xref: /openbmc/qemu/target/sh4/translate.c (revision 764a6ee9)
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 #include "qemu/osdep.h"
21 #include "cpu.h"
22 #include "exec/exec-all.h"
23 #include "tcg/tcg-op.h"
24 #include "exec/helper-proto.h"
25 #include "exec/helper-gen.h"
26 #include "exec/translator.h"
27 #include "exec/log.h"
28 #include "qemu/qemu-print.h"
29 
30 #define HELPER_H "helper.h"
31 #include "exec/helper-info.c.inc"
32 #undef  HELPER_H
33 
34 
35 typedef struct DisasContext {
36     DisasContextBase base;
37 
38     uint32_t tbflags;  /* should stay unmodified during the TB translation */
39     uint32_t envflags; /* should stay in sync with env->flags using TCG ops */
40     int memidx;
41     int gbank;
42     int fbank;
43     uint32_t delayed_pc;
44     uint32_t features;
45 
46     uint16_t opcode;
47 
48     bool has_movcal;
49 } DisasContext;
50 
51 #if defined(CONFIG_USER_ONLY)
52 #define IS_USER(ctx) 1
53 #define UNALIGN(C)   (ctx->tbflags & TB_FLAG_UNALIGN ? MO_UNALN : MO_ALIGN)
54 #else
55 #define IS_USER(ctx) (!(ctx->tbflags & (1u << SR_MD)))
56 #define UNALIGN(C)   0
57 #endif
58 
59 /* Target-specific values for ctx->base.is_jmp.  */
60 /* We want to exit back to the cpu loop for some reason.
61    Usually this is to recognize interrupts immediately.  */
62 #define DISAS_STOP    DISAS_TARGET_0
63 
64 /* global register indexes */
65 static TCGv cpu_gregs[32];
66 static TCGv cpu_sr, cpu_sr_m, cpu_sr_q, cpu_sr_t;
67 static TCGv cpu_pc, cpu_ssr, cpu_spc, cpu_gbr;
68 static TCGv cpu_vbr, cpu_sgr, cpu_dbr, cpu_mach, cpu_macl;
69 static TCGv cpu_pr, cpu_fpscr, cpu_fpul;
70 static TCGv cpu_lock_addr, cpu_lock_value;
71 static TCGv cpu_fregs[32];
72 
73 /* internal register indexes */
74 static TCGv cpu_flags, cpu_delayed_pc, cpu_delayed_cond;
75 
76 void sh4_translate_init(void)
77 {
78     int i;
79     static const char * const gregnames[24] = {
80         "R0_BANK0", "R1_BANK0", "R2_BANK0", "R3_BANK0",
81         "R4_BANK0", "R5_BANK0", "R6_BANK0", "R7_BANK0",
82         "R8", "R9", "R10", "R11", "R12", "R13", "R14", "R15",
83         "R0_BANK1", "R1_BANK1", "R2_BANK1", "R3_BANK1",
84         "R4_BANK1", "R5_BANK1", "R6_BANK1", "R7_BANK1"
85     };
86     static const char * const fregnames[32] = {
87          "FPR0_BANK0",  "FPR1_BANK0",  "FPR2_BANK0",  "FPR3_BANK0",
88          "FPR4_BANK0",  "FPR5_BANK0",  "FPR6_BANK0",  "FPR7_BANK0",
89          "FPR8_BANK0",  "FPR9_BANK0", "FPR10_BANK0", "FPR11_BANK0",
90         "FPR12_BANK0", "FPR13_BANK0", "FPR14_BANK0", "FPR15_BANK0",
91          "FPR0_BANK1",  "FPR1_BANK1",  "FPR2_BANK1",  "FPR3_BANK1",
92          "FPR4_BANK1",  "FPR5_BANK1",  "FPR6_BANK1",  "FPR7_BANK1",
93          "FPR8_BANK1",  "FPR9_BANK1", "FPR10_BANK1", "FPR11_BANK1",
94         "FPR12_BANK1", "FPR13_BANK1", "FPR14_BANK1", "FPR15_BANK1",
95     };
96 
97     for (i = 0; i < 24; i++) {
98         cpu_gregs[i] = tcg_global_mem_new_i32(tcg_env,
99                                               offsetof(CPUSH4State, gregs[i]),
100                                               gregnames[i]);
101     }
102     memcpy(cpu_gregs + 24, cpu_gregs + 8, 8 * sizeof(TCGv));
103 
104     cpu_pc = tcg_global_mem_new_i32(tcg_env,
105                                     offsetof(CPUSH4State, pc), "PC");
106     cpu_sr = tcg_global_mem_new_i32(tcg_env,
107                                     offsetof(CPUSH4State, sr), "SR");
108     cpu_sr_m = tcg_global_mem_new_i32(tcg_env,
109                                       offsetof(CPUSH4State, sr_m), "SR_M");
110     cpu_sr_q = tcg_global_mem_new_i32(tcg_env,
111                                       offsetof(CPUSH4State, sr_q), "SR_Q");
112     cpu_sr_t = tcg_global_mem_new_i32(tcg_env,
113                                       offsetof(CPUSH4State, sr_t), "SR_T");
114     cpu_ssr = tcg_global_mem_new_i32(tcg_env,
115                                      offsetof(CPUSH4State, ssr), "SSR");
116     cpu_spc = tcg_global_mem_new_i32(tcg_env,
117                                      offsetof(CPUSH4State, spc), "SPC");
118     cpu_gbr = tcg_global_mem_new_i32(tcg_env,
119                                      offsetof(CPUSH4State, gbr), "GBR");
120     cpu_vbr = tcg_global_mem_new_i32(tcg_env,
121                                      offsetof(CPUSH4State, vbr), "VBR");
122     cpu_sgr = tcg_global_mem_new_i32(tcg_env,
123                                      offsetof(CPUSH4State, sgr), "SGR");
124     cpu_dbr = tcg_global_mem_new_i32(tcg_env,
125                                      offsetof(CPUSH4State, dbr), "DBR");
126     cpu_mach = tcg_global_mem_new_i32(tcg_env,
127                                       offsetof(CPUSH4State, mach), "MACH");
128     cpu_macl = tcg_global_mem_new_i32(tcg_env,
129                                       offsetof(CPUSH4State, macl), "MACL");
130     cpu_pr = tcg_global_mem_new_i32(tcg_env,
131                                     offsetof(CPUSH4State, pr), "PR");
132     cpu_fpscr = tcg_global_mem_new_i32(tcg_env,
133                                        offsetof(CPUSH4State, fpscr), "FPSCR");
134     cpu_fpul = tcg_global_mem_new_i32(tcg_env,
135                                       offsetof(CPUSH4State, fpul), "FPUL");
136 
137     cpu_flags = tcg_global_mem_new_i32(tcg_env,
138                                        offsetof(CPUSH4State, flags), "_flags_");
139     cpu_delayed_pc = tcg_global_mem_new_i32(tcg_env,
140                                             offsetof(CPUSH4State, delayed_pc),
141                                             "_delayed_pc_");
142     cpu_delayed_cond = tcg_global_mem_new_i32(tcg_env,
143                                               offsetof(CPUSH4State,
144                                                        delayed_cond),
145                                               "_delayed_cond_");
146     cpu_lock_addr = tcg_global_mem_new_i32(tcg_env,
147                                            offsetof(CPUSH4State, lock_addr),
148                                            "_lock_addr_");
149     cpu_lock_value = tcg_global_mem_new_i32(tcg_env,
150                                             offsetof(CPUSH4State, lock_value),
151                                             "_lock_value_");
152 
153     for (i = 0; i < 32; i++)
154         cpu_fregs[i] = tcg_global_mem_new_i32(tcg_env,
155                                               offsetof(CPUSH4State, fregs[i]),
156                                               fregnames[i]);
157 }
158 
159 void superh_cpu_dump_state(CPUState *cs, FILE *f, int flags)
160 {
161     CPUSH4State *env = cpu_env(cs);
162     int i;
163 
164     qemu_fprintf(f, "pc=0x%08x sr=0x%08x pr=0x%08x fpscr=0x%08x\n",
165                  env->pc, cpu_read_sr(env), env->pr, env->fpscr);
166     qemu_fprintf(f, "spc=0x%08x ssr=0x%08x gbr=0x%08x vbr=0x%08x\n",
167                  env->spc, env->ssr, env->gbr, env->vbr);
168     qemu_fprintf(f, "sgr=0x%08x dbr=0x%08x delayed_pc=0x%08x fpul=0x%08x\n",
169                  env->sgr, env->dbr, env->delayed_pc, env->fpul);
170     for (i = 0; i < 24; i += 4) {
171         qemu_fprintf(f, "r%d=0x%08x r%d=0x%08x r%d=0x%08x r%d=0x%08x\n",
172                      i, env->gregs[i], i + 1, env->gregs[i + 1],
173                      i + 2, env->gregs[i + 2], i + 3, env->gregs[i + 3]);
174     }
175     if (env->flags & TB_FLAG_DELAY_SLOT) {
176         qemu_fprintf(f, "in delay slot (delayed_pc=0x%08x)\n",
177                      env->delayed_pc);
178     } else if (env->flags & TB_FLAG_DELAY_SLOT_COND) {
179         qemu_fprintf(f, "in conditional delay slot (delayed_pc=0x%08x)\n",
180                      env->delayed_pc);
181     } else if (env->flags & TB_FLAG_DELAY_SLOT_RTE) {
182         qemu_fprintf(f, "in rte delay slot (delayed_pc=0x%08x)\n",
183                      env->delayed_pc);
184     }
185 }
186 
187 static void gen_read_sr(TCGv dst)
188 {
189     TCGv t0 = tcg_temp_new();
190     tcg_gen_shli_i32(t0, cpu_sr_q, SR_Q);
191     tcg_gen_or_i32(dst, dst, t0);
192     tcg_gen_shli_i32(t0, cpu_sr_m, SR_M);
193     tcg_gen_or_i32(dst, dst, t0);
194     tcg_gen_shli_i32(t0, cpu_sr_t, SR_T);
195     tcg_gen_or_i32(dst, cpu_sr, t0);
196 }
197 
198 static void gen_write_sr(TCGv src)
199 {
200     tcg_gen_andi_i32(cpu_sr, src,
201                      ~((1u << SR_Q) | (1u << SR_M) | (1u << SR_T)));
202     tcg_gen_extract_i32(cpu_sr_q, src, SR_Q, 1);
203     tcg_gen_extract_i32(cpu_sr_m, src, SR_M, 1);
204     tcg_gen_extract_i32(cpu_sr_t, src, SR_T, 1);
205 }
206 
207 static inline void gen_save_cpu_state(DisasContext *ctx, bool save_pc)
208 {
209     if (save_pc) {
210         tcg_gen_movi_i32(cpu_pc, ctx->base.pc_next);
211     }
212     if (ctx->delayed_pc != (uint32_t) -1) {
213         tcg_gen_movi_i32(cpu_delayed_pc, ctx->delayed_pc);
214     }
215     if ((ctx->tbflags & TB_FLAG_ENVFLAGS_MASK) != ctx->envflags) {
216         tcg_gen_movi_i32(cpu_flags, ctx->envflags);
217     }
218 }
219 
220 static inline bool use_exit_tb(DisasContext *ctx)
221 {
222     return (ctx->tbflags & TB_FLAG_GUSA_EXCLUSIVE) != 0;
223 }
224 
225 static bool use_goto_tb(DisasContext *ctx, target_ulong dest)
226 {
227     if (use_exit_tb(ctx)) {
228         return false;
229     }
230     return translator_use_goto_tb(&ctx->base, dest);
231 }
232 
233 static void gen_goto_tb(DisasContext *ctx, int n, target_ulong dest)
234 {
235     if (use_goto_tb(ctx, dest)) {
236         tcg_gen_goto_tb(n);
237         tcg_gen_movi_i32(cpu_pc, dest);
238         tcg_gen_exit_tb(ctx->base.tb, n);
239     } else {
240         tcg_gen_movi_i32(cpu_pc, dest);
241         if (use_exit_tb(ctx)) {
242             tcg_gen_exit_tb(NULL, 0);
243         } else {
244             tcg_gen_lookup_and_goto_ptr();
245         }
246     }
247     ctx->base.is_jmp = DISAS_NORETURN;
248 }
249 
250 static void gen_jump(DisasContext * ctx)
251 {
252     if (ctx->delayed_pc == -1) {
253         /* Target is not statically known, it comes necessarily from a
254            delayed jump as immediate jump are conditinal jumps */
255         tcg_gen_mov_i32(cpu_pc, cpu_delayed_pc);
256         tcg_gen_discard_i32(cpu_delayed_pc);
257         if (use_exit_tb(ctx)) {
258             tcg_gen_exit_tb(NULL, 0);
259         } else {
260             tcg_gen_lookup_and_goto_ptr();
261         }
262         ctx->base.is_jmp = DISAS_NORETURN;
263     } else {
264         gen_goto_tb(ctx, 0, ctx->delayed_pc);
265     }
266 }
267 
268 /* Immediate conditional jump (bt or bf) */
269 static void gen_conditional_jump(DisasContext *ctx, target_ulong dest,
270                                  bool jump_if_true)
271 {
272     TCGLabel *l1 = gen_new_label();
273     TCGCond cond_not_taken = jump_if_true ? TCG_COND_EQ : TCG_COND_NE;
274 
275     if (ctx->tbflags & TB_FLAG_GUSA_EXCLUSIVE) {
276         /* When in an exclusive region, we must continue to the end.
