2 * New-style TCG opcode generator for i386 instructions
4 * Copyright (c) 2022 Red Hat, Inc.
6 * Author: Paolo Bonzini <pbonzini@redhat.com>
8 * This library is free software; you can redistribute it and/or
9 * modify it under the terms of the GNU Lesser General Public
10 * License as published by the Free Software Foundation; either
11 * version 2.1 of the License, or (at your option) any later version.
13 * This library is distributed in the hope that it will be useful,
14 * but WITHOUT ANY WARRANTY; without even the implied warranty of
15 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
16 * Lesser General Public License for more details.
18 * You should have received a copy of the GNU Lesser General Public
19 * License along with this library; if not, see <http://www.gnu.org/licenses/>.
22 #define ZMM_OFFSET(reg) offsetof(CPUX86State, xmm_regs[reg])
24 typedef void (*SSEFunc_i_ep)(TCGv_i32 val, TCGv_ptr env, TCGv_ptr reg);
25 typedef void (*SSEFunc_l_ep)(TCGv_i64 val, TCGv_ptr env, TCGv_ptr reg);
26 typedef void (*SSEFunc_0_epp)(TCGv_ptr env, TCGv_ptr reg_a, TCGv_ptr reg_b);
27 typedef void (*SSEFunc_0_eppp)(TCGv_ptr env, TCGv_ptr reg_a, TCGv_ptr reg_b,
29 typedef void (*SSEFunc_0_epppp)(TCGv_ptr env, TCGv_ptr reg_a, TCGv_ptr reg_b,
30 TCGv_ptr reg_c, TCGv_ptr reg_d);
31 typedef void (*SSEFunc_0_eppi)(TCGv_ptr env, TCGv_ptr reg_a, TCGv_ptr reg_b,
33 typedef void (*SSEFunc_0_epppi)(TCGv_ptr env, TCGv_ptr reg_a, TCGv_ptr reg_b,
34 TCGv_ptr reg_c, TCGv_i32 val);
35 typedef void (*SSEFunc_0_ppi)(TCGv_ptr reg_a, TCGv_ptr reg_b, TCGv_i32 val);
36 typedef void (*SSEFunc_0_pppi)(TCGv_ptr reg_a, TCGv_ptr reg_b, TCGv_ptr reg_c,
38 typedef void (*SSEFunc_0_eppt)(TCGv_ptr env, TCGv_ptr reg_a, TCGv_ptr reg_b,
40 typedef void (*SSEFunc_0_epppti)(TCGv_ptr env, TCGv_ptr reg_a, TCGv_ptr reg_b,
41 TCGv_ptr reg_c, TCGv a0, TCGv_i32 scale);
42 typedef void (*SSEFunc_0_eppppi)(TCGv_ptr env, TCGv_ptr reg_a, TCGv_ptr reg_b,
43 TCGv_ptr reg_c, TCGv_ptr reg_d, TCGv_i32 flags);
44 typedef void (*SSEFunc_0_eppppii)(TCGv_ptr env, TCGv_ptr reg_a, TCGv_ptr reg_b,
45 TCGv_ptr reg_c, TCGv_ptr reg_d, TCGv_i32 even,
48 static inline TCGv_i32 tcg_constant8u_i32(uint8_t val)
50 return tcg_constant_i32(val);
53 static void gen_NM_exception(DisasContext *s)
55 gen_exception(s, EXCP07_PREX);
58 static void gen_load_ea(DisasContext *s, AddressParts *mem, bool is_vsib)
60 TCGv ea = gen_lea_modrm_1(s, *mem, is_vsib);
61 gen_lea_v_seg(s, s->aflag, ea, mem->def_seg, s->override);
64 static inline int mmx_offset(MemOp ot)
68 return offsetof(MMXReg, MMX_B(0));
70 return offsetof(MMXReg, MMX_W(0));
72 return offsetof(MMXReg, MMX_L(0));
74 return offsetof(MMXReg, MMX_Q(0));
76 g_assert_not_reached();
80 static inline int xmm_offset(MemOp ot)
84 return offsetof(ZMMReg, ZMM_B(0));
86 return offsetof(ZMMReg, ZMM_W(0));
88 return offsetof(ZMMReg, ZMM_L(0));
90 return offsetof(ZMMReg, ZMM_Q(0));
92 return offsetof(ZMMReg, ZMM_X(0));
94 return offsetof(ZMMReg, ZMM_Y(0));
96 g_assert_not_reached();
100 static int vector_reg_offset(X86DecodedOp *op)
102 assert(op->unit == X86_OP_MMX || op->unit == X86_OP_SSE);
104 if (op->unit == X86_OP_MMX) {
105 return op->offset - mmx_offset(op->ot);
107 return op->offset - xmm_offset(op->ot);
111 static int vector_elem_offset(X86DecodedOp *op, MemOp ot, int n)
113 int base_ofs = vector_reg_offset(op);
116 if (op->unit == X86_OP_MMX) {
117 return base_ofs + offsetof(MMXReg, MMX_B(n));
119 return base_ofs + offsetof(ZMMReg, ZMM_B(n));
122 if (op->unit == X86_OP_MMX) {
123 return base_ofs + offsetof(MMXReg, MMX_W(n));
125 return base_ofs + offsetof(ZMMReg, ZMM_W(n));
128 if (op->unit == X86_OP_MMX) {
129 return base_ofs + offsetof(MMXReg, MMX_L(n));
131 return base_ofs + offsetof(ZMMReg, ZMM_L(n));
134 if (op->unit == X86_OP_MMX) {
137 return base_ofs + offsetof(ZMMReg, ZMM_Q(n));
140 assert(op->unit == X86_OP_SSE);
141 return base_ofs + offsetof(ZMMReg, ZMM_X(n));
143 assert(op->unit == X86_OP_SSE);
144 return base_ofs + offsetof(ZMMReg, ZMM_Y(n));
146 g_assert_not_reached();
150 static void compute_mmx_offset(X86DecodedOp *op)
153 op->offset = offsetof(CPUX86State, fpregs[op->n].mmx) + mmx_offset(op->ot);
155 op->offset = offsetof(CPUX86State, mmx_t0) + mmx_offset(op->ot);
159 static void compute_xmm_offset(X86DecodedOp *op)
162 op->offset = ZMM_OFFSET(op->n) + xmm_offset(op->ot);
164 op->offset = offsetof(CPUX86State, xmm_t0) + xmm_offset(op->ot);
168 static void gen_load_sse(DisasContext *s, TCGv temp, MemOp ot, int dest_ofs, bool aligned)
172 gen_op_ld_v(s, MO_8, temp, s->A0);
173 tcg_gen_st8_tl(temp, tcg_env, dest_ofs);
176 gen_op_ld_v(s, MO_16, temp, s->A0);
177 tcg_gen_st16_tl(temp, tcg_env, dest_ofs);
180 gen_op_ld_v(s, MO_32, temp, s->A0);
181 tcg_gen_st32_tl(temp, tcg_env, dest_ofs);
184 gen_ldq_env_A0(s, dest_ofs);
187 gen_ldo_env_A0(s, dest_ofs, aligned);
190 gen_ldy_env_A0(s, dest_ofs, aligned);
193 g_assert_not_reached();
197 static bool sse_needs_alignment(DisasContext *s, X86DecodedInsn *decode, MemOp ot)
199 switch (decode->e.vex_class) {
202 if ((s->prefix & PREFIX_VEX) ||
203 decode->e.vex_special == X86_VEX_SSEUnaligned) {
204 /* MOST legacy SSE instructions require aligned memory operands, but not all. */
216 static void gen_load(DisasContext *s, X86DecodedInsn *decode, int opn, TCGv v)
218 X86DecodedOp *op = &decode->op[opn];
224 tcg_gen_ld32u_tl(v, tcg_env,
225 offsetof(CPUX86State,segs[op->n].selector));
228 tcg_gen_ld_tl(v, tcg_env, offsetof(CPUX86State, cr[op->n]));
231 tcg_gen_ld_tl(v, tcg_env, offsetof(CPUX86State, dr[op->n]));
235 gen_op_ld_v(s, op->ot, v, s->A0);
237 gen_op_mov_v_reg(s, op->ot, v, op->n);
241 tcg_gen_movi_tl(v, decode->immediate);
245 compute_mmx_offset(op);
249 compute_xmm_offset(op);
252 bool aligned = sse_needs_alignment(s, decode, op->ot);
253 gen_load_sse(s, v, op->ot, op->offset, aligned);
258 g_assert_not_reached();
262 static TCGv_ptr op_ptr(X86DecodedInsn *decode, int opn)
264 X86DecodedOp *op = &decode->op[opn];
268 op->v_ptr = tcg_temp_new_ptr();
270 /* The temporary points to the MMXReg or ZMMReg. */
271 tcg_gen_addi_ptr(op->v_ptr, tcg_env, vector_reg_offset(op));
275 #define OP_PTR0 op_ptr(decode, 0)
276 #define OP_PTR1 op_ptr(decode, 1)
277 #define OP_PTR2 op_ptr(decode, 2)
279 static void gen_writeback(DisasContext *s, X86DecodedInsn *decode, int opn, TCGv v)
281 X86DecodedOp *op = &decode->op[opn];
286 /* Note that gen_movl_seg_T0 takes care of interrupt shadow and TF. */
287 gen_movl_seg_T0(s, op->n);
291 gen_op_st_v(s, op->ot, v, s->A0);
293 gen_op_mov_reg_v(s, op->ot, op->n, v);
299 if (!op->has_ea && (s->prefix & PREFIX_VEX) && op->ot <= MO_128) {
300 tcg_gen_gvec_dup_imm(MO_64,
301 offsetof(CPUX86State, xmm_regs[op->n].ZMM_X(1)),
308 g_assert_not_reached();
312 static inline int vector_len(DisasContext *s, X86DecodedInsn *decode)
314 if (decode->e.special == X86_SPECIAL_MMX &&
315 !(s->prefix & (PREFIX_DATA | PREFIX_REPZ | PREFIX_REPNZ))) {
318 return s->vex_l ? 32 : 16;
321 static void gen_store_sse(DisasContext *s, X86DecodedInsn *decode, int src_ofs)
323 MemOp ot = decode->op[0].ot;
324 int vec_len = vector_len(s, decode);
325 bool aligned = sse_needs_alignment(s, decode, ot);
327 if (!decode->op[0].has_ea) {
328 tcg_gen_gvec_mov(MO_64, decode->op[0].offset, src_ofs, vec_len, vec_len);
334 gen_stq_env_A0(s, src_ofs);
337 gen_sto_env_A0(s, src_ofs, aligned);
340 gen_sty_env_A0(s, src_ofs, aligned);
343 g_assert_not_reached();
347 static void gen_helper_pavgusb(TCGv_ptr env, TCGv_ptr reg_a, TCGv_ptr reg_b)
349 gen_helper_pavgb_mmx(env, reg_a, reg_a, reg_b);
352 #define FN_3DNOW_MOVE ((SSEFunc_0_epp) (uintptr_t) 1)
353 static const SSEFunc_0_epp fns_3dnow[] = {
354 [0x0c] = gen_helper_pi2fw,
355 [0x0d] = gen_helper_pi2fd,
356 [0x1c] = gen_helper_pf2iw,
357 [0x1d] = gen_helper_pf2id,
358 [0x8a] = gen_helper_pfnacc,
359 [0x8e] = gen_helper_pfpnacc,
360 [0x90] = gen_helper_pfcmpge,
361 [0x94] = gen_helper_pfmin,
362 [0x96] = gen_helper_pfrcp,
363 [0x97] = gen_helper_pfrsqrt,
364 [0x9a] = gen_helper_pfsub,
365 [0x9e] = gen_helper_pfadd,
366 [0xa0] = gen_helper_pfcmpgt,
367 [0xa4] = gen_helper_pfmax,
368 [0xa6] = FN_3DNOW_MOVE, /* PFRCPIT1; no need to actually increase precision */
369 [0xa7] = FN_3DNOW_MOVE, /* PFRSQIT1 */
370 [0xb6] = FN_3DNOW_MOVE, /* PFRCPIT2 */
371 [0xaa] = gen_helper_pfsubr,
