| 1 |
/* { dg-do compile } */ |
| 2 |
/* { dg-options "-march=rv64gcv_zvfh_zvl4096b -mabi=lp64d -mrvv-max-lmul=m8 -O3 -fno-schedule-insns -fno-schedule-insns2" } */ |
| 3 |
/* { dg-final { check-function-bodies "**" "" } } */ |
| 4 |
|
| 5 |
#include "def.h" |
| 6 |
|
| 7 |
/* |
| 8 |
** mov0: |
| 9 |
** vsetivli\s+zero,\s*1,\s*e16,\s*mf4,\s*t[au],\s*m[au] |
| 10 |
** vle16\.v\tv1,0\s*\([a-x0-9]+\) |
| 11 |
** ... |
| 12 |
** vmv1r\.v\tv2,v1 |
| 13 |
** ... |
| 14 |
** ret |
| 15 |
*/ |
| 16 |
void mov0 (_Float16 *in, _Float16 *out) |
| 17 |
{ |
| 18 |
register v1hf v1 asm("v1") = *(v1hf*)in; |
| 19 |
asm volatile ("# %0"::"vr"(v1)); |
| 20 |
register v1hf v2 asm("v2") = v1; |
| 21 |
*(v1hf*)out = v2; |
| 22 |
asm volatile ("# %0"::"vr"(v2)); |
| 23 |
} |
| 24 |
|
| 25 |
/* |
| 26 |
** mov1: |
| 27 |
** vsetivli\s+zero,\s*2,\s*e16,\s*mf4,\s*t[au],\s*m[au] |
| 28 |
** vle16\.v\tv1,0\s*\([a-x0-9]+\) |
| 29 |
** ... |
| 30 |
** vmv1r\.v\tv2,v1 |
| 31 |
** ... |
| 32 |
** ret |
| 33 |
*/ |
| 34 |
void mov1 (_Float16 *in, _Float16 *out) |
| 35 |
{ |
| 36 |
register v2hf v1 asm("v1") = *(v2hf*)in; |
| 37 |
asm volatile ("# %0"::"vr"(v1)); |
| 38 |
register v2hf v2 asm("v2") = v1; |
| 39 |
*(v2hf*)out = v2; |
| 40 |
asm volatile ("# %0"::"vr"(v2)); |
| 41 |
} |
| 42 |
|
| 43 |
/* |
| 44 |
** mov2: |
| 45 |
** vsetivli\s+zero,\s*4,\s*e16,\s*mf4,\s*t[au],\s*m[au] |
| 46 |
** vle16\.v\tv1,0\s*\([a-x0-9]+\) |
| 47 |
** ... |
| 48 |
** vmv1r\.v\tv2,v1 |
| 49 |
** ... |
| 50 |
** ret |
| 51 |
*/ |
| 52 |
void mov2 (_Float16 *in, _Float16 *out) |
| 53 |
{ |
| 54 |
register v4hf v1 asm("v1") = *(v4hf*)in; |
| 55 |
asm volatile ("# %0"::"vr"(v1)); |
| 56 |
register v4hf v2 asm("v2") = v1; |
| 57 |
*(v4hf*)out = v2; |
| 58 |
asm volatile ("# %0"::"vr"(v2)); |
| 59 |
} |
| 60 |
|
| 61 |
/* |
| 62 |
** mov3: |
| 63 |
** vsetivli\s+zero,\s*8,\s*e16,\s*mf4,\s*t[au],\s*m[au] |
| 64 |
** vle16\.v\tv1,0\s*\([a-x0-9]+\) |
| 65 |
** ... |
| 66 |
** vmv1r\.v\tv2,v1 |
| 67 |
** ... |
| 68 |
** ret |
| 69 |
*/ |
| 70 |
void mov3 (_Float16 *in, _Float16 *out) |
| 71 |
{ |
| 72 |
register v8hf v1 asm("v1") = *(v8hf*)in; |
| 73 |
asm volatile ("# %0"::"vr"(v1)); |
| 74 |
register v8hf v2 asm("v2") = v1; |
| 75 |
*(v8hf*)out = v2; |
| 76 |
asm volatile ("# %0"::"vr"(v2)); |
| 77 |
