src/gallium/drivers/nouveau/nv50/nv84_video_bsp.c

   1 /*
   2  * Copyright 2013 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/nv84_video.h"
  24
  25 struct iparm {
  26    struct iseqparm {
  27       uint32_t chroma_format_idc; // 00
  28       uint32_t pad[(0x128 - 0x4) / 4];
  29       uint32_t log2_max_frame_num_minus4; // 128
  30       uint32_t pic_order_cnt_type; // 12c
  31       uint32_t log2_max_pic_order_cnt_lsb_minus4; // 130
  32       uint32_t delta_pic_order_always_zero_flag; // 134
  33       uint32_t num_ref_frames; // 138
  34       uint32_t pic_width_in_mbs_minus1; // 13c
  35       uint32_t pic_height_in_map_units_minus1; // 140
  36       uint32_t frame_mbs_only_flag; // 144
  37       uint32_t mb_adaptive_frame_field_flag; // 148
  38       uint32_t direct_8x8_inference_flag; // 14c
  39    } iseqparm; // 000
  40    struct ipicparm {
  41       uint32_t entropy_coding_mode_flag; // 00
  42       uint32_t pic_order_present_flag; // 04
  43       uint32_t num_slice_groups_minus1; // 08
  44       uint32_t slice_group_map_type; // 0c
  45       uint32_t pad1[0x60 / 4];
  46       uint32_t u70; // 70
  47       uint32_t u74; // 74
  48       uint32_t u78; // 78
  49       uint32_t num_ref_idx_l0_active_minus1; // 7c
  50       uint32_t num_ref_idx_l1_active_minus1; // 80
  51       uint32_t weighted_pred_flag; // 84
  52       uint32_t weighted_bipred_idc; // 88
  53       uint32_t pic_init_qp_minus26; // 8c
  54       uint32_t chroma_qp_index_offset; // 90
  55       uint32_t deblocking_filter_control_present_flag; // 94
  56       uint32_t constrained_intra_pred_flag; // 98
  57       uint32_t redundant_pic_cnt_present_flag; // 9c
  58       uint32_t transform_8x8_mode_flag; // a0
  59       uint32_t pad2[(0x1c8 - 0xa0 - 4) / 4];
  60       uint32_t second_chroma_qp_index_offset; // 1c8
  61       uint32_t u1cc; // 1cc
  62       uint32_t curr_pic_order_cnt; // 1d0
  63       uint32_t field_order_cnt[2]; // 1d4
  64       uint32_t curr_mvidx; // 1dc
  65       struct iref {
  66          uint32_t u00; // 00
  67          uint32_t field_is_ref; // 04 // bit0: top, bit1: bottom
  68          uint8_t is_long_term; // 08
  69          uint8_t non_existing; // 09
  70          uint32_t frame_idx; // 0c
  71          uint32_t field_order_cnt[2]; // 10
  72          uint32_t mvidx; // 18
  73          uint8_t field_pic_flag; // 1c
  74          // 20
  75       } refs[0x10]; // 1e0
  76    } ipicparm; // 150
  77 };
  78
  79 int
  80 nv84_decoder_bsp(struct nv84_decoder *dec,
  81                  struct pipe_h264_picture_desc *desc,
  82                  unsigned num_buffers,
  83                  const void *const *data,
  84                  const unsigned *num_bytes,
  85                  struct nv84_video_buffer *dest)
  86 {
  87    struct iparm params;
  88    uint32_t more_params[0x44 / 4] = {0};
  89    unsigned total_bytes = 0;
  90    int i;
  91    static const uint32_t end[] = {0x0b010000, 0, 0x0b010000, 0};
  92    char indexes[17] = {0};
  93    struct nouveau_pushbuf *push = dec->bsp_pushbuf;
  94    struct nouveau_pushbuf_refn bo_refs[] = {
  95       { dec->vpring, NOUVEAU_BO_RDWR | NOUVEAU_BO_VRAM },
  96       { dec->mbring, NOUVEAU_BO_RDWR | NOUVEAU_BO_VRAM },
  97       { dec->bitstream, NOUVEAU_BO_RDWR | NOUVEAU_BO_GART },
  98       { dec->fence, NOUVEAU_BO_RDWR | NOUVEAU_BO_VRAM },
  99    };
 100
 101    nouveau_bo_wait(dec->fence, NOUVEAU_BO_RDWR, dec->client);
 102
 103    STATIC_ASSERT(sizeof(struct iparm) == 0x530);
 104
 105    memset(&params, 0, sizeof(params));
 106
 107    dest->frame_num = dest->frame_num_max = desc->frame_num;
 108
 109    for (i = 0; i < 16; i++) {
 110       struct iref *ref = &params.ipicparm.refs[i];
 111       struct nv84_video_buffer *frame = (struct nv84_video_buffer *)desc->ref[i];
 112       if (!frame) break;
 113       /* The frame index is relative to the last IDR frame. So once the frame
 114        * num goes back to 0, previous reference frames need to have a negative
 115        * index.
