vl_mpeg12_decoder.c revision 0a2310b375068694d5700395aededc3fe68a0f3a
1/************************************************************************** 2 * 3 * Copyright 2009 Younes Manton. 4 * All Rights Reserved. 5 * 6 * Permission is hereby granted, free of charge, to any person obtaining a 7 * copy of this software and associated documentation files (the 8 * "Software"), to deal in the Software without restriction, including 9 * without limitation the rights to use, copy, modify, merge, publish, 10 * distribute, sub license, and/or sell copies of the Software, and to 11 * permit persons to whom the Software is furnished to do so, subject to 12 * the following conditions: 13 * 14 * The above copyright notice and this permission notice (including the 15 * next paragraph) shall be included in all copies or substantial portions 16 * of the Software. 17 * 18 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS 19 * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF 20 * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. 21 * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR 22 * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, 23 * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE 24 * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. 25 * 26 **************************************************************************/ 27 28#include <math.h> 29#include <assert.h> 30 31#include <util/u_memory.h> 32#include <util/u_rect.h> 33#include <util/u_video.h> 34 35#include "vl_mpeg12_decoder.h" 36#include "vl_defines.h" 37 38#define SCALE_FACTOR_SNORM (32768.0f / 256.0f) 39#define SCALE_FACTOR_SSCALED (1.0f / 256.0f) 40 41static const unsigned const_empty_block_mask_420[3][2][2] = { 42 { { 0x20, 0x10 }, { 0x08, 0x04 } }, 43 { { 0x02, 0x02 }, { 0x02, 0x02 } }, 44 { { 0x01, 0x01 }, { 0x01, 0x01 } } 45}; 46 47static const enum pipe_format const_idct_source_formats[] = { 48 PIPE_FORMAT_R16G16B16A16_SNORM, 49 PIPE_FORMAT_R16G16B16A16_SSCALED 50}; 51 52static const unsigned num_idct_source_formats = 53 sizeof(const_idct_source_formats) / sizeof(enum pipe_format); 54 55static const enum pipe_format const_idct_intermediate_formats[] = { 56 PIPE_FORMAT_R16G16B16A16_FLOAT, 57 PIPE_FORMAT_R16G16B16A16_SNORM, 58 PIPE_FORMAT_R16G16B16A16_SSCALED, 59 PIPE_FORMAT_R32G32B32A32_FLOAT 60}; 61 62static const unsigned num_idct_intermediate_formats = 63 sizeof(const_idct_intermediate_formats) / sizeof(enum pipe_format); 64 65static const enum pipe_format const_mc_source_formats[] = { 66 PIPE_FORMAT_R16_SNORM, 67 PIPE_FORMAT_R16_SSCALED 68}; 69 70static const unsigned num_mc_source_formats = 71 sizeof(const_mc_source_formats) / sizeof(enum pipe_format); 72 73static void 74map_buffers(struct vl_mpeg12_decoder *ctx, struct vl_mpeg12_buffer *buffer) 75{ 76 struct pipe_sampler_view **sampler_views; 77 struct pipe_resource *tex; 78 