Leon Scroggins III | 3993b37 | 2018-07-16 10:43:45 -0400 | [diff] [blame^] | 1 | /* |
| 2 | * jsimd_mips.c |
| 3 | * |
| 4 | * Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB |
| 5 | * Copyright (C) 2009-2011, 2014, 2016, 2018, D. R. Commander. |
| 6 | * Copyright (C) 2013-2014, MIPS Technologies, Inc., California. |
| 7 | * Copyright (C) 2015-2016, 2018, Matthieu Darbois. |
| 8 | * |
| 9 | * Based on the x86 SIMD extension for IJG JPEG library, |
| 10 | * Copyright (C) 1999-2006, MIYASAKA Masaru. |
| 11 | * For conditions of distribution and use, see copyright notice in jsimdext.inc |
| 12 | * |
| 13 | * This file contains the interface between the "normal" portions |
| 14 | * of the library and the SIMD implementations when running on a |
| 15 | * MIPS architecture. |
| 16 | */ |
| 17 | |
| 18 | #define JPEG_INTERNALS |
| 19 | #include "../../jinclude.h" |
| 20 | #include "../../jpeglib.h" |
| 21 | #include "../../jsimd.h" |
| 22 | #include "../../jdct.h" |
| 23 | #include "../../jsimddct.h" |
| 24 | #include "../jsimd.h" |
| 25 | |
| 26 | #include <stdio.h> |
| 27 | #include <string.h> |
| 28 | #include <ctype.h> |
| 29 | |
| 30 | static unsigned int simd_support = ~0; |
| 31 | |
| 32 | #if defined(__linux__) |
| 33 | |
| 34 | LOCAL(int) |
| 35 | parse_proc_cpuinfo(const char *search_string) |
| 36 | { |
| 37 | const char *file_name = "/proc/cpuinfo"; |
| 38 | char cpuinfo_line[256]; |
| 39 | FILE *f = NULL; |
| 40 | |
| 41 | simd_support = 0; |
| 42 | |
| 43 | if ((f = fopen(file_name, "r")) != NULL) { |
| 44 | while (fgets(cpuinfo_line, sizeof(cpuinfo_line), f) != NULL) { |
| 45 | if (strstr(cpuinfo_line, search_string) != NULL) { |
| 46 | fclose(f); |
| 47 | simd_support |= JSIMD_DSPR2; |
| 48 | return 1; |
| 49 | } |
| 50 | } |
| 51 | fclose(f); |
| 52 | } |
| 53 | /* Did not find string in the proc file, or not Linux ELF. */ |
| 54 | return 0; |
| 55 | } |
| 56 | |
| 57 | #endif |
| 58 | |
| 59 | /* |
| 60 | * Check what SIMD accelerations are supported. |
| 61 | * |
| 62 | * FIXME: This code is racy under a multi-threaded environment. |
| 63 | */ |
| 64 | LOCAL(void) |
| 65 | init_simd(void) |
| 66 | { |
| 67 | #ifndef NO_GETENV |
| 68 | char *env = NULL; |
| 69 | #endif |
| 70 | |
| 71 | if (simd_support != ~0U) |
| 72 | return; |
| 73 | |
| 74 | simd_support = 0; |
| 75 | |
| 76 | #if defined(__MIPSEL__) && defined(__mips_dsp) && (__mips_dsp_rev >= 2) |
| 77 | simd_support |= JSIMD_DSPR2; |
| 78 | #elif defined(__linux__) |
| 79 | /* We still have a chance to use MIPS DSPR2 regardless of globally used |
| 80 | * -mdspr2 options passed to gcc by performing runtime detection via |
| 81 | * /proc/cpuinfo parsing on linux */ |
| 82 | if (!parse_proc_cpuinfo("MIPS 74K")) |
| 83 | return; |
| 84 | #endif |
| 85 | |
| 86 | #ifndef NO_GETENV |
| 87 | /* Force different settings through environment variables */ |
| 88 | env = getenv("JSIMD_FORCEDSPR2"); |
| 89 | if ((env != NULL) && (strcmp(env, "1") == 0)) |
| 90 | simd_support = JSIMD_DSPR2; |
| 91 | env = getenv("JSIMD_FORCENONE"); |
| 92 | if ((env != NULL) && (strcmp(env, "1") == 0)) |
| 93 | simd_support = 0; |
| 94 | #endif |
| 95 | } |
| 96 | |
| 97 | static const int mips_idct_ifast_coefs[4] = { |
| 98 | 0x45404540, /* FIX( 1.082392200 / 2) = 17734 = 0x4546 */ |
| 99 | 0x5A805A80, /* FIX( 1.414213562 / 2) = 23170 = 0x5A82 */ |
| 100 | 0x76407640, /* FIX( 1.847759065 / 2) = 30274 = 0x7642 */ |
| 101 | 0xAC60AC60 /* FIX(-2.613125930 / 4) = -21407 = 0xAC61 */ |
| 102 | }; |
| 103 | |
| 104 | /* The following struct is borrowed from jdsample.c */ |
| 105 | typedef void (*upsample1_ptr) (j_decompress_ptr cinfo, |
| 106 | jpeg_component_info *compptr, |
| 107 | JSAMPARRAY input_data, |
| 108 | JSAMPARRAY *output_data_ptr); |
| 109 | typedef struct { |
| 110 | struct jpeg_upsampler pub; |
| 111 | JSAMPARRAY color_buf[MAX_COMPONENTS]; |
| 112 | upsample1_ptr methods[MAX_COMPONENTS]; |
| 113 | int next_row_out; |
| 114 | JDIMENSION rows_to_go; |
| 115 | int rowgroup_height[MAX_COMPONENTS]; |
| 116 | UINT8 h_expand[MAX_COMPONENTS]; |
| 117 | UINT8 v_expand[MAX_COMPONENTS]; |
| 118 | } my_upsampler; |
| 119 | |
| 120 | typedef my_upsampler *my_upsample_ptr; |
