| 1 |
#include <arm_neon.h> |
| 2 |
#include "arm-neon-ref.h" |
| 3 |
#include "compute-ref-data.h" |
| 4 |
|
| 5 |
/* Expected results. */ |
| 6 |
|
| 7 |
/* vld2/chunk 0. */ |
| 8 |
VECT_VAR_DECL(expected_vld2_0,int,8,8) [] = { 0xaa, 0xaa, 0xaa, 0xaa, |
| 9 |
0xaa, 0xaa, 0xaa, 0xaa }; |
| 10 |
VECT_VAR_DECL(expected_vld2_0,int,16,4) [] = { 0xaaaa, 0xaaaa, 0xaaaa, 0xaaaa }; |
| 11 |
VECT_VAR_DECL(expected_vld2_0,int,32,2) [] = { 0xfffffff0, 0xfffffff1 }; |
| 12 |
VECT_VAR_DECL(expected_vld2_0,uint,8,8) [] = { 0xaa, 0xaa, 0xaa, 0xaa, |
| 13 |
0xaa, 0xaa, 0xaa, 0xaa }; |
| 14 |
VECT_VAR_DECL(expected_vld2_0,uint,16,4) [] = { 0xaaaa, 0xaaaa, |
| 15 |
0xaaaa, 0xaaaa }; |
| 16 |
VECT_VAR_DECL(expected_vld2_0,uint,32,2) [] = { 0xaaaaaaaa, 0xaaaaaaaa }; |
| 17 |
VECT_VAR_DECL(expected_vld2_0,poly,8,8) [] = { 0xaa, 0xaa, 0xaa, 0xaa, |
| 18 |
0xaa, 0xaa, 0xaa, 0xaa }; |
| 19 |
VECT_VAR_DECL(expected_vld2_0,poly,16,4) [] = { 0xaaaa, 0xaaaa, |
| 20 |
0xaaaa, 0xaaaa }; |
| 21 |
#if MFLOAT8_SUPPORTED |
| 22 |
VECT_VAR_DECL(expected_vld2_0,hmfloat,8,8) [] = { 0xaa, 0xaa, 0xaa, 0xaa, |
| 23 |
0xaa, 0xaa, 0xaa, 0xaa }; |
| 24 |
#endif |
| 25 |
VECT_VAR_DECL(expected_vld2_0,hfloat,16,4) [] = { 0xaaaa, 0xaaaa, 0xaaaa, 0xaaaa }; |
| 26 |
VECT_VAR_DECL(expected_vld2_0,hfloat,32,2) [] = { 0xc1800000, 0xc1700000 }; |
| 27 |
VECT_VAR_DECL(expected_vld2_0,int,16,8) [] = { 0xaaaa, 0xaaaa, 0xaaaa, 0xaaaa, |
| 28 |
0xaaaa, 0xaaaa, 0xaaaa, 0xaaaa }; |
| 29 |
VECT_VAR_DECL(expected_vld2_0,int,32,4) [] = { 0xaaaaaaaa, 0xaaaaaaaa, |
| 30 |
0xaaaaaaaa, 0xaaaaaaaa }; |
| 31 |
VECT_VAR_DECL(expected_vld2_0,uint,16,8) [] = { 0xaaaa, 0xaaaa, 0xaaaa, 0xaaaa, |
| 32 |
0xaaaa, 0xaaaa, 0xaaaa, 0xaaaa }; |
| 33 |
VECT_VAR_DECL(expected_vld2_0,uint,32,4) [] = { 0xfffffff0, 0xfffffff1, |
| 34 |
0xaaaaaaaa, 0xaaaaaaaa }; |
| 35 |
VECT_VAR_DECL(expected_vld2_0,poly,16,8) [] = { 0xaaaa, 0xaaaa, 0xaaaa, 0xaaaa, |
| 36 |
0xaaaa, 0xaaaa, 0xaaaa, 0xaaaa }; |
| 37 |
VECT_VAR_DECL(expected_vld2_0,hfloat,16,8) [] = { 0xaaaa, 0xaaaa, 0xaaaa, 0xaaaa, |
| 38 |
0xaaaa, 0xaaaa, 0xaaaa, 0xaaaa } ; |
| 39 |
VECT_VAR_DECL(expected_vld2_0,hfloat,32,4) [] = { 0xaaaaaaaa, 0xaaaaaaaa, |
| 40 |
0xaaaaaaaa, 0xaaaaaaaa }; |
| 41 |
|
| 42 |
/* vld2/chunk 1. */ |
| 43 |
VECT_VAR_DECL(expected_vld2_1,int,8,8) [] = { 0xaa, 0xaa, 0xaa, 0xaa, |
| 44 |
0xaa, 0xaa, 0xf0, 0xf1 }; |
| 45 |
VECT_VAR_DECL(expected_vld2_1,int,16,4) [] = { 0xfff0, 0xfff1, 0xaaaa, 0xaaaa }; |
| 46 |
VECT_VAR_DECL(expected_vld2_1,int,32,2) [] = { 0xaaaaaaaa, 0xaaaaaaaa }; |
| 47 |
VECT_VAR_DECL(expected_vld2_1,uint,8,8) [] = { 0xf0, 0xf1, 0xaa, 0xaa, |
| 48 |
0xaa, 0xaa, 0xaa, 0xaa }; |
| 49 |
VECT_VAR_DECL(expected_vld2_1,uint,16,4) [] = { 0xaaaa, 0xaaaa, 0xfff0, 0xfff1 }; |
| 50 |
VECT_VAR_DECL(expected_vld2_1,uint,32,2) [] = { 0xfffffff0, 0xfffffff1 }; |
| 51 |
VECT_VAR_DECL(expected_vld2_1,poly,8,8) [] = { 0xf0, 0xf1, 0xaa, 0xaa, |
| 52 |
0xaa, 0xaa, 0xaa, 0xaa }; |
| 53 |
VECT_VAR_DECL(expected_vld2_1,poly,16,4) [] = { 0xaaaa, 0xaaaa, 0xfff0, 0xfff1 }; |
| 54 |
#if MFLOAT8_SUPPORTED |
| 55 |
VECT_VAR_DECL(expected_vld2_1,hmfloat,8,8) [] = { 0xaa, 0xaa, 0xaa, 0xaa, |
| 56 |
0xaa, 0xaa, 0xf0, 0xf1 }; |
| 57 |
#endif |
| 58 |
VECT_VAR_DECL(expected_vld2_1,hfloat,16,4) [] = { 0xcc00, 0xcb80, 0xaaaa, 0xaaaa }; |
| 59 |
VECT_VAR_DECL(expected_vld2_1,hfloat,32,2) [] = { 0xaaaaaaaa, 0xaaaaaaaa }; |
| 60 |
VECT_VAR_DECL(expected_vld2_1,int,16,8) [] = { 0xaaaa, 0xaaaa, 0xaaaa, 0xaaaa, |
| 61 |
0xfff0, 0xfff1, 0xaaaa, 0xaaaa }; |
| 62 |
VECT_VAR_DECL(expected_vld2_1,int,32,4) [] = { 0xfffffff0, 0xfffffff1, |
| 63 |
0xaaaaaaaa, 0xaaaaaaaa }; |
| 64 |
VECT_VAR_DECL(expected_vld2_1,uint,16,8) [] = { 0xaaaa, 0xaaaa, 0xfff0, 0xfff1, |
| 65 |
0xaaaa, 0xaaaa, 0xaaaa, 0xaaaa }; |
| 66 |
VECT_VAR_DECL(expected_vld2_1,uint,32,4) [] = { 0xaaaaaaaa, 0xaaaaaaaa, |
| 67 |
0xaaaaaaaa, 0xaaaaaaaa }; |
| 68 |
VECT_VAR_DECL(expected_vld2_1,poly,16,8) [] = { 0xaaaa, 0xaaaa, 0xfff0, 0xfff1, |
| 69 |
0xaaaa, 0xaaaa, 0xaaaa, 0xaaaa }; |
| 70 |
VECT_VAR_DECL(expected_vld2_1,hfloat,16,8) [] = { 0xaaaa, 0xaaaa, 0xaaaa, 0xaaaa, |
| 71 |
0xcc00, 0xcb80, 0xaaaa, 0xaaaa }; |
| 72 |
VECT_VAR_DECL(expected_vld2_1,hfloat,32,4) [] = { 0xc1800000, 0xc1700000, |
