]>
Commit | Line | Data |
---|---|---|
2c0262af FB |
1 | /* |
2 | * ARM translation | |
5fafdf24 | 3 | * |
2c0262af | 4 | * Copyright (c) 2003 Fabrice Bellard |
9ee6e8bb | 5 | * Copyright (c) 2005-2007 CodeSourcery |
18c9b560 | 6 | * Copyright (c) 2007 OpenedHand, Ltd. |
2c0262af FB |
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 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 | |
8167ee88 | 19 | * License along with this library; if not, see <http://www.gnu.org/licenses/>. |
2c0262af | 20 | */ |
74c21bd0 | 21 | #include "qemu/osdep.h" |
2c0262af FB |
22 | |
23 | #include "cpu.h" | |
ccd38087 | 24 | #include "internals.h" |
76cad711 | 25 | #include "disas/disas.h" |
63c91552 | 26 | #include "exec/exec-all.h" |
57fec1fe | 27 | #include "tcg-op.h" |
36a71934 | 28 | #include "tcg-op-gvec.h" |
1de7afc9 | 29 | #include "qemu/log.h" |
534df156 | 30 | #include "qemu/bitops.h" |
1d854765 | 31 | #include "arm_ldst.h" |
f1672e6f | 32 | #include "hw/semihosting/semihost.h" |
1497c961 | 33 | |
2ef6175a RH |
34 | #include "exec/helper-proto.h" |
35 | #include "exec/helper-gen.h" | |
2c0262af | 36 | |
a7e30d84 | 37 | #include "trace-tcg.h" |
508127e2 | 38 | #include "exec/log.h" |
a7e30d84 LV |
39 | |
40 | ||
2b51668f PM |
41 | #define ENABLE_ARCH_4T arm_dc_feature(s, ARM_FEATURE_V4T) |
42 | #define ENABLE_ARCH_5 arm_dc_feature(s, ARM_FEATURE_V5) | |
be5e7a76 | 43 | /* currently all emulated v5 cores are also v5TE, so don't bother */ |
2b51668f | 44 | #define ENABLE_ARCH_5TE arm_dc_feature(s, ARM_FEATURE_V5) |
09cbd501 | 45 | #define ENABLE_ARCH_5J dc_isar_feature(jazelle, s) |
2b51668f PM |
46 | #define ENABLE_ARCH_6 arm_dc_feature(s, ARM_FEATURE_V6) |
47 | #define ENABLE_ARCH_6K arm_dc_feature(s, ARM_FEATURE_V6K) | |
48 | #define ENABLE_ARCH_6T2 arm_dc_feature(s, ARM_FEATURE_THUMB2) | |
49 | #define ENABLE_ARCH_7 arm_dc_feature(s, ARM_FEATURE_V7) | |
50 | #define ENABLE_ARCH_8 arm_dc_feature(s, ARM_FEATURE_V8) | |
b5ff1b31 | 51 | |
86753403 | 52 | #define ARCH(x) do { if (!ENABLE_ARCH_##x) goto illegal_op; } while(0) |
b5ff1b31 | 53 | |
f570c61e | 54 | #include "translate.h" |
e12ce78d | 55 | |
b5ff1b31 FB |
56 | #if defined(CONFIG_USER_ONLY) |
57 | #define IS_USER(s) 1 | |
58 | #else | |
59 | #define IS_USER(s) (s->user) | |
60 | #endif | |
61 | ||
ad69471c | 62 | /* We reuse the same 64-bit temporaries for efficiency. */ |
a7812ae4 | 63 | static TCGv_i64 cpu_V0, cpu_V1, cpu_M0; |
155c3eac | 64 | static TCGv_i32 cpu_R[16]; |
78bcaa3e RH |
65 | TCGv_i32 cpu_CF, cpu_NF, cpu_VF, cpu_ZF; |
66 | TCGv_i64 cpu_exclusive_addr; | |
67 | TCGv_i64 cpu_exclusive_val; | |
ad69471c | 68 | |
022c62cb | 69 | #include "exec/gen-icount.h" |
2e70f6ef | 70 | |
308e5636 | 71 | static const char * const regnames[] = |
155c3eac FN |
72 | { "r0", "r1", "r2", "r3", "r4", "r5", "r6", "r7", |
73 | "r8", "r9", "r10", "r11", "r12", "r13", "r14", "pc" }; | |
74 | ||
61adacc8 RH |
75 | /* Function prototypes for gen_ functions calling Neon helpers. */ |
76 | typedef void NeonGenThreeOpEnvFn(TCGv_i32, TCGv_env, TCGv_i32, | |
77 | TCGv_i32, TCGv_i32); | |
c253dd78 PM |
78 | /* Function prototypes for gen_ functions for fix point conversions */ |
79 | typedef void VFPGenFixPointFn(TCGv_i32, TCGv_i32, TCGv_i32, TCGv_ptr); | |
61adacc8 | 80 | |
b26eefb6 PB |
81 | /* initialize TCG globals. */ |
82 | void arm_translate_init(void) | |
83 | { | |
155c3eac FN |
84 | int i; |
85 | ||
155c3eac | 86 | for (i = 0; i < 16; i++) { |
e1ccc054 | 87 | cpu_R[i] = tcg_global_mem_new_i32(cpu_env, |
0ecb72a5 | 88 | offsetof(CPUARMState, regs[i]), |
155c3eac FN |
89 | regnames[i]); |
90 | } | |
e1ccc054 RH |
91 | cpu_CF = tcg_global_mem_new_i32(cpu_env, offsetof(CPUARMState, CF), "CF"); |
92 | cpu_NF = tcg_global_mem_new_i32(cpu_env, offsetof(CPUARMState, NF), "NF"); | |
93 | cpu_VF = tcg_global_mem_new_i32(cpu_env, offsetof(CPUARMState, VF), "VF"); | |
94 | cpu_ZF = tcg_global_mem_new_i32(cpu_env, offsetof(CPUARMState, ZF), "ZF"); | |
66c374de | 95 | |
e1ccc054 | 96 | cpu_exclusive_addr = tcg_global_mem_new_i64(cpu_env, |
0ecb72a5 | 97 | offsetof(CPUARMState, exclusive_addr), "exclusive_addr"); |
e1ccc054 | 98 | cpu_exclusive_val = tcg_global_mem_new_i64(cpu_env, |
0ecb72a5 | 99 | offsetof(CPUARMState, exclusive_val), "exclusive_val"); |
155c3eac | 100 | |
14ade10f | 101 | a64_translate_init(); |
b26eefb6 PB |
102 | } |
103 | ||
9bb6558a PM |
104 | /* Flags for the disas_set_da_iss info argument: |
105 | * lower bits hold the Rt register number, higher bits are flags. | |
106 | */ | |
107 | typedef enum ISSInfo { | |
108 | ISSNone = 0, | |
109 | ISSRegMask = 0x1f, | |
110 | ISSInvalid = (1 << 5), | |
111 | ISSIsAcqRel = (1 << 6), | |
112 | ISSIsWrite = (1 << 7), | |
113 | ISSIs16Bit = (1 << 8), | |
114 | } ISSInfo; | |
115 | ||
116 | /* Save the syndrome information for a Data Abort */ | |
14776ab5 | 117 | static void disas_set_da_iss(DisasContext *s, MemOp memop, ISSInfo issinfo) |
9bb6558a PM |
118 | { |
119 | uint32_t syn; | |
120 | int sas = memop & MO_SIZE; | |
121 | bool sse = memop & MO_SIGN; | |
122 | bool is_acqrel = issinfo & ISSIsAcqRel; | |
123 | bool is_write = issinfo & ISSIsWrite; | |
124 | bool is_16bit = issinfo & ISSIs16Bit; | |
125 | int srt = issinfo & ISSRegMask; | |
126 | ||
127 | if (issinfo & ISSInvalid) { | |
128 | /* Some callsites want to conditionally provide ISS info, | |
129 | * eg "only if this was not a writeback" | |
130 | */ | |
131 | return; | |
132 | } | |
133 | ||
134 | if (srt == 15) { | |
135 | /* For AArch32, insns where the src/dest is R15 never generate | |
136 | * ISS information. Catching that here saves checking at all | |
137 | * the call sites. | |
138 | */ | |
139 | return; | |
140 | } | |
141 | ||
142 | syn = syn_data_abort_with_iss(0, sas, sse, srt, 0, is_acqrel, | |
143 | 0, 0, 0, is_write, 0, is_16bit); | |
144 | disas_set_insn_syndrome(s, syn); | |
145 | } | |
146 | ||
8bd5c820 | 147 | static inline int get_a32_user_mem_index(DisasContext *s) |
579d21cc | 148 | { |
8bd5c820 | 149 | /* Return the core mmu_idx to use for A32/T32 "unprivileged load/store" |
579d21cc PM |
150 | * insns: |
151 | * if PL2, UNPREDICTABLE (we choose to implement as if PL0) | |
152 | * otherwise, access as if at PL0. | |
153 | */ | |
154 | switch (s->mmu_idx) { | |
155 | case ARMMMUIdx_S1E2: /* this one is UNPREDICTABLE */ | |
156 | case ARMMMUIdx_S12NSE0: | |
157 | case ARMMMUIdx_S12NSE1: | |
8bd5c820 | 158 | return arm_to_core_mmu_idx(ARMMMUIdx_S12NSE0); |
579d21cc PM |
159 | case ARMMMUIdx_S1E3: |
160 | case ARMMMUIdx_S1SE0: | |
161 | case ARMMMUIdx_S1SE1: | |
8bd5c820 | 162 | return arm_to_core_mmu_idx(ARMMMUIdx_S1SE0); |
e7b921c2 PM |
163 | case ARMMMUIdx_MUser: |
164 | case ARMMMUIdx_MPriv: | |
165 | return arm_to_core_mmu_idx(ARMMMUIdx_MUser); | |
62593718 PM |
166 | case ARMMMUIdx_MUserNegPri: |
167 | case ARMMMUIdx_MPrivNegPri: | |
168 | return arm_to_core_mmu_idx(ARMMMUIdx_MUserNegPri); | |
b9f587d6 PM |
169 | case ARMMMUIdx_MSUser: |
170 | case ARMMMUIdx_MSPriv: | |
b9f587d6 | 171 | return arm_to_core_mmu_idx(ARMMMUIdx_MSUser); |
62593718 PM |
172 | case ARMMMUIdx_MSUserNegPri: |
173 | case ARMMMUIdx_MSPrivNegPri: | |
174 | return arm_to_core_mmu_idx(ARMMMUIdx_MSUserNegPri); | |
579d21cc PM |
175 | case ARMMMUIdx_S2NS: |
176 | default: | |
177 | g_assert_not_reached(); | |
178 | } | |
179 | } | |
180 | ||
39d5492a | 181 | static inline TCGv_i32 load_cpu_offset(int offset) |
d9ba4830 | 182 | { |
39d5492a | 183 | TCGv_i32 tmp = tcg_temp_new_i32(); |
d9ba4830 PB |
184 | tcg_gen_ld_i32(tmp, cpu_env, offset); |
185 | return tmp; | |
186 | } | |
187 | ||
0ecb72a5 | 188 | #define load_cpu_field(name) load_cpu_offset(offsetof(CPUARMState, name)) |
d9ba4830 | 189 | |
39d5492a | 190 | static inline void store_cpu_offset(TCGv_i32 var, int offset) |
d9ba4830 PB |
191 | { |
192 | tcg_gen_st_i32(var, cpu_env, offset); | |
7d1b0095 | 193 | tcg_temp_free_i32(var); |
d9ba4830 PB |
194 | } |
195 | ||
196 | #define store_cpu_field(var, name) \ | |
0ecb72a5 | 197 | store_cpu_offset(var, offsetof(CPUARMState, name)) |
d9ba4830 | 198 | |
fdbcf632 RH |
199 | /* The architectural value of PC. */ |
200 | static uint32_t read_pc(DisasContext *s) | |
201 | { | |
202 | return s->pc_curr + (s->thumb ? 4 : 8); | |
203 | } | |
204 | ||
b26eefb6 | 205 | /* Set a variable to the value of a CPU register. */ |
39d5492a | 206 | static void load_reg_var(DisasContext *s, TCGv_i32 var, int reg) |
b26eefb6 PB |
207 | { |
208 | if (reg == 15) { | |
fdbcf632 | 209 | tcg_gen_movi_i32(var, read_pc(s)); |
b26eefb6 | 210 | } else { |
155c3eac | 211 | tcg_gen_mov_i32(var, cpu_R[reg]); |
b26eefb6 PB |
212 | } |
213 | } | |
214 | ||
215 | /* Create a new temporary and set it to the value of a CPU register. */ | |
39d5492a | 216 | static inline TCGv_i32 load_reg(DisasContext *s, int reg) |
b26eefb6 | 217 | { |
39d5492a | 218 | TCGv_i32 tmp = tcg_temp_new_i32(); |
b26eefb6 PB |
219 | load_reg_var(s, tmp, reg); |
220 | return tmp; | |
221 | } | |
222 | ||
16e0d823 RH |
223 | /* |
224 | * Create a new temp, REG + OFS, except PC is ALIGN(PC, 4). | |
225 | * This is used for load/store for which use of PC implies (literal), | |
226 | * or ADD that implies ADR. | |
227 | */ | |
228 | static TCGv_i32 add_reg_for_lit(DisasContext *s, int reg, int ofs) | |
229 | { | |
230 | TCGv_i32 tmp = tcg_temp_new_i32(); | |
231 | ||
232 | if (reg == 15) { | |
233 | tcg_gen_movi_i32(tmp, (read_pc(s) & ~3) + ofs); | |
234 | } else { | |
235 | tcg_gen_addi_i32(tmp, cpu_R[reg], ofs); | |
236 | } | |
237 | return tmp; | |
238 | } | |
239 | ||
b26eefb6 PB |
240 | /* Set a CPU register. The source must be a temporary and will be |
241 | marked as dead. */ | |
39d5492a | 242 | static void store_reg(DisasContext *s, int reg, TCGv_i32 var) |
b26eefb6 PB |
243 | { |
244 | if (reg == 15) { | |
9b6a3ea7 PM |
245 | /* In Thumb mode, we must ignore bit 0. |
246 | * In ARM mode, for ARMv4 and ARMv5, it is UNPREDICTABLE if bits [1:0] | |
247 | * are not 0b00, but for ARMv6 and above, we must ignore bits [1:0]. | |
248 | * We choose to ignore [1:0] in ARM mode for all architecture versions. | |
249 | */ | |
250 | tcg_gen_andi_i32(var, var, s->thumb ? ~1 : ~3); | |
dcba3a8d | 251 | s->base.is_jmp = DISAS_JUMP; |
b26eefb6 | 252 | } |
155c3eac | 253 | tcg_gen_mov_i32(cpu_R[reg], var); |
7d1b0095 | 254 | tcg_temp_free_i32(var); |
b26eefb6 PB |
255 | } |
256 | ||
55203189 PM |
257 | /* |
258 | * Variant of store_reg which applies v8M stack-limit checks before updating | |
259 | * SP. If the check fails this will result in an exception being taken. | |
260 | * We disable the stack checks for CONFIG_USER_ONLY because we have | |
261 | * no idea what the stack limits should be in that case. | |
262 | * If stack checking is not being done this just acts like store_reg(). | |
263 | */ | |
264 | static void store_sp_checked(DisasContext *s, TCGv_i32 var) | |
265 | { | |
266 | #ifndef CONFIG_USER_ONLY | |
267 | if (s->v8m_stackcheck) { | |
268 | gen_helper_v8m_stackcheck(cpu_env, var); | |
269 | } | |
270 | #endif | |
271 | store_reg(s, 13, var); | |
272 | } | |
273 | ||
b26eefb6 | 274 | /* Value extensions. */ |
86831435 PB |
275 | #define gen_uxtb(var) tcg_gen_ext8u_i32(var, var) |
276 | #define gen_uxth(var) tcg_gen_ext16u_i32(var, var) | |
b26eefb6 PB |
277 | #define gen_sxtb(var) tcg_gen_ext8s_i32(var, var) |
278 | #define gen_sxth(var) tcg_gen_ext16s_i32(var, var) | |
279 | ||
1497c961 PB |
280 | #define gen_sxtb16(var) gen_helper_sxtb16(var, var) |
281 | #define gen_uxtb16(var) gen_helper_uxtb16(var, var) | |
8f01245e | 282 | |
b26eefb6 | 283 | |
39d5492a | 284 | static inline void gen_set_cpsr(TCGv_i32 var, uint32_t mask) |
b75263d6 | 285 | { |
39d5492a | 286 | TCGv_i32 tmp_mask = tcg_const_i32(mask); |
1ce94f81 | 287 | gen_helper_cpsr_write(cpu_env, var, tmp_mask); |
b75263d6 JR |
288 | tcg_temp_free_i32(tmp_mask); |
289 | } | |
d9ba4830 PB |
290 | /* Set NZCV flags from the high 4 bits of var. */ |
291 | #define gen_set_nzcv(var) gen_set_cpsr(var, CPSR_NZCV) | |
292 | ||
d4a2dc67 | 293 | static void gen_exception_internal(int excp) |
d9ba4830 | 294 | { |
d4a2dc67 PM |
295 | TCGv_i32 tcg_excp = tcg_const_i32(excp); |
296 | ||
297 | assert(excp_is_internal(excp)); | |
298 | gen_helper_exception_internal(cpu_env, tcg_excp); | |
299 | tcg_temp_free_i32(tcg_excp); | |
300 | } | |
301 | ||
50225ad0 PM |
302 | static void gen_step_complete_exception(DisasContext *s) |
303 | { | |
304 | /* We just completed step of an insn. Move from Active-not-pending | |
305 | * to Active-pending, and then also take the swstep exception. | |
306 | * This corresponds to making the (IMPDEF) choice to prioritize | |
307 | * swstep exceptions over asynchronous exceptions taken to an exception | |
308 | * level where debug is disabled. This choice has the advantage that | |
309 | * we do not need to maintain internal state corresponding to the | |
310 | * ISV/EX syndrome bits between completion of the step and generation | |
311 | * of the exception, and our syndrome information is always correct. | |
312 | */ | |
313 | gen_ss_advance(s); | |
c1d5f50f | 314 | gen_swstep_exception(s, 1, s->is_ldex); |
dcba3a8d | 315 | s->base.is_jmp = DISAS_NORETURN; |
50225ad0 PM |
316 | } |
317 | ||
5425415e PM |
318 | static void gen_singlestep_exception(DisasContext *s) |
319 | { | |
320 | /* Generate the right kind of exception for singlestep, which is | |
321 | * either the architectural singlestep or EXCP_DEBUG for QEMU's | |
322 | * gdb singlestepping. | |
323 | */ | |
324 | if (s->ss_active) { | |
325 | gen_step_complete_exception(s); | |
326 | } else { | |
327 | gen_exception_internal(EXCP_DEBUG); | |
328 | } | |
329 | } | |
330 | ||
b636649f PM |
331 | static inline bool is_singlestepping(DisasContext *s) |
332 | { | |
333 | /* Return true if we are singlestepping either because of | |
334 | * architectural singlestep or QEMU gdbstub singlestep. This does | |
335 | * not include the command line '-singlestep' mode which is rather | |
336 | * misnamed as it only means "one instruction per TB" and doesn't | |
337 | * affect the code we generate. | |
338 | */ | |
dcba3a8d | 339 | return s->base.singlestep_enabled || s->ss_active; |
b636649f PM |
340 | } |
341 | ||
39d5492a | 342 | static void gen_smul_dual(TCGv_i32 a, TCGv_i32 b) |
3670669c | 343 | { |
39d5492a PM |
344 | TCGv_i32 tmp1 = tcg_temp_new_i32(); |
345 | TCGv_i32 tmp2 = tcg_temp_new_i32(); | |
22478e79 AZ |
346 | tcg_gen_ext16s_i32(tmp1, a); |
347 | tcg_gen_ext16s_i32(tmp2, b); | |
3670669c | 348 | tcg_gen_mul_i32(tmp1, tmp1, tmp2); |
7d1b0095 | 349 | tcg_temp_free_i32(tmp2); |
3670669c PB |
350 | tcg_gen_sari_i32(a, a, 16); |
351 | tcg_gen_sari_i32(b, b, 16); | |
352 | tcg_gen_mul_i32(b, b, a); | |
353 | tcg_gen_mov_i32(a, tmp1); | |
7d1b0095 | 354 | tcg_temp_free_i32(tmp1); |
3670669c PB |
355 | } |
356 | ||
357 | /* Byteswap each halfword. */ | |
46497f6a | 358 | static void gen_rev16(TCGv_i32 dest, TCGv_i32 var) |
3670669c | 359 | { |
39d5492a | 360 | TCGv_i32 tmp = tcg_temp_new_i32(); |
68cedf73 | 361 | TCGv_i32 mask = tcg_const_i32(0x00ff00ff); |
3670669c | 362 | tcg_gen_shri_i32(tmp, var, 8); |
68cedf73 AJ |
363 | tcg_gen_and_i32(tmp, tmp, mask); |
364 | tcg_gen_and_i32(var, var, mask); | |
3670669c | 365 | tcg_gen_shli_i32(var, var, 8); |
46497f6a | 366 | tcg_gen_or_i32(dest, var, tmp); |
68cedf73 | 367 | tcg_temp_free_i32(mask); |
7d1b0095 | 368 | tcg_temp_free_i32(tmp); |
3670669c PB |
369 | } |
370 | ||
371 | /* Byteswap low halfword and sign extend. */ | |
46497f6a | 372 | static void gen_revsh(TCGv_i32 dest, TCGv_i32 var) |
3670669c | 373 | { |
1a855029 AJ |
374 | tcg_gen_ext16u_i32(var, var); |
375 | tcg_gen_bswap16_i32(var, var); | |
46497f6a | 376 | tcg_gen_ext16s_i32(dest, var); |
3670669c PB |
377 | } |
378 | ||
5e3f878a | 379 | /* 32x32->64 multiply. Marks inputs as dead. */ |
39d5492a | 380 | static TCGv_i64 gen_mulu_i64_i32(TCGv_i32 a, TCGv_i32 b) |
5e3f878a | 381 | { |
39d5492a PM |
382 | TCGv_i32 lo = tcg_temp_new_i32(); |
383 | TCGv_i32 hi = tcg_temp_new_i32(); | |
831d7fe8 | 384 | TCGv_i64 ret; |
5e3f878a | 385 | |
831d7fe8 | 386 | tcg_gen_mulu2_i32(lo, hi, a, b); |
7d1b0095 | 387 | tcg_temp_free_i32(a); |
7d1b0095 | 388 | tcg_temp_free_i32(b); |
831d7fe8 RH |
389 | |
390 | ret = tcg_temp_new_i64(); | |
391 | tcg_gen_concat_i32_i64(ret, lo, hi); | |
39d5492a PM |
392 | tcg_temp_free_i32(lo); |
393 | tcg_temp_free_i32(hi); | |
831d7fe8 RH |
394 | |
395 | return ret; | |
5e3f878a PB |
396 | } |
397 | ||
39d5492a | 398 | static TCGv_i64 gen_muls_i64_i32(TCGv_i32 a, TCGv_i32 b) |
5e3f878a | 399 | { |
39d5492a PM |
400 | TCGv_i32 lo = tcg_temp_new_i32(); |
401 | TCGv_i32 hi = tcg_temp_new_i32(); | |
831d7fe8 | 402 | TCGv_i64 ret; |
5e3f878a | 403 | |
831d7fe8 | 404 | tcg_gen_muls2_i32(lo, hi, a, b); |
7d1b0095 | 405 | tcg_temp_free_i32(a); |
7d1b0095 | 406 | tcg_temp_free_i32(b); |
831d7fe8 RH |
407 | |
408 | ret = tcg_temp_new_i64(); | |
409 | tcg_gen_concat_i32_i64(ret, lo, hi); | |
39d5492a PM |
410 | tcg_temp_free_i32(lo); |
411 | tcg_temp_free_i32(hi); | |
831d7fe8 RH |
412 | |
413 | return ret; | |
5e3f878a PB |
414 | } |
415 | ||
8f01245e | 416 | /* Swap low and high halfwords. */ |
39d5492a | 417 | static void gen_swap_half(TCGv_i32 var) |
8f01245e | 418 | { |
adefba76 | 419 | tcg_gen_rotri_i32(var, var, 16); |
8f01245e PB |
420 | } |
421 | ||
b26eefb6 PB |
422 | /* Dual 16-bit add. Result placed in t0 and t1 is marked as dead. |
423 | tmp = (t0 ^ t1) & 0x8000; | |
424 | t0 &= ~0x8000; | |
425 | t1 &= ~0x8000; | |
426 | t0 = (t0 + t1) ^ tmp; | |
427 | */ | |
428 | ||
46497f6a | 429 | static void gen_add16(TCGv_i32 dest, TCGv_i32 t0, TCGv_i32 t1) |
b26eefb6 | 430 | { |
39d5492a | 431 | TCGv_i32 tmp = tcg_temp_new_i32(); |
b26eefb6 PB |
432 | tcg_gen_xor_i32(tmp, t0, t1); |
433 | tcg_gen_andi_i32(tmp, tmp, 0x8000); | |
434 | tcg_gen_andi_i32(t0, t0, ~0x8000); | |
435 | tcg_gen_andi_i32(t1, t1, ~0x8000); | |
436 | tcg_gen_add_i32(t0, t0, t1); | |
46497f6a | 437 | tcg_gen_xor_i32(dest, t0, tmp); |
7d1b0095 | 438 | tcg_temp_free_i32(tmp); |
b26eefb6 PB |
439 | } |
440 | ||
b26eefb6 | 441 | /* Set N and Z flags from var. */ |
39d5492a | 442 | static inline void gen_logic_CC(TCGv_i32 var) |
b26eefb6 | 443 | { |
66c374de AJ |
444 | tcg_gen_mov_i32(cpu_NF, var); |
445 | tcg_gen_mov_i32(cpu_ZF, var); | |
b26eefb6 PB |
446 | } |
447 | ||
e9bb4aa9 | 448 | /* dest = T0 + T1 + CF. */ |
39d5492a | 449 | static void gen_add_carry(TCGv_i32 dest, TCGv_i32 t0, TCGv_i32 t1) |
e9bb4aa9 | 450 | { |
e9bb4aa9 | 451 | tcg_gen_add_i32(dest, t0, t1); |
66c374de | 452 | tcg_gen_add_i32(dest, dest, cpu_CF); |
e9bb4aa9 JR |
453 | } |
454 | ||
3670669c | 455 | /* dest = T0 - T1 + CF - 1. */ |
39d5492a | 456 | static void gen_sub_carry(TCGv_i32 dest, TCGv_i32 t0, TCGv_i32 t1) |
3670669c | 457 | { |
3670669c | 458 | tcg_gen_sub_i32(dest, t0, t1); |
66c374de | 459 | tcg_gen_add_i32(dest, dest, cpu_CF); |
3670669c | 460 | tcg_gen_subi_i32(dest, dest, 1); |
3670669c PB |
461 | } |
462 | ||
72485ec4 | 463 | /* dest = T0 + T1. Compute C, N, V and Z flags */ |
39d5492a | 464 | static void gen_add_CC(TCGv_i32 dest, TCGv_i32 t0, TCGv_i32 t1) |
72485ec4 | 465 | { |
39d5492a | 466 | TCGv_i32 tmp = tcg_temp_new_i32(); |
e3482cb8 RH |
467 | tcg_gen_movi_i32(tmp, 0); |
468 | tcg_gen_add2_i32(cpu_NF, cpu_CF, t0, tmp, t1, tmp); | |
72485ec4 | 469 | tcg_gen_mov_i32(cpu_ZF, cpu_NF); |
72485ec4 | 470 | tcg_gen_xor_i32(cpu_VF, cpu_NF, t0); |
72485ec4 AJ |
471 | tcg_gen_xor_i32(tmp, t0, t1); |
472 | tcg_gen_andc_i32(cpu_VF, cpu_VF, tmp); | |
473 | tcg_temp_free_i32(tmp); | |
474 | tcg_gen_mov_i32(dest, cpu_NF); | |
475 | } | |
476 | ||
49b4c31e | 477 | /* dest = T0 + T1 + CF. Compute C, N, V and Z flags */ |
39d5492a | 478 | static void gen_adc_CC(TCGv_i32 dest, TCGv_i32 t0, TCGv_i32 t1) |
49b4c31e | 479 | { |
39d5492a | 480 | TCGv_i32 tmp = tcg_temp_new_i32(); |
49b4c31e RH |
481 | if (TCG_TARGET_HAS_add2_i32) { |
482 | tcg_gen_movi_i32(tmp, 0); | |
483 | tcg_gen_add2_i32(cpu_NF, cpu_CF, t0, tmp, cpu_CF, tmp); | |
8c3ac601 | 484 | tcg_gen_add2_i32(cpu_NF, cpu_CF, cpu_NF, cpu_CF, t1, tmp); |
49b4c31e RH |
485 | } else { |
486 | TCGv_i64 q0 = tcg_temp_new_i64(); | |
487 | TCGv_i64 q1 = tcg_temp_new_i64(); | |
488 | tcg_gen_extu_i32_i64(q0, t0); | |
489 | tcg_gen_extu_i32_i64(q1, t1); | |
490 | tcg_gen_add_i64(q0, q0, q1); | |
491 | tcg_gen_extu_i32_i64(q1, cpu_CF); | |
492 | tcg_gen_add_i64(q0, q0, q1); | |
493 | tcg_gen_extr_i64_i32(cpu_NF, cpu_CF, q0); | |
494 | tcg_temp_free_i64(q0); | |
495 | tcg_temp_free_i64(q1); | |
496 | } | |
497 | tcg_gen_mov_i32(cpu_ZF, cpu_NF); | |
498 | tcg_gen_xor_i32(cpu_VF, cpu_NF, t0); | |
499 | tcg_gen_xor_i32(tmp, t0, t1); | |
500 | tcg_gen_andc_i32(cpu_VF, cpu_VF, tmp); | |
501 | tcg_temp_free_i32(tmp); | |
502 | tcg_gen_mov_i32(dest, cpu_NF); | |
503 | } | |
504 | ||
72485ec4 | 505 | /* dest = T0 - T1. Compute C, N, V and Z flags */ |
39d5492a | 506 | static void gen_sub_CC(TCGv_i32 dest, TCGv_i32 t0, TCGv_i32 t1) |
72485ec4 | 507 | { |
39d5492a | 508 | TCGv_i32 tmp; |
72485ec4 AJ |
509 | tcg_gen_sub_i32(cpu_NF, t0, t1); |
510 | tcg_gen_mov_i32(cpu_ZF, cpu_NF); | |
511 | tcg_gen_setcond_i32(TCG_COND_GEU, cpu_CF, t0, t1); | |
512 | tcg_gen_xor_i32(cpu_VF, cpu_NF, t0); | |
513 | tmp = tcg_temp_new_i32(); | |
514 | tcg_gen_xor_i32(tmp, t0, t1); | |
515 | tcg_gen_and_i32(cpu_VF, cpu_VF, tmp); | |
516 | tcg_temp_free_i32(tmp); | |
517 | tcg_gen_mov_i32(dest, cpu_NF); | |
518 | } | |
519 | ||
e77f0832 | 520 | /* dest = T0 + ~T1 + CF. Compute C, N, V and Z flags */ |
39d5492a | 521 | static void gen_sbc_CC(TCGv_i32 dest, TCGv_i32 t0, TCGv_i32 t1) |
2de68a49 | 522 | { |
39d5492a | 523 | TCGv_i32 tmp = tcg_temp_new_i32(); |
e77f0832 RH |
524 | tcg_gen_not_i32(tmp, t1); |
525 | gen_adc_CC(dest, t0, tmp); | |
39d5492a | 526 | tcg_temp_free_i32(tmp); |
2de68a49 RH |
527 | } |
528 | ||
365af80e | 529 | #define GEN_SHIFT(name) \ |
39d5492a | 530 | static void gen_##name(TCGv_i32 dest, TCGv_i32 t0, TCGv_i32 t1) \ |
365af80e | 531 | { \ |
39d5492a | 532 | TCGv_i32 tmp1, tmp2, tmp3; \ |
365af80e AJ |
533 | tmp1 = tcg_temp_new_i32(); \ |
534 | tcg_gen_andi_i32(tmp1, t1, 0xff); \ | |
535 | tmp2 = tcg_const_i32(0); \ | |
536 | tmp3 = tcg_const_i32(0x1f); \ | |
537 | tcg_gen_movcond_i32(TCG_COND_GTU, tmp2, tmp1, tmp3, tmp2, t0); \ | |
538 | tcg_temp_free_i32(tmp3); \ | |
539 | tcg_gen_andi_i32(tmp1, tmp1, 0x1f); \ | |
540 | tcg_gen_##name##_i32(dest, tmp2, tmp1); \ | |
541 | tcg_temp_free_i32(tmp2); \ | |
542 | tcg_temp_free_i32(tmp1); \ | |
543 | } | |
544 | GEN_SHIFT(shl) | |
545 | GEN_SHIFT(shr) | |
546 | #undef GEN_SHIFT | |
547 | ||
39d5492a | 548 | static void gen_sar(TCGv_i32 dest, TCGv_i32 t0, TCGv_i32 t1) |
365af80e | 549 | { |
39d5492a | 550 | TCGv_i32 tmp1, tmp2; |
365af80e AJ |
551 | tmp1 = tcg_temp_new_i32(); |
552 | tcg_gen_andi_i32(tmp1, t1, 0xff); | |
553 | tmp2 = tcg_const_i32(0x1f); | |
554 | tcg_gen_movcond_i32(TCG_COND_GTU, tmp1, tmp1, tmp2, tmp2, tmp1); | |
555 | tcg_temp_free_i32(tmp2); | |
556 | tcg_gen_sar_i32(dest, t0, tmp1); | |
557 | tcg_temp_free_i32(tmp1); | |
558 | } | |
559 | ||
39d5492a | 560 | static void shifter_out_im(TCGv_i32 var, int shift) |
b26eefb6 | 561 | { |
191f4bfe | 562 | tcg_gen_extract_i32(cpu_CF, var, shift, 1); |
9a119ff6 | 563 | } |
b26eefb6 | 564 | |
9a119ff6 | 565 | /* Shift by immediate. Includes special handling for shift == 0. */ |
39d5492a PM |
566 | static inline void gen_arm_shift_im(TCGv_i32 var, int shiftop, |
567 | int shift, int flags) | |
9a119ff6 PB |
568 | { |
569 | switch (shiftop) { | |
570 | case 0: /* LSL */ | |
571 | if (shift != 0) { | |
572 | if (flags) | |
573 | shifter_out_im(var, 32 - shift); | |
574 | tcg_gen_shli_i32(var, var, shift); | |
575 | } | |
576 | break; | |
577 | case 1: /* LSR */ | |
578 | if (shift == 0) { | |
579 | if (flags) { | |
66c374de | 580 | tcg_gen_shri_i32(cpu_CF, var, 31); |
9a119ff6 PB |
581 | } |
582 | tcg_gen_movi_i32(var, 0); | |
583 | } else { | |
584 | if (flags) | |
585 | shifter_out_im(var, shift - 1); | |
586 | tcg_gen_shri_i32(var, var, shift); | |
587 | } | |
588 | break; | |
589 | case 2: /* ASR */ | |
590 | if (shift == 0) | |
591 | shift = 32; | |
592 | if (flags) | |
593 | shifter_out_im(var, shift - 1); | |
594 | if (shift == 32) | |
595 | shift = 31; | |
596 | tcg_gen_sari_i32(var, var, shift); | |
597 | break; | |
598 | case 3: /* ROR/RRX */ | |
599 | if (shift != 0) { | |
600 | if (flags) | |
601 | shifter_out_im(var, shift - 1); | |
f669df27 | 602 | tcg_gen_rotri_i32(var, var, shift); break; |
9a119ff6 | 603 | } else { |
39d5492a | 604 | TCGv_i32 tmp = tcg_temp_new_i32(); |
b6348f29 | 605 | tcg_gen_shli_i32(tmp, cpu_CF, 31); |
9a119ff6 PB |
606 | if (flags) |
607 | shifter_out_im(var, 0); | |
608 | tcg_gen_shri_i32(var, var, 1); | |
b26eefb6 | 609 | tcg_gen_or_i32(var, var, tmp); |
7d1b0095 | 610 | tcg_temp_free_i32(tmp); |
b26eefb6 PB |
611 | } |
612 | } | |
613 | }; | |
614 | ||
39d5492a PM |
615 | static inline void gen_arm_shift_reg(TCGv_i32 var, int shiftop, |
616 | TCGv_i32 shift, int flags) | |
8984bd2e PB |
617 | { |
618 | if (flags) { | |
619 | switch (shiftop) { | |
9ef39277 BS |
620 | case 0: gen_helper_shl_cc(var, cpu_env, var, shift); break; |
621 | case 1: gen_helper_shr_cc(var, cpu_env, var, shift); break; | |
622 | case 2: gen_helper_sar_cc(var, cpu_env, var, shift); break; | |
623 | case 3: gen_helper_ror_cc(var, cpu_env, var, shift); break; | |
8984bd2e PB |
624 | } |
625 | } else { | |
626 | switch (shiftop) { | |
365af80e AJ |
627 | case 0: |
628 | gen_shl(var, var, shift); | |
629 | break; | |
630 | case 1: | |
631 | gen_shr(var, var, shift); | |
632 | break; | |
633 | case 2: | |
634 | gen_sar(var, var, shift); | |
635 | break; | |
f669df27 AJ |
636 | case 3: tcg_gen_andi_i32(shift, shift, 0x1f); |
637 | tcg_gen_rotr_i32(var, var, shift); break; | |
8984bd2e PB |
638 | } |
639 | } | |
7d1b0095 | 640 | tcg_temp_free_i32(shift); |
8984bd2e PB |
641 | } |
642 | ||
39fb730a | 643 | /* |
6c2c63d3 | 644 | * Generate a conditional based on ARM condition code cc. |
39fb730a AG |
645 | * This is common between ARM and Aarch64 targets. |
646 | */ | |
6c2c63d3 | 647 | void arm_test_cc(DisasCompare *cmp, int cc) |
d9ba4830 | 648 | { |
6c2c63d3 RH |
649 | TCGv_i32 value; |
650 | TCGCond cond; | |
651 | bool global = true; | |
d9ba4830 | 652 | |
d9ba4830 PB |
653 | switch (cc) { |
654 | case 0: /* eq: Z */ | |
d9ba4830 | 655 | case 1: /* ne: !Z */ |
6c2c63d3 RH |
656 | cond = TCG_COND_EQ; |
657 | value = cpu_ZF; | |
d9ba4830 | 658 | break; |
6c2c63d3 | 659 | |
d9ba4830 | 660 | case 2: /* cs: C */ |
d9ba4830 | 661 | case 3: /* cc: !C */ |
6c2c63d3 RH |
662 | cond = TCG_COND_NE; |
663 | value = cpu_CF; | |
d9ba4830 | 664 | break; |
6c2c63d3 | 665 | |
d9ba4830 | 666 | case 4: /* mi: N */ |
d9ba4830 | 667 | case 5: /* pl: !N */ |
6c2c63d3 RH |
668 | cond = TCG_COND_LT; |
669 | value = cpu_NF; | |
d9ba4830 | 670 | break; |
6c2c63d3 | 671 | |
d9ba4830 | 672 | case 6: /* vs: V */ |
d9ba4830 | 673 | case 7: /* vc: !V */ |
6c2c63d3 RH |
674 | cond = TCG_COND_LT; |
675 | value = cpu_VF; | |
d9ba4830 | 676 | break; |
6c2c63d3 | 677 | |
d9ba4830 | 678 | case 8: /* hi: C && !Z */ |
6c2c63d3 RH |
679 | case 9: /* ls: !C || Z -> !(C && !Z) */ |
680 | cond = TCG_COND_NE; | |
681 | value = tcg_temp_new_i32(); | |
682 | global = false; | |
683 | /* CF is 1 for C, so -CF is an all-bits-set mask for C; | |
684 | ZF is non-zero for !Z; so AND the two subexpressions. */ | |
685 | tcg_gen_neg_i32(value, cpu_CF); | |
686 | tcg_gen_and_i32(value, value, cpu_ZF); | |
d9ba4830 | 687 | break; |
6c2c63d3 | 688 | |
d9ba4830 | 689 | case 10: /* ge: N == V -> N ^ V == 0 */ |
d9ba4830 | 690 | case 11: /* lt: N != V -> N ^ V != 0 */ |
6c2c63d3 RH |
691 | /* Since we're only interested in the sign bit, == 0 is >= 0. */ |
692 | cond = TCG_COND_GE; | |
693 | value = tcg_temp_new_i32(); | |
694 | global = false; | |
695 | tcg_gen_xor_i32(value, cpu_VF, cpu_NF); | |
d9ba4830 | 696 | break; |
6c2c63d3 | 697 | |
d9ba4830 | 698 | case 12: /* gt: !Z && N == V */ |
d9ba4830 | 699 | case 13: /* le: Z || N != V */ |
6c2c63d3 RH |
700 | cond = TCG_COND_NE; |
701 | value = tcg_temp_new_i32(); | |
702 | global = false; | |
703 | /* (N == V) is equal to the sign bit of ~(NF ^ VF). Propagate | |
704 | * the sign bit then AND with ZF to yield the result. */ | |
705 | tcg_gen_xor_i32(value, cpu_VF, cpu_NF); | |
706 | tcg_gen_sari_i32(value, value, 31); | |
707 | tcg_gen_andc_i32(value, cpu_ZF, value); | |
d9ba4830 | 708 | break; |
6c2c63d3 | 709 | |
9305eac0 RH |
710 | case 14: /* always */ |
711 | case 15: /* always */ | |
712 | /* Use the ALWAYS condition, which will fold early. | |
713 | * It doesn't matter what we use for the value. */ | |
714 | cond = TCG_COND_ALWAYS; | |
715 | value = cpu_ZF; | |
716 | goto no_invert; | |
717 | ||
d9ba4830 PB |
718 | default: |
719 | fprintf(stderr, "Bad condition code 0x%x\n", cc); | |
720 | abort(); | |
721 | } | |
6c2c63d3 RH |
722 | |
723 | if (cc & 1) { | |
724 | cond = tcg_invert_cond(cond); | |
725 | } | |
726 | ||
9305eac0 | 727 | no_invert: |
6c2c63d3 RH |
728 | cmp->cond = cond; |
729 | cmp->value = value; | |
730 | cmp->value_global = global; | |
731 | } | |
732 | ||
733 | void arm_free_cc(DisasCompare *cmp) | |
734 | { | |
735 | if (!cmp->value_global) { | |
736 | tcg_temp_free_i32(cmp->value); | |
737 | } | |
738 | } | |
739 | ||
740 | void arm_jump_cc(DisasCompare *cmp, TCGLabel *label) | |
741 | { | |
742 | tcg_gen_brcondi_i32(cmp->cond, cmp->value, 0, label); | |
743 | } | |
744 | ||
745 | void arm_gen_test_cc(int cc, TCGLabel *label) | |
746 | { | |
747 | DisasCompare cmp; | |
748 | arm_test_cc(&cmp, cc); | |
749 | arm_jump_cc(&cmp, label); | |
750 | arm_free_cc(&cmp); | |
d9ba4830 | 751 | } |
2c0262af | 752 | |
4d5e8c96 PM |
753 | static inline void gen_set_condexec(DisasContext *s) |
754 | { | |
755 | if (s->condexec_mask) { | |
756 | uint32_t val = (s->condexec_cond << 4) | (s->condexec_mask >> 1); | |
757 | TCGv_i32 tmp = tcg_temp_new_i32(); | |
758 | tcg_gen_movi_i32(tmp, val); | |
759 | store_cpu_field(tmp, condexec_bits); | |
760 | } | |
761 | } | |
762 | ||
763 | static inline void gen_set_pc_im(DisasContext *s, target_ulong val) | |
764 | { | |
765 | tcg_gen_movi_i32(cpu_R[15], val); | |
766 | } | |
767 | ||
d9ba4830 | 768 | /* Set PC and Thumb state from var. var is marked as dead. */ |
39d5492a | 769 | static inline void gen_bx(DisasContext *s, TCGv_i32 var) |
d9ba4830 | 770 | { |
dcba3a8d | 771 | s->base.is_jmp = DISAS_JUMP; |
155c3eac FN |
772 | tcg_gen_andi_i32(cpu_R[15], var, ~1); |
773 | tcg_gen_andi_i32(var, var, 1); | |
774 | store_cpu_field(var, thumb); | |
d9ba4830 PB |
775 | } |
776 | ||
5e5584c8 PM |
777 | /* |
778 | * Set PC and Thumb state from var. var is marked as dead. | |
3bb8a96f PM |
779 | * For M-profile CPUs, include logic to detect exception-return |
780 | * branches and handle them. This is needed for Thumb POP/LDM to PC, LDR to PC, | |
781 | * and BX reg, and no others, and happens only for code in Handler mode. | |
5e5584c8 PM |
782 | * The Security Extension also requires us to check for the FNC_RETURN |
783 | * which signals a function return from non-secure state; this can happen | |
784 | * in both Handler and Thread mode. | |
785 | * To avoid having to do multiple comparisons in inline generated code, | |
786 | * we make the check we do here loose, so it will match for EXC_RETURN | |
787 | * in Thread mode. For system emulation do_v7m_exception_exit() checks | |
788 | * for these spurious cases and returns without doing anything (giving | |
789 | * the same behaviour as for a branch to a non-magic address). | |
790 | * | |
791 | * In linux-user mode it is unclear what the right behaviour for an | |
792 | * attempted FNC_RETURN should be, because in real hardware this will go | |
793 | * directly to Secure code (ie not the Linux kernel) which will then treat | |
794 | * the error in any way it chooses. For QEMU we opt to make the FNC_RETURN | |
795 | * attempt behave the way it would on a CPU without the security extension, | |
796 | * which is to say "like a normal branch". That means we can simply treat | |
797 | * all branches as normal with no magic address behaviour. | |
3bb8a96f PM |
798 | */ |
799 | static inline void gen_bx_excret(DisasContext *s, TCGv_i32 var) | |
800 | { | |
801 | /* Generate the same code here as for a simple bx, but flag via | |
dcba3a8d | 802 | * s->base.is_jmp that we need to do the rest of the work later. |
3bb8a96f PM |
803 | */ |
804 | gen_bx(s, var); | |
5e5584c8 | 805 | #ifndef CONFIG_USER_ONLY |
d02a8698 PM |
806 | if (arm_dc_feature(s, ARM_FEATURE_M_SECURITY) || |
807 | (s->v7m_handler_mode && arm_dc_feature(s, ARM_FEATURE_M))) { | |
dcba3a8d | 808 | s->base.is_jmp = DISAS_BX_EXCRET; |
3bb8a96f | 809 | } |
5e5584c8 | 810 | #endif |
3bb8a96f PM |
811 | } |
812 | ||
813 | static inline void gen_bx_excret_final_code(DisasContext *s) | |
814 | { | |
815 | /* Generate the code to finish possible exception return and end the TB */ | |
816 | TCGLabel *excret_label = gen_new_label(); | |
d02a8698 PM |
817 | uint32_t min_magic; |
818 | ||
819 | if (arm_dc_feature(s, ARM_FEATURE_M_SECURITY)) { | |
820 | /* Covers FNC_RETURN and EXC_RETURN magic */ | |
821 | min_magic = FNC_RETURN_MIN_MAGIC; | |
822 | } else { | |
823 | /* EXC_RETURN magic only */ | |
824 | min_magic = EXC_RETURN_MIN_MAGIC; | |
825 | } | |
3bb8a96f PM |
826 | |
827 | /* Is the new PC value in the magic range indicating exception return? */ | |
d02a8698 | 828 | tcg_gen_brcondi_i32(TCG_COND_GEU, cpu_R[15], min_magic, excret_label); |
3bb8a96f PM |
829 | /* No: end the TB as we would for a DISAS_JMP */ |
830 | if (is_singlestepping(s)) { | |
831 | gen_singlestep_exception(s); | |
832 | } else { | |
07ea28b4 | 833 | tcg_gen_exit_tb(NULL, 0); |
3bb8a96f PM |
834 | } |
835 | gen_set_label(excret_label); | |
836 | /* Yes: this is an exception return. | |
837 | * At this point in runtime env->regs[15] and env->thumb will hold | |
838 | * the exception-return magic number, which do_v7m_exception_exit() | |
839 | * will read. Nothing else will be able to see those values because | |
840 | * the cpu-exec main loop guarantees that we will always go straight | |
841 | * from raising the exception to the exception-handling code. | |
842 | * | |
843 | * gen_ss_advance(s) does nothing on M profile currently but | |
844 | * calling it is conceptually the right thing as we have executed | |
845 | * this instruction (compare SWI, HVC, SMC handling). | |
846 | */ | |
847 | gen_ss_advance(s); | |
848 | gen_exception_internal(EXCP_EXCEPTION_EXIT); | |
849 | } | |
850 | ||
fb602cb7 PM |
851 | static inline void gen_bxns(DisasContext *s, int rm) |
852 | { | |
853 | TCGv_i32 var = load_reg(s, rm); | |
854 | ||
855 | /* The bxns helper may raise an EXCEPTION_EXIT exception, so in theory | |
856 | * we need to sync state before calling it, but: | |
857 | * - we don't need to do gen_set_pc_im() because the bxns helper will | |
858 | * always set the PC itself | |
859 | * - we don't need to do gen_set_condexec() because BXNS is UNPREDICTABLE | |
860 | * unless it's outside an IT block or the last insn in an IT block, | |
861 | * so we know that condexec == 0 (already set at the top of the TB) | |
862 | * is correct in the non-UNPREDICTABLE cases, and we can choose | |
863 | * "zeroes the IT bits" as our UNPREDICTABLE behaviour otherwise. | |
864 | */ | |
865 | gen_helper_v7m_bxns(cpu_env, var); | |
866 | tcg_temp_free_i32(var); | |
ef475b5d | 867 | s->base.is_jmp = DISAS_EXIT; |
fb602cb7 PM |
868 | } |
869 | ||
3e3fa230 PM |
870 | static inline void gen_blxns(DisasContext *s, int rm) |
871 | { | |
872 | TCGv_i32 var = load_reg(s, rm); | |
873 | ||
874 | /* We don't need to sync condexec state, for the same reason as bxns. | |
875 | * We do however need to set the PC, because the blxns helper reads it. | |
876 | * The blxns helper may throw an exception. | |
877 | */ | |
a0415916 | 878 | gen_set_pc_im(s, s->base.pc_next); |
3e3fa230 PM |
879 | gen_helper_v7m_blxns(cpu_env, var); |
880 | tcg_temp_free_i32(var); | |
881 | s->base.is_jmp = DISAS_EXIT; | |
882 | } | |
883 | ||
21aeb343 JR |
884 | /* Variant of store_reg which uses branch&exchange logic when storing |
885 | to r15 in ARM architecture v7 and above. The source must be a temporary | |
886 | and will be marked as dead. */ | |
7dcc1f89 | 887 | static inline void store_reg_bx(DisasContext *s, int reg, TCGv_i32 var) |
21aeb343 JR |
888 | { |
889 | if (reg == 15 && ENABLE_ARCH_7) { | |
890 | gen_bx(s, var); | |
891 | } else { | |
892 | store_reg(s, reg, var); | |
893 | } | |
894 | } | |
895 | ||
be5e7a76 DES |
896 | /* Variant of store_reg which uses branch&exchange logic when storing |
897 | * to r15 in ARM architecture v5T and above. This is used for storing | |
898 | * the results of a LDR/LDM/POP into r15, and corresponds to the cases | |
899 | * in the ARM ARM which use the LoadWritePC() pseudocode function. */ | |
7dcc1f89 | 900 | static inline void store_reg_from_load(DisasContext *s, int reg, TCGv_i32 var) |
be5e7a76 DES |
901 | { |
902 | if (reg == 15 && ENABLE_ARCH_5) { | |
3bb8a96f | 903 | gen_bx_excret(s, var); |
be5e7a76 DES |
904 | } else { |
905 | store_reg(s, reg, var); | |
906 | } | |
907 | } | |
908 | ||
e334bd31 PB |
909 | #ifdef CONFIG_USER_ONLY |
910 | #define IS_USER_ONLY 1 | |
911 | #else | |
912 | #define IS_USER_ONLY 0 | |
913 | #endif | |
914 | ||
08307563 PM |
915 | /* Abstractions of "generate code to do a guest load/store for |
916 | * AArch32", where a vaddr is always 32 bits (and is zero | |
917 | * extended if we're a 64 bit core) and data is also | |
918 | * 32 bits unless specifically doing a 64 bit access. | |
919 | * These functions work like tcg_gen_qemu_{ld,st}* except | |
09f78135 | 920 | * that the address argument is TCGv_i32 rather than TCGv. |
08307563 | 921 | */ |
08307563 | 922 | |
14776ab5 | 923 | static inline TCGv gen_aa32_addr(DisasContext *s, TCGv_i32 a32, MemOp op) |
08307563 | 924 | { |
7f5616f5 RH |
925 | TCGv addr = tcg_temp_new(); |
926 | tcg_gen_extu_i32_tl(addr, a32); | |
927 | ||
e334bd31 | 928 | /* Not needed for user-mode BE32, where we use MO_BE instead. */ |
7f5616f5 RH |
929 | if (!IS_USER_ONLY && s->sctlr_b && (op & MO_SIZE) < MO_32) { |
930 | tcg_gen_xori_tl(addr, addr, 4 - (1 << (op & MO_SIZE))); | |
e334bd31 | 931 | } |
7f5616f5 | 932 | return addr; |
08307563 PM |
933 | } |
934 | ||
7f5616f5 | 935 | static void gen_aa32_ld_i32(DisasContext *s, TCGv_i32 val, TCGv_i32 a32, |
14776ab5 | 936 | int index, MemOp opc) |
08307563 | 937 | { |
2aeba0d0 JS |
938 | TCGv addr; |
939 | ||
940 | if (arm_dc_feature(s, ARM_FEATURE_M) && | |
941 | !arm_dc_feature(s, ARM_FEATURE_M_MAIN)) { | |
942 | opc |= MO_ALIGN; | |
943 | } | |
944 | ||
945 | addr = gen_aa32_addr(s, a32, opc); | |
7f5616f5 RH |
946 | tcg_gen_qemu_ld_i32(val, addr, index, opc); |
947 | tcg_temp_free(addr); | |
08307563 PM |
948 | } |
949 | ||
7f5616f5 | 950 | static void gen_aa32_st_i32(DisasContext *s, TCGv_i32 val, TCGv_i32 a32, |
14776ab5 | 951 | int index, MemOp opc) |
7f5616f5 | 952 | { |
2aeba0d0 JS |
953 | TCGv addr; |
954 | ||
955 | if (arm_dc_feature(s, ARM_FEATURE_M) && | |
956 | !arm_dc_feature(s, ARM_FEATURE_M_MAIN)) { | |
957 | opc |= MO_ALIGN; | |
958 | } | |
959 | ||
960 | addr = gen_aa32_addr(s, a32, opc); | |
7f5616f5 RH |
961 | tcg_gen_qemu_st_i32(val, addr, index, opc); |
962 | tcg_temp_free(addr); | |
963 | } | |
08307563 | 964 | |
7f5616f5 | 965 | #define DO_GEN_LD(SUFF, OPC) \ |
12dcc321 | 966 | static inline void gen_aa32_ld##SUFF(DisasContext *s, TCGv_i32 val, \ |
7f5616f5 | 967 | TCGv_i32 a32, int index) \ |
08307563 | 968 | { \ |
7f5616f5 | 969 | gen_aa32_ld_i32(s, val, a32, index, OPC | s->be_data); \ |
08307563 PM |
970 | } |
971 | ||
7f5616f5 | 972 | #define DO_GEN_ST(SUFF, OPC) \ |
12dcc321 | 973 | static inline void gen_aa32_st##SUFF(DisasContext *s, TCGv_i32 val, \ |
7f5616f5 | 974 | TCGv_i32 a32, int index) \ |
08307563 | 975 | { \ |
7f5616f5 | 976 | gen_aa32_st_i32(s, val, a32, index, OPC | s->be_data); \ |
08307563 PM |
977 | } |
978 | ||
7f5616f5 | 979 | static inline void gen_aa32_frob64(DisasContext *s, TCGv_i64 val) |
08307563 | 980 | { |
e334bd31 PB |
981 | /* Not needed for user-mode BE32, where we use MO_BE instead. */ |
982 | if (!IS_USER_ONLY && s->sctlr_b) { | |
983 | tcg_gen_rotri_i64(val, val, 32); | |
984 | } | |
08307563 PM |
985 | } |
986 | ||
7f5616f5 | 987 | static void gen_aa32_ld_i64(DisasContext *s, TCGv_i64 val, TCGv_i32 a32, |
14776ab5 | 988 | int index, MemOp opc) |
08307563 | 989 | { |
7f5616f5 RH |
990 | TCGv addr = gen_aa32_addr(s, a32, opc); |
991 | tcg_gen_qemu_ld_i64(val, addr, index, opc); | |
992 | gen_aa32_frob64(s, val); | |
993 | tcg_temp_free(addr); | |
994 | } | |
995 | ||
996 | static inline void gen_aa32_ld64(DisasContext *s, TCGv_i64 val, | |
997 | TCGv_i32 a32, int index) | |
998 | { | |
999 | gen_aa32_ld_i64(s, val, a32, index, MO_Q | s->be_data); | |
1000 | } | |
1001 | ||
1002 | static void gen_aa32_st_i64(DisasContext *s, TCGv_i64 val, TCGv_i32 a32, | |
14776ab5 | 1003 | int index, MemOp opc) |
7f5616f5 RH |
1004 | { |
1005 | TCGv addr = gen_aa32_addr(s, a32, opc); | |
e334bd31 PB |
1006 | |
1007 | /* Not needed for user-mode BE32, where we use MO_BE instead. */ | |
1008 | if (!IS_USER_ONLY && s->sctlr_b) { | |
7f5616f5 | 1009 | TCGv_i64 tmp = tcg_temp_new_i64(); |
e334bd31 | 1010 | tcg_gen_rotri_i64(tmp, val, 32); |
7f5616f5 RH |
1011 | tcg_gen_qemu_st_i64(tmp, addr, index, opc); |
1012 | tcg_temp_free_i64(tmp); | |
e334bd31 | 1013 | } else { |
7f5616f5 | 1014 | tcg_gen_qemu_st_i64(val, addr, index, opc); |
e334bd31 | 1015 | } |
7f5616f5 | 1016 | tcg_temp_free(addr); |
08307563 PM |
1017 | } |
1018 | ||
7f5616f5 RH |
1019 | static inline void gen_aa32_st64(DisasContext *s, TCGv_i64 val, |
1020 | TCGv_i32 a32, int index) | |
1021 | { | |
1022 | gen_aa32_st_i64(s, val, a32, index, MO_Q | s->be_data); | |
1023 | } | |
08307563 | 1024 | |
7f5616f5 | 1025 | DO_GEN_LD(8u, MO_UB) |
7f5616f5 RH |
1026 | DO_GEN_LD(16u, MO_UW) |
1027 | DO_GEN_LD(32u, MO_UL) | |
7f5616f5 RH |
1028 | DO_GEN_ST(8, MO_UB) |
1029 | DO_GEN_ST(16, MO_UW) | |
1030 | DO_GEN_ST(32, MO_UL) | |
08307563 | 1031 | |
37e6456e PM |
1032 | static inline void gen_hvc(DisasContext *s, int imm16) |
1033 | { | |
1034 | /* The pre HVC helper handles cases when HVC gets trapped | |
1035 | * as an undefined insn by runtime configuration (ie before | |
1036 | * the insn really executes). | |
1037 | */ | |
43722a6d | 1038 | gen_set_pc_im(s, s->pc_curr); |
37e6456e PM |
1039 | gen_helper_pre_hvc(cpu_env); |
1040 | /* Otherwise we will treat this as a real exception which | |
1041 | * happens after execution of the insn. (The distinction matters | |
1042 | * for the PC value reported to the exception handler and also | |
1043 | * for single stepping.) | |
1044 | */ | |
1045 | s->svc_imm = imm16; | |
a0415916 | 1046 | gen_set_pc_im(s, s->base.pc_next); |
dcba3a8d | 1047 | s->base.is_jmp = DISAS_HVC; |
37e6456e PM |
1048 | } |
1049 | ||
1050 | static inline void gen_smc(DisasContext *s) | |
1051 | { | |
1052 | /* As with HVC, we may take an exception either before or after | |
1053 | * the insn executes. | |
1054 | */ | |
1055 | TCGv_i32 tmp; | |
1056 | ||
43722a6d | 1057 | gen_set_pc_im(s, s->pc_curr); |
37e6456e PM |
1058 | tmp = tcg_const_i32(syn_aa32_smc()); |
1059 | gen_helper_pre_smc(cpu_env, tmp); | |
1060 | tcg_temp_free_i32(tmp); | |
a0415916 | 1061 | gen_set_pc_im(s, s->base.pc_next); |
dcba3a8d | 1062 | s->base.is_jmp = DISAS_SMC; |
37e6456e PM |
1063 | } |
1064 | ||
aee828e7 | 1065 | static void gen_exception_internal_insn(DisasContext *s, uint32_t pc, int excp) |
d4a2dc67 PM |
1066 | { |
1067 | gen_set_condexec(s); | |
aee828e7 | 1068 | gen_set_pc_im(s, pc); |
d4a2dc67 | 1069 | gen_exception_internal(excp); |
dcba3a8d | 1070 | s->base.is_jmp = DISAS_NORETURN; |
d4a2dc67 PM |
1071 | } |
1072 | ||
a767fac8 | 1073 | static void gen_exception_insn(DisasContext *s, uint32_t pc, int excp, |
73710361 | 1074 | int syn, uint32_t target_el) |
d4a2dc67 PM |
1075 | { |
1076 | gen_set_condexec(s); | |
a767fac8 | 1077 | gen_set_pc_im(s, pc); |
73710361 | 1078 | gen_exception(excp, syn, target_el); |
dcba3a8d | 1079 | s->base.is_jmp = DISAS_NORETURN; |
d4a2dc67 PM |
1080 | } |
1081 | ||
06bcbda3 | 1082 | static void gen_exception_bkpt_insn(DisasContext *s, uint32_t syn) |
c900a2e6 PM |
1083 | { |
1084 | TCGv_i32 tcg_syn; | |
1085 | ||
1086 | gen_set_condexec(s); | |
06bcbda3 | 1087 | gen_set_pc_im(s, s->pc_curr); |
c900a2e6 PM |
1088 | tcg_syn = tcg_const_i32(syn); |
1089 | gen_helper_exception_bkpt_insn(cpu_env, tcg_syn); | |
1090 | tcg_temp_free_i32(tcg_syn); | |
1091 | s->base.is_jmp = DISAS_NORETURN; | |
1092 | } | |
1093 | ||
1ce21ba1 RH |
1094 | static void unallocated_encoding(DisasContext *s) |
1095 | { | |
1096 | /* Unallocated and reserved encodings are uncategorized */ | |
1097 | gen_exception_insn(s, s->pc_curr, EXCP_UDEF, syn_uncategorized(), | |
1098 | default_exception_el(s)); | |
1099 | } | |
1100 | ||
b5ff1b31 FB |
1101 | /* Force a TB lookup after an instruction that changes the CPU state. */ |
1102 | static inline void gen_lookup_tb(DisasContext *s) | |
1103 | { | |
a0415916 | 1104 | tcg_gen_movi_i32(cpu_R[15], s->base.pc_next); |
dcba3a8d | 1105 | s->base.is_jmp = DISAS_EXIT; |
b5ff1b31 FB |
1106 | } |
1107 | ||
19a6e31c PM |
1108 | static inline void gen_hlt(DisasContext *s, int imm) |
1109 | { | |
1110 | /* HLT. This has two purposes. | |
1111 | * Architecturally, it is an external halting debug instruction. | |
1112 | * Since QEMU doesn't implement external debug, we treat this as | |
1113 | * it is required for halting debug disabled: it will UNDEF. | |
1114 | * Secondly, "HLT 0x3C" is a T32 semihosting trap instruction, | |
1115 | * and "HLT 0xF000" is an A32 semihosting syscall. These traps | |
1116 | * must trigger semihosting even for ARMv7 and earlier, where | |
1117 | * HLT was an undefined encoding. | |
1118 | * In system mode, we don't allow userspace access to | |
1119 | * semihosting, to provide some semblance of security | |
1120 | * (and for consistency with our 32-bit semihosting). | |
1121 | */ | |
1122 | if (semihosting_enabled() && | |
1123 | #ifndef CONFIG_USER_ONLY | |
1124 | s->current_el != 0 && | |
1125 | #endif | |
1126 | (imm == (s->thumb ? 0x3c : 0xf000))) { | |
aee828e7 | 1127 | gen_exception_internal_insn(s, s->base.pc_next, EXCP_SEMIHOST); |
19a6e31c PM |
1128 | return; |
1129 | } | |
1130 | ||
1ce21ba1 | 1131 | unallocated_encoding(s); |
19a6e31c PM |
1132 | } |
1133 | ||
5aaebd13 PM |
1134 | static TCGv_ptr get_fpstatus_ptr(int neon) |
1135 | { | |
1136 | TCGv_ptr statusptr = tcg_temp_new_ptr(); | |
1137 | int offset; | |
1138 | if (neon) { | |
0ecb72a5 | 1139 | offset = offsetof(CPUARMState, vfp.standard_fp_status); |
5aaebd13 | 1140 | } else { |
0ecb72a5 | 1141 | offset = offsetof(CPUARMState, vfp.fp_status); |
5aaebd13 PM |
1142 | } |
1143 | tcg_gen_addi_ptr(statusptr, cpu_env, offset); | |
1144 | return statusptr; | |
1145 | } | |
1146 | ||
c39c2b90 | 1147 | static inline long vfp_reg_offset(bool dp, unsigned reg) |
8e96005d | 1148 | { |
9a2b5256 | 1149 | if (dp) { |
c39c2b90 | 1150 | return offsetof(CPUARMState, vfp.zregs[reg >> 1].d[reg & 1]); |
8e96005d | 1151 | } else { |
c39c2b90 | 1152 | long ofs = offsetof(CPUARMState, vfp.zregs[reg >> 2].d[(reg >> 1) & 1]); |
9a2b5256 RH |
1153 | if (reg & 1) { |
1154 | ofs += offsetof(CPU_DoubleU, l.upper); | |
1155 | } else { | |
1156 | ofs += offsetof(CPU_DoubleU, l.lower); | |
1157 | } | |
1158 | return ofs; | |
8e96005d FB |
1159 | } |
1160 | } | |
9ee6e8bb PB |
1161 | |
1162 | /* Return the offset of a 32-bit piece of a NEON register. | |
1163 | zero is the least significant end of the register. */ | |
1164 | static inline long | |
1165 | neon_reg_offset (int reg, int n) | |
1166 | { | |
1167 | int sreg; | |
1168 | sreg = reg * 2 + n; | |
1169 | return vfp_reg_offset(0, sreg); | |
1170 | } | |
1171 | ||
32f91fb7 RH |
1172 | /* Return the offset of a 2**SIZE piece of a NEON register, at index ELE, |
1173 | * where 0 is the least significant end of the register. | |
1174 | */ | |
1175 | static inline long | |
14776ab5 | 1176 | neon_element_offset(int reg, int element, MemOp size) |
32f91fb7 RH |
1177 | { |
1178 | int element_size = 1 << size; | |
1179 | int ofs = element * element_size; | |
1180 | #ifdef HOST_WORDS_BIGENDIAN | |
1181 | /* Calculate the offset assuming fully little-endian, | |
1182 | * then XOR to account for the order of the 8-byte units. | |
1183 | */ | |
1184 | if (element_size < 8) { | |
1185 | ofs ^= 8 - element_size; | |
1186 | } | |
1187 | #endif | |
1188 | return neon_reg_offset(reg, 0) + ofs; | |
1189 | } | |
1190 | ||
39d5492a | 1191 | static TCGv_i32 neon_load_reg(int reg, int pass) |
8f8e3aa4 | 1192 | { |
39d5492a | 1193 | TCGv_i32 tmp = tcg_temp_new_i32(); |
8f8e3aa4 PB |
1194 | tcg_gen_ld_i32(tmp, cpu_env, neon_reg_offset(reg, pass)); |
1195 | return tmp; | |
1196 | } | |
1197 | ||
14776ab5 | 1198 | static void neon_load_element(TCGv_i32 var, int reg, int ele, MemOp mop) |
2d6ac920 RH |
1199 | { |
1200 | long offset = neon_element_offset(reg, ele, mop & MO_SIZE); | |
1201 | ||
1202 | switch (mop) { | |
1203 | case MO_UB: | |
1204 | tcg_gen_ld8u_i32(var, cpu_env, offset); | |
1205 | break; | |
1206 | case MO_UW: | |
1207 | tcg_gen_ld16u_i32(var, cpu_env, offset); | |
1208 | break; | |
1209 | case MO_UL: | |
1210 | tcg_gen_ld_i32(var, cpu_env, offset); | |
1211 | break; | |
1212 | default: | |
1213 | g_assert_not_reached(); | |
1214 | } | |
1215 | } | |
1216 | ||
14776ab5 | 1217 | static void neon_load_element64(TCGv_i64 var, int reg, int ele, MemOp mop) |
ac55d007 RH |
1218 | { |
1219 | long offset = neon_element_offset(reg, ele, mop & MO_SIZE); | |
1220 | ||
1221 | switch (mop) { | |
1222 | case MO_UB: | |
1223 | tcg_gen_ld8u_i64(var, cpu_env, offset); | |
1224 | break; | |
1225 | case MO_UW: | |
1226 | tcg_gen_ld16u_i64(var, cpu_env, offset); | |
1227 | break; | |
1228 | case MO_UL: | |
1229 | tcg_gen_ld32u_i64(var, cpu_env, offset); | |
1230 | break; | |
1231 | case MO_Q: | |
1232 | tcg_gen_ld_i64(var, cpu_env, offset); | |
1233 | break; | |
1234 | default: | |
1235 | g_assert_not_reached(); | |
1236 | } | |
1237 | } | |
1238 | ||
39d5492a | 1239 | static void neon_store_reg(int reg, int pass, TCGv_i32 var) |
8f8e3aa4 PB |
1240 | { |
1241 | tcg_gen_st_i32(var, cpu_env, neon_reg_offset(reg, pass)); | |
7d1b0095 | 1242 | tcg_temp_free_i32(var); |
8f8e3aa4 PB |
1243 | } |
1244 | ||
14776ab5 | 1245 | static void neon_store_element(int reg, int ele, MemOp size, TCGv_i32 var) |
2d6ac920 RH |
1246 | { |
1247 | long offset = neon_element_offset(reg, ele, size); | |
1248 | ||
1249 | switch (size) { | |
1250 | case MO_8: | |
1251 | tcg_gen_st8_i32(var, cpu_env, offset); | |
1252 | break; | |
1253 | case MO_16: | |
1254 | tcg_gen_st16_i32(var, cpu_env, offset); | |
1255 | break; | |
1256 | case MO_32: | |
1257 | tcg_gen_st_i32(var, cpu_env, offset); | |
1258 | break; | |
1259 | default: | |
1260 | g_assert_not_reached(); | |
1261 | } | |
1262 | } | |
1263 | ||
14776ab5 | 1264 | static void neon_store_element64(int reg, int ele, MemOp size, TCGv_i64 var) |
ac55d007 RH |
1265 | { |
1266 | long offset = neon_element_offset(reg, ele, size); | |
1267 | ||
1268 | switch (size) { | |
1269 | case MO_8: | |
1270 | tcg_gen_st8_i64(var, cpu_env, offset); | |
1271 | break; | |
1272 | case MO_16: | |
1273 | tcg_gen_st16_i64(var, cpu_env, offset); | |
1274 | break; | |
1275 | case MO_32: | |
1276 | tcg_gen_st32_i64(var, cpu_env, offset); | |
1277 | break; | |
1278 | case MO_64: | |
1279 | tcg_gen_st_i64(var, cpu_env, offset); | |
1280 | break; | |
1281 | default: | |
1282 | g_assert_not_reached(); | |
1283 | } | |
1284 | } | |
1285 | ||
a7812ae4 | 1286 | static inline void neon_load_reg64(TCGv_i64 var, int reg) |
ad69471c PB |
1287 | { |
1288 | tcg_gen_ld_i64(var, cpu_env, vfp_reg_offset(1, reg)); | |
1289 | } | |
1290 | ||
a7812ae4 | 1291 | static inline void neon_store_reg64(TCGv_i64 var, int reg) |
ad69471c PB |
1292 | { |
1293 | tcg_gen_st_i64(var, cpu_env, vfp_reg_offset(1, reg)); | |
1294 | } | |
1295 | ||
160f3b64 PM |
1296 | static inline void neon_load_reg32(TCGv_i32 var, int reg) |
1297 | { | |
1298 | tcg_gen_ld_i32(var, cpu_env, vfp_reg_offset(false, reg)); | |
1299 | } | |
1300 | ||
1301 | static inline void neon_store_reg32(TCGv_i32 var, int reg) | |
1302 | { | |