| 121 | |
| 122 | GLOBAL(int) |
| 123 | jsimd_can_rgb_ycc(void) |
| 124 | { |
| 125 | init_simd(); |
| 126 | |
| 127 | /* The code is optimised for these values only */ |
| 128 | if (BITS_IN_JSAMPLE != 8) |
| 129 | return 0; |
| 130 | if (sizeof(JDIMENSION) != 4) |
| 131 | return 0; |
| 132 | if ((RGB_PIXELSIZE != 3) && (RGB_PIXELSIZE != 4)) |
| 133 | return 0; |
| 134 | |
| 135 | if (simd_support & JSIMD_DSPR2) |
| 136 | return 1; |
| 137 | |
| 138 | return 0; |
| 139 | } |
| 140 | |
| 141 | GLOBAL(int) |
| 142 | jsimd_can_rgb_gray(void) |
| 143 | { |
| 144 | init_simd(); |
| 145 | |
| 146 | /* The code is optimised for these values only */ |
| 147 | if (BITS_IN_JSAMPLE != 8) |
| 148 | return 0; |
| 149 | if (sizeof(JDIMENSION) != 4) |
| 150 | return 0; |
| 151 | if ((RGB_PIXELSIZE != 3) && (RGB_PIXELSIZE != 4)) |
| 152 | return 0; |
| 153 | |
| 154 | if (simd_support & JSIMD_DSPR2) |
| 155 | return 1; |
| 156 | |
| 157 | return 0; |
| 158 | } |
| 159 | |
| 160 | GLOBAL(int) |
| 161 | jsimd_can_ycc_rgb(void) |
| 162 | { |
| 163 | init_simd(); |
| 164 | |
| 165 | /* The code is optimised for these values only */ |
| 166 | if (BITS_IN_JSAMPLE != 8) |
| 167 | return 0; |
| 168 | if (sizeof(JDIMENSION) != 4) |
| 169 | return 0; |
| 170 | if ((RGB_PIXELSIZE != 3) && (RGB_PIXELSIZE != 4)) |
| 171 | return 0; |
| 172 | |
| 173 | if (simd_support & JSIMD_DSPR2) |
| 174 | return 1; |
| 175 | |
| 176 | return 0; |
| 177 | } |
| 178 | |
| 179 | GLOBAL(int) |
| 180 | jsimd_can_ycc_rgb565(void) |
| 181 | { |
| 182 | return 0; |
| 183 | } |
| 184 | |
| 185 | GLOBAL(int) |
| 186 | jsimd_c_can_null_convert(void) |
| 187 | { |
| 188 | init_simd(); |
| 189 | |
| 190 | /* The code is optimised for these values only */ |
| 191 | if (BITS_IN_JSAMPLE != 8) |
| 192 | return 0; |
| 193 | if (sizeof(JDIMENSION) != 4) |
| 194 | return 0; |
| 195 | |
| 196 | if (simd_support & JSIMD_DSPR2) |
| 197 | return 1; |
| 198 | |
| 199 | return 0; |
| 200 | } |
| 201 | |
| 202 | GLOBAL(void) |
| 203 | jsimd_rgb_ycc_convert(j_compress_ptr cinfo, JSAMPARRAY input_buf, |
| 204 | JSAMPIMAGE output_buf, JDIMENSION output_row, |
| 205 | int num_rows) |
| 206 | { |
| 207 | void (*dspr2fct) (JDIMENSION, JSAMPARRAY, JSAMPIMAGE, JDIMENSION, int); |
| 208 | |
| 209 | switch (cinfo->in_color_space) { |
| 210 | case JCS_EXT_RGB: |
| 211 | dspr2fct = jsimd_extrgb_ycc_convert_dspr2; |
| 212 | break; |
| 213 | case JCS_EXT_RGBX: |
| 214 | case JCS_EXT_RGBA: |
| 215 | dspr2fct = jsimd_extrgbx_ycc_convert_dspr2; |
| 216 | break; |
| 217 | case JCS_EXT_BGR: |
| 218 | dspr2fct = jsimd_extbgr_ycc_convert_dspr2; |
| 219 | break; |
| 220 | case JCS_EXT_BGRX: |
| 221 | case JCS_EXT_BGRA: |
| 222 | dspr2fct = jsimd_extbgrx_ycc_convert_dspr2; |
| 223 | break; |
| 224 | case JCS_EXT_XBGR: |
| 225 | case JCS_EXT_ABGR: |
| 226 | dspr2fct = jsimd_extxbgr_ycc_convert_dspr2; |
| 227 | break; |
| 228 | case JCS_EXT_XRGB: |
| 229 | case JCS_EXT_ARGB: |
| 230 | dspr2fct = jsimd_extxrgb_ycc_convert_dspr2; |
| 231 | break; |
| 232 | default: |
| 233 | dspr2fct = jsimd_extrgb_ycc_convert_dspr2; |
| 234 | break; |
| 235 | } |
| 236 | |
| 237 | dspr2fct(cinfo->image_width, input_buf, output_buf, output_row, num_rows); |
| 238 | } |
| 239 | |
| 240 | GLOBAL(void) |
| 241 | jsimd_rgb_gray_convert(j_compress_ptr cinfo, JSAMPARRAY input_buf, |
| 242 | JSAMPIMAGE output_buf, JDIMENSION output_row, |
| 243 | int num_rows) |
| 244 | { |
| 245 | void (*dspr2fct) (JDIMENSION, JSAMPARRAY, JSAMPIMAGE, JDIMENSION, int); |
| 246 | |
| 247 | switch (cinfo->in_color_space) { |
| 248 | case JCS_EXT_RGB: |
| 249 | dspr2fct = jsimd_extrgb_gray_convert_dspr2; |
| 250 | break; |
| 251 | case JCS_EXT_RGBX: |
| 252 | case JCS_EXT_RGBA: |
| 253 | dspr2fct = jsimd_extrgbx_gray_convert_dspr2; |
| 254 | break; |
| 255 | case JCS_EXT_BGR: |
| 256 | dspr2fct = jsimd_extbgr_gray_convert_dspr2; |
| 257 | break; |
| 258 | case JCS_EXT_BGRX: |
| 259 | case JCS_EXT_BGRA: |
| 260 | dspr2fct = jsimd_extbgrx_gray_convert_dspr2; |
| 261 | break; |
| 262 | case JCS_EXT_XBGR: |
| 263 | case JCS_EXT_ABGR: |
| 264 | dspr2fct = jsimd_extxbgr_gray_convert_dspr2; |
| 265 | break; |
| 266 | case JCS_EXT_XRGB: |
