int i_lambda2;
int i_qp;
int16_t *p_cost_mv;
- int b_mbrd;
+ int i_mbrd;
/* I: Intra part */
static void x264_mb_analyse_init( x264_t *h, x264_mb_analysis_t *a, int i_qp )
{
+ int i = h->param.analyse.i_subpel_refine - (h->sh.i_type == SLICE_TYPE_B);
+ /* mbrd == 1 -> RD mode decision */
+ /* mbrd == 2 -> RD refinement */
+ a->i_mbrd = (i>=6) + (i>=8);
/* conduct the analysis using this lamda and QP */
a->i_qp = h->mb.i_qp = i_qp;
h->mb.i_chroma_qp = h->chroma_qp_table[i_qp];
a->i_lambda = x264_lambda_tab[i_qp];
a->i_lambda2 = x264_lambda2_tab[i_qp];
- a->b_mbrd = h->param.analyse.i_subpel_refine >= 6 &&
- ( h->sh.i_type != SLICE_TYPE_B || h->param.analyse.b_bframe_rdo );
-
h->mb.i_me_method = h->param.analyse.i_me_method;
h->mb.i_subpel_refine = h->param.analyse.i_subpel_refine;
h->mb.b_chroma_me = h->param.analyse.b_chroma_me && h->sh.i_type == SLICE_TYPE_P
&& h->mb.i_subpel_refine >= 5;
- h->mb.b_trellis = h->param.analyse.i_trellis > 1 && a->b_mbrd;
+ h->mb.b_trellis = h->param.analyse.i_trellis > 1 && a->i_mbrd;
h->mb.b_transform_8x8 = 0;
h->mb.b_noise_reduction = 0;
a->i_satd_i4x4 =
a->i_satd_i8x8chroma = COST_MAX;
- /* non-RD PCM decision is inaccurate, so don't do it */
- a->i_satd_pcm = a->b_mbrd ? ((uint64_t)X264_PCM_COST*a->i_lambda2 + 128) >> 8 : COST_MAX;
+ /* non-RD PCM decision is inaccurate (as is psy-rd), so don't do it */
+ a->i_satd_pcm = !h->mb.i_psy_rd && a->i_mbrd ? ((uint64_t)X264_PCM_COST*a->i_lambda2 + 128) >> 8 : COST_MAX;
a->b_fast_intra = 0;
h->mb.i_skip_intra =
h->mb.b_lossless ? 0 :
- a->b_mbrd ? 2 :
+ a->i_mbrd ? 2 :
!h->param.analyse.i_trellis && !h->param.analyse.i_noise_reduction;
/* II: Inter part P/B frame */
int i_mode = predict_mode[i];
/* we do the prediction */
- h->predict_8x8c[i_mode]( p_dstc[0] );
- h->predict_8x8c[i_mode]( p_dstc[1] );
+ if( h->mb.b_lossless )
+ x264_predict_lossless_8x8_chroma( h, i_mode );
+ else
+ {
+ h->predict_8x8c[i_mode]( p_dstc[0] );
+ h->predict_8x8c[i_mode]( p_dstc[1] );
+ }
/* we calculate the cost */
i_satd = h->pixf.mbcmp[PIXEL_8x8]( p_dstc[0], FDEC_STRIDE,
int i, idx;
int i_max;
int predict_mode[9];
- int b_merged_satd = !!h->pixf.intra_mbcmp_x3_16x16;
+ int b_merged_satd = !!h->pixf.intra_mbcmp_x3_16x16 && !h->mb.b_lossless;
/*---------------- Try all mode and calculate their score ---------------*/
{
int i_satd;
int i_mode = predict_mode[i];
- h->predict_16x16[i_mode]( p_dst );
+
+ if( h->mb.b_lossless )
+ x264_predict_lossless_16x16( h, i_mode );
+ else
+ h->predict_16x16[i_mode]( p_dst );
i_satd = h->pixf.mbcmp[PIXEL_16x16]( p_dst, FDEC_STRIDE, p_src, FENC_STRIDE ) +
a->i_lambda * bs_size_ue( x264_mb_pred_mode16x16_fix[i_mode] );
{
DECLARE_ALIGNED_16( uint8_t edge[33] );