1303 | tcg_gen_st_i32(var, cpu_env, vfp_reg_offset(false, reg)); | |
1304 | } | |
1305 | ||
1a66ac61 RH |
1306 | static TCGv_ptr vfp_reg_ptr(bool dp, int reg) |
1307 | { | |
1308 | TCGv_ptr ret = tcg_temp_new_ptr(); | |
1309 | tcg_gen_addi_ptr(ret, cpu_env, vfp_reg_offset(dp, reg)); | |
1310 | return ret; | |
1311 | } | |
1312 | ||
d00584b7 | 1313 | #define ARM_CP_RW_BIT (1 << 20) |
18c9b560 | 1314 | |
78e138bc PM |
1315 | /* Include the VFP decoder */ |
1316 | #include "translate-vfp.inc.c" | |
1317 | ||
a7812ae4 | 1318 | static inline void iwmmxt_load_reg(TCGv_i64 var, int reg) |
e677137d | 1319 | { |
0ecb72a5 | 1320 | tcg_gen_ld_i64(var, cpu_env, offsetof(CPUARMState, iwmmxt.regs[reg])); |
e677137d PB |
1321 | } |
1322 | ||
a7812ae4 | 1323 | static inline void iwmmxt_store_reg(TCGv_i64 var, int reg) |
e677137d | 1324 | { |
0ecb72a5 | 1325 | tcg_gen_st_i64(var, cpu_env, offsetof(CPUARMState, iwmmxt.regs[reg])); |
e677137d PB |
1326 | } |
1327 | ||
39d5492a | 1328 | static inline TCGv_i32 iwmmxt_load_creg(int reg) |
e677137d | 1329 | { |
39d5492a | 1330 | TCGv_i32 var = tcg_temp_new_i32(); |
0ecb72a5 | 1331 | tcg_gen_ld_i32(var, cpu_env, offsetof(CPUARMState, iwmmxt.cregs[reg])); |
da6b5335 | 1332 | return var; |
e677137d PB |
1333 | } |
1334 | ||
39d5492a | 1335 | static inline void iwmmxt_store_creg(int reg, TCGv_i32 var) |
e677137d | 1336 | { |
0ecb72a5 | 1337 | tcg_gen_st_i32(var, cpu_env, offsetof(CPUARMState, iwmmxt.cregs[reg])); |
7d1b0095 | 1338 | tcg_temp_free_i32(var); |
e677137d PB |
1339 | } |
1340 | ||
1341 | static inline void gen_op_iwmmxt_movq_wRn_M0(int rn) | |
1342 | { | |
1343 | iwmmxt_store_reg(cpu_M0, rn); | |
1344 | } | |
1345 | ||
1346 | static inline void gen_op_iwmmxt_movq_M0_wRn(int rn) | |
1347 | { | |
1348 | iwmmxt_load_reg(cpu_M0, rn); | |
1349 | } | |
1350 | ||
1351 | static inline void gen_op_iwmmxt_orq_M0_wRn(int rn) | |
1352 | { | |
1353 | iwmmxt_load_reg(cpu_V1, rn); | |
1354 | tcg_gen_or_i64(cpu_M0, cpu_M0, cpu_V1); | |
1355 | } | |
1356 | ||
1357 | static inline void gen_op_iwmmxt_andq_M0_wRn(int rn) | |
1358 | { | |
1359 | iwmmxt_load_reg(cpu_V1, rn); | |
1360 | tcg_gen_and_i64(cpu_M0, cpu_M0, cpu_V1); | |
1361 | } | |
1362 | ||
1363 | static inline void gen_op_iwmmxt_xorq_M0_wRn(int rn) | |
1364 | { | |
1365 | iwmmxt_load_reg(cpu_V1, rn); | |
1366 | tcg_gen_xor_i64(cpu_M0, cpu_M0, cpu_V1); | |
1367 | } | |
1368 | ||
1369 | #define IWMMXT_OP(name) \ | |
1370 | static inline void gen_op_iwmmxt_##name##_M0_wRn(int rn) \ | |
1371 | { \ | |
1372 | iwmmxt_load_reg(cpu_V1, rn); \ | |
1373 | gen_helper_iwmmxt_##name(cpu_M0, cpu_M0, cpu_V1); \ | |
1374 | } | |
1375 | ||
477955bd PM |
1376 | #define IWMMXT_OP_ENV(name) \ |
1377 | static inline void gen_op_iwmmxt_##name##_M0_wRn(int rn) \ | |
1378 | { \ | |
1379 | iwmmxt_load_reg(cpu_V1, rn); \ | |
1380 | gen_helper_iwmmxt_##name(cpu_M0, cpu_env, cpu_M0, cpu_V1); \ | |
1381 | } | |
1382 | ||
1383 | #define IWMMXT_OP_ENV_SIZE(name) \ | |
1384 | IWMMXT_OP_ENV(name##b) \ | |
1385 | IWMMXT_OP_ENV(name##w) \ | |
1386 | IWMMXT_OP_ENV(name##l) | |
e677137d | 1387 | |
477955bd | 1388 | #define IWMMXT_OP_ENV1(name) \ |
e677137d PB |
1389 | static inline void gen_op_iwmmxt_##name##_M0(void) \ |
1390 | { \ | |
477955bd | 1391 | gen_helper_iwmmxt_##name(cpu_M0, cpu_env, cpu_M0); \ |
e677137d PB |
1392 | } |
1393 | ||
1394 | IWMMXT_OP(maddsq) | |
1395 | IWMMXT_OP(madduq) | |
1396 | IWMMXT_OP(sadb) | |
1397 | IWMMXT_OP(sadw) | |
1398 | IWMMXT_OP(mulslw) | |
1399 | IWMMXT_OP(mulshw) | |
1400 | IWMMXT_OP(mululw) | |
1401 | IWMMXT_OP(muluhw) | |
1402 | IWMMXT_OP(macsw) | |
1403 | IWMMXT_OP(macuw) | |
1404 | ||
477955bd PM |
1405 | IWMMXT_OP_ENV_SIZE(unpackl) |
1406 | IWMMXT_OP_ENV_SIZE(unpackh) | |
1407 | ||
1408 | IWMMXT_OP_ENV1(unpacklub) | |
1409 | IWMMXT_OP_ENV1(unpackluw) | |
1410 | IWMMXT_OP_ENV1(unpacklul) | |
1411 | IWMMXT_OP_ENV1(unpackhub) | |
1412 | IWMMXT_OP_ENV1(unpackhuw) | |
1413 | IWMMXT_OP_ENV1(unpackhul) | |
1414 | IWMMXT_OP_ENV1(unpacklsb) | |
1415 | IWMMXT_OP_ENV1(unpacklsw) | |
1416 | IWMMXT_OP_ENV1(unpacklsl) | |
1417 | IWMMXT_OP_ENV1(unpackhsb) | |
1418 | IWMMXT_OP_ENV1(unpackhsw) | |
1419 | IWMMXT_OP_ENV1(unpackhsl) | |
1420 | ||
1421 | IWMMXT_OP_ENV_SIZE(cmpeq) | |
1422 | IWMMXT_OP_ENV_SIZE(cmpgtu) | |
1423 | IWMMXT_OP_ENV_SIZE(cmpgts) | |
1424 | ||
1425 | IWMMXT_OP_ENV_SIZE(mins) | |
1426 | IWMMXT_OP_ENV_SIZE(minu) | |
1427 | IWMMXT_OP_ENV_SIZE(maxs) | |
1428 | IWMMXT_OP_ENV_SIZE(maxu) | |
1429 | ||
1430 | IWMMXT_OP_ENV_SIZE(subn) | |
1431 | IWMMXT_OP_ENV_SIZE(addn) | |
1432 | IWMMXT_OP_ENV_SIZE(subu) | |
1433 | IWMMXT_OP_ENV_SIZE(addu) | |
1434 | IWMMXT_OP_ENV_SIZE(subs) | |
1435 | IWMMXT_OP_ENV_SIZE(adds) | |
1436 | ||
1437 | IWMMXT_OP_ENV(avgb0) | |
1438 | IWMMXT_OP_ENV(avgb1) | |
1439 | IWMMXT_OP_ENV(avgw0) | |
1440 | IWMMXT_OP_ENV(avgw1) | |
e677137d | 1441 | |
477955bd PM |
1442 | IWMMXT_OP_ENV(packuw) |
1443 | IWMMXT_OP_ENV(packul) | |
1444 | IWMMXT_OP_ENV(packuq) | |
1445 | IWMMXT_OP_ENV(packsw) | |
1446 | IWMMXT_OP_ENV(packsl) | |
1447 | IWMMXT_OP_ENV(packsq) | |
e677137d | 1448 | |
e677137d PB |
1449 | static void gen_op_iwmmxt_set_mup(void) |
1450 | { | |
39d5492a | 1451 | TCGv_i32 tmp; |
e677137d PB |
1452 | tmp = load_cpu_field(iwmmxt.cregs[ARM_IWMMXT_wCon]); |
1453 | tcg_gen_ori_i32(tmp, tmp, 2); | |
1454 | store_cpu_field(tmp, iwmmxt.cregs[ARM_IWMMXT_wCon]); | |
1455 | } | |
1456 | ||
1457 | static void gen_op_iwmmxt_set_cup(void) | |
1458 | { | |
39d5492a | 1459 | TCGv_i32 tmp; |
e677137d PB |
1460 | tmp = load_cpu_field(iwmmxt.cregs[ARM_IWMMXT_wCon]); |
1461 | tcg_gen_ori_i32(tmp, tmp, 1); | |
1462 | store_cpu_field(tmp, iwmmxt.cregs[ARM_IWMMXT_wCon]); | |
1463 | } | |
1464 | ||
1465 | static void gen_op_iwmmxt_setpsr_nz(void) | |
1466 | { | |
39d5492a | 1467 | TCGv_i32 tmp = tcg_temp_new_i32(); |
e677137d PB |
1468 | gen_helper_iwmmxt_setpsr_nz(tmp, cpu_M0); |
1469 | store_cpu_field(tmp, iwmmxt.cregs[ARM_IWMMXT_wCASF]); | |
1470 | } | |
1471 | ||
1472 | static inline void gen_op_iwmmxt_addl_M0_wRn(int rn) | |
1473 | { | |
1474 | iwmmxt_load_reg(cpu_V1, rn); | |
86831435 | 1475 | tcg_gen_ext32u_i64(cpu_V1, cpu_V1); |
e677137d PB |
1476 | tcg_gen_add_i64(cpu_M0, cpu_M0, cpu_V1); |
1477 | } | |
1478 | ||
39d5492a PM |
1479 | static inline int gen_iwmmxt_address(DisasContext *s, uint32_t insn, |
1480 | TCGv_i32 dest) | |
18c9b560 AZ |
1481 | { |
1482 | int rd; | |
1483 | uint32_t offset; | |
39d5492a | 1484 | TCGv_i32 tmp; |
18c9b560 AZ |
1485 | |
1486 | rd = (insn >> 16) & 0xf; | |
da6b5335 | 1487 | tmp = load_reg(s, rd); |
18c9b560 AZ |
1488 | |
1489 | offset = (insn & 0xff) << ((insn >> 7) & 2); | |
1490 | if (insn & (1 << 24)) { | |
1491 | /* Pre indexed */ | |
1492 | if (insn & (1 << 23)) | |
da6b5335 | 1493 | tcg_gen_addi_i32(tmp, tmp, offset); |
18c9b560 | 1494 | else |
da6b5335 FN |
1495 | tcg_gen_addi_i32(tmp, tmp, -offset); |
1496 | tcg_gen_mov_i32(dest, tmp); | |
18c9b560 | 1497 | if (insn & (1 << 21)) |
da6b5335 FN |
1498 | store_reg(s, rd, tmp); |
1499 | else | |
7d1b0095 | 1500 | tcg_temp_free_i32(tmp); |
18c9b560 AZ |
1501 | } else if (insn & (1 << 21)) { |
1502 | /* Post indexed */ | |
da6b5335 | 1503 | tcg_gen_mov_i32(dest, tmp); |
18c9b560 | 1504 | if (insn & (1 << 23)) |
da6b5335 | 1505 | tcg_gen_addi_i32(tmp, tmp, offset); |
18c9b560 | 1506 | else |
da6b5335 FN |
1507 | tcg_gen_addi_i32(tmp, tmp, -offset); |
1508 | store_reg(s, rd, tmp); | |
18c9b560 AZ |
1509 | } else if (!(insn & (1 << 23))) |
1510 | return 1; | |
1511 | return 0; | |
1512 | } | |
1513 | ||
39d5492a | 1514 | static inline int gen_iwmmxt_shift(uint32_t insn, uint32_t mask, TCGv_i32 dest) |
18c9b560 AZ |
1515 | { |
1516 | int rd = (insn >> 0) & 0xf; | |
39d5492a | 1517 | TCGv_i32 tmp; |
18c9b560 | 1518 | |
da6b5335 FN |
1519 | if (insn & (1 << 8)) { |
1520 | if (rd < ARM_IWMMXT_wCGR0 || rd > ARM_IWMMXT_wCGR3) { | |
18c9b560 | 1521 | return 1; |
da6b5335 FN |
1522 | } else { |
1523 | tmp = iwmmxt_load_creg(rd); | |
1524 | } | |
1525 | } else { | |
7d1b0095 | 1526 | tmp = tcg_temp_new_i32(); |
da6b5335 | 1527 | iwmmxt_load_reg(cpu_V0, rd); |
ecc7b3aa | 1528 | tcg_gen_extrl_i64_i32(tmp, cpu_V0); |
da6b5335 FN |
1529 | } |
1530 | tcg_gen_andi_i32(tmp, tmp, mask); | |
1531 | tcg_gen_mov_i32(dest, tmp); | |
7d1b0095 | 1532 | tcg_temp_free_i32(tmp); |
18c9b560 AZ |
1533 | return 0; |
1534 | } | |
1535 | ||
a1c7273b | 1536 | /* Disassemble an iwMMXt instruction. Returns nonzero if an error occurred |
18c9b560 | 1537 | (ie. an undefined instruction). */ |
7dcc1f89 | 1538 | static int disas_iwmmxt_insn(DisasContext *s, uint32_t insn) |
18c9b560 AZ |
1539 | { |
1540 | int rd, wrd; | |
1541 | int rdhi, rdlo, rd0, rd1, i; | |
39d5492a PM |
1542 | TCGv_i32 addr; |
1543 | TCGv_i32 tmp, tmp2, tmp3; | |
18c9b560 AZ |
1544 | |
1545 | if ((insn & 0x0e000e00) == 0x0c000000) { | |
1546 | if ((insn & 0x0fe00ff0) == 0x0c400000) { | |
1547 | wrd = insn & 0xf; | |
1548 | rdlo = (insn >> 12) & 0xf; | |
1549 | rdhi = (insn >> 16) & 0xf; | |
d00584b7 | 1550 | if (insn & ARM_CP_RW_BIT) { /* TMRRC */ |
da6b5335 | 1551 | iwmmxt_load_reg(cpu_V0, wrd); |
ecc7b3aa | 1552 | tcg_gen_extrl_i64_i32(cpu_R[rdlo], cpu_V0); |
664b7e3b | 1553 | tcg_gen_extrh_i64_i32(cpu_R[rdhi], cpu_V0); |
d00584b7 | 1554 | } else { /* TMCRR */ |
da6b5335 FN |
1555 | tcg_gen_concat_i32_i64(cpu_V0, cpu_R[rdlo], cpu_R[rdhi]); |
1556 | iwmmxt_store_reg(cpu_V0, wrd); | |
18c9b560 AZ |
1557 | gen_op_iwmmxt_set_mup(); |
1558 | } | |
1559 | return 0; | |
1560 | } | |
1561 | ||
1562 | wrd = (insn >> 12) & 0xf; | |
7d1b0095 | 1563 | addr = tcg_temp_new_i32(); |
da6b5335 | 1564 | if (gen_iwmmxt_address(s, insn, addr)) { |
7d1b0095 | 1565 | tcg_temp_free_i32(addr); |
18c9b560 | 1566 | return 1; |
da6b5335 | 1567 | } |
18c9b560 | 1568 | if (insn & ARM_CP_RW_BIT) { |
d00584b7 | 1569 | if ((insn >> 28) == 0xf) { /* WLDRW wCx */ |
7d1b0095 | 1570 | tmp = tcg_temp_new_i32(); |
12dcc321 | 1571 | gen_aa32_ld32u(s, tmp, addr, get_mem_index(s)); |
da6b5335 | 1572 | iwmmxt_store_creg(wrd, tmp); |
18c9b560 | 1573 | } else { |
e677137d PB |
1574 | i = 1; |
1575 | if (insn & (1 << 8)) { | |
d00584b7 | 1576 | if (insn & (1 << 22)) { /* WLDRD */ |
12dcc321 | 1577 | gen_aa32_ld64(s, cpu_M0, addr, get_mem_index(s)); |
e677137d | 1578 | i = 0; |
d00584b7 | 1579 | } else { /* WLDRW wRd */ |
29531141 | 1580 | tmp = tcg_temp_new_i32(); |
12dcc321 | 1581 | gen_aa32_ld32u(s, tmp, addr, get_mem_index(s)); |
e677137d PB |
1582 | } |
1583 | } else { | |
29531141 | 1584 | tmp = tcg_temp_new_i32(); |
d00584b7 | 1585 | if (insn & (1 << 22)) { /* WLDRH */ |
12dcc321 | 1586 | gen_aa32_ld16u(s, tmp, addr, get_mem_index(s)); |
d00584b7 | 1587 | } else { /* WLDRB */ |
12dcc321 | 1588 | gen_aa32_ld8u(s, tmp, addr, get_mem_index(s)); |
e677137d PB |
1589 | } |
1590 | } | |
1591 | if (i) { | |
1592 | tcg_gen_extu_i32_i64(cpu_M0, tmp); | |
7d1b0095 | 1593 | tcg_temp_free_i32(tmp); |
e677137d | 1594 | } |
18c9b560 AZ |
1595 | gen_op_iwmmxt_movq_wRn_M0(wrd); |
1596 | } | |
1597 | } else { | |
d00584b7 | 1598 | if ((insn >> 28) == 0xf) { /* WSTRW wCx */ |
da6b5335 | 1599 | tmp = iwmmxt_load_creg(wrd); |
12dcc321 | 1600 | gen_aa32_st32(s, tmp, addr, get_mem_index(s)); |
18c9b560 AZ |
1601 | } else { |
1602 | gen_op_iwmmxt_movq_M0_wRn(wrd); | |
7d1b0095 | 1603 | tmp = tcg_temp_new_i32(); |
e677137d | 1604 | if (insn & (1 << 8)) { |
d00584b7 | 1605 | if (insn & (1 << 22)) { /* WSTRD */ |
12dcc321 | 1606 | gen_aa32_st64(s, cpu_M0, addr, get_mem_index(s)); |
d00584b7 | 1607 | } else { /* WSTRW wRd */ |
ecc7b3aa | 1608 | tcg_gen_extrl_i64_i32(tmp, cpu_M0); |
12dcc321 | 1609 | gen_aa32_st32(s, tmp, addr, get_mem_index(s)); |
e677137d PB |
1610 | } |
1611 | } else { | |
d00584b7 | 1612 | if (insn & (1 << 22)) { /* WSTRH */ |
ecc7b3aa | 1613 | tcg_gen_extrl_i64_i32(tmp, cpu_M0); |
12dcc321 | 1614 | gen_aa32_st16(s, tmp, addr, get_mem_index(s)); |
d00584b7 | 1615 | } else { /* WSTRB */ |
ecc7b3aa | 1616 | tcg_gen_extrl_i64_i32(tmp, cpu_M0); |
12dcc321 | 1617 | gen_aa32_st8(s, tmp, addr, get_mem_index(s)); |
e677137d PB |
1618 | } |
1619 | } | |
18c9b560 | 1620 | } |
29531141 | 1621 | tcg_temp_free_i32(tmp); |
18c9b560 | 1622 | } |
7d1b0095 | 1623 | tcg_temp_free_i32(addr); |
18c9b560 AZ |
1624 | return 0; |
1625 | } | |
1626 | ||
1627 | if ((insn & 0x0f000000) != 0x0e000000) | |
1628 | return 1; | |
1629 | ||
1630 | switch (((insn >> 12) & 0xf00) | ((insn >> 4) & 0xff)) { | |
d00584b7 | 1631 | case 0x000: /* WOR */ |
18c9b560 AZ |
1632 | wrd = (insn >> 12) & 0xf; |
1633 | rd0 = (insn >> 0) & 0xf; | |
1634 | rd1 = (insn >> 16) & 0xf; | |
1635 | gen_op_iwmmxt_movq_M0_wRn(rd0); | |
1636 | gen_op_iwmmxt_orq_M0_wRn(rd1); | |
1637 | gen_op_iwmmxt_setpsr_nz(); | |
1638 | gen_op_iwmmxt_movq_wRn_M0(wrd); | |
1639 | gen_op_iwmmxt_set_mup(); | |
1640 | gen_op_iwmmxt_set_cup(); | |
1641 | break; | |
d00584b7 | 1642 | case 0x011: /* TMCR */ |
18c9b560 AZ |
1643 | if (insn & 0xf) |
1644 | return 1; | |
1645 | rd = (insn >> 12) & 0xf; | |
1646 | wrd = (insn >> 16) & 0xf; | |
1647 | switch (wrd) { | |
1648 | case ARM_IWMMXT_wCID: | |
1649 | case ARM_IWMMXT_wCASF: | |
1650 | break; | |
1651 | case ARM_IWMMXT_wCon: | |
1652 | gen_op_iwmmxt_set_cup(); | |
1653 | /* Fall through. */ | |
1654 | case ARM_IWMMXT_wCSSF: | |
da6b5335 FN |
1655 | tmp = iwmmxt_load_creg(wrd); |
1656 | tmp2 = load_reg(s, rd); | |
f669df27 | 1657 | tcg_gen_andc_i32(tmp, tmp, tmp2); |
7d1b0095 | 1658 | tcg_temp_free_i32(tmp2); |
da6b5335 | 1659 | iwmmxt_store_creg(wrd, tmp); |
18c9b560 AZ |
1660 | break; |
1661 | case ARM_IWMMXT_wCGR0: | |
1662 | case ARM_IWMMXT_wCGR1: | |
1663 | case ARM_IWMMXT_wCGR2: | |
1664 | case ARM_IWMMXT_wCGR3: | |
1665 | gen_op_iwmmxt_set_cup(); | |
da6b5335 FN |
1666 | tmp = load_reg(s, rd); |
1667 | iwmmxt_store_creg(wrd, tmp); | |
18c9b560 AZ |
1668 | break; |
1669 | default: | |
1670 | return 1; | |
1671 | } | |
1672 | break; | |
d00584b7 | 1673 | case 0x100: /* WXOR */ |
18c9b560 AZ |
1674 | wrd = (insn >> 12) & 0xf; |
1675 | rd0 = (insn >> 0) & 0xf; | |
1676 | rd1 = (insn >> 16) & 0xf; | |
1677 | gen_op_iwmmxt_movq_M0_wRn(rd0); | |
1678 | gen_op_iwmmxt_xorq_M0_wRn(rd1); | |
1679 | gen_op_iwmmxt_setpsr_nz(); | |
1680 | gen_op_iwmmxt_movq_wRn_M0(wrd); | |
1681 | gen_op_iwmmxt_set_mup(); | |
1682 | gen_op_iwmmxt_set_cup(); | |
1683 | break; | |
d00584b7 | 1684 | case 0x111: /* TMRC */ |
18c9b560 AZ |
1685 | if (insn & 0xf) |
1686 | return 1; | |
1687 | rd = (insn >> 12) & 0xf; | |
1688 | wrd = (insn >> 16) & 0xf; | |
da6b5335 FN |
1689 | tmp = iwmmxt_load_creg(wrd); |
1690 | store_reg(s, rd, tmp); | |
18c9b560 | 1691 | break; |
d00584b7 | 1692 | case 0x300: /* WANDN */ |
18c9b560 AZ |
1693 | wrd = (insn >> 12) & 0xf; |
1694 | rd0 = (insn >> 0) & 0xf; | |
1695 | rd1 = (insn >> 16) & 0xf; | |
1696 | gen_op_iwmmxt_movq_M0_wRn(rd0); | |
e677137d | 1697 | tcg_gen_neg_i64(cpu_M0, cpu_M0); |
18c9b560 AZ |
1698 | gen_op_iwmmxt_andq_M0_wRn(rd1); |
1699 | gen_op_iwmmxt_setpsr_nz(); | |
1700 | gen_op_iwmmxt_movq_wRn_M0(wrd); | |
1701 | gen_op_iwmmxt_set_mup(); | |
1702 | gen_op_iwmmxt_set_cup(); | |
1703 | break; | |
d00584b7 | 1704 | case 0x200: /* WAND */ |
18c9b560 AZ |
1705 | wrd = (insn >> 12) & 0xf; |
1706 | rd0 = (insn >> 0) & 0xf; | |
1707 | rd1 = (insn >> 16) & 0xf; | |
1708 | gen_op_iwmmxt_movq_M0_wRn(rd0); | |
1709 | gen_op_iwmmxt_andq_M0_wRn(rd1); | |
1710 | gen_op_iwmmxt_setpsr_nz(); | |
1711 | gen_op_iwmmxt_movq_wRn_M0(wrd); | |
1712 | gen_op_iwmmxt_set_mup(); | |
1713 | gen_op_iwmmxt_set_cup(); | |
1714 | break; | |
d00584b7 | 1715 | case 0x810: case 0xa10: /* WMADD */ |
18c9b560 AZ |
1716 | wrd = (insn >> 12) & 0xf; |
1717 | rd0 = (insn >> 0) & 0xf; | |
1718 | rd1 = (insn >> 16) & 0xf; | |
1719 | gen_op_iwmmxt_movq_M0_wRn(rd0); | |
1720 | if (insn & (1 << 21)) | |
1721 | gen_op_iwmmxt_maddsq_M0_wRn(rd1); | |
1722 | else | |
1723 | gen_op_iwmmxt_madduq_M0_wRn(rd1); | |
1724 | gen_op_iwmmxt_movq_wRn_M0(wrd); | |
1725 | gen_op_iwmmxt_set_mup(); | |
1726 | break; | |
d00584b7 | 1727 | case 0x10e: case 0x50e: case 0x90e: case 0xd0e: /* WUNPCKIL */ |
18c9b560 AZ |
1728 | wrd = (insn >> 12) & 0xf; |
1729 | rd0 = (insn >> 16) & 0xf; | |
1730 | rd1 = (insn >> 0) & 0xf; | |
1731 | gen_op_iwmmxt_movq_M0_wRn(rd0); | |
1732 | switch ((insn >> 22) & 3) { | |
1733 | case 0: | |
1734 | gen_op_iwmmxt_unpacklb_M0_wRn(rd1); | |
1735 | break; | |
1736 | case 1: | |
1737 | gen_op_iwmmxt_unpacklw_M0_wRn(rd1); | |
1738 | break; | |
1739 | case 2: | |
1740 | gen_op_iwmmxt_unpackll_M0_wRn(rd1); | |
1741 | break; | |
1742 | case 3: | |
1743 | return 1; | |
1744 | } | |
1745 | gen_op_iwmmxt_movq_wRn_M0(wrd); | |
1746 | gen_op_iwmmxt_set_mup(); | |
1747 | gen_op_iwmmxt_set_cup(); | |
1748 | break; | |
d00584b7 | 1749 | case 0x10c: case 0x50c: case 0x90c: case 0xd0c: /* WUNPCKIH */ |
18c9b560 AZ |
1750 | wrd = (insn >> 12) & 0xf; |
1751 | rd0 = (insn >> 16) & 0xf; | |
1752 | rd1 = (insn >> 0) & 0xf; | |
1753 | gen_op_iwmmxt_movq_M0_wRn(rd0); | |
1754 | switch ((insn >> 22) & 3) { | |
1755 | case 0: | |
1756 | gen_op_iwmmxt_unpackhb_M0_wRn(rd1); | |
1757 | break; | |
1758 | case 1: | |
1759 | gen_op_iwmmxt_unpackhw_M0_wRn(rd1); | |
1760 | break; | |
1761 | case 2: | |
1762 | gen_op_iwmmxt_unpackhl_M0_wRn(rd1); | |
1763 | break; | |
1764 | case 3: | |
1765 | return 1; | |
1766 | } | |
1767 | gen_op_iwmmxt_movq_wRn_M0(wrd); | |
1768 | gen_op_iwmmxt_set_mup(); | |
1769 | gen_op_iwmmxt_set_cup(); | |
1770 | break; | |
d00584b7 | 1771 | case 0x012: case 0x112: case 0x412: case 0x512: /* WSAD */ |
18c9b560 AZ |
1772 | wrd = (insn >> 12) & 0xf; |
1773 | rd0 = (insn >> 16) & 0xf; | |
1774 | rd1 = (insn >> 0) & 0xf; | |
1775 | gen_op_iwmmxt_movq_M0_wRn(rd0); | |
1776 | if (insn & (1 << 22)) | |
1777 | gen_op_iwmmxt_sadw_M0_wRn(rd1); | |
1778 | else | |
1779 | gen_op_iwmmxt_sadb_M0_wRn(rd1); | |
1780 | if (!(insn & (1 << 20))) | |
1781 | gen_op_iwmmxt_addl_M0_wRn(wrd); | |
1782 | gen_op_iwmmxt_movq_wRn_M0(wrd); | |
1783 | gen_op_iwmmxt_set_mup(); | |
1784 | break; | |
d00584b7 | 1785 | case 0x010: case 0x110: case 0x210: case 0x310: /* WMUL */ |
18c9b560 AZ |
1786 | wrd = (insn >> 12) & 0xf; |
1787 | rd0 = (insn >> 16) & 0xf; | |
1788 | rd1 = (insn >> 0) & 0xf; | |
1789 | gen_op_iwmmxt_movq_M0_wRn(rd0); | |
e677137d PB |
1790 | if (insn & (1 << 21)) { |
1791 | if (insn & (1 << 20)) | |
1792 | gen_op_iwmmxt_mulshw_M0_wRn(rd1); | |
1793 | else | |
1794 | gen_op_iwmmxt_mulslw_M0_wRn(rd1); | |
1795 | } else { | |
1796 | if (insn & (1 << 20)) | |
1797 | gen_op_iwmmxt_muluhw_M0_wRn(rd1); | |
1798 | else | |
1799 | gen_op_iwmmxt_mululw_M0_wRn(rd1); | |
1800 | } | |
18c9b560 AZ |
1801 | gen_op_iwmmxt_movq_wRn_M0(wrd); |
1802 | gen_op_iwmmxt_set_mup(); | |
1803 | break; | |
d00584b7 | 1804 | case 0x410: case 0x510: case 0x610: case 0x710: /* WMAC */ |
18c9b560 AZ |
1805 | wrd = (insn >> 12) & 0xf; |
1806 | rd0 = (insn >> 16) & 0xf; | |
1807 | rd1 = (insn >> 0) & 0xf; | |
1808 | gen_op_iwmmxt_movq_M0_wRn(rd0); | |
1809 | if (insn & (1 << 21)) | |
1810 | gen_op_iwmmxt_macsw_M0_wRn(rd1); | |
1811 | else | |
1812 | gen_op_iwmmxt_macuw_M0_wRn(rd1); | |
1813 | if (!(insn & (1 << 20))) { | |
e677137d PB |
1814 | iwmmxt_load_reg(cpu_V1, wrd); |
1815 | tcg_gen_add_i64(cpu_M0, cpu_M0, cpu_V1); | |
18c9b560 AZ |
1816 | } |
1817 | gen_op_iwmmxt_movq_wRn_M0(wrd); | |
1818 | gen_op_iwmmxt_set_mup(); | |
1819 | break; | |
d00584b7 | 1820 | case 0x006: case 0x406: case 0x806: case 0xc06: /* WCMPEQ */ |
18c9b560 AZ |
1821 | wrd = (insn >> 12) & 0xf; |
1822 | rd0 = (insn >> 16) & 0xf; | |
1823 | rd1 = (insn >> 0) & 0xf; | |
1824 | gen_op_iwmmxt_movq_M0_wRn(rd0); | |
1825 | switch ((insn >> 22) & 3) { | |
1826 | case 0: | |
1827 | gen_op_iwmmxt_cmpeqb_M0_wRn(rd1); | |
1828 | break; | |
1829 | case 1: | |
1830 | gen_op_iwmmxt_cmpeqw_M0_wRn(rd1); | |
1831 | break; | |
1832 | case 2: | |
1833 | gen_op_iwmmxt_cmpeql_M0_wRn(rd1); | |
1834 | break; | |
1835 | case 3: | |
1836 | return 1; | |
1837 | } | |
1838 | gen_op_iwmmxt_movq_wRn_M0(wrd); | |
1839 | gen_op_iwmmxt_set_mup(); | |
1840 | gen_op_iwmmxt_set_cup(); | |
1841 | break; | |
d00584b7 | 1842 | case 0x800: case 0x900: case 0xc00: case 0xd00: /* WAVG2 */ |
18c9b560 AZ |
1843 | wrd = (insn >> 12) & 0xf; |
1844 | rd0 = (insn >> 16) & 0xf; | |
1845 | rd1 = (insn >> 0) & 0xf; | |
1846 | gen_op_iwmmxt_movq_M0_wRn(rd0); | |
e677137d PB |
1847 | if (insn & (1 << 22)) { |
1848 | if (insn & (1 << 20)) | |
1849 | gen_op_iwmmxt_avgw1_M0_wRn(rd1); | |
1850 | else | |
1851 | gen_op_iwmmxt_avgw0_M0_wRn(rd1); | |
1852 | } else { | |
1853 | if (insn & (1 << 20)) | |
1854 | gen_op_iwmmxt_avgb1_M0_wRn(rd1); | |
1855 | else | |
1856 | gen_op_iwmmxt_avgb0_M0_wRn(rd1); | |
1857 | } | |
18c9b560 AZ |
1858 | gen_op_iwmmxt_movq_wRn_M0(wrd); |
1859 | gen_op_iwmmxt_set_mup(); | |
1860 | gen_op_iwmmxt_set_cup(); | |
1861 | break; | |
d00584b7 | 1862 | case 0x802: case 0x902: case 0xa02: case 0xb02: /* WALIGNR */ |
18c9b560 AZ |
1863 | wrd = (insn >> 12) & 0xf; |
1864 | rd0 = (insn >> 16) & 0xf; | |
1865 | rd1 = (insn >> 0) & 0xf; | |
1866 | gen_op_iwmmxt_movq_M0_wRn(rd0); | |
da6b5335 FN |
1867 | tmp = iwmmxt_load_creg(ARM_IWMMXT_wCGR0 + ((insn >> 20) & 3)); |
1868 | tcg_gen_andi_i32(tmp, tmp, 7); | |
1869 | iwmmxt_load_reg(cpu_V1, rd1); | |
1870 | gen_helper_iwmmxt_align(cpu_M0, cpu_M0, cpu_V1, tmp); | |
7d1b0095 | 1871 | tcg_temp_free_i32(tmp); |
18c9b560 AZ |
1872 | gen_op_iwmmxt_movq_wRn_M0(wrd); |
1873 | gen_op_iwmmxt_set_mup(); | |
1874 | break; | |
d00584b7 | 1875 | case 0x601: case 0x605: case 0x609: case 0x60d: /* TINSR */ |
da6b5335 FN |
1876 | if (((insn >> 6) & 3) == 3) |
1877 | return 1; | |
18c9b560 AZ |
1878 | rd = (insn >> 12) & 0xf; |
1879 | wrd = (insn >> 16) & 0xf; | |
da6b5335 | 1880 | tmp = load_reg(s, rd); |
18c9b560 AZ |
1881 | gen_op_iwmmxt_movq_M0_wRn(wrd); |
1882 | switch ((insn >> 6) & 3) { | |
1883 | case 0: | |
da6b5335 FN |
1884 | tmp2 = tcg_const_i32(0xff); |
1885 | tmp3 = tcg_const_i32((insn & 7) << 3); | |
18c9b560 AZ |
1886 | break; |
1887 | case 1: | |
da6b5335 FN |
1888 | tmp2 = tcg_const_i32(0xffff); |
1889 | tmp3 = tcg_const_i32((insn & 3) << 4); | |
18c9b560 AZ |
1890 | break; |
1891 | case 2: | |
da6b5335 FN |
1892 | tmp2 = tcg_const_i32(0xffffffff); |
1893 | tmp3 = tcg_const_i32((insn & 1) << 5); | |
18c9b560 | 1894 | break; |
da6b5335 | 1895 | default: |
f764718d RH |
1896 | tmp2 = NULL; |
1897 | tmp3 = NULL; | |
18c9b560 | 1898 | } |
da6b5335 | 1899 | gen_helper_iwmmxt_insr(cpu_M0, cpu_M0, tmp, tmp2, tmp3); |
39d5492a PM |
1900 | tcg_temp_free_i32(tmp3); |
1901 | tcg_temp_free_i32(tmp2); | |
7d1b0095 | 1902 | tcg_temp_free_i32(tmp); |
18c9b560 AZ |
1903 | gen_op_iwmmxt_movq_wRn_M0(wrd); |
1904 | gen_op_iwmmxt_set_mup(); | |
1905 | break; | |
d00584b7 | 1906 | case 0x107: case 0x507: case 0x907: case 0xd07: /* TEXTRM */ |
18c9b560 AZ |
1907 | rd = (insn >> 12) & 0xf; |
1908 | wrd = (insn >> 16) & 0xf; | |
da6b5335 | 1909 | if (rd == 15 || ((insn >> 22) & 3) == 3) |
18c9b560 AZ |
1910 | return 1; |
1911 | gen_op_iwmmxt_movq_M0_wRn(wrd); | |
7d1b0095 | 1912 | tmp = tcg_temp_new_i32(); |
18c9b560 AZ |
1913 | switch ((insn >> 22) & 3) { |
1914 | case 0: | |
da6b5335 | 1915 | tcg_gen_shri_i64(cpu_M0, cpu_M0, (insn & 7) << 3); |
ecc7b3aa | 1916 | tcg_gen_extrl_i64_i32(tmp, cpu_M0); |
da6b5335 FN |
1917 | if (insn & 8) { |
1918 | tcg_gen_ext8s_i32(tmp, tmp); | |
1919 | } else { | |
1920 | tcg_gen_andi_i32(tmp, tmp, 0xff); | |
18c9b560 AZ |
1921 | } |
1922 | break; | |
1923 | case 1: | |
da6b5335 | 1924 | tcg_gen_shri_i64(cpu_M0, cpu_M0, (insn & 3) << 4); |
ecc7b3aa | 1925 | tcg_gen_extrl_i64_i32(tmp, cpu_M0); |
da6b5335 FN |
1926 | if (insn & 8) { |
1927 | tcg_gen_ext16s_i32(tmp, tmp); | |
1928 | } else { | |
1929 | tcg_gen_andi_i32(tmp, tmp, 0xffff); | |
18c9b560 AZ |
1930 | } |
1931 | break; | |
1932 | case 2: | |
da6b5335 | 1933 | tcg_gen_shri_i64(cpu_M0, cpu_M0, (insn & 1) << 5); |
ecc7b3aa | 1934 | tcg_gen_extrl_i64_i32(tmp, cpu_M0); |
18c9b560 | 1935 | break; |
18c9b560 | 1936 | } |
da6b5335 | 1937 | store_reg(s, rd, tmp); |
18c9b560 | 1938 | break; |
d00584b7 | 1939 | case 0x117: case 0x517: case 0x917: case 0xd17: /* TEXTRC */ |
da6b5335 | 1940 | if ((insn & 0x000ff008) != 0x0003f000 || ((insn >> 22) & 3) == 3) |
18c9b560 | 1941 | return 1; |
da6b5335 | 1942 | tmp = iwmmxt_load_creg(ARM_IWMMXT_wCASF); |
18c9b560 AZ |
1943 | switch ((insn >> 22) & 3) { |
1944 | case 0: | |
da6b5335 | 1945 | tcg_gen_shri_i32(tmp, tmp, ((insn & 7) << 2) + 0); |
18c9b560 AZ |
1946 | break; |
1947 | case 1: | |
da6b5335 | 1948 | tcg_gen_shri_i32(tmp, tmp, ((insn & 3) << 3) + 4); |
18c9b560 AZ |
1949 | break; |
1950 | case 2: | |
da6b5335 | 1951 | tcg_gen_shri_i32(tmp, tmp, ((insn & 1) << 4) + 12); |
18c9b560 | 1952 | break; |
18c9b560 | 1953 | } |
da6b5335 FN |
1954 | tcg_gen_shli_i32(tmp, tmp, 28); |
1955 | gen_set_nzcv(tmp); | |
7d1b0095 | 1956 | tcg_temp_free_i32(tmp); |
18c9b560 | 1957 | break; |
d00584b7 | 1958 | case 0x401: case 0x405: case 0x409: case 0x40d: /* TBCST */ |
da6b5335 FN |
1959 | if (((insn >> 6) & 3) == 3) |
1960 | return 1; | |
18c9b560 AZ |
1961 | rd = (insn >> 12) & 0xf; |
1962 | wrd = (insn >> 16) & 0xf; | |
da6b5335 | 1963 | tmp = load_reg(s, rd); |
18c9b560 AZ |
1964 | switch ((insn >> 6) & 3) { |
1965 | case 0: | |
da6b5335 | 1966 | gen_helper_iwmmxt_bcstb(cpu_M0, tmp); |
18c9b560 AZ |
1967 | break; |
1968 | case 1: | |
da6b5335 | 1969 | gen_helper_iwmmxt_bcstw(cpu_M0, tmp); |
18c9b560 AZ |
1970 | break; |
1971 | case 2: | |
da6b5335 | 1972 | gen_helper_iwmmxt_bcstl(cpu_M0, tmp); |
18c9b560 | 1973 | break; |
18c9b560 | 1974 | } |
7d1b0095 | 1975 | tcg_temp_free_i32(tmp); |
18c9b560 AZ |
1976 | gen_op_iwmmxt_movq_wRn_M0(wrd); |
1977 | gen_op_iwmmxt_set_mup(); | |
1978 | break; | |
d00584b7 | 1979 | case 0x113: case 0x513: case 0x913: case 0xd13: /* TANDC */ |
da6b5335 | 1980 | if ((insn & 0x000ff00f) != 0x0003f000 || ((insn >> 22) & 3) == 3) |
18c9b560 | 1981 | return 1; |
da6b5335 | 1982 | tmp = iwmmxt_load_creg(ARM_IWMMXT_wCASF); |
7d1b0095 | 1983 | tmp2 = tcg_temp_new_i32(); |
da6b5335 | 1984 | tcg_gen_mov_i32(tmp2, tmp); |
18c9b560 AZ |
1985 | switch ((insn >> 22) & 3) { |
1986 | case 0: | |
1987 | for (i = 0; i < 7; i ++) { | |
da6b5335 FN |
1988 | tcg_gen_shli_i32(tmp2, tmp2, 4); |
1989 | tcg_gen_and_i32(tmp, tmp, tmp2); | |
18c9b560 AZ |
1990 | } |
1991 | break; | |
1992 | case 1: | |
1993 | for (i = 0; i < 3; i ++) { | |
da6b5335 FN |
1994 | tcg_gen_shli_i32(tmp2, tmp2, 8); |
1995 | tcg_gen_and_i32(tmp, tmp, tmp2); | |
18c9b560 AZ |
1996 | } |
1997 | break; | |
1998 | case 2: | |
da6b5335 FN |
1999 | tcg_gen_shli_i32(tmp2, tmp2, 16); |
2000 | tcg_gen_and_i32(tmp, tmp, tmp2); | |
18c9b560 | 2001 | break; |
18c9b560 | 2002 | } |
da6b5335 | 2003 | gen_set_nzcv(tmp); |
7d1b0095 PM |
2004 | tcg_temp_free_i32(tmp2); |
2005 | tcg_temp_free_i32(tmp); | |
18c9b560 | 2006 | break; |
d00584b7 | 2007 | case 0x01c: case 0x41c: case 0x81c: case 0xc1c: /* WACC */ |
18c9b560 AZ |
2008 | wrd = (insn >> 12) & 0xf; |
2009 | rd0 = (insn >> 16) & 0xf; | |
2010 | gen_op_iwmmxt_movq_M0_wRn(rd0); | |
2011 | switch ((insn >> 22) & 3) { | |
2012 | case 0: | |
e677137d | 2013 | gen_helper_iwmmxt_addcb(cpu_M0, cpu_M0); |
18c9b560 AZ |
2014 | break; |
2015 | case 1: | |
e677137d | 2016 | gen_helper_iwmmxt_addcw(cpu_M0, cpu_M0); |
18c9b560 AZ |
2017 | break; |
2018 | case 2: | |
e677137d | 2019 | gen_helper_iwmmxt_addcl(cpu_M0, cpu_M0); |
18c9b560 AZ |
2020 | break; |
2021 | case 3: | |
2022 | return 1; | |
2023 | } | |
2024 | gen_op_iwmmxt_movq_wRn_M0(wrd); | |
2025 | gen_op_iwmmxt_set_mup(); | |
2026 | break; | |
d00584b7 | 2027 | case 0x115: case 0x515: case 0x915: case 0xd15: /* TORC */ |
da6b5335 | 2028 | if ((insn & 0x000ff00f) != 0x0003f000 || ((insn >> 22) & 3) == 3) |
18c9b560 | 2029 | return 1; |
da6b5335 | 2030 | tmp = iwmmxt_load_creg(ARM_IWMMXT_wCASF); |
7d1b0095 | 2031 | tmp2 = tcg_temp_new_i32(); |
da6b5335 | 2032 | tcg_gen_mov_i32(tmp2, tmp); |
18c9b560 AZ |
2033 | switch ((insn >> 22) & 3) { |
2034 | case 0: | |
2035 | for (i = 0; i < 7; i ++) { | |
da6b5335 FN |
2036 | tcg_gen_shli_i32(tmp2, tmp2, 4); |
2037 | tcg_gen_or_i32(tmp, tmp, tmp2); | |
18c9b560 AZ |
2038 | } |
2039 | break; | |
2040 | case 1: | |
2041 | for (i = 0; i < 3; i ++) { | |
da6b5335 FN |
2042 | tcg_gen_shli_i32(tmp2, tmp2, 8); |
2043 | tcg_gen_or_i32(tmp, tmp, tmp2); | |
18c9b560 AZ |
2044 | } |
2045 | break; | |
2046 | case 2: | |
da6b5335 FN |
2047 | tcg_gen_shli_i32(tmp2, tmp2, 16); |
2048 | tcg_gen_or_i32(tmp, tmp, tmp2); | |
18c9b560 | 2049 | break; |
18c9b560 | 2050 | } |
da6b5335 | 2051 | gen_set_nzcv(tmp); |
7d1b0095 PM |
2052 | tcg_temp_free_i32(tmp2); |
2053 | tcg_temp_free_i32(tmp); | |
18c9b560 | 2054 | break; |
d00584b7 | 2055 | case 0x103: case 0x503: case 0x903: case 0xd03: /* TMOVMSK */ |
18c9b560 AZ |
2056 | rd = (insn >> 12) & 0xf; |
2057 | rd0 = (insn >> 16) & 0xf; | |
da6b5335 | 2058 | if ((insn & 0xf) != 0 || ((insn >> 22) & 3) == 3) |
18c9b560 AZ |
2059 | return 1; |
2060 | gen_op_iwmmxt_movq_M0_wRn(rd0); | |
7d1b0095 | 2061 | tmp = tcg_temp_new_i32(); |
18c9b560 AZ |
2062 | switch ((insn >> 22) & 3) { |
2063 | case 0: | |
da6b5335 | 2064 | gen_helper_iwmmxt_msbb(tmp, cpu_M0); |
18c9b560 AZ |
2065 | break; |
2066 | case 1: | |
da6b5335 | 2067 | gen_helper_iwmmxt_msbw(tmp, cpu_M0); |
18c9b560 AZ |
2068 | break; |
2069 | case 2: | |
da6b5335 | 2070 | gen_helper_iwmmxt_msbl(tmp, cpu_M0); |
18c9b560 | 2071 | break; |
18c9b560 | 2072 | } |
da6b5335 | 2073 | store_reg(s, rd, tmp); |
18c9b560 | 2074 | break; |
d00584b7 | 2075 | case 0x106: case 0x306: case 0x506: case 0x706: /* WCMPGT */ |
18c9b560 AZ |
2076 | case 0x906: case 0xb06: case 0xd06: case 0xf06: |
2077 | wrd = (insn >> 12) & 0xf; | |
2078 | rd0 = (insn >> 16) & 0xf; | |
2079 | rd1 = (insn >> 0) & 0xf; | |
2080 | gen_op_iwmmxt_movq_M0_wRn(rd0); | |
2081 | switch ((insn >> 22) & 3) { | |
2082 | case 0: | |
2083 | if (insn & (1 << 21)) | |
2084 | gen_op_iwmmxt_cmpgtsb_M0_wRn(rd1); | |
2085 | else | |
2086 | gen_op_iwmmxt_cmpgtub_M0_wRn(rd1); | |
2087 | break; | |
2088 | case 1: | |
2089 | if (insn & (1 << 21)) | |
2090 | gen_op_iwmmxt_cmpgtsw_M0_wRn(rd1); | |
2091 | else | |
2092 | gen_op_iwmmxt_cmpgtuw_M0_wRn(rd1); | |
2093 | break; | |
2094 | case 2: | |
2095 | if (insn & (1 << 21)) | |
2096 | gen_op_iwmmxt_cmpgtsl_M0_wRn(rd1); | |
2097 | else | |
2098 | gen_op_iwmmxt_cmpgtul_M0_wRn(rd1); | |
2099 | break; | |
2100 | case 3: | |
2101 | return 1; | |
2102 | } | |
2103 | gen_op_iwmmxt_movq_wRn_M0(wrd); | |
2104 | gen_op_iwmmxt_set_mup(); | |
2105 | gen_op_iwmmxt_set_cup(); | |
2106 | break; | |
d00584b7 | 2107 | case 0x00e: case 0x20e: case 0x40e: case 0x60e: /* WUNPCKEL */ |
18c9b560 AZ |
2108 | case 0x80e: case 0xa0e: case 0xc0e: case 0xe0e: |
2109 | wrd = (insn >> 12) & 0xf; | |
2110 | rd0 = (insn >> 16) & 0xf; | |
2111 | gen_op_iwmmxt_movq_M0_wRn(rd0); | |
2112 | switch ((insn >> 22) & 3) { | |
2113 | case 0: | |
2114 | if (insn & (1 << 21)) | |
2115 | gen_op_iwmmxt_unpacklsb_M0(); | |
2116 | else | |
2117 | gen_op_iwmmxt_unpacklub_M0(); | |
2118 | break; | |
2119 | case 1: | |
2120 | if (insn & (1 << 21)) | |
2121 | gen_op_iwmmxt_unpacklsw_M0(); | |
2122 | else | |
2123 | gen_op_iwmmxt_unpackluw_M0(); | |
2124 | break; | |
2125 | case 2: | |
2126 | if (insn & (1 << 21)) | |
2127 | gen_op_iwmmxt_unpacklsl_M0(); | |
2128 | else | |
2129 | gen_op_iwmmxt_unpacklul_M0(); | |
2130 | break; | |
2131 | case 3: | |
2132 | return 1; | |
2133 | } | |
2134 | gen_op_iwmmxt_movq_wRn_M0(wrd); | |
2135 | gen_op_iwmmxt_set_mup(); | |
2136 | gen_op_iwmmxt_set_cup(); | |
2137 | break; | |
d00584b7 | 2138 | case 0x00c: case 0x20c: case 0x40c: case 0x60c: /* WUNPCKEH */ |
18c9b560 AZ |
2139 | case 0x80c: case 0xa0c: case 0xc0c: case 0xe0c: |
2140 | wrd = (insn >> 12) & 0xf; | |
2141 | rd0 = (insn >> 16) & 0xf; | |
2142 | gen_op_iwmmxt_movq_M0_wRn(rd0); | |
2143 | switch ((insn >> 22) & 3) { | |
2144 | case 0: | |
2145 | if (insn & (1 << 21)) | |
2146 | gen_op_iwmmxt_unpackhsb_M0(); | |
2147 | else | |
2148 | gen_op_iwmmxt_unpackhub_M0(); | |
2149 | break; | |
2150 | case 1: | |
2151 | if (insn & (1 << 21)) | |
2152 | gen_op_iwmmxt_unpackhsw_M0(); | |
2153 | else | |
2154 | gen_op_iwmmxt_unpackhuw_M0(); | |
2155 | break; | |
2156 | case 2: | |
2157 | if (insn & (1 << 21)) | |
2158 | gen_op_iwmmxt_unpackhsl_M0(); | |
2159 | else | |
2160 | gen_op_iwmmxt_unpackhul_M0(); | |
2161 | break; | |
2162 | case 3: | |
2163 | return 1; | |
2164 | } | |
2165 | gen_op_iwmmxt_movq_wRn_M0(wrd); | |
2166 | gen_op_iwmmxt_set_mup(); | |
2167 | gen_op_iwmmxt_set_cup(); | |
2168 | break; | |
d00584b7 | 2169 | case 0x204: case 0x604: case 0xa04: case 0xe04: /* WSRL */ |
18c9b560 | 2170 | case 0x214: case 0x614: case 0xa14: case 0xe14: |
da6b5335 FN |
2171 | if (((insn >> 22) & 3) == 0) |
2172 | return 1; | |
18c9b560 AZ |
2173 | wrd = (insn >> 12) & 0xf; |
2174 | rd0 = (insn >> 16) & 0xf; | |
2175 | gen_op_iwmmxt_movq_M0_wRn(rd0); | |
7d1b0095 | 2176 | tmp = tcg_temp_new_i32(); |
da6b5335 | 2177 | if (gen_iwmmxt_shift(insn, 0xff, tmp)) { |
7d1b0095 | 2178 | tcg_temp_free_i32(tmp); |
18c9b560 | 2179 | return 1; |
da6b5335 | 2180 | } |
18c9b560 | 2181 | switch ((insn >> 22) & 3) { |
18c9b560 | 2182 | case 1: |
477955bd | 2183 | gen_helper_iwmmxt_srlw(cpu_M0, cpu_env, cpu_M0, tmp); |
18c9b560 AZ |
2184 | break; |
2185 | case 2: | |
477955bd | 2186 | gen_helper_iwmmxt_srll(cpu_M0, cpu_env, cpu_M0, tmp); |
18c9b560 AZ |
2187 | break; |
2188 | case 3: | |
477955bd | 2189 | gen_helper_iwmmxt_srlq(cpu_M0, cpu_env, cpu_M0, tmp); |
18c9b560 AZ |
2190 | break; |
2191 | } | |
7d1b0095 | 2192 | tcg_temp_free_i32(tmp); |
18c9b560 AZ |
2193 | gen_op_iwmmxt_movq_wRn_M0(wrd); |
2194 | gen_op_iwmmxt_set_mup(); | |
2195 | gen_op_iwmmxt_set_cup(); | |
2196 | break; | |
d00584b7 | 2197 | case 0x004: case 0x404: case 0x804: case 0xc04: /* WSRA */ |
18c9b560 | 2198 | case 0x014: case 0x414: case 0x814: case 0xc14: |
da6b5335 FN |
2199 | if (((insn >> 22) & 3) == 0) |
2200 | return 1; | |
18c9b560 AZ |
2201 | wrd = (insn >> 12) & 0xf; |
2202 | rd0 = (insn >> 16) & 0xf; | |
2203 | gen_op_iwmmxt_movq_M0_wRn(rd0); | |
7d1b0095 | 2204 | tmp = tcg_temp_new_i32(); |
da6b5335 | 2205 | if (gen_iwmmxt_shift(insn, 0xff, tmp)) { |
7d1b0095 | 2206 | tcg_temp_free_i32(tmp); |
18c9b560 | 2207 | return 1; |
da6b5335 | 2208 | } |
18c9b560 | 2209 | switch ((insn >> 22) & 3) { |
18c9b560 | 2210 | case 1: |
477955bd | 2211 | gen_helper_iwmmxt_sraw(cpu_M0, cpu_env, cpu_M0, tmp); |
18c9b560 AZ |
2212 | break; |
2213 | case 2: | |
477955bd | 2214 | gen_helper_iwmmxt_sral(cpu_M0, cpu_env, cpu_M0, tmp); |
18c9b560 AZ |
2215 | break; |
2216 | case 3: | |
477955bd | 2217 | gen_helper_iwmmxt_sraq(cpu_M0, cpu_env, cpu_M0, tmp); |
18c9b560 AZ |
2218 | break; |
2219 | } | |
7d1b0095 | 2220 | tcg_temp_free_i32(tmp); |
18c9b560 AZ |
2221 | gen_op_iwmmxt_movq_wRn_M0(wrd); |
2222 | gen_op_iwmmxt_set_mup(); | |
2223 | gen_op_iwmmxt_set_cup(); | |
2224 | break; | |
d00584b7 | 2225 | case 0x104: case 0x504: case 0x904: case 0xd04: /* WSLL */ |
18c9b560 | 2226 | case 0x114: case 0x514: case 0x914: case 0xd14: |
da6b5335 FN |
2227 | if (((insn >> 22) & 3) == 0) |
2228 | return 1; | |
18c9b560 AZ |
2229 | wrd = (insn >> 12) & 0xf; |
2230 | rd0 = (insn >> 16) & 0xf; | |
2231 | gen_op_iwmmxt_movq_M0_wRn(rd0); | |
7d1b0095 | 2232 | tmp = tcg_temp_new_i32(); |
da6b5335 | 2233 | if (gen_iwmmxt_shift(insn, 0xff, tmp)) { |
7d1b0095 | 2234 | tcg_temp_free_i32(tmp); |
18c9b560 | 2235 | return 1; |
da6b5335 | 2236 | } |
18c9b560 | 2237 | switch ((insn >> 22) & 3) { |
18c9b560 | 2238 | case 1: |
477955bd | 2239 | gen_helper_iwmmxt_sllw(cpu_M0, cpu_env, cpu_M0, tmp); |
18c9b560 AZ |
2240 | break; |
2241 | case 2: | |
477955bd | 2242 | gen_helper_iwmmxt_slll(cpu_M0, cpu_env, cpu_M0, tmp); |
18c9b560 AZ |
2243 | break; |
2244 | case 3: | |
477955bd | 2245 | gen_helper_iwmmxt_sllq(cpu_M0, cpu_env, cpu_M0, tmp); |
18c9b560 AZ |
2246 | break; |
2247 | } | |
7d1b0095 | 2248 | tcg_temp_free_i32(tmp); |
18c9b560 AZ |
2249 | gen_op_iwmmxt_movq_wRn_M0(wrd); |
2250 | gen_op_iwmmxt_set_mup(); | |
2251 | gen_op_iwmmxt_set_cup(); | |
2252 | break; | |
d00584b7 | 2253 | case 0x304: case 0x704: case 0xb04: case 0xf04: /* WROR */ |
18c9b560 | 2254 | case 0x314: case 0x714: case 0xb14: case 0xf14: |
da6b5335 FN |
2255 | if (((insn >> 22) & 3) == 0) |
2256 | return 1; | |
18c9b560 AZ |
2257 | wrd = (insn >> 12) & 0xf; |
2258 | rd0 = (insn >> 16) & 0xf; | |
2259 | gen_op_iwmmxt_movq_M0_wRn(rd0); | |
7d1b0095 | 2260 | tmp = tcg_temp_new_i32(); |
18c9b560 | 2261 | switch ((insn >> 22) & 3) { |
18c9b560 | 2262 | case 1: |
da6b5335 | 2263 | if (gen_iwmmxt_shift(insn, 0xf, tmp)) { |
7d1b0095 | 2264 | tcg_temp_free_i32(tmp); |
18c9b560 | 2265 | return 1; |
da6b5335 | 2266 | } |
477955bd | 2267 | gen_helper_iwmmxt_rorw(cpu_M0, cpu_env, cpu_M0, tmp); |
18c9b560 AZ |
2268 | break; |
2269 | case 2: | |
da6b5335 | 2270 | if (gen_iwmmxt_shift(insn, 0x1f, tmp)) { |
7d1b0095 | 2271 | tcg_temp_free_i32(tmp); |
18c9b560 | 2272 | return 1; |
da6b5335 | 2273 | } |
477955bd | 2274 | gen_helper_iwmmxt_rorl(cpu_M0, cpu_env, cpu_M0, tmp); |
18c9b560 AZ |
2275 | break; |
2276 | case 3: | |
da6b5335 | 2277 | if (gen_iwmmxt_shift(insn, 0x3f, tmp)) { |
7d1b0095 | 2278 | tcg_temp_free_i32(tmp); |
18c9b560 | 2279 | return 1; |
da6b5335 | 2280 | } |
477955bd | 2281 | gen_helper_iwmmxt_rorq(cpu_M0, cpu_env, cpu_M0, tmp); |
18c9b560 AZ |
2282 | break; |
2283 | } | |
7d1b0095 | 2284 | tcg_temp_free_i32(tmp); |
18c9b560 AZ |
2285 | gen_op_iwmmxt_movq_wRn_M0(wrd); |
2286 | gen_op_iwmmxt_set_mup(); | |
2287 | gen_op_iwmmxt_set_cup(); | |
2288 | break; | |
d00584b7 | 2289 | case 0x116: case 0x316: case 0x516: case 0x716: /* WMIN */ |
18c9b560 AZ |
2290 | case 0x916: case 0xb16: case 0xd16: case 0xf16: |
2291 | wrd = (insn >> 12) & 0xf; | |
2292 | rd0 = (insn >> 16) & 0xf; | |
2293 | rd1 = (insn >> 0) & 0xf; | |
2294 | gen_op_iwmmxt_movq_M0_wRn(rd0); | |
2295 | switch ((insn >> 22) & 3) { | |
2296 | case 0: | |
2297 | if (insn & (1 << 21)) | |
2298 | gen_op_iwmmxt_minsb_M0_wRn(rd1); | |
2299 | else | |
2300 | gen_op_iwmmxt_minub_M0_wRn(rd1); | |
2301 | break; | |
2302 | case 1: | |
2303 | if (insn & (1 << 21)) | |
2304 | gen_op_iwmmxt_minsw_M0_wRn(rd1); | |
2305 | else | |
2306 | gen_op_iwmmxt_minuw_M0_wRn(rd1); | |
2307 | break; | |
2308 | case 2: | |
2309 | if (insn & (1 << 21)) | |
2310 | gen_op_iwmmxt_minsl_M0_wRn(rd1); | |
2311 | else | |
2312 | gen_op_iwmmxt_minul_M0_wRn(rd1); | |
2313 | break; | |
2314 | case 3: | |
2315 | return 1; | |
2316 | } | |
2317 | gen_op_iwmmxt_movq_wRn_M0(wrd); | |
2318 | gen_op_iwmmxt_set_mup(); | |
2319 | break; | |
d00584b7 | 2320 | case 0x016: case 0x216: case 0x416: case 0x616: /* WMAX */ |
18c9b560 AZ |
2321 | case 0x816: case 0xa16: case 0xc16: case 0xe16: |
2322 | wrd = (insn >> 12) & 0xf; | |
2323 | rd0 = (insn >> 16) & 0xf; | |
2324 | rd1 = (insn >> 0) & 0xf; | |
2325 | gen_op_iwmmxt_movq_M0_wRn(rd0); | |
2326 | switch ((insn >> 22) & 3) { | |
2327 | case 0: | |
2328 | if (insn & (1 << 21)) | |
2329 | gen_op_iwmmxt_maxsb_M0_wRn(rd1); | |
2330 | else | |
2331 | gen_op_iwmmxt_maxub_M0_wRn(rd1); | |
2332 | break; | |
2333 | case 1: | |
2334 | if (insn & (1 << 21)) | |
2335 | gen_op_iwmmxt_maxsw_M0_wRn(rd1); | |
2336 | else | |
2337 | gen_op_iwmmxt_maxuw_M0_wRn(rd1); | |
2338 | break; | |
2339 | case 2: | |
2340 | if (insn & (1 << 21)) | |
2341 | gen_op_iwmmxt_maxsl_M0_wRn(rd1); | |
2342 | else | |
2343 | gen_op_iwmmxt_maxul_M0_wRn(rd1); | |
2344 | break; | |
2345 | case 3: | |
2346 | return 1; | |
2347 | } | |
2348 | gen_op_iwmmxt_movq_wRn_M0(wrd); | |
2349 | gen_op_iwmmxt_set_mup(); | |
2350 | break; | |
d00584b7 | 2351 | case 0x002: case 0x102: case 0x202: case 0x302: /* WALIGNI */ |
18c9b560 AZ |
2352 | case 0x402: case 0x502: case 0x602: case 0x702: |
2353 | wrd = (insn >> 12) & 0xf; | |
2354 | rd0 = (insn >> 16) & 0xf; | |
2355 | rd1 = (insn >> 0) & 0xf; | |
2356 | gen_op_iwmmxt_movq_M0_wRn(rd0); | |
da6b5335 FN |
2357 | tmp = tcg_const_i32((insn >> 20) & 3); |
2358 | iwmmxt_load_reg(cpu_V1, rd1); | |
2359 | gen_helper_iwmmxt_align(cpu_M0, cpu_M0, cpu_V1, tmp); | |
39d5492a | 2360 | tcg_temp_free_i32(tmp); |
18c9b560 AZ |
2361 | gen_op_iwmmxt_movq_wRn_M0(wrd); |
2362 | gen_op_iwmmxt_set_mup(); | |
2363 | break; | |
d00584b7 | 2364 | case 0x01a: case 0x11a: case 0x21a: case 0x31a: /* WSUB */ |
18c9b560 AZ |
2365 | case 0x41a: case 0x51a: case 0x61a: case 0x71a: |
2366 | case 0x81a: case 0x91a: case 0xa1a: case 0xb1a: | |
2367 | case 0xc1a: case 0xd1a: case 0xe1a: case 0xf1a: | |
2368 | wrd = (insn >> 12) & 0xf; | |
2369 | rd0 = (insn >> 16) & 0xf; | |
2370 | rd1 = (insn >> 0) & 0xf; | |
2371 | gen_op_iwmmxt_movq_M0_wRn(rd0); | |
2372 | switch ((insn >> 20) & 0xf) { | |
2373 | case 0x0: | |
2374 | gen_op_iwmmxt_subnb_M0_wRn(rd1); | |
2375 | break; | |
2376 | case 0x1: | |
2377 | gen_op_iwmmxt_subub_M0_wRn(rd1); | |
2378 | break; | |
2379 | case 0x3: | |
2380 | gen_op_iwmmxt_subsb_M0_wRn(rd1); | |
2381 | break; | |
2382 | case 0x4: | |
2383 | gen_op_iwmmxt_subnw_M0_wRn(rd1); | |
2384 | break; | |
2385 | case 0x5: | |
2386 | gen_op_iwmmxt_subuw_M0_wRn(rd1); | |
2387 | break; | |
2388 | case 0x7: | |
2389 | gen_op_iwmmxt_subsw_M0_wRn(rd1); | |
2390 | break; | |
2391 | case 0x8: | |
2392 | gen_op_iwmmxt_subnl_M0_wRn(rd1); | |
2393 | break; | |
2394 | case 0x9: | |
2395 | gen_op_iwmmxt_subul_M0_wRn(rd1); | |
2396 | break; | |
2397 | case 0xb: | |
2398 | gen_op_iwmmxt_subsl_M0_wRn(rd1); | |
2399 | break; | |
2400 | default: | |
2401 | return 1; | |
2402 | } | |
2403 | gen_op_iwmmxt_movq_wRn_M0(wrd); | |
2404 | gen_op_iwmmxt_set_mup(); | |
2405 | gen_op_iwmmxt_set_cup(); | |
2406 | break; | |
d00584b7 | 2407 | case 0x01e: case 0x11e: case 0x21e: case 0x31e: /* WSHUFH */ |
18c9b560 AZ |
2408 | case 0x41e: case 0x51e: case 0x61e: case 0x71e: |
2409 | case 0x81e: case 0x91e: case 0xa1e: case 0xb1e: | |
2410 | case 0xc1e: case 0xd1e: case 0xe1e: case 0xf1e: | |
2411 | wrd = (insn >> 12) & 0xf; | |
2412 | rd0 = (insn >> 16) & 0xf; | |
2413 | gen_op_iwmmxt_movq_M0_wRn(rd0); | |
da6b5335 | 2414 | tmp = tcg_const_i32(((insn >> 16) & 0xf0) | (insn & 0x0f)); |
477955bd | 2415 | gen_helper_iwmmxt_shufh(cpu_M0, cpu_env, cpu_M0, tmp); |
39d5492a | 2416 | tcg_temp_free_i32(tmp); |
18c9b560 AZ |
2417 | gen_op_iwmmxt_movq_wRn_M0(wrd); |
2418 | gen_op_iwmmxt_set_mup(); | |
2419 | gen_op_iwmmxt_set_cup(); | |
2420 | break; | |
d00584b7 | 2421 | case 0x018: case 0x118: case 0x218: case 0x318: /* WADD */ |
18c9b560 AZ |
2422 | case 0x418: case 0x518: case 0x618: case 0x718: |
2423 | case 0x818: case 0x918: case 0xa18: case 0xb18: | |
2424 | case 0xc18: case 0xd18: case 0xe18: case 0xf18: | |
2425 | wrd = (insn >> 12) & 0xf; | |
2426 | rd0 = (insn >> 16) & 0xf; | |
2427 | rd1 = (insn >> 0) & 0xf; | |
2428 | gen_op_iwmmxt_movq_M0_wRn(rd0); | |
2429 | switch ((insn >> 20) & 0xf) { | |
2430 | case 0x0: | |
2431 | gen_op_iwmmxt_addnb_M0_wRn(rd1); | |
2432 | break; | |
2433 | case 0x1: | |
2434 | gen_op_iwmmxt_addub_M0_wRn(rd1); | |
2435 | break; | |
2436 | case 0x3: | |
2437 | gen_op_iwmmxt_addsb_M0_wRn(rd1); | |
2438 | break; | |
2439 | case 0x4: | |
2440 | gen_op_iwmmxt_addnw_M0_wRn(rd1); | |
2441 | break; | |
2442 | case 0x5: | |
2443 | gen_op_iwmmxt_adduw_M0_wRn(rd1); | |
2444 | break; | |
2445 | case 0x7: | |
2446 | gen_op_iwmmxt_addsw_M0_wRn(rd1); | |
2447 | break; | |
2448 | case 0x8: | |
2449 | gen_op_iwmmxt_addnl_M0_wRn(rd1); | |
2450 | break; | |
2451 | case 0x9: | |
2452 | gen_op_iwmmxt_addul_M0_wRn(rd1); | |
2453 | break; | |
2454 | case 0xb: | |
2455 | gen_op_iwmmxt_addsl_M0_wRn(rd1); | |
2456 | break; | |
2457 | default: | |
2458 | return 1; | |
2459 | } | |
2460 | gen_op_iwmmxt_movq_wRn_M0(wrd); | |
2461 | gen_op_iwmmxt_set_mup(); | |
2462 | gen_op_iwmmxt_set_cup(); | |
2463 | break; | |
d00584b7 | 2464 | case 0x008: case 0x108: case 0x208: case 0x308: /* WPACK */ |
18c9b560 AZ |
2465 | case 0x408: case 0x508: case 0x608: case 0x708: |
2466 | case 0x808: case 0x908: case 0xa08: case 0xb08: | |
2467 | case 0xc08: case 0xd08: case 0xe08: case 0xf08: | |
da6b5335 FN |
2468 | if (!(insn & (1 << 20)) || ((insn >> 22) & 3) == 0) |
2469 | return 1; | |
18c9b560 AZ |
2470 | wrd = (insn >> 12) & 0xf; |
2471 | rd0 = (insn >> 16) & 0xf; | |
2472 | rd1 = (insn >> 0) & 0xf; | |
2473 | gen_op_iwmmxt_movq_M0_wRn(rd0); | |
18c9b560 | 2474 | switch ((insn >> 22) & 3) { |
18c9b560 AZ |
2475 | case 1: |
2476 | if (insn & (1 << 21)) | |
2477 | gen_op_iwmmxt_packsw_M0_wRn(rd1); | |
2478 | else | |
2479 | gen_op_iwmmxt_packuw_M0_wRn(rd1); | |
2480 | break; | |
2481 | case 2: | |
2482 | if (insn & (1 << 21)) | |
2483 | gen_op_iwmmxt_packsl_M0_wRn(rd1); | |
2484 | else | |
2485 | gen_op_iwmmxt_packul_M0_wRn(rd1); | |
2486 | break; | |
2487 | case 3: | |
2488 | if (insn & (1 << 21)) | |
2489 | gen_op_iwmmxt_packsq_M0_wRn(rd1); | |
2490 | else | |
2491 | gen_op_iwmmxt_packuq_M0_wRn(rd1); | |
2492 | break; | |
2493 | } | |
2494 | gen_op_iwmmxt_movq_wRn_M0(wrd); | |
2495 | gen_op_iwmmxt_set_mup(); | |
2496 | gen_op_iwmmxt_set_cup(); | |
2497 | break; | |
2498 | case 0x201: case 0x203: case 0x205: case 0x207: | |
2499 | case 0x209: case 0x20b: case 0x20d: case 0x20f: | |
2500 | case 0x211: case 0x213: case 0x215: case 0x217: | |
2501 | case 0x219: case 0x21b: case 0x21d: case 0x21f: | |
2502 | wrd = (insn >> 5) & 0xf; | |
2503 | rd0 = (insn >> 12) & 0xf; | |
2504 | rd1 = (insn >> 0) & 0xf; | |
2505 | if (rd0 == 0xf || rd1 == 0xf) | |
2506 | return 1; | |
2507 | gen_op_iwmmxt_movq_M0_wRn(wrd); | |
da6b5335 FN |
2508 | tmp = load_reg(s, rd0); |
2509 | tmp2 = load_reg(s, rd1); | |
18c9b560 | 2510 | switch ((insn >> 16) & 0xf) { |
d00584b7 | 2511 | case 0x0: /* TMIA */ |
da6b5335 | 2512 | gen_helper_iwmmxt_muladdsl(cpu_M0, cpu_M0, tmp, tmp2); |
18c9b560 | 2513 | break; |
d00584b7 | 2514 | case 0x8: /* TMIAPH */ |
da6b5335 | 2515 | gen_helper_iwmmxt_muladdsw(cpu_M0, cpu_M0, tmp, tmp2); |
18c9b560 | 2516 | break; |
d00584b7 | 2517 | case 0xc: case 0xd: case 0xe: case 0xf: /* TMIAxy */ |
18c9b560 | 2518 | if (insn & (1 << 16)) |
da6b5335 | 2519 | tcg_gen_shri_i32(tmp, tmp, 16); |
18c9b560 | 2520 | if (insn & (1 << 17)) |
da6b5335 FN |
2521 | tcg_gen_shri_i32(tmp2, tmp2, 16); |
2522 | gen_helper_iwmmxt_muladdswl(cpu_M0, cpu_M0, tmp, tmp2); | |
18c9b560 AZ |
2523 | break; |
2524 | default: | |
7d1b0095 PM |
2525 | tcg_temp_free_i32(tmp2); |
2526 | tcg_temp_free_i32(tmp); | |
18c9b560 AZ |
2527 | return 1; |
2528 | } | |
7d1b0095 PM |
2529 | tcg_temp_free_i32(tmp2); |
2530 | tcg_temp_free_i32(tmp); | |
18c9b560 AZ |
2531 | gen_op_iwmmxt_movq_wRn_M0(wrd); |
2532 | gen_op_iwmmxt_set_mup(); | |
2533 | break; | |
2534 | default: | |
2535 | return 1; | |
2536 | } | |
2537 | ||
2538 | return 0; | |
2539 | } | |
2540 | ||
a1c7273b | 2541 | /* Disassemble an XScale DSP instruction. Returns nonzero if an error occurred |
18c9b560 | 2542 | (ie. an undefined instruction). */ |
7dcc1f89 | 2543 | static int disas_dsp_insn(DisasContext *s, uint32_t insn) |
18c9b560 AZ |
2544 | { |
2545 | int acc, rd0, rd1, rdhi, rdlo; | |
39d5492a | 2546 | TCGv_i32 tmp, tmp2; |
18c9b560 AZ |
2547 | |
2548 | if ((insn & 0x0ff00f10) == 0x0e200010) { | |
2549 | /* Multiply with Internal Accumulate Format */ | |
2550 | rd0 = (insn >> 12) & 0xf; | |
2551 | rd1 = insn & 0xf; | |
2552 | acc = (insn >> 5) & 7; | |
2553 | ||
2554 | if (acc != 0) | |
2555 | return 1; | |
2556 | ||
3a554c0f FN |
2557 | tmp = load_reg(s, rd0); |
2558 | tmp2 = load_reg(s, rd1); | |
18c9b560 | 2559 | switch ((insn >> 16) & 0xf) { |
d00584b7 | 2560 | case 0x0: /* MIA */ |
3a554c0f | 2561 | gen_helper_iwmmxt_muladdsl(cpu_M0, cpu_M0, tmp, tmp2); |
18c9b560 | 2562 | break; |
d00584b7 | 2563 | case 0x8: /* MIAPH */ |
3a554c0f | 2564 | gen_helper_iwmmxt_muladdsw(cpu_M0, cpu_M0, tmp, tmp2); |
18c9b560 | 2565 | break; |
d00584b7 PM |
2566 | case 0xc: /* MIABB */ |
2567 | case 0xd: /* MIABT */ | |
2568 | case 0xe: /* MIATB */ | |
2569 | case 0xf: /* MIATT */ | |
18c9b560 | 2570 | if (insn & (1 << 16)) |
3a554c0f | 2571 | tcg_gen_shri_i32(tmp, tmp, 16); |
18c9b560 | 2572 | if (insn & (1 << 17)) |
3a554c0f FN |
2573 | tcg_gen_shri_i32(tmp2, tmp2, 16); |
2574 | gen_helper_iwmmxt_muladdswl(cpu_M0, cpu_M0, tmp, tmp2); | |
18c9b560 AZ |
2575 | break; |
2576 | default: | |
2577 | return 1; | |
2578 | } | |
7d1b0095 PM |
2579 | tcg_temp_free_i32(tmp2); |
2580 | tcg_temp_free_i32(tmp); | |
18c9b560 AZ |
2581 | |
2582 | gen_op_iwmmxt_movq_wRn_M0(acc); | |
2583 | return 0; | |
2584 | } | |
2585 | ||
2586 | if ((insn & 0x0fe00ff8) == 0x0c400000) { | |
2587 | /* Internal Accumulator Access Format */ | |
2588 | rdhi = (insn >> 16) & 0xf; | |
2589 | rdlo = (insn >> 12) & 0xf; | |
2590 | acc = insn & 7; | |
2591 | ||
2592 | if (acc != 0) | |
2593 | return 1; | |
2594 | ||
d00584b7 | 2595 | if (insn & ARM_CP_RW_BIT) { /* MRA */ |
3a554c0f | 2596 | iwmmxt_load_reg(cpu_V0, acc); |
ecc7b3aa | 2597 | tcg_gen_extrl_i64_i32(cpu_R[rdlo], cpu_V0); |
664b7e3b | 2598 | tcg_gen_extrh_i64_i32(cpu_R[rdhi], cpu_V0); |
3a554c0f | 2599 | tcg_gen_andi_i32(cpu_R[rdhi], cpu_R[rdhi], (1 << (40 - 32)) - 1); |
d00584b7 | 2600 | } else { /* MAR */ |
3a554c0f FN |
2601 | tcg_gen_concat_i32_i64(cpu_V0, cpu_R[rdlo], cpu_R[rdhi]); |
2602 | iwmmxt_store_reg(cpu_V0, acc); | |
18c9b560 AZ |
2603 | } |
2604 | return 0; | |
2605 | } | |
2606 | ||
2607 | return 1; | |
2608 | } | |
2609 | ||
9ee6e8bb PB |
2610 | #define VFP_REG_SHR(x, n) (((n) > 0) ? (x) >> (n) : (x) << -(n)) |
2611 | #define VFP_SREG(insn, bigbit, smallbit) \ | |
2612 | ((VFP_REG_SHR(insn, bigbit - 1) & 0x1e) | (((insn) >> (smallbit)) & 1)) | |
2613 | #define VFP_DREG(reg, insn, bigbit, smallbit) do { \ | |
d614a513 | 2614 | if (arm_dc_feature(s, ARM_FEATURE_VFP3)) { \ |
9ee6e8bb PB |
2615 | reg = (((insn) >> (bigbit)) & 0x0f) \ |
2616 | | (((insn) >> ((smallbit) - 4)) & 0x10); \ | |
2617 | } else { \ | |
2618 | if (insn & (1 << (smallbit))) \ | |
2619 | return 1; \ | |
2620 | reg = ((insn) >> (bigbit)) & 0x0f; \ | |
2621 | }} while (0) | |
2622 | ||
2623 | #define VFP_SREG_D(insn) VFP_SREG(insn, 12, 22) | |
2624 | #define VFP_DREG_D(reg, insn) VFP_DREG(reg, insn, 12, 22) | |
2625 | #define VFP_SREG_N(insn) VFP_SREG(insn, 16, 7) | |
2626 | #define VFP_DREG_N(reg, insn) VFP_DREG(reg, insn, 16, 7) | |
2627 | #define VFP_SREG_M(insn) VFP_SREG(insn, 0, 5) | |
2628 | #define VFP_DREG_M(reg, insn) VFP_DREG(reg, insn, 0, 5) | |
2629 | ||
39d5492a | 2630 | static void gen_neon_dup_low16(TCGv_i32 var) |
ad69471c | 2631 | { |
39d5492a | 2632 | TCGv_i32 tmp = tcg_temp_new_i32(); |
86831435 | 2633 | tcg_gen_ext16u_i32(var, var); |
ad69471c PB |
2634 | tcg_gen_shli_i32(tmp, var, 16); |
2635 | tcg_gen_or_i32(var, var, tmp); | |
7d1b0095 | 2636 | tcg_temp_free_i32(tmp); |
ad69471c PB |
2637 | } |
2638 | ||
39d5492a | 2639 | static void gen_neon_dup_high16(TCGv_i32 var) |
ad69471c | 2640 | { |
39d5492a | 2641 | TCGv_i32 tmp = tcg_temp_new_i32(); |
ad69471c PB |
2642 | tcg_gen_andi_i32(var, var, 0xffff0000); |
2643 | tcg_gen_shri_i32(tmp, var, 16); | |
2644 | tcg_gen_or_i32(var, var, tmp); | |
7d1b0095 | 2645 | tcg_temp_free_i32(tmp); |
ad69471c PB |
2646 | } |
2647 | ||
06db8196 PM |
2648 | /* |
2649 | * Disassemble a VFP instruction. Returns nonzero if an error occurred | |
2650 | * (ie. an undefined instruction). | |
2651 | */ | |
7dcc1f89 | 2652 | static int disas_vfp_insn(DisasContext *s, uint32_t insn) |
b7bcbe95 | 2653 | { |
d614a513 | 2654 | if (!arm_dc_feature(s, ARM_FEATURE_VFP)) { |
40f137e1 | 2655 | return 1; |
d614a513 | 2656 | } |
40f137e1 | 2657 | |
78e138bc PM |
2658 | /* |
2659 | * If the decodetree decoder handles this insn it will always | |
2660 | * emit code to either execute the insn or generate an appropriate | |
2661 | * exception; so we don't need to ever return non-zero to tell | |
2662 | * the calling code to emit an UNDEF exception. | |
2663 | */ | |
2664 | if (extract32(insn, 28, 4) == 0xf) { | |
2665 | if (disas_vfp_uncond(s, insn)) { | |
2666 | return 0; | |
2667 | } | |
2668 | } else { | |
2669 | if (disas_vfp(s, insn)) { | |
2670 | return 0; | |
2671 | } | |
2672 | } | |
3111bfc2 PM |
2673 | /* If the decodetree decoder didn't handle this insn, it must be UNDEF */ |
2674 | return 1; | |
b7bcbe95 FB |
2675 | } |
2676 | ||
90aa39a1 | 2677 | static inline bool use_goto_tb(DisasContext *s, target_ulong dest) |
c53be334 | 2678 | { |
90aa39a1 | 2679 | #ifndef CONFIG_USER_ONLY |
dcba3a8d | 2680 | return (s->base.tb->pc & TARGET_PAGE_MASK) == (dest & TARGET_PAGE_MASK) || |
a0415916 | 2681 | ((s->base.pc_next - 1) & TARGET_PAGE_MASK) == (dest & TARGET_PAGE_MASK); |
90aa39a1 SF |
2682 | #else |