| 267 | case JCS_EXT_ARGB: |
| 268 | dspr2fct = jsimd_extxrgb_gray_convert_dspr2; |
| 269 | break; |
| 270 | default: |
| 271 | dspr2fct = jsimd_extrgb_gray_convert_dspr2; |
| 272 | break; |
| 273 | } |
| 274 | |
| 275 | dspr2fct(cinfo->image_width, input_buf, output_buf, output_row, num_rows); |
| 276 | } |
| 277 | |
| 278 | GLOBAL(void) |
| 279 | jsimd_ycc_rgb_convert(j_decompress_ptr cinfo, JSAMPIMAGE input_buf, |
| 280 | JDIMENSION input_row, JSAMPARRAY output_buf, |
| 281 | int num_rows) |
| 282 | { |
| 283 | void (*dspr2fct) (JDIMENSION, JSAMPIMAGE, JDIMENSION, JSAMPARRAY, int); |
| 284 | |
| 285 | switch (cinfo->out_color_space) { |
| 286 | case JCS_EXT_RGB: |
| 287 | dspr2fct = jsimd_ycc_extrgb_convert_dspr2; |
| 288 | break; |
| 289 | case JCS_EXT_RGBX: |
| 290 | case JCS_EXT_RGBA: |
| 291 | dspr2fct = jsimd_ycc_extrgbx_convert_dspr2; |
| 292 | break; |
| 293 | case JCS_EXT_BGR: |
| 294 | dspr2fct = jsimd_ycc_extbgr_convert_dspr2; |
| 295 | break; |
| 296 | case JCS_EXT_BGRX: |
| 297 | case JCS_EXT_BGRA: |
| 298 | dspr2fct = jsimd_ycc_extbgrx_convert_dspr2; |
| 299 | break; |
| 300 | case JCS_EXT_XBGR: |
| 301 | case JCS_EXT_ABGR: |
| 302 | dspr2fct = jsimd_ycc_extxbgr_convert_dspr2; |
| 303 | break; |
| 304 | case JCS_EXT_XRGB: |
| 305 | case JCS_EXT_ARGB: |
| 306 | dspr2fct = jsimd_ycc_extxrgb_convert_dspr2; |
| 307 | break; |
| 308 | default: |
| 309 | dspr2fct = jsimd_ycc_extrgb_convert_dspr2; |
| 310 | break; |
| 311 | } |
| 312 | |
| 313 | dspr2fct(cinfo->output_width, input_buf, input_row, output_buf, num_rows); |
| 314 | } |
| 315 | |
| 316 | GLOBAL(void) |
| 317 | jsimd_ycc_rgb565_convert(j_decompress_ptr cinfo, JSAMPIMAGE input_buf, |
| 318 | JDIMENSION input_row, JSAMPARRAY output_buf, |
| 319 | int num_rows) |
| 320 | { |
| 321 | } |
| 322 | |
| 323 | GLOBAL(void) |
| 324 | jsimd_c_null_convert(j_compress_ptr cinfo, JSAMPARRAY input_buf, |
| 325 | JSAMPIMAGE output_buf, JDIMENSION output_row, |
| 326 | int num_rows) |
| 327 | { |
| 328 | jsimd_c_null_convert_dspr2(cinfo->image_width, input_buf, output_buf, |
| 329 | output_row, num_rows, cinfo->num_components); |
| 330 | } |
| 331 | |
| 332 | GLOBAL(int) |
| 333 | jsimd_can_h2v2_downsample(void) |
| 334 | { |
| 335 | init_simd(); |
| 336 | |
| 337 | /* The code is optimised for these values only */ |
| 338 | if (BITS_IN_JSAMPLE != 8) |
| 339 | return 0; |
| 340 | if (sizeof(JDIMENSION) != 4) |
| 341 | return 0; |
| 342 | |
| 343 | if (simd_support & JSIMD_DSPR2) |
| 344 | return 1; |
| 345 | |
| 346 | return 0; |
| 347 | } |
| 348 | |
| 349 | GLOBAL(int) |
| 350 | jsimd_can_h2v2_smooth_downsample(void) |
| 351 | { |
| 352 | init_simd(); |
| 353 | |
| 354 | /* The code is optimised for these values only */ |
| 355 | if (BITS_IN_JSAMPLE != 8) |
| 356 | return 0; |
| 357 | if (sizeof(JDIMENSION) != 4) |
| 358 | return 0; |
| 359 | if (DCTSIZE != 8) |
| 360 | return 0; |
| 361 | |
| 362 | if (simd_support & JSIMD_DSPR2) |
| 363 | return 1; |
| 364 | |
| 365 | return 0; |
| 366 | } |
| 367 | |
| 368 | GLOBAL(int) |
| 369 | jsimd_can_h2v1_downsample(void) |
| 370 | { |
| 371 | init_simd(); |
| 372 | |
| 373 | /* The code is optimised for these values only */ |
| 374 | if (BITS_IN_JSAMPLE != 8) |
| 375 | return 0; |
| 376 | if (sizeof(JDIMENSION) != 4) |
| 377 | return 0; |
| 378 | |
| 379 | if (simd_support & JSIMD_DSPR2) |
| 380 | return 1; |
| 381 | |
| 382 | return 0; |
| 383 | } |
| 384 | |
| 385 | GLOBAL(void) |
| 386 | jsimd_h2v2_downsample(j_compress_ptr cinfo, jpeg_component_info *compptr, |
| 387 | JSAMPARRAY input_data, JSAMPARRAY output_data) |
| 388 | { |
| 389 | jsimd_h2v2_downsample_dspr2(cinfo->image_width, cinfo->max_v_samp_factor, |
| 390 | compptr->v_samp_factor, compptr->width_in_blocks, |
| 391 | input_data, output_data); |
| 392 | } |
| 393 | |
| 394 | GLOBAL(void) |
| 395 | jsimd_h2v2_smooth_downsample(j_compress_ptr cinfo, |
| 396 | jpeg_component_info *compptr, |
| 397 | JSAMPARRAY input_data, JSAMPARRAY output_data) |
| 398 | { |
| 399 | jsimd_h2v2_smooth_downsample_dspr2(input_data, output_data, |
| 400 | compptr->v_samp_factor, |
| 401 | cinfo->max_v_samp_factor, |
| 402 | cinfo->smoothing_factor, |
