| 1 |
#include <arm_neon.h> |
| 2 |
#include "arm-neon-ref.h" |
| 3 |
#include "compute-ref-data.h" |
| 4 |
#include <math.h> |
| 5 |
|
| 6 |
/* Expected results. */ |
| 7 |
VECT_VAR_DECL(expected,int,8,8) [] = { 0x11, 0x10, 0xf, 0xe, |
| 8 |
0xd, 0xc, 0xb, 0xa }; |
| 9 |
VECT_VAR_DECL(expected,int,16,4) [] = { 0x3, 0x2, 0x1, 0x0 }; |
| 10 |
VECT_VAR_DECL(expected,int,32,2) [] = { 0x18, 0x17 }; |
| 11 |
VECT_VAR_DECL(expected,uint,8,8) [] = { 0xef, 0xf0, 0xf1, 0xf2, |
| 12 |
0xf3, 0xf4, 0xf5, 0xf6 }; |
| 13 |
VECT_VAR_DECL(expected,uint,16,4) [] = { 0xffe3, 0xffe4, 0xffe5, 0xffe6 }; |
| 14 |
VECT_VAR_DECL(expected,uint,32,2) [] = { 0xffffffe8, 0xffffffe9 }; |
| 15 |
VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0x41c26666, 0x41ba6666 }; |
| 16 |
VECT_VAR_DECL(expected,int,8,16) [] = { 0x1a, 0x19, 0x18, 0x17, |
| 17 |
0x16, 0x15, 0x14, 0x13, |
| 18 |
0x12, 0x11, 0x10, 0xf, |
| 19 |
0xe, 0xd, 0xc, 0xb }; |
| 20 |
VECT_VAR_DECL(expected,int,16,8) [] = { 0x4, 0x3, 0x2, 0x1, |
| 21 |
0x0, 0x1, 0x2, 0x3 }; |
| 22 |
VECT_VAR_DECL(expected,int,32,4) [] = { 0x30, 0x2f, 0x2e, 0x2d }; |
| 23 |
VECT_VAR_DECL(expected,uint,8,16) [] = { 0xe6, 0xe7, 0xe8, 0xe9, |
| 24 |
0xea, 0xeb, 0xec, 0xed, |
| 25 |
0xee, 0xef, 0xf0, 0xf1, |
| 26 |
0xf2, 0xf3, 0xf4, 0xf5 }; |
| 27 |
VECT_VAR_DECL(expected,uint,16,8) [] = { 0xffe4, 0xffe5, 0xffe6, 0xffe7, |
| 28 |
0xffe8, 0xffe9, 0xffea, 0xffeb }; |
| 29 |
VECT_VAR_DECL(expected,uint,32,4) [] = { 0xffffffd0, 0xffffffd1, |
| 30 |
0xffffffd2, 0xffffffd3 }; |
| 31 |
VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0x42407ae1, 0x423c7ae1, |
| 32 |
0x42387ae1, 0x42347ae1 }; |
| 33 |
#if defined (__ARM_FEATURE_FP16_VECTOR_ARITHMETIC) |
| 34 |
VECT_VAR_DECL(expected, hfloat, 16, 4) [] = { 0x4e13, 0x4dd3, |
| 35 |
0x4d93, 0x4d53 }; |
| 36 |
VECT_VAR_DECL(expected, hfloat, 16, 8) [] = { 0x5204, 0x51e4, 0x51c4, 0x51a4, |
| 37 |
0x5184, 0x5164, 0x5144, 0x5124 }; |
| 38 |
#endif |
| 39 |
|
| 40 |
/* Additional expected results for float32 variants with specially |
| 41 |
chosen input values. */ |
| 42 |
VECT_VAR_DECL(expected_float32,hfloat,32,4) [] = { 0x0, 0x0, 0x0, 0x0 }; |
| 43 |
#if defined (__ARM_FEATURE_FP16_VECTOR_ARITHMETIC) |
| 44 |
VECT_VAR_DECL(expected_float16, hfloat, 16, 8) [] = { 0x0, 0x0, 0x0, 0x0, |
| 45 |
