]>
Commit | Line | Data |
---|---|---|
2e8cf49e NC |
1 | /* simulator.c -- Interface for the AArch64 simulator. |
2 | ||
618f726f | 3 | Copyright (C) 2015-2016 Free Software Foundation, Inc. |
2e8cf49e NC |
4 | |
5 | Contributed by Red Hat. | |
6 | ||
7 | This file is part of GDB. | |
8 | ||
9 | This program is free software; you can redistribute it and/or modify | |
10 | it under the terms of the GNU General Public License as published by | |
11 | the Free Software Foundation; either version 3 of the License, or | |
12 | (at your option) any later version. | |
13 | ||
14 | This program is distributed in the hope that it will be useful, | |
15 | but WITHOUT ANY WARRANTY; without even the implied warranty of | |
16 | MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the | |
17 | GNU General Public License for more details. | |
18 | ||
19 | You should have received a copy of the GNU General Public License | |
20 | along with this program. If not, see <http://www.gnu.org/licenses/>. */ | |
21 | ||
22 | #include "config.h" | |
23 | #include <stdlib.h> | |
24 | #include <stdio.h> | |
25 | #include <string.h> | |
26 | #include <sys/types.h> | |
2e8cf49e NC |
27 | #include <math.h> |
28 | #include <time.h> | |
29 | #include <limits.h> | |
30 | ||
31 | #include "dis-asm.h" | |
32 | ||
33 | #include "simulator.h" | |
34 | #include "cpustate.h" | |
35 | #include "memory.h" | |
36 | ||
37 | #define NO_SP 0 | |
38 | #define SP_OK 1 | |
39 | ||
40 | bfd_boolean disas = FALSE; | |
41 | ||
42 | #define TST(_flag) (aarch64_test_CPSR_bit (cpu, _flag)) | |
43 | #define IS_SET(_X) ( TST (( _X ))) | |
44 | #define IS_CLEAR(_X) (!TST (( _X ))) | |
45 | ||
46 | #define HALT_UNALLOC \ | |
47 | do \ | |
48 | { \ | |
49 | if (TRACE_INSN_P (cpu)) \ | |
50 | { \ | |
51 | aarch64_print_insn (CPU_STATE (cpu), aarch64_get_PC (cpu)); \ | |
52 | TRACE_INSN (cpu, \ | |
53 | "Unallocated instruction detected at sim line %d,"\ | |
54 | " exe addr %" PRIx64, \ | |
55 | __LINE__, aarch64_get_PC (cpu)); \ | |
56 | } \ | |
57 | sim_engine_halt (CPU_STATE (cpu), cpu, NULL, aarch64_get_PC (cpu),\ | |
58 | sim_stopped, SIM_SIGILL); \ | |
59 | } \ | |
60 | while (0) | |
61 | ||
62 | #define HALT_NYI \ | |
63 | do \ | |
64 | { \ | |
65 | if (TRACE_INSN_P (cpu)) \ | |
66 | { \ | |
67 | aarch64_print_insn (CPU_STATE (cpu), aarch64_get_PC (cpu)); \ | |
68 | TRACE_INSN (cpu, \ | |
69 | "Unimplemented instruction detected at sim line %d,"\ | |
70 | " exe addr %" PRIx64, \ | |
71 | __LINE__, aarch64_get_PC (cpu)); \ | |
72 | } \ | |
73 | sim_engine_halt (CPU_STATE (cpu), cpu, NULL, aarch64_get_PC (cpu),\ | |
74 | sim_stopped, SIM_SIGABRT); \ | |
75 | } \ | |
76 | while (0) | |
77 | ||
78 | #define NYI_assert(HI, LO, EXPECTED) \ | |
79 | do \ | |
80 | { \ | |
81 | if (uimm (aarch64_get_instr (cpu), (HI), (LO)) != (EXPECTED)) \ | |
82 | HALT_NYI; \ | |
83 | } \ | |
84 | while (0) | |
85 | ||
86 | #define HALT_UNREACHABLE \ | |
87 | do \ | |
88 | { \ | |
89 | TRACE_EVENTS (cpu, "ISE: unreachable code point"); \ | |
90 | sim_engine_abort (NULL, cpu, aarch64_get_PC (cpu), "Internal Error"); \ | |
91 | } \ | |
92 | while (0) | |
93 | ||
94 | /* Helper functions used by expandLogicalImmediate. */ | |
95 | ||
96 | /* for i = 1, ... N result<i-1> = 1 other bits are zero */ | |
97 | static inline uint64_t | |
98 | ones (int N) | |
99 | { | |
100 | return (N == 64 ? (uint64_t)-1UL : ((1UL << N) - 1)); | |
101 | } | |
102 | ||
103 | /* result<0> to val<N> */ | |
104 | static inline uint64_t | |
105 | pickbit (uint64_t val, int N) | |
106 | { | |
107 | return pickbits64 (val, N, N); | |
108 | } | |
109 | ||
110 | static uint64_t | |
111 | expand_logical_immediate (uint32_t S, uint32_t R, uint32_t N) | |
112 | { | |
113 | uint64_t mask; | |
114 | uint64_t imm; | |
115 | unsigned simd_size; | |
116 | ||
117 | /* The immediate value is S+1 bits to 1, left rotated by SIMDsize - R | |
118 | (in other words, right rotated by R), then replicated. */ | |
119 | if (N != 0) | |
120 | { | |
121 | simd_size = 64; | |
122 | mask = 0xffffffffffffffffull; | |
123 | } | |
124 | else | |
125 | { | |
126 | switch (S) | |
127 | { | |
128 | case 0x00 ... 0x1f: /* 0xxxxx */ simd_size = 32; break; | |
129 | case 0x20 ... 0x2f: /* 10xxxx */ simd_size = 16; S &= 0xf; break; | |
130 | case 0x30 ... 0x37: /* 110xxx */ simd_size = 8; S &= 0x7; break; | |
131 | case 0x38 ... 0x3b: /* 1110xx */ simd_size = 4; S &= 0x3; break; | |
132 | case 0x3c ... 0x3d: /* 11110x */ simd_size = 2; S &= 0x1; break; | |
133 | default: return 0; | |
134 | } | |
135 | mask = (1ull << simd_size) - 1; | |
136 | /* Top bits are IGNORED. */ | |
137 | R &= simd_size - 1; | |
138 | } | |
139 | ||
140 | /* NOTE: if S = simd_size - 1 we get 0xf..f which is rejected. */ | |
141 | if (S == simd_size - 1) | |
142 | return 0; | |
143 | ||
144 | /* S+1 consecutive bits to 1. */ | |
145 | /* NOTE: S can't be 63 due to detection above. */ | |
146 | imm = (1ull << (S + 1)) - 1; | |
147 | ||
148 | /* Rotate to the left by simd_size - R. */ | |
149 | if (R != 0) | |
150 | imm = ((imm << (simd_size - R)) & mask) | (imm >> R); | |
151 | ||
152 | /* Replicate the value according to SIMD size. */ | |
153 | switch (simd_size) | |
154 | { | |
155 | case 2: imm = (imm << 2) | imm; | |
156 | case 4: imm = (imm << 4) | imm; | |
157 | case 8: imm = (imm << 8) | imm; | |
158 | case 16: imm = (imm << 16) | imm; | |
159 | case 32: imm = (imm << 32) | imm; | |
160 | case 64: break; | |
161 | default: return 0; | |
162 | } | |
163 | ||
164 | return imm; | |
165 | } | |
166 | ||
167 | /* Instr[22,10] encodes N immr and imms. we want a lookup table | |
168 | for each possible combination i.e. 13 bits worth of int entries. */ | |
169 | #define LI_TABLE_SIZE (1 << 13) | |
170 | static uint64_t LITable[LI_TABLE_SIZE]; | |
171 | ||
172 | void | |
173 | aarch64_init_LIT_table (void) | |
174 | { | |
175 | unsigned index; | |
176 | ||
177 | for (index = 0; index < LI_TABLE_SIZE; index++) | |
178 | { | |
179 | uint32_t N = uimm (index, 12, 12); | |
180 | uint32_t immr = uimm (index, 11, 6); | |
181 | uint32_t imms = uimm (index, 5, 0); | |
182 | ||
183 | LITable [index] = expand_logical_immediate (imms, immr, N); | |
184 | } | |
185 | } | |
186 | ||
187 | static void | |
188 | dexNotify (sim_cpu *cpu) | |
189 | { | |
190 | /* instr[14,0] == type : 0 ==> method entry, 1 ==> method reentry | |
191 | 2 ==> exit Java, 3 ==> start next bytecode. */ | |
192 | uint32_t type = uimm (aarch64_get_instr (cpu), 14, 0); | |
193 | ||
194 | TRACE_EVENTS (cpu, "Notify Insn encountered, type = 0x%x", type); | |
195 | ||
196 | switch (type) | |
197 | { | |
198 | case 0: | |
199 | /* aarch64_notifyMethodEntry (aarch64_get_reg_u64 (cpu, R23, 0), | |
200 | aarch64_get_reg_u64 (cpu, R22, 0)); */ | |
201 | break; | |
202 | case 1: | |
203 | /* aarch64_notifyMethodReentry (aarch64_get_reg_u64 (cpu, R23, 0), | |
204 | aarch64_get_reg_u64 (cpu, R22, 0)); */ | |
205 | break; | |
206 | case 2: | |
207 | /* aarch64_notifyMethodExit (); */ | |
208 | break; | |
209 | case 3: | |
210 | /* aarch64_notifyBCStart (aarch64_get_reg_u64 (cpu, R23, 0), | |
211 | aarch64_get_reg_u64 (cpu, R22, 0)); */ | |
212 | break; | |
213 | } | |
214 | } | |
215 | ||
216 | /* secondary decode within top level groups */ | |
217 | ||
218 | static void | |
219 | dexPseudo (sim_cpu *cpu) | |
220 | { | |
221 | /* assert instr[28,27] = 00 | |
222 | ||
223 | We provide 2 pseudo instructions: | |
224 | ||
225 | HALT stops execution of the simulator causing an immediate | |
226 | return to the x86 code which entered it. | |
227 | ||
228 | CALLOUT initiates recursive entry into x86 code. A register | |
229 | argument holds the address of the x86 routine. Immediate | |
230 | values in the instruction identify the number of general | |
231 | purpose and floating point register arguments to be passed | |
232 | and the type of any value to be returned. */ | |
233 | ||
234 | uint32_t PSEUDO_HALT = 0xE0000000U; | |
235 | uint32_t PSEUDO_CALLOUT = 0x00018000U; | |
236 | uint32_t PSEUDO_CALLOUTR = 0x00018001U; | |
237 | uint32_t PSEUDO_NOTIFY = 0x00014000U; | |
238 | uint32_t dispatch; | |
239 | ||
240 | if (aarch64_get_instr (cpu) == PSEUDO_HALT) | |
241 | { | |
242 | TRACE_EVENTS (cpu, " Pseudo Halt Instruction"); | |
243 | sim_engine_halt (CPU_STATE (cpu), cpu, NULL, aarch64_get_PC (cpu), | |
244 | sim_stopped, SIM_SIGTRAP); | |
245 | } | |
246 | ||
247 | dispatch = uimm (aarch64_get_instr (cpu), 31, 15); | |
248 | ||
249 | /* We do not handle callouts at the moment. */ | |
250 | if (dispatch == PSEUDO_CALLOUT || dispatch == PSEUDO_CALLOUTR) | |
251 | { | |
252 | TRACE_EVENTS (cpu, " Callout"); | |
253 | sim_engine_halt (CPU_STATE (cpu), cpu, NULL, aarch64_get_PC (cpu), | |
254 | sim_stopped, SIM_SIGABRT); | |
255 | } | |
256 | ||
257 | else if (dispatch == PSEUDO_NOTIFY) | |
258 | dexNotify (cpu); | |
259 | ||
260 | else | |
261 | HALT_UNALLOC; | |
262 | } | |
263 | ||
264 | /* Load-store single register (unscaled offset) | |
265 | These instructions employ a base register plus an unscaled signed | |
266 | 9 bit offset. | |
267 | ||
268 | N.B. the base register (source) can be Xn or SP. all other | |
269 | registers may not be SP. */ | |
270 | ||
271 | /* 32 bit load 32 bit unscaled signed 9 bit. */ | |
272 | static void | |
273 | ldur32 (sim_cpu *cpu, int32_t offset) | |
274 | { | |
275 | unsigned rn = uimm (aarch64_get_instr (cpu), 9, 5); | |
276 | unsigned rt = uimm (aarch64_get_instr (cpu), 4, 0); | |
277 | ||
278 | aarch64_set_reg_u64 (cpu, rt, NO_SP, aarch64_get_mem_u32 | |
279 | (cpu, aarch64_get_reg_u64 (cpu, rn, SP_OK) | |
280 | + offset)); | |
281 | } | |
282 | ||
283 | /* 64 bit load 64 bit unscaled signed 9 bit. */ | |
284 | static void | |
285 | ldur64 (sim_cpu *cpu, int32_t offset) | |
286 | { | |
287 | unsigned rn = uimm (aarch64_get_instr (cpu), 9, 5); | |
288 | unsigned rt = uimm (aarch64_get_instr (cpu), 4, 0); | |
289 | ||
290 | aarch64_set_reg_u64 (cpu, rt, NO_SP, aarch64_get_mem_u64 | |
291 | (cpu, aarch64_get_reg_u64 (cpu, rn, SP_OK) | |
292 | + offset)); | |
293 | } | |
294 | ||
295 | /* 32 bit load zero-extended byte unscaled signed 9 bit. */ | |
296 | static void | |
297 | ldurb32 (sim_cpu *cpu, int32_t offset) | |
298 | { | |
299 | unsigned rn = uimm (aarch64_get_instr (cpu), 9, 5); | |
300 | unsigned rt = uimm (aarch64_get_instr (cpu), 4, 0); | |
301 | ||
302 | aarch64_set_reg_u64 (cpu, rt, NO_SP, aarch64_get_mem_u8 | |
303 | (cpu, aarch64_get_reg_u64 (cpu, rn, SP_OK) | |
304 | + offset)); | |
305 | } | |
306 | ||
307 | /* 32 bit load sign-extended byte unscaled signed 9 bit. */ | |
308 | static void | |
309 | ldursb32 (sim_cpu *cpu, int32_t offset) | |
310 | { | |
311 | unsigned rn = uimm (aarch64_get_instr (cpu), 9, 5); | |
312 | unsigned rt = uimm (aarch64_get_instr (cpu), 4, 0); | |
313 | ||
314 | aarch64_set_reg_u64 (cpu, rt, NO_SP, (uint32_t) aarch64_get_mem_s8 | |
315 | (cpu, aarch64_get_reg_u64 (cpu, rn, SP_OK) | |
316 | + offset)); | |
317 | } | |
318 | ||
319 | /* 64 bit load sign-extended byte unscaled signed 9 bit. */ | |
320 | static void | |
321 | ldursb64 (sim_cpu *cpu, int32_t offset) | |
322 | { | |
323 | unsigned rn = uimm (aarch64_get_instr (cpu), 9, 5); | |
324 | unsigned rt = uimm (aarch64_get_instr (cpu), 4, 0); | |
325 | ||
326 | aarch64_set_reg_s64 (cpu, rt, NO_SP, aarch64_get_mem_s8 | |
327 | (cpu, aarch64_get_reg_u64 (cpu, rn, SP_OK) | |
328 | + offset)); | |
329 | } | |
330 | ||
331 | /* 32 bit load zero-extended short unscaled signed 9 bit */ | |
332 | static void | |
333 | ldurh32 (sim_cpu *cpu, int32_t offset) | |
334 | { | |
335 | unsigned rn = uimm (aarch64_get_instr (cpu), 9, 5); | |
336 | unsigned rd = uimm (aarch64_get_instr (cpu), 4, 0); | |
337 | ||
338 | aarch64_set_reg_u64 (cpu, rd, NO_SP, aarch64_get_mem_u16 | |
339 | (cpu, aarch64_get_reg_u64 (cpu, rn, SP_OK) | |
340 | + offset)); | |
341 | } | |
342 | ||
343 | /* 32 bit load sign-extended short unscaled signed 9 bit */ | |
344 | static void | |
345 | ldursh32 (sim_cpu *cpu, int32_t offset) | |
346 | { | |
347 | unsigned rn = uimm (aarch64_get_instr (cpu), 9, 5); | |
348 | unsigned rd = uimm (aarch64_get_instr (cpu), 4, 0); | |
349 | ||
350 | aarch64_set_reg_u64 (cpu, rd, NO_SP, (uint32_t) aarch64_get_mem_s16 | |
351 | (cpu, aarch64_get_reg_u64 (cpu, rn, SP_OK) | |
352 | + offset)); | |
353 | } | |
354 | ||
355 | /* 64 bit load sign-extended short unscaled signed 9 bit */ | |
356 | static void | |
357 | ldursh64 (sim_cpu *cpu, int32_t offset) | |
358 | { | |
359 | unsigned rn = uimm (aarch64_get_instr (cpu), 9, 5); | |
360 | unsigned rt = uimm (aarch64_get_instr (cpu), 4, 0); | |
361 | ||
362 | aarch64_set_reg_s64 (cpu, rt, NO_SP, aarch64_get_mem_s16 | |
363 | (cpu, aarch64_get_reg_u64 (cpu, rn, SP_OK) | |
364 | + offset)); | |
365 | } | |
366 | ||
367 | /* 64 bit load sign-extended word unscaled signed 9 bit */ | |
368 | static void | |
369 | ldursw (sim_cpu *cpu, int32_t offset) | |
370 | { | |
371 | unsigned rn = uimm (aarch64_get_instr (cpu), 9, 5); | |
372 | unsigned rd = uimm (aarch64_get_instr (cpu), 4, 0); | |
373 | ||
374 | aarch64_set_reg_u64 (cpu, rd, NO_SP, (uint32_t) aarch64_get_mem_s32 | |
375 | (cpu, aarch64_get_reg_u64 (cpu, rn, SP_OK) | |
376 | + offset)); | |
377 | } | |
378 | ||
379 | /* N.B. with stores the value in source is written to the address | |
380 | identified by source2 modified by offset. */ | |
381 | ||
382 | /* 32 bit store 32 bit unscaled signed 9 bit. */ | |
383 | static void | |
384 | stur32 (sim_cpu *cpu, int32_t offset) | |
385 | { | |
386 | unsigned rn = uimm (aarch64_get_instr (cpu), 9, 5); | |
387 | unsigned rd = uimm (aarch64_get_instr (cpu), 4, 0); | |
388 | ||
389 | aarch64_set_mem_u32 (cpu, | |
390 | aarch64_get_reg_u64 (cpu, rn, SP_OK) + offset, | |
391 | aarch64_get_reg_u32 (cpu, rd, NO_SP)); | |
392 | } | |
393 | ||
394 | /* 64 bit store 64 bit unscaled signed 9 bit */ | |
395 | static void | |
396 | stur64 (sim_cpu *cpu, int32_t offset) | |
397 | { | |
398 | unsigned rn = uimm (aarch64_get_instr (cpu), 9, 5); | |
399 | unsigned rd = uimm (aarch64_get_instr (cpu), 4, 0); | |
400 | ||
401 | aarch64_set_mem_u64 (cpu, | |
402 | aarch64_get_reg_u64 (cpu, rn, SP_OK) + offset, | |
403 | aarch64_get_reg_u64 (cpu, rd, NO_SP)); | |
404 | } | |
405 | ||
406 | /* 32 bit store byte unscaled signed 9 bit */ | |
407 | static void | |
408 | sturb (sim_cpu *cpu, int32_t offset) | |
409 | { | |
410 | unsigned rn = uimm (aarch64_get_instr (cpu), 9, 5); | |
411 | unsigned rd = uimm (aarch64_get_instr (cpu), 4, 0); | |
412 | ||
413 | aarch64_set_mem_u8 (cpu, | |
414 | aarch64_get_reg_u64 (cpu, rn, SP_OK) + offset, | |
415 | aarch64_get_reg_u8 (cpu, rd, NO_SP)); | |
416 | } | |
417 | ||
418 | /* 32 bit store short unscaled signed 9 bit */ | |
419 | static void | |
420 | sturh (sim_cpu *cpu, int32_t offset) | |
421 | { | |
422 | unsigned rn = uimm (aarch64_get_instr (cpu), 9, 5); | |
423 | unsigned rd = uimm (aarch64_get_instr (cpu), 4, 0); | |
424 | ||
425 | aarch64_set_mem_u16 (cpu, | |
426 | aarch64_get_reg_u64 (cpu, rn, SP_OK) + offset, | |
427 | aarch64_get_reg_u16 (cpu, rd, NO_SP)); | |
428 | } | |
429 | ||
430 | /* Load single register pc-relative label | |
431 | Offset is a signed 19 bit immediate count in words | |
432 | rt may not be SP. */ | |
433 | ||
434 | /* 32 bit pc-relative load */ | |
435 | static void | |
436 | ldr32_pcrel (sim_cpu *cpu, int32_t offset) | |
437 | { | |
438 | unsigned rd = uimm (aarch64_get_instr (cpu), 4, 0); | |
439 | ||
440 | aarch64_set_reg_u64 (cpu, rd, NO_SP, | |
441 | aarch64_get_mem_u32 | |
442 | (cpu, aarch64_get_PC (cpu) + offset * 4)); | |
443 | } | |
444 | ||
445 | /* 64 bit pc-relative load */ | |
446 | static void | |
447 | ldr_pcrel (sim_cpu *cpu, int32_t offset) | |
448 | { | |
449 | unsigned rd = uimm (aarch64_get_instr (cpu), 4, 0); | |
450 | ||
451 | aarch64_set_reg_u64 (cpu, rd, NO_SP, | |
452 | aarch64_get_mem_u64 | |
453 | (cpu, aarch64_get_PC (cpu) + offset * 4)); | |
454 | } | |
455 | ||
456 | /* sign extended 32 bit pc-relative load */ | |
457 | static void | |
458 | ldrsw_pcrel (sim_cpu *cpu, int32_t offset) | |
459 | { | |
460 | unsigned rd = uimm (aarch64_get_instr (cpu), 4, 0); | |
461 | ||
462 | aarch64_set_reg_u64 (cpu, rd, NO_SP, | |
463 | aarch64_get_mem_s32 | |
464 | (cpu, aarch64_get_PC (cpu) + offset * 4)); | |
465 | } | |
466 | ||
467 | /* float pc-relative load */ | |
468 | static void | |
469 | fldrs_pcrel (sim_cpu *cpu, int32_t offset) | |
470 | { | |
471 | unsigned int rd = uimm (aarch64_get_instr (cpu), 4, 0); | |
472 | ||
473 | aarch64_set_FP_float (cpu, rd, | |
474 | aarch64_get_mem_float | |
475 | (cpu, aarch64_get_PC (cpu) + offset * 4)); | |
476 | } | |
477 | ||
478 | /* double pc-relative load */ | |
479 | static void | |
480 | fldrd_pcrel (sim_cpu *cpu, int32_t offset) | |
481 | { | |
482 | unsigned int st = uimm (aarch64_get_instr (cpu), 4, 0); | |
483 | ||
484 | aarch64_set_FP_double (cpu, st, | |
485 | aarch64_get_mem_double | |
486 | (cpu, aarch64_get_PC (cpu) + offset * 4)); | |
487 | } | |
488 | ||
489 | /* long double pc-relative load. */ | |
490 | static void | |
491 | fldrq_pcrel (sim_cpu *cpu, int32_t offset) | |
492 | { | |
493 | unsigned int st = uimm (aarch64_get_instr (cpu), 4, 0); | |
494 | uint64_t addr = aarch64_get_PC (cpu) + offset * 4; | |
495 | FRegister a; | |
496 | ||
497 | aarch64_get_mem_long_double (cpu, addr, & a); | |
498 | aarch64_set_FP_long_double (cpu, st, a); | |
499 | } | |
500 | ||
501 | /* This can be used to scale an offset by applying | |
502 | the requisite shift. the second argument is either | |
503 | 16, 32 or 64. */ | |
504 | ||
505 | #define SCALE(_offset, _elementSize) \ | |
506 | ((_offset) << ScaleShift ## _elementSize) | |
507 | ||
508 | /* This can be used to optionally scale a register derived offset | |
509 | by applying the requisite shift as indicated by the Scaling | |
510 | argument. the second argument is either Byte, Short, Word | |
511 | or Long. The third argument is either Scaled or Unscaled. | |
512 | N.B. when _Scaling is Scaled the shift gets ANDed with | |
513 | all 1s while when it is Unscaled it gets ANDed with 0. */ | |
514 | ||
515 | #define OPT_SCALE(_offset, _elementType, _Scaling) \ | |
516 | ((_offset) << (_Scaling ? ScaleShift ## _elementType : 0)) | |
517 | ||
518 | /* This can be used to zero or sign extend a 32 bit register derived | |
519 | value to a 64 bit value. the first argument must be the value as | |
520 | a uint32_t and the second must be either UXTW or SXTW. The result | |
521 | is returned as an int64_t. */ | |
522 | ||
523 | static inline int64_t | |
524 | extend (uint32_t value, Extension extension) | |
525 | { | |
526 | union | |
527 | { | |
528 | uint32_t u; | |
529 | int32_t n; | |
530 | } x; | |
531 | ||
532 | /* A branchless variant of this ought to be possible. */ | |
533 | if (extension == UXTW || extension == NoExtension) | |
534 | return value; | |
535 | ||
536 | x.u = value; | |
537 | return x.n; | |
538 | } | |
539 | ||
540 | /* Scalar Floating Point | |
541 | ||
542 | FP load/store single register (4 addressing modes) | |
543 | ||
544 | N.B. the base register (source) can be the stack pointer. | |
545 | The secondary source register (source2) can only be an Xn register. */ | |
546 | ||
547 | /* Load 32 bit unscaled signed 9 bit with pre- or post-writeback. */ | |
548 | static void | |
549 | fldrs_wb (sim_cpu *cpu, int32_t offset, WriteBack wb) | |
550 | { | |
551 | unsigned rn = uimm (aarch64_get_instr (cpu), 9, 5); | |
552 | unsigned st = uimm (aarch64_get_instr (cpu), 4, 0); | |
553 | uint64_t address = aarch64_get_reg_u64 (cpu, rn, SP_OK); | |
554 | ||
555 | if (wb != Post) | |
556 | address += offset; | |
557 | ||
558 | aarch64_set_FP_float (cpu, st, aarch64_get_mem_float (cpu, address)); | |
559 | if (wb == Post) | |
560 | address += offset; | |
561 | ||
562 | if (wb != NoWriteBack) | |
563 | aarch64_set_reg_u64 (cpu, rn, SP_OK, address); | |
564 | } | |
565 | ||
566 | /* Load 32 bit scaled unsigned 12 bit. */ | |
567 | static void | |
568 | fldrs_abs (sim_cpu *cpu, uint32_t offset) | |
569 | { | |
570 | unsigned st = uimm (aarch64_get_instr (cpu), 4, 0); | |
571 | unsigned rn = uimm (aarch64_get_instr (cpu), 9, 5); | |
572 | ||
573 | aarch64_set_FP_float (cpu, st, | |
574 | aarch64_get_mem_float | |
575 | (cpu, aarch64_get_reg_u64 (cpu, rn, SP_OK) | |
576 | + SCALE (offset, 32))); | |
577 | } | |
578 | ||
579 | /* Load 32 bit scaled or unscaled zero- or sign-extended | |
580 | 32-bit register offset. */ | |
581 | static void | |
582 | fldrs_scale_ext (sim_cpu *cpu, Scaling scaling, Extension extension) | |
583 | { | |
584 | unsigned rm = uimm (aarch64_get_instr (cpu), 20, 16); | |
585 | unsigned rn = uimm (aarch64_get_instr (cpu), 9, 5); | |
586 | unsigned st = uimm (aarch64_get_instr (cpu), 4, 0); | |
587 | uint64_t address = aarch64_get_reg_u64 (cpu, rn, SP_OK); | |
588 | int64_t extended = extend (aarch64_get_reg_u32 (cpu, rm, NO_SP), extension); | |
589 | uint64_t displacement = OPT_SCALE (extended, 32, scaling); | |
590 | ||
591 | aarch64_set_FP_float (cpu, st, | |
592 | aarch64_get_mem_float | |
593 | (cpu, address + displacement)); | |
594 | } | |
595 | ||
596 | /* Load 64 bit unscaled signed 9 bit with pre- or post-writeback. */ | |
597 | static void | |
598 | fldrd_wb (sim_cpu *cpu, int32_t offset, WriteBack wb) | |
599 | { | |
600 | unsigned rn = uimm (aarch64_get_instr (cpu), 9, 5); | |
601 | unsigned st = uimm (aarch64_get_instr (cpu), 4, 0); | |
602 | uint64_t address = aarch64_get_reg_u64 (cpu, rn, SP_OK); | |
603 | ||
604 | if (wb != Post) | |
605 | address += offset; | |
606 | ||
607 | aarch64_set_FP_double (cpu, st, aarch64_get_mem_double (cpu, address)); | |
608 | ||
609 | if (wb == Post) | |
610 | address += offset; | |
611 | ||
612 | if (wb != NoWriteBack) | |
613 | aarch64_set_reg_u64 (cpu, rn, SP_OK, address); | |
614 | } | |
615 | ||
616 | /* Load 64 bit scaled unsigned 12 bit. */ | |
617 | static void | |
618 | fldrd_abs (sim_cpu *cpu, uint32_t offset) | |
619 | { | |
620 | unsigned rn = uimm (aarch64_get_instr (cpu), 9, 5); | |
621 | unsigned st = uimm (aarch64_get_instr (cpu), 4, 0); | |
622 | uint64_t address = aarch64_get_reg_u64 (cpu, rn, SP_OK) + SCALE (offset, 64); | |
623 | ||
624 | aarch64_set_FP_double (cpu, st, aarch64_get_mem_double (cpu, address)); | |
625 | } | |
626 | ||
627 | /* Load 64 bit scaled or unscaled zero- or sign-extended 32-bit register offset. */ | |
628 | static void | |
629 | fldrd_scale_ext (sim_cpu *cpu, Scaling scaling, Extension extension) | |
630 | { | |
631 | unsigned rm = uimm (aarch64_get_instr (cpu), 20, 16); | |
632 | int64_t extended = extend (aarch64_get_reg_u32 (cpu, rm, NO_SP), extension); | |
633 | uint64_t displacement = OPT_SCALE (extended, 64, scaling); | |
634 | ||
635 | fldrd_wb (cpu, displacement, NoWriteBack); | |
636 | } | |
637 | ||
638 | /* Load 128 bit unscaled signed 9 bit with pre- or post-writeback. */ | |
639 | static void | |
640 | fldrq_wb (sim_cpu *cpu, int32_t offset, WriteBack wb) | |
641 | { | |
642 | FRegister a; | |
643 | unsigned rn = uimm (aarch64_get_instr (cpu), 9, 5); | |
644 | unsigned st = uimm (aarch64_get_instr (cpu), 4, 0); | |
645 | uint64_t address = aarch64_get_reg_u64 (cpu, rn, SP_OK); | |
646 | ||
647 | if (wb != Post) | |
648 | address += offset; | |
649 | ||
650 | aarch64_get_mem_long_double (cpu, address, & a); | |
651 | aarch64_set_FP_long_double (cpu, st, a); | |
652 | ||
653 | if (wb == Post) | |
654 | address += offset; | |
655 | ||
656 | if (wb != NoWriteBack) | |
657 | aarch64_set_reg_u64 (cpu, rn, SP_OK, address); | |
658 | } | |
659 | ||
660 | /* Load 128 bit scaled unsigned 12 bit. */ | |
661 | static void | |
662 | fldrq_abs (sim_cpu *cpu, uint32_t offset) | |
663 | { | |
664 | FRegister a; | |
665 | unsigned rn = uimm (aarch64_get_instr (cpu), 9, 5); | |
666 | unsigned st = uimm (aarch64_get_instr (cpu), 4, 0); | |
667 | uint64_t address = aarch64_get_reg_u64 (cpu, rn, SP_OK) + SCALE (offset, 128); | |
668 | ||
669 | aarch64_get_mem_long_double (cpu, address, & a); | |
670 | aarch64_set_FP_long_double (cpu, st, a); | |
671 | } | |
672 | ||
673 | /* Load 128 bit scaled or unscaled zero- or sign-extended 32-bit register offset */ | |
674 | static void | |
675 | fldrq_scale_ext (sim_cpu *cpu, Scaling scaling, Extension extension) | |
676 | { | |
677 | unsigned rm = uimm (aarch64_get_instr (cpu), 20, 16); | |
678 | int64_t extended = extend (aarch64_get_reg_u32 (cpu, rm, NO_SP), extension); | |
679 | uint64_t displacement = OPT_SCALE (extended, 128, scaling); | |
680 | ||
681 | fldrq_wb (cpu, displacement, NoWriteBack); | |
682 | } | |
683 | ||
684 | /* Memory Access | |
685 | ||
686 | load-store single register | |
687 | There are four addressing modes available here which all employ a | |
688 | 64 bit source (base) register. | |
689 | ||
690 | N.B. the base register (source) can be the stack pointer. | |
691 | The secondary source register (source2)can only be an Xn register. | |
692 | ||
693 | Scaled, 12-bit, unsigned immediate offset, without pre- and | |
694 | post-index options. | |
695 | Unscaled, 9-bit, signed immediate offset with pre- or post-index | |
696 | writeback. | |
697 | scaled or unscaled 64-bit register offset. | |
698 | scaled or unscaled 32-bit extended register offset. | |
699 | ||
700 | All offsets are assumed to be raw from the decode i.e. the | |
701 | simulator is expected to adjust scaled offsets based on the | |
702 | accessed data size with register or extended register offset | |
703 | versions the same applies except that in the latter case the | |
704 | operation may also require a sign extend. | |
705 | ||
706 | A separate method is provided for each possible addressing mode. */ | |
707 | ||
708 | /* 32 bit load 32 bit scaled unsigned 12 bit */ | |
709 | static void | |
710 | ldr32_abs (sim_cpu *cpu, uint32_t offset) | |
711 | { | |
712 | unsigned rn = uimm (aarch64_get_instr (cpu), 9, 5); | |
713 | unsigned rt = uimm (aarch64_get_instr (cpu), 4, 0); | |
714 | ||
715 | /* The target register may not be SP but the source may be. */ | |
716 | aarch64_set_reg_u64 (cpu, rt, NO_SP, aarch64_get_mem_u32 | |
717 | (cpu, aarch64_get_reg_u64 (cpu, rn, SP_OK) | |
718 | + SCALE (offset, 32))); | |
719 | } | |
720 | ||
721 | /* 32 bit load 32 bit unscaled signed 9 bit with pre- or post-writeback. */ | |
722 | static void | |
723 | ldr32_wb (sim_cpu *cpu, int32_t offset, WriteBack wb) | |
724 | { | |
725 | unsigned rn = uimm (aarch64_get_instr (cpu), 9, 5); | |
726 | unsigned rt = uimm (aarch64_get_instr (cpu), 4, 0); | |
727 | uint64_t address; | |
728 | ||
729 | if (rn == rt && wb != NoWriteBack) | |
730 | HALT_UNALLOC; | |
731 | ||
732 | address = aarch64_get_reg_u64 (cpu, rn, SP_OK); | |
733 | ||
734 | if (wb != Post) | |
735 | address += offset; | |
736 | ||
737 | aarch64_set_reg_u64 (cpu, rt, NO_SP, aarch64_get_mem_u32 (cpu, address)); | |
738 | ||
739 | if (wb == Post) | |
740 | address += offset; | |
741 | ||
742 | if (wb != NoWriteBack) | |
743 | aarch64_set_reg_u64 (cpu, rn, SP_OK, address); | |
744 | } | |
745 | ||
746 | /* 32 bit load 32 bit scaled or unscaled | |
747 | zero- or sign-extended 32-bit register offset */ | |
748 | static void | |
749 | ldr32_scale_ext (sim_cpu *cpu, Scaling scaling, Extension extension) | |
750 | { | |
751 | unsigned rm = uimm (aarch64_get_instr (cpu), 20, 16); | |
752 | unsigned rn = uimm (aarch64_get_instr (cpu), 9, 5); | |
753 | unsigned rt = uimm (aarch64_get_instr (cpu), 4, 0); | |
754 | /* rn may reference SP, rm and rt must reference ZR */ | |
755 | ||
756 | uint64_t address = aarch64_get_reg_u64 (cpu, rn, SP_OK); | |
757 | int64_t extended = extend (aarch64_get_reg_u32 (cpu, rm, NO_SP), extension); | |
758 | uint64_t displacement = OPT_SCALE (extended, 32, scaling); | |
759 | ||
760 | aarch64_set_reg_u64 (cpu, rt, NO_SP, | |
761 | aarch64_get_mem_u32 (cpu, address + displacement)); | |
762 | } | |
763 | ||
764 | /* 64 bit load 64 bit scaled unsigned 12 bit */ | |
765 | static void | |
766 | ldr_abs (sim_cpu *cpu, uint32_t offset) | |
767 | { | |
768 | unsigned rn = uimm (aarch64_get_instr (cpu), 9, 5); | |
769 | unsigned rt = uimm (aarch64_get_instr (cpu), 4, 0); | |
770 | ||
771 | /* The target register may not be SP but the source may be. */ | |
772 | aarch64_set_reg_u64 (cpu, rt, NO_SP, aarch64_get_mem_u64 | |
773 | (cpu, aarch64_get_reg_u64 (cpu, rn, SP_OK) | |
774 | + SCALE (offset, 64))); | |
775 | } | |
776 | ||
777 | /* 64 bit load 64 bit unscaled signed 9 bit with pre- or post-writeback. */ | |
778 | static void | |
779 | ldr_wb (sim_cpu *cpu, int32_t offset, WriteBack wb) | |
780 | { | |
781 | unsigned rn = uimm (aarch64_get_instr (cpu), 9, 5); | |
782 | unsigned rt = uimm (aarch64_get_instr (cpu), 4, 0); | |
783 | uint64_t address; | |
784 | ||
785 | if (rn == rt && wb != NoWriteBack) | |
786 | HALT_UNALLOC; | |
787 | ||
788 | address = aarch64_get_reg_u64 (cpu, rn, SP_OK); | |
789 | ||
790 | if (wb != Post) | |
791 | address += offset; | |
792 | ||
793 | aarch64_set_reg_u64 (cpu, rt, NO_SP, aarch64_get_mem_u64 (cpu, address)); | |
794 | ||
795 | if (wb == Post) | |
796 | address += offset; | |
797 | ||
798 | if (wb != NoWriteBack) | |
799 | aarch64_set_reg_u64 (cpu, rn, SP_OK, address); | |
800 | } | |
801 | ||
802 | /* 64 bit load 64 bit scaled or unscaled zero- | |
803 | or sign-extended 32-bit register offset. */ | |
804 | static void | |
805 | ldr_scale_ext (sim_cpu *cpu, Scaling scaling, Extension extension) | |
806 | { | |
807 | unsigned rm = uimm (aarch64_get_instr (cpu), 20, 16); | |
808 | unsigned rn = uimm (aarch64_get_instr (cpu), 9, 5); | |
809 | unsigned rt = uimm (aarch64_get_instr (cpu), 4, 0); | |
810 | /* rn may reference SP, rm and rt must reference ZR */ | |
811 | ||
812 | uint64_t address = aarch64_get_reg_u64 (cpu, rn, SP_OK); | |
813 | int64_t extended = extend (aarch64_get_reg_u32 (cpu, rm, NO_SP), extension); | |
814 | uint64_t displacement = OPT_SCALE (extended, 64, scaling); | |
815 | ||
816 | aarch64_set_reg_u64 (cpu, rt, NO_SP, | |
817 | aarch64_get_mem_u64 (cpu, address + displacement)); | |
818 | } | |
819 | ||
820 | /* 32 bit load zero-extended byte scaled unsigned 12 bit. */ | |
821 | static void | |
822 | ldrb32_abs (sim_cpu *cpu, uint32_t offset) | |
823 | { | |
824 | unsigned rn = uimm (aarch64_get_instr (cpu), 9, 5); | |
825 | unsigned rt = uimm (aarch64_get_instr (cpu), 4, 0); | |
826 | ||
827 | /* The target register may not be SP but the source may be | |
828 | there is no scaling required for a byte load. */ | |
829 | aarch64_set_reg_u64 (cpu, rt, NO_SP, | |
830 | aarch64_get_mem_u8 | |
831 | (cpu, aarch64_get_reg_u64 (cpu, rn, SP_OK) + offset)); | |
832 | } | |
833 | ||
834 | /* 32 bit load zero-extended byte unscaled signed 9 bit with pre- or post-writeback. */ | |
835 | static void | |
836 | ldrb32_wb (sim_cpu *cpu, int32_t offset, WriteBack wb) | |
837 | { | |
838 | unsigned rn = uimm (aarch64_get_instr (cpu), 9, 5); | |
839 | unsigned rt = uimm (aarch64_get_instr (cpu), 4, 0); | |
840 | uint64_t address; | |
841 | ||
842 | if (rn == rt && wb != NoWriteBack) | |
843 | HALT_UNALLOC; | |
844 | ||
845 | address = aarch64_get_reg_u64 (cpu, rn, SP_OK); | |
846 | ||
847 | if (wb != Post) | |
848 | address += offset; | |
849 | ||
850 | aarch64_set_reg_u64 (cpu, rt, NO_SP, aarch64_get_mem_u8 (cpu, address)); | |
851 | ||
852 | if (wb == Post) | |
853 | address += offset; | |
854 | ||
855 | if (wb != NoWriteBack) | |
856 | aarch64_set_reg_u64 (cpu, rn, SP_OK, address); | |
857 | } | |
858 | ||
859 | /* 32 bit load zero-extended byte scaled or unscaled zero- | |
860 | or sign-extended 32-bit register offset. */ | |
861 | static void | |
862 | ldrb32_scale_ext (sim_cpu *cpu, Scaling scaling, Extension extension) | |
863 | { | |
864 | unsigned rm = uimm (aarch64_get_instr (cpu), 20, 16); | |
865 | unsigned rn = uimm (aarch64_get_instr (cpu), 9, 5); | |
866 | unsigned rt = uimm (aarch64_get_instr (cpu), 4, 0); | |
867 | /* rn may reference SP, rm and rt must reference ZR */ | |
868 | ||
869 | uint64_t address = aarch64_get_reg_u64 (cpu, rn, SP_OK); | |
870 | int64_t displacement = extend (aarch64_get_reg_u32 (cpu, rm, NO_SP), | |
871 | extension); | |
872 | ||
873 | /* There is no scaling required for a byte load. */ | |
874 | aarch64_set_reg_u64 (cpu, rt, NO_SP, | |
875 | aarch64_get_mem_u8 (cpu, address + displacement)); | |
876 | } | |
877 | ||
878 | /* 64 bit load sign-extended byte unscaled signed 9 bit | |
879 | with pre- or post-writeback. */ | |
880 | static void | |
881 | ldrsb_wb (sim_cpu *cpu, int32_t offset, WriteBack wb) | |
882 | { | |
883 | unsigned rn = uimm (aarch64_get_instr (cpu), 9, 5); | |
884 | unsigned rt = uimm (aarch64_get_instr (cpu), 4, 0); | |
885 | uint64_t address; | |
886 | ||
887 | if (rn == rt && wb != NoWriteBack) | |
888 | HALT_UNALLOC; | |
889 | ||
890 | address = aarch64_get_reg_u64 (cpu, rn, SP_OK); | |
891 | ||
892 | if (wb != Post) | |
893 | address += offset; | |
894 | ||
895 | aarch64_set_reg_u64 (cpu, rt, NO_SP, aarch64_get_mem_s8 (cpu, address)); | |
896 | ||
897 | if (wb == Post) | |
898 | address += offset; | |
899 | ||
900 | if (wb != NoWriteBack) | |
901 | aarch64_set_reg_u64 (cpu, rn, SP_OK, address); | |
902 | } | |
903 | ||
904 | /* 64 bit load sign-extended byte scaled unsigned 12 bit. */ | |
905 | static void | |
906 | ldrsb_abs (sim_cpu *cpu, uint32_t offset) | |
907 | { | |
908 | ldrsb_wb (cpu, offset, NoWriteBack); | |
909 | } | |
910 | ||
911 | /* 64 bit load sign-extended byte scaled or unscaled zero- | |
912 | or sign-extended 32-bit register offset. */ | |
913 | static void | |
914 | ldrsb_scale_ext (sim_cpu *cpu, Scaling scaling, Extension extension) | |
915 | { | |
916 | unsigned rm = uimm (aarch64_get_instr (cpu), 20, 16); | |
917 | unsigned rn = uimm (aarch64_get_instr (cpu), 9, 5); | |
918 | unsigned rt = uimm (aarch64_get_instr (cpu), 4, 0); | |
919 | /* rn may reference SP, rm and rt must reference ZR */ | |
920 | ||
921 | uint64_t address = aarch64_get_reg_u64 (cpu, rn, SP_OK); | |
922 | int64_t displacement = extend (aarch64_get_reg_u32 (cpu, rm, NO_SP), | |
923 | extension); | |
924 | /* There is no scaling required for a byte load. */ | |
925 | aarch64_set_reg_u64 (cpu, rt, NO_SP, | |
926 | aarch64_get_mem_s8 (cpu, address + displacement)); | |
927 | } | |
928 | ||
929 | /* 32 bit load zero-extended short scaled unsigned 12 bit. */ | |
930 | static void | |
931 | ldrh32_abs (sim_cpu *cpu, uint32_t offset) | |
932 | { | |
933 | unsigned rn = uimm (aarch64_get_instr (cpu), 9, 5); | |
934 | unsigned rt = uimm (aarch64_get_instr (cpu), 4, 0); | |
935 | ||
936 | /* The target register may not be SP but the source may be. */ | |
937 | aarch64_set_reg_u64 (cpu, rt, NO_SP, aarch64_get_mem_u16 | |
938 | (cpu, aarch64_get_reg_u64 (cpu, rn, SP_OK) | |
939 | + SCALE (offset, 16))); | |
940 | } | |
941 | ||
942 | /* 32 bit load zero-extended short unscaled signed 9 bit | |
943 | with pre- or post-writeback. */ | |
944 | static void | |
945 | ldrh32_wb (sim_cpu *cpu, int32_t offset, WriteBack wb) | |
946 | { | |
947 | unsigned rn = uimm (aarch64_get_instr (cpu), 9, 5); | |
948 | unsigned rt = uimm (aarch64_get_instr (cpu), 4, 0); | |
949 | uint64_t address; | |
950 | ||
951 | if (rn == rt && wb != NoWriteBack) | |
952 | HALT_UNALLOC; | |
953 | ||
954 | address = aarch64_get_reg_u64 (cpu, rn, SP_OK); | |
955 | ||
956 | if (wb != Post) | |
957 | address += offset; | |
958 | ||
959 | aarch64_set_reg_u64 (cpu, rt, NO_SP, aarch64_get_mem_u16 (cpu, address)); | |
960 | ||
961 | if (wb == Post) | |
962 | address += offset; | |
963 | ||
964 | if (wb != NoWriteBack) | |
965 | aarch64_set_reg_u64 (cpu, rn, SP_OK, address); | |
966 | } | |
967 | ||
968 | /* 32 bit load zero-extended short scaled or unscaled zero- | |
969 | or sign-extended 32-bit register offset. */ | |
970 | static void | |
971 | ldrh32_scale_ext (sim_cpu *cpu, Scaling scaling, Extension extension) | |
972 | { | |
973 | unsigned rm = uimm (aarch64_get_instr (cpu), 20, 16); | |
974 | unsigned rn = uimm (aarch64_get_instr (cpu), 9, 5); | |
975 | unsigned rt = uimm (aarch64_get_instr (cpu), 4, 0); | |
976 | /* rn may reference SP, rm and rt must reference ZR */ | |
977 | ||
978 | uint64_t address = aarch64_get_reg_u64 (cpu, rn, SP_OK); | |
979 | int64_t extended = extend (aarch64_get_reg_u32 (cpu, rm, NO_SP), extension); | |
980 | uint64_t displacement = OPT_SCALE (extended, 16, scaling); | |
981 | ||
982 | aarch64_set_reg_u64 (cpu, rt, NO_SP, | |
983 | aarch64_get_mem_u16 (cpu, address + displacement)); | |
984 | } | |
985 | ||
986 | /* 32 bit load sign-extended short scaled unsigned 12 bit. */ | |
987 | static void | |
988 | ldrsh32_abs (sim_cpu *cpu, uint32_t offset) | |
989 | { | |
990 | unsigned rn = uimm (aarch64_get_instr (cpu), 9, 5); | |
991 | unsigned rt = uimm (aarch64_get_instr (cpu), 4, 0); | |
992 | ||
993 | /* The target register may not be SP but the source may be. */ | |
994 | aarch64_set_reg_u64 (cpu, rt, NO_SP, (uint32_t) aarch64_get_mem_s16 | |
995 | (cpu, | |
996 | aarch64_get_reg_u64 (cpu, rn, SP_OK) | |
997 | + SCALE (offset, 16))); | |
998 | } | |
999 | ||
1000 | /* 32 bit load sign-extended short unscaled signed 9 bit | |
1001 | with pre- or post-writeback. */ | |
1002 | static void | |
1003 | ldrsh32_wb (sim_cpu *cpu, int32_t offset, WriteBack wb) | |
1004 | { | |
1005 | unsigned rn = uimm (aarch64_get_instr (cpu), 9, 5); | |
1006 | unsigned rt = uimm (aarch64_get_instr (cpu), 4, 0); | |
1007 | uint64_t address; | |
1008 | ||
1009 | if (rn == rt && wb != NoWriteBack) | |
1010 | HALT_UNALLOC; | |
1011 | ||
1012 | address = aarch64_get_reg_u64 (cpu, rn, SP_OK); | |
1013 | ||
1014 | if (wb != Post) | |
1015 | address += offset; | |
1016 | ||
1017 | aarch64_set_reg_u64 (cpu, rt, NO_SP, | |
1018 | (uint32_t) aarch64_get_mem_s16 (cpu, address)); | |
1019 | ||
1020 | if (wb == Post) | |
1021 | address += offset; | |
1022 | ||
1023 | if (wb != NoWriteBack) | |
1024 | aarch64_set_reg_u64 (cpu, rn, SP_OK, address); | |
1025 | } | |
1026 | ||
1027 | /* 32 bit load sign-extended short scaled or unscaled zero- | |
1028 | or sign-extended 32-bit register offset. */ | |
1029 | static void | |
1030 | ldrsh32_scale_ext (sim_cpu *cpu, Scaling scaling, Extension extension) | |
1031 | { | |
1032 | unsigned rm = uimm (aarch64_get_instr (cpu), 20, 16); | |
1033 | unsigned rn = uimm (aarch64_get_instr (cpu), 9, 5); | |
1034 | unsigned rt = uimm (aarch64_get_instr (cpu), 4, 0); | |
1035 | /* rn may reference SP, rm and rt must reference ZR */ | |
1036 | ||
1037 | uint64_t address = aarch64_get_reg_u64 (cpu, rn, SP_OK); | |
1038 | int64_t extended = extend (aarch64_get_reg_u32 (cpu, rm, NO_SP), extension); | |
1039 | uint64_t displacement = OPT_SCALE (extended, 16, scaling); | |
1040 | ||
1041 | aarch64_set_reg_u64 (cpu, rt, NO_SP, | |
1042 | (uint32_t) aarch64_get_mem_s16 | |
1043 | (cpu, address + displacement)); | |
1044 | } | |
1045 | ||
1046 | /* 64 bit load sign-extended short scaled unsigned 12 bit. */ | |
1047 | static void | |
1048 | ldrsh_abs (sim_cpu *cpu, uint32_t offset) | |
1049 | { | |
1050 | unsigned rn = uimm (aarch64_get_instr (cpu), 9, 5); | |
1051 | unsigned rt = uimm (aarch64_get_instr (cpu), 4, 0); | |
1052 | ||
1053 | /* The target register may not be SP but the source may be. */ | |
1054 | aarch64_set_reg_u64 (cpu, rt, NO_SP, aarch64_get_mem_s16 | |
1055 | (cpu, aarch64_get_reg_u64 (cpu, rn, SP_OK) | |
1056 | + SCALE (offset, 16))); | |
1057 | } | |
1058 | ||
1059 | /* 64 bit load sign-extended short unscaled signed 9 bit | |
1060 | with pre- or post-writeback. */ | |
1061 | static void | |
1062 | ldrsh64_wb (sim_cpu *cpu, int32_t offset, WriteBack wb) | |
1063 | { | |
1064 | unsigned rn = uimm (aarch64_get_instr (cpu), 9, 5); | |
1065 | unsigned rt = uimm (aarch64_get_instr (cpu), 4, 0); | |
1066 | uint64_t address; | |
1067 | ||
1068 | if (rn == rt && wb != NoWriteBack) | |
1069 | HALT_UNALLOC; | |
1070 | ||
1071 | address = aarch64_get_reg_u64 (cpu, rn, SP_OK); | |
1072 | ||
1073 | if (wb != Post) | |
1074 | address += offset; | |
1075 | ||
1076 | aarch64_set_reg_u64 (cpu, rt, NO_SP, aarch64_get_mem_s16 (cpu, address)); | |
1077 | ||
1078 | if (wb == Post) | |
1079 | address += offset; | |
1080 | ||
1081 | if (wb != NoWriteBack) | |
1082 | aarch64_set_reg_u64 (cpu, rn, SP_OK, address); | |
1083 | } | |
1084 | ||
1085 | /* 64 bit load sign-extended short scaled or unscaled zero- | |
1086 | or sign-extended 32-bit register offset. */ | |
1087 | static void | |
1088 | ldrsh_scale_ext (sim_cpu *cpu, Scaling scaling, Extension extension) | |
1089 | { | |
1090 | unsigned rm = uimm (aarch64_get_instr (cpu), 20, 16); | |
1091 | unsigned rn = uimm (aarch64_get_instr (cpu), 9, 5); | |
1092 | unsigned rt = uimm (aarch64_get_instr (cpu), 4, 0); | |
1093 | /* rn may reference SP, rm and rt must reference ZR */ | |
1094 | ||
1095 | uint64_t address = aarch64_get_reg_u64 (cpu, rn, SP_OK); | |
1096 | int64_t extended = extend (aarch64_get_reg_u32 (cpu, rm, NO_SP), extension); | |
1097 | uint64_t displacement = OPT_SCALE (extended, 16, scaling); | |
1098 | ||
1099 | aarch64_set_reg_u64 (cpu, rt, NO_SP, | |
1100 | aarch64_get_mem_s16 (cpu, address + displacement)); | |
1101 | } | |
1102 | ||
1103 | /* 64 bit load sign-extended 32 bit scaled unsigned 12 bit. */ | |
1104 | static void | |
1105 | ldrsw_abs (sim_cpu *cpu, uint32_t offset) | |
1106 | { | |
1107 | unsigned rn = uimm (aarch64_get_instr (cpu), 9, 5); | |
1108 | unsigned rt = uimm (aarch64_get_instr (cpu), 4, 0); | |
1109 | ||
1110 | /* The target register may not be SP but the source may be. */ | |
1111 | return aarch64_set_reg_s64 (cpu, rt, NO_SP, aarch64_get_mem_s32 | |
1112 | (cpu, aarch64_get_reg_u64 (cpu, rn, SP_OK) | |
1113 | + SCALE (offset, 32))); | |
1114 | } | |
1115 | ||
1116 | /* 64 bit load sign-extended 32 bit unscaled signed 9 bit | |
1117 | with pre- or post-writeback. */ | |
1118 | static void | |
1119 | ldrsw_wb (sim_cpu *cpu, int32_t offset, WriteBack wb) | |
1120 | { | |
1121 | unsigned rn = uimm (aarch64_get_instr (cpu), 9, 5); | |
1122 | unsigned rt = uimm (aarch64_get_instr (cpu), 4, 0); | |
1123 | uint64_t address; | |
1124 | ||
1125 | if (rn == rt && wb != NoWriteBack) | |
1126 | HALT_UNALLOC; | |
1127 | ||
1128 | address = aarch64_get_reg_u64 (cpu, rn, SP_OK); | |
1129 | ||
1130 | if (wb != Post) | |
1131 | address += offset; | |
1132 | ||
1133 | aarch64_set_reg_s64 (cpu, rt, NO_SP, aarch64_get_mem_s32 (cpu, address)); | |
1134 | ||
1135 | if (wb == Post) | |
1136 | address += offset; | |
1137 | ||
1138 | if (wb != NoWriteBack) | |
1139 | aarch64_set_reg_u64 (cpu, rn, SP_OK, address); | |
1140 | } | |
1141 | ||
1142 | /* 64 bit load sign-extended 32 bit scaled or unscaled zero- | |
1143 | or sign-extended 32-bit register offset. */ | |
1144 | static void | |
1145 | ldrsw_scale_ext (sim_cpu *cpu, Scaling scaling, Extension extension) | |
1146 | { | |
1147 | unsigned rm = uimm (aarch64_get_instr (cpu), 20, 16); | |
1148 | unsigned rn = uimm (aarch64_get_instr (cpu), 9, 5); | |
1149 | unsigned rt = uimm (aarch64_get_instr (cpu), 4, 0); | |
1150 | /* rn may reference SP, rm and rt must reference ZR */ | |
1151 | ||
1152 | uint64_t address = aarch64_get_reg_u64 (cpu, rn, SP_OK); | |
1153 | int64_t extended = extend (aarch64_get_reg_u32 (cpu, rm, NO_SP), extension); | |
1154 | uint64_t displacement = OPT_SCALE (extended, 32, scaling); | |
1155 | ||
1156 | aarch64_set_reg_s64 (cpu, rt, NO_SP, | |
1157 | aarch64_get_mem_s32 (cpu, address + displacement)); | |
1158 | } | |
1159 | ||
1160 | /* N.B. with stores the value in source is written to the | |
1161 | address identified by source2 modified by source3/offset. */ | |
1162 | ||
1163 | /* 32 bit store scaled unsigned 12 bit. */ | |
1164 | static void | |
1165 | str32_abs (sim_cpu *cpu, uint32_t offset) | |
1166 | { | |
1167 | unsigned rn = uimm (aarch64_get_instr (cpu), 9, 5); | |
1168 | unsigned rt = uimm (aarch64_get_instr (cpu), 4, 0); | |
1169 | ||
1170 | /* The target register may not be SP but the source may be. */ | |
1171 | aarch64_set_mem_u32 (cpu, (aarch64_get_reg_u64 (cpu, rn, SP_OK) | |
1172 | + SCALE (offset, 32)), | |
1173 | aarch64_get_reg_u32 (cpu, rt, NO_SP)); | |
1174 | } | |
1175 | ||
1176 | /* 32 bit store unscaled signed 9 bit with pre- or post-writeback. */ | |
1177 | static void | |
1178 | str32_wb (sim_cpu *cpu, int32_t offset, WriteBack wb) | |
1179 | { | |
1180 | unsigned rn = uimm (aarch64_get_instr (cpu), 9, 5); | |
1181 | unsigned rt = uimm (aarch64_get_instr (cpu), 4, 0); | |
1182 | uint64_t address; | |
1183 | ||
1184 | if (rn == rt && wb != NoWriteBack) | |
1185 | HALT_UNALLOC; | |
1186 | ||
1187 | address = aarch64_get_reg_u64 (cpu, rn, SP_OK); | |
1188 | if (wb != Post) | |
1189 | address += offset; | |
1190 | ||
1191 | aarch64_set_mem_u32 (cpu, address, aarch64_get_reg_u32 (cpu, rt, NO_SP)); | |
1192 | ||
1193 | if (wb == Post) | |
1194 | address += offset; | |
1195 | ||
1196 | if (wb != NoWriteBack) | |
1197 | aarch64_set_reg_u64 (cpu, rn, SP_OK, address); | |
1198 | } | |
1199 | ||
1200 | /* 32 bit store scaled or unscaled zero- or | |
1201 | sign-extended 32-bit register offset. */ | |
1202 | static void | |
1203 | str32_scale_ext (sim_cpu *cpu, Scaling scaling, Extension extension) | |
1204 | { | |
1205 | unsigned rm = uimm (aarch64_get_instr (cpu), 20, 16); | |
1206 | unsigned rn = uimm (aarch64_get_instr (cpu), 9, 5); | |
1207 | unsigned rt = uimm (aarch64_get_instr (cpu), 4, 0); | |
1208 | ||
1209 | uint64_t address = aarch64_get_reg_u64 (cpu, rn, SP_OK); | |
1210 | int64_t extended = extend (aarch64_get_reg_u32 (cpu, rm, NO_SP), extension); | |
1211 | uint64_t displacement = OPT_SCALE (extended, 32, scaling); | |
1212 | ||
1213 | aarch64_set_mem_u32 (cpu, address + displacement, | |
1214 | aarch64_get_reg_u64 (cpu, rt, NO_SP)); | |
1215 | } | |
1216 | ||
1217 | /* 64 bit store scaled unsigned 12 bit. */ | |
1218 | static void | |
1219 | str_abs (sim_cpu *cpu, uint32_t offset) | |
1220 | { | |
1221 | unsigned rn = uimm (aarch64_get_instr (cpu), 9, 5); | |
1222 | unsigned rt = uimm (aarch64_get_instr (cpu), 4, 0); | |
1223 | ||
1224 | aarch64_set_mem_u64 (cpu, | |
1225 | aarch64_get_reg_u64 (cpu, rn, SP_OK) | |
1226 | + SCALE (offset, 64), | |
1227 | aarch64_get_reg_u64 (cpu, rt, NO_SP)); | |
1228 | } | |
1229 | ||
1230 | /* 64 bit store unscaled signed 9 bit with pre- or post-writeback. */ | |
1231 | static void | |
1232 | str_wb (sim_cpu *cpu, int32_t offset, WriteBack wb) | |
1233 | { | |
1234 | unsigned rn = uimm (aarch64_get_instr (cpu), 9, 5); | |
1235 | unsigned rt = uimm (aarch64_get_instr (cpu), 4, 0); | |
1236 | uint64_t address; | |
1237 | ||
1238 | if (rn == rt && wb != NoWriteBack) | |
1239 | HALT_UNALLOC; | |
1240 | ||
1241 | address = aarch64_get_reg_u64 (cpu, rn, SP_OK); | |
1242 | ||
1243 | if (wb != Post) | |
1244 | address += offset; | |
1245 | ||
1246 | aarch64_set_mem_u64 (cpu, address, aarch64_get_reg_u64 (cpu, rt, NO_SP)); | |
1247 | ||
1248 | if (wb == Post) | |
1249 | address += offset; | |
1250 | ||
1251 | if (wb != NoWriteBack) | |
1252 | aarch64_set_reg_u64 (cpu, rn, SP_OK, address); | |
1253 | } | |
1254 | ||
1255 | /* 64 bit store scaled or unscaled zero- | |
1256 | or sign-extended 32-bit register offset. */ | |
1257 | static void | |
1258 | str_scale_ext (sim_cpu *cpu, Scaling scaling, Extension extension) | |
1259 | { | |
1260 | unsigned rm = uimm (aarch64_get_instr (cpu), 20, 16); | |
1261 | unsigned rn = uimm (aarch64_get_instr (cpu), 9, 5); | |
1262 | unsigned rt = uimm (aarch64_get_instr (cpu), 4, 0); | |
1263 | /* rn may reference SP, rm and rt must reference ZR */ | |
1264 | ||
1265 | uint64_t address = aarch64_get_reg_u64 (cpu, rn, SP_OK); | |
1266 | int64_t extended = extend (aarch64_get_reg_u32 (cpu, rm, NO_SP), | |
1267 | extension); | |
1268 | uint64_t displacement = OPT_SCALE (extended, 64, scaling); | |
1269 | ||
1270 | aarch64_set_mem_u64 (cpu, address + displacement, | |
1271 | aarch64_get_reg_u64 (cpu, rt, NO_SP)); | |
1272 | } | |
1273 | ||
1274 | /* 32 bit store byte scaled unsigned 12 bit. */ | |
1275 | static void | |
1276 | strb_abs (sim_cpu *cpu, uint32_t offset) | |
1277 | { | |
1278 | unsigned rn = uimm (aarch64_get_instr (cpu), 9, 5); | |
1279 | unsigned rt = uimm (aarch64_get_instr (cpu), 4, 0); | |
1280 | ||
1281 | /* The target register may not be SP but the source may be. | |
1282 | There is no scaling required for a byte load. */ | |
1283 | aarch64_set_mem_u8 (cpu, | |
1284 | aarch64_get_reg_u64 (cpu, rn, SP_OK) + offset, | |
1285 | aarch64_get_reg_u8 (cpu, rt, NO_SP)); | |
1286 | } | |
1287 | ||
1288 | /* 32 bit store byte unscaled signed 9 bit with pre- or post-writeback. */ | |
1289 | static void | |
1290 | strb_wb (sim_cpu *cpu, int32_t offset, WriteBack wb) | |
1291 | { | |
1292 | unsigned rn = uimm (aarch64_get_instr (cpu), 9, 5); | |
1293 | unsigned rt = uimm (aarch64_get_instr (cpu), 4, 0); | |
1294 | uint64_t address; | |
1295 | ||
1296 | if (rn == rt && wb != NoWriteBack) | |
1297 | HALT_UNALLOC; | |
1298 | ||
1299 | address = aarch64_get_reg_u64 (cpu, rn, SP_OK); | |
1300 | ||
1301 | if (wb != Post) | |
1302 | address += offset; | |
1303 | ||
1304 | aarch64_set_mem_u8 (cpu, address, aarch64_get_reg_u8 (cpu, rt, NO_SP)); | |
1305 | ||
1306 | if (wb == Post) | |
1307 | address += offset; | |
1308 | ||
1309 | if (wb != NoWriteBack) | |
1310 | aarch64_set_reg_u64 (cpu, rn, SP_OK, address); | |
1311 | } | |
1312 | ||
1313 | /* 32 bit store byte scaled or unscaled zero- | |
1314 | or sign-extended 32-bit register offset. */ | |
1315 | static void | |
1316 | strb_scale_ext (sim_cpu *cpu, Scaling scaling, Extension extension) | |
1317 | { | |
1318 | unsigned rm = uimm (aarch64_get_instr (cpu), 20, 16); | |
1319 | unsigned rn = uimm (aarch64_get_instr (cpu), 9, 5); | |
1320 | unsigned rt = uimm (aarch64_get_instr (cpu), 4, 0); | |
1321 | /* rn may reference SP, rm and rt must reference ZR */ | |
1322 | ||
1323 | uint64_t address = aarch64_get_reg_u64 (cpu, rn, SP_OK); | |
1324 | int64_t displacement = extend (aarch64_get_reg_u32 (cpu, rm, NO_SP), | |
1325 | extension); | |
1326 | ||
1327 | /* There is no scaling required for a byte load. */ | |
1328 | aarch64_set_mem_u8 (cpu, address + displacement, | |
1329 | aarch64_get_reg_u8 (cpu, rt, NO_SP)); | |
1330 | } | |
1331 | ||
1332 | /* 32 bit store short scaled unsigned 12 bit. */ | |
1333 | static void | |
1334 | strh_abs (sim_cpu *cpu, uint32_t offset) | |
1335 | { | |
1336 | unsigned rn = uimm (aarch64_get_instr (cpu), 9, 5); | |
1337 | unsigned rt = uimm (aarch64_get_instr (cpu), 4, 0); | |
1338 | ||
1339 | /* The target register may not be SP but the source may be. */ | |
1340 | aarch64_set_mem_u16 (cpu, aarch64_get_reg_u64 (cpu, rn, SP_OK) | |
1341 | + SCALE (offset, 16), | |
1342 | aarch64_get_reg_u16 (cpu, rt, NO_SP)); | |
1343 | } | |
1344 | ||
1345 | /* 32 bit store short unscaled signed 9 bit with pre- or post-writeback. */ | |
1346 | static void | |
1347 | strh_wb (sim_cpu *cpu, int32_t offset, WriteBack wb) | |
1348 | { | |
1349 | unsigned rn = uimm (aarch64_get_instr (cpu), 9, 5); | |
1350 | unsigned rt = uimm (aarch64_get_instr (cpu), 4, 0); | |
1351 | uint64_t address; | |
1352 | ||
1353 | if (rn == rt && wb != NoWriteBack) | |
1354 | HALT_UNALLOC; | |
1355 | ||
1356 | address = aarch64_get_reg_u64 (cpu, rn, SP_OK); | |
1357 | ||
1358 | if (wb != Post) | |
1359 | address += offset; | |
1360 | ||
1361 | aarch64_set_mem_u16 (cpu, address, aarch64_get_reg_u16 (cpu, rt, NO_SP)); | |
1362 | ||
1363 | if (wb == Post) | |
1364 | address += offset; | |
1365 | ||
1366 | if (wb != NoWriteBack) | |
1367 | aarch64_set_reg_u64 (cpu, rn, SP_OK, address); | |
1368 | } | |
1369 | ||
1370 | /* 32 bit store short scaled or unscaled zero- | |
1371 | or sign-extended 32-bit register offset. */ | |
1372 | static void | |
1373 | strh_scale_ext (sim_cpu *cpu, Scaling scaling, Extension extension) | |
1374 | { | |
1375 | unsigned rm = uimm (aarch64_get_instr (cpu), 20, 16); | |
1376 | unsigned rn = uimm (aarch64_get_instr (cpu), 9, 5); | |
1377 | unsigned rt = uimm (aarch64_get_instr (cpu), 4, 0); | |
1378 | /* rn may reference SP, rm and rt must reference ZR */ | |
1379 | ||
1380 | uint64_t address = aarch64_get_reg_u64 (cpu, rn, SP_OK); | |
1381 | int64_t extended = extend (aarch64_get_reg_u32 (cpu, rm, NO_SP), extension); | |
1382 | uint64_t displacement = OPT_SCALE (extended, 16, scaling); | |
1383 | ||
1384 | aarch64_set_mem_u16 (cpu, address + displacement, | |
1385 | aarch64_get_reg_u16 (cpu, rt, NO_SP)); | |
1386 | } | |
1387 | ||
1388 | /* Prefetch unsigned 12 bit. */ | |
1389 | static void | |
1390 | prfm_abs (sim_cpu *cpu, uint32_t offset) | |
1391 | { | |
1392 | /* instr[4,0] = prfop : 00000 ==> PLDL1KEEP, 00001 ==> PLDL1STRM, | |
1393 | 00010 ==> PLDL2KEEP, 00001 ==> PLDL2STRM, | |
1394 | 00100 ==> PLDL3KEEP, 00101 ==> PLDL3STRM, | |
1395 | 10000 ==> PSTL1KEEP, 10001 ==> PSTL1STRM, | |
1396 | 10010 ==> PSTL2KEEP, 10001 ==> PSTL2STRM, | |
1397 | 10100 ==> PSTL3KEEP, 10101 ==> PSTL3STRM, | |
1398 | ow ==> UNALLOC | |
1399 | PrfOp prfop = prfop (aarch64_get_instr (cpu), 4, 0); | |
1400 | uint64_t address = aarch64_get_reg_u64 (cpu, rn, SP_OK) | |
1401 | + SCALE (offset, 64). */ | |
1402 | ||
1403 | /* TODO : implement prefetch of address. */ | |
1404 | } | |
1405 | ||
1406 | /* Prefetch scaled or unscaled zero- or sign-extended 32-bit register offset. */ | |
1407 | static void | |
1408 | prfm_scale_ext (sim_cpu *cpu, Scaling scaling, Extension extension) | |
1409 | { | |
1410 | /* instr[4,0] = prfop : 00000 ==> PLDL1KEEP, 00001 ==> PLDL1STRM, | |
1411 | 00010 ==> PLDL2KEEP, 00001 ==> PLDL2STRM, | |
1412 | 00100 ==> PLDL3KEEP, 00101 ==> PLDL3STRM, | |
1413 | 10000 ==> PSTL1KEEP, 10001 ==> PSTL1STRM, | |
1414 | 10010 ==> PSTL2KEEP, 10001 ==> PSTL2STRM, | |
1415 | 10100 ==> PSTL3KEEP, 10101 ==> PSTL3STRM, | |
1416 | ow ==> UNALLOC | |
1417 | rn may reference SP, rm may only reference ZR | |
1418 | PrfOp prfop = prfop (aarch64_get_instr (cpu), 4, 0); | |
1419 | uint64_t base = aarch64_get_reg_u64 (cpu, rn, SP_OK); | |
1420 | int64_t extended = extend (aarch64_get_reg_u32 (cpu, rm, NO_SP), | |
1421 | extension); | |
1422 | uint64_t displacement = OPT_SCALE (extended, 64, scaling); | |
1423 | uint64_t address = base + displacement. */ | |
1424 | ||
1425 | /* TODO : implement prefetch of address */ | |
1426 | } | |
1427 | ||
1428 | /* 64 bit pc-relative prefetch. */ | |
1429 | static void | |
1430 | prfm_pcrel (sim_cpu *cpu, int32_t offset) | |
1431 | { | |
1432 | /* instr[4,0] = prfop : 00000 ==> PLDL1KEEP, 00001 ==> PLDL1STRM, | |
1433 | 00010 ==> PLDL2KEEP, 00001 ==> PLDL2STRM, | |
1434 | 00100 ==> PLDL3KEEP, 00101 ==> PLDL3STRM, | |
1435 | 10000 ==> PSTL1KEEP, 10001 ==> PSTL1STRM, | |
1436 | 10010 ==> PSTL2KEEP, 10001 ==> PSTL2STRM, | |
1437 | 10100 ==> PSTL3KEEP, 10101 ==> PSTL3STRM, | |
1438 | ow ==> UNALLOC | |
1439 | PrfOp prfop = prfop (aarch64_get_instr (cpu), 4, 0); | |
1440 | uint64_t address = aarch64_get_PC (cpu) + offset. */ | |
1441 | ||
1442 | /* TODO : implement this */ | |
1443 | } | |
1444 | ||
1445 | /* Load-store exclusive. */ | |
1446 | ||
1447 | static void | |
1448 | ldxr (sim_cpu *cpu) | |
1449 | { | |
1450 | unsigned rn = uimm (aarch64_get_instr (cpu), 9, 5); | |
1451 | unsigned rt = uimm (aarch64_get_instr (cpu), 4, 0); | |
1452 | uint64_t address = aarch64_get_reg_u64 (cpu, rn, SP_OK); | |
1453 | int size = uimm (aarch64_get_instr (cpu), 31, 30); | |
1454 | /* int ordered = uimm (aarch64_get_instr (cpu), 15, 15); */ | |
1455 | /* int exclusive = ! uimm (aarch64_get_instr (cpu), 23, 23); */ | |
1456 | ||
1457 | switch (size) | |
1458 | { | |
1459 | case 0: | |
1460 | aarch64_set_reg_u64 (cpu, rt, NO_SP, aarch64_get_mem_u8 (cpu, address)); | |
1461 | break; | |
1462 | case 1: | |
1463 | aarch64_set_reg_u64 (cpu, rt, NO_SP, aarch64_get_mem_u16 (cpu, address)); | |
1464 | break; | |
1465 | case 2: | |
1466 | aarch64_set_reg_u64 (cpu, rt, NO_SP, aarch64_get_mem_u32 (cpu, address)); | |
1467 | break; | |
1468 | case 3: | |
1469 | aarch64_set_reg_u64 (cpu, rt, NO_SP, aarch64_get_mem_u64 (cpu, address)); | |
1470 | break; | |
1471 | default: | |
1472 | HALT_UNALLOC; | |
1473 | } | |
1474 | } | |
1475 | ||
1476 | static void | |
1477 | stxr (sim_cpu *cpu) | |
1478 | { | |
1479 | unsigned rn = uimm (aarch64_get_instr (cpu), 9, 5); | |
1480 | unsigned rt = uimm (aarch64_get_instr (cpu), 4, 0); | |
1481 | unsigned rs = uimm (aarch64_get_instr (cpu), 20, 16); | |
1482 | uint64_t address = aarch64_get_reg_u64 (cpu, rn, SP_OK); | |
1483 | int size = uimm (aarch64_get_instr (cpu), 31, 30); | |
1484 | uint64_t data = aarch64_get_reg_u64 (cpu, rt, NO_SP); | |
1485 | ||
1486 | switch (size) | |
1487 | { | |
1488 | case 0: aarch64_set_mem_u8 (cpu, address, data); break; | |
1489 | case 1: aarch64_set_mem_u16 (cpu, address, data); break; | |
1490 | case 2: aarch64_set_mem_u32 (cpu, address, data); break; | |
1491 | case 3: aarch64_set_mem_u64 (cpu, address, data); break; | |
1492 | default: HALT_UNALLOC; | |
1493 | } | |
1494 | ||
1495 | aarch64_set_reg_u64 (cpu, rs, NO_SP, 0); /* Always exclusive... */ | |
1496 | } | |
1497 | ||
1498 | static void | |
1499 | dexLoadLiteral (sim_cpu *cpu) | |
1500 | { | |
1501 | /* instr[29,27] == 011 | |
1502 | instr[25,24] == 00 | |
1503 | instr[31,30:26] = opc: 000 ==> LDRW, 001 ==> FLDRS | |
1504 | 010 ==> LDRX, 011 ==> FLDRD | |
1505 | 100 ==> LDRSW, 101 ==> FLDRQ | |
1506 | 110 ==> PRFM, 111 ==> UNALLOC | |
1507 | instr[26] ==> V : 0 ==> GReg, 1 ==> FReg | |
1508 | instr[23, 5] == simm19 */ | |
1509 | ||
1510 | /* unsigned rt = uimm (aarch64_get_instr (cpu), 4, 0); */ | |
1511 | uint32_t dispatch = ( (uimm (aarch64_get_instr (cpu), 31, 30) << 1) | |
1512 | | uimm (aarch64_get_instr (cpu), 26, 26)); | |
1513 | int32_t imm = simm32 (aarch64_get_instr (cpu), 23, 5); | |