2683 | return true; | |
2684 | #endif | |
2685 | } | |
6e256c93 | 2686 | |
8a6b28c7 EC |
2687 | static void gen_goto_ptr(void) |
2688 | { | |
7f11636d | 2689 | tcg_gen_lookup_and_goto_ptr(); |
8a6b28c7 EC |
2690 | } |
2691 | ||
4cae8f56 AB |
2692 | /* This will end the TB but doesn't guarantee we'll return to |
2693 | * cpu_loop_exec. Any live exit_requests will be processed as we | |
2694 | * enter the next TB. | |
2695 | */ | |
8a6b28c7 | 2696 | static void gen_goto_tb(DisasContext *s, int n, target_ulong dest) |
90aa39a1 SF |
2697 | { |
2698 | if (use_goto_tb(s, dest)) { | |
57fec1fe | 2699 | tcg_gen_goto_tb(n); |
eaed129d | 2700 | gen_set_pc_im(s, dest); |
07ea28b4 | 2701 | tcg_gen_exit_tb(s->base.tb, n); |
6e256c93 | 2702 | } else { |
eaed129d | 2703 | gen_set_pc_im(s, dest); |
8a6b28c7 | 2704 | gen_goto_ptr(); |
6e256c93 | 2705 | } |
dcba3a8d | 2706 | s->base.is_jmp = DISAS_NORETURN; |
c53be334 FB |
2707 | } |
2708 | ||
8aaca4c0 FB |
2709 | static inline void gen_jmp (DisasContext *s, uint32_t dest) |
2710 | { | |
b636649f | 2711 | if (unlikely(is_singlestepping(s))) { |
8aaca4c0 | 2712 | /* An indirect jump so that we still trigger the debug exception. */ |
eac2f396 RH |
2713 | gen_set_pc_im(s, dest); |
2714 | s->base.is_jmp = DISAS_JUMP; | |
8aaca4c0 | 2715 | } else { |
6e256c93 | 2716 | gen_goto_tb(s, 0, dest); |
8aaca4c0 FB |
2717 | } |
2718 | } | |
2719 | ||
39d5492a | 2720 | static inline void gen_mulxy(TCGv_i32 t0, TCGv_i32 t1, int x, int y) |
b5ff1b31 | 2721 | { |
ee097184 | 2722 | if (x) |
d9ba4830 | 2723 | tcg_gen_sari_i32(t0, t0, 16); |
b5ff1b31 | 2724 | else |
d9ba4830 | 2725 | gen_sxth(t0); |
ee097184 | 2726 | if (y) |
d9ba4830 | 2727 | tcg_gen_sari_i32(t1, t1, 16); |
b5ff1b31 | 2728 | else |
d9ba4830 PB |
2729 | gen_sxth(t1); |
2730 | tcg_gen_mul_i32(t0, t0, t1); | |
b5ff1b31 FB |
2731 | } |
2732 | ||
2733 | /* Return the mask of PSR bits set by a MSR instruction. */ | |
7dcc1f89 PM |
2734 | static uint32_t msr_mask(DisasContext *s, int flags, int spsr) |
2735 | { | |
b5ff1b31 FB |
2736 | uint32_t mask; |
2737 | ||
2738 | mask = 0; | |
2739 | if (flags & (1 << 0)) | |
2740 | mask |= 0xff; | |
2741 | if (flags & (1 << 1)) | |
2742 | mask |= 0xff00; | |
2743 | if (flags & (1 << 2)) | |
2744 | mask |= 0xff0000; | |
2745 | if (flags & (1 << 3)) | |
2746 | mask |= 0xff000000; | |
9ee6e8bb | 2747 | |
2ae23e75 | 2748 | /* Mask out undefined bits. */ |
9ee6e8bb | 2749 | mask &= ~CPSR_RESERVED; |
d614a513 | 2750 | if (!arm_dc_feature(s, ARM_FEATURE_V4T)) { |
be5e7a76 | 2751 | mask &= ~CPSR_T; |
d614a513 PM |
2752 | } |
2753 | if (!arm_dc_feature(s, ARM_FEATURE_V5)) { | |
be5e7a76 | 2754 | mask &= ~CPSR_Q; /* V5TE in reality*/ |
d614a513 PM |
2755 | } |
2756 | if (!arm_dc_feature(s, ARM_FEATURE_V6)) { | |
e160c51c | 2757 | mask &= ~(CPSR_E | CPSR_GE); |
d614a513 PM |
2758 | } |
2759 | if (!arm_dc_feature(s, ARM_FEATURE_THUMB2)) { | |
e160c51c | 2760 | mask &= ~CPSR_IT; |
d614a513 | 2761 | } |
4051e12c PM |
2762 | /* Mask out execution state and reserved bits. */ |
2763 | if (!spsr) { | |
2764 | mask &= ~(CPSR_EXEC | CPSR_RESERVED); | |
2765 | } | |
b5ff1b31 FB |
2766 | /* Mask out privileged bits. */ |
2767 | if (IS_USER(s)) | |
9ee6e8bb | 2768 | mask &= CPSR_USER; |
b5ff1b31 FB |
2769 | return mask; |
2770 | } | |
2771 | ||
2fbac54b | 2772 | /* Returns nonzero if access to the PSR is not permitted. Marks t0 as dead. */ |
39d5492a | 2773 | static int gen_set_psr(DisasContext *s, uint32_t mask, int spsr, TCGv_i32 t0) |
b5ff1b31 | 2774 | { |
39d5492a | 2775 | TCGv_i32 tmp; |
b5ff1b31 FB |
2776 | if (spsr) { |
2777 | /* ??? This is also undefined in system mode. */ | |
2778 | if (IS_USER(s)) | |
2779 | return 1; | |
d9ba4830 PB |
2780 | |
2781 | tmp = load_cpu_field(spsr); | |
2782 | tcg_gen_andi_i32(tmp, tmp, ~mask); | |
2fbac54b FN |
2783 | tcg_gen_andi_i32(t0, t0, mask); |
2784 | tcg_gen_or_i32(tmp, tmp, t0); | |
d9ba4830 | 2785 | store_cpu_field(tmp, spsr); |
b5ff1b31 | 2786 | } else { |
2fbac54b | 2787 | gen_set_cpsr(t0, mask); |
b5ff1b31 | 2788 | } |
7d1b0095 | 2789 | tcg_temp_free_i32(t0); |
b5ff1b31 FB |
2790 | gen_lookup_tb(s); |
2791 | return 0; | |
2792 | } | |
2793 | ||
2fbac54b FN |
2794 | /* Returns nonzero if access to the PSR is not permitted. */ |
2795 | static int gen_set_psr_im(DisasContext *s, uint32_t mask, int spsr, uint32_t val) | |
2796 | { | |
39d5492a | 2797 | TCGv_i32 tmp; |
7d1b0095 | 2798 | tmp = tcg_temp_new_i32(); |
2fbac54b FN |
2799 | tcg_gen_movi_i32(tmp, val); |
2800 | return gen_set_psr(s, mask, spsr, tmp); | |
2801 | } | |
2802 | ||
8bfd0550 PM |
2803 | static bool msr_banked_access_decode(DisasContext *s, int r, int sysm, int rn, |
2804 | int *tgtmode, int *regno) | |
2805 | { | |
2806 | /* Decode the r and sysm fields of MSR/MRS banked accesses into | |
2807 | * the target mode and register number, and identify the various | |
2808 | * unpredictable cases. | |
2809 | * MSR (banked) and MRS (banked) are CONSTRAINED UNPREDICTABLE if: | |
2810 | * + executed in user mode | |
2811 | * + using R15 as the src/dest register | |
2812 | * + accessing an unimplemented register | |
2813 | * + accessing a register that's inaccessible at current PL/security state* | |
2814 | * + accessing a register that you could access with a different insn | |
2815 | * We choose to UNDEF in all these cases. | |
2816 | * Since we don't know which of the various AArch32 modes we are in | |
2817 | * we have to defer some checks to runtime. | |
2818 | * Accesses to Monitor mode registers from Secure EL1 (which implies | |
2819 | * that EL3 is AArch64) must trap to EL3. | |
2820 | * | |
2821 | * If the access checks fail this function will emit code to take | |
2822 | * an exception and return false. Otherwise it will return true, | |
2823 | * and set *tgtmode and *regno appropriately. | |
2824 | */ | |
2825 | int exc_target = default_exception_el(s); | |
2826 | ||
2827 | /* These instructions are present only in ARMv8, or in ARMv7 with the | |
2828 | * Virtualization Extensions. | |
2829 | */ | |
2830 | if (!arm_dc_feature(s, ARM_FEATURE_V8) && | |
2831 | !arm_dc_feature(s, ARM_FEATURE_EL2)) { | |
2832 | goto undef; | |
2833 | } | |
2834 | ||
2835 | if (IS_USER(s) || rn == 15) { | |
2836 | goto undef; | |
2837 | } | |
2838 | ||
2839 | /* The table in the v8 ARM ARM section F5.2.3 describes the encoding | |
2840 | * of registers into (r, sysm). | |
2841 | */ | |
2842 | if (r) { | |
2843 | /* SPSRs for other modes */ | |
2844 | switch (sysm) { | |
2845 | case 0xe: /* SPSR_fiq */ | |
2846 | *tgtmode = ARM_CPU_MODE_FIQ; | |
2847 | break; | |
2848 | case 0x10: /* SPSR_irq */ | |
2849 | *tgtmode = ARM_CPU_MODE_IRQ; | |
2850 | break; | |
2851 | case 0x12: /* SPSR_svc */ | |
2852 | *tgtmode = ARM_CPU_MODE_SVC; | |
2853 | break; | |
2854 | case 0x14: /* SPSR_abt */ | |
2855 | *tgtmode = ARM_CPU_MODE_ABT; | |
2856 | break; | |
2857 | case 0x16: /* SPSR_und */ | |
2858 | *tgtmode = ARM_CPU_MODE_UND; | |
2859 | break; | |
2860 | case 0x1c: /* SPSR_mon */ | |
2861 | *tgtmode = ARM_CPU_MODE_MON; | |
2862 | break; | |
2863 | case 0x1e: /* SPSR_hyp */ | |
2864 | *tgtmode = ARM_CPU_MODE_HYP; | |
2865 | break; | |
2866 | default: /* unallocated */ | |
2867 | goto undef; | |
2868 | } | |
2869 | /* We arbitrarily assign SPSR a register number of 16. */ | |
2870 | *regno = 16; | |
2871 | } else { | |
2872 | /* general purpose registers for other modes */ | |
2873 | switch (sysm) { | |
2874 | case 0x0 ... 0x6: /* 0b00xxx : r8_usr ... r14_usr */ | |
2875 | *tgtmode = ARM_CPU_MODE_USR; | |
2876 | *regno = sysm + 8; | |
2877 | break; | |
2878 | case 0x8 ... 0xe: /* 0b01xxx : r8_fiq ... r14_fiq */ | |
2879 | *tgtmode = ARM_CPU_MODE_FIQ; | |
2880 | *regno = sysm; | |
2881 | break; | |
2882 | case 0x10 ... 0x11: /* 0b1000x : r14_irq, r13_irq */ | |
2883 | *tgtmode = ARM_CPU_MODE_IRQ; | |
2884 | *regno = sysm & 1 ? 13 : 14; | |
2885 | break; | |
2886 | case 0x12 ... 0x13: /* 0b1001x : r14_svc, r13_svc */ | |
2887 | *tgtmode = ARM_CPU_MODE_SVC; | |
2888 | *regno = sysm & 1 ? 13 : 14; | |
2889 | break; | |
2890 | case 0x14 ... 0x15: /* 0b1010x : r14_abt, r13_abt */ | |
2891 | *tgtmode = ARM_CPU_MODE_ABT; | |
2892 | *regno = sysm & 1 ? 13 : 14; | |
2893 | break; | |
2894 | case 0x16 ... 0x17: /* 0b1011x : r14_und, r13_und */ | |
2895 | *tgtmode = ARM_CPU_MODE_UND; | |
2896 | *regno = sysm & 1 ? 13 : 14; | |
2897 | break; | |
2898 | case 0x1c ... 0x1d: /* 0b1110x : r14_mon, r13_mon */ | |
2899 | *tgtmode = ARM_CPU_MODE_MON; | |
2900 | *regno = sysm & 1 ? 13 : 14; | |
2901 | break; | |
2902 | case 0x1e ... 0x1f: /* 0b1111x : elr_hyp, r13_hyp */ | |
2903 | *tgtmode = ARM_CPU_MODE_HYP; | |
2904 | /* Arbitrarily pick 17 for ELR_Hyp (which is not a banked LR!) */ | |
2905 | *regno = sysm & 1 ? 13 : 17; | |
2906 | break; | |
2907 | default: /* unallocated */ | |
2908 | goto undef; | |
2909 | } | |
2910 | } | |
2911 | ||
2912 | /* Catch the 'accessing inaccessible register' cases we can detect | |
2913 | * at translate time. | |
2914 | */ | |
2915 | switch (*tgtmode) { | |
2916 | case ARM_CPU_MODE_MON: | |
2917 | if (!arm_dc_feature(s, ARM_FEATURE_EL3) || s->ns) { | |
2918 | goto undef; | |
2919 | } | |
2920 | if (s->current_el == 1) { | |
2921 | /* If we're in Secure EL1 (which implies that EL3 is AArch64) | |
2922 | * then accesses to Mon registers trap to EL3 | |
2923 | */ | |
2924 | exc_target = 3; | |
2925 | goto undef; | |
2926 | } | |
2927 | break; | |
2928 | case ARM_CPU_MODE_HYP: | |
aec4dd09 PM |
2929 | /* |
2930 | * SPSR_hyp and r13_hyp can only be accessed from Monitor mode | |
2931 | * (and so we can forbid accesses from EL2 or below). elr_hyp | |
2932 | * can be accessed also from Hyp mode, so forbid accesses from | |
2933 | * EL0 or EL1. | |
8bfd0550 | 2934 | */ |
aec4dd09 PM |
2935 | if (!arm_dc_feature(s, ARM_FEATURE_EL2) || s->current_el < 2 || |
2936 | (s->current_el < 3 && *regno != 17)) { | |
8bfd0550 PM |
2937 | goto undef; |
2938 | } | |
2939 | break; | |
2940 | default: | |
2941 | break; | |
2942 | } | |
2943 | ||
2944 | return true; | |
2945 | ||
2946 | undef: | |
2947 | /* If we get here then some access check did not pass */ | |
a767fac8 RH |
2948 | gen_exception_insn(s, s->pc_curr, EXCP_UDEF, |
2949 | syn_uncategorized(), exc_target); | |
8bfd0550 PM |
2950 | return false; |
2951 | } | |
2952 | ||
2953 | static void gen_msr_banked(DisasContext *s, int r, int sysm, int rn) | |
2954 | { | |
2955 | TCGv_i32 tcg_reg, tcg_tgtmode, tcg_regno; | |
2956 | int tgtmode = 0, regno = 0; | |
2957 | ||
2958 | if (!msr_banked_access_decode(s, r, sysm, rn, &tgtmode, ®no)) { | |
2959 | return; | |
2960 | } | |
2961 | ||
2962 | /* Sync state because msr_banked() can raise exceptions */ | |
2963 | gen_set_condexec(s); | |
43722a6d | 2964 | gen_set_pc_im(s, s->pc_curr); |
8bfd0550 PM |
2965 | tcg_reg = load_reg(s, rn); |
2966 | tcg_tgtmode = tcg_const_i32(tgtmode); | |
2967 | tcg_regno = tcg_const_i32(regno); | |
2968 | gen_helper_msr_banked(cpu_env, tcg_reg, tcg_tgtmode, tcg_regno); | |
2969 | tcg_temp_free_i32(tcg_tgtmode); | |
2970 | tcg_temp_free_i32(tcg_regno); | |
2971 | tcg_temp_free_i32(tcg_reg); | |
dcba3a8d | 2972 | s->base.is_jmp = DISAS_UPDATE; |
8bfd0550 PM |
2973 | } |
2974 | ||
2975 | static void gen_mrs_banked(DisasContext *s, int r, int sysm, int rn) | |
2976 | { | |
2977 | TCGv_i32 tcg_reg, tcg_tgtmode, tcg_regno; | |
2978 | int tgtmode = 0, regno = 0; | |
2979 | ||
2980 | if (!msr_banked_access_decode(s, r, sysm, rn, &tgtmode, ®no)) { | |
2981 | return; | |
2982 | } | |
2983 | ||
2984 | /* Sync state because mrs_banked() can raise exceptions */ | |
2985 | gen_set_condexec(s); | |
43722a6d | 2986 | gen_set_pc_im(s, s->pc_curr); |
8bfd0550 PM |
2987 | tcg_reg = tcg_temp_new_i32(); |
2988 | tcg_tgtmode = tcg_const_i32(tgtmode); | |
2989 | tcg_regno = tcg_const_i32(regno); | |
2990 | gen_helper_mrs_banked(tcg_reg, cpu_env, tcg_tgtmode, tcg_regno); | |
2991 | tcg_temp_free_i32(tcg_tgtmode); | |
2992 | tcg_temp_free_i32(tcg_regno); | |
2993 | store_reg(s, rn, tcg_reg); | |
dcba3a8d | 2994 | s->base.is_jmp = DISAS_UPDATE; |
8bfd0550 PM |
2995 | } |
2996 | ||
fb0e8e79 PM |
2997 | /* Store value to PC as for an exception return (ie don't |
2998 | * mask bits). The subsequent call to gen_helper_cpsr_write_eret() | |
2999 | * will do the masking based on the new value of the Thumb bit. | |
3000 | */ | |
3001 | static void store_pc_exc_ret(DisasContext *s, TCGv_i32 pc) | |
b5ff1b31 | 3002 | { |
fb0e8e79 PM |
3003 | tcg_gen_mov_i32(cpu_R[15], pc); |
3004 | tcg_temp_free_i32(pc); | |
b5ff1b31 FB |
3005 | } |
3006 | ||
b0109805 | 3007 | /* Generate a v6 exception return. Marks both values as dead. */ |
39d5492a | 3008 | static void gen_rfe(DisasContext *s, TCGv_i32 pc, TCGv_i32 cpsr) |
2c0262af | 3009 | { |
fb0e8e79 PM |
3010 | store_pc_exc_ret(s, pc); |
3011 | /* The cpsr_write_eret helper will mask the low bits of PC | |
3012 | * appropriately depending on the new Thumb bit, so it must | |
3013 | * be called after storing the new PC. | |
3014 | */ | |
e69ad9df AL |
3015 | if (tb_cflags(s->base.tb) & CF_USE_ICOUNT) { |
3016 | gen_io_start(); | |
3017 | } | |
235ea1f5 | 3018 | gen_helper_cpsr_write_eret(cpu_env, cpsr); |
7d1b0095 | 3019 | tcg_temp_free_i32(cpsr); |
b29fd33d | 3020 | /* Must exit loop to check un-masked IRQs */ |
dcba3a8d | 3021 | s->base.is_jmp = DISAS_EXIT; |
9ee6e8bb | 3022 | } |
3b46e624 | 3023 | |
fb0e8e79 PM |
3024 | /* Generate an old-style exception return. Marks pc as dead. */ |
3025 | static void gen_exception_return(DisasContext *s, TCGv_i32 pc) | |
3026 | { | |
3027 | gen_rfe(s, pc, load_cpu_field(spsr)); | |
3028 | } | |
3029 | ||
ad69471c | 3030 | #define CPU_V001 cpu_V0, cpu_V0, cpu_V1 |
9ee6e8bb | 3031 | |
39d5492a | 3032 | static inline void gen_neon_add(int size, TCGv_i32 t0, TCGv_i32 t1) |
9ee6e8bb PB |
3033 | { |
3034 | switch (size) { | |
dd8fbd78 FN |
3035 | case 0: gen_helper_neon_add_u8(t0, t0, t1); break; |
3036 | case 1: gen_helper_neon_add_u16(t0, t0, t1); break; | |
3037 | case 2: tcg_gen_add_i32(t0, t0, t1); break; | |
62698be3 | 3038 | default: abort(); |
9ee6e8bb | 3039 | } |
9ee6e8bb PB |
3040 | } |
3041 | ||
39d5492a | 3042 | static inline void gen_neon_rsb(int size, TCGv_i32 t0, TCGv_i32 t1) |
ad69471c PB |
3043 | { |
3044 | switch (size) { | |
dd8fbd78 FN |
3045 | case 0: gen_helper_neon_sub_u8(t0, t1, t0); break; |
3046 | case 1: gen_helper_neon_sub_u16(t0, t1, t0); break; | |
3047 | case 2: tcg_gen_sub_i32(t0, t1, t0); break; | |
ad69471c PB |
3048 | default: return; |
3049 | } | |
3050 | } | |
3051 | ||
3052 | /* 32-bit pairwise ops end up the same as the elementwise versions. */ | |
9ecd3c5c RH |
3053 | #define gen_helper_neon_pmax_s32 tcg_gen_smax_i32 |
3054 | #define gen_helper_neon_pmax_u32 tcg_gen_umax_i32 | |
3055 | #define gen_helper_neon_pmin_s32 tcg_gen_smin_i32 | |
3056 | #define gen_helper_neon_pmin_u32 tcg_gen_umin_i32 | |
ad69471c | 3057 | |
ad69471c PB |
3058 | #define GEN_NEON_INTEGER_OP_ENV(name) do { \ |
3059 | switch ((size << 1) | u) { \ | |
3060 | case 0: \ | |
dd8fbd78 | 3061 | gen_helper_neon_##name##_s8(tmp, cpu_env, tmp, tmp2); \ |
ad69471c PB |
3062 | break; \ |
3063 | case 1: \ | |
dd8fbd78 | 3064 | gen_helper_neon_##name##_u8(tmp, cpu_env, tmp, tmp2); \ |
ad69471c PB |
3065 | break; \ |
3066 | case 2: \ | |
dd8fbd78 | 3067 | gen_helper_neon_##name##_s16(tmp, cpu_env, tmp, tmp2); \ |
ad69471c PB |
3068 | break; \ |
3069 | case 3: \ | |
dd8fbd78 | 3070 | gen_helper_neon_##name##_u16(tmp, cpu_env, tmp, tmp2); \ |
ad69471c PB |
3071 | break; \ |
3072 | case 4: \ | |
dd8fbd78 | 3073 | gen_helper_neon_##name##_s32(tmp, cpu_env, tmp, tmp2); \ |
ad69471c PB |
3074 | break; \ |
3075 | case 5: \ | |
dd8fbd78 | 3076 | gen_helper_neon_##name##_u32(tmp, cpu_env, tmp, tmp2); \ |
ad69471c PB |
3077 | break; \ |
3078 | default: return 1; \ | |
3079 | }} while (0) | |
9ee6e8bb PB |
3080 | |
3081 | #define GEN_NEON_INTEGER_OP(name) do { \ | |
3082 | switch ((size << 1) | u) { \ | |
ad69471c | 3083 | case 0: \ |
dd8fbd78 | 3084 | gen_helper_neon_##name##_s8(tmp, tmp, tmp2); \ |
ad69471c PB |
3085 | break; \ |
3086 | case 1: \ | |
dd8fbd78 | 3087 | gen_helper_neon_##name##_u8(tmp, tmp, tmp2); \ |
ad69471c PB |
3088 | break; \ |
3089 | case 2: \ | |
dd8fbd78 | 3090 | gen_helper_neon_##name##_s16(tmp, tmp, tmp2); \ |
ad69471c PB |
3091 | break; \ |
3092 | case 3: \ | |
dd8fbd78 | 3093 | gen_helper_neon_##name##_u16(tmp, tmp, tmp2); \ |
ad69471c PB |
3094 | break; \ |
3095 | case 4: \ | |
dd8fbd78 | 3096 | gen_helper_neon_##name##_s32(tmp, tmp, tmp2); \ |
ad69471c PB |
3097 | break; \ |
3098 | case 5: \ | |
dd8fbd78 | 3099 | gen_helper_neon_##name##_u32(tmp, tmp, tmp2); \ |
ad69471c | 3100 | break; \ |
9ee6e8bb PB |
3101 | default: return 1; \ |
3102 | }} while (0) | |
3103 | ||
39d5492a | 3104 | static TCGv_i32 neon_load_scratch(int scratch) |
9ee6e8bb | 3105 | { |
39d5492a | 3106 | TCGv_i32 tmp = tcg_temp_new_i32(); |
dd8fbd78 FN |
3107 | tcg_gen_ld_i32(tmp, cpu_env, offsetof(CPUARMState, vfp.scratch[scratch])); |
3108 | return tmp; | |
9ee6e8bb PB |
3109 | } |
3110 | ||
39d5492a | 3111 | static void neon_store_scratch(int scratch, TCGv_i32 var) |
9ee6e8bb | 3112 | { |
dd8fbd78 | 3113 | tcg_gen_st_i32(var, cpu_env, offsetof(CPUARMState, vfp.scratch[scratch])); |
7d1b0095 | 3114 | tcg_temp_free_i32(var); |
9ee6e8bb PB |
3115 | } |
3116 | ||
39d5492a | 3117 | static inline TCGv_i32 neon_get_scalar(int size, int reg) |
9ee6e8bb | 3118 | { |
39d5492a | 3119 | TCGv_i32 tmp; |
9ee6e8bb | 3120 | if (size == 1) { |
0fad6efc PM |
3121 | tmp = neon_load_reg(reg & 7, reg >> 4); |
3122 | if (reg & 8) { | |
dd8fbd78 | 3123 | gen_neon_dup_high16(tmp); |
0fad6efc PM |
3124 | } else { |
3125 | gen_neon_dup_low16(tmp); | |
dd8fbd78 | 3126 | } |
0fad6efc PM |
3127 | } else { |
3128 | tmp = neon_load_reg(reg & 15, reg >> 4); | |
9ee6e8bb | 3129 | } |
dd8fbd78 | 3130 | return tmp; |
9ee6e8bb PB |
3131 | } |
3132 | ||
02acedf9 | 3133 | static int gen_neon_unzip(int rd, int rm, int size, int q) |
19457615 | 3134 | { |
b13708bb RH |
3135 | TCGv_ptr pd, pm; |
3136 | ||
600b828c | 3137 | if (!q && size == 2) { |
02acedf9 PM |
3138 | return 1; |
3139 | } | |
b13708bb RH |
3140 | pd = vfp_reg_ptr(true, rd); |
3141 | pm = vfp_reg_ptr(true, rm); | |
02acedf9 PM |
3142 | if (q) { |
3143 | switch (size) { | |
3144 | case 0: | |
b13708bb | 3145 | gen_helper_neon_qunzip8(pd, pm); |
02acedf9 PM |
3146 | break; |
3147 | case 1: | |
b13708bb | 3148 | gen_helper_neon_qunzip16(pd, pm); |
02acedf9 PM |
3149 | break; |
3150 | case 2: | |
b13708bb | 3151 | gen_helper_neon_qunzip32(pd, pm); |
02acedf9 PM |
3152 | break; |
3153 | default: | |
3154 | abort(); | |
3155 | } | |
3156 | } else { | |
3157 | switch (size) { | |
3158 | case 0: | |
b13708bb | 3159 | gen_helper_neon_unzip8(pd, pm); |
02acedf9 PM |
3160 | break; |
3161 | case 1: | |
b13708bb | 3162 | gen_helper_neon_unzip16(pd, pm); |
02acedf9 PM |
3163 | break; |
3164 | default: | |
3165 | abort(); | |
3166 | } | |
3167 | } | |
b13708bb RH |
3168 | tcg_temp_free_ptr(pd); |
3169 | tcg_temp_free_ptr(pm); | |
02acedf9 | 3170 | return 0; |
19457615 FN |
3171 | } |
3172 | ||
d68a6f3a | 3173 | static int gen_neon_zip(int rd, int rm, int size, int q) |
19457615 | 3174 | { |
b13708bb RH |
3175 | TCGv_ptr pd, pm; |
3176 | ||
600b828c | 3177 | if (!q && size == 2) { |
d68a6f3a PM |
3178 | return 1; |
3179 | } | |
b13708bb RH |
3180 | pd = vfp_reg_ptr(true, rd); |
3181 | pm = vfp_reg_ptr(true, rm); | |
d68a6f3a PM |
3182 | if (q) { |
3183 | switch (size) { | |
3184 | case 0: | |
b13708bb | 3185 | gen_helper_neon_qzip8(pd, pm); |
d68a6f3a PM |
3186 | break; |
3187 | case 1: | |
b13708bb | 3188 | gen_helper_neon_qzip16(pd, pm); |
d68a6f3a PM |
3189 | break; |
3190 | case 2: | |
b13708bb | 3191 | gen_helper_neon_qzip32(pd, pm); |
d68a6f3a PM |
3192 | break; |
3193 | default: | |
3194 | abort(); | |
3195 | } | |
3196 | } else { | |
3197 | switch (size) { | |
3198 | case 0: | |
b13708bb | 3199 | gen_helper_neon_zip8(pd, pm); |
d68a6f3a PM |
3200 | break; |
3201 | case 1: | |
b13708bb | 3202 | gen_helper_neon_zip16(pd, pm); |
d68a6f3a PM |
3203 | break; |
3204 | default: | |
3205 | abort(); | |
3206 | } | |
3207 | } | |
b13708bb RH |
3208 | tcg_temp_free_ptr(pd); |
3209 | tcg_temp_free_ptr(pm); | |
d68a6f3a | 3210 | return 0; |
19457615 FN |
3211 | } |
3212 | ||
39d5492a | 3213 | static void gen_neon_trn_u8(TCGv_i32 t0, TCGv_i32 t1) |
19457615 | 3214 | { |
39d5492a | 3215 | TCGv_i32 rd, tmp; |
19457615 | 3216 | |
7d1b0095 PM |
3217 | rd = tcg_temp_new_i32(); |
3218 | tmp = tcg_temp_new_i32(); | |
19457615 FN |
3219 | |
3220 | tcg_gen_shli_i32(rd, t0, 8); | |
3221 | tcg_gen_andi_i32(rd, rd, 0xff00ff00); | |
3222 | tcg_gen_andi_i32(tmp, t1, 0x00ff00ff); | |
3223 | tcg_gen_or_i32(rd, rd, tmp); | |
3224 | ||
3225 | tcg_gen_shri_i32(t1, t1, 8); | |
3226 | tcg_gen_andi_i32(t1, t1, 0x00ff00ff); | |
3227 | tcg_gen_andi_i32(tmp, t0, 0xff00ff00); | |
3228 | tcg_gen_or_i32(t1, t1, tmp); | |
3229 | tcg_gen_mov_i32(t0, rd); | |
3230 | ||
7d1b0095 PM |
3231 | tcg_temp_free_i32(tmp); |
3232 | tcg_temp_free_i32(rd); | |
19457615 FN |
3233 | } |
3234 | ||
39d5492a | 3235 | static void gen_neon_trn_u16(TCGv_i32 t0, TCGv_i32 t1) |
19457615 | 3236 | { |
39d5492a | 3237 | TCGv_i32 rd, tmp; |
19457615 | 3238 | |
7d1b0095 PM |
3239 | rd = tcg_temp_new_i32(); |
3240 | tmp = tcg_temp_new_i32(); | |
19457615 FN |
3241 | |
3242 | tcg_gen_shli_i32(rd, t0, 16); | |
3243 | tcg_gen_andi_i32(tmp, t1, 0xffff); | |
3244 | tcg_gen_or_i32(rd, rd, tmp); | |
3245 | tcg_gen_shri_i32(t1, t1, 16); | |
3246 | tcg_gen_andi_i32(tmp, t0, 0xffff0000); | |
3247 | tcg_gen_or_i32(t1, t1, tmp); | |
3248 | tcg_gen_mov_i32(t0, rd); | |
3249 | ||
7d1b0095 PM |
3250 | tcg_temp_free_i32(tmp); |
3251 | tcg_temp_free_i32(rd); | |
19457615 FN |
3252 | } |
3253 | ||
3254 | ||
9ee6e8bb PB |
3255 | static struct { |
3256 | int nregs; | |
3257 | int interleave; | |
3258 | int spacing; | |
308e5636 | 3259 | } const neon_ls_element_type[11] = { |
ac55d007 RH |
3260 | {1, 4, 1}, |
3261 | {1, 4, 2}, | |
9ee6e8bb | 3262 | {4, 1, 1}, |
ac55d007 RH |
3263 | {2, 2, 2}, |
3264 | {1, 3, 1}, | |
3265 | {1, 3, 2}, | |
9ee6e8bb PB |
3266 | {3, 1, 1}, |
3267 | {1, 1, 1}, | |
ac55d007 RH |
3268 | {1, 2, 1}, |
3269 | {1, 2, 2}, | |
9ee6e8bb PB |
3270 | {2, 1, 1} |
3271 | }; | |
3272 | ||
3273 | /* Translate a NEON load/store element instruction. Return nonzero if the | |
3274 | instruction is invalid. */ | |
7dcc1f89 | 3275 | static int disas_neon_ls_insn(DisasContext *s, uint32_t insn) |
9ee6e8bb PB |
3276 | { |
3277 | int rd, rn, rm; | |
3278 | int op; | |
3279 | int nregs; | |
3280 | int interleave; | |
84496233 | 3281 | int spacing; |
9ee6e8bb PB |
3282 | int stride; |
3283 | int size; | |
3284 | int reg; | |
9ee6e8bb | 3285 | int load; |
9ee6e8bb | 3286 | int n; |
7377c2c9 | 3287 | int vec_size; |
ac55d007 | 3288 | int mmu_idx; |
14776ab5 | 3289 | MemOp endian; |
39d5492a PM |
3290 | TCGv_i32 addr; |
3291 | TCGv_i32 tmp; | |
3292 | TCGv_i32 tmp2; | |
84496233 | 3293 | TCGv_i64 tmp64; |
9ee6e8bb | 3294 | |
2c7ffc41 PM |
3295 | /* FIXME: this access check should not take precedence over UNDEF |
3296 | * for invalid encodings; we will generate incorrect syndrome information | |
3297 | * for attempts to execute invalid vfp/neon encodings with FP disabled. | |
3298 | */ | |
9dbbc748 | 3299 | if (s->fp_excp_el) { |
a767fac8 | 3300 | gen_exception_insn(s, s->pc_curr, EXCP_UDEF, |
4be42f40 | 3301 | syn_simd_access_trap(1, 0xe, false), s->fp_excp_el); |
2c7ffc41 PM |
3302 | return 0; |
3303 | } | |
3304 | ||
5df8bac1 | 3305 | if (!s->vfp_enabled) |
9ee6e8bb PB |
3306 | return 1; |
3307 | VFP_DREG_D(rd, insn); | |
3308 | rn = (insn >> 16) & 0xf; | |
3309 | rm = insn & 0xf; | |
3310 | load = (insn & (1 << 21)) != 0; | |
ac55d007 RH |
3311 | endian = s->be_data; |
3312 | mmu_idx = get_mem_index(s); | |
9ee6e8bb PB |
3313 | if ((insn & (1 << 23)) == 0) { |
3314 | /* Load store all elements. */ | |
3315 | op = (insn >> 8) & 0xf; | |
3316 | size = (insn >> 6) & 3; | |
84496233 | 3317 | if (op > 10) |
9ee6e8bb | 3318 | return 1; |
f2dd89d0 PM |
3319 | /* Catch UNDEF cases for bad values of align field */ |
3320 | switch (op & 0xc) { | |
3321 | case 4: | |
3322 | if (((insn >> 5) & 1) == 1) { | |
3323 | return 1; | |
3324 | } | |
3325 | break; | |
3326 | case 8: | |
3327 | if (((insn >> 4) & 3) == 3) { | |
3328 | return 1; | |
3329 | } | |
3330 | break; | |
3331 | default: | |
3332 | break; | |
3333 | } | |
9ee6e8bb PB |
3334 | nregs = neon_ls_element_type[op].nregs; |
3335 | interleave = neon_ls_element_type[op].interleave; | |
84496233 | 3336 | spacing = neon_ls_element_type[op].spacing; |
ac55d007 | 3337 | if (size == 3 && (interleave | spacing) != 1) { |
84496233 | 3338 | return 1; |
ac55d007 | 3339 | } |
e23f12b3 RH |
3340 | /* For our purposes, bytes are always little-endian. */ |
3341 | if (size == 0) { | |
3342 | endian = MO_LE; | |
3343 | } | |
3344 | /* Consecutive little-endian elements from a single register | |
3345 | * can be promoted to a larger little-endian operation. | |
3346 | */ | |
3347 | if (interleave == 1 && endian == MO_LE) { | |
3348 | size = 3; | |
3349 | } | |
ac55d007 | 3350 | tmp64 = tcg_temp_new_i64(); |
e318a60b | 3351 | addr = tcg_temp_new_i32(); |
ac55d007 | 3352 | tmp2 = tcg_const_i32(1 << size); |
dcc65026 | 3353 | load_reg_var(s, addr, rn); |
9ee6e8bb | 3354 | for (reg = 0; reg < nregs; reg++) { |
ac55d007 RH |
3355 | for (n = 0; n < 8 >> size; n++) { |
3356 | int xs; | |
3357 | for (xs = 0; xs < interleave; xs++) { | |
3358 | int tt = rd + reg + spacing * xs; | |
3359 | ||
3360 | if (load) { | |
3361 | gen_aa32_ld_i64(s, tmp64, addr, mmu_idx, endian | size); | |
3362 | neon_store_element64(tt, n, size, tmp64); | |
3363 | } else { | |
3364 | neon_load_element64(tmp64, tt, n, size); | |
3365 | gen_aa32_st_i64(s, tmp64, addr, mmu_idx, endian | size); | |
9ee6e8bb | 3366 | } |
ac55d007 | 3367 | tcg_gen_add_i32(addr, addr, tmp2); |
9ee6e8bb PB |
3368 | } |
3369 | } | |
9ee6e8bb | 3370 | } |
e318a60b | 3371 | tcg_temp_free_i32(addr); |
ac55d007 RH |
3372 | tcg_temp_free_i32(tmp2); |
3373 | tcg_temp_free_i64(tmp64); | |
3374 | stride = nregs * interleave * 8; | |
9ee6e8bb PB |
3375 | } else { |
3376 | size = (insn >> 10) & 3; | |
3377 | if (size == 3) { | |
3378 | /* Load single element to all lanes. */ | |
8e18cde3 PM |
3379 | int a = (insn >> 4) & 1; |
3380 | if (!load) { | |
9ee6e8bb | 3381 | return 1; |
8e18cde3 | 3382 | } |
9ee6e8bb PB |
3383 | size = (insn >> 6) & 3; |
3384 | nregs = ((insn >> 8) & 3) + 1; | |
8e18cde3 PM |
3385 | |
3386 | if (size == 3) { | |
3387 | if (nregs != 4 || a == 0) { | |
9ee6e8bb | 3388 | return 1; |
99c475ab | 3389 | } |
8e18cde3 PM |
3390 | /* For VLD4 size==3 a == 1 means 32 bits at 16 byte alignment */ |
3391 | size = 2; | |
3392 | } | |
3393 | if (nregs == 1 && a == 1 && size == 0) { | |
3394 | return 1; | |
3395 | } | |
3396 | if (nregs == 3 && a == 1) { | |
3397 | return 1; | |
3398 | } | |
e318a60b | 3399 | addr = tcg_temp_new_i32(); |
8e18cde3 | 3400 | load_reg_var(s, addr, rn); |
7377c2c9 RH |
3401 | |
3402 | /* VLD1 to all lanes: bit 5 indicates how many Dregs to write. | |
3403 | * VLD2/3/4 to all lanes: bit 5 indicates register stride. | |
3404 | */ | |
3405 | stride = (insn & (1 << 5)) ? 2 : 1; | |
3406 | vec_size = nregs == 1 ? stride * 8 : 8; | |
3407 | ||
3408 | tmp = tcg_temp_new_i32(); | |
3409 | for (reg = 0; reg < nregs; reg++) { | |
3410 | gen_aa32_ld_i32(s, tmp, addr, get_mem_index(s), | |
3411 | s->be_data | size); | |
3412 | if ((rd & 1) && vec_size == 16) { | |
3413 | /* We cannot write 16 bytes at once because the | |
3414 | * destination is unaligned. | |
3415 | */ | |
3416 | tcg_gen_gvec_dup_i32(size, neon_reg_offset(rd, 0), | |
3417 | 8, 8, tmp); | |
3418 | tcg_gen_gvec_mov(0, neon_reg_offset(rd + 1, 0), | |
3419 | neon_reg_offset(rd, 0), 8, 8); | |
3420 | } else { | |
3421 | tcg_gen_gvec_dup_i32(size, neon_reg_offset(rd, 0), | |
3422 | vec_size, vec_size, tmp); | |
8e18cde3 | 3423 | } |
7377c2c9 RH |
3424 | tcg_gen_addi_i32(addr, addr, 1 << size); |
3425 | rd += stride; | |
9ee6e8bb | 3426 | } |
7377c2c9 | 3427 | tcg_temp_free_i32(tmp); |
e318a60b | 3428 | tcg_temp_free_i32(addr); |
9ee6e8bb PB |
3429 | stride = (1 << size) * nregs; |
3430 | } else { | |
3431 | /* Single element. */ | |
93262b16 | 3432 | int idx = (insn >> 4) & 0xf; |
2d6ac920 | 3433 | int reg_idx; |
9ee6e8bb PB |
3434 | switch (size) { |
3435 | case 0: | |
2d6ac920 | 3436 | reg_idx = (insn >> 5) & 7; |
9ee6e8bb PB |
3437 | stride = 1; |
3438 | break; | |
3439 | case 1: | |
2d6ac920 | 3440 | reg_idx = (insn >> 6) & 3; |
9ee6e8bb PB |
3441 | stride = (insn & (1 << 5)) ? 2 : 1; |
3442 | break; | |
3443 | case 2: | |
2d6ac920 | 3444 | reg_idx = (insn >> 7) & 1; |
9ee6e8bb PB |
3445 | stride = (insn & (1 << 6)) ? 2 : 1; |
3446 | break; | |
3447 | default: | |
3448 | abort(); | |
3449 | } | |
3450 | nregs = ((insn >> 8) & 3) + 1; | |
93262b16 PM |
3451 | /* Catch the UNDEF cases. This is unavoidably a bit messy. */ |
3452 | switch (nregs) { | |
3453 | case 1: | |
3454 | if (((idx & (1 << size)) != 0) || | |
3455 | (size == 2 && ((idx & 3) == 1 || (idx & 3) == 2))) { | |
3456 | return 1; | |
3457 | } | |
3458 | break; | |
3459 | case 3: | |
3460 | if ((idx & 1) != 0) { | |
3461 | return 1; | |
3462 | } | |
3463 | /* fall through */ | |
3464 | case 2: | |
3465 | if (size == 2 && (idx & 2) != 0) { | |
3466 | return 1; | |
3467 | } | |
3468 | break; | |
3469 | case 4: | |
3470 | if ((size == 2) && ((idx & 3) == 3)) { | |
3471 | return 1; | |
3472 | } | |
3473 | break; | |
3474 | default: | |
3475 | abort(); | |
3476 | } | |
3477 | if ((rd + stride * (nregs - 1)) > 31) { | |
3478 | /* Attempts to write off the end of the register file | |
3479 | * are UNPREDICTABLE; we choose to UNDEF because otherwise | |
3480 | * the neon_load_reg() would write off the end of the array. | |
3481 | */ | |
3482 | return 1; | |
3483 | } | |
2d6ac920 | 3484 | tmp = tcg_temp_new_i32(); |
e318a60b | 3485 | addr = tcg_temp_new_i32(); |
dcc65026 | 3486 | load_reg_var(s, addr, rn); |
9ee6e8bb PB |
3487 | for (reg = 0; reg < nregs; reg++) { |
3488 | if (load) { | |
2d6ac920 RH |
3489 | gen_aa32_ld_i32(s, tmp, addr, get_mem_index(s), |
3490 | s->be_data | size); | |
3491 | neon_store_element(rd, reg_idx, size, tmp); | |
9ee6e8bb | 3492 | } else { /* Store */ |
2d6ac920 RH |
3493 | neon_load_element(tmp, rd, reg_idx, size); |
3494 | gen_aa32_st_i32(s, tmp, addr, get_mem_index(s), | |
3495 | s->be_data | size); | |
99c475ab | 3496 | } |
9ee6e8bb | 3497 | rd += stride; |
1b2b1e54 | 3498 | tcg_gen_addi_i32(addr, addr, 1 << size); |
99c475ab | 3499 | } |
e318a60b | 3500 | tcg_temp_free_i32(addr); |
2d6ac920 | 3501 | tcg_temp_free_i32(tmp); |
9ee6e8bb | 3502 | stride = nregs * (1 << size); |
99c475ab | 3503 | } |
9ee6e8bb PB |
3504 | } |
3505 | if (rm != 15) { | |
39d5492a | 3506 | TCGv_i32 base; |
b26eefb6 PB |
3507 | |
3508 | base = load_reg(s, rn); | |
9ee6e8bb | 3509 | if (rm == 13) { |
b26eefb6 | 3510 | tcg_gen_addi_i32(base, base, stride); |
9ee6e8bb | 3511 | } else { |
39d5492a | 3512 | TCGv_i32 index; |
b26eefb6 PB |
3513 | index = load_reg(s, rm); |
3514 | tcg_gen_add_i32(base, base, index); | |
7d1b0095 | 3515 | tcg_temp_free_i32(index); |
9ee6e8bb | 3516 | } |
b26eefb6 | 3517 | store_reg(s, rn, base); |
9ee6e8bb PB |
3518 | } |
3519 | return 0; | |
3520 | } | |
3b46e624 | 3521 | |
39d5492a | 3522 | static inline void gen_neon_narrow(int size, TCGv_i32 dest, TCGv_i64 src) |
ad69471c PB |
3523 | { |
3524 | switch (size) { | |
3525 | case 0: gen_helper_neon_narrow_u8(dest, src); break; | |
3526 | case 1: gen_helper_neon_narrow_u16(dest, src); break; | |
ecc7b3aa | 3527 | case 2: tcg_gen_extrl_i64_i32(dest, src); break; |
ad69471c PB |
3528 | default: abort(); |
3529 | } | |
3530 | } | |
3531 | ||
39d5492a | 3532 | static inline void gen_neon_narrow_sats(int size, TCGv_i32 dest, TCGv_i64 src) |
ad69471c PB |
3533 | { |
3534 | switch (size) { | |
02da0b2d PM |
3535 | case 0: gen_helper_neon_narrow_sat_s8(dest, cpu_env, src); break; |
3536 | case 1: gen_helper_neon_narrow_sat_s16(dest, cpu_env, src); break; | |
3537 | case 2: gen_helper_neon_narrow_sat_s32(dest, cpu_env, src); break; | |
ad69471c PB |
3538 | default: abort(); |
3539 | } | |
3540 | } | |
3541 | ||
39d5492a | 3542 | static inline void gen_neon_narrow_satu(int size, TCGv_i32 dest, TCGv_i64 src) |
ad69471c PB |
3543 | { |
3544 | switch (size) { | |
02da0b2d PM |
3545 | case 0: gen_helper_neon_narrow_sat_u8(dest, cpu_env, src); break; |
3546 | case 1: gen_helper_neon_narrow_sat_u16(dest, cpu_env, src); break; | |
3547 | case 2: gen_helper_neon_narrow_sat_u32(dest, cpu_env, src); break; | |
ad69471c PB |
3548 | default: abort(); |
3549 | } | |
3550 | } | |
3551 | ||
39d5492a | 3552 | static inline void gen_neon_unarrow_sats(int size, TCGv_i32 dest, TCGv_i64 src) |
af1bbf30 JR |
3553 | { |
3554 | switch (size) { | |
02da0b2d PM |
3555 | case 0: gen_helper_neon_unarrow_sat8(dest, cpu_env, src); break; |
3556 | case 1: gen_helper_neon_unarrow_sat16(dest, cpu_env, src); break; | |
3557 | case 2: gen_helper_neon_unarrow_sat32(dest, cpu_env, src); break; | |
af1bbf30 JR |
3558 | default: abort(); |
3559 | } | |
3560 | } | |
3561 | ||
39d5492a | 3562 | static inline void gen_neon_shift_narrow(int size, TCGv_i32 var, TCGv_i32 shift, |
ad69471c PB |
3563 | int q, int u) |
3564 | { | |
3565 | if (q) { | |
3566 | if (u) { | |
3567 | switch (size) { | |
3568 | case 1: gen_helper_neon_rshl_u16(var, var, shift); break; | |
3569 | case 2: gen_helper_neon_rshl_u32(var, var, shift); break; | |
3570 | default: abort(); | |
3571 | } | |
3572 | } else { | |
3573 | switch (size) { | |
3574 | case 1: gen_helper_neon_rshl_s16(var, var, shift); break; | |
3575 | case 2: gen_helper_neon_rshl_s32(var, var, shift); break; | |
3576 | default: abort(); | |
3577 | } | |
3578 | } | |
3579 | } else { | |
3580 | if (u) { | |
3581 | switch (size) { | |
b408a9b0 CL |
3582 | case 1: gen_helper_neon_shl_u16(var, var, shift); break; |
3583 | case 2: gen_helper_neon_shl_u32(var, var, shift); break; | |
ad69471c PB |
3584 | default: abort(); |
3585 | } | |
3586 | } else { | |
3587 | switch (size) { | |
3588 | case 1: gen_helper_neon_shl_s16(var, var, shift); break; | |
3589 | case 2: gen_helper_neon_shl_s32(var, var, shift); break; | |
3590 | default: abort(); | |
3591 | } | |
3592 | } | |
3593 | } | |
3594 | } | |
3595 | ||
39d5492a | 3596 | static inline void gen_neon_widen(TCGv_i64 dest, TCGv_i32 src, int size, int u) |
ad69471c PB |
3597 | { |
3598 | if (u) { | |
3599 | switch (size) { | |
3600 | case 0: gen_helper_neon_widen_u8(dest, src); break; | |
3601 | case 1: gen_helper_neon_widen_u16(dest, src); break; | |
3602 | case 2: tcg_gen_extu_i32_i64(dest, src); break; | |
3603 | default: abort(); | |
3604 | } | |
3605 | } else { | |
3606 | switch (size) { | |
3607 | case 0: gen_helper_neon_widen_s8(dest, src); break; | |
3608 | case 1: gen_helper_neon_widen_s16(dest, src); break; | |
3609 | case 2: tcg_gen_ext_i32_i64(dest, src); break; | |
3610 | default: abort(); | |
3611 | } | |
3612 | } | |
7d1b0095 | 3613 | tcg_temp_free_i32(src); |
ad69471c PB |
3614 | } |
3615 | ||
3616 | static inline void gen_neon_addl(int size) | |
3617 | { | |
3618 | switch (size) { | |
3619 | case 0: gen_helper_neon_addl_u16(CPU_V001); break; | |
3620 | case 1: gen_helper_neon_addl_u32(CPU_V001); break; | |
3621 | case 2: tcg_gen_add_i64(CPU_V001); break; | |
3622 | default: abort(); | |
3623 | } | |
3624 | } | |
3625 | ||
3626 | static inline void gen_neon_subl(int size) | |
3627 | { | |
3628 | switch (size) { | |
3629 | case 0: gen_helper_neon_subl_u16(CPU_V001); break; | |
3630 | case 1: gen_helper_neon_subl_u32(CPU_V001); break; | |
3631 | case 2: tcg_gen_sub_i64(CPU_V001); break; | |
3632 | default: abort(); | |
3633 | } | |
3634 | } | |
3635 | ||
a7812ae4 | 3636 | static inline void gen_neon_negl(TCGv_i64 var, int size) |
ad69471c PB |
3637 | { |
3638 | switch (size) { | |
3639 | case 0: gen_helper_neon_negl_u16(var, var); break; | |
3640 | case 1: gen_helper_neon_negl_u32(var, var); break; | |
ee6fa559 PM |
3641 | case 2: |
3642 | tcg_gen_neg_i64(var, var); | |
3643 | break; | |
ad69471c PB |
3644 | default: abort(); |
3645 | } | |
3646 | } | |
3647 | ||
a7812ae4 | 3648 | static inline void gen_neon_addl_saturate(TCGv_i64 op0, TCGv_i64 op1, int size) |
ad69471c PB |
3649 | { |
3650 | switch (size) { | |
02da0b2d PM |
3651 | case 1: gen_helper_neon_addl_saturate_s32(op0, cpu_env, op0, op1); break; |
3652 | case 2: gen_helper_neon_addl_saturate_s64(op0, cpu_env, op0, op1); break; | |
ad69471c PB |
3653 | default: abort(); |
3654 | } | |
3655 | } | |
3656 | ||
39d5492a PM |
3657 | static inline void gen_neon_mull(TCGv_i64 dest, TCGv_i32 a, TCGv_i32 b, |
3658 | int size, int u) | |
ad69471c | 3659 | { |
a7812ae4 | 3660 | TCGv_i64 tmp; |
ad69471c PB |
3661 | |
3662 | switch ((size << 1) | u) { | |
3663 | case 0: gen_helper_neon_mull_s8(dest, a, b); break; | |
3664 | case 1: gen_helper_neon_mull_u8(dest, a, b); break; | |
3665 | case 2: gen_helper_neon_mull_s16(dest, a, b); break; | |
3666 | case 3: gen_helper_neon_mull_u16(dest, a, b); break; | |
3667 | case 4: | |
3668 | tmp = gen_muls_i64_i32(a, b); | |
3669 | tcg_gen_mov_i64(dest, tmp); | |
7d2aabe2 | 3670 | tcg_temp_free_i64(tmp); |
ad69471c PB |
3671 | break; |
3672 | case 5: | |
3673 | tmp = gen_mulu_i64_i32(a, b); | |
3674 | tcg_gen_mov_i64(dest, tmp); | |
7d2aabe2 | 3675 | tcg_temp_free_i64(tmp); |
ad69471c PB |
3676 | break; |
3677 | default: abort(); | |
3678 | } | |
c6067f04 CL |
3679 | |
3680 | /* gen_helper_neon_mull_[su]{8|16} do not free their parameters. | |
3681 | Don't forget to clean them now. */ | |
3682 | if (size < 2) { | |
7d1b0095 PM |
3683 | tcg_temp_free_i32(a); |
3684 | tcg_temp_free_i32(b); | |
c6067f04 | 3685 | } |
ad69471c PB |
3686 | } |
3687 | ||
39d5492a PM |
3688 | static void gen_neon_narrow_op(int op, int u, int size, |
3689 | TCGv_i32 dest, TCGv_i64 src) | |
c33171c7 PM |
3690 | { |
3691 | if (op) { | |
3692 | if (u) { | |
3693 | gen_neon_unarrow_sats(size, dest, src); | |
3694 | } else { | |
3695 | gen_neon_narrow(size, dest, src); | |
3696 | } | |
3697 | } else { | |
3698 | if (u) { | |
3699 | gen_neon_narrow_satu(size, dest, src); | |
3700 | } else { | |
3701 | gen_neon_narrow_sats(size, dest, src); | |
3702 | } | |
3703 | } | |
3704 | } | |
3705 | ||
62698be3 PM |
3706 | /* Symbolic constants for op fields for Neon 3-register same-length. |
3707 | * The values correspond to bits [11:8,4]; see the ARM ARM DDI0406B | |
3708 | * table A7-9. | |
3709 | */ | |
3710 | #define NEON_3R_VHADD 0 | |
3711 | #define NEON_3R_VQADD 1 | |
3712 | #define NEON_3R_VRHADD 2 | |
3713 | #define NEON_3R_LOGIC 3 /* VAND,VBIC,VORR,VMOV,VORN,VEOR,VBIF,VBIT,VBSL */ | |
3714 | #define NEON_3R_VHSUB 4 | |
3715 | #define NEON_3R_VQSUB 5 | |
3716 | #define NEON_3R_VCGT 6 | |
3717 | #define NEON_3R_VCGE 7 | |
3718 | #define NEON_3R_VSHL 8 | |
3719 | #define NEON_3R_VQSHL 9 | |
3720 | #define NEON_3R_VRSHL 10 | |
3721 | #define NEON_3R_VQRSHL 11 | |
3722 | #define NEON_3R_VMAX 12 | |
3723 | #define NEON_3R_VMIN 13 | |
3724 | #define NEON_3R_VABD 14 | |
3725 | #define NEON_3R_VABA 15 | |
3726 | #define NEON_3R_VADD_VSUB 16 | |
3727 | #define NEON_3R_VTST_VCEQ 17 | |
4a7832b0 | 3728 | #define NEON_3R_VML 18 /* VMLA, VMLS */ |
62698be3 PM |
3729 | #define NEON_3R_VMUL 19 |
3730 | #define NEON_3R_VPMAX 20 | |
3731 | #define NEON_3R_VPMIN 21 | |
3732 | #define NEON_3R_VQDMULH_VQRDMULH 22 | |
36a71934 | 3733 | #define NEON_3R_VPADD_VQRDMLAH 23 |
f1ecb913 | 3734 | #define NEON_3R_SHA 24 /* SHA1C,SHA1P,SHA1M,SHA1SU0,SHA256H{2},SHA256SU1 */ |
36a71934 | 3735 | #define NEON_3R_VFM_VQRDMLSH 25 /* VFMA, VFMS, VQRDMLSH */ |
62698be3 PM |
3736 | #define NEON_3R_FLOAT_ARITH 26 /* float VADD, VSUB, VPADD, VABD */ |
3737 | #define NEON_3R_FLOAT_MULTIPLY 27 /* float VMLA, VMLS, VMUL */ | |
3738 | #define NEON_3R_FLOAT_CMP 28 /* float VCEQ, VCGE, VCGT */ | |
3739 | #define NEON_3R_FLOAT_ACMP 29 /* float VACGE, VACGT, VACLE, VACLT */ | |
3740 | #define NEON_3R_FLOAT_MINMAX 30 /* float VMIN, VMAX */ | |
505935fc | 3741 | #define NEON_3R_FLOAT_MISC 31 /* float VRECPS, VRSQRTS, VMAXNM/MINNM */ |
62698be3 PM |
3742 | |
3743 | static const uint8_t neon_3r_sizes[] = { | |
3744 | [NEON_3R_VHADD] = 0x7, | |
3745 | [NEON_3R_VQADD] = 0xf, | |
3746 | [NEON_3R_VRHADD] = 0x7, | |
3747 | [NEON_3R_LOGIC] = 0xf, /* size field encodes op type */ | |
3748 | [NEON_3R_VHSUB] = 0x7, | |
3749 | [NEON_3R_VQSUB] = 0xf, | |
3750 | [NEON_3R_VCGT] = 0x7, | |
3751 | [NEON_3R_VCGE] = 0x7, | |
3752 | [NEON_3R_VSHL] = 0xf, | |
3753 | [NEON_3R_VQSHL] = 0xf, | |
3754 | [NEON_3R_VRSHL] = 0xf, | |
3755 | [NEON_3R_VQRSHL] = 0xf, | |
3756 | [NEON_3R_VMAX] = 0x7, | |
3757 | [NEON_3R_VMIN] = 0x7, | |
3758 | [NEON_3R_VABD] = 0x7, | |
3759 | [NEON_3R_VABA] = 0x7, | |
3760 | [NEON_3R_VADD_VSUB] = 0xf, | |
3761 | [NEON_3R_VTST_VCEQ] = 0x7, | |
3762 | [NEON_3R_VML] = 0x7, | |
3763 | [NEON_3R_VMUL] = 0x7, | |
3764 | [NEON_3R_VPMAX] = 0x7, | |
3765 | [NEON_3R_VPMIN] = 0x7, | |
3766 | [NEON_3R_VQDMULH_VQRDMULH] = 0x6, | |
36a71934 | 3767 | [NEON_3R_VPADD_VQRDMLAH] = 0x7, |
f1ecb913 | 3768 | [NEON_3R_SHA] = 0xf, /* size field encodes op type */ |
36a71934 | 3769 | [NEON_3R_VFM_VQRDMLSH] = 0x7, /* For VFM, size bit 1 encodes op */ |
62698be3 PM |
3770 | [NEON_3R_FLOAT_ARITH] = 0x5, /* size bit 1 encodes op */ |
3771 | [NEON_3R_FLOAT_MULTIPLY] = 0x5, /* size bit 1 encodes op */ | |
3772 | [NEON_3R_FLOAT_CMP] = 0x5, /* size bit 1 encodes op */ | |
3773 | [NEON_3R_FLOAT_ACMP] = 0x5, /* size bit 1 encodes op */ | |
3774 | [NEON_3R_FLOAT_MINMAX] = 0x5, /* size bit 1 encodes op */ | |
505935fc | 3775 | [NEON_3R_FLOAT_MISC] = 0x5, /* size bit 1 encodes op */ |
62698be3 PM |
3776 | }; |
3777 | ||
600b828c PM |
3778 | /* Symbolic constants for op fields for Neon 2-register miscellaneous. |
3779 | * The values correspond to bits [17:16,10:7]; see the ARM ARM DDI0406B | |
3780 | * table A7-13. | |
3781 | */ | |
3782 | #define NEON_2RM_VREV64 0 | |
3783 | #define NEON_2RM_VREV32 1 | |
3784 | #define NEON_2RM_VREV16 2 | |
3785 | #define NEON_2RM_VPADDL 4 | |
3786 | #define NEON_2RM_VPADDL_U 5 | |
9d935509 AB |
3787 | #define NEON_2RM_AESE 6 /* Includes AESD */ |
3788 | #define NEON_2RM_AESMC 7 /* Includes AESIMC */ | |
600b828c PM |
3789 | #define NEON_2RM_VCLS 8 |
3790 | #define NEON_2RM_VCLZ 9 | |
3791 | #define NEON_2RM_VCNT 10 | |
3792 | #define NEON_2RM_VMVN 11 | |
3793 | #define NEON_2RM_VPADAL 12 | |
3794 | #define NEON_2RM_VPADAL_U 13 | |
3795 | #define NEON_2RM_VQABS 14 | |
3796 | #define NEON_2RM_VQNEG 15 | |
3797 | #define NEON_2RM_VCGT0 16 | |
3798 | #define NEON_2RM_VCGE0 17 | |
3799 | #define NEON_2RM_VCEQ0 18 | |
3800 | #define NEON_2RM_VCLE0 19 | |
3801 | #define NEON_2RM_VCLT0 20 | |
f1ecb913 | 3802 | #define NEON_2RM_SHA1H 21 |
600b828c PM |
3803 | #define NEON_2RM_VABS 22 |
3804 | #define NEON_2RM_VNEG 23 | |
3805 | #define NEON_2RM_VCGT0_F 24 | |
3806 | #define NEON_2RM_VCGE0_F 25 | |
3807 | #define NEON_2RM_VCEQ0_F 26 | |
3808 | #define NEON_2RM_VCLE0_F 27 | |
3809 | #define NEON_2RM_VCLT0_F 28 | |
3810 | #define NEON_2RM_VABS_F 30 | |
3811 | #define NEON_2RM_VNEG_F 31 | |
3812 | #define NEON_2RM_VSWP 32 | |
3813 | #define NEON_2RM_VTRN 33 | |
3814 | #define NEON_2RM_VUZP 34 | |
3815 | #define NEON_2RM_VZIP 35 | |
3816 | #define NEON_2RM_VMOVN 36 /* Includes VQMOVN, VQMOVUN */ | |
3817 | #define NEON_2RM_VQMOVN 37 /* Includes VQMOVUN */ | |
3818 | #define NEON_2RM_VSHLL 38 | |
f1ecb913 | 3819 | #define NEON_2RM_SHA1SU1 39 /* Includes SHA256SU0 */ |
34f7b0a2 | 3820 | #define NEON_2RM_VRINTN 40 |
2ce70625 | 3821 | #define NEON_2RM_VRINTX 41 |
34f7b0a2 WN |
3822 | #define NEON_2RM_VRINTA 42 |
3823 | #define NEON_2RM_VRINTZ 43 | |
600b828c | 3824 | #define NEON_2RM_VCVT_F16_F32 44 |
34f7b0a2 | 3825 | #define NEON_2RM_VRINTM 45 |
600b828c | 3826 | #define NEON_2RM_VCVT_F32_F16 46 |
34f7b0a2 | 3827 | #define NEON_2RM_VRINTP 47 |
901ad525 WN |
3828 | #define NEON_2RM_VCVTAU 48 |
3829 | #define NEON_2RM_VCVTAS 49 | |
3830 | #define NEON_2RM_VCVTNU 50 | |
3831 | #define NEON_2RM_VCVTNS 51 | |
3832 | #define NEON_2RM_VCVTPU 52 | |
3833 | #define NEON_2RM_VCVTPS 53 | |
3834 | #define NEON_2RM_VCVTMU 54 | |
3835 | #define NEON_2RM_VCVTMS 55 | |
600b828c PM |
3836 | #define NEON_2RM_VRECPE 56 |
3837 | #define NEON_2RM_VRSQRTE 57 | |
3838 | #define NEON_2RM_VRECPE_F 58 | |
3839 | #define NEON_2RM_VRSQRTE_F 59 | |
3840 | #define NEON_2RM_VCVT_FS 60 | |
3841 | #define NEON_2RM_VCVT_FU 61 | |
3842 | #define NEON_2RM_VCVT_SF 62 | |
3843 | #define NEON_2RM_VCVT_UF 63 | |
3844 | ||
fe8fcf3d PM |
3845 | static bool neon_2rm_is_v8_op(int op) |
3846 | { | |
3847 | /* Return true if this neon 2reg-misc op is ARMv8 and up */ | |
3848 | switch (op) { | |
3849 | case NEON_2RM_VRINTN: | |
3850 | case NEON_2RM_VRINTA: | |
3851 | case NEON_2RM_VRINTM: | |
3852 | case NEON_2RM_VRINTP: | |
3853 | case NEON_2RM_VRINTZ: | |
3854 | case NEON_2RM_VRINTX: | |
3855 | case NEON_2RM_VCVTAU: | |
3856 | case NEON_2RM_VCVTAS: | |
3857 | case NEON_2RM_VCVTNU: | |
3858 | case NEON_2RM_VCVTNS: | |
3859 | case NEON_2RM_VCVTPU: | |
3860 | case NEON_2RM_VCVTPS: | |
3861 | case NEON_2RM_VCVTMU: | |
3862 | case NEON_2RM_VCVTMS: | |
3863 | return true; | |
3864 | default: | |
3865 | return false; | |
3866 | } | |
3867 | } | |
3868 | ||
600b828c PM |
3869 | /* Each entry in this array has bit n set if the insn allows |
3870 | * size value n (otherwise it will UNDEF). Since unallocated | |
3871 | * op values will have no bits set they always UNDEF. | |
3872 | */ | |
3873 | static const uint8_t neon_2rm_sizes[] = { | |
3874 | [NEON_2RM_VREV64] = 0x7, | |
3875 | [NEON_2RM_VREV32] = 0x3, | |
3876 | [NEON_2RM_VREV16] = 0x1, | |
3877 | [NEON_2RM_VPADDL] = 0x7, | |
3878 | [NEON_2RM_VPADDL_U] = 0x7, | |
9d935509 AB |
3879 | [NEON_2RM_AESE] = 0x1, |
3880 | [NEON_2RM_AESMC] = 0x1, | |
600b828c PM |
3881 | [NEON_2RM_VCLS] = 0x7, |
3882 | [NEON_2RM_VCLZ] = 0x7, | |
3883 | [NEON_2RM_VCNT] = 0x1, | |
3884 | [NEON_2RM_VMVN] = 0x1, | |
3885 | [NEON_2RM_VPADAL] = 0x7, | |
3886 | [NEON_2RM_VPADAL_U] = 0x7, | |
3887 | [NEON_2RM_VQABS] = 0x7, | |
3888 | [NEON_2RM_VQNEG] = 0x7, | |
3889 | [NEON_2RM_VCGT0] = 0x7, | |
3890 | [NEON_2RM_VCGE0] = 0x7, | |
3891 | [NEON_2RM_VCEQ0] = 0x7, | |
3892 | [NEON_2RM_VCLE0] = 0x7, | |
3893 | [NEON_2RM_VCLT0] = 0x7, | |
f1ecb913 | 3894 | [NEON_2RM_SHA1H] = 0x4, |
600b828c PM |
3895 | [NEON_2RM_VABS] = 0x7, |
3896 | [NEON_2RM_VNEG] = 0x7, | |
3897 | [NEON_2RM_VCGT0_F] = 0x4, | |
3898 | [NEON_2RM_VCGE0_F] = 0x4, | |
3899 | [NEON_2RM_VCEQ0_F] = 0x4, | |
3900 | [NEON_2RM_VCLE0_F] = 0x4, | |
3901 | [NEON_2RM_VCLT0_F] = 0x4, | |
3902 | [NEON_2RM_VABS_F] = 0x4, | |
3903 | [NEON_2RM_VNEG_F] = 0x4, | |
3904 | [NEON_2RM_VSWP] = 0x1, | |
3905 | [NEON_2RM_VTRN] = 0x7, | |
3906 | [NEON_2RM_VUZP] = 0x7, | |
3907 | [NEON_2RM_VZIP] = 0x7, | |
3908 | [NEON_2RM_VMOVN] = 0x7, | |
3909 | [NEON_2RM_VQMOVN] = 0x7, | |
3910 | [NEON_2RM_VSHLL] = 0x7, | |
f1ecb913 | 3911 | [NEON_2RM_SHA1SU1] = 0x4, |
34f7b0a2 | 3912 | [NEON_2RM_VRINTN] = 0x4, |
2ce70625 | 3913 | [NEON_2RM_VRINTX] = 0x4, |
34f7b0a2 WN |
3914 | [NEON_2RM_VRINTA] = 0x4, |
3915 | [NEON_2RM_VRINTZ] = 0x4, | |
600b828c | 3916 | [NEON_2RM_VCVT_F16_F32] = 0x2, |
34f7b0a2 | 3917 | [NEON_2RM_VRINTM] = 0x4, |
600b828c | 3918 | [NEON_2RM_VCVT_F32_F16] = 0x2, |
34f7b0a2 | 3919 | [NEON_2RM_VRINTP] = 0x4, |
901ad525 WN |
3920 | [NEON_2RM_VCVTAU] = 0x4, |
3921 | [NEON_2RM_VCVTAS] = 0x4, | |
3922 | [NEON_2RM_VCVTNU] = 0x4, | |
3923 | [NEON_2RM_VCVTNS] = 0x4, | |
3924 | [NEON_2RM_VCVTPU] = 0x4, | |
3925 | [NEON_2RM_VCVTPS] = 0x4, | |
3926 | [NEON_2RM_VCVTMU] = 0x4, | |
3927 | [NEON_2RM_VCVTMS] = 0x4, | |
600b828c PM |
3928 | [NEON_2RM_VRECPE] = 0x4, |
3929 | [NEON_2RM_VRSQRTE] = 0x4, | |
3930 | [NEON_2RM_VRECPE_F] = 0x4, | |
3931 | [NEON_2RM_VRSQRTE_F] = 0x4, | |
3932 | [NEON_2RM_VCVT_FS] = 0x4, | |
3933 | [NEON_2RM_VCVT_FU] = 0x4, | |
3934 | [NEON_2RM_VCVT_SF] = 0x4, | |
3935 | [NEON_2RM_VCVT_UF] = 0x4, | |
3936 | }; | |
3937 | ||
36a71934 RH |
3938 | |
3939 | /* Expand v8.1 simd helper. */ | |
3940 | static int do_v81_helper(DisasContext *s, gen_helper_gvec_3_ptr *fn, | |
3941 | int q, int rd, int rn, int rm) | |
3942 | { | |
962fcbf2 | 3943 | if (dc_isar_feature(aa32_rdm, s)) { |
36a71934 RH |
3944 | int opr_sz = (1 + q) * 8; |
3945 | tcg_gen_gvec_3_ptr(vfp_reg_offset(1, rd), | |
3946 | vfp_reg_offset(1, rn), | |
3947 | vfp_reg_offset(1, rm), cpu_env, | |
3948 | opr_sz, opr_sz, 0, fn); | |
3949 | return 0; | |
3950 | } | |
3951 | return 1; | |
3952 | } | |
3953 | ||
41f6c113 RH |
3954 | static void gen_ssra8_i64(TCGv_i64 d, TCGv_i64 a, int64_t shift) |
3955 | { | |
3956 | tcg_gen_vec_sar8i_i64(a, a, shift); | |
3957 | tcg_gen_vec_add8_i64(d, d, a); | |
3958 | } | |
3959 | ||
3960 | static void gen_ssra16_i64(TCGv_i64 d, TCGv_i64 a, int64_t shift) | |
3961 | { | |
3962 | tcg_gen_vec_sar16i_i64(a, a, shift); | |
3963 | tcg_gen_vec_add16_i64(d, d, a); | |
3964 | } | |
3965 | ||
3966 | static void gen_ssra32_i32(TCGv_i32 d, TCGv_i32 a, int32_t shift) | |
3967 | { | |
3968 | tcg_gen_sari_i32(a, a, shift); | |
3969 | tcg_gen_add_i32(d, d, a); | |
3970 | } | |
3971 | ||
3972 | static void gen_ssra64_i64(TCGv_i64 d, TCGv_i64 a, int64_t shift) | |
3973 | { | |
3974 | tcg_gen_sari_i64(a, a, shift); | |
3975 | tcg_gen_add_i64(d, d, a); | |
3976 | } | |
3977 | ||
3978 | static void gen_ssra_vec(unsigned vece, TCGv_vec d, TCGv_vec a, int64_t sh) | |
3979 | { | |
3980 | tcg_gen_sari_vec(vece, a, a, sh); | |
3981 | tcg_gen_add_vec(vece, d, d, a); | |
3982 | } | |
3983 | ||
53229a77 RH |
3984 | static const TCGOpcode vecop_list_ssra[] = { |
3985 | INDEX_op_sari_vec, INDEX_op_add_vec, 0 | |
3986 | }; | |
3987 | ||
41f6c113 RH |
3988 | const GVecGen2i ssra_op[4] = { |
3989 | { .fni8 = gen_ssra8_i64, | |
3990 | .fniv = gen_ssra_vec, | |
3991 | .load_dest = true, | |
53229a77 | 3992 | .opt_opc = vecop_list_ssra, |
41f6c113 RH |
3993 | .vece = MO_8 }, |
3994 | { .fni8 = gen_ssra16_i64, | |
3995 | .fniv = gen_ssra_vec, | |
3996 | .load_dest = true, | |
53229a77 | 3997 | .opt_opc = vecop_list_ssra, |
41f6c113 RH |
3998 | .vece = MO_16 }, |
3999 | { .fni4 = gen_ssra32_i32, | |
4000 | .fniv = gen_ssra_vec, | |
4001 | .load_dest = true, | |
53229a77 | 4002 | .opt_opc = vecop_list_ssra, |
41f6c113 RH |
4003 | .vece = MO_32 }, |
4004 | { .fni8 = gen_ssra64_i64, | |
4005 | .fniv = gen_ssra_vec, | |
4006 | .prefer_i64 = TCG_TARGET_REG_BITS == 64, | |
53229a77 | 4007 | .opt_opc = vecop_list_ssra, |
41f6c113 | 4008 | .load_dest = true, |
41f6c113 RH |
4009 | .vece = MO_64 }, |
4010 | }; | |
4011 | ||
4012 | static void gen_usra8_i64(TCGv_i64 d, TCGv_i64 a, int64_t shift) | |
4013 | { | |
4014 | tcg_gen_vec_shr8i_i64(a, a, shift); | |
4015 | tcg_gen_vec_add8_i64(d, d, a); | |
4016 | } | |
4017 | ||
4018 | static void gen_usra16_i64(TCGv_i64 d, TCGv_i64 a, int64_t shift) | |
4019 | { | |
4020 | tcg_gen_vec_shr16i_i64(a, a, shift); | |
4021 | tcg_gen_vec_add16_i64(d, d, a); | |
4022 | } | |
4023 | ||
4024 | static void gen_usra32_i32(TCGv_i32 d, TCGv_i32 a, int32_t shift) | |
4025 | { | |
4026 | tcg_gen_shri_i32(a, a, shift); | |
4027 | tcg_gen_add_i32(d, d, a); | |
4028 | } | |
4029 | ||
4030 | static void gen_usra64_i64(TCGv_i64 d, TCGv_i64 a, int64_t shift) | |
4031 | { | |
4032 | tcg_gen_shri_i64(a, a, shift); | |
4033 | tcg_gen_add_i64(d, d, a); | |
4034 | } | |
4035 | ||
4036 | static void gen_usra_vec(unsigned vece, TCGv_vec d, TCGv_vec a, int64_t sh) | |
4037 | { | |
4038 | tcg_gen_shri_vec(vece, a, a, sh); | |
4039 | tcg_gen_add_vec(vece, d, d, a); | |
4040 | } | |
4041 | ||
53229a77 RH |
4042 | static const TCGOpcode vecop_list_usra[] = { |
4043 | INDEX_op_shri_vec, INDEX_op_add_vec, 0 | |
4044 | }; | |
4045 | ||
41f6c113 RH |
4046 | const GVecGen2i usra_op[4] = { |
4047 | { .fni8 = gen_usra8_i64, | |
4048 | .fniv = gen_usra_vec, | |
4049 | .load_dest = true, | |
53229a77 | 4050 | .opt_opc = vecop_list_usra, |
41f6c113 RH |
4051 | .vece = MO_8, }, |
4052 | { .fni8 = gen_usra16_i64, | |
4053 | .fniv = gen_usra_vec, | |
4054 | .load_dest = true, | |
53229a77 | 4055 | .opt_opc = vecop_list_usra, |
41f6c113 RH |
4056 | .vece = MO_16, }, |
4057 | { .fni4 = gen_usra32_i32, | |
4058 | .fniv = gen_usra_vec, | |
4059 | .load_dest = true, | |
53229a77 | 4060 | .opt_opc = vecop_list_usra, |
41f6c113 RH |
4061 | .vece = MO_32, }, |
4062 | { .fni8 = gen_usra64_i64, | |
4063 | .fniv = gen_usra_vec, | |
