xref: /openbmc/qemu/target/riscv/vector_helper.c (revision b7aee481)
1 /*
2  * RISC-V Vector Extension Helpers for QEMU.
3  *
4  * Copyright (c) 2020 T-Head Semiconductor Co., Ltd. All rights reserved.
5  *
6  * This program is free software; you can redistribute it and/or modify it
7  * under the terms and conditions of the GNU General Public License,
8  * version 2 or later, as published by the Free Software Foundation.
9  *
10  * This program is distributed in the hope it will be useful, but WITHOUT
11  * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
12  * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License for
13  * more details.
14  *
15  * You should have received a copy of the GNU General Public License along with
16  * this program.  If not, see <http://www.gnu.org/licenses/>.
17  */
18 
19 #include "qemu/osdep.h"
20 #include "cpu.h"
21 #include "exec/memop.h"
22 #include "exec/exec-all.h"
23 #include "exec/helper-proto.h"
24 #include "tcg/tcg-gvec-desc.h"
25 #include "internals.h"
26 #include <math.h>
27 
28 target_ulong HELPER(vsetvl)(CPURISCVState *env, target_ulong s1,
29                             target_ulong s2)
30 {
31     int vlmax, vl;
32     RISCVCPU *cpu = env_archcpu(env);
33     uint16_t sew = 8 << FIELD_EX64(s2, VTYPE, VSEW);
34     uint8_t ediv = FIELD_EX64(s2, VTYPE, VEDIV);
35     bool vill = FIELD_EX64(s2, VTYPE, VILL);
36     target_ulong reserved = FIELD_EX64(s2, VTYPE, RESERVED);
37 
38     if ((sew > cpu->cfg.elen) || vill || (ediv != 0) || (reserved != 0)) {
39         /* only set vill bit. */
40         env->vtype = FIELD_DP64(0, VTYPE, VILL, 1);
41         env->vl = 0;
42         env->vstart = 0;
43         return 0;
44     }
45 
46     vlmax = vext_get_vlmax(cpu, s2);
47     if (s1 <= vlmax) {
48         vl = s1;
49     } else {
50         vl = vlmax;
51     }
52     env->vl = vl;
53     env->vtype = s2;
54     env->vstart = 0;
55     return vl;
56 }
57 
58 /*
59  * Note that vector data is stored in host-endian 64-bit chunks,
60  * so addressing units smaller than that needs a host-endian fixup.
61  */
62 #ifdef HOST_WORDS_BIGENDIAN
63 #define H1(x)   ((x) ^ 7)
64 #define H1_2(x) ((x) ^ 6)
65 #define H1_4(x) ((x) ^ 4)
66 #define H2(x)   ((x) ^ 3)
67 #define H4(x)   ((x) ^ 1)
68 #define H8(x)   ((x))
69 #else
70 #define H1(x)   (x)
71 #define H1_2(x) (x)
72 #define H1_4(x) (x)
73 #define H2(x)   (x)
74 #define H4(x)   (x)
75 #define H8(x)   (x)
76 #endif
77 
78 static inline uint32_t vext_nf(uint32_t desc)
79 {
80     return FIELD_EX32(simd_data(desc), VDATA, NF);
81 }
82 
83 static inline uint32_t vext_mlen(uint32_t desc)
84 {
85     return FIELD_EX32(simd_data(desc), VDATA, MLEN);
86 }
87 
88 static inline uint32_t vext_vm(uint32_t desc)
89 {
90     return FIELD_EX32(simd_data(desc), VDATA, VM);
91 }
92 
93 static inline uint32_t vext_lmul(uint32_t desc)
94 {
95     return FIELD_EX32(simd_data(desc), VDATA, LMUL);
96 }
97 
98 static uint32_t vext_wd(uint32_t desc)
99 {
100     return (simd_data(desc) >> 11) & 0x1;
101 }
102 
103 /*
104  * Get vector group length in bytes. Its range is [64, 2048].
105  *
106  * As simd_desc support at most 256, the max vlen is 512 bits.
107  * So vlen in bytes is encoded as maxsz.
108  */
109 static inline uint32_t vext_maxsz(uint32_t desc)
110 {
111     return simd_maxsz(desc) << vext_lmul(desc);
112 }
113 
114 /*
115  * This function checks watchpoint before real load operation.
116  *
117  * In softmmu mode, the TLB API probe_access is enough for watchpoint check.
118  * In user mode, there is no watchpoint support now.
119  *
120  * It will trigger an exception if there is no mapping in TLB
121  * and page table walk can't fill the TLB entry. Then the guest
122  * software can return here after process the exception or never return.
123  */
124 static void probe_pages(CPURISCVState *env, target_ulong addr,
125                         target_ulong len, uintptr_t ra,
126                         MMUAccessType access_type)
127 {
128     target_ulong pagelen = -(addr | TARGET_PAGE_MASK);
129     target_ulong curlen = MIN(pagelen, len);
130 
131     probe_access(env, addr, curlen, access_type,
132                  cpu_mmu_index(env, false), ra);
133     if (len > curlen) {
134         addr += curlen;
135         curlen = len - curlen;
136         probe_access(env, addr, curlen, access_type,
137                      cpu_mmu_index(env, false), ra);
138     }
139 }
140 
141 #ifdef HOST_WORDS_BIGENDIAN
142 static void vext_clear(void *tail, uint32_t cnt, uint32_t tot)
143 {
144     /*
145      * Split the remaining range to two parts.
146      * The first part is in the last uint64_t unit.
147      * The second part start from the next uint64_t unit.
148      */
149     int part1 = 0, part2 = tot - cnt;
150     if (cnt % 8) {
151         part1 = 8 - (cnt % 8);
152         part2 = tot - cnt - part1;
153         memset((void *)((uintptr_t)tail & ~(7ULL)), 0, part1);
154         memset((void *)(((uintptr_t)tail + 8) & ~(7ULL)), 0, part2);
155     } else {
156         memset(tail, 0, part2);
157     }
158 }
159 #else
160 static void vext_clear(void *tail, uint32_t cnt, uint32_t tot)
161 {
162     memset(tail, 0, tot - cnt);
163 }
164 #endif
165 
166 static void clearb(void *vd, uint32_t idx, uint32_t cnt, uint32_t tot)
167 {
168     int8_t *cur = ((int8_t *)vd + H1(idx));
169     vext_clear(cur, cnt, tot);
170 }
171 
172 static void clearh(void *vd, uint32_t idx, uint32_t cnt, uint32_t tot)
173 {
174     int16_t *cur = ((int16_t *)vd + H2(idx));
175     vext_clear(cur, cnt, tot);
176 }
177 
178 static void clearl(void *vd, uint32_t idx, uint32_t cnt, uint32_t tot)
179 {
180     int32_t *cur = ((int32_t *)vd + H4(idx));
181     vext_clear(cur, cnt, tot);
182 }
183 
184 static void clearq(void *vd, uint32_t idx, uint32_t cnt, uint32_t tot)
185 {
186     int64_t *cur = (int64_t *)vd + idx;
187     vext_clear(cur, cnt, tot);
188 }
189 
190 static inline void vext_set_elem_mask(void *v0, int mlen, int index,
191         uint8_t value)
192 {
193     int idx = (index * mlen) / 64;
194     int pos = (index * mlen) % 64;
195     uint64_t old = ((uint64_t *)v0)[idx];
196     ((uint64_t *)v0)[idx] = deposit64(old, pos, mlen, value);
197 }
198 
199 static inline int vext_elem_mask(void *v0, int mlen, int index)
200 {
201     int idx = (index * mlen) / 64;
202     int pos = (index * mlen) % 64;
203     return (((uint64_t *)v0)[idx] >> pos) & 1;
204 }
205 
206 /* elements operations for load and store */
207 typedef void vext_ldst_elem_fn(CPURISCVState *env, target_ulong addr,
208                                uint32_t idx, void *vd, uintptr_t retaddr);
209 typedef void clear_fn(void *vd, uint32_t idx, uint32_t cnt, uint32_t tot);
210 
211 #define GEN_VEXT_LD_ELEM(NAME, MTYPE, ETYPE, H, LDSUF)     \
212 static void NAME(CPURISCVState *env, abi_ptr addr,         \
213                  uint32_t idx, void *vd, uintptr_t retaddr)\
214 {                                                          \
215     MTYPE data;                                            \
216     ETYPE *cur = ((ETYPE *)vd + H(idx));                   \
217     data = cpu_##LDSUF##_data_ra(env, addr, retaddr);      \
218     *cur = data;                                           \
219 }                                                          \
220 
221 GEN_VEXT_LD_ELEM(ldb_b, int8_t,  int8_t,  H1, ldsb)
222 GEN_VEXT_LD_ELEM(ldb_h, int8_t,  int16_t, H2, ldsb)
223 GEN_VEXT_LD_ELEM(ldb_w, int8_t,  int32_t, H4, ldsb)
224 GEN_VEXT_LD_ELEM(ldb_d, int8_t,  int64_t, H8, ldsb)
225 GEN_VEXT_LD_ELEM(ldh_h, int16_t, int16_t, H2, ldsw)
226 GEN_VEXT_LD_ELEM(ldh_w, int16_t, int32_t, H4, ldsw)
227 GEN_VEXT_LD_ELEM(ldh_d, int16_t, int64_t, H8, ldsw)
228 GEN_VEXT_LD_ELEM(ldw_w, int32_t, int32_t, H4, ldl)
229 GEN_VEXT_LD_ELEM(ldw_d, int32_t, int64_t, H8, ldl)
230 GEN_VEXT_LD_ELEM(lde_b, int8_t,  int8_t,  H1, ldsb)
231 GEN_VEXT_LD_ELEM(lde_h, int16_t, int16_t, H2, ldsw)
232 GEN_VEXT_LD_ELEM(lde_w, int32_t, int32_t, H4, ldl)
233 GEN_VEXT_LD_ELEM(lde_d, int64_t, int64_t, H8, ldq)
234 GEN_VEXT_LD_ELEM(ldbu_b, uint8_t,  uint8_t,  H1, ldub)
235 GEN_VEXT_LD_ELEM(ldbu_h, uint8_t,  uint16_t, H2, ldub)
236 GEN_VEXT_LD_ELEM(ldbu_w, uint8_t,  uint32_t, H4, ldub)
237 GEN_VEXT_LD_ELEM(ldbu_d, uint8_t,  uint64_t, H8, ldub)
238 GEN_VEXT_LD_ELEM(ldhu_h, uint16_t, uint16_t, H2, lduw)
239 GEN_VEXT_LD_ELEM(ldhu_w, uint16_t, uint32_t, H4, lduw)
240 GEN_VEXT_LD_ELEM(ldhu_d, uint16_t, uint64_t, H8, lduw)
241 GEN_VEXT_LD_ELEM(ldwu_w, uint32_t, uint32_t, H4, ldl)
242 GEN_VEXT_LD_ELEM(ldwu_d, uint32_t, uint64_t, H8, ldl)
243 
244 #define GEN_VEXT_ST_ELEM(NAME, ETYPE, H, STSUF)            \
245 static void NAME(CPURISCVState *env, abi_ptr addr,         \
246                  uint32_t idx, void *vd, uintptr_t retaddr)\
247 {                                                          \
248     ETYPE data = *((ETYPE *)vd + H(idx));                  \
249     cpu_##STSUF##_data_ra(env, addr, data, retaddr);       \
250 }
251 
252 GEN_VEXT_ST_ELEM(stb_b, int8_t,  H1, stb)
253 GEN_VEXT_ST_ELEM(stb_h, int16_t, H2, stb)
254 GEN_VEXT_ST_ELEM(stb_w, int32_t, H4, stb)
255 GEN_VEXT_ST_ELEM(stb_d, int64_t, H8, stb)
256 GEN_VEXT_ST_ELEM(sth_h, int16_t, H2, stw)
257 GEN_VEXT_ST_ELEM(sth_w, int32_t, H4, stw)
258 GEN_VEXT_ST_ELEM(sth_d, int64_t, H8, stw)
259 GEN_VEXT_ST_ELEM(stw_w, int32_t, H4, stl)
260 GEN_VEXT_ST_ELEM(stw_d, int64_t, H8, stl)
261 GEN_VEXT_ST_ELEM(ste_b, int8_t,  H1, stb)
262 GEN_VEXT_ST_ELEM(ste_h, int16_t, H2, stw)
263 GEN_VEXT_ST_ELEM(ste_w, int32_t, H4, stl)
264 GEN_VEXT_ST_ELEM(ste_d, int64_t, H8, stq)
265 
266 /*
267  *** stride: access vector element from strided memory
268  */
269 static void
270 vext_ldst_stride(void *vd, void *v0, target_ulong base,
271                  target_ulong stride, CPURISCVState *env,
272                  uint32_t desc, uint32_t vm,
273                  vext_ldst_elem_fn *ldst_elem, clear_fn *clear_elem,
274                  uint32_t esz, uint32_t msz, uintptr_t ra,
275                  MMUAccessType access_type)
276 {
277     uint32_t i, k;
278     uint32_t nf = vext_nf(desc);
279     uint32_t mlen = vext_mlen(desc);
280     uint32_t vlmax = vext_maxsz(desc) / esz;
281 
282     /* probe every access*/
283     for (i = 0; i < env->vl; i++) {
284         if (!vm && !vext_elem_mask(v0, mlen, i)) {
285             continue;
286         }
287         probe_pages(env, base + stride * i, nf * msz, ra, access_type);
288     }
289     /* do real access */
290     for (i = 0; i < env->vl; i++) {
291         k = 0;
292         if (!vm && !vext_elem_mask(v0, mlen, i)) {
293             continue;
294         }
295         while (k < nf) {
296             target_ulong addr = base + stride * i + k * msz;
297             ldst_elem(env, addr, i + k * vlmax, vd, ra);
298             k++;
299         }
300     }
301     /* clear tail elements */
302     if (clear_elem) {
303         for (k = 0; k < nf; k++) {
304             clear_elem(vd, env->vl + k * vlmax, env->vl * esz, vlmax * esz);
305         }
306     }
307 }
308 
309 #define GEN_VEXT_LD_STRIDE(NAME, MTYPE, ETYPE, LOAD_FN, CLEAR_FN)       \
310 void HELPER(NAME)(void *vd, void * v0, target_ulong base,               \
311                   target_ulong stride, CPURISCVState *env,              \
312                   uint32_t desc)                                        \
313 {                                                                       \
314     uint32_t vm = vext_vm(desc);                                        \
315     vext_ldst_stride(vd, v0, base, stride, env, desc, vm, LOAD_FN,      \
316                      CLEAR_FN, sizeof(ETYPE), sizeof(MTYPE),            \
317                      GETPC(), MMU_DATA_LOAD);                           \
318 }
319 
320 GEN_VEXT_LD_STRIDE(vlsb_v_b,  int8_t,   int8_t,   ldb_b,  clearb)
321 GEN_VEXT_LD_STRIDE(vlsb_v_h,  int8_t,   int16_t,  ldb_h,  clearh)
322 GEN_VEXT_LD_STRIDE(vlsb_v_w,  int8_t,   int32_t,  ldb_w,  clearl)
323 GEN_VEXT_LD_STRIDE(vlsb_v_d,  int8_t,   int64_t,  ldb_d,  clearq)
324 GEN_VEXT_LD_STRIDE(vlsh_v_h,  int16_t,  int16_t,  ldh_h,  clearh)
325 GEN_VEXT_LD_STRIDE(vlsh_v_w,  int16_t,  int32_t,  ldh_w,  clearl)
326 GEN_VEXT_LD_STRIDE(vlsh_v_d,  int16_t,  int64_t,  ldh_d,  clearq)
327 GEN_VEXT_LD_STRIDE(vlsw_v_w,  int32_t,  int32_t,  ldw_w,  clearl)
328 GEN_VEXT_LD_STRIDE(vlsw_v_d,  int32_t,  int64_t,  ldw_d,  clearq)
329 GEN_VEXT_LD_STRIDE(vlse_v_b,  int8_t,   int8_t,   lde_b,  clearb)
330 GEN_VEXT_LD_STRIDE(vlse_v_h,  int16_t,  int16_t,  lde_h,  clearh)
331 GEN_VEXT_LD_STRIDE(vlse_v_w,  int32_t,  int32_t,  lde_w,  clearl)
332 GEN_VEXT_LD_STRIDE(vlse_v_d,  int64_t,  int64_t,  lde_d,  clearq)
333 GEN_VEXT_LD_STRIDE(vlsbu_v_b, uint8_t,  uint8_t,  ldbu_b, clearb)
334 GEN_VEXT_LD_STRIDE(vlsbu_v_h, uint8_t,  uint16_t, ldbu_h, clearh)
335 GEN_VEXT_LD_STRIDE(vlsbu_v_w, uint8_t,  uint32_t, ldbu_w, clearl)
336 GEN_VEXT_LD_STRIDE(vlsbu_v_d, uint8_t,  uint64_t, ldbu_d, clearq)
337 GEN_VEXT_LD_STRIDE(vlshu_v_h, uint16_t, uint16_t, ldhu_h, clearh)
338 GEN_VEXT_LD_STRIDE(vlshu_v_w, uint16_t, uint32_t, ldhu_w, clearl)
339 GEN_VEXT_LD_STRIDE(vlshu_v_d, uint16_t, uint64_t, ldhu_d, clearq)
340 GEN_VEXT_LD_STRIDE(vlswu_v_w, uint32_t, uint32_t, ldwu_w, clearl)
341 GEN_VEXT_LD_STRIDE(vlswu_v_d, uint32_t, uint64_t, ldwu_d, clearq)
342 
343 #define GEN_VEXT_ST_STRIDE(NAME, MTYPE, ETYPE, STORE_FN)                \
344 void HELPER(NAME)(void *vd, void *v0, target_ulong base,                \
345                   target_ulong stride, CPURISCVState *env,              \
346                   uint32_t desc)                                        \
347 {                                                                       \
348     uint32_t vm = vext_vm(desc);                                        \
349     vext_ldst_stride(vd, v0, base, stride, env, desc, vm, STORE_FN,     \
350                      NULL, sizeof(ETYPE), sizeof(MTYPE),                \
351                      GETPC(), MMU_DATA_STORE);                          \
352 }
353 
354 GEN_VEXT_ST_STRIDE(vssb_v_b, int8_t,  int8_t,  stb_b)
355 GEN_VEXT_ST_STRIDE(vssb_v_h, int8_t,  int16_t, stb_h)
356 GEN_VEXT_ST_STRIDE(vssb_v_w, int8_t,  int32_t, stb_w)
357 GEN_VEXT_ST_STRIDE(vssb_v_d, int8_t,  int64_t, stb_d)
358 GEN_VEXT_ST_STRIDE(vssh_v_h, int16_t, int16_t, sth_h)
359 GEN_VEXT_ST_STRIDE(vssh_v_w, int16_t, int32_t, sth_w)
360 GEN_VEXT_ST_STRIDE(vssh_v_d, int16_t, int64_t, sth_d)
361 GEN_VEXT_ST_STRIDE(vssw_v_w, int32_t, int32_t, stw_w)
362 GEN_VEXT_ST_STRIDE(vssw_v_d, int32_t, int64_t, stw_d)
363 GEN_VEXT_ST_STRIDE(vsse_v_b, int8_t,  int8_t,  ste_b)
364 GEN_VEXT_ST_STRIDE(vsse_v_h, int16_t, int16_t, ste_h)
365 GEN_VEXT_ST_STRIDE(vsse_v_w, int32_t, int32_t, ste_w)
366 GEN_VEXT_ST_STRIDE(vsse_v_d, int64_t, int64_t, ste_d)
367 
368 /*
369  *** unit-stride: access elements stored contiguously in memory