372 [0xae] = gen_helper_pfacc,
373 [0xb0] = gen_helper_pfcmpeq,
374 [0xb4] = gen_helper_pfmul,
375 [0xb7] = gen_helper_pmulhrw_mmx,
376 [0xbb] = gen_helper_pswapd,
377 [0xbf] = gen_helper_pavgusb,
380 static void gen_3dnow(DisasContext *s, CPUX86State *env, X86DecodedInsn *decode)
382 uint8_t b = decode->immediate;
383 SSEFunc_0_epp fn = b < ARRAY_SIZE(fns_3dnow) ? fns_3dnow[b] : NULL;
386 gen_illegal_opcode(s);
389 if (s->flags & HF_TS_MASK) {
393 if (s->flags & HF_EM_MASK) {
394 gen_illegal_opcode(s);
398 gen_helper_enter_mmx(tcg_env);
399 if (fn == FN_3DNOW_MOVE) {
400 tcg_gen_ld_i64(s->tmp1_i64, tcg_env, decode->op[1].offset);
401 tcg_gen_st_i64(s->tmp1_i64, tcg_env, decode->op[0].offset);
403 fn(tcg_env, OP_PTR0, OP_PTR1);
408 * 00 = v*ps Vps, Hps, Wpd
409 * 66 = v*pd Vpd, Hpd, Wps
410 * f3 = v*ss Vss, Hss, Wps
411 * f2 = v*sd Vsd, Hsd, Wps
413 static inline void gen_unary_fp_sse(DisasContext *s, CPUX86State *env, X86DecodedInsn *decode,
414 SSEFunc_0_epp pd_xmm, SSEFunc_0_epp ps_xmm,
415 SSEFunc_0_epp pd_ymm, SSEFunc_0_epp ps_ymm,
416 SSEFunc_0_eppp sd, SSEFunc_0_eppp ss)
418 if ((s->prefix & (PREFIX_REPZ | PREFIX_REPNZ)) != 0) {
419 SSEFunc_0_eppp fn = s->prefix & PREFIX_REPZ ? ss : sd;
421 gen_illegal_opcode(s);
424 fn(tcg_env, OP_PTR0, OP_PTR1, OP_PTR2);
426 SSEFunc_0_epp ps, pd, fn;
427 ps = s->vex_l ? ps_ymm : ps_xmm;
428 pd = s->vex_l ? pd_ymm : pd_xmm;
429 fn = s->prefix & PREFIX_DATA ? pd : ps;
431 gen_illegal_opcode(s);
434 fn(tcg_env, OP_PTR0, OP_PTR2);
437 #define UNARY_FP_SSE(uname, lname) \
438 static void gen_##uname(DisasContext *s, CPUX86State *env, X86DecodedInsn *decode) \
440 gen_unary_fp_sse(s, env, decode, \
441 gen_helper_##lname##pd_xmm, \
442 gen_helper_##lname##ps_xmm, \
443 gen_helper_##lname##pd_ymm, \
444 gen_helper_##lname##ps_ymm, \
445 gen_helper_##lname##sd, \
446 gen_helper_##lname##ss); \
448 UNARY_FP_SSE(VSQRT, sqrt)
451 * 00 = v*ps Vps, Hps, Wpd
452 * 66 = v*pd Vpd, Hpd, Wps
453 * f3 = v*ss Vss, Hss, Wps
454 * f2 = v*sd Vsd, Hsd, Wps
456 static inline void gen_fp_sse(DisasContext *s, CPUX86State *env, X86DecodedInsn *decode,
457 SSEFunc_0_eppp pd_xmm, SSEFunc_0_eppp ps_xmm,
458 SSEFunc_0_eppp pd_ymm, SSEFunc_0_eppp ps_ymm,
459 SSEFunc_0_eppp sd, SSEFunc_0_eppp ss)
461 SSEFunc_0_eppp ps, pd, fn;
462 if ((s->prefix & (PREFIX_REPZ | PREFIX_REPNZ)) != 0) {
463 fn = s->prefix & PREFIX_REPZ ? ss : sd;
465 ps = s->vex_l ? ps_ymm : ps_xmm;
466 pd = s->vex_l ? pd_ymm : pd_xmm;
467 fn = s->prefix & PREFIX_DATA ? pd : ps;
470 fn(tcg_env, OP_PTR0, OP_PTR1, OP_PTR2);
472 gen_illegal_opcode(s);
476 #define FP_SSE(uname, lname) \
477 static void gen_##uname(DisasContext *s, CPUX86State *env, X86DecodedInsn *decode) \
479 gen_fp_sse(s, env, decode, \
480 gen_helper_##lname##pd_xmm, \
481 gen_helper_##lname##ps_xmm, \
482 gen_helper_##lname##pd_ymm, \
483 gen_helper_##lname##ps_ymm, \
484 gen_helper_##lname##sd, \
485 gen_helper_##lname##ss); \
494 #define FMA_SSE_PACKED(uname, ptr0, ptr1, ptr2, even, odd) \
495 static void gen_##uname##Px(DisasContext *s, CPUX86State *env, X86DecodedInsn *decode) \
497 SSEFunc_0_eppppii xmm = s->vex_w ? gen_helper_fma4pd_xmm : gen_helper_fma4ps_xmm; \
498 SSEFunc_0_eppppii ymm = s->vex_w ? gen_helper_fma4pd_ymm : gen_helper_fma4ps_ymm; \
499 SSEFunc_0_eppppii fn = s->vex_l ? ymm : xmm; \
501 fn(tcg_env, OP_PTR0, ptr0, ptr1, ptr2, \
502 tcg_constant_i32(even), \
503 tcg_constant_i32((even) ^ (odd))); \
506 #define FMA_SSE(uname, ptr0, ptr1, ptr2, flags) \
507 FMA_SSE_PACKED(uname, ptr0, ptr1, ptr2, flags, flags) \
508 static void gen_##uname##Sx(DisasContext *s, CPUX86State *env, X86DecodedInsn *decode) \
510 SSEFunc_0_eppppi fn = s->vex_w ? gen_helper_fma4sd : gen_helper_fma4ss; \
512 fn(tcg_env, OP_PTR0, ptr0, ptr1, ptr2, \
513 tcg_constant_i32(flags)); \
516 FMA_SSE(VFMADD231, OP_PTR1, OP_PTR2, OP_PTR0, 0)
517 FMA_SSE(VFMADD213, OP_PTR1, OP_PTR0, OP_PTR2, 0)
518 FMA_SSE(VFMADD132, OP_PTR0, OP_PTR2, OP_PTR1, 0)
520 FMA_SSE(VFNMADD231, OP_PTR1, OP_PTR2, OP_PTR0, float_muladd_negate_product)
521 FMA_SSE(VFNMADD213, OP_PTR1, OP_PTR0, OP_PTR2, float_muladd_negate_product)
522 FMA_SSE(VFNMADD132, OP_PTR0, OP_PTR2, OP_PTR1, float_muladd_negate_product)
524 FMA_SSE(VFMSUB231, OP_PTR1, OP_PTR2, OP_PTR0, float_muladd_negate_c)
525 FMA_SSE(VFMSUB213, OP_PTR1, OP_PTR0, OP_PTR2, float_muladd_negate_c)
526 FMA_SSE(VFMSUB132, OP_PTR0, OP_PTR2, OP_PTR1, float_muladd_negate_c)
528 FMA_SSE(VFNMSUB231, OP_PTR1, OP_PTR2, OP_PTR0, float_muladd_negate_c|float_muladd_negate_product)
529 FMA_SSE(VFNMSUB213, OP_PTR1, OP_PTR0, OP_PTR2, float_muladd_negate_c|float_muladd_negate_product)
530 FMA_SSE(VFNMSUB132, OP_PTR0, OP_PTR2, OP_PTR1, float_muladd_negate_c|float_muladd_negate_product)
532 FMA_SSE_PACKED(VFMADDSUB231, OP_PTR1, OP_PTR2, OP_PTR0, float_muladd_negate_c, 0)
533 FMA_SSE_PACKED(VFMADDSUB213, OP_PTR1, OP_PTR0, OP_PTR2, float_muladd_negate_c, 0)
534 FMA_SSE_PACKED(VFMADDSUB132, OP_PTR0, OP_PTR2, OP_PTR1, float_muladd_negate_c, 0)
536 FMA_SSE_PACKED(VFMSUBADD231, OP_PTR1, OP_PTR2, OP_PTR0, 0, float_muladd_negate_c)
537 FMA_SSE_PACKED(VFMSUBADD213, OP_PTR1, OP_PTR0, OP_PTR2, 0, float_muladd_negate_c)
538 FMA_SSE_PACKED(VFMSUBADD132, OP_PTR0, OP_PTR2, OP_PTR1, 0, float_muladd_negate_c)
540 #define FP_UNPACK_SSE(uname, lname) \
541 static void gen_##uname(DisasContext *s, CPUX86State *env, X86DecodedInsn *decode) \
543 /* PS maps to the DQ integer instruction, PD maps to QDQ. */ \
544 gen_fp_sse(s, env, decode, \
545 gen_helper_##lname##qdq_xmm, \
546 gen_helper_##lname##dq_xmm, \
547 gen_helper_##lname##qdq_ymm, \
548 gen_helper_##lname##dq_ymm, \
551 FP_UNPACK_SSE(VUNPCKLPx, punpckl)
552 FP_UNPACK_SSE(VUNPCKHPx, punpckh)
558 static inline void gen_unary_fp32_sse(DisasContext *s, CPUX86State *env, X86DecodedInsn *decode,
559 SSEFunc_0_epp ps_xmm,
560 SSEFunc_0_epp ps_ymm,
563 if ((s->prefix & (PREFIX_DATA | PREFIX_REPNZ)) != 0) {
565 } else if (s->prefix & PREFIX_REPZ) {
569 ss(tcg_env, OP_PTR0, OP_PTR1, OP_PTR2);
571 SSEFunc_0_epp fn = s->vex_l ? ps_ymm : ps_xmm;
575 fn(tcg_env, OP_PTR0, OP_PTR2);
580 gen_illegal_opcode(s);
582 #define UNARY_FP32_SSE(uname, lname) \
583 static void gen_##uname(DisasContext *s, CPUX86State *env, X86DecodedInsn *decode) \
585 gen_unary_fp32_sse(s, env, decode, \
586 gen_helper_##lname##ps_xmm, \
587 gen_helper_##lname##ps_ymm, \
588 gen_helper_##lname##ss); \
590 UNARY_FP32_SSE(VRSQRT, rsqrt)
591 UNARY_FP32_SSE(VRCP, rcp)
594 * 66 = v*pd Vpd, Hpd, Wpd
595 * f2 = v*ps Vps, Hps, Wps
597 static inline void gen_horizontal_fp_sse(DisasContext *s, CPUX86State *env, X86DecodedInsn *decode,
598 SSEFunc_0_eppp pd_xmm, SSEFunc_0_eppp ps_xmm,
599 SSEFunc_0_eppp pd_ymm, SSEFunc_0_eppp ps_ymm)
601 SSEFunc_0_eppp ps, pd, fn;
602 ps = s->vex_l ? ps_ymm : ps_xmm;
603 pd = s->vex_l ? pd_ymm : pd_xmm;
604 fn = s->prefix & PREFIX_DATA ? pd : ps;
605 fn(tcg_env, OP_PTR0, OP_PTR1, OP_PTR2);
607 #define HORIZONTAL_FP_SSE(uname, lname) \
608 static void gen_##uname(DisasContext *s, CPUX86State *env, X86DecodedInsn *decode) \
610 gen_horizontal_fp_sse(s, env, decode, \
611 gen_helper_##lname##pd_xmm, gen_helper_##lname##ps_xmm, \
612 gen_helper_##lname##pd_ymm, gen_helper_##lname##ps_ymm); \
614 HORIZONTAL_FP_SSE(VHADD, hadd)
615 HORIZONTAL_FP_SSE(VHSUB, hsub)
616 HORIZONTAL_FP_SSE(VADDSUB, addsub)
618 static inline void gen_ternary_sse(DisasContext *s, CPUX86State *env, X86DecodedInsn *decode,
619 int op3, SSEFunc_0_epppp xmm, SSEFunc_0_epppp ymm)
621 SSEFunc_0_epppp fn = s->vex_l ? ymm : xmm;
622 TCGv_ptr ptr3 = tcg_temp_new_ptr();
624 /* The format of the fourth input is Lx */
625 tcg_gen_addi_ptr(ptr3, tcg_env, ZMM_OFFSET(op3));
626 fn(tcg_env, OP_PTR0, OP_PTR1, OP_PTR2, ptr3);
628 #define TERNARY_SSE(uname, uvname, lname) \
629 static void gen_##uvname(DisasContext *s, CPUX86State *env, X86DecodedInsn *decode) \
631 gen_ternary_sse(s, env, decode, (uint8_t)decode->immediate >> 4, \
632 gen_helper_##lname##_xmm, gen_helper_##lname##_ymm); \
634 static void gen_##uname(DisasContext *s, CPUX86State *env, X86DecodedInsn *decode) \
636 gen_ternary_sse(s, env, decode, 0, \
637 gen_helper_##lname##_xmm, gen_helper_##lname##_ymm); \
639 TERNARY_SSE(BLENDVPS, VBLENDVPS, blendvps)
640 TERNARY_SSE(BLENDVPD, VBLENDVPD, blendvpd)
641 TERNARY_SSE(PBLENDVB, VPBLENDVB, pblendvb)
643 static inline void gen_binary_imm_sse(DisasContext *s, CPUX86State *env, X86DecodedInsn *decode,