1514 | ||
1515 | switch (dispatch) | |
1516 | { | |
1517 | case 0: ldr32_pcrel (cpu, imm); break; | |
1518 | case 1: fldrs_pcrel (cpu, imm); break; | |
1519 | case 2: ldr_pcrel (cpu, imm); break; | |
1520 | case 3: fldrd_pcrel (cpu, imm); break; | |
1521 | case 4: ldrsw_pcrel (cpu, imm); break; | |
1522 | case 5: fldrq_pcrel (cpu, imm); break; | |
1523 | case 6: prfm_pcrel (cpu, imm); break; | |
1524 | case 7: | |
1525 | default: | |
1526 | HALT_UNALLOC; | |
1527 | } | |
1528 | } | |
1529 | ||
1530 | /* Immediate arithmetic | |
1531 | The aimm argument is a 12 bit unsigned value or a 12 bit unsigned | |
1532 | value left shifted by 12 bits (done at decode). | |
1533 | ||
1534 | N.B. the register args (dest, source) can normally be Xn or SP. | |
1535 | the exception occurs for flag setting instructions which may | |
1536 | only use Xn for the output (dest). */ | |
1537 | ||
1538 | /* 32 bit add immediate. */ | |
1539 | static void | |
1540 | add32 (sim_cpu *cpu, uint32_t aimm) | |
1541 | { | |
1542 | unsigned rn = uimm (aarch64_get_instr (cpu), 9, 5); | |
1543 | unsigned rd = uimm (aarch64_get_instr (cpu), 4, 0); | |
1544 | ||
1545 | aarch64_set_reg_u64 (cpu, rd, SP_OK, | |
1546 | aarch64_get_reg_u32 (cpu, rn, SP_OK) + aimm); | |
1547 | } | |
1548 | ||
1549 | /* 64 bit add immediate. */ | |
1550 | static void | |
1551 | add64 (sim_cpu *cpu, uint32_t aimm) | |
1552 | { | |
1553 | unsigned rn = uimm (aarch64_get_instr (cpu), 9, 5); | |
1554 | unsigned rd = uimm (aarch64_get_instr (cpu), 4, 0); | |
1555 | ||
1556 | aarch64_set_reg_u64 (cpu, rd, SP_OK, | |
1557 | aarch64_get_reg_u64 (cpu, rn, SP_OK) + aimm); | |
1558 | } | |
1559 | ||
1560 | static void | |
1561 | set_flags_for_add32 (sim_cpu *cpu, int32_t value1, int32_t value2) | |
1562 | { | |
1563 | int32_t result = value1 + value2; | |
1564 | int64_t sresult = (int64_t) value1 + (int64_t) value2; | |
1565 | uint64_t uresult = (uint64_t)(uint32_t) value1 | |
1566 | + (uint64_t)(uint32_t) value2; | |
1567 | uint32_t flags = 0; | |
1568 | ||
1569 | if (result == 0) | |
1570 | flags |= Z; | |
1571 | ||
1572 | if (result & (1 << 31)) | |
1573 | flags |= N; | |
1574 | ||
1575 | if (uresult != result) | |
1576 | flags |= C; | |
1577 | ||
1578 | if (sresult != result) | |
1579 | flags |= V; | |
1580 | ||
1581 | aarch64_set_CPSR (cpu, flags); | |
1582 | } | |
1583 | ||
1584 | static void | |
1585 | set_flags_for_add64 (sim_cpu *cpu, uint64_t value1, uint64_t value2) | |
1586 | { | |
1587 | int64_t sval1 = value1; | |
1588 | int64_t sval2 = value2; | |
1589 | uint64_t result = value1 + value2; | |
1590 | int64_t sresult = sval1 + sval2; | |
1591 | uint32_t flags = 0; | |
1592 | ||
1593 | if (result == 0) | |
1594 | flags |= Z; | |
1595 | ||
1596 | if (result & (1ULL << 63)) | |
1597 | flags |= N; | |
1598 | ||
1599 | if (sval1 < 0) | |
1600 | { | |
1601 | if (sval2 < 0) | |
1602 | { | |
1603 | /* Negative plus a negative. Overflow happens if | |
1604 | the result is greater than either of the operands. */ | |
1605 | if (sresult > sval1 || sresult > sval2) | |
1606 | flags |= V; | |
1607 | } | |
1608 | /* else Negative plus a positive. Overflow cannot happen. */ | |
1609 | } | |
1610 | else /* value1 is +ve. */ | |
1611 | { | |
1612 | if (sval2 < 0) | |
1613 | { | |
1614 | /* Overflow can only occur if we computed "0 - MININT". */ | |
1615 | if (sval1 == 0 && sval2 == (1LL << 63)) | |
1616 | flags |= V; | |
1617 | } | |
1618 | else | |
1619 | { | |
1620 | /* Postive plus positive - overflow has happened if the | |
1621 | result is smaller than either of the operands. */ | |
1622 | if (result < value1 || result < value2) | |
1623 | flags |= V | C; | |
1624 | } | |
1625 | } | |
1626 | ||
1627 | aarch64_set_CPSR (cpu, flags); | |
1628 | } | |
1629 | ||
1630 | #define NEG(a) (((a) & signbit) == signbit) | |
1631 | #define POS(a) (((a) & signbit) == 0) | |
1632 | ||
1633 | static void | |
1634 | set_flags_for_sub32 (sim_cpu *cpu, uint32_t value1, uint32_t value2) | |
1635 | { | |
1636 | uint32_t result = value1 - value2; | |
1637 | uint32_t flags = 0; | |
1638 | uint32_t signbit = 1ULL << 31; | |
1639 | ||
1640 | if (result == 0) | |
1641 | flags |= Z; | |
1642 | ||
1643 | if (NEG (result)) | |
1644 | flags |= N; | |
1645 | ||
1646 | if ( (NEG (value1) && POS (value2)) | |
1647 | || (NEG (value1) && POS (result)) | |
1648 | || (POS (value2) && POS (result))) | |
1649 | flags |= C; | |
1650 | ||
1651 | if ( (NEG (value1) && POS (value2) && POS (result)) | |
1652 | || (POS (value1) && NEG (value2) && NEG (result))) | |
1653 | flags |= V; | |
1654 | ||
1655 | aarch64_set_CPSR (cpu, flags); | |
1656 | } | |
1657 | ||
1658 | static void | |
1659 | set_flags_for_sub64 (sim_cpu *cpu, uint64_t value1, uint64_t value2) | |
1660 | { | |
1661 | uint64_t result = value1 - value2; | |
1662 | uint32_t flags = 0; | |
1663 | uint64_t signbit = 1ULL << 63; | |
1664 | ||
1665 | if (result == 0) | |
1666 | flags |= Z; | |
1667 | ||
1668 | if (NEG (result)) | |
1669 | flags |= N; | |
1670 | ||
1671 | if ( (NEG (value1) && POS (value2)) | |
1672 | || (NEG (value1) && POS (result)) | |
1673 | || (POS (value2) && POS (result))) | |
1674 | flags |= C; | |
1675 | ||
1676 | if ( (NEG (value1) && POS (value2) && POS (result)) | |
1677 | || (POS (value1) && NEG (value2) && NEG (result))) | |
1678 | flags |= V; | |
1679 | ||
1680 | aarch64_set_CPSR (cpu, flags); | |
1681 | } | |
1682 | ||
1683 | static void | |
1684 | set_flags_for_binop32 (sim_cpu *cpu, uint32_t result) | |
1685 | { | |
1686 | uint32_t flags = 0; | |
1687 | ||
1688 | if (result == 0) | |
1689 | flags |= Z; | |
1690 | else | |
1691 | flags &= ~ Z; | |
1692 | ||
1693 | if (result & (1 << 31)) | |
1694 | flags |= N; | |
1695 | else | |
1696 | flags &= ~ N; | |
1697 | ||
1698 | aarch64_set_CPSR (cpu, flags); | |
1699 | } | |
1700 | ||
1701 | static void | |
1702 | set_flags_for_binop64 (sim_cpu *cpu, uint64_t result) | |
1703 | { | |
1704 | uint32_t flags = 0; | |
1705 | ||
1706 | if (result == 0) | |
1707 | flags |= Z; | |
1708 | else | |
1709 | flags &= ~ Z; | |
1710 | ||
1711 | if (result & (1ULL << 63)) | |
1712 | flags |= N; | |
1713 | else | |
1714 | flags &= ~ N; | |
1715 | ||
1716 | aarch64_set_CPSR (cpu, flags); | |
1717 | } | |
1718 | ||
1719 | /* 32 bit add immediate set flags. */ | |
1720 | static void | |
1721 | adds32 (sim_cpu *cpu, uint32_t aimm) | |
1722 | { | |
1723 | unsigned rn = uimm (aarch64_get_instr (cpu), 9, 5); | |
1724 | unsigned rd = uimm (aarch64_get_instr (cpu), 4, 0); | |
1725 | /* TODO : do we need to worry about signs here? */ | |
1726 | int32_t value1 = aarch64_get_reg_s32 (cpu, rn, SP_OK); | |
1727 | ||
1728 | aarch64_set_reg_u64 (cpu, rd, NO_SP, value1 + aimm); | |
1729 | set_flags_for_add32 (cpu, value1, aimm); | |
1730 | } | |
1731 | ||
1732 | /* 64 bit add immediate set flags. */ | |
1733 | static void | |
1734 | adds64 (sim_cpu *cpu, uint32_t aimm) | |
1735 | { | |
1736 | unsigned rn = uimm (aarch64_get_instr (cpu), 9, 5); | |
1737 | unsigned rd = uimm (aarch64_get_instr (cpu), 4, 0); | |
1738 | uint64_t value1 = aarch64_get_reg_u64 (cpu, rn, SP_OK); | |
1739 | uint64_t value2 = aimm; | |
1740 | ||
1741 | aarch64_set_reg_u64 (cpu, rd, NO_SP, value1 + value2); | |
1742 | set_flags_for_add64 (cpu, value1, value2); | |
1743 | } | |
1744 | ||
1745 | /* 32 bit sub immediate. */ | |
1746 | static void | |
1747 | sub32 (sim_cpu *cpu, uint32_t aimm) | |
1748 | { | |
1749 | unsigned rn = uimm (aarch64_get_instr (cpu), 9, 5); | |
1750 | unsigned rd = uimm (aarch64_get_instr (cpu), 4, 0); | |
1751 | ||
1752 | aarch64_set_reg_u64 (cpu, rd, SP_OK, | |
1753 | aarch64_get_reg_u32 (cpu, rn, SP_OK) - aimm); | |
1754 | } | |
1755 | ||
1756 | /* 64 bit sub immediate. */ | |
1757 | static void | |
1758 | sub64 (sim_cpu *cpu, uint32_t aimm) | |
1759 | { | |
1760 | unsigned rn = uimm (aarch64_get_instr (cpu), 9, 5); | |
1761 | unsigned rd = uimm (aarch64_get_instr (cpu), 4, 0); | |
1762 | ||
1763 | aarch64_set_reg_u64 (cpu, rd, SP_OK, | |
1764 | aarch64_get_reg_u64 (cpu, rn, SP_OK) - aimm); | |
1765 | } | |
1766 | ||
1767 | /* 32 bit sub immediate set flags. */ | |
1768 | static void | |
1769 | subs32 (sim_cpu *cpu, uint32_t aimm) | |
1770 | { | |
1771 | unsigned rn = uimm (aarch64_get_instr (cpu), 9, 5); | |
1772 | unsigned rd = uimm (aarch64_get_instr (cpu), 4, 0); | |
1773 | uint32_t value1 = aarch64_get_reg_u64 (cpu, rn, SP_OK); | |
1774 | uint32_t value2 = aimm; | |
1775 | ||
1776 | aarch64_set_reg_u64 (cpu, rd, NO_SP, value1 - value2); | |
1777 | set_flags_for_sub32 (cpu, value1, value2); | |
1778 | } | |
1779 | ||
1780 | /* 64 bit sub immediate set flags. */ | |
1781 | static void | |
1782 | subs64 (sim_cpu *cpu, uint32_t aimm) | |
1783 | { | |
1784 | unsigned rn = uimm (aarch64_get_instr (cpu), 9, 5); | |
1785 | unsigned rd = uimm (aarch64_get_instr (cpu), 4, 0); | |
1786 | uint64_t value1 = aarch64_get_reg_u64 (cpu, rn, SP_OK); | |
1787 | uint32_t value2 = aimm; | |
1788 | ||
1789 | aarch64_set_reg_u64 (cpu, rd, NO_SP, value1 - value2); | |
1790 | set_flags_for_sub64 (cpu, value1, value2); | |
1791 | } | |
1792 | ||
1793 | /* Data Processing Register. */ | |
1794 | ||
1795 | /* First two helpers to perform the shift operations. */ | |
1796 | ||
1797 | static inline uint32_t | |
1798 | shifted32 (uint32_t value, Shift shift, uint32_t count) | |
1799 | { | |
1800 | switch (shift) | |
1801 | { | |
1802 | default: | |
1803 | case LSL: | |
1804 | return (value << count); | |
1805 | case LSR: | |
1806 | return (value >> count); | |
1807 | case ASR: | |
1808 | { | |
1809 | int32_t svalue = value; | |
1810 | return (svalue >> count); | |
1811 | } | |
1812 | case ROR: | |
1813 | { | |
1814 | uint32_t top = value >> count; | |
1815 | uint32_t bottom = value << (32 - count); | |
1816 | return (bottom | top); | |
1817 | } | |
1818 | } | |
1819 | } | |
1820 | ||
1821 | static inline uint64_t | |
1822 | shifted64 (uint64_t value, Shift shift, uint32_t count) | |
1823 | { | |
1824 | switch (shift) | |
1825 | { | |
1826 | default: | |
1827 | case LSL: | |
1828 | return (value << count); | |
1829 | case LSR: | |
1830 | return (value >> count); | |
1831 | case ASR: | |
1832 | { | |
1833 | int64_t svalue = value; | |
1834 | return (svalue >> count); | |
1835 | } | |
1836 | case ROR: | |
1837 | { | |
1838 | uint64_t top = value >> count; | |
1839 | uint64_t bottom = value << (64 - count); | |
1840 | return (bottom | top); | |
1841 | } | |
1842 | } | |
1843 | } | |
1844 | ||
1845 | /* Arithmetic shifted register. | |
1846 | These allow an optional LSL, ASR or LSR to the second source | |
1847 | register with a count up to the register bit count. | |
1848 | ||
1849 | N.B register args may not be SP. */ | |
1850 | ||
1851 | /* 32 bit ADD shifted register. */ | |
1852 | static void | |
1853 | add32_shift (sim_cpu *cpu, Shift shift, uint32_t count) | |
1854 | { | |
1855 | unsigned rm = uimm (aarch64_get_instr (cpu), 20, 16); | |
1856 | unsigned rn = uimm (aarch64_get_instr (cpu), 9, 5); | |
1857 | unsigned rd = uimm (aarch64_get_instr (cpu), 4, 0); | |
1858 | ||
1859 | aarch64_set_reg_u64 (cpu, rd, NO_SP, | |
1860 | aarch64_get_reg_u32 (cpu, rn, NO_SP) | |
1861 | + shifted32 (aarch64_get_reg_u32 (cpu, rm, NO_SP), | |
1862 | shift, count)); | |
1863 | } | |
1864 | ||
1865 | /* 64 bit ADD shifted register. */ | |
1866 | static void | |
1867 | add64_shift (sim_cpu *cpu, Shift shift, uint32_t count) | |
1868 | { | |
1869 | unsigned rm = uimm (aarch64_get_instr (cpu), 20, 16); | |
1870 | unsigned rn = uimm (aarch64_get_instr (cpu), 9, 5); | |
1871 | unsigned rd = uimm (aarch64_get_instr (cpu), 4, 0); | |
1872 | ||
1873 | aarch64_set_reg_u64 (cpu, rd, NO_SP, | |
1874 | aarch64_get_reg_u64 (cpu, rn, NO_SP) | |
1875 | + shifted64 (aarch64_get_reg_u64 (cpu, rm, NO_SP), | |
1876 | shift, count)); | |
1877 | } | |
1878 | ||
1879 | /* 32 bit ADD shifted register setting flags. */ | |
1880 | static void | |
1881 | adds32_shift (sim_cpu *cpu, Shift shift, uint32_t count) | |
1882 | { | |
1883 | unsigned rm = uimm (aarch64_get_instr (cpu), 20, 16); | |
1884 | unsigned rn = uimm (aarch64_get_instr (cpu), 9, 5); | |
1885 | unsigned rd = uimm (aarch64_get_instr (cpu), 4, 0); | |
1886 | ||
1887 | uint32_t value1 = aarch64_get_reg_u32 (cpu, rn, NO_SP); | |
1888 | uint32_t value2 = shifted32 (aarch64_get_reg_u32 (cpu, rm, NO_SP), | |
1889 | shift, count); | |
1890 | ||
1891 | aarch64_set_reg_u64 (cpu, rd, NO_SP, value1 + value2); | |
1892 | set_flags_for_add32 (cpu, value1, value2); | |
1893 | } | |
1894 | ||
1895 | /* 64 bit ADD shifted register setting flags. */ | |
1896 | static void | |
1897 | adds64_shift (sim_cpu *cpu, Shift shift, uint32_t count) | |
1898 | { | |
1899 | unsigned rm = uimm (aarch64_get_instr (cpu), 20, 16); | |
1900 | unsigned rn = uimm (aarch64_get_instr (cpu), 9, 5); | |
1901 | unsigned rd = uimm (aarch64_get_instr (cpu), 4, 0); | |
1902 | ||
1903 | uint64_t value1 = aarch64_get_reg_u64 (cpu, rn, NO_SP); | |
1904 | uint64_t value2 = shifted64 (aarch64_get_reg_u64 (cpu, rm, NO_SP), | |
1905 | shift, count); | |
1906 | ||
1907 | aarch64_set_reg_u64 (cpu, rd, NO_SP, value1 + value2); | |
1908 | set_flags_for_add64 (cpu, value1, value2); | |
1909 | } | |
1910 | ||
1911 | /* 32 bit SUB shifted register. */ | |
1912 | static void | |
1913 | sub32_shift (sim_cpu *cpu, Shift shift, uint32_t count) | |
1914 | { | |
1915 | unsigned rm = uimm (aarch64_get_instr (cpu), 20, 16); | |
1916 | unsigned rn = uimm (aarch64_get_instr (cpu), 9, 5); | |
1917 | unsigned rd = uimm (aarch64_get_instr (cpu), 4, 0); | |
1918 | ||
1919 | aarch64_set_reg_u64 (cpu, rd, NO_SP, | |
1920 | aarch64_get_reg_u32 (cpu, rn, NO_SP) | |
1921 | - shifted32 (aarch64_get_reg_u32 (cpu, rm, NO_SP), | |
1922 | shift, count)); | |
1923 | } | |
1924 | ||
1925 | /* 64 bit SUB shifted register. */ | |
1926 | static void | |
1927 | sub64_shift (sim_cpu *cpu, Shift shift, uint32_t count) | |
1928 | { | |
1929 | unsigned rm = uimm (aarch64_get_instr (cpu), 20, 16); | |
1930 | unsigned rn = uimm (aarch64_get_instr (cpu), 9, 5); | |
1931 | unsigned rd = uimm (aarch64_get_instr (cpu), 4, 0); | |
1932 | ||
1933 | aarch64_set_reg_u64 (cpu, rd, NO_SP, | |
1934 | aarch64_get_reg_u64 (cpu, rn, NO_SP) | |
1935 | - shifted64 (aarch64_get_reg_u64 (cpu, rm, NO_SP), | |
1936 | shift, count)); | |
1937 | } | |
1938 | ||
1939 | /* 32 bit SUB shifted register setting flags. */ | |
1940 | static void | |
1941 | subs32_shift (sim_cpu *cpu, Shift shift, uint32_t count) | |
1942 | { | |
1943 | unsigned rm = uimm (aarch64_get_instr (cpu), 20, 16); | |
1944 | unsigned rn = uimm (aarch64_get_instr (cpu), 9, 5); | |
1945 | unsigned rd = uimm (aarch64_get_instr (cpu), 4, 0); | |
1946 | ||
1947 | uint32_t value1 = aarch64_get_reg_u32 (cpu, rn, NO_SP); | |
1948 | uint32_t value2 = shifted32 (aarch64_get_reg_u32 (cpu, rm, NO_SP), | |
1949 | shift, count); | |
1950 | ||
1951 | aarch64_set_reg_u64 (cpu, rd, NO_SP, value1 - value2); | |
1952 | set_flags_for_sub32 (cpu, value1, value2); | |
1953 | } | |
1954 | ||
1955 | /* 64 bit SUB shifted register setting flags. */ | |
1956 | static void | |
1957 | subs64_shift (sim_cpu *cpu, Shift shift, uint32_t count) | |
1958 | { | |
1959 | unsigned rm = uimm (aarch64_get_instr (cpu), 20, 16); | |
1960 | unsigned rn = uimm (aarch64_get_instr (cpu), 9, 5); | |
1961 | unsigned rd = uimm (aarch64_get_instr (cpu), 4, 0); | |
1962 | ||
1963 | uint64_t value1 = aarch64_get_reg_u64 (cpu, rn, NO_SP); | |
1964 | uint64_t value2 = shifted64 (aarch64_get_reg_u64 (cpu, rm, NO_SP), | |
1965 | shift, count); | |
1966 | ||
1967 | aarch64_set_reg_u64 (cpu, rd, NO_SP, value1 - value2); | |
1968 | set_flags_for_sub64 (cpu, value1, value2); | |
1969 | } | |
1970 | ||
1971 | /* First a couple more helpers to fetch the | |
1972 | relevant source register element either | |
1973 | sign or zero extended as required by the | |
1974 | extension value. */ | |
1975 | ||
1976 | static uint32_t | |
1977 | extreg32 (sim_cpu *cpu, unsigned int lo, Extension extension) | |
1978 | { | |
1979 | switch (extension) | |
1980 | { | |
1981 | case UXTB: return aarch64_get_reg_u8 (cpu, lo, NO_SP); | |
1982 | case UXTH: return aarch64_get_reg_u16 (cpu, lo, NO_SP); | |
1983 | case UXTW: /* Fall through. */ | |
1984 | case UXTX: return aarch64_get_reg_u32 (cpu, lo, NO_SP); | |
1985 | case SXTB: return aarch64_get_reg_s8 (cpu, lo, NO_SP); | |
1986 | case SXTH: return aarch64_get_reg_s16 (cpu, lo, NO_SP); | |
1987 | case SXTW: /* Fall through. */ | |
1988 | case SXTX: /* Fall through. */ | |
1989 | default: return aarch64_get_reg_s32 (cpu, lo, NO_SP); | |
1990 | } | |
1991 | } | |
1992 | ||
1993 | static uint64_t | |
1994 | extreg64 (sim_cpu *cpu, unsigned int lo, Extension extension) | |
1995 | { | |
1996 | switch (extension) | |
1997 | { | |
1998 | case UXTB: return aarch64_get_reg_u8 (cpu, lo, NO_SP); | |
1999 | case UXTH: return aarch64_get_reg_u16 (cpu, lo, NO_SP); | |
2000 | case UXTW: return aarch64_get_reg_u32 (cpu, lo, NO_SP); | |
2001 | case UXTX: return aarch64_get_reg_u64 (cpu, lo, NO_SP); | |
2002 | case SXTB: return aarch64_get_reg_s8 (cpu, lo, NO_SP); | |
2003 | case SXTH: return aarch64_get_reg_s16 (cpu, lo, NO_SP); | |
2004 | case SXTW: return aarch64_get_reg_s32 (cpu, lo, NO_SP); | |
2005 | case SXTX: | |
2006 | default: return aarch64_get_reg_s64 (cpu, lo, NO_SP); | |
2007 | } | |
2008 | } | |
2009 | ||
2010 | /* Arithmetic extending register | |
2011 | These allow an optional sign extension of some portion of the | |
2012 | second source register followed by an optional left shift of | |
2013 | between 1 and 4 bits (i.e. a shift of 0-4 bits???) | |
2014 | ||
2015 | N.B output (dest) and first input arg (source) may normally be Xn | |
2016 | or SP. However, for flag setting operations dest can only be | |
2017 | Xn. Second input registers are always Xn. */ | |
2018 | ||
2019 | /* 32 bit ADD extending register. */ | |
2020 | static void | |
2021 | add32_ext (sim_cpu *cpu, Extension extension, uint32_t shift) | |
2022 | { | |
2023 | unsigned rm = uimm (aarch64_get_instr (cpu), 20, 16); | |
2024 | unsigned rn = uimm (aarch64_get_instr (cpu), 9, 5); | |
2025 | unsigned rd = uimm (aarch64_get_instr (cpu), 4, 0); | |
2026 | ||
2027 | aarch64_set_reg_u64 (cpu, rd, SP_OK, | |
2028 | aarch64_get_reg_u32 (cpu, rn, SP_OK) | |
2029 | + (extreg32 (cpu, rm, extension) << shift)); | |
2030 | } | |
2031 | ||
2032 | /* 64 bit ADD extending register. | |
2033 | N.B. This subsumes the case with 64 bit source2 and UXTX #n or LSL #0. */ | |
2034 | static void | |
2035 | add64_ext (sim_cpu *cpu, Extension extension, uint32_t shift) | |
2036 | { | |
2037 | unsigned rm = uimm (aarch64_get_instr (cpu), 20, 16); | |
2038 | unsigned rn = uimm (aarch64_get_instr (cpu), 9, 5); | |
2039 | unsigned rd = uimm (aarch64_get_instr (cpu), 4, 0); | |
2040 | ||
2041 | aarch64_set_reg_u64 (cpu, rd, SP_OK, | |
2042 | aarch64_get_reg_u64 (cpu, rn, SP_OK) | |
2043 | + (extreg64 (cpu, rm, extension) << shift)); | |
2044 | } | |
2045 | ||
2046 | /* 32 bit ADD extending register setting flags. */ | |
2047 | static void | |
2048 | adds32_ext (sim_cpu *cpu, Extension extension, uint32_t shift) | |
2049 | { | |
2050 | unsigned rm = uimm (aarch64_get_instr (cpu), 20, 16); | |
2051 | unsigned rn = uimm (aarch64_get_instr (cpu), 9, 5); | |
2052 | unsigned rd = uimm (aarch64_get_instr (cpu), 4, 0); | |
2053 | ||
2054 | uint32_t value1 = aarch64_get_reg_u32 (cpu, rn, SP_OK); | |
2055 | uint32_t value2 = extreg32 (cpu, rm, extension) << shift; | |
2056 | ||
2057 | aarch64_set_reg_u64 (cpu, rd, NO_SP, value1 + value2); | |
2058 | set_flags_for_add32 (cpu, value1, value2); | |
2059 | } | |
2060 | ||
2061 | /* 64 bit ADD extending register setting flags */ | |
2062 | /* N.B. this subsumes the case with 64 bit source2 and UXTX #n or LSL #0 */ | |
2063 | static void | |
2064 | adds64_ext (sim_cpu *cpu, Extension extension, uint32_t shift) | |
2065 | { | |
2066 | unsigned rm = uimm (aarch64_get_instr (cpu), 20, 16); | |
2067 | unsigned rn = uimm (aarch64_get_instr (cpu), 9, 5); | |
2068 | unsigned rd = uimm (aarch64_get_instr (cpu), 4, 0); | |
2069 | ||
2070 | uint64_t value1 = aarch64_get_reg_u64 (cpu, rn, SP_OK); | |
2071 | uint64_t value2 = extreg64 (cpu, rm, extension) << shift; | |
2072 | ||
2073 | aarch64_set_reg_u64 (cpu, rd, NO_SP, value1 + value2); | |
2074 | set_flags_for_add64 (cpu, value1, value2); | |
2075 | } | |
2076 | ||
2077 | /* 32 bit SUB extending register. */ | |
2078 | static void | |
2079 | sub32_ext (sim_cpu *cpu, Extension extension, uint32_t shift) | |
2080 | { | |
2081 | unsigned rm = uimm (aarch64_get_instr (cpu), 20, 16); | |
2082 | unsigned rn = uimm (aarch64_get_instr (cpu), 9, 5); | |
2083 | unsigned rd = uimm (aarch64_get_instr (cpu), 4, 0); | |
2084 | ||
2085 | aarch64_set_reg_u64 (cpu, rd, SP_OK, | |
2086 | aarch64_get_reg_u32 (cpu, rn, SP_OK) | |
2087 | - (extreg32 (cpu, rm, extension) << shift)); | |
2088 | } | |
2089 | ||
2090 | /* 64 bit SUB extending register. */ | |
2091 | /* N.B. this subsumes the case with 64 bit source2 and UXTX #n or LSL #0. */ | |
2092 | static void | |
2093 | sub64_ext (sim_cpu *cpu, Extension extension, uint32_t shift) | |
2094 | { | |
2095 | unsigned rm = uimm (aarch64_get_instr (cpu), 20, 16); | |
2096 | unsigned rn = uimm (aarch64_get_instr (cpu), 9, 5); | |
2097 | unsigned rd = uimm (aarch64_get_instr (cpu), 4, 0); | |
2098 | ||
2099 | aarch64_set_reg_u64 (cpu, rd, SP_OK, | |
2100 | aarch64_get_reg_u64 (cpu, rn, SP_OK) | |
2101 | - (extreg64 (cpu, rm, extension) << shift)); | |
2102 | } | |
2103 | ||
2104 | /* 32 bit SUB extending register setting flags. */ | |
2105 | static void | |
2106 | subs32_ext (sim_cpu *cpu, Extension extension, uint32_t shift) | |
2107 | { | |
2108 | unsigned rm = uimm (aarch64_get_instr (cpu), 20, 16); | |
2109 | unsigned rn = uimm (aarch64_get_instr (cpu), 9, 5); | |
2110 | unsigned rd = uimm (aarch64_get_instr (cpu), 4, 0); | |
2111 | ||
2112 | uint32_t value1 = aarch64_get_reg_u32 (cpu, rn, SP_OK); | |
2113 | uint32_t value2 = extreg32 (cpu, rm, extension) << shift; | |
2114 | ||
2115 | aarch64_set_reg_u64 (cpu, rd, NO_SP, value1 - value2); | |
2116 | set_flags_for_sub32 (cpu, value1, value2); | |
2117 | } | |
2118 | ||
2119 | /* 64 bit SUB extending register setting flags */ | |
2120 | /* N.B. this subsumes the case with 64 bit source2 and UXTX #n or LSL #0 */ | |
2121 | static void | |
2122 | subs64_ext (sim_cpu *cpu, Extension extension, uint32_t shift) | |
2123 | { | |
2124 | unsigned rm = uimm (aarch64_get_instr (cpu), 20, 16); | |
2125 | unsigned rn = uimm (aarch64_get_instr (cpu), 9, 5); | |
2126 | unsigned rd = uimm (aarch64_get_instr (cpu), 4, 0); | |
2127 | ||
2128 | uint64_t value1 = aarch64_get_reg_u64 (cpu, rn, SP_OK); | |
2129 | uint64_t value2 = extreg64 (cpu, rm, extension) << shift; | |
2130 | ||
2131 | aarch64_set_reg_u64 (cpu, rd, NO_SP, value1 - value2); | |
2132 | set_flags_for_sub64 (cpu, value1, value2); | |
2133 | } | |
2134 | ||
2135 | static void | |
2136 | dexAddSubtractImmediate (sim_cpu *cpu) | |
2137 | { | |
2138 | /* instr[31] = size : 0 ==> 32 bit, 1 ==> 64 bit | |
2139 | instr[30] = op : 0 ==> ADD, 1 ==> SUB | |
2140 | instr[29] = set : 0 ==> no flags, 1 ==> set flags | |
2141 | instr[28,24] = 10001 | |
2142 | instr[23,22] = shift : 00 == LSL#0, 01 = LSL#12 1x = UNALLOC | |
2143 | instr[21,10] = uimm12 | |
2144 | instr[9,5] = Rn | |
2145 | instr[4,0] = Rd */ | |
2146 | ||
2147 | /* N.B. the shift is applied at decode before calling the add/sub routine. */ | |
2148 | uint32_t shift = uimm (aarch64_get_instr (cpu), 23, 22); | |
2149 | uint32_t imm = uimm (aarch64_get_instr (cpu), 21, 10); | |
2150 | uint32_t dispatch = uimm (aarch64_get_instr (cpu), 31, 29); | |
2151 | ||
2152 | NYI_assert (28, 24, 0x11); | |
2153 | ||
2154 | if (shift > 1) | |
2155 | HALT_UNALLOC; | |
2156 | ||
2157 | if (shift) | |
2158 | imm <<= 12; | |
2159 | ||
2160 | switch (dispatch) | |
2161 | { | |
2162 | case 0: add32 (cpu, imm); break; | |
2163 | case 1: adds32 (cpu, imm); break; | |
2164 | case 2: sub32 (cpu, imm); break; | |
2165 | case 3: subs32 (cpu, imm); break; | |
2166 | case 4: add64 (cpu, imm); break; | |
2167 | case 5: adds64 (cpu, imm); break; | |
2168 | case 6: sub64 (cpu, imm); break; | |
2169 | case 7: subs64 (cpu, imm); break; | |
2170 | default: | |
2171 | HALT_UNALLOC; | |
2172 | } | |
2173 | } | |
2174 | ||
2175 | static void | |
2176 | dexAddSubtractShiftedRegister (sim_cpu *cpu) | |
2177 | { | |
2178 | /* instr[31] = size : 0 ==> 32 bit, 1 ==> 64 bit | |
2179 | instr[30,29] = op : 00 ==> ADD, 01 ==> ADDS, 10 ==> SUB, 11 ==> SUBS | |
2180 | instr[28,24] = 01011 | |
2181 | instr[23,22] = shift : 0 ==> LSL, 1 ==> LSR, 2 ==> ASR, 3 ==> UNALLOC | |
2182 | instr[21] = 0 | |
2183 | instr[20,16] = Rm | |
2184 | instr[15,10] = count : must be 0xxxxx for 32 bit | |
2185 | instr[9,5] = Rn | |
2186 | instr[4,0] = Rd */ | |
2187 | ||
2188 | uint32_t size = uimm (aarch64_get_instr (cpu), 31, 31); | |
2189 | /* 32 bit operations must have count[5] = 0 | |
2190 | or else we have an UNALLOC. */ | |
2191 | uint32_t count = uimm (aarch64_get_instr (cpu), 15, 10); | |
2192 | /* Shift encoded as ROR is unallocated. */ | |
2193 | Shift shiftType = shift (aarch64_get_instr (cpu), 22); | |
2194 | /* Dispatch on size:op i.e aarch64_get_instr (cpu)[31,29]. */ | |
2195 | uint32_t dispatch = uimm (aarch64_get_instr (cpu), 31, 29); | |
2196 | ||
2197 | NYI_assert (28, 24, 0x0B); | |
2198 | NYI_assert (21, 21, 0); | |
2199 | ||
2200 | if (shiftType == ROR) | |
2201 | HALT_UNALLOC; | |
2202 | ||
2203 | if (!size && uimm (count, 5, 5)) | |
2204 | HALT_UNALLOC; | |
2205 | ||
2206 | switch (dispatch) | |
2207 | { | |
2208 | case 0: add32_shift (cpu, shiftType, count); break; | |
2209 | case 1: adds32_shift (cpu, shiftType, count); break; | |
2210 | case 2: sub32_shift (cpu, shiftType, count); break; | |
2211 | case 3: subs32_shift (cpu, shiftType, count); break; | |
2212 | case 4: add64_shift (cpu, shiftType, count); break; | |
2213 | case 5: adds64_shift (cpu, shiftType, count); break; | |
2214 | case 6: sub64_shift (cpu, shiftType, count); break; | |
2215 | case 7: subs64_shift (cpu, shiftType, count); break; | |
2216 | default: | |
2217 | HALT_UNALLOC; | |
2218 | } | |
2219 | } | |
2220 | ||
2221 | static void | |
2222 | dexAddSubtractExtendedRegister (sim_cpu *cpu) | |
2223 | { | |
2224 | /* instr[31] = size : 0 ==> 32 bit, 1 ==> 64 bit | |
2225 | instr[30] = op : 0 ==> ADD, 1 ==> SUB | |
2226 | instr[29] = set? : 0 ==> no flags, 1 ==> set flags | |
2227 | instr[28,24] = 01011 | |
2228 | instr[23,22] = opt : 0 ==> ok, 1,2,3 ==> UNALLOC | |
2229 | instr[21] = 1 | |
2230 | instr[20,16] = Rm | |
2231 | instr[15,13] = option : 000 ==> UXTB, 001 ==> UXTH, | |
2232 | 000 ==> LSL|UXTW, 001 ==> UXTZ, | |
2233 | 000 ==> SXTB, 001 ==> SXTH, | |
2234 | 000 ==> SXTW, 001 ==> SXTX, | |
2235 | instr[12,10] = shift : 0,1,2,3,4 ==> ok, 5,6,7 ==> UNALLOC | |
2236 | instr[9,5] = Rn | |
2237 | instr[4,0] = Rd */ | |
2238 | ||
2239 | Extension extensionType = extension (aarch64_get_instr (cpu), 13); | |
2240 | uint32_t shift = uimm (aarch64_get_instr (cpu), 12, 10); | |
2241 | /* dispatch on size:op:set? i.e aarch64_get_instr (cpu)[31,29] */ | |
2242 | uint32_t dispatch = uimm (aarch64_get_instr (cpu), 31, 29); | |
2243 | ||
2244 | NYI_assert (28, 24, 0x0B); | |
2245 | NYI_assert (21, 21, 1); | |
2246 | ||
2247 | /* Shift may not exceed 4. */ | |
2248 | if (shift > 4) | |
2249 | HALT_UNALLOC; | |
2250 | ||
2251 | switch (dispatch) | |
2252 | { | |
2253 | case 0: add32_ext (cpu, extensionType, shift); break; | |
2254 | case 1: adds32_ext (cpu, extensionType, shift); break; | |
2255 | case 2: sub32_ext (cpu, extensionType, shift); break; | |
2256 | case 3: subs32_ext (cpu, extensionType, shift); break; | |
2257 | case 4: add64_ext (cpu, extensionType, shift); break; | |
2258 | case 5: adds64_ext (cpu, extensionType, shift); break; | |
2259 | case 6: sub64_ext (cpu, extensionType, shift); break; | |
2260 | case 7: subs64_ext (cpu, extensionType, shift); break; | |
2261 | default: HALT_UNALLOC; | |
2262 | } | |
2263 | } | |
2264 | ||
2265 | /* Conditional data processing | |
2266 | Condition register is implicit 3rd source. */ | |
2267 | ||
2268 | /* 32 bit add with carry. */ | |
2269 | /* N.B register args may not be SP. */ | |
2270 | ||
2271 | static void | |
2272 | adc32 (sim_cpu *cpu) | |
2273 | { | |
2274 | unsigned rm = uimm (aarch64_get_instr (cpu), 20, 16); | |
2275 | unsigned rn = uimm (aarch64_get_instr (cpu), 9, 5); | |
2276 | unsigned rd = uimm (aarch64_get_instr (cpu), 4, 0); | |
2277 | ||
2278 | aarch64_set_reg_u64 (cpu, rd, NO_SP, | |
2279 | aarch64_get_reg_u32 (cpu, rn, NO_SP) | |
2280 | + aarch64_get_reg_u32 (cpu, rm, NO_SP) | |
2281 | + IS_SET (C)); | |
2282 | } | |
2283 | ||
2284 | /* 64 bit add with carry */ | |
2285 | static void | |
2286 | adc64 (sim_cpu *cpu) | |
2287 | { | |
2288 | unsigned rm = uimm (aarch64_get_instr (cpu), 20, 16); | |
2289 | unsigned rn = uimm (aarch64_get_instr (cpu), 9, 5); | |
2290 | unsigned rd = uimm (aarch64_get_instr (cpu), 4, 0); | |
2291 | ||
2292 | aarch64_set_reg_u64 (cpu, rd, NO_SP, | |
2293 | aarch64_get_reg_u64 (cpu, rn, NO_SP) | |
2294 | + aarch64_get_reg_u64 (cpu, rm, NO_SP) | |
2295 | + IS_SET (C)); | |
2296 | } | |
2297 | ||
2298 | /* 32 bit add with carry setting flags. */ | |
2299 | static void | |
2300 | adcs32 (sim_cpu *cpu) | |
2301 | { | |
2302 | unsigned rm = uimm (aarch64_get_instr (cpu), 20, 16); | |
2303 | unsigned rn = uimm (aarch64_get_instr (cpu), 9, 5); | |
2304 | unsigned rd = uimm (aarch64_get_instr (cpu), 4, 0); | |
2305 | ||
2306 | uint32_t value1 = aarch64_get_reg_u32 (cpu, rn, NO_SP); | |
2307 | uint32_t value2 = aarch64_get_reg_u32 (cpu, rm, NO_SP); | |
2308 | uint32_t carry = IS_SET (C); | |
2309 | ||
2310 | aarch64_set_reg_u64 (cpu, rd, NO_SP, value1 + value2 + carry); | |
2311 | set_flags_for_add32 (cpu, value1, value2 + carry); | |
2312 | } | |
2313 | ||
2314 | /* 64 bit add with carry setting flags. */ | |
2315 | static void | |
2316 | adcs64 (sim_cpu *cpu) | |
2317 | { | |
2318 | unsigned rm = uimm (aarch64_get_instr (cpu), 20, 16); | |
2319 | unsigned rn = uimm (aarch64_get_instr (cpu), 9, 5); | |
2320 | unsigned rd = uimm (aarch64_get_instr (cpu), 4, 0); | |
2321 | ||
2322 | uint64_t value1 = aarch64_get_reg_u64 (cpu, rn, NO_SP); | |
2323 | uint64_t value2 = aarch64_get_reg_u64 (cpu, rm, NO_SP); | |
2324 | uint64_t carry = IS_SET (C); | |
2325 | ||
2326 | aarch64_set_reg_u64 (cpu, rd, NO_SP, value1 + value2 + carry); | |
2327 | set_flags_for_add64 (cpu, value1, value2 + carry); | |
2328 | } | |
2329 | ||
2330 | /* 32 bit sub with carry. */ | |
2331 | static void | |
2332 | sbc32 (sim_cpu *cpu) | |
2333 | { | |
2334 | unsigned rm = uimm (aarch64_get_instr (cpu), 20, 16); | |
2335 | unsigned rn = uimm (aarch64_get_instr (cpu), 9, 5); | |
2336 | unsigned rd = uimm (aarch64_get_instr (cpu), 4, 0); | |
2337 | ||
2338 | aarch64_set_reg_u64 (cpu, rd, NO_SP, | |
2339 | aarch64_get_reg_u32 (cpu, rn, NO_SP) | |
2340 | - aarch64_get_reg_u32 (cpu, rm, NO_SP) | |
2341 | - 1 + IS_SET (C)); | |
2342 | } | |
2343 | ||
2344 | /* 64 bit sub with carry */ | |
2345 | static void | |
2346 | sbc64 (sim_cpu *cpu) | |
2347 | { | |
2348 | unsigned rm = uimm (aarch64_get_instr (cpu), 20, 16); | |
2349 | unsigned rn = uimm (aarch64_get_instr (cpu), 9, 5); | |
2350 | unsigned rd = uimm (aarch64_get_instr (cpu), 4, 0); | |
2351 | ||
2352 | aarch64_set_reg_u64 (cpu, rd, NO_SP, | |
2353 | aarch64_get_reg_u64 (cpu, rn, NO_SP) | |
2354 | - aarch64_get_reg_u64 (cpu, rm, NO_SP) | |
2355 | - 1 + IS_SET (C)); | |
2356 | } | |
2357 | ||
2358 | /* 32 bit sub with carry setting flags */ | |
2359 | static void | |
2360 | sbcs32 (sim_cpu *cpu) | |
2361 | { | |
2362 | unsigned rm = uimm (aarch64_get_instr (cpu), 20, 16); | |
2363 | unsigned rn = uimm (aarch64_get_instr (cpu), 9, 5); | |
2364 | unsigned rd = uimm (aarch64_get_instr (cpu), 4, 0); | |
2365 | ||
2366 | uint32_t value1 = aarch64_get_reg_u32 (cpu, rn, NO_SP); | |
2367 | uint32_t value2 = aarch64_get_reg_u32 (cpu, rm, NO_SP); | |
2368 | uint32_t carry = IS_SET (C); | |
2369 | uint32_t result = value1 - value2 + 1 - carry; | |
2370 | ||
2371 | aarch64_set_reg_u64 (cpu, rd, NO_SP, result); | |
2372 | set_flags_for_sub32 (cpu, value1, value2 + 1 - carry); | |
2373 | } | |
2374 | ||
2375 | /* 64 bit sub with carry setting flags */ | |
2376 | static void | |
2377 | sbcs64 (sim_cpu *cpu) | |
2378 | { | |
2379 | unsigned rm = uimm (aarch64_get_instr (cpu), 20, 16); | |
2380 | unsigned rn = uimm (aarch64_get_instr (cpu), 9, 5); | |
2381 | unsigned rd = uimm (aarch64_get_instr (cpu), 4, 0); | |
2382 | ||
2383 | uint64_t value1 = aarch64_get_reg_u64 (cpu, rn, NO_SP); | |
2384 | uint64_t value2 = aarch64_get_reg_u64 (cpu, rm, NO_SP); | |
2385 | uint64_t carry = IS_SET (C); | |
2386 | uint64_t result = value1 - value2 + 1 - carry; | |
2387 | ||
2388 | aarch64_set_reg_u64 (cpu, rd, NO_SP, result); | |
2389 | set_flags_for_sub64 (cpu, value1, value2 + 1 - carry); | |
2390 | } | |
2391 | ||
2392 | static void | |
2393 | dexAddSubtractWithCarry (sim_cpu *cpu) | |
2394 | { | |
2395 | /* instr[31] = size : 0 ==> 32 bit, 1 ==> 64 bit | |
2396 | instr[30] = op : 0 ==> ADC, 1 ==> SBC | |
2397 | instr[29] = set? : 0 ==> no flags, 1 ==> set flags | |
2398 | instr[28,21] = 1 1010 000 | |
2399 | instr[20,16] = Rm | |
2400 | instr[15,10] = op2 : 00000 ==> ok, ow ==> UNALLOC | |
2401 | instr[9,5] = Rn | |
2402 | instr[4,0] = Rd */ | |
2403 | ||
2404 | uint32_t op2 = uimm (aarch64_get_instr (cpu), 15, 10); | |
2405 | /* Dispatch on size:op:set? i.e aarch64_get_instr (cpu)[31,29] */ | |
2406 | uint32_t dispatch = uimm (aarch64_get_instr (cpu), 31, 29); | |
2407 | ||
2408 | NYI_assert (28, 21, 0xD0); | |
2409 | ||
2410 | if (op2 != 0) | |
2411 | HALT_UNALLOC; | |
2412 | ||
2413 | switch (dispatch) | |
2414 | { | |
2415 | case 0: adc32 (cpu); break; | |
2416 | case 1: adcs32 (cpu); break; | |
2417 | case 2: sbc32 (cpu); break; | |
2418 | case 3: sbcs32 (cpu); break; | |
2419 | case 4: adc64 (cpu); break; | |
2420 | case 5: adcs64 (cpu); break; | |
2421 | case 6: sbc64 (cpu); break; | |
2422 | case 7: sbcs64 (cpu); break; | |
2423 | default: HALT_UNALLOC; | |
2424 | } | |
2425 | } | |
2426 | ||
2427 | static uint32_t | |
2428 | testConditionCode (sim_cpu *cpu, CondCode cc) | |
2429 | { | |
2430 | /* This should be reduceable to branchless logic | |
2431 | by some careful testing of bits in CC followed | |
2432 | by the requisite masking and combining of bits | |
2433 | from the flag register. | |
2434 | ||
2435 | For now we do it with a switch. */ | |
2436 | int res; | |
2437 | ||
2438 | switch (cc) | |
2439 | { | |
2440 | case EQ: res = IS_SET (Z); break; | |
2441 | case NE: res = IS_CLEAR (Z); break; | |
2442 | case CS: res = IS_SET (C); break; | |
2443 | case CC: res = IS_CLEAR (C); break; | |
2444 | case MI: res = IS_SET (N); break; | |
2445 | case PL: res = IS_CLEAR (N); break; | |
2446 | case VS: res = IS_SET (V); break; | |
2447 | case VC: res = IS_CLEAR (V); break; | |
2448 | case HI: res = IS_SET (C) && IS_CLEAR (Z); break; | |
2449 | case LS: res = IS_CLEAR (C) || IS_SET (Z); break; | |
2450 | case GE: res = IS_SET (N) == IS_SET (V); break; | |
2451 | case LT: res = IS_SET (N) != IS_SET (V); break; | |
2452 | case GT: res = IS_CLEAR (Z) && (IS_SET (N) == IS_SET (V)); break; | |
2453 | case LE: res = IS_SET (Z) || (IS_SET (N) != IS_SET (V)); break; | |
2454 | case AL: | |
2455 | case NV: | |
2456 | default: | |
2457 | res = 1; | |
2458 | break; | |
2459 | } | |
2460 | return res; | |
2461 | } | |
2462 | ||
2463 | static void | |
2464 | CondCompare (sim_cpu *cpu) /* aka: ccmp and ccmn */ | |
2465 | { | |
2466 | /* instr[31] = size : 0 ==> 32 bit, 1 ==> 64 bit | |
2467 | instr[30] = compare with positive (0) or negative value (1) | |
2468 | instr[29,21] = 1 1101 0010 | |
2469 | instr[20,16] = Rm or const | |
2470 | instr[15,12] = cond | |
2471 | instr[11] = compare reg (0) or const (1) | |
2472 | instr[10] = 0 | |
2473 | instr[9,5] = Rn | |
2474 | instr[4] = 0 | |
2475 | instr[3,0] = value for CPSR bits if the comparison does not take place. */ | |
2476 | signed int negate; | |
2477 | unsigned rm; | |
2478 | unsigned rn; | |
2479 | ||
2480 | NYI_assert (29, 21, 0x1d2); | |
2481 | NYI_assert (10, 10, 0); | |
2482 | NYI_assert (4, 4, 0); | |
2483 | ||
2484 | if (! testConditionCode (cpu, uimm (aarch64_get_instr (cpu), 15, 12))) | |
2485 | { | |
2486 | aarch64_set_CPSR (cpu, uimm (aarch64_get_instr (cpu), 3, 0)); | |
2487 | return; | |
2488 | } | |
2489 | ||
2490 | negate = uimm (aarch64_get_instr (cpu), 30, 30) ? -1 : 1; | |
2491 | rm = uimm (aarch64_get_instr (cpu), 20, 16); | |
2492 | rn = uimm (aarch64_get_instr (cpu), 9, 5); | |
2493 | ||
2494 | if (uimm (aarch64_get_instr (cpu), 31, 31)) | |
2495 | { | |
2496 | if (uimm (aarch64_get_instr (cpu), 11, 11)) | |
2497 | set_flags_for_sub64 (cpu, aarch64_get_reg_u64 (cpu, rn, SP_OK), | |
2498 | negate * (uint64_t) rm); | |
2499 | else | |
2500 | set_flags_for_sub64 (cpu, aarch64_get_reg_u64 (cpu, rn, SP_OK), | |
2501 | negate * aarch64_get_reg_u64 (cpu, rm, SP_OK)); | |
2502 | } | |
2503 | else | |
2504 | { | |
2505 | if (uimm (aarch64_get_instr (cpu), 11, 11)) | |
2506 | set_flags_for_sub32 (cpu, aarch64_get_reg_u32 (cpu, rn, SP_OK), | |
2507 | negate * rm); | |
2508 | else | |
2509 | set_flags_for_sub32 (cpu, aarch64_get_reg_u32 (cpu, rn, SP_OK), | |
2510 | negate * aarch64_get_reg_u32 (cpu, rm, SP_OK)); | |
2511 | } | |
2512 | } | |
2513 | ||
2514 | static void | |
2515 | do_vec_MOV_whole_vector (sim_cpu *cpu) | |
2516 | { | |
2517 | /* MOV Vd.T, Vs.T (alias for ORR Vd.T, Vn.T, Vm.T where Vn == Vm) | |
2518 | ||
2519 | instr[31] = 0 | |
2520 | instr[30] = half(0)/full(1) | |
2521 | instr[29,21] = 001110101 | |
2522 | instr[20,16] = Vs | |
2523 | instr[15,10] = 000111 | |
2524 | instr[9,5] = Vs | |
2525 | instr[4,0] = Vd */ | |
2526 | ||
2527 | unsigned vs = uimm (aarch64_get_instr (cpu), 9, 5); | |
2528 | unsigned vd = uimm (aarch64_get_instr (cpu), 4, 0); | |
2529 | ||
2530 | NYI_assert (29, 21, 0x075); | |
2531 | NYI_assert (15, 10, 0x07); | |
2532 | ||
2533 | if (uimm (aarch64_get_instr (cpu), 20, 16) != vs) | |
2534 | HALT_NYI; | |
2535 | ||
2536 | if (uimm (aarch64_get_instr (cpu), 30, 30)) | |
2537 | aarch64_set_vec_u64 (cpu, vd, 1, aarch64_get_vec_u64 (cpu, vs, 1)); | |
2538 | ||
2539 | aarch64_set_vec_u64 (cpu, vd, 0, aarch64_get_vec_u64 (cpu, vs, 0)); | |
2540 | } | |
2541 | ||
2542 | static void | |
2543 | do_vec_MOV_into_scalar (sim_cpu *cpu) | |
2544 | { | |
2545 | /* instr[31] = 0 | |
2546 | instr[30] = word(0)/long(1) | |
2547 | instr[29,21] = 00 1110 000 | |
2548 | instr[20,18] = element size and index | |
2549 | instr[17,10] = 00 0011 11 | |
2550 | instr[9,5] = V source | |
2551 | instr[4,0] = R dest */ | |
2552 | ||
2553 | unsigned vs = uimm (aarch64_get_instr (cpu), 9, 5); | |
2554 | unsigned rd = uimm (aarch64_get_instr (cpu), 4, 0); | |
2555 | ||
2556 | NYI_assert (29, 21, 0x070); | |
2557 | NYI_assert (17, 10, 0x0F); | |
2558 | ||
2559 | switch (uimm (aarch64_get_instr (cpu), 20, 18)) | |
2560 | { | |
2561 | case 0x2: | |
2562 | aarch64_set_reg_u64 (cpu, rd, NO_SP, aarch64_get_vec_u64 (cpu, vs, 0)); | |
2563 | break; | |
2564 | ||
2565 | case 0x6: | |
2566 | aarch64_set_reg_u64 (cpu, rd, NO_SP, aarch64_get_vec_u64 (cpu, vs, 1)); | |
2567 | break; | |
2568 | ||
2569 | case 0x1: | |
2570 | case 0x3: | |
2571 | case 0x5: | |
2572 | case 0x7: | |
2573 | aarch64_set_reg_u64 (cpu, rd, NO_SP, aarch64_get_vec_u32 | |
2574 | (cpu, vs, uimm (aarch64_get_instr (cpu), 20, 19))); | |
2575 | break; | |
2576 | ||
2577 | default: | |
2578 | HALT_NYI; | |
2579 | } | |
2580 | } | |
2581 | ||
2582 | static void | |
2583 | do_vec_INS (sim_cpu *cpu) | |
2584 | { | |
2585 | /* instr[31,21] = 01001110000 | |
2586 | instr[20,16] = element size and index | |
2587 | instr[15,10] = 000111 | |
2588 | instr[9,5] = W source | |
2589 | instr[4,0] = V dest */ | |
2590 | ||
2591 | int index; | |
2592 | unsigned rs = uimm (aarch64_get_instr (cpu), 9, 5); | |
2593 | unsigned vd = uimm (aarch64_get_instr (cpu), 4, 0); | |
2594 | ||
2595 | NYI_assert (31, 21, 0x270); | |
2596 | NYI_assert (15, 10, 0x07); | |
2597 | ||
2598 | if (uimm (aarch64_get_instr (cpu), 16, 16)) | |
2599 | { | |
2600 | index = uimm (aarch64_get_instr (cpu), 20, 17); | |
2601 | aarch64_set_vec_u8 (cpu, vd, index, | |
2602 | aarch64_get_reg_u8 (cpu, rs, NO_SP)); | |
2603 | } | |
2604 | else if (uimm (aarch64_get_instr (cpu), 17, 17)) | |
2605 | { | |
2606 | index = uimm (aarch64_get_instr (cpu), 20, 18); | |
2607 | aarch64_set_vec_u16 (cpu, vd, index, | |
2608 | aarch64_get_reg_u16 (cpu, rs, NO_SP)); | |
2609 | } | |
2610 | else if (uimm (aarch64_get_instr (cpu), 18, 18)) | |
2611 | { | |
2612 | index = uimm (aarch64_get_instr (cpu), 20, 19); | |
2613 | aarch64_set_vec_u32 (cpu, vd, index, | |
2614 | aarch64_get_reg_u32 (cpu, rs, NO_SP)); | |
2615 | } | |
2616 | else if (uimm (aarch64_get_instr (cpu), 19, 19)) | |
2617 | { | |
2618 | index = uimm (aarch64_get_instr (cpu), 20, 20); | |
2619 | aarch64_set_vec_u64 (cpu, vd, index, | |
2620 | aarch64_get_reg_u64 (cpu, rs, NO_SP)); | |
2621 | } | |
2622 | else | |
2623 | HALT_NYI; | |
2624 | } | |
2625 | ||
2626 | static void | |
2627 | do_vec_DUP_vector_into_vector (sim_cpu *cpu) | |
2628 | { | |
2629 | /* instr[31] = 0 | |
2630 | instr[30] = half(0)/full(1) | |
2631 | instr[29,21] = 00 1110 000 | |
2632 | instr[20,16] = element size and index | |
2633 | instr[15,10] = 0000 01 | |
2634 | instr[9,5] = V source | |
2635 | instr[4,0] = V dest. */ | |
2636 | ||
2637 | unsigned full = uimm (aarch64_get_instr (cpu), 30, 30); | |
2638 | unsigned vs = uimm (aarch64_get_instr (cpu), 9, 5); | |
2639 | unsigned vd = uimm (aarch64_get_instr (cpu), 4, 0); | |
2640 | int i, index; | |
2641 | ||
2642 | NYI_assert (29, 21, 0x070); | |
2643 | NYI_assert (15, 10, 0x01); | |
2644 | ||
2645 | if (uimm (aarch64_get_instr (cpu), 16, 16)) | |
2646 | { | |
2647 | index = uimm (aarch64_get_instr (cpu), 20, 17); | |
2648 | ||
2649 | for (i = 0; i < (full ? 16 : 8); i++) | |
2650 | aarch64_set_vec_u8 (cpu, vd, i, aarch64_get_vec_u8 (cpu, vs, index)); | |
2651 | } | |
2652 | else if (uimm (aarch64_get_instr (cpu), 17, 17)) | |
2653 | { | |
2654 | index = uimm (aarch64_get_instr (cpu), 20, 18); | |
2655 | ||
2656 | for (i = 0; i < (full ? 8 : 4); i++) | |
2657 | aarch64_set_vec_u16 (cpu, vd, i, aarch64_get_vec_u16 (cpu, vs, index)); | |
2658 | } | |
2659 | else if (uimm (aarch64_get_instr (cpu), 18, 18)) | |
2660 | { | |
2661 | index = uimm (aarch64_get_instr (cpu), 20, 19); | |
2662 | ||
2663 | for (i = 0; i < (full ? 4 : 2); i++) | |
2664 | aarch64_set_vec_u32 (cpu, vd, i, aarch64_get_vec_u32 (cpu, vs, index)); | |
2665 | } | |
2666 | else | |
2667 | { | |
2668 | if (uimm (aarch64_get_instr (cpu), 19, 19) == 0) | |
2669 | HALT_UNALLOC; | |
2670 | ||
2671 | if (! full) | |
2672 | HALT_UNALLOC; | |
2673 | ||
2674 | index = uimm (aarch64_get_instr (cpu), 20, 20); | |
2675 | ||
2676 | for (i = 0; i < 2; i++) | |
2677 | aarch64_set_vec_u64 (cpu, vd, i, aarch64_get_vec_u64 (cpu, vs, index)); | |
2678 | } | |
2679 | } | |
2680 | ||
2681 | static void | |
2682 | do_vec_TBL (sim_cpu *cpu) | |
2683 | { | |
2684 | /* instr[31] = 0 | |
2685 | instr[30] = half(0)/full(1) | |
2686 | instr[29,21] = 00 1110 000 | |
2687 | instr[20,16] = Vm | |
2688 | instr[15] = 0 | |
2689 | instr[14,13] = vec length | |
2690 | instr[12,10] = 000 | |
2691 | instr[9,5] = V start | |
2692 | instr[4,0] = V dest */ | |
2693 | ||
2694 | int full = uimm (aarch64_get_instr (cpu), 30, 30); | |
2695 | int len = uimm (aarch64_get_instr (cpu), 14, 13) + 1; | |
2696 | unsigned vm = uimm (aarch64_get_instr (cpu), 20, 16); | |
2697 | unsigned vn = uimm (aarch64_get_instr (cpu), 9, 5); | |
2698 | unsigned vd = uimm (aarch64_get_instr (cpu), 4, 0); | |
2699 | unsigned i; | |
2700 | ||
2701 | NYI_assert (29, 21, 0x070); | |
2702 | NYI_assert (12, 10, 0); | |
2703 | ||
2704 | for (i = 0; i < (full ? 16 : 8); i++) | |
2705 | { | |
2706 | unsigned int selector = aarch64_get_vec_u8 (cpu, vm, i); | |
2707 | uint8_t val; | |
2708 | ||
2709 | if (selector < 16) | |
2710 | val = aarch64_get_vec_u8 (cpu, vn, selector); | |
2711 | else if (selector < 32) | |
2712 | val = len < 2 ? 0 : aarch64_get_vec_u8 (cpu, vn + 1, selector - 16); | |
2713 | else if (selector < 48) | |
2714 | val = len < 3 ? 0 : aarch64_get_vec_u8 (cpu, vn + 2, selector - 32); | |
2715 | else if (selector < 64) | |
2716 | val = len < 4 ? 0 : aarch64_get_vec_u8 (cpu, vn + 3, selector - 48); | |
2717 | else | |
2718 | val = 0; | |
2719 | ||
2720 | aarch64_set_vec_u8 (cpu, vd, i, val); | |
2721 | } | |
2722 | } | |
2723 | ||
2724 | static void | |
2725 | do_vec_TRN (sim_cpu *cpu) | |
2726 | { | |
2727 | /* instr[31] = 0 | |
2728 | instr[30] = half(0)/full(1) | |
2729 | instr[29,24] = 00 1110 | |
2730 | instr[23,22] = size | |
2731 | instr[21] = 0 | |
2732 | instr[20,16] = Vm | |
2733 | instr[15] = 0 | |
2734 | instr[14] = TRN1 (0) / TRN2 (1) | |
2735 | instr[13,10] = 1010 | |
2736 | instr[9,5] = V source | |
2737 | instr[4,0] = V dest. */ | |
2738 | ||
2739 | int full = uimm (aarch64_get_instr (cpu), 30, 30); | |
2740 | int second = uimm (aarch64_get_instr (cpu), 14, 14); | |
2741 | unsigned vm = uimm (aarch64_get_instr (cpu), 20, 16); | |
2742 | unsigned vn = uimm (aarch64_get_instr (cpu), 9, 5); | |
2743 | unsigned vd = uimm (aarch64_get_instr (cpu), 4, 0); | |
2744 | unsigned i; | |
2745 | ||
2746 | NYI_assert (29, 24, 0x0E); | |
2747 | NYI_assert (13, 10, 0xA); | |
2748 | ||
2749 | switch (uimm (aarch64_get_instr (cpu), 23, 22)) | |
2750 | { | |
2751 | case 0: | |
2752 | for (i = 0; i < (full ? 8 : 4); i++) | |
2753 | { | |
2754 | aarch64_set_vec_u8 | |
2755 | (cpu, vd, i * 2, | |
2756 | aarch64_get_vec_u8 (cpu, second ? vm : vn, i * 2)); | |
2757 | aarch64_set_vec_u8 | |
2758 | (cpu, vd, 1 * 2 + 1, | |
2759 | aarch64_get_vec_u8 (cpu, second ? vn : vm, i * 2 + 1)); | |
2760 | } | |
2761 | break; | |
2762 | ||
2763 | case 1: | |
2764 | for (i = 0; i < (full ? 4 : 2); i++) | |
2765 | { | |
2766 | aarch64_set_vec_u16 | |
2767 | (cpu, vd, i * 2, | |
2768 | aarch64_get_vec_u16 (cpu, second ? vm : vn, i * 2)); | |
2769 | aarch64_set_vec_u16 | |
2770 | (cpu, vd, 1 * 2 + 1, | |
2771 | aarch64_get_vec_u16 (cpu, second ? vn : vm, i * 2 + 1)); | |
2772 | } | |
2773 | break; | |
2774 | ||
2775 | case 2: | |
2776 | aarch64_set_vec_u32 | |
2777 | (cpu, vd, 0, aarch64_get_vec_u32 (cpu, second ? vm : vn, 0)); | |
2778 | aarch64_set_vec_u32 | |
2779 | (cpu, vd, 1, aarch64_get_vec_u32 (cpu, second ? vn : vm, 1)); | |
2780 | aarch64_set_vec_u32 | |
2781 | (cpu, vd, 2, aarch64_get_vec_u32 (cpu, second ? vm : vn, 2)); | |
2782 | aarch64_set_vec_u32 | |
2783 | (cpu, vd, 3, aarch64_get_vec_u32 (cpu, second ? vn : vm, 3)); | |
2784 | break; | |
2785 | ||
2786 | case 3: | |
2787 | if (! full) | |
2788 | HALT_UNALLOC; | |
2789 | ||
2790 | aarch64_set_vec_u64 (cpu, vd, 0, | |
2791 | aarch64_get_vec_u64 (cpu, second ? vm : vn, 0)); | |
2792 | aarch64_set_vec_u64 (cpu, vd, 1, | |
2793 | aarch64_get_vec_u64 (cpu, second ? vn : vm, 1)); | |
2794 | break; | |
2795 | ||
2796 | default: | |
2797 | HALT_UNALLOC; | |
2798 | } | |
2799 | } | |
2800 | ||
2801 | static void | |
2802 | do_vec_DUP_scalar_into_vector (sim_cpu *cpu) | |
2803 | { | |
2804 | /* instr[31] = 0 | |
2805 | instr[30] = 0=> zero top 64-bits, 1=> duplicate into top 64-bits | |
2806 | [must be 1 for 64-bit xfer] | |
2807 | instr[29,20] = 00 1110 0000 | |
2808 | instr[19,16] = element size: 0001=> 8-bits, 0010=> 16-bits, | |
2809 | 0100=> 32-bits. 1000=>64-bits | |
2810 | instr[15,10] = 0000 11 | |
2811 | instr[9,5] = W source | |
2812 | instr[4,0] = V dest. */ | |
2813 | ||
2814 | unsigned i; | |
2815 | unsigned Vd = uimm (aarch64_get_instr (cpu), 4, 0); | |
2816 | unsigned Rs = uimm (aarch64_get_instr (cpu), 9, 5); | |
2817 | int both = uimm (aarch64_get_instr (cpu), 30, 30); | |
2818 | ||
2819 | NYI_assert (29, 20, 0x0E0); | |
2820 | NYI_assert (15, 10, 0x03); | |
2821 | ||
2822 | switch (uimm (aarch64_get_instr (cpu), 19, 16)) | |
2823 | { | |
2824 | case 1: | |
2825 | for (i = 0; i < (both ? 16 : 8); i++) | |
2826 | aarch64_set_vec_u8 (cpu, Vd, i, aarch64_get_reg_u8 (cpu, Rs, NO_SP)); | |
2827 | break; | |
2828 | ||
2829 | case 2: | |
2830 | for (i = 0; i < (both ? 8 : 4); i++) | |
2831 | aarch64_set_vec_u16 (cpu, Vd, i, aarch64_get_reg_u16 (cpu, Rs, NO_SP)); | |
2832 | break; | |
2833 | ||
2834 | case 4: | |
2835 | for (i = 0; i < (both ? 4 : 2); i++) | |
2836 | aarch64_set_vec_u32 (cpu, Vd, i, aarch64_get_reg_u32 (cpu, Rs, NO_SP)); | |
2837 | break; | |
2838 | ||
2839 | case 8: | |
2840 | if (!both) | |
2841 | HALT_NYI; | |
2842 | aarch64_set_vec_u64 (cpu, Vd, 0, aarch64_get_reg_u64 (cpu, Rs, NO_SP)); | |
2843 | aarch64_set_vec_u64 (cpu, Vd, 1, aarch64_get_reg_u64 (cpu, Rs, NO_SP)); | |
2844 | break; | |
2845 | ||
2846 | default: | |
2847 | HALT_NYI; | |
2848 | } | |
2849 | } | |
2850 | ||
2851 | static void | |
2852 | do_vec_UZP (sim_cpu *cpu) | |
2853 | { | |
2854 | /* instr[31] = 0 | |
2855 | instr[30] = half(0)/full(1) | |
2856 | instr[29,24] = 00 1110 | |
2857 | instr[23,22] = size: byte(00), half(01), word (10), long (11) | |
2858 | instr[21] = 0 | |
2859 | instr[20,16] = Vm | |
2860 | instr[15] = 0 | |
2861 | instr[14] = lower (0) / upper (1) | |
2862 | instr[13,10] = 0110 | |
2863 | instr[9,5] = Vn | |
2864 | instr[4,0] = Vd. */ | |
2865 | ||
2866 | int full = uimm (aarch64_get_instr (cpu), 30, 30); | |
2867 | int upper = uimm (aarch64_get_instr (cpu), 14, 14); | |
2868 | ||
2869 | unsigned vm = uimm (aarch64_get_instr (cpu), 20, 16); | |
2870 | unsigned vn = uimm (aarch64_get_instr (cpu), 9, 5); | |
2871 | unsigned vd = uimm (aarch64_get_instr (cpu), 4, 0); | |
2872 | ||
2873 | uint64_t val_m1 = aarch64_get_vec_u64 (cpu, vm, 0); | |
2874 | uint64_t val_m2 = aarch64_get_vec_u64 (cpu, vm, 1); | |
2875 | uint64_t val_n1 = aarch64_get_vec_u64 (cpu, vn, 0); | |
2876 | uint64_t val_n2 = aarch64_get_vec_u64 (cpu, vn, 1); | |
2877 | ||
2878 | uint64_t val1 = 0; | |
2879 | uint64_t val2 = 0; | |
2880 | ||
2881 | uint64_t input1 = upper ? val_n1 : val_m1; | |
2882 | uint64_t input2 = upper ? val_n2 : val_m2; | |
2883 | unsigned i; | |
2884 | ||
2885 | NYI_assert (29, 24, 0x0E); | |
2886 | NYI_assert (21, 21, 0); | |
2887 | NYI_assert (15, 15, 0); | |
2888 | NYI_assert (13, 10, 6); | |
2889 | ||
2890 | switch (uimm (aarch64_get_instr (cpu), 23, 23)) | |
2891 | { | |
2892 | case 0: | |
2893 | for (i = 0; i < 8; i++) | |
2894 | { | |
2895 | val1 |= (input1 >> (i * 8)) & (0xFFULL << (i * 8)); | |
2896 | val2 |= (input2 >> (i * 8)) & (0xFFULL << (i * 8)); | |
2897 | } | |
2898 | break; | |
2899 | ||
2900 | case 1: | |
2901 | for (i = 0; i < 4; i++) | |
2902 | { | |
2903 | val1 |= (input1 >> (i * 16)) & (0xFFFFULL << (i * 16)); | |
2904 | val2 |= (input2 >> (i * 16)) & (0xFFFFULL << (i * 16)); | |
2905 | } | |
2906 | break; | |
2907 | ||
2908 | case 2: | |
2909 | val1 = ((input1 & 0xFFFFFFFF) | ((input1 >> 32) & 0xFFFFFFFF00000000ULL)); | |
2910 | val2 = ((input2 & 0xFFFFFFFF) | ((input2 >> 32) & 0xFFFFFFFF00000000ULL)); | |
2911 | ||
2912 | case 3: | |
2913 | val1 = input1; | |
2914 | val2 = input2; | |
2915 | break; | |
2916 | } | |
2917 | ||
2918 | aarch64_set_vec_u64 (cpu, vd, 0, val1); | |
2919 | if (full) | |
2920 | aarch64_set_vec_u64 (cpu, vd, 1, val2); | |
2921 | } | |
2922 | ||
2923 | static void | |
2924 | do_vec_ZIP (sim_cpu *cpu) | |
2925 | { | |
2926 | /* instr[31] = 0 | |
2927 | instr[30] = half(0)/full(1) | |
2928 | instr[29,24] = 00 1110 | |
2929 | instr[23,22] = size: byte(00), hald(01), word (10), long (11) | |
2930 | instr[21] = 0 | |
2931 | instr[20,16] = Vm | |
2932 | instr[15] = 0 | |
2933 | instr[14] = lower (0) / upper (1) | |
2934 | instr[13,10] = 1110 | |
2935 | instr[9,5] = Vn | |
2936 | instr[4,0] = Vd. */ | |
2937 | ||
2938 | int full = uimm (aarch64_get_instr (cpu), 30, 30); | |
2939 | int upper = uimm (aarch64_get_instr (cpu), 14, 14); | |
2940 | ||
2941 | unsigned vm = uimm (aarch64_get_instr (cpu), 20, 16); | |
2942 | unsigned vn = uimm (aarch64_get_instr (cpu), 9, 5); | |
2943 | unsigned vd = uimm (aarch64_get_instr (cpu), 4, 0); | |
2944 | ||
2945 | uint64_t val_m1 = aarch64_get_vec_u64 (cpu, vm, 0); | |
2946 | uint64_t val_m2 = aarch64_get_vec_u64 (cpu, vm, 1); | |
2947 | uint64_t val_n1 = aarch64_get_vec_u64 (cpu, vn, 0); | |
2948 | uint64_t val_n2 = aarch64_get_vec_u64 (cpu, vn, 1); | |
2949 | ||
2950 | uint64_t val1 = 0; | |
2951 | uint64_t val2 = 0; | |
2952 | ||
2953 | uint64_t input1 = upper ? val_n1 : val_m1; | |
2954 | uint64_t input2 = upper ? val_n2 : val_m2; | |
2955 | ||
2956 | NYI_assert (29, 24, 0x0E); | |
2957 | NYI_assert (21, 21, 0); | |
2958 | NYI_assert (15, 15, 0); | |
2959 | NYI_assert (13, 10, 0xE); | |
2960 | ||
2961 | switch (uimm (aarch64_get_instr (cpu), 23, 23)) | |
2962 | { | |
2963 | case 0: | |
2964 | val1 = | |
2965 | ((input1 << 0) & (0xFF << 0)) | |
2966 | | ((input2 << 8) & (0xFF << 8)) | |
2967 | | ((input1 << 8) & (0xFF << 16)) | |
2968 | | ((input2 << 16) & (0xFF << 24)) | |
2969 | | ((input1 << 16) & (0xFFULL << 32)) | |
2970 | | ((input2 << 24) & (0xFFULL << 40)) | |
2971 | | ((input1 << 24) & (0xFFULL << 48)) | |
2972 | | ((input2 << 32) & (0xFFULL << 56)); | |
2973 | ||
2974 | val2 = | |
2975 | ((input1 >> 32) & (0xFF << 0)) | |
2976 | | ((input2 >> 24) & (0xFF << 8)) | |
2977 | | ((input1 >> 24) & (0xFF << 16)) | |
2978 | | ((input2 >> 16) & (0xFF << 24)) | |
2979 | | ((input1 >> 16) & (0xFFULL << 32)) | |
2980 | | ((input2 >> 8) & (0xFFULL << 40)) | |
2981 | | ((input1 >> 8) & (0xFFULL << 48)) | |
2982 | | ((input2 >> 0) & (0xFFULL << 56)); | |
2983 | break; | |
2984 | ||
2985 | case 1: | |
2986 | val1 = | |
2987 | ((input1 << 0) & (0xFFFF << 0)) | |
2988 | | ((input2 << 16) & (0xFFFF << 16)) | |
2989 | | ((input1 << 16) & (0xFFFFULL << 32)) | |
2990 | | ((input2 << 32) & (0xFFFFULL << 48)); | |
2991 | ||
2992 | val2 = | |
2993 | ((input1 >> 32) & (0xFFFF << 0)) | |
2994 | | ((input2 >> 16) & (0xFFFF << 16)) | |
2995 | | ((input1 >> 16) & (0xFFFFULL << 32)) | |
2996 | | ((input2 >> 0) & (0xFFFFULL << 48)); | |
2997 | break; | |
2998 | ||
2999 | case 2: | |
3000 | val1 = (input1 & 0xFFFFFFFFULL) | (input2 << 32); | |
3001 | val2 = (input2 & 0xFFFFFFFFULL) | (input1 << 32); | |
3002 | break; | |
3003 | ||
3004 | case 3: | |
3005 | val1 = input1; | |
3006 | val2 = input2; | |
3007 | break; | |
3008 | } | |
3009 | ||
3010 | aarch64_set_vec_u64 (cpu, vd, 0, val1); | |
3011 | if (full) | |
3012 | aarch64_set_vec_u64 (cpu, vd, 1, val2); | |
3013 | } | |
3014 | ||
3015 | /* Floating point immediates are encoded in 8 bits. | |
3016 | fpimm[7] = sign bit. | |
3017 | fpimm[6:4] = signed exponent. | |
3018 | fpimm[3:0] = fraction (assuming leading 1). | |
3019 | i.e. F = s * 1.f * 2^(e - b). */ | |
3020 | ||
3021 | static float | |
3022 | fp_immediate_for_encoding_32 (uint32_t imm8) | |
3023 | { | |
3024 | float u; | |
3025 | uint32_t s, e, f, i; | |
3026 | ||
3027 | s = (imm8 >> 7) & 0x1; | |
3028 | e = (imm8 >> 4) & 0x7; | |
3029 | f = imm8 & 0xf; | |
3030 | ||
3031 | /* The fp value is s * n/16 * 2r where n is 16+e. */ | |
3032 | u = (16.0 + f) / 16.0; | |
3033 | ||
3034 | /* N.B. exponent is signed. */ | |
3035 | if (e < 4) | |
3036 | { | |
3037 | int epos = e; | |
3038 | ||
3039 | for (i = 0; i <= epos; i++) | |
3040 | u *= 2.0; | |
3041 | } | |
3042 | else | |
3043 | { | |
3044 | int eneg = 7 - e; | |
3045 | ||
3046 | for (i = 0; i < eneg; i++) | |
3047 | u /= 2.0; | |
3048 | } | |
3049 | ||
3050 | if (s) | |
3051 | u = - u; | |
3052 | ||
3053 | return u; | |
3054 | } | |
3055 | ||
3056 | static double | |
3057 | fp_immediate_for_encoding_64 (uint32_t imm8) | |
3058 | { | |
3059 | double u; | |
3060 | uint32_t s, e, f, i; | |
3061 | ||
3062 | s = (imm8 >> 7) & 0x1; | |
3063 | e = (imm8 >> 4) & 0x7; | |
3064 | f = imm8 & 0xf; | |
3065 | ||
3066 | /* The fp value is s * n/16 * 2r where n is 16+e. */ | |
3067 | u = (16.0 + f) / 16.0; | |
3068 | ||
3069 | /* N.B. exponent is signed. */ | |
3070 | if (e < 4) | |
3071 | { | |
3072 | int epos = e; | |
3073 | ||
3074 | for (i = 0; i <= epos; i++) | |
3075 | u *= 2.0; | |
3076 | } | |
3077 | else | |
3078 | { | |
3079 | int eneg = 7 - e; | |
3080 | ||
3081 | for (i = 0; i < eneg; i++) | |
3082 | u /= 2.0; | |
3083 | } | |
3084 | ||
3085 | if (s) | |
3086 | u = - u; | |
3087 | ||
3088 | return u; | |
3089 | } | |
3090 | ||
3091 | static void | |
3092 | do_vec_MOV_immediate (sim_cpu *cpu) | |
3093 | { | |
3094 | /* instr[31] = 0 | |
3095 | instr[30] = full/half selector | |
3096 | instr[29,19] = 00111100000 | |
3097 | instr[18,16] = high 3 bits of uimm8 | |
3098 | instr[15,12] = size & shift: | |
3099 | 0000 => 32-bit | |
3100 | 0010 => 32-bit + LSL#8 | |
3101 | 0100 => 32-bit + LSL#16 | |
3102 | 0110 => 32-bit + LSL#24 | |
3103 | 1010 => 16-bit + LSL#8 | |
3104 | 1000 => 16-bit | |
3105 | 1101 => 32-bit + MSL#16 | |
3106 | 1100 => 32-bit + MSL#8 | |
3107 | 1110 => 8-bit | |
3108 | 1111 => double | |
3109 | instr[11,10] = 01 | |
3110 | instr[9,5] = low 5-bits of uimm8 | |
3111 | instr[4,0] = Vd. */ | |
3112 | ||
3113 | int full = uimm (aarch64_get_instr (cpu), 30, 30); | |
3114 | unsigned vd = uimm (aarch64_get_instr (cpu), 4, 0); | |
3115 | unsigned val = uimm (aarch64_get_instr (cpu), 18, 16) << 5 | |
3116 | | uimm (aarch64_get_instr (cpu), 9, 5); | |
3117 | unsigned i; | |
3118 | ||
3119 | NYI_assert (29, 19, 0x1E0); | |
3120 | NYI_assert (11, 10, 1); | |
3121 | ||
3122 | switch (uimm (aarch64_get_instr (cpu), 15, 12)) | |
3123 | { | |
3124 | case 0x0: /* 32-bit, no shift. */ | |
3125 | case 0x2: /* 32-bit, shift by 8. */ | |
3126 | case 0x4: /* 32-bit, shift by 16. */ | |
3127 | case 0x6: /* 32-bit, shift by 24. */ | |
