| 1 |
/* { dg-do compile } */ |
| 2 |
/* { dg-options "-march=rv64gcv -mabi=lp64d -O3 -fno-sched-pressure" } */ |
| 3 |
|
| 4 |
#include "riscv_vector.h" |
| 5 |
|
| 6 |
void f0 (void *base1,void *base2,void *base3,void *base4,void *out,size_t vl, uint16_t x) |
| 7 |
{ |
| 8 |
vuint64m8_t v1 = __riscv_vle64_v_u64m8 (base1, vl); |
| 9 |
vbool8_t m1 = __riscv_vlm_v_b8 (base3, vl); |
| 10 |
asm volatile("#" :: |
| 11 |
: "v1", "v2", "v3", "v4", "v5", "v6", "v7", "v16", "v17", |
| 12 |
"v18", "v19", "v20", "v21", "v22", "v23","v24","v25", |
| 13 |
"v26", "v27", "v28", "v29", "v30", "v31"); |
| 14 |
|
| 15 |
vbool8_t v = __riscv_vmseq_vx_u64m8_b8_m(m1,v1,x,vl); |
| 16 |
asm volatile("#" :: |
| 17 |
: "v1", "v2", "v3", "v4", "v5", "v6", "v7", "v8", "v9", |
| 18 |
"v10", "v11", "v12", "v13", "v14", "v15", "v16", "v17", |
| 19 |
"v18", "v19", "v20", "v21", "v22", "v23", "v24", "v25", |
| 20 |
"v26", "v27", "v28", "v29", "v30", "v31"); |
| 21 |
|
| 22 |
__riscv_vsm_v_b8 (out,v,vl); |
| 23 |
} |
| 24 |
|
| 25 |
void f1 (void *base1,void *base2,void *base3,void *base4,void *out,size_t vl, uint16_t x) |
| 26 |
{ |
| 27 |
vuint64m8_t v1 = __riscv_vle64_v_u64m8 (base1, vl); |
| 28 |
vbool8_t m1 = __riscv_vlm_v_b8 (base3, vl); |
| 29 |
asm volatile("#" :: |
| 30 |
: "v1", "v2", "v3", "v4", "v5", "v6", "v7", "v16", "v17", |
| 31 |
"v18", "v19", "v20", "v21", "v22", "v23","v24","v25", |
| 32 |
"v26", "v27", "v28", "v29", "v30", "v31"); |
| 33 |
|
| 34 |
vbool8_t v = __riscv_vmseq_vx_u64m8_b8_mu(m1,m1,v1,x,vl); |
| 35 |
asm volatile("#" :: |
| 36 |
: "v1", "v2", "v3", "v4", "v5", "v6", "v7", "v8", "v9", |
| 37 |
"v10", "v11", "v12", "v13", "v14", "v15", "v16", "v17", |
| 38 |
"v18", "v19", "v20", "v21", "v22", "v23", "v24", "v25", |
| 39 |
"v26", "v27", "v28", "v29", "v30", "v31"); |
| 40 |
|
| 41 |
__riscv_vsm_v_b8 (out,v,vl); |
| 42 |
} |
| 43 |
|
| 44 |
void f2 (void *base1,void *base2,void *base3,void *base4,void *out,size_t vl, uint16_t x) |
| 45 |
{ |
| 46 |
vuint64m8_t v1 = __riscv_vle64_v_u64m8 (base1, vl); |
| 47 |
vbool8_t m1 = __riscv_vlm_v_b8 (base3, vl); |
| 48 |
asm volatile("#" :: |
| 49 |
: "v1", "v2", "v3", "v4", "v5", "v6", "v7", "v16", "v17", |
| 50 |
"v18", "v19", "v20", "v21", "v22", "v23","v24","v25", |
| 51 |
"v26", "v27", "v28", "v29", "v30", "v31"); |
| 52 |
|
| 53 |
vbool8_t v = __riscv_vmseq_vx_u64m8_b8_m(m1,v1,x,vl); |
| 54 |
asm volatile("#" :: |
| 55 |
: "v0", "v1", "v2", "v3", "v4", "v5", "v6", "v7", "v9", |
| 56 |
"v10", "v11", "v12", "v13", "v14", "v15", "v16", "v17", |
| 57 |
"v18", "v19", "v20", "v21", "v22", "v23", "v24", "v25", |
