src/gallium/auxiliary/vl/vl_mpeg12_decoder.c

   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
  41 static const enum pipe_format const_zscan_source_formats[] = {
  42    PIPE_FORMAT_R16_SNORM,
  43    PIPE_FORMAT_R16_SSCALED
  44 };
  45
  46 static const unsigned num_zscan_source_formats =
  47    sizeof(const_zscan_source_formats) / sizeof(enum pipe_format);
  48
  49 static const enum pipe_format const_idct_source_formats[] = {
  50    PIPE_FORMAT_R16G16B16A16_SNORM,
  51    PIPE_FORMAT_R16G16B16A16_SSCALED
  52 };
  53
  54 static const unsigned num_idct_source_formats =
  55    sizeof(const_idct_source_formats) / sizeof(enum pipe_format);
  56
  57 static const enum pipe_format const_idct_intermediate_formats[] = {
  58    PIPE_FORMAT_R16G16B16A16_FLOAT,
  59    PIPE_FORMAT_R16G16B16A16_SNORM,
  60    PIPE_FORMAT_R16G16B16A16_SSCALED,
  61    PIPE_FORMAT_R32G32B32A32_FLOAT
  62 };
  63
  64 static const unsigned num_idct_intermediate_formats =
  65    sizeof(const_idct_intermediate_formats) / sizeof(enum pipe_format);
  66
  67 static const enum pipe_format const_mc_source_formats[] = {
  68    PIPE_FORMAT_R16_SNORM,
  69    PIPE_FORMAT_R16_SSCALED
  70 };
  71
  72 static const unsigned num_mc_source_formats =
  73    sizeof(const_mc_source_formats) / sizeof(enum pipe_format);
  74
  75 static bool
  76 init_zscan_buffer(struct vl_mpeg12_buffer *buffer)
  77 {
  78    enum pipe_format formats[3];
  79
  80    struct pipe_sampler_view **source;
  81    struct pipe_surface **destination;
  82
  83    struct vl_mpeg12_decoder *dec;
  84
  85    unsigned i;
  86
  87    assert(buffer);
  88
  89    dec = (struct vl_mpeg12_decoder*)buffer->base.decoder;
  90
  91    formats[0] = formats[1] = formats[2] = dec->zscan_source_format;
  92    buffer->zscan_source = vl_video_buffer_init(dec->base.context, dec->pipe,
  93                                                dec->blocks_per_line * BLOCK_WIDTH * BLOCK_HEIGHT,
  94                                                dec->max_blocks / dec->blocks_per_line,
  95                                                1, PIPE_VIDEO_CHROMA_FORMAT_444,
  96                                                formats, PIPE_USAGE_STATIC);
  97    if (!buffer->zscan_source)
  98       goto error_source;
  99
 100    source = buffer->zscan_source->get_sampler_views(buffer->zscan_source);
 101    if (!source)
 102       goto error_sampler;
 103
 104    if (dec->base.entrypoint <= PIPE_VIDEO_ENTRYPOINT_IDCT)
 105       destination = buffer->idct_source->get_surfaces(buffer->idct_source);
 106    else
 107       destination = buffer->mc_source->get_surfaces(buffer->mc_source);
 108
 109    if (!destination)
 110       goto error_surface;
 111
 112    for (i = 0; i < VL_MAX_PLANES; ++i)
 113       if (!vl_zscan_init_buffer(i == 0 ? &dec->zscan_y : &dec->zscan_c,
 114                                 &buffer->zscan[i], source[i], destination[i]))
 115          goto error_plane;
 116
 117    return true;
 118
 119 error_plane:
 120    for (; i > 0; --i)
 121       vl_zscan_cleanup_buffer(&buffer->zscan[i - 1]);
 122
 123 error_surface:
 124 error_sampler:
 125    buffer->zscan_source->destroy(buffer->zscan_source);
 126
 127 error_source:
 128    return false;
 129 }
 130
 131 static void
 132 cleanup_zscan_buffer(struct vl_mpeg12_buffer *buffer)
 133 {