3128 | val <<= (8 * uimm (aarch64_get_instr (cpu), 14, 13)); | |
3129 | for (i = 0; i < (full ? 4 : 2); i++) | |
3130 | aarch64_set_vec_u32 (cpu, vd, i, val); | |
3131 | break; | |
3132 | ||
3133 | case 0xa: /* 16-bit, shift by 8. */ | |
3134 | val <<= 8; | |
3135 | /* Fall through. */ | |
3136 | case 0x8: /* 16-bit, no shift. */ | |
3137 | for (i = 0; i < (full ? 8 : 4); i++) | |
3138 | aarch64_set_vec_u16 (cpu, vd, i, val); | |
3139 | /* Fall through. */ | |
3140 | case 0xd: /* 32-bit, mask shift by 16. */ | |
3141 | val <<= 8; | |
3142 | val |= 0xFF; | |
3143 | /* Fall through. */ | |
3144 | case 0xc: /* 32-bit, mask shift by 8. */ | |
3145 | val <<= 8; | |
3146 | val |= 0xFF; | |
3147 | for (i = 0; i < (full ? 4 : 2); i++) | |
3148 | aarch64_set_vec_u32 (cpu, vd, i, val); | |
3149 | break; | |
3150 | ||
3151 | case 0xe: /* 8-bit, no shift. */ | |
3152 | for (i = 0; i < (full ? 16 : 8); i++) | |
3153 | aarch64_set_vec_u8 (cpu, vd, i, val); | |
3154 | break; | |
3155 | ||
3156 | case 0xf: /* FMOV Vs.{2|4}S, #fpimm. */ | |
3157 | { | |
3158 | float u = fp_immediate_for_encoding_32 (val); | |
3159 | for (i = 0; i < (full ? 4 : 2); i++) | |
3160 | aarch64_set_vec_float (cpu, vd, i, u); | |
3161 | break; | |
3162 | } | |
3163 | ||
3164 | default: | |
3165 | HALT_NYI; | |
3166 | } | |
3167 | } | |
3168 | ||
3169 | static void | |
3170 | do_vec_MVNI (sim_cpu *cpu) | |
3171 | { | |
3172 | /* instr[31] = 0 | |
3173 | instr[30] = full/half selector | |
3174 | instr[29,19] = 10111100000 | |
3175 | instr[18,16] = high 3 bits of uimm8 | |
3176 | instr[15,12] = selector | |
3177 | instr[11,10] = 01 | |
3178 | instr[9,5] = low 5-bits of uimm8 | |
3179 | instr[4,0] = Vd. */ | |
3180 | ||
3181 | int full = uimm (aarch64_get_instr (cpu), 30, 30); | |
3182 | unsigned vd = uimm (aarch64_get_instr (cpu), 4, 0); | |
3183 | unsigned val = uimm (aarch64_get_instr (cpu), 18, 16) << 5 | |
3184 | | uimm (aarch64_get_instr (cpu), 9, 5); | |
3185 | unsigned i; | |
3186 | ||
3187 | NYI_assert (29, 19, 0x5E0); | |
3188 | NYI_assert (11, 10, 1); | |
3189 | ||
3190 | switch (uimm (aarch64_get_instr (cpu), 15, 12)) | |
3191 | { | |
3192 | case 0x0: /* 32-bit, no shift. */ | |
3193 | case 0x2: /* 32-bit, shift by 8. */ | |
3194 | case 0x4: /* 32-bit, shift by 16. */ | |
3195 | case 0x6: /* 32-bit, shift by 24. */ | |
3196 | val <<= (8 * uimm (aarch64_get_instr (cpu), 14, 13)); | |
3197 | val = ~ val; | |
3198 | for (i = 0; i < (full ? 4 : 2); i++) | |
3199 | aarch64_set_vec_u32 (cpu, vd, i, val); | |
3200 | return; | |
3201 | ||
3202 | case 0xa: /* 16-bit, 8 bit shift. */ | |
3203 | val <<= 8; | |
3204 | case 0x8: /* 16-bit, no shift. */ | |
3205 | val = ~ val; | |
3206 | for (i = 0; i < (full ? 8 : 4); i++) | |
3207 | aarch64_set_vec_u16 (cpu, vd, i, val); | |
3208 | return; | |
3209 | ||
3210 | case 0xd: /* 32-bit, mask shift by 16. */ | |
3211 | val <<= 8; | |
3212 | val |= 0xFF; | |
3213 | case 0xc: /* 32-bit, mask shift by 8. */ | |
3214 | val <<= 8; | |
3215 | val |= 0xFF; | |
3216 | val = ~ val; | |
3217 | for (i = 0; i < (full ? 4 : 2); i++) | |
3218 | aarch64_set_vec_u32 (cpu, vd, i, val); | |
3219 | return; | |
3220 | ||
3221 | case 0xE: /* MOVI Dn, #mask64 */ | |
3222 | { | |
3223 | uint64_t mask = 0; | |
3224 | ||
3225 | for (i = 0; i < 8; i++) | |
3226 | if (val & (1 << i)) | |
3227 | mask |= (0xF << (i * 4)); | |
3228 | aarch64_set_vec_u64 (cpu, vd, 0, mask); | |
3229 | aarch64_set_vec_u64 (cpu, vd, 1, 0); | |
3230 | return; | |
3231 | } | |
3232 | ||
3233 | case 0xf: /* FMOV Vd.2D, #fpimm. */ | |
3234 | { | |
3235 | double u = fp_immediate_for_encoding_64 (val); | |
3236 | ||
3237 | if (! full) | |
3238 | HALT_UNALLOC; | |
3239 | ||
3240 | aarch64_set_vec_double (cpu, vd, 0, u); | |
3241 | aarch64_set_vec_double (cpu, vd, 1, u); | |
3242 | return; | |
3243 | } | |
3244 | ||
3245 | default: | |
3246 | HALT_NYI; | |
3247 | } | |
3248 | } | |
3249 | ||
3250 | #define ABS(A) ((A) < 0 ? - (A) : (A)) | |
3251 | ||
3252 | static void | |
3253 | do_vec_ABS (sim_cpu *cpu) | |
3254 | { | |
3255 | /* instr[31] = 0 | |
3256 | instr[30] = half(0)/full(1) | |
3257 | instr[29,24] = 00 1110 | |
3258 | instr[23,22] = size: 00=> 8-bit, 01=> 16-bit, 10=> 32-bit, 11=> 64-bit | |
3259 | instr[21,10] = 10 0000 1011 10 | |
3260 | instr[9,5] = Vn | |
3261 | instr[4.0] = Vd. */ | |
3262 | ||
3263 | unsigned vn = uimm (aarch64_get_instr (cpu), 9, 5); | |
3264 | unsigned vd = uimm (aarch64_get_instr (cpu), 4, 0); | |
3265 | unsigned full = uimm (aarch64_get_instr (cpu), 30, 30); | |
3266 | unsigned i; | |
3267 | ||
3268 | NYI_assert (29, 24, 0x0E); | |
3269 | NYI_assert (21, 10, 0x82E); | |
3270 | ||
3271 | switch (uimm (aarch64_get_instr (cpu), 23, 22)) | |
3272 | { | |
3273 | case 0: | |
3274 | for (i = 0; i < (full ? 16 : 8); i++) | |
3275 | aarch64_set_vec_s8 (cpu, vd, i, | |
3276 | ABS (aarch64_get_vec_s8 (cpu, vn, i))); | |
3277 | break; | |
3278 | ||
3279 | case 1: | |
3280 | for (i = 0; i < (full ? 8 : 4); i++) | |
3281 | aarch64_set_vec_s16 (cpu, vd, i, | |
3282 | ABS (aarch64_get_vec_s16 (cpu, vn, i))); | |
3283 | break; | |
3284 | ||
3285 | case 2: | |
3286 | for (i = 0; i < (full ? 4 : 2); i++) | |
3287 | aarch64_set_vec_s32 (cpu, vd, i, | |
3288 | ABS (aarch64_get_vec_s32 (cpu, vn, i))); | |
3289 | break; | |
3290 | ||
3291 | case 3: | |
3292 | if (! full) | |
3293 | HALT_NYI; | |
3294 | for (i = 0; i < 2; i++) | |
3295 | aarch64_set_vec_s64 (cpu, vd, i, | |
3296 | ABS (aarch64_get_vec_s64 (cpu, vn, i))); | |
3297 | break; | |
3298 | } | |
3299 | } | |
3300 | ||
3301 | static void | |
3302 | do_vec_ADDV (sim_cpu *cpu) | |
3303 | { | |
3304 | /* instr[31] = 0 | |
3305 | instr[30] = full/half selector | |
3306 | instr[29,24] = 00 1110 | |
3307 | instr[23,22] = size: 00=> 8-bit, 01=> 16-bit, 10=> 32-bit, 11=> 64-bit | |
3308 | instr[21,10] = 11 0001 1011 10 | |
3309 | instr[9,5] = Vm | |
3310 | instr[4.0] = Rd. */ | |
3311 | ||
3312 | unsigned vm = uimm (aarch64_get_instr (cpu), 9, 5); | |
3313 | unsigned rd = uimm (aarch64_get_instr (cpu), 4, 0); | |
3314 | unsigned i; | |
3315 | uint64_t val = 0; | |
3316 | int full = uimm (aarch64_get_instr (cpu), 30, 30); | |
3317 | ||
3318 | NYI_assert (29, 24, 0x0E); | |
3319 | NYI_assert (21, 10, 0xC6E); | |
3320 | ||
3321 | switch (uimm (aarch64_get_instr (cpu), 23, 22)) | |
3322 | { | |
3323 | case 0: | |
3324 | for (i = 0; i < (full ? 16 : 8); i++) | |
3325 | val += aarch64_get_vec_u8 (cpu, vm, i); | |
3326 | aarch64_set_reg_u64 (cpu, rd, NO_SP, val); | |
3327 | return; | |
3328 | ||
3329 | case 1: | |
3330 | for (i = 0; i < (full ? 8 : 4); i++) | |
3331 | val += aarch64_get_vec_u16 (cpu, vm, i); | |
3332 | aarch64_set_reg_u64 (cpu, rd, NO_SP, val); | |
3333 | return; | |
3334 | ||
3335 | case 2: | |
3336 | for (i = 0; i < (full ? 4 : 2); i++) | |
3337 | val += aarch64_get_vec_u32 (cpu, vm, i); | |
3338 | aarch64_set_reg_u64 (cpu, rd, NO_SP, val); | |
3339 | return; | |
3340 | ||
3341 | case 3: | |
3342 | if (! full) | |
3343 | HALT_UNALLOC; | |
3344 | val = aarch64_get_vec_u64 (cpu, vm, 0); | |
3345 | val += aarch64_get_vec_u64 (cpu, vm, 1); | |
3346 | aarch64_set_reg_u64 (cpu, rd, NO_SP, val); | |
3347 | return; | |
3348 | ||
3349 | default: | |
3350 | HALT_UNREACHABLE; | |
3351 | } | |
3352 | } | |
3353 | ||
3354 | static void | |
3355 | do_vec_ins_2 (sim_cpu *cpu) | |
3356 | { | |
3357 | /* instr[31,21] = 01001110000 | |
3358 | instr[20,18] = size & element selector | |
3359 | instr[17,14] = 0000 | |
3360 | instr[13] = direction: to vec(0), from vec (1) | |
3361 | instr[12,10] = 111 | |
3362 | instr[9,5] = Vm | |
3363 | instr[4,0] = Vd. */ | |
3364 | ||
3365 | unsigned elem; | |
3366 | unsigned vm = uimm (aarch64_get_instr (cpu), 9, 5); | |
3367 | unsigned vd = uimm (aarch64_get_instr (cpu), 4, 0); | |
3368 | ||
3369 | NYI_assert (31, 21, 0x270); | |
3370 | NYI_assert (17, 14, 0); | |
3371 | NYI_assert (12, 10, 7); | |
3372 | ||
3373 | if (uimm (aarch64_get_instr (cpu), 13, 13) == 1) | |
3374 | { | |
3375 | if (uimm (aarch64_get_instr (cpu), 18, 18) == 1) | |
3376 | { | |
3377 | /* 32-bit moves. */ | |
3378 | elem = uimm (aarch64_get_instr (cpu), 20, 19); | |
3379 | aarch64_set_reg_u64 (cpu, vd, NO_SP, | |
3380 | aarch64_get_vec_u32 (cpu, vm, elem)); | |
3381 | } | |
3382 | else | |
3383 | { | |
3384 | /* 64-bit moves. */ | |
3385 | if (uimm (aarch64_get_instr (cpu), 19, 19) != 1) | |
3386 | HALT_NYI; | |
3387 | ||
3388 | elem = uimm (aarch64_get_instr (cpu), 20, 20); | |
3389 | aarch64_set_reg_u64 (cpu, vd, NO_SP, | |
3390 | aarch64_get_vec_u64 (cpu, vm, elem)); | |
3391 | } | |
3392 | } | |
3393 | else | |
3394 | { | |
3395 | if (uimm (aarch64_get_instr (cpu), 18, 18) == 1) | |
3396 | { | |
3397 | /* 32-bit moves. */ | |
3398 | elem = uimm (aarch64_get_instr (cpu), 20, 19); | |
3399 | aarch64_set_vec_u32 (cpu, vd, elem, | |
3400 | aarch64_get_reg_u32 (cpu, vm, NO_SP)); | |
3401 | } | |
3402 | else | |
3403 | { | |
3404 | /* 64-bit moves. */ | |
3405 | if (uimm (aarch64_get_instr (cpu), 19, 19) != 1) | |
3406 | HALT_NYI; | |
3407 | ||
3408 | elem = uimm (aarch64_get_instr (cpu), 20, 20); | |
3409 | aarch64_set_vec_u64 (cpu, vd, elem, | |
3410 | aarch64_get_reg_u64 (cpu, vm, NO_SP)); | |
3411 | } | |
3412 | } | |
3413 | } | |
3414 | ||
3415 | static void | |
3416 | do_vec_mull (sim_cpu *cpu) | |
3417 | { | |
3418 | /* instr[31] = 0 | |
3419 | instr[30] = lower(0)/upper(1) selector | |
3420 | instr[29] = signed(0)/unsigned(1) | |
3421 | instr[28,24] = 0 1110 | |
3422 | instr[23,22] = size: 8-bit (00), 16-bit (01), 32-bit (10) | |
3423 | instr[21] = 1 | |
3424 | instr[20,16] = Vm | |
3425 | instr[15,10] = 11 0000 | |
3426 | instr[9,5] = Vn | |
3427 | instr[4.0] = Vd. */ | |
3428 | ||
3429 | int unsign = uimm (aarch64_get_instr (cpu), 29, 29); | |
3430 | int bias = uimm (aarch64_get_instr (cpu), 30, 30); | |
3431 | unsigned vm = uimm (aarch64_get_instr (cpu), 20, 16); | |
3432 | unsigned vn = uimm (aarch64_get_instr (cpu), 9, 5); | |
3433 | unsigned vd = uimm (aarch64_get_instr (cpu), 4, 0); | |
3434 | unsigned i; | |
3435 | ||
3436 | NYI_assert (28, 24, 0x0E); | |
3437 | NYI_assert (15, 10, 0x30); | |
3438 | ||
3439 | switch (uimm (aarch64_get_instr (cpu), 23, 22)) | |
3440 | { | |
3441 | case 0: | |
3442 | if (bias) | |
3443 | bias = 8; | |
3444 | if (unsign) | |
3445 | for (i = 0; i < 8; i++) | |
3446 | aarch64_set_vec_u16 (cpu, vd, i, | |
3447 | aarch64_get_vec_u8 (cpu, vn, i + bias) | |
3448 | * aarch64_get_vec_u8 (cpu, vm, i + bias)); | |
3449 | else | |
3450 | for (i = 0; i < 8; i++) | |
3451 | aarch64_set_vec_s16 (cpu, vd, i, | |
3452 | aarch64_get_vec_s8 (cpu, vn, i + bias) | |
3453 | * aarch64_get_vec_s8 (cpu, vm, i + bias)); | |
3454 | return; | |
3455 | ||
3456 | case 1: | |
3457 | if (bias) | |
3458 | bias = 4; | |
3459 | if (unsign) | |
3460 | for (i = 0; i < 4; i++) | |
3461 | aarch64_set_vec_u32 (cpu, vd, i, | |
3462 | aarch64_get_vec_u16 (cpu, vn, i + bias) | |
3463 | * aarch64_get_vec_u16 (cpu, vm, i + bias)); | |
3464 | else | |
3465 | for (i = 0; i < 4; i++) | |
3466 | aarch64_set_vec_s32 (cpu, vd, i, | |
3467 | aarch64_get_vec_s16 (cpu, vn, i + bias) | |
3468 | * aarch64_get_vec_s16 (cpu, vm, i + bias)); | |
3469 | return; | |
3470 | ||
3471 | case 2: | |
3472 | if (bias) | |
3473 | bias = 2; | |
3474 | if (unsign) | |
3475 | for (i = 0; i < 2; i++) | |
3476 | aarch64_set_vec_u64 (cpu, vd, i, | |
3477 | (uint64_t) aarch64_get_vec_u32 (cpu, vn, | |
3478 | i + bias) | |
3479 | * (uint64_t) aarch64_get_vec_u32 (cpu, vm, | |
3480 | i + bias)); | |
3481 | else | |
3482 | for (i = 0; i < 2; i++) | |
3483 | aarch64_set_vec_s64 (cpu, vd, i, | |
3484 | aarch64_get_vec_s32 (cpu, vn, i + bias) | |
3485 | * aarch64_get_vec_s32 (cpu, vm, i + bias)); | |
3486 | return; | |
3487 | ||
3488 | case 3: | |
3489 | default: | |
3490 | HALT_NYI; | |
3491 | } | |
3492 | } | |
3493 | ||
3494 | static void | |
3495 | do_vec_fadd (sim_cpu *cpu) | |
3496 | { | |
3497 | /* instr[31] = 0 | |
3498 | instr[30] = half(0)/full(1) | |
3499 | instr[29,24] = 001110 | |
3500 | instr[23] = FADD(0)/FSUB(1) | |
3501 | instr[22] = float (0)/double(1) | |
3502 | instr[21] = 1 | |
3503 | instr[20,16] = Vm | |
3504 | instr[15,10] = 110101 | |
3505 | instr[9,5] = Vn | |
3506 | instr[4.0] = Vd. */ | |
3507 | ||
3508 | unsigned vm = uimm (aarch64_get_instr (cpu), 20, 16); | |
3509 | unsigned vn = uimm (aarch64_get_instr (cpu), 9, 5); | |
3510 | unsigned vd = uimm (aarch64_get_instr (cpu), 4, 0); | |
3511 | unsigned i; | |
3512 | int full = uimm (aarch64_get_instr (cpu), 30, 30); | |
3513 | ||
3514 | NYI_assert (29, 24, 0x0E); | |
3515 | NYI_assert (21, 21, 1); | |
3516 | NYI_assert (15, 10, 0x35); | |
3517 | ||
3518 | if (uimm (aarch64_get_instr (cpu), 23, 23)) | |
3519 | { | |
3520 | if (uimm (aarch64_get_instr (cpu), 22, 22)) | |
3521 | { | |
3522 | if (! full) | |
3523 | HALT_NYI; | |
3524 | ||
3525 | for (i = 0; i < 2; i++) | |
3526 | aarch64_set_vec_double (cpu, vd, i, | |
3527 | aarch64_get_vec_double (cpu, vn, i) | |
3528 | - aarch64_get_vec_double (cpu, vm, i)); | |
3529 | } | |
3530 | else | |
3531 | { | |
3532 | for (i = 0; i < (full ? 4 : 2); i++) | |
3533 | aarch64_set_vec_float (cpu, vd, i, | |
3534 | aarch64_get_vec_float (cpu, vn, i) | |
3535 | - aarch64_get_vec_float (cpu, vm, i)); | |
3536 | } | |
3537 | } | |
3538 | else | |
3539 | { | |
3540 | if (uimm (aarch64_get_instr (cpu), 22, 22)) | |
3541 | { | |
3542 | if (! full) | |
3543 | HALT_NYI; | |
3544 | ||
3545 | for (i = 0; i < 2; i++) | |
3546 | aarch64_set_vec_double (cpu, vd, i, | |
3547 | aarch64_get_vec_double (cpu, vm, i) | |
3548 | + aarch64_get_vec_double (cpu, vn, i)); | |
3549 | } | |
3550 | else | |
3551 | { | |
3552 | for (i = 0; i < (full ? 4 : 2); i++) | |
3553 | aarch64_set_vec_float (cpu, vd, i, | |
3554 | aarch64_get_vec_float (cpu, vm, i) | |
3555 | + aarch64_get_vec_float (cpu, vn, i)); | |
3556 | } | |
3557 | } | |
3558 | } | |
3559 | ||
3560 | static void | |
3561 | do_vec_add (sim_cpu *cpu) | |
3562 | { | |
3563 | /* instr[31] = 0 | |
3564 | instr[30] = full/half selector | |
3565 | instr[29,24] = 001110 | |
3566 | instr[23,22] = size: 00=> 8-bit, 01=> 16-bit, 10=> 32-bit, 11=> 64-bit | |
3567 | instr[21] = 1 | |
3568 | instr[20,16] = Vn | |
3569 | instr[15,10] = 100001 | |
3570 | instr[9,5] = Vm | |
3571 | instr[4.0] = Vd. */ | |
3572 | ||
3573 | unsigned vm = uimm (aarch64_get_instr (cpu), 20, 16); | |
3574 | unsigned vn = uimm (aarch64_get_instr (cpu), 9, 5); | |
3575 | unsigned vd = uimm (aarch64_get_instr (cpu), 4, 0); | |
3576 | unsigned i; | |
3577 | int full = uimm (aarch64_get_instr (cpu), 30, 30); | |
3578 | ||
3579 | NYI_assert (29, 24, 0x0E); | |
3580 | NYI_assert (21, 21, 1); | |
3581 | NYI_assert (15, 10, 0x21); | |
3582 | ||
3583 | switch (uimm (aarch64_get_instr (cpu), 23, 22)) | |
3584 | { | |
3585 | case 0: | |
3586 | for (i = 0; i < (full ? 16 : 8); i++) | |
3587 | aarch64_set_vec_u8 (cpu, vd, i, aarch64_get_vec_u8 (cpu, vn, i) | |
3588 | + aarch64_get_vec_u8 (cpu, vm, i)); | |
3589 | return; | |
3590 | ||
3591 | case 1: | |
3592 | for (i = 0; i < (full ? 8 : 4); i++) | |
3593 | aarch64_set_vec_u16 (cpu, vd, i, aarch64_get_vec_u16 (cpu, vn, i) | |
3594 | + aarch64_get_vec_u16 (cpu, vm, i)); | |
3595 | return; | |
3596 | ||
3597 | case 2: | |
3598 | for (i = 0; i < (full ? 4 : 2); i++) | |
3599 | aarch64_set_vec_u32 (cpu, vd, i, aarch64_get_vec_u32 (cpu, vn, i) | |
3600 | + aarch64_get_vec_u32 (cpu, vm, i)); | |
3601 | return; | |
3602 | ||
3603 | case 3: | |
3604 | if (! full) | |
3605 | HALT_UNALLOC; | |
3606 | aarch64_set_vec_u64 (cpu, vd, 0, aarch64_get_vec_u64 (cpu, vn, 0) | |
3607 | + aarch64_get_vec_u64 (cpu, vm, 0)); | |
3608 | aarch64_set_vec_u64 (cpu, vd, 1, | |
3609 | aarch64_get_vec_u64 (cpu, vn, 1) | |
3610 | + aarch64_get_vec_u64 (cpu, vm, 1)); | |
3611 | return; | |
3612 | ||
3613 | default: | |
3614 | HALT_UNREACHABLE; | |
3615 | } | |
3616 | } | |
3617 | ||
3618 | static void | |
3619 | do_vec_mul (sim_cpu *cpu) | |
3620 | { | |
3621 | /* instr[31] = 0 | |
3622 | instr[30] = full/half selector | |
3623 | instr[29,24] = 00 1110 | |
3624 | instr[23,22] = size: 00=> 8-bit, 01=> 16-bit, 10=> 32-bit | |
3625 | instr[21] = 1 | |
3626 | instr[20,16] = Vn | |
3627 | instr[15,10] = 10 0111 | |
3628 | instr[9,5] = Vm | |
3629 | instr[4.0] = Vd. */ | |
3630 | ||
3631 | unsigned vm = uimm (aarch64_get_instr (cpu), 20, 16); | |
3632 | unsigned vn = uimm (aarch64_get_instr (cpu), 9, 5); | |
3633 | unsigned vd = uimm (aarch64_get_instr (cpu), 4, 0); | |
3634 | unsigned i; | |
3635 | int full = uimm (aarch64_get_instr (cpu), 30, 30); | |
3636 | ||
3637 | NYI_assert (29, 24, 0x0E); | |
3638 | NYI_assert (21, 21, 1); | |
3639 | NYI_assert (15, 10, 0x27); | |
3640 | ||
3641 | switch (uimm (aarch64_get_instr (cpu), 23, 22)) | |
3642 | { | |
3643 | case 0: | |
3644 | for (i = 0; i < (full ? 16 : 8); i++) | |
3645 | { | |
3646 | uint16_t val = aarch64_get_vec_u8 (cpu, vn, i); | |
3647 | val *= aarch64_get_vec_u8 (cpu, vm, i); | |
3648 | ||
3649 | aarch64_set_vec_u16 (cpu, vd, i, val); | |
3650 | } | |
3651 | return; | |
3652 | ||
3653 | case 1: | |
3654 | for (i = 0; i < (full ? 8 : 4); i++) | |
3655 | { | |
3656 | uint32_t val = aarch64_get_vec_u16 (cpu, vn, i); | |
3657 | val *= aarch64_get_vec_u16 (cpu, vm, i); | |
3658 | ||
3659 | aarch64_set_vec_u32 (cpu, vd, i, val); | |
3660 | } | |
3661 | return; | |
3662 | ||
3663 | case 2: | |
3664 | for (i = 0; i < (full ? 4 : 2); i++) | |
3665 | { | |
3666 | uint64_t val = aarch64_get_vec_u32 (cpu, vn, i); | |
3667 | val *= aarch64_get_vec_u32 (cpu, vm, i); | |
3668 | ||
3669 | aarch64_set_vec_u64 (cpu, vd, i, val); | |
3670 | } | |
3671 | return; | |
3672 | ||
3673 | default: | |
3674 | case 3: | |
3675 | HALT_UNALLOC; | |
3676 | } | |
3677 | } | |
3678 | ||
3679 | static void | |
3680 | do_vec_MLA (sim_cpu *cpu) | |
3681 | { | |
3682 | /* instr[31] = 0 | |
3683 | instr[30] = full/half selector | |
3684 | instr[29,24] = 00 1110 | |
3685 | instr[23,22] = size: 00=> 8-bit, 01=> 16-bit, 10=> 32-bit | |
3686 | instr[21] = 1 | |
3687 | instr[20,16] = Vn | |
3688 | instr[15,10] = 1001 01 | |
3689 | instr[9,5] = Vm | |
3690 | instr[4.0] = Vd. */ | |
3691 | ||
3692 | unsigned vm = uimm (aarch64_get_instr (cpu), 20, 16); | |
3693 | unsigned vn = uimm (aarch64_get_instr (cpu), 9, 5); | |
3694 | unsigned vd = uimm (aarch64_get_instr (cpu), 4, 0); | |
3695 | unsigned i; | |
3696 | int full = uimm (aarch64_get_instr (cpu), 30, 30); | |
3697 | ||
3698 | NYI_assert (29, 24, 0x0E); | |
3699 | NYI_assert (21, 21, 1); | |
3700 | NYI_assert (15, 10, 0x25); | |
3701 | ||
3702 | switch (uimm (aarch64_get_instr (cpu), 23, 22)) | |
3703 | { | |
3704 | case 0: | |
3705 | for (i = 0; i < (full ? 16 : 8); i++) | |
3706 | { | |
3707 | uint16_t val = aarch64_get_vec_u8 (cpu, vn, i); | |
3708 | val *= aarch64_get_vec_u8 (cpu, vm, i); | |
3709 | val += aarch64_get_vec_u8 (cpu, vd, i); | |
3710 | ||
3711 | aarch64_set_vec_u16 (cpu, vd, i, val); | |
3712 | } | |
3713 | return; | |
3714 | ||
3715 | case 1: | |
3716 | for (i = 0; i < (full ? 8 : 4); i++) | |
3717 | { | |
3718 | uint32_t val = aarch64_get_vec_u16 (cpu, vn, i); | |
3719 | val *= aarch64_get_vec_u16 (cpu, vm, i); | |
3720 | val += aarch64_get_vec_u16 (cpu, vd, i); | |
3721 | ||
3722 | aarch64_set_vec_u32 (cpu, vd, i, val); | |
3723 | } | |
3724 | return; | |
3725 | ||
3726 | case 2: | |
3727 | for (i = 0; i < (full ? 4 : 2); i++) | |
3728 | { | |
3729 | uint64_t val = aarch64_get_vec_u32 (cpu, vn, i); | |
3730 | val *= aarch64_get_vec_u32 (cpu, vm, i); | |
3731 | val += aarch64_get_vec_u32 (cpu, vd, i); | |
3732 | ||
3733 | aarch64_set_vec_u64 (cpu, vd, i, val); | |
3734 | } | |
3735 | return; | |
3736 | ||
3737 | default: | |
3738 | case 3: | |
3739 | HALT_UNALLOC; | |
3740 | } | |
3741 | } | |
3742 | ||
3743 | static float | |
3744 | fmaxnm (float a, float b) | |
3745 | { | |
3746 | if (fpclassify (a) == FP_NORMAL) | |
3747 | { | |
3748 | if (fpclassify (b) == FP_NORMAL) | |
3749 | return a > b ? a : b; | |
3750 | return a; | |
3751 | } | |
3752 | else if (fpclassify (b) == FP_NORMAL) | |
3753 | return b; | |
3754 | return a; | |
3755 | } | |
3756 | ||
3757 | static float | |
3758 | fminnm (float a, float b) | |
3759 | { | |
3760 | if (fpclassify (a) == FP_NORMAL) | |
3761 | { | |
3762 | if (fpclassify (b) == FP_NORMAL) | |
3763 | return a < b ? a : b; | |
3764 | return a; | |
3765 | } | |
3766 | else if (fpclassify (b) == FP_NORMAL) | |
3767 | return b; | |
3768 | return a; | |
3769 | } | |
3770 | ||
3771 | static double | |
3772 | dmaxnm (double a, double b) | |
3773 | { | |
3774 | if (fpclassify (a) == FP_NORMAL) | |
3775 | { | |
3776 | if (fpclassify (b) == FP_NORMAL) | |
3777 | return a > b ? a : b; | |
3778 | return a; | |
3779 | } | |
3780 | else if (fpclassify (b) == FP_NORMAL) | |
3781 | return b; | |
3782 | return a; | |
3783 | } | |
3784 | ||
3785 | static double | |
3786 | dminnm (double a, double b) | |
3787 | { | |
3788 | if (fpclassify (a) == FP_NORMAL) | |
3789 | { | |
3790 | if (fpclassify (b) == FP_NORMAL) | |
3791 | return a < b ? a : b; | |
3792 | return a; | |
3793 | } | |
3794 | else if (fpclassify (b) == FP_NORMAL) | |
3795 | return b; | |
3796 | return a; | |
3797 | } | |
3798 | ||
3799 | static void | |
3800 | do_vec_FminmaxNMP (sim_cpu *cpu) | |
3801 | { | |
3802 | /* aarch64_get_instr (cpu)[31] = 0 | |
3803 | aarch64_get_instr (cpu)[30] = half (0)/full (1) | |
3804 | aarch64_get_instr (cpu)[29,24] = 10 1110 | |
3805 | aarch64_get_instr (cpu)[23] = max(0)/min(1) | |
3806 | aarch64_get_instr (cpu)[22] = float (0)/double (1) | |
3807 | aarch64_get_instr (cpu)[21] = 1 | |
3808 | aarch64_get_instr (cpu)[20,16] = Vn | |
3809 | aarch64_get_instr (cpu)[15,10] = 1100 01 | |
3810 | aarch64_get_instr (cpu)[9,5] = Vm | |
3811 | aarch64_get_instr (cpu)[4.0] = Vd. */ | |
3812 | ||
3813 | unsigned vm = uimm (aarch64_get_instr (cpu), 20, 16); | |
3814 | unsigned vn = uimm (aarch64_get_instr (cpu), 9, 5); | |
3815 | unsigned vd = uimm (aarch64_get_instr (cpu), 4, 0); | |
3816 | int full = uimm (aarch64_get_instr (cpu), 30, 30); | |
3817 | ||
3818 | NYI_assert (29, 24, 0x2E); | |
3819 | NYI_assert (21, 21, 1); | |
3820 | NYI_assert (15, 10, 0x31); | |
3821 | ||
3822 | if (uimm (aarch64_get_instr (cpu), 22, 22)) | |
3823 | { | |
3824 | double (* fn)(double, double) = uimm (aarch64_get_instr (cpu), 23, 23) | |
3825 | ? dminnm : dmaxnm; | |
3826 | ||
3827 | if (! full) | |
3828 | HALT_NYI; | |
3829 | aarch64_set_vec_double (cpu, vd, 0, | |
3830 | fn (aarch64_get_vec_double (cpu, vn, 0), | |
3831 | aarch64_get_vec_double (cpu, vn, 1))); | |
3832 | aarch64_set_vec_double (cpu, vd, 0, | |
3833 | fn (aarch64_get_vec_double (cpu, vm, 0), | |
3834 | aarch64_get_vec_double (cpu, vm, 1))); | |
3835 | } | |
3836 | else | |
3837 | { | |
3838 | float (* fn)(float, float) = uimm (aarch64_get_instr (cpu), 23, 23) | |
3839 | ? fminnm : fmaxnm; | |
3840 | ||
3841 | aarch64_set_vec_float (cpu, vd, 0, | |
3842 | fn (aarch64_get_vec_float (cpu, vn, 0), | |
3843 | aarch64_get_vec_float (cpu, vn, 1))); | |
3844 | if (full) | |
3845 | aarch64_set_vec_float (cpu, vd, 1, | |
3846 | fn (aarch64_get_vec_float (cpu, vn, 2), | |
3847 | aarch64_get_vec_float (cpu, vn, 3))); | |
3848 | ||
3849 | aarch64_set_vec_float (cpu, vd, (full ? 2 : 1), | |
3850 | fn (aarch64_get_vec_float (cpu, vm, 0), | |
3851 | aarch64_get_vec_float (cpu, vm, 1))); | |
3852 | if (full) | |
3853 | aarch64_set_vec_float (cpu, vd, 3, | |
3854 | fn (aarch64_get_vec_float (cpu, vm, 2), | |
3855 | aarch64_get_vec_float (cpu, vm, 3))); | |
3856 | } | |
3857 | } | |
3858 | ||
3859 | static void | |
3860 | do_vec_AND (sim_cpu *cpu) | |
3861 | { | |
3862 | /* instr[31] = 0 | |
3863 | instr[30] = half (0)/full (1) | |
3864 | instr[29,21] = 001110001 | |
3865 | instr[20,16] = Vm | |
3866 | instr[15,10] = 000111 | |
3867 | instr[9,5] = Vn | |
3868 | instr[4.0] = Vd. */ | |
3869 | ||
3870 | unsigned vm = uimm (aarch64_get_instr (cpu), 20, 16); | |
3871 | unsigned vn = uimm (aarch64_get_instr (cpu), 9, 5); | |
3872 | unsigned vd = uimm (aarch64_get_instr (cpu), 4, 0); | |
3873 | unsigned i; | |
3874 | int full = uimm (aarch64_get_instr (cpu), 30, 30); | |
3875 | ||
3876 | NYI_assert (29, 21, 0x071); | |
3877 | NYI_assert (15, 10, 0x07); | |
3878 | ||
3879 | for (i = 0; i < (full ? 4 : 2); i++) | |
3880 | aarch64_set_vec_u32 (cpu, vd, i, | |
3881 | aarch64_get_vec_u32 (cpu, vn, i) | |
3882 | & aarch64_get_vec_u32 (cpu, vm, i)); | |
3883 | } | |
3884 | ||
3885 | static void | |
3886 | do_vec_BSL (sim_cpu *cpu) | |
3887 | { | |
3888 | /* instr[31] = 0 | |
3889 | instr[30] = half (0)/full (1) | |
3890 | instr[29,21] = 101110011 | |
3891 | instr[20,16] = Vm | |
3892 | instr[15,10] = 000111 | |
3893 | instr[9,5] = Vn | |
3894 | instr[4.0] = Vd. */ | |
3895 | ||
3896 | unsigned vm = uimm (aarch64_get_instr (cpu), 20, 16); | |
3897 | unsigned vn = uimm (aarch64_get_instr (cpu), 9, 5); | |
3898 | unsigned vd = uimm (aarch64_get_instr (cpu), 4, 0); | |
3899 | unsigned i; | |
3900 | int full = uimm (aarch64_get_instr (cpu), 30, 30); | |
3901 | ||
3902 | NYI_assert (29, 21, 0x173); | |
3903 | NYI_assert (15, 10, 0x07); | |
3904 | ||
3905 | for (i = 0; i < (full ? 16 : 8); i++) | |
3906 | aarch64_set_vec_u8 (cpu, vd, i, | |
3907 | ( aarch64_get_vec_u8 (cpu, vd, i) | |
3908 | & aarch64_get_vec_u8 (cpu, vn, i)) | |
3909 | | ((~ aarch64_get_vec_u8 (cpu, vd, i)) | |
3910 | & aarch64_get_vec_u8 (cpu, vm, i))); | |
3911 | } | |
3912 | ||
3913 | static void | |
3914 | do_vec_EOR (sim_cpu *cpu) | |
3915 | { | |
3916 | /* instr[31] = 0 | |
3917 | instr[30] = half (0)/full (1) | |
3918 | instr[29,21] = 10 1110 001 | |
3919 | instr[20,16] = Vm | |
3920 | instr[15,10] = 000111 | |
3921 | instr[9,5] = Vn | |
3922 | instr[4.0] = Vd. */ | |
3923 | ||
3924 | unsigned vm = uimm (aarch64_get_instr (cpu), 20, 16); | |
3925 | unsigned vn = uimm (aarch64_get_instr (cpu), 9, 5); | |
3926 | unsigned vd = uimm (aarch64_get_instr (cpu), 4, 0); | |
3927 | unsigned i; | |
3928 | int full = uimm (aarch64_get_instr (cpu), 30, 30); | |
3929 | ||
3930 | NYI_assert (29, 21, 0x171); | |
3931 | NYI_assert (15, 10, 0x07); | |
3932 | ||
3933 | for (i = 0; i < (full ? 4 : 2); i++) | |
3934 | aarch64_set_vec_u32 (cpu, vd, i, | |
3935 | aarch64_get_vec_u32 (cpu, vn, i) | |
3936 | ^ aarch64_get_vec_u32 (cpu, vm, i)); | |
3937 | } | |
3938 | ||
3939 | static void | |
3940 | do_vec_bit (sim_cpu *cpu) | |
3941 | { | |
3942 | /* instr[31] = 0 | |
3943 | instr[30] = half (0)/full (1) | |
3944 | instr[29,23] = 10 1110 1 | |
3945 | instr[22] = BIT (0) / BIF (1) | |
3946 | instr[21] = 1 | |
3947 | instr[20,16] = Vm | |
3948 | instr[15,10] = 0001 11 | |
3949 | instr[9,5] = Vn | |
3950 | instr[4.0] = Vd. */ | |
3951 | ||
3952 | unsigned vm = uimm (aarch64_get_instr (cpu), 20, 16); | |
3953 | unsigned vn = uimm (aarch64_get_instr (cpu), 9, 5); | |
3954 | unsigned vd = uimm (aarch64_get_instr (cpu), 4, 0); | |
3955 | unsigned full = uimm (aarch64_get_instr (cpu), 30, 30); | |
3956 | unsigned test_false = uimm (aarch64_get_instr (cpu), 22, 22); | |
3957 | unsigned i; | |
3958 | ||
3959 | NYI_assert (29, 23, 0x5D); | |
3960 | NYI_assert (21, 21, 1); | |
3961 | NYI_assert (15, 10, 0x07); | |
3962 | ||
3963 | if (test_false) | |
3964 | { | |
3965 | for (i = 0; i < (full ? 16 : 8); i++) | |
3966 | if (aarch64_get_vec_u32 (cpu, vn, i) == 0) | |
3967 | aarch64_set_vec_u32 (cpu, vd, i, aarch64_get_vec_u32 (cpu, vm, i)); | |
3968 | } | |
3969 | else | |
3970 | { | |
3971 | for (i = 0; i < (full ? 16 : 8); i++) | |
3972 | if (aarch64_get_vec_u32 (cpu, vn, i) != 0) | |
3973 | aarch64_set_vec_u32 (cpu, vd, i, aarch64_get_vec_u32 (cpu, vm, i)); | |
3974 | } | |
3975 | } | |
3976 | ||
3977 | static void | |
3978 | do_vec_ORN (sim_cpu *cpu) | |
3979 | { | |
3980 | /* instr[31] = 0 | |
3981 | instr[30] = half (0)/full (1) | |
3982 | instr[29,21] = 00 1110 111 | |
3983 | instr[20,16] = Vm | |
3984 | instr[15,10] = 00 0111 | |
3985 | instr[9,5] = Vn | |
3986 | instr[4.0] = Vd. */ | |
3987 | ||
3988 | unsigned vm = uimm (aarch64_get_instr (cpu), 20, 16); | |
3989 | unsigned vn = uimm (aarch64_get_instr (cpu), 9, 5); | |
3990 | unsigned vd = uimm (aarch64_get_instr (cpu), 4, 0); | |
3991 | unsigned i; | |
3992 | int full = uimm (aarch64_get_instr (cpu), 30, 30); | |
3993 | ||
3994 | NYI_assert (29, 21, 0x077); | |
3995 | NYI_assert (15, 10, 0x07); | |
3996 | ||
3997 | for (i = 0; i < (full ? 16 : 8); i++) | |
3998 | aarch64_set_vec_u8 (cpu, vd, i, | |
3999 | aarch64_get_vec_u8 (cpu, vn, i) | |
4000 | | ~ aarch64_get_vec_u8 (cpu, vm, i)); | |
4001 | } | |
4002 | ||
4003 | static void | |
4004 | do_vec_ORR (sim_cpu *cpu) | |
4005 | { | |
4006 | /* instr[31] = 0 | |
4007 | instr[30] = half (0)/full (1) | |
4008 | instr[29,21] = 00 1110 101 | |
4009 | instr[20,16] = Vm | |
4010 | instr[15,10] = 0001 11 | |
4011 | instr[9,5] = Vn | |
4012 | instr[4.0] = Vd. */ | |
4013 | ||
4014 | unsigned vm = uimm (aarch64_get_instr (cpu), 20, 16); | |
4015 | unsigned vn = uimm (aarch64_get_instr (cpu), 9, 5); | |
4016 | unsigned vd = uimm (aarch64_get_instr (cpu), 4, 0); | |
4017 | unsigned i; | |
4018 | int full = uimm (aarch64_get_instr (cpu), 30, 30); | |
4019 | ||
4020 | NYI_assert (29, 21, 0x075); | |
4021 | NYI_assert (15, 10, 0x07); | |
4022 | ||
4023 | for (i = 0; i < (full ? 16 : 8); i++) | |
4024 | aarch64_set_vec_u8 (cpu, vd, i, | |
4025 | aarch64_get_vec_u8 (cpu, vn, i) | |
4026 | | aarch64_get_vec_u8 (cpu, vm, i)); | |
4027 | } | |
4028 | ||
4029 | static void | |
4030 | do_vec_BIC (sim_cpu *cpu) | |
4031 | { | |
4032 | /* instr[31] = 0 | |
4033 | instr[30] = half (0)/full (1) | |
4034 | instr[29,21] = 00 1110 011 | |
4035 | instr[20,16] = Vm | |
4036 | instr[15,10] = 00 0111 | |
4037 | instr[9,5] = Vn | |
4038 | instr[4.0] = Vd. */ | |
4039 | ||
4040 | unsigned vm = uimm (aarch64_get_instr (cpu), 20, 16); | |
4041 | unsigned vn = uimm (aarch64_get_instr (cpu), 9, 5); | |
4042 | unsigned vd = uimm (aarch64_get_instr (cpu), 4, 0); | |
4043 | unsigned i; | |
4044 | int full = uimm (aarch64_get_instr (cpu), 30, 30); | |
4045 | ||
4046 | NYI_assert (29, 21, 0x073); | |
4047 | NYI_assert (15, 10, 0x07); | |
4048 | ||
4049 | for (i = 0; i < (full ? 16 : 8); i++) | |
4050 | aarch64_set_vec_u8 (cpu, vd, i, | |
4051 | aarch64_get_vec_u8 (cpu, vn, i) | |
4052 | & ~ aarch64_get_vec_u8 (cpu, vm, i)); | |
4053 | } | |
4054 | ||
4055 | static void | |
4056 | do_vec_XTN (sim_cpu *cpu) | |
4057 | { | |
4058 | /* instr[31] = 0 | |
4059 | instr[30] = first part (0)/ second part (1) | |
4060 | instr[29,24] = 00 1110 | |
4061 | instr[23,22] = size: byte(00), half(01), word (10) | |
4062 | instr[21,10] = 1000 0100 1010 | |
4063 | instr[9,5] = Vs | |
4064 | instr[4,0] = Vd. */ | |
4065 | ||
4066 | unsigned vs = uimm (aarch64_get_instr (cpu), 9, 5); | |
4067 | unsigned vd = uimm (aarch64_get_instr (cpu), 4, 0); | |
4068 | unsigned bias = uimm (aarch64_get_instr (cpu), 30, 30); | |
4069 | unsigned i; | |
4070 | ||
4071 | NYI_assert (29, 24, 0x0E); | |
4072 | NYI_assert (21, 10, 0x84A); | |
4073 | ||
4074 | switch (uimm (aarch64_get_instr (cpu), 23, 22)) | |
4075 | { | |
4076 | case 0: | |
4077 | if (bias) | |
4078 | for (i = 0; i < 8; i++) | |
4079 | aarch64_set_vec_u8 (cpu, vd, i + 8, | |
4080 | aarch64_get_vec_u16 (cpu, vs, i) >> 8); | |
4081 | else | |
4082 | for (i = 0; i < 8; i++) | |
4083 | aarch64_set_vec_u8 (cpu, vd, i, aarch64_get_vec_u16 (cpu, vs, i)); | |
4084 | return; | |
4085 | ||
4086 | case 1: | |
4087 | if (bias) | |
4088 | for (i = 0; i < 4; i++) | |
4089 | aarch64_set_vec_u16 (cpu, vd, i + 4, | |
4090 | aarch64_get_vec_u32 (cpu, vs, i) >> 16); | |
4091 | else | |
4092 | for (i = 0; i < 4; i++) | |
4093 | aarch64_set_vec_u16 (cpu, vd, i, aarch64_get_vec_u32 (cpu, vs, i)); | |
4094 | return; | |
4095 | ||
4096 | case 2: | |
4097 | if (bias) | |
4098 | for (i = 0; i < 2; i++) | |
4099 | aarch64_set_vec_u32 (cpu, vd, i + 4, | |
4100 | aarch64_get_vec_u64 (cpu, vs, i) >> 32); | |
4101 | else | |
4102 | for (i = 0; i < 2; i++) | |
4103 | aarch64_set_vec_u32 (cpu, vd, i, aarch64_get_vec_u64 (cpu, vs, i)); | |
4104 | return; | |
4105 | ||
4106 | default: | |
4107 | HALT_UNALLOC; | |
4108 | } | |
4109 | } | |
4110 | ||
2e8cf49e NC |
4111 | static void |
4112 | do_vec_maxv (sim_cpu *cpu) | |
4113 | { | |
4114 | /* instr[31] = 0 | |
4115 | instr[30] = half(0)/full(1) | |
4116 | instr[29] = signed (0)/unsigned(1) | |
4117 | instr[28,24] = 0 1110 | |
4118 | instr[23,22] = size: byte(00), half(01), word (10) | |
4119 | instr[21] = 1 | |
4120 | instr[20,17] = 1 000 | |
4121 | instr[16] = max(0)/min(1) | |
4122 | instr[15,10] = 1010 10 | |
4123 | instr[9,5] = V source | |
4124 | instr[4.0] = R dest. */ | |
4125 | ||
4126 | unsigned vs = uimm (aarch64_get_instr (cpu), 9, 5); | |
4127 | unsigned rd = uimm (aarch64_get_instr (cpu), 4, 0); | |
4128 | unsigned full = uimm (aarch64_get_instr (cpu), 30, 30); | |
4129 | unsigned i; | |
4130 | ||
4131 | NYI_assert (28, 24, 0x0E); | |
4132 | NYI_assert (21, 21, 1); | |
4133 | NYI_assert (20, 17, 8); | |
4134 | NYI_assert (15, 10, 0x2A); | |
4135 | ||
4136 | switch ((uimm (aarch64_get_instr (cpu), 29, 29) << 1) | |
4137 | | uimm (aarch64_get_instr (cpu), 16, 16)) | |
4138 | { | |
4139 | case 0: /* SMAXV. */ | |
4140 | { | |
4141 | int64_t smax; | |
4142 | switch (uimm (aarch64_get_instr (cpu), 23, 22)) | |
4143 | { | |
4144 | case 0: | |
4145 | smax = aarch64_get_vec_s8 (cpu, vs, 0); | |
4146 | for (i = 1; i < (full ? 16 : 8); i++) | |
bc273e17 | 4147 | smax = max (smax, aarch64_get_vec_s8 (cpu, vs, i)); |
2e8cf49e NC |
4148 | break; |
4149 | case 1: | |
4150 | smax = aarch64_get_vec_s16 (cpu, vs, 0); | |
4151 | for (i = 1; i < (full ? 8 : 4); i++) | |
bc273e17 | 4152 | smax = max (smax, aarch64_get_vec_s16 (cpu, vs, i)); |
2e8cf49e NC |
4153 | break; |
4154 | case 2: | |
4155 | smax = aarch64_get_vec_s32 (cpu, vs, 0); | |
4156 | for (i = 1; i < (full ? 4 : 2); i++) | |
bc273e17 | 4157 | smax = max (smax, aarch64_get_vec_s32 (cpu, vs, i)); |
2e8cf49e NC |
4158 | break; |
4159 | default: | |
4160 | case 3: | |
4161 | HALT_UNALLOC; | |
4162 | } | |
4163 | aarch64_set_reg_s64 (cpu, rd, NO_SP, smax); | |
4164 | return; | |
4165 | } | |
4166 | ||
4167 | case 1: /* SMINV. */ | |
4168 | { | |
4169 | int64_t smin; | |
4170 | switch (uimm (aarch64_get_instr (cpu), 23, 22)) | |
4171 | { | |
4172 | case 0: | |
4173 | smin = aarch64_get_vec_s8 (cpu, vs, 0); | |
4174 | for (i = 1; i < (full ? 16 : 8); i++) | |
bc273e17 | 4175 | smin = min (smin, aarch64_get_vec_s8 (cpu, vs, i)); |
2e8cf49e NC |
4176 | break; |
4177 | case 1: | |
4178 | smin = aarch64_get_vec_s16 (cpu, vs, 0); | |
4179 | for (i = 1; i < (full ? 8 : 4); i++) | |
bc273e17 | 4180 | smin = min (smin, aarch64_get_vec_s16 (cpu, vs, i)); |
2e8cf49e NC |
4181 | break; |
4182 | case 2: | |
4183 | smin = aarch64_get_vec_s32 (cpu, vs, 0); | |
4184 | for (i = 1; i < (full ? 4 : 2); i++) | |
bc273e17 | 4185 | smin = min (smin, aarch64_get_vec_s32 (cpu, vs, i)); |
2e8cf49e NC |
4186 | break; |
4187 | default: | |
4188 | case 3: | |
4189 | HALT_UNALLOC; | |
4190 | } | |
4191 | aarch64_set_reg_s64 (cpu, rd, NO_SP, smin); | |
4192 | return; | |
4193 | } | |
4194 | ||
4195 | case 2: /* UMAXV. */ | |
4196 | { | |
4197 | uint64_t umax; | |
4198 | switch (uimm (aarch64_get_instr (cpu), 23, 22)) | |
4199 | { | |
4200 | case 0: | |
4201 | umax = aarch64_get_vec_u8 (cpu, vs, 0); | |
4202 | for (i = 1; i < (full ? 16 : 8); i++) | |
bc273e17 | 4203 | umax = max (umax, aarch64_get_vec_u8 (cpu, vs, i)); |
2e8cf49e NC |
4204 | break; |
4205 | case 1: | |
4206 | umax = aarch64_get_vec_u16 (cpu, vs, 0); | |
4207 | for (i = 1; i < (full ? 8 : 4); i++) | |
bc273e17 | 4208 | umax = max (umax, aarch64_get_vec_u16 (cpu, vs, i)); |
2e8cf49e NC |
4209 | break; |
4210 | case 2: | |
4211 | umax = aarch64_get_vec_u32 (cpu, vs, 0); | |
4212 | for (i = 1; i < (full ? 4 : 2); i++) | |
bc273e17 | 4213 | umax = max (umax, aarch64_get_vec_u32 (cpu, vs, i)); |
2e8cf49e NC |
4214 | break; |
4215 | default: | |
4216 | case 3: | |
4217 | HALT_UNALLOC; | |
4218 | } | |
4219 | aarch64_set_reg_u64 (cpu, rd, NO_SP, umax); | |
4220 | return; | |
4221 | } | |
4222 | ||
4223 | case 3: /* UMINV. */ | |
4224 | { | |
4225 | uint64_t umin; | |
4226 | switch (uimm (aarch64_get_instr (cpu), 23, 22)) | |
4227 | { | |
4228 | case 0: | |
4229 | umin = aarch64_get_vec_u8 (cpu, vs, 0); | |
4230 | for (i = 1; i < (full ? 16 : 8); i++) | |
bc273e17 | 4231 | umin = min (umin, aarch64_get_vec_u8 (cpu, vs, i)); |
2e8cf49e NC |
4232 | break; |
4233 | case 1: | |
4234 | umin = aarch64_get_vec_u16 (cpu, vs, 0); | |
4235 | for (i = 1; i < (full ? 8 : 4); i++) | |
bc273e17 | 4236 | umin = min (umin, aarch64_get_vec_u16 (cpu, vs, i)); |
2e8cf49e NC |
4237 | break; |
4238 | case 2: | |
4239 | umin = aarch64_get_vec_u32 (cpu, vs, 0); | |
4240 | for (i = 1; i < (full ? 4 : 2); i++) | |
bc273e17 | 4241 | umin = min (umin, aarch64_get_vec_u32 (cpu, vs, i)); |
2e8cf49e NC |
4242 | break; |
4243 | default: | |
4244 | case 3: | |
4245 | HALT_UNALLOC; | |
4246 | } | |
4247 | aarch64_set_reg_u64 (cpu, rd, NO_SP, umin); | |
4248 | return; | |
4249 | } | |
4250 | ||
4251 | default: | |
4252 | HALT_UNALLOC; | |
4253 | } | |
4254 | } | |
4255 | ||
4256 | static void | |
4257 | do_vec_fminmaxV (sim_cpu *cpu) | |
4258 | { | |
4259 | /* instr[31,24] = 0110 1110 | |
4260 | instr[23] = max(0)/min(1) | |
4261 | instr[22,14] = 011 0000 11 | |
4262 | instr[13,12] = nm(00)/normal(11) | |
4263 | instr[11,10] = 10 | |
4264 | instr[9,5] = V source | |
4265 | instr[4.0] = R dest. */ | |
4266 | ||
4267 | unsigned vs = uimm (aarch64_get_instr (cpu), 9, 5); | |
4268 | unsigned rd = uimm (aarch64_get_instr (cpu), 4, 0); | |
4269 | unsigned i; | |
4270 | float res = aarch64_get_vec_float (cpu, vs, 0); | |
4271 | ||
4272 | NYI_assert (31, 24, 0x6E); | |
4273 | NYI_assert (22, 14, 0x0C3); | |
4274 | NYI_assert (11, 10, 2); | |
4275 | ||
4276 | if (uimm (aarch64_get_instr (cpu), 23, 23)) | |
4277 | { | |
4278 | switch (uimm (aarch64_get_instr (cpu), 13, 12)) | |
4279 | { | |
4280 | case 0: /* FMNINNMV. */ | |
4281 | for (i = 1; i < 4; i++) | |
4282 | res = fminnm (res, aarch64_get_vec_float (cpu, vs, i)); | |
4283 | break; | |
4284 | ||
4285 | case 3: /* FMINV. */ | |
4286 | for (i = 1; i < 4; i++) | |
bc273e17 | 4287 | res = min (res, aarch64_get_vec_float (cpu, vs, i)); |
2e8cf49e NC |
4288 | break; |
4289 | ||
4290 | default: | |
4291 | HALT_NYI; | |
4292 | } | |
4293 | } | |
4294 | else | |
4295 | { | |
4296 | switch (uimm (aarch64_get_instr (cpu), 13, 12)) | |
4297 | { | |
4298 | case 0: /* FMNAXNMV. */ | |
4299 | for (i = 1; i < 4; i++) | |
4300 | res = fmaxnm (res, aarch64_get_vec_float (cpu, vs, i)); | |
4301 | break; | |
4302 | ||
4303 | case 3: /* FMAXV. */ | |
4304 | for (i = 1; i < 4; i++) | |
bc273e17 | 4305 | res = max (res, aarch64_get_vec_float (cpu, vs, i)); |
2e8cf49e NC |
4306 | break; |
4307 | ||
4308 | default: | |
4309 | HALT_NYI; | |
4310 | } | |
4311 | } | |
4312 | ||
4313 | aarch64_set_FP_float (cpu, rd, res); | |
4314 | } | |
4315 | ||
4316 | static void | |
4317 | do_vec_Fminmax (sim_cpu *cpu) | |
4318 | { | |
4319 | /* instr[31] = 0 | |
4320 | instr[30] = half(0)/full(1) | |
4321 | instr[29,24] = 00 1110 | |
4322 | instr[23] = max(0)/min(1) | |
4323 | instr[22] = float(0)/double(1) | |
4324 | instr[21] = 1 | |
4325 | instr[20,16] = Vm | |
4326 | instr[15,14] = 11 | |
4327 | instr[13,12] = nm(00)/normal(11) | |
4328 | instr[11,10] = 01 | |
4329 | instr[9,5] = Vn | |
4330 | instr[4,0] = Vd. */ | |
4331 | ||
4332 | unsigned vm = uimm (aarch64_get_instr (cpu), 20, 16); | |
4333 | unsigned vn = uimm (aarch64_get_instr (cpu), 9, 5); | |
4334 | unsigned vd = uimm (aarch64_get_instr (cpu), 4, 0); | |
4335 | unsigned full = uimm (aarch64_get_instr (cpu), 30, 30); | |
4336 | unsigned min = uimm (aarch64_get_instr (cpu), 23, 23); | |
4337 | unsigned i; | |
4338 | ||
4339 | NYI_assert (29, 24, 0x0E); | |
4340 | NYI_assert (21, 21, 1); | |
4341 | NYI_assert (15, 14, 3); | |
4342 | NYI_assert (11, 10, 1); | |
4343 | ||
4344 | if (uimm (aarch64_get_instr (cpu), 22, 22)) | |
4345 | { | |
4346 | double (* func)(double, double); | |
4347 | ||
4348 | if (! full) | |
4349 | HALT_NYI; | |
4350 | ||
4351 | if (uimm (aarch64_get_instr (cpu), 13, 12) == 0) | |
4352 | func = min ? dminnm : dmaxnm; | |
4353 | else if (uimm (aarch64_get_instr (cpu), 13, 12) == 3) | |
4354 | func = min ? fmin : fmax; | |
4355 | else | |
4356 | HALT_NYI; | |
4357 | ||
4358 | for (i = 0; i < 2; i++) | |
4359 | aarch64_set_vec_double (cpu, vd, i, | |
4360 | func (aarch64_get_vec_double (cpu, vn, i), | |
4361 | aarch64_get_vec_double (cpu, vm, i))); | |
4362 | } | |
4363 | else | |
4364 | { | |
4365 | float (* func)(float, float); | |
4366 | ||
4367 | if (uimm (aarch64_get_instr (cpu), 13, 12) == 0) | |
4368 | func = min ? fminnm : fmaxnm; | |
4369 | else if (uimm (aarch64_get_instr (cpu), 13, 12) == 3) | |
4370 | func = min ? fminf : fmaxf; | |
4371 | else | |
4372 | HALT_NYI; | |
4373 | ||
4374 | for (i = 0; i < (full ? 4 : 2); i++) | |
4375 | aarch64_set_vec_float (cpu, vd, i, | |
4376 | func (aarch64_get_vec_float (cpu, vn, i), | |
4377 | aarch64_get_vec_float (cpu, vm, i))); | |
4378 | } | |
4379 | } | |
4380 | ||
4381 | static void | |
4382 | do_vec_SCVTF (sim_cpu *cpu) | |
4383 | { | |
4384 | /* instr[31] = 0 | |
4385 | instr[30] = Q | |
4386 | instr[29,23] = 00 1110 0 | |
4387 | instr[22] = float(0)/double(1) | |
4388 | instr[21,10] = 10 0001 1101 10 | |
4389 | instr[9,5] = Vn | |
4390 | instr[4,0] = Vd. */ | |
4391 | ||
4392 | unsigned vn = uimm (aarch64_get_instr (cpu), 9, 5); | |
4393 | unsigned vd = uimm (aarch64_get_instr (cpu), 4, 0); | |
4394 | unsigned full = uimm (aarch64_get_instr (cpu), 30, 30); | |
4395 | unsigned size = uimm (aarch64_get_instr (cpu), 22, 22); | |
4396 | unsigned i; | |
4397 | ||
4398 | NYI_assert (29, 23, 0x1C); | |
4399 | NYI_assert (21, 10, 0x876); | |
4400 | ||
4401 | if (size) | |
4402 | { | |
4403 | if (! full) | |
4404 | HALT_UNALLOC; | |
4405 | ||
4406 | for (i = 0; i < 2; i++) | |
4407 | { | |
4408 | double val = (double) aarch64_get_vec_u64 (cpu, vn, i); | |
4409 | aarch64_set_vec_double (cpu, vd, i, val); | |
4410 | } | |
4411 | } | |
4412 | else | |
4413 | { | |
4414 | for (i = 0; i < (full ? 4 : 2); i++) | |
4415 | { | |
4416 | float val = (float) aarch64_get_vec_u32 (cpu, vn, i); | |
4417 | aarch64_set_vec_float (cpu, vd, i, val); | |
4418 | } | |
4419 | } | |
4420 | } | |
4421 | ||
4422 | #define VEC_CMP(SOURCE, CMP) \ | |
4423 | do \ | |
4424 | { \ | |
4425 | switch (size) \ | |
4426 | { \ | |
4427 | case 0: \ | |
4428 | for (i = 0; i < (full ? 16 : 8); i++) \ | |
4429 | aarch64_set_vec_u8 (cpu, vd, i, \ | |
4430 | aarch64_get_vec_##SOURCE##8 (cpu, vn, i) \ | |
4431 | CMP \ | |
4432 | aarch64_get_vec_##SOURCE##8 (cpu, vm, i) \ | |
4433 | ? -1 : 0); \ | |
4434 | return; \ | |
4435 | case 1: \ | |
4436 | for (i = 0; i < (full ? 8 : 4); i++) \ | |
4437 | aarch64_set_vec_u16 (cpu, vd, i, \ | |
4438 | aarch64_get_vec_##SOURCE##16 (cpu, vn, i) \ | |
4439 | CMP \ | |
4440 | aarch64_get_vec_##SOURCE##16 (cpu, vm, i) \ | |
4441 | ? -1 : 0); \ | |
4442 | return; \ | |
4443 | case 2: \ | |
4444 | for (i = 0; i < (full ? 4 : 2); i++) \ | |
4445 | aarch64_set_vec_u32 (cpu, vd, i, \ | |
4446 | aarch64_get_vec_##SOURCE##32 (cpu, vn, i) \ | |
4447 | CMP \ | |
4448 | aarch64_get_vec_##SOURCE##32 (cpu, vm, i) \ | |
4449 | ? -1 : 0); \ | |
4450 | return; \ | |
4451 | case 3: \ | |
4452 | if (! full) \ | |
4453 | HALT_UNALLOC; \ | |
4454 | for (i = 0; i < 2; i++) \ | |
4455 | aarch64_set_vec_u64 (cpu, vd, i, \ | |
4456 | aarch64_get_vec_##SOURCE##64 (cpu, vn, i) \ | |
4457 | CMP \ | |
4458 | aarch64_get_vec_##SOURCE##64 (cpu, vm, i) \ | |
4459 | ? -1ULL : 0); \ | |
4460 | return; \ | |
4461 | default: \ | |
4462 | HALT_UNALLOC; \ | |
4463 | } \ | |
4464 | } \ | |
4465 | while (0) | |
4466 | ||
4467 | #define VEC_CMP0(SOURCE, CMP) \ | |
4468 | do \ | |
4469 | { \ | |
4470 | switch (size) \ | |
4471 | { \ | |
4472 | case 0: \ | |
4473 | for (i = 0; i < (full ? 16 : 8); i++) \ | |
4474 | aarch64_set_vec_u8 (cpu, vd, i, \ | |
4475 | aarch64_get_vec_##SOURCE##8 (cpu, vn, i) \ | |
4476 | CMP 0 ? -1 : 0); \ | |
4477 | return; \ | |
4478 | case 1: \ | |
4479 | for (i = 0; i < (full ? 8 : 4); i++) \ | |
4480 | aarch64_set_vec_u16 (cpu, vd, i, \ | |
4481 | aarch64_get_vec_##SOURCE##16 (cpu, vn, i) \ | |
4482 | CMP 0 ? -1 : 0); \ | |
4483 | return; \ | |
4484 | case 2: \ | |
4485 | for (i = 0; i < (full ? 4 : 2); i++) \ | |
4486 | aarch64_set_vec_u32 (cpu, vd, i, \ | |
4487 | aarch64_get_vec_##SOURCE##32 (cpu, vn, i) \ | |
4488 | CMP 0 ? -1 : 0); \ | |
4489 | return; \ | |
4490 | case 3: \ | |
4491 | if (! full) \ | |
4492 | HALT_UNALLOC; \ | |
4493 | for (i = 0; i < 2; i++) \ | |
4494 | aarch64_set_vec_u64 (cpu, vd, i, \ | |
4495 | aarch64_get_vec_##SOURCE##64 (cpu, vn, i) \ | |
4496 | CMP 0 ? -1ULL : 0); \ | |
4497 | return; \ | |
4498 | default: \ | |
4499 | HALT_UNALLOC; \ | |
4500 | } \ | |
4501 | } \ | |
4502 | while (0) | |
4503 | ||
4504 | #define VEC_FCMP0(CMP) \ | |
4505 | do \ | |
4506 | { \ | |
4507 | if (vm != 0) \ | |
4508 | HALT_NYI; \ | |
4509 | if (uimm (aarch64_get_instr (cpu), 22, 22)) \ | |
4510 | { \ | |
4511 | if (! full) \ | |
4512 | HALT_NYI; \ | |
4513 | for (i = 0; i < 2; i++) \ | |
4514 | aarch64_set_vec_u64 (cpu, vd, i, \ | |
4515 | aarch64_get_vec_double (cpu, vn, i) \ | |
4516 | CMP 0.0 ? -1 : 0); \ | |
4517 | } \ | |
4518 | else \ | |
4519 | { \ | |
4520 | for (i = 0; i < (full ? 4 : 2); i++) \ | |
4521 | aarch64_set_vec_u32 (cpu, vd, i, \ | |
4522 | aarch64_get_vec_float (cpu, vn, i) \ | |
4523 | CMP 0.0 ? -1 : 0); \ | |
4524 | } \ | |
4525 | return; \ | |
4526 | } \ | |
4527 | while (0) | |
4528 | ||
4529 | #define VEC_FCMP(CMP) \ | |
4530 | do \ | |
4531 | { \ | |
4532 | if (uimm (aarch64_get_instr (cpu), 22, 22)) \ | |
4533 | { \ | |
4534 | if (! full) \ | |
4535 | HALT_NYI; \ | |
4536 | for (i = 0; i < 2; i++) \ | |
4537 | aarch64_set_vec_u64 (cpu, vd, i, \ | |
4538 | aarch64_get_vec_double (cpu, vn, i) \ | |
4539 | CMP \ | |
4540 | aarch64_get_vec_double (cpu, vm, i) \ | |
4541 | ? -1 : 0); \ | |
4542 | } \ | |
4543 | else \ | |
4544 | { \ | |
4545 | for (i = 0; i < (full ? 4 : 2); i++) \ | |
4546 | aarch64_set_vec_u32 (cpu, vd, i, \ | |
4547 | aarch64_get_vec_float (cpu, vn, i) \ | |
4548 | CMP \ | |
4549 | aarch64_get_vec_float (cpu, vm, i) \ | |
4550 | ? -1 : 0); \ | |
4551 | } \ | |
4552 | return; \ | |
4553 | } \ | |
4554 | while (0) | |
4555 | ||
4556 | static void | |
4557 | do_vec_compare (sim_cpu *cpu) | |
4558 | { | |
4559 | /* instr[31] = 0 | |
4560 | instr[30] = half(0)/full(1) | |
4561 | instr[29] = part-of-comparison-type | |
4562 | instr[28,24] = 0 1110 | |
4563 | instr[23,22] = size of integer compares: byte(00), half(01), word (10), long (11) | |
4564 | type of float compares: single (-0) / double (-1) | |
4565 | instr[21] = 1 | |
4566 | instr[20,16] = Vm or 00000 (compare vs 0) | |
4567 | instr[15,10] = part-of-comparison-type | |
4568 | instr[9,5] = Vn | |
4569 | instr[4.0] = Vd. */ | |
4570 | ||
4571 | int full = uimm (aarch64_get_instr (cpu), 30, 30); | |
4572 | int size = uimm (aarch64_get_instr (cpu), 23, 22); | |
4573 | unsigned vm = uimm (aarch64_get_instr (cpu), 20, 16); | |
4574 | unsigned vn = uimm (aarch64_get_instr (cpu), 9, 5); | |
4575 | unsigned vd = uimm (aarch64_get_instr (cpu), 4, 0); | |
4576 | unsigned i; | |
4577 | ||
4578 | NYI_assert (28, 24, 0x0E); | |
4579 | NYI_assert (21, 21, 1); | |
4580 | ||
4581 | if ((uimm (aarch64_get_instr (cpu), 11, 11) | |
4582 | && uimm (aarch64_get_instr (cpu), 14, 14)) | |
4583 | || ((uimm (aarch64_get_instr (cpu), 11, 11) == 0 | |
4584 | && uimm (aarch64_get_instr (cpu), 10, 10) == 0))) | |
4585 | { | |
4586 | /* A compare vs 0. */ | |
4587 | if (vm != 0) | |
4588 | { | |
4589 | if (uimm (aarch64_get_instr (cpu), 15, 10) == 0x2A) | |
4590 | do_vec_maxv (cpu); | |
4591 | else if (uimm (aarch64_get_instr (cpu), 15, 10) == 0x32 | |
4592 | || uimm (aarch64_get_instr (cpu), 15, 10) == 0x3E) | |
4593 | do_vec_fminmaxV (cpu); | |
4594 | else if (uimm (aarch64_get_instr (cpu), 29, 23) == 0x1C | |
4595 | && uimm (aarch64_get_instr (cpu), 21, 10) == 0x876) | |
4596 | do_vec_SCVTF (cpu); | |
4597 | else | |
4598 | HALT_NYI; | |
4599 | return; | |
4600 | } | |
4601 | } | |
4602 | ||
4603 | if (uimm (aarch64_get_instr (cpu), 14, 14)) | |
4604 | { | |
4605 | /* A floating point compare. */ | |
4606 | unsigned decode = (uimm (aarch64_get_instr (cpu), 29, 29) << 5) | |
4607 | | (uimm (aarch64_get_instr (cpu), 23, 23) << 4) | |
4608 | | uimm (aarch64_get_instr (cpu), 13, 10); | |
4609 | ||
4610 | NYI_assert (15, 15, 1); | |
4611 | ||
4612 | switch (decode) | |
4613 | { | |
4614 | case /* 0b010010: GT#0 */ 0x12: VEC_FCMP0 (>); | |
4615 | case /* 0b110010: GE#0 */ 0x32: VEC_FCMP0 (>=); | |
4616 | case /* 0b010110: EQ#0 */ 0x16: VEC_FCMP0 (==); | |
4617 | case /* 0b110110: LE#0 */ 0x36: VEC_FCMP0 (<=); | |
4618 | case /* 0b011010: LT#0 */ 0x1A: VEC_FCMP0 (<); | |
4619 | case /* 0b111001: GT */ 0x39: VEC_FCMP (>); | |
4620 | case /* 0b101001: GE */ 0x29: VEC_FCMP (>=); | |
4621 | case /* 0b001001: EQ */ 0x09: VEC_FCMP (==); | |
4622 | ||
4623 | default: | |
4624 | HALT_NYI; | |
4625 | } | |
4626 | } | |
4627 | else | |
4628 | { | |
4629 | unsigned decode = (uimm (aarch64_get_instr (cpu), 29, 29) << 6) | |
4630 | | uimm (aarch64_get_instr (cpu), 15, 10); | |
4631 | ||
4632 | switch (decode) | |
4633 | { | |
4634 | case 0x0D: /* 0001101 GT */ VEC_CMP (s, > ); | |
4635 | case 0x0F: /* 0001111 GE */ VEC_CMP (s, >= ); | |
4636 | case 0x22: /* 0100010 GT #0 */ VEC_CMP0 (s, > ); | |
4637 | case 0x26: /* 0100110 EQ #0 */ VEC_CMP0 (s, == ); | |
4638 | case 0x2A: /* 0101010 LT #0 */ VEC_CMP0 (s, < ); | |
4639 | case 0x4D: /* 1001101 HI */ VEC_CMP (u, > ); | |
4640 | case 0x4F: /* 1001111 HS */ VEC_CMP (u, >= ); | |
4641 | case 0x62: /* 1100010 GE #0 */ VEC_CMP0 (s, >= ); | |
4642 | case 0x63: /* 1100011 EQ */ VEC_CMP (u, == ); | |
4643 | case 0x66: /* 1100110 LE #0 */ VEC_CMP0 (s, <= ); | |
4644 | default: | |
4645 | if (vm == 0) | |
4646 | HALT_NYI; | |
4647 | do_vec_maxv (cpu); | |
4648 | } | |
4649 | } | |
4650 | } | |
4651 | ||
4652 | static void | |
4653 | do_vec_SSHL (sim_cpu *cpu) | |
4654 | { | |
4655 | /* instr[31] = 0 | |
4656 | instr[30] = first part (0)/ second part (1) | |
4657 | instr[29,24] = 00 1110 | |
4658 | instr[23,22] = size: byte(00), half(01), word (10), long (11) | |
4659 | instr[21] = 1 | |
4660 | instr[20,16] = Vm | |
4661 | instr[15,10] = 0100 01 | |
4662 | instr[9,5] = Vn | |
4663 | instr[4,0] = Vd. */ | |
4664 | ||
4665 | unsigned full = uimm (aarch64_get_instr (cpu), 30, 30); | |
4666 | unsigned vm = uimm (aarch64_get_instr (cpu), 20, 16); | |
4667 | unsigned vn = uimm (aarch64_get_instr (cpu), 9, 5); | |
4668 | unsigned vd = uimm (aarch64_get_instr (cpu), 4, 0); | |
4669 | unsigned i; | |
4670 | ||
4671 | NYI_assert (29, 24, 0x0E); | |
4672 | NYI_assert (21, 21, 1); | |
4673 | NYI_assert (15, 10, 0x11); | |
4674 | ||
4675 | /* FIXME: What is a signed shift left in this context ?. */ | |
4676 | ||
4677 | switch (uimm (aarch64_get_instr (cpu), 23, 22)) | |
4678 | { | |
4679 | case 0: | |
4680 | for (i = 0; i < (full ? 16 : 8); i++) | |
4681 | aarch64_set_vec_s8 (cpu, vd, i, aarch64_get_vec_s8 (cpu, vn, i) | |
4682 | << aarch64_get_vec_s8 (cpu, vm, i)); | |
4683 | return; | |
4684 | ||
4685 | case 1: | |
4686 | for (i = 0; i < (full ? 8 : 4); i++) | |
4687 | aarch64_set_vec_s16 (cpu, vd, i, aarch64_get_vec_s16 (cpu, vn, i) | |
4688 | << aarch64_get_vec_s16 (cpu, vm, i)); | |
4689 | return; | |
4690 | ||
4691 | case 2: | |
4692 | for (i = 0; i < (full ? 4 : 2); i++) | |
4693 | aarch64_set_vec_s32 (cpu, vd, i, aarch64_get_vec_s32 (cpu, vn, i) | |
4694 | << aarch64_get_vec_s32 (cpu, vm, i)); | |
4695 | return; | |
4696 | ||
4697 | case 3: | |
4698 | if (! full) | |
4699 | HALT_UNALLOC; | |
4700 | for (i = 0; i < 2; i++) | |
4701 | aarch64_set_vec_s64 (cpu, vd, i, aarch64_get_vec_s64 (cpu, vn, i) | |
4702 | << aarch64_get_vec_s64 (cpu, vm, i)); | |
4703 | return; | |
4704 | ||
4705 | default: | |
4706 | HALT_NYI; | |
4707 | } | |
4708 | } | |
4709 | ||
4710 | static void | |
4711 | do_vec_USHL (sim_cpu *cpu) | |
4712 | { | |
4713 | /* instr[31] = 0 | |
4714 | instr[30] = first part (0)/ second part (1) | |
4715 | instr[29,24] = 10 1110 | |
4716 | instr[23,22] = size: byte(00), half(01), word (10), long (11) | |
4717 | instr[21] = 1 | |
4718 | instr[20,16] = Vm | |
4719 | instr[15,10] = 0100 01 | |
4720 | instr[9,5] = Vn | |
4721 | instr[4,0] = Vd */ | |
4722 | ||
4723 | unsigned full = uimm (aarch64_get_instr (cpu), 30, 30); | |
4724 | unsigned vm = uimm (aarch64_get_instr (cpu), 20, 16); | |
4725 | unsigned vn = uimm (aarch64_get_instr (cpu), 9, 5); | |
4726 | unsigned vd = uimm (aarch64_get_instr (cpu), 4, 0); | |
4727 | unsigned i; | |
4728 | ||
4729 | NYI_assert (29, 24, 0x2E); | |
4730 | NYI_assert (15, 10, 0x11); | |
4731 | ||
4732 | switch (uimm (aarch64_get_instr (cpu), 23, 22)) | |
4733 | { | |
4734 | case 0: | |
4735 | for (i = 0; i < (full ? 16 : 8); i++) | |
4736 | aarch64_set_vec_u8 (cpu, vd, i, aarch64_get_vec_u8 (cpu, vn, i) | |
4737 | << aarch64_get_vec_u8 (cpu, vm, i)); | |
4738 | return; | |
4739 | ||
4740 | case 1: | |
4741 | for (i = 0; i < (full ? 8 : 4); i++) | |
4742 | aarch64_set_vec_u16 (cpu, vd, i, aarch64_get_vec_u16 (cpu, vn, i) | |
4743 | << aarch64_get_vec_u16 (cpu, vm, i)); | |
4744 | return; | |
4745 | ||
4746 | case 2: | |
4747 | for (i = 0; i < (full ? 4 : 2); i++) | |
4748 | aarch64_set_vec_u32 (cpu, vd, i, aarch64_get_vec_u32 (cpu, vn, i) | |
4749 | << aarch64_get_vec_u32 (cpu, vm, i)); | |
4750 | return; | |
4751 | ||
4752 | case 3: | |
4753 | if (! full) | |
4754 | HALT_UNALLOC; | |
4755 | for (i = 0; i < 2; i++) | |
4756 | aarch64_set_vec_u64 (cpu, vd, i, aarch64_get_vec_u64 (cpu, vn, i) | |
4757 | << aarch64_get_vec_u64 (cpu, vm, i)); | |
4758 | return; | |
4759 | ||
4760 | default: | |
4761 | HALT_NYI; | |
4762 | } | |
4763 | } | |
4764 | ||
4765 | static void | |
4766 | do_vec_FMLA (sim_cpu *cpu) | |
4767 | { | |
4768 | /* instr[31] = 0 | |
4769 | instr[30] = full/half selector | |
4770 | instr[29,23] = 0011100 | |
4771 | instr[22] = size: 0=>float, 1=>double | |
4772 | instr[21] = 1 | |
4773 | instr[20,16] = Vn | |
4774 | instr[15,10] = 1100 11 | |
4775 | instr[9,5] = Vm | |
4776 | instr[4.0] = Vd. */ | |
4777 | ||
4778 | unsigned vm = uimm (aarch64_get_instr (cpu), 20, 16); | |
4779 | unsigned vn = uimm (aarch64_get_instr (cpu), 9, 5); | |
4780 | unsigned vd = uimm (aarch64_get_instr (cpu), 4, 0); | |
4781 | unsigned i; | |
4782 | int full = uimm (aarch64_get_instr (cpu), 30, 30); | |
4783 | ||
4784 | NYI_assert (29, 23, 0x1C); | |
4785 | NYI_assert (21, 21, 1); | |
4786 | NYI_assert (15, 10, 0x33); | |
4787 | ||
4788 | if (uimm (aarch64_get_instr (cpu), 22, 22)) | |
4789 | { | |
4790 | if (! full) | |
4791 | HALT_UNALLOC; | |
4792 | for (i = 0; i < 2; i++) | |
4793 | aarch64_set_vec_double (cpu, vd, i, | |
4794 | aarch64_get_vec_double (cpu, vn, i) * | |
4795 | aarch64_get_vec_double (cpu, vm, i) + | |
4796 | aarch64_get_vec_double (cpu, vd, i)); | |
4797 | } | |
4798 | else | |
4799 | { | |
4800 | for (i = 0; i < (full ? 4 : 2); i++) | |
4801 | aarch64_set_vec_float (cpu, vd, i, | |
4802 | aarch64_get_vec_float (cpu, vn, i) * | |
4803 | aarch64_get_vec_float (cpu, vm, i) + | |
4804 | aarch64_get_vec_float (cpu, vd, i)); | |
4805 | } | |
4806 | } | |
4807 | ||
4808 | static void | |
4809 | do_vec_max (sim_cpu *cpu) | |
4810 | { | |
4811 | /* instr[31] = 0 | |
4812 | instr[30] = full/half selector | |
4813 | instr[29] = SMAX (0) / UMAX (1) | |
4814 | instr[28,24] = 0 1110 | |
4815 | instr[23,22] = size: 00=> 8-bit, 01=> 16-bit, 10=> 32-bit | |
4816 | instr[21] = 1 | |
4817 | instr[20,16] = Vn | |
4818 | instr[15,10] = 0110 01 | |
4819 | instr[9,5] = Vm | |
4820 | instr[4.0] = Vd. */ | |
4821 | ||
4822 | unsigned vm = uimm (aarch64_get_instr (cpu), 20, 16); | |
4823 | unsigned vn = uimm (aarch64_get_instr (cpu), 9, 5); | |
4824 | unsigned vd = uimm (aarch64_get_instr (cpu), 4, 0); | |
4825 | unsigned i; | |
4826 | int full = uimm (aarch64_get_instr (cpu), 30, 30); | |
4827 | ||
4828 | NYI_assert (28, 24, 0x0E); | |
4829 | NYI_assert (21, 21, 1); | |
4830 | NYI_assert (15, 10, 0x19); | |
4831 | ||
4832 | if (uimm (aarch64_get_instr (cpu), 29, 29)) | |
4833 | { | |
4834 | switch (uimm (aarch64_get_instr (cpu), 23, 22)) | |
4835 | { | |
4836 | case 0: | |
4837 | for (i = 0; i < (full ? 16 : 8); i++) | |
4838 | aarch64_set_vec_u8 (cpu, vd, i, | |
4839 | aarch64_get_vec_u8 (cpu, vn, i) | |
4840 | > aarch64_get_vec_u8 (cpu, vm, i) | |
4841 | ? aarch64_get_vec_u8 (cpu, vn, i) | |
4842 | : aarch64_get_vec_u8 (cpu, vm, i)); | |
4843 | return; | |
4844 | ||
4845 | case 1: | |
4846 | for (i = 0; i < (full ? 8 : 4); i++) | |
4847 | aarch64_set_vec_u16 (cpu, vd, i, | |
4848 | aarch64_get_vec_u16 (cpu, vn, i) | |
4849 | > aarch64_get_vec_u16 (cpu, vm, i) | |
4850 | ? aarch64_get_vec_u16 (cpu, vn, i) | |
4851 | : aarch64_get_vec_u16 (cpu, vm, i)); | |
4852 | return; | |
4853 | ||
4854 | case 2: | |
4855 | for (i = 0; i < (full ? 4 : 2); i++) | |
4856 | aarch64_set_vec_u32 (cpu, vd, i, | |
4857 | aarch64_get_vec_u32 (cpu, vn, i) | |
4858 | > aarch64_get_vec_u32 (cpu, vm, i) | |
4859 | ? aarch64_get_vec_u32 (cpu, vn, i) | |
4860 | : aarch64_get_vec_u32 (cpu, vm, i)); | |
4861 | return; | |
4862 | ||
4863 | default: | |
4864 | case 3: | |
4865 | HALT_UNALLOC; | |
4866 | } | |
4867 | } | |
4868 | else | |
4869 | { | |
4870 | switch (uimm (aarch64_get_instr (cpu), 23, 22)) | |
4871 | { | |
4872 | case 0: | |
4873 | for (i = 0; i < (full ? 16 : 8); i++) | |
4874 | aarch64_set_vec_s8 (cpu, vd, i, | |
4875 | aarch64_get_vec_s8 (cpu, vn, i) | |
4876 | > aarch64_get_vec_s8 (cpu, vm, i) | |
4877 | ? aarch64_get_vec_s8 (cpu, vn, i) | |
4878 | : aarch64_get_vec_s8 (cpu, vm, i)); | |
