xref: /openbmc/qemu/target/riscv/vector_helper.c (revision 139c1837)
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 "fpu/softfloat.h"
25 #include "tcg/tcg-gvec-desc.h"
26 #include "internals.h"
27 #include <math.h>
28 
29 target_ulong HELPER(vsetvl)(CPURISCVState *env, target_ulong s1,
30                             target_ulong s2)
31 {
32     int vlmax, vl;
33     RISCVCPU *cpu = env_archcpu(env);
34     uint16_t sew = 8 << FIELD_EX64(s2, VTYPE, VSEW);
35     uint8_t ediv = FIELD_EX64(s2, VTYPE, VEDIV);
36     bool vill = FIELD_EX64(s2, VTYPE, VILL);
37     target_ulong reserved = FIELD_EX64(s2, VTYPE, RESERVED);
38 
39     if ((sew > cpu->cfg.elen) || vill || (ediv != 0) || (reserved != 0)) {
40         /* only set vill bit. */
41         env->vtype = FIELD_DP64(0, VTYPE, VILL, 1);
42         env->vl = 0;
43         env->vstart = 0;
44         return 0;
45     }
46 
47     vlmax = vext_get_vlmax(cpu, s2);
48     if (s1 <= vlmax) {
49         vl = s1;
50     } else {
51         vl = vlmax;
52     }
53     env->vl = vl;
54     env->vtype = s2;
55     env->vstart = 0;
56     return vl;
57 }
58 
59 /*
60  * Note that vector data is stored in host-endian 64-bit chunks,
61  * so addressing units smaller than that needs a host-endian fixup.
62  */
63 #ifdef HOST_WORDS_BIGENDIAN
64 #define H1(x)   ((x) ^ 7)
65 #define H1_2(x) ((x) ^ 6)
66 #define H1_4(x) ((x) ^ 4)
67 #define H2(x)   ((x) ^ 3)
68 #define H4(x)   ((x) ^ 1)
69 #define H8(x)   ((x))
70 #else
71 #define H1(x)   (x)
72 #define H1_2(x) (x)
73 #define H1_4(x) (x)
74 #define H2(x)   (x)
75 #define H4(x)   (x)
76 #define H8(x)   (x)
77 #endif
78 
79 static inline uint32_t vext_nf(uint32_t desc)
80 {
81     return FIELD_EX32(simd_data(desc), VDATA, NF);
82 }
83 
84 static inline uint32_t vext_mlen(uint32_t desc)
85 {
86     return FIELD_EX32(simd_data(desc), VDATA, MLEN);
87 }
88 
89 static inline uint32_t vext_vm(uint32_t desc)
90 {
91     return FIELD_EX32(simd_data(desc), VDATA, VM);
92 }
93 
94 static inline uint32_t vext_lmul(uint32_t desc)
95 {
96     return FIELD_EX32(simd_data(desc), VDATA, LMUL);
97 }
98 
99 static uint32_t vext_wd(uint32_t desc)
100 {
101     return (simd_data(desc) >> 11) & 0x1;
102 }
103 
104 /*
105  * Get vector group length in bytes. Its range is [64, 2048].
106  *
107  * As simd_desc support at most 256, the max vlen is 512 bits.
108  * So vlen in bytes is encoded as maxsz.
109  */
110 static inline uint32_t vext_maxsz(uint32_t desc)
111 {
112     return simd_maxsz(desc) << vext_lmul(desc);
113 }
114 
115 /*
116  * This function checks watchpoint before real load operation.
117  *
118  * In softmmu mode, the TLB API probe_access is enough for watchpoint check.
119  * In user mode, there is no watchpoint support now.
120  *
121  * It will trigger an exception if there is no mapping in TLB
122  * and page table walk can't fill the TLB entry. Then the guest
123  * software can return here after process the exception or never return.
124  */
125 static void probe_pages(CPURISCVState *env, target_ulong addr,
126                         target_ulong len, uintptr_t ra,
127                         MMUAccessType access_type)
128 {
129     target_ulong pagelen = -(addr | TARGET_PAGE_MASK);
130     target_ulong curlen = MIN(pagelen, len);
131 
132     probe_access(env, addr, curlen, access_type,
133                  cpu_mmu_index(env, false), ra);
134     if (len > curlen) {
135         addr += curlen;
136         curlen = len - curlen;
137         probe_access(env, addr, curlen, access_type,
138                      cpu_mmu_index(env, false), ra);
139     }
140 }
141 
142 #ifdef HOST_WORDS_BIGENDIAN
143 static void vext_clear(void *tail, uint32_t cnt, uint32_t tot)
144 {
145     /*
146      * Split the remaining range to two parts.
147      * The first part is in the last uint64_t unit.
148      * The second part start from the next uint64_t unit.
149      */
150     int part1 = 0, part2 = tot - cnt;
151     if (cnt % 8) {
152         part1 = 8 - (cnt % 8);
153         part2 = tot - cnt - part1;
154         memset(QEMU_ALIGN_PTR_DOWN(tail, 8), 0, part1);
155         memset(QEMU_ALIGN_PTR_UP(tail, 8), 0, part2);
156     } else {
157         memset(tail, 0, part2);
158     }
159 }
160 #else
161 static void vext_clear(void *tail, uint32_t cnt, uint32_t tot)
162 {
163     memset(tail, 0, tot - cnt);
164 }
165 #endif
166 
167 static void clearb(void *vd, uint32_t idx, uint32_t cnt, uint32_t tot)
168 {
169     int8_t *cur = ((int8_t *)vd + H1(idx));
170     vext_clear(cur, cnt, tot);
171 }
172 
173 static void clearh(void *vd, uint32_t idx, uint32_t cnt, uint32_t tot)
174 {
175     int16_t *cur = ((int16_t *)vd + H2(idx));
176     vext_clear(cur, cnt, tot);
177 }
178 
179 static void clearl(void *vd, uint32_t idx, uint32_t cnt, uint32_t tot)
180 {
181     int32_t *cur = ((int32_t *)vd + H4(idx));
182     vext_clear(cur, cnt, tot);
183 }
184 
185 static void clearq(void *vd, uint32_t idx, uint32_t cnt, uint32_t tot)
186 {
187     int64_t *cur = (int64_t *)vd + idx;
188     vext_clear(cur, cnt, tot);
189 }
190 
191 static inline void vext_set_elem_mask(void *v0, int mlen, int index,
192         uint8_t value)
193 {
194     int idx = (index * mlen) / 64;
195     int pos = (index * mlen) % 64;
196     uint64_t old = ((uint64_t *)v0)[idx];
197     ((uint64_t *)v0)[idx] = deposit64(old, pos, mlen, value);
198 }
199 
200 static inline int vext_elem_mask(void *v0, int mlen, int index)
201 {
202     int idx = (index * mlen) / 64;
203     int pos = (index * mlen) % 64;
204     return (((uint64_t *)v0)[idx] >> pos) & 1;
205 }
206 
207 /* elements operations for load and store */
208 typedef void vext_ldst_elem_fn(CPURISCVState *env, target_ulong addr,
209                                uint32_t idx, void *vd, uintptr_t retaddr);
210 typedef void clear_fn(void *vd, uint32_t idx, uint32_t cnt, uint32_t tot);
211 
212 #define GEN_VEXT_LD_ELEM(NAME, MTYPE, ETYPE, H, LDSUF)     \
213 static void NAME(CPURISCVState *env, abi_ptr addr,         \
214                  uint32_t idx, void *vd, uintptr_t retaddr)\
215 {                                                          \
216     MTYPE data;                                            \
217     ETYPE *cur = ((ETYPE *)vd + H(idx));                   \
218     data = cpu_##LDSUF##_data_ra(env, addr, retaddr);      \
219     *cur = data;                                           \
220 }                                                          \
221 
222 GEN_VEXT_LD_ELEM(ldb_b, int8_t,  int8_t,  H1, ldsb)
223 GEN_VEXT_LD_ELEM(ldb_h, int8_t,  int16_t, H2, ldsb)
224 GEN_VEXT_LD_ELEM(ldb_w, int8_t,  int32_t, H4, ldsb)
225 GEN_VEXT_LD_ELEM(ldb_d, int8_t,  int64_t, H8, ldsb)
226 GEN_VEXT_LD_ELEM(ldh_h, int16_t, int16_t, H2, ldsw)
227 GEN_VEXT_LD_ELEM(ldh_w, int16_t, int32_t, H4, ldsw)
228 GEN_VEXT_LD_ELEM(ldh_d, int16_t, int64_t, H8, ldsw)
229 GEN_VEXT_LD_ELEM(ldw_w, int32_t, int32_t, H4, ldl)
230 GEN_VEXT_LD_ELEM(ldw_d, int32_t, int64_t, H8, ldl)
231 GEN_VEXT_LD_ELEM(lde_b, int8_t,  int8_t,  H1, ldsb)
232 GEN_VEXT_LD_ELEM(lde_h, int16_t, int16_t, H2, ldsw)
233 GEN_VEXT_LD_ELEM(lde_w, int32_t, int32_t, H4, ldl)
234 GEN_VEXT_LD_ELEM(lde_d, int64_t, int64_t, H8, ldq)
235 GEN_VEXT_LD_ELEM(ldbu_b, uint8_t,  uint8_t,  H1, ldub)
236 GEN_VEXT_LD_ELEM(ldbu_h, uint8_t,  uint16_t, H2, ldub)
237 GEN_VEXT_LD_ELEM(ldbu_w, uint8_t,  uint32_t, H4, ldub)
238 GEN_VEXT_LD_ELEM(ldbu_d, uint8_t,  uint64_t, H8, ldub)
239 GEN_VEXT_LD_ELEM(ldhu_h, uint16_t, uint16_t, H2, lduw)
240 GEN_VEXT_LD_ELEM(ldhu_w, uint16_t, uint32_t, H4, lduw)
241 GEN_VEXT_LD_ELEM(ldhu_d, uint16_t, uint64_t, H8, lduw)
242 GEN_VEXT_LD_ELEM(ldwu_w, uint32_t, uint32_t, H4, ldl)
243 GEN_VEXT_LD_ELEM(ldwu_d, uint32_t, uint64_t, H8, ldl)
244 
245 #define GEN_VEXT_ST_ELEM(NAME, ETYPE, H, STSUF)            \
246 static void NAME(CPURISCVState *env, abi_ptr addr,         \
247                  uint32_t idx, void *vd, uintptr_t retaddr)\
248 {                                                          \
249     ETYPE data = *((ETYPE *)vd + H(idx));                  \
250     cpu_##STSUF##_data_ra(env, addr, data, retaddr);       \
251 }
252 
253 GEN_VEXT_ST_ELEM(stb_b, int8_t,  H1, stb)
254 GEN_VEXT_ST_ELEM(stb_h, int16_t, H2, stb)
255 GEN_VEXT_ST_ELEM(stb_w, int32_t, H4, stb)
256 GEN_VEXT_ST_ELEM(stb_d, int64_t, H8, stb)
257 GEN_VEXT_ST_ELEM(sth_h, int16_t, H2, stw)
258 GEN_VEXT_ST_ELEM(sth_w, int32_t, H4, stw)
259 GEN_VEXT_ST_ELEM(sth_d, int64_t, H8, stw)
260 GEN_VEXT_ST_ELEM(stw_w, int32_t, H4, stl)
261 GEN_VEXT_ST_ELEM(stw_d, int64_t, H8, stl)
262 GEN_VEXT_ST_ELEM(ste_b, int8_t,  H1, stb)
263 GEN_VEXT_ST_ELEM(ste_h, int16_t, H2, stw)
264 GEN_VEXT_ST_ELEM(ste_w, int32_t, H4, stl)
265 GEN_VEXT_ST_ELEM(ste_d, int64_t, H8, stq)
266 
267 /*
268  *** stride: access vector element from strided memory
269  */
270 static void
271 vext_ldst_stride(void *vd, void *v0, target_ulong base,
272                  target_ulong stride, CPURISCVState *env,
273                  uint32_t desc, uint32_t vm,
274                  vext_ldst_elem_fn *ldst_elem, clear_fn *clear_elem,
275                  uint32_t esz, uint32_t msz, uintptr_t ra,
276                  MMUAccessType access_type)
277 {
278     uint32_t i, k;
279     uint32_t nf = vext_nf(desc);
280     uint32_t mlen = vext_mlen(desc);
281     uint32_t vlmax = vext_maxsz(desc) / esz;
282 
283     /* probe every access*/
284     for (i = 0; i < env->vl; i++) {
285         if (!vm && !vext_elem_mask(v0, mlen, i)) {
286             continue;
287         }
288         probe_pages(env, base + stride * i, nf * msz, ra, access_type);
289     }
290     /* do real access */
291     for (i = 0; i < env->vl; i++) {
292         k = 0;
293         if (!vm && !vext_elem_mask(v0, mlen, i)) {
294             continue;
295         }
296         while (k < nf) {
297             target_ulong addr = base + stride * i + k * msz;
298             ldst_elem(env, addr, i + k * vlmax, vd, ra);
299             k++;
300         }
301     }
302     /* clear tail elements */
303     if (clear_elem) {
304         for (k = 0; k < nf; k++) {
305             clear_elem(vd, env->vl + k * vlmax, env->vl * esz, vlmax * esz);
306         }
307     }
308 }
309 
310 #define GEN_VEXT_LD_STRIDE(NAME, MTYPE, ETYPE, LOAD_FN, CLEAR_FN)       \
311 void HELPER(NAME)(void *vd, void * v0, target_ulong base,               \
312                   target_ulong stride, CPURISCVState *env,              \
313                   uint32_t desc)                                        \
314 {                                                                       \
315     uint32_t vm = vext_vm(desc);                                        \
316     vext_ldst_stride(vd, v0, base, stride, env, desc, vm, LOAD_FN,      \
317                      CLEAR_FN, sizeof(ETYPE), sizeof(MTYPE),            \
318                      GETPC(), MMU_DATA_LOAD);                           \
319 }
320 
321 GEN_VEXT_LD_STRIDE(vlsb_v_b,  int8_t,   int8_t,   ldb_b,  clearb)
322 GEN_VEXT_LD_STRIDE(vlsb_v_h,  int8_t,   int16_t,  ldb_h,  clearh)
323 GEN_VEXT_LD_STRIDE(vlsb_v_w,  int8_t,   int32_t,  ldb_w,  clearl)
324 GEN_VEXT_LD_STRIDE(vlsb_v_d,  int8_t,   int64_t,  ldb_d,  clearq)
325 GEN_VEXT_LD_STRIDE(vlsh_v_h,  int16_t,  int16_t,  ldh_h,  clearh)
326 GEN_VEXT_LD_STRIDE(vlsh_v_w,  int16_t,  int32_t,  ldh_w,  clearl)
327 GEN_VEXT_LD_STRIDE(vlsh_v_d,  int16_t,  int64_t,  ldh_d,  clearq)
328 GEN_VEXT_LD_STRIDE(vlsw_v_w,  int32_t,  int32_t,  ldw_w,  clearl)
329 GEN_VEXT_LD_STRIDE(vlsw_v_d,  int32_t,  int64_t,  ldw_d,  clearq)
330 GEN_VEXT_LD_STRIDE(vlse_v_b,  int8_t,   int8_t,   lde_b,  clearb)
331 GEN_VEXT_LD_STRIDE(vlse_v_h,  int16_t,  int16_t,  lde_h,  clearh)
332 GEN_VEXT_LD_STRIDE(vlse_v_w,  int32_t,  int32_t,  lde_w,  clearl)
333 GEN_VEXT_LD_STRIDE(vlse_v_d,  int64_t,  int64_t,  lde_d,  clearq)
334 GEN_VEXT_LD_STRIDE(vlsbu_v_b, uint8_t,  uint8_t,  ldbu_b, clearb)
335 GEN_VEXT_LD_STRIDE(vlsbu_v_h, uint8_t,  uint16_t, ldbu_h, clearh)
336 GEN_VEXT_LD_STRIDE(vlsbu_v_w, uint8_t,  uint32_t, ldbu_w, clearl)
337 GEN_VEXT_LD_STRIDE(vlsbu_v_d, uint8_t,  uint64_t, ldbu_d, clearq)
338 GEN_VEXT_LD_STRIDE(vlshu_v_h, uint16_t, uint16_t, ldhu_h, clearh)
339 GEN_VEXT_LD_STRIDE(vlshu_v_w, uint16_t, uint32_t, ldhu_w, clearl)
340 GEN_VEXT_LD_STRIDE(vlshu_v_d, uint16_t, uint64_t, ldhu_d, clearq)
341 GEN_VEXT_LD_STRIDE(vlswu_v_w, uint32_t, uint32_t, ldwu_w, clearl)
342 GEN_VEXT_LD_STRIDE(vlswu_v_d, uint32_t, uint64_t, ldwu_d, clearq)
343 
344 #define GEN_VEXT_ST_STRIDE(NAME, MTYPE, ETYPE, STORE_FN)                \
345 void HELPER(NAME)(void *vd, void *v0, target_ulong base,                \
346                   target_ulong stride, CPURISCVState *env,              \
347                   uint32_t desc)                                        \
348 {                                                                       \
349     uint32_t vm = vext_vm(desc);                                        \
350     vext_ldst_stride(vd, v0, base, stride, env, desc, vm, STORE_FN,     \
351                      NULL, sizeof(ETYPE), sizeof(MTYPE),                \
352                      GETPC(), MMU_DATA_STORE);                          \
353 }
354 
355 GEN_VEXT_ST_STRIDE(vssb_v_b, int8_t,  int8_t,  stb_b)
356 GEN_VEXT_ST_STRIDE(vssb_v_h, int8_t,  int16_t, stb_h)
357 GEN_VEXT_ST_STRIDE(vssb_v_w, int8_t,  int32_t, stb_w)
358 GEN_VEXT_ST_STRIDE(vssb_v_d, int8_t,  int64_t, stb_d)
359 GEN_VEXT_ST_STRIDE(vssh_v_h, int16_t, int16_t, sth_h)
360 GEN_VEXT_ST_STRIDE(vssh_v_w, int16_t, int32_t, sth_w)
361 GEN_VEXT_ST_STRIDE(vssh_v_d, int16_t, int64_t, sth_d)
362 GEN_VEXT_ST_STRIDE(vssw_v_w, int32_t, int32_t, stw_w)
363 GEN_VEXT_ST_STRIDE(vssw_v_d, int32_t, int64_t, stw_d)
364 GEN_VEXT_ST_STRIDE(vsse_v_b, int8_t,  int8_t,  ste_b)
365 GEN_VEXT_ST_STRIDE(vsse_v_h, int16_t, int16_t, ste_h)
366 GEN_VEXT_ST_STRIDE(vsse_v_w, int32_t, int32_t, ste_w)
367 GEN_VEXT_ST_STRIDE(vsse_v_d, int64_t, int64_t, ste_d)
368 
369 /*
370  *** unit-stride: access elements stored contiguously in memory
371  */
372 
373 /* unmasked unit-stride load and store operation*/
374 static void
375 vext_ldst_us(void *vd, target_ulong base, CPURISCVState *env, uint32_t desc,
376              vext_ldst_elem_fn *ldst_elem, clear_fn *clear_elem,
377              uint32_t esz, uint32_t msz, uintptr_t ra,
378              MMUAccessType access_type)
379 {
380     uint32_t i, k;
381     uint32_t nf = vext_nf(desc);
382     uint32_t vlmax = vext_maxsz(desc) / esz;
383 
384     /* probe every access */
385     probe_pages(env, base, env->vl * nf * msz, ra, access_type);
386     /* load bytes from guest memory */
387     for (i = 0; i < env->vl; i++) {
388         k = 0;
389         while (k < nf) {
390             target_ulong addr = base + (i * nf + k) * msz;
391             ldst_elem(env, addr, i + k * vlmax, vd, ra);
392             k++;
393         }
394     }
395     /* clear tail elements */
396     if (clear_elem) {
397         for (k = 0; k < nf; k++) {
398             clear_elem(vd, env->vl + k * vlmax, env->vl * esz, vlmax * esz);
399         }
400     }
401 }
402 
403 /*
404  * masked unit-stride load and store operation will be a special case of stride,
405  * stride = NF * sizeof (MTYPE)
406  */
407 
408 #define GEN_VEXT_LD_US(NAME, MTYPE, ETYPE, LOAD_FN, CLEAR_FN)           \
409 void HELPER(NAME##_mask)(void *vd, void *v0, target_ulong base,         \
410                          CPURISCVState *env, uint32_t desc)             \
411 {                                                                       \
412     uint32_t stride = vext_nf(desc) * sizeof(MTYPE);                    \
413     vext_ldst_stride(vd, v0, base, stride, env, desc, false, LOAD_FN,   \
414                      CLEAR_FN, sizeof(ETYPE), sizeof(MTYPE),            \
415                      GETPC(), MMU_DATA_LOAD);                           \
416 }                                                                       \
417                                                                         \
418 void HELPER(NAME)(void *vd, void *v0, target_ulong base,                \
419                   CPURISCVState *env, uint32_t desc)                    \
420 {                                                                       \
421     vext_ldst_us(vd, base, env, desc, LOAD_FN, CLEAR_FN,                \
422                  sizeof(ETYPE), sizeof(MTYPE), GETPC(), MMU_DATA_LOAD); \
423 }
424 
425 GEN_VEXT_LD_US(vlb_v_b,  int8_t,   int8_t,   ldb_b,  clearb)
426 GEN_VEXT_LD_US(vlb_v_h,  int8_t,   int16_t,  ldb_h,  clearh)
427 GEN_VEXT_LD_US(vlb_v_w,  int8_t,   int32_t,  ldb_w,  clearl)
428 GEN_VEXT_LD_US(vlb_v_d,  int8_t,   int64_t,  ldb_d,  clearq)
429 GEN_VEXT_LD_US(vlh_v_h,  int16_t,  int16_t,  ldh_h,  clearh)
430 GEN_VEXT_LD_US(vlh_v_w,  int16_t,  int32_t,  ldh_w,  clearl)
431 GEN_VEXT_LD_US(vlh_v_d,  int16_t,  int64_t,  ldh_d,  clearq)
432 GEN_VEXT_LD_US(vlw_v_w,  int32_t,  int32_t,  ldw_w,  clearl)
433 GEN_VEXT_LD_US(vlw_v_d,  int32_t,  int64_t,  ldw_d,  clearq)
434 GEN_VEXT_LD_US(vle_v_b,  int8_t,   int8_t,   lde_b,  clearb)
435 GEN_VEXT_LD_US(vle_v_h,  int16_t,  int16_t,  lde_h,  clearh)
436 GEN_VEXT_LD_US(vle_v_w,  int32_t,  int32_t,  lde_w,  clearl)
437 GEN_VEXT_LD_US(vle_v_d,  int64_t,  int64_t,  lde_d,  clearq)
438 GEN_VEXT_LD_US(vlbu_v_b, uint8_t,  uint8_t,  ldbu_b, clearb)
439 GEN_VEXT_LD_US(vlbu_v_h, uint8_t,  uint16_t, ldbu_h, clearh)
440 GEN_VEXT_LD_US(vlbu_v_w, uint8_t,  uint32_t, ldbu_w, clearl)
441 GEN_VEXT_LD_US(vlbu_v_d, uint8_t,  uint64_t, ldbu_d, clearq)
442 GEN_VEXT_LD_US(vlhu_v_h, uint16_t, uint16_t, ldhu_h, clearh)
443 GEN_VEXT_LD_US(vlhu_v_w, uint16_t, uint32_t, ldhu_w, clearl)
444 GEN_VEXT_LD_US(vlhu_v_d, uint16_t, uint64_t, ldhu_d, clearq)
445 GEN_VEXT_LD_US(vlwu_v_w, uint32_t, uint32_t, ldwu_w, clearl)
446 GEN_VEXT_LD_US(vlwu_v_d, uint32_t, uint64_t, ldwu_d, clearq)
447 
448 #define GEN_VEXT_ST_US(NAME, MTYPE, ETYPE, STORE_FN)                    \
449 void HELPER(NAME##_mask)(void *vd, void *v0, target_ulong base,         \
450                          CPURISCVState *env, uint32_t desc)             \
451 {                                                                       \
452     uint32_t stride = vext_nf(desc) * sizeof(MTYPE);                    \
453     vext_ldst_stride(vd, v0, base, stride, env, desc, false, STORE_FN,  \
454                      NULL, sizeof(ETYPE), sizeof(MTYPE),                \
455                      GETPC(), MMU_DATA_STORE);                          \
456 }                                                                       \
457                                                                         \
458 void HELPER(NAME)(void *vd, void *v0, target_ulong base,                \
459                   CPURISCVState *env, uint32_t desc)                    \
460 {                                                                       \
461     vext_ldst_us(vd, base, env, desc, STORE_FN, NULL,                   \
462                  sizeof(ETYPE), sizeof(MTYPE), GETPC(), MMU_DATA_STORE);\
463 }
464 
465 GEN_VEXT_ST_US(vsb_v_b, int8_t,  int8_t , stb_b)
466 GEN_VEXT_ST_US(vsb_v_h, int8_t,  int16_t, stb_h)
467 GEN_VEXT_ST_US(vsb_v_w, int8_t,  int32_t, stb_w)
468 GEN_VEXT_ST_US(vsb_v_d, int8_t,  int64_t, stb_d)
469 GEN_VEXT_ST_US(vsh_v_h, int16_t, int16_t, sth_h)
470 GEN_VEXT_ST_US(vsh_v_w, int16_t, int32_t, sth_w)
471 GEN_VEXT_ST_US(vsh_v_d, int16_t, int64_t, sth_d)
472 GEN_VEXT_ST_US(vsw_v_w, int32_t, int32_t, stw_w)
473 GEN_VEXT_ST_US(vsw_v_d, int32_t, int64_t, stw_d)
474 GEN_VEXT_ST_US(vse_v_b, int8_t,  int8_t , ste_b)
475 GEN_VEXT_ST_US(vse_v_h, int16_t, int16_t, ste_h)
476 GEN_VEXT_ST_US(vse_v_w, int32_t, int32_t, ste_w)
477 GEN_VEXT_ST_US(vse_v_d, int64_t, int64_t, ste_d)
478 
479 /*
480  *** index: access vector element from indexed memory
481  */
482 typedef target_ulong vext_get_index_addr(target_ulong base,
483         uint32_t idx, void *vs2);
484 
485 #define GEN_VEXT_GET_INDEX_ADDR(NAME, ETYPE, H)        \
486 static target_ulong NAME(target_ulong base,            \
487                          uint32_t idx, void *vs2)      \
488 {                                                      \
489     return (base + *((ETYPE *)vs2 + H(idx)));          \
490 }
491 
492 GEN_VEXT_GET_INDEX_ADDR(idx_b, int8_t,  H1)
493 GEN_VEXT_GET_INDEX_ADDR(idx_h, int16_t, H2)
494 GEN_VEXT_GET_INDEX_ADDR(idx_w, int32_t, H4)
495 GEN_VEXT_GET_INDEX_ADDR(idx_d, int64_t, H8)
496 
497 static inline void
498 vext_ldst_index(void *vd, void *v0, target_ulong base,
499                 void *vs2, CPURISCVState *env, uint32_t desc,
500                 vext_get_index_addr get_index_addr,
501                 vext_ldst_elem_fn *ldst_elem,
502                 clear_fn *clear_elem,
503                 uint32_t esz, uint32_t msz, uintptr_t ra,
504                 MMUAccessType access_type)
505 {
506     uint32_t i, k;
507     uint32_t nf = vext_nf(desc);
508     uint32_t vm = vext_vm(desc);
509     uint32_t mlen = vext_mlen(desc);
510     uint32_t vlmax = vext_maxsz(desc) / esz;
511 
512     /* probe every access*/
513     for (i = 0; i < env->vl; i++) {
514         if (!vm && !vext_elem_mask(v0, mlen, i)) {
515             continue;
516         }
517         probe_pages(env, get_index_addr(base, i, vs2), nf * msz, ra,
518                     access_type);
519     }
520     /* load bytes from guest memory */
521     for (i = 0; i < env->vl; i++) {
522         k = 0;
523         if (!vm && !vext_elem_mask(v0, mlen, i)) {
524             continue;
525         }
526         while (k < nf) {
527             abi_ptr addr = get_index_addr(base, i, vs2) + k * msz;
528             ldst_elem(env, addr, i + k * vlmax, vd, ra);
529             k++;
530         }
531     }
532     /* clear tail elements */
533     if (clear_elem) {
534         for (k = 0; k < nf; k++) {
535             clear_elem(vd, env->vl + k * vlmax, env->vl * esz, vlmax * esz);
536         }
537     }
538 }
539 
540 #define GEN_VEXT_LD_INDEX(NAME, MTYPE, ETYPE, INDEX_FN, LOAD_FN, CLEAR_FN) \
541 void HELPER(NAME)(void *vd, void *v0, target_ulong base,                   \
542                   void *vs2, CPURISCVState *env, uint32_t desc)            \
543 {                                                                          \
544     vext_ldst_index(vd, v0, base, vs2, env, desc, INDEX_FN,                \
545                     LOAD_FN, CLEAR_FN, sizeof(ETYPE), sizeof(MTYPE),       \
546                     GETPC(), MMU_DATA_LOAD);                               \
547 }
548 
549 GEN_VEXT_LD_INDEX(vlxb_v_b,  int8_t,   int8_t,   idx_b, ldb_b,  clearb)
550 GEN_VEXT_LD_INDEX(vlxb_v_h,  int8_t,   int16_t,  idx_h, ldb_h,  clearh)
551 GEN_VEXT_LD_INDEX(vlxb_v_w,  int8_t,   int32_t,  idx_w, ldb_w,  clearl)
552 GEN_VEXT_LD_INDEX(vlxb_v_d,  int8_t,   int64_t,  idx_d, ldb_d,  clearq)
553 GEN_VEXT_LD_INDEX(vlxh_v_h,  int16_t,  int16_t,  idx_h, ldh_h,  clearh)
554 GEN_VEXT_LD_INDEX(vlxh_v_w,  int16_t,  int32_t,  idx_w, ldh_w,  clearl)
555 GEN_VEXT_LD_INDEX(vlxh_v_d,  int16_t,  int64_t,  idx_d, ldh_d,  clearq)
556 GEN_VEXT_LD_INDEX(vlxw_v_w,  int32_t,  int32_t,  idx_w, ldw_w,  clearl)
557 GEN_VEXT_LD_INDEX(vlxw_v_d,  int32_t,  int64_t,  idx_d, ldw_d,  clearq)
558 GEN_VEXT_LD_INDEX(vlxe_v_b,  int8_t,   int8_t,   idx_b, lde_b,  clearb)
559 GEN_VEXT_LD_INDEX(vlxe_v_h,  int16_t,  int16_t,  idx_h, lde_h,  clearh)
560 GEN_VEXT_LD_INDEX(vlxe_v_w,  int32_t,  int32_t,  idx_w, lde_w,  clearl)
561 GEN_VEXT_LD_INDEX(vlxe_v_d,  int64_t,  int64_t,  idx_d, lde_d,  clearq)
562 GEN_VEXT_LD_INDEX(vlxbu_v_b, uint8_t,  uint8_t,  idx_b, ldbu_b, clearb)
563 GEN_VEXT_LD_INDEX(vlxbu_v_h, uint8_t,  uint16_t, idx_h, ldbu_h, clearh)
564 GEN_VEXT_LD_INDEX(vlxbu_v_w, uint8_t,  uint32_t, idx_w, ldbu_w, clearl)
565 GEN_VEXT_LD_INDEX(vlxbu_v_d, uint8_t,  uint64_t, idx_d, ldbu_d, clearq)
566 GEN_VEXT_LD_INDEX(vlxhu_v_h, uint16_t, uint16_t, idx_h, ldhu_h, clearh)
567 GEN_VEXT_LD_INDEX(vlxhu_v_w, uint16_t, uint32_t, idx_w, ldhu_w, clearl)
568 GEN_VEXT_LD_INDEX(vlxhu_v_d, uint16_t, uint64_t, idx_d, ldhu_d, clearq)
569 GEN_VEXT_LD_INDEX(vlxwu_v_w, uint32_t, uint32_t, idx_w, ldwu_w, clearl)
570 GEN_VEXT_LD_INDEX(vlxwu_v_d, uint32_t, uint64_t, idx_d, ldwu_d, clearq)
571 
572 #define GEN_VEXT_ST_INDEX(NAME, MTYPE, ETYPE, INDEX_FN, STORE_FN)\
573 void HELPER(NAME)(void *vd, void *v0, target_ulong base,         \
574                   void *vs2, CPURISCVState *env, uint32_t desc)  \
575 {                                                                \
576     vext_ldst_index(vd, v0, base, vs2, env, desc, INDEX_FN,      \
577                     STORE_FN, NULL, sizeof(ETYPE), sizeof(MTYPE),\
578                     GETPC(), MMU_DATA_STORE);                    \
579 }
580 
581 GEN_VEXT_ST_INDEX(vsxb_v_b, int8_t,  int8_t,  idx_b, stb_b)
582 GEN_VEXT_ST_INDEX(vsxb_v_h, int8_t,  int16_t, idx_h, stb_h)
583 GEN_VEXT_ST_INDEX(vsxb_v_w, int8_t,  int32_t, idx_w, stb_w)
584 GEN_VEXT_ST_INDEX(vsxb_v_d, int8_t,  int64_t, idx_d, stb_d)
585 GEN_VEXT_ST_INDEX(vsxh_v_h, int16_t, int16_t, idx_h, sth_h)
586 GEN_VEXT_ST_INDEX(vsxh_v_w, int16_t, int32_t, idx_w, sth_w)
587 GEN_VEXT_ST_INDEX(vsxh_v_d, int16_t, int64_t, idx_d, sth_d)
588 GEN_VEXT_ST_INDEX(vsxw_v_w, int32_t, int32_t, idx_w, stw_w)
589 GEN_VEXT_ST_INDEX(vsxw_v_d, int32_t, int64_t, idx_d, stw_d)
590 GEN_VEXT_ST_INDEX(vsxe_v_b, int8_t,  int8_t,  idx_b, ste_b)
591 GEN_VEXT_ST_INDEX(vsxe_v_h, int16_t, int16_t, idx_h, ste_h)
592 GEN_VEXT_ST_INDEX(vsxe_v_w, int32_t, int32_t, idx_w, ste_w)
593 GEN_VEXT_ST_INDEX(vsxe_v_d, int64_t, int64_t, idx_d, ste_d)
594 
595 /*
596  *** unit-stride fault-only-fisrt load instructions
597  */
598 static inline void
599 vext_ldff(void *vd, void *v0, target_ulong base,
600           CPURISCVState *env, uint32_t desc,
601           vext_ldst_elem_fn *ldst_elem,
602           clear_fn *clear_elem,
603           uint32_t esz, uint32_t msz, uintptr_t ra)
604 {
605     void *host;