 134    unsigned i;
 135
 136    assert(buffer);
 137
 138    for (i = 0; i < VL_MAX_PLANES; ++i)
 139       vl_zscan_cleanup_buffer(&buffer->zscan[i]);
 140    buffer->zscan_source->destroy(buffer->zscan_source);
 141 }
 142
 143 static bool
 144 init_idct_buffer(struct vl_mpeg12_buffer *buffer)
 145 {
 146    enum pipe_format formats[3];
 147
 148    struct pipe_sampler_view **idct_source_sv, **idct_intermediate_sv;
 149    struct pipe_surface **idct_surfaces;
 150
 151    struct vl_mpeg12_decoder *dec;
 152
 153    unsigned i;
 154
 155    assert(buffer);
 156
 157    dec = (struct vl_mpeg12_decoder*)buffer->base.decoder;
 158
 159    formats[0] = formats[1] = formats[2] = dec->idct_source_format;
 160    buffer->idct_source = vl_video_buffer_init(dec->base.context, dec->pipe,
 161                                               dec->base.width / 4, dec->base.height, 1,
 162                                               dec->base.chroma_format,
 163                                               formats, PIPE_USAGE_STATIC);
 164    if (!buffer->idct_source)
 165       goto error_source;
 166
 167    formats[0] = formats[1] = formats[2] = dec->idct_intermediate_format;
 168    buffer->idct_intermediate = vl_video_buffer_init(dec->base.context, dec->pipe,
 169                                                     dec->base.width / dec->nr_of_idct_render_targets,
 170                                                     dec->base.height / 4, dec->nr_of_idct_render_targets,
 171                                                     dec->base.chroma_format,
 172                                                     formats, PIPE_USAGE_STATIC);
 173
 174    if (!buffer->idct_intermediate)
 175       goto error_intermediate;
 176
 177    idct_source_sv = buffer->idct_source->get_sampler_views(buffer->idct_source);
 178    if (!idct_source_sv)
 179       goto error_source_sv;
 180
 181    idct_intermediate_sv = buffer->idct_intermediate->get_sampler_views(buffer->idct_intermediate);
 182    if (!idct_intermediate_sv)
 183       goto error_intermediate_sv;
 184
 185    idct_surfaces = buffer->mc_source->get_surfaces(buffer->mc_source);
 186    if (!idct_surfaces)
 187       goto error_surfaces;
 188
 189    for (i = 0; i < 3; ++i)
 190       if (!vl_idct_init_buffer(i == 0 ? &dec->idct_y : &dec->idct_c,
 191                                &buffer->idct[i], idct_source_sv[i],
 192                                idct_intermediate_sv[i], idct_surfaces[i]))
 193          goto error_plane;
 194
 195    return true;
 196
 197 error_plane:
 198    for (; i > 0; --i)
 199       vl_idct_cleanup_buffer(i == 1 ? &dec->idct_c : &dec->idct_y, &buffer->idct[i - 1]);
 200
 201 error_surfaces:
 202 error_intermediate_sv:
 203 error_source_sv:
 204    buffer->idct_intermediate->destroy(buffer->idct_intermediate);
 205
 206 error_intermediate:
 207    buffer->idct_source->destroy(buffer->idct_source);
 208
 209 error_source:
 210    return false;
 211 }
 212
 213 static void
 214 cleanup_idct_buffer(struct vl_mpeg12_buffer *buf)
 215 {
 216    struct vl_mpeg12_decoder *dec;
 217    assert(buf);
 218
 219    dec = (struct vl_mpeg12_decoder*)buf->base.decoder;
 220    assert(dec);
 221
 222    vl_idct_cleanup_buffer(&dec->idct_y, &buf->idct[0]);
 223    vl_idct_cleanup_buffer(&dec->idct_c, &buf->idct[1]);
