| 1 |
#include <arm_neon.h> |
| 2 |
#include "arm-neon-ref.h" |
| 3 |
#include "compute-ref-data.h" |
| 4 |
|
| 5 |
/* Expected results. */ |
| 6 |
VECT_VAR_DECL(expected,int,8,8) [] = { 0xf1, 0xf1, 0xf1, 0xf1, |
| 7 |
0xf1, 0xf1, 0xf1, 0xf1 }; |
| 8 |
VECT_VAR_DECL(expected,int,16,4) [] = { 0xfff2, 0xfff2, 0xfff2, 0xfff2 }; |
| 9 |
VECT_VAR_DECL(expected,int,32,2) [] = { 0xfffffff1, 0xfffffff1 }; |
| 10 |
VECT_VAR_DECL(expected,int,64,1) [] = { 0xfffffffffffffff0 }; |
| 11 |
VECT_VAR_DECL(expected,uint,8,8) [] = { 0xf7, 0xf7, 0xf7, 0xf7, |
| 12 |
0xf7, 0xf7, 0xf7, 0xf7 }; |
| 13 |
VECT_VAR_DECL(expected,uint,16,4) [] = { 0xfff3, 0xfff3, 0xfff3, 0xfff3 }; |
| 14 |
VECT_VAR_DECL(expected,uint,32,2) [] = { 0xfffffff1, 0xfffffff1 }; |
| 15 |
VECT_VAR_DECL(expected,uint,64,1) [] = { 0xfffffffffffffff0 }; |
| 16 |
VECT_VAR_DECL(expected,poly,8,8) [] = { 0xf7, 0xf7, 0xf7, 0xf7, |
| 17 |
0xf7, 0xf7, 0xf7, 0xf7 }; |
| 18 |
VECT_VAR_DECL(expected,poly,16,4) [] = { 0xfff3, 0xfff3, 0xfff3, 0xfff3 }; |
| 19 |
#if MFLOAT8_SUPPORTED |
| 20 |
VECT_VAR_DECL(expected,hmfloat,8,8) [] = { 0xf6, 0xf6, 0xf6, 0xf6, |
| 21 |
0xf6, 0xf6, 0xf6, 0xf6 }; |
| 22 |
#endif |
| 23 |
VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0xc1700000, 0xc1700000 }; |
| 24 |
#if defined (FP16_SUPPORTED) |
| 25 |
VECT_VAR_DECL (expected, hfloat, 16, 4) [] = { 0xca80, 0xca80, |
| 26 |
0xca80, 0xca80 }; |
| 27 |
#endif |
| 28 |
VECT_VAR_DECL(expected,int,8,16) [] = { 0xf2, 0xf2, 0xf2, 0xf2, |
| 29 |
0xf2, 0xf2, 0xf2, 0xf2, |
| 30 |
0xf2, 0xf2, 0xf2, 0xf2, |
| 31 |
0xf2, 0xf2, 0xf2, 0xf2 }; |
| 32 |
VECT_VAR_DECL(expected,int,16,8) [] = { 0xfff3, 0xfff3, 0xfff3, 0xfff3, |
| 33 |
0xfff3, 0xfff3, 0xfff3, 0xfff3 }; |
| 34 |
VECT_VAR_DECL(expected,int,32,4) [] = { 0xfffffff1, 0xfffffff1, |
| 35 |
0xfffffff1, 0xfffffff1 }; |
| 36 |
VECT_VAR_DECL(expected,int,64,2) [] = { 0xfffffffffffffff0, |
| 37 |
0xfffffffffffffff0 }; |
| 38 |
VECT_VAR_DECL(expected,uint,8,16) [] = { 0xf5, 0xf5, 0xf5, 0xf5, |
| 39 |
0xf5, 0xf5, 0xf5, 0xf5, |
| 40 |
0xf5, 0xf5, 0xf5, 0xf5, |
| 41 |
0xf5, 0xf5, 0xf5, 0xf5 }; |
| 42 |
VECT_VAR_DECL(expected,uint,16,8) [] = { 0xfff1, 0xfff1, 0xfff1, 0xfff1, |
| 43 |
0xfff1, 0xfff1, 0xfff1, 0xfff1 }; |
| 44 |
VECT_VAR_DECL(expected,uint,32,4) [] = { 0xfffffff0, 0xfffffff0, |
| 45 |
0xfffffff0, 0xfffffff0 }; |
| 46 |
VECT_VAR_DECL(expected,uint,64,2) [] = { 0xfffffffffffffff0, |
| 47 |
0xfffffffffffffff0 }; |
| 48 |
