unplugged-system/external/libavc/encoder/svc/isvce_mc.c

481 lines
17 KiB
C
Raw Normal View History

/******************************************************************************
*
* Copyright (C) 2022 The Android Open Source Project
*
* Licensed under the Apache License, Version 2.0 (the "License");
* you may not use this file except in compliance with the License.
* You may obtain a copy of the License at:
*
* http://www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an "AS IS" BASIS,
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*
*****************************************************************************
* Originally developed and contributed by Ittiam Systems Pvt. Ltd, Bangalore
*/
/**
*******************************************************************************
* @file
* isvce_mc.c
*
* @brief
* Contains definition of functions for motion compensation
*
* @author
* ittiam
*
* @par List of Functions:
* - isvce_motion_comp_luma()
* - isvce_motion_comp_chroma()
*
* @remarks
* None
*
*******************************************************************************
*/
/*****************************************************************************/
/* File Includes */
/*****************************************************************************/
/* System include files */
#include <stdio.h>
/* User include files */
#include "ih264_typedefs.h"
#include "ih264_debug.h"
#include "isvc_defs.h"
#include "iv2.h"
#include "ive2.h"
#include "ime_distortion_metrics.h"
#include "ime_defs.h"
#include "ime_structs.h"
#include "isvc_structs.h"
#include "isvc_inter_pred_filters.h"
#include "isvc_mem_fns.h"
#include "ih264_padding.h"
#include "ih264_intra_pred_filters.h"
#include "ih264_deblk_edge_filters.h"
#include "isvc_trans_quant_itrans_iquant.h"
#include "isvc_cabac_tables.h"
#include "isvce_defs.h"
#include "ih264e_error.h"
#include "ih264e_bitstream.h"
#include "irc_cntrl_param.h"
#include "irc_frame_info_collector.h"
#include "isvce_rate_control.h"
#include "isvce_cabac_structs.h"
#include "isvce_structs.h"
#include "isvce_mc.h"
#include "ih264e_half_pel.h"
#include "isvce_ibl_eval.h"
/*****************************************************************************/
/* Function Definitions */
/*****************************************************************************/
/**
******************************************************************************
*
* @brief
* performs motion compensation for a luma mb for the given mv.
*
* @par Description
* This routine performs motion compensation of an inter mb. When the inter
* mb mode is P16x16, there is no need to copy 16x16 unit from reference buffer
* to pred buffer. In this case the function returns pointer and stride of the
* ref. buffer and this info is used in place of pred buffer else where.
* In other cases, the pred buffer is populated via copy / filtering + copy
* (q pel cases) and returned.
*
* @param[in] ps_proc
* pointer to current proc ctxt
*
* @return none
*
* @remarks Assumes half pel buffers for the entire frame are populated.
*
******************************************************************************
*/
void isvce_motion_comp_luma(isvce_process_ctxt_t *ps_proc, buffer_container_t *ps_pred)
{
/* codec context */
isvce_codec_t *ps_codec = ps_proc->ps_codec;
/* me ctxt */
isvce_me_ctxt_t *ps_me_ctxt = &ps_proc->s_me_ctxt;
isa_dependent_fxns_t *ps_isa_dependent_fxns = &ps_codec->s_isa_dependent_fxns;
inter_pred_fxns_t *ps_inter_pred_fxns = &ps_isa_dependent_fxns->s_inter_pred_fxns;
/* Pointer to the structure having motion vectors, size and position of curr
* partitions */
isvce_enc_pu_t *ps_curr_pu;
/* pointers to full pel, half pel x, half pel y, half pel xy reference buffer
*/
UWORD8 *pu1_ref[4];
/* pred buffer ptr */
UWORD8 *pu1_pred;
/* strides of full pel, half pel x, half pel y, half pel xy reference buffer
*/
WORD32 i4_ref_strd[4];
/* pred buffer stride */
WORD32 i4_pred_strd = ps_proc->i4_pred_strd;
/* full pel motion vectors */
WORD32 u4_mv_x_full, u4_mv_y_full;
/* half pel motion vectors */
WORD32 u4_mv_x_hpel, u4_mv_y_hpel;
/* quarter pel motion vectors */
WORD32 u4_mv_x_qpel, u4_mv_y_qpel;
/* width & height of the partition */
UWORD32 wd, ht;