277            Therefore, exit the region on a taken branch, but otherwise
278            fall through to the next instruction.  */
279         tcg_gen_brcondi_i32(cond_not_taken, cpu_sr_t, 0, l1);
280         tcg_gen_movi_i32(cpu_flags, ctx->envflags & ~TB_FLAG_GUSA_MASK);
281         /* Note that this won't actually use a goto_tb opcode because we
282            disallow it in use_goto_tb, but it handles exit + singlestep.  */
283         gen_goto_tb(ctx, 0, dest);
284         gen_set_label(l1);
285         ctx->base.is_jmp = DISAS_NEXT;
286         return;
287     }
288 
289     gen_save_cpu_state(ctx, false);
290     tcg_gen_brcondi_i32(cond_not_taken, cpu_sr_t, 0, l1);
291     gen_goto_tb(ctx, 0, dest);
292     gen_set_label(l1);
293     gen_goto_tb(ctx, 1, ctx->base.pc_next + 2);
294     ctx->base.is_jmp = DISAS_NORETURN;
295 }
296 
297 /* Delayed conditional jump (bt or bf) */
298 static void gen_delayed_conditional_jump(DisasContext * ctx)
299 {
300     TCGLabel *l1 = gen_new_label();
301     TCGv ds = tcg_temp_new();
302 
303     tcg_gen_mov_i32(ds, cpu_delayed_cond);
304     tcg_gen_discard_i32(cpu_delayed_cond);
305 
306     if (ctx->tbflags & TB_FLAG_GUSA_EXCLUSIVE) {
307         /* When in an exclusive region, we must continue to the end.
308            Therefore, exit the region on a taken branch, but otherwise
309            fall through to the next instruction.  */
310         tcg_gen_brcondi_i32(TCG_COND_EQ, ds, 0, l1);
311 
312         /* Leave the gUSA region.  */
313         tcg_gen_movi_i32(cpu_flags, ctx->envflags & ~TB_FLAG_GUSA_MASK);
314         gen_jump(ctx);
315 
316         gen_set_label(l1);
317         ctx->base.is_jmp = DISAS_NEXT;
318         return;
319     }
320 
321     tcg_gen_brcondi_i32(TCG_COND_NE, ds, 0, l1);
322     gen_goto_tb(ctx, 1, ctx->base.pc_next + 2);
323     gen_set_label(l1);
324     gen_jump(ctx);
325 }
326 
327 static inline void gen_load_fpr64(DisasContext *ctx, TCGv_i64 t, int reg)
328 {
329     /* We have already signaled illegal instruction for odd Dr.  */
330     tcg_debug_assert((reg & 1) == 0);
331     reg ^= ctx->fbank;
332     tcg_gen_concat_i32_i64(t, cpu_fregs[reg + 1], cpu_fregs[reg]);
333 }
334 
335 static inline void gen_store_fpr64(DisasContext *ctx, TCGv_i64 t, int reg)
336 {
337     /* We have already signaled illegal instruction for odd Dr.  */
338     tcg_debug_assert((reg & 1) == 0);
339     reg ^= ctx->fbank;
340     tcg_gen_extr_i64_i32(cpu_fregs[reg + 1], cpu_fregs[reg], t);
341 }
342 
343 #define B3_0 (ctx->opcode & 0xf)
344 #define B6_4 ((ctx->opcode >> 4) & 0x7)
345 #define B7_4 ((ctx->opcode >> 4) & 0xf)
346 #define B7_0 (ctx->opcode & 0xff)
347 #define B7_0s ((int32_t) (int8_t) (ctx->opcode & 0xff))
348 #define B11_0s (ctx->opcode & 0x800 ? 0xfffff000 | (ctx->opcode & 0xfff) : \
349   (ctx->opcode & 0xfff))
350 #define B11_8 ((ctx->opcode >> 8) & 0xf)
351 #define B15_12 ((ctx->opcode >> 12) & 0xf)
352 
353 #define REG(x)     cpu_gregs[(x) ^ ctx->gbank]
354 #define ALTREG(x)  cpu_gregs[(x) ^ ctx->gbank ^ 0x10]
355 #define FREG(x)    cpu_fregs[(x) ^ ctx->fbank]
356 
357 #define XHACK(x) ((((x) & 1 ) << 4) | ((x) & 0xe))
358 
359 #define CHECK_NOT_DELAY_SLOT \
360     if (ctx->envflags & TB_FLAG_DELAY_SLOT_MASK) {  \
361         goto do_illegal_slot;                       \
362     }
363 
364 #define CHECK_PRIVILEGED \
365     if (IS_USER(ctx)) {                     \
366         goto do_illegal;                    \
367     }
368 
369 #define CHECK_FPU_ENABLED \
370     if (ctx->tbflags & (1u << SR_FD)) {     \
371         goto do_fpu_disabled;               \
372     }
373 
374 #define CHECK_FPSCR_PR_0 \
375     if (ctx->tbflags & FPSCR_PR) {          \
376         goto do_illegal;                    \
377     }
378 
379 #define CHECK_FPSCR_PR_1 \
380     if (!(ctx->tbflags & FPSCR_PR)) {       \
381         goto do_illegal;                    \
382     }
383 
384 #define CHECK_SH4A \
385     if (!(ctx->features & SH_FEATURE_SH4A)) { \
386         goto do_illegal;                      \
387     }
388 
389 static void _decode_opc(DisasContext * ctx)
390 {
391     /* This code tries to make movcal emulation sufficiently
392        accurate for Linux purposes.  This instruction writes
393        memory, and prior to that, always allocates a cache line.
394        It is used in two contexts:
395        - in memcpy, where data is copied in blocks, the first write
396        of to a block uses movca.l for performance.
397        - in arch/sh/mm/cache-sh4.c, movcal.l + ocbi combination is used
398        to flush the cache. Here, the data written by movcal.l is never
399        written to memory, and the data written is just bogus.
400 
401        To simulate this, we simulate movcal.l, we store the value to memory,
402        but we also remember the previous content. If we see ocbi, we check
403        if movcal.l for that address was done previously. If so, the write should
404        not have hit the memory, so we restore the previous content.
405        When we see an instruction that is neither movca.l
406        nor ocbi, the previous content is discarded.
407 
408        To optimize, we only try to flush stores when we're at the start of
409        TB, or if we already saw movca.l in this TB and did not flush stores
410        yet.  */
411     if (ctx->has_movcal)
412     {
413         int opcode = ctx->opcode & 0xf0ff;
414         if (opcode != 0x0093 /* ocbi */
415             && opcode != 0x00c3 /* movca.l */)
416         {
417             gen_helper_discard_movcal_backup(tcg_env);
418             ctx->has_movcal = 0;
419         }
420     }
421 
422 #if 0
423     fprintf(stderr, "Translating opcode 0x%04x\n", ctx->opcode);
424 #endif
425 
426     switch (ctx->opcode) {
427     case 0x0019: /* div0u */
428         tcg_gen_movi_i32(cpu_sr_m, 0);
429         tcg_gen_movi_i32(cpu_sr_q, 0);
430         tcg_gen_movi_i32(cpu_sr_t, 0);
431         return;
432     case 0x000b: /* rts */
433         CHECK_NOT_DELAY_SLOT
434         tcg_gen_mov_i32(cpu_delayed_pc, cpu_pr);
435         ctx->envflags |= TB_FLAG_DELAY_SLOT;
436         ctx->delayed_pc = (uint32_t) - 1;
437         return;
438     case 0x0028: /* clrmac */
439         tcg_gen_movi_i32(cpu_mach, 0);
440         tcg_gen_movi_i32(cpu_macl, 0);
441         return;
442     case 0x0048: /* clrs */
443         tcg_gen_andi_i32(cpu_sr, cpu_sr, ~(1u << SR_S));
444         return;
445     case 0x0008: /* clrt */
446         tcg_gen_movi_i32(cpu_sr_t, 0);
447         return;
448     case 0x0038: /* ldtlb */
449         CHECK_PRIVILEGED
450         gen_helper_ldtlb(tcg_env);
451         return;
452     case 0x002b: /* rte */
453         CHECK_PRIVILEGED
454         CHECK_NOT_DELAY_SLOT
455         gen_write_sr(cpu_ssr);
456         tcg_gen_mov_i32(cpu_delayed_pc, cpu_spc);
457         ctx->envflags |= TB_FLAG_DELAY_SLOT_RTE;
458         ctx->delayed_pc = (uint32_t) - 1;
459         ctx->base.is_jmp = DISAS_STOP;
460         return;
461     case 0x0058: /* sets */
462         tcg_gen_ori_i32(cpu_sr, cpu_sr, (1u << SR_S));
463         return;
464     case 0x0018: /* sett */
465         tcg_gen_movi_i32(cpu_sr_t, 1);
466         return;
467     case 0xfbfd: /* frchg */
468         CHECK_FPSCR_PR_0
469         tcg_gen_xori_i32(cpu_fpscr, cpu_fpscr, FPSCR_FR);
470         ctx->base.is_jmp = DISAS_STOP;
471         return;
472     case 0xf3fd: /* fschg */
473         CHECK_FPSCR_PR_0
474         tcg_gen_xori_i32(cpu_fpscr, cpu_fpscr, FPSCR_SZ);
475         ctx->base.is_jmp = DISAS_STOP;
476         return;
477     case 0xf7fd: /* fpchg */
478         CHECK_SH4A
479         tcg_gen_xori_i32(cpu_fpscr, cpu_fpscr, FPSCR_PR);
480         ctx->base.is_jmp = DISAS_STOP;
481         return;
482     case 0x0009: /* nop */
483         return;
484     case 0x001b: /* sleep */
485         CHECK_PRIVILEGED
486         tcg_gen_movi_i32(cpu_pc, ctx->base.pc_next + 2);
487         gen_helper_sleep(tcg_env);
488         return;
489     }
490 
491     switch (ctx->opcode & 0xf000) {
492     case 0x1000: /* mov.l Rm,@(disp,Rn) */
493         {
494             TCGv addr = tcg_temp_new();
495             tcg_gen_addi_i32(addr, REG(B11_8), B3_0 * 4);
496             tcg_gen_qemu_st_i32(REG(B7_4), addr, ctx->memidx,
497                                 MO_TEUL | UNALIGN(ctx));
498         }
499         return;
500     case 0x5000: /* mov.l @(disp,Rm),Rn */
501         {
502             TCGv addr = tcg_temp_new();
503             tcg_gen_addi_i32(addr, REG(B7_4), B3_0 * 4);
504             tcg_gen_qemu_ld_i32(REG(B11_8), addr, ctx->memidx,
505                                 MO_TESL | UNALIGN(ctx));
506         }
507         return;
508     case 0xe000: /* mov #imm,Rn */
509 #ifdef CONFIG_USER_ONLY
510         /*
511          * Detect the start of a gUSA region (mov #-n, r15).
512          * If so, update envflags and end the TB.  This will allow us
513          * to see the end of the region (stored in R0) in the next TB.