4879 | return; | |
4880 | ||
4881 | case 1: | |
4882 | for (i = 0; i < (full ? 8 : 4); i++) | |
4883 | aarch64_set_vec_s16 (cpu, vd, i, | |
4884 | aarch64_get_vec_s16 (cpu, vn, i) | |
4885 | > aarch64_get_vec_s16 (cpu, vm, i) | |
4886 | ? aarch64_get_vec_s16 (cpu, vn, i) | |
4887 | : aarch64_get_vec_s16 (cpu, vm, i)); | |
4888 | return; | |
4889 | ||
4890 | case 2: | |
4891 | for (i = 0; i < (full ? 4 : 2); i++) | |
4892 | aarch64_set_vec_s32 (cpu, vd, i, | |
4893 | aarch64_get_vec_s32 (cpu, vn, i) | |
4894 | > aarch64_get_vec_s32 (cpu, vm, i) | |
4895 | ? aarch64_get_vec_s32 (cpu, vn, i) | |
4896 | : aarch64_get_vec_s32 (cpu, vm, i)); | |
4897 | return; | |
4898 | ||
4899 | default: | |
4900 | case 3: | |
4901 | HALT_UNALLOC; | |
4902 | } | |
4903 | } | |
4904 | } | |
4905 | ||
4906 | static void | |
4907 | do_vec_min (sim_cpu *cpu) | |
4908 | { | |
4909 | /* instr[31] = 0 | |
4910 | instr[30] = full/half selector | |
4911 | instr[29] = SMIN (0) / UMIN (1) | |
4912 | instr[28,24] = 0 1110 | |
4913 | instr[23,22] = size: 00=> 8-bit, 01=> 16-bit, 10=> 32-bit | |
4914 | instr[21] = 1 | |
4915 | instr[20,16] = Vn | |
4916 | instr[15,10] = 0110 11 | |
4917 | instr[9,5] = Vm | |
4918 | instr[4.0] = Vd. */ | |
4919 | ||
4920 | unsigned vm = uimm (aarch64_get_instr (cpu), 20, 16); | |
4921 | unsigned vn = uimm (aarch64_get_instr (cpu), 9, 5); | |
4922 | unsigned vd = uimm (aarch64_get_instr (cpu), 4, 0); | |
4923 | unsigned i; | |
4924 | int full = uimm (aarch64_get_instr (cpu), 30, 30); | |
4925 | ||
4926 | NYI_assert (28, 24, 0x0E); | |
4927 | NYI_assert (21, 21, 1); | |
4928 | NYI_assert (15, 10, 0x1B); | |
4929 | ||
4930 | if (uimm (aarch64_get_instr (cpu), 29, 29)) | |
4931 | { | |
4932 | switch (uimm (aarch64_get_instr (cpu), 23, 22)) | |
4933 | { | |
4934 | case 0: | |
4935 | for (i = 0; i < (full ? 16 : 8); i++) | |
4936 | aarch64_set_vec_u8 (cpu, vd, i, | |
4937 | aarch64_get_vec_u8 (cpu, vn, i) | |
4938 | < aarch64_get_vec_u8 (cpu, vm, i) | |
4939 | ? aarch64_get_vec_u8 (cpu, vn, i) | |
4940 | : aarch64_get_vec_u8 (cpu, vm, i)); | |
4941 | return; | |
4942 | ||
4943 | case 1: | |
4944 | for (i = 0; i < (full ? 8 : 4); i++) | |
4945 | aarch64_set_vec_u16 (cpu, vd, i, | |
4946 | aarch64_get_vec_u16 (cpu, vn, i) | |
4947 | < aarch64_get_vec_u16 (cpu, vm, i) | |
4948 | ? aarch64_get_vec_u16 (cpu, vn, i) | |
4949 | : aarch64_get_vec_u16 (cpu, vm, i)); | |
4950 | return; | |
4951 | ||
4952 | case 2: | |
4953 | for (i = 0; i < (full ? 4 : 2); i++) | |
4954 | aarch64_set_vec_u32 (cpu, vd, i, | |
4955 | aarch64_get_vec_u32 (cpu, vn, i) | |
4956 | < aarch64_get_vec_u32 (cpu, vm, i) | |
4957 | ? aarch64_get_vec_u32 (cpu, vn, i) | |
4958 | : aarch64_get_vec_u32 (cpu, vm, i)); | |
4959 | return; | |
4960 | ||
4961 | default: | |
4962 | case 3: | |
4963 | HALT_UNALLOC; | |
4964 | } | |
4965 | } | |
4966 | else | |
4967 | { | |
4968 | switch (uimm (aarch64_get_instr (cpu), 23, 22)) | |
4969 | { | |
4970 | case 0: | |
4971 | for (i = 0; i < (full ? 16 : 8); i++) | |
4972 | aarch64_set_vec_s8 (cpu, vd, i, | |
4973 | aarch64_get_vec_s8 (cpu, vn, i) | |
4974 | < aarch64_get_vec_s8 (cpu, vm, i) | |
4975 | ? aarch64_get_vec_s8 (cpu, vn, i) | |
4976 | : aarch64_get_vec_s8 (cpu, vm, i)); | |
4977 | return; | |
4978 | ||
4979 | case 1: | |
4980 | for (i = 0; i < (full ? 8 : 4); i++) | |
4981 | aarch64_set_vec_s16 (cpu, vd, i, | |
4982 | aarch64_get_vec_s16 (cpu, vn, i) | |
4983 | < aarch64_get_vec_s16 (cpu, vm, i) | |
4984 | ? aarch64_get_vec_s16 (cpu, vn, i) | |
4985 | : aarch64_get_vec_s16 (cpu, vm, i)); | |
4986 | return; | |
4987 | ||
4988 | case 2: | |
4989 | for (i = 0; i < (full ? 4 : 2); i++) | |
4990 | aarch64_set_vec_s32 (cpu, vd, i, | |
4991 | aarch64_get_vec_s32 (cpu, vn, i) | |
4992 | < aarch64_get_vec_s32 (cpu, vm, i) | |
4993 | ? aarch64_get_vec_s32 (cpu, vn, i) | |
4994 | : aarch64_get_vec_s32 (cpu, vm, i)); | |
4995 | return; | |
4996 | ||
4997 | default: | |
4998 | case 3: | |
4999 | HALT_UNALLOC; | |
5000 | } | |
5001 | } | |
5002 | } | |
5003 | ||
5004 | static void | |
5005 | do_vec_sub_long (sim_cpu *cpu) | |
5006 | { | |
5007 | /* instr[31] = 0 | |
5008 | instr[30] = lower (0) / upper (1) | |
5009 | instr[29] = signed (0) / unsigned (1) | |
5010 | instr[28,24] = 0 1110 | |
5011 | instr[23,22] = size: bytes (00), half (01), word (10) | |
5012 | instr[21] = 1 | |
5013 | insrt[20,16] = Vm | |
5014 | instr[15,10] = 0010 00 | |
5015 | instr[9,5] = Vn | |
5016 | instr[4,0] = V dest. */ | |
5017 | ||
5018 | unsigned size = uimm (aarch64_get_instr (cpu), 23, 22); | |
5019 | unsigned vm = uimm (aarch64_get_instr (cpu), 20, 16); | |
5020 | unsigned vn = uimm (aarch64_get_instr (cpu), 9, 5); | |
5021 | unsigned vd = uimm (aarch64_get_instr (cpu), 4, 0); | |
5022 | unsigned bias = 0; | |
5023 | unsigned i; | |
5024 | ||
5025 | NYI_assert (28, 24, 0x0E); | |
5026 | NYI_assert (21, 21, 1); | |
5027 | NYI_assert (15, 10, 0x08); | |
5028 | ||
5029 | if (size == 3) | |
5030 | HALT_UNALLOC; | |
5031 | ||
5032 | switch (uimm (aarch64_get_instr (cpu), 30, 29)) | |
5033 | { | |
5034 | case 2: /* SSUBL2. */ | |
5035 | bias = 2; | |
5036 | case 0: /* SSUBL. */ | |
5037 | switch (size) | |
5038 | { | |
5039 | case 0: | |
5040 | bias *= 3; | |
5041 | for (i = 0; i < 8; i++) | |
5042 | aarch64_set_vec_s16 (cpu, vd, i, | |
5043 | aarch64_get_vec_s8 (cpu, vn, i + bias) | |
5044 | - aarch64_get_vec_s8 (cpu, vm, i + bias)); | |
5045 | break; | |
5046 | ||
5047 | case 1: | |
5048 | bias *= 2; | |
5049 | for (i = 0; i < 4; i++) | |
5050 | aarch64_set_vec_s32 (cpu, vd, i, | |
5051 | aarch64_get_vec_s16 (cpu, vn, i + bias) | |
5052 | - aarch64_get_vec_s16 (cpu, vm, i + bias)); | |
5053 | break; | |
5054 | ||
5055 | case 2: | |
5056 | for (i = 0; i < 2; i++) | |
5057 | aarch64_set_vec_s64 (cpu, vd, i, | |
5058 | aarch64_get_vec_s32 (cpu, vn, i + bias) | |
5059 | - aarch64_get_vec_s32 (cpu, vm, i + bias)); | |
5060 | break; | |
5061 | ||
5062 | default: | |
5063 | HALT_UNALLOC; | |
5064 | } | |
5065 | break; | |
5066 | ||
5067 | case 3: /* USUBL2. */ | |
5068 | bias = 2; | |
5069 | case 1: /* USUBL. */ | |
5070 | switch (size) | |
5071 | { | |
5072 | case 0: | |
5073 | bias *= 3; | |
5074 | for (i = 0; i < 8; i++) | |
5075 | aarch64_set_vec_u16 (cpu, vd, i, | |
5076 | aarch64_get_vec_u8 (cpu, vn, i + bias) | |
5077 | - aarch64_get_vec_u8 (cpu, vm, i + bias)); | |
5078 | break; | |
5079 | ||
5080 | case 1: | |
5081 | bias *= 2; | |
5082 | for (i = 0; i < 4; i++) | |
5083 | aarch64_set_vec_u32 (cpu, vd, i, | |
5084 | aarch64_get_vec_u16 (cpu, vn, i + bias) | |
5085 | - aarch64_get_vec_u16 (cpu, vm, i + bias)); | |
5086 | break; | |
5087 | ||
5088 | case 2: | |
5089 | for (i = 0; i < 2; i++) | |
5090 | aarch64_set_vec_u64 (cpu, vd, i, | |
5091 | aarch64_get_vec_u32 (cpu, vn, i + bias) | |
5092 | - aarch64_get_vec_u32 (cpu, vm, i + bias)); | |
5093 | break; | |
5094 | ||
5095 | default: | |
5096 | HALT_UNALLOC; | |
5097 | } | |
5098 | break; | |
5099 | } | |
5100 | } | |
5101 | ||
5102 | #define DO_ADDP(FN) \ | |
5103 | do \ | |
5104 | { \ | |
5105 | for (i = 0; i < range; i++) \ | |
5106 | { \ | |
5107 | aarch64_set_vec_##FN (cpu, vd, i, \ | |
5108 | aarch64_get_vec_##FN (cpu, vn, i * 2) \ | |
5109 | + aarch64_get_vec_##FN (cpu, vn, i * 2 + 1)); \ | |
5110 | aarch64_set_vec_##FN (cpu, vd, i + range, \ | |
5111 | aarch64_get_vec_##FN (cpu, vm, i * 2) \ | |
5112 | + aarch64_get_vec_##FN (cpu, vm, i * 2 + 1)); \ | |
5113 | } \ | |
5114 | } \ | |
5115 | while (0) | |
5116 | ||
5117 | static void | |
5118 | do_vec_ADDP (sim_cpu *cpu) | |
5119 | { | |
5120 | /* instr[31] = 0 | |
5121 | instr[30] = half(0)/full(1) | |
5122 | instr[29,24] = 00 1110 | |
5123 | instr[23,22] = size: bytes (00), half (01), word (10), long (11) | |
5124 | instr[21] = 1 | |
5125 | insrt[20,16] = Vm | |
5126 | instr[15,10] = 1011 11 | |
5127 | instr[9,5] = Vn | |
5128 | instr[4,0] = V dest. */ | |
5129 | ||
5130 | unsigned full = uimm (aarch64_get_instr (cpu), 30, 30); | |
5131 | unsigned size = uimm (aarch64_get_instr (cpu), 23, 22); | |
5132 | unsigned vm = uimm (aarch64_get_instr (cpu), 20, 16); | |
5133 | unsigned vn = uimm (aarch64_get_instr (cpu), 9, 5); | |
5134 | unsigned vd = uimm (aarch64_get_instr (cpu), 4, 0); | |
5135 | unsigned i, range; | |
5136 | ||
5137 | NYI_assert (29, 24, 0x0E); | |
5138 | NYI_assert (21, 21, 1); | |
5139 | NYI_assert (15, 10, 0x2F); | |
5140 | ||
5141 | switch (size) | |
5142 | { | |
5143 | case 0: | |
5144 | range = full ? 8 : 4; | |
5145 | DO_ADDP (u8); | |
5146 | return; | |
5147 | ||
5148 | case 1: | |
5149 | range = full ? 4 : 2; | |
5150 | DO_ADDP (u16); | |
5151 | return; | |
5152 | ||
5153 | case 2: | |
5154 | range = full ? 2 : 1; | |
5155 | DO_ADDP (u32); | |
5156 | return; | |
5157 | ||
5158 | case 3: | |
5159 | if (! full) | |
5160 | HALT_UNALLOC; | |
5161 | range = 1; | |
5162 | DO_ADDP (u64); | |
5163 | return; | |
5164 | ||
5165 | default: | |
5166 | HALT_NYI; | |
5167 | } | |
5168 | } | |
5169 | ||
5170 | static void | |
5171 | do_vec_UMOV (sim_cpu *cpu) | |
5172 | { | |
5173 | /* instr[31] = 0 | |
5174 | instr[30] = 32-bit(0)/64-bit(1) | |
5175 | instr[29,21] = 00 1110 000 | |
5176 | insrt[20,16] = size & index | |
5177 | instr[15,10] = 0011 11 | |
5178 | instr[9,5] = V source | |
5179 | instr[4,0] = R dest. */ | |
5180 | ||
5181 | unsigned vs = uimm (aarch64_get_instr (cpu), 9, 5); | |
5182 | unsigned rd = uimm (aarch64_get_instr (cpu), 4, 0); | |
5183 | unsigned index; | |
5184 | ||
5185 | NYI_assert (29, 21, 0x070); | |
5186 | NYI_assert (15, 10, 0x0F); | |
5187 | ||
5188 | if (uimm (aarch64_get_instr (cpu), 16, 16)) | |
5189 | { | |
5190 | /* Byte transfer. */ | |
5191 | index = uimm (aarch64_get_instr (cpu), 20, 17); | |
5192 | aarch64_set_reg_u64 (cpu, rd, NO_SP, | |
5193 | aarch64_get_vec_u8 (cpu, vs, index)); | |
5194 | } | |
5195 | else if (uimm (aarch64_get_instr (cpu), 17, 17)) | |
5196 | { | |
5197 | index = uimm (aarch64_get_instr (cpu), 20, 18); | |
5198 | aarch64_set_reg_u64 (cpu, rd, NO_SP, | |
5199 | aarch64_get_vec_u16 (cpu, vs, index)); | |
5200 | } | |
5201 | else if (uimm (aarch64_get_instr (cpu), 18, 18)) | |
5202 | { | |
5203 | index = uimm (aarch64_get_instr (cpu), 20, 19); | |
5204 | aarch64_set_reg_u64 (cpu, rd, NO_SP, | |
5205 | aarch64_get_vec_u32 (cpu, vs, index)); | |
5206 | } | |
5207 | else | |
5208 | { | |
5209 | if (uimm (aarch64_get_instr (cpu), 30, 30) != 1) | |
5210 | HALT_UNALLOC; | |
5211 | ||
5212 | index = uimm (aarch64_get_instr (cpu), 20, 20); | |
5213 | aarch64_set_reg_u64 (cpu, rd, NO_SP, | |
5214 | aarch64_get_vec_u64 (cpu, vs, index)); | |
5215 | } | |
5216 | } | |
5217 | ||
5218 | static void | |
5219 | do_vec_FABS (sim_cpu *cpu) | |
5220 | { | |
5221 | /* instr[31] = 0 | |
5222 | instr[30] = half(0)/full(1) | |
5223 | instr[29,23] = 00 1110 1 | |
5224 | instr[22] = float(0)/double(1) | |
5225 | instr[21,16] = 10 0000 | |
5226 | instr[15,10] = 1111 10 | |
5227 | instr[9,5] = Vn | |
5228 | instr[4,0] = Vd. */ | |
5229 | ||
5230 | unsigned vn = uimm (aarch64_get_instr (cpu), 9, 5); | |
5231 | unsigned vd = uimm (aarch64_get_instr (cpu), 4, 0); | |
5232 | unsigned full = uimm (aarch64_get_instr (cpu), 30, 30); | |
5233 | unsigned i; | |
5234 | ||
5235 | NYI_assert (29, 23, 0x1D); | |
5236 | NYI_assert (21, 10, 0x83E); | |
5237 | ||
5238 | if (uimm (aarch64_get_instr (cpu), 22, 22)) | |
5239 | { | |
5240 | if (! full) | |
5241 | HALT_NYI; | |
5242 | ||
5243 | for (i = 0; i < 2; i++) | |
5244 | aarch64_set_vec_double (cpu, vd, i, | |
5245 | fabs (aarch64_get_vec_double (cpu, vn, i))); | |
5246 | } | |
5247 | else | |
5248 | { | |
5249 | for (i = 0; i < (full ? 4 : 2); i++) | |
5250 | aarch64_set_vec_float (cpu, vd, i, | |
5251 | fabsf (aarch64_get_vec_float (cpu, vn, i))); | |
5252 | } | |
5253 | } | |
5254 | ||
5255 | static void | |
5256 | do_vec_FCVTZS (sim_cpu *cpu) | |
5257 | { | |
5258 | /* instr[31] = 0 | |
5259 | instr[30] = half (0) / all (1) | |
5260 | instr[29,23] = 00 1110 1 | |
5261 | instr[22] = single (0) / double (1) | |
5262 | instr[21,10] = 10 0001 1011 10 | |
5263 | instr[9,5] = Rn | |
5264 | instr[4,0] = Rd. */ | |
5265 | ||
5266 | unsigned rn = uimm (aarch64_get_instr (cpu), 9, 5); | |
5267 | unsigned rd = uimm (aarch64_get_instr (cpu), 4, 0); | |
5268 | unsigned full = uimm (aarch64_get_instr (cpu), 30, 30); | |
5269 | unsigned i; | |
5270 | ||
5271 | NYI_assert (31, 31, 0); | |
5272 | NYI_assert (29, 23, 0x1D); | |
5273 | NYI_assert (21, 10, 0x86E); | |
5274 | ||
5275 | if (uimm (aarch64_get_instr (cpu), 22, 22)) | |
5276 | { | |
5277 | if (! full) | |
5278 | HALT_UNALLOC; | |
5279 | ||
5280 | for (i = 0; i < 2; i++) | |
5281 | aarch64_set_vec_s64 (cpu, rd, i, | |
5282 | (int64_t) aarch64_get_vec_double (cpu, rn, i)); | |
5283 | } | |
5284 | else | |
5285 | for (i = 0; i < (full ? 4 : 2); i++) | |
5286 | aarch64_set_vec_s32 (cpu, rd, i, | |
5287 | (int32_t) aarch64_get_vec_float (cpu, rn, i)); | |
5288 | } | |
5289 | ||
5290 | static void | |
5291 | do_vec_op1 (sim_cpu *cpu) | |
5292 | { | |
5293 | /* instr[31] = 0 | |
5294 | instr[30] = half/full | |
5295 | instr[29,24] = 00 1110 | |
5296 | instr[23,21] = ??? | |
5297 | instr[20,16] = Vm | |
5298 | instr[15,10] = sub-opcode | |
5299 | instr[9,5] = Vn | |
5300 | instr[4,0] = Vd */ | |
5301 | NYI_assert (29, 24, 0x0E); | |
5302 | ||
5303 | if (uimm (aarch64_get_instr (cpu), 21, 21) == 0) | |
5304 | { | |
5305 | if (uimm (aarch64_get_instr (cpu), 23, 22) == 0) | |
5306 | { | |
5307 | if (uimm (aarch64_get_instr (cpu), 30, 30) == 1 | |
5308 | && uimm (aarch64_get_instr (cpu), 17, 14) == 0 | |
5309 | && uimm (aarch64_get_instr (cpu), 12, 10) == 7) | |
5310 | return do_vec_ins_2 (cpu); | |
5311 | ||
5312 | switch (uimm (aarch64_get_instr (cpu), 15, 10)) | |
5313 | { | |
5314 | case 0x01: do_vec_DUP_vector_into_vector (cpu); return; | |
5315 | case 0x03: do_vec_DUP_scalar_into_vector (cpu); return; | |
5316 | case 0x07: do_vec_INS (cpu); return; | |
5317 | case 0x0A: do_vec_TRN (cpu); return; | |
5318 | ||
5319 | case 0x0F: | |
5320 | if (uimm (aarch64_get_instr (cpu), 17, 16) == 0) | |
5321 | { | |
5322 | do_vec_MOV_into_scalar (cpu); | |
5323 | return; | |
5324 | } | |
5325 | break; | |
5326 | ||
5327 | case 0x00: | |
5328 | case 0x08: | |
5329 | case 0x10: | |
5330 | case 0x18: | |
5331 | do_vec_TBL (cpu); return; | |
5332 | ||
5333 | case 0x06: | |
5334 | case 0x16: | |
5335 | do_vec_UZP (cpu); return; | |
5336 | ||
5337 | case 0x0E: | |
5338 | case 0x1E: | |
5339 | do_vec_ZIP (cpu); return; | |
5340 | ||
5341 | default: | |
5342 | HALT_NYI; | |
5343 | } | |
5344 | } | |
5345 | ||
5346 | switch (uimm (aarch64_get_instr (cpu), 13, 10)) | |
5347 | { | |
5348 | case 0x6: do_vec_UZP (cpu); return; | |
5349 | case 0xE: do_vec_ZIP (cpu); return; | |
5350 | case 0xA: do_vec_TRN (cpu); return; | |
5351 | case 0xF: do_vec_UMOV (cpu); return; | |
5352 | default: HALT_NYI; | |
5353 | } | |
5354 | } | |
5355 | ||
5356 | switch (uimm (aarch64_get_instr (cpu), 15, 10)) | |
5357 | { | |
5358 | case 0x07: | |
5359 | switch (uimm (aarch64_get_instr (cpu), 23, 21)) | |
5360 | { | |
5361 | case 1: do_vec_AND (cpu); return; | |
5362 | case 3: do_vec_BIC (cpu); return; | |
5363 | case 5: do_vec_ORR (cpu); return; | |
5364 | case 7: do_vec_ORN (cpu); return; | |
5365 | default: HALT_NYI; | |
5366 | } | |
5367 | ||
5368 | case 0x08: do_vec_sub_long (cpu); return; | |
5369 | case 0x0a: do_vec_XTN (cpu); return; | |
5370 | case 0x11: do_vec_SSHL (cpu); return; | |
5371 | case 0x19: do_vec_max (cpu); return; | |
5372 | case 0x1B: do_vec_min (cpu); return; | |
5373 | case 0x21: do_vec_add (cpu); return; | |
5374 | case 0x25: do_vec_MLA (cpu); return; | |
5375 | case 0x27: do_vec_mul (cpu); return; | |
5376 | case 0x2F: do_vec_ADDP (cpu); return; | |
5377 | case 0x30: do_vec_mull (cpu); return; | |
5378 | case 0x33: do_vec_FMLA (cpu); return; | |
5379 | case 0x35: do_vec_fadd (cpu); return; | |
5380 | ||
5381 | case 0x2E: | |
5382 | switch (uimm (aarch64_get_instr (cpu), 20, 16)) | |
5383 | { | |
5384 | case 0x00: do_vec_ABS (cpu); return; | |
5385 | case 0x01: do_vec_FCVTZS (cpu); return; | |
5386 | case 0x11: do_vec_ADDV (cpu); return; | |
5387 | default: HALT_NYI; | |
5388 | } | |
5389 | ||
5390 | case 0x31: | |
5391 | case 0x3B: | |
5392 | do_vec_Fminmax (cpu); return; | |
5393 | ||
5394 | case 0x0D: | |
5395 | case 0x0F: | |
5396 | case 0x22: | |
5397 | case 0x23: | |
5398 | case 0x26: | |
5399 | case 0x2A: | |
5400 | case 0x32: | |
5401 | case 0x36: | |
5402 | case 0x39: | |
5403 | case 0x3A: | |
5404 | do_vec_compare (cpu); return; | |
5405 | ||
5406 | case 0x3E: | |
5407 | do_vec_FABS (cpu); return; | |
5408 | ||
5409 | default: | |
5410 | HALT_NYI; | |
5411 | } | |
5412 | } | |
5413 | ||
5414 | static void | |
5415 | do_vec_xtl (sim_cpu *cpu) | |
5416 | { | |
5417 | /* instr[31] = 0 | |
5418 | instr[30,29] = SXTL (00), UXTL (01), SXTL2 (10), UXTL2 (11) | |
5419 | instr[28,22] = 0 1111 00 | |
5420 | instr[21,16] = size & shift (USHLL, SSHLL, USHLL2, SSHLL2) | |
5421 | instr[15,10] = 1010 01 | |
5422 | instr[9,5] = V source | |
5423 | instr[4,0] = V dest. */ | |
5424 | ||
5425 | unsigned vs = uimm (aarch64_get_instr (cpu), 9, 5); | |
5426 | unsigned vd = uimm (aarch64_get_instr (cpu), 4, 0); | |
5427 | unsigned i, shift, bias = 0; | |
5428 | ||
5429 | NYI_assert (28, 22, 0x3C); | |
5430 | NYI_assert (15, 10, 0x29); | |
5431 | ||
5432 | switch (uimm (aarch64_get_instr (cpu), 30, 29)) | |
5433 | { | |
5434 | case 2: /* SXTL2, SSHLL2. */ | |
5435 | bias = 2; | |
5436 | case 0: /* SXTL, SSHLL. */ | |
5437 | if (uimm (aarch64_get_instr (cpu), 21, 21)) | |
5438 | { | |
5439 | shift = uimm (aarch64_get_instr (cpu), 20, 16); | |
5440 | aarch64_set_vec_s64 | |
5441 | (cpu, vd, 0, aarch64_get_vec_s32 (cpu, vs, bias) << shift); | |
5442 | aarch64_set_vec_s64 | |
5443 | (cpu, vd, 1, aarch64_get_vec_s32 (cpu, vs, bias + 1) << shift); | |
5444 | } | |
5445 | else if (uimm (aarch64_get_instr (cpu), 20, 20)) | |
5446 | { | |
5447 | shift = uimm (aarch64_get_instr (cpu), 19, 16); | |
5448 | bias *= 2; | |
5449 | for (i = 0; i < 4; i++) | |
5450 | aarch64_set_vec_s32 | |
5451 | (cpu, vd, i, aarch64_get_vec_s16 (cpu, vs, i + bias) << shift); | |
5452 | } | |
5453 | else | |
5454 | { | |
5455 | NYI_assert (19, 19, 1); | |
5456 | ||
5457 | shift = uimm (aarch64_get_instr (cpu), 18, 16); | |
5458 | bias *= 3; | |
5459 | for (i = 0; i < 8; i++) | |
5460 | aarch64_set_vec_s16 | |
5461 | (cpu, vd, i, aarch64_get_vec_s8 (cpu, vs, i + bias) << shift); | |
5462 | } | |
5463 | return; | |
5464 | ||
5465 | case 3: /* UXTL2, USHLL2. */ | |
5466 | bias = 2; | |
5467 | case 1: /* UXTL, USHLL. */ | |
5468 | if (uimm (aarch64_get_instr (cpu), 21, 21)) | |
5469 | { | |
5470 | shift = uimm (aarch64_get_instr (cpu), 20, 16); | |
5471 | aarch64_set_vec_u64 | |
5472 | (cpu, vd, 0, aarch64_get_vec_u32 (cpu, vs, bias) << shift); | |
5473 | aarch64_set_vec_u64 | |
5474 | (cpu, vd, 1, aarch64_get_vec_u32 (cpu, vs, bias + 1) << shift); | |
5475 | } | |
5476 | else if (uimm (aarch64_get_instr (cpu), 20, 20)) | |
5477 | { | |
5478 | shift = uimm (aarch64_get_instr (cpu), 19, 16); | |
5479 | bias *= 2; | |
5480 | for (i = 0; i < 4; i++) | |
5481 | aarch64_set_vec_u32 | |
5482 | (cpu, vd, i, aarch64_get_vec_u16 (cpu, vs, i + bias) << shift); | |
5483 | } | |
5484 | else | |
5485 | { | |
5486 | NYI_assert (19, 19, 1); | |
5487 | ||
5488 | shift = uimm (aarch64_get_instr (cpu), 18, 16); | |
5489 | bias *= 3; | |
5490 | for (i = 0; i < 8; i++) | |
5491 | aarch64_set_vec_u16 | |
5492 | (cpu, vd, i, aarch64_get_vec_u8 (cpu, vs, i + bias) << shift); | |
5493 | } | |
5494 | return; | |
5495 | ||
5496 | default: | |
5497 | HALT_NYI; | |
5498 | } | |
5499 | } | |
5500 | ||
5501 | static void | |
5502 | do_vec_SHL (sim_cpu *cpu) | |
5503 | { | |
5504 | /* instr [31] = 0 | |
5505 | instr [30] = half(0)/full(1) | |
5506 | instr [29,23] = 001 1110 | |
5507 | instr [22,16] = size and shift amount | |
5508 | instr [15,10] = 01 0101 | |
5509 | instr [9, 5] = Vs | |
5510 | instr [4, 0] = Vd. */ | |
5511 | ||
5512 | int shift; | |
5513 | int full = uimm (aarch64_get_instr (cpu), 30, 30); | |
5514 | unsigned vs = uimm (aarch64_get_instr (cpu), 9, 5); | |
5515 | unsigned vd = uimm (aarch64_get_instr (cpu), 4, 0); | |
5516 | unsigned i; | |
5517 | ||
5518 | NYI_assert (29, 23, 0x1E); | |
5519 | NYI_assert (15, 10, 0x15); | |
5520 | ||
5521 | if (uimm (aarch64_get_instr (cpu), 22, 22)) | |
5522 | { | |
5523 | shift = uimm (aarch64_get_instr (cpu), 21, 16) - 1; | |
5524 | ||
5525 | if (full == 0) | |
5526 | HALT_UNALLOC; | |
5527 | ||
5528 | for (i = 0; i < 2; i++) | |
5529 | { | |
5530 | uint64_t val = aarch64_get_vec_u64 (cpu, vs, i); | |
5531 | aarch64_set_vec_u64 (cpu, vd, i, val << shift); | |
5532 | } | |
5533 | ||
5534 | return; | |
5535 | } | |
5536 | ||
5537 | if (uimm (aarch64_get_instr (cpu), 21, 21)) | |
5538 | { | |
5539 | shift = uimm (aarch64_get_instr (cpu), 20, 16) - 1; | |
5540 | ||
5541 | for (i = 0; i < (full ? 4 : 2); i++) | |
5542 | { | |
5543 | uint32_t val = aarch64_get_vec_u32 (cpu, vs, i); | |
5544 | aarch64_set_vec_u32 (cpu, vd, i, val << shift); | |
5545 | } | |
5546 | ||
5547 | return; | |
5548 | } | |
5549 | ||
5550 | if (uimm (aarch64_get_instr (cpu), 20, 20)) | |
5551 | { | |
5552 | shift = uimm (aarch64_get_instr (cpu), 19, 16) - 1; | |
5553 | ||
5554 | for (i = 0; i < (full ? 8 : 4); i++) | |
5555 | { | |
5556 | uint16_t val = aarch64_get_vec_u16 (cpu, vs, i); | |
5557 | aarch64_set_vec_u16 (cpu, vd, i, val << shift); | |
5558 | } | |
5559 | ||
5560 | return; | |
5561 | } | |
5562 | ||
5563 | if (uimm (aarch64_get_instr (cpu), 19, 19) == 0) | |
5564 | HALT_UNALLOC; | |
5565 | ||
5566 | shift = uimm (aarch64_get_instr (cpu), 18, 16) - 1; | |
5567 | ||
5568 | for (i = 0; i < (full ? 16 : 8); i++) | |
5569 | { | |
5570 | uint8_t val = aarch64_get_vec_u8 (cpu, vs, i); | |
5571 | aarch64_set_vec_u8 (cpu, vd, i, val << shift); | |
5572 | } | |
5573 | } | |
5574 | ||
5575 | static void | |
5576 | do_vec_SSHR_USHR (sim_cpu *cpu) | |
5577 | { | |
5578 | /* instr [31] = 0 | |
5579 | instr [30] = half(0)/full(1) | |
5580 | instr [29] = signed(0)/unsigned(1) | |
5581 | instr [28,23] = 01 1110 | |
5582 | instr [22,16] = size and shift amount | |
5583 | instr [15,10] = 0000 01 | |
5584 | instr [9, 5] = Vs | |
5585 | instr [4, 0] = Vd. */ | |
5586 | ||
5587 | int shift; | |
5588 | int full = uimm (aarch64_get_instr (cpu), 30, 30); | |
5589 | int sign = uimm (aarch64_get_instr (cpu), 29, 29); | |
5590 | unsigned vs = uimm (aarch64_get_instr (cpu), 9, 5); | |
5591 | unsigned vd = uimm (aarch64_get_instr (cpu), 4, 0); | |
5592 | unsigned i; | |
5593 | ||
5594 | NYI_assert (28, 23, 0x1E); | |
5595 | NYI_assert (15, 10, 0x01); | |
5596 | ||
5597 | if (uimm (aarch64_get_instr (cpu), 22, 22)) | |
5598 | { | |
5599 | shift = uimm (aarch64_get_instr (cpu), 21, 16); | |
5600 | ||
5601 | if (full == 0) | |
5602 | HALT_UNALLOC; | |
5603 | ||
5604 | if (sign) | |
5605 | for (i = 0; i < 2; i++) | |
5606 | { | |
5607 | int64_t val = aarch64_get_vec_s64 (cpu, vs, i); | |
5608 | aarch64_set_vec_s64 (cpu, vd, i, val >> shift); | |
5609 | } | |
5610 | else | |
5611 | for (i = 0; i < 2; i++) | |
5612 | { | |
5613 | uint64_t val = aarch64_get_vec_u64 (cpu, vs, i); | |
5614 | aarch64_set_vec_u64 (cpu, vd, i, val >> shift); | |
5615 | } | |
5616 | ||
5617 | return; | |
5618 | } | |
5619 | ||
5620 | if (uimm (aarch64_get_instr (cpu), 21, 21)) | |
5621 | { | |
5622 | shift = uimm (aarch64_get_instr (cpu), 20, 16); | |
5623 | ||
5624 | if (sign) | |
5625 | for (i = 0; i < (full ? 4 : 2); i++) | |
5626 | { | |
5627 | int32_t val = aarch64_get_vec_s32 (cpu, vs, i); | |
5628 | aarch64_set_vec_s32 (cpu, vd, i, val >> shift); | |
5629 | } | |
5630 | else | |
5631 | for (i = 0; i < (full ? 4 : 2); i++) | |
5632 | { | |
5633 | uint32_t val = aarch64_get_vec_u32 (cpu, vs, i); | |
5634 | aarch64_set_vec_u32 (cpu, vd, i, val >> shift); | |
5635 | } | |
5636 | ||
5637 | return; | |
5638 | } | |
5639 | ||
5640 | if (uimm (aarch64_get_instr (cpu), 20, 20)) | |
5641 | { | |
5642 | shift = uimm (aarch64_get_instr (cpu), 19, 16); | |
5643 | ||
5644 | if (sign) | |
5645 | for (i = 0; i < (full ? 8 : 4); i++) | |
5646 | { | |
5647 | int16_t val = aarch64_get_vec_s16 (cpu, vs, i); | |
5648 | aarch64_set_vec_s16 (cpu, vd, i, val >> shift); | |
5649 | } | |
5650 | else | |
5651 | for (i = 0; i < (full ? 8 : 4); i++) | |
5652 | { | |
5653 | uint16_t val = aarch64_get_vec_u16 (cpu, vs, i); | |
5654 | aarch64_set_vec_u16 (cpu, vd, i, val >> shift); | |
5655 | } | |
5656 | ||
5657 | return; | |
5658 | } | |
5659 | ||
5660 | if (uimm (aarch64_get_instr (cpu), 19, 19) == 0) | |
5661 | HALT_UNALLOC; | |
5662 | ||
5663 | shift = uimm (aarch64_get_instr (cpu), 18, 16); | |
5664 | ||
5665 | if (sign) | |
5666 | for (i = 0; i < (full ? 16 : 8); i++) | |
5667 | { | |
5668 | int8_t val = aarch64_get_vec_s8 (cpu, vs, i); | |
5669 | aarch64_set_vec_s8 (cpu, vd, i, val >> shift); | |
5670 | } | |
5671 | else | |
5672 | for (i = 0; i < (full ? 16 : 8); i++) | |
5673 | { | |
5674 | uint8_t val = aarch64_get_vec_u8 (cpu, vs, i); | |
5675 | aarch64_set_vec_u8 (cpu, vd, i, val >> shift); | |
5676 | } | |
5677 | } | |
5678 | ||
5679 | static void | |
5680 | do_vec_op2 (sim_cpu *cpu) | |
5681 | { | |
5682 | /* instr[31] = 0 | |
5683 | instr[30] = half/full | |
5684 | instr[29,24] = 00 1111 | |
5685 | instr[23] = ? | |
5686 | instr[22,16] = element size & index | |
5687 | instr[15,10] = sub-opcode | |
5688 | instr[9,5] = Vm | |
5689 | instr[4.0] = Vd */ | |
5690 | ||
5691 | NYI_assert (29, 24, 0x0F); | |
5692 | ||
5693 | if (uimm (aarch64_get_instr (cpu), 23, 23) != 0) | |
5694 | HALT_NYI; | |
5695 | ||
5696 | switch (uimm (aarch64_get_instr (cpu), 15, 10)) | |
5697 | { | |
5698 | case 0x01: do_vec_SSHR_USHR (cpu); return; | |
5699 | case 0x15: do_vec_SHL (cpu); return; | |
5700 | case 0x29: do_vec_xtl (cpu); return; | |
5701 | default: HALT_NYI; | |
5702 | } | |
5703 | } | |
5704 | ||
5705 | static void | |
5706 | do_vec_neg (sim_cpu *cpu) | |
5707 | { | |
5708 | /* instr[31] = 0 | |
5709 | instr[30] = full(1)/half(0) | |
5710 | instr[29,24] = 10 1110 | |
5711 | instr[23,22] = size: byte(00), half (01), word (10), long (11) | |
5712 | instr[21,10] = 1000 0010 1110 | |
5713 | instr[9,5] = Vs | |
5714 | instr[4,0] = Vd */ | |
5715 | ||
5716 | int full = uimm (aarch64_get_instr (cpu), 30, 30); | |
5717 | unsigned vs = uimm (aarch64_get_instr (cpu), 9, 5); | |
5718 | unsigned vd = uimm (aarch64_get_instr (cpu), 4, 0); | |
5719 | unsigned i; | |
5720 | ||
5721 | NYI_assert (29, 24, 0x2E); | |
5722 | NYI_assert (21, 10, 0x82E); | |
5723 | ||
5724 | switch (uimm (aarch64_get_instr (cpu), 23, 22)) | |
5725 | { | |
5726 | case 0: | |
5727 | for (i = 0; i < (full ? 16 : 8); i++) | |
5728 | aarch64_set_vec_s8 (cpu, vd, i, - aarch64_get_vec_s8 (cpu, vs, i)); | |
5729 | return; | |
5730 | ||
5731 | case 1: | |
5732 | for (i = 0; i < (full ? 8 : 4); i++) | |
5733 | aarch64_set_vec_s16 (cpu, vd, i, - aarch64_get_vec_s16 (cpu, vs, i)); | |
5734 | return; | |
5735 | ||
5736 | case 2: | |
5737 | for (i = 0; i < (full ? 4 : 2); i++) | |
5738 | aarch64_set_vec_s32 (cpu, vd, i, - aarch64_get_vec_s32 (cpu, vs, i)); | |
5739 | return; | |
5740 | ||
5741 | case 3: | |
5742 | if (! full) | |
5743 | HALT_NYI; | |
5744 | for (i = 0; i < 2; i++) | |
5745 | aarch64_set_vec_s64 (cpu, vd, i, - aarch64_get_vec_s64 (cpu, vs, i)); | |
5746 | return; | |
5747 | ||
5748 | default: | |
5749 | HALT_UNREACHABLE; | |
5750 | } | |
5751 | } | |
5752 | ||
5753 | static void | |
5754 | do_vec_sqrt (sim_cpu *cpu) | |
5755 | { | |
5756 | /* instr[31] = 0 | |
5757 | instr[30] = full(1)/half(0) | |
5758 | instr[29,23] = 101 1101 | |
5759 | instr[22] = single(0)/double(1) | |
5760 | instr[21,10] = 1000 0111 1110 | |
5761 | instr[9,5] = Vs | |
5762 | instr[4,0] = Vd. */ | |
5763 | ||
5764 | int full = uimm (aarch64_get_instr (cpu), 30, 30); | |
5765 | unsigned vs = uimm (aarch64_get_instr (cpu), 9, 5); | |
5766 | unsigned vd = uimm (aarch64_get_instr (cpu), 4, 0); | |
5767 | unsigned i; | |
5768 | ||
5769 | NYI_assert (29, 23, 0x5B); | |
5770 | NYI_assert (21, 10, 0x87E); | |
5771 | ||
5772 | if (uimm (aarch64_get_instr (cpu), 22, 22) == 0) | |
5773 | for (i = 0; i < (full ? 4 : 2); i++) | |
5774 | aarch64_set_vec_float (cpu, vd, i, | |
5775 | sqrtf (aarch64_get_vec_float (cpu, vs, i))); | |
5776 | else | |
5777 | for (i = 0; i < 2; i++) | |
5778 | aarch64_set_vec_double (cpu, vd, i, | |
5779 | sqrt (aarch64_get_vec_double (cpu, vs, i))); | |
5780 | } | |
5781 | ||
5782 | static void | |
5783 | do_vec_mls_indexed (sim_cpu *cpu) | |
5784 | { | |
5785 | /* instr[31] = 0 | |
5786 | instr[30] = half(0)/full(1) | |
5787 | instr[29,24] = 10 1111 | |
5788 | instr[23,22] = 16-bit(01)/32-bit(10) | |
5789 | instr[21,20+11] = index (if 16-bit) | |
5790 | instr[21+11] = index (if 32-bit) | |
5791 | instr[20,16] = Vm | |
5792 | instr[15,12] = 0100 | |
5793 | instr[11] = part of index | |
5794 | instr[10] = 0 | |
5795 | instr[9,5] = Vs | |
5796 | instr[4,0] = Vd. */ | |
5797 | ||
5798 | int full = uimm (aarch64_get_instr (cpu), 30, 30); | |
5799 | unsigned vs = uimm (aarch64_get_instr (cpu), 9, 5); | |
5800 | unsigned vd = uimm (aarch64_get_instr (cpu), 4, 0); | |
5801 | unsigned vm = uimm (aarch64_get_instr (cpu), 20, 16); | |
5802 | unsigned i; | |
5803 | ||
5804 | NYI_assert (15, 12, 4); | |
5805 | NYI_assert (10, 10, 0); | |
5806 | ||
5807 | switch (uimm (aarch64_get_instr (cpu), 23, 22)) | |
5808 | { | |
5809 | case 1: | |
5810 | { | |
5811 | unsigned elem; | |
5812 | uint32_t val; | |
5813 | ||
5814 | if (vm > 15) | |
5815 | HALT_NYI; | |
5816 | ||
5817 | elem = (uimm (aarch64_get_instr (cpu), 21, 20) << 1) | |
5818 | | uimm (aarch64_get_instr (cpu), 11, 11); | |
5819 | val = aarch64_get_vec_u16 (cpu, vm, elem); | |
5820 | ||
5821 | for (i = 0; i < (full ? 8 : 4); i++) | |
5822 | aarch64_set_vec_u32 (cpu, vd, i, | |
5823 | aarch64_get_vec_u32 (cpu, vd, i) - | |
5824 | (aarch64_get_vec_u32 (cpu, vs, i) * val)); | |
5825 | return; | |
5826 | } | |
5827 | ||
5828 | case 2: | |
5829 | { | |
5830 | unsigned elem = (uimm (aarch64_get_instr (cpu), 21, 21) << 1) | |
5831 | | uimm (aarch64_get_instr (cpu), 11, 11); | |
5832 | uint64_t val = aarch64_get_vec_u32 (cpu, vm, elem); | |
5833 | ||
5834 | for (i = 0; i < (full ? 4 : 2); i++) | |
5835 | aarch64_set_vec_u64 (cpu, vd, i, | |
5836 | aarch64_get_vec_u64 (cpu, vd, i) - | |
5837 | (aarch64_get_vec_u64 (cpu, vs, i) * val)); | |
5838 | return; | |
5839 | } | |
5840 | ||
5841 | case 0: | |
5842 | case 3: | |
5843 | default: | |
5844 | HALT_NYI; | |
5845 | } | |
5846 | } | |
5847 | ||
5848 | static void | |
5849 | do_vec_SUB (sim_cpu *cpu) | |
5850 | { | |
5851 | /* instr [31] = 0 | |
5852 | instr [30] = half(0)/full(1) | |
5853 | instr [29,24] = 10 1110 | |
5854 | instr [23,22] = size: byte(00, half(01), word (10), long (11) | |
5855 | instr [21] = 1 | |
5856 | instr [20,16] = Vm | |
5857 | instr [15,10] = 10 0001 | |
5858 | instr [9, 5] = Vn | |
5859 | instr [4, 0] = Vd. */ | |
5860 | ||
5861 | unsigned full = uimm (aarch64_get_instr (cpu), 30, 30); | |
5862 | unsigned vm = uimm (aarch64_get_instr (cpu), 20, 16); | |
5863 | unsigned vn = uimm (aarch64_get_instr (cpu), 9, 5); | |
5864 | unsigned vd = uimm (aarch64_get_instr (cpu), 4, 0); | |
5865 | unsigned i; | |
5866 | ||
5867 | NYI_assert (29, 24, 0x2E); | |
5868 | NYI_assert (21, 21, 1); | |
5869 | NYI_assert (15, 10, 0x21); | |
5870 | ||
5871 | switch (uimm (aarch64_get_instr (cpu), 23, 22)) | |
5872 | { | |
5873 | case 0: | |
5874 | for (i = 0; i < (full ? 16 : 8); i++) | |
5875 | aarch64_set_vec_s8 (cpu, vd, i, | |
5876 | aarch64_get_vec_s8 (cpu, vn, i) | |
5877 | - aarch64_get_vec_s8 (cpu, vm, i)); | |
5878 | return; | |
5879 | ||
5880 | case 1: | |
5881 | for (i = 0; i < (full ? 8 : 4); i++) | |
5882 | aarch64_set_vec_s16 (cpu, vd, i, | |
5883 | aarch64_get_vec_s16 (cpu, vn, i) | |
5884 | - aarch64_get_vec_s16 (cpu, vm, i)); | |
5885 | return; | |
5886 | ||
5887 | case 2: | |
5888 | for (i = 0; i < (full ? 4 : 2); i++) | |
5889 | aarch64_set_vec_s32 (cpu, vd, i, | |
5890 | aarch64_get_vec_s32 (cpu, vn, i) | |
5891 | - aarch64_get_vec_s32 (cpu, vm, i)); | |
5892 | return; | |
5893 | ||
5894 | case 3: | |
5895 | if (full == 0) | |
5896 | HALT_UNALLOC; | |
5897 | ||
5898 | for (i = 0; i < 2; i++) | |
5899 | aarch64_set_vec_s64 (cpu, vd, i, | |
5900 | aarch64_get_vec_s64 (cpu, vn, i) | |
5901 | - aarch64_get_vec_s64 (cpu, vm, i)); | |
5902 | return; | |
5903 | ||
5904 | default: | |
5905 | HALT_UNREACHABLE; | |
5906 | } | |
5907 | } | |
5908 | ||
5909 | static void | |
5910 | do_vec_MLS (sim_cpu *cpu) | |
5911 | { | |
5912 | /* instr [31] = 0 | |
5913 | instr [30] = half(0)/full(1) | |
5914 | instr [29,24] = 10 1110 | |
5915 | instr [23,22] = size: byte(00, half(01), word (10) | |
5916 | instr [21] = 1 | |
5917 | instr [20,16] = Vm | |
5918 | instr [15,10] = 10 0101 | |
5919 | instr [9, 5] = Vn | |
5920 | instr [4, 0] = Vd. */ | |
5921 | ||
5922 | unsigned full = uimm (aarch64_get_instr (cpu), 30, 30); | |
5923 | unsigned vm = uimm (aarch64_get_instr (cpu), 20, 16); | |
5924 | unsigned vn = uimm (aarch64_get_instr (cpu), 9, 5); | |
5925 | unsigned vd = uimm (aarch64_get_instr (cpu), 4, 0); | |
5926 | unsigned i; | |
5927 | ||
5928 | NYI_assert (29, 24, 0x2E); | |
5929 | NYI_assert (21, 21, 1); | |
5930 | NYI_assert (15, 10, 0x25); | |
5931 | ||
5932 | switch (uimm (aarch64_get_instr (cpu), 23, 22)) | |
5933 | { | |
5934 | case 0: | |
5935 | for (i = 0; i < (full ? 16 : 8); i++) | |
5936 | aarch64_set_vec_u8 (cpu, vd, i, | |
5937 | (aarch64_get_vec_u8 (cpu, vn, i) | |
5938 | * aarch64_get_vec_u8 (cpu, vm, i)) | |
5939 | - aarch64_get_vec_u8 (cpu, vd, i)); | |
5940 | return; | |
5941 | ||
5942 | case 1: | |
5943 | for (i = 0; i < (full ? 8 : 4); i++) | |
5944 | aarch64_set_vec_u16 (cpu, vd, i, | |
5945 | (aarch64_get_vec_u16 (cpu, vn, i) | |
5946 | * aarch64_get_vec_u16 (cpu, vm, i)) | |
5947 | - aarch64_get_vec_u16 (cpu, vd, i)); | |
5948 | return; | |
5949 | ||
5950 | case 2: | |
5951 | for (i = 0; i < (full ? 4 : 2); i++) | |
5952 | aarch64_set_vec_u32 (cpu, vd, i, | |
5953 | (aarch64_get_vec_u32 (cpu, vn, i) | |
5954 | * aarch64_get_vec_u32 (cpu, vm, i)) | |
5955 | - aarch64_get_vec_u32 (cpu, vd, i)); | |
5956 | return; | |
5957 | ||
5958 | default: | |
5959 | HALT_UNALLOC; | |
5960 | } | |
5961 | } | |
5962 | ||
5963 | static void | |
5964 | do_vec_FDIV (sim_cpu *cpu) | |
5965 | { | |
5966 | /* instr [31] = 0 | |
5967 | instr [30] = half(0)/full(1) | |
5968 | instr [29,23] = 10 1110 0 | |
5969 | instr [22] = float()/double(1) | |
5970 | instr [21] = 1 | |
5971 | instr [20,16] = Vm | |
5972 | instr [15,10] = 1111 11 | |
5973 | instr [9, 5] = Vn | |
5974 | instr [4, 0] = Vd. */ | |
5975 | ||
5976 | unsigned full = uimm (aarch64_get_instr (cpu), 30, 30); | |
5977 | unsigned vm = uimm (aarch64_get_instr (cpu), 20, 16); | |
5978 | unsigned vn = uimm (aarch64_get_instr (cpu), 9, 5); | |
5979 | unsigned vd = uimm (aarch64_get_instr (cpu), 4, 0); | |
5980 | unsigned i; | |
5981 | ||
5982 | NYI_assert (29, 23, 0x5C); | |
5983 | NYI_assert (21, 21, 1); | |
5984 | NYI_assert (15, 10, 0x3F); | |
5985 | ||
5986 | if (uimm (aarch64_get_instr (cpu), 22, 22)) | |
5987 | { | |
5988 | if (! full) | |
5989 | HALT_UNALLOC; | |
5990 | ||
5991 | for (i = 0; i < 2; i++) | |
5992 | aarch64_set_vec_double (cpu, vd, i, | |
5993 | aarch64_get_vec_double (cpu, vn, i) | |
5994 | / aarch64_get_vec_double (cpu, vm, i)); | |
5995 | } | |
5996 | else | |
5997 | for (i = 0; i < (full ? 4 : 2); i++) | |
5998 | aarch64_set_vec_float (cpu, vd, i, | |
5999 | aarch64_get_vec_float (cpu, vn, i) | |
6000 | / aarch64_get_vec_float (cpu, vm, i)); | |
6001 | } | |
6002 | ||
6003 | static void | |
6004 | do_vec_FMUL (sim_cpu *cpu) | |
6005 | { | |
6006 | /* instr [31] = 0 | |
6007 | instr [30] = half(0)/full(1) | |
6008 | instr [29,23] = 10 1110 0 | |
6009 | instr [22] = float(0)/double(1) | |
6010 | instr [21] = 1 | |
6011 | instr [20,16] = Vm | |
6012 | instr [15,10] = 1101 11 | |
6013 | instr [9, 5] = Vn | |
6014 | instr [4, 0] = Vd. */ | |
6015 | ||
6016 | unsigned full = uimm (aarch64_get_instr (cpu), 30, 30); | |
6017 | unsigned vm = uimm (aarch64_get_instr (cpu), 20, 16); | |
6018 | unsigned vn = uimm (aarch64_get_instr (cpu), 9, 5); | |
6019 | unsigned vd = uimm (aarch64_get_instr (cpu), 4, 0); | |
6020 | unsigned i; | |
6021 | ||
6022 | NYI_assert (29, 23, 0x5C); | |
6023 | NYI_assert (21, 21, 1); | |
6024 | NYI_assert (15, 10, 0x37); | |
6025 | ||
6026 | if (uimm (aarch64_get_instr (cpu), 22, 22)) | |
6027 | { | |
6028 | if (! full) | |
6029 | HALT_UNALLOC; | |
6030 | ||
6031 | for (i = 0; i < 2; i++) | |
6032 | aarch64_set_vec_double (cpu, vd, i, | |
6033 | aarch64_get_vec_double (cpu, vn, i) | |
6034 | * aarch64_get_vec_double (cpu, vm, i)); | |
6035 | } | |
6036 | else | |
6037 | for (i = 0; i < (full ? 4 : 2); i++) | |
6038 | aarch64_set_vec_float (cpu, vd, i, | |
6039 | aarch64_get_vec_float (cpu, vn, i) | |
6040 | * aarch64_get_vec_float (cpu, vm, i)); | |
6041 | } | |
6042 | ||
6043 | static void | |
6044 | do_vec_FADDP (sim_cpu *cpu) | |
6045 | { | |
6046 | /* instr [31] = 0 | |
6047 | instr [30] = half(0)/full(1) | |
6048 | instr [29,23] = 10 1110 0 | |
6049 | instr [22] = float(0)/double(1) | |
6050 | instr [21] = 1 | |
6051 | instr [20,16] = Vm | |
6052 | instr [15,10] = 1101 01 | |
6053 | instr [9, 5] = Vn | |
6054 | instr [4, 0] = Vd. */ | |
6055 | ||
6056 | unsigned full = uimm (aarch64_get_instr (cpu), 30, 30); | |
6057 | unsigned vm = uimm (aarch64_get_instr (cpu), 20, 16); | |
6058 | unsigned vn = uimm (aarch64_get_instr (cpu), 9, 5); | |
6059 | unsigned vd = uimm (aarch64_get_instr (cpu), 4, 0); | |
6060 | ||
6061 | NYI_assert (29, 23, 0x5C); | |
6062 | NYI_assert (21, 21, 1); | |
6063 | NYI_assert (15, 10, 0x35); | |
6064 | ||
6065 | if (uimm (aarch64_get_instr (cpu), 22, 22)) | |
6066 | { | |
6067 | if (! full) | |
6068 | HALT_UNALLOC; | |
6069 | ||
6070 | aarch64_set_vec_double (cpu, vd, 0, aarch64_get_vec_double (cpu, vn, 0) | |
6071 | + aarch64_get_vec_double (cpu, vn, 1)); | |
6072 | aarch64_set_vec_double (cpu, vd, 1, aarch64_get_vec_double (cpu, vm, 0) | |
6073 | + aarch64_get_vec_double (cpu, vm, 1)); | |
6074 | } | |
6075 | else | |
6076 | { | |
6077 | aarch64_set_vec_float (cpu, vd, 0, aarch64_get_vec_float (cpu, vn, 0) | |
6078 | + aarch64_get_vec_float (cpu, vn, 1)); | |
6079 | if (full) | |
6080 | aarch64_set_vec_float (cpu, vd, 1, aarch64_get_vec_float (cpu, vn, 2) | |
6081 | + aarch64_get_vec_float (cpu, vn, 3)); | |
6082 | aarch64_set_vec_float (cpu, vd, full ? 2 : 1, | |
6083 | aarch64_get_vec_float (cpu, vm, 0) | |
6084 | + aarch64_get_vec_float (cpu, vm, 1)); | |
6085 | if (full) | |
6086 | aarch64_set_vec_float (cpu, vd, 3, | |
6087 | aarch64_get_vec_float (cpu, vm, 2) | |
6088 | + aarch64_get_vec_float (cpu, vm, 3)); | |
6089 | } | |
6090 | } | |
6091 | ||
6092 | static void | |
6093 | do_vec_FSQRT (sim_cpu *cpu) | |
6094 | { | |
6095 | /* instr[31] = 0 | |
6096 | instr[30] = half(0)/full(1) | |
6097 | instr[29,23] = 10 1110 1 | |
6098 | instr[22] = single(0)/double(1) | |
6099 | instr[21,10] = 10 0001 1111 10 | |
6100 | instr[9,5] = Vsrc | |
6101 | instr[4,0] = Vdest. */ | |
6102 | ||
6103 | unsigned vn = uimm (aarch64_get_instr (cpu), 9, 5); | |
6104 | unsigned vd = uimm (aarch64_get_instr (cpu), 4, 0); | |
6105 | unsigned full = uimm (aarch64_get_instr (cpu), 30, 30); | |
6106 | int i; | |
6107 | ||
6108 | NYI_assert (29, 23, 0x5D); | |
6109 | NYI_assert (21, 10, 0x87E); | |
6110 | ||
6111 | if (uimm (aarch64_get_instr (cpu), 22, 22)) | |
6112 | { | |
6113 | if (! full) | |
6114 | HALT_UNALLOC; | |
6115 | ||
6116 | for (i = 0; i < 2; i++) | |
6117 | aarch64_set_vec_double (cpu, vd, i, | |
6118 | sqrt (aarch64_get_vec_double (cpu, vn, i))); | |
6119 | } | |
6120 | else | |
6121 | { | |
6122 | for (i = 0; i < (full ? 4 : 2); i++) | |
6123 | aarch64_set_vec_float (cpu, vd, i, | |
6124 | sqrtf (aarch64_get_vec_float (cpu, vn, i))); | |
6125 | } | |
6126 | } | |
6127 | ||
6128 | static void | |
6129 | do_vec_FNEG (sim_cpu *cpu) | |
6130 | { | |
6131 | /* instr[31] = 0 | |
6132 | instr[30] = half (0)/full (1) | |
6133 | instr[29,23] = 10 1110 1 | |
6134 | instr[22] = single (0)/double (1) | |
6135 | instr[21,10] = 10 0000 1111 10 | |
6136 | instr[9,5] = Vsrc | |
6137 | instr[4,0] = Vdest. */ | |
6138 | ||
6139 | unsigned vn = uimm (aarch64_get_instr (cpu), 9, 5); | |
6140 | unsigned vd = uimm (aarch64_get_instr (cpu), 4, 0); | |
6141 | unsigned full = uimm (aarch64_get_instr (cpu), 30, 30); | |
6142 | int i; | |
6143 | ||
6144 | NYI_assert (29, 23, 0x5D); | |
6145 | NYI_assert (21, 10, 0x83E); | |
6146 | ||
6147 | if (uimm (aarch64_get_instr (cpu), 22, 22)) | |
6148 | { | |
6149 | if (! full) | |
6150 | HALT_UNALLOC; | |
6151 | ||
6152 | for (i = 0; i < 2; i++) | |
6153 | aarch64_set_vec_double (cpu, vd, i, | |
6154 | - aarch64_get_vec_double (cpu, vn, i)); | |
6155 | } | |
6156 | else | |
6157 | { | |
6158 | for (i = 0; i < (full ? 4 : 2); i++) | |
6159 | aarch64_set_vec_float (cpu, vd, i, | |
6160 | - aarch64_get_vec_float (cpu, vn, i)); | |
6161 | } | |
6162 | } | |
6163 | ||
6164 | static void | |
6165 | do_vec_NOT (sim_cpu *cpu) | |
6166 | { | |
6167 | /* instr[31] = 0 | |
6168 | instr[30] = half (0)/full (1) | |
6169 | instr[29,21] = 10 1110 001 | |
6170 | instr[20,16] = 0 0000 | |
6171 | instr[15,10] = 0101 10 | |
6172 | instr[9,5] = Vn | |
6173 | instr[4.0] = Vd. */ | |
6174 | ||
6175 | unsigned vn = uimm (aarch64_get_instr (cpu), 9, 5); | |
6176 | unsigned vd = uimm (aarch64_get_instr (cpu), 4, 0); | |
6177 | unsigned i; | |
6178 | int full = uimm (aarch64_get_instr (cpu), 30, 30); | |
6179 | ||
6180 | NYI_assert (29, 10, 0xB8816); | |
6181 | ||
6182 | for (i = 0; i < (full ? 16 : 8); i++) | |
6183 | aarch64_set_vec_u8 (cpu, vd, i, ~ aarch64_get_vec_u8 (cpu, vn, i)); | |
6184 | } | |
6185 | ||
6186 | static void | |
6187 | do_vec_MOV_element (sim_cpu *cpu) | |
6188 | { | |
6189 | /* instr[31,21] = 0110 1110 000 | |
6190 | instr[20,16] = size & dest index | |
6191 | instr[15] = 0 | |
6192 | instr[14,11] = source index | |
6193 | instr[10] = 1 | |
6194 | instr[9,5] = Vs | |
6195 | instr[4.0] = Vd. */ | |
6196 | ||
6197 | unsigned vs = uimm (aarch64_get_instr (cpu), 9, 5); | |
6198 | unsigned vd = uimm (aarch64_get_instr (cpu), 4, 0); | |
6199 | unsigned src_index; | |
6200 | unsigned dst_index; | |
6201 | ||
6202 | NYI_assert (31, 21, 0x370); | |
6203 | NYI_assert (15, 15, 0); | |
6204 | NYI_assert (10, 10, 1); | |
6205 | ||
6206 | if (uimm (aarch64_get_instr (cpu), 16, 16)) | |
6207 | { | |
6208 | /* Move a byte. */ | |
6209 | src_index = uimm (aarch64_get_instr (cpu), 14, 11); | |
6210 | dst_index = uimm (aarch64_get_instr (cpu), 20, 17); | |
6211 | aarch64_set_vec_u8 (cpu, vd, dst_index, | |
6212 | aarch64_get_vec_u8 (cpu, vs, src_index)); | |
6213 | } | |
6214 | else if (uimm (aarch64_get_instr (cpu), 17, 17)) | |
6215 | { | |
6216 | /* Move 16-bits. */ | |
6217 | NYI_assert (11, 11, 0); | |
6218 | src_index = uimm (aarch64_get_instr (cpu), 14, 12); | |
6219 | dst_index = uimm (aarch64_get_instr (cpu), 20, 18); | |
6220 | aarch64_set_vec_u16 (cpu, vd, dst_index, | |
6221 | aarch64_get_vec_u16 (cpu, vs, src_index)); | |
6222 | } | |
6223 | else if (uimm (aarch64_get_instr (cpu), 18, 18)) | |
6224 | { | |
6225 | /* Move 32-bits. */ | |
6226 | NYI_assert (12, 11, 0); | |
6227 | src_index = uimm (aarch64_get_instr (cpu), 14, 13); | |
6228 | dst_index = uimm (aarch64_get_instr (cpu), 20, 19); | |
6229 | aarch64_set_vec_u32 (cpu, vd, dst_index, | |
6230 | aarch64_get_vec_u32 (cpu, vs, src_index)); | |
6231 | } | |
6232 | else | |
6233 | { | |
6234 | NYI_assert (19, 19, 1); | |
6235 | NYI_assert (13, 11, 0); | |
6236 | src_index = uimm (aarch64_get_instr (cpu), 14, 14); | |
6237 | dst_index = uimm (aarch64_get_instr (cpu), 20, 20); | |
6238 | aarch64_set_vec_u64 (cpu, vd, dst_index, | |
6239 | aarch64_get_vec_u64 (cpu, vs, src_index)); | |
6240 | } | |
6241 | } | |
6242 | ||
6243 | static void | |
6244 | dexAdvSIMD0 (sim_cpu *cpu) | |
6245 | { | |
6246 | /* instr [28,25] = 0 111. */ | |
6247 | if ( uimm (aarch64_get_instr (cpu), 15, 10) == 0x07 | |
6248 | && (uimm (aarch64_get_instr (cpu), 9, 5) == | |
6249 | uimm (aarch64_get_instr (cpu), 20, 16))) | |
6250 | { | |
6251 | if (uimm (aarch64_get_instr (cpu), 31, 21) == 0x075 | |
6252 | || uimm (aarch64_get_instr (cpu), 31, 21) == 0x275) | |
6253 | { | |
6254 | do_vec_MOV_whole_vector (cpu); | |
6255 | return; | |
6256 | } | |
6257 | } | |
6258 | ||
6259 | if (uimm (aarch64_get_instr (cpu), 29, 19) == 0x1E0) | |
6260 | { | |
6261 | do_vec_MOV_immediate (cpu); | |
6262 | return; | |
6263 | } | |
6264 | ||
6265 | if (uimm (aarch64_get_instr (cpu), 29, 19) == 0x5E0) | |
6266 | { | |
6267 | do_vec_MVNI (cpu); | |
6268 | return; | |
6269 | } | |
6270 | ||
6271 | if (uimm (aarch64_get_instr (cpu), 29, 19) == 0x1C0 | |
6272 | || uimm (aarch64_get_instr (cpu), 29, 19) == 0x1C1) | |
6273 | { | |
6274 | if (uimm (aarch64_get_instr (cpu), 15, 10) == 0x03) | |
6275 | { | |
6276 | do_vec_DUP_scalar_into_vector (cpu); | |
6277 | return; | |
6278 | } | |
6279 | } | |
6280 | ||
6281 | switch (uimm (aarch64_get_instr (cpu), 29, 24)) | |
6282 | { | |
6283 | case 0x0E: do_vec_op1 (cpu); return; | |
6284 | case 0x0F: do_vec_op2 (cpu); return; | |
6285 | ||
6286 | case 0x2f: | |
6287 | switch (uimm (aarch64_get_instr (cpu), 15, 10)) | |
6288 | { | |
6289 | case 0x01: do_vec_SSHR_USHR (cpu); return; | |
6290 | case 0x10: | |
6291 | case 0x12: do_vec_mls_indexed (cpu); return; | |
6292 | case 0x29: do_vec_xtl (cpu); return; | |
6293 | default: | |
6294 | HALT_NYI; | |
6295 | } | |
6296 | ||
6297 | case 0x2E: | |
6298 | if (uimm (aarch64_get_instr (cpu), 21, 21) == 1) | |
6299 | { | |
6300 | switch (uimm (aarch64_get_instr (cpu), 15, 10)) | |
6301 | { | |
6302 | case 0x07: | |
6303 | switch (uimm (aarch64_get_instr (cpu), 23, 22)) | |
6304 | { | |
6305 | case 0: do_vec_EOR (cpu); return; | |
6306 | case 1: do_vec_BSL (cpu); return; | |
6307 | case 2: | |
6308 | case 3: do_vec_bit (cpu); return; | |
6309 | } | |
6310 | break; | |
6311 | ||
6312 | case 0x08: do_vec_sub_long (cpu); return; | |
6313 | case 0x11: do_vec_USHL (cpu); return; | |
6314 | case 0x16: do_vec_NOT (cpu); return; | |
6315 | case 0x19: do_vec_max (cpu); return; | |
6316 | case 0x1B: do_vec_min (cpu); return; | |
6317 | case 0x21: do_vec_SUB (cpu); return; | |
6318 | case 0x25: do_vec_MLS (cpu); return; | |
6319 | case 0x31: do_vec_FminmaxNMP (cpu); return; | |
6320 | case 0x35: do_vec_FADDP (cpu); return; | |
6321 | case 0x37: do_vec_FMUL (cpu); return; | |
6322 | case 0x3F: do_vec_FDIV (cpu); return; | |
6323 | ||
6324 | case 0x3E: | |
6325 | switch (uimm (aarch64_get_instr (cpu), 20, 16)) | |
6326 | { | |
6327 | case 0x00: do_vec_FNEG (cpu); return; | |
6328 | case 0x01: do_vec_FSQRT (cpu); return; | |
6329 | default: HALT_NYI; | |
6330 | } | |
6331 | ||
6332 | case 0x0D: | |
6333 | case 0x0F: | |
6334 | case 0x22: | |
6335 | case 0x23: | |
6336 | case 0x26: | |
6337 | case 0x2A: | |
6338 | case 0x32: | |
6339 | case 0x36: | |
6340 | case 0x39: | |
6341 | case 0x3A: | |
6342 | do_vec_compare (cpu); return; | |
6343 | ||
6344 | default: break; | |
6345 | } | |
6346 | } | |
6347 | ||
6348 | if (uimm (aarch64_get_instr (cpu), 31, 21) == 0x370) | |
6349 | { | |
6350 | do_vec_MOV_element (cpu); | |
6351 | return; | |
6352 | } | |
6353 | ||
6354 | switch (uimm (aarch64_get_instr (cpu), 21, 10)) | |
6355 | { | |
6356 | case 0x82E: do_vec_neg (cpu); return; | |
6357 | case 0x87E: do_vec_sqrt (cpu); return; | |
6358 | default: | |
6359 | if (uimm (aarch64_get_instr (cpu), 15, 10) == 0x30) | |
6360 | { | |
6361 | do_vec_mull (cpu); | |
6362 | return; | |
6363 | } | |
6364 | break; | |
6365 | } | |
6366 | break; | |
6367 | ||
6368 | default: | |
6369 | break; | |
6370 | } | |
6371 | ||
6372 | HALT_NYI; | |
6373 | } | |
6374 | ||
6375 | /* 3 sources. */ | |
6376 | ||
6377 | /* Float multiply add. */ | |
6378 | static void | |
6379 | fmadds (sim_cpu *cpu) | |
6380 | { | |
6381 | unsigned sa = uimm (aarch64_get_instr (cpu), 14, 10); | |
6382 | unsigned sm = uimm (aarch64_get_instr (cpu), 20, 16); | |
6383 | unsigned sn = uimm (aarch64_get_instr (cpu), 9, 5); | |
6384 | unsigned sd = uimm (aarch64_get_instr (cpu), 4, 0); | |
6385 | ||
6386 | aarch64_set_FP_float (cpu, sd, aarch64_get_FP_float (cpu, sa) | |
6387 | + aarch64_get_FP_float (cpu, sn) | |
6388 | * aarch64_get_FP_float (cpu, sm)); | |
6389 | } | |
6390 | ||
6391 | /* Double multiply add. */ | |
6392 | static void | |
6393 | fmaddd (sim_cpu *cpu) | |
6394 | { | |
6395 | unsigned sa = uimm (aarch64_get_instr (cpu), 14, 10); | |
6396 | unsigned sm = uimm (aarch64_get_instr (cpu), 20, 16); | |
6397 | unsigned sn = uimm (aarch64_get_instr (cpu), 9, 5); | |
6398 | unsigned sd = uimm (aarch64_get_instr (cpu), 4, 0); | |
6399 | ||
6400 | aarch64_set_FP_double (cpu, sd, aarch64_get_FP_double (cpu, sa) | |
6401 | + aarch64_get_FP_double (cpu, sn) | |
6402 | * aarch64_get_FP_double (cpu, sm)); | |
6403 | } | |
6404 | ||
6405 | /* Float multiply subtract. */ | |
6406 | static void | |
6407 | fmsubs (sim_cpu *cpu) | |
6408 | { | |
6409 | unsigned sa = uimm (aarch64_get_instr (cpu), 14, 10); | |
6410 | unsigned sm = uimm (aarch64_get_instr (cpu), 20, 16); | |
6411 | unsigned sn = uimm (aarch64_get_instr (cpu), 9, 5); | |
6412 | unsigned sd = uimm (aarch64_get_instr (cpu), 4, 0); | |
6413 | ||
6414 | aarch64_set_FP_float (cpu, sd, aarch64_get_FP_float (cpu, sa) | |
6415 | - aarch64_get_FP_float (cpu, sn) | |
6416 | * aarch64_get_FP_float (cpu, sm)); | |
6417 | } | |
6418 | ||
6419 | /* Double multiply subtract. */ | |
6420 | static void | |
6421 | fmsubd (sim_cpu *cpu) | |
6422 | { | |
6423 | unsigned sa = uimm (aarch64_get_instr (cpu), 14, 10); | |
6424 | unsigned sm = uimm (aarch64_get_instr (cpu), 20, 16); | |
6425 | unsigned sn = uimm (aarch64_get_instr (cpu), 9, 5); | |
6426 | unsigned sd = uimm (aarch64_get_instr (cpu), 4, 0); | |
6427 | ||
6428 | aarch64_set_FP_double (cpu, sd, aarch64_get_FP_double (cpu, sa) | |
6429 | - aarch64_get_FP_double (cpu, sn) | |
6430 | * aarch64_get_FP_double (cpu, sm)); | |
6431 | } | |
6432 | ||
6433 | /* Float negative multiply add. */ | |
6434 | static void | |
6435 | fnmadds (sim_cpu *cpu) | |
6436 | { | |
6437 | unsigned sa = uimm (aarch64_get_instr (cpu), 14, 10); | |
6438 | unsigned sm = uimm (aarch64_get_instr (cpu), 20, 16); | |
6439 | unsigned sn = uimm (aarch64_get_instr (cpu), 9, 5); | |
6440 | unsigned sd = uimm (aarch64_get_instr (cpu), 4, 0); | |
6441 | ||
6442 | aarch64_set_FP_float (cpu, sd, - aarch64_get_FP_float (cpu, sa) | |
6443 | + (- aarch64_get_FP_float (cpu, sn)) | |
6444 | * aarch64_get_FP_float (cpu, sm)); | |
6445 | } | |
6446 | ||
6447 | /* Double negative multiply add. */ | |
6448 | static void | |
6449 | fnmaddd (sim_cpu *cpu) | |
6450 | { | |
6451 | unsigned sa = uimm (aarch64_get_instr (cpu), 14, 10); | |
6452 | unsigned sm = uimm (aarch64_get_instr (cpu), 20, 16); | |
6453 | unsigned sn = uimm (aarch64_get_instr (cpu), 9, 5); | |
6454 | unsigned sd = uimm (aarch64_get_instr (cpu), 4, 0); | |
6455 | ||
6456 | aarch64_set_FP_double (cpu, sd, - aarch64_get_FP_double (cpu, sa) | |
6457 | + (- aarch64_get_FP_double (cpu, sn)) | |
6458 | * aarch64_get_FP_double (cpu, sm)); | |
6459 | } | |
6460 | ||
6461 | /* Float negative multiply subtract. */ | |
6462 | static void | |
6463 | fnmsubs (sim_cpu *cpu) | |
6464 | { | |
6465 | unsigned sa = uimm (aarch64_get_instr (cpu), 14, 10); | |
6466 | unsigned sm = uimm (aarch64_get_instr (cpu), 20, 16); | |
6467 | unsigned sn = uimm (aarch64_get_instr (cpu), 9, 5); | |
6468 | unsigned sd = uimm (aarch64_get_instr (cpu), 4, 0); | |
6469 | ||
6470 | aarch64_set_FP_float (cpu, sd, - aarch64_get_FP_float (cpu, sa) | |
6471 | + aarch64_get_FP_float (cpu, sn) | |
6472 | * aarch64_get_FP_float (cpu, sm)); | |
6473 | } | |
6474 | ||
6475 | /* Double negative multiply subtract. */ | |
6476 | static void | |
6477 | fnmsubd (sim_cpu *cpu) | |
6478 | { | |
6479 | unsigned sa = uimm (aarch64_get_instr (cpu), 14, 10); | |
6480 | unsigned sm = uimm (aarch64_get_instr (cpu), 20, 16); | |
6481 | unsigned sn = uimm (aarch64_get_instr (cpu), 9, 5); | |
6482 | unsigned sd = uimm (aarch64_get_instr (cpu), 4, 0); | |
6483 | ||
6484 | aarch64_set_FP_double (cpu, sd, - aarch64_get_FP_double (cpu, sa) | |
6485 | + aarch64_get_FP_double (cpu, sn) | |
6486 | * aarch64_get_FP_double (cpu, sm)); | |
6487 | } | |
6488 | ||
6489 | static void | |
6490 | dexSimpleFPDataProc3Source (sim_cpu *cpu) | |
6491 | { | |
6492 | /* instr[31] ==> M : 0 ==> OK, 1 ==> UNALLOC | |
6493 | instr[30] = 0 | |
6494 | instr[29] ==> S : 0 ==> OK, 1 ==> UNALLOC | |
6495 | instr[28,25] = 1111 | |
6496 | instr[24] = 1 | |
6497 | instr[23,22] ==> type : 0 ==> single, 01 ==> double, 1x ==> UNALLOC | |
6498 | instr[21] ==> o1 : 0 ==> unnegated, 1 ==> negated | |
6499 | instr[15] ==> o2 : 0 ==> ADD, 1 ==> SUB */ | |
6500 | ||
6501 | uint32_t M_S = (uimm (aarch64_get_instr (cpu), 31, 31) << 1) | |
6502 | | uimm (aarch64_get_instr (cpu), 29, 29); | |
6503 | /* dispatch on combined type:o1:o2. */ | |
6504 | uint32_t dispatch = (uimm (aarch64_get_instr (cpu), 23, 21) << 1) | |
6505 | | uimm (aarch64_get_instr (cpu), 15, 15); | |
6506 | ||
6507 | if (M_S != 0) | |
6508 | HALT_UNALLOC; | |
6509 | ||
6510 | switch (dispatch) | |
6511 | { | |
6512 | case 0: fmadds (cpu); return; | |
6513 | case 1: fmsubs (cpu); return; | |
6514 | case 2: fnmadds (cpu); return; | |
6515 | case 3: fnmsubs (cpu); return; | |
6516 | case 4: fmaddd (cpu); return; | |
6517 | case 5: fmsubd (cpu); return; | |
6518 | case 6: fnmaddd (cpu); return; | |
6519 | case 7: fnmsubd (cpu); return; | |
6520 | default: | |
6521 | /* type > 1 is currently unallocated. */ | |
6522 | HALT_UNALLOC; | |
6523 | } | |
6524 | } | |
6525 | ||
6526 | static void | |
6527 | dexSimpleFPFixedConvert (sim_cpu *cpu) | |
6528 | { | |
6529 | HALT_NYI; | |
6530 | } | |
6531 | ||
6532 | static void | |
6533 | dexSimpleFPCondCompare (sim_cpu *cpu) | |
6534 | { | |
6535 | HALT_NYI; | |
6536 | } | |
6537 | ||
6538 | /* 2 sources. */ | |
6539 | ||
6540 | /* Float add. */ | |
6541 | static void | |
6542 | fadds (sim_cpu *cpu) | |
6543 | { | |
6544 | unsigned sm = uimm (aarch64_get_instr (cpu), 20, 16); | |
6545 | unsigned sn = uimm (aarch64_get_instr (cpu), 9, 5); | |
6546 | unsigned sd = uimm (aarch64_get_instr (cpu), 4, 0); | |
6547 | ||
6548 | aarch64_set_FP_float (cpu, sd, aarch64_get_FP_float (cpu, sn) | |
6549 | + aarch64_get_FP_float (cpu, sm)); | |
6550 | } | |
6551 | ||
6552 | /* Double add. */ | |
6553 | static void | |
6554 | faddd (sim_cpu *cpu) | |
6555 | { | |
6556 | unsigned sm = uimm (aarch64_get_instr (cpu), 20, 16); | |
6557 | unsigned sn = uimm (aarch64_get_instr (cpu), 9, 5); | |
6558 | unsigned sd = uimm (aarch64_get_instr (cpu), 4, 0); | |
6559 | ||
6560 | aarch64_set_FP_double (cpu, sd, aarch64_get_FP_double (cpu, sn) | |
6561 | + aarch64_get_FP_double (cpu, sm)); | |
6562 | } | |
6563 | ||
6564 | /* Float divide. */ | |
6565 | static void | |
6566 | fdivs (sim_cpu *cpu) | |
6567 | { | |
6568 | unsigned sm = uimm (aarch64_get_instr (cpu), 20, 16); | |
6569 | unsigned sn = uimm (aarch64_get_instr (cpu), 9, 5); | |
6570 | unsigned sd = uimm (aarch64_get_instr (cpu), 4, 0); | |
6571 | ||
6572 | aarch64_set_FP_float (cpu, sd, aarch64_get_FP_float (cpu, sn) | |
6573 | / aarch64_get_FP_float (cpu, sm)); | |
6574 | } | |
6575 | ||
6576 | /* Double divide. */ | |
6577 | static void | |
6578 | fdivd (sim_cpu *cpu) | |
6579 | { | |
6580 | unsigned sm = uimm (aarch64_get_instr (cpu), 20, 16); | |
6581 | unsigned sn = uimm (aarch64_get_instr (cpu), 9, 5); | |
6582 | unsigned sd = uimm (aarch64_get_instr (cpu), 4, 0); | |
6583 | ||
6584 | aarch64_set_FP_double (cpu, sd, aarch64_get_FP_double (cpu, sn) | |
6585 | / aarch64_get_FP_double (cpu, sm)); | |
6586 | } | |
6587 | ||
6588 | /* Float multiply. */ | |
6589 | static void | |
6590 | fmuls (sim_cpu *cpu) | |
6591 | { | |
6592 | unsigned sm = uimm (aarch64_get_instr (cpu), 20, 16); | |
6593 | unsigned sn = uimm (aarch64_get_instr (cpu), 9, 5); | |
6594 | unsigned sd = uimm (aarch64_get_instr (cpu), 4, 0); | |
6595 | ||
6596 | aarch64_set_FP_float (cpu, sd, aarch64_get_FP_float (cpu, sn) | |
6597 | * aarch64_get_FP_float (cpu, sm)); | |
6598 | } | |
6599 | ||
6600 | /* Double multiply. */ | |
6601 | static void | |
6602 | fmuld (sim_cpu *cpu) | |
6603 | { | |
6604 | unsigned sm = uimm (aarch64_get_instr (cpu), 20, 16); | |
6605 | unsigned sn = uimm (aarch64_get_instr (cpu), 9, 5); | |
6606 | unsigned sd = uimm (aarch64_get_instr (cpu), 4, 0); | |
6607 | ||
6608 | aarch64_set_FP_double (cpu, sd, aarch64_get_FP_double (cpu, sn) | |
6609 | * aarch64_get_FP_double (cpu, sm)); | |
6610 | } | |
6611 | ||
6612 | /* Float negate and multiply. */ | |
6613 | static void | |
6614 | fnmuls (sim_cpu *cpu) | |
6615 | { | |
6616 | unsigned sm = uimm (aarch64_get_instr (cpu), 20, 16); | |
6617 | unsigned sn = uimm (aarch64_get_instr (cpu), 9, 5); | |
6618 | unsigned sd = uimm (aarch64_get_instr (cpu), 4, 0); | |
6619 | ||
6620 | aarch64_set_FP_float (cpu, sd, - (aarch64_get_FP_float (cpu, sn) | |
6621 | * aarch64_get_FP_float (cpu, sm))); | |
6622 | } | |
6623 | ||
6624 | /* Double negate and multiply. */ | |
6625 | static void | |
6626 | fnmuld (sim_cpu *cpu) | |
6627 | { | |
6628 | unsigned sm = uimm (aarch64_get_instr (cpu), 20, 16); | |
6629 | unsigned sn = uimm (aarch64_get_instr (cpu), 9, 5); | |
6630 | unsigned sd = uimm (aarch64_get_instr (cpu), 4, 0); | |
6631 | ||
6632 | aarch64_set_FP_double (cpu, sd, - (aarch64_get_FP_double (cpu, sn) | |
6633 | * aarch64_get_FP_double (cpu, sm))); | |
6634 | } | |
6635 | ||
6636 | /* Float subtract. */ | |
6637 | static void | |