 224    vl_idct_cleanup_buffer(&dec->idct_c, &buf->idct[2]);
 225    buf->idct_source->destroy(buf->idct_source);
 226    buf->idct_intermediate->destroy(buf->idct_intermediate);
 227 }
 228
 229 static void
 230 vl_mpeg12_buffer_destroy(struct pipe_video_decode_buffer *buffer)
 231 {
 232    struct vl_mpeg12_buffer *buf = (struct vl_mpeg12_buffer*)buffer;
 233    struct vl_mpeg12_decoder *dec;
 234    unsigned i;
 235
 236    assert(buf);
 237
 238    dec = (struct vl_mpeg12_decoder*)buf->base.decoder;
 239    assert(dec);
 240
 241    cleanup_zscan_buffer(buf);
 242
 243    if (dec->base.entrypoint <= PIPE_VIDEO_ENTRYPOINT_IDCT)
 244       cleanup_idct_buffer(buf);
 245
 246    buf->mc_source->destroy(buf->mc_source);
 247    vl_vb_cleanup(&buf->vertex_stream);
 248    for (i = 0; i < VL_MAX_PLANES; ++i)
 249       vl_mc_cleanup_buffer(&buf->mc[i]);
 250
 251    FREE(buf);
 252 }
 253
 254 static void
 255 vl_mpeg12_buffer_map(struct pipe_video_decode_buffer *buffer)
 256 {
 257    struct vl_mpeg12_buffer *buf = (struct vl_mpeg12_buffer*)buffer;
 258    struct vl_mpeg12_decoder *dec;
 259
 260    struct pipe_sampler_view **sampler_views;
 261    unsigned i;
 262
 263    assert(buf);
 264
 265    dec = (struct vl_mpeg12_decoder *)buf->base.decoder;
 266    assert(dec);
 267
 268    vl_vb_map(&buf->vertex_stream, dec->pipe);
 269
 270    sampler_views = buf->zscan_source->get_sampler_views(buf->zscan_source);
 271
 272    assert(sampler_views);
 273
 274    for (i = 0; i < VL_MAX_PLANES; ++i) {
 275       struct pipe_resource *tex = sampler_views[i]->texture;
 276       struct pipe_box rect =
 277       {
 278          0, 0, 0,
 279          tex->width0,
 280          tex->height0,
 281          1
 282       };
 283
 284       buf->tex_transfer[i] = dec->pipe->get_transfer
 285       (
 286          dec->pipe, tex,
 287          0, PIPE_TRANSFER_WRITE | PIPE_TRANSFER_DISCARD,
 288          &rect
 289       );
 290
 291       buf->texels[i] = dec->pipe->transfer_map(dec->pipe, buf->tex_transfer[i]);
 292    }
 293 }
 294
 295 static struct pipe_ycbcr_block *
 296 vl_mpeg12_buffer_get_ycbcr_stream(struct pipe_video_decode_buffer *buffer, int component)
 297 {
 298    struct vl_mpeg12_buffer *buf = (struct vl_mpeg12_buffer*)buffer;
 299
 300    assert(buf);
 301
 302    return vl_vb_get_ycbcr_stream(&buf->vertex_stream, component);
 303 }
 304
 305 static short *
 306 vl_mpeg12_buffer_get_ycbcr_buffer(struct pipe_video_decode_buffer *buffer, int component)
 307 {
 308    struct vl_mpeg12_buffer *buf = (struct vl_mpeg12_buffer*)buffer;
 309
 310    assert(buf);
 311    assert(component < VL_MAX_PLANES);
 312
 313    return buf->texels[component];
 314 }
 315
 316 static unsigned
 317 vl_mpeg12_buffer_get_mv_stream_stride(struct pipe_video_decode_buffer *buffer)
 318 {
 319    struct vl_mpeg12_buffer *buf = (struct vl_mpeg12_buffer*)buffer;
 320
 321    assert(buf);
 322
 323    return vl_vb_get_mv_stream_stride(&buf->vertex_stream);
 324 }
 325
 326 static struct pipe_motionvector *
 327 vl_mpeg12_buffer_get_mv_stream(struct pipe_video_decode_buffer *buffer, int ref_frame)
 328 {
 329    struct vl_mpeg12_buffer *buf = (struct vl_mpeg12_buffer*)buffer;
 330
 331    assert(buf);
 332
 333    return vl_vb_get_mv_stream(&buf->vertex_stream, ref_frame);
 334 }
 335
 336 static void