606     uint32_t i, k, vl = 0;
607     uint32_t mlen = vext_mlen(desc);
608     uint32_t nf = vext_nf(desc);
609     uint32_t vm = vext_vm(desc);
610     uint32_t vlmax = vext_maxsz(desc) / esz;
611     target_ulong addr, offset, remain;
612 
613     /* probe every access*/
614     for (i = 0; i < env->vl; i++) {
615         if (!vm && !vext_elem_mask(v0, mlen, i)) {
616             continue;
617         }
618         addr = base + nf * i * msz;
619         if (i == 0) {
620             probe_pages(env, addr, nf * msz, ra, MMU_DATA_LOAD);
621         } else {
622             /* if it triggers an exception, no need to check watchpoint */
623             remain = nf * msz;
624             while (remain > 0) {
625                 offset = -(addr | TARGET_PAGE_MASK);
626                 host = tlb_vaddr_to_host(env, addr, MMU_DATA_LOAD,
627                                          cpu_mmu_index(env, false));
628                 if (host) {
629 #ifdef CONFIG_USER_ONLY
630                     if (page_check_range(addr, nf * msz, PAGE_READ) < 0) {
631                         vl = i;
632                         goto ProbeSuccess;
633                     }
634 #else
635                     probe_pages(env, addr, nf * msz, ra, MMU_DATA_LOAD);
636 #endif
637                 } else {
638                     vl = i;
639                     goto ProbeSuccess;
640                 }
641                 if (remain <=  offset) {
642                     break;
643                 }
644                 remain -= offset;
645                 addr += offset;
646             }
647         }
648     }
649 ProbeSuccess:
650     /* load bytes from guest memory */
651     if (vl != 0) {
652         env->vl = vl;
653     }
654     for (i = 0; i < env->vl; i++) {
655         k = 0;
656         if (!vm && !vext_elem_mask(v0, mlen, i)) {
657             continue;
658         }
659         while (k < nf) {
660             target_ulong addr = base + (i * nf + k) * msz;
661             ldst_elem(env, addr, i + k * vlmax, vd, ra);
662             k++;
663         }
664     }
665     /* clear tail elements */
666     if (vl != 0) {
667         return;
668     }
669     for (k = 0; k < nf; k++) {
670         clear_elem(vd, env->vl + k * vlmax, env->vl * esz, vlmax * esz);
671     }
672 }
673 
674 #define GEN_VEXT_LDFF(NAME, MTYPE, ETYPE, LOAD_FN, CLEAR_FN)     \
675 void HELPER(NAME)(void *vd, void *v0, target_ulong base,         \
676                   CPURISCVState *env, uint32_t desc)             \
677 {                                                                \
678     vext_ldff(vd, v0, base, env, desc, LOAD_FN, CLEAR_FN,        \
679               sizeof(ETYPE), sizeof(MTYPE), GETPC());            \
680 }
681 
682 GEN_VEXT_LDFF(vlbff_v_b,  int8_t,   int8_t,   ldb_b,  clearb)
683 GEN_VEXT_LDFF(vlbff_v_h,  int8_t,   int16_t,  ldb_h,  clearh)
684 GEN_VEXT_LDFF(vlbff_v_w,  int8_t,   int32_t,  ldb_w,  clearl)
685 GEN_VEXT_LDFF(vlbff_v_d,  int8_t,   int64_t,  ldb_d,  clearq)
686 GEN_VEXT_LDFF(vlhff_v_h,  int16_t,  int16_t,  ldh_h,  clearh)
687 GEN_VEXT_LDFF(vlhff_v_w,  int16_t,  int32_t,  ldh_w,  clearl)
688 GEN_VEXT_LDFF(vlhff_v_d,  int16_t,  int64_t,  ldh_d,  clearq)
689 GEN_VEXT_LDFF(vlwff_v_w,  int32_t,  int32_t,  ldw_w,  clearl)
690 GEN_VEXT_LDFF(vlwff_v_d,  int32_t,  int64_t,  ldw_d,  clearq)
691 GEN_VEXT_LDFF(vleff_v_b,  int8_t,   int8_t,   lde_b,  clearb)
692 GEN_VEXT_LDFF(vleff_v_h,  int16_t,  int16_t,  lde_h,  clearh)
693 GEN_VEXT_LDFF(vleff_v_w,  int32_t,  int32_t,  lde_w,  clearl)
694 GEN_VEXT_LDFF(vleff_v_d,  int64_t,  int64_t,  lde_d,  clearq)
695 GEN_VEXT_LDFF(vlbuff_v_b, uint8_t,  uint8_t,  ldbu_b, clearb)
696 GEN_VEXT_LDFF(vlbuff_v_h, uint8_t,  uint16_t, ldbu_h, clearh)
697 GEN_VEXT_LDFF(vlbuff_v_w, uint8_t,  uint32_t, ldbu_w, clearl)
698 GEN_VEXT_LDFF(vlbuff_v_d, uint8_t,  uint64_t, ldbu_d, clearq)
699 GEN_VEXT_LDFF(vlhuff_v_h, uint16_t, uint16_t, ldhu_h, clearh)
700 GEN_VEXT_LDFF(vlhuff_v_w, uint16_t, uint32_t, ldhu_w, clearl)
701 GEN_VEXT_LDFF(vlhuff_v_d, uint16_t, uint64_t, ldhu_d, clearq)
702 GEN_VEXT_LDFF(vlwuff_v_w, uint32_t, uint32_t, ldwu_w, clearl)
703 GEN_VEXT_LDFF(vlwuff_v_d, uint32_t, uint64_t, ldwu_d, clearq)
704 
705 /*
706  *** Vector AMO Operations (Zvamo)
707  */
708 typedef void vext_amo_noatomic_fn(void *vs3, target_ulong addr,
709                                   uint32_t wd, uint32_t idx, CPURISCVState *env,
710                                   uintptr_t retaddr);
711 
712 /* no atomic opreation for vector atomic insructions */
713 #define DO_SWAP(N, M) (M)
714 #define DO_AND(N, M)  (N & M)
715 #define DO_XOR(N, M)  (N ^ M)
716 #define DO_OR(N, M)   (N | M)
717 #define DO_ADD(N, M)  (N + M)
718 
719 #define GEN_VEXT_AMO_NOATOMIC_OP(NAME, ESZ, MSZ, H, DO_OP, SUF) \
720 static void                                                     \
721 vext_##NAME##_noatomic_op(void *vs3, target_ulong addr,         \
722                           uint32_t wd, uint32_t idx,            \
723                           CPURISCVState *env, uintptr_t retaddr)\
724 {                                                               \
725     typedef int##ESZ##_t ETYPE;                                 \
726     typedef int##MSZ##_t MTYPE;                                 \
727     typedef uint##MSZ##_t UMTYPE __attribute__((unused));       \
728     ETYPE *pe3 = (ETYPE *)vs3 + H(idx);                         \
729     MTYPE  a = cpu_ld##SUF##_data(env, addr), b = *pe3;         \
730                                                                 \
731     cpu_st##SUF##_data(env, addr, DO_OP(a, b));                 \
732     if (wd) {                                                   \
733         *pe3 = a;                                               \
734     }                                                           \
735 }
736 
737 /* Signed min/max */
738 #define DO_MAX(N, M)  ((N) >= (M) ? (N) : (M))
739 #define DO_MIN(N, M)  ((N) >= (M) ? (M) : (N))
740 
741 /* Unsigned min/max */
742 #define DO_MAXU(N, M) DO_MAX((UMTYPE)N, (UMTYPE)M)
743 #define DO_MINU(N, M) DO_MIN((UMTYPE)N, (UMTYPE)M)
744 
745 GEN_VEXT_AMO_NOATOMIC_OP(vamoswapw_v_w, 32, 32, H4, DO_SWAP, l)
746 GEN_VEXT_AMO_NOATOMIC_OP(vamoaddw_v_w,  32, 32, H4, DO_ADD,  l)
747 GEN_VEXT_AMO_NOATOMIC_OP(vamoxorw_v_w,  32, 32, H4, DO_XOR,  l)
748 GEN_VEXT_AMO_NOATOMIC_OP(vamoandw_v_w,  32, 32, H4, DO_AND,  l)
749 GEN_VEXT_AMO_NOATOMIC_OP(vamoorw_v_w,   32, 32, H4, DO_OR,   l)
750 GEN_VEXT_AMO_NOATOMIC_OP(vamominw_v_w,  32, 32, H4, DO_MIN,  l)
751 GEN_VEXT_AMO_NOATOMIC_OP(vamomaxw_v_w,  32, 32, H4, DO_MAX,  l)
752 GEN_VEXT_AMO_NOATOMIC_OP(vamominuw_v_w, 32, 32, H4, DO_MINU, l)
753 GEN_VEXT_AMO_NOATOMIC_OP(vamomaxuw_v_w, 32, 32, H4, DO_MAXU, l)
754 #ifdef TARGET_RISCV64
755 GEN_VEXT_AMO_NOATOMIC_OP(vamoswapw_v_d, 64, 32, H8, DO_SWAP, l)
756 GEN_VEXT_AMO_NOATOMIC_OP(vamoswapd_v_d, 64, 64, H8, DO_SWAP, q)
757 GEN_VEXT_AMO_NOATOMIC_OP(vamoaddw_v_d,  64, 32, H8, DO_ADD,  l)
758 GEN_VEXT_AMO_NOATOMIC_OP(vamoaddd_v_d,  64, 64, H8, DO_ADD,  q)
759 GEN_VEXT_AMO_NOATOMIC_OP(vamoxorw_v_d,  64, 32, H8, DO_XOR,  l)
760 GEN_VEXT_AMO_NOATOMIC_OP(vamoxord_v_d,  64, 64, H8, DO_XOR,  q)
761 GEN_VEXT_AMO_NOATOMIC_OP(vamoandw_v_d,  64, 32, H8, DO_AND,  l)
762 GEN_VEXT_AMO_NOATOMIC_OP(vamoandd_v_d,  64, 64, H8, DO_AND,  q)
763 GEN_VEXT_AMO_NOATOMIC_OP(vamoorw_v_d,   64, 32, H8, DO_OR,   l)
764 GEN_VEXT_AMO_NOATOMIC_OP(vamoord_v_d,   64, 64, H8, DO_OR,   q)
765 GEN_VEXT_AMO_NOATOMIC_OP(vamominw_v_d,  64, 32, H8, DO_MIN,  l)
766 GEN_VEXT_AMO_NOATOMIC_OP(vamomind_v_d,  64, 64, H8, DO_MIN,  q)
767 GEN_VEXT_AMO_NOATOMIC_OP(vamomaxw_v_d,  64, 32, H8, DO_MAX,  l)
768 GEN_VEXT_AMO_NOATOMIC_OP(vamomaxd_v_d,  64, 64, H8, DO_MAX,  q)
769 GEN_VEXT_AMO_NOATOMIC_OP(vamominuw_v_d, 64, 32, H8, DO_MINU, l)
770 GEN_VEXT_AMO_NOATOMIC_OP(vamominud_v_d, 64, 64, H8, DO_MINU, q)
771 GEN_VEXT_AMO_NOATOMIC_OP(vamomaxuw_v_d, 64, 32, H8, DO_MAXU, l)
772 GEN_VEXT_AMO_NOATOMIC_OP(vamomaxud_v_d, 64, 64, H8, DO_MAXU, q)
773 #endif
774 
775 static inline void
776 vext_amo_noatomic(void *vs3, void *v0, target_ulong base,
777                   void *vs2, CPURISCVState *env, uint32_t desc,
778                   vext_get_index_addr get_index_addr,
779                   vext_amo_noatomic_fn *noatomic_op,
780                   clear_fn *clear_elem,
781                   uint32_t esz, uint32_t msz, uintptr_t ra)
782 {
783     uint32_t i;
784     target_long addr;
785     uint32_t wd = vext_wd(desc);
786     uint32_t vm = vext_vm(desc);
787     uint32_t mlen = vext_mlen(desc);
788     uint32_t vlmax = vext_maxsz(desc) / esz;
789 
790     for (i = 0; i < env->vl; i++) {
791         if (!vm && !vext_elem_mask(v0, mlen, i)) {
792             continue;
793         }
794         probe_pages(env, get_index_addr(base, i, vs2), msz, ra, MMU_DATA_LOAD);
795         probe_pages(env, get_index_addr(base, i, vs2), msz, ra, MMU_DATA_STORE);
796     }
797     for (i = 0; i < env->vl; i++) {
798         if (!vm && !vext_elem_mask(v0, mlen, i)) {
799             continue;
800         }
801         addr = get_index_addr(base, i, vs2);
802         noatomic_op(vs3, addr, wd, i, env, ra);
803     }
804     clear_elem(vs3, env->vl, env->vl * esz, vlmax * esz);
805 }
806 
807 #define GEN_VEXT_AMO(NAME, MTYPE, ETYPE, INDEX_FN, CLEAR_FN)    \
808 void HELPER(NAME)(void *vs3, void *v0, target_ulong base,       \
809                   void *vs2, CPURISCVState *env, uint32_t desc) \
810 {                                                               \
811     vext_amo_noatomic(vs3, v0, base, vs2, env, desc,            \
812                       INDEX_FN, vext_##NAME##_noatomic_op,      \
813                       CLEAR_FN, sizeof(ETYPE), sizeof(MTYPE),   \
814                       GETPC());                                 \
815 }
816 
817 #ifdef TARGET_RISCV64
818 GEN_VEXT_AMO(vamoswapw_v_d, int32_t,  int64_t,  idx_d, clearq)
819 GEN_VEXT_AMO(vamoswapd_v_d, int64_t,  int64_t,  idx_d, clearq)
820 GEN_VEXT_AMO(vamoaddw_v_d,  int32_t,  int64_t,  idx_d, clearq)
821 GEN_VEXT_AMO(vamoaddd_v_d,  int64_t,  int64_t,  idx_d, clearq)
822 GEN_VEXT_AMO(vamoxorw_v_d,  int32_t,  int64_t,  idx_d, clearq)
823 GEN_VEXT_AMO(vamoxord_v_d,  int64_t,  int64_t,  idx_d, clearq)
824 GEN_VEXT_AMO(vamoandw_v_d,  int32_t,  int64_t,  idx_d, clearq)
825 GEN_VEXT_AMO(vamoandd_v_d,  int64_t,  int64_t,  idx_d, clearq)
826 GEN_VEXT_AMO(vamoorw_v_d,   int32_t,  int64_t,  idx_d, clearq)
827 GEN_VEXT_AMO(vamoord_v_d,   int64_t,  int64_t,  idx_d, clearq)
828 GEN_VEXT_AMO(vamominw_v_d,  int32_t,  int64_t,  idx_d, clearq)
829 GEN_VEXT_AMO(vamomind_v_d,  int64_t,  int64_t,  idx_d, clearq)
830 GEN_VEXT_AMO(vamomaxw_v_d,  int32_t,  int64_t,  idx_d, clearq)
831 GEN_VEXT_AMO(vamomaxd_v_d,  int64_t,  int64_t,  idx_d, clearq)
832 GEN_VEXT_AMO(vamominuw_v_d, uint32_t, uint64_t, idx_d, clearq)
833 GEN_VEXT_AMO(vamominud_v_d, uint64_t, uint64_t, idx_d, clearq)
834 GEN_VEXT_AMO(vamomaxuw_v_d, uint32_t, uint64_t, idx_d, clearq)
835 GEN_VEXT_AMO(vamomaxud_v_d, uint64_t, uint64_t, idx_d, clearq)
836 #endif
837 GEN_VEXT_AMO(vamoswapw_v_w, int32_t,  int32_t,  idx_w, clearl)
838 GEN_VEXT_AMO(vamoaddw_v_w,  int32_t,  int32_t,  idx_w, clearl)
839 GEN_VEXT_AMO(vamoxorw_v_w,  int32_t,  int32_t,  idx_w, clearl)
840 GEN_VEXT_AMO(vamoandw_v_w,  int32_t,  int32_t,  idx_w, clearl)
841 GEN_VEXT_AMO(vamoorw_v_w,   int32_t,  int32_t,  idx_w, clearl)
842 GEN_VEXT_AMO(vamominw_v_w,  int32_t,  int32_t,  idx_w, clearl)
843 GEN_VEXT_AMO(vamomaxw_v_w,  int32_t,  int32_t,  idx_w, clearl)
844 GEN_VEXT_AMO(vamominuw_v_w, uint32_t, uint32_t, idx_w, clearl)
845 GEN_VEXT_AMO(vamomaxuw_v_w, uint32_t, uint32_t, idx_w, clearl)
846 
847 /*
848  *** Vector Integer Arithmetic Instructions
849  */
850 
851 /* expand macro args before macro */
852 #define RVVCALL(macro, ...)  macro(__VA_ARGS__)
853 
854 /* (TD, T1, T2, TX1, TX2) */
855 #define OP_SSS_B int8_t, int8_t, int8_t, int8_t, int8_t
856 #define OP_SSS_H int16_t, int16_t, int16_t, int16_t, int16_t
857 #define OP_SSS_W int32_t, int32_t, int32_t, int32_t, int32_t
858 #define OP_SSS_D int64_t, int64_t, int64_t, int64_t, int64_t
859 #define OP_UUU_B uint8_t, uint8_t, uint8_t, uint8_t, uint8_t
860 #define OP_UUU_H uint16_t, uint16_t, uint16_t, uint16_t, uint16_t
861 #define OP_UUU_W uint32_t, uint32_t, uint32_t, uint32_t, uint32_t
862 #define OP_UUU_D uint64_t, uint64_t, uint64_t, uint64_t, uint64_t
863 #define OP_SUS_B int8_t, uint8_t, int8_t, uint8_t, int8_t
864 #define OP_SUS_H int16_t, uint16_t, int16_t, uint16_t, int16_t
865 #define OP_SUS_W int32_t, uint32_t, int32_t, uint32_t, int32_t
866 #define OP_SUS_D int64_t, uint64_t, int64_t, uint64_t, int64_t
867 #define WOP_UUU_B uint16_t, uint8_t, uint8_t, uint16_t, uint16_t
868 #define WOP_UUU_H uint32_t, uint16_t, uint16_t, uint32_t, uint32_t
869 #define WOP_UUU_W uint64_t, uint32_t, uint32_t, uint64_t, uint64_t
870 #define WOP_SSS_B int16_t, int8_t, int8_t, int16_t, int16_t
871 #define WOP_SSS_H int32_t, int16_t, int16_t, int32_t, int32_t
872 #define WOP_SSS_W int64_t, int32_t, int32_t, int64_t, int64_t
873 #define WOP_SUS_B int16_t, uint8_t, int8_t, uint16_t, int16_t
874 #define WOP_SUS_H int32_t, uint16_t, int16_t, uint32_t, int32_t
875 #define WOP_SUS_W int64_t, uint32_t, int32_t, uint64_t, int64_t
876 #define WOP_SSU_B int16_t, int8_t, uint8_t, int16_t, uint16_t
877 #define WOP_SSU_H int32_t, int16_t, uint16_t, int32_t, uint32_t
878 #define WOP_SSU_W int64_t, int32_t, uint32_t, int64_t, uint64_t
879 #define NOP_SSS_B int8_t, int8_t, int16_t, int8_t, int16_t
880 #define NOP_SSS_H int16_t, int16_t, int32_t, int16_t, int32_t
881 #define NOP_SSS_W int32_t, int32_t, int64_t, int32_t, int64_t
882 #define NOP_UUU_B uint8_t, uint8_t, uint16_t, uint8_t, uint16_t
883 #define NOP_UUU_H uint16_t, uint16_t, uint32_t, uint16_t, uint32_t
884 #define NOP_UUU_W uint32_t, uint32_t, uint64_t, uint32_t, uint64_t
885 
886 /* operation of two vector elements */
887 typedef void opivv2_fn(void *vd, void *vs1, void *vs2, int i);
888 
889 #define OPIVV2(NAME, TD, T1, T2, TX1, TX2, HD, HS1, HS2, OP)    \
890 static void do_##NAME(void *vd, void *vs1, void *vs2, int i)    \
891 {                                                               \
892     TX1 s1 = *((T1 *)vs1 + HS1(i));                             \
893     TX2 s2 = *((T2 *)vs2 + HS2(i));                             \
894     *((TD *)vd + HD(i)) = OP(s2, s1);                           \
895 }
896 #define DO_SUB(N, M) (N - M)
897 #define DO_RSUB(N, M) (M - N)
898 
899 RVVCALL(OPIVV2, vadd_vv_b, OP_SSS_B, H1, H1, H1, DO_ADD)
900 RVVCALL(OPIVV2, vadd_vv_h, OP_SSS_H, H2, H2, H2, DO_ADD)
901 RVVCALL(OPIVV2, vadd_vv_w, OP_SSS_W, H4, H4, H4, DO_ADD)
902 RVVCALL(OPIVV2, vadd_vv_d, OP_SSS_D, H8, H8, H8, DO_ADD)
903 RVVCALL(OPIVV2, vsub_vv_b, OP_SSS_B, H1, H1, H1, DO_SUB)
904 RVVCALL(OPIVV2, vsub_vv_h, OP_SSS_H, H2, H2, H2, DO_SUB)
905 RVVCALL(OPIVV2, vsub_vv_w, OP_SSS_W, H4, H4, H4, DO_SUB)
906 RVVCALL(OPIVV2, vsub_vv_d, OP_SSS_D, H8, H8, H8, DO_SUB)
907 
908 static void do_vext_vv(void *vd, void *v0, void *vs1, void *vs2,
909                        CPURISCVState *env, uint32_t desc,
910                        uint32_t esz, uint32_t dsz,
911                        opivv2_fn *fn, clear_fn *clearfn)
912 {
913     uint32_t vlmax = vext_maxsz(desc) / esz;
914     uint32_t mlen = vext_mlen(desc);
915     uint32_t vm = vext_vm(desc);
916     uint32_t vl = env->vl;
917     uint32_t i;
918 
919     for (i = 0; i < vl; i++) {
920         if (!vm && !vext_elem_mask(v0, mlen, i)) {
921             continue;
922         }
923         fn(vd, vs1, vs2, i);
924     }
925     clearfn(vd, vl, vl * dsz,  vlmax * dsz);
926 }
927 
928 /* generate the helpers for OPIVV */
929 #define GEN_VEXT_VV(NAME, ESZ, DSZ, CLEAR_FN)             \
930 void HELPER(NAME)(void *vd, void *v0, void *vs1,          \
931                   void *vs2, CPURISCVState *env,          \
932                   uint32_t desc)                          \
933 {                                                         \
934     do_vext_vv(vd, v0, vs1, vs2, env, desc, ESZ, DSZ,     \
935                do_##NAME, CLEAR_FN);                      \
936 }
937 
938 GEN_VEXT_VV(vadd_vv_b, 1, 1, clearb)
939 GEN_VEXT_VV(vadd_vv_h, 2, 2, clearh)
940 GEN_VEXT_VV(vadd_vv_w, 4, 4, clearl)
941 GEN_VEXT_VV(vadd_vv_d, 8, 8, clearq)
942 GEN_VEXT_VV(vsub_vv_b, 1, 1, clearb)
943 GEN_VEXT_VV(vsub_vv_h, 2, 2, clearh)
944 GEN_VEXT_VV(vsub_vv_w, 4, 4, clearl)
945 GEN_VEXT_VV(vsub_vv_d, 8, 8, clearq)
946 
947 typedef void opivx2_fn(void *vd, target_long s1, void *vs2, int i);
948 
949 /*
950  * (T1)s1 gives the real operator type.
951  * (TX1)(T1)s1 expands the operator type of widen or narrow operations.
952  */
953 #define OPIVX2(NAME, TD, T1, T2, TX1, TX2, HD, HS2, OP)             \
954 static void do_##NAME(void *vd, target_long s1, void *vs2, int i)   \
955 {                                                                   \
956     TX2 s2 = *((T2 *)vs2 + HS2(i));                                 \
957     *((TD *)vd + HD(i)) = OP(s2, (TX1)(T1)s1);                      \
958 }
959 
960 RVVCALL(OPIVX2, vadd_vx_b, OP_SSS_B, H1, H1, DO_ADD)
961 RVVCALL(OPIVX2, vadd_vx_h, OP_SSS_H, H2, H2, DO_ADD)
962 RVVCALL(OPIVX2, vadd_vx_w, OP_SSS_W, H4, H4, DO_ADD)
963 RVVCALL(OPIVX2, vadd_vx_d, OP_SSS_D, H8, H8, DO_ADD)
964 RVVCALL(OPIVX2, vsub_vx_b, OP_SSS_B, H1, H1, DO_SUB)
965 RVVCALL(OPIVX2, vsub_vx_h, OP_SSS_H, H2, H2, DO_SUB)
966 RVVCALL(OPIVX2, vsub_vx_w, OP_SSS_W, H4, H4, DO_SUB)
967 RVVCALL(OPIVX2, vsub_vx_d, OP_SSS_D, H8, H8, DO_SUB)
968 RVVCALL(OPIVX2, vrsub_vx_b, OP_SSS_B, H1, H1, DO_RSUB)
969 RVVCALL(OPIVX2, vrsub_vx_h, OP_SSS_H, H2, H2, DO_RSUB)
970 RVVCALL(OPIVX2, vrsub_vx_w, OP_SSS_W, H4, H4, DO_RSUB)
971 RVVCALL(OPIVX2, vrsub_vx_d, OP_SSS_D, H8, H8, DO_RSUB)
972 
973 static void do_vext_vx(void *vd, void *v0, target_long s1, void *vs2,
974                        CPURISCVState *env, uint32_t desc,
975                        uint32_t esz, uint32_t dsz,
976                        opivx2_fn fn, clear_fn *clearfn)
977 {
978     uint32_t vlmax = vext_maxsz(desc) / esz;
979     uint32_t mlen = vext_mlen(desc);
980     uint32_t vm = vext_vm(desc);
981     uint32_t vl = env->vl;
982     uint32_t i;
983 
984     for (i = 0; i < vl; i++) {
985         if (!vm && !vext_elem_mask(v0, mlen, i)) {
986             continue;
987         }
988         fn(vd, s1, vs2, i);
989     }
990     clearfn(vd, vl, vl * dsz,  vlmax * dsz);
991 }
992 
993 /* generate the helpers for OPIVX */
994 #define GEN_VEXT_VX(NAME, ESZ, DSZ, CLEAR_FN)             \
995 void HELPER(NAME)(void *vd, void *v0, target_ulong s1,    \
996                   void *vs2, CPURISCVState *env,          \
997                   uint32_t desc)                          \
998 {                                                         \
999     do_vext_vx(vd, v0, s1, vs2, env, desc, ESZ, DSZ,      \
1000                do_##NAME, CLEAR_FN);                      \
1001 }
1002 
1003 GEN_VEXT_VX(vadd_vx_b, 1, 1, clearb)
1004 GEN_VEXT_VX(vadd_vx_h, 2, 2, clearh)
1005 GEN_VEXT_VX(vadd_vx_w, 4, 4, clearl)
1006 GEN_VEXT_VX(vadd_vx_d, 8, 8, clearq)
1007 GEN_VEXT_VX(vsub_vx_b, 1, 1, clearb)
1008 GEN_VEXT_VX(vsub_vx_h, 2, 2, clearh)
1009 GEN_VEXT_VX(vsub_vx_w, 4, 4, clearl)
1010 GEN_VEXT_VX(vsub_vx_d, 8, 8, clearq)
1011 GEN_VEXT_VX(vrsub_vx_b, 1, 1, clearb)
1012 GEN_VEXT_VX(vrsub_vx_h, 2, 2, clearh)
1013 GEN_VEXT_VX(vrsub_vx_w, 4, 4, clearl)
1014 GEN_VEXT_VX(vrsub_vx_d, 8, 8, clearq)
1015 
1016 void HELPER(vec_rsubs8)(void *d, void *a, uint64_t b, uint32_t desc)
1017 {
1018     intptr_t oprsz = simd_oprsz(desc);
1019     intptr_t i;
1020 
1021     for (i = 0; i < oprsz; i += sizeof(uint8_t)) {
1022         *(uint8_t *)(d + i) = (uint8_t)b - *(uint8_t *)(a + i);
1023     }
1024 }
1025 
1026 void HELPER(vec_rsubs16)(void *d, void *a, uint64_t b, uint32_t desc)
1027 {
1028     intptr_t oprsz = simd_oprsz(desc);
1029     intptr_t i;
1030 
1031     for (i = 0; i < oprsz; i += sizeof(uint16_t)) {
1032         *(uint16_t *)(d + i) = (uint16_t)b - *(uint16_t *)(a + i);
1033     }
1034 }
1035 
1036 void HELPER(vec_rsubs32)(void *d, void *a, uint64_t b, uint32_t desc)
1037 {
1038     intptr_t oprsz = simd_oprsz(desc);
1039     intptr_t i;
1040 
1041     for (i = 0; i < oprsz; i += sizeof(uint32_t)) {
1042         *(uint32_t *)(d + i) = (uint32_t)b - *(uint32_t *)(a + i);
1043     }
1044 }
1045 
1046 void HELPER(vec_rsubs64)(void *d, void *a, uint64_t b, uint32_t desc)
1047 {
1048     intptr_t oprsz = simd_oprsz(desc);
1049     intptr_t i;
1050 
1051     for (i = 0; i < oprsz; i += sizeof(uint64_t)) {
1052         *(uint64_t *)(d + i) = b - *(uint64_t *)(a + i);
1053     }
1054 }
1055 
1056 /* Vector Widening Integer Add/Subtract */
1057 #define WOP_UUU_B uint16_t, uint8_t, uint8_t, uint16_t, uint16_t
1058 #define WOP_UUU_H uint32_t, uint16_t, uint16_t, uint32_t, uint32_t
1059 #define WOP_UUU_W uint64_t, uint32_t, uint32_t, uint64_t, uint64_t
1060 #define WOP_SSS_B int16_t, int8_t, int8_t, int16_t, int16_t
1061 #define WOP_SSS_H int32_t, int16_t, int16_t, int32_t, int32_t
1062 #define WOP_SSS_W int64_t, int32_t, int32_t, int64_t, int64_t
1063 #define WOP_WUUU_B  uint16_t, uint8_t, uint16_t, uint16_t, uint16_t
1064 #define WOP_WUUU_H  uint32_t, uint16_t, uint32_t, uint32_t, uint32_t
1065 #define WOP_WUUU_W  uint64_t, uint32_t, uint64_t, uint64_t, uint64_t
1066 #define WOP_WSSS_B  int16_t, int8_t, int16_t, int16_t, int16_t
1067 #define WOP_WSSS_H  int32_t, int16_t, int32_t, int32_t, int32_t
1068 #define WOP_WSSS_W  int64_t, int32_t, int64_t, int64_t, int64_t
1069 RVVCALL(OPIVV2, vwaddu_vv_b, WOP_UUU_B, H2, H1, H1, DO_ADD)
1070 RVVCALL(OPIVV2, vwaddu_vv_h, WOP_UUU_H, H4, H2, H2, DO_ADD)
1071 RVVCALL(OPIVV2, vwaddu_vv_w, WOP_UUU_W, H8, H4, H4, DO_ADD)
1072 RVVCALL(OPIVV2, vwsubu_vv_b, WOP_UUU_B, H2, H1, H1, DO_SUB)
1073 RVVCALL(OPIVV2, vwsubu_vv_h, WOP_UUU_H, H4, H2, H2, DO_SUB)
1074 RVVCALL(OPIVV2, vwsubu_vv_w, WOP_UUU_W, H8, H4, H4, DO_SUB)
1075 RVVCALL(OPIVV2, vwadd_vv_b, WOP_SSS_B, H2, H1, H1, DO_ADD)
1076 RVVCALL(OPIVV2, vwadd_vv_h, WOP_SSS_H, H4, H2, H2, DO_ADD)
1077 RVVCALL(OPIVV2, vwadd_vv_w, WOP_SSS_W, H8, H4, H4, DO_ADD)
1078 RVVCALL(OPIVV2, vwsub_vv_b, WOP_SSS_B, H2, H1, H1, DO_SUB)
1079 RVVCALL(OPIVV2, vwsub_vv_h, WOP_SSS_H, H4, H2, H2, DO_SUB)
1080 RVVCALL(OPIVV2, vwsub_vv_w, WOP_SSS_W, H8, H4, H4, DO_SUB)
1081 RVVCALL(OPIVV2, vwaddu_wv_b, WOP_WUUU_B, H2, H1, H1, DO_ADD)
1082 RVVCALL(OPIVV2, vwaddu_wv_h, WOP_WUUU_H, H4, H2, H2, DO_ADD)
1083 RVVCALL(OPIVV2, vwaddu_wv_w, WOP_WUUU_W, H8, H4, H4, DO_ADD)
1084 RVVCALL(OPIVV2, vwsubu_wv_b, WOP_WUUU_B, H2, H1, H1, DO_SUB)
1085 RVVCALL(OPIVV2, vwsubu_wv_h, WOP_WUUU_H, H4, H2, H2, DO_SUB)
1086 RVVCALL(OPIVV2, vwsubu_wv_w, WOP_WUUU_W, H8, H4, H4, DO_SUB)
1087 RVVCALL(OPIVV2, vwadd_wv_b, WOP_WSSS_B, H2, H1, H1, DO_ADD)
1088 RVVCALL(OPIVV2, vwadd_wv_h, WOP_WSSS_H, H4, H2, H2, DO_ADD)
1089 RVVCALL(OPIVV2, vwadd_wv_w, WOP_WSSS_W, H8, H4, H4, DO_ADD)
1090 RVVCALL(OPIVV2, vwsub_wv_b, WOP_WSSS_B, H2, H1, H1, DO_SUB)
1091 RVVCALL(OPIVV2, vwsub_wv_h, WOP_WSSS_H, H4, H2, H2, DO_SUB)
1092 RVVCALL(OPIVV2, vwsub_wv_w, WOP_WSSS_W, H8, H4, H4, DO_SUB)
1093 GEN_VEXT_VV(vwaddu_vv_b, 1, 2, clearh)
1094 GEN_VEXT_VV(vwaddu_vv_h, 2, 4, clearl)
1095 GEN_VEXT_VV(vwaddu_vv_w, 4, 8, clearq)
1096 GEN_VEXT_VV(vwsubu_vv_b, 1, 2, clearh)
1097 GEN_VEXT_VV(vwsubu_vv_h, 2, 4, clearl)
1098 GEN_VEXT_VV(vwsubu_vv_w, 4, 8, clearq)
1099 GEN_VEXT_VV(vwadd_vv_b, 1, 2, clearh)
1100 GEN_VEXT_VV(vwadd_vv_h, 2, 4, clearl)
1101 GEN_VEXT_VV(vwadd_vv_w, 4, 8, clearq)
1102 GEN_VEXT_VV(vwsub_vv_b, 1, 2, clearh)
1103 GEN_VEXT_VV(vwsub_vv_h, 2, 4, clearl)
1104 GEN_VEXT_VV(vwsub_vv_w, 4, 8, clearq)
1105 GEN_VEXT_VV(vwaddu_wv_b, 1, 2, clearh)
1106 GEN_VEXT_VV(vwaddu_wv_h, 2, 4, clearl)
1107 GEN_VEXT_VV(vwaddu_wv_w, 4, 8, clearq)
1108 GEN_VEXT_VV(vwsubu_wv_b, 1, 2, clearh)
1109 GEN_VEXT_VV(vwsubu_wv_h, 2, 4, clearl)
1110 GEN_VEXT_VV(vwsubu_wv_w, 4, 8, clearq)
1111 GEN_VEXT_VV(vwadd_wv_b, 1, 2, clearh)
1112 GEN_VEXT_VV(vwadd_wv_h, 2, 4, clearl)
1113 GEN_VEXT_VV(vwadd_wv_w, 4, 8, clearq)
1114 GEN_VEXT_VV(vwsub_wv_b, 1, 2, clearh)
1115 GEN_VEXT_VV(vwsub_wv_h, 2, 4, clearl)
1116 GEN_VEXT_VV(vwsub_wv_w, 4, 8, clearq)
1117 
1118 RVVCALL(OPIVX2, vwaddu_vx_b, WOP_UUU_B, H2, H1, DO_ADD)
1119 RVVCALL(OPIVX2, vwaddu_vx_h, WOP_UUU_H, H4, H2, DO_ADD)
1120 RVVCALL(OPIVX2, vwaddu_vx_w, WOP_UUU_W, H8, H4, DO_ADD)
1121 RVVCALL(OPIVX2, vwsubu_vx_b, WOP_UUU_B, H2, H1, DO_SUB)
1122 RVVCALL(OPIVX2, vwsubu_vx_h, WOP_UUU_H, H4, H2, DO_SUB)
1123 RVVCALL(OPIVX2, vwsubu_vx_w, WOP_UUU_W, H8, H4, DO_SUB)
1124 RVVCALL(OPIVX2, vwadd_vx_b, WOP_SSS_B, H2, H1, DO_ADD)
1125 RVVCALL(OPIVX2, vwadd_vx_h, WOP_SSS_H, H4, H2, DO_ADD)
1126 RVVCALL(OPIVX2, vwadd_vx_w, WOP_SSS_W, H8, H4, DO_ADD)
1127 RVVCALL(OPIVX2, vwsub_vx_b, WOP_SSS_B, H2, H1, DO_SUB)
1128 RVVCALL(OPIVX2, vwsub_vx_h, WOP_SSS_H, H4, H2, DO_SUB)
1129 RVVCALL(OPIVX2, vwsub_vx_w, WOP_SSS_W, H8, H4, DO_SUB)
1130 RVVCALL(OPIVX2, vwaddu_wx_b, WOP_WUUU_B, H2, H1, DO_ADD)
1131 RVVCALL(OPIVX2, vwaddu_wx_h, WOP_WUUU_H, H4, H2, DO_ADD)
1132 RVVCALL(OPIVX2, vwaddu_wx_w, WOP_WUUU_W, H8, H4, DO_ADD)
1133 RVVCALL(OPIVX2, vwsubu_wx_b, WOP_WUUU_B, H2, H1, DO_SUB)
1134 RVVCALL(OPIVX2, vwsubu_wx_h, WOP_WUUU_H, H4, H2, DO_SUB)
1135 RVVCALL(OPIVX2, vwsubu_wx_w, WOP_WUUU_W, H8, H4, DO_SUB)
1136 RVVCALL(OPIVX2, vwadd_wx_b, WOP_WSSS_B, H2, H1, DO_ADD)
1137 RVVCALL(OPIVX2, vwadd_wx_h, WOP_WSSS_H, H4, H2, DO_ADD)
1138 RVVCALL(OPIVX2, vwadd_wx_w, WOP_WSSS_W, H8, H4, DO_ADD)
1139 RVVCALL(OPIVX2, vwsub_wx_b, WOP_WSSS_B, H2, H1, DO_SUB)
1140 RVVCALL(OPIVX2, vwsub_wx_h, WOP_WSSS_H, H4, H2, DO_SUB)
1141 RVVCALL(OPIVX2, vwsub_wx_w, WOP_WSSS_W, H8, H4, DO_SUB)
1142 GEN_VEXT_VX(vwaddu_vx_b, 1, 2, clearh)
1143 GEN_VEXT_VX(vwaddu_vx_h, 2, 4, clearl)
1144 GEN_VEXT_VX(vwaddu_vx_w, 4, 8, clearq)
1145 GEN_VEXT_VX(vwsubu_vx_b, 1, 2, clearh)
1146 GEN_VEXT_VX(vwsubu_vx_h, 2, 4, clearl)
1147 GEN_VEXT_VX(vwsubu_vx_w, 4, 8, clearq)