644 SSEFunc_0_epppi xmm, SSEFunc_0_epppi ymm)
646 TCGv_i32 imm = tcg_constant8u_i32(decode->immediate);
648 xmm(tcg_env, OP_PTR0, OP_PTR1, OP_PTR2, imm);
650 ymm(tcg_env, OP_PTR0, OP_PTR1, OP_PTR2, imm);
654 #define BINARY_IMM_SSE(uname, lname) \
655 static void gen_##uname(DisasContext *s, CPUX86State *env, X86DecodedInsn *decode) \
657 gen_binary_imm_sse(s, env, decode, \
658 gen_helper_##lname##_xmm, \
659 gen_helper_##lname##_ymm); \
662 BINARY_IMM_SSE(VBLENDPD, blendpd)
663 BINARY_IMM_SSE(VBLENDPS, blendps)
664 BINARY_IMM_SSE(VPBLENDW, pblendw)
665 BINARY_IMM_SSE(VDDPS, dpps)
666 #define gen_helper_dppd_ymm NULL
667 BINARY_IMM_SSE(VDDPD, dppd)
668 BINARY_IMM_SSE(VMPSADBW, mpsadbw)
669 BINARY_IMM_SSE(PCLMULQDQ, pclmulqdq)
672 #define UNARY_INT_GVEC(uname, func, ...) \
673 static void gen_##uname(DisasContext *s, CPUX86State *env, X86DecodedInsn *decode) \
675 int vec_len = vector_len(s, decode); \
677 func(__VA_ARGS__, decode->op[0].offset, \
678 decode->op[2].offset, vec_len, vec_len); \
680 UNARY_INT_GVEC(PABSB, tcg_gen_gvec_abs, MO_8)
681 UNARY_INT_GVEC(PABSW, tcg_gen_gvec_abs, MO_16)
682 UNARY_INT_GVEC(PABSD, tcg_gen_gvec_abs, MO_32)
683 UNARY_INT_GVEC(VBROADCASTx128, tcg_gen_gvec_dup_mem, MO_128)
684 UNARY_INT_GVEC(VPBROADCASTB, tcg_gen_gvec_dup_mem, MO_8)
685 UNARY_INT_GVEC(VPBROADCASTW, tcg_gen_gvec_dup_mem, MO_16)
686 UNARY_INT_GVEC(VPBROADCASTD, tcg_gen_gvec_dup_mem, MO_32)
687 UNARY_INT_GVEC(VPBROADCASTQ, tcg_gen_gvec_dup_mem, MO_64)
690 #define BINARY_INT_GVEC(uname, func, ...) \
691 static void gen_##uname(DisasContext *s, CPUX86State *env, X86DecodedInsn *decode) \
693 int vec_len = vector_len(s, decode); \
696 decode->op[0].offset, decode->op[1].offset, \
697 decode->op[2].offset, vec_len, vec_len); \
700 BINARY_INT_GVEC(PADDB, tcg_gen_gvec_add, MO_8)
701 BINARY_INT_GVEC(PADDW, tcg_gen_gvec_add, MO_16)
702 BINARY_INT_GVEC(PADDD, tcg_gen_gvec_add, MO_32)
703 BINARY_INT_GVEC(PADDQ, tcg_gen_gvec_add, MO_64)
704 BINARY_INT_GVEC(PADDSB, tcg_gen_gvec_ssadd, MO_8)
705 BINARY_INT_GVEC(PADDSW, tcg_gen_gvec_ssadd, MO_16)
706 BINARY_INT_GVEC(PADDUSB, tcg_gen_gvec_usadd, MO_8)
707 BINARY_INT_GVEC(PADDUSW, tcg_gen_gvec_usadd, MO_16)
708 BINARY_INT_GVEC(PAND, tcg_gen_gvec_and, MO_64)
709 BINARY_INT_GVEC(PCMPEQB, tcg_gen_gvec_cmp, TCG_COND_EQ, MO_8)
710 BINARY_INT_GVEC(PCMPEQD, tcg_gen_gvec_cmp, TCG_COND_EQ, MO_32)
711 BINARY_INT_GVEC(PCMPEQW, tcg_gen_gvec_cmp, TCG_COND_EQ, MO_16)
712 BINARY_INT_GVEC(PCMPEQQ, tcg_gen_gvec_cmp, TCG_COND_EQ, MO_64)
713 BINARY_INT_GVEC(PCMPGTB, tcg_gen_gvec_cmp, TCG_COND_GT, MO_8)
714 BINARY_INT_GVEC(PCMPGTW, tcg_gen_gvec_cmp, TCG_COND_GT, MO_16)
715 BINARY_INT_GVEC(PCMPGTD, tcg_gen_gvec_cmp, TCG_COND_GT, MO_32)
716 BINARY_INT_GVEC(PCMPGTQ, tcg_gen_gvec_cmp, TCG_COND_GT, MO_64)
717 BINARY_INT_GVEC(PMAXSB, tcg_gen_gvec_smax, MO_8)
718 BINARY_INT_GVEC(PMAXSW, tcg_gen_gvec_smax, MO_16)
719 BINARY_INT_GVEC(PMAXSD, tcg_gen_gvec_smax, MO_32)
720 BINARY_INT_GVEC(PMAXUB, tcg_gen_gvec_umax, MO_8)
721 BINARY_INT_GVEC(PMAXUW, tcg_gen_gvec_umax, MO_16)
722 BINARY_INT_GVEC(PMAXUD, tcg_gen_gvec_umax, MO_32)
723 BINARY_INT_GVEC(PMINSB, tcg_gen_gvec_smin, MO_8)
724 BINARY_INT_GVEC(PMINSW, tcg_gen_gvec_smin, MO_16)
725 BINARY_INT_GVEC(PMINSD, tcg_gen_gvec_smin, MO_32)
726 BINARY_INT_GVEC(PMINUB, tcg_gen_gvec_umin, MO_8)
727 BINARY_INT_GVEC(PMINUW, tcg_gen_gvec_umin, MO_16)
728 BINARY_INT_GVEC(PMINUD, tcg_gen_gvec_umin, MO_32)
729 BINARY_INT_GVEC(PMULLW, tcg_gen_gvec_mul, MO_16)
730 BINARY_INT_GVEC(PMULLD, tcg_gen_gvec_mul, MO_32)
731 BINARY_INT_GVEC(POR, tcg_gen_gvec_or, MO_64)
732 BINARY_INT_GVEC(PSUBB, tcg_gen_gvec_sub, MO_8)
733 BINARY_INT_GVEC(PSUBW, tcg_gen_gvec_sub, MO_16)
734 BINARY_INT_GVEC(PSUBD, tcg_gen_gvec_sub, MO_32)
735 BINARY_INT_GVEC(PSUBQ, tcg_gen_gvec_sub, MO_64)
736 BINARY_INT_GVEC(PSUBSB, tcg_gen_gvec_sssub, MO_8)
737 BINARY_INT_GVEC(PSUBSW, tcg_gen_gvec_sssub, MO_16)
738 BINARY_INT_GVEC(PSUBUSB, tcg_gen_gvec_ussub, MO_8)
739 BINARY_INT_GVEC(PSUBUSW, tcg_gen_gvec_ussub, MO_16)
740 BINARY_INT_GVEC(PXOR, tcg_gen_gvec_xor, MO_64)
744 * 00 = p* Pq, Qq (if mmx not NULL; no VEX)
745 * 66 = vp* Vx, Hx, Wx
747 * These are really the same encoding, because 1) V is the same as P when VEX.V
748 * is not present 2) P and Q are the same as H and W apart from MM/XMM
750 static inline void gen_binary_int_sse(DisasContext *s, CPUX86State *env, X86DecodedInsn *decode,
751 SSEFunc_0_eppp mmx, SSEFunc_0_eppp xmm, SSEFunc_0_eppp ymm)
753 assert(!!mmx == !!(decode->e.special == X86_SPECIAL_MMX));
755 if (mmx && (s->prefix & PREFIX_VEX) && !(s->prefix & PREFIX_DATA)) {
756 /* VEX encoding is not applicable to MMX instructions. */
757 gen_illegal_opcode(s);
760 if (!(s->prefix & PREFIX_DATA)) {
761 mmx(tcg_env, OP_PTR0, OP_PTR1, OP_PTR2);
762 } else if (!s->vex_l) {
763 xmm(tcg_env, OP_PTR0, OP_PTR1, OP_PTR2);
765 ymm(tcg_env, OP_PTR0, OP_PTR1, OP_PTR2);
770 #define BINARY_INT_MMX(uname, lname) \
771 static void gen_##uname(DisasContext *s, CPUX86State *env, X86DecodedInsn *decode) \
773 gen_binary_int_sse(s, env, decode, \
774 gen_helper_##lname##_mmx, \
775 gen_helper_##lname##_xmm, \
776 gen_helper_##lname##_ymm); \
778 BINARY_INT_MMX(PUNPCKLBW, punpcklbw)
779 BINARY_INT_MMX(PUNPCKLWD, punpcklwd)
780 BINARY_INT_MMX(PUNPCKLDQ, punpckldq)
781 BINARY_INT_MMX(PACKSSWB, packsswb)
782 BINARY_INT_MMX(PACKUSWB, packuswb)
783 BINARY_INT_MMX(PUNPCKHBW, punpckhbw)
784 BINARY_INT_MMX(PUNPCKHWD, punpckhwd)
785 BINARY_INT_MMX(PUNPCKHDQ, punpckhdq)
786 BINARY_INT_MMX(PACKSSDW, packssdw)
788 BINARY_INT_MMX(PAVGB, pavgb)
789 BINARY_INT_MMX(PAVGW, pavgw)
790 BINARY_INT_MMX(PMADDWD, pmaddwd)
791 BINARY_INT_MMX(PMULHUW, pmulhuw)
792 BINARY_INT_MMX(PMULHW, pmulhw)
793 BINARY_INT_MMX(PMULUDQ, pmuludq)
794 BINARY_INT_MMX(PSADBW, psadbw)
796 BINARY_INT_MMX(PSLLW_r, psllw)
797 BINARY_INT_MMX(PSLLD_r, pslld)
798 BINARY_INT_MMX(PSLLQ_r, psllq)
799 BINARY_INT_MMX(PSRLW_r, psrlw)
800 BINARY_INT_MMX(PSRLD_r, psrld)
801 BINARY_INT_MMX(PSRLQ_r, psrlq)
802 BINARY_INT_MMX(PSRAW_r, psraw)
803 BINARY_INT_MMX(PSRAD_r, psrad)
805 BINARY_INT_MMX(PHADDW, phaddw)
806 BINARY_INT_MMX(PHADDSW, phaddsw)
807 BINARY_INT_MMX(PHADDD, phaddd)
808 BINARY_INT_MMX(PHSUBW, phsubw)
809 BINARY_INT_MMX(PHSUBSW, phsubsw)
810 BINARY_INT_MMX(PHSUBD, phsubd)
811 BINARY_INT_MMX(PMADDUBSW, pmaddubsw)
812 BINARY_INT_MMX(PSHUFB, pshufb)
813 BINARY_INT_MMX(PSIGNB, psignb)
814 BINARY_INT_MMX(PSIGNW, psignw)
815 BINARY_INT_MMX(PSIGND, psignd)
816 BINARY_INT_MMX(PMULHRSW, pmulhrsw)
818 /* Instructions with no MMX equivalent. */
819 #define BINARY_INT_SSE(uname, lname) \
820 static void gen_##uname(DisasContext *s, CPUX86State *env, X86DecodedInsn *decode) \
822 gen_binary_int_sse(s, env, decode, \
824 gen_helper_##lname##_xmm, \
825 gen_helper_##lname##_ymm); \
828 /* Instructions with no MMX equivalent. */
829 BINARY_INT_SSE(PUNPCKLQDQ, punpcklqdq)
830 BINARY_INT_SSE(PUNPCKHQDQ, punpckhqdq)
831 BINARY_INT_SSE(VPACKUSDW, packusdw)
832 BINARY_INT_SSE(VPERMILPS, vpermilps)
833 BINARY_INT_SSE(VPERMILPD, vpermilpd)
834 BINARY_INT_SSE(VMASKMOVPS, vpmaskmovd)
835 BINARY_INT_SSE(VMASKMOVPD, vpmaskmovq)
837 BINARY_INT_SSE(PMULDQ, pmuldq)
839 BINARY_INT_SSE(VAESDEC, aesdec)
840 BINARY_INT_SSE(VAESDECLAST, aesdeclast)
841 BINARY_INT_SSE(VAESENC, aesenc)
842 BINARY_INT_SSE(VAESENCLAST, aesenclast)
844 #define UNARY_CMP_SSE(uname, lname) \
845 static void gen_##uname(DisasContext *s, CPUX86State *env, X86DecodedInsn *decode) \
848 gen_helper_##lname##_xmm(tcg_env, OP_PTR1, OP_PTR2); \
850 gen_helper_##lname##_ymm(tcg_env, OP_PTR1, OP_PTR2); \
852 set_cc_op(s, CC_OP_EFLAGS); \
854 UNARY_CMP_SSE(VPTEST, ptest)
855 UNARY_CMP_SSE(VTESTPS, vtestps)
856 UNARY_CMP_SSE(VTESTPD, vtestpd)
858 static inline void gen_unary_int_sse(DisasContext *s, CPUX86State *env, X86DecodedInsn *decode,
859 SSEFunc_0_epp xmm, SSEFunc_0_epp ymm)
862 xmm(tcg_env, OP_PTR0, OP_PTR2);
864 ymm(tcg_env, OP_PTR0, OP_PTR2);
868 #define UNARY_INT_SSE(uname, lname) \
869 static void gen_##uname(DisasContext *s, CPUX86State *env, X86DecodedInsn *decode) \
871 gen_unary_int_sse(s, env, decode, \
872 gen_helper_##lname##_xmm, \
873 gen_helper_##lname##_ymm); \
876 UNARY_INT_SSE(VPMOVSXBW, pmovsxbw)
877 UNARY_INT_SSE(VPMOVSXBD, pmovsxbd)
878 UNARY_INT_SSE(VPMOVSXBQ, pmovsxbq)
879 UNARY_INT_SSE(VPMOVSXWD, pmovsxwd)
