]>
Commit | Line | Data |
---|---|---|
b3e22b23 PB |
1 | /* |
2 | * New-style decoder for i386 instructions | |
3 | * | |
4 | * Copyright (c) 2022 Red Hat, Inc. | |
5 | * | |
6 | * Author: Paolo Bonzini <pbonzini@redhat.com> | |
7 | * | |
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. | |
12 | * | |
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. | |
17 | * | |
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/>. | |
20 | */ | |
21 | ||
22 | /* | |
23 | * The decoder is mostly based on tables copied from the Intel SDM. As | |
24 | * a result, most operand load and writeback is done entirely in common | |
25 | * table-driven code using the same operand type (X86_TYPE_*) and | |
26 | * size (X86_SIZE_*) codes used in the manual. | |
27 | * | |
28 | * The main difference is that the V, U and W types are extended to | |
29 | * cover MMX as well; if an instruction is like | |
30 | * | |
31 | * por Pq, Qq | |
32 | * 66 por Vx, Hx, Wx | |
33 | * | |
34 | * only the second row is included and the instruction is marked as a | |
35 | * valid MMX instruction. The MMX flag directs the decoder to rewrite | |
36 | * the V/U/H/W types to P/N/P/Q if there is no prefix, as well as changing | |
37 | * "x" to "q" if there is no prefix. | |
38 | * | |
39 | * In addition, the ss/ps/sd/pd types are sometimes mushed together as "x" | |
40 | * if the difference is expressed via prefixes. Individual instructions | |
41 | * are separated by prefix in the generator functions. | |
42 | * | |
43 | * There are a couple cases in which instructions (e.g. MOVD) write the | |
44 | * whole XMM or MM register but are established incorrectly in the manual | |
45 | * as "d" or "q". These have to be fixed for the decoder to work correctly. | |
46 | */ | |
47 | ||
48 | #define X86_OP_NONE { 0 }, | |
49 | ||
50 | #define X86_OP_GROUP3(op, op0_, s0_, op1_, s1_, op2_, s2_, ...) { \ | |
51 | .decode = glue(decode_, op), \ | |
52 | .op0 = glue(X86_TYPE_, op0_), \ | |
53 | .s0 = glue(X86_SIZE_, s0_), \ | |
54 | .op1 = glue(X86_TYPE_, op1_), \ | |
55 | .s1 = glue(X86_SIZE_, s1_), \ | |
56 | .op2 = glue(X86_TYPE_, op2_), \ | |
57 | .s2 = glue(X86_SIZE_, s2_), \ | |
58 | .is_decode = true, \ | |
59 | ## __VA_ARGS__ \ | |
60 | } | |
61 | ||
62 | #define X86_OP_GROUP2(op, op0, s0, op1, s1, ...) \ | |
63 | X86_OP_GROUP3(op, op0, s0, 2op, s0, op1, s1, ## __VA_ARGS__) | |
64 | #define X86_OP_GROUP0(op, ...) \ | |
65 | X86_OP_GROUP3(op, None, None, None, None, None, None, ## __VA_ARGS__) | |
66 | ||
67 | #define X86_OP_ENTRY3(op, op0_, s0_, op1_, s1_, op2_, s2_, ...) { \ | |
68 | .gen = glue(gen_, op), \ | |
69 | .op0 = glue(X86_TYPE_, op0_), \ | |
70 | .s0 = glue(X86_SIZE_, s0_), \ | |
71 | .op1 = glue(X86_TYPE_, op1_), \ | |
72 | .s1 = glue(X86_SIZE_, s1_), \ | |
73 | .op2 = glue(X86_TYPE_, op2_), \ | |
74 | .s2 = glue(X86_SIZE_, s2_), \ | |
75 | ## __VA_ARGS__ \ | |
76 | } | |
77 | ||
78 | #define X86_OP_ENTRY4(op, op0_, s0_, op1_, s1_, op2_, s2_, ...) \ | |
79 | X86_OP_ENTRY3(op, op0_, s0_, op1_, s1_, op2_, s2_, \ | |
80 | .op3 = X86_TYPE_I, .s3 = X86_SIZE_b, \ | |
81 | ## __VA_ARGS__) | |
82 | ||
83 | #define X86_OP_ENTRY2(op, op0, s0, op1, s1, ...) \ | |
84 | X86_OP_ENTRY3(op, op0, s0, 2op, s0, op1, s1, ## __VA_ARGS__) | |
57f6bba0 PB |
85 | #define X86_OP_ENTRYw(op, op0, s0, ...) \ |
86 | X86_OP_ENTRY3(op, op0, s0, None, None, None, None, ## __VA_ARGS__) | |
87 | #define X86_OP_ENTRYr(op, op0, s0, ...) \ | |
88 | X86_OP_ENTRY3(op, None, None, None, None, op0, s0, ## __VA_ARGS__) | |
b3e22b23 PB |
89 | #define X86_OP_ENTRY0(op, ...) \ |
90 | X86_OP_ENTRY3(op, None, None, None, None, None, None, ## __VA_ARGS__) | |
91 | ||
caa01fad | 92 | #define cpuid(feat) .cpuid = X86_FEAT_##feat, |
b3e22b23 PB |
93 | #define i64 .special = X86_SPECIAL_i64, |
94 | #define o64 .special = X86_SPECIAL_o64, | |
95 | #define xchg .special = X86_SPECIAL_Locked, | |
96 | #define mmx .special = X86_SPECIAL_MMX, | |
97 | #define zext0 .special = X86_SPECIAL_ZExtOp0, | |
98 | #define zext2 .special = X86_SPECIAL_ZExtOp2, | |
16fc5726 | 99 | #define avx_movx .special = X86_SPECIAL_AVXExtMov, |
b3e22b23 | 100 | |
20581aad PB |
101 | #define vex1 .vex_class = 1, |
102 | #define vex1_rep3 .vex_class = 1, .vex_special = X86_VEX_REPScalar, | |
103 | #define vex2 .vex_class = 2, | |
104 | #define vex2_rep3 .vex_class = 2, .vex_special = X86_VEX_REPScalar, | |
105 | #define vex3 .vex_class = 3, | |
106 | #define vex4 .vex_class = 4, | |
107 | #define vex4_unal .vex_class = 4, .vex_special = X86_VEX_SSEUnaligned, | |
3d304620 | 108 | #define vex4_rep5 .vex_class = 4, .vex_special = X86_VEX_REPScalar, |
20581aad PB |
109 | #define vex5 .vex_class = 5, |
110 | #define vex6 .vex_class = 6, | |
111 | #define vex7 .vex_class = 7, | |
112 | #define vex8 .vex_class = 8, | |
113 | #define vex11 .vex_class = 11, | |
114 | #define vex12 .vex_class = 12, | |
115 | #define vex13 .vex_class = 13, | |
116 | ||
117 | #define avx2_256 .vex_special = X86_VEX_AVX2_256, | |
118 | ||
55a33286 PB |
119 | #define P_00 1 |
120 | #define P_66 (1 << PREFIX_DATA) | |
121 | #define P_F3 (1 << PREFIX_REPZ) | |
122 | #define P_F2 (1 << PREFIX_REPNZ) | |
123 | ||
124 | #define p_00 .valid_prefix = P_00, | |
125 | #define p_66 .valid_prefix = P_66, | |
126 | #define p_f3 .valid_prefix = P_F3, | |
127 | #define p_f2 .valid_prefix = P_F2, | |
128 | #define p_00_66 .valid_prefix = P_00 | P_66, | |
129 | #define p_00_f3 .valid_prefix = P_00 | P_F3, | |
130 | #define p_66_f2 .valid_prefix = P_66 | P_F2, | |
131 | #define p_00_66_f3 .valid_prefix = P_00 | P_66 | P_F3, | |
132 | #define p_66_f3_f2 .valid_prefix = P_66 | P_F3 | P_F2, | |
133 | #define p_00_66_f3_f2 .valid_prefix = P_00 | P_66 | P_F3 | P_F2, | |
134 | ||
b3e22b23 PB |
135 | static uint8_t get_modrm(DisasContext *s, CPUX86State *env) |
136 | { | |
137 | if (!s->has_modrm) { | |
138 | s->modrm = x86_ldub_code(env, s); | |
139 | s->has_modrm = true; | |
140 | } | |
141 | return s->modrm; | |
142 | } | |
143 | ||
92ec056a PB |
144 | static inline const X86OpEntry *decode_by_prefix(DisasContext *s, const X86OpEntry entries[4]) |
145 | { | |
146 | if (s->prefix & PREFIX_REPNZ) { | |
147 | return &entries[3]; | |
148 | } else if (s->prefix & PREFIX_REPZ) { | |
149 | return &entries[2]; | |
150 | } else if (s->prefix & PREFIX_DATA) { | |
151 | return &entries[1]; | |
152 | } else { | |
153 | return &entries[0]; | |
154 | } | |
155 | } | |
156 | ||
57f6bba0 PB |
157 | static void decode_group15(DisasContext *s, CPUX86State *env, X86OpEntry *entry, uint8_t *b) |
158 | { | |
159 | /* only includes ldmxcsr and stmxcsr, because they have AVX variants. */ | |
160 | static const X86OpEntry group15_reg[8] = { | |
161 | }; | |
162 | ||
163 | static const X86OpEntry group15_mem[8] = { | |
164 | [2] = X86_OP_ENTRYr(LDMXCSR, E,d, vex5), | |
165 | [3] = X86_OP_ENTRYw(STMXCSR, E,d, vex5), | |
166 | }; | |
167 | ||
168 | uint8_t modrm = get_modrm(s, env); | |
169 | if ((modrm >> 6) == 3) { | |
170 | *entry = group15_reg[(modrm >> 3) & 7]; | |
171 | } else { | |
172 | *entry = group15_mem[(modrm >> 3) & 7]; | |
173 | } | |
174 | } | |
175 | ||
1d0b9261 PB |
176 | static void decode_group17(DisasContext *s, CPUX86State *env, X86OpEntry *entry, uint8_t *b) |
177 | { | |
178 | static const X86GenFunc group17_gen[8] = { | |
179 | NULL, gen_BLSR, gen_BLSMSK, gen_BLSI, | |
180 | }; | |
181 | int op = (get_modrm(s, env) >> 3) & 7; | |
182 | entry->gen = group17_gen[op]; | |
183 | } | |
184 | ||
ce4fcb94 PB |
185 | static void decode_group12(DisasContext *s, CPUX86State *env, X86OpEntry *entry, uint8_t *b) |
186 | { | |
187 | static const X86OpEntry opcodes_group12[8] = { | |
188 | {}, | |
189 | {}, | |
190 | X86_OP_ENTRY3(PSRLW_i, H,x, U,x, I,b, vex7 mmx avx2_256 p_00_66), | |
191 | {}, | |
192 | X86_OP_ENTRY3(PSRAW_i, H,x, U,x, I,b, vex7 mmx avx2_256 p_00_66), | |
193 | {}, | |
194 | X86_OP_ENTRY3(PSLLW_i, H,x, U,x, I,b, vex7 mmx avx2_256 p_00_66), | |
195 | {}, | |
196 | }; | |
197 | ||
198 | int op = (get_modrm(s, env) >> 3) & 7; | |
199 | *entry = opcodes_group12[op]; | |
200 | } | |
201 | ||
202 | static void decode_group13(DisasContext *s, CPUX86State *env, X86OpEntry *entry, uint8_t *b) | |
203 | { | |
204 | static const X86OpEntry opcodes_group13[8] = { | |
205 | {}, | |
206 | {}, | |
207 | X86_OP_ENTRY3(PSRLD_i, H,x, U,x, I,b, vex7 mmx avx2_256 p_00_66), | |
208 | {}, | |
209 | X86_OP_ENTRY3(PSRAD_i, H,x, U,x, I,b, vex7 mmx avx2_256 p_00_66), | |
210 | {}, | |
211 | X86_OP_ENTRY3(PSLLD_i, H,x, U,x, I,b, vex7 mmx avx2_256 p_00_66), | |
212 | {}, | |
213 | }; | |
214 | ||
215 | int op = (get_modrm(s, env) >> 3) & 7; | |
216 | *entry = opcodes_group13[op]; | |
217 | } | |
218 | ||
219 | static void decode_group14(DisasContext *s, CPUX86State *env, X86OpEntry *entry, uint8_t *b) | |
220 | { | |
221 | static const X86OpEntry opcodes_group14[8] = { | |
222 | /* grp14 */ | |
223 | {}, | |
224 | {}, | |
225 | X86_OP_ENTRY3(PSRLQ_i, H,x, U,x, I,b, vex7 mmx avx2_256 p_00_66), | |
226 | X86_OP_ENTRY3(PSRLDQ_i, H,x, U,x, I,b, vex7 avx2_256 p_66), | |
227 | {}, | |
228 | {}, | |
229 | X86_OP_ENTRY3(PSLLQ_i, H,x, U,x, I,b, vex7 mmx avx2_256 p_00_66), | |
230 | X86_OP_ENTRY3(PSLLDQ_i, H,x, U,x, I,b, vex7 avx2_256 p_66), | |
231 | }; | |
232 | ||
233 | int op = (get_modrm(s, env) >> 3) & 7; | |
234 | *entry = opcodes_group14[op]; | |
235 | } | |
236 | ||
92ec056a PB |
237 | static void decode_0F6F(DisasContext *s, CPUX86State *env, X86OpEntry *entry, uint8_t *b) |
238 | { | |
239 | static const X86OpEntry opcodes_0F6F[4] = { | |
cab529b0 | 240 | X86_OP_ENTRY3(MOVDQ, P,q, None,None, Q,q, vex5 mmx), /* movq */ |
92ec056a PB |
241 | X86_OP_ENTRY3(MOVDQ, V,x, None,None, W,x, vex1), /* movdqa */ |
242 | X86_OP_ENTRY3(MOVDQ, V,x, None,None, W,x, vex4_unal), /* movdqu */ | |
243 | {}, | |
244 | }; | |
245 | *entry = *decode_by_prefix(s, opcodes_0F6F); | |
246 | } | |
247 | ||
ce4fcb94 PB |
248 | static void decode_0F70(DisasContext *s, CPUX86State *env, X86OpEntry *entry, uint8_t *b) |
249 | { | |
250 | static const X86OpEntry pshufw[4] = { | |
251 | X86_OP_ENTRY3(PSHUFW, P,q, Q,q, I,b, vex4 mmx), | |
