1/************************************************************************** 2 * 3 * Copyright © 2010 Luca Barbieri 4 * 5 * Permission is hereby granted, free of charge, to any person obtaining a 6 * copy of this software and associated documentation files (the "Software"), 7 * to deal in the Software without restriction, including without limitation 8 * the rights to use, copy, modify, merge, publish, distribute, sublicense, 9 * and/or sell copies of the Software, and to permit persons to whom the 10 * Software is furnished to do so, subject to the following conditions: 11 * 12 * The above copyright notice and this permission notice (including the next 13 * paragraph) shall be included in all copies or substantial portions of the 14 * 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 19 * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER 20 * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING 21 * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER 22 * DEALINGS IN THE SOFTWARE. 23 * 24 **************************************************************************/ 25 26#include <stdio.h> 27#include "translate/translate.h" 28#include "util/u_memory.h" 29#include "util/format/u_format.h" 30#include "util/half_float.h" 31#include "util/u_cpu_detect.h" 32#include "rtasm/rtasm_cpu.h" 33 34/* don't use this for serious use */ 35static double rand_double() 36{ 37 const double rm = (double)RAND_MAX + 1; 38 double div = 1; 39 double v = 0; 40 unsigned i; 41 for(i = 0; i < 4; ++i) 42 { 43 div *= rm; 44 v += (double)rand() / div; 45 } 46 return v; 47} 48 49int main(int argc, char** argv) 50{ 51 struct translate *(*create_fn)(const struct translate_key *key) = 0; 52 53 extern struct util_cpu_caps_t util_cpu_caps; 54 struct translate_key key; 55 unsigned output_format; 56 unsigned input_format; 57 unsigned buffer_size = 4096; 58 unsigned char* buffer[5]; 59 unsigned char* byte_buffer; 60 float* float_buffer; 61 double* double_buffer; 62 uint16_t *half_buffer; 63 unsigned * elts; 64 unsigned count = 4; 65 unsigned i, j, k; 66 unsigned passed = 0; 67 unsigned total = 0; 68 const float error = 0.03125; 69 70 create_fn = 0; 71 72 util_cpu_detect(); 73 74 if (argc <= 1 || 75 !strcmp(argv[1], "default") ) 76 create_fn = translate_create; 77 else if (!strcmp(argv[1], "generic")) 78 create_fn = translate_generic_create; 79 else if (!strcmp(argv[1], "x86")) 80 create_fn = translate_sse2_create; 81 else if (!strcmp(argv[1], "nosse")) 82 { 83 util_cpu_caps.has_sse = 0; 84 util_cpu_caps.has_sse2 = 0; 85 util_cpu_caps.has_sse3 = 0; 86 util_cpu_caps.has_sse4_1 = 0; 87 create_fn = translate_sse2_create; 88 } 89 else if (!strcmp(argv[1], "sse")) 90 { 91 if(!util_get_cpu_caps()->has_sse || !rtasm_cpu_has_sse()) 92 { 93 printf("Error: CPU doesn't support SSE (test with qemu)\n"); 94 return 2; 95 } 96 util_cpu_caps.has_sse2 = 0; 97 util_cpu_caps.has_sse3 = 0; 98 util_cpu_caps.has_sse4_1 = 0; 99 create_fn = translate_sse2_create; 100 } 101 else if (!strcmp(argv[1], "sse2")) 102 { 103 if(!util_get_cpu_caps()->has_sse2 || !rtasm_cpu_has_sse()) 104 { 105 printf("Error: CPU doesn't support SSE2 (test with qemu)\n"); 106 return 2; 107 } 108 util_cpu_caps.has_sse3 = 0; 109 util_cpu_caps.has_sse4_1 = 0; 110 create_fn = translate_sse2_create; 111 } 112 else if (!strcmp(argv[1], "sse3")) 113 { 114 if(!util_get_cpu_caps()->has_sse3 || !rtasm_cpu_has_sse()) 115 { 116 printf("Error: CPU doesn't support SSE3 (test with qemu)\n"); 117 return 2; 118 } 119 util_cpu_caps.has_sse4_1 = 0; 120 create_fn = translate_sse2_create; 