| 1 |
/* { dg-do compile } */ |
| 2 |
/* { dg-options "-march=rv64gcv -mabi=lp64d -O3" } */ |
| 3 |
|
| 4 |
#include "riscv_vector.h" |
| 5 |
|
| 6 |
size_t __attribute__ ((noinline)) |
| 7 |
sumation (size_t sum0, size_t sum1, size_t sum2, size_t sum3, size_t sum4, |
| 8 |
size_t sum5, size_t sum6, size_t sum7, |
| 9 |
size_t sum0_2, size_t sum1_2, size_t sum2_2, size_t sum3_2, size_t sum4_2, |
| 10 |
size_t sum5_2, size_t sum6_2, size_t sum7_2) |
| 11 |
{ |
| 12 |
return sum0 + sum1 + sum2 + sum3 + sum4 + sum5 + sum6 + sum7 |
| 13 |
+ sum0_2 + sum1_2 + sum2_2 + sum3_2 + sum4_2 + sum5_2 + sum6_2 + sum7_2; |
| 14 |
} |
| 15 |
|
| 16 |
size_t |
| 17 |
foo (char const *buf, size_t len) |
| 18 |
{ |
| 19 |
size_t 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 |
vint8m2_t v0 = __riscv_vle8_v_i8m2 ((void *) it, vl); |
| 26 |
it += vl; |
| 27 |
vint8m2_t v1 = __riscv_vle8_v_i8m2 ((void *) it, vl); |
| 28 |
it += vl; |
| 29 |
vint8m2_t v2 = __riscv_vle8_v_i8m2 ((void *) it, vl); |
| 30 |
it += vl; |
| 31 |
vint8m2_t v3 = __riscv_vle8_v_i8m2 ((void *) it, vl); |
| 32 |
it += vl; |
| 33 |
vint8m2_t v4 = __riscv_vle8_v_i8m2 ((void *) it, vl); |
| 34 |
it += vl; |
| 35 |
vint8m2_t v5 = __riscv_vle8_v_i8m2 ((void *) it, vl); |
| 36 |
it += vl; |
| 37 |
vint8m2_t v6 = __riscv_vle8_v_i8m2 ((void *) it, vl); |
| 38 |
it += vl; |
| 39 |
vint8m2_t v7 = __riscv_vle8_v_i8m2 ((void *) it, vl); |
| 40 |
it += vl; |
| 41 |
|
| 42 |
vint16m1_t vw0 = __riscv_vle16_v_i16m1 ((void *) it, vl); |
| 43 |
it += vl; |
| 44 |
vint16m1_t vw1 = __riscv_vle16_v_i16m1 ((void *) it, vl); |
| 45 |
it += vl; |
| 46 |
vint16m1_t vw2 = __riscv_vle16_v_i16m1 ((void *) it, vl); |
| 47 |
it += vl; |
| 48 |
vint16m1_t vw3 = __riscv_vle16_v_i16m1 ((void *) it, vl); |
| 49 |
it += vl; |
| 50 |
vint16m1_t vw4 = __riscv_vle16_v_i16m1 ((void *) it, vl); |
| 51 |
it += vl; |
| 52 |
vint16m1_t vw5 = __riscv_vle16_v_i16m1 ((void *) it, vl); |
| 53 |
it += vl; |
| 54 |
vint16m1_t vw6 = __riscv_vle16_v_i16m1 ((void *) it, vl); |
| 55 |
it += vl; |
| 56 |
vint16m1_t vw7 = __riscv_vle16_v_i16m1 ((void *) it, vl); |
| 57 |
it += vl; |
| 58 |
|
| 59 |
asm volatile("nop" ::: "memory"); |
| 60 |
vint16m1_t vw0_2 = __riscv_vwredsum_vs_i8m2_i16m1 (v0, vw0, vl); |
| 61 |
vint16m1_t vw1_2 = __riscv_vwredsum_vs_i8m2_i16m1 (v1, vw1, vl); |
| 62 |
vint16m1_t vw2_2 = __riscv_vwredsum_vs_i8m2_i16m1 (v2, vw2, vl); |
