Subversion Repositories Kolibri OS

Rev

Blame | Last modification | View Log | RSS feed

  1. /*
  2.  * Copyright 2011-2013 Maarten Lankhorst, Ilia Mirkin
  3.  *
  4.  * Permission is hereby granted, free of charge, to any person obtaining a
  5.  * copy of this software and associated documentation files (the "Software"),
  6.  * to deal in the Software without restriction, including without limitation
  7.  * the rights to use, copy, modify, merge, publish, distribute, sublicense,
  8.  * and/or sell copies of the Software, and to permit persons to whom the
  9.  * Software is furnished to do so, subject to the following conditions:
  10.  *
  11.  * The above copyright notice and this permission notice shall be included in
  12.  * all copies or substantial portions of the Software.
  13.  *
  14.  * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
  15.  * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
  16.  * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
  17.  * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR
  18.  * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
  19.  * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
  20.  * OTHER DEALINGS IN THE SOFTWARE.
  21.  */
  22.  
  23. #include "nv50/nv98_video.h"
  24.  
  25. #include "util/u_sampler.h"
  26. #include "util/u_format.h"
  27.  
  28. static void
  29. nv98_decoder_decode_bitstream(struct pipe_video_codec *decoder,
  30.                               struct pipe_video_buffer *video_target,
  31.                               struct pipe_picture_desc *picture,
  32.                               unsigned num_buffers,
  33.                               const void *const *data,
  34.                               const unsigned *num_bytes)
  35. {
  36.    struct nouveau_vp3_decoder *dec = (struct nouveau_vp3_decoder *)decoder;
  37.    struct nouveau_vp3_video_buffer *target = (struct nouveau_vp3_video_buffer *)video_target;
  38.    uint32_t comm_seq = ++dec->fence_seq;
  39.    union pipe_desc desc;
  40.  
  41.    unsigned vp_caps, is_ref, ret;
  42.    struct nouveau_vp3_video_buffer *refs[16] = {};
  43.  
  44.    desc.base = picture;
  45.  
  46.    assert(target->base.buffer_format == PIPE_FORMAT_NV12);
  47.  
  48.    ret = nv98_decoder_bsp(dec, desc, target, comm_seq,
  49.                           num_buffers, data, num_bytes,
  50.                           &vp_caps, &is_ref, refs);
  51.  
  52.    /* did we decode bitstream correctly? */
  53.    assert(ret == 2);
  54.  
  55.    nv98_decoder_vp(dec, desc, target, comm_seq, vp_caps, is_ref, refs);
  56.    nv98_decoder_ppp(dec, desc, target, comm_seq);
  57. }
  58.  
  59. struct pipe_video_codec *
  60. nv98_create_decoder(struct pipe_context *context,
  61.                     const struct pipe_video_codec *templ)
  62. {
  63.    struct nouveau_screen *screen = &((struct nv50_context *)context)->screen->base;
  64.    struct nouveau_vp3_decoder *dec;
  65.    struct nouveau_pushbuf **push;
  66.    struct nv04_fifo nv04_data = {.vram = 0xbeef0201, .gart = 0xbeef0202};
  67.  
  68.    int ret, i;
  69.    uint32_t codec = 1, ppp_codec = 3;
  70.    uint32_t timeout;
  71.    u32 tmp_size = 0;
  72.  
  73.    if (getenv("XVMC_VL"))
  74.        return vl_create_decoder(context, templ);
  75.  
  76.    if (templ->entrypoint != PIPE_VIDEO_ENTRYPOINT_BITSTREAM) {
  77.       debug_printf("%x\n", templ->entrypoint);
  78.       return NULL;
  79.    }
  80.  
  81.    dec = CALLOC_STRUCT(nouveau_vp3_decoder);
  82.    if (!dec)
  83.       return NULL;
  84.    dec->client = screen->client;
  85.    dec->base = *templ;
  86.    nouveau_vp3_decoder_init_common(&dec->base);
  87.  
  88.    dec->bsp_idx = 5;
  89.    dec->vp_idx = 6;
  90.    dec->ppp_idx = 7;
  91.  
  92.    ret = nouveau_object_new(&screen->device->object, 0,
  93.                             NOUVEAU_FIFO_CHANNEL_CLASS,
  94.                             &nv04_data, sizeof(nv04_data), &dec->channel[0]);
  95.  
  96.    if (!ret)
  97.       ret = nouveau_pushbuf_new(screen->client, dec->channel[0], 4,
  98.                                 32 * 1024, true, &dec->pushbuf[0]);
  99.  
  100.    for (i = 1; i < 3; ++i) {
  101.       dec->channel[i] = dec->channel[0];
  102.       dec->pushbuf[i] = dec->pushbuf[0];
  103.    }
  104.    push = dec->pushbuf;
  105.  
