xref: /openbmc/qemu/target/sh4/translate.c (revision 327d4b7f)
1 /*
2  *  SH4 translation
3  *
4  *  Copyright (c) 2005 Samuel Tardieu
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 <http://www.gnu.org/licenses/>.
18  */
19 
20 #define DEBUG_DISAS
21 
22 #include "qemu/osdep.h"
23 #include "cpu.h"
24 #include "disas/disas.h"
25 #include "exec/exec-all.h"
26 #include "tcg/tcg-op.h"
27 #include "exec/cpu_ldst.h"
28 #include "exec/helper-proto.h"
29 #include "exec/helper-gen.h"
30 #include "exec/translator.h"
31 #include "trace-tcg.h"
32 #include "exec/log.h"
33 #include "qemu/qemu-print.h"
34 
35 
36 typedef struct DisasContext {
37     DisasContextBase base;
38 
39     uint32_t tbflags;  /* should stay unmodified during the TB translation */
40     uint32_t envflags; /* should stay in sync with env->flags using TCG ops */
41     int memidx;
42     int gbank;
43     int fbank;
44     uint32_t delayed_pc;
45     uint32_t features;
46 
47     uint16_t opcode;
48 
49     bool has_movcal;
50 } DisasContext;
51 
52 #if defined(CONFIG_USER_ONLY)
53 #define IS_USER(ctx) 1
54 #else
55 #define IS_USER(ctx) (!(ctx->tbflags & (1u << SR_MD)))
56 #endif
57 
58 /* Target-specific values for ctx->base.is_jmp.  */
59 /* We want to exit back to the cpu loop for some reason.
60    Usually this is to recognize interrupts immediately.  */
61 #define DISAS_STOP    DISAS_TARGET_0
62 
63 /* global register indexes */
64 static TCGv cpu_gregs[32];
65 static TCGv cpu_sr, cpu_sr_m, cpu_sr_q, cpu_sr_t;
66 static TCGv cpu_pc, cpu_ssr, cpu_spc, cpu_gbr;
67 static TCGv cpu_vbr, cpu_sgr, cpu_dbr, cpu_mach, cpu_macl;
68 static TCGv cpu_pr, cpu_fpscr, cpu_fpul;
69 static TCGv cpu_lock_addr, cpu_lock_value;
70 static TCGv cpu_fregs[32];
71 
72 /* internal register indexes */
73 static TCGv cpu_flags, cpu_delayed_pc, cpu_delayed_cond;
74 
75 #include "exec/gen-icount.h"
76 
77 void sh4_translate_init(void)
78 {
79     int i;
80     static const char * const gregnames[24] = {
81         "R0_BANK0", "R1_BANK0", "R2_BANK0", "R3_BANK0",
82         "R4_BANK0", "R5_BANK0", "R6_BANK0", "R7_BANK0",
83         "R8", "R9", "R10", "R11", "R12", "R13", "R14", "R15",
84         "R0_BANK1", "R1_BANK1", "R2_BANK1", "R3_BANK1",
85         "R4_BANK1", "R5_BANK1", "R6_BANK1", "R7_BANK1"
86     };
87     static const char * const fregnames[32] = {
88          "FPR0_BANK0",  "FPR1_BANK0",  "FPR2_BANK0",  "FPR3_BANK0",
89          "FPR4_BANK0",  "FPR5_BANK0",  "FPR6_BANK0",  "FPR7_BANK0",
90          "FPR8_BANK0",  "FPR9_BANK0", "FPR10_BANK0", "FPR11_BANK0",
91         "FPR12_BANK0", "FPR13_BANK0", "FPR14_BANK0", "FPR15_BANK0",
92          "FPR0_BANK1",  "FPR1_BANK1",  "FPR2_BANK1",  "FPR3_BANK1",
93          "FPR4_BANK1",  "FPR5_BANK1",  "FPR6_BANK1",  "FPR7_BANK1",
94          "FPR8_BANK1",  "FPR9_BANK1", "FPR10_BANK1", "FPR11_BANK1",
95         "FPR12_BANK1", "FPR13_BANK1", "FPR14_BANK1", "FPR15_BANK1",
96     };
97 
98     for (i = 0; i < 24; i++) {
99         cpu_gregs[i] = tcg_global_mem_new_i32(cpu_env,
100                                               offsetof(CPUSH4State, gregs[i]),
101                                               gregnames[i]);
102     }
103     memcpy(cpu_gregs + 24, cpu_gregs + 8, 8 * sizeof(TCGv));
104 
105     cpu_pc = tcg_global_mem_new_i32(cpu_env,
106                                     offsetof(CPUSH4State, pc), "PC");
107     cpu_sr = tcg_global_mem_new_i32(cpu_env,
108                                     offsetof(CPUSH4State, sr), "SR");
109     cpu_sr_m = tcg_global_mem_new_i32(cpu_env,
110                                       offsetof(CPUSH4State, sr_m), "SR_M");
111     cpu_sr_q = tcg_global_mem_new_i32(cpu_env,
112                                       offsetof(CPUSH4State, sr_q), "SR_Q");
113     cpu_sr_t = tcg_global_mem_new_i32(cpu_env,
114                                       offsetof(CPUSH4State, sr_t), "SR_T");
115     cpu_ssr = tcg_global_mem_new_i32(cpu_env,
116                                      offsetof(CPUSH4State, ssr), "SSR");
117     cpu_spc = tcg_global_mem_new_i32(cpu_env,
118                                      offsetof(CPUSH4State, spc), "SPC");
119     cpu_gbr = tcg_global_mem_new_i32(cpu_env,
120                                      offsetof(CPUSH4State, gbr), "GBR");
121     cpu_vbr = tcg_global_mem_new_i32(cpu_env,
122                                      offsetof(CPUSH4State, vbr), "VBR");
123     cpu_sgr = tcg_global_mem_new_i32(cpu_env,
124                                      offsetof(CPUSH4State, sgr), "SGR");
125     cpu_dbr = tcg_global_mem_new_i32(cpu_env,
126                                      offsetof(CPUSH4State, dbr), "DBR");
127     cpu_mach = tcg_global_mem_new_i32(cpu_env,
128                                       offsetof(CPUSH4State, mach), "MACH");
129     cpu_macl = tcg_global_mem_new_i32(cpu_env,
130                                       offsetof(CPUSH4State, macl), "MACL");
131     cpu_pr = tcg_global_mem_new_i32(cpu_env,
132                                     offsetof(CPUSH4State, pr), "PR");
133     cpu_fpscr = tcg_global_mem_new_i32(cpu_env,
134                                        offsetof(CPUSH4State, fpscr), "FPSCR");
135     cpu_fpul = tcg_global_mem_new_i32(cpu_env,
136                                       offsetof(CPUSH4State, fpul), "FPUL");
137 
138     cpu_flags = tcg_global_mem_new_i32(cpu_env,
139 				       offsetof(CPUSH4State, flags), "_flags_");
140     cpu_delayed_pc = tcg_global_mem_new_i32(cpu_env,
141 					    offsetof(CPUSH4State, delayed_pc),
142 					    "_delayed_pc_");
143     cpu_delayed_cond = tcg_global_mem_new_i32(cpu_env,
144                                               offsetof(CPUSH4State,
145                                                        delayed_cond),
146                                               "_delayed_cond_");
147     cpu_lock_addr = tcg_global_mem_new_i32(cpu_env,
148                                            offsetof(CPUSH4State, lock_addr),
149                                            "_lock_addr_");
150     cpu_lock_value = tcg_global_mem_new_i32(cpu_env,
151                                             offsetof(CPUSH4State, lock_value),
152                                             "_lock_value_");
153 
154     for (i = 0; i < 32; i++)
155         cpu_fregs[i] = tcg_global_mem_new_i32(cpu_env,
156                                               offsetof(CPUSH4State, fregs[i]),
157                                               fregnames[i]);
158 }
159 
160 void superh_cpu_dump_state(CPUState *cs, FILE *f, int flags)
161 {
162     SuperHCPU *cpu = SUPERH_CPU(cs);
163     CPUSH4State *env = &cpu->env;
164     int i;
165 
166     qemu_fprintf(f, "pc=0x%08x sr=0x%08x pr=0x%08x fpscr=0x%08x\n",
167                  env->pc, cpu_read_sr(env), env->pr, env->fpscr);
168     qemu_fprintf(f, "spc=0x%08x ssr=0x%08x gbr=0x%08x vbr=0x%08x\n",
169                  env->spc, env->ssr, env->gbr, env->vbr);
170     qemu_fprintf(f, "sgr=0x%08x dbr=0x%08x delayed_pc=0x%08x fpul=0x%08x\n",
171                  env->sgr, env->dbr, env->delayed_pc, env->fpul);
172     for (i = 0; i < 24; i += 4) {
173         qemu_printf("r%d=0x%08x r%d=0x%08x r%d=0x%08x r%d=0x%08x\n",
174 		    i, env->gregs[i], i + 1, env->gregs[i + 1],
175 		    i + 2, env->gregs[i + 2], i + 3, env->gregs[i + 3]);
176     }
177     if (env->flags & DELAY_SLOT) {
178         qemu_printf("in delay slot (delayed_pc=0x%08x)\n",
179 		    env->delayed_pc);
180     } else if (env->flags & DELAY_SLOT_CONDITIONAL) {
181         qemu_printf("in conditional delay slot (delayed_pc=0x%08x)\n",
182 		    env->delayed_pc);
183     } else if (env->flags & DELAY_SLOT_RTE) {
184         qemu_fprintf(f, "in rte delay slot (delayed_pc=0x%08x)\n",
185                      env->delayed_pc);
186     }
187 }
188 
189 static void gen_read_sr(TCGv dst)
190 {
191     TCGv t0 = tcg_temp_new();
192     tcg_gen_shli_i32(t0, cpu_sr_q, SR_Q);
193     tcg_gen_or_i32(dst, dst, t0);
194     tcg_gen_shli_i32(t0, cpu_sr_m, SR_M);
195     tcg_gen_or_i32(dst, dst, t0);
196     tcg_gen_shli_i32(t0, cpu_sr_t, SR_T);
197     tcg_gen_or_i32(dst, cpu_sr, t0);
198     tcg_temp_free_i32(t0);
199 }
200 
201 static void gen_write_sr(TCGv src)
202 {
203     tcg_gen_andi_i32(cpu_sr, src,
204                      ~((1u << SR_Q) | (1u << SR_M) | (1u << SR_T)));
205     tcg_gen_extract_i32(cpu_sr_q, src, SR_Q, 1);
206     tcg_gen_extract_i32(cpu_sr_m, src, SR_M, 1);
207     tcg_gen_extract_i32(cpu_sr_t, src, SR_T, 1);
208 }
209 
210 static inline void gen_save_cpu_state(DisasContext *ctx, bool save_pc)
211 {
212     if (save_pc) {
213         tcg_gen_movi_i32(cpu_pc, ctx->base.pc_next);
214     }
215     if (ctx->delayed_pc != (uint32_t) -1) {
216         tcg_gen_movi_i32(cpu_delayed_pc, ctx->delayed_pc);
217     }
218     if ((ctx->tbflags & TB_FLAG_ENVFLAGS_MASK) != ctx->envflags) {
219         tcg_gen_movi_i32(cpu_flags, ctx->envflags);
220     }
221 }
222 
223 static inline bool use_exit_tb(DisasContext *ctx)
224 {
225     return (ctx->tbflags & GUSA_EXCLUSIVE) != 0;
226 }
227 
228 static inline bool use_goto_tb(DisasContext *ctx, target_ulong dest)
229 {
230     /* Use a direct jump if in same page and singlestep not enabled */
231     if (unlikely(ctx->base.singlestep_enabled || use_exit_tb(ctx))) {
232         return false;
233     }
234 #ifndef CONFIG_USER_ONLY
235     return (ctx->base.tb->pc & TARGET_PAGE_MASK) == (dest & TARGET_PAGE_MASK);
236 #else
237     return true;
238 #endif
239 }
240 
241 static void gen_goto_tb(DisasContext *ctx, int n, target_ulong dest)
242 {
243     if (use_goto_tb(ctx, dest)) {
244         tcg_gen_goto_tb(n);
245         tcg_gen_movi_i32(cpu_pc, dest);
246         tcg_gen_exit_tb(ctx->base.tb, n);
247     } else {
248         tcg_gen_movi_i32(cpu_pc, dest);
249         if (ctx->base.singlestep_enabled) {
250             gen_helper_debug(cpu_env);
251         } else if (use_exit_tb(ctx)) {
252             tcg_gen_exit_tb(NULL, 0);
253         } else {
254             tcg_gen_lookup_and_goto_ptr();
255         }
256     }
257     ctx->base.is_jmp = DISAS_NORETURN;
258 }
259 
260 static void gen_jump(DisasContext * ctx)
261 {
262     if (ctx->delayed_pc == -1) {
263 	/* Target is not statically known, it comes necessarily from a
264 	   delayed jump as immediate jump are conditinal jumps */
265 	tcg_gen_mov_i32(cpu_pc, cpu_delayed_pc);
266         tcg_gen_discard_i32(cpu_delayed_pc);
267         if (ctx->base.singlestep_enabled) {
268             gen_helper_debug(cpu_env);
269         } else if (use_exit_tb(ctx)) {
270             tcg_gen_exit_tb(NULL, 0);
271         } else {
272             tcg_gen_lookup_and_goto_ptr();
273         }
274         ctx->base.is_jmp = DISAS_NORETURN;
275     } else {
276 	gen_goto_tb(ctx, 0, ctx->delayed_pc);
277     }
278 }
279 
280 /* Immediate conditional jump (bt or bf) */
281 static void gen_conditional_jump(DisasContext *ctx, target_ulong dest,
282                                  bool jump_if_true)
283 {
284     TCGLabel *l1 = gen_new_label();
285     TCGCond cond_not_taken = jump_if_true ? TCG_COND_EQ : TCG_COND_NE;
286 
287     if (ctx->tbflags & GUSA_EXCLUSIVE) {
288         /* When in an exclusive region, we must continue to the end.
