view vaapi_h264.c @ 11575:5a7797213359 libavcodec

Reduce the size of the bidir refine hashtable from 4096 to 256 and change the hash function to something more sane and simple. about 1/3 faster, no meassureable change in psnr or size and i gues its even same md5 with my test file but i forgot testing that.

/*
 * H.264 HW decode acceleration through VA API
 *
 * Copyright (C) 2008-2009 Splitted-Desktop Systems
 *
 * This file is part of FFmpeg.
 *
 * FFmpeg is free software; you can redistribute it and/or
 * modify it under the terms of the GNU Lesser General Public
 * License as published by the Free Software Foundation; either
 * version 2.1 of the License, or (at your option) any later version.
 *
 * FFmpeg is distributed in the hope that it will be useful,
 * but WITHOUT ANY WARRANTY; without even the implied warranty of
 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
 * Lesser General Public License for more details.
 *
 * You should have received a copy of the GNU Lesser General Public
 * License along with FFmpeg; if not, write to the Free Software
 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
 */

#include "vaapi_internal.h"
#include "h264.h"

/** @file
 *  This file implements the glue code between FFmpeg's and VA API's
 *  structures for H.264 decoding.
 */

/**
 * Initializes an empty VA API picture.
 *
 * VA API requires a fixed-size reference picture array.
 */
static void init_vaapi_pic(VAPictureH264 *va_pic)
{
    va_pic->picture_id          = VA_INVALID_ID;
    va_pic->flags               = VA_PICTURE_H264_INVALID;
    va_pic->TopFieldOrderCnt    = 0;
    va_pic->BottomFieldOrderCnt = 0;
}

/**
 * Translates an FFmpeg Picture into its VA API form.
 *
 * @param[out] va_pic          A pointer to VA API's own picture struct
 * @param[in]  pic             A pointer to the FFmpeg picture struct to convert
 * @param[in]  pic_structure   The picture field type (as defined in mpegvideo.h),
 *                             supersedes pic's field type if nonzero.
 */
static void fill_vaapi_pic(VAPictureH264 *va_pic,
                           Picture       *pic,
                           int            pic_structure)
{
    if (pic_structure == 0)
        pic_structure = pic->reference;

    va_pic->picture_id = ff_vaapi_get_surface_id(pic);
    va_pic->frame_idx  = pic->long_ref ? pic->pic_id : pic->frame_num;

    va_pic->flags      = 0;
    if (pic_structure != PICT_FRAME)
        va_pic->flags |= (pic_structure & PICT_TOP_FIELD) ? VA_PICTURE_H264_TOP_FIELD : VA_PICTURE_H264_BOTTOM_FIELD;
    if (pic->reference)
        va_pic->flags |= pic->long_ref ? VA_PICTURE_H264_LONG_TERM_REFERENCE : VA_PICTURE_H264_SHORT_TERM_REFERENCE;

    va_pic->TopFieldOrderCnt = 0;
    if (pic->field_poc[0] != INT_MAX)
        va_pic->TopFieldOrderCnt = pic->field_poc[0];

    va_pic->BottomFieldOrderCnt = 0;
    if (pic->field_poc[1] != INT_MAX)
        va_pic->BottomFieldOrderCnt = pic->field_poc[1];
}

/** Decoded Picture Buffer (DPB). */
typedef struct DPB {
    int            size;        ///< Current number of reference frames in the DPB
    int            max_size;    ///< Max number of reference frames. This is FF_ARRAY_ELEMS(VAPictureParameterBufferH264.ReferenceFrames)
    VAPictureH264 *va_pics;     ///< Pointer to VAPictureParameterBufferH264.ReferenceFrames array
} DPB;

/**
 * Appends picture to the decoded picture buffer, in a VA API form that
 * merges the second field picture attributes with the first, if
 * available.  The decoded picture buffer's size must be large enough
 * to receive the new VA API picture object.
 */
static int dpb_add(DPB *dpb, Picture *pic)
{
    int i;

    if (dpb->size >= dpb->max_size)
        return -1;

    for (i = 0; i < dpb->size; i++) {
        VAPictureH264 * const va_pic = &dpb->va_pics[i];
        if (va_pic->picture_id == ff_vaapi_get_surface_id(pic)) {
            VAPictureH264 temp_va_pic;
            fill_vaapi_pic(&temp_va_pic, pic, 0);