/* partition idx */
UWORD32 u4_num_prtn;
/* half / qpel coefficient */
UWORD32 u4_subpel_factor;
/* BIPRED Flag */
WORD32 i4_bipred_flag;
/* temp var */
UWORD32 u4_lkup_idx1;
if((ps_proc->ps_mb_info->u2_mb_type == BASE_MODE) && ps_proc->ps_mb_info->u1_is_intra)
{
svc_intra_pred_ctxt_t *ps_intra_pred_ctxt = ps_proc->ps_intra_pred_ctxt;
ps_pred->pv_data =
(UWORD8 *) (ps_intra_pred_ctxt->s_intra_pred_outputs.s_pred_buf.as_component_bufs[Y]
.pv_data);
ps_pred->i4_data_stride =
ps_intra_pred_ctxt->s_intra_pred_outputs.s_pred_buf.as_component_bufs[Y].i4_data_stride;
return;
}
/* Init */
i4_ref_strd[0] = ps_proc->as_ref_buf_props[0].as_component_bufs[0].i4_data_stride;
i4_ref_strd[1] = i4_ref_strd[2] = i4_ref_strd[3] = ps_me_ctxt->u4_subpel_buf_strd;
for(u4_num_prtn = 0; u4_num_prtn < ps_proc->u4_num_sub_partitions; u4_num_prtn++)
{
mv_t *ps_curr_mv;
/* update ptr to curr partition */
ps_curr_pu = ps_proc->ps_mb_info->as_pu + u4_num_prtn;
/* Set no no bipred */
i4_bipred_flag = 0;
switch(ps_curr_pu->u1_pred_mode)
{
case PRED_L0:
ps_curr_mv = &ps_curr_pu->as_me_info[0].s_mv;
pu1_ref[0] = ps_proc->as_ref_buf_props[0].as_component_bufs[0].pv_data;
break;
case PRED_L1:
ps_curr_mv = &ps_curr_pu->as_me_info[1].s_mv;
pu1_ref[0] = ps_proc->as_ref_buf_props[1].as_component_bufs[0].pv_data;
break;
case PRED_BI:
/*
* In case of PRED_BI, we only need to ensure that
* the reference buffer that gets selected is
* ps_proc->pu1_best_subpel_buf
*/
/* Dummy */
ps_curr_mv = &ps_curr_pu->as_me_info[0].s_mv;
pu1_ref[0] = ps_proc->as_ref_buf_props[0].as_component_bufs[0].pv_data;
i4_bipred_flag = 1;
break;
default:
ps_curr_mv = &ps_curr_pu->as_me_info[0].s_mv;
pu1_ref[0] = ps_proc->as_ref_buf_props[0].as_component_bufs[0].pv_data;
break;
}
/* get full pel mv's (full pel units) */
u4_mv_x_full = ps_curr_mv->i2_mvx >> 2;
u4_mv_y_full = ps_curr_mv->i2_mvy >> 2;
/* get half pel mv's */
u4_mv_x_hpel = (ps_curr_mv->i2_mvx & 0x2) >> 1;
u4_mv_y_hpel = (ps_curr_mv->i2_mvy & 0x2) >> 1;
/* get quarter pel mv's */
u4_mv_x_qpel = (ps_curr_mv->i2_mvx & 0x1);
u4_mv_y_qpel = (ps_curr_mv->i2_mvy & 0x1);
/* width and height of partition */
wd = (ps_curr_pu->u1_wd_in_4x4_m1 + 1) << 2;
ht = (ps_curr_pu->u1_ht_in_4x4_m1 + 1) << 2;
/* decision ? qpel/hpel, fpel */
u4_subpel_factor =
(u4_mv_y_hpel << 3) + (u4_mv_x_hpel << 2) + (u4_mv_y_qpel << 1) + (u4_mv_x_qpel);
/* Move ref to position given by MV */
pu1_ref[0] += ((u4_mv_y_full * i4_ref_strd[0]) + u4_mv_x_full);
/* Sub pel ptrs/ Biperd pointers init */
pu1_ref[1] = ps_proc->pu1_best_subpel_buf;
i4_ref_strd[1] = ps_proc->u4_bst_spel_buf_strd;
/* update pred buff ptr */
pu1_pred = ps_proc->pu1_pred_mb + 4 * ps_curr_pu->u1_pos_y_in_4x4 * i4_pred_strd +
4 * ps_curr_pu->u1_pos_x_in_4x4;
/* u4_lkup_idx1 will be non zero for half pel and bipred */
u4_lkup_idx1 = ((u4_subpel_factor >> 2) != 0) || i4_bipred_flag;
{
/********************************************************************/
/* if the block is P16x16 MB and mv are not quarter pel motion */
/* vectors, there is no need to copy 16x16 unit from reference frame*/
/* to pred buffer. We might as well send the reference frame buffer */
/* pointer as pred buffer (ofc with updated stride) to fwd transform*/
/* and inverse transform unit. */
/********************************************************************/
if(ps_proc->u4_num_sub_partitions == 1)
{
ps_pred->pv_data = pu1_ref[u4_lkup_idx1];
ps_pred->i4_data_stride = i4_ref_strd[u4_lkup_idx1];
}
/*
* Copying half pel or full pel to prediction buffer
* Currently ps_proc->u4_num_sub_partitions will always be 1 as we only
* support 16x16 in P mbs
*/
else
{
ps_inter_pred_fxns->pf_inter_pred_luma_copy(pu1_ref[u4_lkup_idx1], pu1_pred,
i4_ref_strd[u4_lkup_idx1], i4_pred_strd,
ht, wd, NULL, 0);
}
}
}
}
/**
******************************************************************************
*
* @brief
* performs motion compensation for chroma mb
*
* @par Description
* Copies a MB of data from the reference buffer (Full pel, half pel or q pel)
* according to the motion vectors given
*
* @param[in] ps_proc
* pointer to current proc ctxt
*
* @return none
*
* @remarks Assumes half pel and quarter pel buffers for the entire frame are
* populated.