VECT_VAR_DECL(expected,poly,8,16) [] = { 0xf5, 0xf5, 0xf5, 0xf5, |
| 49 |
0xf5, 0xf5, 0xf5, 0xf5, |
| 50 |
0xf5, 0xf5, 0xf5, 0xf5, |
| 51 |
0xf5, 0xf5, 0xf5, 0xf5 }; |
| 52 |
VECT_VAR_DECL(expected,poly,16,8) [] = { 0xfff1, 0xfff1, 0xfff1, 0xfff1, |
| 53 |
0xfff1, 0xfff1, 0xfff1, 0xfff1 }; |
| 54 |
#if MFLOAT8_SUPPORTED |
| 55 |
VECT_VAR_DECL(expected,hmfloat,8,16) [] = { 0xf7, 0xf7, 0xf7, 0xf7, |
| 56 |
0xf7, 0xf7, 0xf7, 0xf7, |
| 57 |
0xf7, 0xf7, 0xf7, 0xf7, |
| 58 |
0xf7, 0xf7, 0xf7, 0xf7 }; |
| 59 |
#endif |
| 60 |
#if defined (FP16_SUPPORTED) |
| 61 |
VECT_VAR_DECL (expected, hfloat, 16, 8) [] = { 0xca80, 0xca80, |
| 62 |
0xca80, 0xca80, |
| 63 |
0xca80, 0xca80, |
| 64 |
0xca80, 0xca80 }; |
| 65 |
#endif |
| 66 |
VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0xc1700000, 0xc1700000, |
| 67 |
0xc1700000, 0xc1700000 }; |
| 68 |
|
| 69 |
#define TEST_MSG "VDUP_LANE/VDUPQ_LANE" |
| 70 |
void exec_vdup_lane (void) |
| 71 |
{ |
| 72 |
/* Basic test: vec1=vdup_lane(vec2, lane), then store the result. */ |
| 73 |
#define TEST_VDUP_LANE(Q, T1, T2, W, N, N2, L) \ |
| 74 |
VECT_VAR(vector_res, T1, W, N) = \ |
| 75 |
vdup##Q##_lane_##T2##W(VECT_VAR(vector, T1, W, N2), L); \ |
| 76 |
vst1##Q##_##T2##W(VECT_VAR(result, T1, W, N), VECT_VAR(vector_res, T1, W, N)) |
| 77 |
|
| 78 |
/* Input vector can only have 64 bits. */ |
| 79 |
DECL_VARIABLE_64BITS_VARIANTS(vector); |
| 80 |
|
| 81 |
DECL_VARIABLE_ALL_VARIANTS(vector_res); |
| 82 |
|
| 83 |
clean_results (); |
| 84 |
|
| 85 |
TEST_MACRO_64BITS_VARIANTS_2_5(VLOAD, vector, buffer); |
| 86 |
MFLOAT8_ONLY(VLOAD(vector, buffer, , mfloat, mf, 8, 8);) |
| 87 |
#if defined (FP16_SUPPORTED) |
| 88 |
VLOAD(vector, buffer, , float, f, 16, 4); |
| 89 |
#endif |
| 90 |
VLOAD(vector, buffer, , float, f, 32, 2); |
| 91 |
|
| 92 |
/* Choose lane arbitrarily. */ |
| 93 |
TEST_VDUP_LANE(, int, s, 8, 8, 8, 1); |
| 94 |
TEST_VDUP_LANE(, int, s, 16, 4, 4, 2); |
| 95 |
TEST_VDUP_LANE(, int, s, 32, 2, 2, 1); |
| 96 |
TEST_VDUP_LANE(, int, s, 64, 1, 1, 0); |
| 97 |
TEST_VDUP_LANE(, uint, u, 8, 8, 8, 7); |
| 98 |
TEST_VDUP_LANE(, uint, u, 16, 4, 4, 3); |
| 99 |
TEST_VDUP_LANE(, uint, u, 32, 2, 2, 1); |
| 100 |
TEST_VDUP_LANE(, uint, u, 64, 1, 1, 0); |
| 101 |
TEST_VDUP_LANE(, poly, p, 8, 8, 8, 7); |
| 102 |
TEST_VDUP_LANE(, poly, p, 16, 4, 4, 3); |
| 103 |
MFLOAT8_ONLY(TEST_VDUP_LANE(, mfloat, mf, 8, 8, 8, 6);) |
| 104 |
#if defined (FP16_SUPPORTED) |
| 105 |
