virtio-scsi: suppress virtqueue kick during processing
[qemu.git] / target / tilegx / translate.c
1 /*
2 * QEMU TILE-Gx CPU
3 *
4 * Copyright (c) 2015 Chen Gang
5 *
6 * This library is free software; you can redistribute it and/or
7 * modify it under the terms of the GNU Lesser General Public
8 * License as published by the Free Software Foundation; either
9 * version 2.1 of the License, or (at your option) any later version.
10 *
11 * This library is distributed in the hope that it will be useful,
12 * but WITHOUT ANY WARRANTY; without even the implied warranty of
13 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
14 * Lesser General Public License for more details.
15 *
16 * You should have received a copy of the GNU Lesser General Public
17 * License along with this library; if not, see
18 * <http://www.gnu.org/licenses/lgpl-2.1.html>
19 */
20
21 #include "qemu/osdep.h"
22 #include "cpu.h"
23 #include "qemu/log.h"
24 #include "exec/log.h"
25 #include "disas/disas.h"
26 #include "exec/exec-all.h"
27 #include "tcg-op.h"
28 #include "exec/cpu_ldst.h"
29 #include "linux-user/syscall_defs.h"
30
31 #include "opcode_tilegx.h"
32 #include "spr_def_64.h"
33
34 #define FMT64X "%016" PRIx64
35
36 static TCGv_env cpu_env;
37 static TCGv cpu_pc;
38 static TCGv cpu_regs[TILEGX_R_COUNT];
39
40 static const char * const reg_names[64] = {
41 "r0", "r1", "r2", "r3", "r4", "r5", "r6", "r7",
42 "r8", "r9", "r10", "r11", "r12", "r13", "r14", "r15",
43 "r16", "r17", "r18", "r19", "r20", "r21", "r22", "r23",
44 "r24", "r25", "r26", "r27", "r28", "r29", "r30", "r31",
45 "r32", "r33", "r34", "r35", "r36", "r37", "r38", "r39",
46 "r40", "r41", "r42", "r43", "r44", "r45", "r46", "r47",
47 "r48", "r49", "r50", "r51", "bp", "tp", "sp", "lr",
48 "sn", "idn0", "idn1", "udn0", "udn1", "udn2", "udn2", "zero"
49 };
50
51 /* Modified registers are cached in temporaries until the end of the bundle. */
52 typedef struct {
53 unsigned reg;
54 TCGv val;
55 } DisasContextTemp;
56
57 #define MAX_WRITEBACK 4
58
59 /* This is the state at translation time. */
60 typedef struct {
61 uint64_t pc; /* Current pc */
62
63 TCGv zero; /* For zero register */
64
65 DisasContextTemp wb[MAX_WRITEBACK];
66 int num_wb;
67 int mmuidx;
68 bool exit_tb;
69 TileExcp atomic_excp;
70
71 struct {
72 TCGCond cond; /* branch condition */
73 TCGv dest; /* branch destination */
74 TCGv val1; /* value to be compared against zero, for cond */
75 } jmp; /* Jump object, only once in each TB block */
76 } DisasContext;
77
78 #include "exec/gen-icount.h"
79
80 /* Differentiate the various pipe encodings. */
81 #define TY_X0 0
82 #define TY_X1 1
83 #define TY_Y0 2
84 #define TY_Y1 3
85
86 /* Remerge the base opcode and extension fields for switching.
87 The X opcode fields are 3 bits; Y0/Y1 opcode fields are 4 bits;
88 Y2 opcode field is 2 bits. */
89 #define OE(OP, EXT, XY) (TY_##XY + OP * 4 + EXT * 64)
90
91 /* Similar, but for Y2 only. */
92 #define OEY2(OP, MODE) (OP + MODE * 4)
93
94 /* Similar, but make sure opcode names match up. */
95 #define OE_RR_X0(E) OE(RRR_0_OPCODE_X0, E##_UNARY_OPCODE_X0, X0)
96 #define OE_RR_X1(E) OE(RRR_0_OPCODE_X1, E##_UNARY_OPCODE_X1, X1)
97 #define OE_RR_Y0(E) OE(RRR_1_OPCODE_Y0, E##_UNARY_OPCODE_Y0, Y0)
98 #define OE_RR_Y1(E) OE(RRR_1_OPCODE_Y1, E##_UNARY_OPCODE_Y1, Y1)
99 #define OE_RRR(E,N,XY) OE(RRR_##N##_OPCODE_##XY, E##_RRR_##N##_OPCODE_##XY, XY)
100 #define OE_IM(E,XY) OE(IMM8_OPCODE_##XY, E##_IMM8_OPCODE_##XY, XY)
101 #define OE_SH(E,XY) OE(SHIFT_OPCODE_##XY, E##_SHIFT_OPCODE_##XY, XY)
102
103 #define V1_IMM(X) (((X) & 0xff) * 0x0101010101010101ull)
104 #define V2_IMM(X) (((X) & 0xffff) * 0x0001000100010001ull)
105
106
107 static void gen_exception(DisasContext *dc, TileExcp num)
108 {
109 TCGv_i32 tmp;
110
111 tcg_gen_movi_tl(cpu_pc, dc->pc + TILEGX_BUNDLE_SIZE_IN_BYTES);
112
113 tmp = tcg_const_i32(num);
114 gen_helper_exception(cpu_env, tmp);
115 tcg_temp_free_i32(tmp);
116 dc->exit_tb = true;
117 }
118
119 static bool check_gr(DisasContext *dc, uint8_t reg)
120 {
121 if (likely(reg < TILEGX_R_COUNT)) {
122 return true;
123 }
124
125 switch (reg) {
126 case TILEGX_R_SN:
127 case TILEGX_R_ZERO:
128 break;
129 case TILEGX_R_IDN0:
130 case TILEGX_R_IDN1:
131 gen_exception(dc, TILEGX_EXCP_REG_IDN_ACCESS);
132 break;
133 case TILEGX_R_UDN0:
134 case TILEGX_R_UDN1:
135 case TILEGX_R_UDN2:
136 case TILEGX_R_UDN3:
137 gen_exception(dc, TILEGX_EXCP_REG_UDN_ACCESS);
138 break;
139 default:
140 g_assert_not_reached();
141 }
142 return false;
143 }
144
145 static TCGv load_zero(DisasContext *dc)
146 {
147 if (TCGV_IS_UNUSED_I64(dc->zero)) {
148 dc->zero = tcg_const_i64(0);
149 }
150 return dc->zero;
151 }
152
153 static TCGv load_gr(DisasContext *dc, unsigned reg)
154 {
155 if (check_gr(dc, reg)) {
156 return cpu_regs[reg];
157 }
158 return load_zero(dc);
159 }
160
161 static TCGv dest_gr(DisasContext *dc, unsigned reg)
162 {
163 int n;
164
165 /* Skip the result, mark the exception if necessary, and continue */
166 check_gr(dc, reg);
167
168 n = dc->num_wb++;
169 dc->wb[n].reg = reg;
170 return dc->wb[n].val = tcg_temp_new_i64();
171 }
172
173 static void gen_saturate_op(TCGv tdest, TCGv tsrca, TCGv tsrcb,
174 void (*operate)(TCGv, TCGv, TCGv))
175 {
176 TCGv t0 = tcg_temp_new();
177
178 tcg_gen_ext32s_tl(tdest, tsrca);
179 tcg_gen_ext32s_tl(t0, tsrcb);
180 operate(tdest, tdest, t0);
181
182 tcg_gen_movi_tl(t0, 0x7fffffff);
183 tcg_gen_movcond_tl(TCG_COND_GT, tdest, tdest, t0, t0, tdest);
184 tcg_gen_movi_tl(t0, -0x80000000LL);
185 tcg_gen_movcond_tl(TCG_COND_LT, tdest, tdest, t0, t0, tdest);
186
187 tcg_temp_free(t0);
188 }
189
190 static void gen_atomic_excp(DisasContext *dc, unsigned dest, TCGv tdest,
191 TCGv tsrca, TCGv tsrcb, TileExcp excp)
192 {
193 #ifdef CONFIG_USER_ONLY
194 TCGv_i32 t;
195
196 tcg_gen_st_tl(tsrca, cpu_env, offsetof(CPUTLGState, atomic_srca));
197 tcg_gen_st_tl(tsrcb, cpu_env, offsetof(CPUTLGState, atomic_srcb));
198 t = tcg_const_i32(dest);
199 tcg_gen_st_i32(t, cpu_env, offsetof(CPUTLGState, atomic_dstr));
200 tcg_temp_free_i32(t);
201
202 /* We're going to write the real result in the exception. But in
203 the meantime we've already created a writeback register, and
204 we don't want that to remain uninitialized. */
205 tcg_gen_movi_tl(tdest, 0);
206
207 /* Note that we need to delay issuing the exception that implements
208 the atomic operation until after writing back the results of the
209 instruction occupying the X0 pipe. */
210 dc->atomic_excp = excp;
211 #else
212 gen_exception(dc, TILEGX_EXCP_OPCODE_UNIMPLEMENTED);
213 #endif
214 }
215
216 /* Shift the 128-bit value TSRCA:TSRCD right by the number of bytes
217 specified by the bottom 3 bits of TSRCB, and set TDEST to the
218 low 64 bits of the resulting value. */
219 static void gen_dblalign(TCGv tdest, TCGv tsrcd, TCGv tsrca, TCGv tsrcb)
220 {
221 TCGv t0 = tcg_temp_new();
222
223 tcg_gen_andi_tl(t0, tsrcb, 7);
224 tcg_gen_shli_tl(t0, t0, 3);
225 tcg_gen_shr_tl(tdest, tsrcd, t0);
226
227 /* We want to do "t0 = tsrca << (64 - t0)". Two's complement
228 arithmetic on a 6-bit field tells us that 64 - t0 is equal
229 to (t0 ^ 63) + 1. So we can do the shift in two parts,
230 neither of which will be an invalid shift by 64. */
231 tcg_gen_xori_tl(t0, t0, 63);
232 tcg_gen_shl_tl(t0, tsrca, t0);
233 tcg_gen_shli_tl(t0, t0, 1);
234 tcg_gen_or_tl(tdest, tdest, t0);
235
236 tcg_temp_free(t0);
237 }
238
239 /* Similarly, except that the 128-bit value is TSRCA:TSRCB, and the
240 right shift is an immediate. */
241 static void gen_dblaligni(TCGv tdest, TCGv tsrca, TCGv tsrcb, int shr)
242 {
243 TCGv t0 = tcg_temp_new();
244
245 tcg_gen_shri_tl(t0, tsrcb, shr);
246 tcg_gen_shli_tl(tdest, tsrca, 64 - shr);
247 tcg_gen_or_tl(tdest, tdest, t0);
248
249 tcg_temp_free(t0);
250 }
251
252 typedef enum {
253 LU, LS, HU, HS
254 } MulHalf;
255
256 static void gen_ext_half(TCGv d, TCGv s, MulHalf h)
257 {
258 switch (h) {
259 case LU:
260 tcg_gen_ext32u_tl(d, s);
261 break;
262 case LS:
263 tcg_gen_ext32s_tl(d, s);
264 break;
265 case HU:
266 tcg_gen_shri_tl(d, s, 32);
267 break;
268 case HS:
269 tcg_gen_sari_tl(d, s, 32);
270 break;
271 }
272 }
273
274 static void gen_mul_half(TCGv tdest, TCGv tsrca, TCGv tsrcb,
275 MulHalf ha, MulHalf hb)
276 {
277 TCGv t = tcg_temp_new();
278 gen_ext_half(t, tsrca, ha);
279 gen_ext_half(tdest, tsrcb, hb);
280 tcg_gen_mul_tl(tdest, tdest, t);
281 tcg_temp_free(t);
282 }
283
284 static void gen_cmul2(TCGv tdest, TCGv tsrca, TCGv tsrcb, int sh, int rd)
285 {
286 TCGv_i32 tsh = tcg_const_i32(sh);
287 TCGv_i32 trd = tcg_const_i32(rd);
288 gen_helper_cmul2(tdest, tsrca, tsrcb, tsh, trd);
289 tcg_temp_free_i32(tsh);
290 tcg_temp_free_i32(trd);
291 }
292
293 static TileExcp gen_st_opcode(DisasContext *dc, unsigned dest, unsigned srca,
294 unsigned srcb, TCGMemOp memop, const char *name)
295 {
296 if (dest) {
297 return TILEGX_EXCP_OPCODE_UNKNOWN;
298 }
299
300 tcg_gen_qemu_st_tl(load_gr(dc, srcb), load_gr(dc, srca),
301 dc->mmuidx, memop);
302
303 qemu_log_mask(CPU_LOG_TB_IN_ASM, "%s %s, %s", name,
304 reg_names[srca], reg_names[srcb]);
305 return TILEGX_EXCP_NONE;
306 }
307
308 static TileExcp gen_st_add_opcode(DisasContext *dc, unsigned srca, unsigned srcb,
309 int imm, TCGMemOp memop, const char *name)
310 {
311 TCGv tsrca = load_gr(dc, srca);
312 TCGv tsrcb = load_gr(dc, srcb);
313
314 tcg_gen_qemu_st_tl(tsrcb, tsrca, dc->mmuidx, memop);