******************************************************************************
*/
void isvce_motion_comp_chroma(isvce_process_ctxt_t *ps_proc, buffer_container_t *ps_pred)
{
/* codec context */
isvce_codec_t *ps_codec = ps_proc->ps_codec;
isa_dependent_fxns_t *ps_isa_dependent_fxns = &ps_codec->s_isa_dependent_fxns;
inter_pred_fxns_t *ps_inter_pred_fxns = &ps_isa_dependent_fxns->s_inter_pred_fxns;
/* Pointer to the structure having motion vectors, size and position of curr
* partitions */
isvce_enc_pu_t *ps_curr_pu;
/* pointers to full pel, half pel x, half pel y, half pel xy reference buffer
*/
UWORD8 *pu1_ref;
/* pred buffer ptr */
UWORD8 *pu1_pred;
/* strides of full pel reference buffer */
WORD32 i4_ref_strd;
/* pred buffer stride */
WORD32 i4_pred_strd = ps_proc->i4_pred_strd;
/* full pel motion vectors */
WORD32 u4_mv_x_full, u4_mv_y_full;
/* half pel motion vectors */
WORD32 u4_mv_x_hpel, u4_mv_y_hpel;
/* quarter pel motion vectors */
WORD32 u4_mv_x_qpel, u4_mv_y_qpel;
/* width & height of the partition */
UWORD32 wd, ht;
/* partition idx */
UWORD32 u4_num_prtn;
WORD32 u4_mv_x;
WORD32 u4_mv_y;
UWORD8 u1_dx, u1_dy;
ASSERT(ps_proc->u4_num_sub_partitions <= ENC_MAX_PU_IN_MB);
if((ps_proc->ps_mb_info->u2_mb_type == BASE_MODE) && ps_proc->ps_mb_info->u1_is_intra)
{
svc_intra_pred_ctxt_t *ps_intra_pred_ctxt = ps_proc->ps_intra_pred_ctxt;
ps_pred->pv_data =
(UWORD8 *) (ps_intra_pred_ctxt->s_intra_pred_outputs.s_pred_buf.as_component_bufs[UV]
.pv_data);
ps_pred->i4_data_stride =
ps_intra_pred_ctxt->s_intra_pred_outputs.s_pred_buf.as_component_bufs[UV]
.i4_data_stride;
return;
}
else
{
ps_pred->pv_data = ps_proc->pu1_pred_mb;
ps_pred->i4_data_stride = ps_proc->i4_pred_strd;
}
for(u4_num_prtn = 0; u4_num_prtn < ps_proc->u4_num_sub_partitions; u4_num_prtn++)
{
mv_t *ps_curr_mv;
ps_curr_pu = ps_proc->ps_mb_info->as_pu + u4_num_prtn;
if(ps_curr_pu->u1_pred_mode != BI)
{
ps_curr_mv = &ps_curr_pu->as_me_info[ps_curr_pu->u1_pred_mode].s_mv;
pu1_ref =
ps_proc->as_ref_buf_props[ps_curr_pu->u1_pred_mode].as_component_bufs[1].pv_data;
i4_ref_strd = ps_proc->as_ref_buf_props[ps_curr_pu->u1_pred_mode]
.as_component_bufs[1]
.i4_data_stride;
u4_mv_x = ps_curr_mv->i2_mvx >> 3;
u4_mv_y = ps_curr_mv->i2_mvy >> 3;
/* corresponds to full pel motion vector in luma, but in chroma
* corresponds to pel formed wiith dx, dy =4 */
u4_mv_x_full = (ps_curr_mv->i2_mvx & 0x4) >> 2;
u4_mv_y_full = (ps_curr_mv->i2_mvy & 0x4) >> 2;
/* get half pel mv's */
u4_mv_x_hpel = (ps_curr_mv->i2_mvx & 0x2) >> 1;
u4_mv_y_hpel = (ps_curr_mv->i2_mvy & 0x2) >> 1;
/* get quarter pel mv's */
u4_mv_x_qpel = (ps_curr_mv->i2_mvx & 0x1);
u4_mv_y_qpel = (ps_curr_mv->i2_mvy & 0x1);
/* width and height of sub macro block */