x264_pixel_cmp_t sa8d = (h->pixf.mbcmp[0] == h->pixf.satd[0]) ? h->pixf.sa8d[PIXEL_8x8] : h->pixf.mbcmp[PIXEL_8x8];
- int i_satd_thresh = a->b_mbrd ? COST_MAX : X264_MIN( i_satd_inter, a->i_satd_i16x16 );
+ int i_satd_thresh = a->i_mbrd ? COST_MAX : X264_MIN( i_satd_inter, a->i_satd_i16x16 );
int i_cost = 0;
b_merged_satd = h->pixf.intra_sa8d_x3_8x8 && h->pixf.mbcmp[0] == h->pixf.satd[0];
int i_satd;
int i_mode = predict_mode[i];
- h->predict_8x8[i_mode]( p_dst_by, edge );
+ if( h->mb.b_lossless )
+ x264_predict_lossless_8x8( h, p_dst_by, idx, i_mode, edge );
+ else
+ h->predict_8x8[i_mode]( p_dst_by, edge );
i_satd = sa8d( p_dst_by, FDEC_STRIDE, p_src_by, FENC_STRIDE )
+ a->i_lambda * (i_pred_mode == x264_mb_pred_mode4x4_fix(i_mode) ? 1 : 4);
a->i_satd_i8x8 = COST_MAX;
i_cost = i_cost * 4/(idx+1);
}
- if( X264_MIN(i_cost, a->i_satd_i16x16) > i_satd_inter*(5+a->b_mbrd)/4 )
+ if( X264_MIN(i_cost, a->i_satd_i16x16) > i_satd_inter*(5+!!a->i_mbrd)/4 )
return;
}
int i_cost;
int i_satd_thresh = X264_MIN3( i_satd_inter, a->i_satd_i16x16, a->i_satd_i8x8 );
b_merged_satd = h->pixf.intra_satd_x3_4x4 && h->pixf.mbcmp[0] == h->pixf.satd[0];
- if( a->b_mbrd )
+ if( a->i_mbrd )
i_satd_thresh = i_satd_thresh * (10-a->b_fast_intra)/8;
i_cost = a->i_lambda * 24; /* from JVT (SATD0) */
{
int i_satd;
int i_mode = predict_mode[i];
-
- h->predict_4x4[i_mode]( p_dst_by );
+ if( h->mb.b_lossless )
+ x264_predict_lossless_4x4( h, p_dst_by, idx, i_mode );
+ else
+ h->predict_4x4[i_mode]( p_dst_by );
i_satd = h->pixf.mbcmp[PIXEL_4x4]( p_dst_by, FDEC_STRIDE,
p_src_by, FENC_STRIDE )
for( i = 0; i < i_max; i++ )
{
i_mode = predict_mode[i];
- h->predict_4x4[i_mode]( p_dst_by );
+ if( h->mb.b_lossless )
+ x264_predict_lossless_4x4( h, p_dst_by, idx, i_mode );
+ else
+ h->predict_4x4[i_mode]( p_dst_by );
i_satd = x264_rd_cost_i4x4( h, a->i_lambda2, idx, i_mode );
if( i_best > i_satd )
i_mode = predict_mode[i];
if( a->i_satd_i8x8_dir[i_mode][idx] > i_thresh )
continue;
- h->predict_8x8[i_mode]( p_dst_by, edge );
+ if( h->mb.b_lossless )
+ x264_predict_lossless_8x8( h, p_dst_by, idx, i_mode, edge );
+ else
+ h->predict_8x8[i_mode]( p_dst_by, edge );
i_satd = x264_rd_cost_i8x8( h, a->i_lambda2, idx, i_mode );
if( i_best > i_satd )
for( i = 0; i < i_max; i++ )
{
i_mode = predict_mode[i];
- h->predict_8x8c[i_mode]( h->mb.pic.p_fdec[1] );
- h->predict_8x8c[i_mode]( h->mb.pic.p_fdec[2] );
+ if( h->mb.b_lossless )
+ x264_predict_lossless_8x8_chroma( h, i_mode );
+ else
+ {
+ h->predict_8x8c[i_mode]( h->mb.pic.p_fdec[1] );
+ h->predict_8x8c[i_mode]( h->mb.pic.p_fdec[2] );
+ }
/* if we've already found a mode that needs no residual, then
* probably any mode with a residual will be worse.