289            Therefore, exit the region on a taken branch, but otherwise
290            fall through to the next instruction.  */
291         tcg_gen_brcondi_i32(cond_not_taken, cpu_sr_t, 0, l1);
292         tcg_gen_movi_i32(cpu_flags, ctx->envflags & ~GUSA_MASK);
293         /* Note that this won't actually use a goto_tb opcode because we
294            disallow it in use_goto_tb, but it handles exit + singlestep.  */
295         gen_goto_tb(ctx, 0, dest);
296         gen_set_label(l1);
297         ctx->base.is_jmp = DISAS_NEXT;
298         return;
299     }
300 
301     gen_save_cpu_state(ctx, false);
302     tcg_gen_brcondi_i32(cond_not_taken, cpu_sr_t, 0, l1);
303     gen_goto_tb(ctx, 0, dest);
304     gen_set_label(l1);
305     gen_goto_tb(ctx, 1, ctx->base.pc_next + 2);
306     ctx->base.is_jmp = DISAS_NORETURN;
307 }
308 
309 /* Delayed conditional jump (bt or bf) */
310 static void gen_delayed_conditional_jump(DisasContext * ctx)
311 {
312     TCGLabel *l1 = gen_new_label();
313     TCGv ds = tcg_temp_new();
314 
315     tcg_gen_mov_i32(ds, cpu_delayed_cond);
316     tcg_gen_discard_i32(cpu_delayed_cond);
317 
318     if (ctx->tbflags & GUSA_EXCLUSIVE) {
319         /* When in an exclusive region, we must continue to the end.
320            Therefore, exit the region on a taken branch, but otherwise
321            fall through to the next instruction.  */
322         tcg_gen_brcondi_i32(TCG_COND_EQ, ds, 0, l1);
323 
324         /* Leave the gUSA region.  */
325         tcg_gen_movi_i32(cpu_flags, ctx->envflags & ~GUSA_MASK);
326         gen_jump(ctx);
327 
328         gen_set_label(l1);
329         ctx->base.is_jmp = DISAS_NEXT;
330         return;
331     }
332 
333     tcg_gen_brcondi_i32(TCG_COND_NE, ds, 0, l1);
334     gen_goto_tb(ctx, 1, ctx->base.pc_next + 2);
335     gen_set_label(l1);
336     gen_jump(ctx);
337 }
338 
339 static inline void gen_load_fpr64(DisasContext *ctx, TCGv_i64 t, int reg)
340 {
341     /* We have already signaled illegal instruction for odd Dr.  */
342     tcg_debug_assert((reg & 1) == 0);
343     reg ^= ctx->fbank;
344     tcg_gen_concat_i32_i64(t, cpu_fregs[reg + 1], cpu_fregs[reg]);
345 }
346 
347 static inline void gen_store_fpr64(DisasContext *ctx, TCGv_i64 t, int reg)
348 {
349     /* We have already signaled illegal instruction for odd Dr.  */
350     tcg_debug_assert((reg & 1) == 0);
351     reg ^= ctx->fbank;
352     tcg_gen_extr_i64_i32(cpu_fregs[reg + 1], cpu_fregs[reg], t);
353 }
354 
355 #define B3_0 (ctx->opcode & 0xf)
356 #define B6_4 ((ctx->opcode >> 4) & 0x7)
357 #define B7_4 ((ctx->opcode >> 4) & 0xf)
358 #define B7_0 (ctx->opcode & 0xff)
359 #define B7_0s ((int32_t) (int8_t) (ctx->opcode & 0xff))
360 #define B11_0s (ctx->opcode & 0x800 ? 0xfffff000 | (ctx->opcode & 0xfff) : \
361   (ctx->opcode & 0xfff))
362 #define B11_8 ((ctx->opcode >> 8) & 0xf)
363 #define B15_12 ((ctx->opcode >> 12) & 0xf)
364 
365 #define REG(x)     cpu_gregs[(x) ^ ctx->gbank]
366 #define ALTREG(x)  cpu_gregs[(x) ^ ctx->gbank ^ 0x10]
367 #define FREG(x)    cpu_fregs[(x) ^ ctx->fbank]
368 
369 #define XHACK(x) ((((x) & 1 ) << 4) | ((x) & 0xe))
370 
371 #define CHECK_NOT_DELAY_SLOT \
372     if (ctx->envflags & DELAY_SLOT_MASK) {  \
373         goto do_illegal_slot;               \
374     }
375 
376 #define CHECK_PRIVILEGED \
377     if (IS_USER(ctx)) {                     \
378         goto do_illegal;                    \
379     }
380 
381 #define CHECK_FPU_ENABLED \
382     if (ctx->tbflags & (1u << SR_FD)) {     \
383         goto do_fpu_disabled;               \
384     }
385 
386 #define CHECK_FPSCR_PR_0 \
387     if (ctx->tbflags & FPSCR_PR) {          \
388         goto do_illegal;                    \
389     }
390 
391 #define CHECK_FPSCR_PR_1 \
392     if (!(ctx->tbflags & FPSCR_PR)) {       \
393         goto do_illegal;                    \
394     }
395 
396 #define CHECK_SH4A \
397     if (!(ctx->features & SH_FEATURE_SH4A)) { \
398         goto do_illegal;                      \
399     }
400 
401 static void _decode_opc(DisasContext * ctx)
402 {
403     /* This code tries to make movcal emulation sufficiently
404        accurate for Linux purposes.  This instruction writes
405        memory, and prior to that, always allocates a cache line.
406        It is used in two contexts:
407        - in memcpy, where data is copied in blocks, the first write
408        of to a block uses movca.l for performance.
409        - in arch/sh/mm/cache-sh4.c, movcal.l + ocbi combination is used
410        to flush the cache. Here, the data written by movcal.l is never
411        written to memory, and the data written is just bogus.
412 
413        To simulate this, we simulate movcal.l, we store the value to memory,
414        but we also remember the previous content. If we see ocbi, we check
415        if movcal.l for that address was done previously. If so, the write should
416        not have hit the memory, so we restore the previous content.
417        When we see an instruction that is neither movca.l
418        nor ocbi, the previous content is discarded.
419 
420        To optimize, we only try to flush stores when we're at the start of
421        TB, or if we already saw movca.l in this TB and did not flush stores
422        yet.  */
423     if (ctx->has_movcal)
424 	{
425 	  int opcode = ctx->opcode & 0xf0ff;
426 	  if (opcode != 0x0093 /* ocbi */
427 	      && opcode != 0x00c3 /* movca.l */)
428 	      {
429                   gen_helper_discard_movcal_backup(cpu_env);
430 		  ctx->has_movcal = 0;
431 	      }
432 	}
433 
434 #if 0
435     fprintf(stderr, "Translating opcode 0x%04x\n", ctx->opcode);
436 #endif
437 
438     switch (ctx->opcode) {
439     case 0x0019:		/* div0u */
440         tcg_gen_movi_i32(cpu_sr_m, 0);
441         tcg_gen_movi_i32(cpu_sr_q, 0);
442         tcg_gen_movi_i32(cpu_sr_t, 0);
443 	return;
444     case 0x000b:		/* rts */
445 	CHECK_NOT_DELAY_SLOT
446 	tcg_gen_mov_i32(cpu_delayed_pc, cpu_pr);
447         ctx->envflags |= DELAY_SLOT;
448 	ctx->delayed_pc = (uint32_t) - 1;
449 	return;
450     case 0x0028:		/* clrmac */
451 	tcg_gen_movi_i32(cpu_mach, 0);
452 	tcg_gen_movi_i32(cpu_macl, 0);
453 	return;
454     case 0x0048:		/* clrs */
455         tcg_gen_andi_i32(cpu_sr, cpu_sr, ~(1u << SR_S));
456 	return;
457     case 0x0008:		/* clrt */
458         tcg_gen_movi_i32(cpu_sr_t, 0);
459 	return;
460     case 0x0038:		/* ldtlb */
461 	CHECK_PRIVILEGED
462         gen_helper_ldtlb(cpu_env);
463 	return;
464     case 0x002b:		/* rte */
465 	CHECK_PRIVILEGED
466 	CHECK_NOT_DELAY_SLOT
467         gen_write_sr(cpu_ssr);
468 	tcg_gen_mov_i32(cpu_delayed_pc, cpu_spc);
469         ctx->envflags |= DELAY_SLOT_RTE;
470 	ctx->delayed_pc = (uint32_t) - 1;
471         ctx->base.is_jmp = DISAS_STOP;
472 	return;
473     case 0x0058:		/* sets */
474         tcg_gen_ori_i32(cpu_sr, cpu_sr, (1u << SR_S));
475 	return;
476     case 0x0018:		/* sett */
477         tcg_gen_movi_i32(cpu_sr_t, 1);
478 	return;
479     case 0xfbfd:		/* frchg */
480         CHECK_FPSCR_PR_0
481 	tcg_gen_xori_i32(cpu_fpscr, cpu_fpscr, FPSCR_FR);
482         ctx->base.is_jmp = DISAS_STOP;
483 	return;
484     case 0xf3fd:		/* fschg */
485         CHECK_FPSCR_PR_0
486         tcg_gen_xori_i32(cpu_fpscr, cpu_fpscr, FPSCR_SZ);
487         ctx->base.is_jmp = DISAS_STOP;
488 	return;
489     case 0xf7fd:                /* fpchg */
490         CHECK_SH4A
491         tcg_gen_xori_i32(cpu_fpscr, cpu_fpscr, FPSCR_PR);
492         ctx->base.is_jmp = DISAS_STOP;
493         return;
494     case 0x0009:		/* nop */
495 	return;
496     case 0x001b:		/* sleep */
497 	CHECK_PRIVILEGED
498         tcg_gen_movi_i32(cpu_pc, ctx->base.pc_next + 2);
499         gen_helper_sleep(cpu_env);
500 	return;
501     }
502 
503     switch (ctx->opcode & 0xf000) {
504     case 0x1000:		/* mov.l Rm,@(disp,Rn) */
505 	{
506 	    TCGv addr = tcg_temp_new();
507 	    tcg_gen_addi_i32(addr, REG(B11_8), B3_0 * 4);
508             tcg_gen_qemu_st_i32(REG(B7_4), addr, ctx->memidx, MO_TEUL);
509 	    tcg_temp_free(addr);
510 	}
511 	return;
512     case 0x5000:		/* mov.l @(disp,Rm),Rn */
513 	{
514 	    TCGv addr = tcg_temp_new();
515 	    tcg_gen_addi_i32(addr, REG(B7_4), B3_0 * 4);
516             tcg_gen_qemu_ld_i32(REG(B11_8), addr, ctx->memidx, MO_TESL);
517 	    tcg_temp_free(addr);
518 	}
519 	return;
520     case 0xe000:		/* mov #imm,Rn */
521 #ifdef CONFIG_USER_ONLY
522         /* Detect the start of a gUSA region.  If so, update envflags
523            and end the TB.  This will allow us to see the end of the
524            region (stored in R0) in the next TB.  */
525         if (B11_8 == 15 && B7_0s < 0 &&
526             (tb_cflags(ctx->base.tb) & CF_PARALLEL)) {
527             ctx->envflags = deposit32(ctx->envflags, GUSA_SHIFT, 8, B7_0s);
528             ctx->base.is_jmp = DISAS_STOP;
529         }
530 #endif
531 	tcg_gen_movi_i32(REG(B11_8), B7_0s);
532 	return;
533     case 0x9000:		/* mov.w @(disp,PC),Rn */
534 	{
535             TCGv addr = tcg_const_i32(ctx->base.pc_next + 4 + B7_0 * 2);
536             tcg_gen_qemu_ld_i32(REG(B11_8), addr, ctx->memidx, MO_TESW);
537 	    tcg_temp_free(addr);
538 	}
539 	return;
540     case 0xd000:		/* mov.l @(disp,PC),Rn */
541 	{
542             TCGv addr = tcg_const_i32((ctx->base.pc_next + 4 + B7_0 * 4) & ~3);
543             tcg_gen_qemu_ld_i32(REG(B11_8), addr, ctx->memidx, MO_TESL);
544 	    tcg_temp_free(addr);
545 	}
546 	return;
547     case 0x7000:		/* add #imm,Rn */
548 	tcg_gen_addi_i32(REG(B11_8), REG(B11_8), B7_0s);
549 	return;
550     case 0xa000:		/* bra disp */
551 	CHECK_NOT_DELAY_SLOT
552         ctx->delayed_pc = ctx->base.pc_next + 4 + B11_0s * 2;
553         ctx->envflags |= DELAY_SLOT;
554 	return;
555     case 0xb000:		/* bsr disp */
556 	CHECK_NOT_DELAY_SLOT
557         tcg_gen_movi_i32(cpu_pr, ctx->base.pc_next + 4);
558         ctx->delayed_pc = ctx->base.pc_next + 4 + B11_0s * 2;
559         ctx->envflags |= DELAY_SLOT;
560 	return;
561     }
562 
563     switch (ctx->opcode & 0xf00f) {
564     case 0x6003:		/* mov Rm,Rn */
565 	tcg_gen_mov_i32(REG(B11_8), REG(B7_4));
566 	return;
567     case 0x2000:		/* mov.b Rm,@Rn */
568         tcg_gen_qemu_st_i32(REG(B7_4), REG(B11_8), ctx->memidx, MO_UB);
569 	return;
570     case 0x2001:		/* mov.w Rm,@Rn */
571         tcg_gen_qemu_st_i32(REG(B7_4), REG(B11_8), ctx->memidx, MO_TEUW);
572 	return;
573     case 0x2002:		/* mov.l Rm,@Rn */
574         tcg_gen_qemu_st_i32(REG(B7_4), REG(B11_8), ctx->memidx, MO_TEUL);
575 	return;
576     case 0x6000:		/* mov.b @Rm,Rn */
577         tcg_gen_qemu_ld_i32(REG(B11_8), REG(B7_4), ctx->memidx, MO_SB);
578 	return;
579     case 0x6001:		/* mov.w @Rm,Rn */
580         tcg_gen_qemu_ld_i32(REG(B11_8), REG(B7_4), ctx->memidx, MO_TESW);
581 	return;
582     case 0x6002:		/* mov.l @Rm,Rn */
583         tcg_gen_qemu_ld_i32(REG(B11_8), REG(B7_4), ctx->memidx, MO_TESL);
584 	return;
585     case 0x2004:		/* mov.b Rm,@-Rn */
586 	{
587 	    TCGv addr = tcg_temp_new();
588 	    tcg_gen_subi_i32(addr, REG(B11_8), 1);
589             /* might cause re-execution */
590             tcg_gen_qemu_st_i32(REG(B7_4), addr, ctx->memidx, MO_UB);
591 	    tcg_gen_mov_i32(REG(B11_8), addr);			/* modify register status */