4064 | .prefer_i64 = TCG_TARGET_REG_BITS == 64, | |
4065 | .load_dest = true, | |
53229a77 | 4066 | .opt_opc = vecop_list_usra, |
41f6c113 RH |
4067 | .vece = MO_64, }, |
4068 | }; | |
eabcd6fa | 4069 | |
f3cd8218 RH |
4070 | static void gen_shr8_ins_i64(TCGv_i64 d, TCGv_i64 a, int64_t shift) |
4071 | { | |
4072 | uint64_t mask = dup_const(MO_8, 0xff >> shift); | |
4073 | TCGv_i64 t = tcg_temp_new_i64(); | |
4074 | ||
4075 | tcg_gen_shri_i64(t, a, shift); | |
4076 | tcg_gen_andi_i64(t, t, mask); | |
4077 | tcg_gen_andi_i64(d, d, ~mask); | |
4078 | tcg_gen_or_i64(d, d, t); | |
4079 | tcg_temp_free_i64(t); | |
4080 | } | |
4081 | ||
4082 | static void gen_shr16_ins_i64(TCGv_i64 d, TCGv_i64 a, int64_t shift) | |
4083 | { | |
4084 | uint64_t mask = dup_const(MO_16, 0xffff >> shift); | |
4085 | TCGv_i64 t = tcg_temp_new_i64(); | |
4086 | ||
4087 | tcg_gen_shri_i64(t, a, shift); | |
4088 | tcg_gen_andi_i64(t, t, mask); | |
4089 | tcg_gen_andi_i64(d, d, ~mask); | |
4090 | tcg_gen_or_i64(d, d, t); | |
4091 | tcg_temp_free_i64(t); | |
4092 | } | |
4093 | ||
4094 | static void gen_shr32_ins_i32(TCGv_i32 d, TCGv_i32 a, int32_t shift) | |
4095 | { | |
4096 | tcg_gen_shri_i32(a, a, shift); | |
4097 | tcg_gen_deposit_i32(d, d, a, 0, 32 - shift); | |
4098 | } | |
4099 | ||
4100 | static void gen_shr64_ins_i64(TCGv_i64 d, TCGv_i64 a, int64_t shift) | |
4101 | { | |
4102 | tcg_gen_shri_i64(a, a, shift); | |
4103 | tcg_gen_deposit_i64(d, d, a, 0, 64 - shift); | |
4104 | } | |
4105 | ||
4106 | static void gen_shr_ins_vec(unsigned vece, TCGv_vec d, TCGv_vec a, int64_t sh) | |
4107 | { | |
4108 | if (sh == 0) { | |
4109 | tcg_gen_mov_vec(d, a); | |
4110 | } else { | |
4111 | TCGv_vec t = tcg_temp_new_vec_matching(d); | |
4112 | TCGv_vec m = tcg_temp_new_vec_matching(d); | |
4113 | ||
4114 | tcg_gen_dupi_vec(vece, m, MAKE_64BIT_MASK((8 << vece) - sh, sh)); | |
4115 | tcg_gen_shri_vec(vece, t, a, sh); | |
4116 | tcg_gen_and_vec(vece, d, d, m); | |
4117 | tcg_gen_or_vec(vece, d, d, t); | |
4118 | ||
4119 | tcg_temp_free_vec(t); | |
4120 | tcg_temp_free_vec(m); | |
4121 | } | |
4122 | } | |
4123 | ||
53229a77 RH |
4124 | static const TCGOpcode vecop_list_sri[] = { INDEX_op_shri_vec, 0 }; |
4125 | ||
f3cd8218 RH |
4126 | const GVecGen2i sri_op[4] = { |
4127 | { .fni8 = gen_shr8_ins_i64, | |
4128 | .fniv = gen_shr_ins_vec, | |
4129 | .load_dest = true, | |
53229a77 | 4130 | .opt_opc = vecop_list_sri, |
f3cd8218 RH |
4131 | .vece = MO_8 }, |
4132 | { .fni8 = gen_shr16_ins_i64, | |
4133 | .fniv = gen_shr_ins_vec, | |
4134 | .load_dest = true, | |
53229a77 | 4135 | .opt_opc = vecop_list_sri, |
f3cd8218 RH |
4136 | .vece = MO_16 }, |
4137 | { .fni4 = gen_shr32_ins_i32, | |
4138 | .fniv = gen_shr_ins_vec, | |
4139 | .load_dest = true, | |
53229a77 | 4140 | .opt_opc = vecop_list_sri, |
f3cd8218 RH |
4141 | .vece = MO_32 }, |
4142 | { .fni8 = gen_shr64_ins_i64, | |
4143 | .fniv = gen_shr_ins_vec, | |
4144 | .prefer_i64 = TCG_TARGET_REG_BITS == 64, | |
4145 | .load_dest = true, | |
53229a77 | 4146 | .opt_opc = vecop_list_sri, |
f3cd8218 RH |
4147 | .vece = MO_64 }, |
4148 | }; | |
4149 | ||
4150 | static void gen_shl8_ins_i64(TCGv_i64 d, TCGv_i64 a, int64_t shift) | |
4151 | { | |
4152 | uint64_t mask = dup_const(MO_8, 0xff << shift); | |
4153 | TCGv_i64 t = tcg_temp_new_i64(); | |
4154 | ||
4155 | tcg_gen_shli_i64(t, a, shift); | |
4156 | tcg_gen_andi_i64(t, t, mask); | |
4157 | tcg_gen_andi_i64(d, d, ~mask); | |
4158 | tcg_gen_or_i64(d, d, t); | |
4159 | tcg_temp_free_i64(t); | |
4160 | } | |
4161 | ||
4162 | static void gen_shl16_ins_i64(TCGv_i64 d, TCGv_i64 a, int64_t shift) | |
4163 | { | |
4164 | uint64_t mask = dup_const(MO_16, 0xffff << shift); | |
4165 | TCGv_i64 t = tcg_temp_new_i64(); | |
4166 | ||
4167 | tcg_gen_shli_i64(t, a, shift); | |
4168 | tcg_gen_andi_i64(t, t, mask); | |
4169 | tcg_gen_andi_i64(d, d, ~mask); | |
4170 | tcg_gen_or_i64(d, d, t); | |
4171 | tcg_temp_free_i64(t); | |
4172 | } | |
4173 | ||
4174 | static void gen_shl32_ins_i32(TCGv_i32 d, TCGv_i32 a, int32_t shift) | |
4175 | { | |
4176 | tcg_gen_deposit_i32(d, d, a, shift, 32 - shift); | |
4177 | } | |
4178 | ||
4179 | static void gen_shl64_ins_i64(TCGv_i64 d, TCGv_i64 a, int64_t shift) | |
4180 | { | |
4181 | tcg_gen_deposit_i64(d, d, a, shift, 64 - shift); | |
4182 | } | |
4183 | ||
4184 | static void gen_shl_ins_vec(unsigned vece, TCGv_vec d, TCGv_vec a, int64_t sh) | |
4185 | { | |
4186 | if (sh == 0) { | |
4187 | tcg_gen_mov_vec(d, a); | |
4188 | } else { | |
4189 | TCGv_vec t = tcg_temp_new_vec_matching(d); | |
4190 | TCGv_vec m = tcg_temp_new_vec_matching(d); | |
4191 | ||
4192 | tcg_gen_dupi_vec(vece, m, MAKE_64BIT_MASK(0, sh)); | |
4193 | tcg_gen_shli_vec(vece, t, a, sh); | |
4194 | tcg_gen_and_vec(vece, d, d, m); | |
4195 | tcg_gen_or_vec(vece, d, d, t); | |
4196 | ||
4197 | tcg_temp_free_vec(t); | |
4198 | tcg_temp_free_vec(m); | |
4199 | } | |
4200 | } | |
4201 | ||
53229a77 RH |
4202 | static const TCGOpcode vecop_list_sli[] = { INDEX_op_shli_vec, 0 }; |
4203 | ||
f3cd8218 RH |
4204 | const GVecGen2i sli_op[4] = { |
4205 | { .fni8 = gen_shl8_ins_i64, | |
4206 | .fniv = gen_shl_ins_vec, | |
4207 | .load_dest = true, | |
53229a77 | 4208 | .opt_opc = vecop_list_sli, |
f3cd8218 RH |
4209 | .vece = MO_8 }, |
4210 | { .fni8 = gen_shl16_ins_i64, | |
4211 | .fniv = gen_shl_ins_vec, | |
4212 | .load_dest = true, | |
53229a77 | 4213 | .opt_opc = vecop_list_sli, |
f3cd8218 RH |
4214 | .vece = MO_16 }, |
4215 | { .fni4 = gen_shl32_ins_i32, | |
4216 | .fniv = gen_shl_ins_vec, | |
4217 | .load_dest = true, | |
53229a77 | 4218 | .opt_opc = vecop_list_sli, |
f3cd8218 RH |
4219 | .vece = MO_32 }, |
4220 | { .fni8 = gen_shl64_ins_i64, | |
4221 | .fniv = gen_shl_ins_vec, | |
4222 | .prefer_i64 = TCG_TARGET_REG_BITS == 64, | |
4223 | .load_dest = true, | |
53229a77 | 4224 | .opt_opc = vecop_list_sli, |
f3cd8218 RH |
4225 | .vece = MO_64 }, |
4226 | }; | |
4227 | ||
4a7832b0 RH |
4228 | static void gen_mla8_i32(TCGv_i32 d, TCGv_i32 a, TCGv_i32 b) |
4229 | { | |
4230 | gen_helper_neon_mul_u8(a, a, b); | |
4231 | gen_helper_neon_add_u8(d, d, a); | |
4232 | } | |
4233 | ||
4234 | static void gen_mls8_i32(TCGv_i32 d, TCGv_i32 a, TCGv_i32 b) | |
4235 | { | |
4236 | gen_helper_neon_mul_u8(a, a, b); | |
4237 | gen_helper_neon_sub_u8(d, d, a); | |
4238 | } | |
4239 | ||
4240 | static void gen_mla16_i32(TCGv_i32 d, TCGv_i32 a, TCGv_i32 b) | |
4241 | { | |
4242 | gen_helper_neon_mul_u16(a, a, b); | |
4243 | gen_helper_neon_add_u16(d, d, a); | |
4244 | } | |
4245 | ||
4246 | static void gen_mls16_i32(TCGv_i32 d, TCGv_i32 a, TCGv_i32 b) | |
4247 | { | |
4248 | gen_helper_neon_mul_u16(a, a, b); | |
4249 | gen_helper_neon_sub_u16(d, d, a); | |
4250 | } | |
4251 | ||
4252 | static void gen_mla32_i32(TCGv_i32 d, TCGv_i32 a, TCGv_i32 b) | |
4253 | { | |
4254 | tcg_gen_mul_i32(a, a, b); | |
4255 | tcg_gen_add_i32(d, d, a); | |
4256 | } | |
4257 | ||
4258 | static void gen_mls32_i32(TCGv_i32 d, TCGv_i32 a, TCGv_i32 b) | |
4259 | { | |
4260 | tcg_gen_mul_i32(a, a, b); | |
4261 | tcg_gen_sub_i32(d, d, a); | |
4262 | } | |
4263 | ||
4264 | static void gen_mla64_i64(TCGv_i64 d, TCGv_i64 a, TCGv_i64 b) | |
4265 | { | |
4266 | tcg_gen_mul_i64(a, a, b); | |
4267 | tcg_gen_add_i64(d, d, a); | |
4268 | } | |
4269 | ||
4270 | static void gen_mls64_i64(TCGv_i64 d, TCGv_i64 a, TCGv_i64 b) | |
4271 | { | |
4272 | tcg_gen_mul_i64(a, a, b); | |
4273 | tcg_gen_sub_i64(d, d, a); | |
4274 | } | |
4275 | ||
4276 | static void gen_mla_vec(unsigned vece, TCGv_vec d, TCGv_vec a, TCGv_vec b) | |
4277 | { | |
4278 | tcg_gen_mul_vec(vece, a, a, b); | |
4279 | tcg_gen_add_vec(vece, d, d, a); | |
4280 | } | |
4281 | ||
4282 | static void gen_mls_vec(unsigned vece, TCGv_vec d, TCGv_vec a, TCGv_vec b) | |
4283 | { | |
4284 | tcg_gen_mul_vec(vece, a, a, b); | |
4285 | tcg_gen_sub_vec(vece, d, d, a); | |
4286 | } | |
4287 | ||
4288 | /* Note that while NEON does not support VMLA and VMLS as 64-bit ops, | |
4289 | * these tables are shared with AArch64 which does support them. | |
4290 | */ | |
53229a77 RH |
4291 | |
4292 | static const TCGOpcode vecop_list_mla[] = { | |
4293 | INDEX_op_mul_vec, INDEX_op_add_vec, 0 | |
4294 | }; | |
4295 | ||
4296 | static const TCGOpcode vecop_list_mls[] = { | |
4297 | INDEX_op_mul_vec, INDEX_op_sub_vec, 0 | |
4298 | }; | |
4299 | ||
4a7832b0 RH |
4300 | const GVecGen3 mla_op[4] = { |
4301 | { .fni4 = gen_mla8_i32, | |
4302 | .fniv = gen_mla_vec, | |
4a7832b0 | 4303 | .load_dest = true, |
53229a77 | 4304 | .opt_opc = vecop_list_mla, |
4a7832b0 RH |
4305 | .vece = MO_8 }, |
4306 | { .fni4 = gen_mla16_i32, | |
4307 | .fniv = gen_mla_vec, | |
4a7832b0 | 4308 | .load_dest = true, |
53229a77 | 4309 | .opt_opc = vecop_list_mla, |
4a7832b0 RH |
4310 | .vece = MO_16 }, |
4311 | { .fni4 = gen_mla32_i32, | |
4312 | .fniv = gen_mla_vec, | |
4a7832b0 | 4313 | .load_dest = true, |
53229a77 | 4314 | .opt_opc = vecop_list_mla, |
4a7832b0 RH |
4315 | .vece = MO_32 }, |
4316 | { .fni8 = gen_mla64_i64, | |
4317 | .fniv = gen_mla_vec, | |
4a7832b0 RH |
4318 | .prefer_i64 = TCG_TARGET_REG_BITS == 64, |
4319 | .load_dest = true, | |
53229a77 | 4320 | .opt_opc = vecop_list_mla, |
4a7832b0 RH |
4321 | .vece = MO_64 }, |
4322 | }; | |
4323 | ||
4324 | const GVecGen3 mls_op[4] = { | |
4325 | { .fni4 = gen_mls8_i32, | |
4326 | .fniv = gen_mls_vec, | |
4a7832b0 | 4327 | .load_dest = true, |
53229a77 | 4328 | .opt_opc = vecop_list_mls, |
4a7832b0 RH |
4329 | .vece = MO_8 }, |
4330 | { .fni4 = gen_mls16_i32, | |
4331 | .fniv = gen_mls_vec, | |
4a7832b0 | 4332 | .load_dest = true, |
53229a77 | 4333 | .opt_opc = vecop_list_mls, |
4a7832b0 RH |
4334 | .vece = MO_16 }, |
4335 | { .fni4 = gen_mls32_i32, | |
4336 | .fniv = gen_mls_vec, | |
4a7832b0 | 4337 | .load_dest = true, |
53229a77 | 4338 | .opt_opc = vecop_list_mls, |
4a7832b0 RH |
4339 | .vece = MO_32 }, |
4340 | { .fni8 = gen_mls64_i64, | |
4341 | .fniv = gen_mls_vec, | |
4a7832b0 RH |
4342 | .prefer_i64 = TCG_TARGET_REG_BITS == 64, |
4343 | .load_dest = true, | |
53229a77 | 4344 | .opt_opc = vecop_list_mls, |
4a7832b0 RH |
4345 | .vece = MO_64 }, |
4346 | }; | |
4347 | ||
ea580fa3 RH |
4348 | /* CMTST : test is "if (X & Y != 0)". */ |
4349 | static void gen_cmtst_i32(TCGv_i32 d, TCGv_i32 a, TCGv_i32 b) | |
4350 | { | |
4351 | tcg_gen_and_i32(d, a, b); | |
4352 | tcg_gen_setcondi_i32(TCG_COND_NE, d, d, 0); | |
4353 | tcg_gen_neg_i32(d, d); | |
4354 | } | |
4355 | ||
4356 | void gen_cmtst_i64(TCGv_i64 d, TCGv_i64 a, TCGv_i64 b) | |
4357 | { | |
4358 | tcg_gen_and_i64(d, a, b); | |
4359 | tcg_gen_setcondi_i64(TCG_COND_NE, d, d, 0); | |
4360 | tcg_gen_neg_i64(d, d); | |
4361 | } | |
4362 | ||
4363 | static void gen_cmtst_vec(unsigned vece, TCGv_vec d, TCGv_vec a, TCGv_vec b) | |
4364 | { | |
4365 | tcg_gen_and_vec(vece, d, a, b); | |
4366 | tcg_gen_dupi_vec(vece, a, 0); | |
4367 | tcg_gen_cmp_vec(TCG_COND_NE, vece, d, d, a); | |
4368 | } | |
4369 | ||
53229a77 RH |
4370 | static const TCGOpcode vecop_list_cmtst[] = { INDEX_op_cmp_vec, 0 }; |
4371 | ||
ea580fa3 RH |
4372 | const GVecGen3 cmtst_op[4] = { |
4373 | { .fni4 = gen_helper_neon_tst_u8, | |
4374 | .fniv = gen_cmtst_vec, | |
53229a77 | 4375 | .opt_opc = vecop_list_cmtst, |
ea580fa3 RH |
4376 | .vece = MO_8 }, |
4377 | { .fni4 = gen_helper_neon_tst_u16, | |
4378 | .fniv = gen_cmtst_vec, | |
53229a77 | 4379 | .opt_opc = vecop_list_cmtst, |
ea580fa3 RH |
4380 | .vece = MO_16 }, |
4381 | { .fni4 = gen_cmtst_i32, | |
4382 | .fniv = gen_cmtst_vec, | |
53229a77 | 4383 | .opt_opc = vecop_list_cmtst, |
ea580fa3 RH |
4384 | .vece = MO_32 }, |
4385 | { .fni8 = gen_cmtst_i64, | |
4386 | .fniv = gen_cmtst_vec, | |
4387 | .prefer_i64 = TCG_TARGET_REG_BITS == 64, | |
53229a77 | 4388 | .opt_opc = vecop_list_cmtst, |
ea580fa3 RH |
4389 | .vece = MO_64 }, |
4390 | }; | |
4391 | ||
89e68b57 RH |
4392 | static void gen_uqadd_vec(unsigned vece, TCGv_vec t, TCGv_vec sat, |
4393 | TCGv_vec a, TCGv_vec b) | |
4394 | { | |
4395 | TCGv_vec x = tcg_temp_new_vec_matching(t); | |
4396 | tcg_gen_add_vec(vece, x, a, b); | |
4397 | tcg_gen_usadd_vec(vece, t, a, b); | |
4398 | tcg_gen_cmp_vec(TCG_COND_NE, vece, x, x, t); | |
4399 | tcg_gen_or_vec(vece, sat, sat, x); | |
4400 | tcg_temp_free_vec(x); | |
4401 | } | |
4402 | ||
53229a77 RH |
4403 | static const TCGOpcode vecop_list_uqadd[] = { |
4404 | INDEX_op_usadd_vec, INDEX_op_cmp_vec, INDEX_op_add_vec, 0 | |
4405 | }; | |
4406 | ||
89e68b57 RH |
4407 | const GVecGen4 uqadd_op[4] = { |
4408 | { .fniv = gen_uqadd_vec, | |
4409 | .fno = gen_helper_gvec_uqadd_b, | |
89e68b57 | 4410 | .write_aofs = true, |
53229a77 | 4411 | .opt_opc = vecop_list_uqadd, |
89e68b57 RH |
4412 | .vece = MO_8 }, |
4413 | { .fniv = gen_uqadd_vec, | |
4414 | .fno = gen_helper_gvec_uqadd_h, | |
89e68b57 | 4415 | .write_aofs = true, |
53229a77 | 4416 | .opt_opc = vecop_list_uqadd, |
89e68b57 RH |
4417 | .vece = MO_16 }, |
4418 | { .fniv = gen_uqadd_vec, | |
4419 | .fno = gen_helper_gvec_uqadd_s, | |
89e68b57 | 4420 | .write_aofs = true, |
53229a77 | 4421 | .opt_opc = vecop_list_uqadd, |
89e68b57 RH |
4422 | .vece = MO_32 }, |
4423 | { .fniv = gen_uqadd_vec, | |
4424 | .fno = gen_helper_gvec_uqadd_d, | |
89e68b57 | 4425 | .write_aofs = true, |
53229a77 | 4426 | .opt_opc = vecop_list_uqadd, |
89e68b57 RH |
4427 | .vece = MO_64 }, |
4428 | }; | |
4429 | ||
4430 | static void gen_sqadd_vec(unsigned vece, TCGv_vec t, TCGv_vec sat, | |
4431 | TCGv_vec a, TCGv_vec b) | |
4432 | { | |
4433 | TCGv_vec x = tcg_temp_new_vec_matching(t); | |
4434 | tcg_gen_add_vec(vece, x, a, b); | |
4435 | tcg_gen_ssadd_vec(vece, t, a, b); | |
4436 | tcg_gen_cmp_vec(TCG_COND_NE, vece, x, x, t); | |
4437 | tcg_gen_or_vec(vece, sat, sat, x); | |
4438 | tcg_temp_free_vec(x); | |
4439 | } | |
4440 | ||
53229a77 RH |
4441 | static const TCGOpcode vecop_list_sqadd[] = { |
4442 | INDEX_op_ssadd_vec, INDEX_op_cmp_vec, INDEX_op_add_vec, 0 | |
4443 | }; | |
4444 | ||
89e68b57 RH |
4445 | const GVecGen4 sqadd_op[4] = { |
4446 | { .fniv = gen_sqadd_vec, | |
4447 | .fno = gen_helper_gvec_sqadd_b, | |
53229a77 | 4448 | .opt_opc = vecop_list_sqadd, |
89e68b57 RH |
4449 | .write_aofs = true, |
4450 | .vece = MO_8 }, | |
4451 | { .fniv = gen_sqadd_vec, | |
4452 | .fno = gen_helper_gvec_sqadd_h, | |
53229a77 | 4453 | .opt_opc = vecop_list_sqadd, |
89e68b57 RH |
4454 | .write_aofs = true, |
4455 | .vece = MO_16 }, | |
4456 | { .fniv = gen_sqadd_vec, | |
4457 | .fno = gen_helper_gvec_sqadd_s, | |
53229a77 | 4458 | .opt_opc = vecop_list_sqadd, |
89e68b57 RH |
4459 | .write_aofs = true, |
4460 | .vece = MO_32 }, | |
4461 | { .fniv = gen_sqadd_vec, | |
4462 | .fno = gen_helper_gvec_sqadd_d, | |
53229a77 | 4463 | .opt_opc = vecop_list_sqadd, |
89e68b57 RH |
4464 | .write_aofs = true, |
4465 | .vece = MO_64 }, | |
4466 | }; | |
4467 | ||
4468 | static void gen_uqsub_vec(unsigned vece, TCGv_vec t, TCGv_vec sat, | |
4469 | TCGv_vec a, TCGv_vec b) | |
4470 | { | |
4471 | TCGv_vec x = tcg_temp_new_vec_matching(t); | |
4472 | tcg_gen_sub_vec(vece, x, a, b); | |
4473 | tcg_gen_ussub_vec(vece, t, a, b); | |
4474 | tcg_gen_cmp_vec(TCG_COND_NE, vece, x, x, t); | |
4475 | tcg_gen_or_vec(vece, sat, sat, x); | |
4476 | tcg_temp_free_vec(x); | |
4477 | } | |
4478 | ||
53229a77 RH |
4479 | static const TCGOpcode vecop_list_uqsub[] = { |
4480 | INDEX_op_ussub_vec, INDEX_op_cmp_vec, INDEX_op_sub_vec, 0 | |
4481 | }; | |
4482 | ||
89e68b57 RH |
4483 | const GVecGen4 uqsub_op[4] = { |
4484 | { .fniv = gen_uqsub_vec, | |
4485 | .fno = gen_helper_gvec_uqsub_b, | |
53229a77 | 4486 | .opt_opc = vecop_list_uqsub, |
89e68b57 RH |
4487 | .write_aofs = true, |
4488 | .vece = MO_8 }, | |
4489 | { .fniv = gen_uqsub_vec, | |
4490 | .fno = gen_helper_gvec_uqsub_h, | |
53229a77 | 4491 | .opt_opc = vecop_list_uqsub, |
89e68b57 RH |
4492 | .write_aofs = true, |
4493 | .vece = MO_16 }, | |
4494 | { .fniv = gen_uqsub_vec, | |
4495 | .fno = gen_helper_gvec_uqsub_s, | |
53229a77 | 4496 | .opt_opc = vecop_list_uqsub, |
89e68b57 RH |
4497 | .write_aofs = true, |
4498 | .vece = MO_32 }, | |
4499 | { .fniv = gen_uqsub_vec, | |
4500 | .fno = gen_helper_gvec_uqsub_d, | |
53229a77 | 4501 | .opt_opc = vecop_list_uqsub, |
89e68b57 RH |
4502 | .write_aofs = true, |
4503 | .vece = MO_64 }, | |
4504 | }; | |
4505 | ||
4506 | static void gen_sqsub_vec(unsigned vece, TCGv_vec t, TCGv_vec sat, | |
4507 | TCGv_vec a, TCGv_vec b) | |
4508 | { | |
4509 | TCGv_vec x = tcg_temp_new_vec_matching(t); | |
4510 | tcg_gen_sub_vec(vece, x, a, b); | |
4511 | tcg_gen_sssub_vec(vece, t, a, b); | |
4512 | tcg_gen_cmp_vec(TCG_COND_NE, vece, x, x, t); | |
4513 | tcg_gen_or_vec(vece, sat, sat, x); | |
4514 | tcg_temp_free_vec(x); | |
4515 | } | |
4516 | ||
53229a77 RH |
4517 | static const TCGOpcode vecop_list_sqsub[] = { |
4518 | INDEX_op_sssub_vec, INDEX_op_cmp_vec, INDEX_op_sub_vec, 0 | |
4519 | }; | |
4520 | ||
89e68b57 RH |
4521 | const GVecGen4 sqsub_op[4] = { |
4522 | { .fniv = gen_sqsub_vec, | |
4523 | .fno = gen_helper_gvec_sqsub_b, | |
53229a77 | 4524 | .opt_opc = vecop_list_sqsub, |
89e68b57 RH |
4525 | .write_aofs = true, |
4526 | .vece = MO_8 }, | |
4527 | { .fniv = gen_sqsub_vec, | |
4528 | .fno = gen_helper_gvec_sqsub_h, | |
53229a77 | 4529 | .opt_opc = vecop_list_sqsub, |
89e68b57 RH |
4530 | .write_aofs = true, |
4531 | .vece = MO_16 }, | |
4532 | { .fniv = gen_sqsub_vec, | |
4533 | .fno = gen_helper_gvec_sqsub_s, | |
53229a77 | 4534 | .opt_opc = vecop_list_sqsub, |
89e68b57 RH |
4535 | .write_aofs = true, |
4536 | .vece = MO_32 }, | |
4537 | { .fniv = gen_sqsub_vec, | |
4538 | .fno = gen_helper_gvec_sqsub_d, | |
53229a77 | 4539 | .opt_opc = vecop_list_sqsub, |
89e68b57 RH |
4540 | .write_aofs = true, |
4541 | .vece = MO_64 }, | |
4542 | }; | |
4543 | ||
9ee6e8bb PB |
4544 | /* Translate a NEON data processing instruction. Return nonzero if the |
4545 | instruction is invalid. | |
ad69471c PB |
4546 | We process data in a mixture of 32-bit and 64-bit chunks. |
4547 | Mostly we use 32-bit chunks so we can use normal scalar instructions. */ | |
2c0262af | 4548 | |
7dcc1f89 | 4549 | static int disas_neon_data_insn(DisasContext *s, uint32_t insn) |
9ee6e8bb PB |
4550 | { |
4551 | int op; | |
4552 | int q; | |
eabcd6fa | 4553 | int rd, rn, rm, rd_ofs, rn_ofs, rm_ofs; |
9ee6e8bb PB |
4554 | int size; |
4555 | int shift; | |
4556 | int pass; | |
4557 | int count; | |
4558 | int pairwise; | |
4559 | int u; | |
eabcd6fa | 4560 | int vec_size; |
f3cd8218 | 4561 | uint32_t imm; |
39d5492a | 4562 | TCGv_i32 tmp, tmp2, tmp3, tmp4, tmp5; |
1a66ac61 | 4563 | TCGv_ptr ptr1, ptr2, ptr3; |
a7812ae4 | 4564 | TCGv_i64 tmp64; |
9ee6e8bb | 4565 | |
2c7ffc41 PM |
4566 | /* FIXME: this access check should not take precedence over UNDEF |
4567 | * for invalid encodings; we will generate incorrect syndrome information | |
4568 | * for attempts to execute invalid vfp/neon encodings with FP disabled. | |
4569 | */ | |
9dbbc748 | 4570 | if (s->fp_excp_el) { |
a767fac8 | 4571 | gen_exception_insn(s, s->pc_curr, EXCP_UDEF, |
4be42f40 | 4572 | syn_simd_access_trap(1, 0xe, false), s->fp_excp_el); |
2c7ffc41 PM |
4573 | return 0; |
4574 | } | |
4575 | ||
5df8bac1 | 4576 | if (!s->vfp_enabled) |
9ee6e8bb PB |
4577 | return 1; |
4578 | q = (insn & (1 << 6)) != 0; | |
4579 | u = (insn >> 24) & 1; | |
4580 | VFP_DREG_D(rd, insn); | |
4581 | VFP_DREG_N(rn, insn); | |
4582 | VFP_DREG_M(rm, insn); | |
4583 | size = (insn >> 20) & 3; | |
eabcd6fa RH |
4584 | vec_size = q ? 16 : 8; |
4585 | rd_ofs = neon_reg_offset(rd, 0); | |
4586 | rn_ofs = neon_reg_offset(rn, 0); | |
4587 | rm_ofs = neon_reg_offset(rm, 0); | |
4588 | ||
9ee6e8bb PB |
4589 | if ((insn & (1 << 23)) == 0) { |
4590 | /* Three register same length. */ | |
4591 | op = ((insn >> 7) & 0x1e) | ((insn >> 4) & 1); | |
62698be3 PM |
4592 | /* Catch invalid op and bad size combinations: UNDEF */ |
4593 | if ((neon_3r_sizes[op] & (1 << size)) == 0) { | |
4594 | return 1; | |
4595 | } | |
25f84f79 PM |
4596 | /* All insns of this form UNDEF for either this condition or the |
4597 | * superset of cases "Q==1"; we catch the latter later. | |
4598 | */ | |
4599 | if (q && ((rd | rn | rm) & 1)) { | |
4600 | return 1; | |
4601 | } | |
36a71934 RH |
4602 | switch (op) { |
4603 | case NEON_3R_SHA: | |
4604 | /* The SHA-1/SHA-256 3-register instructions require special | |
4605 | * treatment here, as their size field is overloaded as an | |
4606 | * op type selector, and they all consume their input in a | |
4607 | * single pass. | |
4608 | */ | |
f1ecb913 AB |
4609 | if (!q) { |
4610 | return 1; | |
4611 | } | |
4612 | if (!u) { /* SHA-1 */ | |
962fcbf2 | 4613 | if (!dc_isar_feature(aa32_sha1, s)) { |
f1ecb913 AB |
4614 | return 1; |
4615 | } | |
1a66ac61 RH |
4616 | ptr1 = vfp_reg_ptr(true, rd); |
4617 | ptr2 = vfp_reg_ptr(true, rn); | |
4618 | ptr3 = vfp_reg_ptr(true, rm); | |
f1ecb913 | 4619 | tmp4 = tcg_const_i32(size); |
1a66ac61 | 4620 | gen_helper_crypto_sha1_3reg(ptr1, ptr2, ptr3, tmp4); |
f1ecb913 AB |
4621 | tcg_temp_free_i32(tmp4); |
4622 | } else { /* SHA-256 */ | |
962fcbf2 | 4623 | if (!dc_isar_feature(aa32_sha2, s) || size == 3) { |
f1ecb913 AB |
4624 | return 1; |
4625 | } | |
1a66ac61 RH |
4626 | ptr1 = vfp_reg_ptr(true, rd); |
4627 | ptr2 = vfp_reg_ptr(true, rn); | |
4628 | ptr3 = vfp_reg_ptr(true, rm); | |
f1ecb913 AB |
4629 | switch (size) { |
4630 | case 0: | |
1a66ac61 | 4631 | gen_helper_crypto_sha256h(ptr1, ptr2, ptr3); |
f1ecb913 AB |
4632 | break; |
4633 | case 1: | |
1a66ac61 | 4634 | gen_helper_crypto_sha256h2(ptr1, ptr2, ptr3); |
f1ecb913 AB |
4635 | break; |
4636 | case 2: | |
1a66ac61 | 4637 | gen_helper_crypto_sha256su1(ptr1, ptr2, ptr3); |
f1ecb913 AB |
4638 | break; |
4639 | } | |
4640 | } | |
1a66ac61 RH |
4641 | tcg_temp_free_ptr(ptr1); |
4642 | tcg_temp_free_ptr(ptr2); | |
4643 | tcg_temp_free_ptr(ptr3); | |
f1ecb913 | 4644 | return 0; |
36a71934 RH |
4645 | |
4646 | case NEON_3R_VPADD_VQRDMLAH: | |
4647 | if (!u) { | |
4648 | break; /* VPADD */ | |
4649 | } | |
4650 | /* VQRDMLAH */ | |
4651 | switch (size) { | |
4652 | case 1: | |
4653 | return do_v81_helper(s, gen_helper_gvec_qrdmlah_s16, | |
4654 | q, rd, rn, rm); | |
4655 | case 2: | |
4656 | return do_v81_helper(s, gen_helper_gvec_qrdmlah_s32, | |
4657 | q, rd, rn, rm); | |
4658 | } | |
4659 | return 1; | |
4660 | ||
4661 | case NEON_3R_VFM_VQRDMLSH: | |
4662 | if (!u) { | |
4663 | /* VFM, VFMS */ | |
4664 | if (size == 1) { | |
4665 | return 1; | |
4666 | } | |
4667 | break; | |
4668 | } | |
4669 | /* VQRDMLSH */ | |
4670 | switch (size) { | |
4671 | case 1: | |
4672 | return do_v81_helper(s, gen_helper_gvec_qrdmlsh_s16, | |
4673 | q, rd, rn, rm); | |
4674 | case 2: | |
4675 | return do_v81_helper(s, gen_helper_gvec_qrdmlsh_s32, | |
4676 | q, rd, rn, rm); | |
4677 | } | |
4678 | return 1; | |
eabcd6fa RH |
4679 | |
4680 | case NEON_3R_LOGIC: /* Logic ops. */ | |
4681 | switch ((u << 2) | size) { | |
4682 | case 0: /* VAND */ | |
4683 | tcg_gen_gvec_and(0, rd_ofs, rn_ofs, rm_ofs, | |
4684 | vec_size, vec_size); | |
4685 | break; | |
4686 | case 1: /* VBIC */ | |
4687 | tcg_gen_gvec_andc(0, rd_ofs, rn_ofs, rm_ofs, | |
4688 | vec_size, vec_size); | |
4689 | break; | |
2900847f RH |
4690 | case 2: /* VORR */ |
4691 | tcg_gen_gvec_or(0, rd_ofs, rn_ofs, rm_ofs, | |
4692 | vec_size, vec_size); | |
eabcd6fa RH |
4693 | break; |
4694 | case 3: /* VORN */ | |
4695 | tcg_gen_gvec_orc(0, rd_ofs, rn_ofs, rm_ofs, | |
4696 | vec_size, vec_size); | |
4697 | break; | |
4698 | case 4: /* VEOR */ | |
4699 | tcg_gen_gvec_xor(0, rd_ofs, rn_ofs, rm_ofs, | |
4700 | vec_size, vec_size); | |
4701 | break; | |
4702 | case 5: /* VBSL */ | |
3a7a2b4e RH |
4703 | tcg_gen_gvec_bitsel(MO_8, rd_ofs, rd_ofs, rn_ofs, rm_ofs, |
4704 | vec_size, vec_size); | |
eabcd6fa RH |
4705 | break; |
4706 | case 6: /* VBIT */ | |
3a7a2b4e RH |
4707 | tcg_gen_gvec_bitsel(MO_8, rd_ofs, rm_ofs, rn_ofs, rd_ofs, |
4708 | vec_size, vec_size); | |
eabcd6fa RH |
4709 | break; |
4710 | case 7: /* VBIF */ | |
3a7a2b4e RH |
4711 | tcg_gen_gvec_bitsel(MO_8, rd_ofs, rm_ofs, rd_ofs, rn_ofs, |
4712 | vec_size, vec_size); | |
eabcd6fa RH |
4713 | break; |
4714 | } | |
4715 | return 0; | |
e4717ae0 RH |
4716 | |
4717 | case NEON_3R_VADD_VSUB: | |
4718 | if (u) { | |
4719 | tcg_gen_gvec_sub(size, rd_ofs, rn_ofs, rm_ofs, | |
4720 | vec_size, vec_size); | |
4721 | } else { | |
4722 | tcg_gen_gvec_add(size, rd_ofs, rn_ofs, rm_ofs, | |
4723 | vec_size, vec_size); | |
4724 | } | |
4725 | return 0; | |
82083184 | 4726 | |
89e68b57 RH |
4727 | case NEON_3R_VQADD: |
4728 | tcg_gen_gvec_4(rd_ofs, offsetof(CPUARMState, vfp.qc), | |
4729 | rn_ofs, rm_ofs, vec_size, vec_size, | |
4730 | (u ? uqadd_op : sqadd_op) + size); | |
2f143d3a | 4731 | return 0; |
89e68b57 RH |
4732 | |
4733 | case NEON_3R_VQSUB: | |
4734 | tcg_gen_gvec_4(rd_ofs, offsetof(CPUARMState, vfp.qc), | |
4735 | rn_ofs, rm_ofs, vec_size, vec_size, | |
4736 | (u ? uqsub_op : sqsub_op) + size); | |
2f143d3a | 4737 | return 0; |
89e68b57 | 4738 | |
82083184 RH |
4739 | case NEON_3R_VMUL: /* VMUL */ |
4740 | if (u) { | |
4741 | /* Polynomial case allows only P8 and is handled below. */ | |
4742 | if (size != 0) { | |
4743 | return 1; | |
4744 | } | |
4745 | } else { | |
4746 | tcg_gen_gvec_mul(size, rd_ofs, rn_ofs, rm_ofs, | |
4747 | vec_size, vec_size); | |
4748 | return 0; | |
4749 | } | |
4750 | break; | |
4a7832b0 RH |
4751 | |
4752 | case NEON_3R_VML: /* VMLA, VMLS */ | |
4753 | tcg_gen_gvec_3(rd_ofs, rn_ofs, rm_ofs, vec_size, vec_size, | |
4754 | u ? &mls_op[size] : &mla_op[size]); | |
4755 | return 0; | |
ea580fa3 RH |
4756 | |
4757 | case NEON_3R_VTST_VCEQ: | |
4758 | if (u) { /* VCEQ */ | |
4759 | tcg_gen_gvec_cmp(TCG_COND_EQ, size, rd_ofs, rn_ofs, rm_ofs, | |
4760 | vec_size, vec_size); | |
4761 | } else { /* VTST */ | |
4762 | tcg_gen_gvec_3(rd_ofs, rn_ofs, rm_ofs, | |
4763 | vec_size, vec_size, &cmtst_op[size]); | |
4764 | } | |
4765 | return 0; | |
4766 | ||
4767 | case NEON_3R_VCGT: | |
4768 | tcg_gen_gvec_cmp(u ? TCG_COND_GTU : TCG_COND_GT, size, | |
4769 | rd_ofs, rn_ofs, rm_ofs, vec_size, vec_size); | |
4770 | return 0; | |
4771 | ||
4772 | case NEON_3R_VCGE: | |
4773 | tcg_gen_gvec_cmp(u ? TCG_COND_GEU : TCG_COND_GE, size, | |
4774 | rd_ofs, rn_ofs, rm_ofs, vec_size, vec_size); | |
4775 | return 0; | |
6f278221 RH |
4776 | |
4777 | case NEON_3R_VMAX: | |
4778 | if (u) { | |
4779 | tcg_gen_gvec_umax(size, rd_ofs, rn_ofs, rm_ofs, | |
4780 | vec_size, vec_size); | |
4781 | } else { | |
4782 | tcg_gen_gvec_smax(size, rd_ofs, rn_ofs, rm_ofs, | |
4783 | vec_size, vec_size); | |
4784 | } | |
4785 | return 0; | |
4786 | case NEON_3R_VMIN: | |
4787 | if (u) { | |
4788 | tcg_gen_gvec_umin(size, rd_ofs, rn_ofs, rm_ofs, | |
4789 | vec_size, vec_size); | |
4790 | } else { | |
4791 | tcg_gen_gvec_smin(size, rd_ofs, rn_ofs, rm_ofs, | |
4792 | vec_size, vec_size); | |
4793 | } | |
4794 | return 0; | |
f1ecb913 | 4795 | } |
4a7832b0 | 4796 | |
eabcd6fa | 4797 | if (size == 3) { |
62698be3 | 4798 | /* 64-bit element instructions. */ |
9ee6e8bb | 4799 | for (pass = 0; pass < (q ? 2 : 1); pass++) { |
ad69471c PB |
4800 | neon_load_reg64(cpu_V0, rn + pass); |
4801 | neon_load_reg64(cpu_V1, rm + pass); | |
9ee6e8bb | 4802 | switch (op) { |
62698be3 | 4803 | case NEON_3R_VSHL: |
ad69471c PB |
4804 | if (u) { |
4805 | gen_helper_neon_shl_u64(cpu_V0, cpu_V1, cpu_V0); | |
4806 | } else { | |
4807 | gen_helper_neon_shl_s64(cpu_V0, cpu_V1, cpu_V0); | |
4808 | } | |
4809 | break; | |
62698be3 | 4810 | case NEON_3R_VQSHL: |
ad69471c | 4811 | if (u) { |
02da0b2d PM |
4812 | gen_helper_neon_qshl_u64(cpu_V0, cpu_env, |
4813 | cpu_V1, cpu_V0); | |
ad69471c | 4814 | } else { |
02da0b2d PM |
4815 | gen_helper_neon_qshl_s64(cpu_V0, cpu_env, |
4816 | cpu_V1, cpu_V0); | |
ad69471c PB |
4817 | } |
4818 | break; | |
62698be3 | 4819 | case NEON_3R_VRSHL: |
ad69471c PB |
4820 | if (u) { |
4821 | gen_helper_neon_rshl_u64(cpu_V0, cpu_V1, cpu_V0); | |
1e8d4eec | 4822 | } else { |
ad69471c PB |
4823 | gen_helper_neon_rshl_s64(cpu_V0, cpu_V1, cpu_V0); |
4824 | } | |
4825 | break; | |
62698be3 | 4826 | case NEON_3R_VQRSHL: |
ad69471c | 4827 | if (u) { |
02da0b2d PM |
4828 | gen_helper_neon_qrshl_u64(cpu_V0, cpu_env, |
4829 | cpu_V1, cpu_V0); | |
ad69471c | 4830 | } else { |
02da0b2d PM |
4831 | gen_helper_neon_qrshl_s64(cpu_V0, cpu_env, |
4832 | cpu_V1, cpu_V0); | |
1e8d4eec | 4833 | } |
9ee6e8bb | 4834 | break; |
9ee6e8bb PB |
4835 | default: |
4836 | abort(); | |
2c0262af | 4837 | } |
ad69471c | 4838 | neon_store_reg64(cpu_V0, rd + pass); |
2c0262af | 4839 | } |
9ee6e8bb | 4840 | return 0; |
2c0262af | 4841 | } |
25f84f79 | 4842 | pairwise = 0; |
9ee6e8bb | 4843 | switch (op) { |
62698be3 PM |
4844 | case NEON_3R_VSHL: |
4845 | case NEON_3R_VQSHL: | |
4846 | case NEON_3R_VRSHL: | |
4847 | case NEON_3R_VQRSHL: | |
9ee6e8bb | 4848 | { |
ad69471c PB |
4849 | int rtmp; |
4850 | /* Shift instruction operands are reversed. */ | |
4851 | rtmp = rn; | |
9ee6e8bb | 4852 | rn = rm; |
ad69471c | 4853 | rm = rtmp; |
9ee6e8bb | 4854 | } |
2c0262af | 4855 | break; |
36a71934 | 4856 | case NEON_3R_VPADD_VQRDMLAH: |
62698be3 PM |
4857 | case NEON_3R_VPMAX: |
4858 | case NEON_3R_VPMIN: | |
9ee6e8bb | 4859 | pairwise = 1; |
2c0262af | 4860 | break; |
25f84f79 PM |
4861 | case NEON_3R_FLOAT_ARITH: |
4862 | pairwise = (u && size < 2); /* if VPADD (float) */ | |
4863 | break; | |
4864 | case NEON_3R_FLOAT_MINMAX: | |
4865 | pairwise = u; /* if VPMIN/VPMAX (float) */ | |
4866 | break; | |
4867 | case NEON_3R_FLOAT_CMP: | |
4868 | if (!u && size) { | |
4869 | /* no encoding for U=0 C=1x */ | |
4870 | return 1; | |
4871 | } | |
4872 | break; | |
4873 | case NEON_3R_FLOAT_ACMP: | |
4874 | if (!u) { | |
4875 | return 1; | |
4876 | } | |
4877 | break; | |
505935fc WN |
4878 | case NEON_3R_FLOAT_MISC: |
4879 | /* VMAXNM/VMINNM in ARMv8 */ | |
d614a513 | 4880 | if (u && !arm_dc_feature(s, ARM_FEATURE_V8)) { |
25f84f79 PM |
4881 | return 1; |
4882 | } | |
2c0262af | 4883 | break; |
36a71934 RH |
4884 | case NEON_3R_VFM_VQRDMLSH: |
4885 | if (!arm_dc_feature(s, ARM_FEATURE_VFP4)) { | |
da97f52c PM |
4886 | return 1; |
4887 | } | |
4888 | break; | |
9ee6e8bb | 4889 | default: |
2c0262af | 4890 | break; |
9ee6e8bb | 4891 | } |
dd8fbd78 | 4892 | |
25f84f79 PM |
4893 | if (pairwise && q) { |
4894 | /* All the pairwise insns UNDEF if Q is set */ | |
4895 | return 1; | |
4896 | } | |
4897 | ||
9ee6e8bb PB |
4898 | for (pass = 0; pass < (q ? 4 : 2); pass++) { |
4899 | ||
4900 | if (pairwise) { | |
4901 | /* Pairwise. */ | |
a5a14945 JR |
4902 | if (pass < 1) { |
4903 | tmp = neon_load_reg(rn, 0); | |
4904 | tmp2 = neon_load_reg(rn, 1); | |
9ee6e8bb | 4905 | } else { |
a5a14945 JR |
4906 | tmp = neon_load_reg(rm, 0); |
4907 | tmp2 = neon_load_reg(rm, 1); | |
9ee6e8bb PB |
4908 | } |
4909 | } else { | |
4910 | /* Elementwise. */ | |
dd8fbd78 FN |
4911 | tmp = neon_load_reg(rn, pass); |
4912 | tmp2 = neon_load_reg(rm, pass); | |
9ee6e8bb PB |
4913 | } |
4914 | switch (op) { | |
62698be3 | 4915 | case NEON_3R_VHADD: |
9ee6e8bb PB |
4916 | GEN_NEON_INTEGER_OP(hadd); |
4917 | break; | |
62698be3 | 4918 | case NEON_3R_VRHADD: |
9ee6e8bb | 4919 | GEN_NEON_INTEGER_OP(rhadd); |
2c0262af | 4920 | break; |
62698be3 | 4921 | case NEON_3R_VHSUB: |
9ee6e8bb PB |
4922 | GEN_NEON_INTEGER_OP(hsub); |
4923 | break; | |
62698be3 | 4924 | case NEON_3R_VSHL: |
ad69471c | 4925 | GEN_NEON_INTEGER_OP(shl); |
2c0262af | 4926 | break; |
62698be3 | 4927 | case NEON_3R_VQSHL: |
02da0b2d | 4928 | GEN_NEON_INTEGER_OP_ENV(qshl); |
2c0262af | 4929 | break; |
62698be3 | 4930 | case NEON_3R_VRSHL: |
ad69471c | 4931 | GEN_NEON_INTEGER_OP(rshl); |
2c0262af | 4932 | break; |
62698be3 | 4933 | case NEON_3R_VQRSHL: |
02da0b2d | 4934 | GEN_NEON_INTEGER_OP_ENV(qrshl); |
9ee6e8bb | 4935 | break; |
62698be3 | 4936 | case NEON_3R_VABD: |
9ee6e8bb PB |
4937 | GEN_NEON_INTEGER_OP(abd); |
4938 | break; | |
62698be3 | 4939 | case NEON_3R_VABA: |
9ee6e8bb | 4940 | GEN_NEON_INTEGER_OP(abd); |
7d1b0095 | 4941 | tcg_temp_free_i32(tmp2); |
dd8fbd78 FN |
4942 | tmp2 = neon_load_reg(rd, pass); |
4943 | gen_neon_add(size, tmp, tmp2); | |
9ee6e8bb | 4944 | break; |
62698be3 | 4945 | case NEON_3R_VMUL: |
82083184 RH |
4946 | /* VMUL.P8; other cases already eliminated. */ |
4947 | gen_helper_neon_mul_p8(tmp, tmp, tmp2); | |
9ee6e8bb | 4948 | break; |
62698be3 | 4949 | case NEON_3R_VPMAX: |
9ee6e8bb PB |
4950 | GEN_NEON_INTEGER_OP(pmax); |
4951 | break; | |
62698be3 | 4952 | case NEON_3R_VPMIN: |
9ee6e8bb PB |
4953 | GEN_NEON_INTEGER_OP(pmin); |
4954 | break; | |
62698be3 | 4955 | case NEON_3R_VQDMULH_VQRDMULH: /* Multiply high. */ |
9ee6e8bb PB |
4956 | if (!u) { /* VQDMULH */ |
4957 | switch (size) { | |
02da0b2d PM |
4958 | case 1: |
4959 | gen_helper_neon_qdmulh_s16(tmp, cpu_env, tmp, tmp2); | |
4960 | break; | |
4961 | case 2: | |
4962 | gen_helper_neon_qdmulh_s32(tmp, cpu_env, tmp, tmp2); | |
4963 | break; | |
62698be3 | 4964 | default: abort(); |
9ee6e8bb | 4965 | } |
62698be3 | 4966 | } else { /* VQRDMULH */ |
9ee6e8bb | 4967 | switch (size) { |
02da0b2d PM |
4968 | case 1: |
4969 | gen_helper_neon_qrdmulh_s16(tmp, cpu_env, tmp, tmp2); | |
4970 | break; | |
4971 | case 2: | |
4972 | gen_helper_neon_qrdmulh_s32(tmp, cpu_env, tmp, tmp2); | |
4973 | break; | |
62698be3 | 4974 | default: abort(); |
9ee6e8bb PB |
4975 | } |
4976 | } | |
4977 | break; | |
36a71934 | 4978 | case NEON_3R_VPADD_VQRDMLAH: |
9ee6e8bb | 4979 | switch (size) { |
dd8fbd78 FN |
4980 | case 0: gen_helper_neon_padd_u8(tmp, tmp, tmp2); break; |
4981 | case 1: gen_helper_neon_padd_u16(tmp, tmp, tmp2); break; | |
4982 | case 2: tcg_gen_add_i32(tmp, tmp, tmp2); break; | |
62698be3 | 4983 | default: abort(); |
9ee6e8bb PB |
4984 | } |
4985 | break; | |
62698be3 | 4986 | case NEON_3R_FLOAT_ARITH: /* Floating point arithmetic. */ |
aa47cfdd PM |
4987 | { |
4988 | TCGv_ptr fpstatus = get_fpstatus_ptr(1); | |
9ee6e8bb PB |
4989 | switch ((u << 2) | size) { |
4990 | case 0: /* VADD */ | |
aa47cfdd PM |
4991 | case 4: /* VPADD */ |
4992 | gen_helper_vfp_adds(tmp, tmp, tmp2, fpstatus); | |
9ee6e8bb PB |
4993 | break; |
4994 | case 2: /* VSUB */ | |
aa47cfdd | 4995 | gen_helper_vfp_subs(tmp, tmp, tmp2, fpstatus); |
9ee6e8bb PB |
4996 | break; |
4997 | case 6: /* VABD */ | |
aa47cfdd | 4998 | gen_helper_neon_abd_f32(tmp, tmp, tmp2, fpstatus); |
9ee6e8bb PB |
4999 | break; |
5000 | default: | |
62698be3 | 5001 | abort(); |
9ee6e8bb | 5002 | } |
aa47cfdd | 5003 | tcg_temp_free_ptr(fpstatus); |
9ee6e8bb | 5004 | break; |
aa47cfdd | 5005 | } |
62698be3 | 5006 | case NEON_3R_FLOAT_MULTIPLY: |
aa47cfdd PM |
5007 | { |
5008 | TCGv_ptr fpstatus = get_fpstatus_ptr(1); | |
5009 | gen_helper_vfp_muls(tmp, tmp, tmp2, fpstatus); | |
9ee6e8bb | 5010 | if (!u) { |
7d1b0095 | 5011 | tcg_temp_free_i32(tmp2); |
dd8fbd78 | 5012 | tmp2 = neon_load_reg(rd, pass); |
9ee6e8bb | 5013 | if (size == 0) { |
aa47cfdd | 5014 | gen_helper_vfp_adds(tmp, tmp, tmp2, fpstatus); |
9ee6e8bb | 5015 | } else { |
aa47cfdd | 5016 | gen_helper_vfp_subs(tmp, tmp2, tmp, fpstatus); |
9ee6e8bb PB |
5017 | } |
5018 | } | |
aa47cfdd | 5019 | tcg_temp_free_ptr(fpstatus); |
9ee6e8bb | 5020 | break; |
aa47cfdd | 5021 | } |
62698be3 | 5022 | case NEON_3R_FLOAT_CMP: |
aa47cfdd PM |
5023 | { |
5024 | TCGv_ptr fpstatus = get_fpstatus_ptr(1); | |
9ee6e8bb | 5025 | if (!u) { |
aa47cfdd | 5026 | gen_helper_neon_ceq_f32(tmp, tmp, tmp2, fpstatus); |
b5ff1b31 | 5027 | } else { |
aa47cfdd PM |
5028 | if (size == 0) { |
5029 | gen_helper_neon_cge_f32(tmp, tmp, tmp2, fpstatus); | |
5030 | } else { | |
5031 | gen_helper_neon_cgt_f32(tmp, tmp, tmp2, fpstatus); | |
5032 | } | |
b5ff1b31 | 5033 | } |
aa47cfdd | 5034 | tcg_temp_free_ptr(fpstatus); |
2c0262af | 5035 | break; |
aa47cfdd | 5036 | } |
62698be3 | 5037 | case NEON_3R_FLOAT_ACMP: |
aa47cfdd PM |
5038 | { |
5039 | TCGv_ptr fpstatus = get_fpstatus_ptr(1); | |
5040 | if (size == 0) { | |
5041 | gen_helper_neon_acge_f32(tmp, tmp, tmp2, fpstatus); | |
5042 | } else { | |
5043 | gen_helper_neon_acgt_f32(tmp, tmp, tmp2, fpstatus); | |
5044 | } | |
5045 | tcg_temp_free_ptr(fpstatus); | |
2c0262af | 5046 | break; |
aa47cfdd | 5047 | } |
62698be3 | 5048 | case NEON_3R_FLOAT_MINMAX: |
aa47cfdd PM |
5049 | { |
5050 | TCGv_ptr fpstatus = get_fpstatus_ptr(1); | |
5051 | if (size == 0) { | |
f71a2ae5 | 5052 | gen_helper_vfp_maxs(tmp, tmp, tmp2, fpstatus); |
aa47cfdd | 5053 | } else { |
f71a2ae5 | 5054 | gen_helper_vfp_mins(tmp, tmp, tmp2, fpstatus); |
aa47cfdd PM |
5055 | } |
5056 | tcg_temp_free_ptr(fpstatus); | |
9ee6e8bb | 5057 | break; |
aa47cfdd | 5058 | } |
505935fc WN |
5059 | case NEON_3R_FLOAT_MISC: |
5060 | if (u) { | |
5061 | /* VMAXNM/VMINNM */ | |
5062 | TCGv_ptr fpstatus = get_fpstatus_ptr(1); | |
5063 | if (size == 0) { | |
f71a2ae5 | 5064 | gen_helper_vfp_maxnums(tmp, tmp, tmp2, fpstatus); |
505935fc | 5065 | } else { |
f71a2ae5 | 5066 | gen_helper_vfp_minnums(tmp, tmp, tmp2, fpstatus); |
505935fc WN |
5067 | } |
5068 | tcg_temp_free_ptr(fpstatus); | |
5069 | } else { | |
5070 | if (size == 0) { | |
5071 | gen_helper_recps_f32(tmp, tmp, tmp2, cpu_env); | |
5072 | } else { | |
5073 | gen_helper_rsqrts_f32(tmp, tmp, tmp2, cpu_env); | |
5074 | } | |
5075 | } | |
2c0262af | 5076 | break; |
36a71934 | 5077 | case NEON_3R_VFM_VQRDMLSH: |
da97f52c PM |
5078 | { |
5079 | /* VFMA, VFMS: fused multiply-add */ | |
5080 | TCGv_ptr fpstatus = get_fpstatus_ptr(1); | |
5081 | TCGv_i32 tmp3 = neon_load_reg(rd, pass); | |
5082 | if (size) { | |
5083 | /* VFMS */ | |
5084 | gen_helper_vfp_negs(tmp, tmp); | |
5085 | } | |
5086 | gen_helper_vfp_muladds(tmp, tmp, tmp2, tmp3, fpstatus); | |
5087 | tcg_temp_free_i32(tmp3); | |
5088 | tcg_temp_free_ptr(fpstatus); | |
5089 | break; | |
5090 | } | |
9ee6e8bb PB |
5091 | default: |
5092 | abort(); | |
2c0262af | 5093 | } |
7d1b0095 | 5094 | tcg_temp_free_i32(tmp2); |
dd8fbd78 | 5095 | |
9ee6e8bb PB |
5096 | /* Save the result. For elementwise operations we can put it |
5097 | straight into the destination register. For pairwise operations | |
5098 | we have to be careful to avoid clobbering the source operands. */ | |
5099 | if (pairwise && rd == rm) { | |
dd8fbd78 | 5100 | neon_store_scratch(pass, tmp); |
9ee6e8bb | 5101 | } else { |
dd8fbd78 | 5102 | neon_store_reg(rd, pass, tmp); |
9ee6e8bb PB |
5103 | } |
5104 | ||
5105 | } /* for pass */ | |
5106 | if (pairwise && rd == rm) { | |
5107 | for (pass = 0; pass < (q ? 4 : 2); pass++) { | |
dd8fbd78 FN |
5108 | tmp = neon_load_scratch(pass); |
5109 | neon_store_reg(rd, pass, tmp); | |
9ee6e8bb PB |
5110 | } |
5111 | } | |
ad69471c | 5112 | /* End of 3 register same size operations. */ |
9ee6e8bb PB |
5113 | } else if (insn & (1 << 4)) { |
5114 | if ((insn & 0x00380080) != 0) { | |
5115 | /* Two registers and shift. */ | |
5116 | op = (insn >> 8) & 0xf; | |
5117 | if (insn & (1 << 7)) { | |
cc13115b PM |
5118 | /* 64-bit shift. */ |
5119 | if (op > 7) { | |
5120 | return 1; | |
5121 | } | |
9ee6e8bb PB |
5122 | size = 3; |
5123 | } else { | |
5124 | size = 2; | |
5125 | while ((insn & (1 << (size + 19))) == 0) | |
5126 | size--; | |
5127 | } | |
5128 | shift = (insn >> 16) & ((1 << (3 + size)) - 1); | |
9ee6e8bb PB |
5129 | if (op < 8) { |
5130 | /* Shift by immediate: | |
5131 | VSHR, VSRA, VRSHR, VRSRA, VSRI, VSHL, VQSHL, VQSHLU. */ | |
cc13115b PM |
5132 | if (q && ((rd | rm) & 1)) { |
5133 | return 1; | |
5134 | } | |
5135 | if (!u && (op == 4 || op == 6)) { | |
5136 | return 1; | |
5137 | } | |
9ee6e8bb PB |
5138 | /* Right shifts are encoded as N - shift, where N is the |
5139 | element size in bits. */ | |
1dc8425e | 5140 | if (op <= 4) { |
9ee6e8bb | 5141 | shift = shift - (1 << (size + 3)); |
1dc8425e RH |
5142 | } |
5143 | ||
5144 | switch (op) { | |
5145 | case 0: /* VSHR */ | |
5146 | /* Right shift comes here negative. */ | |
5147 | shift = -shift; | |
5148 | /* Shifts larger than the element size are architecturally | |
5149 | * valid. Unsigned results in all zeros; signed results | |
5150 | * in all sign bits. | |
5151 | */ | |
5152 | if (!u) { | |
5153 | tcg_gen_gvec_sari(size, rd_ofs, rm_ofs, | |
5154 | MIN(shift, (8 << size) - 1), | |
5155 | vec_size, vec_size); | |
5156 | } else if (shift >= 8 << size) { | |
5157 | tcg_gen_gvec_dup8i(rd_ofs, vec_size, vec_size, 0); | |
5158 | } else { | |
5159 | tcg_gen_gvec_shri(size, rd_ofs, rm_ofs, shift, | |
5160 | vec_size, vec_size); | |
5161 | } | |
5162 | return 0; | |
5163 | ||
41f6c113 RH |
5164 | case 1: /* VSRA */ |
5165 | /* Right shift comes here negative. */ | |
5166 | shift = -shift; | |
5167 | /* Shifts larger than the element size are architecturally | |
5168 | * valid. Unsigned results in all zeros; signed results | |
5169 | * in all sign bits. | |
5170 | */ | |
5171 | if (!u) { | |
5172 | tcg_gen_gvec_2i(rd_ofs, rm_ofs, vec_size, vec_size, | |
5173 | MIN(shift, (8 << size) - 1), | |
5174 | &ssra_op[size]); | |
5175 | } else if (shift >= 8 << size) { | |
5176 | /* rd += 0 */ | |
5177 | } else { | |
5178 | tcg_gen_gvec_2i(rd_ofs, rm_ofs, vec_size, vec_size, | |
5179 | shift, &usra_op[size]); | |
5180 | } | |
5181 | return 0; | |
5182 | ||
f3cd8218 RH |
5183 | case 4: /* VSRI */ |
5184 | if (!u) { | |
5185 | return 1; | |
5186 | } | |
5187 | /* Right shift comes here negative. */ | |
5188 | shift = -shift; | |
5189 | /* Shift out of range leaves destination unchanged. */ | |
5190 | if (shift < 8 << size) { | |
5191 | tcg_gen_gvec_2i(rd_ofs, rm_ofs, vec_size, vec_size, | |
5192 | shift, &sri_op[size]); | |
5193 | } | |
5194 | return 0; | |
5195 | ||
1dc8425e | 5196 | case 5: /* VSHL, VSLI */ |
f3cd8218 RH |
5197 | if (u) { /* VSLI */ |
5198 | /* Shift out of range leaves destination unchanged. */ | |
5199 | if (shift < 8 << size) { | |
5200 | tcg_gen_gvec_2i(rd_ofs, rm_ofs, vec_size, | |
5201 | vec_size, shift, &sli_op[size]); | |
5202 | } | |
5203 | } else { /* VSHL */ | |
1dc8425e RH |
5204 | /* Shifts larger than the element size are |
5205 | * architecturally valid and results in zero. | |
5206 | */ | |
5207 | if (shift >= 8 << size) { | |
5208 | tcg_gen_gvec_dup8i(rd_ofs, vec_size, vec_size, 0); | |
5209 | } else { | |
5210 | tcg_gen_gvec_shli(size, rd_ofs, rm_ofs, shift, | |
5211 | vec_size, vec_size); | |
5212 | } | |
1dc8425e | 5213 | } |
f3cd8218 | 5214 | return 0; |
1dc8425e RH |
5215 | } |
5216 | ||
9ee6e8bb PB |
5217 | if (size == 3) { |
5218 | count = q + 1; | |
5219 | } else { | |
5220 | count = q ? 4: 2; | |
5221 | } | |
1dc8425e RH |
5222 | |
5223 | /* To avoid excessive duplication of ops we implement shift | |
5224 | * by immediate using the variable shift operations. | |
5225 | */ | |
5226 | imm = dup_const(size, shift); | |
9ee6e8bb PB |
5227 | |
5228 | for (pass = 0; pass < count; pass++) { | |
ad69471c PB |
5229 | if (size == 3) { |
5230 | neon_load_reg64(cpu_V0, rm + pass); | |
5231 | tcg_gen_movi_i64(cpu_V1, imm); | |
5232 | switch (op) { | |
ad69471c PB |
5233 | case 2: /* VRSHR */ |
5234 | case 3: /* VRSRA */ | |
5235 | if (u) | |
5236 | gen_helper_neon_rshl_u64(cpu_V0, cpu_V0, cpu_V1); | |
9ee6e8bb | 5237 | else |
ad69471c | 5238 | gen_helper_neon_rshl_s64(cpu_V0, cpu_V0, cpu_V1); |
9ee6e8bb | 5239 | break; |
0322b26e | 5240 | case 6: /* VQSHLU */ |
02da0b2d PM |
5241 | gen_helper_neon_qshlu_s64(cpu_V0, cpu_env, |
5242 | cpu_V0, cpu_V1); | |
ad69471c | 5243 | break; |
0322b26e PM |
5244 | case 7: /* VQSHL */ |
5245 | if (u) { | |
02da0b2d | 5246 | gen_helper_neon_qshl_u64(cpu_V0, cpu_env, |
0322b26e PM |
5247 | cpu_V0, cpu_V1); |
5248 | } else { | |
02da0b2d | 5249 | gen_helper_neon_qshl_s64(cpu_V0, cpu_env, |
0322b26e PM |
5250 | cpu_V0, cpu_V1); |
5251 | } | |
9ee6e8bb | 5252 | break; |
1dc8425e RH |
5253 | default: |
5254 | g_assert_not_reached(); | |
9ee6e8bb | 5255 | } |
41f6c113 | 5256 | if (op == 3) { |
ad69471c | 5257 | /* Accumulate. */ |
5371cb81 | 5258 | neon_load_reg64(cpu_V1, rd + pass); |
ad69471c | 5259 | tcg_gen_add_i64(cpu_V0, cpu_V0, cpu_V1); |
ad69471c PB |
5260 | } |
5261 | neon_store_reg64(cpu_V0, rd + pass); | |
5262 | } else { /* size < 3 */ | |
5263 | /* Operands in T0 and T1. */ | |
dd8fbd78 | 5264 | tmp = neon_load_reg(rm, pass); |
7d1b0095 | 5265 | tmp2 = tcg_temp_new_i32(); |
dd8fbd78 | 5266 | tcg_gen_movi_i32(tmp2, imm); |
ad69471c | 5267 | switch (op) { |
ad69471c PB |
5268 | case 2: /* VRSHR */ |
5269 | case 3: /* VRSRA */ | |
5270 | GEN_NEON_INTEGER_OP(rshl); | |
5271 | break; | |
0322b26e | 5272 | case 6: /* VQSHLU */ |
ad69471c | 5273 | switch (size) { |
0322b26e | 5274 | case 0: |
02da0b2d PM |
5275 | gen_helper_neon_qshlu_s8(tmp, cpu_env, |
5276 | tmp, tmp2); | |
0322b26e PM |
5277 | break; |
5278 | case 1: | |
02da0b2d PM |
5279 | gen_helper_neon_qshlu_s16(tmp, cpu_env, |
5280 | tmp, tmp2); | |
0322b26e PM |
5281 | break; |
5282 | case 2: | |
02da0b2d PM |
5283 | gen_helper_neon_qshlu_s32(tmp, cpu_env, |
5284 | tmp, tmp2); | |
0322b26e PM |
5285 | break; |
5286 | default: | |
cc13115b | 5287 | abort(); |
ad69471c PB |
5288 | } |
5289 | break; | |
0322b26e | 5290 | case 7: /* VQSHL */ |
02da0b2d | 5291 | GEN_NEON_INTEGER_OP_ENV(qshl); |
0322b26e | 5292 | break; |
1dc8425e RH |
5293 | default: |
5294 | g_assert_not_reached(); | |
ad69471c | 5295 | } |
7d1b0095 | 5296 | tcg_temp_free_i32(tmp2); |
ad69471c | 5297 | |
41f6c113 | 5298 | if (op == 3) { |
ad69471c | 5299 | /* Accumulate. */ |
dd8fbd78 | 5300 | tmp2 = neon_load_reg(rd, pass); |
5371cb81 | 5301 | gen_neon_add(size, tmp, tmp2); |
7d1b0095 | 5302 | tcg_temp_free_i32(tmp2); |
ad69471c | 5303 | } |
dd8fbd78 | 5304 | neon_store_reg(rd, pass, tmp); |
9ee6e8bb PB |
5305 | } |
5306 | } /* for pass */ | |
5307 | } else if (op < 10) { | |
ad69471c | 5308 | /* Shift by immediate and narrow: |
9ee6e8bb | 5309 | VSHRN, VRSHRN, VQSHRN, VQRSHRN. */ |
0b36f4cd | 5310 | int input_unsigned = (op == 8) ? !u : u; |
cc13115b PM |
5311 | if (rm & 1) { |
5312 | return 1; | |
5313 | } | |
9ee6e8bb PB |
5314 | shift = shift - (1 << (size + 3)); |
5315 | size++; | |
92cdfaeb | 5316 | if (size == 3) { |
a7812ae4 | 5317 | tmp64 = tcg_const_i64(shift); |
92cdfaeb PM |
5318 | neon_load_reg64(cpu_V0, rm); |
5319 | neon_load_reg64(cpu_V1, rm + 1); | |
5320 | for (pass = 0; pass < 2; pass++) { | |
5321 | TCGv_i64 in; | |
5322 | if (pass == 0) { | |
5323 | in = cpu_V0; | |
5324 | } else { | |
5325 | in = cpu_V1; | |
5326 | } | |
ad69471c | 5327 | if (q) { |
0b36f4cd | 5328 | if (input_unsigned) { |
92cdfaeb | 5329 | gen_helper_neon_rshl_u64(cpu_V0, in, tmp64); |
0b36f4cd | 5330 | } else { |
92cdfaeb | 5331 | gen_helper_neon_rshl_s64(cpu_V0, in, tmp64); |
0b36f4cd | 5332 | } |
ad69471c | 5333 | } else { |
0b36f4cd | 5334 | if (input_unsigned) { |
92cdfaeb | 5335 | gen_helper_neon_shl_u64(cpu_V0, in, tmp64); |
0b36f4cd | 5336 | } else { |
92cdfaeb | 5337 | gen_helper_neon_shl_s64(cpu_V0, in, tmp64); |
0b36f4cd | 5338 | } |
ad69471c | 5339 | } |
7d1b0095 | 5340 | tmp = tcg_temp_new_i32(); |
92cdfaeb PM |
5341 | gen_neon_narrow_op(op == 8, u, size - 1, tmp, cpu_V0); |
5342 | neon_store_reg(rd, pass, tmp); | |
5343 | } /* for pass */ | |
5344 | tcg_temp_free_i64(tmp64); | |
5345 | } else { | |
5346 | if (size == 1) { | |
5347 | imm = (uint16_t)shift; | |
5348 | imm |= imm << 16; | |
2c0262af | 5349 | } else { |
92cdfaeb PM |
5350 | /* size == 2 */ |
5351 | imm = (uint32_t)shift; | |
5352 | } | |
5353 | tmp2 = tcg_const_i32(imm); | |
5354 | tmp4 = neon_load_reg(rm + 1, 0); | |
5355 | tmp5 = neon_load_reg(rm + 1, 1); | |
5356 | for (pass = 0; pass < 2; pass++) { | |
5357 | if (pass == 0) { | |
5358 | tmp = neon_load_reg(rm, 0); | |
5359 | } else { | |
5360 | tmp = tmp4; | |
5361 | } | |
0b36f4cd CL |
5362 | gen_neon_shift_narrow(size, tmp, tmp2, q, |
5363 | input_unsigned); | |
92cdfaeb PM |
5364 | if (pass == 0) { |
5365 | tmp3 = neon_load_reg(rm, 1); | |
5366 | } else { | |
5367 | tmp3 = tmp5; | |
5368 | } | |
0b36f4cd CL |
5369 | gen_neon_shift_narrow(size, tmp3, tmp2, q, |
5370 | input_unsigned); | |
36aa55dc | 5371 | tcg_gen_concat_i32_i64(cpu_V0, tmp, tmp3); |
7d1b0095 PM |
5372 | tcg_temp_free_i32(tmp); |
5373 | tcg_temp_free_i32(tmp3); | |
5374 | tmp = tcg_temp_new_i32(); | |
92cdfaeb PM |
5375 | gen_neon_narrow_op(op == 8, u, size - 1, tmp, cpu_V0); |
5376 | neon_store_reg(rd, pass, tmp); | |
5377 | } /* for pass */ | |
c6067f04 | 5378 | tcg_temp_free_i32(tmp2); |
b75263d6 | 5379 | } |
9ee6e8bb | 5380 | } else if (op == 10) { |
cc13115b PM |
5381 | /* VSHLL, VMOVL */ |
5382 | if (q || (rd & 1)) { | |
9ee6e8bb | 5383 | return 1; |
cc13115b | 5384 | } |
ad69471c PB |
5385 | tmp = neon_load_reg(rm, 0); |
5386 | tmp2 = neon_load_reg(rm, 1); | |
9ee6e8bb | 5387 | for (pass = 0; pass < 2; pass++) { |
ad69471c PB |
5388 | if (pass == 1) |
5389 | tmp = tmp2; | |
5390 | ||
5391 | gen_neon_widen(cpu_V0, tmp, size, u); | |
9ee6e8bb | 5392 | |
9ee6e8bb PB |
5393 | if (shift != 0) { |
5394 | /* The shift is less than the width of the source | |
ad69471c PB |
5395 | type, so we can just shift the whole register. */ |
5396 | tcg_gen_shli_i64(cpu_V0, cpu_V0, shift); | |
acdf01ef CL |
5397 | /* Widen the result of shift: we need to clear |
5398 | * the potential overflow bits resulting from | |
5399 | * left bits of the narrow input appearing as | |
5400 | * right bits of left the neighbour narrow | |
5401 | * input. */ | |
ad69471c PB |
5402 | if (size < 2 || !u) { |
5403 | uint64_t imm64; | |
5404 | if (size == 0) { | |
5405 | imm = (0xffu >> (8 - shift)); | |
5406 | imm |= imm << 16; | |
acdf01ef | 5407 | } else if (size == 1) { |
ad69471c | 5408 | imm = 0xffff >> (16 - shift); |
acdf01ef CL |
5409 | } else { |
5410 | /* size == 2 */ | |
5411 | imm = 0xffffffff >> (32 - shift); | |
5412 | } | |
5413 | if (size < 2) { | |
5414 | imm64 = imm | (((uint64_t)imm) << 32); | |
5415 | } else { | |
5416 | imm64 = imm; | |
9ee6e8bb | 5417 | } |
acdf01ef | 5418 | tcg_gen_andi_i64(cpu_V0, cpu_V0, ~imm64); |
9ee6e8bb PB |
5419 | } |
5420 | } | |
ad69471c | 5421 | neon_store_reg64(cpu_V0, rd + pass); |
9ee6e8bb | 5422 | } |
f73534a5 | 5423 | } else if (op >= 14) { |
9ee6e8bb | 5424 | /* VCVT fixed-point. */ |
c253dd78 PM |
5425 | TCGv_ptr fpst; |
5426 | TCGv_i32 shiftv; | |
5427 | VFPGenFixPointFn *fn; | |
5428 | ||
cc13115b PM |
5429 | if (!(insn & (1 << 21)) || (q && ((rd | rm) & 1))) { |
5430 | return 1; | |
5431 | } | |
c253dd78 PM |
5432 | |
5433 | if (!(op & 1)) { | |
5434 | if (u) { | |
5435 | fn = gen_helper_vfp_ultos; | |
5436 | } else { | |
5437 | fn = gen_helper_vfp_sltos; | |
5438 | } | |
5439 | } else { | |
5440 | if (u) { | |
5441 | fn = gen_helper_vfp_touls_round_to_zero; | |
5442 | } else { | |
5443 | fn = gen_helper_vfp_tosls_round_to_zero; | |
5444 | } | |
5445 | } | |
5446 | ||
f73534a5 PM |
5447 | /* We have already masked out the must-be-1 top bit of imm6, |
5448 | * hence this 32-shift where the ARM ARM has 64-imm6. | |
5449 | */ | |
5450 | shift = 32 - shift; | |
c253dd78 PM |
5451 | fpst = get_fpstatus_ptr(1); |
5452 | shiftv = tcg_const_i32(shift); | |
9ee6e8bb | 5453 | for (pass = 0; pass < (q ? 4 : 2); pass++) { |
c253dd78 PM |
5454 | TCGv_i32 tmpf = neon_load_reg(rm, pass); |
5455 | fn(tmpf, tmpf, shiftv, fpst); | |
5456 | neon_store_reg(rd, pass, tmpf); | |
2c0262af | 5457 | } |
c253dd78 PM |
5458 | tcg_temp_free_ptr(fpst); |
5459 | tcg_temp_free_i32(shiftv); | |
2c0262af | 5460 | } else { |
9ee6e8bb PB |
5461 | return 1; |
5462 | } | |
5463 | } else { /* (insn & 0x00380080) == 0 */ | |
246fa4ac RH |
5464 | int invert, reg_ofs, vec_size; |
5465 | ||
7d80fee5 PM |
5466 | if (q && (rd & 1)) { |
5467 | return 1; | |
5468 | } | |
9ee6e8bb PB |
5469 | |
5470 | op = (insn >> 8) & 0xf; | |
5471 | /* One register and immediate. */ | |
5472 | imm = (u << 7) | ((insn >> 12) & 0x70) | (insn & 0xf); | |
5473 | invert = (insn & (1 << 5)) != 0; | |
7d80fee5 PM |
5474 | /* Note that op = 2,3,4,5,6,7,10,11,12,13 imm=0 is UNPREDICTABLE. |
5475 | * We choose to not special-case this and will behave as if a | |
5476 | * valid constant encoding of 0 had been given. | |
5477 | */ | |
9ee6e8bb PB |
5478 | switch (op) { |
5479 | case 0: case 1: | |
5480 | /* no-op */ | |
5481 | break; | |
5482 | case 2: case 3: | |
5483 | imm <<= 8; | |
5484 | break; | |
5485 | case 4: case 5: | |
5486 | imm <<= 16; | |
5487 | break; | |
5488 | case 6: case 7: | |
5489 | imm <<= 24; | |
5490 | break; | |
5491 | case 8: case 9: | |
5492 | imm |= imm << 16; | |
5493 | break; | |
5494 | case 10: case 11: | |
5495 | imm = (imm << 8) | (imm << 24); | |
5496 | break; | |
5497 | case 12: | |
8e31209e | 5498 | imm = (imm << 8) | 0xff; |
9ee6e8bb PB |
5499 | break; |
5500 | case 13: | |
5501 | imm = (imm << 16) | 0xffff; | |
5502 | break; | |
5503 | case 14: | |