880 UNARY_INT_SSE(VPMOVSXWQ, pmovsxwq)
881 UNARY_INT_SSE(VPMOVSXDQ, pmovsxdq)
883 UNARY_INT_SSE(VPMOVZXBW, pmovzxbw)
884 UNARY_INT_SSE(VPMOVZXBD, pmovzxbd)
885 UNARY_INT_SSE(VPMOVZXBQ, pmovzxbq)
886 UNARY_INT_SSE(VPMOVZXWD, pmovzxwd)
887 UNARY_INT_SSE(VPMOVZXWQ, pmovzxwq)
888 UNARY_INT_SSE(VPMOVZXDQ, pmovzxdq)
890 UNARY_INT_SSE(VMOVSLDUP, pmovsldup)
891 UNARY_INT_SSE(VMOVSHDUP, pmovshdup)
892 UNARY_INT_SSE(VMOVDDUP, pmovdldup)
894 UNARY_INT_SSE(VCVTDQ2PD, cvtdq2pd)
895 UNARY_INT_SSE(VCVTPD2DQ, cvtpd2dq)
896 UNARY_INT_SSE(VCVTTPD2DQ, cvttpd2dq)
897 UNARY_INT_SSE(VCVTDQ2PS, cvtdq2ps)
898 UNARY_INT_SSE(VCVTPS2DQ, cvtps2dq)
899 UNARY_INT_SSE(VCVTTPS2DQ, cvttps2dq)
900 UNARY_INT_SSE(VCVTPH2PS, cvtph2ps)
903 static inline void gen_unary_imm_sse(DisasContext *s, CPUX86State *env, X86DecodedInsn *decode,
904 SSEFunc_0_ppi xmm, SSEFunc_0_ppi ymm)
906 TCGv_i32 imm = tcg_constant8u_i32(decode->immediate);
908 xmm(OP_PTR0, OP_PTR1, imm);
910 ymm(OP_PTR0, OP_PTR1, imm);
914 #define UNARY_IMM_SSE(uname, lname) \
915 static void gen_##uname(DisasContext *s, CPUX86State *env, X86DecodedInsn *decode) \
917 gen_unary_imm_sse(s, env, decode, \
918 gen_helper_##lname##_xmm, \
919 gen_helper_##lname##_ymm); \
922 UNARY_IMM_SSE(PSHUFD, pshufd)
923 UNARY_IMM_SSE(PSHUFHW, pshufhw)
924 UNARY_IMM_SSE(PSHUFLW, pshuflw)
925 #define gen_helper_vpermq_xmm NULL
926 UNARY_IMM_SSE(VPERMQ, vpermq)
927 UNARY_IMM_SSE(VPERMILPS_i, vpermilps_imm)
928 UNARY_IMM_SSE(VPERMILPD_i, vpermilpd_imm)
930 static inline void gen_unary_imm_fp_sse(DisasContext *s, CPUX86State *env, X86DecodedInsn *decode,
931 SSEFunc_0_eppi xmm, SSEFunc_0_eppi ymm)
933 TCGv_i32 imm = tcg_constant8u_i32(decode->immediate);
935 xmm(tcg_env, OP_PTR0, OP_PTR1, imm);
937 ymm(tcg_env, OP_PTR0, OP_PTR1, imm);
941 #define UNARY_IMM_FP_SSE(uname, lname) \
942 static void gen_##uname(DisasContext *s, CPUX86State *env, X86DecodedInsn *decode) \
944 gen_unary_imm_fp_sse(s, env, decode, \
945 gen_helper_##lname##_xmm, \
946 gen_helper_##lname##_ymm); \
949 UNARY_IMM_FP_SSE(VROUNDPS, roundps)
950 UNARY_IMM_FP_SSE(VROUNDPD, roundpd)
952 static inline void gen_vexw_avx(DisasContext *s, CPUX86State *env, X86DecodedInsn *decode,
953 SSEFunc_0_eppp d_xmm, SSEFunc_0_eppp q_xmm,
954 SSEFunc_0_eppp d_ymm, SSEFunc_0_eppp q_ymm)
956 SSEFunc_0_eppp d = s->vex_l ? d_ymm : d_xmm;
957 SSEFunc_0_eppp q = s->vex_l ? q_ymm : q_xmm;
958 SSEFunc_0_eppp fn = s->vex_w ? q : d;
959 fn(tcg_env, OP_PTR0, OP_PTR1, OP_PTR2);
962 /* VEX.W affects whether to operate on 32- or 64-bit elements. */
963 #define VEXW_AVX(uname, lname) \
964 static void gen_##uname(DisasContext *s, CPUX86State *env, X86DecodedInsn *decode) \
966 gen_vexw_avx(s, env, decode, \
967 gen_helper_##lname##d_xmm, gen_helper_##lname##q_xmm, \
968 gen_helper_##lname##d_ymm, gen_helper_##lname##q_ymm); \
970 VEXW_AVX(VPSLLV, vpsllv)
971 VEXW_AVX(VPSRLV, vpsrlv)
972 VEXW_AVX(VPSRAV, vpsrav)
973 VEXW_AVX(VPMASKMOV, vpmaskmov)
975 /* Same as above, but with extra arguments to the helper. */
976 static inline void gen_vsib_avx(DisasContext *s, CPUX86State *env, X86DecodedInsn *decode,
977 SSEFunc_0_epppti d_xmm, SSEFunc_0_epppti q_xmm,
978 SSEFunc_0_epppti d_ymm, SSEFunc_0_epppti q_ymm)
980 SSEFunc_0_epppti d = s->vex_l ? d_ymm : d_xmm;
981 SSEFunc_0_epppti q = s->vex_l ? q_ymm : q_xmm;
982 SSEFunc_0_epppti fn = s->vex_w ? q : d;
983 TCGv_i32 scale = tcg_constant_i32(decode->mem.scale);
984 TCGv_ptr index = tcg_temp_new_ptr();
986 /* Pass third input as (index, base, scale) */
987 tcg_gen_addi_ptr(index, tcg_env, ZMM_OFFSET(decode->mem.index));
988 fn(tcg_env, OP_PTR0, OP_PTR1, index, s->A0, scale);
991 * There are two output operands, so zero OP1's high 128 bits
992 * in the VEX.128 case.
995 int ymmh_ofs = vector_elem_offset(&decode->op[1], MO_128, 1);
996 tcg_gen_gvec_dup_imm(MO_64, ymmh_ofs, 16, 16, 0);
999 #define VSIB_AVX(uname, lname) \
1000 static void gen_##uname(DisasContext *s, CPUX86State *env, X86DecodedInsn *decode) \
1002 gen_vsib_avx(s, env, decode, \
1003 gen_helper_##lname##d_xmm, gen_helper_##lname##q_xmm, \
1004 gen_helper_##lname##d_ymm, gen_helper_##lname##q_ymm); \
1006 VSIB_AVX(VPGATHERD, vpgatherd)
1007 VSIB_AVX(VPGATHERQ, vpgatherq)
1009 static void gen_ADCOX(DisasContext *s, CPUX86State *env, MemOp ot, int cc_op)
1012 TCGv carry_in = NULL;
1013 TCGv carry_out = (cc_op == CC_OP_ADCX ? cpu_cc_dst : cpu_cc_src2);
1016 if (cc_op == s->cc_op || s->cc_op == CC_OP_ADCOX) {
1017 /* Re-use the carry-out from a previous round. */
1018 carry_in = carry_out;
1020 /* We don't have a carry-in, get it out of EFLAGS. */
1021 if (s->cc_op != CC_OP_ADCX && s->cc_op != CC_OP_ADOX) {
1022 gen_compute_eflags(s);
1025 tcg_gen_extract_tl(carry_in, cpu_cc_src,
1026 ctz32(cc_op == CC_OP_ADCX ? CC_C : CC_O), 1);
1030 #ifdef TARGET_X86_64
1032 /* If TL is 64-bit just do everything in 64-bit arithmetic. */
1033 tcg_gen_ext32u_tl(s->T0, s->T0);
1034 tcg_gen_ext32u_tl(s->T1, s->T1);
1035 tcg_gen_add_i64(s->T0, s->T0, s->T1);
1036 tcg_gen_add_i64(s->T0, s->T0, carry_in);
1037 tcg_gen_shri_i64(carry_out, s->T0, 32);
1041 zero = tcg_constant_tl(0);
1042 tcg_gen_add2_tl(s->T0, carry_out, s->T0, zero, carry_in, zero);
1043 tcg_gen_add2_tl(s->T0, carry_out, s->T0, carry_out, s->T1, zero);
1047 opposite_cc_op = cc_op == CC_OP_ADCX ? CC_OP_ADOX : CC_OP_ADCX;
1048 if (s->cc_op == CC_OP_ADCOX || s->cc_op == opposite_cc_op) {
1049 /* Merge with the carry-out from the opposite instruction. */
1050 set_cc_op(s, CC_OP_ADCOX);
1052 set_cc_op(s, cc_op);
1056 static void gen_ADCX(DisasContext *s, CPUX86State *env, X86DecodedInsn *decode)
1058 gen_ADCOX(s, env, decode->op[0].ot, CC_OP_ADCX);
1061 static void gen_ADOX(DisasContext *s, CPUX86State *env, X86DecodedInsn *decode)
1063 gen_ADCOX(s, env, decode->op[0].ot, CC_OP_ADOX);
1066 static void gen_ANDN(DisasContext *s, CPUX86State *env, X86DecodedInsn *decode)
1068 MemOp ot = decode->op[0].ot;
1070 tcg_gen_andc_tl(s->T0, s->T1, s->T0);
1071 gen_op_update1_cc(s);
1072 set_cc_op(s, CC_OP_LOGICB + ot);
1075 static void gen_BEXTR(DisasContext *s, CPUX86State *env, X86DecodedInsn *decode)
1077 MemOp ot = decode->op[0].ot;
1078 TCGv bound = tcg_constant_tl(ot == MO_64 ? 63 : 31);
1079 TCGv zero = tcg_constant_tl(0);
1080 TCGv mone = tcg_constant_tl(-1);
1083 * Extract START, and shift the operand.
1084 * Shifts larger than operand size get zeros.
1086 tcg_gen_ext8u_tl(s->A0, s->T1);
1087 if (TARGET_LONG_BITS == 64 && ot == MO_32) {
1088 tcg_gen_ext32u_tl(s->T0, s->T0);
1090 tcg_gen_shr_tl(s->T0, s->T0, s->A0);
1092 tcg_gen_movcond_tl(TCG_COND_LEU, s->T0, s->A0, bound, s->T0, zero);
1095 * Extract the LEN into an inverse mask. Lengths larger than
1096 * operand size get all zeros, length 0 gets all ones.
1098 tcg_gen_extract_tl(s->A0, s->T1, 8, 8);
1099 tcg_gen_shl_tl(s->T1, mone, s->A0);
1100 tcg_gen_movcond_tl(TCG_COND_LEU, s->T1, s->A0, bound, s->T1, zero);
1101 tcg_gen_andc_tl(s->T0, s->T0, s->T1);
1103 gen_op_update1_cc(s);
1104 set_cc_op(s, CC_OP_LOGICB + ot);
1107 static void gen_BLSI(DisasContext *s, CPUX86State *env, X86DecodedInsn *decode)
1109 MemOp ot = decode->op[0].ot;
1111 tcg_gen_mov_tl(cpu_cc_src, s->T0);
1112 tcg_gen_neg_tl(s->T1, s->T0);
1113 tcg_gen_and_tl(s->T0, s->T0, s->T1);
1114 tcg_gen_mov_tl(cpu_cc_dst, s->T0);
1115 set_cc_op(s, CC_OP_BMILGB + ot);
1118 static void gen_BLSMSK(DisasContext *s, CPUX86State *env, X86DecodedInsn *decode)
1120 MemOp ot = decode->op[0].ot;
1122 tcg_gen_mov_tl(cpu_cc_src, s->T0);
1123 tcg_gen_subi_tl(s->T1, s->T0, 1);
1124 tcg_gen_xor_tl(s->T0, s->T0, s->T1);
1125 tcg_gen_mov_tl(cpu_cc_dst, s->T0);
1126 set_cc_op(s, CC_OP_BMILGB + ot);
1129 static void gen_BLSR(DisasContext *s, CPUX86State *env, X86DecodedInsn *decode)
1131 MemOp ot = decode->op[0].ot;
1133 tcg_gen_mov_tl(cpu_cc_src, s->T0);
1134 tcg_gen_subi_tl(s->T1, s->T0, 1);
1135 tcg_gen_and_tl(s->T0, s->T0, s->T1);
1136 tcg_gen_mov_tl(cpu_cc_dst, s->T0);
1137 set_cc_op(s, CC_OP_BMILGB + ot);
1140 static void gen_BZHI(DisasContext *s, CPUX86State *env, X86DecodedInsn *decode)
1142 MemOp ot = decode->op[0].ot;
1143 TCGv bound = tcg_constant_tl(ot == MO_64 ? 63 : 31);
1144 TCGv zero = tcg_constant_tl(0);
1145 TCGv mone = tcg_constant_tl(-1);
1147 tcg_gen_ext8u_tl(s->T1, s->T1);
1150 * Note that since we're using BMILG (in order to get O
1151 * cleared) we need to store the inverse into C.
