| 1 |
/* { dg-do compile } */ |
| 2 |
/* { dg-options "-march=rv32gc_zve64d -mabi=ilp32 -mpreferred-stack-boundary=3 -O3 -fno-schedule-insns -fno-schedule-insns2" } */ |
| 3 |
/* { dg-final { check-function-bodies "**" "" } } */ |
| 4 |
|
| 5 |
#include "riscv_vector.h" |
| 6 |
#include "macro.h" |
| 7 |
|
| 8 |
/* |
| 9 |
** spill_3: |
| 10 |
** csrr\tt0,vlenb |
| 11 |
** sub\tsp,sp,t0 |
| 12 |
** vsetvli\ta5,zero,e32,mf2,ta,ma |
| 13 |
** vle32.v\tv[0-9]+,0\(a0\) |
| 14 |
** csrr\t[a-x0-9]+,vlenb |
| 15 |
** srli\t[a-x0-9]+,[a-x0-9]+,1 |
| 16 |
** add\t[a-x0-9]+,[a-x0-9]+,sp |
| 17 |
** vse32.v\tv[0-9]+,0\([a-x0-9]+\) |
| 18 |
** ... |
| 19 |
** csrr\t[a-x0-9]+,vlenb |
| 20 |
** srli\t[a-x0-9]+,[a-x0-9]+,1 |
| 21 |
** add\t[a-x0-9]+,[a-x0-9]+,sp |
| 22 |
** vle32.v\tv[0-9]+,0\([a-x0-9]+\) |
| 23 |
** vse32.v\tv[0-9]+,0\([a-x0-9]+\) |
| 24 |
** csrr\tt0,vlenb |
| 25 |
** add\tsp,sp,t0 |
| 26 |
** ... |
| 27 |
** jr\tra |
| 28 |
*/ |
| 29 |
void |
| 30 |
spill_3 (int32_t *in, int32_t *out) |
| 31 |
{ |
| 32 |
vint32mf2_t v1 = *(vint32mf2_t*)in; |
| 33 |
exhaust_vector_regs (); |
| 34 |
*(vint32mf2_t*)out = v1; |
| 35 |
} |
| 36 |
|
| 37 |
/* |
| 38 |
** spill_4: |
| 39 |
** csrr\tt0,vlenb |
| 40 |
** sub\tsp,sp,t0 |
| 41 |
** ... |
| 42 |
** vs1r.v\tv[0-9]+,0\(sp\) |
| 43 |
** ... |
| 44 |
** vl1re32.v\tv2,0\(sp\) |
| 45 |
** vs1r.v\tv2,0\([a-x0-9]+\) |
| 46 |
** ... |
| 47 |
** jr\tra |
| 48 |
*/ |
| 49 |
void |
| 50 |
spill_4 (int32_t *in, int32_t *out) |
| 51 |
{ |
| 52 |
register vint32m1_t v1 asm("v1") = *(vint32m1_t*)in; |
| 53 |
asm volatile ("# %0"::"vr"(v1)); |
| 54 |
exhaust_vector_regs (); |
| 55 |
register vint32m1_t v2 asm("v2") = v1; |
| 56 |
*(vint32m1_t*)out = v2; |
| 57 |
asm volatile ("# %0"::"vr"(v2)); |
| 58 |
} |
| 59 |
|
| 60 |
/* |
| 61 |
** spill_5: |
| 62 |
** csrr\tt0,vlenb |
| 63 |
** slli\tt1,t0,1 |
| 64 |
** sub\tsp,sp,t1 |
| 65 |
** ... |
| 66 |
** vs2r.v\tv[0-9]+,0\(sp\) |
| 67 |
** ... |
| 68 |
** vl2re32.v\tv4,0\(sp\) |
| 69 |
** vs2r.v\tv4,0\([a-x0-9]+\) |
| 70 |
** ... |
| 71 |
** jr\tra |
| 72 |
*/ |
| 73 |
void |
| 74 |
spill_5 (int32_t *in, int32_t *out) |
| 75 |
{ |
| 76 |
register vint32m2_t v2 asm("v2") = *(vint32m2_t*)in; |
| 77 |
asm volatile ("# %0"::"vr"(v2)); |
| 78 |
exhaust_vector_regs (); |
| 79 |
register vint32m2_t v4 asm("v4") = v2; |
| 80 |
*(vint32m2_t*)out = v4; |
| 81 |
asm volatile ("# %0"::"vr"(v4)); |
| 82 |
} |
| 83 |
|
| 84 |
/* |
| 85 |