0x0, 0x0, 0x0, 0x0 }; |
| 46 |
#endif |
| 47 |
|
| 48 |
#define TEST_MSG "VABD/VABDQ" |
| 49 |
void exec_vabd (void) |
| 50 |
{ |
| 51 |
/* Basic test: v4=vabd(v1,v2), then store the result. */ |
| 52 |
#define TEST_VABD(Q, T1, T2, W, N) \ |
| 53 |
VECT_VAR(vector_res, T1, W, N) = \ |
| 54 |
vabd##Q##_##T2##W(VECT_VAR(vector1, T1, W, N), \ |
| 55 |
VECT_VAR(vector2, T1, W, N)); \ |
| 56 |
vst1##Q##_##T2##W(VECT_VAR(result, T1, W, N), VECT_VAR(vector_res, T1, W, N)) |
| 57 |
|
| 58 |
#define DECL_VABD_VAR(VAR) \ |
| 59 |
DECL_VARIABLE(VAR, int, 8, 8); \ |
| 60 |
DECL_VARIABLE(VAR, int, 16, 4); \ |
| 61 |
DECL_VARIABLE(VAR, int, 32, 2); \ |
| 62 |
DECL_VARIABLE(VAR, uint, 8, 8); \ |
| 63 |
DECL_VARIABLE(VAR, uint, 16, 4); \ |
| 64 |
DECL_VARIABLE(VAR, uint, 32, 2); \ |
| 65 |
DECL_VARIABLE(VAR, float, 32, 2); \ |
| 66 |
DECL_VARIABLE(VAR, int, 8, 16); \ |
| 67 |
DECL_VARIABLE(VAR, int, 16, 8); \ |
| 68 |
DECL_VARIABLE(VAR, int, 32, 4); \ |
| 69 |
DECL_VARIABLE(VAR, uint, 8, 16); \ |
| 70 |
DECL_VARIABLE(VAR, uint, 16, 8); \ |
| 71 |
DECL_VARIABLE(VAR, uint, 32, 4); \ |
| 72 |
DECL_VARIABLE(VAR, float, 32, 4) |
| 73 |
|
| 74 |
DECL_VABD_VAR(vector1); |
| 75 |
DECL_VABD_VAR(vector2); |
| 76 |
DECL_VABD_VAR(vector_res); |
| 77 |
|
| 78 |
#if defined (__ARM_FEATURE_FP16_VECTOR_ARITHMETIC) |
| 79 |
DECL_VARIABLE(vector1, float, 16, 4); |
| 80 |
DECL_VARIABLE(vector1, float, 16, 8); |
| 81 |
|
| 82 |
DECL_VARIABLE(vector2, float, 16, 4); |
| 83 |
DECL_VARIABLE(vector2, float, 16, 8); |
| 84 |
|
| 85 |
DECL_VARIABLE(vector_res, float, 16, 4); |
| 86 |
DECL_VARIABLE(vector_res, float, 16, 8); |
| 87 |
#endif |
| 88 |
|
| 89 |
clean_results (); |
| 90 |
|
| 91 |
/* Initialize input "vector1" from "buffer". */ |
| 92 |
VLOAD(vector1, buffer, , int, s, 8, 8); |
| 93 |
VLOAD(vector1, buffer, , int, s, 16, 4); |
| 94 |
VLOAD(vector1, buffer, , int, s, 32, 2); |
| 95 |
VLOAD(vector1, buffer, , uint, u, 8, 8); |
| 96 |
VLOAD(vector1, buffer, , uint, u, 16, 4); |
| 97 |
VLOAD(vector1, buffer, , uint, u, 32, 2); |
| 98 |
VLOAD(vector1, buffer, , float, f, 32, 2); |
| 99 |
VLOAD(vector1, buffer, q, int, s, 8, 16); |
| 100 |
VLOAD(vector1, buffer, q, int, s, 16, 8); |
| 101 |
VLOAD(vector1, buffer, q, int, s, 32, 4); |
| 102 |
VLOAD(vector1, buffer, q, uint, u, 8, 16); |
| 103 |