1148 GEN_VEXT_VX(vwadd_vx_b, 1, 2, clearh)
1149 GEN_VEXT_VX(vwadd_vx_h, 2, 4, clearl)
1150 GEN_VEXT_VX(vwadd_vx_w, 4, 8, clearq)
1151 GEN_VEXT_VX(vwsub_vx_b, 1, 2, clearh)
1152 GEN_VEXT_VX(vwsub_vx_h, 2, 4, clearl)
1153 GEN_VEXT_VX(vwsub_vx_w, 4, 8, clearq)
1154 GEN_VEXT_VX(vwaddu_wx_b, 1, 2, clearh)
1155 GEN_VEXT_VX(vwaddu_wx_h, 2, 4, clearl)
1156 GEN_VEXT_VX(vwaddu_wx_w, 4, 8, clearq)
1157 GEN_VEXT_VX(vwsubu_wx_b, 1, 2, clearh)
1158 GEN_VEXT_VX(vwsubu_wx_h, 2, 4, clearl)
1159 GEN_VEXT_VX(vwsubu_wx_w, 4, 8, clearq)
1160 GEN_VEXT_VX(vwadd_wx_b, 1, 2, clearh)
1161 GEN_VEXT_VX(vwadd_wx_h, 2, 4, clearl)
1162 GEN_VEXT_VX(vwadd_wx_w, 4, 8, clearq)
1163 GEN_VEXT_VX(vwsub_wx_b, 1, 2, clearh)
1164 GEN_VEXT_VX(vwsub_wx_h, 2, 4, clearl)
1165 GEN_VEXT_VX(vwsub_wx_w, 4, 8, clearq)
1166 
1167 /* Vector Integer Add-with-Carry / Subtract-with-Borrow Instructions */
1168 #define DO_VADC(N, M, C) (N + M + C)
1169 #define DO_VSBC(N, M, C) (N - M - C)
1170 
1171 #define GEN_VEXT_VADC_VVM(NAME, ETYPE, H, DO_OP, CLEAR_FN)    \
1172 void HELPER(NAME)(void *vd, void *v0, void *vs1, void *vs2,   \
1173                   CPURISCVState *env, uint32_t desc)          \
1174 {                                                             \
1175     uint32_t mlen = vext_mlen(desc);                          \
1176     uint32_t vl = env->vl;                                    \
1177     uint32_t esz = sizeof(ETYPE);                             \
1178     uint32_t vlmax = vext_maxsz(desc) / esz;                  \
1179     uint32_t i;                                               \
1180                                                               \
1181     for (i = 0; i < vl; i++) {                                \
1182         ETYPE s1 = *((ETYPE *)vs1 + H(i));                    \
1183         ETYPE s2 = *((ETYPE *)vs2 + H(i));                    \
1184         uint8_t carry = vext_elem_mask(v0, mlen, i);          \
1185                                                               \
1186         *((ETYPE *)vd + H(i)) = DO_OP(s2, s1, carry);         \
1187     }                                                         \
1188     CLEAR_FN(vd, vl, vl * esz, vlmax * esz);                  \
1189 }
1190 
1191 GEN_VEXT_VADC_VVM(vadc_vvm_b, uint8_t,  H1, DO_VADC, clearb)
1192 GEN_VEXT_VADC_VVM(vadc_vvm_h, uint16_t, H2, DO_VADC, clearh)
1193 GEN_VEXT_VADC_VVM(vadc_vvm_w, uint32_t, H4, DO_VADC, clearl)
1194 GEN_VEXT_VADC_VVM(vadc_vvm_d, uint64_t, H8, DO_VADC, clearq)
1195 
1196 GEN_VEXT_VADC_VVM(vsbc_vvm_b, uint8_t,  H1, DO_VSBC, clearb)
1197 GEN_VEXT_VADC_VVM(vsbc_vvm_h, uint16_t, H2, DO_VSBC, clearh)
1198 GEN_VEXT_VADC_VVM(vsbc_vvm_w, uint32_t, H4, DO_VSBC, clearl)
1199 GEN_VEXT_VADC_VVM(vsbc_vvm_d, uint64_t, H8, DO_VSBC, clearq)
1200 
1201 #define GEN_VEXT_VADC_VXM(NAME, ETYPE, H, DO_OP, CLEAR_FN)               \
1202 void HELPER(NAME)(void *vd, void *v0, target_ulong s1, void *vs2,        \
1203                   CPURISCVState *env, uint32_t desc)                     \
1204 {                                                                        \
1205     uint32_t mlen = vext_mlen(desc);                                     \
1206     uint32_t vl = env->vl;                                               \
1207     uint32_t esz = sizeof(ETYPE);                                        \
1208     uint32_t vlmax = vext_maxsz(desc) / esz;                             \
1209     uint32_t i;                                                          \
1210                                                                          \
1211     for (i = 0; i < vl; i++) {                                           \
1212         ETYPE s2 = *((ETYPE *)vs2 + H(i));                               \
1213         uint8_t carry = vext_elem_mask(v0, mlen, i);                     \
1214                                                                          \
1215         *((ETYPE *)vd + H(i)) = DO_OP(s2, (ETYPE)(target_long)s1, carry);\
1216     }                                                                    \
1217     CLEAR_FN(vd, vl, vl * esz, vlmax * esz);                             \
1218 }
1219 
1220 GEN_VEXT_VADC_VXM(vadc_vxm_b, uint8_t,  H1, DO_VADC, clearb)
1221 GEN_VEXT_VADC_VXM(vadc_vxm_h, uint16_t, H2, DO_VADC, clearh)
1222 GEN_VEXT_VADC_VXM(vadc_vxm_w, uint32_t, H4, DO_VADC, clearl)
1223 GEN_VEXT_VADC_VXM(vadc_vxm_d, uint64_t, H8, DO_VADC, clearq)
1224 
1225 GEN_VEXT_VADC_VXM(vsbc_vxm_b, uint8_t,  H1, DO_VSBC, clearb)
1226 GEN_VEXT_VADC_VXM(vsbc_vxm_h, uint16_t, H2, DO_VSBC, clearh)
1227 GEN_VEXT_VADC_VXM(vsbc_vxm_w, uint32_t, H4, DO_VSBC, clearl)
1228 GEN_VEXT_VADC_VXM(vsbc_vxm_d, uint64_t, H8, DO_VSBC, clearq)
1229 
1230 #define DO_MADC(N, M, C) (C ? (__typeof(N))(N + M + 1) <= N :           \
1231                           (__typeof(N))(N + M) < N)
1232 #define DO_MSBC(N, M, C) (C ? N <= M : N < M)
1233 
1234 #define GEN_VEXT_VMADC_VVM(NAME, ETYPE, H, DO_OP)             \
1235 void HELPER(NAME)(void *vd, void *v0, void *vs1, void *vs2,   \
1236                   CPURISCVState *env, uint32_t desc)          \
1237 {                                                             \
1238     uint32_t mlen = vext_mlen(desc);                          \
1239     uint32_t vl = env->vl;                                    \
1240     uint32_t vlmax = vext_maxsz(desc) / sizeof(ETYPE);        \
1241     uint32_t i;                                               \
1242                                                               \
1243     for (i = 0; i < vl; i++) {                                \
1244         ETYPE s1 = *((ETYPE *)vs1 + H(i));                    \
1245         ETYPE s2 = *((ETYPE *)vs2 + H(i));                    \
1246         uint8_t carry = vext_elem_mask(v0, mlen, i);          \
1247                                                               \
1248         vext_set_elem_mask(vd, mlen, i, DO_OP(s2, s1, carry));\
1249     }                                                         \
1250     for (; i < vlmax; i++) {                                  \
1251         vext_set_elem_mask(vd, mlen, i, 0);                   \
1252     }                                                         \
1253 }
1254 
1255 GEN_VEXT_VMADC_VVM(vmadc_vvm_b, uint8_t,  H1, DO_MADC)
1256 GEN_VEXT_VMADC_VVM(vmadc_vvm_h, uint16_t, H2, DO_MADC)
1257 GEN_VEXT_VMADC_VVM(vmadc_vvm_w, uint32_t, H4, DO_MADC)
1258 GEN_VEXT_VMADC_VVM(vmadc_vvm_d, uint64_t, H8, DO_MADC)
1259 
1260 GEN_VEXT_VMADC_VVM(vmsbc_vvm_b, uint8_t,  H1, DO_MSBC)
1261 GEN_VEXT_VMADC_VVM(vmsbc_vvm_h, uint16_t, H2, DO_MSBC)
1262 GEN_VEXT_VMADC_VVM(vmsbc_vvm_w, uint32_t, H4, DO_MSBC)
1263 GEN_VEXT_VMADC_VVM(vmsbc_vvm_d, uint64_t, H8, DO_MSBC)
1264 
1265 #define GEN_VEXT_VMADC_VXM(NAME, ETYPE, H, DO_OP)               \
1266 void HELPER(NAME)(void *vd, void *v0, target_ulong s1,          \
1267                   void *vs2, CPURISCVState *env, uint32_t desc) \
1268 {                                                               \
1269     uint32_t mlen = vext_mlen(desc);                            \
1270     uint32_t vl = env->vl;                                      \
1271     uint32_t vlmax = vext_maxsz(desc) / sizeof(ETYPE);          \
1272     uint32_t i;                                                 \
1273                                                                 \
1274     for (i = 0; i < vl; i++) {                                  \
1275         ETYPE s2 = *((ETYPE *)vs2 + H(i));                      \
1276         uint8_t carry = vext_elem_mask(v0, mlen, i);            \
1277                                                                 \
1278         vext_set_elem_mask(vd, mlen, i,                         \
1279                 DO_OP(s2, (ETYPE)(target_long)s1, carry));      \
1280     }                                                           \
1281     for (; i < vlmax; i++) {                                    \
1282         vext_set_elem_mask(vd, mlen, i, 0);                     \
1283     }                                                           \
1284 }
1285 
1286 GEN_VEXT_VMADC_VXM(vmadc_vxm_b, uint8_t,  H1, DO_MADC)
1287 GEN_VEXT_VMADC_VXM(vmadc_vxm_h, uint16_t, H2, DO_MADC)
1288 GEN_VEXT_VMADC_VXM(vmadc_vxm_w, uint32_t, H4, DO_MADC)
1289 GEN_VEXT_VMADC_VXM(vmadc_vxm_d, uint64_t, H8, DO_MADC)
1290 
1291 GEN_VEXT_VMADC_VXM(vmsbc_vxm_b, uint8_t,  H1, DO_MSBC)
1292 GEN_VEXT_VMADC_VXM(vmsbc_vxm_h, uint16_t, H2, DO_MSBC)
1293 GEN_VEXT_VMADC_VXM(vmsbc_vxm_w, uint32_t, H4, DO_MSBC)
1294 GEN_VEXT_VMADC_VXM(vmsbc_vxm_d, uint64_t, H8, DO_MSBC)
1295 
1296 /* Vector Bitwise Logical Instructions */
1297 RVVCALL(OPIVV2, vand_vv_b, OP_SSS_B, H1, H1, H1, DO_AND)
1298 RVVCALL(OPIVV2, vand_vv_h, OP_SSS_H, H2, H2, H2, DO_AND)
1299 RVVCALL(OPIVV2, vand_vv_w, OP_SSS_W, H4, H4, H4, DO_AND)
1300 RVVCALL(OPIVV2, vand_vv_d, OP_SSS_D, H8, H8, H8, DO_AND)
1301 RVVCALL(OPIVV2, vor_vv_b, OP_SSS_B, H1, H1, H1, DO_OR)
1302 RVVCALL(OPIVV2, vor_vv_h, OP_SSS_H, H2, H2, H2, DO_OR)
1303 RVVCALL(OPIVV2, vor_vv_w, OP_SSS_W, H4, H4, H4, DO_OR)
1304 RVVCALL(OPIVV2, vor_vv_d, OP_SSS_D, H8, H8, H8, DO_OR)
1305 RVVCALL(OPIVV2, vxor_vv_b, OP_SSS_B, H1, H1, H1, DO_XOR)
1306 RVVCALL(OPIVV2, vxor_vv_h, OP_SSS_H, H2, H2, H2, DO_XOR)
1307 RVVCALL(OPIVV2, vxor_vv_w, OP_SSS_W, H4, H4, H4, DO_XOR)
1308 RVVCALL(OPIVV2, vxor_vv_d, OP_SSS_D, H8, H8, H8, DO_XOR)
1309 GEN_VEXT_VV(vand_vv_b, 1, 1, clearb)
1310 GEN_VEXT_VV(vand_vv_h, 2, 2, clearh)
1311 GEN_VEXT_VV(vand_vv_w, 4, 4, clearl)
1312 GEN_VEXT_VV(vand_vv_d, 8, 8, clearq)
1313 GEN_VEXT_VV(vor_vv_b, 1, 1, clearb)
1314 GEN_VEXT_VV(vor_vv_h, 2, 2, clearh)
1315 GEN_VEXT_VV(vor_vv_w, 4, 4, clearl)
1316 GEN_VEXT_VV(vor_vv_d, 8, 8, clearq)
1317 GEN_VEXT_VV(vxor_vv_b, 1, 1, clearb)
1318 GEN_VEXT_VV(vxor_vv_h, 2, 2, clearh)
1319 GEN_VEXT_VV(vxor_vv_w, 4, 4, clearl)
1320 GEN_VEXT_VV(vxor_vv_d, 8, 8, clearq)
1321 
1322 RVVCALL(OPIVX2, vand_vx_b, OP_SSS_B, H1, H1, DO_AND)
1323 RVVCALL(OPIVX2, vand_vx_h, OP_SSS_H, H2, H2, DO_AND)
1324 RVVCALL(OPIVX2, vand_vx_w, OP_SSS_W, H4, H4, DO_AND)
1325 RVVCALL(OPIVX2, vand_vx_d, OP_SSS_D, H8, H8, DO_AND)
1326 RVVCALL(OPIVX2, vor_vx_b, OP_SSS_B, H1, H1, DO_OR)
1327 RVVCALL(OPIVX2, vor_vx_h, OP_SSS_H, H2, H2, DO_OR)
1328 RVVCALL(OPIVX2, vor_vx_w, OP_SSS_W, H4, H4, DO_OR)
1329 RVVCALL(OPIVX2, vor_vx_d, OP_SSS_D, H8, H8, DO_OR)
1330 RVVCALL(OPIVX2, vxor_vx_b, OP_SSS_B, H1, H1, DO_XOR)
1331 RVVCALL(OPIVX2, vxor_vx_h, OP_SSS_H, H2, H2, DO_XOR)
1332 RVVCALL(OPIVX2, vxor_vx_w, OP_SSS_W, H4, H4, DO_XOR)
1333 RVVCALL(OPIVX2, vxor_vx_d, OP_SSS_D, H8, H8, DO_XOR)
1334 GEN_VEXT_VX(vand_vx_b, 1, 1, clearb)
1335 GEN_VEXT_VX(vand_vx_h, 2, 2, clearh)
1336 GEN_VEXT_VX(vand_vx_w, 4, 4, clearl)
1337 GEN_VEXT_VX(vand_vx_d, 8, 8, clearq)
1338 GEN_VEXT_VX(vor_vx_b, 1, 1, clearb)
1339 GEN_VEXT_VX(vor_vx_h, 2, 2, clearh)
1340 GEN_VEXT_VX(vor_vx_w, 4, 4, clearl)
1341 GEN_VEXT_VX(vor_vx_d, 8, 8, clearq)
1342 GEN_VEXT_VX(vxor_vx_b, 1, 1, clearb)
1343 GEN_VEXT_VX(vxor_vx_h, 2, 2, clearh)
1344 GEN_VEXT_VX(vxor_vx_w, 4, 4, clearl)
1345 GEN_VEXT_VX(vxor_vx_d, 8, 8, clearq)
1346 
1347 /* Vector Single-Width Bit Shift Instructions */
1348 #define DO_SLL(N, M)  (N << (M))
1349 #define DO_SRL(N, M)  (N >> (M))
1350 
1351 /* generate the helpers for shift instructions with two vector operators */
1352 #define GEN_VEXT_SHIFT_VV(NAME, TS1, TS2, HS1, HS2, OP, MASK, CLEAR_FN)   \
1353 void HELPER(NAME)(void *vd, void *v0, void *vs1,                          \
1354                   void *vs2, CPURISCVState *env, uint32_t desc)           \
1355 {                                                                         \
1356     uint32_t mlen = vext_mlen(desc);                                      \
1357     uint32_t vm = vext_vm(desc);                                          \
1358     uint32_t vl = env->vl;                                                \
1359     uint32_t esz = sizeof(TS1);                                           \
1360     uint32_t vlmax = vext_maxsz(desc) / esz;                              \
1361     uint32_t i;                                                           \
1362                                                                           \
1363     for (i = 0; i < vl; i++) {                                            \
1364         if (!vm && !vext_elem_mask(v0, mlen, i)) {                        \
1365             continue;                                                     \
1366         }                                                                 \
1367         TS1 s1 = *((TS1 *)vs1 + HS1(i));                                  \
1368         TS2 s2 = *((TS2 *)vs2 + HS2(i));                                  \
1369         *((TS1 *)vd + HS1(i)) = OP(s2, s1 & MASK);                        \
1370     }                                                                     \
1371     CLEAR_FN(vd, vl, vl * esz, vlmax * esz);                              \
1372 }
1373 
1374 GEN_VEXT_SHIFT_VV(vsll_vv_b, uint8_t,  uint8_t, H1, H1, DO_SLL, 0x7, clearb)
1375 GEN_VEXT_SHIFT_VV(vsll_vv_h, uint16_t, uint16_t, H2, H2, DO_SLL, 0xf, clearh)
1376 GEN_VEXT_SHIFT_VV(vsll_vv_w, uint32_t, uint32_t, H4, H4, DO_SLL, 0x1f, clearl)
1377 GEN_VEXT_SHIFT_VV(vsll_vv_d, uint64_t, uint64_t, H8, H8, DO_SLL, 0x3f, clearq)
1378 
1379 GEN_VEXT_SHIFT_VV(vsrl_vv_b, uint8_t, uint8_t, H1, H1, DO_SRL, 0x7, clearb)
1380 GEN_VEXT_SHIFT_VV(vsrl_vv_h, uint16_t, uint16_t, H2, H2, DO_SRL, 0xf, clearh)
1381 GEN_VEXT_SHIFT_VV(vsrl_vv_w, uint32_t, uint32_t, H4, H4, DO_SRL, 0x1f, clearl)
1382 GEN_VEXT_SHIFT_VV(vsrl_vv_d, uint64_t, uint64_t, H8, H8, DO_SRL, 0x3f, clearq)
1383 
1384 GEN_VEXT_SHIFT_VV(vsra_vv_b, uint8_t,  int8_t, H1, H1, DO_SRL, 0x7, clearb)
1385 GEN_VEXT_SHIFT_VV(vsra_vv_h, uint16_t, int16_t, H2, H2, DO_SRL, 0xf, clearh)
1386 GEN_VEXT_SHIFT_VV(vsra_vv_w, uint32_t, int32_t, H4, H4, DO_SRL, 0x1f, clearl)
1387 GEN_VEXT_SHIFT_VV(vsra_vv_d, uint64_t, int64_t, H8, H8, DO_SRL, 0x3f, clearq)
1388 
1389 /* generate the helpers for shift instructions with one vector and one scalar */
1390 #define GEN_VEXT_SHIFT_VX(NAME, TD, TS2, HD, HS2, OP, MASK, CLEAR_FN) \
1391 void HELPER(NAME)(void *vd, void *v0, target_ulong s1,                \
1392         void *vs2, CPURISCVState *env, uint32_t desc)                 \
1393 {                                                                     \
1394     uint32_t mlen = vext_mlen(desc);                                  \
1395     uint32_t vm = vext_vm(desc);                                      \
1396     uint32_t vl = env->vl;                                            \
1397     uint32_t esz = sizeof(TD);                                        \
1398     uint32_t vlmax = vext_maxsz(desc) / esz;                          \
1399     uint32_t i;                                                       \
1400                                                                       \
1401     for (i = 0; i < vl; i++) {                                        \
1402         if (!vm && !vext_elem_mask(v0, mlen, i)) {                    \
1403             continue;                                                 \
1404         }                                                             \
1405         TS2 s2 = *((TS2 *)vs2 + HS2(i));                              \
1406         *((TD *)vd + HD(i)) = OP(s2, s1 & MASK);                      \
1407     }                                                                 \
1408     CLEAR_FN(vd, vl, vl * esz, vlmax * esz);                          \
1409 }
1410 
1411 GEN_VEXT_SHIFT_VX(vsll_vx_b, uint8_t, int8_t, H1, H1, DO_SLL, 0x7, clearb)
1412 GEN_VEXT_SHIFT_VX(vsll_vx_h, uint16_t, int16_t, H2, H2, DO_SLL, 0xf, clearh)
1413 GEN_VEXT_SHIFT_VX(vsll_vx_w, uint32_t, int32_t, H4, H4, DO_SLL, 0x1f, clearl)
1414 GEN_VEXT_SHIFT_VX(vsll_vx_d, uint64_t, int64_t, H8, H8, DO_SLL, 0x3f, clearq)
1415 
1416 GEN_VEXT_SHIFT_VX(vsrl_vx_b, uint8_t, uint8_t, H1, H1, DO_SRL, 0x7, clearb)
1417 GEN_VEXT_SHIFT_VX(vsrl_vx_h, uint16_t, uint16_t, H2, H2, DO_SRL, 0xf, clearh)
1418 GEN_VEXT_SHIFT_VX(vsrl_vx_w, uint32_t, uint32_t, H4, H4, DO_SRL, 0x1f, clearl)
1419 GEN_VEXT_SHIFT_VX(vsrl_vx_d, uint64_t, uint64_t, H8, H8, DO_SRL, 0x3f, clearq)
1420 
1421 GEN_VEXT_SHIFT_VX(vsra_vx_b, int8_t, int8_t, H1, H1, DO_SRL, 0x7, clearb)
1422 GEN_VEXT_SHIFT_VX(vsra_vx_h, int16_t, int16_t, H2, H2, DO_SRL, 0xf, clearh)
1423 GEN_VEXT_SHIFT_VX(vsra_vx_w, int32_t, int32_t, H4, H4, DO_SRL, 0x1f, clearl)
1424 GEN_VEXT_SHIFT_VX(vsra_vx_d, int64_t, int64_t, H8, H8, DO_SRL, 0x3f, clearq)
1425 
1426 /* Vector Narrowing Integer Right Shift Instructions */
1427 GEN_VEXT_SHIFT_VV(vnsrl_vv_b, uint8_t,  uint16_t, H1, H2, DO_SRL, 0xf, clearb)
1428 GEN_VEXT_SHIFT_VV(vnsrl_vv_h, uint16_t, uint32_t, H2, H4, DO_SRL, 0x1f, clearh)
1429 GEN_VEXT_SHIFT_VV(vnsrl_vv_w, uint32_t, uint64_t, H4, H8, DO_SRL, 0x3f, clearl)
1430 GEN_VEXT_SHIFT_VV(vnsra_vv_b, uint8_t,  int16_t, H1, H2, DO_SRL, 0xf, clearb)
1431 GEN_VEXT_SHIFT_VV(vnsra_vv_h, uint16_t, int32_t, H2, H4, DO_SRL, 0x1f, clearh)
1432 GEN_VEXT_SHIFT_VV(vnsra_vv_w, uint32_t, int64_t, H4, H8, DO_SRL, 0x3f, clearl)
1433 GEN_VEXT_SHIFT_VX(vnsrl_vx_b, uint8_t, uint16_t, H1, H2, DO_SRL, 0xf, clearb)
1434 GEN_VEXT_SHIFT_VX(vnsrl_vx_h, uint16_t, uint32_t, H2, H4, DO_SRL, 0x1f, clearh)
1435 GEN_VEXT_SHIFT_VX(vnsrl_vx_w, uint32_t, uint64_t, H4, H8, DO_SRL, 0x3f, clearl)
1436 GEN_VEXT_SHIFT_VX(vnsra_vx_b, int8_t, int16_t, H1, H2, DO_SRL, 0xf, clearb)
1437 GEN_VEXT_SHIFT_VX(vnsra_vx_h, int16_t, int32_t, H2, H4, DO_SRL, 0x1f, clearh)
1438 GEN_VEXT_SHIFT_VX(vnsra_vx_w, int32_t, int64_t, H4, H8, DO_SRL, 0x3f, clearl)
1439 
1440 /* Vector Integer Comparison Instructions */
1441 #define DO_MSEQ(N, M) (N == M)
1442 #define DO_MSNE(N, M) (N != M)
1443 #define DO_MSLT(N, M) (N < M)
1444 #define DO_MSLE(N, M) (N <= M)
1445 #define DO_MSGT(N, M) (N > M)
1446 
1447 #define GEN_VEXT_CMP_VV(NAME, ETYPE, H, DO_OP)                \
1448 void HELPER(NAME)(void *vd, void *v0, void *vs1, void *vs2,   \
1449                   CPURISCVState *env, uint32_t desc)          \
1450 {                                                             \
1451     uint32_t mlen = vext_mlen(desc);                          \
1452     uint32_t vm = vext_vm(desc);                              \
1453     uint32_t vl = env->vl;                                    \
1454     uint32_t vlmax = vext_maxsz(desc) / sizeof(ETYPE);        \
1455     uint32_t i;                                               \
1456                                                               \
1457     for (i = 0; i < vl; i++) {                                \
1458         ETYPE s1 = *((ETYPE *)vs1 + H(i));                    \
1459         ETYPE s2 = *((ETYPE *)vs2 + H(i));                    \
1460         if (!vm && !vext_elem_mask(v0, mlen, i)) {            \
1461             continue;                                         \
1462         }                                                     \
1463         vext_set_elem_mask(vd, mlen, i, DO_OP(s2, s1));       \
1464     }                                                         \
1465     for (; i < vlmax; i++) {                                  \
1466         vext_set_elem_mask(vd, mlen, i, 0);                   \
1467     }                                                         \
1468 }
1469 
1470 GEN_VEXT_CMP_VV(vmseq_vv_b, uint8_t,  H1, DO_MSEQ)
1471 GEN_VEXT_CMP_VV(vmseq_vv_h, uint16_t, H2, DO_MSEQ)
1472 GEN_VEXT_CMP_VV(vmseq_vv_w, uint32_t, H4, DO_MSEQ)
1473 GEN_VEXT_CMP_VV(vmseq_vv_d, uint64_t, H8, DO_MSEQ)
1474 
1475 GEN_VEXT_CMP_VV(vmsne_vv_b, uint8_t,  H1, DO_MSNE)
1476 GEN_VEXT_CMP_VV(vmsne_vv_h, uint16_t, H2, DO_MSNE)
1477 GEN_VEXT_CMP_VV(vmsne_vv_w, uint32_t, H4, DO_MSNE)
1478 GEN_VEXT_CMP_VV(vmsne_vv_d, uint64_t, H8, DO_MSNE)
1479 
1480 GEN_VEXT_CMP_VV(vmsltu_vv_b, uint8_t,  H1, DO_MSLT)
1481 GEN_VEXT_CMP_VV(vmsltu_vv_h, uint16_t, H2, DO_MSLT)
1482 GEN_VEXT_CMP_VV(vmsltu_vv_w, uint32_t, H4, DO_MSLT)
1483 GEN_VEXT_CMP_VV(vmsltu_vv_d, uint64_t, H8, DO_MSLT)
1484 
1485 GEN_VEXT_CMP_VV(vmslt_vv_b, int8_t,  H1, DO_MSLT)
1486 GEN_VEXT_CMP_VV(vmslt_vv_h, int16_t, H2, DO_MSLT)
1487 GEN_VEXT_CMP_VV(vmslt_vv_w, int32_t, H4, DO_MSLT)
1488 GEN_VEXT_CMP_VV(vmslt_vv_d, int64_t, H8, DO_MSLT)
1489 
1490 GEN_VEXT_CMP_VV(vmsleu_vv_b, uint8_t,  H1, DO_MSLE)
1491 GEN_VEXT_CMP_VV(vmsleu_vv_h, uint16_t, H2, DO_MSLE)
1492 GEN_VEXT_CMP_VV(vmsleu_vv_w, uint32_t, H4, DO_MSLE)
1493 GEN_VEXT_CMP_VV(vmsleu_vv_d, uint64_t, H8, DO_MSLE)
1494 
1495 GEN_VEXT_CMP_VV(vmsle_vv_b, int8_t,  H1, DO_MSLE)
1496 GEN_VEXT_CMP_VV(vmsle_vv_h, int16_t, H2, DO_MSLE)
1497 GEN_VEXT_CMP_VV(vmsle_vv_w, int32_t, H4, DO_MSLE)
1498 GEN_VEXT_CMP_VV(vmsle_vv_d, int64_t, H8, DO_MSLE)
1499 
1500 #define GEN_VEXT_CMP_VX(NAME, ETYPE, H, DO_OP)                      \
1501 void HELPER(NAME)(void *vd, void *v0, target_ulong s1, void *vs2,   \
1502                   CPURISCVState *env, uint32_t desc)                \
1503 {                                                                   \
1504     uint32_t mlen = vext_mlen(desc);                                \
1505     uint32_t vm = vext_vm(desc);                                    \
1506     uint32_t vl = env->vl;                                          \
1507     uint32_t vlmax = vext_maxsz(desc) / sizeof(ETYPE);              \
1508     uint32_t i;                                                     \
1509                                                                     \
1510     for (i = 0; i < vl; i++) {                                      \
1511         ETYPE s2 = *((ETYPE *)vs2 + H(i));                          \
1512         if (!vm && !vext_elem_mask(v0, mlen, i)) {                  \
1513             continue;                                               \
1514         }                                                           \
1515         vext_set_elem_mask(vd, mlen, i,                             \
1516                 DO_OP(s2, (ETYPE)(target_long)s1));                 \
1517     }                                                               \
1518     for (; i < vlmax; i++) {                                        \
1519         vext_set_elem_mask(vd, mlen, i, 0);                         \
1520     }                                                               \
1521 }
1522 
1523 GEN_VEXT_CMP_VX(vmseq_vx_b, uint8_t,  H1, DO_MSEQ)
1524 GEN_VEXT_CMP_VX(vmseq_vx_h, uint16_t, H2, DO_MSEQ)
1525 GEN_VEXT_CMP_VX(vmseq_vx_w, uint32_t, H4, DO_MSEQ)
1526 GEN_VEXT_CMP_VX(vmseq_vx_d, uint64_t, H8, DO_MSEQ)
1527 
1528 GEN_VEXT_CMP_VX(vmsne_vx_b, uint8_t,  H1, DO_MSNE)
1529 GEN_VEXT_CMP_VX(vmsne_vx_h, uint16_t, H2, DO_MSNE)
1530 GEN_VEXT_CMP_VX(vmsne_vx_w, uint32_t, H4, DO_MSNE)
1531 GEN_VEXT_CMP_VX(vmsne_vx_d, uint64_t, H8, DO_MSNE)
1532 
1533 GEN_VEXT_CMP_VX(vmsltu_vx_b, uint8_t,  H1, DO_MSLT)
1534 GEN_VEXT_CMP_VX(vmsltu_vx_h, uint16_t, H2, DO_MSLT)
1535 GEN_VEXT_CMP_VX(vmsltu_vx_w, uint32_t, H4, DO_MSLT)
1536 GEN_VEXT_CMP_VX(vmsltu_vx_d, uint64_t, H8, DO_MSLT)
1537 
1538 GEN_VEXT_CMP_VX(vmslt_vx_b, int8_t,  H1, DO_MSLT)
1539 GEN_VEXT_CMP_VX(vmslt_vx_h, int16_t, H2, DO_MSLT)
1540 GEN_VEXT_CMP_VX(vmslt_vx_w, int32_t, H4, DO_MSLT)
1541 GEN_VEXT_CMP_VX(vmslt_vx_d, int64_t, H8, DO_MSLT)
1542 
1543 GEN_VEXT_CMP_VX(vmsleu_vx_b, uint8_t,  H1, DO_MSLE)
1544 GEN_VEXT_CMP_VX(vmsleu_vx_h, uint16_t, H2, DO_MSLE)
1545 GEN_VEXT_CMP_VX(vmsleu_vx_w, uint32_t, H4, DO_MSLE)
1546 GEN_VEXT_CMP_VX(vmsleu_vx_d, uint64_t, H8, DO_MSLE)
1547 
1548 GEN_VEXT_CMP_VX(vmsle_vx_b, int8_t,  H1, DO_MSLE)
1549 GEN_VEXT_CMP_VX(vmsle_vx_h, int16_t, H2, DO_MSLE)
1550 GEN_VEXT_CMP_VX(vmsle_vx_w, int32_t, H4, DO_MSLE)
1551 GEN_VEXT_CMP_VX(vmsle_vx_d, int64_t, H8, DO_MSLE)
1552 
1553 GEN_VEXT_CMP_VX(vmsgtu_vx_b, uint8_t,  H1, DO_MSGT)
1554 GEN_VEXT_CMP_VX(vmsgtu_vx_h, uint16_t, H2, DO_MSGT)
1555 GEN_VEXT_CMP_VX(vmsgtu_vx_w, uint32_t, H4, DO_MSGT)
1556 GEN_VEXT_CMP_VX(vmsgtu_vx_d, uint64_t, H8, DO_MSGT)
1557 
1558 GEN_VEXT_CMP_VX(vmsgt_vx_b, int8_t,  H1, DO_MSGT)
1559 GEN_VEXT_CMP_VX(vmsgt_vx_h, int16_t, H2, DO_MSGT)
1560 GEN_VEXT_CMP_VX(vmsgt_vx_w, int32_t, H4, DO_MSGT)
1561 GEN_VEXT_CMP_VX(vmsgt_vx_d, int64_t, H8, DO_MSGT)
1562 
1563 /* Vector Integer Min/Max Instructions */
1564 RVVCALL(OPIVV2, vminu_vv_b, OP_UUU_B, H1, H1, H1, DO_MIN)
1565 RVVCALL(OPIVV2, vminu_vv_h, OP_UUU_H, H2, H2, H2, DO_MIN)
1566 RVVCALL(OPIVV2, vminu_vv_w, OP_UUU_W, H4, H4, H4, DO_MIN)
1567 RVVCALL(OPIVV2, vminu_vv_d, OP_UUU_D, H8, H8, H8, DO_MIN)
1568 RVVCALL(OPIVV2, vmin_vv_b, OP_SSS_B, H1, H1, H1, DO_MIN)
1569 RVVCALL(OPIVV2, vmin_vv_h, OP_SSS_H, H2, H2, H2, DO_MIN)
1570 RVVCALL(OPIVV2, vmin_vv_w, OP_SSS_W, H4, H4, H4, DO_MIN)
1571 RVVCALL(OPIVV2, vmin_vv_d, OP_SSS_D, H8, H8, H8, DO_MIN)
1572 RVVCALL(OPIVV2, vmaxu_vv_b, OP_UUU_B, H1, H1, H1, DO_MAX)
1573 RVVCALL(OPIVV2, vmaxu_vv_h, OP_UUU_H, H2, H2, H2, DO_MAX)
1574 RVVCALL(OPIVV2, vmaxu_vv_w, OP_UUU_W, H4, H4, H4, DO_MAX)
1575 RVVCALL(OPIVV2, vmaxu_vv_d, OP_UUU_D, H8, H8, H8, DO_MAX)
1576 RVVCALL(OPIVV2, vmax_vv_b, OP_SSS_B, H1, H1, H1, DO_MAX)
1577 RVVCALL(OPIVV2, vmax_vv_h, OP_SSS_H, H2, H2, H2, DO_MAX)
1578 RVVCALL(OPIVV2, vmax_vv_w, OP_SSS_W, H4, H4, H4, DO_MAX)
1579 RVVCALL(OPIVV2, vmax_vv_d, OP_SSS_D, H8, H8, H8, DO_MAX)
1580 GEN_VEXT_VV(vminu_vv_b, 1, 1, clearb)
1581 GEN_VEXT_VV(vminu_vv_h, 2, 2, clearh)
1582 GEN_VEXT_VV(vminu_vv_w, 4, 4, clearl)
1583 GEN_VEXT_VV(vminu_vv_d, 8, 8, clearq)
1584 GEN_VEXT_VV(vmin_vv_b, 1, 1, clearb)
1585 GEN_VEXT_VV(vmin_vv_h, 2, 2, clearh)
1586 GEN_VEXT_VV(vmin_vv_w, 4, 4, clearl)
1587 GEN_VEXT_VV(vmin_vv_d, 8, 8, clearq)
1588 GEN_VEXT_VV(vmaxu_vv_b, 1, 1, clearb)
1589 GEN_VEXT_VV(vmaxu_vv_h, 2, 2, clearh)
1590 GEN_VEXT_VV(vmaxu_vv_w, 4, 4, clearl)
1591 GEN_VEXT_VV(vmaxu_vv_d, 8, 8, clearq)
1592 GEN_VEXT_VV(vmax_vv_b, 1, 1, clearb)
1593 GEN_VEXT_VV(vmax_vv_h, 2, 2, clearh)
1594 GEN_VEXT_VV(vmax_vv_w, 4, 4, clearl)
1595 GEN_VEXT_VV(vmax_vv_d, 8, 8, clearq)
1596 
1597 RVVCALL(OPIVX2, vminu_vx_b, OP_UUU_B, H1, H1, DO_MIN)
1598 RVVCALL(OPIVX2, vminu_vx_h, OP_UUU_H, H2, H2, DO_MIN)