 116        */
 117       if (desc->frame_num >= frame->frame_num_max) {
 118          frame->frame_num_max = desc->frame_num;
 119       } else {
 120          frame->frame_num -= frame->frame_num_max + 1;
 121          frame->frame_num_max = desc->frame_num;
 122       }
 123       ref->non_existing = 0;
 124       ref->field_is_ref = (desc->top_is_reference[i] ? 1 : 0) |
 125          (desc->bottom_is_reference[i] ? 2 : 0);
 126       ref->is_long_term = desc->is_long_term[i];
 127       ref->field_order_cnt[0] = desc->field_order_cnt_list[i][0];
 128       ref->field_order_cnt[1] = desc->field_order_cnt_list[i][1];
 129       ref->frame_idx = frame->frame_num;
 130       ref->u00 = ref->mvidx = frame->mvidx;
 131       ref->field_pic_flag = desc->field_pic_flag;
 132       indexes[frame->mvidx] = 1;
 133    }
 134
 135    /* Needs to be adjusted if we ever support non-4:2:0 videos */
 136    params.iseqparm.chroma_format_idc = 1;
 137
 138    params.iseqparm.pic_width_in_mbs_minus1 = mb(dec->base.width) - 1;
 139    if (desc->field_pic_flag || desc->pps->sps->mb_adaptive_frame_field_flag)
 140       params.iseqparm.pic_height_in_map_units_minus1 = mb_half(dec->base.height) - 1;
 141    else
 142       params.iseqparm.pic_height_in_map_units_minus1 = mb(dec->base.height) - 1;
 143
 144    if (desc->bottom_field_flag)
 145       params.ipicparm.curr_pic_order_cnt = desc->field_order_cnt[1];
 146    else
 147       params.ipicparm.curr_pic_order_cnt = desc->field_order_cnt[0];
 148    params.ipicparm.field_order_cnt[0] = desc->field_order_cnt[0];
 149    params.ipicparm.field_order_cnt[1] = desc->field_order_cnt[1];
 150    if (desc->is_reference) {
 151       if (dest->mvidx < 0) {
 152          for (i = 0; i < desc->num_ref_frames + 1; i++) {
 153             if (!indexes[i]) {
 154                dest->mvidx = i;
 155                break;
 156             }
 157          }
 158          assert(i != desc->num_ref_frames + 1);
 159       }
 160
 161       params.ipicparm.u1cc = params.ipicparm.curr_mvidx = dest->mvidx;
 162    }
 163
 164    params.iseqparm.num_ref_frames = desc->num_ref_frames;
 165    params.iseqparm.mb_adaptive_frame_field_flag = desc->pps->sps->mb_adaptive_frame_field_flag;
 166    params.ipicparm.constrained_intra_pred_flag = desc->pps->constrained_intra_pred_flag;
 167    params.ipicparm.weighted_pred_flag = desc->pps->weighted_pred_flag;
 168    params.ipicparm.weighted_bipred_idc = desc->pps->weighted_bipred_idc;
 169    params.iseqparm.frame_mbs_only_flag = desc->pps->sps->frame_mbs_only_flag;
 170    params.ipicparm.transform_8x8_mode_flag = desc->pps->transform_8x8_mode_flag;
 171    params.ipicparm.chroma_qp_index_offset = desc->pps->chroma_qp_index_offset;
 172    params.ipicparm.second_chroma_qp_index_offset = desc->pps->second_chroma_qp_index_offset;
 173    params.ipicparm.pic_init_qp_minus26 = desc->pps->pic_init_qp_minus26;
 174    params.ipicparm.num_ref_idx_l0_active_minus1 = desc->num_ref_idx_l0_active_minus1;
 175    params.ipicparm.num_ref_idx_l1_active_minus1 = desc->num_ref_idx_l1_active_minus1;
 176    params.iseqparm.log2_max_frame_num_minus4 = desc->pps->sps->log2_max_frame_num_minus4;
 177    params.iseqparm.pic_order_cnt_type = desc->pps->sps->pic_order_cnt_type;
 178    params.iseqparm.log2_max_pic_order_cnt_lsb_minus4 = desc->pps->sps->log2_max_pic_order_cnt_lsb_minus4;