592 	    tcg_temp_free(addr);
593 	}
594 	return;
595     case 0x2005:		/* mov.w Rm,@-Rn */
596 	{
597 	    TCGv addr = tcg_temp_new();
598 	    tcg_gen_subi_i32(addr, REG(B11_8), 2);
599             tcg_gen_qemu_st_i32(REG(B7_4), addr, ctx->memidx, MO_TEUW);
600 	    tcg_gen_mov_i32(REG(B11_8), addr);
601 	    tcg_temp_free(addr);
602 	}
603 	return;
604     case 0x2006:		/* mov.l Rm,@-Rn */
605 	{
606 	    TCGv addr = tcg_temp_new();
607 	    tcg_gen_subi_i32(addr, REG(B11_8), 4);
608             tcg_gen_qemu_st_i32(REG(B7_4), addr, ctx->memidx, MO_TEUL);
609 	    tcg_gen_mov_i32(REG(B11_8), addr);
610         tcg_temp_free(addr);
611 	}
612 	return;
613     case 0x6004:		/* mov.b @Rm+,Rn */
614         tcg_gen_qemu_ld_i32(REG(B11_8), REG(B7_4), ctx->memidx, MO_SB);
615 	if ( B11_8 != B7_4 )
616 		tcg_gen_addi_i32(REG(B7_4), REG(B7_4), 1);
617 	return;
618     case 0x6005:		/* mov.w @Rm+,Rn */
619         tcg_gen_qemu_ld_i32(REG(B11_8), REG(B7_4), ctx->memidx, MO_TESW);
620 	if ( B11_8 != B7_4 )
621 		tcg_gen_addi_i32(REG(B7_4), REG(B7_4), 2);
622 	return;
623     case 0x6006:		/* mov.l @Rm+,Rn */
624         tcg_gen_qemu_ld_i32(REG(B11_8), REG(B7_4), ctx->memidx, MO_TESL);
625 	if ( B11_8 != B7_4 )
626 		tcg_gen_addi_i32(REG(B7_4), REG(B7_4), 4);
627 	return;
628     case 0x0004:		/* mov.b Rm,@(R0,Rn) */
629 	{
630 	    TCGv addr = tcg_temp_new();
631 	    tcg_gen_add_i32(addr, REG(B11_8), REG(0));
632             tcg_gen_qemu_st_i32(REG(B7_4), addr, ctx->memidx, MO_UB);
633 	    tcg_temp_free(addr);
634 	}
635 	return;
636     case 0x0005:		/* mov.w Rm,@(R0,Rn) */
637 	{
638 	    TCGv addr = tcg_temp_new();
639 	    tcg_gen_add_i32(addr, REG(B11_8), REG(0));
640             tcg_gen_qemu_st_i32(REG(B7_4), addr, ctx->memidx, MO_TEUW);
641 	    tcg_temp_free(addr);
642 	}
643 	return;
644     case 0x0006:		/* mov.l Rm,@(R0,Rn) */
645 	{
646 	    TCGv addr = tcg_temp_new();
647 	    tcg_gen_add_i32(addr, REG(B11_8), REG(0));
648             tcg_gen_qemu_st_i32(REG(B7_4), addr, ctx->memidx, MO_TEUL);
649 	    tcg_temp_free(addr);
650 	}
651 	return;
652     case 0x000c:		/* mov.b @(R0,Rm),Rn */
653 	{
654 	    TCGv addr = tcg_temp_new();
655 	    tcg_gen_add_i32(addr, REG(B7_4), REG(0));
656             tcg_gen_qemu_ld_i32(REG(B11_8), addr, ctx->memidx, MO_SB);
657 	    tcg_temp_free(addr);
658 	}
659 	return;
660     case 0x000d:		/* mov.w @(R0,Rm),Rn */
661 	{
662 	    TCGv addr = tcg_temp_new();
663 	    tcg_gen_add_i32(addr, REG(B7_4), REG(0));
664             tcg_gen_qemu_ld_i32(REG(B11_8), addr, ctx->memidx, MO_TESW);
665 	    tcg_temp_free(addr);
666 	}
667 	return;
668     case 0x000e:		/* mov.l @(R0,Rm),Rn */
669 	{
670 	    TCGv addr = tcg_temp_new();
671 	    tcg_gen_add_i32(addr, REG(B7_4), REG(0));
672             tcg_gen_qemu_ld_i32(REG(B11_8), addr, ctx->memidx, MO_TESL);
673 	    tcg_temp_free(addr);
674 	}
675 	return;
676     case 0x6008:		/* swap.b Rm,Rn */
677 	{
678             TCGv low = tcg_temp_new();
679             tcg_gen_bswap16_i32(low, REG(B7_4), 0);
680             tcg_gen_deposit_i32(REG(B11_8), REG(B7_4), low, 0, 16);
681 	    tcg_temp_free(low);
682 	}
683 	return;
684     case 0x6009:		/* swap.w Rm,Rn */
685         tcg_gen_rotli_i32(REG(B11_8), REG(B7_4), 16);
686 	return;
687     case 0x200d:		/* xtrct Rm,Rn */
688 	{
689 	    TCGv high, low;
690 	    high = tcg_temp_new();
691 	    tcg_gen_shli_i32(high, REG(B7_4), 16);
692 	    low = tcg_temp_new();
693 	    tcg_gen_shri_i32(low, REG(B11_8), 16);
694 	    tcg_gen_or_i32(REG(B11_8), high, low);
695 	    tcg_temp_free(low);
696 	    tcg_temp_free(high);
697 	}
698 	return;
699     case 0x300c:		/* add Rm,Rn */
700 	tcg_gen_add_i32(REG(B11_8), REG(B11_8), REG(B7_4));
701 	return;
702     case 0x300e:		/* addc Rm,Rn */
703         {
704             TCGv t0, t1;
705             t0 = tcg_const_tl(0);
706             t1 = tcg_temp_new();
707             tcg_gen_add2_i32(t1, cpu_sr_t, cpu_sr_t, t0, REG(B7_4), t0);
708             tcg_gen_add2_i32(REG(B11_8), cpu_sr_t,
709                              REG(B11_8), t0, t1, cpu_sr_t);
710             tcg_temp_free(t0);
711             tcg_temp_free(t1);
712         }
713 	return;
714     case 0x300f:		/* addv Rm,Rn */
715         {
716             TCGv t0, t1, t2;
717             t0 = tcg_temp_new();
718             tcg_gen_add_i32(t0, REG(B7_4), REG(B11_8));
719             t1 = tcg_temp_new();
720             tcg_gen_xor_i32(t1, t0, REG(B11_8));
721             t2 = tcg_temp_new();
722             tcg_gen_xor_i32(t2, REG(B7_4), REG(B11_8));
723             tcg_gen_andc_i32(cpu_sr_t, t1, t2);
724             tcg_temp_free(t2);
725             tcg_gen_shri_i32(cpu_sr_t, cpu_sr_t, 31);
726             tcg_temp_free(t1);
727             tcg_gen_mov_i32(REG(B7_4), t0);
728             tcg_temp_free(t0);
729         }
730 	return;
731     case 0x2009:		/* and Rm,Rn */
732 	tcg_gen_and_i32(REG(B11_8), REG(B11_8), REG(B7_4));
733 	return;
734     case 0x3000:		/* cmp/eq Rm,Rn */
735         tcg_gen_setcond_i32(TCG_COND_EQ, cpu_sr_t, REG(B11_8), REG(B7_4));
736 	return;
737     case 0x3003:		/* cmp/ge Rm,Rn */
738         tcg_gen_setcond_i32(TCG_COND_GE, cpu_sr_t, REG(B11_8), REG(B7_4));
739 	return;
740     case 0x3007:		/* cmp/gt Rm,Rn */
741         tcg_gen_setcond_i32(TCG_COND_GT, cpu_sr_t, REG(B11_8), REG(B7_4));
742 	return;
743     case 0x3006:		/* cmp/hi Rm,Rn */
744         tcg_gen_setcond_i32(TCG_COND_GTU, cpu_sr_t, REG(B11_8), REG(B7_4));
745 	return;
746     case 0x3002:		/* cmp/hs Rm,Rn */
747         tcg_gen_setcond_i32(TCG_COND_GEU, cpu_sr_t, REG(B11_8), REG(B7_4));
748 	return;
749     case 0x200c:		/* cmp/str Rm,Rn */
750 	{
751 	    TCGv cmp1 = tcg_temp_new();
752 	    TCGv cmp2 = tcg_temp_new();
753             tcg_gen_xor_i32(cmp2, REG(B7_4), REG(B11_8));
754             tcg_gen_subi_i32(cmp1, cmp2, 0x01010101);
755             tcg_gen_andc_i32(cmp1, cmp1, cmp2);
756             tcg_gen_andi_i32(cmp1, cmp1, 0x80808080);
757             tcg_gen_setcondi_i32(TCG_COND_NE, cpu_sr_t, cmp1, 0);
758 	    tcg_temp_free(cmp2);
759 	    tcg_temp_free(cmp1);
760 	}
761 	return;
762     case 0x2007:		/* div0s Rm,Rn */
763         tcg_gen_shri_i32(cpu_sr_q, REG(B11_8), 31);         /* SR_Q */
764         tcg_gen_shri_i32(cpu_sr_m, REG(B7_4), 31);          /* SR_M */
765         tcg_gen_xor_i32(cpu_sr_t, cpu_sr_q, cpu_sr_m);      /* SR_T */
766 	return;
767     case 0x3004:		/* div1 Rm,Rn */
768         {
769             TCGv t0 = tcg_temp_new();
770             TCGv t1 = tcg_temp_new();
771             TCGv t2 = tcg_temp_new();
772             TCGv zero = tcg_const_i32(0);
773 
774             /* shift left arg1, saving the bit being pushed out and inserting
775                T on the right */
776             tcg_gen_shri_i32(t0, REG(B11_8), 31);
777             tcg_gen_shli_i32(REG(B11_8), REG(B11_8), 1);
778             tcg_gen_or_i32(REG(B11_8), REG(B11_8), cpu_sr_t);
779 
780             /* Add or subtract arg0 from arg1 depending if Q == M. To avoid
781                using 64-bit temps, we compute arg0's high part from q ^ m, so
782                that it is 0x00000000 when adding the value or 0xffffffff when
783                subtracting it. */
784             tcg_gen_xor_i32(t1, cpu_sr_q, cpu_sr_m);
785             tcg_gen_subi_i32(t1, t1, 1);
786             tcg_gen_neg_i32(t2, REG(B7_4));
787             tcg_gen_movcond_i32(TCG_COND_EQ, t2, t1, zero, REG(B7_4), t2);
788             tcg_gen_add2_i32(REG(B11_8), t1, REG(B11_8), zero, t2, t1);
789 
790             /* compute T and Q depending on carry */
791             tcg_gen_andi_i32(t1, t1, 1);
792             tcg_gen_xor_i32(t1, t1, t0);
793             tcg_gen_xori_i32(cpu_sr_t, t1, 1);
794             tcg_gen_xor_i32(cpu_sr_q, cpu_sr_m, t1);
795 
796             tcg_temp_free(zero);
797             tcg_temp_free(t2);
798             tcg_temp_free(t1);
799             tcg_temp_free(t0);
800         }
801 	return;
802     case 0x300d:		/* dmuls.l Rm,Rn */
803         tcg_gen_muls2_i32(cpu_macl, cpu_mach, REG(B7_4), REG(B11_8));
804 	return;
805     case 0x3005:		/* dmulu.l Rm,Rn */
806         tcg_gen_mulu2_i32(cpu_macl, cpu_mach, REG(B7_4), REG(B11_8));
807 	return;
808     case 0x600e:		/* exts.b Rm,Rn */
809 	tcg_gen_ext8s_i32(REG(B11_8), REG(B7_4));
810 	return;
811     case 0x600f:		/* exts.w Rm,Rn */
812 	tcg_gen_ext16s_i32(REG(B11_8), REG(B7_4));
813 	return;
814     case 0x600c:		/* extu.b Rm,Rn */
815 	tcg_gen_ext8u_i32(REG(B11_8), REG(B7_4));
816 	return;
817     case 0x600d:		/* extu.w Rm,Rn */
818 	tcg_gen_ext16u_i32(REG(B11_8), REG(B7_4));
819 	return;
820     case 0x000f:		/* mac.l @Rm+,@Rn+ */
821 	{
822 	    TCGv arg0, arg1;
823 	    arg0 = tcg_temp_new();
824             tcg_gen_qemu_ld_i32(arg0, REG(B7_4), ctx->memidx, MO_TESL);
825 	    arg1 = tcg_temp_new();
826             tcg_gen_qemu_ld_i32(arg1, REG(B11_8), ctx->memidx, MO_TESL);
827             gen_helper_macl(cpu_env, arg0, arg1);
828 	    tcg_temp_free(arg1);
829 	    tcg_temp_free(arg0);
830 	    tcg_gen_addi_i32(REG(B7_4), REG(B7_4), 4);
831 	    tcg_gen_addi_i32(REG(B11_8), REG(B11_8), 4);
832 	}
833 	return;
834     case 0x400f:		/* mac.w @Rm+,@Rn+ */
835 	{
836 	    TCGv arg0, arg1;
837 	    arg0 = tcg_temp_new();
838             tcg_gen_qemu_ld_i32(arg0, REG(B7_4), ctx->memidx, MO_TESL);
839 	    arg1 = tcg_temp_new();
840             tcg_gen_qemu_ld_i32(arg1, REG(B11_8), ctx->memidx, MO_TESL);
841             gen_helper_macw(cpu_env, arg0, arg1);
842 	    tcg_temp_free(arg1);
843 	    tcg_temp_free(arg0);
844 	    tcg_gen_addi_i32(REG(B11_8), REG(B11_8), 2);
845 	    tcg_gen_addi_i32(REG(B7_4), REG(B7_4), 2);
846 	}
847 	return;
848     case 0x0007:		/* mul.l Rm,Rn */
849 	tcg_gen_mul_i32(cpu_macl, REG(B7_4), REG(B11_8));
850 	return;
851     case 0x200f:		/* muls.w Rm,Rn */
852 	{
853 	    TCGv arg0, arg1;
854 	    arg0 = tcg_temp_new();
855 	    tcg_gen_ext16s_i32(arg0, REG(B7_4));
856 	    arg1 = tcg_temp_new();
857 	    tcg_gen_ext16s_i32(arg1, REG(B11_8));
858 	    tcg_gen_mul_i32(cpu_macl, arg0, arg1);
859 	    tcg_temp_free(arg1);
860 	    tcg_temp_free(arg0);
861 	}
862 	return;
863     case 0x200e:		/* mulu.w Rm,Rn */
864 	{
865 	    TCGv arg0, arg1;
866 	    arg0 = tcg_temp_new();
867 	    tcg_gen_ext16u_i32(arg0, REG(B7_4));
868 	    arg1 = tcg_temp_new();
869 	    tcg_gen_ext16u_i32(arg1, REG(B11_8));
870 	    tcg_gen_mul_i32(cpu_macl, arg0, arg1);
871 	    tcg_temp_free(arg1);
872 	    tcg_temp_free(arg0);
873 	}
874 	return;
875     case 0x600b:		/* neg Rm,Rn */
876 	tcg_gen_neg_i32(REG(B11_8), REG(B7_4));
877 	return;
878     case 0x600a:		/* negc Rm,Rn */
879         {
880             TCGv t0 = tcg_const_i32(0);
881             tcg_gen_add2_i32(REG(B11_8), cpu_sr_t,
882                              REG(B7_4), t0, cpu_sr_t, t0);
883             tcg_gen_sub2_i32(REG(B11_8), cpu_sr_t,
884                              t0, t0, REG(B11_8), cpu_sr_t);
885             tcg_gen_andi_i32(cpu_sr_t, cpu_sr_t, 1);
886             tcg_temp_free(t0);
887         }
888 	return;
889     case 0x6007:		/* not Rm,Rn */
890 	tcg_gen_not_i32(REG(B11_8), REG(B7_4));
891 	return;
892     case 0x200b:		/* or Rm,Rn */
893 	tcg_gen_or_i32(REG(B11_8), REG(B11_8), REG(B7_4));
894 	return;