| 73 |
0xaaaaaaaa, 0xaaaaaaaa }; |
| 74 |
|
| 75 |
/* vld3/chunk 0. */ |
| 76 |
VECT_VAR_DECL(expected_vld3_0,int,8,8) [] = { 0xaa, 0xaa, 0xaa, 0xaa, |
| 77 |
0xaa, 0xaa, 0xaa, 0xaa }; |
| 78 |
VECT_VAR_DECL(expected_vld3_0,int,16,4) [] = { 0xaaaa, 0xaaaa, 0xaaaa, 0xaaaa }; |
| 79 |
VECT_VAR_DECL(expected_vld3_0,int,32,2) [] = { 0xfffffff0, 0xfffffff1 }; |
| 80 |
VECT_VAR_DECL(expected_vld3_0,uint,8,8) [] = { 0xaa, 0xaa, 0xaa, 0xaa, |
| 81 |
0xaa, 0xaa, 0xaa, 0xaa }; |
| 82 |
VECT_VAR_DECL(expected_vld3_0,uint,16,4) [] = { 0xaaaa, 0xaaaa, 0xaaaa, 0xaaaa }; |
| 83 |
VECT_VAR_DECL(expected_vld3_0,uint,32,2) [] = { 0xaaaaaaaa, 0xaaaaaaaa }; |
| 84 |
VECT_VAR_DECL(expected_vld3_0,poly,8,8) [] = { 0xaa, 0xaa, 0xaa, 0xaa, |
| 85 |
0xaa, 0xaa, 0xaa, 0xaa }; |
| 86 |
VECT_VAR_DECL(expected_vld3_0,poly,16,4) [] = { 0xaaaa, 0xaaaa, 0xaaaa, 0xaaaa }; |
| 87 |
#if MFLOAT8_SUPPORTED |
| 88 |
VECT_VAR_DECL(expected_vld3_0,hmfloat,8,8) [] = { 0xaa, 0xaa, 0xaa, 0xaa, |
| 89 |
0xaa, 0xaa, 0xaa, 0xaa }; |
| 90 |
#endif |
| 91 |
VECT_VAR_DECL(expected_vld3_0,hfloat,16,4) [] = { 0xaaaa, 0xaaaa, 0xaaaa, 0xaaaa }; |
| 92 |
VECT_VAR_DECL(expected_vld3_0,hfloat,32,2) [] = { 0xc1800000, 0xc1700000 }; |
| 93 |
VECT_VAR_DECL(expected_vld3_0,int,16,8) [] = { 0xaaaa, 0xaaaa, 0xaaaa, 0xaaaa, |
| 94 |
0xaaaa, 0xaaaa, 0xaaaa, 0xaaaa }; |
| 95 |
VECT_VAR_DECL(expected_vld3_0,int,32,4) [] = { 0xaaaaaaaa, 0xaaaaaaaa, |
| 96 |
0xaaaaaaaa, 0xaaaaaaaa }; |
| 97 |
VECT_VAR_DECL(expected_vld3_0,uint,16,8) [] = { 0xaaaa, 0xaaaa, 0xaaaa, 0xaaaa, |
| 98 |
0xaaaa, 0xaaaa, 0xaaaa, 0xaaaa }; |
| 99 |
VECT_VAR_DECL(expected_vld3_0,uint,32,4) [] = { 0xfffffff0, 0xfffffff1, |
| 100 |
0xfffffff2, 0xaaaaaaaa }; |
| 101 |
VECT_VAR_DECL(expected_vld3_0,poly,16,8) [] = { 0xaaaa, 0xaaaa, 0xaaaa, 0xaaaa, |
| 102 |
0xaaaa, 0xaaaa, 0xaaaa, 0xaaaa }; |
| 103 |
VECT_VAR_DECL(expected_vld3_0,hfloat,16,8) [] = { 0xaaaa, 0xaaaa, 0xaaaa, 0xaaaa, |
| 104 |
0xaaaa, 0xaaaa, 0xaaaa, 0xaaaa }; |
| 105 |
VECT_VAR_DECL(expected_vld3_0,hfloat,32,4) [] = { 0xaaaaaaaa, 0xaaaaaaaa, |
| 106 |
0xaaaaaaaa, 0xaaaaaaaa }; |
| 107 |
|
| 108 |
/* vld3/chunk 1. */ |
| 109 |
VECT_VAR_DECL(expected_vld3_1,int,8,8) [] = { 0xaa, 0xaa, 0xaa, 0xaa, |
| 110 |
0xaa, 0xaa, 0xaa, 0xaa }; |
| 111 |
VECT_VAR_DECL(expected_vld3_1,int,16,4) [] = { 0xaaaa, 0xaaaa, 0xfff0, 0xfff1 }; |
| 112 |
VECT_VAR_DECL(expected_vld3_1,int,32,2) [] = { 0xfffffff2, 0xaaaaaaaa }; |
| 113 |
VECT_VAR_DECL(expected_vld3_1,uint,8,8) [] = { 0xaa, 0xaa, 0xaa, 0xaa, |
| 114 |
0xf0, 0xf1, 0xf2, 0xaa }; |
| 115 |
VECT_VAR_DECL(expected_vld3_1,uint,16,4) [] = { 0xaaaa, 0xaaaa, 0xaaaa, 0xaaaa }; |
| 116 |
VECT_VAR_DECL(expected_vld3_1,uint,32,2) [] = { 0xaaaaaaaa, 0xfffffff0 }; |
| 117 |
VECT_VAR_DECL(expected_vld3_1,poly,8,8) [] = { 0xaa, 0xaa, 0xaa, 0xaa, |
| 118 |
0xf0, 0xf1, 0xf2, 0xaa }; |
| 119 |
VECT_VAR_DECL(expected_vld3_1,poly,16,4) [] = { 0xaaaa, 0xaaaa, 0xaaaa, 0xaaaa }; |
| 120 |
#if MFLOAT8_SUPPORTED |
| 121 |
VECT_VAR_DECL(expected_vld3_1,hmfloat,8,8) [] = { 0xaa, 0xaa, 0xaa, 0xaa, |
| 122 |
0xaa, 0xaa, 0xaa, 0xaa }; |
| 123 |
#endif |
| 124 |
VECT_VAR_DECL(expected_vld3_1,hfloat,16,4) [] = { 0xaaaa, 0xaaaa, 0xcc00, 0xcb80 }; |
| 125 |
VECT_VAR_DECL(expected_vld3_1,hfloat,32,2) [] = { 0xc1600000, 0xaaaaaaaa }; |
| 126 |
VECT_VAR_DECL(expected_vld3_1,int,16,8) [] = { 0xaaaa, 0xaaaa, 0xaaaa, 0xaaaa, |
| 127 |
0xaaaa, 0xaaaa, 0xaaaa, 0xaaaa }; |
| 128 |
VECT_VAR_DECL(expected_vld3_1,int,32,4) [] = { 0xaaaaaaaa, 0xaaaaaaaa, |
| 129 |
0xfffffff0, 0xfffffff1 }; |
| 130 |
VECT_VAR_DECL(expected_vld3_1,uint,16,8) [] = { 0xaaaa, 0xaaaa, 0xaaaa, 0xaaaa, |
| 131 |
0xaaaa, 0xaaaa, 0xaaaa, 0xfff0 }; |
| 132 |
VECT_VAR_DECL(expected_vld3_1,uint,32,4) [] = { 0xaaaaaaaa, 0xaaaaaaaa, |
| 133 |
0xaaaaaaaa, 0xaaaaaaaa }; |
| 134 |
VECT_VAR_DECL(expected_vld3_1,poly,16,8) [] = { 0xaaaa, 0xaaaa, 0xaaaa, 0xaaaa, |
| 135 |
0xaaaa, 0xaaaa, 0xaaaa, 0xfff0 }; |
| 136 |
VECT_VAR_DECL(expected_vld3_1,hfloat,16,8) [] = { 0xaaaa, 0xaaaa, 0xaaaa, 0xaaaa, |
| 137 |
0xaaaa, 0xaaaa, 0xaaaa, 0xaaaa }; |
| 138 |
VECT_VAR_DECL(expected_vld3_1,hfloat,32,4) [] = { 0xaaaaaaaa, 0xaaaaaaaa, |
| 139 |