252 | X86_OP_ENTRY3(PSHUFD, V,x, W,x, I,b, vex4 avx2_256), | |
253 | X86_OP_ENTRY3(PSHUFHW, V,x, W,x, I,b, vex4 avx2_256), | |
254 | X86_OP_ENTRY3(PSHUFLW, V,x, W,x, I,b, vex4 avx2_256), | |
255 | }; | |
256 | ||
257 | *entry = *decode_by_prefix(s, pshufw); | |
258 | } | |
259 | ||
260 | static void decode_0F77(DisasContext *s, CPUX86State *env, X86OpEntry *entry, uint8_t *b) | |
261 | { | |
262 | if (!(s->prefix & PREFIX_VEX)) { | |
263 | entry->gen = gen_EMMS; | |
264 | } else if (!s->vex_l) { | |
265 | entry->gen = gen_VZEROUPPER; | |
266 | entry->vex_class = 8; | |
267 | } else { | |
268 | entry->gen = gen_VZEROALL; | |
269 | entry->vex_class = 8; | |
270 | } | |
271 | } | |
272 | ||
d1c1a422 PB |
273 | static void decode_0F78(DisasContext *s, CPUX86State *env, X86OpEntry *entry, uint8_t *b) |
274 | { | |
275 | static const X86OpEntry opcodes_0F78[4] = { | |
276 | {}, | |
afa94dab | 277 | X86_OP_ENTRY3(EXTRQ_i, V,x, None,None, I,w, cpuid(SSE4A)), /* AMD extension */ |
d1c1a422 | 278 | {}, |
afa94dab | 279 | X86_OP_ENTRY3(INSERTQ_i, V,x, U,x, I,w, cpuid(SSE4A)), /* AMD extension */ |
d1c1a422 PB |
280 | }; |
281 | *entry = *decode_by_prefix(s, opcodes_0F78); | |
282 | } | |
283 | ||
284 | static void decode_0F79(DisasContext *s, CPUX86State *env, X86OpEntry *entry, uint8_t *b) | |
285 | { | |
286 | if (s->prefix & PREFIX_REPNZ) { | |
afa94dab | 287 | entry->gen = gen_INSERTQ_r; /* AMD extension */ |
d1c1a422 | 288 | } else if (s->prefix & PREFIX_DATA) { |
afa94dab | 289 | entry->gen = gen_EXTRQ_r; /* AMD extension */ |
d1c1a422 PB |
290 | } else { |
291 | entry->gen = NULL; | |
292 | }; | |
293 | } | |
294 | ||
295 | static void decode_0F7E(DisasContext *s, CPUX86State *env, X86OpEntry *entry, uint8_t *b) | |
296 | { | |
297 | static const X86OpEntry opcodes_0F7E[4] = { | |
298 | X86_OP_ENTRY3(MOVD_from, E,y, None,None, P,y, vex5 mmx), | |
299 | X86_OP_ENTRY3(MOVD_from, E,y, None,None, V,y, vex5), | |
300 | X86_OP_ENTRY3(MOVQ, V,x, None,None, W,q, vex5), /* wrong dest Vy on SDM! */ | |
301 | {}, | |
302 | }; | |
303 | *entry = *decode_by_prefix(s, opcodes_0F7E); | |
304 | } | |
305 | ||
306 | static void decode_0F7F(DisasContext *s, CPUX86State *env, X86OpEntry *entry, uint8_t *b) | |
307 | { | |
308 | static const X86OpEntry opcodes_0F7F[4] = { | |
cab529b0 | 309 | X86_OP_ENTRY3(MOVDQ, W,x, None,None, V,x, vex5 mmx), /* movq */ |
d1c1a422 PB |
310 | X86_OP_ENTRY3(MOVDQ, W,x, None,None, V,x, vex1), /* movdqa */ |
311 | X86_OP_ENTRY3(MOVDQ, W,x, None,None, V,x, vex4_unal), /* movdqu */ | |
312 | {}, | |
313 | }; | |
314 | *entry = *decode_by_prefix(s, opcodes_0F7F); | |
315 | } | |
316 | ||
6bbeb98d PB |
317 | static void decode_0FD6(DisasContext *s, CPUX86State *env, X86OpEntry *entry, uint8_t *b) |
318 | { | |
319 | static const X86OpEntry movq[4] = { | |
320 | {}, | |
321 | X86_OP_ENTRY3(MOVQ, W,x, None, None, V,q, vex5), | |
322 | X86_OP_ENTRY3(MOVq_dq, V,dq, None, None, N,q), | |
323 | X86_OP_ENTRY3(MOVq_dq, P,q, None, None, U,q), | |
324 | }; | |
325 | ||
326 | *entry = *decode_by_prefix(s, movq); | |
327 | } | |
328 | ||
b3e22b23 | 329 | static const X86OpEntry opcodes_0F38_00toEF[240] = { |
16fc5726 PB |
330 | [0x00] = X86_OP_ENTRY3(PSHUFB, V,x, H,x, W,x, vex4 cpuid(SSSE3) mmx avx2_256 p_00_66), |
331 | [0x01] = X86_OP_ENTRY3(PHADDW, V,x, H,x, W,x, vex4 cpuid(SSSE3) mmx avx2_256 p_00_66), | |
332 | [0x02] = X86_OP_ENTRY3(PHADDD, V,x, H,x, W,x, vex4 cpuid(SSSE3) mmx avx2_256 p_00_66), | |
333 | [0x03] = X86_OP_ENTRY3(PHADDSW, V,x, H,x, W,x, vex4 cpuid(SSSE3) mmx avx2_256 p_00_66), | |
334 | [0x04] = X86_OP_ENTRY3(PMADDUBSW, V,x, H,x, W,x, vex4 cpuid(SSSE3) mmx avx2_256 p_00_66), | |
335 | [0x05] = X86_OP_ENTRY3(PHSUBW, V,x, H,x, W,x, vex4 cpuid(SSSE3) mmx avx2_256 p_00_66), | |
336 | [0x06] = X86_OP_ENTRY3(PHSUBD, V,x, H,x, W,x, vex4 cpuid(SSSE3) mmx avx2_256 p_00_66), | |
337 | [0x07] = X86_OP_ENTRY3(PHSUBSW, V,x, H,x, W,x, vex4 cpuid(SSSE3) mmx avx2_256 p_00_66), | |
338 | ||
339 | [0x10] = X86_OP_ENTRY2(PBLENDVB, V,x, W,x, vex4 cpuid(SSE41) avx2_256 p_66), | |
a48b2697 | 340 | [0x13] = X86_OP_ENTRY2(VCVTPH2PS, V,x, W,xh, vex11 cpuid(F16C) p_66), |
16fc5726 PB |
341 | [0x14] = X86_OP_ENTRY2(BLENDVPS, V,x, W,x, vex4 cpuid(SSE41) p_66), |
342 | [0x15] = X86_OP_ENTRY2(BLENDVPD, V,x, W,x, vex4 cpuid(SSE41) p_66), | |
343 | /* Listed incorrectly as type 4 */ | |
344 | [0x16] = X86_OP_ENTRY3(VPERMD, V,qq, H,qq, W,qq, vex6 cpuid(AVX2) p_66), | |
345 | [0x17] = X86_OP_ENTRY3(VPTEST, None,None, V,x, W,x, vex4 cpuid(SSE41) p_66), | |
346 | ||
347 | /* | |
348 | * Source operand listed as Mq/Ux and similar in the manual; incorrectly listed | |
349 | * as 128-bit only in 2-17. | |
350 | */ | |
351 | [0x20] = X86_OP_ENTRY3(VPMOVSXBW, V,x, None,None, W,q, vex5 cpuid(SSE41) avx_movx avx2_256 p_66), | |
352 | [0x21] = X86_OP_ENTRY3(VPMOVSXBD, V,x, None,None, W,d, vex5 cpuid(SSE41) avx_movx avx2_256 p_66), | |
353 | [0x22] = X86_OP_ENTRY3(VPMOVSXBQ, V,x, None,None, W,w, vex5 cpuid(SSE41) avx_movx avx2_256 p_66), | |
354 | [0x23] = X86_OP_ENTRY3(VPMOVSXWD, V,x, None,None, W,q, vex5 cpuid(SSE41) avx_movx avx2_256 p_66), | |
355 | [0x24] = X86_OP_ENTRY3(VPMOVSXWQ, V,x, None,None, W,d, vex5 cpuid(SSE41) avx_movx avx2_256 p_66), | |
356 | [0x25] = X86_OP_ENTRY3(VPMOVSXDQ, V,x, None,None, W,q, vex5 cpuid(SSE41) avx_movx avx2_256 p_66), | |
357 | ||
358 | /* Same as PMOVSX. */ | |
359 | [0x30] = X86_OP_ENTRY3(VPMOVZXBW, V,x, None,None, W,q, vex5 cpuid(SSE41) avx_movx avx2_256 p_66), | |
360 | [0x31] = X86_OP_ENTRY3(VPMOVZXBD, V,x, None,None, W,d, vex5 cpuid(SSE41) avx_movx avx2_256 p_66), | |
361 | [0x32] = X86_OP_ENTRY3(VPMOVZXBQ, V,x, None,None, W,w, vex5 cpuid(SSE41) avx_movx avx2_256 p_66), | |
362 | [0x33] = X86_OP_ENTRY3(VPMOVZXWD, V,x, None,None, W,q, vex5 cpuid(SSE41) avx_movx avx2_256 p_66), | |
363 | [0x34] = X86_OP_ENTRY3(VPMOVZXWQ, V,x, None,None, W,d, vex5 cpuid(SSE41) avx_movx avx2_256 p_66), | |
364 | [0x35] = X86_OP_ENTRY3(VPMOVZXDQ, V,x, None,None, W,q, vex5 cpuid(SSE41) avx_movx avx2_256 p_66), | |
365 | [0x36] = X86_OP_ENTRY3(VPERMD, V,qq, H,qq, W,qq, vex6 cpuid(AVX2) p_66), | |
366 | [0x37] = X86_OP_ENTRY3(PCMPGTQ, V,x, H,x, W,x, vex4 cpuid(SSE42) avx2_256 p_66), | |
367 | ||
368 | [0x40] = X86_OP_ENTRY3(PMULLD, V,x, H,x, W,x, vex4 cpuid(SSE41) avx2_256 p_66), | |
369 | [0x41] = X86_OP_ENTRY3(VPHMINPOSUW, V,dq, None,None, W,dq, vex4 cpuid(SSE41) p_66), | |
370 | /* Listed incorrectly as type 4 */ | |
371 | [0x45] = X86_OP_ENTRY3(VPSRLV, V,x, H,x, W,x, vex6 cpuid(AVX2) p_66), | |
372 | [0x46] = X86_OP_ENTRY3(VPSRAV, V,x, H,x, W,x, vex6 cpuid(AVX2) p_66), | |
373 | [0x47] = X86_OP_ENTRY3(VPSLLV, V,x, H,x, W,x, vex6 cpuid(AVX2) p_66), | |
374 | ||
375 | [0x90] = X86_OP_ENTRY3(VPGATHERD, V,x, H,x, M,d, vex12 cpuid(AVX2) p_66), /* vpgatherdd/q */ | |
376 | [0x91] = X86_OP_ENTRY3(VPGATHERQ, V,x, H,x, M,q, vex12 cpuid(AVX2) p_66), /* vpgatherqd/q */ | |
377 | [0x92] = X86_OP_ENTRY3(VPGATHERD, V,x, H,x, M,d, vex12 cpuid(AVX2) p_66), /* vgatherdps/d */ | |
378 | [0x93] = X86_OP_ENTRY3(VPGATHERQ, V,x, H,x, M,q, vex12 cpuid(AVX2) p_66), /* vgatherqps/d */ | |
379 | ||
2872b0f3 PB |
380 | /* Should be exception type 2 but they do not have legacy SSE equivalents? */ |
381 | [0x96] = X86_OP_ENTRY3(VFMADDSUB132Px, V,x, H,x, W,x, vex6 cpuid(FMA) p_66), | |
382 | [0x97] = X86_OP_ENTRY3(VFMSUBADD132Px, V,x, H,x, W,x, vex6 cpuid(FMA) p_66), | |
383 | ||
384 | [0xa6] = X86_OP_ENTRY3(VFMADDSUB213Px, V,x, H,x, W,x, vex6 cpuid(FMA) p_66), | |
385 | [0xa7] = X86_OP_ENTRY3(VFMSUBADD213Px, V,x, H,x, W,x, vex6 cpuid(FMA) p_66), | |
386 | ||
387 | [0xb6] = X86_OP_ENTRY3(VFMADDSUB231Px, V,x, H,x, W,x, vex6 cpuid(FMA) p_66), | |
388 | [0xb7] = X86_OP_ENTRY3(VFMSUBADD231Px, V,x, H,x, W,x, vex6 cpuid(FMA) p_66), | |
389 | ||
16fc5726 PB |
390 | [0x08] = X86_OP_ENTRY3(PSIGNB, V,x, H,x, W,x, vex4 cpuid(SSSE3) mmx avx2_256 p_00_66), |
391 | [0x09] = X86_OP_ENTRY3(PSIGNW, V,x, H,x, W,x, vex4 cpuid(SSSE3) mmx avx2_256 p_00_66), | |
392 | [0x0a] = X86_OP_ENTRY3(PSIGND, V,x, H,x, W,x, vex4 cpuid(SSSE3) mmx avx2_256 p_00_66), | |
393 | [0x0b] = X86_OP_ENTRY3(PMULHRSW, V,x, H,x, W,x, vex4 cpuid(SSSE3) mmx avx2_256 p_00_66), | |
394 | [0x0c] = X86_OP_ENTRY3(VPERMILPS, V,x, H,x, W,x, vex4 cpuid(AVX) p_00_66), | |
395 | [0x0d] = X86_OP_ENTRY3(VPERMILPD, V,x, H,x, W,x, vex4 cpuid(AVX) p_66), | |
396 | [0x0e] = X86_OP_ENTRY3(VTESTPS, None,None, V,x, W,x, vex4 cpuid(AVX) p_66), | |
397 | [0x0f] = X86_OP_ENTRY3(VTESTPD, None,None, V,x, W,x, vex4 cpuid(AVX) p_66), | |
398 | ||
399 | [0x18] = X86_OP_ENTRY3(VPBROADCASTD, V,x, None,None, W,d, vex6 cpuid(AVX) p_66), /* vbroadcastss */ | |
400 | [0x19] = X86_OP_ENTRY3(VPBROADCASTQ, V,qq, None,None, W,q, vex6 cpuid(AVX) p_66), /* vbroadcastsd */ | |
401 | [0x1a] = X86_OP_ENTRY3(VBROADCASTx128, V,qq, None,None, WM,dq,vex6 cpuid(AVX) p_66), | |
402 | [0x1c] = X86_OP_ENTRY3(PABSB, V,x, None,None, W,x, vex4 cpuid(SSSE3) mmx avx2_256 p_00_66), | |
403 | [0x1d] = X86_OP_ENTRY3(PABSW, V,x, None,None, W,x, vex4 cpuid(SSSE3) mmx avx2_256 p_00_66), | |
404 | [0x1e] = X86_OP_ENTRY3(PABSD, V,x, None,None, W,x, vex4 cpuid(SSSE3) mmx avx2_256 p_00_66), | |
405 | ||
406 | [0x28] = X86_OP_ENTRY3(PMULDQ, V,x, H,x, W,x, vex4 cpuid(SSE41) avx2_256 p_66), | |
407 | [0x29] = X86_OP_ENTRY3(PCMPEQQ, V,x, H,x, W,x, vex4 cpuid(SSE41) avx2_256 p_66), | |
408 | [0x2a] = X86_OP_ENTRY3(MOVDQ, V,x, None,None, WM,x, vex1 cpuid(SSE41) avx2_256 p_66), /* movntdqa */ | |
409 | [0x2b] = X86_OP_ENTRY3(VPACKUSDW, V,x, H,x, W,x, vex4 cpuid(SSE41) avx2_256 p_66), | |
410 | [0x2c] = X86_OP_ENTRY3(VMASKMOVPS, V,x, H,x, WM,x, vex6 cpuid(AVX) p_66), | |
411 | [0x2d] = X86_OP_ENTRY3(VMASKMOVPD, V,x, H,x, WM,x, vex6 cpuid(AVX) p_66), | |
412 | /* Incorrectly listed as Mx,Hx,Vx in the manual */ | |
413 | [0x2e] = X86_OP_ENTRY3(VMASKMOVPS_st, M,x, V,x, H,x, vex6 cpuid(AVX) p_66), | |
414 | [0x2f] = X86_OP_ENTRY3(VMASKMOVPD_st, M,x, V,x, H,x, vex6 cpuid(AVX) p_66), | |
415 | ||
416 | [0x38] = X86_OP_ENTRY3(PMINSB, V,x, H,x, W,x, vex4 cpuid(SSE41) avx2_256 p_66), | |
417 | [0x39] = X86_OP_ENTRY3(PMINSD, V,x, H,x, W,x, vex4 cpuid(SSE41) avx2_256 p_66), | |