6638 | fsubs (sim_cpu *cpu) | |
6639 | { | |
6640 | unsigned sm = uimm (aarch64_get_instr (cpu), 20, 16); | |
6641 | unsigned sn = uimm (aarch64_get_instr (cpu), 9, 5); | |
6642 | unsigned sd = uimm (aarch64_get_instr (cpu), 4, 0); | |
6643 | ||
6644 | aarch64_set_FP_float (cpu, sd, aarch64_get_FP_float (cpu, sn) | |
6645 | - aarch64_get_FP_float (cpu, sm)); | |
6646 | } | |
6647 | ||
6648 | /* Double subtract. */ | |
6649 | static void | |
6650 | fsubd (sim_cpu *cpu) | |
6651 | { | |
6652 | unsigned sm = uimm (aarch64_get_instr (cpu), 20, 16); | |
6653 | unsigned sn = uimm (aarch64_get_instr (cpu), 9, 5); | |
6654 | unsigned sd = uimm (aarch64_get_instr (cpu), 4, 0); | |
6655 | ||
6656 | aarch64_set_FP_double (cpu, sd, aarch64_get_FP_double (cpu, sn) | |
6657 | - aarch64_get_FP_double (cpu, sm)); | |
6658 | } | |
6659 | ||
6660 | static void | |
6661 | do_FMINNM (sim_cpu *cpu) | |
6662 | { | |
6663 | /* instr[31,23] = 0 0011 1100 | |
6664 | instr[22] = float(0)/double(1) | |
6665 | instr[21] = 1 | |
6666 | instr[20,16] = Sm | |
6667 | instr[15,10] = 01 1110 | |
6668 | instr[9,5] = Sn | |
6669 | instr[4,0] = Cpu */ | |
6670 | ||
6671 | unsigned sm = uimm (aarch64_get_instr (cpu), 20, 16); | |
6672 | unsigned sn = uimm (aarch64_get_instr (cpu), 9, 5); | |
6673 | unsigned sd = uimm (aarch64_get_instr (cpu), 4, 0); | |
6674 | ||
6675 | NYI_assert (31, 23, 0x03C); | |
6676 | NYI_assert (15, 10, 0x1E); | |
6677 | ||
6678 | if (uimm (aarch64_get_instr (cpu), 22, 22)) | |
6679 | aarch64_set_FP_double (cpu, sd, | |
6680 | dminnm (aarch64_get_FP_double (cpu, sn), | |
6681 | aarch64_get_FP_double (cpu, sm))); | |
6682 | else | |
6683 | aarch64_set_FP_float (cpu, sd, | |
6684 | fminnm (aarch64_get_FP_float (cpu, sn), | |
6685 | aarch64_get_FP_float (cpu, sm))); | |
6686 | } | |
6687 | ||
6688 | static void | |
6689 | do_FMAXNM (sim_cpu *cpu) | |
6690 | { | |
6691 | /* instr[31,23] = 0 0011 1100 | |
6692 | instr[22] = float(0)/double(1) | |
6693 | instr[21] = 1 | |
6694 | instr[20,16] = Sm | |
6695 | instr[15,10] = 01 1010 | |
6696 | instr[9,5] = Sn | |
6697 | instr[4,0] = Cpu */ | |
6698 | ||
6699 | unsigned sm = uimm (aarch64_get_instr (cpu), 20, 16); | |
6700 | unsigned sn = uimm (aarch64_get_instr (cpu), 9, 5); | |
6701 | unsigned sd = uimm (aarch64_get_instr (cpu), 4, 0); | |
6702 | ||
6703 | NYI_assert (31, 23, 0x03C); | |
6704 | NYI_assert (15, 10, 0x1A); | |
6705 | ||
6706 | if (uimm (aarch64_get_instr (cpu), 22, 22)) | |
6707 | aarch64_set_FP_double (cpu, sd, | |
6708 | dmaxnm (aarch64_get_FP_double (cpu, sn), | |
6709 | aarch64_get_FP_double (cpu, sm))); | |
6710 | else | |
6711 | aarch64_set_FP_float (cpu, sd, | |
6712 | fmaxnm (aarch64_get_FP_float (cpu, sn), | |
6713 | aarch64_get_FP_float (cpu, sm))); | |
6714 | } | |
6715 | ||
6716 | static void | |
6717 | dexSimpleFPDataProc2Source (sim_cpu *cpu) | |
6718 | { | |
6719 | /* instr[31] ==> M : 0 ==> OK, 1 ==> UNALLOC | |
6720 | instr[30] = 0 | |
6721 | instr[29] ==> S : 0 ==> OK, 1 ==> UNALLOC | |
6722 | instr[28,25] = 1111 | |
6723 | instr[24] = 0 | |
6724 | instr[23,22] ==> type : 0 ==> single, 01 ==> double, 1x ==> UNALLOC | |
6725 | instr[21] = 1 | |
6726 | instr[20,16] = Vm | |
6727 | instr[15,12] ==> opcode : 0000 ==> FMUL, 0001 ==> FDIV | |
6728 | 0010 ==> FADD, 0011 ==> FSUB, | |
6729 | 0100 ==> FMAX, 0101 ==> FMIN | |
6730 | 0110 ==> FMAXNM, 0111 ==> FMINNM | |
6731 | 1000 ==> FNMUL, ow ==> UNALLOC | |
6732 | instr[11,10] = 10 | |
6733 | instr[9,5] = Vn | |
6734 | instr[4,0] = Vd */ | |
6735 | ||
6736 | uint32_t M_S = (uimm (aarch64_get_instr (cpu), 31, 31) << 1) | |
6737 | | uimm (aarch64_get_instr (cpu), 29, 29); | |
6738 | uint32_t type = uimm (aarch64_get_instr (cpu), 23, 22); | |
6739 | /* Dispatch on opcode. */ | |
6740 | uint32_t dispatch = uimm (aarch64_get_instr (cpu), 15, 12); | |
6741 | ||
6742 | if (type > 1) | |
6743 | HALT_UNALLOC; | |
6744 | ||
6745 | if (M_S != 0) | |
6746 | HALT_UNALLOC; | |
6747 | ||
6748 | if (type) | |
6749 | switch (dispatch) | |
6750 | { | |
6751 | case 0: fmuld (cpu); return; | |
6752 | case 1: fdivd (cpu); return; | |
6753 | case 2: faddd (cpu); return; | |
6754 | case 3: fsubd (cpu); return; | |
6755 | case 6: do_FMAXNM (cpu); return; | |
6756 | case 7: do_FMINNM (cpu); return; | |
6757 | case 8: fnmuld (cpu); return; | |
6758 | ||
6759 | /* Have not yet implemented fmax and fmin. */ | |
6760 | case 4: | |
6761 | case 5: | |
6762 | HALT_NYI; | |
6763 | ||
6764 | default: | |
6765 | HALT_UNALLOC; | |
6766 | } | |
6767 | else /* type == 0 => floats. */ | |
6768 | switch (dispatch) | |
6769 | { | |
6770 | case 0: fmuls (cpu); return; | |
6771 | case 1: fdivs (cpu); return; | |
6772 | case 2: fadds (cpu); return; | |
6773 | case 3: fsubs (cpu); return; | |
6774 | case 6: do_FMAXNM (cpu); return; | |
6775 | case 7: do_FMINNM (cpu); return; | |
6776 | case 8: fnmuls (cpu); return; | |
6777 | ||
6778 | case 4: | |
6779 | case 5: | |
6780 | HALT_NYI; | |
6781 | ||
6782 | default: | |
6783 | HALT_UNALLOC; | |
6784 | } | |
6785 | } | |
6786 | ||
6787 | static void | |
6788 | dexSimpleFPCondSelect (sim_cpu *cpu) | |
6789 | { | |
6790 | /* FCSEL | |
6791 | instr[31,23] = 0 0011 1100 | |
6792 | instr[22] = 0=>single 1=>double | |
6793 | instr[21] = 1 | |
6794 | instr[20,16] = Sm | |
6795 | instr[15,12] = cond | |
6796 | instr[11,10] = 11 | |
6797 | instr[9,5] = Sn | |
6798 | instr[4,0] = Cpu */ | |
6799 | unsigned sm = uimm (aarch64_get_instr (cpu), 20, 16); | |
6800 | unsigned sn = uimm (aarch64_get_instr (cpu), 9, 5); | |
6801 | unsigned sd = uimm (aarch64_get_instr (cpu), 4, 0); | |
6802 | uint32_t set = testConditionCode (cpu, uimm (aarch64_get_instr (cpu), 15, 12)); | |
6803 | ||
6804 | NYI_assert (31, 23, 0x03C); | |
6805 | NYI_assert (11, 10, 0x3); | |
6806 | ||
6807 | if (uimm (aarch64_get_instr (cpu), 22, 22)) | |
6808 | aarch64_set_FP_double (cpu, sd, set ? sn : sm); | |
6809 | else | |
6810 | aarch64_set_FP_float (cpu, sd, set ? sn : sm); | |
6811 | } | |
6812 | ||
6813 | /* Store 32 bit unscaled signed 9 bit. */ | |
6814 | static void | |
6815 | fsturs (sim_cpu *cpu, int32_t offset) | |
6816 | { | |
6817 | unsigned int rn = uimm (aarch64_get_instr (cpu), 9, 5); | |
6818 | unsigned int st = uimm (aarch64_get_instr (cpu), 4, 0); | |
6819 | ||
6820 | aarch64_set_mem_float (cpu, aarch64_get_reg_u64 (cpu, st, 1) + offset, | |
6821 | aarch64_get_FP_float (cpu, rn)); | |
6822 | } | |
6823 | ||
6824 | /* Store 64 bit unscaled signed 9 bit. */ | |
6825 | static void | |
6826 | fsturd (sim_cpu *cpu, int32_t offset) | |
6827 | { | |
6828 | unsigned int rn = uimm (aarch64_get_instr (cpu), 9, 5); | |
6829 | unsigned int st = uimm (aarch64_get_instr (cpu), 4, 0); | |
6830 | ||
6831 | aarch64_set_mem_double (cpu, aarch64_get_reg_u64 (cpu, st, 1) + offset, | |
6832 | aarch64_get_FP_double (cpu, rn)); | |
6833 | } | |
6834 | ||
6835 | /* Store 128 bit unscaled signed 9 bit. */ | |
6836 | static void | |
6837 | fsturq (sim_cpu *cpu, int32_t offset) | |
6838 | { | |
6839 | unsigned int rn = uimm (aarch64_get_instr (cpu), 9, 5); | |
6840 | unsigned int st = uimm (aarch64_get_instr (cpu), 4, 0); | |
6841 | FRegister a; | |
6842 | ||
6843 | aarch64_get_FP_long_double (cpu, rn, & a); | |
6844 | aarch64_set_mem_long_double (cpu, | |
6845 | aarch64_get_reg_u64 (cpu, st, 1) | |
6846 | + offset, a); | |
6847 | } | |
6848 | ||
6849 | /* TODO FP move register. */ | |
6850 | ||
6851 | /* 32 bit fp to fp move register. */ | |
6852 | static void | |
6853 | ffmovs (sim_cpu *cpu) | |
6854 | { | |
6855 | unsigned int rn = uimm (aarch64_get_instr (cpu), 9, 5); | |
6856 | unsigned int st = uimm (aarch64_get_instr (cpu), 4, 0); | |
6857 | ||
6858 | aarch64_set_FP_float (cpu, st, aarch64_get_FP_float (cpu, rn)); | |
6859 | } | |
6860 | ||
6861 | /* 64 bit fp to fp move register. */ | |
6862 | static void | |
6863 | ffmovd (sim_cpu *cpu) | |
6864 | { | |
6865 | unsigned int rn = uimm (aarch64_get_instr (cpu), 9, 5); | |
6866 | unsigned int st = uimm (aarch64_get_instr (cpu), 4, 0); | |
6867 | ||
6868 | aarch64_set_FP_double (cpu, st, aarch64_get_FP_double (cpu, rn)); | |
6869 | } | |
6870 | ||
6871 | /* 32 bit GReg to Vec move register. */ | |
6872 | static void | |
6873 | fgmovs (sim_cpu *cpu) | |
6874 | { | |
6875 | unsigned int rn = uimm (aarch64_get_instr (cpu), 9, 5); | |
6876 | unsigned int st = uimm (aarch64_get_instr (cpu), 4, 0); | |
6877 | ||
6878 | aarch64_set_vec_u32 (cpu, st, 0, aarch64_get_reg_u32 (cpu, rn, NO_SP)); | |
6879 | } | |
6880 | ||
6881 | /* 64 bit g to fp move register. */ | |
6882 | static void | |
6883 | fgmovd (sim_cpu *cpu) | |
6884 | { | |
6885 | unsigned int rn = uimm (aarch64_get_instr (cpu), 9, 5); | |
6886 | unsigned int st = uimm (aarch64_get_instr (cpu), 4, 0); | |
6887 | ||
6888 | aarch64_set_vec_u64 (cpu, st, 0, aarch64_get_reg_u64 (cpu, rn, NO_SP)); | |
6889 | } | |
6890 | ||
6891 | /* 32 bit fp to g move register. */ | |
6892 | static void | |
6893 | gfmovs (sim_cpu *cpu) | |
6894 | { | |
6895 | unsigned int rn = uimm (aarch64_get_instr (cpu), 9, 5); | |
6896 | unsigned int st = uimm (aarch64_get_instr (cpu), 4, 0); | |
6897 | ||
6898 | aarch64_set_reg_u64 (cpu, st, NO_SP, aarch64_get_vec_u32 (cpu, rn, 0)); | |
6899 | } | |
6900 | ||
6901 | /* 64 bit fp to g move register. */ | |
6902 | static void | |
6903 | gfmovd (sim_cpu *cpu) | |
6904 | { | |
6905 | unsigned int rn = uimm (aarch64_get_instr (cpu), 9, 5); | |
6906 | unsigned int st = uimm (aarch64_get_instr (cpu), 4, 0); | |
6907 | ||
6908 | aarch64_set_reg_u64 (cpu, st, NO_SP, aarch64_get_vec_u64 (cpu, rn, 0)); | |
6909 | } | |
6910 | ||
6911 | /* FP move immediate | |
6912 | ||
6913 | These install an immediate 8 bit value in the target register | |
6914 | where the 8 bits comprise 1 sign bit, 4 bits of fraction and a 3 | |
6915 | bit exponent. */ | |
6916 | ||
6917 | static void | |
6918 | fmovs (sim_cpu *cpu) | |
6919 | { | |
6920 | unsigned int sd = uimm (aarch64_get_instr (cpu), 4, 0); | |
6921 | uint32_t imm = uimm (aarch64_get_instr (cpu), 20, 13); | |
6922 | float f = fp_immediate_for_encoding_32 (imm); | |
6923 | ||
6924 | aarch64_set_FP_float (cpu, sd, f); | |
6925 | } | |
6926 | ||
6927 | static void | |
6928 | fmovd (sim_cpu *cpu) | |
6929 | { | |
6930 | unsigned int sd = uimm (aarch64_get_instr (cpu), 4, 0); | |
6931 | uint32_t imm = uimm (aarch64_get_instr (cpu), 20, 13); | |
6932 | double d = fp_immediate_for_encoding_64 (imm); | |
6933 | ||
6934 | aarch64_set_FP_double (cpu, sd, d); | |
6935 | } | |
6936 | ||
6937 | static void | |
6938 | dexSimpleFPImmediate (sim_cpu *cpu) | |
6939 | { | |
6940 | /* instr[31,23] == 00111100 | |
6941 | instr[22] == type : single(0)/double(1) | |
6942 | instr[21] == 1 | |
6943 | instr[20,13] == imm8 | |
6944 | instr[12,10] == 100 | |
6945 | instr[9,5] == imm5 : 00000 ==> PK, ow ==> UNALLOC | |
6946 | instr[4,0] == Rd */ | |
6947 | uint32_t imm5 = uimm (aarch64_get_instr (cpu), 9, 5); | |
6948 | ||
6949 | NYI_assert (31, 23, 0x3C); | |
6950 | ||
6951 | if (imm5 != 0) | |
6952 | HALT_UNALLOC; | |
6953 | ||
6954 | if (uimm (aarch64_get_instr (cpu), 22, 22)) | |
6955 | fmovd (cpu); | |
6956 | else | |
6957 | fmovs (cpu); | |
6958 | } | |
6959 | ||
6960 | /* TODO specific decode and execute for group Load Store. */ | |
6961 | ||
6962 | /* TODO FP load/store single register (unscaled offset). */ | |
6963 | ||
6964 | /* TODO load 8 bit unscaled signed 9 bit. */ | |
6965 | /* TODO load 16 bit unscaled signed 9 bit. */ | |
6966 | ||
6967 | /* Load 32 bit unscaled signed 9 bit. */ | |
6968 | static void | |
6969 | fldurs (sim_cpu *cpu, int32_t offset) | |
6970 | { | |
6971 | unsigned int rn = uimm (aarch64_get_instr (cpu), 9, 5); | |
6972 | unsigned int st = uimm (aarch64_get_instr (cpu), 4, 0); | |
6973 | ||
6974 | aarch64_set_FP_float (cpu, st, aarch64_get_mem_float | |
6975 | (cpu, aarch64_get_reg_u64 (cpu, rn, SP_OK) + offset)); | |
6976 | } | |
6977 | ||
6978 | /* Load 64 bit unscaled signed 9 bit. */ | |
6979 | static void | |
6980 | fldurd (sim_cpu *cpu, int32_t offset) | |
6981 | { | |
6982 | unsigned int rn = uimm (aarch64_get_instr (cpu), 9, 5); | |
6983 | unsigned int st = uimm (aarch64_get_instr (cpu), 4, 0); | |
6984 | ||
6985 | aarch64_set_FP_double (cpu, st, aarch64_get_mem_double | |
6986 | (cpu, aarch64_get_reg_u64 (cpu, rn, SP_OK) + offset)); | |
6987 | } | |
6988 | ||
6989 | /* Load 128 bit unscaled signed 9 bit. */ | |
6990 | static void | |
6991 | fldurq (sim_cpu *cpu, int32_t offset) | |
6992 | { | |
6993 | unsigned int rn = uimm (aarch64_get_instr (cpu), 9, 5); | |
6994 | unsigned int st = uimm (aarch64_get_instr (cpu), 4, 0); | |
6995 | FRegister a; | |
6996 | uint64_t addr = aarch64_get_reg_u64 (cpu, rn, SP_OK) + offset; | |
6997 | ||
6998 | aarch64_get_mem_long_double (cpu, addr, & a); | |
6999 | aarch64_set_FP_long_double (cpu, st, a); | |
7000 | } | |
7001 | ||
7002 | /* TODO store 8 bit unscaled signed 9 bit. */ | |
7003 | /* TODO store 16 bit unscaled signed 9 bit. */ | |
7004 | ||
7005 | ||
7006 | /* 1 source. */ | |
7007 | ||
7008 | /* Float absolute value. */ | |
7009 | static void | |
7010 | fabss (sim_cpu *cpu) | |
7011 | { | |
7012 | unsigned sn = uimm (aarch64_get_instr (cpu), 9, 5); | |
7013 | unsigned sd = uimm (aarch64_get_instr (cpu), 4, 0); | |
7014 | float value = aarch64_get_FP_float (cpu, sn); | |
7015 | ||
7016 | aarch64_set_FP_float (cpu, sd, fabsf (value)); | |
7017 | } | |
7018 | ||
7019 | /* Double absolute value. */ | |
7020 | static void | |
7021 | fabcpu (sim_cpu *cpu) | |
7022 | { | |
7023 | unsigned sn = uimm (aarch64_get_instr (cpu), 9, 5); | |
7024 | unsigned sd = uimm (aarch64_get_instr (cpu), 4, 0); | |
7025 | double value = aarch64_get_FP_double (cpu, sn); | |
7026 | ||
7027 | aarch64_set_FP_double (cpu, sd, fabs (value)); | |
7028 | } | |
7029 | ||
7030 | /* Float negative value. */ | |
7031 | static void | |
7032 | fnegs (sim_cpu *cpu) | |
7033 | { | |
7034 | unsigned sn = uimm (aarch64_get_instr (cpu), 9, 5); | |
7035 | unsigned sd = uimm (aarch64_get_instr (cpu), 4, 0); | |
7036 | ||
7037 | aarch64_set_FP_float (cpu, sd, - aarch64_get_FP_float (cpu, sn)); | |
7038 | } | |
7039 | ||
7040 | /* Double negative value. */ | |
7041 | static void | |
7042 | fnegd (sim_cpu *cpu) | |
7043 | { | |
7044 | unsigned sn = uimm (aarch64_get_instr (cpu), 9, 5); | |
7045 | unsigned sd = uimm (aarch64_get_instr (cpu), 4, 0); | |
7046 | ||
7047 | aarch64_set_FP_double (cpu, sd, - aarch64_get_FP_double (cpu, sn)); | |
7048 | } | |
7049 | ||
7050 | /* Float square root. */ | |
7051 | static void | |
7052 | fsqrts (sim_cpu *cpu) | |
7053 | { | |
7054 | unsigned sn = uimm (aarch64_get_instr (cpu), 9, 5); | |
7055 | unsigned sd = uimm (aarch64_get_instr (cpu), 4, 0); | |
7056 | ||
7057 | aarch64_set_FP_float (cpu, sd, sqrt (aarch64_get_FP_float (cpu, sn))); | |
7058 | } | |
7059 | ||
7060 | /* Double square root. */ | |
7061 | static void | |
7062 | fsqrtd (sim_cpu *cpu) | |
7063 | { | |
7064 | unsigned sn = uimm (aarch64_get_instr (cpu), 9, 5); | |
7065 | unsigned sd = uimm (aarch64_get_instr (cpu), 4, 0); | |
7066 | ||
7067 | aarch64_set_FP_double (cpu, sd, | |
7068 | sqrt (aarch64_get_FP_double (cpu, sn))); | |
7069 | } | |
7070 | ||
7071 | /* Convert double to float. */ | |
7072 | static void | |
7073 | fcvtds (sim_cpu *cpu) | |
7074 | { | |
7075 | unsigned sn = uimm (aarch64_get_instr (cpu), 9, 5); | |
7076 | unsigned sd = uimm (aarch64_get_instr (cpu), 4, 0); | |
7077 | ||
7078 | aarch64_set_FP_float (cpu, sd, (float) aarch64_get_FP_double (cpu, sn)); | |
7079 | } | |
7080 | ||
7081 | /* Convert float to double. */ | |
7082 | static void | |
7083 | fcvtcpu (sim_cpu *cpu) | |
7084 | { | |
7085 | unsigned sn = uimm (aarch64_get_instr (cpu), 9, 5); | |
7086 | unsigned sd = uimm (aarch64_get_instr (cpu), 4, 0); | |
7087 | ||
7088 | aarch64_set_FP_double (cpu, sd, (double) aarch64_get_FP_float (cpu, sn)); | |
7089 | } | |
7090 | ||
7091 | static void | |
7092 | do_FRINT (sim_cpu *cpu) | |
7093 | { | |
7094 | /* instr[31,23] = 0001 1110 0 | |
7095 | instr[22] = single(0)/double(1) | |
7096 | instr[21,18] = 1001 | |
7097 | instr[17,15] = rounding mode | |
7098 | instr[14,10] = 10000 | |
7099 | instr[9,5] = source | |
7100 | instr[4,0] = dest */ | |
7101 | ||
7102 | float val; | |
7103 | unsigned rs = uimm (aarch64_get_instr (cpu), 9, 5); | |
7104 | unsigned rd = uimm (aarch64_get_instr (cpu), 4, 0); | |
7105 | unsigned int rmode = uimm (aarch64_get_instr (cpu), 17, 15); | |
7106 | ||
7107 | NYI_assert (31, 23, 0x03C); | |
7108 | NYI_assert (21, 18, 0x9); | |
7109 | NYI_assert (14, 10, 0x10); | |
7110 | ||
7111 | if (rmode == 6 || rmode == 7) | |
7112 | /* FIXME: Add support for rmode == 6 exactness check. */ | |
7113 | rmode = uimm (aarch64_get_FPSR (cpu), 23, 22); | |
7114 | ||
7115 | if (uimm (aarch64_get_instr (cpu), 22, 22)) | |
7116 | { | |
7117 | double val = aarch64_get_FP_double (cpu, rs); | |
7118 | ||
7119 | switch (rmode) | |
7120 | { | |
7121 | case 0: /* mode N: nearest or even. */ | |
7122 | { | |
7123 | double rval = round (val); | |
7124 | ||
7125 | if (val - rval == 0.5) | |
7126 | { | |
7127 | if (((rval / 2.0) * 2.0) != rval) | |
7128 | rval += 1.0; | |
7129 | } | |
7130 | ||
7131 | aarch64_set_FP_double (cpu, rd, round (val)); | |
7132 | return; | |
7133 | } | |
7134 | ||
7135 | case 1: /* mode P: towards +inf. */ | |
7136 | if (val < 0.0) | |
7137 | aarch64_set_FP_double (cpu, rd, trunc (val)); | |
7138 | else | |
7139 | aarch64_set_FP_double (cpu, rd, round (val)); | |
7140 | return; | |
7141 | ||
7142 | case 2: /* mode M: towards -inf. */ | |
7143 | if (val < 0.0) | |
7144 | aarch64_set_FP_double (cpu, rd, round (val)); | |
7145 | else | |
7146 | aarch64_set_FP_double (cpu, rd, trunc (val)); | |
7147 | return; | |
7148 | ||
7149 | case 3: /* mode Z: towards 0. */ | |
7150 | aarch64_set_FP_double (cpu, rd, trunc (val)); | |
7151 | return; | |
7152 | ||
7153 | case 4: /* mode A: away from 0. */ | |
7154 | aarch64_set_FP_double (cpu, rd, round (val)); | |
7155 | return; | |
7156 | ||
7157 | case 6: /* mode X: use FPCR with exactness check. */ | |
7158 | case 7: /* mode I: use FPCR mode. */ | |
7159 | HALT_NYI; | |
7160 | ||
7161 | default: | |
7162 | HALT_UNALLOC; | |
7163 | } | |
7164 | } | |
7165 | ||
7166 | val = aarch64_get_FP_float (cpu, rs); | |
7167 | ||
7168 | switch (rmode) | |
7169 | { | |
7170 | case 0: /* mode N: nearest or even. */ | |
7171 | { | |
7172 | float rval = roundf (val); | |
7173 | ||
7174 | if (val - rval == 0.5) | |
7175 | { | |
7176 | if (((rval / 2.0) * 2.0) != rval) | |
7177 | rval += 1.0; | |
7178 | } | |
7179 | ||
7180 | aarch64_set_FP_float (cpu, rd, rval); | |
7181 | return; | |
7182 | } | |
7183 | ||
7184 | case 1: /* mode P: towards +inf. */ | |
7185 | if (val < 0.0) | |
7186 | aarch64_set_FP_float (cpu, rd, truncf (val)); | |
7187 | else | |
7188 | aarch64_set_FP_float (cpu, rd, roundf (val)); | |
7189 | return; | |
7190 | ||
7191 | case 2: /* mode M: towards -inf. */ | |
7192 | if (val < 0.0) | |
7193 | aarch64_set_FP_float (cpu, rd, truncf (val)); | |
7194 | else | |
7195 | aarch64_set_FP_float (cpu, rd, roundf (val)); | |
7196 | return; | |
7197 | ||
7198 | case 3: /* mode Z: towards 0. */ | |
7199 | aarch64_set_FP_float (cpu, rd, truncf (val)); | |
7200 | return; | |
7201 | ||
7202 | case 4: /* mode A: away from 0. */ | |
7203 | aarch64_set_FP_float (cpu, rd, roundf (val)); | |
7204 | return; | |
7205 | ||
7206 | case 6: /* mode X: use FPCR with exactness check. */ | |
7207 | case 7: /* mode I: use FPCR mode. */ | |
7208 | HALT_NYI; | |
7209 | ||
7210 | default: | |
7211 | HALT_UNALLOC; | |
7212 | } | |
7213 | } | |
7214 | ||
7215 | static void | |
7216 | dexSimpleFPDataProc1Source (sim_cpu *cpu) | |
7217 | { | |
7218 | /* instr[31] ==> M : 0 ==> OK, 1 ==> UNALLOC | |
7219 | instr[30] = 0 | |
7220 | instr[29] ==> S : 0 ==> OK, 1 ==> UNALLOC | |
7221 | instr[28,25] = 1111 | |
7222 | instr[24] = 0 | |
7223 | instr[23,22] ==> type : 00 ==> source is single, | |
7224 | 01 ==> source is double | |
7225 | 10 ==> UNALLOC | |
7226 | 11 ==> UNALLOC or source is half | |
7227 | instr[21] = 1 | |
7228 | instr[20,15] ==> opcode : with type 00 or 01 | |
7229 | 000000 ==> FMOV, 000001 ==> FABS, | |
7230 | 000010 ==> FNEG, 000011 ==> FSQRT, | |
7231 | 000100 ==> UNALLOC, 000101 ==> FCVT,(to single/double) | |
7232 | 000110 ==> UNALLOC, 000111 ==> FCVT (to half) | |
7233 | 001000 ==> FRINTN, 001001 ==> FRINTP, | |
7234 | 001010 ==> FRINTM, 001011 ==> FRINTZ, | |
7235 | 001100 ==> FRINTA, 001101 ==> UNALLOC | |
7236 | 001110 ==> FRINTX, 001111 ==> FRINTI | |
7237 | with type 11 | |
7238 | 000100 ==> FCVT (half-to-single) | |
7239 | 000101 ==> FCVT (half-to-double) | |
7240 | instr[14,10] = 10000. */ | |
7241 | ||
7242 | uint32_t M_S = (uimm (aarch64_get_instr (cpu), 31, 31) << 1) | |
7243 | | uimm (aarch64_get_instr (cpu), 29, 29); | |
7244 | uint32_t type = uimm (aarch64_get_instr (cpu), 23, 22); | |
7245 | uint32_t opcode = uimm (aarch64_get_instr (cpu), 20, 15); | |
7246 | ||
7247 | if (M_S != 0) | |
7248 | HALT_UNALLOC; | |
7249 | ||
7250 | if (type == 3) | |
7251 | { | |
7252 | if (opcode == 4 || opcode == 5) | |
7253 | HALT_NYI; | |
7254 | else | |
7255 | HALT_UNALLOC; | |
7256 | } | |
7257 | ||
7258 | if (type == 2) | |
7259 | HALT_UNALLOC; | |
7260 | ||
7261 | switch (opcode) | |
7262 | { | |
7263 | case 0: | |
7264 | if (type) | |
7265 | ffmovd (cpu); | |
7266 | else | |
7267 | ffmovs (cpu); | |
7268 | return; | |
7269 | ||
7270 | case 1: | |
7271 | if (type) | |
7272 | fabcpu (cpu); | |
7273 | else | |
7274 | fabss (cpu); | |
7275 | return; | |
7276 | ||
7277 | case 2: | |
7278 | if (type) | |
7279 | fnegd (cpu); | |
7280 | else | |
7281 | fnegs (cpu); | |
7282 | return; | |
7283 | ||
7284 | case 3: | |
7285 | if (type) | |
7286 | fsqrtd (cpu); | |
7287 | else | |
7288 | fsqrts (cpu); | |
7289 | return; | |
7290 | ||
7291 | case 4: | |
7292 | if (type) | |
7293 | fcvtds (cpu); | |
7294 | else | |
7295 | HALT_UNALLOC; | |
7296 | return; | |
7297 | ||
7298 | case 5: | |
7299 | if (type) | |
7300 | HALT_UNALLOC; | |
7301 | fcvtcpu (cpu); | |
7302 | return; | |
7303 | ||
7304 | case 8: /* FRINTN etc. */ | |
7305 | case 9: | |
7306 | case 10: | |
7307 | case 11: | |
7308 | case 12: | |
7309 | case 14: | |
7310 | case 15: | |
7311 | do_FRINT (cpu); | |
7312 | return; | |
7313 | ||
7314 | case 7: /* FCVT double/single to half precision. */ | |
7315 | case 13: | |
7316 | HALT_NYI; | |
7317 | ||
7318 | default: | |
7319 | HALT_UNALLOC; | |
7320 | } | |
7321 | } | |
7322 | ||
7323 | /* 32 bit signed int to float. */ | |
7324 | static void | |
7325 | scvtf32 (sim_cpu *cpu) | |
7326 | { | |
7327 | unsigned rn = uimm (aarch64_get_instr (cpu), 9, 5); | |
7328 | unsigned sd = uimm (aarch64_get_instr (cpu), 4, 0); | |
7329 | ||
7330 | aarch64_set_FP_float | |
7331 | (cpu, sd, (float) aarch64_get_reg_s32 (cpu, rn, NO_SP)); | |
7332 | } | |
7333 | ||
7334 | /* signed int to float. */ | |
7335 | static void | |
7336 | scvtf (sim_cpu *cpu) | |
7337 | { | |
7338 | unsigned rn = uimm (aarch64_get_instr (cpu), 9, 5); | |
7339 | unsigned sd = uimm (aarch64_get_instr (cpu), 4, 0); | |
7340 | ||
7341 | aarch64_set_FP_float | |
7342 | (cpu, sd, (float) aarch64_get_reg_s64 (cpu, rn, NO_SP)); | |
7343 | } | |
7344 | ||
7345 | /* 32 bit signed int to double. */ | |
7346 | static void | |
7347 | scvtd32 (sim_cpu *cpu) | |
7348 | { | |
7349 | unsigned rn = uimm (aarch64_get_instr (cpu), 9, 5); | |
7350 | unsigned sd = uimm (aarch64_get_instr (cpu), 4, 0); | |
7351 | ||
7352 | aarch64_set_FP_double | |
7353 | (cpu, sd, (double) aarch64_get_reg_s32 (cpu, rn, NO_SP)); | |
7354 | } | |
7355 | ||
7356 | /* signed int to double. */ | |
7357 | static void | |
7358 | scvtd (sim_cpu *cpu) | |
7359 | { | |
7360 | unsigned rn = uimm (aarch64_get_instr (cpu), 9, 5); | |
7361 | unsigned sd = uimm (aarch64_get_instr (cpu), 4, 0); | |
7362 | ||
7363 | aarch64_set_FP_double | |
7364 | (cpu, sd, (double) aarch64_get_reg_s64 (cpu, rn, NO_SP)); | |
7365 | } | |
7366 | ||
7367 | static const float FLOAT_INT_MAX = (float) INT_MAX; | |
7368 | static const float FLOAT_INT_MIN = (float) INT_MIN; | |
7369 | static const double DOUBLE_INT_MAX = (double) INT_MAX; | |
7370 | static const double DOUBLE_INT_MIN = (double) INT_MIN; | |
7371 | static const float FLOAT_LONG_MAX = (float) LONG_MAX; | |
7372 | static const float FLOAT_LONG_MIN = (float) LONG_MIN; | |
7373 | static const double DOUBLE_LONG_MAX = (double) LONG_MAX; | |
7374 | static const double DOUBLE_LONG_MIN = (double) LONG_MIN; | |
7375 | ||
7376 | /* Check for FP exception conditions: | |
7377 | NaN raises IO | |
7378 | Infinity raises IO | |
7379 | Out of Range raises IO and IX and saturates value | |
7380 | Denormal raises ID and IX and sets to zero. */ | |
7381 | #define RAISE_EXCEPTIONS(F, VALUE, FTYPE, ITYPE) \ | |
7382 | do \ | |
7383 | { \ | |
7384 | switch (fpclassify (F)) \ | |
7385 | { \ | |
7386 | case FP_INFINITE: \ | |
7387 | case FP_NAN: \ | |
7388 | aarch64_set_FPSR (cpu, IO); \ | |
7389 | if (signbit (F)) \ | |
7390 | VALUE = ITYPE##_MAX; \ | |
7391 | else \ | |
7392 | VALUE = ITYPE##_MIN; \ | |
7393 | break; \ | |
7394 | \ | |
7395 | case FP_NORMAL: \ | |
7396 | if (F >= FTYPE##_##ITYPE##_MAX) \ | |
7397 | { \ | |
7398 | aarch64_set_FPSR_bits (cpu, IO | IX, IO | IX); \ | |
7399 | VALUE = ITYPE##_MAX; \ | |
7400 | } \ | |
7401 | else if (F <= FTYPE##_##ITYPE##_MIN) \ | |
7402 | { \ | |
7403 | aarch64_set_FPSR_bits (cpu, IO | IX, IO | IX); \ | |
7404 | VALUE = ITYPE##_MIN; \ | |
7405 | } \ | |
7406 | break; \ | |
7407 | \ | |
7408 | case FP_SUBNORMAL: \ | |
7409 | aarch64_set_FPSR_bits (cpu, IO | IX | ID, IX | ID); \ | |
7410 | VALUE = 0; \ | |
7411 | break; \ | |
7412 | \ | |
7413 | default: \ | |
7414 | case FP_ZERO: \ | |
7415 | VALUE = 0; \ | |
7416 | break; \ | |
7417 | } \ | |
7418 | } \ | |
7419 | while (0) | |
7420 | ||
7421 | /* 32 bit convert float to signed int truncate towards zero. */ | |
7422 | static void | |
7423 | fcvtszs32 (sim_cpu *cpu) | |
7424 | { | |
7425 | unsigned sn = uimm (aarch64_get_instr (cpu), 9, 5); | |
7426 | unsigned rd = uimm (aarch64_get_instr (cpu), 4, 0); | |
7427 | /* TODO : check that this rounds toward zero. */ | |
7428 | float f = aarch64_get_FP_float (cpu, sn); | |
7429 | int32_t value = (int32_t) f; | |
7430 | ||
7431 | RAISE_EXCEPTIONS (f, value, FLOAT, INT); | |
7432 | ||
7433 | /* Avoid sign extension to 64 bit. */ | |
7434 | aarch64_set_reg_u64 (cpu, rd, NO_SP, (uint32_t) value); | |
7435 | } | |
7436 | ||
7437 | /* 64 bit convert float to signed int truncate towards zero. */ | |
7438 | static void | |
7439 | fcvtszs (sim_cpu *cpu) | |
7440 | { | |
7441 | unsigned sn = uimm (aarch64_get_instr (cpu), 9, 5); | |
7442 | unsigned rd = uimm (aarch64_get_instr (cpu), 4, 0); | |
7443 | float f = aarch64_get_FP_float (cpu, sn); | |
7444 | int64_t value = (int64_t) f; | |
7445 | ||
7446 | RAISE_EXCEPTIONS (f, value, FLOAT, LONG); | |
7447 | ||
7448 | aarch64_set_reg_s64 (cpu, rd, NO_SP, value); | |
7449 | } | |
7450 | ||
7451 | /* 32 bit convert double to signed int truncate towards zero. */ | |
7452 | static void | |
7453 | fcvtszd32 (sim_cpu *cpu) | |
7454 | { | |
7455 | unsigned sn = uimm (aarch64_get_instr (cpu), 9, 5); | |
7456 | unsigned rd = uimm (aarch64_get_instr (cpu), 4, 0); | |
7457 | /* TODO : check that this rounds toward zero. */ | |
7458 | double d = aarch64_get_FP_double (cpu, sn); | |
7459 | int32_t value = (int32_t) d; | |
7460 | ||
7461 | RAISE_EXCEPTIONS (d, value, DOUBLE, INT); | |
7462 | ||
7463 | /* Avoid sign extension to 64 bit. */ | |
7464 | aarch64_set_reg_u64 (cpu, rd, NO_SP, (uint32_t) value); | |
7465 | } | |
7466 | ||
7467 | /* 64 bit convert double to signed int truncate towards zero. */ | |
7468 | static void | |
7469 | fcvtszd (sim_cpu *cpu) | |
7470 | { | |
7471 | unsigned sn = uimm (aarch64_get_instr (cpu), 9, 5); | |
7472 | unsigned rd = uimm (aarch64_get_instr (cpu), 4, 0); | |
7473 | /* TODO : check that this rounds toward zero. */ | |
7474 | double d = aarch64_get_FP_double (cpu, sn); | |
7475 | int64_t value; | |
7476 | ||
7477 | value = (int64_t) d; | |
7478 | ||
7479 | RAISE_EXCEPTIONS (d, value, DOUBLE, LONG); | |
7480 | ||
7481 | aarch64_set_reg_s64 (cpu, rd, NO_SP, value); | |
7482 | } | |
7483 | ||
7484 | static void | |
7485 | do_fcvtzu (sim_cpu *cpu) | |
7486 | { | |
7487 | /* instr[31] = size: 32-bit (0), 64-bit (1) | |
7488 | instr[30,23] = 00111100 | |
7489 | instr[22] = type: single (0)/ double (1) | |
7490 | instr[21] = enable (0)/disable(1) precision | |
7491 | instr[20,16] = 11001 | |
7492 | instr[15,10] = precision | |
7493 | instr[9,5] = Rs | |
7494 | instr[4,0] = Rd. */ | |
7495 | ||
7496 | unsigned rs = uimm (aarch64_get_instr (cpu), 9, 5); | |
7497 | unsigned rd = uimm (aarch64_get_instr (cpu), 4, 0); | |
7498 | ||
7499 | NYI_assert (30, 23, 0x3C); | |
7500 | NYI_assert (20, 16, 0x19); | |
7501 | ||
7502 | if (uimm (aarch64_get_instr (cpu), 21, 21) != 1) | |
7503 | /* Convert to fixed point. */ | |
7504 | HALT_NYI; | |
7505 | ||
7506 | if (uimm (aarch64_get_instr (cpu), 31, 31)) | |
7507 | { | |
7508 | /* Convert to unsigned 64-bit integer. */ | |
7509 | if (uimm (aarch64_get_instr (cpu), 22, 22)) | |
7510 | { | |
7511 | double d = aarch64_get_FP_double (cpu, rs); | |
7512 | uint64_t value = (uint64_t) d; | |
7513 | ||
7514 | /* Do not raise an exception if we have reached ULONG_MAX. */ | |
7515 | if (value != (1UL << 63)) | |
7516 | RAISE_EXCEPTIONS (d, value, DOUBLE, LONG); | |
7517 | ||
7518 | aarch64_set_reg_u64 (cpu, rd, NO_SP, value); | |
7519 | } | |
7520 | else | |
7521 | { | |
7522 | float f = aarch64_get_FP_float (cpu, rs); | |
7523 | uint64_t value = (uint64_t) f; | |
7524 | ||
7525 | /* Do not raise an exception if we have reached ULONG_MAX. */ | |
7526 | if (value != (1UL << 63)) | |
7527 | RAISE_EXCEPTIONS (f, value, FLOAT, LONG); | |
7528 | ||
7529 | aarch64_set_reg_u64 (cpu, rd, NO_SP, value); | |
7530 | } | |
7531 | } | |
7532 | else | |
7533 | { | |
7534 | uint32_t value; | |
7535 | ||
7536 | /* Convert to unsigned 32-bit integer. */ | |
7537 | if (uimm (aarch64_get_instr (cpu), 22, 22)) | |
7538 | { | |
7539 | double d = aarch64_get_FP_double (cpu, rs); | |
7540 | ||
7541 | value = (uint32_t) d; | |
7542 | /* Do not raise an exception if we have reached UINT_MAX. */ | |
7543 | if (value != (1UL << 31)) | |
7544 | RAISE_EXCEPTIONS (d, value, DOUBLE, INT); | |
7545 | } | |
7546 | else | |
7547 | { | |
7548 | float f = aarch64_get_FP_float (cpu, rs); | |
7549 | ||
7550 | value = (uint32_t) f; | |
7551 | /* Do not raise an exception if we have reached UINT_MAX. */ | |
7552 | if (value != (1UL << 31)) | |
7553 | RAISE_EXCEPTIONS (f, value, FLOAT, INT); | |
7554 | } | |
7555 | ||
7556 | aarch64_set_reg_u64 (cpu, rd, NO_SP, value); | |
7557 | } | |
7558 | } | |
7559 | ||
7560 | static void | |
7561 | do_UCVTF (sim_cpu *cpu) | |
7562 | { | |
7563 | /* instr[31] = size: 32-bit (0), 64-bit (1) | |
7564 | instr[30,23] = 001 1110 0 | |
7565 | instr[22] = type: single (0)/ double (1) | |
7566 | instr[21] = enable (0)/disable(1) precision | |
7567 | instr[20,16] = 0 0011 | |
7568 | instr[15,10] = precision | |
7569 | instr[9,5] = Rs | |
7570 | instr[4,0] = Rd. */ | |
7571 | ||
7572 | unsigned rs = uimm (aarch64_get_instr (cpu), 9, 5); | |
7573 | unsigned rd = uimm (aarch64_get_instr (cpu), 4, 0); | |
7574 | ||
7575 | NYI_assert (30, 23, 0x3C); | |
7576 | NYI_assert (20, 16, 0x03); | |
7577 | ||
7578 | if (uimm (aarch64_get_instr (cpu), 21, 21) != 1) | |
7579 | HALT_NYI; | |
7580 | ||
7581 | /* FIXME: Add exception raising. */ | |
7582 | if (uimm (aarch64_get_instr (cpu), 31, 31)) | |
7583 | { | |
7584 | uint64_t value = aarch64_get_reg_u64 (cpu, rs, NO_SP); | |
7585 | ||
7586 | if (uimm (aarch64_get_instr (cpu), 22, 22)) | |
7587 | aarch64_set_FP_double (cpu, rd, (double) value); | |
7588 | else | |
7589 | aarch64_set_FP_float (cpu, rd, (float) value); | |
7590 | } | |
7591 | else | |
7592 | { | |
7593 | uint32_t value = aarch64_get_reg_u32 (cpu, rs, NO_SP); | |
7594 | ||
7595 | if (uimm (aarch64_get_instr (cpu), 22, 22)) | |
7596 | aarch64_set_FP_double (cpu, rd, (double) value); | |
7597 | else | |
7598 | aarch64_set_FP_float (cpu, rd, (float) value); | |
7599 | } | |
7600 | } | |
7601 | ||
7602 | static void | |
7603 | float_vector_move (sim_cpu *cpu) | |
7604 | { | |
7605 | /* instr[31,17] == 100 1111 0101 0111 | |
7606 | instr[16] ==> direction 0=> to GR, 1=> from GR | |
7607 | instr[15,10] => ??? | |
7608 | instr[9,5] ==> source | |
7609 | instr[4,0] ==> dest. */ | |
7610 | ||
7611 | unsigned rn = uimm (aarch64_get_instr (cpu), 9, 5); | |
7612 | unsigned rd = uimm (aarch64_get_instr (cpu), 4, 0); | |
7613 | ||
7614 | NYI_assert (31, 17, 0x4F57); | |
7615 | ||
7616 | if (uimm (aarch64_get_instr (cpu), 15, 10) != 0) | |
7617 | HALT_UNALLOC; | |
7618 | ||
7619 | if (uimm (aarch64_get_instr (cpu), 16, 16)) | |
7620 | aarch64_set_vec_u64 (cpu, rd, 1, aarch64_get_reg_u64 (cpu, rn, NO_SP)); | |
7621 | else | |
7622 | aarch64_set_reg_u64 (cpu, rd, NO_SP, aarch64_get_vec_u64 (cpu, rn, 1)); | |
7623 | } | |
7624 | ||
7625 | static void | |
7626 | dexSimpleFPIntegerConvert (sim_cpu *cpu) | |
7627 | { | |
7628 | /* instr[31] = size : 0 ==> 32 bit, 1 ==> 64 bit | |
7629 | instr[30 = 0 | |
7630 | instr[29] = S : 0 ==> OK, 1 ==> UNALLOC | |
7631 | instr[28,25] = 1111 | |
7632 | instr[24] = 0 | |
7633 | instr[23,22] = type : 00 ==> single, 01 ==> double, 1x ==> UNALLOC | |
7634 | instr[21] = 1 | |
7635 | instr[20,19] = rmode | |
7636 | instr[18,16] = opcode | |
7637 | instr[15,10] = 10 0000 */ | |
7638 | ||
7639 | uint32_t rmode_opcode; | |
7640 | uint32_t size_type; | |
7641 | uint32_t type; | |
7642 | uint32_t size; | |
7643 | uint32_t S; | |
7644 | ||
7645 | if (uimm (aarch64_get_instr (cpu), 31, 17) == 0x4F57) | |
7646 | { | |
7647 | float_vector_move (cpu); | |
7648 | return; | |
7649 | } | |
7650 | ||
7651 | size = uimm (aarch64_get_instr (cpu), 31, 31); | |
7652 | S = uimm (aarch64_get_instr (cpu), 29, 29); | |
7653 | if (S != 0) | |
7654 | HALT_UNALLOC; | |
7655 | ||
7656 | type = uimm (aarch64_get_instr (cpu), 23, 22); | |
7657 | if (type > 1) | |
7658 | HALT_UNALLOC; | |
7659 | ||
7660 | rmode_opcode = uimm (aarch64_get_instr (cpu), 20, 16); | |
7661 | size_type = (size << 1) | type; /* 0==32f, 1==32d, 2==64f, 3==64d. */ | |
7662 | ||
7663 | switch (rmode_opcode) | |
7664 | { | |
7665 | case 2: /* SCVTF. */ | |
7666 | switch (size_type) | |
7667 | { | |
7668 | case 0: scvtf32 (cpu); return; | |
7669 | case 1: scvtd32 (cpu); return; | |
7670 | case 2: scvtf (cpu); return; | |
7671 | case 3: scvtd (cpu); return; | |
7672 | default: | |
7673 | HALT_UNREACHABLE; | |
7674 | } | |
7675 | ||
7676 | case 6: /* FMOV GR, Vec. */ | |
7677 | switch (size_type) | |
7678 | { | |
7679 | case 0: gfmovs (cpu); return; | |
7680 | case 3: gfmovd (cpu); return; | |
7681 | default: HALT_UNALLOC; | |
7682 | } | |
7683 | ||
7684 | case 7: /* FMOV vec, GR. */ | |
7685 | switch (size_type) | |
7686 | { | |
7687 | case 0: fgmovs (cpu); return; | |
7688 | case 3: fgmovd (cpu); return; | |
7689 | default: HALT_UNALLOC; | |
7690 | } | |
7691 | ||
7692 | case 24: /* FCVTZS. */ | |
7693 | switch (size_type) | |
7694 | { | |
7695 | case 0: fcvtszs32 (cpu); return; | |
7696 | case 1: fcvtszd32 (cpu); return; | |
7697 | case 2: fcvtszs (cpu); return; | |
7698 | case 3: fcvtszd (cpu); return; | |
7699 | default: HALT_UNREACHABLE; | |
7700 | } | |
7701 | ||
7702 | case 25: do_fcvtzu (cpu); return; | |
7703 | case 3: do_UCVTF (cpu); return; | |
7704 | ||
7705 | case 0: /* FCVTNS. */ | |
7706 | case 1: /* FCVTNU. */ | |
7707 | case 4: /* FCVTAS. */ | |
7708 | case 5: /* FCVTAU. */ | |
7709 | case 8: /* FCVPTS. */ | |
7710 | case 9: /* FCVTPU. */ | |
7711 | case 16: /* FCVTMS. */ | |
7712 | case 17: /* FCVTMU. */ | |
7713 | default: | |
7714 | HALT_NYI; | |
7715 | } | |
7716 | } | |
7717 | ||
7718 | static void | |
7719 | set_flags_for_float_compare (sim_cpu *cpu, float fvalue1, float fvalue2) | |
7720 | { | |
7721 | uint32_t flags; | |
7722 | ||
7723 | if (isnan (fvalue1) || isnan (fvalue2)) | |
7724 | flags = C|V; | |
7725 | else | |
7726 | { | |
7727 | float result = fvalue1 - fvalue2; | |
7728 | ||
7729 | if (result == 0.0) | |
7730 | flags = Z|C; | |
7731 | else if (result < 0) | |
7732 | flags = N; | |
7733 | else /* (result > 0). */ | |
7734 | flags = C; | |
7735 | } | |
7736 | ||
7737 | aarch64_set_CPSR (cpu, flags); | |
7738 | } | |
7739 | ||
7740 | static void | |
7741 | fcmps (sim_cpu *cpu) | |
7742 | { | |
7743 | unsigned sm = uimm (aarch64_get_instr (cpu), 20, 16); | |
7744 | unsigned sn = uimm (aarch64_get_instr (cpu), 9, 5); | |
7745 | ||
7746 | float fvalue1 = aarch64_get_FP_float (cpu, sn); | |
7747 | float fvalue2 = aarch64_get_FP_float (cpu, sm); | |
7748 | ||
7749 | set_flags_for_float_compare (cpu, fvalue1, fvalue2); | |
7750 | } | |
7751 | ||
7752 | /* Float compare to zero -- Invalid Operation exception | |
7753 | only on signaling NaNs. */ | |
7754 | static void | |
7755 | fcmpzs (sim_cpu *cpu) | |
7756 | { | |
7757 | unsigned sn = uimm (aarch64_get_instr (cpu), 9, 5); | |
7758 | float fvalue1 = aarch64_get_FP_float (cpu, sn); | |
7759 | ||
7760 | set_flags_for_float_compare (cpu, fvalue1, 0.0f); | |
7761 | } | |
7762 | ||
7763 | /* Float compare -- Invalid Operation exception on all NaNs. */ | |
7764 | static void | |
7765 | fcmpes (sim_cpu *cpu) | |
7766 | { | |
7767 | unsigned sm = uimm (aarch64_get_instr (cpu), 20, 16); | |
7768 | unsigned sn = uimm (aarch64_get_instr (cpu), 9, 5); | |
7769 | ||
7770 | float fvalue1 = aarch64_get_FP_float (cpu, sn); | |
7771 | float fvalue2 = aarch64_get_FP_float (cpu, sm); | |
7772 | ||
7773 | set_flags_for_float_compare (cpu, fvalue1, fvalue2); | |
7774 | } | |
7775 | ||
7776 | /* Float compare to zero -- Invalid Operation exception on all NaNs. */ | |
7777 | static void | |
7778 | fcmpzes (sim_cpu *cpu) | |
7779 | { | |
7780 | unsigned sn = uimm (aarch64_get_instr (cpu), 9, 5); | |
7781 | float fvalue1 = aarch64_get_FP_float (cpu, sn); | |
7782 | ||
7783 | set_flags_for_float_compare (cpu, fvalue1, 0.0f); | |
7784 | } | |
7785 | ||
7786 | static void | |
7787 | set_flags_for_double_compare (sim_cpu *cpu, double dval1, double dval2) | |
7788 | { | |
7789 | uint32_t flags; | |
7790 | ||
7791 | if (isnan (dval1) || isnan (dval2)) | |
7792 | flags = C|V; | |
7793 | else | |
7794 | { | |
7795 | double result = dval1 - dval2; | |
7796 | ||
7797 | if (result == 0.0) | |
7798 | flags = Z|C; | |
7799 | else if (result < 0) | |
7800 | flags = N; | |
7801 | else /* (result > 0). */ | |
7802 | flags = C; | |
7803 | } | |
7804 | ||
7805 | aarch64_set_CPSR (cpu, flags); | |
7806 | } | |
7807 | ||
7808 | /* Double compare -- Invalid Operation exception only on signaling NaNs. */ | |
7809 | static void | |
7810 | fcmpd (sim_cpu *cpu) | |
7811 | { | |
7812 | unsigned sm = uimm (aarch64_get_instr (cpu), 20, 16); | |
7813 | unsigned sn = uimm (aarch64_get_instr (cpu), 9, 5); | |
7814 | ||
7815 | double dvalue1 = aarch64_get_FP_double (cpu, sn); | |
7816 | double dvalue2 = aarch64_get_FP_double (cpu, sm); | |
7817 | ||
7818 | set_flags_for_double_compare (cpu, dvalue1, dvalue2); | |
7819 | } | |
7820 | ||
7821 | /* Double compare to zero -- Invalid Operation exception | |
7822 | only on signaling NaNs. */ | |
7823 | static void | |
7824 | fcmpzd (sim_cpu *cpu) | |
7825 | { | |
7826 | unsigned sn = uimm (aarch64_get_instr (cpu), 9, 5); | |
7827 | double dvalue1 = aarch64_get_FP_double (cpu, sn); | |
7828 | ||
7829 | set_flags_for_double_compare (cpu, dvalue1, 0.0); | |
7830 | } | |
7831 | ||
7832 | /* Double compare -- Invalid Operation exception on all NaNs. */ | |
7833 | static void | |
7834 | fcmped (sim_cpu *cpu) | |
7835 | { | |
7836 | unsigned sm = uimm (aarch64_get_instr (cpu), 20, 16); | |
7837 | unsigned sn = uimm (aarch64_get_instr (cpu), 9, 5); | |
7838 | ||
7839 | double dvalue1 = aarch64_get_FP_double (cpu, sn); | |
7840 | double dvalue2 = aarch64_get_FP_double (cpu, sm); | |
7841 | ||
7842 | set_flags_for_double_compare (cpu, dvalue1, dvalue2); | |
7843 | } | |
7844 | ||
7845 | /* Double compare to zero -- Invalid Operation exception on all NaNs. */ | |
7846 | static void | |
7847 | fcmpzed (sim_cpu *cpu) | |
7848 | { | |
7849 | unsigned sn = uimm (aarch64_get_instr (cpu), 9, 5); | |
7850 | double dvalue1 = aarch64_get_FP_double (cpu, sn); | |
7851 | ||
7852 | set_flags_for_double_compare (cpu, dvalue1, 0.0); | |
7853 | } | |
7854 | ||
7855 | static void | |
7856 | dexSimpleFPCompare (sim_cpu *cpu) | |
7857 | { | |
7858 | /* assert instr[28,25] == 1111 | |
7859 | instr[30:24:21:13,10] = 0011000 | |
7860 | instr[31] = M : 0 ==> OK, 1 ==> UNALLOC | |
7861 | instr[29] ==> S : 0 ==> OK, 1 ==> UNALLOC | |
7862 | instr[23,22] ==> type : 0 ==> single, 01 ==> double, 1x ==> UNALLOC | |
7863 | instr[15,14] ==> op : 00 ==> OK, ow ==> UNALLOC | |
7864 | instr[4,0] ==> opcode2 : 00000 ==> FCMP, 10000 ==> FCMPE, | |
7865 | 01000 ==> FCMPZ, 11000 ==> FCMPEZ, | |
7866 | ow ==> UNALLOC */ | |
7867 | uint32_t dispatch; | |
7868 | uint32_t M_S = (uimm (aarch64_get_instr (cpu), 31, 31) << 1) | |
7869 | | uimm (aarch64_get_instr (cpu), 29, 29); | |
7870 | uint32_t type = uimm (aarch64_get_instr (cpu), 23, 22); | |
7871 | uint32_t op = uimm (aarch64_get_instr (cpu), 15, 14); | |
7872 | uint32_t op2_2_0 = uimm (aarch64_get_instr (cpu), 2, 0); | |
7873 | ||
7874 | if (op2_2_0 != 0) | |
7875 | HALT_UNALLOC; | |
7876 | ||
7877 | if (M_S != 0) | |
7878 | HALT_UNALLOC; | |
7879 | ||
7880 | if (type > 1) | |
7881 | HALT_UNALLOC; | |
7882 | ||
7883 | if (op != 0) | |
7884 | HALT_UNALLOC; | |
7885 | ||
7886 | /* dispatch on type and top 2 bits of opcode. */ | |
7887 | dispatch = (type << 2) | uimm (aarch64_get_instr (cpu), 4, 3); | |
7888 | ||
7889 | switch (dispatch) | |
7890 | { | |
7891 | case 0: fcmps (cpu); return; | |
7892 | case 1: fcmpzs (cpu); return; | |
7893 | case 2: fcmpes (cpu); return; | |
7894 | case 3: fcmpzes (cpu); return; | |
7895 | case 4: fcmpd (cpu); return; | |
7896 | case 5: fcmpzd (cpu); return; | |
7897 | case 6: fcmped (cpu); return; | |
7898 | case 7: fcmpzed (cpu); return; | |
7899 | default: HALT_UNREACHABLE; | |
7900 | } | |
7901 | } | |
7902 | ||
7903 | static void | |
7904 | do_scalar_FADDP (sim_cpu *cpu) | |
7905 | { | |
7906 | /* instr [31,23] = 011111100 | |
7907 | instr [22] = single(0)/double(1) | |
7908 | instr [21,10] = 1100 0011 0110 | |
7909 | instr [9,5] = Fn | |
7910 | instr [4,0] = Fd. */ | |
7911 | ||
7912 | unsigned Fn = uimm (aarch64_get_instr (cpu), 9, 5); | |
7913 | unsigned Fd = uimm (aarch64_get_instr (cpu), 4, 0); | |
7914 | ||
7915 | NYI_assert (31, 23, 0x0FC); | |
7916 | NYI_assert (21, 10, 0xC36); | |
7917 | ||
7918 | if (uimm (aarch64_get_instr (cpu), 22, 22)) | |
7919 | { | |
7920 | double val1 = aarch64_get_vec_double (cpu, Fn, 0); | |
7921 | double val2 = aarch64_get_vec_double (cpu, Fn, 1); | |
7922 | ||
7923 | aarch64_set_FP_double (cpu, Fd, val1 + val2); | |
7924 | } | |
7925 | else | |
7926 | { | |
7927 | float val1 = aarch64_get_vec_float (cpu, Fn, 0); | |
7928 | float val2 = aarch64_get_vec_float (cpu, Fn, 1); | |
7929 | ||
7930 | aarch64_set_FP_float (cpu, Fd, val1 + val2); | |
7931 | } | |
7932 | } | |
7933 | ||
7934 | /* Floating point absolute difference. */ | |
7935 | ||
7936 | static void | |
7937 | do_scalar_FABD (sim_cpu *cpu) | |
7938 | { | |
7939 | /* instr [31,23] = 0111 1110 1 | |
7940 | instr [22] = float(0)/double(1) | |
7941 | instr [21] = 1 | |
7942 | instr [20,16] = Rm | |
7943 | instr [15,10] = 1101 01 | |
7944 | instr [9, 5] = Rn | |
7945 | instr [4, 0] = Rd. */ | |
7946 | ||
7947 | unsigned rm = uimm (aarch64_get_instr (cpu), 20, 16); | |
7948 | unsigned rn = uimm (aarch64_get_instr (cpu), 9, 5); | |
7949 | unsigned rd = uimm (aarch64_get_instr (cpu), 4, 0); | |
7950 | ||
7951 | NYI_assert (31, 23, 0x0FD); | |
7952 | NYI_assert (21, 21, 1); | |
7953 | NYI_assert (15, 10, 0x35); | |
7954 | ||
7955 | if (uimm (aarch64_get_instr (cpu), 22, 22)) | |
7956 | aarch64_set_FP_double (cpu, rd, | |
7957 | fabs (aarch64_get_FP_double (cpu, rn) | |
7958 | - aarch64_get_FP_double (cpu, rm))); | |
7959 | else | |
7960 | aarch64_set_FP_float (cpu, rd, | |
7961 | fabsf (aarch64_get_FP_float (cpu, rn) | |
7962 | - aarch64_get_FP_float (cpu, rm))); | |
7963 | } | |
7964 | ||
7965 | static void | |
7966 | do_scalar_CMGT (sim_cpu *cpu) | |
7967 | { | |
7968 | /* instr [31,21] = 0101 1110 111 | |
7969 | instr [20,16] = Rm | |
7970 | instr [15,10] = 00 1101 | |
7971 | instr [9, 5] = Rn | |
7972 | instr [4, 0] = Rd. */ | |
7973 | ||
7974 | unsigned rm = uimm (aarch64_get_instr (cpu), 20, 16); | |
7975 | unsigned rn = uimm (aarch64_get_instr (cpu), 9, 5); | |
7976 | unsigned rd = uimm (aarch64_get_instr (cpu), 4, 0); | |
7977 | ||
7978 | NYI_assert (31, 21, 0x2F7); | |
7979 | NYI_assert (15, 10, 0x0D); | |
7980 | ||
7981 | aarch64_set_vec_u64 (cpu, rd, 0, | |
7982 | aarch64_get_vec_u64 (cpu, rn, 0) > | |
7983 | aarch64_get_vec_u64 (cpu, rm, 0) ? -1L : 0L); | |
7984 | } | |
7985 | ||
7986 | static void | |
7987 | do_scalar_USHR (sim_cpu *cpu) | |
7988 | { | |
7989 | /* instr [31,23] = 0111 1111 0 | |
7990 | instr [22,16] = shift amount | |
7991 | instr [15,10] = 0000 01 | |
7992 | instr [9, 5] = Rn | |
7993 | instr [4, 0] = Rd. */ | |
7994 | ||
7995 | unsigned amount = 128 - uimm (aarch64_get_instr (cpu), 22, 16); | |
7996 | unsigned rn = uimm (aarch64_get_instr (cpu), 9, 5); | |
7997 | unsigned rd = uimm (aarch64_get_instr (cpu), 4, 0); | |
7998 | ||
7999 | NYI_assert (31, 23, 0x0FE); | |
8000 | NYI_assert (15, 10, 0x01); | |
8001 | ||
8002 | aarch64_set_vec_u64 (cpu, rd, 0, | |
8003 | aarch64_get_vec_u64 (cpu, rn, 0) >> amount); | |
8004 | } | |
8005 | ||
8006 | static void | |
8007 | do_scalar_SHL (sim_cpu *cpu) | |
8008 | { | |
8009 | /* instr [31,23] = 0111 1101 0 | |
8010 | instr [22,16] = shift amount | |
8011 | instr [15,10] = 0101 01 | |
8012 | instr [9, 5] = Rn | |
8013 | instr [4, 0] = Rd. */ | |
8014 | ||
8015 | unsigned amount = uimm (aarch64_get_instr (cpu), 22, 16) - 64; | |
8016 | unsigned rn = uimm (aarch64_get_instr (cpu), 9, 5); | |
8017 | unsigned rd = uimm (aarch64_get_instr (cpu), 4, 0); | |
8018 | ||
8019 | NYI_assert (31, 23, 0x0BE); | |
8020 | NYI_assert (15, 10, 0x15); | |
8021 | ||
8022 | if (uimm (aarch64_get_instr (cpu), 22, 22) == 0) | |
8023 | HALT_UNALLOC; | |
8024 | ||
8025 | aarch64_set_vec_u64 (cpu, rd, 0, | |
8026 | aarch64_get_vec_u64 (cpu, rn, 0) << amount); | |
8027 | } | |
8028 | ||
8029 | /* FCMEQ FCMGT FCMGE. */ | |
8030 | static void | |
8031 | do_scalar_FCM (sim_cpu *cpu) | |
8032 | { | |
8033 | /* instr [31,30] = 01 | |
8034 | instr [29] = U | |
8035 | instr [28,24] = 1 1110 | |
8036 | instr [23] = E | |
8037 | instr [22] = size | |
8038 | instr [21] = 1 | |
8039 | instr [20,16] = Rm | |
8040 | instr [15,12] = 1110 | |
8041 | instr [11] = AC | |
8042 | instr [10] = 1 | |
8043 | instr [9, 5] = Rn | |
8044 | instr [4, 0] = Rd. */ | |
8045 | ||
8046 | unsigned rm = uimm (aarch64_get_instr (cpu), 20, 16); | |
8047 | unsigned rn = uimm (aarch64_get_instr (cpu), 9, 5); | |
8048 | unsigned rd = uimm (aarch64_get_instr (cpu), 4, 0); | |
8049 | unsigned EUac = (uimm (aarch64_get_instr (cpu), 23, 23) << 2) | |
8050 | | (uimm (aarch64_get_instr (cpu), 29, 29) << 1) | |
8051 | | uimm (aarch64_get_instr (cpu), 11, 11); | |
8052 | unsigned result; | |
8053 | float val1; | |
8054 | float val2; | |
8055 | ||
8056 | NYI_assert (31, 30, 1); | |
8057 | NYI_assert (28, 24, 0x1E); | |
8058 | NYI_assert (21, 21, 1); | |
8059 | NYI_assert (15, 12, 0xE); | |
8060 | NYI_assert (10, 10, 1); | |
8061 | ||
8062 | if (uimm (aarch64_get_instr (cpu), 22, 22)) | |
8063 | { | |
8064 | double val1 = aarch64_get_FP_double (cpu, rn); | |
8065 | double val2 = aarch64_get_FP_double (cpu, rm); | |
8066 | ||
8067 | switch (EUac) | |
8068 | { | |
8069 | case 0: /* 000 */ | |
8070 | result = val1 == val2; | |
8071 | break; | |
8072 | ||
8073 | case 3: /* 011 */ | |
8074 | val1 = fabs (val1); | |
8075 | val2 = fabs (val2); | |
8076 | /* Fall through. */ | |
8077 | case 2: /* 010 */ | |
8078 | result = val1 >= val2; | |
8079 | break; | |
8080 | ||
8081 | case 7: /* 111 */ | |
8082 | val1 = fabs (val1); | |
8083 | val2 = fabs (val2); | |
8084 | /* Fall through. */ | |
8085 | case 6: /* 110 */ | |
8086 | result = val1 > val2; | |
8087 | break; | |
8088 | ||
8089 | default: | |
8090 | HALT_UNALLOC; | |
8091 | } | |
8092 | ||
8093 | aarch64_set_vec_u32 (cpu, rd, 0, result ? -1 : 0); | |
8094 | return; | |
8095 | } | |
8096 | ||
8097 | val1 = aarch64_get_FP_float (cpu, rn); | |
8098 | val2 = aarch64_get_FP_float (cpu, rm); | |
8099 | ||
8100 | switch (EUac) | |
8101 | { | |
8102 | case 0: /* 000 */ | |
8103 | result = val1 == val2; | |
8104 | break; | |
8105 | ||
8106 | case 3: /* 011 */ | |
8107 | val1 = fabsf (val1); | |
8108 | val2 = fabsf (val2); | |
8109 | /* Fall through. */ | |
8110 | case 2: /* 010 */ | |
8111 | result = val1 >= val2; | |
8112 | break; | |
8113 | ||
8114 | case 7: /* 111 */ | |
8115 | val1 = fabsf (val1); | |
8116 | val2 = fabsf (val2); | |
8117 | /* Fall through. */ | |
8118 | case 6: /* 110 */ | |
8119 | result = val1 > val2; | |
8120 | break; | |
8121 | ||
8122 | default: | |
8123 | HALT_UNALLOC; | |
8124 | } | |
8125 | ||
8126 | aarch64_set_vec_u32 (cpu, rd, 0, result ? -1 : 0); | |
8127 | } | |
8128 | ||
8129 | /* An alias of DUP. */ | |
8130 | static void | |
8131 | do_scalar_MOV (sim_cpu *cpu) | |
8132 | { | |
8133 | /* instr [31,21] = 0101 1110 000 | |
8134 | instr [20,16] = imm5 | |
8135 | instr [15,10] = 0000 01 | |
8136 | instr [9, 5] = Rn | |
8137 | instr [4, 0] = Rd. */ | |
8138 | ||
8139 | unsigned rn = uimm (aarch64_get_instr (cpu), 9, 5); | |
8140 | unsigned rd = uimm (aarch64_get_instr (cpu), 4, 0); | |
8141 | unsigned index; | |
8142 | ||
8143 | NYI_assert (31, 21, 0x2F0); | |
8144 | NYI_assert (15, 10, 0x01); | |
8145 | ||
8146 | if (uimm (aarch64_get_instr (cpu), 16, 16)) | |
8147 | { | |
8148 | /* 8-bit. */ | |
8149 | index = uimm (aarch64_get_instr (cpu), 20, 17); | |
8150 | aarch64_set_vec_u8 | |
8151 | (cpu, rd, 0, aarch64_get_vec_u8 (cpu, rn, index)); | |
8152 | } | |
8153 | else if (uimm (aarch64_get_instr (cpu), 17, 17)) | |
8154 | { | |
8155 | /* 16-bit. */ | |
8156 | index = uimm (aarch64_get_instr (cpu), 20, 18); | |
8157 | aarch64_set_vec_u16 | |
8158 | (cpu, rd, 0, aarch64_get_vec_u16 (cpu, rn, index)); | |
8159 | } | |
8160 | else if (uimm (aarch64_get_instr (cpu), 18, 18)) | |
8161 | { | |
8162 | /* 32-bit. */ | |
8163 | index = uimm (aarch64_get_instr (cpu), 20, 19); | |
8164 | aarch64_set_vec_u32 | |
8165 | (cpu, rd, 0, aarch64_get_vec_u32 (cpu, rn, index)); | |
8166 | } | |
8167 | else if (uimm (aarch64_get_instr (cpu), 19, 19)) | |
8168 | { | |
8169 | /* 64-bit. */ | |
8170 | index = uimm (aarch64_get_instr (cpu), 20, 20); | |
8171 | aarch64_set_vec_u64 | |
8172 | (cpu, rd, 0, aarch64_get_vec_u64 (cpu, rn, index)); | |
8173 | } | |
8174 | else | |
8175 | HALT_UNALLOC; | |
8176 | } | |
8177 | ||
8178 | static void | |
8179 | do_double_add (sim_cpu *cpu) | |
8180 | { | |
8181 | /* instr [28,25] = 1111. */ | |
8182 | unsigned Fd; | |
8183 | unsigned Fm; | |
8184 | unsigned Fn; | |
8185 | double val1; | |
8186 | double val2; | |
8187 | ||
8188 | switch (uimm (aarch64_get_instr (cpu), 31, 23)) | |
8189 | { | |
8190 | case 0xBC: | |
8191 | switch (uimm (aarch64_get_instr (cpu), 15, 10)) | |
8192 | { | |
8193 | case 0x01: do_scalar_MOV (cpu); return; | |
8194 | case 0x39: do_scalar_FCM (cpu); return; | |
8195 | case 0x3B: do_scalar_FCM (cpu); return; | |
8196 | } | |
8197 | break; | |
8198 | ||
8199 | case 0xBE: do_scalar_SHL (cpu); return; | |
8200 | ||
8201 | case 0xFC: | |
8202 | switch (uimm (aarch64_get_instr (cpu), 15, 10)) | |
8203 | { | |
8204 | case 0x36: do_scalar_FADDP (cpu); return; | |
8205 | case 0x39: do_scalar_FCM (cpu); return; | |
8206 | case 0x3B: do_scalar_FCM (cpu); return; | |
8207 | } | |
8208 | break; | |
8209 | ||
8210 | case 0xFD: | |
8211 | switch (uimm (aarch64_get_instr (cpu), 15, 10)) | |
8212 | { | |
8213 | case 0x0D: do_scalar_CMGT (cpu); return; | |
8214 | case 0x35: do_scalar_FABD (cpu); return; | |
8215 | case 0x39: do_scalar_FCM (cpu); return; | |
8216 | case 0x3B: do_scalar_FCM (cpu); return; | |
8217 | default: | |
8218 | HALT_NYI; | |
8219 | } | |
8220 | ||
8221 | case 0xFE: do_scalar_USHR (cpu); return; | |
8222 | default: | |
8223 | break; | |
8224 | } | |
8225 | ||
8226 | /* instr [31,21] = 0101 1110 111 | |
8227 | instr [20,16] = Fn | |
8228 | instr [15,10] = 1000 01 | |
8229 | instr [9,5] = Fm | |
8230 | instr [4,0] = Fd. */ | |
8231 | if (uimm (aarch64_get_instr (cpu), 31, 21) != 0x2F7 | |
8232 | || uimm (aarch64_get_instr (cpu), 15, 10) != 0x21) | |
8233 | HALT_NYI; | |
8234 | ||
8235 | Fd = uimm (aarch64_get_instr (cpu), 4, 0); | |
8236 | Fm = uimm (aarch64_get_instr (cpu), 9, 5); | |
8237 | Fn = uimm (aarch64_get_instr (cpu), 20, 16); | |
8238 | ||
8239 | val1 = aarch64_get_FP_double (cpu, Fm); | |
8240 | val2 = aarch64_get_FP_double (cpu, Fn); | |
8241 | ||
8242 | aarch64_set_FP_double (cpu, Fd, val1 + val2); | |
8243 | } | |
8244 | ||
8245 | static void | |
8246 | dexAdvSIMD1 (sim_cpu *cpu) | |
8247 | { | |
8248 | /* instr [28,25] = 1 111. */ | |
8249 | ||
8250 | /* we are currently only interested in the basic | |
8251 | scalar fp routines which all have bit 30 = 0. */ | |
8252 | if (uimm (aarch64_get_instr (cpu), 30, 30)) | |
8253 | do_double_add (cpu); | |
8254 | ||
8255 | /* instr[24] is set for FP data processing 3-source and clear for | |
8256 | all other basic scalar fp instruction groups. */ | |
8257 | else if (uimm (aarch64_get_instr (cpu), 24, 24)) | |
8258 | dexSimpleFPDataProc3Source (cpu); | |
8259 | ||
8260 | /* instr[21] is clear for floating <-> fixed conversions and set for | |
8261 | all other basic scalar fp instruction groups. */ | |
8262 | else if (!uimm (aarch64_get_instr (cpu), 21, 21)) | |
8263 | dexSimpleFPFixedConvert (cpu); | |
8264 | ||
8265 | /* instr[11,10] : 01 ==> cond compare, 10 ==> Data Proc 2 Source | |
8266 | 11 ==> cond select, 00 ==> other. */ | |
8267 | else | |
8268 | switch (uimm (aarch64_get_instr (cpu), 11, 10)) | |
8269 | { | |
8270 | case 1: dexSimpleFPCondCompare (cpu); return; | |
8271 | case 2: dexSimpleFPDataProc2Source (cpu); return; | |
8272 | case 3: dexSimpleFPCondSelect (cpu); return; | |
8273 | ||
8274 | default: | |
8275 | /* Now an ordered cascade of tests. | |
8276 | FP immediate has aarch64_get_instr (cpu)[12] == 1. | |
8277 | FP compare has aarch64_get_instr (cpu)[13] == 1. | |
8278 | FP Data Proc 1 Source has aarch64_get_instr (cpu)[14] == 1. | |
8279 | FP floating <--> integer conversions has aarch64_get_instr (cpu)[15] == 0. */ | |
8280 | if (uimm (aarch64_get_instr (cpu), 12, 12)) | |
8281 | dexSimpleFPImmediate (cpu); | |
8282 | ||
8283 | else if (uimm (aarch64_get_instr (cpu), 13, 13)) | |
8284 | dexSimpleFPCompare (cpu); | |
8285 | ||
8286 | else if (uimm (aarch64_get_instr (cpu), 14, 14)) | |
8287 | dexSimpleFPDataProc1Source (cpu); | |
8288 | ||
8289 | else if (!uimm (aarch64_get_instr (cpu), 15, 15)) | |
8290 | dexSimpleFPIntegerConvert (cpu); | |
8291 | ||
8292 | else | |
8293 | /* If we get here then instr[15] == 1 which means UNALLOC. */ | |
8294 | HALT_UNALLOC; | |
8295 | } | |
8296 | } | |
8297 | ||
8298 | /* PC relative addressing. */ | |
8299 | ||
8300 | static void | |
8301 | pcadr (sim_cpu *cpu) | |
8302 | { | |
8303 | /* instr[31] = op : 0 ==> ADR, 1 ==> ADRP | |
8304 | instr[30,29] = immlo | |
8305 | instr[23,5] = immhi. */ | |
8306 | uint64_t address; | |
8307 | unsigned rd = uimm (aarch64_get_instr (cpu), 4, 0); | |
8308 | uint32_t isPage = uimm (aarch64_get_instr (cpu), 31, 31); | |
8309 | union { int64_t u64; uint64_t s64; } imm; | |
8310 | uint64_t offset; | |
8311 | ||
8312 | imm.s64 = simm64 (aarch64_get_instr (cpu), 23, 5); | |
8313 | offset = imm.u64; | |
8314 | offset = (offset << 2) | uimm (aarch64_get_instr (cpu), 30, 29); | |
8315 | ||
8316 | address = aarch64_get_PC (cpu); | |
8317 | ||
8318 | if (isPage) | |
8319 | { | |
8320 | offset <<= 12; | |
8321 | address &= ~0xfff; | |
8322 | } | |
8323 | ||
8324 | aarch64_set_reg_u64 (cpu, rd, NO_SP, address + offset); | |
8325 | } | |
8326 | ||
8327 | /* Specific decode and execute for group Data Processing Immediate. */ | |
8328 | ||
8329 | static void | |
8330 | dexPCRelAddressing (sim_cpu *cpu) | |
8331 | { | |
8332 | /* assert instr[28,24] = 10000. */ | |
8333 | pcadr (cpu); | |
8334 | } | |
8335 | ||
8336 | /* Immediate logical. | |
8337 | The bimm32/64 argument is constructed by replicating a 2, 4, 8, | |
8338 | 16, 32 or 64 bit sequence pulled out at decode and possibly | |
8339 | inverting it.. | |
8340 | ||
8341 | N.B. the output register (dest) can normally be Xn or SP | |
8342 | the exception occurs for flag setting instructions which may | |
8343 | only use Xn for the output (dest). The input register can | |
8344 | never be SP. */ | |
8345 | ||
8346 | /* 32 bit and immediate. */ | |
8347 | static void | |
8348 | and32 (sim_cpu *cpu, uint32_t bimm) | |
8349 | { | |
8350 | unsigned rn = uimm (aarch64_get_instr (cpu), 9, 5); | |
8351 | unsigned rd = uimm (aarch64_get_instr (cpu), 4, 0); | |
8352 | ||
8353 | aarch64_set_reg_u64 (cpu, rd, SP_OK, | |
8354 | aarch64_get_reg_u32 (cpu, rn, NO_SP) & bimm); | |
8355 | } | |
8356 | ||
8357 | /* 64 bit and immediate. */ | |
8358 | static void | |
8359 | and64 (sim_cpu *cpu, uint64_t bimm) | |
8360 | { | |
8361 | unsigned rn = uimm (aarch64_get_instr (cpu), 9, 5); | |
8362 | unsigned rd = uimm (aarch64_get_instr (cpu), 4, 0); | |
8363 | ||
8364 | aarch64_set_reg_u64 (cpu, rd, SP_OK, | |
8365 | aarch64_get_reg_u64 (cpu, rn, NO_SP) & bimm); | |
8366 | } | |
8367 | ||
8368 | /* 32 bit and immediate set flags. */ | |
8369 | static void | |
8370 | ands32 (sim_cpu *cpu, uint32_t bimm) | |
8371 | { | |