315 tcg_gen_addi_tl(dest_gr(dc, srca), tsrca, imm);
316
317 qemu_log_mask(CPU_LOG_TB_IN_ASM, "%s %s, %s, %d", name,
318 reg_names[srca], reg_names[srcb], imm);
319 return TILEGX_EXCP_NONE;
320 }
321
322 /* Equality comparison with zero can be done quickly and efficiently. */
323 static void gen_v1cmpeq0(TCGv v)
324 {
325 TCGv m = tcg_const_tl(V1_IMM(0x7f));
326 TCGv c = tcg_temp_new();
327
328 /* ~(((v & m) + m) | m | v). Sets the msb for each byte == 0. */
329 tcg_gen_and_tl(c, v, m);
330 tcg_gen_add_tl(c, c, m);
331 tcg_gen_or_tl(c, c, m);
332 tcg_gen_nor_tl(c, c, v);
333 tcg_temp_free(m);
334
335 /* Shift the msb down to form the lsb boolean result. */
336 tcg_gen_shri_tl(v, c, 7);
337 tcg_temp_free(c);
338 }
339
340 static void gen_v1cmpne0(TCGv v)
341 {
342 TCGv m = tcg_const_tl(V1_IMM(0x7f));
343 TCGv c = tcg_temp_new();
344
345 /* (((v & m) + m) | v) & ~m. Sets the msb for each byte != 0. */
346 tcg_gen_and_tl(c, v, m);
347 tcg_gen_add_tl(c, c, m);
348 tcg_gen_or_tl(c, c, v);
349 tcg_gen_andc_tl(c, c, m);
350 tcg_temp_free(m);
351
352 /* Shift the msb down to form the lsb boolean result. */
353 tcg_gen_shri_tl(v, c, 7);
354 tcg_temp_free(c);
355 }
356
357 /* Vector addition can be performed via arithmetic plus masking. It is
358 efficient this way only for 4 or more elements. */
359 static void gen_v12add(TCGv tdest, TCGv tsrca, TCGv tsrcb, uint64_t sign)
360 {
361 TCGv tmask = tcg_const_tl(~sign);
362 TCGv t0 = tcg_temp_new();
363 TCGv t1 = tcg_temp_new();
364
365 /* ((a & ~sign) + (b & ~sign)) ^ ((a ^ b) & sign). */
366 tcg_gen_and_tl(t0, tsrca, tmask);
367 tcg_gen_and_tl(t1, tsrcb, tmask);
368 tcg_gen_add_tl(tdest, t0, t1);
369 tcg_gen_xor_tl(t0, tsrca, tsrcb);
370 tcg_gen_andc_tl(t0, t0, tmask);
371 tcg_gen_xor_tl(tdest, tdest, t0);
372
373 tcg_temp_free(t1);
374 tcg_temp_free(t0);
375 tcg_temp_free(tmask);
376 }
377
378 /* Similarly for vector subtraction. */
379 static void gen_v12sub(TCGv tdest, TCGv tsrca, TCGv tsrcb, uint64_t sign)
380 {
381 TCGv tsign = tcg_const_tl(sign);
382 TCGv t0 = tcg_temp_new();
383 TCGv t1 = tcg_temp_new();
384
385 /* ((a | sign) - (b & ~sign)) ^ ((a ^ ~b) & sign). */
386 tcg_gen_or_tl(t0, tsrca, tsign);
387 tcg_gen_andc_tl(t1, tsrcb, tsign);
388 tcg_gen_sub_tl(tdest, t0, t1);
389 tcg_gen_eqv_tl(t0, tsrca, tsrcb);
390 tcg_gen_and_tl(t0, t0, tsign);
391 tcg_gen_xor_tl(tdest, tdest, t0);
392
393 tcg_temp_free(t1);
394 tcg_temp_free(t0);
395 tcg_temp_free(tsign);
396 }
397
398 static void gen_v4sh(TCGv d64, TCGv a64, TCGv b64,
399 void (*generate)(TCGv_i32, TCGv_i32, TCGv_i32))
400 {
401 TCGv_i32 al = tcg_temp_new_i32();
402 TCGv_i32 ah = tcg_temp_new_i32();
403 TCGv_i32 bl = tcg_temp_new_i32();
404
405 tcg_gen_extr_i64_i32(al, ah, a64);
406 tcg_gen_extrl_i64_i32(bl, b64);
407 tcg_gen_andi_i32(bl, bl, 31);
408 generate(al, al, bl);
409 generate(ah, ah, bl);
410 tcg_gen_concat_i32_i64(d64, al, ah);
411
412 tcg_temp_free_i32(al);
413 tcg_temp_free_i32(ah);
414 tcg_temp_free_i32(bl);
415 }
416
417 static void gen_v4op(TCGv d64, TCGv a64, TCGv b64,
418 void (*generate)(TCGv_i32, TCGv_i32, TCGv_i32))
419 {
420 TCGv_i32 al = tcg_temp_new_i32();
421 TCGv_i32 ah = tcg_temp_new_i32();
422 TCGv_i32 bl = tcg_temp_new_i32();
423 TCGv_i32 bh = tcg_temp_new_i32();
424
425 tcg_gen_extr_i64_i32(al, ah, a64);
426 tcg_gen_extr_i64_i32(bl, bh, b64);
427 generate(al, al, bl);
428 generate(ah, ah, bh);
429 tcg_gen_concat_i32_i64(d64, al, ah);
430
431 tcg_temp_free_i32(al);
432 tcg_temp_free_i32(ah);
433 tcg_temp_free_i32(bl);
434 tcg_temp_free_i32(bh);
435 }
436
437 static TileExcp gen_signal(DisasContext *dc, int signo, int sigcode,
438 const char *mnemonic)
439 {
440 TCGv_i32 t0 = tcg_const_i32(signo);
441 TCGv_i32 t1 = tcg_const_i32(sigcode);
442
443 tcg_gen_st_i32(t0, cpu_env, offsetof(CPUTLGState, signo));
444 tcg_gen_st_i32(t1, cpu_env, offsetof(CPUTLGState, sigcode));
445
446 tcg_temp_free_i32(t1);
447 tcg_temp_free_i32(t0);
448
449 qemu_log_mask(CPU_LOG_TB_IN_ASM, "%s", mnemonic);
450 return TILEGX_EXCP_SIGNAL;
451 }
452
453 static bool parse_from_addli(uint64_t bundle, int *signo, int *sigcode)
454 {
455 int imm;
456
457 if ((get_Opcode_X0(bundle) != ADDLI_OPCODE_X0)
458 || (get_Dest_X0(bundle) != TILEGX_R_ZERO)
459 || (get_SrcA_X0(bundle) != TILEGX_R_ZERO)) {
460 return false;
461 }
462
463 imm = get_Imm16_X0(bundle);
464 *signo = imm & 0x3f;
465 *sigcode = (imm >> 6) & 0xf;
466
467 /* ??? The linux kernel validates both signo and the sigcode vs the
468 known max for each signal. Don't bother here. */
469 return true;
470 }
471
472 static TileExcp gen_specill(DisasContext *dc, unsigned dest, unsigned srca,
473 uint64_t bundle)
474 {
475 const char *mnemonic;
476 int signo;
477 int sigcode;
478
479 if (dest == 0x1c && srca == 0x25) {
480 signo = TARGET_SIGTRAP;
481 sigcode = TARGET_TRAP_BRKPT;
482 mnemonic = "bpt";
483 } else if (dest == 0x1d && srca == 0x25
484 && parse_from_addli(bundle, &signo, &sigcode)) {
485 mnemonic = "raise";
486 } else {
487 signo = TARGET_SIGILL;
488 sigcode = TARGET_ILL_ILLOPC;
489 mnemonic = "ill";
490 }
491
492 return gen_signal(dc, signo, sigcode, mnemonic);
493 }
494
495 static TileExcp gen_rr_opcode(DisasContext *dc, unsigned opext,
496 unsigned dest, unsigned srca, uint64_t bundle)
497 {
498 TCGv tdest, tsrca;
499 const char *mnemonic;
500 TCGMemOp memop;
501 TileExcp ret = TILEGX_EXCP_NONE;
502 bool prefetch_nofault = false;
503
504 /* Eliminate instructions with no output before doing anything else. */
505 switch (opext) {
506 case OE_RR_Y0(NOP):
507 case OE_RR_Y1(NOP):
508 case OE_RR_X0(NOP):
509 case OE_RR_X1(NOP):
510 mnemonic = "nop";
511 goto done0;
512 case OE_RR_Y0(FNOP):
513 case OE_RR_Y1(FNOP):
514 case OE_RR_X0(FNOP):
515 case OE_RR_X1(FNOP):
516 mnemonic = "fnop";
517 goto done0;
518 case OE_RR_X1(DRAIN):
519 mnemonic = "drain";
520 goto done0;
521 case OE_RR_X1(FLUSHWB):
522 mnemonic = "flushwb";
523 goto done0;
524 case OE_RR_X1(ILL):
525 return gen_specill(dc, dest, srca, bundle);
526 case OE_RR_Y1(ILL):
527 return gen_signal(dc, TARGET_SIGILL, TARGET_ILL_ILLOPC, "ill");
528 case OE_RR_X1(MF):
529 mnemonic = "mf";
530 goto done0;
531 case OE_RR_X1(NAP):
532 /* ??? This should yield, especially in system mode. */
533 mnemonic = "nap";
534 goto done0;
535 case OE_RR_X1(IRET):
536 gen_helper_ext01_ics(cpu_env);
537 dc->jmp.cond = TCG_COND_ALWAYS;
538 dc->jmp.dest = tcg_temp_new();
539 tcg_gen_ld_tl(dc->jmp.dest, cpu_env,
540 offsetof(CPUTLGState, spregs[TILEGX_SPR_EX_CONTEXT_0_0]));
541 tcg_gen_andi_tl(dc->jmp.dest, dc->jmp.dest, ~7);
542 mnemonic = "iret";
543 goto done0;
544 case OE_RR_X1(SWINT0):
545 case OE_RR_X1(SWINT2):
546 case OE_RR_X1(SWINT3):
547 return TILEGX_EXCP_OPCODE_UNIMPLEMENTED;
548 case OE_RR_X1(SWINT1):
549 ret = TILEGX_EXCP_SYSCALL;
550 mnemonic = "swint1";
551 done0:
552 if (srca || dest) {
553 return TILEGX_EXCP_OPCODE_UNKNOWN;
554 }
555 qemu_log_mask(CPU_LOG_TB_IN_ASM, "%s", mnemonic);
556 return ret;
557
558 case OE_RR_X1(DTLBPR):
559 return TILEGX_EXCP_OPCODE_UNIMPLEMENTED;
560 case OE_RR_X1(FINV):
561 mnemonic = "finv";
562 goto done1;
563 case OE_RR_X1(FLUSH):
564 mnemonic = "flush";
565 goto done1;
566 case OE_RR_X1(ICOH):
567 mnemonic = "icoh";
568 goto done1;
569 case OE_RR_X1(INV):
570 mnemonic = "inv";
571 goto done1;
572 case OE_RR_X1(WH64):
573 mnemonic = "wh64";
574 goto done1;
575 case OE_RR_X1(JRP):
576 case OE_RR_Y1(JRP):
577 mnemonic = "jrp";
578 goto do_jr;
579 case OE_RR_X1(JR):
580 case OE_RR_Y1(JR):
581 mnemonic = "jr";
582 goto do_jr;
583 case OE_RR_X1(JALRP):
584 case OE_RR_Y1(JALRP):
585 mnemonic = "jalrp";
586 goto do_jalr;
587 case OE_RR_X1(JALR):
588 case OE_RR_Y1(JALR):
589 mnemonic = "jalr";
590 do_jalr:
591 tcg_gen_movi_tl(dest_gr(dc, TILEGX_R_LR),
592 dc->pc + TILEGX_BUNDLE_SIZE_IN_BYTES);
593 do_jr:
594 dc->jmp.cond = TCG_COND_ALWAYS;
595 dc->jmp.dest = tcg_temp_new();
596 tcg_gen_andi_tl(dc->jmp.dest, load_gr(dc, srca), ~7);
597 done1:
598 if (dest) {
599 return TILEGX_EXCP_OPCODE_UNKNOWN;
600 }
601 qemu_log_mask(CPU_LOG_TB_IN_ASM, "%s %s", mnemonic, reg_names[srca]);
602 return ret;
603 }
604
605 tdest = dest_gr(dc, dest);
606 tsrca = load_gr(dc, srca);
607
608 switch (opext) {
609 case OE_RR_X0(CNTLZ):
610 case OE_RR_Y0(CNTLZ):
611 gen_helper_cntlz(tdest, tsrca);
612 mnemonic = "cntlz";
613 break;
614 case OE_RR_X0(CNTTZ):
615 case OE_RR_Y0(CNTTZ):
616 gen_helper_cnttz(tdest, tsrca);
617 mnemonic = "cnttz";
618 break;
619 case OE_RR_X0(FSINGLE_PACK1):
620 case OE_RR_Y0(FSINGLE_PACK1):
621 return TILEGX_EXCP_OPCODE_UNIMPLEMENTED;
622 case OE_RR_X1(LD1S):
623 memop = MO_SB;
624 mnemonic = "ld1s"; /* prefetch_l1_fault */
625 goto do_load;
626 case OE_RR_X1(LD1U):
627 memop = MO_UB;
628 mnemonic = "ld1u"; /* prefetch, prefetch_l1 */
629 prefetch_nofault = (dest == TILEGX_R_ZERO);
630 goto do_load;
631 case OE_RR_X1(LD2S):
632 memop = MO_TESW;
633 mnemonic = "ld2s"; /* prefetch_l2_fault */
634 goto do_load;
635 case OE_RR_X1(LD2U):
636 memop = MO_TEUW;
637 mnemonic = "ld2u"; /* prefetch_l2 */
638 prefetch_nofault = (dest == TILEGX_R_ZERO);
639 goto do_load;
640 case OE_RR_X1(LD4S):
641 memop = MO_TESL;
642 mnemonic = "ld4s"; /* prefetch_l3_fault */
643 goto do_load;
644 case OE_RR_X1(LD4U):
645 memop = MO_TEUL;
646 mnemonic = "ld4u"; /* prefetch_l3 */
647 prefetch_nofault = (dest == TILEGX_R_ZERO);
648 goto do_load;
649 case OE_RR_X1(LDNT1S):
