lea_modrm(env, s, modrm);
7376 gen_helper_invlpg(cpu_env, cpu_A0);
7377 gen_jmp_im(s->pc - s->cs_base);
7378 gen_eob(s);
7380 } else {
7381 switch (rm) {
7382 case 0: /* swapgs */
7383 #ifdef TARGET_X86_64
7384 if (CODE64(s)) {
7385 if (s->cpl != 0) {
7386 gen_exception(s, EXCP0D_GPF, pc_start - s->cs_base);
7387 } else {
7388 tcg_gen_ld_tl(cpu_T[0], cpu_env,
7389 offsetof(CPUX86State,segs[R_GS].base));
7390 tcg_gen_ld_tl(cpu_T[1], cpu_env,
7391 offsetof(CPUX86State,kernelgsbase));
7392 tcg_gen_st_tl(cpu_T[1], cpu_env,
7393 offsetof(CPUX86State,segs[R_GS].base));
7394 tcg_gen_st_tl(cpu_T[0], cpu_env,
7395 offsetof(CPUX86State,kernelgsbase));
7397 } else
7398 #endif
7400 goto illegal_op;
7402 break;
7403 case 1: /* rdtscp */
7404 if (!(s->cpuid_ext2_features & CPUID_EXT2_RDTSCP))
7405 goto illegal_op;
7406 gen_update_cc_op(s);
7407 gen_jmp_im(pc_start - s->cs_base);
7408 if (use_icount)
7409 gen_io_start();
7410 gen_helper_rdtscp(cpu_env);
7411 if (use_icount) {
7412 gen_io_end();
7413 gen_jmp(s, s->pc - s->cs_base);
7415 break;
7416 default:
7417 goto illegal_op;
7420 break;
7421 default:
7422 goto illegal_op;
7424 break;
7425 case 0x108: /* invd */
7426 case 0x109: /* wbinvd */
7427 if (s->cpl != 0) {
7428 gen_exception(s, EXCP0D_GPF, pc_start - s->cs_base);
7429 } else {
7430 gen_svm_check_intercept(s, pc_start, (b & 2) ? SVM_EXIT_INVD : SVM_EXIT_WBINVD);
7431 /* nothing to do */
7433 break;
7434 case 0x63: /* arpl or movslS (x86_64) */
7435 #ifdef TARGET_X86_64
7436 if (CODE64(s)) {
7437 int d_ot;
7438 /* d_ot is the size of destination */
7439 d_ot = dflag;
7441 modrm = cpu_ldub_code(env, s->pc++);
7442 reg = ((modrm >> 3) & 7) | rex_r;
7443 mod = (modrm >> 6) & 3;
7444 rm = (modrm & 7) | REX_B(s);
7446 if (mod == 3) {
7447 gen_op_mov_v_reg(MO_32, cpu_T[0], rm);
7448 /* sign extend */
7449 if (d_ot == MO_64) {
7450 tcg_gen_ext32s_tl(cpu_T[0], cpu_T[0]);
7452 gen_op_mov_reg_v(d_ot, reg, cpu_T[0]);
7453 } else {
7454 gen_lea_modrm(env, s, modrm);
7455 gen_op_ld_v(s, MO_32 | MO_SIGN, cpu_T[0], cpu_A0);
7456 gen_op_mov_reg_v(d_ot, reg, cpu_T[0]);
7458 } else
7459 #endif
7461 int label1;
7462 TCGv t0, t1, t2, a0;
7464 if (!s->pe || s->vm86)
7465 goto illegal_op;
7466 t0 = tcg_temp_local_new();
7467 t1 = tcg_temp_local_new();
7468 t2 = tcg_temp_local_new();
7469 ot = MO_16;
7470 modrm = cpu_ldub_code(env, s->pc++);
7471 reg = (modrm >> 3) & 7;
7472 mod = (modrm >> 6) & 3;
7473 rm = modrm & 7;
7474 if (mod != 3) {
7475 gen_lea_modrm(env, s, modrm);
7476 gen_op_ld_v(s, ot, t0, cpu_A0);
7477 a0 = tcg_temp_local_new();
7478 tcg_gen_mov_tl(a0, cpu_A0);