| 58 |
"v26", "v27", "v28", "v29", "v30", "v31"); |
| 59 |
|
| 60 |
__riscv_vsm_v_b8 (out,v,vl); |
| 61 |
} |
| 62 |
|
| 63 |
void f3 (void *base1,void *base2,void *base3,void *base4,void *out,size_t vl, uint16_t x) |
| 64 |
{ |
| 65 |
vuint64m8_t v1 = __riscv_vle64_v_u64m8 (base1, vl); |
| 66 |
vbool8_t m1 = __riscv_vlm_v_b8 (base3, vl); |
| 67 |
asm volatile("#" :: |
| 68 |
: "v1", "v2", "v3", "v4", "v5", "v6", "v7", "v16", "v17", |
| 69 |
"v18", "v19", "v20", "v21", "v22", "v23","v24","v25", |
| 70 |
"v26", "v27", "v28", "v29", "v30", "v31"); |
| 71 |
|
| 72 |
vbool8_t v = __riscv_vmseq_vx_u64m8_b8_m(m1,v1,x,vl); |
| 73 |
asm volatile("#" :: |
| 74 |
: "v0", "v1", "v2", "v3", "v4", "v5", "v6", "v7", "v8", "v9", |
| 75 |
"v10", "v11", "v12", "v13", "v14", "v15", "v17", |
| 76 |
"v18", "v19", "v20", "v21", "v22", "v23", "v24", "v25", |
| 77 |
"v26", "v27", "v28", "v29", "v30", "v31"); |
| 78 |
|
| 79 |
__riscv_vsm_v_b8 (out,v,vl); |
| 80 |
} |
| 81 |
|
| 82 |
void f4 (void *base1,void *base2,void *base3,void *base4,void *out,size_t vl, uint16_t x) |
| 83 |
{ |
| 84 |
vuint64m8_t v1 = __riscv_vle64_v_u64m8 (base1, vl); |
| 85 |
vbool8_t m1 = __riscv_vlm_v_b8 (base3, vl); |
| 86 |
asm volatile("#" :: |
| 87 |
: "v1", "v2", "v3", "v4", "v5", "v6", "v7", "v8", "v9", |
| 88 |
"v10", "v11", "v12", "v13", "v14", "v15", "v24", "v25", |
| 89 |
"v26", "v27", "v28", "v29", "v30", "v31"); |
| 90 |
|
| 91 |
vbool8_t v = __riscv_vmseq_vx_u64m8_b8_m(m1,v1,x,vl); |
| 92 |
asm volatile("#" :: |
| 93 |
: "v0", "v1", "v2", "v3", "v4", "v5", "v6", "v7", "v8", "v9", |
| 94 |
"v10", "v11", "v12", "v13", "v14", "v15", "v17", |
| 95 |
"v18", "v19", "v20", "v21", "v22", "v23", "v24", "v25", |
| 96 |
"v26", "v27", "v28", "v29", "v30", "v31"); |
| 97 |
|
| 98 |
__riscv_vsm_v_b8 (out,v,vl); |
| 99 |
} |
| 100 |
|
| 101 |
void f5 (void *base1,void *base2,void *base3,void *base4,void *out,size_t vl, uint16_t x) |
| 102 |
{ |
| 103 |
vuint64m8_t v1 = __riscv_vle64_v_u64m8 (base1, vl); |
| 104 |
vbool8_t m1 = __riscv_vlm_v_b8 (base3, vl); |
| 105 |
asm volatile("#" :: |
| 106 |
: "v1", "v2", "v3", "v4", "v5", "v6", "v7", "v16", "v17", |
| 107 |
"v18", "v19", "v20", "v21", "v22", "v23","v24","v25", |
| 108 |
"v26", "v27", "v28", "v29", "v30", "v31"); |
| 109 |
|
| 110 |
vbool8_t v = __riscv_vmseq_vx_u64m8_b8_m(m1,v1,x,vl); |
| 111 |
asm volatile("#" :: |
| 112 |
: "v0", "v1", "v2", "v3", "v4", "v5", "v6", "v7", "v8", "v9", |
| 113 |
"v10", "v11", "v12", "v13", "v14", "v15", "v16", "v17", |
| 114 |
