xref: /openbmc/qemu/target/arm/tcg/translate-sve.c (revision bd46e45a)
1 /*
2  * AArch64 SVE translation
3  *
4  * Copyright (c) 2018 Linaro, Ltd
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 "tcg/tcg-op-gvec.h"
25 #include "tcg/tcg-gvec-desc.h"
26 #include "qemu/log.h"
27 #include "arm_ldst.h"
28 #include "translate.h"
29 #include "internals.h"
30 #include "exec/helper-proto.h"
31 #include "exec/helper-gen.h"
32 #include "exec/log.h"
33 #include "translate-a64.h"
34 #include "fpu/softfloat.h"
35 
36 
37 typedef void GVecGen2sFn(unsigned, uint32_t, uint32_t,
38                          TCGv_i64, uint32_t, uint32_t);
39 
40 typedef void gen_helper_gvec_flags_3(TCGv_i32, TCGv_ptr, TCGv_ptr,
41                                      TCGv_ptr, TCGv_i32);
42 typedef void gen_helper_gvec_flags_4(TCGv_i32, TCGv_ptr, TCGv_ptr,
43                                      TCGv_ptr, TCGv_ptr, TCGv_i32);
44 
45 typedef void gen_helper_gvec_mem(TCGv_env, TCGv_ptr, TCGv_i64, TCGv_i32);
46 typedef void gen_helper_gvec_mem_scatter(TCGv_env, TCGv_ptr, TCGv_ptr,
47                                          TCGv_ptr, TCGv_i64, TCGv_i32);
48 
49 /*
50  * Helpers for extracting complex instruction fields.
51  */
52 
53 /* See e.g. ASR (immediate, predicated).
54  * Returns -1 for unallocated encoding; diagnose later.
55  */
56 static int tszimm_esz(DisasContext *s, int x)
57 {
58     x >>= 3;  /* discard imm3 */
59     return 31 - clz32(x);
60 }
61 
62 static int tszimm_shr(DisasContext *s, int x)
63 {
64     return (16 << tszimm_esz(s, x)) - x;
65 }
66 
67 /* See e.g. LSL (immediate, predicated).  */
68 static int tszimm_shl(DisasContext *s, int x)
69 {
70     return x - (8 << tszimm_esz(s, x));
71 }
72 
73 /* The SH bit is in bit 8.  Extract the low 8 and shift.  */
74 static inline int expand_imm_sh8s(DisasContext *s, int x)
75 {
76     return (int8_t)x << (x & 0x100 ? 8 : 0);
77 }
78 
79 static inline int expand_imm_sh8u(DisasContext *s, int x)
80 {
81     return (uint8_t)x << (x & 0x100 ? 8 : 0);
82 }
83 
84 /* Convert a 2-bit memory size (msz) to a 4-bit data type (dtype)
85  * with unsigned data.  C.f. SVE Memory Contiguous Load Group.
86  */
87 static inline int msz_dtype(DisasContext *s, int msz)
88 {
89     static const uint8_t dtype[4] = { 0, 5, 10, 15 };
90     return dtype[msz];
91 }
92 
93 /*
94  * Include the generated decoder.
95  */
96 
97 #include "decode-sve.c.inc"
98 
99 /*
100  * Implement all of the translator functions referenced by the decoder.
101  */
102 
103 /* Invoke an out-of-line helper on 2 Zregs. */
104 static bool gen_gvec_ool_zz(DisasContext *s, gen_helper_gvec_2 *fn,
105                             int rd, int rn, int data)
106 {
107     if (fn == NULL) {
108         return false;
109     }
110     if (sve_access_check(s)) {
111         unsigned vsz = vec_full_reg_size(s);
112         tcg_gen_gvec_2_ool(vec_full_reg_offset(s, rd),
113                            vec_full_reg_offset(s, rn),
114                            vsz, vsz, data, fn);
115     }
116     return true;
117 }
118 
119 static bool gen_gvec_fpst_zz(DisasContext *s, gen_helper_gvec_2_ptr *fn,
120                              int rd, int rn, int data,
121                              ARMFPStatusFlavour flavour)
122 {
123     if (fn == NULL) {
124         return false;
125     }
126     if (sve_access_check(s)) {
127         unsigned vsz = vec_full_reg_size(s);
128         TCGv_ptr status = fpstatus_ptr(flavour);
129 
130         tcg_gen_gvec_2_ptr(vec_full_reg_offset(s, rd),
131                            vec_full_reg_offset(s, rn),
132                            status, vsz, vsz, data, fn);
133     }
134     return true;
135 }
136 
137 static bool gen_gvec_fpst_arg_zz(DisasContext *s, gen_helper_gvec_2_ptr *fn,
138                                  arg_rr_esz *a, int data)
139 {
140     return gen_gvec_fpst_zz(s, fn, a->rd, a->rn, data,
141                             a->esz == MO_16 ? FPST_FPCR_F16 : FPST_FPCR);
142 }
143 
144 /* Invoke an out-of-line helper on 3 Zregs. */
145 static bool gen_gvec_ool_zzz(DisasContext *s, gen_helper_gvec_3 *fn,
146                              int rd, int rn, int rm, int data)
147 {
148     if (fn == NULL) {
149         return false;
150     }
151     if (sve_access_check(s)) {
152         unsigned vsz = vec_full_reg_size(s);
153         tcg_gen_gvec_3_ool(vec_full_reg_offset(s, rd),
154                            vec_full_reg_offset(s, rn),
155                            vec_full_reg_offset(s, rm),
156                            vsz, vsz, data, fn);
157     }
158     return true;
159 }
160 
161 static bool gen_gvec_ool_arg_zzz(DisasContext *s, gen_helper_gvec_3 *fn,
162                                  arg_rrr_esz *a, int data)
163 {
164     return gen_gvec_ool_zzz(s, fn, a->rd, a->rn, a->rm, data);
165 }
166 
167 /* Invoke an out-of-line helper on 3 Zregs, plus float_status. */
168 static bool gen_gvec_fpst_zzz(DisasContext *s, gen_helper_gvec_3_ptr *fn,
169                               int rd, int rn, int rm,
170                               int data, ARMFPStatusFlavour flavour)
171 {
172     if (fn == NULL) {
173         return false;
174     }
175     if (sve_access_check(s)) {
176         unsigned vsz = vec_full_reg_size(s);
177         TCGv_ptr status = fpstatus_ptr(flavour);
178 
179         tcg_gen_gvec_3_ptr(vec_full_reg_offset(s, rd),
180                            vec_full_reg_offset(s, rn),
181                            vec_full_reg_offset(s, rm),
182                            status, vsz, vsz, data, fn);
183     }
184     return true;
185 }
186 
187 static bool gen_gvec_fpst_arg_zzz(DisasContext *s, gen_helper_gvec_3_ptr *fn,
188                                   arg_rrr_esz *a, int data)
189 {
190     return gen_gvec_fpst_zzz(s, fn, a->rd, a->rn, a->rm, data,
191                              a->esz == MO_16 ? FPST_FPCR_F16 : FPST_FPCR);
192 }
193 
194 /* Invoke an out-of-line helper on 4 Zregs. */
195 static bool gen_gvec_ool_zzzz(DisasContext *s, gen_helper_gvec_4 *fn,
196                               int rd, int rn, int rm, int ra, int data)
197 {
198     if (fn == NULL) {
199         return false;
200     }
201     if (sve_access_check(s)) {
202         unsigned vsz = vec_full_reg_size(s);
203         tcg_gen_gvec_4_ool(vec_full_reg_offset(s, rd),
204                            vec_full_reg_offset(s, rn),
205                            vec_full_reg_offset(s, rm),
206                            vec_full_reg_offset(s, ra),
207                            vsz, vsz, data, fn);
208     }
209     return true;
210 }
211 
212 static bool gen_gvec_ool_arg_zzzz(DisasContext *s, gen_helper_gvec_4 *fn,
213                                   arg_rrrr_esz *a, int data)
214 {
215     return gen_gvec_ool_zzzz(s, fn, a->rd, a->rn, a->rm, a->ra, data);
216 }
217 
218 static bool gen_gvec_ool_arg_zzxz(DisasContext *s, gen_helper_gvec_4 *fn,
219                                   arg_rrxr_esz *a)
220 {
221     return gen_gvec_ool_zzzz(s, fn, a->rd, a->rn, a->rm, a->ra, a->index);
222 }
223 
224 /* Invoke an out-of-line helper on 4 Zregs, plus a pointer. */
225 static bool gen_gvec_ptr_zzzz(DisasContext *s, gen_helper_gvec_4_ptr *fn,
226                               int rd, int rn, int rm, int ra,
227                               int data, TCGv_ptr ptr)
228 {
229     if (fn == NULL) {
230         return false;
231     }
232     if (sve_access_check(s)) {
233         unsigned vsz = vec_full_reg_size(s);
234         tcg_gen_gvec_4_ptr(vec_full_reg_offset(s, rd),
235                            vec_full_reg_offset(s, rn),
236                            vec_full_reg_offset(s, rm),
237                            vec_full_reg_offset(s, ra),
238                            ptr, vsz, vsz, data, fn);
239     }
240     return true;
241 }
242 
243 static bool gen_gvec_fpst_zzzz(DisasContext *s, gen_helper_gvec_4_ptr *fn,
244                                int rd, int rn, int rm, int ra,
245                                int data, ARMFPStatusFlavour flavour)
246 {
247     TCGv_ptr status = fpstatus_ptr(flavour);
248     bool ret = gen_gvec_ptr_zzzz(s, fn, rd, rn, rm, ra, data, status);
249     return ret;
250 }
251 
252 /* Invoke an out-of-line helper on 4 Zregs, 1 Preg, plus fpst. */
253 static bool gen_gvec_fpst_zzzzp(DisasContext *s, gen_helper_gvec_5_ptr *fn,
254                                 int rd, int rn, int rm, int ra, int pg,
255                                 int data, ARMFPStatusFlavour flavour)
256 {
257     if (fn == NULL) {
258         return false;
259     }
260     if (sve_access_check(s)) {
261         unsigned vsz = vec_full_reg_size(s);
262         TCGv_ptr status = fpstatus_ptr(flavour);
263 
264         tcg_gen_gvec_5_ptr(vec_full_reg_offset(s, rd),
265                            vec_full_reg_offset(s, rn),
266                            vec_full_reg_offset(s, rm),
267                            vec_full_reg_offset(s, ra),
268                            pred_full_reg_offset(s, pg),
269                            status, vsz, vsz, data, fn);
270     }
271     return true;
272 }
273 
274 /* Invoke an out-of-line helper on 2 Zregs and a predicate. */
275 static bool gen_gvec_ool_zzp(DisasContext *s, gen_helper_gvec_3 *fn,
276                              int rd, int rn, int pg, int data)
277 {
278     if (fn == NULL) {
279         return false;
280     }
281     if (sve_access_check(s)) {
282         unsigned vsz = vec_full_reg_size(s);
283         tcg_gen_gvec_3_ool(vec_full_reg_offset(s, rd),
284                            vec_full_reg_offset(s, rn),
285                            pred_full_reg_offset(s, pg),
286                            vsz, vsz, data, fn);
287     }
288     return true;
289 }
290 
291 static bool gen_gvec_ool_arg_zpz(DisasContext *s, gen_helper_gvec_3 *fn,
292                                  arg_rpr_esz *a, int data)
293 {
294     return gen_gvec_ool_zzp(s, fn, a->rd, a->rn, a->pg, data);
295 }
296 
297 static bool gen_gvec_ool_arg_zpzi(DisasContext *s, gen_helper_gvec_3 *fn,
298                                   arg_rpri_esz *a)
299 {
300     return gen_gvec_ool_zzp(s, fn, a->rd, a->rn, a->pg, a->imm);
301 }
302 
303 static bool gen_gvec_fpst_zzp(DisasContext *s, gen_helper_gvec_3_ptr *fn,
304                               int rd, int rn, int pg, int data,
305                               ARMFPStatusFlavour flavour)
306 {
307     if (fn == NULL) {
308         return false;
309     }
310     if (sve_access_check(s)) {
311         unsigned vsz = vec_full_reg_size(s);
312         TCGv_ptr status = fpstatus_ptr(flavour);
313 
314         tcg_gen_gvec_3_ptr(vec_full_reg_offset(s, rd),
315                            vec_full_reg_offset(s, rn),
316                            pred_full_reg_offset(s, pg),
317                            status, vsz, vsz, data, fn);
318     }
319     return true;
320 }
321 
322 static bool gen_gvec_fpst_arg_zpz(DisasContext *s, gen_helper_gvec_3_ptr *fn,
323                                   arg_rpr_esz *a, int data,
324                                   ARMFPStatusFlavour flavour)
325 {
326     return gen_gvec_fpst_zzp(s, fn, a->rd, a->rn, a->pg, data, flavour);
327 }
328 
329 /* Invoke an out-of-line helper on 3 Zregs and a predicate. */
330 static bool gen_gvec_ool_zzzp(DisasContext *s, gen_helper_gvec_4 *fn,
331                               int rd, int rn, int rm, int pg, int data)
332 {
333     if (fn == NULL) {
334         return false;
335     }
336     if (sve_access_check(s)) {
337         unsigned vsz = vec_full_reg_size(s);
338         tcg_gen_gvec_4_ool(vec_full_reg_offset(s, rd),
339                            vec_full_reg_offset(s, rn),
340                            vec_full_reg_offset(s, rm),
341                            pred_full_reg_offset(s, pg),
342                            vsz, vsz, data, fn);
343     }
344     return true;
345 }
346 
347 static bool gen_gvec_ool_arg_zpzz(DisasContext *s, gen_helper_gvec_4 *fn,
348                                   arg_rprr_esz *a, int data)
349 {
350     return gen_gvec_ool_zzzp(s, fn, a->rd, a->rn, a->rm, a->pg, data);
351 }
352 
353 /* Invoke an out-of-line helper on 3 Zregs and a predicate. */
354 static bool gen_gvec_fpst_zzzp(DisasContext *s, gen_helper_gvec_4_ptr *fn,
355                                int rd, int rn, int rm, int pg, int data,
356                                ARMFPStatusFlavour flavour)
357 {
358     if (fn == NULL) {
359         return false;
360     }
361     if (sve_access_check(s)) {
362         unsigned vsz = vec_full_reg_size(s);
363         TCGv_ptr status = fpstatus_ptr(flavour);
364 
365         tcg_gen_gvec_4_ptr(vec_full_reg_offset(s, rd),
366                            vec_full_reg_offset(s, rn),
367                            vec_full_reg_offset(s, rm),
368                            pred_full_reg_offset(s, pg),
369                            status, vsz, vsz, data, fn);
370     }
371     return true;
372 }
373 
374 static bool gen_gvec_fpst_arg_zpzz(DisasContext *s, gen_helper_gvec_4_ptr *fn,
375                                    arg_rprr_esz *a)
376 {
377     return gen_gvec_fpst_zzzp(s, fn, a->rd, a->rn, a->rm, a->pg, 0,
378                               a->esz == MO_16 ? FPST_FPCR_F16 : FPST_FPCR);
379 }
380 
381 /* Invoke a vector expander on two Zregs and an immediate.  */
382 static bool gen_gvec_fn_zzi(DisasContext *s, GVecGen2iFn *gvec_fn,
383                             int esz, int rd, int rn, uint64_t imm)
384 {
385     if (gvec_fn == NULL) {
386         return false;
387     }
388     if (sve_access_check(s)) {
389         unsigned vsz = vec_full_reg_size(s);
390         gvec_fn(esz, vec_full_reg_offset(s, rd),
391                 vec_full_reg_offset(s, rn), imm, vsz, vsz);
392     }
393     return true;
394 }
395 
396 static bool gen_gvec_fn_arg_zzi(DisasContext *s, GVecGen2iFn *gvec_fn,
397                                 arg_rri_esz *a)
398 {
399     if (a->esz < 0) {
400         /* Invalid tsz encoding -- see tszimm_esz. */
401         return false;
402     }
403     return gen_gvec_fn_zzi(s, gvec_fn, a->esz, a->rd, a->rn, a->imm);
404 }
405 
406 /* Invoke a vector expander on three Zregs.  */
407 static bool gen_gvec_fn_zzz(DisasContext *s, GVecGen3Fn *gvec_fn,
408                             int esz, int rd, int rn, int rm)
409 {
410     if (gvec_fn == NULL) {
411         return false;
412     }
413     if (sve_access_check(s)) {
414         unsigned vsz = vec_full_reg_size(s);
415         gvec_fn(esz, vec_full_reg_offset(s, rd),
416                 vec_full_reg_offset(s, rn),
417                 vec_full_reg_offset(s, rm), vsz, vsz);
418     }
419     return true;
420 }
421 
422 static bool gen_gvec_fn_arg_zzz(DisasContext *s, GVecGen3Fn *fn,
423                                 arg_rrr_esz *a)
424 {
425     return gen_gvec_fn_zzz(s, fn, a->esz, a->rd, a->rn, a->rm);
426 }
427 
428 /* Invoke a vector expander on four Zregs.  */
429 static bool gen_gvec_fn_arg_zzzz(DisasContext *s, GVecGen4Fn *gvec_fn,
430                                  arg_rrrr_esz *a)
431 {
432     if (gvec_fn == NULL) {
433         return false;
434     }
435     if (sve_access_check(s)) {
436         unsigned vsz = vec_full_reg_size(s);
437         gvec_fn(a->esz, vec_full_reg_offset(s, a->rd),
438                 vec_full_reg_offset(s, a->rn),
439                 vec_full_reg_offset(s, a->rm),
440                 vec_full_reg_offset(s, a->ra), vsz, vsz);
441     }
442     return true;
443 }
444 
445 /* Invoke a vector move on two Zregs.  */
446 static bool do_mov_z(DisasContext *s, int rd, int rn)
447 {
448     if (sve_access_check(s)) {
449         unsigned vsz = vec_full_reg_size(s);
450         tcg_gen_gvec_mov(MO_8, vec_full_reg_offset(s, rd),
451                          vec_full_reg_offset(s, rn), vsz, vsz);
452     }
453     return true;
454 }
455 
456 /* Initialize a Zreg with replications of a 64-bit immediate.  */
457 static void do_dupi_z(DisasContext *s, int rd, uint64_t word)
458 {
459     unsigned vsz = vec_full_reg_size(s);
460     tcg_gen_gvec_dup_imm(MO_64, vec_full_reg_offset(s, rd), vsz, vsz, word);
461 }
462 
463 /* Invoke a vector expander on three Pregs.  */
464 static bool gen_gvec_fn_ppp(DisasContext *s, GVecGen3Fn *gvec_fn,
465                             int rd, int rn, int rm)
466 {
467     if (sve_access_check(s)) {
468         unsigned psz = pred_gvec_reg_size(s);
469         gvec_fn(MO_64, pred_full_reg_offset(s, rd),
470                 pred_full_reg_offset(s, rn),
471                 pred_full_reg_offset(s, rm), psz, psz);
472     }
473     return true;
474 }
475 
476 /* Invoke a vector move on two Pregs.  */
477 static bool do_mov_p(DisasContext *s, int rd, int rn)
478 {
479     if (sve_access_check(s)) {
480         unsigned psz = pred_gvec_reg_size(s);
481         tcg_gen_gvec_mov(MO_8, pred_full_reg_offset(s, rd),
482                          pred_full_reg_offset(s, rn), psz, psz);
483     }
484     return true;
485 }
486 
487 /* Set the cpu flags as per a return from an SVE helper.  */
488 static void do_pred_flags(TCGv_i32 t)
489 {
490     tcg_gen_mov_i32(cpu_NF, t);
491     tcg_gen_andi_i32(cpu_ZF, t, 2);
492     tcg_gen_andi_i32(cpu_CF, t, 1);
493     tcg_gen_movi_i32(cpu_VF, 0);
494 }
495 
496 /* Subroutines computing the ARM PredTest psuedofunction.  */
497 static void do_predtest1(TCGv_i64 d, TCGv_i64 g)
498 {
499     TCGv_i32 t = tcg_temp_new_i32();
500 
501     gen_helper_sve_predtest1(t, d, g);
502     do_pred_flags(t);
503 }
504 
505 static void do_predtest(DisasContext *s, int dofs, int gofs, int words)
506 {
507     TCGv_ptr dptr = tcg_temp_new_ptr();
508     TCGv_ptr gptr = tcg_temp_new_ptr();
509     TCGv_i32 t = tcg_temp_new_i32();
510 
511     tcg_gen_addi_ptr(dptr, cpu_env, dofs);
512     tcg_gen_addi_ptr(gptr, cpu_env, gofs);
513 
514     gen_helper_sve_predtest(t, dptr, gptr, tcg_constant_i32(words));
515 
516     do_pred_flags(t);
517 }
518 
519 /* For each element size, the bits within a predicate word that are active.  */
520 const uint64_t pred_esz_masks[5] = {
521     0xffffffffffffffffull, 0x5555555555555555ull,
522     0x1111111111111111ull, 0x0101010101010101ull,
523     0x0001000100010001ull,
524 };
525 
526 static bool trans_INVALID(DisasContext *s, arg_INVALID *a)
527 {
528     unallocated_encoding(s);
529     return true;
530 }
531 
532 /*
533  *** SVE Logical - Unpredicated Group
534  */
535 
536 TRANS_FEAT(AND_zzz, aa64_sve, gen_gvec_fn_arg_zzz, tcg_gen_gvec_and, a)
537 TRANS_FEAT(ORR_zzz, aa64_sve, gen_gvec_fn_arg_zzz, tcg_gen_gvec_or, a)
538 TRANS_FEAT(EOR_zzz, aa64_sve, gen_gvec_fn_arg_zzz, tcg_gen_gvec_xor, a)
539 TRANS_FEAT(BIC_zzz, aa64_sve, gen_gvec_fn_arg_zzz, tcg_gen_gvec_andc, a)
540 
541 static void gen_xar8_i64(TCGv_i64 d, TCGv_i64 n, TCGv_i64 m, int64_t sh)
542 {
543     TCGv_i64 t = tcg_temp_new_i64();
544     uint64_t mask = dup_const(MO_8, 0xff >> sh);
545 
546     tcg_gen_xor_i64(t, n, m);
547     tcg_gen_shri_i64(d, t, sh);
548     tcg_gen_shli_i64(t, t, 8 - sh);
549     tcg_gen_andi_i64(d, d, mask);
550     tcg_gen_andi_i64(t, t, ~mask);
551     tcg_gen_or_i64(d, d, t);
552 }
553 
554 static void gen_xar16_i64(TCGv_i64 d, TCGv_i64 n, TCGv_i64 m, int64_t sh)
555 {
556     TCGv_i64 t = tcg_temp_new_i64();
557     uint64_t mask = dup_const(MO_16, 0xffff >> sh);
558 
559     tcg_gen_xor_i64(t, n, m);
560     tcg_gen_shri_i64(d, t, sh);
561     tcg_gen_shli_i64(t, t, 16 - sh);
562     tcg_gen_andi_i64(d, d, mask);
563     tcg_gen_andi_i64(t, t, ~mask);
564     tcg_gen_or_i64(d, d, t);
565 }
566 
567 static void gen_xar_i32(TCGv_i32 d, TCGv_i32 n, TCGv_i32 m, int32_t sh)
568 {
569     tcg_gen_xor_i32(d, n, m);
570     tcg_gen_rotri_i32(d, d, sh);
571 }
572 
573 static void gen_xar_i64(TCGv_i64 d, TCGv_i64 n, TCGv_i64 m, int64_t sh)
574 {
575     tcg_gen_xor_i64(d, n, m);
576     tcg_gen_rotri_i64(d, d, sh);
577 }
578 
579 static void gen_xar_vec(unsigned vece, TCGv_vec d, TCGv_vec n,
580                         TCGv_vec m, int64_t sh)
581 {
582     tcg_gen_xor_vec(vece, d, n, m);
583     tcg_gen_rotri_vec(vece, d, d, sh);
584 }
585 
586 void gen_gvec_xar(unsigned vece, uint32_t rd_ofs, uint32_t rn_ofs,
587                   uint32_t rm_ofs, int64_t shift,
588                   uint32_t opr_sz, uint32_t max_sz)
589 {
590     static const TCGOpcode vecop[] = { INDEX_op_rotli_vec, 0 };
591     static const GVecGen3i ops[4] = {
592         { .fni8 = gen_xar8_i64,
593           .fniv = gen_xar_vec,
594           .fno = gen_helper_sve2_xar_b,
595           .opt_opc = vecop,
596           .vece = MO_8 },
597         { .fni8 = gen_xar16_i64,
598           .fniv = gen_xar_vec,
599           .fno = gen_helper_sve2_xar_h,
600           .opt_opc = vecop,
601           .vece = MO_16 },
602         { .fni4 = gen_xar_i32,
603           .fniv = gen_xar_vec,
604           .fno = gen_helper_sve2_xar_s,
605           .opt_opc = vecop,
606           .vece = MO_32 },
607         { .fni8 = gen_xar_i64,
608           .fniv = gen_xar_vec,
609           .fno = gen_helper_gvec_xar_d,
610           .opt_opc = vecop,
611           .vece = MO_64 }
612     };
613     int esize = 8 << vece;
614 
615     /* The SVE2 range is 1 .. esize; the AdvSIMD range is 0 .. esize-1. */
616     tcg_debug_assert(shift >= 0);
617     tcg_debug_assert(shift <= esize);
618     shift &= esize - 1;
619 
620     if (shift == 0) {
621         /* xar with no rotate devolves to xor. */
622         tcg_gen_gvec_xor(vece, rd_ofs, rn_ofs, rm_ofs, opr_sz, max_sz);
623     } else {
624         tcg_gen_gvec_3i(rd_ofs, rn_ofs, rm_ofs, opr_sz, max_sz,
625                         shift, &ops[vece]);
626     }
627 }
628 
629 static bool trans_XAR(DisasContext *s, arg_rrri_esz *a)
630 {
631     if (a->esz < 0 || !dc_isar_feature(aa64_sve2, s)) {
632         return false;
633     }
634     if (sve_access_check(s)) {
635         unsigned vsz = vec_full_reg_size(s);
636         gen_gvec_xar(a->esz, vec_full_reg_offset(s, a->rd),
637                      vec_full_reg_offset(s, a->rn),
638                      vec_full_reg_offset(s, a->rm), a->imm, vsz, vsz);
639     }
640     return true;
641 }
642 
643 static void gen_eor3_i64(TCGv_i64 d, TCGv_i64 n, TCGv_i64 m, TCGv_i64 k)
644 {
645     tcg_gen_xor_i64(d, n, m);
646     tcg_gen_xor_i64(d, d, k);
647 }
648 
649 static void gen_eor3_vec(unsigned vece, TCGv_vec d, TCGv_vec n,
650                          TCGv_vec m, TCGv_vec k)
651 {
652     tcg_gen_xor_vec(vece, d, n, m);
653     tcg_gen_xor_vec(vece, d, d, k);
654 }
655 
656 static void gen_eor3(unsigned vece, uint32_t d, uint32_t n, uint32_t m,
657                      uint32_t a, uint32_t oprsz, uint32_t maxsz)
658 {
659     static const GVecGen4 op = {
660         .fni8 = gen_eor3_i64,
661         .fniv = gen_eor3_vec,
662         .fno = gen_helper_sve2_eor3,
663         .vece = MO_64,
664         .prefer_i64 = TCG_TARGET_REG_BITS == 64,
665     };
666     tcg_gen_gvec_4(d, n, m, a, oprsz, maxsz, &op);
667 }
668 
669 TRANS_FEAT(EOR3, aa64_sve2, gen_gvec_fn_arg_zzzz, gen_eor3, a)
670 
671 static void gen_bcax_i64(TCGv_i64 d, TCGv_i64 n, TCGv_i64 m, TCGv_i64 k)
672 {
673     tcg_gen_andc_i64(d, m, k);
674     tcg_gen_xor_i64(d, d, n);
675 }
676 
677 static void gen_bcax_vec(unsigned vece, TCGv_vec d, TCGv_vec n,
678                          TCGv_vec m, TCGv_vec k)
679 {
680     tcg_gen_andc_vec(vece, d, m, k);
681     tcg_gen_xor_vec(vece, d, d, n);
682 }
683 
684 static void gen_bcax(unsigned vece, uint32_t d, uint32_t n, uint32_t m,
685                      uint32_t a, uint32_t oprsz, uint32_t maxsz)
686 {
687     static const GVecGen4 op = {
688         .fni8 = gen_bcax_i64,
689         .fniv = gen_bcax_vec,
690         .fno = gen_helper_sve2_bcax,
691         .vece = MO_64,
692         .prefer_i64 = TCG_TARGET_REG_BITS == 64,
693     };
694     tcg_gen_gvec_4(d, n, m, a, oprsz, maxsz, &op);
695 }
696 
697 TRANS_FEAT(BCAX, aa64_sve2, gen_gvec_fn_arg_zzzz, gen_bcax, a)
698 
699 static void gen_bsl(unsigned vece, uint32_t d, uint32_t n, uint32_t m,
700                     uint32_t a, uint32_t oprsz, uint32_t maxsz)
701 {
702     /* BSL differs from the generic bitsel in argument ordering. */
703     tcg_gen_gvec_bitsel(vece, d, a, n, m, oprsz, maxsz);
704 }
705 
706 TRANS_FEAT(BSL, aa64_sve2, gen_gvec_fn_arg_zzzz, gen_bsl, a)
707 
708 static void gen_bsl1n_i64(TCGv_i64 d, TCGv_i64 n, TCGv_i64 m, TCGv_i64 k)
709 {
710     tcg_gen_andc_i64(n, k, n);
711     tcg_gen_andc_i64(m, m, k);
712     tcg_gen_or_i64(d, n, m);
713 }
714 
715 static void gen_bsl1n_vec(unsigned vece, TCGv_vec d, TCGv_vec n,
716                           TCGv_vec m, TCGv_vec k)
717 {
718     if (TCG_TARGET_HAS_bitsel_vec) {
719         tcg_gen_not_vec(vece, n, n);
720         tcg_gen_bitsel_vec(vece, d, k, n, m);
721     } else {
722         tcg_gen_andc_vec(vece, n, k, n);
723         tcg_gen_andc_vec(vece, m, m, k);
724         tcg_gen_or_vec(vece, d, n, m);
725     }
726 }
727 
728 static void gen_bsl1n(unsigned vece, uint32_t d, uint32_t n, uint32_t m,
729                       uint32_t a, uint32_t oprsz, uint32_t maxsz)
730 {
731     static const GVecGen4 op = {
732         .fni8 = gen_bsl1n_i64,
733         .fniv = gen_bsl1n_vec,
734         .fno = gen_helper_sve2_bsl1n,
735         .vece = MO_64,
736         .prefer_i64 = TCG_TARGET_REG_BITS == 64,
737     };
738     tcg_gen_gvec_4(d, n, m, a, oprsz, maxsz, &op);
739 }
740 
741 TRANS_FEAT(BSL1N, aa64_sve2, gen_gvec_fn_arg_zzzz, gen_bsl1n, a)
742 
743 static void gen_bsl2n_i64(TCGv_i64 d, TCGv_i64 n, TCGv_i64 m, TCGv_i64 k)
744 {
745     /*
746      * Z[dn] = (n & k) | (~m & ~k)
747      *       =         | ~(m | k)
748      */
749     tcg_gen_and_i64(n, n, k);
750     if (TCG_TARGET_HAS_orc_i64) {
751         tcg_gen_or_i64(m, m, k);
752         tcg_gen_orc_i64(d, n, m);
753     } else {
754         tcg_gen_nor_i64(m, m, k);
755         tcg_gen_or_i64(d, n, m);
756     }
757 }
758 
759 static void gen_bsl2n_vec(unsigned vece, TCGv_vec d, TCGv_vec n,
760                           TCGv_vec m, TCGv_vec k)
761 {
762     if (TCG_TARGET_HAS_bitsel_vec) {
763         tcg_gen_not_vec(vece, m, m);
764         tcg_gen_bitsel_vec(vece, d, k, n, m);
765     } else {
766         tcg_gen_and_vec(vece, n, n, k);
767         tcg_gen_or_vec(vece, m, m, k);
768         tcg_gen_orc_vec(vece, d, n, m);
769     }
770 }
771 
772 static void gen_bsl2n(unsigned vece, uint32_t d, uint32_t n, uint32_t m,
773                       uint32_t a, uint32_t oprsz, uint32_t maxsz)
774 {
775     static const GVecGen4 op = {
776         .fni8 = gen_bsl2n_i64,
777         .fniv = gen_bsl2n_vec,
778         .fno = gen_helper_sve2_bsl2n,
779         .vece = MO_64,
780         .prefer_i64 = TCG_TARGET_REG_BITS == 64,
781     };
782     tcg_gen_gvec_4(d, n, m, a, oprsz, maxsz, &op);
783 }
784 
785 TRANS_FEAT(BSL2N, aa64_sve2, gen_gvec_fn_arg_zzzz, gen_bsl2n, a)
786 
787 static void gen_nbsl_i64(TCGv_i64 d, TCGv_i64 n, TCGv_i64 m, TCGv_i64 k)
788 {
789     tcg_gen_and_i64(n, n, k);
790     tcg_gen_andc_i64(m, m, k);
791     tcg_gen_nor_i64(d, n, m);
792 }
793 
794 static void gen_nbsl_vec(unsigned vece, TCGv_vec d, TCGv_vec n,
795                           TCGv_vec m, TCGv_vec k)
796 {
797     tcg_gen_bitsel_vec(vece, d, k, n, m);
798     tcg_gen_not_vec(vece, d, d);
799 }
800 
801 static void gen_nbsl(unsigned vece, uint32_t d, uint32_t n, uint32_t m,
802                      uint32_t a, uint32_t oprsz, uint32_t maxsz)
803 {
804     static const GVecGen4 op = {
805         .fni8 = gen_nbsl_i64,
806         .fniv = gen_nbsl_vec,
807         .fno = gen_helper_sve2_nbsl,
808         .vece = MO_64,
809         .prefer_i64 = TCG_TARGET_REG_BITS == 64,
810     };
811     tcg_gen_gvec_4(d, n, m, a, oprsz, maxsz, &op);
812 }
813 
814 TRANS_FEAT(NBSL, aa64_sve2, gen_gvec_fn_arg_zzzz, gen_nbsl, a)
815 
816 /*
817  *** SVE Integer Arithmetic - Unpredicated Group
818  */
819 
820 TRANS_FEAT(ADD_zzz, aa64_sve, gen_gvec_fn_arg_zzz, tcg_gen_gvec_add, a)
821 TRANS_FEAT(SUB_zzz, aa64_sve, gen_gvec_fn_arg_zzz, tcg_gen_gvec_sub, a)
822 TRANS_FEAT(SQADD_zzz, aa64_sve, gen_gvec_fn_arg_zzz, tcg_gen_gvec_ssadd, a)
823 TRANS_FEAT(SQSUB_zzz, aa64_sve, gen_gvec_fn_arg_zzz, tcg_gen_gvec_sssub, a)
824 TRANS_FEAT(UQADD_zzz, aa64_sve, gen_gvec_fn_arg_zzz, tcg_gen_gvec_usadd, a)
825 TRANS_FEAT(UQSUB_zzz, aa64_sve, gen_gvec_fn_arg_zzz, tcg_gen_gvec_ussub, a)
826 
827 /*
828  *** SVE Integer Arithmetic - Binary Predicated Group
829  */
830 
831 /* Select active elememnts from Zn and inactive elements from Zm,
832  * storing the result in Zd.
833  */
834 static bool do_sel_z(DisasContext *s, int rd, int rn, int rm, int pg, int esz)
835 {
836     static gen_helper_gvec_4 * const fns[4] = {
837         gen_helper_sve_sel_zpzz_b, gen_helper_sve_sel_zpzz_h,
838         gen_helper_sve_sel_zpzz_s, gen_helper_sve_sel_zpzz_d
839     };
840     return gen_gvec_ool_zzzp(s, fns[esz], rd, rn, rm, pg, 0);
841 }
842 
843 #define DO_ZPZZ(NAME, FEAT, name) \
844     static gen_helper_gvec_4 * const name##_zpzz_fns[4] = {               \
845         gen_helper_##name##_zpzz_b, gen_helper_##name##_zpzz_h,           \
846         gen_helper_##name##_zpzz_s, gen_helper_##name##_zpzz_d,           \
847     };                                                                    \
848     TRANS_FEAT(NAME, FEAT, gen_gvec_ool_arg_zpzz,                         \
849                name##_zpzz_fns[a->esz], a, 0)
850 
851 DO_ZPZZ(AND_zpzz, aa64_sve, sve_and)
852 DO_ZPZZ(EOR_zpzz, aa64_sve, sve_eor)
853 DO_ZPZZ(ORR_zpzz, aa64_sve, sve_orr)
854 DO_ZPZZ(BIC_zpzz, aa64_sve, sve_bic)
855 
856 DO_ZPZZ(ADD_zpzz, aa64_sve, sve_add)
857 DO_ZPZZ(SUB_zpzz, aa64_sve, sve_sub)
858 
859 DO_ZPZZ(SMAX_zpzz, aa64_sve, sve_smax)
860 DO_ZPZZ(UMAX_zpzz, aa64_sve, sve_umax)
861 DO_ZPZZ(SMIN_zpzz, aa64_sve, sve_smin)
862 DO_ZPZZ(UMIN_zpzz, aa64_sve, sve_umin)
863 DO_ZPZZ(SABD_zpzz, aa64_sve, sve_sabd)
864 DO_ZPZZ(UABD_zpzz, aa64_sve, sve_uabd)
865 
866 DO_ZPZZ(MUL_zpzz, aa64_sve, sve_mul)
867 DO_ZPZZ(SMULH_zpzz, aa64_sve, sve_smulh)
868 DO_ZPZZ(UMULH_zpzz, aa64_sve, sve_umulh)
869 
870 DO_ZPZZ(ASR_zpzz, aa64_sve, sve_asr)
871 DO_ZPZZ(LSR_zpzz, aa64_sve, sve_lsr)
872 DO_ZPZZ(LSL_zpzz, aa64_sve, sve_lsl)
873 
874 static gen_helper_gvec_4 * const sdiv_fns[4] = {
875     NULL, NULL, gen_helper_sve_sdiv_zpzz_s, gen_helper_sve_sdiv_zpzz_d
876 };
877 TRANS_FEAT(SDIV_zpzz, aa64_sve, gen_gvec_ool_arg_zpzz, sdiv_fns[a->esz], a, 0)
878 
879 static gen_helper_gvec_4 * const udiv_fns[4] = {
880     NULL, NULL, gen_helper_sve_udiv_zpzz_s, gen_helper_sve_udiv_zpzz_d
881 };
882 TRANS_FEAT(UDIV_zpzz, aa64_sve, gen_gvec_ool_arg_zpzz, udiv_fns[a->esz], a, 0)
883 
884 TRANS_FEAT(SEL_zpzz, aa64_sve, do_sel_z, a->rd, a->rn, a->rm, a->pg, a->esz)
885 
886 /*
887  *** SVE Integer Arithmetic - Unary Predicated Group
888  */
889 
890 #define DO_ZPZ(NAME, FEAT, name) \
891     static gen_helper_gvec_3 * const name##_fns[4] = {              \
892         gen_helper_##name##_b, gen_helper_##name##_h,               \
893         gen_helper_##name##_s, gen_helper_##name##_d,               \
894     };                                                              \
895     TRANS_FEAT(NAME, FEAT, gen_gvec_ool_arg_zpz, name##_fns[a->esz], a, 0)
896 
897 DO_ZPZ(CLS, aa64_sve, sve_cls)
898 DO_ZPZ(CLZ, aa64_sve, sve_clz)
899 DO_ZPZ(CNT_zpz, aa64_sve, sve_cnt_zpz)
900 DO_ZPZ(CNOT, aa64_sve, sve_cnot)
901 DO_ZPZ(NOT_zpz, aa64_sve, sve_not_zpz)
902 DO_ZPZ(ABS, aa64_sve, sve_abs)
903 DO_ZPZ(NEG, aa64_sve, sve_neg)
904 DO_ZPZ(RBIT, aa64_sve, sve_rbit)
905 
906 static gen_helper_gvec_3 * const fabs_fns[4] = {
907     NULL,                  gen_helper_sve_fabs_h,
908     gen_helper_sve_fabs_s, gen_helper_sve_fabs_d,
909 };
910 TRANS_FEAT(FABS, aa64_sve, gen_gvec_ool_arg_zpz, fabs_fns[a->esz], a, 0)
911 
912 static gen_helper_gvec_3 * const fneg_fns[4] = {
913     NULL,                  gen_helper_sve_fneg_h,
914     gen_helper_sve_fneg_s, gen_helper_sve_fneg_d,
915 };
916 TRANS_FEAT(FNEG, aa64_sve, gen_gvec_ool_arg_zpz, fneg_fns[a->esz], a, 0)
917 
918 static gen_helper_gvec_3 * const sxtb_fns[4] = {
919     NULL,                  gen_helper_sve_sxtb_h,
920     gen_helper_sve_sxtb_s, gen_helper_sve_sxtb_d,
921 };
922 TRANS_FEAT(SXTB, aa64_sve, gen_gvec_ool_arg_zpz, sxtb_fns[a->esz], a, 0)
923 
924 static gen_helper_gvec_3 * const uxtb_fns[4] = {
925     NULL,                  gen_helper_sve_uxtb_h,
926     gen_helper_sve_uxtb_s, gen_helper_sve_uxtb_d,
927 };
928 TRANS_FEAT(UXTB, aa64_sve, gen_gvec_ool_arg_zpz, uxtb_fns[a->esz], a, 0)
929 
930 static gen_helper_gvec_3 * const sxth_fns[4] = {
931     NULL, NULL, gen_helper_sve_sxth_s, gen_helper_sve_sxth_d
932 };
933 TRANS_FEAT(SXTH, aa64_sve, gen_gvec_ool_arg_zpz, sxth_fns[a->esz], a, 0)
934 
935 static gen_helper_gvec_3 * const uxth_fns[4] = {
936     NULL, NULL, gen_helper_sve_uxth_s, gen_helper_sve_uxth_d
937 };
938 TRANS_FEAT(UXTH, aa64_sve, gen_gvec_ool_arg_zpz, uxth_fns[a->esz], a, 0)
939 
940 TRANS_FEAT(SXTW, aa64_sve, gen_gvec_ool_arg_zpz,
941            a->esz == 3 ? gen_helper_sve_sxtw_d : NULL, a, 0)
942 TRANS_FEAT(UXTW, aa64_sve, gen_gvec_ool_arg_zpz,
943            a->esz == 3 ? gen_helper_sve_uxtw_d : NULL, a, 0)
944 
945 /*
946  *** SVE Integer Reduction Group
947  */
948 
949 typedef void gen_helper_gvec_reduc(TCGv_i64, TCGv_ptr, TCGv_ptr, TCGv_i32);
950 static bool do_vpz_ool(DisasContext *s, arg_rpr_esz *a,
951                        gen_helper_gvec_reduc *fn)
952 {
953     unsigned vsz = vec_full_reg_size(s);
954     TCGv_ptr t_zn, t_pg;
955     TCGv_i32 desc;
956     TCGv_i64 temp;
957 
958     if (fn == NULL) {
959         return false;
960     }
961     if (!sve_access_check(s)) {
962         return true;
963     }
964 
965     desc = tcg_constant_i32(simd_desc(vsz, vsz, 0));
966     temp = tcg_temp_new_i64();
967     t_zn = tcg_temp_new_ptr();
968     t_pg = tcg_temp_new_ptr();
969 
970     tcg_gen_addi_ptr(t_zn, cpu_env, vec_full_reg_offset(s, a->rn));
971     tcg_gen_addi_ptr(t_pg, cpu_env, pred_full_reg_offset(s, a->pg));
972     fn(temp, t_zn, t_pg, desc);
973 
974     write_fp_dreg(s, a->rd, temp);
975     return true;
976 }
977 
978 #define DO_VPZ(NAME, name) \
979     static gen_helper_gvec_reduc * const name##_fns[4] = {               \
980         gen_helper_sve_##name##_b, gen_helper_sve_##name##_h,            \
981         gen_helper_sve_##name##_s, gen_helper_sve_##name##_d,            \
982     };                                                                   \
983     TRANS_FEAT(NAME, aa64_sve, do_vpz_ool, a, name##_fns[a->esz])
984 
985 DO_VPZ(ORV, orv)
986 DO_VPZ(ANDV, andv)
987 DO_VPZ(EORV, eorv)
988 
989 DO_VPZ(UADDV, uaddv)
990 DO_VPZ(SMAXV, smaxv)
991 DO_VPZ(UMAXV, umaxv)
992 DO_VPZ(SMINV, sminv)
993 DO_VPZ(UMINV, uminv)
994 
995 static gen_helper_gvec_reduc * const saddv_fns[4] = {
996     gen_helper_sve_saddv_b, gen_helper_sve_saddv_h,
997     gen_helper_sve_saddv_s, NULL
998 };
999 TRANS_FEAT(SADDV, aa64_sve, do_vpz_ool, a, saddv_fns[a->esz])
1000 
1001 #undef DO_VPZ
1002 
1003 /*
1004  *** SVE Shift by Immediate - Predicated Group
1005  */
1006 
1007 /*
1008  * Copy Zn into Zd, storing zeros into inactive elements.
1009  * If invert, store zeros into the active elements.
1010  */
1011 static bool do_movz_zpz(DisasContext *s, int rd, int rn, int pg,
1012                         int esz, bool invert)
1013 {
1014     static gen_helper_gvec_3 * const fns[4] = {
1015         gen_helper_sve_movz_b, gen_helper_sve_movz_h,
1016         gen_helper_sve_movz_s, gen_helper_sve_movz_d,
1017     };
1018     return gen_gvec_ool_zzp(s, fns[esz], rd, rn, pg, invert);
1019 }
1020 
1021 static bool do_shift_zpzi(DisasContext *s, arg_rpri_esz *a, bool asr,
1022                           gen_helper_gvec_3 * const fns[4])
1023 {
1024     int max;
1025 
1026     if (a->esz < 0) {
1027         /* Invalid tsz encoding -- see tszimm_esz. */
1028         return false;
1029     }
1030 
1031     /*
1032      * Shift by element size is architecturally valid.