 337 vl_mpeg12_buffer_unmap(struct pipe_video_decode_buffer *buffer)
 338 {
 339    struct vl_mpeg12_buffer *buf = (struct vl_mpeg12_buffer*)buffer;
 340    struct vl_mpeg12_decoder *dec;
 341    unsigned i;
 342
 343    assert(buf);
 344
 345    dec = (struct vl_mpeg12_decoder *)buf->base.decoder;
 346    assert(dec);
 347
 348    vl_vb_unmap(&buf->vertex_stream, dec->pipe);
 349
 350    for (i = 0; i < VL_MAX_PLANES; ++i) {
 351       dec->pipe->transfer_unmap(dec->pipe, buf->tex_transfer[i]);
 352       dec->pipe->transfer_destroy(dec->pipe, buf->tex_transfer[i]);
 353    }
 354 }
 355
 356 static void
 357 vl_mpeg12_destroy(struct pipe_video_decoder *decoder)
 358 {
 359    struct vl_mpeg12_decoder *dec = (struct vl_mpeg12_decoder*)decoder;
 360
 361    assert(decoder);
 362
 363    /* Asserted in softpipe_delete_fs_state() for some reason */
 364    dec->pipe->bind_vs_state(dec->pipe, NULL);
 365    dec->pipe->bind_fs_state(dec->pipe, NULL);
 366
 367    dec->pipe->delete_depth_stencil_alpha_state(dec->pipe, dec->dsa);
 368
 369    vl_mc_cleanup(&dec->mc_y);
 370    vl_mc_cleanup(&dec->mc_c);
 371
 372    if (dec->base.entrypoint <= PIPE_VIDEO_ENTRYPOINT_IDCT) {
 373       vl_idct_cleanup(&dec->idct_y);
 374       vl_idct_cleanup(&dec->idct_c);
 375    }
 376
 377    vl_zscan_cleanup(&dec->zscan_y);
 378    vl_zscan_cleanup(&dec->zscan_c);
 379
 380    dec->pipe->delete_vertex_elements_state(dec->pipe, dec->ves_ycbcr);
 381    dec->pipe->delete_vertex_elements_state(dec->pipe, dec->ves_mv);
 382
 383    pipe_resource_reference(&dec->quads.buffer, NULL);
 384    pipe_resource_reference(&dec->pos.buffer, NULL);
 385
 386    FREE(dec);
 387 }
 388
 389 static struct pipe_video_decode_buffer *
 390 vl_mpeg12_create_buffer(struct pipe_video_decoder *decoder)
 391 {
 392    enum pipe_format formats[3];
 393
 394    struct vl_mpeg12_decoder *dec = (struct vl_mpeg12_decoder*)decoder;
 395    struct vl_mpeg12_buffer *buffer;
 396
 397    struct pipe_sampler_view **mc_source_sv;
 398
 399    assert(dec);
 400
 401    buffer = CALLOC_STRUCT(vl_mpeg12_buffer);
 402    if (buffer == NULL)
 403       return NULL;
 404
 405    buffer->base.decoder = decoder;
 406    buffer->base.destroy = vl_mpeg12_buffer_destroy;
 407    buffer->base.map = vl_mpeg12_buffer_map;
 408    buffer->base.get_ycbcr_stream = vl_mpeg12_buffer_get_ycbcr_stream;
 409    buffer->base.get_ycbcr_buffer = vl_mpeg12_buffer_get_ycbcr_buffer;
 410    buffer->base.get_mv_stream_stride = vl_mpeg12_buffer_get_mv_stream_stride;
 411    buffer->base.get_mv_stream = vl_mpeg12_buffer_get_mv_stream;
 412    buffer->base.unmap = vl_mpeg12_buffer_unmap;
 413
 414    vl_vb_init(&buffer->vertex_stream, dec->pipe,
 415               dec->base.width / MACROBLOCK_WIDTH,
 416               dec->base.height / MACROBLOCK_HEIGHT);
 417
 418    formats[0] = formats[1] = formats[2] =dec->mc_source_format;
 419    buffer->mc_source = vl_video_buffer_init(dec->base.context, dec->pipe,
 420                                             dec->base.width, dec->base.height, 1,
 421                                             dec->base.chroma_format,
 422                                             formats, PIPE_USAGE_STATIC);
 423
 424    if (!buffer->mc_source)
 425       goto error_mc_source;
 426
 427    mc_source_sv = buffer->mc_source->get_sampler_views(buffer->mc_source);