VLOAD(vector1, buffer, q, uint, u, 16, 8); |
| 104 |
VLOAD(vector1, buffer, q, uint, u, 32, 4); |
| 105 |
VLOAD(vector1, buffer, q, float, f, 32, 4); |
| 106 |
#if defined (__ARM_FEATURE_FP16_VECTOR_ARITHMETIC) |
| 107 |
VLOAD(vector1, buffer, , float, f, 16, 4); |
| 108 |
VLOAD(vector1, buffer, , float, f, 16, 4); |
| 109 |
VLOAD(vector1, buffer, q, float, f, 16, 8); |
| 110 |
VLOAD(vector1, buffer, q, float, f, 16, 8); |
| 111 |
#endif |
| 112 |
|
| 113 |
/* Choose init value arbitrarily. */ |
| 114 |
VDUP(vector2, , int, s, 8, 8, 1); |
| 115 |
VDUP(vector2, , int, s, 16, 4, -13); |
| 116 |
VDUP(vector2, , int, s, 32, 2, 8); |
| 117 |
VDUP(vector2, , uint, u, 8, 8, 1); |
| 118 |
VDUP(vector2, , uint, u, 16, 4, 13); |
| 119 |
VDUP(vector2, , uint, u, 32, 2, 8); |
| 120 |
VDUP(vector2, , float, f, 32, 2, 8.3f); |
| 121 |
VDUP(vector2, q, int, s, 8, 16, 10); |
| 122 |
VDUP(vector2, q, int, s, 16, 8, -12); |
| 123 |
VDUP(vector2, q, int, s, 32, 4, 32); |
| 124 |
VDUP(vector2, q, uint, u, 8, 16, 10); |
| 125 |
VDUP(vector2, q, uint, u, 16, 8, 12); |
| 126 |
VDUP(vector2, q, uint, u, 32, 4, 32); |
| 127 |
VDUP(vector2, q, float, f, 32, 4, 32.12f); |
| 128 |
#if defined (__ARM_FEATURE_FP16_VECTOR_ARITHMETIC) |
| 129 |
VDUP(vector2, , float, f, 16, 4, 8.3f); |
| 130 |
VDUP(vector2, q, float, f, 16, 8, 32.12f); |
| 131 |
#endif |
| 132 |
|
| 133 |
/* Execute the tests. */ |
| 134 |
TEST_VABD(, int, s, 8, 8); |
| 135 |
TEST_VABD(, int, s, 16, 4); |
| 136 |
TEST_VABD(, int, s, 32, 2); |
| 137 |
TEST_VABD(, uint, u, 8, 8); |
| 138 |
TEST_VABD(, uint, u, 16, 4); |
| 139 |
TEST_VABD(, uint, u, 32, 2); |
| 140 |
TEST_VABD(, float, f, 32, 2); |
| 141 |
TEST_VABD(q, int, s, 8, 16); |
| 142 |
TEST_VABD(q, int, s, 16, 8); |
| 143 |
TEST_VABD(q, int, s, 32, 4); |
| 144 |
TEST_VABD(q, uint, u, 8, 16); |
| 145 |
TEST_VABD(q, uint, u, 16, 8); |
| 146 |
TEST_VABD(q, uint, u, 32, 4); |
| 147 |
TEST_VABD(q, float, f, 32, 4); |
| 148 |
|
| 149 |
#if defined (__ARM_FEATURE_FP16_VECTOR_ARITHMETIC) |
| 150 |
TEST_VABD(, float, f, 16, 4); |
| 151 |
TEST_VABD(q, float, f, 16, 8); |
| 152 |
#endif |
| 153 |
|
| 154 |
CHECK(TEST_MSG, int, 8, 8, PRIx8, expected, ""); |
| 155 |
CHECK(TEST_MSG, int, 16, 4, PRIx16, expected, ""); |
| 156 |
CHECK(TEST_MSG, int, 32, 2, PRIx32, expected, ""); |