1033      * For arithmetic right-shift, it's the same as by one less.
1034      * For logical shifts and ASRD, it is a zeroing operation.
1035      */
1036     max = 8 << a->esz;
1037     if (a->imm >= max) {
1038         if (asr) {
1039             a->imm = max - 1;
1040         } else {
1041             return do_movz_zpz(s, a->rd, a->rd, a->pg, a->esz, true);
1042         }
1043     }
1044     return gen_gvec_ool_arg_zpzi(s, fns[a->esz], a);
1045 }
1046 
1047 static gen_helper_gvec_3 * const asr_zpzi_fns[4] = {
1048     gen_helper_sve_asr_zpzi_b, gen_helper_sve_asr_zpzi_h,
1049     gen_helper_sve_asr_zpzi_s, gen_helper_sve_asr_zpzi_d,
1050 };
1051 TRANS_FEAT(ASR_zpzi, aa64_sve, do_shift_zpzi, a, true, asr_zpzi_fns)
1052 
1053 static gen_helper_gvec_3 * const lsr_zpzi_fns[4] = {
1054     gen_helper_sve_lsr_zpzi_b, gen_helper_sve_lsr_zpzi_h,
1055     gen_helper_sve_lsr_zpzi_s, gen_helper_sve_lsr_zpzi_d,
1056 };
1057 TRANS_FEAT(LSR_zpzi, aa64_sve, do_shift_zpzi, a, false, lsr_zpzi_fns)
1058 
1059 static gen_helper_gvec_3 * const lsl_zpzi_fns[4] = {
1060     gen_helper_sve_lsl_zpzi_b, gen_helper_sve_lsl_zpzi_h,
1061     gen_helper_sve_lsl_zpzi_s, gen_helper_sve_lsl_zpzi_d,
1062 };
1063 TRANS_FEAT(LSL_zpzi, aa64_sve, do_shift_zpzi, a, false, lsl_zpzi_fns)
1064 
1065 static gen_helper_gvec_3 * const asrd_fns[4] = {
1066     gen_helper_sve_asrd_b, gen_helper_sve_asrd_h,
1067     gen_helper_sve_asrd_s, gen_helper_sve_asrd_d,
1068 };
1069 TRANS_FEAT(ASRD, aa64_sve, do_shift_zpzi, a, false, asrd_fns)
1070 
1071 static gen_helper_gvec_3 * const sqshl_zpzi_fns[4] = {
1072     gen_helper_sve2_sqshl_zpzi_b, gen_helper_sve2_sqshl_zpzi_h,
1073     gen_helper_sve2_sqshl_zpzi_s, gen_helper_sve2_sqshl_zpzi_d,
1074 };
1075 TRANS_FEAT(SQSHL_zpzi, aa64_sve2, gen_gvec_ool_arg_zpzi,
1076            a->esz < 0 ? NULL : sqshl_zpzi_fns[a->esz], a)
1077 
1078 static gen_helper_gvec_3 * const uqshl_zpzi_fns[4] = {
1079     gen_helper_sve2_uqshl_zpzi_b, gen_helper_sve2_uqshl_zpzi_h,
1080     gen_helper_sve2_uqshl_zpzi_s, gen_helper_sve2_uqshl_zpzi_d,
1081 };
1082 TRANS_FEAT(UQSHL_zpzi, aa64_sve2, gen_gvec_ool_arg_zpzi,
1083            a->esz < 0 ? NULL : uqshl_zpzi_fns[a->esz], a)
1084 
1085 static gen_helper_gvec_3 * const srshr_fns[4] = {
1086     gen_helper_sve2_srshr_b, gen_helper_sve2_srshr_h,
1087     gen_helper_sve2_srshr_s, gen_helper_sve2_srshr_d,
1088 };
1089 TRANS_FEAT(SRSHR, aa64_sve2, gen_gvec_ool_arg_zpzi,
1090            a->esz < 0 ? NULL : srshr_fns[a->esz], a)
1091 
1092 static gen_helper_gvec_3 * const urshr_fns[4] = {
1093     gen_helper_sve2_urshr_b, gen_helper_sve2_urshr_h,
1094     gen_helper_sve2_urshr_s, gen_helper_sve2_urshr_d,
1095 };
1096 TRANS_FEAT(URSHR, aa64_sve2, gen_gvec_ool_arg_zpzi,
1097            a->esz < 0 ? NULL : urshr_fns[a->esz], a)
1098 
1099 static gen_helper_gvec_3 * const sqshlu_fns[4] = {
1100     gen_helper_sve2_sqshlu_b, gen_helper_sve2_sqshlu_h,
1101     gen_helper_sve2_sqshlu_s, gen_helper_sve2_sqshlu_d,
1102 };
1103 TRANS_FEAT(SQSHLU, aa64_sve2, gen_gvec_ool_arg_zpzi,
1104            a->esz < 0 ? NULL : sqshlu_fns[a->esz], a)
1105 
1106 /*
1107  *** SVE Bitwise Shift - Predicated Group
1108  */
1109 
1110 #define DO_ZPZW(NAME, name) \
1111     static gen_helper_gvec_4 * const name##_zpzw_fns[4] = {               \
1112         gen_helper_sve_##name##_zpzw_b, gen_helper_sve_##name##_zpzw_h,   \
1113         gen_helper_sve_##name##_zpzw_s, NULL                              \
1114     };                                                                    \
1115     TRANS_FEAT(NAME##_zpzw, aa64_sve, gen_gvec_ool_arg_zpzz,              \
1116                a->esz < 0 ? NULL : name##_zpzw_fns[a->esz], a, 0)
1117 
1118 DO_ZPZW(ASR, asr)
1119 DO_ZPZW(LSR, lsr)
1120 DO_ZPZW(LSL, lsl)
1121 
1122 #undef DO_ZPZW
1123 
1124 /*
1125  *** SVE Bitwise Shift - Unpredicated Group
1126  */
1127 
1128 static bool do_shift_imm(DisasContext *s, arg_rri_esz *a, bool asr,
1129                          void (*gvec_fn)(unsigned, uint32_t, uint32_t,
1130                                          int64_t, uint32_t, uint32_t))
1131 {
1132     if (a->esz < 0) {
1133         /* Invalid tsz encoding -- see tszimm_esz. */
1134         return false;
1135     }
1136     if (sve_access_check(s)) {
1137         unsigned vsz = vec_full_reg_size(s);
1138         /* Shift by element size is architecturally valid.  For
1139            arithmetic right-shift, it's the same as by one less.
1140            Otherwise it is a zeroing operation.  */
1141         if (a->imm >= 8 << a->esz) {
1142             if (asr) {
1143                 a->imm = (8 << a->esz) - 1;
1144             } else {
1145                 do_dupi_z(s, a->rd, 0);
1146                 return true;
1147             }
1148         }
1149         gvec_fn(a->esz, vec_full_reg_offset(s, a->rd),
1150                 vec_full_reg_offset(s, a->rn), a->imm, vsz, vsz);
1151     }
1152     return true;
1153 }
1154 
1155 TRANS_FEAT(ASR_zzi, aa64_sve, do_shift_imm, a, true, tcg_gen_gvec_sari)
1156 TRANS_FEAT(LSR_zzi, aa64_sve, do_shift_imm, a, false, tcg_gen_gvec_shri)
1157 TRANS_FEAT(LSL_zzi, aa64_sve, do_shift_imm, a, false, tcg_gen_gvec_shli)
1158 
1159 #define DO_ZZW(NAME, name) \
1160     static gen_helper_gvec_3 * const name##_zzw_fns[4] = {                \
1161         gen_helper_sve_##name##_zzw_b, gen_helper_sve_##name##_zzw_h,     \
1162         gen_helper_sve_##name##_zzw_s, NULL                               \
1163     };                                                                    \
1164     TRANS_FEAT(NAME, aa64_sve, gen_gvec_ool_arg_zzz,                      \
1165                name##_zzw_fns[a->esz], a, 0)
1166 
1167 DO_ZZW(ASR_zzw, asr)
1168 DO_ZZW(LSR_zzw, lsr)
1169 DO_ZZW(LSL_zzw, lsl)
1170 
1171 #undef DO_ZZW
1172 
1173 /*
1174  *** SVE Integer Multiply-Add Group
1175  */
1176 
1177 static bool do_zpzzz_ool(DisasContext *s, arg_rprrr_esz *a,
1178                          gen_helper_gvec_5 *fn)
1179 {
1180     if (sve_access_check(s)) {
1181         unsigned vsz = vec_full_reg_size(s);
1182         tcg_gen_gvec_5_ool(vec_full_reg_offset(s, a->rd),
1183                            vec_full_reg_offset(s, a->ra),
1184                            vec_full_reg_offset(s, a->rn),
1185                            vec_full_reg_offset(s, a->rm),
1186                            pred_full_reg_offset(s, a->pg),
1187                            vsz, vsz, 0, fn);
1188     }
1189     return true;
1190 }
1191 
1192 static gen_helper_gvec_5 * const mla_fns[4] = {
1193     gen_helper_sve_mla_b, gen_helper_sve_mla_h,
1194     gen_helper_sve_mla_s, gen_helper_sve_mla_d,
1195 };
1196 TRANS_FEAT(MLA, aa64_sve, do_zpzzz_ool, a, mla_fns[a->esz])
1197 
1198 static gen_helper_gvec_5 * const mls_fns[4] = {
1199     gen_helper_sve_mls_b, gen_helper_sve_mls_h,
1200     gen_helper_sve_mls_s, gen_helper_sve_mls_d,
1201 };
1202 TRANS_FEAT(MLS, aa64_sve, do_zpzzz_ool, a, mls_fns[a->esz])
1203 
1204 /*
1205  *** SVE Index Generation Group
1206  */
1207 
1208 static bool do_index(DisasContext *s, int esz, int rd,
1209                      TCGv_i64 start, TCGv_i64 incr)
1210 {
1211     unsigned vsz;
1212     TCGv_i32 desc;
1213     TCGv_ptr t_zd;
1214 
1215     if (!sve_access_check(s)) {
1216         return true;
1217     }
1218 
1219     vsz = vec_full_reg_size(s);
1220     desc = tcg_constant_i32(simd_desc(vsz, vsz, 0));
1221     t_zd = tcg_temp_new_ptr();
1222 
1223     tcg_gen_addi_ptr(t_zd, cpu_env, vec_full_reg_offset(s, rd));
1224     if (esz == 3) {
1225         gen_helper_sve_index_d(t_zd, start, incr, desc);
1226     } else {
1227         typedef void index_fn(TCGv_ptr, TCGv_i32, TCGv_i32, TCGv_i32);
1228         static index_fn * const fns[3] = {
1229             gen_helper_sve_index_b,
1230             gen_helper_sve_index_h,
1231             gen_helper_sve_index_s,
1232         };
1233         TCGv_i32 s32 = tcg_temp_new_i32();
1234         TCGv_i32 i32 = tcg_temp_new_i32();
1235 
1236         tcg_gen_extrl_i64_i32(s32, start);
1237         tcg_gen_extrl_i64_i32(i32, incr);
1238         fns[esz](t_zd, s32, i32, desc);
1239     }
1240     return true;
1241 }
1242 
1243 TRANS_FEAT(INDEX_ii, aa64_sve, do_index, a->esz, a->rd,
1244            tcg_constant_i64(a->imm1), tcg_constant_i64(a->imm2))
1245 TRANS_FEAT(INDEX_ir, aa64_sve, do_index, a->esz, a->rd,
1246            tcg_constant_i64(a->imm), cpu_reg(s, a->rm))
1247 TRANS_FEAT(INDEX_ri, aa64_sve, do_index, a->esz, a->rd,
1248            cpu_reg(s, a->rn), tcg_constant_i64(a->imm))
1249 TRANS_FEAT(INDEX_rr, aa64_sve, do_index, a->esz, a->rd,
1250            cpu_reg(s, a->rn), cpu_reg(s, a->rm))
1251 
1252 /*
1253  *** SVE Stack Allocation Group
1254  */
1255 
1256 static bool trans_ADDVL(DisasContext *s, arg_ADDVL *a)
1257 {
1258     if (!dc_isar_feature(aa64_sve, s)) {
1259         return false;
1260     }
1261     if (sve_access_check(s)) {
1262         TCGv_i64 rd = cpu_reg_sp(s, a->rd);
1263         TCGv_i64 rn = cpu_reg_sp(s, a->rn);
1264         tcg_gen_addi_i64(rd, rn, a->imm * vec_full_reg_size(s));
1265     }
1266     return true;
1267 }
1268 
1269 static bool trans_ADDSVL(DisasContext *s, arg_ADDSVL *a)
1270 {
1271     if (!dc_isar_feature(aa64_sme, s)) {
1272         return false;
1273     }
1274     if (sme_enabled_check(s)) {
1275         TCGv_i64 rd = cpu_reg_sp(s, a->rd);
1276         TCGv_i64 rn = cpu_reg_sp(s, a->rn);
1277         tcg_gen_addi_i64(rd, rn, a->imm * streaming_vec_reg_size(s));
1278     }
1279     return true;
1280 }
1281 
1282 static bool trans_ADDPL(DisasContext *s, arg_ADDPL *a)
1283 {
1284     if (!dc_isar_feature(aa64_sve, s)) {
1285         return false;
1286     }
1287     if (sve_access_check(s)) {
1288         TCGv_i64 rd = cpu_reg_sp(s, a->rd);
1289         TCGv_i64 rn = cpu_reg_sp(s, a->rn);
1290         tcg_gen_addi_i64(rd, rn, a->imm * pred_full_reg_size(s));
1291     }
1292     return true;
1293 }
1294 
1295 static bool trans_ADDSPL(DisasContext *s, arg_ADDSPL *a)
1296 {
1297     if (!dc_isar_feature(aa64_sme, s)) {
1298         return false;
1299     }
1300     if (sme_enabled_check(s)) {
1301         TCGv_i64 rd = cpu_reg_sp(s, a->rd);
1302         TCGv_i64 rn = cpu_reg_sp(s, a->rn);
1303         tcg_gen_addi_i64(rd, rn, a->imm * streaming_pred_reg_size(s));
1304     }
1305     return true;
1306 }
1307 
1308 static bool trans_RDVL(DisasContext *s, arg_RDVL *a)
1309 {
1310     if (!dc_isar_feature(aa64_sve, s)) {
1311         return false;
1312     }
1313     if (sve_access_check(s)) {
1314         TCGv_i64 reg = cpu_reg(s, a->rd);
1315         tcg_gen_movi_i64(reg, a->imm * vec_full_reg_size(s));
1316     }
1317     return true;
1318 }
1319 
1320 static bool trans_RDSVL(DisasContext *s, arg_RDSVL *a)
1321 {
1322     if (!dc_isar_feature(aa64_sme, s)) {
1323         return false;
1324     }
1325     if (sme_enabled_check(s)) {
1326         TCGv_i64 reg = cpu_reg(s, a->rd);
1327         tcg_gen_movi_i64(reg, a->imm * streaming_vec_reg_size(s));
1328     }
1329     return true;
1330 }
1331 
1332 /*
1333  *** SVE Compute Vector Address Group
1334  */
1335 
1336 static bool do_adr(DisasContext *s, arg_rrri *a, gen_helper_gvec_3 *fn)
1337 {
1338     return gen_gvec_ool_zzz(s, fn, a->rd, a->rn, a->rm, a->imm);
1339 }
1340 
1341 TRANS_FEAT_NONSTREAMING(ADR_p32, aa64_sve, do_adr, a, gen_helper_sve_adr_p32)
1342 TRANS_FEAT_NONSTREAMING(ADR_p64, aa64_sve, do_adr, a, gen_helper_sve_adr_p64)
1343 TRANS_FEAT_NONSTREAMING(ADR_s32, aa64_sve, do_adr, a, gen_helper_sve_adr_s32)
1344 TRANS_FEAT_NONSTREAMING(ADR_u32, aa64_sve, do_adr, a, gen_helper_sve_adr_u32)
1345 
1346 /*
1347  *** SVE Integer Misc - Unpredicated Group
1348  */
1349 
1350 static gen_helper_gvec_2 * const fexpa_fns[4] = {
1351     NULL,                   gen_helper_sve_fexpa_h,
1352     gen_helper_sve_fexpa_s, gen_helper_sve_fexpa_d,
1353 };
1354 TRANS_FEAT_NONSTREAMING(FEXPA, aa64_sve, gen_gvec_ool_zz,
1355                         fexpa_fns[a->esz], a->rd, a->rn, 0)
1356 
1357 static gen_helper_gvec_3 * const ftssel_fns[4] = {
1358     NULL,                    gen_helper_sve_ftssel_h,
1359     gen_helper_sve_ftssel_s, gen_helper_sve_ftssel_d,
1360 };
1361 TRANS_FEAT_NONSTREAMING(FTSSEL, aa64_sve, gen_gvec_ool_arg_zzz,
1362                         ftssel_fns[a->esz], a, 0)
1363 
1364 /*
1365  *** SVE Predicate Logical Operations Group
1366  */
1367 
1368 static bool do_pppp_flags(DisasContext *s, arg_rprr_s *a,
1369                           const GVecGen4 *gvec_op)
1370 {
1371     if (!sve_access_check(s)) {
1372         return true;
1373     }
1374 
1375     unsigned psz = pred_gvec_reg_size(s);
1376     int dofs = pred_full_reg_offset(s, a->rd);
1377     int nofs = pred_full_reg_offset(s, a->rn);
1378     int mofs = pred_full_reg_offset(s, a->rm);
1379     int gofs = pred_full_reg_offset(s, a->pg);
1380 
1381     if (!a->s) {
1382         tcg_gen_gvec_4(dofs, nofs, mofs, gofs, psz, psz, gvec_op);
1383         return true;
1384     }
1385 
1386     if (psz == 8) {
1387         /* Do the operation and the flags generation in temps.  */
1388         TCGv_i64 pd = tcg_temp_new_i64();
1389         TCGv_i64 pn = tcg_temp_new_i64();
1390         TCGv_i64 pm = tcg_temp_new_i64();
1391         TCGv_i64 pg = tcg_temp_new_i64();
1392 
1393         tcg_gen_ld_i64(pn, cpu_env, nofs);
1394         tcg_gen_ld_i64(pm, cpu_env, mofs);
1395         tcg_gen_ld_i64(pg, cpu_env, gofs);
1396 
1397         gvec_op->fni8(pd, pn, pm, pg);
1398         tcg_gen_st_i64(pd, cpu_env, dofs);
1399 
1400         do_predtest1(pd, pg);
1401     } else {
1402         /* The operation and flags generation is large.  The computation
1403          * of the flags depends on the original contents of the guarding
1404          * predicate.  If the destination overwrites the guarding predicate,
1405          * then the easiest way to get this right is to save a copy.
1406           */
1407         int tofs = gofs;
1408         if (a->rd == a->pg) {
1409             tofs = offsetof(CPUARMState, vfp.preg_tmp);
1410             tcg_gen_gvec_mov(0, tofs, gofs, psz, psz);
1411         }
1412 
1413         tcg_gen_gvec_4(dofs, nofs, mofs, gofs, psz, psz, gvec_op);
1414         do_predtest(s, dofs, tofs, psz / 8);
1415     }
1416     return true;
1417 }
1418 
1419 static void gen_and_pg_i64(TCGv_i64 pd, TCGv_i64 pn, TCGv_i64 pm, TCGv_i64 pg)
1420 {
1421     tcg_gen_and_i64(pd, pn, pm);
1422     tcg_gen_and_i64(pd, pd, pg);
1423 }
1424 
1425 static void gen_and_pg_vec(unsigned vece, TCGv_vec pd, TCGv_vec pn,
1426                            TCGv_vec pm, TCGv_vec pg)
1427 {
1428     tcg_gen_and_vec(vece, pd, pn, pm);
1429     tcg_gen_and_vec(vece, pd, pd, pg);
1430 }
1431 
1432 static bool trans_AND_pppp(DisasContext *s, arg_rprr_s *a)
1433 {
1434     static const GVecGen4 op = {
1435         .fni8 = gen_and_pg_i64,
1436         .fniv = gen_and_pg_vec,
1437         .fno = gen_helper_sve_and_pppp,
1438         .prefer_i64 = TCG_TARGET_REG_BITS == 64,
1439     };
1440 
1441     if (!dc_isar_feature(aa64_sve, s)) {
1442         return false;
1443     }
1444     if (!a->s) {
1445         if (a->rn == a->rm) {
1446             if (a->pg == a->rn) {
1447                 return do_mov_p(s, a->rd, a->rn);
1448             }
1449             return gen_gvec_fn_ppp(s, tcg_gen_gvec_and, a->rd, a->rn, a->pg);
1450         } else if (a->pg == a->rn || a->pg == a->rm) {
1451             return gen_gvec_fn_ppp(s, tcg_gen_gvec_and, a->rd, a->rn, a->rm);
1452         }
1453     }
1454     return do_pppp_flags(s, a, &op);
1455 }
1456 
1457 static void gen_bic_pg_i64(TCGv_i64 pd, TCGv_i64 pn, TCGv_i64 pm, TCGv_i64 pg)
1458 {
1459     tcg_gen_andc_i64(pd, pn, pm);
1460     tcg_gen_and_i64(pd, pd, pg);
1461 }
1462 
1463 static void gen_bic_pg_vec(unsigned vece, TCGv_vec pd, TCGv_vec pn,
1464                            TCGv_vec pm, TCGv_vec pg)
1465 {
1466     tcg_gen_andc_vec(vece, pd, pn, pm);
1467     tcg_gen_and_vec(vece, pd, pd, pg);
1468 }
1469 
1470 static bool trans_BIC_pppp(DisasContext *s, arg_rprr_s *a)
1471 {
1472     static const GVecGen4 op = {
1473         .fni8 = gen_bic_pg_i64,
1474         .fniv = gen_bic_pg_vec,
1475         .fno = gen_helper_sve_bic_pppp,
1476         .prefer_i64 = TCG_TARGET_REG_BITS == 64,
1477     };
1478 
1479     if (!dc_isar_feature(aa64_sve, s)) {
1480         return false;
1481     }
1482     if (!a->s && a->pg == a->rn) {
1483         return gen_gvec_fn_ppp(s, tcg_gen_gvec_andc, a->rd, a->rn, a->rm);
1484     }
1485     return do_pppp_flags(s, a, &op);
1486 }
1487 
1488 static void gen_eor_pg_i64(TCGv_i64 pd, TCGv_i64 pn, TCGv_i64 pm, TCGv_i64 pg)
1489 {
1490     tcg_gen_xor_i64(pd, pn, pm);
1491     tcg_gen_and_i64(pd, pd, pg);
1492 }
1493 
1494 static void gen_eor_pg_vec(unsigned vece, TCGv_vec pd, TCGv_vec pn,
1495                            TCGv_vec pm, TCGv_vec pg)
1496 {
1497     tcg_gen_xor_vec(vece, pd, pn, pm);
1498     tcg_gen_and_vec(vece, pd, pd, pg);
1499 }
1500 
1501 static bool trans_EOR_pppp(DisasContext *s, arg_rprr_s *a)
1502 {
1503     static const GVecGen4 op = {
1504         .fni8 = gen_eor_pg_i64,
1505         .fniv = gen_eor_pg_vec,
1506         .fno = gen_helper_sve_eor_pppp,
1507         .prefer_i64 = TCG_TARGET_REG_BITS == 64,
1508     };
1509 
1510     if (!dc_isar_feature(aa64_sve, s)) {
1511         return false;
1512     }
1513     /* Alias NOT (predicate) is EOR Pd.B, Pg/Z, Pn.B, Pg.B */
1514     if (!a->s && a->pg == a->rm) {
1515         return gen_gvec_fn_ppp(s, tcg_gen_gvec_andc, a->rd, a->pg, a->rn);
1516     }
1517     return do_pppp_flags(s, a, &op);
1518 }
1519 
1520 static bool trans_SEL_pppp(DisasContext *s, arg_rprr_s *a)
1521 {
1522     if (a->s || !dc_isar_feature(aa64_sve, s)) {
1523         return false;
1524     }
1525     if (sve_access_check(s)) {
1526         unsigned psz = pred_gvec_reg_size(s);
1527         tcg_gen_gvec_bitsel(MO_8, pred_full_reg_offset(s, a->rd),
1528                             pred_full_reg_offset(s, a->pg),
1529                             pred_full_reg_offset(s, a->rn),
1530                             pred_full_reg_offset(s, a->rm), psz, psz);
1531     }
1532     return true;
1533 }
1534 
1535 static void gen_orr_pg_i64(TCGv_i64 pd, TCGv_i64 pn, TCGv_i64 pm, TCGv_i64 pg)
1536 {
1537     tcg_gen_or_i64(pd, pn, pm);
1538     tcg_gen_and_i64(pd, pd, pg);
1539 }
1540 
1541 static void gen_orr_pg_vec(unsigned vece, TCGv_vec pd, TCGv_vec pn,
1542                            TCGv_vec pm, TCGv_vec pg)
1543 {
1544     tcg_gen_or_vec(vece, pd, pn, pm);
1545     tcg_gen_and_vec(vece, pd, pd, pg);
1546 }
1547 
1548 static bool trans_ORR_pppp(DisasContext *s, arg_rprr_s *a)
1549 {
1550     static const GVecGen4 op = {
1551         .fni8 = gen_orr_pg_i64,
1552         .fniv = gen_orr_pg_vec,
1553         .fno = gen_helper_sve_orr_pppp,
1554         .prefer_i64 = TCG_TARGET_REG_BITS == 64,
1555     };
1556 
1557     if (!dc_isar_feature(aa64_sve, s)) {
1558         return false;
1559     }
1560     if (!a->s && a->pg == a->rn && a->rn == a->rm) {
1561         return do_mov_p(s, a->rd, a->rn);
1562     }
1563     return do_pppp_flags(s, a, &op);
1564 }
1565 
1566 static void gen_orn_pg_i64(TCGv_i64 pd, TCGv_i64 pn, TCGv_i64 pm, TCGv_i64 pg)
1567 {
1568     tcg_gen_orc_i64(pd, pn, pm);
1569     tcg_gen_and_i64(pd, pd, pg);
1570 }
1571 
1572 static void gen_orn_pg_vec(unsigned vece, TCGv_vec pd, TCGv_vec pn,
1573                            TCGv_vec pm, TCGv_vec pg)
1574 {
1575     tcg_gen_orc_vec(vece, pd, pn, pm);
1576     tcg_gen_and_vec(vece, pd, pd, pg);
1577 }
1578 
1579 static bool trans_ORN_pppp(DisasContext *s, arg_rprr_s *a)
1580 {
1581     static const GVecGen4 op = {
1582         .fni8 = gen_orn_pg_i64,
1583         .fniv = gen_orn_pg_vec,
1584         .fno = gen_helper_sve_orn_pppp,
1585         .prefer_i64 = TCG_TARGET_REG_BITS == 64,
1586     };
1587 
1588     if (!dc_isar_feature(aa64_sve, s)) {
1589         return false;
1590     }
1591     return do_pppp_flags(s, a, &op);
1592 }
1593 
1594 static void gen_nor_pg_i64(TCGv_i64 pd, TCGv_i64 pn, TCGv_i64 pm, TCGv_i64 pg)
1595 {
1596     tcg_gen_or_i64(pd, pn, pm);
1597     tcg_gen_andc_i64(pd, pg, pd);
1598 }
1599 
1600 static void gen_nor_pg_vec(unsigned vece, TCGv_vec pd, TCGv_vec pn,
1601                            TCGv_vec pm, TCGv_vec pg)
1602 {
1603     tcg_gen_or_vec(vece, pd, pn, pm);
1604     tcg_gen_andc_vec(vece, pd, pg, pd);
1605 }
1606 
1607 static bool trans_NOR_pppp(DisasContext *s, arg_rprr_s *a)
1608 {
1609     static const GVecGen4 op = {
1610         .fni8 = gen_nor_pg_i64,
1611         .fniv = gen_nor_pg_vec,
1612         .fno = gen_helper_sve_nor_pppp,
1613         .prefer_i64 = TCG_TARGET_REG_BITS == 64,
1614     };
1615 
1616     if (!dc_isar_feature(aa64_sve, s)) {
1617         return false;
1618     }
1619     return do_pppp_flags(s, a, &op);
1620 }
1621 
1622 static void gen_nand_pg_i64(TCGv_i64 pd, TCGv_i64 pn, TCGv_i64 pm, TCGv_i64 pg)
1623 {
1624     tcg_gen_and_i64(pd, pn, pm);
1625     tcg_gen_andc_i64(pd, pg, pd);
1626 }
1627 
1628 static void gen_nand_pg_vec(unsigned vece, TCGv_vec pd, TCGv_vec pn,
1629                            TCGv_vec pm, TCGv_vec pg)
1630 {
1631     tcg_gen_and_vec(vece, pd, pn, pm);
1632     tcg_gen_andc_vec(vece, pd, pg, pd);
1633 }
1634 
1635 static bool trans_NAND_pppp(DisasContext *s, arg_rprr_s *a)
1636 {
1637     static const GVecGen4 op = {
1638         .fni8 = gen_nand_pg_i64,
1639         .fniv = gen_nand_pg_vec,
1640         .fno = gen_helper_sve_nand_pppp,
1641         .prefer_i64 = TCG_TARGET_REG_BITS == 64,
1642     };
1643 
1644     if (!dc_isar_feature(aa64_sve, s)) {
1645         return false;
1646     }
1647     return do_pppp_flags(s, a, &op);
1648 }
1649 
1650 /*
1651  *** SVE Predicate Misc Group
1652  */
1653 
1654 static bool trans_PTEST(DisasContext *s, arg_PTEST *a)
1655 {
1656     if (!dc_isar_feature(aa64_sve, s)) {
1657         return false;
1658     }
1659     if (sve_access_check(s)) {
1660         int nofs = pred_full_reg_offset(s, a->rn);
1661         int gofs = pred_full_reg_offset(s, a->pg);
1662         int words = DIV_ROUND_UP(pred_full_reg_size(s), 8);
1663 
1664         if (words == 1) {
1665             TCGv_i64 pn = tcg_temp_new_i64();
1666             TCGv_i64 pg = tcg_temp_new_i64();
1667 
1668             tcg_gen_ld_i64(pn, cpu_env, nofs);
1669             tcg_gen_ld_i64(pg, cpu_env, gofs);
1670             do_predtest1(pn, pg);
1671         } else {
1672             do_predtest(s, nofs, gofs, words);
1673         }
1674     }
1675     return true;
1676 }
1677 
1678 /* See the ARM pseudocode DecodePredCount.  */
1679 static unsigned decode_pred_count(unsigned fullsz, int pattern, int esz)
1680 {
1681     unsigned elements = fullsz >> esz;
1682     unsigned bound;
1683 
1684     switch (pattern) {
1685     case 0x0: /* POW2 */
1686         return pow2floor(elements);
1687     case 0x1: /* VL1 */
1688     case 0x2: /* VL2 */
1689     case 0x3: /* VL3 */
1690     case 0x4: /* VL4 */
1691     case 0x5: /* VL5 */
1692     case 0x6: /* VL6 */
1693     case 0x7: /* VL7 */
1694     case 0x8: /* VL8 */
1695         bound = pattern;
1696         break;
1697     case 0x9: /* VL16 */
1698     case 0xa: /* VL32 */
1699     case 0xb: /* VL64 */
1700     case 0xc: /* VL128 */
1701     case 0xd: /* VL256 */
1702         bound = 16 << (pattern - 9);
1703         break;
1704     case 0x1d: /* MUL4 */
1705         return elements - elements % 4;
1706     case 0x1e: /* MUL3 */
1707         return elements - elements % 3;
1708     case 0x1f: /* ALL */
1709         return elements;
1710     default:   /* #uimm5 */
1711         return 0;
1712     }
1713     return elements >= bound ? bound : 0;
1714 }
1715 
1716 /* This handles all of the predicate initialization instructions,
1717  * PTRUE, PFALSE, SETFFR.  For PFALSE, we will have set PAT == 32
1718  * so that decode_pred_count returns 0.  For SETFFR, we will have
1719  * set RD == 16 == FFR.
1720  */
1721 static bool do_predset(DisasContext *s, int esz, int rd, int pat, bool setflag)
1722 {
1723     if (!sve_access_check(s)) {
1724         return true;
1725     }
1726 
1727     unsigned fullsz = vec_full_reg_size(s);
1728     unsigned ofs = pred_full_reg_offset(s, rd);
1729     unsigned numelem, setsz, i;
1730     uint64_t word, lastword;
1731     TCGv_i64 t;
1732 
1733     numelem = decode_pred_count(fullsz, pat, esz);
1734 
1735     /* Determine what we must store into each bit, and how many.  */
1736     if (numelem == 0) {
1737         lastword = word = 0;
1738         setsz = fullsz;
1739     } else {
1740         setsz = numelem << esz;
1741         lastword = word = pred_esz_masks[esz];
1742         if (setsz % 64) {
1743             lastword &= MAKE_64BIT_MASK(0, setsz % 64);
1744         }
1745     }
1746 
1747     t = tcg_temp_new_i64();
1748     if (fullsz <= 64) {
1749         tcg_gen_movi_i64(t, lastword);
1750         tcg_gen_st_i64(t, cpu_env, ofs);
1751         goto done;
1752     }
1753 
1754     if (word == lastword) {
1755         unsigned maxsz = size_for_gvec(fullsz / 8);
1756         unsigned oprsz = size_for_gvec(setsz / 8);
1757 
1758         if (oprsz * 8 == setsz) {
1759             tcg_gen_gvec_dup_imm(MO_64, ofs, oprsz, maxsz, word);
1760             goto done;
1761         }
1762     }
1763 
1764     setsz /= 8;
1765     fullsz /= 8;
1766 
1767     tcg_gen_movi_i64(t, word);
1768     for (i = 0; i < QEMU_ALIGN_DOWN(setsz, 8); i += 8) {
1769         tcg_gen_st_i64(t, cpu_env, ofs + i);
1770     }
1771     if (lastword != word) {
1772         tcg_gen_movi_i64(t, lastword);
1773         tcg_gen_st_i64(t, cpu_env, ofs + i);
1774         i += 8;
1775     }
1776     if (i < fullsz) {
1777         tcg_gen_movi_i64(t, 0);
1778         for (; i < fullsz; i += 8) {
1779             tcg_gen_st_i64(t, cpu_env, ofs + i);
1780         }
1781     }
1782 
1783  done:
1784     /* PTRUES */
1785     if (setflag) {
1786         tcg_gen_movi_i32(cpu_NF, -(word != 0));
1787         tcg_gen_movi_i32(cpu_CF, word == 0);
1788         tcg_gen_movi_i32(cpu_VF, 0);
1789         tcg_gen_mov_i32(cpu_ZF, cpu_NF);
1790     }
1791     return true;
1792 }
1793 
1794 TRANS_FEAT(PTRUE, aa64_sve, do_predset, a->esz, a->rd, a->pat, a->s)
1795 
1796 /* Note pat == 31 is #all, to set all elements.  */
1797 TRANS_FEAT_NONSTREAMING(SETFFR, aa64_sve,
1798                         do_predset, 0, FFR_PRED_NUM, 31, false)
1799 
1800 /* Note pat == 32 is #unimp, to set no elements.  */
1801 TRANS_FEAT(PFALSE, aa64_sve, do_predset, 0, a->rd, 32, false)
1802 
1803 static bool trans_RDFFR_p(DisasContext *s, arg_RDFFR_p *a)
1804 {
1805     /* The path through do_pppp_flags is complicated enough to want to avoid
1806      * duplication.  Frob the arguments into the form of a predicated AND.
1807      */
1808     arg_rprr_s alt_a = {
1809         .rd = a->rd, .pg = a->pg, .s = a->s,
1810         .rn = FFR_PRED_NUM, .rm = FFR_PRED_NUM,
1811     };
1812 
1813     s->is_nonstreaming = true;
1814     return trans_AND_pppp(s, &alt_a);
1815 }
1816 
1817 TRANS_FEAT_NONSTREAMING(RDFFR, aa64_sve, do_mov_p, a->rd, FFR_PRED_NUM)
1818 TRANS_FEAT_NONSTREAMING(WRFFR, aa64_sve, do_mov_p, FFR_PRED_NUM, a->rn)
1819 
1820 static bool do_pfirst_pnext(DisasContext *s, arg_rr_esz *a,
1821                             void (*gen_fn)(TCGv_i32, TCGv_ptr,
1822                                            TCGv_ptr, TCGv_i32))
1823 {
1824     if (!sve_access_check(s)) {
1825         return true;
1826     }
1827 
1828     TCGv_ptr t_pd = tcg_temp_new_ptr();
1829     TCGv_ptr t_pg = tcg_temp_new_ptr();
1830     TCGv_i32 t;
1831     unsigned desc = 0;
1832 
1833     desc = FIELD_DP32(desc, PREDDESC, OPRSZ, pred_full_reg_size(s));
1834     desc = FIELD_DP32(desc, PREDDESC, ESZ, a->esz);
1835 
1836     tcg_gen_addi_ptr(t_pd, cpu_env, pred_full_reg_offset(s, a->rd));
1837     tcg_gen_addi_ptr(t_pg, cpu_env, pred_full_reg_offset(s, a->rn));
1838     t = tcg_temp_new_i32();
1839 
1840     gen_fn(t, t_pd, t_pg, tcg_constant_i32(desc));
1841 
1842     do_pred_flags(t);
1843     return true;
1844 }
1845 
1846 TRANS_FEAT(PFIRST, aa64_sve, do_pfirst_pnext, a, gen_helper_sve_pfirst)
1847 TRANS_FEAT(PNEXT, aa64_sve, do_pfirst_pnext, a, gen_helper_sve_pnext)
1848 
1849 /*
1850  *** SVE Element Count Group
1851  */
1852 
1853 /* Perform an inline saturating addition of a 32-bit value within
1854  * a 64-bit register.  The second operand is known to be positive,
1855  * which halves the comparisions we must perform to bound the result.
1856  */
1857 static void do_sat_addsub_32(TCGv_i64 reg, TCGv_i64 val, bool u, bool d)
1858 {
1859     int64_t ibound;
1860 
1861     /* Use normal 64-bit arithmetic to detect 32-bit overflow.  */
1862     if (u) {
1863         tcg_gen_ext32u_i64(reg, reg);
1864     } else {
1865         tcg_gen_ext32s_i64(reg, reg);
1866     }
1867     if (d) {
1868         tcg_gen_sub_i64(reg, reg, val);
1869         ibound = (u ? 0 : INT32_MIN);
1870         tcg_gen_smax_i64(reg, reg, tcg_constant_i64(ibound));
1871     } else {
1872         tcg_gen_add_i64(reg, reg, val);
1873         ibound = (u ? UINT32_MAX : INT32_MAX);
1874         tcg_gen_smin_i64(reg, reg, tcg_constant_i64(ibound));
1875     }
1876 }
1877 
1878 /* Similarly with 64-bit values.  */
1879 static void do_sat_addsub_64(TCGv_i64 reg, TCGv_i64 val, bool u, bool d)
1880 {
1881     TCGv_i64 t0 = tcg_temp_new_i64();
1882     TCGv_i64 t2;
1883 
1884     if (u) {
1885         if (d) {
1886             tcg_gen_sub_i64(t0, reg, val);
1887             t2 = tcg_constant_i64(0);
1888             tcg_gen_movcond_i64(TCG_COND_LTU, reg, reg, val, t2, t0);
1889         } else {
1890             tcg_gen_add_i64(t0, reg, val);
1891             t2 = tcg_constant_i64(-1);
1892             tcg_gen_movcond_i64(TCG_COND_LTU, reg, t0, reg, t2, t0);
1893         }
1894     } else {
1895         TCGv_i64 t1 = tcg_temp_new_i64();
1896         if (d) {
1897             /* Detect signed overflow for subtraction.  */
1898             tcg_gen_xor_i64(t0, reg, val);
1899             tcg_gen_sub_i64(t1, reg, val);
1900             tcg_gen_xor_i64(reg, reg, t1);
1901             tcg_gen_and_i64(t0, t0, reg);
1902 
1903             /* Bound the result.  */
1904             tcg_gen_movi_i64(reg, INT64_MIN);
1905             t2 = tcg_constant_i64(0);
1906             tcg_gen_movcond_i64(TCG_COND_LT, reg, t0, t2, reg, t1);
1907         } else {
1908             /* Detect signed overflow for addition.  */
1909             tcg_gen_xor_i64(t0, reg, val);
1910             tcg_gen_add_i64(reg, reg, val);
1911             tcg_gen_xor_i64(t1, reg, val);
1912             tcg_gen_andc_i64(t0, t1, t0);
1913 
1914             /* Bound the result.  */
1915             tcg_gen_movi_i64(t1, INT64_MAX);
1916             t2 = tcg_constant_i64(0);
1917             tcg_gen_movcond_i64(TCG_COND_LT, reg, t0, t2, t1, reg);
1918         }
1919     }
1920 }
1921 
1922 /* Similarly with a vector and a scalar operand.  */
1923 static void do_sat_addsub_vec(DisasContext *s, int esz, int rd, int rn,
1924                               TCGv_i64 val, bool u, bool d)
1925 {
1926     unsigned vsz = vec_full_reg_size(s);
1927     TCGv_ptr dptr, nptr;
1928     TCGv_i32 t32, desc;
1929     TCGv_i64 t64;
1930 
1931     dptr = tcg_temp_new_ptr();
1932     nptr = tcg_temp_new_ptr();
1933     tcg_gen_addi_ptr(dptr, cpu_env, vec_full_reg_offset(s, rd));
1934     tcg_gen_addi_ptr(nptr, cpu_env, vec_full_reg_offset(s, rn));
1935     desc = tcg_constant_i32(simd_desc(vsz, vsz, 0));
1936 
1937     switch (esz) {
1938     case MO_8:
1939         t32 = tcg_temp_new_i32();
1940         tcg_gen_extrl_i64_i32(t32, val);
1941         if (d) {
1942             tcg_gen_neg_i32(t32, t32);
1943         }
1944         if (u) {
1945             gen_helper_sve_uqaddi_b(dptr, nptr, t32, desc);
1946         } else {
1947             gen_helper_sve_sqaddi_b(dptr, nptr, t32, desc);
1948         }
1949         break;
1950 
1951     case MO_16:
1952         t32 = tcg_temp_new_i32();
1953         tcg_gen_extrl_i64_i32(t32, val);
1954         if (d) {
1955             tcg_gen_neg_i32(t32, t32);
1956         }
1957         if (u) {
1958             gen_helper_sve_uqaddi_h(dptr, nptr, t32, desc);
1959         } else {
1960             gen_helper_sve_sqaddi_h(dptr, nptr, t32, desc);
1961         }
1962         break;
1963 
1964     case MO_32:
1965         t64 = tcg_temp_new_i64();
1966         if (d) {
1967             tcg_gen_neg_i64(t64, val);
1968         } else {
1969             tcg_gen_mov_i64(t64, val);
1970         }
1971         if (u) {
1972             gen_helper_sve_uqaddi_s(dptr, nptr, t64, desc);
1973         } else {
1974             gen_helper_sve_sqaddi_s(dptr, nptr, t64, desc);
1975         }
1976         break;
1977 
1978     case MO_64:
1979         if (u) {
1980             if (d) {
1981                 gen_helper_sve_uqsubi_d(dptr, nptr, val, desc);
1982             } else {
1983                 gen_helper_sve_uqaddi_d(dptr, nptr, val, desc);
1984             }
1985         } else if (d) {
1986             t64 = tcg_temp_new_i64();
1987             tcg_gen_neg_i64(t64, val);
1988             gen_helper_sve_sqaddi_d(dptr, nptr, t64, desc);
1989         } else {
1990             gen_helper_sve_sqaddi_d(dptr, nptr, val, desc);
1991         }
1992         break;
1993 
1994     default:
1995         g_assert_not_reached();
1996     }
1997 }
1998 
1999 static bool trans_CNT_r(DisasContext *s, arg_CNT_r *a)
2000 {
2001     if (!dc_isar_feature(aa64_sve, s)) {
2002         return false;
2003     }
2004     if (sve_access_check(s)) {
2005         unsigned fullsz = vec_full_reg_size(s);
2006         unsigned numelem = decode_pred_count(fullsz, a->pat, a->esz);
2007         tcg_gen_movi_i64(cpu_reg(s, a->rd), numelem * a->imm);
2008     }
2009     return true;
2010 }
2011 
2012 static bool trans_INCDEC_r(DisasContext *s, arg_incdec_cnt *a)
2013 {
2014     if (!dc_isar_feature(aa64_sve, s)) {
2015         return false;
2016     }
2017     if (sve_access_check(s)) {
2018         unsigned fullsz = vec_full_reg_size(s);
2019         unsigned numelem = decode_pred_count(fullsz, a->pat, a->esz);
2020         int inc = numelem * a->imm * (a->d ? -1 : 1);
2021         TCGv_i64 reg = cpu_reg(s, a->rd);
2022 
2023         tcg_gen_addi_i64(reg, reg, inc);
2024     }
2025     return true;
2026 }
2027 
2028 static bool trans_SINCDEC_r_32(DisasContext *s, arg_incdec_cnt *a)
2029 {
2030     if (!dc_isar_feature(aa64_sve, s)) {
2031         return false;
2032     }
2033     if (!sve_access_check(s)) {
2034         return true;
2035     }
2036 
2037     unsigned fullsz = vec_full_reg_size(s);
2038     unsigned numelem = decode_pred_count(fullsz, a->pat, a->esz);
2039     int inc = numelem * a->imm;
2040     TCGv_i64 reg = cpu_reg(s, a->rd);
2041 
2042     /* Use normal 64-bit arithmetic to detect 32-bit overflow.  */
2043     if (inc == 0) {
2044         if (a->u) {
2045             tcg_gen_ext32u_i64(reg, reg);
2046         } else {
2047             tcg_gen_ext32s_i64(reg, reg);
2048         }
2049     } else {
2050         do_sat_addsub_32(reg, tcg_constant_i64(inc), a->u, a->d);
2051     }
2052     return true;
2053 }
2054 
2055 static bool trans_SINCDEC_r_64(DisasContext *s, arg_incdec_cnt *a)
2056 {
2057     if (!dc_isar_feature(aa64_sve, s)) {
2058         return false;
2059     }
2060     if (!sve_access_check(s)) {
2061         return true;
2062     }
2063 
2064     unsigned fullsz = vec_full_reg_size(s);
2065     unsigned numelem = decode_pred_count(fullsz, a->pat, a->esz);
2066     int inc = numelem * a->imm;
2067     TCGv_i64 reg = cpu_reg(s, a->rd);
2068 
2069     if (inc != 0) {
2070         do_sat_addsub_64(reg, tcg_constant_i64(inc), a->u, a->d);
2071     }
2072     return true;
2073 }
2074 
2075 static bool trans_INCDEC_v(DisasContext *s, arg_incdec2_cnt *a)
2076 {
2077     if (a->esz == 0 || !dc_isar_feature(aa64_sve, s)) {
2078         return false;
2079     }
2080 
2081     unsigned fullsz = vec_full_reg_size(s);
2082     unsigned numelem = decode_pred_count(fullsz, a->pat, a->esz);
2083     int inc = numelem * a->imm;
2084 
2085     if (inc != 0) {
2086         if (sve_access_check(s)) {
2087             tcg_gen_gvec_adds(a->esz, vec_full_reg_offset(s, a->rd),
2088                               vec_full_reg_offset(s, a->rn),
2089                               tcg_constant_i64(a->d ? -inc : inc),
2090                               fullsz, fullsz);
2091         }
2092     } else {
2093         do_mov_z(s, a->rd, a->rn);
2094     }
2095     return true;
2096 }
2097 
2098 static bool trans_SINCDEC_v(DisasContext *s, arg_incdec2_cnt *a)
2099 {
2100     if (a->esz == 0 || !dc_isar_feature(aa64_sve, s)) {
2101         return false;
2102     }
2103 
2104     unsigned fullsz = vec_full_reg_size(s);
2105     unsigned numelem = decode_pred_count(fullsz, a->pat, a->esz);
2106     int inc = numelem * a->imm;
2107 
2108     if (inc != 0) {
2109         if (sve_access_check(s)) {
2110             do_sat_addsub_vec(s, a->esz, a->rd, a->rn,
2111                               tcg_constant_i64(inc), a->u, a->d);
2112         }
2113     } else {
2114         do_mov_z(s, a->rd, a->rn);
2115     }
2116     return true;
2117 }
2118 
2119 /*
2120  *** SVE Bitwise Immediate Group
2121  */
2122 
2123 static bool do_zz_dbm(DisasContext *s, arg_rr_dbm *a, GVecGen2iFn *gvec_fn)
2124 {
2125     uint64_t imm;
2126     if (!logic_imm_decode_wmask(&imm, extract32(a->dbm, 12, 1),
2127                                 extract32(a->dbm, 0, 6),
2128                                 extract32(a->dbm, 6, 6))) {
2129         return false;
2130     }
2131     return gen_gvec_fn_zzi(s, gvec_fn, MO_64, a->rd, a->rn, imm);
2132 }
2133 
2134 TRANS_FEAT(AND_zzi, aa64_sve, do_zz_dbm, a, tcg_gen_gvec_andi)
2135 TRANS_FEAT(ORR_zzi, aa64_sve, do_zz_dbm, a, tcg_gen_gvec_ori)
2136 TRANS_FEAT(EOR_zzi, aa64_sve, do_zz_dbm, a, tcg_gen_gvec_xori)
2137 
2138 static bool trans_DUPM(DisasContext *s, arg_DUPM *a)
2139 {
2140     uint64_t imm;
2141 
2142     if (!dc_isar_feature(aa64_sve, s)) {
2143         return false;
2144     }
2145     if (!logic_imm_decode_wmask(&imm, extract32(a->dbm, 12, 1),
2146                                 extract32(a->dbm, 0, 6),
2147                                 extract32(a->dbm, 6, 6))) {
2148         return false;
2149     }
2150     if (sve_access_check(s)) {
2151         do_dupi_z(s, a->rd, imm);
2152     }
2153     return true;
2154 }
2155 
2156 /*
2157  *** SVE Integer Wide Immediate - Predicated Group
2158  */
2159 
2160 /* Implement all merging copies.  This is used for CPY (immediate),
2161  * FCPY, CPY (scalar), CPY (SIMD&FP scalar).
