| 1 |
/* { dg-require-effective-target arm_v8_1m_mve_ok } */ |
| 2 |
/* { dg-add-options arm_v8_1m_mve } */ |
| 3 |
/* { dg-additional-options "-O2" } */ |
| 4 |
|
| 5 |
#include "arm_mve.h" |
| 6 |
void |
| 7 |
foo (uint16_t row_x_col, int8_t *out) |
| 8 |
{ |
| 9 |
for (;;) |
| 10 |
{ |
| 11 |
int32x4_t out_3; |
| 12 |
int8_t *rhs_0; |
| 13 |
int8_t *lhs_3; |
| 14 |
int i_row_x_col; |
| 15 |
for (;i_row_x_col < row_x_col; i_row_x_col++) |
| 16 |
{ |
| 17 |
int32x4_t ker_0 = vldrbq_s32(rhs_0); |
| 18 |
int32x4_t ip_3 = vldrbq_s32(lhs_3); |
| 19 |
out_3 = vmulq_s32(ip_3, ker_0); |
| 20 |
} |
| 21 |
vstrbq_s32(out, out_3); |
| 22 |
} |
| 23 |
} |
| 24 |
|
| 25 |
void |
| 26 |
foo1 (uint16_t row_x_col, int8_t *out) |
| 27 |
{ |
| 28 |
for (;;) |
| 29 |
{ |
| 30 |
int16x8_t out_3; |
| 31 |
int8_t *rhs_0; |
| 32 |
int8_t *lhs_3; |
| 33 |
int i_row_x_col; |
| 34 |
for (; i_row_x_col < row_x_col; i_row_x_col++) |
| 35 |
{ |
| 36 |
int16x8_t ker_0 = vldrbq_s16(rhs_0); |
| 37 |
int16x8_t ip_3 = vldrbq_s16(lhs_3); |
| 38 |
out_3 = vmulq_s16(ip_3, ker_0); |
| 39 |
} |
| 40 |
vstrbq_s16(out, out_3); |
| 41 |
} |
| 42 |
} |
| 43 |
|
| 44 |
void |
| 45 |
foo2 (uint16_t row_x_col, int16_t *out) |
| 46 |
{ |
| 47 |
for (;;) |
| 48 |
{ |
| 49 |
int32x4_t out_3; |
| 50 |
int16_t *rhs_0; |
| 51 |
int16_t *lhs_3; |
| 52 |
int i_row_x_col; |
| 53 |
for (; i_row_x_col < row_x_col; i_row_x_col++) |
| 54 |
{ |
| 55 |
int32x4_t ker_0 = vldrhq_s32(rhs_0); |
| 56 |
int32x4_t ip_3 = vldrhq_s32(lhs_3); |
| 57 |
out_3 = vmulq_s32(ip_3, ker_0); |
| 58 |
} |
| 59 |
vstrhq_s32(out, out_3); |
| 60 |
} |
| 61 |
} |