514          */
515         if (B11_8 == 15 && B7_0s < 0 &&
516             (tb_cflags(ctx->base.tb) & CF_PARALLEL)) {
517             ctx->envflags =
518                 deposit32(ctx->envflags, TB_FLAG_GUSA_SHIFT, 8, B7_0s);
519             ctx->base.is_jmp = DISAS_STOP;
520         }
521 #endif
522         tcg_gen_movi_i32(REG(B11_8), B7_0s);
523         return;
524     case 0x9000: /* mov.w @(disp,PC),Rn */
525         CHECK_NOT_DELAY_SLOT
526         {
527             TCGv addr = tcg_constant_i32(ctx->base.pc_next + 4 + B7_0 * 2);
528             tcg_gen_qemu_ld_i32(REG(B11_8), addr, ctx->memidx,
529                                 MO_TESW | MO_ALIGN);
530         }
531         return;
532     case 0xd000: /* mov.l @(disp,PC),Rn */
533         CHECK_NOT_DELAY_SLOT
534         {
535             TCGv addr = tcg_constant_i32((ctx->base.pc_next + 4 + B7_0 * 4) & ~3);
536             tcg_gen_qemu_ld_i32(REG(B11_8), addr, ctx->memidx,
537                                 MO_TESL | MO_ALIGN);
538         }
539         return;
540     case 0x7000: /* add #imm,Rn */
541         tcg_gen_addi_i32(REG(B11_8), REG(B11_8), B7_0s);
542         return;
543     case 0xa000: /* bra disp */
544         CHECK_NOT_DELAY_SLOT
545         ctx->delayed_pc = ctx->base.pc_next + 4 + B11_0s * 2;
546         ctx->envflags |= TB_FLAG_DELAY_SLOT;
547         return;
548     case 0xb000: /* bsr disp */
549         CHECK_NOT_DELAY_SLOT
550         tcg_gen_movi_i32(cpu_pr, ctx->base.pc_next + 4);
551         ctx->delayed_pc = ctx->base.pc_next + 4 + B11_0s * 2;
552         ctx->envflags |= TB_FLAG_DELAY_SLOT;
553         return;
554     }
555 
556     switch (ctx->opcode & 0xf00f) {
557     case 0x6003: /* mov Rm,Rn */
558         tcg_gen_mov_i32(REG(B11_8), REG(B7_4));
559         return;
560     case 0x2000: /* mov.b Rm,@Rn */
561         tcg_gen_qemu_st_i32(REG(B7_4), REG(B11_8), ctx->memidx, MO_UB);
562         return;
563     case 0x2001: /* mov.w Rm,@Rn */
564         tcg_gen_qemu_st_i32(REG(B7_4), REG(B11_8), ctx->memidx,
565                             MO_TEUW | UNALIGN(ctx));
566         return;
567     case 0x2002: /* mov.l Rm,@Rn */
568         tcg_gen_qemu_st_i32(REG(B7_4), REG(B11_8), ctx->memidx,
569                             MO_TEUL | UNALIGN(ctx));
570         return;
571     case 0x6000: /* mov.b @Rm,Rn */
572         tcg_gen_qemu_ld_i32(REG(B11_8), REG(B7_4), ctx->memidx, MO_SB);
573         return;
574     case 0x6001: /* mov.w @Rm,Rn */
575         tcg_gen_qemu_ld_i32(REG(B11_8), REG(B7_4), ctx->memidx,
576                             MO_TESW | UNALIGN(ctx));
577         return;
578     case 0x6002: /* mov.l @Rm,Rn */
579         tcg_gen_qemu_ld_i32(REG(B11_8), REG(B7_4), ctx->memidx,
580                             MO_TESL | UNALIGN(ctx));
581         return;
582     case 0x2004: /* mov.b Rm,@-Rn */
583         {
584             TCGv addr = tcg_temp_new();
585             tcg_gen_subi_i32(addr, REG(B11_8), 1);
586             /* might cause re-execution */
587             tcg_gen_qemu_st_i32(REG(B7_4), addr, ctx->memidx, MO_UB);
588             tcg_gen_mov_i32(REG(B11_8), addr); /* modify register status */
589         }
590         return;
591     case 0x2005: /* mov.w Rm,@-Rn */
592         {
593             TCGv addr = tcg_temp_new();
594             tcg_gen_subi_i32(addr, REG(B11_8), 2);
595             tcg_gen_qemu_st_i32(REG(B7_4), addr, ctx->memidx,
596                                 MO_TEUW | UNALIGN(ctx));
597             tcg_gen_mov_i32(REG(B11_8), addr);
598         }
599         return;
600     case 0x2006: /* mov.l Rm,@-Rn */
601         {
602             TCGv addr = tcg_temp_new();
603             tcg_gen_subi_i32(addr, REG(B11_8), 4);
604             tcg_gen_qemu_st_i32(REG(B7_4), addr, ctx->memidx,
605                                 MO_TEUL | UNALIGN(ctx));
606             tcg_gen_mov_i32(REG(B11_8), addr);
607         }
608         return;
609     case 0x6004: /* mov.b @Rm+,Rn */
610         tcg_gen_qemu_ld_i32(REG(B11_8), REG(B7_4), ctx->memidx, MO_SB);
611         if ( B11_8 != B7_4 )
612                 tcg_gen_addi_i32(REG(B7_4), REG(B7_4), 1);
613         return;
614     case 0x6005: /* mov.w @Rm+,Rn */
615         tcg_gen_qemu_ld_i32(REG(B11_8), REG(B7_4), ctx->memidx,
616                             MO_TESW | UNALIGN(ctx));
617         if ( B11_8 != B7_4 )
618                 tcg_gen_addi_i32(REG(B7_4), REG(B7_4), 2);
619         return;
620     case 0x6006: /* mov.l @Rm+,Rn */
621         tcg_gen_qemu_ld_i32(REG(B11_8), REG(B7_4), ctx->memidx,
622                             MO_TESL | UNALIGN(ctx));
623         if ( B11_8 != B7_4 )
624                 tcg_gen_addi_i32(REG(B7_4), REG(B7_4), 4);
625         return;
626     case 0x0004: /* mov.b Rm,@(R0,Rn) */
627         {
628             TCGv addr = tcg_temp_new();
629             tcg_gen_add_i32(addr, REG(B11_8), REG(0));
630             tcg_gen_qemu_st_i32(REG(B7_4), addr, ctx->memidx, MO_UB);
631         }
632         return;
633     case 0x0005: /* mov.w Rm,@(R0,Rn) */
634         {
635             TCGv addr = tcg_temp_new();
636             tcg_gen_add_i32(addr, REG(B11_8), REG(0));
637             tcg_gen_qemu_st_i32(REG(B7_4), addr, ctx->memidx,
638                                 MO_TEUW | UNALIGN(ctx));
639         }
640         return;
641     case 0x0006: /* mov.l Rm,@(R0,Rn) */
642         {
643             TCGv addr = tcg_temp_new();
644             tcg_gen_add_i32(addr, REG(B11_8), REG(0));
645             tcg_gen_qemu_st_i32(REG(B7_4), addr, ctx->memidx,
646                                 MO_TEUL | UNALIGN(ctx));
647         }
648         return;
649     case 0x000c: /* mov.b @(R0,Rm),Rn */
650         {
651             TCGv addr = tcg_temp_new();
652             tcg_gen_add_i32(addr, REG(B7_4), REG(0));
653             tcg_gen_qemu_ld_i32(REG(B11_8), addr, ctx->memidx, MO_SB);
654         }
655         return;
656     case 0x000d: /* mov.w @(R0,Rm),Rn */
657         {
658             TCGv addr = tcg_temp_new();
659             tcg_gen_add_i32(addr, REG(B7_4), REG(0));
660             tcg_gen_qemu_ld_i32(REG(B11_8), addr, ctx->memidx,
661                                 MO_TESW | UNALIGN(ctx));
662         }
663         return;
664     case 0x000e: /* mov.l @(R0,Rm),Rn */
665         {
666             TCGv addr = tcg_temp_new();
667             tcg_gen_add_i32(addr, REG(B7_4), REG(0));
668             tcg_gen_qemu_ld_i32(REG(B11_8), addr, ctx->memidx,
669                                 MO_TESL | UNALIGN(ctx));
670         }
671         return;
672     case 0x6008: /* swap.b Rm,Rn */
673         {
674             TCGv low = tcg_temp_new();
675             tcg_gen_bswap16_i32(low, REG(B7_4), 0);
676             tcg_gen_deposit_i32(REG(B11_8), REG(B7_4), low, 0, 16);
677         }
678         return;
679     case 0x6009: /* swap.w Rm,Rn */
680         tcg_gen_rotli_i32(REG(B11_8), REG(B7_4), 16);
681         return;
682     case 0x200d: /* xtrct Rm,Rn */
683         {
684             TCGv high, low;
685             high = tcg_temp_new();
686             tcg_gen_shli_i32(high, REG(B7_4), 16);
687             low = tcg_temp_new();
688             tcg_gen_shri_i32(low, REG(B11_8), 16);
689             tcg_gen_or_i32(REG(B11_8), high, low);
690         }
691         return;
692     case 0x300c: /* add Rm,Rn */
693         tcg_gen_add_i32(REG(B11_8), REG(B11_8), REG(B7_4));
694         return;
695     case 0x300e: /* addc Rm,Rn */
696         {
697             TCGv t0, t1;
698             t0 = tcg_constant_tl(0);
699             t1 = tcg_temp_new();
700             tcg_gen_add2_i32(t1, cpu_sr_t, cpu_sr_t, t0, REG(B7_4), t0);
701             tcg_gen_add2_i32(REG(B11_8), cpu_sr_t,
702                              REG(B11_8), t0, t1, cpu_sr_t);
703         }
704         return;
705     case 0x300f: /* addv Rm,Rn */
706         {
707             TCGv Rn = REG(B11_8);
708             TCGv Rm = REG(B7_4);
709             TCGv result, t1, t2;
710 
711             result = tcg_temp_new();
712             t1 = tcg_temp_new();
713             t2 = tcg_temp_new();
714             tcg_gen_add_i32(result, Rm, Rn);
715             /* T = ((Rn ^ Rm) & (Result ^ Rn)) >> 31 */
716             tcg_gen_xor_i32(t1, result, Rn);
717             tcg_gen_xor_i32(t2, Rm, Rn);
718             tcg_gen_andc_i32(cpu_sr_t, t1, t2);
719             tcg_gen_shri_i32(cpu_sr_t, cpu_sr_t, 31);
720             tcg_gen_mov_i32(Rn, result);
721         }
722         return;
723     case 0x2009: /* and Rm,Rn */
724         tcg_gen_and_i32(REG(B11_8), REG(B11_8), REG(B7_4));
725         return;
726     case 0x3000: /* cmp/eq Rm,Rn */
727         tcg_gen_setcond_i32(TCG_COND_EQ, cpu_sr_t, REG(B11_8), REG(B7_4));
728         return;
729     case 0x3003: /* cmp/ge Rm,Rn */
730         tcg_gen_setcond_i32(TCG_COND_GE, cpu_sr_t, REG(B11_8), REG(B7_4));
731         return;
732     case 0x3007: /* cmp/gt Rm,Rn */
733         tcg_gen_setcond_i32(TCG_COND_GT, cpu_sr_t, REG(B11_8), REG(B7_4));
734         return;
735     case 0x3006: /* cmp/hi Rm,Rn */
736         tcg_gen_setcond_i32(TCG_COND_GTU, cpu_sr_t, REG(B11_8), REG(B7_4));
737         return;
738     case 0x3002: /* cmp/hs Rm,Rn */
739         tcg_gen_setcond_i32(TCG_COND_GEU, cpu_sr_t, REG(B11_8), REG(B7_4));
740         return;
741     case 0x200c: /* cmp/str Rm,Rn */
742         {
743             TCGv cmp1 = tcg_temp_new();
744             TCGv cmp2 = tcg_temp_new();
745             tcg_gen_xor_i32(cmp2, REG(B7_4), REG(B11_8));
746             tcg_gen_subi_i32(cmp1, cmp2, 0x01010101);
747             tcg_gen_andc_i32(cmp1, cmp1, cmp2);
748             tcg_gen_andi_i32(cmp1, cmp1, 0x80808080);
749             tcg_gen_setcondi_i32(TCG_COND_NE, cpu_sr_t, cmp1, 0);
750         }
751         return;
752     case 0x2007: /* div0s Rm,Rn */
753         tcg_gen_shri_i32(cpu_sr_q, REG(B11_8), 31);         /* SR_Q */
754         tcg_gen_shri_i32(cpu_sr_m, REG(B7_4), 31);          /* SR_M */
755         tcg_gen_xor_i32(cpu_sr_t, cpu_sr_q, cpu_sr_m);      /* SR_T */
756         return;
757     case 0x3004: /* div1 Rm,Rn */
758         {
759             TCGv t0 = tcg_temp_new();
760             TCGv t1 = tcg_temp_new();
761             TCGv t2 = tcg_temp_new();
762             TCGv zero = tcg_constant_i32(0);
763 
764             /* shift left arg1, saving the bit being pushed out and inserting
765                T on the right */
766             tcg_gen_shri_i32(t0, REG(B11_8), 31);
767             tcg_gen_shli_i32(REG(B11_8), REG(B11_8), 1);
768             tcg_gen_or_i32(REG(B11_8), REG(B11_8), cpu_sr_t);
769 
770             /* Add or subtract arg0 from arg1 depending if Q == M. To avoid
771                using 64-bit temps, we compute arg0's high part from q ^ m, so
772                that it is 0x00000000 when adding the value or 0xffffffff when
773                subtracting it. */
774             tcg_gen_xor_i32(t1, cpu_sr_q, cpu_sr_m);
775             tcg_gen_subi_i32(t1, t1, 1);
776             tcg_gen_neg_i32(t2, REG(B7_4));
777             tcg_gen_movcond_i32(TCG_COND_EQ, t2, t1, zero, REG(B7_4), t2);
778             tcg_gen_add2_i32(REG(B11_8), t1, REG(B11_8), zero, t2, t1);
779 
780             /* compute T and Q depending on carry */
781             tcg_gen_andi_i32(t1, t1, 1);
782             tcg_gen_xor_i32(t1, t1, t0);
783             tcg_gen_xori_i32(cpu_sr_t, t1, 1);
784             tcg_gen_xor_i32(cpu_sr_q, cpu_sr_m, t1);
785         }
786         return;
787     case 0x300d: /* dmuls.l Rm,Rn */
788         tcg_gen_muls2_i32(cpu_macl, cpu_mach, REG(B7_4), REG(B11_8));
789         return;
790     case 0x3005: /* dmulu.l Rm,Rn */
791         tcg_gen_mulu2_i32(cpu_macl, cpu_mach, REG(B7_4), REG(B11_8));
792         return;
793     case 0x600e: /* exts.b Rm,Rn */