418 | [0x3a] = X86_OP_ENTRY3(PMINUW, V,x, H,x, W,x, vex4 cpuid(SSE41) avx2_256 p_66), | |
419 | [0x3b] = X86_OP_ENTRY3(PMINUD, V,x, H,x, W,x, vex4 cpuid(SSE41) avx2_256 p_66), | |
420 | [0x3c] = X86_OP_ENTRY3(PMAXSB, V,x, H,x, W,x, vex4 cpuid(SSE41) avx2_256 p_66), | |
421 | [0x3d] = X86_OP_ENTRY3(PMAXSD, V,x, H,x, W,x, vex4 cpuid(SSE41) avx2_256 p_66), | |
422 | [0x3e] = X86_OP_ENTRY3(PMAXUW, V,x, H,x, W,x, vex4 cpuid(SSE41) avx2_256 p_66), | |
423 | [0x3f] = X86_OP_ENTRY3(PMAXUD, V,x, H,x, W,x, vex4 cpuid(SSE41) avx2_256 p_66), | |
424 | ||
425 | [0x58] = X86_OP_ENTRY3(VPBROADCASTD, V,x, None,None, W,d, vex6 cpuid(AVX2) p_66), | |
426 | [0x59] = X86_OP_ENTRY3(VPBROADCASTQ, V,x, None,None, W,q, vex6 cpuid(AVX2) p_66), | |
427 | [0x5a] = X86_OP_ENTRY3(VBROADCASTx128, V,qq, None,None, WM,dq,vex6 cpuid(AVX2) p_66), | |
428 | ||
429 | [0x78] = X86_OP_ENTRY3(VPBROADCASTB, V,x, None,None, W,b, vex6 cpuid(AVX2) p_66), | |
430 | [0x79] = X86_OP_ENTRY3(VPBROADCASTW, V,x, None,None, W,w, vex6 cpuid(AVX2) p_66), | |
431 | ||
432 | [0x8c] = X86_OP_ENTRY3(VPMASKMOV, V,x, H,x, WM,x, vex6 cpuid(AVX2) p_66), | |
433 | [0x8e] = X86_OP_ENTRY3(VPMASKMOV_st, M,x, V,x, H,x, vex6 cpuid(AVX2) p_66), | |
434 | ||
2872b0f3 PB |
435 | /* Should be exception type 2 or 3 but they do not have legacy SSE equivalents? */ |
436 | [0x98] = X86_OP_ENTRY3(VFMADD132Px, V,x, H,x, W,x, vex6 cpuid(FMA) p_66), | |
437 | [0x99] = X86_OP_ENTRY3(VFMADD132Sx, V,x, H,x, W,x, vex6 cpuid(FMA) p_66), | |
438 | [0x9a] = X86_OP_ENTRY3(VFMSUB132Px, V,x, H,x, W,x, vex6 cpuid(FMA) p_66), | |
439 | [0x9b] = X86_OP_ENTRY3(VFMSUB132Sx, V,x, H,x, W,x, vex6 cpuid(FMA) p_66), | |
440 | [0x9c] = X86_OP_ENTRY3(VFNMADD132Px, V,x, H,x, W,x, vex6 cpuid(FMA) p_66), | |
441 | [0x9d] = X86_OP_ENTRY3(VFNMADD132Sx, V,x, H,x, W,x, vex6 cpuid(FMA) p_66), | |
442 | [0x9e] = X86_OP_ENTRY3(VFNMSUB132Px, V,x, H,x, W,x, vex6 cpuid(FMA) p_66), | |
443 | [0x9f] = X86_OP_ENTRY3(VFNMSUB132Sx, V,x, H,x, W,x, vex6 cpuid(FMA) p_66), | |
444 | ||
445 | [0xa8] = X86_OP_ENTRY3(VFMADD213Px, V,x, H,x, W,x, vex6 cpuid(FMA) p_66), | |
446 | [0xa9] = X86_OP_ENTRY3(VFMADD213Sx, V,x, H,x, W,x, vex6 cpuid(FMA) p_66), | |
447 | [0xaa] = X86_OP_ENTRY3(VFMSUB213Px, V,x, H,x, W,x, vex6 cpuid(FMA) p_66), | |
448 | [0xab] = X86_OP_ENTRY3(VFMSUB213Sx, V,x, H,x, W,x, vex6 cpuid(FMA) p_66), | |
449 | [0xac] = X86_OP_ENTRY3(VFNMADD213Px, V,x, H,x, W,x, vex6 cpuid(FMA) p_66), | |
450 | [0xad] = X86_OP_ENTRY3(VFNMADD213Sx, V,x, H,x, W,x, vex6 cpuid(FMA) p_66), | |
451 | [0xae] = X86_OP_ENTRY3(VFNMSUB213Px, V,x, H,x, W,x, vex6 cpuid(FMA) p_66), | |
452 | [0xaf] = X86_OP_ENTRY3(VFNMSUB213Sx, V,x, H,x, W,x, vex6 cpuid(FMA) p_66), | |
453 | ||
454 | [0xb8] = X86_OP_ENTRY3(VFMADD231Px, V,x, H,x, W,x, vex6 cpuid(FMA) p_66), | |
455 | [0xb9] = X86_OP_ENTRY3(VFMADD231Sx, V,x, H,x, W,x, vex6 cpuid(FMA) p_66), | |
456 | [0xba] = X86_OP_ENTRY3(VFMSUB231Px, V,x, H,x, W,x, vex6 cpuid(FMA) p_66), | |
457 | [0xbb] = X86_OP_ENTRY3(VFMSUB231Sx, V,x, H,x, W,x, vex6 cpuid(FMA) p_66), | |
458 | [0xbc] = X86_OP_ENTRY3(VFNMADD231Px, V,x, H,x, W,x, vex6 cpuid(FMA) p_66), | |
459 | [0xbd] = X86_OP_ENTRY3(VFNMADD231Sx, V,x, H,x, W,x, vex6 cpuid(FMA) p_66), | |
460 | [0xbe] = X86_OP_ENTRY3(VFNMSUB231Px, V,x, H,x, W,x, vex6 cpuid(FMA) p_66), | |
461 | [0xbf] = X86_OP_ENTRY3(VFNMSUB231Sx, V,x, H,x, W,x, vex6 cpuid(FMA) p_66), | |
462 | ||
16fc5726 PB |
463 | [0xdb] = X86_OP_ENTRY3(VAESIMC, V,dq, None,None, W,dq, vex4 cpuid(AES) p_66), |
464 | [0xdc] = X86_OP_ENTRY3(VAESENC, V,x, H,x, W,x, vex4 cpuid(AES) p_66), | |
465 | [0xdd] = X86_OP_ENTRY3(VAESENCLAST, V,x, H,x, W,x, vex4 cpuid(AES) p_66), | |
466 | [0xde] = X86_OP_ENTRY3(VAESDEC, V,x, H,x, W,x, vex4 cpuid(AES) p_66), | |
467 | [0xdf] = X86_OP_ENTRY3(VAESDECLAST, V,x, H,x, W,x, vex4 cpuid(AES) p_66), | |
b3e22b23 PB |
468 | }; |
469 | ||
470 | /* five rows for no prefix, 66, F3, F2, 66+F2 */ | |
471 | static const X86OpEntry opcodes_0F38_F0toFF[16][5] = { | |
1d0b9261 PB |
472 | [0] = { |
473 | X86_OP_ENTRY3(MOVBE, G,y, M,y, None,None, cpuid(MOVBE)), | |
474 | X86_OP_ENTRY3(MOVBE, G,w, M,w, None,None, cpuid(MOVBE)), | |
475 | {}, | |
476 | X86_OP_ENTRY2(CRC32, G,d, E,b, cpuid(SSE42)), | |
477 | X86_OP_ENTRY2(CRC32, G,d, E,b, cpuid(SSE42)), | |
478 | }, | |
479 | [1] = { | |
480 | X86_OP_ENTRY3(MOVBE, M,y, G,y, None,None, cpuid(MOVBE)), | |
481 | X86_OP_ENTRY3(MOVBE, M,w, G,w, None,None, cpuid(MOVBE)), | |
482 | {}, | |
483 | X86_OP_ENTRY2(CRC32, G,d, E,y, cpuid(SSE42)), | |
484 | X86_OP_ENTRY2(CRC32, G,d, E,w, cpuid(SSE42)), | |
485 | }, | |
486 | [2] = { | |
487 | X86_OP_ENTRY3(ANDN, G,y, B,y, E,y, vex13 cpuid(BMI1)), | |
488 | {}, | |
489 | {}, | |
490 | {}, | |
491 | {}, | |
492 | }, | |
493 | [3] = { | |
494 | X86_OP_GROUP3(group17, B,y, E,y, None,None, vex13 cpuid(BMI1)), | |
495 | {}, | |
496 | {}, | |
497 | {}, | |
498 | {}, | |
499 | }, | |
500 | [5] = { | |
501 | X86_OP_ENTRY3(BZHI, G,y, E,y, B,y, vex13 cpuid(BMI1)), | |
502 | {}, | |
503 | X86_OP_ENTRY3(PEXT, G,y, B,y, E,y, vex13 cpuid(BMI2)), | |
504 | X86_OP_ENTRY3(PDEP, G,y, B,y, E,y, vex13 cpuid(BMI2)), | |
505 | {}, | |
506 | }, | |
507 | [6] = { | |
508 | {}, | |
509 | X86_OP_ENTRY2(ADCX, G,y, E,y, cpuid(ADX)), | |
510 | X86_OP_ENTRY2(ADOX, G,y, E,y, cpuid(ADX)), | |
511 | X86_OP_ENTRY3(MULX, /* B,y, */ G,y, E,y, 2,y, vex13 cpuid(BMI2)), | |
512 | {}, | |
513 | }, | |
514 | [7] = { | |
515 | X86_OP_ENTRY3(BEXTR, G,y, E,y, B,y, vex13 cpuid(BMI1)), | |
516 | X86_OP_ENTRY3(SHLX, G,y, E,y, B,y, vex13 cpuid(BMI1)), | |
517 | X86_OP_ENTRY3(SARX, G,y, E,y, B,y, vex13 cpuid(BMI1)), | |
518 | X86_OP_ENTRY3(SHRX, G,y, E,y, B,y, vex13 cpuid(BMI1)), | |
519 | {}, | |
520 | }, | |
b3e22b23 PB |
521 | }; |
522 | ||
523 | static void decode_0F38(DisasContext *s, CPUX86State *env, X86OpEntry *entry, uint8_t *b) | |
524 | { | |
525 | *b = x86_ldub_code(env, s); | |
526 | if (*b < 0xf0) { | |
527 | *entry = opcodes_0F38_00toEF[*b]; | |
528 | } else { | |
529 | int row = 0; | |
530 | if (s->prefix & PREFIX_REPZ) { | |
531 | /* The REPZ (F3) prefix has priority over 66 */ | |
532 | row = 2; | |
533 | } else { | |
534 | row += s->prefix & PREFIX_REPNZ ? 3 : 0; | |
535 | row += s->prefix & PREFIX_DATA ? 1 : 0; | |
536 | } | |
537 | *entry = opcodes_0F38_F0toFF[*b & 15][row]; | |
538 | } | |
539 | } | |
540 | ||
79068477 PB |
541 | static void decode_VINSERTPS(DisasContext *s, CPUX86State *env, X86OpEntry *entry, uint8_t *b) |
542 | { | |
543 | static const X86OpEntry | |
544 | vinsertps_reg = X86_OP_ENTRY4(VINSERTPS_r, V,dq, H,dq, U,dq, vex5 cpuid(SSE41) p_66), | |
545 | vinsertps_mem = X86_OP_ENTRY4(VINSERTPS_m, V,dq, H,dq, M,d, vex5 cpuid(SSE41) p_66); | |
546 | ||
547 | int modrm = get_modrm(s, env); | |
548 | *entry = (modrm >> 6) == 3 ? vinsertps_reg : vinsertps_mem; | |
549 | } | |
550 | ||
b3e22b23 | 551 | static const X86OpEntry opcodes_0F3A[256] = { |
79068477 PB |
552 | /* |
553 | * These are VEX-only, but incorrectly listed in the manual as exception type 4. | |
554 | * Also the "qq" instructions are sometimes omitted by Table 2-17, but are VEX256 | |
555 | * only. | |
556 | */ | |
557 | [0x00] = X86_OP_ENTRY3(VPERMQ, V,qq, W,qq, I,b, vex6 cpuid(AVX2) p_66), | |
558 | [0x01] = X86_OP_ENTRY3(VPERMQ, V,qq, W,qq, I,b, vex6 cpuid(AVX2) p_66), /* VPERMPD */ | |
559 | [0x02] = X86_OP_ENTRY4(VBLENDPS, V,x, H,x, W,x, vex6 cpuid(AVX2) p_66), /* VPBLENDD */ | |
560 | [0x04] = X86_OP_ENTRY3(VPERMILPS_i, V,x, W,x, I,b, vex6 cpuid(AVX) p_66), | |
561 | [0x05] = X86_OP_ENTRY3(VPERMILPD_i, V,x, W,x, I,b, vex6 cpuid(AVX) p_66), | |
562 | [0x06] = X86_OP_ENTRY4(VPERM2x128, V,qq, H,qq, W,qq, vex6 cpuid(AVX) p_66), | |
563 | ||
564 | [0x14] = X86_OP_ENTRY3(PEXTRB, E,b, V,dq, I,b, vex5 cpuid(SSE41) zext0 p_66), | |
565 | [0x15] = X86_OP_ENTRY3(PEXTRW, E,w, V,dq, I,b, vex5 cpuid(SSE41) zext0 p_66), | |
566 | [0x16] = X86_OP_ENTRY3(PEXTR, E,y, V,dq, I,b, vex5 cpuid(SSE41) p_66), | |
567 | [0x17] = X86_OP_ENTRY3(VEXTRACTPS, E,d, V,dq, I,b, vex5 cpuid(SSE41) p_66), | |
a48b2697 | 568 | [0x1d] = X86_OP_ENTRY3(VCVTPS2PH, W,xh, V,x, I,b, vex11 cpuid(F16C) p_66), |
79068477 PB |
569 | |
570 | [0x20] = X86_OP_ENTRY4(PINSRB, V,dq, H,dq, E,b, vex5 cpuid(SSE41) zext2 p_66), | |
571 | [0x21] = X86_OP_GROUP0(VINSERTPS), | |
572 | [0x22] = X86_OP_ENTRY4(PINSR, V,dq, H,dq, E,y, vex5 cpuid(SSE41) p_66), | |
573 | ||
574 | [0x40] = X86_OP_ENTRY4(VDDPS, V,x, H,x, W,x, vex2 cpuid(SSE41) p_66), | |
575 | [0x41] = X86_OP_ENTRY4(VDDPD, V,dq, H,dq, W,dq, vex2 cpuid(SSE41) p_66), | |
576 | [0x42] = X86_OP_ENTRY4(VMPSADBW, V,x, H,x, W,x, vex2 cpuid(SSE41) avx2_256 p_66), | |
577 | [0x44] = X86_OP_ENTRY4(PCLMULQDQ, V,dq, H,dq, W,dq, vex4 cpuid(PCLMULQDQ) p_66), | |
578 | [0x46] = X86_OP_ENTRY4(VPERM2x128, V,qq, H,qq, W,qq, vex6 cpuid(AVX2) p_66), | |
579 | ||
580 | [0x60] = X86_OP_ENTRY4(PCMPESTRM, None,None, V,dq, W,dq, vex4_unal cpuid(SSE42) p_66), | |
581 | [0x61] = X86_OP_ENTRY4(PCMPESTRI, None,None, V,dq, W,dq, vex4_unal cpuid(SSE42) p_66), | |
582 | [0x62] = X86_OP_ENTRY4(PCMPISTRM, None,None, V,dq, W,dq, vex4_unal cpuid(SSE42) p_66), | |
583 | [0x63] = X86_OP_ENTRY4(PCMPISTRI, None,None, V,dq, W,dq, vex4_unal cpuid(SSE42) p_66), | |
584 | ||
585 | [0x08] = X86_OP_ENTRY3(VROUNDPS, V,x, W,x, I,b, vex2 cpuid(SSE41) p_66), | |
586 | [0x09] = X86_OP_ENTRY3(VROUNDPD, V,x, W,x, I,b, vex2 cpuid(SSE41) p_66), | |
587 | /* | |
588 | * Not listed as four operand in the manual. Also writes and reads 128-bits | |
589 | * from the first two operands due to the V operand picking higher entries of | |
590 | * the H operand; the "Vss,Hss,Wss" description from the manual is incorrect. | |
591 | * For other unary operations such as VSQRTSx this is hidden by the "REPScalar" | |
592 | * value of vex_special, because the table lists the operand types of VSQRTPx. | |
593 | */ | |
594 | [0x0a] = X86_OP_ENTRY4(VROUNDSS, V,x, H,x, W,ss, vex3 cpuid(SSE41) p_66), | |
595 | [0x0b] = X86_OP_ENTRY4(VROUNDSD, V,x, H,x, W,sd, vex3 cpuid(SSE41) p_66), | |
596 | [0x0c] = X86_OP_ENTRY4(VBLENDPS, V,x, H,x, W,x, vex4 cpuid(SSE41) p_66), | |