"v18", "v19", "v20", "v21", "v22", "v23", "v24", "v25", |
| 115 |
"v26", "v28", "v29", "v30", "v31"); |
| 116 |
|
| 117 |
__riscv_vsm_v_b8 (out,v,vl); |
| 118 |
} |
| 119 |
|
| 120 |
void f6 (void *base1,void *base2,void *base3,void *base4,void *out,size_t vl, uint16_t x) |
| 121 |
{ |
| 122 |
vuint64m8_t v1 = __riscv_vle64_v_u64m8 (base1, vl); |
| 123 |
vbool8_t m1 = __riscv_vlm_v_b8 (base3, vl); |
| 124 |
vbool8_t m2 = __riscv_vlm_v_b8 (base4, vl); |
| 125 |
asm volatile("#" :: |
| 126 |
: "v1", "v2", "v3", "v4", "v5", "v6", "v7", "v16", "v17", |
| 127 |
"v18", "v19", "v20", "v21", "v22", "v23","v24","v25", |
| 128 |
"v26", "v28", "v29", "v30", "v31"); |
| 129 |
|
| 130 |
vbool8_t v = __riscv_vmseq_vx_u64m8_b8_mu(m1,m2,v1,x,vl); |
| 131 |
asm volatile("#" :: |
| 132 |
: "v0", "v1", "v2", "v3", "v4", "v5", "v6", "v7", "v8", "v9", |
| 133 |
"v10", "v11", "v12", "v13", "v14", "v15", "v16", "v17", |
| 134 |
"v18", "v19", "v20", "v21", "v22", "v23", "v24", "v25", |
| 135 |
"v26", "v28", "v29", "v30", "v31"); |
| 136 |
|
| 137 |
__riscv_vsm_v_b8 (out,v,vl); |
| 138 |
} |
| 139 |
|
| 140 |
void f7 (void * in, void *out, int32_t x) |
| 141 |
{ |
| 142 |
vbool8_t mask = *(vbool8_t*)in; |
| 143 |
asm volatile ("":::"memory"); |
| 144 |
vint64m8_t v = __riscv_vle64_v_i64m8 (in, 4); |
| 145 |
vint64m8_t v2 = __riscv_vle64_v_i64m8_m (mask, in, 4); |
| 146 |
vbool8_t m3 = __riscv_vmseq_vx_i64m8_b8 (v, x, 4); |
| 147 |
vbool8_t m4 = __riscv_vmseq_vx_i64m8_b8_mu (m3, m3, v2, x, 4); |
| 148 |
__riscv_vsm_v_b8 (out, m4, 4); |
| 149 |
} |
| 150 |
|
| 151 |
void f8 (void * in, void *out, int32_t x) |
| 152 |
{ |
| 153 |
vbool8_t mask = *(vbool8_t*)in; |
| 154 |
asm volatile ("":::"memory"); |
| 155 |
vint64m8_t v = __riscv_vle64_v_i64m8 (in, 4); |
| 156 |
vint64m8_t v2 = __riscv_vle64_v_i64m8_m (mask, in, 4); |
| 157 |
vbool8_t m3 = __riscv_vmseq_vx_i64m8_b8 (v, x, 4); |
| 158 |
vbool8_t m4 = __riscv_vmseq_vx_i64m8_b8_m (m3, v2, x, 4); |
| 159 |
__riscv_vsm_v_b8 (out, m3, 4); |
| 160 |
__riscv_vsm_v_b8 (out, m4, 4); |
| 161 |
} |
| 162 |
|
| 163 |
void f9 (void * in, void *out, int32_t x) |
| 164 |
{ |
| 165 |
vbool8_t mask = *(vbool8_t*)in; |
| 166 |
asm volatile ("":::"memory"); |
| 167 |
vint64m8_t v = __riscv_vle64_v_i64m8 (in, 4); |
| 168 |
vint64m8_t v2 = __riscv_vle64_v_i64m8_m (mask, in, 4); |
| 169 |
vbool8_t m3 = __riscv_vmseq_vx_i64m8_b8 (v, x, 4); |
| 170 |
vbool8_t m4 = __riscv_vmseq_vx_i64m8_b8_m (m3, v2, x, 4); |
| 171 |
__riscv_vsm_v_b8 (out, m4, 4); |
| 172 |
vbool8_t m5 = __riscv_vmseq_vx_i64m8_b8_mu (m3, m4, v2, x, 4); |
| 173 |
