| 1 |
/* Test the vmulxq_laneq_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 |
float32x4_t __attribute__ ((noinline)) |
| 11 |
test_vmulxq_laneq_f32_lane0 (float32x4_t vec1_1, float32x4_t vec1_2) |
| 12 |
{ |
| 13 |
return vmulxq_laneq_f32 (vec1_1, vec1_2, 0); |
| 14 |
} |
| 15 |
|
| 16 |
float32x4_t __attribute__ ((noinline)) |
| 17 |
test_vmulxq_laneq_f32_lane1 (float32x4_t vec1_1, float32x4_t vec1_2) |
| 18 |
{ |
| 19 |
return vmulxq_laneq_f32 (vec1_1, vec1_2, 1); |
| 20 |
} |
| 21 |
|
| 22 |
float32x4_t __attribute__ ((noinline)) |
| 23 |
test_vmulxq_laneq_f32_lane2 (float32x4_t vec1_1, float32x4_t vec1_2) |
| 24 |
{ |
| 25 |
return vmulxq_laneq_f32 (vec1_1, vec1_2, 2); |
| 26 |
} |
| 27 |
|
| 28 |
float32x4_t __attribute__ ((noinline)) |
| 29 |
test_vmulxq_laneq_f32_lane3 (float32x4_t vec1_1, float32x4_t vec1_2) |
| 30 |
{ |
| 31 |
return vmulxq_laneq_f32 (vec1_1, vec1_2, 3); |
| 32 |
} |
| 33 |
|
| 34 |
#define PASS_ARRAY(...) {__VA_ARGS__} |
| 35 |
|
| 36 |
#define SETUP_VEC(V1_D, V2_D, EXP0, EXP1, EXP2, EXP3, I) \ |
| 37 |
void set_and_test_case##I () \ |
| 38 |
{ \ |
| 39 |
int i; \ |
| 40 |
float32_t vec1_data[] = V1_D; \ |
| 41 |
float32x4_t vec1 = vld1q_f32 (vec1_data); \ |
| 42 |
float32_t vec2_data[] = V2_D; \ |
| 43 |
float32x4_t vec2 = vld1q_f32 (vec2_data); \ |
| 44 |
\ |
| 45 |
float32_t expected_lane0[] = EXP0; \ |
| 46 |
float32_t expected_lane1[] = EXP1; \ |
| 47 |
float32_t expected_lane2[] = EXP2; \ |
| 48 |
float32_t expected_lane3[] = EXP3; \ |
| 49 |
\ |
| 50 |
float32x4_t actual_lane0_v = \ |
| 51 |
test_vmulxq_laneq_f32_lane0 (vec1, vec2); \ |
| 52 |
float32_t actual_lane0[4]; \ |
| 53 |
vst1q_f32 (actual_lane0, actual_lane0_v); \ |
| 54 |
for (i = 0; i < 4; ++i) \ |
| 55 |
if (actual_lane0[i] != expected_lane0[i]) \ |
| 56 |
abort (); \ |
| 57 |
\ |
| 58 |
float32x4_t actual_lane1_v = \ |
| 59 |
test_vmulxq_laneq_f32_lane1 (vec1, vec2); \ |
| 60 |
float32_t actual_lane1[4]; \ |
| 61 |
vst1q_f32 (actual_lane1, actual_lane1_v); \ |
| 62 |
for (i = 0; i < 4; ++i) \ |
| 63 |
if (actual_lane1[i] != expected_lane1[i]) \ |
| 64 |
abort (); \ |
| 65 |
\ |
| 66 |
float32x4_t actual_lane2_v = \ |
| 67 |
test_vmulxq_laneq_f32_lane2 (vec1, vec2); \ |
| 68 |
float32_t actual_lane2[4]; \ |
| 69 |