650 memop = MO_SB;
651 mnemonic = "ldnt1s";
652 goto do_load;
653 case OE_RR_X1(LDNT1U):
654 memop = MO_UB;
655 mnemonic = "ldnt1u";
656 goto do_load;
657 case OE_RR_X1(LDNT2S):
658 memop = MO_TESW;
659 mnemonic = "ldnt2s";
660 goto do_load;
661 case OE_RR_X1(LDNT2U):
662 memop = MO_TEUW;
663 mnemonic = "ldnt2u";
664 goto do_load;
665 case OE_RR_X1(LDNT4S):
666 memop = MO_TESL;
667 mnemonic = "ldnt4s";
668 goto do_load;
669 case OE_RR_X1(LDNT4U):
670 memop = MO_TEUL;
671 mnemonic = "ldnt4u";
672 goto do_load;
673 case OE_RR_X1(LDNT):
674 memop = MO_TEQ;
675 mnemonic = "ldnt";
676 goto do_load;
677 case OE_RR_X1(LD):
678 memop = MO_TEQ;
679 mnemonic = "ld";
680 do_load:
681 if (!prefetch_nofault) {
682 tcg_gen_qemu_ld_tl(tdest, tsrca, dc->mmuidx, memop);
683 }
684 break;
685 case OE_RR_X1(LDNA):
686 tcg_gen_andi_tl(tdest, tsrca, ~7);
687 tcg_gen_qemu_ld_tl(tdest, tdest, dc->mmuidx, MO_TEQ);
688 mnemonic = "ldna";
689 break;
690 case OE_RR_X1(LNK):
691 case OE_RR_Y1(LNK):
692 if (srca) {
693 return TILEGX_EXCP_OPCODE_UNKNOWN;
694 }
695 tcg_gen_movi_tl(tdest, dc->pc + TILEGX_BUNDLE_SIZE_IN_BYTES);
696 mnemonic = "lnk";
697 break;
698 case OE_RR_X0(PCNT):
699 case OE_RR_Y0(PCNT):
700 gen_helper_pcnt(tdest, tsrca);
701 mnemonic = "pcnt";
702 break;
703 case OE_RR_X0(REVBITS):
704 case OE_RR_Y0(REVBITS):
705 gen_helper_revbits(tdest, tsrca);
706 mnemonic = "revbits";
707 break;
708 case OE_RR_X0(REVBYTES):
709 case OE_RR_Y0(REVBYTES):
710 tcg_gen_bswap64_tl(tdest, tsrca);
711 mnemonic = "revbytes";
712 break;
713 case OE_RR_X0(TBLIDXB0):
714 case OE_RR_Y0(TBLIDXB0):
715 tcg_gen_deposit_tl(tdest, load_gr(dc, dest), tsrca, 2, 8);
716 mnemonic = "tblidxb0";
717 break;
718 case OE_RR_X0(TBLIDXB1):
719 case OE_RR_Y0(TBLIDXB1):
720 tcg_gen_shri_tl(tdest, tsrca, 8);
721 tcg_gen_deposit_tl(tdest, load_gr(dc, dest), tdest, 2, 8);
722 mnemonic = "tblidxb1";
723 break;
724 case OE_RR_X0(TBLIDXB2):
725 case OE_RR_Y0(TBLIDXB2):
726 tcg_gen_shri_tl(tdest, tsrca, 16);
727 tcg_gen_deposit_tl(tdest, load_gr(dc, dest), tdest, 2, 8);
728 mnemonic = "tblidxb2";
729 break;
730 case OE_RR_X0(TBLIDXB3):
731 case OE_RR_Y0(TBLIDXB3):
732 tcg_gen_shri_tl(tdest, tsrca, 24);
733 tcg_gen_deposit_tl(tdest, load_gr(dc, dest), tdest, 2, 8);
734 mnemonic = "tblidxb3";
735 break;
736 default:
737 return TILEGX_EXCP_OPCODE_UNKNOWN;
738 }
739
740 qemu_log_mask(CPU_LOG_TB_IN_ASM, "%s %s, %s", mnemonic,
741 reg_names[dest], reg_names[srca]);
742 return ret;
743 }
744
745 static TileExcp gen_rrr_opcode(DisasContext *dc, unsigned opext,
746 unsigned dest, unsigned srca, unsigned srcb)
747 {
748 TCGv tdest = dest_gr(dc, dest);
749 TCGv tsrca = load_gr(dc, srca);
750 TCGv tsrcb = load_gr(dc, srcb);
751 TCGv t0;
752 const char *mnemonic;
753
754 switch (opext) {
755 case OE_RRR(ADDXSC, 0, X0):
756 case OE_RRR(ADDXSC, 0, X1):
757 gen_saturate_op(tdest, tsrca, tsrcb, tcg_gen_add_tl);
758 mnemonic = "addxsc";
759 break;
760 case OE_RRR(ADDX, 0, X0):
761 case OE_RRR(ADDX, 0, X1):
762 case OE_RRR(ADDX, 0, Y0):
763 case OE_RRR(ADDX, 0, Y1):
764 tcg_gen_add_tl(tdest, tsrca, tsrcb);
765 tcg_gen_ext32s_tl(tdest, tdest);
766 mnemonic = "addx";
767 break;
768 case OE_RRR(ADD, 0, X0):
769 case OE_RRR(ADD, 0, X1):
770 case OE_RRR(ADD, 0, Y0):
771 case OE_RRR(ADD, 0, Y1):
772 tcg_gen_add_tl(tdest, tsrca, tsrcb);
773 mnemonic = "add";
774 break;
775 case OE_RRR(AND, 0, X0):
776 case OE_RRR(AND, 0, X1):
777 case OE_RRR(AND, 5, Y0):
778 case OE_RRR(AND, 5, Y1):
779 tcg_gen_and_tl(tdest, tsrca, tsrcb);
780 mnemonic = "and";
781 break;
782 case OE_RRR(CMOVEQZ, 0, X0):
783 case OE_RRR(CMOVEQZ, 4, Y0):
784 tcg_gen_movcond_tl(TCG_COND_EQ, tdest, tsrca, load_zero(dc),
785 tsrcb, load_gr(dc, dest));
786 mnemonic = "cmoveqz";
787 break;
788 case OE_RRR(CMOVNEZ, 0, X0):
789 case OE_RRR(CMOVNEZ, 4, Y0):
790 tcg_gen_movcond_tl(TCG_COND_NE, tdest, tsrca, load_zero(dc),
791 tsrcb, load_gr(dc, dest));
792 mnemonic = "cmovnez";
793 break;
794 case OE_RRR(CMPEQ, 0, X0):
795 case OE_RRR(CMPEQ, 0, X1):
796 case OE_RRR(CMPEQ, 3, Y0):
797 case OE_RRR(CMPEQ, 3, Y1):
798 tcg_gen_setcond_tl(TCG_COND_EQ, tdest, tsrca, tsrcb);
799 mnemonic = "cmpeq";
800 break;
801 case OE_RRR(CMPEXCH4, 0, X1):
802 gen_atomic_excp(dc, dest, tdest, tsrca, tsrcb,
803 TILEGX_EXCP_OPCODE_CMPEXCH4);
804 mnemonic = "cmpexch4";
805 break;
806 case OE_RRR(CMPEXCH, 0, X1):
807 gen_atomic_excp(dc, dest, tdest, tsrca, tsrcb,
808 TILEGX_EXCP_OPCODE_CMPEXCH);
809 mnemonic = "cmpexch";
810 break;
811 case OE_RRR(CMPLES, 0, X0):
812 case OE_RRR(CMPLES, 0, X1):
813 case OE_RRR(CMPLES, 2, Y0):
814 case OE_RRR(CMPLES, 2, Y1):
815 tcg_gen_setcond_tl(TCG_COND_LE, tdest, tsrca, tsrcb);
816 mnemonic = "cmples";
817 break;
818 case OE_RRR(CMPLEU, 0, X0):
819 case OE_RRR(CMPLEU, 0, X1):
820 case OE_RRR(CMPLEU, 2, Y0):
821 case OE_RRR(CMPLEU, 2, Y1):
822 tcg_gen_setcond_tl(TCG_COND_LEU, tdest, tsrca, tsrcb);
823 mnemonic = "cmpleu";
824 break;
825 case OE_RRR(CMPLTS, 0, X0):
826 case OE_RRR(CMPLTS, 0, X1):
827 case OE_RRR(CMPLTS, 2, Y0):
828 case OE_RRR(CMPLTS, 2, Y1):
829 tcg_gen_setcond_tl(TCG_COND_LT, tdest, tsrca, tsrcb);
830 mnemonic = "cmplts";
831 break;
832 case OE_RRR(CMPLTU, 0, X0):
833 case OE_RRR(CMPLTU, 0, X1):
834 case OE_RRR(CMPLTU, 2, Y0):
835 case OE_RRR(CMPLTU, 2, Y1):
836 tcg_gen_setcond_tl(TCG_COND_LTU, tdest, tsrca, tsrcb);
837 mnemonic = "cmpltu";
838 break;
839 case OE_RRR(CMPNE, 0, X0):
840 case OE_RRR(CMPNE, 0, X1):
841 case OE_RRR(CMPNE, 3, Y0):
842 case OE_RRR(CMPNE, 3, Y1):
843 tcg_gen_setcond_tl(TCG_COND_NE, tdest, tsrca, tsrcb);
844 mnemonic = "cmpne";
845 break;
846 case OE_RRR(CMULAF, 0, X0):
847 gen_helper_cmulaf(tdest, load_gr(dc, dest), tsrca, tsrcb);
848 mnemonic = "cmulaf";
849 break;
850 case OE_RRR(CMULA, 0, X0):
851 gen_helper_cmula(tdest, load_gr(dc, dest), tsrca, tsrcb);
852 mnemonic = "cmula";
853 break;
854 case OE_RRR(CMULFR, 0, X0):
855 gen_cmul2(tdest, tsrca, tsrcb, 15, 1 << 14);
856 mnemonic = "cmulfr";
857 break;
858 case OE_RRR(CMULF, 0, X0):
859 gen_cmul2(tdest, tsrca, tsrcb, 15, 0);
860 mnemonic = "cmulf";
861 break;
862 case OE_RRR(CMULHR, 0, X0):
863 gen_cmul2(tdest, tsrca, tsrcb, 16, 1 << 15);
864 mnemonic = "cmulhr";
865 break;
866 case OE_RRR(CMULH, 0, X0):
867 gen_cmul2(tdest, tsrca, tsrcb, 16, 0);
868 mnemonic = "cmulh";
869 break;
870 case OE_RRR(CMUL, 0, X0):
871 gen_helper_cmula(tdest, load_zero(dc), tsrca, tsrcb);
872 mnemonic = "cmul";
873 break;
874 case OE_RRR(CRC32_32, 0, X0):
875 gen_helper_crc32_32(tdest, tsrca, tsrcb);
876 mnemonic = "crc32_32";
877 break;
878 case OE_RRR(CRC32_8, 0, X0):
879 gen_helper_crc32_8(tdest, tsrca, tsrcb);
880 mnemonic = "crc32_8";
881 break;
882 case OE_RRR(DBLALIGN2, 0, X0):
883 case OE_RRR(DBLALIGN2, 0, X1):
884 gen_dblaligni(tdest, tsrca, tsrcb, 16);
885 mnemonic = "dblalign2";
886 break;
887 case OE_RRR(DBLALIGN4, 0, X0):
888 case OE_RRR(DBLALIGN4, 0, X1):
889 gen_dblaligni(tdest, tsrca, tsrcb, 32);
890 mnemonic = "dblalign4";
891 break;
892 case OE_RRR(DBLALIGN6, 0, X0):
893 case OE_RRR(DBLALIGN6, 0, X1):
894 gen_dblaligni(tdest, tsrca, tsrcb, 48);
895 mnemonic = "dblalign6";
896 break;
897 case OE_RRR(DBLALIGN, 0, X0):
898 gen_dblalign(tdest, load_gr(dc, dest), tsrca, tsrcb);
899 mnemonic = "dblalign";
900 break;
901 case OE_RRR(EXCH4, 0, X1):
902 gen_atomic_excp(dc, dest, tdest, tsrca, tsrcb,
903 TILEGX_EXCP_OPCODE_EXCH4);
904 mnemonic = "exch4";
905 break;
906 case OE_RRR(EXCH, 0, X1):
907 gen_atomic_excp(dc, dest, tdest, tsrca, tsrcb,
908 TILEGX_EXCP_OPCODE_EXCH);
909 mnemonic = "exch";
910 break;
911 case OE_RRR(FDOUBLE_ADDSUB, 0, X0):
912 case OE_RRR(FDOUBLE_ADD_FLAGS, 0, X0):
913 case OE_RRR(FDOUBLE_MUL_FLAGS, 0, X0):
914 case OE_RRR(FDOUBLE_PACK1, 0, X0):
915 case OE_RRR(FDOUBLE_PACK2, 0, X0):
916 case OE_RRR(FDOUBLE_SUB_FLAGS, 0, X0):
917 case OE_RRR(FDOUBLE_UNPACK_MAX, 0, X0):
918 case OE_RRR(FDOUBLE_UNPACK_MIN, 0, X0):
919 return TILEGX_EXCP_OPCODE_UNIMPLEMENTED;
920 case OE_RRR(FETCHADD4, 0, X1):
921 gen_atomic_excp(dc, dest, tdest, tsrca, tsrcb,
922 TILEGX_EXCP_OPCODE_FETCHADD4);
923 mnemonic = "fetchadd4";
924 break;
925 case OE_RRR(FETCHADDGEZ4, 0, X1):
926 gen_atomic_excp(dc, dest, tdest, tsrca, tsrcb,
927 TILEGX_EXCP_OPCODE_FETCHADDGEZ4);
928 mnemonic = "fetchaddgez4";
929 break;
930 case OE_RRR(FETCHADDGEZ, 0, X1):
931 gen_atomic_excp(dc, dest, tdest, tsrca, tsrcb,
932 TILEGX_EXCP_OPCODE_FETCHADDGEZ);
933 mnemonic = "fetchaddgez";
934 break;
935 case OE_RRR(FETCHADD, 0, X1):
936 gen_atomic_excp(dc, dest, tdest, tsrca, tsrcb,
937 TILEGX_EXCP_OPCODE_FETCHADD);
938 mnemonic = "fetchadd";
939 break;
940 case OE_RRR(FETCHAND4, 0, X1):
941 gen_atomic_excp(dc, dest, tdest, tsrca, tsrcb,
942 TILEGX_EXCP_OPCODE_FETCHAND4);
943 mnemonic = "fetchand4";
944 break;
945 case OE_RRR(FETCHAND, 0, X1):
946 gen_atomic_excp(dc, dest, tdest, tsrca, tsrcb,
947 TILEGX_EXCP_OPCODE_FETCHAND);
948 mnemonic = "fetchand";
949 break;
950 case OE_RRR(FETCHOR4, 0, X1):
951 gen_atomic_excp(dc, dest, tdest, tsrca, tsrcb,
952 TILEGX_EXCP_OPCODE_FETCHOR4);
953 mnemonic = "fetchor4";
954 break;
955 case OE_RRR(FETCHOR, 0, X1):
956 gen_atomic_excp(dc, dest, tdest, tsrca, tsrcb,
957 TILEGX_EXCP_OPCODE_FETCHOR);
958 mnemonic = "fetchor";
959 break;
960 case OE_RRR(FSINGLE_ADD1, 0, X0):
961 case OE_RRR(FSINGLE_ADDSUB2, 0, X0):
962 case OE_RRR(FSINGLE_MUL1, 0, X0):
963 case OE_RRR(FSINGLE_MUL2, 0, X0):
964 case OE_RRR(FSINGLE_PACK2, 0, X0):