895     case 0x400c:		/* shad Rm,Rn */
896 	{
897             TCGv t0 = tcg_temp_new();
898             TCGv t1 = tcg_temp_new();
899             TCGv t2 = tcg_temp_new();
900 
901             tcg_gen_andi_i32(t0, REG(B7_4), 0x1f);
902 
903             /* positive case: shift to the left */
904             tcg_gen_shl_i32(t1, REG(B11_8), t0);
905 
906             /* negative case: shift to the right in two steps to
907                correctly handle the -32 case */
908             tcg_gen_xori_i32(t0, t0, 0x1f);
909             tcg_gen_sar_i32(t2, REG(B11_8), t0);
910             tcg_gen_sari_i32(t2, t2, 1);
911 
912             /* select between the two cases */
913             tcg_gen_movi_i32(t0, 0);
914             tcg_gen_movcond_i32(TCG_COND_GE, REG(B11_8), REG(B7_4), t0, t1, t2);
915 
916             tcg_temp_free(t0);
917             tcg_temp_free(t1);
918             tcg_temp_free(t2);
919 	}
920 	return;
921     case 0x400d:		/* shld Rm,Rn */
922 	{
923             TCGv t0 = tcg_temp_new();
924             TCGv t1 = tcg_temp_new();
925             TCGv t2 = tcg_temp_new();
926 
927             tcg_gen_andi_i32(t0, REG(B7_4), 0x1f);
928 
929             /* positive case: shift to the left */
930             tcg_gen_shl_i32(t1, REG(B11_8), t0);
931 
932             /* negative case: shift to the right in two steps to
933                correctly handle the -32 case */
934             tcg_gen_xori_i32(t0, t0, 0x1f);
935             tcg_gen_shr_i32(t2, REG(B11_8), t0);
936             tcg_gen_shri_i32(t2, t2, 1);
937 
938             /* select between the two cases */
939             tcg_gen_movi_i32(t0, 0);
940             tcg_gen_movcond_i32(TCG_COND_GE, REG(B11_8), REG(B7_4), t0, t1, t2);
941 
942             tcg_temp_free(t0);
943             tcg_temp_free(t1);
944             tcg_temp_free(t2);
945 	}
946 	return;
947     case 0x3008:		/* sub Rm,Rn */
948 	tcg_gen_sub_i32(REG(B11_8), REG(B11_8), REG(B7_4));
949 	return;
950     case 0x300a:		/* subc Rm,Rn */
951         {
952             TCGv t0, t1;
953             t0 = tcg_const_tl(0);
954             t1 = tcg_temp_new();
955             tcg_gen_add2_i32(t1, cpu_sr_t, cpu_sr_t, t0, REG(B7_4), t0);
956             tcg_gen_sub2_i32(REG(B11_8), cpu_sr_t,
957                              REG(B11_8), t0, t1, cpu_sr_t);
958             tcg_gen_andi_i32(cpu_sr_t, cpu_sr_t, 1);
959             tcg_temp_free(t0);
960             tcg_temp_free(t1);
961         }
962 	return;
963     case 0x300b:		/* subv Rm,Rn */
964         {
965             TCGv t0, t1, t2;
966             t0 = tcg_temp_new();
967             tcg_gen_sub_i32(t0, REG(B11_8), REG(B7_4));
968             t1 = tcg_temp_new();
969             tcg_gen_xor_i32(t1, t0, REG(B7_4));
970             t2 = tcg_temp_new();
971             tcg_gen_xor_i32(t2, REG(B11_8), REG(B7_4));
972             tcg_gen_and_i32(t1, t1, t2);
973             tcg_temp_free(t2);
974             tcg_gen_shri_i32(cpu_sr_t, t1, 31);
975             tcg_temp_free(t1);
976             tcg_gen_mov_i32(REG(B11_8), t0);
977             tcg_temp_free(t0);
978         }
979 	return;
980     case 0x2008:		/* tst Rm,Rn */
981 	{
982 	    TCGv val = tcg_temp_new();
983 	    tcg_gen_and_i32(val, REG(B7_4), REG(B11_8));
984             tcg_gen_setcondi_i32(TCG_COND_EQ, cpu_sr_t, val, 0);
985 	    tcg_temp_free(val);
986 	}
987 	return;
988     case 0x200a:		/* xor Rm,Rn */
989 	tcg_gen_xor_i32(REG(B11_8), REG(B11_8), REG(B7_4));
990 	return;
991     case 0xf00c: /* fmov {F,D,X}Rm,{F,D,X}Rn - FPSCR: Nothing */
992 	CHECK_FPU_ENABLED
993         if (ctx->tbflags & FPSCR_SZ) {
994             int xsrc = XHACK(B7_4);
995             int xdst = XHACK(B11_8);
996             tcg_gen_mov_i32(FREG(xdst), FREG(xsrc));
997             tcg_gen_mov_i32(FREG(xdst + 1), FREG(xsrc + 1));
998 	} else {
999             tcg_gen_mov_i32(FREG(B11_8), FREG(B7_4));
1000 	}
1001 	return;
1002     case 0xf00a: /* fmov {F,D,X}Rm,@Rn - FPSCR: Nothing */
1003 	CHECK_FPU_ENABLED
1004         if (ctx->tbflags & FPSCR_SZ) {
1005             TCGv_i64 fp = tcg_temp_new_i64();
1006             gen_load_fpr64(ctx, fp, XHACK(B7_4));
1007             tcg_gen_qemu_st_i64(fp, REG(B11_8), ctx->memidx, MO_TEQ);
1008             tcg_temp_free_i64(fp);
1009 	} else {
1010             tcg_gen_qemu_st_i32(FREG(B7_4), REG(B11_8), ctx->memidx, MO_TEUL);
1011 	}
1012 	return;
1013     case 0xf008: /* fmov @Rm,{F,D,X}Rn - FPSCR: Nothing */
1014 	CHECK_FPU_ENABLED
1015         if (ctx->tbflags & FPSCR_SZ) {
1016             TCGv_i64 fp = tcg_temp_new_i64();
1017             tcg_gen_qemu_ld_i64(fp, REG(B7_4), ctx->memidx, MO_TEQ);
1018             gen_store_fpr64(ctx, fp, XHACK(B11_8));
1019             tcg_temp_free_i64(fp);
1020 	} else {
1021             tcg_gen_qemu_ld_i32(FREG(B11_8), REG(B7_4), ctx->memidx, MO_TEUL);
1022 	}
1023 	return;
1024     case 0xf009: /* fmov @Rm+,{F,D,X}Rn - FPSCR: Nothing */
1025 	CHECK_FPU_ENABLED
1026         if (ctx->tbflags & FPSCR_SZ) {
1027             TCGv_i64 fp = tcg_temp_new_i64();
1028             tcg_gen_qemu_ld_i64(fp, REG(B7_4), ctx->memidx, MO_TEQ);
1029             gen_store_fpr64(ctx, fp, XHACK(B11_8));
1030             tcg_temp_free_i64(fp);
1031             tcg_gen_addi_i32(REG(B7_4), REG(B7_4), 8);
1032 	} else {
1033             tcg_gen_qemu_ld_i32(FREG(B11_8), REG(B7_4), ctx->memidx, MO_TEUL);
1034 	    tcg_gen_addi_i32(REG(B7_4), REG(B7_4), 4);
1035 	}
1036 	return;
1037     case 0xf00b: /* fmov {F,D,X}Rm,@-Rn - FPSCR: Nothing */
1038 	CHECK_FPU_ENABLED
1039         {
1040             TCGv addr = tcg_temp_new_i32();
1041             if (ctx->tbflags & FPSCR_SZ) {
1042                 TCGv_i64 fp = tcg_temp_new_i64();
1043                 gen_load_fpr64(ctx, fp, XHACK(B7_4));
1044                 tcg_gen_subi_i32(addr, REG(B11_8), 8);
1045                 tcg_gen_qemu_st_i64(fp, addr, ctx->memidx, MO_TEQ);
1046                 tcg_temp_free_i64(fp);
1047             } else {
1048                 tcg_gen_subi_i32(addr, REG(B11_8), 4);
1049                 tcg_gen_qemu_st_i32(FREG(B7_4), addr, ctx->memidx, MO_TEUL);
1050             }
1051             tcg_gen_mov_i32(REG(B11_8), addr);
1052             tcg_temp_free(addr);
1053         }
1054 	return;
1055     case 0xf006: /* fmov @(R0,Rm),{F,D,X}Rm - FPSCR: Nothing */
1056 	CHECK_FPU_ENABLED
1057 	{
1058 	    TCGv addr = tcg_temp_new_i32();
1059 	    tcg_gen_add_i32(addr, REG(B7_4), REG(0));
1060             if (ctx->tbflags & FPSCR_SZ) {
1061                 TCGv_i64 fp = tcg_temp_new_i64();
1062                 tcg_gen_qemu_ld_i64(fp, addr, ctx->memidx, MO_TEQ);
1063                 gen_store_fpr64(ctx, fp, XHACK(B11_8));
1064                 tcg_temp_free_i64(fp);
1065 	    } else {
1066                 tcg_gen_qemu_ld_i32(FREG(B11_8), addr, ctx->memidx, MO_TEUL);
1067 	    }
1068 	    tcg_temp_free(addr);
1069 	}
1070 	return;
1071     case 0xf007: /* fmov {F,D,X}Rn,@(R0,Rn) - FPSCR: Nothing */
1072 	CHECK_FPU_ENABLED
1073 	{
1074 	    TCGv addr = tcg_temp_new();
1075 	    tcg_gen_add_i32(addr, REG(B11_8), REG(0));
1076             if (ctx->tbflags & FPSCR_SZ) {
1077                 TCGv_i64 fp = tcg_temp_new_i64();
1078                 gen_load_fpr64(ctx, fp, XHACK(B7_4));
1079                 tcg_gen_qemu_st_i64(fp, addr, ctx->memidx, MO_TEQ);
1080                 tcg_temp_free_i64(fp);
1081 	    } else {
1082                 tcg_gen_qemu_st_i32(FREG(B7_4), addr, ctx->memidx, MO_TEUL);
1083 	    }
1084 	    tcg_temp_free(addr);
1085 	}
1086 	return;
1087     case 0xf000: /* fadd Rm,Rn - FPSCR: R[PR,Enable.O/U/I]/W[Cause,Flag] */
1088     case 0xf001: /* fsub Rm,Rn - FPSCR: R[PR,Enable.O/U/I]/W[Cause,Flag] */
1089     case 0xf002: /* fmul Rm,Rn - FPSCR: R[PR,Enable.O/U/I]/W[Cause,Flag] */
1090     case 0xf003: /* fdiv Rm,Rn - FPSCR: R[PR,Enable.O/U/I]/W[Cause,Flag] */
1091     case 0xf004: /* fcmp/eq Rm,Rn - FPSCR: R[PR,Enable.V]/W[Cause,Flag] */
1092     case 0xf005: /* fcmp/gt Rm,Rn - FPSCR: R[PR,Enable.V]/W[Cause,Flag] */
1093 	{
1094 	    CHECK_FPU_ENABLED
1095             if (ctx->tbflags & FPSCR_PR) {
1096                 TCGv_i64 fp0, fp1;
1097 
1098                 if (ctx->opcode & 0x0110) {
1099                     goto do_illegal;
1100                 }
1101 		fp0 = tcg_temp_new_i64();
1102 		fp1 = tcg_temp_new_i64();
1103                 gen_load_fpr64(ctx, fp0, B11_8);
1104                 gen_load_fpr64(ctx, fp1, B7_4);
1105                 switch (ctx->opcode & 0xf00f) {
1106                 case 0xf000:		/* fadd Rm,Rn */
1107                     gen_helper_fadd_DT(fp0, cpu_env, fp0, fp1);
1108                     break;
1109                 case 0xf001:		/* fsub Rm,Rn */
1110                     gen_helper_fsub_DT(fp0, cpu_env, fp0, fp1);
1111                     break;
1112                 case 0xf002:		/* fmul Rm,Rn */
1113                     gen_helper_fmul_DT(fp0, cpu_env, fp0, fp1);
1114                     break;
1115                 case 0xf003:		/* fdiv Rm,Rn */
1116                     gen_helper_fdiv_DT(fp0, cpu_env, fp0, fp1);
1117                     break;
1118                 case 0xf004:		/* fcmp/eq Rm,Rn */
1119                     gen_helper_fcmp_eq_DT(cpu_sr_t, cpu_env, fp0, fp1);
1120                     return;
1121                 case 0xf005:		/* fcmp/gt Rm,Rn */
1122                     gen_helper_fcmp_gt_DT(cpu_sr_t, cpu_env, fp0, fp1);
1123                     return;
1124                 }
1125                 gen_store_fpr64(ctx, fp0, B11_8);
1126                 tcg_temp_free_i64(fp0);
1127                 tcg_temp_free_i64(fp1);
1128 	    } else {
1129                 switch (ctx->opcode & 0xf00f) {
1130                 case 0xf000:		/* fadd Rm,Rn */
1131                     gen_helper_fadd_FT(FREG(B11_8), cpu_env,
1132                                        FREG(B11_8), FREG(B7_4));
1133                     break;
1134                 case 0xf001:		/* fsub Rm,Rn */
1135                     gen_helper_fsub_FT(FREG(B11_8), cpu_env,
1136                                        FREG(B11_8), FREG(B7_4));
1137                     break;
1138                 case 0xf002:		/* fmul Rm,Rn */
1139                     gen_helper_fmul_FT(FREG(B11_8), cpu_env,
1140                                        FREG(B11_8), FREG(B7_4));
1141                     break;
1142                 case 0xf003:		/* fdiv Rm,Rn */
1143                     gen_helper_fdiv_FT(FREG(B11_8), cpu_env,
1144                                        FREG(B11_8), FREG(B7_4));
1145                     break;
1146                 case 0xf004:		/* fcmp/eq Rm,Rn */
1147                     gen_helper_fcmp_eq_FT(cpu_sr_t, cpu_env,
1148                                           FREG(B11_8), FREG(B7_4));
1149                     return;
1150                 case 0xf005:		/* fcmp/gt Rm,Rn */
1151                     gen_helper_fcmp_gt_FT(cpu_sr_t, cpu_env,
1152                                           FREG(B11_8), FREG(B7_4));
1153                     return;
1154                 }
1155 	    }
1156 	}
1157 	return;
1158     case 0xf00e: /* fmac FR0,RM,Rn */
1159         CHECK_FPU_ENABLED
1160         CHECK_FPSCR_PR_0
1161         gen_helper_fmac_FT(FREG(B11_8), cpu_env,
1162                            FREG(0), FREG(B7_4), FREG(B11_8));