0xc1800000, 0xc1700000 }; |
| 140 |
|
| 141 |
/* vld3/chunk 2. */ |
| 142 |
VECT_VAR_DECL(expected_vld3_2,int,8,8) [] = { 0xaa, 0xaa, 0xaa, 0xaa, |
| 143 |
0xaa, 0xf0, 0xf1, 0xf2 }; |
| 144 |
VECT_VAR_DECL(expected_vld3_2,int,16,4) [] = { 0xfff2, 0xaaaa, 0xaaaa, 0xaaaa }; |
| 145 |
VECT_VAR_DECL(expected_vld3_2,int,32,2) [] = { 0xaaaaaaaa, 0xaaaaaaaa }; |
| 146 |
VECT_VAR_DECL(expected_vld3_2,uint,8,8) [] = { 0xaa, 0xaa, 0xaa, 0xaa, |
| 147 |
0xaa, 0xaa, 0xaa, 0xaa }; |
| 148 |
VECT_VAR_DECL(expected_vld3_2,uint,16,4) [] = { 0xaaaa, 0xfff0, 0xfff1, 0xfff2 }; |
| 149 |
VECT_VAR_DECL(expected_vld3_2,uint,32,2) [] = { 0xfffffff1, 0xfffffff2 }; |
| 150 |
VECT_VAR_DECL(expected_vld3_2,poly,8,8) [] = { 0xaa, 0xaa, 0xaa, 0xaa, |
| 151 |
0xaa, 0xaa, 0xaa, 0xaa }; |
| 152 |
VECT_VAR_DECL(expected_vld3_2,poly,16,4) [] = { 0xaaaa, 0xfff0, 0xfff1, 0xfff2 }; |
| 153 |
#if MFLOAT8_SUPPORTED |
| 154 |
VECT_VAR_DECL(expected_vld3_2,hmfloat,8,8) [] = { 0xaa, 0xaa, 0xaa, 0xaa, |
| 155 |
0xaa, 0xf0, 0xf1, 0xf2 }; |
| 156 |
#endif |
| 157 |
VECT_VAR_DECL(expected_vld3_2,hfloat,16,4) [] = { 0xcb00, 0xaaaa, 0xaaaa, 0xaaaa }; |
| 158 |
VECT_VAR_DECL(expected_vld3_2,hfloat,32,2) [] = { 0xaaaaaaaa, 0xaaaaaaaa }; |
| 159 |
VECT_VAR_DECL(expected_vld3_2,int,16,8) [] = { 0xaaaa, 0xaaaa, 0xfff0, 0xfff1, |
| 160 |
0xfff2, 0xaaaa, 0xaaaa, 0xaaaa }; |
| 161 |
VECT_VAR_DECL(expected_vld3_2,int,32,4) [] = { 0xfffffff2, 0xaaaaaaaa, |
| 162 |
0xaaaaaaaa, 0xaaaaaaaa }; |
| 163 |
VECT_VAR_DECL(expected_vld3_2,uint,16,8) [] = { 0xfff1, 0xfff2, 0xaaaa, 0xaaaa, |
| 164 |
0xaaaa, 0xaaaa, 0xaaaa, 0xaaaa }; |
| 165 |
VECT_VAR_DECL(expected_vld3_2,uint,32,4) [] = { 0xaaaaaaaa, 0xaaaaaaaa, |
| 166 |
0xaaaaaaaa, 0xaaaaaaaa }; |
| 167 |
VECT_VAR_DECL(expected_vld3_2,poly,16,8) [] = { 0xfff1, 0xfff2, 0xaaaa, 0xaaaa, |
| 168 |
0xaaaa, 0xaaaa, 0xaaaa, 0xaaaa }; |
| 169 |
VECT_VAR_DECL(expected_vld3_2,hfloat,16,8) [] = { 0xaaaa, 0xaaaa, 0xcc00, 0xcb80, |
| 170 |
0xcb00, 0xaaaa, 0xaaaa, 0xaaaa }; |
| 171 |
VECT_VAR_DECL(expected_vld3_2,hfloat,32,4) [] = { 0xc1600000, 0xaaaaaaaa, |
| 172 |
0xaaaaaaaa, 0xaaaaaaaa }; |
| 173 |
|
| 174 |
/* vld4/chunk 0. */ |
| 175 |
VECT_VAR_DECL(expected_vld4_0,int,8,8) [] = { 0xaa, 0xaa, 0xaa, 0xaa, |
| 176 |
0xaa, 0xaa, 0xaa, 0xaa }; |
| 177 |
VECT_VAR_DECL(expected_vld4_0,int,16,4) [] = { 0xaaaa, 0xaaaa, 0xaaaa, 0xaaaa }; |
| 178 |
VECT_VAR_DECL(expected_vld4_0,int,32,2) [] = { 0xfffffff0, 0xfffffff1 }; |
| 179 |
VECT_VAR_DECL(expected_vld4_0,uint,8,8) [] = { 0xaa, 0xaa, 0xaa, 0xaa, |
| 180 |
0xaa, 0xaa, 0xaa, 0xaa }; |
| 181 |
VECT_VAR_DECL(expected_vld4_0,uint,16,4) [] = { 0xaaaa, 0xaaaa, 0xaaaa, 0xaaaa }; |
| 182 |
VECT_VAR_DECL(expected_vld4_0,uint,32,2) [] = { 0xaaaaaaaa, 0xaaaaaaaa }; |
| 183 |
VECT_VAR_DECL(expected_vld4_0,poly,8,8) [] = { 0xaa, 0xaa, 0xaa, 0xaa, |
| 184 |
0xaa, 0xaa, 0xaa, 0xaa }; |
| 185 |
VECT_VAR_DECL(expected_vld4_0,poly,16,4) [] = { 0xaaaa, 0xaaaa, 0xaaaa, 0xaaaa }; |
| 186 |
VECT_VAR_DECL(expected_vld4_0,hfloat,16,4) [] = { 0xaaaa, 0xaaaa, 0xaaaa, 0xaaaa }; |
| 187 |
#if MFLOAT8_SUPPORTED |
| 188 |
VECT_VAR_DECL(expected_vld4_0,hmfloat,8,8) [] = { 0xaa, 0xaa, 0xaa, 0xaa, |
| 189 |
0xaa, 0xaa, 0xaa, 0xaa }; |
| 190 |
#endif |
| 191 |
VECT_VAR_DECL(expected_vld4_0,hfloat,32,2) [] = { 0xc1800000, 0xc1700000 }; |
| 192 |
VECT_VAR_DECL(expected_vld4_0,int,16,8) [] = { 0xaaaa, 0xaaaa, 0xaaaa, 0xaaaa, |
| 193 |
0xaaaa, 0xaaaa, 0xaaaa, 0xaaaa }; |
| 194 |
VECT_VAR_DECL(expected_vld4_0,int,32,4) [] = { 0xaaaaaaaa, 0xaaaaaaaa, |
| 195 |
0xaaaaaaaa, 0xaaaaaaaa }; |
| 196 |
VECT_VAR_DECL(expected_vld4_0,uint,16,8) [] = { 0xaaaa, 0xaaaa, 0xaaaa, 0xaaaa, |
| 197 |
0xaaaa, 0xaaaa, 0xaaaa, 0xaaaa }; |
| 198 |
VECT_VAR_DECL(expected_vld4_0,uint,32,4) [] = { 0xfffffff0, 0xfffffff1, |
| 199 |
0xfffffff2, 0xfffffff3 }; |
| 200 |
VECT_VAR_DECL(expected_vld4_0,poly,16,8) [] = { 0xaaaa, 0xaaaa, 0xaaaa, 0xaaaa, |
| 201 |
0xaaaa, 0xaaaa, 0xaaaa, 0xaaaa }; |
| 202 |
VECT_VAR_DECL(expected_vld4_0,hfloat,16,8) [] = { 0xaaaa, 0xaaaa, 0xaaaa, 0xaaaa, |
| 203 |
0xaaaa, 0xaaaa, 0xaaaa, 0xaaaa }; |
| 204 |
VECT_VAR_DECL(expected_vld4_0,hfloat,32,4) [] = { 0xaaaaaaaa, 0xaaaaaaaa, |
| 205 |
0xaaaaaaaa, 0xaaaaaaaa }; |
