1 /* 2 3 Copyright (c) 2009, 2010, 2011, 2013 STMicroelectronics 4 Written by Christophe Lyon 5 6 Permission is hereby granted, free of charge, to any person obtaining a copy 7 of this software and associated documentation files (the "Software"), to deal 8 in the Software without restriction, including without limitation the rights 9 to use, copy, modify, merge, publish, distribute, sublicense, and/or sell 10 copies of the Software, and to permit persons to whom the Software is 11 furnished to do so, subject to the following conditions: 12 13 The above copyright notice and this permission notice shall be included in 14 all copies or substantial portions of the Software. 15 16 THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR 17 IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, 18 FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE 19 AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER 20 LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, 21 OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN 22 THE SOFTWARE. 23 24 */ 25 26 #if defined(__arm__) || defined(__aarch64__) 27 #include <arm_neon.h> 28 #else 29 #include "stm-arm-neon.h" 30 #endif 31 32 #include "stm-arm-neon-ref.h" 33 34 void exec_vldX (void) 35 { 36 /* In this case, input variables are arrays of vectors */ 37 #define DECL_VLDX(T1, W, N, X) \ 38 VECT_ARRAY_TYPE(T1, W, N, X) VECT_ARRAY_VAR(vector, T1, W, N, X); \ 39 VECT_VAR_DECL(result_bis_##X, T1, W, N)[X * N] 40 41 /* We need to use a temporary result buffer (result_bis), because 42 the one used for other tests is not large enough. A subset of the 43 result data is moved from result_bis to result, and it is this 44 subset which is used to check the actual behaviour. The next 45 macro enables to move another chunk of data from result_bis to 46 result. */ 47 #define TEST_VLDX(Q, T1, T2, W, N, X) \ 48 VECT_ARRAY_VAR(vector, T1, W, N, X) = \ 49 /* Use dedicated init buffer, of size X */ \ 50 vld##X##Q##_##T2##W(VECT_ARRAY_VAR(buffer_vld##X, T1, W, N, X)); \ 51 vst##X##Q##_##T2##W(VECT_VAR(result_bis_##X, T1, W, N), \ 52 VECT_ARRAY_VAR(vector, T1, W, N, X)); \ 53 memcpy(VECT_VAR(result, T1, W, N), VECT_VAR(result_bis_##X, T1, W, N), \ 54 sizeof(VECT_VAR(result, T1, W, N))); 55 56 /* Overwrite "result" with the contents of "result_bis"[Y] */ 57 #define TEST_EXTRA_CHUNK(T1, W, N, X,Y) \ 58 memcpy(VECT_VAR(result, T1, W, N), \ 59 &(VECT_VAR(result_bis_##X, T1, W, N)[Y*N]), \ 60 sizeof(VECT_VAR(result, T1, W, N))); 61 62 /* With ARM RVCT, we need to declare variables before any executable 63 statement */ 64 65 /* We need all variants in 64 bits, but there is no 64x2 variant */ 66 #define DECL_ALL_VLDX(X) \ 67 DECL_VLDX(int, 8, 8, X); \ 68 DECL_VLDX(int, 16, 4, X); \ 69 DECL_VLDX(int, 32, 2, X); \ 70 DECL_VLDX(int, 64, 1, X); \ 71 