1163         return;
1164     }
1165 
1166     switch (ctx->opcode & 0xff00) {
1167     case 0xc900:		/* and #imm,R0 */
1168 	tcg_gen_andi_i32(REG(0), REG(0), B7_0);
1169 	return;
1170     case 0xcd00:		/* and.b #imm,@(R0,GBR) */
1171 	{
1172 	    TCGv addr, val;
1173 	    addr = tcg_temp_new();
1174 	    tcg_gen_add_i32(addr, REG(0), cpu_gbr);
1175 	    val = tcg_temp_new();
1176             tcg_gen_qemu_ld_i32(val, addr, ctx->memidx, MO_UB);
1177 	    tcg_gen_andi_i32(val, val, B7_0);
1178             tcg_gen_qemu_st_i32(val, addr, ctx->memidx, MO_UB);
1179 	    tcg_temp_free(val);
1180 	    tcg_temp_free(addr);
1181 	}
1182 	return;
1183     case 0x8b00:		/* bf label */
1184 	CHECK_NOT_DELAY_SLOT
1185         gen_conditional_jump(ctx, ctx->base.pc_next + 4 + B7_0s * 2, false);
1186 	return;
1187     case 0x8f00:		/* bf/s label */
1188 	CHECK_NOT_DELAY_SLOT
1189         tcg_gen_xori_i32(cpu_delayed_cond, cpu_sr_t, 1);
1190         ctx->delayed_pc = ctx->base.pc_next + 4 + B7_0s * 2;
1191         ctx->envflags |= DELAY_SLOT_CONDITIONAL;
1192 	return;
1193     case 0x8900:		/* bt label */
1194 	CHECK_NOT_DELAY_SLOT
1195         gen_conditional_jump(ctx, ctx->base.pc_next + 4 + B7_0s * 2, true);
1196 	return;
1197     case 0x8d00:		/* bt/s label */
1198 	CHECK_NOT_DELAY_SLOT
1199         tcg_gen_mov_i32(cpu_delayed_cond, cpu_sr_t);
1200         ctx->delayed_pc = ctx->base.pc_next + 4 + B7_0s * 2;
1201         ctx->envflags |= DELAY_SLOT_CONDITIONAL;
1202 	return;
1203     case 0x8800:		/* cmp/eq #imm,R0 */
1204         tcg_gen_setcondi_i32(TCG_COND_EQ, cpu_sr_t, REG(0), B7_0s);
1205 	return;
1206     case 0xc400:		/* mov.b @(disp,GBR),R0 */
1207 	{
1208 	    TCGv addr = tcg_temp_new();
1209 	    tcg_gen_addi_i32(addr, cpu_gbr, B7_0);
1210             tcg_gen_qemu_ld_i32(REG(0), addr, ctx->memidx, MO_SB);
1211 	    tcg_temp_free(addr);
1212 	}
1213 	return;
1214     case 0xc500:		/* mov.w @(disp,GBR),R0 */
1215 	{
1216 	    TCGv addr = tcg_temp_new();
1217 	    tcg_gen_addi_i32(addr, cpu_gbr, B7_0 * 2);
1218             tcg_gen_qemu_ld_i32(REG(0), addr, ctx->memidx, MO_TESW);
1219 	    tcg_temp_free(addr);
1220 	}
1221 	return;
1222     case 0xc600:		/* mov.l @(disp,GBR),R0 */
1223 	{
1224 	    TCGv addr = tcg_temp_new();
1225 	    tcg_gen_addi_i32(addr, cpu_gbr, B7_0 * 4);
1226             tcg_gen_qemu_ld_i32(REG(0), addr, ctx->memidx, MO_TESL);
1227 	    tcg_temp_free(addr);
1228 	}
1229 	return;
1230     case 0xc000:		/* mov.b R0,@(disp,GBR) */
1231 	{
1232 	    TCGv addr = tcg_temp_new();
1233 	    tcg_gen_addi_i32(addr, cpu_gbr, B7_0);
1234             tcg_gen_qemu_st_i32(REG(0), addr, ctx->memidx, MO_UB);
1235 	    tcg_temp_free(addr);
1236 	}
1237 	return;
1238     case 0xc100:		/* mov.w R0,@(disp,GBR) */
1239 	{
1240 	    TCGv addr = tcg_temp_new();
1241 	    tcg_gen_addi_i32(addr, cpu_gbr, B7_0 * 2);
1242             tcg_gen_qemu_st_i32(REG(0), addr, ctx->memidx, MO_TEUW);
1243 	    tcg_temp_free(addr);
1244 	}
1245 	return;
1246     case 0xc200:		/* mov.l R0,@(disp,GBR) */
1247 	{
1248 	    TCGv addr = tcg_temp_new();
1249 	    tcg_gen_addi_i32(addr, cpu_gbr, B7_0 * 4);
1250             tcg_gen_qemu_st_i32(REG(0), addr, ctx->memidx, MO_TEUL);
1251 	    tcg_temp_free(addr);
1252 	}
1253 	return;
1254     case 0x8000:		/* mov.b R0,@(disp,Rn) */
1255 	{
1256 	    TCGv addr = tcg_temp_new();
1257 	    tcg_gen_addi_i32(addr, REG(B7_4), B3_0);
1258             tcg_gen_qemu_st_i32(REG(0), addr, ctx->memidx, MO_UB);
1259 	    tcg_temp_free(addr);
1260 	}
1261 	return;
1262     case 0x8100:		/* mov.w R0,@(disp,Rn) */
1263 	{
1264 	    TCGv addr = tcg_temp_new();
1265 	    tcg_gen_addi_i32(addr, REG(B7_4), B3_0 * 2);
1266             tcg_gen_qemu_st_i32(REG(0), addr, ctx->memidx, MO_TEUW);
1267 	    tcg_temp_free(addr);
1268 	}
1269 	return;
1270     case 0x8400:		/* mov.b @(disp,Rn),R0 */
1271 	{
1272 	    TCGv addr = tcg_temp_new();
1273 	    tcg_gen_addi_i32(addr, REG(B7_4), B3_0);
1274             tcg_gen_qemu_ld_i32(REG(0), addr, ctx->memidx, MO_SB);
1275 	    tcg_temp_free(addr);
1276 	}
1277 	return;
1278     case 0x8500:		/* mov.w @(disp,Rn),R0 */
1279 	{
1280 	    TCGv addr = tcg_temp_new();
1281 	    tcg_gen_addi_i32(addr, REG(B7_4), B3_0 * 2);
1282             tcg_gen_qemu_ld_i32(REG(0), addr, ctx->memidx, MO_TESW);
1283 	    tcg_temp_free(addr);
1284 	}
1285 	return;
1286     case 0xc700:		/* mova @(disp,PC),R0 */
1287         tcg_gen_movi_i32(REG(0), ((ctx->base.pc_next & 0xfffffffc) +
1288                                   4 + B7_0 * 4) & ~3);
1289 	return;
1290     case 0xcb00:		/* or #imm,R0 */
1291 	tcg_gen_ori_i32(REG(0), REG(0), B7_0);
1292 	return;
1293     case 0xcf00:		/* or.b #imm,@(R0,GBR) */
1294 	{
1295 	    TCGv addr, val;
1296 	    addr = tcg_temp_new();
1297 	    tcg_gen_add_i32(addr, REG(0), cpu_gbr);
1298 	    val = tcg_temp_new();
1299             tcg_gen_qemu_ld_i32(val, addr, ctx->memidx, MO_UB);
1300 	    tcg_gen_ori_i32(val, val, B7_0);
1301             tcg_gen_qemu_st_i32(val, addr, ctx->memidx, MO_UB);
1302 	    tcg_temp_free(val);
1303 	    tcg_temp_free(addr);
1304 	}
1305 	return;
1306     case 0xc300:		/* trapa #imm */
1307 	{
1308 	    TCGv imm;
1309 	    CHECK_NOT_DELAY_SLOT
1310             gen_save_cpu_state(ctx, true);
1311 	    imm = tcg_const_i32(B7_0);
1312             gen_helper_trapa(cpu_env, imm);
1313 	    tcg_temp_free(imm);
1314             ctx->base.is_jmp = DISAS_NORETURN;
1315 	}
1316 	return;
1317     case 0xc800:		/* tst #imm,R0 */
1318 	{
1319 	    TCGv val = tcg_temp_new();
1320 	    tcg_gen_andi_i32(val, REG(0), B7_0);
1321             tcg_gen_setcondi_i32(TCG_COND_EQ, cpu_sr_t, val, 0);
1322 	    tcg_temp_free(val);
1323 	}
1324 	return;
1325     case 0xcc00:		/* tst.b #imm,@(R0,GBR) */
1326 	{
1327 	    TCGv val = tcg_temp_new();
1328 	    tcg_gen_add_i32(val, REG(0), cpu_gbr);
1329             tcg_gen_qemu_ld_i32(val, val, ctx->memidx, MO_UB);
1330 	    tcg_gen_andi_i32(val, val, B7_0);
1331             tcg_gen_setcondi_i32(TCG_COND_EQ, cpu_sr_t, val, 0);
1332 	    tcg_temp_free(val);
1333 	}
1334 	return;
1335     case 0xca00:		/* xor #imm,R0 */
1336 	tcg_gen_xori_i32(REG(0), REG(0), B7_0);
1337 	return;
1338     case 0xce00:		/* xor.b #imm,@(R0,GBR) */
1339 	{
1340 	    TCGv addr, val;
1341 	    addr = tcg_temp_new();
1342 	    tcg_gen_add_i32(addr, REG(0), cpu_gbr);
1343 	    val = tcg_temp_new();
1344             tcg_gen_qemu_ld_i32(val, addr, ctx->memidx, MO_UB);
1345 	    tcg_gen_xori_i32(val, val, B7_0);
1346             tcg_gen_qemu_st_i32(val, addr, ctx->memidx, MO_UB);
1347 	    tcg_temp_free(val);
1348 	    tcg_temp_free(addr);
1349 	}
1350 	return;
1351     }
1352 
1353     switch (ctx->opcode & 0xf08f) {
1354     case 0x408e:		/* ldc Rm,Rn_BANK */
1355 	CHECK_PRIVILEGED
1356 	tcg_gen_mov_i32(ALTREG(B6_4), REG(B11_8));
1357 	return;
1358     case 0x4087:		/* ldc.l @Rm+,Rn_BANK */
1359 	CHECK_PRIVILEGED
1360         tcg_gen_qemu_ld_i32(ALTREG(B6_4), REG(B11_8), ctx->memidx, MO_TESL);
1361 	tcg_gen_addi_i32(REG(B11_8), REG(B11_8), 4);
1362 	return;
1363     case 0x0082:		/* stc Rm_BANK,Rn */
1364 	CHECK_PRIVILEGED
1365 	tcg_gen_mov_i32(REG(B11_8), ALTREG(B6_4));
1366 	return;
1367     case 0x4083:		/* stc.l Rm_BANK,@-Rn */
1368 	CHECK_PRIVILEGED
1369 	{
1370 	    TCGv addr = tcg_temp_new();
1371 	    tcg_gen_subi_i32(addr, REG(B11_8), 4);
1372             tcg_gen_qemu_st_i32(ALTREG(B6_4), addr, ctx->memidx, MO_TEUL);
1373 	    tcg_gen_mov_i32(REG(B11_8), addr);
1374 	    tcg_temp_free(addr);
1375 	}
1376 	return;
1377     }
1378 
1379     switch (ctx->opcode & 0xf0ff) {
1380     case 0x0023:		/* braf Rn */
1381 	CHECK_NOT_DELAY_SLOT
1382         tcg_gen_addi_i32(cpu_delayed_pc, REG(B11_8), ctx->base.pc_next + 4);
1383         ctx->envflags |= DELAY_SLOT;
1384 	ctx->delayed_pc = (uint32_t) - 1;
1385 	return;
1386     case 0x0003:		/* bsrf Rn */
1387 	CHECK_NOT_DELAY_SLOT
1388         tcg_gen_movi_i32(cpu_pr, ctx->base.pc_next + 4);
1389 	tcg_gen_add_i32(cpu_delayed_pc, REG(B11_8), cpu_pr);
1390         ctx->envflags |= DELAY_SLOT;
1391 	ctx->delayed_pc = (uint32_t) - 1;
1392 	return;
1393     case 0x4015:		/* cmp/pl Rn */
1394         tcg_gen_setcondi_i32(TCG_COND_GT, cpu_sr_t, REG(B11_8), 0);
1395 	return;
1396     case 0x4011:		/* cmp/pz Rn */
1397         tcg_gen_setcondi_i32(TCG_COND_GE, cpu_sr_t, REG(B11_8), 0);
1398 	return;
1399     case 0x4010:		/* dt Rn */
1400 	tcg_gen_subi_i32(REG(B11_8), REG(B11_8), 1);
1401         tcg_gen_setcondi_i32(TCG_COND_EQ, cpu_sr_t, REG(B11_8), 0);
1402 	return;
1403     case 0x402b:		/* jmp @Rn */
1404 	CHECK_NOT_DELAY_SLOT
1405 	tcg_gen_mov_i32(cpu_delayed_pc, REG(B11_8));
1406         ctx->envflags |= DELAY_SLOT;
1407 	ctx->delayed_pc = (uint32_t) - 1;
1408 	return;
1409     case 0x400b:		/* jsr @Rn */
1410 	CHECK_NOT_DELAY_SLOT
1411         tcg_gen_movi_i32(cpu_pr, ctx->base.pc_next + 4);
1412 	tcg_gen_mov_i32(cpu_delayed_pc, REG(B11_8));
1413         ctx->envflags |= DELAY_SLOT;
1414 	ctx->delayed_pc = (uint32_t) - 1;
1415 	return;
1416     case 0x400e:		/* ldc Rm,SR */
1417 	CHECK_PRIVILEGED
1418         {
1419             TCGv val = tcg_temp_new();
1420             tcg_gen_andi_i32(val, REG(B11_8), 0x700083f3);
1421             gen_write_sr(val);
1422             tcg_temp_free(val);
1423             ctx->base.is_jmp = DISAS_STOP;
1424         }
1425 	return;
1426     case 0x4007:		/* ldc.l @Rm+,SR */
1427 	CHECK_PRIVILEGED
1428 	{
1429 	    TCGv val = tcg_temp_new();
1430             tcg_gen_qemu_ld_i32(val, REG(B11_8), ctx->memidx, MO_TESL);
1431             tcg_gen_andi_i32(val, val, 0x700083f3);
1432             gen_write_sr(val);
1433 	    tcg_temp_free(val);
1434 	    tcg_gen_addi_i32(REG(B11_8), REG(B11_8), 4);
1435             ctx->base.is_jmp = DISAS_STOP;
1436 	}
1437 	return;
1438     case 0x0002:		/* stc SR,Rn */
1439 	CHECK_PRIVILEGED
1440         gen_read_sr(REG(B11_8));
1441 	return;
1442     case 0x4003:		/* stc SR,@-Rn */
1443 	CHECK_PRIVILEGED
1444 	{
1445 	    TCGv addr = tcg_temp_new();
1446             TCGv val = tcg_temp_new();
1447 	    tcg_gen_subi_i32(addr, REG(B11_8), 4);
1448             gen_read_sr(val);
1449             tcg_gen_qemu_st_i32(val, addr, ctx->memidx, MO_TEUL);
1450 	    tcg_gen_mov_i32(REG(B11_8), addr);
1451             tcg_temp_free(val);
1452 	    tcg_temp_free(addr);
1453 	}
1454 	return;
1455 #define LD(reg,ldnum,ldpnum,prechk)		\
1456   case ldnum:							\
1457     prechk    							\
1458     tcg_gen_mov_i32 (cpu_##reg, REG(B11_8));			\
1459     return;							\
1460   case ldpnum:							\
1461     prechk    							\
1462     tcg_gen_qemu_ld_i32(cpu_##reg, REG(B11_8), ctx->memidx, MO_TESL); \
1463     tcg_gen_addi_i32(REG(B11_8), REG(B11_8), 4);		\
1464     return;
1465 #define ST(reg,stnum,stpnum,prechk)		\
1466   case stnum:							\
1467     prechk    							\
1468     tcg_gen_mov_i32 (REG(B11_8), cpu_##reg);			\
1469     return;							\
1470   case stpnum:							\
1471     prechk    							\
1472     {								\
1473 	TCGv addr = tcg_temp_new();				\
1474 	tcg_gen_subi_i32(addr, REG(B11_8), 4);			\
1475         tcg_gen_qemu_st_i32(cpu_##reg, addr, ctx->memidx, MO_TEUL); \
1476 	tcg_gen_mov_i32(REG(B11_8), addr);			\
1477 	tcg_temp_free(addr);					\
1478     }								\
1479     return;