| 206 |
|
| 207 |
/* vld4/chunk 1. */ |
| 208 |
VECT_VAR_DECL(expected_vld4_1,int,8,8) [] = { 0xaa, 0xaa, 0xaa, 0xaa, |
| 209 |
0xaa, 0xaa, 0xaa, 0xaa }; |
| 210 |
VECT_VAR_DECL(expected_vld4_1,int,16,4) [] = { 0xaaaa, 0xaaaa, 0xaaaa, 0xaaaa }; |
| 211 |
VECT_VAR_DECL(expected_vld4_1,int,32,2) [] = { 0xfffffff2, 0xfffffff3 }; |
| 212 |
VECT_VAR_DECL(expected_vld4_1,uint,8,8) [] = { 0xaa, 0xaa, 0xaa, 0xaa, |
| 213 |
0xaa, 0xaa, 0xaa, 0xaa }; |
| 214 |
VECT_VAR_DECL(expected_vld4_1,uint,16,4) [] = { 0xaaaa, 0xaaaa, 0xaaaa, 0xaaaa }; |
| 215 |
VECT_VAR_DECL(expected_vld4_1,uint,32,2) [] = { 0xaaaaaaaa, 0xaaaaaaaa }; |
| 216 |
VECT_VAR_DECL(expected_vld4_1,poly,8,8) [] = { 0xaa, 0xaa, 0xaa, 0xaa, |
| 217 |
0xaa, 0xaa, 0xaa, 0xaa }; |
| 218 |
VECT_VAR_DECL(expected_vld4_1,poly,16,4) [] = { 0xaaaa, 0xaaaa, 0xaaaa, 0xaaaa }; |
| 219 |
#if MFLOAT8_SUPPORTED |
| 220 |
VECT_VAR_DECL(expected_vld4_1,hmfloat,8,8) [] = { 0xaa, 0xaa, 0xaa, 0xaa, |
| 221 |
0xaa, 0xaa, 0xaa, 0xaa }; |
| 222 |
#endif |
| 223 |
VECT_VAR_DECL(expected_vld4_1,hfloat,16,4) [] = { 0xaaaa, 0xaaaa, 0xaaaa, 0xaaaa }; |
| 224 |
VECT_VAR_DECL(expected_vld4_1,hfloat,32,2) [] = { 0xc1600000, 0xc1500000 }; |
| 225 |
VECT_VAR_DECL(expected_vld4_1,int,16,8) [] = { 0xaaaa, 0xaaaa, 0xaaaa, 0xaaaa, |
| 226 |
0xaaaa, 0xaaaa, 0xaaaa, 0xaaaa }; |
| 227 |
VECT_VAR_DECL(expected_vld4_1,int,32,4) [] = { 0xaaaaaaaa, 0xaaaaaaaa, |
| 228 |
0xaaaaaaaa, 0xaaaaaaaa }; |
| 229 |
VECT_VAR_DECL(expected_vld4_1,uint,16,8) [] = { 0xaaaa, 0xaaaa, 0xaaaa, 0xaaaa, |
| 230 |
0xaaaa, 0xaaaa, 0xaaaa, 0xaaaa }; |
| 231 |
VECT_VAR_DECL(expected_vld4_1,uint,32,4) [] = { 0xaaaaaaaa, 0xaaaaaaaa, |
| 232 |
0xaaaaaaaa, 0xaaaaaaaa }; |
| 233 |
VECT_VAR_DECL(expected_vld4_1,poly,16,8) [] = { 0xaaaa, 0xaaaa, 0xaaaa, 0xaaaa, |
| 234 |
0xaaaa, 0xaaaa, 0xaaaa, 0xaaaa }; |
| 235 |
VECT_VAR_DECL(expected_vld4_1,hfloat,16,8) [] = { 0xaaaa, 0xaaaa, 0xaaaa, 0xaaaa, |
| 236 |
0xaaaa, 0xaaaa, 0xaaaa, 0xaaaa }; |
| 237 |
VECT_VAR_DECL(expected_vld4_1,hfloat,32,4) [] = { 0xaaaaaaaa, 0xaaaaaaaa, |
| 238 |
0xaaaaaaaa, 0xaaaaaaaa }; |
| 239 |
|
| 240 |
/* vld4/chunk 2. */ |
| 241 |
VECT_VAR_DECL(expected_vld4_2,int,8,8) [] = { 0xaa, 0xaa, 0xaa, 0xaa, |
| 242 |
0xaa, 0xaa, 0xaa, 0xaa }; |
| 243 |
VECT_VAR_DECL(expected_vld4_2,int,16,4) [] = { 0xfff0, 0xfff1, 0xfff2, 0xfff3 }; |
| 244 |
VECT_VAR_DECL(expected_vld4_2,int,32,2) [] = { 0xaaaaaaaa, 0xaaaaaaaa }; |
| 245 |
VECT_VAR_DECL(expected_vld4_2,uint,8,8) [] = { 0xf0, 0xf1, 0xf2, 0xf3, |
| 246 |
0xaa, 0xaa, 0xaa, 0xaa }; |
| 247 |
VECT_VAR_DECL(expected_vld4_2,uint,16,4) [] = { 0xaaaa, 0xaaaa, 0xaaaa, 0xaaaa }; |
| 248 |
VECT_VAR_DECL(expected_vld4_2,uint,32,2) [] = { 0xfffffff0, 0xfffffff1 }; |
| 249 |
VECT_VAR_DECL(expected_vld4_2,poly,8,8) [] = { 0xf0, 0xf1, 0xf2, 0xf3, |
| 250 |
0xaa, 0xaa, 0xaa, 0xaa }; |
| 251 |
VECT_VAR_DECL(expected_vld4_2,poly,16,4) [] = { 0xaaaa, 0xaaaa, 0xaaaa, 0xaaaa }; |
| 252 |
#if MFLOAT8_SUPPORTED |
| 253 |
VECT_VAR_DECL(expected_vld4_2,hmfloat,8,8) [] = { 0xaa, 0xaa, 0xaa, 0xaa, |
| 254 |
0xaa, 0xaa, 0xaa, 0xaa }; |
| 255 |
#endif |
| 256 |
VECT_VAR_DECL(expected_vld4_2,hfloat,16,4) [] = { 0xcc00, 0xcb80, 0xcb00, 0xca80 }; |
| 257 |
VECT_VAR_DECL(expected_vld4_2,hfloat,32,2) [] = { 0xaaaaaaaa, 0xaaaaaaaa }; |
| 258 |
VECT_VAR_DECL(expected_vld4_2,int,16,8) [] = { 0xaaaa, 0xaaaa, 0xaaaa, 0xaaaa, |
| 259 |
0xaaaa, 0xaaaa, 0xaaaa, 0xaaaa }; |
| 260 |
VECT_VAR_DECL(expected_vld4_2,int,32,4) [] = { 0xfffffff0, 0xfffffff1, |
| 261 |
0xfffffff2, 0xfffffff3 }; |
| 262 |
VECT_VAR_DECL(expected_vld4_2,uint,16,8) [] = { 0xaaaa, 0xaaaa, 0xaaaa, 0xaaaa, |
| 263 |
0xfff0, 0xfff1, 0xfff2, 0xfff3 }; |
| 264 |
VECT_VAR_DECL(expected_vld4_2,uint,32,4) [] = { 0xaaaaaaaa, 0xaaaaaaaa, |
| 265 |
0xaaaaaaaa, 0xaaaaaaaa }; |
| 266 |
VECT_VAR_DECL(expected_vld4_2,poly,16,8) [] = { 0xaaaa, 0xaaaa, 0xaaaa, 0xaaaa, |
| 267 |
0xfff0, 0xfff1, 0xfff2, 0xfff3 }; |
| 268 |
VECT_VAR_DECL(expected_vld4_2,hfloat,16,8) [] = { 0xaaaa, 0xaaaa, 0xaaaa, 0xaaaa, |
| 269 |
0xaaaa, 0xaaaa, 0xaaaa, 0xaaaa }; |
| 270 |
VECT_VAR_DECL(expected_vld4_2,hfloat,32,4) [] = { 0xc1800000, 0xc1700000, |
| 271 |
0xc1600000, 0xc1500000 }; |
| 272 |
|
| 273 |