1599 RVVCALL(OPIVX2, vminu_vx_w, OP_UUU_W, H4, H4, DO_MIN)
1600 RVVCALL(OPIVX2, vminu_vx_d, OP_UUU_D, H8, H8, DO_MIN)
1601 RVVCALL(OPIVX2, vmin_vx_b, OP_SSS_B, H1, H1, DO_MIN)
1602 RVVCALL(OPIVX2, vmin_vx_h, OP_SSS_H, H2, H2, DO_MIN)
1603 RVVCALL(OPIVX2, vmin_vx_w, OP_SSS_W, H4, H4, DO_MIN)
1604 RVVCALL(OPIVX2, vmin_vx_d, OP_SSS_D, H8, H8, DO_MIN)
1605 RVVCALL(OPIVX2, vmaxu_vx_b, OP_UUU_B, H1, H1, DO_MAX)
1606 RVVCALL(OPIVX2, vmaxu_vx_h, OP_UUU_H, H2, H2, DO_MAX)
1607 RVVCALL(OPIVX2, vmaxu_vx_w, OP_UUU_W, H4, H4, DO_MAX)
1608 RVVCALL(OPIVX2, vmaxu_vx_d, OP_UUU_D, H8, H8, DO_MAX)
1609 RVVCALL(OPIVX2, vmax_vx_b, OP_SSS_B, H1, H1, DO_MAX)
1610 RVVCALL(OPIVX2, vmax_vx_h, OP_SSS_H, H2, H2, DO_MAX)
1611 RVVCALL(OPIVX2, vmax_vx_w, OP_SSS_W, H4, H4, DO_MAX)
1612 RVVCALL(OPIVX2, vmax_vx_d, OP_SSS_D, H8, H8, DO_MAX)
1613 GEN_VEXT_VX(vminu_vx_b, 1, 1, clearb)
1614 GEN_VEXT_VX(vminu_vx_h, 2, 2, clearh)
1615 GEN_VEXT_VX(vminu_vx_w, 4, 4, clearl)
1616 GEN_VEXT_VX(vminu_vx_d, 8, 8, clearq)
1617 GEN_VEXT_VX(vmin_vx_b, 1, 1, clearb)
1618 GEN_VEXT_VX(vmin_vx_h, 2, 2, clearh)
1619 GEN_VEXT_VX(vmin_vx_w, 4, 4, clearl)
1620 GEN_VEXT_VX(vmin_vx_d, 8, 8, clearq)
1621 GEN_VEXT_VX(vmaxu_vx_b, 1, 1, clearb)
1622 GEN_VEXT_VX(vmaxu_vx_h, 2, 2, clearh)
1623 GEN_VEXT_VX(vmaxu_vx_w, 4, 4, clearl)
1624 GEN_VEXT_VX(vmaxu_vx_d, 8, 8,  clearq)
1625 GEN_VEXT_VX(vmax_vx_b, 1, 1, clearb)
1626 GEN_VEXT_VX(vmax_vx_h, 2, 2, clearh)
1627 GEN_VEXT_VX(vmax_vx_w, 4, 4, clearl)
1628 GEN_VEXT_VX(vmax_vx_d, 8, 8, clearq)
1629 
1630 /* Vector Single-Width Integer Multiply Instructions */
1631 #define DO_MUL(N, M) (N * M)
1632 RVVCALL(OPIVV2, vmul_vv_b, OP_SSS_B, H1, H1, H1, DO_MUL)
1633 RVVCALL(OPIVV2, vmul_vv_h, OP_SSS_H, H2, H2, H2, DO_MUL)
1634 RVVCALL(OPIVV2, vmul_vv_w, OP_SSS_W, H4, H4, H4, DO_MUL)
1635 RVVCALL(OPIVV2, vmul_vv_d, OP_SSS_D, H8, H8, H8, DO_MUL)
1636 GEN_VEXT_VV(vmul_vv_b, 1, 1, clearb)
1637 GEN_VEXT_VV(vmul_vv_h, 2, 2, clearh)
1638 GEN_VEXT_VV(vmul_vv_w, 4, 4, clearl)
1639 GEN_VEXT_VV(vmul_vv_d, 8, 8, clearq)
1640 
1641 static int8_t do_mulh_b(int8_t s2, int8_t s1)
1642 {
1643     return (int16_t)s2 * (int16_t)s1 >> 8;
1644 }
1645 
1646 static int16_t do_mulh_h(int16_t s2, int16_t s1)
1647 {
1648     return (int32_t)s2 * (int32_t)s1 >> 16;
1649 }
1650 
1651 static int32_t do_mulh_w(int32_t s2, int32_t s1)
1652 {
1653     return (int64_t)s2 * (int64_t)s1 >> 32;
1654 }
1655 
1656 static int64_t do_mulh_d(int64_t s2, int64_t s1)
1657 {
1658     uint64_t hi_64, lo_64;
1659 
1660     muls64(&lo_64, &hi_64, s1, s2);
1661     return hi_64;
1662 }
1663 
1664 static uint8_t do_mulhu_b(uint8_t s2, uint8_t s1)
1665 {
1666     return (uint16_t)s2 * (uint16_t)s1 >> 8;
1667 }
1668 
1669 static uint16_t do_mulhu_h(uint16_t s2, uint16_t s1)
1670 {
1671     return (uint32_t)s2 * (uint32_t)s1 >> 16;
1672 }
1673 
1674 static uint32_t do_mulhu_w(uint32_t s2, uint32_t s1)
1675 {
1676     return (uint64_t)s2 * (uint64_t)s1 >> 32;
1677 }
1678 
1679 static uint64_t do_mulhu_d(uint64_t s2, uint64_t s1)
1680 {
1681     uint64_t hi_64, lo_64;
1682 
1683     mulu64(&lo_64, &hi_64, s2, s1);
1684     return hi_64;
1685 }
1686 
1687 static int8_t do_mulhsu_b(int8_t s2, uint8_t s1)
1688 {
1689     return (int16_t)s2 * (uint16_t)s1 >> 8;
1690 }
1691 
1692 static int16_t do_mulhsu_h(int16_t s2, uint16_t s1)
1693 {
1694     return (int32_t)s2 * (uint32_t)s1 >> 16;
1695 }
1696 
1697 static int32_t do_mulhsu_w(int32_t s2, uint32_t s1)
1698 {
1699     return (int64_t)s2 * (uint64_t)s1 >> 32;
1700 }
1701 
1702 /*
1703  * Let  A = signed operand,
1704  *      B = unsigned operand
1705  *      P = mulu64(A, B), unsigned product
1706  *
1707  * LET  X = 2 ** 64  - A, 2's complement of A
1708  *      SP = signed product
1709  * THEN
1710  *      IF A < 0
1711  *          SP = -X * B
1712  *             = -(2 ** 64 - A) * B
1713  *             = A * B - 2 ** 64 * B
1714  *             = P - 2 ** 64 * B
1715  *      ELSE
1716  *          SP = P
1717  * THEN
1718  *      HI_P -= (A < 0 ? B : 0)
1719  */
1720 
1721 static int64_t do_mulhsu_d(int64_t s2, uint64_t s1)
1722 {
1723     uint64_t hi_64, lo_64;
1724 
1725     mulu64(&lo_64, &hi_64, s2, s1);
1726 
1727     hi_64 -= s2 < 0 ? s1 : 0;
1728     return hi_64;
1729 }
1730 
1731 RVVCALL(OPIVV2, vmulh_vv_b, OP_SSS_B, H1, H1, H1, do_mulh_b)
1732 RVVCALL(OPIVV2, vmulh_vv_h, OP_SSS_H, H2, H2, H2, do_mulh_h)
1733 RVVCALL(OPIVV2, vmulh_vv_w, OP_SSS_W, H4, H4, H4, do_mulh_w)
1734 RVVCALL(OPIVV2, vmulh_vv_d, OP_SSS_D, H8, H8, H8, do_mulh_d)
1735 RVVCALL(OPIVV2, vmulhu_vv_b, OP_UUU_B, H1, H1, H1, do_mulhu_b)
1736 RVVCALL(OPIVV2, vmulhu_vv_h, OP_UUU_H, H2, H2, H2, do_mulhu_h)
1737 RVVCALL(OPIVV2, vmulhu_vv_w, OP_UUU_W, H4, H4, H4, do_mulhu_w)
1738 RVVCALL(OPIVV2, vmulhu_vv_d, OP_UUU_D, H8, H8, H8, do_mulhu_d)
1739 RVVCALL(OPIVV2, vmulhsu_vv_b, OP_SUS_B, H1, H1, H1, do_mulhsu_b)
1740 RVVCALL(OPIVV2, vmulhsu_vv_h, OP_SUS_H, H2, H2, H2, do_mulhsu_h)
1741 RVVCALL(OPIVV2, vmulhsu_vv_w, OP_SUS_W, H4, H4, H4, do_mulhsu_w)
1742 RVVCALL(OPIVV2, vmulhsu_vv_d, OP_SUS_D, H8, H8, H8, do_mulhsu_d)
1743 GEN_VEXT_VV(vmulh_vv_b, 1, 1, clearb)
1744 GEN_VEXT_VV(vmulh_vv_h, 2, 2, clearh)
1745 GEN_VEXT_VV(vmulh_vv_w, 4, 4, clearl)
1746 GEN_VEXT_VV(vmulh_vv_d, 8, 8, clearq)
1747 GEN_VEXT_VV(vmulhu_vv_b, 1, 1, clearb)
1748 GEN_VEXT_VV(vmulhu_vv_h, 2, 2, clearh)
1749 GEN_VEXT_VV(vmulhu_vv_w, 4, 4, clearl)
1750 GEN_VEXT_VV(vmulhu_vv_d, 8, 8, clearq)
1751 GEN_VEXT_VV(vmulhsu_vv_b, 1, 1, clearb)
1752 GEN_VEXT_VV(vmulhsu_vv_h, 2, 2, clearh)
1753 GEN_VEXT_VV(vmulhsu_vv_w, 4, 4, clearl)
1754 GEN_VEXT_VV(vmulhsu_vv_d, 8, 8, clearq)
1755 
1756 RVVCALL(OPIVX2, vmul_vx_b, OP_SSS_B, H1, H1, DO_MUL)
1757 RVVCALL(OPIVX2, vmul_vx_h, OP_SSS_H, H2, H2, DO_MUL)
1758 RVVCALL(OPIVX2, vmul_vx_w, OP_SSS_W, H4, H4, DO_MUL)
1759 RVVCALL(OPIVX2, vmul_vx_d, OP_SSS_D, H8, H8, DO_MUL)
1760 RVVCALL(OPIVX2, vmulh_vx_b, OP_SSS_B, H1, H1, do_mulh_b)
1761 RVVCALL(OPIVX2, vmulh_vx_h, OP_SSS_H, H2, H2, do_mulh_h)
1762 RVVCALL(OPIVX2, vmulh_vx_w, OP_SSS_W, H4, H4, do_mulh_w)
1763 RVVCALL(OPIVX2, vmulh_vx_d, OP_SSS_D, H8, H8, do_mulh_d)
1764 RVVCALL(OPIVX2, vmulhu_vx_b, OP_UUU_B, H1, H1, do_mulhu_b)
1765 RVVCALL(OPIVX2, vmulhu_vx_h, OP_UUU_H, H2, H2, do_mulhu_h)
1766 RVVCALL(OPIVX2, vmulhu_vx_w, OP_UUU_W, H4, H4, do_mulhu_w)
1767 RVVCALL(OPIVX2, vmulhu_vx_d, OP_UUU_D, H8, H8, do_mulhu_d)
1768 RVVCALL(OPIVX2, vmulhsu_vx_b, OP_SUS_B, H1, H1, do_mulhsu_b)
1769 RVVCALL(OPIVX2, vmulhsu_vx_h, OP_SUS_H, H2, H2, do_mulhsu_h)
1770 RVVCALL(OPIVX2, vmulhsu_vx_w, OP_SUS_W, H4, H4, do_mulhsu_w)
1771 RVVCALL(OPIVX2, vmulhsu_vx_d, OP_SUS_D, H8, H8, do_mulhsu_d)
1772 GEN_VEXT_VX(vmul_vx_b, 1, 1, clearb)
1773 GEN_VEXT_VX(vmul_vx_h, 2, 2, clearh)
1774 GEN_VEXT_VX(vmul_vx_w, 4, 4, clearl)
1775 GEN_VEXT_VX(vmul_vx_d, 8, 8, clearq)
1776 GEN_VEXT_VX(vmulh_vx_b, 1, 1, clearb)
1777 GEN_VEXT_VX(vmulh_vx_h, 2, 2, clearh)
1778 GEN_VEXT_VX(vmulh_vx_w, 4, 4, clearl)
1779 GEN_VEXT_VX(vmulh_vx_d, 8, 8, clearq)
1780 GEN_VEXT_VX(vmulhu_vx_b, 1, 1, clearb)
1781 GEN_VEXT_VX(vmulhu_vx_h, 2, 2, clearh)
1782 GEN_VEXT_VX(vmulhu_vx_w, 4, 4, clearl)
1783 GEN_VEXT_VX(vmulhu_vx_d, 8, 8, clearq)
1784 GEN_VEXT_VX(vmulhsu_vx_b, 1, 1, clearb)
1785 GEN_VEXT_VX(vmulhsu_vx_h, 2, 2, clearh)
1786 GEN_VEXT_VX(vmulhsu_vx_w, 4, 4, clearl)
1787 GEN_VEXT_VX(vmulhsu_vx_d, 8, 8, clearq)
1788 
1789 /* Vector Integer Divide Instructions */
1790 #define DO_DIVU(N, M) (unlikely(M == 0) ? (__typeof(N))(-1) : N / M)
1791 #define DO_REMU(N, M) (unlikely(M == 0) ? N : N % M)
1792 #define DO_DIV(N, M)  (unlikely(M == 0) ? (__typeof(N))(-1) :\
1793         unlikely((N == -N) && (M == (__typeof(N))(-1))) ? N : N / M)
1794 #define DO_REM(N, M)  (unlikely(M == 0) ? N :\
1795         unlikely((N == -N) && (M == (__typeof(N))(-1))) ? 0 : N % M)
1796 
1797 RVVCALL(OPIVV2, vdivu_vv_b, OP_UUU_B, H1, H1, H1, DO_DIVU)
1798 RVVCALL(OPIVV2, vdivu_vv_h, OP_UUU_H, H2, H2, H2, DO_DIVU)
1799 RVVCALL(OPIVV2, vdivu_vv_w, OP_UUU_W, H4, H4, H4, DO_DIVU)
1800 RVVCALL(OPIVV2, vdivu_vv_d, OP_UUU_D, H8, H8, H8, DO_DIVU)
1801 RVVCALL(OPIVV2, vdiv_vv_b, OP_SSS_B, H1, H1, H1, DO_DIV)
1802 RVVCALL(OPIVV2, vdiv_vv_h, OP_SSS_H, H2, H2, H2, DO_DIV)
1803 RVVCALL(OPIVV2, vdiv_vv_w, OP_SSS_W, H4, H4, H4, DO_DIV)
1804 RVVCALL(OPIVV2, vdiv_vv_d, OP_SSS_D, H8, H8, H8, DO_DIV)
1805 RVVCALL(OPIVV2, vremu_vv_b, OP_UUU_B, H1, H1, H1, DO_REMU)
1806 RVVCALL(OPIVV2, vremu_vv_h, OP_UUU_H, H2, H2, H2, DO_REMU)
1807 RVVCALL(OPIVV2, vremu_vv_w, OP_UUU_W, H4, H4, H4, DO_REMU)
1808 RVVCALL(OPIVV2, vremu_vv_d, OP_UUU_D, H8, H8, H8, DO_REMU)
1809 RVVCALL(OPIVV2, vrem_vv_b, OP_SSS_B, H1, H1, H1, DO_REM)
1810 RVVCALL(OPIVV2, vrem_vv_h, OP_SSS_H, H2, H2, H2, DO_REM)
1811 RVVCALL(OPIVV2, vrem_vv_w, OP_SSS_W, H4, H4, H4, DO_REM)
1812 RVVCALL(OPIVV2, vrem_vv_d, OP_SSS_D, H8, H8, H8, DO_REM)
1813 GEN_VEXT_VV(vdivu_vv_b, 1, 1, clearb)
1814 GEN_VEXT_VV(vdivu_vv_h, 2, 2, clearh)
1815 GEN_VEXT_VV(vdivu_vv_w, 4, 4, clearl)
1816 GEN_VEXT_VV(vdivu_vv_d, 8, 8, clearq)
1817 GEN_VEXT_VV(vdiv_vv_b, 1, 1, clearb)
1818 GEN_VEXT_VV(vdiv_vv_h, 2, 2, clearh)
1819 GEN_VEXT_VV(vdiv_vv_w, 4, 4, clearl)
1820 GEN_VEXT_VV(vdiv_vv_d, 8, 8, clearq)
1821 GEN_VEXT_VV(vremu_vv_b, 1, 1, clearb)
1822 GEN_VEXT_VV(vremu_vv_h, 2, 2, clearh)
1823 GEN_VEXT_VV(vremu_vv_w, 4, 4, clearl)
1824 GEN_VEXT_VV(vremu_vv_d, 8, 8, clearq)
1825 GEN_VEXT_VV(vrem_vv_b, 1, 1, clearb)
1826 GEN_VEXT_VV(vrem_vv_h, 2, 2, clearh)
1827 GEN_VEXT_VV(vrem_vv_w, 4, 4, clearl)
1828 GEN_VEXT_VV(vrem_vv_d, 8, 8, clearq)
1829 
1830 RVVCALL(OPIVX2, vdivu_vx_b, OP_UUU_B, H1, H1, DO_DIVU)
1831 RVVCALL(OPIVX2, vdivu_vx_h, OP_UUU_H, H2, H2, DO_DIVU)
1832 RVVCALL(OPIVX2, vdivu_vx_w, OP_UUU_W, H4, H4, DO_DIVU)
1833 RVVCALL(OPIVX2, vdivu_vx_d, OP_UUU_D, H8, H8, DO_DIVU)
1834 RVVCALL(OPIVX2, vdiv_vx_b, OP_SSS_B, H1, H1, DO_DIV)
1835 RVVCALL(OPIVX2, vdiv_vx_h, OP_SSS_H, H2, H2, DO_DIV)
1836 RVVCALL(OPIVX2, vdiv_vx_w, OP_SSS_W, H4, H4, DO_DIV)
1837 RVVCALL(OPIVX2, vdiv_vx_d, OP_SSS_D, H8, H8, DO_DIV)
1838 RVVCALL(OPIVX2, vremu_vx_b, OP_UUU_B, H1, H1, DO_REMU)
1839 RVVCALL(OPIVX2, vremu_vx_h, OP_UUU_H, H2, H2, DO_REMU)
1840 RVVCALL(OPIVX2, vremu_vx_w, OP_UUU_W, H4, H4, DO_REMU)
1841 RVVCALL(OPIVX2, vremu_vx_d, OP_UUU_D, H8, H8, DO_REMU)
1842 RVVCALL(OPIVX2, vrem_vx_b, OP_SSS_B, H1, H1, DO_REM)
1843 RVVCALL(OPIVX2, vrem_vx_h, OP_SSS_H, H2, H2, DO_REM)
1844 RVVCALL(OPIVX2, vrem_vx_w, OP_SSS_W, H4, H4, DO_REM)
1845 RVVCALL(OPIVX2, vrem_vx_d, OP_SSS_D, H8, H8, DO_REM)
1846 GEN_VEXT_VX(vdivu_vx_b, 1, 1, clearb)
1847 GEN_VEXT_VX(vdivu_vx_h, 2, 2, clearh)
1848 GEN_VEXT_VX(vdivu_vx_w, 4, 4, clearl)
1849 GEN_VEXT_VX(vdivu_vx_d, 8, 8, clearq)
1850 GEN_VEXT_VX(vdiv_vx_b, 1, 1, clearb)
1851 GEN_VEXT_VX(vdiv_vx_h, 2, 2, clearh)
1852 GEN_VEXT_VX(vdiv_vx_w, 4, 4, clearl)
1853 GEN_VEXT_VX(vdiv_vx_d, 8, 8, clearq)
1854 GEN_VEXT_VX(vremu_vx_b, 1, 1, clearb)
1855 GEN_VEXT_VX(vremu_vx_h, 2, 2, clearh)
1856 GEN_VEXT_VX(vremu_vx_w, 4, 4, clearl)
1857 GEN_VEXT_VX(vremu_vx_d, 8, 8, clearq)
1858 GEN_VEXT_VX(vrem_vx_b, 1, 1, clearb)
1859 GEN_VEXT_VX(vrem_vx_h, 2, 2, clearh)
1860 GEN_VEXT_VX(vrem_vx_w, 4, 4, clearl)
1861 GEN_VEXT_VX(vrem_vx_d, 8, 8, clearq)
1862 
1863 /* Vector Widening Integer Multiply Instructions */
1864 RVVCALL(OPIVV2, vwmul_vv_b, WOP_SSS_B, H2, H1, H1, DO_MUL)
1865 RVVCALL(OPIVV2, vwmul_vv_h, WOP_SSS_H, H4, H2, H2, DO_MUL)
1866 RVVCALL(OPIVV2, vwmul_vv_w, WOP_SSS_W, H8, H4, H4, DO_MUL)
1867 RVVCALL(OPIVV2, vwmulu_vv_b, WOP_UUU_B, H2, H1, H1, DO_MUL)
1868 RVVCALL(OPIVV2, vwmulu_vv_h, WOP_UUU_H, H4, H2, H2, DO_MUL)
1869 RVVCALL(OPIVV2, vwmulu_vv_w, WOP_UUU_W, H8, H4, H4, DO_MUL)
1870 RVVCALL(OPIVV2, vwmulsu_vv_b, WOP_SUS_B, H2, H1, H1, DO_MUL)
1871 RVVCALL(OPIVV2, vwmulsu_vv_h, WOP_SUS_H, H4, H2, H2, DO_MUL)
1872 RVVCALL(OPIVV2, vwmulsu_vv_w, WOP_SUS_W, H8, H4, H4, DO_MUL)
1873 GEN_VEXT_VV(vwmul_vv_b, 1, 2, clearh)
1874 GEN_VEXT_VV(vwmul_vv_h, 2, 4, clearl)
1875 GEN_VEXT_VV(vwmul_vv_w, 4, 8, clearq)
1876 GEN_VEXT_VV(vwmulu_vv_b, 1, 2, clearh)
1877 GEN_VEXT_VV(vwmulu_vv_h, 2, 4, clearl)
1878 GEN_VEXT_VV(vwmulu_vv_w, 4, 8, clearq)
1879 GEN_VEXT_VV(vwmulsu_vv_b, 1, 2, clearh)
1880 GEN_VEXT_VV(vwmulsu_vv_h, 2, 4, clearl)
1881 GEN_VEXT_VV(vwmulsu_vv_w, 4, 8, clearq)
1882 
1883 RVVCALL(OPIVX2, vwmul_vx_b, WOP_SSS_B, H2, H1, DO_MUL)
1884 RVVCALL(OPIVX2, vwmul_vx_h, WOP_SSS_H, H4, H2, DO_MUL)
1885 RVVCALL(OPIVX2, vwmul_vx_w, WOP_SSS_W, H8, H4, DO_MUL)
1886 RVVCALL(OPIVX2, vwmulu_vx_b, WOP_UUU_B, H2, H1, DO_MUL)
1887 RVVCALL(OPIVX2, vwmulu_vx_h, WOP_UUU_H, H4, H2, DO_MUL)
1888 RVVCALL(OPIVX2, vwmulu_vx_w, WOP_UUU_W, H8, H4, DO_MUL)
1889 RVVCALL(OPIVX2, vwmulsu_vx_b, WOP_SUS_B, H2, H1, DO_MUL)
1890 RVVCALL(OPIVX2, vwmulsu_vx_h, WOP_SUS_H, H4, H2, DO_MUL)
1891 RVVCALL(OPIVX2, vwmulsu_vx_w, WOP_SUS_W, H8, H4, DO_MUL)
1892 GEN_VEXT_VX(vwmul_vx_b, 1, 2, clearh)
1893 GEN_VEXT_VX(vwmul_vx_h, 2, 4, clearl)
1894 GEN_VEXT_VX(vwmul_vx_w, 4, 8, clearq)
1895 GEN_VEXT_VX(vwmulu_vx_b, 1, 2, clearh)
1896 GEN_VEXT_VX(vwmulu_vx_h, 2, 4, clearl)
1897 GEN_VEXT_VX(vwmulu_vx_w, 4, 8, clearq)
1898 GEN_VEXT_VX(vwmulsu_vx_b, 1, 2, clearh)
1899 GEN_VEXT_VX(vwmulsu_vx_h, 2, 4, clearl)
1900 GEN_VEXT_VX(vwmulsu_vx_w, 4, 8, clearq)
1901 
1902 /* Vector Single-Width Integer Multiply-Add Instructions */
1903 #define OPIVV3(NAME, TD, T1, T2, TX1, TX2, HD, HS1, HS2, OP)   \
1904 static void do_##NAME(void *vd, void *vs1, void *vs2, int i)       \
1905 {                                                                  \
1906     TX1 s1 = *((T1 *)vs1 + HS1(i));                                \
1907     TX2 s2 = *((T2 *)vs2 + HS2(i));                                \
1908     TD d = *((TD *)vd + HD(i));                                    \
1909     *((TD *)vd + HD(i)) = OP(s2, s1, d);                           \
1910 }
1911 
1912 #define DO_MACC(N, M, D) (M * N + D)
1913 #define DO_NMSAC(N, M, D) (-(M * N) + D)
1914 #define DO_MADD(N, M, D) (M * D + N)
1915 #define DO_NMSUB(N, M, D) (-(M * D) + N)
1916 RVVCALL(OPIVV3, vmacc_vv_b, OP_SSS_B, H1, H1, H1, DO_MACC)
1917 RVVCALL(OPIVV3, vmacc_vv_h, OP_SSS_H, H2, H2, H2, DO_MACC)
1918 RVVCALL(OPIVV3, vmacc_vv_w, OP_SSS_W, H4, H4, H4, DO_MACC)
1919 RVVCALL(OPIVV3, vmacc_vv_d, OP_SSS_D, H8, H8, H8, DO_MACC)
1920 RVVCALL(OPIVV3, vnmsac_vv_b, OP_SSS_B, H1, H1, H1, DO_NMSAC)
1921 RVVCALL(OPIVV3, vnmsac_vv_h, OP_SSS_H, H2, H2, H2, DO_NMSAC)
1922 RVVCALL(OPIVV3, vnmsac_vv_w, OP_SSS_W, H4, H4, H4, DO_NMSAC)
1923 RVVCALL(OPIVV3, vnmsac_vv_d, OP_SSS_D, H8, H8, H8, DO_NMSAC)
1924 RVVCALL(OPIVV3, vmadd_vv_b, OP_SSS_B, H1, H1, H1, DO_MADD)
1925 RVVCALL(OPIVV3, vmadd_vv_h, OP_SSS_H, H2, H2, H2, DO_MADD)
1926 RVVCALL(OPIVV3, vmadd_vv_w, OP_SSS_W, H4, H4, H4, DO_MADD)
1927 RVVCALL(OPIVV3, vmadd_vv_d, OP_SSS_D, H8, H8, H8, DO_MADD)
1928 RVVCALL(OPIVV3, vnmsub_vv_b, OP_SSS_B, H1, H1, H1, DO_NMSUB)
1929 RVVCALL(OPIVV3, vnmsub_vv_h, OP_SSS_H, H2, H2, H2, DO_NMSUB)
1930 RVVCALL(OPIVV3, vnmsub_vv_w, OP_SSS_W, H4, H4, H4, DO_NMSUB)
1931 RVVCALL(OPIVV3, vnmsub_vv_d, OP_SSS_D, H8, H8, H8, DO_NMSUB)
1932 GEN_VEXT_VV(vmacc_vv_b, 1, 1, clearb)
1933 GEN_VEXT_VV(vmacc_vv_h, 2, 2, clearh)
1934 GEN_VEXT_VV(vmacc_vv_w, 4, 4, clearl)
1935 GEN_VEXT_VV(vmacc_vv_d, 8, 8, clearq)
1936 GEN_VEXT_VV(vnmsac_vv_b, 1, 1, clearb)
1937 GEN_VEXT_VV(vnmsac_vv_h, 2, 2, clearh)
1938 GEN_VEXT_VV(vnmsac_vv_w, 4, 4, clearl)
1939 GEN_VEXT_VV(vnmsac_vv_d, 8, 8, clearq)
1940 GEN_VEXT_VV(vmadd_vv_b, 1, 1, clearb)
1941 GEN_VEXT_VV(vmadd_vv_h, 2, 2, clearh)
1942 GEN_VEXT_VV(vmadd_vv_w, 4, 4, clearl)
1943 GEN_VEXT_VV(vmadd_vv_d, 8, 8, clearq)
1944 GEN_VEXT_VV(vnmsub_vv_b, 1, 1, clearb)
1945 GEN_VEXT_VV(vnmsub_vv_h, 2, 2, clearh)
1946 GEN_VEXT_VV(vnmsub_vv_w, 4, 4, clearl)
1947 GEN_VEXT_VV(vnmsub_vv_d, 8, 8, clearq)
1948 
1949 #define OPIVX3(NAME, TD, T1, T2, TX1, TX2, HD, HS2, OP)             \
1950 static void do_##NAME(void *vd, target_long s1, void *vs2, int i)   \
1951 {                                                                   \
1952     TX2 s2 = *((T2 *)vs2 + HS2(i));                                 \
1953     TD d = *((TD *)vd + HD(i));                                     \
1954     *((TD *)vd + HD(i)) = OP(s2, (TX1)(T1)s1, d);                   \
1955 }
1956 
1957 RVVCALL(OPIVX3, vmacc_vx_b, OP_SSS_B, H1, H1, DO_MACC)
1958 RVVCALL(OPIVX3, vmacc_vx_h, OP_SSS_H, H2, H2, DO_MACC)
1959 RVVCALL(OPIVX3, vmacc_vx_w, OP_SSS_W, H4, H4, DO_MACC)
1960 RVVCALL(OPIVX3, vmacc_vx_d, OP_SSS_D, H8, H8, DO_MACC)
1961 RVVCALL(OPIVX3, vnmsac_vx_b, OP_SSS_B, H1, H1, DO_NMSAC)
1962 RVVCALL(OPIVX3, vnmsac_vx_h, OP_SSS_H, H2, H2, DO_NMSAC)
1963 RVVCALL(OPIVX3, vnmsac_vx_w, OP_SSS_W, H4, H4, DO_NMSAC)
1964 RVVCALL(OPIVX3, vnmsac_vx_d, OP_SSS_D, H8, H8, DO_NMSAC)
1965 RVVCALL(OPIVX3, vmadd_vx_b, OP_SSS_B, H1, H1, DO_MADD)
1966 RVVCALL(OPIVX3, vmadd_vx_h, OP_SSS_H, H2, H2, DO_MADD)
1967 RVVCALL(OPIVX3, vmadd_vx_w, OP_SSS_W, H4, H4, DO_MADD)
1968 RVVCALL(OPIVX3, vmadd_vx_d, OP_SSS_D, H8, H8, DO_MADD)
1969 RVVCALL(OPIVX3, vnmsub_vx_b, OP_SSS_B, H1, H1, DO_NMSUB)
1970 RVVCALL(OPIVX3, vnmsub_vx_h, OP_SSS_H, H2, H2, DO_NMSUB)
1971 RVVCALL(OPIVX3, vnmsub_vx_w, OP_SSS_W, H4, H4, DO_NMSUB)
1972 RVVCALL(OPIVX3, vnmsub_vx_d, OP_SSS_D, H8, H8, DO_NMSUB)
1973 GEN_VEXT_VX(vmacc_vx_b, 1, 1, clearb)
1974 GEN_VEXT_VX(vmacc_vx_h, 2, 2, clearh)
1975 GEN_VEXT_VX(vmacc_vx_w, 4, 4, clearl)
1976 GEN_VEXT_VX(vmacc_vx_d, 8, 8, clearq)
1977 GEN_VEXT_VX(vnmsac_vx_b, 1, 1, clearb)
1978 GEN_VEXT_VX(vnmsac_vx_h, 2, 2, clearh)
1979 GEN_VEXT_VX(vnmsac_vx_w, 4, 4, clearl)
1980 GEN_VEXT_VX(vnmsac_vx_d, 8, 8, clearq)
1981 GEN_VEXT_VX(vmadd_vx_b, 1, 1, clearb)
1982 GEN_VEXT_VX(vmadd_vx_h, 2, 2, clearh)
1983 GEN_VEXT_VX(vmadd_vx_w, 4, 4, clearl)
1984 GEN_VEXT_VX(vmadd_vx_d, 8, 8, clearq)
1985 GEN_VEXT_VX(vnmsub_vx_b, 1, 1, clearb)
1986 GEN_VEXT_VX(vnmsub_vx_h, 2, 2, clearh)
1987 GEN_VEXT_VX(vnmsub_vx_w, 4, 4, clearl)
1988 GEN_VEXT_VX(vnmsub_vx_d, 8, 8, clearq)
1989 
1990 /* Vector Widening Integer Multiply-Add Instructions */
1991 RVVCALL(OPIVV3, vwmaccu_vv_b, WOP_UUU_B, H2, H1, H1, DO_MACC)
1992 RVVCALL(OPIVV3, vwmaccu_vv_h, WOP_UUU_H, H4, H2, H2, DO_MACC)
1993 RVVCALL(OPIVV3, vwmaccu_vv_w, WOP_UUU_W, H8, H4, H4, DO_MACC)
1994 RVVCALL(OPIVV3, vwmacc_vv_b, WOP_SSS_B, H2, H1, H1, DO_MACC)
1995 RVVCALL(OPIVV3, vwmacc_vv_h, WOP_SSS_H, H4, H2, H2, DO_MACC)
1996 RVVCALL(OPIVV3, vwmacc_vv_w, WOP_SSS_W, H8, H4, H4, DO_MACC)
1997 RVVCALL(OPIVV3, vwmaccsu_vv_b, WOP_SSU_B, H2, H1, H1, DO_MACC)
1998 RVVCALL(OPIVV3, vwmaccsu_vv_h, WOP_SSU_H, H4, H2, H2, DO_MACC)
1999 RVVCALL(OPIVV3, vwmaccsu_vv_w, WOP_SSU_W, H8, H4, H4, DO_MACC)
2000 GEN_VEXT_VV(vwmaccu_vv_b, 1, 2, clearh)
2001 GEN_VEXT_VV(vwmaccu_vv_h, 2, 4, clearl)
2002 GEN_VEXT_VV(vwmaccu_vv_w, 4, 8, clearq)
2003 GEN_VEXT_VV(vwmacc_vv_b, 1, 2, clearh)
2004 GEN_VEXT_VV(vwmacc_vv_h, 2, 4, clearl)
2005 GEN_VEXT_VV(vwmacc_vv_w, 4, 8, clearq)
2006 GEN_VEXT_VV(vwmaccsu_vv_b, 1, 2, clearh)
2007 GEN_VEXT_VV(vwmaccsu_vv_h, 2, 4, clearl)
2008 GEN_VEXT_VV(vwmaccsu_vv_w, 4, 8, clearq)
2009 
2010 RVVCALL(OPIVX3, vwmaccu_vx_b, WOP_UUU_B, H2, H1, DO_MACC)
2011 RVVCALL(OPIVX3, vwmaccu_vx_h, WOP_UUU_H, H4, H2, DO_MACC)
2012 RVVCALL(OPIVX3, vwmaccu_vx_w, WOP_UUU_W, H8, H4, DO_MACC)
2013 RVVCALL(OPIVX3, vwmacc_vx_b, WOP_SSS_B, H2, H1, DO_MACC)
2014 RVVCALL(OPIVX3, vwmacc_vx_h, WOP_SSS_H, H4, H2, DO_MACC)
2015 RVVCALL(OPIVX3, vwmacc_vx_w, WOP_SSS_W, H8, H4, DO_MACC)
2016 RVVCALL(OPIVX3, vwmaccsu_vx_b, WOP_SSU_B, H2, H1, DO_MACC)
2017 RVVCALL(OPIVX3, vwmaccsu_vx_h, WOP_SSU_H, H4, H2, DO_MACC)
2018 RVVCALL(OPIVX3, vwmaccsu_vx_w, WOP_SSU_W, H8, H4, DO_MACC)
2019 RVVCALL(OPIVX3, vwmaccus_vx_b, WOP_SUS_B, H2, H1, DO_MACC)
2020 RVVCALL(OPIVX3, vwmaccus_vx_h, WOP_SUS_H, H4, H2, DO_MACC)
2021 RVVCALL(OPIVX3, vwmaccus_vx_w, WOP_SUS_W, H8, H4, DO_MACC)
2022 GEN_VEXT_VX(vwmaccu_vx_b, 1, 2, clearh)
2023 GEN_VEXT_VX(vwmaccu_vx_h, 2, 4, clearl)
2024 GEN_VEXT_VX(vwmaccu_vx_w, 4, 8, clearq)
2025 GEN_VEXT_VX(vwmacc_vx_b, 1, 2, clearh)
2026 GEN_VEXT_VX(vwmacc_vx_h, 2, 4, clearl)
2027 GEN_VEXT_VX(vwmacc_vx_w, 4, 8, clearq)
2028 GEN_VEXT_VX(vwmaccsu_vx_b, 1, 2, clearh)
2029 GEN_VEXT_VX(vwmaccsu_vx_h, 2, 4, clearl)
2030 GEN_VEXT_VX(vwmaccsu_vx_w, 4, 8, clearq)
2031 GEN_VEXT_VX(vwmaccus_vx_b, 1, 2, clearh)
2032 GEN_VEXT_VX(vwmaccus_vx_h, 2, 4, clearl)
2033 GEN_VEXT_VX(vwmaccus_vx_w, 4, 8, clearq)
2034 
2035 /* Vector Integer Merge and Move Instructions */
2036 #define GEN_VEXT_VMV_VV(NAME, ETYPE, H, CLEAR_FN)                    \
2037 void HELPER(NAME)(void *vd, void *vs1, CPURISCVState *env,           \
2038                   uint32_t desc)                                     \
2039 {                                                                    \
2040     uint32_t vl = env->vl;                                           \
2041     uint32_t esz = sizeof(ETYPE);                                    \
2042     uint32_t vlmax = vext_maxsz(desc) / esz;                         \
2043     uint32_t i;                                                      \
2044                                                                      \
2045     for (i = 0; i < vl; i++) {                                       \
2046         ETYPE s1 = *((ETYPE *)vs1 + H(i));                           \
2047         *((ETYPE *)vd + H(i)) = s1;                                  \
2048     }                                                                \
2049     CLEAR_FN(vd, vl, vl * esz, vlmax * esz);                         \
2050 }
2051 
2052 GEN_VEXT_VMV_VV(vmv_v_v_b, int8_t,  H1, clearb)
2053 GEN_VEXT_VMV_VV(vmv_v_v_h, int16_t, H2, clearh)
2054 GEN_VEXT_VMV_VV(vmv_v_v_w, int32_t, H4, clearl)
2055 GEN_VEXT_VMV_VV(vmv_v_v_d, int64_t, H8, clearq)
2056 
2057 #define GEN_VEXT_VMV_VX(NAME, ETYPE, H, CLEAR_FN)                    \
2058 void HELPER(NAME)(void *vd, uint64_t s1, CPURISCVState *env,         \
2059                   uint32_t desc)                                     \
2060 {                                                                    \
2061     uint32_t vl = env->vl;                                           \
2062     uint32_t esz = sizeof(ETYPE);                                    \
2063     uint32_t vlmax = vext_maxsz(desc) / esz;                         \
2064     uint32_t i;                                                      \
2065                                                                      \
2066     for (i = 0; i < vl; i++) {                                       \
2067         *((ETYPE *)vd + H(i)) = (ETYPE)s1;                           \
2068     }                                                                \
2069     CLEAR_FN(vd, vl, vl * esz, vlmax * esz);                         \
2070 }
2071 
2072 GEN_VEXT_VMV_VX(vmv_v_x_b, int8_t,  H1, clearb)
2073 GEN_VEXT_VMV_VX(vmv_v_x_h, int16_t, H2, clearh)
2074 GEN_VEXT_VMV_VX(vmv_v_x_w, int32_t, H4, clearl)
2075 GEN_VEXT_VMV_VX(vmv_v_x_d, int64_t, H8, clearq)
2076 
2077 #define GEN_VEXT_VMERGE_VV(NAME, ETYPE, H, CLEAR_FN)                 \
2078 void HELPER(NAME)(void *vd, void *v0, void *vs1, void *vs2,          \
2079                   CPURISCVState *env, uint32_t desc)                 \
2080 {                                                                    \
2081     uint32_t mlen = vext_mlen(desc);                                 \
2082     uint32_t vl = env->vl;                                           \
2083     uint32_t esz = sizeof(ETYPE);                                    \
2084     uint32_t vlmax = vext_maxsz(desc) / esz;                         \
2085     uint32_t i;                                                      \
2086                                                                      \
2087     for (i = 0; i < vl; i++) {                                       \
2088         ETYPE *vt = (!vext_elem_mask(v0, mlen, i) ? vs2 : vs1);      \
2089         *((ETYPE *)vd + H(i)) = *(vt + H(i));                        \
2090     }                                                                \
2091     CLEAR_FN(vd, vl, vl * esz, vlmax * esz);                         \
2092 }
2093 
2094 GEN_VEXT_VMERGE_VV(vmerge_vvm_b, int8_t,  H1, clearb)
2095 GEN_VEXT_VMERGE_VV(vmerge_vvm_h, int16_t, H2, clearh)
2096 GEN_VEXT_VMERGE_VV(vmerge_vvm_w, int32_t, H4, clearl)
2097 GEN_VEXT_VMERGE_VV(vmerge_vvm_d, int64_t, H8, clearq)
2098 
2099 #define GEN_VEXT_VMERGE_VX(NAME, ETYPE, H, CLEAR_FN)                 \
2100 void HELPER(NAME)(void *vd, void *v0, target_ulong s1,               \
2101                   void *vs2, CPURISCVState *env, uint32_t desc)      \
2102 {                                                                    \
2103     uint32_t mlen = vext_mlen(desc);                                 \
2104     uint32_t vl = env->vl;                                           \
2105     uint32_t esz = sizeof(ETYPE);                                    \
2106     uint32_t vlmax = vext_maxsz(desc) / esz;                         \
2107     uint32_t i;                                                      \
2108                                                                      \
2109     for (i = 0; i < vl; i++) {                                       \
2110         ETYPE s2 = *((ETYPE *)vs2 + H(i));                           \
2111         ETYPE d = (!vext_elem_mask(v0, mlen, i) ? s2 :               \
2112                    (ETYPE)(target_long)s1);                          \
2113         *((ETYPE *)vd + H(i)) = d;                                   \
2114     }                                                                \
2115     CLEAR_FN(vd, vl, vl * esz, vlmax * esz);                         \
2116 }
2117 
2118 GEN_VEXT_VMERGE_VX(vmerge_vxm_b, int8_t,  H1, clearb)
2119 GEN_VEXT_VMERGE_VX(vmerge_vxm_h, int16_t, H2, clearh)
2120 GEN_VEXT_VMERGE_VX(vmerge_vxm_w, int32_t, H4, clearl)
2121 GEN_VEXT_VMERGE_VX(vmerge_vxm_d, int64_t, H8, clearq)
2122 
2123 /*
2124  *** Vector Fixed-Point Arithmetic Instructions
2125  */
2126 
2127 /* Vector Single-Width Saturating Add and Subtract */
2128 
2129 /*
2130  * As fixed point instructions probably have round mode and saturation,
2131  * define common macros for fixed point here.