            if ((temp_va_pic.flags ^ va_pic->flags) & (VA_PICTURE_H264_TOP_FIELD | VA_PICTURE_H264_BOTTOM_FIELD)) {
                va_pic->flags |= temp_va_pic.flags & (VA_PICTURE_H264_TOP_FIELD | VA_PICTURE_H264_BOTTOM_FIELD);
                /* Merge second field */
                if (temp_va_pic.flags & VA_PICTURE_H264_TOP_FIELD) {
                    va_pic->TopFieldOrderCnt    = temp_va_pic.TopFieldOrderCnt;
                } else {
                    va_pic->BottomFieldOrderCnt = temp_va_pic.BottomFieldOrderCnt;
                }
            }
            return 0;
        }
    }

    fill_vaapi_pic(&dpb->va_pics[dpb->size++], pic, 0);
    return 0;
}

/** Fills in VA API reference frames array. */
static int fill_vaapi_ReferenceFrames(VAPictureParameterBufferH264 *pic_param,
                                      H264Context                  *h)
{
    DPB dpb;
    int i;

    dpb.size     = 0;
    dpb.max_size = FF_ARRAY_ELEMS(pic_param->ReferenceFrames);
    dpb.va_pics  = pic_param->ReferenceFrames;
    for (i = 0; i < dpb.max_size; i++)
        init_vaapi_pic(&dpb.va_pics[i]);

    for (i = 0; i < h->short_ref_count; i++) {
        Picture * const pic = h->short_ref[i];
        if (pic && pic->reference && dpb_add(&dpb, pic) < 0)
            return -1;
    }

    for (i = 0; i < h->long_ref_count; i++) {
        Picture * const pic = h->long_ref[i];
        if (pic && pic->reference && dpb_add(&dpb, pic) < 0)
            return -1;
    }
    return 0;
}

/**
 * Fills in VA API reference picture lists from the FFmpeg reference
 * picture list.
 *
 * @param[out] RefPicList  VA API internal reference picture list
 * @param[in]  ref_list    A pointer to the FFmpeg reference list
 * @param[in]  ref_count   The number of reference pictures in ref_list
 */
static void fill_vaapi_RefPicList(VAPictureH264 RefPicList[32],
                                  Picture      *ref_list,
                                  unsigned int  ref_count)
{
    unsigned int i, n = 0;
    for (i = 0; i < ref_count; i++)
        if (ref_list[i].reference)
            fill_vaapi_pic(&RefPicList[n++], &ref_list[i], 0);

    for (; n < 32; n++)
        init_vaapi_pic(&RefPicList[n]);
}

/**
 * Fills in prediction weight table.
 *
 * VA API requires a plain prediction weight table as it does not infer
 * any value.
 *
 * @param[in]  h                   A pointer to the current H.264 context
 * @param[in]  list                The reference frame list index to use
 * @param[out] luma_weight_flag    VA API plain luma weight flag
 * @param[out] luma_weight         VA API plain luma weight table
 * @param[out] luma_offset         VA API plain luma offset table
 * @param[out] chroma_weight_flag  VA API plain chroma weight flag
 * @param[out] chroma_weight       VA API plain chroma weight table
 * @param[out] chroma_offset       VA API plain chroma offset table
 */
static void fill_vaapi_plain_pred_weight_table(H264Context   *h,
                                               int            list,
                                               unsigned char *luma_weight_flag,
                                               short          luma_weight[32],
                                               short          luma_offset[32],
                                               unsigned char *chroma_weight_flag,
                                               short          chroma_weight[32][2],
                                               short          chroma_offset[32][2])
{
    unsigned int i, j;

    *luma_weight_flag    = h->luma_weight_flag[list];
    *chroma_weight_flag  = h->chroma_weight_flag[list];

    for (i = 0; i < h->ref_count[list]; i++) {
        /* VA API also wants the inferred (default) values, not
           only what is available in the bitstream (7.4.3.2). */
        if (h->luma_weight_flag[list]) {
            luma_weight[i] = h->luma_weight[i][list][0];
            luma_offset[i] = h->luma_weight[i][list][1];
        } else {
            luma_weight[i] = 1 << h->luma_log2_weight_denom;
            luma_offset[i] = 0;
        }
        for (j = 0; j < 2; j++) {
            if (h->chroma_weight_flag[list]) {
                chroma_weight[i][j] = h->chroma_weight[i][list][j][0];
                chroma_offset[i][j] = h->chroma_weight[i][list][j][1];
            } else {
                chroma_weight[i][j] = 1 << h->chroma_log2_weight_denom;
                chroma_offset[i][j] = 0;
            }
        }
    }
}