2162  */
2163 static void do_cpy_m(DisasContext *s, int esz, int rd, int rn, int pg,
2164                      TCGv_i64 val)
2165 {
2166     typedef void gen_cpy(TCGv_ptr, TCGv_ptr, TCGv_ptr, TCGv_i64, TCGv_i32);
2167     static gen_cpy * const fns[4] = {
2168         gen_helper_sve_cpy_m_b, gen_helper_sve_cpy_m_h,
2169         gen_helper_sve_cpy_m_s, gen_helper_sve_cpy_m_d,
2170     };
2171     unsigned vsz = vec_full_reg_size(s);
2172     TCGv_i32 desc = tcg_constant_i32(simd_desc(vsz, vsz, 0));
2173     TCGv_ptr t_zd = tcg_temp_new_ptr();
2174     TCGv_ptr t_zn = tcg_temp_new_ptr();
2175     TCGv_ptr t_pg = tcg_temp_new_ptr();
2176 
2177     tcg_gen_addi_ptr(t_zd, cpu_env, vec_full_reg_offset(s, rd));
2178     tcg_gen_addi_ptr(t_zn, cpu_env, vec_full_reg_offset(s, rn));
2179     tcg_gen_addi_ptr(t_pg, cpu_env, pred_full_reg_offset(s, pg));
2180 
2181     fns[esz](t_zd, t_zn, t_pg, val, desc);
2182 }
2183 
2184 static bool trans_FCPY(DisasContext *s, arg_FCPY *a)
2185 {
2186     if (a->esz == 0 || !dc_isar_feature(aa64_sve, s)) {
2187         return false;
2188     }
2189     if (sve_access_check(s)) {
2190         /* Decode the VFP immediate.  */
2191         uint64_t imm = vfp_expand_imm(a->esz, a->imm);
2192         do_cpy_m(s, a->esz, a->rd, a->rn, a->pg, tcg_constant_i64(imm));
2193     }
2194     return true;
2195 }
2196 
2197 static bool trans_CPY_m_i(DisasContext *s, arg_rpri_esz *a)
2198 {
2199     if (!dc_isar_feature(aa64_sve, s)) {
2200         return false;
2201     }
2202     if (sve_access_check(s)) {
2203         do_cpy_m(s, a->esz, a->rd, a->rn, a->pg, tcg_constant_i64(a->imm));
2204     }
2205     return true;
2206 }
2207 
2208 static bool trans_CPY_z_i(DisasContext *s, arg_CPY_z_i *a)
2209 {
2210     static gen_helper_gvec_2i * const fns[4] = {
2211         gen_helper_sve_cpy_z_b, gen_helper_sve_cpy_z_h,
2212         gen_helper_sve_cpy_z_s, gen_helper_sve_cpy_z_d,
2213     };
2214 
2215     if (!dc_isar_feature(aa64_sve, s)) {
2216         return false;
2217     }
2218     if (sve_access_check(s)) {
2219         unsigned vsz = vec_full_reg_size(s);
2220         tcg_gen_gvec_2i_ool(vec_full_reg_offset(s, a->rd),
2221                             pred_full_reg_offset(s, a->pg),
2222                             tcg_constant_i64(a->imm),
2223                             vsz, vsz, 0, fns[a->esz]);
2224     }
2225     return true;
2226 }
2227 
2228 /*
2229  *** SVE Permute Extract Group
2230  */
2231 
2232 static bool do_EXT(DisasContext *s, int rd, int rn, int rm, int imm)
2233 {
2234     if (!sve_access_check(s)) {
2235         return true;
2236     }
2237 
2238     unsigned vsz = vec_full_reg_size(s);
2239     unsigned n_ofs = imm >= vsz ? 0 : imm;
2240     unsigned n_siz = vsz - n_ofs;
2241     unsigned d = vec_full_reg_offset(s, rd);
2242     unsigned n = vec_full_reg_offset(s, rn);
2243     unsigned m = vec_full_reg_offset(s, rm);
2244 
2245     /* Use host vector move insns if we have appropriate sizes
2246      * and no unfortunate overlap.
2247      */
2248     if (m != d
2249         && n_ofs == size_for_gvec(n_ofs)
2250         && n_siz == size_for_gvec(n_siz)
2251         && (d != n || n_siz <= n_ofs)) {
2252         tcg_gen_gvec_mov(0, d, n + n_ofs, n_siz, n_siz);
2253         if (n_ofs != 0) {
2254             tcg_gen_gvec_mov(0, d + n_siz, m, n_ofs, n_ofs);
2255         }
2256     } else {
2257         tcg_gen_gvec_3_ool(d, n, m, vsz, vsz, n_ofs, gen_helper_sve_ext);
2258     }
2259     return true;
2260 }
2261 
2262 TRANS_FEAT(EXT, aa64_sve, do_EXT, a->rd, a->rn, a->rm, a->imm)
2263 TRANS_FEAT(EXT_sve2, aa64_sve2, do_EXT, a->rd, a->rn, (a->rn + 1) % 32, a->imm)
2264 
2265 /*
2266  *** SVE Permute - Unpredicated Group
2267  */
2268 
2269 static bool trans_DUP_s(DisasContext *s, arg_DUP_s *a)
2270 {
2271     if (!dc_isar_feature(aa64_sve, s)) {
2272         return false;
2273     }
2274     if (sve_access_check(s)) {
2275         unsigned vsz = vec_full_reg_size(s);
2276         tcg_gen_gvec_dup_i64(a->esz, vec_full_reg_offset(s, a->rd),
2277                              vsz, vsz, cpu_reg_sp(s, a->rn));
2278     }
2279     return true;
2280 }
2281 
2282 static bool trans_DUP_x(DisasContext *s, arg_DUP_x *a)
2283 {
2284     if (!dc_isar_feature(aa64_sve, s)) {
2285         return false;
2286     }
2287     if ((a->imm & 0x1f) == 0) {
2288         return false;
2289     }
2290     if (sve_access_check(s)) {
2291         unsigned vsz = vec_full_reg_size(s);
2292         unsigned dofs = vec_full_reg_offset(s, a->rd);
2293         unsigned esz, index;
2294 
2295         esz = ctz32(a->imm);
2296         index = a->imm >> (esz + 1);
2297 
2298         if ((index << esz) < vsz) {
2299             unsigned nofs = vec_reg_offset(s, a->rn, index, esz);
2300             tcg_gen_gvec_dup_mem(esz, dofs, nofs, vsz, vsz);
2301         } else {
2302             /*
2303              * While dup_mem handles 128-bit elements, dup_imm does not.
2304              * Thankfully element size doesn't matter for splatting zero.
2305              */
2306             tcg_gen_gvec_dup_imm(MO_64, dofs, vsz, vsz, 0);
2307         }
2308     }
2309     return true;
2310 }
2311 
2312 static void do_insr_i64(DisasContext *s, arg_rrr_esz *a, TCGv_i64 val)
2313 {
2314     typedef void gen_insr(TCGv_ptr, TCGv_ptr, TCGv_i64, TCGv_i32);
2315     static gen_insr * const fns[4] = {
2316         gen_helper_sve_insr_b, gen_helper_sve_insr_h,
2317         gen_helper_sve_insr_s, gen_helper_sve_insr_d,
2318     };
2319     unsigned vsz = vec_full_reg_size(s);
2320     TCGv_i32 desc = tcg_constant_i32(simd_desc(vsz, vsz, 0));
2321     TCGv_ptr t_zd = tcg_temp_new_ptr();
2322     TCGv_ptr t_zn = tcg_temp_new_ptr();
2323 
2324     tcg_gen_addi_ptr(t_zd, cpu_env, vec_full_reg_offset(s, a->rd));
2325     tcg_gen_addi_ptr(t_zn, cpu_env, vec_full_reg_offset(s, a->rn));
2326 
2327     fns[a->esz](t_zd, t_zn, val, desc);
2328 }
2329 
2330 static bool trans_INSR_f(DisasContext *s, arg_rrr_esz *a)
2331 {
2332     if (!dc_isar_feature(aa64_sve, s)) {
2333         return false;
2334     }
2335     if (sve_access_check(s)) {
2336         TCGv_i64 t = tcg_temp_new_i64();
2337         tcg_gen_ld_i64(t, cpu_env, vec_reg_offset(s, a->rm, 0, MO_64));
2338         do_insr_i64(s, a, t);
2339     }
2340     return true;
2341 }
2342 
2343 static bool trans_INSR_r(DisasContext *s, arg_rrr_esz *a)
2344 {
2345     if (!dc_isar_feature(aa64_sve, s)) {
2346         return false;
2347     }
2348     if (sve_access_check(s)) {
2349         do_insr_i64(s, a, cpu_reg(s, a->rm));
2350     }
2351     return true;
2352 }
2353 
2354 static gen_helper_gvec_2 * const rev_fns[4] = {
2355     gen_helper_sve_rev_b, gen_helper_sve_rev_h,
2356     gen_helper_sve_rev_s, gen_helper_sve_rev_d
2357 };
2358 TRANS_FEAT(REV_v, aa64_sve, gen_gvec_ool_zz, rev_fns[a->esz], a->rd, a->rn, 0)
2359 
2360 static gen_helper_gvec_3 * const sve_tbl_fns[4] = {
2361     gen_helper_sve_tbl_b, gen_helper_sve_tbl_h,
2362     gen_helper_sve_tbl_s, gen_helper_sve_tbl_d
2363 };
2364 TRANS_FEAT(TBL, aa64_sve, gen_gvec_ool_arg_zzz, sve_tbl_fns[a->esz], a, 0)
2365 
2366 static gen_helper_gvec_4 * const sve2_tbl_fns[4] = {
2367     gen_helper_sve2_tbl_b, gen_helper_sve2_tbl_h,
2368     gen_helper_sve2_tbl_s, gen_helper_sve2_tbl_d
2369 };
2370 TRANS_FEAT(TBL_sve2, aa64_sve2, gen_gvec_ool_zzzz, sve2_tbl_fns[a->esz],
2371            a->rd, a->rn, (a->rn + 1) % 32, a->rm, 0)
2372 
2373 static gen_helper_gvec_3 * const tbx_fns[4] = {
2374     gen_helper_sve2_tbx_b, gen_helper_sve2_tbx_h,
2375     gen_helper_sve2_tbx_s, gen_helper_sve2_tbx_d
2376 };
2377 TRANS_FEAT(TBX, aa64_sve2, gen_gvec_ool_arg_zzz, tbx_fns[a->esz], a, 0)
2378 
2379 static bool trans_UNPK(DisasContext *s, arg_UNPK *a)
2380 {
2381     static gen_helper_gvec_2 * const fns[4][2] = {
2382         { NULL, NULL },
2383         { gen_helper_sve_sunpk_h, gen_helper_sve_uunpk_h },
2384         { gen_helper_sve_sunpk_s, gen_helper_sve_uunpk_s },
2385         { gen_helper_sve_sunpk_d, gen_helper_sve_uunpk_d },
2386     };
2387 
2388     if (a->esz == 0 || !dc_isar_feature(aa64_sve, s)) {
2389         return false;
2390     }
2391     if (sve_access_check(s)) {
2392         unsigned vsz = vec_full_reg_size(s);
2393         tcg_gen_gvec_2_ool(vec_full_reg_offset(s, a->rd),
2394                            vec_full_reg_offset(s, a->rn)
2395                            + (a->h ? vsz / 2 : 0),
2396                            vsz, vsz, 0, fns[a->esz][a->u]);
2397     }
2398     return true;
2399 }
2400 
2401 /*
2402  *** SVE Permute - Predicates Group
2403  */
2404 
2405 static bool do_perm_pred3(DisasContext *s, arg_rrr_esz *a, bool high_odd,
2406                           gen_helper_gvec_3 *fn)
2407 {
2408     if (!sve_access_check(s)) {
2409         return true;
2410     }
2411 
2412     unsigned vsz = pred_full_reg_size(s);
2413 
2414     TCGv_ptr t_d = tcg_temp_new_ptr();
2415     TCGv_ptr t_n = tcg_temp_new_ptr();
2416     TCGv_ptr t_m = tcg_temp_new_ptr();
2417     uint32_t desc = 0;
2418 
2419     desc = FIELD_DP32(desc, PREDDESC, OPRSZ, vsz);
2420     desc = FIELD_DP32(desc, PREDDESC, ESZ, a->esz);
2421     desc = FIELD_DP32(desc, PREDDESC, DATA, high_odd);
2422 
2423     tcg_gen_addi_ptr(t_d, cpu_env, pred_full_reg_offset(s, a->rd));
2424     tcg_gen_addi_ptr(t_n, cpu_env, pred_full_reg_offset(s, a->rn));
2425     tcg_gen_addi_ptr(t_m, cpu_env, pred_full_reg_offset(s, a->rm));
2426 
2427     fn(t_d, t_n, t_m, tcg_constant_i32(desc));
2428     return true;
2429 }
2430 
2431 static bool do_perm_pred2(DisasContext *s, arg_rr_esz *a, bool high_odd,
2432                           gen_helper_gvec_2 *fn)
2433 {
2434     if (!sve_access_check(s)) {
2435         return true;
2436     }
2437 
2438     unsigned vsz = pred_full_reg_size(s);
2439     TCGv_ptr t_d = tcg_temp_new_ptr();
2440     TCGv_ptr t_n = tcg_temp_new_ptr();
2441     uint32_t desc = 0;
2442 
2443     tcg_gen_addi_ptr(t_d, cpu_env, pred_full_reg_offset(s, a->rd));
2444     tcg_gen_addi_ptr(t_n, cpu_env, pred_full_reg_offset(s, a->rn));
2445 
2446     desc = FIELD_DP32(desc, PREDDESC, OPRSZ, vsz);
2447     desc = FIELD_DP32(desc, PREDDESC, ESZ, a->esz);
2448     desc = FIELD_DP32(desc, PREDDESC, DATA, high_odd);
2449 
2450     fn(t_d, t_n, tcg_constant_i32(desc));
2451     return true;
2452 }
2453 
2454 TRANS_FEAT(ZIP1_p, aa64_sve, do_perm_pred3, a, 0, gen_helper_sve_zip_p)
2455 TRANS_FEAT(ZIP2_p, aa64_sve, do_perm_pred3, a, 1, gen_helper_sve_zip_p)
2456 TRANS_FEAT(UZP1_p, aa64_sve, do_perm_pred3, a, 0, gen_helper_sve_uzp_p)
2457 TRANS_FEAT(UZP2_p, aa64_sve, do_perm_pred3, a, 1, gen_helper_sve_uzp_p)
2458 TRANS_FEAT(TRN1_p, aa64_sve, do_perm_pred3, a, 0, gen_helper_sve_trn_p)
2459 TRANS_FEAT(TRN2_p, aa64_sve, do_perm_pred3, a, 1, gen_helper_sve_trn_p)
2460 
2461 TRANS_FEAT(REV_p, aa64_sve, do_perm_pred2, a, 0, gen_helper_sve_rev_p)
2462 TRANS_FEAT(PUNPKLO, aa64_sve, do_perm_pred2, a, 0, gen_helper_sve_punpk_p)
2463 TRANS_FEAT(PUNPKHI, aa64_sve, do_perm_pred2, a, 1, gen_helper_sve_punpk_p)
2464 
2465 /*
2466  *** SVE Permute - Interleaving Group
2467  */
2468 
2469 static gen_helper_gvec_3 * const zip_fns[4] = {
2470     gen_helper_sve_zip_b, gen_helper_sve_zip_h,
2471     gen_helper_sve_zip_s, gen_helper_sve_zip_d,
2472 };
2473 TRANS_FEAT(ZIP1_z, aa64_sve, gen_gvec_ool_arg_zzz,
2474            zip_fns[a->esz], a, 0)
2475 TRANS_FEAT(ZIP2_z, aa64_sve, gen_gvec_ool_arg_zzz,
2476            zip_fns[a->esz], a, vec_full_reg_size(s) / 2)
2477 
2478 TRANS_FEAT(ZIP1_q, aa64_sve_f64mm, gen_gvec_ool_arg_zzz,
2479            gen_helper_sve2_zip_q, a, 0)
2480 TRANS_FEAT(ZIP2_q, aa64_sve_f64mm, gen_gvec_ool_arg_zzz,
2481            gen_helper_sve2_zip_q, a,
2482            QEMU_ALIGN_DOWN(vec_full_reg_size(s), 32) / 2)
2483 
2484 static gen_helper_gvec_3 * const uzp_fns[4] = {
2485     gen_helper_sve_uzp_b, gen_helper_sve_uzp_h,
2486     gen_helper_sve_uzp_s, gen_helper_sve_uzp_d,
2487 };
2488 
2489 TRANS_FEAT(UZP1_z, aa64_sve, gen_gvec_ool_arg_zzz,
2490            uzp_fns[a->esz], a, 0)
2491 TRANS_FEAT(UZP2_z, aa64_sve, gen_gvec_ool_arg_zzz,
2492            uzp_fns[a->esz], a, 1 << a->esz)
2493 
2494 TRANS_FEAT(UZP1_q, aa64_sve_f64mm, gen_gvec_ool_arg_zzz,
2495            gen_helper_sve2_uzp_q, a, 0)
2496 TRANS_FEAT(UZP2_q, aa64_sve_f64mm, gen_gvec_ool_arg_zzz,
2497            gen_helper_sve2_uzp_q, a, 16)
2498 
2499 static gen_helper_gvec_3 * const trn_fns[4] = {
2500     gen_helper_sve_trn_b, gen_helper_sve_trn_h,
2501     gen_helper_sve_trn_s, gen_helper_sve_trn_d,
2502 };
2503 
2504 TRANS_FEAT(TRN1_z, aa64_sve, gen_gvec_ool_arg_zzz,
2505            trn_fns[a->esz], a, 0)
2506 TRANS_FEAT(TRN2_z, aa64_sve, gen_gvec_ool_arg_zzz,
2507            trn_fns[a->esz], a, 1 << a->esz)
2508 
2509 TRANS_FEAT(TRN1_q, aa64_sve_f64mm, gen_gvec_ool_arg_zzz,
2510            gen_helper_sve2_trn_q, a, 0)
2511 TRANS_FEAT(TRN2_q, aa64_sve_f64mm, gen_gvec_ool_arg_zzz,
2512            gen_helper_sve2_trn_q, a, 16)
2513 
2514 /*
2515  *** SVE Permute Vector - Predicated Group
2516  */
2517 
2518 static gen_helper_gvec_3 * const compact_fns[4] = {
2519     NULL, NULL, gen_helper_sve_compact_s, gen_helper_sve_compact_d
2520 };
2521 TRANS_FEAT_NONSTREAMING(COMPACT, aa64_sve, gen_gvec_ool_arg_zpz,
2522                         compact_fns[a->esz], a, 0)
2523 
2524 /* Call the helper that computes the ARM LastActiveElement pseudocode
2525  * function, scaled by the element size.  This includes the not found
2526  * indication; e.g. not found for esz=3 is -8.
2527  */
2528 static void find_last_active(DisasContext *s, TCGv_i32 ret, int esz, int pg)
2529 {
2530     /* Predicate sizes may be smaller and cannot use simd_desc.  We cannot
2531      * round up, as we do elsewhere, because we need the exact size.
2532      */
2533     TCGv_ptr t_p = tcg_temp_new_ptr();
2534     unsigned desc = 0;
2535 
2536     desc = FIELD_DP32(desc, PREDDESC, OPRSZ, pred_full_reg_size(s));
2537     desc = FIELD_DP32(desc, PREDDESC, ESZ, esz);
2538 
2539     tcg_gen_addi_ptr(t_p, cpu_env, pred_full_reg_offset(s, pg));
2540 
2541     gen_helper_sve_last_active_element(ret, t_p, tcg_constant_i32(desc));
2542 }
2543 
2544 /* Increment LAST to the offset of the next element in the vector,
2545  * wrapping around to 0.
2546  */
2547 static void incr_last_active(DisasContext *s, TCGv_i32 last, int esz)
2548 {
2549     unsigned vsz = vec_full_reg_size(s);
2550 
2551     tcg_gen_addi_i32(last, last, 1 << esz);
2552     if (is_power_of_2(vsz)) {
2553         tcg_gen_andi_i32(last, last, vsz - 1);
2554     } else {
2555         TCGv_i32 max = tcg_constant_i32(vsz);
2556         TCGv_i32 zero = tcg_constant_i32(0);
2557         tcg_gen_movcond_i32(TCG_COND_GEU, last, last, max, zero, last);
2558     }
2559 }
2560 
2561 /* If LAST < 0, set LAST to the offset of the last element in the vector.  */
2562 static void wrap_last_active(DisasContext *s, TCGv_i32 last, int esz)
2563 {
2564     unsigned vsz = vec_full_reg_size(s);
2565 
2566     if (is_power_of_2(vsz)) {
2567         tcg_gen_andi_i32(last, last, vsz - 1);
2568     } else {
2569         TCGv_i32 max = tcg_constant_i32(vsz - (1 << esz));
2570         TCGv_i32 zero = tcg_constant_i32(0);
2571         tcg_gen_movcond_i32(TCG_COND_LT, last, last, zero, max, last);
2572     }
2573 }
2574 
2575 /* Load an unsigned element of ESZ from BASE+OFS.  */
2576 static TCGv_i64 load_esz(TCGv_ptr base, int ofs, int esz)
2577 {
2578     TCGv_i64 r = tcg_temp_new_i64();
2579 
2580     switch (esz) {
2581     case 0:
2582         tcg_gen_ld8u_i64(r, base, ofs);
2583         break;
2584     case 1:
2585         tcg_gen_ld16u_i64(r, base, ofs);
2586         break;
2587     case 2:
2588         tcg_gen_ld32u_i64(r, base, ofs);
2589         break;
2590     case 3:
2591         tcg_gen_ld_i64(r, base, ofs);
2592         break;
2593     default:
2594         g_assert_not_reached();
2595     }
2596     return r;
2597 }
2598 
2599 /* Load an unsigned element of ESZ from RM[LAST].  */
2600 static TCGv_i64 load_last_active(DisasContext *s, TCGv_i32 last,
2601                                  int rm, int esz)
2602 {
2603     TCGv_ptr p = tcg_temp_new_ptr();
2604 
2605     /* Convert offset into vector into offset into ENV.
2606      * The final adjustment for the vector register base
2607      * is added via constant offset to the load.
2608      */
2609 #if HOST_BIG_ENDIAN
2610     /* Adjust for element ordering.  See vec_reg_offset.  */
2611     if (esz < 3) {
2612         tcg_gen_xori_i32(last, last, 8 - (1 << esz));
2613     }
2614 #endif
2615     tcg_gen_ext_i32_ptr(p, last);
2616     tcg_gen_add_ptr(p, p, cpu_env);
2617 
2618     return load_esz(p, vec_full_reg_offset(s, rm), esz);
2619 }
2620 
2621 /* Compute CLAST for a Zreg.  */
2622 static bool do_clast_vector(DisasContext *s, arg_rprr_esz *a, bool before)
2623 {
2624     TCGv_i32 last;
2625     TCGLabel *over;
2626     TCGv_i64 ele;
2627     unsigned vsz, esz = a->esz;
2628 
2629     if (!sve_access_check(s)) {
2630         return true;
2631     }
2632 
2633     last = tcg_temp_new_i32();
2634     over = gen_new_label();
2635 
2636     find_last_active(s, last, esz, a->pg);
2637 
2638     /* There is of course no movcond for a 2048-bit vector,
2639      * so we must branch over the actual store.
2640      */
2641     tcg_gen_brcondi_i32(TCG_COND_LT, last, 0, over);
2642 
2643     if (!before) {
2644         incr_last_active(s, last, esz);
2645     }
2646 
2647     ele = load_last_active(s, last, a->rm, esz);
2648 
2649     vsz = vec_full_reg_size(s);
2650     tcg_gen_gvec_dup_i64(esz, vec_full_reg_offset(s, a->rd), vsz, vsz, ele);
2651 
2652     /* If this insn used MOVPRFX, we may need a second move.  */
2653     if (a->rd != a->rn) {
2654         TCGLabel *done = gen_new_label();
2655         tcg_gen_br(done);
2656 
2657         gen_set_label(over);
2658         do_mov_z(s, a->rd, a->rn);
2659 
2660         gen_set_label(done);
2661     } else {
2662         gen_set_label(over);
2663     }
2664     return true;
2665 }
2666 
2667 TRANS_FEAT(CLASTA_z, aa64_sve, do_clast_vector, a, false)
2668 TRANS_FEAT(CLASTB_z, aa64_sve, do_clast_vector, a, true)
2669 
2670 /* Compute CLAST for a scalar.  */
2671 static void do_clast_scalar(DisasContext *s, int esz, int pg, int rm,
2672                             bool before, TCGv_i64 reg_val)
2673 {
2674     TCGv_i32 last = tcg_temp_new_i32();
2675     TCGv_i64 ele, cmp;
2676 
2677     find_last_active(s, last, esz, pg);
2678 
2679     /* Extend the original value of last prior to incrementing.  */
2680     cmp = tcg_temp_new_i64();
2681     tcg_gen_ext_i32_i64(cmp, last);
2682 
2683     if (!before) {
2684         incr_last_active(s, last, esz);
2685     }
2686 
2687     /* The conceit here is that while last < 0 indicates not found, after
2688      * adjusting for cpu_env->vfp.zregs[rm], it is still a valid address
2689      * from which we can load garbage.  We then discard the garbage with
2690      * a conditional move.
2691      */
2692     ele = load_last_active(s, last, rm, esz);
2693 
2694     tcg_gen_movcond_i64(TCG_COND_GE, reg_val, cmp, tcg_constant_i64(0),
2695                         ele, reg_val);
2696 }
2697 
2698 /* Compute CLAST for a Vreg.  */
2699 static bool do_clast_fp(DisasContext *s, arg_rpr_esz *a, bool before)
2700 {
2701     if (sve_access_check(s)) {
2702         int esz = a->esz;
2703         int ofs = vec_reg_offset(s, a->rd, 0, esz);
2704         TCGv_i64 reg = load_esz(cpu_env, ofs, esz);
2705 
2706         do_clast_scalar(s, esz, a->pg, a->rn, before, reg);
2707         write_fp_dreg(s, a->rd, reg);
2708     }
2709     return true;
2710 }
2711 
2712 TRANS_FEAT(CLASTA_v, aa64_sve, do_clast_fp, a, false)
2713 TRANS_FEAT(CLASTB_v, aa64_sve, do_clast_fp, a, true)
2714 
2715 /* Compute CLAST for a Xreg.  */
2716 static bool do_clast_general(DisasContext *s, arg_rpr_esz *a, bool before)
2717 {
2718     TCGv_i64 reg;
2719 
2720     if (!sve_access_check(s)) {
2721         return true;
2722     }
2723 
2724     reg = cpu_reg(s, a->rd);
2725     switch (a->esz) {
2726     case 0:
2727         tcg_gen_ext8u_i64(reg, reg);
2728         break;
2729     case 1:
2730         tcg_gen_ext16u_i64(reg, reg);
2731         break;
2732     case 2:
2733         tcg_gen_ext32u_i64(reg, reg);
2734         break;
2735     case 3:
2736         break;
2737     default:
2738         g_assert_not_reached();
2739     }
2740 
2741     do_clast_scalar(s, a->esz, a->pg, a->rn, before, reg);
2742     return true;
2743 }
2744 
2745 TRANS_FEAT(CLASTA_r, aa64_sve, do_clast_general, a, false)
2746 TRANS_FEAT(CLASTB_r, aa64_sve, do_clast_general, a, true)
2747 
2748 /* Compute LAST for a scalar.  */
2749 static TCGv_i64 do_last_scalar(DisasContext *s, int esz,
2750                                int pg, int rm, bool before)
2751 {
2752     TCGv_i32 last = tcg_temp_new_i32();
2753 
2754     find_last_active(s, last, esz, pg);
2755     if (before) {
2756         wrap_last_active(s, last, esz);
2757     } else {
2758         incr_last_active(s, last, esz);
2759     }
2760 
2761     return load_last_active(s, last, rm, esz);
2762 }
2763 
2764 /* Compute LAST for a Vreg.  */
2765 static bool do_last_fp(DisasContext *s, arg_rpr_esz *a, bool before)
2766 {
2767     if (sve_access_check(s)) {
2768         TCGv_i64 val = do_last_scalar(s, a->esz, a->pg, a->rn, before);
2769         write_fp_dreg(s, a->rd, val);
2770     }
2771     return true;
2772 }
2773 
2774 TRANS_FEAT(LASTA_v, aa64_sve, do_last_fp, a, false)
2775 TRANS_FEAT(LASTB_v, aa64_sve, do_last_fp, a, true)
2776 
2777 /* Compute LAST for a Xreg.  */
2778 static bool do_last_general(DisasContext *s, arg_rpr_esz *a, bool before)
2779 {
2780     if (sve_access_check(s)) {
2781         TCGv_i64 val = do_last_scalar(s, a->esz, a->pg, a->rn, before);
2782         tcg_gen_mov_i64(cpu_reg(s, a->rd), val);
2783     }
2784     return true;
2785 }
2786 
2787 TRANS_FEAT(LASTA_r, aa64_sve, do_last_general, a, false)
2788 TRANS_FEAT(LASTB_r, aa64_sve, do_last_general, a, true)
2789 
2790 static bool trans_CPY_m_r(DisasContext *s, arg_rpr_esz *a)
2791 {
2792     if (!dc_isar_feature(aa64_sve, s)) {
2793         return false;
2794     }
2795     if (sve_access_check(s)) {
2796         do_cpy_m(s, a->esz, a->rd, a->rd, a->pg, cpu_reg_sp(s, a->rn));
2797     }
2798     return true;
2799 }
2800 
2801 static bool trans_CPY_m_v(DisasContext *s, arg_rpr_esz *a)
2802 {
2803     if (!dc_isar_feature(aa64_sve, s)) {
2804         return false;
2805     }
2806     if (sve_access_check(s)) {
2807         int ofs = vec_reg_offset(s, a->rn, 0, a->esz);
2808         TCGv_i64 t = load_esz(cpu_env, ofs, a->esz);
2809         do_cpy_m(s, a->esz, a->rd, a->rd, a->pg, t);
2810     }
2811     return true;
2812 }
2813 
2814 static gen_helper_gvec_3 * const revb_fns[4] = {
2815     NULL,                  gen_helper_sve_revb_h,
2816     gen_helper_sve_revb_s, gen_helper_sve_revb_d,
2817 };
2818 TRANS_FEAT(REVB, aa64_sve, gen_gvec_ool_arg_zpz, revb_fns[a->esz], a, 0)
2819 
2820 static gen_helper_gvec_3 * const revh_fns[4] = {
2821     NULL, NULL, gen_helper_sve_revh_s, gen_helper_sve_revh_d,
2822 };
2823 TRANS_FEAT(REVH, aa64_sve, gen_gvec_ool_arg_zpz, revh_fns[a->esz], a, 0)
2824 
2825 TRANS_FEAT(REVW, aa64_sve, gen_gvec_ool_arg_zpz,
2826            a->esz == 3 ? gen_helper_sve_revw_d : NULL, a, 0)
2827 
2828 TRANS_FEAT(REVD, aa64_sme, gen_gvec_ool_arg_zpz, gen_helper_sme_revd_q, a, 0)
2829 
2830 TRANS_FEAT(SPLICE, aa64_sve, gen_gvec_ool_arg_zpzz,
2831            gen_helper_sve_splice, a, a->esz)
2832 
2833 TRANS_FEAT(SPLICE_sve2, aa64_sve2, gen_gvec_ool_zzzp, gen_helper_sve_splice,
2834            a->rd, a->rn, (a->rn + 1) % 32, a->pg, a->esz)
2835 
2836 /*
2837  *** SVE Integer Compare - Vectors Group
2838  */
2839 
2840 static bool do_ppzz_flags(DisasContext *s, arg_rprr_esz *a,
2841                           gen_helper_gvec_flags_4 *gen_fn)
2842 {
2843     TCGv_ptr pd, zn, zm, pg;
2844     unsigned vsz;
2845     TCGv_i32 t;
2846 
2847     if (gen_fn == NULL) {
2848         return false;
2849     }
2850     if (!sve_access_check(s)) {
2851         return true;
2852     }
2853 
2854     vsz = vec_full_reg_size(s);
2855     t = tcg_temp_new_i32();
2856     pd = tcg_temp_new_ptr();
2857     zn = tcg_temp_new_ptr();
2858     zm = tcg_temp_new_ptr();
2859     pg = tcg_temp_new_ptr();
2860 
2861     tcg_gen_addi_ptr(pd, cpu_env, pred_full_reg_offset(s, a->rd));
2862     tcg_gen_addi_ptr(zn, cpu_env, vec_full_reg_offset(s, a->rn));
2863     tcg_gen_addi_ptr(zm, cpu_env, vec_full_reg_offset(s, a->rm));
2864     tcg_gen_addi_ptr(pg, cpu_env, pred_full_reg_offset(s, a->pg));
2865 
2866     gen_fn(t, pd, zn, zm, pg, tcg_constant_i32(simd_desc(vsz, vsz, 0)));
2867 
2868     do_pred_flags(t);
2869     return true;
2870 }
2871 
2872 #define DO_PPZZ(NAME, name) \
2873     static gen_helper_gvec_flags_4 * const name##_ppzz_fns[4] = {       \
2874         gen_helper_sve_##name##_ppzz_b, gen_helper_sve_##name##_ppzz_h, \
2875         gen_helper_sve_##name##_ppzz_s, gen_helper_sve_##name##_ppzz_d, \
2876     };                                                                  \
2877     TRANS_FEAT(NAME##_ppzz, aa64_sve, do_ppzz_flags,                    \
2878                a, name##_ppzz_fns[a->esz])
2879 
2880 DO_PPZZ(CMPEQ, cmpeq)
2881 DO_PPZZ(CMPNE, cmpne)
2882 DO_PPZZ(CMPGT, cmpgt)
2883 DO_PPZZ(CMPGE, cmpge)
2884 DO_PPZZ(CMPHI, cmphi)
2885 DO_PPZZ(CMPHS, cmphs)
2886 
2887 #undef DO_PPZZ
2888 
2889 #define DO_PPZW(NAME, name) \
2890     static gen_helper_gvec_flags_4 * const name##_ppzw_fns[4] = {       \
2891         gen_helper_sve_##name##_ppzw_b, gen_helper_sve_##name##_ppzw_h, \
2892         gen_helper_sve_##name##_ppzw_s, NULL                            \
2893     };                                                                  \
2894     TRANS_FEAT(NAME##_ppzw, aa64_sve, do_ppzz_flags,                    \
2895                a, name##_ppzw_fns[a->esz])
2896 
2897 DO_PPZW(CMPEQ, cmpeq)
2898 DO_PPZW(CMPNE, cmpne)
2899 DO_PPZW(CMPGT, cmpgt)
2900 DO_PPZW(CMPGE, cmpge)
2901 DO_PPZW(CMPHI, cmphi)
2902 DO_PPZW(CMPHS, cmphs)
2903 DO_PPZW(CMPLT, cmplt)
2904 DO_PPZW(CMPLE, cmple)
2905 DO_PPZW(CMPLO, cmplo)
2906 DO_PPZW(CMPLS, cmpls)
2907 
2908 #undef DO_PPZW
2909 
2910 /*
2911  *** SVE Integer Compare - Immediate Groups
2912  */
2913 
2914 static bool do_ppzi_flags(DisasContext *s, arg_rpri_esz *a,
2915                           gen_helper_gvec_flags_3 *gen_fn)
2916 {
2917     TCGv_ptr pd, zn, pg;
2918     unsigned vsz;
2919     TCGv_i32 t;
2920 
2921     if (gen_fn == NULL) {
2922         return false;
2923     }
2924     if (!sve_access_check(s)) {
2925         return true;
2926     }
2927 
2928     vsz = vec_full_reg_size(s);
2929     t = tcg_temp_new_i32();
2930     pd = tcg_temp_new_ptr();
2931     zn = tcg_temp_new_ptr();
2932     pg = tcg_temp_new_ptr();
2933 
2934     tcg_gen_addi_ptr(pd, cpu_env, pred_full_reg_offset(s, a->rd));
2935     tcg_gen_addi_ptr(zn, cpu_env, vec_full_reg_offset(s, a->rn));
2936     tcg_gen_addi_ptr(pg, cpu_env, pred_full_reg_offset(s, a->pg));
2937 
2938     gen_fn(t, pd, zn, pg, tcg_constant_i32(simd_desc(vsz, vsz, a->imm)));
2939 
2940     do_pred_flags(t);
2941     return true;
2942 }
2943 
2944 #define DO_PPZI(NAME, name) \
2945     static gen_helper_gvec_flags_3 * const name##_ppzi_fns[4] = {         \
2946         gen_helper_sve_##name##_ppzi_b, gen_helper_sve_##name##_ppzi_h,   \
2947         gen_helper_sve_##name##_ppzi_s, gen_helper_sve_##name##_ppzi_d,   \
2948     };                                                                    \
2949     TRANS_FEAT(NAME##_ppzi, aa64_sve, do_ppzi_flags, a,                   \
2950                name##_ppzi_fns[a->esz])
2951 
2952 DO_PPZI(CMPEQ, cmpeq)
2953 DO_PPZI(CMPNE, cmpne)
2954 DO_PPZI(CMPGT, cmpgt)
2955 DO_PPZI(CMPGE, cmpge)
2956 DO_PPZI(CMPHI, cmphi)
2957 DO_PPZI(CMPHS, cmphs)
2958 DO_PPZI(CMPLT, cmplt)
2959 DO_PPZI(CMPLE, cmple)
2960 DO_PPZI(CMPLO, cmplo)
2961 DO_PPZI(CMPLS, cmpls)
2962 
2963 #undef DO_PPZI
2964 
2965 /*
2966  *** SVE Partition Break Group
2967  */
2968 
2969 static bool do_brk3(DisasContext *s, arg_rprr_s *a,
2970                     gen_helper_gvec_4 *fn, gen_helper_gvec_flags_4 *fn_s)
2971 {
2972     if (!sve_access_check(s)) {
2973         return true;
2974     }
2975 
2976     unsigned vsz = pred_full_reg_size(s);
2977 
2978     /* Predicate sizes may be smaller and cannot use simd_desc.  */
2979     TCGv_ptr d = tcg_temp_new_ptr();
2980     TCGv_ptr n = tcg_temp_new_ptr();
2981     TCGv_ptr m = tcg_temp_new_ptr();
2982     TCGv_ptr g = tcg_temp_new_ptr();
2983     TCGv_i32 desc = tcg_constant_i32(FIELD_DP32(0, PREDDESC, OPRSZ, vsz));
2984 
2985     tcg_gen_addi_ptr(d, cpu_env, pred_full_reg_offset(s, a->rd));
2986     tcg_gen_addi_ptr(n, cpu_env, pred_full_reg_offset(s, a->rn));
2987     tcg_gen_addi_ptr(m, cpu_env, pred_full_reg_offset(s, a->rm));
2988     tcg_gen_addi_ptr(g, cpu_env, pred_full_reg_offset(s, a->pg));
2989 
2990     if (a->s) {
2991         TCGv_i32 t = tcg_temp_new_i32();
2992         fn_s(t, d, n, m, g, desc);
2993         do_pred_flags(t);
2994     } else {
2995         fn(d, n, m, g, desc);
2996     }
2997     return true;
2998 }
2999 
3000 static bool do_brk2(DisasContext *s, arg_rpr_s *a,
3001                     gen_helper_gvec_3 *fn, gen_helper_gvec_flags_3 *fn_s)
3002 {
3003     if (!sve_access_check(s)) {
3004         return true;
3005     }
3006 
3007     unsigned vsz = pred_full_reg_size(s);
3008 
3009     /* Predicate sizes may be smaller and cannot use simd_desc.  */
3010     TCGv_ptr d = tcg_temp_new_ptr();
3011     TCGv_ptr n = tcg_temp_new_ptr();
3012     TCGv_ptr g = tcg_temp_new_ptr();
3013     TCGv_i32 desc = tcg_constant_i32(FIELD_DP32(0, PREDDESC, OPRSZ, vsz));
3014 
3015     tcg_gen_addi_ptr(d, cpu_env, pred_full_reg_offset(s, a->rd));
3016     tcg_gen_addi_ptr(n, cpu_env, pred_full_reg_offset(s, a->rn));
3017     tcg_gen_addi_ptr(g, cpu_env, pred_full_reg_offset(s, a->pg));
3018 
3019     if (a->s) {
3020         TCGv_i32 t = tcg_temp_new_i32();
3021         fn_s(t, d, n, g, desc);
3022         do_pred_flags(t);
3023     } else {
3024         fn(d, n, g, desc);
3025     }
3026     return true;
3027 }
3028 
3029 TRANS_FEAT(BRKPA, aa64_sve, do_brk3, a,
3030            gen_helper_sve_brkpa, gen_helper_sve_brkpas)
3031 TRANS_FEAT(BRKPB, aa64_sve, do_brk3, a,
3032            gen_helper_sve_brkpb, gen_helper_sve_brkpbs)
3033 
3034 TRANS_FEAT(BRKA_m, aa64_sve, do_brk2, a,
3035            gen_helper_sve_brka_m, gen_helper_sve_brkas_m)
3036 TRANS_FEAT(BRKB_m, aa64_sve, do_brk2, a,
3037            gen_helper_sve_brkb_m, gen_helper_sve_brkbs_m)
3038 
3039 TRANS_FEAT(BRKA_z, aa64_sve, do_brk2, a,
3040            gen_helper_sve_brka_z, gen_helper_sve_brkas_z)
3041 TRANS_FEAT(BRKB_z, aa64_sve, do_brk2, a,
3042            gen_helper_sve_brkb_z, gen_helper_sve_brkbs_z)
3043 
3044 TRANS_FEAT(BRKN, aa64_sve, do_brk2, a,
3045            gen_helper_sve_brkn, gen_helper_sve_brkns)
3046 
3047 /*
3048  *** SVE Predicate Count Group
3049  */
3050 
3051 static void do_cntp(DisasContext *s, TCGv_i64 val, int esz, int pn, int pg)
3052 {
3053     unsigned psz = pred_full_reg_size(s);
3054 
3055     if (psz <= 8) {
3056         uint64_t psz_mask;
3057 
3058         tcg_gen_ld_i64(val, cpu_env, pred_full_reg_offset(s, pn));
3059         if (pn != pg) {
3060             TCGv_i64 g = tcg_temp_new_i64();
3061             tcg_gen_ld_i64(g, cpu_env, pred_full_reg_offset(s, pg));
3062             tcg_gen_and_i64(val, val, g);
3063         }
3064 
3065         /* Reduce the pred_esz_masks value simply to reduce the
3066          * size of the code generated here.