unsigned i; 79 80 assert(ctx && buffer); 81 82 if (ctx->base.entrypoint <= PIPE_VIDEO_ENTRYPOINT_IDCT) 83 sampler_views = buffer->idct_source->get_sampler_views(buffer->idct_source); 84 else 85 sampler_views = buffer->mc_source->get_sampler_views(buffer->mc_source); 86 assert(sampler_views); 87 88 for (i = 0; i < VL_MAX_PLANES; ++i) { 89 tex = sampler_views[i]->texture; 90 91 struct pipe_box rect = 92 { 93 0, 0, 0, 94 tex->width0, 95 tex->height0, 96 1 97 }; 98 99 buffer->tex_transfer[i] = ctx->pipe->get_transfer 100 ( 101 ctx->pipe, tex, 102 0, PIPE_TRANSFER_WRITE | PIPE_TRANSFER_DISCARD, 103 &rect 104 ); 105 106 buffer->texels[i] = ctx->pipe->transfer_map(ctx->pipe, buffer->tex_transfer[i]); 107 } 108} 109 110static void 111upload_block(struct vl_mpeg12_buffer *buffer, unsigned plane, unsigned x, unsigned y, short *block) 112{ 113 unsigned tex_pitch; 114 short *texels; 115 116 unsigned i; 117 118 assert(buffer); 119 assert(block); 120 121 tex_pitch = buffer->tex_transfer[plane]->stride / sizeof(short); 122 texels = buffer->texels[plane] + y * tex_pitch * BLOCK_HEIGHT + x * BLOCK_WIDTH; 123 124 for (i = 0; i < BLOCK_HEIGHT; ++i) 125 memcpy(texels + i * tex_pitch, block + i * BLOCK_WIDTH, BLOCK_WIDTH * sizeof(short)); 126} 127 128static void 129upload_buffer(struct vl_mpeg12_decoder *ctx, 130 struct vl_mpeg12_buffer *buffer, 131 struct pipe_mpeg12_macroblock *mb) 132{ 133 short *blocks; 134 unsigned tb, x, y; 135 136 assert(ctx); 137 assert(buffer); 138 assert(mb); 139 140 blocks = mb->blocks; 141 142 for (y = 0; y < 2; ++y) { 143 for (x = 0; x < 2; ++x, ++tb) { 144 if (mb->cbp & (*ctx->empty_block_mask)[0][y][x]) { 145 upload_block(buffer, 0, mb->mbx * 2 + x, mb->mby * 2 + y, blocks); 146 blocks += BLOCK_WIDTH * BLOCK_HEIGHT; 147 } 148 } 149 } 150 151 /* TODO: Implement 422, 444 */ 152 assert(ctx->base.chroma_format == PIPE_VIDEO_CHROMA_FORMAT_420); 153 154 for (tb = 1; tb < 3; ++tb) { 155 if (mb->cbp & (*ctx->empty_block_mask)[tb][0][0]) { 156 upload_block(buffer, tb, mb->mbx, mb->mby, blocks); 157 blocks += BLOCK_WIDTH * BLOCK_HEIGHT; 158 } 159 } 160} 161 162static void 163unmap_buffers(struct vl_mpeg12_decoder *ctx, struct vl_mpeg12_buffer *buffer) 164{ 165 unsigned i; 166 167 assert(ctx && buffer); 168 169 for (i = 0; i < VL_MAX_PLANES; ++i) { 170 ctx->pipe->transfer_unmap(ctx->pipe, buffer->tex_transfer[i]); 171 ctx->pipe->transfer_destroy(ctx->pipe, buffer->tex_transfer[i]); 172 } 173} 174 175static void 176cleanup_idct_buffer(struct vl_mpeg12_buffer *buf) 177{ 178 struct vl_mpeg12_decoder *dec; 179 assert(buf); 180 181 dec = (struct vl_mpeg12_decoder*)buf->base.decoder; 182 assert(dec); 183 184 buf->idct_source->destroy(buf->idct_source); 185 buf->idct_intermediate->destroy(buf->idct_intermediate); 186 vl_idct_cleanup_buffer(&dec->idct_y, &buf->idct[0]); 187 vl_idct_cleanup_buffer(&dec->idct_c, &buf->idct[1]); 