7479 } else {
7480 gen_op_mov_v_reg(ot, t0, rm);
7481 TCGV_UNUSED(a0);
7483 gen_op_mov_v_reg(ot, t1, reg);
7484 tcg_gen_andi_tl(cpu_tmp0, t0, 3);
7485 tcg_gen_andi_tl(t1, t1, 3);
7486 tcg_gen_movi_tl(t2, 0);
7487 label1 = gen_new_label();
7488 tcg_gen_brcond_tl(TCG_COND_GE, cpu_tmp0, t1, label1);
7489 tcg_gen_andi_tl(t0, t0, ~3);
7490 tcg_gen_or_tl(t0, t0, t1);
7491 tcg_gen_movi_tl(t2, CC_Z);
7492 gen_set_label(label1);
7493 if (mod != 3) {
7494 gen_op_st_v(s, ot, t0, a0);
7495 tcg_temp_free(a0);
7496 } else {
7497 gen_op_mov_reg_v(ot, rm, t0);
7499 gen_compute_eflags(s);
7500 tcg_gen_andi_tl(cpu_cc_src, cpu_cc_src, ~CC_Z);
7501 tcg_gen_or_tl(cpu_cc_src, cpu_cc_src, t2);
7502 tcg_temp_free(t0);
7503 tcg_temp_free(t1);
7504 tcg_temp_free(t2);
7506 break;
7507 case 0x102: /* lar */
7508 case 0x103: /* lsl */
7510 int label1;
7511 TCGv t0;
7512 if (!s->pe || s->vm86)
7513 goto illegal_op;
7514 ot = dflag != MO_16 ? MO_32 : MO_16;
7515 modrm = cpu_ldub_code(env, s->pc++);
7516 reg = ((modrm >> 3) & 7) | rex_r;
7517 gen_ldst_modrm(env, s, modrm, MO_16, OR_TMP0, 0);
7518 t0 = tcg_temp_local_new();
7519 gen_update_cc_op(s);
7520 if (b == 0x102) {
7521 gen_helper_lar(t0, cpu_env, cpu_T[0]);
7522 } else {
7523 gen_helper_lsl(t0, cpu_env, cpu_T[0]);
7525 tcg_gen_andi_tl(cpu_tmp0, cpu_cc_src, CC_Z);
7526 label1 = gen_new_label();
7527 tcg_gen_brcondi_tl(TCG_COND_EQ, cpu_tmp0, 0, label1);
7528 gen_op_mov_reg_v(ot, reg, t0);
7529 gen_set_label(label1);
7530 set_cc_op(s, CC_OP_EFLAGS);
7531 tcg_temp_free(t0);
7533 break;
7534 case 0x118:
7535 modrm = cpu_ldub_code(env, s->pc++);
7536 mod = (modrm >> 6) & 3;
7537 op = (modrm >> 3) & 7;
7538 switch(op) {
7539 case 0: /* prefetchnta */
7540 case 1: /* prefetchnt0 */
7541 case 2: /* prefetchnt0 */
7542 case 3: /* prefetchnt0 */
7543 if (mod == 3)
7544 goto illegal_op;
7545 gen_lea_modrm(env, s, modrm);
7546 /* nothing more to do */
7547 break;
7548 default: /* nop (multi byte) */
7549 gen_nop_modrm(env, s, modrm);
7550 break;
7552 break;
7553 case 0x119 ... 0x11f: /* nop (multi byte) */
7554 modrm = cpu_ldub_code(env, s->pc++);
7555 gen_nop_modrm(env, s, modrm);
7556 break;
7557 case 0x120: /* mov reg, crN */
7558 case 0x122: /* mov crN, reg */
7559 if (s->cpl != 0) {
7560 gen_exception(s, EXCP0D_GPF, pc_start - s->cs_base);
7561 } else {
7562 modrm = cpu_ldub_code(env, s->pc++);
7563 /* Ignore the mod bits (assume (modrm&0xc0)==0xc0).
7564 * AMD documentation (24594.pdf) and testing of
7565 * intel 386 and 486 processors all show that the mod bits
7566 * are assumed to be 1's, regardless of actual values.