} |
| 78 |
|
| 79 |
/* |
| 80 |
** mov4: |
| 81 |
** vsetivli\s+zero,\s*16,\s*e16,\s*mf4,\s*t[au],\s*m[au] |
| 82 |
** vle16\.v\tv1,0\s*\([a-x0-9]+\) |
| 83 |
** ... |
| 84 |
** vmv1r\.v\tv2,v1 |
| 85 |
** ... |
| 86 |
** ret |
| 87 |
*/ |
| 88 |
void mov4 (_Float16 *in, _Float16 *out) |
| 89 |
{ |
| 90 |
register v16hf v1 asm("v1") = *(v16hf*)in; |
| 91 |
asm volatile ("# %0"::"vr"(v1)); |
| 92 |
register v16hf v2 asm("v2") = v1; |
| 93 |
*(v16hf*)out = v2; |
| 94 |
asm volatile ("# %0"::"vr"(v2)); |
| 95 |
} |
| 96 |
|
| 97 |
/* |
| 98 |
** mov5: |
| 99 |
** li\s+[a-x0-9]+,32 |
| 100 |
** vsetvli\s+zero,\s*[a-x0-9]+,\s*e16,\s*mf4,\s*t[au],\s*m[au] |
| 101 |
** vle16\.v\tv1,0\s*\([a-x0-9]+\) |
| 102 |
** ... |
| 103 |
** vmv1r\.v\tv2,v1 |
| 104 |
** ... |
| 105 |
** ret |
| 106 |
*/ |
| 107 |
void mov5 (_Float16 *in, _Float16 *out) |
| 108 |
{ |
| 109 |
register v32hf v1 asm("v1") = *(v32hf*)in; |
| 110 |
asm volatile ("# %0"::"vr"(v1)); |
| 111 |
register v32hf v2 asm("v2") = v1; |
| 112 |
*(v32hf*)out = v2; |
| 113 |
asm volatile ("# %0"::"vr"(v2)); |
| 114 |
} |
| 115 |
|
| 116 |
/* |
| 117 |
** mov6: |
| 118 |
** li\s+[a-x0-9]+,64 |
| 119 |
** vsetvli\s+zero,\s*[a-x0-9]+,\s*e16,\s*mf4,\s*t[au],\s*m[au] |
| 120 |
** vle16\.v\tv1,0\s*\([a-x0-9]+\) |
| 121 |
** ... |
| 122 |
** vmv1r\.v\tv2,v1 |
| 123 |
** ... |
| 124 |
** ret |
| 125 |
*/ |
| 126 |
void mov6 (_Float16 *in, _Float16 *out) |
| 127 |
{ |
| 128 |
register v64hf v1 asm("v1") = *(v64hf*)in; |
| 129 |
asm volatile ("# %0"::"vr"(v1)); |
| 130 |
register v64hf v2 asm("v2") = v1; |
| 131 |
*(v64hf*)out = v2; |
| 132 |
asm volatile ("# %0"::"vr"(v2)); |
| 133 |
} |
| 134 |
|
| 135 |
/* |
| 136 |
** mov7: |
| 137 |
** li\s+[a-x0-9]+,128 |
| 138 |
** vsetvli\s+zero,\s*[a-x0-9]+,\s*e16,\s*mf2,\s*t[au],\s*m[au] |
| 139 |
** vle16\.v\tv1,0\s*\([a-x0-9]+\) |
| 140 |
** ... |
| 141 |
** vmv1r\.v\tv2,v1 |
| 142 |
** ... |
| 143 |
** ret |
| 144 |
*/ |
| 145 |
void mov7 (_Float16 *in, _Float16 *out) |
| 146 |
{ |
| 147 |
register v128hf v1 asm("v1") = *(v128hf*)in; |
| 148 |
asm volatile ("# %0"::"vr"(v1)); |
| 149 |
register v128hf v2 asm("v2") = v1; |
| 150 |
*(v128hf*)out = v2; |
| 151 |
asm volatile ("# %0"::"vr"(v2)); |
| 152 |
} |
| 153 |
|
| 154 |
/* |
| 155 |
** mov8: |
| 156 |