/* vld4/chunk 3. */ |
| 274 |
VECT_VAR_DECL(expected_vld4_3,int,8,8) [] = { 0xaa, 0xaa, 0xaa, 0xaa, |
| 275 |
0xf0, 0xf1, 0xf2, 0xf3 }; |
| 276 |
VECT_VAR_DECL(expected_vld4_3,int,16,4) [] = { 0xaaaa, 0xaaaa, 0xaaaa, 0xaaaa }; |
| 277 |
VECT_VAR_DECL(expected_vld4_3,int,32,2) [] = { 0xaaaaaaaa, 0xaaaaaaaa }; |
| 278 |
VECT_VAR_DECL(expected_vld4_3,uint,8,8) [] = { 0xaa, 0xaa, 0xaa, 0xaa, |
| 279 |
0xaa, 0xaa, 0xaa, 0xaa }; |
| 280 |
VECT_VAR_DECL(expected_vld4_3,uint,16,4) [] = { 0xfff0, 0xfff1, 0xfff2, 0xfff3 }; |
| 281 |
VECT_VAR_DECL(expected_vld4_3,uint,32,2) [] = { 0xfffffff2, 0xfffffff3 }; |
| 282 |
VECT_VAR_DECL(expected_vld4_3,poly,8,8) [] = { 0xaa, 0xaa, 0xaa, 0xaa, |
| 283 |
0xaa, 0xaa, 0xaa, 0xaa }; |
| 284 |
VECT_VAR_DECL(expected_vld4_3,poly,16,4) [] = { 0xfff0, 0xfff1, 0xfff2, 0xfff3 }; |
| 285 |
#if MFLOAT8_SUPPORTED |
| 286 |
VECT_VAR_DECL(expected_vld4_3,hmfloat,8,8) [] = { 0xaa, 0xaa, 0xaa, 0xaa, |
| 287 |
0xf0, 0xf1, 0xf2, 0xf3 }; |
| 288 |
#endif |
| 289 |
VECT_VAR_DECL(expected_vld4_3,hfloat,16,4) [] = { 0xaaaa, 0xaaaa, 0xaaaa, 0xaaaa }; |
| 290 |
VECT_VAR_DECL(expected_vld4_3,hfloat,32,2) [] = { 0xaaaaaaaa, 0xaaaaaaaa }; |
| 291 |
VECT_VAR_DECL(expected_vld4_3,int,16,8) [] = { 0xfff0, 0xfff1, 0xfff2, 0xfff3, |
| 292 |
0xaaaa, 0xaaaa, 0xaaaa, 0xaaaa }; |
| 293 |
VECT_VAR_DECL(expected_vld4_3,int,32,4) [] = { 0xaaaaaaaa, 0xaaaaaaaa, |
| 294 |
0xaaaaaaaa, 0xaaaaaaaa }; |
| 295 |
VECT_VAR_DECL(expected_vld4_3,uint,16,8) [] = { 0xaaaa, 0xaaaa, 0xaaaa, 0xaaaa, |
| 296 |
0xaaaa, 0xaaaa, 0xaaaa, 0xaaaa }; |
| 297 |
VECT_VAR_DECL(expected_vld4_3,uint,32,4) [] = { 0xaaaaaaaa, 0xaaaaaaaa, |
| 298 |
0xaaaaaaaa, 0xaaaaaaaa }; |
| 299 |
VECT_VAR_DECL(expected_vld4_3,poly,16,8) [] = { 0xaaaa, 0xaaaa, 0xaaaa, 0xaaaa, |
| 300 |
0xaaaa, 0xaaaa, 0xaaaa, 0xaaaa }; |
| 301 |
VECT_VAR_DECL(expected_vld4_3,hfloat,16,8) [] = { 0xcc00, 0xcb80, 0xcb00, 0xca80, |
| 302 |
0xaaaa, 0xaaaa, 0xaaaa, 0xaaaa }; |
| 303 |
VECT_VAR_DECL(expected_vld4_3,hfloat,32,4) [] = { 0xaaaaaaaa, 0xaaaaaaaa, |
| 304 |
0xaaaaaaaa, 0xaaaaaaaa }; |
| 305 |
|
| 306 |
/* Declare additional input buffers as needed. */ |
| 307 |
/* Input buffers for vld2_lane */ |
| 308 |
VECT_VAR_DECL_INIT(buffer_vld2_lane, int, 8, 2); |
| 309 |
VECT_VAR_DECL_INIT(buffer_vld2_lane, int, 16, 2); |
| 310 |
VECT_VAR_DECL_INIT(buffer_vld2_lane, int, 32, 2); |
| 311 |
VECT_VAR_DECL_INIT(buffer_vld2_lane, int, 64, 2); |
| 312 |
VECT_VAR_DECL_INIT(buffer_vld2_lane, uint, 8, 2); |
| 313 |
VECT_VAR_DECL_INIT(buffer_vld2_lane, uint, 16, 2); |
| 314 |
VECT_VAR_DECL_INIT(buffer_vld2_lane, uint, 32, 2); |
| 315 |
VECT_VAR_DECL_INIT(buffer_vld2_lane, uint, 64, 2); |
| 316 |
VECT_VAR_DECL_INIT(buffer_vld2_lane, poly, 8, 2); |
| 317 |
VECT_VAR_DECL_INIT(buffer_vld2_lane, poly, 16, 2); |
| 318 |
#if MFLOAT8_SUPPORTED |
| 319 |
VECT_VAR_DECL(buffer_vld2_lane, mfloat, 8, 2)[2]; |
| 320 |
#endif |
| 321 |
#if defined (__ARM_FP16_FORMAT_IEEE) || defined (__ARM_FP16_FORMAT_ALTERNATIVE) |
| 322 |
VECT_VAR_DECL_INIT(buffer_vld2_lane, float, 16, 2); |
| 323 |
#endif |
| 324 |
VECT_VAR_DECL_INIT(buffer_vld2_lane, float, 32, 2); |
| 325 |
|
| 326 |
/* Input buffers for vld3_lane */ |
| 327 |
VECT_VAR_DECL_INIT(buffer_vld3_lane, int, 8, 3); |
| 328 |
VECT_VAR_DECL_INIT(buffer_vld3_lane, int, 16, 3); |
| 329 |
VECT_VAR_DECL_INIT(buffer_vld3_lane, int, 32, 3); |
| 330 |
VECT_VAR_DECL_INIT(buffer_vld3_lane, int, 64, 3); |
| 331 |
VECT_VAR_DECL_INIT(buffer_vld3_lane, uint, 8, 3); |
| 332 |
VECT_VAR_DECL_INIT(buffer_vld3_lane, uint, 16, 3); |
| 333 |
VECT_VAR_DECL_INIT(buffer_vld3_lane, uint, 32, 3); |
| 334 |
VECT_VAR_DECL_INIT(buffer_vld3_lane, uint, 64, 3); |
| 335 |
VECT_VAR_DECL_INIT(buffer_vld3_lane, poly, 8, 3); |
| 336 |
VECT_VAR_DECL_INIT(buffer_vld3_lane, poly, 16, 3); |
| 337 |
#if MFLOAT8_SUPPORTED |
| 338 |
VECT_VAR_DECL(buffer_vld3_lane, mfloat, 8, 3)[3]; |
| 339 |
#endif |
| 340 |
#if defined (__ARM_FP16_FORMAT_IEEE) || defined (__ARM_FP16_FORMAT_ALTERNATIVE) |
| 341 |
VECT_VAR_DECL_INIT(buffer_vld3_lane, float, 16, 3); |
| 342 |
#endif |
| 343 |
VECT_VAR_DECL_INIT(buffer_vld3_lane, float, 32, 3); |
| 344 |
|
| 345 |
/* Input buffers for vld4_lane */ |