5504 | imm |= (imm << 8) | (imm << 16) | (imm << 24); | |
246fa4ac | 5505 | if (invert) { |
9ee6e8bb | 5506 | imm = ~imm; |
246fa4ac | 5507 | } |
9ee6e8bb PB |
5508 | break; |
5509 | case 15: | |
7d80fee5 PM |
5510 | if (invert) { |
5511 | return 1; | |
5512 | } | |
9ee6e8bb PB |
5513 | imm = ((imm & 0x80) << 24) | ((imm & 0x3f) << 19) |
5514 | | ((imm & 0x40) ? (0x1f << 25) : (1 << 30)); | |
5515 | break; | |
5516 | } | |
246fa4ac | 5517 | if (invert) { |
9ee6e8bb | 5518 | imm = ~imm; |
246fa4ac | 5519 | } |
9ee6e8bb | 5520 | |
246fa4ac RH |
5521 | reg_ofs = neon_reg_offset(rd, 0); |
5522 | vec_size = q ? 16 : 8; | |
5523 | ||
5524 | if (op & 1 && op < 12) { | |
5525 | if (invert) { | |
5526 | /* The immediate value has already been inverted, | |
5527 | * so BIC becomes AND. | |
5528 | */ | |
5529 | tcg_gen_gvec_andi(MO_32, reg_ofs, reg_ofs, imm, | |
5530 | vec_size, vec_size); | |
9ee6e8bb | 5531 | } else { |
246fa4ac RH |
5532 | tcg_gen_gvec_ori(MO_32, reg_ofs, reg_ofs, imm, |
5533 | vec_size, vec_size); | |
5534 | } | |
5535 | } else { | |
5536 | /* VMOV, VMVN. */ | |
5537 | if (op == 14 && invert) { | |
5538 | TCGv_i64 t64 = tcg_temp_new_i64(); | |
5539 | ||
5540 | for (pass = 0; pass <= q; ++pass) { | |
5541 | uint64_t val = 0; | |
a5a14945 | 5542 | int n; |
246fa4ac RH |
5543 | |
5544 | for (n = 0; n < 8; n++) { | |
5545 | if (imm & (1 << (n + pass * 8))) { | |
5546 | val |= 0xffull << (n * 8); | |
5547 | } | |
9ee6e8bb | 5548 | } |
246fa4ac RH |
5549 | tcg_gen_movi_i64(t64, val); |
5550 | neon_store_reg64(t64, rd + pass); | |
9ee6e8bb | 5551 | } |
246fa4ac RH |
5552 | tcg_temp_free_i64(t64); |
5553 | } else { | |
5554 | tcg_gen_gvec_dup32i(reg_ofs, vec_size, vec_size, imm); | |
9ee6e8bb PB |
5555 | } |
5556 | } | |
5557 | } | |
e4b3861d | 5558 | } else { /* (insn & 0x00800010 == 0x00800000) */ |
9ee6e8bb PB |
5559 | if (size != 3) { |
5560 | op = (insn >> 8) & 0xf; | |
5561 | if ((insn & (1 << 6)) == 0) { | |
5562 | /* Three registers of different lengths. */ | |
5563 | int src1_wide; | |
5564 | int src2_wide; | |
5565 | int prewiden; | |
526d0096 PM |
5566 | /* undefreq: bit 0 : UNDEF if size == 0 |
5567 | * bit 1 : UNDEF if size == 1 | |
5568 | * bit 2 : UNDEF if size == 2 | |
5569 | * bit 3 : UNDEF if U == 1 | |
5570 | * Note that [2:0] set implies 'always UNDEF' | |
695272dc PM |
5571 | */ |
5572 | int undefreq; | |
5573 | /* prewiden, src1_wide, src2_wide, undefreq */ | |
5574 | static const int neon_3reg_wide[16][4] = { | |
5575 | {1, 0, 0, 0}, /* VADDL */ | |
5576 | {1, 1, 0, 0}, /* VADDW */ | |
5577 | {1, 0, 0, 0}, /* VSUBL */ | |
5578 | {1, 1, 0, 0}, /* VSUBW */ | |
5579 | {0, 1, 1, 0}, /* VADDHN */ | |
5580 | {0, 0, 0, 0}, /* VABAL */ | |
5581 | {0, 1, 1, 0}, /* VSUBHN */ | |
5582 | {0, 0, 0, 0}, /* VABDL */ | |
5583 | {0, 0, 0, 0}, /* VMLAL */ | |
526d0096 | 5584 | {0, 0, 0, 9}, /* VQDMLAL */ |
695272dc | 5585 | {0, 0, 0, 0}, /* VMLSL */ |
526d0096 | 5586 | {0, 0, 0, 9}, /* VQDMLSL */ |
695272dc | 5587 | {0, 0, 0, 0}, /* Integer VMULL */ |
526d0096 | 5588 | {0, 0, 0, 1}, /* VQDMULL */ |
4e624eda | 5589 | {0, 0, 0, 0xa}, /* Polynomial VMULL */ |
526d0096 | 5590 | {0, 0, 0, 7}, /* Reserved: always UNDEF */ |
9ee6e8bb PB |
5591 | }; |
5592 | ||
5593 | prewiden = neon_3reg_wide[op][0]; | |
5594 | src1_wide = neon_3reg_wide[op][1]; | |
5595 | src2_wide = neon_3reg_wide[op][2]; | |
695272dc | 5596 | undefreq = neon_3reg_wide[op][3]; |
9ee6e8bb | 5597 | |
526d0096 PM |
5598 | if ((undefreq & (1 << size)) || |
5599 | ((undefreq & 8) && u)) { | |
695272dc PM |
5600 | return 1; |
5601 | } | |
5602 | if ((src1_wide && (rn & 1)) || | |
5603 | (src2_wide && (rm & 1)) || | |
5604 | (!src2_wide && (rd & 1))) { | |
ad69471c | 5605 | return 1; |
695272dc | 5606 | } |
ad69471c | 5607 | |
4e624eda PM |
5608 | /* Handle VMULL.P64 (Polynomial 64x64 to 128 bit multiply) |
5609 | * outside the loop below as it only performs a single pass. | |
5610 | */ | |
5611 | if (op == 14 && size == 2) { | |
5612 | TCGv_i64 tcg_rn, tcg_rm, tcg_rd; | |
5613 | ||
962fcbf2 | 5614 | if (!dc_isar_feature(aa32_pmull, s)) { |
4e624eda PM |
5615 | return 1; |
5616 | } | |
5617 | tcg_rn = tcg_temp_new_i64(); | |
5618 | tcg_rm = tcg_temp_new_i64(); | |
5619 | tcg_rd = tcg_temp_new_i64(); | |
5620 | neon_load_reg64(tcg_rn, rn); | |
5621 | neon_load_reg64(tcg_rm, rm); | |
5622 | gen_helper_neon_pmull_64_lo(tcg_rd, tcg_rn, tcg_rm); | |
5623 | neon_store_reg64(tcg_rd, rd); | |
5624 | gen_helper_neon_pmull_64_hi(tcg_rd, tcg_rn, tcg_rm); | |
5625 | neon_store_reg64(tcg_rd, rd + 1); | |
5626 | tcg_temp_free_i64(tcg_rn); | |
5627 | tcg_temp_free_i64(tcg_rm); | |
5628 | tcg_temp_free_i64(tcg_rd); | |
5629 | return 0; | |
5630 | } | |
5631 | ||
9ee6e8bb PB |
5632 | /* Avoid overlapping operands. Wide source operands are |
5633 | always aligned so will never overlap with wide | |
5634 | destinations in problematic ways. */ | |
8f8e3aa4 | 5635 | if (rd == rm && !src2_wide) { |
dd8fbd78 FN |
5636 | tmp = neon_load_reg(rm, 1); |
5637 | neon_store_scratch(2, tmp); | |
8f8e3aa4 | 5638 | } else if (rd == rn && !src1_wide) { |
dd8fbd78 FN |
5639 | tmp = neon_load_reg(rn, 1); |
5640 | neon_store_scratch(2, tmp); | |
9ee6e8bb | 5641 | } |
f764718d | 5642 | tmp3 = NULL; |
9ee6e8bb | 5643 | for (pass = 0; pass < 2; pass++) { |
ad69471c PB |
5644 | if (src1_wide) { |
5645 | neon_load_reg64(cpu_V0, rn + pass); | |
f764718d | 5646 | tmp = NULL; |
9ee6e8bb | 5647 | } else { |
ad69471c | 5648 | if (pass == 1 && rd == rn) { |
dd8fbd78 | 5649 | tmp = neon_load_scratch(2); |
9ee6e8bb | 5650 | } else { |
ad69471c PB |
5651 | tmp = neon_load_reg(rn, pass); |
5652 | } | |
5653 | if (prewiden) { | |
5654 | gen_neon_widen(cpu_V0, tmp, size, u); | |
9ee6e8bb PB |
5655 | } |
5656 | } | |
ad69471c PB |
5657 | if (src2_wide) { |
5658 | neon_load_reg64(cpu_V1, rm + pass); | |
f764718d | 5659 | tmp2 = NULL; |
9ee6e8bb | 5660 | } else { |
ad69471c | 5661 | if (pass == 1 && rd == rm) { |
dd8fbd78 | 5662 | tmp2 = neon_load_scratch(2); |
9ee6e8bb | 5663 | } else { |
ad69471c PB |
5664 | tmp2 = neon_load_reg(rm, pass); |
5665 | } | |
5666 | if (prewiden) { | |
5667 | gen_neon_widen(cpu_V1, tmp2, size, u); | |
9ee6e8bb | 5668 | } |
9ee6e8bb PB |
5669 | } |
5670 | switch (op) { | |
5671 | case 0: case 1: case 4: /* VADDL, VADDW, VADDHN, VRADDHN */ | |
ad69471c | 5672 | gen_neon_addl(size); |
9ee6e8bb | 5673 | break; |
79b0e534 | 5674 | case 2: case 3: case 6: /* VSUBL, VSUBW, VSUBHN, VRSUBHN */ |
ad69471c | 5675 | gen_neon_subl(size); |
9ee6e8bb PB |
5676 | break; |
5677 | case 5: case 7: /* VABAL, VABDL */ | |
5678 | switch ((size << 1) | u) { | |
ad69471c PB |
5679 | case 0: |
5680 | gen_helper_neon_abdl_s16(cpu_V0, tmp, tmp2); | |
5681 | break; | |
5682 | case 1: | |
5683 | gen_helper_neon_abdl_u16(cpu_V0, tmp, tmp2); | |
5684 | break; | |
5685 | case 2: | |
5686 | gen_helper_neon_abdl_s32(cpu_V0, tmp, tmp2); | |
5687 | break; | |
5688 | case 3: | |
5689 | gen_helper_neon_abdl_u32(cpu_V0, tmp, tmp2); | |
5690 | break; | |
5691 | case 4: | |
5692 | gen_helper_neon_abdl_s64(cpu_V0, tmp, tmp2); | |
5693 | break; | |
5694 | case 5: | |
5695 | gen_helper_neon_abdl_u64(cpu_V0, tmp, tmp2); | |
5696 | break; | |
9ee6e8bb PB |
5697 | default: abort(); |
5698 | } | |
7d1b0095 PM |
5699 | tcg_temp_free_i32(tmp2); |
5700 | tcg_temp_free_i32(tmp); | |
9ee6e8bb PB |
5701 | break; |
5702 | case 8: case 9: case 10: case 11: case 12: case 13: | |
5703 | /* VMLAL, VQDMLAL, VMLSL, VQDMLSL, VMULL, VQDMULL */ | |
ad69471c | 5704 | gen_neon_mull(cpu_V0, tmp, tmp2, size, u); |
9ee6e8bb PB |
5705 | break; |
5706 | case 14: /* Polynomial VMULL */ | |
e5ca24cb | 5707 | gen_helper_neon_mull_p8(cpu_V0, tmp, tmp2); |
7d1b0095 PM |
5708 | tcg_temp_free_i32(tmp2); |
5709 | tcg_temp_free_i32(tmp); | |
e5ca24cb | 5710 | break; |
695272dc PM |
5711 | default: /* 15 is RESERVED: caught earlier */ |
5712 | abort(); | |
9ee6e8bb | 5713 | } |
ebcd88ce PM |
5714 | if (op == 13) { |
5715 | /* VQDMULL */ | |
5716 | gen_neon_addl_saturate(cpu_V0, cpu_V0, size); | |
5717 | neon_store_reg64(cpu_V0, rd + pass); | |
5718 | } else if (op == 5 || (op >= 8 && op <= 11)) { | |
9ee6e8bb | 5719 | /* Accumulate. */ |
ebcd88ce | 5720 | neon_load_reg64(cpu_V1, rd + pass); |
9ee6e8bb | 5721 | switch (op) { |
4dc064e6 PM |
5722 | case 10: /* VMLSL */ |
5723 | gen_neon_negl(cpu_V0, size); | |
5724 | /* Fall through */ | |
5725 | case 5: case 8: /* VABAL, VMLAL */ | |
ad69471c | 5726 | gen_neon_addl(size); |
9ee6e8bb PB |
5727 | break; |
5728 | case 9: case 11: /* VQDMLAL, VQDMLSL */ | |
ad69471c | 5729 | gen_neon_addl_saturate(cpu_V0, cpu_V0, size); |
4dc064e6 PM |
5730 | if (op == 11) { |
5731 | gen_neon_negl(cpu_V0, size); | |
5732 | } | |
ad69471c PB |
5733 | gen_neon_addl_saturate(cpu_V0, cpu_V1, size); |
5734 | break; | |
9ee6e8bb PB |
5735 | default: |
5736 | abort(); | |
5737 | } | |
ad69471c | 5738 | neon_store_reg64(cpu_V0, rd + pass); |
9ee6e8bb PB |
5739 | } else if (op == 4 || op == 6) { |
5740 | /* Narrowing operation. */ | |
7d1b0095 | 5741 | tmp = tcg_temp_new_i32(); |
79b0e534 | 5742 | if (!u) { |
9ee6e8bb | 5743 | switch (size) { |
ad69471c PB |
5744 | case 0: |
5745 | gen_helper_neon_narrow_high_u8(tmp, cpu_V0); | |
5746 | break; | |
5747 | case 1: | |
5748 | gen_helper_neon_narrow_high_u16(tmp, cpu_V0); | |
5749 | break; | |
5750 | case 2: | |
664b7e3b | 5751 | tcg_gen_extrh_i64_i32(tmp, cpu_V0); |
ad69471c | 5752 | break; |
9ee6e8bb PB |
5753 | default: abort(); |
5754 | } | |
5755 | } else { | |
5756 | switch (size) { | |
ad69471c PB |
5757 | case 0: |
5758 | gen_helper_neon_narrow_round_high_u8(tmp, cpu_V0); | |
5759 | break; | |
5760 | case 1: | |
5761 | gen_helper_neon_narrow_round_high_u16(tmp, cpu_V0); | |
5762 | break; | |
5763 | case 2: | |
5764 | tcg_gen_addi_i64(cpu_V0, cpu_V0, 1u << 31); | |
664b7e3b | 5765 | tcg_gen_extrh_i64_i32(tmp, cpu_V0); |
ad69471c | 5766 | break; |
9ee6e8bb PB |
5767 | default: abort(); |
5768 | } | |
5769 | } | |
ad69471c PB |
5770 | if (pass == 0) { |
5771 | tmp3 = tmp; | |
5772 | } else { | |
5773 | neon_store_reg(rd, 0, tmp3); | |
5774 | neon_store_reg(rd, 1, tmp); | |
5775 | } | |
9ee6e8bb PB |
5776 | } else { |
5777 | /* Write back the result. */ | |
ad69471c | 5778 | neon_store_reg64(cpu_V0, rd + pass); |
9ee6e8bb PB |
5779 | } |
5780 | } | |
5781 | } else { | |
3e3326df PM |
5782 | /* Two registers and a scalar. NB that for ops of this form |
5783 | * the ARM ARM labels bit 24 as Q, but it is in our variable | |
5784 | * 'u', not 'q'. | |
5785 | */ | |
5786 | if (size == 0) { | |
5787 | return 1; | |
5788 | } | |
9ee6e8bb | 5789 | switch (op) { |
9ee6e8bb | 5790 | case 1: /* Float VMLA scalar */ |
9ee6e8bb | 5791 | case 5: /* Floating point VMLS scalar */ |
9ee6e8bb | 5792 | case 9: /* Floating point VMUL scalar */ |
3e3326df PM |
5793 | if (size == 1) { |
5794 | return 1; | |
5795 | } | |
5796 | /* fall through */ | |
5797 | case 0: /* Integer VMLA scalar */ | |
5798 | case 4: /* Integer VMLS scalar */ | |
5799 | case 8: /* Integer VMUL scalar */ | |
9ee6e8bb PB |
5800 | case 12: /* VQDMULH scalar */ |
5801 | case 13: /* VQRDMULH scalar */ | |
3e3326df PM |
5802 | if (u && ((rd | rn) & 1)) { |
5803 | return 1; | |
5804 | } | |
dd8fbd78 FN |
5805 | tmp = neon_get_scalar(size, rm); |
5806 | neon_store_scratch(0, tmp); | |
9ee6e8bb | 5807 | for (pass = 0; pass < (u ? 4 : 2); pass++) { |
dd8fbd78 FN |
5808 | tmp = neon_load_scratch(0); |
5809 | tmp2 = neon_load_reg(rn, pass); | |
9ee6e8bb PB |
5810 | if (op == 12) { |
5811 | if (size == 1) { | |
02da0b2d | 5812 | gen_helper_neon_qdmulh_s16(tmp, cpu_env, tmp, tmp2); |
9ee6e8bb | 5813 | } else { |
02da0b2d | 5814 | gen_helper_neon_qdmulh_s32(tmp, cpu_env, tmp, tmp2); |
9ee6e8bb PB |
5815 | } |
5816 | } else if (op == 13) { | |
5817 | if (size == 1) { | |
02da0b2d | 5818 | gen_helper_neon_qrdmulh_s16(tmp, cpu_env, tmp, tmp2); |
9ee6e8bb | 5819 | } else { |
02da0b2d | 5820 | gen_helper_neon_qrdmulh_s32(tmp, cpu_env, tmp, tmp2); |
9ee6e8bb PB |
5821 | } |
5822 | } else if (op & 1) { | |
aa47cfdd PM |
5823 | TCGv_ptr fpstatus = get_fpstatus_ptr(1); |
5824 | gen_helper_vfp_muls(tmp, tmp, tmp2, fpstatus); | |
5825 | tcg_temp_free_ptr(fpstatus); | |
9ee6e8bb PB |
5826 | } else { |
5827 | switch (size) { | |
dd8fbd78 FN |
5828 | case 0: gen_helper_neon_mul_u8(tmp, tmp, tmp2); break; |
5829 | case 1: gen_helper_neon_mul_u16(tmp, tmp, tmp2); break; | |
5830 | case 2: tcg_gen_mul_i32(tmp, tmp, tmp2); break; | |
3e3326df | 5831 | default: abort(); |
9ee6e8bb PB |
5832 | } |
5833 | } | |
7d1b0095 | 5834 | tcg_temp_free_i32(tmp2); |
9ee6e8bb PB |
5835 | if (op < 8) { |
5836 | /* Accumulate. */ | |
dd8fbd78 | 5837 | tmp2 = neon_load_reg(rd, pass); |
9ee6e8bb PB |
5838 | switch (op) { |
5839 | case 0: | |
dd8fbd78 | 5840 | gen_neon_add(size, tmp, tmp2); |
9ee6e8bb PB |
5841 | break; |
5842 | case 1: | |
aa47cfdd PM |
5843 | { |
5844 | TCGv_ptr fpstatus = get_fpstatus_ptr(1); | |
5845 | gen_helper_vfp_adds(tmp, tmp, tmp2, fpstatus); | |
5846 | tcg_temp_free_ptr(fpstatus); | |
9ee6e8bb | 5847 | break; |
aa47cfdd | 5848 | } |
9ee6e8bb | 5849 | case 4: |
dd8fbd78 | 5850 | gen_neon_rsb(size, tmp, tmp2); |
9ee6e8bb PB |
5851 | break; |
5852 | case 5: | |
aa47cfdd PM |
5853 | { |
5854 | TCGv_ptr fpstatus = get_fpstatus_ptr(1); | |
5855 | gen_helper_vfp_subs(tmp, tmp2, tmp, fpstatus); | |
5856 | tcg_temp_free_ptr(fpstatus); | |
9ee6e8bb | 5857 | break; |
aa47cfdd | 5858 | } |
9ee6e8bb PB |
5859 | default: |
5860 | abort(); | |
5861 | } | |
7d1b0095 | 5862 | tcg_temp_free_i32(tmp2); |
9ee6e8bb | 5863 | } |
dd8fbd78 | 5864 | neon_store_reg(rd, pass, tmp); |
9ee6e8bb PB |
5865 | } |
5866 | break; | |
9ee6e8bb | 5867 | case 3: /* VQDMLAL scalar */ |
9ee6e8bb | 5868 | case 7: /* VQDMLSL scalar */ |
9ee6e8bb | 5869 | case 11: /* VQDMULL scalar */ |
3e3326df | 5870 | if (u == 1) { |
ad69471c | 5871 | return 1; |
3e3326df PM |
5872 | } |
5873 | /* fall through */ | |
5874 | case 2: /* VMLAL sclar */ | |
5875 | case 6: /* VMLSL scalar */ | |
5876 | case 10: /* VMULL scalar */ | |
5877 | if (rd & 1) { | |
5878 | return 1; | |
5879 | } | |
dd8fbd78 | 5880 | tmp2 = neon_get_scalar(size, rm); |
c6067f04 CL |
5881 | /* We need a copy of tmp2 because gen_neon_mull |
5882 | * deletes it during pass 0. */ | |
7d1b0095 | 5883 | tmp4 = tcg_temp_new_i32(); |
c6067f04 | 5884 | tcg_gen_mov_i32(tmp4, tmp2); |
dd8fbd78 | 5885 | tmp3 = neon_load_reg(rn, 1); |
ad69471c | 5886 | |
9ee6e8bb | 5887 | for (pass = 0; pass < 2; pass++) { |
ad69471c PB |
5888 | if (pass == 0) { |
5889 | tmp = neon_load_reg(rn, 0); | |
9ee6e8bb | 5890 | } else { |
dd8fbd78 | 5891 | tmp = tmp3; |
c6067f04 | 5892 | tmp2 = tmp4; |
9ee6e8bb | 5893 | } |
ad69471c | 5894 | gen_neon_mull(cpu_V0, tmp, tmp2, size, u); |
ad69471c PB |
5895 | if (op != 11) { |
5896 | neon_load_reg64(cpu_V1, rd + pass); | |
9ee6e8bb | 5897 | } |
9ee6e8bb | 5898 | switch (op) { |
4dc064e6 PM |
5899 | case 6: |
5900 | gen_neon_negl(cpu_V0, size); | |
5901 | /* Fall through */ | |
5902 | case 2: | |
ad69471c | 5903 | gen_neon_addl(size); |
9ee6e8bb PB |
5904 | break; |
5905 | case 3: case 7: | |
ad69471c | 5906 | gen_neon_addl_saturate(cpu_V0, cpu_V0, size); |
4dc064e6 PM |
5907 | if (op == 7) { |
5908 | gen_neon_negl(cpu_V0, size); | |
5909 | } | |
ad69471c | 5910 | gen_neon_addl_saturate(cpu_V0, cpu_V1, size); |
9ee6e8bb PB |
5911 | break; |
5912 | case 10: | |
5913 | /* no-op */ | |
5914 | break; | |
5915 | case 11: | |
ad69471c | 5916 | gen_neon_addl_saturate(cpu_V0, cpu_V0, size); |
9ee6e8bb PB |
5917 | break; |
5918 | default: | |
5919 | abort(); | |
5920 | } | |
ad69471c | 5921 | neon_store_reg64(cpu_V0, rd + pass); |
9ee6e8bb | 5922 | } |
61adacc8 RH |
5923 | break; |
5924 | case 14: /* VQRDMLAH scalar */ | |
5925 | case 15: /* VQRDMLSH scalar */ | |
5926 | { | |
5927 | NeonGenThreeOpEnvFn *fn; | |
dd8fbd78 | 5928 | |
962fcbf2 | 5929 | if (!dc_isar_feature(aa32_rdm, s)) { |
61adacc8 RH |
5930 | return 1; |
5931 | } | |
5932 | if (u && ((rd | rn) & 1)) { | |
5933 | return 1; | |
5934 | } | |
5935 | if (op == 14) { | |
5936 | if (size == 1) { | |
5937 | fn = gen_helper_neon_qrdmlah_s16; | |
5938 | } else { | |
5939 | fn = gen_helper_neon_qrdmlah_s32; | |
5940 | } | |
5941 | } else { | |
5942 | if (size == 1) { | |
5943 | fn = gen_helper_neon_qrdmlsh_s16; | |
5944 | } else { | |
5945 | fn = gen_helper_neon_qrdmlsh_s32; | |
5946 | } | |
5947 | } | |
dd8fbd78 | 5948 | |
61adacc8 RH |
5949 | tmp2 = neon_get_scalar(size, rm); |
5950 | for (pass = 0; pass < (u ? 4 : 2); pass++) { | |
5951 | tmp = neon_load_reg(rn, pass); | |
5952 | tmp3 = neon_load_reg(rd, pass); | |
5953 | fn(tmp, cpu_env, tmp, tmp2, tmp3); | |
5954 | tcg_temp_free_i32(tmp3); | |
5955 | neon_store_reg(rd, pass, tmp); | |
5956 | } | |
5957 | tcg_temp_free_i32(tmp2); | |
5958 | } | |
9ee6e8bb | 5959 | break; |
61adacc8 RH |
5960 | default: |
5961 | g_assert_not_reached(); | |
9ee6e8bb PB |
5962 | } |
5963 | } | |
5964 | } else { /* size == 3 */ | |
5965 | if (!u) { | |
5966 | /* Extract. */ | |
9ee6e8bb | 5967 | imm = (insn >> 8) & 0xf; |
ad69471c PB |
5968 | |
5969 | if (imm > 7 && !q) | |
5970 | return 1; | |
5971 | ||
52579ea1 PM |
5972 | if (q && ((rd | rn | rm) & 1)) { |
5973 | return 1; | |
5974 | } | |
5975 | ||
ad69471c PB |
5976 | if (imm == 0) { |
5977 | neon_load_reg64(cpu_V0, rn); | |
5978 | if (q) { | |
5979 | neon_load_reg64(cpu_V1, rn + 1); | |
9ee6e8bb | 5980 | } |
ad69471c PB |
5981 | } else if (imm == 8) { |
5982 | neon_load_reg64(cpu_V0, rn + 1); | |
5983 | if (q) { | |
5984 | neon_load_reg64(cpu_V1, rm); | |
9ee6e8bb | 5985 | } |
ad69471c | 5986 | } else if (q) { |
a7812ae4 | 5987 | tmp64 = tcg_temp_new_i64(); |
ad69471c PB |
5988 | if (imm < 8) { |
5989 | neon_load_reg64(cpu_V0, rn); | |
a7812ae4 | 5990 | neon_load_reg64(tmp64, rn + 1); |
ad69471c PB |
5991 | } else { |
5992 | neon_load_reg64(cpu_V0, rn + 1); | |
a7812ae4 | 5993 | neon_load_reg64(tmp64, rm); |
ad69471c PB |
5994 | } |
5995 | tcg_gen_shri_i64(cpu_V0, cpu_V0, (imm & 7) * 8); | |
a7812ae4 | 5996 | tcg_gen_shli_i64(cpu_V1, tmp64, 64 - ((imm & 7) * 8)); |
ad69471c PB |
5997 | tcg_gen_or_i64(cpu_V0, cpu_V0, cpu_V1); |
5998 | if (imm < 8) { | |
5999 | neon_load_reg64(cpu_V1, rm); | |
9ee6e8bb | 6000 | } else { |
ad69471c PB |
6001 | neon_load_reg64(cpu_V1, rm + 1); |
6002 | imm -= 8; | |
9ee6e8bb | 6003 | } |
ad69471c | 6004 | tcg_gen_shli_i64(cpu_V1, cpu_V1, 64 - (imm * 8)); |
a7812ae4 PB |
6005 | tcg_gen_shri_i64(tmp64, tmp64, imm * 8); |
6006 | tcg_gen_or_i64(cpu_V1, cpu_V1, tmp64); | |
b75263d6 | 6007 | tcg_temp_free_i64(tmp64); |
ad69471c | 6008 | } else { |
a7812ae4 | 6009 | /* BUGFIX */ |
ad69471c | 6010 | neon_load_reg64(cpu_V0, rn); |
a7812ae4 | 6011 | tcg_gen_shri_i64(cpu_V0, cpu_V0, imm * 8); |
ad69471c | 6012 | neon_load_reg64(cpu_V1, rm); |
a7812ae4 | 6013 | tcg_gen_shli_i64(cpu_V1, cpu_V1, 64 - (imm * 8)); |
ad69471c PB |
6014 | tcg_gen_or_i64(cpu_V0, cpu_V0, cpu_V1); |
6015 | } | |
6016 | neon_store_reg64(cpu_V0, rd); | |
6017 | if (q) { | |
6018 | neon_store_reg64(cpu_V1, rd + 1); | |
9ee6e8bb PB |
6019 | } |
6020 | } else if ((insn & (1 << 11)) == 0) { | |
6021 | /* Two register misc. */ | |
6022 | op = ((insn >> 12) & 0x30) | ((insn >> 7) & 0xf); | |
6023 | size = (insn >> 18) & 3; | |
600b828c PM |
6024 | /* UNDEF for unknown op values and bad op-size combinations */ |
6025 | if ((neon_2rm_sizes[op] & (1 << size)) == 0) { | |
6026 | return 1; | |
6027 | } | |
fe8fcf3d PM |
6028 | if (neon_2rm_is_v8_op(op) && |
6029 | !arm_dc_feature(s, ARM_FEATURE_V8)) { | |
6030 | return 1; | |
6031 | } | |
fc2a9b37 PM |
6032 | if ((op != NEON_2RM_VMOVN && op != NEON_2RM_VQMOVN) && |
6033 | q && ((rm | rd) & 1)) { | |
6034 | return 1; | |
6035 | } | |
9ee6e8bb | 6036 | switch (op) { |
600b828c | 6037 | case NEON_2RM_VREV64: |
9ee6e8bb | 6038 | for (pass = 0; pass < (q ? 2 : 1); pass++) { |
dd8fbd78 FN |
6039 | tmp = neon_load_reg(rm, pass * 2); |
6040 | tmp2 = neon_load_reg(rm, pass * 2 + 1); | |
9ee6e8bb | 6041 | switch (size) { |
dd8fbd78 FN |
6042 | case 0: tcg_gen_bswap32_i32(tmp, tmp); break; |
6043 | case 1: gen_swap_half(tmp); break; | |
9ee6e8bb PB |
6044 | case 2: /* no-op */ break; |
6045 | default: abort(); | |
6046 | } | |
dd8fbd78 | 6047 | neon_store_reg(rd, pass * 2 + 1, tmp); |
9ee6e8bb | 6048 | if (size == 2) { |
dd8fbd78 | 6049 | neon_store_reg(rd, pass * 2, tmp2); |
9ee6e8bb | 6050 | } else { |
9ee6e8bb | 6051 | switch (size) { |
dd8fbd78 FN |
6052 | case 0: tcg_gen_bswap32_i32(tmp2, tmp2); break; |
6053 | case 1: gen_swap_half(tmp2); break; | |
9ee6e8bb PB |
6054 | default: abort(); |
6055 | } | |
dd8fbd78 | 6056 | neon_store_reg(rd, pass * 2, tmp2); |
9ee6e8bb PB |
6057 | } |
6058 | } | |
6059 | break; | |
600b828c PM |
6060 | case NEON_2RM_VPADDL: case NEON_2RM_VPADDL_U: |
6061 | case NEON_2RM_VPADAL: case NEON_2RM_VPADAL_U: | |
ad69471c PB |
6062 | for (pass = 0; pass < q + 1; pass++) { |
6063 | tmp = neon_load_reg(rm, pass * 2); | |
6064 | gen_neon_widen(cpu_V0, tmp, size, op & 1); | |
6065 | tmp = neon_load_reg(rm, pass * 2 + 1); | |
6066 | gen_neon_widen(cpu_V1, tmp, size, op & 1); | |
6067 | switch (size) { | |
6068 | case 0: gen_helper_neon_paddl_u16(CPU_V001); break; | |
6069 | case 1: gen_helper_neon_paddl_u32(CPU_V001); break; | |
6070 | case 2: tcg_gen_add_i64(CPU_V001); break; | |
6071 | default: abort(); | |
6072 | } | |
600b828c | 6073 | if (op >= NEON_2RM_VPADAL) { |
9ee6e8bb | 6074 | /* Accumulate. */ |
ad69471c PB |
6075 | neon_load_reg64(cpu_V1, rd + pass); |
6076 | gen_neon_addl(size); | |
9ee6e8bb | 6077 | } |
ad69471c | 6078 | neon_store_reg64(cpu_V0, rd + pass); |
9ee6e8bb PB |
6079 | } |
6080 | break; | |
600b828c | 6081 | case NEON_2RM_VTRN: |
9ee6e8bb | 6082 | if (size == 2) { |
a5a14945 | 6083 | int n; |
9ee6e8bb | 6084 | for (n = 0; n < (q ? 4 : 2); n += 2) { |
dd8fbd78 FN |
6085 | tmp = neon_load_reg(rm, n); |
6086 | tmp2 = neon_load_reg(rd, n + 1); | |
6087 | neon_store_reg(rm, n, tmp2); | |
6088 | neon_store_reg(rd, n + 1, tmp); | |
9ee6e8bb PB |
6089 | } |
6090 | } else { | |
6091 | goto elementwise; | |
6092 | } | |
6093 | break; | |
600b828c | 6094 | case NEON_2RM_VUZP: |
02acedf9 | 6095 | if (gen_neon_unzip(rd, rm, size, q)) { |
9ee6e8bb | 6096 | return 1; |
9ee6e8bb PB |
6097 | } |
6098 | break; | |
600b828c | 6099 | case NEON_2RM_VZIP: |
d68a6f3a | 6100 | if (gen_neon_zip(rd, rm, size, q)) { |
9ee6e8bb | 6101 | return 1; |
9ee6e8bb PB |
6102 | } |
6103 | break; | |
600b828c PM |
6104 | case NEON_2RM_VMOVN: case NEON_2RM_VQMOVN: |
6105 | /* also VQMOVUN; op field and mnemonics don't line up */ | |
fc2a9b37 PM |
6106 | if (rm & 1) { |
6107 | return 1; | |
6108 | } | |
f764718d | 6109 | tmp2 = NULL; |
9ee6e8bb | 6110 | for (pass = 0; pass < 2; pass++) { |
ad69471c | 6111 | neon_load_reg64(cpu_V0, rm + pass); |
7d1b0095 | 6112 | tmp = tcg_temp_new_i32(); |
600b828c PM |
6113 | gen_neon_narrow_op(op == NEON_2RM_VMOVN, q, size, |
6114 | tmp, cpu_V0); | |
ad69471c PB |
6115 | if (pass == 0) { |
6116 | tmp2 = tmp; | |
6117 | } else { | |
6118 | neon_store_reg(rd, 0, tmp2); | |
6119 | neon_store_reg(rd, 1, tmp); | |
9ee6e8bb | 6120 | } |
9ee6e8bb PB |
6121 | } |
6122 | break; | |
600b828c | 6123 | case NEON_2RM_VSHLL: |
fc2a9b37 | 6124 | if (q || (rd & 1)) { |
9ee6e8bb | 6125 | return 1; |
600b828c | 6126 | } |
ad69471c PB |
6127 | tmp = neon_load_reg(rm, 0); |
6128 | tmp2 = neon_load_reg(rm, 1); | |
9ee6e8bb | 6129 | for (pass = 0; pass < 2; pass++) { |
ad69471c PB |
6130 | if (pass == 1) |
6131 | tmp = tmp2; | |
6132 | gen_neon_widen(cpu_V0, tmp, size, 1); | |
30d11a2a | 6133 | tcg_gen_shli_i64(cpu_V0, cpu_V0, 8 << size); |
ad69471c | 6134 | neon_store_reg64(cpu_V0, rd + pass); |
9ee6e8bb PB |
6135 | } |
6136 | break; | |
600b828c | 6137 | case NEON_2RM_VCVT_F16_F32: |
486624fc AB |
6138 | { |
6139 | TCGv_ptr fpst; | |
6140 | TCGv_i32 ahp; | |
6141 | ||
602f6e42 | 6142 | if (!dc_isar_feature(aa32_fp16_spconv, s) || |
fc2a9b37 PM |
6143 | q || (rm & 1)) { |
6144 | return 1; | |
6145 | } | |
486624fc AB |
6146 | fpst = get_fpstatus_ptr(true); |
6147 | ahp = get_ahp_flag(); | |
58f2682e PM |
6148 | tmp = neon_load_reg(rm, 0); |
6149 | gen_helper_vfp_fcvt_f32_to_f16(tmp, tmp, fpst, ahp); | |
6150 | tmp2 = neon_load_reg(rm, 1); | |
6151 | gen_helper_vfp_fcvt_f32_to_f16(tmp2, tmp2, fpst, ahp); | |
60011498 PB |
6152 | tcg_gen_shli_i32(tmp2, tmp2, 16); |
6153 | tcg_gen_or_i32(tmp2, tmp2, tmp); | |
58f2682e PM |
6154 | tcg_temp_free_i32(tmp); |
6155 | tmp = neon_load_reg(rm, 2); | |
6156 | gen_helper_vfp_fcvt_f32_to_f16(tmp, tmp, fpst, ahp); | |
6157 | tmp3 = neon_load_reg(rm, 3); | |
60011498 | 6158 | neon_store_reg(rd, 0, tmp2); |
58f2682e PM |
6159 | gen_helper_vfp_fcvt_f32_to_f16(tmp3, tmp3, fpst, ahp); |
6160 | tcg_gen_shli_i32(tmp3, tmp3, 16); | |
6161 | tcg_gen_or_i32(tmp3, tmp3, tmp); | |
6162 | neon_store_reg(rd, 1, tmp3); | |
7d1b0095 | 6163 | tcg_temp_free_i32(tmp); |
486624fc AB |
6164 | tcg_temp_free_i32(ahp); |
6165 | tcg_temp_free_ptr(fpst); | |
60011498 | 6166 | break; |
486624fc | 6167 | } |
600b828c | 6168 | case NEON_2RM_VCVT_F32_F16: |
486624fc AB |
6169 | { |
6170 | TCGv_ptr fpst; | |
6171 | TCGv_i32 ahp; | |
602f6e42 | 6172 | if (!dc_isar_feature(aa32_fp16_spconv, s) || |
fc2a9b37 PM |
6173 | q || (rd & 1)) { |
6174 | return 1; | |
6175 | } | |
486624fc AB |
6176 | fpst = get_fpstatus_ptr(true); |
6177 | ahp = get_ahp_flag(); | |
7d1b0095 | 6178 | tmp3 = tcg_temp_new_i32(); |
60011498 PB |
6179 | tmp = neon_load_reg(rm, 0); |
6180 | tmp2 = neon_load_reg(rm, 1); | |
6181 | tcg_gen_ext16u_i32(tmp3, tmp); | |
b66f6b99 PM |
6182 | gen_helper_vfp_fcvt_f16_to_f32(tmp3, tmp3, fpst, ahp); |
6183 | neon_store_reg(rd, 0, tmp3); | |
6184 | tcg_gen_shri_i32(tmp, tmp, 16); | |
6185 | gen_helper_vfp_fcvt_f16_to_f32(tmp, tmp, fpst, ahp); | |
6186 | neon_store_reg(rd, 1, tmp); | |
6187 | tmp3 = tcg_temp_new_i32(); | |
60011498 | 6188 | tcg_gen_ext16u_i32(tmp3, tmp2); |
b66f6b99 PM |
6189 | gen_helper_vfp_fcvt_f16_to_f32(tmp3, tmp3, fpst, ahp); |
6190 | neon_store_reg(rd, 2, tmp3); | |
6191 | tcg_gen_shri_i32(tmp2, tmp2, 16); | |
6192 | gen_helper_vfp_fcvt_f16_to_f32(tmp2, tmp2, fpst, ahp); | |
6193 | neon_store_reg(rd, 3, tmp2); | |
486624fc AB |
6194 | tcg_temp_free_i32(ahp); |
6195 | tcg_temp_free_ptr(fpst); | |
60011498 | 6196 | break; |
486624fc | 6197 | } |
9d935509 | 6198 | case NEON_2RM_AESE: case NEON_2RM_AESMC: |
962fcbf2 | 6199 | if (!dc_isar_feature(aa32_aes, s) || ((rm | rd) & 1)) { |
9d935509 AB |
6200 | return 1; |
6201 | } | |
1a66ac61 RH |
6202 | ptr1 = vfp_reg_ptr(true, rd); |
6203 | ptr2 = vfp_reg_ptr(true, rm); | |
9d935509 AB |
6204 | |
6205 | /* Bit 6 is the lowest opcode bit; it distinguishes between | |
6206 | * encryption (AESE/AESMC) and decryption (AESD/AESIMC) | |
6207 | */ | |
6208 | tmp3 = tcg_const_i32(extract32(insn, 6, 1)); | |
6209 | ||
6210 | if (op == NEON_2RM_AESE) { | |
1a66ac61 | 6211 | gen_helper_crypto_aese(ptr1, ptr2, tmp3); |
9d935509 | 6212 | } else { |
1a66ac61 | 6213 | gen_helper_crypto_aesmc(ptr1, ptr2, tmp3); |
9d935509 | 6214 | } |
1a66ac61 RH |
6215 | tcg_temp_free_ptr(ptr1); |
6216 | tcg_temp_free_ptr(ptr2); | |
9d935509 AB |
6217 | tcg_temp_free_i32(tmp3); |
6218 | break; | |
f1ecb913 | 6219 | case NEON_2RM_SHA1H: |
962fcbf2 | 6220 | if (!dc_isar_feature(aa32_sha1, s) || ((rm | rd) & 1)) { |
f1ecb913 AB |
6221 | return 1; |
6222 | } | |
1a66ac61 RH |
6223 | ptr1 = vfp_reg_ptr(true, rd); |
6224 | ptr2 = vfp_reg_ptr(true, rm); | |
f1ecb913 | 6225 | |
1a66ac61 | 6226 | gen_helper_crypto_sha1h(ptr1, ptr2); |
f1ecb913 | 6227 | |
1a66ac61 RH |
6228 | tcg_temp_free_ptr(ptr1); |
6229 | tcg_temp_free_ptr(ptr2); | |
f1ecb913 AB |
6230 | break; |
6231 | case NEON_2RM_SHA1SU1: | |
6232 | if ((rm | rd) & 1) { | |
6233 | return 1; | |
6234 | } | |
6235 | /* bit 6 (q): set -> SHA256SU0, cleared -> SHA1SU1 */ | |
6236 | if (q) { | |
962fcbf2 | 6237 | if (!dc_isar_feature(aa32_sha2, s)) { |
f1ecb913 AB |
6238 | return 1; |
6239 | } | |
962fcbf2 | 6240 | } else if (!dc_isar_feature(aa32_sha1, s)) { |
f1ecb913 AB |
6241 | return 1; |
6242 | } | |
1a66ac61 RH |
6243 | ptr1 = vfp_reg_ptr(true, rd); |
6244 | ptr2 = vfp_reg_ptr(true, rm); | |
f1ecb913 | 6245 | if (q) { |
1a66ac61 | 6246 | gen_helper_crypto_sha256su0(ptr1, ptr2); |
f1ecb913 | 6247 | } else { |
1a66ac61 | 6248 | gen_helper_crypto_sha1su1(ptr1, ptr2); |
f1ecb913 | 6249 | } |
1a66ac61 RH |
6250 | tcg_temp_free_ptr(ptr1); |
6251 | tcg_temp_free_ptr(ptr2); | |
f1ecb913 | 6252 | break; |
4bf940be RH |
6253 | |
6254 | case NEON_2RM_VMVN: | |
6255 | tcg_gen_gvec_not(0, rd_ofs, rm_ofs, vec_size, vec_size); | |
6256 | break; | |
6257 | case NEON_2RM_VNEG: | |
6258 | tcg_gen_gvec_neg(size, rd_ofs, rm_ofs, vec_size, vec_size); | |
6259 | break; | |
4e027a71 RH |
6260 | case NEON_2RM_VABS: |
6261 | tcg_gen_gvec_abs(size, rd_ofs, rm_ofs, vec_size, vec_size); | |
6262 | break; | |
4bf940be | 6263 | |
9ee6e8bb PB |
6264 | default: |
6265 | elementwise: | |
6266 | for (pass = 0; pass < (q ? 4 : 2); pass++) { | |
60737ed5 | 6267 | tmp = neon_load_reg(rm, pass); |
9ee6e8bb | 6268 | switch (op) { |
600b828c | 6269 | case NEON_2RM_VREV32: |
9ee6e8bb | 6270 | switch (size) { |
dd8fbd78 FN |
6271 | case 0: tcg_gen_bswap32_i32(tmp, tmp); break; |
6272 | case 1: gen_swap_half(tmp); break; | |
600b828c | 6273 | default: abort(); |
9ee6e8bb PB |
6274 | } |
6275 | break; | |
600b828c | 6276 | case NEON_2RM_VREV16: |
46497f6a | 6277 | gen_rev16(tmp, tmp); |
9ee6e8bb | 6278 | break; |
600b828c | 6279 | case NEON_2RM_VCLS: |
9ee6e8bb | 6280 | switch (size) { |
dd8fbd78 FN |
6281 | case 0: gen_helper_neon_cls_s8(tmp, tmp); break; |
6282 | case 1: gen_helper_neon_cls_s16(tmp, tmp); break; | |
6283 | case 2: gen_helper_neon_cls_s32(tmp, tmp); break; | |
600b828c | 6284 | default: abort(); |
9ee6e8bb PB |
6285 | } |
6286 | break; | |
600b828c | 6287 | case NEON_2RM_VCLZ: |
9ee6e8bb | 6288 | switch (size) { |
dd8fbd78 FN |
6289 | case 0: gen_helper_neon_clz_u8(tmp, tmp); break; |
6290 | case 1: gen_helper_neon_clz_u16(tmp, tmp); break; | |
7539a012 | 6291 | case 2: tcg_gen_clzi_i32(tmp, tmp, 32); break; |
600b828c | 6292 | default: abort(); |
9ee6e8bb PB |
6293 | } |
6294 | break; | |
600b828c | 6295 | case NEON_2RM_VCNT: |
dd8fbd78 | 6296 | gen_helper_neon_cnt_u8(tmp, tmp); |
9ee6e8bb | 6297 | break; |
600b828c | 6298 | case NEON_2RM_VQABS: |
9ee6e8bb | 6299 | switch (size) { |
02da0b2d PM |
6300 | case 0: |
6301 | gen_helper_neon_qabs_s8(tmp, cpu_env, tmp); | |
6302 | break; | |
6303 | case 1: | |
6304 | gen_helper_neon_qabs_s16(tmp, cpu_env, tmp); | |
6305 | break; | |
6306 | case 2: | |
6307 | gen_helper_neon_qabs_s32(tmp, cpu_env, tmp); | |
6308 | break; | |
600b828c | 6309 | default: abort(); |
9ee6e8bb PB |
6310 | } |
6311 | break; | |
600b828c | 6312 | case NEON_2RM_VQNEG: |
9ee6e8bb | 6313 | switch (size) { |
02da0b2d PM |
6314 | case 0: |
6315 | gen_helper_neon_qneg_s8(tmp, cpu_env, tmp); | |
6316 | break; | |
6317 | case 1: | |
6318 | gen_helper_neon_qneg_s16(tmp, cpu_env, tmp); | |
6319 | break; | |
6320 | case 2: | |
6321 | gen_helper_neon_qneg_s32(tmp, cpu_env, tmp); | |
6322 | break; | |
600b828c | 6323 | default: abort(); |
9ee6e8bb PB |
6324 | } |
6325 | break; | |
600b828c | 6326 | case NEON_2RM_VCGT0: case NEON_2RM_VCLE0: |
dd8fbd78 | 6327 | tmp2 = tcg_const_i32(0); |
9ee6e8bb | 6328 | switch(size) { |
dd8fbd78 FN |
6329 | case 0: gen_helper_neon_cgt_s8(tmp, tmp, tmp2); break; |
6330 | case 1: gen_helper_neon_cgt_s16(tmp, tmp, tmp2); break; | |
6331 | case 2: gen_helper_neon_cgt_s32(tmp, tmp, tmp2); break; | |
600b828c | 6332 | default: abort(); |
9ee6e8bb | 6333 | } |
39d5492a | 6334 | tcg_temp_free_i32(tmp2); |
600b828c | 6335 | if (op == NEON_2RM_VCLE0) { |
dd8fbd78 | 6336 | tcg_gen_not_i32(tmp, tmp); |
600b828c | 6337 | } |
9ee6e8bb | 6338 | break; |
600b828c | 6339 | case NEON_2RM_VCGE0: case NEON_2RM_VCLT0: |
dd8fbd78 | 6340 | tmp2 = tcg_const_i32(0); |
9ee6e8bb | 6341 | switch(size) { |
dd8fbd78 FN |
6342 | case 0: gen_helper_neon_cge_s8(tmp, tmp, tmp2); break; |
6343 | case 1: gen_helper_neon_cge_s16(tmp, tmp, tmp2); break; | |
6344 | case 2: gen_helper_neon_cge_s32(tmp, tmp, tmp2); break; | |
600b828c | 6345 | default: abort(); |
9ee6e8bb | 6346 | } |
39d5492a | 6347 | tcg_temp_free_i32(tmp2); |
600b828c | 6348 | if (op == NEON_2RM_VCLT0) { |
dd8fbd78 | 6349 | tcg_gen_not_i32(tmp, tmp); |
600b828c | 6350 | } |
9ee6e8bb | 6351 | break; |
600b828c | 6352 | case NEON_2RM_VCEQ0: |
dd8fbd78 | 6353 | tmp2 = tcg_const_i32(0); |
9ee6e8bb | 6354 | switch(size) { |
dd8fbd78 FN |
6355 | case 0: gen_helper_neon_ceq_u8(tmp, tmp, tmp2); break; |
6356 | case 1: gen_helper_neon_ceq_u16(tmp, tmp, tmp2); break; | |
6357 | case 2: gen_helper_neon_ceq_u32(tmp, tmp, tmp2); break; | |
600b828c | 6358 | default: abort(); |
9ee6e8bb | 6359 | } |
39d5492a | 6360 | tcg_temp_free_i32(tmp2); |
9ee6e8bb | 6361 | break; |
600b828c | 6362 | case NEON_2RM_VCGT0_F: |
aa47cfdd PM |
6363 | { |
6364 | TCGv_ptr fpstatus = get_fpstatus_ptr(1); | |
dd8fbd78 | 6365 | tmp2 = tcg_const_i32(0); |
aa47cfdd | 6366 | gen_helper_neon_cgt_f32(tmp, tmp, tmp2, fpstatus); |
39d5492a | 6367 | tcg_temp_free_i32(tmp2); |
aa47cfdd | 6368 | tcg_temp_free_ptr(fpstatus); |
9ee6e8bb | 6369 | break; |
aa47cfdd | 6370 | } |
600b828c | 6371 | case NEON_2RM_VCGE0_F: |
aa47cfdd PM |
6372 | { |
6373 | TCGv_ptr fpstatus = get_fpstatus_ptr(1); | |
dd8fbd78 | 6374 | tmp2 = tcg_const_i32(0); |
aa47cfdd | 6375 | gen_helper_neon_cge_f32(tmp, tmp, tmp2, fpstatus); |
39d5492a | 6376 | tcg_temp_free_i32(tmp2); |
aa47cfdd | 6377 | tcg_temp_free_ptr(fpstatus); |
9ee6e8bb | 6378 | break; |
aa47cfdd | 6379 | } |
600b828c | 6380 | case NEON_2RM_VCEQ0_F: |
aa47cfdd PM |
6381 | { |
6382 | TCGv_ptr fpstatus = get_fpstatus_ptr(1); | |
dd8fbd78 | 6383 | tmp2 = tcg_const_i32(0); |
aa47cfdd | 6384 | gen_helper_neon_ceq_f32(tmp, tmp, tmp2, fpstatus); |
39d5492a | 6385 | tcg_temp_free_i32(tmp2); |
aa47cfdd | 6386 | tcg_temp_free_ptr(fpstatus); |
9ee6e8bb | 6387 | break; |
aa47cfdd | 6388 | } |
600b828c | 6389 | case NEON_2RM_VCLE0_F: |
aa47cfdd PM |
6390 | { |
6391 | TCGv_ptr fpstatus = get_fpstatus_ptr(1); | |
0e326109 | 6392 | tmp2 = tcg_const_i32(0); |
aa47cfdd | 6393 | gen_helper_neon_cge_f32(tmp, tmp2, tmp, fpstatus); |
39d5492a | 6394 | tcg_temp_free_i32(tmp2); |
aa47cfdd | 6395 | tcg_temp_free_ptr(fpstatus); |
0e326109 | 6396 | break; |
aa47cfdd | 6397 | } |
600b828c | 6398 | case NEON_2RM_VCLT0_F: |
aa47cfdd PM |
6399 | { |
6400 | TCGv_ptr fpstatus = get_fpstatus_ptr(1); | |
0e326109 | 6401 | tmp2 = tcg_const_i32(0); |
aa47cfdd | 6402 | gen_helper_neon_cgt_f32(tmp, tmp2, tmp, fpstatus); |
39d5492a | 6403 | tcg_temp_free_i32(tmp2); |
aa47cfdd | 6404 | tcg_temp_free_ptr(fpstatus); |
0e326109 | 6405 | break; |
aa47cfdd | 6406 | } |
600b828c | 6407 | case NEON_2RM_VABS_F: |
fd8a68cd | 6408 | gen_helper_vfp_abss(tmp, tmp); |
9ee6e8bb | 6409 | break; |
600b828c | 6410 | case NEON_2RM_VNEG_F: |
cedcc96f | 6411 | gen_helper_vfp_negs(tmp, tmp); |
9ee6e8bb | 6412 | break; |
600b828c | 6413 | case NEON_2RM_VSWP: |
dd8fbd78 FN |
6414 | tmp2 = neon_load_reg(rd, pass); |
6415 | neon_store_reg(rm, pass, tmp2); | |
9ee6e8bb | 6416 | break; |
600b828c | 6417 | case NEON_2RM_VTRN: |
dd8fbd78 | 6418 | tmp2 = neon_load_reg(rd, pass); |
9ee6e8bb | 6419 | switch (size) { |
dd8fbd78 FN |
6420 | case 0: gen_neon_trn_u8(tmp, tmp2); break; |
6421 | case 1: gen_neon_trn_u16(tmp, tmp2); break; | |
600b828c | 6422 | default: abort(); |
9ee6e8bb | 6423 | } |
dd8fbd78 | 6424 | neon_store_reg(rm, pass, tmp2); |
9ee6e8bb | 6425 | break; |
34f7b0a2 WN |
6426 | case NEON_2RM_VRINTN: |
6427 | case NEON_2RM_VRINTA: | |
6428 | case NEON_2RM_VRINTM: | |
6429 | case NEON_2RM_VRINTP: | |
6430 | case NEON_2RM_VRINTZ: | |
6431 | { | |
6432 | TCGv_i32 tcg_rmode; | |
6433 | TCGv_ptr fpstatus = get_fpstatus_ptr(1); | |
6434 | int rmode; | |
6435 | ||
6436 | if (op == NEON_2RM_VRINTZ) { | |
6437 | rmode = FPROUNDING_ZERO; | |
6438 | } else { | |
6439 | rmode = fp_decode_rm[((op & 0x6) >> 1) ^ 1]; | |
6440 | } | |
6441 | ||
6442 | tcg_rmode = tcg_const_i32(arm_rmode_to_sf(rmode)); | |
6443 | gen_helper_set_neon_rmode(tcg_rmode, tcg_rmode, | |
6444 | cpu_env); | |
3b52ad1f | 6445 | gen_helper_rints(tmp, tmp, fpstatus); |
34f7b0a2 WN |
6446 | gen_helper_set_neon_rmode(tcg_rmode, tcg_rmode, |
6447 | cpu_env); | |
6448 | tcg_temp_free_ptr(fpstatus); | |
6449 | tcg_temp_free_i32(tcg_rmode); | |
6450 | break; | |
6451 | } | |
2ce70625 WN |
6452 | case NEON_2RM_VRINTX: |
6453 | { | |
6454 | TCGv_ptr fpstatus = get_fpstatus_ptr(1); | |
3b52ad1f | 6455 | gen_helper_rints_exact(tmp, tmp, fpstatus); |
2ce70625 WN |
6456 | tcg_temp_free_ptr(fpstatus); |
6457 | break; | |
6458 | } | |
901ad525 WN |
6459 | case NEON_2RM_VCVTAU: |
6460 | case NEON_2RM_VCVTAS: | |
6461 | case NEON_2RM_VCVTNU: | |
6462 | case NEON_2RM_VCVTNS: | |
6463 | case NEON_2RM_VCVTPU: | |
6464 | case NEON_2RM_VCVTPS: | |
6465 | case NEON_2RM_VCVTMU: | |
6466 | case NEON_2RM_VCVTMS: | |
6467 | { | |
6468 | bool is_signed = !extract32(insn, 7, 1); | |
6469 | TCGv_ptr fpst = get_fpstatus_ptr(1); | |
6470 | TCGv_i32 tcg_rmode, tcg_shift; | |
6471 | int rmode = fp_decode_rm[extract32(insn, 8, 2)]; | |
6472 | ||
6473 | tcg_shift = tcg_const_i32(0); | |
6474 | tcg_rmode = tcg_const_i32(arm_rmode_to_sf(rmode)); | |
6475 | gen_helper_set_neon_rmode(tcg_rmode, tcg_rmode, | |
6476 | cpu_env); | |
6477 | ||
6478 | if (is_signed) { | |
30bf0a01 | 6479 | gen_helper_vfp_tosls(tmp, tmp, |
901ad525 WN |
6480 | tcg_shift, fpst); |
6481 | } else { | |
30bf0a01 | 6482 | gen_helper_vfp_touls(tmp, tmp, |
901ad525 WN |
6483 | tcg_shift, fpst); |
6484 | } | |
6485 | ||
6486 | gen_helper_set_neon_rmode(tcg_rmode, tcg_rmode, | |
6487 | cpu_env); | |
6488 | tcg_temp_free_i32(tcg_rmode); | |
6489 | tcg_temp_free_i32(tcg_shift); | |
6490 | tcg_temp_free_ptr(fpst); | |
6491 | break; | |
6492 | } | |
600b828c | 6493 | case NEON_2RM_VRECPE: |
b6d4443a AB |
6494 | { |
6495 | TCGv_ptr fpstatus = get_fpstatus_ptr(1); | |
6496 | gen_helper_recpe_u32(tmp, tmp, fpstatus); | |
6497 | tcg_temp_free_ptr(fpstatus); | |
9ee6e8bb | 6498 | break; |
b6d4443a | 6499 | } |
600b828c | 6500 | case NEON_2RM_VRSQRTE: |
c2fb418e AB |
6501 | { |
6502 | TCGv_ptr fpstatus = get_fpstatus_ptr(1); | |
6503 | gen_helper_rsqrte_u32(tmp, tmp, fpstatus); | |
6504 | tcg_temp_free_ptr(fpstatus); | |
9ee6e8bb | 6505 | break; |
c2fb418e | 6506 | } |
600b828c | 6507 | case NEON_2RM_VRECPE_F: |
b6d4443a AB |
6508 | { |
6509 | TCGv_ptr fpstatus = get_fpstatus_ptr(1); | |
9a011fec | 6510 | gen_helper_recpe_f32(tmp, tmp, fpstatus); |
b6d4443a | 6511 | tcg_temp_free_ptr(fpstatus); |
9ee6e8bb | 6512 | break; |
b6d4443a | 6513 | } |
600b828c | 6514 | case NEON_2RM_VRSQRTE_F: |
c2fb418e AB |
6515 | { |
6516 | TCGv_ptr fpstatus = get_fpstatus_ptr(1); | |
9a011fec | 6517 | gen_helper_rsqrte_f32(tmp, tmp, fpstatus); |
c2fb418e | 6518 | tcg_temp_free_ptr(fpstatus); |
9ee6e8bb | 6519 | break; |
c2fb418e | 6520 | } |
600b828c | 6521 | case NEON_2RM_VCVT_FS: /* VCVT.F32.S32 */ |
60737ed5 PM |
6522 | { |
6523 | TCGv_ptr fpstatus = get_fpstatus_ptr(1); | |
6524 | gen_helper_vfp_sitos(tmp, tmp, fpstatus); | |
6525 | tcg_temp_free_ptr(fpstatus); | |
9ee6e8bb | 6526 | break; |
60737ed5 | 6527 | } |
600b828c | 6528 | case NEON_2RM_VCVT_FU: /* VCVT.F32.U32 */ |
60737ed5 PM |
6529 | { |
6530 | TCGv_ptr fpstatus = get_fpstatus_ptr(1); | |
6531 | gen_helper_vfp_uitos(tmp, tmp, fpstatus); | |
6532 | tcg_temp_free_ptr(fpstatus); | |
9ee6e8bb | 6533 | break; |
60737ed5 | 6534 | } |
600b828c | 6535 | case NEON_2RM_VCVT_SF: /* VCVT.S32.F32 */ |
60737ed5 PM |
6536 | { |
6537 | TCGv_ptr fpstatus = get_fpstatus_ptr(1); | |
6538 | gen_helper_vfp_tosizs(tmp, tmp, fpstatus); | |
6539 | tcg_temp_free_ptr(fpstatus); | |
9ee6e8bb | 6540 | break; |
60737ed5 | 6541 | } |
600b828c | 6542 | case NEON_2RM_VCVT_UF: /* VCVT.U32.F32 */ |
60737ed5 PM |
6543 | { |
6544 | TCGv_ptr fpstatus = get_fpstatus_ptr(1); | |
6545 | gen_helper_vfp_touizs(tmp, tmp, fpstatus); | |
6546 | tcg_temp_free_ptr(fpstatus); | |
9ee6e8bb | 6547 | break; |
60737ed5 | 6548 | } |
9ee6e8bb | 6549 | default: |
600b828c PM |
6550 | /* Reserved op values were caught by the |
6551 | * neon_2rm_sizes[] check earlier. | |
6552 | */ | |
6553 | abort(); | |
9ee6e8bb | 6554 | } |
60737ed5 | 6555 | neon_store_reg(rd, pass, tmp); |
9ee6e8bb PB |
6556 | } |
6557 | break; | |
6558 | } | |
6559 | } else if ((insn & (1 << 10)) == 0) { | |
6560 | /* VTBL, VTBX. */ | |
56907d77 PM |
6561 | int n = ((insn >> 8) & 3) + 1; |
6562 | if ((rn + n) > 32) { | |
6563 | /* This is UNPREDICTABLE; we choose to UNDEF to avoid the | |
6564 | * helper function running off the end of the register file. | |
6565 | */ | |
6566 | return 1; | |
6567 | } | |
6568 | n <<= 3; | |
9ee6e8bb | 6569 | if (insn & (1 << 6)) { |
8f8e3aa4 | 6570 | tmp = neon_load_reg(rd, 0); |
9ee6e8bb | 6571 | } else { |
7d1b0095 | 6572 | tmp = tcg_temp_new_i32(); |
8f8e3aa4 | 6573 | tcg_gen_movi_i32(tmp, 0); |
9ee6e8bb | 6574 | } |
8f8e3aa4 | 6575 | tmp2 = neon_load_reg(rm, 0); |
e7c06c4e | 6576 | ptr1 = vfp_reg_ptr(true, rn); |
b75263d6 | 6577 | tmp5 = tcg_const_i32(n); |
e7c06c4e | 6578 | gen_helper_neon_tbl(tmp2, tmp2, tmp, ptr1, tmp5); |
7d1b0095 | 6579 | tcg_temp_free_i32(tmp); |
9ee6e8bb | 6580 | if (insn & (1 << 6)) { |
8f8e3aa4 | 6581 | tmp = neon_load_reg(rd, 1); |
9ee6e8bb | 6582 | } else { |
7d1b0095 | 6583 | tmp = tcg_temp_new_i32(); |
8f8e3aa4 | 6584 | tcg_gen_movi_i32(tmp, 0); |
9ee6e8bb | 6585 | } |
8f8e3aa4 | 6586 | tmp3 = neon_load_reg(rm, 1); |
e7c06c4e | 6587 | gen_helper_neon_tbl(tmp3, tmp3, tmp, ptr1, tmp5); |
25aeb69b | 6588 | tcg_temp_free_i32(tmp5); |
e7c06c4e | 6589 | tcg_temp_free_ptr(ptr1); |
8f8e3aa4 | 6590 | neon_store_reg(rd, 0, tmp2); |
3018f259 | 6591 | neon_store_reg(rd, 1, tmp3); |
7d1b0095 | 6592 | tcg_temp_free_i32(tmp); |
9ee6e8bb PB |
6593 | } else if ((insn & 0x380) == 0) { |
6594 | /* VDUP */ | |
32f91fb7 | 6595 | int element; |
14776ab5 | 6596 | MemOp size; |
32f91fb7 | 6597 | |
133da6aa JR |
6598 | if ((insn & (7 << 16)) == 0 || (q && (rd & 1))) { |
6599 | return 1; | |
6600 | } | |
9ee6e8bb | 6601 | if (insn & (1 << 16)) { |
32f91fb7 RH |
6602 | size = MO_8; |
6603 | element = (insn >> 17) & 7; | |
9ee6e8bb | 6604 | } else if (insn & (1 << 17)) { |
32f91fb7 RH |
6605 | size = MO_16; |
6606 | element = (insn >> 18) & 3; | |
6607 | } else { | |
6608 | size = MO_32; | |
6609 | element = (insn >> 19) & 1; | |
9ee6e8bb | 6610 | } |
32f91fb7 RH |
6611 | tcg_gen_gvec_dup_mem(size, neon_reg_offset(rd, 0), |
6612 | neon_element_offset(rm, element, size), | |
6613 | q ? 16 : 8, q ? 16 : 8); | |
9ee6e8bb PB |
6614 | } else { |
6615 | return 1; | |
6616 | } | |
6617 | } | |
6618 | } | |
6619 | return 0; | |
6620 | } | |
6621 | ||
8b7209fa RH |
6622 | /* Advanced SIMD three registers of the same length extension. |
6623 | * 31 25 23 22 20 16 12 11 10 9 8 3 0 | |
6624 | * +---------------+-----+---+-----+----+----+---+----+---+----+---------+----+ | |
6625 | * | 1 1 1 1 1 1 0 | op1 | D | op2 | Vn | Vd | 1 | o3 | 0 | o4 | N Q M U | Vm | | |
6626 | * +---------------+-----+---+-----+----+----+---+----+---+----+---------+----+ | |
6627 | */ | |
6628 | static int disas_neon_insn_3same_ext(DisasContext *s, uint32_t insn) | |
6629 | { | |
26c470a7 RH |
6630 | gen_helper_gvec_3 *fn_gvec = NULL; |
6631 | gen_helper_gvec_3_ptr *fn_gvec_ptr = NULL; | |
6632 | int rd, rn, rm, opr_sz; | |
6633 | int data = 0; | |
87732318 RH |
6634 | int off_rn, off_rm; |
6635 | bool is_long = false, q = extract32(insn, 6, 1); | |
6636 | bool ptr_is_env = false; | |
8b7209fa RH |
6637 | |
6638 | if ((insn & 0xfe200f10) == 0xfc200800) { | |
6639 | /* VCMLA -- 1111 110R R.1S .... .... 1000 ...0 .... */ | |
26c470a7 RH |
6640 | int size = extract32(insn, 20, 1); |
6641 | data = extract32(insn, 23, 2); /* rot */ | |
962fcbf2 | 6642 | if (!dc_isar_feature(aa32_vcma, s) |
5763190f | 6643 | || (!size && !dc_isar_feature(aa32_fp16_arith, s))) { |
8b7209fa RH |
6644 | return 1; |
6645 | } | |
6646 | fn_gvec_ptr = size ? gen_helper_gvec_fcmlas : gen_helper_gvec_fcmlah; | |
6647 | } else if ((insn & 0xfea00f10) == 0xfc800800) { | |
6648 | /* VCADD -- 1111 110R 1.0S .... .... 1000 ...0 .... */ | |
26c470a7 RH |
6649 | int size = extract32(insn, 20, 1); |
6650 | data = extract32(insn, 24, 1); /* rot */ | |
962fcbf2 | 6651 | if (!dc_isar_feature(aa32_vcma, s) |
5763190f | 6652 | || (!size && !dc_isar_feature(aa32_fp16_arith, s))) { |
8b7209fa RH |
6653 | return 1; |
6654 | } | |
6655 | fn_gvec_ptr = size ? gen_helper_gvec_fcadds : gen_helper_gvec_fcaddh; | |
26c470a7 RH |
6656 | } else if ((insn & 0xfeb00f00) == 0xfc200d00) { |
6657 | /* V[US]DOT -- 1111 1100 0.10 .... .... 1101 .Q.U .... */ | |
6658 | bool u = extract32(insn, 4, 1); | |
962fcbf2 | 6659 | if (!dc_isar_feature(aa32_dp, s)) { |
26c470a7 RH |
6660 | return 1; |
6661 | } | |
6662 | fn_gvec = u ? gen_helper_gvec_udot_b : gen_helper_gvec_sdot_b; | |
87732318 RH |
6663 | } else if ((insn & 0xff300f10) == 0xfc200810) { |
6664 | /* VFM[AS]L -- 1111 1100 S.10 .... .... 1000 .Q.1 .... */ | |
6665 | int is_s = extract32(insn, 23, 1); | |
6666 | if (!dc_isar_feature(aa32_fhm, s)) { | |
6667 | return 1; | |
6668 | } | |
6669 | is_long = true; | |
6670 | data = is_s; /* is_2 == 0 */ | |
6671 | fn_gvec_ptr = gen_helper_gvec_fmlal_a32; | |
6672 | ptr_is_env = true; | |
8b7209fa RH |
6673 | } else { |
6674 | return 1; | |
6675 | } | |
6676 | ||
87732318 RH |
6677 | VFP_DREG_D(rd, insn); |
6678 | if (rd & q) { | |
6679 | return 1; | |
6680 | } | |
6681 | if (q || !is_long) { | |
6682 | VFP_DREG_N(rn, insn); | |
6683 | VFP_DREG_M(rm, insn); | |
6684 | if ((rn | rm) & q & !is_long) { | |
6685 | return 1; | |
6686 | } | |
6687 | off_rn = vfp_reg_offset(1, rn); | |
6688 | off_rm = vfp_reg_offset(1, rm); | |
6689 | } else { | |
6690 | rn = VFP_SREG_N(insn); | |
6691 | rm = VFP_SREG_M(insn); | |
6692 | off_rn = vfp_reg_offset(0, rn); | |
6693 | off_rm = vfp_reg_offset(0, rm); | |
6694 | } | |
6695 | ||
8b7209fa | 6696 | if (s->fp_excp_el) { |
a767fac8 | 6697 | gen_exception_insn(s, s->pc_curr, EXCP_UDEF, |
4be42f40 | 6698 | syn_simd_access_trap(1, 0xe, false), s->fp_excp_el); |
8b7209fa RH |
6699 | return 0; |
6700 | } | |
6701 | if (!s->vfp_enabled) { | |
6702 | return 1; | |
6703 | } | |
6704 | ||
6705 | opr_sz = (1 + q) * 8; | |
26c470a7 | 6706 | if (fn_gvec_ptr) { |
87732318 RH |
6707 | TCGv_ptr ptr; |
6708 | if (ptr_is_env) { | |
6709 | ptr = cpu_env; | |
6710 | } else { | |
6711 | ptr = get_fpstatus_ptr(1); | |
6712 | } | |
6713 | tcg_gen_gvec_3_ptr(vfp_reg_offset(1, rd), off_rn, off_rm, ptr, | |
26c470a7 | 6714 | opr_sz, opr_sz, data, fn_gvec_ptr); |
87732318 RH |
6715 | if (!ptr_is_env) { |
6716 | tcg_temp_free_ptr(ptr); | |
6717 | } | |
26c470a7 | 6718 | } else { |
87732318 | 6719 | tcg_gen_gvec_3_ool(vfp_reg_offset(1, rd), off_rn, off_rm, |
26c470a7 RH |
6720 | opr_sz, opr_sz, data, fn_gvec); |
6721 | } | |
8b7209fa RH |
6722 | return 0; |
6723 | } | |
6724 | ||
638808ff RH |
6725 | /* Advanced SIMD two registers and a scalar extension. |
6726 | * 31 24 23 22 20 16 12 11 10 9 8 3 0 | |
6727 | * +-----------------+----+---+----+----+----+---+----+---+----+---------+----+ | |
6728 | * | 1 1 1 1 1 1 1 0 | o1 | D | o2 | Vn | Vd | 1 | o3 | 0 | o4 | N Q M U | Vm | | |
6729 | * +-----------------+----+---+----+----+----+---+----+---+----+---------+----+ | |
6730 | * | |
6731 | */ | |
6732 | ||
6733 | static int disas_neon_insn_2reg_scalar_ext(DisasContext *s, uint32_t insn) | |
6734 | { | |
26c470a7 RH |
6735 | gen_helper_gvec_3 *fn_gvec = NULL; |
6736 | gen_helper_gvec_3_ptr *fn_gvec_ptr = NULL; | |
2cc99919 | 6737 | int rd, rn, rm, opr_sz, data; |
87732318 RH |
6738 | int off_rn, off_rm; |
6739 | bool is_long = false, q = extract32(insn, 6, 1); | |
6740 | bool ptr_is_env = false; | |
638808ff RH |
6741 | |
6742 | if ((insn & 0xff000f10) == 0xfe000800) { | |
6743 | /* VCMLA (indexed) -- 1111 1110 S.RR .... .... 1000 ...0 .... */ | |
2cc99919 RH |
6744 | int rot = extract32(insn, 20, 2); |
6745 | int size = extract32(insn, 23, 1); | |
6746 | int index; | |
6747 | ||
962fcbf2 | 6748 | if (!dc_isar_feature(aa32_vcma, s)) { |
638808ff RH |
6749 | return 1; |
6750 | } | |
2cc99919 | 6751 | if (size == 0) { |
5763190f | 6752 | if (!dc_isar_feature(aa32_fp16_arith, s)) { |
2cc99919 RH |
6753 | return 1; |
6754 | } | |
6755 | /* For fp16, rm is just Vm, and index is M. */ | |
6756 | rm = extract32(insn, 0, 4); | |
6757 | index = extract32(insn, 5, 1); | |
6758 | } else { | |
6759 | /* For fp32, rm is the usual M:Vm, and index is 0. */ | |
6760 | VFP_DREG_M(rm, insn); | |
6761 | index = 0; | |
6762 | } | |
6763 | data = (index << 2) | rot; | |
6764 | fn_gvec_ptr = (size ? gen_helper_gvec_fcmlas_idx | |
6765 | : gen_helper_gvec_fcmlah_idx); | |
26c470a7 RH |
6766 | } else if ((insn & 0xffb00f00) == 0xfe200d00) { |
6767 | /* V[US]DOT -- 1111 1110 0.10 .... .... 1101 .Q.U .... */ | |
6768 | int u = extract32(insn, 4, 1); | |
87732318 | 6769 | |
962fcbf2 | 6770 | if (!dc_isar_feature(aa32_dp, s)) { |
26c470a7 RH |
6771 | return 1; |
6772 | } | |
6773 | fn_gvec = u ? gen_helper_gvec_udot_idx_b : gen_helper_gvec_sdot_idx_b; | |
6774 | /* rm is just Vm, and index is M. */ | |
6775 | data = extract32(insn, 5, 1); /* index */ | |
6776 | rm = extract32(insn, 0, 4); | |
87732318 RH |
6777 | } else if ((insn & 0xffa00f10) == 0xfe000810) { |
6778 | /* VFM[AS]L -- 1111 1110 0.0S .... .... 1000 .Q.1 .... */ | |
6779 | int is_s = extract32(insn, 20, 1); | |
6780 | int vm20 = extract32(insn, 0, 3); | |
6781 | int vm3 = extract32(insn, 3, 1); | |
6782 | int m = extract32(insn, 5, 1); | |
6783 | int index; | |
6784 | ||
6785 | if (!dc_isar_feature(aa32_fhm, s)) { | |
6786 | return 1; | |
6787 | } | |
6788 | if (q) { | |
6789 | rm = vm20; | |
6790 | index = m * 2 + vm3; | |
6791 | } else { | |
6792 | rm = vm20 * 2 + m; | |
6793 | index = vm3; | |
6794 | } | |
6795 | is_long = true; | |
6796 | data = (index << 2) | is_s; /* is_2 == 0 */ | |
6797 | fn_gvec_ptr = gen_helper_gvec_fmlal_idx_a32; | |
6798 | ptr_is_env = true; | |
638808ff RH |
6799 | } else { |
6800 | return 1; | |
6801 | } | |
6802 | ||
87732318 RH |
6803 | VFP_DREG_D(rd, insn); |
6804 | if (rd & q) { | |
6805 | return 1; | |
6806 | } | |
6807 | if (q || !is_long) { | |
6808 | VFP_DREG_N(rn, insn); | |
6809 | if (rn & q & !is_long) { | |
6810 | return 1; | |
6811 | } | |
6812 | off_rn = vfp_reg_offset(1, rn); | |
6813 | off_rm = vfp_reg_offset(1, rm); | |
6814 | } else { | |
6815 | rn = VFP_SREG_N(insn); | |
6816 | off_rn = vfp_reg_offset(0, rn); | |
6817 | off_rm = vfp_reg_offset(0, rm); | |
6818 | } | |
638808ff | 6819 | if (s->fp_excp_el) { |
a767fac8 | 6820 | gen_exception_insn(s, s->pc_curr, EXCP_UDEF, |
4be42f40 | 6821 | syn_simd_access_trap(1, 0xe, false), s->fp_excp_el); |
638808ff RH |
6822 | return 0; |
6823 | } | |
6824 | if (!s->vfp_enabled) { | |
6825 | return 1; | |
6826 | } | |
6827 | ||
6828 | opr_sz = (1 + q) * 8; | |
26c470a7 | 6829 | if (fn_gvec_ptr) { |
87732318 RH |
6830 | TCGv_ptr ptr; |
6831 | if (ptr_is_env) { | |
6832 | ptr = cpu_env; | |
6833 | } else { | |
6834 | ptr = get_fpstatus_ptr(1); | |
6835 | } | |
6836 | tcg_gen_gvec_3_ptr(vfp_reg_offset(1, rd), off_rn, off_rm, ptr, | |
26c470a7 | 6837 | opr_sz, opr_sz, data, fn_gvec_ptr); |
87732318 RH |
6838 | if (!ptr_is_env) { |
6839 | tcg_temp_free_ptr(ptr); | |
6840 | } | |
26c470a7 | 6841 | } else { |
87732318 | 6842 | tcg_gen_gvec_3_ool(vfp_reg_offset(1, rd), off_rn, off_rm, |
26c470a7 RH |
6843 | opr_sz, opr_sz, data, fn_gvec); |
6844 | } | |
638808ff RH |
6845 | return 0; |
6846 | } | |
6847 | ||
7dcc1f89 | 6848 | static int disas_coproc_insn(DisasContext *s, uint32_t insn) |
9ee6e8bb | 6849 | { |
4b6a83fb PM |
6850 | int cpnum, is64, crn, crm, opc1, opc2, isread, rt, rt2; |
6851 | const ARMCPRegInfo *ri; | |
9ee6e8bb PB |
6852 | |
6853 | cpnum = (insn >> 8) & 0xf; | |
c0f4af17 PM |
6854 | |
6855 | /* First check for coprocessor space used for XScale/iwMMXt insns */ | |
d614a513 | 6856 | if (arm_dc_feature(s, ARM_FEATURE_XSCALE) && (cpnum < 2)) { |
c0f4af17 PM |
6857 | if (extract32(s->c15_cpar, cpnum, 1) == 0) { |
6858 | return 1; | |
6859 | } | |
d614a513 | 6860 | if (arm_dc_feature(s, ARM_FEATURE_IWMMXT)) { |
7dcc1f89 | 6861 | return disas_iwmmxt_insn(s, insn); |
d614a513 | 6862 | } else if (arm_dc_feature(s, ARM_FEATURE_XSCALE)) { |
7dcc1f89 | 6863 | return disas_dsp_insn(s, insn); |
c0f4af17 PM |
6864 | } |
6865 | return 1; | |
4b6a83fb PM |
6866 | } |
6867 | ||
6868 | /* Otherwise treat as a generic register access */ | |