597 | [0x0d] = X86_OP_ENTRY4(VBLENDPD, V,x, H,x, W,x, vex4 cpuid(SSE41) p_66), | |
16fc5726 PB |
598 | [0x0e] = X86_OP_ENTRY4(VPBLENDW, V,x, H,x, W,x, vex4 cpuid(SSE41) avx2_256 p_66), |
599 | [0x0f] = X86_OP_ENTRY4(PALIGNR, V,x, H,x, W,x, vex4 cpuid(SSSE3) mmx avx2_256 p_00_66), | |
79068477 PB |
600 | |
601 | [0x18] = X86_OP_ENTRY4(VINSERTx128, V,qq, H,qq, W,qq, vex6 cpuid(AVX) p_66), | |
602 | [0x19] = X86_OP_ENTRY3(VEXTRACTx128, W,dq, V,qq, I,b, vex6 cpuid(AVX) p_66), | |
603 | ||
604 | [0x38] = X86_OP_ENTRY4(VINSERTx128, V,qq, H,qq, W,qq, vex6 cpuid(AVX2) p_66), | |
605 | [0x39] = X86_OP_ENTRY3(VEXTRACTx128, W,dq, V,qq, I,b, vex6 cpuid(AVX2) p_66), | |
606 | ||
607 | /* Listed incorrectly as type 4 */ | |
608 | [0x4a] = X86_OP_ENTRY4(VBLENDVPS, V,x, H,x, W,x, vex6 cpuid(AVX) p_66), | |
609 | [0x4b] = X86_OP_ENTRY4(VBLENDVPD, V,x, H,x, W,x, vex6 cpuid(AVX) p_66), | |
610 | [0x4c] = X86_OP_ENTRY4(VPBLENDVB, V,x, H,x, W,x, vex6 cpuid(AVX) p_66 avx2_256), | |
611 | ||
612 | [0xdf] = X86_OP_ENTRY3(VAESKEYGEN, V,dq, W,dq, I,b, vex4 cpuid(AES) p_66), | |
613 | ||
1d0b9261 | 614 | [0xF0] = X86_OP_ENTRY3(RORX, G,y, E,y, I,b, vex13 cpuid(BMI2) p_f2), |
b3e22b23 PB |
615 | }; |
616 | ||
617 | static void decode_0F3A(DisasContext *s, CPUX86State *env, X86OpEntry *entry, uint8_t *b) | |
618 | { | |
619 | *b = x86_ldub_code(env, s); | |
620 | *entry = opcodes_0F3A[*b]; | |
621 | } | |
622 | ||
7170a17e PB |
623 | /* |
624 | * There are some mistakes in the operands in the manual, and the load/store/register | |
625 | * cases are easiest to keep separate, so the entries for 10-17 follow simplicity and | |
626 | * efficiency of implementation rather than copying what the manual says. | |
627 | * | |
628 | * In particular: | |
629 | * | |
630 | * 1) "VMOVSS m32, xmm1" and "VMOVSD m64, xmm1" do not support VEX.vvvv != 1111b, | |
631 | * but this is not mentioned in the tables. | |
632 | * | |
633 | * 2) MOVHLPS, MOVHPS, MOVHPD, MOVLPD, MOVLPS read the high quadword of one of their | |
634 | * operands, which must therefore be dq; MOVLPD and MOVLPS also write the high | |
635 | * quadword of the V operand. | |
636 | */ | |
637 | static void decode_0F10(DisasContext *s, CPUX86State *env, X86OpEntry *entry, uint8_t *b) | |
638 | { | |
639 | static const X86OpEntry opcodes_0F10_reg[4] = { | |
640 | X86_OP_ENTRY3(MOVDQ, V,x, None,None, W,x, vex4_unal), /* MOVUPS */ | |
641 | X86_OP_ENTRY3(MOVDQ, V,x, None,None, W,x, vex4_unal), /* MOVUPD */ | |
cab529b0 RZ |
642 | X86_OP_ENTRY3(VMOVSS, V,x, H,x, W,x, vex5), |
643 | X86_OP_ENTRY3(VMOVLPx, V,x, H,x, W,x, vex5), /* MOVSD */ | |
7170a17e PB |
644 | }; |
645 | ||
646 | static const X86OpEntry opcodes_0F10_mem[4] = { | |
647 | X86_OP_ENTRY3(MOVDQ, V,x, None,None, W,x, vex4_unal), /* MOVUPS */ | |
648 | X86_OP_ENTRY3(MOVDQ, V,x, None,None, W,x, vex4_unal), /* MOVUPD */ | |
cab529b0 RZ |
649 | X86_OP_ENTRY3(VMOVSS_ld, V,x, H,x, M,ss, vex5), |
650 | X86_OP_ENTRY3(VMOVSD_ld, V,x, H,x, M,sd, vex5), | |
7170a17e PB |
651 | }; |
652 | ||
653 | if ((get_modrm(s, env) >> 6) == 3) { | |
654 | *entry = *decode_by_prefix(s, opcodes_0F10_reg); | |
655 | } else { | |
656 | *entry = *decode_by_prefix(s, opcodes_0F10_mem); | |
657 | } | |
658 | } | |
659 | ||
660 | static void decode_0F11(DisasContext *s, CPUX86State *env, X86OpEntry *entry, uint8_t *b) | |
661 | { | |
662 | static const X86OpEntry opcodes_0F11_reg[4] = { | |
afa94dab RZ |
663 | X86_OP_ENTRY3(MOVDQ, W,x, None,None, V,x, vex4), /* MOVUPS */ |
664 | X86_OP_ENTRY3(MOVDQ, W,x, None,None, V,x, vex4), /* MOVUPD */ | |
cab529b0 RZ |
665 | X86_OP_ENTRY3(VMOVSS, W,x, H,x, V,x, vex5), |
666 | X86_OP_ENTRY3(VMOVLPx, W,x, H,x, V,q, vex5), /* MOVSD */ | |
7170a17e PB |
667 | }; |
668 | ||
669 | static const X86OpEntry opcodes_0F11_mem[4] = { | |
afa94dab RZ |
670 | X86_OP_ENTRY3(MOVDQ, W,x, None,None, V,x, vex4), /* MOVUPS */ |
671 | X86_OP_ENTRY3(MOVDQ, W,x, None,None, V,x, vex4), /* MOVUPD */ | |
cab529b0 RZ |
672 | X86_OP_ENTRY3(VMOVSS_st, M,ss, None,None, V,x, vex5), |
673 | X86_OP_ENTRY3(VMOVLPx_st, M,sd, None,None, V,x, vex5), /* MOVSD */ | |
7170a17e PB |
674 | }; |
675 | ||
676 | if ((get_modrm(s, env) >> 6) == 3) { | |
677 | *entry = *decode_by_prefix(s, opcodes_0F11_reg); | |
678 | } else { | |
679 | *entry = *decode_by_prefix(s, opcodes_0F11_mem); | |
680 | } | |
681 | } | |
682 | ||
683 | static void decode_0F12(DisasContext *s, CPUX86State *env, X86OpEntry *entry, uint8_t *b) | |
684 | { | |
685 | static const X86OpEntry opcodes_0F12_mem[4] = { | |
686 | /* | |
687 | * Use dq for operand for compatibility with gen_MOVSD and | |
688 | * to allow VEX128 only. | |
689 | */ | |
cab529b0 RZ |
690 | X86_OP_ENTRY3(VMOVLPx_ld, V,dq, H,dq, M,q, vex5), /* MOVLPS */ |
691 | X86_OP_ENTRY3(VMOVLPx_ld, V,dq, H,dq, M,q, vex5), /* MOVLPD */ | |
7170a17e | 692 | X86_OP_ENTRY3(VMOVSLDUP, V,x, None,None, W,x, vex4 cpuid(SSE3)), |
cab529b0 | 693 | X86_OP_ENTRY3(VMOVDDUP, V,x, None,None, WM,q, vex5 cpuid(SSE3)), /* qq if VEX.256 */ |
7170a17e PB |
694 | }; |
695 | static const X86OpEntry opcodes_0F12_reg[4] = { | |
cab529b0 RZ |
696 | X86_OP_ENTRY3(VMOVHLPS, V,dq, H,dq, U,dq, vex7), |
697 | X86_OP_ENTRY3(VMOVLPx, W,x, H,x, U,q, vex5), /* MOVLPD */ | |
7170a17e | 698 | X86_OP_ENTRY3(VMOVSLDUP, V,x, None,None, U,x, vex4 cpuid(SSE3)), |
cab529b0 | 699 | X86_OP_ENTRY3(VMOVDDUP, V,x, None,None, U,x, vex5 cpuid(SSE3)), |
7170a17e PB |
700 | }; |
701 | ||
702 | if ((get_modrm(s, env) >> 6) == 3) { | |
703 | *entry = *decode_by_prefix(s, opcodes_0F12_reg); | |
704 | } else { | |
705 | *entry = *decode_by_prefix(s, opcodes_0F12_mem); | |
706 | if ((s->prefix & PREFIX_REPNZ) && s->vex_l) { | |
707 | entry->s2 = X86_SIZE_qq; | |
708 | } | |
709 | } | |
710 | } | |
711 | ||
712 | static void decode_0F16(DisasContext *s, CPUX86State *env, X86OpEntry *entry, uint8_t *b) | |
713 | { | |
714 | static const X86OpEntry opcodes_0F16_mem[4] = { | |
715 | /* | |
716 | * Operand 1 technically only reads the low 64 bits, but uses dq so that | |
717 | * it is easier to check for op0 == op1 in an endianness-neutral manner. | |
718 | */ | |
cab529b0 RZ |
719 | X86_OP_ENTRY3(VMOVHPx_ld, V,dq, H,dq, M,q, vex5), /* MOVHPS */ |
720 | X86_OP_ENTRY3(VMOVHPx_ld, V,dq, H,dq, M,q, vex5), /* MOVHPD */ | |
7170a17e PB |
721 | X86_OP_ENTRY3(VMOVSHDUP, V,x, None,None, W,x, vex4 cpuid(SSE3)), |
722 | {}, | |
723 | }; | |
724 | static const X86OpEntry opcodes_0F16_reg[4] = { | |
725 | /* Same as above, operand 1 could be Hq if it wasn't for big-endian. */ | |
cab529b0 RZ |
726 | X86_OP_ENTRY3(VMOVLHPS, V,dq, H,dq, U,q, vex7), |
727 | X86_OP_ENTRY3(VMOVHPx, V,x, H,x, U,x, vex5), /* MOVHPD */ | |
7170a17e PB |
728 | X86_OP_ENTRY3(VMOVSHDUP, V,x, None,None, U,x, vex4 cpuid(SSE3)), |
729 | {}, | |
730 | }; | |
731 | ||
732 | if ((get_modrm(s, env) >> 6) == 3) { | |
733 | *entry = *decode_by_prefix(s, opcodes_0F16_reg); | |
734 | } else { | |
735 | *entry = *decode_by_prefix(s, opcodes_0F16_mem); | |
736 | } | |
737 | } | |
738 | ||
f8d19eec PB |
739 | static void decode_0F2A(DisasContext *s, CPUX86State *env, X86OpEntry *entry, uint8_t *b) |
740 | { | |
741 | static const X86OpEntry opcodes_0F2A[4] = { | |
742 | X86_OP_ENTRY3(CVTPI2Px, V,x, None,None, Q,q), | |
743 | X86_OP_ENTRY3(CVTPI2Px, V,x, None,None, Q,q), | |
744 | X86_OP_ENTRY3(VCVTSI2Sx, V,x, H,x, E,y, vex3), | |
745 | X86_OP_ENTRY3(VCVTSI2Sx, V,x, H,x, E,y, vex3), | |
746 | }; | |
747 | *entry = *decode_by_prefix(s, opcodes_0F2A); | |
748 | } | |
749 | ||
750 | static void decode_0F2B(DisasContext *s, CPUX86State *env, X86OpEntry *entry, uint8_t *b) | |
751 | { | |
752 | static const X86OpEntry opcodes_0F2B[4] = { | |
8bf171c2 RZ |
753 | X86_OP_ENTRY3(MOVDQ, M,x, None,None, V,x, vex1), /* MOVNTPS */ |
754 | X86_OP_ENTRY3(MOVDQ, M,x, None,None, V,x, vex1), /* MOVNTPD */ | |
755 | /* AMD extensions */ | |
f8d19eec PB |
756 | X86_OP_ENTRY3(VMOVSS_st, M,ss, None,None, V,x, vex4 cpuid(SSE4A)), /* MOVNTSS */ |
757 | X86_OP_ENTRY3(VMOVLPx_st, M,sd, None,None, V,x, vex4 cpuid(SSE4A)), /* MOVNTSD */ | |
758 | }; | |
759 | ||
760 | *entry = *decode_by_prefix(s, opcodes_0F2B); | |
761 | } | |
762 | ||
763 | static void decode_0F2C(DisasContext *s, CPUX86State *env, X86OpEntry *entry, uint8_t *b) | |
764 | { | |
765 | static const X86OpEntry opcodes_0F2C[4] = { | |
766 | /* Listed as ps/pd in the manual, but CVTTPS2PI only reads 64-bit. */ | |
767 | X86_OP_ENTRY3(CVTTPx2PI, P,q, None,None, W,q), | |
768 | X86_OP_ENTRY3(CVTTPx2PI, P,q, None,None, W,dq), | |
769 | X86_OP_ENTRY3(VCVTTSx2SI, G,y, None,None, W,ss, vex3), | |
770 | X86_OP_ENTRY3(VCVTTSx2SI, G,y, None,None, W,sd, vex3), | |
771 | }; | |
772 | *entry = *decode_by_prefix(s, opcodes_0F2C); | |
773 | } | |
774 | ||
775 | static void decode_0F2D(DisasContext *s, CPUX86State *env, X86OpEntry *entry, uint8_t *b) | |
776 | { | |
777 | static const X86OpEntry opcodes_0F2D[4] = { | |
778 | /* Listed as ps/pd in the manual, but CVTPS2PI only reads 64-bit. */ | |
779 | X86_OP_ENTRY3(CVTPx2PI, P,q, None,None, W,q), | |
780 | X86_OP_ENTRY3(CVTPx2PI, P,q, None,None, W,dq), | |
781 | X86_OP_ENTRY3(VCVTSx2SI, G,y, None,None, W,ss, vex3), | |
782 | X86_OP_ENTRY3(VCVTSx2SI, G,y, None,None, W,sd, vex3), | |
783 | }; | |
784 | *entry = *decode_by_prefix(s, opcodes_0F2D); | |
785 | } | |
786 | ||
2b55e479 PB |
787 | static void decode_VxCOMISx(DisasContext *s, CPUX86State *env, X86OpEntry *entry, uint8_t *b) |
788 | { | |
789 | /* | |
790 | * VUCOMISx and VCOMISx are different and use no-prefix and 0x66 for SS and SD | |
791 | * respectively. Scalar values usually are associated with 0xF2 and 0xF3, for | |
792 | * which X86_VEX_REPScalar exists, but here it has to be decoded by hand. | |
793 | */ | |
794 | entry->s1 = entry->s2 = (s->prefix & PREFIX_DATA ? X86_SIZE_sd : X86_SIZE_ss); | |
795 | entry->gen = (*b == 0x2E ? gen_VUCOMI : gen_VCOMI); | |
796 | } | |
797 | ||
03b45880 PB |
798 | static void decode_sse_unary(DisasContext *s, CPUX86State *env, X86OpEntry *entry, uint8_t *b) |
799 | { | |
800 | if (!(s->prefix & (PREFIX_REPZ | PREFIX_REPNZ))) { | |
801 | entry->op1 = X86_TYPE_None; | |
802 | entry->s1 = X86_SIZE_None; | |
803 | } | |