3067          */
3068         psz_mask = MAKE_64BIT_MASK(0, psz * 8);
3069         tcg_gen_andi_i64(val, val, pred_esz_masks[esz] & psz_mask);
3070 
3071         tcg_gen_ctpop_i64(val, val);
3072     } else {
3073         TCGv_ptr t_pn = tcg_temp_new_ptr();
3074         TCGv_ptr t_pg = tcg_temp_new_ptr();
3075         unsigned desc = 0;
3076 
3077         desc = FIELD_DP32(desc, PREDDESC, OPRSZ, psz);
3078         desc = FIELD_DP32(desc, PREDDESC, ESZ, esz);
3079 
3080         tcg_gen_addi_ptr(t_pn, cpu_env, pred_full_reg_offset(s, pn));
3081         tcg_gen_addi_ptr(t_pg, cpu_env, pred_full_reg_offset(s, pg));
3082 
3083         gen_helper_sve_cntp(val, t_pn, t_pg, tcg_constant_i32(desc));
3084     }
3085 }
3086 
3087 static bool trans_CNTP(DisasContext *s, arg_CNTP *a)
3088 {
3089     if (!dc_isar_feature(aa64_sve, s)) {
3090         return false;
3091     }
3092     if (sve_access_check(s)) {
3093         do_cntp(s, cpu_reg(s, a->rd), a->esz, a->rn, a->pg);
3094     }
3095     return true;
3096 }
3097 
3098 static bool trans_INCDECP_r(DisasContext *s, arg_incdec_pred *a)
3099 {
3100     if (!dc_isar_feature(aa64_sve, s)) {
3101         return false;
3102     }
3103     if (sve_access_check(s)) {
3104         TCGv_i64 reg = cpu_reg(s, a->rd);
3105         TCGv_i64 val = tcg_temp_new_i64();
3106 
3107         do_cntp(s, val, a->esz, a->pg, a->pg);
3108         if (a->d) {
3109             tcg_gen_sub_i64(reg, reg, val);
3110         } else {
3111             tcg_gen_add_i64(reg, reg, val);
3112         }
3113     }
3114     return true;
3115 }
3116 
3117 static bool trans_INCDECP_z(DisasContext *s, arg_incdec2_pred *a)
3118 {
3119     if (a->esz == 0 || !dc_isar_feature(aa64_sve, s)) {
3120         return false;
3121     }
3122     if (sve_access_check(s)) {
3123         unsigned vsz = vec_full_reg_size(s);
3124         TCGv_i64 val = tcg_temp_new_i64();
3125         GVecGen2sFn *gvec_fn = a->d ? tcg_gen_gvec_subs : tcg_gen_gvec_adds;
3126 
3127         do_cntp(s, val, a->esz, a->pg, a->pg);
3128         gvec_fn(a->esz, vec_full_reg_offset(s, a->rd),
3129                 vec_full_reg_offset(s, a->rn), val, vsz, vsz);
3130     }
3131     return true;
3132 }
3133 
3134 static bool trans_SINCDECP_r_32(DisasContext *s, arg_incdec_pred *a)
3135 {
3136     if (!dc_isar_feature(aa64_sve, s)) {
3137         return false;
3138     }
3139     if (sve_access_check(s)) {
3140         TCGv_i64 reg = cpu_reg(s, a->rd);
3141         TCGv_i64 val = tcg_temp_new_i64();
3142 
3143         do_cntp(s, val, a->esz, a->pg, a->pg);
3144         do_sat_addsub_32(reg, val, a->u, a->d);
3145     }
3146     return true;
3147 }
3148 
3149 static bool trans_SINCDECP_r_64(DisasContext *s, arg_incdec_pred *a)
3150 {
3151     if (!dc_isar_feature(aa64_sve, s)) {
3152         return false;
3153     }
3154     if (sve_access_check(s)) {
3155         TCGv_i64 reg = cpu_reg(s, a->rd);
3156         TCGv_i64 val = tcg_temp_new_i64();
3157 
3158         do_cntp(s, val, a->esz, a->pg, a->pg);
3159         do_sat_addsub_64(reg, val, a->u, a->d);
3160     }
3161     return true;
3162 }
3163 
3164 static bool trans_SINCDECP_z(DisasContext *s, arg_incdec2_pred *a)
3165 {
3166     if (a->esz == 0 || !dc_isar_feature(aa64_sve, s)) {
3167         return false;
3168     }
3169     if (sve_access_check(s)) {
3170         TCGv_i64 val = tcg_temp_new_i64();
3171         do_cntp(s, val, a->esz, a->pg, a->pg);
3172         do_sat_addsub_vec(s, a->esz, a->rd, a->rn, val, a->u, a->d);
3173     }
3174     return true;
3175 }
3176 
3177 /*
3178  *** SVE Integer Compare Scalars Group
3179  */
3180 
3181 static bool trans_CTERM(DisasContext *s, arg_CTERM *a)
3182 {
3183     if (!dc_isar_feature(aa64_sve, s)) {
3184         return false;
3185     }
3186     if (!sve_access_check(s)) {
3187         return true;
3188     }
3189 
3190     TCGCond cond = (a->ne ? TCG_COND_NE : TCG_COND_EQ);
3191     TCGv_i64 rn = read_cpu_reg(s, a->rn, a->sf);
3192     TCGv_i64 rm = read_cpu_reg(s, a->rm, a->sf);
3193     TCGv_i64 cmp = tcg_temp_new_i64();
3194 
3195     tcg_gen_setcond_i64(cond, cmp, rn, rm);
3196     tcg_gen_extrl_i64_i32(cpu_NF, cmp);
3197 
3198     /* VF = !NF & !CF.  */
3199     tcg_gen_xori_i32(cpu_VF, cpu_NF, 1);
3200     tcg_gen_andc_i32(cpu_VF, cpu_VF, cpu_CF);
3201 
3202     /* Both NF and VF actually look at bit 31.  */
3203     tcg_gen_neg_i32(cpu_NF, cpu_NF);
3204     tcg_gen_neg_i32(cpu_VF, cpu_VF);
3205     return true;
3206 }
3207 
3208 static bool trans_WHILE(DisasContext *s, arg_WHILE *a)
3209 {
3210     TCGv_i64 op0, op1, t0, t1, tmax;
3211     TCGv_i32 t2;
3212     TCGv_ptr ptr;
3213     unsigned vsz = vec_full_reg_size(s);
3214     unsigned desc = 0;
3215     TCGCond cond;
3216     uint64_t maxval;
3217     /* Note that GE/HS has a->eq == 0 and GT/HI has a->eq == 1. */
3218     bool eq = a->eq == a->lt;
3219 
3220     /* The greater-than conditions are all SVE2. */
3221     if (a->lt
3222         ? !dc_isar_feature(aa64_sve, s)
3223         : !dc_isar_feature(aa64_sve2, s)) {
3224         return false;
3225     }
3226     if (!sve_access_check(s)) {
3227         return true;
3228     }
3229 
3230     op0 = read_cpu_reg(s, a->rn, 1);
3231     op1 = read_cpu_reg(s, a->rm, 1);
3232 
3233     if (!a->sf) {
3234         if (a->u) {
3235             tcg_gen_ext32u_i64(op0, op0);
3236             tcg_gen_ext32u_i64(op1, op1);
3237         } else {
3238             tcg_gen_ext32s_i64(op0, op0);
3239             tcg_gen_ext32s_i64(op1, op1);
3240         }
3241     }
3242 
3243     /* For the helper, compress the different conditions into a computation
3244      * of how many iterations for which the condition is true.
3245      */
3246     t0 = tcg_temp_new_i64();
3247     t1 = tcg_temp_new_i64();
3248 
3249     if (a->lt) {
3250         tcg_gen_sub_i64(t0, op1, op0);
3251         if (a->u) {
3252             maxval = a->sf ? UINT64_MAX : UINT32_MAX;
3253             cond = eq ? TCG_COND_LEU : TCG_COND_LTU;
3254         } else {
3255             maxval = a->sf ? INT64_MAX : INT32_MAX;
3256             cond = eq ? TCG_COND_LE : TCG_COND_LT;
3257         }
3258     } else {
3259         tcg_gen_sub_i64(t0, op0, op1);
3260         if (a->u) {
3261             maxval = 0;
3262             cond = eq ? TCG_COND_GEU : TCG_COND_GTU;
3263         } else {
3264             maxval = a->sf ? INT64_MIN : INT32_MIN;
3265             cond = eq ? TCG_COND_GE : TCG_COND_GT;
3266         }
3267     }
3268 
3269     tmax = tcg_constant_i64(vsz >> a->esz);
3270     if (eq) {
3271         /* Equality means one more iteration.  */
3272         tcg_gen_addi_i64(t0, t0, 1);
3273 
3274         /*
3275          * For the less-than while, if op1 is maxval (and the only time
3276          * the addition above could overflow), then we produce an all-true
3277          * predicate by setting the count to the vector length.  This is
3278          * because the pseudocode is described as an increment + compare
3279          * loop, and the maximum integer would always compare true.
3280          * Similarly, the greater-than while has the same issue with the
3281          * minimum integer due to the decrement + compare loop.
3282          */
3283         tcg_gen_movi_i64(t1, maxval);
3284         tcg_gen_movcond_i64(TCG_COND_EQ, t0, op1, t1, tmax, t0);
3285     }
3286 
3287     /* Bound to the maximum.  */
3288     tcg_gen_umin_i64(t0, t0, tmax);
3289 
3290     /* Set the count to zero if the condition is false.  */
3291     tcg_gen_movi_i64(t1, 0);
3292     tcg_gen_movcond_i64(cond, t0, op0, op1, t0, t1);
3293 
3294     /* Since we're bounded, pass as a 32-bit type.  */
3295     t2 = tcg_temp_new_i32();
3296     tcg_gen_extrl_i64_i32(t2, t0);
3297 
3298     /* Scale elements to bits.  */
3299     tcg_gen_shli_i32(t2, t2, a->esz);
3300 
3301     desc = FIELD_DP32(desc, PREDDESC, OPRSZ, vsz / 8);
3302     desc = FIELD_DP32(desc, PREDDESC, ESZ, a->esz);
3303 
3304     ptr = tcg_temp_new_ptr();
3305     tcg_gen_addi_ptr(ptr, cpu_env, pred_full_reg_offset(s, a->rd));
3306 
3307     if (a->lt) {
3308         gen_helper_sve_whilel(t2, ptr, t2, tcg_constant_i32(desc));
3309     } else {
3310         gen_helper_sve_whileg(t2, ptr, t2, tcg_constant_i32(desc));
3311     }
3312     do_pred_flags(t2);
3313     return true;
3314 }
3315 
3316 static bool trans_WHILE_ptr(DisasContext *s, arg_WHILE_ptr *a)
3317 {
3318     TCGv_i64 op0, op1, diff, t1, tmax;
3319     TCGv_i32 t2;
3320     TCGv_ptr ptr;
3321     unsigned vsz = vec_full_reg_size(s);
3322     unsigned desc = 0;
3323 
3324     if (!dc_isar_feature(aa64_sve2, s)) {
3325         return false;
3326     }
3327     if (!sve_access_check(s)) {
3328         return true;
3329     }
3330 
3331     op0 = read_cpu_reg(s, a->rn, 1);
3332     op1 = read_cpu_reg(s, a->rm, 1);
3333 
3334     tmax = tcg_constant_i64(vsz);
3335     diff = tcg_temp_new_i64();
3336 
3337     if (a->rw) {
3338         /* WHILERW */
3339         /* diff = abs(op1 - op0), noting that op0/1 are unsigned. */
3340         t1 = tcg_temp_new_i64();
3341         tcg_gen_sub_i64(diff, op0, op1);
3342         tcg_gen_sub_i64(t1, op1, op0);
3343         tcg_gen_movcond_i64(TCG_COND_GEU, diff, op0, op1, diff, t1);
3344         /* Round down to a multiple of ESIZE.  */
3345         tcg_gen_andi_i64(diff, diff, -1 << a->esz);
3346         /* If op1 == op0, diff == 0, and the condition is always true. */
3347         tcg_gen_movcond_i64(TCG_COND_EQ, diff, op0, op1, tmax, diff);
3348     } else {
3349         /* WHILEWR */
3350         tcg_gen_sub_i64(diff, op1, op0);
3351         /* Round down to a multiple of ESIZE.  */
3352         tcg_gen_andi_i64(diff, diff, -1 << a->esz);
3353         /* If op0 >= op1, diff <= 0, the condition is always true. */
3354         tcg_gen_movcond_i64(TCG_COND_GEU, diff, op0, op1, tmax, diff);
3355     }
3356 
3357     /* Bound to the maximum.  */
3358     tcg_gen_umin_i64(diff, diff, tmax);
3359 
3360     /* Since we're bounded, pass as a 32-bit type.  */
3361     t2 = tcg_temp_new_i32();
3362     tcg_gen_extrl_i64_i32(t2, diff);
3363 
3364     desc = FIELD_DP32(desc, PREDDESC, OPRSZ, vsz / 8);
3365     desc = FIELD_DP32(desc, PREDDESC, ESZ, a->esz);
3366 
3367     ptr = tcg_temp_new_ptr();
3368     tcg_gen_addi_ptr(ptr, cpu_env, pred_full_reg_offset(s, a->rd));
3369 
3370     gen_helper_sve_whilel(t2, ptr, t2, tcg_constant_i32(desc));
3371     do_pred_flags(t2);
3372     return true;
3373 }
3374 
3375 /*
3376  *** SVE Integer Wide Immediate - Unpredicated Group
3377  */
3378 
3379 static bool trans_FDUP(DisasContext *s, arg_FDUP *a)
3380 {
3381     if (a->esz == 0 || !dc_isar_feature(aa64_sve, s)) {
3382         return false;
3383     }
3384     if (sve_access_check(s)) {
3385         unsigned vsz = vec_full_reg_size(s);
3386         int dofs = vec_full_reg_offset(s, a->rd);
3387         uint64_t imm;
3388 
3389         /* Decode the VFP immediate.  */
3390         imm = vfp_expand_imm(a->esz, a->imm);
3391         tcg_gen_gvec_dup_imm(a->esz, dofs, vsz, vsz, imm);
3392     }
3393     return true;
3394 }
3395 
3396 static bool trans_DUP_i(DisasContext *s, arg_DUP_i *a)
3397 {
3398     if (!dc_isar_feature(aa64_sve, s)) {
3399         return false;
3400     }
3401     if (sve_access_check(s)) {
3402         unsigned vsz = vec_full_reg_size(s);
3403         int dofs = vec_full_reg_offset(s, a->rd);
3404         tcg_gen_gvec_dup_imm(a->esz, dofs, vsz, vsz, a->imm);
3405     }
3406     return true;
3407 }
3408 
3409 TRANS_FEAT(ADD_zzi, aa64_sve, gen_gvec_fn_arg_zzi, tcg_gen_gvec_addi, a)
3410 
3411 static bool trans_SUB_zzi(DisasContext *s, arg_rri_esz *a)
3412 {
3413     a->imm = -a->imm;
3414     return trans_ADD_zzi(s, a);
3415 }
3416 
3417 static bool trans_SUBR_zzi(DisasContext *s, arg_rri_esz *a)
3418 {
3419     static const TCGOpcode vecop_list[] = { INDEX_op_sub_vec, 0 };
3420     static const GVecGen2s op[4] = {
3421         { .fni8 = tcg_gen_vec_sub8_i64,
3422           .fniv = tcg_gen_sub_vec,
3423           .fno = gen_helper_sve_subri_b,
3424           .opt_opc = vecop_list,
3425           .vece = MO_8,
3426           .scalar_first = true },
3427         { .fni8 = tcg_gen_vec_sub16_i64,
3428           .fniv = tcg_gen_sub_vec,
3429           .fno = gen_helper_sve_subri_h,
3430           .opt_opc = vecop_list,
3431           .vece = MO_16,
3432           .scalar_first = true },
3433         { .fni4 = tcg_gen_sub_i32,
3434           .fniv = tcg_gen_sub_vec,
3435           .fno = gen_helper_sve_subri_s,
3436           .opt_opc = vecop_list,
3437           .vece = MO_32,
3438           .scalar_first = true },
3439         { .fni8 = tcg_gen_sub_i64,
3440           .fniv = tcg_gen_sub_vec,
3441           .fno = gen_helper_sve_subri_d,
3442           .opt_opc = vecop_list,
3443           .prefer_i64 = TCG_TARGET_REG_BITS == 64,
3444           .vece = MO_64,
3445           .scalar_first = true }
3446     };
3447 
3448     if (!dc_isar_feature(aa64_sve, s)) {
3449         return false;
3450     }
3451     if (sve_access_check(s)) {
3452         unsigned vsz = vec_full_reg_size(s);
3453         tcg_gen_gvec_2s(vec_full_reg_offset(s, a->rd),
3454                         vec_full_reg_offset(s, a->rn),
3455                         vsz, vsz, tcg_constant_i64(a->imm), &op[a->esz]);
3456     }
3457     return true;
3458 }
3459 
3460 TRANS_FEAT(MUL_zzi, aa64_sve, gen_gvec_fn_arg_zzi, tcg_gen_gvec_muli, a)
3461 
3462 static bool do_zzi_sat(DisasContext *s, arg_rri_esz *a, bool u, bool d)
3463 {
3464     if (sve_access_check(s)) {
3465         do_sat_addsub_vec(s, a->esz, a->rd, a->rn,
3466                           tcg_constant_i64(a->imm), u, d);
3467     }
3468     return true;
3469 }
3470 
3471 TRANS_FEAT(SQADD_zzi, aa64_sve, do_zzi_sat, a, false, false)
3472 TRANS_FEAT(UQADD_zzi, aa64_sve, do_zzi_sat, a, true, false)
3473 TRANS_FEAT(SQSUB_zzi, aa64_sve, do_zzi_sat, a, false, true)
3474 TRANS_FEAT(UQSUB_zzi, aa64_sve, do_zzi_sat, a, true, true)
3475 
3476 static bool do_zzi_ool(DisasContext *s, arg_rri_esz *a, gen_helper_gvec_2i *fn)
3477 {
3478     if (sve_access_check(s)) {
3479         unsigned vsz = vec_full_reg_size(s);
3480         tcg_gen_gvec_2i_ool(vec_full_reg_offset(s, a->rd),
3481                             vec_full_reg_offset(s, a->rn),
3482                             tcg_constant_i64(a->imm), vsz, vsz, 0, fn);
3483     }
3484     return true;
3485 }
3486 
3487 #define DO_ZZI(NAME, name) \
3488     static gen_helper_gvec_2i * const name##i_fns[4] = {                \
3489         gen_helper_sve_##name##i_b, gen_helper_sve_##name##i_h,         \
3490         gen_helper_sve_##name##i_s, gen_helper_sve_##name##i_d,         \
3491     };                                                                  \
3492     TRANS_FEAT(NAME##_zzi, aa64_sve, do_zzi_ool, a, name##i_fns[a->esz])
3493 
3494 DO_ZZI(SMAX, smax)
3495 DO_ZZI(UMAX, umax)
3496 DO_ZZI(SMIN, smin)
3497 DO_ZZI(UMIN, umin)
3498 
3499 #undef DO_ZZI
3500 
3501 static gen_helper_gvec_4 * const dot_fns[2][2] = {
3502     { gen_helper_gvec_sdot_b, gen_helper_gvec_sdot_h },
3503     { gen_helper_gvec_udot_b, gen_helper_gvec_udot_h }
3504 };
3505 TRANS_FEAT(DOT_zzzz, aa64_sve, gen_gvec_ool_zzzz,
3506            dot_fns[a->u][a->sz], a->rd, a->rn, a->rm, a->ra, 0)
3507 
3508 /*
3509  * SVE Multiply - Indexed
3510  */
3511 
3512 TRANS_FEAT(SDOT_zzxw_s, aa64_sve, gen_gvec_ool_arg_zzxz,
3513            gen_helper_gvec_sdot_idx_b, a)
3514 TRANS_FEAT(SDOT_zzxw_d, aa64_sve, gen_gvec_ool_arg_zzxz,
3515            gen_helper_gvec_sdot_idx_h, a)
3516 TRANS_FEAT(UDOT_zzxw_s, aa64_sve, gen_gvec_ool_arg_zzxz,
3517            gen_helper_gvec_udot_idx_b, a)
3518 TRANS_FEAT(UDOT_zzxw_d, aa64_sve, gen_gvec_ool_arg_zzxz,
3519            gen_helper_gvec_udot_idx_h, a)
3520 
3521 TRANS_FEAT(SUDOT_zzxw_s, aa64_sve_i8mm, gen_gvec_ool_arg_zzxz,
3522            gen_helper_gvec_sudot_idx_b, a)
3523 TRANS_FEAT(USDOT_zzxw_s, aa64_sve_i8mm, gen_gvec_ool_arg_zzxz,
3524            gen_helper_gvec_usdot_idx_b, a)
3525 
3526 #define DO_SVE2_RRX(NAME, FUNC) \
3527     TRANS_FEAT(NAME, aa64_sve, gen_gvec_ool_zzz, FUNC,          \
3528                a->rd, a->rn, a->rm, a->index)
3529 
3530 DO_SVE2_RRX(MUL_zzx_h, gen_helper_gvec_mul_idx_h)
3531 DO_SVE2_RRX(MUL_zzx_s, gen_helper_gvec_mul_idx_s)
3532 DO_SVE2_RRX(MUL_zzx_d, gen_helper_gvec_mul_idx_d)
3533 
3534 DO_SVE2_RRX(SQDMULH_zzx_h, gen_helper_sve2_sqdmulh_idx_h)
3535 DO_SVE2_RRX(SQDMULH_zzx_s, gen_helper_sve2_sqdmulh_idx_s)
3536 DO_SVE2_RRX(SQDMULH_zzx_d, gen_helper_sve2_sqdmulh_idx_d)
3537 
3538 DO_SVE2_RRX(SQRDMULH_zzx_h, gen_helper_sve2_sqrdmulh_idx_h)
3539 DO_SVE2_RRX(SQRDMULH_zzx_s, gen_helper_sve2_sqrdmulh_idx_s)
3540 DO_SVE2_RRX(SQRDMULH_zzx_d, gen_helper_sve2_sqrdmulh_idx_d)
3541 
3542 #undef DO_SVE2_RRX
3543 
3544 #define DO_SVE2_RRX_TB(NAME, FUNC, TOP) \
3545     TRANS_FEAT(NAME, aa64_sve, gen_gvec_ool_zzz, FUNC,          \
3546                a->rd, a->rn, a->rm, (a->index << 1) | TOP)
3547 
3548 DO_SVE2_RRX_TB(SQDMULLB_zzx_s, gen_helper_sve2_sqdmull_idx_s, false)
3549 DO_SVE2_RRX_TB(SQDMULLB_zzx_d, gen_helper_sve2_sqdmull_idx_d, false)
3550 DO_SVE2_RRX_TB(SQDMULLT_zzx_s, gen_helper_sve2_sqdmull_idx_s, true)
3551 DO_SVE2_RRX_TB(SQDMULLT_zzx_d, gen_helper_sve2_sqdmull_idx_d, true)
3552 
3553 DO_SVE2_RRX_TB(SMULLB_zzx_s, gen_helper_sve2_smull_idx_s, false)
3554 DO_SVE2_RRX_TB(SMULLB_zzx_d, gen_helper_sve2_smull_idx_d, false)
3555 DO_SVE2_RRX_TB(SMULLT_zzx_s, gen_helper_sve2_smull_idx_s, true)
3556 DO_SVE2_RRX_TB(SMULLT_zzx_d, gen_helper_sve2_smull_idx_d, true)
3557 
3558 DO_SVE2_RRX_TB(UMULLB_zzx_s, gen_helper_sve2_umull_idx_s, false)
3559 DO_SVE2_RRX_TB(UMULLB_zzx_d, gen_helper_sve2_umull_idx_d, false)
3560 DO_SVE2_RRX_TB(UMULLT_zzx_s, gen_helper_sve2_umull_idx_s, true)
3561 DO_SVE2_RRX_TB(UMULLT_zzx_d, gen_helper_sve2_umull_idx_d, true)
3562 
3563 #undef DO_SVE2_RRX_TB
3564 
3565 #define DO_SVE2_RRXR(NAME, FUNC) \
3566     TRANS_FEAT(NAME, aa64_sve2, gen_gvec_ool_arg_zzxz, FUNC, a)
3567 
3568 DO_SVE2_RRXR(MLA_zzxz_h, gen_helper_gvec_mla_idx_h)
3569 DO_SVE2_RRXR(MLA_zzxz_s, gen_helper_gvec_mla_idx_s)
3570 DO_SVE2_RRXR(MLA_zzxz_d, gen_helper_gvec_mla_idx_d)
3571 
3572 DO_SVE2_RRXR(MLS_zzxz_h, gen_helper_gvec_mls_idx_h)
3573 DO_SVE2_RRXR(MLS_zzxz_s, gen_helper_gvec_mls_idx_s)
3574 DO_SVE2_RRXR(MLS_zzxz_d, gen_helper_gvec_mls_idx_d)
3575 
3576 DO_SVE2_RRXR(SQRDMLAH_zzxz_h, gen_helper_sve2_sqrdmlah_idx_h)
3577 DO_SVE2_RRXR(SQRDMLAH_zzxz_s, gen_helper_sve2_sqrdmlah_idx_s)
3578 DO_SVE2_RRXR(SQRDMLAH_zzxz_d, gen_helper_sve2_sqrdmlah_idx_d)
3579 
3580 DO_SVE2_RRXR(SQRDMLSH_zzxz_h, gen_helper_sve2_sqrdmlsh_idx_h)
3581 DO_SVE2_RRXR(SQRDMLSH_zzxz_s, gen_helper_sve2_sqrdmlsh_idx_s)
3582 DO_SVE2_RRXR(SQRDMLSH_zzxz_d, gen_helper_sve2_sqrdmlsh_idx_d)
3583 
3584 #undef DO_SVE2_RRXR
3585 
3586 #define DO_SVE2_RRXR_TB(NAME, FUNC, TOP) \
3587     TRANS_FEAT(NAME, aa64_sve2, gen_gvec_ool_zzzz, FUNC,        \
3588                a->rd, a->rn, a->rm, a->ra, (a->index << 1) | TOP)
3589 
3590 DO_SVE2_RRXR_TB(SQDMLALB_zzxw_s, gen_helper_sve2_sqdmlal_idx_s, false)
3591 DO_SVE2_RRXR_TB(SQDMLALB_zzxw_d, gen_helper_sve2_sqdmlal_idx_d, false)
3592 DO_SVE2_RRXR_TB(SQDMLALT_zzxw_s, gen_helper_sve2_sqdmlal_idx_s, true)
3593 DO_SVE2_RRXR_TB(SQDMLALT_zzxw_d, gen_helper_sve2_sqdmlal_idx_d, true)
3594 
3595 DO_SVE2_RRXR_TB(SQDMLSLB_zzxw_s, gen_helper_sve2_sqdmlsl_idx_s, false)
3596 DO_SVE2_RRXR_TB(SQDMLSLB_zzxw_d, gen_helper_sve2_sqdmlsl_idx_d, false)
3597 DO_SVE2_RRXR_TB(SQDMLSLT_zzxw_s, gen_helper_sve2_sqdmlsl_idx_s, true)
3598 DO_SVE2_RRXR_TB(SQDMLSLT_zzxw_d, gen_helper_sve2_sqdmlsl_idx_d, true)
3599 
3600 DO_SVE2_RRXR_TB(SMLALB_zzxw_s, gen_helper_sve2_smlal_idx_s, false)
3601 DO_SVE2_RRXR_TB(SMLALB_zzxw_d, gen_helper_sve2_smlal_idx_d, false)
3602 DO_SVE2_RRXR_TB(SMLALT_zzxw_s, gen_helper_sve2_smlal_idx_s, true)
3603 DO_SVE2_RRXR_TB(SMLALT_zzxw_d, gen_helper_sve2_smlal_idx_d, true)
3604 
3605 DO_SVE2_RRXR_TB(UMLALB_zzxw_s, gen_helper_sve2_umlal_idx_s, false)
3606 DO_SVE2_RRXR_TB(UMLALB_zzxw_d, gen_helper_sve2_umlal_idx_d, false)
3607 DO_SVE2_RRXR_TB(UMLALT_zzxw_s, gen_helper_sve2_umlal_idx_s, true)
3608 DO_SVE2_RRXR_TB(UMLALT_zzxw_d, gen_helper_sve2_umlal_idx_d, true)
3609 
3610 DO_SVE2_RRXR_TB(SMLSLB_zzxw_s, gen_helper_sve2_smlsl_idx_s, false)
3611 DO_SVE2_RRXR_TB(SMLSLB_zzxw_d, gen_helper_sve2_smlsl_idx_d, false)
3612 DO_SVE2_RRXR_TB(SMLSLT_zzxw_s, gen_helper_sve2_smlsl_idx_s, true)
3613 DO_SVE2_RRXR_TB(SMLSLT_zzxw_d, gen_helper_sve2_smlsl_idx_d, true)
3614 
3615 DO_SVE2_RRXR_TB(UMLSLB_zzxw_s, gen_helper_sve2_umlsl_idx_s, false)
3616 DO_SVE2_RRXR_TB(UMLSLB_zzxw_d, gen_helper_sve2_umlsl_idx_d, false)
3617 DO_SVE2_RRXR_TB(UMLSLT_zzxw_s, gen_helper_sve2_umlsl_idx_s, true)
3618 DO_SVE2_RRXR_TB(UMLSLT_zzxw_d, gen_helper_sve2_umlsl_idx_d, true)
3619 
3620 #undef DO_SVE2_RRXR_TB
3621 
3622 #define DO_SVE2_RRXR_ROT(NAME, FUNC) \
3623     TRANS_FEAT(NAME, aa64_sve2, gen_gvec_ool_zzzz, FUNC,           \
3624                a->rd, a->rn, a->rm, a->ra, (a->index << 2) | a->rot)
3625 
3626 DO_SVE2_RRXR_ROT(CMLA_zzxz_h, gen_helper_sve2_cmla_idx_h)
3627 DO_SVE2_RRXR_ROT(CMLA_zzxz_s, gen_helper_sve2_cmla_idx_s)
3628 
3629 DO_SVE2_RRXR_ROT(SQRDCMLAH_zzxz_h, gen_helper_sve2_sqrdcmlah_idx_h)
3630 DO_SVE2_RRXR_ROT(SQRDCMLAH_zzxz_s, gen_helper_sve2_sqrdcmlah_idx_s)
3631 
3632 DO_SVE2_RRXR_ROT(CDOT_zzxw_s, gen_helper_sve2_cdot_idx_s)
3633 DO_SVE2_RRXR_ROT(CDOT_zzxw_d, gen_helper_sve2_cdot_idx_d)
3634 
3635 #undef DO_SVE2_RRXR_ROT
3636 
3637 /*
3638  *** SVE Floating Point Multiply-Add Indexed Group
3639  */
3640 
3641 static bool do_FMLA_zzxz(DisasContext *s, arg_rrxr_esz *a, bool sub)
3642 {
3643     static gen_helper_gvec_4_ptr * const fns[4] = {
3644         NULL,
3645         gen_helper_gvec_fmla_idx_h,
3646         gen_helper_gvec_fmla_idx_s,
3647         gen_helper_gvec_fmla_idx_d,
3648     };
3649     return gen_gvec_fpst_zzzz(s, fns[a->esz], a->rd, a->rn, a->rm, a->ra,
3650                               (a->index << 1) | sub,
3651                               a->esz == MO_16 ? FPST_FPCR_F16 : FPST_FPCR);
3652 }
3653 
3654 TRANS_FEAT(FMLA_zzxz, aa64_sve, do_FMLA_zzxz, a, false)
3655 TRANS_FEAT(FMLS_zzxz, aa64_sve, do_FMLA_zzxz, a, true)
3656 
3657 /*
3658  *** SVE Floating Point Multiply Indexed Group
3659  */
3660 
3661 static gen_helper_gvec_3_ptr * const fmul_idx_fns[4] = {
3662     NULL,                       gen_helper_gvec_fmul_idx_h,
3663     gen_helper_gvec_fmul_idx_s, gen_helper_gvec_fmul_idx_d,
3664 };
3665 TRANS_FEAT(FMUL_zzx, aa64_sve, gen_gvec_fpst_zzz,
3666            fmul_idx_fns[a->esz], a->rd, a->rn, a->rm, a->index,
3667            a->esz == MO_16 ? FPST_FPCR_F16 : FPST_FPCR)
3668 
3669 /*
3670  *** SVE Floating Point Fast Reduction Group
3671  */
3672 
3673 typedef void gen_helper_fp_reduce(TCGv_i64, TCGv_ptr, TCGv_ptr,
3674                                   TCGv_ptr, TCGv_i32);
3675 
3676 static bool do_reduce(DisasContext *s, arg_rpr_esz *a,
3677                       gen_helper_fp_reduce *fn)
3678 {
3679     unsigned vsz, p2vsz;
3680     TCGv_i32 t_desc;
3681     TCGv_ptr t_zn, t_pg, status;
3682     TCGv_i64 temp;
3683 
3684     if (fn == NULL) {
3685         return false;
3686     }
3687     if (!sve_access_check(s)) {
3688         return true;
3689     }
3690 
3691     vsz = vec_full_reg_size(s);
3692     p2vsz = pow2ceil(vsz);
3693     t_desc = tcg_constant_i32(simd_desc(vsz, vsz, p2vsz));
3694     temp = tcg_temp_new_i64();
3695     t_zn = tcg_temp_new_ptr();
3696     t_pg = tcg_temp_new_ptr();
3697 
3698     tcg_gen_addi_ptr(t_zn, cpu_env, vec_full_reg_offset(s, a->rn));
3699     tcg_gen_addi_ptr(t_pg, cpu_env, pred_full_reg_offset(s, a->pg));
3700     status = fpstatus_ptr(a->esz == MO_16 ? FPST_FPCR_F16 : FPST_FPCR);
3701 
3702     fn(temp, t_zn, t_pg, status, t_desc);
3703 
3704     write_fp_dreg(s, a->rd, temp);
3705     return true;
3706 }
3707 
3708 #define DO_VPZ(NAME, name) \
3709     static gen_helper_fp_reduce * const name##_fns[4] = {                \
3710         NULL,                      gen_helper_sve_##name##_h,            \
3711         gen_helper_sve_##name##_s, gen_helper_sve_##name##_d,            \
3712     };                                                                   \
3713     TRANS_FEAT(NAME, aa64_sve, do_reduce, a, name##_fns[a->esz])
3714 
3715 DO_VPZ(FADDV, faddv)
3716 DO_VPZ(FMINNMV, fminnmv)
3717 DO_VPZ(FMAXNMV, fmaxnmv)
3718 DO_VPZ(FMINV, fminv)
3719 DO_VPZ(FMAXV, fmaxv)
3720 
3721 #undef DO_VPZ
3722 
3723 /*
3724  *** SVE Floating Point Unary Operations - Unpredicated Group
3725  */
3726 
3727 static gen_helper_gvec_2_ptr * const frecpe_fns[] = {
3728     NULL,                     gen_helper_gvec_frecpe_h,
3729     gen_helper_gvec_frecpe_s, gen_helper_gvec_frecpe_d,
3730 };
3731 TRANS_FEAT(FRECPE, aa64_sve, gen_gvec_fpst_arg_zz, frecpe_fns[a->esz], a, 0)
3732 
3733 static gen_helper_gvec_2_ptr * const frsqrte_fns[] = {
3734     NULL,                      gen_helper_gvec_frsqrte_h,
3735     gen_helper_gvec_frsqrte_s, gen_helper_gvec_frsqrte_d,
3736 };
3737 TRANS_FEAT(FRSQRTE, aa64_sve, gen_gvec_fpst_arg_zz, frsqrte_fns[a->esz], a, 0)
3738 
3739 /*
3740  *** SVE Floating Point Compare with Zero Group
3741  */
3742 
3743 static bool do_ppz_fp(DisasContext *s, arg_rpr_esz *a,
3744                       gen_helper_gvec_3_ptr *fn)
3745 {
3746     if (fn == NULL) {
3747         return false;
3748     }
3749     if (sve_access_check(s)) {
3750         unsigned vsz = vec_full_reg_size(s);
3751         TCGv_ptr status =
3752             fpstatus_ptr(a->esz == MO_16 ? FPST_FPCR_F16 : FPST_FPCR);
3753 
3754         tcg_gen_gvec_3_ptr(pred_full_reg_offset(s, a->rd),
3755                            vec_full_reg_offset(s, a->rn),
3756                            pred_full_reg_offset(s, a->pg),
3757                            status, vsz, vsz, 0, fn);
3758     }
3759     return true;
3760 }
3761 
3762 #define DO_PPZ(NAME, name) \
3763     static gen_helper_gvec_3_ptr * const name##_fns[] = {         \
3764         NULL,                      gen_helper_sve_##name##_h,     \
3765         gen_helper_sve_##name##_s, gen_helper_sve_##name##_d,     \
3766     };                                                            \
3767     TRANS_FEAT(NAME, aa64_sve, do_ppz_fp, a, name##_fns[a->esz])
3768 
3769 DO_PPZ(FCMGE_ppz0, fcmge0)
3770 DO_PPZ(FCMGT_ppz0, fcmgt0)
3771 DO_PPZ(FCMLE_ppz0, fcmle0)
3772 DO_PPZ(FCMLT_ppz0, fcmlt0)
3773 DO_PPZ(FCMEQ_ppz0, fcmeq0)
3774 DO_PPZ(FCMNE_ppz0, fcmne0)
3775 
3776 #undef DO_PPZ
3777 
3778 /*
3779  *** SVE floating-point trig multiply-add coefficient
3780  */
3781 
3782 static gen_helper_gvec_3_ptr * const ftmad_fns[4] = {
3783     NULL,                   gen_helper_sve_ftmad_h,
3784     gen_helper_sve_ftmad_s, gen_helper_sve_ftmad_d,
3785 };
3786 TRANS_FEAT_NONSTREAMING(FTMAD, aa64_sve, gen_gvec_fpst_zzz,
3787                         ftmad_fns[a->esz], a->rd, a->rn, a->rm, a->imm,
3788                         a->esz == MO_16 ? FPST_FPCR_F16 : FPST_FPCR)
3789 
3790 /*
3791  *** SVE Floating Point Accumulating Reduction Group
3792  */
3793 
3794 static bool trans_FADDA(DisasContext *s, arg_rprr_esz *a)
3795 {
3796     typedef void fadda_fn(TCGv_i64, TCGv_i64, TCGv_ptr,
3797                           TCGv_ptr, TCGv_ptr, TCGv_i32);
3798     static fadda_fn * const fns[3] = {
3799         gen_helper_sve_fadda_h,
3800         gen_helper_sve_fadda_s,
3801         gen_helper_sve_fadda_d,
3802     };
3803     unsigned vsz = vec_full_reg_size(s);
3804     TCGv_ptr t_rm, t_pg, t_fpst;
3805     TCGv_i64 t_val;
3806     TCGv_i32 t_desc;
3807 
3808     if (a->esz == 0 || !dc_isar_feature(aa64_sve, s)) {
3809         return false;
3810     }
3811     s->is_nonstreaming = true;
3812     if (!sve_access_check(s)) {
3813         return true;
3814     }
3815 
3816     t_val = load_esz(cpu_env, vec_reg_offset(s, a->rn, 0, a->esz), a->esz);
3817     t_rm = tcg_temp_new_ptr();
3818     t_pg = tcg_temp_new_ptr();
3819     tcg_gen_addi_ptr(t_rm, cpu_env, vec_full_reg_offset(s, a->rm));
3820     tcg_gen_addi_ptr(t_pg, cpu_env, pred_full_reg_offset(s, a->pg));
3821     t_fpst = fpstatus_ptr(a->esz == MO_16 ? FPST_FPCR_F16 : FPST_FPCR);
3822     t_desc = tcg_constant_i32(simd_desc(vsz, vsz, 0));
3823 
3824     fns[a->esz - 1](t_val, t_val, t_rm, t_pg, t_fpst, t_desc);
3825 
3826     write_fp_dreg(s, a->rd, t_val);
3827     return true;
3828 }
3829 
3830 /*
3831  *** SVE Floating Point Arithmetic - Unpredicated Group
3832  */
3833 
3834 #define DO_FP3(NAME, name) \
3835     static gen_helper_gvec_3_ptr * const name##_fns[4] = {          \
3836         NULL, gen_helper_gvec_##name##_h,                           \
3837         gen_helper_gvec_##name##_s, gen_helper_gvec_##name##_d      \
3838     };                                                              \
3839     TRANS_FEAT(NAME, aa64_sve, gen_gvec_fpst_arg_zzz, name##_fns[a->esz], a, 0)
3840 
3841 DO_FP3(FADD_zzz, fadd)
3842 DO_FP3(FSUB_zzz, fsub)
3843 DO_FP3(FMUL_zzz, fmul)
3844 DO_FP3(FRECPS, recps)
3845 DO_FP3(FRSQRTS, rsqrts)
3846 
3847 #undef DO_FP3
3848 
3849 static gen_helper_gvec_3_ptr * const ftsmul_fns[4] = {
3850     NULL,                     gen_helper_gvec_ftsmul_h,
3851     gen_helper_gvec_ftsmul_s, gen_helper_gvec_ftsmul_d
3852 };
3853 TRANS_FEAT_NONSTREAMING(FTSMUL, aa64_sve, gen_gvec_fpst_arg_zzz,
3854                         ftsmul_fns[a->esz], a, 0)
3855 
3856 /*
3857  *** SVE Floating Point Arithmetic - Predicated Group
3858  */
3859 
3860 #define DO_ZPZZ_FP(NAME, FEAT, name) \
3861     static gen_helper_gvec_4_ptr * const name##_zpzz_fns[4] = { \
3862         NULL,                  gen_helper_##name##_h,           \
3863         gen_helper_##name##_s, gen_helper_##name##_d            \
3864     };                                                          \
3865     TRANS_FEAT(NAME, FEAT, gen_gvec_fpst_arg_zpzz, name##_zpzz_fns[a->esz], a)
3866 
3867 DO_ZPZZ_FP(FADD_zpzz, aa64_sve, sve_fadd)
3868 DO_ZPZZ_FP(FSUB_zpzz, aa64_sve, sve_fsub)
3869 DO_ZPZZ_FP(FMUL_zpzz, aa64_sve, sve_fmul)
3870 DO_ZPZZ_FP(FMIN_zpzz, aa64_sve, sve_fmin)
3871 DO_ZPZZ_FP(FMAX_zpzz, aa64_sve, sve_fmax)
3872 DO_ZPZZ_FP(FMINNM_zpzz, aa64_sve, sve_fminnum)
3873 DO_ZPZZ_FP(FMAXNM_zpzz, aa64_sve, sve_fmaxnum)
3874 DO_ZPZZ_FP(FABD, aa64_sve, sve_fabd)
3875 DO_ZPZZ_FP(FSCALE, aa64_sve, sve_fscalbn)
3876 DO_ZPZZ_FP(FDIV, aa64_sve, sve_fdiv)
3877 DO_ZPZZ_FP(FMULX, aa64_sve, sve_fmulx)
3878 
3879 typedef void gen_helper_sve_fp2scalar(TCGv_ptr, TCGv_ptr, TCGv_ptr,
3880                                       TCGv_i64, TCGv_ptr, TCGv_i32);
3881 
3882 static void do_fp_scalar(DisasContext *s, int zd, int zn, int pg, bool is_fp16,
3883                          TCGv_i64 scalar, gen_helper_sve_fp2scalar *fn)
3884 {
3885     unsigned vsz = vec_full_reg_size(s);
3886     TCGv_ptr t_zd, t_zn, t_pg, status;
3887     TCGv_i32 desc;
3888 
3889     t_zd = tcg_temp_new_ptr();
3890     t_zn = tcg_temp_new_ptr();
3891     t_pg = tcg_temp_new_ptr();
3892     tcg_gen_addi_ptr(t_zd, cpu_env, vec_full_reg_offset(s, zd));
3893     tcg_gen_addi_ptr(t_zn, cpu_env, vec_full_reg_offset(s, zn));
3894     tcg_gen_addi_ptr(t_pg, cpu_env, pred_full_reg_offset(s, pg));
3895 
3896     status = fpstatus_ptr(is_fp16 ? FPST_FPCR_F16 : FPST_FPCR);
3897     desc = tcg_constant_i32(simd_desc(vsz, vsz, 0));
3898     fn(t_zd, t_zn, t_pg, scalar, status, desc);
3899 }
3900 
3901 static bool do_fp_imm(DisasContext *s, arg_rpri_esz *a, uint64_t imm,
3902                       gen_helper_sve_fp2scalar *fn)
3903 {
3904     if (fn == NULL) {
3905         return false;
3906     }
3907     if (sve_access_check(s)) {
3908         do_fp_scalar(s, a->rd, a->rn, a->pg, a->esz == MO_16,
3909                      tcg_constant_i64(imm), fn);
3910     }
3911     return true;
3912 }
3913 
3914 #define DO_FP_IMM(NAME, name, const0, const1)                           \
3915     static gen_helper_sve_fp2scalar * const name##_fns[4] = {           \
3916         NULL, gen_helper_sve_##name##_h,                                \
3917         gen_helper_sve_##name##_s,                                      \
3918         gen_helper_sve_##name##_d                                       \
3919     };                                                                  \
3920     static uint64_t const name##_const[4][2] = {                        \
3921         { -1, -1 },                                                     \
3922         { float16_##const0, float16_##const1 },                         \
3923         { float32_##const0, float32_##const1 },                         \
3924         { float64_##const0, float64_##const1 },                         \
3925     };                                                                  \
3926     TRANS_FEAT(NAME##_zpzi, aa64_sve, do_fp_imm, a,                     \
3927                name##_const[a->esz][a->imm], name##_fns[a->esz])
3928 
3929 DO_FP_IMM(FADD, fadds, half, one)
3930 DO_FP_IMM(FSUB, fsubs, half, one)
3931 DO_FP_IMM(FMUL, fmuls, half, two)
3932 DO_FP_IMM(FSUBR, fsubrs, half, one)
3933 DO_FP_IMM(FMAXNM, fmaxnms, zero, one)
3934 DO_FP_IMM(FMINNM, fminnms, zero, one)
3935 DO_FP_IMM(FMAX, fmaxs, zero, one)
3936 DO_FP_IMM(FMIN, fmins, zero, one)
3937 
3938 #undef DO_FP_IMM
3939 
3940 static bool do_fp_cmp(DisasContext *s, arg_rprr_esz *a,
3941                       gen_helper_gvec_4_ptr *fn)
3942 {
3943     if (fn == NULL) {
3944         return false;
3945     }
3946     if (sve_access_check(s)) {
3947         unsigned vsz = vec_full_reg_size(s);
3948         TCGv_ptr status = fpstatus_ptr(a->esz == MO_16 ? FPST_FPCR_F16 : FPST_FPCR);
3949         tcg_gen_gvec_4_ptr(pred_full_reg_offset(s, a->rd),
3950                            vec_full_reg_offset(s, a->rn),
3951                            vec_full_reg_offset(s, a->rm),
3952                            pred_full_reg_offset(s, a->pg),
3953                            status, vsz, vsz, 0, fn);
3954     }
3955     return true;
3956 }
3957 
3958 #define DO_FPCMP(NAME, name) \
3959     static gen_helper_gvec_4_ptr * const name##_fns[4] = {            \
3960         NULL, gen_helper_sve_##name##_h,                              \
3961         gen_helper_sve_##name##_s, gen_helper_sve_##name##_d          \
3962     };                                                                \
3963     TRANS_FEAT(NAME##_ppzz, aa64_sve, do_fp_cmp, a, name##_fns[a->esz])
3964 
3965 DO_FPCMP(FCMGE, fcmge)
3966 DO_FPCMP(FCMGT, fcmgt)
3967 DO_FPCMP(FCMEQ, fcmeq)
3968 DO_FPCMP(FCMNE, fcmne)
3969 DO_FPCMP(FCMUO, fcmuo)
3970 DO_FPCMP(FACGE, facge)