794         tcg_gen_ext8s_i32(REG(B11_8), REG(B7_4));
795         return;
796     case 0x600f: /* exts.w Rm,Rn */
797         tcg_gen_ext16s_i32(REG(B11_8), REG(B7_4));
798         return;
799     case 0x600c: /* extu.b Rm,Rn */
800         tcg_gen_ext8u_i32(REG(B11_8), REG(B7_4));
801         return;
802     case 0x600d: /* extu.w Rm,Rn */
803         tcg_gen_ext16u_i32(REG(B11_8), REG(B7_4));
804         return;
805     case 0x000f: /* mac.l @Rm+,@Rn+ */
806         {
807             TCGv arg0, arg1;
808             arg0 = tcg_temp_new();
809             tcg_gen_qemu_ld_i32(arg0, REG(B7_4), ctx->memidx,
810                                 MO_TESL | MO_ALIGN);
811             arg1 = tcg_temp_new();
812             tcg_gen_qemu_ld_i32(arg1, REG(B11_8), ctx->memidx,
813                                 MO_TESL | MO_ALIGN);
814             gen_helper_macl(tcg_env, arg0, arg1);
815             tcg_gen_addi_i32(REG(B7_4), REG(B7_4), 4);
816             tcg_gen_addi_i32(REG(B11_8), REG(B11_8), 4);
817         }
818         return;
819     case 0x400f: /* mac.w @Rm+,@Rn+ */
820         {
821             TCGv arg0, arg1;
822             arg0 = tcg_temp_new();
823             tcg_gen_qemu_ld_i32(arg0, REG(B7_4), ctx->memidx,
824                                 MO_TESW | MO_ALIGN);
825             arg1 = tcg_temp_new();
826             tcg_gen_qemu_ld_i32(arg1, REG(B11_8), ctx->memidx,
827                                 MO_TESW | MO_ALIGN);
828             gen_helper_macw(tcg_env, arg0, arg1);
829             tcg_gen_addi_i32(REG(B11_8), REG(B11_8), 2);
830             tcg_gen_addi_i32(REG(B7_4), REG(B7_4), 2);
831         }
832         return;
833     case 0x0007: /* mul.l Rm,Rn */
834         tcg_gen_mul_i32(cpu_macl, REG(B7_4), REG(B11_8));
835         return;
836     case 0x200f: /* muls.w Rm,Rn */
837         {
838             TCGv arg0, arg1;
839             arg0 = tcg_temp_new();
840             tcg_gen_ext16s_i32(arg0, REG(B7_4));
841             arg1 = tcg_temp_new();
842             tcg_gen_ext16s_i32(arg1, REG(B11_8));
843             tcg_gen_mul_i32(cpu_macl, arg0, arg1);
844         }
845         return;
846     case 0x200e: /* mulu.w Rm,Rn */
847         {
848             TCGv arg0, arg1;
849             arg0 = tcg_temp_new();
850             tcg_gen_ext16u_i32(arg0, REG(B7_4));
851             arg1 = tcg_temp_new();
852             tcg_gen_ext16u_i32(arg1, REG(B11_8));
853             tcg_gen_mul_i32(cpu_macl, arg0, arg1);
854         }
855         return;
856     case 0x600b: /* neg Rm,Rn */
857         tcg_gen_neg_i32(REG(B11_8), REG(B7_4));
858         return;
859     case 0x600a: /* negc Rm,Rn */
860         {
861             TCGv t0 = tcg_constant_i32(0);
862             tcg_gen_add2_i32(REG(B11_8), cpu_sr_t,
863                              REG(B7_4), t0, cpu_sr_t, t0);
864             tcg_gen_sub2_i32(REG(B11_8), cpu_sr_t,
865                              t0, t0, REG(B11_8), cpu_sr_t);
866             tcg_gen_andi_i32(cpu_sr_t, cpu_sr_t, 1);
867         }
868         return;
869     case 0x6007: /* not Rm,Rn */
870         tcg_gen_not_i32(REG(B11_8), REG(B7_4));
871         return;
872     case 0x200b: /* or Rm,Rn */
873         tcg_gen_or_i32(REG(B11_8), REG(B11_8), REG(B7_4));
874         return;
875     case 0x400c: /* shad Rm,Rn */
876         {
877             TCGv t0 = tcg_temp_new();
878             TCGv t1 = tcg_temp_new();
879             TCGv t2 = tcg_temp_new();
880 
881             tcg_gen_andi_i32(t0, REG(B7_4), 0x1f);
882 
883             /* positive case: shift to the left */
884             tcg_gen_shl_i32(t1, REG(B11_8), t0);
885 
886             /* negative case: shift to the right in two steps to
887                correctly handle the -32 case */
888             tcg_gen_xori_i32(t0, t0, 0x1f);
889             tcg_gen_sar_i32(t2, REG(B11_8), t0);
890             tcg_gen_sari_i32(t2, t2, 1);
891 
892             /* select between the two cases */
893             tcg_gen_movi_i32(t0, 0);
894             tcg_gen_movcond_i32(TCG_COND_GE, REG(B11_8), REG(B7_4), t0, t1, t2);
895         }
896         return;
897     case 0x400d: /* shld Rm,Rn */
898         {
899             TCGv t0 = tcg_temp_new();
900             TCGv t1 = tcg_temp_new();
901             TCGv t2 = tcg_temp_new();
902 
903             tcg_gen_andi_i32(t0, REG(B7_4), 0x1f);
904 
905             /* positive case: shift to the left */
906             tcg_gen_shl_i32(t1, REG(B11_8), t0);
907 
908             /* negative case: shift to the right in two steps to
909                correctly handle the -32 case */
910             tcg_gen_xori_i32(t0, t0, 0x1f);
911             tcg_gen_shr_i32(t2, REG(B11_8), t0);
912             tcg_gen_shri_i32(t2, t2, 1);
913 
914             /* select between the two cases */
915             tcg_gen_movi_i32(t0, 0);
916             tcg_gen_movcond_i32(TCG_COND_GE, REG(B11_8), REG(B7_4), t0, t1, t2);
917         }
918         return;
919     case 0x3008: /* sub Rm,Rn */
920         tcg_gen_sub_i32(REG(B11_8), REG(B11_8), REG(B7_4));
921         return;
922     case 0x300a: /* subc Rm,Rn */
923         {
924             TCGv t0, t1;
925             t0 = tcg_constant_tl(0);
926             t1 = tcg_temp_new();
927             tcg_gen_add2_i32(t1, cpu_sr_t, cpu_sr_t, t0, REG(B7_4), t0);
928             tcg_gen_sub2_i32(REG(B11_8), cpu_sr_t,
929                              REG(B11_8), t0, t1, cpu_sr_t);
930             tcg_gen_andi_i32(cpu_sr_t, cpu_sr_t, 1);
931         }
932         return;
933     case 0x300b: /* subv Rm,Rn */
934         {
935             TCGv Rn = REG(B11_8);
936             TCGv Rm = REG(B7_4);
937             TCGv result, t1, t2;
938 
939             result = tcg_temp_new();
940             t1 = tcg_temp_new();
941             t2 = tcg_temp_new();
942             tcg_gen_sub_i32(result, Rn, Rm);
943             /* T = ((Rn ^ Rm) & (Result ^ Rn)) >> 31 */
944             tcg_gen_xor_i32(t1, result, Rn);
945             tcg_gen_xor_i32(t2, Rn, Rm);
946             tcg_gen_and_i32(t1, t1, t2);
947             tcg_gen_shri_i32(cpu_sr_t, t1, 31);
948             tcg_gen_mov_i32(Rn, result);
949         }
950         return;
951     case 0x2008: /* tst Rm,Rn */
952         {
953             TCGv val = tcg_temp_new();
954             tcg_gen_and_i32(val, REG(B7_4), REG(B11_8));
955             tcg_gen_setcondi_i32(TCG_COND_EQ, cpu_sr_t, val, 0);
956         }
957         return;
958     case 0x200a: /* xor Rm,Rn */
959         tcg_gen_xor_i32(REG(B11_8), REG(B11_8), REG(B7_4));
960         return;
961     case 0xf00c: /* fmov {F,D,X}Rm,{F,D,X}Rn - FPSCR: Nothing */
962         CHECK_FPU_ENABLED
963         if (ctx->tbflags & FPSCR_SZ) {
964             int xsrc = XHACK(B7_4);
965             int xdst = XHACK(B11_8);
966             tcg_gen_mov_i32(FREG(xdst), FREG(xsrc));
967             tcg_gen_mov_i32(FREG(xdst + 1), FREG(xsrc + 1));
968         } else {
969             tcg_gen_mov_i32(FREG(B11_8), FREG(B7_4));
970         }
971         return;
972     case 0xf00a: /* fmov {F,D,X}Rm,@Rn - FPSCR: Nothing */
973         CHECK_FPU_ENABLED
974         if (ctx->tbflags & FPSCR_SZ) {
975             TCGv_i64 fp = tcg_temp_new_i64();
976             gen_load_fpr64(ctx, fp, XHACK(B7_4));
977             tcg_gen_qemu_st_i64(fp, REG(B11_8), ctx->memidx,
978                                 MO_TEUQ | MO_ALIGN);
979         } else {
980             tcg_gen_qemu_st_i32(FREG(B7_4), REG(B11_8), ctx->memidx,
981                                 MO_TEUL | MO_ALIGN);
982         }
983         return;
984     case 0xf008: /* fmov @Rm,{F,D,X}Rn - FPSCR: Nothing */
985         CHECK_FPU_ENABLED
986         if (ctx->tbflags & FPSCR_SZ) {
987             TCGv_i64 fp = tcg_temp_new_i64();
988             tcg_gen_qemu_ld_i64(fp, REG(B7_4), ctx->memidx,
989                                 MO_TEUQ | MO_ALIGN);
990             gen_store_fpr64(ctx, fp, XHACK(B11_8));
991         } else {
992             tcg_gen_qemu_ld_i32(FREG(B11_8), REG(B7_4), ctx->memidx,
993                                 MO_TEUL | MO_ALIGN);
994         }
995         return;
996     case 0xf009: /* fmov @Rm+,{F,D,X}Rn - FPSCR: Nothing */
997         CHECK_FPU_ENABLED
998         if (ctx->tbflags & FPSCR_SZ) {
999             TCGv_i64 fp = tcg_temp_new_i64();
1000             tcg_gen_qemu_ld_i64(fp, REG(B7_4), ctx->memidx,
1001                                 MO_TEUQ | MO_ALIGN);
1002             gen_store_fpr64(ctx, fp, XHACK(B11_8));
1003             tcg_gen_addi_i32(REG(B7_4), REG(B7_4), 8);
1004         } else {
1005             tcg_gen_qemu_ld_i32(FREG(B11_8), REG(B7_4), ctx->memidx,
1006                                 MO_TEUL | MO_ALIGN);
1007             tcg_gen_addi_i32(REG(B7_4), REG(B7_4), 4);
1008         }
1009         return;
1010     case 0xf00b: /* fmov {F,D,X}Rm,@-Rn - FPSCR: Nothing */
1011         CHECK_FPU_ENABLED
1012         {
1013             TCGv addr = tcg_temp_new_i32();
1014             if (ctx->tbflags & FPSCR_SZ) {
1015                 TCGv_i64 fp = tcg_temp_new_i64();
1016                 gen_load_fpr64(ctx, fp, XHACK(B7_4));
1017                 tcg_gen_subi_i32(addr, REG(B11_8), 8);
1018                 tcg_gen_qemu_st_i64(fp, addr, ctx->memidx,
1019                                     MO_TEUQ | MO_ALIGN);
1020             } else {
1021                 tcg_gen_subi_i32(addr, REG(B11_8), 4);
1022                 tcg_gen_qemu_st_i32(FREG(B7_4), addr, ctx->memidx,
1023                                     MO_TEUL | MO_ALIGN);
1024             }
1025             tcg_gen_mov_i32(REG(B11_8), addr);
1026         }
1027         return;
1028     case 0xf006: /* fmov @(R0,Rm),{F,D,X}Rm - FPSCR: Nothing */
1029         CHECK_FPU_ENABLED
1030         {
1031             TCGv addr = tcg_temp_new_i32();
1032             tcg_gen_add_i32(addr, REG(B7_4), REG(0));
1033             if (ctx->tbflags & FPSCR_SZ) {
1034                 TCGv_i64 fp = tcg_temp_new_i64();
1035                 tcg_gen_qemu_ld_i64(fp, addr, ctx->memidx,
1036                                     MO_TEUQ | MO_ALIGN);
1037                 gen_store_fpr64(ctx, fp, XHACK(B11_8));
1038             } else {
1039                 tcg_gen_qemu_ld_i32(FREG(B11_8), addr, ctx->memidx,
1040                                     MO_TEUL | MO_ALIGN);
1041             }
1042         }
1043         return;
1044     case 0xf007: /* fmov {F,D,X}Rn,@(R0,Rn) - FPSCR: Nothing */
1045         CHECK_FPU_ENABLED
1046         {
1047             TCGv addr = tcg_temp_new();
1048             tcg_gen_add_i32(addr, REG(B11_8), REG(0));
1049             if (ctx->tbflags & FPSCR_SZ) {
1050                 TCGv_i64 fp = tcg_temp_new_i64();
1051                 gen_load_fpr64(ctx, fp, XHACK(B7_4));
1052                 tcg_gen_qemu_st_i64(fp, addr, ctx->memidx,
1053                                     MO_TEUQ | MO_ALIGN);
1054             } else {
1055                 tcg_gen_qemu_st_i32(FREG(B7_4), addr, ctx->memidx,
1056                                     MO_TEUL | MO_ALIGN);
1057             }
1058         }
1059         return;
1060     case 0xf000: /* fadd Rm,Rn - FPSCR: R[PR,Enable.O/U/I]/W[Cause,Flag] */
1061     case 0xf001: /* fsub Rm,Rn - FPSCR: R[PR,Enable.O/U/I]/W[Cause,Flag] */
1062     case 0xf002: /* fmul Rm,Rn - FPSCR: R[PR,Enable.O/U/I]/W[Cause,Flag] */
1063     case 0xf003: /* fdiv Rm,Rn - FPSCR: R[PR,Enable.O/U/I]/W[Cause,Flag] */
1064     case 0xf004: /* fcmp/eq Rm,Rn - FPSCR: R[PR,Enable.V]/W[Cause,Flag] */
1065     case 0xf005: /* fcmp/gt Rm,Rn - FPSCR: R[PR,Enable.V]/W[Cause,Flag] */
1066         {
1067             CHECK_FPU_ENABLED
1068             if (ctx->tbflags & FPSCR_PR) {
1069                 TCGv_i64 fp0, fp1;
1070 
1071                 if (ctx->opcode & 0x0110) {
1072                     goto do_illegal;
1073                 }