** li\s+[a-x0-9]+,256 |
| 157 |
** vsetvli\s+zero,\s*[a-x0-9]+,\s*e16,\s*m1,\s*t[au],\s*m[au] |
| 158 |
** vle16\.v\tv1,0\s*\([a-x0-9]+\) |
| 159 |
** ... |
| 160 |
** vmv1r\.v\tv2,v1 |
| 161 |
** ... |
| 162 |
** ret |
| 163 |
*/ |
| 164 |
void mov8 (_Float16 *in, _Float16 *out) |
| 165 |
{ |
| 166 |
register v256hf v1 asm("v1") = *(v256hf*)in; |
| 167 |
asm volatile ("# %0"::"vr"(v1)); |
| 168 |
register v256hf v2 asm("v2") = v1; |
| 169 |
*(v256hf*)out = v2; |
| 170 |
asm volatile ("# %0"::"vr"(v2)); |
| 171 |
} |
| 172 |
|
| 173 |
/* |
| 174 |
** mov9: |
| 175 |
** li\s+[a-x0-9]+,512 |
| 176 |
** vsetvli\s+zero,\s*[a-x0-9]+,\s*e16,\s*m2,\s*t[au],\s*m[au] |
| 177 |
** vle16\.v\tv2,0\s*\([a-x0-9]+\) |
| 178 |
** ... |
| 179 |
** vmv2r\.v\tv4,v2 |
| 180 |
** ... |
| 181 |
** ret |
| 182 |
*/ |
| 183 |
void mov9 (_Float16 *in, _Float16 *out) |
| 184 |
{ |
| 185 |
register v512hf v1 asm("v2") = *(v512hf*)in; |
| 186 |
asm volatile ("# %0"::"vr"(v1)); |
| 187 |
register v512hf v2 asm("v4") = v1; |
| 188 |
*(v512hf*)out = v2; |
| 189 |
asm volatile ("# %0"::"vr"(v2)); |
| 190 |
} |
| 191 |
|
| 192 |
/* |
| 193 |
** mov10: |
| 194 |
** li\s+[a-x0-9]+,1024 |
| 195 |
** vsetvli\s+zero,\s*[a-x0-9]+,\s*e16,\s*m4,\s*t[au],\s*m[au] |
| 196 |
** vle16\.v\tv4,0\s*\([a-x0-9]+\) |
| 197 |
** ... |
| 198 |
** vmv4r\.v\tv8,v4 |
| 199 |
** ... |
| 200 |
** ret |
| 201 |
*/ |
| 202 |
void mov10 (_Float16 *in, _Float16 *out) |
| 203 |
{ |
| 204 |
register v1024hf v2 asm("v4") = *(v1024hf*)in; |
| 205 |
asm volatile ("# %0"::"vr"(v2)); |
| 206 |
register v1024hf v4 asm("v8") = v2; |
| 207 |
*(v1024hf*)out = v4; |
| 208 |
asm volatile ("# %0"::"vr"(v4)); |
| 209 |
} |
| 210 |
|
| 211 |
/* |
| 212 |
** mov11: |
| 213 |
** li\s+[a-x0-9]+,4096 |
| 214 |
** addi\s+[a-x0-9]+,[a-x0-9]+,-2048 |
| 215 |
** vsetvli\s+zero,\s*[a-x0-9]+,\s*e16,\s*m8,\s*t[au],\s*m[au] |
| 216 |
** vle16\.v\tv8,0\s*\([a-x0-9]+\) |
| 217 |
** ... |
| 218 |
** vmv8r\.v\tv16,v8 |
| 219 |
** ... |
| 220 |
** ret |
| 221 |
*/ |
| 222 |
void mov11 (_Float16 *in, _Float16 *out) |
| 223 |
{ |
| 224 |
register v2048hf v4 asm("v8") = *(v2048hf*)in; |
| 225 |
asm volatile ("# %0"::"vr"(v4)); |
| 226 |
register v2048hf v8 asm("v16") = v4; |
| 227 |
*(v2048hf*)out = v8; |
| 228 |
asm volatile ("# %0"::"vr"(v8)); |
| 229 |
} |