370  */
371 
372 /* unmasked unit-stride load and store operation*/
373 static void
374 vext_ldst_us(void *vd, target_ulong base, CPURISCVState *env, uint32_t desc,
375              vext_ldst_elem_fn *ldst_elem, clear_fn *clear_elem,
376              uint32_t esz, uint32_t msz, uintptr_t ra,
377              MMUAccessType access_type)
378 {
379     uint32_t i, k;
380     uint32_t nf = vext_nf(desc);
381     uint32_t vlmax = vext_maxsz(desc) / esz;
382 
383     /* probe every access */
384     probe_pages(env, base, env->vl * nf * msz, ra, access_type);
385     /* load bytes from guest memory */
386     for (i = 0; i < env->vl; i++) {
387         k = 0;
388         while (k < nf) {
389             target_ulong addr = base + (i * nf + k) * msz;
390             ldst_elem(env, addr, i + k * vlmax, vd, ra);
391             k++;
392         }
393     }
394     /* clear tail elements */
395     if (clear_elem) {
396         for (k = 0; k < nf; k++) {
397             clear_elem(vd, env->vl + k * vlmax, env->vl * esz, vlmax * esz);
398         }
399     }
400 }
401 
402 /*
403  * masked unit-stride load and store operation will be a special case of stride,
404  * stride = NF * sizeof (MTYPE)
405  */
406 
407 #define GEN_VEXT_LD_US(NAME, MTYPE, ETYPE, LOAD_FN, CLEAR_FN)           \
408 void HELPER(NAME##_mask)(void *vd, void *v0, target_ulong base,         \
409                          CPURISCVState *env, uint32_t desc)             \
410 {                                                                       \
411     uint32_t stride = vext_nf(desc) * sizeof(MTYPE);                    \
412     vext_ldst_stride(vd, v0, base, stride, env, desc, false, LOAD_FN,   \
413                      CLEAR_FN, sizeof(ETYPE), sizeof(MTYPE),            \
414                      GETPC(), MMU_DATA_LOAD);                           \
415 }                                                                       \
416                                                                         \
417 void HELPER(NAME)(void *vd, void *v0, target_ulong base,                \
418                   CPURISCVState *env, uint32_t desc)                    \
419 {                                                                       \
420     vext_ldst_us(vd, base, env, desc, LOAD_FN, CLEAR_FN,                \
421                  sizeof(ETYPE), sizeof(MTYPE), GETPC(), MMU_DATA_LOAD); \
422 }
423 
424 GEN_VEXT_LD_US(vlb_v_b,  int8_t,   int8_t,   ldb_b,  clearb)
425 GEN_VEXT_LD_US(vlb_v_h,  int8_t,   int16_t,  ldb_h,  clearh)
426 GEN_VEXT_LD_US(vlb_v_w,  int8_t,   int32_t,  ldb_w,  clearl)
427 GEN_VEXT_LD_US(vlb_v_d,  int8_t,   int64_t,  ldb_d,  clearq)
428 GEN_VEXT_LD_US(vlh_v_h,  int16_t,  int16_t,  ldh_h,  clearh)
429 GEN_VEXT_LD_US(vlh_v_w,  int16_t,  int32_t,  ldh_w,  clearl)
430 GEN_VEXT_LD_US(vlh_v_d,  int16_t,  int64_t,  ldh_d,  clearq)
431 GEN_VEXT_LD_US(vlw_v_w,  int32_t,  int32_t,  ldw_w,  clearl)
432 GEN_VEXT_LD_US(vlw_v_d,  int32_t,  int64_t,  ldw_d,  clearq)
433 GEN_VEXT_LD_US(vle_v_b,  int8_t,   int8_t,   lde_b,  clearb)
434 GEN_VEXT_LD_US(vle_v_h,  int16_t,  int16_t,  lde_h,  clearh)
435 GEN_VEXT_LD_US(vle_v_w,  int32_t,  int32_t,  lde_w,  clearl)
436 GEN_VEXT_LD_US(vle_v_d,  int64_t,  int64_t,  lde_d,  clearq)
437 GEN_VEXT_LD_US(vlbu_v_b, uint8_t,  uint8_t,  ldbu_b, clearb)
438 GEN_VEXT_LD_US(vlbu_v_h, uint8_t,  uint16_t, ldbu_h, clearh)
439 GEN_VEXT_LD_US(vlbu_v_w, uint8_t,  uint32_t, ldbu_w, clearl)
440 GEN_VEXT_LD_US(vlbu_v_d, uint8_t,  uint64_t, ldbu_d, clearq)
441 GEN_VEXT_LD_US(vlhu_v_h, uint16_t, uint16_t, ldhu_h, clearh)
442 GEN_VEXT_LD_US(vlhu_v_w, uint16_t, uint32_t, ldhu_w, clearl)
443 GEN_VEXT_LD_US(vlhu_v_d, uint16_t, uint64_t, ldhu_d, clearq)
444 GEN_VEXT_LD_US(vlwu_v_w, uint32_t, uint32_t, ldwu_w, clearl)
445 GEN_VEXT_LD_US(vlwu_v_d, uint32_t, uint64_t, ldwu_d, clearq)
446 
447 #define GEN_VEXT_ST_US(NAME, MTYPE, ETYPE, STORE_FN)                    \
448 void HELPER(NAME##_mask)(void *vd, void *v0, target_ulong base,         \
449                          CPURISCVState *env, uint32_t desc)             \
450 {                                                                       \
451     uint32_t stride = vext_nf(desc) * sizeof(MTYPE);                    \
452     vext_ldst_stride(vd, v0, base, stride, env, desc, false, STORE_FN,  \
453                      NULL, sizeof(ETYPE), sizeof(MTYPE),                \
454                      GETPC(), MMU_DATA_STORE);                          \
455 }                                                                       \
456                                                                         \
457 void HELPER(NAME)(void *vd, void *v0, target_ulong base,                \
458                   CPURISCVState *env, uint32_t desc)                    \
459 {                                                                       \
460     vext_ldst_us(vd, base, env, desc, STORE_FN, NULL,                   \
461                  sizeof(ETYPE), sizeof(MTYPE), GETPC(), MMU_DATA_STORE);\
462 }
463 
464 GEN_VEXT_ST_US(vsb_v_b, int8_t,  int8_t , stb_b)
465 GEN_VEXT_ST_US(vsb_v_h, int8_t,  int16_t, stb_h)
466 GEN_VEXT_ST_US(vsb_v_w, int8_t,  int32_t, stb_w)
467 GEN_VEXT_ST_US(vsb_v_d, int8_t,  int64_t, stb_d)
468 GEN_VEXT_ST_US(vsh_v_h, int16_t, int16_t, sth_h)
469 GEN_VEXT_ST_US(vsh_v_w, int16_t, int32_t, sth_w)
470 GEN_VEXT_ST_US(vsh_v_d, int16_t, int64_t, sth_d)
471 GEN_VEXT_ST_US(vsw_v_w, int32_t, int32_t, stw_w)
472 GEN_VEXT_ST_US(vsw_v_d, int32_t, int64_t, stw_d)
473 GEN_VEXT_ST_US(vse_v_b, int8_t,  int8_t , ste_b)
474 GEN_VEXT_ST_US(vse_v_h, int16_t, int16_t, ste_h)
475 GEN_VEXT_ST_US(vse_v_w, int32_t, int32_t, ste_w)
476 GEN_VEXT_ST_US(vse_v_d, int64_t, int64_t, ste_d)
477 
478 /*
479  *** index: access vector element from indexed memory
480  */
481 typedef target_ulong vext_get_index_addr(target_ulong base,
482         uint32_t idx, void *vs2);
483 
484 #define GEN_VEXT_GET_INDEX_ADDR(NAME, ETYPE, H)        \
485 static target_ulong NAME(target_ulong base,            \
486                          uint32_t idx, void *vs2)      \
487 {                                                      \
488     return (base + *((ETYPE *)vs2 + H(idx)));          \
489 }
490 
491 GEN_VEXT_GET_INDEX_ADDR(idx_b, int8_t,  H1)
492 GEN_VEXT_GET_INDEX_ADDR(idx_h, int16_t, H2)
493 GEN_VEXT_GET_INDEX_ADDR(idx_w, int32_t, H4)
494 GEN_VEXT_GET_INDEX_ADDR(idx_d, int64_t, H8)
495 
496 static inline void
497 vext_ldst_index(void *vd, void *v0, target_ulong base,
498                 void *vs2, CPURISCVState *env, uint32_t desc,
499                 vext_get_index_addr get_index_addr,
500                 vext_ldst_elem_fn *ldst_elem,
501                 clear_fn *clear_elem,
502                 uint32_t esz, uint32_t msz, uintptr_t ra,
503                 MMUAccessType access_type)
504 {
505     uint32_t i, k;
506     uint32_t nf = vext_nf(desc);
507     uint32_t vm = vext_vm(desc);
508     uint32_t mlen = vext_mlen(desc);
509     uint32_t vlmax = vext_maxsz(desc) / esz;
510 
511     /* probe every access*/
512     for (i = 0; i < env->vl; i++) {
513         if (!vm && !vext_elem_mask(v0, mlen, i)) {
514             continue;
515         }
516         probe_pages(env, get_index_addr(base, i, vs2), nf * msz, ra,
517                     access_type);
518     }
519     /* load bytes from guest memory */
520     for (i = 0; i < env->vl; i++) {
521         k = 0;
522         if (!vm && !vext_elem_mask(v0, mlen, i)) {
523             continue;
524         }
525         while (k < nf) {
526             abi_ptr addr = get_index_addr(base, i, vs2) + k * msz;
527             ldst_elem(env, addr, i + k * vlmax, vd, ra);
528             k++;
529         }
530     }
531     /* clear tail elements */
532     if (clear_elem) {
533         for (k = 0; k < nf; k++) {
534             clear_elem(vd, env->vl + k * vlmax, env->vl * esz, vlmax * esz);
535         }
536     }
537 }
538 
539 #define GEN_VEXT_LD_INDEX(NAME, MTYPE, ETYPE, INDEX_FN, LOAD_FN, CLEAR_FN) \
540 void HELPER(NAME)(void *vd, void *v0, target_ulong base,                   \
541                   void *vs2, CPURISCVState *env, uint32_t desc)            \
542 {                                                                          \
543     vext_ldst_index(vd, v0, base, vs2, env, desc, INDEX_FN,                \
544                     LOAD_FN, CLEAR_FN, sizeof(ETYPE), sizeof(MTYPE),       \
545                     GETPC(), MMU_DATA_LOAD);                               \
546 }
547 
548 GEN_VEXT_LD_INDEX(vlxb_v_b,  int8_t,   int8_t,   idx_b, ldb_b,  clearb)
549 GEN_VEXT_LD_INDEX(vlxb_v_h,  int8_t,   int16_t,  idx_h, ldb_h,  clearh)
550 GEN_VEXT_LD_INDEX(vlxb_v_w,  int8_t,   int32_t,  idx_w, ldb_w,  clearl)
551 GEN_VEXT_LD_INDEX(vlxb_v_d,  int8_t,   int64_t,  idx_d, ldb_d,  clearq)
552 GEN_VEXT_LD_INDEX(vlxh_v_h,  int16_t,  int16_t,  idx_h, ldh_h,  clearh)
553 GEN_VEXT_LD_INDEX(vlxh_v_w,  int16_t,  int32_t,  idx_w, ldh_w,  clearl)
554 GEN_VEXT_LD_INDEX(vlxh_v_d,  int16_t,  int64_t,  idx_d, ldh_d,  clearq)
555 GEN_VEXT_LD_INDEX(vlxw_v_w,  int32_t,  int32_t,  idx_w, ldw_w,  clearl)
556 GEN_VEXT_LD_INDEX(vlxw_v_d,  int32_t,  int64_t,  idx_d, ldw_d,  clearq)
557 GEN_VEXT_LD_INDEX(vlxe_v_b,  int8_t,   int8_t,   idx_b, lde_b,  clearb)
558 GEN_VEXT_LD_INDEX(vlxe_v_h,  int16_t,  int16_t,  idx_h, lde_h,  clearh)
559 GEN_VEXT_LD_INDEX(vlxe_v_w,  int32_t,  int32_t,  idx_w, lde_w,  clearl)
560 GEN_VEXT_LD_INDEX(vlxe_v_d,  int64_t,  int64_t,  idx_d, lde_d,  clearq)
561 GEN_VEXT_LD_INDEX(vlxbu_v_b, uint8_t,  uint8_t,  idx_b, ldbu_b, clearb)
562 GEN_VEXT_LD_INDEX(vlxbu_v_h, uint8_t,  uint16_t, idx_h, ldbu_h, clearh)
563 GEN_VEXT_LD_INDEX(vlxbu_v_w, uint8_t,  uint32_t, idx_w, ldbu_w, clearl)
564 GEN_VEXT_LD_INDEX(vlxbu_v_d, uint8_t,  uint64_t, idx_d, ldbu_d, clearq)
565 GEN_VEXT_LD_INDEX(vlxhu_v_h, uint16_t, uint16_t, idx_h, ldhu_h, clearh)
566 GEN_VEXT_LD_INDEX(vlxhu_v_w, uint16_t, uint32_t, idx_w, ldhu_w, clearl)
567 GEN_VEXT_LD_INDEX(vlxhu_v_d, uint16_t, uint64_t, idx_d, ldhu_d, clearq)
568 GEN_VEXT_LD_INDEX(vlxwu_v_w, uint32_t, uint32_t, idx_w, ldwu_w, clearl)
569 GEN_VEXT_LD_INDEX(vlxwu_v_d, uint32_t, uint64_t, idx_d, ldwu_d, clearq)
570 
571 #define GEN_VEXT_ST_INDEX(NAME, MTYPE, ETYPE, INDEX_FN, STORE_FN)\
572 void HELPER(NAME)(void *vd, void *v0, target_ulong base,         \
573                   void *vs2, CPURISCVState *env, uint32_t desc)  \
574 {                                                                \
575     vext_ldst_index(vd, v0, base, vs2, env, desc, INDEX_FN,      \
576                     STORE_FN, NULL, sizeof(ETYPE), sizeof(MTYPE),\
577                     GETPC(), MMU_DATA_STORE);                    \
578 }
579 
580 GEN_VEXT_ST_INDEX(vsxb_v_b, int8_t,  int8_t,  idx_b, stb_b)
581 GEN_VEXT_ST_INDEX(vsxb_v_h, int8_t,  int16_t, idx_h, stb_h)
582 GEN_VEXT_ST_INDEX(vsxb_v_w, int8_t,  int32_t, idx_w, stb_w)
583 GEN_VEXT_ST_INDEX(vsxb_v_d, int8_t,  int64_t, idx_d, stb_d)
584 GEN_VEXT_ST_INDEX(vsxh_v_h, int16_t, int16_t, idx_h, sth_h)
585 GEN_VEXT_ST_INDEX(vsxh_v_w, int16_t, int32_t, idx_w, sth_w)
586 GEN_VEXT_ST_INDEX(vsxh_v_d, int16_t, int64_t, idx_d, sth_d)
587 GEN_VEXT_ST_INDEX(vsxw_v_w, int32_t, int32_t, idx_w, stw_w)
588 GEN_VEXT_ST_INDEX(vsxw_v_d, int32_t, int64_t, idx_d, stw_d)
589 GEN_VEXT_ST_INDEX(vsxe_v_b, int8_t,  int8_t,  idx_b, ste_b)
590 GEN_VEXT_ST_INDEX(vsxe_v_h, int16_t, int16_t, idx_h, ste_h)
591 GEN_VEXT_ST_INDEX(vsxe_v_w, int32_t, int32_t, idx_w, ste_w)
592 GEN_VEXT_ST_INDEX(vsxe_v_d, int64_t, int64_t, idx_d, ste_d)
593 
594 /*
595  *** unit-stride fault-only-fisrt load instructions
596  */
597 static inline void
598 vext_ldff(void *vd, void *v0, target_ulong base,
599           CPURISCVState *env, uint32_t desc,
600           vext_ldst_elem_fn *ldst_elem,
601           clear_fn *clear_elem,
602           uint32_t esz, uint32_t msz, uintptr_t ra)
603 {
604     void *host;
605     uint32_t i, k, vl = 0;
606     uint32_t mlen = vext_mlen(desc);
607     uint32_t nf = vext_nf(desc);
608     uint32_t vm = vext_vm(desc);
609     uint32_t vlmax = vext_maxsz(desc) / esz;
610     target_ulong addr, offset, remain;
611 
612     /* probe every access*/
613     for (i = 0; i < env->vl; i++) {
614         if (!vm && !vext_elem_mask(v0, mlen, i)) {
615             continue;
616         }
617         addr = base + nf * i * msz;
618         if (i == 0) {
619             probe_pages(env, addr, nf * msz, ra, MMU_DATA_LOAD);
620         } else {
621             /* if it triggers an exception, no need to check watchpoint */
622             remain = nf * msz;
623             while (remain > 0) {
624                 offset = -(addr | TARGET_PAGE_MASK);
625                 host = tlb_vaddr_to_host(env, addr, MMU_DATA_LOAD,
626                                          cpu_mmu_index(env, false));
627                 if (host) {
628 #ifdef CONFIG_USER_ONLY
629                     if (page_check_range(addr, nf * msz, PAGE_READ) < 0) {
630                         vl = i;
631                         goto ProbeSuccess;
632                     }
633 #else
634                     probe_pages(env, addr, nf * msz, ra, MMU_DATA_LOAD);
635 #endif
636                 } else {
637                     vl = i;
638                     goto ProbeSuccess;
639                 }
640                 if (remain <=  offset) {
641                     break;
642                 }
643                 remain -= offset;
644                 addr += offset;
645             }
646         }
647     }
648 ProbeSuccess:
649     /* load bytes from guest memory */
650     if (vl != 0) {
651         env->vl = vl;
652     }
653     for (i = 0; i < env->vl; i++) {
654         k = 0;
655         if (!vm && !vext_elem_mask(v0, mlen, i)) {
656             continue;
657         }
658         while (k < nf) {
659             target_ulong addr = base + (i * nf + k) * msz;
660             ldst_elem(env, addr, i + k * vlmax, vd, ra);
661             k++;
662         }
663     }
664     /* clear tail elements */
665     if (vl != 0) {
666         return;
667     }
668     for (k = 0; k < nf; k++) {
669         clear_elem(vd, env->vl + k * vlmax, env->vl * esz, vlmax * esz);
670     }
671 }
672 
673 #define GEN_VEXT_LDFF(NAME, MTYPE, ETYPE, LOAD_FN, CLEAR_FN)     \
674 void HELPER(NAME)(void *vd, void *v0, target_ulong base,         \
675                   CPURISCVState *env, uint32_t desc)             \
676 {                                                                \
677     vext_ldff(vd, v0, base, env, desc, LOAD_FN, CLEAR_FN,        \
678               sizeof(ETYPE), sizeof(MTYPE), GETPC());            \
679 }
680 
681 GEN_VEXT_LDFF(vlbff_v_b,  int8_t,   int8_t,   ldb_b,  clearb)
682 GEN_VEXT_LDFF(vlbff_v_h,  int8_t,   int16_t,  ldb_h,  clearh)
683 GEN_VEXT_LDFF(vlbff_v_w,  int8_t,   int32_t,  ldb_w,  clearl)
684 GEN_VEXT_LDFF(vlbff_v_d,  int8_t,   int64_t,  ldb_d,  clearq)