188 vl_idct_cleanup_buffer(&dec->idct_c, &buf->idct[2]); 189} 190 191static void 192vl_mpeg12_buffer_destroy(struct pipe_video_decode_buffer *buffer) 193{ 194 struct vl_mpeg12_buffer *buf = (struct vl_mpeg12_buffer*)buffer; 195 struct vl_mpeg12_decoder *dec; 196 unsigned i; 197 198 assert(buf); 199 200 dec = (struct vl_mpeg12_decoder*)buf->base.decoder; 201 assert(dec); 202 203 if (dec->base.entrypoint <= PIPE_VIDEO_ENTRYPOINT_IDCT) 204 cleanup_idct_buffer(buf); 205 206 buf->mc_source->destroy(buf->mc_source); 207 vl_vb_cleanup(&buf->vertex_stream); 208 for (i = 0; i < VL_MAX_PLANES; ++i) 209 vl_mc_cleanup_buffer(&buf->mc[i]); 210 211 FREE(buf); 212} 213 214static void 215vl_mpeg12_buffer_map(struct pipe_video_decode_buffer *buffer) 216{ 217 struct vl_mpeg12_buffer *buf = (struct vl_mpeg12_buffer*)buffer; 218 struct vl_mpeg12_decoder *dec; 219 assert(buf); 220 221 dec = (struct vl_mpeg12_decoder *)buf->base.decoder; 222 assert(dec); 223 224 vl_vb_map(&buf->vertex_stream, dec->pipe); 225 map_buffers(dec, buf); 226} 227 228static void 229vl_mpeg12_buffer_add_macroblocks(struct pipe_video_decode_buffer *buffer, 230 unsigned num_macroblocks, 231 struct pipe_macroblock *macroblocks) 232{ 233 struct pipe_mpeg12_macroblock *mb = (struct pipe_mpeg12_macroblock*)macroblocks; 234 struct vl_mpeg12_buffer *buf = (struct vl_mpeg12_buffer*)buffer; 235 struct vl_mpeg12_decoder *dec; 236 unsigned i; 237 238 assert(buf); 239 240 dec = (struct vl_mpeg12_decoder*)buf->base.decoder; 241 assert(dec); 242 243 assert(num_macroblocks); 244 assert(macroblocks); 245 assert(macroblocks->codec == PIPE_VIDEO_CODEC_MPEG12); 246 247 for ( i = 0; i < num_macroblocks; ++i ) { 248 vl_vb_add_block(&buf->vertex_stream, &mb[i], dec->empty_block_mask); 249 upload_buffer(dec, buf, &mb[i]); 250 } 251} 252 253static void 254vl_mpeg12_buffer_unmap(struct pipe_video_decode_buffer *buffer) 255{ 256 struct vl_mpeg12_buffer *buf = (struct vl_mpeg12_buffer*)buffer; 257 struct vl_mpeg12_decoder *dec; 258 assert(buf); 259 260 dec = (struct vl_mpeg12_decoder *)buf->base.decoder; 261 assert(dec); 262 263 vl_vb_unmap(&buf->vertex_stream, dec->pipe); 264 unmap_buffers(dec, buf); 265} 266 267static void 268vl_mpeg12_destroy(struct pipe_video_decoder *decoder) 269{ 270 struct vl_mpeg12_decoder *dec = (struct vl_mpeg12_decoder*)decoder; 271 unsigned i; 272 273 assert(decoder); 274 275 /* Asserted in softpipe_delete_fs_state() for some reason */ 276 dec->pipe->bind_vs_state(dec->pipe, NULL); 277 dec->pipe->bind_fs_state(dec->pipe, NULL); 278 279 dec->pipe->delete_depth_stencil_alpha_state(dec->pipe, dec->dsa); 280 281 vl_mc_cleanup(&dec->mc_y); 282 vl_mc_cleanup(&dec->mc_c); 283 284 if (dec->base.entrypoint <= PIPE_VIDEO_ENTRYPOINT_IDCT) { 285 vl_idct_cleanup(&dec->idct_y); 286 vl_idct_cleanup(&dec->idct_c); 287 } 288 289 for (i = 0; i < VL_MAX_PLANES; ++i) 