3971 DO_FPCMP(FACGT, facgt)
3972 
3973 #undef DO_FPCMP
3974 
3975 static gen_helper_gvec_4_ptr * const fcadd_fns[] = {
3976     NULL,                   gen_helper_sve_fcadd_h,
3977     gen_helper_sve_fcadd_s, gen_helper_sve_fcadd_d,
3978 };
3979 TRANS_FEAT(FCADD, aa64_sve, gen_gvec_fpst_zzzp, fcadd_fns[a->esz],
3980            a->rd, a->rn, a->rm, a->pg, a->rot,
3981            a->esz == MO_16 ? FPST_FPCR_F16 : FPST_FPCR)
3982 
3983 #define DO_FMLA(NAME, name) \
3984     static gen_helper_gvec_5_ptr * const name##_fns[4] = {              \
3985         NULL, gen_helper_sve_##name##_h,                                \
3986         gen_helper_sve_##name##_s, gen_helper_sve_##name##_d            \
3987     };                                                                  \
3988     TRANS_FEAT(NAME, aa64_sve, gen_gvec_fpst_zzzzp, name##_fns[a->esz], \
3989                a->rd, a->rn, a->rm, a->ra, a->pg, 0,                    \
3990                a->esz == MO_16 ? FPST_FPCR_F16 : FPST_FPCR)
3991 
3992 DO_FMLA(FMLA_zpzzz, fmla_zpzzz)
3993 DO_FMLA(FMLS_zpzzz, fmls_zpzzz)
3994 DO_FMLA(FNMLA_zpzzz, fnmla_zpzzz)
3995 DO_FMLA(FNMLS_zpzzz, fnmls_zpzzz)
3996 
3997 #undef DO_FMLA
3998 
3999 static gen_helper_gvec_5_ptr * const fcmla_fns[4] = {
4000     NULL,                         gen_helper_sve_fcmla_zpzzz_h,
4001     gen_helper_sve_fcmla_zpzzz_s, gen_helper_sve_fcmla_zpzzz_d,
4002 };
4003 TRANS_FEAT(FCMLA_zpzzz, aa64_sve, gen_gvec_fpst_zzzzp, fcmla_fns[a->esz],
4004            a->rd, a->rn, a->rm, a->ra, a->pg, a->rot,
4005            a->esz == MO_16 ? FPST_FPCR_F16 : FPST_FPCR)
4006 
4007 static gen_helper_gvec_4_ptr * const fcmla_idx_fns[4] = {
4008     NULL, gen_helper_gvec_fcmlah_idx, gen_helper_gvec_fcmlas_idx, NULL
4009 };
4010 TRANS_FEAT(FCMLA_zzxz, aa64_sve, gen_gvec_fpst_zzzz, fcmla_idx_fns[a->esz],
4011            a->rd, a->rn, a->rm, a->ra, a->index * 4 + a->rot,
4012            a->esz == MO_16 ? FPST_FPCR_F16 : FPST_FPCR)
4013 
4014 /*
4015  *** SVE Floating Point Unary Operations Predicated Group
4016  */
4017 
4018 TRANS_FEAT(FCVT_sh, aa64_sve, gen_gvec_fpst_arg_zpz,
4019            gen_helper_sve_fcvt_sh, a, 0, FPST_FPCR)
4020 TRANS_FEAT(FCVT_hs, aa64_sve, gen_gvec_fpst_arg_zpz,
4021            gen_helper_sve_fcvt_hs, a, 0, FPST_FPCR)
4022 
4023 TRANS_FEAT(BFCVT, aa64_sve_bf16, gen_gvec_fpst_arg_zpz,
4024            gen_helper_sve_bfcvt, a, 0, FPST_FPCR)
4025 
4026 TRANS_FEAT(FCVT_dh, aa64_sve, gen_gvec_fpst_arg_zpz,
4027            gen_helper_sve_fcvt_dh, a, 0, FPST_FPCR)
4028 TRANS_FEAT(FCVT_hd, aa64_sve, gen_gvec_fpst_arg_zpz,
4029            gen_helper_sve_fcvt_hd, a, 0, FPST_FPCR)
4030 TRANS_FEAT(FCVT_ds, aa64_sve, gen_gvec_fpst_arg_zpz,
4031            gen_helper_sve_fcvt_ds, a, 0, FPST_FPCR)
4032 TRANS_FEAT(FCVT_sd, aa64_sve, gen_gvec_fpst_arg_zpz,
4033            gen_helper_sve_fcvt_sd, a, 0, FPST_FPCR)
4034 
4035 TRANS_FEAT(FCVTZS_hh, aa64_sve, gen_gvec_fpst_arg_zpz,
4036            gen_helper_sve_fcvtzs_hh, a, 0, FPST_FPCR_F16)
4037 TRANS_FEAT(FCVTZU_hh, aa64_sve, gen_gvec_fpst_arg_zpz,
4038            gen_helper_sve_fcvtzu_hh, a, 0, FPST_FPCR_F16)
4039 TRANS_FEAT(FCVTZS_hs, aa64_sve, gen_gvec_fpst_arg_zpz,
4040            gen_helper_sve_fcvtzs_hs, a, 0, FPST_FPCR_F16)
4041 TRANS_FEAT(FCVTZU_hs, aa64_sve, gen_gvec_fpst_arg_zpz,
4042            gen_helper_sve_fcvtzu_hs, a, 0, FPST_FPCR_F16)
4043 TRANS_FEAT(FCVTZS_hd, aa64_sve, gen_gvec_fpst_arg_zpz,
4044            gen_helper_sve_fcvtzs_hd, a, 0, FPST_FPCR_F16)
4045 TRANS_FEAT(FCVTZU_hd, aa64_sve, gen_gvec_fpst_arg_zpz,
4046            gen_helper_sve_fcvtzu_hd, a, 0, FPST_FPCR_F16)
4047 
4048 TRANS_FEAT(FCVTZS_ss, aa64_sve, gen_gvec_fpst_arg_zpz,
4049            gen_helper_sve_fcvtzs_ss, a, 0, FPST_FPCR)
4050 TRANS_FEAT(FCVTZU_ss, aa64_sve, gen_gvec_fpst_arg_zpz,
4051            gen_helper_sve_fcvtzu_ss, a, 0, FPST_FPCR)
4052 TRANS_FEAT(FCVTZS_sd, aa64_sve, gen_gvec_fpst_arg_zpz,
4053            gen_helper_sve_fcvtzs_sd, a, 0, FPST_FPCR)
4054 TRANS_FEAT(FCVTZU_sd, aa64_sve, gen_gvec_fpst_arg_zpz,
4055            gen_helper_sve_fcvtzu_sd, a, 0, FPST_FPCR)
4056 TRANS_FEAT(FCVTZS_ds, aa64_sve, gen_gvec_fpst_arg_zpz,
4057            gen_helper_sve_fcvtzs_ds, a, 0, FPST_FPCR)
4058 TRANS_FEAT(FCVTZU_ds, aa64_sve, gen_gvec_fpst_arg_zpz,
4059            gen_helper_sve_fcvtzu_ds, a, 0, FPST_FPCR)
4060 
4061 TRANS_FEAT(FCVTZS_dd, aa64_sve, gen_gvec_fpst_arg_zpz,
4062            gen_helper_sve_fcvtzs_dd, a, 0, FPST_FPCR)
4063 TRANS_FEAT(FCVTZU_dd, aa64_sve, gen_gvec_fpst_arg_zpz,
4064            gen_helper_sve_fcvtzu_dd, a, 0, FPST_FPCR)
4065 
4066 static gen_helper_gvec_3_ptr * const frint_fns[] = {
4067     NULL,
4068     gen_helper_sve_frint_h,
4069     gen_helper_sve_frint_s,
4070     gen_helper_sve_frint_d
4071 };
4072 TRANS_FEAT(FRINTI, aa64_sve, gen_gvec_fpst_arg_zpz, frint_fns[a->esz],
4073            a, 0, a->esz == MO_16 ? FPST_FPCR_F16 : FPST_FPCR)
4074 
4075 static gen_helper_gvec_3_ptr * const frintx_fns[] = {
4076     NULL,
4077     gen_helper_sve_frintx_h,
4078     gen_helper_sve_frintx_s,
4079     gen_helper_sve_frintx_d
4080 };
4081 TRANS_FEAT(FRINTX, aa64_sve, gen_gvec_fpst_arg_zpz, frintx_fns[a->esz],
4082            a, 0, a->esz == MO_16 ? FPST_FPCR_F16 : FPST_FPCR);
4083 
4084 static bool do_frint_mode(DisasContext *s, arg_rpr_esz *a,
4085                           int mode, gen_helper_gvec_3_ptr *fn)
4086 {
4087     unsigned vsz;
4088     TCGv_i32 tmode;
4089     TCGv_ptr status;
4090 
4091     if (fn == NULL) {
4092         return false;
4093     }
4094     if (!sve_access_check(s)) {
4095         return true;
4096     }
4097 
4098     vsz = vec_full_reg_size(s);
4099     tmode = tcg_const_i32(mode);
4100     status = fpstatus_ptr(a->esz == MO_16 ? FPST_FPCR_F16 : FPST_FPCR);
4101 
4102     gen_helper_set_rmode(tmode, tmode, status);
4103 
4104     tcg_gen_gvec_3_ptr(vec_full_reg_offset(s, a->rd),
4105                        vec_full_reg_offset(s, a->rn),
4106                        pred_full_reg_offset(s, a->pg),
4107                        status, vsz, vsz, 0, fn);
4108 
4109     gen_helper_set_rmode(tmode, tmode, status);
4110     return true;
4111 }
4112 
4113 TRANS_FEAT(FRINTN, aa64_sve, do_frint_mode, a,
4114            float_round_nearest_even, frint_fns[a->esz])
4115 TRANS_FEAT(FRINTP, aa64_sve, do_frint_mode, a,
4116            float_round_up, frint_fns[a->esz])
4117 TRANS_FEAT(FRINTM, aa64_sve, do_frint_mode, a,
4118            float_round_down, frint_fns[a->esz])
4119 TRANS_FEAT(FRINTZ, aa64_sve, do_frint_mode, a,
4120            float_round_to_zero, frint_fns[a->esz])
4121 TRANS_FEAT(FRINTA, aa64_sve, do_frint_mode, a,
4122            float_round_ties_away, frint_fns[a->esz])
4123 
4124 static gen_helper_gvec_3_ptr * const frecpx_fns[] = {
4125     NULL,                    gen_helper_sve_frecpx_h,
4126     gen_helper_sve_frecpx_s, gen_helper_sve_frecpx_d,
4127 };
4128 TRANS_FEAT(FRECPX, aa64_sve, gen_gvec_fpst_arg_zpz, frecpx_fns[a->esz],
4129            a, 0, a->esz == MO_16 ? FPST_FPCR_F16 : FPST_FPCR)
4130 
4131 static gen_helper_gvec_3_ptr * const fsqrt_fns[] = {
4132     NULL,                   gen_helper_sve_fsqrt_h,
4133     gen_helper_sve_fsqrt_s, gen_helper_sve_fsqrt_d,
4134 };
4135 TRANS_FEAT(FSQRT, aa64_sve, gen_gvec_fpst_arg_zpz, fsqrt_fns[a->esz],
4136            a, 0, a->esz == MO_16 ? FPST_FPCR_F16 : FPST_FPCR)
4137 
4138 TRANS_FEAT(SCVTF_hh, aa64_sve, gen_gvec_fpst_arg_zpz,
4139            gen_helper_sve_scvt_hh, a, 0, FPST_FPCR_F16)
4140 TRANS_FEAT(SCVTF_sh, aa64_sve, gen_gvec_fpst_arg_zpz,
4141            gen_helper_sve_scvt_sh, a, 0, FPST_FPCR_F16)
4142 TRANS_FEAT(SCVTF_dh, aa64_sve, gen_gvec_fpst_arg_zpz,
4143            gen_helper_sve_scvt_dh, a, 0, FPST_FPCR_F16)
4144 
4145 TRANS_FEAT(SCVTF_ss, aa64_sve, gen_gvec_fpst_arg_zpz,
4146            gen_helper_sve_scvt_ss, a, 0, FPST_FPCR)
4147 TRANS_FEAT(SCVTF_ds, aa64_sve, gen_gvec_fpst_arg_zpz,
4148            gen_helper_sve_scvt_ds, a, 0, FPST_FPCR)
4149 
4150 TRANS_FEAT(SCVTF_sd, aa64_sve, gen_gvec_fpst_arg_zpz,
4151            gen_helper_sve_scvt_sd, a, 0, FPST_FPCR)
4152 TRANS_FEAT(SCVTF_dd, aa64_sve, gen_gvec_fpst_arg_zpz,
4153            gen_helper_sve_scvt_dd, a, 0, FPST_FPCR)
4154 
4155 TRANS_FEAT(UCVTF_hh, aa64_sve, gen_gvec_fpst_arg_zpz,
4156            gen_helper_sve_ucvt_hh, a, 0, FPST_FPCR_F16)
4157 TRANS_FEAT(UCVTF_sh, aa64_sve, gen_gvec_fpst_arg_zpz,
4158            gen_helper_sve_ucvt_sh, a, 0, FPST_FPCR_F16)
4159 TRANS_FEAT(UCVTF_dh, aa64_sve, gen_gvec_fpst_arg_zpz,
4160            gen_helper_sve_ucvt_dh, a, 0, FPST_FPCR_F16)
4161 
4162 TRANS_FEAT(UCVTF_ss, aa64_sve, gen_gvec_fpst_arg_zpz,
4163            gen_helper_sve_ucvt_ss, a, 0, FPST_FPCR)
4164 TRANS_FEAT(UCVTF_ds, aa64_sve, gen_gvec_fpst_arg_zpz,
4165            gen_helper_sve_ucvt_ds, a, 0, FPST_FPCR)
4166 TRANS_FEAT(UCVTF_sd, aa64_sve, gen_gvec_fpst_arg_zpz,
4167            gen_helper_sve_ucvt_sd, a, 0, FPST_FPCR)
4168 
4169 TRANS_FEAT(UCVTF_dd, aa64_sve, gen_gvec_fpst_arg_zpz,
4170            gen_helper_sve_ucvt_dd, a, 0, FPST_FPCR)
4171 
4172 /*
4173  *** SVE Memory - 32-bit Gather and Unsized Contiguous Group
4174  */
4175 
4176 /* Subroutine loading a vector register at VOFS of LEN bytes.
4177  * The load should begin at the address Rn + IMM.
4178  */
4179 
4180 void gen_sve_ldr(DisasContext *s, TCGv_ptr base, int vofs,
4181                  int len, int rn, int imm)
4182 {
4183     int len_align = QEMU_ALIGN_DOWN(len, 8);
4184     int len_remain = len % 8;
4185     int nparts = len / 8 + ctpop8(len_remain);
4186     int midx = get_mem_index(s);
4187     TCGv_i64 dirty_addr, clean_addr, t0, t1;
4188 
4189     dirty_addr = tcg_temp_new_i64();
4190     tcg_gen_addi_i64(dirty_addr, cpu_reg_sp(s, rn), imm);
4191     clean_addr = gen_mte_checkN(s, dirty_addr, false, rn != 31, len);
4192 
4193     /*
4194      * Note that unpredicated load/store of vector/predicate registers
4195      * are defined as a stream of bytes, which equates to little-endian
4196      * operations on larger quantities.
4197      * Attempt to keep code expansion to a minimum by limiting the
4198      * amount of unrolling done.
4199      */
4200     if (nparts <= 4) {
4201         int i;
4202 
4203         t0 = tcg_temp_new_i64();
4204         for (i = 0; i < len_align; i += 8) {
4205             tcg_gen_qemu_ld_i64(t0, clean_addr, midx, MO_LEUQ);
4206             tcg_gen_st_i64(t0, base, vofs + i);
4207             tcg_gen_addi_i64(clean_addr, clean_addr, 8);
4208         }
4209     } else {
4210         TCGLabel *loop = gen_new_label();
4211         TCGv_ptr tp, i = tcg_const_ptr(0);
4212 
4213         gen_set_label(loop);
4214 
4215         t0 = tcg_temp_new_i64();
4216         tcg_gen_qemu_ld_i64(t0, clean_addr, midx, MO_LEUQ);
4217         tcg_gen_addi_i64(clean_addr, clean_addr, 8);
4218 
4219         tp = tcg_temp_new_ptr();
4220         tcg_gen_add_ptr(tp, base, i);
4221         tcg_gen_addi_ptr(i, i, 8);
4222         tcg_gen_st_i64(t0, tp, vofs);
4223 
4224         tcg_gen_brcondi_ptr(TCG_COND_LTU, i, len_align, loop);
4225     }
4226 
4227     /*
4228      * Predicate register loads can be any multiple of 2.
4229      * Note that we still store the entire 64-bit unit into cpu_env.
4230      */
4231     if (len_remain) {
4232         t0 = tcg_temp_new_i64();
4233         switch (len_remain) {
4234         case 2:
4235         case 4:
4236         case 8:
4237             tcg_gen_qemu_ld_i64(t0, clean_addr, midx,
4238                                 MO_LE | ctz32(len_remain));
4239             break;
4240 
4241         case 6:
4242             t1 = tcg_temp_new_i64();
4243             tcg_gen_qemu_ld_i64(t0, clean_addr, midx, MO_LEUL);
4244             tcg_gen_addi_i64(clean_addr, clean_addr, 4);
4245             tcg_gen_qemu_ld_i64(t1, clean_addr, midx, MO_LEUW);
4246             tcg_gen_deposit_i64(t0, t0, t1, 32, 32);
4247             break;
4248 
4249         default:
4250             g_assert_not_reached();
4251         }
4252         tcg_gen_st_i64(t0, base, vofs + len_align);
4253     }
4254 }
4255 
4256 /* Similarly for stores.  */
4257 void gen_sve_str(DisasContext *s, TCGv_ptr base, int vofs,
4258                  int len, int rn, int imm)
4259 {
4260     int len_align = QEMU_ALIGN_DOWN(len, 8);
4261     int len_remain = len % 8;
4262     int nparts = len / 8 + ctpop8(len_remain);
4263     int midx = get_mem_index(s);
4264     TCGv_i64 dirty_addr, clean_addr, t0;
4265 
4266     dirty_addr = tcg_temp_new_i64();
4267     tcg_gen_addi_i64(dirty_addr, cpu_reg_sp(s, rn), imm);
4268     clean_addr = gen_mte_checkN(s, dirty_addr, false, rn != 31, len);
4269 
4270     /* Note that unpredicated load/store of vector/predicate registers
4271      * are defined as a stream of bytes, which equates to little-endian
4272      * operations on larger quantities.  There is no nice way to force
4273      * a little-endian store for aarch64_be-linux-user out of line.
4274      *
4275      * Attempt to keep code expansion to a minimum by limiting the
4276      * amount of unrolling done.
4277      */
4278     if (nparts <= 4) {
4279         int i;
4280 
4281         t0 = tcg_temp_new_i64();
4282         for (i = 0; i < len_align; i += 8) {
4283             tcg_gen_ld_i64(t0, base, vofs + i);
4284             tcg_gen_qemu_st_i64(t0, clean_addr, midx, MO_LEUQ);
4285             tcg_gen_addi_i64(clean_addr, clean_addr, 8);
4286         }
4287     } else {
4288         TCGLabel *loop = gen_new_label();
4289         TCGv_ptr tp, i = tcg_const_ptr(0);
4290 
4291         gen_set_label(loop);
4292 
4293         t0 = tcg_temp_new_i64();
4294         tp = tcg_temp_new_ptr();
4295         tcg_gen_add_ptr(tp, base, i);
4296         tcg_gen_ld_i64(t0, tp, vofs);
4297         tcg_gen_addi_ptr(i, i, 8);
4298 
4299         tcg_gen_qemu_st_i64(t0, clean_addr, midx, MO_LEUQ);
4300         tcg_gen_addi_i64(clean_addr, clean_addr, 8);
4301 
4302         tcg_gen_brcondi_ptr(TCG_COND_LTU, i, len_align, loop);
4303     }
4304 
4305     /* Predicate register stores can be any multiple of 2.  */
4306     if (len_remain) {
4307         t0 = tcg_temp_new_i64();
4308         tcg_gen_ld_i64(t0, base, vofs + len_align);
4309 
4310         switch (len_remain) {
4311         case 2:
4312         case 4:
4313         case 8:
4314             tcg_gen_qemu_st_i64(t0, clean_addr, midx,
4315                                 MO_LE | ctz32(len_remain));
4316             break;
4317 
4318         case 6:
4319             tcg_gen_qemu_st_i64(t0, clean_addr, midx, MO_LEUL);
4320             tcg_gen_addi_i64(clean_addr, clean_addr, 4);
4321             tcg_gen_shri_i64(t0, t0, 32);
4322             tcg_gen_qemu_st_i64(t0, clean_addr, midx, MO_LEUW);
4323             break;
4324 
4325         default:
4326             g_assert_not_reached();
4327         }
4328     }
4329 }
4330 
4331 static bool trans_LDR_zri(DisasContext *s, arg_rri *a)
4332 {
4333     if (!dc_isar_feature(aa64_sve, s)) {
4334         return false;
4335     }
4336     if (sve_access_check(s)) {
4337         int size = vec_full_reg_size(s);
4338         int off = vec_full_reg_offset(s, a->rd);
4339         gen_sve_ldr(s, cpu_env, off, size, a->rn, a->imm * size);
4340     }
4341     return true;
4342 }
4343 
4344 static bool trans_LDR_pri(DisasContext *s, arg_rri *a)
4345 {
4346     if (!dc_isar_feature(aa64_sve, s)) {
4347         return false;
4348     }
4349     if (sve_access_check(s)) {
4350         int size = pred_full_reg_size(s);
4351         int off = pred_full_reg_offset(s, a->rd);
4352         gen_sve_ldr(s, cpu_env, off, size, a->rn, a->imm * size);
4353     }
4354     return true;
4355 }
4356 
4357 static bool trans_STR_zri(DisasContext *s, arg_rri *a)
4358 {
4359     if (!dc_isar_feature(aa64_sve, s)) {
4360         return false;
4361     }
4362     if (sve_access_check(s)) {
4363         int size = vec_full_reg_size(s);
4364         int off = vec_full_reg_offset(s, a->rd);
4365         gen_sve_str(s, cpu_env, off, size, a->rn, a->imm * size);
4366     }
4367     return true;
4368 }
4369 
4370 static bool trans_STR_pri(DisasContext *s, arg_rri *a)
4371 {
4372     if (!dc_isar_feature(aa64_sve, s)) {
4373         return false;
4374     }
4375     if (sve_access_check(s)) {
4376         int size = pred_full_reg_size(s);
4377         int off = pred_full_reg_offset(s, a->rd);
4378         gen_sve_str(s, cpu_env, off, size, a->rn, a->imm * size);
4379     }
4380     return true;
4381 }
4382 
4383 /*
4384  *** SVE Memory - Contiguous Load Group
4385  */
4386 
4387 /* The memory mode of the dtype.  */
4388 static const MemOp dtype_mop[16] = {
4389     MO_UB, MO_UB, MO_UB, MO_UB,
4390     MO_SL, MO_UW, MO_UW, MO_UW,
4391     MO_SW, MO_SW, MO_UL, MO_UL,
4392     MO_SB, MO_SB, MO_SB, MO_UQ
4393 };
4394 
4395 #define dtype_msz(x)  (dtype_mop[x] & MO_SIZE)
4396 
4397 /* The vector element size of dtype.  */
4398 static const uint8_t dtype_esz[16] = {
4399     0, 1, 2, 3,
4400     3, 1, 2, 3,
4401     3, 2, 2, 3,
4402     3, 2, 1, 3
4403 };
4404 
4405 static void do_mem_zpa(DisasContext *s, int zt, int pg, TCGv_i64 addr,
4406                        int dtype, uint32_t mte_n, bool is_write,
4407                        gen_helper_gvec_mem *fn)
4408 {
4409     unsigned vsz = vec_full_reg_size(s);
4410     TCGv_ptr t_pg;
4411     int desc = 0;
4412 
4413     /*
4414      * For e.g. LD4, there are not enough arguments to pass all 4
4415      * registers as pointers, so encode the regno into the data field.
4416      * For consistency, do this even for LD1.
4417      */
4418     if (s->mte_active[0]) {
4419         int msz = dtype_msz(dtype);
4420 
4421         desc = FIELD_DP32(desc, MTEDESC, MIDX, get_mem_index(s));
4422         desc = FIELD_DP32(desc, MTEDESC, TBI, s->tbid);
4423         desc = FIELD_DP32(desc, MTEDESC, TCMA, s->tcma);
4424         desc = FIELD_DP32(desc, MTEDESC, WRITE, is_write);
4425         desc = FIELD_DP32(desc, MTEDESC, SIZEM1, (mte_n << msz) - 1);
4426         desc <<= SVE_MTEDESC_SHIFT;
4427     } else {
4428         addr = clean_data_tbi(s, addr);
4429     }
4430 
4431     desc = simd_desc(vsz, vsz, zt | desc);
4432     t_pg = tcg_temp_new_ptr();
4433 
4434     tcg_gen_addi_ptr(t_pg, cpu_env, pred_full_reg_offset(s, pg));
4435     fn(cpu_env, t_pg, addr, tcg_constant_i32(desc));
4436 }
4437 
4438 /* Indexed by [mte][be][dtype][nreg] */
4439 static gen_helper_gvec_mem * const ldr_fns[2][2][16][4] = {
4440     { /* mte inactive, little-endian */
4441       { { gen_helper_sve_ld1bb_r, gen_helper_sve_ld2bb_r,
4442           gen_helper_sve_ld3bb_r, gen_helper_sve_ld4bb_r },
4443         { gen_helper_sve_ld1bhu_r, NULL, NULL, NULL },
4444         { gen_helper_sve_ld1bsu_r, NULL, NULL, NULL },
4445         { gen_helper_sve_ld1bdu_r, NULL, NULL, NULL },
4446 
4447         { gen_helper_sve_ld1sds_le_r, NULL, NULL, NULL },
4448         { gen_helper_sve_ld1hh_le_r, gen_helper_sve_ld2hh_le_r,
4449           gen_helper_sve_ld3hh_le_r, gen_helper_sve_ld4hh_le_r },
4450         { gen_helper_sve_ld1hsu_le_r, NULL, NULL, NULL },
4451         { gen_helper_sve_ld1hdu_le_r, NULL, NULL, NULL },
4452 
4453         { gen_helper_sve_ld1hds_le_r, NULL, NULL, NULL },
4454         { gen_helper_sve_ld1hss_le_r, NULL, NULL, NULL },
4455         { gen_helper_sve_ld1ss_le_r, gen_helper_sve_ld2ss_le_r,
4456           gen_helper_sve_ld3ss_le_r, gen_helper_sve_ld4ss_le_r },
4457         { gen_helper_sve_ld1sdu_le_r, NULL, NULL, NULL },
4458 
4459         { gen_helper_sve_ld1bds_r, NULL, NULL, NULL },
4460         { gen_helper_sve_ld1bss_r, NULL, NULL, NULL },
4461         { gen_helper_sve_ld1bhs_r, NULL, NULL, NULL },
4462         { gen_helper_sve_ld1dd_le_r, gen_helper_sve_ld2dd_le_r,
4463           gen_helper_sve_ld3dd_le_r, gen_helper_sve_ld4dd_le_r } },
4464 
4465       /* mte inactive, big-endian */
4466       { { gen_helper_sve_ld1bb_r, gen_helper_sve_ld2bb_r,
4467           gen_helper_sve_ld3bb_r, gen_helper_sve_ld4bb_r },
4468         { gen_helper_sve_ld1bhu_r, NULL, NULL, NULL },
4469         { gen_helper_sve_ld1bsu_r, NULL, NULL, NULL },
4470         { gen_helper_sve_ld1bdu_r, NULL, NULL, NULL },
4471 
4472         { gen_helper_sve_ld1sds_be_r, NULL, NULL, NULL },
4473         { gen_helper_sve_ld1hh_be_r, gen_helper_sve_ld2hh_be_r,
4474           gen_helper_sve_ld3hh_be_r, gen_helper_sve_ld4hh_be_r },
4475         { gen_helper_sve_ld1hsu_be_r, NULL, NULL, NULL },
4476         { gen_helper_sve_ld1hdu_be_r, NULL, NULL, NULL },
4477 
4478         { gen_helper_sve_ld1hds_be_r, NULL, NULL, NULL },
4479         { gen_helper_sve_ld1hss_be_r, NULL, NULL, NULL },
4480         { gen_helper_sve_ld1ss_be_r, gen_helper_sve_ld2ss_be_r,
4481           gen_helper_sve_ld3ss_be_r, gen_helper_sve_ld4ss_be_r },
4482         { gen_helper_sve_ld1sdu_be_r, NULL, NULL, NULL },
4483 
4484         { gen_helper_sve_ld1bds_r, NULL, NULL, NULL },
4485         { gen_helper_sve_ld1bss_r, NULL, NULL, NULL },
4486         { gen_helper_sve_ld1bhs_r, NULL, NULL, NULL },
4487         { gen_helper_sve_ld1dd_be_r, gen_helper_sve_ld2dd_be_r,
4488           gen_helper_sve_ld3dd_be_r, gen_helper_sve_ld4dd_be_r } } },
4489 
4490     { /* mte active, little-endian */
4491       { { gen_helper_sve_ld1bb_r_mte,
4492           gen_helper_sve_ld2bb_r_mte,
4493           gen_helper_sve_ld3bb_r_mte,
4494           gen_helper_sve_ld4bb_r_mte },
4495         { gen_helper_sve_ld1bhu_r_mte, NULL, NULL, NULL },
4496         { gen_helper_sve_ld1bsu_r_mte, NULL, NULL, NULL },
4497         { gen_helper_sve_ld1bdu_r_mte, NULL, NULL, NULL },
4498 
4499         { gen_helper_sve_ld1sds_le_r_mte, NULL, NULL, NULL },
4500         { gen_helper_sve_ld1hh_le_r_mte,
4501           gen_helper_sve_ld2hh_le_r_mte,
4502           gen_helper_sve_ld3hh_le_r_mte,
4503           gen_helper_sve_ld4hh_le_r_mte },
4504         { gen_helper_sve_ld1hsu_le_r_mte, NULL, NULL, NULL },
4505         { gen_helper_sve_ld1hdu_le_r_mte, NULL, NULL, NULL },
4506 
4507         { gen_helper_sve_ld1hds_le_r_mte, NULL, NULL, NULL },
4508         { gen_helper_sve_ld1hss_le_r_mte, NULL, NULL, NULL },
4509         { gen_helper_sve_ld1ss_le_r_mte,
4510           gen_helper_sve_ld2ss_le_r_mte,
4511           gen_helper_sve_ld3ss_le_r_mte,
4512           gen_helper_sve_ld4ss_le_r_mte },
4513         { gen_helper_sve_ld1sdu_le_r_mte, NULL, NULL, NULL },
4514 
4515         { gen_helper_sve_ld1bds_r_mte, NULL, NULL, NULL },
4516         { gen_helper_sve_ld1bss_r_mte, NULL, NULL, NULL },
4517         { gen_helper_sve_ld1bhs_r_mte, NULL, NULL, NULL },
4518         { gen_helper_sve_ld1dd_le_r_mte,
4519           gen_helper_sve_ld2dd_le_r_mte,
4520           gen_helper_sve_ld3dd_le_r_mte,
4521           gen_helper_sve_ld4dd_le_r_mte } },
4522 
4523       /* mte active, big-endian */
4524       { { gen_helper_sve_ld1bb_r_mte,
4525           gen_helper_sve_ld2bb_r_mte,
4526           gen_helper_sve_ld3bb_r_mte,
4527           gen_helper_sve_ld4bb_r_mte },
4528         { gen_helper_sve_ld1bhu_r_mte, NULL, NULL, NULL },
4529         { gen_helper_sve_ld1bsu_r_mte, NULL, NULL, NULL },
4530         { gen_helper_sve_ld1bdu_r_mte, NULL, NULL, NULL },
4531 
4532         { gen_helper_sve_ld1sds_be_r_mte, NULL, NULL, NULL },
4533         { gen_helper_sve_ld1hh_be_r_mte,
4534           gen_helper_sve_ld2hh_be_r_mte,
4535           gen_helper_sve_ld3hh_be_r_mte,
4536           gen_helper_sve_ld4hh_be_r_mte },
4537         { gen_helper_sve_ld1hsu_be_r_mte, NULL, NULL, NULL },
4538         { gen_helper_sve_ld1hdu_be_r_mte, NULL, NULL, NULL },
4539 
4540         { gen_helper_sve_ld1hds_be_r_mte, NULL, NULL, NULL },
4541         { gen_helper_sve_ld1hss_be_r_mte, NULL, NULL, NULL },
4542         { gen_helper_sve_ld1ss_be_r_mte,
4543           gen_helper_sve_ld2ss_be_r_mte,
4544           gen_helper_sve_ld3ss_be_r_mte,
4545           gen_helper_sve_ld4ss_be_r_mte },
4546         { gen_helper_sve_ld1sdu_be_r_mte, NULL, NULL, NULL },
4547 
4548         { gen_helper_sve_ld1bds_r_mte, NULL, NULL, NULL },
4549         { gen_helper_sve_ld1bss_r_mte, NULL, NULL, NULL },
4550         { gen_helper_sve_ld1bhs_r_mte, NULL, NULL, NULL },
4551         { gen_helper_sve_ld1dd_be_r_mte,
4552           gen_helper_sve_ld2dd_be_r_mte,
4553           gen_helper_sve_ld3dd_be_r_mte,
4554           gen_helper_sve_ld4dd_be_r_mte } } },
4555 };
4556 
4557 static void do_ld_zpa(DisasContext *s, int zt, int pg,
4558                       TCGv_i64 addr, int dtype, int nreg)
4559 {
4560     gen_helper_gvec_mem *fn
4561         = ldr_fns[s->mte_active[0]][s->be_data == MO_BE][dtype][nreg];
4562 
4563     /*
4564      * While there are holes in the table, they are not
4565      * accessible via the instruction encoding.
4566      */
4567     assert(fn != NULL);
4568     do_mem_zpa(s, zt, pg, addr, dtype, nreg, false, fn);
4569 }
4570 
4571 static bool trans_LD_zprr(DisasContext *s, arg_rprr_load *a)
4572 {
4573     if (a->rm == 31 || !dc_isar_feature(aa64_sve, s)) {
4574         return false;
4575     }
4576     if (sve_access_check(s)) {
4577         TCGv_i64 addr = tcg_temp_new_i64();
4578         tcg_gen_shli_i64(addr, cpu_reg(s, a->rm), dtype_msz(a->dtype));
4579         tcg_gen_add_i64(addr, addr, cpu_reg_sp(s, a->rn));
4580         do_ld_zpa(s, a->rd, a->pg, addr, a->dtype, a->nreg);
4581     }
4582     return true;
4583 }
4584 
4585 static bool trans_LD_zpri(DisasContext *s, arg_rpri_load *a)
4586 {
4587     if (!dc_isar_feature(aa64_sve, s)) {
4588         return false;
4589     }
4590     if (sve_access_check(s)) {
4591         int vsz = vec_full_reg_size(s);
4592         int elements = vsz >> dtype_esz[a->dtype];
4593         TCGv_i64 addr = tcg_temp_new_i64();
4594 
4595         tcg_gen_addi_i64(addr, cpu_reg_sp(s, a->rn),
4596                          (a->imm * elements * (a->nreg + 1))
4597                          << dtype_msz(a->dtype));
4598         do_ld_zpa(s, a->rd, a->pg, addr, a->dtype, a->nreg);
4599     }
4600     return true;
4601 }
4602 
4603 static bool trans_LDFF1_zprr(DisasContext *s, arg_rprr_load *a)
4604 {
4605     static gen_helper_gvec_mem * const fns[2][2][16] = {
4606         { /* mte inactive, little-endian */
4607           { gen_helper_sve_ldff1bb_r,
4608             gen_helper_sve_ldff1bhu_r,
4609             gen_helper_sve_ldff1bsu_r,
4610             gen_helper_sve_ldff1bdu_r,
4611 
4612             gen_helper_sve_ldff1sds_le_r,
4613             gen_helper_sve_ldff1hh_le_r,
4614             gen_helper_sve_ldff1hsu_le_r,
4615             gen_helper_sve_ldff1hdu_le_r,
4616 
4617             gen_helper_sve_ldff1hds_le_r,
4618             gen_helper_sve_ldff1hss_le_r,
4619             gen_helper_sve_ldff1ss_le_r,
4620             gen_helper_sve_ldff1sdu_le_r,
4621 
4622             gen_helper_sve_ldff1bds_r,
4623             gen_helper_sve_ldff1bss_r,
4624             gen_helper_sve_ldff1bhs_r,
4625             gen_helper_sve_ldff1dd_le_r },
4626 
4627           /* mte inactive, big-endian */
4628           { gen_helper_sve_ldff1bb_r,
4629             gen_helper_sve_ldff1bhu_r,
4630             gen_helper_sve_ldff1bsu_r,
4631             gen_helper_sve_ldff1bdu_r,
4632 
4633             gen_helper_sve_ldff1sds_be_r,
4634             gen_helper_sve_ldff1hh_be_r,
4635             gen_helper_sve_ldff1hsu_be_r,
4636             gen_helper_sve_ldff1hdu_be_r,
4637 
4638             gen_helper_sve_ldff1hds_be_r,
4639             gen_helper_sve_ldff1hss_be_r,
4640             gen_helper_sve_ldff1ss_be_r,
4641             gen_helper_sve_ldff1sdu_be_r,
4642 
4643             gen_helper_sve_ldff1bds_r,
4644             gen_helper_sve_ldff1bss_r,
4645             gen_helper_sve_ldff1bhs_r,
4646             gen_helper_sve_ldff1dd_be_r } },
4647 
4648         { /* mte active, little-endian */
4649           { gen_helper_sve_ldff1bb_r_mte,
4650             gen_helper_sve_ldff1bhu_r_mte,
4651             gen_helper_sve_ldff1bsu_r_mte,
4652             gen_helper_sve_ldff1bdu_r_mte,
4653 
4654             gen_helper_sve_ldff1sds_le_r_mte,
4655             gen_helper_sve_ldff1hh_le_r_mte,
4656             gen_helper_sve_ldff1hsu_le_r_mte,
4657             gen_helper_sve_ldff1hdu_le_r_mte,
4658 
4659             gen_helper_sve_ldff1hds_le_r_mte,
4660             gen_helper_sve_ldff1hss_le_r_mte,
4661             gen_helper_sve_ldff1ss_le_r_mte,
4662             gen_helper_sve_ldff1sdu_le_r_mte,
4663 
4664             gen_helper_sve_ldff1bds_r_mte,
4665             gen_helper_sve_ldff1bss_r_mte,
4666             gen_helper_sve_ldff1bhs_r_mte,
4667             gen_helper_sve_ldff1dd_le_r_mte },
4668 
4669           /* mte active, big-endian */
4670           { gen_helper_sve_ldff1bb_r_mte,
4671             gen_helper_sve_ldff1bhu_r_mte,
4672             gen_helper_sve_ldff1bsu_r_mte,
4673             gen_helper_sve_ldff1bdu_r_mte,
4674 
4675             gen_helper_sve_ldff1sds_be_r_mte,
4676             gen_helper_sve_ldff1hh_be_r_mte,
4677             gen_helper_sve_ldff1hsu_be_r_mte,
4678             gen_helper_sve_ldff1hdu_be_r_mte,
4679 
4680             gen_helper_sve_ldff1hds_be_r_mte,
4681             gen_helper_sve_ldff1hss_be_r_mte,
4682             gen_helper_sve_ldff1ss_be_r_mte,
4683             gen_helper_sve_ldff1sdu_be_r_mte,
4684 
4685             gen_helper_sve_ldff1bds_r_mte,
4686             gen_helper_sve_ldff1bss_r_mte,
4687             gen_helper_sve_ldff1bhs_r_mte,
4688             gen_helper_sve_ldff1dd_be_r_mte } },
4689     };
4690 
4691     if (!dc_isar_feature(aa64_sve, s)) {
4692         return false;
4693     }
4694     s->is_nonstreaming = true;
4695     if (sve_access_check(s)) {
4696         TCGv_i64 addr = tcg_temp_new_i64();
4697         tcg_gen_shli_i64(addr, cpu_reg(s, a->rm), dtype_msz(a->dtype));
4698         tcg_gen_add_i64(addr, addr, cpu_reg_sp(s, a->rn));
4699         do_mem_zpa(s, a->rd, a->pg, addr, a->dtype, 1, false,
4700                    fns[s->mte_active[0]][s->be_data == MO_BE][a->dtype]);
4701     }
4702     return true;
4703 }
4704 
4705 static bool trans_LDNF1_zpri(DisasContext *s, arg_rpri_load *a)
4706 {
4707     static gen_helper_gvec_mem * const fns[2][2][16] = {
4708         { /* mte inactive, little-endian */
4709           { gen_helper_sve_ldnf1bb_r,
4710             gen_helper_sve_ldnf1bhu_r,
4711             gen_helper_sve_ldnf1bsu_r,
4712             gen_helper_sve_ldnf1bdu_r,
4713 
4714             gen_helper_sve_ldnf1sds_le_r,
4715             gen_helper_sve_ldnf1hh_le_r,
4716             gen_helper_sve_ldnf1hsu_le_r,
4717             gen_helper_sve_ldnf1hdu_le_r,
4718 
4719             gen_helper_sve_ldnf1hds_le_r,
4720             gen_helper_sve_ldnf1hss_le_r,
4721             gen_helper_sve_ldnf1ss_le_r,
4722             gen_helper_sve_ldnf1sdu_le_r,
4723 
4724             gen_helper_sve_ldnf1bds_r,
4725             gen_helper_sve_ldnf1bss_r,
4726             gen_helper_sve_ldnf1bhs_r,
4727             gen_helper_sve_ldnf1dd_le_r },
4728 
4729           /* mte inactive, big-endian */
4730           { gen_helper_sve_ldnf1bb_r,
4731             gen_helper_sve_ldnf1bhu_r,
4732             gen_helper_sve_ldnf1bsu_r,
4733             gen_helper_sve_ldnf1bdu_r,
4734 
4735             gen_helper_sve_ldnf1sds_be_r,
4736             gen_helper_sve_ldnf1hh_be_r,
4737             gen_helper_sve_ldnf1hsu_be_r,
4738             gen_helper_sve_ldnf1hdu_be_r,
4739 
4740             gen_helper_sve_ldnf1hds_be_r,
4741             gen_helper_sve_ldnf1hss_be_r,
4742             gen_helper_sve_ldnf1ss_be_r,
4743             gen_helper_sve_ldnf1sdu_be_r,
4744 
4745             gen_helper_sve_ldnf1bds_r,
4746             gen_helper_sve_ldnf1bss_r,
4747             gen_helper_sve_ldnf1bhs_r,
4748             gen_helper_sve_ldnf1dd_be_r } },
4749 
4750         { /* mte inactive, little-endian */
4751           { gen_helper_sve_ldnf1bb_r_mte,
4752             gen_helper_sve_ldnf1bhu_r_mte,
4753             gen_helper_sve_ldnf1bsu_r_mte,
4754             gen_helper_sve_ldnf1bdu_r_mte,
4755 
4756             gen_helper_sve_ldnf1sds_le_r_mte,
4757             gen_helper_sve_ldnf1hh_le_r_mte,
4758             gen_helper_sve_ldnf1hsu_le_r_mte,
4759             gen_helper_sve_ldnf1hdu_le_r_mte,
4760 
4761             gen_helper_sve_ldnf1hds_le_r_mte,
4762             gen_helper_sve_ldnf1hss_le_r_mte,
4763             gen_helper_sve_ldnf1ss_le_r_mte,
4764             gen_helper_sve_ldnf1sdu_le_r_mte,
4765 
4766             gen_helper_sve_ldnf1bds_r_mte,
4767             gen_helper_sve_ldnf1bss_r_mte,
4768             gen_helper_sve_ldnf1bhs_r_mte,
4769             gen_helper_sve_ldnf1dd_le_r_mte },
4770 
4771           /* mte inactive, big-endian */
4772           { gen_helper_sve_ldnf1bb_r_mte,
4773             gen_helper_sve_ldnf1bhu_r_mte,
4774             gen_helper_sve_ldnf1bsu_r_mte,
4775             gen_helper_sve_ldnf1bdu_r_mte,
4776 
4777             gen_helper_sve_ldnf1sds_be_r_mte,
4778             gen_helper_sve_ldnf1hh_be_r_mte,
4779             gen_helper_sve_ldnf1hsu_be_r_mte,
4780             gen_helper_sve_ldnf1hdu_be_r_mte,
4781 
4782             gen_helper_sve_ldnf1hds_be_r_mte,
4783             gen_helper_sve_ldnf1hss_be_r_mte,
4784             gen_helper_sve_ldnf1ss_be_r_mte,
4785             gen_helper_sve_ldnf1sdu_be_r_mte,
4786 
4787             gen_helper_sve_ldnf1bds_r_mte,
4788             gen_helper_sve_ldnf1bss_r_mte,
4789             gen_helper_sve_ldnf1bhs_r_mte,
4790             gen_helper_sve_ldnf1dd_be_r_mte } },
4791     };
4792 
4793     if (!dc_isar_feature(aa64_sve, s)) {
4794         return false;
4795     }
4796     s->is_nonstreaming = true;
4797     if (sve_access_check(s)) {
4798         int vsz = vec_full_reg_size(s);
4799         int elements = vsz >> dtype_esz[a->dtype];
4800         int off = (a->imm * elements) << dtype_msz(a->dtype);
4801         TCGv_i64 addr = tcg_temp_new_i64();
4802 
4803         tcg_gen_addi_i64(addr, cpu_reg_sp(s, a->rn), off);
4804         do_mem_zpa(s, a->rd, a->pg, addr, a->dtype, 1, false,
4805                    fns[s->mte_active[0]][s->be_data == MO_BE][a->dtype]);
4806     }
4807     return true;
4808 }
4809 
4810 static void do_ldrq(DisasContext *s, int zt, int pg, TCGv_i64 addr, int dtype)
4811 {
4812     unsigned vsz = vec_full_reg_size(s);
4813     TCGv_ptr t_pg;
4814     int poff;
4815 
4816     /* Load the first quadword using the normal predicated load helpers.  */
4817     poff = pred_full_reg_offset(s, pg);
4818     if (vsz > 16) {
4819         /*
4820          * Zero-extend the first 16 bits of the predicate into a temporary.