1153 tcg_gen_setcond_tl(TCG_COND_LEU, cpu_cc_src, s->T1, bound);
1155 tcg_gen_shl_tl(s->A0, mone, s->T1);
1156 tcg_gen_movcond_tl(TCG_COND_LEU, s->A0, s->T1, bound, s->A0, zero);
1157 tcg_gen_andc_tl(s->T0, s->T0, s->A0);
1159 gen_op_update1_cc(s);
1160 set_cc_op(s, CC_OP_BMILGB + ot);
1163 static void gen_CRC32(DisasContext *s, CPUX86State *env, X86DecodedInsn *decode)
1165 MemOp ot = decode->op[2].ot;
1167 tcg_gen_trunc_tl_i32(s->tmp2_i32, s->T0);
1168 gen_helper_crc32(s->T0, s->tmp2_i32, s->T1, tcg_constant_i32(8 << ot));
1171 static void gen_CVTPI2Px(DisasContext *s, CPUX86State *env, X86DecodedInsn *decode)
1173 gen_helper_enter_mmx(tcg_env);
1174 if (s->prefix & PREFIX_DATA) {
1175 gen_helper_cvtpi2pd(tcg_env, OP_PTR0, OP_PTR2);
1177 gen_helper_cvtpi2ps(tcg_env, OP_PTR0, OP_PTR2);
1181 static void gen_CVTPx2PI(DisasContext *s, CPUX86State *env, X86DecodedInsn *decode)
1183 gen_helper_enter_mmx(tcg_env);
1184 if (s->prefix & PREFIX_DATA) {
1185 gen_helper_cvtpd2pi(tcg_env, OP_PTR0, OP_PTR2);
1187 gen_helper_cvtps2pi(tcg_env, OP_PTR0, OP_PTR2);
1191 static void gen_CVTTPx2PI(DisasContext *s, CPUX86State *env, X86DecodedInsn *decode)
1193 gen_helper_enter_mmx(tcg_env);
1194 if (s->prefix & PREFIX_DATA) {
1195 gen_helper_cvttpd2pi(tcg_env, OP_PTR0, OP_PTR2);
1197 gen_helper_cvttps2pi(tcg_env, OP_PTR0, OP_PTR2);
1201 static void gen_EMMS(DisasContext *s, CPUX86State *env, X86DecodedInsn *decode)
1203 gen_helper_emms(tcg_env);
1206 static void gen_EXTRQ_i(DisasContext *s, CPUX86State *env, X86DecodedInsn *decode)
1208 TCGv_i32 length = tcg_constant_i32(decode->immediate & 63);
1209 TCGv_i32 index = tcg_constant_i32((decode->immediate >> 8) & 63);
1211 gen_helper_extrq_i(tcg_env, OP_PTR0, index, length);
1214 static void gen_EXTRQ_r(DisasContext *s, CPUX86State *env, X86DecodedInsn *decode)
1216 gen_helper_extrq_r(tcg_env, OP_PTR0, OP_PTR2);
1219 static void gen_INSERTQ_i(DisasContext *s, CPUX86State *env, X86DecodedInsn *decode)
1221 TCGv_i32 length = tcg_constant_i32(decode->immediate & 63);
1222 TCGv_i32 index = tcg_constant_i32((decode->immediate >> 8) & 63);
1224 gen_helper_insertq_i(tcg_env, OP_PTR0, OP_PTR1, index, length);
1227 static void gen_INSERTQ_r(DisasContext *s, CPUX86State *env, X86DecodedInsn *decode)
1229 gen_helper_insertq_r(tcg_env, OP_PTR0, OP_PTR2);
1232 static void gen_LDMXCSR(DisasContext *s, CPUX86State *env, X86DecodedInsn *decode)
1234 tcg_gen_trunc_tl_i32(s->tmp2_i32, s->T1);
1235 gen_helper_ldmxcsr(tcg_env, s->tmp2_i32);
1238 static void gen_MASKMOV(DisasContext *s, CPUX86State *env, X86DecodedInsn *decode)
1240 gen_lea_v_seg(s, s->aflag, cpu_regs[R_EDI], R_DS, s->override);
1242 if (s->prefix & PREFIX_DATA) {
1243 gen_helper_maskmov_xmm(tcg_env, OP_PTR1, OP_PTR2, s->A0);
1245 gen_helper_maskmov_mmx(tcg_env, OP_PTR1, OP_PTR2, s->A0);
1249 static void gen_MOVBE(DisasContext *s, CPUX86State *env, X86DecodedInsn *decode)
1251 MemOp ot = decode->op[0].ot;
1253 /* M operand type does not load/store */
1254 if (decode->e.op0 == X86_TYPE_M) {
1255 tcg_gen_qemu_st_tl(s->T0, s->A0, s->mem_index, ot | MO_BE);
1257 tcg_gen_qemu_ld_tl(s->T0, s->A0, s->mem_index, ot | MO_BE);
1261 static void gen_MOVD_from(DisasContext *s, CPUX86State *env, X86DecodedInsn *decode)
1263 MemOp ot = decode->op[2].ot;
1267 #ifdef TARGET_X86_64
1268 tcg_gen_ld32u_tl(s->T0, tcg_env, decode->op[2].offset);
1272 tcg_gen_ld_tl(s->T0, tcg_env, decode->op[2].offset);
1279 static void gen_MOVD_to(DisasContext *s, CPUX86State *env, X86DecodedInsn *decode)
1281 MemOp ot = decode->op[2].ot;
1282 int vec_len = vector_len(s, decode);
1283 int lo_ofs = vector_elem_offset(&decode->op[0], ot, 0);
1285 tcg_gen_gvec_dup_imm(MO_64, decode->op[0].offset, vec_len, vec_len, 0);
1289 #ifdef TARGET_X86_64
1290 tcg_gen_st32_tl(s->T1, tcg_env, lo_ofs);
1294 tcg_gen_st_tl(s->T1, tcg_env, lo_ofs);
1297 g_assert_not_reached();
1301 static void gen_MOVDQ(DisasContext *s, CPUX86State *env, X86DecodedInsn *decode)
1303 gen_store_sse(s, decode, decode->op[2].offset);
1306 static void gen_MOVMSK(DisasContext *s, CPUX86State *env, X86DecodedInsn *decode)
1308 typeof(gen_helper_movmskps_ymm) *ps, *pd, *fn;
1309 ps = s->vex_l ? gen_helper_movmskps_ymm : gen_helper_movmskps_xmm;
1310 pd = s->vex_l ? gen_helper_movmskpd_ymm : gen_helper_movmskpd_xmm;
1311 fn = s->prefix & PREFIX_DATA ? pd : ps;
1312 fn(s->tmp2_i32, tcg_env, OP_PTR2);
1313 tcg_gen_extu_i32_tl(s->T0, s->tmp2_i32);
1316 static void gen_MOVQ(DisasContext *s, CPUX86State *env, X86DecodedInsn *decode)
1318 int vec_len = vector_len(s, decode);
1319 int lo_ofs = vector_elem_offset(&decode->op[0], MO_64, 0);
1321 tcg_gen_ld_i64(s->tmp1_i64, tcg_env, decode->op[2].offset);
1322 if (decode->op[0].has_ea) {
1323 tcg_gen_qemu_st_i64(s->tmp1_i64, s->A0, s->mem_index, MO_LEUQ);
1326 * tcg_gen_gvec_dup_i64(MO_64, op0.offset, 8, vec_len, s->tmp1_64) would
1327 * seem to work, but it does not on big-endian platforms; the cleared parts
1328 * are always at higher addresses, but cross-endian emulation inverts the
1329 * byte order so that the cleared parts need to be at *lower* addresses.
1330 * Because oprsz is 8, we see this here even for SSE; but more in general,
1331 * it disqualifies using oprsz < maxsz to emulate VEX128.
1333 tcg_gen_gvec_dup_imm(MO_64, decode->op[0].offset, vec_len, vec_len, 0);
1334 tcg_gen_st_i64(s->tmp1_i64, tcg_env, lo_ofs);
1338 static void gen_MOVq_dq(DisasContext *s, CPUX86State *env, X86DecodedInsn *decode)
1340 gen_helper_enter_mmx(tcg_env);
1341 /* Otherwise the same as any other movq. */
1342 return gen_MOVQ(s, env, decode);
1345 static void gen_MULX(DisasContext *s, CPUX86State *env, X86DecodedInsn *decode)
1347 MemOp ot = decode->op[0].ot;
1349 /* low part of result in VEX.vvvv, high in MODRM */
1352 #ifdef TARGET_X86_64
1353 tcg_gen_trunc_tl_i32(s->tmp2_i32, s->T0);
1354 tcg_gen_trunc_tl_i32(s->tmp3_i32, s->T1);
1355 tcg_gen_mulu2_i32(s->tmp2_i32, s->tmp3_i32,
1356 s->tmp2_i32, s->tmp3_i32);
1357 tcg_gen_extu_i32_tl(cpu_regs[s->vex_v], s->tmp2_i32);
1358 tcg_gen_extu_i32_tl(s->T0, s->tmp3_i32);
1363 tcg_gen_mulu2_tl(cpu_regs[s->vex_v], s->T0, s->T0, s->T1);
1367 g_assert_not_reached();
1371 static void gen_PALIGNR(DisasContext *s, CPUX86State *env, X86DecodedInsn *decode)
1373 TCGv_i32 imm = tcg_constant8u_i32(decode->immediate);
1374 if (!(s->prefix & PREFIX_DATA)) {
1375 gen_helper_palignr_mmx(tcg_env, OP_PTR0, OP_PTR1, OP_PTR2, imm);
1376 } else if (!s->vex_l) {
1377 gen_helper_palignr_xmm(tcg_env, OP_PTR0, OP_PTR1, OP_PTR2, imm);
1379 gen_helper_palignr_ymm(tcg_env, OP_PTR0, OP_PTR1, OP_PTR2, imm);
1383 static void gen_PANDN(DisasContext *s, CPUX86State *env, X86DecodedInsn *decode)
1385 int vec_len = vector_len(s, decode);
1387 /* Careful, operand order is reversed! */
1388 tcg_gen_gvec_andc(MO_64,
1389 decode->op[0].offset, decode->op[2].offset,
1390 decode->op[1].offset, vec_len, vec_len);
1393 static void gen_PCMPESTRI(DisasContext *s, CPUX86State *env, X86DecodedInsn *decode)
1395 TCGv_i32 imm = tcg_constant8u_i32(decode->immediate);
1396 gen_helper_pcmpestri_xmm(tcg_env, OP_PTR1, OP_PTR2, imm);
1397 set_cc_op(s, CC_OP_EFLAGS);
1400 static void gen_PCMPESTRM(DisasContext *s, CPUX86State *env, X86DecodedInsn *decode)
1402 TCGv_i32 imm = tcg_constant8u_i32(decode->immediate);
1403 gen_helper_pcmpestrm_xmm(tcg_env, OP_PTR1, OP_PTR2, imm);
1404 set_cc_op(s, CC_OP_EFLAGS);
1405 if ((s->prefix & PREFIX_VEX) && !s->vex_l) {
1406 tcg_gen_gvec_dup_imm(MO_64, offsetof(CPUX86State, xmm_regs[0].ZMM_X(1)),
1411 static void gen_PCMPISTRI(DisasContext *s, CPUX86State *env, X86DecodedInsn *decode)
1413 TCGv_i32 imm = tcg_constant8u_i32(decode->immediate);
1414 gen_helper_pcmpistri_xmm(tcg_env, OP_PTR1, OP_PTR2, imm);
1415 set_cc_op(s, CC_OP_EFLAGS);
1418 static void gen_PCMPISTRM(DisasContext *s, CPUX86State *env, X86DecodedInsn *decode)
1420 TCGv_i32 imm = tcg_constant8u_i32(decode->immediate);
1421 gen_helper_pcmpistrm_xmm(tcg_env, OP_PTR1, OP_PTR2, imm);
1422 set_cc_op(s, CC_OP_EFLAGS);
1423 if ((s->prefix & PREFIX_VEX) && !s->vex_l) {
1424 tcg_gen_gvec_dup_imm(MO_64, offsetof(CPUX86State, xmm_regs[0].ZMM_X(1)),
1429 static void gen_PDEP(DisasContext *s, CPUX86State *env, X86DecodedInsn *decode)
1431 MemOp ot = decode->op[1].ot;
1433 tcg_gen_ext32u_tl(s->T0, s->T0);
1435 gen_helper_pdep(s->T0, s->T0, s->T1);
1438 static void gen_PEXT(DisasContext *s, CPUX86State *env, X86DecodedInsn *decode)
1440 MemOp ot = decode->op[1].ot;
1442 tcg_gen_ext32u_tl(s->T0, s->T0);
1444 gen_helper_pext(s->T0, s->T0, s->T1);
1447 static inline void gen_pextr(DisasContext *s, CPUX86State *env, X86DecodedInsn *decode, MemOp ot)
1449 int vec_len = vector_len(s, decode);
1450 int mask = (vec_len >> ot) - 1;
1451 int val = decode->immediate & mask;
1455 tcg_gen_ld8u_tl(s->T0, tcg_env, vector_elem_offset(&decode->op[1], ot, val));
1458 tcg_gen_ld16u_tl(s->T0, tcg_env, vector_elem_offset(&decode->op[1], ot, val));
1461 #ifdef TARGET_X86_64
1462 tcg_gen_ld32u_tl(s->T0, tcg_env, vector_elem_offset(&decode->op[1], ot, val));
1466 tcg_gen_ld_tl(s->T0, tcg_env, vector_elem_offset(&decode->op[1], ot, val));
1473 static void gen_PEXTRB(DisasContext *s, CPUX86State *env, X86DecodedInsn *decode)
1475 gen_pextr(s, env, decode, MO_8);
1478 static void gen_PEXTRW(DisasContext *s, CPUX86State *env, X86DecodedInsn *decode)
1480 gen_pextr(s, env, decode, MO_16);
1483 static void gen_PEXTR(DisasContext *s, CPUX86State *env, X86DecodedInsn *decode)
1485 MemOp ot = decode->op[0].ot;
1486 gen_pextr(s, env, decode, ot);
1489 static inline void gen_pinsr(DisasContext *s, CPUX86State *env, X86DecodedInsn *decode, MemOp ot)
1491 int vec_len = vector_len(s, decode);
1492 int mask = (vec_len >> ot) - 1;
1493 int val = decode->immediate & mask;
1495 if (decode->op[1].offset != decode->op[0].offset) {
1496 assert(vec_len == 16);
1497 gen_store_sse(s, decode, decode->op[1].offset);
1502 tcg_gen_st8_tl(s->T1, tcg_env, vector_elem_offset(&decode->op[0], ot, val));
1505 tcg_gen_st16_tl(s->T1, tcg_env, vector_elem_offset(&decode->op[0], ot, val));
1508 #ifdef TARGET_X86_64
1509 tcg_gen_st32_tl(s->T1, tcg_env, vector_elem_offset(&decode->op[0], ot, val));
1513 tcg_gen_st_tl(s->T1, tcg_env, vector_elem_offset(&decode->op[0], ot, val));
1520 static void gen_PINSRB(DisasContext *s, CPUX86State *env, X86DecodedInsn *decode)
1522 gen_pinsr(s, env, decode, MO_8);
1525 static void gen_PINSRW(DisasContext *s, CPUX86State *env, X86DecodedInsn *decode)
1527 gen_pinsr(s, env, decode, MO_16);
1530 static void gen_PINSR(DisasContext *s, CPUX86State *env, X86DecodedInsn *decode)
1532 gen_pinsr(s, env, decode, decode->op[2].ot);
1535 static void gen_pmovmskb_i64(TCGv_i64 d, TCGv_i64 s)
1537 TCGv_i64 t = tcg_temp_new_i64();
1539 tcg_gen_andi_i64(d, s, 0x8080808080808080ull);
1542 * After each shift+or pair:
1543 * 0: a.......b.......c.......d.......e.......f.......g.......h.......