685 GEN_VEXT_LDFF(vlhff_v_h,  int16_t,  int16_t,  ldh_h,  clearh)
686 GEN_VEXT_LDFF(vlhff_v_w,  int16_t,  int32_t,  ldh_w,  clearl)
687 GEN_VEXT_LDFF(vlhff_v_d,  int16_t,  int64_t,  ldh_d,  clearq)
688 GEN_VEXT_LDFF(vlwff_v_w,  int32_t,  int32_t,  ldw_w,  clearl)
689 GEN_VEXT_LDFF(vlwff_v_d,  int32_t,  int64_t,  ldw_d,  clearq)
690 GEN_VEXT_LDFF(vleff_v_b,  int8_t,   int8_t,   lde_b,  clearb)
691 GEN_VEXT_LDFF(vleff_v_h,  int16_t,  int16_t,  lde_h,  clearh)
692 GEN_VEXT_LDFF(vleff_v_w,  int32_t,  int32_t,  lde_w,  clearl)
693 GEN_VEXT_LDFF(vleff_v_d,  int64_t,  int64_t,  lde_d,  clearq)
694 GEN_VEXT_LDFF(vlbuff_v_b, uint8_t,  uint8_t,  ldbu_b, clearb)
695 GEN_VEXT_LDFF(vlbuff_v_h, uint8_t,  uint16_t, ldbu_h, clearh)
696 GEN_VEXT_LDFF(vlbuff_v_w, uint8_t,  uint32_t, ldbu_w, clearl)
697 GEN_VEXT_LDFF(vlbuff_v_d, uint8_t,  uint64_t, ldbu_d, clearq)
698 GEN_VEXT_LDFF(vlhuff_v_h, uint16_t, uint16_t, ldhu_h, clearh)
699 GEN_VEXT_LDFF(vlhuff_v_w, uint16_t, uint32_t, ldhu_w, clearl)
700 GEN_VEXT_LDFF(vlhuff_v_d, uint16_t, uint64_t, ldhu_d, clearq)
701 GEN_VEXT_LDFF(vlwuff_v_w, uint32_t, uint32_t, ldwu_w, clearl)
702 GEN_VEXT_LDFF(vlwuff_v_d, uint32_t, uint64_t, ldwu_d, clearq)
703 
704 /*
705  *** Vector AMO Operations (Zvamo)
706  */
707 typedef void vext_amo_noatomic_fn(void *vs3, target_ulong addr,
708                                   uint32_t wd, uint32_t idx, CPURISCVState *env,
709                                   uintptr_t retaddr);
710 
711 /* no atomic opreation for vector atomic insructions */
712 #define DO_SWAP(N, M) (M)
713 #define DO_AND(N, M)  (N & M)
714 #define DO_XOR(N, M)  (N ^ M)
715 #define DO_OR(N, M)   (N | M)
716 #define DO_ADD(N, M)  (N + M)
717 
718 #define GEN_VEXT_AMO_NOATOMIC_OP(NAME, ESZ, MSZ, H, DO_OP, SUF) \
719 static void                                                     \
720 vext_##NAME##_noatomic_op(void *vs3, target_ulong addr,         \
721                           uint32_t wd, uint32_t idx,            \
722                           CPURISCVState *env, uintptr_t retaddr)\
723 {                                                               \
724     typedef int##ESZ##_t ETYPE;                                 \
725     typedef int##MSZ##_t MTYPE;                                 \
726     typedef uint##MSZ##_t UMTYPE __attribute__((unused));       \
727     ETYPE *pe3 = (ETYPE *)vs3 + H(idx);                         \
728     MTYPE  a = cpu_ld##SUF##_data(env, addr), b = *pe3;         \
729                                                                 \
730     cpu_st##SUF##_data(env, addr, DO_OP(a, b));                 \
731     if (wd) {                                                   \
732         *pe3 = a;                                               \
733     }                                                           \
734 }
735 
736 /* Signed min/max */
737 #define DO_MAX(N, M)  ((N) >= (M) ? (N) : (M))
738 #define DO_MIN(N, M)  ((N) >= (M) ? (M) : (N))
739 
740 /* Unsigned min/max */
741 #define DO_MAXU(N, M) DO_MAX((UMTYPE)N, (UMTYPE)M)
742 #define DO_MINU(N, M) DO_MIN((UMTYPE)N, (UMTYPE)M)
743 
744 GEN_VEXT_AMO_NOATOMIC_OP(vamoswapw_v_w, 32, 32, H4, DO_SWAP, l)
745 GEN_VEXT_AMO_NOATOMIC_OP(vamoaddw_v_w,  32, 32, H4, DO_ADD,  l)
746 GEN_VEXT_AMO_NOATOMIC_OP(vamoxorw_v_w,  32, 32, H4, DO_XOR,  l)
747 GEN_VEXT_AMO_NOATOMIC_OP(vamoandw_v_w,  32, 32, H4, DO_AND,  l)
748 GEN_VEXT_AMO_NOATOMIC_OP(vamoorw_v_w,   32, 32, H4, DO_OR,   l)
749 GEN_VEXT_AMO_NOATOMIC_OP(vamominw_v_w,  32, 32, H4, DO_MIN,  l)
750 GEN_VEXT_AMO_NOATOMIC_OP(vamomaxw_v_w,  32, 32, H4, DO_MAX,  l)
751 GEN_VEXT_AMO_NOATOMIC_OP(vamominuw_v_w, 32, 32, H4, DO_MINU, l)
752 GEN_VEXT_AMO_NOATOMIC_OP(vamomaxuw_v_w, 32, 32, H4, DO_MAXU, l)
753 #ifdef TARGET_RISCV64
754 GEN_VEXT_AMO_NOATOMIC_OP(vamoswapw_v_d, 64, 32, H8, DO_SWAP, l)
755 GEN_VEXT_AMO_NOATOMIC_OP(vamoswapd_v_d, 64, 64, H8, DO_SWAP, q)
756 GEN_VEXT_AMO_NOATOMIC_OP(vamoaddw_v_d,  64, 32, H8, DO_ADD,  l)
757 GEN_VEXT_AMO_NOATOMIC_OP(vamoaddd_v_d,  64, 64, H8, DO_ADD,  q)
758 GEN_VEXT_AMO_NOATOMIC_OP(vamoxorw_v_d,  64, 32, H8, DO_XOR,  l)
759 GEN_VEXT_AMO_NOATOMIC_OP(vamoxord_v_d,  64, 64, H8, DO_XOR,  q)
760 GEN_VEXT_AMO_NOATOMIC_OP(vamoandw_v_d,  64, 32, H8, DO_AND,  l)
761 GEN_VEXT_AMO_NOATOMIC_OP(vamoandd_v_d,  64, 64, H8, DO_AND,  q)
762 GEN_VEXT_AMO_NOATOMIC_OP(vamoorw_v_d,   64, 32, H8, DO_OR,   l)
763 GEN_VEXT_AMO_NOATOMIC_OP(vamoord_v_d,   64, 64, H8, DO_OR,   q)
764 GEN_VEXT_AMO_NOATOMIC_OP(vamominw_v_d,  64, 32, H8, DO_MIN,  l)
765 GEN_VEXT_AMO_NOATOMIC_OP(vamomind_v_d,  64, 64, H8, DO_MIN,  q)
766 GEN_VEXT_AMO_NOATOMIC_OP(vamomaxw_v_d,  64, 32, H8, DO_MAX,  l)
767 GEN_VEXT_AMO_NOATOMIC_OP(vamomaxd_v_d,  64, 64, H8, DO_MAX,  q)
768 GEN_VEXT_AMO_NOATOMIC_OP(vamominuw_v_d, 64, 32, H8, DO_MINU, l)
769 GEN_VEXT_AMO_NOATOMIC_OP(vamominud_v_d, 64, 64, H8, DO_MINU, q)
770 GEN_VEXT_AMO_NOATOMIC_OP(vamomaxuw_v_d, 64, 32, H8, DO_MAXU, l)
771 GEN_VEXT_AMO_NOATOMIC_OP(vamomaxud_v_d, 64, 64, H8, DO_MAXU, q)
772 #endif
773 
774 static inline void
775 vext_amo_noatomic(void *vs3, void *v0, target_ulong base,
776                   void *vs2, CPURISCVState *env, uint32_t desc,
777                   vext_get_index_addr get_index_addr,
778                   vext_amo_noatomic_fn *noatomic_op,
779                   clear_fn *clear_elem,
780                   uint32_t esz, uint32_t msz, uintptr_t ra)
781 {
782     uint32_t i;
783     target_long addr;
784     uint32_t wd = vext_wd(desc);
785     uint32_t vm = vext_vm(desc);
786     uint32_t mlen = vext_mlen(desc);
787     uint32_t vlmax = vext_maxsz(desc) / esz;
788 
789     for (i = 0; i < env->vl; i++) {
790         if (!vm && !vext_elem_mask(v0, mlen, i)) {
791             continue;
792         }
793         probe_pages(env, get_index_addr(base, i, vs2), msz, ra, MMU_DATA_LOAD);
794         probe_pages(env, get_index_addr(base, i, vs2), msz, ra, MMU_DATA_STORE);
795     }
796     for (i = 0; i < env->vl; i++) {
797         if (!vm && !vext_elem_mask(v0, mlen, i)) {
798             continue;
799         }
800         addr = get_index_addr(base, i, vs2);
801         noatomic_op(vs3, addr, wd, i, env, ra);
802     }
803     clear_elem(vs3, env->vl, env->vl * esz, vlmax * esz);
804 }
805 
806 #define GEN_VEXT_AMO(NAME, MTYPE, ETYPE, INDEX_FN, CLEAR_FN)    \
807 void HELPER(NAME)(void *vs3, void *v0, target_ulong base,       \
808                   void *vs2, CPURISCVState *env, uint32_t desc) \
809 {                                                               \
810     vext_amo_noatomic(vs3, v0, base, vs2, env, desc,            \
811                       INDEX_FN, vext_##NAME##_noatomic_op,      \
812                       CLEAR_FN, sizeof(ETYPE), sizeof(MTYPE),   \
813                       GETPC());                                 \
814 }
815 
816 #ifdef TARGET_RISCV64
817 GEN_VEXT_AMO(vamoswapw_v_d, int32_t,  int64_t,  idx_d, clearq)
818 GEN_VEXT_AMO(vamoswapd_v_d, int64_t,  int64_t,  idx_d, clearq)
819 GEN_VEXT_AMO(vamoaddw_v_d,  int32_t,  int64_t,  idx_d, clearq)
820 GEN_VEXT_AMO(vamoaddd_v_d,  int64_t,  int64_t,  idx_d, clearq)
821 GEN_VEXT_AMO(vamoxorw_v_d,  int32_t,  int64_t,  idx_d, clearq)
822 GEN_VEXT_AMO(vamoxord_v_d,  int64_t,  int64_t,  idx_d, clearq)
823 GEN_VEXT_AMO(vamoandw_v_d,  int32_t,  int64_t,  idx_d, clearq)
824 GEN_VEXT_AMO(vamoandd_v_d,  int64_t,  int64_t,  idx_d, clearq)
825 GEN_VEXT_AMO(vamoorw_v_d,   int32_t,  int64_t,  idx_d, clearq)
826 GEN_VEXT_AMO(vamoord_v_d,   int64_t,  int64_t,  idx_d, clearq)
827 GEN_VEXT_AMO(vamominw_v_d,  int32_t,  int64_t,  idx_d, clearq)
828 GEN_VEXT_AMO(vamomind_v_d,  int64_t,  int64_t,  idx_d, clearq)
829 GEN_VEXT_AMO(vamomaxw_v_d,  int32_t,  int64_t,  idx_d, clearq)
830 GEN_VEXT_AMO(vamomaxd_v_d,  int64_t,  int64_t,  idx_d, clearq)
831 GEN_VEXT_AMO(vamominuw_v_d, uint32_t, uint64_t, idx_d, clearq)
832 GEN_VEXT_AMO(vamominud_v_d, uint64_t, uint64_t, idx_d, clearq)
833 GEN_VEXT_AMO(vamomaxuw_v_d, uint32_t, uint64_t, idx_d, clearq)
834 GEN_VEXT_AMO(vamomaxud_v_d, uint64_t, uint64_t, idx_d, clearq)
835 #endif
836 GEN_VEXT_AMO(vamoswapw_v_w, int32_t,  int32_t,  idx_w, clearl)
837 GEN_VEXT_AMO(vamoaddw_v_w,  int32_t,  int32_t,  idx_w, clearl)
838 GEN_VEXT_AMO(vamoxorw_v_w,  int32_t,  int32_t,  idx_w, clearl)
839 GEN_VEXT_AMO(vamoandw_v_w,  int32_t,  int32_t,  idx_w, clearl)
840 GEN_VEXT_AMO(vamoorw_v_w,   int32_t,  int32_t,  idx_w, clearl)
841 GEN_VEXT_AMO(vamominw_v_w,  int32_t,  int32_t,  idx_w, clearl)
842 GEN_VEXT_AMO(vamomaxw_v_w,  int32_t,  int32_t,  idx_w, clearl)
843 GEN_VEXT_AMO(vamominuw_v_w, uint32_t, uint32_t, idx_w, clearl)
844 GEN_VEXT_AMO(vamomaxuw_v_w, uint32_t, uint32_t, idx_w, clearl)
845 
846 /*
847  *** Vector Integer Arithmetic Instructions
848  */
849 
850 /* expand macro args before macro */
851 #define RVVCALL(macro, ...)  macro(__VA_ARGS__)
852 
853 /* (TD, T1, T2, TX1, TX2) */
854 #define OP_SSS_B int8_t, int8_t, int8_t, int8_t, int8_t
855 #define OP_SSS_H int16_t, int16_t, int16_t, int16_t, int16_t
856 #define OP_SSS_W int32_t, int32_t, int32_t, int32_t, int32_t
857 #define OP_SSS_D int64_t, int64_t, int64_t, int64_t, int64_t
858 #define OP_UUU_B uint8_t, uint8_t, uint8_t, uint8_t, uint8_t
859 #define OP_UUU_H uint16_t, uint16_t, uint16_t, uint16_t, uint16_t
860 #define OP_UUU_W uint32_t, uint32_t, uint32_t, uint32_t, uint32_t
861 #define OP_UUU_D uint64_t, uint64_t, uint64_t, uint64_t, uint64_t
862 #define OP_SUS_B int8_t, uint8_t, int8_t, uint8_t, int8_t
863 #define OP_SUS_H int16_t, uint16_t, int16_t, uint16_t, int16_t
864 #define OP_SUS_W int32_t, uint32_t, int32_t, uint32_t, int32_t
865 #define OP_SUS_D int64_t, uint64_t, int64_t, uint64_t, int64_t
866 #define WOP_UUU_B uint16_t, uint8_t, uint8_t, uint16_t, uint16_t
867 #define WOP_UUU_H uint32_t, uint16_t, uint16_t, uint32_t, uint32_t
868 #define WOP_UUU_W uint64_t, uint32_t, uint32_t, uint64_t, uint64_t
869 #define WOP_SSS_B int16_t, int8_t, int8_t, int16_t, int16_t
870 #define WOP_SSS_H int32_t, int16_t, int16_t, int32_t, int32_t
871 #define WOP_SSS_W int64_t, int32_t, int32_t, int64_t, int64_t
872 #define WOP_SUS_B int16_t, uint8_t, int8_t, uint16_t, int16_t
873 #define WOP_SUS_H int32_t, uint16_t, int16_t, uint32_t, int32_t
874 #define WOP_SUS_W int64_t, uint32_t, int32_t, uint64_t, int64_t
875 #define WOP_SSU_B int16_t, int8_t, uint8_t, int16_t, uint16_t
876 #define WOP_SSU_H int32_t, int16_t, uint16_t, int32_t, uint32_t
877 #define WOP_SSU_W int64_t, int32_t, uint32_t, int64_t, uint64_t
878 
879 /* operation of two vector elements */
880 typedef void opivv2_fn(void *vd, void *vs1, void *vs2, int i);
881 
882 #define OPIVV2(NAME, TD, T1, T2, TX1, TX2, HD, HS1, HS2, OP)    \
883 static void do_##NAME(void *vd, void *vs1, void *vs2, int i)    \
884 {                                                               \
885     TX1 s1 = *((T1 *)vs1 + HS1(i));                             \
886     TX2 s2 = *((T2 *)vs2 + HS2(i));                             \
887     *((TD *)vd + HD(i)) = OP(s2, s1);                           \
888 }
889 #define DO_SUB(N, M) (N - M)
890 #define DO_RSUB(N, M) (M - N)
891 
892 RVVCALL(OPIVV2, vadd_vv_b, OP_SSS_B, H1, H1, H1, DO_ADD)
893 RVVCALL(OPIVV2, vadd_vv_h, OP_SSS_H, H2, H2, H2, DO_ADD)
894 RVVCALL(OPIVV2, vadd_vv_w, OP_SSS_W, H4, H4, H4, DO_ADD)
895 RVVCALL(OPIVV2, vadd_vv_d, OP_SSS_D, H8, H8, H8, DO_ADD)
896 RVVCALL(OPIVV2, vsub_vv_b, OP_SSS_B, H1, H1, H1, DO_SUB)
897 RVVCALL(OPIVV2, vsub_vv_h, OP_SSS_H, H2, H2, H2, DO_SUB)
898 RVVCALL(OPIVV2, vsub_vv_w, OP_SSS_W, H4, H4, H4, DO_SUB)
899 RVVCALL(OPIVV2, vsub_vv_d, OP_SSS_D, H8, H8, H8, DO_SUB)
900 
901 static void do_vext_vv(void *vd, void *v0, void *vs1, void *vs2,
902                        CPURISCVState *env, uint32_t desc,
903                        uint32_t esz, uint32_t dsz,
904                        opivv2_fn *fn, clear_fn *clearfn)
905 {
906     uint32_t vlmax = vext_maxsz(desc) / esz;
907     uint32_t mlen = vext_mlen(desc);
908     uint32_t vm = vext_vm(desc);
909     uint32_t vl = env->vl;
910     uint32_t i;
911 
912     for (i = 0; i < vl; i++) {
913         if (!vm && !vext_elem_mask(v0, mlen, i)) {
914             continue;
915         }
916         fn(vd, vs1, vs2, i);
917     }
918     clearfn(vd, vl, vl * dsz,  vlmax * dsz);
919 }
920 
921 /* generate the helpers for OPIVV */
922 #define GEN_VEXT_VV(NAME, ESZ, DSZ, CLEAR_FN)             \
923 void HELPER(NAME)(void *vd, void *v0, void *vs1,          \
924                   void *vs2, CPURISCVState *env,          \
925                   uint32_t desc)                          \
926 {                                                         \
927     do_vext_vv(vd, v0, vs1, vs2, env, desc, ESZ, DSZ,     \
928                do_##NAME, CLEAR_FN);                      \
929 }
930 
931 GEN_VEXT_VV(vadd_vv_b, 1, 1, clearb)
932 GEN_VEXT_VV(vadd_vv_h, 2, 2, clearh)
933 GEN_VEXT_VV(vadd_vv_w, 4, 4, clearl)
934 GEN_VEXT_VV(vadd_vv_d, 8, 8, clearq)
935 GEN_VEXT_VV(vsub_vv_b, 1, 1, clearb)
936 GEN_VEXT_VV(vsub_vv_h, 2, 2, clearh)
937 GEN_VEXT_VV(vsub_vv_w, 4, 4, clearl)
938 GEN_VEXT_VV(vsub_vv_d, 8, 8, clearq)
939 
940 typedef void opivx2_fn(void *vd, target_long s1, void *vs2, int i);
941 
942 /*
943  * (T1)s1 gives the real operator type.
944  * (TX1)(T1)s1 expands the operator type of widen or narrow operations.
945  */
946 #define OPIVX2(NAME, TD, T1, T2, TX1, TX2, HD, HS2, OP)             \
947 static void do_##NAME(void *vd, target_long s1, void *vs2, int i)   \
948 {                                                                   \
949     TX2 s2 = *((T2 *)vs2 + HS2(i));                                 \
950     *((TD *)vd + HD(i)) = OP(s2, (TX1)(T1)s1);                      \
951 }
952 
953 RVVCALL(OPIVX2, vadd_vx_b, OP_SSS_B, H1, H1, DO_ADD)
954 RVVCALL(OPIVX2, vadd_vx_h, OP_SSS_H, H2, H2, DO_ADD)
955 RVVCALL(OPIVX2, vadd_vx_w, OP_SSS_W, H4, H4, DO_ADD)
956 RVVCALL(OPIVX2, vadd_vx_d, OP_SSS_D, H8, H8, DO_ADD)
957 RVVCALL(OPIVX2, vsub_vx_b, OP_SSS_B, H1, H1, DO_SUB)
958 RVVCALL(OPIVX2, vsub_vx_h, OP_SSS_H, H2, H2, DO_SUB)
959 RVVCALL(OPIVX2, vsub_vx_w, OP_SSS_W, H4, H4, DO_SUB)
960 RVVCALL(OPIVX2, vsub_vx_d, OP_SSS_D, H8, H8, DO_SUB)
961 RVVCALL(OPIVX2, vrsub_vx_b, OP_SSS_B, H1, H1, DO_RSUB)
962 RVVCALL(OPIVX2, vrsub_vx_h, OP_SSS_H, H2, H2, DO_RSUB)
963 RVVCALL(OPIVX2, vrsub_vx_w, OP_SSS_W, H4, H4, DO_RSUB)
964 RVVCALL(OPIVX2, vrsub_vx_d, OP_SSS_D, H8, H8, DO_RSUB)
965 
966 static void do_vext_vx(void *vd, void *v0, target_long s1, void *vs2,
967                        CPURISCVState *env, uint32_t desc,
968                        uint32_t esz, uint32_t dsz,
969                        opivx2_fn fn, clear_fn *clearfn)
970 {
971     uint32_t vlmax = vext_maxsz(desc) / esz;
972     uint32_t mlen = vext_mlen(desc);
973     uint32_t vm = vext_vm(desc);
974     uint32_t vl = env->vl;
975     uint32_t i;
976 
977     for (i = 0; i < vl; i++) {