 428    if (!mc_source_sv)
 429       goto error_mc_source_sv;
 430
 431    if(!vl_mc_init_buffer(&dec->mc_y, &buffer->mc[0], mc_source_sv[0]))
 432       goto error_mc_y;
 433
 434    if(!vl_mc_init_buffer(&dec->mc_c, &buffer->mc[1], mc_source_sv[1]))
 435       goto error_mc_cb;
 436
 437    if(!vl_mc_init_buffer(&dec->mc_c, &buffer->mc[2], mc_source_sv[2]))
 438       goto error_mc_cr;
 439
 440    if (dec->base.entrypoint <= PIPE_VIDEO_ENTRYPOINT_IDCT)
 441       if (!init_idct_buffer(buffer))
 442          goto error_idct;
 443
 444    if (!init_zscan_buffer(buffer))
 445       goto error_zscan;
 446
 447    return &buffer->base;
 448
 449 error_zscan:
 450    // TODO Cleanup error handling
 451
 452 error_mc_cr:
 453    vl_mc_cleanup_buffer(&buffer->mc[1]);
 454
 455 error_mc_cb:
 456    vl_mc_cleanup_buffer(&buffer->mc[0]);
 457
 458 error_mc_y:
 459 error_mc_source_sv:
 460    if (dec->base.entrypoint <= PIPE_VIDEO_ENTRYPOINT_IDCT)
 461       cleanup_idct_buffer(buffer);
 462
 463 error_idct:
 464    buffer->mc_source->destroy(buffer->mc_source);
 465
 466 error_mc_source:
 467    vl_vb_cleanup(&buffer->vertex_stream);
 468
 469 error_vertex_stream:
 470    FREE(buffer);
 471    return NULL;
 472 }
 473
 474 static void
 475 vl_mpeg12_decoder_flush_buffer(struct pipe_video_decode_buffer *buffer,
 476                                unsigned num_ycbcr_blocks[3],
 477                                struct pipe_video_buffer *refs[2],
 478                                struct pipe_video_buffer *dst,
 479                                struct pipe_fence_handle **fence)
 480 {
 481    struct vl_mpeg12_buffer *buf = (struct vl_mpeg12_buffer *)buffer;
 482    struct vl_mpeg12_decoder *dec;
 483
 484    struct pipe_sampler_view **sv[2];
 485    struct pipe_surface **surfaces;
 486
 487    struct pipe_vertex_buffer vb[3];
 488
 489    unsigned i, j;
 490
 491    assert(buf);
 492
 493    dec = (struct vl_mpeg12_decoder *)buf->base.decoder;
 494    assert(dec);
 495
 496    for (i = 0; i < 2; ++i)
 497       sv[i] = refs[i] ? refs[i]->get_sampler_views(refs[i]) : NULL;
 498
 499    surfaces = dst->get_surfaces(dst);
 500
 501    vb[0] = dec->quads;
 502    vb[1] = dec->pos;
 503
 504    dec->pipe->bind_vertex_elements_state(dec->pipe, dec->ves_mv);
 505    for (i = 0; i < VL_MAX_PLANES; ++i) {
 506       vl_mc_set_surface(&buf->mc[i], surfaces[i]);
 507
 508       for (j = 0; j < 2; ++j) {
 509          if (sv[j] == NULL) continue;
 510
 511          vb[2] = vl_vb_get_mv(&buf->vertex_stream, j);;
 512          dec->pipe->set_vertex_buffers(dec->pipe, 3, vb);
 513
 514          vl_mc_render_ref(&buf->mc[i], sv[j][i]);
 515       }
 516    }
 517
 518    dec->pipe->bind_vertex_elements_state(dec->pipe, dec->ves_ycbcr);
 519    for (i = 0; i < VL_MAX_PLANES; ++i) {
 520       if (num_ycbcr_blocks[i] == 0) continue;
 521
 522       vb[1] = vl_vb_get_ycbcr(&buf->vertex_stream, i);
 523       dec->pipe->set_vertex_buffers(dec->pipe, 2, vb);
 524
 525       vl_zscan_render(&buf->zscan[i] , num_ycbcr_blocks[i]);
 526
 527       if (dec->base.entrypoint <= PIPE_VIDEO_ENTRYPOINT_IDCT)
 528          vl_idct_flush(i == 0 ? &dec->idct_y : &dec->idct_c, &buf->idct[i], num_ycbcr_blocks[i]);
 529
 530       vl_mc_render_ycbcr(&buf->mc[i], num_ycbcr_blocks[i]);
 531    }
 532
 533    dec->pipe->flush(dec->pipe, fence);