804 | switch (*b) { | |
805 | case 0x51: entry->gen = gen_VSQRT; break; | |
806 | case 0x52: entry->gen = gen_VRSQRT; break; | |
807 | case 0x53: entry->gen = gen_VRCP; break; | |
03b45880 PB |
808 | } |
809 | } | |
810 | ||
abd41884 PB |
811 | static void decode_0F5A(DisasContext *s, CPUX86State *env, X86OpEntry *entry, uint8_t *b) |
812 | { | |
813 | static const X86OpEntry opcodes_0F5A[4] = { | |
814 | X86_OP_ENTRY2(VCVTPS2PD, V,x, W,xh, vex2), /* VCVTPS2PD */ | |
815 | X86_OP_ENTRY2(VCVTPD2PS, V,x, W,x, vex2), /* VCVTPD2PS */ | |
816 | X86_OP_ENTRY3(VCVTSS2SD, V,x, H,x, W,x, vex2_rep3), /* VCVTSS2SD */ | |
817 | X86_OP_ENTRY3(VCVTSD2SS, V,x, H,x, W,x, vex2_rep3), /* VCVTSD2SS */ | |
818 | }; | |
819 | *entry = *decode_by_prefix(s, opcodes_0F5A); | |
820 | } | |
821 | ||
03b45880 PB |
822 | static void decode_0F5B(DisasContext *s, CPUX86State *env, X86OpEntry *entry, uint8_t *b) |
823 | { | |
824 | static const X86OpEntry opcodes_0F5B[4] = { | |
825 | X86_OP_ENTRY2(VCVTDQ2PS, V,x, W,x, vex2), | |
826 | X86_OP_ENTRY2(VCVTPS2DQ, V,x, W,x, vex2), | |
827 | X86_OP_ENTRY2(VCVTTPS2DQ, V,x, W,x, vex2), | |
828 | {}, | |
829 | }; | |
830 | *entry = *decode_by_prefix(s, opcodes_0F5B); | |
831 | } | |
832 | ||
6bbeb98d PB |
833 | static void decode_0FE6(DisasContext *s, CPUX86State *env, X86OpEntry *entry, uint8_t *b) |
834 | { | |
835 | static const X86OpEntry opcodes_0FE6[4] = { | |
836 | {}, | |
837 | X86_OP_ENTRY2(VCVTTPD2DQ, V,x, W,x, vex2), | |
cab529b0 | 838 | X86_OP_ENTRY2(VCVTDQ2PD, V,x, W,x, vex5), |
6bbeb98d PB |
839 | X86_OP_ENTRY2(VCVTPD2DQ, V,x, W,x, vex2), |
840 | }; | |
841 | *entry = *decode_by_prefix(s, opcodes_0FE6); | |
842 | } | |
843 | ||
b3e22b23 | 844 | static const X86OpEntry opcodes_0F[256] = { |
71a0891d PB |
845 | [0x0E] = X86_OP_ENTRY0(EMMS, cpuid(3DNOW)), /* femms */ |
846 | /* | |
847 | * 3DNow!'s opcode byte comes *after* modrm and displacements, making it | |
848 | * more like an Ib operand. Dispatch to the right helper in a single gen_* | |
849 | * function. | |
850 | */ | |
851 | [0x0F] = X86_OP_ENTRY3(3dnow, P,q, Q,q, I,b, cpuid(3DNOW)), | |
852 | ||
7170a17e PB |
853 | [0x10] = X86_OP_GROUP0(0F10), |
854 | [0x11] = X86_OP_GROUP0(0F11), | |
855 | [0x12] = X86_OP_GROUP0(0F12), | |
cab529b0 | 856 | [0x13] = X86_OP_ENTRY3(VMOVLPx_st, M,q, None,None, V,q, vex5 p_00_66), |
7170a17e PB |
857 | [0x14] = X86_OP_ENTRY3(VUNPCKLPx, V,x, H,x, W,x, vex4 p_00_66), |
858 | [0x15] = X86_OP_ENTRY3(VUNPCKHPx, V,x, H,x, W,x, vex4 p_00_66), | |
859 | [0x16] = X86_OP_GROUP0(0F16), | |
860 | /* Incorrectly listed as Mq,Vq in the manual */ | |
cab529b0 | 861 | [0x17] = X86_OP_ENTRY3(VMOVHPx_st, M,q, None,None, V,dq, vex5 p_00_66), |
7170a17e | 862 | |
03b45880 | 863 | [0x50] = X86_OP_ENTRY3(MOVMSK, G,y, None,None, U,x, vex7 p_00_66), |
afa94dab RZ |
864 | [0x51] = X86_OP_GROUP3(sse_unary, V,x, H,x, W,x, vex2_rep3 p_00_66_f3_f2), /* sqrtps */ |
865 | [0x52] = X86_OP_GROUP3(sse_unary, V,x, H,x, W,x, vex4_rep5 p_00_f3), /* rsqrtps */ | |
866 | [0x53] = X86_OP_GROUP3(sse_unary, V,x, H,x, W,x, vex4_rep5 p_00_f3), /* rcpps */ | |
03b45880 PB |
867 | [0x54] = X86_OP_ENTRY3(PAND, V,x, H,x, W,x, vex4 p_00_66), /* vand */ |
868 | [0x55] = X86_OP_ENTRY3(PANDN, V,x, H,x, W,x, vex4 p_00_66), /* vandn */ | |
869 | [0x56] = X86_OP_ENTRY3(POR, V,x, H,x, W,x, vex4 p_00_66), /* vor */ | |
870 | [0x57] = X86_OP_ENTRY3(PXOR, V,x, H,x, W,x, vex4 p_00_66), /* vxor */ | |
871 | ||
92ec056a PB |
872 | [0x60] = X86_OP_ENTRY3(PUNPCKLBW, V,x, H,x, W,x, vex4 mmx avx2_256 p_00_66), |
873 | [0x61] = X86_OP_ENTRY3(PUNPCKLWD, V,x, H,x, W,x, vex4 mmx avx2_256 p_00_66), | |
874 | [0x62] = X86_OP_ENTRY3(PUNPCKLDQ, V,x, H,x, W,x, vex4 mmx avx2_256 p_00_66), | |
875 | [0x63] = X86_OP_ENTRY3(PACKSSWB, V,x, H,x, W,x, vex4 mmx avx2_256 p_00_66), | |
876 | [0x64] = X86_OP_ENTRY3(PCMPGTB, V,x, H,x, W,x, vex4 mmx avx2_256 p_00_66), | |
877 | [0x65] = X86_OP_ENTRY3(PCMPGTW, V,x, H,x, W,x, vex4 mmx avx2_256 p_00_66), | |
878 | [0x66] = X86_OP_ENTRY3(PCMPGTD, V,x, H,x, W,x, vex4 mmx avx2_256 p_00_66), | |
879 | [0x67] = X86_OP_ENTRY3(PACKUSWB, V,x, H,x, W,x, vex4 mmx avx2_256 p_00_66), | |
880 | ||
ce4fcb94 PB |
881 | [0x70] = X86_OP_GROUP0(0F70), |
882 | [0x71] = X86_OP_GROUP0(group12), | |
883 | [0x72] = X86_OP_GROUP0(group13), | |
884 | [0x73] = X86_OP_GROUP0(group14), | |
885 | [0x74] = X86_OP_ENTRY3(PCMPEQB, V,x, H,x, W,x, vex4 mmx avx2_256 p_00_66), | |
886 | [0x75] = X86_OP_ENTRY3(PCMPEQW, V,x, H,x, W,x, vex4 mmx avx2_256 p_00_66), | |
887 | [0x76] = X86_OP_ENTRY3(PCMPEQD, V,x, H,x, W,x, vex4 mmx avx2_256 p_00_66), | |
888 | [0x77] = X86_OP_GROUP0(0F77), | |
889 | ||
f8d19eec PB |
890 | [0x28] = X86_OP_ENTRY3(MOVDQ, V,x, None,None, W,x, vex1 p_00_66), /* MOVAPS */ |
891 | [0x29] = X86_OP_ENTRY3(MOVDQ, W,x, None,None, V,x, vex1 p_00_66), /* MOVAPS */ | |
892 | [0x2A] = X86_OP_GROUP0(0F2A), | |
893 | [0x2B] = X86_OP_GROUP0(0F2B), | |
894 | [0x2C] = X86_OP_GROUP0(0F2C), | |
895 | [0x2D] = X86_OP_GROUP0(0F2D), | |
2b55e479 PB |
896 | [0x2E] = X86_OP_GROUP3(VxCOMISx, None,None, V,x, W,x, vex3 p_00_66), /* VUCOMISS/SD */ |
897 | [0x2F] = X86_OP_GROUP3(VxCOMISx, None,None, V,x, W,x, vex3 p_00_66), /* VCOMISS/SD */ | |
f8d19eec | 898 | |
b3e22b23 PB |
899 | [0x38] = X86_OP_GROUP0(0F38), |
900 | [0x3a] = X86_OP_GROUP0(0F3A), | |
92ec056a | 901 | |
03b45880 PB |
902 | [0x58] = X86_OP_ENTRY3(VADD, V,x, H,x, W,x, vex2_rep3 p_00_66_f3_f2), |
903 | [0x59] = X86_OP_ENTRY3(VMUL, V,x, H,x, W,x, vex2_rep3 p_00_66_f3_f2), | |
abd41884 | 904 | [0x5a] = X86_OP_GROUP0(0F5A), |
03b45880 PB |
905 | [0x5b] = X86_OP_GROUP0(0F5B), |
906 | [0x5c] = X86_OP_ENTRY3(VSUB, V,x, H,x, W,x, vex2_rep3 p_00_66_f3_f2), | |
907 | [0x5d] = X86_OP_ENTRY3(VMIN, V,x, H,x, W,x, vex2_rep3 p_00_66_f3_f2), | |
908 | [0x5e] = X86_OP_ENTRY3(VDIV, V,x, H,x, W,x, vex2_rep3 p_00_66_f3_f2), | |
909 | [0x5f] = X86_OP_ENTRY3(VMAX, V,x, H,x, W,x, vex2_rep3 p_00_66_f3_f2), | |
910 | ||
92ec056a PB |
911 | [0x68] = X86_OP_ENTRY3(PUNPCKHBW, V,x, H,x, W,x, vex4 mmx avx2_256 p_00_66), |
912 | [0x69] = X86_OP_ENTRY3(PUNPCKHWD, V,x, H,x, W,x, vex4 mmx avx2_256 p_00_66), | |
913 | [0x6a] = X86_OP_ENTRY3(PUNPCKHDQ, V,x, H,x, W,x, vex4 mmx avx2_256 p_00_66), | |
914 | [0x6b] = X86_OP_ENTRY3(PACKSSDW, V,x, H,x, W,x, vex4 mmx avx2_256 p_00_66), | |
915 | [0x6c] = X86_OP_ENTRY3(PUNPCKLQDQ, V,x, H,x, W,x, vex4 p_66 avx2_256), | |
916 | [0x6d] = X86_OP_ENTRY3(PUNPCKHQDQ, V,x, H,x, W,x, vex4 p_66 avx2_256), | |
917 | [0x6e] = X86_OP_ENTRY3(MOVD_to, V,x, None,None, E,y, vex5 mmx p_00_66), /* wrong dest Vy on SDM! */ | |
918 | [0x6f] = X86_OP_GROUP0(0F6F), | |
1d0efbdb | 919 | |
d1c1a422 PB |
920 | [0x78] = X86_OP_GROUP0(0F78), |
921 | [0x79] = X86_OP_GROUP2(0F79, V,x, U,x, cpuid(SSE4A)), | |
922 | [0x7c] = X86_OP_ENTRY3(VHADD, V,x, H,x, W,x, vex2 cpuid(SSE3) p_66_f2), | |
923 | [0x7d] = X86_OP_ENTRY3(VHSUB, V,x, H,x, W,x, vex2 cpuid(SSE3) p_66_f2), | |
924 | [0x7e] = X86_OP_GROUP0(0F7E), | |
925 | [0x7f] = X86_OP_GROUP0(0F7F), | |
926 | ||
57f6bba0 PB |
927 | [0xae] = X86_OP_GROUP0(group15), |
928 | ||
aba2b8ec PB |
929 | [0xc2] = X86_OP_ENTRY4(VCMP, V,x, H,x, W,x, vex2_rep3 p_00_66_f3_f2), |
930 | [0xc4] = X86_OP_ENTRY4(PINSRW, V,dq,H,dq,E,w, vex5 mmx p_00_66), | |
931 | [0xc5] = X86_OP_ENTRY3(PEXTRW, G,d, U,dq,I,b, vex5 mmx p_00_66), | |
932 | [0xc6] = X86_OP_ENTRY4(VSHUF, V,x, H,x, W,x, vex4 p_00_66), | |
933 | ||
6bbeb98d PB |
934 | [0xd0] = X86_OP_ENTRY3(VADDSUB, V,x, H,x, W,x, vex2 cpuid(SSE3) p_66_f2), |
935 | [0xd1] = X86_OP_ENTRY3(PSRLW_r, V,x, H,x, W,x, vex4 mmx avx2_256 p_00_66), | |
936 | [0xd2] = X86_OP_ENTRY3(PSRLD_r, V,x, H,x, W,x, vex4 mmx avx2_256 p_00_66), | |
937 | [0xd3] = X86_OP_ENTRY3(PSRLQ_r, V,x, H,x, W,x, vex4 mmx avx2_256 p_00_66), | |
938 | [0xd4] = X86_OP_ENTRY3(PADDQ, V,x, H,x, W,x, vex4 mmx avx2_256 p_00_66), | |
939 | [0xd5] = X86_OP_ENTRY3(PMULLW, V,x, H,x, W,x, vex4 mmx avx2_256 p_00_66), | |
940 | [0xd6] = X86_OP_GROUP0(0FD6), | |
941 | [0xd7] = X86_OP_ENTRY3(PMOVMSKB, G,d, None,None, U,x, vex7 mmx avx2_256 p_00_66), | |
942 | ||
943 | [0xe0] = X86_OP_ENTRY3(PAVGB, V,x, H,x, W,x, vex4 mmx avx2_256 p_00_66), | |
944 | [0xe1] = X86_OP_ENTRY3(PSRAW_r, V,x, H,x, W,x, vex7 mmx avx2_256 p_00_66), | |
945 | [0xe2] = X86_OP_ENTRY3(PSRAD_r, V,x, H,x, W,x, vex7 mmx avx2_256 p_00_66), | |
946 | [0xe3] = X86_OP_ENTRY3(PAVGW, V,x, H,x, W,x, vex4 mmx avx2_256 p_00_66), | |
947 | [0xe4] = X86_OP_ENTRY3(PMULHUW, V,x, H,x, W,x, vex4 mmx avx2_256 p_00_66), | |
948 | [0xe5] = X86_OP_ENTRY3(PMULHW, V,x, H,x, W,x, vex4 mmx avx2_256 p_00_66), | |
949 | [0xe6] = X86_OP_GROUP0(0FE6), | |
950 | [0xe7] = X86_OP_ENTRY3(MOVDQ, W,x, None,None, V,x, vex1 mmx p_00_66), /* MOVNTQ/MOVNTDQ */ | |
951 | ||
952 | [0xf0] = X86_OP_ENTRY3(MOVDQ, V,x, None,None, WM,x, vex4_unal cpuid(SSE3) p_f2), /* LDDQU */ | |
953 | [0xf1] = X86_OP_ENTRY3(PSLLW_r, V,x, H,x, W,x, vex7 mmx avx2_256 p_00_66), | |
954 | [0xf2] = X86_OP_ENTRY3(PSLLD_r, V,x, H,x, W,x, vex7 mmx avx2_256 p_00_66), | |
955 | [0xf3] = X86_OP_ENTRY3(PSLLQ_r, V,x, H,x, W,x, vex7 mmx avx2_256 p_00_66), | |
956 | [0xf4] = X86_OP_ENTRY3(PMULUDQ, V,x, H,x, W,x, vex4 mmx avx2_256 p_00_66), | |
957 | [0xf5] = X86_OP_ENTRY3(PMADDWD, V,x, H,x, W,x, vex4 mmx avx2_256 p_00_66), | |
958 | [0xf6] = X86_OP_ENTRY3(PSADBW, V,x, H,x, W,x, vex4 mmx avx2_256 p_00_66), | |
959 | [0xf7] = X86_OP_ENTRY3(MASKMOV, None,None, V,dq, U,dq, vex4_unal avx2_256 mmx p_00_66), | |
960 | ||
1d0efbdb PB |
961 | /* Incorrectly missing from 2-17 */ |
962 | [0xd8] = X86_OP_ENTRY3(PSUBUSB, V,x, H,x, W,x, vex4 mmx avx2_256 p_00_66), | |
963 | [0xd9] = X86_OP_ENTRY3(PSUBUSW, V,x, H,x, W,x, vex4 mmx avx2_256 p_00_66), | |
964 | [0xda] = X86_OP_ENTRY3(PMINUB, V,x, H,x, W,x, vex4 mmx avx2_256 p_00_66), | |
965 | [0xdb] = X86_OP_ENTRY3(PAND, V,x, H,x, W,x, vex4 mmx avx2_256 p_00_66), | |
966 | [0xdc] = X86_OP_ENTRY3(PADDUSB, V,x, H,x, W,x, vex4 mmx avx2_256 p_00_66), | |
967 | [0xdd] = X86_OP_ENTRY3(PADDUSW, V,x, H,x, W,x, vex4 mmx avx2_256 p_00_66), | |
968 | [0xde] = X86_OP_ENTRY3(PMAXUB, V,x, H,x, W,x, vex4 mmx avx2_256 p_00_66), | |
969 | [0xdf] = X86_OP_ENTRY3(PANDN, V,x, H,x, W,x, vex4 mmx avx2_256 p_00_66), | |