1544 * 7: ab......bc......cd......de......ef......fg......gh......h.......
1545 * 14: abcd....bcde....cdef....defg....efgh....fgh.....gh......h.......
1546 * 28: abcdefghbcdefgh.cdefgh..defgh...efgh....fgh.....gh......h.......
1547 * The result is left in the high bits of the word.
1549 tcg_gen_shli_i64(t, d, 7);
1550 tcg_gen_or_i64(d, d, t);
1551 tcg_gen_shli_i64(t, d, 14);
1552 tcg_gen_or_i64(d, d, t);
1553 tcg_gen_shli_i64(t, d, 28);
1554 tcg_gen_or_i64(d, d, t);
1557 static void gen_pmovmskb_vec(unsigned vece, TCGv_vec d, TCGv_vec s)
1559 TCGv_vec t = tcg_temp_new_vec_matching(d);
1560 TCGv_vec m = tcg_constant_vec_matching(d, MO_8, 0x80);
1563 tcg_gen_and_vec(vece, d, s, m);
1564 tcg_gen_shli_vec(vece, t, d, 7);
1565 tcg_gen_or_vec(vece, d, d, t);
1566 tcg_gen_shli_vec(vece, t, d, 14);
1567 tcg_gen_or_vec(vece, d, d, t);
1568 tcg_gen_shli_vec(vece, t, d, 28);
1569 tcg_gen_or_vec(vece, d, d, t);
1572 #ifdef TARGET_X86_64
1573 #define TCG_TARGET_HAS_extract2_tl TCG_TARGET_HAS_extract2_i64
1575 #define TCG_TARGET_HAS_extract2_tl TCG_TARGET_HAS_extract2_i32
1578 static void gen_PMOVMSKB(DisasContext *s, CPUX86State *env, X86DecodedInsn *decode)
1580 static const TCGOpcode vecop_list[] = { INDEX_op_shli_vec, 0 };
1581 static const GVecGen2 g = {
1582 .fni8 = gen_pmovmskb_i64,
1583 .fniv = gen_pmovmskb_vec,
1584 .opt_opc = vecop_list,
1586 .prefer_i64 = TCG_TARGET_REG_BITS == 64
1588 MemOp ot = decode->op[2].ot;
1589 int vec_len = vector_len(s, decode);
1590 TCGv t = tcg_temp_new();
1592 tcg_gen_gvec_2(offsetof(CPUX86State, xmm_t0) + xmm_offset(ot), decode->op[2].offset,
1593 vec_len, vec_len, &g);
1594 tcg_gen_ld8u_tl(s->T0, tcg_env, offsetof(CPUX86State, xmm_t0.ZMM_B(vec_len - 1)));
1595 while (vec_len > 8) {
1597 if (TCG_TARGET_HAS_extract2_tl) {
1599 * Load the next byte of the result into the high byte of T.
1600 * TCG does a similar expansion of deposit to shl+extract2; by
1601 * loading the whole word, the shift left is avoided.
1603 #ifdef TARGET_X86_64
1604 tcg_gen_ld_tl(t, tcg_env, offsetof(CPUX86State, xmm_t0.ZMM_Q((vec_len - 1) / 8)));
1606 tcg_gen_ld_tl(t, tcg_env, offsetof(CPUX86State, xmm_t0.ZMM_L((vec_len - 1) / 4)));
1609 tcg_gen_extract2_tl(s->T0, t, s->T0, TARGET_LONG_BITS - 8);
1612 * The _previous_ value is deposited into bits 8 and higher of t. Because
1613 * those bits are known to be zero after ld8u, this becomes a shift+or
1614 * if deposit is not available.
1616 tcg_gen_ld8u_tl(t, tcg_env, offsetof(CPUX86State, xmm_t0.ZMM_B(vec_len - 1)));
1617 tcg_gen_deposit_tl(s->T0, t, s->T0, 8, TARGET_LONG_BITS - 8);
1622 static void gen_PSHUFW(DisasContext *s, CPUX86State *env, X86DecodedInsn *decode)
1624 TCGv_i32 imm = tcg_constant8u_i32(decode->immediate);
1625 gen_helper_pshufw_mmx(OP_PTR0, OP_PTR1, imm);
1628 static void gen_PSRLW_i(DisasContext *s, CPUX86State *env, X86DecodedInsn *decode)
1630 int vec_len = vector_len(s, decode);
1632 if (decode->immediate >= 16) {
1633 tcg_gen_gvec_dup_imm(MO_64, decode->op[0].offset, vec_len, vec_len, 0);
1635 tcg_gen_gvec_shri(MO_16,
1636 decode->op[0].offset, decode->op[1].offset,
1637 decode->immediate, vec_len, vec_len);
1641 static void gen_PSLLW_i(DisasContext *s, CPUX86State *env, X86DecodedInsn *decode)
1643 int vec_len = vector_len(s, decode);
1645 if (decode->immediate >= 16) {
1646 tcg_gen_gvec_dup_imm(MO_64, decode->op[0].offset, vec_len, vec_len, 0);
1648 tcg_gen_gvec_shli(MO_16,
1649 decode->op[0].offset, decode->op[1].offset,
1650 decode->immediate, vec_len, vec_len);
1654 static void gen_PSRAW_i(DisasContext *s, CPUX86State *env, X86DecodedInsn *decode)
1656 int vec_len = vector_len(s, decode);
1658 if (decode->immediate >= 16) {
1659 decode->immediate = 15;
1661 tcg_gen_gvec_sari(MO_16,
1662 decode->op[0].offset, decode->op[1].offset,
1663 decode->immediate, vec_len, vec_len);
1666 static void gen_PSRLD_i(DisasContext *s, CPUX86State *env, X86DecodedInsn *decode)
1668 int vec_len = vector_len(s, decode);
1670 if (decode->immediate >= 32) {
1671 tcg_gen_gvec_dup_imm(MO_64, decode->op[0].offset, vec_len, vec_len, 0);
1673 tcg_gen_gvec_shri(MO_32,
1674 decode->op[0].offset, decode->op[1].offset,
1675 decode->immediate, vec_len, vec_len);
1679 static void gen_PSLLD_i(DisasContext *s, CPUX86State *env, X86DecodedInsn *decode)
1681 int vec_len = vector_len(s, decode);
1683 if (decode->immediate >= 32) {
1684 tcg_gen_gvec_dup_imm(MO_64, decode->op[0].offset, vec_len, vec_len, 0);
1686 tcg_gen_gvec_shli(MO_32,
1687 decode->op[0].offset, decode->op[1].offset,
1688 decode->immediate, vec_len, vec_len);
1692 static void gen_PSRAD_i(DisasContext *s, CPUX86State *env, X86DecodedInsn *decode)
1694 int vec_len = vector_len(s, decode);
1696 if (decode->immediate >= 32) {
1697 decode->immediate = 31;
1699 tcg_gen_gvec_sari(MO_32,
1700 decode->op[0].offset, decode->op[1].offset,
1701 decode->immediate, vec_len, vec_len);
1704 static void gen_PSRLQ_i(DisasContext *s, CPUX86State *env, X86DecodedInsn *decode)
1706 int vec_len = vector_len(s, decode);
1708 if (decode->immediate >= 64) {
1709 tcg_gen_gvec_dup_imm(MO_64, decode->op[0].offset, vec_len, vec_len, 0);
1711 tcg_gen_gvec_shri(MO_64,
1712 decode->op[0].offset, decode->op[1].offset,
1713 decode->immediate, vec_len, vec_len);
1717 static void gen_PSLLQ_i(DisasContext *s, CPUX86State *env, X86DecodedInsn *decode)
1719 int vec_len = vector_len(s, decode);
1721 if (decode->immediate >= 64) {
1722 tcg_gen_gvec_dup_imm(MO_64, decode->op[0].offset, vec_len, vec_len, 0);
1724 tcg_gen_gvec_shli(MO_64,
1725 decode->op[0].offset, decode->op[1].offset,
1726 decode->immediate, vec_len, vec_len);
1730 static TCGv_ptr make_imm8u_xmm_vec(uint8_t imm, int vec_len)
1732 MemOp ot = vec_len == 16 ? MO_128 : MO_256;
1733 TCGv_i32 imm_v = tcg_constant8u_i32(imm);
1734 TCGv_ptr ptr = tcg_temp_new_ptr();
1736 tcg_gen_gvec_dup_imm(MO_64, offsetof(CPUX86State, xmm_t0) + xmm_offset(ot),
1737 vec_len, vec_len, 0);
1739 tcg_gen_addi_ptr(ptr, tcg_env, offsetof(CPUX86State, xmm_t0));
1740 tcg_gen_st_i32(imm_v, tcg_env, offsetof(CPUX86State, xmm_t0.ZMM_L(0)));
1744 static void gen_PSRLDQ_i(DisasContext *s, CPUX86State *env, X86DecodedInsn *decode)
1746 int vec_len = vector_len(s, decode);
1747 TCGv_ptr imm_vec = make_imm8u_xmm_vec(decode->immediate, vec_len);
1750 gen_helper_psrldq_ymm(tcg_env, OP_PTR0, OP_PTR1, imm_vec);
1752 gen_helper_psrldq_xmm(tcg_env, OP_PTR0, OP_PTR1, imm_vec);
1756 static void gen_PSLLDQ_i(DisasContext *s, CPUX86State *env, X86DecodedInsn *decode)
1758 int vec_len = vector_len(s, decode);
1759 TCGv_ptr imm_vec = make_imm8u_xmm_vec(decode->immediate, vec_len);
1762 gen_helper_pslldq_ymm(tcg_env, OP_PTR0, OP_PTR1, imm_vec);
1764 gen_helper_pslldq_xmm(tcg_env, OP_PTR0, OP_PTR1, imm_vec);
1768 static void gen_RORX(DisasContext *s, CPUX86State *env, X86DecodedInsn *decode)
1770 MemOp ot = decode->op[0].ot;
1771 int mask = ot == MO_64 ? 63 : 31;
1772 int b = decode->immediate & mask;
1776 #ifdef TARGET_X86_64
1777 tcg_gen_trunc_tl_i32(s->tmp2_i32, s->T0);
1778 tcg_gen_rotri_i32(s->tmp2_i32, s->tmp2_i32, b);
1779 tcg_gen_extu_i32_tl(s->T0, s->tmp2_i32);
1784 tcg_gen_rotri_tl(s->T0, s->T0, b);
1788 g_assert_not_reached();
1792 static void gen_SARX(DisasContext *s, CPUX86State *env, X86DecodedInsn *decode)
1794 MemOp ot = decode->op[0].ot;
1797 mask = ot == MO_64 ? 63 : 31;
1798 tcg_gen_andi_tl(s->T1, s->T1, mask);
1800 tcg_gen_ext32s_tl(s->T0, s->T0);
1802 tcg_gen_sar_tl(s->T0, s->T0, s->T1);
1805 static void gen_SHA1NEXTE(DisasContext *s, CPUX86State *env, X86DecodedInsn *decode)
1807 gen_helper_sha1nexte(OP_PTR0, OP_PTR1, OP_PTR2);
1810 static void gen_SHA1MSG1(DisasContext *s, CPUX86State *env, X86DecodedInsn *decode)
1812 gen_helper_sha1msg1(OP_PTR0, OP_PTR1, OP_PTR2);
1815 static void gen_SHA1MSG2(DisasContext *s, CPUX86State *env, X86DecodedInsn *decode)
1817 gen_helper_sha1msg2(OP_PTR0, OP_PTR1, OP_PTR2);