290 dec->pipe->delete_vertex_elements_state(dec->pipe, dec->ves_eb[i]); 291 292 for (i = 0; i < 2; ++i) 293 dec->pipe->delete_vertex_elements_state(dec->pipe, dec->ves_mv[i]); 294 295 pipe_resource_reference(&dec->quads.buffer, NULL); 296 297 FREE(dec); 298} 299 300static bool 301init_idct_buffer(struct vl_mpeg12_buffer *buffer) 302{ 303 enum pipe_format formats[3]; 304 305 struct pipe_sampler_view **idct_source_sv, **idct_intermediate_sv; 306 struct pipe_surface **idct_surfaces; 307 308 struct vl_mpeg12_decoder *dec; 309 310 unsigned i; 311 312 assert(buffer); 313 314 dec = (struct vl_mpeg12_decoder*)buffer->base.decoder; 315 316 formats[0] = formats[1] = formats[2] = dec->idct_source_format; 317 buffer->idct_source = vl_video_buffer_init(dec->base.context, dec->pipe, 318 dec->base.width / 4, dec->base.height, 1, 319 dec->base.chroma_format, 320 formats, PIPE_USAGE_STREAM); 321 if (!buffer->idct_source) 322 goto error_source; 323 324 formats[0] = formats[1] = formats[2] = dec->idct_intermediate_format; 325 buffer->idct_intermediate = vl_video_buffer_init(dec->base.context, dec->pipe, 326 dec->base.width / dec->nr_of_idct_render_targets, 327 dec->base.height / 4, dec->nr_of_idct_render_targets, 328 dec->base.chroma_format, 329 formats, PIPE_USAGE_STATIC); 330 331 if (!buffer->idct_intermediate) 332 goto error_intermediate; 333 334 idct_source_sv = buffer->idct_source->get_sampler_views(buffer->idct_source); 335 if (!idct_source_sv) 336 goto error_source_sv; 337 338 idct_intermediate_sv = buffer->idct_intermediate->get_sampler_views(buffer->idct_intermediate); 339 if (!idct_intermediate_sv) 340 goto error_intermediate_sv; 341 342 idct_surfaces = buffer->mc_source->get_surfaces(buffer->mc_source); 343 if (!idct_surfaces) 344 goto error_surfaces; 345 346 for (i = 0; i < 3; ++i) 347 if (!vl_idct_init_buffer(i == 0 ? &dec->idct_y : &dec->idct_c, 348 &buffer->idct[i], idct_source_sv[i], 349 idct_intermediate_sv[i], idct_surfaces[i])) 350 goto error_plane; 351 352 return true; 353 354error_plane: 355 for (; i > 0; --i) 356 vl_idct_cleanup_buffer(i == 1 ? &dec->idct_c : &dec->idct_y, &buffer->idct[i - 1]); 357 358error_surfaces: 359error_intermediate_sv: 360error_source_sv: 361 buffer->idct_intermediate->destroy(buffer->idct_intermediate); 362 363error_intermediate: 364 buffer->idct_source->destroy(buffer->idct_source); 365 366error_source: 367 return false; 368} 369 370static struct pipe_video_decode_buffer * 371vl_mpeg12_create_buffer(struct pipe_video_decoder *decoder) 372{ 373 enum pipe_format formats[3]; 374 375 struct vl_mpeg12_decoder *dec = (struct vl_mpeg12_decoder*)decoder; 376 struct vl_mpeg12_buffer *buffer; 377 378 struct pipe_sampler_view **mc_source_sv; 379 380 assert(dec); 381 382 buffer = CALLOC_STRUCT(vl_mpeg12_buffer); 383 if (buffer == NULL) 384 return NULL; 385 386 buffer->base.decoder = decoder; 387 