 179    params.iseqparm.delta_pic_order_always_zero_flag = desc->pps->sps->delta_pic_order_always_zero_flag;
 180    params.iseqparm.direct_8x8_inference_flag = desc->pps->sps->direct_8x8_inference_flag;
 181    params.ipicparm.entropy_coding_mode_flag = desc->pps->entropy_coding_mode_flag;
 182    params.ipicparm.pic_order_present_flag = desc->pps->bottom_field_pic_order_in_frame_present_flag;
 183    params.ipicparm.deblocking_filter_control_present_flag = desc->pps->deblocking_filter_control_present_flag;
 184    params.ipicparm.redundant_pic_cnt_present_flag = desc->pps->redundant_pic_cnt_present_flag;
 185
 186    memcpy(dec->bitstream->map, &params, sizeof(params));
 187    for (i = 0; i < num_buffers; i++) {
 188       assert(total_bytes + num_bytes[i] < dec->bitstream->size / 2 - 0x700);
 189       memcpy(dec->bitstream->map + 0x700 + total_bytes, data[i], num_bytes[i]);
 190       total_bytes += num_bytes[i];
 191    }
 192    memcpy(dec->bitstream->map + 0x700 + total_bytes, end, sizeof(end));
 193    total_bytes += sizeof(end);
 194    more_params[1] = total_bytes;
 195    memcpy(dec->bitstream->map + 0x600, more_params, sizeof(more_params));
 196
 197    PUSH_SPACE(push, 5 + 21 + 3 + 2 + 4 + 2);
 198    nouveau_pushbuf_refn(push, bo_refs, sizeof(bo_refs)/sizeof(bo_refs[0]));
 199
 200    /* Wait for the fence = 1 */
 201    BEGIN_NV04(push, SUBC_BSP(0x10), 4);
 202    PUSH_DATAh(push, dec->fence->offset);
 203    PUSH_DATA (push, dec->fence->offset);
 204    PUSH_DATA (push, 1);
 205    PUSH_DATA (push, 1);
 206
 207    /* TODO: Use both halves of bitstream/vpring for alternating frames */
 208
 209    /* Kick off the BSP */
 210    BEGIN_NV04(push, SUBC_BSP(0x400), 20);
 211    PUSH_DATA (push, dec->bitstream->offset >> 8);
 212    PUSH_DATA (push, (dec->bitstream->offset >> 8) + 7);
 213    PUSH_DATA (push, dec->bitstream->size / 2 - 0x700);
 214    PUSH_DATA (push, (dec->bitstream->offset >> 8) + 6);
 215    PUSH_DATA (push, 1);
 216    PUSH_DATA (push, dec->mbring->offset >> 8);
 217    PUSH_DATA (push, dec->frame_size);
 218    PUSH_DATA (push, (dec->mbring->offset + dec->frame_size) >> 8);
 219    PUSH_DATA (push, dec->vpring->offset >> 8);
 220    PUSH_DATA (push, dec->vpring->size / 2);
 221    PUSH_DATA (push, dec->vpring_residual);
 222    PUSH_DATA (push, dec->vpring_ctrl);
 223    PUSH_DATA (push, 0);
 224    PUSH_DATA (push, dec->vpring_residual);
 225    PUSH_DATA (push, dec->vpring_residual + dec->vpring_ctrl);
 226    PUSH_DATA (push, dec->vpring_deblock);
 227    PUSH_DATA (push, (dec->vpring->offset + dec->vpring_ctrl +
 228                      dec->vpring_residual + dec->vpring_deblock) >> 8);
 229    PUSH_DATA (push, 0x654321);
 230    PUSH_DATA (push, 0);
 231    PUSH_DATA (push, 0x100008);
 232
 233    BEGIN_NV04(push, SUBC_BSP(0x620), 2);
 234    PUSH_DATA (push, 0);
 235    PUSH_DATA (push, 0);
 236
 237    BEGIN_NV04(push, SUBC_BSP(0x300), 1);
 238    PUSH_DATA (push, 0);
 239
 240    /* Write fence = 2, intr */
 241    BEGIN_NV04(push, SUBC_BSP(0x610), 3);
 242    PUSH_DATAh(push, dec->fence->offset);
 243    PUSH_DATA (push, dec->fence->offset);
 244    PUSH_DATA (push, 2);
 245
 246    BEGIN_NV04(push, SUBC_BSP(0x304), 1);
 247    PUSH_DATA (push, 0x101);
 248    PUSH_KICK (push);
 249    return 0;
 250 }