| 346 |
VECT_VAR_DECL_INIT(buffer_vld4_lane, int, 8, 4); |
| 347 |
VECT_VAR_DECL_INIT(buffer_vld4_lane, int, 16, 4); |
| 348 |
VECT_VAR_DECL_INIT(buffer_vld4_lane, int, 32, 4); |
| 349 |
VECT_VAR_DECL_INIT(buffer_vld4_lane, int, 64, 4); |
| 350 |
VECT_VAR_DECL_INIT(buffer_vld4_lane, uint, 8, 4); |
| 351 |
VECT_VAR_DECL_INIT(buffer_vld4_lane, uint, 16, 4); |
| 352 |
VECT_VAR_DECL_INIT(buffer_vld4_lane, uint, 32, 4); |
| 353 |
VECT_VAR_DECL_INIT(buffer_vld4_lane, uint, 64, 4); |
| 354 |
VECT_VAR_DECL_INIT(buffer_vld4_lane, poly, 8, 4); |
| 355 |
VECT_VAR_DECL_INIT(buffer_vld4_lane, poly, 16, 4); |
| 356 |
#if MFLOAT8_SUPPORTED |
| 357 |
VECT_VAR_DECL(buffer_vld4_lane, mfloat, 8, 4)[4]; |
| 358 |
#endif |
| 359 |
#if defined (__ARM_FP16_FORMAT_IEEE) || defined (__ARM_FP16_FORMAT_ALTERNATIVE) |
| 360 |
VECT_VAR_DECL_INIT(buffer_vld4_lane, float, 16, 4); |
| 361 |
#endif |
| 362 |
VECT_VAR_DECL_INIT(buffer_vld4_lane, float, 32, 4); |
| 363 |
|
| 364 |
void exec_vldX_lane (void) |
| 365 |
{ |
| 366 |
/* In this case, input variables are arrays of vectors. */ |
| 367 |
#define DECL_VLDX_LANE(T1, W, N, X) \ |
| 368 |
VECT_ARRAY_TYPE(T1, W, N, X) VECT_ARRAY_VAR(vector, T1, W, N, X); \ |
| 369 |
VECT_ARRAY_TYPE(T1, W, N, X) VECT_ARRAY_VAR(vector_src, T1, W, N, X); \ |
| 370 |
VECT_VAR_DECL(result_bis_##X, T1, W, N)[X * N] |
| 371 |
|
| 372 |
/* We need to use a temporary result buffer (result_bis), because |
| 373 |
the one used for other tests is not large enough. A subset of the |
| 374 |
result data is moved from result_bis to result, and it is this |
| 375 |
subset which is used to check the actual behavior. The next |
| 376 |
macro enables to move another chunk of data from result_bis to |
| 377 |
result. */ |
| 378 |
/* We also use another extra input buffer (buffer_src), which we |
| 379 |
fill with 0xAA, and which it used to load a vector from which we |
| 380 |
read a given lane. */ |
| 381 |
#define TEST_VLDX_LANE(Q, T1, T2, W, N, X, L) \ |
| 382 |
memset (VECT_VAR(buffer_src, T1, W, N), 0xAA, \ |
| 383 |
sizeof(VECT_VAR(buffer_src, T1, W, N))); \ |
| 384 |
\ |
| 385 |
VECT_ARRAY_VAR(vector_src, T1, W, N, X) = \ |
| 386 |
vld##X##Q##_##T2##W(VECT_VAR(buffer_src, T1, W, N)); \ |
| 387 |
\ |
| 388 |
VECT_ARRAY_VAR(vector, T1, W, N, X) = \ |
| 389 |
/* Use dedicated init buffer, of size. X */ \ |
| 390 |
vld##X##Q##_lane_##T2##W(VECT_VAR(buffer_vld##X##_lane, T1, W, X), \ |
| 391 |
VECT_ARRAY_VAR(vector_src, T1, W, N, X), \ |
| 392 |
L); \ |
| 393 |
vst##X##Q##_##T2##W(VECT_VAR(result_bis_##X, T1, W, N), \ |
| 394 |
VECT_ARRAY_VAR(vector, T1, W, N, X)); \ |
| 395 |
memcpy(VECT_VAR(result, T1, W, N), VECT_VAR(result_bis_##X, T1, W, N), \ |
| 396 |
sizeof(VECT_VAR(result, T1, W, N))) |
| 397 |
|
| 398 |
/* Overwrite "result" with the contents of "result_bis"[Y]. */ |
| 399 |
#define TEST_EXTRA_CHUNK(T1, W, N, X, Y) \ |
| 400 |
memcpy(VECT_VAR(result, T1, W, N), \ |
| 401 |
&(VECT_VAR(result_bis_##X, T1, W, N)[Y*N]), \ |
| 402 |
sizeof(VECT_VAR(result, T1, W, N))); |
| 403 |
|
| 404 |
/* We need all variants in 64 bits, but there is no 64x2 variant. */ |
| 405 |
#define DECL_ALL_VLDX_LANE_NO_FP16(X) \ |
| 406 |
DECL_VLDX_LANE(int, 8, 8, X); \ |
| 407 |
DECL_VLDX_LANE(int, 16, 4, X); \ |
| 408 |
DECL_VLDX_LANE(int, 32, 2, X); \ |
| 409 |
DECL_VLDX_LANE(uint, 8, 8, X); \ |
| 410 |
DECL_VLDX_LANE(uint, 16, 4, X); \ |
| 411 |
DECL_VLDX_LANE(uint, 32, 2, X); \ |
| 412 |
DECL_VLDX_LANE(poly, 8, 8, X); \ |
| 413 |
DECL_VLDX_LANE(poly, 16, 4, X); \ |
| 414 |
DECL_VLDX_LANE(int, 16, 8, X); \ |
| 415 |
DECL_VLDX_LANE(int, 32, 4, X); \ |
| 416 |
DECL_VLDX_LANE(uint, 16, 8, X); \ |
| 417 |
DECL_VLDX_LANE(uint, 32, 4, X); \ |
| 418 |
DECL_VLDX_LANE(poly, 16, 8, X); \ |
| 419 |
MFLOAT8_ONLY(DECL_VLDX_LANE(mfloat, 8, 8, X)); \ |
| 420 |
DECL_VLDX_LANE(float, 32, 2, X); \ |
| 421 |
DECL_VLDX_LANE(float, 32, 4, X) |
| 422 |
|
| 423 |
#if defined (__ARM_FP16_FORMAT_IEEE) || defined (__ARM_FP16_FORMAT_ALTERNATIVE) |
| 424 |
#define DECL_ALL_VLDX_LANE(X) \ |
| 425 |
DECL_ALL_VLDX_LANE_NO_FP16(X); \ |
| 426 |
DECL_VLDX_LANE(float, 16, 4, X); \ |
| 427 |
DECL_VLDX_LANE(float, 16, 8, X) |
| 428 |
#else |
| 429 |