8372 | unsigned rn = uimm (aarch64_get_instr (cpu), 9, 5); | |
8373 | unsigned rd = uimm (aarch64_get_instr (cpu), 4, 0); | |
8374 | ||
8375 | uint32_t value1 = aarch64_get_reg_u32 (cpu, rn, NO_SP); | |
8376 | uint32_t value2 = bimm; | |
8377 | ||
8378 | aarch64_set_reg_u64 (cpu, rd, NO_SP, value1 & value2); | |
8379 | set_flags_for_binop32 (cpu, value1 & value2); | |
8380 | } | |
8381 | ||
8382 | /* 64 bit and immediate set flags. */ | |
8383 | static void | |
8384 | ands64 (sim_cpu *cpu, uint64_t bimm) | |
8385 | { | |
8386 | unsigned rn = uimm (aarch64_get_instr (cpu), 9, 5); | |
8387 | unsigned rd = uimm (aarch64_get_instr (cpu), 4, 0); | |
8388 | ||
8389 | uint64_t value1 = aarch64_get_reg_u64 (cpu, rn, NO_SP); | |
8390 | uint64_t value2 = bimm; | |
8391 | ||
8392 | aarch64_set_reg_u64 (cpu, rd, NO_SP, value1 & value2); | |
8393 | set_flags_for_binop64 (cpu, value1 & value2); | |
8394 | } | |
8395 | ||
8396 | /* 32 bit exclusive or immediate. */ | |
8397 | static void | |
8398 | eor32 (sim_cpu *cpu, uint32_t bimm) | |
8399 | { | |
8400 | unsigned rn = uimm (aarch64_get_instr (cpu), 9, 5); | |
8401 | unsigned rd = uimm (aarch64_get_instr (cpu), 4, 0); | |
8402 | ||
8403 | aarch64_set_reg_u64 (cpu, rd, SP_OK, | |
8404 | aarch64_get_reg_u32 (cpu, rn, NO_SP) ^ bimm); | |
8405 | } | |
8406 | ||
8407 | /* 64 bit exclusive or immediate. */ | |
8408 | static void | |
8409 | eor64 (sim_cpu *cpu, uint64_t bimm) | |
8410 | { | |
8411 | unsigned rn = uimm (aarch64_get_instr (cpu), 9, 5); | |
8412 | unsigned rd = uimm (aarch64_get_instr (cpu), 4, 0); | |
8413 | ||
8414 | aarch64_set_reg_u64 (cpu, rd, SP_OK, | |
8415 | aarch64_get_reg_u64 (cpu, rn, NO_SP) ^ bimm); | |
8416 | } | |
8417 | ||
8418 | /* 32 bit or immediate. */ | |
8419 | static void | |
8420 | orr32 (sim_cpu *cpu, uint32_t bimm) | |
8421 | { | |
8422 | unsigned rn = uimm (aarch64_get_instr (cpu), 9, 5); | |
8423 | unsigned rd = uimm (aarch64_get_instr (cpu), 4, 0); | |
8424 | ||
8425 | aarch64_set_reg_u64 (cpu, rd, SP_OK, | |
8426 | aarch64_get_reg_u32 (cpu, rn, NO_SP) | bimm); | |
8427 | } | |
8428 | ||
8429 | /* 64 bit or immediate. */ | |
8430 | static void | |
8431 | orr64 (sim_cpu *cpu, uint64_t bimm) | |
8432 | { | |
8433 | unsigned rn = uimm (aarch64_get_instr (cpu), 9, 5); | |
8434 | unsigned rd = uimm (aarch64_get_instr (cpu), 4, 0); | |
8435 | ||
8436 | aarch64_set_reg_u64 (cpu, rd, SP_OK, | |
8437 | aarch64_get_reg_u64 (cpu, rn, NO_SP) | bimm); | |
8438 | } | |
8439 | ||
8440 | /* Logical shifted register. | |
8441 | These allow an optional LSL, ASR, LSR or ROR to the second source | |
8442 | register with a count up to the register bit count. | |
8443 | N.B register args may not be SP. */ | |
8444 | ||
8445 | /* 32 bit AND shifted register. */ | |
8446 | static void | |
8447 | and32_shift (sim_cpu *cpu, Shift shift, uint32_t count) | |
8448 | { | |
8449 | unsigned rm = uimm (aarch64_get_instr (cpu), 20, 16); | |
8450 | unsigned rn = uimm (aarch64_get_instr (cpu), 9, 5); | |
8451 | unsigned rd = uimm (aarch64_get_instr (cpu), 4, 0); | |
8452 | ||
8453 | aarch64_set_reg_u64 | |
8454 | (cpu, rd, NO_SP, aarch64_get_reg_u32 (cpu, rn, NO_SP) | |
8455 | & shifted32 (aarch64_get_reg_u32 (cpu, rm, NO_SP), shift, count)); | |
8456 | } | |
8457 | ||
8458 | /* 64 bit AND shifted register. */ | |
8459 | static void | |
8460 | and64_shift (sim_cpu *cpu, Shift shift, uint32_t count) | |
8461 | { | |
8462 | unsigned rm = uimm (aarch64_get_instr (cpu), 20, 16); | |
8463 | unsigned rn = uimm (aarch64_get_instr (cpu), 9, 5); | |
8464 | unsigned rd = uimm (aarch64_get_instr (cpu), 4, 0); | |
8465 | ||
8466 | aarch64_set_reg_u64 | |
8467 | (cpu, rd, NO_SP, aarch64_get_reg_u64 (cpu, rn, NO_SP) | |
8468 | & shifted64 (aarch64_get_reg_u64 (cpu, rm, NO_SP), shift, count)); | |
8469 | } | |
8470 | ||
8471 | /* 32 bit AND shifted register setting flags. */ | |
8472 | static void | |
8473 | ands32_shift (sim_cpu *cpu, Shift shift, uint32_t count) | |
8474 | { | |
8475 | unsigned rm = uimm (aarch64_get_instr (cpu), 20, 16); | |
8476 | unsigned rn = uimm (aarch64_get_instr (cpu), 9, 5); | |
8477 | unsigned rd = uimm (aarch64_get_instr (cpu), 4, 0); | |
8478 | ||
8479 | uint32_t value1 = aarch64_get_reg_u32 (cpu, rn, NO_SP); | |
8480 | uint32_t value2 = shifted32 (aarch64_get_reg_u32 (cpu, rm, NO_SP), | |
8481 | shift, count); | |
8482 | ||
8483 | aarch64_set_reg_u64 (cpu, rd, NO_SP, value1 & value2); | |
8484 | set_flags_for_binop32 (cpu, value1 & value2); | |
8485 | } | |
8486 | ||
8487 | /* 64 bit AND shifted register setting flags. */ | |
8488 | static void | |
8489 | ands64_shift (sim_cpu *cpu, Shift shift, uint32_t count) | |
8490 | { | |
8491 | unsigned rm = uimm (aarch64_get_instr (cpu), 20, 16); | |
8492 | unsigned rn = uimm (aarch64_get_instr (cpu), 9, 5); | |
8493 | unsigned rd = uimm (aarch64_get_instr (cpu), 4, 0); | |
8494 | ||
8495 | uint64_t value1 = aarch64_get_reg_u64 (cpu, rn, NO_SP); | |
8496 | uint64_t value2 = shifted64 (aarch64_get_reg_u64 (cpu, rm, NO_SP), | |
8497 | shift, count); | |
8498 | ||
8499 | aarch64_set_reg_u64 (cpu, rd, NO_SP, value1 & value2); | |
8500 | set_flags_for_binop64 (cpu, value1 & value2); | |
8501 | } | |
8502 | ||
8503 | /* 32 bit BIC shifted register. */ | |
8504 | static void | |
8505 | bic32_shift (sim_cpu *cpu, Shift shift, uint32_t count) | |
8506 | { | |
8507 | unsigned rm = uimm (aarch64_get_instr (cpu), 20, 16); | |
8508 | unsigned rn = uimm (aarch64_get_instr (cpu), 9, 5); | |
8509 | unsigned rd = uimm (aarch64_get_instr (cpu), 4, 0); | |
8510 | ||
8511 | aarch64_set_reg_u64 | |
8512 | (cpu, rd, NO_SP, aarch64_get_reg_u32 (cpu, rn, NO_SP) | |
8513 | & ~ shifted32 (aarch64_get_reg_u32 (cpu, rm, NO_SP), shift, count)); | |
8514 | } | |
8515 | ||
8516 | /* 64 bit BIC shifted register. */ | |
8517 | static void | |
8518 | bic64_shift (sim_cpu *cpu, Shift shift, uint32_t count) | |
8519 | { | |
8520 | unsigned rm = uimm (aarch64_get_instr (cpu), 20, 16); | |
8521 | unsigned rn = uimm (aarch64_get_instr (cpu), 9, 5); | |
8522 | unsigned rd = uimm (aarch64_get_instr (cpu), 4, 0); | |
8523 | ||
8524 | aarch64_set_reg_u64 | |
8525 | (cpu, rd, NO_SP, aarch64_get_reg_u64 (cpu, rn, NO_SP) | |
8526 | & ~ shifted64 (aarch64_get_reg_u64 (cpu, rm, NO_SP), shift, count)); | |
8527 | } | |
8528 | ||
8529 | /* 32 bit BIC shifted register setting flags. */ | |
8530 | static void | |
8531 | bics32_shift (sim_cpu *cpu, Shift shift, uint32_t count) | |
8532 | { | |
8533 | unsigned rm = uimm (aarch64_get_instr (cpu), 20, 16); | |
8534 | unsigned rn = uimm (aarch64_get_instr (cpu), 9, 5); | |
8535 | unsigned rd = uimm (aarch64_get_instr (cpu), 4, 0); | |
8536 | ||
8537 | uint32_t value1 = aarch64_get_reg_u32 (cpu, rn, NO_SP); | |
8538 | uint32_t value2 = ~ shifted32 (aarch64_get_reg_u32 (cpu, rm, NO_SP), | |
8539 | shift, count); | |
8540 | ||
8541 | aarch64_set_reg_u64 (cpu, rd, NO_SP, value1 & value2); | |
8542 | set_flags_for_binop32 (cpu, value1 & value2); | |
8543 | } | |
8544 | ||
8545 | /* 64 bit BIC shifted register setting flags. */ | |
8546 | static void | |
8547 | bics64_shift (sim_cpu *cpu, Shift shift, uint32_t count) | |
8548 | { | |
8549 | unsigned rm = uimm (aarch64_get_instr (cpu), 20, 16); | |
8550 | unsigned rn = uimm (aarch64_get_instr (cpu), 9, 5); | |
8551 | unsigned rd = uimm (aarch64_get_instr (cpu), 4, 0); | |
8552 | ||
8553 | uint64_t value1 = aarch64_get_reg_u64 (cpu, rn, NO_SP); | |
8554 | uint64_t value2 = ~ shifted64 (aarch64_get_reg_u64 (cpu, rm, NO_SP), | |
8555 | shift, count); | |
8556 | ||
8557 | aarch64_set_reg_u64 (cpu, rd, NO_SP, value1 & value2); | |
8558 | set_flags_for_binop64 (cpu, value1 & value2); | |
8559 | } | |
8560 | ||
8561 | /* 32 bit EON shifted register. */ | |
8562 | static void | |
8563 | eon32_shift (sim_cpu *cpu, Shift shift, uint32_t count) | |
8564 | { | |
8565 | unsigned rm = uimm (aarch64_get_instr (cpu), 20, 16); | |
8566 | unsigned rn = uimm (aarch64_get_instr (cpu), 9, 5); | |
8567 | unsigned rd = uimm (aarch64_get_instr (cpu), 4, 0); | |
8568 | ||
8569 | aarch64_set_reg_u64 | |
8570 | (cpu, rd, NO_SP, aarch64_get_reg_u32 (cpu, rn, NO_SP) | |
8571 | ^ ~ shifted32 (aarch64_get_reg_u32 (cpu, rm, NO_SP), shift, count)); | |
8572 | } | |
8573 | ||
8574 | /* 64 bit EON shifted register. */ | |
8575 | static void | |
8576 | eon64_shift (sim_cpu *cpu, Shift shift, uint32_t count) | |
8577 | { | |
8578 | unsigned rm = uimm (aarch64_get_instr (cpu), 20, 16); | |
8579 | unsigned rn = uimm (aarch64_get_instr (cpu), 9, 5); | |
8580 | unsigned rd = uimm (aarch64_get_instr (cpu), 4, 0); | |
8581 | ||
8582 | aarch64_set_reg_u64 | |
8583 | (cpu, rd, NO_SP, aarch64_get_reg_u64 (cpu, rn, NO_SP) | |
8584 | ^ ~ shifted64 (aarch64_get_reg_u64 (cpu, rm, NO_SP), shift, count)); | |
8585 | } | |
8586 | ||
8587 | /* 32 bit EOR shifted register. */ | |
8588 | static void | |
8589 | eor32_shift (sim_cpu *cpu, Shift shift, uint32_t count) | |
8590 | { | |
8591 | unsigned rm = uimm (aarch64_get_instr (cpu), 20, 16); | |
8592 | unsigned rn = uimm (aarch64_get_instr (cpu), 9, 5); | |
8593 | unsigned rd = uimm (aarch64_get_instr (cpu), 4, 0); | |
8594 | ||
8595 | aarch64_set_reg_u64 | |
8596 | (cpu, rd, NO_SP, aarch64_get_reg_u32 (cpu, rn, NO_SP) | |
8597 | ^ shifted32 (aarch64_get_reg_u32 (cpu, rm, NO_SP), shift, count)); | |
8598 | } | |
8599 | ||
8600 | /* 64 bit EOR shifted register. */ | |
8601 | static void | |
8602 | eor64_shift (sim_cpu *cpu, Shift shift, uint32_t count) | |
8603 | { | |
8604 | unsigned rm = uimm (aarch64_get_instr (cpu), 20, 16); | |
8605 | unsigned rn = uimm (aarch64_get_instr (cpu), 9, 5); | |
8606 | unsigned rd = uimm (aarch64_get_instr (cpu), 4, 0); | |
8607 | ||
8608 | aarch64_set_reg_u64 | |
8609 | (cpu, rd, NO_SP, aarch64_get_reg_u64 (cpu, rn, NO_SP) | |
8610 | ^ shifted64 (aarch64_get_reg_u64 (cpu, rm, NO_SP), shift, count)); | |
8611 | } | |
8612 | ||
8613 | /* 32 bit ORR shifted register. */ | |
8614 | static void | |
8615 | orr32_shift (sim_cpu *cpu, Shift shift, uint32_t count) | |
8616 | { | |
8617 | unsigned rm = uimm (aarch64_get_instr (cpu), 20, 16); | |
8618 | unsigned rn = uimm (aarch64_get_instr (cpu), 9, 5); | |
8619 | unsigned rd = uimm (aarch64_get_instr (cpu), 4, 0); | |
8620 | ||
8621 | aarch64_set_reg_u64 | |
8622 | (cpu, rd, NO_SP, aarch64_get_reg_u32 (cpu, rn, NO_SP) | |
8623 | | shifted32 (aarch64_get_reg_u32 (cpu, rm, NO_SP), shift, count)); | |
8624 | } | |
8625 | ||
8626 | /* 64 bit ORR shifted register. */ | |
8627 | static void | |
8628 | orr64_shift (sim_cpu *cpu, Shift shift, uint32_t count) | |
8629 | { | |
8630 | unsigned rm = uimm (aarch64_get_instr (cpu), 20, 16); | |
8631 | unsigned rn = uimm (aarch64_get_instr (cpu), 9, 5); | |
8632 | unsigned rd = uimm (aarch64_get_instr (cpu), 4, 0); | |
8633 | ||
8634 | aarch64_set_reg_u64 | |
8635 | (cpu, rd, NO_SP, aarch64_get_reg_u64 (cpu, rn, NO_SP) | |
8636 | | shifted64 (aarch64_get_reg_u64 (cpu, rm, NO_SP), shift, count)); | |
8637 | } | |
8638 | ||
8639 | /* 32 bit ORN shifted register. */ | |
8640 | static void | |
8641 | orn32_shift (sim_cpu *cpu, Shift shift, uint32_t count) | |
8642 | { | |
8643 | unsigned rm = uimm (aarch64_get_instr (cpu), 20, 16); | |
8644 | unsigned rn = uimm (aarch64_get_instr (cpu), 9, 5); | |
8645 | unsigned rd = uimm (aarch64_get_instr (cpu), 4, 0); | |
8646 | ||
8647 | aarch64_set_reg_u64 | |
8648 | (cpu, rd, NO_SP, aarch64_get_reg_u32 (cpu, rn, NO_SP) | |
8649 | | ~ shifted32 (aarch64_get_reg_u32 (cpu, rm, NO_SP), shift, count)); | |
8650 | } | |
8651 | ||
8652 | /* 64 bit ORN shifted register. */ | |
8653 | static void | |
8654 | orn64_shift (sim_cpu *cpu, Shift shift, uint32_t count) | |
8655 | { | |
8656 | unsigned rm = uimm (aarch64_get_instr (cpu), 20, 16); | |
8657 | unsigned rn = uimm (aarch64_get_instr (cpu), 9, 5); | |
8658 | unsigned rd = uimm (aarch64_get_instr (cpu), 4, 0); | |
8659 | ||
8660 | aarch64_set_reg_u64 | |
8661 | (cpu, rd, NO_SP, aarch64_get_reg_u64 (cpu, rn, NO_SP) | |
8662 | | ~ shifted64 (aarch64_get_reg_u64 (cpu, rm, NO_SP), shift, count)); | |
8663 | } | |
8664 | ||
8665 | static void | |
8666 | dexLogicalImmediate (sim_cpu *cpu) | |
8667 | { | |
8668 | /* assert instr[28,23] = 1001000 | |
8669 | instr[31] = size : 0 ==> 32 bit, 1 ==> 64 bit | |
8670 | instr[30,29] = op : 0 ==> AND, 1 ==> ORR, 2 ==> EOR, 3 ==> ANDS | |
8671 | instr[22] = N : used to construct immediate mask | |
8672 | instr[21,16] = immr | |
8673 | instr[15,10] = imms | |
8674 | instr[9,5] = Rn | |
8675 | instr[4,0] = Rd */ | |
8676 | ||
8677 | /* 32 bit operations must have N = 0 or else we have an UNALLOC. */ | |
8678 | uint32_t size = uimm (aarch64_get_instr (cpu), 31, 31); | |
8679 | uint32_t N = uimm (aarch64_get_instr (cpu), 22, 22); | |
8680 | /* uint32_t immr = uimm (aarch64_get_instr (cpu), 21, 16);. */ | |
8681 | /* uint32_t imms = uimm (aarch64_get_instr (cpu), 15, 10);. */ | |
8682 | uint32_t index = uimm (aarch64_get_instr (cpu), 22, 10); | |
8683 | uint64_t bimm64 = LITable [index]; | |
8684 | uint32_t dispatch = uimm (aarch64_get_instr (cpu), 30, 29); | |
8685 | ||
8686 | if (~size & N) | |
8687 | HALT_UNALLOC; | |
8688 | ||
8689 | if (!bimm64) | |
8690 | HALT_UNALLOC; | |
8691 | ||
8692 | if (size == 0) | |
8693 | { | |
8694 | uint32_t bimm = (uint32_t) bimm64; | |
8695 | ||
8696 | switch (dispatch) | |
8697 | { | |
8698 | case 0: and32 (cpu, bimm); return; | |
8699 | case 1: orr32 (cpu, bimm); return; | |
8700 | case 2: eor32 (cpu, bimm); return; | |
8701 | case 3: ands32 (cpu, bimm); return; | |
8702 | } | |
8703 | } | |
8704 | else | |
8705 | { | |
8706 | switch (dispatch) | |
8707 | { | |
8708 | case 0: and64 (cpu, bimm64); return; | |
8709 | case 1: orr64 (cpu, bimm64); return; | |
8710 | case 2: eor64 (cpu, bimm64); return; | |
8711 | case 3: ands64 (cpu, bimm64); return; | |
8712 | } | |
8713 | } | |
8714 | HALT_UNALLOC; | |
8715 | } | |
8716 | ||
8717 | /* Immediate move. | |
8718 | The uimm argument is a 16 bit value to be inserted into the | |
8719 | target register the pos argument locates the 16 bit word in the | |
8720 | dest register i.e. it is in {0, 1} for 32 bit and {0, 1, 2, | |
8721 | 3} for 64 bit. | |
8722 | N.B register arg may not be SP so it should be. | |
8723 | accessed using the setGZRegisterXXX accessors. */ | |
8724 | ||
8725 | /* 32 bit move 16 bit immediate zero remaining shorts. */ | |
8726 | static void | |
8727 | movz32 (sim_cpu *cpu, uint32_t val, uint32_t pos) | |
8728 | { | |
8729 | unsigned rd = uimm (aarch64_get_instr (cpu), 4, 0); | |
8730 | ||
8731 | aarch64_set_reg_u64 (cpu, rd, NO_SP, val << (pos * 16)); | |
8732 | } | |
8733 | ||
8734 | /* 64 bit move 16 bit immediate zero remaining shorts. */ | |
8735 | static void | |
8736 | movz64 (sim_cpu *cpu, uint32_t val, uint32_t pos) | |
8737 | { | |
8738 | unsigned rd = uimm (aarch64_get_instr (cpu), 4, 0); | |
8739 | ||
8740 | aarch64_set_reg_u64 (cpu, rd, NO_SP, ((uint64_t) val) << (pos * 16)); | |
8741 | } | |
8742 | ||
8743 | /* 32 bit move 16 bit immediate negated. */ | |
8744 | static void | |
8745 | movn32 (sim_cpu *cpu, uint32_t val, uint32_t pos) | |
8746 | { | |
8747 | unsigned rd = uimm (aarch64_get_instr (cpu), 4, 0); | |
8748 | ||
8749 | aarch64_set_reg_u64 (cpu, rd, NO_SP, ((val << (pos * 16)) ^ 0xffffffffU)); | |
8750 | } | |
8751 | ||
8752 | /* 64 bit move 16 bit immediate negated. */ | |
8753 | static void | |
8754 | movn64 (sim_cpu *cpu, uint32_t val, uint32_t pos) | |
8755 | { | |
8756 | unsigned rd = uimm (aarch64_get_instr (cpu), 4, 0); | |
8757 | ||
8758 | aarch64_set_reg_u64 | |
8759 | (cpu, rd, NO_SP, ((((uint64_t) val) << (pos * 16)) | |
8760 | ^ 0xffffffffffffffffULL)); | |
8761 | } | |
8762 | ||
8763 | /* 32 bit move 16 bit immediate keep remaining shorts. */ | |
8764 | static void | |
8765 | movk32 (sim_cpu *cpu, uint32_t val, uint32_t pos) | |
8766 | { | |
8767 | unsigned rd = uimm (aarch64_get_instr (cpu), 4, 0); | |
8768 | uint32_t current = aarch64_get_reg_u32 (cpu, rd, NO_SP); | |
8769 | uint32_t value = val << (pos * 16); | |
8770 | uint32_t mask = ~(0xffffU << (pos * 16)); | |
8771 | ||
8772 | aarch64_set_reg_u64 (cpu, rd, NO_SP, (value | (current & mask))); | |
8773 | } | |
8774 | ||
8775 | /* 64 bit move 16 it immediate keep remaining shorts. */ | |
8776 | static void | |
8777 | movk64 (sim_cpu *cpu, uint32_t val, uint32_t pos) | |
8778 | { | |
8779 | unsigned rd = uimm (aarch64_get_instr (cpu), 4, 0); | |
8780 | uint64_t current = aarch64_get_reg_u64 (cpu, rd, NO_SP); | |
8781 | uint64_t value = (uint64_t) val << (pos * 16); | |
8782 | uint64_t mask = ~(0xffffULL << (pos * 16)); | |
8783 | ||
8784 | aarch64_set_reg_u64 (cpu, rd, NO_SP, (value | (current & mask))); | |
8785 | } | |
8786 | ||
8787 | static void | |
8788 | dexMoveWideImmediate (sim_cpu *cpu) | |
8789 | { | |
8790 | /* assert instr[28:23] = 100101 | |
8791 | instr[31] = size : 0 ==> 32 bit, 1 ==> 64 bit | |
8792 | instr[30,29] = op : 0 ==> MOVN, 1 ==> UNALLOC, 2 ==> MOVZ, 3 ==> MOVK | |
8793 | instr[22,21] = shift : 00 == LSL#0, 01 = LSL#16, 10 = LSL#32, 11 = LSL#48 | |
8794 | instr[20,5] = uimm16 | |
8795 | instr[4,0] = Rd */ | |
8796 | ||
8797 | /* N.B. the (multiple of 16) shift is applied by the called routine, | |
8798 | we just pass the multiplier. */ | |
8799 | ||
8800 | uint32_t imm; | |
8801 | uint32_t size = uimm (aarch64_get_instr (cpu), 31, 31); | |
8802 | uint32_t op = uimm (aarch64_get_instr (cpu), 30, 29); | |
8803 | uint32_t shift = uimm (aarch64_get_instr (cpu), 22, 21); | |
8804 | ||
8805 | /* 32 bit can only shift 0 or 1 lot of 16. | |
8806 | anything else is an unallocated instruction. */ | |
8807 | if (size == 0 && (shift > 1)) | |
8808 | HALT_UNALLOC; | |
8809 | ||
8810 | if (op == 1) | |
8811 | HALT_UNALLOC; | |
8812 | ||
8813 | imm = uimm (aarch64_get_instr (cpu), 20, 5); | |
8814 | ||
8815 | if (size == 0) | |
8816 | { | |
8817 | if (op == 0) | |
8818 | movn32 (cpu, imm, shift); | |
8819 | else if (op == 2) | |
8820 | movz32 (cpu, imm, shift); | |
8821 | else | |
8822 | movk32 (cpu, imm, shift); | |
8823 | } | |
8824 | else | |
8825 | { | |
8826 | if (op == 0) | |
8827 | movn64 (cpu, imm, shift); | |
8828 | else if (op == 2) | |
8829 | movz64 (cpu, imm, shift); | |
8830 | else | |
8831 | movk64 (cpu, imm, shift); | |
8832 | } | |
8833 | } | |
8834 | ||
8835 | /* Bitfield operations. | |
8836 | These take a pair of bit positions r and s which are in {0..31} | |
8837 | or {0..63} depending on the instruction word size. | |
8838 | N.B register args may not be SP. */ | |
8839 | ||
8840 | /* OK, we start with ubfm which just needs to pick | |
8841 | some bits out of source zero the rest and write | |
8842 | the result to dest. Just need two logical shifts. */ | |
8843 | ||
8844 | /* 32 bit bitfield move, left and right of affected zeroed | |
8845 | if r <= s Wd<s-r:0> = Wn<s:r> else Wd<32+s-r,32-r> = Wn<s:0>. */ | |
8846 | static void | |
8847 | ubfm32 (sim_cpu *cpu, uint32_t r, uint32_t s) | |
8848 | { | |
8849 | unsigned rd; | |
8850 | unsigned rn = uimm (aarch64_get_instr (cpu), 9, 5); | |
8851 | uint32_t value = aarch64_get_reg_u32 (cpu, rn, NO_SP); | |
8852 | ||
8853 | /* Pick either s+1-r or s+1 consecutive bits out of the original word. */ | |
8854 | if (r <= s) | |
8855 | { | |
8856 | /* 31:...:s:xxx:r:...:0 ==> 31:...:s-r:xxx:0. | |
8857 | We want only bits s:xxx:r at the bottom of the word | |
8858 | so we LSL bit s up to bit 31 i.e. by 31 - s | |
8859 | and then we LSR to bring bit 31 down to bit s - r | |
8860 | i.e. by 31 + r - s. */ | |
8861 | value <<= 31 - s; | |
8862 | value >>= 31 + r - s; | |
8863 | } | |
8864 | else | |
8865 | { | |
8866 | /* 31:...:s:xxx:0 ==> 31:...:31-(r-1)+s:xxx:31-(r-1):...:0 | |
8867 | We want only bits s:xxx:0 starting at it 31-(r-1) | |
8868 | so we LSL bit s up to bit 31 i.e. by 31 - s | |
8869 | and then we LSL to bring bit 31 down to 31-(r-1)+s | |
8870 | i.e. by r - (s + 1). */ | |
8871 | value <<= 31 - s; | |
8872 | value >>= r - (s + 1); | |
8873 | } | |
8874 | ||
8875 | rd = uimm (aarch64_get_instr (cpu), 4, 0); | |
8876 | aarch64_set_reg_u64 (cpu, rd, NO_SP, value); | |
8877 | } | |
8878 | ||
8879 | /* 64 bit bitfield move, left and right of affected zeroed | |
8880 | if r <= s Wd<s-r:0> = Wn<s:r> else Wd<64+s-r,64-r> = Wn<s:0>. */ | |
8881 | static void | |
8882 | ubfm (sim_cpu *cpu, uint32_t r, uint32_t s) | |
8883 | { | |
8884 | unsigned rd; | |
8885 | unsigned rn = uimm (aarch64_get_instr (cpu), 9, 5); | |
8886 | uint64_t value = aarch64_get_reg_u64 (cpu, rn, NO_SP); | |
8887 | ||
8888 | if (r <= s) | |
8889 | { | |
8890 | /* 63:...:s:xxx:r:...:0 ==> 63:...:s-r:xxx:0. | |
8891 | We want only bits s:xxx:r at the bottom of the word. | |
8892 | So we LSL bit s up to bit 63 i.e. by 63 - s | |
8893 | and then we LSR to bring bit 63 down to bit s - r | |
8894 | i.e. by 63 + r - s. */ | |
8895 | value <<= 63 - s; | |
8896 | value >>= 63 + r - s; | |
8897 | } | |
8898 | else | |
8899 | { | |
8900 | /* 63:...:s:xxx:0 ==> 63:...:63-(r-1)+s:xxx:63-(r-1):...:0. | |
8901 | We want only bits s:xxx:0 starting at it 63-(r-1). | |
8902 | So we LSL bit s up to bit 63 i.e. by 63 - s | |
8903 | and then we LSL to bring bit 63 down to 63-(r-1)+s | |
8904 | i.e. by r - (s + 1). */ | |
8905 | value <<= 63 - s; | |
8906 | value >>= r - (s + 1); | |
8907 | } | |
8908 | ||
8909 | rd = uimm (aarch64_get_instr (cpu), 4, 0); | |
8910 | aarch64_set_reg_u64 (cpu, rd, NO_SP, value); | |
8911 | } | |
8912 | ||
8913 | /* The signed versions need to insert sign bits | |
8914 | on the left of the inserted bit field. so we do | |
8915 | much the same as the unsigned version except we | |
8916 | use an arithmetic shift right -- this just means | |
8917 | we need to operate on signed values. */ | |
8918 | ||
8919 | /* 32 bit bitfield move, left of affected sign-extended, right zeroed. */ | |
8920 | /* If r <= s Wd<s-r:0> = Wn<s:r> else Wd<32+s-r,32-r> = Wn<s:0>. */ | |
8921 | static void | |
8922 | sbfm32 (sim_cpu *cpu, uint32_t r, uint32_t s) | |
8923 | { | |
8924 | unsigned rd; | |
8925 | unsigned rn = uimm (aarch64_get_instr (cpu), 9, 5); | |
8926 | /* as per ubfm32 but use an ASR instead of an LSR. */ | |
8927 | int32_t value = aarch64_get_reg_s32 (cpu, rn, NO_SP); | |
8928 | ||
8929 | if (r <= s) | |
8930 | { | |
8931 | value <<= 31 - s; | |
8932 | value >>= 31 + r - s; | |
8933 | } | |
8934 | else | |
8935 | { | |
8936 | value <<= 31 - s; | |
8937 | value >>= r - (s + 1); | |
8938 | } | |
8939 | ||
8940 | rd = uimm (aarch64_get_instr (cpu), 4, 0); | |
8941 | aarch64_set_reg_u64 (cpu, rd, NO_SP, (uint32_t) value); | |
8942 | } | |
8943 | ||
8944 | /* 64 bit bitfield move, left of affected sign-extended, right zeroed. */ | |
8945 | /* If r <= s Wd<s-r:0> = Wn<s:r> else Wd<64+s-r,64-r> = Wn<s:0>. */ | |
8946 | static void | |
8947 | sbfm (sim_cpu *cpu, uint32_t r, uint32_t s) | |
8948 | { | |
8949 | unsigned rd; | |
8950 | unsigned rn = uimm (aarch64_get_instr (cpu), 9, 5); | |
8951 | /* acpu per ubfm but use an ASR instead of an LSR. */ | |
8952 | int64_t value = aarch64_get_reg_s64 (cpu, rn, NO_SP); | |
8953 | ||
8954 | if (r <= s) | |
8955 | { | |
8956 | value <<= 63 - s; | |
8957 | value >>= 63 + r - s; | |
8958 | } | |
8959 | else | |
8960 | { | |
8961 | value <<= 63 - s; | |
8962 | value >>= r - (s + 1); | |
8963 | } | |
8964 | ||
8965 | rd = uimm (aarch64_get_instr (cpu), 4, 0); | |
8966 | aarch64_set_reg_s64 (cpu, rd, NO_SP, value); | |
8967 | } | |
8968 | ||
8969 | /* Finally, these versions leave non-affected bits | |
8970 | as is. so we need to generate the bits as per | |
8971 | ubfm and also generate a mask to pick the | |
8972 | bits from the original and computed values. */ | |
8973 | ||
8974 | /* 32 bit bitfield move, non-affected bits left as is. | |
8975 | If r <= s Wd<s-r:0> = Wn<s:r> else Wd<32+s-r,32-r> = Wn<s:0>. */ | |
8976 | static void | |
8977 | bfm32 (sim_cpu *cpu, uint32_t r, uint32_t s) | |
8978 | { | |
8979 | unsigned rn = uimm (aarch64_get_instr (cpu), 9, 5); | |
8980 | uint32_t value = aarch64_get_reg_u32 (cpu, rn, NO_SP); | |
8981 | uint32_t mask = -1; | |
8982 | unsigned rd; | |
8983 | uint32_t value2; | |
8984 | ||
8985 | /* Pick either s+1-r or s+1 consecutive bits out of the original word. */ | |
8986 | if (r <= s) | |
8987 | { | |
8988 | /* 31:...:s:xxx:r:...:0 ==> 31:...:s-r:xxx:0. | |
8989 | We want only bits s:xxx:r at the bottom of the word | |
8990 | so we LSL bit s up to bit 31 i.e. by 31 - s | |
8991 | and then we LSR to bring bit 31 down to bit s - r | |
8992 | i.e. by 31 + r - s. */ | |
8993 | value <<= 31 - s; | |
8994 | value >>= 31 + r - s; | |
8995 | /* the mask must include the same bits. */ | |
8996 | mask <<= 31 - s; | |
8997 | mask >>= 31 + r - s; | |
8998 | } | |
8999 | else | |
9000 | { | |
9001 | /* 31:...:s:xxx:0 ==> 31:...:31-(r-1)+s:xxx:31-(r-1):...:0. | |
9002 | We want only bits s:xxx:0 starting at it 31-(r-1) | |
9003 | so we LSL bit s up to bit 31 i.e. by 31 - s | |
9004 | and then we LSL to bring bit 31 down to 31-(r-1)+s | |
9005 | i.e. by r - (s + 1). */ | |
9006 | value <<= 31 - s; | |
9007 | value >>= r - (s + 1); | |
9008 | /* The mask must include the same bits. */ | |
9009 | mask <<= 31 - s; | |
9010 | mask >>= r - (s + 1); | |
9011 | } | |
9012 | ||
9013 | rd = uimm (aarch64_get_instr (cpu), 4, 0); | |
9014 | value2 = aarch64_get_reg_u32 (cpu, rd, NO_SP); | |
9015 | ||
9016 | value2 &= ~mask; | |
9017 | value2 |= value; | |
9018 | ||
9019 | aarch64_set_reg_u64 | |
9020 | (cpu, rd, NO_SP, (aarch64_get_reg_u32 (cpu, rd, NO_SP) & ~mask) | value); | |
9021 | } | |
9022 | ||
9023 | /* 64 bit bitfield move, non-affected bits left as is. | |
9024 | If r <= s Wd<s-r:0> = Wn<s:r> else Wd<64+s-r,64-r> = Wn<s:0>. */ | |
9025 | static void | |
9026 | bfm (sim_cpu *cpu, uint32_t r, uint32_t s) | |
9027 | { | |
9028 | unsigned rd; | |
9029 | unsigned rn = uimm (aarch64_get_instr (cpu), 9, 5); | |
9030 | uint64_t value = aarch64_get_reg_u64 (cpu, rn, NO_SP); | |
9031 | uint64_t mask = 0xffffffffffffffffULL; | |
9032 | ||
9033 | if (r <= s) | |
9034 | { | |
9035 | /* 63:...:s:xxx:r:...:0 ==> 63:...:s-r:xxx:0. | |
9036 | We want only bits s:xxx:r at the bottom of the word | |
9037 | so we LSL bit s up to bit 63 i.e. by 63 - s | |
9038 | and then we LSR to bring bit 63 down to bit s - r | |
9039 | i.e. by 63 + r - s. */ | |
9040 | value <<= 63 - s; | |
9041 | value >>= 63 + r - s; | |
9042 | /* The mask must include the same bits. */ | |
9043 | mask <<= 63 - s; | |
9044 | mask >>= 63 + r - s; | |
9045 | } | |
9046 | else | |
9047 | { | |
9048 | /* 63:...:s:xxx:0 ==> 63:...:63-(r-1)+s:xxx:63-(r-1):...:0 | |
9049 | We want only bits s:xxx:0 starting at it 63-(r-1) | |
9050 | so we LSL bit s up to bit 63 i.e. by 63 - s | |
9051 | and then we LSL to bring bit 63 down to 63-(r-1)+s | |
9052 | i.e. by r - (s + 1). */ | |
9053 | value <<= 63 - s; | |
9054 | value >>= r - (s + 1); | |
9055 | /* The mask must include the same bits. */ | |
9056 | mask <<= 63 - s; | |
9057 | mask >>= r - (s + 1); | |
9058 | } | |
9059 | ||
9060 | rd = uimm (aarch64_get_instr (cpu), 4, 0); | |
9061 | aarch64_set_reg_u64 | |
9062 | (cpu, rd, NO_SP, (aarch64_get_reg_u64 (cpu, rd, NO_SP) & ~mask) | value); | |
9063 | } | |
9064 | ||
9065 | static void | |
9066 | dexBitfieldImmediate (sim_cpu *cpu) | |
9067 | { | |
9068 | /* assert instr[28:23] = 100110 | |
9069 | instr[31] = size : 0 ==> 32 bit, 1 ==> 64 bit | |
9070 | instr[30,29] = op : 0 ==> SBFM, 1 ==> BFM, 2 ==> UBFM, 3 ==> UNALLOC | |
9071 | instr[22] = N : must be 0 for 32 bit, 1 for 64 bit ow UNALLOC | |
9072 | instr[21,16] = immr : 0xxxxx for 32 bit, xxxxxx for 64 bit | |
9073 | instr[15,10] = imms : 0xxxxx for 32 bit, xxxxxx for 64 bit | |
9074 | instr[9,5] = Rn | |
9075 | instr[4,0] = Rd */ | |
9076 | ||
9077 | /* 32 bit operations must have N = 0 or else we have an UNALLOC. */ | |
9078 | uint32_t dispatch; | |
9079 | uint32_t imms; | |
9080 | uint32_t size = uimm (aarch64_get_instr (cpu), 31, 31); | |
9081 | uint32_t N = uimm (aarch64_get_instr (cpu), 22, 22); | |
9082 | /* 32 bit operations must have immr[5] = 0 and imms[5] = 0. */ | |
9083 | /* or else we have an UNALLOC. */ | |
9084 | uint32_t immr = uimm (aarch64_get_instr (cpu), 21, 16); | |
9085 | ||
9086 | if (~size & N) | |
9087 | HALT_UNALLOC; | |
9088 | ||
9089 | if (!size && uimm (immr, 5, 5)) | |
9090 | HALT_UNALLOC; | |
9091 | ||
9092 | imms = uimm (aarch64_get_instr (cpu), 15, 10); | |
9093 | if (!size && uimm (imms, 5, 5)) | |
9094 | HALT_UNALLOC; | |
9095 | ||
9096 | /* Switch on combined size and op. */ | |
9097 | dispatch = uimm (aarch64_get_instr (cpu), 31, 29); | |
9098 | switch (dispatch) | |
9099 | { | |
9100 | case 0: sbfm32 (cpu, immr, imms); return; | |
9101 | case 1: bfm32 (cpu, immr, imms); return; | |
9102 | case 2: ubfm32 (cpu, immr, imms); return; | |
9103 | case 4: sbfm (cpu, immr, imms); return; | |
9104 | case 5: bfm (cpu, immr, imms); return; | |
9105 | case 6: ubfm (cpu, immr, imms); return; | |
9106 | default: HALT_UNALLOC; | |
9107 | } | |
9108 | } | |
9109 | ||
9110 | static void | |
9111 | do_EXTR_32 (sim_cpu *cpu) | |
9112 | { | |
9113 | /* instr[31:21] = 00010011100 | |
9114 | instr[20,16] = Rm | |
9115 | instr[15,10] = imms : 0xxxxx for 32 bit | |
9116 | instr[9,5] = Rn | |
9117 | instr[4,0] = Rd */ | |
9118 | unsigned rm = uimm (aarch64_get_instr (cpu), 20, 16); | |
9119 | unsigned imms = uimm (aarch64_get_instr (cpu), 15, 10) & 31; | |
9120 | unsigned rn = uimm (aarch64_get_instr (cpu), 9, 5); | |
9121 | unsigned rd = uimm (aarch64_get_instr (cpu), 4, 0); | |
9122 | uint64_t val1; | |
9123 | uint64_t val2; | |
9124 | ||
9125 | val1 = aarch64_get_reg_u32 (cpu, rm, NO_SP); | |
9126 | val1 >>= imms; | |
9127 | val2 = aarch64_get_reg_u32 (cpu, rn, NO_SP); | |
9128 | val2 <<= (32 - imms); | |
9129 | ||
9130 | aarch64_set_reg_u64 (cpu, rd, NO_SP, val1 | val2); | |
9131 | } | |
9132 | ||
9133 | static void | |
9134 | do_EXTR_64 (sim_cpu *cpu) | |
9135 | { | |
9136 | /* instr[31:21] = 10010011100 | |
9137 | instr[20,16] = Rm | |
9138 | instr[15,10] = imms | |
9139 | instr[9,5] = Rn | |
9140 | instr[4,0] = Rd */ | |
9141 | unsigned rm = uimm (aarch64_get_instr (cpu), 20, 16); | |
9142 | unsigned imms = uimm (aarch64_get_instr (cpu), 15, 10) & 63; | |
9143 | unsigned rn = uimm (aarch64_get_instr (cpu), 9, 5); | |
9144 | unsigned rd = uimm (aarch64_get_instr (cpu), 4, 0); | |
9145 | uint64_t val; | |
9146 | ||
9147 | val = aarch64_get_reg_u64 (cpu, rm, NO_SP); | |
9148 | val >>= imms; | |
9149 | val |= (aarch64_get_reg_u64 (cpu, rn, NO_SP) << (64 - imms)); | |
9150 | ||
9151 | aarch64_set_reg_u64 (cpu, rd, NO_SP, val); | |
9152 | } | |
9153 | ||
9154 | static void | |
9155 | dexExtractImmediate (sim_cpu *cpu) | |
9156 | { | |
9157 | /* assert instr[28:23] = 100111 | |
9158 | instr[31] = size : 0 ==> 32 bit, 1 ==> 64 bit | |
9159 | instr[30,29] = op21 : 0 ==> EXTR, 1,2,3 ==> UNALLOC | |
9160 | instr[22] = N : must be 0 for 32 bit, 1 for 64 bit or UNALLOC | |
9161 | instr[21] = op0 : must be 0 or UNALLOC | |
9162 | instr[20,16] = Rm | |
9163 | instr[15,10] = imms : 0xxxxx for 32 bit, xxxxxx for 64 bit | |
9164 | instr[9,5] = Rn | |
9165 | instr[4,0] = Rd */ | |
9166 | ||
9167 | /* 32 bit operations must have N = 0 or else we have an UNALLOC. */ | |
9168 | /* 64 bit operations must have N = 1 or else we have an UNALLOC. */ | |
9169 | uint32_t dispatch; | |
9170 | uint32_t size = uimm (aarch64_get_instr (cpu), 31, 31); | |
9171 | uint32_t N = uimm (aarch64_get_instr (cpu), 22, 22); | |
9172 | /* 32 bit operations must have imms[5] = 0 | |
9173 | or else we have an UNALLOC. */ | |
9174 | uint32_t imms = uimm (aarch64_get_instr (cpu), 15, 10); | |
9175 | ||
9176 | if (size ^ N) | |
9177 | HALT_UNALLOC; | |
9178 | ||
9179 | if (!size && uimm (imms, 5, 5)) | |
9180 | HALT_UNALLOC; | |
9181 | ||
9182 | /* Switch on combined size and op. */ | |
9183 | dispatch = uimm (aarch64_get_instr (cpu), 31, 29); | |
9184 | ||
9185 | if (dispatch == 0) | |
9186 | do_EXTR_32 (cpu); | |
9187 | ||
9188 | else if (dispatch == 4) | |
9189 | do_EXTR_64 (cpu); | |
9190 | ||
9191 | else if (dispatch == 1) | |
9192 | HALT_NYI; | |
9193 | else | |
9194 | HALT_UNALLOC; | |
9195 | } | |
9196 | ||
9197 | static void | |
9198 | dexDPImm (sim_cpu *cpu) | |
9199 | { | |
9200 | /* uint32_t group = dispatchGroup (aarch64_get_instr (cpu)); | |
9201 | assert group == GROUP_DPIMM_1000 || grpoup == GROUP_DPIMM_1001 | |
9202 | bits [25,23] of a DPImm are the secondary dispatch vector. */ | |
9203 | uint32_t group2 = dispatchDPImm (aarch64_get_instr (cpu)); | |
9204 | ||
9205 | switch (group2) | |
9206 | { | |
9207 | case DPIMM_PCADR_000: | |
9208 | case DPIMM_PCADR_001: | |
9209 | dexPCRelAddressing (cpu); | |
9210 | return; | |
9211 | ||
9212 | case DPIMM_ADDSUB_010: | |
9213 | case DPIMM_ADDSUB_011: | |
9214 | dexAddSubtractImmediate (cpu); | |
9215 | return; | |
9216 | ||
9217 | case DPIMM_LOG_100: | |
9218 | dexLogicalImmediate (cpu); | |
9219 | return; | |
9220 | ||
9221 | case DPIMM_MOV_101: | |
9222 | dexMoveWideImmediate (cpu); | |
9223 | return; | |
9224 | ||
9225 | case DPIMM_BITF_110: | |
9226 | dexBitfieldImmediate (cpu); | |
9227 | return; | |
9228 | ||
9229 | case DPIMM_EXTR_111: | |
9230 | dexExtractImmediate (cpu); | |
9231 | return; | |
9232 | ||
9233 | default: | |
9234 | /* Should never reach here. */ | |
9235 | HALT_NYI; | |
9236 | } | |
9237 | } | |
9238 | ||
9239 | static void | |
9240 | dexLoadUnscaledImmediate (sim_cpu *cpu) | |
9241 | { | |
9242 | /* instr[29,24] == 111_00 | |
9243 | instr[21] == 0 | |
9244 | instr[11,10] == 00 | |
9245 | instr[31,30] = size | |
9246 | instr[26] = V | |
9247 | instr[23,22] = opc | |
9248 | instr[20,12] = simm9 | |
9249 | instr[9,5] = rn may be SP. */ | |
9250 | /* unsigned rt = uimm (aarch64_get_instr (cpu), 4, 0); */ | |
9251 | uint32_t V = uimm (aarch64_get_instr (cpu), 26, 26); | |
9252 | uint32_t dispatch = ( (uimm (aarch64_get_instr (cpu), 31, 30) << 2) | |
9253 | | uimm (aarch64_get_instr (cpu), 23, 22)); | |
9254 | int32_t imm = simm32 (aarch64_get_instr (cpu), 20, 12); | |
9255 | ||
9256 | if (!V) | |
9257 | { | |
9258 | /* GReg operations. */ | |
9259 | switch (dispatch) | |
9260 | { | |
9261 | case 0: sturb (cpu, imm); return; | |
9262 | case 1: ldurb32 (cpu, imm); return; | |
9263 | case 2: ldursb64 (cpu, imm); return; | |
9264 | case 3: ldursb32 (cpu, imm); return; | |
9265 | case 4: sturh (cpu, imm); return; | |
9266 | case 5: ldurh32 (cpu, imm); return; | |
9267 | case 6: ldursh64 (cpu, imm); return; | |
9268 | case 7: ldursh32 (cpu, imm); return; | |
9269 | case 8: stur32 (cpu, imm); return; | |
9270 | case 9: ldur32 (cpu, imm); return; | |
9271 | case 10: ldursw (cpu, imm); return; | |
9272 | case 12: stur64 (cpu, imm); return; | |
9273 | case 13: ldur64 (cpu, imm); return; | |
9274 | ||
9275 | case 14: | |
9276 | /* PRFUM NYI. */ | |
9277 | HALT_NYI; | |
9278 | ||
9279 | default: | |
9280 | case 11: | |
9281 | case 15: | |
9282 | HALT_UNALLOC; | |
9283 | } | |
9284 | } | |
9285 | ||
9286 | /* FReg operations. */ | |
9287 | switch (dispatch) | |
9288 | { | |
9289 | case 2: fsturq (cpu, imm); return; | |
9290 | case 3: fldurq (cpu, imm); return; | |
9291 | case 8: fsturs (cpu, imm); return; | |
9292 | case 9: fldurs (cpu, imm); return; | |
9293 | case 12: fsturd (cpu, imm); return; | |
9294 | case 13: fldurd (cpu, imm); return; | |
9295 | ||
9296 | case 0: /* STUR 8 bit FP. */ | |
9297 | case 1: /* LDUR 8 bit FP. */ | |
9298 | case 4: /* STUR 16 bit FP. */ | |
9299 | case 5: /* LDUR 8 bit FP. */ | |
9300 | HALT_NYI; | |
9301 | ||
9302 | default: | |
9303 | case 6: | |
9304 | case 7: | |
9305 | case 10: | |
9306 | case 11: | |
9307 | case 14: | |
9308 | case 15: | |
9309 | HALT_UNALLOC; | |
9310 | } | |
9311 | } | |
9312 | ||
9313 | /* N.B. A preliminary note regarding all the ldrs<x>32 | |
9314 | instructions | |
9315 | ||
9316 | The signed value loaded by these instructions is cast to unsigned | |
9317 | before being assigned to aarch64_get_reg_u64 (cpu, N) i.e. to the | |
9318 | 64 bit element of the GReg union. this performs a 32 bit sign extension | |
9319 | (as required) but avoids 64 bit sign extension, thus ensuring that the | |
9320 | top half of the register word is zero. this is what the spec demands | |
9321 | when a 32 bit load occurs. */ | |
9322 | ||
9323 | /* 32 bit load sign-extended byte scaled unsigned 12 bit. */ | |
9324 | static void | |
9325 | ldrsb32_abs (sim_cpu *cpu, uint32_t offset) | |
9326 | { | |
9327 | unsigned int rn = uimm (aarch64_get_instr (cpu), 9, 5); | |
9328 | unsigned int rt = uimm (aarch64_get_instr (cpu), 4, 0); | |
9329 | ||
9330 | /* The target register may not be SP but the source may be | |
9331 | there is no scaling required for a byte load. */ | |
9332 | uint64_t address = aarch64_get_reg_u64 (cpu, rn, SP_OK) + offset; | |
9333 | aarch64_set_reg_u64 (cpu, rt, NO_SP, | |
9334 | (int64_t) aarch64_get_mem_s8 (cpu, address)); | |
9335 | } | |
9336 | ||
9337 | /* 32 bit load sign-extended byte scaled or unscaled zero- | |
9338 | or sign-extended 32-bit register offset. */ | |
9339 | static void | |
9340 | ldrsb32_scale_ext (sim_cpu *cpu, Scaling scaling, Extension extension) | |
9341 | { | |
9342 | unsigned int rm = uimm (aarch64_get_instr (cpu), 20, 16); | |
9343 | unsigned int rn = uimm (aarch64_get_instr (cpu), 9, 5); | |
9344 | unsigned int rt = uimm (aarch64_get_instr (cpu), 4, 0); | |
9345 | ||
9346 | /* rn may reference SP, rm and rt must reference ZR. */ | |
9347 | ||
9348 | uint64_t address = aarch64_get_reg_u64 (cpu, rn, SP_OK); | |
9349 | int64_t displacement = extend (aarch64_get_reg_u32 (cpu, rm, NO_SP), | |
9350 | extension); | |
9351 | ||
9352 | /* There is no scaling required for a byte load. */ | |
9353 | aarch64_set_reg_u64 | |
9354 | (cpu, rt, NO_SP, (int64_t) aarch64_get_mem_s8 (cpu, address | |
9355 | + displacement)); | |
9356 | } | |
9357 | ||
9358 | /* 32 bit load sign-extended byte unscaled signed 9 bit with | |
9359 | pre- or post-writeback. */ | |
9360 | static void | |
9361 | ldrsb32_wb (sim_cpu *cpu, int32_t offset, WriteBack wb) | |
9362 | { | |
9363 | uint64_t address; | |
9364 | unsigned int rn = uimm (aarch64_get_instr (cpu), 9, 5); | |
9365 | unsigned int rt = uimm (aarch64_get_instr (cpu), 4, 0); | |
9366 | ||
9367 | if (rn == rt && wb != NoWriteBack) | |
9368 | HALT_UNALLOC; | |
9369 | ||
9370 | address = aarch64_get_reg_u64 (cpu, rn, SP_OK); | |
9371 | ||
9372 | if (wb == Pre) | |
9373 | address += offset; | |
9374 | ||
9375 | aarch64_set_reg_u64 (cpu, rt, NO_SP, | |
9376 | (int64_t) aarch64_get_mem_s8 (cpu, address)); | |
9377 | ||
9378 | if (wb == Post) | |
9379 | address += offset; | |
9380 | ||
9381 | if (wb != NoWriteBack) | |
9382 | aarch64_set_reg_u64 (cpu, rn, NO_SP, address); | |
9383 | } | |
9384 | ||
9385 | /* 8 bit store scaled. */ | |
9386 | static void | |
9387 | fstrb_abs (sim_cpu *cpu, uint32_t offset) | |
9388 | { | |
9389 | unsigned st = uimm (aarch64_get_instr (cpu), 4, 0); | |
9390 | unsigned rn = uimm (aarch64_get_instr (cpu), 9, 5); | |
9391 | ||
9392 | aarch64_set_mem_u8 (cpu, | |
9393 | aarch64_get_reg_u64 (cpu, rn, SP_OK) + offset, | |
9394 | aarch64_get_vec_u8 (cpu, st, 0)); | |
9395 | } | |
9396 | ||
9397 | /* 8 bit store scaled or unscaled zero- or | |
9398 | sign-extended 8-bit register offset. */ | |
9399 | static void | |
9400 | fstrb_scale_ext (sim_cpu *cpu, Scaling scaling, Extension extension) | |
9401 | { | |
9402 | unsigned rm = uimm (aarch64_get_instr (cpu), 20, 16); | |
9403 | unsigned rn = uimm (aarch64_get_instr (cpu), 9, 5); | |
9404 | unsigned st = uimm (aarch64_get_instr (cpu), 4, 0); | |
9405 | ||
9406 | uint64_t address = aarch64_get_reg_u64 (cpu, rn, SP_OK); | |
9407 | int64_t extended = extend (aarch64_get_reg_u32 (cpu, rm, NO_SP), | |
9408 | extension); | |
9409 | uint64_t displacement = OPT_SCALE (extended, 32, scaling); | |
9410 | ||
9411 | aarch64_set_mem_u8 | |
9412 | (cpu, address + displacement, aarch64_get_vec_u8 (cpu, st, 0)); | |
9413 | } | |
9414 | ||
9415 | /* 16 bit store scaled. */ | |
9416 | static void | |
9417 | fstrh_abs (sim_cpu *cpu, uint32_t offset) | |
9418 | { | |
9419 | unsigned st = uimm (aarch64_get_instr (cpu), 4, 0); | |
9420 | unsigned rn = uimm (aarch64_get_instr (cpu), 9, 5); | |
9421 | ||
9422 | aarch64_set_mem_u16 | |
9423 | (cpu, | |
9424 | aarch64_get_reg_u64 (cpu, rn, SP_OK) + SCALE (offset, 16), | |
9425 | aarch64_get_vec_u16 (cpu, st, 0)); | |
9426 | } | |
9427 | ||
9428 | /* 16 bit store scaled or unscaled zero- | |
9429 | or sign-extended 16-bit register offset. */ | |
9430 | static void | |
9431 | fstrh_scale_ext (sim_cpu *cpu, Scaling scaling, Extension extension) | |
9432 | { | |
9433 | unsigned rm = uimm (aarch64_get_instr (cpu), 20, 16); | |
9434 | unsigned rn = uimm (aarch64_get_instr (cpu), 9, 5); | |
9435 | unsigned st = uimm (aarch64_get_instr (cpu), 4, 0); | |
9436 | ||
9437 | uint64_t address = aarch64_get_reg_u64 (cpu, rn, SP_OK); | |
9438 | int64_t extended = extend (aarch64_get_reg_u32 (cpu, rm, NO_SP), | |
9439 | extension); | |
9440 | uint64_t displacement = OPT_SCALE (extended, 32, scaling); | |
9441 | ||
9442 | aarch64_set_mem_u16 | |
9443 | (cpu, address + displacement, aarch64_get_vec_u16 (cpu, st, 0)); | |
9444 | } | |
9445 | ||
9446 | /* 32 bit store scaled unsigned 12 bit. */ | |
9447 | static void | |
9448 | fstrs_abs (sim_cpu *cpu, uint32_t offset) | |
9449 | { | |
9450 | unsigned st = uimm (aarch64_get_instr (cpu), 4, 0); | |
9451 | unsigned rn = uimm (aarch64_get_instr (cpu), 9, 5); | |
9452 | ||
9453 | aarch64_set_mem_float | |
9454 | (cpu, | |
9455 | aarch64_get_reg_u64 (cpu, rn, SP_OK) + SCALE (offset, 32), | |
9456 | aarch64_get_FP_float (cpu, st)); | |
9457 | } | |
9458 | ||
9459 | /* 32 bit store unscaled signed 9 bit with pre- or post-writeback. */ | |
9460 | static void | |
9461 | fstrs_wb (sim_cpu *cpu, int32_t offset, WriteBack wb) | |
9462 | { | |
9463 | unsigned rn = uimm (aarch64_get_instr (cpu), 9, 5); | |
9464 | unsigned st = uimm (aarch64_get_instr (cpu), 4, 0); | |
9465 | ||
9466 | uint64_t address = aarch64_get_reg_u64 (cpu, rn, SP_OK); | |
9467 | ||
9468 | if (wb != Post) | |
9469 | address += offset; | |
9470 | ||
9471 | aarch64_set_mem_float (cpu, address, aarch64_get_FP_float (cpu, st)); | |
9472 | ||
9473 | if (wb == Post) | |
9474 | address += offset; | |
9475 | ||
9476 | if (wb != NoWriteBack) | |
9477 | aarch64_set_reg_u64 (cpu, rn, SP_OK, address); | |
9478 | } | |
9479 | ||
9480 | /* 32 bit store scaled or unscaled zero- | |
9481 | or sign-extended 32-bit register offset. */ | |
9482 | static void | |
9483 | fstrs_scale_ext (sim_cpu *cpu, Scaling scaling, Extension extension) | |
9484 | { | |
9485 | unsigned rm = uimm (aarch64_get_instr (cpu), 20, 16); | |
9486 | unsigned rn = uimm (aarch64_get_instr (cpu), 9, 5); | |
9487 | unsigned st = uimm (aarch64_get_instr (cpu), 4, 0); | |
9488 | ||
9489 | uint64_t address = aarch64_get_reg_u64 (cpu, rn, SP_OK); | |
9490 | int64_t extended = extend (aarch64_get_reg_u32 (cpu, rm, NO_SP), | |
9491 | extension); | |
9492 | uint64_t displacement = OPT_SCALE (extended, 32, scaling); | |
9493 | ||
9494 | aarch64_set_mem_float | |
9495 | (cpu, address + displacement, aarch64_get_FP_float (cpu, st)); | |
9496 | } | |
9497 | ||
9498 | /* 64 bit store scaled unsigned 12 bit. */ | |
9499 | static void | |
9500 | fstrd_abs (sim_cpu *cpu, uint32_t offset) | |
9501 | { | |
9502 | unsigned st = uimm (aarch64_get_instr (cpu), 4, 0); | |
9503 | unsigned rn = uimm (aarch64_get_instr (cpu), 9, 5); | |
9504 | ||
9505 | aarch64_set_mem_double | |
9506 | (cpu, | |
9507 | aarch64_get_reg_u64 (cpu, rn, SP_OK) + SCALE (offset, 64), | |
9508 | aarch64_get_FP_double (cpu, st)); | |
9509 | } | |
9510 | ||
9511 | /* 64 bit store unscaled signed 9 bit with pre- or post-writeback. */ | |
9512 | static void | |
9513 | fstrd_wb (sim_cpu *cpu, int32_t offset, WriteBack wb) | |
9514 | { | |
9515 | unsigned rn = uimm (aarch64_get_instr (cpu), 9, 5); | |
9516 | unsigned st = uimm (aarch64_get_instr (cpu), 4, 0); | |
9517 | ||
9518 | uint64_t address = aarch64_get_reg_u64 (cpu, rn, SP_OK); | |
9519 | ||
9520 | if (wb != Post) | |
9521 | address += offset; | |
9522 | ||
9523 | aarch64_set_mem_double (cpu, address, aarch64_get_FP_double (cpu, st)); | |
9524 | ||
9525 | if (wb == Post) | |
9526 | address += offset; | |
9527 | ||
9528 | if (wb != NoWriteBack) | |
9529 | aarch64_set_reg_u64 (cpu, rn, SP_OK, address); | |
9530 | } | |
9531 | ||
9532 | /* 64 bit store scaled or unscaled zero- | |
9533 | or sign-extended 32-bit register offset. */ | |
9534 | static void | |
9535 | fstrd_scale_ext (sim_cpu *cpu, Scaling scaling, Extension extension) | |
9536 | { | |
9537 | unsigned rm = uimm (aarch64_get_instr (cpu), 20, 16); | |
9538 | unsigned rn = uimm (aarch64_get_instr (cpu), 9, 5); | |
9539 | unsigned st = uimm (aarch64_get_instr (cpu), 4, 0); | |
9540 | ||
9541 | uint64_t address = aarch64_get_reg_u64 (cpu, rn, SP_OK); | |
9542 | int64_t extended = extend (aarch64_get_reg_u32 (cpu, rm, NO_SP), | |
9543 | extension); | |
9544 | uint64_t displacement = OPT_SCALE (extended, 64, scaling); | |
9545 | ||
9546 | aarch64_set_mem_double | |
9547 | (cpu, address + displacement, aarch64_get_FP_double (cpu, st)); | |
9548 | } | |
9549 | ||
9550 | /* 128 bit store scaled unsigned 12 bit. */ | |
9551 | static void | |
9552 | fstrq_abs (sim_cpu *cpu, uint32_t offset) | |
9553 | { | |
9554 | FRegister a; | |
9555 | unsigned st = uimm (aarch64_get_instr (cpu), 4, 0); | |
9556 | unsigned rn = uimm (aarch64_get_instr (cpu), 9, 5); | |
9557 | uint64_t addr; | |
9558 | ||
9559 | aarch64_get_FP_long_double (cpu, st, & a); | |
9560 | ||
9561 | addr = aarch64_get_reg_u64 (cpu, rn, SP_OK) + SCALE (offset, 128); | |
9562 | aarch64_set_mem_long_double (cpu, addr, a); | |
9563 | } | |
9564 | ||
9565 | /* 128 bit store unscaled signed 9 bit with pre- or post-writeback. */ | |
9566 | static void | |
9567 | fstrq_wb (sim_cpu *cpu, int32_t offset, WriteBack wb) | |
9568 | { | |
9569 | FRegister a; | |
9570 | unsigned rn = uimm (aarch64_get_instr (cpu), 9, 5); | |
9571 | unsigned st = uimm (aarch64_get_instr (cpu), 4, 0); | |
9572 | uint64_t address = aarch64_get_reg_u64 (cpu, rn, SP_OK); | |
9573 | ||
9574 | if (wb != Post) | |
9575 | address += offset; | |
9576 | ||
9577 | aarch64_get_FP_long_double (cpu, st, & a); | |
9578 | aarch64_set_mem_long_double (cpu, address, a); | |
9579 | ||
9580 | if (wb == Post) | |
9581 | address += offset; | |
9582 | ||
9583 | if (wb != NoWriteBack) | |
9584 | aarch64_set_reg_u64 (cpu, rn, SP_OK, address); | |
9585 | } | |
9586 | ||
9587 | /* 128 bit store scaled or unscaled zero- | |
9588 | or sign-extended 32-bit register offset. */ | |
9589 | static void | |
9590 | fstrq_scale_ext (sim_cpu *cpu, Scaling scaling, Extension extension) | |
9591 | { | |
9592 | unsigned rm = uimm (aarch64_get_instr (cpu), 20, 16); | |
9593 | unsigned rn = uimm (aarch64_get_instr (cpu), 9, 5); | |
9594 | unsigned st = uimm (aarch64_get_instr (cpu), 4, 0); | |
9595 | ||
9596 | uint64_t address = aarch64_get_reg_u64 (cpu, rn, SP_OK); | |
9597 | int64_t extended = extend (aarch64_get_reg_u32 (cpu, rm, NO_SP), | |
9598 | extension); | |
9599 | uint64_t displacement = OPT_SCALE (extended, 128, scaling); | |
9600 | ||
9601 | FRegister a; | |
9602 | ||
9603 | aarch64_get_FP_long_double (cpu, st, & a); | |
9604 | aarch64_set_mem_long_double (cpu, address + displacement, a); | |
9605 | } | |
9606 | ||
9607 | static void | |
9608 | dexLoadImmediatePrePost (sim_cpu *cpu) | |
9609 | { | |
9610 | /* instr[29,24] == 111_00 | |
9611 | instr[21] == 0 | |
9612 | instr[11,10] == 00 | |
9613 | instr[31,30] = size | |
9614 | instr[26] = V | |
9615 | instr[23,22] = opc | |
9616 | instr[20,12] = simm9 | |
9617 | instr[11] = wb : 0 ==> Post, 1 ==> Pre | |
9618 | instr[9,5] = rn may be SP. */ | |
9619 | /* unsigned rt = uimm (aarch64_get_instr (cpu), 4, 0); */ | |
9620 | uint32_t V = uimm (aarch64_get_instr (cpu), 26, 26); | |
9621 | uint32_t dispatch = ( (uimm (aarch64_get_instr (cpu), 31, 30) << 2) | |
9622 | | uimm (aarch64_get_instr (cpu), 23, 22)); | |
9623 | int32_t imm = simm32 (aarch64_get_instr (cpu), 20, 12); | |
9624 | WriteBack wb = writeback (aarch64_get_instr (cpu), 11); | |
9625 | ||
9626 | if (!V) | |
9627 | { | |
9628 | /* GReg operations. */ | |
9629 | switch (dispatch) | |
9630 | { | |
9631 | case 0: strb_wb (cpu, imm, wb); return; | |
9632 | case 1: ldrb32_wb (cpu, imm, wb); return; | |
9633 | case 2: ldrsb_wb (cpu, imm, wb); return; | |
9634 | case 3: ldrsb32_wb (cpu, imm, wb); return; | |
9635 | case 4: strh_wb (cpu, imm, wb); return; | |
9636 | case 5: ldrh32_wb (cpu, imm, wb); return; | |
9637 | case 6: ldrsh64_wb (cpu, imm, wb); return; | |
9638 | case 7: ldrsh32_wb (cpu, imm, wb); return; | |
9639 | case 8: str32_wb (cpu, imm, wb); return; | |
9640 | case 9: ldr32_wb (cpu, imm, wb); return; | |
9641 | case 10: ldrsw_wb (cpu, imm, wb); return; | |
9642 | case 12: str_wb (cpu, imm, wb); return; | |
9643 | case 13: ldr_wb (cpu, imm, wb); return; | |
9644 | ||
9645 | default: | |
9646 | case 11: | |
9647 | case 14: | |
9648 | case 15: | |
9649 | HALT_UNALLOC; | |
9650 | } | |
9651 | } | |
9652 | ||
9653 | /* FReg operations. */ | |
9654 | switch (dispatch) | |
9655 | { | |
9656 | case 2: fstrq_wb (cpu, imm, wb); return; | |
9657 | case 3: fldrq_wb (cpu, imm, wb); return; | |
9658 | case 8: fstrs_wb (cpu, imm, wb); return; | |
9659 | case 9: fldrs_wb (cpu, imm, wb); return; | |
9660 | case 12: fstrd_wb (cpu, imm, wb); return; | |
9661 | case 13: fldrd_wb (cpu, imm, wb); return; | |
9662 | ||
9663 | case 0: /* STUR 8 bit FP. */ | |
9664 | case 1: /* LDUR 8 bit FP. */ | |
9665 | case 4: /* STUR 16 bit FP. */ | |
9666 | case 5: /* LDUR 8 bit FP. */ | |
9667 | HALT_NYI; | |
9668 | ||
9669 | default: | |
9670 | case 6: | |
9671 | case 7: | |
9672 | case 10: | |
9673 | case 11: | |
9674 | case 14: | |
9675 | case 15: | |
9676 | HALT_UNALLOC; | |
9677 | } | |
9678 | } | |
9679 | ||
9680 | static void | |
9681 | dexLoadRegisterOffset (sim_cpu *cpu) | |
9682 | { | |
9683 | /* instr[31,30] = size | |
9684 | instr[29,27] = 111 | |
9685 | instr[26] = V | |
9686 | instr[25,24] = 00 | |
9687 | instr[23,22] = opc | |
9688 | instr[21] = 1 | |
9689 | instr[20,16] = rm | |
9690 | instr[15,13] = option : 010 ==> UXTW, 011 ==> UXTX/LSL, | |
9691 | 110 ==> SXTW, 111 ==> SXTX, | |
9692 | ow ==> RESERVED | |
9693 | instr[12] = scaled | |
9694 | instr[11,10] = 10 | |
9695 | instr[9,5] = rn | |
9696 | instr[4,0] = rt. */ | |
9697 | ||
9698 | uint32_t V = uimm (aarch64_get_instr (cpu), 26,26); | |
9699 | uint32_t dispatch = ( (uimm (aarch64_get_instr (cpu), 31, 30) << 2) | |
9700 | | uimm (aarch64_get_instr (cpu), 23, 22)); | |
9701 | Scaling scale = scaling (aarch64_get_instr (cpu), 12); | |
9702 | Extension extensionType = extension (aarch64_get_instr (cpu), 13); | |
9703 | ||
9704 | /* Check for illegal extension types. */ | |
9705 | if (uimm (extensionType, 1, 1) == 0) | |
9706 | HALT_UNALLOC; | |
9707 | ||
9708 | if (extensionType == UXTX || extensionType == SXTX) | |
9709 | extensionType = NoExtension; | |
9710 | ||
9711 | if (!V) | |
9712 | { | |
9713 | /* GReg operations. */ | |
9714 | switch (dispatch) | |
9715 | { | |
9716 | case 0: strb_scale_ext (cpu, scale, extensionType); return; | |
9717 | case 1: ldrb32_scale_ext (cpu, scale, extensionType); return; | |
9718 | case 2: ldrsb_scale_ext (cpu, scale, extensionType); return; | |
9719 | case 3: ldrsb32_scale_ext (cpu, scale, extensionType); return; | |
9720 | case 4: strh_scale_ext (cpu, scale, extensionType); return; | |
9721 | case 5: ldrh32_scale_ext (cpu, scale, extensionType); return; | |
9722 | case 6: ldrsh_scale_ext (cpu, scale, extensionType); return; | |
9723 | case 7: ldrsh32_scale_ext (cpu, scale, extensionType); return; | |
9724 | case 8: str32_scale_ext (cpu, scale, extensionType); return; | |
9725 | case 9: ldr32_scale_ext (cpu, scale, extensionType); return; | |
9726 | case 10: ldrsw_scale_ext (cpu, scale, extensionType); return; | |
9727 | case 12: str_scale_ext (cpu, scale, extensionType); return; | |
9728 | case 13: ldr_scale_ext (cpu, scale, extensionType); return; | |
9729 | case 14: prfm_scale_ext (cpu, scale, extensionType); return; | |
9730 | ||
9731 | default: | |
9732 | case 11: | |
9733 | case 15: | |
9734 | HALT_UNALLOC; | |
9735 | } | |
9736 | } | |
9737 | ||
9738 | /* FReg operations. */ | |
9739 | switch (dispatch) | |
9740 | { | |
9741 | case 1: /* LDUR 8 bit FP. */ | |
9742 | HALT_NYI; | |
9743 | case 3: fldrq_scale_ext (cpu, scale, extensionType); return; | |
9744 | case 5: /* LDUR 8 bit FP. */ | |
9745 | HALT_NYI; | |
9746 | case 9: fldrs_scale_ext (cpu, scale, extensionType); return; | |
9747 | case 13: fldrd_scale_ext (cpu, scale, extensionType); return; | |
9748 | ||
9749 | case 0: fstrb_scale_ext (cpu, scale, extensionType); return; | |
9750 | case 2: fstrq_scale_ext (cpu, scale, extensionType); return; | |
9751 | case 4: fstrh_scale_ext (cpu, scale, extensionType); return; | |
9752 | case 8: fstrs_scale_ext (cpu, scale, extensionType); return; | |
9753 | case 12: fstrd_scale_ext (cpu, scale, extensionType); return; | |
9754 | ||
9755 | default: | |
9756 | case 6: | |
9757 | case 7: | |
9758 | case 10: | |
9759 | case 11: | |
9760 | case 14: | |
9761 | case 15: | |
9762 | HALT_UNALLOC; | |
9763 | } | |
9764 | } | |
9765 | ||
9766 | static void | |
9767 | dexLoadUnsignedImmediate (sim_cpu *cpu) | |
9768 | { | |
9769 | /* assert instr[29,24] == 111_01 | |
9770 | instr[31,30] = size | |
9771 | instr[26] = V | |
9772 | instr[23,22] = opc | |
9773 | instr[21,10] = uimm12 : unsigned immediate offset | |
9774 | instr[9,5] = rn may be SP. */ | |
9775 | /* unsigned rt = uimm (aarch64_get_instr (cpu), 4, 0); */ | |
9776 | uint32_t V = uimm (aarch64_get_instr (cpu), 26,26); | |
9777 | uint32_t dispatch = ( (uimm (aarch64_get_instr (cpu), 31, 30) << 2) | |
9778 | | uimm (aarch64_get_instr (cpu), 23, 22)); | |
9779 | uint32_t imm = uimm (aarch64_get_instr (cpu), 21, 10); | |
9780 | ||
9781 | if (!V) | |
9782 | { | |
9783 | /* GReg operations. */ | |
9784 | switch (dispatch) | |
9785 | { | |
9786 | case 0: strb_abs (cpu, imm); return; | |
9787 | case 1: ldrb32_abs (cpu, imm); return; | |
9788 | case 2: ldrsb_abs (cpu, imm); return; | |
9789 | case 3: ldrsb32_abs (cpu, imm); return; | |
9790 | case 4: strh_abs (cpu, imm); return; | |
9791 | case 5: ldrh32_abs (cpu, imm); return; | |
9792 | case 6: ldrsh_abs (cpu, imm); return; | |
9793 | case 7: ldrsh32_abs (cpu, imm); return; | |
9794 | case 8: str32_abs (cpu, imm); return; | |
9795 | case 9: ldr32_abs (cpu, imm); return; | |
9796 | case 10: ldrsw_abs (cpu, imm); return; | |
9797 | case 12: str_abs (cpu, imm); return; | |
9798 | case 13: ldr_abs (cpu, imm); return; | |
9799 | case 14: prfm_abs (cpu, imm); return; | |
9800 | ||
9801 | default: | |
9802 | case 11: | |
9803 | case 15: | |
9804 | HALT_UNALLOC; | |
9805 | } | |
9806 | } | |
9807 | ||
9808 | /* FReg operations. */ | |
9809 | switch (dispatch) | |
9810 | { | |
9811 | case 3: fldrq_abs (cpu, imm); return; | |
9812 | case 9: fldrs_abs (cpu, imm); return; | |
9813 | case 13: fldrd_abs (cpu, imm); return; | |
9814 | ||
9815 | case 0: fstrb_abs (cpu, imm); return; | |
9816 | case 2: fstrq_abs (cpu, imm); return; | |
9817 | case 4: fstrh_abs (cpu, imm); return; | |
9818 | case 8: fstrs_abs (cpu, imm); return; | |
9819 | case 12: fstrd_abs (cpu, imm); return; | |
9820 | ||
9821 | case 1: /* LDR 8 bit FP. */ | |
9822 | case 5: /* LDR 8 bit FP. */ | |
9823 | HALT_NYI; | |
9824 | ||
9825 | default: | |
9826 | case 6: | |
9827 | case 7: | |
9828 | case 10: | |
9829 | case 11: | |
9830 | case 14: | |
9831 | case 15: | |
9832 | HALT_UNALLOC; | |
9833 | } | |
9834 | } | |
9835 | ||
9836 | static void | |
9837 | dexLoadExclusive (sim_cpu *cpu) | |
9838 | { | |
9839 | /* assert instr[29:24] = 001000; | |
9840 | instr[31,30] = size | |
9841 | instr[23] = 0 if exclusive | |
9842 | instr[22] = L : 1 if load, 0 if store | |
9843 | instr[21] = 1 if pair | |
9844 | instr[20,16] = Rs | |
9845 | instr[15] = o0 : 1 if ordered | |
9846 | instr[14,10] = Rt2 | |
9847 | instr[9,5] = Rn | |
9848 | instr[4.0] = Rt. */ | |
9849 | ||
9850 | switch (uimm (aarch64_get_instr (cpu), 22, 21)) | |
9851 | { | |
9852 | case 2: ldxr (cpu); return; | |
9853 | case 0: stxr (cpu); return; | |
9854 | default: HALT_NYI; | |
9855 | } | |
9856 | } | |
9857 | ||
9858 | static void | |
9859 | dexLoadOther (sim_cpu *cpu) | |
9860 | { | |
9861 | uint32_t dispatch; | |
9862 | ||
9863 | /* instr[29,25] = 111_0 | |
9864 | instr[24] == 0 ==> dispatch, 1 ==> ldst reg unsigned immediate | |
9865 | instr[21:11,10] is the secondary dispatch. */ | |
9866 | if (uimm (aarch64_get_instr (cpu), 24, 24)) | |
9867 | { | |
9868 | dexLoadUnsignedImmediate (cpu); | |
9869 | return; | |
9870 | } | |
9871 | ||
9872 | dispatch = ( (uimm (aarch64_get_instr (cpu), 21, 21) << 2) | |
9873 | | uimm (aarch64_get_instr (cpu), 11, 10)); | |
9874 | switch (dispatch) | |
9875 | { | |
9876 | case 0: dexLoadUnscaledImmediate (cpu); return; | |
9877 | case 1: dexLoadImmediatePrePost (cpu); return; | |
9878 | case 3: dexLoadImmediatePrePost (cpu); return; | |
9879 | case 6: dexLoadRegisterOffset (cpu); return; | |
9880 | ||
9881 | default: | |
9882 | case 2: | |
9883 | case 4: | |
9884 | case 5: | |
9885 | case 7: | |
9886 | HALT_NYI; | |
9887 | } | |
9888 | } | |
9889 | ||
9890 | static void | |
9891 | store_pair_u32 (sim_cpu *cpu, int32_t offset, WriteBack wb) | |
9892 | { | |
9893 | unsigned rn = uimm (aarch64_get_instr (cpu), 14, 10); | |
9894 | unsigned rd = uimm (aarch64_get_instr (cpu), 9, 5); | |
9895 | unsigned rm = uimm (aarch64_get_instr (cpu), 4, 0); | |
9896 | uint64_t address = aarch64_get_reg_u64 (cpu, rd, SP_OK); | |
9897 | ||
9898 | if ((rn == rd || rm == rd) && wb != NoWriteBack) | |
9899 | HALT_UNALLOC; /* ??? */ | |
9900 | ||
9901 | offset <<= 2; | |
9902 | ||
9903 | if (wb != Post) | |