1480 #define LDST(reg,ldnum,ldpnum,stnum,stpnum,prechk)		\
1481 	LD(reg,ldnum,ldpnum,prechk)				\
1482 	ST(reg,stnum,stpnum,prechk)
1483 	LDST(gbr,  0x401e, 0x4017, 0x0012, 0x4013, {})
1484 	LDST(vbr,  0x402e, 0x4027, 0x0022, 0x4023, CHECK_PRIVILEGED)
1485 	LDST(ssr,  0x403e, 0x4037, 0x0032, 0x4033, CHECK_PRIVILEGED)
1486 	LDST(spc,  0x404e, 0x4047, 0x0042, 0x4043, CHECK_PRIVILEGED)
1487 	ST(sgr,  0x003a, 0x4032, CHECK_PRIVILEGED)
1488         LD(sgr,  0x403a, 0x4036, CHECK_PRIVILEGED CHECK_SH4A)
1489 	LDST(dbr,  0x40fa, 0x40f6, 0x00fa, 0x40f2, CHECK_PRIVILEGED)
1490 	LDST(mach, 0x400a, 0x4006, 0x000a, 0x4002, {})
1491 	LDST(macl, 0x401a, 0x4016, 0x001a, 0x4012, {})
1492 	LDST(pr,   0x402a, 0x4026, 0x002a, 0x4022, {})
1493 	LDST(fpul, 0x405a, 0x4056, 0x005a, 0x4052, {CHECK_FPU_ENABLED})
1494     case 0x406a:		/* lds Rm,FPSCR */
1495 	CHECK_FPU_ENABLED
1496         gen_helper_ld_fpscr(cpu_env, REG(B11_8));
1497         ctx->base.is_jmp = DISAS_STOP;
1498 	return;
1499     case 0x4066:		/* lds.l @Rm+,FPSCR */
1500 	CHECK_FPU_ENABLED
1501 	{
1502 	    TCGv addr = tcg_temp_new();
1503             tcg_gen_qemu_ld_i32(addr, REG(B11_8), ctx->memidx, MO_TESL);
1504 	    tcg_gen_addi_i32(REG(B11_8), REG(B11_8), 4);
1505             gen_helper_ld_fpscr(cpu_env, addr);
1506 	    tcg_temp_free(addr);
1507             ctx->base.is_jmp = DISAS_STOP;
1508 	}
1509 	return;
1510     case 0x006a:		/* sts FPSCR,Rn */
1511 	CHECK_FPU_ENABLED
1512 	tcg_gen_andi_i32(REG(B11_8), cpu_fpscr, 0x003fffff);
1513 	return;
1514     case 0x4062:		/* sts FPSCR,@-Rn */
1515 	CHECK_FPU_ENABLED
1516 	{
1517 	    TCGv addr, val;
1518 	    val = tcg_temp_new();
1519 	    tcg_gen_andi_i32(val, cpu_fpscr, 0x003fffff);
1520 	    addr = tcg_temp_new();
1521 	    tcg_gen_subi_i32(addr, REG(B11_8), 4);
1522             tcg_gen_qemu_st_i32(val, addr, ctx->memidx, MO_TEUL);
1523 	    tcg_gen_mov_i32(REG(B11_8), addr);
1524 	    tcg_temp_free(addr);
1525 	    tcg_temp_free(val);
1526 	}
1527 	return;
1528     case 0x00c3:		/* movca.l R0,@Rm */
1529         {
1530             TCGv val = tcg_temp_new();
1531             tcg_gen_qemu_ld_i32(val, REG(B11_8), ctx->memidx, MO_TEUL);
1532             gen_helper_movcal(cpu_env, REG(B11_8), val);
1533             tcg_gen_qemu_st_i32(REG(0), REG(B11_8), ctx->memidx, MO_TEUL);
1534             tcg_temp_free(val);
1535         }
1536         ctx->has_movcal = 1;
1537 	return;
1538     case 0x40a9:                /* movua.l @Rm,R0 */
1539         CHECK_SH4A
1540         /* Load non-boundary-aligned data */
1541         tcg_gen_qemu_ld_i32(REG(0), REG(B11_8), ctx->memidx,
1542                             MO_TEUL | MO_UNALN);
1543         return;
1544     case 0x40e9:                /* movua.l @Rm+,R0 */
1545         CHECK_SH4A
1546         /* Load non-boundary-aligned data */
1547         tcg_gen_qemu_ld_i32(REG(0), REG(B11_8), ctx->memidx,
1548                             MO_TEUL | MO_UNALN);
1549         tcg_gen_addi_i32(REG(B11_8), REG(B11_8), 4);
1550         return;
1551     case 0x0029:		/* movt Rn */
1552         tcg_gen_mov_i32(REG(B11_8), cpu_sr_t);
1553 	return;
1554     case 0x0073:
1555         /* MOVCO.L
1556          *     LDST -> T
1557          *     If (T == 1) R0 -> (Rn)
1558          *     0 -> LDST
1559          *
1560          * The above description doesn't work in a parallel context.
1561          * Since we currently support no smp boards, this implies user-mode.
1562          * But we can still support the official mechanism while user-mode
1563          * is single-threaded.  */
1564         CHECK_SH4A
1565         {
1566             TCGLabel *fail = gen_new_label();
1567             TCGLabel *done = gen_new_label();
1568 
1569             if ((tb_cflags(ctx->base.tb) & CF_PARALLEL)) {
1570                 TCGv tmp;
1571 
1572                 tcg_gen_brcond_i32(TCG_COND_NE, REG(B11_8),
1573                                    cpu_lock_addr, fail);
1574                 tmp = tcg_temp_new();
1575                 tcg_gen_atomic_cmpxchg_i32(tmp, REG(B11_8), cpu_lock_value,
1576                                            REG(0), ctx->memidx, MO_TEUL);
1577                 tcg_gen_setcond_i32(TCG_COND_EQ, cpu_sr_t, tmp, cpu_lock_value);
1578                 tcg_temp_free(tmp);
1579             } else {
1580                 tcg_gen_brcondi_i32(TCG_COND_EQ, cpu_lock_addr, -1, fail);
1581                 tcg_gen_qemu_st_i32(REG(0), REG(B11_8), ctx->memidx, MO_TEUL);
1582                 tcg_gen_movi_i32(cpu_sr_t, 1);
1583             }
1584             tcg_gen_br(done);
1585 
1586             gen_set_label(fail);
1587             tcg_gen_movi_i32(cpu_sr_t, 0);
1588 
1589             gen_set_label(done);
1590             tcg_gen_movi_i32(cpu_lock_addr, -1);
1591         }
1592         return;
1593     case 0x0063:
1594         /* MOVLI.L @Rm,R0
1595          *     1 -> LDST
1596          *     (Rm) -> R0
1597          *     When interrupt/exception
1598          *     occurred 0 -> LDST
1599          *
1600          * In a parallel context, we must also save the loaded value
1601          * for use with the cmpxchg that we'll use with movco.l.  */
1602         CHECK_SH4A
1603         if ((tb_cflags(ctx->base.tb) & CF_PARALLEL)) {
1604             TCGv tmp = tcg_temp_new();
1605             tcg_gen_mov_i32(tmp, REG(B11_8));
1606             tcg_gen_qemu_ld_i32(REG(0), REG(B11_8), ctx->memidx, MO_TESL);
1607             tcg_gen_mov_i32(cpu_lock_value, REG(0));
1608             tcg_gen_mov_i32(cpu_lock_addr, tmp);
1609             tcg_temp_free(tmp);
1610         } else {
1611             tcg_gen_qemu_ld_i32(REG(0), REG(B11_8), ctx->memidx, MO_TESL);
1612             tcg_gen_movi_i32(cpu_lock_addr, 0);
1613         }
1614         return;
1615     case 0x0093:		/* ocbi @Rn */
1616 	{
1617             gen_helper_ocbi(cpu_env, REG(B11_8));
1618 	}
1619 	return;
1620     case 0x00a3:		/* ocbp @Rn */
1621     case 0x00b3:		/* ocbwb @Rn */
1622         /* These instructions are supposed to do nothing in case of
1623            a cache miss. Given that we only partially emulate caches
1624            it is safe to simply ignore them. */
1625 	return;
1626     case 0x0083:		/* pref @Rn */
1627 	return;
1628     case 0x00d3:		/* prefi @Rn */
1629         CHECK_SH4A
1630         return;
1631     case 0x00e3:		/* icbi @Rn */
1632         CHECK_SH4A
1633         return;
1634     case 0x00ab:		/* synco */
1635         CHECK_SH4A
1636         tcg_gen_mb(TCG_MO_ALL | TCG_BAR_SC);
1637         return;
1638     case 0x4024:		/* rotcl Rn */
1639 	{
1640 	    TCGv tmp = tcg_temp_new();
1641             tcg_gen_mov_i32(tmp, cpu_sr_t);
1642             tcg_gen_shri_i32(cpu_sr_t, REG(B11_8), 31);
1643 	    tcg_gen_shli_i32(REG(B11_8), REG(B11_8), 1);
1644             tcg_gen_or_i32(REG(B11_8), REG(B11_8), tmp);
1645 	    tcg_temp_free(tmp);
1646 	}
1647 	return;
1648     case 0x4025:		/* rotcr Rn */
1649 	{
1650 	    TCGv tmp = tcg_temp_new();
1651             tcg_gen_shli_i32(tmp, cpu_sr_t, 31);
1652             tcg_gen_andi_i32(cpu_sr_t, REG(B11_8), 1);
1653 	    tcg_gen_shri_i32(REG(B11_8), REG(B11_8), 1);
1654             tcg_gen_or_i32(REG(B11_8), REG(B11_8), tmp);
1655 	    tcg_temp_free(tmp);
1656 	}
1657 	return;
1658     case 0x4004:		/* rotl Rn */
1659 	tcg_gen_rotli_i32(REG(B11_8), REG(B11_8), 1);
1660         tcg_gen_andi_i32(cpu_sr_t, REG(B11_8), 0);
1661 	return;
1662     case 0x4005:		/* rotr Rn */
1663         tcg_gen_andi_i32(cpu_sr_t, REG(B11_8), 0);
1664 	tcg_gen_rotri_i32(REG(B11_8), REG(B11_8), 1);
1665 	return;
1666     case 0x4000:		/* shll Rn */
1667     case 0x4020:		/* shal Rn */
1668         tcg_gen_shri_i32(cpu_sr_t, REG(B11_8), 31);
1669 	tcg_gen_shli_i32(REG(B11_8), REG(B11_8), 1);
1670 	return;
1671     case 0x4021:		/* shar Rn */
1672         tcg_gen_andi_i32(cpu_sr_t, REG(B11_8), 1);
1673 	tcg_gen_sari_i32(REG(B11_8), REG(B11_8), 1);
1674 	return;
1675     case 0x4001:		/* shlr Rn */
1676         tcg_gen_andi_i32(cpu_sr_t, REG(B11_8), 1);
1677 	tcg_gen_shri_i32(REG(B11_8), REG(B11_8), 1);
1678 	return;
1679     case 0x4008:		/* shll2 Rn */
1680 	tcg_gen_shli_i32(REG(B11_8), REG(B11_8), 2);
1681 	return;
1682     case 0x4018:		/* shll8 Rn */
1683 	tcg_gen_shli_i32(REG(B11_8), REG(B11_8), 8);
1684 	return;
1685     case 0x4028:		/* shll16 Rn */
1686 	tcg_gen_shli_i32(REG(B11_8), REG(B11_8), 16);
1687 	return;
1688     case 0x4009:		/* shlr2 Rn */
1689 	tcg_gen_shri_i32(REG(B11_8), REG(B11_8), 2);
1690 	return;
1691     case 0x4019:		/* shlr8 Rn */
1692 	tcg_gen_shri_i32(REG(B11_8), REG(B11_8), 8);
1693 	return;
1694     case 0x4029:		/* shlr16 Rn */
1695 	tcg_gen_shri_i32(REG(B11_8), REG(B11_8), 16);
1696 	return;
1697     case 0x401b:		/* tas.b @Rn */
1698         {
1699             TCGv val = tcg_const_i32(0x80);
1700             tcg_gen_atomic_fetch_or_i32(val, REG(B11_8), val,
1701                                         ctx->memidx, MO_UB);
1702             tcg_gen_setcondi_i32(TCG_COND_EQ, cpu_sr_t, val, 0);
1703             tcg_temp_free(val);
1704         }
1705         return;
1706     case 0xf00d: /* fsts FPUL,FRn - FPSCR: Nothing */
1707 	CHECK_FPU_ENABLED
1708         tcg_gen_mov_i32(FREG(B11_8), cpu_fpul);
1709 	return;
1710     case 0xf01d: /* flds FRm,FPUL - FPSCR: Nothing */
1711 	CHECK_FPU_ENABLED
1712         tcg_gen_mov_i32(cpu_fpul, FREG(B11_8));
1713 	return;
1714     case 0xf02d: /* float FPUL,FRn/DRn - FPSCR: R[PR,Enable.I]/W[Cause,Flag] */
1715 	CHECK_FPU_ENABLED
1716         if (ctx->tbflags & FPSCR_PR) {
1717 	    TCGv_i64 fp;
1718             if (ctx->opcode & 0x0100) {
1719                 goto do_illegal;
1720             }
1721 	    fp = tcg_temp_new_i64();
1722             gen_helper_float_DT(fp, cpu_env, cpu_fpul);
1723             gen_store_fpr64(ctx, fp, B11_8);
1724 	    tcg_temp_free_i64(fp);
1725 	}
1726 	else {
1727             gen_helper_float_FT(FREG(B11_8), cpu_env, cpu_fpul);
1728 	}
1729 	return;
1730     case 0xf03d: /* ftrc FRm/DRm,FPUL - FPSCR: R[PR,Enable.V]/W[Cause,Flag] */
1731 	CHECK_FPU_ENABLED
1732         if (ctx->tbflags & FPSCR_PR) {
1733 	    TCGv_i64 fp;
1734             if (ctx->opcode & 0x0100) {
1735                 goto do_illegal;
1736             }
1737 	    fp = tcg_temp_new_i64();
1738             gen_load_fpr64(ctx, fp, B11_8);
1739             gen_helper_ftrc_DT(cpu_fpul, cpu_env, fp);
1740 	    tcg_temp_free_i64(fp);
1741 	}
1742 	else {
1743             gen_helper_ftrc_FT(cpu_fpul, cpu_env, FREG(B11_8));
1744 	}
1745 	return;
1746     case 0xf04d: /* fneg FRn/DRn - FPSCR: Nothing */
1747 	CHECK_FPU_ENABLED
1748         tcg_gen_xori_i32(FREG(B11_8), FREG(B11_8), 0x80000000);
1749 	return;
1750     case 0xf05d: /* fabs FRn/DRn - FPCSR: Nothing */
1751 	CHECK_FPU_ENABLED
1752         tcg_gen_andi_i32(FREG(B11_8), FREG(B11_8), 0x7fffffff);
1753 	return;
1754     case 0xf06d: /* fsqrt FRn */
1755 	CHECK_FPU_ENABLED
1756         if (ctx->tbflags & FPSCR_PR) {
1757             if (ctx->opcode & 0x0100) {
1758                 goto do_illegal;
1759             }
1760 	    TCGv_i64 fp = tcg_temp_new_i64();
1761             gen_load_fpr64(ctx, fp, B11_8);
1762             gen_helper_fsqrt_DT(fp, cpu_env, fp);
1763             gen_store_fpr64(ctx, fp, B11_8);
1764 	    tcg_temp_free_i64(fp);
1765 	} else {