978         if (!vm && !vext_elem_mask(v0, mlen, i)) {
979             continue;
980         }
981         fn(vd, s1, vs2, i);
982     }
983     clearfn(vd, vl, vl * dsz,  vlmax * dsz);
984 }
985 
986 /* generate the helpers for OPIVX */
987 #define GEN_VEXT_VX(NAME, ESZ, DSZ, CLEAR_FN)             \
988 void HELPER(NAME)(void *vd, void *v0, target_ulong s1,    \
989                   void *vs2, CPURISCVState *env,          \
990                   uint32_t desc)                          \
991 {                                                         \
992     do_vext_vx(vd, v0, s1, vs2, env, desc, ESZ, DSZ,      \
993                do_##NAME, CLEAR_FN);                      \
994 }
995 
996 GEN_VEXT_VX(vadd_vx_b, 1, 1, clearb)
997 GEN_VEXT_VX(vadd_vx_h, 2, 2, clearh)
998 GEN_VEXT_VX(vadd_vx_w, 4, 4, clearl)
999 GEN_VEXT_VX(vadd_vx_d, 8, 8, clearq)
1000 GEN_VEXT_VX(vsub_vx_b, 1, 1, clearb)
1001 GEN_VEXT_VX(vsub_vx_h, 2, 2, clearh)
1002 GEN_VEXT_VX(vsub_vx_w, 4, 4, clearl)
1003 GEN_VEXT_VX(vsub_vx_d, 8, 8, clearq)
1004 GEN_VEXT_VX(vrsub_vx_b, 1, 1, clearb)
1005 GEN_VEXT_VX(vrsub_vx_h, 2, 2, clearh)
1006 GEN_VEXT_VX(vrsub_vx_w, 4, 4, clearl)
1007 GEN_VEXT_VX(vrsub_vx_d, 8, 8, clearq)
1008 
1009 void HELPER(vec_rsubs8)(void *d, void *a, uint64_t b, uint32_t desc)
1010 {
1011     intptr_t oprsz = simd_oprsz(desc);
1012     intptr_t i;
1013 
1014     for (i = 0; i < oprsz; i += sizeof(uint8_t)) {
1015         *(uint8_t *)(d + i) = (uint8_t)b - *(uint8_t *)(a + i);
1016     }
1017 }
1018 
1019 void HELPER(vec_rsubs16)(void *d, void *a, uint64_t b, uint32_t desc)
1020 {
1021     intptr_t oprsz = simd_oprsz(desc);
1022     intptr_t i;
1023 
1024     for (i = 0; i < oprsz; i += sizeof(uint16_t)) {
1025         *(uint16_t *)(d + i) = (uint16_t)b - *(uint16_t *)(a + i);
1026     }
1027 }
1028 
1029 void HELPER(vec_rsubs32)(void *d, void *a, uint64_t b, uint32_t desc)
1030 {
1031     intptr_t oprsz = simd_oprsz(desc);
1032     intptr_t i;
1033 
1034     for (i = 0; i < oprsz; i += sizeof(uint32_t)) {
1035         *(uint32_t *)(d + i) = (uint32_t)b - *(uint32_t *)(a + i);
1036     }
1037 }
1038 
1039 void HELPER(vec_rsubs64)(void *d, void *a, uint64_t b, uint32_t desc)
1040 {
1041     intptr_t oprsz = simd_oprsz(desc);
1042     intptr_t i;
1043 
1044     for (i = 0; i < oprsz; i += sizeof(uint64_t)) {
1045         *(uint64_t *)(d + i) = b - *(uint64_t *)(a + i);
1046     }
1047 }
1048 
1049 /* Vector Widening Integer Add/Subtract */
1050 #define WOP_UUU_B uint16_t, uint8_t, uint8_t, uint16_t, uint16_t
1051 #define WOP_UUU_H uint32_t, uint16_t, uint16_t, uint32_t, uint32_t
1052 #define WOP_UUU_W uint64_t, uint32_t, uint32_t, uint64_t, uint64_t
1053 #define WOP_SSS_B int16_t, int8_t, int8_t, int16_t, int16_t
1054 #define WOP_SSS_H int32_t, int16_t, int16_t, int32_t, int32_t
1055 #define WOP_SSS_W int64_t, int32_t, int32_t, int64_t, int64_t
1056 #define WOP_WUUU_B  uint16_t, uint8_t, uint16_t, uint16_t, uint16_t
1057 #define WOP_WUUU_H  uint32_t, uint16_t, uint32_t, uint32_t, uint32_t
1058 #define WOP_WUUU_W  uint64_t, uint32_t, uint64_t, uint64_t, uint64_t
1059 #define WOP_WSSS_B  int16_t, int8_t, int16_t, int16_t, int16_t
1060 #define WOP_WSSS_H  int32_t, int16_t, int32_t, int32_t, int32_t
1061 #define WOP_WSSS_W  int64_t, int32_t, int64_t, int64_t, int64_t
1062 RVVCALL(OPIVV2, vwaddu_vv_b, WOP_UUU_B, H2, H1, H1, DO_ADD)
1063 RVVCALL(OPIVV2, vwaddu_vv_h, WOP_UUU_H, H4, H2, H2, DO_ADD)
1064 RVVCALL(OPIVV2, vwaddu_vv_w, WOP_UUU_W, H8, H4, H4, DO_ADD)
1065 RVVCALL(OPIVV2, vwsubu_vv_b, WOP_UUU_B, H2, H1, H1, DO_SUB)
1066 RVVCALL(OPIVV2, vwsubu_vv_h, WOP_UUU_H, H4, H2, H2, DO_SUB)
1067 RVVCALL(OPIVV2, vwsubu_vv_w, WOP_UUU_W, H8, H4, H4, DO_SUB)
1068 RVVCALL(OPIVV2, vwadd_vv_b, WOP_SSS_B, H2, H1, H1, DO_ADD)
1069 RVVCALL(OPIVV2, vwadd_vv_h, WOP_SSS_H, H4, H2, H2, DO_ADD)
1070 RVVCALL(OPIVV2, vwadd_vv_w, WOP_SSS_W, H8, H4, H4, DO_ADD)
1071 RVVCALL(OPIVV2, vwsub_vv_b, WOP_SSS_B, H2, H1, H1, DO_SUB)
1072 RVVCALL(OPIVV2, vwsub_vv_h, WOP_SSS_H, H4, H2, H2, DO_SUB)
1073 RVVCALL(OPIVV2, vwsub_vv_w, WOP_SSS_W, H8, H4, H4, DO_SUB)
1074 RVVCALL(OPIVV2, vwaddu_wv_b, WOP_WUUU_B, H2, H1, H1, DO_ADD)
1075 RVVCALL(OPIVV2, vwaddu_wv_h, WOP_WUUU_H, H4, H2, H2, DO_ADD)
1076 RVVCALL(OPIVV2, vwaddu_wv_w, WOP_WUUU_W, H8, H4, H4, DO_ADD)
1077 RVVCALL(OPIVV2, vwsubu_wv_b, WOP_WUUU_B, H2, H1, H1, DO_SUB)
1078 RVVCALL(OPIVV2, vwsubu_wv_h, WOP_WUUU_H, H4, H2, H2, DO_SUB)
1079 RVVCALL(OPIVV2, vwsubu_wv_w, WOP_WUUU_W, H8, H4, H4, DO_SUB)
1080 RVVCALL(OPIVV2, vwadd_wv_b, WOP_WSSS_B, H2, H1, H1, DO_ADD)
1081 RVVCALL(OPIVV2, vwadd_wv_h, WOP_WSSS_H, H4, H2, H2, DO_ADD)
1082 RVVCALL(OPIVV2, vwadd_wv_w, WOP_WSSS_W, H8, H4, H4, DO_ADD)
1083 RVVCALL(OPIVV2, vwsub_wv_b, WOP_WSSS_B, H2, H1, H1, DO_SUB)
1084 RVVCALL(OPIVV2, vwsub_wv_h, WOP_WSSS_H, H4, H2, H2, DO_SUB)
1085 RVVCALL(OPIVV2, vwsub_wv_w, WOP_WSSS_W, H8, H4, H4, DO_SUB)
1086 GEN_VEXT_VV(vwaddu_vv_b, 1, 2, clearh)
1087 GEN_VEXT_VV(vwaddu_vv_h, 2, 4, clearl)
1088 GEN_VEXT_VV(vwaddu_vv_w, 4, 8, clearq)
1089 GEN_VEXT_VV(vwsubu_vv_b, 1, 2, clearh)
1090 GEN_VEXT_VV(vwsubu_vv_h, 2, 4, clearl)
1091 GEN_VEXT_VV(vwsubu_vv_w, 4, 8, clearq)
1092 GEN_VEXT_VV(vwadd_vv_b, 1, 2, clearh)
1093 GEN_VEXT_VV(vwadd_vv_h, 2, 4, clearl)
1094 GEN_VEXT_VV(vwadd_vv_w, 4, 8, clearq)
1095 GEN_VEXT_VV(vwsub_vv_b, 1, 2, clearh)
1096 GEN_VEXT_VV(vwsub_vv_h, 2, 4, clearl)
1097 GEN_VEXT_VV(vwsub_vv_w, 4, 8, clearq)
1098 GEN_VEXT_VV(vwaddu_wv_b, 1, 2, clearh)
1099 GEN_VEXT_VV(vwaddu_wv_h, 2, 4, clearl)
1100 GEN_VEXT_VV(vwaddu_wv_w, 4, 8, clearq)
1101 GEN_VEXT_VV(vwsubu_wv_b, 1, 2, clearh)
1102 GEN_VEXT_VV(vwsubu_wv_h, 2, 4, clearl)
1103 GEN_VEXT_VV(vwsubu_wv_w, 4, 8, clearq)
1104 GEN_VEXT_VV(vwadd_wv_b, 1, 2, clearh)
1105 GEN_VEXT_VV(vwadd_wv_h, 2, 4, clearl)
1106 GEN_VEXT_VV(vwadd_wv_w, 4, 8, clearq)
1107 GEN_VEXT_VV(vwsub_wv_b, 1, 2, clearh)
1108 GEN_VEXT_VV(vwsub_wv_h, 2, 4, clearl)
1109 GEN_VEXT_VV(vwsub_wv_w, 4, 8, clearq)
1110 
1111 RVVCALL(OPIVX2, vwaddu_vx_b, WOP_UUU_B, H2, H1, DO_ADD)
1112 RVVCALL(OPIVX2, vwaddu_vx_h, WOP_UUU_H, H4, H2, DO_ADD)
1113 RVVCALL(OPIVX2, vwaddu_vx_w, WOP_UUU_W, H8, H4, DO_ADD)
1114 RVVCALL(OPIVX2, vwsubu_vx_b, WOP_UUU_B, H2, H1, DO_SUB)
1115 RVVCALL(OPIVX2, vwsubu_vx_h, WOP_UUU_H, H4, H2, DO_SUB)
1116 RVVCALL(OPIVX2, vwsubu_vx_w, WOP_UUU_W, H8, H4, DO_SUB)
1117 RVVCALL(OPIVX2, vwadd_vx_b, WOP_SSS_B, H2, H1, DO_ADD)
1118 RVVCALL(OPIVX2, vwadd_vx_h, WOP_SSS_H, H4, H2, DO_ADD)
1119 RVVCALL(OPIVX2, vwadd_vx_w, WOP_SSS_W, H8, H4, DO_ADD)
1120 RVVCALL(OPIVX2, vwsub_vx_b, WOP_SSS_B, H2, H1, DO_SUB)
1121 RVVCALL(OPIVX2, vwsub_vx_h, WOP_SSS_H, H4, H2, DO_SUB)
1122 RVVCALL(OPIVX2, vwsub_vx_w, WOP_SSS_W, H8, H4, DO_SUB)
1123 RVVCALL(OPIVX2, vwaddu_wx_b, WOP_WUUU_B, H2, H1, DO_ADD)
1124 RVVCALL(OPIVX2, vwaddu_wx_h, WOP_WUUU_H, H4, H2, DO_ADD)
1125 RVVCALL(OPIVX2, vwaddu_wx_w, WOP_WUUU_W, H8, H4, DO_ADD)
1126 RVVCALL(OPIVX2, vwsubu_wx_b, WOP_WUUU_B, H2, H1, DO_SUB)
1127 RVVCALL(OPIVX2, vwsubu_wx_h, WOP_WUUU_H, H4, H2, DO_SUB)
1128 RVVCALL(OPIVX2, vwsubu_wx_w, WOP_WUUU_W, H8, H4, DO_SUB)
1129 RVVCALL(OPIVX2, vwadd_wx_b, WOP_WSSS_B, H2, H1, DO_ADD)
1130 RVVCALL(OPIVX2, vwadd_wx_h, WOP_WSSS_H, H4, H2, DO_ADD)
1131 RVVCALL(OPIVX2, vwadd_wx_w, WOP_WSSS_W, H8, H4, DO_ADD)
1132 RVVCALL(OPIVX2, vwsub_wx_b, WOP_WSSS_B, H2, H1, DO_SUB)
1133 RVVCALL(OPIVX2, vwsub_wx_h, WOP_WSSS_H, H4, H2, DO_SUB)
1134 RVVCALL(OPIVX2, vwsub_wx_w, WOP_WSSS_W, H8, H4, DO_SUB)
1135 GEN_VEXT_VX(vwaddu_vx_b, 1, 2, clearh)
1136 GEN_VEXT_VX(vwaddu_vx_h, 2, 4, clearl)
1137 GEN_VEXT_VX(vwaddu_vx_w, 4, 8, clearq)
1138 GEN_VEXT_VX(vwsubu_vx_b, 1, 2, clearh)
1139 GEN_VEXT_VX(vwsubu_vx_h, 2, 4, clearl)
1140 GEN_VEXT_VX(vwsubu_vx_w, 4, 8, clearq)
1141 GEN_VEXT_VX(vwadd_vx_b, 1, 2, clearh)
1142 GEN_VEXT_VX(vwadd_vx_h, 2, 4, clearl)
1143 GEN_VEXT_VX(vwadd_vx_w, 4, 8, clearq)
1144 GEN_VEXT_VX(vwsub_vx_b, 1, 2, clearh)
1145 GEN_VEXT_VX(vwsub_vx_h, 2, 4, clearl)
1146 GEN_VEXT_VX(vwsub_vx_w, 4, 8, clearq)
1147 GEN_VEXT_VX(vwaddu_wx_b, 1, 2, clearh)
1148 GEN_VEXT_VX(vwaddu_wx_h, 2, 4, clearl)
1149 GEN_VEXT_VX(vwaddu_wx_w, 4, 8, clearq)
1150 GEN_VEXT_VX(vwsubu_wx_b, 1, 2, clearh)
1151 GEN_VEXT_VX(vwsubu_wx_h, 2, 4, clearl)
1152 GEN_VEXT_VX(vwsubu_wx_w, 4, 8, clearq)
1153 GEN_VEXT_VX(vwadd_wx_b, 1, 2, clearh)
1154 GEN_VEXT_VX(vwadd_wx_h, 2, 4, clearl)
1155 GEN_VEXT_VX(vwadd_wx_w, 4, 8, clearq)
1156 GEN_VEXT_VX(vwsub_wx_b, 1, 2, clearh)
1157 GEN_VEXT_VX(vwsub_wx_h, 2, 4, clearl)
1158 GEN_VEXT_VX(vwsub_wx_w, 4, 8, clearq)
1159 
1160 /* Vector Integer Add-with-Carry / Subtract-with-Borrow Instructions */
1161 #define DO_VADC(N, M, C) (N + M + C)
1162 #define DO_VSBC(N, M, C) (N - M - C)
1163 
1164 #define GEN_VEXT_VADC_VVM(NAME, ETYPE, H, DO_OP, CLEAR_FN)    \
1165 void HELPER(NAME)(void *vd, void *v0, void *vs1, void *vs2,   \
1166                   CPURISCVState *env, uint32_t desc)          \
1167 {                                                             \
1168     uint32_t mlen = vext_mlen(desc);                          \
1169     uint32_t vl = env->vl;                                    \
1170     uint32_t esz = sizeof(ETYPE);                             \
1171     uint32_t vlmax = vext_maxsz(desc) / esz;                  \
1172     uint32_t i;                                               \
1173                                                               \
1174     for (i = 0; i < vl; i++) {                                \
1175         ETYPE s1 = *((ETYPE *)vs1 + H(i));                    \
1176         ETYPE s2 = *((ETYPE *)vs2 + H(i));                    \
1177         uint8_t carry = vext_elem_mask(v0, mlen, i);          \
1178                                                               \
1179         *((ETYPE *)vd + H(i)) = DO_OP(s2, s1, carry);         \
1180     }                                                         \
1181     CLEAR_FN(vd, vl, vl * esz, vlmax * esz);                  \
1182 }
1183 
1184 GEN_VEXT_VADC_VVM(vadc_vvm_b, uint8_t,  H1, DO_VADC, clearb)
1185 GEN_VEXT_VADC_VVM(vadc_vvm_h, uint16_t, H2, DO_VADC, clearh)
1186 GEN_VEXT_VADC_VVM(vadc_vvm_w, uint32_t, H4, DO_VADC, clearl)
1187 GEN_VEXT_VADC_VVM(vadc_vvm_d, uint64_t, H8, DO_VADC, clearq)
1188 
1189 GEN_VEXT_VADC_VVM(vsbc_vvm_b, uint8_t,  H1, DO_VSBC, clearb)
1190 GEN_VEXT_VADC_VVM(vsbc_vvm_h, uint16_t, H2, DO_VSBC, clearh)
1191 GEN_VEXT_VADC_VVM(vsbc_vvm_w, uint32_t, H4, DO_VSBC, clearl)
1192 GEN_VEXT_VADC_VVM(vsbc_vvm_d, uint64_t, H8, DO_VSBC, clearq)
1193 
1194 #define GEN_VEXT_VADC_VXM(NAME, ETYPE, H, DO_OP, CLEAR_FN)               \
1195 void HELPER(NAME)(void *vd, void *v0, target_ulong s1, void *vs2,        \
1196                   CPURISCVState *env, uint32_t desc)                     \
1197 {                                                                        \
1198     uint32_t mlen = vext_mlen(desc);                                     \
1199     uint32_t vl = env->vl;                                               \
1200     uint32_t esz = sizeof(ETYPE);                                        \
1201     uint32_t vlmax = vext_maxsz(desc) / esz;                             \
1202     uint32_t i;                                                          \
1203                                                                          \
1204     for (i = 0; i < vl; i++) {                                           \
1205         ETYPE s2 = *((ETYPE *)vs2 + H(i));                               \
1206         uint8_t carry = vext_elem_mask(v0, mlen, i);                     \
1207                                                                          \
1208         *((ETYPE *)vd + H(i)) = DO_OP(s2, (ETYPE)(target_long)s1, carry);\
1209     }                                                                    \
1210     CLEAR_FN(vd, vl, vl * esz, vlmax * esz);                             \
1211 }
1212 
1213 GEN_VEXT_VADC_VXM(vadc_vxm_b, uint8_t,  H1, DO_VADC, clearb)
1214 GEN_VEXT_VADC_VXM(vadc_vxm_h, uint16_t, H2, DO_VADC, clearh)
1215 GEN_VEXT_VADC_VXM(vadc_vxm_w, uint32_t, H4, DO_VADC, clearl)
1216 GEN_VEXT_VADC_VXM(vadc_vxm_d, uint64_t, H8, DO_VADC, clearq)
1217 
1218 GEN_VEXT_VADC_VXM(vsbc_vxm_b, uint8_t,  H1, DO_VSBC, clearb)
1219 GEN_VEXT_VADC_VXM(vsbc_vxm_h, uint16_t, H2, DO_VSBC, clearh)
1220 GEN_VEXT_VADC_VXM(vsbc_vxm_w, uint32_t, H4, DO_VSBC, clearl)
1221 GEN_VEXT_VADC_VXM(vsbc_vxm_d, uint64_t, H8, DO_VSBC, clearq)
1222 
1223 #define DO_MADC(N, M, C) (C ? (__typeof(N))(N + M + 1) <= N :           \
1224                           (__typeof(N))(N + M) < N)
1225 #define DO_MSBC(N, M, C) (C ? N <= M : N < M)
1226 
1227 #define GEN_VEXT_VMADC_VVM(NAME, ETYPE, H, DO_OP)             \
1228 void HELPER(NAME)(void *vd, void *v0, void *vs1, void *vs2,   \
1229                   CPURISCVState *env, uint32_t desc)          \
1230 {                                                             \
1231     uint32_t mlen = vext_mlen(desc);                          \
1232     uint32_t vl = env->vl;                                    \
1233     uint32_t vlmax = vext_maxsz(desc) / sizeof(ETYPE);        \
1234     uint32_t i;                                               \
1235                                                               \
1236     for (i = 0; i < vl; i++) {                                \
1237         ETYPE s1 = *((ETYPE *)vs1 + H(i));                    \
1238         ETYPE s2 = *((ETYPE *)vs2 + H(i));                    \
1239         uint8_t carry = vext_elem_mask(v0, mlen, i);          \
1240                                                               \
1241         vext_set_elem_mask(vd, mlen, i, DO_OP(s2, s1, carry));\
1242     }                                                         \
1243     for (; i < vlmax; i++) {                                  \
1244         vext_set_elem_mask(vd, mlen, i, 0);                   \
1245     }                                                         \
1246 }
1247 
1248 GEN_VEXT_VMADC_VVM(vmadc_vvm_b, uint8_t,  H1, DO_MADC)
1249 GEN_VEXT_VMADC_VVM(vmadc_vvm_h, uint16_t, H2, DO_MADC)
1250 GEN_VEXT_VMADC_VVM(vmadc_vvm_w, uint32_t, H4, DO_MADC)
1251 GEN_VEXT_VMADC_VVM(vmadc_vvm_d, uint64_t, H8, DO_MADC)
1252 
1253 GEN_VEXT_VMADC_VVM(vmsbc_vvm_b, uint8_t,  H1, DO_MSBC)
1254 GEN_VEXT_VMADC_VVM(vmsbc_vvm_h, uint16_t, H2, DO_MSBC)
1255 GEN_VEXT_VMADC_VVM(vmsbc_vvm_w, uint32_t, H4, DO_MSBC)
1256 GEN_VEXT_VMADC_VVM(vmsbc_vvm_d, uint64_t, H8, DO_MSBC)
1257 
1258 #define GEN_VEXT_VMADC_VXM(NAME, ETYPE, H, DO_OP)               \
1259 void HELPER(NAME)(void *vd, void *v0, target_ulong s1,          \
1260                   void *vs2, CPURISCVState *env, uint32_t desc) \
1261 {                                                               \
1262     uint32_t mlen = vext_mlen(desc);                            \
1263     uint32_t vl = env->vl;                                      \
1264     uint32_t vlmax = vext_maxsz(desc) / sizeof(ETYPE);          \
1265     uint32_t i;                                                 \