#define DECL_ALL_VLDX_LANE(X) DECL_ALL_VLDX_LANE_NO_FP16(X) |
| 430 |
#endif |
| 431 |
|
| 432 |
/* Add some padding to try to catch out of bound accesses. */ |
| 433 |
#define ARRAY1(V, T, W, N) VECT_VAR_DECL(V,T,W,N)[1]={CONVERT(T##W##_t,42)} |
| 434 |
#define DUMMY_ARRAY(V, T, W, N, L) \ |
| 435 |
VECT_VAR_DECL(V,T,W,N)[N*L]={}; \ |
| 436 |
ARRAY1(V##_pad,T,W,N) |
| 437 |
|
| 438 |
/* Use the same lanes regardless of the size of the array (X), for |
| 439 |
simplicity. */ |
| 440 |
#define TEST_ALL_VLDX_LANE_NO_FP16(X) \ |
| 441 |
TEST_VLDX_LANE(, int, s, 8, 8, X, 7); \ |
| 442 |
TEST_VLDX_LANE(, int, s, 16, 4, X, 2); \ |
| 443 |
TEST_VLDX_LANE(, int, s, 32, 2, X, 0); \ |
| 444 |
TEST_VLDX_LANE(, uint, u, 8, 8, X, 4); \ |
| 445 |
TEST_VLDX_LANE(, uint, u, 16, 4, X, 3); \ |
| 446 |
TEST_VLDX_LANE(, uint, u, 32, 2, X, 1); \ |
| 447 |
TEST_VLDX_LANE(, poly, p, 8, 8, X, 4); \ |
| 448 |
TEST_VLDX_LANE(, poly, p, 16, 4, X, 3); \ |
| 449 |
TEST_VLDX_LANE(q, int, s, 16, 8, X, 6); \ |
| 450 |
TEST_VLDX_LANE(q, int, s, 32, 4, X, 2); \ |
| 451 |
TEST_VLDX_LANE(q, uint, u, 16, 8, X, 5); \ |
| 452 |
TEST_VLDX_LANE(q, uint, u, 32, 4, X, 0); \ |
| 453 |
TEST_VLDX_LANE(q, poly, p, 16, 8, X, 5); \ |
| 454 |
MFLOAT8_ONLY(TEST_VLDX_LANE(, mfloat, mf, 8, 8, X, 7)); \ |
| 455 |
TEST_VLDX_LANE(, float, f, 32, 2, X, 0); \ |
| 456 |
TEST_VLDX_LANE(q, float, f, 32, 4, X, 2) |
| 457 |
|
| 458 |
#if defined (__ARM_FP16_FORMAT_IEEE) || defined (__ARM_FP16_FORMAT_ALTERNATIVE) |
| 459 |
#define TEST_ALL_VLDX_LANE(X) \ |
| 460 |
TEST_ALL_VLDX_LANE_NO_FP16(X); \ |
| 461 |
TEST_VLDX_LANE(, float, f, 16, 4, X, 2); \ |
| 462 |
TEST_VLDX_LANE(q, float, f, 16, 8, X, 6) |
| 463 |
#else |
| 464 |
#define TEST_ALL_VLDX_LANE(X) TEST_ALL_VLDX_LANE_NO_FP16(X) |
| 465 |
#endif |
| 466 |
|
| 467 |
#define TEST_ALL_EXTRA_CHUNKS_NO_FP16(X,Y) \ |
| 468 |
TEST_EXTRA_CHUNK(int, 8, 8, X, Y); \ |
| 469 |
TEST_EXTRA_CHUNK(int, 16, 4, X, Y); \ |
| 470 |
TEST_EXTRA_CHUNK(int, 32, 2, X, Y); \ |
| 471 |
TEST_EXTRA_CHUNK(uint, 8, 8, X, Y); \ |
| 472 |
TEST_EXTRA_CHUNK(uint, 16, 4, X, Y); \ |
| 473 |
TEST_EXTRA_CHUNK(uint, 32, 2, X, Y); \ |
| 474 |
TEST_EXTRA_CHUNK(poly, 8, 8, X, Y); \ |
| 475 |
TEST_EXTRA_CHUNK(poly, 16, 4, X, Y); \ |
| 476 |
TEST_EXTRA_CHUNK(int, 16, 8, X, Y); \ |
| 477 |
TEST_EXTRA_CHUNK(int, 32, 4, X, Y); \ |
| 478 |
TEST_EXTRA_CHUNK(uint, 16, 8, X, Y); \ |
| 479 |
TEST_EXTRA_CHUNK(uint, 32, 4, X, Y); \ |
| 480 |
TEST_EXTRA_CHUNK(poly, 16, 8, X, Y); \ |
| 481 |
MFLOAT8_ONLY(TEST_EXTRA_CHUNK(mfloat, 8, 8, X, Y)); \ |
| 482 |
TEST_EXTRA_CHUNK(float, 32, 2, X, Y); \ |
| 483 |
TEST_EXTRA_CHUNK(float, 32, 4, X, Y) |
| 484 |
|
| 485 |
#if defined (__ARM_FP16_FORMAT_IEEE) || defined (__ARM_FP16_FORMAT_ALTERNATIVE) |
| 486 |
#define TEST_ALL_EXTRA_CHUNKS(X,Y) \ |
| 487 |
TEST_ALL_EXTRA_CHUNKS_NO_FP16(X, Y); \ |
| 488 |
TEST_EXTRA_CHUNK(float, 16, 4, X, Y); \ |
| 489 |
TEST_EXTRA_CHUNK(float, 16, 8, X, Y) |
| 490 |
#else |
| 491 |
#define TEST_ALL_EXTRA_CHUNKS(X,Y) TEST_ALL_EXTRA_CHUNKS_NO_FP16(X, Y) |
| 492 |
#endif |
| 493 |
|
| 494 |
/* vldX_lane supports only a subset of all variants. */ |
| 495 |
#define CHECK_RESULTS_VLDX_LANE_NO_FP16(test_name,EXPECTED,comment) \ |
| 496 |
CHECK(test_name, int, 8, 8, PRIx8, EXPECTED, comment); \ |
| 497 |
CHECK(test_name, int, 16, 4, PRIx16, EXPECTED, comment); \ |
| 498 |
CHECK(test_name, int, 32, 2, PRIx32, EXPECTED, comment); \ |
| 499 |
CHECK(test_name, uint, 8, 8, PRIx8, EXPECTED, comment); \ |
| 500 |
CHECK(test_name, uint, 16, 4, PRIx16, EXPECTED, comment); \ |
| 501 |
CHECK(test_name, uint, 32, 2, PRIx32, EXPECTED, comment); \ |
| 502 |
CHECK_POLY(test_name, poly, 8, 8, PRIx8, EXPECTED, comment); \ |
| 503 |
CHECK_POLY(test_name, poly, 16, 4, PRIx16, EXPECTED, comment); \ |
| 504 |
MFLOAT8_ONLY(CHECK_FP(test_name, mfloat, 8, 8, PRIx8, EXPECTED, comment)); \ |
| 505 |
CHECK_FP(test_name, float, 32, 2, PRIx32, EXPECTED, comment); \ |
| 506 |
CHECK(test_name, int, 16, 8, PRIx16, EXPECTED, comment); \ |
| 507 |
CHECK(test_name, int, 32, 4, PRIx32, EXPECTED, comment); \ |
| 508 |
CHECK(test_name, uint, 16, 8, PRIx16, EXPECTED, comment); \ |
| 509 |
CHECK(test_name, uint, 32, 4, PRIx32, EXPECTED, comment); \ |