2132  */
2133 typedef void opivv2_rm_fn(void *vd, void *vs1, void *vs2, int i,
2134                           CPURISCVState *env, int vxrm);
2135 
2136 #define OPIVV2_RM(NAME, TD, T1, T2, TX1, TX2, HD, HS1, HS2, OP)     \
2137 static inline void                                                  \
2138 do_##NAME(void *vd, void *vs1, void *vs2, int i,                    \
2139           CPURISCVState *env, int vxrm)                             \
2140 {                                                                   \
2141     TX1 s1 = *((T1 *)vs1 + HS1(i));                                 \
2142     TX2 s2 = *((T2 *)vs2 + HS2(i));                                 \
2143     *((TD *)vd + HD(i)) = OP(env, vxrm, s2, s1);                    \
2144 }
2145 
2146 static inline void
2147 vext_vv_rm_1(void *vd, void *v0, void *vs1, void *vs2,
2148              CPURISCVState *env,
2149              uint32_t vl, uint32_t vm, uint32_t mlen, int vxrm,
2150              opivv2_rm_fn *fn)
2151 {
2152     for (uint32_t i = 0; i < vl; i++) {
2153         if (!vm && !vext_elem_mask(v0, mlen, i)) {
2154             continue;
2155         }
2156         fn(vd, vs1, vs2, i, env, vxrm);
2157     }
2158 }
2159 
2160 static inline void
2161 vext_vv_rm_2(void *vd, void *v0, void *vs1, void *vs2,
2162              CPURISCVState *env,
2163              uint32_t desc, uint32_t esz, uint32_t dsz,
2164              opivv2_rm_fn *fn, clear_fn *clearfn)
2165 {
2166     uint32_t vlmax = vext_maxsz(desc) / esz;
2167     uint32_t mlen = vext_mlen(desc);
2168     uint32_t vm = vext_vm(desc);
2169     uint32_t vl = env->vl;
2170 
2171     switch (env->vxrm) {
2172     case 0: /* rnu */
2173         vext_vv_rm_1(vd, v0, vs1, vs2,
2174                      env, vl, vm, mlen, 0, fn);
2175         break;
2176     case 1: /* rne */
2177         vext_vv_rm_1(vd, v0, vs1, vs2,
2178                      env, vl, vm, mlen, 1, fn);
2179         break;
2180     case 2: /* rdn */
2181         vext_vv_rm_1(vd, v0, vs1, vs2,
2182                      env, vl, vm, mlen, 2, fn);
2183         break;
2184     default: /* rod */
2185         vext_vv_rm_1(vd, v0, vs1, vs2,
2186                      env, vl, vm, mlen, 3, fn);
2187         break;
2188     }
2189 
2190     clearfn(vd, vl, vl * dsz,  vlmax * dsz);
2191 }
2192 
2193 /* generate helpers for fixed point instructions with OPIVV format */
2194 #define GEN_VEXT_VV_RM(NAME, ESZ, DSZ, CLEAR_FN)                \
2195 void HELPER(NAME)(void *vd, void *v0, void *vs1, void *vs2,     \
2196                   CPURISCVState *env, uint32_t desc)            \
2197 {                                                               \
2198     vext_vv_rm_2(vd, v0, vs1, vs2, env, desc, ESZ, DSZ,         \
2199                  do_##NAME, CLEAR_FN);                          \
2200 }
2201 
2202 static inline uint8_t saddu8(CPURISCVState *env, int vxrm, uint8_t a, uint8_t b)
2203 {
2204     uint8_t res = a + b;
2205     if (res < a) {
2206         res = UINT8_MAX;
2207         env->vxsat = 0x1;
2208     }
2209     return res;
2210 }
2211 
2212 static inline uint16_t saddu16(CPURISCVState *env, int vxrm, uint16_t a,
2213                                uint16_t b)
2214 {
2215     uint16_t res = a + b;
2216     if (res < a) {
2217         res = UINT16_MAX;
2218         env->vxsat = 0x1;
2219     }
2220     return res;
2221 }
2222 
2223 static inline uint32_t saddu32(CPURISCVState *env, int vxrm, uint32_t a,
2224                                uint32_t b)
2225 {
2226     uint32_t res = a + b;
2227     if (res < a) {
2228         res = UINT32_MAX;
2229         env->vxsat = 0x1;
2230     }
2231     return res;
2232 }
2233 
2234 static inline uint64_t saddu64(CPURISCVState *env, int vxrm, uint64_t a,
2235                                uint64_t b)
2236 {
2237     uint64_t res = a + b;
2238     if (res < a) {
2239         res = UINT64_MAX;
2240         env->vxsat = 0x1;
2241     }
2242     return res;
2243 }
2244 
2245 RVVCALL(OPIVV2_RM, vsaddu_vv_b, OP_UUU_B, H1, H1, H1, saddu8)
2246 RVVCALL(OPIVV2_RM, vsaddu_vv_h, OP_UUU_H, H2, H2, H2, saddu16)
2247 RVVCALL(OPIVV2_RM, vsaddu_vv_w, OP_UUU_W, H4, H4, H4, saddu32)
2248 RVVCALL(OPIVV2_RM, vsaddu_vv_d, OP_UUU_D, H8, H8, H8, saddu64)
2249 GEN_VEXT_VV_RM(vsaddu_vv_b, 1, 1, clearb)
2250 GEN_VEXT_VV_RM(vsaddu_vv_h, 2, 2, clearh)
2251 GEN_VEXT_VV_RM(vsaddu_vv_w, 4, 4, clearl)
2252 GEN_VEXT_VV_RM(vsaddu_vv_d, 8, 8, clearq)
2253 
2254 typedef void opivx2_rm_fn(void *vd, target_long s1, void *vs2, int i,
2255                           CPURISCVState *env, int vxrm);
2256 
2257 #define OPIVX2_RM(NAME, TD, T1, T2, TX1, TX2, HD, HS2, OP)          \
2258 static inline void                                                  \
2259 do_##NAME(void *vd, target_long s1, void *vs2, int i,               \
2260           CPURISCVState *env, int vxrm)                             \
2261 {                                                                   \
2262     TX2 s2 = *((T2 *)vs2 + HS2(i));                                 \
2263     *((TD *)vd + HD(i)) = OP(env, vxrm, s2, (TX1)(T1)s1);           \
2264 }
2265 
2266 static inline void
2267 vext_vx_rm_1(void *vd, void *v0, target_long s1, void *vs2,
2268              CPURISCVState *env,
2269              uint32_t vl, uint32_t vm, uint32_t mlen, int vxrm,
2270              opivx2_rm_fn *fn)
2271 {
2272     for (uint32_t i = 0; i < vl; i++) {
2273         if (!vm && !vext_elem_mask(v0, mlen, i)) {
2274             continue;
2275         }
2276         fn(vd, s1, vs2, i, env, vxrm);
2277     }
2278 }
2279 
2280 static inline void
2281 vext_vx_rm_2(void *vd, void *v0, target_long s1, void *vs2,
2282              CPURISCVState *env,
2283              uint32_t desc, uint32_t esz, uint32_t dsz,
2284              opivx2_rm_fn *fn, clear_fn *clearfn)
2285 {
2286     uint32_t vlmax = vext_maxsz(desc) / esz;
2287     uint32_t mlen = vext_mlen(desc);
2288     uint32_t vm = vext_vm(desc);
2289     uint32_t vl = env->vl;
2290 
2291     switch (env->vxrm) {
2292     case 0: /* rnu */
2293         vext_vx_rm_1(vd, v0, s1, vs2,
2294                      env, vl, vm, mlen, 0, fn);
2295         break;
2296     case 1: /* rne */
2297         vext_vx_rm_1(vd, v0, s1, vs2,
2298                      env, vl, vm, mlen, 1, fn);
2299         break;
2300     case 2: /* rdn */
2301         vext_vx_rm_1(vd, v0, s1, vs2,
2302                      env, vl, vm, mlen, 2, fn);
2303         break;
2304     default: /* rod */
2305         vext_vx_rm_1(vd, v0, s1, vs2,
2306                      env, vl, vm, mlen, 3, fn);
2307         break;
2308     }
2309 
2310     clearfn(vd, vl, vl * dsz,  vlmax * dsz);
2311 }
2312 
2313 /* generate helpers for fixed point instructions with OPIVX format */
2314 #define GEN_VEXT_VX_RM(NAME, ESZ, DSZ, CLEAR_FN)          \
2315 void HELPER(NAME)(void *vd, void *v0, target_ulong s1,    \
2316         void *vs2, CPURISCVState *env, uint32_t desc)     \
2317 {                                                         \
2318     vext_vx_rm_2(vd, v0, s1, vs2, env, desc, ESZ, DSZ,    \
2319                  do_##NAME, CLEAR_FN);                    \
2320 }
2321 
2322 RVVCALL(OPIVX2_RM, vsaddu_vx_b, OP_UUU_B, H1, H1, saddu8)
2323 RVVCALL(OPIVX2_RM, vsaddu_vx_h, OP_UUU_H, H2, H2, saddu16)
2324 RVVCALL(OPIVX2_RM, vsaddu_vx_w, OP_UUU_W, H4, H4, saddu32)
2325 RVVCALL(OPIVX2_RM, vsaddu_vx_d, OP_UUU_D, H8, H8, saddu64)
2326 GEN_VEXT_VX_RM(vsaddu_vx_b, 1, 1, clearb)
2327 GEN_VEXT_VX_RM(vsaddu_vx_h, 2, 2, clearh)
2328 GEN_VEXT_VX_RM(vsaddu_vx_w, 4, 4, clearl)
2329 GEN_VEXT_VX_RM(vsaddu_vx_d, 8, 8, clearq)
2330 
2331 static inline int8_t sadd8(CPURISCVState *env, int vxrm, int8_t a, int8_t b)
2332 {
2333     int8_t res = a + b;
2334     if ((res ^ a) & (res ^ b) & INT8_MIN) {
2335         res = a > 0 ? INT8_MAX : INT8_MIN;
2336         env->vxsat = 0x1;
2337     }
2338     return res;
2339 }
2340 
2341 static inline int16_t sadd16(CPURISCVState *env, int vxrm, int16_t a, int16_t b)
2342 {
2343     int16_t res = a + b;
2344     if ((res ^ a) & (res ^ b) & INT16_MIN) {
2345         res = a > 0 ? INT16_MAX : INT16_MIN;
2346         env->vxsat = 0x1;
2347     }
2348     return res;
2349 }
2350 
2351 static inline int32_t sadd32(CPURISCVState *env, int vxrm, int32_t a, int32_t b)
2352 {
2353     int32_t res = a + b;
2354     if ((res ^ a) & (res ^ b) & INT32_MIN) {
2355         res = a > 0 ? INT32_MAX : INT32_MIN;
2356         env->vxsat = 0x1;
2357     }
2358     return res;
2359 }
2360 
2361 static inline int64_t sadd64(CPURISCVState *env, int vxrm, int64_t a, int64_t b)
2362 {
2363     int64_t res = a + b;
2364     if ((res ^ a) & (res ^ b) & INT64_MIN) {
2365         res = a > 0 ? INT64_MAX : INT64_MIN;
2366         env->vxsat = 0x1;
2367     }
2368     return res;
2369 }
2370 
2371 RVVCALL(OPIVV2_RM, vsadd_vv_b, OP_SSS_B, H1, H1, H1, sadd8)
2372 RVVCALL(OPIVV2_RM, vsadd_vv_h, OP_SSS_H, H2, H2, H2, sadd16)
2373 RVVCALL(OPIVV2_RM, vsadd_vv_w, OP_SSS_W, H4, H4, H4, sadd32)
2374 RVVCALL(OPIVV2_RM, vsadd_vv_d, OP_SSS_D, H8, H8, H8, sadd64)
2375 GEN_VEXT_VV_RM(vsadd_vv_b, 1, 1, clearb)
2376 GEN_VEXT_VV_RM(vsadd_vv_h, 2, 2, clearh)
2377 GEN_VEXT_VV_RM(vsadd_vv_w, 4, 4, clearl)
2378 GEN_VEXT_VV_RM(vsadd_vv_d, 8, 8, clearq)
2379 
2380 RVVCALL(OPIVX2_RM, vsadd_vx_b, OP_SSS_B, H1, H1, sadd8)
2381 RVVCALL(OPIVX2_RM, vsadd_vx_h, OP_SSS_H, H2, H2, sadd16)
2382 RVVCALL(OPIVX2_RM, vsadd_vx_w, OP_SSS_W, H4, H4, sadd32)
2383 RVVCALL(OPIVX2_RM, vsadd_vx_d, OP_SSS_D, H8, H8, sadd64)
2384 GEN_VEXT_VX_RM(vsadd_vx_b, 1, 1, clearb)
2385 GEN_VEXT_VX_RM(vsadd_vx_h, 2, 2, clearh)
2386 GEN_VEXT_VX_RM(vsadd_vx_w, 4, 4, clearl)
2387 GEN_VEXT_VX_RM(vsadd_vx_d, 8, 8, clearq)
2388 
2389 static inline uint8_t ssubu8(CPURISCVState *env, int vxrm, uint8_t a, uint8_t b)
2390 {
2391     uint8_t res = a - b;
2392     if (res > a) {
2393         res = 0;
2394         env->vxsat = 0x1;
2395     }
2396     return res;
2397 }
2398 
2399 static inline uint16_t ssubu16(CPURISCVState *env, int vxrm, uint16_t a,
2400                                uint16_t b)
2401 {
2402     uint16_t res = a - b;
2403     if (res > a) {
2404         res = 0;
2405         env->vxsat = 0x1;
2406     }
2407     return res;
2408 }
2409 
2410 static inline uint32_t ssubu32(CPURISCVState *env, int vxrm, uint32_t a,
2411                                uint32_t b)
2412 {
2413     uint32_t res = a - b;
2414     if (res > a) {
2415         res = 0;
2416         env->vxsat = 0x1;
2417     }
2418     return res;
2419 }
2420 
2421 static inline uint64_t ssubu64(CPURISCVState *env, int vxrm, uint64_t a,
2422                                uint64_t b)
2423 {
2424     uint64_t res = a - b;
2425     if (res > a) {
2426         res = 0;
2427         env->vxsat = 0x1;
2428     }
2429     return res;
2430 }
2431 
2432 RVVCALL(OPIVV2_RM, vssubu_vv_b, OP_UUU_B, H1, H1, H1, ssubu8)
2433 RVVCALL(OPIVV2_RM, vssubu_vv_h, OP_UUU_H, H2, H2, H2, ssubu16)
2434 RVVCALL(OPIVV2_RM, vssubu_vv_w, OP_UUU_W, H4, H4, H4, ssubu32)
2435 RVVCALL(OPIVV2_RM, vssubu_vv_d, OP_UUU_D, H8, H8, H8, ssubu64)
2436 GEN_VEXT_VV_RM(vssubu_vv_b, 1, 1, clearb)
2437 GEN_VEXT_VV_RM(vssubu_vv_h, 2, 2, clearh)
2438 GEN_VEXT_VV_RM(vssubu_vv_w, 4, 4, clearl)
2439 GEN_VEXT_VV_RM(vssubu_vv_d, 8, 8, clearq)
2440 
2441 RVVCALL(OPIVX2_RM, vssubu_vx_b, OP_UUU_B, H1, H1, ssubu8)
2442 RVVCALL(OPIVX2_RM, vssubu_vx_h, OP_UUU_H, H2, H2, ssubu16)
2443 RVVCALL(OPIVX2_RM, vssubu_vx_w, OP_UUU_W, H4, H4, ssubu32)
2444 RVVCALL(OPIVX2_RM, vssubu_vx_d, OP_UUU_D, H8, H8, ssubu64)
2445 GEN_VEXT_VX_RM(vssubu_vx_b, 1, 1, clearb)
2446 GEN_VEXT_VX_RM(vssubu_vx_h, 2, 2, clearh)
2447 GEN_VEXT_VX_RM(vssubu_vx_w, 4, 4, clearl)
2448 GEN_VEXT_VX_RM(vssubu_vx_d, 8, 8, clearq)
2449 
2450 static inline int8_t ssub8(CPURISCVState *env, int vxrm, int8_t a, int8_t b)
2451 {
2452     int8_t res = a - b;
2453     if ((res ^ a) & (a ^ b) & INT8_MIN) {
2454         res = a > 0 ? INT8_MAX : INT8_MIN;
2455         env->vxsat = 0x1;
2456     }
2457     return res;
2458 }
2459 
2460 static inline int16_t ssub16(CPURISCVState *env, int vxrm, int16_t a, int16_t b)
2461 {
2462     int16_t res = a - b;
2463     if ((res ^ a) & (a ^ b) & INT16_MIN) {
2464         res = a > 0 ? INT16_MAX : INT16_MIN;
2465         env->vxsat = 0x1;
2466     }
2467     return res;
2468 }
2469 
2470 static inline int32_t ssub32(CPURISCVState *env, int vxrm, int32_t a, int32_t b)
2471 {
2472     int32_t res = a - b;
2473     if ((res ^ a) & (a ^ b) & INT32_MIN) {
2474         res = a > 0 ? INT32_MAX : INT32_MIN;
2475         env->vxsat = 0x1;
2476     }
2477     return res;
2478 }
2479 
2480 static inline int64_t ssub64(CPURISCVState *env, int vxrm, int64_t a, int64_t b)
2481 {
2482     int64_t res = a - b;
2483     if ((res ^ a) & (a ^ b) & INT64_MIN) {
2484         res = a > 0 ? INT64_MAX : INT64_MIN;
2485         env->vxsat = 0x1;
2486     }
2487     return res;
2488 }
2489 
2490 RVVCALL(OPIVV2_RM, vssub_vv_b, OP_SSS_B, H1, H1, H1, ssub8)
2491 RVVCALL(OPIVV2_RM, vssub_vv_h, OP_SSS_H, H2, H2, H2, ssub16)
2492 RVVCALL(OPIVV2_RM, vssub_vv_w, OP_SSS_W, H4, H4, H4, ssub32)
2493 RVVCALL(OPIVV2_RM, vssub_vv_d, OP_SSS_D, H8, H8, H8, ssub64)
2494 GEN_VEXT_VV_RM(vssub_vv_b, 1, 1, clearb)
2495 GEN_VEXT_VV_RM(vssub_vv_h, 2, 2, clearh)
2496 GEN_VEXT_VV_RM(vssub_vv_w, 4, 4, clearl)
2497 GEN_VEXT_VV_RM(vssub_vv_d, 8, 8, clearq)
2498 
2499 RVVCALL(OPIVX2_RM, vssub_vx_b, OP_SSS_B, H1, H1, ssub8)
2500 RVVCALL(OPIVX2_RM, vssub_vx_h, OP_SSS_H, H2, H2, ssub16)
2501 RVVCALL(OPIVX2_RM, vssub_vx_w, OP_SSS_W, H4, H4, ssub32)
2502 RVVCALL(OPIVX2_RM, vssub_vx_d, OP_SSS_D, H8, H8, ssub64)
2503 GEN_VEXT_VX_RM(vssub_vx_b, 1, 1, clearb)
2504 GEN_VEXT_VX_RM(vssub_vx_h, 2, 2, clearh)
2505 GEN_VEXT_VX_RM(vssub_vx_w, 4, 4, clearl)
2506 GEN_VEXT_VX_RM(vssub_vx_d, 8, 8, clearq)
2507 
2508 /* Vector Single-Width Averaging Add and Subtract */
2509 static inline uint8_t get_round(int vxrm, uint64_t v, uint8_t shift)
2510 {
2511     uint8_t d = extract64(v, shift, 1);
2512     uint8_t d1;
2513     uint64_t D1, D2;
2514 
2515     if (shift == 0 || shift > 64) {
2516         return 0;
2517     }
2518 
2519     d1 = extract64(v, shift - 1, 1);
2520     D1 = extract64(v, 0, shift);
2521     if (vxrm == 0) { /* round-to-nearest-up (add +0.5 LSB) */
2522         return d1;
2523     } else if (vxrm == 1) { /* round-to-nearest-even */
2524         if (shift > 1) {
2525             D2 = extract64(v, 0, shift - 1);
2526             return d1 & ((D2 != 0) | d);
2527         } else {
2528             return d1 & d;
2529         }
2530     } else if (vxrm == 3) { /* round-to-odd (OR bits into LSB, aka "jam") */
2531         return !d & (D1 != 0);
2532     }
2533     return 0; /* round-down (truncate) */
2534 }
2535 
2536 static inline int32_t aadd32(CPURISCVState *env, int vxrm, int32_t a, int32_t b)
2537 {
2538     int64_t res = (int64_t)a + b;
2539     uint8_t round = get_round(vxrm, res, 1);
2540 
2541     return (res >> 1) + round;
2542 }
2543 
2544 static inline int64_t aadd64(CPURISCVState *env, int vxrm, int64_t a, int64_t b)
2545 {
2546     int64_t res = a + b;
2547     uint8_t round = get_round(vxrm, res, 1);
2548     int64_t over = (res ^ a) & (res ^ b) & INT64_MIN;
2549 
2550     /* With signed overflow, bit 64 is inverse of bit 63. */
2551     return ((res >> 1) ^ over) + round;
2552 }
2553 
2554 RVVCALL(OPIVV2_RM, vaadd_vv_b, OP_SSS_B, H1, H1, H1, aadd32)
2555 RVVCALL(OPIVV2_RM, vaadd_vv_h, OP_SSS_H, H2, H2, H2, aadd32)
2556 RVVCALL(OPIVV2_RM, vaadd_vv_w, OP_SSS_W, H4, H4, H4, aadd32)
2557 RVVCALL(OPIVV2_RM, vaadd_vv_d, OP_SSS_D, H8, H8, H8, aadd64)
2558 GEN_VEXT_VV_RM(vaadd_vv_b, 1, 1, clearb)
2559 GEN_VEXT_VV_RM(vaadd_vv_h, 2, 2, clearh)
2560 GEN_VEXT_VV_RM(vaadd_vv_w, 4, 4, clearl)
2561 GEN_VEXT_VV_RM(vaadd_vv_d, 8, 8, clearq)
2562 
2563 RVVCALL(OPIVX2_RM, vaadd_vx_b, OP_SSS_B, H1, H1, aadd32)
2564 RVVCALL(OPIVX2_RM, vaadd_vx_h, OP_SSS_H, H2, H2, aadd32)
2565 RVVCALL(OPIVX2_RM, vaadd_vx_w, OP_SSS_W, H4, H4, aadd32)
2566 RVVCALL(OPIVX2_RM, vaadd_vx_d, OP_SSS_D, H8, H8, aadd64)
2567 GEN_VEXT_VX_RM(vaadd_vx_b, 1, 1, clearb)
2568 GEN_VEXT_VX_RM(vaadd_vx_h, 2, 2, clearh)
2569 GEN_VEXT_VX_RM(vaadd_vx_w, 4, 4, clearl)
2570 GEN_VEXT_VX_RM(vaadd_vx_d, 8, 8, clearq)
2571 
2572 static inline int32_t asub32(CPURISCVState *env, int vxrm, int32_t a, int32_t b)
2573 {
2574     int64_t res = (int64_t)a - b;
2575     uint8_t round = get_round(vxrm, res, 1);
2576 
2577     return (res >> 1) + round;
2578 }
2579 
2580 static inline int64_t asub64(CPURISCVState *env, int vxrm, int64_t a, int64_t b)
2581 {
2582     int64_t res = (int64_t)a - b;
2583     uint8_t round = get_round(vxrm, res, 1);
2584     int64_t over = (res ^ a) & (a ^ b) & INT64_MIN;
2585 
2586     /* With signed overflow, bit 64 is inverse of bit 63. */
2587     return ((res >> 1) ^ over) + round;
2588 }
2589 
2590 RVVCALL(OPIVV2_RM, vasub_vv_b, OP_SSS_B, H1, H1, H1, asub32)
2591 RVVCALL(OPIVV2_RM, vasub_vv_h, OP_SSS_H, H2, H2, H2, asub32)
2592 RVVCALL(OPIVV2_RM, vasub_vv_w, OP_SSS_W, H4, H4, H4, asub32)
2593 RVVCALL(OPIVV2_RM, vasub_vv_d, OP_SSS_D, H8, H8, H8, asub64)
2594 GEN_VEXT_VV_RM(vasub_vv_b, 1, 1, clearb)
2595 GEN_VEXT_VV_RM(vasub_vv_h, 2, 2, clearh)
2596 GEN_VEXT_VV_RM(vasub_vv_w, 4, 4, clearl)
2597 GEN_VEXT_VV_RM(vasub_vv_d, 8, 8, clearq)
2598 
2599 RVVCALL(OPIVX2_RM, vasub_vx_b, OP_SSS_B, H1, H1, asub32)
2600 RVVCALL(OPIVX2_RM, vasub_vx_h, OP_SSS_H, H2, H2, asub32)
2601 RVVCALL(OPIVX2_RM, vasub_vx_w, OP_SSS_W, H4, H4, asub32)
2602 RVVCALL(OPIVX2_RM, vasub_vx_d, OP_SSS_D, H8, H8, asub64)
2603 GEN_VEXT_VX_RM(vasub_vx_b, 1, 1, clearb)
2604 GEN_VEXT_VX_RM(vasub_vx_h, 2, 2, clearh)
2605 GEN_VEXT_VX_RM(vasub_vx_w, 4, 4, clearl)
2606 GEN_VEXT_VX_RM(vasub_vx_d, 8, 8, clearq)
2607 
2608 /* Vector Single-Width Fractional Multiply with Rounding and Saturation */
2609 static inline int8_t vsmul8(CPURISCVState *env, int vxrm, int8_t a, int8_t b)
2610 {
2611     uint8_t round;
2612     int16_t res;
2613 
2614     res = (int16_t)a * (int16_t)b;
2615     round = get_round(vxrm, res, 7);
2616     res   = (res >> 7) + round;
2617 
2618     if (res > INT8_MAX) {
2619         env->vxsat = 0x1;
2620         return INT8_MAX;
2621     } else if (res < INT8_MIN) {
2622         env->vxsat = 0x1;
2623         return INT8_MIN;
2624     } else {
2625         return res;
2626     }
2627 }
2628 
2629 static int16_t vsmul16(CPURISCVState *env, int vxrm, int16_t a, int16_t b)
2630 {
2631     uint8_t round;
2632     int32_t res;
2633 
2634     res = (int32_t)a * (int32_t)b;
2635     round = get_round(vxrm, res, 15);
2636     res   = (res >> 15) + round;
2637 
2638     if (res > INT16_MAX) {
2639         env->vxsat = 0x1;
2640         return INT16_MAX;
2641     } else if (res < INT16_MIN) {
2642         env->vxsat = 0x1;
2643         return INT16_MIN;
2644     } else {
2645         return res;
2646     }
2647 }
2648 
2649 static int32_t vsmul32(CPURISCVState *env, int vxrm, int32_t a, int32_t b)
2650 {
2651     uint8_t round;
2652     int64_t res;
2653 
2654     res = (int64_t)a * (int64_t)b;
2655     round = get_round(vxrm, res, 31);
2656     res   = (res >> 31) + round;
2657 
2658     if (res > INT32_MAX) {
2659         env->vxsat = 0x1;
2660         return INT32_MAX;
2661     } else if (res < INT32_MIN) {
2662         env->vxsat = 0x1;
2663         return INT32_MIN;
2664     } else {
2665         return res;
2666     }
2667 }
2668 
2669 static int64_t vsmul64(CPURISCVState *env, int vxrm, int64_t a, int64_t b)
2670 {
2671     uint8_t round;
2672     uint64_t hi_64, lo_64;
2673     int64_t res;
2674 
2675     if (a == INT64_MIN && b == INT64_MIN) {
2676         env->vxsat = 1;
2677         return INT64_MAX;
2678     }
2679 
2680     muls64(&lo_64, &hi_64, a, b);
2681     round = get_round(vxrm, lo_64, 63);
2682     /*
2683      * Cannot overflow, as there are always
2684      * 2 sign bits after multiply.
2685      */
2686     res = (hi_64 << 1) | (lo_64 >> 63);
2687     if (round) {
2688         if (res == INT64_MAX) {
2689             env->vxsat = 1;
2690         } else {
2691             res += 1;
2692         }
2693     }
2694     return res;
2695 }
2696 
2697 RVVCALL(OPIVV2_RM, vsmul_vv_b, OP_SSS_B, H1, H1, H1, vsmul8)
2698 RVVCALL(OPIVV2_RM, vsmul_vv_h, OP_SSS_H, H2, H2, H2, vsmul16)
2699 RVVCALL(OPIVV2_RM, vsmul_vv_w, OP_SSS_W, H4, H4, H4, vsmul32)
2700 RVVCALL(OPIVV2_RM, vsmul_vv_d, OP_SSS_D, H8, H8, H8, vsmul64)
2701 GEN_VEXT_VV_RM(vsmul_vv_b, 1, 1, clearb)
2702 GEN_VEXT_VV_RM(vsmul_vv_h, 2, 2, clearh)
2703 GEN_VEXT_VV_RM(vsmul_vv_w, 4, 4, clearl)
2704 GEN_VEXT_VV_RM(vsmul_vv_d, 8, 8, clearq)
2705 
2706 RVVCALL(OPIVX2_RM, vsmul_vx_b, OP_SSS_B, H1, H1, vsmul8)
2707 RVVCALL(OPIVX2_RM, vsmul_vx_h, OP_SSS_H, H2, H2, vsmul16)
2708 RVVCALL(OPIVX2_RM, vsmul_vx_w, OP_SSS_W, H4, H4, vsmul32)
2709 RVVCALL(OPIVX2_RM, vsmul_vx_d, OP_SSS_D, H8, H8, vsmul64)
2710 GEN_VEXT_VX_RM(vsmul_vx_b, 1, 1, clearb)
2711 GEN_VEXT_VX_RM(vsmul_vx_h, 2, 2, clearh)
2712 GEN_VEXT_VX_RM(vsmul_vx_w, 4, 4, clearl)
2713 GEN_VEXT_VX_RM(vsmul_vx_d, 8, 8, clearq)
2714 
2715 /* Vector Widening Saturating Scaled Multiply-Add */
2716 static inline uint16_t
2717 vwsmaccu8(CPURISCVState *env, int vxrm, uint8_t a, uint8_t b,
2718           uint16_t c)
2719 {
2720     uint8_t round;
2721     uint16_t res = (uint16_t)a * b;
2722 
2723     round = get_round(vxrm, res, 4);
2724     res   = (res >> 4) + round;
2725     return saddu16(env, vxrm, c, res);
2726 }
2727 
2728 static inline uint32_t
2729 vwsmaccu16(CPURISCVState *env, int vxrm, uint16_t a, uint16_t b,
2730            uint32_t c)
2731 {
2732     uint8_t round;
2733     uint32_t res = (uint32_t)a * b;
2734 
2735     round = get_round(vxrm, res, 8);
2736     res   = (res >> 8) + round;
2737     return saddu32(env, vxrm, c, res);
2738 }
2739 
2740 static inline uint64_t
2741 vwsmaccu32(CPURISCVState *env, int vxrm, uint32_t a, uint32_t b,
2742            uint64_t c)
2743 {
2744     uint8_t round;
2745     uint64_t res = (uint64_t)a * b;
2746 
2747     round = get_round(vxrm, res, 16);
2748     res   = (res >> 16) + round;
2749     return saddu64(env, vxrm, c, res);
2750 }
2751 
2752 #define OPIVV3_RM(NAME, TD, T1, T2, TX1, TX2, HD, HS1, HS2, OP)    \
2753 static inline void                                                 \
2754 do_##NAME(void *vd, void *vs1, void *vs2, int i,                   \
2755           CPURISCVState *env, int vxrm)                            \
2756 {                                                                  \
2757     TX1 s1 = *((T1 *)vs1 + HS1(i));                                \
2758     TX2 s2 = *((T2 *)vs2 + HS2(i));                                \
2759     TD d = *((TD *)vd + HD(i));                                    \
2760     *((TD *)vd + HD(i)) = OP(env, vxrm, s2, s1, d);                \
2761 }
2762 
2763 RVVCALL(OPIVV3_RM, vwsmaccu_vv_b, WOP_UUU_B, H2, H1, H1, vwsmaccu8)
2764 RVVCALL(OPIVV3_RM, vwsmaccu_vv_h, WOP_UUU_H, H4, H2, H2, vwsmaccu16)
2765 RVVCALL(OPIVV3_RM, vwsmaccu_vv_w, WOP_UUU_W, H8, H4, H4, vwsmaccu32)
2766 GEN_VEXT_VV_RM(vwsmaccu_vv_b, 1, 2, clearh)
2767 GEN_VEXT_VV_RM(vwsmaccu_vv_h, 2, 4, clearl)
2768 GEN_VEXT_VV_RM(vwsmaccu_vv_w, 4, 8, clearq)
2769 
2770 #define OPIVX3_RM(NAME, TD, T1, T2, TX1, TX2, HD, HS2, OP)         \
2771 static inline void                                                 \
2772 do_##NAME(void *vd, target_long s1, void *vs2, int i,              \
2773           CPURISCVState *env, int vxrm)                            \
2774 {                                                                  \
2775     TX2 s2 = *((T2 *)vs2 + HS2(i));                                \
2776     TD d = *((TD *)vd + HD(i));                                    \
2777     *((TD *)vd + HD(i)) = OP(env, vxrm, s2, (TX1)(T1)s1, d);       \
2778 }
2779 
2780 RVVCALL(OPIVX3_RM, vwsmaccu_vx_b, WOP_UUU_B, H2, H1, vwsmaccu8)
2781 RVVCALL(OPIVX3_RM, vwsmaccu_vx_h, WOP_UUU_H, H4, H2, vwsmaccu16)
2782 RVVCALL(OPIVX3_RM, vwsmaccu_vx_w, WOP_UUU_W, H8, H4, vwsmaccu32)
2783 GEN_VEXT_VX_RM(vwsmaccu_vx_b, 1, 2, clearh)
2784 GEN_VEXT_VX_RM(vwsmaccu_vx_h, 2, 4, clearl)
2785 GEN_VEXT_VX_RM(vwsmaccu_vx_w, 4, 8, clearq)
2786 
2787 static inline int16_t
2788 vwsmacc8(CPURISCVState *env, int vxrm, int8_t a, int8_t b, int16_t c)
2789 {
2790     uint8_t round;
2791     int16_t res = (int16_t)a * b;
2792 
2793     round = get_round(vxrm, res, 4);
2794     res   = (res >> 4) + round;
2795     return sadd16(env, vxrm, c, res);
2796 }
2797 
2798 static inline int32_t
2799 vwsmacc16(CPURISCVState *env, int vxrm, int16_t a, int16_t b, int32_t c)
2800 {
2801     uint8_t round;
2802     int32_t res = (int32_t)a * b;
2803 
2804     round = get_round(vxrm, res, 8);
2805     res   = (res >> 8) + round;
2806     return sadd32(env, vxrm, c, res);
2807 
2808 }
2809 
2810 static inline int64_t
2811 vwsmacc32(CPURISCVState *env, int vxrm, int32_t a, int32_t b, int64_t c)
2812 {
2813     uint8_t round;
2814     int64_t res = (int64_t)a * b;
2815 
2816     round = get_round(vxrm, res, 16);
2817     res   = (res >> 16) + round;
2818     return sadd64(env, vxrm, c, res);
2819 }
2820 
2821 RVVCALL(OPIVV3_RM, vwsmacc_vv_b, WOP_SSS_B, H2, H1, H1, vwsmacc8)
2822 RVVCALL(OPIVV3_RM, vwsmacc_vv_h, WOP_SSS_H, H4, H2, H2, vwsmacc16)
2823 RVVCALL(OPIVV3_RM, vwsmacc_vv_w, WOP_SSS_W, H8, H4, H4, vwsmacc32)
2824 GEN_VEXT_VV_RM(vwsmacc_vv_b, 1, 2, clearh)
2825 GEN_VEXT_VV_RM(vwsmacc_vv_h, 2, 4, clearl)
2826 GEN_VEXT_VV_RM(vwsmacc_vv_w, 4, 8, clearq)
2827 RVVCALL(OPIVX3_RM, vwsmacc_vx_b, WOP_SSS_B, H2, H1, vwsmacc8)
2828 RVVCALL(OPIVX3_RM, vwsmacc_vx_h, WOP_SSS_H, H4, H2, vwsmacc16)
2829 RVVCALL(OPIVX3_RM, vwsmacc_vx_w, WOP_SSS_W, H8, H4, vwsmacc32)
2830 GEN_VEXT_VX_RM(vwsmacc_vx_b, 1, 2, clearh)
2831 GEN_VEXT_VX_RM(vwsmacc_vx_h, 2, 4, clearl)
2832 GEN_VEXT_VX_RM(vwsmacc_vx_w, 4, 8, clearq)
2833 
2834 static inline int16_t
2835 vwsmaccsu8(CPURISCVState *env, int vxrm, uint8_t a, int8_t b, int16_t c)
2836 {
2837     uint8_t round;
2838     int16_t res = a * (int16_t)b;
2839 
2840     round = get_round(vxrm, res, 4);
2841     res   = (res >> 4) + round;
2842     return ssub16(env, vxrm, c, res);
2843 }
2844 
2845 static inline int32_t
2846 vwsmaccsu16(CPURISCVState *env, int vxrm, uint16_t a, int16_t b, uint32_t c)
2847 {
2848     uint8_t round;
2849     int32_t res = a * (int32_t)b;
2850 
2851     round = get_round(vxrm, res, 8);
2852     res   = (res >> 8) + round;
2853     return ssub32(env, vxrm, c, res);
2854 }
2855 
2856 static inline int64_t
2857 vwsmaccsu32(CPURISCVState *env, int vxrm, uint32_t a, int32_t b, int64_t c)
2858 {
2859     uint8_t round;
2860     int64_t res = a * (int64_t)b;
2861 
2862     round = get_round(vxrm, res, 16);
2863     res   = (res >> 16) + round;
2864     return ssub64(env, vxrm, c, res);
2865 }
2866 
2867 RVVCALL(OPIVV3_RM, vwsmaccsu_vv_b, WOP_SSU_B, H2, H1, H1, vwsmaccsu8)
2868 RVVCALL(OPIVV3_RM, vwsmaccsu_vv_h, WOP_SSU_H, H4, H2, H2, vwsmaccsu16)
2869 RVVCALL(OPIVV3_RM, vwsmaccsu_vv_w, WOP_SSU_W, H8, H4, H4, vwsmaccsu32)
2870 GEN_VEXT_VV_RM(vwsmaccsu_vv_b, 1, 2, clearh)
2871 GEN_VEXT_VV_RM(vwsmaccsu_vv_h, 2, 4, clearl)
2872 GEN_VEXT_VV_RM(vwsmaccsu_vv_w, 4, 8, clearq)
2873 RVVCALL(OPIVX3_RM, vwsmaccsu_vx_b, WOP_SSU_B, H2, H1, vwsmaccsu8)
2874 RVVCALL(OPIVX3_RM, vwsmaccsu_vx_h, WOP_SSU_H, H4, H2, vwsmaccsu16)
2875 RVVCALL(OPIVX3_RM, vwsmaccsu_vx_w, WOP_SSU_W, H8, H4, vwsmaccsu32)
2876 GEN_VEXT_VX_RM(vwsmaccsu_vx_b, 1, 2, clearh)
2877 GEN_VEXT_VX_RM(vwsmaccsu_vx_h, 2, 4, clearl)
2878 GEN_VEXT_VX_RM(vwsmaccsu_vx_w, 4, 8, clearq)
2879 
2880 static inline int16_t
2881 vwsmaccus8(CPURISCVState *env, int vxrm, int8_t a, uint8_t b, int16_t c)
2882 {
2883     uint8_t round;
2884     int16_t res = (int16_t)a * b;
2885 
2886     round = get_round(vxrm, res, 4);
2887     res   = (res >> 4) + round;
2888     return ssub16(env, vxrm, c, res);
2889 }
2890 
2891 static inline int32_t
2892 vwsmaccus16(CPURISCVState *env, int vxrm, int16_t a, uint16_t b, int32_t c)
2893 {
2894     uint8_t round;
2895     int32_t res = (int32_t)a * b;
2896 
2897     round = get_round(vxrm, res, 8);
2898     res   = (res >> 8) + round;
2899     return ssub32(env, vxrm, c, res);
2900 }
2901 
2902 static inline int64_t
2903 vwsmaccus32(CPURISCVState *env, int vxrm, int32_t a, uint32_t b, int64_t c)
2904 {
2905     uint8_t round;
2906     int64_t res = (int64_t)a * b;
2907 
2908     round = get_round(vxrm, res, 16);
2909     res   = (res >> 16) + round;
2910     return ssub64(env, vxrm, c, res);
2911 }
2912 
2913 RVVCALL(OPIVX3_RM, vwsmaccus_vx_b, WOP_SUS_B, H2, H1, vwsmaccus8)
2914 RVVCALL(OPIVX3_RM, vwsmaccus_vx_h, WOP_SUS_H, H4, H2, vwsmaccus16)
2915 RVVCALL(OPIVX3_RM, vwsmaccus_vx_w, WOP_SUS_W, H8, H4, vwsmaccus32)
2916 GEN_VEXT_VX_RM(vwsmaccus_vx_b, 1, 2, clearh)
2917 GEN_VEXT_VX_RM(vwsmaccus_vx_h, 2, 4, clearl)
2918 GEN_VEXT_VX_RM(vwsmaccus_vx_w, 4, 8, clearq)
2919 
2920 /* Vector Single-Width Scaling Shift Instructions */
2921 static inline uint8_t
2922 vssrl8(CPURISCVState *env, int vxrm, uint8_t a, uint8_t b)
2923 {
2924     uint8_t round, shift = b & 0x7;
2925     uint8_t res;
2926 
2927     round = get_round(vxrm, a, shift);
2928     res   = (a >> shift)  + round;
2929     return res;
2930 }
2931 static inline uint16_t
2932 vssrl16(CPURISCVState *env, int vxrm, uint16_t a, uint16_t b)
2933 {
2934     uint8_t round, shift = b & 0xf;
2935     uint16_t res;
2936 