1266                                                                 \
1267     for (i = 0; i < vl; i++) {                                  \
1268         ETYPE s2 = *((ETYPE *)vs2 + H(i));                      \
1269         uint8_t carry = vext_elem_mask(v0, mlen, i);            \
1270                                                                 \
1271         vext_set_elem_mask(vd, mlen, i,                         \
1272                 DO_OP(s2, (ETYPE)(target_long)s1, carry));      \
1273     }                                                           \
1274     for (; i < vlmax; i++) {                                    \
1275         vext_set_elem_mask(vd, mlen, i, 0);                     \
1276     }                                                           \
1277 }
1278 
1279 GEN_VEXT_VMADC_VXM(vmadc_vxm_b, uint8_t,  H1, DO_MADC)
1280 GEN_VEXT_VMADC_VXM(vmadc_vxm_h, uint16_t, H2, DO_MADC)
1281 GEN_VEXT_VMADC_VXM(vmadc_vxm_w, uint32_t, H4, DO_MADC)
1282 GEN_VEXT_VMADC_VXM(vmadc_vxm_d, uint64_t, H8, DO_MADC)
1283 
1284 GEN_VEXT_VMADC_VXM(vmsbc_vxm_b, uint8_t,  H1, DO_MSBC)
1285 GEN_VEXT_VMADC_VXM(vmsbc_vxm_h, uint16_t, H2, DO_MSBC)
1286 GEN_VEXT_VMADC_VXM(vmsbc_vxm_w, uint32_t, H4, DO_MSBC)
1287 GEN_VEXT_VMADC_VXM(vmsbc_vxm_d, uint64_t, H8, DO_MSBC)
1288 
1289 /* Vector Bitwise Logical Instructions */
1290 RVVCALL(OPIVV2, vand_vv_b, OP_SSS_B, H1, H1, H1, DO_AND)
1291 RVVCALL(OPIVV2, vand_vv_h, OP_SSS_H, H2, H2, H2, DO_AND)
1292 RVVCALL(OPIVV2, vand_vv_w, OP_SSS_W, H4, H4, H4, DO_AND)
1293 RVVCALL(OPIVV2, vand_vv_d, OP_SSS_D, H8, H8, H8, DO_AND)
1294 RVVCALL(OPIVV2, vor_vv_b, OP_SSS_B, H1, H1, H1, DO_OR)
1295 RVVCALL(OPIVV2, vor_vv_h, OP_SSS_H, H2, H2, H2, DO_OR)
1296 RVVCALL(OPIVV2, vor_vv_w, OP_SSS_W, H4, H4, H4, DO_OR)
1297 RVVCALL(OPIVV2, vor_vv_d, OP_SSS_D, H8, H8, H8, DO_OR)
1298 RVVCALL(OPIVV2, vxor_vv_b, OP_SSS_B, H1, H1, H1, DO_XOR)
1299 RVVCALL(OPIVV2, vxor_vv_h, OP_SSS_H, H2, H2, H2, DO_XOR)
1300 RVVCALL(OPIVV2, vxor_vv_w, OP_SSS_W, H4, H4, H4, DO_XOR)
1301 RVVCALL(OPIVV2, vxor_vv_d, OP_SSS_D, H8, H8, H8, DO_XOR)
1302 GEN_VEXT_VV(vand_vv_b, 1, 1, clearb)
1303 GEN_VEXT_VV(vand_vv_h, 2, 2, clearh)
1304 GEN_VEXT_VV(vand_vv_w, 4, 4, clearl)
1305 GEN_VEXT_VV(vand_vv_d, 8, 8, clearq)
1306 GEN_VEXT_VV(vor_vv_b, 1, 1, clearb)
1307 GEN_VEXT_VV(vor_vv_h, 2, 2, clearh)
1308 GEN_VEXT_VV(vor_vv_w, 4, 4, clearl)
1309 GEN_VEXT_VV(vor_vv_d, 8, 8, clearq)
1310 GEN_VEXT_VV(vxor_vv_b, 1, 1, clearb)
1311 GEN_VEXT_VV(vxor_vv_h, 2, 2, clearh)
1312 GEN_VEXT_VV(vxor_vv_w, 4, 4, clearl)
1313 GEN_VEXT_VV(vxor_vv_d, 8, 8, clearq)
1314 
1315 RVVCALL(OPIVX2, vand_vx_b, OP_SSS_B, H1, H1, DO_AND)
1316 RVVCALL(OPIVX2, vand_vx_h, OP_SSS_H, H2, H2, DO_AND)
1317 RVVCALL(OPIVX2, vand_vx_w, OP_SSS_W, H4, H4, DO_AND)
1318 RVVCALL(OPIVX2, vand_vx_d, OP_SSS_D, H8, H8, DO_AND)
1319 RVVCALL(OPIVX2, vor_vx_b, OP_SSS_B, H1, H1, DO_OR)
1320 RVVCALL(OPIVX2, vor_vx_h, OP_SSS_H, H2, H2, DO_OR)
1321 RVVCALL(OPIVX2, vor_vx_w, OP_SSS_W, H4, H4, DO_OR)
1322 RVVCALL(OPIVX2, vor_vx_d, OP_SSS_D, H8, H8, DO_OR)
1323 RVVCALL(OPIVX2, vxor_vx_b, OP_SSS_B, H1, H1, DO_XOR)
1324 RVVCALL(OPIVX2, vxor_vx_h, OP_SSS_H, H2, H2, DO_XOR)
1325 RVVCALL(OPIVX2, vxor_vx_w, OP_SSS_W, H4, H4, DO_XOR)
1326 RVVCALL(OPIVX2, vxor_vx_d, OP_SSS_D, H8, H8, DO_XOR)
1327 GEN_VEXT_VX(vand_vx_b, 1, 1, clearb)
1328 GEN_VEXT_VX(vand_vx_h, 2, 2, clearh)
1329 GEN_VEXT_VX(vand_vx_w, 4, 4, clearl)
1330 GEN_VEXT_VX(vand_vx_d, 8, 8, clearq)
1331 GEN_VEXT_VX(vor_vx_b, 1, 1, clearb)
1332 GEN_VEXT_VX(vor_vx_h, 2, 2, clearh)
1333 GEN_VEXT_VX(vor_vx_w, 4, 4, clearl)
1334 GEN_VEXT_VX(vor_vx_d, 8, 8, clearq)
1335 GEN_VEXT_VX(vxor_vx_b, 1, 1, clearb)
1336 GEN_VEXT_VX(vxor_vx_h, 2, 2, clearh)
1337 GEN_VEXT_VX(vxor_vx_w, 4, 4, clearl)
1338 GEN_VEXT_VX(vxor_vx_d, 8, 8, clearq)
1339 
1340 /* Vector Single-Width Bit Shift Instructions */
1341 #define DO_SLL(N, M)  (N << (M))
1342 #define DO_SRL(N, M)  (N >> (M))
1343 
1344 /* generate the helpers for shift instructions with two vector operators */
1345 #define GEN_VEXT_SHIFT_VV(NAME, TS1, TS2, HS1, HS2, OP, MASK, CLEAR_FN)   \
1346 void HELPER(NAME)(void *vd, void *v0, void *vs1,                          \
1347                   void *vs2, CPURISCVState *env, uint32_t desc)           \
1348 {                                                                         \
1349     uint32_t mlen = vext_mlen(desc);                                      \
1350     uint32_t vm = vext_vm(desc);                                          \
1351     uint32_t vl = env->vl;                                                \
1352     uint32_t esz = sizeof(TS1);                                           \
1353     uint32_t vlmax = vext_maxsz(desc) / esz;                              \
1354     uint32_t i;                                                           \
1355                                                                           \
1356     for (i = 0; i < vl; i++) {                                            \
1357         if (!vm && !vext_elem_mask(v0, mlen, i)) {                        \
1358             continue;                                                     \
1359         }                                                                 \
1360         TS1 s1 = *((TS1 *)vs1 + HS1(i));                                  \
1361         TS2 s2 = *((TS2 *)vs2 + HS2(i));                                  \
1362         *((TS1 *)vd + HS1(i)) = OP(s2, s1 & MASK);                        \
1363     }                                                                     \
1364     CLEAR_FN(vd, vl, vl * esz, vlmax * esz);                              \
1365 }
1366 
1367 GEN_VEXT_SHIFT_VV(vsll_vv_b, uint8_t,  uint8_t, H1, H1, DO_SLL, 0x7, clearb)
1368 GEN_VEXT_SHIFT_VV(vsll_vv_h, uint16_t, uint16_t, H2, H2, DO_SLL, 0xf, clearh)
1369 GEN_VEXT_SHIFT_VV(vsll_vv_w, uint32_t, uint32_t, H4, H4, DO_SLL, 0x1f, clearl)
1370 GEN_VEXT_SHIFT_VV(vsll_vv_d, uint64_t, uint64_t, H8, H8, DO_SLL, 0x3f, clearq)
1371 
1372 GEN_VEXT_SHIFT_VV(vsrl_vv_b, uint8_t, uint8_t, H1, H1, DO_SRL, 0x7, clearb)
1373 GEN_VEXT_SHIFT_VV(vsrl_vv_h, uint16_t, uint16_t, H2, H2, DO_SRL, 0xf, clearh)
1374 GEN_VEXT_SHIFT_VV(vsrl_vv_w, uint32_t, uint32_t, H4, H4, DO_SRL, 0x1f, clearl)
1375 GEN_VEXT_SHIFT_VV(vsrl_vv_d, uint64_t, uint64_t, H8, H8, DO_SRL, 0x3f, clearq)
1376 
1377 GEN_VEXT_SHIFT_VV(vsra_vv_b, uint8_t,  int8_t, H1, H1, DO_SRL, 0x7, clearb)
1378 GEN_VEXT_SHIFT_VV(vsra_vv_h, uint16_t, int16_t, H2, H2, DO_SRL, 0xf, clearh)
1379 GEN_VEXT_SHIFT_VV(vsra_vv_w, uint32_t, int32_t, H4, H4, DO_SRL, 0x1f, clearl)
1380 GEN_VEXT_SHIFT_VV(vsra_vv_d, uint64_t, int64_t, H8, H8, DO_SRL, 0x3f, clearq)
1381 
1382 /* generate the helpers for shift instructions with one vector and one scalar */
1383 #define GEN_VEXT_SHIFT_VX(NAME, TD, TS2, HD, HS2, OP, MASK, CLEAR_FN) \
1384 void HELPER(NAME)(void *vd, void *v0, target_ulong s1,                \
1385         void *vs2, CPURISCVState *env, uint32_t desc)                 \
1386 {                                                                     \
1387     uint32_t mlen = vext_mlen(desc);                                  \
1388     uint32_t vm = vext_vm(desc);                                      \
1389     uint32_t vl = env->vl;                                            \
1390     uint32_t esz = sizeof(TD);                                        \
1391     uint32_t vlmax = vext_maxsz(desc) / esz;                          \
1392     uint32_t i;                                                       \
1393                                                                       \
1394     for (i = 0; i < vl; i++) {                                        \
1395         if (!vm && !vext_elem_mask(v0, mlen, i)) {                    \
1396             continue;                                                 \
1397         }                                                             \
1398         TS2 s2 = *((TS2 *)vs2 + HS2(i));                              \
1399         *((TD *)vd + HD(i)) = OP(s2, s1 & MASK);                      \
1400     }                                                                 \
1401     CLEAR_FN(vd, vl, vl * esz, vlmax * esz);                          \
1402 }
1403 
1404 GEN_VEXT_SHIFT_VX(vsll_vx_b, uint8_t, int8_t, H1, H1, DO_SLL, 0x7, clearb)
1405 GEN_VEXT_SHIFT_VX(vsll_vx_h, uint16_t, int16_t, H2, H2, DO_SLL, 0xf, clearh)
1406 GEN_VEXT_SHIFT_VX(vsll_vx_w, uint32_t, int32_t, H4, H4, DO_SLL, 0x1f, clearl)
1407 GEN_VEXT_SHIFT_VX(vsll_vx_d, uint64_t, int64_t, H8, H8, DO_SLL, 0x3f, clearq)
1408 
1409 GEN_VEXT_SHIFT_VX(vsrl_vx_b, uint8_t, uint8_t, H1, H1, DO_SRL, 0x7, clearb)
1410 GEN_VEXT_SHIFT_VX(vsrl_vx_h, uint16_t, uint16_t, H2, H2, DO_SRL, 0xf, clearh)
1411 GEN_VEXT_SHIFT_VX(vsrl_vx_w, uint32_t, uint32_t, H4, H4, DO_SRL, 0x1f, clearl)
1412 GEN_VEXT_SHIFT_VX(vsrl_vx_d, uint64_t, uint64_t, H8, H8, DO_SRL, 0x3f, clearq)
1413 
1414 GEN_VEXT_SHIFT_VX(vsra_vx_b, int8_t, int8_t, H1, H1, DO_SRL, 0x7, clearb)
1415 GEN_VEXT_SHIFT_VX(vsra_vx_h, int16_t, int16_t, H2, H2, DO_SRL, 0xf, clearh)
1416 GEN_VEXT_SHIFT_VX(vsra_vx_w, int32_t, int32_t, H4, H4, DO_SRL, 0x1f, clearl)
1417 GEN_VEXT_SHIFT_VX(vsra_vx_d, int64_t, int64_t, H8, H8, DO_SRL, 0x3f, clearq)
1418 
1419 /* Vector Narrowing Integer Right Shift Instructions */
1420 GEN_VEXT_SHIFT_VV(vnsrl_vv_b, uint8_t,  uint16_t, H1, H2, DO_SRL, 0xf, clearb)
1421 GEN_VEXT_SHIFT_VV(vnsrl_vv_h, uint16_t, uint32_t, H2, H4, DO_SRL, 0x1f, clearh)
1422 GEN_VEXT_SHIFT_VV(vnsrl_vv_w, uint32_t, uint64_t, H4, H8, DO_SRL, 0x3f, clearl)
1423 GEN_VEXT_SHIFT_VV(vnsra_vv_b, uint8_t,  int16_t, H1, H2, DO_SRL, 0xf, clearb)
1424 GEN_VEXT_SHIFT_VV(vnsra_vv_h, uint16_t, int32_t, H2, H4, DO_SRL, 0x1f, clearh)
1425 GEN_VEXT_SHIFT_VV(vnsra_vv_w, uint32_t, int64_t, H4, H8, DO_SRL, 0x3f, clearl)
1426 GEN_VEXT_SHIFT_VX(vnsrl_vx_b, uint8_t, uint16_t, H1, H2, DO_SRL, 0xf, clearb)
1427 GEN_VEXT_SHIFT_VX(vnsrl_vx_h, uint16_t, uint32_t, H2, H4, DO_SRL, 0x1f, clearh)
1428 GEN_VEXT_SHIFT_VX(vnsrl_vx_w, uint32_t, uint64_t, H4, H8, DO_SRL, 0x3f, clearl)
1429 GEN_VEXT_SHIFT_VX(vnsra_vx_b, int8_t, int16_t, H1, H2, DO_SRL, 0xf, clearb)
1430 GEN_VEXT_SHIFT_VX(vnsra_vx_h, int16_t, int32_t, H2, H4, DO_SRL, 0x1f, clearh)
1431 GEN_VEXT_SHIFT_VX(vnsra_vx_w, int32_t, int64_t, H4, H8, DO_SRL, 0x3f, clearl)
1432 
1433 /* Vector Integer Comparison Instructions */
1434 #define DO_MSEQ(N, M) (N == M)
1435 #define DO_MSNE(N, M) (N != M)
1436 #define DO_MSLT(N, M) (N < M)
1437 #define DO_MSLE(N, M) (N <= M)
1438 #define DO_MSGT(N, M) (N > M)
1439 
1440 #define GEN_VEXT_CMP_VV(NAME, ETYPE, H, DO_OP)                \
1441 void HELPER(NAME)(void *vd, void *v0, void *vs1, void *vs2,   \
1442                   CPURISCVState *env, uint32_t desc)          \
1443 {                                                             \
1444     uint32_t mlen = vext_mlen(desc);                          \
1445     uint32_t vm = vext_vm(desc);                              \
1446     uint32_t vl = env->vl;                                    \
1447     uint32_t vlmax = vext_maxsz(desc) / sizeof(ETYPE);        \
1448     uint32_t i;                                               \
1449                                                               \
1450     for (i = 0; i < vl; i++) {                                \
1451         ETYPE s1 = *((ETYPE *)vs1 + H(i));                    \
1452         ETYPE s2 = *((ETYPE *)vs2 + H(i));                    \
1453         if (!vm && !vext_elem_mask(v0, mlen, i)) {            \
1454             continue;                                         \
1455         }                                                     \
1456         vext_set_elem_mask(vd, mlen, i, DO_OP(s2, s1));       \
1457     }                                                         \
1458     for (; i < vlmax; i++) {                                  \
1459         vext_set_elem_mask(vd, mlen, i, 0);                   \
1460     }                                                         \
1461 }
1462 
1463 GEN_VEXT_CMP_VV(vmseq_vv_b, uint8_t,  H1, DO_MSEQ)
1464 GEN_VEXT_CMP_VV(vmseq_vv_h, uint16_t, H2, DO_MSEQ)
1465 GEN_VEXT_CMP_VV(vmseq_vv_w, uint32_t, H4, DO_MSEQ)
1466 GEN_VEXT_CMP_VV(vmseq_vv_d, uint64_t, H8, DO_MSEQ)
1467 
1468 GEN_VEXT_CMP_VV(vmsne_vv_b, uint8_t,  H1, DO_MSNE)
1469 GEN_VEXT_CMP_VV(vmsne_vv_h, uint16_t, H2, DO_MSNE)
1470 GEN_VEXT_CMP_VV(vmsne_vv_w, uint32_t, H4, DO_MSNE)
1471 GEN_VEXT_CMP_VV(vmsne_vv_d, uint64_t, H8, DO_MSNE)
1472 
1473 GEN_VEXT_CMP_VV(vmsltu_vv_b, uint8_t,  H1, DO_MSLT)
1474 GEN_VEXT_CMP_VV(vmsltu_vv_h, uint16_t, H2, DO_MSLT)
1475 GEN_VEXT_CMP_VV(vmsltu_vv_w, uint32_t, H4, DO_MSLT)
1476 GEN_VEXT_CMP_VV(vmsltu_vv_d, uint64_t, H8, DO_MSLT)
1477 
1478 GEN_VEXT_CMP_VV(vmslt_vv_b, int8_t,  H1, DO_MSLT)
1479 GEN_VEXT_CMP_VV(vmslt_vv_h, int16_t, H2, DO_MSLT)
1480 GEN_VEXT_CMP_VV(vmslt_vv_w, int32_t, H4, DO_MSLT)
1481 GEN_VEXT_CMP_VV(vmslt_vv_d, int64_t, H8, DO_MSLT)
1482 
1483 GEN_VEXT_CMP_VV(vmsleu_vv_b, uint8_t,  H1, DO_MSLE)
1484 GEN_VEXT_CMP_VV(vmsleu_vv_h, uint16_t, H2, DO_MSLE)
1485 GEN_VEXT_CMP_VV(vmsleu_vv_w, uint32_t, H4, DO_MSLE)
1486 GEN_VEXT_CMP_VV(vmsleu_vv_d, uint64_t, H8, DO_MSLE)
1487 
1488 GEN_VEXT_CMP_VV(vmsle_vv_b, int8_t,  H1, DO_MSLE)
1489 GEN_VEXT_CMP_VV(vmsle_vv_h, int16_t, H2, DO_MSLE)
1490 GEN_VEXT_CMP_VV(vmsle_vv_w, int32_t, H4, DO_MSLE)
1491 GEN_VEXT_CMP_VV(vmsle_vv_d, int64_t, H8, DO_MSLE)
1492 
1493 #define GEN_VEXT_CMP_VX(NAME, ETYPE, H, DO_OP)                      \
1494 void HELPER(NAME)(void *vd, void *v0, target_ulong s1, void *vs2,   \
1495                   CPURISCVState *env, uint32_t desc)                \
1496 {                                                                   \
1497     uint32_t mlen = vext_mlen(desc);                                \
1498     uint32_t vm = vext_vm(desc);                                    \
1499     uint32_t vl = env->vl;                                          \
1500     uint32_t vlmax = vext_maxsz(desc) / sizeof(ETYPE);              \
1501     uint32_t i;                                                     \
1502                                                                     \
1503     for (i = 0; i < vl; i++) {                                      \
1504         ETYPE s2 = *((ETYPE *)vs2 + H(i));                          \
1505         if (!vm && !vext_elem_mask(v0, mlen, i)) {                  \
1506             continue;                                               \
1507         }                                                           \
1508         vext_set_elem_mask(vd, mlen, i,                             \
1509                 DO_OP(s2, (ETYPE)(target_long)s1));                 \
1510     }                                                               \
1511     for (; i < vlmax; i++) {                                        \
1512         vext_set_elem_mask(vd, mlen, i, 0);                         \