7568 rm = (modrm & 7) | REX_B(s);
7569 reg = ((modrm >> 3) & 7) | rex_r;
7570 if (CODE64(s))
7571 ot = MO_64;
7572 else
7573 ot = MO_32;
7574 if ((prefixes & PREFIX_LOCK) && (reg == 0) &&
7575 (s->cpuid_ext3_features & CPUID_EXT3_CR8LEG)) {
7576 reg = 8;
7578 switch(reg) {
7579 case 0:
7580 case 2:
7581 case 3:
7582 case 4:
7583 case 8:
7584 gen_update_cc_op(s);
7585 gen_jmp_im(pc_start - s->cs_base);
7586 if (b & 2) {
7587 gen_op_mov_v_reg(ot, cpu_T[0], rm);
7588 gen_helper_write_crN(cpu_env, tcg_const_i32(reg),
7589 cpu_T[0]);
7590 gen_jmp_im(s->pc - s->cs_base);
7591 gen_eob(s);
7592 } else {
7593 gen_helper_read_crN(cpu_T[0], cpu_env, tcg_const_i32(reg));
7594 gen_op_mov_reg_v(ot, rm, cpu_T[0]);
7596 break;
7597 default:
7598 goto illegal_op;
7601 break;
7602 case 0x121: /* mov reg, drN */
7603 case 0x123: /* mov drN, reg */
7604 if (s->cpl != 0) {
7605 gen_exception(s, EXCP0D_GPF, pc_start - s->cs_base);
7606 } else {
7607 modrm = cpu_ldub_code(env, s->pc++);
7608 /* Ignore the mod bits (assume (modrm&0xc0)==0xc0).
7609 * AMD documentation (24594.pdf) and testing of
7610 * intel 386 and 486 processors all show that the mod bits
7611 * are assumed to be 1's, regardless of actual values.
7613 rm = (modrm & 7) | REX_B(s);
7614 reg = ((modrm >> 3) & 7) | rex_r;
7615 if (CODE64(s))
7616 ot = MO_64;
7617 else
7618 ot = MO_32;
7619 /* XXX: do it dynamically with CR4.DE bit */
7620 if (reg == 4 || reg == 5 || reg >= 8)
7621 goto illegal_op;
7622 if (b & 2) {
7623 gen_svm_check_intercept(s, pc_start, SVM_EXIT_WRITE_DR0 + reg);
7624 gen_op_mov_v_reg(ot, cpu_T[0], rm);
7625 gen_helper_movl_drN_T0(cpu_env, tcg_const_i32(reg), cpu_T[0]);
7626 gen_jmp_im(s->pc - s->cs_base);
7627 gen_eob(s);
7628 } else {
7629 gen_svm_check_intercept(s, pc_start, SVM_EXIT_READ_DR0 + reg);
7630 tcg_gen_ld_tl(cpu_T[0], cpu_env, offsetof(CPUX86State,dr[reg]));
7631 gen_op_mov_reg_v(ot, rm, cpu_T[0]);
7634 break;
7635 case 0x106: /* clts */
7636 if (s->cpl != 0) {
7637 gen_exception(s, EXCP0D_GPF, pc_start - s->cs_base);
7638 } else {
7639 gen_svm_check_intercept(s, pc_start, SVM_EXIT_WRITE_CR0);
7640 gen_helper_clts(cpu_env);
7641 /* abort block because static cpu state changed */
7642 gen_jmp_im(s->pc - s->cs_base);
7643 gen_eob(s);
7645 break;
7646 /* MMX/3DNow!/SSE/SSE2/SSE3/SSSE3/SSE4 support */
7647 case 0x1c3: /* MOVNTI reg, mem */
7648 if (!(s->cpuid_features & CPUID_SSE2))
7649 goto illegal_op;
7650 ot = mo_64_32(dflag);
7651 modrm = cpu_ldub_code(env, s->pc++);
7652 mod = (modrm >> 6) & 3;
7653 if (mod == 3)
7654 goto illegal_op;
7655 reg = ((modrm >> 3) & 7) | rex_r;
7656 /* generate a generic store */
7657 gen_ldst_modrm(env, s, modrm, ot, reg, 1);
7658 break;
7659 case 0x1ae:
7660 modrm = cpu_ldub_code(env, s->pc++);
7661 mod = (modrm >> 6) & 3;
7662 op = (modrm >> 3) & 7;
7663 switch(op) {
7664 case 0: /* fxsave */