1820 static void gen_SHA1RNDS4(DisasContext *s, CPUX86State *env, X86DecodedInsn *decode)
1822 switch(decode->immediate & 3) {
1824 gen_helper_sha1rnds4_f0(OP_PTR0, OP_PTR0, OP_PTR1);
1827 gen_helper_sha1rnds4_f1(OP_PTR0, OP_PTR0, OP_PTR1);
1830 gen_helper_sha1rnds4_f2(OP_PTR0, OP_PTR0, OP_PTR1);
1833 gen_helper_sha1rnds4_f3(OP_PTR0, OP_PTR0, OP_PTR1);
1838 static void gen_SHA256MSG1(DisasContext *s, CPUX86State *env, X86DecodedInsn *decode)
1840 gen_helper_sha256msg1(OP_PTR0, OP_PTR1, OP_PTR2);
1843 static void gen_SHA256MSG2(DisasContext *s, CPUX86State *env, X86DecodedInsn *decode)
1845 gen_helper_sha256msg2(OP_PTR0, OP_PTR1, OP_PTR2);
1848 static void gen_SHA256RNDS2(DisasContext *s, CPUX86State *env, X86DecodedInsn *decode)
1850 TCGv_i32 wk0 = tcg_temp_new_i32();
1851 TCGv_i32 wk1 = tcg_temp_new_i32();
1853 tcg_gen_ld_i32(wk0, tcg_env, ZMM_OFFSET(0) + offsetof(ZMMReg, ZMM_L(0)));
1854 tcg_gen_ld_i32(wk1, tcg_env, ZMM_OFFSET(0) + offsetof(ZMMReg, ZMM_L(1)));
1856 gen_helper_sha256rnds2(OP_PTR0, OP_PTR1, OP_PTR2, wk0, wk1);
1859 static void gen_SHLX(DisasContext *s, CPUX86State *env, X86DecodedInsn *decode)
1861 MemOp ot = decode->op[0].ot;
1864 mask = ot == MO_64 ? 63 : 31;
1865 tcg_gen_andi_tl(s->T1, s->T1, mask);
1866 tcg_gen_shl_tl(s->T0, s->T0, s->T1);
1869 static void gen_SHRX(DisasContext *s, CPUX86State *env, X86DecodedInsn *decode)
1871 MemOp ot = decode->op[0].ot;
1874 mask = ot == MO_64 ? 63 : 31;
1875 tcg_gen_andi_tl(s->T1, s->T1, mask);
1877 tcg_gen_ext32u_tl(s->T0, s->T0);
1879 tcg_gen_shr_tl(s->T0, s->T0, s->T1);
1882 static void gen_VAESKEYGEN(DisasContext *s, CPUX86State *env, X86DecodedInsn *decode)
1884 TCGv_i32 imm = tcg_constant8u_i32(decode->immediate);
1886 gen_helper_aeskeygenassist_xmm(tcg_env, OP_PTR0, OP_PTR1, imm);
1889 static void gen_STMXCSR(DisasContext *s, CPUX86State *env, X86DecodedInsn *decode)
1891 gen_helper_update_mxcsr(tcg_env);
1892 tcg_gen_ld32u_tl(s->T0, tcg_env, offsetof(CPUX86State, mxcsr));
1895 static void gen_VAESIMC(DisasContext *s, CPUX86State *env, X86DecodedInsn *decode)
1898 gen_helper_aesimc_xmm(tcg_env, OP_PTR0, OP_PTR2);
1902 * 00 = v*ps Vps, Hps, Wpd
1903 * 66 = v*pd Vpd, Hpd, Wps
1904 * f3 = v*ss Vss, Hss, Wps
1905 * f2 = v*sd Vsd, Hsd, Wps
1907 #define SSE_CMP(x) { \
1908 gen_helper_ ## x ## ps ## _xmm, gen_helper_ ## x ## pd ## _xmm, \
1909 gen_helper_ ## x ## ss, gen_helper_ ## x ## sd, \
1910 gen_helper_ ## x ## ps ## _ymm, gen_helper_ ## x ## pd ## _ymm}
1911 static const SSEFunc_0_eppp gen_helper_cmp_funcs[32][6] = {
1950 static void gen_VCMP(DisasContext *s, CPUX86State *env, X86DecodedInsn *decode)
1952 int index = decode->immediate & (s->prefix & PREFIX_VEX ? 31 : 7);
1954 s->prefix & PREFIX_REPZ ? 2 /* ss */ :
1955 s->prefix & PREFIX_REPNZ ? 3 /* sd */ :
1956 !!(s->prefix & PREFIX_DATA) /* pd */ + (s->vex_l << 2);
1958 gen_helper_cmp_funcs[index][b](tcg_env, OP_PTR0, OP_PTR1, OP_PTR2);
1961 static void gen_VCOMI(DisasContext *s, CPUX86State *env, X86DecodedInsn *decode)
1964 fn = s->prefix & PREFIX_DATA ? gen_helper_comisd : gen_helper_comiss;
1965 fn(tcg_env, OP_PTR1, OP_PTR2);
1966 set_cc_op(s, CC_OP_EFLAGS);
1969 static void gen_VCVTPD2PS(DisasContext *s, CPUX86State *env, X86DecodedInsn *decode)
1972 gen_helper_cvtpd2ps_ymm(tcg_env, OP_PTR0, OP_PTR2);
1974 gen_helper_cvtpd2ps_xmm(tcg_env, OP_PTR0, OP_PTR2);
1978 static void gen_VCVTPS2PD(DisasContext *s, CPUX86State *env, X86DecodedInsn *decode)
1981 gen_helper_cvtps2pd_ymm(tcg_env, OP_PTR0, OP_PTR2);
1983 gen_helper_cvtps2pd_xmm(tcg_env, OP_PTR0, OP_PTR2);
1987 static void gen_VCVTPS2PH(DisasContext *s, CPUX86State *env, X86DecodedInsn *decode)
1989 gen_unary_imm_fp_sse(s, env, decode,
1990 gen_helper_cvtps2ph_xmm,
1991 gen_helper_cvtps2ph_ymm);
1993 * VCVTPS2PH is the only instruction that performs an operation on a
1994 * register source and then *stores* into memory.
1996 if (decode->op[0].has_ea) {
1997 gen_store_sse(s, decode, decode->op[0].offset);
2001 static void gen_VCVTSD2SS(DisasContext *s, CPUX86State *env, X86DecodedInsn *decode)
2003 gen_helper_cvtsd2ss(tcg_env, OP_PTR0, OP_PTR1, OP_PTR2);
2006 static void gen_VCVTSS2SD(DisasContext *s, CPUX86State *env, X86DecodedInsn *decode)
2008 gen_helper_cvtss2sd(tcg_env, OP_PTR0, OP_PTR1, OP_PTR2);
2011 static void gen_VCVTSI2Sx(DisasContext *s, CPUX86State *env, X86DecodedInsn *decode)
2013 int vec_len = vector_len(s, decode);
2016 tcg_gen_gvec_mov(MO_64, decode->op[0].offset, decode->op[1].offset, vec_len, vec_len);
2018 #ifdef TARGET_X86_64
2019 MemOp ot = decode->op[2].ot;
2021 if (s->prefix & PREFIX_REPNZ) {
2022 gen_helper_cvtsq2sd(tcg_env, OP_PTR0, s->T1);
2024 gen_helper_cvtsq2ss(tcg_env, OP_PTR0, s->T1);
2029 tcg_gen_trunc_tl_i32(in, s->T1);
2034 if (s->prefix & PREFIX_REPNZ) {
2035 gen_helper_cvtsi2sd(tcg_env, OP_PTR0, in);
2037 gen_helper_cvtsi2ss(tcg_env, OP_PTR0, in);
2041 static inline void gen_VCVTtSx2SI(DisasContext *s, CPUX86State *env, X86DecodedInsn *decode,
2042 SSEFunc_i_ep ss2si, SSEFunc_l_ep ss2sq,
2043 SSEFunc_i_ep sd2si, SSEFunc_l_ep sd2sq)
2047 #ifdef TARGET_X86_64
2048 MemOp ot = decode->op[0].ot;
2050 if (s->prefix & PREFIX_REPNZ) {
2051 sd2sq(s->T0, tcg_env, OP_PTR2);
2053 ss2sq(s->T0, tcg_env, OP_PTR2);
2062 if (s->prefix & PREFIX_REPNZ) {
2063 sd2si(out, tcg_env, OP_PTR2);
2065 ss2si(out, tcg_env, OP_PTR2);
2067 #ifdef TARGET_X86_64
2068 tcg_gen_extu_i32_tl(s->T0, out);
2072 #ifndef TARGET_X86_64
2073 #define gen_helper_cvtss2sq NULL
2074 #define gen_helper_cvtsd2sq NULL
2075 #define gen_helper_cvttss2sq NULL
2076 #define gen_helper_cvttsd2sq NULL
2079 static void gen_VCVTSx2SI(DisasContext *s, CPUX86State *env, X86DecodedInsn *decode)
2081 gen_VCVTtSx2SI(s, env, decode,
2082 gen_helper_cvtss2si, gen_helper_cvtss2sq,
2083 gen_helper_cvtsd2si, gen_helper_cvtsd2sq);
2086 static void gen_VCVTTSx2SI(DisasContext *s, CPUX86State *env, X86DecodedInsn *decode)
2088 gen_VCVTtSx2SI(s, env, decode,
2089 gen_helper_cvttss2si, gen_helper_cvttss2sq,
2090 gen_helper_cvttsd2si, gen_helper_cvttsd2sq);
2093 static void gen_VEXTRACTx128(DisasContext *s, CPUX86State *env, X86DecodedInsn *decode)
2095 int mask = decode->immediate & 1;
2096 int src_ofs = vector_elem_offset(&decode->op[1], MO_128, mask);
2097 if (decode->op[0].has_ea) {
2098 /* VEX-only instruction, no alignment requirements. */
2099 gen_sto_env_A0(s, src_ofs, false);
2101 tcg_gen_gvec_mov(MO_64, decode->op[0].offset, src_ofs, 16, 16);
2105 static void gen_VEXTRACTPS(DisasContext *s, CPUX86State *env, X86DecodedInsn *decode)
2107 gen_pextr(s, env, decode, MO_32);
2110 static void gen_vinsertps(DisasContext *s, CPUX86State *env, X86DecodedInsn *decode)
2112 int val = decode->immediate;
2113 int dest_word = (val >> 4) & 3;
2114 int new_mask = (val & 15) | (1 << dest_word);
2119 if (new_mask == 15) {
2120 /* All zeroes except possibly for the inserted element */
2121 tcg_gen_gvec_dup_imm(MO_64, decode->op[0].offset, vec_len, vec_len, 0);
2122 } else if (decode->op[1].offset != decode->op[0].offset) {
2123 gen_store_sse(s, decode, decode->op[1].offset);
2126 if (new_mask != (val & 15)) {
2127 tcg_gen_st_i32(s->tmp2_i32, tcg_env,
2128 vector_elem_offset(&decode->op[0], MO_32, dest_word));
2131 if (new_mask != 15) {
2132 TCGv_i32 zero = tcg_constant_i32(0); /* float32_zero */
2134 for (i = 0; i < 4; i++) {
2135 if ((val >> i) & 1) {
2136 tcg_gen_st_i32(zero, tcg_env,
2137 vector_elem_offset(&decode->op[0], MO_32, i));
2143 static void gen_VINSERTPS_r(DisasContext *s, CPUX86State *env, X86DecodedInsn *decode)
2145 int val = decode->immediate;
2146 tcg_gen_ld_i32(s->tmp2_i32, tcg_env,
2147 vector_elem_offset(&decode->op[2], MO_32, (val >> 6) & 3));
2148 gen_vinsertps(s, env, decode);
2151 static void gen_VINSERTPS_m(DisasContext *s, CPUX86State *env, X86DecodedInsn *decode)
2153 tcg_gen_qemu_ld_i32(s->tmp2_i32, s->A0, s->mem_index, MO_LEUL);
2154 gen_vinsertps(s, env, decode);
2157 static void gen_VINSERTx128(DisasContext *s, CPUX86State *env, X86DecodedInsn *decode)