970 | ||
971 | [0xe8] = X86_OP_ENTRY3(PSUBSB, V,x, H,x, W,x, vex4 mmx avx2_256 p_00_66), | |
972 | [0xe9] = X86_OP_ENTRY3(PSUBSW, V,x, H,x, W,x, vex4 mmx avx2_256 p_00_66), | |
973 | [0xea] = X86_OP_ENTRY3(PMINSW, V,x, H,x, W,x, vex4 mmx avx2_256 p_00_66), | |
974 | [0xeb] = X86_OP_ENTRY3(POR, V,x, H,x, W,x, vex4 mmx avx2_256 p_00_66), | |
975 | [0xec] = X86_OP_ENTRY3(PADDSB, V,x, H,x, W,x, vex4 mmx avx2_256 p_00_66), | |
976 | [0xed] = X86_OP_ENTRY3(PADDSW, V,x, H,x, W,x, vex4 mmx avx2_256 p_00_66), | |
977 | [0xee] = X86_OP_ENTRY3(PMAXSW, V,x, H,x, W,x, vex4 mmx avx2_256 p_00_66), | |
978 | [0xef] = X86_OP_ENTRY3(PXOR, V,x, H,x, W,x, vex4 mmx avx2_256 p_00_66), | |
979 | ||
980 | [0xf8] = X86_OP_ENTRY3(PSUBB, V,x, H,x, W,x, vex4 mmx avx2_256 p_00_66), | |
981 | [0xf9] = X86_OP_ENTRY3(PSUBW, V,x, H,x, W,x, vex4 mmx avx2_256 p_00_66), | |
982 | [0xfa] = X86_OP_ENTRY3(PSUBD, V,x, H,x, W,x, vex4 mmx avx2_256 p_00_66), | |
983 | [0xfb] = X86_OP_ENTRY3(PSUBQ, V,x, H,x, W,x, vex4 mmx avx2_256 p_00_66), | |
984 | [0xfc] = X86_OP_ENTRY3(PADDB, V,x, H,x, W,x, vex4 mmx avx2_256 p_00_66), | |
985 | [0xfd] = X86_OP_ENTRY3(PADDW, V,x, H,x, W,x, vex4 mmx avx2_256 p_00_66), | |
986 | [0xfe] = X86_OP_ENTRY3(PADDD, V,x, H,x, W,x, vex4 mmx avx2_256 p_00_66), | |
987 | /* 0xff = UD0 */ | |
b3e22b23 PB |
988 | }; |
989 | ||
990 | static void do_decode_0F(DisasContext *s, CPUX86State *env, X86OpEntry *entry, uint8_t *b) | |
991 | { | |
992 | *entry = opcodes_0F[*b]; | |
993 | } | |
994 | ||
995 | static void decode_0F(DisasContext *s, CPUX86State *env, X86OpEntry *entry, uint8_t *b) | |
996 | { | |
997 | *b = x86_ldub_code(env, s); | |
998 | do_decode_0F(s, env, entry, b); | |
999 | } | |
1000 | ||
1001 | static const X86OpEntry opcodes_root[256] = { | |
1002 | [0x0F] = X86_OP_GROUP0(0F), | |
1003 | }; | |
1004 | ||
1005 | #undef mmx | |
20581aad PB |
1006 | #undef vex1 |
1007 | #undef vex2 | |
1008 | #undef vex3 | |
1009 | #undef vex4 | |
1010 | #undef vex4_unal | |
1011 | #undef vex5 | |
1012 | #undef vex6 | |
1013 | #undef vex7 | |
1014 | #undef vex8 | |
1015 | #undef vex11 | |
1016 | #undef vex12 | |
1017 | #undef vex13 | |
b3e22b23 PB |
1018 | |
1019 | /* | |
1020 | * Decode the fixed part of the opcode and place the last | |
1021 | * in b. | |
1022 | */ | |
1023 | static void decode_root(DisasContext *s, CPUX86State *env, X86OpEntry *entry, uint8_t *b) | |
1024 | { | |
1025 | *entry = opcodes_root[*b]; | |
1026 | } | |
1027 | ||
1028 | ||
1029 | static int decode_modrm(DisasContext *s, CPUX86State *env, X86DecodedInsn *decode, | |
1030 | X86DecodedOp *op, X86OpType type) | |
1031 | { | |
1032 | int modrm = get_modrm(s, env); | |
1033 | if ((modrm >> 6) == 3) { | |
1034 | if (s->prefix & PREFIX_LOCK) { | |
1035 | decode->e.gen = gen_illegal; | |
1036 | return 0xff; | |
1037 | } | |
1038 | op->n = (modrm & 7); | |
1039 | if (type != X86_TYPE_Q && type != X86_TYPE_N) { | |
1040 | op->n |= REX_B(s); | |
1041 | } | |
1042 | } else { | |
1043 | op->has_ea = true; | |
1044 | op->n = -1; | |
1045 | decode->mem = gen_lea_modrm_0(env, s, get_modrm(s, env)); | |
1046 | } | |
1047 | return modrm; | |
1048 | } | |
1049 | ||
1050 | static bool decode_op_size(DisasContext *s, X86OpEntry *e, X86OpSize size, MemOp *ot) | |
1051 | { | |
1052 | switch (size) { | |
1053 | case X86_SIZE_b: /* byte */ | |
1054 | *ot = MO_8; | |
1055 | return true; | |
1056 | ||
1057 | case X86_SIZE_d: /* 32-bit */ | |
1058 | case X86_SIZE_ss: /* SSE/AVX scalar single precision */ | |
1059 | *ot = MO_32; | |
1060 | return true; | |
1061 | ||
1062 | case X86_SIZE_p: /* Far pointer, return offset size */ | |
1063 | case X86_SIZE_s: /* Descriptor, return offset size */ | |
1064 | case X86_SIZE_v: /* 16/32/64-bit, based on operand size */ | |
1065 | *ot = s->dflag; | |
1066 | return true; | |
1067 | ||
1068 | case X86_SIZE_pi: /* MMX */ | |
1069 | case X86_SIZE_q: /* 64-bit */ | |
1070 | case X86_SIZE_sd: /* SSE/AVX scalar double precision */ | |
1071 | *ot = MO_64; | |
1072 | return true; | |
1073 | ||
1074 | case X86_SIZE_w: /* 16-bit */ | |
1075 | *ot = MO_16; | |
1076 | return true; | |
1077 | ||
1078 | case X86_SIZE_y: /* 32/64-bit, based on operand size */ | |
1079 | *ot = s->dflag == MO_16 ? MO_32 : s->dflag; | |
1080 | return true; | |
1081 | ||
1082 | case X86_SIZE_z: /* 16-bit for 16-bit operand size, else 32-bit */ | |
1083 | *ot = s->dflag == MO_16 ? MO_16 : MO_32; | |
1084 | return true; | |
1085 | ||
1086 | case X86_SIZE_dq: /* SSE/AVX 128-bit */ | |
1087 | if (e->special == X86_SPECIAL_MMX && | |
1088 | !(s->prefix & (PREFIX_DATA | PREFIX_REPZ | PREFIX_REPNZ))) { | |
1089 | *ot = MO_64; | |
1090 | return true; | |
1091 | } | |
1092 | if (s->vex_l && e->s0 != X86_SIZE_qq && e->s1 != X86_SIZE_qq) { | |
1093 | return false; | |
1094 | } | |
1095 | *ot = MO_128; | |
1096 | return true; | |
1097 | ||
1098 | case X86_SIZE_qq: /* AVX 256-bit */ | |
1099 | if (!s->vex_l) { | |
1100 | return false; | |
1101 | } | |
1102 | *ot = MO_256; | |
1103 | return true; | |
1104 | ||
1105 | case X86_SIZE_x: /* 128/256-bit, based on operand size */ | |
1106 | if (e->special == X86_SPECIAL_MMX && | |
1107 | !(s->prefix & (PREFIX_DATA | PREFIX_REPZ | PREFIX_REPNZ))) { | |
1108 | *ot = MO_64; | |
1109 | return true; | |
1110 | } | |
1111 | /* fall through */ | |
1112 | case X86_SIZE_ps: /* SSE/AVX packed single precision */ | |
1113 | case X86_SIZE_pd: /* SSE/AVX packed double precision */ | |
1114 | *ot = s->vex_l ? MO_256 : MO_128; | |
1115 | return true; | |
1116 | ||
a48b2697 | 1117 | case X86_SIZE_xh: /* SSE/AVX packed half register */ |
cf5ec664 PB |
1118 | *ot = s->vex_l ? MO_128 : MO_64; |
1119 | return true; | |
1120 | ||
b3e22b23 PB |
1121 | case X86_SIZE_d64: /* Default to 64-bit in 64-bit mode */ |
1122 | *ot = CODE64(s) && s->dflag == MO_32 ? MO_64 : s->dflag; | |
1123 | return true; | |
1124 | ||
1125 | case X86_SIZE_f64: /* Ignore size override prefix in 64-bit mode */ | |
1126 | *ot = CODE64(s) ? MO_64 : s->dflag; | |
1127 | return true; | |
1128 | ||
1129 | default: | |
1130 | *ot = -1; | |
1131 | return true; | |
1132 | } | |
1133 | } | |
1134 | ||
1135 | static bool decode_op(DisasContext *s, CPUX86State *env, X86DecodedInsn *decode, | |
1136 | X86DecodedOp *op, X86OpType type, int b) | |
1137 | { | |
1138 | int modrm; | |
1139 | ||
1140 | switch (type) { | |
1141 | case X86_TYPE_None: /* Implicit or absent */ | |
1142 | case X86_TYPE_A: /* Implicit */ | |
1143 | case X86_TYPE_F: /* EFLAGS/RFLAGS */ | |
1144 | break; | |
1145 | ||
1146 | case X86_TYPE_B: /* VEX.vvvv selects a GPR */ | |
1147 | op->unit = X86_OP_INT; | |
1148 | op->n = s->vex_v; | |
1149 | break; | |
1150 | ||
1151 | case X86_TYPE_C: /* REG in the modrm byte selects a control register */ | |
1152 | op->unit = X86_OP_CR; | |
1153 | goto get_reg; | |
1154 | ||
1155 | case X86_TYPE_D: /* REG in the modrm byte selects a debug register */ | |
1156 | op->unit = X86_OP_DR; | |
1157 | goto get_reg; | |
1158 | ||
1159 | case X86_TYPE_G: /* REG in the modrm byte selects a GPR */ | |
1160 | op->unit = X86_OP_INT; | |
1161 | goto get_reg; | |
1162 | ||
1163 | case X86_TYPE_S: /* reg selects a segment register */ | |
1164 | op->unit = X86_OP_SEG; | |
1165 | goto get_reg; | |
1166 | ||
1167 | case X86_TYPE_P: | |
1168 | op->unit = X86_OP_MMX; | |
1169 | goto get_reg; | |
1170 | ||
1171 | case X86_TYPE_V: /* reg in the modrm byte selects an XMM/YMM register */ | |
1172 | if (decode->e.special == X86_SPECIAL_MMX && | |
1173 | !(s->prefix & (PREFIX_DATA | PREFIX_REPZ | PREFIX_REPNZ))) { | |
1174 | op->unit = X86_OP_MMX; | |
1175 | } else { | |
1176 | op->unit = X86_OP_SSE; | |
1177 | } | |
1178 | get_reg: | |
1179 | op->n = ((get_modrm(s, env) >> 3) & 7) | REX_R(s); | |
1180 | break; | |
1181 | ||
1182 | case X86_TYPE_E: /* ALU modrm operand */ | |
1183 | op->unit = X86_OP_INT; | |
1184 | goto get_modrm; | |
1185 | ||
1186 | case X86_TYPE_Q: /* MMX modrm operand */ | |
1187 | op->unit = X86_OP_MMX; | |
1188 | goto get_modrm; | |
1189 | ||
1190 | case X86_TYPE_W: /* XMM/YMM modrm operand */ | |
1191 | if (decode->e.special == X86_SPECIAL_MMX && | |
1192 | !(s->prefix & (PREFIX_DATA | PREFIX_REPZ | PREFIX_REPNZ))) { | |
1193 | op->unit = X86_OP_MMX; | |
1194 | } else { | |
1195 | op->unit = X86_OP_SSE; | |
1196 | } | |
1197 | goto get_modrm; | |
1198 | ||
1199 | case X86_TYPE_N: /* R/M in the modrm byte selects an MMX register */ | |
1200 | op->unit = X86_OP_MMX; | |
1201 | goto get_modrm_reg; | |
1202 | ||
1203 | case X86_TYPE_U: /* R/M in the modrm byte selects an XMM/YMM register */ | |
1204 | if (decode->e.special == X86_SPECIAL_MMX && | |
1205 | !(s->prefix & (PREFIX_DATA | PREFIX_REPZ | PREFIX_REPNZ))) { | |
1206 | op->unit = X86_OP_MMX; | |
1207 | } else { | |
1208 | op->unit = X86_OP_SSE; | |
1209 | } | |
1210 | goto get_modrm_reg; | |
1211 | ||
1212 | case X86_TYPE_R: /* R/M in the modrm byte selects a register */ | |
1213 | op->unit = X86_OP_INT; | |
1214 | get_modrm_reg: | |
1215 | modrm = get_modrm(s, env); | |
1216 | if ((modrm >> 6) != 3) { | |
1217 | return false; | |
1218 | } | |
1219 | goto get_modrm; | |
1220 | ||
6bbeb98d PB |
1221 | case X86_TYPE_WM: /* modrm byte selects an XMM/YMM memory operand */ |
1222 | op->unit = X86_OP_SSE; | |
1223 | /* fall through */ | |
b3e22b23 PB |
1224 | case X86_TYPE_M: /* modrm byte selects a memory operand */ |
1225 | modrm = get_modrm(s, env); | |
1226 | if ((modrm >> 6) == 3) { | |
1227 | return false; | |
1228 | } | |
1229 | get_modrm: | |
1230 | decode_modrm(s, env, decode, op, type); | |
1231 | break; | |
1232 | ||
1233 | case X86_TYPE_O: /* Absolute address encoded in the instruction */ | |
1234 | op->unit = X86_OP_INT; | |
1235 | op->has_ea = true; | |
1236 | op->n = -1; | |
1237 | decode->mem = (AddressParts) { | |
1238 | .def_seg = R_DS, | |
1239 | .base = -1, | |
1240 | .index = -1, | |
1241 | .disp = insn_get_addr(env, s, s->aflag) | |
1242 | }; | |
1243 | break; | |
1244 | ||
1245 | case X86_TYPE_H: /* For AVX, VEX.vvvv selects an XMM/YMM register */ | |
1246 | if ((s->prefix & PREFIX_VEX)) { | |
1247 | op->unit = X86_OP_SSE; | |