buffer->base.destroy = vl_mpeg12_buffer_destroy; 388 buffer->base.map = vl_mpeg12_buffer_map; 389 buffer->base.add_macroblocks = vl_mpeg12_buffer_add_macroblocks; 390 buffer->base.unmap = vl_mpeg12_buffer_unmap; 391 392 buffer->vertex_bufs.individual.quad.stride = dec->quads.stride; 393 buffer->vertex_bufs.individual.quad.buffer_offset = dec->quads.buffer_offset; 394 pipe_resource_reference(&buffer->vertex_bufs.individual.quad.buffer, dec->quads.buffer); 395 396 buffer->vertex_bufs.individual.stream = vl_vb_init(&buffer->vertex_stream, dec->pipe, 397 dec->base.width / MACROBLOCK_WIDTH * 398 dec->base.height / MACROBLOCK_HEIGHT); 399 if (!buffer->vertex_bufs.individual.stream.buffer) 400 goto error_vertex_stream; 401 402 formats[0] = formats[1] = formats[2] =dec->mc_source_format; 403 buffer->mc_source = vl_video_buffer_init(dec->base.context, dec->pipe, 404 dec->base.width, dec->base.height, 1, 405 dec->base.chroma_format, 406 formats, PIPE_USAGE_STATIC); 407 408 if (!buffer->mc_source) 409 goto error_mc_source; 410 411 if (dec->base.entrypoint <= PIPE_VIDEO_ENTRYPOINT_IDCT) 412 if (!init_idct_buffer(buffer)) 413 goto error_idct; 414 415 mc_source_sv = buffer->mc_source->get_sampler_views(buffer->mc_source); 416 if (!mc_source_sv) 417 goto error_mc_source_sv; 418 419 if(!vl_mc_init_buffer(&dec->mc_y, &buffer->mc[0], mc_source_sv[0])) 420 goto error_mc_y; 421 422 if(!vl_mc_init_buffer(&dec->mc_c, &buffer->mc[1], mc_source_sv[1])) 423 goto error_mc_cb; 424 425 if(!vl_mc_init_buffer(&dec->mc_c, &buffer->mc[2], mc_source_sv[2])) 426 goto error_mc_cr; 427 428 return &buffer->base; 429 430error_mc_cr: 431 vl_mc_cleanup_buffer(&buffer->mc[1]); 432 433error_mc_cb: 434 vl_mc_cleanup_buffer(&buffer->mc[0]); 435 436error_mc_y: 437error_mc_source_sv: 438 if (dec->base.entrypoint <= PIPE_VIDEO_ENTRYPOINT_IDCT) 439 cleanup_idct_buffer(buffer); 440 441error_idct: 442 buffer->mc_source->destroy(buffer->mc_source); 443 444error_mc_source: 445 vl_vb_cleanup(&buffer->vertex_stream); 446 447error_vertex_stream: 448 FREE(buffer); 449 return NULL; 450} 451 452static void 453vl_mpeg12_decoder_flush_buffer(struct pipe_video_decode_buffer *buffer, 454 struct pipe_video_buffer *refs[2], 455 struct pipe_video_buffer *dst, 456 struct pipe_fence_handle **fence) 457{ 458 struct vl_mpeg12_buffer *buf = (struct vl_mpeg12_buffer *)buffer; 459 struct vl_mpeg12_decoder *dec; 460 461 struct pipe_sampler_view **sv[2]; 462 struct pipe_surface **surfaces; 463 464 unsigned ne_start, ne_num, e_start, e_num; 465 unsigned i, j; 466 467 assert(buf); 468 469 dec = (struct vl_mpeg12_decoder *)buf->base.decoder; 470 assert(dec); 471 472 for (i = 0; i < 2; ++i) 473 sv[i] = refs[i] ? refs[i]->get_sampler_views(refs[i]) : NULL; 474 475 surfaces = dst->get_surfaces(dst); 476 477 vl_vb_restart(&buf->vertex_stream, &ne_start, &ne_num, &e_start, &e_num); 478 479 