* so avoid dct on the remaining modes to improve speed. */
{
x264_me_t m;
int i_ref, i_mvc;
- DECLARE_ALIGNED_4( int16_t mvc[7][2] );
+ DECLARE_ALIGNED_4( int16_t mvc[8][2] );
int i_halfpel_thresh = INT_MAX;
int *p_halfpel_thresh = h->mb.pic.i_fref[0]>1 ? &i_halfpel_thresh : NULL;
assert( a->l0.me16x16.mv[1] <= h->mb.mv_max_spel[1] || h->param.i_threads == 1 );
h->mb.i_type = P_L0;
- if( a->b_mbrd )
+ if( a->i_mbrd )
{
x264_mb_cache_fenc_satd( h );
if( a->l0.me16x16.i_ref == 0 && *(uint32_t*)a->l0.me16x16.mv == *(uint32_t*)h->mb.cache.pskip_mv )
}
}
-#define WEIGHTED_AVG( size, pix1, stride1, src2, stride2 ) \
- { \
- if( h->param.analyse.b_weighted_bipred ) \
- h->mc.avg_weight[size]( pix1, stride1, src2, stride2, \
- h->mb.bipred_weight[a->l0.i_ref][a->l1.i_ref] ); \
- else \
- h->mc.avg[size]( pix1, stride1, src2, stride2 ); \
- }
+#define WEIGHTED_AVG( size, pix, stride, src1, stride1, src2, stride2 ) \
+{ \
+ h->mc.avg[size]( pix, stride, src1, stride1, src2, stride2, h->mb.bipred_weight[a->l0.i_ref][a->l1.i_ref] ); \
+}
static void x264_mb_analyse_inter_b16x16( x264_t *h, x264_mb_analysis_t *a )
{
+ DECLARE_ALIGNED_16( uint8_t pix0[16*16] );
DECLARE_ALIGNED_16( uint8_t pix1[16*16] );
- DECLARE_ALIGNED_16( uint8_t pix2[16*16] );
- uint8_t *src2;
- int stride2 = 16;
- int weight;
+ uint8_t *src0, *src1;
+ int stride0 = 16, stride1 = 16;
x264_me_t m;
int i_ref, i_mvc;
- DECLARE_ALIGNED_4( int16_t mvc[8][2] );
+ DECLARE_ALIGNED_4( int16_t mvc[9][2] );
int i_halfpel_thresh = INT_MAX;
int *p_halfpel_thresh = h->mb.pic.i_fref[0]>1 ? &i_halfpel_thresh : NULL;
x264_macroblock_cache_ref( h, 0, 0, 4, 4, 1, a->l1.i_ref );
/* get cost of BI mode */
- weight = h->mb.bipred_weight[a->l0.i_ref][a->l1.i_ref];
- if ( (*(uint32_t*)a->l0.me16x16.mv & 0x10001) == 0 )
- {
- /* l0 reference is halfpel, so get_ref on it will make it faster */
- src2 =
- h->mc.get_ref( pix2, &stride2,
- h->mb.pic.p_fref[0][a->l0.i_ref], h->mb.pic.i_stride[0],
- a->l0.me16x16.mv[0], a->l0.me16x16.mv[1],
- 16, 16 );
- h->mc.mc_luma( pix1, 16,
- h->mb.pic.p_fref[1][a->l1.i_ref], h->mb.pic.i_stride[0],
- a->l1.me16x16.mv[0], a->l1.me16x16.mv[1],
- 16, 16 );
- weight = 64 - weight;
- }
- else
- {
- /* if l0 was qpel, we'll use get_ref on l1 instead */
- h->mc.mc_luma( pix1, 16,
- h->mb.pic.p_fref[0][a->l0.i_ref], h->mb.pic.i_stride[0],
- a->l0.me16x16.mv[0], a->l0.me16x16.mv[1],
- 16, 16 );
- src2 =
- h->mc.get_ref( pix2, &stride2,
- h->mb.pic.p_fref[1][a->l1.i_ref], h->mb.pic.i_stride[0],
