xref: /openbmc/qemu/target/arm/tcg/translate.h (revision 5a572dd2cb4c96e51c7c0d6a549050ed33dea269)
1 #ifndef TARGET_ARM_TRANSLATE_H
2 #define TARGET_ARM_TRANSLATE_H
3 
4 #include "cpu.h"
5 #include "tcg/tcg-op.h"
6 #include "tcg/tcg-op-gvec.h"
7 #include "exec/translator.h"
8 #include "exec/translation-block.h"
9 #include "exec/helper-gen.h"
10 #include "internals.h"
11 #include "cpu-features.h"
12 
13 /* internal defines */
14 
15 /*
16  * Save pc_save across a branch, so that we may restore the value from
17  * before the branch at the point the label is emitted.
18  */
19 typedef struct DisasLabel {
20     TCGLabel *label;
21     target_ulong pc_save;
22 } DisasLabel;
23 
24 /*
25  * Emit an exception call out of line.
26  */
27 typedef struct DisasDelayException {
28     struct DisasDelayException *next;
29     TCGLabel *lab;
30     target_long pc_curr;
31     target_long pc_save;
32     int condexec_mask;
33     int condexec_cond;
34     uint32_t excp;
35     uint32_t syn;
36     uint32_t target_el;
37 } DisasDelayException;
38 
39 typedef struct DisasContext {
40     DisasContextBase base;
41     const ARMISARegisters *isar;
42     DisasDelayException *delay_excp_list;
43 
44     /* The address of the current instruction being translated. */
45     target_ulong pc_curr;
46     /*
47      * For CF_PCREL, the full value of cpu_pc is not known
48      * (although the page offset is known).  For convenience, the
49      * translation loop uses the full virtual address that triggered
50      * the translation, from base.pc_start through pc_curr.
51      * For efficiency, we do not update cpu_pc for every instruction.
52      * Instead, pc_save has the value of pc_curr at the time of the
53      * last update to cpu_pc, which allows us to compute the addend
54      * needed to bring cpu_pc current: pc_curr - pc_save.
55      * If cpu_pc now contains the destination of an indirect branch,
56      * pc_save contains -1 to indicate that relative updates are no
57      * longer possible.
58      */
59     target_ulong pc_save;
60     target_ulong page_start;
61     uint32_t insn;
62     /* Nonzero if this instruction has been conditionally skipped.  */
63     int condjmp;
64     /* The label that will be jumped to when the instruction is skipped.  */
65     DisasLabel condlabel;
66     /* Thumb-2 conditional execution bits.  */
67     int condexec_mask;
68     int condexec_cond;
69     /* M-profile ECI/ICI exception-continuable instruction state */
70     int eci;
71     /*
72      * trans_ functions for insns which are continuable should set this true
73      * after decode (ie after any UNDEF checks)
74      */
75     bool eci_handled;
76     int sctlr_b;
77     MemOp be_data;
78 #if !defined(CONFIG_USER_ONLY)
79     int user;
80 #endif
81     ARMMMUIdx mmu_idx; /* MMU index to use for normal loads/stores */
82     uint8_t tbii;      /* TBI1|TBI0 for insns */
83     uint8_t tbid;      /* TBI1|TBI0 for data */
84     uint8_t tcma;      /* TCMA1|TCMA0 for MTE */
85     bool ns;        /* Use non-secure CPREG bank on access */
86     int fp_excp_el; /* FP exception EL or 0 if enabled */
87     int sve_excp_el; /* SVE exception EL or 0 if enabled */
88     int sme_excp_el; /* SME exception EL or 0 if enabled */
89     int zt0_excp_el; /* ZT0 exception EL or 0 if enabled */
90     int vl;          /* current vector length in bytes */
91     int svl;         /* current streaming vector length in bytes */
92     int max_svl;     /* maximum implemented streaming vector length */
93     bool vfp_enabled; /* FP enabled via FPSCR.EN */
94     int vec_len;
95     int vec_stride;
96     bool v7m_handler_mode;
97     bool v8m_secure; /* true if v8M and we're in Secure mode */
98     bool v8m_stackcheck; /* true if we need to perform v8M stack limit checks */
99     bool v8m_fpccr_s_wrong; /* true if v8M FPCCR.S != v8m_secure */
100     bool v7m_new_fp_ctxt_needed; /* ASPEN set but no active FP context */
101     bool v7m_lspact; /* FPCCR.LSPACT set */
102     /* Immediate value in AArch32 SVC insn; must be set if is_jmp == DISAS_SWI
103      * so that top level loop can generate correct syndrome information.
104      */
105     uint32_t svc_imm;
106     int current_el;
107     GHashTable *cp_regs;
108     uint64_t features; /* CPU features bits */
109     bool aarch64;
110     bool thumb;
111     bool lse2;
112     /*
113      * Because unallocated encodings generate different exception syndrome
114      * information from traps due to FP being disabled, we can't do a single
115      * "is fp access disabled" check at a high level in the decode tree.
116      * To help in catching bugs where the access check was forgotten in some
117      * code path, we set this flag when the access check is done, and assert
118      * that it is set at the point where we actually touch the FP regs.
119      *   0: not checked,
120      *   1: checked, access ok
121      *  -1: checked, access denied
122      */
123     int8_t fp_access_checked;
124     int8_t sve_access_checked;
125     /* ARMv8 single-step state (this is distinct from the QEMU gdbstub
126      * single-step support).
127      */
128     bool ss_active;
129     bool pstate_ss;
130     /* True if the insn just emitted was a load-exclusive instruction
131      * (necessary for syndrome information for single step exceptions),
132      * ie A64 LDX*, LDAX*, A32/T32 LDREX*, LDAEX*.