wd = (ps_curr_pu->u1_wd_in_4x4_m1 + 1) << 1;
ht = (ps_curr_pu->u1_ht_in_4x4_m1 + 1) << 1;
/* move the pointers so that they point to the motion compensated
* locations */
pu1_ref += ((u4_mv_y * i4_ref_strd) + (u4_mv_x << 1));
pu1_pred = ps_proc->pu1_pred_mb + 4 * ps_curr_pu->u1_pos_y_in_4x4 * i4_pred_strd +
2 * ps_curr_pu->u1_pos_x_in_4x4;
u1_dx = (u4_mv_x_full << 2) + (u4_mv_x_hpel << 1) + (u4_mv_x_qpel);
u1_dy = (u4_mv_y_full << 2) + (u4_mv_y_hpel << 1) + (u4_mv_y_qpel);
/* cases where u1_dx = 0 or u1_dy = 0 are dealt separately in neon with
* separate functions for better performance
*
* isvc_inter_pred_chroma_dx_zero_a9q
* and
* isvc_inter_pred_chroma_dy_zero_a9q
*/
ps_inter_pred_fxns->pf_inter_pred_chroma(pu1_ref, pu1_pred, i4_ref_strd, i4_pred_strd,
u1_dx, u1_dy, ht, wd);
}
else
{
/*
* We need to interpolate the L0 and L1 ref pics with the chorma MV
* then use them to average for bilinrar interpred
*/
WORD32 i4_predmode;
UWORD8 *pu1_ref_buf[2];
/* Temporary buffers to store the interpolated value from L0 and L1 */
pu1_ref_buf[L0] = ps_proc->apu1_subpel_buffs[0];
pu1_ref_buf[L1] = ps_proc->apu1_subpel_buffs[1];
for(i4_predmode = 0; i4_predmode < BI; i4_predmode++)
{
ps_curr_mv = &ps_curr_pu->as_me_info[i4_predmode].s_mv;
pu1_ref = ps_proc->as_ref_buf_props[i4_predmode].as_component_bufs[1].pv_data;
i4_ref_strd =
ps_proc->as_ref_buf_props[i4_predmode].as_component_bufs[1].i4_data_stride;
u4_mv_x = ps_curr_mv->i2_mvx >> 3;
u4_mv_y = ps_curr_mv->i2_mvy >> 3;
/*
* corresponds to full pel motion vector in luma, but in chroma
* corresponds to pel formed wiith dx, dy =4
*/
u4_mv_x_full = (ps_curr_mv->i2_mvx & 0x4) >> 2;
u4_mv_y_full = (ps_curr_mv->i2_mvy & 0x4) >> 2;
/* get half pel mv's */
u4_mv_x_hpel = (ps_curr_mv->i2_mvx & 0x2) >> 1;
u4_mv_y_hpel = (ps_curr_mv->i2_mvy & 0x2) >> 1;
/* get quarter pel mv's */
u4_mv_x_qpel = (ps_curr_mv->i2_mvx & 0x1);
u4_mv_y_qpel = (ps_curr_mv->i2_mvy & 0x1);
/* width and height of sub macro block */
wd = (ps_curr_pu->u1_wd_in_4x4_m1 + 1) << 1;
ht = (ps_curr_pu->u1_ht_in_4x4_m1 + 1) << 1;
/* move the pointers so that they point to the motion compensated
* locations */
pu1_ref += ((u4_mv_y * i4_ref_strd) + (u4_mv_x << 1));
pu1_pred = ps_proc->pu1_pred_mb + 4 * ps_curr_pu->u1_pos_y_in_4x4 * i4_pred_strd +
2 * ps_curr_pu->u1_pos_x_in_4x4;
u1_dx = (u4_mv_x_full << 2) + (u4_mv_x_hpel << 1) + (u4_mv_x_qpel);
u1_dy = (u4_mv_y_full << 2) + (u4_mv_y_hpel << 1) + (u4_mv_y_qpel);
ps_inter_pred_fxns->pf_inter_pred_chroma(
pu1_ref, pu1_ref_buf[i4_predmode], i4_ref_strd, MB_SIZE, u1_dx, u1_dy, ht, wd);
}
ps_inter_pred_fxns->pf_inter_pred_luma_bilinear(pu1_ref_buf[L0], pu1_ref_buf[L1],
pu1_pred, MB_SIZE, MB_SIZE,
i4_pred_strd, MB_SIZE >> 1, MB_SIZE);
}
}
}