965 case OE_RRR(FSINGLE_SUB1, 0, X0):
966 return TILEGX_EXCP_OPCODE_UNIMPLEMENTED;
967 case OE_RRR(MNZ, 0, X0):
968 case OE_RRR(MNZ, 0, X1):
969 case OE_RRR(MNZ, 4, Y0):
970 case OE_RRR(MNZ, 4, Y1):
971 t0 = load_zero(dc);
972 tcg_gen_movcond_tl(TCG_COND_NE, tdest, tsrca, t0, tsrcb, t0);
973 mnemonic = "mnz";
974 break;
975 case OE_RRR(MULAX, 0, X0):
976 case OE_RRR(MULAX, 3, Y0):
977 tcg_gen_mul_tl(tdest, tsrca, tsrcb);
978 tcg_gen_add_tl(tdest, tdest, load_gr(dc, dest));
979 tcg_gen_ext32s_tl(tdest, tdest);
980 mnemonic = "mulax";
981 break;
982 case OE_RRR(MULA_HS_HS, 0, X0):
983 case OE_RRR(MULA_HS_HS, 9, Y0):
984 gen_mul_half(tdest, tsrca, tsrcb, HS, HS);
985 tcg_gen_add_tl(tdest, tdest, load_gr(dc, dest));
986 mnemonic = "mula_hs_hs";
987 break;
988 case OE_RRR(MULA_HS_HU, 0, X0):
989 gen_mul_half(tdest, tsrca, tsrcb, HS, HU);
990 tcg_gen_add_tl(tdest, tdest, load_gr(dc, dest));
991 mnemonic = "mula_hs_hu";
992 break;
993 case OE_RRR(MULA_HS_LS, 0, X0):
994 gen_mul_half(tdest, tsrca, tsrcb, HS, LS);
995 tcg_gen_add_tl(tdest, tdest, load_gr(dc, dest));
996 mnemonic = "mula_hs_ls";
997 break;
998 case OE_RRR(MULA_HS_LU, 0, X0):
999 gen_mul_half(tdest, tsrca, tsrcb, HS, LU);
1000 tcg_gen_add_tl(tdest, tdest, load_gr(dc, dest));
1001 mnemonic = "mula_hs_lu";
1002 break;
1003 case OE_RRR(MULA_HU_HU, 0, X0):
1004 case OE_RRR(MULA_HU_HU, 9, Y0):
1005 gen_mul_half(tdest, tsrca, tsrcb, HU, HU);
1006 tcg_gen_add_tl(tdest, tdest, load_gr(dc, dest));
1007 mnemonic = "mula_hu_hu";
1008 break;
1009 case OE_RRR(MULA_HU_LS, 0, X0):
1010 gen_mul_half(tdest, tsrca, tsrcb, HU, LS);
1011 tcg_gen_add_tl(tdest, tdest, load_gr(dc, dest));
1012 mnemonic = "mula_hu_ls";
1013 break;
1014 case OE_RRR(MULA_HU_LU, 0, X0):
1015 gen_mul_half(tdest, tsrca, tsrcb, HU, LU);
1016 tcg_gen_add_tl(tdest, tdest, load_gr(dc, dest));
1017 mnemonic = "mula_hu_lu";
1018 break;
1019 case OE_RRR(MULA_LS_LS, 0, X0):
1020 case OE_RRR(MULA_LS_LS, 9, Y0):
1021 gen_mul_half(tdest, tsrca, tsrcb, LS, LS);
1022 tcg_gen_add_tl(tdest, tdest, load_gr(dc, dest));
1023 mnemonic = "mula_ls_ls";
1024 break;
1025 case OE_RRR(MULA_LS_LU, 0, X0):
1026 gen_mul_half(tdest, tsrca, tsrcb, LS, LU);
1027 tcg_gen_add_tl(tdest, tdest, load_gr(dc, dest));
1028 mnemonic = "mula_ls_lu";
1029 break;
1030 case OE_RRR(MULA_LU_LU, 0, X0):
1031 case OE_RRR(MULA_LU_LU, 9, Y0):
1032 gen_mul_half(tdest, tsrca, tsrcb, LU, LU);
1033 tcg_gen_add_tl(tdest, tdest, load_gr(dc, dest));
1034 mnemonic = "mula_lu_lu";
1035 break;
1036 case OE_RRR(MULX, 0, X0):
1037 case OE_RRR(MULX, 3, Y0):
1038 tcg_gen_mul_tl(tdest, tsrca, tsrcb);
1039 tcg_gen_ext32s_tl(tdest, tdest);
1040 mnemonic = "mulx";
1041 break;
1042 case OE_RRR(MUL_HS_HS, 0, X0):
1043 case OE_RRR(MUL_HS_HS, 8, Y0):
1044 gen_mul_half(tdest, tsrca, tsrcb, HS, HS);
1045 mnemonic = "mul_hs_hs";
1046 break;
1047 case OE_RRR(MUL_HS_HU, 0, X0):
1048 gen_mul_half(tdest, tsrca, tsrcb, HS, HU);
1049 mnemonic = "mul_hs_hu";
1050 break;
1051 case OE_RRR(MUL_HS_LS, 0, X0):
1052 gen_mul_half(tdest, tsrca, tsrcb, HS, LS);
1053 mnemonic = "mul_hs_ls";
1054 break;
1055 case OE_RRR(MUL_HS_LU, 0, X0):
1056 gen_mul_half(tdest, tsrca, tsrcb, HS, LU);
1057 mnemonic = "mul_hs_lu";
1058 break;
1059 case OE_RRR(MUL_HU_HU, 0, X0):
1060 case OE_RRR(MUL_HU_HU, 8, Y0):
1061 gen_mul_half(tdest, tsrca, tsrcb, HU, HU);
1062 mnemonic = "mul_hu_hu";
1063 break;
1064 case OE_RRR(MUL_HU_LS, 0, X0):
1065 gen_mul_half(tdest, tsrca, tsrcb, HU, LS);
1066 mnemonic = "mul_hu_ls";
1067 break;
1068 case OE_RRR(MUL_HU_LU, 0, X0):
1069 gen_mul_half(tdest, tsrca, tsrcb, HU, LU);
1070 mnemonic = "mul_hu_lu";
1071 break;
1072 case OE_RRR(MUL_LS_LS, 0, X0):
1073 case OE_RRR(MUL_LS_LS, 8, Y0):
1074 gen_mul_half(tdest, tsrca, tsrcb, LS, LS);
1075 mnemonic = "mul_ls_ls";
1076 break;
1077 case OE_RRR(MUL_LS_LU, 0, X0):
1078 gen_mul_half(tdest, tsrca, tsrcb, LS, LU);
1079 mnemonic = "mul_ls_lu";
1080 break;
1081 case OE_RRR(MUL_LU_LU, 0, X0):
1082 case OE_RRR(MUL_LU_LU, 8, Y0):
1083 gen_mul_half(tdest, tsrca, tsrcb, LU, LU);
1084 mnemonic = "mul_lu_lu";
1085 break;
1086 case OE_RRR(MZ, 0, X0):
1087 case OE_RRR(MZ, 0, X1):
1088 case OE_RRR(MZ, 4, Y0):
1089 case OE_RRR(MZ, 4, Y1):
1090 t0 = load_zero(dc);
1091 tcg_gen_movcond_tl(TCG_COND_EQ, tdest, tsrca, t0, tsrcb, t0);
1092 mnemonic = "mz";
1093 break;
1094 case OE_RRR(NOR, 0, X0):
1095 case OE_RRR(NOR, 0, X1):
1096 case OE_RRR(NOR, 5, Y0):
1097 case OE_RRR(NOR, 5, Y1):
1098 tcg_gen_nor_tl(tdest, tsrca, tsrcb);
1099 mnemonic = "nor";
1100 break;
1101 case OE_RRR(OR, 0, X0):
1102 case OE_RRR(OR, 0, X1):
1103 case OE_RRR(OR, 5, Y0):
1104 case OE_RRR(OR, 5, Y1):
1105 tcg_gen_or_tl(tdest, tsrca, tsrcb);
1106 mnemonic = "or";
1107 break;
1108 case OE_RRR(ROTL, 0, X0):
1109 case OE_RRR(ROTL, 0, X1):
1110 case OE_RRR(ROTL, 6, Y0):
1111 case OE_RRR(ROTL, 6, Y1):
1112 tcg_gen_andi_tl(tdest, tsrcb, 63);
1113 tcg_gen_rotl_tl(tdest, tsrca, tdest);
1114 mnemonic = "rotl";
1115 break;
1116 case OE_RRR(SHL1ADDX, 0, X0):
1117 case OE_RRR(SHL1ADDX, 0, X1):
1118 case OE_RRR(SHL1ADDX, 7, Y0):
1119 case OE_RRR(SHL1ADDX, 7, Y1):
1120 tcg_gen_shli_tl(tdest, tsrca, 1);
1121 tcg_gen_add_tl(tdest, tdest, tsrcb);
1122 tcg_gen_ext32s_tl(tdest, tdest);
1123 mnemonic = "shl1addx";
1124 break;
1125 case OE_RRR(SHL1ADD, 0, X0):
1126 case OE_RRR(SHL1ADD, 0, X1):
1127 case OE_RRR(SHL1ADD, 1, Y0):
1128 case OE_RRR(SHL1ADD, 1, Y1):
1129 tcg_gen_shli_tl(tdest, tsrca, 1);
1130 tcg_gen_add_tl(tdest, tdest, tsrcb);
1131 mnemonic = "shl1add";
1132 break;
1133 case OE_RRR(SHL2ADDX, 0, X0):
1134 case OE_RRR(SHL2ADDX, 0, X1):
1135 case OE_RRR(SHL2ADDX, 7, Y0):
1136 case OE_RRR(SHL2ADDX, 7, Y1):
1137 tcg_gen_shli_tl(tdest, tsrca, 2);
1138 tcg_gen_add_tl(tdest, tdest, tsrcb);
1139 tcg_gen_ext32s_tl(tdest, tdest);
1140 mnemonic = "shl2addx";
1141 break;
1142 case OE_RRR(SHL2ADD, 0, X0):
1143 case OE_RRR(SHL2ADD, 0, X1):
1144 case OE_RRR(SHL2ADD, 1, Y0):
1145 case OE_RRR(SHL2ADD, 1, Y1):
1146 tcg_gen_shli_tl(tdest, tsrca, 2);
1147 tcg_gen_add_tl(tdest, tdest, tsrcb);
1148 mnemonic = "shl2add";
1149 break;
1150 case OE_RRR(SHL3ADDX, 0, X0):
1151 case OE_RRR(SHL3ADDX, 0, X1):
1152 case OE_RRR(SHL3ADDX, 7, Y0):
1153 case OE_RRR(SHL3ADDX, 7, Y1):
1154 tcg_gen_shli_tl(tdest, tsrca, 3);
1155 tcg_gen_add_tl(tdest, tdest, tsrcb);
1156 tcg_gen_ext32s_tl(tdest, tdest);
1157 mnemonic = "shl3addx";
1158 break;
1159 case OE_RRR(SHL3ADD, 0, X0):
1160 case OE_RRR(SHL3ADD, 0, X1):
1161 case OE_RRR(SHL3ADD, 1, Y0):
1162 case OE_RRR(SHL3ADD, 1, Y1):
1163 tcg_gen_shli_tl(tdest, tsrca, 3);
1164 tcg_gen_add_tl(tdest, tdest, tsrcb);
1165 mnemonic = "shl3add";
1166 break;
1167 case OE_RRR(SHLX, 0, X0):
1168 case OE_RRR(SHLX, 0, X1):
1169 tcg_gen_andi_tl(tdest, tsrcb, 31);
1170 tcg_gen_shl_tl(tdest, tsrca, tdest);
1171 tcg_gen_ext32s_tl(tdest, tdest);
1172 mnemonic = "shlx";
1173 break;
1174 case OE_RRR(SHL, 0, X0):
1175 case OE_RRR(SHL, 0, X1):
1176 case OE_RRR(SHL, 6, Y0):
1177 case OE_RRR(SHL, 6, Y1):
1178 tcg_gen_andi_tl(tdest, tsrcb, 63);
1179 tcg_gen_shl_tl(tdest, tsrca, tdest);
1180 mnemonic = "shl";
1181 break;
1182 case OE_RRR(SHRS, 0, X0):
1183 case OE_RRR(SHRS, 0, X1):
1184 case OE_RRR(SHRS, 6, Y0):
1185 case OE_RRR(SHRS, 6, Y1):
1186 tcg_gen_andi_tl(tdest, tsrcb, 63);
1187 tcg_gen_sar_tl(tdest, tsrca, tdest);
1188 mnemonic = "shrs";
1189 break;
1190 case OE_RRR(SHRUX, 0, X0):
1191 case OE_RRR(SHRUX, 0, X1):
1192 t0 = tcg_temp_new();
1193 tcg_gen_andi_tl(t0, tsrcb, 31);
1194 tcg_gen_ext32u_tl(tdest, tsrca);
1195 tcg_gen_shr_tl(tdest, tdest, t0);
1196 tcg_gen_ext32s_tl(tdest, tdest);
1197 tcg_temp_free(t0);
1198 mnemonic = "shrux";
1199 break;
1200 case OE_RRR(SHRU, 0, X0):
1201 case OE_RRR(SHRU, 0, X1):
1202 case OE_RRR(SHRU, 6, Y0):
1203 case OE_RRR(SHRU, 6, Y1):
1204 tcg_gen_andi_tl(tdest, tsrcb, 63);
1205 tcg_gen_shr_tl(tdest, tsrca, tdest);
1206 mnemonic = "shru";
1207 break;
1208 case OE_RRR(SHUFFLEBYTES, 0, X0):
1209 gen_helper_shufflebytes(tdest, load_gr(dc, dest), tsrca, tsrca);
1210 mnemonic = "shufflebytes";
1211 break;
1212 case OE_RRR(SUBXSC, 0, X0):
1213 case OE_RRR(SUBXSC, 0, X1):
1214 gen_saturate_op(tdest, tsrca, tsrcb, tcg_gen_sub_tl);
1215 mnemonic = "subxsc";
1216 break;
1217 case OE_RRR(SUBX, 0, X0):
1218 case OE_RRR(SUBX, 0, X1):
1219 case OE_RRR(SUBX, 0, Y0):
1220 case OE_RRR(SUBX, 0, Y1):
1221 tcg_gen_sub_tl(tdest, tsrca, tsrcb);
1222 tcg_gen_ext32s_tl(tdest, tdest);
1223 mnemonic = "subx";
1224 break;
1225 case OE_RRR(SUB, 0, X0):
1226 case OE_RRR(SUB, 0, X1):
1227 case OE_RRR(SUB, 0, Y0):
1228 case OE_RRR(SUB, 0, Y1):
1229 tcg_gen_sub_tl(tdest, tsrca, tsrcb);
1230 mnemonic = "sub";
1231 break;
1232 case OE_RRR(V1ADDUC, 0, X0):
1233 case OE_RRR(V1ADDUC, 0, X1):
1234 return TILEGX_EXCP_OPCODE_UNIMPLEMENTED;
1235 case OE_RRR(V1ADD, 0, X0):
1236 case OE_RRR(V1ADD, 0, X1):
1237 gen_v12add(tdest, tsrca, tsrcb, V1_IMM(0x80));
1238 mnemonic = "v1add";
1239 break;
1240 case OE_RRR(V1ADIFFU, 0, X0):
1241 case OE_RRR(V1AVGU, 0, X0):
1242 return TILEGX_EXCP_OPCODE_UNIMPLEMENTED;
1243 case OE_RRR(V1CMPEQ, 0, X0):
1244 case OE_RRR(V1CMPEQ, 0, X1):
1245 tcg_gen_xor_tl(tdest, tsrca, tsrcb);
1246 gen_v1cmpeq0(tdest);
1247 mnemonic = "v1cmpeq";
1248 break;
1249 case OE_RRR(V1CMPLES, 0, X0):
1250 case OE_RRR(V1CMPLES, 0, X1):
1251 case OE_RRR(V1CMPLEU, 0, X0):
1252 case OE_RRR(V1CMPLEU, 0, X1):
1253 case OE_RRR(V1CMPLTS, 0, X0):