2937     round = get_round(vxrm, a, shift);
2938     res   = (a >> shift)  + round;
2939     return res;
2940 }
2941 static inline uint32_t
2942 vssrl32(CPURISCVState *env, int vxrm, uint32_t a, uint32_t b)
2943 {
2944     uint8_t round, shift = b & 0x1f;
2945     uint32_t res;
2946 
2947     round = get_round(vxrm, a, shift);
2948     res   = (a >> shift)  + round;
2949     return res;
2950 }
2951 static inline uint64_t
2952 vssrl64(CPURISCVState *env, int vxrm, uint64_t a, uint64_t b)
2953 {
2954     uint8_t round, shift = b & 0x3f;
2955     uint64_t res;
2956 
2957     round = get_round(vxrm, a, shift);
2958     res   = (a >> shift)  + round;
2959     return res;
2960 }
2961 RVVCALL(OPIVV2_RM, vssrl_vv_b, OP_UUU_B, H1, H1, H1, vssrl8)
2962 RVVCALL(OPIVV2_RM, vssrl_vv_h, OP_UUU_H, H2, H2, H2, vssrl16)
2963 RVVCALL(OPIVV2_RM, vssrl_vv_w, OP_UUU_W, H4, H4, H4, vssrl32)
2964 RVVCALL(OPIVV2_RM, vssrl_vv_d, OP_UUU_D, H8, H8, H8, vssrl64)
2965 GEN_VEXT_VV_RM(vssrl_vv_b, 1, 1, clearb)
2966 GEN_VEXT_VV_RM(vssrl_vv_h, 2, 2, clearh)
2967 GEN_VEXT_VV_RM(vssrl_vv_w, 4, 4, clearl)
2968 GEN_VEXT_VV_RM(vssrl_vv_d, 8, 8, clearq)
2969 
2970 RVVCALL(OPIVX2_RM, vssrl_vx_b, OP_UUU_B, H1, H1, vssrl8)
2971 RVVCALL(OPIVX2_RM, vssrl_vx_h, OP_UUU_H, H2, H2, vssrl16)
2972 RVVCALL(OPIVX2_RM, vssrl_vx_w, OP_UUU_W, H4, H4, vssrl32)
2973 RVVCALL(OPIVX2_RM, vssrl_vx_d, OP_UUU_D, H8, H8, vssrl64)
2974 GEN_VEXT_VX_RM(vssrl_vx_b, 1, 1, clearb)
2975 GEN_VEXT_VX_RM(vssrl_vx_h, 2, 2, clearh)
2976 GEN_VEXT_VX_RM(vssrl_vx_w, 4, 4, clearl)
2977 GEN_VEXT_VX_RM(vssrl_vx_d, 8, 8, clearq)
2978 
2979 static inline int8_t
2980 vssra8(CPURISCVState *env, int vxrm, int8_t a, int8_t b)
2981 {
2982     uint8_t round, shift = b & 0x7;
2983     int8_t res;
2984 
2985     round = get_round(vxrm, a, shift);
2986     res   = (a >> shift)  + round;
2987     return res;
2988 }
2989 static inline int16_t
2990 vssra16(CPURISCVState *env, int vxrm, int16_t a, int16_t b)
2991 {
2992     uint8_t round, shift = b & 0xf;
2993     int16_t res;
2994 
2995     round = get_round(vxrm, a, shift);
2996     res   = (a >> shift)  + round;
2997     return res;
2998 }
2999 static inline int32_t
3000 vssra32(CPURISCVState *env, int vxrm, int32_t a, int32_t b)
3001 {
3002     uint8_t round, shift = b & 0x1f;
3003     int32_t res;
3004 
3005     round = get_round(vxrm, a, shift);
3006     res   = (a >> shift)  + round;
3007     return res;
3008 }
3009 static inline int64_t
3010 vssra64(CPURISCVState *env, int vxrm, int64_t a, int64_t b)
3011 {
3012     uint8_t round, shift = b & 0x3f;
3013     int64_t res;
3014 
3015     round = get_round(vxrm, a, shift);
3016     res   = (a >> shift)  + round;
3017     return res;
3018 }
3019 
3020 RVVCALL(OPIVV2_RM, vssra_vv_b, OP_SSS_B, H1, H1, H1, vssra8)
3021 RVVCALL(OPIVV2_RM, vssra_vv_h, OP_SSS_H, H2, H2, H2, vssra16)
3022 RVVCALL(OPIVV2_RM, vssra_vv_w, OP_SSS_W, H4, H4, H4, vssra32)
3023 RVVCALL(OPIVV2_RM, vssra_vv_d, OP_SSS_D, H8, H8, H8, vssra64)
3024 GEN_VEXT_VV_RM(vssra_vv_b, 1, 1, clearb)
3025 GEN_VEXT_VV_RM(vssra_vv_h, 2, 2, clearh)
3026 GEN_VEXT_VV_RM(vssra_vv_w, 4, 4, clearl)
3027 GEN_VEXT_VV_RM(vssra_vv_d, 8, 8, clearq)
3028 
3029 RVVCALL(OPIVX2_RM, vssra_vx_b, OP_SSS_B, H1, H1, vssra8)
3030 RVVCALL(OPIVX2_RM, vssra_vx_h, OP_SSS_H, H2, H2, vssra16)
3031 RVVCALL(OPIVX2_RM, vssra_vx_w, OP_SSS_W, H4, H4, vssra32)
3032 RVVCALL(OPIVX2_RM, vssra_vx_d, OP_SSS_D, H8, H8, vssra64)
3033 GEN_VEXT_VX_RM(vssra_vx_b, 1, 1, clearb)
3034 GEN_VEXT_VX_RM(vssra_vx_h, 2, 2, clearh)
3035 GEN_VEXT_VX_RM(vssra_vx_w, 4, 4, clearl)
3036 GEN_VEXT_VX_RM(vssra_vx_d, 8, 8, clearq)
3037 
3038 /* Vector Narrowing Fixed-Point Clip Instructions */
3039 static inline int8_t
3040 vnclip8(CPURISCVState *env, int vxrm, int16_t a, int8_t b)
3041 {
3042     uint8_t round, shift = b & 0xf;
3043     int16_t res;
3044 
3045     round = get_round(vxrm, a, shift);
3046     res   = (a >> shift)  + round;
3047     if (res > INT8_MAX) {
3048         env->vxsat = 0x1;
3049         return INT8_MAX;
3050     } else if (res < INT8_MIN) {
3051         env->vxsat = 0x1;
3052         return INT8_MIN;
3053     } else {
3054         return res;
3055     }
3056 }
3057 
3058 static inline int16_t
3059 vnclip16(CPURISCVState *env, int vxrm, int32_t a, int16_t b)
3060 {
3061     uint8_t round, shift = b & 0x1f;
3062     int32_t res;
3063 
3064     round = get_round(vxrm, a, shift);
3065     res   = (a >> shift)  + round;
3066     if (res > INT16_MAX) {
3067         env->vxsat = 0x1;
3068         return INT16_MAX;
3069     } else if (res < INT16_MIN) {
3070         env->vxsat = 0x1;
3071         return INT16_MIN;
3072     } else {
3073         return res;
3074     }
3075 }
3076 
3077 static inline int32_t
3078 vnclip32(CPURISCVState *env, int vxrm, int64_t a, int32_t b)
3079 {
3080     uint8_t round, shift = b & 0x3f;
3081     int64_t res;
3082 
3083     round = get_round(vxrm, a, shift);
3084     res   = (a >> shift)  + round;
3085     if (res > INT32_MAX) {
3086         env->vxsat = 0x1;
3087         return INT32_MAX;
3088     } else if (res < INT32_MIN) {
3089         env->vxsat = 0x1;
3090         return INT32_MIN;
3091     } else {
3092         return res;
3093     }
3094 }
3095 
3096 RVVCALL(OPIVV2_RM, vnclip_vv_b, NOP_SSS_B, H1, H2, H1, vnclip8)
3097 RVVCALL(OPIVV2_RM, vnclip_vv_h, NOP_SSS_H, H2, H4, H2, vnclip16)
3098 RVVCALL(OPIVV2_RM, vnclip_vv_w, NOP_SSS_W, H4, H8, H4, vnclip32)
3099 GEN_VEXT_VV_RM(vnclip_vv_b, 1, 1, clearb)
3100 GEN_VEXT_VV_RM(vnclip_vv_h, 2, 2, clearh)
3101 GEN_VEXT_VV_RM(vnclip_vv_w, 4, 4, clearl)
3102 
3103 RVVCALL(OPIVX2_RM, vnclip_vx_b, NOP_SSS_B, H1, H2, vnclip8)
3104 RVVCALL(OPIVX2_RM, vnclip_vx_h, NOP_SSS_H, H2, H4, vnclip16)
3105 RVVCALL(OPIVX2_RM, vnclip_vx_w, NOP_SSS_W, H4, H8, vnclip32)
3106 GEN_VEXT_VX_RM(vnclip_vx_b, 1, 1, clearb)
3107 GEN_VEXT_VX_RM(vnclip_vx_h, 2, 2, clearh)
3108 GEN_VEXT_VX_RM(vnclip_vx_w, 4, 4, clearl)
3109 
3110 static inline uint8_t
3111 vnclipu8(CPURISCVState *env, int vxrm, uint16_t a, uint8_t b)
3112 {
3113     uint8_t round, shift = b & 0xf;
3114     uint16_t res;
3115 
3116     round = get_round(vxrm, a, shift);
3117     res   = (a >> shift)  + round;
3118     if (res > UINT8_MAX) {
3119         env->vxsat = 0x1;
3120         return UINT8_MAX;
3121     } else {
3122         return res;
3123     }
3124 }
3125 
3126 static inline uint16_t
3127 vnclipu16(CPURISCVState *env, int vxrm, uint32_t a, uint16_t b)
3128 {
3129     uint8_t round, shift = b & 0x1f;
3130     uint32_t res;
3131 
3132     round = get_round(vxrm, a, shift);
3133     res   = (a >> shift)  + round;
3134     if (res > UINT16_MAX) {
3135         env->vxsat = 0x1;
3136         return UINT16_MAX;
3137     } else {
3138         return res;
3139     }
3140 }
3141 
3142 static inline uint32_t
3143 vnclipu32(CPURISCVState *env, int vxrm, uint64_t a, uint32_t b)
3144 {
3145     uint8_t round, shift = b & 0x3f;
3146     int64_t res;
3147 
3148     round = get_round(vxrm, a, shift);
3149     res   = (a >> shift)  + round;
3150     if (res > UINT32_MAX) {
3151         env->vxsat = 0x1;
3152         return UINT32_MAX;
3153     } else {
3154         return res;
3155     }
3156 }
3157 
3158 RVVCALL(OPIVV2_RM, vnclipu_vv_b, NOP_UUU_B, H1, H2, H1, vnclipu8)
3159 RVVCALL(OPIVV2_RM, vnclipu_vv_h, NOP_UUU_H, H2, H4, H2, vnclipu16)
3160 RVVCALL(OPIVV2_RM, vnclipu_vv_w, NOP_UUU_W, H4, H8, H4, vnclipu32)
3161 GEN_VEXT_VV_RM(vnclipu_vv_b, 1, 1, clearb)
3162 GEN_VEXT_VV_RM(vnclipu_vv_h, 2, 2, clearh)
3163 GEN_VEXT_VV_RM(vnclipu_vv_w, 4, 4, clearl)
3164 
3165 RVVCALL(OPIVX2_RM, vnclipu_vx_b, NOP_UUU_B, H1, H2, vnclipu8)
3166 RVVCALL(OPIVX2_RM, vnclipu_vx_h, NOP_UUU_H, H2, H4, vnclipu16)
3167 RVVCALL(OPIVX2_RM, vnclipu_vx_w, NOP_UUU_W, H4, H8, vnclipu32)
3168 GEN_VEXT_VX_RM(vnclipu_vx_b, 1, 1, clearb)
3169 GEN_VEXT_VX_RM(vnclipu_vx_h, 2, 2, clearh)
3170 GEN_VEXT_VX_RM(vnclipu_vx_w, 4, 4, clearl)
3171 
3172 /*
3173  *** Vector Float Point Arithmetic Instructions
3174  */
3175 /* Vector Single-Width Floating-Point Add/Subtract Instructions */
3176 #define OPFVV2(NAME, TD, T1, T2, TX1, TX2, HD, HS1, HS2, OP)   \
3177 static void do_##NAME(void *vd, void *vs1, void *vs2, int i,   \
3178                       CPURISCVState *env)                      \
3179 {                                                              \
3180     TX1 s1 = *((T1 *)vs1 + HS1(i));                            \
3181     TX2 s2 = *((T2 *)vs2 + HS2(i));                            \
3182     *((TD *)vd + HD(i)) = OP(s2, s1, &env->fp_status);         \
3183 }
3184 
3185 #define GEN_VEXT_VV_ENV(NAME, ESZ, DSZ, CLEAR_FN)         \
3186 void HELPER(NAME)(void *vd, void *v0, void *vs1,          \
3187                   void *vs2, CPURISCVState *env,          \
3188                   uint32_t desc)                          \
3189 {                                                         \
3190     uint32_t vlmax = vext_maxsz(desc) / ESZ;              \
3191     uint32_t mlen = vext_mlen(desc);                      \
3192     uint32_t vm = vext_vm(desc);                          \
3193     uint32_t vl = env->vl;                                \
3194     uint32_t i;                                           \
3195                                                           \
3196     for (i = 0; i < vl; i++) {                            \
3197         if (!vm && !vext_elem_mask(v0, mlen, i)) {        \
3198             continue;                                     \
3199         }                                                 \
3200         do_##NAME(vd, vs1, vs2, i, env);                  \
3201     }                                                     \
3202     CLEAR_FN(vd, vl, vl * DSZ,  vlmax * DSZ);             \
3203 }
3204 
3205 RVVCALL(OPFVV2, vfadd_vv_h, OP_UUU_H, H2, H2, H2, float16_add)
3206 RVVCALL(OPFVV2, vfadd_vv_w, OP_UUU_W, H4, H4, H4, float32_add)
3207 RVVCALL(OPFVV2, vfadd_vv_d, OP_UUU_D, H8, H8, H8, float64_add)
3208 GEN_VEXT_VV_ENV(vfadd_vv_h, 2, 2, clearh)
3209 GEN_VEXT_VV_ENV(vfadd_vv_w, 4, 4, clearl)
3210 GEN_VEXT_VV_ENV(vfadd_vv_d, 8, 8, clearq)
3211 
3212 #define OPFVF2(NAME, TD, T1, T2, TX1, TX2, HD, HS2, OP)        \
3213 static void do_##NAME(void *vd, uint64_t s1, void *vs2, int i, \
3214                       CPURISCVState *env)                      \
3215 {                                                              \
3216     TX2 s2 = *((T2 *)vs2 + HS2(i));                            \
3217     *((TD *)vd + HD(i)) = OP(s2, (TX1)(T1)s1, &env->fp_status);\
3218 }
3219 
3220 #define GEN_VEXT_VF(NAME, ESZ, DSZ, CLEAR_FN)             \
3221 void HELPER(NAME)(void *vd, void *v0, uint64_t s1,        \
3222                   void *vs2, CPURISCVState *env,          \
3223                   uint32_t desc)                          \
3224 {                                                         \
3225     uint32_t vlmax = vext_maxsz(desc) / ESZ;              \
3226     uint32_t mlen = vext_mlen(desc);                      \
3227     uint32_t vm = vext_vm(desc);                          \
3228     uint32_t vl = env->vl;                                \
3229     uint32_t i;                                           \
3230                                                           \
3231     for (i = 0; i < vl; i++) {                            \
3232         if (!vm && !vext_elem_mask(v0, mlen, i)) {        \
3233             continue;                                     \
3234         }                                                 \
3235         do_##NAME(vd, s1, vs2, i, env);                   \
3236     }                                                     \
3237     CLEAR_FN(vd, vl, vl * DSZ,  vlmax * DSZ);             \
3238 }
3239 
3240 RVVCALL(OPFVF2, vfadd_vf_h, OP_UUU_H, H2, H2, float16_add)
3241 RVVCALL(OPFVF2, vfadd_vf_w, OP_UUU_W, H4, H4, float32_add)
3242 RVVCALL(OPFVF2, vfadd_vf_d, OP_UUU_D, H8, H8, float64_add)
3243 GEN_VEXT_VF(vfadd_vf_h, 2, 2, clearh)
3244 GEN_VEXT_VF(vfadd_vf_w, 4, 4, clearl)
3245 GEN_VEXT_VF(vfadd_vf_d, 8, 8, clearq)
3246 
3247 RVVCALL(OPFVV2, vfsub_vv_h, OP_UUU_H, H2, H2, H2, float16_sub)
3248 RVVCALL(OPFVV2, vfsub_vv_w, OP_UUU_W, H4, H4, H4, float32_sub)
3249 RVVCALL(OPFVV2, vfsub_vv_d, OP_UUU_D, H8, H8, H8, float64_sub)
3250 GEN_VEXT_VV_ENV(vfsub_vv_h, 2, 2, clearh)
3251 GEN_VEXT_VV_ENV(vfsub_vv_w, 4, 4, clearl)
3252 GEN_VEXT_VV_ENV(vfsub_vv_d, 8, 8, clearq)
3253 RVVCALL(OPFVF2, vfsub_vf_h, OP_UUU_H, H2, H2, float16_sub)
3254 RVVCALL(OPFVF2, vfsub_vf_w, OP_UUU_W, H4, H4, float32_sub)
3255 RVVCALL(OPFVF2, vfsub_vf_d, OP_UUU_D, H8, H8, float64_sub)
3256 GEN_VEXT_VF(vfsub_vf_h, 2, 2, clearh)
3257 GEN_VEXT_VF(vfsub_vf_w, 4, 4, clearl)
3258 GEN_VEXT_VF(vfsub_vf_d, 8, 8, clearq)
3259 
3260 static uint16_t float16_rsub(uint16_t a, uint16_t b, float_status *s)
3261 {
3262     return float16_sub(b, a, s);
3263 }
3264 
3265 static uint32_t float32_rsub(uint32_t a, uint32_t b, float_status *s)
3266 {
3267     return float32_sub(b, a, s);
3268 }
3269 
3270 static uint64_t float64_rsub(uint64_t a, uint64_t b, float_status *s)
3271 {
3272     return float64_sub(b, a, s);
3273 }
3274 
3275 RVVCALL(OPFVF2, vfrsub_vf_h, OP_UUU_H, H2, H2, float16_rsub)
3276 RVVCALL(OPFVF2, vfrsub_vf_w, OP_UUU_W, H4, H4, float32_rsub)
3277 RVVCALL(OPFVF2, vfrsub_vf_d, OP_UUU_D, H8, H8, float64_rsub)
3278 GEN_VEXT_VF(vfrsub_vf_h, 2, 2, clearh)
3279 GEN_VEXT_VF(vfrsub_vf_w, 4, 4, clearl)
3280 GEN_VEXT_VF(vfrsub_vf_d, 8, 8, clearq)
3281 
3282 /* Vector Widening Floating-Point Add/Subtract Instructions */
3283 static uint32_t vfwadd16(uint16_t a, uint16_t b, float_status *s)
3284 {
3285     return float32_add(float16_to_float32(a, true, s),
3286             float16_to_float32(b, true, s), s);
3287 }
3288 
3289 static uint64_t vfwadd32(uint32_t a, uint32_t b, float_status *s)
3290 {
3291     return float64_add(float32_to_float64(a, s),
3292             float32_to_float64(b, s), s);
3293 
3294 }
3295 
3296 RVVCALL(OPFVV2, vfwadd_vv_h, WOP_UUU_H, H4, H2, H2, vfwadd16)
3297 RVVCALL(OPFVV2, vfwadd_vv_w, WOP_UUU_W, H8, H4, H4, vfwadd32)
3298 GEN_VEXT_VV_ENV(vfwadd_vv_h, 2, 4, clearl)
3299 GEN_VEXT_VV_ENV(vfwadd_vv_w, 4, 8, clearq)
3300 RVVCALL(OPFVF2, vfwadd_vf_h, WOP_UUU_H, H4, H2, vfwadd16)
3301 RVVCALL(OPFVF2, vfwadd_vf_w, WOP_UUU_W, H8, H4, vfwadd32)
3302 GEN_VEXT_VF(vfwadd_vf_h, 2, 4, clearl)
3303 GEN_VEXT_VF(vfwadd_vf_w, 4, 8, clearq)
3304 
3305 static uint32_t vfwsub16(uint16_t a, uint16_t b, float_status *s)
3306 {
3307     return float32_sub(float16_to_float32(a, true, s),
3308             float16_to_float32(b, true, s), s);
3309 }
3310 
3311 static uint64_t vfwsub32(uint32_t a, uint32_t b, float_status *s)
3312 {
3313     return float64_sub(float32_to_float64(a, s),
3314             float32_to_float64(b, s), s);
3315 
3316 }
3317 
3318 RVVCALL(OPFVV2, vfwsub_vv_h, WOP_UUU_H, H4, H2, H2, vfwsub16)
3319 RVVCALL(OPFVV2, vfwsub_vv_w, WOP_UUU_W, H8, H4, H4, vfwsub32)
3320 GEN_VEXT_VV_ENV(vfwsub_vv_h, 2, 4, clearl)
3321 GEN_VEXT_VV_ENV(vfwsub_vv_w, 4, 8, clearq)
3322 RVVCALL(OPFVF2, vfwsub_vf_h, WOP_UUU_H, H4, H2, vfwsub16)
3323 RVVCALL(OPFVF2, vfwsub_vf_w, WOP_UUU_W, H8, H4, vfwsub32)
3324 GEN_VEXT_VF(vfwsub_vf_h, 2, 4, clearl)
3325 GEN_VEXT_VF(vfwsub_vf_w, 4, 8, clearq)
3326 
3327 static uint32_t vfwaddw16(uint32_t a, uint16_t b, float_status *s)
3328 {
3329     return float32_add(a, float16_to_float32(b, true, s), s);
3330 }
3331 
3332 static uint64_t vfwaddw32(uint64_t a, uint32_t b, float_status *s)
3333 {
3334     return float64_add(a, float32_to_float64(b, s), s);
3335 }
3336 
3337 RVVCALL(OPFVV2, vfwadd_wv_h, WOP_WUUU_H, H4, H2, H2, vfwaddw16)
3338 RVVCALL(OPFVV2, vfwadd_wv_w, WOP_WUUU_W, H8, H4, H4, vfwaddw32)
3339 GEN_VEXT_VV_ENV(vfwadd_wv_h, 2, 4, clearl)
3340 GEN_VEXT_VV_ENV(vfwadd_wv_w, 4, 8, clearq)
3341 RVVCALL(OPFVF2, vfwadd_wf_h, WOP_WUUU_H, H4, H2, vfwaddw16)
3342 RVVCALL(OPFVF2, vfwadd_wf_w, WOP_WUUU_W, H8, H4, vfwaddw32)
3343 GEN_VEXT_VF(vfwadd_wf_h, 2, 4, clearl)
3344 GEN_VEXT_VF(vfwadd_wf_w, 4, 8, clearq)
3345 
3346 static uint32_t vfwsubw16(uint32_t a, uint16_t b, float_status *s)
3347 {
3348     return float32_sub(a, float16_to_float32(b, true, s), s);
3349 }
3350 
3351 static uint64_t vfwsubw32(uint64_t a, uint32_t b, float_status *s)
3352 {
3353     return float64_sub(a, float32_to_float64(b, s), s);
3354 }
3355 
3356 RVVCALL(OPFVV2, vfwsub_wv_h, WOP_WUUU_H, H4, H2, H2, vfwsubw16)
3357 RVVCALL(OPFVV2, vfwsub_wv_w, WOP_WUUU_W, H8, H4, H4, vfwsubw32)
3358 GEN_VEXT_VV_ENV(vfwsub_wv_h, 2, 4, clearl)
3359 GEN_VEXT_VV_ENV(vfwsub_wv_w, 4, 8, clearq)
3360 RVVCALL(OPFVF2, vfwsub_wf_h, WOP_WUUU_H, H4, H2, vfwsubw16)
3361 RVVCALL(OPFVF2, vfwsub_wf_w, WOP_WUUU_W, H8, H4, vfwsubw32)
3362 GEN_VEXT_VF(vfwsub_wf_h, 2, 4, clearl)
3363 GEN_VEXT_VF(vfwsub_wf_w, 4, 8, clearq)
3364 
3365 /* Vector Single-Width Floating-Point Multiply/Divide Instructions */
3366 RVVCALL(OPFVV2, vfmul_vv_h, OP_UUU_H, H2, H2, H2, float16_mul)
3367 RVVCALL(OPFVV2, vfmul_vv_w, OP_UUU_W, H4, H4, H4, float32_mul)
3368 RVVCALL(OPFVV2, vfmul_vv_d, OP_UUU_D, H8, H8, H8, float64_mul)
3369 GEN_VEXT_VV_ENV(vfmul_vv_h, 2, 2, clearh)
3370 GEN_VEXT_VV_ENV(vfmul_vv_w, 4, 4, clearl)
3371 GEN_VEXT_VV_ENV(vfmul_vv_d, 8, 8, clearq)
3372 RVVCALL(OPFVF2, vfmul_vf_h, OP_UUU_H, H2, H2, float16_mul)
3373 RVVCALL(OPFVF2, vfmul_vf_w, OP_UUU_W, H4, H4, float32_mul)
3374 RVVCALL(OPFVF2, vfmul_vf_d, OP_UUU_D, H8, H8, float64_mul)
3375 GEN_VEXT_VF(vfmul_vf_h, 2, 2, clearh)
3376 GEN_VEXT_VF(vfmul_vf_w, 4, 4, clearl)
3377 GEN_VEXT_VF(vfmul_vf_d, 8, 8, clearq)
3378 
3379 RVVCALL(OPFVV2, vfdiv_vv_h, OP_UUU_H, H2, H2, H2, float16_div)
3380 RVVCALL(OPFVV2, vfdiv_vv_w, OP_UUU_W, H4, H4, H4, float32_div)
3381 RVVCALL(OPFVV2, vfdiv_vv_d, OP_UUU_D, H8, H8, H8, float64_div)
3382 GEN_VEXT_VV_ENV(vfdiv_vv_h, 2, 2, clearh)
3383 GEN_VEXT_VV_ENV(vfdiv_vv_w, 4, 4, clearl)
3384 GEN_VEXT_VV_ENV(vfdiv_vv_d, 8, 8, clearq)
3385 RVVCALL(OPFVF2, vfdiv_vf_h, OP_UUU_H, H2, H2, float16_div)
3386 RVVCALL(OPFVF2, vfdiv_vf_w, OP_UUU_W, H4, H4, float32_div)
3387 RVVCALL(OPFVF2, vfdiv_vf_d, OP_UUU_D, H8, H8, float64_div)
3388 GEN_VEXT_VF(vfdiv_vf_h, 2, 2, clearh)
3389 GEN_VEXT_VF(vfdiv_vf_w, 4, 4, clearl)
3390 GEN_VEXT_VF(vfdiv_vf_d, 8, 8, clearq)
3391 
3392 static uint16_t float16_rdiv(uint16_t a, uint16_t b, float_status *s)
3393 {
3394     return float16_div(b, a, s);
3395 }
3396 
3397 static uint32_t float32_rdiv(uint32_t a, uint32_t b, float_status *s)
3398 {
3399     return float32_div(b, a, s);
3400 }
3401 
3402 static uint64_t float64_rdiv(uint64_t a, uint64_t b, float_status *s)
3403 {
3404     return float64_div(b, a, s);
3405 }
3406 
3407 RVVCALL(OPFVF2, vfrdiv_vf_h, OP_UUU_H, H2, H2, float16_rdiv)
3408 RVVCALL(OPFVF2, vfrdiv_vf_w, OP_UUU_W, H4, H4, float32_rdiv)
3409 RVVCALL(OPFVF2, vfrdiv_vf_d, OP_UUU_D, H8, H8, float64_rdiv)
3410 GEN_VEXT_VF(vfrdiv_vf_h, 2, 2, clearh)
3411 GEN_VEXT_VF(vfrdiv_vf_w, 4, 4, clearl)
3412 GEN_VEXT_VF(vfrdiv_vf_d, 8, 8, clearq)
3413 
3414 /* Vector Widening Floating-Point Multiply */
3415 static uint32_t vfwmul16(uint16_t a, uint16_t b, float_status *s)
3416 {
3417     return float32_mul(float16_to_float32(a, true, s),
3418             float16_to_float32(b, true, s), s);
3419 }
3420 
3421 static uint64_t vfwmul32(uint32_t a, uint32_t b, float_status *s)
3422 {
3423     return float64_mul(float32_to_float64(a, s),
3424             float32_to_float64(b, s), s);
3425 
3426 }
3427 RVVCALL(OPFVV2, vfwmul_vv_h, WOP_UUU_H, H4, H2, H2, vfwmul16)
3428 RVVCALL(OPFVV2, vfwmul_vv_w, WOP_UUU_W, H8, H4, H4, vfwmul32)
3429 GEN_VEXT_VV_ENV(vfwmul_vv_h, 2, 4, clearl)
3430 GEN_VEXT_VV_ENV(vfwmul_vv_w, 4, 8, clearq)
3431 RVVCALL(OPFVF2, vfwmul_vf_h, WOP_UUU_H, H4, H2, vfwmul16)
3432 RVVCALL(OPFVF2, vfwmul_vf_w, WOP_UUU_W, H8, H4, vfwmul32)
3433 GEN_VEXT_VF(vfwmul_vf_h, 2, 4, clearl)
3434 GEN_VEXT_VF(vfwmul_vf_w, 4, 8, clearq)
3435 
3436 /* Vector Single-Width Floating-Point Fused Multiply-Add Instructions */
3437 #define OPFVV3(NAME, TD, T1, T2, TX1, TX2, HD, HS1, HS2, OP)       \
3438 static void do_##NAME(void *vd, void *vs1, void *vs2, int i,       \
3439         CPURISCVState *env)                                        \
3440 {                                                                  \
3441     TX1 s1 = *((T1 *)vs1 + HS1(i));                                \
3442     TX2 s2 = *((T2 *)vs2 + HS2(i));                                \
3443     TD d = *((TD *)vd + HD(i));                                    \
3444     *((TD *)vd + HD(i)) = OP(s2, s1, d, &env->fp_status);          \
3445 }
3446 
3447 static uint16_t fmacc16(uint16_t a, uint16_t b, uint16_t d, float_status *s)
3448 {
3449     return float16_muladd(a, b, d, 0, s);
3450 }
3451 
3452 static uint32_t fmacc32(uint32_t a, uint32_t b, uint32_t d, float_status *s)
3453 {
3454     return float32_muladd(a, b, d, 0, s);
3455 }
3456 
3457 static uint64_t fmacc64(uint64_t a, uint64_t b, uint64_t d, float_status *s)
3458 {
3459     return float64_muladd(a, b, d, 0, s);
3460 }
3461 
3462 RVVCALL(OPFVV3, vfmacc_vv_h, OP_UUU_H, H2, H2, H2, fmacc16)
3463 RVVCALL(OPFVV3, vfmacc_vv_w, OP_UUU_W, H4, H4, H4, fmacc32)
3464 RVVCALL(OPFVV3, vfmacc_vv_d, OP_UUU_D, H8, H8, H8, fmacc64)
3465 GEN_VEXT_VV_ENV(vfmacc_vv_h, 2, 2, clearh)
3466 GEN_VEXT_VV_ENV(vfmacc_vv_w, 4, 4, clearl)
3467 GEN_VEXT_VV_ENV(vfmacc_vv_d, 8, 8, clearq)
3468 
3469 #define OPFVF3(NAME, TD, T1, T2, TX1, TX2, HD, HS2, OP)           \
3470 static void do_##NAME(void *vd, uint64_t s1, void *vs2, int i,    \
3471         CPURISCVState *env)                                       \
3472 {                                                                 \
3473     TX2 s2 = *((T2 *)vs2 + HS2(i));                               \
3474     TD d = *((TD *)vd + HD(i));                                   \
3475     *((TD *)vd + HD(i)) = OP(s2, (TX1)(T1)s1, d, &env->fp_status);\
3476 }
3477 
3478 RVVCALL(OPFVF3, vfmacc_vf_h, OP_UUU_H, H2, H2, fmacc16)
3479 RVVCALL(OPFVF3, vfmacc_vf_w, OP_UUU_W, H4, H4, fmacc32)
3480 RVVCALL(OPFVF3, vfmacc_vf_d, OP_UUU_D, H8, H8, fmacc64)
3481 GEN_VEXT_VF(vfmacc_vf_h, 2, 2, clearh)
3482 GEN_VEXT_VF(vfmacc_vf_w, 4, 4, clearl)
3483 GEN_VEXT_VF(vfmacc_vf_d, 8, 8, clearq)
3484 
3485 static uint16_t fnmacc16(uint16_t a, uint16_t b, uint16_t d, float_status *s)
3486 {
3487     return float16_muladd(a, b, d,
3488             float_muladd_negate_c | float_muladd_negate_product, s);
3489 }
3490 
3491 static uint32_t fnmacc32(uint32_t a, uint32_t b, uint32_t d, float_status *s)
3492 {
3493     return float32_muladd(a, b, d,
3494             float_muladd_negate_c | float_muladd_negate_product, s);
3495 }
3496 
3497 static uint64_t fnmacc64(uint64_t a, uint64_t b, uint64_t d, float_status *s)
3498 {
3499     return float64_muladd(a, b, d,
3500             float_muladd_negate_c | float_muladd_negate_product, s);
3501 }
3502 
3503 RVVCALL(OPFVV3, vfnmacc_vv_h, OP_UUU_H, H2, H2, H2, fnmacc16)
3504 RVVCALL(OPFVV3, vfnmacc_vv_w, OP_UUU_W, H4, H4, H4, fnmacc32)
3505 RVVCALL(OPFVV3, vfnmacc_vv_d, OP_UUU_D, H8, H8, H8, fnmacc64)
3506 GEN_VEXT_VV_ENV(vfnmacc_vv_h, 2, 2, clearh)
3507 GEN_VEXT_VV_ENV(vfnmacc_vv_w, 4, 4, clearl)
3508 GEN_VEXT_VV_ENV(vfnmacc_vv_d, 8, 8, clearq)
3509 RVVCALL(OPFVF3, vfnmacc_vf_h, OP_UUU_H, H2, H2, fnmacc16)
3510 RVVCALL(OPFVF3, vfnmacc_vf_w, OP_UUU_W, H4, H4, fnmacc32)
3511 RVVCALL(OPFVF3, vfnmacc_vf_d, OP_UUU_D, H8, H8, fnmacc64)
3512 GEN_VEXT_VF(vfnmacc_vf_h, 2, 2, clearh)
3513 GEN_VEXT_VF(vfnmacc_vf_w, 4, 4, clearl)
3514 GEN_VEXT_VF(vfnmacc_vf_d, 8, 8, clearq)
3515 
3516 static uint16_t fmsac16(uint16_t a, uint16_t b, uint16_t d, float_status *s)
3517 {
3518     return float16_muladd(a, b, d, float_muladd_negate_c, s);
3519 }
3520 
3521 static uint32_t fmsac32(uint32_t a, uint32_t b, uint32_t d, float_status *s)
3522 {
3523     return float32_muladd(a, b, d, float_muladd_negate_c, s);
3524 }
3525 
3526 static uint64_t fmsac64(uint64_t a, uint64_t b, uint64_t d, float_status *s)
3527 {
3528     return float64_muladd(a, b, d, float_muladd_negate_c, s);
3529 }
3530 
3531 RVVCALL(OPFVV3, vfmsac_vv_h, OP_UUU_H, H2, H2, H2, fmsac16)
3532 RVVCALL(OPFVV3, vfmsac_vv_w, OP_UUU_W, H4, H4, H4, fmsac32)
3533 RVVCALL(OPFVV3, vfmsac_vv_d, OP_UUU_D, H8, H8, H8, fmsac64)
3534 GEN_VEXT_VV_ENV(vfmsac_vv_h, 2, 2, clearh)
3535 GEN_VEXT_VV_ENV(vfmsac_vv_w, 4, 4, clearl)
3536 GEN_VEXT_VV_ENV(vfmsac_vv_d, 8, 8, clearq)
3537 RVVCALL(OPFVF3, vfmsac_vf_h, OP_UUU_H, H2, H2, fmsac16)
3538 RVVCALL(OPFVF3, vfmsac_vf_w, OP_UUU_W, H4, H4, fmsac32)
3539 RVVCALL(OPFVF3, vfmsac_vf_d, OP_UUU_D, H8, H8, fmsac64)
3540 GEN_VEXT_VF(vfmsac_vf_h, 2, 2, clearh)
3541 GEN_VEXT_VF(vfmsac_vf_w, 4, 4, clearl)
3542 GEN_VEXT_VF(vfmsac_vf_d, 8, 8, clearq)
3543 
3544 static uint16_t fnmsac16(uint16_t a, uint16_t b, uint16_t d, float_status *s)
3545 {
3546     return float16_muladd(a, b, d, float_muladd_negate_product, s);
3547 }
3548 
3549 static uint32_t fnmsac32(uint32_t a, uint32_t b, uint32_t d, float_status *s)
3550 {
3551     return float32_muladd(a, b, d, float_muladd_negate_product, s);
3552 }
3553 
3554 static uint64_t fnmsac64(uint64_t a, uint64_t b, uint64_t d, float_status *s)
3555 {
3556     return float64_muladd(a, b, d, float_muladd_negate_product, s);
3557 }
3558 
3559 RVVCALL(OPFVV3, vfnmsac_vv_h, OP_UUU_H, H2, H2, H2, fnmsac16)
3560 RVVCALL(OPFVV3, vfnmsac_vv_w, OP_UUU_W, H4, H4, H4, fnmsac32)
3561 RVVCALL(OPFVV3, vfnmsac_vv_d, OP_UUU_D, H8, H8, H8, fnmsac64)
3562 GEN_VEXT_VV_ENV(vfnmsac_vv_h, 2, 2, clearh)
3563 GEN_VEXT_VV_ENV(vfnmsac_vv_w, 4, 4, clearl)
3564 GEN_VEXT_VV_ENV(vfnmsac_vv_d, 8, 8, clearq)
3565 RVVCALL(OPFVF3, vfnmsac_vf_h, OP_UUU_H, H2, H2, fnmsac16)
3566 RVVCALL(OPFVF3, vfnmsac_vf_w, OP_UUU_W, H4, H4, fnmsac32)
3567 RVVCALL(OPFVF3, vfnmsac_vf_d, OP_UUU_D, H8, H8, fnmsac64)
3568 GEN_VEXT_VF(vfnmsac_vf_h, 2, 2, clearh)
3569 GEN_VEXT_VF(vfnmsac_vf_w, 4, 4, clearl)
3570 GEN_VEXT_VF(vfnmsac_vf_d, 8, 8, clearq)
3571 
3572 static uint16_t fmadd16(uint16_t a, uint16_t b, uint16_t d, float_status *s)
3573 {
3574     return float16_muladd(d, b, a, 0, s);
3575 }
3576 
3577 static uint32_t fmadd32(uint32_t a, uint32_t b, uint32_t d, float_status *s)
3578 {
3579     return float32_muladd(d, b, a, 0, s);
3580 }
3581 
3582 static uint64_t fmadd64(uint64_t a, uint64_t b, uint64_t d, float_status *s)
3583 {
3584     return float64_muladd(d, b, a, 0, s);
3585 }
3586 
3587 RVVCALL(OPFVV3, vfmadd_vv_h, OP_UUU_H, H2, H2, H2, fmadd16)
3588 RVVCALL(OPFVV3, vfmadd_vv_w, OP_UUU_W, H4, H4, H4, fmadd32)
3589 RVVCALL(OPFVV3, vfmadd_vv_d, OP_UUU_D, H8, H8, H8, fmadd64)
3590 GEN_VEXT_VV_ENV(vfmadd_vv_h, 2, 2, clearh)
3591 GEN_VEXT_VV_ENV(vfmadd_vv_w, 4, 4, clearl)
3592 GEN_VEXT_VV_ENV(vfmadd_vv_d, 8, 8, clearq)
3593 RVVCALL(OPFVF3, vfmadd_vf_h, OP_UUU_H, H2, H2, fmadd16)
3594 RVVCALL(OPFVF3, vfmadd_vf_w, OP_UUU_W, H4, H4, fmadd32)
3595 RVVCALL(OPFVF3, vfmadd_vf_d, OP_UUU_D, H8, H8, fmadd64)
3596 GEN_VEXT_VF(vfmadd_vf_h, 2, 2, clearh)
3597 GEN_VEXT_VF(vfmadd_vf_w, 4, 4, clearl)
3598 GEN_VEXT_VF(vfmadd_vf_d, 8, 8, clearq)
3599 
3600 static uint16_t fnmadd16(uint16_t a, uint16_t b, uint16_t d, float_status *s)
3601 {
3602     return float16_muladd(d, b, a,
3603             float_muladd_negate_c | float_muladd_negate_product, s);
3604 }
3605 
3606 static uint32_t fnmadd32(uint32_t a, uint32_t b, uint32_t d, float_status *s)
3607 {
3608     return float32_muladd(d, b, a,
3609             float_muladd_negate_c | float_muladd_negate_product, s);
3610 }
3611 
3612 static uint64_t fnmadd64(uint64_t a, uint64_t b, uint64_t d, float_status *s)
3613 {
3614     return float64_muladd(d, b, a,
3615             float_muladd_negate_c | float_muladd_negate_product, s);
3616 }
3617 
3618 RVVCALL(OPFVV3, vfnmadd_vv_h, OP_UUU_H, H2, H2, H2, fnmadd16)
3619 RVVCALL(OPFVV3, vfnmadd_vv_w, OP_UUU_W, H4, H4, H4, fnmadd32)
3620 RVVCALL(OPFVV3, vfnmadd_vv_d, OP_UUU_D, H8, H8, H8, fnmadd64)
3621 GEN_VEXT_VV_ENV(vfnmadd_vv_h, 2, 2, clearh)
3622 GEN_VEXT_VV_ENV(vfnmadd_vv_w, 4, 4, clearl)
3623 GEN_VEXT_VV_ENV(vfnmadd_vv_d, 8, 8, clearq)
3624 RVVCALL(OPFVF3, vfnmadd_vf_h, OP_UUU_H, H2, H2, fnmadd16)
3625 RVVCALL(OPFVF3, vfnmadd_vf_w, OP_UUU_W, H4, H4, fnmadd32)
3626 RVVCALL(OPFVF3, vfnmadd_vf_d, OP_UUU_D, H8, H8, fnmadd64)
3627 GEN_VEXT_VF(vfnmadd_vf_h, 2, 2, clearh)
3628 GEN_VEXT_VF(vfnmadd_vf_w, 4, 4, clearl)
3629 GEN_VEXT_VF(vfnmadd_vf_d, 8, 8, clearq)
3630 
3631 static uint16_t fmsub16(uint16_t a, uint16_t b, uint16_t d, float_status *s)
3632 {
3633     return float16_muladd(d, b, a, float_muladd_negate_c, s);
3634 }
3635 
3636 static uint32_t fmsub32(uint32_t a, uint32_t b, uint32_t d, float_status *s)
3637 {
3638     return float32_muladd(d, b, a, float_muladd_negate_c, s);
3639 }
3640 
3641 static uint64_t fmsub64(uint64_t a, uint64_t b, uint64_t d, float_status *s)
3642 {
3643     return float64_muladd(d, b, a, float_muladd_negate_c, s);
3644 }
3645 
3646 RVVCALL(OPFVV3, vfmsub_vv_h, OP_UUU_H, H2, H2, H2, fmsub16)
3647 RVVCALL(OPFVV3, vfmsub_vv_w, OP_UUU_W, H4, H4, H4, fmsub32)
3648 RVVCALL(OPFVV3, vfmsub_vv_d, OP_UUU_D, H8, H8, H8, fmsub64)
3649 GEN_VEXT_VV_ENV(vfmsub_vv_h, 2, 2, clearh)
3650 GEN_VEXT_VV_ENV(vfmsub_vv_w, 4, 4, clearl)
3651 GEN_VEXT_VV_ENV(vfmsub_vv_d, 8, 8, clearq)
3652 RVVCALL(OPFVF3, vfmsub_vf_h, OP_UUU_H, H2, H2, fmsub16)
3653 RVVCALL(OPFVF3, vfmsub_vf_w, OP_UUU_W, H4, H4, fmsub32)
3654 RVVCALL(OPFVF3, vfmsub_vf_d, OP_UUU_D, H8, H8, fmsub64)
3655 GEN_VEXT_VF(vfmsub_vf_h, 2, 2, clearh)