1074                 fp0 = tcg_temp_new_i64();
1075                 fp1 = tcg_temp_new_i64();
1076                 gen_load_fpr64(ctx, fp0, B11_8);
1077                 gen_load_fpr64(ctx, fp1, B7_4);
1078                 switch (ctx->opcode & 0xf00f) {
1079                 case 0xf000: /* fadd Rm,Rn */
1080                     gen_helper_fadd_DT(fp0, tcg_env, fp0, fp1);
1081                     break;
1082                 case 0xf001: /* fsub Rm,Rn */
1083                     gen_helper_fsub_DT(fp0, tcg_env, fp0, fp1);
1084                     break;
1085                 case 0xf002: /* fmul Rm,Rn */
1086                     gen_helper_fmul_DT(fp0, tcg_env, fp0, fp1);
1087                     break;
1088                 case 0xf003: /* fdiv Rm,Rn */
1089                     gen_helper_fdiv_DT(fp0, tcg_env, fp0, fp1);
1090                     break;
1091                 case 0xf004: /* fcmp/eq Rm,Rn */
1092                     gen_helper_fcmp_eq_DT(cpu_sr_t, tcg_env, fp0, fp1);
1093                     return;
1094                 case 0xf005: /* fcmp/gt Rm,Rn */
1095                     gen_helper_fcmp_gt_DT(cpu_sr_t, tcg_env, fp0, fp1);
1096                     return;
1097                 }
1098                 gen_store_fpr64(ctx, fp0, B11_8);
1099             } else {
1100                 switch (ctx->opcode & 0xf00f) {
1101                 case 0xf000: /* fadd Rm,Rn */
1102                     gen_helper_fadd_FT(FREG(B11_8), tcg_env,
1103                                        FREG(B11_8), FREG(B7_4));
1104                     break;
1105                 case 0xf001: /* fsub Rm,Rn */
1106                     gen_helper_fsub_FT(FREG(B11_8), tcg_env,
1107                                        FREG(B11_8), FREG(B7_4));
1108                     break;
1109                 case 0xf002: /* fmul Rm,Rn */
1110                     gen_helper_fmul_FT(FREG(B11_8), tcg_env,
1111                                        FREG(B11_8), FREG(B7_4));
1112                     break;
1113                 case 0xf003: /* fdiv Rm,Rn */
1114                     gen_helper_fdiv_FT(FREG(B11_8), tcg_env,
1115                                        FREG(B11_8), FREG(B7_4));
1116                     break;
1117                 case 0xf004: /* fcmp/eq Rm,Rn */
1118                     gen_helper_fcmp_eq_FT(cpu_sr_t, tcg_env,
1119                                           FREG(B11_8), FREG(B7_4));
1120                     return;
1121                 case 0xf005: /* fcmp/gt Rm,Rn */
1122                     gen_helper_fcmp_gt_FT(cpu_sr_t, tcg_env,
1123                                           FREG(B11_8), FREG(B7_4));
1124                     return;
1125                 }
1126             }
1127         }
1128         return;
1129     case 0xf00e: /* fmac FR0,RM,Rn */
1130         CHECK_FPU_ENABLED
1131         CHECK_FPSCR_PR_0
1132         gen_helper_fmac_FT(FREG(B11_8), tcg_env,
1133                            FREG(0), FREG(B7_4), FREG(B11_8));
1134         return;
1135     }
1136 
1137     switch (ctx->opcode & 0xff00) {
1138     case 0xc900: /* and #imm,R0 */
1139         tcg_gen_andi_i32(REG(0), REG(0), B7_0);
1140         return;
1141     case 0xcd00: /* and.b #imm,@(R0,GBR) */
1142         {
1143             TCGv addr, val;
1144             addr = tcg_temp_new();
1145             tcg_gen_add_i32(addr, REG(0), cpu_gbr);
1146             val = tcg_temp_new();
1147             tcg_gen_qemu_ld_i32(val, addr, ctx->memidx, MO_UB);
1148             tcg_gen_andi_i32(val, val, B7_0);
1149             tcg_gen_qemu_st_i32(val, addr, ctx->memidx, MO_UB);
1150         }
1151         return;
1152     case 0x8b00: /* bf label */
1153         CHECK_NOT_DELAY_SLOT
1154         gen_conditional_jump(ctx, ctx->base.pc_next + 4 + B7_0s * 2, false);
1155         return;
1156     case 0x8f00: /* bf/s label */
1157         CHECK_NOT_DELAY_SLOT
1158         tcg_gen_xori_i32(cpu_delayed_cond, cpu_sr_t, 1);
1159         ctx->delayed_pc = ctx->base.pc_next + 4 + B7_0s * 2;
1160         ctx->envflags |= TB_FLAG_DELAY_SLOT_COND;
1161         return;
1162     case 0x8900: /* bt label */
1163         CHECK_NOT_DELAY_SLOT
1164         gen_conditional_jump(ctx, ctx->base.pc_next + 4 + B7_0s * 2, true);
1165         return;
1166     case 0x8d00: /* bt/s label */
1167         CHECK_NOT_DELAY_SLOT
1168         tcg_gen_mov_i32(cpu_delayed_cond, cpu_sr_t);
1169         ctx->delayed_pc = ctx->base.pc_next + 4 + B7_0s * 2;
1170         ctx->envflags |= TB_FLAG_DELAY_SLOT_COND;
1171         return;
1172     case 0x8800: /* cmp/eq #imm,R0 */
1173         tcg_gen_setcondi_i32(TCG_COND_EQ, cpu_sr_t, REG(0), B7_0s);
1174         return;
1175     case 0xc400: /* mov.b @(disp,GBR),R0 */
1176         {
1177             TCGv addr = tcg_temp_new();
1178             tcg_gen_addi_i32(addr, cpu_gbr, B7_0);
1179             tcg_gen_qemu_ld_i32(REG(0), addr, ctx->memidx, MO_SB);
1180         }
1181         return;
1182     case 0xc500: /* mov.w @(disp,GBR),R0 */
1183         {
1184             TCGv addr = tcg_temp_new();
1185             tcg_gen_addi_i32(addr, cpu_gbr, B7_0 * 2);
1186             tcg_gen_qemu_ld_i32(REG(0), addr, ctx->memidx, MO_TESW | MO_ALIGN);
1187         }
1188         return;
1189     case 0xc600: /* mov.l @(disp,GBR),R0 */
1190         {
1191             TCGv addr = tcg_temp_new();
1192             tcg_gen_addi_i32(addr, cpu_gbr, B7_0 * 4);
1193             tcg_gen_qemu_ld_i32(REG(0), addr, ctx->memidx, MO_TESL | MO_ALIGN);
1194         }
1195         return;
1196     case 0xc000: /* mov.b R0,@(disp,GBR) */
1197         {
1198             TCGv addr = tcg_temp_new();
1199             tcg_gen_addi_i32(addr, cpu_gbr, B7_0);
1200             tcg_gen_qemu_st_i32(REG(0), addr, ctx->memidx, MO_UB);
1201         }
1202         return;
1203     case 0xc100: /* mov.w R0,@(disp,GBR) */
1204         {
1205             TCGv addr = tcg_temp_new();
1206             tcg_gen_addi_i32(addr, cpu_gbr, B7_0 * 2);
1207             tcg_gen_qemu_st_i32(REG(0), addr, ctx->memidx, MO_TEUW | MO_ALIGN);
1208         }
1209         return;
1210     case 0xc200: /* mov.l R0,@(disp,GBR) */
1211         {
1212             TCGv addr = tcg_temp_new();
1213             tcg_gen_addi_i32(addr, cpu_gbr, B7_0 * 4);
1214             tcg_gen_qemu_st_i32(REG(0), addr, ctx->memidx, MO_TEUL | MO_ALIGN);
1215         }
1216         return;
1217     case 0x8000: /* mov.b R0,@(disp,Rn) */
1218         {
1219             TCGv addr = tcg_temp_new();
1220             tcg_gen_addi_i32(addr, REG(B7_4), B3_0);
1221             tcg_gen_qemu_st_i32(REG(0), addr, ctx->memidx, MO_UB);
1222         }
1223         return;
1224     case 0x8100: /* mov.w R0,@(disp,Rn) */
1225         {
1226             TCGv addr = tcg_temp_new();
1227             tcg_gen_addi_i32(addr, REG(B7_4), B3_0 * 2);
1228             tcg_gen_qemu_st_i32(REG(0), addr, ctx->memidx,
1229                                 MO_TEUW | UNALIGN(ctx));
1230         }
1231         return;
1232     case 0x8400: /* mov.b @(disp,Rn),R0 */
1233         {
1234             TCGv addr = tcg_temp_new();
1235             tcg_gen_addi_i32(addr, REG(B7_4), B3_0);
1236             tcg_gen_qemu_ld_i32(REG(0), addr, ctx->memidx, MO_SB);
1237         }
1238         return;
1239     case 0x8500: /* mov.w @(disp,Rn),R0 */
1240         {
1241             TCGv addr = tcg_temp_new();
1242             tcg_gen_addi_i32(addr, REG(B7_4), B3_0 * 2);
1243             tcg_gen_qemu_ld_i32(REG(0), addr, ctx->memidx,
1244                                 MO_TESW | UNALIGN(ctx));
1245         }
1246         return;
1247     case 0xc700: /* mova @(disp,PC),R0 */
1248         CHECK_NOT_DELAY_SLOT
1249         tcg_gen_movi_i32(REG(0), ((ctx->base.pc_next & 0xfffffffc) +
1250                                   4 + B7_0 * 4) & ~3);
1251         return;
1252     case 0xcb00: /* or #imm,R0 */
1253         tcg_gen_ori_i32(REG(0), REG(0), B7_0);
1254         return;
1255     case 0xcf00: /* or.b #imm,@(R0,GBR) */
1256         {
1257             TCGv addr, val;
1258             addr = tcg_temp_new();
1259             tcg_gen_add_i32(addr, REG(0), cpu_gbr);
1260             val = tcg_temp_new();
1261             tcg_gen_qemu_ld_i32(val, addr, ctx->memidx, MO_UB);
1262             tcg_gen_ori_i32(val, val, B7_0);
1263             tcg_gen_qemu_st_i32(val, addr, ctx->memidx, MO_UB);
1264         }
1265         return;
1266     case 0xc300: /* trapa #imm */
1267         {
1268             TCGv imm;
1269             CHECK_NOT_DELAY_SLOT
1270             gen_save_cpu_state(ctx, true);
1271             imm = tcg_constant_i32(B7_0);
1272             gen_helper_trapa(tcg_env, imm);
1273             ctx->base.is_jmp = DISAS_NORETURN;
1274         }
1275         return;
1276     case 0xc800: /* tst #imm,R0 */
1277         {
1278             TCGv val = tcg_temp_new();
1279             tcg_gen_andi_i32(val, REG(0), B7_0);
1280             tcg_gen_setcondi_i32(TCG_COND_EQ, cpu_sr_t, val, 0);
1281         }
1282         return;
1283     case 0xcc00: /* tst.b #imm,@(R0,GBR) */
1284         {
1285             TCGv val = tcg_temp_new();
1286             tcg_gen_add_i32(val, REG(0), cpu_gbr);
1287             tcg_gen_qemu_ld_i32(val, val, ctx->memidx, MO_UB);
1288             tcg_gen_andi_i32(val, val, B7_0);
1289             tcg_gen_setcondi_i32(TCG_COND_EQ, cpu_sr_t, val, 0);
1290         }
1291         return;
1292     case 0xca00: /* xor #imm,R0 */
1293         tcg_gen_xori_i32(REG(0), REG(0), B7_0);
1294         return;
1295     case 0xce00: /* xor.b #imm,@(R0,GBR) */
1296         {
1297             TCGv addr, val;
1298             addr = tcg_temp_new();
1299             tcg_gen_add_i32(addr, REG(0), cpu_gbr);
1300             val = tcg_temp_new();
1301             tcg_gen_qemu_ld_i32(val, addr, ctx->memidx, MO_UB);
1302             tcg_gen_xori_i32(val, val, B7_0);
1303             tcg_gen_qemu_st_i32(val, addr, ctx->memidx, MO_UB);
1304         }
1305         return;
1306     }
1307 
1308     switch (ctx->opcode & 0xf08f) {
1309     case 0x408e: /* ldc Rm,Rn_BANK */
1310         CHECK_PRIVILEGED
1311         tcg_gen_mov_i32(ALTREG(B6_4), REG(B11_8));
1312         return;
1313     case 0x4087: /* ldc.l @Rm+,Rn_BANK */
1314         CHECK_PRIVILEGED
1315         tcg_gen_qemu_ld_i32(ALTREG(B6_4), REG(B11_8), ctx->memidx,
1316                             MO_TESL | MO_ALIGN);
1317         tcg_gen_addi_i32(REG(B11_8), REG(B11_8), 4);
1318         return;
1319     case 0x0082: /* stc Rm_BANK,Rn */
1320         CHECK_PRIVILEGED
1321         tcg_gen_mov_i32(REG(B11_8), ALTREG(B6_4));
1322         return;
1323     case 0x4083: /* stc.l Rm_BANK,@-Rn */
1324         CHECK_PRIVILEGED
1325         {
1326             TCGv addr = tcg_temp_new();
1327             tcg_gen_subi_i32(addr, REG(B11_8), 4);
1328             tcg_gen_qemu_st_i32(ALTREG(B6_4), addr, ctx->memidx,
1329                                 MO_TEUL | MO_ALIGN);
1330             tcg_gen_mov_i32(REG(B11_8), addr);
1331         }
1332         return;
1333     }
1334 
1335     switch (ctx->opcode & 0xf0ff) {
1336     case 0x0023: /* braf Rn */
1337         CHECK_NOT_DELAY_SLOT
1338         tcg_gen_addi_i32(cpu_delayed_pc, REG(B11_8), ctx->base.pc_next + 4);
1339         ctx->envflags |= TB_FLAG_DELAY_SLOT;
1340         ctx->delayed_pc = (uint32_t) - 1;
1341         return;
1342     case 0x0003: /* bsrf Rn */
1343         CHECK_NOT_DELAY_SLOT
1344         tcg_gen_movi_i32(cpu_pr, ctx->base.pc_next + 4);
1345         tcg_gen_add_i32(cpu_delayed_pc, REG(B11_8), cpu_pr);
1346         ctx->envflags |= TB_FLAG_DELAY_SLOT;
1347         ctx->delayed_pc = (uint32_t) - 1;
1348         return;
1349     case 0x4015: /* cmp/pl Rn */
1350         tcg_gen_setcondi_i32(TCG_COND_GT, cpu_sr_t, REG(B11_8), 0);
1351         return;
1352     case 0x4011: /* cmp/pz Rn */