6869 | is64 = (insn & (1 << 25)) == 0; | |
6870 | if (!is64 && ((insn & (1 << 4)) == 0)) { | |
6871 | /* cdp */ | |
6872 | return 1; | |
6873 | } | |
6874 | ||
6875 | crm = insn & 0xf; | |
6876 | if (is64) { | |
6877 | crn = 0; | |
6878 | opc1 = (insn >> 4) & 0xf; | |
6879 | opc2 = 0; | |
6880 | rt2 = (insn >> 16) & 0xf; | |
6881 | } else { | |
6882 | crn = (insn >> 16) & 0xf; | |
6883 | opc1 = (insn >> 21) & 7; | |
6884 | opc2 = (insn >> 5) & 7; | |
6885 | rt2 = 0; | |
6886 | } | |
6887 | isread = (insn >> 20) & 1; | |
6888 | rt = (insn >> 12) & 0xf; | |
6889 | ||
60322b39 | 6890 | ri = get_arm_cp_reginfo(s->cp_regs, |
51a79b03 | 6891 | ENCODE_CP_REG(cpnum, is64, s->ns, crn, crm, opc1, opc2)); |
4b6a83fb PM |
6892 | if (ri) { |
6893 | /* Check access permissions */ | |
dcbff19b | 6894 | if (!cp_access_ok(s->current_el, ri, isread)) { |
4b6a83fb PM |
6895 | return 1; |
6896 | } | |
6897 | ||
c0f4af17 | 6898 | if (ri->accessfn || |
d614a513 | 6899 | (arm_dc_feature(s, ARM_FEATURE_XSCALE) && cpnum < 14)) { |
f59df3f2 PM |
6900 | /* Emit code to perform further access permissions checks at |
6901 | * runtime; this may result in an exception. | |
c0f4af17 PM |
6902 | * Note that on XScale all cp0..c13 registers do an access check |
6903 | * call in order to handle c15_cpar. | |
f59df3f2 PM |
6904 | */ |
6905 | TCGv_ptr tmpptr; | |
3f208fd7 | 6906 | TCGv_i32 tcg_syn, tcg_isread; |
8bcbf37c PM |
6907 | uint32_t syndrome; |
6908 | ||
6909 | /* Note that since we are an implementation which takes an | |
6910 | * exception on a trapped conditional instruction only if the | |
6911 | * instruction passes its condition code check, we can take | |
6912 | * advantage of the clause in the ARM ARM that allows us to set | |
6913 | * the COND field in the instruction to 0xE in all cases. | |
6914 | * We could fish the actual condition out of the insn (ARM) | |
6915 | * or the condexec bits (Thumb) but it isn't necessary. | |
6916 | */ | |
6917 | switch (cpnum) { | |
6918 | case 14: | |
6919 | if (is64) { | |
6920 | syndrome = syn_cp14_rrt_trap(1, 0xe, opc1, crm, rt, rt2, | |
4df32259 | 6921 | isread, false); |
8bcbf37c PM |
6922 | } else { |
6923 | syndrome = syn_cp14_rt_trap(1, 0xe, opc1, opc2, crn, crm, | |
4df32259 | 6924 | rt, isread, false); |
8bcbf37c PM |
6925 | } |
6926 | break; | |
6927 | case 15: | |
6928 | if (is64) { | |
6929 | syndrome = syn_cp15_rrt_trap(1, 0xe, opc1, crm, rt, rt2, | |
4df32259 | 6930 | isread, false); |
8bcbf37c PM |
6931 | } else { |
6932 | syndrome = syn_cp15_rt_trap(1, 0xe, opc1, opc2, crn, crm, | |
4df32259 | 6933 | rt, isread, false); |
8bcbf37c PM |
6934 | } |
6935 | break; | |
6936 | default: | |
6937 | /* ARMv8 defines that only coprocessors 14 and 15 exist, | |
6938 | * so this can only happen if this is an ARMv7 or earlier CPU, | |
6939 | * in which case the syndrome information won't actually be | |
6940 | * guest visible. | |
6941 | */ | |
d614a513 | 6942 | assert(!arm_dc_feature(s, ARM_FEATURE_V8)); |
8bcbf37c PM |
6943 | syndrome = syn_uncategorized(); |
6944 | break; | |
6945 | } | |
6946 | ||
43bfa4a1 | 6947 | gen_set_condexec(s); |
43722a6d | 6948 | gen_set_pc_im(s, s->pc_curr); |
f59df3f2 | 6949 | tmpptr = tcg_const_ptr(ri); |
8bcbf37c | 6950 | tcg_syn = tcg_const_i32(syndrome); |
3f208fd7 PM |
6951 | tcg_isread = tcg_const_i32(isread); |
6952 | gen_helper_access_check_cp_reg(cpu_env, tmpptr, tcg_syn, | |
6953 | tcg_isread); | |
f59df3f2 | 6954 | tcg_temp_free_ptr(tmpptr); |
8bcbf37c | 6955 | tcg_temp_free_i32(tcg_syn); |
3f208fd7 | 6956 | tcg_temp_free_i32(tcg_isread); |
37ff584c PM |
6957 | } else if (ri->type & ARM_CP_RAISES_EXC) { |
6958 | /* | |
6959 | * The readfn or writefn might raise an exception; | |
6960 | * synchronize the CPU state in case it does. | |
6961 | */ | |
6962 | gen_set_condexec(s); | |
6963 | gen_set_pc_im(s, s->pc_curr); | |
f59df3f2 PM |
6964 | } |
6965 | ||
4b6a83fb PM |
6966 | /* Handle special cases first */ |
6967 | switch (ri->type & ~(ARM_CP_FLAG_MASK & ~ARM_CP_SPECIAL)) { | |
6968 | case ARM_CP_NOP: | |
6969 | return 0; | |
6970 | case ARM_CP_WFI: | |
6971 | if (isread) { | |
6972 | return 1; | |
6973 | } | |
a0415916 | 6974 | gen_set_pc_im(s, s->base.pc_next); |
dcba3a8d | 6975 | s->base.is_jmp = DISAS_WFI; |
2bee5105 | 6976 | return 0; |
4b6a83fb PM |
6977 | default: |
6978 | break; | |
6979 | } | |
6980 | ||
c5a49c63 | 6981 | if ((tb_cflags(s->base.tb) & CF_USE_ICOUNT) && (ri->type & ARM_CP_IO)) { |
2452731c PM |
6982 | gen_io_start(); |
6983 | } | |
6984 | ||
4b6a83fb PM |
6985 | if (isread) { |
6986 | /* Read */ | |
6987 | if (is64) { | |
6988 | TCGv_i64 tmp64; | |
6989 | TCGv_i32 tmp; | |
6990 | if (ri->type & ARM_CP_CONST) { | |
6991 | tmp64 = tcg_const_i64(ri->resetvalue); | |
6992 | } else if (ri->readfn) { | |
6993 | TCGv_ptr tmpptr; | |
4b6a83fb PM |
6994 | tmp64 = tcg_temp_new_i64(); |
6995 | tmpptr = tcg_const_ptr(ri); | |
6996 | gen_helper_get_cp_reg64(tmp64, cpu_env, tmpptr); | |
6997 | tcg_temp_free_ptr(tmpptr); | |
6998 | } else { | |
6999 | tmp64 = tcg_temp_new_i64(); | |
7000 | tcg_gen_ld_i64(tmp64, cpu_env, ri->fieldoffset); | |
7001 | } | |
7002 | tmp = tcg_temp_new_i32(); | |
ecc7b3aa | 7003 | tcg_gen_extrl_i64_i32(tmp, tmp64); |
4b6a83fb | 7004 | store_reg(s, rt, tmp); |
ed336850 | 7005 | tmp = tcg_temp_new_i32(); |
664b7e3b | 7006 | tcg_gen_extrh_i64_i32(tmp, tmp64); |
ed336850 | 7007 | tcg_temp_free_i64(tmp64); |
4b6a83fb PM |
7008 | store_reg(s, rt2, tmp); |
7009 | } else { | |
39d5492a | 7010 | TCGv_i32 tmp; |
4b6a83fb PM |
7011 | if (ri->type & ARM_CP_CONST) { |
7012 | tmp = tcg_const_i32(ri->resetvalue); | |
7013 | } else if (ri->readfn) { | |
7014 | TCGv_ptr tmpptr; | |
4b6a83fb PM |
7015 | tmp = tcg_temp_new_i32(); |
7016 | tmpptr = tcg_const_ptr(ri); | |
7017 | gen_helper_get_cp_reg(tmp, cpu_env, tmpptr); | |
7018 | tcg_temp_free_ptr(tmpptr); | |
7019 | } else { | |
7020 | tmp = load_cpu_offset(ri->fieldoffset); | |
7021 | } | |
7022 | if (rt == 15) { | |
7023 | /* Destination register of r15 for 32 bit loads sets | |
7024 | * the condition codes from the high 4 bits of the value | |
7025 | */ | |
7026 | gen_set_nzcv(tmp); | |
7027 | tcg_temp_free_i32(tmp); | |
7028 | } else { | |
7029 | store_reg(s, rt, tmp); | |
7030 | } | |
7031 | } | |
7032 | } else { | |
7033 | /* Write */ | |
7034 | if (ri->type & ARM_CP_CONST) { | |
7035 | /* If not forbidden by access permissions, treat as WI */ | |
7036 | return 0; | |
7037 | } | |
7038 | ||
7039 | if (is64) { | |
39d5492a | 7040 | TCGv_i32 tmplo, tmphi; |
4b6a83fb PM |
7041 | TCGv_i64 tmp64 = tcg_temp_new_i64(); |
7042 | tmplo = load_reg(s, rt); | |
7043 | tmphi = load_reg(s, rt2); | |
7044 | tcg_gen_concat_i32_i64(tmp64, tmplo, tmphi); | |
7045 | tcg_temp_free_i32(tmplo); | |
7046 | tcg_temp_free_i32(tmphi); | |
7047 | if (ri->writefn) { | |
7048 | TCGv_ptr tmpptr = tcg_const_ptr(ri); | |
4b6a83fb PM |
7049 | gen_helper_set_cp_reg64(cpu_env, tmpptr, tmp64); |
7050 | tcg_temp_free_ptr(tmpptr); | |
7051 | } else { | |
7052 | tcg_gen_st_i64(tmp64, cpu_env, ri->fieldoffset); | |
7053 | } | |
7054 | tcg_temp_free_i64(tmp64); | |
7055 | } else { | |
7056 | if (ri->writefn) { | |
39d5492a | 7057 | TCGv_i32 tmp; |
4b6a83fb | 7058 | TCGv_ptr tmpptr; |
4b6a83fb PM |
7059 | tmp = load_reg(s, rt); |
7060 | tmpptr = tcg_const_ptr(ri); | |
7061 | gen_helper_set_cp_reg(cpu_env, tmpptr, tmp); | |
7062 | tcg_temp_free_ptr(tmpptr); | |
7063 | tcg_temp_free_i32(tmp); | |
7064 | } else { | |
39d5492a | 7065 | TCGv_i32 tmp = load_reg(s, rt); |
4b6a83fb PM |
7066 | store_cpu_offset(tmp, ri->fieldoffset); |
7067 | } | |
7068 | } | |
2452731c PM |
7069 | } |
7070 | ||
c5a49c63 | 7071 | if ((tb_cflags(s->base.tb) & CF_USE_ICOUNT) && (ri->type & ARM_CP_IO)) { |
2452731c | 7072 | /* I/O operations must end the TB here (whether read or write) */ |
2452731c PM |
7073 | gen_lookup_tb(s); |
7074 | } else if (!isread && !(ri->type & ARM_CP_SUPPRESS_TB_END)) { | |
4b6a83fb PM |
7075 | /* We default to ending the TB on a coprocessor register write, |
7076 | * but allow this to be suppressed by the register definition | |
7077 | * (usually only necessary to work around guest bugs). | |
7078 | */ | |
2452731c | 7079 | gen_lookup_tb(s); |
4b6a83fb | 7080 | } |
2452731c | 7081 | |
4b6a83fb PM |
7082 | return 0; |
7083 | } | |
7084 | ||
626187d8 PM |
7085 | /* Unknown register; this might be a guest error or a QEMU |
7086 | * unimplemented feature. | |
7087 | */ | |
7088 | if (is64) { | |
7089 | qemu_log_mask(LOG_UNIMP, "%s access to unsupported AArch32 " | |
51a79b03 PM |
7090 | "64 bit system register cp:%d opc1: %d crm:%d " |
7091 | "(%s)\n", | |
7092 | isread ? "read" : "write", cpnum, opc1, crm, | |
7093 | s->ns ? "non-secure" : "secure"); | |
626187d8 PM |
7094 | } else { |
7095 | qemu_log_mask(LOG_UNIMP, "%s access to unsupported AArch32 " | |
51a79b03 PM |
7096 | "system register cp:%d opc1:%d crn:%d crm:%d opc2:%d " |
7097 | "(%s)\n", | |
7098 | isread ? "read" : "write", cpnum, opc1, crn, crm, opc2, | |
7099 | s->ns ? "non-secure" : "secure"); | |
626187d8 PM |
7100 | } |
7101 | ||
4a9a539f | 7102 | return 1; |
9ee6e8bb PB |
7103 | } |
7104 | ||
5e3f878a PB |
7105 | |
7106 | /* Store a 64-bit value to a register pair. Clobbers val. */ | |
a7812ae4 | 7107 | static void gen_storeq_reg(DisasContext *s, int rlow, int rhigh, TCGv_i64 val) |
5e3f878a | 7108 | { |
39d5492a | 7109 | TCGv_i32 tmp; |
7d1b0095 | 7110 | tmp = tcg_temp_new_i32(); |
ecc7b3aa | 7111 | tcg_gen_extrl_i64_i32(tmp, val); |
5e3f878a | 7112 | store_reg(s, rlow, tmp); |
7d1b0095 | 7113 | tmp = tcg_temp_new_i32(); |
664b7e3b | 7114 | tcg_gen_extrh_i64_i32(tmp, val); |
5e3f878a PB |
7115 | store_reg(s, rhigh, tmp); |
7116 | } | |
7117 | ||
5e3f878a | 7118 | /* load and add a 64-bit value from a register pair. */ |
a7812ae4 | 7119 | static void gen_addq(DisasContext *s, TCGv_i64 val, int rlow, int rhigh) |
5e3f878a | 7120 | { |
a7812ae4 | 7121 | TCGv_i64 tmp; |
39d5492a PM |
7122 | TCGv_i32 tmpl; |
7123 | TCGv_i32 tmph; | |
5e3f878a PB |
7124 | |
7125 | /* Load 64-bit value rd:rn. */ | |
36aa55dc PB |
7126 | tmpl = load_reg(s, rlow); |
7127 | tmph = load_reg(s, rhigh); | |
a7812ae4 | 7128 | tmp = tcg_temp_new_i64(); |
36aa55dc | 7129 | tcg_gen_concat_i32_i64(tmp, tmpl, tmph); |
7d1b0095 PM |
7130 | tcg_temp_free_i32(tmpl); |
7131 | tcg_temp_free_i32(tmph); | |
5e3f878a | 7132 | tcg_gen_add_i64(val, val, tmp); |
b75263d6 | 7133 | tcg_temp_free_i64(tmp); |
5e3f878a PB |
7134 | } |
7135 | ||
c9f10124 | 7136 | /* Set N and Z flags from hi|lo. */ |
39d5492a | 7137 | static void gen_logicq_cc(TCGv_i32 lo, TCGv_i32 hi) |
5e3f878a | 7138 | { |
c9f10124 RH |
7139 | tcg_gen_mov_i32(cpu_NF, hi); |
7140 | tcg_gen_or_i32(cpu_ZF, lo, hi); | |
5e3f878a PB |
7141 | } |
7142 | ||
426f5abc PB |
7143 | /* Load/Store exclusive instructions are implemented by remembering |
7144 | the value/address loaded, and seeing if these are the same | |
354161b3 | 7145 | when the store is performed. This should be sufficient to implement |
426f5abc | 7146 | the architecturally mandated semantics, and avoids having to monitor |
354161b3 EC |
7147 | regular stores. The compare vs the remembered value is done during |
7148 | the cmpxchg operation, but we must compare the addresses manually. */ | |
426f5abc | 7149 | static void gen_load_exclusive(DisasContext *s, int rt, int rt2, |
39d5492a | 7150 | TCGv_i32 addr, int size) |
426f5abc | 7151 | { |
94ee24e7 | 7152 | TCGv_i32 tmp = tcg_temp_new_i32(); |
14776ab5 | 7153 | MemOp opc = size | MO_ALIGN | s->be_data; |
426f5abc | 7154 | |
50225ad0 PM |
7155 | s->is_ldex = true; |
7156 | ||
426f5abc | 7157 | if (size == 3) { |
39d5492a | 7158 | TCGv_i32 tmp2 = tcg_temp_new_i32(); |
354161b3 | 7159 | TCGv_i64 t64 = tcg_temp_new_i64(); |
03d05e2d | 7160 | |
3448d47b PM |
7161 | /* For AArch32, architecturally the 32-bit word at the lowest |
7162 | * address is always Rt and the one at addr+4 is Rt2, even if | |
7163 | * the CPU is big-endian. That means we don't want to do a | |
7164 | * gen_aa32_ld_i64(), which invokes gen_aa32_frob64() as if | |
7165 | * for an architecturally 64-bit access, but instead do a | |
7166 | * 64-bit access using MO_BE if appropriate and then split | |
7167 | * the two halves. | |
7168 | * This only makes a difference for BE32 user-mode, where | |
7169 | * frob64() must not flip the two halves of the 64-bit data | |
7170 | * but this code must treat BE32 user-mode like BE32 system. | |
7171 | */ | |
7172 | TCGv taddr = gen_aa32_addr(s, addr, opc); | |
7173 | ||
7174 | tcg_gen_qemu_ld_i64(t64, taddr, get_mem_index(s), opc); | |
7175 | tcg_temp_free(taddr); | |
354161b3 | 7176 | tcg_gen_mov_i64(cpu_exclusive_val, t64); |
3448d47b PM |
7177 | if (s->be_data == MO_BE) { |
7178 | tcg_gen_extr_i64_i32(tmp2, tmp, t64); | |
7179 | } else { | |
7180 | tcg_gen_extr_i64_i32(tmp, tmp2, t64); | |
7181 | } | |
354161b3 EC |
7182 | tcg_temp_free_i64(t64); |
7183 | ||
7184 | store_reg(s, rt2, tmp2); | |
03d05e2d | 7185 | } else { |
354161b3 | 7186 | gen_aa32_ld_i32(s, tmp, addr, get_mem_index(s), opc); |
03d05e2d | 7187 | tcg_gen_extu_i32_i64(cpu_exclusive_val, tmp); |
426f5abc | 7188 | } |
03d05e2d PM |
7189 | |
7190 | store_reg(s, rt, tmp); | |
7191 | tcg_gen_extu_i32_i64(cpu_exclusive_addr, addr); | |
426f5abc PB |
7192 | } |
7193 | ||
7194 | static void gen_clrex(DisasContext *s) | |
7195 | { | |
03d05e2d | 7196 | tcg_gen_movi_i64(cpu_exclusive_addr, -1); |
426f5abc PB |
7197 | } |
7198 | ||
426f5abc | 7199 | static void gen_store_exclusive(DisasContext *s, int rd, int rt, int rt2, |
39d5492a | 7200 | TCGv_i32 addr, int size) |
426f5abc | 7201 | { |
354161b3 EC |
7202 | TCGv_i32 t0, t1, t2; |
7203 | TCGv_i64 extaddr; | |
7204 | TCGv taddr; | |
42a268c2 RH |
7205 | TCGLabel *done_label; |
7206 | TCGLabel *fail_label; | |
14776ab5 | 7207 | MemOp opc = size | MO_ALIGN | s->be_data; |
426f5abc PB |
7208 | |
7209 | /* if (env->exclusive_addr == addr && env->exclusive_val == [addr]) { | |
7210 | [addr] = {Rt}; | |
7211 | {Rd} = 0; | |
7212 | } else { | |
7213 | {Rd} = 1; | |
7214 | } */ | |
7215 | fail_label = gen_new_label(); | |
7216 | done_label = gen_new_label(); | |
03d05e2d PM |
7217 | extaddr = tcg_temp_new_i64(); |
7218 | tcg_gen_extu_i32_i64(extaddr, addr); | |
7219 | tcg_gen_brcond_i64(TCG_COND_NE, extaddr, cpu_exclusive_addr, fail_label); | |
7220 | tcg_temp_free_i64(extaddr); | |
7221 | ||
354161b3 EC |
7222 | taddr = gen_aa32_addr(s, addr, opc); |
7223 | t0 = tcg_temp_new_i32(); | |
7224 | t1 = load_reg(s, rt); | |
426f5abc | 7225 | if (size == 3) { |
354161b3 EC |
7226 | TCGv_i64 o64 = tcg_temp_new_i64(); |
7227 | TCGv_i64 n64 = tcg_temp_new_i64(); | |
03d05e2d | 7228 | |
354161b3 | 7229 | t2 = load_reg(s, rt2); |
3448d47b PM |
7230 | /* For AArch32, architecturally the 32-bit word at the lowest |
7231 | * address is always Rt and the one at addr+4 is Rt2, even if | |
7232 | * the CPU is big-endian. Since we're going to treat this as a | |
7233 | * single 64-bit BE store, we need to put the two halves in the | |
7234 | * opposite order for BE to LE, so that they end up in the right | |
7235 | * places. | |
7236 | * We don't want gen_aa32_frob64() because that does the wrong | |
7237 | * thing for BE32 usermode. | |
7238 | */ | |
7239 | if (s->be_data == MO_BE) { | |
7240 | tcg_gen_concat_i32_i64(n64, t2, t1); | |
7241 | } else { | |
7242 | tcg_gen_concat_i32_i64(n64, t1, t2); | |
7243 | } | |
354161b3 | 7244 | tcg_temp_free_i32(t2); |
03d05e2d | 7245 | |
354161b3 EC |
7246 | tcg_gen_atomic_cmpxchg_i64(o64, taddr, cpu_exclusive_val, n64, |
7247 | get_mem_index(s), opc); | |
7248 | tcg_temp_free_i64(n64); | |
7249 | ||
354161b3 EC |
7250 | tcg_gen_setcond_i64(TCG_COND_NE, o64, o64, cpu_exclusive_val); |
7251 | tcg_gen_extrl_i64_i32(t0, o64); | |
7252 | ||
7253 | tcg_temp_free_i64(o64); | |
7254 | } else { | |
7255 | t2 = tcg_temp_new_i32(); | |
7256 | tcg_gen_extrl_i64_i32(t2, cpu_exclusive_val); | |
7257 | tcg_gen_atomic_cmpxchg_i32(t0, taddr, t2, t1, get_mem_index(s), opc); | |
7258 | tcg_gen_setcond_i32(TCG_COND_NE, t0, t0, t2); | |
7259 | tcg_temp_free_i32(t2); | |
426f5abc | 7260 | } |
354161b3 EC |
7261 | tcg_temp_free_i32(t1); |
7262 | tcg_temp_free(taddr); | |
7263 | tcg_gen_mov_i32(cpu_R[rd], t0); | |
7264 | tcg_temp_free_i32(t0); | |
426f5abc | 7265 | tcg_gen_br(done_label); |
354161b3 | 7266 | |
426f5abc PB |
7267 | gen_set_label(fail_label); |
7268 | tcg_gen_movi_i32(cpu_R[rd], 1); | |
7269 | gen_set_label(done_label); | |
03d05e2d | 7270 | tcg_gen_movi_i64(cpu_exclusive_addr, -1); |
426f5abc | 7271 | } |
426f5abc | 7272 | |
81465888 PM |
7273 | /* gen_srs: |
7274 | * @env: CPUARMState | |
7275 | * @s: DisasContext | |
7276 | * @mode: mode field from insn (which stack to store to) | |
7277 | * @amode: addressing mode (DA/IA/DB/IB), encoded as per P,U bits in ARM insn | |
7278 | * @writeback: true if writeback bit set | |
7279 | * | |
7280 | * Generate code for the SRS (Store Return State) insn. | |
7281 | */ | |
7282 | static void gen_srs(DisasContext *s, | |
7283 | uint32_t mode, uint32_t amode, bool writeback) | |
7284 | { | |
7285 | int32_t offset; | |
cbc0326b PM |
7286 | TCGv_i32 addr, tmp; |
7287 | bool undef = false; | |
7288 | ||
7289 | /* SRS is: | |
7290 | * - trapped to EL3 if EL3 is AArch64 and we are at Secure EL1 | |
ba63cf47 | 7291 | * and specified mode is monitor mode |
cbc0326b PM |
7292 | * - UNDEFINED in Hyp mode |
7293 | * - UNPREDICTABLE in User or System mode | |
7294 | * - UNPREDICTABLE if the specified mode is: | |
7295 | * -- not implemented | |
7296 | * -- not a valid mode number | |
7297 | * -- a mode that's at a higher exception level | |
7298 | * -- Monitor, if we are Non-secure | |
f01377f5 | 7299 | * For the UNPREDICTABLE cases we choose to UNDEF. |
cbc0326b | 7300 | */ |
ba63cf47 | 7301 | if (s->current_el == 1 && !s->ns && mode == ARM_CPU_MODE_MON) { |
a767fac8 | 7302 | gen_exception_insn(s, s->pc_curr, EXCP_UDEF, syn_uncategorized(), 3); |
cbc0326b PM |
7303 | return; |
7304 | } | |
7305 | ||
7306 | if (s->current_el == 0 || s->current_el == 2) { | |
7307 | undef = true; | |
7308 | } | |
7309 | ||
7310 | switch (mode) { | |
7311 | case ARM_CPU_MODE_USR: | |
7312 | case ARM_CPU_MODE_FIQ: | |
7313 | case ARM_CPU_MODE_IRQ: | |
7314 | case ARM_CPU_MODE_SVC: | |
7315 | case ARM_CPU_MODE_ABT: | |
7316 | case ARM_CPU_MODE_UND: | |
7317 | case ARM_CPU_MODE_SYS: | |
7318 | break; | |
7319 | case ARM_CPU_MODE_HYP: | |
7320 | if (s->current_el == 1 || !arm_dc_feature(s, ARM_FEATURE_EL2)) { | |
7321 | undef = true; | |
7322 | } | |
7323 | break; | |
7324 | case ARM_CPU_MODE_MON: | |
7325 | /* No need to check specifically for "are we non-secure" because | |
7326 | * we've already made EL0 UNDEF and handled the trap for S-EL1; | |
7327 | * so if this isn't EL3 then we must be non-secure. | |
7328 | */ | |
7329 | if (s->current_el != 3) { | |
7330 | undef = true; | |
7331 | } | |
7332 | break; | |
7333 | default: | |
7334 | undef = true; | |
7335 | } | |
7336 | ||
7337 | if (undef) { | |
1ce21ba1 | 7338 | unallocated_encoding(s); |
cbc0326b PM |
7339 | return; |
7340 | } | |
7341 | ||
7342 | addr = tcg_temp_new_i32(); | |
7343 | tmp = tcg_const_i32(mode); | |
f01377f5 PM |
7344 | /* get_r13_banked() will raise an exception if called from System mode */ |
7345 | gen_set_condexec(s); | |
43722a6d | 7346 | gen_set_pc_im(s, s->pc_curr); |
81465888 PM |
7347 | gen_helper_get_r13_banked(addr, cpu_env, tmp); |
7348 | tcg_temp_free_i32(tmp); | |
7349 | switch (amode) { | |
7350 | case 0: /* DA */ | |
7351 | offset = -4; | |
7352 | break; | |
7353 | case 1: /* IA */ | |
7354 | offset = 0; | |
7355 | break; | |
7356 | case 2: /* DB */ | |
7357 | offset = -8; | |
7358 | break; | |
7359 | case 3: /* IB */ | |
7360 | offset = 4; | |
7361 | break; | |
7362 | default: | |
7363 | abort(); | |
7364 | } | |
7365 | tcg_gen_addi_i32(addr, addr, offset); | |
7366 | tmp = load_reg(s, 14); | |
12dcc321 | 7367 | gen_aa32_st32(s, tmp, addr, get_mem_index(s)); |
5a839c0d | 7368 | tcg_temp_free_i32(tmp); |
81465888 PM |
7369 | tmp = load_cpu_field(spsr); |
7370 | tcg_gen_addi_i32(addr, addr, 4); | |
12dcc321 | 7371 | gen_aa32_st32(s, tmp, addr, get_mem_index(s)); |
5a839c0d | 7372 | tcg_temp_free_i32(tmp); |
81465888 PM |
7373 | if (writeback) { |
7374 | switch (amode) { | |
7375 | case 0: | |
7376 | offset = -8; | |
7377 | break; | |
7378 | case 1: | |
7379 | offset = 4; | |
7380 | break; | |
7381 | case 2: | |
7382 | offset = -4; | |
7383 | break; | |
7384 | case 3: | |
7385 | offset = 0; | |
7386 | break; | |
7387 | default: | |
7388 | abort(); | |
7389 | } | |
7390 | tcg_gen_addi_i32(addr, addr, offset); | |
7391 | tmp = tcg_const_i32(mode); | |
7392 | gen_helper_set_r13_banked(cpu_env, tmp, addr); | |
7393 | tcg_temp_free_i32(tmp); | |
7394 | } | |
7395 | tcg_temp_free_i32(addr); | |
dcba3a8d | 7396 | s->base.is_jmp = DISAS_UPDATE; |
81465888 PM |
7397 | } |
7398 | ||
c2d9644e RK |
7399 | /* Generate a label used for skipping this instruction */ |
7400 | static void arm_gen_condlabel(DisasContext *s) | |
7401 | { | |
7402 | if (!s->condjmp) { | |
7403 | s->condlabel = gen_new_label(); | |
7404 | s->condjmp = 1; | |
7405 | } | |
7406 | } | |
7407 | ||
7408 | /* Skip this instruction if the ARM condition is false */ | |
7409 | static void arm_skip_unless(DisasContext *s, uint32_t cond) | |
7410 | { | |
7411 | arm_gen_condlabel(s); | |
7412 | arm_gen_test_cc(cond ^ 1, s->condlabel); | |
7413 | } | |
7414 | ||
581c6ebd RH |
7415 | |
7416 | /* | |
7417 | * Constant expanders for the decoders. | |
7418 | */ | |
7419 | ||
145952e8 RH |
7420 | static int negate(DisasContext *s, int x) |
7421 | { | |
7422 | return -x; | |
7423 | } | |
7424 | ||
20556e7b RH |
7425 | static int plus_2(DisasContext *s, int x) |
7426 | { | |
7427 | return x + 2; | |
7428 | } | |
7429 | ||
581c6ebd RH |
7430 | static int times_2(DisasContext *s, int x) |
7431 | { | |
7432 | return x * 2; | |
7433 | } | |
7434 | ||
5e291fe1 RH |
7435 | static int times_4(DisasContext *s, int x) |
7436 | { | |
7437 | return x * 4; | |
7438 | } | |
7439 | ||
581c6ebd RH |
7440 | /* Return only the rotation part of T32ExpandImm. */ |
7441 | static int t32_expandimm_rot(DisasContext *s, int x) | |
7442 | { | |
7443 | return x & 0xc00 ? extract32(x, 7, 5) : 0; | |
7444 | } | |
7445 | ||
7446 | /* Return the unrotated immediate from T32ExpandImm. */ | |
7447 | static int t32_expandimm_imm(DisasContext *s, int x) | |
7448 | { | |
7449 | int imm = extract32(x, 0, 8); | |
7450 | ||
7451 | switch (extract32(x, 8, 4)) { | |
7452 | case 0: /* XY */ | |
7453 | /* Nothing to do. */ | |
7454 | break; | |
7455 | case 1: /* 00XY00XY */ | |
7456 | imm *= 0x00010001; | |
7457 | break; | |
7458 | case 2: /* XY00XY00 */ | |
7459 | imm *= 0x01000100; | |
7460 | break; | |
7461 | case 3: /* XYXYXYXY */ | |
7462 | imm *= 0x01010101; | |
7463 | break; | |
7464 | default: | |
7465 | /* Rotated constant. */ | |
7466 | imm |= 0x80; | |
7467 | break; | |
7468 | } | |
7469 | return imm; | |
7470 | } | |
7471 | ||
360144f3 RH |
7472 | static int t32_branch24(DisasContext *s, int x) |
7473 | { | |
7474 | /* Convert J1:J2 at x[22:21] to I2:I1, which involves I=J^~S. */ | |
7475 | x ^= !(x < 0) * (3 << 21); | |
7476 | /* Append the final zero. */ | |
7477 | return x << 1; | |
7478 | } | |
7479 | ||
080c4ead RH |
7480 | static int t16_setflags(DisasContext *s) |
7481 | { | |
7482 | return s->condexec_mask == 0; | |
7483 | } | |
7484 | ||
564b125f RH |
7485 | static int t16_push_list(DisasContext *s, int x) |
7486 | { | |
7487 | return (x & 0xff) | (x & 0x100) << (14 - 8); | |
7488 | } | |
7489 | ||
7490 | static int t16_pop_list(DisasContext *s, int x) | |
7491 | { | |
7492 | return (x & 0xff) | (x & 0x100) << (15 - 8); | |
7493 | } | |
7494 | ||
51409b9e RH |
7495 | /* |
7496 | * Include the generated decoders. | |
7497 | */ | |
7498 | ||
7499 | #include "decode-a32.inc.c" | |
7500 | #include "decode-a32-uncond.inc.c" | |
7501 | #include "decode-t32.inc.c" | |
f97b454e | 7502 | #include "decode-t16.inc.c" |
51409b9e | 7503 | |
25ae32c5 RH |
7504 | /* Helpers to swap operands for reverse-subtract. */ |
7505 | static void gen_rsb(TCGv_i32 dst, TCGv_i32 a, TCGv_i32 b) | |
7506 | { | |
7507 | tcg_gen_sub_i32(dst, b, a); | |
7508 | } | |
7509 | ||
7510 | static void gen_rsb_CC(TCGv_i32 dst, TCGv_i32 a, TCGv_i32 b) | |
7511 | { | |
7512 | gen_sub_CC(dst, b, a); | |
7513 | } | |
7514 | ||
7515 | static void gen_rsc(TCGv_i32 dest, TCGv_i32 a, TCGv_i32 b) | |
7516 | { | |
7517 | gen_sub_carry(dest, b, a); | |
7518 | } | |
7519 | ||
7520 | static void gen_rsc_CC(TCGv_i32 dest, TCGv_i32 a, TCGv_i32 b) | |
7521 | { | |
7522 | gen_sbc_CC(dest, b, a); | |
7523 | } | |
7524 | ||
7525 | /* | |
7526 | * Helpers for the data processing routines. | |
7527 | * | |
7528 | * After the computation store the results back. | |
7529 | * This may be suppressed altogether (STREG_NONE), require a runtime | |
7530 | * check against the stack limits (STREG_SP_CHECK), or generate an | |
7531 | * exception return. Oh, or store into a register. | |
7532 | * | |
7533 | * Always return true, indicating success for a trans_* function. | |
7534 | */ | |
7535 | typedef enum { | |
7536 | STREG_NONE, | |
7537 | STREG_NORMAL, | |
7538 | STREG_SP_CHECK, | |
7539 | STREG_EXC_RET, | |
7540 | } StoreRegKind; | |
7541 | ||
7542 | static bool store_reg_kind(DisasContext *s, int rd, | |
7543 | TCGv_i32 val, StoreRegKind kind) | |
7544 | { | |
7545 | switch (kind) { | |
7546 | case STREG_NONE: | |
7547 | tcg_temp_free_i32(val); | |
7548 | return true; | |
7549 | case STREG_NORMAL: | |
7550 | /* See ALUWritePC: Interworking only from a32 mode. */ | |
7551 | if (s->thumb) { | |
7552 | store_reg(s, rd, val); | |
7553 | } else { | |
7554 | store_reg_bx(s, rd, val); | |
7555 | } | |
7556 | return true; | |
7557 | case STREG_SP_CHECK: | |
7558 | store_sp_checked(s, val); | |
7559 | return true; | |
7560 | case STREG_EXC_RET: | |
7561 | gen_exception_return(s, val); | |
7562 | return true; | |
7563 | } | |
7564 | g_assert_not_reached(); | |
7565 | } | |
7566 | ||
7567 | /* | |
7568 | * Data Processing (register) | |
7569 | * | |
7570 | * Operate, with set flags, one register source, | |
7571 | * one immediate shifted register source, and a destination. | |
7572 | */ | |
7573 | static bool op_s_rrr_shi(DisasContext *s, arg_s_rrr_shi *a, | |
7574 | void (*gen)(TCGv_i32, TCGv_i32, TCGv_i32), | |
7575 | int logic_cc, StoreRegKind kind) | |
7576 | { | |
7577 | TCGv_i32 tmp1, tmp2; | |
7578 | ||
7579 | tmp2 = load_reg(s, a->rm); | |
7580 | gen_arm_shift_im(tmp2, a->shty, a->shim, logic_cc); | |
7581 | tmp1 = load_reg(s, a->rn); | |
7582 | ||
7583 | gen(tmp1, tmp1, tmp2); | |
7584 | tcg_temp_free_i32(tmp2); | |
7585 | ||
7586 | if (logic_cc) { | |
7587 | gen_logic_CC(tmp1); | |
7588 | } | |
7589 | return store_reg_kind(s, a->rd, tmp1, kind); | |
7590 | } | |
7591 | ||
7592 | static bool op_s_rxr_shi(DisasContext *s, arg_s_rrr_shi *a, | |
7593 | void (*gen)(TCGv_i32, TCGv_i32), | |
7594 | int logic_cc, StoreRegKind kind) | |
7595 | { | |
7596 | TCGv_i32 tmp; | |
7597 | ||
7598 | tmp = load_reg(s, a->rm); | |
7599 | gen_arm_shift_im(tmp, a->shty, a->shim, logic_cc); | |
7600 | ||
7601 | gen(tmp, tmp); | |
7602 | if (logic_cc) { | |
7603 | gen_logic_CC(tmp); | |
7604 | } | |
7605 | return store_reg_kind(s, a->rd, tmp, kind); | |
7606 | } | |
7607 | ||
5be2c123 RH |
7608 | /* |
7609 | * Data-processing (register-shifted register) | |
7610 | * | |
7611 | * Operate, with set flags, one register source, | |
7612 | * one register shifted register source, and a destination. | |
7613 | */ | |
7614 | static bool op_s_rrr_shr(DisasContext *s, arg_s_rrr_shr *a, | |
7615 | void (*gen)(TCGv_i32, TCGv_i32, TCGv_i32), | |
7616 | int logic_cc, StoreRegKind kind) | |
7617 | { | |
7618 | TCGv_i32 tmp1, tmp2; | |
7619 | ||
7620 | tmp1 = load_reg(s, a->rs); | |
7621 | tmp2 = load_reg(s, a->rm); | |
7622 | gen_arm_shift_reg(tmp2, a->shty, tmp1, logic_cc); | |
7623 | tmp1 = load_reg(s, a->rn); | |
7624 | ||
7625 | gen(tmp1, tmp1, tmp2); | |
7626 | tcg_temp_free_i32(tmp2); | |
7627 | ||
7628 | if (logic_cc) { | |
7629 | gen_logic_CC(tmp1); | |
7630 | } | |
7631 | return store_reg_kind(s, a->rd, tmp1, kind); | |
7632 | } | |
7633 | ||
7634 | static bool op_s_rxr_shr(DisasContext *s, arg_s_rrr_shr *a, | |
7635 | void (*gen)(TCGv_i32, TCGv_i32), | |
7636 | int logic_cc, StoreRegKind kind) | |
7637 | { | |
7638 | TCGv_i32 tmp1, tmp2; | |
7639 | ||
7640 | tmp1 = load_reg(s, a->rs); | |
7641 | tmp2 = load_reg(s, a->rm); | |
7642 | gen_arm_shift_reg(tmp2, a->shty, tmp1, logic_cc); | |
7643 | ||
7644 | gen(tmp2, tmp2); | |
7645 | if (logic_cc) { | |
7646 | gen_logic_CC(tmp2); | |
7647 | } | |
7648 | return store_reg_kind(s, a->rd, tmp2, kind); | |
7649 | } | |
7650 | ||
581c6ebd RH |
7651 | /* |
7652 | * Data-processing (immediate) | |
7653 | * | |
7654 | * Operate, with set flags, one register source, | |
7655 | * one rotated immediate, and a destination. | |
7656 | * | |
7657 | * Note that logic_cc && a->rot setting CF based on the msb of the | |
7658 | * immediate is the reason why we must pass in the unrotated form | |
7659 | * of the immediate. | |
7660 | */ | |
7661 | static bool op_s_rri_rot(DisasContext *s, arg_s_rri_rot *a, | |
7662 | void (*gen)(TCGv_i32, TCGv_i32, TCGv_i32), | |
7663 | int logic_cc, StoreRegKind kind) | |
7664 | { | |
7665 | TCGv_i32 tmp1, tmp2; | |
7666 | uint32_t imm; | |
7667 | ||
7668 | imm = ror32(a->imm, a->rot); | |
7669 | if (logic_cc && a->rot) { | |
7670 | tcg_gen_movi_i32(cpu_CF, imm >> 31); | |
7671 | } | |
7672 | tmp2 = tcg_const_i32(imm); | |
7673 | tmp1 = load_reg(s, a->rn); | |
7674 | ||
7675 | gen(tmp1, tmp1, tmp2); | |
7676 | tcg_temp_free_i32(tmp2); | |
7677 | ||
7678 | if (logic_cc) { | |
7679 | gen_logic_CC(tmp1); | |
7680 | } | |
7681 | return store_reg_kind(s, a->rd, tmp1, kind); | |
7682 | } | |
7683 | ||
7684 | static bool op_s_rxi_rot(DisasContext *s, arg_s_rri_rot *a, | |
7685 | void (*gen)(TCGv_i32, TCGv_i32), | |
7686 | int logic_cc, StoreRegKind kind) | |
7687 | { | |
7688 | TCGv_i32 tmp; | |
7689 | uint32_t imm; | |
7690 | ||
7691 | imm = ror32(a->imm, a->rot); | |
7692 | if (logic_cc && a->rot) { | |
7693 | tcg_gen_movi_i32(cpu_CF, imm >> 31); | |
7694 | } | |
7695 | tmp = tcg_const_i32(imm); | |
7696 | ||
7697 | gen(tmp, tmp); | |
7698 | if (logic_cc) { | |
7699 | gen_logic_CC(tmp); | |
7700 | } | |
7701 | return store_reg_kind(s, a->rd, tmp, kind); | |
7702 | } | |
7703 | ||
25ae32c5 RH |
7704 | #define DO_ANY3(NAME, OP, L, K) \ |
7705 | static bool trans_##NAME##_rrri(DisasContext *s, arg_s_rrr_shi *a) \ | |
5be2c123 RH |
7706 | { StoreRegKind k = (K); return op_s_rrr_shi(s, a, OP, L, k); } \ |
7707 | static bool trans_##NAME##_rrrr(DisasContext *s, arg_s_rrr_shr *a) \ | |
581c6ebd RH |
7708 | { StoreRegKind k = (K); return op_s_rrr_shr(s, a, OP, L, k); } \ |
7709 | static bool trans_##NAME##_rri(DisasContext *s, arg_s_rri_rot *a) \ | |
7710 | { StoreRegKind k = (K); return op_s_rri_rot(s, a, OP, L, k); } | |
25ae32c5 RH |
7711 | |
7712 | #define DO_ANY2(NAME, OP, L, K) \ | |
7713 | static bool trans_##NAME##_rxri(DisasContext *s, arg_s_rrr_shi *a) \ | |
5be2c123 RH |
7714 | { StoreRegKind k = (K); return op_s_rxr_shi(s, a, OP, L, k); } \ |
7715 | static bool trans_##NAME##_rxrr(DisasContext *s, arg_s_rrr_shr *a) \ | |
581c6ebd RH |
7716 | { StoreRegKind k = (K); return op_s_rxr_shr(s, a, OP, L, k); } \ |
7717 | static bool trans_##NAME##_rxi(DisasContext *s, arg_s_rri_rot *a) \ | |
7718 | { StoreRegKind k = (K); return op_s_rxi_rot(s, a, OP, L, k); } | |
25ae32c5 RH |
7719 | |
7720 | #define DO_CMP2(NAME, OP, L) \ | |
7721 | static bool trans_##NAME##_xrri(DisasContext *s, arg_s_rrr_shi *a) \ | |
5be2c123 RH |
7722 | { return op_s_rrr_shi(s, a, OP, L, STREG_NONE); } \ |
7723 | static bool trans_##NAME##_xrrr(DisasContext *s, arg_s_rrr_shr *a) \ | |
581c6ebd RH |
7724 | { return op_s_rrr_shr(s, a, OP, L, STREG_NONE); } \ |
7725 | static bool trans_##NAME##_xri(DisasContext *s, arg_s_rri_rot *a) \ | |
7726 | { return op_s_rri_rot(s, a, OP, L, STREG_NONE); } | |
25ae32c5 RH |
7727 | |
7728 | DO_ANY3(AND, tcg_gen_and_i32, a->s, STREG_NORMAL) | |
7729 | DO_ANY3(EOR, tcg_gen_xor_i32, a->s, STREG_NORMAL) | |
7730 | DO_ANY3(ORR, tcg_gen_or_i32, a->s, STREG_NORMAL) | |
7731 | DO_ANY3(BIC, tcg_gen_andc_i32, a->s, STREG_NORMAL) | |
7732 | ||
7733 | DO_ANY3(RSB, a->s ? gen_rsb_CC : gen_rsb, false, STREG_NORMAL) | |
7734 | DO_ANY3(ADC, a->s ? gen_adc_CC : gen_add_carry, false, STREG_NORMAL) | |
7735 | DO_ANY3(SBC, a->s ? gen_sbc_CC : gen_sub_carry, false, STREG_NORMAL) | |
7736 | DO_ANY3(RSC, a->s ? gen_rsc_CC : gen_rsc, false, STREG_NORMAL) | |
7737 | ||
7738 | DO_CMP2(TST, tcg_gen_and_i32, true) | |
7739 | DO_CMP2(TEQ, tcg_gen_xor_i32, true) | |
7740 | DO_CMP2(CMN, gen_add_CC, false) | |
7741 | DO_CMP2(CMP, gen_sub_CC, false) | |
7742 | ||
7743 | DO_ANY3(ADD, a->s ? gen_add_CC : tcg_gen_add_i32, false, | |
7744 | a->rd == 13 && a->rn == 13 ? STREG_SP_CHECK : STREG_NORMAL) | |
7745 | ||
7746 | /* | |
7747 | * Note for the computation of StoreRegKind we return out of the | |
7748 | * middle of the functions that are expanded by DO_ANY3, and that | |
7749 | * we modify a->s via that parameter before it is used by OP. | |
7750 | */ | |
7751 | DO_ANY3(SUB, a->s ? gen_sub_CC : tcg_gen_sub_i32, false, | |
7752 | ({ | |
7753 | StoreRegKind ret = STREG_NORMAL; | |
7754 | if (a->rd == 15 && a->s) { | |
7755 | /* | |
7756 | * See ALUExceptionReturn: | |
7757 | * In User mode, UNPREDICTABLE; we choose UNDEF. | |
7758 | * In Hyp mode, UNDEFINED. | |
7759 | */ | |
7760 | if (IS_USER(s) || s->current_el == 2) { | |
7761 | unallocated_encoding(s); | |
7762 | return true; | |
7763 | } | |
7764 | /* There is no writeback of nzcv to PSTATE. */ | |
7765 | a->s = 0; | |
7766 | ret = STREG_EXC_RET; | |
7767 | } else if (a->rd == 13 && a->rn == 13) { | |
7768 | ret = STREG_SP_CHECK; | |
7769 | } | |
7770 | ret; | |
7771 | })) | |
7772 | ||
7773 | DO_ANY2(MOV, tcg_gen_mov_i32, a->s, | |
7774 | ({ | |
7775 | StoreRegKind ret = STREG_NORMAL; | |
7776 | if (a->rd == 15 && a->s) { | |
7777 | /* | |
7778 | * See ALUExceptionReturn: | |
7779 | * In User mode, UNPREDICTABLE; we choose UNDEF. | |
7780 | * In Hyp mode, UNDEFINED. | |
7781 | */ | |
7782 | if (IS_USER(s) || s->current_el == 2) { | |
7783 | unallocated_encoding(s); | |
7784 | return true; | |
7785 | } | |
7786 | /* There is no writeback of nzcv to PSTATE. */ | |
7787 | a->s = 0; | |
7788 | ret = STREG_EXC_RET; | |
7789 | } else if (a->rd == 13) { | |
7790 | ret = STREG_SP_CHECK; | |
7791 | } | |
7792 | ret; | |
7793 | })) | |
7794 | ||
7795 | DO_ANY2(MVN, tcg_gen_not_i32, a->s, STREG_NORMAL) | |
7796 | ||
7797 | /* | |
7798 | * ORN is only available with T32, so there is no register-shifted-register | |
7799 | * form of the insn. Using the DO_ANY3 macro would create an unused function. | |
7800 | */ | |
7801 | static bool trans_ORN_rrri(DisasContext *s, arg_s_rrr_shi *a) | |
7802 | { | |
7803 | return op_s_rrr_shi(s, a, tcg_gen_orc_i32, a->s, STREG_NORMAL); | |
7804 | } | |
7805 | ||
581c6ebd RH |
7806 | static bool trans_ORN_rri(DisasContext *s, arg_s_rri_rot *a) |
7807 | { | |
7808 | return op_s_rri_rot(s, a, tcg_gen_orc_i32, a->s, STREG_NORMAL); | |
7809 | } | |
7810 | ||
25ae32c5 RH |
7811 | #undef DO_ANY3 |
7812 | #undef DO_ANY2 | |
7813 | #undef DO_CMP2 | |
7814 | ||
145952e8 RH |
7815 | static bool trans_ADR(DisasContext *s, arg_ri *a) |
7816 | { | |
7817 | store_reg_bx(s, a->rd, add_reg_for_lit(s, 15, a->imm)); | |
7818 | return true; | |
7819 | } | |
7820 | ||
8f445127 RH |
7821 | static bool trans_MOVW(DisasContext *s, arg_MOVW *a) |
7822 | { | |
7823 | TCGv_i32 tmp; | |
7824 | ||
7825 | if (!ENABLE_ARCH_6T2) { | |
7826 | return false; | |
7827 | } | |
7828 | ||
7829 | tmp = tcg_const_i32(a->imm); | |
7830 | store_reg(s, a->rd, tmp); | |
7831 | return true; | |
7832 | } | |
7833 | ||
7834 | static bool trans_MOVT(DisasContext *s, arg_MOVW *a) | |
7835 | { | |
7836 | TCGv_i32 tmp; | |
7837 | ||
7838 | if (!ENABLE_ARCH_6T2) { | |
7839 | return false; | |
7840 | } | |
7841 | ||
7842 | tmp = load_reg(s, a->rd); | |
7843 | tcg_gen_ext16u_i32(tmp, tmp); | |
7844 | tcg_gen_ori_i32(tmp, tmp, a->imm << 16); | |
7845 | store_reg(s, a->rd, tmp); | |
7846 | return true; | |
7847 | } | |
7848 | ||
bd92fe35 RH |
7849 | /* |
7850 | * Multiply and multiply accumulate | |
7851 | */ | |
7852 | ||
7853 | static bool op_mla(DisasContext *s, arg_s_rrrr *a, bool add) | |
7854 | { | |
7855 | TCGv_i32 t1, t2; | |
7856 | ||
7857 | t1 = load_reg(s, a->rn); | |
7858 | t2 = load_reg(s, a->rm); | |
7859 | tcg_gen_mul_i32(t1, t1, t2); | |
7860 | tcg_temp_free_i32(t2); | |
7861 | if (add) { | |
7862 | t2 = load_reg(s, a->ra); | |
7863 | tcg_gen_add_i32(t1, t1, t2); | |
7864 | tcg_temp_free_i32(t2); | |
7865 | } | |
7866 | if (a->s) { | |
7867 | gen_logic_CC(t1); | |
7868 | } | |
7869 | store_reg(s, a->rd, t1); | |
7870 | return true; | |
7871 | } | |
7872 | ||
7873 | static bool trans_MUL(DisasContext *s, arg_MUL *a) | |
7874 | { | |
7875 | return op_mla(s, a, false); | |
7876 | } | |
7877 | ||
7878 | static bool trans_MLA(DisasContext *s, arg_MLA *a) | |
7879 | { | |
7880 | return op_mla(s, a, true); | |
7881 | } | |
7882 | ||
7883 | static bool trans_MLS(DisasContext *s, arg_MLS *a) | |
7884 | { | |
7885 | TCGv_i32 t1, t2; | |
7886 | ||
7887 | if (!ENABLE_ARCH_6T2) { | |
7888 | return false; | |
7889 | } | |
7890 | t1 = load_reg(s, a->rn); | |
7891 | t2 = load_reg(s, a->rm); | |
7892 | tcg_gen_mul_i32(t1, t1, t2); | |
7893 | tcg_temp_free_i32(t2); | |
7894 | t2 = load_reg(s, a->ra); | |
7895 | tcg_gen_sub_i32(t1, t2, t1); | |
7896 | tcg_temp_free_i32(t2); | |
7897 | store_reg(s, a->rd, t1); | |
7898 | return true; | |
7899 | } | |
7900 | ||
7901 | static bool op_mlal(DisasContext *s, arg_s_rrrr *a, bool uns, bool add) | |
7902 | { | |
7903 | TCGv_i32 t0, t1, t2, t3; | |
7904 | ||
7905 | t0 = load_reg(s, a->rm); | |
7906 | t1 = load_reg(s, a->rn); | |
7907 | if (uns) { | |
7908 | tcg_gen_mulu2_i32(t0, t1, t0, t1); | |
7909 | } else { | |
7910 | tcg_gen_muls2_i32(t0, t1, t0, t1); | |
7911 | } | |
7912 | if (add) { | |
7913 | t2 = load_reg(s, a->ra); | |
7914 | t3 = load_reg(s, a->rd); | |
7915 | tcg_gen_add2_i32(t0, t1, t0, t1, t2, t3); | |
7916 | tcg_temp_free_i32(t2); | |
7917 | tcg_temp_free_i32(t3); | |
7918 | } | |
7919 | if (a->s) { | |
7920 | gen_logicq_cc(t0, t1); | |
7921 | } | |
7922 | store_reg(s, a->ra, t0); | |
7923 | store_reg(s, a->rd, t1); | |
7924 | return true; | |
7925 | } | |
7926 | ||
7927 | static bool trans_UMULL(DisasContext *s, arg_UMULL *a) | |
7928 | { | |
7929 | return op_mlal(s, a, true, false); | |
7930 | } | |
7931 | ||
7932 | static bool trans_SMULL(DisasContext *s, arg_SMULL *a) | |
7933 | { | |
7934 | return op_mlal(s, a, false, false); | |
7935 | } | |
7936 | ||
7937 | static bool trans_UMLAL(DisasContext *s, arg_UMLAL *a) | |
7938 | { | |
7939 | return op_mlal(s, a, true, true); | |
7940 | } | |
7941 | ||
7942 | static bool trans_SMLAL(DisasContext *s, arg_SMLAL *a) | |
7943 | { | |
7944 | return op_mlal(s, a, false, true); | |
7945 | } | |
7946 | ||
7947 | static bool trans_UMAAL(DisasContext *s, arg_UMAAL *a) | |
7948 | { | |
2409d564 | 7949 | TCGv_i32 t0, t1, t2, zero; |
bd92fe35 RH |
7950 | |
7951 | if (s->thumb | |
7952 | ? !arm_dc_feature(s, ARM_FEATURE_THUMB_DSP) | |
7953 | : !ENABLE_ARCH_6) { | |
7954 | return false; | |
7955 | } | |
7956 | ||
7957 | t0 = load_reg(s, a->rm); | |
7958 | t1 = load_reg(s, a->rn); | |
2409d564 RH |
7959 | tcg_gen_mulu2_i32(t0, t1, t0, t1); |
7960 | zero = tcg_const_i32(0); | |
7961 | t2 = load_reg(s, a->ra); | |
7962 | tcg_gen_add2_i32(t0, t1, t0, t1, t2, zero); | |
7963 | tcg_temp_free_i32(t2); | |
7964 | t2 = load_reg(s, a->rd); | |
7965 | tcg_gen_add2_i32(t0, t1, t0, t1, t2, zero); | |
7966 | tcg_temp_free_i32(t2); | |
7967 | tcg_temp_free_i32(zero); | |
7968 | store_reg(s, a->ra, t0); | |
7969 | store_reg(s, a->rd, t1); | |
bd92fe35 RH |
7970 | return true; |
7971 | } | |
7972 | ||
6d0730a8 RH |
7973 | /* |
7974 | * Saturating addition and subtraction | |
7975 | */ | |
7976 | ||
7977 | static bool op_qaddsub(DisasContext *s, arg_rrr *a, bool add, bool doub) | |
7978 | { | |
7979 | TCGv_i32 t0, t1; | |
7980 | ||
7981 | if (s->thumb | |
7982 | ? !arm_dc_feature(s, ARM_FEATURE_THUMB_DSP) | |
7983 | : !ENABLE_ARCH_5TE) { | |
7984 | return false; | |
7985 | } | |
7986 | ||
7987 | t0 = load_reg(s, a->rm); | |
7988 | t1 = load_reg(s, a->rn); | |
7989 | if (doub) { | |
7990 | gen_helper_add_saturate(t1, cpu_env, t1, t1); | |
7991 | } | |
7992 | if (add) { | |
7993 | gen_helper_add_saturate(t0, cpu_env, t0, t1); | |
7994 | } else { | |
7995 | gen_helper_sub_saturate(t0, cpu_env, t0, t1); | |
7996 | } | |
7997 | tcg_temp_free_i32(t1); | |
7998 | store_reg(s, a->rd, t0); | |
7999 | return true; | |
8000 | } | |
8001 | ||
8002 | #define DO_QADDSUB(NAME, ADD, DOUB) \ | |
8003 | static bool trans_##NAME(DisasContext *s, arg_rrr *a) \ | |
8004 | { \ | |
8005 | return op_qaddsub(s, a, ADD, DOUB); \ | |
8006 | } | |
8007 | ||
8008 | DO_QADDSUB(QADD, true, false) | |
8009 | DO_QADDSUB(QSUB, false, false) | |
8010 | DO_QADDSUB(QDADD, true, true) | |
8011 | DO_QADDSUB(QDSUB, false, true) | |
8012 | ||
8013 | #undef DO_QADDSUB | |
8014 | ||
26c6923d RH |
8015 | /* |
8016 | * Halfword multiply and multiply accumulate | |
8017 | */ | |
8018 | ||
8019 | static bool op_smlaxxx(DisasContext *s, arg_rrrr *a, | |
8020 | int add_long, bool nt, bool mt) | |
8021 | { | |
ea96b374 | 8022 | TCGv_i32 t0, t1, tl, th; |
26c6923d RH |
8023 | |
8024 | if (s->thumb | |
8025 | ? !arm_dc_feature(s, ARM_FEATURE_THUMB_DSP) | |
8026 | : !ENABLE_ARCH_5TE) { | |
8027 | return false; | |
8028 | } | |
8029 | ||
8030 | t0 = load_reg(s, a->rn); | |
8031 | t1 = load_reg(s, a->rm); | |
8032 | gen_mulxy(t0, t1, nt, mt); | |
8033 | tcg_temp_free_i32(t1); | |
8034 | ||
8035 | switch (add_long) { | |
8036 | case 0: | |
8037 | store_reg(s, a->rd, t0); | |
8038 | break; | |
8039 | case 1: | |
8040 | t1 = load_reg(s, a->ra); | |
8041 | gen_helper_add_setq(t0, cpu_env, t0, t1); | |
8042 | tcg_temp_free_i32(t1); | |
8043 | store_reg(s, a->rd, t0); | |
8044 | break; | |
8045 | case 2: | |
ea96b374 RH |
8046 | tl = load_reg(s, a->ra); |
8047 | th = load_reg(s, a->rd); | |
8048 | t1 = tcg_const_i32(0); | |
8049 | tcg_gen_add2_i32(tl, th, tl, th, t0, t1); | |
26c6923d | 8050 | tcg_temp_free_i32(t0); |
ea96b374 RH |
8051 | tcg_temp_free_i32(t1); |
8052 | store_reg(s, a->ra, tl); | |
8053 | store_reg(s, a->rd, th); | |
26c6923d RH |
8054 | break; |
8055 | default: | |
8056 | g_assert_not_reached(); | |
8057 | } | |
8058 | return true; | |
8059 | } | |
8060 | ||
8061 | #define DO_SMLAX(NAME, add, nt, mt) \ | |
8062 | static bool trans_##NAME(DisasContext *s, arg_rrrr *a) \ | |
8063 | { \ | |
8064 | return op_smlaxxx(s, a, add, nt, mt); \ | |
8065 | } | |
8066 | ||
8067 | DO_SMLAX(SMULBB, 0, 0, 0) | |
8068 | DO_SMLAX(SMULBT, 0, 0, 1) | |
8069 | DO_SMLAX(SMULTB, 0, 1, 0) | |
8070 | DO_SMLAX(SMULTT, 0, 1, 1) | |
8071 | ||
8072 | DO_SMLAX(SMLABB, 1, 0, 0) | |
8073 | DO_SMLAX(SMLABT, 1, 0, 1) | |
8074 | DO_SMLAX(SMLATB, 1, 1, 0) | |
8075 | DO_SMLAX(SMLATT, 1, 1, 1) | |
8076 | ||
8077 | DO_SMLAX(SMLALBB, 2, 0, 0) | |
8078 | DO_SMLAX(SMLALBT, 2, 0, 1) | |
8079 | DO_SMLAX(SMLALTB, 2, 1, 0) | |
8080 | DO_SMLAX(SMLALTT, 2, 1, 1) | |
8081 | ||
8082 | #undef DO_SMLAX | |
8083 | ||
8084 | static bool op_smlawx(DisasContext *s, arg_rrrr *a, bool add, bool mt) | |
8085 | { | |
8086 | TCGv_i32 t0, t1; | |
26c6923d RH |
8087 | |
8088 | if (!ENABLE_ARCH_5TE) { | |
8089 | return false; | |
8090 | } | |
8091 | ||
8092 | t0 = load_reg(s, a->rn); | |
8093 | t1 = load_reg(s, a->rm); | |
485b607d RH |
8094 | /* |
8095 | * Since the nominal result is product<47:16>, shift the 16-bit | |
8096 | * input up by 16 bits, so that the result is at product<63:32>. | |
8097 | */ | |
26c6923d | 8098 | if (mt) { |
485b607d | 8099 | tcg_gen_andi_i32(t1, t1, 0xffff0000); |
26c6923d | 8100 | } else { |
485b607d | 8101 | tcg_gen_shli_i32(t1, t1, 16); |
26c6923d | 8102 | } |
485b607d RH |
8103 | tcg_gen_muls2_i32(t0, t1, t0, t1); |
8104 | tcg_temp_free_i32(t0); | |
26c6923d RH |
8105 | if (add) { |
8106 | t0 = load_reg(s, a->ra); | |
8107 | gen_helper_add_setq(t1, cpu_env, t1, t0); | |
8108 | tcg_temp_free_i32(t0); | |
8109 | } | |
8110 | store_reg(s, a->rd, t1); | |
8111 | return true; | |
8112 | } | |
8113 | ||
8114 | #define DO_SMLAWX(NAME, add, mt) \ | |
8115 | static bool trans_##NAME(DisasContext *s, arg_rrrr *a) \ | |
8116 | { \ | |
8117 | return op_smlawx(s, a, add, mt); \ | |
8118 | } | |
8119 | ||
8120 | DO_SMLAWX(SMULWB, 0, 0) | |
8121 | DO_SMLAWX(SMULWT, 0, 1) | |
8122 | DO_SMLAWX(SMLAWB, 1, 0) | |
8123 | DO_SMLAWX(SMLAWT, 1, 1) | |
8124 | ||
8125 | #undef DO_SMLAWX | |
8126 | ||
63130596 RH |
8127 | /* |
8128 | * MSR (immediate) and hints | |
8129 | */ | |
8130 | ||
8131 | static bool trans_YIELD(DisasContext *s, arg_YIELD *a) | |
8132 | { | |
279de61a RH |
8133 | /* |
8134 | * When running single-threaded TCG code, use the helper to ensure that | |
8135 | * the next round-robin scheduled vCPU gets a crack. When running in | |
8136 | * MTTCG we don't generate jumps to the helper as it won't affect the | |
8137 | * scheduling of other vCPUs. | |
8138 | */ | |
8139 | if (!(tb_cflags(s->base.tb) & CF_PARALLEL)) { | |
8140 | gen_set_pc_im(s, s->base.pc_next); | |
8141 | s->base.is_jmp = DISAS_YIELD; | |
8142 | } | |
63130596 RH |
8143 | return true; |
8144 | } | |
8145 | ||
8146 | static bool trans_WFE(DisasContext *s, arg_WFE *a) | |
8147 | { | |
279de61a RH |
8148 | /* |
8149 | * When running single-threaded TCG code, use the helper to ensure that | |
8150 | * the next round-robin scheduled vCPU gets a crack. In MTTCG mode we | |
8151 | * just skip this instruction. Currently the SEV/SEVL instructions, | |
8152 | * which are *one* of many ways to wake the CPU from WFE, are not | |
8153 | * implemented so we can't sleep like WFI does. | |
8154 | */ | |
8155 | if (!(tb_cflags(s->base.tb) & CF_PARALLEL)) { | |
8156 | gen_set_pc_im(s, s->base.pc_next); | |
8157 | s->base.is_jmp = DISAS_WFE; | |
8158 | } | |
63130596 RH |
8159 | return true; |
8160 | } | |
8161 | ||
8162 | static bool trans_WFI(DisasContext *s, arg_WFI *a) | |
8163 | { | |
279de61a RH |
8164 | /* For WFI, halt the vCPU until an IRQ. */ |
8165 | gen_set_pc_im(s, s->base.pc_next); | |
8166 | s->base.is_jmp = DISAS_WFI; | |
63130596 RH |
8167 | return true; |
8168 | } | |
8169 | ||
8170 | static bool trans_NOP(DisasContext *s, arg_NOP *a) | |
8171 | { | |
8172 | return true; | |
8173 | } | |
8174 | ||
8175 | static bool trans_MSR_imm(DisasContext *s, arg_MSR_imm *a) | |
8176 | { | |
8177 | uint32_t val = ror32(a->imm, a->rot * 2); | |
8178 | uint32_t mask = msr_mask(s, a->mask, a->r); | |
8179 | ||
8180 | if (gen_set_psr_im(s, mask, a->r, val)) { | |
8181 | unallocated_encoding(s); | |
8182 | } | |
8183 | return true; | |
8184 | } | |
8185 | ||
6c35d53f RH |
8186 | /* |
8187 | * Cyclic Redundancy Check | |
8188 | */ | |
8189 | ||
8190 | static bool op_crc32(DisasContext *s, arg_rrr *a, bool c, MemOp sz) | |
8191 | { | |
8192 | TCGv_i32 t1, t2, t3; | |
8193 | ||
8194 | if (!dc_isar_feature(aa32_crc32, s)) { | |
8195 | return false; | |
8196 | } | |
8197 | ||
8198 | t1 = load_reg(s, a->rn); | |
8199 | t2 = load_reg(s, a->rm); | |
8200 | switch (sz) { | |
8201 | case MO_8: | |
8202 | gen_uxtb(t2); | |
8203 | break; | |
8204 | case MO_16: | |
8205 | gen_uxth(t2); | |
8206 | break; | |
8207 | case MO_32: | |
8208 | break; | |
8209 | default: | |
8210 | g_assert_not_reached(); | |
8211 | } | |
8212 | t3 = tcg_const_i32(1 << sz); | |
8213 | if (c) { | |
8214 | gen_helper_crc32c(t1, t1, t2, t3); | |
8215 | } else { | |
8216 | gen_helper_crc32(t1, t1, t2, t3); | |
8217 | } | |
8218 | tcg_temp_free_i32(t2); | |
8219 | tcg_temp_free_i32(t3); | |
8220 | store_reg(s, a->rd, t1); | |
8221 | return true; | |
8222 | } | |
8223 | ||
8224 | #define DO_CRC32(NAME, c, sz) \ | |
8225 | static bool trans_##NAME(DisasContext *s, arg_rrr *a) \ | |
8226 | { return op_crc32(s, a, c, sz); } | |
8227 | ||
8228 | DO_CRC32(CRC32B, false, MO_8) | |
8229 | DO_CRC32(CRC32H, false, MO_16) | |
8230 | DO_CRC32(CRC32W, false, MO_32) | |
8231 | DO_CRC32(CRC32CB, true, MO_8) | |
8232 | DO_CRC32(CRC32CH, true, MO_16) | |
8233 | DO_CRC32(CRC32CW, true, MO_32) | |
8234 | ||
8235 | #undef DO_CRC32 | |
8236 | ||
d0b26644 RH |
8237 | /* |
8238 | * Miscellaneous instructions | |
8239 | */ | |
8240 | ||
8241 | static bool trans_MRS_bank(DisasContext *s, arg_MRS_bank *a) | |
8242 | { | |
8243 | if (arm_dc_feature(s, ARM_FEATURE_M)) { | |
8244 | return false; | |
8245 | } | |
8246 | gen_mrs_banked(s, a->r, a->sysm, a->rd); | |
8247 | return true; | |
8248 | } | |
8249 | ||
8250 | static bool trans_MSR_bank(DisasContext *s, arg_MSR_bank *a) | |
8251 | { | |
8252 | if (arm_dc_feature(s, ARM_FEATURE_M)) { | |
8253 | return false; | |
8254 | } | |
8255 | gen_msr_banked(s, a->r, a->sysm, a->rn); | |
8256 | return true; | |
8257 | } | |
8258 | ||
8259 | static bool trans_MRS_reg(DisasContext *s, arg_MRS_reg *a) | |
8260 | { | |
8261 | TCGv_i32 tmp; | |
8262 | ||
8263 | if (arm_dc_feature(s, ARM_FEATURE_M)) { | |
8264 | return false; | |
8265 | } | |
8266 | if (a->r) { | |
8267 | if (IS_USER(s)) { | |
8268 | unallocated_encoding(s); | |
8269 | return true; | |
8270 | } | |
8271 | tmp = load_cpu_field(spsr); | |
8272 | } else { | |
8273 | tmp = tcg_temp_new_i32(); | |
8274 | gen_helper_cpsr_read(tmp, cpu_env); | |
8275 | } | |
8276 | store_reg(s, a->rd, tmp); | |
8277 | return true; | |
8278 | } | |
8279 | ||
8280 | static bool trans_MSR_reg(DisasContext *s, arg_MSR_reg *a) | |
8281 | { | |
8282 | TCGv_i32 tmp; | |
8283 | uint32_t mask = msr_mask(s, a->mask, a->r); | |
8284 | ||
8285 | if (arm_dc_feature(s, ARM_FEATURE_M)) { | |
8286 | return false; | |
8287 | } | |
8288 | tmp = load_reg(s, a->rn); | |
8289 | if (gen_set_psr(s, mask, a->r, tmp)) { | |
8290 | unallocated_encoding(s); | |
8291 | } | |
8292 | return true; | |
8293 | } | |
8294 | ||
8295 | static bool trans_MRS_v7m(DisasContext *s, arg_MRS_v7m *a) | |
8296 | { | |
8297 | TCGv_i32 tmp; | |
8298 | ||
8299 | if (!arm_dc_feature(s, ARM_FEATURE_M)) { | |
8300 | return false; | |
8301 | } | |
8302 | tmp = tcg_const_i32(a->sysm); | |
8303 | gen_helper_v7m_mrs(tmp, cpu_env, tmp); | |
8304 | store_reg(s, a->rd, tmp); | |
8305 | return true; | |
8306 | } | |
8307 | ||
8308 | static bool trans_MSR_v7m(DisasContext *s, arg_MSR_v7m *a) | |
8309 | { | |
8310 | TCGv_i32 addr, reg; | |
8311 | ||
8312 | if (!arm_dc_feature(s, ARM_FEATURE_M)) { | |
8313 | return false; | |
8314 | } | |
8315 | addr = tcg_const_i32((a->mask << 10) | a->sysm); | |
8316 | reg = load_reg(s, a->rn); | |
8317 | gen_helper_v7m_msr(cpu_env, addr, reg); | |
8318 | tcg_temp_free_i32(addr); | |
8319 | tcg_temp_free_i32(reg); | |
8320 | gen_lookup_tb(s); | |
8321 | return true; | |
8322 | } | |
8323 | ||
4ed95abd RH |
8324 | static bool trans_BX(DisasContext *s, arg_BX *a) |
8325 | { | |
8326 | if (!ENABLE_ARCH_4T) { | |
8327 | return false; | |
8328 | } | |
a0ef0774 | 8329 | gen_bx_excret(s, load_reg(s, a->rm)); |
4ed95abd RH |
8330 | return true; |
8331 | } | |
8332 | ||
8333 | static bool trans_BXJ(DisasContext *s, arg_BXJ *a) | |
8334 | { | |
8335 | if (!ENABLE_ARCH_5J || arm_dc_feature(s, ARM_FEATURE_M)) { | |
8336 | return false; | |
8337 | } | |
8338 | /* Trivial implementation equivalent to bx. */ | |
8339 | gen_bx(s, load_reg(s, a->rm)); | |
8340 | return true; | |
8341 | } | |
8342 | ||
8343 | static bool trans_BLX_r(DisasContext *s, arg_BLX_r *a) | |
8344 | { | |
8345 | TCGv_i32 tmp; | |
8346 | ||
8347 | if (!ENABLE_ARCH_5) { | |
8348 | return false; | |
8349 | } | |
8350 | tmp = load_reg(s, a->rm); | |
8351 | tcg_gen_movi_i32(cpu_R[14], s->base.pc_next | s->thumb); | |
8352 | gen_bx(s, tmp); | |
8353 | return true; | |
8354 | } | |
8355 | ||
a0ef0774 RH |
8356 | /* |
8357 | * BXNS/BLXNS: only exist for v8M with the security extensions, | |
8358 | * and always UNDEF if NonSecure. We don't implement these in | |
8359 | * the user-only mode either (in theory you can use them from | |
8360 | * Secure User mode but they are too tied in to system emulation). | |
8361 | */ | |
8362 | static bool trans_BXNS(DisasContext *s, arg_BXNS *a) | |
8363 | { | |
8364 | if (!s->v8m_secure || IS_USER_ONLY) { | |
8365 | unallocated_encoding(s); | |