7665 if (mod == 3 || !(s->cpuid_features & CPUID_FXSR) ||
7666 (s->prefix & PREFIX_LOCK))
7667 goto illegal_op;
7668 if ((s->flags & HF_EM_MASK) || (s->flags & HF_TS_MASK)) {
7669 gen_exception(s, EXCP07_PREX, pc_start - s->cs_base);
7670 break;
7672 gen_lea_modrm(env, s, modrm);
7673 gen_update_cc_op(s);
7674 gen_jmp_im(pc_start - s->cs_base);
7675 gen_helper_fxsave(cpu_env, cpu_A0, tcg_const_i32(dflag == MO_64));
7676 break;
7677 case 1: /* fxrstor */
7678 if (mod == 3 || !(s->cpuid_features & CPUID_FXSR) ||
7679 (s->prefix & PREFIX_LOCK))
7680 goto illegal_op;
7681 if ((s->flags & HF_EM_MASK) || (s->flags & HF_TS_MASK)) {
7682 gen_exception(s, EXCP07_PREX, pc_start - s->cs_base);
7683 break;
7685 gen_lea_modrm(env, s, modrm);
7686 gen_update_cc_op(s);
7687 gen_jmp_im(pc_start - s->cs_base);
7688 gen_helper_fxrstor(cpu_env, cpu_A0, tcg_const_i32(dflag == MO_64));
7689 break;
7690 case 2: /* ldmxcsr */
7691 case 3: /* stmxcsr */
7692 if (s->flags & HF_TS_MASK) {
7693 gen_exception(s, EXCP07_PREX, pc_start - s->cs_base);
7694 break;
7696 if ((s->flags & HF_EM_MASK) || !(s->flags & HF_OSFXSR_MASK) ||
7697 mod == 3)
7698 goto illegal_op;
7699 gen_lea_modrm(env, s, modrm);
7700 if (op == 2) {
7701 tcg_gen_qemu_ld_i32(cpu_tmp2_i32, cpu_A0,
7702 s->mem_index, MO_LEUL);
7703 gen_helper_ldmxcsr(cpu_env, cpu_tmp2_i32);
7704 } else {
7705 tcg_gen_ld32u_tl(cpu_T[0], cpu_env, offsetof(CPUX86State, mxcsr));
7706 gen_op_st_v(s, MO_32, cpu_T[0], cpu_A0);
7708 break;
7709 case 5: /* lfence */
7710 case 6: /* mfence */
7711 if ((modrm & 0xc7) != 0xc0 || !(s->cpuid_features & CPUID_SSE2))
7712 goto illegal_op;
7713 break;
7714 case 7: /* sfence / clflush */
7715 if ((modrm & 0xc7) == 0xc0) {
7716 /* sfence */
7717 /* XXX: also check for cpuid_ext2_features & CPUID_EXT2_EMMX */
7718 if (!(s->cpuid_features & CPUID_SSE))
7719 goto illegal_op;
7720 } else {
7721 /* clflush */
7722 if (!(s->cpuid_features & CPUID_CLFLUSH))
7723 goto illegal_op;
7724 gen_lea_modrm(env, s, modrm);
7726 break;
7727 default:
7728 goto illegal_op;
7730 break;
7731 case 0x10d: /* 3DNow! prefetch(w) */
7732 modrm = cpu_ldub_code(env, s->pc++);
7733 mod = (modrm >> 6) & 3;
7734 if (mod == 3)
7735 goto illegal_op;
7736 gen_lea_modrm(env, s, modrm);
7737 /* ignore for now */
7738 break;
7739 case 0x1aa: /* rsm */
7740 gen_svm_check_intercept(s, pc_start, SVM_EXIT_RSM);
7741 if (!(s->flags & HF_SMM_MASK))
7742 goto illegal_op;
7743 gen_update_cc_op(s);
7744 gen_jmp_im(s->pc - s->cs_base);
7745 gen_helper_rsm(cpu_env);
7746 gen_eob(s);
7747 break;
7748 case 0x1b8: /* SSE4.2 popcnt */
7749 if ((prefixes & (PREFIX_REPZ | PREFIX_LOCK | PREFIX_REPNZ)) !=
7750 PREFIX_REPZ)
7751 goto illegal_op;
7752 if (!(s->cpuid_ext_features & CPUID_EXT_POPCNT))
7753 goto illegal_op;
7755 modrm = cpu_ldub_code(env, s->pc++);
7756 reg = ((modrm >> 3) & 7) | rex_r;
