| 1 |
/* { dg-do compile } */ |
| 2 |
/* { dg-options "-mrvv-vector-bits=scalable -march=rv32gcv -mabi=ilp32 -fno-schedule-insns -fno-schedule-insns2" } */ |
| 3 |
|
| 4 |
#include "riscv_vector.h" |
| 5 |
|
| 6 |
void f(void *base, void *out, void *mask_in, |
| 7 |
size_t m, size_t n, size_t a, size_t b) { |
| 8 |
|
| 9 |
size_t vl = 666; |
| 10 |
for (size_t i = 0; i < m; i++) { |
| 11 |
if (i % 2 == 0) { |
| 12 |
for (size_t j = 0; j < n; j++){ |
| 13 |
if (j % 2 == 0) { |
| 14 |
vint8mf8_t v0 = __riscv_vle8_v_i8mf8(base + i + 200 + j, vl); |
| 15 |
__riscv_vse8_v_i8mf8 (out + i + 200, v0, vl); |
| 16 |
} else { |
| 17 |
vint8mf8_t v0 = __riscv_vle8_v_i8mf8(base + i + 300 + j, vl); |
| 18 |
__riscv_vse8_v_i8mf8 (out + i + 300, v0, vl); |
| 19 |
} |
| 20 |
} |
| 21 |
} else { |
| 22 |
for (size_t j = 0; j < vl; j++){ |
| 23 |
if (j % 2 == 0) { |
| 24 |
for (size_t k = 0; k < n; k++) { |
| 25 |
for (size_t i_a = 0; i_a < a; i_a++){ |
| 26 |
for (size_t i_b = 0; i_b < b; i_b++){ |
| 27 |
vint8mf8_t v0 = __riscv_vle8_v_i8mf8(base + i + 500 + k + j + i_a + i_b, vl); |
| 28 |
vint8mf8_t v1 = __riscv_vle8_v_i8mf8_tu(v0, base + i + 600 + k + j + i_a + i_b, vl); |
| 29 |
__riscv_vse8_v_i8mf8 (out + i + 600 + j + k + i_a + i_b, v1, vl); |
| 30 |
} |
| 31 |
} |
| 32 |
} |
| 33 |
} else { |
| 34 |
vint8mf8_t v0 = __riscv_vle8_v_i8mf8(base + i + 700, vl); |
| 35 |
__riscv_vse8_v_i8mf8 (out + i + 800, v0, vl); |
| 36 |
} |
| 37 |
} |
| 38 |
} |
| 39 |
} |
| 40 |
} |
| 41 |
|
| 42 |
/* { dg-final { scan-assembler-times {vsetvli\s+zero,\s*[a-x0-9]+,\s*e8,\s*mf8,\s*tu,\s*m[au]} 1 { target { no-opts "-O0" no-opts "-g" no-opts "-funroll-loops" } } } } */ |
| 43 |
/* { dg-final { scan-assembler-times {vsetvli} 1 { target { no-opts "-O0" no-opts "-g" no-opts "-funroll-loops" } } } } */ |