| 403 | compptr->width_in_blocks, |
| 404 | cinfo->image_width); |
| 405 | } |
| 406 | |
| 407 | GLOBAL(void) |
| 408 | jsimd_h2v1_downsample(j_compress_ptr cinfo, jpeg_component_info *compptr, |
| 409 | JSAMPARRAY input_data, JSAMPARRAY output_data) |
| 410 | { |
| 411 | jsimd_h2v1_downsample_dspr2(cinfo->image_width, cinfo->max_v_samp_factor, |
| 412 | compptr->v_samp_factor, compptr->width_in_blocks, |
| 413 | input_data, output_data); |
| 414 | } |
| 415 | |
| 416 | GLOBAL(int) |
| 417 | jsimd_can_h2v2_upsample(void) |
| 418 | { |
| 419 | init_simd(); |
| 420 | |
| 421 | /* The code is optimised for these values only */ |
| 422 | if (BITS_IN_JSAMPLE != 8) |
| 423 | return 0; |
| 424 | if (sizeof(JDIMENSION) != 4) |
| 425 | return 0; |
| 426 | |
| 427 | if (simd_support & JSIMD_DSPR2) |
| 428 | return 1; |
| 429 | |
| 430 | return 0; |
| 431 | } |
| 432 | |
| 433 | GLOBAL(int) |
| 434 | jsimd_can_h2v1_upsample(void) |
| 435 | { |
| 436 | init_simd(); |
| 437 | |
| 438 | /* The code is optimised for these values only */ |
| 439 | if (BITS_IN_JSAMPLE != 8) |
| 440 | return 0; |
| 441 | if (sizeof(JDIMENSION) != 4) |
| 442 | return 0; |
| 443 | |
| 444 | if (simd_support & JSIMD_DSPR2) |
| 445 | return 1; |
| 446 | |
| 447 | return 0; |
| 448 | } |
| 449 | |
| 450 | GLOBAL(int) |
| 451 | jsimd_can_int_upsample(void) |
| 452 | { |
| 453 | init_simd(); |
| 454 | |
| 455 | /* The code is optimised for these values only */ |
| 456 | if (BITS_IN_JSAMPLE != 8) |
| 457 | return 0; |
| 458 | if (sizeof(JDIMENSION) != 4) |
| 459 | return 0; |
| 460 | |
| 461 | if (simd_support & JSIMD_DSPR2) |
| 462 | return 1; |
| 463 | |
| 464 | return 0; |
| 465 | } |
| 466 | |
| 467 | GLOBAL(void) |
| 468 | jsimd_h2v2_upsample(j_decompress_ptr cinfo, jpeg_component_info *compptr, |
| 469 | JSAMPARRAY input_data, JSAMPARRAY *output_data_ptr) |
| 470 | { |
| 471 | jsimd_h2v2_upsample_dspr2(cinfo->max_v_samp_factor, cinfo->output_width, |
| 472 | input_data, output_data_ptr); |
| 473 | } |
| 474 | |
| 475 | GLOBAL(void) |
| 476 | jsimd_h2v1_upsample(j_decompress_ptr cinfo, jpeg_component_info *compptr, |
| 477 | JSAMPARRAY input_data, JSAMPARRAY *output_data_ptr) |
| 478 | { |
| 479 | jsimd_h2v1_upsample_dspr2(cinfo->max_v_samp_factor, cinfo->output_width, |
| 480 | input_data, output_data_ptr); |
| 481 | } |
| 482 | |
| 483 | GLOBAL(void) |
| 484 | jsimd_int_upsample(j_decompress_ptr cinfo, jpeg_component_info *compptr, |
| 485 | JSAMPARRAY input_data, JSAMPARRAY *output_data_ptr) |
| 486 | { |
| 487 | my_upsample_ptr upsample = (my_upsample_ptr)cinfo->upsample; |
| 488 | |
| 489 | jsimd_int_upsample_dspr2(upsample->h_expand[compptr->component_index], |
| 490 | upsample->v_expand[compptr->component_index], |
| 491 | input_data, output_data_ptr, cinfo->output_width, |
| 492 | cinfo->max_v_samp_factor); |
| 493 | } |
| 494 | |
| 495 | GLOBAL(int) |
| 496 | jsimd_can_h2v2_fancy_upsample(void) |
| 497 | { |
| 498 | init_simd(); |
| 499 | |
| 500 | /* The code is optimised for these values only */ |
| 501 | if (BITS_IN_JSAMPLE != 8) |
| 502 | return 0; |
| 503 | if (sizeof(JDIMENSION) != 4) |
| 504 | return 0; |
| 505 | |
| 506 | if (simd_support & JSIMD_DSPR2) |
| 507 | return 1; |
| 508 | |
| 509 | return 0; |
| 510 | } |
| 511 | |
| 512 | GLOBAL(int) |
| 513 | jsimd_can_h2v1_fancy_upsample(void) |
| 514 | { |
| 515 | init_simd(); |
| 516 | |
| 517 | /* The code is optimised for these values only */ |
| 518 | if (BITS_IN_JSAMPLE != 8) |
| 519 | return 0; |
| 520 | if (sizeof(JDIMENSION) != 4) |
| 521 | return 0; |
| 522 | |
| 523 | if (simd_support & JSIMD_DSPR2) |
| 524 | return 1; |
| 525 | |
| 526 | return 0; |
| 527 | } |
| 528 | |
| 529 | GLOBAL(void) |
| 530 | jsimd_h2v2_fancy_upsample(j_decompress_ptr cinfo, jpeg_component_info *compptr, |
| 531 | JSAMPARRAY input_data, JSAMPARRAY *output_data_ptr) |
| 532 | { |
| 533 | jsimd_h2v2_fancy_upsample_dspr2(cinfo->max_v_samp_factor, |
| 534 | compptr->downsampled_width, input_data, |
| 535 | output_data_ptr); |
| 536 | } |
| 537 | |
| 538 | GLOBAL(void) |
| 539 | jsimd_h2v1_fancy_upsample(j_decompress_ptr cinfo, jpeg_component_info *compptr, |
| 540 | JSAMPARRAY input_data, JSAMPARRAY *output_data_ptr) |
