| 1 |
/* Test the vmulx_f32 AArch64 SIMD intrinsic. */ |
| 2 |
|
| 3 |
/* { dg-do run } */ |
| 4 |
/* { dg-options "-save-temps -O3" } */ |
| 5 |
|
| 6 |
#include "arm_neon.h" |
| 7 |
|
| 8 |
extern void abort (void); |
| 9 |
|
| 10 |
void __attribute__ ((noinline)) |
| 11 |
test_case (float32_t v1[2], float32_t v2[2], float32_t e[2]) |
| 12 |
{ |
| 13 |
int i; |
| 14 |
float32x2_t vec1_1 = vld1_f32 (v1); |
| 15 |
float32x2_t vec1_2 = vld1_f32 (v2); |
| 16 |
|
| 17 |
float32x2_t actual1 = vmulx_f32 (vec1_1, vec1_2); |
| 18 |
float32_t actual[2]; |
| 19 |
vst1_f32 (actual, actual1); |
| 20 |
|
| 21 |
for (i = 0; i < 2; ++i) |
| 22 |
if (actual[i] != e[i]) |
| 23 |
abort (); |
| 24 |
} |
| 25 |
|
| 26 |
int |
| 27 |
main (void) |
| 28 |
{ |
| 29 |
float32_t v1 = 3.14159265359; |
| 30 |
float32_t v2 = 1.383894; |
| 31 |
float32_t v3 = -2.71828; |
| 32 |
float32_t v4 = -3.4891931; |
| 33 |
|
| 34 |
float32_t v1_1[] = {v1, v2}; |
| 35 |
float32_t v1_2[] = {v3, v4}; |
| 36 |
float32_t e1[] = {v1 * v3, v2 * v4}; |
| 37 |
test_case (v1_1, v1_2, e1); |
| 38 |
|
| 39 |
float32_t v2_1[] = {0, -0.0}; |
| 40 |
float32_t v2_2[] = {__builtin_huge_valf (), __builtin_huge_valf ()}; |
| 41 |
float32_t e2[] = {2.0, -2.0}; |
| 42 |
test_case (v2_1, v2_2, e2); |
| 43 |
|
| 44 |
float32_t v3_1[] = {0, -0.0}; |
| 45 |
float32_t v3_2[] = {-__builtin_huge_valf (), -__builtin_huge_valf ()}; |
| 46 |
float32_t e3[] = {-2.0, 2.0}; |
| 47 |
test_case (v3_1, v3_2, e3); |
| 48 |
|
| 49 |
return 0; |
| 50 |
} |
| 51 |
|
| 52 |
/* { dg-final { scan-assembler-times "fmulx\[ \t\]+\[vV\]\[0-9\]+\.2s, ?\[vV\]\[0-9\]+\.2s, ?\[vV\]\[0-9\]+\.2s\n" 1 } } */ |