dec->pipe->set_vertex_buffers(dec->pipe, 2, buf->vertex_bufs.all); 480 481 for (i = 0; i < VL_MAX_PLANES; ++i) { 482 vl_mc_set_surface(&buf->mc[i], surfaces[i]); 483 484 for (j = 0; j < 2; ++j) { 485 if (sv[j] == NULL) continue; 486 487 dec->pipe->bind_vertex_elements_state(dec->pipe, dec->ves_mv[j]); 488 vl_mc_render_ref(&buf->mc[i], sv[j][i], ne_start, ne_num, e_start, e_num); 489 } 490 491 dec->pipe->bind_vertex_elements_state(dec->pipe, dec->ves_eb[i]); 492 493 if (dec->base.entrypoint <= PIPE_VIDEO_ENTRYPOINT_IDCT) 494 vl_idct_flush(i == 0 ? &dec->idct_y : &dec->idct_c, &buf->idct[i], ne_num); 495 496 vl_mc_render_ycbcr(&buf->mc[i], ne_start, ne_num); 497 498 } 499 dec->pipe->flush(dec->pipe, fence); 500} 501 502static void 503vl_mpeg12_decoder_clear_buffer(struct pipe_video_decode_buffer *buffer) 504{ 505 struct vl_mpeg12_buffer *buf = (struct vl_mpeg12_buffer *)buffer; 506 unsigned ne_start, ne_num, e_start, e_num; 507 508 assert(buf); 509 510 vl_vb_restart(&buf->vertex_stream, &ne_start, &ne_num, &e_start, &e_num); 511} 512 513static bool 514init_pipe_state(struct vl_mpeg12_decoder *dec) 515{ 516 struct pipe_depth_stencil_alpha_state dsa; 517 unsigned i; 518 519 assert(dec); 520 521 memset(&dsa, 0, sizeof dsa); 522 dsa.depth.enabled = 0; 523 dsa.depth.writemask = 0; 524 dsa.depth.func = PIPE_FUNC_ALWAYS; 525 for (i = 0; i < 2; ++i) { 526 dsa.stencil[i].enabled = 0; 527 dsa.stencil[i].func = PIPE_FUNC_ALWAYS; 528 dsa.stencil[i].fail_op = PIPE_STENCIL_OP_KEEP; 529 dsa.stencil[i].zpass_op = PIPE_STENCIL_OP_KEEP; 530 dsa.stencil[i].zfail_op = PIPE_STENCIL_OP_KEEP; 531 dsa.stencil[i].valuemask = 0; 532 dsa.stencil[i].writemask = 0; 533 } 534 dsa.alpha.enabled = 0; 535 dsa.alpha.func = PIPE_FUNC_ALWAYS; 536 dsa.alpha.ref_value = 0; 537 dec->dsa = dec->pipe->create_depth_stencil_alpha_state(dec->pipe, &dsa); 538 dec->pipe->bind_depth_stencil_alpha_state(dec->pipe, dec->dsa); 539 540 return true; 541} 542 543static enum pipe_format 544find_first_supported_format(struct vl_mpeg12_decoder *dec, 545 const enum pipe_format formats[], 546 unsigned num_formats, 547 enum pipe_texture_target target) 548{ 549 struct pipe_screen *screen; 550 unsigned i; 551 552 assert(dec); 553 554 screen = dec->pipe->screen; 555 556 for (i = 0; i < num_formats; ++i) 557 if (screen->is_format_supported(dec->pipe->screen, formats[i], target, 1, 558 PIPE_BIND_SAMPLER_VIEW | PIPE_BIND_RENDER_TARGET)) 559 return formats[i]; 560 561 return PIPE_FORMAT_NONE; 562} 563 564static bool 565init_idct(struct vl_mpeg12_decoder *dec, unsigned buffer_width, unsigned buffer_height) 566{ 567 unsigned chroma_width, chroma_height, chroma_blocks_x, chroma_blocks_y; 568 struct pipe_sampler_view *matrix, *transpose; 569 float matrix_scale, transpose_scale; 570 571 dec->nr_of_idct_render_targets = dec->pipe->screen->get_param(dec->pipe->screen, PIPE_CAP_MAX_RENDER_TARGETS); 