1254 case OE_RRR(V1CMPLTS, 0, X1):
1255 case OE_RRR(V1CMPLTU, 0, X0):
1256 case OE_RRR(V1CMPLTU, 0, X1):
1257 return TILEGX_EXCP_OPCODE_UNIMPLEMENTED;
1258 case OE_RRR(V1CMPNE, 0, X0):
1259 case OE_RRR(V1CMPNE, 0, X1):
1260 tcg_gen_xor_tl(tdest, tsrca, tsrcb);
1261 gen_v1cmpne0(tdest);
1262 mnemonic = "v1cmpne";
1263 break;
1264 case OE_RRR(V1DDOTPUA, 0, X0):
1265 case OE_RRR(V1DDOTPUSA, 0, X0):
1266 case OE_RRR(V1DDOTPUS, 0, X0):
1267 case OE_RRR(V1DDOTPU, 0, X0):
1268 case OE_RRR(V1DOTPA, 0, X0):
1269 case OE_RRR(V1DOTPUA, 0, X0):
1270 case OE_RRR(V1DOTPUSA, 0, X0):
1271 case OE_RRR(V1DOTPUS, 0, X0):
1272 case OE_RRR(V1DOTPU, 0, X0):
1273 case OE_RRR(V1DOTP, 0, X0):
1274 return TILEGX_EXCP_OPCODE_UNIMPLEMENTED;
1275 case OE_RRR(V1INT_H, 0, X0):
1276 case OE_RRR(V1INT_H, 0, X1):
1277 gen_helper_v1int_h(tdest, tsrca, tsrcb);
1278 mnemonic = "v1int_h";
1279 break;
1280 case OE_RRR(V1INT_L, 0, X0):
1281 case OE_RRR(V1INT_L, 0, X1):
1282 gen_helper_v1int_l(tdest, tsrca, tsrcb);
1283 mnemonic = "v1int_l";
1284 break;
1285 case OE_RRR(V1MAXU, 0, X0):
1286 case OE_RRR(V1MAXU, 0, X1):
1287 case OE_RRR(V1MINU, 0, X0):
1288 case OE_RRR(V1MINU, 0, X1):
1289 case OE_RRR(V1MNZ, 0, X0):
1290 case OE_RRR(V1MNZ, 0, X1):
1291 return TILEGX_EXCP_OPCODE_UNIMPLEMENTED;
1292 case OE_RRR(V1MULTU, 0, X0):
1293 gen_helper_v1multu(tdest, tsrca, tsrcb);
1294 mnemonic = "v1multu";
1295 break;
1296 case OE_RRR(V1MULUS, 0, X0):
1297 case OE_RRR(V1MULU, 0, X0):
1298 case OE_RRR(V1MZ, 0, X0):
1299 case OE_RRR(V1MZ, 0, X1):
1300 case OE_RRR(V1SADAU, 0, X0):
1301 case OE_RRR(V1SADU, 0, X0):
1302 return TILEGX_EXCP_OPCODE_UNIMPLEMENTED;
1303 case OE_RRR(V1SHL, 0, X0):
1304 case OE_RRR(V1SHL, 0, X1):
1305 gen_helper_v1shl(tdest, tsrca, tsrcb);
1306 mnemonic = "v1shl";
1307 break;
1308 case OE_RRR(V1SHRS, 0, X0):
1309 case OE_RRR(V1SHRS, 0, X1):
1310 gen_helper_v1shrs(tdest, tsrca, tsrcb);
1311 mnemonic = "v1shrs";
1312 break;
1313 case OE_RRR(V1SHRU, 0, X0):
1314 case OE_RRR(V1SHRU, 0, X1):
1315 gen_helper_v1shru(tdest, tsrca, tsrcb);
1316 mnemonic = "v1shru";
1317 break;
1318 case OE_RRR(V1SUBUC, 0, X0):
1319 case OE_RRR(V1SUBUC, 0, X1):
1320 return TILEGX_EXCP_OPCODE_UNIMPLEMENTED;
1321 case OE_RRR(V1SUB, 0, X0):
1322 case OE_RRR(V1SUB, 0, X1):
1323 gen_v12sub(tdest, tsrca, tsrcb, V1_IMM(0x80));
1324 mnemonic = "v1sub";
1325 break;
1326 case OE_RRR(V2ADDSC, 0, X0):
1327 case OE_RRR(V2ADDSC, 0, X1):
1328 return TILEGX_EXCP_OPCODE_UNIMPLEMENTED;
1329 case OE_RRR(V2ADD, 0, X0):
1330 case OE_RRR(V2ADD, 0, X1):
1331 gen_v12add(tdest, tsrca, tsrcb, V2_IMM(0x8000));
1332 mnemonic = "v2add";
1333 break;
1334 case OE_RRR(V2ADIFFS, 0, X0):
1335 case OE_RRR(V2AVGS, 0, X0):
1336 case OE_RRR(V2CMPEQ, 0, X0):
1337 case OE_RRR(V2CMPEQ, 0, X1):
1338 case OE_RRR(V2CMPLES, 0, X0):
1339 case OE_RRR(V2CMPLES, 0, X1):
1340 case OE_RRR(V2CMPLEU, 0, X0):
1341 case OE_RRR(V2CMPLEU, 0, X1):
1342 case OE_RRR(V2CMPLTS, 0, X0):
1343 case OE_RRR(V2CMPLTS, 0, X1):
1344 case OE_RRR(V2CMPLTU, 0, X0):
1345 case OE_RRR(V2CMPLTU, 0, X1):
1346 case OE_RRR(V2CMPNE, 0, X0):
1347 case OE_RRR(V2CMPNE, 0, X1):
1348 case OE_RRR(V2DOTPA, 0, X0):
1349 case OE_RRR(V2DOTP, 0, X0):
1350 return TILEGX_EXCP_OPCODE_UNIMPLEMENTED;
1351 case OE_RRR(V2INT_H, 0, X0):
1352 case OE_RRR(V2INT_H, 0, X1):
1353 gen_helper_v2int_h(tdest, tsrca, tsrcb);
1354 mnemonic = "v2int_h";
1355 break;
1356 case OE_RRR(V2INT_L, 0, X0):
1357 case OE_RRR(V2INT_L, 0, X1):
1358 gen_helper_v2int_l(tdest, tsrca, tsrcb);
1359 mnemonic = "v2int_l";
1360 break;
1361 case OE_RRR(V2MAXS, 0, X0):
1362 case OE_RRR(V2MAXS, 0, X1):
1363 case OE_RRR(V2MINS, 0, X0):
1364 case OE_RRR(V2MINS, 0, X1):
1365 case OE_RRR(V2MNZ, 0, X0):
1366 case OE_RRR(V2MNZ, 0, X1):
1367 case OE_RRR(V2MULFSC, 0, X0):
1368 case OE_RRR(V2MULS, 0, X0):
1369 return TILEGX_EXCP_OPCODE_UNIMPLEMENTED;
1370 case OE_RRR(V2MULTS, 0, X0):
1371 gen_helper_v2mults(tdest, tsrca, tsrcb);
1372 mnemonic = "v2mults";
1373 break;
1374 case OE_RRR(V2MZ, 0, X0):
1375 case OE_RRR(V2MZ, 0, X1):
1376 case OE_RRR(V2PACKH, 0, X0):
1377 case OE_RRR(V2PACKH, 0, X1):
1378 case OE_RRR(V2PACKL, 0, X0):
1379 case OE_RRR(V2PACKL, 0, X1):
1380 case OE_RRR(V2PACKUC, 0, X0):
1381 case OE_RRR(V2PACKUC, 0, X1):
1382 case OE_RRR(V2SADAS, 0, X0):
1383 case OE_RRR(V2SADAU, 0, X0):
1384 case OE_RRR(V2SADS, 0, X0):
1385 case OE_RRR(V2SADU, 0, X0):
1386 case OE_RRR(V2SHLSC, 0, X0):
1387 case OE_RRR(V2SHLSC, 0, X1):
1388 return TILEGX_EXCP_OPCODE_UNIMPLEMENTED;
1389 case OE_RRR(V2SHL, 0, X0):
1390 case OE_RRR(V2SHL, 0, X1):
1391 gen_helper_v2shl(tdest, tsrca, tsrcb);
1392 mnemonic = "v2shl";
1393 break;
1394 case OE_RRR(V2SHRS, 0, X0):
1395 case OE_RRR(V2SHRS, 0, X1):
1396 gen_helper_v2shrs(tdest, tsrca, tsrcb);
1397 mnemonic = "v2shrs";
1398 break;
1399 case OE_RRR(V2SHRU, 0, X0):
1400 case OE_RRR(V2SHRU, 0, X1):
1401 gen_helper_v2shru(tdest, tsrca, tsrcb);
1402 mnemonic = "v2shru";
1403 break;
1404 case OE_RRR(V2SUBSC, 0, X0):
1405 case OE_RRR(V2SUBSC, 0, X1):
1406 return TILEGX_EXCP_OPCODE_UNIMPLEMENTED;
1407 case OE_RRR(V2SUB, 0, X0):
1408 case OE_RRR(V2SUB, 0, X1):
1409 gen_v12sub(tdest, tsrca, tsrcb, V2_IMM(0x8000));
1410 mnemonic = "v2sub";
1411 break;
1412 case OE_RRR(V4ADDSC, 0, X0):
1413 case OE_RRR(V4ADDSC, 0, X1):
1414 return TILEGX_EXCP_OPCODE_UNIMPLEMENTED;
1415 case OE_RRR(V4ADD, 0, X0):
1416 case OE_RRR(V4ADD, 0, X1):
1417 gen_v4op(tdest, tsrca, tsrcb, tcg_gen_add_i32);
1418 mnemonic = "v4add";
1419 break;
1420 case OE_RRR(V4INT_H, 0, X0):
1421 case OE_RRR(V4INT_H, 0, X1):
1422 tcg_gen_shri_tl(tdest, tsrcb, 32);
1423 tcg_gen_deposit_tl(tdest, tsrca, tdest, 0, 32);
1424 mnemonic = "v4int_h";
1425 break;
1426 case OE_RRR(V4INT_L, 0, X0):
1427 case OE_RRR(V4INT_L, 0, X1):
1428 tcg_gen_deposit_tl(tdest, tsrcb, tsrca, 32, 32);
1429 mnemonic = "v4int_l";
1430 break;
1431 case OE_RRR(V4PACKSC, 0, X0):
1432 case OE_RRR(V4PACKSC, 0, X1):
1433 case OE_RRR(V4SHLSC, 0, X0):
1434 case OE_RRR(V4SHLSC, 0, X1):
1435 return TILEGX_EXCP_OPCODE_UNIMPLEMENTED;
1436 case OE_RRR(V4SHL, 0, X0):
1437 case OE_RRR(V4SHL, 0, X1):
1438 gen_v4sh(tdest, tsrca, tsrcb, tcg_gen_shl_i32);
1439 mnemonic = "v4shl";
1440 break;
1441 case OE_RRR(V4SHRS, 0, X0):
1442 case OE_RRR(V4SHRS, 0, X1):
1443 gen_v4sh(tdest, tsrca, tsrcb, tcg_gen_sar_i32);
1444 mnemonic = "v4shrs";
1445 break;
1446 case OE_RRR(V4SHRU, 0, X0):
1447 case OE_RRR(V4SHRU, 0, X1):
1448 gen_v4sh(tdest, tsrca, tsrcb, tcg_gen_shr_i32);
1449 mnemonic = "v4shru";
1450 break;
1451 case OE_RRR(V4SUBSC, 0, X0):
1452 case OE_RRR(V4SUBSC, 0, X1):
1453 return TILEGX_EXCP_OPCODE_UNIMPLEMENTED;
1454 case OE_RRR(V4SUB, 0, X0):
1455 case OE_RRR(V4SUB, 0, X1):
1456 gen_v4op(tdest, tsrca, tsrcb, tcg_gen_sub_i32);
1457 mnemonic = "v2sub";
1458 break;
1459 case OE_RRR(XOR, 0, X0):
1460 case OE_RRR(XOR, 0, X1):
1461 case OE_RRR(XOR, 5, Y0):
1462 case OE_RRR(XOR, 5, Y1):
1463 tcg_gen_xor_tl(tdest, tsrca, tsrcb);
1464 mnemonic = "xor";
1465 break;
1466 default:
1467 return TILEGX_EXCP_OPCODE_UNKNOWN;
1468 }
1469
1470 qemu_log_mask(CPU_LOG_TB_IN_ASM, "%s %s, %s, %s", mnemonic,
1471 reg_names[dest], reg_names[srca], reg_names[srcb]);
1472 return TILEGX_EXCP_NONE;
1473 }
1474
1475 static TileExcp gen_rri_opcode(DisasContext *dc, unsigned opext,
1476 unsigned dest, unsigned srca, int imm)
1477 {
1478 TCGv tdest = dest_gr(dc, dest);
1479 TCGv tsrca = load_gr(dc, srca);
1480 bool prefetch_nofault = false;
1481 const char *mnemonic;
1482 TCGMemOp memop;
1483 int i2, i3;
1484 TCGv t0;
1485
1486 switch (opext) {
1487 case OE(ADDI_OPCODE_Y0, 0, Y0):
1488 case OE(ADDI_OPCODE_Y1, 0, Y1):
1489 case OE_IM(ADDI, X0):
1490 case OE_IM(ADDI, X1):
1491 tcg_gen_addi_tl(tdest, tsrca, imm);
1492 mnemonic = "addi";
1493 break;
1494 case OE(ADDXI_OPCODE_Y0, 0, Y0):
1495 case OE(ADDXI_OPCODE_Y1, 0, Y1):
1496 case OE_IM(ADDXI, X0):
1497 case OE_IM(ADDXI, X1):
1498 tcg_gen_addi_tl(tdest, tsrca, imm);
1499 tcg_gen_ext32s_tl(tdest, tdest);
1500 mnemonic = "addxi";
1501 break;
1502 case OE(ANDI_OPCODE_Y0, 0, Y0):
1503 case OE(ANDI_OPCODE_Y1, 0, Y1):
1504 case OE_IM(ANDI, X0):
1505 case OE_IM(ANDI, X1):
1506 tcg_gen_andi_tl(tdest, tsrca, imm);
1507 mnemonic = "andi";
1508 break;
1509 case OE(CMPEQI_OPCODE_Y0, 0, Y0):
1510 case OE(CMPEQI_OPCODE_Y1, 0, Y1):
1511 case OE_IM(CMPEQI, X0):
1512 case OE_IM(CMPEQI, X1):
1513 tcg_gen_setcondi_tl(TCG_COND_EQ, tdest, tsrca, imm);
1514 mnemonic = "cmpeqi";
1515 break;
1516 case OE(CMPLTSI_OPCODE_Y0, 0, Y0):
1517 case OE(CMPLTSI_OPCODE_Y1, 0, Y1):
1518 case OE_IM(CMPLTSI, X0):
1519 case OE_IM(CMPLTSI, X1):
1520 tcg_gen_setcondi_tl(TCG_COND_LT, tdest, tsrca, imm);
1521 mnemonic = "cmpltsi";
1522 break;
1523 case OE_IM(CMPLTUI, X0):
1524 case OE_IM(CMPLTUI, X1):
1525 tcg_gen_setcondi_tl(TCG_COND_LTU, tdest, tsrca, imm);
1526 mnemonic = "cmpltui";
1527 break;
1528 case OE_IM(LD1S_ADD, X1):
1529 memop = MO_SB;
1530 mnemonic = "ld1s_add"; /* prefetch_add_l1_fault */
1531 goto do_load_add;
1532 case OE_IM(LD1U_ADD, X1):
1533 memop = MO_UB;
1534 mnemonic = "ld1u_add"; /* prefetch_add_l1 */
1535 prefetch_nofault = (dest == TILEGX_R_ZERO);
1536 goto do_load_add;
1537 case OE_IM(LD2S_ADD, X1):
1538 memop = MO_TESW;
1539 mnemonic = "ld2s_add"; /* prefetch_add_l2_fault */
1540 goto do_load_add;
1541 case OE_IM(LD2U_ADD, X1):
1542 memop = MO_TEUW;
1543 mnemonic = "ld2u_add"; /* prefetch_add_l2 */
1544 prefetch_nofault = (dest == TILEGX_R_ZERO);
1545 goto do_load_add;