TEST_VDUP_LANE(, float, f, 16, 4, 4, 3); |
| 106 |
#endif |
| 107 |
TEST_VDUP_LANE(, float, f, 32, 2, 2, 1); |
| 108 |
|
| 109 |
TEST_VDUP_LANE(q, int, s, 8, 16, 8, 2); |
| 110 |
TEST_VDUP_LANE(q, int, s, 16, 8, 4, 3); |
| 111 |
TEST_VDUP_LANE(q, int, s, 32, 4, 2, 1); |
| 112 |
TEST_VDUP_LANE(q, int, s, 64, 2, 1, 0); |
| 113 |
TEST_VDUP_LANE(q, uint, u, 8, 16, 8, 5); |
| 114 |
TEST_VDUP_LANE(q, uint, u, 16, 8, 4, 1); |
| 115 |
TEST_VDUP_LANE(q, uint, u, 32, 4, 2, 0); |
| 116 |
TEST_VDUP_LANE(q, uint, u, 64, 2, 1, 0); |
| 117 |
TEST_VDUP_LANE(q, poly, p, 8, 16, 8, 5); |
| 118 |
TEST_VDUP_LANE(q, poly, p, 16, 8, 4, 1); |
| 119 |
MFLOAT8_ONLY(TEST_VDUP_LANE(q, mfloat, mf, 8, 16, 8, 7);) |
| 120 |
#if defined (FP16_SUPPORTED) |
| 121 |
TEST_VDUP_LANE(q, float, f, 16, 8, 4, 3); |
| 122 |
#endif |
| 123 |
TEST_VDUP_LANE(q, float, f, 32, 4, 2, 1); |
| 124 |
|
| 125 |
#if defined (FP16_SUPPORTED) |
| 126 |
CHECK_RESULTS (TEST_MSG, ""); |
| 127 |
#else |
| 128 |
CHECK_RESULTS_NO_FP16 (TEST_MSG, ""); |
| 129 |
#endif |
| 130 |
|
| 131 |
#if defined (__aarch64__) |
| 132 |
|
| 133 |
#undef TEST_MSG |
| 134 |
#define TEST_MSG "VDUP_LANEQ/VDUPQ_LANEQ" |
| 135 |
|
| 136 |
/* Expected results for vdup*_laneq tests. */ |
| 137 |
VECT_VAR_DECL(expected2,int,8,8) [] = { 0xfd, 0xfd, 0xfd, 0xfd, |
| 138 |
0xfd, 0xfd, 0xfd, 0xfd }; |
| 139 |
VECT_VAR_DECL(expected2,int,16,4) [] = { 0xfff2, 0xfff2, 0xfff2, 0xfff2 }; |
| 140 |
VECT_VAR_DECL(expected2,int,32,2) [] = { 0xfffffff1, 0xfffffff1 }; |
| 141 |
VECT_VAR_DECL(expected2,int,64,1) [] = { 0xfffffffffffffff0 }; |
| 142 |
VECT_VAR_DECL(expected2,uint,8,8) [] = { 0xff, 0xff, 0xff, 0xff, |
| 143 |
0xff, 0xff, 0xff, 0xff }; |
| 144 |
VECT_VAR_DECL(expected2,uint,16,4) [] = { 0xfff3, 0xfff3, 0xfff3, 0xfff3 }; |
| 145 |
VECT_VAR_DECL(expected2,uint,32,2) [] = { 0xfffffff1, 0xfffffff1 }; |
| 146 |
VECT_VAR_DECL(expected2,uint,64,1) [] = { 0xfffffffffffffff0 }; |
| 147 |
VECT_VAR_DECL(expected2,poly,8,8) [] = { 0xf7, 0xf7, 0xf7, 0xf7, |
| 148 |
0xf7, 0xf7, 0xf7, 0xf7 }; |
| 149 |
VECT_VAR_DECL(expected2,poly,16,4) [] = { 0xfff3, 0xfff3, 0xfff3, 0xfff3 }; |
| 150 |
#if MFLOAT8_SUPPORTED |
| 151 |
VECT_VAR_DECL(expected2,hmfloat,8,8) [] = { 0xfb, 0xfb, 0xfb, 0xfb, |
| 152 |
0xfb, 0xfb, 0xfb, 0xfb }; |
| 153 |
#endif |
| 154 |
VECT_VAR_DECL(expected2,hfloat,32,2) [] = { 0xc1700000, 0xc1700000 }; |
| 155 |
#if defined (FP16_SUPPORTED) |
| 156 |