1353         tcg_gen_setcondi_i32(TCG_COND_GE, cpu_sr_t, REG(B11_8), 0);
1354         return;
1355     case 0x4010: /* dt Rn */
1356         tcg_gen_subi_i32(REG(B11_8), REG(B11_8), 1);
1357         tcg_gen_setcondi_i32(TCG_COND_EQ, cpu_sr_t, REG(B11_8), 0);
1358         return;
1359     case 0x402b: /* jmp @Rn */
1360         CHECK_NOT_DELAY_SLOT
1361         tcg_gen_mov_i32(cpu_delayed_pc, REG(B11_8));
1362         ctx->envflags |= TB_FLAG_DELAY_SLOT;
1363         ctx->delayed_pc = (uint32_t) - 1;
1364         return;
1365     case 0x400b: /* jsr @Rn */
1366         CHECK_NOT_DELAY_SLOT
1367         tcg_gen_movi_i32(cpu_pr, ctx->base.pc_next + 4);
1368         tcg_gen_mov_i32(cpu_delayed_pc, REG(B11_8));
1369         ctx->envflags |= TB_FLAG_DELAY_SLOT;
1370         ctx->delayed_pc = (uint32_t) - 1;
1371         return;
1372     case 0x400e: /* ldc Rm,SR */
1373         CHECK_PRIVILEGED
1374         {
1375             TCGv val = tcg_temp_new();
1376             tcg_gen_andi_i32(val, REG(B11_8), 0x700083f3);
1377             gen_write_sr(val);
1378             ctx->base.is_jmp = DISAS_STOP;
1379         }
1380         return;
1381     case 0x4007: /* ldc.l @Rm+,SR */
1382         CHECK_PRIVILEGED
1383         {
1384             TCGv val = tcg_temp_new();
1385             tcg_gen_qemu_ld_i32(val, REG(B11_8), ctx->memidx,
1386                                 MO_TESL | MO_ALIGN);
1387             tcg_gen_andi_i32(val, val, 0x700083f3);
1388             gen_write_sr(val);
1389             tcg_gen_addi_i32(REG(B11_8), REG(B11_8), 4);
1390             ctx->base.is_jmp = DISAS_STOP;
1391         }
1392         return;
1393     case 0x0002: /* stc SR,Rn */
1394         CHECK_PRIVILEGED
1395         gen_read_sr(REG(B11_8));
1396         return;
1397     case 0x4003: /* stc SR,@-Rn */
1398         CHECK_PRIVILEGED
1399         {
1400             TCGv addr = tcg_temp_new();
1401             TCGv val = tcg_temp_new();
1402             tcg_gen_subi_i32(addr, REG(B11_8), 4);
1403             gen_read_sr(val);
1404             tcg_gen_qemu_st_i32(val, addr, ctx->memidx, MO_TEUL | MO_ALIGN);
1405             tcg_gen_mov_i32(REG(B11_8), addr);
1406         }
1407         return;
1408 #define LD(reg,ldnum,ldpnum,prechk)            \
1409   case ldnum:                                                        \
1410     prechk                                                           \
1411     tcg_gen_mov_i32 (cpu_##reg, REG(B11_8));                         \
1412     return;                                                          \
1413   case ldpnum:                                                       \
1414     prechk                                                           \
1415     tcg_gen_qemu_ld_i32(cpu_##reg, REG(B11_8), ctx->memidx,          \
1416                         MO_TESL | MO_ALIGN);                         \
1417     tcg_gen_addi_i32(REG(B11_8), REG(B11_8), 4);                     \
1418     return;
1419 #define ST(reg,stnum,stpnum,prechk)                \
1420   case stnum:                                                        \
1421     prechk                                                           \
1422     tcg_gen_mov_i32 (REG(B11_8), cpu_##reg);                         \
1423     return;                                                          \
1424   case stpnum:                                                       \
1425     prechk                                                           \
1426     {                                                                \
1427         TCGv addr = tcg_temp_new();                                  \
1428         tcg_gen_subi_i32(addr, REG(B11_8), 4);                       \
1429         tcg_gen_qemu_st_i32(cpu_##reg, addr, ctx->memidx,            \
1430                             MO_TEUL | MO_ALIGN);                     \
1431         tcg_gen_mov_i32(REG(B11_8), addr);                           \
1432     }                                                                \
1433     return;
1434 #define LDST(reg,ldnum,ldpnum,stnum,stpnum,prechk)                \
1435         LD(reg,ldnum,ldpnum,prechk)                               \
1436         ST(reg,stnum,stpnum,prechk)
1437         LDST(gbr,  0x401e, 0x4017, 0x0012, 0x4013, {})
1438         LDST(vbr,  0x402e, 0x4027, 0x0022, 0x4023, CHECK_PRIVILEGED)
1439         LDST(ssr,  0x403e, 0x4037, 0x0032, 0x4033, CHECK_PRIVILEGED)
1440         LDST(spc,  0x404e, 0x4047, 0x0042, 0x4043, CHECK_PRIVILEGED)
1441         ST(sgr,  0x003a, 0x4032, CHECK_PRIVILEGED)
1442         LD(sgr,  0x403a, 0x4036, CHECK_PRIVILEGED CHECK_SH4A)
1443         LDST(dbr,  0x40fa, 0x40f6, 0x00fa, 0x40f2, CHECK_PRIVILEGED)
1444         LDST(mach, 0x400a, 0x4006, 0x000a, 0x4002, {})
1445         LDST(macl, 0x401a, 0x4016, 0x001a, 0x4012, {})
1446         LDST(pr,   0x402a, 0x4026, 0x002a, 0x4022, {})
1447         LDST(fpul, 0x405a, 0x4056, 0x005a, 0x4052, {CHECK_FPU_ENABLED})
1448     case 0x406a: /* lds Rm,FPSCR */
1449         CHECK_FPU_ENABLED
1450         gen_helper_ld_fpscr(tcg_env, REG(B11_8));
1451         ctx->base.is_jmp = DISAS_STOP;
1452         return;
1453     case 0x4066: /* lds.l @Rm+,FPSCR */
1454         CHECK_FPU_ENABLED
1455         {
1456             TCGv addr = tcg_temp_new();
1457             tcg_gen_qemu_ld_i32(addr, REG(B11_8), ctx->memidx,
1458                                 MO_TESL | MO_ALIGN);
1459             tcg_gen_addi_i32(REG(B11_8), REG(B11_8), 4);
1460             gen_helper_ld_fpscr(tcg_env, addr);
1461             ctx->base.is_jmp = DISAS_STOP;
1462         }
1463         return;
1464     case 0x006a: /* sts FPSCR,Rn */
1465         CHECK_FPU_ENABLED
1466         tcg_gen_andi_i32(REG(B11_8), cpu_fpscr, 0x003fffff);
1467         return;
1468     case 0x4062: /* sts FPSCR,@-Rn */
1469         CHECK_FPU_ENABLED
1470         {
1471             TCGv addr, val;
1472             val = tcg_temp_new();
1473             tcg_gen_andi_i32(val, cpu_fpscr, 0x003fffff);
1474             addr = tcg_temp_new();
1475             tcg_gen_subi_i32(addr, REG(B11_8), 4);
1476             tcg_gen_qemu_st_i32(val, addr, ctx->memidx, MO_TEUL | MO_ALIGN);
1477             tcg_gen_mov_i32(REG(B11_8), addr);
1478         }
1479         return;
1480     case 0x00c3: /* movca.l R0,@Rm */
1481         {
1482             TCGv val = tcg_temp_new();
1483             tcg_gen_qemu_ld_i32(val, REG(B11_8), ctx->memidx,
1484                                 MO_TEUL | MO_ALIGN);
1485             gen_helper_movcal(tcg_env, REG(B11_8), val);
1486             tcg_gen_qemu_st_i32(REG(0), REG(B11_8), ctx->memidx,
1487                                 MO_TEUL | MO_ALIGN);
1488         }
1489         ctx->has_movcal = 1;
1490         return;
1491     case 0x40a9: /* movua.l @Rm,R0 */
1492         CHECK_SH4A
1493         /* Load non-boundary-aligned data */
1494         tcg_gen_qemu_ld_i32(REG(0), REG(B11_8), ctx->memidx,
1495                             MO_TEUL | MO_UNALN);
1496         return;
1497     case 0x40e9: /* movua.l @Rm+,R0 */
1498         CHECK_SH4A
1499         /* Load non-boundary-aligned data */
1500         tcg_gen_qemu_ld_i32(REG(0), REG(B11_8), ctx->memidx,
1501                             MO_TEUL | MO_UNALN);
1502         tcg_gen_addi_i32(REG(B11_8), REG(B11_8), 4);
1503         return;
1504     case 0x0029: /* movt Rn */
1505         tcg_gen_mov_i32(REG(B11_8), cpu_sr_t);
1506         return;
1507     case 0x0073:
1508         /* MOVCO.L
1509          *     LDST -> T
1510          *     If (T == 1) R0 -> (Rn)
1511          *     0 -> LDST
1512          *
1513          * The above description doesn't work in a parallel context.
1514          * Since we currently support no smp boards, this implies user-mode.
1515          * But we can still support the official mechanism while user-mode
1516          * is single-threaded.  */
1517         CHECK_SH4A
1518         {
1519             TCGLabel *fail = gen_new_label();
1520             TCGLabel *done = gen_new_label();
1521 
1522             if ((tb_cflags(ctx->base.tb) & CF_PARALLEL)) {
1523                 TCGv tmp;
1524 
1525                 tcg_gen_brcond_i32(TCG_COND_NE, REG(B11_8),
1526                                    cpu_lock_addr, fail);
1527                 tmp = tcg_temp_new();
1528                 tcg_gen_atomic_cmpxchg_i32(tmp, REG(B11_8), cpu_lock_value,
1529                                            REG(0), ctx->memidx,
1530                                            MO_TEUL | MO_ALIGN);
1531                 tcg_gen_setcond_i32(TCG_COND_EQ, cpu_sr_t, tmp, cpu_lock_value);
1532             } else {
1533                 tcg_gen_brcondi_i32(TCG_COND_EQ, cpu_lock_addr, -1, fail);
1534                 tcg_gen_qemu_st_i32(REG(0), REG(B11_8), ctx->memidx,
1535                                     MO_TEUL | MO_ALIGN);
1536                 tcg_gen_movi_i32(cpu_sr_t, 1);
1537             }
1538             tcg_gen_br(done);
1539 
1540             gen_set_label(fail);
1541             tcg_gen_movi_i32(cpu_sr_t, 0);
1542 
1543             gen_set_label(done);
1544             tcg_gen_movi_i32(cpu_lock_addr, -1);
1545         }
1546         return;
1547     case 0x0063:
1548         /* MOVLI.L @Rm,R0
1549          *     1 -> LDST
1550          *     (Rm) -> R0
1551          *     When interrupt/exception
1552          *     occurred 0 -> LDST
1553          *
1554          * In a parallel context, we must also save the loaded value
1555          * for use with the cmpxchg that we'll use with movco.l.  */
1556         CHECK_SH4A
1557         if ((tb_cflags(ctx->base.tb) & CF_PARALLEL)) {
1558             TCGv tmp = tcg_temp_new();
1559             tcg_gen_mov_i32(tmp, REG(B11_8));
1560             tcg_gen_qemu_ld_i32(REG(0), REG(B11_8), ctx->memidx,
1561                                 MO_TESL | MO_ALIGN);
1562             tcg_gen_mov_i32(cpu_lock_value, REG(0));
1563             tcg_gen_mov_i32(cpu_lock_addr, tmp);
1564         } else {
1565             tcg_gen_qemu_ld_i32(REG(0), REG(B11_8), ctx->memidx,
1566                                 MO_TESL | MO_ALIGN);
1567             tcg_gen_movi_i32(cpu_lock_addr, 0);
1568         }
1569         return;
1570     case 0x0093: /* ocbi @Rn */
1571         {
1572             gen_helper_ocbi(tcg_env, REG(B11_8));
1573         }
1574         return;
1575     case 0x00a3: /* ocbp @Rn */
1576     case 0x00b3: /* ocbwb @Rn */
1577         /* These instructions are supposed to do nothing in case of
1578            a cache miss. Given that we only partially emulate caches
1579            it is safe to simply ignore them. */
1580         return;
1581     case 0x0083: /* pref @Rn */
1582         return;
1583     case 0x00d3: /* prefi @Rn */
1584         CHECK_SH4A
1585         return;
1586     case 0x00e3: /* icbi @Rn */
1587         CHECK_SH4A
1588         return;
1589     case 0x00ab: /* synco */
1590         CHECK_SH4A
1591         tcg_gen_mb(TCG_MO_ALL | TCG_BAR_SC);
1592         return;
1593     case 0x4024: /* rotcl Rn */
1594         {
1595             TCGv tmp = tcg_temp_new();
1596             tcg_gen_mov_i32(tmp, cpu_sr_t);
1597             tcg_gen_shri_i32(cpu_sr_t, REG(B11_8), 31);
1598             tcg_gen_shli_i32(REG(B11_8), REG(B11_8), 1);
1599             tcg_gen_or_i32(REG(B11_8), REG(B11_8), tmp);
1600         }
1601         return;
1602     case 0x4025: /* rotcr Rn */
1603         {
1604             TCGv tmp = tcg_temp_new();
1605             tcg_gen_shli_i32(tmp, cpu_sr_t, 31);
1606             tcg_gen_andi_i32(cpu_sr_t, REG(B11_8), 1);
1607             tcg_gen_shri_i32(REG(B11_8), REG(B11_8), 1);
1608             tcg_gen_or_i32(REG(B11_8), REG(B11_8), tmp);
1609         }
1610         return;
1611     case 0x4004: /* rotl Rn */
1612         tcg_gen_rotli_i32(REG(B11_8), REG(B11_8), 1);