__riscv_vsm_v_b8 (out, m3, 4); |
| 174 |
__riscv_vsm_v_b8 (out, m5, 4); |
| 175 |
} |
| 176 |
|
| 177 |
void f10 (void * in, void *out, int32_t x) |
| 178 |
{ |
| 179 |
vbool8_t mask = *(vbool8_t*)in; |
| 180 |
asm volatile ("":::"memory"); |
| 181 |
vint64m8_t v = __riscv_vle64_v_i64m8 (in, 4); |
| 182 |
vint64m8_t v2 = __riscv_vle64_v_i64m8_m (mask, in, 4); |
| 183 |
vbool8_t m3 = __riscv_vmseq_vx_i64m8_b8 (v, x, 4); |
| 184 |
vbool8_t m4 = __riscv_vmseq_vx_i64m8_b8_mu (m3, m3, v2, x, 4); |
| 185 |
__riscv_vsm_v_b8 (out, m4, 4); |
| 186 |
} |
| 187 |
|
| 188 |
void f11 (void * in, void *out, int32_t x) |
| 189 |
{ |
| 190 |
vbool8_t mask = *(vbool8_t*)in; |
| 191 |
asm volatile ("":::"memory"); |
| 192 |
vint64m8_t v = __riscv_vle64_v_i64m8 (in, 4); |
| 193 |
vint64m8_t v2 = __riscv_vle64_v_i64m8_m (mask, in, 4); |
| 194 |
vbool8_t m3 = __riscv_vmseq_vx_i64m8_b8 (v, x, 4); |
| 195 |
vbool8_t m4 = __riscv_vmseq_vx_i64m8_b8_m (m3, v2, x, 4); |
| 196 |
__riscv_vsm_v_b8 (out, m3, 4); |
| 197 |
__riscv_vsm_v_b8 (out, m4, 4); |
| 198 |
} |
| 199 |
|
| 200 |
void f12 (void* base1,void* base2,void* out,int n, int32_t x) |
| 201 |
{ |
| 202 |
vbool8_t mask = *(vbool8_t*)base1; |
| 203 |
vint64m8_t v = __riscv_vle64_v_i64m8 (base1, 4); |
| 204 |
vint64m8_t v2 = __riscv_vle64_v_i64m8_m (mask, base1, 4); |
| 205 |
mask = __riscv_vmseq_vx_i64m8_b8 (v, x, 4); |
| 206 |
for (int i = 0; i < n; i++){ |
| 207 |
vint64m8_t v3 = __riscv_vle64_v_i64m8 (base1 + i, 4); |
| 208 |
vint64m8_t v4 = __riscv_vle64_v_i64m8_m (mask, base1 + i * 2, 4); |
| 209 |
mask = __riscv_vmseq_vx_i64m8_b8_m (mask, v3, x,32); |
| 210 |
mask = __riscv_vmseq_vx_i64m8_b8_mu (mask, mask, v4, x, 32); |
| 211 |
} |
| 212 |
__riscv_vsm_v_b8 (out, mask, 32); |
| 213 |
} |
| 214 |
|
| 215 |
void f13 (void* base1,void* base2,void* out,int n, int32_t x) |
| 216 |
{ |
| 217 |
vbool64_t mask = *(vbool64_t*)base1; |
| 218 |
vint64m1_t v = __riscv_vle64_v_i64m1 (base1, 4); |
| 219 |
vint64m1_t v2 = __riscv_vle64_v_i64m1_m (mask, base1, 4); |
| 220 |
mask = __riscv_vmseq_vx_i64m1_b64 (v, x, 4); |
| 221 |
for (int i = 0; i < n; i++){ |
| 222 |
vint64m1_t v3 = __riscv_vle64_v_i64m1 (base1 + i, 4); |
| 223 |
vint64m1_t v4 = __riscv_vle64_v_i64m1_m (mask, base1 + i * 2, 4); |
| 224 |
mask = __riscv_vmseq_vx_i64m1_b64_m (mask, v3, x,32); |
| 225 |
mask = __riscv_vmseq_vx_i64m1_b64_mu (mask, mask, v4, x, 32); |
| 226 |
} |
| 227 |
__riscv_vsm_v_b64 (out, mask, 32); |
| 228 |
} |
| 229 |
|
| 230 |
/* { dg-final { scan-assembler-not {vmv} } } */ |
| 231 |
/* { dg-final { scan-assembler-not {csrr} } } */ |