VECT_VAR_DECL (expected2, hfloat, 16, 4) [] = { 0xca80, 0xca80, |
| 157 |
0xca80, 0xca80 }; |
| 158 |
#endif |
| 159 |
VECT_VAR_DECL(expected2,int,8,16) [] = { 0xfb, 0xfb, 0xfb, 0xfb, |
| 160 |
0xfb, 0xfb, 0xfb, 0xfb, |
| 161 |
0xfb, 0xfb, 0xfb, 0xfb, |
| 162 |
0xfb, 0xfb, 0xfb, 0xfb }; |
| 163 |
VECT_VAR_DECL(expected2,int,16,8) [] = { 0xfff7, 0xfff7, 0xfff7, 0xfff7, |
| 164 |
0xfff7, 0xfff7, 0xfff7, 0xfff7 }; |
| 165 |
VECT_VAR_DECL(expected2,int,32,4) [] = { 0xfffffff1, 0xfffffff1, |
| 166 |
0xfffffff1, 0xfffffff1 }; |
| 167 |
VECT_VAR_DECL(expected2,int,64,2) [] = { 0xfffffffffffffff0, |
| 168 |
0xfffffffffffffff0 }; |
| 169 |
VECT_VAR_DECL(expected2,uint,8,16) [] = { 0xf5, 0xf5, 0xf5, 0xf5, |
| 170 |
0xf5, 0xf5, 0xf5, 0xf5, |
| 171 |
0xf5, 0xf5, 0xf5, 0xf5, |
| 172 |
0xf5, 0xf5, 0xf5, 0xf5 }; |
| 173 |
VECT_VAR_DECL(expected2,uint,16,8) [] = { 0xfff1, 0xfff1, 0xfff1, 0xfff1, |
| 174 |
0xfff1, 0xfff1, 0xfff1, 0xfff1 }; |
| 175 |
VECT_VAR_DECL(expected2,uint,32,4) [] = { 0xfffffff0, 0xfffffff0, |
| 176 |
0xfffffff0, 0xfffffff0 }; |
| 177 |
VECT_VAR_DECL(expected2,uint,64,2) [] = { 0xfffffffffffffff0, |
| 178 |
0xfffffffffffffff0 }; |
| 179 |
VECT_VAR_DECL(expected2,poly,8,16) [] = { 0xf5, 0xf5, 0xf5, 0xf5, |
| 180 |
0xf5, 0xf5, 0xf5, 0xf5, |
| 181 |
0xf5, 0xf5, 0xf5, 0xf5, |
| 182 |
0xf5, 0xf5, 0xf5, 0xf5 }; |
| 183 |
VECT_VAR_DECL(expected2,poly,16,8) [] = { 0xfff1, 0xfff1, 0xfff1, 0xfff1, |
| 184 |
0xfff1, 0xfff1, 0xfff1, 0xfff1 }; |
| 185 |
#if MFLOAT8_SUPPORTED |
| 186 |
VECT_VAR_DECL(expected2,hmfloat,8,16) [] = { 0xfc, 0xfc, 0xfc, 0xfc, |
| 187 |
0xfc, 0xfc, 0xfc, 0xfc, |
| 188 |
0xfc, 0xfc, 0xfc, 0xfc, |
| 189 |
0xfc, 0xfc, 0xfc, 0xfc }; |
| 190 |
#endif |
| 191 |
#if defined (FP16_SUPPORTED) |
| 192 |
VECT_VAR_DECL (expected2, hfloat, 16, 8) [] = { 0xc880, 0xc880, |
| 193 |
0xc880, 0xc880, |
| 194 |
0xc880, 0xc880, |
| 195 |
0xc880, 0xc880 }; |
| 196 |
#endif |
| 197 |
VECT_VAR_DECL(expected2,hfloat,32,4) [] = { 0xc1700000, 0xc1700000, |
| 198 |
0xc1700000, 0xc1700000 }; |
| 199 |
|
| 200 |
/* Clean all results for vdup*_laneq tests. */ |
| 201 |
clean_results (); |
| 202 |
/* Basic test: vec1=vdup_lane(vec2, lane), then store the result. */ |
| 203 |
#define TEST_VDUP_LANEQ(Q, T1, T2, W, N, N2, L) \ |
| 204 |
VECT_VAR(vector_res, T1, W, N) = \ |
| 205 |
vdup##Q##_laneq_##T2##W(VECT_VAR(vector, T1, W, N2), L); \ |
| 206 |
vst1##Q##_##T2##W(VECT_VAR(result, T1, W, N), VECT_VAR(vector_res, T1, W, N)) |