- a->l1.me16x16.mv[0], a->l1.me16x16.mv[1],
- 16, 16 );
- }
+ src0 = h->mc.get_ref( pix0, &stride0,
+ h->mb.pic.p_fref[0][a->l0.i_ref], h->mb.pic.i_stride[0],
+ a->l0.me16x16.mv[0], a->l0.me16x16.mv[1], 16, 16 );
+ src1 = h->mc.get_ref( pix1, &stride1,
+ h->mb.pic.p_fref[1][a->l1.i_ref], h->mb.pic.i_stride[0],
+ a->l1.me16x16.mv[0], a->l1.me16x16.mv[1], 16, 16 );
- if( h->param.analyse.b_weighted_bipred )
- h->mc.avg_weight[PIXEL_16x16]( pix1, 16, src2, stride2, weight );
- else
- h->mc.avg[PIXEL_16x16]( pix1, 16, src2, stride2 );
+ h->mc.avg[PIXEL_16x16]( pix0, 16, src0, stride0, src1, stride1, h->mb.bipred_weight[a->l0.i_ref][a->l1.i_ref] );
- a->i_cost16x16bi = h->pixf.mbcmp[PIXEL_16x16]( h->mb.pic.p_fenc[0], FENC_STRIDE, pix1, 16 )
+ a->i_cost16x16bi = h->pixf.mbcmp[PIXEL_16x16]( h->mb.pic.p_fenc[0], FENC_STRIDE, pix0, 16 )
+ REF_COST( 0, a->l0.i_ref )
+ REF_COST( 1, a->l1.i_ref )
+ a->l0.me16x16.cost_mv
const int y8 = i/2;
int i_part_cost;
int i_part_cost_bi = 0;
+ int stride[2] = {8,8};
+ uint8_t *src[2];
for( l = 0; l < 2; l++ )
{
x264_macroblock_cache_mv_ptr( h, 2*x8, 2*y8, 2, 2, l, m->mv );
/* BI mode */
- h->mc.mc_luma( pix[l], 8, m->p_fref, m->i_stride[0],
- m->mv[0], m->mv[1], 8, 8 );
+ src[l] = h->mc.get_ref( pix[l], &stride[l], m->p_fref, m->i_stride[0],
+ m->mv[0], m->mv[1], 8, 8 );
i_part_cost_bi += m->cost_mv;
/* FIXME: ref cost */
}
-
- WEIGHTED_AVG( PIXEL_8x8, pix[0], 8, pix[1], 8 );
+ h->mc.avg[PIXEL_8x8]( pix[0], 8, src[0], stride[0], src[1], stride[1], h->mb.bipred_weight[a->l0.i_ref][a->l1.i_ref] );
i_part_cost_bi += h->pixf.mbcmp[PIXEL_8x8]( a->l0.me8x8[i].p_fenc[0], FENC_STRIDE, pix[0], 8 )
+ a->i_lambda * i_sub_mb_b_cost_table[D_BI_8x8];
a->l0.me8x8[i].cost += a->i_lambda * i_sub_mb_b_cost_table[D_L0_8x8];
uint8_t **p_fref[2] =
{ h->mb.pic.p_fref[0][a->l0.i_ref],
h->mb.pic.p_fref[1][a->l1.i_ref] };
- DECLARE_ALIGNED_16( uint8_t pix[2][16*8] );
+ DECLARE_ALIGNED_16( uint8_t pix[2][16*8] );
DECLARE_ALIGNED_4( int16_t mvc[2][2] );
int i, l;
{
int i_part_cost;
int i_part_cost_bi = 0;
+ int stride[2] = {16,16};
+ uint8_t *src[2];
/* TODO: check only the list(s) that were used in b8x8? */
for( l = 0; l < 2; l++ )
x264_me_search( h, m, mvc, 2 );
/* BI mode */
- h->mc.mc_luma( pix[l], 16, m->p_fref, m->i_stride[0],
- m->mv[0], m->mv[1], 16, 8 );
+ src[l] = h->mc.get_ref( pix[l], &stride[l], m->p_fref, m->i_stride[0],
+ m->mv[0], m->mv[1], 16, 8 );
/* FIXME: ref cost */
i_part_cost_bi += m->cost_mv;
}
-
- WEIGHTED_AVG( PIXEL_16x8, pix[0], 16, pix[1], 16 );