1613         tcg_gen_andi_i32(cpu_sr_t, REG(B11_8), 0);
1614         return;
1615     case 0x4005: /* rotr Rn */
1616         tcg_gen_andi_i32(cpu_sr_t, REG(B11_8), 0);
1617         tcg_gen_rotri_i32(REG(B11_8), REG(B11_8), 1);
1618         return;
1619     case 0x4000: /* shll Rn */
1620     case 0x4020: /* shal Rn */
1621         tcg_gen_shri_i32(cpu_sr_t, REG(B11_8), 31);
1622         tcg_gen_shli_i32(REG(B11_8), REG(B11_8), 1);
1623         return;
1624     case 0x4021: /* shar Rn */
1625         tcg_gen_andi_i32(cpu_sr_t, REG(B11_8), 1);
1626         tcg_gen_sari_i32(REG(B11_8), REG(B11_8), 1);
1627         return;
1628     case 0x4001: /* shlr Rn */
1629         tcg_gen_andi_i32(cpu_sr_t, REG(B11_8), 1);
1630         tcg_gen_shri_i32(REG(B11_8), REG(B11_8), 1);
1631         return;
1632     case 0x4008: /* shll2 Rn */
1633         tcg_gen_shli_i32(REG(B11_8), REG(B11_8), 2);
1634         return;
1635     case 0x4018: /* shll8 Rn */
1636         tcg_gen_shli_i32(REG(B11_8), REG(B11_8), 8);
1637         return;
1638     case 0x4028: /* shll16 Rn */
1639         tcg_gen_shli_i32(REG(B11_8), REG(B11_8), 16);
1640         return;
1641     case 0x4009: /* shlr2 Rn */
1642         tcg_gen_shri_i32(REG(B11_8), REG(B11_8), 2);
1643         return;
1644     case 0x4019: /* shlr8 Rn */
1645         tcg_gen_shri_i32(REG(B11_8), REG(B11_8), 8);
1646         return;
1647     case 0x4029: /* shlr16 Rn */
1648         tcg_gen_shri_i32(REG(B11_8), REG(B11_8), 16);
1649         return;
1650     case 0x401b: /* tas.b @Rn */
1651         tcg_gen_atomic_fetch_or_i32(cpu_sr_t, REG(B11_8),
1652                                     tcg_constant_i32(0x80), ctx->memidx, MO_UB);
1653         tcg_gen_setcondi_i32(TCG_COND_EQ, cpu_sr_t, cpu_sr_t, 0);
1654         return;
1655     case 0xf00d: /* fsts FPUL,FRn - FPSCR: Nothing */
1656         CHECK_FPU_ENABLED
1657         tcg_gen_mov_i32(FREG(B11_8), cpu_fpul);
1658         return;
1659     case 0xf01d: /* flds FRm,FPUL - FPSCR: Nothing */
1660         CHECK_FPU_ENABLED
1661         tcg_gen_mov_i32(cpu_fpul, FREG(B11_8));
1662         return;
1663     case 0xf02d: /* float FPUL,FRn/DRn - FPSCR: R[PR,Enable.I]/W[Cause,Flag] */
1664         CHECK_FPU_ENABLED
1665         if (ctx->tbflags & FPSCR_PR) {
1666             TCGv_i64 fp;
1667             if (ctx->opcode & 0x0100) {
1668                 goto do_illegal;
1669             }
1670             fp = tcg_temp_new_i64();
1671             gen_helper_float_DT(fp, tcg_env, cpu_fpul);
1672             gen_store_fpr64(ctx, fp, B11_8);
1673         }
1674         else {
1675             gen_helper_float_FT(FREG(B11_8), tcg_env, cpu_fpul);
1676         }
1677         return;
1678     case 0xf03d: /* ftrc FRm/DRm,FPUL - FPSCR: R[PR,Enable.V]/W[Cause,Flag] */
1679         CHECK_FPU_ENABLED
1680         if (ctx->tbflags & FPSCR_PR) {
1681             TCGv_i64 fp;
1682             if (ctx->opcode & 0x0100) {
1683                 goto do_illegal;
1684             }
1685             fp = tcg_temp_new_i64();
1686             gen_load_fpr64(ctx, fp, B11_8);
1687             gen_helper_ftrc_DT(cpu_fpul, tcg_env, fp);
1688         }
1689         else {
1690             gen_helper_ftrc_FT(cpu_fpul, tcg_env, FREG(B11_8));
1691         }
1692         return;
1693     case 0xf04d: /* fneg FRn/DRn - FPSCR: Nothing */
1694         CHECK_FPU_ENABLED
1695         tcg_gen_xori_i32(FREG(B11_8), FREG(B11_8), 0x80000000);
1696         return;
1697     case 0xf05d: /* fabs FRn/DRn - FPCSR: Nothing */
1698         CHECK_FPU_ENABLED
1699         tcg_gen_andi_i32(FREG(B11_8), FREG(B11_8), 0x7fffffff);
1700         return;
1701     case 0xf06d: /* fsqrt FRn */
1702         CHECK_FPU_ENABLED
1703         if (ctx->tbflags & FPSCR_PR) {
1704             if (ctx->opcode & 0x0100) {
1705                 goto do_illegal;
1706             }
1707             TCGv_i64 fp = tcg_temp_new_i64();
1708             gen_load_fpr64(ctx, fp, B11_8);
1709             gen_helper_fsqrt_DT(fp, tcg_env, fp);
1710             gen_store_fpr64(ctx, fp, B11_8);
1711         } else {
1712             gen_helper_fsqrt_FT(FREG(B11_8), tcg_env, FREG(B11_8));
1713         }
1714         return;
1715     case 0xf07d: /* fsrra FRn */
1716         CHECK_FPU_ENABLED
1717         CHECK_FPSCR_PR_0
1718         gen_helper_fsrra_FT(FREG(B11_8), tcg_env, FREG(B11_8));
1719         break;
1720     case 0xf08d: /* fldi0 FRn - FPSCR: R[PR] */
1721         CHECK_FPU_ENABLED
1722         CHECK_FPSCR_PR_0
1723         tcg_gen_movi_i32(FREG(B11_8), 0);
1724         return;
1725     case 0xf09d: /* fldi1 FRn - FPSCR: R[PR] */
1726         CHECK_FPU_ENABLED
1727         CHECK_FPSCR_PR_0
1728         tcg_gen_movi_i32(FREG(B11_8), 0x3f800000);
1729         return;
1730     case 0xf0ad: /* fcnvsd FPUL,DRn */
1731         CHECK_FPU_ENABLED
1732         {
1733             TCGv_i64 fp = tcg_temp_new_i64();
1734             gen_helper_fcnvsd_FT_DT(fp, tcg_env, cpu_fpul);
1735             gen_store_fpr64(ctx, fp, B11_8);
1736         }
1737         return;
1738     case 0xf0bd: /* fcnvds DRn,FPUL */
1739         CHECK_FPU_ENABLED
1740         {
1741             TCGv_i64 fp = tcg_temp_new_i64();
1742             gen_load_fpr64(ctx, fp, B11_8);
1743             gen_helper_fcnvds_DT_FT(cpu_fpul, tcg_env, fp);
1744         }
1745         return;
1746     case 0xf0ed: /* fipr FVm,FVn */
1747         CHECK_FPU_ENABLED
1748         CHECK_FPSCR_PR_1
1749         {
1750             TCGv m = tcg_constant_i32((ctx->opcode >> 8) & 3);
1751             TCGv n = tcg_constant_i32((ctx->opcode >> 10) & 3);
1752             gen_helper_fipr(tcg_env, m, n);
1753             return;
1754         }
1755         break;
1756     case 0xf0fd: /* ftrv XMTRX,FVn */
1757         CHECK_FPU_ENABLED
1758         CHECK_FPSCR_PR_1
1759         {
1760             if ((ctx->opcode & 0x0300) != 0x0100) {
1761                 goto do_illegal;
1762             }
1763             TCGv n = tcg_constant_i32((ctx->opcode >> 10) & 3);
1764             gen_helper_ftrv(tcg_env, n);
1765             return;
1766         }
1767         break;
1768     }
1769 #if 0
1770     fprintf(stderr, "unknown instruction 0x%04x at pc 0x%08x\n",
1771             ctx->opcode, ctx->base.pc_next);
1772     fflush(stderr);
1773 #endif
1774  do_illegal:
1775     if (ctx->envflags & TB_FLAG_DELAY_SLOT_MASK) {
1776  do_illegal_slot:
1777         gen_save_cpu_state(ctx, true);
1778         gen_helper_raise_slot_illegal_instruction(tcg_env);
1779     } else {
1780         gen_save_cpu_state(ctx, true);
1781         gen_helper_raise_illegal_instruction(tcg_env);
1782     }
1783     ctx->base.is_jmp = DISAS_NORETURN;
1784     return;
1785 
1786  do_fpu_disabled:
1787     gen_save_cpu_state(ctx, true);
1788     if (ctx->envflags & TB_FLAG_DELAY_SLOT_MASK) {
1789         gen_helper_raise_slot_fpu_disable(tcg_env);
1790     } else {
1791         gen_helper_raise_fpu_disable(tcg_env);
1792     }
1793     ctx->base.is_jmp = DISAS_NORETURN;
1794     return;
1795 }
1796 
1797 static void decode_opc(DisasContext * ctx)
1798 {
1799     uint32_t old_flags = ctx->envflags;
1800 
1801     _decode_opc(ctx);
1802 
1803     if (old_flags & TB_FLAG_DELAY_SLOT_MASK) {
1804         /* go out of the delay slot */
1805         ctx->envflags &= ~TB_FLAG_DELAY_SLOT_MASK;
1806 
1807         /* When in an exclusive region, we must continue to the end
1808            for conditional branches.  */
1809         if (ctx->tbflags & TB_FLAG_GUSA_EXCLUSIVE
1810             && old_flags & TB_FLAG_DELAY_SLOT_COND) {
1811             gen_delayed_conditional_jump(ctx);
1812             return;
1813         }
1814         /* Otherwise this is probably an invalid gUSA region.
1815            Drop the GUSA bits so the next TB doesn't see them.  */
1816         ctx->envflags &= ~TB_FLAG_GUSA_MASK;
1817 
1818         tcg_gen_movi_i32(cpu_flags, ctx->envflags);
1819         if (old_flags & TB_FLAG_DELAY_SLOT_COND) {
1820             gen_delayed_conditional_jump(ctx);
1821         } else {
1822             gen_jump(ctx);
1823         }
1824     }
1825 }
1826 
1827 #ifdef CONFIG_USER_ONLY
1828 /*
1829  * Restart with the EXCLUSIVE bit set, within a TB run via
1830  * cpu_exec_step_atomic holding the exclusive lock.
1831  */
1832 static void gen_restart_exclusive(DisasContext *ctx)
1833 {
1834     ctx->envflags |= TB_FLAG_GUSA_EXCLUSIVE;
1835     gen_save_cpu_state(ctx, false);
1836     gen_helper_exclusive(tcg_env);
1837     ctx->base.is_jmp = DISAS_NORETURN;
1838 }
1839 
1840 /* For uniprocessors, SH4 uses optimistic restartable atomic sequences.
1841    Upon an interrupt, a real kernel would simply notice magic values in
1842    the registers and reset the PC to the start of the sequence.
1843 
1844    For QEMU, we cannot do this in quite the same way.  Instead, we notice
1845    the normal start of such a sequence (mov #-x,r15).  While we can handle
1846    any sequence via cpu_exec_step_atomic, we can recognize the "normal"
1847    sequences and transform them into atomic operations as seen by the host.
1848 */
1849 static void decode_gusa(DisasContext *ctx, CPUSH4State *env)
1850 {
1851     uint16_t insns[5];
1852     int ld_adr, ld_dst, ld_mop;
1853     int op_dst, op_src, op_opc;
1854     int mv_src, mt_dst, st_src, st_mop;
1855     TCGv op_arg;
1856     uint32_t pc = ctx->base.pc_next;
1857     uint32_t pc_end = ctx->base.tb->cs_base;
1858     int max_insns = (pc_end - pc) / 2;
1859     int i;
1860 
1861     /* The state machine below will consume only a few insns.
1862        If there are more than that in a region, fail now.  */
1863     if (max_insns > ARRAY_SIZE(insns)) {
1864         goto fail;
1865     }
1866 
1867     /* Read all of the insns for the region.  */
1868     for (i = 0; i < max_insns; ++i) {
1869         insns[i] = translator_lduw(env, &ctx->base, pc + i * 2);
1870     }
1871 
1872     ld_adr = ld_dst = ld_mop = -1;
1873     mv_src = -1;
1874     op_dst = op_src = op_opc = -1;
1875     mt_dst = -1;
1876     st_src = st_mop = -1;
1877     op_arg = NULL;
1878     i = 0;
1879 
1880 #define NEXT_INSN \
1881     do { if (i >= max_insns) goto fail; ctx->opcode = insns[i++]; } while (0)
1882 
1883     /*
1884      * Expect a load to begin the region.
1885      */
1886     NEXT_INSN;
1887     switch (ctx->opcode & 0xf00f) {
1888     case 0x6000: /* mov.b @Rm,Rn */
1889         ld_mop = MO_SB;
1890         break;
1891     case 0x6001: /* mov.w @Rm,Rn */
1892         ld_mop = MO_TESW;
1893         break;
1894     case 0x6002: /* mov.l @Rm,Rn */
1895         ld_mop = MO_TESL;
1896         break;
1897     default:
1898         goto fail;
1899     }
1900     ld_adr = B7_4;
1901     ld_dst = B11_8;
1902     if (ld_adr == ld_dst) {
1903         goto fail;
1904     }
1905     /* Unless we see a mov, any two-operand operation must use ld_dst.  */
1906     op_dst = ld_dst;
1907 
1908     /*
1909      * Expect an optional register move.
1910      */
1911     NEXT_INSN;
1912     switch (ctx->opcode & 0xf00f) {
1913     case 0x6003: /* mov Rm,Rn */
1914         /*
1915          * Here we want to recognize ld_dst being saved for later consumption,
1916          * or for another input register being copied so that ld_dst need not
1917          * be clobbered during the operation.
1918          */
1919         op_dst = B11_8;
1920         mv_src = B7_4;
1921         if (op_dst == ld_dst) {
1922             /* Overwriting the load output.  */
1923             goto fail;
1924         }
1925         if (mv_src != ld_dst) {
1926             /* Copying a new input; constrain op_src to match the load.  */
1927             op_src = ld_dst;
1928         }
1929         break;
1930 
1931     default:
1932         /* Put back and re-examine as operation.  */
1933         --i;
1934     }
1935 
1936     /*
1937      * Expect the operation.