7758 if (s->prefix & PREFIX_DATA) {
7759 ot = MO_16;
7760 } else {
7761 ot = mo_64_32(dflag);
7764 gen_ldst_modrm(env, s, modrm, ot, OR_TMP0, 0);
7765 gen_helper_popcnt(cpu_T[0], cpu_env, cpu_T[0], tcg_const_i32(ot));
7766 gen_op_mov_reg_v(ot, reg, cpu_T[0]);
7768 set_cc_op(s, CC_OP_EFLAGS);
7769 break;
7770 case 0x10e ... 0x10f:
7771 /* 3DNow! instructions, ignore prefixes */
7772 s->prefix &= ~(PREFIX_REPZ | PREFIX_REPNZ | PREFIX_DATA);
7773 case 0x110 ... 0x117:
7774 case 0x128 ... 0x12f:
7775 case 0x138 ... 0x13a:
7776 case 0x150 ... 0x179:
7777 case 0x17c ... 0x17f:
7778 case 0x1c2:
7779 case 0x1c4 ... 0x1c6:
7780 case 0x1d0 ... 0x1fe:
7781 gen_sse(env, s, b, pc_start, rex_r);
7782 break;
7783 default:
7784 goto illegal_op;
7786 /* lock generation */
7787 if (s->prefix & PREFIX_LOCK)
7788 gen_helper_unlock();
7789 return s->pc;
7790 illegal_op:
7791 if (s->prefix & PREFIX_LOCK)
7792 gen_helper_unlock();
7793 /* XXX: ensure that no lock was generated */
7794 gen_exception(s, EXCP06_ILLOP, pc_start - s->cs_base);
7795 return s->pc;
7798 void optimize_flags_init(void)
7800 static const char reg_names[CPU_NB_REGS][4] = {
7801 #ifdef TARGET_X86_64
7802 [R_EAX] = "rax",
7803 [R_EBX] = "rbx",
7804 [R_ECX] = "rcx",
7805 [R_EDX] = "rdx",
7806 [R_ESI] = "rsi",
7807 [R_EDI] = "rdi",
7808 [R_EBP] = "rbp",
7809 [R_ESP] = "rsp",
7810 [8] = "r8",
7811 [9] = "r9",
7812 [10] = "r10",
7813 [11] = "r11",
7814 [12] = "r12",
7815 [13] = "r13",
7816 [14] = "r14",
7817 [15] = "r15",
7818 #else
7819 [R_EAX] = "eax",
7820 [R_EBX] = "ebx",
7821 [R_ECX] = "ecx",
7822 [R_EDX] = "edx",
7823 [R_ESI] = "esi",
7824 [R_EDI] = "edi",
7825 [R_EBP] = "ebp",
7826 [R_ESP] = "esp",
7827 #endif
7829 int i;
7831 cpu_env = tcg_global_reg_new_ptr(TCG_AREG0, "env");
7832 cpu_cc_op = tcg_global_mem_new_i32(TCG_AREG0,
7833 offsetof(CPUX86State, cc_op), "cc_op");
7834 cpu_cc_dst = tcg_global_mem_new(TCG_AREG0, offsetof(CPUX86State, cc_dst),
7835 "cc_dst");
7836 cpu_cc_src = tcg_global_mem_new(TCG_AREG0, offsetof(CPUX86State, cc_src),
7837 "cc_src");
7838 cpu_cc_src2 = tcg_global_mem_new(TCG_AREG0, offsetof(CPUX86State, cc_src2),
7839 "cc_src2");
7841 for (i = 0; i < CPU_NB_REGS; ++i) {
7842 cpu_regs[i] = tcg_global_mem_new(TCG_AREG0,
7843 offsetof(CPUX86State, regs[i]),
7844 reg_names[i]);
7848 /* generate intermediate code in gen_opc_buf and gen_opparam_buf for
7849 basic block 'tb'. If search_pc is TRUE, also generate PC
7850 information for each intermediate instruction. */
7851 static inline void gen_intermediate_code_internal(X86CPU *cpu,
7852 TranslationBlock *tb,
7853 bool search_pc)
7855 CPUState *cs = CPU(cpu);
7856 CPUX86State *env = &cpu->env;
7857 DisasContext dc1, *dc = &dc1;
7858 target_ulong pc_ptr;
7859 uint16_t *gen_opc_end;
7860 CPUBreakpoint *bp;
7861 int j, lj;
7862 uint64_t flags;
7863 target_ulong pc_start;