DECL_VLDX(uint, 8, 8, X); \ 72 DECL_VLDX(uint, 16, 4, X); \ 73 DECL_VLDX(uint, 32, 2, X); \ 74 DECL_VLDX(uint, 64, 1, X); \ 75 DECL_VLDX(poly, 8, 8, X); \ 76 DECL_VLDX(poly, 16, 4, X); \ 77 DECL_VLDX(float, 32, 2, X); \ 78 DECL_VLDX(int, 8, 16, X); \ 79 DECL_VLDX(int, 16, 8, X); \ 80 DECL_VLDX(int, 32, 4, X); \ 81 DECL_VLDX(uint, 8, 16, X); \ 82 DECL_VLDX(uint, 16, 8, X); \ 83 DECL_VLDX(uint, 32, 4, X); \ 84 DECL_VLDX(poly, 8, 16, X); \ 85 DECL_VLDX(poly, 16, 8, X); \ 86 DECL_VLDX(float, 32, 4, X) 87 88 #if defined(__ARM_FP16_FORMAT_IEEE) && ( ((__ARM_FP & 0x2) != 0) || ((__ARM_NEON_FP16_INTRINSICS & 1) != 0) ) 89 #define DECL_ALL_VLDX_FP16(X) \ 90 DECL_VLDX(float, 16, 4, X); \ 91 DECL_VLDX(float, 16, 8, X) 92 #endif 93 94 #define TEST_ALL_VLDX(X) \ 95 TEST_VLDX(, int, s, 8, 8, X); \ 96 TEST_VLDX(, int, s, 16, 4, X); \ 97 TEST_VLDX(, int, s, 32, 2, X); \ 98 TEST_VLDX(, int, s, 64, 1, X); \ 99 TEST_VLDX(, uint, u, 8, 8, X); \ 100 TEST_VLDX(, uint, u, 16, 4, X); \ 101 TEST_VLDX(, uint, u, 32, 2, X); \ 102 TEST_VLDX(, uint, u, 64, 1, X); \ 103 TEST_VLDX(, poly, p, 8, 8, X); \ 104 TEST_VLDX(, poly, p, 16, 4, X); \ 105 TEST_VLDX(, float, f, 32, 2, X); \ 106 TEST_VLDX(q, int, s, 8, 16, X); \ 107 TEST_VLDX(q, int, s, 16, 8, X); \ 108 TEST_VLDX(q, int, s, 32, 4, X); \ 109 TEST_VLDX(q, uint, u, 8, 16, X); \ 110 TEST_VLDX(q, uint, u, 16, 8, X); \ 111 TEST_VLDX(q, uint, u, 32, 4, X); \ 112 TEST_VLDX(q, poly, p, 8, 16, X); \ 113 TEST_VLDX(q, poly, p, 16, 8, X); \ 114 TEST_VLDX(q, float, f, 32, 4, X) 115 116 #if defined(__ARM_FP16_FORMAT_IEEE) && ( ((__ARM_FP & 0x2) != 0) || ((__ARM_NEON_FP16_INTRINSICS & 1) != 0) ) 117 #define TEST_ALL_VLDX_FP16(X) \ 118 TEST_VLDX(, float, f, 16, 4, X); \ 119 TEST_VLDX(q, float, f, 16, 8, X) 120 #endif 121 122 #define TEST_ALL_EXTRA_CHUNKS(X, Y) \ 123 TEST_EXTRA_CHUNK(int, 8, 8, X, Y); \ 124 TEST_EXTRA_CHUNK(int, 16, 4, X, Y); \ 125 TEST_EXTRA_CHUNK(int, 32, 2, X, Y); \ 126 TEST_EXTRA_CHUNK(int, 64, 1, X, Y); \ 127 TEST_EXTRA_CHUNK(uint, 8, 8, X, Y); \ 128 TEST_EXTRA_CHUNK(uint, 16, 4, X, Y); \ 129 TEST_EXTRA_CHUNK(uint, 32, 2, X, Y); \ 130 TEST_EXTRA_CHUNK(uint, 64, 1, X, Y); \ 131 TEST_EXTRA_CHUNK(poly, 8, 8, X, Y); \ 132 TEST_EXTRA_CHUNK(poly, 16, 4, X, Y); \ 133 TEST_EXTRA_CHUNK(float, 32, 2, X, Y); \ 134 TEST_EXTRA_CHUNK(int, 8, 16, X, Y); \ 135 TEST_EXTRA_CHUNK(int, 16, 8, X, Y); \ 136 TEST_EXTRA_CHUNK(int, 32, 4, X, Y); \ 137 TEST_EXTRA_CHUNK(uint, 8, 16, X, Y); \ 138 TEST_EXTRA_CHUNK(uint, 16, 8, X, Y); \ 139 TEST_EXTRA_CHUNK(uint, 32, 4, X, Y); \ 140 TEST_EXTRA_CHUNK(poly, 8, 16, X, Y); \ 141 TEST_EXTRA_CHUNK(poly, 16, 8, X, Y); \ 142 TEST_EXTRA_CHUNK(float, 32, 4, X, Y) 143 144 #if defined(__ARM_FP16_FORMAT_IEEE) && ( ((__ARM_FP & 0x2) != 0) || ((__ARM_NEON_FP16_INTRINSICS & 1) != 0) ) 145 #define TEST_ALL_EXTRA_CHUNKS_FP16(X, Y) \ 