| 541 | { |
| 542 | jsimd_h2v1_fancy_upsample_dspr2(cinfo->max_v_samp_factor, |
| 543 | compptr->downsampled_width, input_data, |
| 544 | output_data_ptr); |
| 545 | } |
| 546 | |
| 547 | GLOBAL(int) |
| 548 | jsimd_can_h2v2_merged_upsample(void) |
| 549 | { |
| 550 | init_simd(); |
| 551 | |
| 552 | /* The code is optimised for these values only */ |
| 553 | if (BITS_IN_JSAMPLE != 8) |
| 554 | return 0; |
| 555 | if (sizeof(JDIMENSION) != 4) |
| 556 | return 0; |
| 557 | |
| 558 | if (simd_support & JSIMD_DSPR2) |
| 559 | return 1; |
| 560 | |
| 561 | return 0; |
| 562 | } |
| 563 | |
| 564 | GLOBAL(int) |
| 565 | jsimd_can_h2v1_merged_upsample(void) |
| 566 | { |
| 567 | init_simd(); |
| 568 | |
| 569 | /* The code is optimised for these values only */ |
| 570 | if (BITS_IN_JSAMPLE != 8) |
| 571 | return 0; |
| 572 | if (sizeof(JDIMENSION) != 4) |
| 573 | return 0; |
| 574 | |
| 575 | if (simd_support & JSIMD_DSPR2) |
| 576 | return 1; |
| 577 | |
| 578 | return 0; |
| 579 | } |
| 580 | |
| 581 | GLOBAL(void) |
| 582 | jsimd_h2v2_merged_upsample(j_decompress_ptr cinfo, JSAMPIMAGE input_buf, |
| 583 | JDIMENSION in_row_group_ctr, JSAMPARRAY output_buf) |
| 584 | { |
| 585 | void (*dspr2fct) (JDIMENSION, JSAMPIMAGE, JDIMENSION, JSAMPARRAY, JSAMPLE *); |
| 586 | |
| 587 | switch (cinfo->out_color_space) { |
| 588 | case JCS_EXT_RGB: |
| 589 | dspr2fct = jsimd_h2v2_extrgb_merged_upsample_dspr2; |
| 590 | break; |
| 591 | case JCS_EXT_RGBX: |
| 592 | case JCS_EXT_RGBA: |
| 593 | dspr2fct = jsimd_h2v2_extrgbx_merged_upsample_dspr2; |
| 594 | break; |
| 595 | case JCS_EXT_BGR: |
| 596 | dspr2fct = jsimd_h2v2_extbgr_merged_upsample_dspr2; |
| 597 | break; |
| 598 | case JCS_EXT_BGRX: |
| 599 | case JCS_EXT_BGRA: |
| 600 | dspr2fct = jsimd_h2v2_extbgrx_merged_upsample_dspr2; |
| 601 | break; |
| 602 | case JCS_EXT_XBGR: |
| 603 | case JCS_EXT_ABGR: |
| 604 | dspr2fct = jsimd_h2v2_extxbgr_merged_upsample_dspr2; |
| 605 | break; |
| 606 | case JCS_EXT_XRGB: |
| 607 | case JCS_EXT_ARGB: |
| 608 | dspr2fct = jsimd_h2v2_extxrgb_merged_upsample_dspr2; |
| 609 | break; |
| 610 | default: |
| 611 | dspr2fct = jsimd_h2v2_extrgb_merged_upsample_dspr2; |
| 612 | break; |
| 613 | } |
| 614 | |
| 615 | dspr2fct(cinfo->output_width, input_buf, in_row_group_ctr, output_buf, |
| 616 | cinfo->sample_range_limit); |
| 617 | } |
| 618 | |
| 619 | GLOBAL(void) |
| 620 | jsimd_h2v1_merged_upsample(j_decompress_ptr cinfo, JSAMPIMAGE input_buf, |
| 621 | JDIMENSION in_row_group_ctr, JSAMPARRAY output_buf) |
| 622 | { |
| 623 | void (*dspr2fct) (JDIMENSION, JSAMPIMAGE, JDIMENSION, JSAMPARRAY, JSAMPLE *); |
| 624 | |
| 625 | switch (cinfo->out_color_space) { |
| 626 | case JCS_EXT_RGB: |
| 627 | dspr2fct = jsimd_h2v1_extrgb_merged_upsample_dspr2; |
| 628 | break; |
| 629 | case JCS_EXT_RGBX: |
| 630 | case JCS_EXT_RGBA: |
| 631 | dspr2fct = jsimd_h2v1_extrgbx_merged_upsample_dspr2; |
| 632 | break; |
| 633 | case JCS_EXT_BGR: |
| 634 | dspr2fct = jsimd_h2v1_extbgr_merged_upsample_dspr2; |
| 635 | break; |
| 636 | case JCS_EXT_BGRX: |
| 637 | case JCS_EXT_BGRA: |
| 638 | dspr2fct = jsimd_h2v1_extbgrx_merged_upsample_dspr2; |
| 639 | break; |
| 640 | case JCS_EXT_XBGR: |
| 641 | case JCS_EXT_ABGR: |
| 642 | dspr2fct = jsimd_h2v1_extxbgr_merged_upsample_dspr2; |
| 643 | break; |
| 644 | case JCS_EXT_XRGB: |
| 645 | case JCS_EXT_ARGB: |
| 646 | dspr2fct = jsimd_h2v1_extxrgb_merged_upsample_dspr2; |
| 647 | break; |
| 648 | default: |
| 649 | dspr2fct = jsimd_h2v1_extrgb_merged_upsample_dspr2; |
| 650 | break; |
| 651 | } |
| 652 | |
| 653 | dspr2fct(cinfo->output_width, input_buf, in_row_group_ctr, output_buf, |
| 654 | cinfo->sample_range_limit); |
| 655 | } |
| 656 | |
| 657 | GLOBAL(int) |
| 658 | jsimd_can_convsamp(void) |
| 659 | { |
| 660 | init_simd(); |
| 661 | |
| 662 | /* The code is optimised for these values only */ |
| 663 | if (DCTSIZE != 8) |
| 664 | return 0; |
| 665 | if (BITS_IN_JSAMPLE != 8) |
| 666 | return 0; |
| 667 | if (sizeof(JDIMENSION) != 4) |
| 668 | return 0; |
| 669 | if (sizeof(DCTELEM) != 2) |
| 670 | return 0; |
| 671 | |
| 672 | if (simd_support & JSIMD_DSPR2) |
| 673 | return 1; |