/** Initializes and starts decoding a frame with VA API. */
static int start_frame(AVCodecContext          *avctx,
                       av_unused const uint8_t *buffer,
                       av_unused uint32_t       size)
{
    H264Context * const h = avctx->priv_data;
    MpegEncContext * const s = &h->s;
    struct vaapi_context * const vactx = avctx->hwaccel_context;
    VAPictureParameterBufferH264 *pic_param;
    VAIQMatrixBufferH264 *iq_matrix;

    dprintf(avctx, "start_frame()\n");

    vactx->slice_param_size = sizeof(VASliceParameterBufferH264);

    /* Fill in VAPictureParameterBufferH264. */
    pic_param = ff_vaapi_alloc_pic_param(vactx, sizeof(VAPictureParameterBufferH264));
    if (!pic_param)
        return -1;
    fill_vaapi_pic(&pic_param->CurrPic, s->current_picture_ptr, s->picture_structure);
    if (fill_vaapi_ReferenceFrames(pic_param, h) < 0)
        return -1;
    pic_param->picture_width_in_mbs_minus1                      = s->mb_width - 1;
    pic_param->picture_height_in_mbs_minus1                     = s->mb_height - 1;
    pic_param->bit_depth_luma_minus8                            = h->sps.bit_depth_luma - 8;
    pic_param->bit_depth_chroma_minus8                          = h->sps.bit_depth_chroma - 8;
    pic_param->num_ref_frames                                   = h->sps.ref_frame_count;
    pic_param->seq_fields.value                                 = 0; /* reset all bits */
    pic_param->seq_fields.bits.chroma_format_idc                = h->sps.chroma_format_idc;
    pic_param->seq_fields.bits.residual_colour_transform_flag   = h->sps.residual_color_transform_flag; /* XXX: only for 4:4:4 high profile? */
    pic_param->seq_fields.bits.gaps_in_frame_num_value_allowed_flag = h->sps.gaps_in_frame_num_allowed_flag;
    pic_param->seq_fields.bits.frame_mbs_only_flag              = h->sps.frame_mbs_only_flag;
    pic_param->seq_fields.bits.mb_adaptive_frame_field_flag     = h->sps.mb_aff;
    pic_param->seq_fields.bits.direct_8x8_inference_flag        = h->sps.direct_8x8_inference_flag;
    pic_param->seq_fields.bits.MinLumaBiPredSize8x8             = h->sps.level_idc >= 31; /* A.3.3.2 */
    pic_param->seq_fields.bits.log2_max_frame_num_minus4        = h->sps.log2_max_frame_num - 4;
    pic_param->seq_fields.bits.pic_order_cnt_type               = h->sps.poc_type;
    pic_param->seq_fields.bits.log2_max_pic_order_cnt_lsb_minus4 = h->sps.log2_max_poc_lsb - 4;
    pic_param->seq_fields.bits.delta_pic_order_always_zero_flag = h->sps.delta_pic_order_always_zero_flag;
    pic_param->num_slice_groups_minus1                          = h->pps.slice_group_count - 1;
    pic_param->slice_group_map_type                             = h->pps.mb_slice_group_map_type;
    pic_param->slice_group_change_rate_minus1                   = 0; /* XXX: unimplemented in FFmpeg */
    pic_param->pic_init_qp_minus26                              = h->pps.init_qp - 26;
    pic_param->pic_init_qs_minus26                              = h->pps.init_qs - 26;
    pic_param->chroma_qp_index_offset                           = h->pps.chroma_qp_index_offset[0];
    pic_param->second_chroma_qp_index_offset                    = h->pps.chroma_qp_index_offset[1];
    pic_param->pic_fields.value                                 = 0; /* reset all bits */
    pic_param->pic_fields.bits.entropy_coding_mode_flag         = h->pps.cabac;
    pic_param->pic_fields.bits.weighted_pred_flag               = h->pps.weighted_pred;
    pic_param->pic_fields.bits.weighted_bipred_idc              = h->pps.weighted_bipred_idc;
    pic_param->pic_fields.bits.transform_8x8_mode_flag          = h->pps.transform_8x8_mode;
    pic_param->pic_fields.bits.field_pic_flag                   = s->picture_structure != PICT_FRAME;
    pic_param->pic_fields.bits.constrained_intra_pred_flag      = h->pps.constrained_intra_pred;
    pic_param->pic_fields.bits.pic_order_present_flag           = h->pps.pic_order_present;
    pic_param->pic_fields.bits.deblocking_filter_control_present_flag = h->pps.deblocking_filter_parameters_present;
    pic_param->pic_fields.bits.redundant_pic_cnt_present_flag   = h->pps.redundant_pic_cnt_present;
    pic_param->pic_fields.bits.reference_pic_flag               = h->nal_ref_idc != 0;
    pic_param->frame_num                                        = h->frame_num;