1513     }                                                               \
1514 }
1515 
1516 GEN_VEXT_CMP_VX(vmseq_vx_b, uint8_t,  H1, DO_MSEQ)
1517 GEN_VEXT_CMP_VX(vmseq_vx_h, uint16_t, H2, DO_MSEQ)
1518 GEN_VEXT_CMP_VX(vmseq_vx_w, uint32_t, H4, DO_MSEQ)
1519 GEN_VEXT_CMP_VX(vmseq_vx_d, uint64_t, H8, DO_MSEQ)
1520 
1521 GEN_VEXT_CMP_VX(vmsne_vx_b, uint8_t,  H1, DO_MSNE)
1522 GEN_VEXT_CMP_VX(vmsne_vx_h, uint16_t, H2, DO_MSNE)
1523 GEN_VEXT_CMP_VX(vmsne_vx_w, uint32_t, H4, DO_MSNE)
1524 GEN_VEXT_CMP_VX(vmsne_vx_d, uint64_t, H8, DO_MSNE)
1525 
1526 GEN_VEXT_CMP_VX(vmsltu_vx_b, uint8_t,  H1, DO_MSLT)
1527 GEN_VEXT_CMP_VX(vmsltu_vx_h, uint16_t, H2, DO_MSLT)
1528 GEN_VEXT_CMP_VX(vmsltu_vx_w, uint32_t, H4, DO_MSLT)
1529 GEN_VEXT_CMP_VX(vmsltu_vx_d, uint64_t, H8, DO_MSLT)
1530 
1531 GEN_VEXT_CMP_VX(vmslt_vx_b, int8_t,  H1, DO_MSLT)
1532 GEN_VEXT_CMP_VX(vmslt_vx_h, int16_t, H2, DO_MSLT)
1533 GEN_VEXT_CMP_VX(vmslt_vx_w, int32_t, H4, DO_MSLT)
1534 GEN_VEXT_CMP_VX(vmslt_vx_d, int64_t, H8, DO_MSLT)
1535 
1536 GEN_VEXT_CMP_VX(vmsleu_vx_b, uint8_t,  H1, DO_MSLE)
1537 GEN_VEXT_CMP_VX(vmsleu_vx_h, uint16_t, H2, DO_MSLE)
1538 GEN_VEXT_CMP_VX(vmsleu_vx_w, uint32_t, H4, DO_MSLE)
1539 GEN_VEXT_CMP_VX(vmsleu_vx_d, uint64_t, H8, DO_MSLE)
1540 
1541 GEN_VEXT_CMP_VX(vmsle_vx_b, int8_t,  H1, DO_MSLE)
1542 GEN_VEXT_CMP_VX(vmsle_vx_h, int16_t, H2, DO_MSLE)
1543 GEN_VEXT_CMP_VX(vmsle_vx_w, int32_t, H4, DO_MSLE)
1544 GEN_VEXT_CMP_VX(vmsle_vx_d, int64_t, H8, DO_MSLE)
1545 
1546 GEN_VEXT_CMP_VX(vmsgtu_vx_b, uint8_t,  H1, DO_MSGT)
1547 GEN_VEXT_CMP_VX(vmsgtu_vx_h, uint16_t, H2, DO_MSGT)
1548 GEN_VEXT_CMP_VX(vmsgtu_vx_w, uint32_t, H4, DO_MSGT)
1549 GEN_VEXT_CMP_VX(vmsgtu_vx_d, uint64_t, H8, DO_MSGT)
1550 
1551 GEN_VEXT_CMP_VX(vmsgt_vx_b, int8_t,  H1, DO_MSGT)
1552 GEN_VEXT_CMP_VX(vmsgt_vx_h, int16_t, H2, DO_MSGT)
1553 GEN_VEXT_CMP_VX(vmsgt_vx_w, int32_t, H4, DO_MSGT)
1554 GEN_VEXT_CMP_VX(vmsgt_vx_d, int64_t, H8, DO_MSGT)
1555 
1556 /* Vector Integer Min/Max Instructions */
1557 RVVCALL(OPIVV2, vminu_vv_b, OP_UUU_B, H1, H1, H1, DO_MIN)
1558 RVVCALL(OPIVV2, vminu_vv_h, OP_UUU_H, H2, H2, H2, DO_MIN)
1559 RVVCALL(OPIVV2, vminu_vv_w, OP_UUU_W, H4, H4, H4, DO_MIN)
1560 RVVCALL(OPIVV2, vminu_vv_d, OP_UUU_D, H8, H8, H8, DO_MIN)
1561 RVVCALL(OPIVV2, vmin_vv_b, OP_SSS_B, H1, H1, H1, DO_MIN)
1562 RVVCALL(OPIVV2, vmin_vv_h, OP_SSS_H, H2, H2, H2, DO_MIN)
1563 RVVCALL(OPIVV2, vmin_vv_w, OP_SSS_W, H4, H4, H4, DO_MIN)
1564 RVVCALL(OPIVV2, vmin_vv_d, OP_SSS_D, H8, H8, H8, DO_MIN)
1565 RVVCALL(OPIVV2, vmaxu_vv_b, OP_UUU_B, H1, H1, H1, DO_MAX)
1566 RVVCALL(OPIVV2, vmaxu_vv_h, OP_UUU_H, H2, H2, H2, DO_MAX)
1567 RVVCALL(OPIVV2, vmaxu_vv_w, OP_UUU_W, H4, H4, H4, DO_MAX)
1568 RVVCALL(OPIVV2, vmaxu_vv_d, OP_UUU_D, H8, H8, H8, DO_MAX)
1569 RVVCALL(OPIVV2, vmax_vv_b, OP_SSS_B, H1, H1, H1, DO_MAX)
1570 RVVCALL(OPIVV2, vmax_vv_h, OP_SSS_H, H2, H2, H2, DO_MAX)
1571 RVVCALL(OPIVV2, vmax_vv_w, OP_SSS_W, H4, H4, H4, DO_MAX)
1572 RVVCALL(OPIVV2, vmax_vv_d, OP_SSS_D, H8, H8, H8, DO_MAX)
1573 GEN_VEXT_VV(vminu_vv_b, 1, 1, clearb)
1574 GEN_VEXT_VV(vminu_vv_h, 2, 2, clearh)
1575 GEN_VEXT_VV(vminu_vv_w, 4, 4, clearl)
1576 GEN_VEXT_VV(vminu_vv_d, 8, 8, clearq)
1577 GEN_VEXT_VV(vmin_vv_b, 1, 1, clearb)
1578 GEN_VEXT_VV(vmin_vv_h, 2, 2, clearh)
1579 GEN_VEXT_VV(vmin_vv_w, 4, 4, clearl)
1580 GEN_VEXT_VV(vmin_vv_d, 8, 8, clearq)
1581 GEN_VEXT_VV(vmaxu_vv_b, 1, 1, clearb)
1582 GEN_VEXT_VV(vmaxu_vv_h, 2, 2, clearh)
1583 GEN_VEXT_VV(vmaxu_vv_w, 4, 4, clearl)
1584 GEN_VEXT_VV(vmaxu_vv_d, 8, 8, clearq)
1585 GEN_VEXT_VV(vmax_vv_b, 1, 1, clearb)
1586 GEN_VEXT_VV(vmax_vv_h, 2, 2, clearh)
1587 GEN_VEXT_VV(vmax_vv_w, 4, 4, clearl)
1588 GEN_VEXT_VV(vmax_vv_d, 8, 8, clearq)
1589 
1590 RVVCALL(OPIVX2, vminu_vx_b, OP_UUU_B, H1, H1, DO_MIN)
1591 RVVCALL(OPIVX2, vminu_vx_h, OP_UUU_H, H2, H2, DO_MIN)
1592 RVVCALL(OPIVX2, vminu_vx_w, OP_UUU_W, H4, H4, DO_MIN)
1593 RVVCALL(OPIVX2, vminu_vx_d, OP_UUU_D, H8, H8, DO_MIN)
1594 RVVCALL(OPIVX2, vmin_vx_b, OP_SSS_B, H1, H1, DO_MIN)
1595 RVVCALL(OPIVX2, vmin_vx_h, OP_SSS_H, H2, H2, DO_MIN)
1596 RVVCALL(OPIVX2, vmin_vx_w, OP_SSS_W, H4, H4, DO_MIN)
1597 RVVCALL(OPIVX2, vmin_vx_d, OP_SSS_D, H8, H8, DO_MIN)
1598 RVVCALL(OPIVX2, vmaxu_vx_b, OP_UUU_B, H1, H1, DO_MAX)
1599 RVVCALL(OPIVX2, vmaxu_vx_h, OP_UUU_H, H2, H2, DO_MAX)
1600 RVVCALL(OPIVX2, vmaxu_vx_w, OP_UUU_W, H4, H4, DO_MAX)
1601 RVVCALL(OPIVX2, vmaxu_vx_d, OP_UUU_D, H8, H8, DO_MAX)
1602 RVVCALL(OPIVX2, vmax_vx_b, OP_SSS_B, H1, H1, DO_MAX)
1603 RVVCALL(OPIVX2, vmax_vx_h, OP_SSS_H, H2, H2, DO_MAX)
1604 RVVCALL(OPIVX2, vmax_vx_w, OP_SSS_W, H4, H4, DO_MAX)
1605 RVVCALL(OPIVX2, vmax_vx_d, OP_SSS_D, H8, H8, DO_MAX)
1606 GEN_VEXT_VX(vminu_vx_b, 1, 1, clearb)
1607 GEN_VEXT_VX(vminu_vx_h, 2, 2, clearh)
1608 GEN_VEXT_VX(vminu_vx_w, 4, 4, clearl)
1609 GEN_VEXT_VX(vminu_vx_d, 8, 8, clearq)
1610 GEN_VEXT_VX(vmin_vx_b, 1, 1, clearb)
1611 GEN_VEXT_VX(vmin_vx_h, 2, 2, clearh)
1612 GEN_VEXT_VX(vmin_vx_w, 4, 4, clearl)
1613 GEN_VEXT_VX(vmin_vx_d, 8, 8, clearq)
1614 GEN_VEXT_VX(vmaxu_vx_b, 1, 1, clearb)
1615 GEN_VEXT_VX(vmaxu_vx_h, 2, 2, clearh)
1616 GEN_VEXT_VX(vmaxu_vx_w, 4, 4, clearl)
1617 GEN_VEXT_VX(vmaxu_vx_d, 8, 8,  clearq)
1618 GEN_VEXT_VX(vmax_vx_b, 1, 1, clearb)
1619 GEN_VEXT_VX(vmax_vx_h, 2, 2, clearh)
1620 GEN_VEXT_VX(vmax_vx_w, 4, 4, clearl)
1621 GEN_VEXT_VX(vmax_vx_d, 8, 8, clearq)
1622 
1623 /* Vector Single-Width Integer Multiply Instructions */
1624 #define DO_MUL(N, M) (N * M)
1625 RVVCALL(OPIVV2, vmul_vv_b, OP_SSS_B, H1, H1, H1, DO_MUL)
1626 RVVCALL(OPIVV2, vmul_vv_h, OP_SSS_H, H2, H2, H2, DO_MUL)
1627 RVVCALL(OPIVV2, vmul_vv_w, OP_SSS_W, H4, H4, H4, DO_MUL)
1628 RVVCALL(OPIVV2, vmul_vv_d, OP_SSS_D, H8, H8, H8, DO_MUL)
1629 GEN_VEXT_VV(vmul_vv_b, 1, 1, clearb)
1630 GEN_VEXT_VV(vmul_vv_h, 2, 2, clearh)
1631 GEN_VEXT_VV(vmul_vv_w, 4, 4, clearl)
1632 GEN_VEXT_VV(vmul_vv_d, 8, 8, clearq)
1633 
1634 static int8_t do_mulh_b(int8_t s2, int8_t s1)
1635 {
1636     return (int16_t)s2 * (int16_t)s1 >> 8;
1637 }
1638 
1639 static int16_t do_mulh_h(int16_t s2, int16_t s1)
1640 {
1641     return (int32_t)s2 * (int32_t)s1 >> 16;
1642 }
1643 
1644 static int32_t do_mulh_w(int32_t s2, int32_t s1)
1645 {
1646     return (int64_t)s2 * (int64_t)s1 >> 32;
1647 }
1648 
1649 static int64_t do_mulh_d(int64_t s2, int64_t s1)
1650 {
1651     uint64_t hi_64, lo_64;
1652 
1653     muls64(&lo_64, &hi_64, s1, s2);
1654     return hi_64;
1655 }
1656 
1657 static uint8_t do_mulhu_b(uint8_t s2, uint8_t s1)
1658 {
1659     return (uint16_t)s2 * (uint16_t)s1 >> 8;
1660 }
1661 
1662 static uint16_t do_mulhu_h(uint16_t s2, uint16_t s1)
1663 {
1664     return (uint32_t)s2 * (uint32_t)s1 >> 16;
1665 }
1666 
1667 static uint32_t do_mulhu_w(uint32_t s2, uint32_t s1)
1668 {
1669     return (uint64_t)s2 * (uint64_t)s1 >> 32;
1670 }
1671 
1672 static uint64_t do_mulhu_d(uint64_t s2, uint64_t s1)
1673 {
1674     uint64_t hi_64, lo_64;
1675 
1676     mulu64(&lo_64, &hi_64, s2, s1);
1677     return hi_64;
1678 }
1679 
1680 static int8_t do_mulhsu_b(int8_t s2, uint8_t s1)
1681 {
1682     return (int16_t)s2 * (uint16_t)s1 >> 8;
1683 }
1684 
1685 static int16_t do_mulhsu_h(int16_t s2, uint16_t s1)
1686 {
1687     return (int32_t)s2 * (uint32_t)s1 >> 16;
1688 }
1689 
1690 static int32_t do_mulhsu_w(int32_t s2, uint32_t s1)
1691 {
1692     return (int64_t)s2 * (uint64_t)s1 >> 32;
1693 }
1694 
1695 /*
1696  * Let  A = signed operand,
1697  *      B = unsigned operand
1698  *      P = mulu64(A, B), unsigned product
1699  *
1700  * LET  X = 2 ** 64  - A, 2's complement of A
1701  *      SP = signed product
1702  * THEN
1703  *      IF A < 0
1704  *          SP = -X * B
1705  *             = -(2 ** 64 - A) * B
1706  *             = A * B - 2 ** 64 * B
1707  *             = P - 2 ** 64 * B
1708  *      ELSE
1709  *          SP = P
1710  * THEN
1711  *      HI_P -= (A < 0 ? B : 0)
1712  */
1713 
1714 static int64_t do_mulhsu_d(int64_t s2, uint64_t s1)
1715 {
1716     uint64_t hi_64, lo_64;
1717 
1718     mulu64(&lo_64, &hi_64, s2, s1);
1719 
1720     hi_64 -= s2 < 0 ? s1 : 0;
1721     return hi_64;
1722 }
1723 
1724 RVVCALL(OPIVV2, vmulh_vv_b, OP_SSS_B, H1, H1, H1, do_mulh_b)
1725 RVVCALL(OPIVV2, vmulh_vv_h, OP_SSS_H, H2, H2, H2, do_mulh_h)
1726 RVVCALL(OPIVV2, vmulh_vv_w, OP_SSS_W, H4, H4, H4, do_mulh_w)
1727 RVVCALL(OPIVV2, vmulh_vv_d, OP_SSS_D, H8, H8, H8, do_mulh_d)
1728 RVVCALL(OPIVV2, vmulhu_vv_b, OP_UUU_B, H1, H1, H1, do_mulhu_b)
1729 RVVCALL(OPIVV2, vmulhu_vv_h, OP_UUU_H, H2, H2, H2, do_mulhu_h)
1730 RVVCALL(OPIVV2, vmulhu_vv_w, OP_UUU_W, H4, H4, H4, do_mulhu_w)
1731 RVVCALL(OPIVV2, vmulhu_vv_d, OP_UUU_D, H8, H8, H8, do_mulhu_d)
1732 RVVCALL(OPIVV2, vmulhsu_vv_b, OP_SUS_B, H1, H1, H1, do_mulhsu_b)
1733 RVVCALL(OPIVV2, vmulhsu_vv_h, OP_SUS_H, H2, H2, H2, do_mulhsu_h)
1734 RVVCALL(OPIVV2, vmulhsu_vv_w, OP_SUS_W, H4, H4, H4, do_mulhsu_w)
1735 RVVCALL(OPIVV2, vmulhsu_vv_d, OP_SUS_D, H8, H8, H8, do_mulhsu_d)
1736 GEN_VEXT_VV(vmulh_vv_b, 1, 1, clearb)
1737 GEN_VEXT_VV(vmulh_vv_h, 2, 2, clearh)
1738 GEN_VEXT_VV(vmulh_vv_w, 4, 4, clearl)
1739 GEN_VEXT_VV(vmulh_vv_d, 8, 8, clearq)
1740 GEN_VEXT_VV(vmulhu_vv_b, 1, 1, clearb)
1741 GEN_VEXT_VV(vmulhu_vv_h, 2, 2, clearh)
1742 GEN_VEXT_VV(vmulhu_vv_w, 4, 4, clearl)
1743 GEN_VEXT_VV(vmulhu_vv_d, 8, 8, clearq)
1744 GEN_VEXT_VV(vmulhsu_vv_b, 1, 1, clearb)
1745 GEN_VEXT_VV(vmulhsu_vv_h, 2, 2, clearh)
1746 GEN_VEXT_VV(vmulhsu_vv_w, 4, 4, clearl)
1747 GEN_VEXT_VV(vmulhsu_vv_d, 8, 8, clearq)
1748 
1749 RVVCALL(OPIVX2, vmul_vx_b, OP_SSS_B, H1, H1, DO_MUL)
1750 RVVCALL(OPIVX2, vmul_vx_h, OP_SSS_H, H2, H2, DO_MUL)
1751 RVVCALL(OPIVX2, vmul_vx_w, OP_SSS_W, H4, H4, DO_MUL)
1752 RVVCALL(OPIVX2, vmul_vx_d, OP_SSS_D, H8, H8, DO_MUL)
1753 RVVCALL(OPIVX2, vmulh_vx_b, OP_SSS_B, H1, H1, do_mulh_b)
1754 RVVCALL(OPIVX2, vmulh_vx_h, OP_SSS_H, H2, H2, do_mulh_h)
1755 RVVCALL(OPIVX2, vmulh_vx_w, OP_SSS_W, H4, H4, do_mulh_w)
1756 RVVCALL(OPIVX2, vmulh_vx_d, OP_SSS_D, H8, H8, do_mulh_d)
1757 RVVCALL(OPIVX2, vmulhu_vx_b, OP_UUU_B, H1, H1, do_mulhu_b)
1758 RVVCALL(OPIVX2, vmulhu_vx_h, OP_UUU_H, H2, H2, do_mulhu_h)
1759 RVVCALL(OPIVX2, vmulhu_vx_w, OP_UUU_W, H4, H4, do_mulhu_w)
1760 RVVCALL(OPIVX2, vmulhu_vx_d, OP_UUU_D, H8, H8, do_mulhu_d)
1761 RVVCALL(OPIVX2, vmulhsu_vx_b, OP_SUS_B, H1, H1, do_mulhsu_b)
1762 RVVCALL(OPIVX2, vmulhsu_vx_h, OP_SUS_H, H2, H2, do_mulhsu_h)
1763 RVVCALL(OPIVX2, vmulhsu_vx_w, OP_SUS_W, H4, H4, do_mulhsu_w)
1764 RVVCALL(OPIVX2, vmulhsu_vx_d, OP_SUS_D, H8, H8, do_mulhsu_d)
1765 GEN_VEXT_VX(vmul_vx_b, 1, 1, clearb)
1766 GEN_VEXT_VX(vmul_vx_h, 2, 2, clearh)
1767 GEN_VEXT_VX(vmul_vx_w, 4, 4, clearl)
1768 GEN_VEXT_VX(vmul_vx_d, 8, 8, clearq)
1769 GEN_VEXT_VX(vmulh_vx_b, 1, 1, clearb)
1770 GEN_VEXT_VX(vmulh_vx_h, 2, 2, clearh)
1771 GEN_VEXT_VX(vmulh_vx_w, 4, 4, clearl)
1772 GEN_VEXT_VX(vmulh_vx_d, 8, 8, clearq)
1773 GEN_VEXT_VX(vmulhu_vx_b, 1, 1, clearb)
1774 GEN_VEXT_VX(vmulhu_vx_h, 2, 2, clearh)
1775 GEN_VEXT_VX(vmulhu_vx_w, 4, 4, clearl)
1776 GEN_VEXT_VX(vmulhu_vx_d, 8, 8, clearq)
1777 GEN_VEXT_VX(vmulhsu_vx_b, 1, 1, clearb)
1778 GEN_VEXT_VX(vmulhsu_vx_h, 2, 2, clearh)
1779 GEN_VEXT_VX(vmulhsu_vx_w, 4, 4, clearl)
1780 GEN_VEXT_VX(vmulhsu_vx_d, 8, 8, clearq)
1781 
1782 /* Vector Integer Divide Instructions */
1783 #define DO_DIVU(N, M) (unlikely(M == 0) ? (__typeof(N))(-1) : N / M)
1784 #define DO_REMU(N, M) (unlikely(M == 0) ? N : N % M)
1785 #define DO_DIV(N, M)  (unlikely(M == 0) ? (__typeof(N))(-1) :\
1786         unlikely((N == -N) && (M == (__typeof(N))(-1))) ? N : N / M)
1787 #define DO_REM(N, M)  (unlikely(M == 0) ? N :\
1788         unlikely((N == -N) && (M == (__typeof(N))(-1))) ? 0 : N % M)
1789 
1790 RVVCALL(OPIVV2, vdivu_vv_b, OP_UUU_B, H1, H1, H1, DO_DIVU)
1791 RVVCALL(OPIVV2, vdivu_vv_h, OP_UUU_H, H2, H2, H2, DO_DIVU)
1792 RVVCALL(OPIVV2, vdivu_vv_w, OP_UUU_W, H4, H4, H4, DO_DIVU)
1793 RVVCALL(OPIVV2, vdivu_vv_d, OP_UUU_D, H8, H8, H8, DO_DIVU)
1794 RVVCALL(OPIVV2, vdiv_vv_b, OP_SSS_B, H1, H1, H1, DO_DIV)
1795 RVVCALL(OPIVV2, vdiv_vv_h, OP_SSS_H, H2, H2, H2, DO_DIV)
1796 RVVCALL(OPIVV2, vdiv_vv_w, OP_SSS_W, H4, H4, H4, DO_DIV)
1797 RVVCALL(OPIVV2, vdiv_vv_d, OP_SSS_D, H8, H8, H8, DO_DIV)
1798 RVVCALL(OPIVV2, vremu_vv_b, OP_UUU_B, H1, H1, H1, DO_REMU)
1799 RVVCALL(OPIVV2, vremu_vv_h, OP_UUU_H, H2, H2, H2, DO_REMU)
1800 RVVCALL(OPIVV2, vremu_vv_w, OP_UUU_W, H4, H4, H4, DO_REMU)
1801 RVVCALL(OPIVV2, vremu_vv_d, OP_UUU_D, H8, H8, H8, DO_REMU)
1802 RVVCALL(OPIVV2, vrem_vv_b, OP_SSS_B, H1, H1, H1, DO_REM)
1803 RVVCALL(OPIVV2, vrem_vv_h, OP_SSS_H, H2, H2, H2, DO_REM)
1804 RVVCALL(OPIVV2, vrem_vv_w, OP_SSS_W, H4, H4, H4, DO_REM)
1805 RVVCALL(OPIVV2, vrem_vv_d, OP_SSS_D, H8, H8, H8, DO_REM)
1806 GEN_VEXT_VV(vdivu_vv_b, 1, 1, clearb)
1807 GEN_VEXT_VV(vdivu_vv_h, 2, 2, clearh)
1808 GEN_VEXT_VV(vdivu_vv_w, 4, 4, clearl)
1809 GEN_VEXT_VV(vdivu_vv_d, 8, 8, clearq)
1810 GEN_VEXT_VV(vdiv_vv_b, 1, 1, clearb)
1811 GEN_VEXT_VV(vdiv_vv_h, 2, 2, clearh)
1812 GEN_VEXT_VV(vdiv_vv_w, 4, 4, clearl)
1813 GEN_VEXT_VV(vdiv_vv_d, 8, 8, clearq)
1814 GEN_VEXT_VV(vremu_vv_b, 1, 1, clearb)
1815 GEN_VEXT_VV(vremu_vv_h, 2, 2, clearh)
1816 GEN_VEXT_VV(vremu_vv_w, 4, 4, clearl)
1817 GEN_VEXT_VV(vremu_vv_d, 8, 8, clearq)
1818 GEN_VEXT_VV(vrem_vv_b, 1, 1, clearb)
1819 GEN_VEXT_VV(vrem_vv_h, 2, 2, clearh)
1820 GEN_VEXT_VV(vrem_vv_w, 4, 4, clearl)
1821 GEN_VEXT_VV(vrem_vv_d, 8, 8, clearq)
1822 
1823 RVVCALL(OPIVX2, vdivu_vx_b, OP_UUU_B, H1, H1, DO_DIVU)
1824 RVVCALL(OPIVX2, vdivu_vx_h, OP_UUU_H, H2, H2, DO_DIVU)
1825 RVVCALL(OPIVX2, vdivu_vx_w, OP_UUU_W, H4, H4, DO_DIVU)
1826 RVVCALL(OPIVX2, vdivu_vx_d, OP_UUU_D, H8, H8, DO_DIVU)
1827 RVVCALL(OPIVX2, vdiv_vx_b, OP_SSS_B, H1, H1, DO_DIV)
1828 RVVCALL(OPIVX2, vdiv_vx_h, OP_SSS_H, H2, H2, DO_DIV)
1829 RVVCALL(OPIVX2, vdiv_vx_w, OP_SSS_W, H4, H4, DO_DIV)
1830 RVVCALL(OPIVX2, vdiv_vx_d, OP_SSS_D, H8, H8, DO_DIV)
1831 RVVCALL(OPIVX2, vremu_vx_b, OP_UUU_B, H1, H1, DO_REMU)
1832 RVVCALL(OPIVX2, vremu_vx_h, OP_UUU_H, H2, H2, DO_REMU)
1833 RVVCALL(OPIVX2, vremu_vx_w, OP_UUU_W, H4, H4, DO_REMU)
1834 RVVCALL(OPIVX2, vremu_vx_d, OP_UUU_D, H8, H8, DO_REMU)
1835 RVVCALL(OPIVX2, vrem_vx_b, OP_SSS_B, H1, H1, DO_REM)
1836 RVVCALL(OPIVX2, vrem_vx_h, OP_SSS_H, H2, H2, DO_REM)
1837 RVVCALL(OPIVX2, vrem_vx_w, OP_SSS_W, H4, H4, DO_REM)
1838 RVVCALL(OPIVX2, vrem_vx_d, OP_SSS_D, H8, H8, DO_REM)
1839 GEN_VEXT_VX(vdivu_vx_b, 1, 1, clearb)
1840 GEN_VEXT_VX(vdivu_vx_h, 2, 2, clearh)
1841 GEN_VEXT_VX(vdivu_vx_w, 4, 4, clearl)
1842 GEN_VEXT_VX(vdivu_vx_d, 8, 8, clearq)
1843 GEN_VEXT_VX(vdiv_vx_b, 1, 1, clearb)
1844 GEN_VEXT_VX(vdiv_vx_h, 2, 2, clearh)
1845 GEN_VEXT_VX(vdiv_vx_w, 4, 4, clearl)
1846 GEN_VEXT_VX(vdiv_vx_d, 8, 8, clearq)
1847 GEN_VEXT_VX(vremu_vx_b, 1, 1, clearb)
1848 GEN_VEXT_VX(vremu_vx_h, 2, 2, clearh)
1849 GEN_VEXT_VX(vremu_vx_w, 4, 4, clearl)
1850 GEN_VEXT_VX(vremu_vx_d, 8, 8, clearq)
1851 GEN_VEXT_VX(vrem_vx_b, 1, 1, clearb)
1852 GEN_VEXT_VX(vrem_vx_h, 2, 2, clearh)
1853 GEN_VEXT_VX(vrem_vx_w, 4, 4, clearl)
1854 GEN_VEXT_VX(vrem_vx_d, 8, 8, clearq)