133      */
134     bool is_ldex;
135     /* True if AccType_UNPRIV should be used for LDTR et al */
136     bool unpriv;
137     /* True if v8.3-PAuth is active.  */
138     bool pauth_active;
139     /* True if v8.5-MTE access to tags is enabled; index with is_unpriv.  */
140     bool ata[2];
141     /* True if v8.5-MTE tag checks affect the PE; index with is_unpriv.  */
142     bool mte_active[2];
143     /* True with v8.5-BTI and SCTLR_ELx.BT* set.  */
144     bool bt;
145     /* True if any CP15 access is trapped by HSTR_EL2 */
146     bool hstr_active;
147     /* True if memory operations require alignment */
148     bool align_mem;
149     /* True if PSTATE.IL is set */
150     bool pstate_il;
151     /* True if PSTATE.SM is set. */
152     bool pstate_sm;
153     /* True if PSTATE.ZA is set. */
154     bool pstate_za;
155     /* True if non-streaming insns should raise an SME Streaming exception. */
156     bool sme_trap_nonstreaming;
157     /* True if the current instruction is non-streaming. */
158     bool is_nonstreaming;
159     /* True if MVE insns are definitely not predicated by VPR or LTPSIZE */
160     bool mve_no_pred;
161     /* True if fine-grained traps are active */
162     bool fgt_active;
163     /* True if fine-grained trap on SVC is enabled */
164     bool fgt_svc;
165     /* True if a trap on ERET is enabled (FGT or NV) */
166     bool trap_eret;
167     /* True if FEAT_LSE2 SCTLR_ELx.nAA is set */
168     bool naa;
169     /* True if HCR_EL2.E2H is set */
170     bool e2h;
171     /* True if FEAT_NV HCR_EL2.NV is enabled */
172     bool nv;
173     /* True if NV enabled and HCR_EL2.NV1 is set */
174     bool nv1;
175     /* True if NV enabled and HCR_EL2.NV2 is set */
176     bool nv2;
177     /* True if NV2 enabled and NV2 RAM accesses use EL2&0 translation regime */
178     bool nv2_mem_e20;
179     /* True if NV2 enabled and NV2 RAM accesses are big-endian */
180     bool nv2_mem_be;
181     /* True if FPCR.AH is 1 (alternate floating point handling) */
182     bool fpcr_ah;
183     /* True if FPCR.NEP is 1 (FEAT_AFP scalar upper-element result handling) */
184     bool fpcr_nep;
185     /* True if GCSEnabled. */
186     bool gcs_en;
187     /* True if GCSReturnValueCheckEnabled. */
188     bool gcs_rvcen;
189     /* GCSSTR exception EL or 0 if enabled */
190     uint8_t gcsstr_el;
191     /*
192      * >= 0, a copy of PSTATE.BTYPE, which will be 0 without v8.5-BTI.
193      *  < 0, set by the current instruction.
194      */
195     int8_t btype;
196     /* A copy of cpu->dcz_blocksize. */
197     uint8_t dcz_blocksize;
198     /* A copy of cpu->gm_blocksize. */
199     uint8_t gm_blocksize;
200     /* True if the current insn_start has been updated. */
201     bool insn_start_updated;
202     /* Offset from VNCR_EL2 when FEAT_NV2 redirects this reg to memory */
203     uint32_t nv2_redirect_offset;
204 } DisasContext;
205 
206 typedef struct DisasCompare {
207     TCGCond cond;
208     TCGv_i32 value;
209 } DisasCompare;
210 
211 /* Share the TCG temporaries common between 32 and 64 bit modes.  */
212 extern TCGv_i32 cpu_NF, cpu_ZF, cpu_CF, cpu_VF;
213 extern TCGv_i64 cpu_exclusive_addr;
214 extern TCGv_i64 cpu_exclusive_val;
215 
216 /*
217  * Constant expanders for the decoders.
218  */
219 
220 static inline int negate(DisasContext *s, int x)
221 {
222     return -x;
223 }
224 
225 static inline int plus_1(DisasContext *s, int x)
226 {
227     return x + 1;
228 }
229 
230 static inline int plus_2(DisasContext *s, int x)
231 {
232     return x + 2;
233 }
234 
235 static inline int plus_8(DisasContext *s, int x)
236 {
237     return x + 8;
238 }
239 
240 static inline int plus_12(DisasContext *s, int x)
241 {
242     return x + 12;
243 }
244 
245 static inline int times_2(DisasContext *s, int x)
246 {
247     return x * 2;
248 }
249 
250 static inline int times_4(DisasContext *s, int x)
251 {
252     return x * 4;
253 }
254 
255 static inline int times_8(DisasContext *s, int x)
256 {
257     return x * 8;
258 }
259 
260 static inline int times_2_plus_1(DisasContext *s, int x)
261 {
262     return x * 2 + 1;
263 }
264 
265 static inline int rsub_64(DisasContext *s, int x)
266 {
267     return 64 - x;
268 }
269 
270 static inline int rsub_32(DisasContext *s, int x)
271 {
272     return 32 - x;
273 }
274 
275 static inline int rsub_16(DisasContext *s, int x)
276 {
277     return 16 - x;
278 }
279 
280 static inline int rsub_8(DisasContext *s, int x)
281 {
282     return 8 - x;
283 }
284 
285 static inline int shl_12(DisasContext *s, int x)
286 {
287     return x << 12;
288 }
289 
290 static inline int xor_2(DisasContext *s, int x)
291 {
292     return x ^ 2;
293 }
294 
295 static inline int neon_3same_fp_size(DisasContext *s, int x)
296 {
297     /* Convert 0==fp32, 1==fp16 into a MO_* value */
298     return MO_32 - x;
299 }
300 
301 static inline int arm_dc_feature(DisasContext *dc, int feature)
302 {
303     return (dc->features & (1ULL << feature)) != 0;
304 }
305 
306 static inline int get_mem_index(DisasContext *s)
307 {
308     return arm_to_core_mmu_idx(s->mmu_idx);
309 }
310 
311 static inline void disas_set_insn_syndrome(DisasContext *s, uint32_t syn)
312 {
313     /* We don't need to save all of the syndrome so we mask and shift
314      * out unneeded bits to help the sleb128 encoder do a better job.