146 TEST_EXTRA_CHUNK(float, 16, 4, X, Y); \ 147 TEST_EXTRA_CHUNK(float, 16, 8, X, Y) 148 #endif 149 150 DECL_ALL_VLDX(2); 151 DECL_ALL_VLDX(3); 152 DECL_ALL_VLDX(4); 153 154 #if defined(__ARM_FP16_FORMAT_IEEE) && ( ((__ARM_FP & 0x2) != 0) || ((__ARM_NEON_FP16_INTRINSICS & 1) != 0) ) 155 DECL_ALL_VLDX_FP16(2); 156 DECL_ALL_VLDX_FP16(3); 157 DECL_ALL_VLDX_FP16(4); 158 #endif 159 160 /* Check vld2/vld2q */ 161 clean_results (); 162 #define TEST_MSG "VLD2/VLD2Q" 163 TEST_ALL_VLDX(2); 164 #if defined(__ARM_FP16_FORMAT_IEEE) && ( ((__ARM_FP & 0x2) != 0) || ((__ARM_NEON_FP16_INTRINSICS & 1) != 0) ) 165 TEST_ALL_VLDX_FP16(2); 166 #endif 167 dump_results_hex2 (TEST_MSG, " chunk 0"); 168 169 TEST_ALL_EXTRA_CHUNKS(2, 1); 170 #if defined(__ARM_FP16_FORMAT_IEEE) && ( ((__ARM_FP & 0x2) != 0) || ((__ARM_NEON_FP16_INTRINSICS & 1) != 0) ) 171 TEST_ALL_EXTRA_CHUNKS_FP16(2, 1); 172 #endif 173 dump_results_hex2 (TEST_MSG, " chunk 1"); 174 175 /* Check vld3/vld3q */ 176 clean_results (); 177 #undef TEST_MSG 178 #define TEST_MSG "VLD3/VLD3Q" 179 TEST_ALL_VLDX(3); 180 #if defined(__ARM_FP16_FORMAT_IEEE) && ( ((__ARM_FP & 0x2) != 0) || ((__ARM_NEON_FP16_INTRINSICS & 1) != 0) ) 181 TEST_ALL_VLDX_FP16(3); 182 #endif 183 dump_results_hex2 (TEST_MSG, " chunk 0"); 184 185 TEST_ALL_EXTRA_CHUNKS(3, 1); 186 #if defined(__ARM_FP16_FORMAT_IEEE) && ( ((__ARM_FP & 0x2) != 0) || ((__ARM_NEON_FP16_INTRINSICS & 1) != 0) ) 187 TEST_ALL_EXTRA_CHUNKS_FP16(3, 1); 188 #endif 189 dump_results_hex2 (TEST_MSG, " chunk 1"); 190 TEST_ALL_EXTRA_CHUNKS(3, 2); 191 #if defined(__ARM_FP16_FORMAT_IEEE) && ( ((__ARM_FP & 0x2) != 0) || ((__ARM_NEON_FP16_INTRINSICS & 1) != 0) ) 192 TEST_ALL_EXTRA_CHUNKS_FP16(3, 2); 193 #endif 194 dump_results_hex2 (TEST_MSG, " chunk 2"); 195 196 /* Check vld4/vld4q */ 197 clean_results (); 198 #undef TEST_MSG 199 #define TEST_MSG "VLD4/VLD4Q" 200 TEST_ALL_VLDX(4); 201 #if defined(__ARM_FP16_FORMAT_IEEE) && ( ((__ARM_FP & 0x2) != 0) || ((__ARM_NEON_FP16_INTRINSICS & 1) != 0) ) 202 TEST_ALL_VLDX_FP16(4); 203 #endif 204 dump_results_hex2 (TEST_MSG, " chunk 0"); 205 206 TEST_ALL_EXTRA_CHUNKS(4, 1); 207 #if defined(__ARM_FP16_FORMAT_IEEE) && ( ((__ARM_FP & 0x2) != 0) || ((__ARM_NEON_FP16_INTRINSICS & 1) != 0) ) 208 TEST_ALL_EXTRA_CHUNKS_FP16(4, 1); 209 #endif 210 dump_results_hex2 (TEST_MSG, " chunk 1"); 211 TEST_ALL_EXTRA_CHUNKS(4, 2); 212 #if defined(__ARM_FP16_FORMAT_IEEE) && ( ((__ARM_FP & 0x2) != 0) || ((__ARM_NEON_FP16_INTRINSICS & 1) != 0) ) 213 TEST_ALL_EXTRA_CHUNKS_FP16(4, 2); 214 #endif 215 dump_results_hex2 (TEST_MSG, " chunk 2"); 216 TEST_ALL_EXTRA_CHUNKS(4, 3); 217 #if defined(__ARM_FP16_FORMAT_IEEE) && ( ((__ARM_FP & 0x2) != 0) || ((__ARM_NEON_FP16_INTRINSICS & 1) != 0) ) 218 TEST_ALL_EXTRA_CHUNKS_FP16(4, 3); 219 #endif 220 dump_results_hex2 (TEST_MSG, " chunk 3"); 221 } 222