+ h->mc.avg[PIXEL_16x8]( pix[0], 16, src[0], stride[0], src[1], stride[1], h->mb.bipred_weight[a->l0.i_ref][a->l1.i_ref] );
i_part_cost_bi += h->pixf.mbcmp[PIXEL_16x8]( a->l0.me16x8[i].p_fenc[0], FENC_STRIDE, pix[0], 16 );
i_part_cost = a->l0.me16x8[i].cost;
{
int i_part_cost;
int i_part_cost_bi = 0;
+ int stride[2] = {8,8};
+ uint8_t *src[2];
for( l = 0; l < 2; l++ )
{
x264_me_search( h, m, mvc, 2 );
/* BI mode */
- h->mc.mc_luma( pix[l], 8, m->p_fref, m->i_stride[0],
- m->mv[0], m->mv[1], 8, 16 );
+ src[l] = h->mc.get_ref( pix[l], &stride[l], m->p_fref, m->i_stride[0],
+ m->mv[0], m->mv[1], 8, 16 );
/* FIXME: ref cost */
i_part_cost_bi += m->cost_mv;
}
- WEIGHTED_AVG( PIXEL_8x16, pix[0], 8, pix[1], 8 );
+ h->mc.avg[PIXEL_8x16]( pix[0], 8, src[0], stride[0], src[1], stride[1], h->mb.bipred_weight[a->l0.i_ref][a->l1.i_ref] );
i_part_cost_bi += h->pixf.mbcmp[PIXEL_8x16]( a->l0.me8x16[i].p_fenc[0], FENC_STRIDE, pix[0], 8 );
i_part_cost = a->l0.me8x16[i].cost;
{
h->mb.i_type = P_8x8;
h->mb.i_partition = D_8x8;
- x264_analyse_update_cache( h, a );
- a->l0.i_cost8x8 = x264_rd_cost_mb( h, a->i_lambda2 );
-
if( h->param.analyse.inter & X264_ANALYSE_PSUB8x8 )
{
- /* FIXME: RD per subpartition */
- int part_bak[4];
- int i, i_cost;
- int b_sub8x8 = 0;
- for( i=0; i<4; i++ )
- {
- part_bak[i] = h->mb.i_sub_partition[i];
- b_sub8x8 |= (part_bak[i] != D_L0_8x8);
- }
- if( b_sub8x8 )
+ int i;
+ x264_macroblock_cache_ref( h, 0, 0, 2, 2, 0, a->l0.me8x8[0].i_ref );
+ x264_macroblock_cache_ref( h, 2, 0, 2, 2, 0, a->l0.me8x8[1].i_ref );
+ x264_macroblock_cache_ref( h, 0, 2, 2, 2, 0, a->l0.me8x8[2].i_ref );
+ x264_macroblock_cache_ref( h, 2, 2, 2, 2, 0, a->l0.me8x8[3].i_ref );
+ for( i = 0; i < 4; i++ )
{
- h->mb.i_sub_partition[0] = h->mb.i_sub_partition[1] =
- h->mb.i_sub_partition[2] = h->mb.i_sub_partition[3] = D_L0_8x8;
- x264_analyse_update_cache( h, a );
- i_cost = x264_rd_cost_mb( h, a->i_lambda2 );
- if( a->l0.i_cost8x8 < i_cost )
+ int costs[4] = {a->l0.i_cost4x4[i], a->l0.i_cost8x4[i], a->l0.i_cost4x8[i], a->l0.me8x8[i].cost};
+ int thresh = X264_MIN4( costs[0], costs[1], costs[2], costs[3] ) * 5 / 4;
+ int subtype, btype = D_L0_8x8;
+ uint64_t bcost = COST_MAX64;
+ for( subtype = D_L0_4x4; subtype <= D_L0_8x8; subtype++ )
{
- for( i=0; i<4; i++ )
- h->mb.i_sub_partition[i] = part_bak[i];
+ uint64_t cost;
+ if( costs[subtype] > thresh || (subtype == D_L0_8x8 && bcost == COST_MAX64) )
+ continue;
+ h->mb.i_sub_partition[i] = subtype;
+ x264_mb_cache_mv_p8x8( h, a, i );