 534 }
 535
 536 static bool
 537 init_pipe_state(struct vl_mpeg12_decoder *dec)
 538 {
 539    struct pipe_depth_stencil_alpha_state dsa;
 540    unsigned i;
 541
 542    assert(dec);
 543
 544    memset(&dsa, 0, sizeof dsa);
 545    dsa.depth.enabled = 0;
 546    dsa.depth.writemask = 0;
 547    dsa.depth.func = PIPE_FUNC_ALWAYS;
 548    for (i = 0; i < 2; ++i) {
 549       dsa.stencil[i].enabled = 0;
 550       dsa.stencil[i].func = PIPE_FUNC_ALWAYS;
 551       dsa.stencil[i].fail_op = PIPE_STENCIL_OP_KEEP;
 552       dsa.stencil[i].zpass_op = PIPE_STENCIL_OP_KEEP;
 553       dsa.stencil[i].zfail_op = PIPE_STENCIL_OP_KEEP;
 554       dsa.stencil[i].valuemask = 0;
 555       dsa.stencil[i].writemask = 0;
 556    }
 557    dsa.alpha.enabled = 0;
 558    dsa.alpha.func = PIPE_FUNC_ALWAYS;
 559    dsa.alpha.ref_value = 0;
 560    dec->dsa = dec->pipe->create_depth_stencil_alpha_state(dec->pipe, &dsa);
 561    dec->pipe->bind_depth_stencil_alpha_state(dec->pipe, dec->dsa);
 562
 563    return true;
 564 }
 565
 566 static enum pipe_format
 567 find_first_supported_format(struct vl_mpeg12_decoder *dec,
 568                             const enum pipe_format formats[],
 569                             unsigned num_formats,
 570                             enum pipe_texture_target target)
 571 {
 572    struct pipe_screen *screen;
 573    unsigned i;
 574
 575    assert(dec);
 576
 577    screen = dec->pipe->screen;
 578
 579    for (i = 0; i < num_formats; ++i)
 580       if (screen->is_format_supported(dec->pipe->screen, formats[i], target, 1,
 581                                       PIPE_BIND_SAMPLER_VIEW | PIPE_BIND_RENDER_TARGET))
 582          return formats[i];
 583
 584    return PIPE_FORMAT_NONE;
 585 }
 586
 587 static bool
 588 init_zscan(struct vl_mpeg12_decoder *dec)
 589 {
 590    struct pipe_sampler_view *layout;
 591
 592    unsigned num_channels;
 593
 594    assert(dec);
 595
 596    dec->blocks_per_line = 4;
 597    dec->max_blocks =
 598       (dec->base.width * dec->base.height) /
 599       (BLOCK_WIDTH * BLOCK_HEIGHT);
 600
 601    dec->zscan_source_format = find_first_supported_format(dec, const_zscan_source_formats,
 602                                                           num_zscan_source_formats, PIPE_TEXTURE_2D);
 603
 604    if (dec->zscan_source_format == PIPE_FORMAT_NONE)
 605       return false;
 606
 607    layout = vl_zscan_linear(dec->pipe, dec->blocks_per_line);
 608
 609    num_channels = dec->base.entrypoint <= PIPE_VIDEO_ENTRYPOINT_IDCT ? 4 : 1;
 610
 611    if (!vl_zscan_init(&dec->zscan_y, dec->pipe, dec->base.width, dec->base.height,
 612                       dec->blocks_per_line, dec->max_blocks, num_channels))
 613       return false;
 614
 615    vl_zscan_set_layout(&dec->zscan_y, layout);
 616
 617    if (!vl_zscan_init(&dec->zscan_c, dec->pipe, dec->chroma_width, dec->chroma_height,
 618                       dec->blocks_per_line, dec->max_blocks, num_channels))
 619       return false;
 620
 621    vl_zscan_set_layout(&dec->zscan_c, layout);
 622
 623    return true;
 624 }
 625
 626 static bool
 627 init_idct(struct vl_mpeg12_decoder *dec)
 628 {
 629    struct pipe_sampler_view *matrix, *transpose;
 630    float matrix_scale, transpose_scale;
 631
 632    dec->nr_of_idct_render_targets = dec->pipe->screen->get_param(dec->pipe->screen, PIPE_CAP_MAX_RENDER_TARGETS);