  106.    if (!ret)
  107.       ret = nouveau_object_new(dec->channel[0], 0x390b1, 0x85b1, NULL, 0, &dec->bsp);
  108.    if (!ret)
  109.       ret = nouveau_object_new(dec->channel[1], 0x190b2, 0x85b2, NULL, 0, &dec->vp);
  110.    if (!ret)
  111.       ret = nouveau_object_new(dec->channel[2], 0x290b3, 0x85b3, NULL, 0, &dec->ppp);
  112.    if (ret)
  113.       goto fail;
  114.  
  115.    BEGIN_NV04(push[0], SUBC_BSP(NV01_SUBCHAN_OBJECT), 1);
  116.    PUSH_DATA (push[0], dec->bsp->handle);
  117.  
  118.    BEGIN_NV04(push[0], SUBC_BSP(0x180), 5);
  119.    for (i = 0; i < 5; i++)
  120.       PUSH_DATA (push[0], nv04_data.vram);
  121.  
  122.    BEGIN_NV04(push[1], SUBC_VP(NV01_SUBCHAN_OBJECT), 1);
  123.    PUSH_DATA (push[1], dec->vp->handle);
  124.  
  125.    BEGIN_NV04(push[1], SUBC_VP(0x180), 6);
  126.    for (i = 0; i < 6; i++)
  127.       PUSH_DATA (push[1], nv04_data.vram);
  128.  
  129.    BEGIN_NV04(push[2], SUBC_PPP(NV01_SUBCHAN_OBJECT), 1);
  130.    PUSH_DATA (push[2], dec->ppp->handle);
  131.  
  132.    BEGIN_NV04(push[2], SUBC_PPP(0x180), 5);
  133.    for (i = 0; i < 5; i++)
  134.       PUSH_DATA (push[2], nv04_data.vram);
  135.  
  136.    dec->base.context = context;
  137.    dec->base.decode_bitstream = nv98_decoder_decode_bitstream;
  138.  
  139.    for (i = 0; i < NOUVEAU_VP3_VIDEO_QDEPTH && !ret; ++i)
  140.       ret = nouveau_bo_new(screen->device, NOUVEAU_BO_VRAM,
  141.                            0, 1 << 20, NULL, &dec->bsp_bo[i]);
  142.    if (!ret)
  143.       ret = nouveau_bo_new(screen->device, NOUVEAU_BO_VRAM,
  144.                            0x100, 4 << 20, NULL, &dec->inter_bo[0]);
  145.    if (!ret)
  146.       nouveau_bo_ref(dec->inter_bo[0], &dec->inter_bo[1]);
  147.    if (ret)
  148.       goto fail;
  149.  
  150.    switch (u_reduce_video_profile(templ->profile)) {
  151.    case PIPE_VIDEO_FORMAT_MPEG12: {
  152.       codec = 1;
  153.       assert(templ->max_references <= 2);
  154.       break;
  155.    }
  156.    case PIPE_VIDEO_FORMAT_MPEG4: {
  157.       codec = 4;
  158.       tmp_size = mb(templ->height)*16 * mb(templ->width)*16;
  159.       assert(templ->max_references <= 2);
  160.       break;
  161.    }
  162.    case PIPE_VIDEO_FORMAT_VC1: {
  163.       ppp_codec = codec = 2;
  164.       tmp_size = mb(templ->height)*16 * mb(templ->width)*16;
  165.       assert(templ->max_references <= 2);
  166.       break;
  167.    }
  168.    case PIPE_VIDEO_FORMAT_MPEG4_AVC: {
  169.       codec = 3;
  170.       dec->tmp_stride = 16 * mb_half(templ->width) * nouveau_vp3_video_align(templ->height) * 3 / 2;
  171.       tmp_size = dec->tmp_stride * (templ->max_references + 1);
  172.       assert(templ->max_references <= 16);
  173.       break;
  174.    }
  175.    default:
  176.       fprintf(stderr, "invalid codec\n");
  177.       goto fail;
  178.    }
  179.  
  180.    ret = nouveau_bo_new(screen->device, NOUVEAU_BO_VRAM, 0,
  181.                            0x4000, NULL, &dec->fw_bo);
  182.    if (ret)
  183.       goto fail;
  184.  
  185.    ret = nouveau_vp3_load_firmware(dec, templ->profile, screen->device->chipset);
  186.    if (ret)
  187.       goto fw_fail;
  188.  
  189.    if (codec != 3) {
  190.       ret = nouveau_bo_new(screen->device, NOUVEAU_BO_VRAM, 0,
  191.                            0x400, NULL, &dec->bitplane_bo);
  192.       if (ret)
  193.          goto fail;
  194.    }
  195.  
  196.    dec->ref_stride = mb(templ->width)*16 * (mb_half(templ->height)*32 + nouveau_vp3_video_align(templ->height)/2);
  197.    ret = nouveau_bo_new(screen->device, NOUVEAU_BO_VRAM, 0,
  198.                         dec->ref_stride * (templ->max_references+2) + tmp_size,
  199.                         NULL, &dec->ref_bo);
  200.    if (ret)
  201.       goto fail;
  202.  
  203.    timeout = 0;
  204.  
  205.    BEGIN_NV04(push[0], SUBC_BSP(0x200), 2);
  206.    PUSH_DATA (push[0], codec);
  207.    PUSH_DATA (push[0], timeout);
  208.  