    /* Fill in VAIQMatrixBufferH264. */
    iq_matrix = ff_vaapi_alloc_iq_matrix(vactx, sizeof(VAIQMatrixBufferH264));
    if (!iq_matrix)
        return -1;
    memcpy(iq_matrix->ScalingList4x4, h->pps.scaling_matrix4, sizeof(iq_matrix->ScalingList4x4));
    memcpy(iq_matrix->ScalingList8x8, h->pps.scaling_matrix8, sizeof(iq_matrix->ScalingList8x8));
    return 0;
}

/** Ends a hardware decoding based frame. */
static int end_frame(AVCodecContext *avctx)
{
    H264Context * const h = avctx->priv_data;

    dprintf(avctx, "end_frame()\n");
    return ff_vaapi_common_end_frame(&h->s);
}

/** Decodes the given H.264 slice with VA API. */
static int decode_slice(AVCodecContext *avctx,
                        const uint8_t  *buffer,
                        uint32_t        size)
{
    H264Context * const h = avctx->priv_data;
    MpegEncContext * const s = &h->s;
    VASliceParameterBufferH264 *slice_param;

    dprintf(avctx, "decode_slice(): buffer %p, size %d\n", buffer, size);

    /* Fill in VASliceParameterBufferH264. */
    slice_param = (VASliceParameterBufferH264 *)ff_vaapi_alloc_slice(avctx->hwaccel_context, buffer, size);
    if (!slice_param)
        return -1;
    slice_param->slice_data_bit_offset          = get_bits_count(&h->s.gb) + 8; /* bit buffer started beyond nal_unit_type */
    slice_param->first_mb_in_slice              = (s->mb_y >> FIELD_OR_MBAFF_PICTURE) * s->mb_width + s->mb_x;
    slice_param->slice_type                     = ff_h264_get_slice_type(h);
    slice_param->direct_spatial_mv_pred_flag    = h->slice_type == FF_B_TYPE ? h->direct_spatial_mv_pred : 0;
    slice_param->num_ref_idx_l0_active_minus1   = h->list_count > 0 ? h->ref_count[0] - 1 : 0;
    slice_param->num_ref_idx_l1_active_minus1   = h->list_count > 1 ? h->ref_count[1] - 1 : 0;
    slice_param->cabac_init_idc                 = h->cabac_init_idc;
    slice_param->slice_qp_delta                 = s->qscale - h->pps.init_qp;
    slice_param->disable_deblocking_filter_idc  = h->deblocking_filter < 2 ? !h->deblocking_filter : h->deblocking_filter;
    slice_param->slice_alpha_c0_offset_div2     = h->slice_alpha_c0_offset / 2 - 26;
    slice_param->slice_beta_offset_div2         = h->slice_beta_offset     / 2 - 26;
    slice_param->luma_log2_weight_denom         = h->luma_log2_weight_denom;
    slice_param->chroma_log2_weight_denom       = h->chroma_log2_weight_denom;

    fill_vaapi_RefPicList(slice_param->RefPicList0, h->ref_list[0], h->list_count > 0 ? h->ref_count[0] : 0);
    fill_vaapi_RefPicList(slice_param->RefPicList1, h->ref_list[1], h->list_count > 1 ? h->ref_count[1] : 0);

    fill_vaapi_plain_pred_weight_table(h, 0,
                                       &slice_param->luma_weight_l0_flag,   slice_param->luma_weight_l0,   slice_param->luma_offset_l0,
                                       &slice_param->chroma_weight_l0_flag, slice_param->chroma_weight_l0, slice_param->chroma_offset_l0);
    fill_vaapi_plain_pred_weight_table(h, 1,
                                       &slice_param->luma_weight_l1_flag,   slice_param->luma_weight_l1,   slice_param->luma_offset_l1,
                                       &slice_param->chroma_weight_l1_flag, slice_param->chroma_weight_l1, slice_param->chroma_offset_l1);
    return 0;
}

AVHWAccel h264_vaapi_hwaccel = {
    .name           = "h264_vaapi",
    .type           = AVMEDIA_TYPE_VIDEO,
    .id             = CODEC_ID_H264,
    .pix_fmt        = PIX_FMT_VAAPI_VLD,
    .capabilities   = 0,
    .start_frame    = start_frame,
    .end_frame      = end_frame,
    .decode_slice   = decode_slice,
    .priv_data_size = 0,
};