 633
 634    // more than 4 render targets usually doesn't makes any seens
 635    dec->nr_of_idct_render_targets = MIN2(dec->nr_of_idct_render_targets, 4);
 636
 637    dec->idct_source_format = find_first_supported_format(dec, const_idct_source_formats,
 638                                                          num_idct_source_formats, PIPE_TEXTURE_2D);
 639
 640    if (dec->idct_source_format == PIPE_FORMAT_NONE)
 641       return false;
 642
 643    dec->idct_intermediate_format = find_first_supported_format(dec, const_idct_intermediate_formats,
 644                                                                num_idct_intermediate_formats, PIPE_TEXTURE_3D);
 645
 646    if (dec->idct_intermediate_format == PIPE_FORMAT_NONE)
 647       return false;
 648
 649    switch (dec->idct_source_format) {
 650    case PIPE_FORMAT_R16G16B16A16_SSCALED:
 651       matrix_scale = SCALE_FACTOR_SSCALED;
 652       break;
 653
 654    case PIPE_FORMAT_R16G16B16A16_SNORM:
 655       matrix_scale = SCALE_FACTOR_SNORM;
 656       break;
 657
 658    default:
 659       assert(0);
 660       return false;
 661    }
 662
 663    if (dec->idct_intermediate_format == PIPE_FORMAT_R16G16B16A16_FLOAT ||
 664        dec->idct_intermediate_format == PIPE_FORMAT_R32G32B32A32_FLOAT)
 665       transpose_scale = 1.0f;
 666    else
 667       transpose_scale = matrix_scale = sqrt(matrix_scale);
 668
 669    if (dec->mc_source_format == PIPE_FORMAT_R16_SSCALED)
 670       transpose_scale /= SCALE_FACTOR_SSCALED;
 671
 672    if (!(matrix = vl_idct_upload_matrix(dec->pipe, matrix_scale)))
 673       goto error_matrix;
 674
 675    if (matrix_scale != transpose_scale) {
 676       if (!(transpose = vl_idct_upload_matrix(dec->pipe, transpose_scale)))
 677          goto error_transpose;
 678    } else
 679       pipe_sampler_view_reference(&transpose, matrix);
 680
 681    if (!vl_idct_init(&dec->idct_y, dec->pipe, dec->base.width, dec->base.height,
 682                      dec->nr_of_idct_render_targets, matrix, transpose))
 683       goto error_y;
 684
 685    if(!vl_idct_init(&dec->idct_c, dec->pipe, dec->chroma_width, dec->chroma_height,
 686                     dec->nr_of_idct_render_targets, matrix, transpose))
 687       goto error_c;
 688
 689    pipe_sampler_view_reference(&matrix, NULL);
 690    pipe_sampler_view_reference(&transpose, NULL);
 691    return true;
 692
 693 error_c:
 694    vl_idct_cleanup(&dec->idct_y);
 695
 696 error_y:
 697    pipe_sampler_view_reference(&transpose, NULL);
 698
 699 error_transpose:
 700    pipe_sampler_view_reference(&matrix, NULL);
 701
 702 error_matrix:
 703    return false;
 704 }
 705
 706 struct pipe_video_decoder *
 707 vl_create_mpeg12_decoder(struct pipe_video_context *context,
 708                          struct pipe_context *pipe,
 709                          enum pipe_video_profile profile,
 710                          enum pipe_video_entrypoint entrypoint,
 711                          enum pipe_video_chroma_format chroma_format,
 712                          unsigned width, unsigned height)
 713 {
 714    struct vl_mpeg12_decoder *dec;
 715    float mc_scale;
 716
 717    assert(u_reduce_video_profile(profile) == PIPE_VIDEO_CODEC_MPEG12);
 718
 719    dec = CALLOC_STRUCT(vl_mpeg12_decoder);
 720