572 573 // more than 4 render targets usually doesn't makes any seens 574 dec->nr_of_idct_render_targets = MIN2(dec->nr_of_idct_render_targets, 4); 575 576 dec->idct_source_format = find_first_supported_format(dec, const_idct_source_formats, 577 num_idct_source_formats, PIPE_TEXTURE_2D); 578 579 if (dec->idct_source_format == PIPE_FORMAT_NONE) 580 return false; 581 582 dec->idct_intermediate_format = find_first_supported_format(dec, const_idct_intermediate_formats, 583 num_idct_intermediate_formats, PIPE_TEXTURE_3D); 584 585 if (dec->idct_intermediate_format == PIPE_FORMAT_NONE) 586 return false; 587 588 switch (dec->idct_source_format) { 589 case PIPE_FORMAT_R16G16B16A16_SSCALED: 590 matrix_scale = SCALE_FACTOR_SSCALED; 591 break; 592 593 case PIPE_FORMAT_R16G16B16A16_SNORM: 594 matrix_scale = SCALE_FACTOR_SNORM; 595 break; 596 597 default: 598 assert(0); 599 return false; 600 } 601 602 if (dec->idct_intermediate_format == PIPE_FORMAT_R16G16B16A16_FLOAT || 603 dec->idct_intermediate_format == PIPE_FORMAT_R32G32B32A32_FLOAT) 604 transpose_scale = 1.0f; 605 else 606 transpose_scale = matrix_scale = sqrt(matrix_scale); 607 608 if (dec->mc_source_format == PIPE_FORMAT_R16_SSCALED) 609 transpose_scale /= SCALE_FACTOR_SSCALED; 610 611 if (!(matrix = vl_idct_upload_matrix(dec->pipe, matrix_scale))) 612 goto error_matrix; 613 614 if (matrix_scale != transpose_scale) { 615 if (!(transpose = vl_idct_upload_matrix(dec->pipe, transpose_scale))) 616 goto error_transpose; 617 } else 618 pipe_sampler_view_reference(&transpose, matrix); 619 620 if (!vl_idct_init(&dec->idct_y, dec->pipe, buffer_width, buffer_height, 621 2, 2, dec->nr_of_idct_render_targets, matrix, transpose)) 622 goto error_y; 623 624 if (dec->base.chroma_format == PIPE_VIDEO_CHROMA_FORMAT_420) { 625 chroma_width = buffer_width / 2; 626 chroma_height = buffer_height / 2; 627 chroma_blocks_x = 1; 628 chroma_blocks_y = 1; 629 } else if (dec->base.chroma_format == PIPE_VIDEO_CHROMA_FORMAT_422) { 630 chroma_width = buffer_width; 631 chroma_height = buffer_height / 2; 632 chroma_blocks_x = 2; 633 chroma_blocks_y = 1; 634 } else { 635 chroma_width = buffer_width; 636 chroma_height = buffer_height; 637 chroma_blocks_x = 2; 638 chroma_blocks_y = 2; 639 } 640 641 if(!vl_idct_init(&dec->idct_c, dec->pipe, chroma_width, chroma_height, 642 chroma_blocks_x, chroma_blocks_y, 643 dec->nr_of_idct_render_targets, matrix, transpose)) 644 goto error_c; 645 646 pipe_sampler_view_reference(&matrix, NULL); 647 pipe_sampler_view_reference(&transpose, NULL); 648 return true; 649 650error_c: 651 vl_idct_cleanup(&dec->idct_y); 652 653error_y: 654 pipe_sampler_view_reference(&transpose, NULL); 655 656error_transpose: 657 pipe_sampler_view_reference(&matrix, NULL); 658 659error_matrix: 660 return false; 661} 662 663struct pipe_video_decoder * 664vl_create_mpeg12_decoder(struct