1766             gen_helper_fsqrt_FT(FREG(B11_8), cpu_env, FREG(B11_8));
1767 	}
1768 	return;
1769     case 0xf07d: /* fsrra FRn */
1770 	CHECK_FPU_ENABLED
1771         CHECK_FPSCR_PR_0
1772         gen_helper_fsrra_FT(FREG(B11_8), cpu_env, FREG(B11_8));
1773 	break;
1774     case 0xf08d: /* fldi0 FRn - FPSCR: R[PR] */
1775 	CHECK_FPU_ENABLED
1776         CHECK_FPSCR_PR_0
1777         tcg_gen_movi_i32(FREG(B11_8), 0);
1778         return;
1779     case 0xf09d: /* fldi1 FRn - FPSCR: R[PR] */
1780 	CHECK_FPU_ENABLED
1781         CHECK_FPSCR_PR_0
1782         tcg_gen_movi_i32(FREG(B11_8), 0x3f800000);
1783         return;
1784     case 0xf0ad: /* fcnvsd FPUL,DRn */
1785 	CHECK_FPU_ENABLED
1786 	{
1787 	    TCGv_i64 fp = tcg_temp_new_i64();
1788             gen_helper_fcnvsd_FT_DT(fp, cpu_env, cpu_fpul);
1789             gen_store_fpr64(ctx, fp, B11_8);
1790 	    tcg_temp_free_i64(fp);
1791 	}
1792 	return;
1793     case 0xf0bd: /* fcnvds DRn,FPUL */
1794 	CHECK_FPU_ENABLED
1795 	{
1796 	    TCGv_i64 fp = tcg_temp_new_i64();
1797             gen_load_fpr64(ctx, fp, B11_8);
1798             gen_helper_fcnvds_DT_FT(cpu_fpul, cpu_env, fp);
1799 	    tcg_temp_free_i64(fp);
1800 	}
1801 	return;
1802     case 0xf0ed: /* fipr FVm,FVn */
1803         CHECK_FPU_ENABLED
1804         CHECK_FPSCR_PR_1
1805         {
1806             TCGv m = tcg_const_i32((ctx->opcode >> 8) & 3);
1807             TCGv n = tcg_const_i32((ctx->opcode >> 10) & 3);
1808             gen_helper_fipr(cpu_env, m, n);
1809             tcg_temp_free(m);
1810             tcg_temp_free(n);
1811             return;
1812         }
1813         break;
1814     case 0xf0fd: /* ftrv XMTRX,FVn */
1815         CHECK_FPU_ENABLED
1816         CHECK_FPSCR_PR_1
1817         {
1818             if ((ctx->opcode & 0x0300) != 0x0100) {
1819                 goto do_illegal;
1820             }
1821             TCGv n = tcg_const_i32((ctx->opcode >> 10) & 3);
1822             gen_helper_ftrv(cpu_env, n);
1823             tcg_temp_free(n);
1824             return;
1825         }
1826         break;
1827     }
1828 #if 0
1829     fprintf(stderr, "unknown instruction 0x%04x at pc 0x%08x\n",
1830             ctx->opcode, ctx->base.pc_next);
1831     fflush(stderr);
1832 #endif
1833  do_illegal:
1834     if (ctx->envflags & DELAY_SLOT_MASK) {
1835  do_illegal_slot:
1836         gen_save_cpu_state(ctx, true);
1837         gen_helper_raise_slot_illegal_instruction(cpu_env);
1838     } else {
1839         gen_save_cpu_state(ctx, true);
1840         gen_helper_raise_illegal_instruction(cpu_env);
1841     }
1842     ctx->base.is_jmp = DISAS_NORETURN;
1843     return;
1844 
1845  do_fpu_disabled:
1846     gen_save_cpu_state(ctx, true);
1847     if (ctx->envflags & DELAY_SLOT_MASK) {
1848         gen_helper_raise_slot_fpu_disable(cpu_env);
1849     } else {
1850         gen_helper_raise_fpu_disable(cpu_env);
1851     }
1852     ctx->base.is_jmp = DISAS_NORETURN;
1853     return;
1854 }
1855 
1856 static void decode_opc(DisasContext * ctx)
1857 {
1858     uint32_t old_flags = ctx->envflags;
1859 
1860     _decode_opc(ctx);
1861 
1862     if (old_flags & DELAY_SLOT_MASK) {
1863         /* go out of the delay slot */
1864         ctx->envflags &= ~DELAY_SLOT_MASK;
1865 
1866         /* When in an exclusive region, we must continue to the end
1867            for conditional branches.  */
1868         if (ctx->tbflags & GUSA_EXCLUSIVE
1869             && old_flags & DELAY_SLOT_CONDITIONAL) {
1870             gen_delayed_conditional_jump(ctx);
1871             return;
1872         }
1873         /* Otherwise this is probably an invalid gUSA region.
1874            Drop the GUSA bits so the next TB doesn't see them.  */
1875         ctx->envflags &= ~GUSA_MASK;
1876 
1877         tcg_gen_movi_i32(cpu_flags, ctx->envflags);
1878         if (old_flags & DELAY_SLOT_CONDITIONAL) {
1879 	    gen_delayed_conditional_jump(ctx);
1880         } else {
1881             gen_jump(ctx);
1882 	}
1883     }
1884 }
1885 
1886 #ifdef CONFIG_USER_ONLY
1887 /* For uniprocessors, SH4 uses optimistic restartable atomic sequences.
1888    Upon an interrupt, a real kernel would simply notice magic values in
1889    the registers and reset the PC to the start of the sequence.
1890 
1891    For QEMU, we cannot do this in quite the same way.  Instead, we notice
1892    the normal start of such a sequence (mov #-x,r15).  While we can handle
1893    any sequence via cpu_exec_step_atomic, we can recognize the "normal"
1894    sequences and transform them into atomic operations as seen by the host.
1895 */
1896 static void decode_gusa(DisasContext *ctx, CPUSH4State *env)
1897 {
1898     uint16_t insns[5];
1899     int ld_adr, ld_dst, ld_mop;
1900     int op_dst, op_src, op_opc;
1901     int mv_src, mt_dst, st_src, st_mop;
1902     TCGv op_arg;
1903     uint32_t pc = ctx->base.pc_next;
1904     uint32_t pc_end = ctx->base.tb->cs_base;
1905     int max_insns = (pc_end - pc) / 2;
1906     int i;
1907 
1908     /* The state machine below will consume only a few insns.
1909        If there are more than that in a region, fail now.  */
1910     if (max_insns > ARRAY_SIZE(insns)) {
1911         goto fail;
1912     }
1913 
1914     /* Read all of the insns for the region.  */
1915     for (i = 0; i < max_insns; ++i) {
1916         insns[i] = translator_lduw(env, pc + i * 2);
1917     }
1918 
1919     ld_adr = ld_dst = ld_mop = -1;
1920     mv_src = -1;
1921     op_dst = op_src = op_opc = -1;
1922     mt_dst = -1;
1923     st_src = st_mop = -1;
1924     op_arg = NULL;
1925     i = 0;
1926 
1927 #define NEXT_INSN \
1928     do { if (i >= max_insns) goto fail; ctx->opcode = insns[i++]; } while (0)
1929 
1930     /*
1931      * Expect a load to begin the region.
1932      */
1933     NEXT_INSN;
1934     switch (ctx->opcode & 0xf00f) {
1935     case 0x6000: /* mov.b @Rm,Rn */
1936         ld_mop = MO_SB;
1937         break;
1938     case 0x6001: /* mov.w @Rm,Rn */
1939         ld_mop = MO_TESW;
1940         break;
1941     case 0x6002: /* mov.l @Rm,Rn */
1942         ld_mop = MO_TESL;
1943         break;
1944     default:
1945         goto fail;
1946     }
1947     ld_adr = B7_4;
1948     ld_dst = B11_8;
1949     if (ld_adr == ld_dst) {
1950         goto fail;
1951     }
1952     /* Unless we see a mov, any two-operand operation must use ld_dst.  */
1953     op_dst = ld_dst;
1954 
1955     /*
1956      * Expect an optional register move.
1957      */
1958     NEXT_INSN;
1959     switch (ctx->opcode & 0xf00f) {
1960     case 0x6003: /* mov Rm,Rn */
1961         /*
1962          * Here we want to recognize ld_dst being saved for later consumption,
1963          * or for another input register being copied so that ld_dst need not
1964          * be clobbered during the operation.
1965          */
1966         op_dst = B11_8;
1967         mv_src = B7_4;
1968         if (op_dst == ld_dst) {
1969             /* Overwriting the load output.  */
1970             goto fail;
1971         }
1972         if (mv_src != ld_dst) {
1973             /* Copying a new input; constrain op_src to match the load.  */
1974             op_src = ld_dst;
1975         }
1976         break;
1977 
1978     default:
1979         /* Put back and re-examine as operation.  */
1980         --i;
1981     }
1982 
1983     /*
1984      * Expect the operation.
1985      */
1986     NEXT_INSN;
1987     switch (ctx->opcode & 0xf00f) {
1988     case 0x300c: /* add Rm,Rn */
1989         op_opc = INDEX_op_add_i32;
1990         goto do_reg_op;
1991     case 0x2009: /* and Rm,Rn */
1992         op_opc = INDEX_op_and_i32;
1993         goto do_reg_op;
1994     case 0x200a: /* xor Rm,Rn */
1995         op_opc = INDEX_op_xor_i32;
1996         goto do_reg_op;
1997     case 0x200b: /* or Rm,Rn */
1998         op_opc = INDEX_op_or_i32;
1999     do_reg_op:
2000         /* The operation register should be as expected, and the
2001            other input cannot depend on the load.  */
2002         if (op_dst != B11_8) {
2003             goto fail;
2004         }
2005         if (op_src < 0) {
2006             /* Unconstrainted input.  */
2007             op_src = B7_4;
2008         } else if (op_src == B7_4) {
2009             /* Constrained input matched load.  All operations are
2010                commutative; "swap" them by "moving" the load output
2011                to the (implicit) first argument and the move source
2012                to the (explicit) second argument.  */
2013             op_src = mv_src;
2014         } else {
2015             goto fail;
2016         }
2017         op_arg = REG(op_src);
2018         break;
2019 
2020     case 0x6007: /* not Rm,Rn */
2021         if (ld_dst != B7_4 || mv_src >= 0) {
2022             goto fail;
2023         }
2024         op_dst = B11_8;
2025         op_opc = INDEX_op_xor_i32;
2026         op_arg = tcg_const_i32(-1);
2027         break;
2028 
2029     case 0x7000 ... 0x700f: /* add #imm,Rn */
2030         if (op_dst != B11_8 || mv_src >= 0) {
2031             goto fail;
2032         }
2033         op_opc = INDEX_op_add_i32;
2034         op_arg = tcg_const_i32(B7_0s);
2035         break;
2036 
2037     case 0x3000: /* cmp/eq Rm,Rn */
2038         /* Looking for the middle of a compare-and-swap sequence,
2039            beginning with the compare.  Operands can be either order,
2040            but with only one overlapping the load.  */
2041         if ((ld_dst == B11_8) + (ld_dst == B7_4) != 1 || mv_src >= 0) {
2042             goto fail;
2043         }
2044         op_opc = INDEX_op_setcond_i32;  /* placeholder */
2045         op_src = (ld_dst == B11_8 ? B7_4 : B11_8);
2046         op_arg = REG(op_src);
2047 
2048         NEXT_INSN;
2049         switch (ctx->opcode & 0xff00) {
2050         case 0x8b00: /* bf label */
2051         case 0x8f00: /* bf/s label */
2052             if (pc + (i + 1 + B7_0s) * 2 != pc_end) {
2053                 goto fail;
2054             }
2055             if ((ctx->opcode & 0xff00) == 0x8b00) { /* bf label */
2056                 break;
2057             }
2058             /* We're looking to unconditionally modify Rn with the
2059                result of the comparison, within the delay slot of
2060                the branch.  This is used by older gcc.  */
2061             NEXT_INSN;
2062             if ((ctx->opcode & 0xf0ff) == 0x0029) { /* movt Rn */
2063                 mt_dst = B11_8;
2064             } else {
2065                 goto fail;
2066             }
2067             break;
2068 
2069         default:
2070             goto fail;
2071         }
2072         break;
2073 
2074     case 0x2008: /* tst Rm,Rn */
2075         /* Looking for a compare-and-swap against zero.  */
2076         if (ld_dst != B11_8 || ld_dst != B7_4 || mv_src >= 0) {
2077             goto fail;
2078         }
2079         op_opc = INDEX_op_setcond_i32;
2080         op_arg = tcg_const_i32(0);
2081 
2082         NEXT_INSN;
2083         if ((ctx->opcode & 0xff00) != 0x8900 /* bt label */
2084             || pc + (i + 1 + B7_0s) * 2 != pc_end) {
2085             goto fail;
2086         }
2087         break;
2088 
2089     default:
2090         /* Put back and re-examine as store.  */
2091         --i;
2092     }
2093 
2094     /*
2095      * Expect the store.