315      */
316     syn &= ARM_INSN_START_WORD2_MASK;
317     syn >>= ARM_INSN_START_WORD2_SHIFT;
318 
319     /* Check for multiple updates.  */
320     assert(!s->insn_start_updated);
321     s->insn_start_updated = true;
322     tcg_set_insn_start_param(s->base.insn_start, 2, syn);
323 }
324 
325 static inline int curr_insn_len(DisasContext *s)
326 {
327     return s->base.pc_next - s->pc_curr;
328 }
329 
330 /* is_jmp field values */
331 #define DISAS_JUMP      DISAS_TARGET_0 /* only pc was modified dynamically */
332 /* CPU state was modified dynamically; exit to main loop for interrupts. */
333 #define DISAS_UPDATE_EXIT  DISAS_TARGET_1
334 /* These instructions trap after executing, so the A32/T32 decoder must
335  * defer them until after the conditional execution state has been updated.
336  * WFI also needs special handling when single-stepping.
337  */
338 #define DISAS_WFI       DISAS_TARGET_2
339 #define DISAS_SWI       DISAS_TARGET_3
340 /* WFE */
341 #define DISAS_WFE       DISAS_TARGET_4
342 #define DISAS_HVC       DISAS_TARGET_5
343 #define DISAS_SMC       DISAS_TARGET_6
344 #define DISAS_YIELD     DISAS_TARGET_7
345 /* M profile branch which might be an exception return (and so needs
346  * custom end-of-TB code)
347  */
348 #define DISAS_BX_EXCRET DISAS_TARGET_8
349 /*
350  * For instructions which want an immediate exit to the main loop, as opposed
351  * to attempting to use lookup_and_goto_ptr.  Unlike DISAS_UPDATE_EXIT, this
352  * doesn't write the PC on exiting the translation loop so you need to ensure
353  * something (gen_a64_update_pc or runtime helper) has done so before we reach
354  * return from cpu_tb_exec.
355  */
356 #define DISAS_EXIT      DISAS_TARGET_9
357 /* CPU state was modified dynamically; no need to exit, but do not chain. */
358 #define DISAS_UPDATE_NOCHAIN  DISAS_TARGET_10
359 
360 #ifdef TARGET_AARCH64
361 void a64_translate_init(void);
362 void gen_a64_update_pc(DisasContext *s, target_long diff);
363 extern const TranslatorOps aarch64_translator_ops;
364 #else
365 static inline void a64_translate_init(void)
366 {
367 }
368 
369 static inline void gen_a64_update_pc(DisasContext *s, target_long diff)
370 {
371 }
372 #endif
373 
374 void arm_test_cc(DisasCompare *cmp, int cc);
375 void arm_jump_cc(DisasCompare *cmp, TCGLabel *label);
376 void arm_gen_test_cc(int cc, TCGLabel *label);
377 MemOp pow2_align(unsigned i);
378 void unallocated_encoding(DisasContext *s);
379 void gen_exception_internal(int excp);
380 void gen_exception_insn_el(DisasContext *s, target_long pc_diff, int excp,
381                            uint32_t syn, uint32_t target_el);
382 void gen_exception_insn(DisasContext *s, target_long pc_diff,
383                         int excp, uint32_t syn);
384 TCGLabel *delay_exception_el(DisasContext *s, int excp,
385                              uint32_t syn, uint32_t target_el);
386 TCGLabel *delay_exception(DisasContext *s, int excp, uint32_t syn);
387 void emit_delayed_exceptions(DisasContext *s);
388 
389 /* Return state of Alternate Half-precision flag, caller frees result */
390 static inline TCGv_i32 get_ahp_flag(void)
391 {
392     TCGv_i32 ret = tcg_temp_new_i32();
393 
394     tcg_gen_ld_i32(ret, tcg_env, offsetoflow32(CPUARMState, vfp.fpcr));
395     tcg_gen_extract_i32(ret, ret, 26, 1);
396 
397     return ret;
398 }
399 
400 /* Set bits within PSTATE.  */
401 static inline void set_pstate_bits(uint64_t bits)
402 {
403     TCGv_i64 p = tcg_temp_new_i64();
404 
405     tcg_debug_assert(!(bits & CACHED_PSTATE_BITS));
406 
407     tcg_gen_ld_i64(p, tcg_env, offsetof(CPUARMState, pstate));
408     tcg_gen_ori_i64(p, p, bits);
409     tcg_gen_st_i64(p, tcg_env, offsetof(CPUARMState, pstate));
410 }
411 
412 /* Clear bits within PSTATE.  */
413 static inline void clear_pstate_bits(uint64_t bits)
414 {
415     TCGv_i64 p = tcg_temp_new_i64();
416 
417     tcg_debug_assert(!(bits & CACHED_PSTATE_BITS));
418 
419     tcg_gen_ld_i64(p, tcg_env, offsetof(CPUARMState, pstate));
420     tcg_gen_andi_i64(p, p, ~bits);
421     tcg_gen_st_i64(p, tcg_env, offsetof(CPUARMState, pstate));
422 }
423 
424 /* If the singlestep state is Active-not-pending, advance to Active-pending. */
425 static inline void gen_ss_advance(DisasContext *s)
426 {
427     if (s->ss_active) {
428         s->pstate_ss = 0;
429         clear_pstate_bits(PSTATE_SS);
430     }
431 }
432 
433 /* Generate an architectural singlestep exception */
434 static inline void gen_swstep_exception(DisasContext *s, int isv, int ex)
435 {
436     /* Fill in the same_el field of the syndrome in the helper. */
437     uint32_t syn = syn_swstep(false, isv, ex);
438     gen_helper_exception_swstep(tcg_env, tcg_constant_i32(syn));
439 }
440 
441 /*
442  * Given a VFP floating point constant encoded into an 8 bit immediate in an
443  * instruction, expand it to the actual constant value of the specified
444  * size, as per the VFPExpandImm() pseudocode in the Arm ARM.