1248 | op->n = s->vex_v; | |
1249 | break; | |
1250 | } | |
1251 | if (op == &decode->op[0]) { | |
1252 | /* shifts place the destination in VEX.vvvv, use modrm */ | |
1253 | return decode_op(s, env, decode, op, decode->e.op1, b); | |
1254 | } else { | |
1255 | return decode_op(s, env, decode, op, decode->e.op0, b); | |
1256 | } | |
1257 | ||
1258 | case X86_TYPE_I: /* Immediate */ | |
1259 | op->unit = X86_OP_IMM; | |
1260 | decode->immediate = insn_get_signed(env, s, op->ot); | |
1261 | break; | |
1262 | ||
1263 | case X86_TYPE_J: /* Relative offset for a jump */ | |
1264 | op->unit = X86_OP_IMM; | |
1265 | decode->immediate = insn_get_signed(env, s, op->ot); | |
1266 | decode->immediate += s->pc - s->cs_base; | |
1267 | if (s->dflag == MO_16) { | |
1268 | decode->immediate &= 0xffff; | |
1269 | } else if (!CODE64(s)) { | |
1270 | decode->immediate &= 0xffffffffu; | |
1271 | } | |
1272 | break; | |
1273 | ||
1274 | case X86_TYPE_L: /* The upper 4 bits of the immediate select a 128-bit register */ | |
1275 | op->n = insn_get(env, s, op->ot) >> 4; | |
1276 | break; | |
1277 | ||
1278 | case X86_TYPE_X: /* string source */ | |
1279 | op->n = -1; | |
1280 | decode->mem = (AddressParts) { | |
1281 | .def_seg = R_DS, | |
1282 | .base = R_ESI, | |
1283 | .index = -1, | |
1284 | }; | |
1285 | break; | |
1286 | ||
1287 | case X86_TYPE_Y: /* string destination */ | |
1288 | op->n = -1; | |
1289 | decode->mem = (AddressParts) { | |
1290 | .def_seg = R_ES, | |
1291 | .base = R_EDI, | |
1292 | .index = -1, | |
1293 | }; | |
1294 | break; | |
1295 | ||
1296 | case X86_TYPE_2op: | |
1297 | *op = decode->op[0]; | |
1298 | break; | |
1299 | ||
1300 | case X86_TYPE_LoBits: | |
1301 | op->n = (b & 7) | REX_B(s); | |
1302 | op->unit = X86_OP_INT; | |
1303 | break; | |
1304 | ||
1305 | case X86_TYPE_0 ... X86_TYPE_7: | |
1306 | op->n = type - X86_TYPE_0; | |
1307 | op->unit = X86_OP_INT; | |
1308 | break; | |
1309 | ||
1310 | case X86_TYPE_ES ... X86_TYPE_GS: | |
1311 | op->n = type - X86_TYPE_ES; | |
1312 | op->unit = X86_OP_SEG; | |
1313 | break; | |
1314 | } | |
1315 | ||
1316 | return true; | |
1317 | } | |
1318 | ||
55a33286 PB |
1319 | static bool validate_sse_prefix(DisasContext *s, X86OpEntry *e) |
1320 | { | |
1321 | uint16_t sse_prefixes; | |
1322 | ||
1323 | if (!e->valid_prefix) { | |
1324 | return true; | |
1325 | } | |
1326 | if (s->prefix & (PREFIX_REPZ | PREFIX_REPNZ)) { | |
1327 | /* In SSE instructions, 0xF3 and 0xF2 cancel 0x66. */ | |
1328 | s->prefix &= ~PREFIX_DATA; | |
1329 | } | |
1330 | ||
1331 | /* Now, either zero or one bit is set in sse_prefixes. */ | |
1332 | sse_prefixes = s->prefix & (PREFIX_REPZ | PREFIX_REPNZ | PREFIX_DATA); | |
1333 | return e->valid_prefix & (1 << sse_prefixes); | |
1334 | } | |
1335 | ||
b3e22b23 PB |
1336 | static bool decode_insn(DisasContext *s, CPUX86State *env, X86DecodeFunc decode_func, |
1337 | X86DecodedInsn *decode) | |
1338 | { | |
1339 | X86OpEntry *e = &decode->e; | |
1340 | ||
1341 | decode_func(s, env, e, &decode->b); | |
1342 | while (e->is_decode) { | |
1343 | e->is_decode = false; | |
1344 | e->decode(s, env, e, &decode->b); | |
1345 | } | |
1346 | ||
55a33286 PB |
1347 | if (!validate_sse_prefix(s, e)) { |
1348 | return false; | |
1349 | } | |
1350 | ||
b3e22b23 PB |
1351 | /* First compute size of operands in order to initialize s->rip_offset. */ |
1352 | if (e->op0 != X86_TYPE_None) { | |
1353 | if (!decode_op_size(s, e, e->s0, &decode->op[0].ot)) { | |
1354 | return false; | |
1355 | } | |
1356 | if (e->op0 == X86_TYPE_I) { | |
1357 | s->rip_offset += 1 << decode->op[0].ot; | |
1358 | } | |
1359 | } | |
1360 | if (e->op1 != X86_TYPE_None) { | |
1361 | if (!decode_op_size(s, e, e->s1, &decode->op[1].ot)) { | |
1362 | return false; | |
1363 | } | |
1364 | if (e->op1 == X86_TYPE_I) { | |
1365 | s->rip_offset += 1 << decode->op[1].ot; | |
1366 | } | |
1367 | } | |
1368 | if (e->op2 != X86_TYPE_None) { | |
1369 | if (!decode_op_size(s, e, e->s2, &decode->op[2].ot)) { | |
1370 | return false; | |
1371 | } | |
1372 | if (e->op2 == X86_TYPE_I) { | |
1373 | s->rip_offset += 1 << decode->op[2].ot; | |
1374 | } | |
1375 | } | |
1376 | if (e->op3 != X86_TYPE_None) { | |
79068477 PB |
1377 | /* |
1378 | * A couple instructions actually use the extra immediate byte for an Lx | |
1379 | * register operand; those are handled in the gen_* functions as one off. | |
1380 | */ | |
b3e22b23 PB |
1381 | assert(e->op3 == X86_TYPE_I && e->s3 == X86_SIZE_b); |
1382 | s->rip_offset += 1; | |
1383 | } | |
1384 | ||
1385 | if (e->op0 != X86_TYPE_None && | |
1386 | !decode_op(s, env, decode, &decode->op[0], e->op0, decode->b)) { | |
1387 | return false; | |
1388 | } | |
1389 | ||
1390 | if (e->op1 != X86_TYPE_None && | |
1391 | !decode_op(s, env, decode, &decode->op[1], e->op1, decode->b)) { | |
1392 | return false; | |
1393 | } | |
1394 | ||
1395 | if (e->op2 != X86_TYPE_None && | |
1396 | !decode_op(s, env, decode, &decode->op[2], e->op2, decode->b)) { | |
1397 | return false; | |
1398 | } | |
1399 | ||
1400 | if (e->op3 != X86_TYPE_None) { | |
1401 | decode->immediate = insn_get_signed(env, s, MO_8); | |
1402 | } | |
1403 | ||
1404 | return true; | |
1405 | } | |
1406 | ||
caa01fad PB |
1407 | static bool has_cpuid_feature(DisasContext *s, X86CPUIDFeature cpuid) |
1408 | { | |
1409 | switch (cpuid) { | |
1410 | case X86_FEAT_None: | |
1411 | return true; | |
cf5ec664 PB |
1412 | case X86_FEAT_F16C: |
1413 | return (s->cpuid_ext_features & CPUID_EXT_F16C); | |
2872b0f3 PB |
1414 | case X86_FEAT_FMA: |
1415 | return (s->cpuid_ext_features & CPUID_EXT_FMA); | |
caa01fad PB |
1416 | case X86_FEAT_MOVBE: |
1417 | return (s->cpuid_ext_features & CPUID_EXT_MOVBE); | |
1418 | case X86_FEAT_PCLMULQDQ: | |
1419 | return (s->cpuid_ext_features & CPUID_EXT_PCLMULQDQ); | |
1420 | case X86_FEAT_SSE: | |
1421 | return (s->cpuid_ext_features & CPUID_SSE); | |
1422 | case X86_FEAT_SSE2: | |
1423 | return (s->cpuid_ext_features & CPUID_SSE2); | |
1424 | case X86_FEAT_SSE3: | |
1425 | return (s->cpuid_ext_features & CPUID_EXT_SSE3); | |
1426 | case X86_FEAT_SSSE3: | |
1427 | return (s->cpuid_ext_features & CPUID_EXT_SSSE3); | |
1428 | case X86_FEAT_SSE41: | |
1429 | return (s->cpuid_ext_features & CPUID_EXT_SSE41); | |
1430 | case X86_FEAT_SSE42: | |
1431 | return (s->cpuid_ext_features & CPUID_EXT_SSE42); | |
1432 | case X86_FEAT_AES: | |
1433 | if (!(s->cpuid_ext_features & CPUID_EXT_AES)) { | |
1434 | return false; | |
1435 | } else if (!(s->prefix & PREFIX_VEX)) { | |
1436 | return true; | |
1437 | } else if (!(s->cpuid_ext_features & CPUID_EXT_AVX)) { | |
1438 | return false; | |
1439 | } else { | |
1440 | return !s->vex_l || (s->cpuid_7_0_ecx_features & CPUID_7_0_ECX_VAES); | |
1441 | } | |
1442 | ||
1443 | case X86_FEAT_AVX: | |
1444 | return (s->cpuid_ext_features & CPUID_EXT_AVX); | |
1445 | ||
71a0891d PB |
1446 | case X86_FEAT_3DNOW: |
1447 | return (s->cpuid_ext2_features & CPUID_EXT2_3DNOW); | |
caa01fad PB |
1448 | case X86_FEAT_SSE4A: |
1449 | return (s->cpuid_ext3_features & CPUID_EXT3_SSE4A); | |
1450 | ||
1451 | case X86_FEAT_ADX: | |
1452 | return (s->cpuid_7_0_ebx_features & CPUID_7_0_EBX_ADX); | |
1453 | case X86_FEAT_BMI1: | |
1454 | return (s->cpuid_7_0_ebx_features & CPUID_7_0_EBX_BMI1); | |
1455 | case X86_FEAT_BMI2: | |
1456 | return (s->cpuid_7_0_ebx_features & CPUID_7_0_EBX_BMI2); | |
1457 | case X86_FEAT_AVX2: | |
1458 | return (s->cpuid_7_0_ebx_features & CPUID_7_0_EBX_AVX2); | |
1459 | } | |
1460 | g_assert_not_reached(); | |
1461 | } | |
1462 | ||
20581aad PB |
1463 | static bool validate_vex(DisasContext *s, X86DecodedInsn *decode) |
1464 | { | |
1465 | X86OpEntry *e = &decode->e; | |
1466 | ||
1467 | switch (e->vex_special) { | |
1468 | case X86_VEX_REPScalar: | |
1469 | /* | |
1470 | * Instructions which differ between 00/66 and F2/F3 in the | |
1471 | * exception classification and the size of the memory operand. | |
1472 | */ | |
3d304620 | 1473 | assert(e->vex_class == 1 || e->vex_class == 2 || e->vex_class == 4); |
20581aad | 1474 | if (s->prefix & (PREFIX_REPZ | PREFIX_REPNZ)) { |
3d304620 | 1475 | e->vex_class = e->vex_class < 4 ? 3 : 5; |
20581aad PB |
1476 | if (s->vex_l) { |
1477 | goto illegal; | |
1478 | } | |
1479 | assert(decode->e.s2 == X86_SIZE_x); | |
1480 | if (decode->op[2].has_ea) { | |
1481 | decode->op[2].ot = s->prefix & PREFIX_REPZ ? MO_32 : MO_64; | |
1482 | } | |
1483 | } | |
1484 | break; | |
1485 | ||
1486 | case X86_VEX_SSEUnaligned: | |
1487 | /* handled in sse_needs_alignment. */ | |
1488 | break; | |
1489 | ||
1490 | case X86_VEX_AVX2_256: | |
1491 | if ((s->prefix & PREFIX_VEX) && s->vex_l && !has_cpuid_feature(s, X86_FEAT_AVX2)) { | |
1492 | goto illegal; | |
1493 | } | |
1494 | } | |
1495 | ||
1496 | /* TODO: instructions that require VEX.W=0 (Table 2-16) */ | |
1497 | ||
1498 | switch (e->vex_class) { | |
1499 | case 0: | |
1500 | if (s->prefix & PREFIX_VEX) { | |
1501 | goto illegal; | |
1502 | } | |
1503 | return true; | |
1504 | case 1: | |
1505 | case 2: | |
1506 | case 3: | |
1507 | case 4: | |
1508 | case 5: | |
1509 | case 7: | |
1510 | if (s->prefix & PREFIX_VEX) { | |
1511 | if (!(s->flags & HF_AVX_EN_MASK)) { | |
1512 | goto illegal; | |
1513 | } | |
38e65936 PB |
1514 | } else if (e->special != X86_SPECIAL_MMX || |
1515 | (s->prefix & (PREFIX_REPZ | PREFIX_REPNZ | PREFIX_DATA))) { | |
20581aad PB |
1516 | if (!(s->flags & HF_OSFXSR_MASK)) { |
1517 | goto illegal; | |
1518 | } | |
1519 | } | |
1520 | break; | |
1521 | case 12: | |
1522 | /* Must have a VSIB byte and no address prefix. */ | |
1523 | assert(s->has_modrm); | |
1524 | if ((s->modrm & 7) != 4 || s->aflag == MO_16) { | |
1525 | goto illegal; | |
1526 | } | |
1527 | ||
1528 | /* Check no overlap between registers. */ | |
1529 | if (!decode->op[0].has_ea && | |
1530 | (decode->op[0].n == decode->mem.index || decode->op[0].n == decode->op[1].n)) { | |
1531 | goto illegal; | |
1532 | } | |
1533 | assert(!decode->op[1].has_ea); | |
1534 | if (decode->op[1].n == decode->mem.index) { | |
1535 | goto illegal; | |
1536 | } | |
1537 | if (!decode->op[2].has_ea && | |