** spill_6: |
| 86 |
** csrr\tt0,vlenb |
| 87 |
** slli\tt1,t0,2 |
| 88 |
** sub\tsp,sp,t1 |
| 89 |
** ... |
| 90 |
** vs4r.v\tv[0-9]+,0\(sp\) |
| 91 |
** ... |
| 92 |
** vl4re32.v\tv8,0\(sp\) |
| 93 |
** vs4r.v\tv8,0\([a-x0-9]+\) |
| 94 |
** ... |
| 95 |
** jr\tra |
| 96 |
*/ |
| 97 |
void |
| 98 |
spill_6 (int32_t *in, int32_t *out) |
| 99 |
{ |
| 100 |
register vint32m4_t v4 asm("v4") = *(vint32m4_t*)in; |
| 101 |
asm volatile ("# %0"::"vr"(v4)); |
| 102 |
exhaust_vector_regs (); |
| 103 |
register vint32m4_t v8 asm("v8") = v4; |
| 104 |
*(vint32m4_t*)out = v8; |
| 105 |
asm volatile ("# %0"::"vr"(v8)); |
| 106 |
} |
| 107 |
|
| 108 |
/* |
| 109 |
** spill_7: |
| 110 |
** csrr\tt0,vlenb |
| 111 |
** slli\tt1,t0,3 |
| 112 |
** sub\tsp,sp,t1 |
| 113 |
** ... |
| 114 |
** vs8r.v\tv[0-9]+,0\(sp\) |
| 115 |
** ... |
| 116 |
** vl8re32.v\tv16,0\(sp\) |
| 117 |
** vs8r.v\tv16,0\([a-x0-9]+\) |
| 118 |
** ... |
| 119 |
** jr\tra |
| 120 |
*/ |
| 121 |
void |
| 122 |
spill_7 (int32_t *in, int32_t *out) |
| 123 |
{ |
| 124 |
register vint32m8_t v8 asm("v8") = *(vint32m8_t*)in; |
| 125 |
asm volatile ("# %0"::"vr"(v8)); |
| 126 |
exhaust_vector_regs (); |
| 127 |
register vint32m8_t v16 asm("v16") = v8; |
| 128 |
*(vint32m8_t*)out = v16; |
| 129 |
asm volatile ("# %0"::"vr"(v16)); |
| 130 |
} |
| 131 |
|
| 132 |
/* |
| 133 |
** spill_10: |
| 134 |
** csrr\tt0,vlenb |
| 135 |
** sub\tsp,sp,t0 |
| 136 |
** vsetvli\ta5,zero,e32,mf2,ta,ma |
| 137 |
** vle32.v\tv[0-9]+,0\(a0\) |
| 138 |
** csrr\t[a-x0-9]+,vlenb |
| 139 |
** srli\t[a-x0-9]+,[a-x0-9]+,1 |
| 140 |
** add\t[a-x0-9]+,[a-x0-9]+,sp |
| 141 |
** vse32.v\tv[0-9]+,0\([a-x0-9]+\) |
| 142 |
** ... |
| 143 |
** csrr\t[a-x0-9]+,vlenb |
| 144 |
** srli\t[a-x0-9]+,[a-x0-9]+,1 |
| 145 |
** add\t[a-x0-9]+,[a-x0-9]+,sp |
| 146 |
** vle32.v\tv[0-9]+,0\([a-x0-9]+\) |
| 147 |
** vse32.v\tv[0-9]+,0\([a-x0-9]+\) |
| 148 |
** csrr\tt0,vlenb |
| 149 |
** add\tsp,sp,t0 |
| 150 |
** ... |
| 151 |
** jr\tra |
| 152 |
*/ |
| 153 |
void |
| 154 |
spill_10 (uint32_t *in, uint32_t *out) |
| 155 |
{ |
| 156 |
vuint32mf2_t v1 = *(vuint32mf2_t*)in; |
| 157 |
exhaust_vector_regs (); |
| 158 |
*(vuint32mf2_t*)out = v1; |
| 159 |
} |
| 160 |
|
| 161 |
/* |
| 162 |
** spill_11: |
| 163 |
** csrr\tt0,vlenb |
| 164 |
** sub\tsp,sp,t0 |
| 165 |
** ... |
| 166 |
** vs1r.v\tv[0-9]+,0\(sp\) |
| 167 |
** ... |
| 168 |
** vl1re32.v\tv2,0\(sp\) |
| 169 |
** vs1r.v\tv2,0\([a-x0-9]+\) |
| 170 |