445  */
446 uint64_t vfp_expand_imm(int size, uint8_t imm8);
447 
448 static inline void gen_vfp_absh(TCGv_i32 d, TCGv_i32 s)
449 {
450     tcg_gen_andi_i32(d, s, INT16_MAX);
451 }
452 
453 static inline void gen_vfp_abss(TCGv_i32 d, TCGv_i32 s)
454 {
455     tcg_gen_andi_i32(d, s, INT32_MAX);
456 }
457 
458 static inline void gen_vfp_absd(TCGv_i64 d, TCGv_i64 s)
459 {
460     tcg_gen_andi_i64(d, s, INT64_MAX);
461 }
462 
463 static inline void gen_vfp_negh(TCGv_i32 d, TCGv_i32 s)
464 {
465     tcg_gen_xori_i32(d, s, 1u << 15);
466 }
467 
468 static inline void gen_vfp_negs(TCGv_i32 d, TCGv_i32 s)
469 {
470     tcg_gen_xori_i32(d, s, 1u << 31);
471 }
472 
473 static inline void gen_vfp_negd(TCGv_i64 d, TCGv_i64 s)
474 {
475     tcg_gen_xori_i64(d, s, 1ull << 63);
476 }
477 
478 /* Vector operations shared between ARM and AArch64.  */
479 void gen_gvec_ceq0(unsigned vece, uint32_t rd_ofs, uint32_t rm_ofs,
480                    uint32_t opr_sz, uint32_t max_sz);
481 void gen_gvec_clt0(unsigned vece, uint32_t rd_ofs, uint32_t rm_ofs,
482                    uint32_t opr_sz, uint32_t max_sz);
483 void gen_gvec_cgt0(unsigned vece, uint32_t rd_ofs, uint32_t rm_ofs,
484                    uint32_t opr_sz, uint32_t max_sz);
485 void gen_gvec_cle0(unsigned vece, uint32_t rd_ofs, uint32_t rm_ofs,
486                    uint32_t opr_sz, uint32_t max_sz);
487 void gen_gvec_cge0(unsigned vece, uint32_t rd_ofs, uint32_t rm_ofs,
488                    uint32_t opr_sz, uint32_t max_sz);
489 
490 void gen_gvec_mla(unsigned vece, uint32_t rd_ofs, uint32_t rn_ofs,
491                   uint32_t rm_ofs, uint32_t opr_sz, uint32_t max_sz);
492 void gen_gvec_mls(unsigned vece, uint32_t rd_ofs, uint32_t rn_ofs,
493                   uint32_t rm_ofs, uint32_t opr_sz, uint32_t max_sz);
494 
495 void gen_gvec_cmtst(unsigned vece, uint32_t rd_ofs, uint32_t rn_ofs,
496                     uint32_t rm_ofs, uint32_t opr_sz, uint32_t max_sz);
497 void gen_gvec_sshl(unsigned vece, uint32_t rd_ofs, uint32_t rn_ofs,
498                    uint32_t rm_ofs, uint32_t opr_sz, uint32_t max_sz);
499 void gen_gvec_ushl(unsigned vece, uint32_t rd_ofs, uint32_t rn_ofs,
500                    uint32_t rm_ofs, uint32_t opr_sz, uint32_t max_sz);
501 void gen_gvec_srshl(unsigned vece, uint32_t rd_ofs, uint32_t rn_ofs,
502                     uint32_t rm_ofs, uint32_t opr_sz, uint32_t max_sz);
503 void gen_gvec_urshl(unsigned vece, uint32_t rd_ofs, uint32_t rn_ofs,
504                     uint32_t rm_ofs, uint32_t opr_sz, uint32_t max_sz);
505 void gen_neon_sqshl(unsigned vece, uint32_t rd_ofs, uint32_t rn_ofs,
506                     uint32_t rm_ofs, uint32_t opr_sz, uint32_t max_sz);
507 void gen_neon_uqshl(unsigned vece, uint32_t rd_ofs, uint32_t rn_ofs,
508                     uint32_t rm_ofs, uint32_t opr_sz, uint32_t max_sz);
509 void gen_neon_sqrshl(unsigned vece, uint32_t rd_ofs, uint32_t rn_ofs,
510                      uint32_t rm_ofs, uint32_t opr_sz, uint32_t max_sz);
511 void gen_neon_uqrshl(unsigned vece, uint32_t rd_ofs, uint32_t rn_ofs,
512                      uint32_t rm_ofs, uint32_t opr_sz, uint32_t max_sz);
513 
514 void gen_neon_sqshli(unsigned vece, uint32_t rd_ofs, uint32_t rn_ofs,
515                      int64_t c, uint32_t opr_sz, uint32_t max_sz);
516 void gen_neon_uqshli(unsigned vece, uint32_t rd_ofs, uint32_t rn_ofs,
517                      int64_t c, uint32_t opr_sz, uint32_t max_sz);
518 void gen_neon_sqshlui(unsigned vece, uint32_t rd_ofs, uint32_t rn_ofs,
519                       int64_t c, uint32_t opr_sz, uint32_t max_sz);
520 
521 void gen_gvec_shadd(unsigned vece, uint32_t rd_ofs, uint32_t rn_ofs,
522                     uint32_t rm_ofs, uint32_t opr_sz, uint32_t max_sz);
523 void gen_gvec_uhadd(unsigned vece, uint32_t rd_ofs, uint32_t rn_ofs,
524                     uint32_t rm_ofs, uint32_t opr_sz, uint32_t max_sz);
525 void gen_gvec_shsub(unsigned vece, uint32_t rd_ofs, uint32_t rn_ofs,
526                     uint32_t rm_ofs, uint32_t opr_sz, uint32_t max_sz);