4821          * This avoids triggering an assert making sure we don't have bits
4822          * set within a predicate beyond VQ, but we have lowered VQ to 1
4823          * for this load operation.
4824          */
4825         TCGv_i64 tmp = tcg_temp_new_i64();
4826 #if HOST_BIG_ENDIAN
4827         poff += 6;
4828 #endif
4829         tcg_gen_ld16u_i64(tmp, cpu_env, poff);
4830 
4831         poff = offsetof(CPUARMState, vfp.preg_tmp);
4832         tcg_gen_st_i64(tmp, cpu_env, poff);
4833     }
4834 
4835     t_pg = tcg_temp_new_ptr();
4836     tcg_gen_addi_ptr(t_pg, cpu_env, poff);
4837 
4838     gen_helper_gvec_mem *fn
4839         = ldr_fns[s->mte_active[0]][s->be_data == MO_BE][dtype][0];
4840     fn(cpu_env, t_pg, addr, tcg_constant_i32(simd_desc(16, 16, zt)));
4841 
4842     /* Replicate that first quadword.  */
4843     if (vsz > 16) {
4844         int doff = vec_full_reg_offset(s, zt);
4845         tcg_gen_gvec_dup_mem(4, doff + 16, doff, vsz - 16, vsz - 16);
4846     }
4847 }
4848 
4849 static bool trans_LD1RQ_zprr(DisasContext *s, arg_rprr_load *a)
4850 {
4851     if (a->rm == 31 || !dc_isar_feature(aa64_sve, s)) {
4852         return false;
4853     }
4854     if (sve_access_check(s)) {
4855         int msz = dtype_msz(a->dtype);
4856         TCGv_i64 addr = tcg_temp_new_i64();
4857         tcg_gen_shli_i64(addr, cpu_reg(s, a->rm), msz);
4858         tcg_gen_add_i64(addr, addr, cpu_reg_sp(s, a->rn));
4859         do_ldrq(s, a->rd, a->pg, addr, a->dtype);
4860     }
4861     return true;
4862 }
4863 
4864 static bool trans_LD1RQ_zpri(DisasContext *s, arg_rpri_load *a)
4865 {
4866     if (!dc_isar_feature(aa64_sve, s)) {
4867         return false;
4868     }
4869     if (sve_access_check(s)) {
4870         TCGv_i64 addr = tcg_temp_new_i64();
4871         tcg_gen_addi_i64(addr, cpu_reg_sp(s, a->rn), a->imm * 16);
4872         do_ldrq(s, a->rd, a->pg, addr, a->dtype);
4873     }
4874     return true;
4875 }
4876 
4877 static void do_ldro(DisasContext *s, int zt, int pg, TCGv_i64 addr, int dtype)
4878 {
4879     unsigned vsz = vec_full_reg_size(s);
4880     unsigned vsz_r32;
4881     TCGv_ptr t_pg;
4882     int poff, doff;
4883 
4884     if (vsz < 32) {
4885         /*
4886          * Note that this UNDEFINED check comes after CheckSVEEnabled()
4887          * in the ARM pseudocode, which is the sve_access_check() done
4888          * in our caller.  We should not now return false from the caller.
4889          */
4890         unallocated_encoding(s);
4891         return;
4892     }
4893 
4894     /* Load the first octaword using the normal predicated load helpers.  */
4895 
4896     poff = pred_full_reg_offset(s, pg);
4897     if (vsz > 32) {
4898         /*
4899          * Zero-extend the first 32 bits of the predicate into a temporary.
4900          * This avoids triggering an assert making sure we don't have bits
4901          * set within a predicate beyond VQ, but we have lowered VQ to 2
4902          * for this load operation.
4903          */
4904         TCGv_i64 tmp = tcg_temp_new_i64();
4905 #if HOST_BIG_ENDIAN
4906         poff += 4;
4907 #endif
4908         tcg_gen_ld32u_i64(tmp, cpu_env, poff);
4909 
4910         poff = offsetof(CPUARMState, vfp.preg_tmp);
4911         tcg_gen_st_i64(tmp, cpu_env, poff);
4912     }
4913 
4914     t_pg = tcg_temp_new_ptr();
4915     tcg_gen_addi_ptr(t_pg, cpu_env, poff);
4916 
4917     gen_helper_gvec_mem *fn
4918         = ldr_fns[s->mte_active[0]][s->be_data == MO_BE][dtype][0];
4919     fn(cpu_env, t_pg, addr, tcg_constant_i32(simd_desc(32, 32, zt)));
4920 
4921     /*
4922      * Replicate that first octaword.
4923      * The replication happens in units of 32; if the full vector size
4924      * is not a multiple of 32, the final bits are zeroed.
4925      */
4926     doff = vec_full_reg_offset(s, zt);
4927     vsz_r32 = QEMU_ALIGN_DOWN(vsz, 32);
4928     if (vsz >= 64) {
4929         tcg_gen_gvec_dup_mem(5, doff + 32, doff, vsz_r32 - 32, vsz_r32 - 32);
4930     }
4931     vsz -= vsz_r32;
4932     if (vsz) {
4933         tcg_gen_gvec_dup_imm(MO_64, doff + vsz_r32, vsz, vsz, 0);
4934     }
4935 }
4936 
4937 static bool trans_LD1RO_zprr(DisasContext *s, arg_rprr_load *a)
4938 {
4939     if (!dc_isar_feature(aa64_sve_f64mm, s)) {
4940         return false;
4941     }
4942     if (a->rm == 31) {
4943         return false;
4944     }
4945     s->is_nonstreaming = true;
4946     if (sve_access_check(s)) {
4947         TCGv_i64 addr = tcg_temp_new_i64();
4948         tcg_gen_shli_i64(addr, cpu_reg(s, a->rm), dtype_msz(a->dtype));
4949         tcg_gen_add_i64(addr, addr, cpu_reg_sp(s, a->rn));
4950         do_ldro(s, a->rd, a->pg, addr, a->dtype);
4951     }
4952     return true;
4953 }
4954 
4955 static bool trans_LD1RO_zpri(DisasContext *s, arg_rpri_load *a)
4956 {
4957     if (!dc_isar_feature(aa64_sve_f64mm, s)) {
4958         return false;
4959     }
4960     s->is_nonstreaming = true;
4961     if (sve_access_check(s)) {
4962         TCGv_i64 addr = tcg_temp_new_i64();
4963         tcg_gen_addi_i64(addr, cpu_reg_sp(s, a->rn), a->imm * 32);
4964         do_ldro(s, a->rd, a->pg, addr, a->dtype);
4965     }
4966     return true;
4967 }
4968 
4969 /* Load and broadcast element.  */
4970 static bool trans_LD1R_zpri(DisasContext *s, arg_rpri_load *a)
4971 {
4972     unsigned vsz = vec_full_reg_size(s);
4973     unsigned psz = pred_full_reg_size(s);
4974     unsigned esz = dtype_esz[a->dtype];
4975     unsigned msz = dtype_msz(a->dtype);
4976     TCGLabel *over;
4977     TCGv_i64 temp, clean_addr;
4978 
4979     if (!dc_isar_feature(aa64_sve, s)) {
4980         return false;
4981     }
4982     if (!sve_access_check(s)) {
4983         return true;
4984     }
4985 
4986     over = gen_new_label();
4987 
4988     /* If the guarding predicate has no bits set, no load occurs.  */
4989     if (psz <= 8) {
4990         /* Reduce the pred_esz_masks value simply to reduce the
4991          * size of the code generated here.
4992          */
4993         uint64_t psz_mask = MAKE_64BIT_MASK(0, psz * 8);
4994         temp = tcg_temp_new_i64();
4995         tcg_gen_ld_i64(temp, cpu_env, pred_full_reg_offset(s, a->pg));
4996         tcg_gen_andi_i64(temp, temp, pred_esz_masks[esz] & psz_mask);
4997         tcg_gen_brcondi_i64(TCG_COND_EQ, temp, 0, over);
4998     } else {
4999         TCGv_i32 t32 = tcg_temp_new_i32();
5000         find_last_active(s, t32, esz, a->pg);
5001         tcg_gen_brcondi_i32(TCG_COND_LT, t32, 0, over);
5002     }
5003 
5004     /* Load the data.  */
5005     temp = tcg_temp_new_i64();
5006     tcg_gen_addi_i64(temp, cpu_reg_sp(s, a->rn), a->imm << msz);
5007     clean_addr = gen_mte_check1(s, temp, false, true, msz);
5008 
5009     tcg_gen_qemu_ld_i64(temp, clean_addr, get_mem_index(s),
5010                         finalize_memop(s, dtype_mop[a->dtype]));
5011 
5012     /* Broadcast to *all* elements.  */
5013     tcg_gen_gvec_dup_i64(esz, vec_full_reg_offset(s, a->rd),
5014                          vsz, vsz, temp);
5015 
5016     /* Zero the inactive elements.  */
5017     gen_set_label(over);
5018     return do_movz_zpz(s, a->rd, a->rd, a->pg, esz, false);
5019 }
5020 
5021 static void do_st_zpa(DisasContext *s, int zt, int pg, TCGv_i64 addr,
5022                       int msz, int esz, int nreg)
5023 {
5024     static gen_helper_gvec_mem * const fn_single[2][2][4][4] = {
5025         { { { gen_helper_sve_st1bb_r,
5026               gen_helper_sve_st1bh_r,
5027               gen_helper_sve_st1bs_r,
5028               gen_helper_sve_st1bd_r },
5029             { NULL,
5030               gen_helper_sve_st1hh_le_r,
5031               gen_helper_sve_st1hs_le_r,
5032               gen_helper_sve_st1hd_le_r },
5033             { NULL, NULL,
5034               gen_helper_sve_st1ss_le_r,
5035               gen_helper_sve_st1sd_le_r },
5036             { NULL, NULL, NULL,
5037               gen_helper_sve_st1dd_le_r } },
5038           { { gen_helper_sve_st1bb_r,
5039               gen_helper_sve_st1bh_r,
5040               gen_helper_sve_st1bs_r,
5041               gen_helper_sve_st1bd_r },
5042             { NULL,
5043               gen_helper_sve_st1hh_be_r,
5044               gen_helper_sve_st1hs_be_r,
5045               gen_helper_sve_st1hd_be_r },
5046             { NULL, NULL,
5047               gen_helper_sve_st1ss_be_r,
5048               gen_helper_sve_st1sd_be_r },
5049             { NULL, NULL, NULL,
5050               gen_helper_sve_st1dd_be_r } } },
5051 
5052         { { { gen_helper_sve_st1bb_r_mte,
5053               gen_helper_sve_st1bh_r_mte,
5054               gen_helper_sve_st1bs_r_mte,
5055               gen_helper_sve_st1bd_r_mte },
5056             { NULL,
5057               gen_helper_sve_st1hh_le_r_mte,
5058               gen_helper_sve_st1hs_le_r_mte,
5059               gen_helper_sve_st1hd_le_r_mte },
5060             { NULL, NULL,
5061               gen_helper_sve_st1ss_le_r_mte,
5062               gen_helper_sve_st1sd_le_r_mte },
5063             { NULL, NULL, NULL,
5064               gen_helper_sve_st1dd_le_r_mte } },
5065           { { gen_helper_sve_st1bb_r_mte,
5066               gen_helper_sve_st1bh_r_mte,
5067               gen_helper_sve_st1bs_r_mte,
5068               gen_helper_sve_st1bd_r_mte },
5069             { NULL,
5070               gen_helper_sve_st1hh_be_r_mte,
5071               gen_helper_sve_st1hs_be_r_mte,
5072               gen_helper_sve_st1hd_be_r_mte },
5073             { NULL, NULL,
5074               gen_helper_sve_st1ss_be_r_mte,
5075               gen_helper_sve_st1sd_be_r_mte },
5076             { NULL, NULL, NULL,
5077               gen_helper_sve_st1dd_be_r_mte } } },
5078     };
5079     static gen_helper_gvec_mem * const fn_multiple[2][2][3][4] = {
5080         { { { gen_helper_sve_st2bb_r,
5081               gen_helper_sve_st2hh_le_r,
5082               gen_helper_sve_st2ss_le_r,
5083               gen_helper_sve_st2dd_le_r },
5084             { gen_helper_sve_st3bb_r,
5085               gen_helper_sve_st3hh_le_r,
5086               gen_helper_sve_st3ss_le_r,
5087               gen_helper_sve_st3dd_le_r },
5088             { gen_helper_sve_st4bb_r,
5089               gen_helper_sve_st4hh_le_r,
5090               gen_helper_sve_st4ss_le_r,
5091               gen_helper_sve_st4dd_le_r } },
5092           { { gen_helper_sve_st2bb_r,
5093               gen_helper_sve_st2hh_be_r,
5094               gen_helper_sve_st2ss_be_r,
5095               gen_helper_sve_st2dd_be_r },
5096             { gen_helper_sve_st3bb_r,
5097               gen_helper_sve_st3hh_be_r,
5098               gen_helper_sve_st3ss_be_r,
5099               gen_helper_sve_st3dd_be_r },
5100             { gen_helper_sve_st4bb_r,
5101               gen_helper_sve_st4hh_be_r,
5102               gen_helper_sve_st4ss_be_r,
5103               gen_helper_sve_st4dd_be_r } } },
5104         { { { gen_helper_sve_st2bb_r_mte,
5105               gen_helper_sve_st2hh_le_r_mte,
5106               gen_helper_sve_st2ss_le_r_mte,
5107               gen_helper_sve_st2dd_le_r_mte },
5108             { gen_helper_sve_st3bb_r_mte,
5109               gen_helper_sve_st3hh_le_r_mte,
5110               gen_helper_sve_st3ss_le_r_mte,
5111               gen_helper_sve_st3dd_le_r_mte },
5112             { gen_helper_sve_st4bb_r_mte,
5113               gen_helper_sve_st4hh_le_r_mte,
5114               gen_helper_sve_st4ss_le_r_mte,
5115               gen_helper_sve_st4dd_le_r_mte } },
5116           { { gen_helper_sve_st2bb_r_mte,
5117               gen_helper_sve_st2hh_be_r_mte,
5118               gen_helper_sve_st2ss_be_r_mte,
5119               gen_helper_sve_st2dd_be_r_mte },
5120             { gen_helper_sve_st3bb_r_mte,
5121               gen_helper_sve_st3hh_be_r_mte,
5122               gen_helper_sve_st3ss_be_r_mte,
5123               gen_helper_sve_st3dd_be_r_mte },
5124             { gen_helper_sve_st4bb_r_mte,
5125               gen_helper_sve_st4hh_be_r_mte,
5126               gen_helper_sve_st4ss_be_r_mte,
5127               gen_helper_sve_st4dd_be_r_mte } } },
5128     };
5129     gen_helper_gvec_mem *fn;
5130     int be = s->be_data == MO_BE;
5131 
5132     if (nreg == 0) {
5133         /* ST1 */
5134         fn = fn_single[s->mte_active[0]][be][msz][esz];
5135         nreg = 1;
5136     } else {
5137         /* ST2, ST3, ST4 -- msz == esz, enforced by encoding */
5138         assert(msz == esz);
5139         fn = fn_multiple[s->mte_active[0]][be][nreg - 1][msz];
5140     }
5141     assert(fn != NULL);
5142     do_mem_zpa(s, zt, pg, addr, msz_dtype(s, msz), nreg, true, fn);
5143 }
5144 
5145 static bool trans_ST_zprr(DisasContext *s, arg_rprr_store *a)
5146 {
5147     if (!dc_isar_feature(aa64_sve, s)) {
5148         return false;
5149     }
5150     if (a->rm == 31 || a->msz > a->esz) {
5151         return false;
5152     }
5153     if (sve_access_check(s)) {
5154         TCGv_i64 addr = tcg_temp_new_i64();
5155         tcg_gen_shli_i64(addr, cpu_reg(s, a->rm), a->msz);
5156         tcg_gen_add_i64(addr, addr, cpu_reg_sp(s, a->rn));
5157         do_st_zpa(s, a->rd, a->pg, addr, a->msz, a->esz, a->nreg);
5158     }
5159     return true;
5160 }
5161 
5162 static bool trans_ST_zpri(DisasContext *s, arg_rpri_store *a)
5163 {
5164     if (!dc_isar_feature(aa64_sve, s)) {
5165         return false;
5166     }
5167     if (a->msz > a->esz) {
5168         return false;
5169     }
5170     if (sve_access_check(s)) {
5171         int vsz = vec_full_reg_size(s);
5172         int elements = vsz >> a->esz;
5173         TCGv_i64 addr = tcg_temp_new_i64();
5174 
5175         tcg_gen_addi_i64(addr, cpu_reg_sp(s, a->rn),
5176                          (a->imm * elements * (a->nreg + 1)) << a->msz);
5177         do_st_zpa(s, a->rd, a->pg, addr, a->msz, a->esz, a->nreg);
5178     }
5179     return true;
5180 }
5181 
5182 /*
5183  *** SVE gather loads / scatter stores
5184  */
5185 
5186 static void do_mem_zpz(DisasContext *s, int zt, int pg, int zm,
5187                        int scale, TCGv_i64 scalar, int msz, bool is_write,
5188                        gen_helper_gvec_mem_scatter *fn)
5189 {
5190     unsigned vsz = vec_full_reg_size(s);
5191     TCGv_ptr t_zm = tcg_temp_new_ptr();
5192     TCGv_ptr t_pg = tcg_temp_new_ptr();
5193     TCGv_ptr t_zt = tcg_temp_new_ptr();
5194     int desc = 0;
5195 
5196     if (s->mte_active[0]) {
5197         desc = FIELD_DP32(desc, MTEDESC, MIDX, get_mem_index(s));
5198         desc = FIELD_DP32(desc, MTEDESC, TBI, s->tbid);
5199         desc = FIELD_DP32(desc, MTEDESC, TCMA, s->tcma);
5200         desc = FIELD_DP32(desc, MTEDESC, WRITE, is_write);
5201         desc = FIELD_DP32(desc, MTEDESC, SIZEM1, (1 << msz) - 1);
5202         desc <<= SVE_MTEDESC_SHIFT;
5203     }
5204     desc = simd_desc(vsz, vsz, desc | scale);
5205 
5206     tcg_gen_addi_ptr(t_pg, cpu_env, pred_full_reg_offset(s, pg));
5207     tcg_gen_addi_ptr(t_zm, cpu_env, vec_full_reg_offset(s, zm));
5208     tcg_gen_addi_ptr(t_zt, cpu_env, vec_full_reg_offset(s, zt));
5209     fn(cpu_env, t_zt, t_pg, t_zm, scalar, tcg_constant_i32(desc));
5210 }
5211 
5212 /* Indexed by [mte][be][ff][xs][u][msz].  */
5213 static gen_helper_gvec_mem_scatter * const
5214 gather_load_fn32[2][2][2][2][2][3] = {
5215     { /* MTE Inactive */
5216         { /* Little-endian */
5217             { { { gen_helper_sve_ldbss_zsu,
5218                   gen_helper_sve_ldhss_le_zsu,
5219                   NULL, },
5220                 { gen_helper_sve_ldbsu_zsu,
5221                   gen_helper_sve_ldhsu_le_zsu,
5222                   gen_helper_sve_ldss_le_zsu, } },
5223               { { gen_helper_sve_ldbss_zss,
5224                   gen_helper_sve_ldhss_le_zss,
5225                   NULL, },
5226                 { gen_helper_sve_ldbsu_zss,
5227                   gen_helper_sve_ldhsu_le_zss,
5228                   gen_helper_sve_ldss_le_zss, } } },
5229 
5230             /* First-fault */
5231             { { { gen_helper_sve_ldffbss_zsu,
5232                   gen_helper_sve_ldffhss_le_zsu,
5233                   NULL, },
5234                 { gen_helper_sve_ldffbsu_zsu,
5235                   gen_helper_sve_ldffhsu_le_zsu,
5236                   gen_helper_sve_ldffss_le_zsu, } },
5237               { { gen_helper_sve_ldffbss_zss,
5238                   gen_helper_sve_ldffhss_le_zss,
5239                   NULL, },
5240                 { gen_helper_sve_ldffbsu_zss,
5241                   gen_helper_sve_ldffhsu_le_zss,
5242                   gen_helper_sve_ldffss_le_zss, } } } },
5243 
5244         { /* Big-endian */
5245             { { { gen_helper_sve_ldbss_zsu,
5246                   gen_helper_sve_ldhss_be_zsu,
5247                   NULL, },
5248                 { gen_helper_sve_ldbsu_zsu,
5249                   gen_helper_sve_ldhsu_be_zsu,
5250                   gen_helper_sve_ldss_be_zsu, } },
5251               { { gen_helper_sve_ldbss_zss,
5252                   gen_helper_sve_ldhss_be_zss,
5253                   NULL, },
5254                 { gen_helper_sve_ldbsu_zss,
5255                   gen_helper_sve_ldhsu_be_zss,
5256                   gen_helper_sve_ldss_be_zss, } } },
5257 
5258             /* First-fault */
5259             { { { gen_helper_sve_ldffbss_zsu,
5260                   gen_helper_sve_ldffhss_be_zsu,
5261                   NULL, },
5262                 { gen_helper_sve_ldffbsu_zsu,
5263                   gen_helper_sve_ldffhsu_be_zsu,
5264                   gen_helper_sve_ldffss_be_zsu, } },
5265               { { gen_helper_sve_ldffbss_zss,
5266                   gen_helper_sve_ldffhss_be_zss,
5267                   NULL, },
5268                 { gen_helper_sve_ldffbsu_zss,
5269                   gen_helper_sve_ldffhsu_be_zss,
5270                   gen_helper_sve_ldffss_be_zss, } } } } },
5271     { /* MTE Active */
5272         { /* Little-endian */
5273             { { { gen_helper_sve_ldbss_zsu_mte,
5274                   gen_helper_sve_ldhss_le_zsu_mte,
5275                   NULL, },
5276                 { gen_helper_sve_ldbsu_zsu_mte,
5277                   gen_helper_sve_ldhsu_le_zsu_mte,
5278                   gen_helper_sve_ldss_le_zsu_mte, } },
5279               { { gen_helper_sve_ldbss_zss_mte,
5280                   gen_helper_sve_ldhss_le_zss_mte,
5281                   NULL, },
5282                 { gen_helper_sve_ldbsu_zss_mte,
5283                   gen_helper_sve_ldhsu_le_zss_mte,
5284                   gen_helper_sve_ldss_le_zss_mte, } } },
5285 
5286             /* First-fault */
5287             { { { gen_helper_sve_ldffbss_zsu_mte,
5288                   gen_helper_sve_ldffhss_le_zsu_mte,
5289                   NULL, },
5290                 { gen_helper_sve_ldffbsu_zsu_mte,
5291                   gen_helper_sve_ldffhsu_le_zsu_mte,
5292                   gen_helper_sve_ldffss_le_zsu_mte, } },
5293               { { gen_helper_sve_ldffbss_zss_mte,
5294                   gen_helper_sve_ldffhss_le_zss_mte,
5295                   NULL, },
5296                 { gen_helper_sve_ldffbsu_zss_mte,
5297                   gen_helper_sve_ldffhsu_le_zss_mte,
5298                   gen_helper_sve_ldffss_le_zss_mte, } } } },
5299 
5300         { /* Big-endian */
5301             { { { gen_helper_sve_ldbss_zsu_mte,
5302                   gen_helper_sve_ldhss_be_zsu_mte,
5303                   NULL, },
5304                 { gen_helper_sve_ldbsu_zsu_mte,
5305                   gen_helper_sve_ldhsu_be_zsu_mte,
5306                   gen_helper_sve_ldss_be_zsu_mte, } },
5307               { { gen_helper_sve_ldbss_zss_mte,
5308                   gen_helper_sve_ldhss_be_zss_mte,
5309                   NULL, },
5310                 { gen_helper_sve_ldbsu_zss_mte,
5311                   gen_helper_sve_ldhsu_be_zss_mte,
5312                   gen_helper_sve_ldss_be_zss_mte, } } },
5313 
5314             /* First-fault */
5315             { { { gen_helper_sve_ldffbss_zsu_mte,
5316                   gen_helper_sve_ldffhss_be_zsu_mte,
5317                   NULL, },
5318                 { gen_helper_sve_ldffbsu_zsu_mte,
5319                   gen_helper_sve_ldffhsu_be_zsu_mte,
5320                   gen_helper_sve_ldffss_be_zsu_mte, } },
5321               { { gen_helper_sve_ldffbss_zss_mte,
5322                   gen_helper_sve_ldffhss_be_zss_mte,
5323                   NULL, },
5324                 { gen_helper_sve_ldffbsu_zss_mte,
5325                   gen_helper_sve_ldffhsu_be_zss_mte,
5326                   gen_helper_sve_ldffss_be_zss_mte, } } } } },
5327 };
5328 
5329 /* Note that we overload xs=2 to indicate 64-bit offset.  */
5330 static gen_helper_gvec_mem_scatter * const
5331 gather_load_fn64[2][2][2][3][2][4] = {
5332     { /* MTE Inactive */
5333         { /* Little-endian */
5334             { { { gen_helper_sve_ldbds_zsu,
5335                   gen_helper_sve_ldhds_le_zsu,
5336                   gen_helper_sve_ldsds_le_zsu,
5337                   NULL, },
5338                 { gen_helper_sve_ldbdu_zsu,
5339                   gen_helper_sve_ldhdu_le_zsu,
5340                   gen_helper_sve_ldsdu_le_zsu,
5341                   gen_helper_sve_lddd_le_zsu, } },
5342               { { gen_helper_sve_ldbds_zss,
5343                   gen_helper_sve_ldhds_le_zss,
5344                   gen_helper_sve_ldsds_le_zss,
5345                   NULL, },
5346                 { gen_helper_sve_ldbdu_zss,
5347                   gen_helper_sve_ldhdu_le_zss,
5348                   gen_helper_sve_ldsdu_le_zss,
5349                   gen_helper_sve_lddd_le_zss, } },
5350               { { gen_helper_sve_ldbds_zd,
5351                   gen_helper_sve_ldhds_le_zd,
5352                   gen_helper_sve_ldsds_le_zd,
5353                   NULL, },
5354                 { gen_helper_sve_ldbdu_zd,
5355                   gen_helper_sve_ldhdu_le_zd,
5356                   gen_helper_sve_ldsdu_le_zd,
5357                   gen_helper_sve_lddd_le_zd, } } },
5358 
5359             /* First-fault */
5360             { { { gen_helper_sve_ldffbds_zsu,
5361                   gen_helper_sve_ldffhds_le_zsu,
5362                   gen_helper_sve_ldffsds_le_zsu,
5363                   NULL, },
5364                 { gen_helper_sve_ldffbdu_zsu,
5365                   gen_helper_sve_ldffhdu_le_zsu,
5366                   gen_helper_sve_ldffsdu_le_zsu,
5367                   gen_helper_sve_ldffdd_le_zsu, } },
5368               { { gen_helper_sve_ldffbds_zss,
5369                   gen_helper_sve_ldffhds_le_zss,
5370                   gen_helper_sve_ldffsds_le_zss,
5371                   NULL, },
5372                 { gen_helper_sve_ldffbdu_zss,
5373                   gen_helper_sve_ldffhdu_le_zss,
5374                   gen_helper_sve_ldffsdu_le_zss,
5375                   gen_helper_sve_ldffdd_le_zss, } },
5376               { { gen_helper_sve_ldffbds_zd,
5377                   gen_helper_sve_ldffhds_le_zd,
5378                   gen_helper_sve_ldffsds_le_zd,
5379                   NULL, },
5380                 { gen_helper_sve_ldffbdu_zd,
5381                   gen_helper_sve_ldffhdu_le_zd,
5382                   gen_helper_sve_ldffsdu_le_zd,
5383                   gen_helper_sve_ldffdd_le_zd, } } } },
5384         { /* Big-endian */
5385             { { { gen_helper_sve_ldbds_zsu,
5386                   gen_helper_sve_ldhds_be_zsu,
5387                   gen_helper_sve_ldsds_be_zsu,
5388                   NULL, },
5389                 { gen_helper_sve_ldbdu_zsu,
5390                   gen_helper_sve_ldhdu_be_zsu,
5391                   gen_helper_sve_ldsdu_be_zsu,
5392                   gen_helper_sve_lddd_be_zsu, } },
5393               { { gen_helper_sve_ldbds_zss,
5394                   gen_helper_sve_ldhds_be_zss,
5395                   gen_helper_sve_ldsds_be_zss,
5396                   NULL, },
5397                 { gen_helper_sve_ldbdu_zss,
5398                   gen_helper_sve_ldhdu_be_zss,
5399                   gen_helper_sve_ldsdu_be_zss,
5400                   gen_helper_sve_lddd_be_zss, } },
5401               { { gen_helper_sve_ldbds_zd,
5402                   gen_helper_sve_ldhds_be_zd,
5403                   gen_helper_sve_ldsds_be_zd,
5404                   NULL, },
5405                 { gen_helper_sve_ldbdu_zd,
5406                   gen_helper_sve_ldhdu_be_zd,
5407                   gen_helper_sve_ldsdu_be_zd,
5408                   gen_helper_sve_lddd_be_zd, } } },
5409 
5410             /* First-fault */
5411             { { { gen_helper_sve_ldffbds_zsu,
5412                   gen_helper_sve_ldffhds_be_zsu,
5413                   gen_helper_sve_ldffsds_be_zsu,
5414                   NULL, },
5415                 { gen_helper_sve_ldffbdu_zsu,
5416                   gen_helper_sve_ldffhdu_be_zsu,
5417                   gen_helper_sve_ldffsdu_be_zsu,
5418                   gen_helper_sve_ldffdd_be_zsu, } },
5419               { { gen_helper_sve_ldffbds_zss,
5420                   gen_helper_sve_ldffhds_be_zss,
5421                   gen_helper_sve_ldffsds_be_zss,
5422                   NULL, },
5423                 { gen_helper_sve_ldffbdu_zss,
5424                   gen_helper_sve_ldffhdu_be_zss,
5425                   gen_helper_sve_ldffsdu_be_zss,
5426                   gen_helper_sve_ldffdd_be_zss, } },
5427               { { gen_helper_sve_ldffbds_zd,
5428                   gen_helper_sve_ldffhds_be_zd,
5429                   gen_helper_sve_ldffsds_be_zd,
5430                   NULL, },
5431                 { gen_helper_sve_ldffbdu_zd,
5432                   gen_helper_sve_ldffhdu_be_zd,
5433                   gen_helper_sve_ldffsdu_be_zd,
5434                   gen_helper_sve_ldffdd_be_zd, } } } } },
5435     { /* MTE Active */
5436         { /* Little-endian */
5437             { { { gen_helper_sve_ldbds_zsu_mte,
5438                   gen_helper_sve_ldhds_le_zsu_mte,
5439                   gen_helper_sve_ldsds_le_zsu_mte,
5440                   NULL, },
5441                 { gen_helper_sve_ldbdu_zsu_mte,
5442                   gen_helper_sve_ldhdu_le_zsu_mte,
5443                   gen_helper_sve_ldsdu_le_zsu_mte,
5444                   gen_helper_sve_lddd_le_zsu_mte, } },
5445               { { gen_helper_sve_ldbds_zss_mte,
5446                   gen_helper_sve_ldhds_le_zss_mte,
5447                   gen_helper_sve_ldsds_le_zss_mte,
5448                   NULL, },
5449                 { gen_helper_sve_ldbdu_zss_mte,
5450                   gen_helper_sve_ldhdu_le_zss_mte,
5451                   gen_helper_sve_ldsdu_le_zss_mte,
5452                   gen_helper_sve_lddd_le_zss_mte, } },
5453               { { gen_helper_sve_ldbds_zd_mte,
5454                   gen_helper_sve_ldhds_le_zd_mte,
5455                   gen_helper_sve_ldsds_le_zd_mte,
5456                   NULL, },
5457                 { gen_helper_sve_ldbdu_zd_mte,
5458                   gen_helper_sve_ldhdu_le_zd_mte,
5459                   gen_helper_sve_ldsdu_le_zd_mte,
5460                   gen_helper_sve_lddd_le_zd_mte, } } },
5461 
5462             /* First-fault */
5463             { { { gen_helper_sve_ldffbds_zsu_mte,
5464                   gen_helper_sve_ldffhds_le_zsu_mte,
5465                   gen_helper_sve_ldffsds_le_zsu_mte,
5466                   NULL, },
5467                 { gen_helper_sve_ldffbdu_zsu_mte,
5468                   gen_helper_sve_ldffhdu_le_zsu_mte,
5469                   gen_helper_sve_ldffsdu_le_zsu_mte,
5470                   gen_helper_sve_ldffdd_le_zsu_mte, } },
5471               { { gen_helper_sve_ldffbds_zss_mte,
5472                   gen_helper_sve_ldffhds_le_zss_mte,
5473                   gen_helper_sve_ldffsds_le_zss_mte,
5474                   NULL, },
5475                 { gen_helper_sve_ldffbdu_zss_mte,
5476                   gen_helper_sve_ldffhdu_le_zss_mte,
5477                   gen_helper_sve_ldffsdu_le_zss_mte,
5478                   gen_helper_sve_ldffdd_le_zss_mte, } },
5479               { { gen_helper_sve_ldffbds_zd_mte,
5480                   gen_helper_sve_ldffhds_le_zd_mte,
5481                   gen_helper_sve_ldffsds_le_zd_mte,
5482                   NULL, },
5483                 { gen_helper_sve_ldffbdu_zd_mte,
5484                   gen_helper_sve_ldffhdu_le_zd_mte,
5485                   gen_helper_sve_ldffsdu_le_zd_mte,
5486                   gen_helper_sve_ldffdd_le_zd_mte, } } } },
5487         { /* Big-endian */
5488             { { { gen_helper_sve_ldbds_zsu_mte,
5489                   gen_helper_sve_ldhds_be_zsu_mte,
5490                   gen_helper_sve_ldsds_be_zsu_mte,
5491                   NULL, },
5492                 { gen_helper_sve_ldbdu_zsu_mte,
5493                   gen_helper_sve_ldhdu_be_zsu_mte,
5494                   gen_helper_sve_ldsdu_be_zsu_mte,
5495                   gen_helper_sve_lddd_be_zsu_mte, } },
5496               { { gen_helper_sve_ldbds_zss_mte,
5497                   gen_helper_sve_ldhds_be_zss_mte,
5498                   gen_helper_sve_ldsds_be_zss_mte,
5499                   NULL, },
5500                 { gen_helper_sve_ldbdu_zss_mte,
5501                   gen_helper_sve_ldhdu_be_zss_mte,
5502                   gen_helper_sve_ldsdu_be_zss_mte,
5503                   gen_helper_sve_lddd_be_zss_mte, } },
5504               { { gen_helper_sve_ldbds_zd_mte,
5505                   gen_helper_sve_ldhds_be_zd_mte,
5506                   gen_helper_sve_ldsds_be_zd_mte,
5507                   NULL, },
5508                 { gen_helper_sve_ldbdu_zd_mte,
5509                   gen_helper_sve_ldhdu_be_zd_mte,
5510                   gen_helper_sve_ldsdu_be_zd_mte,
5511                   gen_helper_sve_lddd_be_zd_mte, } } },
5512 
5513             /* First-fault */
5514             { { { gen_helper_sve_ldffbds_zsu_mte,
5515                   gen_helper_sve_ldffhds_be_zsu_mte,
5516                   gen_helper_sve_ldffsds_be_zsu_mte,
5517                   NULL, },
5518                 { gen_helper_sve_ldffbdu_zsu_mte,
5519                   gen_helper_sve_ldffhdu_be_zsu_mte,
5520                   gen_helper_sve_ldffsdu_be_zsu_mte,
5521                   gen_helper_sve_ldffdd_be_zsu_mte, } },
5522               { { gen_helper_sve_ldffbds_zss_mte,
5523                   gen_helper_sve_ldffhds_be_zss_mte,
5524                   gen_helper_sve_ldffsds_be_zss_mte,
5525                   NULL, },
5526                 { gen_helper_sve_ldffbdu_zss_mte,
5527                   gen_helper_sve_ldffhdu_be_zss_mte,
5528                   gen_helper_sve_ldffsdu_be_zss_mte,
5529                   gen_helper_sve_ldffdd_be_zss_mte, } },
5530               { { gen_helper_sve_ldffbds_zd_mte,
5531                   gen_helper_sve_ldffhds_be_zd_mte,
5532                   gen_helper_sve_ldffsds_be_zd_mte,
5533                   NULL, },
5534                 { gen_helper_sve_ldffbdu_zd_mte,
5535                   gen_helper_sve_ldffhdu_be_zd_mte,
5536                   gen_helper_sve_ldffsdu_be_zd_mte,
5537                   gen_helper_sve_ldffdd_be_zd_mte, } } } } },
5538 };
5539 
5540 static bool trans_LD1_zprz(DisasContext *s, arg_LD1_zprz *a)
5541 {
5542     gen_helper_gvec_mem_scatter *fn = NULL;
5543     bool be = s->be_data == MO_BE;
5544     bool mte = s->mte_active[0];
5545 
5546     if (!dc_isar_feature(aa64_sve, s)) {
5547         return false;
5548     }
5549     s->is_nonstreaming = true;
5550     if (!sve_access_check(s)) {
5551         return true;
5552     }
5553 
5554     switch (a->esz) {
5555     case MO_32:
5556         fn = gather_load_fn32[mte][be][a->ff][a->xs][a->u][a->msz];
5557         break;
5558     case MO_64:
5559         fn = gather_load_fn64[mte][be][a->ff][a->xs][a->u][a->msz];
5560         break;
5561     }
5562     assert(fn != NULL);
5563 
5564     do_mem_zpz(s, a->rd, a->pg, a->rm, a->scale * a->msz,
5565                cpu_reg_sp(s, a->rn), a->msz, false, fn);
5566     return true;
5567 }
5568 
5569 static bool trans_LD1_zpiz(DisasContext *s, arg_LD1_zpiz *a)
5570 {
5571     gen_helper_gvec_mem_scatter *fn = NULL;
5572     bool be = s->be_data == MO_BE;
5573     bool mte = s->mte_active[0];
5574 
5575     if (a->esz < a->msz || (a->esz == a->msz && !a->u)) {
5576         return false;
5577     }
5578     if (!dc_isar_feature(aa64_sve, s)) {
5579         return false;
5580     }
5581     s->is_nonstreaming = true;
5582     if (!sve_access_check(s)) {
5583         return true;
5584     }
5585 
5586     switch (a->esz) {
5587     case MO_32:
5588         fn = gather_load_fn32[mte][be][a->ff][0][a->u][a->msz];
5589         break;
5590     case MO_64:
5591         fn = gather_load_fn64[mte][be][a->ff][2][a->u][a->msz];
5592         break;
5593     }
5594     assert(fn != NULL);
5595 
5596     /* Treat LD1_zpiz (zn[x] + imm) the same way as LD1_zprz (rn + zm[x])
5597      * by loading the immediate into the scalar parameter.
5598      */
5599     do_mem_zpz(s, a->rd, a->pg, a->rn, 0,
5600                tcg_constant_i64(a->imm << a->msz), a->msz, false, fn);
5601     return true;
5602 }
5603 
5604 static bool trans_LDNT1_zprz(DisasContext *s, arg_LD1_zprz *a)
5605 {
5606     gen_helper_gvec_mem_scatter *fn = NULL;
5607     bool be = s->be_data == MO_BE;
5608     bool mte = s->mte_active[0];
5609 
5610     if (a->esz < a->msz + !a->u) {
5611         return false;
5612     }
5613     if (!dc_isar_feature(aa64_sve2, s)) {
5614         return false;
5615     }
5616     s->is_nonstreaming = true;
5617     if (!sve_access_check(s)) {
5618         return true;
5619     }
5620 
5621     switch (a->esz) {
5622     case MO_32:
5623         fn = gather_load_fn32[mte][be][0][0][a->u][a->msz];
5624         break;
5625     case MO_64:
5626         fn = gather_load_fn64[mte][be][0][2][a->u][a->msz];
5627         break;
5628     }
5629     assert(fn != NULL);
5630 
5631     do_mem_zpz(s, a->rd, a->pg, a->rn, 0,
5632                cpu_reg(s, a->rm), a->msz, false, fn);
5633     return true;
5634 }
5635 
5636 /* Indexed by [mte][be][xs][msz].  */
5637 static gen_helper_gvec_mem_scatter * const scatter_store_fn32[2][2][2][3] = {
5638     { /* MTE Inactive */
5639         { /* Little-endian */
5640             { gen_helper_sve_stbs_zsu,
5641               gen_helper_sve_sths_le_zsu,
5642               gen_helper_sve_stss_le_zsu, },
5643             { gen_helper_sve_stbs_zss,
5644               gen_helper_sve_sths_le_zss,
5645               gen_helper_sve_stss_le_zss, } },
5646         { /* Big-endian */
5647             { gen_helper_sve_stbs_zsu,
5648               gen_helper_sve_sths_be_zsu,
5649               gen_helper_sve_stss_be_zsu, },
5650             { gen_helper_sve_stbs_zss,
5651               gen_helper_sve_sths_be_zss,
5652               gen_helper_sve_stss_be_zss, } } },
5653     { /* MTE Active */
5654         { /* Little-endian */
5655             { gen_helper_sve_stbs_zsu_mte,
5656               gen_helper_sve_sths_le_zsu_mte,
5657               gen_helper_sve_stss_le_zsu_mte, },
5658             { gen_helper_sve_stbs_zss_mte,
5659               gen_helper_sve_sths_le_zss_mte,
5660               gen_helper_sve_stss_le_zss_mte, } },
5661         { /* Big-endian */
5662             { gen_helper_sve_stbs_zsu_mte,
5663               gen_helper_sve_sths_be_zsu_mte,
5664               gen_helper_sve_stss_be_zsu_mte, },
5665             { gen_helper_sve_stbs_zss_mte,
5666               gen_helper_sve_sths_be_zss_mte,
5667               gen_helper_sve_stss_be_zss_mte, } } },
5668 };
5669 
5670 /* Note that we overload xs=2 to indicate 64-bit offset.  */
5671 static gen_helper_gvec_mem_scatter * const scatter_store_fn64[2][2][3][4] = {
5672     { /* MTE Inactive */
5673          { /* Little-endian */
5674              { gen_helper_sve_stbd_zsu,
5675                gen_helper_sve_sthd_le_zsu,
5676                gen_helper_sve_stsd_le_zsu,
5677                gen_helper_sve_stdd_le_zsu, },
5678              { gen_helper_sve_stbd_zss,
5679                gen_helper_sve_sthd_le_zss,
5680                gen_helper_sve_stsd_le_zss,
5681                gen_helper_sve_stdd_le_zss, },
5682              { gen_helper_sve_stbd_zd,
5683                gen_helper_sve_sthd_le_zd,
5684                gen_helper_sve_stsd_le_zd,
5685                gen_helper_sve_stdd_le_zd, } },
5686          { /* Big-endian */
5687              { gen_helper_sve_stbd_zsu,
5688                gen_helper_sve_sthd_be_zsu,
5689                gen_helper_sve_stsd_be_zsu,
5690                gen_helper_sve_stdd_be_zsu, },
5691              { gen_helper_sve_stbd_zss,
5692                gen_helper_sve_sthd_be_zss,
5693                gen_helper_sve_stsd_be_zss,
5694                gen_helper_sve_stdd_be_zss, },
5695              { gen_helper_sve_stbd_zd,
5696                gen_helper_sve_sthd_be_zd,
5697                gen_helper_sve_stsd_be_zd,
5698                gen_helper_sve_stdd_be_zd, } } },
5699     { /* MTE Inactive */
5700          { /* Little-endian */
5701              { gen_helper_sve_stbd_zsu_mte,
5702                gen_helper_sve_sthd_le_zsu_mte,
5703                gen_helper_sve_stsd_le_zsu_mte,
5704                gen_helper_sve_stdd_le_zsu_mte, },
5705              { gen_helper_sve_stbd_zss_mte,
5706                gen_helper_sve_sthd_le_zss_mte,
5707                gen_helper_sve_stsd_le_zss_mte,
5708                gen_helper_sve_stdd_le_zss_mte, },
5709              { gen_helper_sve_stbd_zd_mte,
5710                gen_helper_sve_sthd_le_zd_mte,
5711                gen_helper_sve_stsd_le_zd_mte,
5712                gen_helper_sve_stdd_le_zd_mte, } },
5713          { /* Big-endian */
5714              { gen_helper_sve_stbd_zsu_mte,
5715                gen_helper_sve_sthd_be_zsu_mte,
5716                gen_helper_sve_stsd_be_zsu_mte,
5717                gen_helper_sve_stdd_be_zsu_mte, },
5718              { gen_helper_sve_stbd_zss_mte,
5719                gen_helper_sve_sthd_be_zss_mte,
5720                gen_helper_sve_stsd_be_zss_mte,
5721                gen_helper_sve_stdd_be_zss_mte, },
5722              { gen_helper_sve_stbd_zd_mte,
5723                gen_helper_sve_sthd_be_zd_mte,
5724                gen_helper_sve_stsd_be_zd_mte,
5725                gen_helper_sve_stdd_be_zd_mte, } } },
5726 };
5727 
5728 static bool trans_ST1_zprz(DisasContext *s, arg_ST1_zprz *a)
5729 {
5730     gen_helper_gvec_mem_scatter *fn;
5731     bool be = s->be_data == MO_BE;
5732     bool mte = s->mte_active[0];
5733 
5734     if (a->esz < a->msz || (a->msz == 0 && a->scale)) {
5735         return false;
5736     }
5737     if (!dc_isar_feature(aa64_sve, s)) {
5738         return false;
5739     }
5740     s->is_nonstreaming = true;
5741     if (!sve_access_check(s)) {
5742         return true;
5743     }
5744     switch (a->esz) {
5745     case MO_32:
5746         fn = scatter_store_fn32[mte][be][a->xs][a->msz];
5747         break;
5748     case MO_64:
5749         fn = scatter_store_fn64[mte][be][a->xs][a->msz];
5750         break;
5751     default:
5752         g_assert_not_reached();
5753     }
5754     do_mem_zpz(s, a->rd, a->pg, a->rm, a->scale * a->msz,
5755                cpu_reg_sp(s, a->rn), a->msz, true, fn);
5756     return true;
5757 }
5758 
5759 static bool trans_ST1_zpiz(DisasContext *s, arg_ST1_zpiz *a)
5760 {
5761     gen_helper_gvec_mem_scatter *fn = NULL;
5762     bool be = s->be_data == MO_BE;
5763     bool mte = s->mte_active[0];
5764 
5765     if (a->esz < a->msz) {
5766         return false;
5767     }
5768     if (!dc_isar_feature(aa64_sve, s)) {
5769         return false;
5770     }
5771     s->is_nonstreaming = true;
5772     if (!sve_access_check(s)) {
5773         return true;
5774     }
5775 
5776     switch (a->esz) {
5777     case MO_32:
5778         fn = scatter_store_fn32[mte][be][0][a->msz];
5779         break;
5780     case MO_64:
5781         fn = scatter_store_fn64[mte][be][2][a->msz];
5782         break;
5783     }
5784     assert(fn != NULL);
5785 
5786     /* Treat ST1_zpiz (zn[x] + imm) the same way as ST1_zprz (rn + zm[x])
5787      * by loading the immediate into the scalar parameter.