9904 | address += offset; | |
9905 | ||
9906 | aarch64_set_mem_u32 (cpu, address, | |
9907 | aarch64_get_reg_u32 (cpu, rm, NO_SP)); | |
9908 | aarch64_set_mem_u32 (cpu, address + 4, | |
9909 | aarch64_get_reg_u32 (cpu, rn, NO_SP)); | |
9910 | ||
9911 | if (wb == Post) | |
9912 | address += offset; | |
9913 | ||
9914 | if (wb != NoWriteBack) | |
9915 | aarch64_set_reg_u64 (cpu, rd, SP_OK, address); | |
9916 | } | |
9917 | ||
9918 | static void | |
9919 | store_pair_u64 (sim_cpu *cpu, int32_t offset, WriteBack wb) | |
9920 | { | |
9921 | unsigned rn = uimm (aarch64_get_instr (cpu), 14, 10); | |
9922 | unsigned rd = uimm (aarch64_get_instr (cpu), 9, 5); | |
9923 | unsigned rm = uimm (aarch64_get_instr (cpu), 4, 0); | |
9924 | uint64_t address = aarch64_get_reg_u64 (cpu, rd, SP_OK); | |
9925 | ||
9926 | if ((rn == rd || rm == rd) && wb != NoWriteBack) | |
9927 | HALT_UNALLOC; /* ??? */ | |
9928 | ||
9929 | offset <<= 3; | |
9930 | ||
9931 | if (wb != Post) | |
9932 | address += offset; | |
9933 | ||
9934 | aarch64_set_mem_u64 (cpu, address, | |
9935 | aarch64_get_reg_u64 (cpu, rm, SP_OK)); | |
9936 | aarch64_set_mem_u64 (cpu, address + 8, | |
9937 | aarch64_get_reg_u64 (cpu, rn, SP_OK)); | |
9938 | ||
9939 | if (wb == Post) | |
9940 | address += offset; | |
9941 | ||
9942 | if (wb != NoWriteBack) | |
9943 | aarch64_set_reg_u64 (cpu, rd, SP_OK, address); | |
9944 | } | |
9945 | ||
9946 | static void | |
9947 | load_pair_u32 (sim_cpu *cpu, int32_t offset, WriteBack wb) | |
9948 | { | |
9949 | unsigned rn = uimm (aarch64_get_instr (cpu), 14, 10); | |
9950 | unsigned rd = uimm (aarch64_get_instr (cpu), 9, 5); | |
9951 | unsigned rm = uimm (aarch64_get_instr (cpu), 4, 0); | |
9952 | uint64_t address = aarch64_get_reg_u64 (cpu, rd, SP_OK); | |
9953 | ||
9954 | /* treat this as unalloc to make sure we don't do it. */ | |
9955 | if (rn == rm) | |
9956 | HALT_UNALLOC; | |
9957 | ||
9958 | offset <<= 2; | |
9959 | ||
9960 | if (wb != Post) | |
9961 | address += offset; | |
9962 | ||
9963 | aarch64_set_reg_u64 (cpu, rm, SP_OK, aarch64_get_mem_u32 (cpu, address)); | |
9964 | aarch64_set_reg_u64 (cpu, rn, SP_OK, aarch64_get_mem_u32 (cpu, address + 4)); | |
9965 | ||
9966 | if (wb == Post) | |
9967 | address += offset; | |
9968 | ||
9969 | if (wb != NoWriteBack) | |
9970 | aarch64_set_reg_u64 (cpu, rd, SP_OK, address); | |
9971 | } | |
9972 | ||
9973 | static void | |
9974 | load_pair_s32 (sim_cpu *cpu, int32_t offset, WriteBack wb) | |
9975 | { | |
9976 | unsigned rn = uimm (aarch64_get_instr (cpu), 14, 10); | |
9977 | unsigned rd = uimm (aarch64_get_instr (cpu), 9, 5); | |
9978 | unsigned rm = uimm (aarch64_get_instr (cpu), 4, 0); | |
9979 | uint64_t address = aarch64_get_reg_u64 (cpu, rd, SP_OK); | |
9980 | ||
9981 | /* Treat this as unalloc to make sure we don't do it. */ | |
9982 | if (rn == rm) | |
9983 | HALT_UNALLOC; | |
9984 | ||
9985 | offset <<= 2; | |
9986 | ||
9987 | if (wb != Post) | |
9988 | address += offset; | |
9989 | ||
9990 | aarch64_set_reg_s64 (cpu, rm, SP_OK, aarch64_get_mem_s32 (cpu, address)); | |
9991 | aarch64_set_reg_s64 (cpu, rn, SP_OK, aarch64_get_mem_s32 (cpu, address + 4)); | |
9992 | ||
9993 | if (wb == Post) | |
9994 | address += offset; | |
9995 | ||
9996 | if (wb != NoWriteBack) | |
9997 | aarch64_set_reg_u64 (cpu, rd, SP_OK, address); | |
9998 | } | |
9999 | ||
10000 | static void | |
10001 | load_pair_u64 (sim_cpu *cpu, int32_t offset, WriteBack wb) | |
10002 | { | |
10003 | unsigned rn = uimm (aarch64_get_instr (cpu), 14, 10); | |
10004 | unsigned rd = uimm (aarch64_get_instr (cpu), 9, 5); | |
10005 | unsigned rm = uimm (aarch64_get_instr (cpu), 4, 0); | |
10006 | uint64_t address = aarch64_get_reg_u64 (cpu, rd, SP_OK); | |
10007 | ||
10008 | /* Treat this as unalloc to make sure we don't do it. */ | |
10009 | if (rn == rm) | |
10010 | HALT_UNALLOC; | |
10011 | ||
10012 | offset <<= 3; | |
10013 | ||
10014 | if (wb != Post) | |
10015 | address += offset; | |
10016 | ||
10017 | aarch64_set_reg_u64 (cpu, rm, SP_OK, aarch64_get_mem_u64 (cpu, address)); | |
10018 | aarch64_set_reg_u64 (cpu, rn, SP_OK, aarch64_get_mem_u64 (cpu, address + 8)); | |
10019 | ||
10020 | if (wb == Post) | |
10021 | address += offset; | |
10022 | ||
10023 | if (wb != NoWriteBack) | |
10024 | aarch64_set_reg_u64 (cpu, rd, SP_OK, address); | |
10025 | } | |
10026 | ||
10027 | static void | |
10028 | dex_load_store_pair_gr (sim_cpu *cpu) | |
10029 | { | |
10030 | /* instr[31,30] = size (10=> 64-bit, 01=> signed 32-bit, 00=> 32-bit) | |
10031 | instr[29,25] = instruction encoding: 101_0 | |
10032 | instr[26] = V : 1 if fp 0 if gp | |
10033 | instr[24,23] = addressing mode (10=> offset, 01=> post, 11=> pre) | |
10034 | instr[22] = load/store (1=> load) | |
10035 | instr[21,15] = signed, scaled, offset | |
10036 | instr[14,10] = Rn | |
10037 | instr[ 9, 5] = Rd | |
10038 | instr[ 4, 0] = Rm. */ | |
10039 | ||
10040 | uint32_t dispatch = ((uimm (aarch64_get_instr (cpu), 31, 30) << 3) | |
10041 | | uimm (aarch64_get_instr (cpu), 24, 22)); | |
10042 | int32_t offset = simm32 (aarch64_get_instr (cpu), 21, 15); | |
10043 | ||
10044 | switch (dispatch) | |
10045 | { | |
10046 | case 2: store_pair_u32 (cpu, offset, Post); return; | |
10047 | case 3: load_pair_u32 (cpu, offset, Post); return; | |
10048 | case 4: store_pair_u32 (cpu, offset, NoWriteBack); return; | |
10049 | case 5: load_pair_u32 (cpu, offset, NoWriteBack); return; | |
10050 | case 6: store_pair_u32 (cpu, offset, Pre); return; | |
10051 | case 7: load_pair_u32 (cpu, offset, Pre); return; | |
10052 | ||
10053 | case 11: load_pair_s32 (cpu, offset, Post); return; | |
10054 | case 13: load_pair_s32 (cpu, offset, NoWriteBack); return; | |
10055 | case 15: load_pair_s32 (cpu, offset, Pre); return; | |
10056 | ||
10057 | case 18: store_pair_u64 (cpu, offset, Post); return; | |
10058 | case 19: load_pair_u64 (cpu, offset, Post); return; | |
10059 | case 20: store_pair_u64 (cpu, offset, NoWriteBack); return; | |
10060 | case 21: load_pair_u64 (cpu, offset, NoWriteBack); return; | |
10061 | case 22: store_pair_u64 (cpu, offset, Pre); return; | |
10062 | case 23: load_pair_u64 (cpu, offset, Pre); return; | |
10063 | ||
10064 | default: | |
10065 | HALT_UNALLOC; | |
10066 | } | |
10067 | } | |
10068 | ||
10069 | static void | |
10070 | store_pair_float (sim_cpu *cpu, int32_t offset, WriteBack wb) | |
10071 | { | |
10072 | unsigned rn = uimm (aarch64_get_instr (cpu), 14, 10); | |
10073 | unsigned rd = uimm (aarch64_get_instr (cpu), 9, 5); | |
10074 | unsigned rm = uimm (aarch64_get_instr (cpu), 4, 0); | |
10075 | uint64_t address = aarch64_get_reg_u64 (cpu, rd, SP_OK); | |
10076 | ||
10077 | offset <<= 2; | |
10078 | ||
10079 | if (wb != Post) | |
10080 | address += offset; | |
10081 | ||
10082 | aarch64_set_mem_float (cpu, address, aarch64_get_FP_float (cpu, rm)); | |
10083 | aarch64_set_mem_float (cpu, address + 4, aarch64_get_FP_float (cpu, rn)); | |
10084 | ||
10085 | if (wb == Post) | |
10086 | address += offset; | |
10087 | ||
10088 | if (wb != NoWriteBack) | |
10089 | aarch64_set_reg_u64 (cpu, rd, SP_OK, address); | |
10090 | } | |
10091 | ||
10092 | static void | |
10093 | store_pair_double (sim_cpu *cpu, int32_t offset, WriteBack wb) | |
10094 | { | |
10095 | unsigned rn = uimm (aarch64_get_instr (cpu), 14, 10); | |
10096 | unsigned rd = uimm (aarch64_get_instr (cpu), 9, 5); | |
10097 | unsigned rm = uimm (aarch64_get_instr (cpu), 4, 0); | |
10098 | uint64_t address = aarch64_get_reg_u64 (cpu, rd, SP_OK); | |
10099 | ||
10100 | offset <<= 3; | |
10101 | ||
10102 | if (wb != Post) | |
10103 | address += offset; | |
10104 | ||
10105 | aarch64_set_mem_double (cpu, address, aarch64_get_FP_double (cpu, rm)); | |
10106 | aarch64_set_mem_double (cpu, address + 8, aarch64_get_FP_double (cpu, rn)); | |
10107 | ||
10108 | if (wb == Post) | |
10109 | address += offset; | |
10110 | ||
10111 | if (wb != NoWriteBack) | |
10112 | aarch64_set_reg_u64 (cpu, rd, SP_OK, address); | |
10113 | } | |
10114 | ||
10115 | static void | |
10116 | store_pair_long_double (sim_cpu *cpu, int32_t offset, WriteBack wb) | |
10117 | { | |
10118 | FRegister a; | |
10119 | unsigned rn = uimm (aarch64_get_instr (cpu), 14, 10); | |
10120 | unsigned rd = uimm (aarch64_get_instr (cpu), 9, 5); | |
10121 | unsigned rm = uimm (aarch64_get_instr (cpu), 4, 0); | |
10122 | uint64_t address = aarch64_get_reg_u64 (cpu, rd, SP_OK); | |
10123 | ||
10124 | offset <<= 4; | |
10125 | ||
10126 | if (wb != Post) | |
10127 | address += offset; | |
10128 | ||
10129 | aarch64_get_FP_long_double (cpu, rm, & a); | |
10130 | aarch64_set_mem_long_double (cpu, address, a); | |
10131 | aarch64_get_FP_long_double (cpu, rn, & a); | |
10132 | aarch64_set_mem_long_double (cpu, address + 16, a); | |
10133 | ||
10134 | if (wb == Post) | |
10135 | address += offset; | |
10136 | ||
10137 | if (wb != NoWriteBack) | |
10138 | aarch64_set_reg_u64 (cpu, rd, SP_OK, address); | |
10139 | } | |
10140 | ||
10141 | static void | |
10142 | load_pair_float (sim_cpu *cpu, int32_t offset, WriteBack wb) | |
10143 | { | |
10144 | unsigned rn = uimm (aarch64_get_instr (cpu), 14, 10); | |
10145 | unsigned rd = uimm (aarch64_get_instr (cpu), 9, 5); | |
10146 | unsigned rm = uimm (aarch64_get_instr (cpu), 4, 0); | |
10147 | uint64_t address = aarch64_get_reg_u64 (cpu, rd, SP_OK); | |
10148 | ||
10149 | if (rm == rn) | |
10150 | HALT_UNALLOC; | |
10151 | ||
10152 | offset <<= 2; | |
10153 | ||
10154 | if (wb != Post) | |
10155 | address += offset; | |
10156 | ||
10157 | aarch64_set_FP_float (cpu, rm, aarch64_get_mem_float (cpu, address)); | |
10158 | aarch64_set_FP_float (cpu, rn, aarch64_get_mem_float (cpu, address + 4)); | |
10159 | ||
10160 | if (wb == Post) | |
10161 | address += offset; | |
10162 | ||
10163 | if (wb != NoWriteBack) | |
10164 | aarch64_set_reg_u64 (cpu, rd, SP_OK, address); | |
10165 | } | |
10166 | ||
10167 | static void | |
10168 | load_pair_double (sim_cpu *cpu, int32_t offset, WriteBack wb) | |
10169 | { | |
10170 | unsigned rn = uimm (aarch64_get_instr (cpu), 14, 10); | |
10171 | unsigned rd = uimm (aarch64_get_instr (cpu), 9, 5); | |
10172 | unsigned rm = uimm (aarch64_get_instr (cpu), 4, 0); | |
10173 | uint64_t address = aarch64_get_reg_u64 (cpu, rd, SP_OK); | |
10174 | ||
10175 | if (rm == rn) | |
10176 | HALT_UNALLOC; | |
10177 | ||
10178 | offset <<= 3; | |
10179 | ||
10180 | if (wb != Post) | |
10181 | address += offset; | |
10182 | ||
10183 | aarch64_set_FP_double (cpu, rm, aarch64_get_mem_double (cpu, address)); | |
10184 | aarch64_set_FP_double (cpu, rn, aarch64_get_mem_double (cpu, address + 8)); | |
10185 | ||
10186 | if (wb == Post) | |
10187 | address += offset; | |
10188 | ||
10189 | if (wb != NoWriteBack) | |
10190 | aarch64_set_reg_u64 (cpu, rd, SP_OK, address); | |
10191 | } | |
10192 | ||
10193 | static void | |
10194 | load_pair_long_double (sim_cpu *cpu, int32_t offset, WriteBack wb) | |
10195 | { | |
10196 | FRegister a; | |
10197 | unsigned rn = uimm (aarch64_get_instr (cpu), 14, 10); | |
10198 | unsigned rd = uimm (aarch64_get_instr (cpu), 9, 5); | |
10199 | unsigned rm = uimm (aarch64_get_instr (cpu), 4, 0); | |
10200 | uint64_t address = aarch64_get_reg_u64 (cpu, rd, SP_OK); | |
10201 | ||
10202 | if (rm == rn) | |
10203 | HALT_UNALLOC; | |
10204 | ||
10205 | offset <<= 4; | |
10206 | ||
10207 | if (wb != Post) | |
10208 | address += offset; | |
10209 | ||
10210 | aarch64_get_mem_long_double (cpu, address, & a); | |
10211 | aarch64_set_FP_long_double (cpu, rm, a); | |
10212 | aarch64_get_mem_long_double (cpu, address + 16, & a); | |
10213 | aarch64_set_FP_long_double (cpu, rn, a); | |
10214 | ||
10215 | if (wb == Post) | |
10216 | address += offset; | |
10217 | ||
10218 | if (wb != NoWriteBack) | |
10219 | aarch64_set_reg_u64 (cpu, rd, SP_OK, address); | |
10220 | } | |
10221 | ||
10222 | static void | |
10223 | dex_load_store_pair_fp (sim_cpu *cpu) | |
10224 | { | |
10225 | /* instr[31,30] = size (10=> 128-bit, 01=> 64-bit, 00=> 32-bit) | |
10226 | instr[29,25] = instruction encoding | |
10227 | instr[24,23] = addressing mode (10=> offset, 01=> post, 11=> pre) | |
10228 | instr[22] = load/store (1=> load) | |
10229 | instr[21,15] = signed, scaled, offset | |
10230 | instr[14,10] = Rn | |
10231 | instr[ 9, 5] = Rd | |
10232 | instr[ 4, 0] = Rm */ | |
10233 | ||
10234 | uint32_t dispatch = ((uimm (aarch64_get_instr (cpu), 31, 30) << 3) | |
10235 | | uimm (aarch64_get_instr (cpu), 24, 22)); | |
10236 | int32_t offset = simm32 (aarch64_get_instr (cpu), 21, 15); | |
10237 | ||
10238 | switch (dispatch) | |
10239 | { | |
10240 | case 2: store_pair_float (cpu, offset, Post); return; | |
10241 | case 3: load_pair_float (cpu, offset, Post); return; | |
10242 | case 4: store_pair_float (cpu, offset, NoWriteBack); return; | |
10243 | case 5: load_pair_float (cpu, offset, NoWriteBack); return; | |
10244 | case 6: store_pair_float (cpu, offset, Pre); return; | |
10245 | case 7: load_pair_float (cpu, offset, Pre); return; | |
10246 | ||
10247 | case 10: store_pair_double (cpu, offset, Post); return; | |
10248 | case 11: load_pair_double (cpu, offset, Post); return; | |
10249 | case 12: store_pair_double (cpu, offset, NoWriteBack); return; | |
10250 | case 13: load_pair_double (cpu, offset, NoWriteBack); return; | |
10251 | case 14: store_pair_double (cpu, offset, Pre); return; | |
10252 | case 15: load_pair_double (cpu, offset, Pre); return; | |
10253 | ||
10254 | case 18: store_pair_long_double (cpu, offset, Post); return; | |
10255 | case 19: load_pair_long_double (cpu, offset, Post); return; | |
10256 | case 20: store_pair_long_double (cpu, offset, NoWriteBack); return; | |
10257 | case 21: load_pair_long_double (cpu, offset, NoWriteBack); return; | |
10258 | case 22: store_pair_long_double (cpu, offset, Pre); return; | |
10259 | case 23: load_pair_long_double (cpu, offset, Pre); return; | |
10260 | ||
10261 | default: | |
10262 | HALT_UNALLOC; | |
10263 | } | |
10264 | } | |
10265 | ||
10266 | static inline unsigned | |
10267 | vec_reg (unsigned v, unsigned o) | |
10268 | { | |
10269 | return (v + o) & 0x3F; | |
10270 | } | |
10271 | ||
10272 | /* Load multiple N-element structures to N consecutive registers. */ | |
10273 | static void | |
10274 | vec_load (sim_cpu *cpu, uint64_t address, unsigned N) | |
10275 | { | |
10276 | int all = uimm (aarch64_get_instr (cpu), 30, 30); | |
10277 | unsigned size = uimm (aarch64_get_instr (cpu), 11, 10); | |
10278 | unsigned vd = uimm (aarch64_get_instr (cpu), 4, 0); | |
10279 | unsigned i; | |
10280 | ||
10281 | switch (size) | |
10282 | { | |
10283 | case 0: /* 8-bit operations. */ | |
10284 | if (all) | |
10285 | for (i = 0; i < (16 * N); i++) | |
10286 | aarch64_set_vec_u8 (cpu, vec_reg (vd, i >> 4), i & 15, | |
10287 | aarch64_get_mem_u8 (cpu, address + i)); | |
10288 | else | |
10289 | for (i = 0; i < (8 * N); i++) | |
10290 | aarch64_set_vec_u8 (cpu, vec_reg (vd, i >> 3), i & 7, | |
10291 | aarch64_get_mem_u8 (cpu, address + i)); | |
10292 | return; | |
10293 | ||
10294 | case 1: /* 16-bit operations. */ | |
10295 | if (all) | |
10296 | for (i = 0; i < (8 * N); i++) | |
10297 | aarch64_set_vec_u16 (cpu, vec_reg (vd, i >> 3), i & 7, | |
10298 | aarch64_get_mem_u16 (cpu, address + i * 2)); | |
10299 | else | |
10300 | for (i = 0; i < (4 * N); i++) | |
10301 | aarch64_set_vec_u16 (cpu, vec_reg (vd, i >> 2), i & 3, | |
10302 | aarch64_get_mem_u16 (cpu, address + i * 2)); | |
10303 | return; | |
10304 | ||
10305 | case 2: /* 32-bit operations. */ | |
10306 | if (all) | |
10307 | for (i = 0; i < (4 * N); i++) | |
10308 | aarch64_set_vec_u32 (cpu, vec_reg (vd, i >> 2), i & 3, | |
10309 | aarch64_get_mem_u32 (cpu, address + i * 4)); | |
10310 | else | |
10311 | for (i = 0; i < (2 * N); i++) | |
10312 | aarch64_set_vec_u32 (cpu, vec_reg (vd, i >> 1), i & 1, | |
10313 | aarch64_get_mem_u32 (cpu, address + i * 4)); | |
10314 | return; | |
10315 | ||
10316 | case 3: /* 64-bit operations. */ | |
10317 | if (all) | |
10318 | for (i = 0; i < (2 * N); i++) | |
10319 | aarch64_set_vec_u64 (cpu, vec_reg (vd, i >> 1), i & 1, | |
10320 | aarch64_get_mem_u64 (cpu, address + i * 8)); | |
10321 | else | |
10322 | for (i = 0; i < N; i++) | |
10323 | aarch64_set_vec_u64 (cpu, vec_reg (vd, i), 0, | |
10324 | aarch64_get_mem_u64 (cpu, address + i * 8)); | |
10325 | return; | |
10326 | ||
10327 | default: | |
10328 | HALT_UNREACHABLE; | |
10329 | } | |
10330 | } | |
10331 | ||
10332 | /* LD4: load multiple 4-element to four consecutive registers. */ | |
10333 | static void | |
10334 | LD4 (sim_cpu *cpu, uint64_t address) | |
10335 | { | |
10336 | vec_load (cpu, address, 4); | |
10337 | } | |
10338 | ||
10339 | /* LD3: load multiple 3-element structures to three consecutive registers. */ | |
10340 | static void | |
10341 | LD3 (sim_cpu *cpu, uint64_t address) | |
10342 | { | |
10343 | vec_load (cpu, address, 3); | |
10344 | } | |
10345 | ||
10346 | /* LD2: load multiple 2-element structures to two consecutive registers. */ | |
10347 | static void | |
10348 | LD2 (sim_cpu *cpu, uint64_t address) | |
10349 | { | |
10350 | vec_load (cpu, address, 2); | |
10351 | } | |
10352 | ||
10353 | /* Load multiple 1-element structures into one register. */ | |
10354 | static void | |
10355 | LD1_1 (sim_cpu *cpu, uint64_t address) | |
10356 | { | |
10357 | int all = uimm (aarch64_get_instr (cpu), 30, 30); | |
10358 | unsigned size = uimm (aarch64_get_instr (cpu), 11, 10); | |
10359 | unsigned vd = uimm (aarch64_get_instr (cpu), 4, 0); | |
10360 | unsigned i; | |
10361 | ||
10362 | switch (size) | |
10363 | { | |
10364 | case 0: | |
10365 | /* LD1 {Vd.16b}, addr, #16 */ | |
10366 | /* LD1 {Vd.8b}, addr, #8 */ | |
10367 | for (i = 0; i < (all ? 16 : 8); i++) | |
10368 | aarch64_set_vec_u8 (cpu, vd, i, | |
10369 | aarch64_get_mem_u8 (cpu, address + i)); | |
10370 | return; | |
10371 | ||
10372 | case 1: | |
10373 | /* LD1 {Vd.8h}, addr, #16 */ | |
10374 | /* LD1 {Vd.4h}, addr, #8 */ | |
10375 | for (i = 0; i < (all ? 8 : 4); i++) | |
10376 | aarch64_set_vec_u16 (cpu, vd, i, | |
10377 | aarch64_get_mem_u16 (cpu, address + i * 2)); | |
10378 | return; | |
10379 | ||
10380 | case 2: | |
10381 | /* LD1 {Vd.4s}, addr, #16 */ | |
10382 | /* LD1 {Vd.2s}, addr, #8 */ | |
10383 | for (i = 0; i < (all ? 4 : 2); i++) | |
10384 | aarch64_set_vec_u32 (cpu, vd, i, | |
10385 | aarch64_get_mem_u32 (cpu, address + i * 4)); | |
10386 | return; | |
10387 | ||
10388 | case 3: | |
10389 | /* LD1 {Vd.2d}, addr, #16 */ | |
10390 | /* LD1 {Vd.1d}, addr, #8 */ | |
10391 | for (i = 0; i < (all ? 2 : 1); i++) | |
10392 | aarch64_set_vec_u64 (cpu, vd, i, | |
10393 | aarch64_get_mem_u64 (cpu, address + i * 8)); | |
10394 | return; | |
10395 | ||
10396 | default: | |
10397 | HALT_UNREACHABLE; | |
10398 | } | |
10399 | } | |
10400 | ||
10401 | /* Load multiple 1-element structures into two registers. */ | |
10402 | static void | |
10403 | LD1_2 (sim_cpu *cpu, uint64_t address) | |
10404 | { | |
10405 | /* FIXME: This algorithm is *exactly* the same as the LD2 version. | |
10406 | So why have two different instructions ? There must be something | |
10407 | wrong somewhere. */ | |
10408 | vec_load (cpu, address, 2); | |
10409 | } | |
10410 | ||
10411 | /* Load multiple 1-element structures into three registers. */ | |
10412 | static void | |
10413 | LD1_3 (sim_cpu *cpu, uint64_t address) | |
10414 | { | |
10415 | /* FIXME: This algorithm is *exactly* the same as the LD3 version. | |
10416 | So why have two different instructions ? There must be something | |
10417 | wrong somewhere. */ | |
10418 | vec_load (cpu, address, 3); | |
10419 | } | |
10420 | ||
10421 | /* Load multiple 1-element structures into four registers. */ | |
10422 | static void | |
10423 | LD1_4 (sim_cpu *cpu, uint64_t address) | |
10424 | { | |
10425 | /* FIXME: This algorithm is *exactly* the same as the LD4 version. | |
10426 | So why have two different instructions ? There must be something | |
10427 | wrong somewhere. */ | |
10428 | vec_load (cpu, address, 4); | |
10429 | } | |
10430 | ||
10431 | /* Store multiple N-element structures to N consecutive registers. */ | |
10432 | static void | |
10433 | vec_store (sim_cpu *cpu, uint64_t address, unsigned N) | |
10434 | { | |
10435 | int all = uimm (aarch64_get_instr (cpu), 30, 30); | |
10436 | unsigned size = uimm (aarch64_get_instr (cpu), 11, 10); | |
10437 | unsigned vd = uimm (aarch64_get_instr (cpu), 4, 0); | |
10438 | unsigned i; | |
10439 | ||
10440 | switch (size) | |
10441 | { | |
10442 | case 0: /* 8-bit operations. */ | |
10443 | if (all) | |
10444 | for (i = 0; i < (16 * N); i++) | |
10445 | aarch64_set_mem_u8 | |
10446 | (cpu, address + i, | |
10447 | aarch64_get_vec_u8 (cpu, vec_reg (vd, i >> 4), i & 15)); | |
10448 | else | |
10449 | for (i = 0; i < (8 * N); i++) | |
10450 | aarch64_set_mem_u8 | |
10451 | (cpu, address + i, | |
10452 | aarch64_get_vec_u8 (cpu, vec_reg (vd, i >> 3), i & 7)); | |
10453 | return; | |
10454 | ||
10455 | case 1: /* 16-bit operations. */ | |
10456 | if (all) | |
10457 | for (i = 0; i < (8 * N); i++) | |
10458 | aarch64_set_mem_u16 | |
10459 | (cpu, address + i * 2, | |
10460 | aarch64_get_vec_u16 (cpu, vec_reg (vd, i >> 3), i & 7)); | |
10461 | else | |
10462 | for (i = 0; i < (4 * N); i++) | |
10463 | aarch64_set_mem_u16 | |
10464 | (cpu, address + i * 2, | |
10465 | aarch64_get_vec_u16 (cpu, vec_reg (vd, i >> 2), i & 3)); | |
10466 | return; | |
10467 | ||
10468 | case 2: /* 32-bit operations. */ | |
10469 | if (all) | |
10470 | for (i = 0; i < (4 * N); i++) | |
10471 | aarch64_set_mem_u32 | |
10472 | (cpu, address + i * 4, | |
10473 | aarch64_get_vec_u32 (cpu, vec_reg (vd, i >> 2), i & 3)); | |
10474 | else | |
10475 | for (i = 0; i < (2 * N); i++) | |
10476 | aarch64_set_mem_u32 | |
10477 | (cpu, address + i * 4, | |
10478 | aarch64_get_vec_u32 (cpu, vec_reg (vd, i >> 1), i & 1)); | |
10479 | return; | |
10480 | ||
10481 | case 3: /* 64-bit operations. */ | |
10482 | if (all) | |
10483 | for (i = 0; i < (2 * N); i++) | |
10484 | aarch64_set_mem_u64 | |
10485 | (cpu, address + i * 8, | |
10486 | aarch64_get_vec_u64 (cpu, vec_reg (vd, i >> 1), i & 1)); | |
10487 | else | |
10488 | for (i = 0; i < N; i++) | |
10489 | aarch64_set_mem_u64 | |
10490 | (cpu, address + i * 8, | |
10491 | aarch64_get_vec_u64 (cpu, vec_reg (vd, i), 0)); | |
10492 | return; | |
10493 | ||
10494 | default: | |
10495 | HALT_UNREACHABLE; | |
10496 | } | |
10497 | } | |
10498 | ||
10499 | /* Store multiple 4-element structure to four consecutive registers. */ | |
10500 | static void | |
10501 | ST4 (sim_cpu *cpu, uint64_t address) | |
10502 | { | |
10503 | vec_store (cpu, address, 4); | |
10504 | } | |
10505 | ||
10506 | /* Store multiple 3-element structures to three consecutive registers. */ | |
10507 | static void | |
10508 | ST3 (sim_cpu *cpu, uint64_t address) | |
10509 | { | |
10510 | vec_store (cpu, address, 3); | |
10511 | } | |
10512 | ||
10513 | /* Store multiple 2-element structures to two consecutive registers. */ | |
10514 | static void | |
10515 | ST2 (sim_cpu *cpu, uint64_t address) | |
10516 | { | |
10517 | vec_store (cpu, address, 2); | |
10518 | } | |
10519 | ||
10520 | /* Store multiple 1-element structures into one register. */ | |
10521 | static void | |
10522 | ST1_1 (sim_cpu *cpu, uint64_t address) | |
10523 | { | |
10524 | int all = uimm (aarch64_get_instr (cpu), 30, 30); | |
10525 | unsigned size = uimm (aarch64_get_instr (cpu), 11, 10); | |
10526 | unsigned vd = uimm (aarch64_get_instr (cpu), 4, 0); | |
10527 | unsigned i; | |
10528 | ||
10529 | switch (size) | |
10530 | { | |
10531 | case 0: | |
10532 | for (i = 0; i < (all ? 16 : 8); i++) | |
10533 | aarch64_set_mem_u8 (cpu, address + i, | |
10534 | aarch64_get_vec_u8 (cpu, vd, i)); | |
10535 | return; | |
10536 | ||
10537 | case 1: | |
10538 | for (i = 0; i < (all ? 8 : 4); i++) | |
10539 | aarch64_set_mem_u16 (cpu, address + i * 2, | |
10540 | aarch64_get_vec_u16 (cpu, vd, i)); | |
10541 | return; | |
10542 | ||
10543 | case 2: | |
10544 | for (i = 0; i < (all ? 4 : 2); i++) | |
10545 | aarch64_set_mem_u32 (cpu, address + i * 4, | |
10546 | aarch64_get_vec_u32 (cpu, vd, i)); | |
10547 | return; | |
10548 | ||
10549 | case 3: | |
10550 | for (i = 0; i < (all ? 2 : 1); i++) | |
10551 | aarch64_set_mem_u64 (cpu, address + i * 8, | |
10552 | aarch64_get_vec_u64 (cpu, vd, i)); | |
10553 | return; | |
10554 | ||
10555 | default: | |
10556 | HALT_UNREACHABLE; | |
10557 | } | |
10558 | } | |
10559 | ||
10560 | /* Store multiple 1-element structures into two registers. */ | |
10561 | static void | |
10562 | ST1_2 (sim_cpu *cpu, uint64_t address) | |
10563 | { | |
10564 | /* FIXME: This algorithm is *exactly* the same as the ST2 version. | |
10565 | So why have two different instructions ? There must be | |
10566 | something wrong somewhere. */ | |
10567 | vec_store (cpu, address, 2); | |
10568 | } | |
10569 | ||
10570 | /* Store multiple 1-element structures into three registers. */ | |
10571 | static void | |
10572 | ST1_3 (sim_cpu *cpu, uint64_t address) | |
10573 | { | |
10574 | /* FIXME: This algorithm is *exactly* the same as the ST3 version. | |
10575 | So why have two different instructions ? There must be | |
10576 | something wrong somewhere. */ | |
10577 | vec_store (cpu, address, 3); | |
10578 | } | |
10579 | ||
10580 | /* Store multiple 1-element structures into four registers. */ | |
10581 | static void | |
10582 | ST1_4 (sim_cpu *cpu, uint64_t address) | |
10583 | { | |
10584 | /* FIXME: This algorithm is *exactly* the same as the ST4 version. | |
10585 | So why have two different instructions ? There must be | |
10586 | something wrong somewhere. */ | |
10587 | vec_store (cpu, address, 4); | |
10588 | } | |
10589 | ||
10590 | static void | |
10591 | do_vec_LDnR (sim_cpu *cpu, uint64_t address) | |
10592 | { | |
10593 | /* instr[31] = 0 | |
10594 | instr[30] = element selector 0=>half, 1=>all elements | |
10595 | instr[29,24] = 00 1101 | |
10596 | instr[23] = 0=>simple, 1=>post | |
10597 | instr[22] = 1 | |
10598 | instr[21] = width: LD1R-or-LD3R (0) / LD2R-or-LD4R (1) | |
10599 | instr[20,16] = 0 0000 (simple), Vinc (reg-post-inc, no SP), | |
10600 | 11111 (immediate post inc) | |
10601 | instr[15,14] = 11 | |
10602 | instr[13] = width: LD1R-or-LD2R (0) / LD3R-or-LD4R (1) | |
10603 | instr[12] = 0 | |
10604 | instr[11,10] = element size 00=> byte(b), 01=> half(h), | |
10605 | 10=> word(s), 11=> double(d) | |
10606 | instr[9,5] = address | |
10607 | instr[4,0] = Vd */ | |
10608 | ||
10609 | unsigned full = uimm (aarch64_get_instr (cpu), 30, 30); | |
10610 | unsigned vd = uimm (aarch64_get_instr (cpu), 4, 0); | |
10611 | unsigned size = uimm (aarch64_get_instr (cpu), 11, 10); | |
10612 | int i; | |
10613 | ||
10614 | NYI_assert (29, 24, 0x0D); | |
10615 | NYI_assert (22, 22, 1); | |
10616 | NYI_assert (15, 14, 3); | |
10617 | NYI_assert (12, 12, 0); | |
10618 | ||
10619 | switch ((uimm (aarch64_get_instr (cpu), 13, 13) << 1) | |
10620 | | uimm (aarch64_get_instr (cpu), 21, 21)) | |
10621 | { | |
10622 | case 0: /* LD1R. */ | |
10623 | switch (size) | |
10624 | { | |
10625 | case 0: | |
10626 | { | |
10627 | uint8_t val = aarch64_get_mem_u8 (cpu, address); | |
10628 | for (i = 0; i < (full ? 16 : 8); i++) | |
10629 | aarch64_set_vec_u8 (cpu, vd, i, val); | |
10630 | break; | |
10631 | } | |
10632 | ||
10633 | case 1: | |
10634 | { | |
10635 | uint16_t val = aarch64_get_mem_u16 (cpu, address); | |
10636 | for (i = 0; i < (full ? 8 : 4); i++) | |
10637 | aarch64_set_vec_u16 (cpu, vd, i, val); | |
10638 | break; | |
10639 | } | |
10640 | ||
10641 | case 2: | |
10642 | { | |
10643 | uint32_t val = aarch64_get_mem_u32 (cpu, address); | |
10644 | for (i = 0; i < (full ? 4 : 2); i++) | |
10645 | aarch64_set_vec_u32 (cpu, vd, i, val); | |
10646 | break; | |
10647 | } | |
10648 | ||
10649 | case 3: | |
10650 | { | |
10651 | uint64_t val = aarch64_get_mem_u64 (cpu, address); | |
10652 | for (i = 0; i < (full ? 2 : 1); i++) | |
10653 | aarch64_set_vec_u64 (cpu, vd, i, val); | |
10654 | break; | |
10655 | } | |
10656 | ||
10657 | default: | |
10658 | HALT_UNALLOC; | |
10659 | } | |
10660 | break; | |
10661 | ||
10662 | case 1: /* LD2R. */ | |
10663 | switch (size) | |
10664 | { | |
10665 | case 0: | |
10666 | { | |
10667 | uint8_t val1 = aarch64_get_mem_u8 (cpu, address); | |
10668 | uint8_t val2 = aarch64_get_mem_u8 (cpu, address + 1); | |
10669 | ||
10670 | for (i = 0; i < (full ? 16 : 8); i++) | |
10671 | { | |
10672 | aarch64_set_vec_u8 (cpu, vd, 0, val1); | |
10673 | aarch64_set_vec_u8 (cpu, vd + 1, 0, val2); | |
10674 | } | |
10675 | break; | |
10676 | } | |
10677 | ||
10678 | case 1: | |
10679 | { | |
10680 | uint16_t val1 = aarch64_get_mem_u16 (cpu, address); | |
10681 | uint16_t val2 = aarch64_get_mem_u16 (cpu, address + 2); | |
10682 | ||
10683 | for (i = 0; i < (full ? 8 : 4); i++) | |
10684 | { | |
10685 | aarch64_set_vec_u16 (cpu, vd, 0, val1); | |
10686 | aarch64_set_vec_u16 (cpu, vd + 1, 0, val2); | |
10687 | } | |
10688 | break; | |
10689 | } | |
10690 | ||
10691 | case 2: | |
10692 | { | |
10693 | uint32_t val1 = aarch64_get_mem_u32 (cpu, address); | |
10694 | uint32_t val2 = aarch64_get_mem_u32 (cpu, address + 4); | |
10695 | ||
10696 | for (i = 0; i < (full ? 4 : 2); i++) | |
10697 | { | |
10698 | aarch64_set_vec_u32 (cpu, vd, 0, val1); | |
10699 | aarch64_set_vec_u32 (cpu, vd + 1, 0, val2); | |
10700 | } | |
10701 | break; | |
10702 | } | |
10703 | ||
10704 | case 3: | |
10705 | { | |
10706 | uint64_t val1 = aarch64_get_mem_u64 (cpu, address); | |
10707 | uint64_t val2 = aarch64_get_mem_u64 (cpu, address + 8); | |
10708 | ||
10709 | for (i = 0; i < (full ? 2 : 1); i++) | |
10710 | { | |
10711 | aarch64_set_vec_u64 (cpu, vd, 0, val1); | |
10712 | aarch64_set_vec_u64 (cpu, vd + 1, 0, val2); | |
10713 | } | |
10714 | break; | |
10715 | } | |
10716 | ||
10717 | default: | |
10718 | HALT_UNALLOC; | |
10719 | } | |
10720 | break; | |
10721 | ||
10722 | case 2: /* LD3R. */ | |
10723 | switch (size) | |
10724 | { | |
10725 | case 0: | |
10726 | { | |
10727 | uint8_t val1 = aarch64_get_mem_u8 (cpu, address); | |
10728 | uint8_t val2 = aarch64_get_mem_u8 (cpu, address + 1); | |
10729 | uint8_t val3 = aarch64_get_mem_u8 (cpu, address + 2); | |
10730 | ||
10731 | for (i = 0; i < (full ? 16 : 8); i++) | |
10732 | { | |
10733 | aarch64_set_vec_u8 (cpu, vd, 0, val1); | |
10734 | aarch64_set_vec_u8 (cpu, vd + 1, 0, val2); | |
10735 | aarch64_set_vec_u8 (cpu, vd + 2, 0, val3); | |
10736 | } | |
10737 | } | |
10738 | break; | |
10739 | ||
10740 | case 1: | |
10741 | { | |
10742 | uint32_t val1 = aarch64_get_mem_u16 (cpu, address); | |
10743 | uint32_t val2 = aarch64_get_mem_u16 (cpu, address + 2); | |
10744 | uint32_t val3 = aarch64_get_mem_u16 (cpu, address + 4); | |
10745 | ||
10746 | for (i = 0; i < (full ? 8 : 4); i++) | |
10747 | { | |
10748 | aarch64_set_vec_u16 (cpu, vd, 0, val1); | |
10749 | aarch64_set_vec_u16 (cpu, vd + 1, 0, val2); | |
10750 | aarch64_set_vec_u16 (cpu, vd + 2, 0, val3); | |
10751 | } | |
10752 | } | |
10753 | break; | |
10754 | ||
10755 | case 2: | |
10756 | { | |
10757 | uint32_t val1 = aarch64_get_mem_u32 (cpu, address); | |
10758 | uint32_t val2 = aarch64_get_mem_u32 (cpu, address + 4); | |
10759 | uint32_t val3 = aarch64_get_mem_u32 (cpu, address + 8); | |
10760 | ||
10761 | for (i = 0; i < (full ? 4 : 2); i++) | |
10762 | { | |
10763 | aarch64_set_vec_u32 (cpu, vd, 0, val1); | |
10764 | aarch64_set_vec_u32 (cpu, vd + 1, 0, val2); | |
10765 | aarch64_set_vec_u32 (cpu, vd + 2, 0, val3); | |
10766 | } | |
10767 | } | |
10768 | break; | |
10769 | ||
10770 | case 3: | |
10771 | { | |
10772 | uint64_t val1 = aarch64_get_mem_u64 (cpu, address); | |
10773 | uint64_t val2 = aarch64_get_mem_u64 (cpu, address + 8); | |
10774 | uint64_t val3 = aarch64_get_mem_u64 (cpu, address + 16); | |
10775 | ||
10776 | for (i = 0; i < (full ? 2 : 1); i++) | |
10777 | { | |
10778 | aarch64_set_vec_u64 (cpu, vd, 0, val1); | |
10779 | aarch64_set_vec_u64 (cpu, vd + 1, 0, val2); | |
10780 | aarch64_set_vec_u64 (cpu, vd + 2, 0, val3); | |
10781 | } | |
10782 | } | |
10783 | break; | |
10784 | ||
10785 | default: | |
10786 | HALT_UNALLOC; | |
10787 | } | |
10788 | break; | |
10789 | ||
10790 | case 3: /* LD4R. */ | |
10791 | switch (size) | |
10792 | { | |
10793 | case 0: | |
10794 | { | |
10795 | uint8_t val1 = aarch64_get_mem_u8 (cpu, address); | |
10796 | uint8_t val2 = aarch64_get_mem_u8 (cpu, address + 1); | |
10797 | uint8_t val3 = aarch64_get_mem_u8 (cpu, address + 2); | |
10798 | uint8_t val4 = aarch64_get_mem_u8 (cpu, address + 3); | |
10799 | ||
10800 | for (i = 0; i < (full ? 16 : 8); i++) | |
10801 | { | |
10802 | aarch64_set_vec_u8 (cpu, vd, 0, val1); | |
10803 | aarch64_set_vec_u8 (cpu, vd + 1, 0, val2); | |
10804 | aarch64_set_vec_u8 (cpu, vd + 2, 0, val3); | |
10805 | aarch64_set_vec_u8 (cpu, vd + 3, 0, val4); | |
10806 | } | |
10807 | } | |
10808 | break; | |
10809 | ||
10810 | case 1: | |
10811 | { | |
10812 | uint32_t val1 = aarch64_get_mem_u16 (cpu, address); | |
10813 | uint32_t val2 = aarch64_get_mem_u16 (cpu, address + 2); | |
10814 | uint32_t val3 = aarch64_get_mem_u16 (cpu, address + 4); | |
10815 | uint32_t val4 = aarch64_get_mem_u16 (cpu, address + 6); | |
10816 | ||
10817 | for (i = 0; i < (full ? 8 : 4); i++) | |
10818 | { | |
10819 | aarch64_set_vec_u16 (cpu, vd, 0, val1); | |
10820 | aarch64_set_vec_u16 (cpu, vd + 1, 0, val2); | |
10821 | aarch64_set_vec_u16 (cpu, vd + 2, 0, val3); | |
10822 | aarch64_set_vec_u16 (cpu, vd + 3, 0, val4); | |
10823 | } | |
10824 | } | |
10825 | break; | |
10826 | ||
10827 | case 2: | |
10828 | { | |
10829 | uint32_t val1 = aarch64_get_mem_u32 (cpu, address); | |
10830 | uint32_t val2 = aarch64_get_mem_u32 (cpu, address + 4); | |
10831 | uint32_t val3 = aarch64_get_mem_u32 (cpu, address + 8); | |
10832 | uint32_t val4 = aarch64_get_mem_u32 (cpu, address + 12); | |
10833 | ||
10834 | for (i = 0; i < (full ? 4 : 2); i++) | |
10835 | { | |
10836 | aarch64_set_vec_u32 (cpu, vd, 0, val1); | |
10837 | aarch64_set_vec_u32 (cpu, vd + 1, 0, val2); | |
10838 | aarch64_set_vec_u32 (cpu, vd + 2, 0, val3); | |
10839 | aarch64_set_vec_u32 (cpu, vd + 3, 0, val4); | |
10840 | } | |
10841 | } | |
10842 | break; | |
10843 | ||
10844 | case 3: | |
10845 | { | |
10846 | uint64_t val1 = aarch64_get_mem_u64 (cpu, address); | |
10847 | uint64_t val2 = aarch64_get_mem_u64 (cpu, address + 8); | |
10848 | uint64_t val3 = aarch64_get_mem_u64 (cpu, address + 16); | |
10849 | uint64_t val4 = aarch64_get_mem_u64 (cpu, address + 24); | |
10850 | ||
10851 | for (i = 0; i < (full ? 2 : 1); i++) | |
10852 | { | |
10853 | aarch64_set_vec_u64 (cpu, vd, 0, val1); | |
10854 | aarch64_set_vec_u64 (cpu, vd + 1, 0, val2); | |
10855 | aarch64_set_vec_u64 (cpu, vd + 2, 0, val3); | |
10856 | aarch64_set_vec_u64 (cpu, vd + 3, 0, val4); | |
10857 | } | |
10858 | } | |
10859 | break; | |
10860 | ||
10861 | default: | |
10862 | HALT_UNALLOC; | |
10863 | } | |
10864 | break; | |
10865 | ||
10866 | default: | |
10867 | HALT_UNALLOC; | |
10868 | } | |
10869 | } | |
10870 | ||
10871 | static void | |
10872 | do_vec_load_store (sim_cpu *cpu) | |
10873 | { | |
10874 | /* {LD|ST}<N> {Vd..Vd+N}, vaddr | |
10875 | ||
10876 | instr[31] = 0 | |
10877 | instr[30] = element selector 0=>half, 1=>all elements | |
10878 | instr[29,25] = 00110 | |
10879 | instr[24] = ? | |
10880 | instr[23] = 0=>simple, 1=>post | |
10881 | instr[22] = 0=>store, 1=>load | |
10882 | instr[21] = 0 (LDn) / small(0)-large(1) selector (LDnR) | |
10883 | instr[20,16] = 00000 (simple), Vinc (reg-post-inc, no SP), | |
10884 | 11111 (immediate post inc) | |
10885 | instr[15,12] = elements and destinations. eg for load: | |
10886 | 0000=>LD4 => load multiple 4-element to | |
10887 | four consecutive registers | |
10888 | 0100=>LD3 => load multiple 3-element to | |
10889 | three consecutive registers | |
10890 | 1000=>LD2 => load multiple 2-element to | |
10891 | two consecutive registers | |
10892 | 0010=>LD1 => load multiple 1-element to | |
10893 | four consecutive registers | |
10894 | 0110=>LD1 => load multiple 1-element to | |
10895 | three consecutive registers | |
10896 | 1010=>LD1 => load multiple 1-element to | |
10897 | two consecutive registers | |
10898 | 0111=>LD1 => load multiple 1-element to | |
10899 | one register | |
10900 | 1100=>LDR1,LDR2 | |
10901 | 1110=>LDR3,LDR4 | |
10902 | instr[11,10] = element size 00=> byte(b), 01=> half(h), | |
10903 | 10=> word(s), 11=> double(d) | |
10904 | instr[9,5] = Vn, can be SP | |
10905 | instr[4,0] = Vd */ | |
10906 | ||
10907 | int post; | |
10908 | int load; | |
10909 | unsigned vn; | |
10910 | uint64_t address; | |
10911 | int type; | |
10912 | ||
10913 | if (uimm (aarch64_get_instr (cpu), 31, 31) != 0 | |
10914 | || uimm (aarch64_get_instr (cpu), 29, 25) != 0x06) | |
10915 | HALT_NYI; | |
10916 | ||
10917 | type = uimm (aarch64_get_instr (cpu), 15, 12); | |
10918 | if (type != 0xE && type != 0xE && uimm (aarch64_get_instr (cpu), 21, 21) != 0) | |
10919 | HALT_NYI; | |
10920 | ||
10921 | post = uimm (aarch64_get_instr (cpu), 23, 23); | |
10922 | load = uimm (aarch64_get_instr (cpu), 22, 22); | |
10923 | vn = uimm (aarch64_get_instr (cpu), 9, 5); | |
10924 | address = aarch64_get_reg_u64 (cpu, vn, SP_OK); | |
10925 | ||
10926 | if (post) | |
10927 | { | |
10928 | unsigned vm = uimm (aarch64_get_instr (cpu), 20, 16); | |
10929 | ||
10930 | if (vm == R31) | |
10931 | { | |
10932 | unsigned sizeof_operation; | |
10933 | ||
10934 | switch (type) | |
10935 | { | |
10936 | case 0: sizeof_operation = 32; break; | |
10937 | case 4: sizeof_operation = 24; break; | |
10938 | case 8: sizeof_operation = 16; break; | |
10939 | ||
10940 | case 0xC: | |
10941 | sizeof_operation = uimm (aarch64_get_instr (cpu), 21, 21) ? 2 : 1; | |
10942 | sizeof_operation <<= uimm (aarch64_get_instr (cpu), 11, 10); | |
10943 | break; | |
10944 | ||
10945 | case 0xE: | |
10946 | sizeof_operation = uimm (aarch64_get_instr (cpu), 21, 21) ? 8 : 4; | |
10947 | sizeof_operation <<= uimm (aarch64_get_instr (cpu), 11, 10); | |
10948 | break; | |
10949 | ||
10950 | case 2: | |
10951 | case 6: | |
10952 | case 10: | |
10953 | case 7: | |
10954 | sizeof_operation = 2 << uimm (aarch64_get_instr (cpu), 11, 10); | |
10955 | break; | |
10956 | ||
10957 | default: | |
10958 | HALT_UNALLOC; | |
10959 | } | |
10960 | ||
10961 | if (uimm (aarch64_get_instr (cpu), 30, 30)) | |
10962 | sizeof_operation *= 2; | |
10963 | ||
10964 | aarch64_set_reg_u64 (cpu, vn, SP_OK, address + sizeof_operation); | |
10965 | } | |
10966 | else | |
10967 | aarch64_set_reg_u64 (cpu, vn, SP_OK, | |
10968 | address + aarch64_get_reg_u64 (cpu, vm, NO_SP)); | |
10969 | } | |
10970 | else | |
10971 | { | |
10972 | NYI_assert (20, 16, 0); | |
10973 | } | |
10974 | ||
10975 | if (load) | |
10976 | { | |
10977 | switch (type) | |
10978 | { | |
10979 | case 0: LD4 (cpu, address); return; | |
10980 | case 4: LD3 (cpu, address); return; | |
10981 | case 8: LD2 (cpu, address); return; | |
10982 | case 2: LD1_4 (cpu, address); return; | |
10983 | case 6: LD1_3 (cpu, address); return; | |
10984 | case 10: LD1_2 (cpu, address); return; | |
10985 | case 7: LD1_1 (cpu, address); return; | |
10986 | ||
10987 | case 0xE: | |
10988 | case 0xC: do_vec_LDnR (cpu, address); return; | |
10989 | ||
10990 | default: | |
10991 | HALT_NYI; | |
10992 | } | |
10993 | } | |
10994 | ||
10995 | /* Stores. */ | |
10996 | switch (type) | |
10997 | { | |
10998 | case 0: ST4 (cpu, address); return; | |
10999 | case 4: ST3 (cpu, address); return; | |
11000 | case 8: ST2 (cpu, address); return; | |
11001 | case 2: ST1_4 (cpu, address); return; | |
11002 | case 6: ST1_3 (cpu, address); return; | |
11003 | case 10: ST1_2 (cpu, address); return; | |
11004 | case 7: ST1_1 (cpu, address); return; | |
11005 | default: | |
11006 | HALT_NYI; | |
11007 | } | |
11008 | } | |
11009 | ||
11010 | static void | |
11011 | dexLdSt (sim_cpu *cpu) | |
11012 | { | |
11013 | /* uint32_t group = dispatchGroup (aarch64_get_instr (cpu)); | |
11014 | assert group == GROUP_LDST_0100 || group == GROUP_LDST_0110 || | |
11015 | group == GROUP_LDST_1100 || group == GROUP_LDST_1110 | |
11016 | bits [29,28:26] of a LS are the secondary dispatch vector. */ | |
11017 | uint32_t group2 = dispatchLS (aarch64_get_instr (cpu)); | |
11018 | ||
11019 | switch (group2) | |
11020 | { | |
11021 | case LS_EXCL_000: | |
11022 | dexLoadExclusive (cpu); return; | |
11023 | ||
11024 | case LS_LIT_010: | |
11025 | case LS_LIT_011: | |
11026 | dexLoadLiteral (cpu); return; | |
11027 | ||
11028 | case LS_OTHER_110: | |
11029 | case LS_OTHER_111: | |
11030 | dexLoadOther (cpu); return; | |
11031 | ||
11032 | case LS_ADVSIMD_001: | |
11033 | do_vec_load_store (cpu); return; | |
11034 | ||
11035 | case LS_PAIR_100: | |
11036 | dex_load_store_pair_gr (cpu); return; | |
11037 | ||
11038 | case LS_PAIR_101: | |
11039 | dex_load_store_pair_fp (cpu); return; | |
11040 | ||
11041 | default: | |
11042 | /* Should never reach here. */ | |
11043 | HALT_NYI; | |
11044 | } | |
11045 | } | |
11046 | ||
11047 | /* Specific decode and execute for group Data Processing Register. */ | |
11048 | ||
11049 | static void | |
11050 | dexLogicalShiftedRegister (sim_cpu *cpu) | |
11051 | { | |
11052 | /* assert instr[28:24] = 01010 | |
11053 | instr[31] = size : 0 ==> 32 bit, 1 ==> 64 bit | |
11054 | instr[30,29:21] = op,N : 000 ==> AND, 001 ==> BIC, | |
11055 | 010 ==> ORR, 011 ==> ORN | |
11056 | 100 ==> EOR, 101 ==> EON, | |
11057 | 110 ==> ANDS, 111 ==> BICS | |
11058 | instr[23,22] = shift : 0 ==> LSL, 1 ==> LSR, 2 ==> ASR, 3 ==> ROR | |
11059 | instr[15,10] = count : must be 0xxxxx for 32 bit | |
11060 | instr[9,5] = Rn | |
11061 | instr[4,0] = Rd */ | |
11062 | ||
11063 | /* unsigned rm = uimm (aarch64_get_instr (cpu), 20, 16); */ | |
11064 | uint32_t dispatch; | |
11065 | Shift shiftType; | |
11066 | uint32_t size = uimm (aarch64_get_instr (cpu), 31, 31); | |
11067 | ||
11068 | /* 32 bit operations must have count[5] = 0. */ | |
11069 | /* or else we have an UNALLOC. */ | |
11070 | uint32_t count = uimm (aarch64_get_instr (cpu), 15, 10); | |
11071 | ||
11072 | if (!size && uimm (count, 5, 5)) | |
11073 | HALT_UNALLOC; | |
11074 | ||
11075 | shiftType = shift (aarch64_get_instr (cpu), 22); | |
11076 | ||
11077 | /* dispatch on size:op:N i.e aarch64_get_instr (cpu)[31,29:21]. */ | |
11078 | dispatch = ( (uimm (aarch64_get_instr (cpu), 31, 29) << 1) | |
11079 | | uimm (aarch64_get_instr (cpu), 21, 21)); | |
11080 | ||
11081 | switch (dispatch) | |
11082 | { | |
11083 | case 0: and32_shift (cpu, shiftType, count); return; | |
11084 | case 1: bic32_shift (cpu, shiftType, count); return; | |
11085 | case 2: orr32_shift (cpu, shiftType, count); return; | |
11086 | case 3: orn32_shift (cpu, shiftType, count); return; | |
11087 | case 4: eor32_shift (cpu, shiftType, count); return; | |
11088 | case 5: eon32_shift (cpu, shiftType, count); return; | |
11089 | case 6: ands32_shift (cpu, shiftType, count); return; | |
11090 | case 7: bics32_shift (cpu, shiftType, count); return; | |
11091 | case 8: and64_shift (cpu, shiftType, count); return; | |
11092 | case 9: bic64_shift (cpu, shiftType, count); return; | |
11093 | case 10:orr64_shift (cpu, shiftType, count); return; | |
11094 | case 11:orn64_shift (cpu, shiftType, count); return; | |
11095 | case 12:eor64_shift (cpu, shiftType, count); return; | |
11096 | case 13:eon64_shift (cpu, shiftType, count); return; | |
11097 | case 14:ands64_shift (cpu, shiftType, count); return; | |
11098 | case 15:bics64_shift (cpu, shiftType, count); return; | |
11099 | default: HALT_UNALLOC; | |
11100 | } | |
11101 | } | |
11102 | ||
11103 | /* 32 bit conditional select. */ | |
11104 | static void | |
11105 | csel32 (sim_cpu *cpu, CondCode cc) | |
11106 | { | |
11107 | unsigned rm = uimm (aarch64_get_instr (cpu), 20, 16); | |
11108 | unsigned rn = uimm (aarch64_get_instr (cpu), 9, 5); | |
11109 | unsigned rd = uimm (aarch64_get_instr (cpu), 4, 0); | |
11110 | ||
11111 | aarch64_set_reg_u64 (cpu, rd, NO_SP, | |
11112 | testConditionCode (cpu, cc) | |
11113 | ? aarch64_get_reg_u32 (cpu, rn, NO_SP) | |
11114 | : aarch64_get_reg_u32 (cpu, rm, NO_SP)); | |
11115 | } | |
11116 | ||
11117 | /* 64 bit conditional select. */ | |
11118 | static void | |
11119 | csel64 (sim_cpu *cpu, CondCode cc) | |
11120 | { | |
11121 | unsigned rm = uimm (aarch64_get_instr (cpu), 20, 16); | |
11122 | unsigned rn = uimm (aarch64_get_instr (cpu), 9, 5); | |
11123 | unsigned rd = uimm (aarch64_get_instr (cpu), 4, 0); | |
11124 | ||
11125 | aarch64_set_reg_u64 (cpu, rd, NO_SP, | |
11126 | testConditionCode (cpu, cc) | |
11127 | ? aarch64_get_reg_u64 (cpu, rn, NO_SP) | |
11128 | : aarch64_get_reg_u64 (cpu, rm, NO_SP)); | |
11129 | } | |
11130 | ||
11131 | /* 32 bit conditional increment. */ | |
11132 | static void | |
11133 | csinc32 (sim_cpu *cpu, CondCode cc) | |
11134 | { | |
11135 | unsigned rm = uimm (aarch64_get_instr (cpu), 20, 16); | |
11136 | unsigned rn = uimm (aarch64_get_instr (cpu), 9, 5); | |
11137 | unsigned rd = uimm (aarch64_get_instr (cpu), 4, 0); | |
11138 | ||
11139 | aarch64_set_reg_u64 (cpu, rd, NO_SP, | |
11140 | testConditionCode (cpu, cc) | |
11141 | ? aarch64_get_reg_u32 (cpu, rn, NO_SP) | |
11142 | : aarch64_get_reg_u32 (cpu, rm, NO_SP) + 1); | |
11143 | } | |
11144 | ||
11145 | /* 64 bit conditional increment. */ | |
11146 | static void | |
11147 | csinc64 (sim_cpu *cpu, CondCode cc) | |
11148 | { | |
11149 | unsigned rm = uimm (aarch64_get_instr (cpu), 20, 16); | |
11150 | unsigned rn = uimm (aarch64_get_instr (cpu), 9, 5); | |
11151 | unsigned rd = uimm (aarch64_get_instr (cpu), 4, 0); | |
11152 | ||
11153 | aarch64_set_reg_u64 (cpu, rd, NO_SP, | |
11154 | testConditionCode (cpu, cc) | |
11155 | ? aarch64_get_reg_u64 (cpu, rn, NO_SP) | |
11156 | : aarch64_get_reg_u64 (cpu, rm, NO_SP) + 1); | |
11157 | } | |
11158 | ||
11159 | /* 32 bit conditional invert. */ | |
11160 | static void | |
11161 | csinv32 (sim_cpu *cpu, CondCode cc) | |
11162 | { | |
11163 | unsigned rm = uimm (aarch64_get_instr (cpu), 20, 16); | |
11164 | unsigned rn = uimm (aarch64_get_instr (cpu), 9, 5); | |
11165 | unsigned rd = uimm (aarch64_get_instr (cpu), 4, 0); | |
11166 | ||
11167 | aarch64_set_reg_u64 (cpu, rd, NO_SP, | |
11168 | testConditionCode (cpu, cc) | |
11169 | ? aarch64_get_reg_u32 (cpu, rn, NO_SP) | |
11170 | : ~ aarch64_get_reg_u32 (cpu, rm, NO_SP)); | |
11171 | } | |
11172 | ||
11173 | /* 64 bit conditional invert. */ | |
11174 | static void | |
11175 | csinv64 (sim_cpu *cpu, CondCode cc) | |
11176 | { | |
11177 | unsigned rm = uimm (aarch64_get_instr (cpu), 20, 16); | |
11178 | unsigned rn = uimm (aarch64_get_instr (cpu), 9, 5); | |
11179 | unsigned rd = uimm (aarch64_get_instr (cpu), 4, 0); | |
11180 | ||
11181 | aarch64_set_reg_u64 (cpu, rd, NO_SP, | |
11182 | testConditionCode (cpu, cc) | |
11183 | ? aarch64_get_reg_u64 (cpu, rn, NO_SP) | |
11184 | : ~ aarch64_get_reg_u64 (cpu, rm, NO_SP)); | |
11185 | } | |
11186 | ||
11187 | /* 32 bit conditional negate. */ | |
11188 | static void | |
11189 | csneg32 (sim_cpu *cpu, CondCode cc) | |
11190 | { | |
11191 | unsigned rm = uimm (aarch64_get_instr (cpu), 20, 16); | |
11192 | unsigned rn = uimm (aarch64_get_instr (cpu), 9, 5); | |
11193 | unsigned rd = uimm (aarch64_get_instr (cpu), 4, 0); | |
11194 | ||
11195 | aarch64_set_reg_u64 (cpu, rd, NO_SP, | |
11196 | testConditionCode (cpu, cc) | |
11197 | ? aarch64_get_reg_u32 (cpu, rn, NO_SP) | |
11198 | : - aarch64_get_reg_u32 (cpu, rm, NO_SP)); | |
11199 | } | |
11200 | ||
11201 | /* 64 bit conditional negate. */ | |
11202 | static void | |
11203 | csneg64 (sim_cpu *cpu, CondCode cc) | |
11204 | { | |
11205 | unsigned rm = uimm (aarch64_get_instr (cpu), 20, 16); | |
11206 | unsigned rn = uimm (aarch64_get_instr (cpu), 9, 5); | |
11207 | unsigned rd = uimm (aarch64_get_instr (cpu), 4, 0); | |
11208 | ||
11209 | aarch64_set_reg_u64 (cpu, rd, NO_SP, | |
11210 | testConditionCode (cpu, cc) | |
11211 | ? aarch64_get_reg_u64 (cpu, rn, NO_SP) | |
11212 | : - aarch64_get_reg_u64 (cpu, rm, NO_SP)); | |
11213 | } | |
11214 | ||
11215 | static void | |
11216 | dexCondSelect (sim_cpu *cpu) | |
11217 | { | |
11218 | /* assert instr[28,21] = 11011011 | |
11219 | instr[31] = size : 0 ==> 32 bit, 1 ==> 64 bit | |
11220 | instr[30:11,10] = op : 000 ==> CSEL, 001 ==> CSINC, | |
11221 | 100 ==> CSINV, 101 ==> CSNEG, | |
11222 | _1_ ==> UNALLOC | |
11223 | instr[29] = S : 0 ==> ok, 1 ==> UNALLOC | |
11224 | instr[15,12] = cond | |
11225 | instr[29] = S : 0 ==> ok, 1 ==> UNALLOC */ | |
11226 | ||
11227 | CondCode cc; | |
11228 | uint32_t dispatch; | |
11229 | uint32_t S = uimm (aarch64_get_instr (cpu), 29, 29); | |
11230 | uint32_t op2 = uimm (aarch64_get_instr (cpu), 11, 10); | |
11231 | ||
11232 | if (S == 1) | |
11233 | HALT_UNALLOC; | |
11234 | ||
11235 | if (op2 & 0x2) | |
11236 | HALT_UNALLOC; | |
11237 | ||
11238 | cc = condcode (aarch64_get_instr (cpu), 12); | |
11239 | dispatch = ((uimm (aarch64_get_instr (cpu), 31, 30) << 1) | op2); | |
11240 | ||
11241 | switch (dispatch) | |
11242 | { | |
11243 | case 0: csel32 (cpu, cc); return; | |
11244 | case 1: csinc32 (cpu, cc); return; | |
11245 | case 2: csinv32 (cpu, cc); return; | |
11246 | case 3: csneg32 (cpu, cc); return; | |
11247 | case 4: csel64 (cpu, cc); return; | |
11248 | case 5: csinc64 (cpu, cc); return; | |
11249 | case 6: csinv64 (cpu, cc); return; | |
11250 | case 7: csneg64 (cpu, cc); return; | |
11251 | default: HALT_UNALLOC; | |
11252 | } | |
11253 | } | |
11254 | ||
11255 | /* Some helpers for counting leading 1 or 0 bits. */ | |
11256 | ||
11257 | /* Counts the number of leading bits which are the same | |
11258 | in a 32 bit value in the range 1 to 32. */ | |
11259 | static uint32_t | |
11260 | leading32 (uint32_t value) | |
11261 | { | |
11262 | int32_t mask= 0xffff0000; | |
11263 | uint32_t count= 16; /* Counts number of bits set in mask. */ | |
11264 | uint32_t lo = 1; /* Lower bound for number of sign bits. */ | |
11265 | uint32_t hi = 32; /* Upper bound for number of sign bits. */ | |
11266 | ||
11267 | while (lo + 1 < hi) | |
11268 | { | |
11269 | int32_t test = (value & mask); | |
11270 | ||
11271 | if (test == 0 || test == mask) | |
11272 | { | |
11273 | lo = count; | |
11274 | count = (lo + hi) / 2; | |
11275 | mask >>= (count - lo); | |
11276 | } | |
11277 | else | |
11278 | { | |
11279 | hi = count; | |
11280 | count = (lo + hi) / 2; | |
11281 | mask <<= hi - count; | |
11282 | } | |
11283 | } | |
11284 | ||
11285 | if (lo != hi) | |
11286 | { | |
11287 | int32_t test; | |
11288 | ||
11289 | mask >>= 1; | |
11290 | test = (value & mask); | |
11291 | ||
11292 | if (test == 0 || test == mask) | |
11293 | count = hi; | |
11294 | else | |
11295 | count = lo; | |
11296 | } | |
11297 | ||
11298 | return count; | |
11299 | } | |
11300 | ||
11301 | /* Counts the number of leading bits which are the same | |
11302 | in a 64 bit value in the range 1 to 64. */ | |
11303 | static uint64_t | |
11304 | leading64 (uint64_t value) | |
11305 | { | |
11306 | int64_t mask= 0xffffffff00000000LL; | |
11307 | uint64_t count = 32; /* Counts number of bits set in mask. */ | |
11308 | uint64_t lo = 1; /* Lower bound for number of sign bits. */ | |
11309 | uint64_t hi = 64; /* Upper bound for number of sign bits. */ | |
11310 | ||
11311 | while (lo + 1 < hi) | |
11312 | { | |
11313 | int64_t test = (value & mask); | |
11314 | ||
11315 | if (test == 0 || test == mask) | |
11316 | { | |
11317 | lo = count; | |
11318 | count = (lo + hi) / 2; | |
11319 | mask >>= (count - lo); | |
11320 | } | |
11321 | else | |
11322 | { | |
11323 | hi = count; | |
11324 | count = (lo + hi) / 2; | |
11325 | mask <<= hi - count; | |
11326 | } | |
11327 | } | |
11328 | ||
11329 | if (lo != hi) | |
11330 | { | |
11331 | int64_t test; | |
11332 | ||
11333 | mask >>= 1; | |
11334 | test = (value & mask); | |
11335 | ||
11336 | if (test == 0 || test == mask) | |
11337 | count = hi; | |
11338 | else | |
11339 | count = lo; | |
11340 | } | |
11341 | ||
11342 | return count; | |
11343 | } | |
11344 | ||
11345 | /* Bit operations. */ | |
11346 | /* N.B register args may not be SP. */ | |
11347 | ||
11348 | /* 32 bit count leading sign bits. */ | |
11349 | static void | |
11350 | cls32 (sim_cpu *cpu) | |
11351 | { | |
11352 | unsigned rn = uimm (aarch64_get_instr (cpu), 9, 5); | |
11353 | unsigned rd = uimm (aarch64_get_instr (cpu), 4, 0); | |
11354 | ||
11355 | /* N.B. the result needs to exclude the leading bit. */ | |
11356 | aarch64_set_reg_u64 | |
11357 | (cpu, rd, NO_SP, leading32 (aarch64_get_reg_u32 (cpu, rn, NO_SP)) - 1); | |
11358 | } | |
11359 | ||
11360 | /* 64 bit count leading sign bits. */ | |
11361 | static void | |
11362 | cls64 (sim_cpu *cpu) | |
11363 | { | |
11364 | unsigned rn = uimm (aarch64_get_instr (cpu), 9, 5); | |
11365 | unsigned rd = uimm (aarch64_get_instr (cpu), 4, 0); | |
11366 | ||
11367 | /* N.B. the result needs to exclude the leading bit. */ | |
11368 | aarch64_set_reg_u64 | |
11369 | (cpu, rd, NO_SP, leading64 (aarch64_get_reg_u64 (cpu, rn, NO_SP)) - 1); | |
11370 | } | |
11371 | ||
11372 | /* 32 bit count leading zero bits. */ | |
11373 | static void | |
11374 | clz32 (sim_cpu *cpu) | |
11375 | { | |
11376 | unsigned rn = uimm (aarch64_get_instr (cpu), 9, 5); | |
11377 | unsigned rd = uimm (aarch64_get_instr (cpu), 4, 0); | |
11378 | uint32_t value = aarch64_get_reg_u32 (cpu, rn, NO_SP); | |
11379 | ||
11380 | /* if the sign (top) bit is set then the count is 0. */ | |
11381 | if (pick32 (value, 31, 31)) | |
11382 | aarch64_set_reg_u64 (cpu, rd, NO_SP, 0L); | |
11383 | else | |
11384 | aarch64_set_reg_u64 (cpu, rd, NO_SP, leading32 (value)); | |
11385 | } | |
11386 | ||
11387 | /* 64 bit count leading zero bits. */ | |
11388 | static void | |
11389 | clz64 (sim_cpu *cpu) | |
11390 | { | |
11391 | unsigned rn = uimm (aarch64_get_instr (cpu), 9, 5); | |
11392 | unsigned rd = uimm (aarch64_get_instr (cpu), 4, 0); | |
11393 | uint64_t value = aarch64_get_reg_u64 (cpu, rn, NO_SP); | |
11394 | ||
11395 | /* if the sign (top) bit is set then the count is 0. */ | |
11396 | if (pick64 (value, 63, 63)) | |
11397 | aarch64_set_reg_u64 (cpu, rd, NO_SP, 0L); | |
11398 | else | |
11399 | aarch64_set_reg_u64 (cpu, rd, NO_SP, leading64 (value)); | |
11400 | } | |
11401 | ||
11402 | /* 32 bit reverse bits. */ | |
11403 | static void | |
11404 | rbit32 (sim_cpu *cpu) | |
11405 | { | |
11406 | unsigned rn = uimm (aarch64_get_instr (cpu), 9, 5); | |
11407 | unsigned rd = uimm (aarch64_get_instr (cpu), 4, 0); | |
11408 | uint32_t value = aarch64_get_reg_u32 (cpu, rn, NO_SP); | |
11409 | uint32_t result = 0; | |
11410 | int i; | |
11411 | ||
11412 | for (i = 0; i < 32; i++) | |
11413 | { | |
11414 | result <<= 1; | |
11415 | result |= (value & 1); | |
11416 | value >>= 1; | |
11417 | } | |
11418 | aarch64_set_reg_u64 (cpu, rd, NO_SP, result); | |
11419 | } | |
11420 | ||
11421 | /* 64 bit reverse bits. */ | |
11422 | static void | |
11423 | rbit64 (sim_cpu *cpu) | |
11424 | { | |
11425 | unsigned rn = uimm (aarch64_get_instr (cpu), 9, 5); | |
11426 | unsigned rd = uimm (aarch64_get_instr (cpu), 4, 0); | |
11427 | uint64_t value = aarch64_get_reg_u64 (cpu, rn, NO_SP); | |
11428 | uint64_t result = 0; | |
11429 | int i; | |
11430 | ||
11431 | for (i = 0; i < 64; i++) | |
11432 | { | |
11433 | result <<= 1; | |
11434 | result |= (value & 1L); | |
11435 | value >>= 1; | |
11436 | } | |
11437 | aarch64_set_reg_u64 (cpu, rd, NO_SP, result); | |
11438 | } | |
11439 | ||
11440 | /* 32 bit reverse bytes. */ | |
11441 | static void | |
11442 | rev32 (sim_cpu *cpu) | |
11443 | { | |
11444 | unsigned rn = uimm (aarch64_get_instr (cpu), 9, 5); | |
11445 | unsigned rd = uimm (aarch64_get_instr (cpu), 4, 0); | |
11446 | uint32_t value = aarch64_get_reg_u32 (cpu, rn, NO_SP); | |
11447 | uint32_t result = 0; | |
11448 | int i; | |
11449 | ||
11450 | for (i = 0; i < 4; i++) | |
11451 | { | |
11452 | result <<= 8; | |
11453 | result |= (value & 0xff); | |
11454 | value >>= 8; | |
11455 | } | |
11456 | aarch64_set_reg_u64 (cpu, rd, NO_SP, result); | |
11457 | } | |
11458 | ||
11459 | /* 64 bit reverse bytes. */ | |
11460 | static void | |
11461 | rev64 (sim_cpu *cpu) | |
11462 | { | |
11463 | unsigned rn = uimm (aarch64_get_instr (cpu), 9, 5); | |
11464 | unsigned rd = uimm (aarch64_get_instr (cpu), 4, 0); | |
11465 | uint64_t value = aarch64_get_reg_u64 (cpu, rn, NO_SP); | |
11466 | uint64_t result = 0; | |
11467 | int i; | |
11468 | ||
11469 | for (i = 0; i < 8; i++) | |
11470 | { | |
11471 | result <<= 8; | |
11472 | result |= (value & 0xffULL); | |
11473 | value >>= 8; | |
11474 | } | |
11475 | aarch64_set_reg_u64 (cpu, rd, NO_SP, result); | |
11476 | } | |
11477 | ||
11478 | /* 32 bit reverse shorts. */ | |
11479 | /* N.B.this reverses the order of the bytes in each half word. */ | |
11480 | static void | |
11481 | revh32 (sim_cpu *cpu) | |
11482 | { | |
11483 | unsigned rn = uimm (aarch64_get_instr (cpu), 9, 5); | |
11484 | unsigned rd = uimm (aarch64_get_instr (cpu), 4, 0); | |
11485 | uint32_t value = aarch64_get_reg_u32 (cpu, rn, NO_SP); | |
11486 | uint32_t result = 0; | |
11487 | int i; | |
11488 | ||
11489 | for (i = 0; i < 2; i++) | |
11490 | { | |
11491 | result <<= 8; | |
11492 | result |= (value & 0x00ff00ff); | |
11493 | value >>= 8; | |
11494 | } | |
11495 | aarch64_set_reg_u64 (cpu, rd, NO_SP, result); | |
11496 | } | |
11497 | ||
11498 | /* 64 bit reverse shorts. */ | |
11499 | /* N.B.this reverses the order of the bytes in each half word. */ | |
11500 | static void | |
11501 | revh64 (sim_cpu *cpu) | |
11502 | { | |
11503 | unsigned rn = uimm (aarch64_get_instr (cpu), 9, 5); | |
11504 | unsigned rd = uimm (aarch64_get_instr (cpu), 4, 0); | |
11505 | uint64_t value = aarch64_get_reg_u64 (cpu, rn, NO_SP); | |
11506 | uint64_t result = 0; | |
11507 | int i; | |
11508 | ||
11509 | for (i = 0; i < 2; i++) | |