527 void gen_gvec_uhsub(unsigned vece, uint32_t rd_ofs, uint32_t rn_ofs,
528                     uint32_t rm_ofs, uint32_t opr_sz, uint32_t max_sz);
529 void gen_gvec_srhadd(unsigned vece, uint32_t rd_ofs, uint32_t rn_ofs,
530                      uint32_t rm_ofs, uint32_t opr_sz, uint32_t max_sz);
531 void gen_gvec_urhadd(unsigned vece, uint32_t rd_ofs, uint32_t rn_ofs,
532                      uint32_t rm_ofs, uint32_t opr_sz, uint32_t max_sz);
533 
534 void gen_cmtst_i64(TCGv_i64 d, TCGv_i64 a, TCGv_i64 b);
535 void gen_ushl_i32(TCGv_i32 d, TCGv_i32 a, TCGv_i32 b);
536 void gen_sshl_i32(TCGv_i32 d, TCGv_i32 a, TCGv_i32 b);
537 void gen_ushl_i64(TCGv_i64 d, TCGv_i64 a, TCGv_i64 b);
538 void gen_sshl_i64(TCGv_i64 d, TCGv_i64 a, TCGv_i64 b);
539 
540 void gen_uqadd_bhs(TCGv_i64 res, TCGv_i64 qc,
541                    TCGv_i64 a, TCGv_i64 b, MemOp esz);
542 void gen_uqadd_d(TCGv_i64 d, TCGv_i64 q, TCGv_i64 a, TCGv_i64 b);
543 void gen_gvec_uqadd_qc(unsigned vece, uint32_t rd_ofs, uint32_t rn_ofs,
544                        uint32_t rm_ofs, uint32_t opr_sz, uint32_t max_sz);
545 
546 void gen_sqadd_bhs(TCGv_i64 res, TCGv_i64 qc,
547                    TCGv_i64 a, TCGv_i64 b, MemOp esz);
548 void gen_sqadd_d(TCGv_i64 d, TCGv_i64 q, TCGv_i64 a, TCGv_i64 b);
549 void gen_gvec_sqadd_qc(unsigned vece, uint32_t rd_ofs, uint32_t rn_ofs,
550                        uint32_t rm_ofs, uint32_t opr_sz, uint32_t max_sz);
551 
552 void gen_uqsub_bhs(TCGv_i64 res, TCGv_i64 qc,
553                    TCGv_i64 a, TCGv_i64 b, MemOp esz);
554 void gen_uqsub_d(TCGv_i64 d, TCGv_i64 q, TCGv_i64 a, TCGv_i64 b);
555 void gen_gvec_uqsub_qc(unsigned vece, uint32_t rd_ofs, uint32_t rn_ofs,
556                        uint32_t rm_ofs, uint32_t opr_sz, uint32_t max_sz);
557 
558 void gen_sqsub_bhs(TCGv_i64 res, TCGv_i64 qc,
559                    TCGv_i64 a, TCGv_i64 b, MemOp esz);
560 void gen_sqsub_d(TCGv_i64 d, TCGv_i64 q, TCGv_i64 a, TCGv_i64 b);
561 void gen_gvec_sqsub_qc(unsigned vece, uint32_t rd_ofs, uint32_t rn_ofs,
562                        uint32_t rm_ofs, uint32_t opr_sz, uint32_t max_sz);
563 
564 void gen_gvec_sshr(unsigned vece, uint32_t rd_ofs, uint32_t rm_ofs,
565                    int64_t shift, uint32_t opr_sz, uint32_t max_sz);
566 void gen_gvec_ushr(unsigned vece, uint32_t rd_ofs, uint32_t rm_ofs,
567                    int64_t shift, uint32_t opr_sz, uint32_t max_sz);
568 
569 void gen_gvec_ssra(unsigned vece, uint32_t rd_ofs, uint32_t rm_ofs,
570                    int64_t shift, uint32_t opr_sz, uint32_t max_sz);
571 void gen_gvec_usra(unsigned vece, uint32_t rd_ofs, uint32_t rm_ofs,
572                    int64_t shift, uint32_t opr_sz, uint32_t max_sz);
573 
574 void gen_srshr32_i32(TCGv_i32 d, TCGv_i32 a, int32_t sh);
575 void gen_srshr64_i64(TCGv_i64 d, TCGv_i64 a, int64_t sh);
576 void gen_urshr32_i32(TCGv_i32 d, TCGv_i32 a, int32_t sh);
577 void gen_urshr64_i64(TCGv_i64 d, TCGv_i64 a, int64_t sh);
578 
579 void gen_gvec_srshr(unsigned vece, uint32_t rd_ofs, uint32_t rm_ofs,
580                     int64_t shift, uint32_t opr_sz, uint32_t max_sz);
581 void gen_gvec_urshr(unsigned vece, uint32_t rd_ofs, uint32_t rm_ofs,
582                     int64_t shift, uint32_t opr_sz, uint32_t max_sz);
583 void gen_gvec_srsra(unsigned vece, uint32_t rd_ofs, uint32_t rm_ofs,
584                     int64_t shift, uint32_t opr_sz, uint32_t max_sz);
585 void gen_gvec_ursra(unsigned vece, uint32_t rd_ofs, uint32_t rm_ofs,
586                     int64_t shift, uint32_t opr_sz, uint32_t max_sz);
587 
588 void gen_gvec_sri(unsigned vece, uint32_t rd_ofs, uint32_t rm_ofs,
589                   int64_t shift, uint32_t opr_sz, uint32_t max_sz);
590 void gen_gvec_sli(unsigned vece, uint32_t rd_ofs, uint32_t rm_ofs,
591                   int64_t shift, uint32_t opr_sz, uint32_t max_sz);
592 
593 void gen_gvec_sqdmulh_qc(unsigned vece, uint32_t rd_ofs, uint32_t rn_ofs,
