| 1 |
/* { dg-do compile } */ |
| 2 |
/* { dg-require-effective-target elf } */ |
| 3 |
/* { dg-options "-O2 -funwind-tables" } */ |
| 4 |
|
| 5 |
#include <arm_sve.h> |
| 6 |
|
| 7 |
#define T(F, TY1, TY2) \ |
| 8 |
void F##_f (TY1 *base, sv##TY2 data) \ |
| 9 |
{ \ |
| 10 |
return sv##F (svpfalse_b (), base, data); \ |
| 11 |
} |
| 12 |
|
| 13 |
#define D_INTEGER(F, TY) \ |
| 14 |
T (F##_s64, TY, int64_t) \ |
| 15 |
T (F##_u64, TY, uint64_t) |
| 16 |
|
| 17 |
#define SD_INTEGER(F, TY) \ |
| 18 |
D_INTEGER (F, TY) \ |
| 19 |
T (F##_s32, TY, int32_t) \ |
| 20 |
T (F##_u32, TY, uint32_t) |
| 21 |
|
| 22 |
#define HSD_INTEGER(F, TY) \ |
| 23 |
SD_INTEGER (F, TY) \ |
| 24 |
T (F##_s16, TY, int16_t) \ |
| 25 |
T (F##_u16, TY, uint16_t) |
| 26 |
|
| 27 |
#define ALL_DATA(F, A) \ |
| 28 |
T (F##_bf16, bfloat16_t, bfloat16##A) \ |
| 29 |
T (F##_f16, float16_t, float16##A) \ |
| 30 |
T (F##_f32, float32_t, float32##A) \ |
| 31 |
T (F##_f64, float64_t, float64##A) \ |
| 32 |
T (F##_s8, int8_t, int8##A) \ |
| 33 |
T (F##_s16, int16_t, int16##A) \ |
| 34 |
T (F##_s32, int32_t, int32##A) \ |
| 35 |
T (F##_s64, int64_t, int64##A) \ |
| 36 |
T (F##_u8, uint8_t, uint8##A) \ |
| 37 |
T (F##_u16, uint16_t, uint16##A) \ |
| 38 |
T (F##_u32, uint32_t, uint32##A) \ |
| 39 |
T (F##_u64, uint64_t, uint64##A) \ |
| 40 |
|
| 41 |
HSD_INTEGER (st1b, int8_t) |
| 42 |
SD_INTEGER (st1h, int16_t) |
| 43 |
D_INTEGER (st1w, int32_t) |
| 44 |
ALL_DATA (st1, _t) |
| 45 |
ALL_DATA (st2, x2_t) |
| 46 |
ALL_DATA (st3, x3_t) |
| 47 |
ALL_DATA (st4, x4_t) |
| 48 |
|
| 49 |
/* FIXME: Currently, st1/2/3/4 are not folded with a pfalse |
| 50 |
predicate, which is the reason for the 48 missing cases below. Once |
| 51 |
folding is implemented for these intrinsics, the sum should be 60. */ |
| 52 |
/* { dg-final { scan-assembler-times {\t.cfi_startproc\n\tret\n} 12 } } */ |
| 53 |
/* { dg-final { scan-assembler-times {\t.cfi_startproc\n} 60 } } */ |