vst1q_f32 (actual_lane2, actual_lane2_v); \ |
| 70 |
for (i = 0; i < 4; ++i) \ |
| 71 |
if (actual_lane2[i] != expected_lane2[i]) \ |
| 72 |
abort (); \ |
| 73 |
\ |
| 74 |
float32x4_t actual_lane3_v = \ |
| 75 |
test_vmulxq_laneq_f32_lane3 (vec1, vec2); \ |
| 76 |
float32_t actual_lane3[4]; \ |
| 77 |
vst1q_f32 (actual_lane3, actual_lane3_v); \ |
| 78 |
for (i = 0; i < 4; ++i) \ |
| 79 |
if (actual_lane3[i] != expected_lane3[i]) \ |
| 80 |
abort (); \ |
| 81 |
} \ |
| 82 |
|
| 83 |
float32_t v1 = 3.14159265359; |
| 84 |
float32_t v2 = 1.383894; |
| 85 |
float32_t v3 = -2.71828; |
| 86 |
float32_t v4 = -3.4891931; |
| 87 |
|
| 88 |
float32_t v5 = 0.0; |
| 89 |
float32_t v6 = -0.0; |
| 90 |
float32_t v7 = __builtin_huge_valf (); |
| 91 |
float32_t v8 = -__builtin_huge_valf (); |
| 92 |
|
| 93 |
float32_t spec = __builtin_huge_valf () * __builtin_huge_valf (); |
| 94 |
float32_t spec_n = -__builtin_huge_valf () * __builtin_huge_valf (); |
| 95 |
|
| 96 |
SETUP_VEC (PASS_ARRAY (v1, v2, v3, v4), PASS_ARRAY (v1, v2, v3, v4), |
| 97 |
PASS_ARRAY (v1*v1, v1*v2, v1*v3, v1*v4), |
| 98 |
PASS_ARRAY (v1*v2, v2*v2, v2*v3, v2*v4), |
| 99 |
PASS_ARRAY (v1*v3, v2*v3, v3*v3, v4*v3), |
| 100 |
PASS_ARRAY (v1*v4, v2*v4, v3*v4, v4*v4), 1) |
| 101 |
|
| 102 |
SETUP_VEC (PASS_ARRAY (v5, v6, v7, v8), PASS_ARRAY (v5, v6, v7, v8), |
| 103 |
PASS_ARRAY (0.0, -0.0, 2.0, -2.0), |
| 104 |
PASS_ARRAY (-0.0, 0.0, -2.0, 2.0), |
| 105 |
PASS_ARRAY (2.0, -2.0, spec, spec_n), |
| 106 |
PASS_ARRAY (-2.0, 2.0, spec_n, spec), 2) |
| 107 |
|
| 108 |
int |
| 109 |
main (void) |
| 110 |
{ |
| 111 |
set_and_test_case1 (); |
| 112 |
set_and_test_case2 (); |
| 113 |
return 0; |
| 114 |
} |
| 115 |
/* { dg-final { scan-assembler-times "fmulx\[ \t\]+\[vV\]\[0-9\]+\.4\[sS\], ?\[vV\]\[0-9\]+\.4\[sS\], ?\[vV\]\[0-9\]+\.\[sS\]\\\[0\\\]\n" 1 } } */ |
| 116 |
/* { dg-final { scan-assembler-times "fmulx\[ \t\]+\[vV\]\[0-9\]+\.4\[sS\], ?\[vV\]\[0-9\]+\.4\[sS\], ?\[vV\]\[0-9\]+\.\[sS\]\\\[1\\\]\n" 1 } } */ |
| 117 |
/* { dg-final { scan-assembler-times "fmulx\[ \t\]+\[vV\]\[0-9\]+\.4\[sS\], ?\[vV\]\[0-9\]+\.4\[sS\], ?\[vV\]\[0-9\]+\.\[sS\]\\\[2\\\]\n" 1 } } */ |
| 118 |
/* { dg-final { scan-assembler-times "fmulx\[ \t\]+\[vV\]\[0-9\]+\.4\[sS\], ?\[vV\]\[0-9\]+\.4\[sS\], ?\[vV\]\[0-9\]+\.\[sS\]\\\[3\\\]\n" 1 } } */ |