| 63 |
vint16m1_t vw3_2 = __riscv_vwredsum_vs_i8m2_i16m1 (v3, vw3, vl); |
| 64 |
vint16m1_t vw4_2 = __riscv_vwredsum_vs_i8m2_i16m1 (v4, vw4, vl); |
| 65 |
vint16m1_t vw5_2 = __riscv_vwredsum_vs_i8m2_i16m1 (v5, vw5, vl); |
| 66 |
vint16m1_t vw6_2 = __riscv_vwredsum_vs_i8m2_i16m1 (v6, vw6, vl); |
| 67 |
vint16m1_t vw7_2 = __riscv_vwredsum_vs_i8m2_i16m1 (v7, vw7, vl); |
| 68 |
|
| 69 |
vw0 = __riscv_vwredsum_vs_i8m2_i16m1 (v0, vw0_2, vl); |
| 70 |
vw1 = __riscv_vwredsum_vs_i8m2_i16m1 (v1, vw1_2, vl); |
| 71 |
vw2 = __riscv_vwredsum_vs_i8m2_i16m1 (v2, vw2_2, vl); |
| 72 |
vw3 = __riscv_vwredsum_vs_i8m2_i16m1 (v3, vw3_2, vl); |
| 73 |
vw4 = __riscv_vwredsum_vs_i8m2_i16m1 (v4, vw4_2, vl); |
| 74 |
vw5 = __riscv_vwredsum_vs_i8m2_i16m1 (v5, vw5_2, vl); |
| 75 |
vw6 = __riscv_vwredsum_vs_i8m2_i16m1 (v6, vw6_2, vl); |
| 76 |
vw7 = __riscv_vwredsum_vs_i8m2_i16m1 (v7, vw7_2, vl); |
| 77 |
|
| 78 |
asm volatile("nop" ::: "memory"); |
| 79 |
size_t sum0 = __riscv_vmv_x_s_i16m1_i16 (vw0); |
| 80 |
size_t sum1 = __riscv_vmv_x_s_i16m1_i16 (vw1); |
| 81 |
size_t sum2 = __riscv_vmv_x_s_i16m1_i16 (vw2); |
| 82 |
size_t sum3 = __riscv_vmv_x_s_i16m1_i16 (vw3); |
| 83 |
size_t sum4 = __riscv_vmv_x_s_i16m1_i16 (vw4); |
| 84 |
size_t sum5 = __riscv_vmv_x_s_i16m1_i16 (vw5); |
| 85 |
size_t sum6 = __riscv_vmv_x_s_i16m1_i16 (vw6); |
| 86 |
size_t sum7 = __riscv_vmv_x_s_i16m1_i16 (vw7); |
| 87 |
|
| 88 |
size_t sum0_2 = __riscv_vmv_x_s_i16m1_i16 (vw0_2); |
| 89 |
size_t sum1_2 = __riscv_vmv_x_s_i16m1_i16 (vw1_2); |
| 90 |
size_t sum2_2 = __riscv_vmv_x_s_i16m1_i16 (vw2_2); |
| 91 |
size_t sum3_2 = __riscv_vmv_x_s_i16m1_i16 (vw3_2); |
| 92 |
size_t sum4_2 = __riscv_vmv_x_s_i16m1_i16 (vw4_2); |
| 93 |
size_t sum5_2 = __riscv_vmv_x_s_i16m1_i16 (vw5_2); |
| 94 |
size_t sum6_2 = __riscv_vmv_x_s_i16m1_i16 (vw6_2); |
| 95 |
size_t sum7_2 = __riscv_vmv_x_s_i16m1_i16 (vw7_2); |
| 96 |
|
| 97 |
sum += sumation (sum0, sum1, sum2, sum3, sum4, sum5, sum6, sum7, |
| 98 |
sum0_2, sum1_2, sum2_2, sum3_2, sum4_2, sum5_2, sum6_2, sum7_2); |
| 99 |
} |
| 100 |
return sum; |
| 101 |
} |
| 102 |
|
| 103 |
/* { dg-final { scan-assembler-not {vmv1r} } } */ |
| 104 |
/* { dg-final { scan-assembler-not {vmv2r} } } */ |
| 105 |
/* { dg-final { scan-assembler-not {vmv4r} } } */ |
| 106 |
/* { dg-final { scan-assembler-not {vmv8r} } } */ |
| 107 |
/* { dg-final { scan-assembler-not {csrr} } } */ |