| 1 |
/* { dg-do compile } */ |
| 2 |
/* { dg-options "-march=rv64gcv -mabi=lp64d -O3" } */ |
| 3 |
|
| 4 |
#include "riscv_vector.h" |
| 5 |
|
| 6 |
double __attribute__ ((noinline)) |
| 7 |
sumation (double sum0, double sum1, double sum2, double sum3, double sum4, |
| 8 |
double sum5, double sum6, double sum7, double sum8, double sum9, |
| 9 |
double sum10, double sum11, double sum12, double sum13, double sum14, |
| 10 |
double sum15) |
| 11 |
{ |
| 12 |
return sum0 + sum1 + sum2 + sum3 + sum4 + sum5 + sum6 + sum7 + sum8 + sum9 |
| 13 |
+ sum10 + sum11 + sum12 + sum13 + sum14 + sum15; |
| 14 |
} |
| 15 |
|
| 16 |
double |
| 17 |
foo (char const *buf, size_t len) |
| 18 |
{ |
| 19 |
double sum = 0; |
| 20 |
size_t vl = __riscv_vsetvlmax_e8m8 (); |
| 21 |
size_t step = vl * 4; |
| 22 |
const char *it = buf, *end = buf + len; |
| 23 |
for (; it + step <= end;) |
| 24 |
{ |
| 25 |
vfloat32m1_t v0 = __riscv_vle32_v_f32m1 ((void *) it, vl); |
| 26 |
it += vl; |
| 27 |
vfloat32m1_t v1 = __riscv_vle32_v_f32m1 ((void *) it, vl); |
| 28 |
it += vl; |
| 29 |
vfloat32m1_t v2 = __riscv_vle32_v_f32m1 ((void *) it, vl); |
| 30 |
it += vl; |
| 31 |
vfloat32m1_t v3 = __riscv_vle32_v_f32m1 ((void *) it, vl); |
| 32 |
it += vl; |
| 33 |
vfloat32m1_t v4 = __riscv_vle32_v_f32m1 ((void *) it, vl); |
| 34 |
it += vl; |
| 35 |
vfloat32m1_t v5 = __riscv_vle32_v_f32m1 ((void *) it, vl); |
| 36 |
it += vl; |
| 37 |
vfloat32m1_t v6 = __riscv_vle32_v_f32m1 ((void *) it, vl); |
| 38 |
it += vl; |
| 39 |
vfloat32m1_t v7 = __riscv_vle32_v_f32m1 ((void *) it, vl); |
| 40 |
it += vl; |
| 41 |
vfloat32m1_t v8 = __riscv_vle32_v_f32m1 ((void *) it, vl); |
| 42 |
it += vl; |
| 43 |
vfloat32m1_t v9 = __riscv_vle32_v_f32m1 ((void *) it, vl); |
| 44 |
it += vl; |
| 45 |
vfloat32m1_t v10 = __riscv_vle32_v_f32m1 ((void *) it, vl); |
| 46 |
it += vl; |
| 47 |
vfloat32m1_t v11 = __riscv_vle32_v_f32m1 ((void *) it, vl); |
| 48 |
it += vl; |
| 49 |
vfloat32m1_t v12 = __riscv_vle32_v_f32m1 ((void *) it, vl); |
| 50 |
it += vl; |
| 51 |
vfloat32m1_t v13 = __riscv_vle32_v_f32m1 ((void *) it, vl); |
| 52 |
it += vl; |
| 53 |
vfloat32m1_t v14 = __riscv_vle32_v_f32m1 ((void *) it, vl); |
| 54 |
it += vl; |
| 55 |
vfloat32m1_t v15 = __riscv_vle32_v_f32m1 ((void *) it, vl); |
| 56 |
it += vl; |
| 57 |
|
| 58 |
asm volatile("nop" ::: "memory"); |
| 59 |
vint64m2_t vw0 = __riscv_vfwcvt_rtz_x_f_v_i64m2 (v0, vl); |
| 60 |
vint64m2_t vw1 = __riscv_vfwcvt_rtz_x_f_v_i64m2 (v1, vl); |
| 61 |
vint64m2_t vw2 = __riscv_vfwcvt_rtz_x_f_v_i64m2 (v2, vl); |
| 62 |
vint64m2_t vw3 = __riscv_vfwcvt_rtz_x_f_v_i64m2 (v3, vl); |
| 63 |
vint64m2_t vw4 = __riscv_vfwcvt_rtz_x_f_v_i64m2 (v4, vl); |
| 64 |
vint64m2_t vw5 = __riscv_vfwcvt_rtz_x_f_v_i64m2 (v5, vl); |
| 65 |
vint64m2_t vw6 = __riscv_vfwcvt_rtz_x_f_v_i64m2 (v6, vl); |
| 66 |
vint64m2_t vw7 = __riscv_vfwcvt_rtz_x_f_v_i64m2 (v7, vl); |
| 67 |
vint64m2_t vw8 = __riscv_vfwcvt_rtz_x_f_v_i64m2 (v8, vl); |
| 68 |