8366 | } else { | |
8367 | gen_bxns(s, a->rm); | |
8368 | } | |
8369 | return true; | |
8370 | } | |
8371 | ||
8372 | static bool trans_BLXNS(DisasContext *s, arg_BLXNS *a) | |
8373 | { | |
8374 | if (!s->v8m_secure || IS_USER_ONLY) { | |
8375 | unallocated_encoding(s); | |
8376 | } else { | |
8377 | gen_blxns(s, a->rm); | |
8378 | } | |
8379 | return true; | |
8380 | } | |
8381 | ||
4c97f5b2 RH |
8382 | static bool trans_CLZ(DisasContext *s, arg_CLZ *a) |
8383 | { | |
8384 | TCGv_i32 tmp; | |
8385 | ||
8386 | if (!ENABLE_ARCH_5) { | |
8387 | return false; | |
8388 | } | |
8389 | tmp = load_reg(s, a->rm); | |
8390 | tcg_gen_clzi_i32(tmp, tmp, 32); | |
8391 | store_reg(s, a->rd, tmp); | |
8392 | return true; | |
8393 | } | |
8394 | ||
ef11bc3c RH |
8395 | static bool trans_ERET(DisasContext *s, arg_ERET *a) |
8396 | { | |
8397 | TCGv_i32 tmp; | |
8398 | ||
8399 | if (!arm_dc_feature(s, ARM_FEATURE_V7VE)) { | |
8400 | return false; | |
8401 | } | |
8402 | if (IS_USER(s)) { | |
8403 | unallocated_encoding(s); | |
8404 | return true; | |
8405 | } | |
8406 | if (s->current_el == 2) { | |
8407 | /* ERET from Hyp uses ELR_Hyp, not LR */ | |
8408 | tmp = load_cpu_field(elr_el[2]); | |
8409 | } else { | |
8410 | tmp = load_reg(s, 14); | |
8411 | } | |
8412 | gen_exception_return(s, tmp); | |
8413 | return true; | |
8414 | } | |
8415 | ||
2cde9ea5 RH |
8416 | static bool trans_HLT(DisasContext *s, arg_HLT *a) |
8417 | { | |
8418 | gen_hlt(s, a->imm); | |
8419 | return true; | |
8420 | } | |
8421 | ||
8422 | static bool trans_BKPT(DisasContext *s, arg_BKPT *a) | |
8423 | { | |
8424 | if (!ENABLE_ARCH_5) { | |
8425 | return false; | |
8426 | } | |
376214e4 AB |
8427 | if (arm_dc_feature(s, ARM_FEATURE_M) && |
8428 | semihosting_enabled() && | |
8429 | #ifndef CONFIG_USER_ONLY | |
8430 | !IS_USER(s) && | |
8431 | #endif | |
8432 | (a->imm == 0xab)) { | |
8433 | gen_exception_internal_insn(s, s->base.pc_next, EXCP_SEMIHOST); | |
8434 | } else { | |
8435 | gen_exception_bkpt_insn(s, syn_aa32_bkpt(a->imm, false)); | |
8436 | } | |
2cde9ea5 RH |
8437 | return true; |
8438 | } | |
8439 | ||
8440 | static bool trans_HVC(DisasContext *s, arg_HVC *a) | |
8441 | { | |
8442 | if (!ENABLE_ARCH_7 || arm_dc_feature(s, ARM_FEATURE_M)) { | |
8443 | return false; | |
8444 | } | |
8445 | if (IS_USER(s)) { | |
8446 | unallocated_encoding(s); | |
8447 | } else { | |
8448 | gen_hvc(s, a->imm); | |
8449 | } | |
8450 | return true; | |
8451 | } | |
8452 | ||
8453 | static bool trans_SMC(DisasContext *s, arg_SMC *a) | |
8454 | { | |
8455 | if (!ENABLE_ARCH_6K || arm_dc_feature(s, ARM_FEATURE_M)) { | |
8456 | return false; | |
8457 | } | |
8458 | if (IS_USER(s)) { | |
8459 | unallocated_encoding(s); | |
8460 | } else { | |
8461 | gen_smc(s); | |
8462 | } | |
8463 | return true; | |
8464 | } | |
8465 | ||
35d240ac RH |
8466 | static bool trans_SG(DisasContext *s, arg_SG *a) |
8467 | { | |
8468 | if (!arm_dc_feature(s, ARM_FEATURE_M) || | |
8469 | !arm_dc_feature(s, ARM_FEATURE_V8)) { | |
8470 | return false; | |
8471 | } | |
8472 | /* | |
8473 | * SG (v8M only) | |
8474 | * The bulk of the behaviour for this instruction is implemented | |
8475 | * in v7m_handle_execute_nsc(), which deals with the insn when | |
8476 | * it is executed by a CPU in non-secure state from memory | |
8477 | * which is Secure & NonSecure-Callable. | |
8478 | * Here we only need to handle the remaining cases: | |
8479 | * * in NS memory (including the "security extension not | |
8480 | * implemented" case) : NOP | |
8481 | * * in S memory but CPU already secure (clear IT bits) | |
8482 | * We know that the attribute for the memory this insn is | |
8483 | * in must match the current CPU state, because otherwise | |
8484 | * get_phys_addr_pmsav8 would have generated an exception. | |
8485 | */ | |
8486 | if (s->v8m_secure) { | |
8487 | /* Like the IT insn, we don't need to generate any code */ | |
8488 | s->condexec_cond = 0; | |
8489 | s->condexec_mask = 0; | |
8490 | } | |
8491 | return true; | |
8492 | } | |
8493 | ||
d449f174 RH |
8494 | static bool trans_TT(DisasContext *s, arg_TT *a) |
8495 | { | |
8496 | TCGv_i32 addr, tmp; | |
8497 | ||
8498 | if (!arm_dc_feature(s, ARM_FEATURE_M) || | |
8499 | !arm_dc_feature(s, ARM_FEATURE_V8)) { | |
8500 | return false; | |
8501 | } | |
8502 | if (a->rd == 13 || a->rd == 15 || a->rn == 15) { | |
8503 | /* We UNDEF for these UNPREDICTABLE cases */ | |
8504 | unallocated_encoding(s); | |
8505 | return true; | |
8506 | } | |
8507 | if (a->A && !s->v8m_secure) { | |
8508 | /* This case is UNDEFINED. */ | |
8509 | unallocated_encoding(s); | |
8510 | return true; | |
8511 | } | |
8512 | ||
8513 | addr = load_reg(s, a->rn); | |
8514 | tmp = tcg_const_i32((a->A << 1) | a->T); | |
8515 | gen_helper_v7m_tt(tmp, cpu_env, addr, tmp); | |
8516 | tcg_temp_free_i32(addr); | |
8517 | store_reg(s, a->rd, tmp); | |
8518 | return true; | |
8519 | } | |
8520 | ||
5e291fe1 RH |
8521 | /* |
8522 | * Load/store register index | |
8523 | */ | |
8524 | ||
8525 | static ISSInfo make_issinfo(DisasContext *s, int rd, bool p, bool w) | |
8526 | { | |
8527 | ISSInfo ret; | |
8528 | ||
8529 | /* ISS not valid if writeback */ | |
8530 | if (p && !w) { | |
8531 | ret = rd; | |
8532 | } else { | |
8533 | ret = ISSInvalid; | |
8534 | } | |
8535 | return ret; | |
8536 | } | |
8537 | ||
8538 | static TCGv_i32 op_addr_rr_pre(DisasContext *s, arg_ldst_rr *a) | |
8539 | { | |
8540 | TCGv_i32 addr = load_reg(s, a->rn); | |
8541 | ||
8542 | if (s->v8m_stackcheck && a->rn == 13 && a->w) { | |
8543 | gen_helper_v8m_stackcheck(cpu_env, addr); | |
8544 | } | |
8545 | ||
8546 | if (a->p) { | |
8547 | TCGv_i32 ofs = load_reg(s, a->rm); | |
8548 | gen_arm_shift_im(ofs, a->shtype, a->shimm, 0); | |
8549 | if (a->u) { | |
8550 | tcg_gen_add_i32(addr, addr, ofs); | |
8551 | } else { | |
8552 | tcg_gen_sub_i32(addr, addr, ofs); | |
8553 | } | |
8554 | tcg_temp_free_i32(ofs); | |
8555 | } | |
8556 | return addr; | |
8557 | } | |
8558 | ||
8559 | static void op_addr_rr_post(DisasContext *s, arg_ldst_rr *a, | |
8560 | TCGv_i32 addr, int address_offset) | |
8561 | { | |
8562 | if (!a->p) { | |
8563 | TCGv_i32 ofs = load_reg(s, a->rm); | |
8564 | gen_arm_shift_im(ofs, a->shtype, a->shimm, 0); | |
8565 | if (a->u) { | |
8566 | tcg_gen_add_i32(addr, addr, ofs); | |
8567 | } else { | |
8568 | tcg_gen_sub_i32(addr, addr, ofs); | |
8569 | } | |
8570 | tcg_temp_free_i32(ofs); | |
8571 | } else if (!a->w) { | |
8572 | tcg_temp_free_i32(addr); | |
8573 | return; | |
8574 | } | |
8575 | tcg_gen_addi_i32(addr, addr, address_offset); | |
8576 | store_reg(s, a->rn, addr); | |
8577 | } | |
8578 | ||
8579 | static bool op_load_rr(DisasContext *s, arg_ldst_rr *a, | |
8580 | MemOp mop, int mem_idx) | |
8581 | { | |
8582 | ISSInfo issinfo = make_issinfo(s, a->rt, a->p, a->w); | |
8583 | TCGv_i32 addr, tmp; | |
8584 | ||
8585 | addr = op_addr_rr_pre(s, a); | |
8586 | ||
8587 | tmp = tcg_temp_new_i32(); | |
8588 | gen_aa32_ld_i32(s, tmp, addr, mem_idx, mop | s->be_data); | |
8589 | disas_set_da_iss(s, mop, issinfo); | |
8590 | ||
8591 | /* | |
8592 | * Perform base writeback before the loaded value to | |
8593 | * ensure correct behavior with overlapping index registers. | |
8594 | */ | |
8595 | op_addr_rr_post(s, a, addr, 0); | |
8596 | store_reg_from_load(s, a->rt, tmp); | |
8597 | return true; | |
8598 | } | |
8599 | ||
8600 | static bool op_store_rr(DisasContext *s, arg_ldst_rr *a, | |
8601 | MemOp mop, int mem_idx) | |
8602 | { | |
8603 | ISSInfo issinfo = make_issinfo(s, a->rt, a->p, a->w) | ISSIsWrite; | |
8604 | TCGv_i32 addr, tmp; | |
8605 | ||
8606 | addr = op_addr_rr_pre(s, a); | |
8607 | ||
8608 | tmp = load_reg(s, a->rt); | |
8609 | gen_aa32_st_i32(s, tmp, addr, mem_idx, mop | s->be_data); | |
8610 | disas_set_da_iss(s, mop, issinfo); | |
8611 | tcg_temp_free_i32(tmp); | |
8612 | ||
8613 | op_addr_rr_post(s, a, addr, 0); | |
8614 | return true; | |
8615 | } | |
8616 | ||
8617 | static bool trans_LDRD_rr(DisasContext *s, arg_ldst_rr *a) | |
8618 | { | |
8619 | int mem_idx = get_mem_index(s); | |
8620 | TCGv_i32 addr, tmp; | |
8621 | ||
8622 | if (!ENABLE_ARCH_5TE) { | |
8623 | return false; | |
8624 | } | |
8625 | if (a->rt & 1) { | |
8626 | unallocated_encoding(s); | |
8627 | return true; | |
8628 | } | |
8629 | addr = op_addr_rr_pre(s, a); | |
8630 | ||
8631 | tmp = tcg_temp_new_i32(); | |
8632 | gen_aa32_ld_i32(s, tmp, addr, mem_idx, MO_UL | s->be_data); | |
8633 | store_reg(s, a->rt, tmp); | |
8634 | ||
8635 | tcg_gen_addi_i32(addr, addr, 4); | |
8636 | ||
8637 | tmp = tcg_temp_new_i32(); | |
8638 | gen_aa32_ld_i32(s, tmp, addr, mem_idx, MO_UL | s->be_data); | |
8639 | store_reg(s, a->rt + 1, tmp); | |
8640 | ||
8641 | /* LDRD w/ base writeback is undefined if the registers overlap. */ | |
8642 | op_addr_rr_post(s, a, addr, -4); | |
8643 | return true; | |
8644 | } | |
8645 | ||
8646 | static bool trans_STRD_rr(DisasContext *s, arg_ldst_rr *a) | |
8647 | { | |
8648 | int mem_idx = get_mem_index(s); | |
8649 | TCGv_i32 addr, tmp; | |
8650 | ||
8651 | if (!ENABLE_ARCH_5TE) { | |
8652 | return false; | |
8653 | } | |
8654 | if (a->rt & 1) { | |
8655 | unallocated_encoding(s); | |
8656 | return true; | |
8657 | } | |
8658 | addr = op_addr_rr_pre(s, a); | |
8659 | ||
8660 | tmp = load_reg(s, a->rt); | |
8661 | gen_aa32_st_i32(s, tmp, addr, mem_idx, MO_UL | s->be_data); | |
8662 | tcg_temp_free_i32(tmp); | |
8663 | ||
8664 | tcg_gen_addi_i32(addr, addr, 4); | |
8665 | ||
8666 | tmp = load_reg(s, a->rt + 1); | |
8667 | gen_aa32_st_i32(s, tmp, addr, mem_idx, MO_UL | s->be_data); | |
8668 | tcg_temp_free_i32(tmp); | |
8669 | ||
8670 | op_addr_rr_post(s, a, addr, -4); | |
8671 | return true; | |
8672 | } | |
8673 | ||
8674 | /* | |
8675 | * Load/store immediate index | |
8676 | */ | |
8677 | ||
8678 | static TCGv_i32 op_addr_ri_pre(DisasContext *s, arg_ldst_ri *a) | |
8679 | { | |
8680 | int ofs = a->imm; | |
8681 | ||
8682 | if (!a->u) { | |
8683 | ofs = -ofs; | |
8684 | } | |
8685 | ||
8686 | if (s->v8m_stackcheck && a->rn == 13 && a->w) { | |
8687 | /* | |
8688 | * Stackcheck. Here we know 'addr' is the current SP; | |
8689 | * U is set if we're moving SP up, else down. It is | |
8690 | * UNKNOWN whether the limit check triggers when SP starts | |
8691 | * below the limit and ends up above it; we chose to do so. | |
8692 | */ | |
8693 | if (!a->u) { | |
8694 | TCGv_i32 newsp = tcg_temp_new_i32(); | |
8695 | tcg_gen_addi_i32(newsp, cpu_R[13], ofs); | |
8696 | gen_helper_v8m_stackcheck(cpu_env, newsp); | |
8697 | tcg_temp_free_i32(newsp); | |
8698 | } else { | |
8699 | gen_helper_v8m_stackcheck(cpu_env, cpu_R[13]); | |
8700 | } | |
8701 | } | |
8702 | ||
8703 | return add_reg_for_lit(s, a->rn, a->p ? ofs : 0); | |
8704 | } | |
8705 | ||
8706 | static void op_addr_ri_post(DisasContext *s, arg_ldst_ri *a, | |
8707 | TCGv_i32 addr, int address_offset) | |
8708 | { | |
8709 | if (!a->p) { | |
8710 | if (a->u) { | |
8711 | address_offset += a->imm; | |
8712 | } else { | |
8713 | address_offset -= a->imm; | |
8714 | } | |
8715 | } else if (!a->w) { | |
8716 | tcg_temp_free_i32(addr); | |
8717 | return; | |
8718 | } | |
8719 | tcg_gen_addi_i32(addr, addr, address_offset); | |
8720 | store_reg(s, a->rn, addr); | |
8721 | } | |
8722 | ||
8723 | static bool op_load_ri(DisasContext *s, arg_ldst_ri *a, | |
8724 | MemOp mop, int mem_idx) | |
8725 | { | |
8726 | ISSInfo issinfo = make_issinfo(s, a->rt, a->p, a->w); | |
8727 | TCGv_i32 addr, tmp; | |
8728 | ||
8729 | addr = op_addr_ri_pre(s, a); | |
8730 | ||
8731 | tmp = tcg_temp_new_i32(); | |
8732 | gen_aa32_ld_i32(s, tmp, addr, mem_idx, mop | s->be_data); | |
8733 | disas_set_da_iss(s, mop, issinfo); | |
8734 | ||
8735 | /* | |
8736 | * Perform base writeback before the loaded value to | |
8737 | * ensure correct behavior with overlapping index registers. | |
8738 | */ | |
8739 | op_addr_ri_post(s, a, addr, 0); | |
8740 | store_reg_from_load(s, a->rt, tmp); | |
8741 | return true; | |
8742 | } | |
8743 | ||
8744 | static bool op_store_ri(DisasContext *s, arg_ldst_ri *a, | |
8745 | MemOp mop, int mem_idx) | |
8746 | { | |
8747 | ISSInfo issinfo = make_issinfo(s, a->rt, a->p, a->w) | ISSIsWrite; | |
8748 | TCGv_i32 addr, tmp; | |
8749 | ||
8750 | addr = op_addr_ri_pre(s, a); | |
8751 | ||
8752 | tmp = load_reg(s, a->rt); | |
8753 | gen_aa32_st_i32(s, tmp, addr, mem_idx, mop | s->be_data); | |
8754 | disas_set_da_iss(s, mop, issinfo); | |
8755 | tcg_temp_free_i32(tmp); | |
8756 | ||
8757 | op_addr_ri_post(s, a, addr, 0); | |
8758 | return true; | |
8759 | } | |
8760 | ||
8761 | static bool op_ldrd_ri(DisasContext *s, arg_ldst_ri *a, int rt2) | |
8762 | { | |
8763 | int mem_idx = get_mem_index(s); | |
8764 | TCGv_i32 addr, tmp; | |
8765 | ||
8766 | addr = op_addr_ri_pre(s, a); | |
8767 | ||
8768 | tmp = tcg_temp_new_i32(); | |
8769 | gen_aa32_ld_i32(s, tmp, addr, mem_idx, MO_UL | s->be_data); | |
8770 | store_reg(s, a->rt, tmp); | |
8771 | ||
8772 | tcg_gen_addi_i32(addr, addr, 4); | |
8773 | ||
8774 | tmp = tcg_temp_new_i32(); | |
8775 | gen_aa32_ld_i32(s, tmp, addr, mem_idx, MO_UL | s->be_data); | |
8776 | store_reg(s, rt2, tmp); | |
8777 | ||
8778 | /* LDRD w/ base writeback is undefined if the registers overlap. */ | |
8779 | op_addr_ri_post(s, a, addr, -4); | |
8780 | return true; | |
8781 | } | |
8782 | ||
8783 | static bool trans_LDRD_ri_a32(DisasContext *s, arg_ldst_ri *a) | |
8784 | { | |
8785 | if (!ENABLE_ARCH_5TE || (a->rt & 1)) { | |
8786 | return false; | |
8787 | } | |
8788 | return op_ldrd_ri(s, a, a->rt + 1); | |
8789 | } | |
8790 | ||
8791 | static bool trans_LDRD_ri_t32(DisasContext *s, arg_ldst_ri2 *a) | |
8792 | { | |
8793 | arg_ldst_ri b = { | |
8794 | .u = a->u, .w = a->w, .p = a->p, | |
8795 | .rn = a->rn, .rt = a->rt, .imm = a->imm | |
8796 | }; | |
8797 | return op_ldrd_ri(s, &b, a->rt2); | |
8798 | } | |
8799 | ||
8800 | static bool op_strd_ri(DisasContext *s, arg_ldst_ri *a, int rt2) | |
8801 | { | |
8802 | int mem_idx = get_mem_index(s); | |
8803 | TCGv_i32 addr, tmp; | |
8804 | ||
8805 | addr = op_addr_ri_pre(s, a); | |
8806 | ||
8807 | tmp = load_reg(s, a->rt); | |
8808 | gen_aa32_st_i32(s, tmp, addr, mem_idx, MO_UL | s->be_data); | |
8809 | tcg_temp_free_i32(tmp); | |
8810 | ||
8811 | tcg_gen_addi_i32(addr, addr, 4); | |
8812 | ||
8813 | tmp = load_reg(s, rt2); | |
8814 | gen_aa32_st_i32(s, tmp, addr, mem_idx, MO_UL | s->be_data); | |
8815 | tcg_temp_free_i32(tmp); | |
8816 | ||
8817 | op_addr_ri_post(s, a, addr, -4); | |
8818 | return true; | |
8819 | } | |
8820 | ||
8821 | static bool trans_STRD_ri_a32(DisasContext *s, arg_ldst_ri *a) | |
8822 | { | |
8823 | if (!ENABLE_ARCH_5TE || (a->rt & 1)) { | |
8824 | return false; | |
8825 | } | |
8826 | return op_strd_ri(s, a, a->rt + 1); | |
8827 | } | |
8828 | ||
8829 | static bool trans_STRD_ri_t32(DisasContext *s, arg_ldst_ri2 *a) | |
8830 | { | |
8831 | arg_ldst_ri b = { | |
8832 | .u = a->u, .w = a->w, .p = a->p, | |
8833 | .rn = a->rn, .rt = a->rt, .imm = a->imm | |
8834 | }; | |
8835 | return op_strd_ri(s, &b, a->rt2); | |
8836 | } | |
8837 | ||
8838 | #define DO_LDST(NAME, WHICH, MEMOP) \ | |
8839 | static bool trans_##NAME##_ri(DisasContext *s, arg_ldst_ri *a) \ | |
8840 | { \ | |
8841 | return op_##WHICH##_ri(s, a, MEMOP, get_mem_index(s)); \ | |
8842 | } \ | |
8843 | static bool trans_##NAME##T_ri(DisasContext *s, arg_ldst_ri *a) \ | |
8844 | { \ | |
8845 | return op_##WHICH##_ri(s, a, MEMOP, get_a32_user_mem_index(s)); \ | |
8846 | } \ | |
8847 | static bool trans_##NAME##_rr(DisasContext *s, arg_ldst_rr *a) \ | |
8848 | { \ | |
8849 | return op_##WHICH##_rr(s, a, MEMOP, get_mem_index(s)); \ | |
8850 | } \ | |
8851 | static bool trans_##NAME##T_rr(DisasContext *s, arg_ldst_rr *a) \ | |
8852 | { \ | |
8853 | return op_##WHICH##_rr(s, a, MEMOP, get_a32_user_mem_index(s)); \ | |
8854 | } | |
8855 | ||
8856 | DO_LDST(LDR, load, MO_UL) | |
8857 | DO_LDST(LDRB, load, MO_UB) | |
8858 | DO_LDST(LDRH, load, MO_UW) | |
8859 | DO_LDST(LDRSB, load, MO_SB) | |
8860 | DO_LDST(LDRSH, load, MO_SW) | |
8861 | ||
8862 | DO_LDST(STR, store, MO_UL) | |
8863 | DO_LDST(STRB, store, MO_UB) | |
8864 | DO_LDST(STRH, store, MO_UW) | |
8865 | ||
8866 | #undef DO_LDST | |
8867 | ||
1efdd407 RH |
8868 | /* |
8869 | * Synchronization primitives | |
8870 | */ | |
8871 | ||
8872 | static bool op_swp(DisasContext *s, arg_SWP *a, MemOp opc) | |
8873 | { | |
8874 | TCGv_i32 addr, tmp; | |
8875 | TCGv taddr; | |
8876 | ||
8877 | opc |= s->be_data; | |
8878 | addr = load_reg(s, a->rn); | |
8879 | taddr = gen_aa32_addr(s, addr, opc); | |
8880 | tcg_temp_free_i32(addr); | |
8881 | ||
8882 | tmp = load_reg(s, a->rt2); | |
8883 | tcg_gen_atomic_xchg_i32(tmp, taddr, tmp, get_mem_index(s), opc); | |
8884 | tcg_temp_free(taddr); | |
8885 | ||
8886 | store_reg(s, a->rt, tmp); | |
8887 | return true; | |
8888 | } | |
8889 | ||
8890 | static bool trans_SWP(DisasContext *s, arg_SWP *a) | |
8891 | { | |
8892 | return op_swp(s, a, MO_UL | MO_ALIGN); | |
8893 | } | |
8894 | ||
8895 | static bool trans_SWPB(DisasContext *s, arg_SWP *a) | |
8896 | { | |
8897 | return op_swp(s, a, MO_UB); | |
8898 | } | |
8899 | ||
8900 | /* | |
8901 | * Load/Store Exclusive and Load-Acquire/Store-Release | |
8902 | */ | |
8903 | ||
8904 | static bool op_strex(DisasContext *s, arg_STREX *a, MemOp mop, bool rel) | |
8905 | { | |
8906 | TCGv_i32 addr; | |
8907 | ||
af288228 RH |
8908 | /* We UNDEF for these UNPREDICTABLE cases. */ |
8909 | if (a->rd == 15 || a->rn == 15 || a->rt == 15 | |
8910 | || a->rd == a->rn || a->rd == a->rt | |
8911 | || (s->thumb && (a->rd == 13 || a->rt == 13)) | |
8912 | || (mop == MO_64 | |
8913 | && (a->rt2 == 15 | |
8914 | || a->rd == a->rt2 || a->rt == a->rt2 | |
8915 | || (s->thumb && a->rt2 == 13)))) { | |
8916 | unallocated_encoding(s); | |
8917 | return true; | |
8918 | } | |
8919 | ||
1efdd407 RH |
8920 | if (rel) { |
8921 | tcg_gen_mb(TCG_MO_ALL | TCG_BAR_STRL); | |
8922 | } | |
8923 | ||
8924 | addr = tcg_temp_local_new_i32(); | |
8925 | load_reg_var(s, addr, a->rn); | |
8926 | tcg_gen_addi_i32(addr, addr, a->imm); | |
8927 | ||
8928 | gen_store_exclusive(s, a->rd, a->rt, a->rt2, addr, mop); | |
8929 | tcg_temp_free_i32(addr); | |
8930 | return true; | |
8931 | } | |
8932 | ||
8933 | static bool trans_STREX(DisasContext *s, arg_STREX *a) | |
8934 | { | |
8935 | if (!ENABLE_ARCH_6) { | |
8936 | return false; | |
8937 | } | |
8938 | return op_strex(s, a, MO_32, false); | |
8939 | } | |
8940 | ||
8941 | static bool trans_STREXD_a32(DisasContext *s, arg_STREX *a) | |
8942 | { | |
8943 | if (!ENABLE_ARCH_6K) { | |
8944 | return false; | |
8945 | } | |
af288228 | 8946 | /* We UNDEF for these UNPREDICTABLE cases. */ |
1efdd407 RH |
8947 | if (a->rt & 1) { |
8948 | unallocated_encoding(s); | |
8949 | return true; | |
8950 | } | |
8951 | a->rt2 = a->rt + 1; | |
8952 | return op_strex(s, a, MO_64, false); | |
8953 | } | |
8954 | ||
8955 | static bool trans_STREXD_t32(DisasContext *s, arg_STREX *a) | |
8956 | { | |
8957 | return op_strex(s, a, MO_64, false); | |
8958 | } | |
8959 | ||
8960 | static bool trans_STREXB(DisasContext *s, arg_STREX *a) | |
8961 | { | |
8962 | if (s->thumb ? !ENABLE_ARCH_7 : !ENABLE_ARCH_6K) { | |
8963 | return false; | |
8964 | } | |
8965 | return op_strex(s, a, MO_8, false); | |
8966 | } | |
8967 | ||
8968 | static bool trans_STREXH(DisasContext *s, arg_STREX *a) | |
8969 | { | |
8970 | if (s->thumb ? !ENABLE_ARCH_7 : !ENABLE_ARCH_6K) { | |
8971 | return false; | |
8972 | } | |
8973 | return op_strex(s, a, MO_16, false); | |
8974 | } | |
8975 | ||
8976 | static bool trans_STLEX(DisasContext *s, arg_STREX *a) | |
8977 | { | |
8978 | if (!ENABLE_ARCH_8) { | |
8979 | return false; | |
8980 | } | |
8981 | return op_strex(s, a, MO_32, true); | |
8982 | } | |
8983 | ||
8984 | static bool trans_STLEXD_a32(DisasContext *s, arg_STREX *a) | |
8985 | { | |
8986 | if (!ENABLE_ARCH_8) { | |
8987 | return false; | |
8988 | } | |
af288228 | 8989 | /* We UNDEF for these UNPREDICTABLE cases. */ |
1efdd407 RH |
8990 | if (a->rt & 1) { |
8991 | unallocated_encoding(s); | |
8992 | return true; | |
8993 | } | |
8994 | a->rt2 = a->rt + 1; | |
8995 | return op_strex(s, a, MO_64, true); | |
8996 | } | |
8997 | ||
8998 | static bool trans_STLEXD_t32(DisasContext *s, arg_STREX *a) | |
8999 | { | |
9000 | if (!ENABLE_ARCH_8) { | |
9001 | return false; | |
9002 | } | |
9003 | return op_strex(s, a, MO_64, true); | |
9004 | } | |
9005 | ||
9006 | static bool trans_STLEXB(DisasContext *s, arg_STREX *a) | |
9007 | { | |
9008 | if (!ENABLE_ARCH_8) { | |
9009 | return false; | |
9010 | } | |
9011 | return op_strex(s, a, MO_8, true); | |
9012 | } | |
9013 | ||
9014 | static bool trans_STLEXH(DisasContext *s, arg_STREX *a) | |
9015 | { | |
9016 | if (!ENABLE_ARCH_8) { | |
9017 | return false; | |
9018 | } | |
9019 | return op_strex(s, a, MO_16, true); | |
9020 | } | |
9021 | ||
9022 | static bool op_stl(DisasContext *s, arg_STL *a, MemOp mop) | |
9023 | { | |
9024 | TCGv_i32 addr, tmp; | |
9025 | ||
9026 | if (!ENABLE_ARCH_8) { | |
9027 | return false; | |
9028 | } | |
af288228 RH |
9029 | /* We UNDEF for these UNPREDICTABLE cases. */ |
9030 | if (a->rn == 15 || a->rt == 15) { | |
9031 | unallocated_encoding(s); | |
9032 | return true; | |
9033 | } | |
1efdd407 | 9034 | |
af288228 | 9035 | addr = load_reg(s, a->rn); |
1efdd407 RH |
9036 | tmp = load_reg(s, a->rt); |
9037 | tcg_gen_mb(TCG_MO_ALL | TCG_BAR_STRL); | |
9038 | gen_aa32_st_i32(s, tmp, addr, get_mem_index(s), mop | s->be_data); | |
9039 | disas_set_da_iss(s, mop, a->rt | ISSIsAcqRel | ISSIsWrite); | |
9040 | ||
9041 | tcg_temp_free_i32(tmp); | |
9042 | tcg_temp_free_i32(addr); | |
9043 | return true; | |
9044 | } | |
9045 | ||
9046 | static bool trans_STL(DisasContext *s, arg_STL *a) | |
9047 | { | |
9048 | return op_stl(s, a, MO_UL); | |
9049 | } | |
9050 | ||
9051 | static bool trans_STLB(DisasContext *s, arg_STL *a) | |
9052 | { | |
9053 | return op_stl(s, a, MO_UB); | |
9054 | } | |
9055 | ||
9056 | static bool trans_STLH(DisasContext *s, arg_STL *a) | |
9057 | { | |
9058 | return op_stl(s, a, MO_UW); | |
9059 | } | |
9060 | ||
9061 | static bool op_ldrex(DisasContext *s, arg_LDREX *a, MemOp mop, bool acq) | |
9062 | { | |
9063 | TCGv_i32 addr; | |
9064 | ||
af288228 RH |
9065 | /* We UNDEF for these UNPREDICTABLE cases. */ |
9066 | if (a->rn == 15 || a->rt == 15 | |
9067 | || (s->thumb && a->rt == 13) | |
9068 | || (mop == MO_64 | |
9069 | && (a->rt2 == 15 || a->rt == a->rt2 | |
9070 | || (s->thumb && a->rt2 == 13)))) { | |
9071 | unallocated_encoding(s); | |
9072 | return true; | |
9073 | } | |
9074 | ||
1efdd407 RH |
9075 | addr = tcg_temp_local_new_i32(); |
9076 | load_reg_var(s, addr, a->rn); | |
9077 | tcg_gen_addi_i32(addr, addr, a->imm); | |
9078 | ||
9079 | gen_load_exclusive(s, a->rt, a->rt2, addr, mop); | |
9080 | tcg_temp_free_i32(addr); | |
9081 | ||
9082 | if (acq) { | |
9083 | tcg_gen_mb(TCG_MO_ALL | TCG_BAR_LDAQ); | |
9084 | } | |
9085 | return true; | |
9086 | } | |
9087 | ||
9088 | static bool trans_LDREX(DisasContext *s, arg_LDREX *a) | |
9089 | { | |
9090 | if (!ENABLE_ARCH_6) { | |
9091 | return false; | |
9092 | } | |
9093 | return op_ldrex(s, a, MO_32, false); | |
9094 | } | |
9095 | ||
9096 | static bool trans_LDREXD_a32(DisasContext *s, arg_LDREX *a) | |
9097 | { | |
9098 | if (!ENABLE_ARCH_6K) { | |
9099 | return false; | |
9100 | } | |
af288228 | 9101 | /* We UNDEF for these UNPREDICTABLE cases. */ |
1efdd407 RH |
9102 | if (a->rt & 1) { |
9103 | unallocated_encoding(s); | |
9104 | return true; | |
9105 | } | |
9106 | a->rt2 = a->rt + 1; | |
9107 | return op_ldrex(s, a, MO_64, false); | |
9108 | } | |
9109 | ||
9110 | static bool trans_LDREXD_t32(DisasContext *s, arg_LDREX *a) | |
9111 | { | |
9112 | return op_ldrex(s, a, MO_64, false); | |
9113 | } | |
9114 | ||
9115 | static bool trans_LDREXB(DisasContext *s, arg_LDREX *a) | |
9116 | { | |
9117 | if (s->thumb ? !ENABLE_ARCH_7 : !ENABLE_ARCH_6K) { | |
9118 | return false; | |
9119 | } | |
9120 | return op_ldrex(s, a, MO_8, false); | |
9121 | } | |
9122 | ||
9123 | static bool trans_LDREXH(DisasContext *s, arg_LDREX *a) | |
9124 | { | |
9125 | if (s->thumb ? !ENABLE_ARCH_7 : !ENABLE_ARCH_6K) { | |
9126 | return false; | |
9127 | } | |
9128 | return op_ldrex(s, a, MO_16, false); | |
9129 | } | |
9130 | ||
9131 | static bool trans_LDAEX(DisasContext *s, arg_LDREX *a) | |
9132 | { | |
9133 | if (!ENABLE_ARCH_8) { | |
9134 | return false; | |
9135 | } | |
9136 | return op_ldrex(s, a, MO_32, true); | |
9137 | } | |
9138 | ||
9139 | static bool trans_LDAEXD_a32(DisasContext *s, arg_LDREX *a) | |
9140 | { | |
9141 | if (!ENABLE_ARCH_8) { | |
9142 | return false; | |
9143 | } | |
af288228 | 9144 | /* We UNDEF for these UNPREDICTABLE cases. */ |
1efdd407 RH |
9145 | if (a->rt & 1) { |
9146 | unallocated_encoding(s); | |
9147 | return true; | |
9148 | } | |
9149 | a->rt2 = a->rt + 1; | |
9150 | return op_ldrex(s, a, MO_64, true); | |
9151 | } | |
9152 | ||
9153 | static bool trans_LDAEXD_t32(DisasContext *s, arg_LDREX *a) | |
9154 | { | |
9155 | if (!ENABLE_ARCH_8) { | |
9156 | return false; | |
9157 | } | |
9158 | return op_ldrex(s, a, MO_64, true); | |
9159 | } | |
9160 | ||
9161 | static bool trans_LDAEXB(DisasContext *s, arg_LDREX *a) | |
9162 | { | |
9163 | if (!ENABLE_ARCH_8) { | |
9164 | return false; | |
9165 | } | |
9166 | return op_ldrex(s, a, MO_8, true); | |
9167 | } | |
9168 | ||
9169 | static bool trans_LDAEXH(DisasContext *s, arg_LDREX *a) | |
9170 | { | |
9171 | if (!ENABLE_ARCH_8) { | |
9172 | return false; | |
9173 | } | |
9174 | return op_ldrex(s, a, MO_16, true); | |
9175 | } | |
9176 | ||
9177 | static bool op_lda(DisasContext *s, arg_LDA *a, MemOp mop) | |
9178 | { | |
9179 | TCGv_i32 addr, tmp; | |
9180 | ||
9181 | if (!ENABLE_ARCH_8) { | |
9182 | return false; | |
9183 | } | |
af288228 RH |
9184 | /* We UNDEF for these UNPREDICTABLE cases. */ |
9185 | if (a->rn == 15 || a->rt == 15) { | |
9186 | unallocated_encoding(s); | |
9187 | return true; | |
9188 | } | |
1efdd407 | 9189 | |
af288228 | 9190 | addr = load_reg(s, a->rn); |
1efdd407 RH |
9191 | tmp = tcg_temp_new_i32(); |
9192 | gen_aa32_ld_i32(s, tmp, addr, get_mem_index(s), mop | s->be_data); | |
9193 | disas_set_da_iss(s, mop, a->rt | ISSIsAcqRel); | |
9194 | tcg_temp_free_i32(addr); | |
9195 | ||
9196 | store_reg(s, a->rt, tmp); | |
9197 | tcg_gen_mb(TCG_MO_ALL | TCG_BAR_STRL); | |
9198 | return true; | |
9199 | } | |
9200 | ||
9201 | static bool trans_LDA(DisasContext *s, arg_LDA *a) | |
9202 | { | |
9203 | return op_lda(s, a, MO_UL); | |
9204 | } | |
9205 | ||
9206 | static bool trans_LDAB(DisasContext *s, arg_LDA *a) | |
9207 | { | |
9208 | return op_lda(s, a, MO_UB); | |
9209 | } | |
9210 | ||
9211 | static bool trans_LDAH(DisasContext *s, arg_LDA *a) | |
9212 | { | |
9213 | return op_lda(s, a, MO_UW); | |
9214 | } | |
9215 | ||
86d21e4b RH |
9216 | /* |
9217 | * Media instructions | |
9218 | */ | |
9219 | ||
9220 | static bool trans_USADA8(DisasContext *s, arg_USADA8 *a) | |
9221 | { | |
9222 | TCGv_i32 t1, t2; | |
9223 | ||
9224 | if (!ENABLE_ARCH_6) { | |
9225 | return false; | |
9226 | } | |
9227 | ||
9228 | t1 = load_reg(s, a->rn); | |
9229 | t2 = load_reg(s, a->rm); | |
9230 | gen_helper_usad8(t1, t1, t2); | |
9231 | tcg_temp_free_i32(t2); | |
9232 | if (a->ra != 15) { | |
9233 | t2 = load_reg(s, a->ra); | |
9234 | tcg_gen_add_i32(t1, t1, t2); | |
9235 | tcg_temp_free_i32(t2); | |
9236 | } | |
9237 | store_reg(s, a->rd, t1); | |
9238 | return true; | |
9239 | } | |
9240 | ||
9241 | static bool op_bfx(DisasContext *s, arg_UBFX *a, bool u) | |
9242 | { | |
9243 | TCGv_i32 tmp; | |
9244 | int width = a->widthm1 + 1; | |
9245 | int shift = a->lsb; | |
9246 | ||
9247 | if (!ENABLE_ARCH_6T2) { | |
9248 | return false; | |
9249 | } | |
9250 | if (shift + width > 32) { | |
9251 | /* UNPREDICTABLE; we choose to UNDEF */ | |
9252 | unallocated_encoding(s); | |
9253 | return true; | |
9254 | } | |
9255 | ||
9256 | tmp = load_reg(s, a->rn); | |
9257 | if (u) { | |
9258 | tcg_gen_extract_i32(tmp, tmp, shift, width); | |
9259 | } else { | |
9260 | tcg_gen_sextract_i32(tmp, tmp, shift, width); | |
9261 | } | |
9262 | store_reg(s, a->rd, tmp); | |
9263 | return true; | |
9264 | } | |
9265 | ||
9266 | static bool trans_SBFX(DisasContext *s, arg_SBFX *a) | |
9267 | { | |
9268 | return op_bfx(s, a, false); | |
9269 | } | |
9270 | ||
9271 | static bool trans_UBFX(DisasContext *s, arg_UBFX *a) | |
9272 | { | |
9273 | return op_bfx(s, a, true); | |
9274 | } | |
9275 | ||
9276 | static bool trans_BFCI(DisasContext *s, arg_BFCI *a) | |
9277 | { | |
9278 | TCGv_i32 tmp; | |
9279 | int msb = a->msb, lsb = a->lsb; | |
9280 | int width; | |
9281 | ||
9282 | if (!ENABLE_ARCH_6T2) { | |
9283 | return false; | |
9284 | } | |
9285 | if (msb < lsb) { | |
9286 | /* UNPREDICTABLE; we choose to UNDEF */ | |
9287 | unallocated_encoding(s); | |
9288 | return true; | |
9289 | } | |
9290 | ||
9291 | width = msb + 1 - lsb; | |
9292 | if (a->rn == 15) { | |
9293 | /* BFC */ | |
9294 | tmp = tcg_const_i32(0); | |
9295 | } else { | |
9296 | /* BFI */ | |
9297 | tmp = load_reg(s, a->rn); | |
9298 | } | |
9299 | if (width != 32) { | |
9300 | TCGv_i32 tmp2 = load_reg(s, a->rd); | |
9301 | tcg_gen_deposit_i32(tmp, tmp2, tmp, lsb, width); | |
9302 | tcg_temp_free_i32(tmp2); | |
9303 | } | |
9304 | store_reg(s, a->rd, tmp); | |
9305 | return true; | |
9306 | } | |
9307 | ||
9308 | static bool trans_UDF(DisasContext *s, arg_UDF *a) | |
9309 | { | |
9310 | unallocated_encoding(s); | |
9311 | return true; | |
9312 | } | |
9313 | ||
adf1a566 RH |
9314 | /* |
9315 | * Parallel addition and subtraction | |
9316 | */ | |
9317 | ||
9318 | static bool op_par_addsub(DisasContext *s, arg_rrr *a, | |
9319 | void (*gen)(TCGv_i32, TCGv_i32, TCGv_i32)) | |
9320 | { | |
9321 | TCGv_i32 t0, t1; | |
9322 | ||
9323 | if (s->thumb | |
9324 | ? !arm_dc_feature(s, ARM_FEATURE_THUMB_DSP) | |
9325 | : !ENABLE_ARCH_6) { | |
9326 | return false; | |
9327 | } | |
9328 | ||
9329 | t0 = load_reg(s, a->rn); | |
9330 | t1 = load_reg(s, a->rm); | |
9331 | ||
9332 | gen(t0, t0, t1); | |
9333 | ||
9334 | tcg_temp_free_i32(t1); | |
9335 | store_reg(s, a->rd, t0); | |
9336 | return true; | |
9337 | } | |
9338 | ||
9339 | static bool op_par_addsub_ge(DisasContext *s, arg_rrr *a, | |
9340 | void (*gen)(TCGv_i32, TCGv_i32, | |
9341 | TCGv_i32, TCGv_ptr)) | |
9342 | { | |
9343 | TCGv_i32 t0, t1; | |
9344 | TCGv_ptr ge; | |
9345 | ||
9346 | if (s->thumb | |
9347 | ? !arm_dc_feature(s, ARM_FEATURE_THUMB_DSP) | |
9348 | : !ENABLE_ARCH_6) { | |
9349 | return false; | |
9350 | } | |
9351 | ||
9352 | t0 = load_reg(s, a->rn); | |
9353 | t1 = load_reg(s, a->rm); | |
9354 | ||
9355 | ge = tcg_temp_new_ptr(); | |
9356 | tcg_gen_addi_ptr(ge, cpu_env, offsetof(CPUARMState, GE)); | |
9357 | gen(t0, t0, t1, ge); | |
9358 | ||
9359 | tcg_temp_free_ptr(ge); | |
9360 | tcg_temp_free_i32(t1); | |
9361 | store_reg(s, a->rd, t0); | |
9362 | return true; | |
9363 | } | |
9364 | ||
9365 | #define DO_PAR_ADDSUB(NAME, helper) \ | |
9366 | static bool trans_##NAME(DisasContext *s, arg_rrr *a) \ | |
9367 | { \ | |
9368 | return op_par_addsub(s, a, helper); \ | |
9369 | } | |
9370 | ||
9371 | #define DO_PAR_ADDSUB_GE(NAME, helper) \ | |
9372 | static bool trans_##NAME(DisasContext *s, arg_rrr *a) \ | |
9373 | { \ | |
9374 | return op_par_addsub_ge(s, a, helper); \ | |
9375 | } | |
9376 | ||
9377 | DO_PAR_ADDSUB_GE(SADD16, gen_helper_sadd16) | |
9378 | DO_PAR_ADDSUB_GE(SASX, gen_helper_saddsubx) | |
9379 | DO_PAR_ADDSUB_GE(SSAX, gen_helper_ssubaddx) | |
9380 | DO_PAR_ADDSUB_GE(SSUB16, gen_helper_ssub16) | |
9381 | DO_PAR_ADDSUB_GE(SADD8, gen_helper_sadd8) | |
9382 | DO_PAR_ADDSUB_GE(SSUB8, gen_helper_ssub8) | |
9383 | ||
9384 | DO_PAR_ADDSUB_GE(UADD16, gen_helper_uadd16) | |
9385 | DO_PAR_ADDSUB_GE(UASX, gen_helper_uaddsubx) | |
9386 | DO_PAR_ADDSUB_GE(USAX, gen_helper_usubaddx) | |
9387 | DO_PAR_ADDSUB_GE(USUB16, gen_helper_usub16) | |
9388 | DO_PAR_ADDSUB_GE(UADD8, gen_helper_uadd8) | |
9389 | DO_PAR_ADDSUB_GE(USUB8, gen_helper_usub8) | |
9390 | ||
9391 | DO_PAR_ADDSUB(QADD16, gen_helper_qadd16) | |
9392 | DO_PAR_ADDSUB(QASX, gen_helper_qaddsubx) | |
9393 | DO_PAR_ADDSUB(QSAX, gen_helper_qsubaddx) | |
9394 | DO_PAR_ADDSUB(QSUB16, gen_helper_qsub16) | |
9395 | DO_PAR_ADDSUB(QADD8, gen_helper_qadd8) | |
9396 | DO_PAR_ADDSUB(QSUB8, gen_helper_qsub8) | |
9397 | ||
9398 | DO_PAR_ADDSUB(UQADD16, gen_helper_uqadd16) | |
9399 | DO_PAR_ADDSUB(UQASX, gen_helper_uqaddsubx) | |
9400 | DO_PAR_ADDSUB(UQSAX, gen_helper_uqsubaddx) | |
9401 | DO_PAR_ADDSUB(UQSUB16, gen_helper_uqsub16) | |
9402 | DO_PAR_ADDSUB(UQADD8, gen_helper_uqadd8) | |
9403 | DO_PAR_ADDSUB(UQSUB8, gen_helper_uqsub8) | |
9404 | ||
9405 | DO_PAR_ADDSUB(SHADD16, gen_helper_shadd16) | |
9406 | DO_PAR_ADDSUB(SHASX, gen_helper_shaddsubx) | |
9407 | DO_PAR_ADDSUB(SHSAX, gen_helper_shsubaddx) | |
9408 | DO_PAR_ADDSUB(SHSUB16, gen_helper_shsub16) | |
9409 | DO_PAR_ADDSUB(SHADD8, gen_helper_shadd8) | |
9410 | DO_PAR_ADDSUB(SHSUB8, gen_helper_shsub8) | |
9411 | ||
9412 | DO_PAR_ADDSUB(UHADD16, gen_helper_uhadd16) | |
9413 | DO_PAR_ADDSUB(UHASX, gen_helper_uhaddsubx) | |
9414 | DO_PAR_ADDSUB(UHSAX, gen_helper_uhsubaddx) | |
9415 | DO_PAR_ADDSUB(UHSUB16, gen_helper_uhsub16) | |
9416 | DO_PAR_ADDSUB(UHADD8, gen_helper_uhadd8) | |
9417 | DO_PAR_ADDSUB(UHSUB8, gen_helper_uhsub8) | |
9418 | ||
9419 | #undef DO_PAR_ADDSUB | |
9420 | #undef DO_PAR_ADDSUB_GE | |
9421 | ||
46497f6a RH |
9422 | /* |
9423 | * Packing, unpacking, saturation, and reversal | |
9424 | */ | |
9425 | ||
9426 | static bool trans_PKH(DisasContext *s, arg_PKH *a) | |
9427 | { | |
9428 | TCGv_i32 tn, tm; | |
9429 | int shift = a->imm; | |
9430 | ||
9431 | if (s->thumb | |
9432 | ? !arm_dc_feature(s, ARM_FEATURE_THUMB_DSP) | |
9433 | : !ENABLE_ARCH_6) { | |
9434 | return false; | |
9435 | } | |
9436 | ||
9437 | tn = load_reg(s, a->rn); | |
9438 | tm = load_reg(s, a->rm); | |
9439 | if (a->tb) { | |
9440 | /* PKHTB */ | |
9441 | if (shift == 0) { | |
9442 | shift = 31; | |
9443 | } | |
9444 | tcg_gen_sari_i32(tm, tm, shift); | |
9445 | tcg_gen_deposit_i32(tn, tn, tm, 0, 16); | |
9446 | } else { | |
9447 | /* PKHBT */ | |
9448 | tcg_gen_shli_i32(tm, tm, shift); | |
9449 | tcg_gen_deposit_i32(tn, tm, tn, 0, 16); | |
9450 | } | |
9451 | tcg_temp_free_i32(tm); | |
9452 | store_reg(s, a->rd, tn); | |
9453 | return true; | |
9454 | } | |
9455 | ||
9456 | static bool op_sat(DisasContext *s, arg_sat *a, | |
9457 | void (*gen)(TCGv_i32, TCGv_env, TCGv_i32, TCGv_i32)) | |
9458 | { | |
9459 | TCGv_i32 tmp, satimm; | |
9460 | int shift = a->imm; | |
9461 | ||
9462 | if (!ENABLE_ARCH_6) { | |
9463 | return false; | |
9464 | } | |
9465 | ||
9466 | tmp = load_reg(s, a->rn); | |
9467 | if (a->sh) { | |
9468 | tcg_gen_sari_i32(tmp, tmp, shift ? shift : 31); | |
9469 | } else { | |
9470 | tcg_gen_shli_i32(tmp, tmp, shift); | |
9471 | } | |
9472 | ||
9473 | satimm = tcg_const_i32(a->satimm); | |
9474 | gen(tmp, cpu_env, tmp, satimm); | |
9475 | tcg_temp_free_i32(satimm); | |
9476 | ||
9477 | store_reg(s, a->rd, tmp); | |
9478 | return true; | |
9479 | } | |
9480 | ||
9481 | static bool trans_SSAT(DisasContext *s, arg_sat *a) | |
9482 | { | |
9483 | return op_sat(s, a, gen_helper_ssat); | |
9484 | } | |
9485 | ||
9486 | static bool trans_USAT(DisasContext *s, arg_sat *a) | |
9487 | { | |
9488 | return op_sat(s, a, gen_helper_usat); | |
9489 | } | |
9490 | ||
9491 | static bool trans_SSAT16(DisasContext *s, arg_sat *a) | |
9492 | { | |
9493 | if (s->thumb && !arm_dc_feature(s, ARM_FEATURE_THUMB_DSP)) { | |
9494 | return false; | |
9495 | } | |
9496 | return op_sat(s, a, gen_helper_ssat16); | |
9497 | } | |
9498 | ||
9499 | static bool trans_USAT16(DisasContext *s, arg_sat *a) | |
9500 | { | |
9501 | if (s->thumb && !arm_dc_feature(s, ARM_FEATURE_THUMB_DSP)) { | |
9502 | return false; | |
9503 | } | |
9504 | return op_sat(s, a, gen_helper_usat16); | |
9505 | } | |
9506 | ||
9507 | static bool op_xta(DisasContext *s, arg_rrr_rot *a, | |
9508 | void (*gen_extract)(TCGv_i32, TCGv_i32), | |
9509 | void (*gen_add)(TCGv_i32, TCGv_i32, TCGv_i32)) | |
9510 | { | |
9511 | TCGv_i32 tmp; | |
9512 | ||
9513 | if (!ENABLE_ARCH_6) { | |
9514 | return false; | |
9515 | } | |
9516 | ||
9517 | tmp = load_reg(s, a->rm); | |
9518 | /* | |
9519 | * TODO: In many cases we could do a shift instead of a rotate. | |
9520 | * Combined with a simple extend, that becomes an extract. | |
9521 | */ | |
9522 | tcg_gen_rotri_i32(tmp, tmp, a->rot * 8); | |
9523 | gen_extract(tmp, tmp); | |
9524 | ||
9525 | if (a->rn != 15) { | |
9526 | TCGv_i32 tmp2 = load_reg(s, a->rn); | |
9527 | gen_add(tmp, tmp, tmp2); | |
9528 | tcg_temp_free_i32(tmp2); | |
9529 | } | |
9530 | store_reg(s, a->rd, tmp); | |
9531 | return true; | |
9532 | } | |
9533 | ||
9534 | static bool trans_SXTAB(DisasContext *s, arg_rrr_rot *a) | |
9535 | { | |
9536 | return op_xta(s, a, tcg_gen_ext8s_i32, tcg_gen_add_i32); | |
9537 | } | |
9538 | ||
9539 | static bool trans_SXTAH(DisasContext *s, arg_rrr_rot *a) | |
9540 | { | |
9541 | return op_xta(s, a, tcg_gen_ext16s_i32, tcg_gen_add_i32); | |
9542 | } | |
9543 | ||
9544 | static bool trans_SXTAB16(DisasContext *s, arg_rrr_rot *a) | |
9545 | { | |
9546 | if (s->thumb && !arm_dc_feature(s, ARM_FEATURE_THUMB_DSP)) { | |
9547 | return false; | |
9548 | } | |
9549 | return op_xta(s, a, gen_helper_sxtb16, gen_add16); | |
9550 | } | |
9551 | ||
9552 | static bool trans_UXTAB(DisasContext *s, arg_rrr_rot *a) | |
9553 | { | |
9554 | return op_xta(s, a, tcg_gen_ext8u_i32, tcg_gen_add_i32); | |
9555 | } | |
9556 | ||
9557 | static bool trans_UXTAH(DisasContext *s, arg_rrr_rot *a) | |
9558 | { | |
9559 | return op_xta(s, a, tcg_gen_ext16u_i32, tcg_gen_add_i32); | |
9560 | } | |
9561 | ||
9562 | static bool trans_UXTAB16(DisasContext *s, arg_rrr_rot *a) | |
9563 | { | |
9564 | if (s->thumb && !arm_dc_feature(s, ARM_FEATURE_THUMB_DSP)) { | |
9565 | return false; | |
9566 | } | |
9567 | return op_xta(s, a, gen_helper_uxtb16, gen_add16); | |
9568 | } | |
9569 | ||
9570 | static bool trans_SEL(DisasContext *s, arg_rrr *a) | |
9571 | { | |
9572 | TCGv_i32 t1, t2, t3; | |
9573 | ||
9574 | if (s->thumb | |
9575 | ? !arm_dc_feature(s, ARM_FEATURE_THUMB_DSP) | |
9576 | : !ENABLE_ARCH_6) { | |
9577 | return false; | |
9578 | } | |
9579 | ||
9580 | t1 = load_reg(s, a->rn); | |
9581 | t2 = load_reg(s, a->rm); | |
9582 | t3 = tcg_temp_new_i32(); | |
9583 | tcg_gen_ld_i32(t3, cpu_env, offsetof(CPUARMState, GE)); | |
9584 | gen_helper_sel_flags(t1, t3, t1, t2); | |
9585 | tcg_temp_free_i32(t3); | |
9586 | tcg_temp_free_i32(t2); | |
9587 | store_reg(s, a->rd, t1); | |
9588 | return true; | |
9589 | } | |
9590 | ||
9591 | static bool op_rr(DisasContext *s, arg_rr *a, | |
9592 | void (*gen)(TCGv_i32, TCGv_i32)) | |
9593 | { | |
9594 | TCGv_i32 tmp; | |
9595 | ||
9596 | tmp = load_reg(s, a->rm); | |
9597 | gen(tmp, tmp); | |
9598 | store_reg(s, a->rd, tmp); | |
9599 | return true; | |
9600 | } | |
9601 | ||
9602 | static bool trans_REV(DisasContext *s, arg_rr *a) | |
9603 | { | |
9604 | if (!ENABLE_ARCH_6) { | |
9605 | return false; | |
9606 | } | |
9607 | return op_rr(s, a, tcg_gen_bswap32_i32); | |
9608 | } | |
9609 | ||
9610 | static bool trans_REV16(DisasContext *s, arg_rr *a) | |
9611 | { | |
9612 | if (!ENABLE_ARCH_6) { | |
9613 | return false; | |
9614 | } | |
9615 | return op_rr(s, a, gen_rev16); | |
9616 | } | |
9617 | ||
9618 | static bool trans_REVSH(DisasContext *s, arg_rr *a) | |
9619 | { | |
9620 | if (!ENABLE_ARCH_6) { | |
9621 | return false; | |
9622 | } | |
9623 | return op_rr(s, a, gen_revsh); | |
9624 | } | |
9625 | ||
9626 | static bool trans_RBIT(DisasContext *s, arg_rr *a) | |
9627 | { | |
9628 | if (!ENABLE_ARCH_6T2) { | |
9629 | return false; | |
9630 | } | |
9631 | return op_rr(s, a, gen_helper_rbit); | |
9632 | } | |
9633 | ||
2c7c4e09 RH |
9634 | /* |
9635 | * Signed multiply, signed and unsigned divide | |
9636 | */ | |
9637 | ||
9638 | static bool op_smlad(DisasContext *s, arg_rrrr *a, bool m_swap, bool sub) | |
9639 | { | |
9640 | TCGv_i32 t1, t2; | |
9641 | ||
9642 | if (!ENABLE_ARCH_6) { | |
9643 | return false; | |
9644 | } | |
9645 | ||
9646 | t1 = load_reg(s, a->rn); | |
9647 | t2 = load_reg(s, a->rm); | |
9648 | if (m_swap) { | |
9649 | gen_swap_half(t2); | |
9650 | } | |
9651 | gen_smul_dual(t1, t2); | |
9652 | ||
9653 | if (sub) { | |
9654 | /* This subtraction cannot overflow. */ | |
9655 | tcg_gen_sub_i32(t1, t1, t2); | |
9656 | } else { | |
9657 | /* | |
9658 | * This addition cannot overflow 32 bits; however it may | |
9659 | * overflow considered as a signed operation, in which case | |
9660 | * we must set the Q flag. | |
9661 | */ | |
9662 | gen_helper_add_setq(t1, cpu_env, t1, t2); | |
9663 | } | |
9664 | tcg_temp_free_i32(t2); | |
9665 | ||
9666 | if (a->ra != 15) { | |
9667 | t2 = load_reg(s, a->ra); | |
9668 | gen_helper_add_setq(t1, cpu_env, t1, t2); | |
9669 | tcg_temp_free_i32(t2); | |
9670 | } | |
9671 | store_reg(s, a->rd, t1); | |
9672 | return true; | |
9673 | } | |
9674 | ||
9675 | static bool trans_SMLAD(DisasContext *s, arg_rrrr *a) | |
9676 | { | |
9677 | return op_smlad(s, a, false, false); | |
9678 | } | |
9679 | ||
9680 | static bool trans_SMLADX(DisasContext *s, arg_rrrr *a) | |
9681 | { | |
9682 | return op_smlad(s, a, true, false); | |
9683 | } | |
9684 | ||
9685 | static bool trans_SMLSD(DisasContext *s, arg_rrrr *a) | |
9686 | { | |
9687 | return op_smlad(s, a, false, true); | |
9688 | } | |
9689 | ||
9690 | static bool trans_SMLSDX(DisasContext *s, arg_rrrr *a) | |
9691 | { | |
9692 | return op_smlad(s, a, true, true); | |
9693 | } | |
9694 | ||
9695 | static bool op_smlald(DisasContext *s, arg_rrrr *a, bool m_swap, bool sub) | |
9696 | { | |
9697 | TCGv_i32 t1, t2; | |
9698 | TCGv_i64 l1, l2; | |
9699 | ||
9700 | if (!ENABLE_ARCH_6) { | |
9701 | return false; | |
9702 | } | |
9703 | ||
9704 | t1 = load_reg(s, a->rn); | |
9705 | t2 = load_reg(s, a->rm); | |
9706 | if (m_swap) { | |
9707 | gen_swap_half(t2); | |
9708 | } | |
9709 | gen_smul_dual(t1, t2); | |
9710 | ||
9711 | l1 = tcg_temp_new_i64(); | |
9712 | l2 = tcg_temp_new_i64(); | |
9713 | tcg_gen_ext_i32_i64(l1, t1); | |
9714 | tcg_gen_ext_i32_i64(l2, t2); | |
9715 | tcg_temp_free_i32(t1); | |
9716 | tcg_temp_free_i32(t2); | |
9717 | ||
9718 | if (sub) { | |
9719 | tcg_gen_sub_i64(l1, l1, l2); | |
9720 | } else { | |
9721 | tcg_gen_add_i64(l1, l1, l2); | |
9722 | } | |
9723 | tcg_temp_free_i64(l2); | |
9724 | ||
9725 | gen_addq(s, l1, a->ra, a->rd); | |
9726 | gen_storeq_reg(s, a->ra, a->rd, l1); | |
9727 | tcg_temp_free_i64(l1); | |
9728 | return true; | |
9729 | } | |
9730 | ||
9731 | static bool trans_SMLALD(DisasContext *s, arg_rrrr *a) | |
9732 | { | |
9733 | return op_smlald(s, a, false, false); | |
9734 | } | |
9735 | ||
9736 | static bool trans_SMLALDX(DisasContext *s, arg_rrrr *a) | |
9737 | { | |
9738 | return op_smlald(s, a, true, false); | |
9739 | } | |
9740 | ||
9741 | static bool trans_SMLSLD(DisasContext *s, arg_rrrr *a) | |
9742 | { | |
9743 | return op_smlald(s, a, false, true); | |
9744 | } | |
9745 | ||
9746 | static bool trans_SMLSLDX(DisasContext *s, arg_rrrr *a) | |
9747 | { | |
9748 | return op_smlald(s, a, true, true); | |
9749 | } | |
9750 | ||
9751 | static bool op_smmla(DisasContext *s, arg_rrrr *a, bool round, bool sub) | |
9752 | { | |
9753 | TCGv_i32 t1, t2; | |
9754 | ||
9755 | if (s->thumb | |
9756 | ? !arm_dc_feature(s, ARM_FEATURE_THUMB_DSP) | |
9757 | : !ENABLE_ARCH_6) { | |
9758 | return false; | |
9759 | } | |
9760 | ||
9761 | t1 = load_reg(s, a->rn); | |
9762 | t2 = load_reg(s, a->rm); | |
9763 | tcg_gen_muls2_i32(t2, t1, t1, t2); | |
9764 | ||
9765 | if (a->ra != 15) { | |
9766 | TCGv_i32 t3 = load_reg(s, a->ra); | |
9767 | if (sub) { | |
9768 | /* | |
9769 | * For SMMLS, we need a 64-bit subtract. Borrow caused by | |
9770 | * a non-zero multiplicand lowpart, and the correct result | |
9771 | * lowpart for rounding. | |
9772 | */ | |
9773 | TCGv_i32 zero = tcg_const_i32(0); | |
9774 | tcg_gen_sub2_i32(t2, t1, zero, t3, t2, t1); | |
9775 | tcg_temp_free_i32(zero); | |
9776 | } else { | |
9777 | tcg_gen_add_i32(t1, t1, t3); | |
9778 | } | |
9779 | tcg_temp_free_i32(t3); | |
9780 | } | |
9781 | if (round) { | |
9782 | /* | |
9783 | * Adding 0x80000000 to the 64-bit quantity means that we have | |
9784 | * carry in to the high word when the low word has the msb set. | |
9785 | */ | |
9786 | tcg_gen_shri_i32(t2, t2, 31); | |
9787 | tcg_gen_add_i32(t1, t1, t2); | |
9788 | } | |
9789 | tcg_temp_free_i32(t2); | |
9790 | store_reg(s, a->rd, t1); | |
9791 | return true; | |
9792 | } | |
9793 | ||
9794 | static bool trans_SMMLA(DisasContext *s, arg_rrrr *a) | |
9795 | { | |
9796 | return op_smmla(s, a, false, false); | |
9797 | } | |
9798 | ||
9799 | static bool trans_SMMLAR(DisasContext *s, arg_rrrr *a) | |
9800 | { | |
9801 | return op_smmla(s, a, true, false); | |
9802 | } | |
9803 | ||
9804 | static bool trans_SMMLS(DisasContext *s, arg_rrrr *a) | |
9805 | { | |
9806 | return op_smmla(s, a, false, true); | |
9807 | } | |
9808 | ||
9809 | static bool trans_SMMLSR(DisasContext *s, arg_rrrr *a) | |
9810 | { | |
9811 | return op_smmla(s, a, true, true); | |
9812 | } | |
9813 | ||
9814 | static bool op_div(DisasContext *s, arg_rrr *a, bool u) | |
9815 | { | |
9816 | TCGv_i32 t1, t2; | |
9817 | ||
9818 | if (s->thumb | |
9819 | ? !dc_isar_feature(thumb_div, s) | |
9820 | : !dc_isar_feature(arm_div, s)) { | |
9821 | return false; | |
9822 | } | |
9823 | ||
9824 | t1 = load_reg(s, a->rn); | |
9825 | t2 = load_reg(s, a->rm); | |
9826 | if (u) { | |
9827 | gen_helper_udiv(t1, t1, t2); | |
9828 | } else { | |
9829 | gen_helper_sdiv(t1, t1, t2); | |
9830 | } | |
9831 | tcg_temp_free_i32(t2); | |
9832 | store_reg(s, a->rd, t1); | |
9833 | return true; | |
9834 | } | |
9835 | ||
9836 | static bool trans_SDIV(DisasContext *s, arg_rrr *a) | |
9837 | { | |
9838 | return op_div(s, a, false); | |
9839 | } | |
9840 | ||
9841 | static bool trans_UDIV(DisasContext *s, arg_rrr *a) | |
9842 | { | |
9843 | return op_div(s, a, true); | |
9844 | } | |
9845 | ||
c5c426d4 RH |
9846 | /* |
9847 | * Block data transfer | |
9848 | */ | |
9849 | ||
9850 | static TCGv_i32 op_addr_block_pre(DisasContext *s, arg_ldst_block *a, int n) | |
9851 | { | |
9852 | TCGv_i32 addr = load_reg(s, a->rn); | |
9853 | ||
9854 | if (a->b) { | |
9855 | if (a->i) { | |
9856 | /* pre increment */ | |
9857 | tcg_gen_addi_i32(addr, addr, 4); | |
9858 | } else { | |
9859 | /* pre decrement */ | |
9860 | tcg_gen_addi_i32(addr, addr, -(n * 4)); | |
9861 | } | |
9862 | } else if (!a->i && n != 1) { | |
9863 | /* post decrement */ | |
9864 | tcg_gen_addi_i32(addr, addr, -((n - 1) * 4)); | |
9865 | } | |
9866 | ||
9867 | if (s->v8m_stackcheck && a->rn == 13 && a->w) { | |
9868 | /* | |
9869 | * If the writeback is incrementing SP rather than | |
9870 | * decrementing it, and the initial SP is below the | |
9871 | * stack limit but the final written-back SP would | |
9872 | * be above, then then we must not perform any memory | |
9873 | * accesses, but it is IMPDEF whether we generate | |
9874 | * an exception. We choose to do so in this case. | |
9875 | * At this point 'addr' is the lowest address, so | |
9876 | * either the original SP (if incrementing) or our | |
9877 | * final SP (if decrementing), so that's what we check. | |
9878 | */ | |
9879 | gen_helper_v8m_stackcheck(cpu_env, addr); | |
9880 | } | |
9881 | ||
9882 | return addr; | |
9883 | } | |
9884 | ||
9885 | static void op_addr_block_post(DisasContext *s, arg_ldst_block *a, | |
9886 | TCGv_i32 addr, int n) | |
9887 | { | |
9888 | if (a->w) { | |
9889 | /* write back */ | |
9890 | if (!a->b) { | |
9891 | if (a->i) { | |
9892 | /* post increment */ | |
9893 | tcg_gen_addi_i32(addr, addr, 4); | |
9894 | } else { | |
9895 | /* post decrement */ | |
9896 | tcg_gen_addi_i32(addr, addr, -(n * 4)); | |
9897 | } | |
9898 | } else if (!a->i && n != 1) { | |
9899 | /* pre decrement */ | |
9900 | tcg_gen_addi_i32(addr, addr, -((n - 1) * 4)); | |
9901 | } | |
9902 | store_reg(s, a->rn, addr); | |
9903 | } else { | |
9904 | tcg_temp_free_i32(addr); | |
9905 | } | |
9906 | } | |
9907 | ||
4b222545 | 9908 | static bool op_stm(DisasContext *s, arg_ldst_block *a, int min_n) |
c5c426d4 RH |
9909 | { |
9910 | int i, j, n, list, mem_idx; | |
9911 | bool user = a->u; | |
9912 | TCGv_i32 addr, tmp, tmp2; | |
9913 | ||
9914 | if (user) { | |
9915 | /* STM (user) */ | |
9916 | if (IS_USER(s)) { | |
9917 | /* Only usable in supervisor mode. */ | |
9918 | unallocated_encoding(s); | |
9919 | return true; | |
9920 | } | |
9921 | } | |
9922 | ||
9923 | list = a->list; | |
9924 | n = ctpop16(list); | |
b0e382b8 | 9925 | if (n < min_n || a->rn == 15) { |
4b222545 RH |
9926 | unallocated_encoding(s); |
9927 | return true; | |
9928 | } | |
c5c426d4 RH |
9929 | |
9930 | addr = op_addr_block_pre(s, a, n); | |
9931 | mem_idx = get_mem_index(s); | |
9932 | ||
9933 | for (i = j = 0; i < 16; i++) { | |
9934 | if (!(list & (1 << i))) { | |
9935 | continue; | |
9936 | } | |
9937 | ||
9938 | if (user && i != 15) { | |
9939 | tmp = tcg_temp_new_i32(); | |
9940 | tmp2 = tcg_const_i32(i); | |
9941 | gen_helper_get_user_reg(tmp, cpu_env, tmp2); | |
9942 | tcg_temp_free_i32(tmp2); | |
9943 | } else { | |
9944 | tmp = load_reg(s, i); | |
9945 | } | |
9946 | gen_aa32_st32(s, tmp, addr, mem_idx); | |
9947 | tcg_temp_free_i32(tmp); | |
9948 | ||
9949 | /* No need to add after the last transfer. */ | |
9950 | if (++j != n) { | |
9951 | tcg_gen_addi_i32(addr, addr, 4); | |
9952 | } | |
9953 | } | |
9954 | ||
9955 | op_addr_block_post(s, a, addr, n); | |
9956 | return true; | |
9957 | } | |
9958 | ||
9959 | static bool trans_STM(DisasContext *s, arg_ldst_block *a) | |
9960 | { | |
4b222545 RH |
9961 | /* BitCount(list) < 1 is UNPREDICTABLE */ |
9962 | return op_stm(s, a, 1); | |
c5c426d4 RH |
9963 | } |
9964 | ||
9965 | static bool trans_STM_t32(DisasContext *s, arg_ldst_block *a) | |
9966 | { | |
9967 | /* Writeback register in register list is UNPREDICTABLE for T32. */ | |
9968 | if (a->w && (a->list & (1 << a->rn))) { | |
9969 | unallocated_encoding(s); | |
9970 | return true; | |
9971 | } | |
4b222545 RH |
9972 | /* BitCount(list) < 2 is UNPREDICTABLE */ |
9973 | return op_stm(s, a, 2); | |
c5c426d4 RH |
9974 | } |
9975 | ||
4b222545 | 9976 | static bool do_ldm(DisasContext *s, arg_ldst_block *a, int min_n) |
c5c426d4 RH |
9977 | { |
9978 | int i, j, n, list, mem_idx; | |
9979 | bool loaded_base; | |
9980 | bool user = a->u; | |
9981 | bool exc_return = false; | |
9982 | TCGv_i32 addr, tmp, tmp2, loaded_var; | |
9983 | ||
9984 | if (user) { | |
9985 | /* LDM (user), LDM (exception return) */ | |
9986 | if (IS_USER(s)) { | |
9987 | /* Only usable in supervisor mode. */ | |
9988 | unallocated_encoding(s); | |
9989 | return true; | |
9990 | } | |
9991 | if (extract32(a->list, 15, 1)) { | |
9992 | exc_return = true; | |
9993 | user = false; | |
9994 | } else { | |
9995 | /* LDM (user) does not allow writeback. */ | |
9996 | if (a->w) { | |
9997 | unallocated_encoding(s); | |
9998 | return true; | |
9999 | } | |
10000 | } | |
10001 | } | |
10002 | ||
10003 | list = a->list; | |
10004 | n = ctpop16(list); | |
b0e382b8 | 10005 | if (n < min_n || a->rn == 15) { |
4b222545 RH |
10006 | unallocated_encoding(s); |
10007 | return true; | |
10008 | } | |
c5c426d4 RH |
10009 | |
10010 | addr = op_addr_block_pre(s, a, n); | |
10011 | mem_idx = get_mem_index(s); | |
10012 | loaded_base = false; | |
10013 | loaded_var = NULL; | |
10014 | ||
10015 | for (i = j = 0; i < 16; i++) { | |
10016 | if (!(list & (1 << i))) { | |
10017 | continue; | |
10018 | } | |
10019 | ||
10020 | tmp = tcg_temp_new_i32(); | |
10021 | gen_aa32_ld32u(s, tmp, addr, mem_idx); | |
10022 | if (user) { | |
10023 | tmp2 = tcg_const_i32(i); | |
10024 | gen_helper_set_user_reg(cpu_env, tmp2, tmp); | |
10025 | tcg_temp_free_i32(tmp2); | |
10026 | tcg_temp_free_i32(tmp); | |
10027 | } else if (i == a->rn) { | |
10028 | loaded_var = tmp; | |
10029 | loaded_base = true; | |
10030 | } else if (i == 15 && exc_return) { | |
10031 | store_pc_exc_ret(s, tmp); | |
10032 | } else { | |
10033 | store_reg_from_load(s, i, tmp); | |
10034 | } | |
10035 | ||
10036 | /* No need to add after the last transfer. */ | |
10037 | if (++j != n) { | |
10038 | tcg_gen_addi_i32(addr, addr, 4); | |
10039 | } | |
10040 | } | |
10041 | ||
10042 | op_addr_block_post(s, a, addr, n); | |
10043 | ||