** ... |
| 171 |
** jr\tra |
| 172 |
*/ |
| 173 |
void |
| 174 |
spill_11 (uint32_t *in, uint32_t *out) |
| 175 |
{ |
| 176 |
register vuint32m1_t v1 asm("v1") = *(vuint32m1_t*)in; |
| 177 |
asm volatile ("# %0"::"vr"(v1)); |
| 178 |
exhaust_vector_regs (); |
| 179 |
register vuint32m1_t v2 asm("v2") = v1; |
| 180 |
*(vuint32m1_t*)out = v2; |
| 181 |
asm volatile ("# %0"::"vr"(v2)); |
| 182 |
} |
| 183 |
|
| 184 |
/* |
| 185 |
** spill_12: |
| 186 |
** csrr\tt0,vlenb |
| 187 |
** slli\tt1,t0,1 |
| 188 |
** sub\tsp,sp,t1 |
| 189 |
** ... |
| 190 |
** vs2r.v\tv[0-9]+,0\(sp\) |
| 191 |
** ... |
| 192 |
** vl2re32.v\tv4,0\(sp\) |
| 193 |
** vs2r.v\tv4,0\([a-x0-9]+\) |
| 194 |
** ... |
| 195 |
** jr\tra |
| 196 |
*/ |
| 197 |
void |
| 198 |
spill_12 (uint32_t *in, uint32_t *out) |
| 199 |
{ |
| 200 |
register vuint32m2_t v2 asm("v2") = *(vuint32m2_t*)in; |
| 201 |
asm volatile ("# %0"::"vr"(v2)); |
| 202 |
exhaust_vector_regs (); |
| 203 |
register vuint32m2_t v4 asm("v4") = v2; |
| 204 |
*(vuint32m2_t*)out = v4; |
| 205 |
asm volatile ("# %0"::"vr"(v4)); |
| 206 |
} |
| 207 |
|
| 208 |
/* |
| 209 |
** spill_13: |
| 210 |
** csrr\tt0,vlenb |
| 211 |
** slli\tt1,t0,2 |
| 212 |
** sub\tsp,sp,t1 |
| 213 |
** ... |
| 214 |
** vs4r.v\tv[0-9]+,0\(sp\) |
| 215 |
** ... |
| 216 |
** vl4re32.v\tv8,0\(sp\) |
| 217 |
** vs4r.v\tv8,0\([a-x0-9]+\) |
| 218 |
** ... |
| 219 |
** jr\tra |
| 220 |
*/ |
| 221 |
void |
| 222 |
spill_13 (uint32_t *in, uint32_t *out) |
| 223 |
{ |
| 224 |
register vuint32m4_t v4 asm("v4") = *(vuint32m4_t*)in; |
| 225 |
asm volatile ("# %0"::"vr"(v4)); |
| 226 |
exhaust_vector_regs (); |
| 227 |
register vuint32m4_t v8 asm("v8") = v4; |
| 228 |
*(vuint32m4_t*)out = v8; |
| 229 |
asm volatile ("# %0"::"vr"(v8)); |
| 230 |
} |
| 231 |
|
| 232 |
/* |
| 233 |
** spill_14: |
| 234 |
** csrr\tt0,vlenb |
| 235 |
** slli\tt1,t0,3 |
| 236 |
** sub\tsp,sp,t1 |
| 237 |
** ... |
| 238 |
** vs8r.v\tv[0-9]+,0\(sp\) |
| 239 |
** ... |
| 240 |
** vl8re32.v\tv16,0\(sp\) |
| 241 |
** vs8r.v\tv16,0\([a-x0-9]+\) |
| 242 |
** ... |
| 243 |
** jr\tra |
| 244 |
*/ |
| 245 |
void |
| 246 |
spill_14 (uint32_t *in, uint32_t *out) |
| 247 |
{ |
| 248 |
register vuint32m8_t v8 asm("v8") = *(vuint32m8_t*)in; |
| 249 |
asm volatile ("# %0"::"vr"(v8)); |
| 250 |
exhaust_vector_regs (); |
| 251 |
register vuint32m8_t v16 asm("v16") = v8; |
| 252 |
*(vuint32m8_t*)out = v16; |
| 253 |
asm volatile ("# %0"::"vr"(v16)); |
| 254 |
} |