| 674 | |
| 675 | return 0; |
| 676 | } |
| 677 | |
| 678 | GLOBAL(int) |
| 679 | jsimd_can_convsamp_float(void) |
| 680 | { |
| 681 | init_simd(); |
| 682 | |
| 683 | /* The code is optimised for these values only */ |
| 684 | if (DCTSIZE != 8) |
| 685 | return 0; |
| 686 | if (sizeof(JCOEF) != 2) |
| 687 | return 0; |
| 688 | if (BITS_IN_JSAMPLE != 8) |
| 689 | return 0; |
| 690 | if (sizeof(JDIMENSION) != 4) |
| 691 | return 0; |
| 692 | if (sizeof(ISLOW_MULT_TYPE) != 2) |
| 693 | return 0; |
| 694 | |
| 695 | if (simd_support & JSIMD_DSPR2) |
| 696 | return 1; |
| 697 | |
| 698 | return 0; |
| 699 | } |
| 700 | |
| 701 | GLOBAL(void) |
| 702 | jsimd_convsamp(JSAMPARRAY sample_data, JDIMENSION start_col, |
| 703 | DCTELEM *workspace) |
| 704 | { |
| 705 | jsimd_convsamp_dspr2(sample_data, start_col, workspace); |
| 706 | } |
| 707 | |
| 708 | GLOBAL(void) |
| 709 | jsimd_convsamp_float(JSAMPARRAY sample_data, JDIMENSION start_col, |
| 710 | FAST_FLOAT *workspace) |
| 711 | { |
| 712 | jsimd_convsamp_float_dspr2(sample_data, start_col, workspace); |
| 713 | } |
| 714 | |
| 715 | GLOBAL(int) |
| 716 | jsimd_can_fdct_islow(void) |
| 717 | { |
| 718 | init_simd(); |
| 719 | |
| 720 | /* The code is optimised for these values only */ |
| 721 | if (DCTSIZE != 8) |
| 722 | return 0; |
| 723 | if (sizeof(DCTELEM) != 2) |
| 724 | return 0; |
| 725 | |
| 726 | if (simd_support & JSIMD_DSPR2) |
| 727 | return 1; |
| 728 | |
| 729 | return 0; |
| 730 | } |
| 731 | |
| 732 | GLOBAL(int) |
| 733 | jsimd_can_fdct_ifast(void) |
| 734 | { |
| 735 | init_simd(); |
| 736 | |
| 737 | /* The code is optimised for these values only */ |
| 738 | if (DCTSIZE != 8) |
| 739 | return 0; |
| 740 | if (sizeof(DCTELEM) != 2) |
| 741 | return 0; |
| 742 | |
| 743 | if (simd_support & JSIMD_DSPR2) |
| 744 | return 1; |
| 745 | |
| 746 | return 0; |
| 747 | } |
| 748 | |
| 749 | GLOBAL(int) |
| 750 | jsimd_can_fdct_float(void) |
| 751 | { |
| 752 | return 0; |
| 753 | } |
| 754 | |
| 755 | GLOBAL(void) |
| 756 | jsimd_fdct_islow(DCTELEM *data) |
| 757 | { |
| 758 | jsimd_fdct_islow_dspr2(data); |
| 759 | } |
| 760 | |
| 761 | GLOBAL(void) |
| 762 | jsimd_fdct_ifast(DCTELEM *data) |
| 763 | { |
| 764 | jsimd_fdct_ifast_dspr2(data); |
| 765 | } |
| 766 | |
| 767 | GLOBAL(void) |
| 768 | jsimd_fdct_float(FAST_FLOAT *data) |
| 769 | { |
| 770 | } |
| 771 | |
| 772 | GLOBAL(int) |
| 773 | jsimd_can_quantize(void) |
| 774 | { |
| 775 | init_simd(); |
| 776 | |
| 777 | /* The code is optimised for these values only */ |
| 778 | if (DCTSIZE != 8) |
| 779 | return 0; |
| 780 | if (sizeof(JCOEF) != 2) |
| 781 | return 0; |
| 782 | if (sizeof(DCTELEM) != 2) |
| 783 | return 0; |
| 784 | |
| 785 | if (simd_support & JSIMD_DSPR2) |
| 786 | return 1; |
| 787 | |
| 788 | return 0; |
| 789 | } |
| 790 | |
| 791 | GLOBAL(int) |
| 792 | jsimd_can_quantize_float(void) |
| 793 | { |
| 794 | init_simd(); |
| 795 | |
| 796 | /* The code is optimised for these values only */ |
| 797 | if (DCTSIZE != 8) |
| 798 | return 0; |
| 799 | if (sizeof(JCOEF) != 2) |
| 800 | return 0; |
| 801 | if (BITS_IN_JSAMPLE != 8) |
| 802 | return 0; |
| 803 | if (sizeof(JDIMENSION) != 4) |
| 804 | return 0; |
| 805 | if (sizeof(ISLOW_MULT_TYPE) != 2) |
| 806 | return 0; |
| 807 | |
| 808 | if (simd_support & JSIMD_DSPR2) |
| 809 | return 1; |
| 810 | |
| 811 | return 0; |
| 812 | } |
| 813 | |
| 814 | GLOBAL(void) |
| 815 | jsimd_quantize(JCOEFPTR coef_block, DCTELEM *divisors, DCTELEM *workspace) |
| 816 | { |
| 817 | jsimd_quantize_dspr2(coef_block, divisors, workspace); |
| 818 | } |
| 819 | |
| 820 | GLOBAL(void) |
| 821 | jsimd_quantize_float(JCOEFPTR coef_block, FAST_FLOAT *divisors, |
| 822 | FAST_FLOAT *workspace) |
| 823 | { |
| 824 | jsimd_quantize_float_dspr2(coef_block, divisors, workspace); |
| 825 | } |
| 826 | |
| 827 | GLOBAL(int) |
| 828 | jsimd_can_idct_2x2(void) |
| 829 | { |
| 830 | init_simd(); |
| 831 | |
| 832 | /* The code is optimised for these values only */ |
| 833 | if (DCTSIZE != 8) |
| 834 | return 0; |
| 835 | if (sizeof(JCOEF) != 2) |
| 836 | return 0; |
| 837 | if (BITS_IN_JSAMPLE != 8) |
| 838 | return 0; |
| 839 | if (sizeof(JDIMENSION) != 4) |