1546 case OE_IM(LD4S_ADD, X1):
1547 memop = MO_TESL;
1548 mnemonic = "ld4s_add"; /* prefetch_add_l3_fault */
1549 goto do_load_add;
1550 case OE_IM(LD4U_ADD, X1):
1551 memop = MO_TEUL;
1552 mnemonic = "ld4u_add"; /* prefetch_add_l3 */
1553 prefetch_nofault = (dest == TILEGX_R_ZERO);
1554 goto do_load_add;
1555 case OE_IM(LDNT1S_ADD, X1):
1556 memop = MO_SB;
1557 mnemonic = "ldnt1s_add";
1558 goto do_load_add;
1559 case OE_IM(LDNT1U_ADD, X1):
1560 memop = MO_UB;
1561 mnemonic = "ldnt1u_add";
1562 goto do_load_add;
1563 case OE_IM(LDNT2S_ADD, X1):
1564 memop = MO_TESW;
1565 mnemonic = "ldnt2s_add";
1566 goto do_load_add;
1567 case OE_IM(LDNT2U_ADD, X1):
1568 memop = MO_TEUW;
1569 mnemonic = "ldnt2u_add";
1570 goto do_load_add;
1571 case OE_IM(LDNT4S_ADD, X1):
1572 memop = MO_TESL;
1573 mnemonic = "ldnt4s_add";
1574 goto do_load_add;
1575 case OE_IM(LDNT4U_ADD, X1):
1576 memop = MO_TEUL;
1577 mnemonic = "ldnt4u_add";
1578 goto do_load_add;
1579 case OE_IM(LDNT_ADD, X1):
1580 memop = MO_TEQ;
1581 mnemonic = "ldnt_add";
1582 goto do_load_add;
1583 case OE_IM(LD_ADD, X1):
1584 memop = MO_TEQ;
1585 mnemonic = "ld_add";
1586 do_load_add:
1587 if (!prefetch_nofault) {
1588 tcg_gen_qemu_ld_tl(tdest, tsrca, dc->mmuidx, memop);
1589 }
1590 tcg_gen_addi_tl(dest_gr(dc, srca), tsrca, imm);
1591 break;
1592 case OE_IM(LDNA_ADD, X1):
1593 tcg_gen_andi_tl(tdest, tsrca, ~7);
1594 tcg_gen_qemu_ld_tl(tdest, tdest, dc->mmuidx, MO_TEQ);
1595 tcg_gen_addi_tl(dest_gr(dc, srca), tsrca, imm);
1596 mnemonic = "ldna_add";
1597 break;
1598 case OE_IM(ORI, X0):
1599 case OE_IM(ORI, X1):
1600 tcg_gen_ori_tl(tdest, tsrca, imm);
1601 mnemonic = "ori";
1602 break;
1603 case OE_IM(V1ADDI, X0):
1604 case OE_IM(V1ADDI, X1):
1605 t0 = tcg_const_tl(V1_IMM(imm));
1606 gen_v12add(tdest, tsrca, t0, V1_IMM(0x80));
1607 tcg_temp_free(t0);
1608 mnemonic = "v1addi";
1609 break;
1610 case OE_IM(V1CMPEQI, X0):
1611 case OE_IM(V1CMPEQI, X1):
1612 tcg_gen_xori_tl(tdest, tsrca, V1_IMM(imm));
1613 gen_v1cmpeq0(tdest);
1614 mnemonic = "v1cmpeqi";
1615 break;
1616 case OE_IM(V1CMPLTSI, X0):
1617 case OE_IM(V1CMPLTSI, X1):
1618 case OE_IM(V1CMPLTUI, X0):
1619 case OE_IM(V1CMPLTUI, X1):
1620 case OE_IM(V1MAXUI, X0):
1621 case OE_IM(V1MAXUI, X1):
1622 case OE_IM(V1MINUI, X0):
1623 case OE_IM(V1MINUI, X1):
1624 return TILEGX_EXCP_OPCODE_UNIMPLEMENTED;
1625 case OE_IM(V2ADDI, X0):
1626 case OE_IM(V2ADDI, X1):
1627 t0 = tcg_const_tl(V2_IMM(imm));
1628 gen_v12add(tdest, tsrca, t0, V2_IMM(0x8000));
1629 tcg_temp_free(t0);
1630 mnemonic = "v2addi";
1631 break;
1632 case OE_IM(V2CMPEQI, X0):
1633 case OE_IM(V2CMPEQI, X1):
1634 case OE_IM(V2CMPLTSI, X0):
1635 case OE_IM(V2CMPLTSI, X1):
1636 case OE_IM(V2CMPLTUI, X0):
1637 case OE_IM(V2CMPLTUI, X1):
1638 case OE_IM(V2MAXSI, X0):
1639 case OE_IM(V2MAXSI, X1):
1640 case OE_IM(V2MINSI, X0):
1641 case OE_IM(V2MINSI, X1):
1642 return TILEGX_EXCP_OPCODE_UNIMPLEMENTED;
1643 case OE_IM(XORI, X0):
1644 case OE_IM(XORI, X1):
1645 tcg_gen_xori_tl(tdest, tsrca, imm);
1646 mnemonic = "xori";
1647 break;
1648
1649 case OE_SH(ROTLI, X0):
1650 case OE_SH(ROTLI, X1):
1651 case OE_SH(ROTLI, Y0):
1652 case OE_SH(ROTLI, Y1):
1653 tcg_gen_rotli_tl(tdest, tsrca, imm);
1654 mnemonic = "rotli";
1655 break;
1656 case OE_SH(SHLI, X0):
1657 case OE_SH(SHLI, X1):
1658 case OE_SH(SHLI, Y0):
1659 case OE_SH(SHLI, Y1):
1660 tcg_gen_shli_tl(tdest, tsrca, imm);
1661 mnemonic = "shli";
1662 break;
1663 case OE_SH(SHLXI, X0):
1664 case OE_SH(SHLXI, X1):
1665 tcg_gen_shli_tl(tdest, tsrca, imm & 31);
1666 tcg_gen_ext32s_tl(tdest, tdest);
1667 mnemonic = "shlxi";
1668 break;
1669 case OE_SH(SHRSI, X0):
1670 case OE_SH(SHRSI, X1):
1671 case OE_SH(SHRSI, Y0):
1672 case OE_SH(SHRSI, Y1):
1673 tcg_gen_sari_tl(tdest, tsrca, imm);
1674 mnemonic = "shrsi";
1675 break;
1676 case OE_SH(SHRUI, X0):
1677 case OE_SH(SHRUI, X1):
1678 case OE_SH(SHRUI, Y0):
1679 case OE_SH(SHRUI, Y1):
1680 tcg_gen_shri_tl(tdest, tsrca, imm);
1681 mnemonic = "shrui";
1682 break;
1683 case OE_SH(SHRUXI, X0):
1684 case OE_SH(SHRUXI, X1):
1685 if ((imm & 31) == 0) {
1686 tcg_gen_ext32s_tl(tdest, tsrca);
1687 } else {
1688 tcg_gen_ext32u_tl(tdest, tsrca);
1689 tcg_gen_shri_tl(tdest, tdest, imm & 31);
1690 }
1691 mnemonic = "shlxi";
1692 break;
1693 case OE_SH(V1SHLI, X0):
1694 case OE_SH(V1SHLI, X1):
1695 i2 = imm & 7;
1696 i3 = 0xff >> i2;
1697 tcg_gen_andi_tl(tdest, tsrca, V1_IMM(i3));
1698 tcg_gen_shli_tl(tdest, tdest, i2);
1699 mnemonic = "v1shli";
1700 break;
1701 case OE_SH(V1SHRSI, X0):
1702 case OE_SH(V1SHRSI, X1):
1703 t0 = tcg_const_tl(imm & 7);
1704 gen_helper_v1shrs(tdest, tsrca, t0);
1705 tcg_temp_free(t0);
1706 mnemonic = "v1shrsi";
1707 break;
1708 case OE_SH(V1SHRUI, X0):
1709 case OE_SH(V1SHRUI, X1):
1710 i2 = imm & 7;
1711 i3 = (0xff << i2) & 0xff;
1712 tcg_gen_andi_tl(tdest, tsrca, V1_IMM(i3));
1713 tcg_gen_shri_tl(tdest, tdest, i2);
1714 mnemonic = "v1shrui";
1715 break;
1716 case OE_SH(V2SHLI, X0):
1717 case OE_SH(V2SHLI, X1):
1718 i2 = imm & 15;
1719 i3 = 0xffff >> i2;
1720 tcg_gen_andi_tl(tdest, tsrca, V2_IMM(i3));
1721 tcg_gen_shli_tl(tdest, tdest, i2);
1722 mnemonic = "v2shli";
1723 break;
1724 case OE_SH(V2SHRSI, X0):
1725 case OE_SH(V2SHRSI, X1):
1726 t0 = tcg_const_tl(imm & 15);
1727 gen_helper_v2shrs(tdest, tsrca, t0);
1728 tcg_temp_free(t0);
1729 mnemonic = "v2shrsi";
1730 break;
1731 case OE_SH(V2SHRUI, X0):
1732 case OE_SH(V2SHRUI, X1):
1733 i2 = imm & 15;
1734 i3 = (0xffff << i2) & 0xffff;
1735 tcg_gen_andi_tl(tdest, tsrca, V2_IMM(i3));
1736 tcg_gen_shri_tl(tdest, tdest, i2);
1737 mnemonic = "v2shrui";
1738 break;
1739
1740 case OE(ADDLI_OPCODE_X0, 0, X0):
1741 case OE(ADDLI_OPCODE_X1, 0, X1):
1742 tcg_gen_addi_tl(tdest, tsrca, imm);
1743 mnemonic = "addli";
1744 break;
1745 case OE(ADDXLI_OPCODE_X0, 0, X0):
1746 case OE(ADDXLI_OPCODE_X1, 0, X1):
1747 tcg_gen_addi_tl(tdest, tsrca, imm);
1748 tcg_gen_ext32s_tl(tdest, tdest);
1749 mnemonic = "addxli";
1750 break;
1751 case OE(SHL16INSLI_OPCODE_X0, 0, X0):
1752 case OE(SHL16INSLI_OPCODE_X1, 0, X1):
1753 tcg_gen_shli_tl(tdest, tsrca, 16);
1754 tcg_gen_ori_tl(tdest, tdest, imm & 0xffff);
1755 mnemonic = "shl16insli";
1756 break;
1757
1758 default:
1759 return TILEGX_EXCP_OPCODE_UNKNOWN;
1760 }
1761
1762 qemu_log_mask(CPU_LOG_TB_IN_ASM, "%s %s, %s, %d", mnemonic,
1763 reg_names[dest], reg_names[srca], imm);
1764 return TILEGX_EXCP_NONE;
1765 }
1766
1767 static TileExcp gen_bf_opcode_x0(DisasContext *dc, unsigned ext,
1768 unsigned dest, unsigned srca,
1769 unsigned bfs, unsigned bfe)
1770 {
1771 TCGv tdest = dest_gr(dc, dest);
1772 TCGv tsrca = load_gr(dc, srca);
1773 TCGv tsrcd;
1774 int len;
1775 const char *mnemonic;
1776
1777 /* The bitfield is either between E and S inclusive,
1778 or up from S and down from E inclusive. */
1779 if (bfs <= bfe) {
1780 len = bfe - bfs + 1;
1781 } else {
1782 len = (64 - bfs) + (bfe + 1);
1783 }
1784
1785 switch (ext) {
1786 case BFEXTU_BF_OPCODE_X0:
1787 if (bfs == 0 && bfe == 7) {
1788 tcg_gen_ext8u_tl(tdest, tsrca);
1789 } else if (bfs == 0 && bfe == 15) {
1790 tcg_gen_ext16u_tl(tdest, tsrca);
1791 } else if (bfs == 0 && bfe == 31) {
1792 tcg_gen_ext32u_tl(tdest, tsrca);
1793 } else {
1794 int rol = 63 - bfe;
1795 if (bfs <= bfe) {
1796 tcg_gen_shli_tl(tdest, tsrca, rol);
1797 } else {
1798 tcg_gen_rotli_tl(tdest, tsrca, rol);
1799 }
1800 tcg_gen_shri_tl(tdest, tdest, (bfs + rol) & 63);
1801 }
1802 mnemonic = "bfextu";
1803 break;
1804
1805 case BFEXTS_BF_OPCODE_X0:
1806 if (bfs == 0 && bfe == 7) {
1807 tcg_gen_ext8s_tl(tdest, tsrca);
1808 } else if (bfs == 0 && bfe == 15) {
1809 tcg_gen_ext16s_tl(tdest, tsrca);
1810 } else if (bfs == 0 && bfe == 31) {
1811 tcg_gen_ext32s_tl(tdest, tsrca);
1812 } else {
1813 int rol = 63 - bfe;
1814 if (bfs <= bfe) {
1815 tcg_gen_shli_tl(tdest, tsrca, rol);
1816 } else {
1817 tcg_gen_rotli_tl(tdest, tsrca, rol);
1818 }
1819 tcg_gen_sari_tl(tdest, tdest, (bfs + rol) & 63);
1820 }
1821 mnemonic = "bfexts";
1822 break;
1823
1824 case BFINS_BF_OPCODE_X0:
1825 tsrcd = load_gr(dc, dest);
1826 if (bfs <= bfe) {
1827 tcg_gen_deposit_tl(tdest, tsrcd, tsrca, bfs, len);
1828 } else {
1829 tcg_gen_rotri_tl(tdest, tsrcd, bfs);
1830 tcg_gen_deposit_tl(tdest, tdest, tsrca, 0, len);
1831 tcg_gen_rotli_tl(tdest, tdest, bfs);
1832 }
1833 mnemonic = "bfins";
1834 break;
1835
1836 case MM_BF_OPCODE_X0:
1837 tsrcd = load_gr(dc, dest);
1838 if (bfs == 0) {
1839 tcg_gen_deposit_tl(tdest, tsrca, tsrcd, 0, len);
1840 } else {
1841 uint64_t mask = len == 64 ? -1 : rol64((1ULL << len) - 1, bfs);
1842 TCGv tmp = tcg_const_tl(mask);
1843
1844 tcg_gen_and_tl(tdest, tsrcd, tmp);
1845 tcg_gen_andc_tl(tmp, tsrca, tmp);
1846 tcg_gen_or_tl(tdest, tdest, tmp);
1847 tcg_temp_free(tmp);
1848 }
1849 mnemonic = "mm";
1850 break;
1851
1852 default:
1853 return TILEGX_EXCP_OPCODE_UNKNOWN;
1854 }
1855
1856 qemu_log_mask(CPU_LOG_TB_IN_ASM, "%s %s, %s, %u, %u", mnemonic,
1857 reg_names[dest], reg_names[srca], bfs, bfe);
1858 return TILEGX_EXCP_NONE;
1859 }
1860
1861 static TileExcp gen_branch_opcode_x1(DisasContext *dc, unsigned ext,
1862 unsigned srca, int off)
1863 {
1864 target_ulong tgt = dc->pc + off * TILEGX_BUNDLE_SIZE_IN_BYTES;
1865 const char *mnemonic;
1866
1867 dc->jmp.dest = tcg_const_tl(tgt);
1868 dc->jmp.val1 = tcg_temp_new();
1869 tcg_gen_mov_tl(dc->jmp.val1, load_gr(dc, srca));
1870
1871 /* Note that the "predict taken" opcodes have bit 0 clear.