pipe_video_context *context, 665 struct pipe_context *pipe, 666 enum pipe_video_profile profile, 667 enum pipe_video_entrypoint entrypoint, 668 enum pipe_video_chroma_format chroma_format, 669 unsigned width, unsigned height) 670{ 671 struct vl_mpeg12_decoder *dec; 672 float mc_scale; 673 unsigned i; 674 675 assert(u_reduce_video_profile(profile) == PIPE_VIDEO_CODEC_MPEG12); 676 677 dec = CALLOC_STRUCT(vl_mpeg12_decoder); 678 679 if (!dec) 680 return NULL; 681 682 dec->base.context = context; 683 dec->base.profile = profile; 684 dec->base.entrypoint = entrypoint; 685 dec->base.chroma_format = chroma_format; 686 dec->base.width = width; 687 dec->base.height = height; 688 689 dec->base.destroy = vl_mpeg12_destroy; 690 dec->base.create_buffer = vl_mpeg12_create_buffer; 691 dec->base.flush_buffer = vl_mpeg12_decoder_flush_buffer; 692 dec->base.clear_buffer = vl_mpeg12_decoder_clear_buffer; 693 694 dec->pipe = pipe; 695 696 dec->quads = vl_vb_upload_quads(dec->pipe, 2, 2); 697 for (i = 0; i < VL_MAX_PLANES; ++i) 698 dec->ves_eb[i] = vl_vb_get_ves_eb(dec->pipe, i); 699 700 for (i = 0; i < 2; ++i) 701 dec->ves_mv[i] = vl_vb_get_ves_mv(dec->pipe, i); 702 703 dec->base.width = align(width, MACROBLOCK_WIDTH); 704 dec->base.height = align(height, MACROBLOCK_HEIGHT); 705 706 /* TODO: Implement 422, 444 */ 707 assert(dec->base.chroma_format == PIPE_VIDEO_CHROMA_FORMAT_420); 708 dec->empty_block_mask = &const_empty_block_mask_420; 709 710 dec->mc_source_format = find_first_supported_format(dec, const_mc_source_formats, 711 num_mc_source_formats, PIPE_TEXTURE_3D); 712 713 if (dec->mc_source_format == PIPE_FORMAT_NONE) 714 return NULL; 715 716 if (entrypoint <= PIPE_VIDEO_ENTRYPOINT_IDCT) { 717 if (!init_idct(dec, dec->base.width, dec->base.height)) 718 goto error_idct; 719 if (dec->mc_source_format == PIPE_FORMAT_R16_SSCALED) 720 mc_scale = SCALE_FACTOR_SSCALED; 721 else 722 mc_scale = 1.0f; 723 } else { 724 switch (dec->mc_source_format) { 725 case PIPE_FORMAT_R16_SNORM: 726 mc_scale = SCALE_FACTOR_SNORM; 727 break; 728 729 case PIPE_FORMAT_R16_SSCALED: 730 mc_scale = SCALE_FACTOR_SSCALED; 731 break; 732 733 default: 734 assert(0); 735 return NULL; 736 } 737 } 738 739 if (!vl_mc_init(&dec->mc_y, dec->pipe, dec->base.width, dec->base.height, MACROBLOCK_HEIGHT, mc_scale)) 740 goto error_mc_y; 741 742 // TODO 743 if (!vl_mc_init(&dec->mc_c, dec->pipe, dec->base.width, dec->base.height, BLOCK_HEIGHT, mc_scale)) 744 goto error_mc_c; 745 746 if (!init_pipe_state(dec)) 747 goto error_pipe_state; 748 749 return &dec->base; 750 751error_pipe_state: 752 vl_mc_cleanup(&dec->mc_c); 753 754error_mc_c: 755 vl_mc_cleanup(&dec->mc_y); 756 757error_mc_y: 758 if (entrypoint <= PIPE_VIDEO_ENTRYPOINT_IDCT) { 759 vl_idct_cleanup(&dec->idct_y); 760 vl_idct_cleanup(&dec->idct_c); 761 } 762 763error_idct: 764 FREE(dec); 765 return NULL; 766} 767