7864 target_ulong cs_base;
7865 int num_insns;
7866 int max_insns;
7868 /* generate intermediate code */
7869 pc_start = tb->pc;
7870 cs_base = tb->cs_base;
7871 flags = tb->flags;
7873 dc->pe = (flags >> HF_PE_SHIFT) & 1;
7874 dc->code32 = (flags >> HF_CS32_SHIFT) & 1;
7875 dc->ss32 = (flags >> HF_SS32_SHIFT) & 1;
7876 dc->addseg = (flags >> HF_ADDSEG_SHIFT) & 1;
7877 dc->f_st = 0;
7878 dc->vm86 = (flags >> VM_SHIFT) & 1;
7879 dc->cpl = (flags >> HF_CPL_SHIFT) & 3;
7880 dc->iopl = (flags >> IOPL_SHIFT) & 3;
7881 dc->tf = (flags >> TF_SHIFT) & 1;
7882 dc->singlestep_enabled = cs->singlestep_enabled;
7883 dc->cc_op = CC_OP_DYNAMIC;
7884 dc->cc_op_dirty = false;
7885 dc->cs_base = cs_base;
7886 dc->tb = tb;
7887 dc->popl_esp_hack = 0;
7888 /* select memory access functions */
7889 dc->mem_index = 0;
7890 if (flags & HF_SOFTMMU_MASK) {
7891 dc->mem_index = cpu_mmu_index(env);
7893 dc->cpuid_features = env->features[FEAT_1_EDX];
7894 dc->cpuid_ext_features = env->features[FEAT_1_ECX];
7895 dc->cpuid_ext2_features = env->features[FEAT_8000_0001_EDX];
7896 dc->cpuid_ext3_features = env->features[FEAT_8000_0001_ECX];
7897 dc->cpuid_7_0_ebx_features = env->features[FEAT_7_0_EBX];
7898 #ifdef TARGET_X86_64
7899 dc->lma = (flags >> HF_LMA_SHIFT) & 1;
7900 dc->code64 = (flags >> HF_CS64_SHIFT) & 1;
7901 #endif
7902 dc->flags = flags;
7903 dc->jmp_opt = !(dc->tf || cs->singlestep_enabled ||
7904 (flags & HF_INHIBIT_IRQ_MASK)
7905 #ifndef CONFIG_SOFTMMU
7906 || (flags & HF_SOFTMMU_MASK)
7907 #endif
7909 #if 0
7910 /* check addseg logic */
7911 if (!dc->addseg && (dc->vm86 || !dc->pe || !dc->code32))
7912 printf("ERROR addseg\n");
7913 #endif
7915 cpu_T[0] = tcg_temp_new();
7916 cpu_T[1] = tcg_temp_new();
7917 cpu_A0 = tcg_temp_new();
7919 cpu_tmp0 = tcg_temp_new();
7920 cpu_tmp1_i64 = tcg_temp_new_i64();
7921 cpu_tmp2_i32 = tcg_temp_new_i32();
7922 cpu_tmp3_i32 = tcg_temp_new_i32();
7923 cpu_tmp4 = tcg_temp_new();
7924 cpu_ptr0 = tcg_temp_new_ptr();
7925 cpu_ptr1 = tcg_temp_new_ptr();
7926 cpu_cc_srcT = tcg_temp_local_new();
7928 gen_opc_end = tcg_ctx.gen_opc_buf + OPC_MAX_SIZE;
7930 dc->is_jmp = DISAS_NEXT;
7931 pc_ptr = pc_start;
7932 lj = -1;
7933 num_insns = 0;
7934 max_insns = tb->cflags & CF_COUNT_MASK;
7935 if (max_insns == 0)
7936 max_insns = CF_COUNT_MASK;
7938 gen_tb_start();
7939 for(;;) {
7940 if (unlikely(!QTAILQ_EMPTY(&env->breakpoints))) {
7941 QTAILQ_FOREACH(bp, &env->breakpoints, entry) {
7942 if (bp->pc == pc_ptr &&
7943 !((bp->flags & BP_CPU) && (tb->flags & HF_RF_MASK))) {
7944 gen_debug(dc, pc_ptr - dc->cs_base);
7945 break;
7949 if (search_pc) {
7950 j = tcg_ctx.gen_opc_ptr - tcg_ctx.gen_opc_buf;
7951 if (lj < j) {
7952 lj++;
7953 while (lj < j)
7954 tcg_ctx.gen_opc_instr_start[lj++] = 0;
7956 tcg_ctx.gen_opc_pc[lj] = pc_ptr;
7957 gen_opc_cc_op[lj] = dc->cc_op;