1872 Therefore, fold the two cases together by setting bit 0. */
1873 switch (ext | 1) {
1874 case BEQZ_BRANCH_OPCODE_X1:
1875 dc->jmp.cond = TCG_COND_EQ;
1876 mnemonic = "beqz";
1877 break;
1878 case BNEZ_BRANCH_OPCODE_X1:
1879 dc->jmp.cond = TCG_COND_NE;
1880 mnemonic = "bnez";
1881 break;
1882 case BGEZ_BRANCH_OPCODE_X1:
1883 dc->jmp.cond = TCG_COND_GE;
1884 mnemonic = "bgez";
1885 break;
1886 case BGTZ_BRANCH_OPCODE_X1:
1887 dc->jmp.cond = TCG_COND_GT;
1888 mnemonic = "bgtz";
1889 break;
1890 case BLEZ_BRANCH_OPCODE_X1:
1891 dc->jmp.cond = TCG_COND_LE;
1892 mnemonic = "blez";
1893 break;
1894 case BLTZ_BRANCH_OPCODE_X1:
1895 dc->jmp.cond = TCG_COND_LT;
1896 mnemonic = "bltz";
1897 break;
1898 case BLBC_BRANCH_OPCODE_X1:
1899 dc->jmp.cond = TCG_COND_EQ;
1900 tcg_gen_andi_tl(dc->jmp.val1, dc->jmp.val1, 1);
1901 mnemonic = "blbc";
1902 break;
1903 case BLBS_BRANCH_OPCODE_X1:
1904 dc->jmp.cond = TCG_COND_NE;
1905 tcg_gen_andi_tl(dc->jmp.val1, dc->jmp.val1, 1);
1906 mnemonic = "blbs";
1907 break;
1908 default:
1909 return TILEGX_EXCP_OPCODE_UNKNOWN;
1910 }
1911
1912 if (qemu_loglevel_mask(CPU_LOG_TB_IN_ASM)) {
1913 qemu_log("%s%s %s, " TARGET_FMT_lx " <%s>",
1914 mnemonic, ext & 1 ? "" : "t",
1915 reg_names[srca], tgt, lookup_symbol(tgt));
1916 }
1917 return TILEGX_EXCP_NONE;
1918 }
1919
1920 static TileExcp gen_jump_opcode_x1(DisasContext *dc, unsigned ext, int off)
1921 {
1922 target_ulong tgt = dc->pc + off * TILEGX_BUNDLE_SIZE_IN_BYTES;
1923 const char *mnemonic = "j";
1924
1925 /* The extension field is 1 bit, therefore we only have JAL and J. */
1926 if (ext == JAL_JUMP_OPCODE_X1) {
1927 tcg_gen_movi_tl(dest_gr(dc, TILEGX_R_LR),
1928 dc->pc + TILEGX_BUNDLE_SIZE_IN_BYTES);
1929 mnemonic = "jal";
1930 }
1931 dc->jmp.cond = TCG_COND_ALWAYS;
1932 dc->jmp.dest = tcg_const_tl(tgt);
1933
1934 if (qemu_loglevel_mask(CPU_LOG_TB_IN_ASM)) {
1935 qemu_log("%s " TARGET_FMT_lx " <%s>",
1936 mnemonic, tgt, lookup_symbol(tgt));
1937 }
1938 return TILEGX_EXCP_NONE;
1939 }
1940
1941 typedef struct {
1942 const char *name;
1943 intptr_t offset;
1944 void (*get)(TCGv, TCGv_ptr);
1945 void (*put)(TCGv_ptr, TCGv);
1946 } TileSPR;
1947
1948 static const TileSPR *find_spr(unsigned spr)
1949 {
1950 /* Allow the compiler to construct the binary search tree. */
1951 #define D(N, O, G, P) \
1952 case SPR_##N: { static const TileSPR x = { #N, O, G, P }; return &x; }
1953
1954 switch (spr) {
1955 D(CMPEXCH_VALUE,
1956 offsetof(CPUTLGState, spregs[TILEGX_SPR_CMPEXCH]), 0, 0)
1957 D(INTERRUPT_CRITICAL_SECTION,
1958 offsetof(CPUTLGState, spregs[TILEGX_SPR_CRITICAL_SEC]), 0, 0)
1959 D(SIM_CONTROL,
1960 offsetof(CPUTLGState, spregs[TILEGX_SPR_SIM_CONTROL]), 0, 0)
1961 D(EX_CONTEXT_0_0,
1962 offsetof(CPUTLGState, spregs[TILEGX_SPR_EX_CONTEXT_0_0]), 0, 0)
1963 D(EX_CONTEXT_0_1,
1964 offsetof(CPUTLGState, spregs[TILEGX_SPR_EX_CONTEXT_0_1]), 0, 0)
1965 }
1966
1967 #undef D
1968
1969 qemu_log_mask(LOG_UNIMP, "UNIMP SPR %u\n", spr);
1970 return NULL;
1971 }
1972
1973 static TileExcp gen_mtspr_x1(DisasContext *dc, unsigned spr, unsigned srca)
1974 {
1975 const TileSPR *def = find_spr(spr);
1976 TCGv tsrca;
1977
1978 if (def == NULL) {
1979 qemu_log_mask(CPU_LOG_TB_IN_ASM, "mtspr spr[%u], %s", spr, reg_names[srca]);
1980 return TILEGX_EXCP_OPCODE_UNIMPLEMENTED;
1981 }
1982
1983 tsrca = load_gr(dc, srca);
1984 if (def->put) {
1985 def->put(cpu_env, tsrca);
1986 } else {
1987 tcg_gen_st_tl(tsrca, cpu_env, def->offset);
1988 }
1989 qemu_log_mask(CPU_LOG_TB_IN_ASM, "mtspr %s, %s", def->name, reg_names[srca]);
1990 return TILEGX_EXCP_NONE;
1991 }
1992
1993 static TileExcp gen_mfspr_x1(DisasContext *dc, unsigned dest, unsigned spr)
1994 {
1995 const TileSPR *def = find_spr(spr);
1996 TCGv tdest;
1997
1998 if (def == NULL) {
1999 qemu_log_mask(CPU_LOG_TB_IN_ASM, "mtspr %s, spr[%u]", reg_names[dest], spr);
2000 return TILEGX_EXCP_OPCODE_UNIMPLEMENTED;
2001 }
2002
2003 tdest = dest_gr(dc, dest);
2004 if (def->get) {
2005 def->get(tdest, cpu_env);
2006 } else {
2007 tcg_gen_ld_tl(tdest, cpu_env, def->offset);
2008 }
2009 qemu_log_mask(CPU_LOG_TB_IN_ASM, "mfspr %s, %s", reg_names[dest], def->name);
2010 return TILEGX_EXCP_NONE;
2011 }
2012
2013 static TileExcp decode_y0(DisasContext *dc, tilegx_bundle_bits bundle)
2014 {
2015 unsigned opc = get_Opcode_Y0(bundle);
2016 unsigned ext = get_RRROpcodeExtension_Y0(bundle);
2017 unsigned dest = get_Dest_Y0(bundle);
2018 unsigned srca = get_SrcA_Y0(bundle);
2019 unsigned srcb;
2020 int imm;
2021
2022 switch (opc) {
2023 case RRR_1_OPCODE_Y0:
2024 if (ext == UNARY_RRR_1_OPCODE_Y0) {
2025 ext = get_UnaryOpcodeExtension_Y0(bundle);
2026 return gen_rr_opcode(dc, OE(opc, ext, Y0), dest, srca, bundle);
2027 }
2028 /* fallthru */
2029 case RRR_0_OPCODE_Y0:
2030 case RRR_2_OPCODE_Y0:
2031 case RRR_3_OPCODE_Y0:
2032 case RRR_4_OPCODE_Y0:
2033 case RRR_5_OPCODE_Y0:
2034 case RRR_6_OPCODE_Y0:
2035 case RRR_7_OPCODE_Y0:
2036 case RRR_8_OPCODE_Y0:
2037 case RRR_9_OPCODE_Y0:
2038 srcb = get_SrcB_Y0(bundle);
2039 return gen_rrr_opcode(dc, OE(opc, ext, Y0), dest, srca, srcb);
2040
2041 case SHIFT_OPCODE_Y0:
2042 ext = get_ShiftOpcodeExtension_Y0(bundle);
2043 imm = get_ShAmt_Y0(bundle);
2044 return gen_rri_opcode(dc, OE(opc, ext, Y0), dest, srca, imm);
2045
2046 case ADDI_OPCODE_Y0:
2047 case ADDXI_OPCODE_Y0:
2048 case ANDI_OPCODE_Y0:
2049 case CMPEQI_OPCODE_Y0:
2050 case CMPLTSI_OPCODE_Y0:
2051 imm = (int8_t)get_Imm8_Y0(bundle);
2052 return gen_rri_opcode(dc, OE(opc, 0, Y0), dest, srca, imm);
2053
2054 default:
2055 return TILEGX_EXCP_OPCODE_UNKNOWN;
2056 }
2057 }
2058
2059 static TileExcp decode_y1(DisasContext *dc, tilegx_bundle_bits bundle)
2060 {
2061 unsigned opc = get_Opcode_Y1(bundle);
2062 unsigned ext = get_RRROpcodeExtension_Y1(bundle);
2063 unsigned dest = get_Dest_Y1(bundle);
2064 unsigned srca = get_SrcA_Y1(bundle);
2065 unsigned srcb;
2066 int imm;
2067
2068 switch (get_Opcode_Y1(bundle)) {
2069 case RRR_1_OPCODE_Y1:
2070 if (ext == UNARY_RRR_1_OPCODE_Y0) {
2071 ext = get_UnaryOpcodeExtension_Y1(bundle);
2072 return gen_rr_opcode(dc, OE(opc, ext, Y1), dest, srca, bundle);
2073 }
2074 /* fallthru */
2075 case RRR_0_OPCODE_Y1:
2076 case RRR_2_OPCODE_Y1:
2077 case RRR_3_OPCODE_Y1:
2078 case RRR_4_OPCODE_Y1:
2079 case RRR_5_OPCODE_Y1:
2080 case RRR_6_OPCODE_Y1:
2081 case RRR_7_OPCODE_Y1:
2082 srcb = get_SrcB_Y1(bundle);
2083 return gen_rrr_opcode(dc, OE(opc, ext, Y1), dest, srca, srcb);
2084
2085 case SHIFT_OPCODE_Y1:
2086 ext = get_ShiftOpcodeExtension_Y1(bundle);
2087 imm = get_ShAmt_Y1(bundle);
2088 return gen_rri_opcode(dc, OE(opc, ext, Y1), dest, srca, imm);
2089
2090 case ADDI_OPCODE_Y1:
2091 case ADDXI_OPCODE_Y1:
2092 case ANDI_OPCODE_Y1:
2093 case CMPEQI_OPCODE_Y1:
2094 case CMPLTSI_OPCODE_Y1:
2095 imm = (int8_t)get_Imm8_Y1(bundle);
2096 return gen_rri_opcode(dc, OE(opc, 0, Y1), dest, srca, imm);
2097
2098 default:
2099 return TILEGX_EXCP_OPCODE_UNKNOWN;
2100 }
2101 }
2102
2103 static TileExcp decode_y2(DisasContext *dc, tilegx_bundle_bits bundle)
2104 {
2105 unsigned mode = get_Mode(bundle);
2106 unsigned opc = get_Opcode_Y2(bundle);
2107 unsigned srca = get_SrcA_Y2(bundle);
2108 unsigned srcbdest = get_SrcBDest_Y2(bundle);
2109 const char *mnemonic;
2110 TCGMemOp memop;
2111 bool prefetch_nofault = false;
2112
2113 switch (OEY2(opc, mode)) {
2114 case OEY2(LD1S_OPCODE_Y2, MODE_OPCODE_YA2):
2115 memop = MO_SB;
2116 mnemonic = "ld1s"; /* prefetch_l1_fault */
2117 goto do_load;
2118 case OEY2(LD1U_OPCODE_Y2, MODE_OPCODE_YA2):
2119 memop = MO_UB;
2120 mnemonic = "ld1u"; /* prefetch, prefetch_l1 */
2121 prefetch_nofault = (srcbdest == TILEGX_R_ZERO);
2122 goto do_load;
2123 case OEY2(LD2S_OPCODE_Y2, MODE_OPCODE_YA2):
2124 memop = MO_TESW;
2125 mnemonic = "ld2s"; /* prefetch_l2_fault */
2126 goto do_load;
2127 case OEY2(LD2U_OPCODE_Y2, MODE_OPCODE_YA2):
2128 memop = MO_TEUW;
2129 mnemonic = "ld2u"; /* prefetch_l2 */
2130 prefetch_nofault = (srcbdest == TILEGX_R_ZERO);
2131 goto do_load;
2132 case OEY2(LD4S_OPCODE_Y2, MODE_OPCODE_YB2):
2133 memop = MO_TESL;
2134 mnemonic = "ld4s"; /* prefetch_l3_fault */
2135 goto do_load;
2136 case OEY2(LD4U_OPCODE_Y2, MODE_OPCODE_YB2):
2137 memop = MO_TEUL;
2138 mnemonic = "ld4u"; /* prefetch_l3 */
2139 prefetch_nofault = (srcbdest == TILEGX_R_ZERO);
2140 goto do_load;
2141 case OEY2(LD_OPCODE_Y2, MODE_OPCODE_YB2):
2142 memop = MO_TEQ;
2143 mnemonic = "ld";
2144 do_load:
2145 if (!prefetch_nofault) {
2146 tcg_gen_qemu_ld_tl(dest_gr(dc, srcbdest), load_gr(dc, srca),
2147 dc->mmuidx, memop);
2148 }
2149 qemu_log_mask(CPU_LOG_TB_IN_ASM, "%s %s, %s", mnemonic,
2150 reg_names[srcbdest], reg_names[srca]);
2151 return TILEGX_EXCP_NONE;
2152
2153 case OEY2(ST1_OPCODE_Y2, MODE_OPCODE_YC2):
2154 return gen_st_opcode(dc, 0, srca, srcbdest, MO_UB, "st1");
2155 case OEY2(ST2_OPCODE_Y2, MODE_OPCODE_YC2):
2156 return gen_st_opcode(dc, 0, srca, srcbdest, MO_TEUW, "st2");
2157 case OEY2(ST4_OPCODE_Y2, MODE_OPCODE_YC2):
2158 return gen_st_opcode(dc, 0, srca, srcbdest, MO_TEUL, "st4");
2159 case OEY2(ST_OPCODE_Y2, MODE_OPCODE_YC2):
2160 return gen_st_opcode(dc, 0, srca, srcbdest, MO_TEQ, "st");
2161
2162 default:
2163 return TILEGX_EXCP_OPCODE_UNKNOWN;
2164 }
2165 }
2166
2167 static TileExcp decode_x0(DisasContext *dc, tilegx_bundle_bits bundle)
2168 {
2169 unsigned opc = get_Opcode_X0(bundle);
2170 unsigned dest = get_Dest_X0(bundle);
2171 unsigned srca = get_SrcA_X0(bundle);