3656 GEN_VEXT_VF(vfmsub_vf_w, 4, 4, clearl)
3657 GEN_VEXT_VF(vfmsub_vf_d, 8, 8, clearq)
3658 
3659 static uint16_t fnmsub16(uint16_t a, uint16_t b, uint16_t d, float_status *s)
3660 {
3661     return float16_muladd(d, b, a, float_muladd_negate_product, s);
3662 }
3663 
3664 static uint32_t fnmsub32(uint32_t a, uint32_t b, uint32_t d, float_status *s)
3665 {
3666     return float32_muladd(d, b, a, float_muladd_negate_product, s);
3667 }
3668 
3669 static uint64_t fnmsub64(uint64_t a, uint64_t b, uint64_t d, float_status *s)
3670 {
3671     return float64_muladd(d, b, a, float_muladd_negate_product, s);
3672 }
3673 
3674 RVVCALL(OPFVV3, vfnmsub_vv_h, OP_UUU_H, H2, H2, H2, fnmsub16)
3675 RVVCALL(OPFVV3, vfnmsub_vv_w, OP_UUU_W, H4, H4, H4, fnmsub32)
3676 RVVCALL(OPFVV3, vfnmsub_vv_d, OP_UUU_D, H8, H8, H8, fnmsub64)
3677 GEN_VEXT_VV_ENV(vfnmsub_vv_h, 2, 2, clearh)
3678 GEN_VEXT_VV_ENV(vfnmsub_vv_w, 4, 4, clearl)
3679 GEN_VEXT_VV_ENV(vfnmsub_vv_d, 8, 8, clearq)
3680 RVVCALL(OPFVF3, vfnmsub_vf_h, OP_UUU_H, H2, H2, fnmsub16)
3681 RVVCALL(OPFVF3, vfnmsub_vf_w, OP_UUU_W, H4, H4, fnmsub32)
3682 RVVCALL(OPFVF3, vfnmsub_vf_d, OP_UUU_D, H8, H8, fnmsub64)
3683 GEN_VEXT_VF(vfnmsub_vf_h, 2, 2, clearh)
3684 GEN_VEXT_VF(vfnmsub_vf_w, 4, 4, clearl)
3685 GEN_VEXT_VF(vfnmsub_vf_d, 8, 8, clearq)
3686 
3687 /* Vector Widening Floating-Point Fused Multiply-Add Instructions */
3688 static uint32_t fwmacc16(uint16_t a, uint16_t b, uint32_t d, float_status *s)
3689 {
3690     return float32_muladd(float16_to_float32(a, true, s),
3691                         float16_to_float32(b, true, s), d, 0, s);
3692 }
3693 
3694 static uint64_t fwmacc32(uint32_t a, uint32_t b, uint64_t d, float_status *s)
3695 {
3696     return float64_muladd(float32_to_float64(a, s),
3697                         float32_to_float64(b, s), d, 0, s);
3698 }
3699 
3700 RVVCALL(OPFVV3, vfwmacc_vv_h, WOP_UUU_H, H4, H2, H2, fwmacc16)
3701 RVVCALL(OPFVV3, vfwmacc_vv_w, WOP_UUU_W, H8, H4, H4, fwmacc32)
3702 GEN_VEXT_VV_ENV(vfwmacc_vv_h, 2, 4, clearl)
3703 GEN_VEXT_VV_ENV(vfwmacc_vv_w, 4, 8, clearq)
3704 RVVCALL(OPFVF3, vfwmacc_vf_h, WOP_UUU_H, H4, H2, fwmacc16)
3705 RVVCALL(OPFVF3, vfwmacc_vf_w, WOP_UUU_W, H8, H4, fwmacc32)
3706 GEN_VEXT_VF(vfwmacc_vf_h, 2, 4, clearl)
3707 GEN_VEXT_VF(vfwmacc_vf_w, 4, 8, clearq)
3708 
3709 static uint32_t fwnmacc16(uint16_t a, uint16_t b, uint32_t d, float_status *s)
3710 {
3711     return float32_muladd(float16_to_float32(a, true, s),
3712                         float16_to_float32(b, true, s), d,
3713                         float_muladd_negate_c | float_muladd_negate_product, s);
3714 }
3715 
3716 static uint64_t fwnmacc32(uint32_t a, uint32_t b, uint64_t d, float_status *s)
3717 {
3718     return float64_muladd(float32_to_float64(a, s),
3719                         float32_to_float64(b, s), d,
3720                         float_muladd_negate_c | float_muladd_negate_product, s);
3721 }
3722 
3723 RVVCALL(OPFVV3, vfwnmacc_vv_h, WOP_UUU_H, H4, H2, H2, fwnmacc16)
3724 RVVCALL(OPFVV3, vfwnmacc_vv_w, WOP_UUU_W, H8, H4, H4, fwnmacc32)
3725 GEN_VEXT_VV_ENV(vfwnmacc_vv_h, 2, 4, clearl)
3726 GEN_VEXT_VV_ENV(vfwnmacc_vv_w, 4, 8, clearq)
3727 RVVCALL(OPFVF3, vfwnmacc_vf_h, WOP_UUU_H, H4, H2, fwnmacc16)
3728 RVVCALL(OPFVF3, vfwnmacc_vf_w, WOP_UUU_W, H8, H4, fwnmacc32)
3729 GEN_VEXT_VF(vfwnmacc_vf_h, 2, 4, clearl)
3730 GEN_VEXT_VF(vfwnmacc_vf_w, 4, 8, clearq)
3731 
3732 static uint32_t fwmsac16(uint16_t a, uint16_t b, uint32_t d, float_status *s)
3733 {
3734     return float32_muladd(float16_to_float32(a, true, s),
3735                         float16_to_float32(b, true, s), d,
3736                         float_muladd_negate_c, s);
3737 }
3738 
3739 static uint64_t fwmsac32(uint32_t a, uint32_t b, uint64_t d, float_status *s)
3740 {
3741     return float64_muladd(float32_to_float64(a, s),
3742                         float32_to_float64(b, s), d,
3743                         float_muladd_negate_c, s);
3744 }
3745 
3746 RVVCALL(OPFVV3, vfwmsac_vv_h, WOP_UUU_H, H4, H2, H2, fwmsac16)
3747 RVVCALL(OPFVV3, vfwmsac_vv_w, WOP_UUU_W, H8, H4, H4, fwmsac32)
3748 GEN_VEXT_VV_ENV(vfwmsac_vv_h, 2, 4, clearl)
3749 GEN_VEXT_VV_ENV(vfwmsac_vv_w, 4, 8, clearq)
3750 RVVCALL(OPFVF3, vfwmsac_vf_h, WOP_UUU_H, H4, H2, fwmsac16)
3751 RVVCALL(OPFVF3, vfwmsac_vf_w, WOP_UUU_W, H8, H4, fwmsac32)
3752 GEN_VEXT_VF(vfwmsac_vf_h, 2, 4, clearl)
3753 GEN_VEXT_VF(vfwmsac_vf_w, 4, 8, clearq)
3754 
3755 static uint32_t fwnmsac16(uint16_t a, uint16_t b, uint32_t d, float_status *s)
3756 {
3757     return float32_muladd(float16_to_float32(a, true, s),
3758                         float16_to_float32(b, true, s), d,
3759                         float_muladd_negate_product, s);
3760 }
3761 
3762 static uint64_t fwnmsac32(uint32_t a, uint32_t b, uint64_t d, float_status *s)
3763 {
3764     return float64_muladd(float32_to_float64(a, s),
3765                         float32_to_float64(b, s), d,
3766                         float_muladd_negate_product, s);
3767 }
3768 
3769 RVVCALL(OPFVV3, vfwnmsac_vv_h, WOP_UUU_H, H4, H2, H2, fwnmsac16)
3770 RVVCALL(OPFVV3, vfwnmsac_vv_w, WOP_UUU_W, H8, H4, H4, fwnmsac32)
3771 GEN_VEXT_VV_ENV(vfwnmsac_vv_h, 2, 4, clearl)
3772 GEN_VEXT_VV_ENV(vfwnmsac_vv_w, 4, 8, clearq)
3773 RVVCALL(OPFVF3, vfwnmsac_vf_h, WOP_UUU_H, H4, H2, fwnmsac16)
3774 RVVCALL(OPFVF3, vfwnmsac_vf_w, WOP_UUU_W, H8, H4, fwnmsac32)
3775 GEN_VEXT_VF(vfwnmsac_vf_h, 2, 4, clearl)
3776 GEN_VEXT_VF(vfwnmsac_vf_w, 4, 8, clearq)
3777 
3778 /* Vector Floating-Point Square-Root Instruction */
3779 /* (TD, T2, TX2) */
3780 #define OP_UU_H uint16_t, uint16_t, uint16_t
3781 #define OP_UU_W uint32_t, uint32_t, uint32_t
3782 #define OP_UU_D uint64_t, uint64_t, uint64_t
3783 
3784 #define OPFVV1(NAME, TD, T2, TX2, HD, HS2, OP)        \
3785 static void do_##NAME(void *vd, void *vs2, int i,      \
3786         CPURISCVState *env)                            \
3787 {                                                      \
3788     TX2 s2 = *((T2 *)vs2 + HS2(i));                    \
3789     *((TD *)vd + HD(i)) = OP(s2, &env->fp_status);     \
3790 }
3791 
3792 #define GEN_VEXT_V_ENV(NAME, ESZ, DSZ, CLEAR_FN)       \
3793 void HELPER(NAME)(void *vd, void *v0, void *vs2,       \
3794         CPURISCVState *env, uint32_t desc)             \
3795 {                                                      \
3796     uint32_t vlmax = vext_maxsz(desc) / ESZ;           \
3797     uint32_t mlen = vext_mlen(desc);                   \
3798     uint32_t vm = vext_vm(desc);                       \
3799     uint32_t vl = env->vl;                             \
3800     uint32_t i;                                        \
3801                                                        \
3802     if (vl == 0) {                                     \
3803         return;                                        \
3804     }                                                  \
3805     for (i = 0; i < vl; i++) {                         \
3806         if (!vm && !vext_elem_mask(v0, mlen, i)) {     \
3807             continue;                                  \
3808         }                                              \
3809         do_##NAME(vd, vs2, i, env);                    \
3810     }                                                  \
3811     CLEAR_FN(vd, vl, vl * DSZ,  vlmax * DSZ);          \
3812 }
3813 
3814 RVVCALL(OPFVV1, vfsqrt_v_h, OP_UU_H, H2, H2, float16_sqrt)
3815 RVVCALL(OPFVV1, vfsqrt_v_w, OP_UU_W, H4, H4, float32_sqrt)
3816 RVVCALL(OPFVV1, vfsqrt_v_d, OP_UU_D, H8, H8, float64_sqrt)
3817 GEN_VEXT_V_ENV(vfsqrt_v_h, 2, 2, clearh)
3818 GEN_VEXT_V_ENV(vfsqrt_v_w, 4, 4, clearl)
3819 GEN_VEXT_V_ENV(vfsqrt_v_d, 8, 8, clearq)
3820 
3821 /* Vector Floating-Point MIN/MAX Instructions */
3822 RVVCALL(OPFVV2, vfmin_vv_h, OP_UUU_H, H2, H2, H2, float16_minnum)
3823 RVVCALL(OPFVV2, vfmin_vv_w, OP_UUU_W, H4, H4, H4, float32_minnum)
3824 RVVCALL(OPFVV2, vfmin_vv_d, OP_UUU_D, H8, H8, H8, float64_minnum)
3825 GEN_VEXT_VV_ENV(vfmin_vv_h, 2, 2, clearh)
3826 GEN_VEXT_VV_ENV(vfmin_vv_w, 4, 4, clearl)
3827 GEN_VEXT_VV_ENV(vfmin_vv_d, 8, 8, clearq)
3828 RVVCALL(OPFVF2, vfmin_vf_h, OP_UUU_H, H2, H2, float16_minnum)
3829 RVVCALL(OPFVF2, vfmin_vf_w, OP_UUU_W, H4, H4, float32_minnum)
3830 RVVCALL(OPFVF2, vfmin_vf_d, OP_UUU_D, H8, H8, float64_minnum)
3831 GEN_VEXT_VF(vfmin_vf_h, 2, 2, clearh)
3832 GEN_VEXT_VF(vfmin_vf_w, 4, 4, clearl)
3833 GEN_VEXT_VF(vfmin_vf_d, 8, 8, clearq)
3834 
3835 RVVCALL(OPFVV2, vfmax_vv_h, OP_UUU_H, H2, H2, H2, float16_maxnum)
3836 RVVCALL(OPFVV2, vfmax_vv_w, OP_UUU_W, H4, H4, H4, float32_maxnum)
3837 RVVCALL(OPFVV2, vfmax_vv_d, OP_UUU_D, H8, H8, H8, float64_maxnum)
3838 GEN_VEXT_VV_ENV(vfmax_vv_h, 2, 2, clearh)
3839 GEN_VEXT_VV_ENV(vfmax_vv_w, 4, 4, clearl)
3840 GEN_VEXT_VV_ENV(vfmax_vv_d, 8, 8, clearq)
3841 RVVCALL(OPFVF2, vfmax_vf_h, OP_UUU_H, H2, H2, float16_maxnum)
3842 RVVCALL(OPFVF2, vfmax_vf_w, OP_UUU_W, H4, H4, float32_maxnum)
3843 RVVCALL(OPFVF2, vfmax_vf_d, OP_UUU_D, H8, H8, float64_maxnum)
3844 GEN_VEXT_VF(vfmax_vf_h, 2, 2, clearh)
3845 GEN_VEXT_VF(vfmax_vf_w, 4, 4, clearl)
3846 GEN_VEXT_VF(vfmax_vf_d, 8, 8, clearq)
3847 
3848 /* Vector Floating-Point Sign-Injection Instructions */
3849 static uint16_t fsgnj16(uint16_t a, uint16_t b, float_status *s)
3850 {
3851     return deposit64(b, 0, 15, a);
3852 }
3853 
3854 static uint32_t fsgnj32(uint32_t a, uint32_t b, float_status *s)
3855 {
3856     return deposit64(b, 0, 31, a);
3857 }
3858 
3859 static uint64_t fsgnj64(uint64_t a, uint64_t b, float_status *s)
3860 {
3861     return deposit64(b, 0, 63, a);
3862 }
3863 
3864 RVVCALL(OPFVV2, vfsgnj_vv_h, OP_UUU_H, H2, H2, H2, fsgnj16)
3865 RVVCALL(OPFVV2, vfsgnj_vv_w, OP_UUU_W, H4, H4, H4, fsgnj32)
3866 RVVCALL(OPFVV2, vfsgnj_vv_d, OP_UUU_D, H8, H8, H8, fsgnj64)
3867 GEN_VEXT_VV_ENV(vfsgnj_vv_h, 2, 2, clearh)
3868 GEN_VEXT_VV_ENV(vfsgnj_vv_w, 4, 4, clearl)
3869 GEN_VEXT_VV_ENV(vfsgnj_vv_d, 8, 8, clearq)
3870 RVVCALL(OPFVF2, vfsgnj_vf_h, OP_UUU_H, H2, H2, fsgnj16)
3871 RVVCALL(OPFVF2, vfsgnj_vf_w, OP_UUU_W, H4, H4, fsgnj32)
3872 RVVCALL(OPFVF2, vfsgnj_vf_d, OP_UUU_D, H8, H8, fsgnj64)
3873 GEN_VEXT_VF(vfsgnj_vf_h, 2, 2, clearh)
3874 GEN_VEXT_VF(vfsgnj_vf_w, 4, 4, clearl)
3875 GEN_VEXT_VF(vfsgnj_vf_d, 8, 8, clearq)
3876 
3877 static uint16_t fsgnjn16(uint16_t a, uint16_t b, float_status *s)
3878 {
3879     return deposit64(~b, 0, 15, a);
3880 }
3881 
3882 static uint32_t fsgnjn32(uint32_t a, uint32_t b, float_status *s)
3883 {
3884     return deposit64(~b, 0, 31, a);
3885 }
3886 
3887 static uint64_t fsgnjn64(uint64_t a, uint64_t b, float_status *s)
3888 {
3889     return deposit64(~b, 0, 63, a);
3890 }
3891 
3892 RVVCALL(OPFVV2, vfsgnjn_vv_h, OP_UUU_H, H2, H2, H2, fsgnjn16)
3893 RVVCALL(OPFVV2, vfsgnjn_vv_w, OP_UUU_W, H4, H4, H4, fsgnjn32)
3894 RVVCALL(OPFVV2, vfsgnjn_vv_d, OP_UUU_D, H8, H8, H8, fsgnjn64)
3895 GEN_VEXT_VV_ENV(vfsgnjn_vv_h, 2, 2, clearh)
3896 GEN_VEXT_VV_ENV(vfsgnjn_vv_w, 4, 4, clearl)
3897 GEN_VEXT_VV_ENV(vfsgnjn_vv_d, 8, 8, clearq)
3898 RVVCALL(OPFVF2, vfsgnjn_vf_h, OP_UUU_H, H2, H2, fsgnjn16)
3899 RVVCALL(OPFVF2, vfsgnjn_vf_w, OP_UUU_W, H4, H4, fsgnjn32)
3900 RVVCALL(OPFVF2, vfsgnjn_vf_d, OP_UUU_D, H8, H8, fsgnjn64)
3901 GEN_VEXT_VF(vfsgnjn_vf_h, 2, 2, clearh)
3902 GEN_VEXT_VF(vfsgnjn_vf_w, 4, 4, clearl)
3903 GEN_VEXT_VF(vfsgnjn_vf_d, 8, 8, clearq)
3904 
3905 static uint16_t fsgnjx16(uint16_t a, uint16_t b, float_status *s)
3906 {
3907     return deposit64(b ^ a, 0, 15, a);
3908 }
3909 
3910 static uint32_t fsgnjx32(uint32_t a, uint32_t b, float_status *s)
3911 {
3912     return deposit64(b ^ a, 0, 31, a);
3913 }
3914 
3915 static uint64_t fsgnjx64(uint64_t a, uint64_t b, float_status *s)
3916 {
3917     return deposit64(b ^ a, 0, 63, a);
3918 }
3919 
3920 RVVCALL(OPFVV2, vfsgnjx_vv_h, OP_UUU_H, H2, H2, H2, fsgnjx16)
3921 RVVCALL(OPFVV2, vfsgnjx_vv_w, OP_UUU_W, H4, H4, H4, fsgnjx32)
3922 RVVCALL(OPFVV2, vfsgnjx_vv_d, OP_UUU_D, H8, H8, H8, fsgnjx64)
3923 GEN_VEXT_VV_ENV(vfsgnjx_vv_h, 2, 2, clearh)
3924 GEN_VEXT_VV_ENV(vfsgnjx_vv_w, 4, 4, clearl)
3925 GEN_VEXT_VV_ENV(vfsgnjx_vv_d, 8, 8, clearq)
3926 RVVCALL(OPFVF2, vfsgnjx_vf_h, OP_UUU_H, H2, H2, fsgnjx16)
3927 RVVCALL(OPFVF2, vfsgnjx_vf_w, OP_UUU_W, H4, H4, fsgnjx32)
3928 RVVCALL(OPFVF2, vfsgnjx_vf_d, OP_UUU_D, H8, H8, fsgnjx64)
3929 GEN_VEXT_VF(vfsgnjx_vf_h, 2, 2, clearh)
3930 GEN_VEXT_VF(vfsgnjx_vf_w, 4, 4, clearl)
3931 GEN_VEXT_VF(vfsgnjx_vf_d, 8, 8, clearq)
3932 
3933 /* Vector Floating-Point Compare Instructions */
3934 #define GEN_VEXT_CMP_VV_ENV(NAME, ETYPE, H, DO_OP)            \
3935 void HELPER(NAME)(void *vd, void *v0, void *vs1, void *vs2,   \
3936                   CPURISCVState *env, uint32_t desc)          \
3937 {                                                             \
3938     uint32_t mlen = vext_mlen(desc);                          \
3939     uint32_t vm = vext_vm(desc);                              \
3940     uint32_t vl = env->vl;                                    \
3941     uint32_t vlmax = vext_maxsz(desc) / sizeof(ETYPE);        \
3942     uint32_t i;                                               \
3943                                                               \
3944     for (i = 0; i < vl; i++) {                                \
3945         ETYPE s1 = *((ETYPE *)vs1 + H(i));                    \
3946         ETYPE s2 = *((ETYPE *)vs2 + H(i));                    \
3947         if (!vm && !vext_elem_mask(v0, mlen, i)) {            \
3948             continue;                                         \
3949         }                                                     \
3950         vext_set_elem_mask(vd, mlen, i,                       \
3951                            DO_OP(s2, s1, &env->fp_status));   \
3952     }                                                         \
3953     for (; i < vlmax; i++) {                                  \
3954         vext_set_elem_mask(vd, mlen, i, 0);                   \
3955     }                                                         \
3956 }
3957 
3958 static bool float16_eq_quiet(uint16_t a, uint16_t b, float_status *s)
3959 {
3960     FloatRelation compare = float16_compare_quiet(a, b, s);
3961     return compare == float_relation_equal;
3962 }
3963 
3964 GEN_VEXT_CMP_VV_ENV(vmfeq_vv_h, uint16_t, H2, float16_eq_quiet)
3965 GEN_VEXT_CMP_VV_ENV(vmfeq_vv_w, uint32_t, H4, float32_eq_quiet)
3966 GEN_VEXT_CMP_VV_ENV(vmfeq_vv_d, uint64_t, H8, float64_eq_quiet)
3967 
3968 #define GEN_VEXT_CMP_VF(NAME, ETYPE, H, DO_OP)                      \
3969 void HELPER(NAME)(void *vd, void *v0, uint64_t s1, void *vs2,       \
3970                   CPURISCVState *env, uint32_t desc)                \
3971 {                                                                   \
3972     uint32_t mlen = vext_mlen(desc);                                \
3973     uint32_t vm = vext_vm(desc);                                    \
3974     uint32_t vl = env->vl;                                          \
3975     uint32_t vlmax = vext_maxsz(desc) / sizeof(ETYPE);              \
3976     uint32_t i;                                                     \
3977                                                                     \
3978     for (i = 0; i < vl; i++) {                                      \
3979         ETYPE s2 = *((ETYPE *)vs2 + H(i));                          \
3980         if (!vm && !vext_elem_mask(v0, mlen, i)) {                  \
3981             continue;                                               \
3982         }                                                           \
3983         vext_set_elem_mask(vd, mlen, i,                             \
3984                            DO_OP(s2, (ETYPE)s1, &env->fp_status));  \
3985     }                                                               \
3986     for (; i < vlmax; i++) {                                        \
3987         vext_set_elem_mask(vd, mlen, i, 0);                         \
3988     }                                                               \
3989 }
3990 
3991 GEN_VEXT_CMP_VF(vmfeq_vf_h, uint16_t, H2, float16_eq_quiet)
3992 GEN_VEXT_CMP_VF(vmfeq_vf_w, uint32_t, H4, float32_eq_quiet)
3993 GEN_VEXT_CMP_VF(vmfeq_vf_d, uint64_t, H8, float64_eq_quiet)
3994 
3995 static bool vmfne16(uint16_t a, uint16_t b, float_status *s)
3996 {
3997     FloatRelation compare = float16_compare_quiet(a, b, s);
3998     return compare != float_relation_equal;
3999 }
4000 
4001 static bool vmfne32(uint32_t a, uint32_t b, float_status *s)
4002 {
4003     FloatRelation compare = float32_compare_quiet(a, b, s);
4004     return compare != float_relation_equal;
4005 }
4006 
4007 static bool vmfne64(uint64_t a, uint64_t b, float_status *s)
4008 {
4009     FloatRelation compare = float64_compare_quiet(a, b, s);
4010     return compare != float_relation_equal;
4011 }
4012 
4013 GEN_VEXT_CMP_VV_ENV(vmfne_vv_h, uint16_t, H2, vmfne16)
4014 GEN_VEXT_CMP_VV_ENV(vmfne_vv_w, uint32_t, H4, vmfne32)
4015 GEN_VEXT_CMP_VV_ENV(vmfne_vv_d, uint64_t, H8, vmfne64)
4016 GEN_VEXT_CMP_VF(vmfne_vf_h, uint16_t, H2, vmfne16)
4017 GEN_VEXT_CMP_VF(vmfne_vf_w, uint32_t, H4, vmfne32)
4018 GEN_VEXT_CMP_VF(vmfne_vf_d, uint64_t, H8, vmfne64)
4019 
4020 static bool float16_lt(uint16_t a, uint16_t b, float_status *s)
4021 {
4022     FloatRelation compare = float16_compare(a, b, s);
4023     return compare == float_relation_less;
4024 }
4025 
4026 GEN_VEXT_CMP_VV_ENV(vmflt_vv_h, uint16_t, H2, float16_lt)
4027 GEN_VEXT_CMP_VV_ENV(vmflt_vv_w, uint32_t, H4, float32_lt)
4028 GEN_VEXT_CMP_VV_ENV(vmflt_vv_d, uint64_t, H8, float64_lt)
4029 GEN_VEXT_CMP_VF(vmflt_vf_h, uint16_t, H2, float16_lt)
4030 GEN_VEXT_CMP_VF(vmflt_vf_w, uint32_t, H4, float32_lt)
4031 GEN_VEXT_CMP_VF(vmflt_vf_d, uint64_t, H8, float64_lt)
4032 
4033 static bool float16_le(uint16_t a, uint16_t b, float_status *s)
4034 {
4035     FloatRelation compare = float16_compare(a, b, s);
4036     return compare == float_relation_less ||
4037            compare == float_relation_equal;
4038 }
4039 
4040 GEN_VEXT_CMP_VV_ENV(vmfle_vv_h, uint16_t, H2, float16_le)
4041 GEN_VEXT_CMP_VV_ENV(vmfle_vv_w, uint32_t, H4, float32_le)
4042 GEN_VEXT_CMP_VV_ENV(vmfle_vv_d, uint64_t, H8, float64_le)
4043 GEN_VEXT_CMP_VF(vmfle_vf_h, uint16_t, H2, float16_le)
4044 GEN_VEXT_CMP_VF(vmfle_vf_w, uint32_t, H4, float32_le)
4045 GEN_VEXT_CMP_VF(vmfle_vf_d, uint64_t, H8, float64_le)
4046 
4047 static bool vmfgt16(uint16_t a, uint16_t b, float_status *s)
4048 {
4049     FloatRelation compare = float16_compare(a, b, s);
4050     return compare == float_relation_greater;
4051 }
4052 
4053 static bool vmfgt32(uint32_t a, uint32_t b, float_status *s)
4054 {
4055     FloatRelation compare = float32_compare(a, b, s);
4056     return compare == float_relation_greater;
4057 }
4058 
4059 static bool vmfgt64(uint64_t a, uint64_t b, float_status *s)
4060 {
4061     FloatRelation compare = float64_compare(a, b, s);
4062     return compare == float_relation_greater;
4063 }
4064 
4065 GEN_VEXT_CMP_VF(vmfgt_vf_h, uint16_t, H2, vmfgt16)
4066 GEN_VEXT_CMP_VF(vmfgt_vf_w, uint32_t, H4, vmfgt32)
4067 GEN_VEXT_CMP_VF(vmfgt_vf_d, uint64_t, H8, vmfgt64)
4068 
4069 static bool vmfge16(uint16_t a, uint16_t b, float_status *s)
4070 {
4071     FloatRelation compare = float16_compare(a, b, s);
4072     return compare == float_relation_greater ||
4073            compare == float_relation_equal;
4074 }
4075 
4076 static bool vmfge32(uint32_t a, uint32_t b, float_status *s)
4077 {
4078     FloatRelation compare = float32_compare(a, b, s);
4079     return compare == float_relation_greater ||
4080            compare == float_relation_equal;
4081 }
4082 
4083 static bool vmfge64(uint64_t a, uint64_t b, float_status *s)
4084 {
4085     FloatRelation compare = float64_compare(a, b, s);
4086     return compare == float_relation_greater ||
4087            compare == float_relation_equal;
4088 }
4089 
4090 GEN_VEXT_CMP_VF(vmfge_vf_h, uint16_t, H2, vmfge16)
4091 GEN_VEXT_CMP_VF(vmfge_vf_w, uint32_t, H4, vmfge32)
4092 GEN_VEXT_CMP_VF(vmfge_vf_d, uint64_t, H8, vmfge64)
4093 
4094 static bool float16_unordered_quiet(uint16_t a, uint16_t b, float_status *s)
4095 {
4096     FloatRelation compare = float16_compare_quiet(a, b, s);
4097     return compare == float_relation_unordered;
4098 }
4099 
4100 GEN_VEXT_CMP_VV_ENV(vmford_vv_h, uint16_t, H2, !float16_unordered_quiet)
4101 GEN_VEXT_CMP_VV_ENV(vmford_vv_w, uint32_t, H4, !float32_unordered_quiet)
4102 GEN_VEXT_CMP_VV_ENV(vmford_vv_d, uint64_t, H8, !float64_unordered_quiet)
4103 GEN_VEXT_CMP_VF(vmford_vf_h, uint16_t, H2, !float16_unordered_quiet)
4104 GEN_VEXT_CMP_VF(vmford_vf_w, uint32_t, H4, !float32_unordered_quiet)
4105 GEN_VEXT_CMP_VF(vmford_vf_d, uint64_t, H8, !float64_unordered_quiet)
4106 
4107 /* Vector Floating-Point Classify Instruction */
4108 #define OPIVV1(NAME, TD, T2, TX2, HD, HS2, OP)         \
4109 static void do_##NAME(void *vd, void *vs2, int i)      \
4110 {                                                      \
4111     TX2 s2 = *((T2 *)vs2 + HS2(i));                    \
4112     *((TD *)vd + HD(i)) = OP(s2);                      \
4113 }
4114 
4115 #define GEN_VEXT_V(NAME, ESZ, DSZ, CLEAR_FN)           \
4116 void HELPER(NAME)(void *vd, void *v0, void *vs2,       \
4117                   CPURISCVState *env, uint32_t desc)   \
4118 {                                                      \
4119     uint32_t vlmax = vext_maxsz(desc) / ESZ;           \
4120     uint32_t mlen = vext_mlen(desc);                   \
4121     uint32_t vm = vext_vm(desc);                       \
4122     uint32_t vl = env->vl;                             \
4123     uint32_t i;                                        \
4124                                                        \
4125     for (i = 0; i < vl; i++) {                         \
4126         if (!vm && !vext_elem_mask(v0, mlen, i)) {     \
4127             continue;                                  \
4128         }                                              \
4129         do_##NAME(vd, vs2, i);                         \
4130     }                                                  \
4131     CLEAR_FN(vd, vl, vl * DSZ,  vlmax * DSZ);          \
4132 }
4133 
4134 target_ulong fclass_h(uint64_t frs1)
4135 {
4136     float16 f = frs1;
4137     bool sign = float16_is_neg(f);
4138 
4139     if (float16_is_infinity(f)) {
4140         return sign ? 1 << 0 : 1 << 7;
4141     } else if (float16_is_zero(f)) {
4142         return sign ? 1 << 3 : 1 << 4;
4143     } else if (float16_is_zero_or_denormal(f)) {
4144         return sign ? 1 << 2 : 1 << 5;
4145     } else if (float16_is_any_nan(f)) {
4146         float_status s = { }; /* for snan_bit_is_one */
4147         return float16_is_quiet_nan(f, &s) ? 1 << 9 : 1 << 8;
4148     } else {
4149         return sign ? 1 << 1 : 1 << 6;
4150     }
4151 }
4152 
4153 target_ulong fclass_s(uint64_t frs1)
4154 {
4155     float32 f = frs1;
4156     bool sign = float32_is_neg(f);
4157 
4158     if (float32_is_infinity(f)) {
4159         return sign ? 1 << 0 : 1 << 7;
4160     } else if (float32_is_zero(f)) {
4161         return sign ? 1 << 3 : 1 << 4;
4162     } else if (float32_is_zero_or_denormal(f)) {
4163         return sign ? 1 << 2 : 1 << 5;
4164     } else if (float32_is_any_nan(f)) {
4165         float_status s = { }; /* for snan_bit_is_one */
4166         return float32_is_quiet_nan(f, &s) ? 1 << 9 : 1 << 8;
4167     } else {
4168         return sign ? 1 << 1 : 1 << 6;
4169     }
4170 }
4171 
4172 target_ulong fclass_d(uint64_t frs1)
4173 {
4174     float64 f = frs1;
4175     bool sign = float64_is_neg(f);
4176 
4177     if (float64_is_infinity(f)) {
4178         return sign ? 1 << 0 : 1 << 7;
4179     } else if (float64_is_zero(f)) {
4180         return sign ? 1 << 3 : 1 << 4;
4181     } else if (float64_is_zero_or_denormal(f)) {
4182         return sign ? 1 << 2 : 1 << 5;
4183     } else if (float64_is_any_nan(f)) {
4184         float_status s = { }; /* for snan_bit_is_one */
4185         return float64_is_quiet_nan(f, &s) ? 1 << 9 : 1 << 8;
4186     } else {
4187         return sign ? 1 << 1 : 1 << 6;
4188     }
4189 }
4190 
4191 RVVCALL(OPIVV1, vfclass_v_h, OP_UU_H, H2, H2, fclass_h)
4192 RVVCALL(OPIVV1, vfclass_v_w, OP_UU_W, H4, H4, fclass_s)
4193 RVVCALL(OPIVV1, vfclass_v_d, OP_UU_D, H8, H8, fclass_d)
4194 GEN_VEXT_V(vfclass_v_h, 2, 2, clearh)
4195 GEN_VEXT_V(vfclass_v_w, 4, 4, clearl)
4196 GEN_VEXT_V(vfclass_v_d, 8, 8, clearq)
4197 
4198 /* Vector Floating-Point Merge Instruction */
4199 #define GEN_VFMERGE_VF(NAME, ETYPE, H, CLEAR_FN)              \
4200 void HELPER(NAME)(void *vd, void *v0, uint64_t s1, void *vs2, \
4201                   CPURISCVState *env, uint32_t desc)          \
4202 {                                                             \
4203     uint32_t mlen = vext_mlen(desc);                          \
4204     uint32_t vm = vext_vm(desc);                              \
4205     uint32_t vl = env->vl;                                    \
4206     uint32_t esz = sizeof(ETYPE);                             \
4207     uint32_t vlmax = vext_maxsz(desc) / esz;                  \
4208     uint32_t i;                                               \
4209                                                               \
4210     for (i = 0; i < vl; i++) {                                \
4211         ETYPE s2 = *((ETYPE *)vs2 + H(i));                    \
4212         *((ETYPE *)vd + H(i))                                 \
4213           = (!vm && !vext_elem_mask(v0, mlen, i) ? s2 : s1);  \
4214     }                                                         \
4215     CLEAR_FN(vd, vl, vl * esz, vlmax * esz);                  \
4216 }
4217 
4218 GEN_VFMERGE_VF(vfmerge_vfm_h, int16_t, H2, clearh)
4219 GEN_VFMERGE_VF(vfmerge_vfm_w, int32_t, H4, clearl)
4220 GEN_VFMERGE_VF(vfmerge_vfm_d, int64_t, H8, clearq)
4221 
4222 /* Single-Width Floating-Point/Integer Type-Convert Instructions */
4223 /* vfcvt.xu.f.v vd, vs2, vm # Convert float to unsigned integer. */
4224 RVVCALL(OPFVV1, vfcvt_xu_f_v_h, OP_UU_H, H2, H2, float16_to_uint16)
4225 RVVCALL(OPFVV1, vfcvt_xu_f_v_w, OP_UU_W, H4, H4, float32_to_uint32)
4226 RVVCALL(OPFVV1, vfcvt_xu_f_v_d, OP_UU_D, H8, H8, float64_to_uint64)
4227 GEN_VEXT_V_ENV(vfcvt_xu_f_v_h, 2, 2, clearh)
4228 GEN_VEXT_V_ENV(vfcvt_xu_f_v_w, 4, 4, clearl)
4229 GEN_VEXT_V_ENV(vfcvt_xu_f_v_d, 8, 8, clearq)
4230 
4231 /* vfcvt.x.f.v vd, vs2, vm # Convert float to signed integer. */
4232 RVVCALL(OPFVV1, vfcvt_x_f_v_h, OP_UU_H, H2, H2, float16_to_int16)
4233 RVVCALL(OPFVV1, vfcvt_x_f_v_w, OP_UU_W, H4, H4, float32_to_int32)
4234 RVVCALL(OPFVV1, vfcvt_x_f_v_d, OP_UU_D, H8, H8, float64_to_int64)
4235 GEN_VEXT_V_ENV(vfcvt_x_f_v_h, 2, 2, clearh)
4236 GEN_VEXT_V_ENV(vfcvt_x_f_v_w, 4, 4, clearl)
4237 GEN_VEXT_V_ENV(vfcvt_x_f_v_d, 8, 8, clearq)
4238 
4239 /* vfcvt.f.xu.v vd, vs2, vm # Convert unsigned integer to float. */
4240 RVVCALL(OPFVV1, vfcvt_f_xu_v_h, OP_UU_H, H2, H2, uint16_to_float16)
4241 RVVCALL(OPFVV1, vfcvt_f_xu_v_w, OP_UU_W, H4, H4, uint32_to_float32)
4242 RVVCALL(OPFVV1, vfcvt_f_xu_v_d, OP_UU_D, H8, H8, uint64_to_float64)
4243 GEN_VEXT_V_ENV(vfcvt_f_xu_v_h, 2, 2, clearh)
4244 GEN_VEXT_V_ENV(vfcvt_f_xu_v_w, 4, 4, clearl)
4245 GEN_VEXT_V_ENV(vfcvt_f_xu_v_d, 8, 8, clearq)
4246 
4247 /* vfcvt.f.x.v vd, vs2, vm # Convert integer to float. */
4248 RVVCALL(OPFVV1, vfcvt_f_x_v_h, OP_UU_H, H2, H2, int16_to_float16)
4249 RVVCALL(OPFVV1, vfcvt_f_x_v_w, OP_UU_W, H4, H4, int32_to_float32)
4250 RVVCALL(OPFVV1, vfcvt_f_x_v_d, OP_UU_D, H8, H8, int64_to_float64)
4251 GEN_VEXT_V_ENV(vfcvt_f_x_v_h, 2, 2, clearh)
4252 GEN_VEXT_V_ENV(vfcvt_f_x_v_w, 4, 4, clearl)
4253 GEN_VEXT_V_ENV(vfcvt_f_x_v_d, 8, 8, clearq)
4254 
4255 /* Widening Floating-Point/Integer Type-Convert Instructions */
4256 /* (TD, T2, TX2) */
4257 #define WOP_UU_H uint32_t, uint16_t, uint16_t
4258 #define WOP_UU_W uint64_t, uint32_t, uint32_t
4259 /* vfwcvt.xu.f.v vd, vs2, vm # Convert float to double-width unsigned integer.*/
4260 RVVCALL(OPFVV1, vfwcvt_xu_f_v_h, WOP_UU_H, H4, H2, float16_to_uint32)
4261 RVVCALL(OPFVV1, vfwcvt_xu_f_v_w, WOP_UU_W, H8, H4, float32_to_uint64)
4262 GEN_VEXT_V_ENV(vfwcvt_xu_f_v_h, 2, 4, clearl)
4263 GEN_VEXT_V_ENV(vfwcvt_xu_f_v_w, 4, 8, clearq)
4264 
4265 /* vfwcvt.x.f.v vd, vs2, vm # Convert float to double-width signed integer. */
4266 RVVCALL(OPFVV1, vfwcvt_x_f_v_h, WOP_UU_H, H4, H2, float16_to_int32)
4267 RVVCALL(OPFVV1, vfwcvt_x_f_v_w, WOP_UU_W, H8, H4, float32_to_int64)
4268 GEN_VEXT_V_ENV(vfwcvt_x_f_v_h, 2, 4, clearl)
4269 GEN_VEXT_V_ENV(vfwcvt_x_f_v_w, 4, 8, clearq)
4270 
4271 /* vfwcvt.f.xu.v vd, vs2, vm # Convert unsigned integer to double-width float */
4272 RVVCALL(OPFVV1, vfwcvt_f_xu_v_h, WOP_UU_H, H4, H2, uint16_to_float32)
4273 RVVCALL(OPFVV1, vfwcvt_f_xu_v_w, WOP_UU_W, H8, H4, uint32_to_float64)
4274 GEN_VEXT_V_ENV(vfwcvt_f_xu_v_h, 2, 4, clearl)
4275 GEN_VEXT_V_ENV(vfwcvt_f_xu_v_w, 4, 8, clearq)
4276 
4277 /* vfwcvt.f.x.v vd, vs2, vm # Convert integer to double-width float. */
4278 RVVCALL(OPFVV1, vfwcvt_f_x_v_h, WOP_UU_H, H4, H2, int16_to_float32)
4279 RVVCALL(OPFVV1, vfwcvt_f_x_v_w, WOP_UU_W, H8, H4, int32_to_float64)
4280 GEN_VEXT_V_ENV(vfwcvt_f_x_v_h, 2, 4, clearl)
4281 GEN_VEXT_V_ENV(vfwcvt_f_x_v_w, 4, 8, clearq)
4282 
4283 /*
4284  * vfwcvt.f.f.v vd, vs2, vm #
4285  * Convert single-width float to double-width float.