7958 tcg_ctx.gen_opc_instr_start[lj] = 1;
7959 tcg_ctx.gen_opc_icount[lj] = num_insns;
7961 if (num_insns + 1 == max_insns && (tb->cflags & CF_LAST_IO))
7962 gen_io_start();
7964 pc_ptr = disas_insn(env, dc, pc_ptr);
7965 num_insns++;
7966 /* stop translation if indicated */
7967 if (dc->is_jmp)
7968 break;
7969 /* if single step mode, we generate only one instruction and
7970 generate an exception */
7971 /* if irq were inhibited with HF_INHIBIT_IRQ_MASK, we clear
7972 the flag and abort the translation to give the irqs a
7973 change to be happen */
7974 if (dc->tf || dc->singlestep_enabled ||
7975 (flags & HF_INHIBIT_IRQ_MASK)) {
7976 gen_jmp_im(pc_ptr - dc->cs_base);
7977 gen_eob(dc);
7978 break;
7980 /* if too long translation, stop generation too */
7981 if (tcg_ctx.gen_opc_ptr >= gen_opc_end ||
7982 (pc_ptr - pc_start) >= (TARGET_PAGE_SIZE - 32) ||
7983 num_insns >= max_insns) {
7984 gen_jmp_im(pc_ptr - dc->cs_base);
7985 gen_eob(dc);
7986 break;
7988 if (singlestep) {
7989 gen_jmp_im(pc_ptr - dc->cs_base);
7990 gen_eob(dc);
7991 break;
7994 if (tb->cflags & CF_LAST_IO)
7995 gen_io_end();
7996 gen_tb_end(tb, num_insns);
7997 *tcg_ctx.gen_opc_ptr = INDEX_op_end;
7998 /* we don't forget to fill the last values */
7999 if (search_pc) {
8000 j = tcg_ctx.gen_opc_ptr - tcg_ctx.gen_opc_buf;
8001 lj++;
8002 while (lj <= j)
8003 tcg_ctx.gen_opc_instr_start[lj++] = 0;
8006 #ifdef DEBUG_DISAS
8007 if (qemu_loglevel_mask(CPU_LOG_TB_IN_ASM)) {
8008 int disas_flags;
8009 qemu_log("----------------\n");
8010 qemu_log("IN: %s\n", lookup_symbol(pc_start));
8011 #ifdef TARGET_X86_64
8012 if (dc->code64)
8013 disas_flags = 2;
8014 else
8015 #endif
8016 disas_flags = !dc->code32;
8017 log_target_disas(env, pc_start, pc_ptr - pc_start, disas_flags);
8018 qemu_log("\n");
8020 #endif
8022 if (!search_pc) {
8023 tb->size = pc_ptr - pc_start;
8024 tb->icount = num_insns;
8028 void gen_intermediate_code(CPUX86State *env, TranslationBlock *tb)
8030 gen_intermediate_code_internal(x86_env_get_cpu(env), tb, false);
8033 void gen_intermediate_code_pc(CPUX86State *env, TranslationBlock *tb)
8035 gen_intermediate_code_internal(x86_env_get_cpu(env), tb, true);
8038 void restore_state_to_opc(CPUX86State *env, TranslationBlock *tb, int pc_pos)
8040 int cc_op;
8041 #ifdef DEBUG_DISAS
8042 if (qemu_loglevel_mask(CPU_LOG_TB_OP)) {
8043 int i;
8044 qemu_log("RESTORE:\n");
8045 for(i = 0;i <= pc_pos; i++) {
8046 if (tcg_ctx.gen_opc_instr_start[i]) {
8047 qemu_log("0x%04x: " TARGET_FMT_lx "\n", i,
8048 tcg_ctx.gen_opc_pc[i]);
8051 qemu_log("pc_pos=0x%x eip=" TARGET_FMT_lx " cs_base=%x\n",
8052 pc_pos, tcg_ctx.gen_opc_pc[pc_pos] - tb->cs_base,
8053 (uint32_t)tb->cs_base);
8055 #endif
8056 env->eip = tcg_ctx.gen_opc_pc[pc_pos] - tb->cs_base;
8057 cc_op = gen_opc_cc_op[pc_pos];
8058 if (cc_op != CC_OP_DYNAMIC)
8059 env->cc_op = cc_op;