2172 unsigned ext, srcb, bfs, bfe;
2173 int imm;
2174
2175 switch (opc) {
2176 case RRR_0_OPCODE_X0:
2177 ext = get_RRROpcodeExtension_X0(bundle);
2178 if (ext == UNARY_RRR_0_OPCODE_X0) {
2179 ext = get_UnaryOpcodeExtension_X0(bundle);
2180 return gen_rr_opcode(dc, OE(opc, ext, X0), dest, srca, bundle);
2181 }
2182 srcb = get_SrcB_X0(bundle);
2183 return gen_rrr_opcode(dc, OE(opc, ext, X0), dest, srca, srcb);
2184
2185 case SHIFT_OPCODE_X0:
2186 ext = get_ShiftOpcodeExtension_X0(bundle);
2187 imm = get_ShAmt_X0(bundle);
2188 return gen_rri_opcode(dc, OE(opc, ext, X0), dest, srca, imm);
2189
2190 case IMM8_OPCODE_X0:
2191 ext = get_Imm8OpcodeExtension_X0(bundle);
2192 imm = (int8_t)get_Imm8_X0(bundle);
2193 return gen_rri_opcode(dc, OE(opc, ext, X0), dest, srca, imm);
2194
2195 case BF_OPCODE_X0:
2196 ext = get_BFOpcodeExtension_X0(bundle);
2197 bfs = get_BFStart_X0(bundle);
2198 bfe = get_BFEnd_X0(bundle);
2199 return gen_bf_opcode_x0(dc, ext, dest, srca, bfs, bfe);
2200
2201 case ADDLI_OPCODE_X0:
2202 case SHL16INSLI_OPCODE_X0:
2203 case ADDXLI_OPCODE_X0:
2204 imm = (int16_t)get_Imm16_X0(bundle);
2205 return gen_rri_opcode(dc, OE(opc, 0, X0), dest, srca, imm);
2206
2207 default:
2208 return TILEGX_EXCP_OPCODE_UNKNOWN;
2209 }
2210 }
2211
2212 static TileExcp decode_x1(DisasContext *dc, tilegx_bundle_bits bundle)
2213 {
2214 unsigned opc = get_Opcode_X1(bundle);
2215 unsigned dest = get_Dest_X1(bundle);
2216 unsigned srca = get_SrcA_X1(bundle);
2217 unsigned ext, srcb;
2218 int imm;
2219
2220 switch (opc) {
2221 case RRR_0_OPCODE_X1:
2222 ext = get_RRROpcodeExtension_X1(bundle);
2223 srcb = get_SrcB_X1(bundle);
2224 switch (ext) {
2225 case UNARY_RRR_0_OPCODE_X1:
2226 ext = get_UnaryOpcodeExtension_X1(bundle);
2227 return gen_rr_opcode(dc, OE(opc, ext, X1), dest, srca, bundle);
2228 case ST1_RRR_0_OPCODE_X1:
2229 return gen_st_opcode(dc, dest, srca, srcb, MO_UB, "st1");
2230 case ST2_RRR_0_OPCODE_X1:
2231 return gen_st_opcode(dc, dest, srca, srcb, MO_TEUW, "st2");
2232 case ST4_RRR_0_OPCODE_X1:
2233 return gen_st_opcode(dc, dest, srca, srcb, MO_TEUL, "st4");
2234 case STNT1_RRR_0_OPCODE_X1:
2235 return gen_st_opcode(dc, dest, srca, srcb, MO_UB, "stnt1");
2236 case STNT2_RRR_0_OPCODE_X1:
2237 return gen_st_opcode(dc, dest, srca, srcb, MO_TEUW, "stnt2");
2238 case STNT4_RRR_0_OPCODE_X1:
2239 return gen_st_opcode(dc, dest, srca, srcb, MO_TEUL, "stnt4");
2240 case STNT_RRR_0_OPCODE_X1:
2241 return gen_st_opcode(dc, dest, srca, srcb, MO_TEQ, "stnt");
2242 case ST_RRR_0_OPCODE_X1:
2243 return gen_st_opcode(dc, dest, srca, srcb, MO_TEQ, "st");
2244 }
2245 return gen_rrr_opcode(dc, OE(opc, ext, X1), dest, srca, srcb);
2246
2247 case SHIFT_OPCODE_X1:
2248 ext = get_ShiftOpcodeExtension_X1(bundle);
2249 imm = get_ShAmt_X1(bundle);
2250 return gen_rri_opcode(dc, OE(opc, ext, X1), dest, srca, imm);
2251
2252 case IMM8_OPCODE_X1:
2253 ext = get_Imm8OpcodeExtension_X1(bundle);
2254 imm = (int8_t)get_Dest_Imm8_X1(bundle);
2255 srcb = get_SrcB_X1(bundle);
2256 switch (ext) {
2257 case ST1_ADD_IMM8_OPCODE_X1:
2258 return gen_st_add_opcode(dc, srca, srcb, imm, MO_UB, "st1_add");
2259 case ST2_ADD_IMM8_OPCODE_X1:
2260 return gen_st_add_opcode(dc, srca, srcb, imm, MO_TEUW, "st2_add");
2261 case ST4_ADD_IMM8_OPCODE_X1:
2262 return gen_st_add_opcode(dc, srca, srcb, imm, MO_TEUL, "st4_add");
2263 case STNT1_ADD_IMM8_OPCODE_X1:
2264 return gen_st_add_opcode(dc, srca, srcb, imm, MO_UB, "stnt1_add");
2265 case STNT2_ADD_IMM8_OPCODE_X1:
2266 return gen_st_add_opcode(dc, srca, srcb, imm, MO_TEUW, "stnt2_add");
2267 case STNT4_ADD_IMM8_OPCODE_X1:
2268 return gen_st_add_opcode(dc, srca, srcb, imm, MO_TEUL, "stnt4_add");
2269 case STNT_ADD_IMM8_OPCODE_X1:
2270 return gen_st_add_opcode(dc, srca, srcb, imm, MO_TEQ, "stnt_add");
2271 case ST_ADD_IMM8_OPCODE_X1:
2272 return gen_st_add_opcode(dc, srca, srcb, imm, MO_TEQ, "st_add");
2273 case MFSPR_IMM8_OPCODE_X1:
2274 return gen_mfspr_x1(dc, dest, get_MF_Imm14_X1(bundle));
2275 case MTSPR_IMM8_OPCODE_X1:
2276 return gen_mtspr_x1(dc, get_MT_Imm14_X1(bundle), srca);
2277 }
2278 imm = (int8_t)get_Imm8_X1(bundle);
2279 return gen_rri_opcode(dc, OE(opc, ext, X1), dest, srca, imm);
2280
2281 case BRANCH_OPCODE_X1:
2282 ext = get_BrType_X1(bundle);
2283 imm = sextract32(get_BrOff_X1(bundle), 0, 17);
2284 return gen_branch_opcode_x1(dc, ext, srca, imm);
2285
2286 case JUMP_OPCODE_X1:
2287 ext = get_JumpOpcodeExtension_X1(bundle);
2288 imm = sextract32(get_JumpOff_X1(bundle), 0, 27);
2289 return gen_jump_opcode_x1(dc, ext, imm);
2290
2291 case ADDLI_OPCODE_X1:
2292 case SHL16INSLI_OPCODE_X1:
2293 case ADDXLI_OPCODE_X1:
2294 imm = (int16_t)get_Imm16_X1(bundle);
2295 return gen_rri_opcode(dc, OE(opc, 0, X1), dest, srca, imm);
2296
2297 default:
2298 return TILEGX_EXCP_OPCODE_UNKNOWN;
2299 }
2300 }
2301
2302 static void notice_excp(DisasContext *dc, uint64_t bundle,
2303 const char *type, TileExcp excp)
2304 {
2305 if (likely(excp == TILEGX_EXCP_NONE)) {
2306 return;
2307 }
2308 gen_exception(dc, excp);
2309 switch (excp) {
2310 case TILEGX_EXCP_OPCODE_UNIMPLEMENTED:
2311 qemu_log_mask(LOG_UNIMP, "UNIMP %s, [" FMT64X "]\n", type, bundle);
2312 break;
2313 case TILEGX_EXCP_OPCODE_UNKNOWN:
2314 qemu_log_mask(LOG_UNIMP, "UNKNOWN %s, [" FMT64X "]\n", type, bundle);
2315 break;
2316 default:
2317 break;
2318 }
2319 }
2320
2321 static void translate_one_bundle(DisasContext *dc, uint64_t bundle)
2322 {
2323 int i;
2324
2325 for (i = 0; i < ARRAY_SIZE(dc->wb); i++) {
2326 DisasContextTemp *wb = &dc->wb[i];
2327 wb->reg = TILEGX_R_NOREG;
2328 TCGV_UNUSED_I64(wb->val);
2329 }
2330 dc->num_wb = 0;
2331
2332 qemu_log_mask(CPU_LOG_TB_IN_ASM, " %" PRIx64 ": { ", dc->pc);
2333 if (get_Mode(bundle)) {
2334 notice_excp(dc, bundle, "y0", decode_y0(dc, bundle));
2335 qemu_log_mask(CPU_LOG_TB_IN_ASM, " ; ");
2336 notice_excp(dc, bundle, "y1", decode_y1(dc, bundle));
2337 qemu_log_mask(CPU_LOG_TB_IN_ASM, " ; ");
2338 notice_excp(dc, bundle, "y2", decode_y2(dc, bundle));
2339 } else {
2340 notice_excp(dc, bundle, "x0", decode_x0(dc, bundle));
2341 qemu_log_mask(CPU_LOG_TB_IN_ASM, " ; ");
2342 notice_excp(dc, bundle, "x1", decode_x1(dc, bundle));
2343 }
2344 qemu_log_mask(CPU_LOG_TB_IN_ASM, " }\n");
2345
2346 for (i = dc->num_wb - 1; i >= 0; --i) {
2347 DisasContextTemp *wb = &dc->wb[i];
2348 if (wb->reg < TILEGX_R_COUNT) {
2349 tcg_gen_mov_i64(cpu_regs[wb->reg], wb->val);
2350 }
2351 tcg_temp_free_i64(wb->val);
2352 }
2353
2354 if (dc->jmp.cond != TCG_COND_NEVER) {
2355 if (dc->jmp.cond == TCG_COND_ALWAYS) {
2356 tcg_gen_mov_i64(cpu_pc, dc->jmp.dest);
2357 } else {
2358 TCGv next = tcg_const_i64(dc->pc + TILEGX_BUNDLE_SIZE_IN_BYTES);
2359 tcg_gen_movcond_i64(dc->jmp.cond, cpu_pc,
2360 dc->jmp.val1, load_zero(dc),
2361 dc->jmp.dest, next);
2362 tcg_temp_free_i64(dc->jmp.val1);
2363 tcg_temp_free_i64(next);
2364 }
2365 tcg_temp_free_i64(dc->jmp.dest);
2366 tcg_gen_exit_tb(0);
2367 dc->exit_tb = true;
2368 } else if (dc->atomic_excp != TILEGX_EXCP_NONE) {
2369 gen_exception(dc, dc->atomic_excp);
2370 }
2371 }
2372
2373 void gen_intermediate_code(CPUTLGState *env, struct TranslationBlock *tb)
2374 {
2375 TileGXCPU *cpu = tilegx_env_get_cpu(env);
2376 DisasContext ctx;
2377 DisasContext *dc = &ctx;
2378 CPUState *cs = CPU(cpu);
2379 uint64_t pc_start = tb->pc;
2380 uint64_t next_page_start = (pc_start & TARGET_PAGE_MASK) + TARGET_PAGE_SIZE;
2381 int num_insns = 0;
2382 int max_insns = tb->cflags & CF_COUNT_MASK;
2383
2384 dc->pc = pc_start;
2385 dc->mmuidx = 0;
2386 dc->exit_tb = false;
2387 dc->atomic_excp = TILEGX_EXCP_NONE;
2388 dc->jmp.cond = TCG_COND_NEVER;
2389 TCGV_UNUSED_I64(dc->jmp.dest);
2390 TCGV_UNUSED_I64(dc->jmp.val1);
2391 TCGV_UNUSED_I64(dc->zero);
2392
2393 if (qemu_loglevel_mask(CPU_LOG_TB_IN_ASM)) {
2394 qemu_log_lock();
2395 qemu_log("IN: %s\n", lookup_symbol(pc_start));
2396 }
2397 if (!max_insns) {
2398 max_insns = CF_COUNT_MASK;
2399 }
2400 if (cs->singlestep_enabled || singlestep) {
2401 max_insns = 1;
2402 }
2403 if (max_insns > TCG_MAX_INSNS) {
2404 max_insns = TCG_MAX_INSNS;
2405 }
2406 gen_tb_start(tb);
2407
2408 while (1) {
2409 tcg_gen_insn_start(dc->pc);
2410 num_insns++;
2411
2412 translate_one_bundle(dc, cpu_ldq_data(env, dc->pc));
2413
2414 if (dc->exit_tb) {
2415 /* PC updated and EXIT_TB/GOTO_TB/exception emitted. */
2416 break;
2417 }
2418 dc->pc += TILEGX_BUNDLE_SIZE_IN_BYTES;
2419 if (num_insns >= max_insns
2420 || dc->pc >= next_page_start
2421 || tcg_op_buf_full()) {
2422 /* Ending the TB due to TB size or page boundary. Set PC. */
2423 tcg_gen_movi_tl(cpu_pc, dc->pc);
2424 tcg_gen_exit_tb(0);
2425 break;
2426 }
2427 }
2428
2429 gen_tb_end(tb, num_insns);
2430 tb->size = dc->pc - pc_start;
2431 tb->icount = num_insns;
2432
2433 if (qemu_loglevel_mask(CPU_LOG_TB_IN_ASM)) {
2434 qemu_log("\n");
2435 qemu_log_unlock();
2436 }
2437 }
2438
2439 void restore_state_to_opc(CPUTLGState *env, TranslationBlock *tb,
2440 target_ulong *data)
2441 {
2442 env->pc = data[0];
2443 }
2444
2445 void tilegx_tcg_init(void)
2446 {
2447 int i;
2448
2449 cpu_env = tcg_global_reg_new_ptr(TCG_AREG0, "env");
2450 tcg_ctx.tcg_env = cpu_env;
2451 cpu_pc = tcg_global_mem_new_i64(cpu_env, offsetof(CPUTLGState, pc), "pc");
2452 for (i = 0; i < TILEGX_R_COUNT; i++) {
2453 cpu_regs[i] = tcg_global_mem_new_i64(cpu_env,
2454 offsetof(CPUTLGState, regs[i]),
2455 reg_names[i]);
2456 }
2457 }