2159 int mask = decode->immediate & 1;
2160 tcg_gen_gvec_mov(MO_64,
2161 decode->op[0].offset + offsetof(YMMReg, YMM_X(mask)),
2162 decode->op[2].offset + offsetof(YMMReg, YMM_X(0)), 16, 16);
2163 tcg_gen_gvec_mov(MO_64,
2164 decode->op[0].offset + offsetof(YMMReg, YMM_X(!mask)),
2165 decode->op[1].offset + offsetof(YMMReg, YMM_X(!mask)), 16, 16);
2168 static inline void gen_maskmov(DisasContext *s, CPUX86State *env, X86DecodedInsn *decode,
2169 SSEFunc_0_eppt xmm, SSEFunc_0_eppt ymm)
2172 xmm(tcg_env, OP_PTR2, OP_PTR1, s->A0);
2174 ymm(tcg_env, OP_PTR2, OP_PTR1, s->A0);
2178 static void gen_VMASKMOVPD_st(DisasContext *s, CPUX86State *env, X86DecodedInsn *decode)
2180 gen_maskmov(s, env, decode, gen_helper_vpmaskmovq_st_xmm, gen_helper_vpmaskmovq_st_ymm);
2183 static void gen_VMASKMOVPS_st(DisasContext *s, CPUX86State *env, X86DecodedInsn *decode)
2185 gen_maskmov(s, env, decode, gen_helper_vpmaskmovd_st_xmm, gen_helper_vpmaskmovd_st_ymm);
2188 static void gen_VMOVHPx_ld(DisasContext *s, CPUX86State *env, X86DecodedInsn *decode)
2190 gen_ldq_env_A0(s, decode->op[0].offset + offsetof(XMMReg, XMM_Q(1)));
2191 if (decode->op[0].offset != decode->op[1].offset) {
2192 tcg_gen_ld_i64(s->tmp1_i64, tcg_env, decode->op[1].offset + offsetof(XMMReg, XMM_Q(0)));
2193 tcg_gen_st_i64(s->tmp1_i64, tcg_env, decode->op[0].offset + offsetof(XMMReg, XMM_Q(0)));
2197 static void gen_VMOVHPx_st(DisasContext *s, CPUX86State *env, X86DecodedInsn *decode)
2199 gen_stq_env_A0(s, decode->op[2].offset + offsetof(XMMReg, XMM_Q(1)));
2202 static void gen_VMOVHPx(DisasContext *s, CPUX86State *env, X86DecodedInsn *decode)
2204 if (decode->op[0].offset != decode->op[2].offset) {
2205 tcg_gen_ld_i64(s->tmp1_i64, tcg_env, decode->op[2].offset + offsetof(XMMReg, XMM_Q(1)));
2206 tcg_gen_st_i64(s->tmp1_i64, tcg_env, decode->op[0].offset + offsetof(XMMReg, XMM_Q(1)));
2208 if (decode->op[0].offset != decode->op[1].offset) {
2209 tcg_gen_ld_i64(s->tmp1_i64, tcg_env, decode->op[1].offset + offsetof(XMMReg, XMM_Q(0)));
2210 tcg_gen_st_i64(s->tmp1_i64, tcg_env, decode->op[0].offset + offsetof(XMMReg, XMM_Q(0)));
2214 static void gen_VMOVHLPS(DisasContext *s, CPUX86State *env, X86DecodedInsn *decode)
2216 tcg_gen_ld_i64(s->tmp1_i64, tcg_env, decode->op[2].offset + offsetof(XMMReg, XMM_Q(1)));
2217 tcg_gen_st_i64(s->tmp1_i64, tcg_env, decode->op[0].offset + offsetof(XMMReg, XMM_Q(0)));
2218 if (decode->op[0].offset != decode->op[1].offset) {
2219 tcg_gen_ld_i64(s->tmp1_i64, tcg_env, decode->op[1].offset + offsetof(XMMReg, XMM_Q(1)));
2220 tcg_gen_st_i64(s->tmp1_i64, tcg_env, decode->op[0].offset + offsetof(XMMReg, XMM_Q(1)));
2224 static void gen_VMOVLHPS(DisasContext *s, CPUX86State *env, X86DecodedInsn *decode)
2226 tcg_gen_ld_i64(s->tmp1_i64, tcg_env, decode->op[2].offset);
2227 tcg_gen_st_i64(s->tmp1_i64, tcg_env, decode->op[0].offset + offsetof(XMMReg, XMM_Q(1)));
2228 if (decode->op[0].offset != decode->op[1].offset) {
2229 tcg_gen_ld_i64(s->tmp1_i64, tcg_env, decode->op[1].offset + offsetof(XMMReg, XMM_Q(0)));
2230 tcg_gen_st_i64(s->tmp1_i64, tcg_env, decode->op[0].offset + offsetof(XMMReg, XMM_Q(0)));
2235 * Note that MOVLPx supports 256-bit operation unlike MOVHLPx, MOVLHPx, MOXHPx.
2236 * Use a gvec move to move everything above the bottom 64 bits.
2239 static void gen_VMOVLPx(DisasContext *s, CPUX86State *env, X86DecodedInsn *decode)
2241 int vec_len = vector_len(s, decode);
2243 tcg_gen_ld_i64(s->tmp1_i64, tcg_env, decode->op[2].offset + offsetof(XMMReg, XMM_Q(0)));
2244 tcg_gen_gvec_mov(MO_64, decode->op[0].offset, decode->op[1].offset, vec_len, vec_len);
2245 tcg_gen_st_i64(s->tmp1_i64, tcg_env, decode->op[0].offset + offsetof(XMMReg, XMM_Q(0)));
2248 static void gen_VMOVLPx_ld(DisasContext *s, CPUX86State *env, X86DecodedInsn *decode)
2250 int vec_len = vector_len(s, decode);
2252 tcg_gen_qemu_ld_i64(s->tmp1_i64, s->A0, s->mem_index, MO_LEUQ);
2253 tcg_gen_gvec_mov(MO_64, decode->op[0].offset, decode->op[1].offset, vec_len, vec_len);
2254 tcg_gen_st_i64(s->tmp1_i64, OP_PTR0, offsetof(ZMMReg, ZMM_Q(0)));
2257 static void gen_VMOVLPx_st(DisasContext *s, CPUX86State *env, X86DecodedInsn *decode)
2259 tcg_gen_ld_i64(s->tmp1_i64, OP_PTR2, offsetof(ZMMReg, ZMM_Q(0)));
2260 tcg_gen_qemu_st_i64(s->tmp1_i64, s->A0, s->mem_index, MO_LEUQ);
2263 static void gen_VMOVSD_ld(DisasContext *s, CPUX86State *env, X86DecodedInsn *decode)
2265 TCGv_i64 zero = tcg_constant_i64(0);
2267 tcg_gen_qemu_ld_i64(s->tmp1_i64, s->A0, s->mem_index, MO_LEUQ);
2268 tcg_gen_st_i64(zero, OP_PTR0, offsetof(ZMMReg, ZMM_Q(1)));
2269 tcg_gen_st_i64(s->tmp1_i64, OP_PTR0, offsetof(ZMMReg, ZMM_Q(0)));
2272 static void gen_VMOVSS(DisasContext *s, CPUX86State *env, X86DecodedInsn *decode)
2274 int vec_len = vector_len(s, decode);
2276 tcg_gen_ld_i32(s->tmp2_i32, OP_PTR2, offsetof(ZMMReg, ZMM_L(0)));
2277 tcg_gen_gvec_mov(MO_64, decode->op[0].offset, decode->op[1].offset, vec_len, vec_len);
2278 tcg_gen_st_i32(s->tmp2_i32, OP_PTR0, offsetof(ZMMReg, ZMM_L(0)));
2281 static void gen_VMOVSS_ld(DisasContext *s, CPUX86State *env, X86DecodedInsn *decode)
2283 int vec_len = vector_len(s, decode);
2285 tcg_gen_qemu_ld_i32(s->tmp2_i32, s->A0, s->mem_index, MO_LEUL);
2286 tcg_gen_gvec_dup_imm(MO_64, decode->op[0].offset, vec_len, vec_len, 0);
2287 tcg_gen_st_i32(s->tmp2_i32, OP_PTR0, offsetof(ZMMReg, ZMM_L(0)));
2290 static void gen_VMOVSS_st(DisasContext *s, CPUX86State *env, X86DecodedInsn *decode)
2292 tcg_gen_ld_i32(s->tmp2_i32, OP_PTR2, offsetof(ZMMReg, ZMM_L(0)));
2293 tcg_gen_qemu_st_i32(s->tmp2_i32, s->A0, s->mem_index, MO_LEUL);
2296 static void gen_VPMASKMOV_st(DisasContext *s, CPUX86State *env, X86DecodedInsn *decode)
2299 gen_VMASKMOVPD_st(s, env, decode);
2301 gen_VMASKMOVPS_st(s, env, decode);
2305 static void gen_VPERMD(DisasContext *s, CPUX86State *env, X86DecodedInsn *decode)
2308 gen_helper_vpermd_ymm(OP_PTR0, OP_PTR1, OP_PTR2);
2311 static void gen_VPERM2x128(DisasContext *s, CPUX86State *env, X86DecodedInsn *decode)
2313 TCGv_i32 imm = tcg_constant8u_i32(decode->immediate);
2315 gen_helper_vpermdq_ymm(OP_PTR0, OP_PTR1, OP_PTR2, imm);
2318 static void gen_VPHMINPOSUW(DisasContext *s, CPUX86State *env, X86DecodedInsn *decode)
2321 gen_helper_phminposuw_xmm(tcg_env, OP_PTR0, OP_PTR2);
2324 static void gen_VROUNDSD(DisasContext *s, CPUX86State *env, X86DecodedInsn *decode)
2326 TCGv_i32 imm = tcg_constant8u_i32(decode->immediate);
2328 gen_helper_roundsd_xmm(tcg_env, OP_PTR0, OP_PTR1, OP_PTR2, imm);
2331 static void gen_VROUNDSS(DisasContext *s, CPUX86State *env, X86DecodedInsn *decode)
2333 TCGv_i32 imm = tcg_constant8u_i32(decode->immediate);
2335 gen_helper_roundss_xmm(tcg_env, OP_PTR0, OP_PTR1, OP_PTR2, imm);
2338 static void gen_VSHUF(DisasContext *s, CPUX86State *env, X86DecodedInsn *decode)
2340 TCGv_i32 imm = tcg_constant_i32(decode->immediate);
2341 SSEFunc_0_pppi ps, pd, fn;
2342 ps = s->vex_l ? gen_helper_shufps_ymm : gen_helper_shufps_xmm;
2343 pd = s->vex_l ? gen_helper_shufpd_ymm : gen_helper_shufpd_xmm;
2344 fn = s->prefix & PREFIX_DATA ? pd : ps;
2345 fn(OP_PTR0, OP_PTR1, OP_PTR2, imm);
2348 static void gen_VUCOMI(DisasContext *s, CPUX86State *env, X86DecodedInsn *decode)
2351 fn = s->prefix & PREFIX_DATA ? gen_helper_ucomisd : gen_helper_ucomiss;
2352 fn(tcg_env, OP_PTR1, OP_PTR2);
2353 set_cc_op(s, CC_OP_EFLAGS);
2356 static void gen_VZEROALL(DisasContext *s, CPUX86State *env, X86DecodedInsn *decode)
2358 TCGv_ptr ptr = tcg_temp_new_ptr();
2360 tcg_gen_addi_ptr(ptr, tcg_env, offsetof(CPUX86State, xmm_regs));
2361 gen_helper_memset(ptr, ptr, tcg_constant_i32(0),
2362 tcg_constant_ptr(CPU_NB_REGS * sizeof(ZMMReg)));
2365 static void gen_VZEROUPPER(DisasContext *s, CPUX86State *env, X86DecodedInsn *decode)
2369 for (i = 0; i < CPU_NB_REGS; i++) {
2370 int offset = offsetof(CPUX86State, xmm_regs[i].ZMM_X(1));
2371 tcg_gen_gvec_dup_imm(MO_64, offset, 16, 16, 0);