10044 | if (loaded_base) { | |
b0e382b8 | 10045 | /* Note that we reject base == pc above. */ |
c5c426d4 RH |
10046 | store_reg(s, a->rn, loaded_var); |
10047 | } | |
10048 | ||
10049 | if (exc_return) { | |
10050 | /* Restore CPSR from SPSR. */ | |
10051 | tmp = load_cpu_field(spsr); | |
10052 | if (tb_cflags(s->base.tb) & CF_USE_ICOUNT) { | |
10053 | gen_io_start(); | |
10054 | } | |
10055 | gen_helper_cpsr_write_eret(cpu_env, tmp); | |
10056 | if (tb_cflags(s->base.tb) & CF_USE_ICOUNT) { | |
10057 | gen_io_end(); | |
10058 | } | |
10059 | tcg_temp_free_i32(tmp); | |
10060 | /* Must exit loop to check un-masked IRQs */ | |
10061 | s->base.is_jmp = DISAS_EXIT; | |
10062 | } | |
10063 | return true; | |
10064 | } | |
10065 | ||
10066 | static bool trans_LDM_a32(DisasContext *s, arg_ldst_block *a) | |
10067 | { | |
3949f467 RH |
10068 | /* |
10069 | * Writeback register in register list is UNPREDICTABLE | |
10070 | * for ArchVersion() >= 7. Prior to v7, A32 would write | |
10071 | * an UNKNOWN value to the base register. | |
10072 | */ | |
10073 | if (ENABLE_ARCH_7 && a->w && (a->list & (1 << a->rn))) { | |
10074 | unallocated_encoding(s); | |
10075 | return true; | |
10076 | } | |
4b222545 RH |
10077 | /* BitCount(list) < 1 is UNPREDICTABLE */ |
10078 | return do_ldm(s, a, 1); | |
c5c426d4 RH |
10079 | } |
10080 | ||
10081 | static bool trans_LDM_t32(DisasContext *s, arg_ldst_block *a) | |
10082 | { | |
10083 | /* Writeback register in register list is UNPREDICTABLE for T32. */ | |
10084 | if (a->w && (a->list & (1 << a->rn))) { | |
10085 | unallocated_encoding(s); | |
10086 | return true; | |
10087 | } | |
4b222545 RH |
10088 | /* BitCount(list) < 2 is UNPREDICTABLE */ |
10089 | return do_ldm(s, a, 2); | |
c5c426d4 RH |
10090 | } |
10091 | ||
6e8514ba RH |
10092 | static bool trans_LDM_t16(DisasContext *s, arg_ldst_block *a) |
10093 | { | |
10094 | /* Writeback is conditional on the base register not being loaded. */ | |
10095 | a->w = !(a->list & (1 << a->rn)); | |
10096 | /* BitCount(list) < 1 is UNPREDICTABLE */ | |
10097 | return do_ldm(s, a, 1); | |
10098 | } | |
10099 | ||
360144f3 RH |
10100 | /* |
10101 | * Branch, branch with link | |
10102 | */ | |
10103 | ||
10104 | static bool trans_B(DisasContext *s, arg_i *a) | |
10105 | { | |
10106 | gen_jmp(s, read_pc(s) + a->imm); | |
10107 | return true; | |
10108 | } | |
10109 | ||
10110 | static bool trans_B_cond_thumb(DisasContext *s, arg_ci *a) | |
10111 | { | |
10112 | /* This has cond from encoding, required to be outside IT block. */ | |
10113 | if (a->cond >= 0xe) { | |
10114 | return false; | |
10115 | } | |
10116 | if (s->condexec_mask) { | |
10117 | unallocated_encoding(s); | |
10118 | return true; | |
10119 | } | |
10120 | arm_skip_unless(s, a->cond); | |
10121 | gen_jmp(s, read_pc(s) + a->imm); | |
10122 | return true; | |
10123 | } | |
10124 | ||
10125 | static bool trans_BL(DisasContext *s, arg_i *a) | |
10126 | { | |
10127 | tcg_gen_movi_i32(cpu_R[14], s->base.pc_next | s->thumb); | |
10128 | gen_jmp(s, read_pc(s) + a->imm); | |
10129 | return true; | |
10130 | } | |
10131 | ||
10132 | static bool trans_BLX_i(DisasContext *s, arg_BLX_i *a) | |
10133 | { | |
eac2f396 RH |
10134 | TCGv_i32 tmp; |
10135 | ||
360144f3 RH |
10136 | /* For A32, ARCH(5) is checked near the start of the uncond block. */ |
10137 | if (s->thumb && (a->imm & 2)) { | |
10138 | return false; | |
10139 | } | |
10140 | tcg_gen_movi_i32(cpu_R[14], s->base.pc_next | s->thumb); | |
eac2f396 RH |
10141 | tmp = tcg_const_i32(!s->thumb); |
10142 | store_cpu_field(tmp, thumb); | |
10143 | gen_jmp(s, (read_pc(s) & ~3) + a->imm); | |
360144f3 RH |
10144 | return true; |
10145 | } | |
10146 | ||
67b54c55 RH |
10147 | static bool trans_BL_BLX_prefix(DisasContext *s, arg_BL_BLX_prefix *a) |
10148 | { | |
10149 | assert(!arm_dc_feature(s, ARM_FEATURE_THUMB2)); | |
10150 | tcg_gen_movi_i32(cpu_R[14], read_pc(s) + (a->imm << 12)); | |
10151 | return true; | |
10152 | } | |
10153 | ||
10154 | static bool trans_BL_suffix(DisasContext *s, arg_BL_suffix *a) | |
10155 | { | |
10156 | TCGv_i32 tmp = tcg_temp_new_i32(); | |
10157 | ||
10158 | assert(!arm_dc_feature(s, ARM_FEATURE_THUMB2)); | |
10159 | tcg_gen_addi_i32(tmp, cpu_R[14], (a->imm << 1) | 1); | |
10160 | tcg_gen_movi_i32(cpu_R[14], s->base.pc_next | 1); | |
10161 | gen_bx(s, tmp); | |
10162 | return true; | |
10163 | } | |
10164 | ||
10165 | static bool trans_BLX_suffix(DisasContext *s, arg_BLX_suffix *a) | |
10166 | { | |
10167 | TCGv_i32 tmp; | |
10168 | ||
10169 | assert(!arm_dc_feature(s, ARM_FEATURE_THUMB2)); | |
10170 | if (!ENABLE_ARCH_5) { | |
10171 | return false; | |
10172 | } | |
10173 | tmp = tcg_temp_new_i32(); | |
10174 | tcg_gen_addi_i32(tmp, cpu_R[14], a->imm << 1); | |
10175 | tcg_gen_andi_i32(tmp, tmp, 0xfffffffc); | |
10176 | tcg_gen_movi_i32(cpu_R[14], s->base.pc_next | 1); | |
10177 | gen_bx(s, tmp); | |
10178 | return true; | |
10179 | } | |
10180 | ||
808092bb RH |
10181 | static bool op_tbranch(DisasContext *s, arg_tbranch *a, bool half) |
10182 | { | |
10183 | TCGv_i32 addr, tmp; | |
10184 | ||
10185 | tmp = load_reg(s, a->rm); | |
10186 | if (half) { | |
10187 | tcg_gen_add_i32(tmp, tmp, tmp); | |
10188 | } | |
10189 | addr = load_reg(s, a->rn); | |
10190 | tcg_gen_add_i32(addr, addr, tmp); | |
10191 | ||
10192 | gen_aa32_ld_i32(s, tmp, addr, get_mem_index(s), | |
10193 | half ? MO_UW | s->be_data : MO_UB); | |
10194 | tcg_temp_free_i32(addr); | |
10195 | ||
10196 | tcg_gen_add_i32(tmp, tmp, tmp); | |
10197 | tcg_gen_addi_i32(tmp, tmp, read_pc(s)); | |
10198 | store_reg(s, 15, tmp); | |
10199 | return true; | |
10200 | } | |
10201 | ||
10202 | static bool trans_TBB(DisasContext *s, arg_tbranch *a) | |
10203 | { | |
10204 | return op_tbranch(s, a, false); | |
10205 | } | |
10206 | ||
10207 | static bool trans_TBH(DisasContext *s, arg_tbranch *a) | |
10208 | { | |
10209 | return op_tbranch(s, a, true); | |
10210 | } | |
10211 | ||
43f7e42c RH |
10212 | static bool trans_CBZ(DisasContext *s, arg_CBZ *a) |
10213 | { | |
10214 | TCGv_i32 tmp = load_reg(s, a->rn); | |
10215 | ||
10216 | arm_gen_condlabel(s); | |
10217 | tcg_gen_brcondi_i32(a->nz ? TCG_COND_EQ : TCG_COND_NE, | |
10218 | tmp, 0, s->condlabel); | |
10219 | tcg_temp_free_i32(tmp); | |
10220 | gen_jmp(s, read_pc(s) + a->imm); | |
10221 | return true; | |
10222 | } | |
10223 | ||
542f5188 | 10224 | /* |
5651697f AB |
10225 | * Supervisor call - both T32 & A32 come here so we need to check |
10226 | * which mode we are in when checking for semihosting. | |
542f5188 RH |
10227 | */ |
10228 | ||
10229 | static bool trans_SVC(DisasContext *s, arg_SVC *a) | |
10230 | { | |
5651697f AB |
10231 | const uint32_t semihost_imm = s->thumb ? 0xab : 0x123456; |
10232 | ||
10233 | if (!arm_dc_feature(s, ARM_FEATURE_M) && semihosting_enabled() && | |
10234 | #ifndef CONFIG_USER_ONLY | |
10235 | !IS_USER(s) && | |
10236 | #endif | |
10237 | (a->imm == semihost_imm)) { | |
10238 | gen_exception_internal_insn(s, s->base.pc_next, EXCP_SEMIHOST); | |
10239 | } else { | |
10240 | gen_set_pc_im(s, s->base.pc_next); | |
10241 | s->svc_imm = a->imm; | |
10242 | s->base.is_jmp = DISAS_SWI; | |
10243 | } | |
542f5188 RH |
10244 | return true; |
10245 | } | |
10246 | ||
885782a7 RH |
10247 | /* |
10248 | * Unconditional system instructions | |
10249 | */ | |
10250 | ||
10251 | static bool trans_RFE(DisasContext *s, arg_RFE *a) | |
10252 | { | |
10253 | static const int8_t pre_offset[4] = { | |
10254 | /* DA */ -4, /* IA */ 0, /* DB */ -8, /* IB */ 4 | |
10255 | }; | |
10256 | static const int8_t post_offset[4] = { | |
10257 | /* DA */ -8, /* IA */ 4, /* DB */ -4, /* IB */ 0 | |
10258 | }; | |
10259 | TCGv_i32 addr, t1, t2; | |
10260 | ||
10261 | if (!ENABLE_ARCH_6 || arm_dc_feature(s, ARM_FEATURE_M)) { | |
10262 | return false; | |
10263 | } | |
10264 | if (IS_USER(s)) { | |
10265 | unallocated_encoding(s); | |
10266 | return true; | |
10267 | } | |
10268 | ||
10269 | addr = load_reg(s, a->rn); | |
10270 | tcg_gen_addi_i32(addr, addr, pre_offset[a->pu]); | |
10271 | ||
10272 | /* Load PC into tmp and CPSR into tmp2. */ | |
10273 | t1 = tcg_temp_new_i32(); | |
10274 | gen_aa32_ld32u(s, t1, addr, get_mem_index(s)); | |
10275 | tcg_gen_addi_i32(addr, addr, 4); | |
10276 | t2 = tcg_temp_new_i32(); | |
10277 | gen_aa32_ld32u(s, t2, addr, get_mem_index(s)); | |
10278 | ||
10279 | if (a->w) { | |
10280 | /* Base writeback. */ | |
10281 | tcg_gen_addi_i32(addr, addr, post_offset[a->pu]); | |
10282 | store_reg(s, a->rn, addr); | |
10283 | } else { | |
10284 | tcg_temp_free_i32(addr); | |
10285 | } | |
10286 | gen_rfe(s, t1, t2); | |
10287 | return true; | |
10288 | } | |
10289 | ||
10290 | static bool trans_SRS(DisasContext *s, arg_SRS *a) | |
10291 | { | |
10292 | if (!ENABLE_ARCH_6 || arm_dc_feature(s, ARM_FEATURE_M)) { | |
10293 | return false; | |
10294 | } | |
10295 | gen_srs(s, a->mode, a->pu, a->w); | |
10296 | return true; | |
10297 | } | |
10298 | ||
52f83b9c RH |
10299 | static bool trans_CPS(DisasContext *s, arg_CPS *a) |
10300 | { | |
10301 | uint32_t mask, val; | |
10302 | ||
20556e7b | 10303 | if (!ENABLE_ARCH_6 || arm_dc_feature(s, ARM_FEATURE_M)) { |
52f83b9c RH |
10304 | return false; |
10305 | } | |
10306 | if (IS_USER(s)) { | |
10307 | /* Implemented as NOP in user mode. */ | |
10308 | return true; | |
10309 | } | |
10310 | /* TODO: There are quite a lot of UNPREDICTABLE argument combinations. */ | |
10311 | ||
10312 | mask = val = 0; | |
10313 | if (a->imod & 2) { | |
10314 | if (a->A) { | |
10315 | mask |= CPSR_A; | |
10316 | } | |
10317 | if (a->I) { | |
10318 | mask |= CPSR_I; | |
10319 | } | |
10320 | if (a->F) { | |
10321 | mask |= CPSR_F; | |
10322 | } | |
10323 | if (a->imod & 1) { | |
10324 | val |= mask; | |
10325 | } | |
10326 | } | |
10327 | if (a->M) { | |
10328 | mask |= CPSR_M; | |
10329 | val |= a->mode; | |
10330 | } | |
10331 | if (mask) { | |
10332 | gen_set_psr_im(s, mask, 0, val); | |
10333 | } | |
10334 | return true; | |
10335 | } | |
10336 | ||
20556e7b RH |
10337 | static bool trans_CPS_v7m(DisasContext *s, arg_CPS_v7m *a) |
10338 | { | |
10339 | TCGv_i32 tmp, addr; | |
10340 | ||
10341 | if (!arm_dc_feature(s, ARM_FEATURE_M)) { | |
10342 | return false; | |
10343 | } | |
10344 | if (IS_USER(s)) { | |
10345 | /* Implemented as NOP in user mode. */ | |
10346 | return true; | |
10347 | } | |
10348 | ||
10349 | tmp = tcg_const_i32(a->im); | |
10350 | /* FAULTMASK */ | |
10351 | if (a->F) { | |
10352 | addr = tcg_const_i32(19); | |
10353 | gen_helper_v7m_msr(cpu_env, addr, tmp); | |
10354 | tcg_temp_free_i32(addr); | |
10355 | } | |
10356 | /* PRIMASK */ | |
10357 | if (a->I) { | |
10358 | addr = tcg_const_i32(16); | |
10359 | gen_helper_v7m_msr(cpu_env, addr, tmp); | |
10360 | tcg_temp_free_i32(addr); | |
10361 | } | |
10362 | tcg_temp_free_i32(tmp); | |
10363 | gen_lookup_tb(s); | |
10364 | return true; | |
10365 | } | |
10366 | ||
519b8471 RH |
10367 | /* |
10368 | * Clear-Exclusive, Barriers | |
10369 | */ | |
10370 | ||
10371 | static bool trans_CLREX(DisasContext *s, arg_CLREX *a) | |
10372 | { | |
10373 | if (s->thumb | |
10374 | ? !ENABLE_ARCH_7 && !arm_dc_feature(s, ARM_FEATURE_M) | |
10375 | : !ENABLE_ARCH_6K) { | |
10376 | return false; | |
10377 | } | |
10378 | gen_clrex(s); | |
10379 | return true; | |
10380 | } | |
10381 | ||
10382 | static bool trans_DSB(DisasContext *s, arg_DSB *a) | |
10383 | { | |
10384 | if (!ENABLE_ARCH_7 && !arm_dc_feature(s, ARM_FEATURE_M)) { | |
10385 | return false; | |
10386 | } | |
10387 | tcg_gen_mb(TCG_MO_ALL | TCG_BAR_SC); | |
10388 | return true; | |
10389 | } | |
10390 | ||
10391 | static bool trans_DMB(DisasContext *s, arg_DMB *a) | |
10392 | { | |
10393 | return trans_DSB(s, NULL); | |
10394 | } | |
10395 | ||
10396 | static bool trans_ISB(DisasContext *s, arg_ISB *a) | |
10397 | { | |
10398 | if (!ENABLE_ARCH_7 && !arm_dc_feature(s, ARM_FEATURE_M)) { | |
10399 | return false; | |
10400 | } | |
10401 | /* | |
10402 | * We need to break the TB after this insn to execute | |
10403 | * self-modifying code correctly and also to take | |
10404 | * any pending interrupts immediately. | |
10405 | */ | |
10406 | gen_goto_tb(s, 0, s->base.pc_next); | |
10407 | return true; | |
10408 | } | |
10409 | ||
10410 | static bool trans_SB(DisasContext *s, arg_SB *a) | |
10411 | { | |
10412 | if (!dc_isar_feature(aa32_sb, s)) { | |
10413 | return false; | |
10414 | } | |
10415 | /* | |
10416 | * TODO: There is no speculation barrier opcode | |
10417 | * for TCG; MB and end the TB instead. | |
10418 | */ | |
10419 | tcg_gen_mb(TCG_MO_ALL | TCG_BAR_SC); | |
10420 | gen_goto_tb(s, 0, s->base.pc_next); | |
10421 | return true; | |
10422 | } | |
10423 | ||
48c04a5d RH |
10424 | static bool trans_SETEND(DisasContext *s, arg_SETEND *a) |
10425 | { | |
10426 | if (!ENABLE_ARCH_6) { | |
10427 | return false; | |
10428 | } | |
10429 | if (a->E != (s->be_data == MO_BE)) { | |
10430 | gen_helper_setend(cpu_env); | |
10431 | s->base.is_jmp = DISAS_UPDATE; | |
10432 | } | |
10433 | return true; | |
10434 | } | |
10435 | ||
beb595f6 RH |
10436 | /* |
10437 | * Preload instructions | |
10438 | * All are nops, contingent on the appropriate arch level. | |
10439 | */ | |
10440 | ||
10441 | static bool trans_PLD(DisasContext *s, arg_PLD *a) | |
10442 | { | |
10443 | return ENABLE_ARCH_5TE; | |
10444 | } | |
10445 | ||
10446 | static bool trans_PLDW(DisasContext *s, arg_PLD *a) | |
10447 | { | |
10448 | return arm_dc_feature(s, ARM_FEATURE_V7MP); | |
10449 | } | |
10450 | ||
10451 | static bool trans_PLI(DisasContext *s, arg_PLD *a) | |
10452 | { | |
10453 | return ENABLE_ARCH_7; | |
10454 | } | |
10455 | ||
43f7e42c RH |
10456 | /* |
10457 | * If-then | |
10458 | */ | |
10459 | ||
10460 | static bool trans_IT(DisasContext *s, arg_IT *a) | |
10461 | { | |
10462 | int cond_mask = a->cond_mask; | |
10463 | ||
10464 | /* | |
10465 | * No actual code generated for this insn, just setup state. | |
10466 | * | |
10467 | * Combinations of firstcond and mask which set up an 0b1111 | |
10468 | * condition are UNPREDICTABLE; we take the CONSTRAINED | |
10469 | * UNPREDICTABLE choice to treat 0b1111 the same as 0b1110, | |
10470 | * i.e. both meaning "execute always". | |
10471 | */ | |
10472 | s->condexec_cond = (cond_mask >> 4) & 0xe; | |
10473 | s->condexec_mask = cond_mask & 0x1f; | |
10474 | return true; | |
10475 | } | |
10476 | ||
51409b9e RH |
10477 | /* |
10478 | * Legacy decoder. | |
10479 | */ | |
10480 | ||
f4df2210 | 10481 | static void disas_arm_insn(DisasContext *s, unsigned int insn) |
9ee6e8bb | 10482 | { |
590057d9 | 10483 | unsigned int cond = insn >> 28; |
9ee6e8bb | 10484 | |
e13886e3 PM |
10485 | /* M variants do not implement ARM mode; this must raise the INVSTATE |
10486 | * UsageFault exception. | |
10487 | */ | |
b53d8923 | 10488 | if (arm_dc_feature(s, ARM_FEATURE_M)) { |
a767fac8 | 10489 | gen_exception_insn(s, s->pc_curr, EXCP_INVSTATE, syn_uncategorized(), |
e13886e3 PM |
10490 | default_exception_el(s)); |
10491 | return; | |
b53d8923 | 10492 | } |
51409b9e RH |
10493 | |
10494 | if (cond == 0xf) { | |
be5e7a76 DES |
10495 | /* In ARMv3 and v4 the NV condition is UNPREDICTABLE; we |
10496 | * choose to UNDEF. In ARMv5 and above the space is used | |
10497 | * for miscellaneous unconditional instructions. | |
10498 | */ | |
10499 | ARCH(5); | |
10500 | ||
9ee6e8bb | 10501 | /* Unconditional instructions. */ |
51409b9e RH |
10502 | if (disas_a32_uncond(s, insn)) { |
10503 | return; | |
10504 | } | |
10505 | /* fall back to legacy decoder */ | |
10506 | ||
9ee6e8bb PB |
10507 | if (((insn >> 25) & 7) == 1) { |
10508 | /* NEON Data processing. */ | |
d614a513 | 10509 | if (!arm_dc_feature(s, ARM_FEATURE_NEON)) { |
9ee6e8bb | 10510 | goto illegal_op; |
d614a513 | 10511 | } |
9ee6e8bb | 10512 | |
7dcc1f89 | 10513 | if (disas_neon_data_insn(s, insn)) { |
9ee6e8bb | 10514 | goto illegal_op; |
7dcc1f89 | 10515 | } |
9ee6e8bb PB |
10516 | return; |
10517 | } | |
10518 | if ((insn & 0x0f100000) == 0x04000000) { | |
10519 | /* NEON load/store. */ | |
d614a513 | 10520 | if (!arm_dc_feature(s, ARM_FEATURE_NEON)) { |
9ee6e8bb | 10521 | goto illegal_op; |
d614a513 | 10522 | } |
9ee6e8bb | 10523 | |
7dcc1f89 | 10524 | if (disas_neon_ls_insn(s, insn)) { |
9ee6e8bb | 10525 | goto illegal_op; |
7dcc1f89 | 10526 | } |
9ee6e8bb PB |
10527 | return; |
10528 | } | |
6a57f3eb WN |
10529 | if ((insn & 0x0f000e10) == 0x0e000a00) { |
10530 | /* VFP. */ | |
7dcc1f89 | 10531 | if (disas_vfp_insn(s, insn)) { |
6a57f3eb WN |
10532 | goto illegal_op; |
10533 | } | |
10534 | return; | |
10535 | } | |
48c04a5d | 10536 | if ((insn & 0x0e000f00) == 0x0c000100) { |
d614a513 | 10537 | if (arm_dc_feature(s, ARM_FEATURE_IWMMXT)) { |
9ee6e8bb | 10538 | /* iWMMXt register transfer. */ |
c0f4af17 | 10539 | if (extract32(s->c15_cpar, 1, 1)) { |
7dcc1f89 | 10540 | if (!disas_iwmmxt_insn(s, insn)) { |
9ee6e8bb | 10541 | return; |
c0f4af17 PM |
10542 | } |
10543 | } | |
9ee6e8bb | 10544 | } |
8b7209fa RH |
10545 | } else if ((insn & 0x0e000a00) == 0x0c000800 |
10546 | && arm_dc_feature(s, ARM_FEATURE_V8)) { | |
10547 | if (disas_neon_insn_3same_ext(s, insn)) { | |
10548 | goto illegal_op; | |
10549 | } | |
10550 | return; | |
638808ff RH |
10551 | } else if ((insn & 0x0f000a00) == 0x0e000800 |
10552 | && arm_dc_feature(s, ARM_FEATURE_V8)) { | |
10553 | if (disas_neon_insn_2reg_scalar_ext(s, insn)) { | |
10554 | goto illegal_op; | |
10555 | } | |
10556 | return; | |
9ee6e8bb PB |
10557 | } |
10558 | goto illegal_op; | |
10559 | } | |
10560 | if (cond != 0xe) { | |
10561 | /* if not always execute, we generate a conditional jump to | |
10562 | next instruction */ | |
c2d9644e | 10563 | arm_skip_unless(s, cond); |
9ee6e8bb | 10564 | } |
51409b9e RH |
10565 | |
10566 | if (disas_a32(s, insn)) { | |
10567 | return; | |
10568 | } | |
10569 | /* fall back to legacy decoder */ | |
10570 | ||
590057d9 RH |
10571 | switch ((insn >> 24) & 0xf) { |
10572 | case 0xc: | |
10573 | case 0xd: | |
10574 | case 0xe: | |
10575 | if (((insn >> 8) & 0xe) == 10) { | |
10576 | /* VFP. */ | |
10577 | if (disas_vfp_insn(s, insn)) { | |
9ee6e8bb | 10578 | goto illegal_op; |
6a57f3eb | 10579 | } |
590057d9 RH |
10580 | } else if (disas_coproc_insn(s, insn)) { |
10581 | /* Coprocessor. */ | |
10582 | goto illegal_op; | |
9ee6e8bb | 10583 | } |
590057d9 RH |
10584 | break; |
10585 | default: | |
10586 | illegal_op: | |
10587 | unallocated_encoding(s); | |
10588 | break; | |
9ee6e8bb PB |
10589 | } |
10590 | } | |
10591 | ||
331b1ca6 | 10592 | static bool thumb_insn_is_16bit(DisasContext *s, uint32_t pc, uint32_t insn) |
296e5a0a | 10593 | { |
331b1ca6 RH |
10594 | /* |
10595 | * Return true if this is a 16 bit instruction. We must be precise | |
10596 | * about this (matching the decode). | |
296e5a0a PM |
10597 | */ |
10598 | if ((insn >> 11) < 0x1d) { | |
10599 | /* Definitely a 16-bit instruction */ | |
10600 | return true; | |
10601 | } | |
10602 | ||
10603 | /* Top five bits 0b11101 / 0b11110 / 0b11111 : this is the | |
10604 | * first half of a 32-bit Thumb insn. Thumb-1 cores might | |
10605 | * end up actually treating this as two 16-bit insns, though, | |
10606 | * if it's half of a bl/blx pair that might span a page boundary. | |
10607 | */ | |
14120108 JS |
10608 | if (arm_dc_feature(s, ARM_FEATURE_THUMB2) || |
10609 | arm_dc_feature(s, ARM_FEATURE_M)) { | |
296e5a0a PM |
10610 | /* Thumb2 cores (including all M profile ones) always treat |
10611 | * 32-bit insns as 32-bit. | |
10612 | */ | |
10613 | return false; | |
10614 | } | |
10615 | ||
331b1ca6 | 10616 | if ((insn >> 11) == 0x1e && pc - s->page_start < TARGET_PAGE_SIZE - 3) { |
296e5a0a PM |
10617 | /* 0b1111_0xxx_xxxx_xxxx : BL/BLX prefix, and the suffix |
10618 | * is not on the next page; we merge this into a 32-bit | |
10619 | * insn. | |
10620 | */ | |
10621 | return false; | |
10622 | } | |
10623 | /* 0b1110_1xxx_xxxx_xxxx : BLX suffix (or UNDEF); | |
10624 | * 0b1111_1xxx_xxxx_xxxx : BL suffix; | |
10625 | * 0b1111_0xxx_xxxx_xxxx : BL/BLX prefix on the end of a page | |
10626 | * -- handle as single 16 bit insn | |
10627 | */ | |
10628 | return true; | |
10629 | } | |
10630 | ||
2eea841c PM |
10631 | /* Translate a 32-bit thumb instruction. */ |
10632 | static void disas_thumb2_insn(DisasContext *s, uint32_t insn) | |
9ee6e8bb | 10633 | { |
14120108 JS |
10634 | /* |
10635 | * ARMv6-M supports a limited subset of Thumb2 instructions. | |
10636 | * Other Thumb1 architectures allow only 32-bit | |
10637 | * combined BL/BLX prefix and suffix. | |
296e5a0a | 10638 | */ |
14120108 JS |
10639 | if (arm_dc_feature(s, ARM_FEATURE_M) && |
10640 | !arm_dc_feature(s, ARM_FEATURE_V7)) { | |
10641 | int i; | |
10642 | bool found = false; | |
8297cb13 JS |
10643 | static const uint32_t armv6m_insn[] = {0xf3808000 /* msr */, |
10644 | 0xf3b08040 /* dsb */, | |
10645 | 0xf3b08050 /* dmb */, | |
10646 | 0xf3b08060 /* isb */, | |
10647 | 0xf3e08000 /* mrs */, | |
10648 | 0xf000d000 /* bl */}; | |
10649 | static const uint32_t armv6m_mask[] = {0xffe0d000, | |
10650 | 0xfff0d0f0, | |
10651 | 0xfff0d0f0, | |
10652 | 0xfff0d0f0, | |
10653 | 0xffe0d000, | |
10654 | 0xf800d000}; | |
14120108 JS |
10655 | |
10656 | for (i = 0; i < ARRAY_SIZE(armv6m_insn); i++) { | |
10657 | if ((insn & armv6m_mask[i]) == armv6m_insn[i]) { | |
10658 | found = true; | |
10659 | break; | |
10660 | } | |
10661 | } | |
10662 | if (!found) { | |
10663 | goto illegal_op; | |
10664 | } | |
10665 | } else if ((insn & 0xf800e800) != 0xf000e800) { | |
9ee6e8bb PB |
10666 | ARCH(6T2); |
10667 | } | |
10668 | ||
51409b9e RH |
10669 | if (disas_t32(s, insn)) { |
10670 | return; | |
10671 | } | |
10672 | /* fall back to legacy decoder */ | |
10673 | ||
9ee6e8bb PB |
10674 | switch ((insn >> 25) & 0xf) { |
10675 | case 0: case 1: case 2: case 3: | |
10676 | /* 16-bit instructions. Should never happen. */ | |
10677 | abort(); | |
9ee6e8bb PB |
10678 | case 6: case 7: case 14: case 15: |
10679 | /* Coprocessor. */ | |
7517748e | 10680 | if (arm_dc_feature(s, ARM_FEATURE_M)) { |
8859ba3c PM |
10681 | /* 0b111x_11xx_xxxx_xxxx_xxxx_xxxx_xxxx_xxxx */ |
10682 | if (extract32(insn, 24, 2) == 3) { | |
10683 | goto illegal_op; /* op0 = 0b11 : unallocated */ | |
10684 | } | |
10685 | ||
10686 | /* | |
10687 | * Decode VLLDM and VLSTM first: these are nonstandard because: | |
10688 | * * if there is no FPU then these insns must NOP in | |
10689 | * Secure state and UNDEF in Nonsecure state | |
10690 | * * if there is an FPU then these insns do not have | |
10691 | * the usual behaviour that disas_vfp_insn() provides of | |
10692 | * being controlled by CPACR/NSACR enable bits or the | |
10693 | * lazy-stacking logic. | |
7517748e | 10694 | */ |
b1e5336a PM |
10695 | if (arm_dc_feature(s, ARM_FEATURE_V8) && |
10696 | (insn & 0xffa00f00) == 0xec200a00) { | |
10697 | /* 0b1110_1100_0x1x_xxxx_xxxx_1010_xxxx_xxxx | |
10698 | * - VLLDM, VLSTM | |
10699 | * We choose to UNDEF if the RAZ bits are non-zero. | |
10700 | */ | |
10701 | if (!s->v8m_secure || (insn & 0x0040f0ff)) { | |
10702 | goto illegal_op; | |
10703 | } | |
019076b0 PM |
10704 | |
10705 | if (arm_dc_feature(s, ARM_FEATURE_VFP)) { | |
f843e771 | 10706 | uint32_t rn = (insn >> 16) & 0xf; |
019076b0 PM |
10707 | TCGv_i32 fptr = load_reg(s, rn); |
10708 | ||
10709 | if (extract32(insn, 20, 1)) { | |
956fe143 | 10710 | gen_helper_v7m_vlldm(cpu_env, fptr); |
019076b0 PM |
10711 | } else { |
10712 | gen_helper_v7m_vlstm(cpu_env, fptr); | |
10713 | } | |
10714 | tcg_temp_free_i32(fptr); | |
10715 | ||
10716 | /* End the TB, because we have updated FP control bits */ | |
10717 | s->base.is_jmp = DISAS_UPDATE; | |
10718 | } | |
b1e5336a PM |
10719 | break; |
10720 | } | |
8859ba3c PM |
10721 | if (arm_dc_feature(s, ARM_FEATURE_VFP) && |
10722 | ((insn >> 8) & 0xe) == 10) { | |
10723 | /* FP, and the CPU supports it */ | |
10724 | if (disas_vfp_insn(s, insn)) { | |
10725 | goto illegal_op; | |
10726 | } | |
10727 | break; | |
10728 | } | |
10729 | ||
b1e5336a | 10730 | /* All other insns: NOCP */ |
a767fac8 | 10731 | gen_exception_insn(s, s->pc_curr, EXCP_NOCP, syn_uncategorized(), |
7517748e PM |
10732 | default_exception_el(s)); |
10733 | break; | |
10734 | } | |
0052087e RH |
10735 | if ((insn & 0xfe000a00) == 0xfc000800 |
10736 | && arm_dc_feature(s, ARM_FEATURE_V8)) { | |
10737 | /* The Thumb2 and ARM encodings are identical. */ | |
10738 | if (disas_neon_insn_3same_ext(s, insn)) { | |
10739 | goto illegal_op; | |
10740 | } | |
10741 | } else if ((insn & 0xff000a00) == 0xfe000800 | |
10742 | && arm_dc_feature(s, ARM_FEATURE_V8)) { | |
10743 | /* The Thumb2 and ARM encodings are identical. */ | |
10744 | if (disas_neon_insn_2reg_scalar_ext(s, insn)) { | |
10745 | goto illegal_op; | |
10746 | } | |
10747 | } else if (((insn >> 24) & 3) == 3) { | |
9ee6e8bb | 10748 | /* Translate into the equivalent ARM encoding. */ |
f06053e3 | 10749 | insn = (insn & 0xe2ffffff) | ((insn & (1 << 28)) >> 4) | (1 << 28); |
7dcc1f89 | 10750 | if (disas_neon_data_insn(s, insn)) { |
9ee6e8bb | 10751 | goto illegal_op; |
7dcc1f89 | 10752 | } |
6a57f3eb | 10753 | } else if (((insn >> 8) & 0xe) == 10) { |
7dcc1f89 | 10754 | if (disas_vfp_insn(s, insn)) { |
6a57f3eb WN |
10755 | goto illegal_op; |
10756 | } | |
9ee6e8bb PB |
10757 | } else { |
10758 | if (insn & (1 << 28)) | |
10759 | goto illegal_op; | |
7dcc1f89 | 10760 | if (disas_coproc_insn(s, insn)) { |
9ee6e8bb | 10761 | goto illegal_op; |
7dcc1f89 | 10762 | } |
9ee6e8bb PB |
10763 | } |
10764 | break; | |
5e291fe1 | 10765 | case 12: |
9ee6e8bb | 10766 | if ((insn & 0x01100000) == 0x01000000) { |
7dcc1f89 | 10767 | if (disas_neon_ls_insn(s, insn)) { |
c1713132 | 10768 | goto illegal_op; |
7dcc1f89 | 10769 | } |
9ee6e8bb PB |
10770 | break; |
10771 | } | |
5e291fe1 | 10772 | goto illegal_op; |
9ee6e8bb | 10773 | default: |
f843e771 RH |
10774 | illegal_op: |
10775 | unallocated_encoding(s); | |
2c0262af FB |
10776 | } |
10777 | } | |
10778 | ||
296e5a0a | 10779 | static void disas_thumb_insn(DisasContext *s, uint32_t insn) |
99c475ab | 10780 | { |
0831403b RH |
10781 | if (!disas_t16(s, insn)) { |
10782 | unallocated_encoding(s); | |
99c475ab | 10783 | } |
99c475ab FB |
10784 | } |
10785 | ||
541ebcd4 PM |
10786 | static bool insn_crosses_page(CPUARMState *env, DisasContext *s) |
10787 | { | |
a0415916 | 10788 | /* Return true if the insn at dc->base.pc_next might cross a page boundary. |
541ebcd4 | 10789 | * (False positives are OK, false negatives are not.) |
5b8d7289 | 10790 | * We know this is a Thumb insn, and our caller ensures we are |
a0415916 | 10791 | * only called if dc->base.pc_next is less than 4 bytes from the page |
5b8d7289 PM |
10792 | * boundary, so we cross the page if the first 16 bits indicate |
10793 | * that this is a 32 bit insn. | |
541ebcd4 | 10794 | */ |
a0415916 | 10795 | uint16_t insn = arm_lduw_code(env, s->base.pc_next, s->sctlr_b); |
541ebcd4 | 10796 | |
a0415916 | 10797 | return !thumb_insn_is_16bit(s, s->base.pc_next, insn); |
541ebcd4 PM |
10798 | } |
10799 | ||
b542683d | 10800 | static void arm_tr_init_disas_context(DisasContextBase *dcbase, CPUState *cs) |
2c0262af | 10801 | { |
1d8a5535 | 10802 | DisasContext *dc = container_of(dcbase, DisasContext, base); |
9c489ea6 | 10803 | CPUARMState *env = cs->env_ptr; |
2fc0cc0e | 10804 | ARMCPU *cpu = env_archcpu(env); |
aad821ac RH |
10805 | uint32_t tb_flags = dc->base.tb->flags; |
10806 | uint32_t condexec, core_mmu_idx; | |
3b46e624 | 10807 | |
962fcbf2 | 10808 | dc->isar = &cpu->isar; |
e50e6a20 | 10809 | dc->condjmp = 0; |
3926cc84 | 10810 | |
40f860cd | 10811 | dc->aarch64 = 0; |
cef9ee70 SS |
10812 | /* If we are coming from secure EL0 in a system with a 32-bit EL3, then |
10813 | * there is no secure EL1, so we route exceptions to EL3. | |
10814 | */ | |
10815 | dc->secure_routed_to_el3 = arm_feature(env, ARM_FEATURE_EL3) && | |
10816 | !arm_el_is_aa64(env, 3); | |
aad821ac RH |
10817 | dc->thumb = FIELD_EX32(tb_flags, TBFLAG_A32, THUMB); |
10818 | dc->sctlr_b = FIELD_EX32(tb_flags, TBFLAG_A32, SCTLR_B); | |
10819 | dc->be_data = FIELD_EX32(tb_flags, TBFLAG_ANY, BE_DATA) ? MO_BE : MO_LE; | |
10820 | condexec = FIELD_EX32(tb_flags, TBFLAG_A32, CONDEXEC); | |
10821 | dc->condexec_mask = (condexec & 0xf) << 1; | |
10822 | dc->condexec_cond = condexec >> 4; | |
10823 | core_mmu_idx = FIELD_EX32(tb_flags, TBFLAG_ANY, MMUIDX); | |
10824 | dc->mmu_idx = core_to_arm_mmu_idx(env, core_mmu_idx); | |
c1e37810 | 10825 | dc->current_el = arm_mmu_idx_to_el(dc->mmu_idx); |
3926cc84 | 10826 | #if !defined(CONFIG_USER_ONLY) |
c1e37810 | 10827 | dc->user = (dc->current_el == 0); |
3926cc84 | 10828 | #endif |
aad821ac RH |
10829 | dc->ns = FIELD_EX32(tb_flags, TBFLAG_A32, NS); |
10830 | dc->fp_excp_el = FIELD_EX32(tb_flags, TBFLAG_ANY, FPEXC_EL); | |
10831 | dc->vfp_enabled = FIELD_EX32(tb_flags, TBFLAG_A32, VFPEN); | |
10832 | dc->vec_len = FIELD_EX32(tb_flags, TBFLAG_A32, VECLEN); | |
ea7ac69d PM |
10833 | if (arm_feature(env, ARM_FEATURE_XSCALE)) { |
10834 | dc->c15_cpar = FIELD_EX32(tb_flags, TBFLAG_A32, XSCALE_CPAR); | |
10835 | dc->vec_stride = 0; | |
10836 | } else { | |
10837 | dc->vec_stride = FIELD_EX32(tb_flags, TBFLAG_A32, VECSTRIDE); | |
10838 | dc->c15_cpar = 0; | |
10839 | } | |
aad821ac | 10840 | dc->v7m_handler_mode = FIELD_EX32(tb_flags, TBFLAG_A32, HANDLER); |
fb602cb7 PM |
10841 | dc->v8m_secure = arm_feature(env, ARM_FEATURE_M_SECURITY) && |
10842 | regime_is_secure(env, dc->mmu_idx); | |
aad821ac | 10843 | dc->v8m_stackcheck = FIELD_EX32(tb_flags, TBFLAG_A32, STACKCHECK); |
6d60c67a | 10844 | dc->v8m_fpccr_s_wrong = FIELD_EX32(tb_flags, TBFLAG_A32, FPCCR_S_WRONG); |
6000531e PM |
10845 | dc->v7m_new_fp_ctxt_needed = |
10846 | FIELD_EX32(tb_flags, TBFLAG_A32, NEW_FP_CTXT_NEEDED); | |
e33cf0f8 | 10847 | dc->v7m_lspact = FIELD_EX32(tb_flags, TBFLAG_A32, LSPACT); |
60322b39 | 10848 | dc->cp_regs = cpu->cp_regs; |
a984e42c | 10849 | dc->features = env->features; |
40f860cd | 10850 | |
50225ad0 PM |
10851 | /* Single step state. The code-generation logic here is: |
10852 | * SS_ACTIVE == 0: | |
10853 | * generate code with no special handling for single-stepping (except | |
10854 | * that anything that can make us go to SS_ACTIVE == 1 must end the TB; | |
10855 | * this happens anyway because those changes are all system register or | |
10856 | * PSTATE writes). | |
10857 | * SS_ACTIVE == 1, PSTATE.SS == 1: (active-not-pending) | |
10858 | * emit code for one insn | |
10859 | * emit code to clear PSTATE.SS | |
10860 | * emit code to generate software step exception for completed step | |
10861 | * end TB (as usual for having generated an exception) | |
10862 | * SS_ACTIVE == 1, PSTATE.SS == 0: (active-pending) | |
10863 | * emit code to generate a software step exception | |
10864 | * end the TB | |
10865 | */ | |
aad821ac RH |
10866 | dc->ss_active = FIELD_EX32(tb_flags, TBFLAG_ANY, SS_ACTIVE); |
10867 | dc->pstate_ss = FIELD_EX32(tb_flags, TBFLAG_ANY, PSTATE_SS); | |
50225ad0 | 10868 | dc->is_ldex = false; |
8bd587c1 PM |
10869 | if (!arm_feature(env, ARM_FEATURE_M)) { |
10870 | dc->debug_target_el = FIELD_EX32(tb_flags, TBFLAG_ANY, DEBUG_TARGET_EL); | |
10871 | } | |
50225ad0 | 10872 | |
bfe7ad5b | 10873 | dc->page_start = dc->base.pc_first & TARGET_PAGE_MASK; |
1d8a5535 | 10874 | |
f7708456 RH |
10875 | /* If architectural single step active, limit to 1. */ |
10876 | if (is_singlestepping(dc)) { | |
b542683d | 10877 | dc->base.max_insns = 1; |
f7708456 RH |
10878 | } |
10879 | ||
d0264d86 RH |
10880 | /* ARM is a fixed-length ISA. Bound the number of insns to execute |
10881 | to those left on the page. */ | |
10882 | if (!dc->thumb) { | |
bfe7ad5b | 10883 | int bound = -(dc->base.pc_first | TARGET_PAGE_MASK) / 4; |
b542683d | 10884 | dc->base.max_insns = MIN(dc->base.max_insns, bound); |
d0264d86 RH |
10885 | } |
10886 | ||
d9eea52c PM |
10887 | cpu_V0 = tcg_temp_new_i64(); |
10888 | cpu_V1 = tcg_temp_new_i64(); | |
e677137d | 10889 | /* FIXME: cpu_M0 can probably be the same as cpu_V0. */ |
a7812ae4 | 10890 | cpu_M0 = tcg_temp_new_i64(); |
1d8a5535 LV |
10891 | } |
10892 | ||
b1476854 LV |
10893 | static void arm_tr_tb_start(DisasContextBase *dcbase, CPUState *cpu) |
10894 | { | |
10895 | DisasContext *dc = container_of(dcbase, DisasContext, base); | |
10896 | ||
10897 | /* A note on handling of the condexec (IT) bits: | |
10898 | * | |
10899 | * We want to avoid the overhead of having to write the updated condexec | |
10900 | * bits back to the CPUARMState for every instruction in an IT block. So: | |
10901 | * (1) if the condexec bits are not already zero then we write | |
10902 | * zero back into the CPUARMState now. This avoids complications trying | |
10903 | * to do it at the end of the block. (For example if we don't do this | |
10904 | * it's hard to identify whether we can safely skip writing condexec | |
10905 | * at the end of the TB, which we definitely want to do for the case | |
10906 | * where a TB doesn't do anything with the IT state at all.) | |
10907 | * (2) if we are going to leave the TB then we call gen_set_condexec() | |
10908 | * which will write the correct value into CPUARMState if zero is wrong. | |
10909 | * This is done both for leaving the TB at the end, and for leaving | |
10910 | * it because of an exception we know will happen, which is done in | |
10911 | * gen_exception_insn(). The latter is necessary because we need to | |
10912 | * leave the TB with the PC/IT state just prior to execution of the | |
10913 | * instruction which caused the exception. | |
10914 | * (3) if we leave the TB unexpectedly (eg a data abort on a load) | |
10915 | * then the CPUARMState will be wrong and we need to reset it. | |
10916 | * This is handled in the same way as restoration of the | |
10917 | * PC in these situations; we save the value of the condexec bits | |
10918 | * for each PC via tcg_gen_insn_start(), and restore_state_to_opc() | |
10919 | * then uses this to restore them after an exception. | |
10920 | * | |
10921 | * Note that there are no instructions which can read the condexec | |
10922 | * bits, and none which can write non-static values to them, so | |
10923 | * we don't need to care about whether CPUARMState is correct in the | |
10924 | * middle of a TB. | |
10925 | */ | |
10926 | ||
10927 | /* Reset the conditional execution bits immediately. This avoids | |
10928 | complications trying to do it at the end of the block. */ | |
10929 | if (dc->condexec_mask || dc->condexec_cond) { | |
10930 | TCGv_i32 tmp = tcg_temp_new_i32(); | |
10931 | tcg_gen_movi_i32(tmp, 0); | |
10932 | store_cpu_field(tmp, condexec_bits); | |
10933 | } | |
10934 | } | |
10935 | ||
f62bd897 LV |
10936 | static void arm_tr_insn_start(DisasContextBase *dcbase, CPUState *cpu) |
10937 | { | |
10938 | DisasContext *dc = container_of(dcbase, DisasContext, base); | |
10939 | ||
a0415916 | 10940 | tcg_gen_insn_start(dc->base.pc_next, |
f62bd897 LV |
10941 | (dc->condexec_cond << 4) | (dc->condexec_mask >> 1), |
10942 | 0); | |
15fa08f8 | 10943 | dc->insn_start = tcg_last_op(); |
f62bd897 LV |
10944 | } |
10945 | ||
a68956ad LV |
10946 | static bool arm_tr_breakpoint_check(DisasContextBase *dcbase, CPUState *cpu, |
10947 | const CPUBreakpoint *bp) | |
10948 | { | |
10949 | DisasContext *dc = container_of(dcbase, DisasContext, base); | |
10950 | ||
10951 | if (bp->flags & BP_CPU) { | |
10952 | gen_set_condexec(dc); | |
a0415916 | 10953 | gen_set_pc_im(dc, dc->base.pc_next); |
a68956ad LV |
10954 | gen_helper_check_breakpoints(cpu_env); |
10955 | /* End the TB early; it's likely not going to be executed */ | |
10956 | dc->base.is_jmp = DISAS_TOO_MANY; | |
10957 | } else { | |
aee828e7 | 10958 | gen_exception_internal_insn(dc, dc->base.pc_next, EXCP_DEBUG); |
a68956ad LV |
10959 | /* The address covered by the breakpoint must be |
10960 | included in [tb->pc, tb->pc + tb->size) in order | |
10961 | to for it to be properly cleared -- thus we | |
10962 | increment the PC here so that the logic setting | |
10963 | tb->size below does the right thing. */ | |
10964 | /* TODO: Advance PC by correct instruction length to | |
10965 | * avoid disassembler error messages */ | |
a0415916 | 10966 | dc->base.pc_next += 2; |
a68956ad LV |
10967 | dc->base.is_jmp = DISAS_NORETURN; |
10968 | } | |
10969 | ||
10970 | return true; | |
10971 | } | |
10972 | ||
722ef0a5 | 10973 | static bool arm_pre_translate_insn(DisasContext *dc) |
13189a90 | 10974 | { |
13189a90 LV |
10975 | #ifdef CONFIG_USER_ONLY |
10976 | /* Intercept jump to the magic kernel page. */ | |
a0415916 | 10977 | if (dc->base.pc_next >= 0xffff0000) { |
13189a90 LV |
10978 | /* We always get here via a jump, so know we are not in a |
10979 | conditional execution block. */ | |
10980 | gen_exception_internal(EXCP_KERNEL_TRAP); | |
10981 | dc->base.is_jmp = DISAS_NORETURN; | |
722ef0a5 | 10982 | return true; |
13189a90 LV |
10983 | } |
10984 | #endif | |
10985 | ||
10986 | if (dc->ss_active && !dc->pstate_ss) { | |
10987 | /* Singlestep state is Active-pending. | |
10988 | * If we're in this state at the start of a TB then either | |
10989 | * a) we just took an exception to an EL which is being debugged | |
10990 | * and this is the first insn in the exception handler | |
10991 | * b) debug exceptions were masked and we just unmasked them | |
10992 | * without changing EL (eg by clearing PSTATE.D) | |
10993 | * In either case we're going to take a swstep exception in the | |
10994 | * "did not step an insn" case, and so the syndrome ISV and EX | |
10995 | * bits should be zero. | |
10996 | */ | |
10997 | assert(dc->base.num_insns == 1); | |
c1d5f50f | 10998 | gen_swstep_exception(dc, 0, 0); |
13189a90 | 10999 | dc->base.is_jmp = DISAS_NORETURN; |
722ef0a5 | 11000 | return true; |
13189a90 LV |
11001 | } |
11002 | ||
722ef0a5 RH |
11003 | return false; |
11004 | } | |
13189a90 | 11005 | |
d0264d86 | 11006 | static void arm_post_translate_insn(DisasContext *dc) |
722ef0a5 | 11007 | { |
13189a90 LV |
11008 | if (dc->condjmp && !dc->base.is_jmp) { |
11009 | gen_set_label(dc->condlabel); | |
11010 | dc->condjmp = 0; | |
11011 | } | |
23169224 | 11012 | translator_loop_temp_check(&dc->base); |
13189a90 LV |
11013 | } |
11014 | ||
722ef0a5 RH |
11015 | static void arm_tr_translate_insn(DisasContextBase *dcbase, CPUState *cpu) |
11016 | { | |
11017 | DisasContext *dc = container_of(dcbase, DisasContext, base); | |
11018 | CPUARMState *env = cpu->env_ptr; | |
11019 | unsigned int insn; | |
11020 | ||
11021 | if (arm_pre_translate_insn(dc)) { | |
11022 | return; | |
11023 | } | |
11024 | ||
a0415916 RH |
11025 | dc->pc_curr = dc->base.pc_next; |
11026 | insn = arm_ldl_code(env, dc->base.pc_next, dc->sctlr_b); | |
58803318 | 11027 | dc->insn = insn; |
a0415916 | 11028 | dc->base.pc_next += 4; |
722ef0a5 RH |
11029 | disas_arm_insn(dc, insn); |
11030 | ||
d0264d86 RH |
11031 | arm_post_translate_insn(dc); |
11032 | ||
11033 | /* ARM is a fixed-length ISA. We performed the cross-page check | |
11034 | in init_disas_context by adjusting max_insns. */ | |
722ef0a5 RH |
11035 | } |
11036 | ||
dcf14dfb PM |
11037 | static bool thumb_insn_is_unconditional(DisasContext *s, uint32_t insn) |
11038 | { | |
11039 | /* Return true if this Thumb insn is always unconditional, | |
11040 | * even inside an IT block. This is true of only a very few | |
11041 | * instructions: BKPT, HLT, and SG. | |
11042 | * | |
11043 | * A larger class of instructions are UNPREDICTABLE if used | |
11044 | * inside an IT block; we do not need to detect those here, because | |
11045 | * what we do by default (perform the cc check and update the IT | |
11046 | * bits state machine) is a permitted CONSTRAINED UNPREDICTABLE | |
11047 | * choice for those situations. | |
11048 | * | |
11049 | * insn is either a 16-bit or a 32-bit instruction; the two are | |
11050 | * distinguishable because for the 16-bit case the top 16 bits | |
11051 | * are zeroes, and that isn't a valid 32-bit encoding. | |
11052 | */ | |
11053 | if ((insn & 0xffffff00) == 0xbe00) { | |
11054 | /* BKPT */ | |
11055 | return true; | |
11056 | } | |
11057 | ||
11058 | if ((insn & 0xffffffc0) == 0xba80 && arm_dc_feature(s, ARM_FEATURE_V8) && | |
11059 | !arm_dc_feature(s, ARM_FEATURE_M)) { | |
11060 | /* HLT: v8A only. This is unconditional even when it is going to | |
11061 | * UNDEF; see the v8A ARM ARM DDI0487B.a H3.3. | |
11062 | * For v7 cores this was a plain old undefined encoding and so | |
11063 | * honours its cc check. (We might be using the encoding as | |
11064 | * a semihosting trap, but we don't change the cc check behaviour | |
11065 | * on that account, because a debugger connected to a real v7A | |
11066 | * core and emulating semihosting traps by catching the UNDEF | |
11067 | * exception would also only see cases where the cc check passed. | |
11068 | * No guest code should be trying to do a HLT semihosting trap | |
11069 | * in an IT block anyway. | |
11070 | */ | |
11071 | return true; | |
11072 | } | |
11073 | ||
11074 | if (insn == 0xe97fe97f && arm_dc_feature(s, ARM_FEATURE_V8) && | |
11075 | arm_dc_feature(s, ARM_FEATURE_M)) { | |
11076 | /* SG: v8M only */ | |
11077 | return true; | |
11078 | } | |
11079 | ||
11080 | return false; | |
11081 | } | |
11082 | ||
722ef0a5 RH |
11083 | static void thumb_tr_translate_insn(DisasContextBase *dcbase, CPUState *cpu) |
11084 | { | |
11085 | DisasContext *dc = container_of(dcbase, DisasContext, base); | |
11086 | CPUARMState *env = cpu->env_ptr; | |
296e5a0a PM |
11087 | uint32_t insn; |
11088 | bool is_16bit; | |
722ef0a5 RH |
11089 | |
11090 | if (arm_pre_translate_insn(dc)) { | |
11091 | return; | |
11092 | } | |
11093 | ||
a0415916 RH |
11094 | dc->pc_curr = dc->base.pc_next; |
11095 | insn = arm_lduw_code(env, dc->base.pc_next, dc->sctlr_b); | |
11096 | is_16bit = thumb_insn_is_16bit(dc, dc->base.pc_next, insn); | |
11097 | dc->base.pc_next += 2; | |
296e5a0a | 11098 | if (!is_16bit) { |
a0415916 | 11099 | uint32_t insn2 = arm_lduw_code(env, dc->base.pc_next, dc->sctlr_b); |
296e5a0a PM |
11100 | |
11101 | insn = insn << 16 | insn2; | |
a0415916 | 11102 | dc->base.pc_next += 2; |
296e5a0a | 11103 | } |
58803318 | 11104 | dc->insn = insn; |
296e5a0a | 11105 | |
dcf14dfb | 11106 | if (dc->condexec_mask && !thumb_insn_is_unconditional(dc, insn)) { |
296e5a0a PM |
11107 | uint32_t cond = dc->condexec_cond; |
11108 | ||
5529de1e PM |
11109 | /* |
11110 | * Conditionally skip the insn. Note that both 0xe and 0xf mean | |
11111 | * "always"; 0xf is not "never". | |
11112 | */ | |
11113 | if (cond < 0x0e) { | |
c2d9644e | 11114 | arm_skip_unless(dc, cond); |
296e5a0a PM |
11115 | } |
11116 | } | |
11117 | ||
11118 | if (is_16bit) { | |
11119 | disas_thumb_insn(dc, insn); | |
11120 | } else { | |
2eea841c | 11121 | disas_thumb2_insn(dc, insn); |
296e5a0a | 11122 | } |
722ef0a5 RH |
11123 | |
11124 | /* Advance the Thumb condexec condition. */ | |
11125 | if (dc->condexec_mask) { | |
11126 | dc->condexec_cond = ((dc->condexec_cond & 0xe) | | |
11127 | ((dc->condexec_mask >> 4) & 1)); | |
11128 | dc->condexec_mask = (dc->condexec_mask << 1) & 0x1f; | |
11129 | if (dc->condexec_mask == 0) { | |
11130 | dc->condexec_cond = 0; | |
11131 | } | |
11132 | } | |
11133 | ||
d0264d86 RH |
11134 | arm_post_translate_insn(dc); |
11135 | ||
11136 | /* Thumb is a variable-length ISA. Stop translation when the next insn | |
11137 | * will touch a new page. This ensures that prefetch aborts occur at | |
11138 | * the right place. | |
11139 | * | |
11140 | * We want to stop the TB if the next insn starts in a new page, | |
11141 | * or if it spans between this page and the next. This means that | |
11142 | * if we're looking at the last halfword in the page we need to | |
11143 | * see if it's a 16-bit Thumb insn (which will fit in this TB) | |
11144 | * or a 32-bit Thumb insn (which won't). | |
11145 | * This is to avoid generating a silly TB with a single 16-bit insn | |
11146 | * in it at the end of this page (which would execute correctly | |
11147 | * but isn't very efficient). | |
11148 | */ | |
11149 | if (dc->base.is_jmp == DISAS_NEXT | |
a0415916 RH |
11150 | && (dc->base.pc_next - dc->page_start >= TARGET_PAGE_SIZE |
11151 | || (dc->base.pc_next - dc->page_start >= TARGET_PAGE_SIZE - 3 | |
d0264d86 RH |
11152 | && insn_crosses_page(env, dc)))) { |
11153 | dc->base.is_jmp = DISAS_TOO_MANY; | |
11154 | } | |
722ef0a5 RH |
11155 | } |
11156 | ||
70d3c035 | 11157 | static void arm_tr_tb_stop(DisasContextBase *dcbase, CPUState *cpu) |
1d8a5535 | 11158 | { |
70d3c035 | 11159 | DisasContext *dc = container_of(dcbase, DisasContext, base); |
2e70f6ef | 11160 | |
c5a49c63 | 11161 | if (tb_cflags(dc->base.tb) & CF_LAST_IO && dc->condjmp) { |
70d3c035 LV |
11162 | /* FIXME: This can theoretically happen with self-modifying code. */ |
11163 | cpu_abort(cpu, "IO on conditional branch instruction"); | |
2e70f6ef | 11164 | } |
9ee6e8bb | 11165 | |
b5ff1b31 | 11166 | /* At this stage dc->condjmp will only be set when the skipped |
9ee6e8bb PB |
11167 | instruction was a conditional branch or trap, and the PC has |
11168 | already been written. */ | |
f021b2c4 | 11169 | gen_set_condexec(dc); |
dcba3a8d | 11170 | if (dc->base.is_jmp == DISAS_BX_EXCRET) { |
3bb8a96f PM |
11171 | /* Exception return branches need some special case code at the |
11172 | * end of the TB, which is complex enough that it has to | |
11173 | * handle the single-step vs not and the condition-failed | |
11174 | * insn codepath itself. | |
11175 | */ | |
11176 | gen_bx_excret_final_code(dc); | |
11177 | } else if (unlikely(is_singlestepping(dc))) { | |
7999a5c8 | 11178 | /* Unconditional and "condition passed" instruction codepath. */ |
dcba3a8d | 11179 | switch (dc->base.is_jmp) { |
7999a5c8 | 11180 | case DISAS_SWI: |
50225ad0 | 11181 | gen_ss_advance(dc); |
73710361 GB |
11182 | gen_exception(EXCP_SWI, syn_aa32_svc(dc->svc_imm, dc->thumb), |
11183 | default_exception_el(dc)); | |
7999a5c8 SF |
11184 | break; |
11185 | case DISAS_HVC: | |
37e6456e | 11186 | gen_ss_advance(dc); |
73710361 | 11187 | gen_exception(EXCP_HVC, syn_aa32_hvc(dc->svc_imm), 2); |
7999a5c8 SF |
11188 | break; |
11189 | case DISAS_SMC: | |
37e6456e | 11190 | gen_ss_advance(dc); |
73710361 | 11191 | gen_exception(EXCP_SMC, syn_aa32_smc(), 3); |
7999a5c8 SF |
11192 | break; |
11193 | case DISAS_NEXT: | |
a68956ad | 11194 | case DISAS_TOO_MANY: |
7999a5c8 | 11195 | case DISAS_UPDATE: |
a0415916 | 11196 | gen_set_pc_im(dc, dc->base.pc_next); |
7999a5c8 SF |
11197 | /* fall through */ |
11198 | default: | |
5425415e PM |
11199 | /* FIXME: Single stepping a WFI insn will not halt the CPU. */ |
11200 | gen_singlestep_exception(dc); | |
a0c231e6 RH |
11201 | break; |
11202 | case DISAS_NORETURN: | |
11203 | break; | |
7999a5c8 | 11204 | } |
8aaca4c0 | 11205 | } else { |
9ee6e8bb PB |
11206 | /* While branches must always occur at the end of an IT block, |
11207 | there are a few other things that can cause us to terminate | |
65626741 | 11208 | the TB in the middle of an IT block: |
9ee6e8bb PB |
11209 | - Exception generating instructions (bkpt, swi, undefined). |
11210 | - Page boundaries. | |
11211 | - Hardware watchpoints. | |
11212 | Hardware breakpoints have already been handled and skip this code. | |
11213 | */ | |
dcba3a8d | 11214 | switch(dc->base.is_jmp) { |
8aaca4c0 | 11215 | case DISAS_NEXT: |
a68956ad | 11216 | case DISAS_TOO_MANY: |
a0415916 | 11217 | gen_goto_tb(dc, 1, dc->base.pc_next); |
8aaca4c0 | 11218 | break; |
577bf808 | 11219 | case DISAS_JUMP: |
8a6b28c7 EC |
11220 | gen_goto_ptr(); |
11221 | break; | |
e8d52302 | 11222 | case DISAS_UPDATE: |
a0415916 | 11223 | gen_set_pc_im(dc, dc->base.pc_next); |
e8d52302 | 11224 | /* fall through */ |
577bf808 | 11225 | default: |
8aaca4c0 | 11226 | /* indicate that the hash table must be used to find the next TB */ |
07ea28b4 | 11227 | tcg_gen_exit_tb(NULL, 0); |
8aaca4c0 | 11228 | break; |
a0c231e6 | 11229 | case DISAS_NORETURN: |
8aaca4c0 FB |
11230 | /* nothing more to generate */ |
11231 | break; | |
9ee6e8bb | 11232 | case DISAS_WFI: |
58803318 SS |
11233 | { |
11234 | TCGv_i32 tmp = tcg_const_i32((dc->thumb && | |
11235 | !(dc->insn & (1U << 31))) ? 2 : 4); | |
11236 | ||
11237 | gen_helper_wfi(cpu_env, tmp); | |
11238 | tcg_temp_free_i32(tmp); | |
84549b6d PM |
11239 | /* The helper doesn't necessarily throw an exception, but we |
11240 | * must go back to the main loop to check for interrupts anyway. | |
11241 | */ | |
07ea28b4 | 11242 | tcg_gen_exit_tb(NULL, 0); |
9ee6e8bb | 11243 | break; |
58803318 | 11244 | } |
72c1d3af PM |
11245 | case DISAS_WFE: |
11246 | gen_helper_wfe(cpu_env); | |
11247 | break; | |
c87e5a61 PM |
11248 | case DISAS_YIELD: |
11249 | gen_helper_yield(cpu_env); | |
11250 | break; | |
9ee6e8bb | 11251 | case DISAS_SWI: |
73710361 GB |
11252 | gen_exception(EXCP_SWI, syn_aa32_svc(dc->svc_imm, dc->thumb), |
11253 | default_exception_el(dc)); | |
9ee6e8bb | 11254 | break; |
37e6456e | 11255 | case DISAS_HVC: |
73710361 | 11256 | gen_exception(EXCP_HVC, syn_aa32_hvc(dc->svc_imm), 2); |
37e6456e PM |
11257 | break; |
11258 | case DISAS_SMC: | |
73710361 | 11259 | gen_exception(EXCP_SMC, syn_aa32_smc(), 3); |
37e6456e | 11260 | break; |
8aaca4c0 | 11261 | } |
f021b2c4 PM |
11262 | } |
11263 | ||
11264 | if (dc->condjmp) { | |
11265 | /* "Condition failed" instruction codepath for the branch/trap insn */ | |
11266 | gen_set_label(dc->condlabel); | |
11267 | gen_set_condexec(dc); | |
b636649f | 11268 | if (unlikely(is_singlestepping(dc))) { |
a0415916 | 11269 | gen_set_pc_im(dc, dc->base.pc_next); |
f021b2c4 PM |
11270 | gen_singlestep_exception(dc); |
11271 | } else { | |
a0415916 | 11272 | gen_goto_tb(dc, 1, dc->base.pc_next); |
e50e6a20 | 11273 | } |
2c0262af | 11274 | } |
70d3c035 LV |
11275 | } |
11276 | ||
4013f7fc LV |
11277 | static void arm_tr_disas_log(const DisasContextBase *dcbase, CPUState *cpu) |
11278 | { | |
11279 | DisasContext *dc = container_of(dcbase, DisasContext, base); | |
11280 | ||
11281 | qemu_log("IN: %s\n", lookup_symbol(dc->base.pc_first)); | |
1d48474d | 11282 | log_target_disas(cpu, dc->base.pc_first, dc->base.tb->size); |
4013f7fc LV |
11283 | } |
11284 | ||
23169224 LV |
11285 | static const TranslatorOps arm_translator_ops = { |
11286 | .init_disas_context = arm_tr_init_disas_context, | |
11287 | .tb_start = arm_tr_tb_start, | |
11288 | .insn_start = arm_tr_insn_start, | |
11289 | .breakpoint_check = arm_tr_breakpoint_check, | |
11290 | .translate_insn = arm_tr_translate_insn, | |
11291 | .tb_stop = arm_tr_tb_stop, | |
11292 | .disas_log = arm_tr_disas_log, | |
11293 | }; | |
11294 | ||
722ef0a5 RH |
11295 | static const TranslatorOps thumb_translator_ops = { |
11296 | .init_disas_context = arm_tr_init_disas_context, | |
11297 | .tb_start = arm_tr_tb_start, | |
11298 | .insn_start = arm_tr_insn_start, | |
11299 | .breakpoint_check = arm_tr_breakpoint_check, | |
11300 | .translate_insn = thumb_tr_translate_insn, | |
11301 | .tb_stop = arm_tr_tb_stop, | |
11302 | .disas_log = arm_tr_disas_log, | |
11303 | }; | |
11304 | ||
70d3c035 | 11305 | /* generate intermediate code for basic block 'tb'. */ |
8b86d6d2 | 11306 | void gen_intermediate_code(CPUState *cpu, TranslationBlock *tb, int max_insns) |
70d3c035 | 11307 | { |
23169224 LV |
11308 | DisasContext dc; |
11309 | const TranslatorOps *ops = &arm_translator_ops; | |
70d3c035 | 11310 | |
aad821ac | 11311 | if (FIELD_EX32(tb->flags, TBFLAG_A32, THUMB)) { |
722ef0a5 RH |
11312 | ops = &thumb_translator_ops; |
11313 | } | |
23169224 | 11314 | #ifdef TARGET_AARCH64 |
aad821ac | 11315 | if (FIELD_EX32(tb->flags, TBFLAG_ANY, AARCH64_STATE)) { |
23169224 | 11316 | ops = &aarch64_translator_ops; |
2c0262af FB |
11317 | } |
11318 | #endif | |
23169224 | 11319 | |
8b86d6d2 | 11320 | translator_loop(ops, &dc.base, cpu, tb, max_insns); |
2c0262af FB |
11321 | } |
11322 | ||
bad729e2 RH |
11323 | void restore_state_to_opc(CPUARMState *env, TranslationBlock *tb, |
11324 | target_ulong *data) | |
d2856f1a | 11325 | { |
3926cc84 | 11326 | if (is_a64(env)) { |
bad729e2 | 11327 | env->pc = data[0]; |
40f860cd | 11328 | env->condexec_bits = 0; |
aaa1f954 | 11329 | env->exception.syndrome = data[2] << ARM_INSN_START_WORD2_SHIFT; |
3926cc84 | 11330 | } else { |
bad729e2 RH |
11331 | env->regs[15] = data[0]; |
11332 | env->condexec_bits = data[1]; | |
aaa1f954 | 11333 | env->exception.syndrome = data[2] << ARM_INSN_START_WORD2_SHIFT; |
3926cc84 | 11334 | } |
d2856f1a | 11335 | } |