  209.    BEGIN_NV04(push[1], SUBC_VP(0x200), 2);
  210.    PUSH_DATA (push[1], codec);
  211.    PUSH_DATA (push[1], timeout);
  212.  
  213.    BEGIN_NV04(push[2], SUBC_PPP(0x200), 2);
  214.    PUSH_DATA (push[2], ppp_codec);
  215.    PUSH_DATA (push[2], timeout);
  216.  
  217.    ++dec->fence_seq;
  218.  
  219. #if NOUVEAU_VP3_DEBUG_FENCE
  220.    ret = nouveau_bo_new(screen->device, NOUVEAU_BO_GART|NOUVEAU_BO_MAP,
  221.                         0, 0x1000, NULL, &dec->fence_bo);
  222.    if (ret)
  223.       goto fail;
  224.  
  225.    nouveau_bo_map(dec->fence_bo, NOUVEAU_BO_RDWR, screen->client);
  226.    dec->fence_map = dec->fence_bo->map;
  227.    dec->fence_map[0] = dec->fence_map[4] = dec->fence_map[8] = 0;
  228.    dec->comm = (struct comm *)(dec->fence_map + (COMM_OFFSET/sizeof(*dec->fence_map)));
  229.  
  230.    /* So lets test if the fence is working? */
  231.    nouveau_pushbuf_space(push[0], 6, 1, 0);
  232.    PUSH_REFN (push[0], dec->fence_bo, NOUVEAU_BO_GART|NOUVEAU_BO_RDWR);
  233.    BEGIN_NV04(push[0], SUBC_BSP(0x240), 3);
  234.    PUSH_DATAh(push[0], dec->fence_bo->offset);
  235.    PUSH_DATA (push[0], dec->fence_bo->offset);
  236.    PUSH_DATA (push[0], dec->fence_seq);
  237.  
  238.    BEGIN_NV04(push[0], SUBC_BSP(0x304), 1);
  239.    PUSH_DATA (push[0], 0);
  240.    PUSH_KICK (push[0]);
  241.  
  242.    nouveau_pushbuf_space(push[1], 6, 1, 0);
  243.    PUSH_REFN (push[1], dec->fence_bo, NOUVEAU_BO_GART|NOUVEAU_BO_RDWR);
  244.    BEGIN_NV04(push[1], SUBC_VP(0x240), 3);
  245.    PUSH_DATAh(push[1], (dec->fence_bo->offset + 0x10));
  246.    PUSH_DATA (push[1], (dec->fence_bo->offset + 0x10));
  247.    PUSH_DATA (push[1], dec->fence_seq);
  248.  
  249.    BEGIN_NV04(push[1], SUBC_VP(0x304), 1);
  250.    PUSH_DATA (push[1], 0);
  251.    PUSH_KICK (push[1]);
  252.  
  253.    nouveau_pushbuf_space(push[2], 6, 1, 0);
  254.    PUSH_REFN (push[2], dec->fence_bo, NOUVEAU_BO_GART|NOUVEAU_BO_RDWR);
  255.    BEGIN_NV04(push[2], SUBC_PPP(0x240), 3);
  256.    PUSH_DATAh(push[2], (dec->fence_bo->offset + 0x20));
  257.    PUSH_DATA (push[2], (dec->fence_bo->offset + 0x20));
  258.    PUSH_DATA (push[2], dec->fence_seq);
  259.  
  260.    BEGIN_NV04(push[2], SUBC_PPP(0x304), 1);
  261.    PUSH_DATA (push[2], 0);
  262.    PUSH_KICK (push[2]);
  263.  
  264.    usleep(100);
  265.    while (dec->fence_seq > dec->fence_map[0] ||
  266.           dec->fence_seq > dec->fence_map[4] ||
  267.           dec->fence_seq > dec->fence_map[8]) {
  268.       debug_printf("%u: %u %u %u\n", dec->fence_seq, dec->fence_map[0], dec->fence_map[4], dec->fence_map[8]);
  269.       usleep(100);
  270.    }
  271.    debug_printf("%u: %u %u %u\n", dec->fence_seq, dec->fence_map[0], dec->fence_map[4], dec->fence_map[8]);
  272. #endif
  273.  
  274.    return &dec->base;
  275.  
  276. fw_fail:
  277.    debug_printf("Cannot create decoder without firmware..\n");
  278.    dec->base.destroy(&dec->base);
  279.    return NULL;
  280.  
  281. fail:
  282.    debug_printf("Creation failed: %s (%i)\n", strerror(-ret), ret);
  283.    dec->base.destroy(&dec->base);
  284.    return NULL;
  285. }
  286.  
  287. struct pipe_video_buffer *
  288. nv98_video_buffer_create(struct pipe_context *pipe,
  289.                          const struct pipe_video_buffer *templat)
  290. {
  291.    return nouveau_vp3_video_buffer_create(
  292.          pipe, templat, NV50_RESOURCE_FLAG_VIDEO);
  293. }
  294.