594                          uint32_t rm_ofs, uint32_t opr_sz, uint32_t max_sz);
595 void gen_gvec_sqrdmulh_qc(unsigned vece, uint32_t rd_ofs, uint32_t rn_ofs,
596                           uint32_t rm_ofs, uint32_t opr_sz, uint32_t max_sz);
597 void gen_gvec_sqrdmlah_qc(unsigned vece, uint32_t rd_ofs, uint32_t rn_ofs,
598                           uint32_t rm_ofs, uint32_t opr_sz, uint32_t max_sz);
599 void gen_gvec_sqrdmlsh_qc(unsigned vece, uint32_t rd_ofs, uint32_t rn_ofs,
600                           uint32_t rm_ofs, uint32_t opr_sz, uint32_t max_sz);
601 
602 void gen_gvec_sabd(unsigned vece, uint32_t rd_ofs, uint32_t rn_ofs,
603                    uint32_t rm_ofs, uint32_t opr_sz, uint32_t max_sz);
604 void gen_gvec_uabd(unsigned vece, uint32_t rd_ofs, uint32_t rn_ofs,
605                    uint32_t rm_ofs, uint32_t opr_sz, uint32_t max_sz);
606 
607 void gen_gvec_saba(unsigned vece, uint32_t rd_ofs, uint32_t rn_ofs,
608                    uint32_t rm_ofs, uint32_t opr_sz, uint32_t max_sz);
609 void gen_gvec_uaba(unsigned vece, uint32_t rd_ofs, uint32_t rn_ofs,
610                    uint32_t rm_ofs, uint32_t opr_sz, uint32_t max_sz);
611 
612 void gen_gvec_addp(unsigned vece, uint32_t rd_ofs, uint32_t rn_ofs,
613                    uint32_t rm_ofs, uint32_t opr_sz, uint32_t max_sz);
614 void gen_gvec_smaxp(unsigned vece, uint32_t rd_ofs, uint32_t rn_ofs,
615                     uint32_t rm_ofs, uint32_t opr_sz, uint32_t max_sz);
616 void gen_gvec_sminp(unsigned vece, uint32_t rd_ofs, uint32_t rn_ofs,
617                     uint32_t rm_ofs, uint32_t opr_sz, uint32_t max_sz);
618 void gen_gvec_umaxp(unsigned vece, uint32_t rd_ofs, uint32_t rn_ofs,
619                     uint32_t rm_ofs, uint32_t opr_sz, uint32_t max_sz);
620 void gen_gvec_uminp(unsigned vece, uint32_t rd_ofs, uint32_t rn_ofs,
621                     uint32_t rm_ofs, uint32_t opr_sz, uint32_t max_sz);
622 
623 void gen_gvec_cls(unsigned vece, uint32_t rd_ofs, uint32_t rn_ofs,
624                   uint32_t opr_sz, uint32_t max_sz);
625 void gen_gvec_clz(unsigned vece, uint32_t rd_ofs, uint32_t rn_ofs,
626                   uint32_t opr_sz, uint32_t max_sz);
627 void gen_gvec_cnt(unsigned vece, uint32_t rd_ofs, uint32_t rn_ofs,
628                   uint32_t opr_sz, uint32_t max_sz);
629 void gen_gvec_rbit(unsigned vece, uint32_t rd_ofs, uint32_t rn_ofs,
630                    uint32_t opr_sz, uint32_t max_sz);
631 void gen_gvec_rev16(unsigned vece, uint32_t rd_ofs, uint32_t rn_ofs,
632                     uint32_t opr_sz, uint32_t max_sz);
633 void gen_gvec_rev32(unsigned vece, uint32_t rd_ofs, uint32_t rn_ofs,
634                     uint32_t opr_sz, uint32_t max_sz);
635 void gen_gvec_rev64(unsigned vece, uint32_t rd_ofs, uint32_t rn_ofs,
636                     uint32_t opr_sz, uint32_t max_sz);
637 
638 void gen_gvec_saddlp(unsigned vece, uint32_t rd_ofs, uint32_t rn_ofs,
639                      uint32_t opr_sz, uint32_t max_sz);
640 void gen_gvec_sadalp(unsigned vece, uint32_t rd_ofs, uint32_t rn_ofs,
641                      uint32_t opr_sz, uint32_t max_sz);
642 void gen_gvec_uaddlp(unsigned vece, uint32_t rd_ofs, uint32_t rn_ofs,
643                      uint32_t opr_sz, uint32_t max_sz);
644 void gen_gvec_uadalp(unsigned vece, uint32_t rd_ofs, uint32_t rn_ofs,
645                      uint32_t opr_sz, uint32_t max_sz);
646 
647 /* These exclusively manipulate the sign bit. */
648 void gen_gvec_fabs(unsigned vece, uint32_t dofs, uint32_t aofs,
649                    uint32_t oprsz, uint32_t maxsz);
650 void gen_gvec_fneg(unsigned vece, uint32_t dofs, uint32_t aofs,
651                    uint32_t oprsz, uint32_t maxsz);
652 
653 void gen_gvec_urecpe(unsigned vece, uint32_t rd_ofs, uint32_t rn_ofs,
654                      uint32_t opr_sz, uint32_t max_sz);
655 void gen_gvec_ursqrte(unsigned vece, uint32_t rd_ofs, uint32_t rn_ofs,
656                       uint32_t opr_sz, uint32_t max_sz);
657 
658 /*
659  * Forward to the isar_feature_* tests given a DisasContext pointer.