5788      */
5789     do_mem_zpz(s, a->rd, a->pg, a->rn, 0,
5790                tcg_constant_i64(a->imm << a->msz), a->msz, true, fn);
5791     return true;
5792 }
5793 
5794 static bool trans_STNT1_zprz(DisasContext *s, arg_ST1_zprz *a)
5795 {
5796     gen_helper_gvec_mem_scatter *fn;
5797     bool be = s->be_data == MO_BE;
5798     bool mte = s->mte_active[0];
5799 
5800     if (a->esz < a->msz) {
5801         return false;
5802     }
5803     if (!dc_isar_feature(aa64_sve2, s)) {
5804         return false;
5805     }
5806     s->is_nonstreaming = true;
5807     if (!sve_access_check(s)) {
5808         return true;
5809     }
5810 
5811     switch (a->esz) {
5812     case MO_32:
5813         fn = scatter_store_fn32[mte][be][0][a->msz];
5814         break;
5815     case MO_64:
5816         fn = scatter_store_fn64[mte][be][2][a->msz];
5817         break;
5818     default:
5819         g_assert_not_reached();
5820     }
5821 
5822     do_mem_zpz(s, a->rd, a->pg, a->rn, 0,
5823                cpu_reg(s, a->rm), a->msz, true, fn);
5824     return true;
5825 }
5826 
5827 /*
5828  * Prefetches
5829  */
5830 
5831 static bool trans_PRF(DisasContext *s, arg_PRF *a)
5832 {
5833     if (!dc_isar_feature(aa64_sve, s)) {
5834         return false;
5835     }
5836     /* Prefetch is a nop within QEMU.  */
5837     (void)sve_access_check(s);
5838     return true;
5839 }
5840 
5841 static bool trans_PRF_rr(DisasContext *s, arg_PRF_rr *a)
5842 {
5843     if (a->rm == 31 || !dc_isar_feature(aa64_sve, s)) {
5844         return false;
5845     }
5846     /* Prefetch is a nop within QEMU.  */
5847     (void)sve_access_check(s);
5848     return true;
5849 }
5850 
5851 static bool trans_PRF_ns(DisasContext *s, arg_PRF_ns *a)
5852 {
5853     if (!dc_isar_feature(aa64_sve, s)) {
5854         return false;
5855     }
5856     /* Prefetch is a nop within QEMU.  */
5857     s->is_nonstreaming = true;
5858     (void)sve_access_check(s);
5859     return true;
5860 }
5861 
5862 /*
5863  * Move Prefix
5864  *
5865  * TODO: The implementation so far could handle predicated merging movprfx.
5866  * The helper functions as written take an extra source register to
5867  * use in the operation, but the result is only written when predication
5868  * succeeds.  For unpredicated movprfx, we need to rearrange the helpers
5869  * to allow the final write back to the destination to be unconditional.
5870  * For predicated zeroing movprfx, we need to rearrange the helpers to
5871  * allow the final write back to zero inactives.
5872  *
5873  * In the meantime, just emit the moves.
5874  */
5875 
5876 TRANS_FEAT(MOVPRFX, aa64_sve, do_mov_z, a->rd, a->rn)
5877 TRANS_FEAT(MOVPRFX_m, aa64_sve, do_sel_z, a->rd, a->rn, a->rd, a->pg, a->esz)
5878 TRANS_FEAT(MOVPRFX_z, aa64_sve, do_movz_zpz, a->rd, a->rn, a->pg, a->esz, false)
5879 
5880 /*
5881  * SVE2 Integer Multiply - Unpredicated
5882  */
5883 
5884 TRANS_FEAT(MUL_zzz, aa64_sve2, gen_gvec_fn_arg_zzz, tcg_gen_gvec_mul, a)
5885 
5886 static gen_helper_gvec_3 * const smulh_zzz_fns[4] = {
5887     gen_helper_gvec_smulh_b, gen_helper_gvec_smulh_h,
5888     gen_helper_gvec_smulh_s, gen_helper_gvec_smulh_d,
5889 };
5890 TRANS_FEAT(SMULH_zzz, aa64_sve2, gen_gvec_ool_arg_zzz,
5891            smulh_zzz_fns[a->esz], a, 0)
5892 
5893 static gen_helper_gvec_3 * const umulh_zzz_fns[4] = {
5894     gen_helper_gvec_umulh_b, gen_helper_gvec_umulh_h,
5895     gen_helper_gvec_umulh_s, gen_helper_gvec_umulh_d,
5896 };
5897 TRANS_FEAT(UMULH_zzz, aa64_sve2, gen_gvec_ool_arg_zzz,
5898            umulh_zzz_fns[a->esz], a, 0)
5899 
5900 TRANS_FEAT(PMUL_zzz, aa64_sve2, gen_gvec_ool_arg_zzz,
5901            gen_helper_gvec_pmul_b, a, 0)
5902 
5903 static gen_helper_gvec_3 * const sqdmulh_zzz_fns[4] = {
5904     gen_helper_sve2_sqdmulh_b, gen_helper_sve2_sqdmulh_h,
5905     gen_helper_sve2_sqdmulh_s, gen_helper_sve2_sqdmulh_d,
5906 };
5907 TRANS_FEAT(SQDMULH_zzz, aa64_sve2, gen_gvec_ool_arg_zzz,
5908            sqdmulh_zzz_fns[a->esz], a, 0)
5909 
5910 static gen_helper_gvec_3 * const sqrdmulh_zzz_fns[4] = {
5911     gen_helper_sve2_sqrdmulh_b, gen_helper_sve2_sqrdmulh_h,
5912     gen_helper_sve2_sqrdmulh_s, gen_helper_sve2_sqrdmulh_d,
5913 };
5914 TRANS_FEAT(SQRDMULH_zzz, aa64_sve2, gen_gvec_ool_arg_zzz,
5915            sqrdmulh_zzz_fns[a->esz], a, 0)
5916 
5917 /*
5918  * SVE2 Integer - Predicated
5919  */
5920 
5921 static gen_helper_gvec_4 * const sadlp_fns[4] = {
5922     NULL,                          gen_helper_sve2_sadalp_zpzz_h,
5923     gen_helper_sve2_sadalp_zpzz_s, gen_helper_sve2_sadalp_zpzz_d,
5924 };
5925 TRANS_FEAT(SADALP_zpzz, aa64_sve2, gen_gvec_ool_arg_zpzz,
5926            sadlp_fns[a->esz], a, 0)
5927 
5928 static gen_helper_gvec_4 * const uadlp_fns[4] = {
5929     NULL,                          gen_helper_sve2_uadalp_zpzz_h,
5930     gen_helper_sve2_uadalp_zpzz_s, gen_helper_sve2_uadalp_zpzz_d,
5931 };
5932 TRANS_FEAT(UADALP_zpzz, aa64_sve2, gen_gvec_ool_arg_zpzz,
5933            uadlp_fns[a->esz], a, 0)
5934 
5935 /*
5936  * SVE2 integer unary operations (predicated)
5937  */
5938 
5939 TRANS_FEAT(URECPE, aa64_sve2, gen_gvec_ool_arg_zpz,
5940            a->esz == 2 ? gen_helper_sve2_urecpe_s : NULL, a, 0)
5941 
5942 TRANS_FEAT(URSQRTE, aa64_sve2, gen_gvec_ool_arg_zpz,
5943            a->esz == 2 ? gen_helper_sve2_ursqrte_s : NULL, a, 0)
5944 
5945 static gen_helper_gvec_3 * const sqabs_fns[4] = {
5946     gen_helper_sve2_sqabs_b, gen_helper_sve2_sqabs_h,
5947     gen_helper_sve2_sqabs_s, gen_helper_sve2_sqabs_d,
5948 };
5949 TRANS_FEAT(SQABS, aa64_sve2, gen_gvec_ool_arg_zpz, sqabs_fns[a->esz], a, 0)
5950 
5951 static gen_helper_gvec_3 * const sqneg_fns[4] = {
5952     gen_helper_sve2_sqneg_b, gen_helper_sve2_sqneg_h,
5953     gen_helper_sve2_sqneg_s, gen_helper_sve2_sqneg_d,
5954 };
5955 TRANS_FEAT(SQNEG, aa64_sve2, gen_gvec_ool_arg_zpz, sqneg_fns[a->esz], a, 0)
5956 
5957 DO_ZPZZ(SQSHL, aa64_sve2, sve2_sqshl)
5958 DO_ZPZZ(SQRSHL, aa64_sve2, sve2_sqrshl)
5959 DO_ZPZZ(SRSHL, aa64_sve2, sve2_srshl)
5960 
5961 DO_ZPZZ(UQSHL, aa64_sve2, sve2_uqshl)
5962 DO_ZPZZ(UQRSHL, aa64_sve2, sve2_uqrshl)
5963 DO_ZPZZ(URSHL, aa64_sve2, sve2_urshl)
5964 
5965 DO_ZPZZ(SHADD, aa64_sve2, sve2_shadd)
5966 DO_ZPZZ(SRHADD, aa64_sve2, sve2_srhadd)
5967 DO_ZPZZ(SHSUB, aa64_sve2, sve2_shsub)
5968 
5969 DO_ZPZZ(UHADD, aa64_sve2, sve2_uhadd)
5970 DO_ZPZZ(URHADD, aa64_sve2, sve2_urhadd)
5971 DO_ZPZZ(UHSUB, aa64_sve2, sve2_uhsub)
5972 
5973 DO_ZPZZ(ADDP, aa64_sve2, sve2_addp)
5974 DO_ZPZZ(SMAXP, aa64_sve2, sve2_smaxp)
5975 DO_ZPZZ(UMAXP, aa64_sve2, sve2_umaxp)
5976 DO_ZPZZ(SMINP, aa64_sve2, sve2_sminp)
5977 DO_ZPZZ(UMINP, aa64_sve2, sve2_uminp)
5978 
5979 DO_ZPZZ(SQADD_zpzz, aa64_sve2, sve2_sqadd)
5980 DO_ZPZZ(UQADD_zpzz, aa64_sve2, sve2_uqadd)
5981 DO_ZPZZ(SQSUB_zpzz, aa64_sve2, sve2_sqsub)
5982 DO_ZPZZ(UQSUB_zpzz, aa64_sve2, sve2_uqsub)
5983 DO_ZPZZ(SUQADD, aa64_sve2, sve2_suqadd)
5984 DO_ZPZZ(USQADD, aa64_sve2, sve2_usqadd)
5985 
5986 /*
5987  * SVE2 Widening Integer Arithmetic
5988  */
5989 
5990 static gen_helper_gvec_3 * const saddl_fns[4] = {
5991     NULL,                    gen_helper_sve2_saddl_h,
5992     gen_helper_sve2_saddl_s, gen_helper_sve2_saddl_d,
5993 };
5994 TRANS_FEAT(SADDLB, aa64_sve2, gen_gvec_ool_arg_zzz,
5995            saddl_fns[a->esz], a, 0)
5996 TRANS_FEAT(SADDLT, aa64_sve2, gen_gvec_ool_arg_zzz,
5997            saddl_fns[a->esz], a, 3)
5998 TRANS_FEAT(SADDLBT, aa64_sve2, gen_gvec_ool_arg_zzz,
5999            saddl_fns[a->esz], a, 2)
6000 
6001 static gen_helper_gvec_3 * const ssubl_fns[4] = {
6002     NULL,                    gen_helper_sve2_ssubl_h,
6003     gen_helper_sve2_ssubl_s, gen_helper_sve2_ssubl_d,
6004 };
6005 TRANS_FEAT(SSUBLB, aa64_sve2, gen_gvec_ool_arg_zzz,
6006            ssubl_fns[a->esz], a, 0)
6007 TRANS_FEAT(SSUBLT, aa64_sve2, gen_gvec_ool_arg_zzz,
6008            ssubl_fns[a->esz], a, 3)
6009 TRANS_FEAT(SSUBLBT, aa64_sve2, gen_gvec_ool_arg_zzz,
6010            ssubl_fns[a->esz], a, 2)
6011 TRANS_FEAT(SSUBLTB, aa64_sve2, gen_gvec_ool_arg_zzz,
6012            ssubl_fns[a->esz], a, 1)
6013 
6014 static gen_helper_gvec_3 * const sabdl_fns[4] = {
6015     NULL,                    gen_helper_sve2_sabdl_h,
6016     gen_helper_sve2_sabdl_s, gen_helper_sve2_sabdl_d,
6017 };
6018 TRANS_FEAT(SABDLB, aa64_sve2, gen_gvec_ool_arg_zzz,
6019            sabdl_fns[a->esz], a, 0)
6020 TRANS_FEAT(SABDLT, aa64_sve2, gen_gvec_ool_arg_zzz,
6021            sabdl_fns[a->esz], a, 3)
6022 
6023 static gen_helper_gvec_3 * const uaddl_fns[4] = {
6024     NULL,                    gen_helper_sve2_uaddl_h,
6025     gen_helper_sve2_uaddl_s, gen_helper_sve2_uaddl_d,
6026 };
6027 TRANS_FEAT(UADDLB, aa64_sve2, gen_gvec_ool_arg_zzz,
6028            uaddl_fns[a->esz], a, 0)
6029 TRANS_FEAT(UADDLT, aa64_sve2, gen_gvec_ool_arg_zzz,
6030            uaddl_fns[a->esz], a, 3)
6031 
6032 static gen_helper_gvec_3 * const usubl_fns[4] = {
6033     NULL,                    gen_helper_sve2_usubl_h,
6034     gen_helper_sve2_usubl_s, gen_helper_sve2_usubl_d,
6035 };
6036 TRANS_FEAT(USUBLB, aa64_sve2, gen_gvec_ool_arg_zzz,
6037            usubl_fns[a->esz], a, 0)
6038 TRANS_FEAT(USUBLT, aa64_sve2, gen_gvec_ool_arg_zzz,
6039            usubl_fns[a->esz], a, 3)
6040 
6041 static gen_helper_gvec_3 * const uabdl_fns[4] = {
6042     NULL,                    gen_helper_sve2_uabdl_h,
6043     gen_helper_sve2_uabdl_s, gen_helper_sve2_uabdl_d,
6044 };
6045 TRANS_FEAT(UABDLB, aa64_sve2, gen_gvec_ool_arg_zzz,
6046            uabdl_fns[a->esz], a, 0)
6047 TRANS_FEAT(UABDLT, aa64_sve2, gen_gvec_ool_arg_zzz,
6048            uabdl_fns[a->esz], a, 3)
6049 
6050 static gen_helper_gvec_3 * const sqdmull_fns[4] = {
6051     NULL,                          gen_helper_sve2_sqdmull_zzz_h,
6052     gen_helper_sve2_sqdmull_zzz_s, gen_helper_sve2_sqdmull_zzz_d,
6053 };
6054 TRANS_FEAT(SQDMULLB_zzz, aa64_sve2, gen_gvec_ool_arg_zzz,
6055            sqdmull_fns[a->esz], a, 0)
6056 TRANS_FEAT(SQDMULLT_zzz, aa64_sve2, gen_gvec_ool_arg_zzz,
6057            sqdmull_fns[a->esz], a, 3)
6058 
6059 static gen_helper_gvec_3 * const smull_fns[4] = {
6060     NULL,                        gen_helper_sve2_smull_zzz_h,
6061     gen_helper_sve2_smull_zzz_s, gen_helper_sve2_smull_zzz_d,
6062 };
6063 TRANS_FEAT(SMULLB_zzz, aa64_sve2, gen_gvec_ool_arg_zzz,
6064            smull_fns[a->esz], a, 0)
6065 TRANS_FEAT(SMULLT_zzz, aa64_sve2, gen_gvec_ool_arg_zzz,
6066            smull_fns[a->esz], a, 3)
6067 
6068 static gen_helper_gvec_3 * const umull_fns[4] = {
6069     NULL,                        gen_helper_sve2_umull_zzz_h,
6070     gen_helper_sve2_umull_zzz_s, gen_helper_sve2_umull_zzz_d,
6071 };
6072 TRANS_FEAT(UMULLB_zzz, aa64_sve2, gen_gvec_ool_arg_zzz,
6073            umull_fns[a->esz], a, 0)
6074 TRANS_FEAT(UMULLT_zzz, aa64_sve2, gen_gvec_ool_arg_zzz,
6075            umull_fns[a->esz], a, 3)
6076 
6077 static gen_helper_gvec_3 * const eoril_fns[4] = {
6078     gen_helper_sve2_eoril_b, gen_helper_sve2_eoril_h,
6079     gen_helper_sve2_eoril_s, gen_helper_sve2_eoril_d,
6080 };
6081 TRANS_FEAT(EORBT, aa64_sve2, gen_gvec_ool_arg_zzz, eoril_fns[a->esz], a, 2)
6082 TRANS_FEAT(EORTB, aa64_sve2, gen_gvec_ool_arg_zzz, eoril_fns[a->esz], a, 1)
6083 
6084 static bool do_trans_pmull(DisasContext *s, arg_rrr_esz *a, bool sel)
6085 {
6086     static gen_helper_gvec_3 * const fns[4] = {
6087         gen_helper_gvec_pmull_q, gen_helper_sve2_pmull_h,
6088         NULL,                    gen_helper_sve2_pmull_d,
6089     };
6090 
6091     if (a->esz == 0) {
6092         if (!dc_isar_feature(aa64_sve2_pmull128, s)) {
6093             return false;
6094         }
6095         s->is_nonstreaming = true;
6096     } else if (!dc_isar_feature(aa64_sve, s)) {
6097         return false;
6098     }
6099     return gen_gvec_ool_arg_zzz(s, fns[a->esz], a, sel);
6100 }
6101 
6102 TRANS_FEAT(PMULLB, aa64_sve2, do_trans_pmull, a, false)
6103 TRANS_FEAT(PMULLT, aa64_sve2, do_trans_pmull, a, true)
6104 
6105 static gen_helper_gvec_3 * const saddw_fns[4] = {
6106     NULL,                    gen_helper_sve2_saddw_h,
6107     gen_helper_sve2_saddw_s, gen_helper_sve2_saddw_d,
6108 };
6109 TRANS_FEAT(SADDWB, aa64_sve2, gen_gvec_ool_arg_zzz, saddw_fns[a->esz], a, 0)
6110 TRANS_FEAT(SADDWT, aa64_sve2, gen_gvec_ool_arg_zzz, saddw_fns[a->esz], a, 1)
6111 
6112 static gen_helper_gvec_3 * const ssubw_fns[4] = {
6113     NULL,                    gen_helper_sve2_ssubw_h,
6114     gen_helper_sve2_ssubw_s, gen_helper_sve2_ssubw_d,
6115 };
6116 TRANS_FEAT(SSUBWB, aa64_sve2, gen_gvec_ool_arg_zzz, ssubw_fns[a->esz], a, 0)
6117 TRANS_FEAT(SSUBWT, aa64_sve2, gen_gvec_ool_arg_zzz, ssubw_fns[a->esz], a, 1)
6118 
6119 static gen_helper_gvec_3 * const uaddw_fns[4] = {
6120     NULL,                    gen_helper_sve2_uaddw_h,
6121     gen_helper_sve2_uaddw_s, gen_helper_sve2_uaddw_d,
6122 };
6123 TRANS_FEAT(UADDWB, aa64_sve2, gen_gvec_ool_arg_zzz, uaddw_fns[a->esz], a, 0)
6124 TRANS_FEAT(UADDWT, aa64_sve2, gen_gvec_ool_arg_zzz, uaddw_fns[a->esz], a, 1)
6125 
6126 static gen_helper_gvec_3 * const usubw_fns[4] = {
6127     NULL,                    gen_helper_sve2_usubw_h,
6128     gen_helper_sve2_usubw_s, gen_helper_sve2_usubw_d,
6129 };
6130 TRANS_FEAT(USUBWB, aa64_sve2, gen_gvec_ool_arg_zzz, usubw_fns[a->esz], a, 0)
6131 TRANS_FEAT(USUBWT, aa64_sve2, gen_gvec_ool_arg_zzz, usubw_fns[a->esz], a, 1)
6132 
6133 static void gen_sshll_vec(unsigned vece, TCGv_vec d, TCGv_vec n, int64_t imm)
6134 {
6135     int top = imm & 1;
6136     int shl = imm >> 1;
6137     int halfbits = 4 << vece;
6138 
6139     if (top) {
6140         if (shl == halfbits) {
6141             TCGv_vec t = tcg_temp_new_vec_matching(d);
6142             tcg_gen_dupi_vec(vece, t, MAKE_64BIT_MASK(halfbits, halfbits));
6143             tcg_gen_and_vec(vece, d, n, t);
6144         } else {
6145             tcg_gen_sari_vec(vece, d, n, halfbits);
6146             tcg_gen_shli_vec(vece, d, d, shl);
6147         }
6148     } else {
6149         tcg_gen_shli_vec(vece, d, n, halfbits);
6150         tcg_gen_sari_vec(vece, d, d, halfbits - shl);
6151     }
6152 }
6153 
6154 static void gen_ushll_i64(unsigned vece, TCGv_i64 d, TCGv_i64 n, int imm)
6155 {
6156     int halfbits = 4 << vece;
6157     int top = imm & 1;
6158     int shl = (imm >> 1);
6159     int shift;
6160     uint64_t mask;
6161 
6162     mask = MAKE_64BIT_MASK(0, halfbits);
6163     mask <<= shl;
6164     mask = dup_const(vece, mask);
6165 
6166     shift = shl - top * halfbits;
6167     if (shift < 0) {
6168         tcg_gen_shri_i64(d, n, -shift);
6169     } else {
6170         tcg_gen_shli_i64(d, n, shift);
6171     }
6172     tcg_gen_andi_i64(d, d, mask);
6173 }
6174 
6175 static void gen_ushll16_i64(TCGv_i64 d, TCGv_i64 n, int64_t imm)
6176 {
6177     gen_ushll_i64(MO_16, d, n, imm);
6178 }
6179 
6180 static void gen_ushll32_i64(TCGv_i64 d, TCGv_i64 n, int64_t imm)
6181 {
6182     gen_ushll_i64(MO_32, d, n, imm);
6183 }
6184 
6185 static void gen_ushll64_i64(TCGv_i64 d, TCGv_i64 n, int64_t imm)
6186 {
6187     gen_ushll_i64(MO_64, d, n, imm);
6188 }
6189 
6190 static void gen_ushll_vec(unsigned vece, TCGv_vec d, TCGv_vec n, int64_t imm)
6191 {
6192     int halfbits = 4 << vece;
6193     int top = imm & 1;
6194     int shl = imm >> 1;
6195 
6196     if (top) {
6197         if (shl == halfbits) {
6198             TCGv_vec t = tcg_temp_new_vec_matching(d);
6199             tcg_gen_dupi_vec(vece, t, MAKE_64BIT_MASK(halfbits, halfbits));
6200             tcg_gen_and_vec(vece, d, n, t);
6201         } else {
6202             tcg_gen_shri_vec(vece, d, n, halfbits);
6203             tcg_gen_shli_vec(vece, d, d, shl);
6204         }
6205     } else {
6206         if (shl == 0) {
6207             TCGv_vec t = tcg_temp_new_vec_matching(d);
6208             tcg_gen_dupi_vec(vece, t, MAKE_64BIT_MASK(0, halfbits));
6209             tcg_gen_and_vec(vece, d, n, t);
6210         } else {
6211             tcg_gen_shli_vec(vece, d, n, halfbits);
6212             tcg_gen_shri_vec(vece, d, d, halfbits - shl);
6213         }
6214     }
6215 }
6216 
6217 static bool do_shll_tb(DisasContext *s, arg_rri_esz *a,
6218                        const GVecGen2i ops[3], bool sel)
6219 {
6220 
6221     if (a->esz < 0 || a->esz > 2) {
6222         return false;
6223     }
6224     if (sve_access_check(s)) {
6225         unsigned vsz = vec_full_reg_size(s);
6226         tcg_gen_gvec_2i(vec_full_reg_offset(s, a->rd),
6227                         vec_full_reg_offset(s, a->rn),
6228                         vsz, vsz, (a->imm << 1) | sel,
6229                         &ops[a->esz]);
6230     }
6231     return true;
6232 }
6233 
6234 static const TCGOpcode sshll_list[] = {
6235     INDEX_op_shli_vec, INDEX_op_sari_vec, 0
6236 };
6237 static const GVecGen2i sshll_ops[3] = {
6238     { .fniv = gen_sshll_vec,
6239       .opt_opc = sshll_list,
6240       .fno = gen_helper_sve2_sshll_h,
6241       .vece = MO_16 },
6242     { .fniv = gen_sshll_vec,
6243       .opt_opc = sshll_list,
6244       .fno = gen_helper_sve2_sshll_s,
6245       .vece = MO_32 },
6246     { .fniv = gen_sshll_vec,
6247       .opt_opc = sshll_list,
6248       .fno = gen_helper_sve2_sshll_d,
6249       .vece = MO_64 }
6250 };
6251 TRANS_FEAT(SSHLLB, aa64_sve2, do_shll_tb, a, sshll_ops, false)
6252 TRANS_FEAT(SSHLLT, aa64_sve2, do_shll_tb, a, sshll_ops, true)
6253 
6254 static const TCGOpcode ushll_list[] = {
6255     INDEX_op_shli_vec, INDEX_op_shri_vec, 0
6256 };
6257 static const GVecGen2i ushll_ops[3] = {
6258     { .fni8 = gen_ushll16_i64,
6259       .fniv = gen_ushll_vec,
6260       .opt_opc = ushll_list,
6261       .fno = gen_helper_sve2_ushll_h,
6262       .vece = MO_16 },
6263     { .fni8 = gen_ushll32_i64,
6264       .fniv = gen_ushll_vec,
6265       .opt_opc = ushll_list,
6266       .fno = gen_helper_sve2_ushll_s,
6267       .vece = MO_32 },
6268     { .fni8 = gen_ushll64_i64,
6269       .fniv = gen_ushll_vec,
6270       .opt_opc = ushll_list,
6271       .fno = gen_helper_sve2_ushll_d,
6272       .vece = MO_64 },
6273 };
6274 TRANS_FEAT(USHLLB, aa64_sve2, do_shll_tb, a, ushll_ops, false)
6275 TRANS_FEAT(USHLLT, aa64_sve2, do_shll_tb, a, ushll_ops, true)
6276 
6277 static gen_helper_gvec_3 * const bext_fns[4] = {
6278     gen_helper_sve2_bext_b, gen_helper_sve2_bext_h,
6279     gen_helper_sve2_bext_s, gen_helper_sve2_bext_d,
6280 };
6281 TRANS_FEAT_NONSTREAMING(BEXT, aa64_sve2_bitperm, gen_gvec_ool_arg_zzz,
6282                         bext_fns[a->esz], a, 0)
6283 
6284 static gen_helper_gvec_3 * const bdep_fns[4] = {
6285     gen_helper_sve2_bdep_b, gen_helper_sve2_bdep_h,
6286     gen_helper_sve2_bdep_s, gen_helper_sve2_bdep_d,
6287 };
6288 TRANS_FEAT_NONSTREAMING(BDEP, aa64_sve2_bitperm, gen_gvec_ool_arg_zzz,
6289                         bdep_fns[a->esz], a, 0)
6290 
6291 static gen_helper_gvec_3 * const bgrp_fns[4] = {
6292     gen_helper_sve2_bgrp_b, gen_helper_sve2_bgrp_h,
6293     gen_helper_sve2_bgrp_s, gen_helper_sve2_bgrp_d,
6294 };
6295 TRANS_FEAT_NONSTREAMING(BGRP, aa64_sve2_bitperm, gen_gvec_ool_arg_zzz,
6296                         bgrp_fns[a->esz], a, 0)
6297 
6298 static gen_helper_gvec_3 * const cadd_fns[4] = {
6299     gen_helper_sve2_cadd_b, gen_helper_sve2_cadd_h,
6300     gen_helper_sve2_cadd_s, gen_helper_sve2_cadd_d,
6301 };
6302 TRANS_FEAT(CADD_rot90, aa64_sve2, gen_gvec_ool_arg_zzz,
6303            cadd_fns[a->esz], a, 0)
6304 TRANS_FEAT(CADD_rot270, aa64_sve2, gen_gvec_ool_arg_zzz,
6305            cadd_fns[a->esz], a, 1)
6306 
6307 static gen_helper_gvec_3 * const sqcadd_fns[4] = {
6308     gen_helper_sve2_sqcadd_b, gen_helper_sve2_sqcadd_h,
6309     gen_helper_sve2_sqcadd_s, gen_helper_sve2_sqcadd_d,
6310 };
6311 TRANS_FEAT(SQCADD_rot90, aa64_sve2, gen_gvec_ool_arg_zzz,
6312            sqcadd_fns[a->esz], a, 0)
6313 TRANS_FEAT(SQCADD_rot270, aa64_sve2, gen_gvec_ool_arg_zzz,
6314            sqcadd_fns[a->esz], a, 1)
6315 
6316 static gen_helper_gvec_4 * const sabal_fns[4] = {
6317     NULL,                    gen_helper_sve2_sabal_h,
6318     gen_helper_sve2_sabal_s, gen_helper_sve2_sabal_d,
6319 };
6320 TRANS_FEAT(SABALB, aa64_sve2, gen_gvec_ool_arg_zzzz, sabal_fns[a->esz], a, 0)
6321 TRANS_FEAT(SABALT, aa64_sve2, gen_gvec_ool_arg_zzzz, sabal_fns[a->esz], a, 1)
6322 
6323 static gen_helper_gvec_4 * const uabal_fns[4] = {
6324     NULL,                    gen_helper_sve2_uabal_h,
6325     gen_helper_sve2_uabal_s, gen_helper_sve2_uabal_d,
6326 };
6327 TRANS_FEAT(UABALB, aa64_sve2, gen_gvec_ool_arg_zzzz, uabal_fns[a->esz], a, 0)
6328 TRANS_FEAT(UABALT, aa64_sve2, gen_gvec_ool_arg_zzzz, uabal_fns[a->esz], a, 1)
6329 
6330 static bool do_adcl(DisasContext *s, arg_rrrr_esz *a, bool sel)
6331 {
6332     static gen_helper_gvec_4 * const fns[2] = {
6333         gen_helper_sve2_adcl_s,
6334         gen_helper_sve2_adcl_d,
6335     };
6336     /*
6337      * Note that in this case the ESZ field encodes both size and sign.
6338      * Split out 'subtract' into bit 1 of the data field for the helper.