| 510 |
CHECK_POLY(test_name, poly, 16, 8, PRIx16, EXPECTED, comment); \ |
| 511 |
CHECK_FP(test_name, float, 32, 4, PRIx32, EXPECTED, comment) |
| 512 |
|
| 513 |
#if defined (__ARM_FP16_FORMAT_IEEE) || defined (__ARM_FP16_FORMAT_ALTERNATIVE) |
| 514 |
#define CHECK_RESULTS_VLDX_LANE(test_name,EXPECTED,comment) \ |
| 515 |
{ \ |
| 516 |
CHECK_RESULTS_VLDX_LANE_NO_FP16(test_name,EXPECTED,comment); \ |
| 517 |
CHECK_FP(test_name, float, 16, 4, PRIx16, EXPECTED, comment); \ |
| 518 |
CHECK_FP(test_name, float, 16, 8, PRIx16, EXPECTED, comment); \ |
| 519 |
} |
| 520 |
#else |
| 521 |
#define CHECK_RESULTS_VLDX_LANE(test_name,EXPECTED,comment) \ |
| 522 |
{ \ |
| 523 |
CHECK_RESULTS_VLDX_LANE_NO_FP16(test_name,EXPECTED,comment); \ |
| 524 |
} |
| 525 |
#endif |
| 526 |
|
| 527 |
#if MFLOAT8_SUPPORTED |
| 528 |
__builtin_memcpy (VECT_VAR(buffer_vld2_lane, mfloat, 8, 2), |
| 529 |
VECT_VAR(buffer_vld2_lane, int, 8, 2), 2); |
| 530 |
__builtin_memcpy (VECT_VAR(buffer_vld3_lane, mfloat, 8, 3), |
| 531 |
VECT_VAR(buffer_vld3_lane, int, 8, 3), 3); |
| 532 |
__builtin_memcpy (VECT_VAR(buffer_vld4_lane, mfloat, 8, 4), |
| 533 |
VECT_VAR(buffer_vld4_lane, int, 8, 4), 4); |
| 534 |
#endif |
| 535 |
|
| 536 |
/* Declare the temporary buffers / variables. */ |
| 537 |
DECL_ALL_VLDX_LANE(2); |
| 538 |
DECL_ALL_VLDX_LANE(3); |
| 539 |
DECL_ALL_VLDX_LANE(4); |
| 540 |
|
| 541 |
/* Define dummy input arrays, large enough for x4 vectors. */ |
| 542 |
DUMMY_ARRAY(buffer_src, int, 8, 8, 4); |
| 543 |
DUMMY_ARRAY(buffer_src, int, 16, 4, 4); |
| 544 |
DUMMY_ARRAY(buffer_src, int, 32, 2, 4); |
| 545 |
DUMMY_ARRAY(buffer_src, uint, 8, 8, 4); |
| 546 |
DUMMY_ARRAY(buffer_src, uint, 16, 4, 4); |
| 547 |
DUMMY_ARRAY(buffer_src, uint, 32, 2, 4); |
| 548 |
DUMMY_ARRAY(buffer_src, poly, 8, 8, 4); |
| 549 |
DUMMY_ARRAY(buffer_src, poly, 16, 4, 4); |
| 550 |
DUMMY_ARRAY(buffer_src, int, 16, 8, 4); |
| 551 |
DUMMY_ARRAY(buffer_src, int, 32, 4, 4); |
| 552 |
DUMMY_ARRAY(buffer_src, uint, 16, 8, 4); |
| 553 |
DUMMY_ARRAY(buffer_src, uint, 32, 4, 4); |
| 554 |
DUMMY_ARRAY(buffer_src, poly, 16, 8, 4); |
| 555 |
#if MFLOAT8_SUPPORTED |
| 556 |
DUMMY_ARRAY(buffer_src, mfloat, 8, 8, 4); |
| 557 |
#endif |
| 558 |
#if defined (__ARM_FP16_FORMAT_IEEE) || defined (__ARM_FP16_FORMAT_ALTERNATIVE) |
| 559 |
DUMMY_ARRAY(buffer_src, float, 16, 4, 4); |
| 560 |
DUMMY_ARRAY(buffer_src, float, 16, 8, 4); |
| 561 |
#endif |
| 562 |
DUMMY_ARRAY(buffer_src, float, 32, 2, 4); |
| 563 |
DUMMY_ARRAY(buffer_src, float, 32, 4, 4); |
| 564 |
|
| 565 |
/* Check vld2_lane/vld2q_lane. */ |
| 566 |
clean_results (); |
| 567 |
#define TEST_MSG "VLD2_LANE/VLD2Q_LANE" |
| 568 |
TEST_ALL_VLDX_LANE(2); |
| 569 |
CHECK_RESULTS_VLDX_LANE (TEST_MSG, expected_vld2_0, " chunk 0"); |
| 570 |
|
| 571 |
TEST_ALL_EXTRA_CHUNKS(2, 1); |
| 572 |
CHECK_RESULTS_VLDX_LANE (TEST_MSG, expected_vld2_1, " chunk 1"); |
| 573 |
|
| 574 |
/* Check vld3_lane/vld3q_lane. */ |
| 575 |
clean_results (); |
| 576 |
#undef TEST_MSG |
| 577 |
#define TEST_MSG "VLD3_LANE/VLD3Q_LANE" |
| 578 |
TEST_ALL_VLDX_LANE(3); |
| 579 |
CHECK_RESULTS_VLDX_LANE (TEST_MSG, expected_vld3_0, " chunk 0"); |
| 580 |
|
| 581 |
TEST_ALL_EXTRA_CHUNKS(3, 1); |
| 582 |
CHECK_RESULTS_VLDX_LANE (TEST_MSG, expected_vld3_1, " chunk 1"); |
| 583 |
|
| 584 |
TEST_ALL_EXTRA_CHUNKS(3, 2); |
| 585 |
CHECK_RESULTS_VLDX_LANE (TEST_MSG, expected_vld3_2, " chunk 2"); |
| 586 |
|
| 587 |
/* Check vld4_lane/vld4q_lane. */ |
| 588 |
clean_results (); |
| 589 |
#undef TEST_MSG |
| 590 |
#define TEST_MSG "VLD4_LANE/VLD4Q_LANE" |
| 591 |
TEST_ALL_VLDX_LANE(4); |
| 592 |
CHECK_RESULTS_VLDX_LANE (TEST_MSG, expected_vld4_0, " chunk 0"); |
| 593 |
|
| 594 |
TEST_ALL_EXTRA_CHUNKS(4, 1); |
| 595 |
CHECK_RESULTS_VLDX_LANE (TEST_MSG, expected_vld4_1, " chunk 1"); |
| 596 |
TEST_ALL_EXTRA_CHUNKS(4, 2); |
| 597 |
|
| 598 |
CHECK_RESULTS_VLDX_LANE (TEST_MSG, expected_vld4_2, " chunk 2"); |
| 599 |
|
| 600 |
TEST_ALL_EXTRA_CHUNKS(4, 3); |
| 601 |
CHECK_RESULTS_VLDX_LANE (TEST_MSG, expected_vld4_3, " chunk 3"); |
| 602 |
} |
| 603 |
|
| 604 |
int main (void) |
| 605 |
{ |
| 606 |
exec_vldX_lane (); |
| 607 |
return 0; |
| 608 |
} |