| 840 | return 0; |
| 841 | if (sizeof(ISLOW_MULT_TYPE) != 2) |
| 842 | return 0; |
| 843 | |
| 844 | if (simd_support & JSIMD_DSPR2) |
| 845 | return 1; |
| 846 | |
| 847 | return 0; |
| 848 | } |
| 849 | |
| 850 | GLOBAL(int) |
| 851 | jsimd_can_idct_4x4(void) |
| 852 | { |
| 853 | init_simd(); |
| 854 | |
| 855 | /* The code is optimised for these values only */ |
| 856 | if (DCTSIZE != 8) |
| 857 | return 0; |
| 858 | if (sizeof(JCOEF) != 2) |
| 859 | return 0; |
| 860 | if (BITS_IN_JSAMPLE != 8) |
| 861 | return 0; |
| 862 | if (sizeof(JDIMENSION) != 4) |
| 863 | return 0; |
| 864 | if (sizeof(ISLOW_MULT_TYPE) != 2) |
| 865 | return 0; |
| 866 | |
| 867 | if (simd_support & JSIMD_DSPR2) |
| 868 | return 1; |
| 869 | |
| 870 | return 0; |
| 871 | } |
| 872 | |
| 873 | GLOBAL(int) |
| 874 | jsimd_can_idct_6x6(void) |
| 875 | { |
| 876 | init_simd(); |
| 877 | |
| 878 | /* The code is optimised for these values only */ |
| 879 | if (DCTSIZE != 8) |
| 880 | return 0; |
| 881 | if (sizeof(JCOEF) != 2) |
| 882 | return 0; |
| 883 | if (BITS_IN_JSAMPLE != 8) |
| 884 | return 0; |
| 885 | if (sizeof(JDIMENSION) != 4) |
| 886 | return 0; |
| 887 | if (sizeof(ISLOW_MULT_TYPE) != 2) |
| 888 | return 0; |
| 889 | |
| 890 | if (simd_support & JSIMD_DSPR2) |
| 891 | return 1; |
| 892 | |
| 893 | return 0; |
| 894 | } |
| 895 | |
| 896 | GLOBAL(int) |
| 897 | jsimd_can_idct_12x12(void) |
| 898 | { |
| 899 | init_simd(); |
| 900 | |
| 901 | if (BITS_IN_JSAMPLE != 8) |
| 902 | return 0; |
| 903 | if (DCTSIZE != 8) |
| 904 | return 0; |
| 905 | if (sizeof(JCOEF) != 2) |
| 906 | return 0; |
| 907 | if (sizeof(JDIMENSION) != 4) |
| 908 | return 0; |
| 909 | if (sizeof(ISLOW_MULT_TYPE) != 2) |
| 910 | return 0; |
| 911 | |
| 912 | if (simd_support & JSIMD_DSPR2) |
| 913 | return 1; |
| 914 | |
| 915 | return 0; |
| 916 | } |
| 917 | |
| 918 | GLOBAL(void) |
| 919 | jsimd_idct_2x2(j_decompress_ptr cinfo, jpeg_component_info *compptr, |
| 920 | JCOEFPTR coef_block, JSAMPARRAY output_buf, |
| 921 | JDIMENSION output_col) |
| 922 | { |
| 923 | jsimd_idct_2x2_dspr2(compptr->dct_table, coef_block, output_buf, output_col); |
| 924 | } |
| 925 | |
| 926 | GLOBAL(void) |
| 927 | jsimd_idct_4x4(j_decompress_ptr cinfo, jpeg_component_info *compptr, |
| 928 | JCOEFPTR coef_block, JSAMPARRAY output_buf, |
| 929 | JDIMENSION output_col) |
| 930 | { |
| 931 | int workspace[DCTSIZE * 4]; /* buffers data between passes */ |
| 932 | |
| 933 | jsimd_idct_4x4_dspr2(compptr->dct_table, coef_block, output_buf, output_col, |
| 934 | workspace); |
| 935 | } |
| 936 | |
| 937 | GLOBAL(void) |
| 938 | jsimd_idct_6x6(j_decompress_ptr cinfo, jpeg_component_info *compptr, |
| 939 | JCOEFPTR coef_block, JSAMPARRAY output_buf, |
| 940 | JDIMENSION output_col) |
| 941 | { |
| 942 | jsimd_idct_6x6_dspr2(compptr->dct_table, coef_block, output_buf, output_col); |
| 943 | } |
| 944 | |
| 945 | GLOBAL(void) |
| 946 | jsimd_idct_12x12(j_decompress_ptr cinfo, jpeg_component_info *compptr, |
| 947 | JCOEFPTR coef_block, JSAMPARRAY output_buf, |
| 948 | JDIMENSION output_col) |
| 949 | { |
| 950 | int workspace[96]; |
| 951 | int output[12] = { |
| 952 | (int)(output_buf[0] + output_col), |
| 953 | (int)(output_buf[1] + output_col), |
| 954 | (int)(output_buf[2] + output_col), |
| 955 | (int)(output_buf[3] + output_col), |
| 956 | (int)(output_buf[4] + output_col), |
| 957 | (int)(output_buf[5] + output_col), |
| 958 | (int)(output_buf[6] + output_col), |
| 959 | (int)(output_buf[7] + output_col), |
| 960 | (int)(output_buf[8] + output_col), |
| 961 | (int)(output_buf[9] + output_col), |
| 962 | (int)(output_buf[10] + output_col), |
| 963 | (int)(output_buf[11] + output_col) |
| 964 | }; |
| 965 | |
| 966 | jsimd_idct_12x12_pass1_dspr2(coef_block, compptr->dct_table, workspace); |
| 967 | jsimd_idct_12x12_pass2_dspr2(workspace, output); |
| 968 | } |
| 969 | |
| 970 | GLOBAL(int) |
| 971 | jsimd_can_idct_islow(void) |
| 972 | { |
| 973 | init_simd(); |
| 974 | |
| 975 | /* The code is optimised for these values only */ |
| 976 | if (DCTSIZE != 8) |
| 977 | return 0; |
| 978 | if (sizeof(JCOEF) != 2) |
| 979 | return 0; |
| 980 | if (BITS_IN_JSAMPLE != 8) |