1938      */
1939     NEXT_INSN;
1940     switch (ctx->opcode & 0xf00f) {
1941     case 0x300c: /* add Rm,Rn */
1942         op_opc = INDEX_op_add_i32;
1943         goto do_reg_op;
1944     case 0x2009: /* and Rm,Rn */
1945         op_opc = INDEX_op_and_i32;
1946         goto do_reg_op;
1947     case 0x200a: /* xor Rm,Rn */
1948         op_opc = INDEX_op_xor_i32;
1949         goto do_reg_op;
1950     case 0x200b: /* or Rm,Rn */
1951         op_opc = INDEX_op_or_i32;
1952     do_reg_op:
1953         /* The operation register should be as expected, and the
1954            other input cannot depend on the load.  */
1955         if (op_dst != B11_8) {
1956             goto fail;
1957         }
1958         if (op_src < 0) {
1959             /* Unconstrainted input.  */
1960             op_src = B7_4;
1961         } else if (op_src == B7_4) {
1962             /* Constrained input matched load.  All operations are
1963                commutative; "swap" them by "moving" the load output
1964                to the (implicit) first argument and the move source
1965                to the (explicit) second argument.  */
1966             op_src = mv_src;
1967         } else {
1968             goto fail;
1969         }
1970         op_arg = REG(op_src);
1971         break;
1972 
1973     case 0x6007: /* not Rm,Rn */
1974         if (ld_dst != B7_4 || mv_src >= 0) {
1975             goto fail;
1976         }
1977         op_dst = B11_8;
1978         op_opc = INDEX_op_xor_i32;
1979         op_arg = tcg_constant_i32(-1);
1980         break;
1981 
1982     case 0x7000 ... 0x700f: /* add #imm,Rn */
1983         if (op_dst != B11_8 || mv_src >= 0) {
1984             goto fail;
1985         }
1986         op_opc = INDEX_op_add_i32;
1987         op_arg = tcg_constant_i32(B7_0s);
1988         break;
1989 
1990     case 0x3000: /* cmp/eq Rm,Rn */
1991         /* Looking for the middle of a compare-and-swap sequence,
1992            beginning with the compare.  Operands can be either order,
1993            but with only one overlapping the load.  */
1994         if ((ld_dst == B11_8) + (ld_dst == B7_4) != 1 || mv_src >= 0) {
1995             goto fail;
1996         }
1997         op_opc = INDEX_op_setcond_i32;  /* placeholder */
1998         op_src = (ld_dst == B11_8 ? B7_4 : B11_8);
1999         op_arg = REG(op_src);
2000 
2001         NEXT_INSN;
2002         switch (ctx->opcode & 0xff00) {
2003         case 0x8b00: /* bf label */
2004         case 0x8f00: /* bf/s label */
2005             if (pc + (i + 1 + B7_0s) * 2 != pc_end) {
2006                 goto fail;
2007             }
2008             if ((ctx->opcode & 0xff00) == 0x8b00) { /* bf label */
2009                 break;
2010             }
2011             /* We're looking to unconditionally modify Rn with the
2012                result of the comparison, within the delay slot of
2013                the branch.  This is used by older gcc.  */
2014             NEXT_INSN;
2015             if ((ctx->opcode & 0xf0ff) == 0x0029) { /* movt Rn */
2016                 mt_dst = B11_8;
2017             } else {
2018                 goto fail;
2019             }
2020             break;
2021 
2022         default:
2023             goto fail;
2024         }
2025         break;
2026 
2027     case 0x2008: /* tst Rm,Rn */
2028         /* Looking for a compare-and-swap against zero.  */
2029         if (ld_dst != B11_8 || ld_dst != B7_4 || mv_src >= 0) {
2030             goto fail;
2031         }
2032         op_opc = INDEX_op_setcond_i32;
2033         op_arg = tcg_constant_i32(0);
2034 
2035         NEXT_INSN;
2036         if ((ctx->opcode & 0xff00) != 0x8900 /* bt label */
2037             || pc + (i + 1 + B7_0s) * 2 != pc_end) {
2038             goto fail;
2039         }
2040         break;
2041 
2042     default:
2043         /* Put back and re-examine as store.  */
2044         --i;
2045     }
2046 
2047     /*
2048      * Expect the store.
2049      */
2050     /* The store must be the last insn.  */
2051     if (i != max_insns - 1) {
2052         goto fail;
2053     }
2054     NEXT_INSN;
2055     switch (ctx->opcode & 0xf00f) {
2056     case 0x2000: /* mov.b Rm,@Rn */
2057         st_mop = MO_UB;
2058         break;
2059     case 0x2001: /* mov.w Rm,@Rn */
2060         st_mop = MO_UW;
2061         break;
2062     case 0x2002: /* mov.l Rm,@Rn */
2063         st_mop = MO_UL;
2064         break;
2065     default:
2066         goto fail;
2067     }
2068     /* The store must match the load.  */
2069     if (ld_adr != B11_8 || st_mop != (ld_mop & MO_SIZE)) {
2070         goto fail;
2071     }
2072     st_src = B7_4;
2073 
2074 #undef NEXT_INSN
2075 
2076     /*
2077      * Emit the operation.
2078      */
2079     switch (op_opc) {
2080     case -1:
2081         /* No operation found.  Look for exchange pattern.  */
2082         if (st_src == ld_dst || mv_src >= 0) {
2083             goto fail;
2084         }
2085         tcg_gen_atomic_xchg_i32(REG(ld_dst), REG(ld_adr), REG(st_src),
2086                                 ctx->memidx, ld_mop);
2087         break;
2088 
2089     case INDEX_op_add_i32:
2090         if (op_dst != st_src) {
2091             goto fail;
2092         }
2093         if (op_dst == ld_dst && st_mop == MO_UL) {
2094             tcg_gen_atomic_add_fetch_i32(REG(ld_dst), REG(ld_adr),
2095                                          op_arg, ctx->memidx, ld_mop);
2096         } else {
2097             tcg_gen_atomic_fetch_add_i32(REG(ld_dst), REG(ld_adr),
2098                                          op_arg, ctx->memidx, ld_mop);
2099             if (op_dst != ld_dst) {
2100                 /* Note that mop sizes < 4 cannot use add_fetch
2101                    because it won't carry into the higher bits.  */
2102                 tcg_gen_add_i32(REG(op_dst), REG(ld_dst), op_arg);
2103             }
2104         }
2105         break;
2106 
2107     case INDEX_op_and_i32:
2108         if (op_dst != st_src) {
2109             goto fail;
2110         }
2111         if (op_dst == ld_dst) {
2112             tcg_gen_atomic_and_fetch_i32(REG(ld_dst), REG(ld_adr),
2113                                          op_arg, ctx->memidx, ld_mop);
2114         } else {
2115             tcg_gen_atomic_fetch_and_i32(REG(ld_dst), REG(ld_adr),
2116                                          op_arg, ctx->memidx, ld_mop);
2117             tcg_gen_and_i32(REG(op_dst), REG(ld_dst), op_arg);
2118         }
2119         break;
2120 
2121     case INDEX_op_or_i32:
2122         if (op_dst != st_src) {
2123             goto fail;
2124         }
2125         if (op_dst == ld_dst) {
2126             tcg_gen_atomic_or_fetch_i32(REG(ld_dst), REG(ld_adr),
2127                                         op_arg, ctx->memidx, ld_mop);
2128         } else {
2129             tcg_gen_atomic_fetch_or_i32(REG(ld_dst), REG(ld_adr),
2130                                         op_arg, ctx->memidx, ld_mop);
2131             tcg_gen_or_i32(REG(op_dst), REG(ld_dst), op_arg);
2132         }
2133         break;
2134 
2135     case INDEX_op_xor_i32:
2136         if (op_dst != st_src) {
2137             goto fail;
2138         }
2139         if (op_dst == ld_dst) {
2140             tcg_gen_atomic_xor_fetch_i32(REG(ld_dst), REG(ld_adr),
2141                                          op_arg, ctx->memidx, ld_mop);
2142         } else {
2143             tcg_gen_atomic_fetch_xor_i32(REG(ld_dst), REG(ld_adr),
2144                                          op_arg, ctx->memidx, ld_mop);
2145             tcg_gen_xor_i32(REG(op_dst), REG(ld_dst), op_arg);
2146         }
2147         break;
2148 
2149     case INDEX_op_setcond_i32:
2150         if (st_src == ld_dst) {
2151             goto fail;
2152         }
2153         tcg_gen_atomic_cmpxchg_i32(REG(ld_dst), REG(ld_adr), op_arg,
2154                                    REG(st_src), ctx->memidx, ld_mop);
2155         tcg_gen_setcond_i32(TCG_COND_EQ, cpu_sr_t, REG(ld_dst), op_arg);
2156         if (mt_dst >= 0) {
2157             tcg_gen_mov_i32(REG(mt_dst), cpu_sr_t);
2158         }
2159         break;
2160 
2161     default:
2162         g_assert_not_reached();
2163     }
2164 
2165     /* The entire region has been translated.  */
2166     ctx->envflags &= ~TB_FLAG_GUSA_MASK;
2167     goto done;
2168 
2169  fail:
2170     qemu_log_mask(LOG_UNIMP, "Unrecognized gUSA sequence %08x-%08x\n",
2171                   pc, pc_end);
2172 
2173     gen_restart_exclusive(ctx);
2174 
2175     /* We're not executing an instruction, but we must report one for the
2176        purposes of accounting within the TB.  We might as well report the
2177        entire region consumed via ctx->base.pc_next so that it's immediately
2178        available in the disassembly dump.  */
2179 
2180  done:
2181     ctx->base.pc_next = pc_end;
2182     ctx->base.num_insns += max_insns - 1;
2183 
2184     /*
2185      * Emit insn_start to cover each of the insns in the region.
2186      * This matches an assert in tcg.c making sure that we have
2187      * tb->icount * insn_start.
2188      */
2189     for (i = 1; i < max_insns; ++i) {
2190         tcg_gen_insn_start(pc + i * 2, ctx->envflags);
2191         ctx->base.insn_start = tcg_last_op();
2192     }
2193 }
2194 #endif
2195 
2196 static void sh4_tr_init_disas_context(DisasContextBase *dcbase, CPUState *cs)
2197 {
2198     DisasContext *ctx = container_of(dcbase, DisasContext, base);
2199     uint32_t tbflags;
2200     int bound;
2201 
2202     ctx->tbflags = tbflags = ctx->base.tb->flags;
2203     ctx->envflags = tbflags & TB_FLAG_ENVFLAGS_MASK;
2204     ctx->memidx = (tbflags & (1u << SR_MD)) == 0 ? 1 : 0;
2205     /* We don't know if the delayed pc came from a dynamic or static branch,
2206        so assume it is a dynamic branch.  */
2207     ctx->delayed_pc = -1; /* use delayed pc from env pointer */
2208     ctx->features = cpu_env(cs)->features;
2209     ctx->has_movcal = (tbflags & TB_FLAG_PENDING_MOVCA);
2210     ctx->gbank = ((tbflags & (1 << SR_MD)) &&
2211                   (tbflags & (1 << SR_RB))) * 0x10;
2212     ctx->fbank = tbflags & FPSCR_FR ? 0x10 : 0;
2213 
2214 #ifdef CONFIG_USER_ONLY
2215     if (tbflags & TB_FLAG_GUSA_MASK) {
2216         /* In gUSA exclusive region. */
2217         uint32_t pc = ctx->base.pc_next;
2218         uint32_t pc_end = ctx->base.tb->cs_base;
2219         int backup = sextract32(ctx->tbflags, TB_FLAG_GUSA_SHIFT, 8);
2220         int max_insns = (pc_end - pc) / 2;
2221 
2222         if (pc != pc_end + backup || max_insns < 2) {
2223             /* This is a malformed gUSA region.  Don't do anything special,
2224                since the interpreter is likely to get confused.  */
2225             ctx->envflags &= ~TB_FLAG_GUSA_MASK;
2226         } else if (tbflags & TB_FLAG_GUSA_EXCLUSIVE) {
2227             /* Regardless of single-stepping or the end of the page,
2228                we must complete execution of the gUSA region while
2229                holding the exclusive lock.  */
2230             ctx->base.max_insns = max_insns;
2231             return;
2232         }
2233     }
2234 #endif
2235 
2236     /* Since the ISA is fixed-width, we can bound by the number
2237        of instructions remaining on the page.  */
2238     bound = -(ctx->base.pc_next | TARGET_PAGE_MASK) / 2;
2239     ctx->base.max_insns = MIN(ctx->base.max_insns, bound);
2240 }
2241 
2242 static void sh4_tr_tb_start(DisasContextBase *dcbase, CPUState *cs)
2243 {
2244 }
2245 
2246 static void sh4_tr_insn_start(DisasContextBase *dcbase, CPUState *cs)
2247 {
2248     DisasContext *ctx = container_of(dcbase, DisasContext, base);
2249 
2250     tcg_gen_insn_start(ctx->base.pc_next, ctx->envflags);
2251 }
2252 
2253 static void sh4_tr_translate_insn(DisasContextBase *dcbase, CPUState *cs)
2254 {
2255     CPUSH4State *env = cpu_env(cs);
2256     DisasContext *ctx = container_of(dcbase, DisasContext, base);
2257 
2258 #ifdef CONFIG_USER_ONLY
2259     if (unlikely(ctx->envflags & TB_FLAG_GUSA_MASK)
2260         && !(ctx->envflags & TB_FLAG_GUSA_EXCLUSIVE)) {
2261         /*
2262          * We're in an gUSA region, and we have not already fallen
2263          * back on using an exclusive region.  Attempt to parse the
2264          * region into a single supported atomic operation.  Failure
2265          * is handled within the parser by raising an exception to
2266          * retry using an exclusive region.
2267          *
2268          * Parsing the region in one block conflicts with plugins,
2269          * so always use exclusive mode if plugins enabled.
2270          */
2271         if (ctx->base.plugin_enabled) {
2272             gen_restart_exclusive(ctx);
2273             ctx->base.pc_next += 2;
2274         } else {
2275             decode_gusa(ctx, env);
2276         }
2277         return;
2278     }
2279 #endif
2280 
2281     ctx->opcode = translator_lduw(env, &ctx->base, ctx->base.pc_next);
2282     decode_opc(ctx);
2283     ctx->base.pc_next += 2;
2284 }
2285 
2286 static void sh4_tr_tb_stop(DisasContextBase *dcbase, CPUState *cs)
2287 {
2288     DisasContext *ctx = container_of(dcbase, DisasContext, base);
2289 
2290     if (ctx->tbflags & TB_FLAG_GUSA_EXCLUSIVE) {
2291         /* Ending the region of exclusivity.  Clear the bits.  */
2292         ctx->envflags &= ~TB_FLAG_GUSA_MASK;
2293     }
2294 
2295     switch (ctx->base.is_jmp) {
2296     case DISAS_STOP:
2297         gen_save_cpu_state(ctx, true);
2298         tcg_gen_exit_tb(NULL, 0);
2299         break;
2300     case DISAS_NEXT:
2301     case DISAS_TOO_MANY:
2302         gen_save_cpu_state(ctx, false);
2303         gen_goto_tb(ctx, 0, ctx->base.pc_next);
2304         break;
2305     case DISAS_NORETURN:
2306         break;
2307     default:
2308         g_assert_not_reached();
2309     }
2310 }
2311 
2312 static const TranslatorOps sh4_tr_ops = {
2313     .init_disas_context = sh4_tr_init_disas_context,
2314     .tb_start           = sh4_tr_tb_start,
2315     .insn_start         = sh4_tr_insn_start,
2316     .translate_insn     = sh4_tr_translate_insn,
2317     .tb_stop            = sh4_tr_tb_stop,
2318 };
2319 
2320 void gen_intermediate_code(CPUState *cs, TranslationBlock *tb, int *max_insns,
2321                            vaddr pc, void *host_pc)
2322 {
2323     DisasContext ctx;
2324 
2325     translator_loop(cs, tb, max_insns, pc, host_pc, &sh4_tr_ops, &ctx.base);
2326 }
2327