660  */
661 #define dc_isar_feature(name, ctx) \
662     ({ DisasContext *ctx_ = (ctx); isar_feature_##name(ctx_->isar); })
663 
664 /* Note that the gvec expanders operate on offsets + sizes.  */
665 typedef void GVecGen2Fn(unsigned, uint32_t, uint32_t, uint32_t, uint32_t);
666 typedef void GVecGen2iFn(unsigned, uint32_t, uint32_t, int64_t,
667                          uint32_t, uint32_t);
668 typedef void GVecGen3Fn(unsigned, uint32_t, uint32_t,
669                         uint32_t, uint32_t, uint32_t);
670 typedef void GVecGen4Fn(unsigned, uint32_t, uint32_t, uint32_t,
671                         uint32_t, uint32_t, uint32_t);
672 typedef void GVecGen3FnVar(unsigned, TCGv_ptr, uint32_t, TCGv_ptr, uint32_t,
673                            TCGv_ptr, uint32_t, uint32_t, uint32_t);
674 
675 /* Function prototype for gen_ functions for calling Neon helpers */
676 typedef void NeonGenOneOpFn(TCGv_i32, TCGv_i32);
677 typedef void NeonGenOneOpEnvFn(TCGv_i32, TCGv_ptr, TCGv_i32);
678 typedef void NeonGenTwoOpFn(TCGv_i32, TCGv_i32, TCGv_i32);
679 typedef void NeonGenTwoOpEnvFn(TCGv_i32, TCGv_ptr, TCGv_i32, TCGv_i32);
680 typedef void NeonGenThreeOpEnvFn(TCGv_i32, TCGv_env, TCGv_i32,
681                                  TCGv_i32, TCGv_i32);
682 typedef void NeonGenTwo64OpFn(TCGv_i64, TCGv_i64, TCGv_i64);
683 typedef void NeonGenTwo64OpEnvFn(TCGv_i64, TCGv_ptr, TCGv_i64, TCGv_i64);
684 typedef void NeonGenNarrowFn(TCGv_i32, TCGv_i64);
685 typedef void NeonGenWidenFn(TCGv_i64, TCGv_i32);
686 typedef void NeonGenTwoOpWidenFn(TCGv_i64, TCGv_i32, TCGv_i32);
687 typedef void NeonGenOneSingleOpFn(TCGv_i32, TCGv_i32, TCGv_ptr);
688 typedef void NeonGenTwoSingleOpFn(TCGv_i32, TCGv_i32, TCGv_i32, TCGv_ptr);
689 typedef void NeonGenTwoDoubleOpFn(TCGv_i64, TCGv_i64, TCGv_i64, TCGv_ptr);
690 typedef void NeonGenOne64OpFn(TCGv_i64, TCGv_i64);
691 typedef void NeonGenOne64OpEnvFn(TCGv_i64, TCGv_env, TCGv_i64);
692 typedef void CryptoTwoOpFn(TCGv_ptr, TCGv_ptr);
693 typedef void CryptoThreeOpIntFn(TCGv_ptr, TCGv_ptr, TCGv_i32);
694 typedef void CryptoThreeOpFn(TCGv_ptr, TCGv_ptr, TCGv_ptr);
695 typedef void AtomicThreeOpFn(TCGv_i64, TCGv_i64, TCGv_i64, TCGArg, MemOp);
696 typedef void WideShiftImmFn(TCGv_i64, TCGv_i64, int64_t shift);
697 typedef void WideShiftFn(TCGv_i64, TCGv_ptr, TCGv_i64, TCGv_i32);
698 typedef void ShiftImmFn(TCGv_i32, TCGv_i32, int32_t shift);
699 typedef void ShiftFn(TCGv_i32, TCGv_ptr, TCGv_i32, TCGv_i32);
700 
701 /**
702  * arm_tbflags_from_tb:
703  * @tb: the TranslationBlock
704  *
705  * Extract the flag values from @tb.
706  */
707 static inline CPUARMTBFlags arm_tbflags_from_tb(const TranslationBlock *tb)
708 {
709     return (CPUARMTBFlags){ tb->flags, tb->cs_base };
710 }
711 
712 /**
713  * fpstatus_ptr: return TCGv_ptr to the specified fp_status field
714  *
715  * We have multiple softfloat float_status fields in the Arm CPU state struct
716  * (see the comment in cpu.h for details). Return a TCGv_ptr which has
717  * been set up to point to the requested field in the CPU state struct.
718  */
719 static inline TCGv_ptr fpstatus_ptr(ARMFPStatusFlavour flavour)
720 {
721     TCGv_ptr statusptr = tcg_temp_new_ptr();
722     int offset = offsetof(CPUARMState, vfp.fp_status[flavour]);
723 
724     tcg_gen_addi_ptr(statusptr, tcg_env, offset);
725     return statusptr;
726 }
727 
728 /**
729  * finalize_memop_atom:
730  * @s: DisasContext
731  * @opc: size+sign+align of the memory operation
732  * @atom: atomicity of the memory operation
733  *
734  * Build the complete MemOp for a memory operation, including alignment,
735  * endianness, and atomicity.
736  *
737  * If (op & MO_AMASK) then the operation already contains the required
738  * alignment, e.g. for AccType_ATOMIC.  Otherwise, this an optionally
739  * unaligned operation, e.g. for AccType_NORMAL.
740  *
741  * In the latter case, there are configuration bits that require alignment,
742  * and this is applied here.  Note that there is no way to indicate that
743  * no alignment should ever be enforced; this must be handled manually.