| 207 |
|
| 208 |
/* Input vector can only have 64 bits. */ |
| 209 |
DECL_VARIABLE_128BITS_VARIANTS(vector); |
| 210 |
|
| 211 |
clean_results (); |
| 212 |
|
| 213 |
TEST_MACRO_128BITS_VARIANTS_2_5(VLOAD, vector, buffer); |
| 214 |
MFLOAT8_ONLY(VLOAD(vector, buffer, q, mfloat, mf, 8, 16);) |
| 215 |
#if defined (FP16_SUPPORTED) |
| 216 |
VLOAD(vector, buffer, q, float, f, 16, 8); |
| 217 |
#endif |
| 218 |
VLOAD(vector, buffer, q, float, f, 32, 4); |
| 219 |
|
| 220 |
/* Choose lane arbitrarily. */ |
| 221 |
TEST_VDUP_LANEQ(, int, s, 8, 8, 16, 13); |
| 222 |
TEST_VDUP_LANEQ(, int, s, 16, 4, 8, 2); |
| 223 |
TEST_VDUP_LANEQ(, int, s, 32, 2, 4, 1); |
| 224 |
TEST_VDUP_LANEQ(, int, s, 64, 1, 2, 0); |
| 225 |
TEST_VDUP_LANEQ(, uint, u, 8, 8, 16, 15); |
| 226 |
TEST_VDUP_LANEQ(, uint, u, 16, 4, 8, 3); |
| 227 |
TEST_VDUP_LANEQ(, uint, u, 32, 2, 4, 1); |
| 228 |
TEST_VDUP_LANEQ(, uint, u, 64, 1, 2, 0); |
| 229 |
TEST_VDUP_LANEQ(, poly, p, 8, 8, 16, 7); |
| 230 |
TEST_VDUP_LANEQ(, poly, p, 16, 4, 8, 3); |
| 231 |
MFLOAT8_ONLY(TEST_VDUP_LANEQ(, mfloat, mf, 8, 8, 16, 11);) |
| 232 |
#if defined (FP16_SUPPORTED) |
| 233 |
TEST_VDUP_LANEQ(, float, f, 16, 4, 8, 3); |
| 234 |
#endif |
| 235 |
TEST_VDUP_LANEQ(, float, f, 32, 2, 4, 1); |
| 236 |
|
| 237 |
TEST_VDUP_LANEQ(q, int, s, 8, 16, 16, 11); |
| 238 |
TEST_VDUP_LANEQ(q, int, s, 16, 8, 8, 7); |
| 239 |
TEST_VDUP_LANEQ(q, int, s, 32, 4, 4, 1); |
| 240 |
TEST_VDUP_LANEQ(q, int, s, 64, 2, 2, 0); |
| 241 |
TEST_VDUP_LANEQ(q, uint, u, 8, 16, 16, 5); |
| 242 |
TEST_VDUP_LANEQ(q, uint, u, 16, 8, 8, 1); |
| 243 |
TEST_VDUP_LANEQ(q, uint, u, 32, 4, 4, 0); |
| 244 |
TEST_VDUP_LANEQ(q, uint, u, 64, 2, 2, 0); |
| 245 |
TEST_VDUP_LANEQ(q, poly, p, 8, 16, 16, 5); |
| 246 |
TEST_VDUP_LANEQ(q, poly, p, 16, 8, 8, 1); |
| 247 |
MFLOAT8_ONLY(TEST_VDUP_LANEQ(q, mfloat, mf, 8, 16, 16, 12);) |
| 248 |
#if defined (FP16_SUPPORTED) |
| 249 |
TEST_VDUP_LANEQ(q, float, f, 16, 8, 8, 7); |
| 250 |
#endif |
| 251 |
TEST_VDUP_LANEQ(q, float, f, 32, 4, 4, 1); |
| 252 |
|
| 253 |
CHECK_RESULTS_NAMED (TEST_MSG, expected2, ""); |
| 254 |
#if defined (FP16_SUPPORTED) |
| 255 |
CHECK_FP(TEST_MSG, float, 16, 4, PRIx16, expected2, ""); |
| 256 |
CHECK_FP(TEST_MSG, float, 16, 8, PRIx16, expected2, ""); |
| 257 |
#endif |
| 258 |
|
| 259 |
#endif /* __aarch64__. */ |
| 260 |
} |
| 261 |
|
| 262 |
int main (void) |
| 263 |
{ |
| 264 |
exec_vdup_lane (); |
| 265 |
return 0; |
| 266 |
} |