| 1 |
/* { dg-do compile } */ |
| 2 |
/* { dg-options "-mrvv-vector-bits=scalable -march=rv32gcv -mabi=ilp32 -fno-tree-vectorize -fno-schedule-insns -fno-schedule-insns2" } */ |
| 3 |
|
| 4 |
#include "riscv_vector.h" |
| 5 |
|
| 6 |
void f (int8_t * restrict in, int8_t * restrict out, int l, int n, int m, size_t cond) |
| 7 |
{ |
| 8 |
size_t vl = in[0] + 555; |
| 9 |
|
| 10 |
if (cond) { |
| 11 |
for (int i = 0; i < l; i++){ |
| 12 |
for (int j = 0; j < m; j++){ |
| 13 |
for (int k = 0; k < n; k++) |
| 14 |
{ |
| 15 |
vint8mf8_t v = __riscv_vle8_v_i8mf8 (in + i + j + k, vl); |
| 16 |
__riscv_vse8_v_i8mf8 (out + i + j + k, v, vl); |
| 17 |
} |
| 18 |
} |
| 19 |
} |
| 20 |
} else { |
| 21 |
for (int i = 0; i < l; i++){ |
| 22 |
for (int j = 0; j < m; j++){ |
| 23 |
for (int k = 0; k < n; k++) |
| 24 |
{ |
| 25 |
vint32mf2_t v = __riscv_vle32_v_i32mf2 ((int32_t *)(in + i + j + k), vl); |
| 26 |
__riscv_vse32_v_i32mf2 ((int32_t *)(out + i + j + k), v, vl); |
| 27 |
} |
| 28 |
} |
| 29 |
} |
| 30 |
} |
| 31 |
|
| 32 |
for (int i = 0; i < l; i++){ |
| 33 |
for (int j = 0; j < m; j++){ |
| 34 |
for (int k = 0; k < n; k++) |
| 35 |
{ |
| 36 |
out[i+j+k+10000000] = out[i+j+k+10000000] + in[i+j+k+10000000]; |
| 37 |
} |
| 38 |
} |
| 39 |
} |
| 40 |
|
| 41 |
for (int i = 0; i < l; i++){ |
| 42 |
for (int j = 0; j < m; j++){ |
| 43 |
for (int k = 0; k < n; k++) |
| 44 |
{ |
| 45 |
vint8mf8_t v = __riscv_vle8_v_i8mf8 (in + i + j + k + 10000, vl); |
| 46 |
v = __riscv_vle8_v_i8mf8_tu (v, in + i + j + k + 20000, vl); |
| 47 |
__riscv_vse8_v_i8mf8 (out + i + j + k + 10000, v, vl); |
| 48 |
} |
| 49 |
} |
| 50 |
} |
| 51 |
} |
| 52 |
|
| 53 |
/* { dg-final { scan-assembler-times {vsetvli\s+zero,\s*[a-x0-9]+,\s*e8,\s*mf8,\s*tu,\s*m[au]} 2 { target { no-opts "-O0" no-opts "-O1" no-opts "-Os" no-opts "-Oz" no-opts "-g" no-opts "-funroll-loops" } } } } */ |
| 54 |
/* { dg-final { scan-assembler-times {vsetvli} 2 { target { no-opts "-O0" no-opts "-O1" no-opts "-Os" no-opts "-Oz" no-opts "-g" no-opts "-funroll-loops" } } } } */ |