vint64m2_t vw9 = __riscv_vfwcvt_rtz_x_f_v_i64m2 (v9, vl); |
| 69 |
vint64m2_t vw10 = __riscv_vfwcvt_rtz_x_f_v_i64m2 (v10, vl); |
| 70 |
vint64m2_t vw11 = __riscv_vfwcvt_rtz_x_f_v_i64m2 (v11, vl); |
| 71 |
vint64m2_t vw12 = __riscv_vfwcvt_rtz_x_f_v_i64m2 (v12, vl); |
| 72 |
vint64m2_t vw13 = __riscv_vfwcvt_rtz_x_f_v_i64m2 (v13, vl); |
| 73 |
vint64m2_t vw14 = __riscv_vfwcvt_rtz_x_f_v_i64m2 (v14, vl); |
| 74 |
vint64m2_t vw15 = __riscv_vfwcvt_rtz_x_f_v_i64m2 (v15, vl); |
| 75 |
|
| 76 |
asm volatile("nop" ::: "memory"); |
| 77 |
double sum0 = __riscv_vmv_x_s_i64m2_i64 (vw0); |
| 78 |
double sum1 = __riscv_vmv_x_s_i64m2_i64 (vw1); |
| 79 |
double sum2 = __riscv_vmv_x_s_i64m2_i64 (vw2); |
| 80 |
double sum3 = __riscv_vmv_x_s_i64m2_i64 (vw3); |
| 81 |
double sum4 = __riscv_vmv_x_s_i64m2_i64 (vw4); |
| 82 |
double sum5 = __riscv_vmv_x_s_i64m2_i64 (vw5); |
| 83 |
double sum6 = __riscv_vmv_x_s_i64m2_i64 (vw6); |
| 84 |
double sum7 = __riscv_vmv_x_s_i64m2_i64 (vw7); |
| 85 |
double sum8 = __riscv_vmv_x_s_i64m2_i64 (vw8); |
| 86 |
double sum9 = __riscv_vmv_x_s_i64m2_i64 (vw9); |
| 87 |
double sum10 = __riscv_vmv_x_s_i64m2_i64 (vw10); |
| 88 |
double sum11 = __riscv_vmv_x_s_i64m2_i64 (vw11); |
| 89 |
double sum12 = __riscv_vmv_x_s_i64m2_i64 (vw12); |
| 90 |
double sum13 = __riscv_vmv_x_s_i64m2_i64 (vw13); |
| 91 |
double sum14 = __riscv_vmv_x_s_i64m2_i64 (vw14); |
| 92 |
double sum15 = __riscv_vmv_x_s_i64m2_i64 (vw15); |
| 93 |
|
| 94 |
sum += sumation (sum0, sum1, sum2, sum3, sum4, sum5, sum6, sum7, sum8, |
| 95 |
sum9, sum10, sum11, sum12, sum13, sum14, sum15); |
| 96 |
} |
| 97 |
return sum; |
| 98 |
} |
| 99 |
|
| 100 |
double |
| 101 |
foo2 (char const *buf, size_t len) |
| 102 |
{ |
| 103 |
double sum = 0; |
| 104 |
size_t vl = __riscv_vsetvlmax_e8m8 (); |
| 105 |
size_t step = vl * 4; |
| 106 |
const char *it = buf, *end = buf + len; |
| 107 |
for (; it + step <= end;) |
| 108 |
{ |
| 109 |
vfloat32m1_t v0 = __riscv_vle32_v_f32m1 ((void *) it, vl); |
| 110 |
it += vl; |
| 111 |
vfloat32m1_t v1 = __riscv_vle32_v_f32m1 ((void *) it, vl); |
| 112 |
it += vl; |
| 113 |
vfloat32m1_t v2 = __riscv_vle32_v_f32m1 ((void *) it, vl); |
| 114 |
it += vl; |
| 115 |
vfloat32m1_t v3 = __riscv_vle32_v_f32m1 ((void *) it, vl); |
| 116 |
it += vl; |
| 117 |
vfloat32m1_t v4 = __riscv_vle32_v_f32m1 ((void *) it, vl); |
| 118 |
it += vl; |
| 119 |
vfloat32m1_t v5 = __riscv_vle32_v_f32m1 ((void *) it, vl); |
| 120 |
it += vl; |
| 121 |
vfloat32m1_t v6 = __riscv_vle32_v_f32m1 ((void *) it, vl); |
| 122 |
it += vl; |
| 123 |
vfloat32m1_t v7 = __riscv_vle32_v_f32m1 ((void *) it, vl); |
| 124 |
it += vl; |
| 125 |
vfloat32m1_t v8 = __riscv_vle32_v_f32m1 ((void *) it, vl); |
| 126 |
it += vl; |
| 127 |
vfloat32m1_t v9 = __riscv_vle32_v_f32m1 ((void *) it, vl); |
| 128 |
it += vl; |
| 129 |
vfloat32m1_t v10 = __riscv_vle32_v_f32m1 ((void *) it, vl); |
| 130 |
it += vl; |
| 131 |
vfloat32m1_t v11 = __riscv_vle32_v_f32m1 ((void *) it, vl); |
| 132 |