11510 | { | |
11511 | result <<= 8; | |
11512 | result |= (value & 0x00ff00ff00ff00ffULL); | |
11513 | value >>= 8; | |
11514 | } | |
11515 | aarch64_set_reg_u64 (cpu, rd, NO_SP, result); | |
11516 | } | |
11517 | ||
11518 | static void | |
11519 | dexDataProc1Source (sim_cpu *cpu) | |
11520 | { | |
11521 | /* assert instr[30] == 1 | |
11522 | aarch64_get_instr (cpu)[28,21] == 111010110 | |
11523 | instr[31] = size : 0 ==> 32 bit, 1 ==> 64 bit | |
11524 | instr[29] = S : 0 ==> ok, 1 ==> UNALLOC | |
11525 | instr[20,16] = opcode2 : 00000 ==> ok, ow ==> UNALLOC | |
11526 | instr[15,10] = opcode : 000000 ==> RBIT, 000001 ==> REV16, | |
11527 | 000010 ==> REV, 000011 ==> UNALLOC | |
11528 | 000100 ==> CLZ, 000101 ==> CLS | |
11529 | ow ==> UNALLOC | |
11530 | instr[9,5] = rn : may not be SP | |
11531 | instr[4,0] = rd : may not be SP. */ | |
11532 | ||
11533 | uint32_t S = uimm (aarch64_get_instr (cpu), 29, 29); | |
11534 | uint32_t opcode2 = uimm (aarch64_get_instr (cpu), 20, 16); | |
11535 | uint32_t opcode = uimm (aarch64_get_instr (cpu), 15, 10); | |
11536 | uint32_t dispatch = ((uimm (aarch64_get_instr (cpu), 31, 31) << 3) | opcode); | |
11537 | ||
11538 | if (S == 1) | |
11539 | HALT_UNALLOC; | |
11540 | ||
11541 | if (opcode2 != 0) | |
11542 | HALT_UNALLOC; | |
11543 | ||
11544 | if (opcode & 0x38) | |
11545 | HALT_UNALLOC; | |
11546 | ||
11547 | switch (dispatch) | |
11548 | { | |
11549 | case 0: rbit32 (cpu); return; | |
11550 | case 1: revh32 (cpu); return; | |
11551 | case 2: rev32 (cpu); return; | |
11552 | case 4: clz32 (cpu); return; | |
11553 | case 5: cls32 (cpu); return; | |
11554 | case 8: rbit64 (cpu); return; | |
11555 | case 9: revh64 (cpu); return; | |
11556 | case 10:rev32 (cpu); return; | |
11557 | case 11:rev64 (cpu); return; | |
11558 | case 12:clz64 (cpu); return; | |
11559 | case 13:cls64 (cpu); return; | |
11560 | default: HALT_UNALLOC; | |
11561 | } | |
11562 | } | |
11563 | ||
11564 | /* Variable shift. | |
11565 | Shifts by count supplied in register. | |
11566 | N.B register args may not be SP. | |
11567 | These all use the shifted auxiliary function for | |
11568 | simplicity and clarity. Writing the actual shift | |
11569 | inline would avoid a branch and so be faster but | |
11570 | would also necessitate getting signs right. */ | |
11571 | ||
11572 | /* 32 bit arithmetic shift right. */ | |
11573 | static void | |
11574 | asrv32 (sim_cpu *cpu) | |
11575 | { | |
11576 | unsigned rm = uimm (aarch64_get_instr (cpu), 20, 16); | |
11577 | unsigned rn = uimm (aarch64_get_instr (cpu), 9, 5); | |
11578 | unsigned rd = uimm (aarch64_get_instr (cpu), 4, 0); | |
11579 | ||
11580 | aarch64_set_reg_u64 | |
11581 | (cpu, rd, NO_SP, | |
11582 | shifted32 (aarch64_get_reg_u32 (cpu, rn, NO_SP), ASR, | |
11583 | (aarch64_get_reg_u32 (cpu, rm, NO_SP) & 0x1f))); | |
11584 | } | |
11585 | ||
11586 | /* 64 bit arithmetic shift right. */ | |
11587 | static void | |
11588 | asrv64 (sim_cpu *cpu) | |
11589 | { | |
11590 | unsigned rm = uimm (aarch64_get_instr (cpu), 20, 16); | |
11591 | unsigned rn = uimm (aarch64_get_instr (cpu), 9, 5); | |
11592 | unsigned rd = uimm (aarch64_get_instr (cpu), 4, 0); | |
11593 | ||
11594 | aarch64_set_reg_u64 | |
11595 | (cpu, rd, NO_SP, | |
11596 | shifted64 (aarch64_get_reg_u64 (cpu, rn, NO_SP), ASR, | |
11597 | (aarch64_get_reg_u64 (cpu, rm, NO_SP) & 0x3f))); | |
11598 | } | |
11599 | ||
11600 | /* 32 bit logical shift left. */ | |
11601 | static void | |
11602 | lslv32 (sim_cpu *cpu) | |
11603 | { | |
11604 | unsigned rm = uimm (aarch64_get_instr (cpu), 20, 16); | |
11605 | unsigned rn = uimm (aarch64_get_instr (cpu), 9, 5); | |
11606 | unsigned rd = uimm (aarch64_get_instr (cpu), 4, 0); | |
11607 | ||
11608 | aarch64_set_reg_u64 | |
11609 | (cpu, rd, NO_SP, | |
11610 | shifted32 (aarch64_get_reg_u32 (cpu, rn, NO_SP), LSL, | |
11611 | (aarch64_get_reg_u32 (cpu, rm, NO_SP) & 0x1f))); | |
11612 | } | |
11613 | ||
11614 | /* 64 bit arithmetic shift left. */ | |
11615 | static void | |
11616 | lslv64 (sim_cpu *cpu) | |
11617 | { | |
11618 | unsigned rm = uimm (aarch64_get_instr (cpu), 20, 16); | |
11619 | unsigned rn = uimm (aarch64_get_instr (cpu), 9, 5); | |
11620 | unsigned rd = uimm (aarch64_get_instr (cpu), 4, 0); | |
11621 | ||
11622 | aarch64_set_reg_u64 | |
11623 | (cpu, rd, NO_SP, | |
11624 | shifted64 (aarch64_get_reg_u64 (cpu, rn, NO_SP), LSL, | |
11625 | (aarch64_get_reg_u64 (cpu, rm, NO_SP) & 0x3f))); | |
11626 | } | |
11627 | ||
11628 | /* 32 bit logical shift right. */ | |
11629 | static void | |
11630 | lsrv32 (sim_cpu *cpu) | |
11631 | { | |
11632 | unsigned rm = uimm (aarch64_get_instr (cpu), 20, 16); | |
11633 | unsigned rn = uimm (aarch64_get_instr (cpu), 9, 5); | |
11634 | unsigned rd = uimm (aarch64_get_instr (cpu), 4, 0); | |
11635 | ||
11636 | aarch64_set_reg_u64 | |
11637 | (cpu, rd, NO_SP, | |
11638 | shifted32 (aarch64_get_reg_u32 (cpu, rn, NO_SP), LSR, | |
11639 | (aarch64_get_reg_u32 (cpu, rm, NO_SP) & 0x1f))); | |
11640 | } | |
11641 | ||
11642 | /* 64 bit logical shift right. */ | |
11643 | static void | |
11644 | lsrv64 (sim_cpu *cpu) | |
11645 | { | |
11646 | unsigned rm = uimm (aarch64_get_instr (cpu), 20, 16); | |
11647 | unsigned rn = uimm (aarch64_get_instr (cpu), 9, 5); | |
11648 | unsigned rd = uimm (aarch64_get_instr (cpu), 4, 0); | |
11649 | ||
11650 | aarch64_set_reg_u64 | |
11651 | (cpu, rd, NO_SP, | |
11652 | shifted64 (aarch64_get_reg_u64 (cpu, rn, NO_SP), LSR, | |
11653 | (aarch64_get_reg_u64 (cpu, rm, NO_SP) & 0x3f))); | |
11654 | } | |
11655 | ||
11656 | /* 32 bit rotate right. */ | |
11657 | static void | |
11658 | rorv32 (sim_cpu *cpu) | |
11659 | { | |
11660 | unsigned rm = uimm (aarch64_get_instr (cpu), 20, 16); | |
11661 | unsigned rn = uimm (aarch64_get_instr (cpu), 9, 5); | |
11662 | unsigned rd = uimm (aarch64_get_instr (cpu), 4, 0); | |
11663 | ||
11664 | aarch64_set_reg_u64 | |
11665 | (cpu, rd, NO_SP, | |
11666 | shifted32 (aarch64_get_reg_u32 (cpu, rn, NO_SP), ROR, | |
11667 | (aarch64_get_reg_u32 (cpu, rm, NO_SP) & 0x1f))); | |
11668 | } | |
11669 | ||
11670 | /* 64 bit rotate right. */ | |
11671 | static void | |
11672 | rorv64 (sim_cpu *cpu) | |
11673 | { | |
11674 | unsigned rm = uimm (aarch64_get_instr (cpu), 20, 16); | |
11675 | unsigned rn = uimm (aarch64_get_instr (cpu), 9, 5); | |
11676 | unsigned rd = uimm (aarch64_get_instr (cpu), 4, 0); | |
11677 | ||
11678 | aarch64_set_reg_u64 | |
11679 | (cpu, rd, NO_SP, | |
11680 | shifted64 (aarch64_get_reg_u64 (cpu, rn, NO_SP), ROR, | |
11681 | (aarch64_get_reg_u64 (cpu, rm, NO_SP) & 0x3f))); | |
11682 | } | |
11683 | ||
11684 | ||
11685 | /* divide. */ | |
11686 | ||
11687 | /* 32 bit signed divide. */ | |
11688 | static void | |
11689 | cpuiv32 (sim_cpu *cpu) | |
11690 | { | |
11691 | unsigned rm = uimm (aarch64_get_instr (cpu), 20, 16); | |
11692 | unsigned rn = uimm (aarch64_get_instr (cpu), 9, 5); | |
11693 | unsigned rd = uimm (aarch64_get_instr (cpu), 4, 0); | |
11694 | /* N.B. the pseudo-code does the divide using 64 bit data. */ | |
11695 | /* TODO : check that this rounds towards zero as required. */ | |
11696 | int64_t dividend = aarch64_get_reg_s32 (cpu, rn, NO_SP); | |
11697 | int64_t divisor = aarch64_get_reg_s32 (cpu, rm, NO_SP); | |
11698 | ||
11699 | aarch64_set_reg_s64 (cpu, rd, NO_SP, | |
11700 | divisor ? ((int32_t) (dividend / divisor)) : 0); | |
11701 | } | |
11702 | ||
11703 | /* 64 bit signed divide. */ | |
11704 | static void | |
11705 | cpuiv64 (sim_cpu *cpu) | |
11706 | { | |
11707 | unsigned rm = uimm (aarch64_get_instr (cpu), 20, 16); | |
11708 | unsigned rn = uimm (aarch64_get_instr (cpu), 9, 5); | |
11709 | unsigned rd = uimm (aarch64_get_instr (cpu), 4, 0); | |
11710 | ||
11711 | /* TODO : check that this rounds towards zero as required. */ | |
11712 | int64_t divisor = aarch64_get_reg_s64 (cpu, rm, NO_SP); | |
11713 | ||
11714 | aarch64_set_reg_s64 | |
11715 | (cpu, rd, NO_SP, | |
11716 | divisor ? (aarch64_get_reg_s64 (cpu, rn, NO_SP) / divisor) : 0); | |
11717 | } | |
11718 | ||
11719 | /* 32 bit unsigned divide. */ | |
11720 | static void | |
11721 | udiv32 (sim_cpu *cpu) | |
11722 | { | |
11723 | unsigned rm = uimm (aarch64_get_instr (cpu), 20, 16); | |
11724 | unsigned rn = uimm (aarch64_get_instr (cpu), 9, 5); | |
11725 | unsigned rd = uimm (aarch64_get_instr (cpu), 4, 0); | |
11726 | ||
11727 | /* N.B. the pseudo-code does the divide using 64 bit data. */ | |
11728 | uint64_t dividend = aarch64_get_reg_u32 (cpu, rn, NO_SP); | |
11729 | uint64_t divisor = aarch64_get_reg_u32 (cpu, rm, NO_SP); | |
11730 | ||
11731 | aarch64_set_reg_u64 (cpu, rd, NO_SP, | |
11732 | divisor ? (uint32_t) (dividend / divisor) : 0); | |
11733 | } | |
11734 | ||
11735 | /* 64 bit unsigned divide. */ | |
11736 | static void | |
11737 | udiv64 (sim_cpu *cpu) | |
11738 | { | |
11739 | unsigned rm = uimm (aarch64_get_instr (cpu), 20, 16); | |
11740 | unsigned rn = uimm (aarch64_get_instr (cpu), 9, 5); | |
11741 | unsigned rd = uimm (aarch64_get_instr (cpu), 4, 0); | |
11742 | ||
11743 | /* TODO : check that this rounds towards zero as required. */ | |
11744 | uint64_t divisor = aarch64_get_reg_u64 (cpu, rm, NO_SP); | |
11745 | ||
11746 | aarch64_set_reg_u64 | |
11747 | (cpu, rd, NO_SP, | |
11748 | divisor ? (aarch64_get_reg_u64 (cpu, rn, NO_SP) / divisor) : 0); | |
11749 | } | |
11750 | ||
11751 | static void | |
11752 | dexDataProc2Source (sim_cpu *cpu) | |
11753 | { | |
11754 | /* assert instr[30] == 0 | |
11755 | instr[28,21] == 11010110 | |
11756 | instr[31] = size : 0 ==> 32 bit, 1 ==> 64 bit | |
11757 | instr[29] = S : 0 ==> ok, 1 ==> UNALLOC | |
11758 | instr[15,10] = opcode : 000010 ==> UDIV, 000011 ==> CPUIV, | |
11759 | 001000 ==> LSLV, 001001 ==> LSRV | |
11760 | 001010 ==> ASRV, 001011 ==> RORV | |
11761 | ow ==> UNALLOC. */ | |
11762 | ||
11763 | uint32_t dispatch; | |
11764 | uint32_t S = uimm (aarch64_get_instr (cpu), 29, 29); | |
11765 | uint32_t opcode = uimm (aarch64_get_instr (cpu), 15, 10); | |
11766 | ||
11767 | if (S == 1) | |
11768 | HALT_UNALLOC; | |
11769 | ||
11770 | if (opcode & 0x34) | |
11771 | HALT_UNALLOC; | |
11772 | ||
11773 | dispatch = ( (uimm (aarch64_get_instr (cpu), 31, 31) << 3) | |
11774 | | (uimm (opcode, 3, 3) << 2) | |
11775 | | uimm (opcode, 1, 0)); | |
11776 | switch (dispatch) | |
11777 | { | |
11778 | case 2: udiv32 (cpu); return; | |
11779 | case 3: cpuiv32 (cpu); return; | |
11780 | case 4: lslv32 (cpu); return; | |
11781 | case 5: lsrv32 (cpu); return; | |
11782 | case 6: asrv32 (cpu); return; | |
11783 | case 7: rorv32 (cpu); return; | |
11784 | case 10: udiv64 (cpu); return; | |
11785 | case 11: cpuiv64 (cpu); return; | |
11786 | case 12: lslv64 (cpu); return; | |
11787 | case 13: lsrv64 (cpu); return; | |
11788 | case 14: asrv64 (cpu); return; | |
11789 | case 15: rorv64 (cpu); return; | |
11790 | default: HALT_UNALLOC; | |
11791 | } | |
11792 | } | |
11793 | ||
11794 | ||
11795 | /* Multiply. */ | |
11796 | ||
11797 | /* 32 bit multiply and add. */ | |
11798 | static void | |
11799 | madd32 (sim_cpu *cpu) | |
11800 | { | |
11801 | unsigned rm = uimm (aarch64_get_instr (cpu), 20, 16); | |
11802 | unsigned ra = uimm (aarch64_get_instr (cpu), 14, 10); | |
11803 | unsigned rn = uimm (aarch64_get_instr (cpu), 9, 5); | |
11804 | unsigned rd = uimm (aarch64_get_instr (cpu), 4, 0); | |
11805 | ||
11806 | aarch64_set_reg_u64 (cpu, rd, NO_SP, | |
11807 | aarch64_get_reg_u32 (cpu, ra, NO_SP) | |
11808 | + aarch64_get_reg_u32 (cpu, rn, NO_SP) | |
11809 | * aarch64_get_reg_u32 (cpu, rm, NO_SP)); | |
11810 | } | |
11811 | ||
11812 | /* 64 bit multiply and add. */ | |
11813 | static void | |
11814 | madd64 (sim_cpu *cpu) | |
11815 | { | |
11816 | unsigned rm = uimm (aarch64_get_instr (cpu), 20, 16); | |
11817 | unsigned ra = uimm (aarch64_get_instr (cpu), 14, 10); | |
11818 | unsigned rn = uimm (aarch64_get_instr (cpu), 9, 5); | |
11819 | unsigned rd = uimm (aarch64_get_instr (cpu), 4, 0); | |
11820 | ||
11821 | aarch64_set_reg_u64 (cpu, rd, NO_SP, | |
11822 | aarch64_get_reg_u64 (cpu, ra, NO_SP) | |
11823 | + aarch64_get_reg_u64 (cpu, rn, NO_SP) | |
11824 | * aarch64_get_reg_u64 (cpu, rm, NO_SP)); | |
11825 | } | |
11826 | ||
11827 | /* 32 bit multiply and sub. */ | |
11828 | static void | |
11829 | msub32 (sim_cpu *cpu) | |
11830 | { | |
11831 | unsigned rm = uimm (aarch64_get_instr (cpu), 20, 16); | |
11832 | unsigned ra = uimm (aarch64_get_instr (cpu), 14, 10); | |
11833 | unsigned rn = uimm (aarch64_get_instr (cpu), 9, 5); | |
11834 | unsigned rd = uimm (aarch64_get_instr (cpu), 4, 0); | |
11835 | ||
11836 | aarch64_set_reg_u64 (cpu, rd, NO_SP, | |
11837 | aarch64_get_reg_u32 (cpu, ra, NO_SP) | |
11838 | - aarch64_get_reg_u32 (cpu, rn, NO_SP) | |
11839 | * aarch64_get_reg_u32 (cpu, rm, NO_SP)); | |
11840 | } | |
11841 | ||
11842 | /* 64 bit multiply and sub. */ | |
11843 | static void | |
11844 | msub64 (sim_cpu *cpu) | |
11845 | { | |
11846 | unsigned rm = uimm (aarch64_get_instr (cpu), 20, 16); | |
11847 | unsigned ra = uimm (aarch64_get_instr (cpu), 14, 10); | |
11848 | unsigned rn = uimm (aarch64_get_instr (cpu), 9, 5); | |
11849 | unsigned rd = uimm (aarch64_get_instr (cpu), 4, 0); | |
11850 | ||
11851 | aarch64_set_reg_u64 (cpu, rd, NO_SP, | |
11852 | aarch64_get_reg_u64 (cpu, ra, NO_SP) | |
11853 | - aarch64_get_reg_u64 (cpu, rn, NO_SP) | |
11854 | * aarch64_get_reg_u64 (cpu, rm, NO_SP)); | |
11855 | } | |
11856 | ||
11857 | /* Signed multiply add long -- source, source2 : 32 bit, source3 : 64 bit. */ | |
11858 | static void | |
11859 | smaddl (sim_cpu *cpu) | |
11860 | { | |
11861 | unsigned rm = uimm (aarch64_get_instr (cpu), 20, 16); | |
11862 | unsigned ra = uimm (aarch64_get_instr (cpu), 14, 10); | |
11863 | unsigned rn = uimm (aarch64_get_instr (cpu), 9, 5); | |
11864 | unsigned rd = uimm (aarch64_get_instr (cpu), 4, 0); | |
11865 | ||
11866 | /* N.B. we need to multiply the signed 32 bit values in rn, rm to | |
11867 | obtain a 64 bit product. */ | |
11868 | aarch64_set_reg_s64 | |
11869 | (cpu, rd, NO_SP, | |
11870 | aarch64_get_reg_s64 (cpu, ra, NO_SP) | |
11871 | + ((int64_t) aarch64_get_reg_s32 (cpu, rn, NO_SP)) | |
11872 | * ((int64_t) aarch64_get_reg_s32 (cpu, rm, NO_SP))); | |
11873 | } | |
11874 | ||
11875 | /* Signed multiply sub long -- source, source2 : 32 bit, source3 : 64 bit. */ | |
11876 | static void | |
11877 | smsubl (sim_cpu *cpu) | |
11878 | { | |
11879 | unsigned rm = uimm (aarch64_get_instr (cpu), 20, 16); | |
11880 | unsigned ra = uimm (aarch64_get_instr (cpu), 14, 10); | |
11881 | unsigned rn = uimm (aarch64_get_instr (cpu), 9, 5); | |
11882 | unsigned rd = uimm (aarch64_get_instr (cpu), 4, 0); | |
11883 | ||
11884 | /* N.B. we need to multiply the signed 32 bit values in rn, rm to | |
11885 | obtain a 64 bit product. */ | |
11886 | aarch64_set_reg_s64 | |
11887 | (cpu, rd, NO_SP, | |
11888 | aarch64_get_reg_s64 (cpu, ra, NO_SP) | |
11889 | - ((int64_t) aarch64_get_reg_s32 (cpu, rn, NO_SP)) | |
11890 | * ((int64_t) aarch64_get_reg_s32 (cpu, rm, NO_SP))); | |
11891 | } | |
11892 | ||
11893 | /* Integer Multiply/Divide. */ | |
11894 | ||
11895 | /* First some macros and a helper function. */ | |
11896 | /* Macros to test or access elements of 64 bit words. */ | |
11897 | ||
11898 | /* Mask used to access lo 32 bits of 64 bit unsigned int. */ | |
11899 | #define LOW_WORD_MASK ((1ULL << 32) - 1) | |
11900 | /* Return the lo 32 bit word of a 64 bit unsigned int as a 64 bit unsigned int. */ | |
11901 | #define lowWordToU64(_value_u64) ((_value_u64) & LOW_WORD_MASK) | |
11902 | /* Return the hi 32 bit word of a 64 bit unsigned int as a 64 bit unsigned int. */ | |
11903 | #define highWordToU64(_value_u64) ((_value_u64) >> 32) | |
11904 | ||
11905 | /* Offset of sign bit in 64 bit signed integger. */ | |
11906 | #define SIGN_SHIFT_U64 63 | |
11907 | /* The sign bit itself -- also identifies the minimum negative int value. */ | |
11908 | #define SIGN_BIT_U64 (1UL << SIGN_SHIFT_U64) | |
11909 | /* Return true if a 64 bit signed int presented as an unsigned int is the | |
11910 | most negative value. */ | |
11911 | #define isMinimumU64(_value_u64) ((_value_u64) == SIGN_BIT_U64) | |
11912 | /* Return true (non-zero) if a 64 bit signed int presented as an unsigned | |
11913 | int has its sign bit set to false. */ | |
11914 | #define isSignSetU64(_value_u64) ((_value_u64) & SIGN_BIT_U64) | |
11915 | /* Return 1L or -1L according to whether a 64 bit signed int presented as | |
11916 | an unsigned int has its sign bit set or not. */ | |
11917 | #define signOfU64(_value_u64) (1L + (((value_u64) >> SIGN_SHIFT_U64) * -2L) | |
11918 | /* Clear the sign bit of a 64 bit signed int presented as an unsigned int. */ | |
11919 | #define clearSignU64(_value_u64) ((_value_u64) &= ~SIGN_BIT_U64) | |
11920 | ||
11921 | /* Multiply two 64 bit ints and return. | |
11922 | the hi 64 bits of the 128 bit product. */ | |
11923 | ||
11924 | static uint64_t | |
11925 | mul64hi (uint64_t value1, uint64_t value2) | |
11926 | { | |
11927 | uint64_t resultmid1; | |
11928 | uint64_t result; | |
11929 | uint64_t value1_lo = lowWordToU64 (value1); | |
11930 | uint64_t value1_hi = highWordToU64 (value1) ; | |
11931 | uint64_t value2_lo = lowWordToU64 (value2); | |
11932 | uint64_t value2_hi = highWordToU64 (value2); | |
11933 | ||
11934 | /* Cross-multiply and collect results. */ | |
11935 | ||
11936 | uint64_t xproductlo = value1_lo * value2_lo; | |
11937 | uint64_t xproductmid1 = value1_lo * value2_hi; | |
11938 | uint64_t xproductmid2 = value1_hi * value2_lo; | |
11939 | uint64_t xproducthi = value1_hi * value2_hi; | |
11940 | uint64_t carry = 0; | |
11941 | /* Start accumulating 64 bit results. */ | |
11942 | /* Drop bottom half of lowest cross-product. */ | |
11943 | uint64_t resultmid = xproductlo >> 32; | |
11944 | /* Add in middle products. */ | |
11945 | resultmid = resultmid + xproductmid1; | |
11946 | ||
11947 | /* Check for overflow. */ | |
11948 | if (resultmid < xproductmid1) | |
11949 | /* Carry over 1 into top cross-product. */ | |
11950 | carry++; | |
11951 | ||
11952 | resultmid1 = resultmid + xproductmid2; | |
11953 | ||
11954 | /* Check for overflow. */ | |
11955 | if (resultmid1 < xproductmid2) | |
11956 | /* Carry over 1 into top cross-product. */ | |
11957 | carry++; | |
11958 | ||
11959 | /* Drop lowest 32 bits of middle cross-product. */ | |
11960 | result = resultmid1 >> 32; | |
11961 | ||
11962 | /* Add top cross-product plus and any carry. */ | |
11963 | result += xproducthi + carry; | |
11964 | ||
11965 | return result; | |
11966 | } | |
11967 | ||
11968 | /* Signed multiply high, source, source2 : | |
11969 | 64 bit, dest <-- high 64-bit of result. */ | |
11970 | static void | |
11971 | smulh (sim_cpu *cpu) | |
11972 | { | |
11973 | uint64_t uresult; | |
11974 | int64_t result; | |
11975 | unsigned rm = uimm (aarch64_get_instr (cpu), 20, 16); | |
11976 | unsigned rn = uimm (aarch64_get_instr (cpu), 9, 5); | |
11977 | unsigned rd = uimm (aarch64_get_instr (cpu), 4, 0); | |
11978 | GReg ra = greg (aarch64_get_instr (cpu), 10); | |
11979 | int64_t value1 = aarch64_get_reg_u64 (cpu, rn, NO_SP); | |
11980 | int64_t value2 = aarch64_get_reg_u64 (cpu, rm, NO_SP); | |
11981 | uint64_t uvalue1; | |
11982 | uint64_t uvalue2; | |
11983 | int64_t signum = 1; | |
11984 | ||
11985 | if (ra != R31) | |
11986 | HALT_UNALLOC; | |
11987 | ||
11988 | /* Convert to unsigned and use the unsigned mul64hi routine | |
11989 | the fix the sign up afterwards. */ | |
11990 | if (value1 < 0) | |
11991 | { | |
11992 | signum *= -1L; | |
11993 | uvalue1 = -value1; | |
11994 | } | |
11995 | else | |
11996 | { | |
11997 | uvalue1 = value1; | |
11998 | } | |
11999 | ||
12000 | if (value2 < 0) | |
12001 | { | |
12002 | signum *= -1L; | |
12003 | uvalue2 = -value2; | |
12004 | } | |
12005 | else | |
12006 | { | |
12007 | uvalue2 = value2; | |
12008 | } | |
12009 | ||
12010 | uresult = mul64hi (uvalue1, uvalue2); | |
12011 | result = uresult; | |
12012 | result *= signum; | |
12013 | ||
12014 | aarch64_set_reg_s64 (cpu, rd, NO_SP, result); | |
12015 | } | |
12016 | ||
12017 | /* Unsigned multiply add long -- source, source2 : | |
12018 | 32 bit, source3 : 64 bit. */ | |
12019 | static void | |
12020 | umaddl (sim_cpu *cpu) | |
12021 | { | |
12022 | unsigned rm = uimm (aarch64_get_instr (cpu), 20, 16); | |
12023 | unsigned ra = uimm (aarch64_get_instr (cpu), 14, 10); | |
12024 | unsigned rn = uimm (aarch64_get_instr (cpu), 9, 5); | |
12025 | unsigned rd = uimm (aarch64_get_instr (cpu), 4, 0); | |
12026 | ||
12027 | /* N.B. we need to multiply the signed 32 bit values in rn, rm to | |
12028 | obtain a 64 bit product. */ | |
12029 | aarch64_set_reg_u64 | |
12030 | (cpu, rd, NO_SP, | |
12031 | aarch64_get_reg_u64 (cpu, ra, NO_SP) | |
12032 | + ((uint64_t) aarch64_get_reg_u32 (cpu, rn, NO_SP)) | |
12033 | * ((uint64_t) aarch64_get_reg_u32 (cpu, rm, NO_SP))); | |
12034 | } | |
12035 | ||
12036 | /* Unsigned multiply sub long -- source, source2 : 32 bit, source3 : 64 bit. */ | |
12037 | static void | |
12038 | umsubl (sim_cpu *cpu) | |
12039 | { | |
12040 | unsigned rm = uimm (aarch64_get_instr (cpu), 20, 16); | |
12041 | unsigned ra = uimm (aarch64_get_instr (cpu), 14, 10); | |
12042 | unsigned rn = uimm (aarch64_get_instr (cpu), 9, 5); | |
12043 | unsigned rd = uimm (aarch64_get_instr (cpu), 4, 0); | |
12044 | ||
12045 | /* N.B. we need to multiply the signed 32 bit values in rn, rm to | |
12046 | obtain a 64 bit product. */ | |
12047 | aarch64_set_reg_u64 | |
12048 | (cpu, rd, NO_SP, | |
12049 | aarch64_get_reg_u64 (cpu, ra, NO_SP) | |
12050 | - ((uint64_t) aarch64_get_reg_u32 (cpu, rn, NO_SP)) | |
12051 | * ((uint64_t) aarch64_get_reg_u32 (cpu, rm, NO_SP))); | |
12052 | } | |
12053 | ||
12054 | /* Unsigned multiply high, source, source2 : | |
12055 | 64 bit, dest <-- high 64-bit of result. */ | |
12056 | static void | |
12057 | umulh (sim_cpu *cpu) | |
12058 | { | |
12059 | unsigned rm = uimm (aarch64_get_instr (cpu), 20, 16); | |
12060 | unsigned rn = uimm (aarch64_get_instr (cpu), 9, 5); | |
12061 | unsigned rd = uimm (aarch64_get_instr (cpu), 4, 0); | |
12062 | GReg ra = greg (aarch64_get_instr (cpu), 10); | |
12063 | ||
12064 | if (ra != R31) | |
12065 | HALT_UNALLOC; | |
12066 | ||
12067 | aarch64_set_reg_u64 (cpu, rd, NO_SP, | |
12068 | mul64hi (aarch64_get_reg_u64 (cpu, rn, NO_SP), | |
12069 | aarch64_get_reg_u64 (cpu, rm, NO_SP))); | |
12070 | } | |
12071 | ||
12072 | static void | |
12073 | dexDataProc3Source (sim_cpu *cpu) | |
12074 | { | |
12075 | /* assert instr[28,24] == 11011. */ | |
12076 | /* instr[31] = size : 0 ==> 32 bit, 1 ==> 64 bit (for rd at least) | |
12077 | instr[30,29] = op54 : 00 ==> ok, ow ==> UNALLOC | |
12078 | instr[23,21] = op31 : 111 ==> UNALLOC, o2 ==> ok | |
12079 | instr[15] = o0 : 0/1 ==> ok | |
12080 | instr[23,21:15] ==> op : 0000 ==> MADD, 0001 ==> MSUB, (32/64 bit) | |
12081 | 0010 ==> SMADDL, 0011 ==> SMSUBL, (64 bit only) | |
12082 | 0100 ==> SMULH, (64 bit only) | |
12083 | 1010 ==> UMADDL, 1011 ==> UNSUBL, (64 bit only) | |
12084 | 1100 ==> UMULH (64 bit only) | |
12085 | ow ==> UNALLOC. */ | |
12086 | ||
12087 | uint32_t dispatch; | |
12088 | uint32_t size = uimm (aarch64_get_instr (cpu), 31, 31); | |
12089 | uint32_t op54 = uimm (aarch64_get_instr (cpu), 30, 29); | |
12090 | uint32_t op31 = uimm (aarch64_get_instr (cpu), 23, 21); | |
12091 | uint32_t o0 = uimm (aarch64_get_instr (cpu), 15, 15); | |
12092 | ||
12093 | if (op54 != 0) | |
12094 | HALT_UNALLOC; | |
12095 | ||
12096 | if (size == 0) | |
12097 | { | |
12098 | if (op31 != 0) | |
12099 | HALT_UNALLOC; | |
12100 | ||
12101 | if (o0 == 0) | |
12102 | madd32 (cpu); | |
12103 | else | |
12104 | msub32 (cpu); | |
12105 | return; | |
12106 | } | |
12107 | ||
12108 | dispatch = (op31 << 1) | o0; | |
12109 | ||
12110 | switch (dispatch) | |
12111 | { | |
12112 | case 0: madd64 (cpu); return; | |
12113 | case 1: msub64 (cpu); return; | |
12114 | case 2: smaddl (cpu); return; | |
12115 | case 3: smsubl (cpu); return; | |
12116 | case 4: smulh (cpu); return; | |
12117 | case 10: umaddl (cpu); return; | |
12118 | case 11: umsubl (cpu); return; | |
12119 | case 12: umulh (cpu); return; | |
12120 | default: HALT_UNALLOC; | |
12121 | } | |
12122 | } | |
12123 | ||
12124 | static void | |
12125 | dexDPReg (sim_cpu *cpu) | |
12126 | { | |
12127 | /* uint32_t group = dispatchGroup (aarch64_get_instr (cpu)); | |
12128 | assert group == GROUP_DPREG_0101 || group == GROUP_DPREG_1101 | |
12129 | bits [28:24:21] of a DPReg are the secondary dispatch vector. */ | |
12130 | uint32_t group2 = dispatchDPReg (aarch64_get_instr (cpu)); | |
12131 | ||
12132 | switch (group2) | |
12133 | { | |
12134 | case DPREG_LOG_000: | |
12135 | case DPREG_LOG_001: | |
12136 | dexLogicalShiftedRegister (cpu); return; | |
12137 | ||
12138 | case DPREG_ADDSHF_010: | |
12139 | dexAddSubtractShiftedRegister (cpu); return; | |
12140 | ||
12141 | case DPREG_ADDEXT_011: | |
12142 | dexAddSubtractExtendedRegister (cpu); return; | |
12143 | ||
12144 | case DPREG_ADDCOND_100: | |
12145 | { | |
12146 | /* This set bundles a variety of different operations. */ | |
12147 | /* Check for. */ | |
12148 | /* 1) add/sub w carry. */ | |
12149 | uint32_t mask1 = 0x1FE00000U; | |
12150 | uint32_t val1 = 0x1A000000U; | |
12151 | /* 2) cond compare register/immediate. */ | |
12152 | uint32_t mask2 = 0x1FE00000U; | |
12153 | uint32_t val2 = 0x1A400000U; | |
12154 | /* 3) cond select. */ | |
12155 | uint32_t mask3 = 0x1FE00000U; | |
12156 | uint32_t val3 = 0x1A800000U; | |
12157 | /* 4) data proc 1/2 source. */ | |
12158 | uint32_t mask4 = 0x1FE00000U; | |
12159 | uint32_t val4 = 0x1AC00000U; | |
12160 | ||
12161 | if ((aarch64_get_instr (cpu) & mask1) == val1) | |
12162 | dexAddSubtractWithCarry (cpu); | |
12163 | ||
12164 | else if ((aarch64_get_instr (cpu) & mask2) == val2) | |
12165 | CondCompare (cpu); | |
12166 | ||
12167 | else if ((aarch64_get_instr (cpu) & mask3) == val3) | |
12168 | dexCondSelect (cpu); | |
12169 | ||
12170 | else if ((aarch64_get_instr (cpu) & mask4) == val4) | |
12171 | { | |
12172 | /* Bit 30 is clear for data proc 2 source | |
12173 | and set for data proc 1 source. */ | |
12174 | if (aarch64_get_instr (cpu) & (1U << 30)) | |
12175 | dexDataProc1Source (cpu); | |
12176 | else | |
12177 | dexDataProc2Source (cpu); | |
12178 | } | |
12179 | ||
12180 | else | |
12181 | /* Should not reach here. */ | |
12182 | HALT_NYI; | |
12183 | ||
12184 | return; | |
12185 | } | |
12186 | ||
12187 | case DPREG_3SRC_110: | |
12188 | dexDataProc3Source (cpu); return; | |
12189 | ||
12190 | case DPREG_UNALLOC_101: | |
12191 | HALT_UNALLOC; | |
12192 | ||
12193 | case DPREG_3SRC_111: | |
12194 | dexDataProc3Source (cpu); return; | |
12195 | ||
12196 | default: | |
12197 | /* Should never reach here. */ | |
12198 | HALT_NYI; | |
12199 | } | |
12200 | } | |
12201 | ||
12202 | /* Unconditional Branch immediate. | |
12203 | Offset is a PC-relative byte offset in the range +/- 128MiB. | |
12204 | The offset is assumed to be raw from the decode i.e. the | |
12205 | simulator is expected to scale them from word offsets to byte. */ | |
12206 | ||
12207 | /* Unconditional branch. */ | |
12208 | static void | |
12209 | buc (sim_cpu *cpu, int32_t offset) | |
12210 | { | |
12211 | aarch64_set_next_PC_by_offset (cpu, offset); | |
12212 | } | |
12213 | ||
12214 | static unsigned stack_depth = 0; | |
12215 | ||
12216 | /* Unconditional branch and link -- writes return PC to LR. */ | |
12217 | static void | |
12218 | bl (sim_cpu *cpu, int32_t offset) | |
12219 | { | |
12220 | aarch64_save_LR (cpu); | |
12221 | aarch64_set_next_PC_by_offset (cpu, offset); | |
12222 | ||
12223 | if (TRACE_BRANCH_P (cpu)) | |
12224 | { | |
12225 | ++ stack_depth; | |
12226 | TRACE_BRANCH (cpu, | |
12227 | " %*scall %" PRIx64 " [%s]" | |
12228 | " [args: %" PRIx64 " %" PRIx64 " %" PRIx64 "]", | |
12229 | stack_depth, " ", aarch64_get_next_PC (cpu), | |
12230 | aarch64_get_func (aarch64_get_next_PC (cpu)), | |
12231 | aarch64_get_reg_u64 (cpu, 0, NO_SP), | |
12232 | aarch64_get_reg_u64 (cpu, 1, NO_SP), | |
12233 | aarch64_get_reg_u64 (cpu, 2, NO_SP) | |
12234 | ); | |
12235 | } | |
12236 | } | |
12237 | ||
12238 | /* Unconditional Branch register. | |
12239 | Branch/return address is in source register. */ | |
12240 | ||
12241 | /* Unconditional branch. */ | |
12242 | static void | |
12243 | br (sim_cpu *cpu) | |
12244 | { | |
12245 | unsigned rn = uimm (aarch64_get_instr (cpu), 9, 5); | |
12246 | aarch64_set_next_PC (cpu, aarch64_get_reg_u64 (cpu, rn, NO_SP)); | |
12247 | } | |
12248 | ||
12249 | /* Unconditional branch and link -- writes return PC to LR. */ | |
12250 | static void | |
12251 | blr (sim_cpu *cpu) | |
12252 | { | |
12253 | unsigned rn = uimm (aarch64_get_instr (cpu), 9, 5); | |
12254 | ||
12255 | /* The pseudo code in the spec says we update LR before fetching. | |
12256 | the value from the rn. */ | |
12257 | aarch64_save_LR (cpu); | |
12258 | aarch64_set_next_PC (cpu, aarch64_get_reg_u64 (cpu, rn, NO_SP)); | |
12259 | ||
12260 | if (TRACE_BRANCH_P (cpu)) | |
12261 | { | |
12262 | ++ stack_depth; | |
12263 | TRACE_BRANCH (cpu, | |
12264 | " %*scall %" PRIx64 " [%s]" | |
12265 | " [args: %" PRIx64 " %" PRIx64 " %" PRIx64 "]", | |
12266 | stack_depth, " ", aarch64_get_next_PC (cpu), | |
12267 | aarch64_get_func (aarch64_get_next_PC (cpu)), | |
12268 | aarch64_get_reg_u64 (cpu, 0, NO_SP), | |
12269 | aarch64_get_reg_u64 (cpu, 1, NO_SP), | |
12270 | aarch64_get_reg_u64 (cpu, 2, NO_SP) | |
12271 | ); | |
12272 | } | |
12273 | } | |
12274 | ||
12275 | /* Return -- assembler will default source to LR this is functionally | |
12276 | equivalent to br but, presumably, unlike br it side effects the | |
12277 | branch predictor. */ | |
12278 | static void | |
12279 | ret (sim_cpu *cpu) | |
12280 | { | |
12281 | unsigned rn = uimm (aarch64_get_instr (cpu), 9, 5); | |
12282 | aarch64_set_next_PC (cpu, aarch64_get_reg_u64 (cpu, rn, NO_SP)); | |
12283 | ||
12284 | if (TRACE_BRANCH_P (cpu)) | |
12285 | { | |
12286 | TRACE_BRANCH (cpu, | |
12287 | " %*sreturn [result: %" PRIx64 "]", | |
12288 | stack_depth, " ", aarch64_get_reg_u64 (cpu, 0, NO_SP)); | |
12289 | -- stack_depth; | |
12290 | } | |
12291 | } | |
12292 | ||
12293 | /* NOP -- we implement this and call it from the decode in case we | |
12294 | want to intercept it later. */ | |
12295 | ||
12296 | static void | |
12297 | nop (sim_cpu *cpu) | |
12298 | { | |
12299 | } | |
12300 | ||
12301 | /* Data synchronization barrier. */ | |
12302 | ||
12303 | static void | |
12304 | dsb (sim_cpu *cpu) | |
12305 | { | |
12306 | } | |
12307 | ||
12308 | /* Data memory barrier. */ | |
12309 | ||
12310 | static void | |
12311 | dmb (sim_cpu *cpu) | |
12312 | { | |
12313 | } | |
12314 | ||
12315 | /* Instruction synchronization barrier. */ | |
12316 | ||
12317 | static void | |
12318 | isb (sim_cpu *cpu) | |
12319 | { | |
12320 | } | |
12321 | ||
12322 | static void | |
12323 | dexBranchImmediate (sim_cpu *cpu) | |
12324 | { | |
12325 | /* assert instr[30,26] == 00101 | |
12326 | instr[31] ==> 0 == B, 1 == BL | |
12327 | instr[25,0] == imm26 branch offset counted in words. */ | |
12328 | ||
12329 | uint32_t top = uimm (aarch64_get_instr (cpu), 31, 31); | |
12330 | /* We have a 26 byte signed word offset which we need to pass to the | |
12331 | execute routine as a signed byte offset. */ | |
12332 | int32_t offset = simm32 (aarch64_get_instr (cpu), 25, 0) << 2; | |
12333 | ||
12334 | if (top) | |
12335 | bl (cpu, offset); | |
12336 | else | |
12337 | buc (cpu, offset); | |
12338 | } | |
12339 | ||
12340 | /* Control Flow. */ | |
12341 | ||
12342 | /* Conditional branch | |
12343 | ||
12344 | Offset is a PC-relative byte offset in the range +/- 1MiB pos is | |
12345 | a bit position in the range 0 .. 63 | |
12346 | ||
12347 | cc is a CondCode enum value as pulled out of the decode | |
12348 | ||
12349 | N.B. any offset register (source) can only be Xn or Wn. */ | |
12350 | ||
12351 | static void | |
12352 | bcc (sim_cpu *cpu, int32_t offset, CondCode cc) | |
12353 | { | |
12354 | /* the test returns TRUE if CC is met. */ | |
12355 | if (testConditionCode (cpu, cc)) | |
12356 | aarch64_set_next_PC_by_offset (cpu, offset); | |
12357 | } | |
12358 | ||
12359 | /* 32 bit branch on register non-zero. */ | |
12360 | static void | |
12361 | cbnz32 (sim_cpu *cpu, int32_t offset) | |
12362 | { | |
12363 | unsigned rt = uimm (aarch64_get_instr (cpu), 4, 0); | |
12364 | ||
12365 | if (aarch64_get_reg_u32 (cpu, rt, NO_SP) != 0) | |
12366 | aarch64_set_next_PC_by_offset (cpu, offset); | |
12367 | } | |
12368 | ||
12369 | /* 64 bit branch on register zero. */ | |
12370 | static void | |
12371 | cbnz (sim_cpu *cpu, int32_t offset) | |
12372 | { | |
12373 | unsigned rt = uimm (aarch64_get_instr (cpu), 4, 0); | |
12374 | ||
12375 | if (aarch64_get_reg_u64 (cpu, rt, NO_SP) != 0) | |
12376 | aarch64_set_next_PC_by_offset (cpu, offset); | |
12377 | } | |
12378 | ||
12379 | /* 32 bit branch on register non-zero. */ | |
12380 | static void | |
12381 | cbz32 (sim_cpu *cpu, int32_t offset) | |
12382 | { | |
12383 | unsigned rt = uimm (aarch64_get_instr (cpu), 4, 0); | |
12384 | ||
12385 | if (aarch64_get_reg_u32 (cpu, rt, NO_SP) == 0) | |
12386 | aarch64_set_next_PC_by_offset (cpu, offset); | |
12387 | } | |
12388 | ||
12389 | /* 64 bit branch on register zero. */ | |
12390 | static void | |
12391 | cbz (sim_cpu *cpu, int32_t offset) | |
12392 | { | |
12393 | unsigned rt = uimm (aarch64_get_instr (cpu), 4, 0); | |
12394 | ||
12395 | if (aarch64_get_reg_u64 (cpu, rt, NO_SP) == 0) | |
12396 | aarch64_set_next_PC_by_offset (cpu, offset); | |
12397 | } | |
12398 | ||
12399 | /* Branch on register bit test non-zero -- one size fits all. */ | |
12400 | static void | |
12401 | tbnz (sim_cpu *cpu, uint32_t pos, int32_t offset) | |
12402 | { | |
12403 | unsigned rt = uimm (aarch64_get_instr (cpu), 4, 0); | |
12404 | ||
12405 | if (aarch64_get_reg_u64 (cpu, rt, NO_SP) & (1 << pos)) | |
12406 | aarch64_set_next_PC_by_offset (cpu, offset); | |
12407 | } | |
12408 | ||
12409 | /* branch on register bit test zero -- one size fits all. */ | |
12410 | static void | |
12411 | tbz (sim_cpu *cpu, uint32_t pos, int32_t offset) | |
12412 | { | |
12413 | unsigned rt = uimm (aarch64_get_instr (cpu), 4, 0); | |
12414 | ||
12415 | if (!(aarch64_get_reg_u64 (cpu, rt, NO_SP) & (1 << pos))) | |
12416 | aarch64_set_next_PC_by_offset (cpu, offset); | |
12417 | } | |
12418 | ||
12419 | static void | |
12420 | dexCompareBranchImmediate (sim_cpu *cpu) | |
12421 | { | |
12422 | /* instr[30,25] = 01 1010 | |
12423 | instr[31] = size : 0 ==> 32, 1 ==> 64 | |
12424 | instr[24] = op : 0 ==> CBZ, 1 ==> CBNZ | |
12425 | instr[23,5] = simm19 branch offset counted in words | |
12426 | instr[4,0] = rt */ | |
12427 | ||
12428 | uint32_t size = uimm (aarch64_get_instr (cpu), 31, 31); | |
12429 | uint32_t op = uimm (aarch64_get_instr (cpu), 24, 24); | |
12430 | int32_t offset = simm32 (aarch64_get_instr (cpu), 23, 5) << 2; | |
12431 | ||
12432 | if (size == 0) | |
12433 | { | |
12434 | if (op == 0) | |
12435 | cbz32 (cpu, offset); | |
12436 | else | |
12437 | cbnz32 (cpu, offset); | |
12438 | } | |
12439 | else | |
12440 | { | |
12441 | if (op == 0) | |
12442 | cbz (cpu, offset); | |
12443 | else | |
12444 | cbnz (cpu, offset); | |
12445 | } | |
12446 | } | |
12447 | ||
12448 | static void | |
12449 | dexTestBranchImmediate (sim_cpu *cpu) | |
12450 | { | |
12451 | /* instr[31] = b5 : bit 5 of test bit idx | |
12452 | instr[30,25] = 01 1011 | |
12453 | instr[24] = op : 0 ==> TBZ, 1 == TBNZ | |
12454 | instr[23,19] = b40 : bits 4 to 0 of test bit idx | |
12455 | instr[18,5] = simm14 : signed offset counted in words | |
12456 | instr[4,0] = uimm5 */ | |
12457 | ||
12458 | uint32_t pos = ((uimm (aarch64_get_instr (cpu), 31, 31) << 4) | |
12459 | | uimm (aarch64_get_instr (cpu), 23,19)); | |
12460 | int32_t offset = simm32 (aarch64_get_instr (cpu), 18, 5) << 2; | |
12461 | ||
12462 | NYI_assert (30, 25, 0x1b); | |
12463 | ||
12464 | if (uimm (aarch64_get_instr (cpu), 24, 24) == 0) | |
12465 | tbz (cpu, pos, offset); | |
12466 | else | |
12467 | tbnz (cpu, pos, offset); | |
12468 | } | |
12469 | ||
12470 | static void | |
12471 | dexCondBranchImmediate (sim_cpu *cpu) | |
12472 | { | |
12473 | /* instr[31,25] = 010 1010 | |
12474 | instr[24] = op1; op => 00 ==> B.cond | |
12475 | instr[23,5] = simm19 : signed offset counted in words | |
12476 | instr[4] = op0 | |
12477 | instr[3,0] = cond */ | |
12478 | ||
12479 | int32_t offset; | |
12480 | CondCode cc; | |
12481 | uint32_t op = ((uimm (aarch64_get_instr (cpu), 24, 24) << 1) | |
12482 | | uimm (aarch64_get_instr (cpu), 4, 4)); | |
12483 | ||
12484 | NYI_assert (31, 25, 0x2a); | |
12485 | ||
12486 | if (op != 0) | |
12487 | HALT_UNALLOC; | |
12488 | ||
12489 | offset = simm32 (aarch64_get_instr (cpu), 23, 5) << 2; | |
12490 | cc = condcode (aarch64_get_instr (cpu), 0); | |
12491 | ||
12492 | bcc (cpu, offset, cc); | |
12493 | } | |
12494 | ||
12495 | static void | |
12496 | dexBranchRegister (sim_cpu *cpu) | |
12497 | { | |
12498 | /* instr[31,25] = 110 1011 | |
12499 | instr[24,21] = op : 0 ==> BR, 1 => BLR, 2 => RET, 3 => ERET, 4 => DRPS | |
12500 | instr[20,16] = op2 : must be 11111 | |
12501 | instr[15,10] = op3 : must be 000000 | |
12502 | instr[4,0] = op2 : must be 11111. */ | |
12503 | ||
12504 | uint32_t op = uimm (aarch64_get_instr (cpu), 24, 21); | |
12505 | uint32_t op2 = uimm (aarch64_get_instr (cpu), 20, 16); | |
12506 | uint32_t op3 = uimm (aarch64_get_instr (cpu), 15, 10); | |
12507 | uint32_t op4 = uimm (aarch64_get_instr (cpu), 4, 0); | |
12508 | ||
12509 | NYI_assert (31, 25, 0x6b); | |
12510 | ||
12511 | if (op2 != 0x1F || op3 != 0 || op4 != 0) | |
12512 | HALT_UNALLOC; | |
12513 | ||
12514 | if (op == 0) | |
12515 | br (cpu); | |
12516 | ||
12517 | else if (op == 1) | |
12518 | blr (cpu); | |
12519 | ||
12520 | else if (op == 2) | |
12521 | ret (cpu); | |
12522 | ||
12523 | else | |
12524 | { | |
12525 | /* ERET and DRPS accept 0b11111 for rn = aarch64_get_instr (cpu)[4,0]. */ | |
12526 | /* anything else is unallocated. */ | |
12527 | uint32_t rn = greg (aarch64_get_instr (cpu), 0); | |
12528 | ||
12529 | if (rn != 0x1f) | |
12530 | HALT_UNALLOC; | |
12531 | ||
12532 | if (op == 4 || op == 5) | |
12533 | HALT_NYI; | |
12534 | ||
12535 | HALT_UNALLOC; | |
12536 | } | |
12537 | } | |
12538 | ||
12539 | /* FIXME: We should get the Angel SWI values from ../../libgloss/aarch64/svc.h | |
12540 | but this may not be available. So instead we define the values we need | |
12541 | here. */ | |
12542 | #define AngelSVC_Reason_Open 0x01 | |
12543 | #define AngelSVC_Reason_Close 0x02 | |
12544 | #define AngelSVC_Reason_Write 0x05 | |
12545 | #define AngelSVC_Reason_Read 0x06 | |
12546 | #define AngelSVC_Reason_IsTTY 0x09 | |
12547 | #define AngelSVC_Reason_Seek 0x0A | |
12548 | #define AngelSVC_Reason_FLen 0x0C | |
12549 | #define AngelSVC_Reason_Remove 0x0E | |
12550 | #define AngelSVC_Reason_Rename 0x0F | |
12551 | #define AngelSVC_Reason_Clock 0x10 | |
12552 | #define AngelSVC_Reason_Time 0x11 | |
12553 | #define AngelSVC_Reason_System 0x12 | |
12554 | #define AngelSVC_Reason_Errno 0x13 | |
12555 | #define AngelSVC_Reason_GetCmdLine 0x15 | |
12556 | #define AngelSVC_Reason_HeapInfo 0x16 | |
12557 | #define AngelSVC_Reason_ReportException 0x18 | |
12558 | #define AngelSVC_Reason_Elapsed 0x30 | |
12559 | ||
12560 | ||
12561 | static void | |
12562 | handle_halt (sim_cpu *cpu, uint32_t val) | |
12563 | { | |
12564 | uint64_t result = 0; | |
12565 | ||
12566 | if (val != 0xf000) | |
12567 | { | |
12568 | TRACE_SYSCALL (cpu, " HLT [0x%x]", val); | |
12569 | sim_engine_halt (CPU_STATE (cpu), cpu, NULL, aarch64_get_PC (cpu), | |
12570 | sim_stopped, SIM_SIGTRAP); | |
12571 | } | |
12572 | ||
12573 | /* We have encountered an Angel SVC call. See if we can process it. */ | |
12574 | switch (aarch64_get_reg_u32 (cpu, 0, NO_SP)) | |
12575 | { | |
12576 | case AngelSVC_Reason_HeapInfo: | |
12577 | { | |
12578 | /* Get the values. */ | |
12579 | uint64_t stack_top = aarch64_get_stack_start (cpu); | |
12580 | uint64_t heap_base = aarch64_get_heap_start (cpu); | |
12581 | ||
12582 | /* Get the pointer */ | |
12583 | uint64_t ptr = aarch64_get_reg_u64 (cpu, 1, SP_OK); | |
12584 | ptr = aarch64_get_mem_u64 (cpu, ptr); | |
12585 | ||
12586 | /* Fill in the memory block. */ | |
12587 | /* Start addr of heap. */ | |
12588 | aarch64_set_mem_u64 (cpu, ptr + 0, heap_base); | |
12589 | /* End addr of heap. */ | |
12590 | aarch64_set_mem_u64 (cpu, ptr + 8, stack_top); | |
12591 | /* Lowest stack addr. */ | |
12592 | aarch64_set_mem_u64 (cpu, ptr + 16, heap_base); | |
12593 | /* Initial stack addr. */ | |
12594 | aarch64_set_mem_u64 (cpu, ptr + 24, stack_top); | |
12595 | ||
12596 | TRACE_SYSCALL (cpu, " AngelSVC: Get Heap Info"); | |
12597 | } | |
12598 | break; | |
12599 | ||
12600 | case AngelSVC_Reason_Open: | |
12601 | { | |
12602 | /* Get the pointer */ | |
12603 | /* uint64_t ptr = aarch64_get_reg_u64 (cpu, 1, SP_OK);. */ | |
12604 | /* FIXME: For now we just assume that we will only be asked | |
12605 | to open the standard file descriptors. */ | |
12606 | static int fd = 0; | |
12607 | result = fd ++; | |
12608 | ||
12609 | TRACE_SYSCALL (cpu, " AngelSVC: Open file %d", fd - 1); | |
12610 | } | |
12611 | break; | |
12612 | ||
12613 | case AngelSVC_Reason_Close: | |
12614 | { | |
12615 | uint64_t fh = aarch64_get_reg_u64 (cpu, 1, SP_OK); | |
12616 | TRACE_SYSCALL (cpu, " AngelSVC: Close file %d", (int) fh); | |
12617 | result = 0; | |
12618 | } | |
12619 | break; | |
12620 | ||
12621 | case AngelSVC_Reason_Errno: | |
12622 | result = 0; | |
12623 | TRACE_SYSCALL (cpu, " AngelSVC: Get Errno"); | |
12624 | break; | |
12625 | ||
12626 | case AngelSVC_Reason_Clock: | |
12627 | result = | |
12628 | #ifdef CLOCKS_PER_SEC | |
12629 | (CLOCKS_PER_SEC >= 100) | |
12630 | ? (clock () / (CLOCKS_PER_SEC / 100)) | |
12631 | : ((clock () * 100) / CLOCKS_PER_SEC) | |
12632 | #else | |
12633 | /* Presume unix... clock() returns microseconds. */ | |
12634 | (clock () / 10000) | |
12635 | #endif | |
12636 | ; | |
12637 | TRACE_SYSCALL (cpu, " AngelSVC: Get Clock"); | |
12638 | break; | |
12639 | ||
12640 | case AngelSVC_Reason_GetCmdLine: | |
12641 | { | |
12642 | /* Get the pointer */ | |
12643 | uint64_t ptr = aarch64_get_reg_u64 (cpu, 1, SP_OK); | |
12644 | ptr = aarch64_get_mem_u64 (cpu, ptr); | |
12645 | ||
12646 | /* FIXME: No command line for now. */ | |
12647 | aarch64_set_mem_u64 (cpu, ptr, 0); | |
12648 | TRACE_SYSCALL (cpu, " AngelSVC: Get Command Line"); | |
12649 | } | |
12650 | break; | |
12651 | ||
12652 | case AngelSVC_Reason_IsTTY: | |
12653 | result = 1; | |
12654 | TRACE_SYSCALL (cpu, " AngelSVC: IsTTY ?"); | |
12655 | break; | |
12656 | ||
12657 | case AngelSVC_Reason_Write: | |
12658 | { | |
12659 | /* Get the pointer */ | |
12660 | uint64_t ptr = aarch64_get_reg_u64 (cpu, 1, SP_OK); | |
12661 | /* Get the write control block. */ | |
12662 | uint64_t fd = aarch64_get_mem_u64 (cpu, ptr); | |
12663 | uint64_t buf = aarch64_get_mem_u64 (cpu, ptr + 8); | |
12664 | uint64_t len = aarch64_get_mem_u64 (cpu, ptr + 16); | |
12665 | ||
12666 | TRACE_SYSCALL (cpu, "write of %" PRIx64 " bytes from %" | |
12667 | PRIx64 " on descriptor %" PRIx64, | |
12668 | len, buf, fd); | |
12669 | ||
12670 | if (len > 1280) | |
12671 | { | |
12672 | TRACE_SYSCALL (cpu, | |
12673 | " AngelSVC: Write: Suspiciously long write: %ld", | |
12674 | (long) len); | |
12675 | sim_engine_halt (CPU_STATE (cpu), cpu, NULL, aarch64_get_PC (cpu), | |
12676 | sim_stopped, SIM_SIGBUS); | |
12677 | } | |
12678 | else if (fd == 1) | |
12679 | { | |
12680 | printf ("%.*s", (int) len, aarch64_get_mem_ptr (cpu, buf)); | |
12681 | if (disas) | |
12682 | /* So that the output stays in sync with trace output. */ | |
12683 | fflush (stdout); | |
12684 | } | |
12685 | else if (fd == 2) | |
12686 | { | |
12687 | TRACE (cpu, 0, "\n"); | |
12688 | sim_io_eprintf (CPU_STATE (cpu), "%.*s", | |
12689 | (int) len, aarch64_get_mem_ptr (cpu, buf)); | |
12690 | TRACE (cpu, 0, "\n"); | |
12691 | } | |
12692 | else | |
12693 | { | |
12694 | TRACE_SYSCALL (cpu, | |
12695 | " AngelSVC: Write: Unexpected file handle: %d", | |
12696 | (int) fd); | |
12697 | sim_engine_halt (CPU_STATE (cpu), cpu, NULL, aarch64_get_PC (cpu), | |
12698 | sim_stopped, SIM_SIGABRT); | |
12699 | } | |
12700 | } | |
12701 | break; | |
12702 | ||
12703 | case AngelSVC_Reason_ReportException: | |
12704 | { | |
12705 | /* Get the pointer */ | |
12706 | uint64_t ptr = aarch64_get_reg_u64 (cpu, 1, SP_OK); | |
12707 | /*ptr = aarch64_get_mem_u64 (cpu, ptr);. */ | |
12708 | uint64_t type = aarch64_get_mem_u64 (cpu, ptr); | |
12709 | uint64_t state = aarch64_get_mem_u64 (cpu, ptr + 8); | |
12710 | ||
12711 | TRACE_SYSCALL (cpu, | |
12712 | "Angel Exception: type 0x%" PRIx64 " state %" PRIx64, | |
12713 | type, state); | |
12714 | ||
12715 | if (type == 0x20026) | |
12716 | sim_engine_halt (CPU_STATE (cpu), cpu, NULL, aarch64_get_PC (cpu), | |
12717 | sim_exited, state); | |
12718 | else | |
12719 | sim_engine_halt (CPU_STATE (cpu), cpu, NULL, aarch64_get_PC (cpu), | |
12720 | sim_stopped, SIM_SIGINT); | |
12721 | } | |
12722 | break; | |
12723 | ||
12724 | case AngelSVC_Reason_Read: | |
12725 | case AngelSVC_Reason_FLen: | |
12726 | case AngelSVC_Reason_Seek: | |
12727 | case AngelSVC_Reason_Remove: | |
12728 | case AngelSVC_Reason_Time: | |
12729 | case AngelSVC_Reason_System: | |
12730 | case AngelSVC_Reason_Rename: | |
12731 | case AngelSVC_Reason_Elapsed: | |
12732 | default: | |
12733 | TRACE_SYSCALL (cpu, " HLT [Unknown angel %x]", | |
12734 | aarch64_get_reg_u32 (cpu, 0, NO_SP)); | |
12735 | sim_engine_halt (CPU_STATE (cpu), cpu, NULL, aarch64_get_PC (cpu), | |
12736 | sim_stopped, SIM_SIGTRAP); | |
12737 | } | |
12738 | ||
12739 | aarch64_set_reg_u64 (cpu, 0, NO_SP, result); | |
12740 | } | |
12741 | ||
12742 | static void | |
12743 | dexExcpnGen (sim_cpu *cpu) | |
12744 | { | |
12745 | /* instr[31:24] = 11010100 | |
12746 | instr[23,21] = opc : 000 ==> GEN EXCPN, 001 ==> BRK | |
12747 | 010 ==> HLT, 101 ==> DBG GEN EXCPN | |
12748 | instr[20,5] = imm16 | |
12749 | instr[4,2] = opc2 000 ==> OK, ow ==> UNALLOC | |
12750 | instr[1,0] = LL : discriminates opc */ | |
12751 | ||
12752 | uint32_t opc = uimm (aarch64_get_instr (cpu), 23, 21); | |
12753 | uint32_t imm16 = uimm (aarch64_get_instr (cpu), 20, 5); | |
12754 | uint32_t opc2 = uimm (aarch64_get_instr (cpu), 4, 2); | |
12755 | uint32_t LL; | |
12756 | ||
12757 | NYI_assert (31, 24, 0xd4); | |
12758 | ||
12759 | if (opc2 != 0) | |
12760 | HALT_UNALLOC; | |
12761 | ||
12762 | LL = uimm (aarch64_get_instr (cpu), 1, 0); | |
12763 | ||
12764 | /* We only implement HLT and BRK for now. */ | |
12765 | if (opc == 1 && LL == 0) | |
12766 | { | |
12767 | TRACE_EVENTS (cpu, " BRK [0x%x]", imm16); | |
12768 | sim_engine_halt (CPU_STATE (cpu), cpu, NULL, aarch64_get_PC (cpu), | |
12769 | sim_exited, aarch64_get_reg_s32 (cpu, R0, SP_OK)); | |
12770 | } | |
12771 | ||
12772 | if (opc == 2 && LL == 0) | |
12773 | handle_halt (cpu, imm16); | |
12774 | ||
12775 | else if (opc == 0 || opc == 5) | |
12776 | HALT_NYI; | |
12777 | ||
12778 | else | |
12779 | HALT_UNALLOC; | |
12780 | } | |
12781 | ||
caa8d700 NC |
12782 | /* Stub for accessing system registers. |
12783 | We implement support for the DCZID register since this is used | |
12784 | by the C library's memset function. */ | |
12785 | ||
12786 | static uint64_t | |
12787 | system_get (sim_cpu *cpu, unsigned op0, unsigned op1, unsigned crn, | |
12788 | unsigned crm, unsigned op2) | |
12789 | { | |
12790 | if (crn == 0 && op1 == 3 && crm == 0 && op2 == 7) | |
12791 | /* DCZID_EL0 - the Data Cache Zero ID register. | |
12792 | We do not support DC ZVA at the moment, so | |
12793 | we return a value with the disable bit set. */ | |
12794 | return ((uint64_t) 1) << 4; | |
12795 | ||
12796 | HALT_NYI; | |
12797 | } | |
12798 | ||
12799 | static void | |
12800 | do_mrs (sim_cpu *cpu) | |
12801 | { | |
12802 | /* instr[31:20] = 1101 01010 0011 | |
12803 | instr[19] = op0 | |
12804 | instr[18,16] = op1 | |
12805 | instr[15,12] = CRn | |
12806 | instr[11,8] = CRm | |
12807 | instr[7,5] = op2 | |
12808 | instr[4,0] = Rt */ | |
12809 | unsigned sys_op0 = uimm (aarch64_get_instr (cpu), 19, 19) + 2; | |
12810 | unsigned sys_op1 = uimm (aarch64_get_instr (cpu), 18, 16); | |
12811 | unsigned sys_crn = uimm (aarch64_get_instr (cpu), 15, 12); | |
12812 | unsigned sys_crm = uimm (aarch64_get_instr (cpu), 11, 8); | |
12813 | unsigned sys_op2 = uimm (aarch64_get_instr (cpu), 7, 5); | |
12814 | unsigned rt = uimm (aarch64_get_instr (cpu), 4, 0); | |
12815 | ||
12816 | aarch64_set_reg_u64 (cpu, rt, NO_SP, | |
12817 | system_get (cpu, sys_op0, sys_op1, sys_crn, sys_crm, sys_op2)); | |
12818 | } | |
12819 | ||
2e8cf49e NC |
12820 | static void |
12821 | dexSystem (sim_cpu *cpu) | |
12822 | { | |
12823 | /* instr[31:22] = 1101 01010 0 | |
12824 | instr[21] = L | |
12825 | instr[20,19] = op0 | |
12826 | instr[18,16] = op1 | |
12827 | instr[15,12] = CRn | |
12828 | instr[11,8] = CRm | |
12829 | instr[7,5] = op2 | |
12830 | instr[4,0] = uimm5 */ | |
12831 | ||
12832 | /* We are interested in HINT, DSB, DMB and ISB | |
12833 | ||
12834 | Hint #0 encodes NOOP (this is the only hint we care about) | |
12835 | L == 0, op0 == 0, op1 = 011, CRn = 0010, Rt = 11111, | |
12836 | CRm op2 != 0000 000 OR CRm op2 == 0000 000 || CRm op > 0000 101 | |
12837 | ||
12838 | DSB, DMB, ISB are data store barrier, data memory barrier and | |
12839 | instruction store barrier, respectively, where | |
12840 | ||
12841 | L == 0, op0 == 0, op1 = 011, CRn = 0011, Rt = 11111, | |
12842 | op2 : DSB ==> 100, DMB ==> 101, ISB ==> 110 | |
12843 | CRm<3:2> ==> domain, CRm<1:0> ==> types, | |
12844 | domain : 00 ==> OuterShareable, 01 ==> Nonshareable, | |
12845 | 10 ==> InerShareable, 11 ==> FullSystem | |
12846 | types : 01 ==> Reads, 10 ==> Writes, | |
12847 | 11 ==> All, 00 ==> All (domain == FullSystem). */ | |
12848 | ||
12849 | unsigned rt = uimm (aarch64_get_instr (cpu), 4, 0); | |
12850 | uint32_t l_op0_op1_crn = uimm (aarch64_get_instr (cpu), 21, 12); | |
12851 | ||
12852 | NYI_assert (31, 22, 0x354); | |
12853 | ||
12854 | switch (l_op0_op1_crn) | |
12855 | { | |
12856 | case 0x032: | |
12857 | if (rt == 0x1F) | |
12858 | { | |
12859 | /* NOP has CRm != 0000 OR. */ | |
12860 | /* (CRm == 0000 AND (op2 == 000 OR op2 > 101)). */ | |
12861 | uint32_t crm = uimm (aarch64_get_instr (cpu), 11, 8); | |
12862 | uint32_t op2 = uimm (aarch64_get_instr (cpu), 7, 5); | |
12863 | ||
12864 | if (crm != 0 || (op2 == 0 || op2 > 5)) | |
12865 | { | |
12866 | /* Actually call nop method so we can reimplement it later. */ | |
12867 | nop (cpu); | |
12868 | return; | |
12869 | } | |
12870 | } | |
12871 | HALT_NYI; | |
12872 | ||
12873 | case 0x033: | |
12874 | { | |
12875 | uint32_t op2 = uimm (aarch64_get_instr (cpu), 7, 5); | |
12876 | ||
12877 | switch (op2) | |
12878 | { | |
caa8d700 | 12879 | case 2: HALT_NYI; |
2e8cf49e NC |
12880 | case 4: dsb (cpu); return; |
12881 | case 5: dmb (cpu); return; | |
12882 | case 6: isb (cpu); return; | |
12883 | case 7: | |
12884 | default: HALT_UNALLOC; | |
12885 | } | |
12886 | } | |
12887 | ||
12888 | case 0x3B0: | |
12889 | /* MRS Wt, sys-reg. */ | |
caa8d700 | 12890 | do_mrs (cpu); |
2e8cf49e NC |
12891 | return; |
12892 | ||
12893 | case 0x3B4: | |
12894 | case 0x3BD: | |
12895 | /* MRS Xt, sys-reg. */ | |
caa8d700 | 12896 | do_mrs (cpu); |
2e8cf49e NC |
12897 | return; |
12898 | ||
12899 | case 0x0B7: | |
12900 | /* DC <type>, x<n>. */ | |
caa8d700 | 12901 | HALT_NYI; |
2e8cf49e NC |
12902 | return; |
12903 | ||
12904 | default: | |
caa8d700 NC |
12905 | /* if (uimm (aarch64_get_instr (cpu), 21, 20) == 0x1) |
12906 | MRS Xt, sys-reg. */ | |
2e8cf49e | 12907 | HALT_NYI; |
caa8d700 | 12908 | return; |
2e8cf49e NC |
12909 | } |
12910 | } | |
12911 | ||
12912 | static void | |
12913 | dexBr (sim_cpu *cpu) | |
12914 | { | |
12915 | /* uint32_t group = dispatchGroup (aarch64_get_instr (cpu)); | |
12916 | assert group == GROUP_BREXSYS_1010 || group == GROUP_BREXSYS_1011 | |
12917 | bits [31,29] of a BrExSys are the secondary dispatch vector. */ | |
12918 | uint32_t group2 = dispatchBrExSys (aarch64_get_instr (cpu)); | |
12919 | ||
12920 | switch (group2) | |
12921 | { | |
12922 | case BR_IMM_000: | |
12923 | return dexBranchImmediate (cpu); | |
12924 | ||
12925 | case BR_IMMCMP_001: | |
12926 | /* Compare has bit 25 clear while test has it set. */ | |
12927 | if (!uimm (aarch64_get_instr (cpu), 25, 25)) | |
12928 | dexCompareBranchImmediate (cpu); | |
12929 | else | |
12930 | dexTestBranchImmediate (cpu); | |
12931 | return; | |
12932 | ||
12933 | case BR_IMMCOND_010: | |
12934 | /* This is a conditional branch if bit 25 is clear otherwise | |
12935 | unallocated. */ | |
12936 | if (!uimm (aarch64_get_instr (cpu), 25, 25)) | |
12937 | dexCondBranchImmediate (cpu); | |
12938 | else | |
12939 | HALT_UNALLOC; | |
12940 | return; | |
12941 | ||
12942 | case BR_UNALLOC_011: | |
12943 | HALT_UNALLOC; | |
12944 | ||
12945 | case BR_IMM_100: | |
12946 | dexBranchImmediate (cpu); | |
12947 | return; | |
12948 | ||
12949 | case BR_IMMCMP_101: | |
12950 | /* Compare has bit 25 clear while test has it set. */ | |
12951 | if (!uimm (aarch64_get_instr (cpu), 25, 25)) | |
12952 | dexCompareBranchImmediate (cpu); | |
12953 | else | |
12954 | dexTestBranchImmediate (cpu); | |
12955 | return; | |
12956 | ||
12957 | case BR_REG_110: | |
12958 | /* Unconditional branch reg has bit 25 set. */ | |
12959 | if (uimm (aarch64_get_instr (cpu), 25, 25)) | |
12960 | dexBranchRegister (cpu); | |
12961 | ||
12962 | /* This includes both Excpn Gen, System and unalloc operations. | |
12963 | We need to decode the Excpn Gen operation BRK so we can plant | |
12964 | debugger entry points. | |
12965 | Excpn Gen operations have aarch64_get_instr (cpu)[24] = 0. | |
12966 | we need to decode at least one of the System operations NOP | |
12967 | which is an alias for HINT #0. | |
12968 | System operations have aarch64_get_instr (cpu)[24,22] = 100. */ | |
12969 | else if (uimm (aarch64_get_instr (cpu), 24, 24) == 0) | |
12970 | dexExcpnGen (cpu); | |
12971 | ||
12972 | else if (uimm (aarch64_get_instr (cpu), 24, 22) == 4) | |
12973 | dexSystem (cpu); | |
12974 | ||
12975 | else | |
12976 | HALT_UNALLOC; | |
12977 | ||
12978 | return; | |
12979 | ||
12980 | case BR_UNALLOC_111: | |
12981 | HALT_UNALLOC; | |
12982 | ||
12983 | default: | |
12984 | /* Should never reach here. */ | |
12985 | HALT_NYI; | |
12986 | } | |
12987 | } | |
12988 | ||
12989 | static void | |
12990 | aarch64_decode_and_execute (sim_cpu *cpu, uint64_t pc) | |
12991 | { | |
12992 | /* We need to check if gdb wants an in here. */ | |
12993 | /* checkBreak (cpu);. */ | |
12994 | ||
12995 | uint64_t group = dispatchGroup (aarch64_get_instr (cpu)); | |
12996 | ||
12997 | switch (group) | |
12998 | { | |
12999 | case GROUP_PSEUDO_0000: dexPseudo (cpu); break; | |
13000 | case GROUP_LDST_0100: dexLdSt (cpu); break; | |
13001 | case GROUP_DPREG_0101: dexDPReg (cpu); break; | |
13002 | case GROUP_LDST_0110: dexLdSt (cpu); break; | |
13003 | case GROUP_ADVSIMD_0111: dexAdvSIMD0 (cpu); break; | |
13004 | case GROUP_DPIMM_1000: dexDPImm (cpu); break; | |
13005 | case GROUP_DPIMM_1001: dexDPImm (cpu); break; | |
13006 | case GROUP_BREXSYS_1010: dexBr (cpu); break; | |
13007 | case GROUP_BREXSYS_1011: dexBr (cpu); break; | |
13008 | case GROUP_LDST_1100: dexLdSt (cpu); break; | |
13009 | case GROUP_DPREG_1101: dexDPReg (cpu); break; | |
13010 | case GROUP_LDST_1110: dexLdSt (cpu); break; | |
13011 | case GROUP_ADVSIMD_1111: dexAdvSIMD1 (cpu); break; | |
13012 | ||
13013 | case GROUP_UNALLOC_0001: | |
13014 | case GROUP_UNALLOC_0010: | |
13015 | case GROUP_UNALLOC_0011: | |
13016 | HALT_UNALLOC; | |
13017 | ||
13018 | default: | |
13019 | /* Should never reach here. */ | |
13020 | HALT_NYI; | |
13021 | } | |
13022 | } | |
13023 | ||
13024 | static bfd_boolean | |
13025 | aarch64_step (sim_cpu *cpu) | |
13026 | { | |
13027 | uint64_t pc = aarch64_get_PC (cpu); | |
13028 | ||
13029 | if (pc == TOP_LEVEL_RETURN_PC) | |
13030 | return FALSE; | |
13031 | ||
13032 | aarch64_set_next_PC (cpu, pc + 4); | |
13033 | aarch64_get_instr (cpu) = aarch64_get_mem_u32 (cpu, pc); | |
13034 | ||
13035 | if (TRACE_INSN_P (cpu)) | |
13036 | { | |
13037 | if (disas) | |
13038 | TRACE_INSN (cpu, " pc = %" PRIx64 " ", pc); | |
13039 | else | |
13040 | TRACE_INSN (cpu, " pc = %" PRIx64 " instr = %x", pc, | |
13041 | aarch64_get_instr (cpu)); | |
13042 | } | |
13043 | else if (disas) | |
13044 | sim_io_eprintf (CPU_STATE (cpu), " %" PRIx64 " ", pc); | |
13045 | ||
13046 | if (disas) | |
13047 | aarch64_print_insn (CPU_STATE (cpu), pc); | |
13048 | ||
13049 | aarch64_decode_and_execute (cpu, pc); | |
13050 | ||
13051 | return TRUE; | |
13052 | } | |
13053 | ||
13054 | void | |
13055 | aarch64_run (SIM_DESC sd) | |
13056 | { | |
13057 | sim_cpu *cpu = STATE_CPU (sd, 0); | |
13058 | ||
13059 | while (aarch64_step (cpu)) | |
13060 | aarch64_update_PC (cpu); | |
13061 | ||
13062 | sim_engine_halt (sd, NULL, NULL, aarch64_get_PC (cpu), | |
13063 | sim_exited, aarch64_get_reg_s32 (cpu, R0, SP_OK)); | |
13064 | } | |
13065 | ||
13066 | void | |
13067 | aarch64_init (sim_cpu *cpu, uint64_t pc) | |
13068 | { | |
13069 | uint64_t sp = aarch64_get_stack_start (cpu); | |
13070 | ||
13071 | /* Install SP, FP and PC and set LR to -20 | |
13072 | so we can detect a top-level return. */ | |
13073 | aarch64_set_reg_u64 (cpu, SP, SP_OK, sp); | |
13074 | aarch64_set_reg_u64 (cpu, FP, SP_OK, sp); | |
13075 | aarch64_set_reg_u64 (cpu, LR, SP_OK, TOP_LEVEL_RETURN_PC); | |
13076 | aarch64_set_next_PC (cpu, pc); | |
13077 | aarch64_update_PC (cpu); | |
13078 | aarch64_init_LIT_table (); | |
13079 | } |