| 981 | return 0; |
| 982 | if (sizeof(JDIMENSION) != 4) |
| 983 | return 0; |
| 984 | if (sizeof(ISLOW_MULT_TYPE) != 2) |
| 985 | return 0; |
| 986 | |
| 987 | if (simd_support & JSIMD_DSPR2) |
| 988 | return 1; |
| 989 | |
| 990 | return 0; |
| 991 | } |
| 992 | |
| 993 | GLOBAL(int) |
| 994 | jsimd_can_idct_ifast(void) |
| 995 | { |
| 996 | init_simd(); |
| 997 | |
| 998 | /* The code is optimised for these values only */ |
| 999 | if (DCTSIZE != 8) |
| 1000 | return 0; |
| 1001 | if (sizeof(JCOEF) != 2) |
| 1002 | return 0; |
| 1003 | if (BITS_IN_JSAMPLE != 8) |
| 1004 | return 0; |
| 1005 | if (sizeof(JDIMENSION) != 4) |
| 1006 | return 0; |
| 1007 | if (sizeof(IFAST_MULT_TYPE) != 2) |
| 1008 | return 0; |
| 1009 | if (IFAST_SCALE_BITS != 2) |
| 1010 | return 0; |
| 1011 | |
| 1012 | if (simd_support & JSIMD_DSPR2) |
| 1013 | return 1; |
| 1014 | |
| 1015 | return 0; |
| 1016 | } |
| 1017 | |
| 1018 | GLOBAL(int) |
| 1019 | jsimd_can_idct_float(void) |
| 1020 | { |
| 1021 | return 0; |
| 1022 | } |
| 1023 | |
| 1024 | GLOBAL(void) |
| 1025 | jsimd_idct_islow(j_decompress_ptr cinfo, jpeg_component_info *compptr, |
| 1026 | JCOEFPTR coef_block, JSAMPARRAY output_buf, |
| 1027 | JDIMENSION output_col) |
| 1028 | { |
| 1029 | int output[8] = { |
| 1030 | (int)(output_buf[0] + output_col), |
| 1031 | (int)(output_buf[1] + output_col), |
| 1032 | (int)(output_buf[2] + output_col), |
| 1033 | (int)(output_buf[3] + output_col), |
| 1034 | (int)(output_buf[4] + output_col), |
| 1035 | (int)(output_buf[5] + output_col), |
| 1036 | (int)(output_buf[6] + output_col), |
| 1037 | (int)(output_buf[7] + output_col) |
| 1038 | }; |
| 1039 | |
| 1040 | jsimd_idct_islow_dspr2(coef_block, compptr->dct_table, output, |
| 1041 | IDCT_range_limit(cinfo)); |
| 1042 | } |
| 1043 | |
| 1044 | GLOBAL(void) |
| 1045 | jsimd_idct_ifast(j_decompress_ptr cinfo, jpeg_component_info *compptr, |
| 1046 | JCOEFPTR coef_block, JSAMPARRAY output_buf, |
| 1047 | JDIMENSION output_col) |
| 1048 | { |
| 1049 | JCOEFPTR inptr; |
| 1050 | IFAST_MULT_TYPE *quantptr; |
| 1051 | DCTELEM workspace[DCTSIZE2]; /* buffers data between passes */ |
| 1052 | |
| 1053 | /* Pass 1: process columns from input, store into work array. */ |
| 1054 | |
| 1055 | inptr = coef_block; |
| 1056 | quantptr = (IFAST_MULT_TYPE *)compptr->dct_table; |
| 1057 | |
| 1058 | jsimd_idct_ifast_cols_dspr2(inptr, quantptr, workspace, |
| 1059 | mips_idct_ifast_coefs); |
| 1060 | |
| 1061 | /* Pass 2: process rows from work array, store into output array. */ |
| 1062 | /* Note that we must descale the results by a factor of 8 == 2**3, */ |
| 1063 | /* and also undo the PASS1_BITS scaling. */ |
| 1064 | |
| 1065 | jsimd_idct_ifast_rows_dspr2(workspace, output_buf, output_col, |
| 1066 | mips_idct_ifast_coefs); |
| 1067 | } |
| 1068 | |
| 1069 | GLOBAL(void) |
| 1070 | jsimd_idct_float(j_decompress_ptr cinfo, jpeg_component_info *compptr, |
| 1071 | JCOEFPTR coef_block, JSAMPARRAY output_buf, |
| 1072 | JDIMENSION output_col) |
| 1073 | { |
| 1074 | } |
| 1075 | |
| 1076 | GLOBAL(int) |
| 1077 | jsimd_can_huff_encode_one_block(void) |
| 1078 | { |
| 1079 | return 0; |
| 1080 | } |
| 1081 | |
| 1082 | GLOBAL(JOCTET *) |
| 1083 | jsimd_huff_encode_one_block(void *state, JOCTET *buffer, JCOEFPTR block, |
| 1084 | int last_dc_val, c_derived_tbl *dctbl, |
| 1085 | c_derived_tbl *actbl) |
| 1086 | { |
| 1087 | return NULL; |
| 1088 | } |
| 1089 | |
| 1090 | GLOBAL(int) |
| 1091 | jsimd_can_encode_mcu_AC_first_prepare(void) |
| 1092 | { |
| 1093 | return 0; |
| 1094 | } |
| 1095 | |
| 1096 | GLOBAL(void) |
| 1097 | jsimd_encode_mcu_AC_first_prepare(const JCOEF *block, |
| 1098 | const int *jpeg_natural_order_start, int Sl, |
| 1099 | int Al, JCOEF *values, size_t *zerobits) |
| 1100 | { |
| 1101 | } |
| 1102 | |
| 1103 | GLOBAL(int) |
| 1104 | jsimd_can_encode_mcu_AC_refine_prepare(void) |
| 1105 | { |
| 1106 | return 0; |
| 1107 | } |
| 1108 | |
| 1109 | GLOBAL(int) |
| 1110 | jsimd_encode_mcu_AC_refine_prepare(const JCOEF *block, |
| 1111 | const int *jpeg_natural_order_start, int Sl, |
| 1112 | int Al, JCOEF *absvalues, size_t *bits) |
| 1113 | { |
| 1114 | return 0; |
| 1115 | } |