744  */
745 static inline MemOp finalize_memop_atom(DisasContext *s, MemOp opc, MemOp atom)
746 {
747     if (!(opc & MO_AMASK)) {
748         opc |= MO_ALIGN | (s->align_mem ? 0 : MO_ALIGN_TLB_ONLY);
749     }
750     return opc | atom | s->be_data;
751 }
752 
753 /**
754  * finalize_memop:
755  * @s: DisasContext
756  * @opc: size+sign+align of the memory operation
757  *
758  * Like finalize_memop_atom, but with default atomicity.
759  */
760 static inline MemOp finalize_memop(DisasContext *s, MemOp opc)
761 {
762     MemOp atom = s->lse2 ? MO_ATOM_WITHIN16 : MO_ATOM_IFALIGN;
763     return finalize_memop_atom(s, opc, atom);
764 }
765 
766 /**
767  * finalize_memop_pair:
768  * @s: DisasContext
769  * @opc: size+sign+align of the memory operation
770  *
771  * Like finalize_memop_atom, but with atomicity for a pair.
772  * C.f. Pseudocode for Mem[], operand ispair.
773  */
774 static inline MemOp finalize_memop_pair(DisasContext *s, MemOp opc)
775 {
776     MemOp atom = s->lse2 ? MO_ATOM_WITHIN16_PAIR : MO_ATOM_IFALIGN_PAIR;
777     return finalize_memop_atom(s, opc, atom);
778 }
779 
780 /**
781  * finalize_memop_asimd:
782  * @s: DisasContext
783  * @opc: size+sign+align of the memory operation
784  *
785  * Like finalize_memop_atom, but with atomicity of AccessType_ASIMD.
786  */
787 static inline MemOp finalize_memop_asimd(DisasContext *s, MemOp opc)
788 {
789     /*
790      * In the pseudocode for Mem[], with AccessType_ASIMD, size == 16,
791      * if IsAligned(8), the first case provides separate atomicity for
792      * the pair of 64-bit accesses.  If !IsAligned(8), the middle cases
793      * do not apply, and we're left with the final case of no atomicity.
794      * Thus MO_ATOM_IFALIGN_PAIR.
795      *
796      * For other sizes, normal LSE2 rules apply.
797      */
798     if ((opc & MO_SIZE) == MO_128) {
799         return finalize_memop_atom(s, opc, MO_ATOM_IFALIGN_PAIR);
800     }
801     return finalize_memop(s, opc);
802 }
803 
804 /**
805  * asimd_imm_const: Expand an encoded SIMD constant value
806  *
807  * Expand a SIMD constant value. This is essentially the pseudocode
808  * AdvSIMDExpandImm, except that we also perform the boolean NOT needed for
809  * VMVN and VBIC (when cmode < 14 && op == 1).
810  *
811  * The combination cmode == 15 op == 1 is a reserved encoding for AArch32;
812  * callers must catch this; we return the 64-bit constant value defined
813  * for AArch64.
814  *
815  * cmode = 2,3,4,5,6,7,10,11,12,13 imm=0 was UNPREDICTABLE in v7A but
816  * is either not unpredictable or merely CONSTRAINED UNPREDICTABLE in v8A;
817  * we produce an immediate constant value of 0 in these cases.
818  */
819 uint64_t asimd_imm_const(uint32_t imm, int cmode, int op);
820 
821 /*
822  * gen_disas_label:
823  * Create a label and cache a copy of pc_save.
824  */
825 static inline DisasLabel gen_disas_label(DisasContext *s)
826 {
827     return (DisasLabel){
828         .label = gen_new_label(),
829         .pc_save = s->pc_save,
830     };
831 }
832 
833 /*
834  * set_disas_label:
835  * Emit a label and restore the cached copy of pc_save.
836  */
837 static inline void set_disas_label(DisasContext *s, DisasLabel l)
838 {
839     gen_set_label(l.label);
840     s->pc_save = l.pc_save;
841 }
842 
843 static inline TCGv_ptr gen_lookup_cp_reg(uint32_t key)
844 {
845     TCGv_ptr ret = tcg_temp_new_ptr();
846     gen_helper_lookup_cp_reg(ret, tcg_env, tcg_constant_i32(key));
847     return ret;
848 }
849 
850 /*
851  * Set and reset rounding mode around another operation.
852  */
853 static inline TCGv_i32 gen_set_rmode(ARMFPRounding rmode, TCGv_ptr fpst)
854 {
855     TCGv_i32 new = tcg_constant_i32(arm_rmode_to_sf(rmode));
856     TCGv_i32 old = tcg_temp_new_i32();
857 
858     gen_helper_set_rmode(old, new, fpst);
859     return old;
860 }
861 
862 static inline void gen_restore_rmode(TCGv_i32 old, TCGv_ptr fpst)
863 {
864     gen_helper_set_rmode(old, old, fpst);
865 }
866 
867 /*
868  * Helpers for implementing sets of trans_* functions.
869  * Defer the implementation of NAME to FUNC, with optional extra arguments.
870  */
871 #define TRANS(NAME, FUNC, ...) \
872     static bool trans_##NAME(DisasContext *s, arg_##NAME *a) \
873     { return FUNC(s, __VA_ARGS__); }
874 #define TRANS_FEAT(NAME, FEAT, FUNC, ...) \
875     static bool trans_##NAME(DisasContext *s, arg_##NAME *a) \
876     { return dc_isar_feature(FEAT, s) && FUNC(s, __VA_ARGS__); }
877 
878 #define TRANS_FEAT_NONSTREAMING(NAME, FEAT, FUNC, ...)            \
879     static bool trans_##NAME(DisasContext *s, arg_##NAME *a)      \
880     {                                                             \
881         s->is_nonstreaming = true;                                \
882         return dc_isar_feature(FEAT, s) && FUNC(s, __VA_ARGS__);  \
883     }
884 
885 #endif /* TARGET_ARM_TRANSLATE_H */
886