1538 | (decode->op[2].n == decode->mem.index || decode->op[2].n == decode->op[1].n)) { | |
1539 | goto illegal; | |
1540 | } | |
1541 | /* fall through */ | |
1542 | case 6: | |
1543 | case 11: | |
1544 | if (!(s->prefix & PREFIX_VEX)) { | |
1545 | goto illegal; | |
1546 | } | |
1547 | if (!(s->flags & HF_AVX_EN_MASK)) { | |
1548 | goto illegal; | |
1549 | } | |
1550 | break; | |
1551 | case 8: | |
ce4fcb94 PB |
1552 | /* Non-VEX case handled in decode_0F77. */ |
1553 | assert(s->prefix & PREFIX_VEX); | |
20581aad PB |
1554 | if (!(s->flags & HF_AVX_EN_MASK)) { |
1555 | goto illegal; | |
1556 | } | |
1557 | break; | |
1558 | case 13: | |
1559 | if (!(s->prefix & PREFIX_VEX)) { | |
1560 | goto illegal; | |
1561 | } | |
1562 | if (s->vex_l) { | |
1563 | goto illegal; | |
1564 | } | |
1565 | /* All integer instructions use VEX.vvvv, so exit. */ | |
1566 | return true; | |
1567 | } | |
1568 | ||
1569 | if (s->vex_v != 0 && | |
1570 | e->op0 != X86_TYPE_H && e->op0 != X86_TYPE_B && | |
1571 | e->op1 != X86_TYPE_H && e->op1 != X86_TYPE_B && | |
1572 | e->op2 != X86_TYPE_H && e->op2 != X86_TYPE_B) { | |
1573 | goto illegal; | |
1574 | } | |
1575 | ||
1576 | if (s->flags & HF_TS_MASK) { | |
1577 | goto nm_exception; | |
1578 | } | |
1579 | if (s->flags & HF_EM_MASK) { | |
1580 | goto illegal; | |
1581 | } | |
1582 | return true; | |
1583 | ||
1584 | nm_exception: | |
1585 | gen_NM_exception(s); | |
1586 | return false; | |
1587 | illegal: | |
1588 | gen_illegal_opcode(s); | |
1589 | return false; | |
1590 | } | |
1591 | ||
b3e22b23 PB |
1592 | /* |
1593 | * Convert one instruction. s->base.is_jmp is set if the translation must | |
1594 | * be stopped. | |
1595 | */ | |
1596 | static void disas_insn_new(DisasContext *s, CPUState *cpu, int b) | |
1597 | { | |
b77af26e | 1598 | CPUX86State *env = cpu_env(cpu); |
b3e22b23 PB |
1599 | bool first = true; |
1600 | X86DecodedInsn decode; | |
1601 | X86DecodeFunc decode_func = decode_root; | |
1602 | ||
b3e22b23 PB |
1603 | s->has_modrm = false; |
1604 | ||
1605 | next_byte: | |
1606 | if (first) { | |
1607 | first = false; | |
1608 | } else { | |
1609 | b = x86_ldub_code(env, s); | |
1610 | } | |
1611 | /* Collect prefixes. */ | |
1612 | switch (b) { | |
1613 | case 0xf3: | |
1614 | s->prefix |= PREFIX_REPZ; | |
1615 | s->prefix &= ~PREFIX_REPNZ; | |
1616 | goto next_byte; | |
1617 | case 0xf2: | |
1618 | s->prefix |= PREFIX_REPNZ; | |
1619 | s->prefix &= ~PREFIX_REPZ; | |
1620 | goto next_byte; | |
1621 | case 0xf0: | |
1622 | s->prefix |= PREFIX_LOCK; | |
1623 | goto next_byte; | |
1624 | case 0x2e: | |
1625 | s->override = R_CS; | |
1626 | goto next_byte; | |
1627 | case 0x36: | |
1628 | s->override = R_SS; | |
1629 | goto next_byte; | |
1630 | case 0x3e: | |
1631 | s->override = R_DS; | |
1632 | goto next_byte; | |
1633 | case 0x26: | |
1634 | s->override = R_ES; | |
1635 | goto next_byte; | |
1636 | case 0x64: | |
1637 | s->override = R_FS; | |
1638 | goto next_byte; | |
1639 | case 0x65: | |
1640 | s->override = R_GS; | |
1641 | goto next_byte; | |
1642 | case 0x66: | |
1643 | s->prefix |= PREFIX_DATA; | |
1644 | goto next_byte; | |
1645 | case 0x67: | |
1646 | s->prefix |= PREFIX_ADR; | |
1647 | goto next_byte; | |
1648 | #ifdef TARGET_X86_64 | |
1649 | case 0x40 ... 0x4f: | |
1650 | if (CODE64(s)) { | |
1651 | /* REX prefix */ | |
1652 | s->prefix |= PREFIX_REX; | |
1653 | s->vex_w = (b >> 3) & 1; | |
1654 | s->rex_r = (b & 0x4) << 1; | |
1655 | s->rex_x = (b & 0x2) << 2; | |
1656 | s->rex_b = (b & 0x1) << 3; | |
1657 | goto next_byte; | |
1658 | } | |
1659 | break; | |
1660 | #endif | |
1661 | case 0xc5: /* 2-byte VEX */ | |
1662 | case 0xc4: /* 3-byte VEX */ | |
1663 | /* | |
1664 | * VEX prefixes cannot be used except in 32-bit mode. | |
1665 | * Otherwise the instruction is LES or LDS. | |
1666 | */ | |
1667 | if (CODE32(s) && !VM86(s)) { | |
1668 | static const int pp_prefix[4] = { | |
1669 | 0, PREFIX_DATA, PREFIX_REPZ, PREFIX_REPNZ | |
1670 | }; | |
1671 | int vex3, vex2 = x86_ldub_code(env, s); | |
1672 | ||
1673 | if (!CODE64(s) && (vex2 & 0xc0) != 0xc0) { | |
1674 | /* | |
1675 | * 4.1.4.6: In 32-bit mode, bits [7:6] must be 11b, | |
1676 | * otherwise the instruction is LES or LDS. | |
1677 | */ | |
1678 | s->pc--; /* rewind the advance_pc() x86_ldub_code() did */ | |
1679 | break; | |
1680 | } | |
1681 | ||
1682 | /* 4.1.1-4.1.3: No preceding lock, 66, f2, f3, or rex prefixes. */ | |
1683 | if (s->prefix & (PREFIX_REPZ | PREFIX_REPNZ | |
1684 | | PREFIX_LOCK | PREFIX_DATA | PREFIX_REX)) { | |
1685 | goto illegal_op; | |
1686 | } | |
1687 | #ifdef TARGET_X86_64 | |
1688 | s->rex_r = (~vex2 >> 4) & 8; | |
1689 | #endif | |
1690 | if (b == 0xc5) { | |
1691 | /* 2-byte VEX prefix: RVVVVlpp, implied 0f leading opcode byte */ | |
1692 | vex3 = vex2; | |
1693 | decode_func = decode_0F; | |
1694 | } else { | |
1695 | /* 3-byte VEX prefix: RXBmmmmm wVVVVlpp */ | |
1696 | vex3 = x86_ldub_code(env, s); | |
1697 | #ifdef TARGET_X86_64 | |
1698 | s->rex_x = (~vex2 >> 3) & 8; | |
1699 | s->rex_b = (~vex2 >> 2) & 8; | |
1700 | #endif | |
1701 | s->vex_w = (vex3 >> 7) & 1; | |
1702 | switch (vex2 & 0x1f) { | |
1703 | case 0x01: /* Implied 0f leading opcode bytes. */ | |
1704 | decode_func = decode_0F; | |
1705 | break; | |
1706 | case 0x02: /* Implied 0f 38 leading opcode bytes. */ | |
1707 | decode_func = decode_0F38; | |
1708 | break; | |
1709 | case 0x03: /* Implied 0f 3a leading opcode bytes. */ | |
1710 | decode_func = decode_0F3A; | |
1711 | break; | |
1712 | default: /* Reserved for future use. */ | |
1713 | goto unknown_op; | |
1714 | } | |
1715 | } | |
1716 | s->vex_v = (~vex3 >> 3) & 0xf; | |
1717 | s->vex_l = (vex3 >> 2) & 1; | |
1718 | s->prefix |= pp_prefix[vex3 & 3] | PREFIX_VEX; | |
1719 | } | |
1720 | break; | |
1721 | default: | |
1722 | if (b >= 0x100) { | |
1723 | b -= 0x100; | |
1724 | decode_func = do_decode_0F; | |
1725 | } | |
1726 | break; | |
1727 | } | |
1728 | ||
1729 | /* Post-process prefixes. */ | |
1730 | if (CODE64(s)) { | |
1731 | /* | |
1732 | * In 64-bit mode, the default data size is 32-bit. Select 64-bit | |
1733 | * data with rex_w, and 16-bit data with 0x66; rex_w takes precedence | |
1734 | * over 0x66 if both are present. | |
1735 | */ | |
1736 | s->dflag = (REX_W(s) ? MO_64 : s->prefix & PREFIX_DATA ? MO_16 : MO_32); | |
1737 | /* In 64-bit mode, 0x67 selects 32-bit addressing. */ | |
1738 | s->aflag = (s->prefix & PREFIX_ADR ? MO_32 : MO_64); | |
1739 | } else { | |
1740 | /* In 16/32-bit mode, 0x66 selects the opposite data size. */ | |
1741 | if (CODE32(s) ^ ((s->prefix & PREFIX_DATA) != 0)) { | |
1742 | s->dflag = MO_32; | |
1743 | } else { | |
1744 | s->dflag = MO_16; | |
1745 | } | |
1746 | /* In 16/32-bit mode, 0x67 selects the opposite addressing. */ | |
1747 | if (CODE32(s) ^ ((s->prefix & PREFIX_ADR) != 0)) { | |
1748 | s->aflag = MO_32; | |
1749 | } else { | |
1750 | s->aflag = MO_16; | |
1751 | } | |
1752 | } | |
1753 | ||
1754 | memset(&decode, 0, sizeof(decode)); | |
1755 | decode.b = b; | |
1756 | if (!decode_insn(s, env, decode_func, &decode)) { | |
1757 | goto illegal_op; | |
1758 | } | |
1759 | if (!decode.e.gen) { | |
1760 | goto unknown_op; | |
1761 | } | |
1762 | ||
caa01fad PB |
1763 | if (!has_cpuid_feature(s, decode.e.cpuid)) { |
1764 | goto illegal_op; | |
1765 | } | |
1766 | ||
b3e22b23 PB |
1767 | switch (decode.e.special) { |
1768 | case X86_SPECIAL_None: | |
1769 | break; | |
1770 | ||
1771 | case X86_SPECIAL_Locked: | |
1772 | if (decode.op[0].has_ea) { | |
1773 | s->prefix |= PREFIX_LOCK; | |
1774 | } | |
1775 | break; | |
1776 | ||
1777 | case X86_SPECIAL_ProtMode: | |
1778 | if (!PE(s) || VM86(s)) { | |
1779 | goto illegal_op; | |
1780 | } | |
1781 | break; | |
1782 | ||
1783 | case X86_SPECIAL_i64: | |
1784 | if (CODE64(s)) { | |
1785 | goto illegal_op; | |
1786 | } | |
1787 | break; | |
1788 | case X86_SPECIAL_o64: | |
1789 | if (!CODE64(s)) { | |
1790 | goto illegal_op; | |
1791 | } | |
1792 | break; | |
1793 | ||
1794 | case X86_SPECIAL_ZExtOp0: | |
1795 | assert(decode.op[0].unit == X86_OP_INT); | |
1796 | if (!decode.op[0].has_ea) { | |
1797 | decode.op[0].ot = MO_32; | |
1798 | } | |
1799 | break; | |
1800 | ||
1801 | case X86_SPECIAL_ZExtOp2: | |
1802 | assert(decode.op[2].unit == X86_OP_INT); | |
1803 | if (!decode.op[2].has_ea) { | |
1804 | decode.op[2].ot = MO_32; | |
1805 | } | |
1806 | break; | |
1807 | ||
16fc5726 PB |
1808 | case X86_SPECIAL_AVXExtMov: |
1809 | if (!decode.op[2].has_ea) { | |
1810 | decode.op[2].ot = s->vex_l ? MO_256 : MO_128; | |
1811 | } else if (s->vex_l) { | |
1812 | decode.op[2].ot++; | |
1813 | } | |
1814 | break; | |
1815 | ||
b2ea6450 | 1816 | default: |
b3e22b23 PB |
1817 | break; |
1818 | } | |
1819 | ||
20581aad PB |
1820 | if (!validate_vex(s, &decode)) { |
1821 | return; | |
1822 | } | |
b2ea6450 MB |
1823 | if (decode.e.special == X86_SPECIAL_MMX && |
1824 | !(s->prefix & (PREFIX_REPZ | PREFIX_REPNZ | PREFIX_DATA))) { | |
ad75a51e | 1825 | gen_helper_enter_mmx(tcg_env); |
b2ea6450 MB |
1826 | } |
1827 | ||
b3e22b23 | 1828 | if (decode.op[0].has_ea || decode.op[1].has_ea || decode.op[2].has_ea) { |
20581aad | 1829 | gen_load_ea(s, &decode.mem, decode.e.vex_class == 12); |
b3e22b23 | 1830 | } |
6ba13999 PB |
1831 | if (s->prefix & PREFIX_LOCK) { |
1832 | if (decode.op[0].unit != X86_OP_INT || !decode.op[0].has_ea) { | |
1833 | goto illegal_op; | |
1834 | } | |
1835 | gen_load(s, &decode, 2, s->T1); | |
1836 | decode.e.gen(s, env, &decode); | |
1837 | } else { | |
1838 | if (decode.op[0].unit == X86_OP_MMX) { | |
1839 | compute_mmx_offset(&decode.op[0]); | |
1840 | } else if (decode.op[0].unit == X86_OP_SSE) { | |
1841 | compute_xmm_offset(&decode.op[0]); | |
1842 | } | |
1843 | gen_load(s, &decode, 1, s->T0); | |
1844 | gen_load(s, &decode, 2, s->T1); | |
1845 | decode.e.gen(s, env, &decode); | |
1846 | gen_writeback(s, &decode, 0, s->T0); | |
1847 | } | |
b3e22b23 PB |
1848 | return; |
1849 | illegal_op: | |
1850 | gen_illegal_opcode(s); | |
1851 | return; | |
1852 | unknown_op: | |
1853 | gen_unknown_opcode(env, s); | |
1854 | } |