121 } 122 else if (!strcmp(argv[1], "sse4.1")) 123 { 124 if(!util_get_cpu_caps()->has_sse4_1 || !rtasm_cpu_has_sse()) 125 { 126 printf("Error: CPU doesn't support SSE4.1 (test with qemu)\n"); 127 return 2; 128 } 129 create_fn = translate_sse2_create; 130 } 131 132 if (!create_fn) 133 { 134 printf("Usage: ./translate_test [default|generic|x86|nosse|sse|sse2|sse3|sse4.1]\n"); 135 return 2; 136 } 137 138 for (i = 1; i < ARRAY_SIZE(buffer); ++i) 139 buffer[i] = align_malloc(buffer_size, 4096); 140 141 byte_buffer = align_malloc(buffer_size, 4096); 142 float_buffer = align_malloc(buffer_size, 4096); 143 double_buffer = align_malloc(buffer_size, 4096); 144 half_buffer = align_malloc(buffer_size, 4096); 145 146 elts = align_malloc(count * sizeof *elts, 4096); 147 148 key.nr_elements = 1; 149 key.element[0].input_buffer = 0; 150 key.element[0].input_offset = 0; 151 key.element[0].output_offset = 0; 152 key.element[0].type = TRANSLATE_ELEMENT_NORMAL; 153 key.element[0].instance_divisor = 0; 154 155 srand(4359025); 156 157 /* avoid negative values that work badly when converted to unsigned format*/ 158 for (i = 0; i < buffer_size; ++i) 159 byte_buffer[i] = rand() & 0x7f7f7f7f; 160 161 for (i = 0; i < buffer_size / sizeof(float); ++i) 162 float_buffer[i] = (float)rand_double(); 163 164 for (i = 0; i < buffer_size / sizeof(double); ++i) 165 double_buffer[i] = rand_double(); 166 167 for (i = 0; i < buffer_size / sizeof(double); ++i) 168 half_buffer[i] = _mesa_float_to_half((float) rand_double()); 169 170 for (i = 0; i < count; ++i) 171 elts[i] = i; 172 173 for (output_format = 1; output_format < PIPE_FORMAT_COUNT; ++output_format) 174 { 175 const struct util_format_description* output_format_desc = util_format_description(output_format); 176 const struct util_format_pack_description* output_format_pack = util_format_pack_description(output_format); 177 util_format_fetch_rgba_func_ptr fetch_rgba = 178 util_format_fetch_rgba_func(output_format); 179 unsigned output_format_size; 180 unsigned output_normalized = 0; 181 182 if (!output_format_desc 183 || !fetch_rgba 184 || !output_format_pack->pack_rgba_float 185 || output_format_desc->colorspace != UTIL_FORMAT_COLORSPACE_RGB 186 || output_format_desc->layout != UTIL_FORMAT_LAYOUT_PLAIN 187 || !translate_is_output_format_supported(output_format)) 188 continue; 189 190 for(i = 0; i < output_format_desc->nr_channels; ++i) 191 { 192 if(output_format_desc->channel[i].type != UTIL_FORMAT_TYPE_FLOAT) 193 output_normalized |= (1 << output_format_desc->channel[i].normalized); 194 } 195 196 output_format_size = util_format_get_stride(output_format, 1); 197 198 for (input_format = 1; input_format < PIPE_FORMAT_COUNT; ++input_format) 199 { 200 const struct util_format_description* input_format_desc = util_format_description(input_format); 201 const struct util_format_pack_description* input_format_pack = util_format_pack_description(input_format); 202 util_format_fetch_rgba_func_ptr fetch_rgba = 203 util_format_fetch_rgba_func(input_format); 204 unsigned input_format_size; 205 struct translate* translate[2]; 206 unsigned fail = 0; 207 unsigned used_generic = 0; 208 unsigned input_normalized = 0; 209 boolean input_is_float = FALSE; 210 211 if (!input_format_desc 212 || !fetch_rgba 213 || !input_format_pack->pack_rgba_float 214 || input_format_desc->colorspace != UTIL_FORMAT_COLORSPACE_RGB 215 || input_format_desc->layout != UTIL_FORMAT_LAYOUT_PLAIN 216 || !translate_is_output_format_supported(input_format)) 217 continue; 218 219 input_format_size = util_format_get_stride(input_format, 1); 220 221 for(i = 0; i < input_format_desc->nr_channels; ++i) 222 { 223 if(input_format_desc->channel[i].type == UTIL_FORMAT_TYPE_FLOAT) 224 { 225 input_is_float = 1; 226 input_normalized |= 1 << 1; 227 } 228 else 229 input_normalized |= (1 << input_format_desc->channel[i].normalized); 230 } 231 232 if(((input_normalized | output_normalized) == 3) 233 || ((input_normalized & 1) && (output_normalized & 1) 234 && input_format_size * output_format_desc->nr_channels > output_format_size * input_format_desc->nr_channels)) 235 continue; 236 237 key.element[0].input_format = input_format; 238 key.element[0].output_format = output_format; 239 key.output_stride = output_format_size; 240 translate[0] = create_fn(&key); 241 if (!translate[0]) 242 continue; 243 244 key.element[0].input_format = output_format; 245 key.element[0].output_format = input_format; 246 key.output_stride = input_format_size; 247 translate[1] = create_fn(&key); 248 if(!translate[1]) 249 { 250 used_generic = 1; 251 translate[1] = translate_generic_create(&key); 252 if(!translate[1]) 253 continue; 254 } 255 256 for(i = 1; i < 5; ++i) 257 memset(buffer[i], 0xcd - (0x22 * i), 4096); 258 259 if(input_is_float && input_format_desc->channel[0].size == 32) 260 buffer[0] = (unsigned char*)float_buffer; 261 else if(input_is_float && input_format_desc->channel[0].size == 64) 262 buffer[0] = (unsigned char*)double_buffer; 263 else if(input_is_float && input_format_desc->channel[0].size == 16) 264 buffer[0] = (unsigned char*)half_buffer; 265 else if(input_is_float) 266 abort(); 267 else 268 buffer[0] = byte_buffer; 269 270 translate[0]->set_buffer(translate[0], 0, buffer[0], input_format_size, count - 1); 271 translate[0]->run_elts(translate[0], elts, count, 0, 0, buffer[1]); 272 translate[1]->set_buffer(translate[1], 0, buffer[1], output_format_size, count - 1); 273 translate[1]->run_elts(translate[1], elts, count, 0, 0, buffer[2]); 274 translate[0]->set_buffer(translate[0], 0, buffer[2], input_format_size, count - 1); 275 translate[0]->run_elts(translate[0], elts, count, 0, 0, buffer[3]); 276 translate[1]->set_buffer(translate[1], 0, buffer[3], output_format_size, count - 1); 277 translate[1]->run_elts(translate[1], elts, count, 0, 0, buffer[4]); 278 279 for (i = 0; i < count; ++i) 280 { 281 float a[4]; 282 float b[4]; 283 fetch_rgba(a, buffer[2] + i * input_format_size, 0, 0); 284 fetch_rgba(b, buffer[4] + i * input_format_size, 0, 0); 285 286 for (j = 0; j < count; ++j) 287 { 288 float d = a[j] - b[j]; 289 if (d > error || d < -error) 290 { 291 fail = 1; 292 break; 293 } 294 } 295 } 296 297 printf("%s%s: %s -> %s -> %s -> %s -> %s\n", 298 fail ? "FAIL" : "PASS", 299 used_generic ? "[GENERIC]" : "", 300 input_format_desc->name, output_format_desc->name, input_format_desc->name, output_format_desc->name, input_format_desc->name); 301 302 if (1) 303 { 304 for (i = 0; i < ARRAY_SIZE(buffer); ++i) 305 { 306 unsigned format_size = (i & 1) ? output_format_size : input_format_size; 307 printf("%c ", (i == 2 || i == 4) ? '*' : ' '); 308 for (j = 0; j < count; ++j) 309 { 310 for (k = 0; k < format_size; ++k) 311 { 312 printf("%02x", buffer[i][j * format_size + k]); 313 } 314 printf(" "); 315 } 316 printf("\n"); 317 } 318 } 319 320 if (!fail) 321 ++passed; 322 ++total; 323 324 if(translate[1]) 325 translate[1]->release(translate[1]); 326 translate[0]->release(translate[0]); 327 } 328 } 329 330 printf("%u/%u tests passed for translate_%s\n", passed, total, argv[1]); 331 return passed != total; 332} 333