4286  */
4287 static uint32_t vfwcvtffv16(uint16_t a, float_status *s)
4288 {
4289     return float16_to_float32(a, true, s);
4290 }
4291 
4292 RVVCALL(OPFVV1, vfwcvt_f_f_v_h, WOP_UU_H, H4, H2, vfwcvtffv16)
4293 RVVCALL(OPFVV1, vfwcvt_f_f_v_w, WOP_UU_W, H8, H4, float32_to_float64)
4294 GEN_VEXT_V_ENV(vfwcvt_f_f_v_h, 2, 4, clearl)
4295 GEN_VEXT_V_ENV(vfwcvt_f_f_v_w, 4, 8, clearq)
4296 
4297 /* Narrowing Floating-Point/Integer Type-Convert Instructions */
4298 /* (TD, T2, TX2) */
4299 #define NOP_UU_H uint16_t, uint32_t, uint32_t
4300 #define NOP_UU_W uint32_t, uint64_t, uint64_t
4301 /* vfncvt.xu.f.v vd, vs2, vm # Convert float to unsigned integer. */
4302 RVVCALL(OPFVV1, vfncvt_xu_f_v_h, NOP_UU_H, H2, H4, float32_to_uint16)
4303 RVVCALL(OPFVV1, vfncvt_xu_f_v_w, NOP_UU_W, H4, H8, float64_to_uint32)
4304 GEN_VEXT_V_ENV(vfncvt_xu_f_v_h, 2, 2, clearh)
4305 GEN_VEXT_V_ENV(vfncvt_xu_f_v_w, 4, 4, clearl)
4306 
4307 /* vfncvt.x.f.v vd, vs2, vm # Convert double-width float to signed integer. */
4308 RVVCALL(OPFVV1, vfncvt_x_f_v_h, NOP_UU_H, H2, H4, float32_to_int16)
4309 RVVCALL(OPFVV1, vfncvt_x_f_v_w, NOP_UU_W, H4, H8, float64_to_int32)
4310 GEN_VEXT_V_ENV(vfncvt_x_f_v_h, 2, 2, clearh)
4311 GEN_VEXT_V_ENV(vfncvt_x_f_v_w, 4, 4, clearl)
4312 
4313 /* vfncvt.f.xu.v vd, vs2, vm # Convert double-width unsigned integer to float */
4314 RVVCALL(OPFVV1, vfncvt_f_xu_v_h, NOP_UU_H, H2, H4, uint32_to_float16)
4315 RVVCALL(OPFVV1, vfncvt_f_xu_v_w, NOP_UU_W, H4, H8, uint64_to_float32)
4316 GEN_VEXT_V_ENV(vfncvt_f_xu_v_h, 2, 2, clearh)
4317 GEN_VEXT_V_ENV(vfncvt_f_xu_v_w, 4, 4, clearl)
4318 
4319 /* vfncvt.f.x.v vd, vs2, vm # Convert double-width integer to float. */
4320 RVVCALL(OPFVV1, vfncvt_f_x_v_h, NOP_UU_H, H2, H4, int32_to_float16)
4321 RVVCALL(OPFVV1, vfncvt_f_x_v_w, NOP_UU_W, H4, H8, int64_to_float32)
4322 GEN_VEXT_V_ENV(vfncvt_f_x_v_h, 2, 2, clearh)
4323 GEN_VEXT_V_ENV(vfncvt_f_x_v_w, 4, 4, clearl)
4324 
4325 /* vfncvt.f.f.v vd, vs2, vm # Convert double float to single-width float. */
4326 static uint16_t vfncvtffv16(uint32_t a, float_status *s)
4327 {
4328     return float32_to_float16(a, true, s);
4329 }
4330 
4331 RVVCALL(OPFVV1, vfncvt_f_f_v_h, NOP_UU_H, H2, H4, vfncvtffv16)
4332 RVVCALL(OPFVV1, vfncvt_f_f_v_w, NOP_UU_W, H4, H8, float64_to_float32)
4333 GEN_VEXT_V_ENV(vfncvt_f_f_v_h, 2, 2, clearh)
4334 GEN_VEXT_V_ENV(vfncvt_f_f_v_w, 4, 4, clearl)
4335 
4336 /*
4337  *** Vector Reduction Operations
4338  */
4339 /* Vector Single-Width Integer Reduction Instructions */
4340 #define GEN_VEXT_RED(NAME, TD, TS2, HD, HS2, OP, CLEAR_FN)\
4341 void HELPER(NAME)(void *vd, void *v0, void *vs1,          \
4342         void *vs2, CPURISCVState *env, uint32_t desc)     \
4343 {                                                         \
4344     uint32_t mlen = vext_mlen(desc);                      \
4345     uint32_t vm = vext_vm(desc);                          \
4346     uint32_t vl = env->vl;                                \
4347     uint32_t i;                                           \
4348     uint32_t tot = env_archcpu(env)->cfg.vlen / 8;        \
4349     TD s1 =  *((TD *)vs1 + HD(0));                        \
4350                                                           \
4351     for (i = 0; i < vl; i++) {                            \
4352         TS2 s2 = *((TS2 *)vs2 + HS2(i));                  \
4353         if (!vm && !vext_elem_mask(v0, mlen, i)) {        \
4354             continue;                                     \
4355         }                                                 \
4356         s1 = OP(s1, (TD)s2);                              \
4357     }                                                     \
4358     *((TD *)vd + HD(0)) = s1;                             \
4359     CLEAR_FN(vd, 1, sizeof(TD), tot);                     \
4360 }
4361 
4362 /* vd[0] = sum(vs1[0], vs2[*]) */
4363 GEN_VEXT_RED(vredsum_vs_b, int8_t, int8_t, H1, H1, DO_ADD, clearb)
4364 GEN_VEXT_RED(vredsum_vs_h, int16_t, int16_t, H2, H2, DO_ADD, clearh)
4365 GEN_VEXT_RED(vredsum_vs_w, int32_t, int32_t, H4, H4, DO_ADD, clearl)
4366 GEN_VEXT_RED(vredsum_vs_d, int64_t, int64_t, H8, H8, DO_ADD, clearq)
4367 
4368 /* vd[0] = maxu(vs1[0], vs2[*]) */
4369 GEN_VEXT_RED(vredmaxu_vs_b, uint8_t, uint8_t, H1, H1, DO_MAX, clearb)
4370 GEN_VEXT_RED(vredmaxu_vs_h, uint16_t, uint16_t, H2, H2, DO_MAX, clearh)
4371 GEN_VEXT_RED(vredmaxu_vs_w, uint32_t, uint32_t, H4, H4, DO_MAX, clearl)
4372 GEN_VEXT_RED(vredmaxu_vs_d, uint64_t, uint64_t, H8, H8, DO_MAX, clearq)
4373 
4374 /* vd[0] = max(vs1[0], vs2[*]) */
4375 GEN_VEXT_RED(vredmax_vs_b, int8_t, int8_t, H1, H1, DO_MAX, clearb)
4376 GEN_VEXT_RED(vredmax_vs_h, int16_t, int16_t, H2, H2, DO_MAX, clearh)
4377 GEN_VEXT_RED(vredmax_vs_w, int32_t, int32_t, H4, H4, DO_MAX, clearl)
4378 GEN_VEXT_RED(vredmax_vs_d, int64_t, int64_t, H8, H8, DO_MAX, clearq)
4379 
4380 /* vd[0] = minu(vs1[0], vs2[*]) */
4381 GEN_VEXT_RED(vredminu_vs_b, uint8_t, uint8_t, H1, H1, DO_MIN, clearb)
4382 GEN_VEXT_RED(vredminu_vs_h, uint16_t, uint16_t, H2, H2, DO_MIN, clearh)
4383 GEN_VEXT_RED(vredminu_vs_w, uint32_t, uint32_t, H4, H4, DO_MIN, clearl)
4384 GEN_VEXT_RED(vredminu_vs_d, uint64_t, uint64_t, H8, H8, DO_MIN, clearq)
4385 
4386 /* vd[0] = min(vs1[0], vs2[*]) */
4387 GEN_VEXT_RED(vredmin_vs_b, int8_t, int8_t, H1, H1, DO_MIN, clearb)
4388 GEN_VEXT_RED(vredmin_vs_h, int16_t, int16_t, H2, H2, DO_MIN, clearh)
4389 GEN_VEXT_RED(vredmin_vs_w, int32_t, int32_t, H4, H4, DO_MIN, clearl)
4390 GEN_VEXT_RED(vredmin_vs_d, int64_t, int64_t, H8, H8, DO_MIN, clearq)
4391 
4392 /* vd[0] = and(vs1[0], vs2[*]) */
4393 GEN_VEXT_RED(vredand_vs_b, int8_t, int8_t, H1, H1, DO_AND, clearb)
4394 GEN_VEXT_RED(vredand_vs_h, int16_t, int16_t, H2, H2, DO_AND, clearh)
4395 GEN_VEXT_RED(vredand_vs_w, int32_t, int32_t, H4, H4, DO_AND, clearl)
4396 GEN_VEXT_RED(vredand_vs_d, int64_t, int64_t, H8, H8, DO_AND, clearq)
4397 
4398 /* vd[0] = or(vs1[0], vs2[*]) */
4399 GEN_VEXT_RED(vredor_vs_b, int8_t, int8_t, H1, H1, DO_OR, clearb)
4400 GEN_VEXT_RED(vredor_vs_h, int16_t, int16_t, H2, H2, DO_OR, clearh)
4401 GEN_VEXT_RED(vredor_vs_w, int32_t, int32_t, H4, H4, DO_OR, clearl)
4402 GEN_VEXT_RED(vredor_vs_d, int64_t, int64_t, H8, H8, DO_OR, clearq)
4403 
4404 /* vd[0] = xor(vs1[0], vs2[*]) */
4405 GEN_VEXT_RED(vredxor_vs_b, int8_t, int8_t, H1, H1, DO_XOR, clearb)
4406 GEN_VEXT_RED(vredxor_vs_h, int16_t, int16_t, H2, H2, DO_XOR, clearh)
4407 GEN_VEXT_RED(vredxor_vs_w, int32_t, int32_t, H4, H4, DO_XOR, clearl)
4408 GEN_VEXT_RED(vredxor_vs_d, int64_t, int64_t, H8, H8, DO_XOR, clearq)
4409 
4410 /* Vector Widening Integer Reduction Instructions */
4411 /* signed sum reduction into double-width accumulator */
4412 GEN_VEXT_RED(vwredsum_vs_b, int16_t, int8_t, H2, H1, DO_ADD, clearh)
4413 GEN_VEXT_RED(vwredsum_vs_h, int32_t, int16_t, H4, H2, DO_ADD, clearl)
4414 GEN_VEXT_RED(vwredsum_vs_w, int64_t, int32_t, H8, H4, DO_ADD, clearq)
4415 
4416 /* Unsigned sum reduction into double-width accumulator */
4417 GEN_VEXT_RED(vwredsumu_vs_b, uint16_t, uint8_t, H2, H1, DO_ADD, clearh)
4418 GEN_VEXT_RED(vwredsumu_vs_h, uint32_t, uint16_t, H4, H2, DO_ADD, clearl)
4419 GEN_VEXT_RED(vwredsumu_vs_w, uint64_t, uint32_t, H8, H4, DO_ADD, clearq)
4420 
4421 /* Vector Single-Width Floating-Point Reduction Instructions */
4422 #define GEN_VEXT_FRED(NAME, TD, TS2, HD, HS2, OP, CLEAR_FN)\
4423 void HELPER(NAME)(void *vd, void *v0, void *vs1,           \
4424                   void *vs2, CPURISCVState *env,           \
4425                   uint32_t desc)                           \
4426 {                                                          \
4427     uint32_t mlen = vext_mlen(desc);                       \
4428     uint32_t vm = vext_vm(desc);                           \
4429     uint32_t vl = env->vl;                                 \
4430     uint32_t i;                                            \
4431     uint32_t tot = env_archcpu(env)->cfg.vlen / 8;         \
4432     TD s1 =  *((TD *)vs1 + HD(0));                         \
4433                                                            \
4434     for (i = 0; i < vl; i++) {                             \
4435         TS2 s2 = *((TS2 *)vs2 + HS2(i));                   \
4436         if (!vm && !vext_elem_mask(v0, mlen, i)) {         \
4437             continue;                                      \
4438         }                                                  \
4439         s1 = OP(s1, (TD)s2, &env->fp_status);              \
4440     }                                                      \
4441     *((TD *)vd + HD(0)) = s1;                              \
4442     CLEAR_FN(vd, 1, sizeof(TD), tot);                      \
4443 }
4444 
4445 /* Unordered sum */
4446 GEN_VEXT_FRED(vfredsum_vs_h, uint16_t, uint16_t, H2, H2, float16_add, clearh)
4447 GEN_VEXT_FRED(vfredsum_vs_w, uint32_t, uint32_t, H4, H4, float32_add, clearl)
4448 GEN_VEXT_FRED(vfredsum_vs_d, uint64_t, uint64_t, H8, H8, float64_add, clearq)
4449 
4450 /* Maximum value */
4451 GEN_VEXT_FRED(vfredmax_vs_h, uint16_t, uint16_t, H2, H2, float16_maxnum, clearh)
4452 GEN_VEXT_FRED(vfredmax_vs_w, uint32_t, uint32_t, H4, H4, float32_maxnum, clearl)
4453 GEN_VEXT_FRED(vfredmax_vs_d, uint64_t, uint64_t, H8, H8, float64_maxnum, clearq)
4454 
4455 /* Minimum value */
4456 GEN_VEXT_FRED(vfredmin_vs_h, uint16_t, uint16_t, H2, H2, float16_minnum, clearh)
4457 GEN_VEXT_FRED(vfredmin_vs_w, uint32_t, uint32_t, H4, H4, float32_minnum, clearl)
4458 GEN_VEXT_FRED(vfredmin_vs_d, uint64_t, uint64_t, H8, H8, float64_minnum, clearq)
4459 
4460 /* Vector Widening Floating-Point Reduction Instructions */
4461 /* Unordered reduce 2*SEW = 2*SEW + sum(promote(SEW)) */
4462 void HELPER(vfwredsum_vs_h)(void *vd, void *v0, void *vs1,
4463                             void *vs2, CPURISCVState *env, uint32_t desc)
4464 {
4465     uint32_t mlen = vext_mlen(desc);
4466     uint32_t vm = vext_vm(desc);
4467     uint32_t vl = env->vl;
4468     uint32_t i;
4469     uint32_t tot = env_archcpu(env)->cfg.vlen / 8;
4470     uint32_t s1 =  *((uint32_t *)vs1 + H4(0));
4471 
4472     for (i = 0; i < vl; i++) {
4473         uint16_t s2 = *((uint16_t *)vs2 + H2(i));
4474         if (!vm && !vext_elem_mask(v0, mlen, i)) {
4475             continue;
4476         }
4477         s1 = float32_add(s1, float16_to_float32(s2, true, &env->fp_status),
4478                          &env->fp_status);
4479     }
4480     *((uint32_t *)vd + H4(0)) = s1;
4481     clearl(vd, 1, sizeof(uint32_t), tot);
4482 }
4483 
4484 void HELPER(vfwredsum_vs_w)(void *vd, void *v0, void *vs1,
4485                             void *vs2, CPURISCVState *env, uint32_t desc)
4486 {
4487     uint32_t mlen = vext_mlen(desc);
4488     uint32_t vm = vext_vm(desc);
4489     uint32_t vl = env->vl;
4490     uint32_t i;
4491     uint32_t tot = env_archcpu(env)->cfg.vlen / 8;
4492     uint64_t s1 =  *((uint64_t *)vs1);
4493 
4494     for (i = 0; i < vl; i++) {
4495         uint32_t s2 = *((uint32_t *)vs2 + H4(i));
4496         if (!vm && !vext_elem_mask(v0, mlen, i)) {
4497             continue;
4498         }
4499         s1 = float64_add(s1, float32_to_float64(s2, &env->fp_status),
4500                          &env->fp_status);
4501     }
4502     *((uint64_t *)vd) = s1;
4503     clearq(vd, 1, sizeof(uint64_t), tot);
4504 }
4505 
4506 /*
4507  *** Vector Mask Operations
4508  */
4509 /* Vector Mask-Register Logical Instructions */
4510 #define GEN_VEXT_MASK_VV(NAME, OP)                        \
4511 void HELPER(NAME)(void *vd, void *v0, void *vs1,          \
4512                   void *vs2, CPURISCVState *env,          \
4513                   uint32_t desc)                          \
4514 {                                                         \
4515     uint32_t mlen = vext_mlen(desc);                      \
4516     uint32_t vlmax = env_archcpu(env)->cfg.vlen / mlen;   \
4517     uint32_t vl = env->vl;                                \
4518     uint32_t i;                                           \
4519     int a, b;                                             \
4520                                                           \
4521     for (i = 0; i < vl; i++) {                            \
4522         a = vext_elem_mask(vs1, mlen, i);                 \
4523         b = vext_elem_mask(vs2, mlen, i);                 \
4524         vext_set_elem_mask(vd, mlen, i, OP(b, a));        \
4525     }                                                     \
4526     for (; i < vlmax; i++) {                              \
4527         vext_set_elem_mask(vd, mlen, i, 0);               \
4528     }                                                     \
4529 }
4530 
4531 #define DO_NAND(N, M)  (!(N & M))
4532 #define DO_ANDNOT(N, M)  (N & !M)
4533 #define DO_NOR(N, M)  (!(N | M))
4534 #define DO_ORNOT(N, M)  (N | !M)
4535 #define DO_XNOR(N, M)  (!(N ^ M))
4536 
4537 GEN_VEXT_MASK_VV(vmand_mm, DO_AND)
4538 GEN_VEXT_MASK_VV(vmnand_mm, DO_NAND)
4539 GEN_VEXT_MASK_VV(vmandnot_mm, DO_ANDNOT)
4540 GEN_VEXT_MASK_VV(vmxor_mm, DO_XOR)
4541 GEN_VEXT_MASK_VV(vmor_mm, DO_OR)
4542 GEN_VEXT_MASK_VV(vmnor_mm, DO_NOR)
4543 GEN_VEXT_MASK_VV(vmornot_mm, DO_ORNOT)
4544 GEN_VEXT_MASK_VV(vmxnor_mm, DO_XNOR)
4545 
4546 /* Vector mask population count vmpopc */
4547 target_ulong HELPER(vmpopc_m)(void *v0, void *vs2, CPURISCVState *env,
4548                               uint32_t desc)
4549 {
4550     target_ulong cnt = 0;
4551     uint32_t mlen = vext_mlen(desc);
4552     uint32_t vm = vext_vm(desc);
4553     uint32_t vl = env->vl;
4554     int i;
4555 
4556     for (i = 0; i < vl; i++) {
4557         if (vm || vext_elem_mask(v0, mlen, i)) {
4558             if (vext_elem_mask(vs2, mlen, i)) {
4559                 cnt++;
4560             }
4561         }
4562     }
4563     return cnt;
4564 }
4565 
4566 /* vmfirst find-first-set mask bit*/
4567 target_ulong HELPER(vmfirst_m)(void *v0, void *vs2, CPURISCVState *env,
4568                                uint32_t desc)
4569 {
4570     uint32_t mlen = vext_mlen(desc);
4571     uint32_t vm = vext_vm(desc);
4572     uint32_t vl = env->vl;
4573     int i;
4574 
4575     for (i = 0; i < vl; i++) {
4576         if (vm || vext_elem_mask(v0, mlen, i)) {
4577             if (vext_elem_mask(vs2, mlen, i)) {
4578                 return i;
4579             }
4580         }
4581     }
4582     return -1LL;
4583 }
4584 
4585 enum set_mask_type {
4586     ONLY_FIRST = 1,
4587     INCLUDE_FIRST,
4588     BEFORE_FIRST,
4589 };
4590 
4591 static void vmsetm(void *vd, void *v0, void *vs2, CPURISCVState *env,
4592                    uint32_t desc, enum set_mask_type type)
4593 {
4594     uint32_t mlen = vext_mlen(desc);
4595     uint32_t vlmax = env_archcpu(env)->cfg.vlen / mlen;
4596     uint32_t vm = vext_vm(desc);
4597     uint32_t vl = env->vl;
4598     int i;
4599     bool first_mask_bit = false;
4600 
4601     for (i = 0; i < vl; i++) {
4602         if (!vm && !vext_elem_mask(v0, mlen, i)) {
4603             continue;
4604         }
4605         /* write a zero to all following active elements */
4606         if (first_mask_bit) {
4607             vext_set_elem_mask(vd, mlen, i, 0);
4608             continue;
4609         }
4610         if (vext_elem_mask(vs2, mlen, i)) {
4611             first_mask_bit = true;
4612             if (type == BEFORE_FIRST) {
4613                 vext_set_elem_mask(vd, mlen, i, 0);
4614             } else {
4615                 vext_set_elem_mask(vd, mlen, i, 1);
4616             }
4617         } else {
4618             if (type == ONLY_FIRST) {
4619                 vext_set_elem_mask(vd, mlen, i, 0);
4620             } else {
4621                 vext_set_elem_mask(vd, mlen, i, 1);
4622             }
4623         }
4624     }
4625     for (; i < vlmax; i++) {
4626         vext_set_elem_mask(vd, mlen, i, 0);
4627     }
4628 }
4629 
4630 void HELPER(vmsbf_m)(void *vd, void *v0, void *vs2, CPURISCVState *env,
4631                      uint32_t desc)
4632 {
4633     vmsetm(vd, v0, vs2, env, desc, BEFORE_FIRST);
4634 }
4635 
4636 void HELPER(vmsif_m)(void *vd, void *v0, void *vs2, CPURISCVState *env,
4637                      uint32_t desc)
4638 {
4639     vmsetm(vd, v0, vs2, env, desc, INCLUDE_FIRST);
4640 }
4641 
4642 void HELPER(vmsof_m)(void *vd, void *v0, void *vs2, CPURISCVState *env,
4643                      uint32_t desc)
4644 {
4645     vmsetm(vd, v0, vs2, env, desc, ONLY_FIRST);
4646 }
4647 
4648 /* Vector Iota Instruction */
4649 #define GEN_VEXT_VIOTA_M(NAME, ETYPE, H, CLEAR_FN)                        \
4650 void HELPER(NAME)(void *vd, void *v0, void *vs2, CPURISCVState *env,      \
4651                   uint32_t desc)                                          \
4652 {                                                                         \
4653     uint32_t mlen = vext_mlen(desc);                                      \
4654     uint32_t vlmax = env_archcpu(env)->cfg.vlen / mlen;                   \
4655     uint32_t vm = vext_vm(desc);                                          \
4656     uint32_t vl = env->vl;                                                \
4657     uint32_t sum = 0;                                                     \
4658     int i;                                                                \
4659                                                                           \
4660     for (i = 0; i < vl; i++) {                                            \
4661         if (!vm && !vext_elem_mask(v0, mlen, i)) {                        \
4662             continue;                                                     \
4663         }                                                                 \
4664         *((ETYPE *)vd + H(i)) = sum;                                      \
4665         if (vext_elem_mask(vs2, mlen, i)) {                               \
4666             sum++;                                                        \
4667         }                                                                 \
4668     }                                                                     \
4669     CLEAR_FN(vd, vl, vl * sizeof(ETYPE), vlmax * sizeof(ETYPE));          \
4670 }
4671 
4672 GEN_VEXT_VIOTA_M(viota_m_b, uint8_t, H1, clearb)
4673 GEN_VEXT_VIOTA_M(viota_m_h, uint16_t, H2, clearh)
4674 GEN_VEXT_VIOTA_M(viota_m_w, uint32_t, H4, clearl)
4675 GEN_VEXT_VIOTA_M(viota_m_d, uint64_t, H8, clearq)
4676 
4677 /* Vector Element Index Instruction */
4678 #define GEN_VEXT_VID_V(NAME, ETYPE, H, CLEAR_FN)                          \
4679 void HELPER(NAME)(void *vd, void *v0, CPURISCVState *env, uint32_t desc)  \
4680 {                                                                         \
4681     uint32_t mlen = vext_mlen(desc);                                      \
4682     uint32_t vlmax = env_archcpu(env)->cfg.vlen / mlen;                   \
4683     uint32_t vm = vext_vm(desc);                                          \
4684     uint32_t vl = env->vl;                                                \
4685     int i;                                                                \
4686                                                                           \
4687     for (i = 0; i < vl; i++) {                                            \
4688         if (!vm && !vext_elem_mask(v0, mlen, i)) {                        \
4689             continue;                                                     \
4690         }                                                                 \
4691         *((ETYPE *)vd + H(i)) = i;                                        \
4692     }                                                                     \
4693     CLEAR_FN(vd, vl, vl * sizeof(ETYPE), vlmax * sizeof(ETYPE));          \
4694 }
4695 
4696 GEN_VEXT_VID_V(vid_v_b, uint8_t, H1, clearb)
4697 GEN_VEXT_VID_V(vid_v_h, uint16_t, H2, clearh)
4698 GEN_VEXT_VID_V(vid_v_w, uint32_t, H4, clearl)
4699 GEN_VEXT_VID_V(vid_v_d, uint64_t, H8, clearq)
4700 
4701 /*
4702  *** Vector Permutation Instructions
4703  */
4704 
4705 /* Vector Slide Instructions */
4706 #define GEN_VEXT_VSLIDEUP_VX(NAME, ETYPE, H, CLEAR_FN)                    \
4707 void HELPER(NAME)(void *vd, void *v0, target_ulong s1, void *vs2,         \
4708                   CPURISCVState *env, uint32_t desc)                      \
4709 {                                                                         \
4710     uint32_t mlen = vext_mlen(desc);                                      \
4711     uint32_t vlmax = env_archcpu(env)->cfg.vlen / mlen;                   \
4712     uint32_t vm = vext_vm(desc);                                          \
4713     uint32_t vl = env->vl;                                                \
4714     target_ulong offset = s1, i;                                          \
4715                                                                           \
4716     for (i = offset; i < vl; i++) {                                       \
4717         if (!vm && !vext_elem_mask(v0, mlen, i)) {                        \
4718             continue;                                                     \
4719         }                                                                 \
4720         *((ETYPE *)vd + H(i)) = *((ETYPE *)vs2 + H(i - offset));          \
4721     }                                                                     \
4722     CLEAR_FN(vd, vl, vl * sizeof(ETYPE), vlmax * sizeof(ETYPE));          \
4723 }
4724 
4725 /* vslideup.vx vd, vs2, rs1, vm # vd[i+rs1] = vs2[i] */
4726 GEN_VEXT_VSLIDEUP_VX(vslideup_vx_b, uint8_t, H1, clearb)
4727 GEN_VEXT_VSLIDEUP_VX(vslideup_vx_h, uint16_t, H2, clearh)
4728 GEN_VEXT_VSLIDEUP_VX(vslideup_vx_w, uint32_t, H4, clearl)
4729 GEN_VEXT_VSLIDEUP_VX(vslideup_vx_d, uint64_t, H8, clearq)
4730 
4731 #define GEN_VEXT_VSLIDEDOWN_VX(NAME, ETYPE, H, CLEAR_FN)                  \
4732 void HELPER(NAME)(void *vd, void *v0, target_ulong s1, void *vs2,         \
4733                   CPURISCVState *env, uint32_t desc)                      \
4734 {                                                                         \
4735     uint32_t mlen = vext_mlen(desc);                                      \
4736     uint32_t vlmax = env_archcpu(env)->cfg.vlen / mlen;                   \
4737     uint32_t vm = vext_vm(desc);                                          \
4738     uint32_t vl = env->vl;                                                \
4739     target_ulong offset = s1, i;                                          \
4740                                                                           \
4741     for (i = 0; i < vl; ++i) {                                            \
4742         target_ulong j = i + offset;                                      \
4743         if (!vm && !vext_elem_mask(v0, mlen, i)) {                        \
4744             continue;                                                     \
4745         }                                                                 \
4746         *((ETYPE *)vd + H(i)) = j >= vlmax ? 0 : *((ETYPE *)vs2 + H(j));  \
4747     }                                                                     \
4748     CLEAR_FN(vd, vl, vl * sizeof(ETYPE), vlmax * sizeof(ETYPE));          \
4749 }
4750 
4751 /* vslidedown.vx vd, vs2, rs1, vm # vd[i] = vs2[i+rs1] */
4752 GEN_VEXT_VSLIDEDOWN_VX(vslidedown_vx_b, uint8_t, H1, clearb)
4753 GEN_VEXT_VSLIDEDOWN_VX(vslidedown_vx_h, uint16_t, H2, clearh)
4754 GEN_VEXT_VSLIDEDOWN_VX(vslidedown_vx_w, uint32_t, H4, clearl)
4755 GEN_VEXT_VSLIDEDOWN_VX(vslidedown_vx_d, uint64_t, H8, clearq)
4756 
4757 #define GEN_VEXT_VSLIDE1UP_VX(NAME, ETYPE, H, CLEAR_FN)                   \
4758 void HELPER(NAME)(void *vd, void *v0, target_ulong s1, void *vs2,         \
4759                   CPURISCVState *env, uint32_t desc)                      \
4760 {                                                                         \
4761     uint32_t mlen = vext_mlen(desc);                                      \
4762     uint32_t vlmax = env_archcpu(env)->cfg.vlen / mlen;                   \
4763     uint32_t vm = vext_vm(desc);                                          \
4764     uint32_t vl = env->vl;                                                \
4765     uint32_t i;                                                           \
4766                                                                           \
4767     for (i = 0; i < vl; i++) {                                            \
4768         if (!vm && !vext_elem_mask(v0, mlen, i)) {                        \
4769             continue;                                                     \
4770         }                                                                 \
4771         if (i == 0) {                                                     \
4772             *((ETYPE *)vd + H(i)) = s1;                                   \
4773         } else {                                                          \
4774             *((ETYPE *)vd + H(i)) = *((ETYPE *)vs2 + H(i - 1));           \
4775         }                                                                 \
4776     }                                                                     \
4777     CLEAR_FN(vd, vl, vl * sizeof(ETYPE), vlmax * sizeof(ETYPE));          \
4778 }
4779 
4780 /* vslide1up.vx vd, vs2, rs1, vm # vd[0]=x[rs1], vd[i+1] = vs2[i] */
4781 GEN_VEXT_VSLIDE1UP_VX(vslide1up_vx_b, uint8_t, H1, clearb)
4782 GEN_VEXT_VSLIDE1UP_VX(vslide1up_vx_h, uint16_t, H2, clearh)
4783 GEN_VEXT_VSLIDE1UP_VX(vslide1up_vx_w, uint32_t, H4, clearl)
4784 GEN_VEXT_VSLIDE1UP_VX(vslide1up_vx_d, uint64_t, H8, clearq)
4785 
4786 #define GEN_VEXT_VSLIDE1DOWN_VX(NAME, ETYPE, H, CLEAR_FN)                 \
4787 void HELPER(NAME)(void *vd, void *v0, target_ulong s1, void *vs2,         \
4788                   CPURISCVState *env, uint32_t desc)                      \
4789 {                                                                         \
4790     uint32_t mlen = vext_mlen(desc);                                      \
4791     uint32_t vlmax = env_archcpu(env)->cfg.vlen / mlen;                   \
4792     uint32_t vm = vext_vm(desc);                                          \
4793     uint32_t vl = env->vl;                                                \
4794     uint32_t i;                                                           \
4795                                                                           \
4796     for (i = 0; i < vl; i++) {                                            \
4797         if (!vm && !vext_elem_mask(v0, mlen, i)) {                        \
4798             continue;                                                     \
4799         }                                                                 \
4800         if (i == vl - 1) {                                                \
4801             *((ETYPE *)vd + H(i)) = s1;                                   \
4802         } else {                                                          \
4803             *((ETYPE *)vd + H(i)) = *((ETYPE *)vs2 + H(i + 1));           \
4804         }                                                                 \
4805     }                                                                     \
4806     CLEAR_FN(vd, vl, vl * sizeof(ETYPE), vlmax * sizeof(ETYPE));          \
4807 }
4808 
4809 /* vslide1down.vx vd, vs2, rs1, vm # vd[i] = vs2[i+1], vd[vl-1]=x[rs1] */
4810 GEN_VEXT_VSLIDE1DOWN_VX(vslide1down_vx_b, uint8_t, H1, clearb)
4811 GEN_VEXT_VSLIDE1DOWN_VX(vslide1down_vx_h, uint16_t, H2, clearh)
4812 GEN_VEXT_VSLIDE1DOWN_VX(vslide1down_vx_w, uint32_t, H4, clearl)
4813 GEN_VEXT_VSLIDE1DOWN_VX(vslide1down_vx_d, uint64_t, H8, clearq)
4814 
4815 /* Vector Register Gather Instruction */
4816 #define GEN_VEXT_VRGATHER_VV(NAME, ETYPE, H, CLEAR_FN)                    \
4817 void HELPER(NAME)(void *vd, void *v0, void *vs1, void *vs2,               \
4818                   CPURISCVState *env, uint32_t desc)                      \
4819 {                                                                         \
4820     uint32_t mlen = vext_mlen(desc);                                      \
4821     uint32_t vlmax = env_archcpu(env)->cfg.vlen / mlen;                   \
4822     uint32_t vm = vext_vm(desc);                                          \
4823     uint32_t vl = env->vl;                                                \
4824     uint32_t index, i;                                                    \
4825                                                                           \
4826     for (i = 0; i < vl; i++) {                                            \
4827         if (!vm && !vext_elem_mask(v0, mlen, i)) {                        \
4828             continue;                                                     \
4829         }                                                                 \
4830         index = *((ETYPE *)vs1 + H(i));                                   \
4831         if (index >= vlmax) {                                             \
4832             *((ETYPE *)vd + H(i)) = 0;                                    \
4833         } else {                                                          \
4834             *((ETYPE *)vd + H(i)) = *((ETYPE *)vs2 + H(index));           \
4835         }                                                                 \
4836     }                                                                     \
4837     CLEAR_FN(vd, vl, vl * sizeof(ETYPE), vlmax * sizeof(ETYPE));          \
4838 }
4839 
4840 /* vd[i] = (vs1[i] >= VLMAX) ? 0 : vs2[vs1[i]]; */
4841 GEN_VEXT_VRGATHER_VV(vrgather_vv_b, uint8_t, H1, clearb)
4842 GEN_VEXT_VRGATHER_VV(vrgather_vv_h, uint16_t, H2, clearh)
4843 GEN_VEXT_VRGATHER_VV(vrgather_vv_w, uint32_t, H4, clearl)
4844 GEN_VEXT_VRGATHER_VV(vrgather_vv_d, uint64_t, H8, clearq)
4845 
4846 #define GEN_VEXT_VRGATHER_VX(NAME, ETYPE, H, CLEAR_FN)                    \
4847 void HELPER(NAME)(void *vd, void *v0, target_ulong s1, void *vs2,         \
4848                   CPURISCVState *env, uint32_t desc)                      \
4849 {                                                                         \
4850     uint32_t mlen = vext_mlen(desc);                                      \
4851     uint32_t vlmax = env_archcpu(env)->cfg.vlen / mlen;                   \
4852     uint32_t vm = vext_vm(desc);                                          \
4853     uint32_t vl = env->vl;                                                \
4854     uint32_t index = s1, i;                                               \
4855                                                                           \
4856     for (i = 0; i < vl; i++) {                                            \
4857         if (!vm && !vext_elem_mask(v0, mlen, i)) {                        \
4858             continue;                                                     \
4859         }                                                                 \
4860         if (index >= vlmax) {                                             \
4861             *((ETYPE *)vd + H(i)) = 0;                                    \
4862         } else {                                                          \
4863             *((ETYPE *)vd + H(i)) = *((ETYPE *)vs2 + H(index));           \
4864         }                                                                 \
4865     }                                                                     \
4866     CLEAR_FN(vd, vl, vl * sizeof(ETYPE), vlmax * sizeof(ETYPE));          \
4867 }
4868 
4869 /* vd[i] = (x[rs1] >= VLMAX) ? 0 : vs2[rs1] */
4870 GEN_VEXT_VRGATHER_VX(vrgather_vx_b, uint8_t, H1, clearb)
4871 GEN_VEXT_VRGATHER_VX(vrgather_vx_h, uint16_t, H2, clearh)
4872 GEN_VEXT_VRGATHER_VX(vrgather_vx_w, uint32_t, H4, clearl)
4873 GEN_VEXT_VRGATHER_VX(vrgather_vx_d, uint64_t, H8, clearq)
4874 
4875 /* Vector Compress Instruction */
4876 #define GEN_VEXT_VCOMPRESS_VM(NAME, ETYPE, H, CLEAR_FN)                   \
4877 void HELPER(NAME)(void *vd, void *v0, void *vs1, void *vs2,               \
4878                   CPURISCVState *env, uint32_t desc)                      \
4879 {                                                                         \
4880     uint32_t mlen = vext_mlen(desc);                                      \
4881     uint32_t vlmax = env_archcpu(env)->cfg.vlen / mlen;                   \
4882     uint32_t vl = env->vl;                                                \
4883     uint32_t num = 0, i;                                                  \
4884                                                                           \
4885     for (i = 0; i < vl; i++) {                                            \
4886         if (!vext_elem_mask(vs1, mlen, i)) {                              \
4887             continue;                                                     \
4888         }                                                                 \
4889         *((ETYPE *)vd + H(num)) = *((ETYPE *)vs2 + H(i));                 \
4890         num++;                                                            \
4891     }                                                                     \
4892     CLEAR_FN(vd, num, num * sizeof(ETYPE), vlmax * sizeof(ETYPE));        \
4893 }
4894 
4895 /* Compress into vd elements of vs2 where vs1 is enabled */
4896 GEN_VEXT_VCOMPRESS_VM(vcompress_vm_b, uint8_t, H1, clearb)
4897 GEN_VEXT_VCOMPRESS_VM(vcompress_vm_h, uint16_t, H2, clearh)
4898 GEN_VEXT_VCOMPRESS_VM(vcompress_vm_w, uint32_t, H4, clearl)
4899 GEN_VEXT_VCOMPRESS_VM(vcompress_vm_d, uint64_t, H8, clearq)
4900