6339      */
6340     return gen_gvec_ool_arg_zzzz(s, fns[a->esz & 1], a, (a->esz & 2) | sel);
6341 }
6342 
6343 TRANS_FEAT(ADCLB, aa64_sve2, do_adcl, a, false)
6344 TRANS_FEAT(ADCLT, aa64_sve2, do_adcl, a, true)
6345 
6346 TRANS_FEAT(SSRA, aa64_sve2, gen_gvec_fn_arg_zzi, gen_gvec_ssra, a)
6347 TRANS_FEAT(USRA, aa64_sve2, gen_gvec_fn_arg_zzi, gen_gvec_usra, a)
6348 TRANS_FEAT(SRSRA, aa64_sve2, gen_gvec_fn_arg_zzi, gen_gvec_srsra, a)
6349 TRANS_FEAT(URSRA, aa64_sve2, gen_gvec_fn_arg_zzi, gen_gvec_ursra, a)
6350 TRANS_FEAT(SRI, aa64_sve2, gen_gvec_fn_arg_zzi, gen_gvec_sri, a)
6351 TRANS_FEAT(SLI, aa64_sve2, gen_gvec_fn_arg_zzi, gen_gvec_sli, a)
6352 
6353 TRANS_FEAT(SABA, aa64_sve2, gen_gvec_fn_arg_zzz, gen_gvec_saba, a)
6354 TRANS_FEAT(UABA, aa64_sve2, gen_gvec_fn_arg_zzz, gen_gvec_uaba, a)
6355 
6356 static bool do_narrow_extract(DisasContext *s, arg_rri_esz *a,
6357                               const GVecGen2 ops[3])
6358 {
6359     if (a->esz < 0 || a->esz > MO_32 || a->imm != 0) {
6360         return false;
6361     }
6362     if (sve_access_check(s)) {
6363         unsigned vsz = vec_full_reg_size(s);
6364         tcg_gen_gvec_2(vec_full_reg_offset(s, a->rd),
6365                         vec_full_reg_offset(s, a->rn),
6366                         vsz, vsz, &ops[a->esz]);
6367     }
6368     return true;
6369 }
6370 
6371 static const TCGOpcode sqxtn_list[] = {
6372     INDEX_op_shli_vec, INDEX_op_smin_vec, INDEX_op_smax_vec, 0
6373 };
6374 
6375 static void gen_sqxtnb_vec(unsigned vece, TCGv_vec d, TCGv_vec n)
6376 {
6377     TCGv_vec t = tcg_temp_new_vec_matching(d);
6378     int halfbits = 4 << vece;
6379     int64_t mask = (1ull << halfbits) - 1;
6380     int64_t min = -1ull << (halfbits - 1);
6381     int64_t max = -min - 1;
6382 
6383     tcg_gen_dupi_vec(vece, t, min);
6384     tcg_gen_smax_vec(vece, d, n, t);
6385     tcg_gen_dupi_vec(vece, t, max);
6386     tcg_gen_smin_vec(vece, d, d, t);
6387     tcg_gen_dupi_vec(vece, t, mask);
6388     tcg_gen_and_vec(vece, d, d, t);
6389 }
6390 
6391 static const GVecGen2 sqxtnb_ops[3] = {
6392     { .fniv = gen_sqxtnb_vec,
6393       .opt_opc = sqxtn_list,
6394       .fno = gen_helper_sve2_sqxtnb_h,
6395       .vece = MO_16 },
6396     { .fniv = gen_sqxtnb_vec,
6397       .opt_opc = sqxtn_list,
6398       .fno = gen_helper_sve2_sqxtnb_s,
6399       .vece = MO_32 },
6400     { .fniv = gen_sqxtnb_vec,
6401       .opt_opc = sqxtn_list,
6402       .fno = gen_helper_sve2_sqxtnb_d,
6403       .vece = MO_64 },
6404 };
6405 TRANS_FEAT(SQXTNB, aa64_sve2, do_narrow_extract, a, sqxtnb_ops)
6406 
6407 static void gen_sqxtnt_vec(unsigned vece, TCGv_vec d, TCGv_vec n)
6408 {
6409     TCGv_vec t = tcg_temp_new_vec_matching(d);
6410     int halfbits = 4 << vece;
6411     int64_t mask = (1ull << halfbits) - 1;
6412     int64_t min = -1ull << (halfbits - 1);
6413     int64_t max = -min - 1;
6414 
6415     tcg_gen_dupi_vec(vece, t, min);
6416     tcg_gen_smax_vec(vece, n, n, t);
6417     tcg_gen_dupi_vec(vece, t, max);
6418     tcg_gen_smin_vec(vece, n, n, t);
6419     tcg_gen_shli_vec(vece, n, n, halfbits);
6420     tcg_gen_dupi_vec(vece, t, mask);
6421     tcg_gen_bitsel_vec(vece, d, t, d, n);
6422 }
6423 
6424 static const GVecGen2 sqxtnt_ops[3] = {
6425     { .fniv = gen_sqxtnt_vec,
6426       .opt_opc = sqxtn_list,
6427       .load_dest = true,
6428       .fno = gen_helper_sve2_sqxtnt_h,
6429       .vece = MO_16 },
6430     { .fniv = gen_sqxtnt_vec,
6431       .opt_opc = sqxtn_list,
6432       .load_dest = true,
6433       .fno = gen_helper_sve2_sqxtnt_s,
6434       .vece = MO_32 },
6435     { .fniv = gen_sqxtnt_vec,
6436       .opt_opc = sqxtn_list,
6437       .load_dest = true,
6438       .fno = gen_helper_sve2_sqxtnt_d,
6439       .vece = MO_64 },
6440 };
6441 TRANS_FEAT(SQXTNT, aa64_sve2, do_narrow_extract, a, sqxtnt_ops)
6442 
6443 static const TCGOpcode uqxtn_list[] = {
6444     INDEX_op_shli_vec, INDEX_op_umin_vec, 0
6445 };
6446 
6447 static void gen_uqxtnb_vec(unsigned vece, TCGv_vec d, TCGv_vec n)
6448 {
6449     TCGv_vec t = tcg_temp_new_vec_matching(d);
6450     int halfbits = 4 << vece;
6451     int64_t max = (1ull << halfbits) - 1;
6452 
6453     tcg_gen_dupi_vec(vece, t, max);
6454     tcg_gen_umin_vec(vece, d, n, t);
6455 }
6456 
6457 static const GVecGen2 uqxtnb_ops[3] = {
6458     { .fniv = gen_uqxtnb_vec,
6459       .opt_opc = uqxtn_list,
6460       .fno = gen_helper_sve2_uqxtnb_h,
6461       .vece = MO_16 },
6462     { .fniv = gen_uqxtnb_vec,
6463       .opt_opc = uqxtn_list,
6464       .fno = gen_helper_sve2_uqxtnb_s,
6465       .vece = MO_32 },
6466     { .fniv = gen_uqxtnb_vec,
6467       .opt_opc = uqxtn_list,
6468       .fno = gen_helper_sve2_uqxtnb_d,
6469       .vece = MO_64 },
6470 };
6471 TRANS_FEAT(UQXTNB, aa64_sve2, do_narrow_extract, a, uqxtnb_ops)
6472 
6473 static void gen_uqxtnt_vec(unsigned vece, TCGv_vec d, TCGv_vec n)
6474 {
6475     TCGv_vec t = tcg_temp_new_vec_matching(d);
6476     int halfbits = 4 << vece;
6477     int64_t max = (1ull << halfbits) - 1;
6478 
6479     tcg_gen_dupi_vec(vece, t, max);
6480     tcg_gen_umin_vec(vece, n, n, t);
6481     tcg_gen_shli_vec(vece, n, n, halfbits);
6482     tcg_gen_bitsel_vec(vece, d, t, d, n);
6483 }
6484 
6485 static const GVecGen2 uqxtnt_ops[3] = {
6486     { .fniv = gen_uqxtnt_vec,
6487       .opt_opc = uqxtn_list,
6488       .load_dest = true,
6489       .fno = gen_helper_sve2_uqxtnt_h,
6490       .vece = MO_16 },
6491     { .fniv = gen_uqxtnt_vec,
6492       .opt_opc = uqxtn_list,
6493       .load_dest = true,
6494       .fno = gen_helper_sve2_uqxtnt_s,
6495       .vece = MO_32 },
6496     { .fniv = gen_uqxtnt_vec,
6497       .opt_opc = uqxtn_list,
6498       .load_dest = true,
6499       .fno = gen_helper_sve2_uqxtnt_d,
6500       .vece = MO_64 },
6501 };
6502 TRANS_FEAT(UQXTNT, aa64_sve2, do_narrow_extract, a, uqxtnt_ops)
6503 
6504 static const TCGOpcode sqxtun_list[] = {
6505     INDEX_op_shli_vec, INDEX_op_umin_vec, INDEX_op_smax_vec, 0
6506 };
6507 
6508 static void gen_sqxtunb_vec(unsigned vece, TCGv_vec d, TCGv_vec n)
6509 {
6510     TCGv_vec t = tcg_temp_new_vec_matching(d);
6511     int halfbits = 4 << vece;
6512     int64_t max = (1ull << halfbits) - 1;
6513 
6514     tcg_gen_dupi_vec(vece, t, 0);
6515     tcg_gen_smax_vec(vece, d, n, t);
6516     tcg_gen_dupi_vec(vece, t, max);
6517     tcg_gen_umin_vec(vece, d, d, t);
6518 }
6519 
6520 static const GVecGen2 sqxtunb_ops[3] = {
6521     { .fniv = gen_sqxtunb_vec,
6522       .opt_opc = sqxtun_list,
6523       .fno = gen_helper_sve2_sqxtunb_h,
6524       .vece = MO_16 },
6525     { .fniv = gen_sqxtunb_vec,
6526       .opt_opc = sqxtun_list,
6527       .fno = gen_helper_sve2_sqxtunb_s,
6528       .vece = MO_32 },
6529     { .fniv = gen_sqxtunb_vec,
6530       .opt_opc = sqxtun_list,
6531       .fno = gen_helper_sve2_sqxtunb_d,
6532       .vece = MO_64 },
6533 };
6534 TRANS_FEAT(SQXTUNB, aa64_sve2, do_narrow_extract, a, sqxtunb_ops)
6535 
6536 static void gen_sqxtunt_vec(unsigned vece, TCGv_vec d, TCGv_vec n)
6537 {
6538     TCGv_vec t = tcg_temp_new_vec_matching(d);
6539     int halfbits = 4 << vece;
6540     int64_t max = (1ull << halfbits) - 1;
6541 
6542     tcg_gen_dupi_vec(vece, t, 0);
6543     tcg_gen_smax_vec(vece, n, n, t);
6544     tcg_gen_dupi_vec(vece, t, max);
6545     tcg_gen_umin_vec(vece, n, n, t);
6546     tcg_gen_shli_vec(vece, n, n, halfbits);
6547     tcg_gen_bitsel_vec(vece, d, t, d, n);
6548 }
6549 
6550 static const GVecGen2 sqxtunt_ops[3] = {
6551     { .fniv = gen_sqxtunt_vec,
6552       .opt_opc = sqxtun_list,
6553       .load_dest = true,
6554       .fno = gen_helper_sve2_sqxtunt_h,
6555       .vece = MO_16 },
6556     { .fniv = gen_sqxtunt_vec,
6557       .opt_opc = sqxtun_list,
6558       .load_dest = true,
6559       .fno = gen_helper_sve2_sqxtunt_s,
6560       .vece = MO_32 },
6561     { .fniv = gen_sqxtunt_vec,
6562       .opt_opc = sqxtun_list,
6563       .load_dest = true,
6564       .fno = gen_helper_sve2_sqxtunt_d,
6565       .vece = MO_64 },
6566 };
6567 TRANS_FEAT(SQXTUNT, aa64_sve2, do_narrow_extract, a, sqxtunt_ops)
6568 
6569 static bool do_shr_narrow(DisasContext *s, arg_rri_esz *a,
6570                           const GVecGen2i ops[3])
6571 {
6572     if (a->esz < 0 || a->esz > MO_32) {
6573         return false;
6574     }
6575     assert(a->imm > 0 && a->imm <= (8 << a->esz));
6576     if (sve_access_check(s)) {
6577         unsigned vsz = vec_full_reg_size(s);
6578         tcg_gen_gvec_2i(vec_full_reg_offset(s, a->rd),
6579                         vec_full_reg_offset(s, a->rn),
6580                         vsz, vsz, a->imm, &ops[a->esz]);
6581     }
6582     return true;
6583 }
6584 
6585 static void gen_shrnb_i64(unsigned vece, TCGv_i64 d, TCGv_i64 n, int shr)
6586 {
6587     int halfbits = 4 << vece;
6588     uint64_t mask = dup_const(vece, MAKE_64BIT_MASK(0, halfbits));
6589 
6590     tcg_gen_shri_i64(d, n, shr);
6591     tcg_gen_andi_i64(d, d, mask);
6592 }
6593 
6594 static void gen_shrnb16_i64(TCGv_i64 d, TCGv_i64 n, int64_t shr)
6595 {
6596     gen_shrnb_i64(MO_16, d, n, shr);
6597 }
6598 
6599 static void gen_shrnb32_i64(TCGv_i64 d, TCGv_i64 n, int64_t shr)
6600 {
6601     gen_shrnb_i64(MO_32, d, n, shr);
6602 }
6603 
6604 static void gen_shrnb64_i64(TCGv_i64 d, TCGv_i64 n, int64_t shr)
6605 {
6606     gen_shrnb_i64(MO_64, d, n, shr);
6607 }
6608 
6609 static void gen_shrnb_vec(unsigned vece, TCGv_vec d, TCGv_vec n, int64_t shr)
6610 {
6611     TCGv_vec t = tcg_temp_new_vec_matching(d);
6612     int halfbits = 4 << vece;
6613     uint64_t mask = MAKE_64BIT_MASK(0, halfbits);
6614 
6615     tcg_gen_shri_vec(vece, n, n, shr);
6616     tcg_gen_dupi_vec(vece, t, mask);
6617     tcg_gen_and_vec(vece, d, n, t);
6618 }
6619 
6620 static const TCGOpcode shrnb_vec_list[] = { INDEX_op_shri_vec, 0 };
6621 static const GVecGen2i shrnb_ops[3] = {
6622     { .fni8 = gen_shrnb16_i64,
6623       .fniv = gen_shrnb_vec,
6624       .opt_opc = shrnb_vec_list,
6625       .fno = gen_helper_sve2_shrnb_h,
6626       .vece = MO_16 },
6627     { .fni8 = gen_shrnb32_i64,
6628       .fniv = gen_shrnb_vec,
6629       .opt_opc = shrnb_vec_list,
6630       .fno = gen_helper_sve2_shrnb_s,
6631       .vece = MO_32 },
6632     { .fni8 = gen_shrnb64_i64,
6633       .fniv = gen_shrnb_vec,
6634       .opt_opc = shrnb_vec_list,
6635       .fno = gen_helper_sve2_shrnb_d,
6636       .vece = MO_64 },
6637 };
6638 TRANS_FEAT(SHRNB, aa64_sve2, do_shr_narrow, a, shrnb_ops)
6639 
6640 static void gen_shrnt_i64(unsigned vece, TCGv_i64 d, TCGv_i64 n, int shr)
6641 {
6642     int halfbits = 4 << vece;
6643     uint64_t mask = dup_const(vece, MAKE_64BIT_MASK(0, halfbits));
6644 
6645     tcg_gen_shli_i64(n, n, halfbits - shr);
6646     tcg_gen_andi_i64(n, n, ~mask);
6647     tcg_gen_andi_i64(d, d, mask);
6648     tcg_gen_or_i64(d, d, n);
6649 }
6650 
6651 static void gen_shrnt16_i64(TCGv_i64 d, TCGv_i64 n, int64_t shr)
6652 {
6653     gen_shrnt_i64(MO_16, d, n, shr);
6654 }
6655 
6656 static void gen_shrnt32_i64(TCGv_i64 d, TCGv_i64 n, int64_t shr)
6657 {
6658     gen_shrnt_i64(MO_32, d, n, shr);
6659 }
6660 
6661 static void gen_shrnt64_i64(TCGv_i64 d, TCGv_i64 n, int64_t shr)
6662 {
6663     tcg_gen_shri_i64(n, n, shr);
6664     tcg_gen_deposit_i64(d, d, n, 32, 32);
6665 }
6666 
6667 static void gen_shrnt_vec(unsigned vece, TCGv_vec d, TCGv_vec n, int64_t shr)
6668 {
6669     TCGv_vec t = tcg_temp_new_vec_matching(d);
6670     int halfbits = 4 << vece;
6671     uint64_t mask = MAKE_64BIT_MASK(0, halfbits);
6672 
6673     tcg_gen_shli_vec(vece, n, n, halfbits - shr);
6674     tcg_gen_dupi_vec(vece, t, mask);
6675     tcg_gen_bitsel_vec(vece, d, t, d, n);
6676 }
6677 
6678 static const TCGOpcode shrnt_vec_list[] = { INDEX_op_shli_vec, 0 };
6679 static const GVecGen2i shrnt_ops[3] = {
6680     { .fni8 = gen_shrnt16_i64,
6681       .fniv = gen_shrnt_vec,
6682       .opt_opc = shrnt_vec_list,
6683       .load_dest = true,
6684       .fno = gen_helper_sve2_shrnt_h,
6685       .vece = MO_16 },
6686     { .fni8 = gen_shrnt32_i64,
6687       .fniv = gen_shrnt_vec,
6688       .opt_opc = shrnt_vec_list,
6689       .load_dest = true,
6690       .fno = gen_helper_sve2_shrnt_s,
6691       .vece = MO_32 },
6692     { .fni8 = gen_shrnt64_i64,
6693       .fniv = gen_shrnt_vec,
6694       .opt_opc = shrnt_vec_list,
6695       .load_dest = true,
6696       .fno = gen_helper_sve2_shrnt_d,
6697       .vece = MO_64 },
6698 };
6699 TRANS_FEAT(SHRNT, aa64_sve2, do_shr_narrow, a, shrnt_ops)
6700 
6701 static const GVecGen2i rshrnb_ops[3] = {
6702     { .fno = gen_helper_sve2_rshrnb_h },
6703     { .fno = gen_helper_sve2_rshrnb_s },
6704     { .fno = gen_helper_sve2_rshrnb_d },
6705 };
6706 TRANS_FEAT(RSHRNB, aa64_sve2, do_shr_narrow, a, rshrnb_ops)
6707 
6708 static const GVecGen2i rshrnt_ops[3] = {
6709     { .fno = gen_helper_sve2_rshrnt_h },
6710     { .fno = gen_helper_sve2_rshrnt_s },
6711     { .fno = gen_helper_sve2_rshrnt_d },
6712 };
6713 TRANS_FEAT(RSHRNT, aa64_sve2, do_shr_narrow, a, rshrnt_ops)
6714 
6715 static void gen_sqshrunb_vec(unsigned vece, TCGv_vec d,
6716                              TCGv_vec n, int64_t shr)
6717 {
6718     TCGv_vec t = tcg_temp_new_vec_matching(d);
6719     int halfbits = 4 << vece;
6720 
6721     tcg_gen_sari_vec(vece, n, n, shr);
6722     tcg_gen_dupi_vec(vece, t, 0);
6723     tcg_gen_smax_vec(vece, n, n, t);
6724     tcg_gen_dupi_vec(vece, t, MAKE_64BIT_MASK(0, halfbits));
6725     tcg_gen_umin_vec(vece, d, n, t);
6726 }
6727 
6728 static const TCGOpcode sqshrunb_vec_list[] = {
6729     INDEX_op_sari_vec, INDEX_op_smax_vec, INDEX_op_umin_vec, 0
6730 };
6731 static const GVecGen2i sqshrunb_ops[3] = {
6732     { .fniv = gen_sqshrunb_vec,
6733       .opt_opc = sqshrunb_vec_list,
6734       .fno = gen_helper_sve2_sqshrunb_h,
6735       .vece = MO_16 },
6736     { .fniv = gen_sqshrunb_vec,
6737       .opt_opc = sqshrunb_vec_list,
6738       .fno = gen_helper_sve2_sqshrunb_s,
6739       .vece = MO_32 },
6740     { .fniv = gen_sqshrunb_vec,
6741       .opt_opc = sqshrunb_vec_list,
6742       .fno = gen_helper_sve2_sqshrunb_d,
6743       .vece = MO_64 },
6744 };
6745 TRANS_FEAT(SQSHRUNB, aa64_sve2, do_shr_narrow, a, sqshrunb_ops)
6746 
6747 static void gen_sqshrunt_vec(unsigned vece, TCGv_vec d,
6748                              TCGv_vec n, int64_t shr)
6749 {
6750     TCGv_vec t = tcg_temp_new_vec_matching(d);
6751     int halfbits = 4 << vece;
6752 
6753     tcg_gen_sari_vec(vece, n, n, shr);
6754     tcg_gen_dupi_vec(vece, t, 0);
6755     tcg_gen_smax_vec(vece, n, n, t);
6756     tcg_gen_dupi_vec(vece, t, MAKE_64BIT_MASK(0, halfbits));
6757     tcg_gen_umin_vec(vece, n, n, t);
6758     tcg_gen_shli_vec(vece, n, n, halfbits);
6759     tcg_gen_bitsel_vec(vece, d, t, d, n);
6760 }
6761 
6762 static const TCGOpcode sqshrunt_vec_list[] = {
6763     INDEX_op_shli_vec, INDEX_op_sari_vec,
6764     INDEX_op_smax_vec, INDEX_op_umin_vec, 0
6765 };
6766 static const GVecGen2i sqshrunt_ops[3] = {
6767     { .fniv = gen_sqshrunt_vec,
6768       .opt_opc = sqshrunt_vec_list,
6769       .load_dest = true,
6770       .fno = gen_helper_sve2_sqshrunt_h,
6771       .vece = MO_16 },
6772     { .fniv = gen_sqshrunt_vec,
6773       .opt_opc = sqshrunt_vec_list,
6774       .load_dest = true,
6775       .fno = gen_helper_sve2_sqshrunt_s,
6776       .vece = MO_32 },
6777     { .fniv = gen_sqshrunt_vec,
6778       .opt_opc = sqshrunt_vec_list,
6779       .load_dest = true,
6780       .fno = gen_helper_sve2_sqshrunt_d,
6781       .vece = MO_64 },
6782 };
6783 TRANS_FEAT(SQSHRUNT, aa64_sve2, do_shr_narrow, a, sqshrunt_ops)
6784 
6785 static const GVecGen2i sqrshrunb_ops[3] = {
6786     { .fno = gen_helper_sve2_sqrshrunb_h },
6787     { .fno = gen_helper_sve2_sqrshrunb_s },
6788     { .fno = gen_helper_sve2_sqrshrunb_d },
6789 };
6790 TRANS_FEAT(SQRSHRUNB, aa64_sve2, do_shr_narrow, a, sqrshrunb_ops)
6791 
6792 static const GVecGen2i sqrshrunt_ops[3] = {
6793     { .fno = gen_helper_sve2_sqrshrunt_h },
6794     { .fno = gen_helper_sve2_sqrshrunt_s },
6795     { .fno = gen_helper_sve2_sqrshrunt_d },
6796 };
6797 TRANS_FEAT(SQRSHRUNT, aa64_sve2, do_shr_narrow, a, sqrshrunt_ops)
6798 
6799 static void gen_sqshrnb_vec(unsigned vece, TCGv_vec d,
6800                             TCGv_vec n, int64_t shr)
6801 {
6802     TCGv_vec t = tcg_temp_new_vec_matching(d);
6803     int halfbits = 4 << vece;
6804     int64_t max = MAKE_64BIT_MASK(0, halfbits - 1);
6805     int64_t min = -max - 1;
6806 
6807     tcg_gen_sari_vec(vece, n, n, shr);
6808     tcg_gen_dupi_vec(vece, t, min);
6809     tcg_gen_smax_vec(vece, n, n, t);
6810     tcg_gen_dupi_vec(vece, t, max);
6811     tcg_gen_smin_vec(vece, n, n, t);
6812     tcg_gen_dupi_vec(vece, t, MAKE_64BIT_MASK(0, halfbits));
6813     tcg_gen_and_vec(vece, d, n, t);
6814 }
6815 
6816 static const TCGOpcode sqshrnb_vec_list[] = {
6817     INDEX_op_sari_vec, INDEX_op_smax_vec, INDEX_op_smin_vec, 0
6818 };
6819 static const GVecGen2i sqshrnb_ops[3] = {
6820     { .fniv = gen_sqshrnb_vec,
6821       .opt_opc = sqshrnb_vec_list,
6822       .fno = gen_helper_sve2_sqshrnb_h,
6823       .vece = MO_16 },
6824     { .fniv = gen_sqshrnb_vec,
6825       .opt_opc = sqshrnb_vec_list,
6826       .fno = gen_helper_sve2_sqshrnb_s,
6827       .vece = MO_32 },
6828     { .fniv = gen_sqshrnb_vec,
6829       .opt_opc = sqshrnb_vec_list,
6830       .fno = gen_helper_sve2_sqshrnb_d,
6831       .vece = MO_64 },
6832 };
6833 TRANS_FEAT(SQSHRNB, aa64_sve2, do_shr_narrow, a, sqshrnb_ops)
6834 
6835 static void gen_sqshrnt_vec(unsigned vece, TCGv_vec d,
6836                              TCGv_vec n, int64_t shr)
6837 {
6838     TCGv_vec t = tcg_temp_new_vec_matching(d);
6839     int halfbits = 4 << vece;
6840     int64_t max = MAKE_64BIT_MASK(0, halfbits - 1);
6841     int64_t min = -max - 1;
6842 
6843     tcg_gen_sari_vec(vece, n, n, shr);
6844     tcg_gen_dupi_vec(vece, t, min);
6845     tcg_gen_smax_vec(vece, n, n, t);
6846     tcg_gen_dupi_vec(vece, t, max);
6847     tcg_gen_smin_vec(vece, n, n, t);
6848     tcg_gen_shli_vec(vece, n, n, halfbits);
6849     tcg_gen_dupi_vec(vece, t, MAKE_64BIT_MASK(0, halfbits));
6850     tcg_gen_bitsel_vec(vece, d, t, d, n);
6851 }
6852 
6853 static const TCGOpcode sqshrnt_vec_list[] = {
6854     INDEX_op_shli_vec, INDEX_op_sari_vec,
6855     INDEX_op_smax_vec, INDEX_op_smin_vec, 0
6856 };
6857 static const GVecGen2i sqshrnt_ops[3] = {
6858     { .fniv = gen_sqshrnt_vec,
6859       .opt_opc = sqshrnt_vec_list,
6860       .load_dest = true,
6861       .fno = gen_helper_sve2_sqshrnt_h,
6862       .vece = MO_16 },
6863     { .fniv = gen_sqshrnt_vec,
6864       .opt_opc = sqshrnt_vec_list,
6865       .load_dest = true,
6866       .fno = gen_helper_sve2_sqshrnt_s,
6867       .vece = MO_32 },
6868     { .fniv = gen_sqshrnt_vec,
6869       .opt_opc = sqshrnt_vec_list,
6870       .load_dest = true,
6871       .fno = gen_helper_sve2_sqshrnt_d,
6872       .vece = MO_64 },
6873 };
6874 TRANS_FEAT(SQSHRNT, aa64_sve2, do_shr_narrow, a, sqshrnt_ops)
6875 
6876 static const GVecGen2i sqrshrnb_ops[3] = {
6877     { .fno = gen_helper_sve2_sqrshrnb_h },
6878     { .fno = gen_helper_sve2_sqrshrnb_s },
6879     { .fno = gen_helper_sve2_sqrshrnb_d },
6880 };
6881 TRANS_FEAT(SQRSHRNB, aa64_sve2, do_shr_narrow, a, sqrshrnb_ops)
6882 
6883 static const GVecGen2i sqrshrnt_ops[3] = {
6884     { .fno = gen_helper_sve2_sqrshrnt_h },
6885     { .fno = gen_helper_sve2_sqrshrnt_s },
6886     { .fno = gen_helper_sve2_sqrshrnt_d },
6887 };
6888 TRANS_FEAT(SQRSHRNT, aa64_sve2, do_shr_narrow, a, sqrshrnt_ops)
6889 
6890 static void gen_uqshrnb_vec(unsigned vece, TCGv_vec d,
6891                             TCGv_vec n, int64_t shr)
6892 {
6893     TCGv_vec t = tcg_temp_new_vec_matching(d);
6894     int halfbits = 4 << vece;
6895 
6896     tcg_gen_shri_vec(vece, n, n, shr);
6897     tcg_gen_dupi_vec(vece, t, MAKE_64BIT_MASK(0, halfbits));
6898     tcg_gen_umin_vec(vece, d, n, t);
6899 }
6900 
6901 static const TCGOpcode uqshrnb_vec_list[] = {
6902     INDEX_op_shri_vec, INDEX_op_umin_vec, 0
6903 };
6904 static const GVecGen2i uqshrnb_ops[3] = {
6905     { .fniv = gen_uqshrnb_vec,
6906       .opt_opc = uqshrnb_vec_list,
6907       .fno = gen_helper_sve2_uqshrnb_h,
6908       .vece = MO_16 },
6909     { .fniv = gen_uqshrnb_vec,
6910       .opt_opc = uqshrnb_vec_list,
6911       .fno = gen_helper_sve2_uqshrnb_s,
6912       .vece = MO_32 },
6913     { .fniv = gen_uqshrnb_vec,
6914       .opt_opc = uqshrnb_vec_list,
6915       .fno = gen_helper_sve2_uqshrnb_d,
6916       .vece = MO_64 },
6917 };
6918 TRANS_FEAT(UQSHRNB, aa64_sve2, do_shr_narrow, a, uqshrnb_ops)
6919 
6920 static void gen_uqshrnt_vec(unsigned vece, TCGv_vec d,
6921                             TCGv_vec n, int64_t shr)
6922 {
6923     TCGv_vec t = tcg_temp_new_vec_matching(d);
6924     int halfbits = 4 << vece;
6925 
6926     tcg_gen_shri_vec(vece, n, n, shr);
6927     tcg_gen_dupi_vec(vece, t, MAKE_64BIT_MASK(0, halfbits));
6928     tcg_gen_umin_vec(vece, n, n, t);
6929     tcg_gen_shli_vec(vece, n, n, halfbits);
6930     tcg_gen_bitsel_vec(vece, d, t, d, n);
6931 }
6932 
6933 static const TCGOpcode uqshrnt_vec_list[] = {
6934     INDEX_op_shli_vec, INDEX_op_shri_vec, INDEX_op_umin_vec, 0
6935 };
6936 static const GVecGen2i uqshrnt_ops[3] = {
6937     { .fniv = gen_uqshrnt_vec,
6938       .opt_opc = uqshrnt_vec_list,
6939       .load_dest = true,
6940       .fno = gen_helper_sve2_uqshrnt_h,
6941       .vece = MO_16 },
6942     { .fniv = gen_uqshrnt_vec,
6943       .opt_opc = uqshrnt_vec_list,
6944       .load_dest = true,
6945       .fno = gen_helper_sve2_uqshrnt_s,
6946       .vece = MO_32 },
6947     { .fniv = gen_uqshrnt_vec,
6948       .opt_opc = uqshrnt_vec_list,
6949       .load_dest = true,
6950       .fno = gen_helper_sve2_uqshrnt_d,
6951       .vece = MO_64 },
6952 };
6953 TRANS_FEAT(UQSHRNT, aa64_sve2, do_shr_narrow, a, uqshrnt_ops)
6954 
6955 static const GVecGen2i uqrshrnb_ops[3] = {
6956     { .fno = gen_helper_sve2_uqrshrnb_h },
6957     { .fno = gen_helper_sve2_uqrshrnb_s },
6958     { .fno = gen_helper_sve2_uqrshrnb_d },
6959 };
6960 TRANS_FEAT(UQRSHRNB, aa64_sve2, do_shr_narrow, a, uqrshrnb_ops)
6961 
6962 static const GVecGen2i uqrshrnt_ops[3] = {
6963     { .fno = gen_helper_sve2_uqrshrnt_h },
6964     { .fno = gen_helper_sve2_uqrshrnt_s },
6965     { .fno = gen_helper_sve2_uqrshrnt_d },
6966 };
6967 TRANS_FEAT(UQRSHRNT, aa64_sve2, do_shr_narrow, a, uqrshrnt_ops)
6968 
6969 #define DO_SVE2_ZZZ_NARROW(NAME, name)                                    \
6970     static gen_helper_gvec_3 * const name##_fns[4] = {                    \
6971         NULL,                       gen_helper_sve2_##name##_h,           \
6972         gen_helper_sve2_##name##_s, gen_helper_sve2_##name##_d,           \
6973     };                                                                    \
6974     TRANS_FEAT(NAME, aa64_sve2, gen_gvec_ool_arg_zzz,                     \
6975                name##_fns[a->esz], a, 0)
6976 
6977 DO_SVE2_ZZZ_NARROW(ADDHNB, addhnb)
6978 DO_SVE2_ZZZ_NARROW(ADDHNT, addhnt)
6979 DO_SVE2_ZZZ_NARROW(RADDHNB, raddhnb)
6980 DO_SVE2_ZZZ_NARROW(RADDHNT, raddhnt)
6981 
6982 DO_SVE2_ZZZ_NARROW(SUBHNB, subhnb)
6983 DO_SVE2_ZZZ_NARROW(SUBHNT, subhnt)
6984 DO_SVE2_ZZZ_NARROW(RSUBHNB, rsubhnb)
6985 DO_SVE2_ZZZ_NARROW(RSUBHNT, rsubhnt)
6986 
6987 static gen_helper_gvec_flags_4 * const match_fns[4] = {
6988     gen_helper_sve2_match_ppzz_b, gen_helper_sve2_match_ppzz_h, NULL, NULL
6989 };
6990 TRANS_FEAT_NONSTREAMING(MATCH, aa64_sve2, do_ppzz_flags, a, match_fns[a->esz])
6991 
6992 static gen_helper_gvec_flags_4 * const nmatch_fns[4] = {
6993     gen_helper_sve2_nmatch_ppzz_b, gen_helper_sve2_nmatch_ppzz_h, NULL, NULL
6994 };
6995 TRANS_FEAT_NONSTREAMING(NMATCH, aa64_sve2, do_ppzz_flags, a, nmatch_fns[a->esz])
6996 
6997 static gen_helper_gvec_4 * const histcnt_fns[4] = {
6998     NULL, NULL, gen_helper_sve2_histcnt_s, gen_helper_sve2_histcnt_d
6999 };
7000 TRANS_FEAT_NONSTREAMING(HISTCNT, aa64_sve2, gen_gvec_ool_arg_zpzz,
7001                         histcnt_fns[a->esz], a, 0)
7002 
7003 TRANS_FEAT_NONSTREAMING(HISTSEG, aa64_sve2, gen_gvec_ool_arg_zzz,
7004                         a->esz == 0 ? gen_helper_sve2_histseg : NULL, a, 0)
7005 
7006 DO_ZPZZ_FP(FADDP, aa64_sve2, sve2_faddp_zpzz)
7007 DO_ZPZZ_FP(FMAXNMP, aa64_sve2, sve2_fmaxnmp_zpzz)
7008 DO_ZPZZ_FP(FMINNMP, aa64_sve2, sve2_fminnmp_zpzz)
7009 DO_ZPZZ_FP(FMAXP, aa64_sve2, sve2_fmaxp_zpzz)
7010 DO_ZPZZ_FP(FMINP, aa64_sve2, sve2_fminp_zpzz)
7011 
7012 /*
7013  * SVE Integer Multiply-Add (unpredicated)
7014  */
7015 
7016 TRANS_FEAT_NONSTREAMING(FMMLA_s, aa64_sve_f32mm, gen_gvec_fpst_zzzz,
7017                         gen_helper_fmmla_s, a->rd, a->rn, a->rm, a->ra,
7018                         0, FPST_FPCR)
7019 TRANS_FEAT_NONSTREAMING(FMMLA_d, aa64_sve_f64mm, gen_gvec_fpst_zzzz,
7020                         gen_helper_fmmla_d, a->rd, a->rn, a->rm, a->ra,
7021                         0, FPST_FPCR)
7022 
7023 static gen_helper_gvec_4 * const sqdmlal_zzzw_fns[] = {
7024     NULL,                           gen_helper_sve2_sqdmlal_zzzw_h,
7025     gen_helper_sve2_sqdmlal_zzzw_s, gen_helper_sve2_sqdmlal_zzzw_d,
7026 };
7027 TRANS_FEAT(SQDMLALB_zzzw, aa64_sve2, gen_gvec_ool_arg_zzzz,
7028            sqdmlal_zzzw_fns[a->esz], a, 0)
7029 TRANS_FEAT(SQDMLALT_zzzw, aa64_sve2, gen_gvec_ool_arg_zzzz,
7030            sqdmlal_zzzw_fns[a->esz], a, 3)
7031 TRANS_FEAT(SQDMLALBT, aa64_sve2, gen_gvec_ool_arg_zzzz,
7032            sqdmlal_zzzw_fns[a->esz], a, 2)
7033 
7034 static gen_helper_gvec_4 * const sqdmlsl_zzzw_fns[] = {
7035     NULL,                           gen_helper_sve2_sqdmlsl_zzzw_h,
7036     gen_helper_sve2_sqdmlsl_zzzw_s, gen_helper_sve2_sqdmlsl_zzzw_d,
7037 };
7038 TRANS_FEAT(SQDMLSLB_zzzw, aa64_sve2, gen_gvec_ool_arg_zzzz,
7039            sqdmlsl_zzzw_fns[a->esz], a, 0)
7040 TRANS_FEAT(SQDMLSLT_zzzw, aa64_sve2, gen_gvec_ool_arg_zzzz,
7041            sqdmlsl_zzzw_fns[a->esz], a, 3)
7042 TRANS_FEAT(SQDMLSLBT, aa64_sve2, gen_gvec_ool_arg_zzzz,
7043            sqdmlsl_zzzw_fns[a->esz], a, 2)
7044 
7045 static gen_helper_gvec_4 * const sqrdmlah_fns[] = {
7046     gen_helper_sve2_sqrdmlah_b, gen_helper_sve2_sqrdmlah_h,
7047     gen_helper_sve2_sqrdmlah_s, gen_helper_sve2_sqrdmlah_d,
7048 };
7049 TRANS_FEAT(SQRDMLAH_zzzz, aa64_sve2, gen_gvec_ool_arg_zzzz,
7050            sqrdmlah_fns[a->esz], a, 0)
7051 
7052 static gen_helper_gvec_4 * const sqrdmlsh_fns[] = {
7053     gen_helper_sve2_sqrdmlsh_b, gen_helper_sve2_sqrdmlsh_h,
7054     gen_helper_sve2_sqrdmlsh_s, gen_helper_sve2_sqrdmlsh_d,
7055 };
7056 TRANS_FEAT(SQRDMLSH_zzzz, aa64_sve2, gen_gvec_ool_arg_zzzz,
7057            sqrdmlsh_fns[a->esz], a, 0)
7058 
7059 static gen_helper_gvec_4 * const smlal_zzzw_fns[] = {
7060     NULL,                         gen_helper_sve2_smlal_zzzw_h,
7061     gen_helper_sve2_smlal_zzzw_s, gen_helper_sve2_smlal_zzzw_d,
7062 };
7063 TRANS_FEAT(SMLALB_zzzw, aa64_sve2, gen_gvec_ool_arg_zzzz,
7064            smlal_zzzw_fns[a->esz], a, 0)
7065 TRANS_FEAT(SMLALT_zzzw, aa64_sve2, gen_gvec_ool_arg_zzzz,
7066            smlal_zzzw_fns[a->esz], a, 1)
7067 
7068 static gen_helper_gvec_4 * const umlal_zzzw_fns[] = {
7069     NULL,                         gen_helper_sve2_umlal_zzzw_h,
7070     gen_helper_sve2_umlal_zzzw_s, gen_helper_sve2_umlal_zzzw_d,
7071 };
7072 TRANS_FEAT(UMLALB_zzzw, aa64_sve2, gen_gvec_ool_arg_zzzz,
7073            umlal_zzzw_fns[a->esz], a, 0)
7074 TRANS_FEAT(UMLALT_zzzw, aa64_sve2, gen_gvec_ool_arg_zzzz,
7075            umlal_zzzw_fns[a->esz], a, 1)
7076 
7077 static gen_helper_gvec_4 * const smlsl_zzzw_fns[] = {
7078     NULL,                         gen_helper_sve2_smlsl_zzzw_h,
7079     gen_helper_sve2_smlsl_zzzw_s, gen_helper_sve2_smlsl_zzzw_d,
7080 };
7081 TRANS_FEAT(SMLSLB_zzzw, aa64_sve2, gen_gvec_ool_arg_zzzz,
7082            smlsl_zzzw_fns[a->esz], a, 0)
7083 TRANS_FEAT(SMLSLT_zzzw, aa64_sve2, gen_gvec_ool_arg_zzzz,
7084            smlsl_zzzw_fns[a->esz], a, 1)
7085 
7086 static gen_helper_gvec_4 * const umlsl_zzzw_fns[] = {
7087     NULL,                         gen_helper_sve2_umlsl_zzzw_h,
7088     gen_helper_sve2_umlsl_zzzw_s, gen_helper_sve2_umlsl_zzzw_d,
7089 };
7090 TRANS_FEAT(UMLSLB_zzzw, aa64_sve2, gen_gvec_ool_arg_zzzz,
7091            umlsl_zzzw_fns[a->esz], a, 0)
7092 TRANS_FEAT(UMLSLT_zzzw, aa64_sve2, gen_gvec_ool_arg_zzzz,
7093            umlsl_zzzw_fns[a->esz], a, 1)
7094 
7095 static gen_helper_gvec_4 * const cmla_fns[] = {
7096     gen_helper_sve2_cmla_zzzz_b, gen_helper_sve2_cmla_zzzz_h,
7097     gen_helper_sve2_cmla_zzzz_s, gen_helper_sve2_cmla_zzzz_d,
7098 };
7099 TRANS_FEAT(CMLA_zzzz, aa64_sve2, gen_gvec_ool_zzzz,
7100            cmla_fns[a->esz], a->rd, a->rn, a->rm, a->ra, a->rot)
7101 
7102 static gen_helper_gvec_4 * const cdot_fns[] = {
7103     NULL, NULL, gen_helper_sve2_cdot_zzzz_s, gen_helper_sve2_cdot_zzzz_d
7104 };
7105 TRANS_FEAT(CDOT_zzzz, aa64_sve2, gen_gvec_ool_zzzz,
7106            cdot_fns[a->esz], a->rd, a->rn, a->rm, a->ra, a->rot)
7107 
7108 static gen_helper_gvec_4 * const sqrdcmlah_fns[] = {
7109     gen_helper_sve2_sqrdcmlah_zzzz_b, gen_helper_sve2_sqrdcmlah_zzzz_h,
7110     gen_helper_sve2_sqrdcmlah_zzzz_s, gen_helper_sve2_sqrdcmlah_zzzz_d,
7111 };
7112 TRANS_FEAT(SQRDCMLAH_zzzz, aa64_sve2, gen_gvec_ool_zzzz,
7113            sqrdcmlah_fns[a->esz], a->rd, a->rn, a->rm, a->ra, a->rot)
7114 
7115 TRANS_FEAT(USDOT_zzzz, aa64_sve_i8mm, gen_gvec_ool_arg_zzzz,
7116            a->esz == 2 ? gen_helper_gvec_usdot_b : NULL, a, 0)
7117 
7118 TRANS_FEAT_NONSTREAMING(AESMC, aa64_sve2_aes, gen_gvec_ool_zz,
7119                         gen_helper_crypto_aesmc, a->rd, a->rd, a->decrypt)
7120 
7121 TRANS_FEAT_NONSTREAMING(AESE, aa64_sve2_aes, gen_gvec_ool_arg_zzz,
7122                         gen_helper_crypto_aese, a, false)
7123 TRANS_FEAT_NONSTREAMING(AESD, aa64_sve2_aes, gen_gvec_ool_arg_zzz,
7124                         gen_helper_crypto_aese, a, true)
7125 
7126 TRANS_FEAT_NONSTREAMING(SM4E, aa64_sve2_sm4, gen_gvec_ool_arg_zzz,
7127                         gen_helper_crypto_sm4e, a, 0)
7128 TRANS_FEAT_NONSTREAMING(SM4EKEY, aa64_sve2_sm4, gen_gvec_ool_arg_zzz,
7129                         gen_helper_crypto_sm4ekey, a, 0)
7130 
7131 TRANS_FEAT_NONSTREAMING(RAX1, aa64_sve2_sha3, gen_gvec_fn_arg_zzz,
7132                         gen_gvec_rax1, a)
7133 
7134 TRANS_FEAT(FCVTNT_sh, aa64_sve2, gen_gvec_fpst_arg_zpz,
7135            gen_helper_sve2_fcvtnt_sh, a, 0, FPST_FPCR)
7136 TRANS_FEAT(FCVTNT_ds, aa64_sve2, gen_gvec_fpst_arg_zpz,
7137            gen_helper_sve2_fcvtnt_ds, a, 0, FPST_FPCR)
7138 
7139 TRANS_FEAT(BFCVTNT, aa64_sve_bf16, gen_gvec_fpst_arg_zpz,
7140            gen_helper_sve_bfcvtnt, a, 0, FPST_FPCR)
7141 
7142 TRANS_FEAT(FCVTLT_hs, aa64_sve2, gen_gvec_fpst_arg_zpz,
7143            gen_helper_sve2_fcvtlt_hs, a, 0, FPST_FPCR)
7144 TRANS_FEAT(FCVTLT_sd, aa64_sve2, gen_gvec_fpst_arg_zpz,
7145            gen_helper_sve2_fcvtlt_sd, a, 0, FPST_FPCR)
7146 
7147 TRANS_FEAT(FCVTX_ds, aa64_sve2, do_frint_mode, a,
7148            float_round_to_odd, gen_helper_sve_fcvt_ds)
7149 TRANS_FEAT(FCVTXNT_ds, aa64_sve2, do_frint_mode, a,
7150            float_round_to_odd, gen_helper_sve2_fcvtnt_ds)
7151 
7152 static gen_helper_gvec_3_ptr * const flogb_fns[] = {
7153     NULL,               gen_helper_flogb_h,
7154     gen_helper_flogb_s, gen_helper_flogb_d
7155 };
7156 TRANS_FEAT(FLOGB, aa64_sve2, gen_gvec_fpst_arg_zpz, flogb_fns[a->esz],
7157            a, 0, a->esz == MO_16 ? FPST_FPCR_F16 : FPST_FPCR)
7158 
7159 static bool do_FMLAL_zzzw(DisasContext *s, arg_rrrr_esz *a, bool sub, bool sel)
7160 {
7161     return gen_gvec_ptr_zzzz(s, gen_helper_sve2_fmlal_zzzw_s,
7162                              a->rd, a->rn, a->rm, a->ra,
7163                              (sel << 1) | sub, cpu_env);
7164 }
7165 
7166 TRANS_FEAT(FMLALB_zzzw, aa64_sve2, do_FMLAL_zzzw, a, false, false)
7167 TRANS_FEAT(FMLALT_zzzw, aa64_sve2, do_FMLAL_zzzw, a, false, true)
7168 TRANS_FEAT(FMLSLB_zzzw, aa64_sve2, do_FMLAL_zzzw, a, true, false)
7169 TRANS_FEAT(FMLSLT_zzzw, aa64_sve2, do_FMLAL_zzzw, a, true, true)
7170 
7171 static bool do_FMLAL_zzxw(DisasContext *s, arg_rrxr_esz *a, bool sub, bool sel)
7172 {
7173     return gen_gvec_ptr_zzzz(s, gen_helper_sve2_fmlal_zzxw_s,
7174                              a->rd, a->rn, a->rm, a->ra,
7175                              (a->index << 2) | (sel << 1) | sub, cpu_env);
7176 }
7177 
7178 TRANS_FEAT(FMLALB_zzxw, aa64_sve2, do_FMLAL_zzxw, a, false, false)
7179 TRANS_FEAT(FMLALT_zzxw, aa64_sve2, do_FMLAL_zzxw, a, false, true)
7180 TRANS_FEAT(FMLSLB_zzxw, aa64_sve2, do_FMLAL_zzxw, a, true, false)
7181 TRANS_FEAT(FMLSLT_zzxw, aa64_sve2, do_FMLAL_zzxw, a, true, true)
7182 
7183 TRANS_FEAT_NONSTREAMING(SMMLA, aa64_sve_i8mm, gen_gvec_ool_arg_zzzz,
7184                         gen_helper_gvec_smmla_b, a, 0)
7185 TRANS_FEAT_NONSTREAMING(USMMLA, aa64_sve_i8mm, gen_gvec_ool_arg_zzzz,
7186                         gen_helper_gvec_usmmla_b, a, 0)
7187 TRANS_FEAT_NONSTREAMING(UMMLA, aa64_sve_i8mm, gen_gvec_ool_arg_zzzz,
7188                         gen_helper_gvec_ummla_b, a, 0)
7189 
7190 TRANS_FEAT(BFDOT_zzzz, aa64_sve_bf16, gen_gvec_ool_arg_zzzz,
7191            gen_helper_gvec_bfdot, a, 0)
7192 TRANS_FEAT(BFDOT_zzxz, aa64_sve_bf16, gen_gvec_ool_arg_zzxz,
7193            gen_helper_gvec_bfdot_idx, a)
7194 
7195 TRANS_FEAT_NONSTREAMING(BFMMLA, aa64_sve_bf16, gen_gvec_ool_arg_zzzz,
7196                         gen_helper_gvec_bfmmla, a, 0)
7197 
7198 static bool do_BFMLAL_zzzw(DisasContext *s, arg_rrrr_esz *a, bool sel)
7199 {
7200     return gen_gvec_fpst_zzzz(s, gen_helper_gvec_bfmlal,
7201                               a->rd, a->rn, a->rm, a->ra, sel, FPST_FPCR);
7202 }
7203 
7204 TRANS_FEAT(BFMLALB_zzzw, aa64_sve_bf16, do_BFMLAL_zzzw, a, false)
7205 TRANS_FEAT(BFMLALT_zzzw, aa64_sve_bf16, do_BFMLAL_zzzw, a, true)
7206 
7207 static bool do_BFMLAL_zzxw(DisasContext *s, arg_rrxr_esz *a, bool sel)
7208 {
7209     return gen_gvec_fpst_zzzz(s, gen_helper_gvec_bfmlal_idx,
7210                               a->rd, a->rn, a->rm, a->ra,
7211                               (a->index << 1) | sel, FPST_FPCR);
7212 }
7213 
7214 TRANS_FEAT(BFMLALB_zzxw, aa64_sve_bf16, do_BFMLAL_zzxw, a, false)
7215 TRANS_FEAT(BFMLALT_zzxw, aa64_sve_bf16, do_BFMLAL_zzxw, a, true)
7216 
7217 static bool trans_PSEL(DisasContext *s, arg_psel *a)
7218 {
7219     int vl = vec_full_reg_size(s);
7220     int pl = pred_gvec_reg_size(s);
7221     int elements = vl >> a->esz;
7222     TCGv_i64 tmp, didx, dbit;
7223     TCGv_ptr ptr;
7224 
7225     if (!dc_isar_feature(aa64_sme, s)) {
7226         return false;
7227     }
7228     if (!sve_access_check(s)) {
7229         return true;
7230     }
7231 
7232     tmp = tcg_temp_new_i64();
7233     dbit = tcg_temp_new_i64();
7234     didx = tcg_temp_new_i64();
7235     ptr = tcg_temp_new_ptr();
7236 
7237     /* Compute the predicate element. */
7238     tcg_gen_addi_i64(tmp, cpu_reg(s, a->rv), a->imm);
7239     if (is_power_of_2(elements)) {
7240         tcg_gen_andi_i64(tmp, tmp, elements - 1);
7241     } else {
7242         tcg_gen_remu_i64(tmp, tmp, tcg_constant_i64(elements));
7243     }
7244 
7245     /* Extract the predicate byte and bit indices. */
7246     tcg_gen_shli_i64(tmp, tmp, a->esz);
7247     tcg_gen_andi_i64(dbit, tmp, 7);
7248     tcg_gen_shri_i64(didx, tmp, 3);
7249     if (HOST_BIG_ENDIAN) {
7250         tcg_gen_xori_i64(didx, didx, 7);
7251     }
7252 
7253     /* Load the predicate word. */
7254     tcg_gen_trunc_i64_ptr(ptr, didx);
7255     tcg_gen_add_ptr(ptr, ptr, cpu_env);
7256     tcg_gen_ld8u_i64(tmp, ptr, pred_full_reg_offset(s, a->pm));
7257 
7258     /* Extract the predicate bit and replicate to MO_64. */
7259     tcg_gen_shr_i64(tmp, tmp, dbit);
7260     tcg_gen_andi_i64(tmp, tmp, 1);
7261     tcg_gen_neg_i64(tmp, tmp);
7262 
7263     /* Apply to either copy the source, or write zeros. */
7264     tcg_gen_gvec_ands(MO_64, pred_full_reg_offset(s, a->pd),
7265                       pred_full_reg_offset(s, a->pn), tmp, pl, pl);
7266     return true;
7267 }
7268 
7269 static void gen_sclamp_i32(TCGv_i32 d, TCGv_i32 n, TCGv_i32 m, TCGv_i32 a)
7270 {
7271     tcg_gen_smax_i32(d, a, n);
7272     tcg_gen_smin_i32(d, d, m);
7273 }
7274 
7275 static void gen_sclamp_i64(TCGv_i64 d, TCGv_i64 n, TCGv_i64 m, TCGv_i64 a)
7276 {
7277     tcg_gen_smax_i64(d, a, n);
7278     tcg_gen_smin_i64(d, d, m);
7279 }
7280 
7281 static void gen_sclamp_vec(unsigned vece, TCGv_vec d, TCGv_vec n,
7282                            TCGv_vec m, TCGv_vec a)
7283 {
7284     tcg_gen_smax_vec(vece, d, a, n);
7285     tcg_gen_smin_vec(vece, d, d, m);
7286 }
7287 
7288 static void gen_sclamp(unsigned vece, uint32_t d, uint32_t n, uint32_t m,
7289                        uint32_t a, uint32_t oprsz, uint32_t maxsz)
7290 {
7291     static const TCGOpcode vecop[] = {
7292         INDEX_op_smin_vec, INDEX_op_smax_vec, 0
7293     };
7294     static const GVecGen4 ops[4] = {
7295         { .fniv = gen_sclamp_vec,
7296           .fno  = gen_helper_gvec_sclamp_b,
7297           .opt_opc = vecop,
7298           .vece = MO_8 },
7299         { .fniv = gen_sclamp_vec,
7300           .fno  = gen_helper_gvec_sclamp_h,
7301           .opt_opc = vecop,
7302           .vece = MO_16 },
7303         { .fni4 = gen_sclamp_i32,
7304           .fniv = gen_sclamp_vec,
7305           .fno  = gen_helper_gvec_sclamp_s,
7306           .opt_opc = vecop,
7307           .vece = MO_32 },
7308         { .fni8 = gen_sclamp_i64,
7309           .fniv = gen_sclamp_vec,
7310           .fno  = gen_helper_gvec_sclamp_d,
7311           .opt_opc = vecop,
7312           .vece = MO_64,
7313           .prefer_i64 = TCG_TARGET_REG_BITS == 64 }
7314     };
7315     tcg_gen_gvec_4(d, n, m, a, oprsz, maxsz, &ops[vece]);
7316 }
7317 
7318 TRANS_FEAT(SCLAMP, aa64_sme, gen_gvec_fn_arg_zzzz, gen_sclamp, a)
7319 
7320 static void gen_uclamp_i32(TCGv_i32 d, TCGv_i32 n, TCGv_i32 m, TCGv_i32 a)
7321 {
7322     tcg_gen_umax_i32(d, a, n);
7323     tcg_gen_umin_i32(d, d, m);
7324 }
7325 
7326 static void gen_uclamp_i64(TCGv_i64 d, TCGv_i64 n, TCGv_i64 m, TCGv_i64 a)
7327 {
7328     tcg_gen_umax_i64(d, a, n);
7329     tcg_gen_umin_i64(d, d, m);
7330 }
7331 
7332 static void gen_uclamp_vec(unsigned vece, TCGv_vec d, TCGv_vec n,
7333                            TCGv_vec m, TCGv_vec a)
7334 {
7335     tcg_gen_umax_vec(vece, d, a, n);
7336     tcg_gen_umin_vec(vece, d, d, m);
7337 }
7338 
7339 static void gen_uclamp(unsigned vece, uint32_t d, uint32_t n, uint32_t m,
7340                        uint32_t a, uint32_t oprsz, uint32_t maxsz)
7341 {
7342     static const TCGOpcode vecop[] = {
7343         INDEX_op_umin_vec, INDEX_op_umax_vec, 0
7344     };
7345     static const GVecGen4 ops[4] = {
7346         { .fniv = gen_uclamp_vec,
7347           .fno  = gen_helper_gvec_uclamp_b,
7348           .opt_opc = vecop,
7349           .vece = MO_8 },
7350         { .fniv = gen_uclamp_vec,
7351           .fno  = gen_helper_gvec_uclamp_h,
7352           .opt_opc = vecop,
7353           .vece = MO_16 },
7354         { .fni4 = gen_uclamp_i32,
7355           .fniv = gen_uclamp_vec,
7356           .fno  = gen_helper_gvec_uclamp_s,
7357           .opt_opc = vecop,
7358           .vece = MO_32 },
7359         { .fni8 = gen_uclamp_i64,
7360           .fniv = gen_uclamp_vec,
7361           .fno  = gen_helper_gvec_uclamp_d,
7362           .opt_opc = vecop,
7363           .vece = MO_64,
7364           .prefer_i64 = TCG_TARGET_REG_BITS == 64 }
7365     };
7366     tcg_gen_gvec_4(d, n, m, a, oprsz, maxsz, &ops[vece]);
7367 }
7368 
7369 TRANS_FEAT(UCLAMP, aa64_sme, gen_gvec_fn_arg_zzzz, gen_uclamp, a)
7370