1855 
1856 /* Vector Widening Integer Multiply Instructions */
1857 RVVCALL(OPIVV2, vwmul_vv_b, WOP_SSS_B, H2, H1, H1, DO_MUL)
1858 RVVCALL(OPIVV2, vwmul_vv_h, WOP_SSS_H, H4, H2, H2, DO_MUL)
1859 RVVCALL(OPIVV2, vwmul_vv_w, WOP_SSS_W, H8, H4, H4, DO_MUL)
1860 RVVCALL(OPIVV2, vwmulu_vv_b, WOP_UUU_B, H2, H1, H1, DO_MUL)
1861 RVVCALL(OPIVV2, vwmulu_vv_h, WOP_UUU_H, H4, H2, H2, DO_MUL)
1862 RVVCALL(OPIVV2, vwmulu_vv_w, WOP_UUU_W, H8, H4, H4, DO_MUL)
1863 RVVCALL(OPIVV2, vwmulsu_vv_b, WOP_SUS_B, H2, H1, H1, DO_MUL)
1864 RVVCALL(OPIVV2, vwmulsu_vv_h, WOP_SUS_H, H4, H2, H2, DO_MUL)
1865 RVVCALL(OPIVV2, vwmulsu_vv_w, WOP_SUS_W, H8, H4, H4, DO_MUL)
1866 GEN_VEXT_VV(vwmul_vv_b, 1, 2, clearh)
1867 GEN_VEXT_VV(vwmul_vv_h, 2, 4, clearl)
1868 GEN_VEXT_VV(vwmul_vv_w, 4, 8, clearq)
1869 GEN_VEXT_VV(vwmulu_vv_b, 1, 2, clearh)
1870 GEN_VEXT_VV(vwmulu_vv_h, 2, 4, clearl)
1871 GEN_VEXT_VV(vwmulu_vv_w, 4, 8, clearq)
1872 GEN_VEXT_VV(vwmulsu_vv_b, 1, 2, clearh)
1873 GEN_VEXT_VV(vwmulsu_vv_h, 2, 4, clearl)
1874 GEN_VEXT_VV(vwmulsu_vv_w, 4, 8, clearq)
1875 
1876 RVVCALL(OPIVX2, vwmul_vx_b, WOP_SSS_B, H2, H1, DO_MUL)
1877 RVVCALL(OPIVX2, vwmul_vx_h, WOP_SSS_H, H4, H2, DO_MUL)
1878 RVVCALL(OPIVX2, vwmul_vx_w, WOP_SSS_W, H8, H4, DO_MUL)
1879 RVVCALL(OPIVX2, vwmulu_vx_b, WOP_UUU_B, H2, H1, DO_MUL)
1880 RVVCALL(OPIVX2, vwmulu_vx_h, WOP_UUU_H, H4, H2, DO_MUL)
1881 RVVCALL(OPIVX2, vwmulu_vx_w, WOP_UUU_W, H8, H4, DO_MUL)
1882 RVVCALL(OPIVX2, vwmulsu_vx_b, WOP_SUS_B, H2, H1, DO_MUL)
1883 RVVCALL(OPIVX2, vwmulsu_vx_h, WOP_SUS_H, H4, H2, DO_MUL)
1884 RVVCALL(OPIVX2, vwmulsu_vx_w, WOP_SUS_W, H8, H4, DO_MUL)
1885 GEN_VEXT_VX(vwmul_vx_b, 1, 2, clearh)
1886 GEN_VEXT_VX(vwmul_vx_h, 2, 4, clearl)
1887 GEN_VEXT_VX(vwmul_vx_w, 4, 8, clearq)
1888 GEN_VEXT_VX(vwmulu_vx_b, 1, 2, clearh)
1889 GEN_VEXT_VX(vwmulu_vx_h, 2, 4, clearl)
1890 GEN_VEXT_VX(vwmulu_vx_w, 4, 8, clearq)
1891 GEN_VEXT_VX(vwmulsu_vx_b, 1, 2, clearh)
1892 GEN_VEXT_VX(vwmulsu_vx_h, 2, 4, clearl)
1893 GEN_VEXT_VX(vwmulsu_vx_w, 4, 8, clearq)
1894 
1895 /* Vector Single-Width Integer Multiply-Add Instructions */
1896 #define OPIVV3(NAME, TD, T1, T2, TX1, TX2, HD, HS1, HS2, OP)   \
1897 static void do_##NAME(void *vd, void *vs1, void *vs2, int i)       \
1898 {                                                                  \
1899     TX1 s1 = *((T1 *)vs1 + HS1(i));                                \
1900     TX2 s2 = *((T2 *)vs2 + HS2(i));                                \
1901     TD d = *((TD *)vd + HD(i));                                    \
1902     *((TD *)vd + HD(i)) = OP(s2, s1, d);                           \
1903 }
1904 
1905 #define DO_MACC(N, M, D) (M * N + D)
1906 #define DO_NMSAC(N, M, D) (-(M * N) + D)
1907 #define DO_MADD(N, M, D) (M * D + N)
1908 #define DO_NMSUB(N, M, D) (-(M * D) + N)
1909 RVVCALL(OPIVV3, vmacc_vv_b, OP_SSS_B, H1, H1, H1, DO_MACC)
1910 RVVCALL(OPIVV3, vmacc_vv_h, OP_SSS_H, H2, H2, H2, DO_MACC)
1911 RVVCALL(OPIVV3, vmacc_vv_w, OP_SSS_W, H4, H4, H4, DO_MACC)
1912 RVVCALL(OPIVV3, vmacc_vv_d, OP_SSS_D, H8, H8, H8, DO_MACC)
1913 RVVCALL(OPIVV3, vnmsac_vv_b, OP_SSS_B, H1, H1, H1, DO_NMSAC)
1914 RVVCALL(OPIVV3, vnmsac_vv_h, OP_SSS_H, H2, H2, H2, DO_NMSAC)
1915 RVVCALL(OPIVV3, vnmsac_vv_w, OP_SSS_W, H4, H4, H4, DO_NMSAC)
1916 RVVCALL(OPIVV3, vnmsac_vv_d, OP_SSS_D, H8, H8, H8, DO_NMSAC)
1917 RVVCALL(OPIVV3, vmadd_vv_b, OP_SSS_B, H1, H1, H1, DO_MADD)
1918 RVVCALL(OPIVV3, vmadd_vv_h, OP_SSS_H, H2, H2, H2, DO_MADD)
1919 RVVCALL(OPIVV3, vmadd_vv_w, OP_SSS_W, H4, H4, H4, DO_MADD)
1920 RVVCALL(OPIVV3, vmadd_vv_d, OP_SSS_D, H8, H8, H8, DO_MADD)
1921 RVVCALL(OPIVV3, vnmsub_vv_b, OP_SSS_B, H1, H1, H1, DO_NMSUB)
1922 RVVCALL(OPIVV3, vnmsub_vv_h, OP_SSS_H, H2, H2, H2, DO_NMSUB)
1923 RVVCALL(OPIVV3, vnmsub_vv_w, OP_SSS_W, H4, H4, H4, DO_NMSUB)
1924 RVVCALL(OPIVV3, vnmsub_vv_d, OP_SSS_D, H8, H8, H8, DO_NMSUB)
1925 GEN_VEXT_VV(vmacc_vv_b, 1, 1, clearb)
1926 GEN_VEXT_VV(vmacc_vv_h, 2, 2, clearh)
1927 GEN_VEXT_VV(vmacc_vv_w, 4, 4, clearl)
1928 GEN_VEXT_VV(vmacc_vv_d, 8, 8, clearq)
1929 GEN_VEXT_VV(vnmsac_vv_b, 1, 1, clearb)
1930 GEN_VEXT_VV(vnmsac_vv_h, 2, 2, clearh)
1931 GEN_VEXT_VV(vnmsac_vv_w, 4, 4, clearl)
1932 GEN_VEXT_VV(vnmsac_vv_d, 8, 8, clearq)
1933 GEN_VEXT_VV(vmadd_vv_b, 1, 1, clearb)
1934 GEN_VEXT_VV(vmadd_vv_h, 2, 2, clearh)
1935 GEN_VEXT_VV(vmadd_vv_w, 4, 4, clearl)
1936 GEN_VEXT_VV(vmadd_vv_d, 8, 8, clearq)
1937 GEN_VEXT_VV(vnmsub_vv_b, 1, 1, clearb)
1938 GEN_VEXT_VV(vnmsub_vv_h, 2, 2, clearh)
1939 GEN_VEXT_VV(vnmsub_vv_w, 4, 4, clearl)
1940 GEN_VEXT_VV(vnmsub_vv_d, 8, 8, clearq)
1941 
1942 #define OPIVX3(NAME, TD, T1, T2, TX1, TX2, HD, HS2, OP)             \
1943 static void do_##NAME(void *vd, target_long s1, void *vs2, int i)   \
1944 {                                                                   \
1945     TX2 s2 = *((T2 *)vs2 + HS2(i));                                 \
1946     TD d = *((TD *)vd + HD(i));                                     \
1947     *((TD *)vd + HD(i)) = OP(s2, (TX1)(T1)s1, d);                   \
1948 }
1949 
1950 RVVCALL(OPIVX3, vmacc_vx_b, OP_SSS_B, H1, H1, DO_MACC)
1951 RVVCALL(OPIVX3, vmacc_vx_h, OP_SSS_H, H2, H2, DO_MACC)
1952 RVVCALL(OPIVX3, vmacc_vx_w, OP_SSS_W, H4, H4, DO_MACC)
1953 RVVCALL(OPIVX3, vmacc_vx_d, OP_SSS_D, H8, H8, DO_MACC)
1954 RVVCALL(OPIVX3, vnmsac_vx_b, OP_SSS_B, H1, H1, DO_NMSAC)
1955 RVVCALL(OPIVX3, vnmsac_vx_h, OP_SSS_H, H2, H2, DO_NMSAC)
1956 RVVCALL(OPIVX3, vnmsac_vx_w, OP_SSS_W, H4, H4, DO_NMSAC)
1957 RVVCALL(OPIVX3, vnmsac_vx_d, OP_SSS_D, H8, H8, DO_NMSAC)
1958 RVVCALL(OPIVX3, vmadd_vx_b, OP_SSS_B, H1, H1, DO_MADD)
1959 RVVCALL(OPIVX3, vmadd_vx_h, OP_SSS_H, H2, H2, DO_MADD)
1960 RVVCALL(OPIVX3, vmadd_vx_w, OP_SSS_W, H4, H4, DO_MADD)
1961 RVVCALL(OPIVX3, vmadd_vx_d, OP_SSS_D, H8, H8, DO_MADD)
1962 RVVCALL(OPIVX3, vnmsub_vx_b, OP_SSS_B, H1, H1, DO_NMSUB)
1963 RVVCALL(OPIVX3, vnmsub_vx_h, OP_SSS_H, H2, H2, DO_NMSUB)
1964 RVVCALL(OPIVX3, vnmsub_vx_w, OP_SSS_W, H4, H4, DO_NMSUB)
1965 RVVCALL(OPIVX3, vnmsub_vx_d, OP_SSS_D, H8, H8, DO_NMSUB)
1966 GEN_VEXT_VX(vmacc_vx_b, 1, 1, clearb)
1967 GEN_VEXT_VX(vmacc_vx_h, 2, 2, clearh)
1968 GEN_VEXT_VX(vmacc_vx_w, 4, 4, clearl)
1969 GEN_VEXT_VX(vmacc_vx_d, 8, 8, clearq)
1970 GEN_VEXT_VX(vnmsac_vx_b, 1, 1, clearb)
1971 GEN_VEXT_VX(vnmsac_vx_h, 2, 2, clearh)
1972 GEN_VEXT_VX(vnmsac_vx_w, 4, 4, clearl)
1973 GEN_VEXT_VX(vnmsac_vx_d, 8, 8, clearq)
1974 GEN_VEXT_VX(vmadd_vx_b, 1, 1, clearb)
1975 GEN_VEXT_VX(vmadd_vx_h, 2, 2, clearh)
1976 GEN_VEXT_VX(vmadd_vx_w, 4, 4, clearl)
1977 GEN_VEXT_VX(vmadd_vx_d, 8, 8, clearq)
1978 GEN_VEXT_VX(vnmsub_vx_b, 1, 1, clearb)
1979 GEN_VEXT_VX(vnmsub_vx_h, 2, 2, clearh)
1980 GEN_VEXT_VX(vnmsub_vx_w, 4, 4, clearl)
1981 GEN_VEXT_VX(vnmsub_vx_d, 8, 8, clearq)
1982 
1983 /* Vector Widening Integer Multiply-Add Instructions */
1984 RVVCALL(OPIVV3, vwmaccu_vv_b, WOP_UUU_B, H2, H1, H1, DO_MACC)
1985 RVVCALL(OPIVV3, vwmaccu_vv_h, WOP_UUU_H, H4, H2, H2, DO_MACC)
1986 RVVCALL(OPIVV3, vwmaccu_vv_w, WOP_UUU_W, H8, H4, H4, DO_MACC)
1987 RVVCALL(OPIVV3, vwmacc_vv_b, WOP_SSS_B, H2, H1, H1, DO_MACC)
1988 RVVCALL(OPIVV3, vwmacc_vv_h, WOP_SSS_H, H4, H2, H2, DO_MACC)
1989 RVVCALL(OPIVV3, vwmacc_vv_w, WOP_SSS_W, H8, H4, H4, DO_MACC)
1990 RVVCALL(OPIVV3, vwmaccsu_vv_b, WOP_SSU_B, H2, H1, H1, DO_MACC)
1991 RVVCALL(OPIVV3, vwmaccsu_vv_h, WOP_SSU_H, H4, H2, H2, DO_MACC)
1992 RVVCALL(OPIVV3, vwmaccsu_vv_w, WOP_SSU_W, H8, H4, H4, DO_MACC)
1993 GEN_VEXT_VV(vwmaccu_vv_b, 1, 2, clearh)
1994 GEN_VEXT_VV(vwmaccu_vv_h, 2, 4, clearl)
1995 GEN_VEXT_VV(vwmaccu_vv_w, 4, 8, clearq)
1996 GEN_VEXT_VV(vwmacc_vv_b, 1, 2, clearh)
1997 GEN_VEXT_VV(vwmacc_vv_h, 2, 4, clearl)
1998 GEN_VEXT_VV(vwmacc_vv_w, 4, 8, clearq)
1999 GEN_VEXT_VV(vwmaccsu_vv_b, 1, 2, clearh)
2000 GEN_VEXT_VV(vwmaccsu_vv_h, 2, 4, clearl)
2001 GEN_VEXT_VV(vwmaccsu_vv_w, 4, 8, clearq)
2002 
2003 RVVCALL(OPIVX3, vwmaccu_vx_b, WOP_UUU_B, H2, H1, DO_MACC)
2004 RVVCALL(OPIVX3, vwmaccu_vx_h, WOP_UUU_H, H4, H2, DO_MACC)
2005 RVVCALL(OPIVX3, vwmaccu_vx_w, WOP_UUU_W, H8, H4, DO_MACC)
2006 RVVCALL(OPIVX3, vwmacc_vx_b, WOP_SSS_B, H2, H1, DO_MACC)
2007 RVVCALL(OPIVX3, vwmacc_vx_h, WOP_SSS_H, H4, H2, DO_MACC)
2008 RVVCALL(OPIVX3, vwmacc_vx_w, WOP_SSS_W, H8, H4, DO_MACC)
2009 RVVCALL(OPIVX3, vwmaccsu_vx_b, WOP_SSU_B, H2, H1, DO_MACC)
2010 RVVCALL(OPIVX3, vwmaccsu_vx_h, WOP_SSU_H, H4, H2, DO_MACC)
2011 RVVCALL(OPIVX3, vwmaccsu_vx_w, WOP_SSU_W, H8, H4, DO_MACC)
2012 RVVCALL(OPIVX3, vwmaccus_vx_b, WOP_SUS_B, H2, H1, DO_MACC)
2013 RVVCALL(OPIVX3, vwmaccus_vx_h, WOP_SUS_H, H4, H2, DO_MACC)
2014 RVVCALL(OPIVX3, vwmaccus_vx_w, WOP_SUS_W, H8, H4, DO_MACC)
2015 GEN_VEXT_VX(vwmaccu_vx_b, 1, 2, clearh)
2016 GEN_VEXT_VX(vwmaccu_vx_h, 2, 4, clearl)
2017 GEN_VEXT_VX(vwmaccu_vx_w, 4, 8, clearq)
2018 GEN_VEXT_VX(vwmacc_vx_b, 1, 2, clearh)
2019 GEN_VEXT_VX(vwmacc_vx_h, 2, 4, clearl)
2020 GEN_VEXT_VX(vwmacc_vx_w, 4, 8, clearq)
2021 GEN_VEXT_VX(vwmaccsu_vx_b, 1, 2, clearh)
2022 GEN_VEXT_VX(vwmaccsu_vx_h, 2, 4, clearl)
2023 GEN_VEXT_VX(vwmaccsu_vx_w, 4, 8, clearq)
2024 GEN_VEXT_VX(vwmaccus_vx_b, 1, 2, clearh)
2025 GEN_VEXT_VX(vwmaccus_vx_h, 2, 4, clearl)
2026 GEN_VEXT_VX(vwmaccus_vx_w, 4, 8, clearq)
2027 
2028 /* Vector Integer Merge and Move Instructions */
2029 #define GEN_VEXT_VMV_VV(NAME, ETYPE, H, CLEAR_FN)                    \
2030 void HELPER(NAME)(void *vd, void *vs1, CPURISCVState *env,           \
2031                   uint32_t desc)                                     \
2032 {                                                                    \
2033     uint32_t vl = env->vl;                                           \
2034     uint32_t esz = sizeof(ETYPE);                                    \
2035     uint32_t vlmax = vext_maxsz(desc) / esz;                         \
2036     uint32_t i;                                                      \
2037                                                                      \
2038     for (i = 0; i < vl; i++) {                                       \
2039         ETYPE s1 = *((ETYPE *)vs1 + H(i));                           \
2040         *((ETYPE *)vd + H(i)) = s1;                                  \
2041     }                                                                \
2042     CLEAR_FN(vd, vl, vl * esz, vlmax * esz);                         \
2043 }
2044 
2045 GEN_VEXT_VMV_VV(vmv_v_v_b, int8_t,  H1, clearb)
2046 GEN_VEXT_VMV_VV(vmv_v_v_h, int16_t, H2, clearh)
2047 GEN_VEXT_VMV_VV(vmv_v_v_w, int32_t, H4, clearl)
2048 GEN_VEXT_VMV_VV(vmv_v_v_d, int64_t, H8, clearq)
2049 
2050 #define GEN_VEXT_VMV_VX(NAME, ETYPE, H, CLEAR_FN)                    \
2051 void HELPER(NAME)(void *vd, uint64_t s1, CPURISCVState *env,         \
2052                   uint32_t desc)                                     \
2053 {                                                                    \
2054     uint32_t vl = env->vl;                                           \
2055     uint32_t esz = sizeof(ETYPE);                                    \
2056     uint32_t vlmax = vext_maxsz(desc) / esz;                         \
2057     uint32_t i;                                                      \
2058                                                                      \
2059     for (i = 0; i < vl; i++) {                                       \
2060         *((ETYPE *)vd + H(i)) = (ETYPE)s1;                           \
2061     }                                                                \
2062     CLEAR_FN(vd, vl, vl * esz, vlmax * esz);                         \
2063 }
2064 
2065 GEN_VEXT_VMV_VX(vmv_v_x_b, int8_t,  H1, clearb)
2066 GEN_VEXT_VMV_VX(vmv_v_x_h, int16_t, H2, clearh)
2067 GEN_VEXT_VMV_VX(vmv_v_x_w, int32_t, H4, clearl)
2068 GEN_VEXT_VMV_VX(vmv_v_x_d, int64_t, H8, clearq)
2069 
2070 #define GEN_VEXT_VMERGE_VV(NAME, ETYPE, H, CLEAR_FN)                 \
2071 void HELPER(NAME)(void *vd, void *v0, void *vs1, void *vs2,          \
2072                   CPURISCVState *env, uint32_t desc)                 \
2073 {                                                                    \
2074     uint32_t mlen = vext_mlen(desc);                                 \
2075     uint32_t vl = env->vl;                                           \
2076     uint32_t esz = sizeof(ETYPE);                                    \
2077     uint32_t vlmax = vext_maxsz(desc) / esz;                         \
2078     uint32_t i;                                                      \
2079                                                                      \
2080     for (i = 0; i < vl; i++) {                                       \
2081         ETYPE *vt = (!vext_elem_mask(v0, mlen, i) ? vs2 : vs1);      \
2082         *((ETYPE *)vd + H(i)) = *(vt + H(i));                        \
2083     }                                                                \
2084     CLEAR_FN(vd, vl, vl * esz, vlmax * esz);                         \
2085 }
2086 
2087 GEN_VEXT_VMERGE_VV(vmerge_vvm_b, int8_t,  H1, clearb)
2088 GEN_VEXT_VMERGE_VV(vmerge_vvm_h, int16_t, H2, clearh)
2089 GEN_VEXT_VMERGE_VV(vmerge_vvm_w, int32_t, H4, clearl)
2090 GEN_VEXT_VMERGE_VV(vmerge_vvm_d, int64_t, H8, clearq)
2091 
2092 #define GEN_VEXT_VMERGE_VX(NAME, ETYPE, H, CLEAR_FN)                 \
2093 void HELPER(NAME)(void *vd, void *v0, target_ulong s1,               \
2094                   void *vs2, CPURISCVState *env, uint32_t desc)      \
2095 {                                                                    \
2096     uint32_t mlen = vext_mlen(desc);                                 \
2097     uint32_t vl = env->vl;                                           \
2098     uint32_t esz = sizeof(ETYPE);                                    \
2099     uint32_t vlmax = vext_maxsz(desc) / esz;                         \
2100     uint32_t i;                                                      \
2101                                                                      \
2102     for (i = 0; i < vl; i++) {                                       \
2103         ETYPE s2 = *((ETYPE *)vs2 + H(i));                           \
2104         ETYPE d = (!vext_elem_mask(v0, mlen, i) ? s2 :               \
2105                    (ETYPE)(target_long)s1);                          \
2106         *((ETYPE *)vd + H(i)) = d;                                   \
2107     }                                                                \
2108     CLEAR_FN(vd, vl, vl * esz, vlmax * esz);                         \
2109 }
2110 
2111 GEN_VEXT_VMERGE_VX(vmerge_vxm_b, int8_t,  H1, clearb)
2112 GEN_VEXT_VMERGE_VX(vmerge_vxm_h, int16_t, H2, clearh)
2113 GEN_VEXT_VMERGE_VX(vmerge_vxm_w, int32_t, H4, clearl)
2114 GEN_VEXT_VMERGE_VX(vmerge_vxm_d, int64_t, H8, clearq)
2115 
2116 /*
2117  *** Vector Fixed-Point Arithmetic Instructions
2118  */
2119 
2120 /* Vector Single-Width Saturating Add and Subtract */
2121 
2122 /*
2123  * As fixed point instructions probably have round mode and saturation,
2124  * define common macros for fixed point here.