2096      */
2097     /* The store must be the last insn.  */
2098     if (i != max_insns - 1) {
2099         goto fail;
2100     }
2101     NEXT_INSN;
2102     switch (ctx->opcode & 0xf00f) {
2103     case 0x2000: /* mov.b Rm,@Rn */
2104         st_mop = MO_UB;
2105         break;
2106     case 0x2001: /* mov.w Rm,@Rn */
2107         st_mop = MO_UW;
2108         break;
2109     case 0x2002: /* mov.l Rm,@Rn */
2110         st_mop = MO_UL;
2111         break;
2112     default:
2113         goto fail;
2114     }
2115     /* The store must match the load.  */
2116     if (ld_adr != B11_8 || st_mop != (ld_mop & MO_SIZE)) {
2117         goto fail;
2118     }
2119     st_src = B7_4;
2120 
2121 #undef NEXT_INSN
2122 
2123     /*
2124      * Emit the operation.
2125      */
2126     switch (op_opc) {
2127     case -1:
2128         /* No operation found.  Look for exchange pattern.  */
2129         if (st_src == ld_dst || mv_src >= 0) {
2130             goto fail;
2131         }
2132         tcg_gen_atomic_xchg_i32(REG(ld_dst), REG(ld_adr), REG(st_src),
2133                                 ctx->memidx, ld_mop);
2134         break;
2135 
2136     case INDEX_op_add_i32:
2137         if (op_dst != st_src) {
2138             goto fail;
2139         }
2140         if (op_dst == ld_dst && st_mop == MO_UL) {
2141             tcg_gen_atomic_add_fetch_i32(REG(ld_dst), REG(ld_adr),
2142                                          op_arg, ctx->memidx, ld_mop);
2143         } else {
2144             tcg_gen_atomic_fetch_add_i32(REG(ld_dst), REG(ld_adr),
2145                                          op_arg, ctx->memidx, ld_mop);
2146             if (op_dst != ld_dst) {
2147                 /* Note that mop sizes < 4 cannot use add_fetch
2148                    because it won't carry into the higher bits.  */
2149                 tcg_gen_add_i32(REG(op_dst), REG(ld_dst), op_arg);
2150             }
2151         }
2152         break;
2153 
2154     case INDEX_op_and_i32:
2155         if (op_dst != st_src) {
2156             goto fail;
2157         }
2158         if (op_dst == ld_dst) {
2159             tcg_gen_atomic_and_fetch_i32(REG(ld_dst), REG(ld_adr),
2160                                          op_arg, ctx->memidx, ld_mop);
2161         } else {
2162             tcg_gen_atomic_fetch_and_i32(REG(ld_dst), REG(ld_adr),
2163                                          op_arg, ctx->memidx, ld_mop);
2164             tcg_gen_and_i32(REG(op_dst), REG(ld_dst), op_arg);
2165         }
2166         break;
2167 
2168     case INDEX_op_or_i32:
2169         if (op_dst != st_src) {
2170             goto fail;
2171         }
2172         if (op_dst == ld_dst) {
2173             tcg_gen_atomic_or_fetch_i32(REG(ld_dst), REG(ld_adr),
2174                                         op_arg, ctx->memidx, ld_mop);
2175         } else {
2176             tcg_gen_atomic_fetch_or_i32(REG(ld_dst), REG(ld_adr),
2177                                         op_arg, ctx->memidx, ld_mop);
2178             tcg_gen_or_i32(REG(op_dst), REG(ld_dst), op_arg);
2179         }
2180         break;
2181 
2182     case INDEX_op_xor_i32:
2183         if (op_dst != st_src) {
2184             goto fail;
2185         }
2186         if (op_dst == ld_dst) {
2187             tcg_gen_atomic_xor_fetch_i32(REG(ld_dst), REG(ld_adr),
2188                                          op_arg, ctx->memidx, ld_mop);
2189         } else {
2190             tcg_gen_atomic_fetch_xor_i32(REG(ld_dst), REG(ld_adr),
2191                                          op_arg, ctx->memidx, ld_mop);
2192             tcg_gen_xor_i32(REG(op_dst), REG(ld_dst), op_arg);
2193         }
2194         break;
2195 
2196     case INDEX_op_setcond_i32:
2197         if (st_src == ld_dst) {
2198             goto fail;
2199         }
2200         tcg_gen_atomic_cmpxchg_i32(REG(ld_dst), REG(ld_adr), op_arg,
2201                                    REG(st_src), ctx->memidx, ld_mop);
2202         tcg_gen_setcond_i32(TCG_COND_EQ, cpu_sr_t, REG(ld_dst), op_arg);
2203         if (mt_dst >= 0) {
2204             tcg_gen_mov_i32(REG(mt_dst), cpu_sr_t);
2205         }
2206         break;
2207 
2208     default:
2209         g_assert_not_reached();
2210     }
2211 
2212     /* If op_src is not a valid register, then op_arg was a constant.  */
2213     if (op_src < 0 && op_arg) {
2214         tcg_temp_free_i32(op_arg);
2215     }
2216 
2217     /* The entire region has been translated.  */
2218     ctx->envflags &= ~GUSA_MASK;
2219     ctx->base.pc_next = pc_end;
2220     ctx->base.num_insns += max_insns - 1;
2221     return;
2222 
2223  fail:
2224     qemu_log_mask(LOG_UNIMP, "Unrecognized gUSA sequence %08x-%08x\n",
2225                   pc, pc_end);
2226 
2227     /* Restart with the EXCLUSIVE bit set, within a TB run via
2228        cpu_exec_step_atomic holding the exclusive lock.  */
2229     ctx->envflags |= GUSA_EXCLUSIVE;
2230     gen_save_cpu_state(ctx, false);
2231     gen_helper_exclusive(cpu_env);
2232     ctx->base.is_jmp = DISAS_NORETURN;
2233 
2234     /* We're not executing an instruction, but we must report one for the
2235        purposes of accounting within the TB.  We might as well report the
2236        entire region consumed via ctx->base.pc_next so that it's immediately
2237        available in the disassembly dump.  */
2238     ctx->base.pc_next = pc_end;
2239     ctx->base.num_insns += max_insns - 1;
2240 }
2241 #endif
2242 
2243 static void sh4_tr_init_disas_context(DisasContextBase *dcbase, CPUState *cs)
2244 {
2245     DisasContext *ctx = container_of(dcbase, DisasContext, base);
2246     CPUSH4State *env = cs->env_ptr;
2247     uint32_t tbflags;
2248     int bound;
2249 
2250     ctx->tbflags = tbflags = ctx->base.tb->flags;
2251     ctx->envflags = tbflags & TB_FLAG_ENVFLAGS_MASK;
2252     ctx->memidx = (tbflags & (1u << SR_MD)) == 0 ? 1 : 0;
2253     /* We don't know if the delayed pc came from a dynamic or static branch,
2254        so assume it is a dynamic branch.  */
2255     ctx->delayed_pc = -1; /* use delayed pc from env pointer */
2256     ctx->features = env->features;
2257     ctx->has_movcal = (tbflags & TB_FLAG_PENDING_MOVCA);
2258     ctx->gbank = ((tbflags & (1 << SR_MD)) &&
2259                   (tbflags & (1 << SR_RB))) * 0x10;
2260     ctx->fbank = tbflags & FPSCR_FR ? 0x10 : 0;
2261 
2262     if (tbflags & GUSA_MASK) {
2263         uint32_t pc = ctx->base.pc_next;
2264         uint32_t pc_end = ctx->base.tb->cs_base;
2265         int backup = sextract32(ctx->tbflags, GUSA_SHIFT, 8);
2266         int max_insns = (pc_end - pc) / 2;
2267 
2268         if (pc != pc_end + backup || max_insns < 2) {
2269             /* This is a malformed gUSA region.  Don't do anything special,
2270                since the interpreter is likely to get confused.  */
2271             ctx->envflags &= ~GUSA_MASK;
2272         } else if (tbflags & GUSA_EXCLUSIVE) {
2273             /* Regardless of single-stepping or the end of the page,
2274                we must complete execution of the gUSA region while
2275                holding the exclusive lock.  */
2276             ctx->base.max_insns = max_insns;
2277             return;
2278         }
2279     }
2280 
2281     /* Since the ISA is fixed-width, we can bound by the number
2282        of instructions remaining on the page.  */
2283     bound = -(ctx->base.pc_next | TARGET_PAGE_MASK) / 2;
2284     ctx->base.max_insns = MIN(ctx->base.max_insns, bound);
2285 }
2286 
2287 static void sh4_tr_tb_start(DisasContextBase *dcbase, CPUState *cs)
2288 {
2289 }
2290 
2291 static void sh4_tr_insn_start(DisasContextBase *dcbase, CPUState *cs)
2292 {
2293     DisasContext *ctx = container_of(dcbase, DisasContext, base);
2294 
2295     tcg_gen_insn_start(ctx->base.pc_next, ctx->envflags);
2296 }
2297 
2298 static bool sh4_tr_breakpoint_check(DisasContextBase *dcbase, CPUState *cs,
2299                                     const CPUBreakpoint *bp)
2300 {
2301     DisasContext *ctx = container_of(dcbase, DisasContext, base);
2302 
2303     /* We have hit a breakpoint - make sure PC is up-to-date */
2304     gen_save_cpu_state(ctx, true);
2305     gen_helper_debug(cpu_env);
2306     ctx->base.is_jmp = DISAS_NORETURN;
2307     /* The address covered by the breakpoint must be included in
2308        [tb->pc, tb->pc + tb->size) in order to for it to be
2309        properly cleared -- thus we increment the PC here so that
2310        the logic setting tb->size below does the right thing.  */
2311     ctx->base.pc_next += 2;
2312     return true;
2313 }
2314 
2315 static void sh4_tr_translate_insn(DisasContextBase *dcbase, CPUState *cs)
2316 {
2317     CPUSH4State *env = cs->env_ptr;
2318     DisasContext *ctx = container_of(dcbase, DisasContext, base);
2319 
2320 #ifdef CONFIG_USER_ONLY
2321     if (unlikely(ctx->envflags & GUSA_MASK)
2322         && !(ctx->envflags & GUSA_EXCLUSIVE)) {
2323         /* We're in an gUSA region, and we have not already fallen
2324            back on using an exclusive region.  Attempt to parse the
2325            region into a single supported atomic operation.  Failure
2326            is handled within the parser by raising an exception to
2327            retry using an exclusive region.  */
2328         decode_gusa(ctx, env);
2329         return;
2330     }
2331 #endif
2332 
2333     ctx->opcode = translator_lduw(env, ctx->base.pc_next);
2334     decode_opc(ctx);
2335     ctx->base.pc_next += 2;
2336 }
2337 
2338 static void sh4_tr_tb_stop(DisasContextBase *dcbase, CPUState *cs)
2339 {
2340     DisasContext *ctx = container_of(dcbase, DisasContext, base);
2341 
2342     if (ctx->tbflags & GUSA_EXCLUSIVE) {
2343         /* Ending the region of exclusivity.  Clear the bits.  */
2344         ctx->envflags &= ~GUSA_MASK;
2345     }
2346 
2347     switch (ctx->base.is_jmp) {
2348     case DISAS_STOP:
2349         gen_save_cpu_state(ctx, true);
2350         if (ctx->base.singlestep_enabled) {
2351             gen_helper_debug(cpu_env);
2352         } else {
2353             tcg_gen_exit_tb(NULL, 0);
2354         }
2355         break;
2356     case DISAS_NEXT:
2357     case DISAS_TOO_MANY:
2358         gen_save_cpu_state(ctx, false);
2359         gen_goto_tb(ctx, 0, ctx->base.pc_next);
2360         break;
2361     case DISAS_NORETURN:
2362         break;
2363     default:
2364         g_assert_not_reached();
2365     }
2366 }
2367 
2368 static void sh4_tr_disas_log(const DisasContextBase *dcbase, CPUState *cs)
2369 {
2370     qemu_log("IN:\n");  /* , lookup_symbol(dcbase->pc_first)); */
2371     log_target_disas(cs, dcbase->pc_first, dcbase->tb->size);
2372 }
2373 
2374 static const TranslatorOps sh4_tr_ops = {
2375     .init_disas_context = sh4_tr_init_disas_context,
2376     .tb_start           = sh4_tr_tb_start,
2377     .insn_start         = sh4_tr_insn_start,
2378     .breakpoint_check   = sh4_tr_breakpoint_check,
2379     .translate_insn     = sh4_tr_translate_insn,
2380     .tb_stop            = sh4_tr_tb_stop,
2381     .disas_log          = sh4_tr_disas_log,
2382 };
2383 
2384 void gen_intermediate_code(CPUState *cs, TranslationBlock *tb, int max_insns)
2385 {
2386     DisasContext ctx;
2387 
2388     translator_loop(&sh4_tr_ops, &ctx.base, cs, tb, max_insns);
2389 }
2390 
2391 void restore_state_to_opc(CPUSH4State *env, TranslationBlock *tb,
2392                           target_ulong *data)
2393 {
2394     env->pc = data[0];
2395     env->flags = data[1];
2396     /* Theoretically delayed_pc should also be restored. In practice the
2397        branch instruction is re-executed after exception, so the delayed
2398        branch target will be recomputed. */
2399 }
2400