+ cost = x264_rd_cost_part( h, a->i_lambda2, i<<2, PIXEL_8x8 );
+ COPY2_IF_LT( bcost, cost, btype, subtype );
}
- else
- a->l0.i_cost8x8 = i_cost;
+ h->mb.i_sub_partition[i] = btype;
+ x264_mb_cache_mv_p8x8( h, a, i );
}
}
+ else
+ x264_analyse_update_cache( h, a );
+ a->l0.i_cost8x8 = x264_rd_cost_mb( h, a->i_lambda2 );
}
else
a->l0.i_cost8x8 = COST_MAX;
}
}
-static void refine_bidir( x264_t *h, x264_mb_analysis_t *a )
+static void x264_refine_bidir( x264_t *h, x264_mb_analysis_t *a )
{
const int i_biweight = h->mb.bipred_weight[a->l0.i_ref][a->l1.i_ref];
int i;
+ if( IS_INTRA(h->mb.i_type) )
+ return;
+
switch( h->mb.i_partition )
{
case D_16x16:
if( h->mb.i_type == B_BI_BI )
- x264_me_refine_bidir( h, &a->l0.me16x16, &a->l1.me16x16, i_biweight );
+ x264_me_refine_bidir_satd( h, &a->l0.me16x16, &a->l1.me16x16, i_biweight );
break;
case D_16x8:
for( i=0; i<2; i++ )
if( a->i_mb_partition16x8[i] == D_BI_8x8 )
- x264_me_refine_bidir( h, &a->l0.me16x8[i], &a->l1.me16x8[i], i_biweight );
+ x264_me_refine_bidir_satd( h, &a->l0.me16x8[i], &a->l1.me16x8[i], i_biweight );
break;
case D_8x16:
for( i=0; i<2; i++ )
if( a->i_mb_partition8x16[i] == D_BI_8x8 )
- x264_me_refine_bidir( h, &a->l0.me8x16[i], &a->l1.me8x16[i], i_biweight );
+ x264_me_refine_bidir_satd( h, &a->l0.me8x16[i], &a->l1.me8x16[i], i_biweight );
break;
case D_8x8:
for( i=0; i<4; i++ )
if( h->mb.i_sub_partition[i] == D_BI_8x8 )
- x264_me_refine_bidir( h, &a->l0.me8x8[i], &a->l1.me8x8[i], i_biweight );
+ x264_me_refine_bidir_satd( h, &a->l0.me8x8[i], &a->l1.me8x8[i], i_biweight );
break;
}
}
static inline void x264_mb_analyse_transform( x264_t *h )
{
- if( x264_mb_transform_8x8_allowed( h ) && h->param.analyse.b_transform_8x8 )
+ if( x264_mb_transform_8x8_allowed( h ) && h->param.analyse.b_transform_8x8 && !h->mb.b_lossless )
{
int i_cost4, i_cost8;
/* Only luma MC is really needed, but the full MC is re-used in macroblock_encode. */
/*--------------------------- Do the analysis ---------------------------*/
if( h->sh.i_type == SLICE_TYPE_I )
{
- if( analysis.b_mbrd )
+ if( analysis.i_mbrd )
x264_mb_cache_fenc_satd( h );
x264_mb_analyse_intra( h, &analysis, COST_MAX );
- if( analysis.b_mbrd )
+ if( analysis.i_mbrd )
x264_intra_rd( h, &analysis, COST_MAX );
i_cost = analysis.i_satd_i16x16;
if( analysis.i_satd_pcm < i_cost )
h->mb.i_type = I_PCM;
- else if( h->mb.i_subpel_refine >= 7 )
+ else if( analysis.i_mbrd >= 2 )
x264_intra_rd_refine( h, &analysis );
}
else if( h->sh.i_type == SLICE_TYPE_P )
/* refine qpel */
//FIXME mb_type costs?