 721    if (!dec)
 722       return NULL;
 723
 724    dec->base.context = context;
 725    dec->base.profile = profile;
 726    dec->base.entrypoint = entrypoint;
 727    dec->base.chroma_format = chroma_format;
 728    dec->base.width = width;
 729    dec->base.height = height;
 730
 731    dec->base.destroy = vl_mpeg12_destroy;
 732    dec->base.create_buffer = vl_mpeg12_create_buffer;
 733    dec->base.flush_buffer = vl_mpeg12_decoder_flush_buffer;
 734
 735    dec->base.width = align(width, MACROBLOCK_WIDTH);
 736    dec->base.height = align(height, MACROBLOCK_HEIGHT);
 737
 738    dec->pipe = pipe;
 739
 740    dec->quads = vl_vb_upload_quads(dec->pipe);
 741    dec->pos = vl_vb_upload_pos(
 742       dec->pipe,
 743       dec->base.width / MACROBLOCK_WIDTH,
 744       dec->base.height / MACROBLOCK_HEIGHT
 745    );
 746
 747    dec->ves_ycbcr = vl_vb_get_ves_ycbcr(dec->pipe);
 748    dec->ves_mv = vl_vb_get_ves_mv(dec->pipe);
 749
 750    /* TODO: Implement 422, 444 */
 751    assert(dec->base.chroma_format == PIPE_VIDEO_CHROMA_FORMAT_420);
 752
 753    dec->mc_source_format = find_first_supported_format(dec, const_mc_source_formats,
 754                                                        num_mc_source_formats, PIPE_TEXTURE_3D);
 755
 756    if (dec->mc_source_format == PIPE_FORMAT_NONE)
 757       return NULL;
 758
 759    if (dec->base.chroma_format == PIPE_VIDEO_CHROMA_FORMAT_420) {
 760       dec->chroma_width = dec->base.width / 2;
 761       dec->chroma_height = dec->base.height / 2;
 762    } else if (dec->base.chroma_format == PIPE_VIDEO_CHROMA_FORMAT_422) {
 763       dec->chroma_width = dec->base.width;
 764       dec->chroma_height = dec->base.height / 2;
 765    } else {
 766       dec->chroma_width = dec->base.width;
 767       dec->chroma_height = dec->base.height;
 768    }
 769
 770    if (!init_zscan(dec))
 771       return NULL; // TODO error handling
 772
 773    if (entrypoint <= PIPE_VIDEO_ENTRYPOINT_IDCT) {
 774       if (!init_idct(dec))
 775          goto error_idct;
 776       if (dec->mc_source_format == PIPE_FORMAT_R16_SSCALED)
 777          mc_scale = SCALE_FACTOR_SSCALED;
 778       else
 779          mc_scale = 1.0f;
 780    } else {
 781       switch (dec->mc_source_format) {
 782       case PIPE_FORMAT_R16_SNORM:
 783          mc_scale = SCALE_FACTOR_SNORM;
 784          break;
 785
 786       case PIPE_FORMAT_R16_SSCALED:
 787          mc_scale = SCALE_FACTOR_SSCALED;
 788          break;
 789
 790       default:
 791          assert(0);
 792          return NULL;
 793       }
 794    }
 795
 796    if (!vl_mc_init(&dec->mc_y, dec->pipe, dec->base.width, dec->base.height, MACROBLOCK_HEIGHT, mc_scale))
 797       goto error_mc_y;
 798
 799    // TODO
 800    if (!vl_mc_init(&dec->mc_c, dec->pipe, dec->base.width, dec->base.height, BLOCK_HEIGHT, mc_scale))
 801       goto error_mc_c;
 802
 803    if (!init_pipe_state(dec))
 804       goto error_pipe_state;
 805
 806    return &dec->base;
 807
 808 error_pipe_state:
 809    vl_mc_cleanup(&dec->mc_c);
 810
 811 error_mc_c:
 812    vl_mc_cleanup(&dec->mc_y);
 813
 814 error_mc_y:
 815    if (entrypoint <= PIPE_VIDEO_ENTRYPOINT_IDCT) {
 816       vl_idct_cleanup(&dec->idct_y);
 817       vl_idct_cleanup(&dec->idct_c);
 818    }
 819
 820 error_idct:
 821    FREE(dec);
 822    return NULL;
 823 }