it += vl; |
| 133 |
vfloat32m1_t v12 = __riscv_vle32_v_f32m1 ((void *) it, vl); |
| 134 |
it += vl; |
| 135 |
vfloat32m1_t v13 = __riscv_vle32_v_f32m1 ((void *) it, vl); |
| 136 |
it += vl; |
| 137 |
vfloat32m1_t v14 = __riscv_vle32_v_f32m1 ((void *) it, vl); |
| 138 |
it += vl; |
| 139 |
vfloat32m1_t v15 = __riscv_vle32_v_f32m1 ((void *) it, vl); |
| 140 |
it += vl; |
| 141 |
|
| 142 |
asm volatile("nop" ::: "memory"); |
| 143 |
vint64m2_t vw0 = __riscv_vfwcvt_x_f_v_i64m2 (v0, vl); |
| 144 |
vint64m2_t vw1 = __riscv_vfwcvt_x_f_v_i64m2 (v1, vl); |
| 145 |
vint64m2_t vw2 = __riscv_vfwcvt_x_f_v_i64m2 (v2, vl); |
| 146 |
vint64m2_t vw3 = __riscv_vfwcvt_x_f_v_i64m2 (v3, vl); |
| 147 |
vint64m2_t vw4 = __riscv_vfwcvt_x_f_v_i64m2 (v4, vl); |
| 148 |
vint64m2_t vw5 = __riscv_vfwcvt_x_f_v_i64m2 (v5, vl); |
| 149 |
vint64m2_t vw6 = __riscv_vfwcvt_x_f_v_i64m2 (v6, vl); |
| 150 |
vint64m2_t vw7 = __riscv_vfwcvt_x_f_v_i64m2 (v7, vl); |
| 151 |
vint64m2_t vw8 = __riscv_vfwcvt_x_f_v_i64m2 (v8, vl); |
| 152 |
vint64m2_t vw9 = __riscv_vfwcvt_x_f_v_i64m2 (v9, vl); |
| 153 |
vint64m2_t vw10 = __riscv_vfwcvt_x_f_v_i64m2 (v10, vl); |
| 154 |
vint64m2_t vw11 = __riscv_vfwcvt_x_f_v_i64m2 (v11, vl); |
| 155 |
vint64m2_t vw12 = __riscv_vfwcvt_x_f_v_i64m2 (v12, vl); |
| 156 |
vint64m2_t vw13 = __riscv_vfwcvt_x_f_v_i64m2 (v13, vl); |
| 157 |
vint64m2_t vw14 = __riscv_vfwcvt_x_f_v_i64m2 (v14, vl); |
| 158 |
vint64m2_t vw15 = __riscv_vfwcvt_x_f_v_i64m2 (v15, vl); |
| 159 |
|
| 160 |
asm volatile("nop" ::: "memory"); |
| 161 |
double sum0 = __riscv_vmv_x_s_i64m2_i64 (vw0); |
| 162 |
double sum1 = __riscv_vmv_x_s_i64m2_i64 (vw1); |
| 163 |
double sum2 = __riscv_vmv_x_s_i64m2_i64 (vw2); |
| 164 |
double sum3 = __riscv_vmv_x_s_i64m2_i64 (vw3); |
| 165 |
double sum4 = __riscv_vmv_x_s_i64m2_i64 (vw4); |
| 166 |
double sum5 = __riscv_vmv_x_s_i64m2_i64 (vw5); |
| 167 |
double sum6 = __riscv_vmv_x_s_i64m2_i64 (vw6); |
| 168 |
double sum7 = __riscv_vmv_x_s_i64m2_i64 (vw7); |
| 169 |
double sum8 = __riscv_vmv_x_s_i64m2_i64 (vw8); |
| 170 |
double sum9 = __riscv_vmv_x_s_i64m2_i64 (vw9); |
| 171 |
double sum10 = __riscv_vmv_x_s_i64m2_i64 (vw10); |
| 172 |
double sum11 = __riscv_vmv_x_s_i64m2_i64 (vw11); |
| 173 |
double sum12 = __riscv_vmv_x_s_i64m2_i64 (vw12); |
| 174 |
double sum13 = __riscv_vmv_x_s_i64m2_i64 (vw13); |
| 175 |
double sum14 = __riscv_vmv_x_s_i64m2_i64 (vw14); |
| 176 |
double sum15 = __riscv_vmv_x_s_i64m2_i64 (vw15); |
| 177 |
|
| 178 |
sum += sumation (sum0, sum1, sum2, sum3, sum4, sum5, sum6, sum7, sum8, |
| 179 |
sum9, sum10, sum11, sum12, sum13, sum14, sum15); |
| 180 |
} |
| 181 |
return sum; |
| 182 |
} |
| 183 |
|
| 184 |
/* { dg-final { scan-assembler-not {vmv1r} } } */ |
| 185 |
/* { dg-final { scan-assembler-not {vmv2r} } } */ |
| 186 |
/* { dg-final { scan-assembler-not {vmv4r} } } */ |
| 187 |
/* { dg-final { scan-assembler-not {vmv8r} } } */ |
| 188 |
/* { dg-final { scan-assembler-not {csrr} { xfail riscv*-*-* } } } */ |