2125  */
2126 typedef void opivv2_rm_fn(void *vd, void *vs1, void *vs2, int i,
2127                           CPURISCVState *env, int vxrm);
2128 
2129 #define OPIVV2_RM(NAME, TD, T1, T2, TX1, TX2, HD, HS1, HS2, OP)     \
2130 static inline void                                                  \
2131 do_##NAME(void *vd, void *vs1, void *vs2, int i,                    \
2132           CPURISCVState *env, int vxrm)                             \
2133 {                                                                   \
2134     TX1 s1 = *((T1 *)vs1 + HS1(i));                                 \
2135     TX2 s2 = *((T2 *)vs2 + HS2(i));                                 \
2136     *((TD *)vd + HD(i)) = OP(env, vxrm, s2, s1);                    \
2137 }
2138 
2139 static inline void
2140 vext_vv_rm_1(void *vd, void *v0, void *vs1, void *vs2,
2141              CPURISCVState *env,
2142              uint32_t vl, uint32_t vm, uint32_t mlen, int vxrm,
2143              opivv2_rm_fn *fn)
2144 {
2145     for (uint32_t i = 0; i < vl; i++) {
2146         if (!vm && !vext_elem_mask(v0, mlen, i)) {
2147             continue;
2148         }
2149         fn(vd, vs1, vs2, i, env, vxrm);
2150     }
2151 }
2152 
2153 static inline void
2154 vext_vv_rm_2(void *vd, void *v0, void *vs1, void *vs2,
2155              CPURISCVState *env,
2156              uint32_t desc, uint32_t esz, uint32_t dsz,
2157              opivv2_rm_fn *fn, clear_fn *clearfn)
2158 {
2159     uint32_t vlmax = vext_maxsz(desc) / esz;
2160     uint32_t mlen = vext_mlen(desc);
2161     uint32_t vm = vext_vm(desc);
2162     uint32_t vl = env->vl;
2163 
2164     switch (env->vxrm) {
2165     case 0: /* rnu */
2166         vext_vv_rm_1(vd, v0, vs1, vs2,
2167                      env, vl, vm, mlen, 0, fn);
2168         break;
2169     case 1: /* rne */
2170         vext_vv_rm_1(vd, v0, vs1, vs2,
2171                      env, vl, vm, mlen, 1, fn);
2172         break;
2173     case 2: /* rdn */
2174         vext_vv_rm_1(vd, v0, vs1, vs2,
2175                      env, vl, vm, mlen, 2, fn);
2176         break;
2177     default: /* rod */
2178         vext_vv_rm_1(vd, v0, vs1, vs2,
2179                      env, vl, vm, mlen, 3, fn);
2180         break;
2181     }
2182 
2183     clearfn(vd, vl, vl * dsz,  vlmax * dsz);
2184 }
2185 
2186 /* generate helpers for fixed point instructions with OPIVV format */
2187 #define GEN_VEXT_VV_RM(NAME, ESZ, DSZ, CLEAR_FN)                \
2188 void HELPER(NAME)(void *vd, void *v0, void *vs1, void *vs2,     \
2189                   CPURISCVState *env, uint32_t desc)            \
2190 {                                                               \
2191     vext_vv_rm_2(vd, v0, vs1, vs2, env, desc, ESZ, DSZ,         \
2192                  do_##NAME, CLEAR_FN);                          \
2193 }
2194 
2195 static inline uint8_t saddu8(CPURISCVState *env, int vxrm, uint8_t a, uint8_t b)
2196 {
2197     uint8_t res = a + b;
2198     if (res < a) {
2199         res = UINT8_MAX;
2200         env->vxsat = 0x1;
2201     }
2202     return res;
2203 }
2204 
2205 static inline uint16_t saddu16(CPURISCVState *env, int vxrm, uint16_t a,
2206                                uint16_t b)
2207 {
2208     uint16_t res = a + b;
2209     if (res < a) {
2210         res = UINT16_MAX;
2211         env->vxsat = 0x1;
2212     }
2213     return res;
2214 }
2215 
2216 static inline uint32_t saddu32(CPURISCVState *env, int vxrm, uint32_t a,
2217                                uint32_t b)
2218 {
2219     uint32_t res = a + b;
2220     if (res < a) {
2221         res = UINT32_MAX;
2222         env->vxsat = 0x1;
2223     }
2224     return res;
2225 }
2226 
2227 static inline uint64_t saddu64(CPURISCVState *env, int vxrm, uint64_t a,
2228                                uint64_t b)
2229 {
2230     uint64_t res = a + b;
2231     if (res < a) {
2232         res = UINT64_MAX;
2233         env->vxsat = 0x1;
2234     }
2235     return res;
2236 }
2237 
2238 RVVCALL(OPIVV2_RM, vsaddu_vv_b, OP_UUU_B, H1, H1, H1, saddu8)
2239 RVVCALL(OPIVV2_RM, vsaddu_vv_h, OP_UUU_H, H2, H2, H2, saddu16)
2240 RVVCALL(OPIVV2_RM, vsaddu_vv_w, OP_UUU_W, H4, H4, H4, saddu32)
2241 RVVCALL(OPIVV2_RM, vsaddu_vv_d, OP_UUU_D, H8, H8, H8, saddu64)
2242 GEN_VEXT_VV_RM(vsaddu_vv_b, 1, 1, clearb)
2243 GEN_VEXT_VV_RM(vsaddu_vv_h, 2, 2, clearh)
2244 GEN_VEXT_VV_RM(vsaddu_vv_w, 4, 4, clearl)
2245 GEN_VEXT_VV_RM(vsaddu_vv_d, 8, 8, clearq)
2246 
2247 typedef void opivx2_rm_fn(void *vd, target_long s1, void *vs2, int i,
2248                           CPURISCVState *env, int vxrm);
2249 
2250 #define OPIVX2_RM(NAME, TD, T1, T2, TX1, TX2, HD, HS2, OP)          \
2251 static inline void                                                  \
2252 do_##NAME(void *vd, target_long s1, void *vs2, int i,               \
2253           CPURISCVState *env, int vxrm)                             \
2254 {                                                                   \
2255     TX2 s2 = *((T2 *)vs2 + HS2(i));                                 \
2256     *((TD *)vd + HD(i)) = OP(env, vxrm, s2, (TX1)(T1)s1);           \
2257 }
2258 
2259 static inline void
2260 vext_vx_rm_1(void *vd, void *v0, target_long s1, void *vs2,
2261              CPURISCVState *env,
2262              uint32_t vl, uint32_t vm, uint32_t mlen, int vxrm,
2263              opivx2_rm_fn *fn)
2264 {
2265     for (uint32_t i = 0; i < vl; i++) {
2266         if (!vm && !vext_elem_mask(v0, mlen, i)) {
2267             continue;
2268         }
2269         fn(vd, s1, vs2, i, env, vxrm);
2270     }
2271 }
2272 
2273 static inline void
2274 vext_vx_rm_2(void *vd, void *v0, target_long s1, void *vs2,
2275              CPURISCVState *env,
2276              uint32_t desc, uint32_t esz, uint32_t dsz,
2277              opivx2_rm_fn *fn, clear_fn *clearfn)
2278 {
2279     uint32_t vlmax = vext_maxsz(desc) / esz;
2280     uint32_t mlen = vext_mlen(desc);
2281     uint32_t vm = vext_vm(desc);
2282     uint32_t vl = env->vl;
2283 
2284     switch (env->vxrm) {
2285     case 0: /* rnu */
2286         vext_vx_rm_1(vd, v0, s1, vs2,
2287                      env, vl, vm, mlen, 0, fn);
2288         break;
2289     case 1: /* rne */
2290         vext_vx_rm_1(vd, v0, s1, vs2,
2291                      env, vl, vm, mlen, 1, fn);
2292         break;
2293     case 2: /* rdn */
2294         vext_vx_rm_1(vd, v0, s1, vs2,
2295                      env, vl, vm, mlen, 2, fn);
2296         break;
2297     default: /* rod */
2298         vext_vx_rm_1(vd, v0, s1, vs2,
2299                      env, vl, vm, mlen, 3, fn);
2300         break;
2301     }
2302 
2303     clearfn(vd, vl, vl * dsz,  vlmax * dsz);
2304 }
2305 
2306 /* generate helpers for fixed point instructions with OPIVX format */
2307 #define GEN_VEXT_VX_RM(NAME, ESZ, DSZ, CLEAR_FN)          \
2308 void HELPER(NAME)(void *vd, void *v0, target_ulong s1,    \
2309         void *vs2, CPURISCVState *env, uint32_t desc)     \
2310 {                                                         \
2311     vext_vx_rm_2(vd, v0, s1, vs2, env, desc, ESZ, DSZ,    \
2312                  do_##NAME, CLEAR_FN);                    \
2313 }
2314 
2315 RVVCALL(OPIVX2_RM, vsaddu_vx_b, OP_UUU_B, H1, H1, saddu8)
2316 RVVCALL(OPIVX2_RM, vsaddu_vx_h, OP_UUU_H, H2, H2, saddu16)
2317 RVVCALL(OPIVX2_RM, vsaddu_vx_w, OP_UUU_W, H4, H4, saddu32)
2318 RVVCALL(OPIVX2_RM, vsaddu_vx_d, OP_UUU_D, H8, H8, saddu64)
2319 GEN_VEXT_VX_RM(vsaddu_vx_b, 1, 1, clearb)
2320 GEN_VEXT_VX_RM(vsaddu_vx_h, 2, 2, clearh)
2321 GEN_VEXT_VX_RM(vsaddu_vx_w, 4, 4, clearl)
2322 GEN_VEXT_VX_RM(vsaddu_vx_d, 8, 8, clearq)
2323 
2324 static inline int8_t sadd8(CPURISCVState *env, int vxrm, int8_t a, int8_t b)
2325 {
2326     int8_t res = a + b;
2327     if ((res ^ a) & (res ^ b) & INT8_MIN) {
2328         res = a > 0 ? INT8_MAX : INT8_MIN;
2329         env->vxsat = 0x1;
2330     }
2331     return res;
2332 }
2333 
2334 static inline int16_t sadd16(CPURISCVState *env, int vxrm, int16_t a, int16_t b)
2335 {
2336     int16_t res = a + b;
2337     if ((res ^ a) & (res ^ b) & INT16_MIN) {
2338         res = a > 0 ? INT16_MAX : INT16_MIN;
2339         env->vxsat = 0x1;
2340     }
2341     return res;
2342 }
2343 
2344 static inline int32_t sadd32(CPURISCVState *env, int vxrm, int32_t a, int32_t b)
2345 {
2346     int32_t res = a + b;
2347     if ((res ^ a) & (res ^ b) & INT32_MIN) {
2348         res = a > 0 ? INT32_MAX : INT32_MIN;
2349         env->vxsat = 0x1;
2350     }
2351     return res;
2352 }
2353 
2354 static inline int64_t sadd64(CPURISCVState *env, int vxrm, int64_t a, int64_t b)
2355 {
2356     int64_t res = a + b;
2357     if ((res ^ a) & (res ^ b) & INT64_MIN) {
2358         res = a > 0 ? INT64_MAX : INT64_MIN;
2359         env->vxsat = 0x1;
2360     }
2361     return res;
2362 }
2363 
2364 RVVCALL(OPIVV2_RM, vsadd_vv_b, OP_SSS_B, H1, H1, H1, sadd8)
2365 RVVCALL(OPIVV2_RM, vsadd_vv_h, OP_SSS_H, H2, H2, H2, sadd16)
2366 RVVCALL(OPIVV2_RM, vsadd_vv_w, OP_SSS_W, H4, H4, H4, sadd32)
2367 RVVCALL(OPIVV2_RM, vsadd_vv_d, OP_SSS_D, H8, H8, H8, sadd64)
2368 GEN_VEXT_VV_RM(vsadd_vv_b, 1, 1, clearb)
2369 GEN_VEXT_VV_RM(vsadd_vv_h, 2, 2, clearh)
2370 GEN_VEXT_VV_RM(vsadd_vv_w, 4, 4, clearl)
2371 GEN_VEXT_VV_RM(vsadd_vv_d, 8, 8, clearq)
2372 
2373 RVVCALL(OPIVX2_RM, vsadd_vx_b, OP_SSS_B, H1, H1, sadd8)
2374 RVVCALL(OPIVX2_RM, vsadd_vx_h, OP_SSS_H, H2, H2, sadd16)
2375 RVVCALL(OPIVX2_RM, vsadd_vx_w, OP_SSS_W, H4, H4, sadd32)
2376 RVVCALL(OPIVX2_RM, vsadd_vx_d, OP_SSS_D, H8, H8, sadd64)
2377 GEN_VEXT_VX_RM(vsadd_vx_b, 1, 1, clearb)
2378 GEN_VEXT_VX_RM(vsadd_vx_h, 2, 2, clearh)
2379 GEN_VEXT_VX_RM(vsadd_vx_w, 4, 4, clearl)
2380 GEN_VEXT_VX_RM(vsadd_vx_d, 8, 8, clearq)
2381 
2382 static inline uint8_t ssubu8(CPURISCVState *env, int vxrm, uint8_t a, uint8_t b)
2383 {
2384     uint8_t res = a - b;
2385     if (res > a) {
2386         res = 0;
2387         env->vxsat = 0x1;
2388     }
2389     return res;
2390 }
2391 
2392 static inline uint16_t ssubu16(CPURISCVState *env, int vxrm, uint16_t a,
2393                                uint16_t b)
2394 {
2395     uint16_t res = a - b;
2396     if (res > a) {
2397         res = 0;
2398         env->vxsat = 0x1;
2399     }
2400     return res;
2401 }
2402 
2403 static inline uint32_t ssubu32(CPURISCVState *env, int vxrm, uint32_t a,
2404                                uint32_t b)
2405 {
2406     uint32_t res = a - b;
2407     if (res > a) {
2408         res = 0;
2409         env->vxsat = 0x1;
2410     }
2411     return res;
2412 }
2413 
2414 static inline uint64_t ssubu64(CPURISCVState *env, int vxrm, uint64_t a,
2415                                uint64_t b)
2416 {
2417     uint64_t res = a - b;
2418     if (res > a) {
2419         res = 0;
2420         env->vxsat = 0x1;
2421     }
2422     return res;
2423 }
2424 
2425 RVVCALL(OPIVV2_RM, vssubu_vv_b, OP_UUU_B, H1, H1, H1, ssubu8)
2426 RVVCALL(OPIVV2_RM, vssubu_vv_h, OP_UUU_H, H2, H2, H2, ssubu16)
2427 RVVCALL(OPIVV2_RM, vssubu_vv_w, OP_UUU_W, H4, H4, H4, ssubu32)
2428 RVVCALL(OPIVV2_RM, vssubu_vv_d, OP_UUU_D, H8, H8, H8, ssubu64)
2429 GEN_VEXT_VV_RM(vssubu_vv_b, 1, 1, clearb)
2430 GEN_VEXT_VV_RM(vssubu_vv_h, 2, 2, clearh)
2431 GEN_VEXT_VV_RM(vssubu_vv_w, 4, 4, clearl)
2432 GEN_VEXT_VV_RM(vssubu_vv_d, 8, 8, clearq)
2433 
2434 RVVCALL(OPIVX2_RM, vssubu_vx_b, OP_UUU_B, H1, H1, ssubu8)
2435 RVVCALL(OPIVX2_RM, vssubu_vx_h, OP_UUU_H, H2, H2, ssubu16)
2436 RVVCALL(OPIVX2_RM, vssubu_vx_w, OP_UUU_W, H4, H4, ssubu32)
2437 RVVCALL(OPIVX2_RM, vssubu_vx_d, OP_UUU_D, H8, H8, ssubu64)
2438 GEN_VEXT_VX_RM(vssubu_vx_b, 1, 1, clearb)
2439 GEN_VEXT_VX_RM(vssubu_vx_h, 2, 2, clearh)
2440 GEN_VEXT_VX_RM(vssubu_vx_w, 4, 4, clearl)
2441 GEN_VEXT_VX_RM(vssubu_vx_d, 8, 8, clearq)
2442 
2443 static inline int8_t ssub8(CPURISCVState *env, int vxrm, int8_t a, int8_t b)
2444 {
2445     int8_t res = a - b;
2446     if ((res ^ a) & (a ^ b) & INT8_MIN) {
2447         res = a > 0 ? INT8_MAX : INT8_MIN;
2448         env->vxsat = 0x1;
2449     }
2450     return res;
2451 }
2452 
2453 static inline int16_t ssub16(CPURISCVState *env, int vxrm, int16_t a, int16_t b)
2454 {
2455     int16_t res = a - b;
2456     if ((res ^ a) & (a ^ b) & INT16_MIN) {
2457         res = a > 0 ? INT16_MAX : INT16_MIN;
2458         env->vxsat = 0x1;
2459     }
2460     return res;
2461 }
2462 
2463 static inline int32_t ssub32(CPURISCVState *env, int vxrm, int32_t a, int32_t b)
2464 {
2465     int32_t res = a - b;
2466     if ((res ^ a) & (a ^ b) & INT32_MIN) {
2467         res = a > 0 ? INT32_MAX : INT32_MIN;
2468         env->vxsat = 0x1;
2469     }
2470     return res;
2471 }
2472 
2473 static inline int64_t ssub64(CPURISCVState *env, int vxrm, int64_t a, int64_t b)
2474 {
2475     int64_t res = a - b;
2476     if ((res ^ a) & (a ^ b) & INT64_MIN) {
2477         res = a > 0 ? INT64_MAX : INT64_MIN;
2478         env->vxsat = 0x1;
2479     }
2480     return res;
2481 }
2482 
2483 RVVCALL(OPIVV2_RM, vssub_vv_b, OP_SSS_B, H1, H1, H1, ssub8)
2484 RVVCALL(OPIVV2_RM, vssub_vv_h, OP_SSS_H, H2, H2, H2, ssub16)
2485 RVVCALL(OPIVV2_RM, vssub_vv_w, OP_SSS_W, H4, H4, H4, ssub32)
2486 RVVCALL(OPIVV2_RM, vssub_vv_d, OP_SSS_D, H8, H8, H8, ssub64)
2487 GEN_VEXT_VV_RM(vssub_vv_b, 1, 1, clearb)
2488 GEN_VEXT_VV_RM(vssub_vv_h, 2, 2, clearh)
2489 GEN_VEXT_VV_RM(vssub_vv_w, 4, 4, clearl)
2490 GEN_VEXT_VV_RM(vssub_vv_d, 8, 8, clearq)
2491 
2492 RVVCALL(OPIVX2_RM, vssub_vx_b, OP_SSS_B, H1, H1, ssub8)
2493 RVVCALL(OPIVX2_RM, vssub_vx_h, OP_SSS_H, H2, H2, ssub16)
2494 RVVCALL(OPIVX2_RM, vssub_vx_w, OP_SSS_W, H4, H4, ssub32)
2495 RVVCALL(OPIVX2_RM, vssub_vx_d, OP_SSS_D, H8, H8, ssub64)
2496 GEN_VEXT_VX_RM(vssub_vx_b, 1, 1, clearb)
2497 GEN_VEXT_VX_RM(vssub_vx_h, 2, 2, clearh)
2498 GEN_VEXT_VX_RM(vssub_vx_w, 4, 4, clearl)
2499 GEN_VEXT_VX_RM(vssub_vx_d, 8, 8, clearq)
2500 
2501 /* Vector Single-Width Averaging Add and Subtract */
2502 static inline uint8_t get_round(int vxrm, uint64_t v, uint8_t shift)
2503 {
2504     uint8_t d = extract64(v, shift, 1);
2505     uint8_t d1;
2506     uint64_t D1, D2;
2507 
2508     if (shift == 0 || shift > 64) {
2509         return 0;
2510     }
2511 
2512     d1 = extract64(v, shift - 1, 1);
2513     D1 = extract64(v, 0, shift);
2514     if (vxrm == 0) { /* round-to-nearest-up (add +0.5 LSB) */
2515         return d1;
2516     } else if (vxrm == 1) { /* round-to-nearest-even */
2517         if (shift > 1) {
2518             D2 = extract64(v, 0, shift - 1);
2519             return d1 & ((D2 != 0) | d);
2520         } else {
2521             return d1 & d;
2522         }
2523     } else if (vxrm == 3) { /* round-to-odd (OR bits into LSB, aka "jam") */
2524         return !d & (D1 != 0);
2525     }
2526     return 0; /* round-down (truncate) */
2527 }
2528 
2529 static inline int32_t aadd32(CPURISCVState *env, int vxrm, int32_t a, int32_t b)
2530 {
2531     int64_t res = (int64_t)a + b;
2532     uint8_t round = get_round(vxrm, res, 1);
2533 
2534     return (res >> 1) + round;
2535 }
2536 
2537 static inline int64_t aadd64(CPURISCVState *env, int vxrm, int64_t a, int64_t b)
2538 {
2539     int64_t res = a + b;
2540     uint8_t round = get_round(vxrm, res, 1);
2541     int64_t over = (res ^ a) & (res ^ b) & INT64_MIN;
2542 
2543     /* With signed overflow, bit 64 is inverse of bit 63. */
2544     return ((res >> 1) ^ over) + round;
2545 }
2546 
2547 RVVCALL(OPIVV2_RM, vaadd_vv_b, OP_SSS_B, H1, H1, H1, aadd32)
2548 RVVCALL(OPIVV2_RM, vaadd_vv_h, OP_SSS_H, H2, H2, H2, aadd32)
2549 RVVCALL(OPIVV2_RM, vaadd_vv_w, OP_SSS_W, H4, H4, H4, aadd32)
2550 RVVCALL(OPIVV2_RM, vaadd_vv_d, OP_SSS_D, H8, H8, H8, aadd64)
2551 GEN_VEXT_VV_RM(vaadd_vv_b, 1, 1, clearb)
2552 GEN_VEXT_VV_RM(vaadd_vv_h, 2, 2, clearh)
2553 GEN_VEXT_VV_RM(vaadd_vv_w, 4, 4, clearl)
2554 GEN_VEXT_VV_RM(vaadd_vv_d, 8, 8, clearq)
2555 
2556 RVVCALL(OPIVX2_RM, vaadd_vx_b, OP_SSS_B, H1, H1, aadd32)
2557 RVVCALL(OPIVX2_RM, vaadd_vx_h, OP_SSS_H, H2, H2, aadd32)
2558 RVVCALL(OPIVX2_RM, vaadd_vx_w, OP_SSS_W, H4, H4, aadd32)
2559 RVVCALL(OPIVX2_RM, vaadd_vx_d, OP_SSS_D, H8, H8, aadd64)
2560 GEN_VEXT_VX_RM(vaadd_vx_b, 1, 1, clearb)
2561 GEN_VEXT_VX_RM(vaadd_vx_h, 2, 2, clearh)
2562 GEN_VEXT_VX_RM(vaadd_vx_w, 4, 4, clearl)
2563 GEN_VEXT_VX_RM(vaadd_vx_d, 8, 8, clearq)
2564 
2565 static inline int32_t asub32(CPURISCVState *env, int vxrm, int32_t a, int32_t b)
2566 {
2567     int64_t res = (int64_t)a - b;
2568     uint8_t round = get_round(vxrm, res, 1);
2569 
2570     return (res >> 1) + round;
2571 }
2572 
2573 static inline int64_t asub64(CPURISCVState *env, int vxrm, int64_t a, int64_t b)
2574 {
2575     int64_t res = (int64_t)a - b;
2576     uint8_t round = get_round(vxrm, res, 1);
2577     int64_t over = (res ^ a) & (a ^ b) & INT64_MIN;
2578 
2579     /* With signed overflow, bit 64 is inverse of bit 63. */
2580     return ((res >> 1) ^ over) + round;
2581 }
2582 
2583 RVVCALL(OPIVV2_RM, vasub_vv_b, OP_SSS_B, H1, H1, H1, asub32)
2584 RVVCALL(OPIVV2_RM, vasub_vv_h, OP_SSS_H, H2, H2, H2, asub32)
2585 RVVCALL(OPIVV2_RM, vasub_vv_w, OP_SSS_W, H4, H4, H4, asub32)
2586 RVVCALL(OPIVV2_RM, vasub_vv_d, OP_SSS_D, H8, H8, H8, asub64)
2587 GEN_VEXT_VV_RM(vasub_vv_b, 1, 1, clearb)
2588 GEN_VEXT_VV_RM(vasub_vv_h, 2, 2, clearh)
2589 GEN_VEXT_VV_RM(vasub_vv_w, 4, 4, clearl)
2590 GEN_VEXT_VV_RM(vasub_vv_d, 8, 8, clearq)
2591 
2592 RVVCALL(OPIVX2_RM, vasub_vx_b, OP_SSS_B, H1, H1, asub32)
2593 RVVCALL(OPIVX2_RM, vasub_vx_h, OP_SSS_H, H2, H2, asub32)
2594 RVVCALL(OPIVX2_RM, vasub_vx_w, OP_SSS_W, H4, H4, asub32)
2595 RVVCALL(OPIVX2_RM, vasub_vx_d, OP_SSS_D, H8, H8, asub64)
2596 GEN_VEXT_VX_RM(vasub_vx_b, 1, 1, clearb)
2597 GEN_VEXT_VX_RM(vasub_vx_h, 2, 2, clearh)
2598 GEN_VEXT_VX_RM(vasub_vx_w, 4, 4, clearl)
2599 GEN_VEXT_VX_RM(vasub_vx_d, 8, 8, clearq)
2600