| 157 |
CHECK(TEST_MSG, uint, 8, 8, PRIx8, expected, ""); |
| 158 |
CHECK(TEST_MSG, uint, 16, 4, PRIx16, expected, ""); |
| 159 |
CHECK(TEST_MSG, uint, 32, 2, PRIx32, expected, ""); |
| 160 |
CHECK_FP(TEST_MSG, float, 32, 2, PRIx32, expected, ""); |
| 161 |
CHECK(TEST_MSG, int, 8, 16, PRIx8, expected, ""); |
| 162 |
CHECK(TEST_MSG, int, 16, 8, PRIx16, expected, ""); |
| 163 |
CHECK(TEST_MSG, int, 32, 4, PRIx32, expected, ""); |
| 164 |
CHECK(TEST_MSG, uint, 8, 16, PRIx8, expected, ""); |
| 165 |
CHECK(TEST_MSG, uint, 16, 8, PRIx16, expected, ""); |
| 166 |
CHECK(TEST_MSG, uint, 32, 4, PRIx32, expected, ""); |
| 167 |
CHECK_FP(TEST_MSG, float, 32, 4, PRIx32, expected, ""); |
| 168 |
#if defined (__ARM_FEATURE_FP16_VECTOR_ARITHMETIC) |
| 169 |
CHECK_FP(TEST_MSG, float, 16, 4, PRIx16, expected, ""); |
| 170 |
CHECK_FP(TEST_MSG, float, 16, 8, PRIx16, expected, ""); |
| 171 |
#endif |
| 172 |
|
| 173 |
/* Extra FP tests with special values (-0.0, ....) */ |
| 174 |
VDUP(vector1, q, float, f, 32, 4, -0.0f); |
| 175 |
VDUP(vector2, q, float, f, 32, 4, 0.0); |
| 176 |
TEST_VABD(q, float, f, 32, 4); |
| 177 |
CHECK_FP(TEST_MSG, float, 32, 4, PRIx32, expected_float32, " FP special (-0.0)"); |
| 178 |
|
| 179 |
#if defined (__ARM_FEATURE_FP16_VECTOR_ARITHMETIC) |
| 180 |
VDUP(vector1, q, float, f, 16, 8, -0.0f); |
| 181 |
VDUP(vector2, q, float, f, 16, 8, 0.0); |
| 182 |
TEST_VABD(q, float, f, 16, 8); |
| 183 |
CHECK_FP(TEST_MSG, float, 16, 8, PRIx16, expected_float16, |
| 184 |
" FP special (-0.0)"); |
| 185 |
#endif |
| 186 |
|
| 187 |
/* Extra FP tests with special values (-0.0, ....) */ |
| 188 |
VDUP(vector1, q, float, f, 32, 4, 0.0f); |
| 189 |
VDUP(vector2, q, float, f, 32, 4, -0.0); |
| 190 |
TEST_VABD(q, float, f, 32, 4); |
| 191 |
CHECK_FP(TEST_MSG, float, 32, 4, PRIx32, expected_float32, " FP special (-0.0)"); |
| 192 |
|
| 193 |
#if defined (__ARM_FEATURE_FP16_VECTOR_ARITHMETIC) |
| 194 |
VDUP(vector1, q, float, f, 16, 8, 0.0f); |
| 195 |
VDUP(vector2, q, float, f, 16, 8, -0.0); |
| 196 |
TEST_VABD(q, float, f, 16, 8); |
| 197 |
CHECK_FP(TEST_MSG, float, 16, 8, PRIx16, expected_float16, |
| 198 |
" FP special (-0.0)"); |
| 199 |
#endif |
| 200 |
} |
| 201 |
|
| 202 |
int main (void) |
| 203 |
{ |
| 204 |
exec_vabd (); |
| 205 |
return 0; |
| 206 |
} |