- if( analysis.b_mbrd )
+ if( analysis.i_mbrd )
{
/* refine later */
}
analysis.i_satd_i8x8,
analysis.i_satd_i4x4 );
- if( analysis.b_mbrd )
+ if( analysis.i_mbrd )
{
x264_mb_analyse_p_rd( h, &analysis, X264_MIN(i_satd_inter, i_satd_intra) );
i_type = P_L0;
h->stat.frame.i_inter_cost += i_cost;
h->stat.frame.i_mbs_analysed++;
- if( h->mb.i_subpel_refine >= 7 && h->mb.i_type != I_PCM )
+ if( analysis.i_mbrd >= 2 && h->mb.i_type != I_PCM )
{
if( IS_INTRA( h->mb.i_type ) )
{
else if( i_partition == D_16x16 )
{
x264_macroblock_cache_ref( h, 0, 0, 4, 4, 0, analysis.l0.me16x16.i_ref );
- x264_me_refine_qpel_rd( h, &analysis.l0.me16x16, analysis.i_lambda2, 0 );
+ x264_me_refine_qpel_rd( h, &analysis.l0.me16x16, analysis.i_lambda2, 0, 0 );
}
else if( i_partition == D_16x8 )
{
h->mb.i_sub_partition[2] = h->mb.i_sub_partition[3] = D_L0_8x8;
x264_macroblock_cache_ref( h, 0, 0, 4, 2, 0, analysis.l0.me16x8[0].i_ref );
x264_macroblock_cache_ref( h, 0, 2, 4, 2, 0, analysis.l0.me16x8[1].i_ref );
- x264_me_refine_qpel_rd( h, &analysis.l0.me16x8[0], analysis.i_lambda2, 0 );
- x264_me_refine_qpel_rd( h, &analysis.l0.me16x8[1], analysis.i_lambda2, 2 );
+ x264_me_refine_qpel_rd( h, &analysis.l0.me16x8[0], analysis.i_lambda2, 0, 0 );
+ x264_me_refine_qpel_rd( h, &analysis.l0.me16x8[1], analysis.i_lambda2, 8, 0 );
}
else if( i_partition == D_8x16 )
{
h->mb.i_sub_partition[2] = h->mb.i_sub_partition[3] = D_L0_8x8;
x264_macroblock_cache_ref( h, 0, 0, 2, 4, 0, analysis.l0.me8x16[0].i_ref );
x264_macroblock_cache_ref( h, 2, 0, 2, 4, 0, analysis.l0.me8x16[1].i_ref );
- x264_me_refine_qpel_rd( h, &analysis.l0.me8x16[0], analysis.i_lambda2, 0 );
- x264_me_refine_qpel_rd( h, &analysis.l0.me8x16[1], analysis.i_lambda2, 1 );
+ x264_me_refine_qpel_rd( h, &analysis.l0.me8x16[0], analysis.i_lambda2, 0, 0 );
+ x264_me_refine_qpel_rd( h, &analysis.l0.me8x16[1], analysis.i_lambda2, 4, 0 );
}
else if( i_partition == D_8x8 )
{
int i8x8;
x264_analyse_update_cache( h, &analysis );
for( i8x8 = 0; i8x8 < 4; i8x8++ )
- if( h->mb.i_sub_partition[i8x8] == D_L0_8x8 )
- x264_me_refine_qpel_rd( h, &analysis.l0.me8x8[i8x8], analysis.i_lambda2, i8x8 );
+ {
+ if( h->mb.i_sub_partition[i8x8] == D_L0_8x8 )
+ {
+ x264_me_refine_qpel_rd( h, &analysis.l0.me8x8[i8x8], analysis.i_lambda2, i8x8*4, 0 );
+ }
+ else if( h->mb.i_sub_partition[i8x8] == D_L0_8x4 )
+ {
+ x264_me_refine_qpel_rd( h, &analysis.l0.me8x4[i8x8][0], analysis.i_lambda2, i8x8*4+0, 0 );
+ x264_me_refine_qpel_rd( h, &analysis.l0.me8x4[i8x8][1], analysis.i_lambda2, i8x8*4+2, 0 );
+ }
+ else if( h->mb.i_sub_partition[i8x8] == D_L0_4x8 )
+ {
+ x264_me_refine_qpel_rd( h, &analysis.l0.me4x8[i8x8][0], analysis.i_lambda2, i8x8*4+0, 0 );
+ x264_me_refine_qpel_rd( h, &analysis.l0.me4x8[i8x8][1], analysis.i_lambda2, i8x8*4+1, 0 );
+ }
+ else if( h->mb.i_sub_partition[i8x8] == D_L0_4x4 )
+ {
+ x264_me_refine_qpel_rd( h, &analysis.l0.me4x4[i8x8][0], analysis.i_lambda2, i8x8*4+0, 0 );
+ x264_me_refine_qpel_rd( h, &analysis.l0.me4x4[i8x8][1], analysis.i_lambda2, i8x8*4+1, 0 );
+ x264_me_refine_qpel_rd( h, &analysis.l0.me4x4[i8x8][2], analysis.i_lambda2, i8x8*4+2, 0 );
+ x264_me_refine_qpel_rd( h, &analysis.l0.me4x4[i8x8][3], analysis.i_lambda2, i8x8*4+3, 0 );
+ }
+ }
}
}
}
int i_bskip_cost = COST_MAX;
int b_skip = 0;
- if( analysis.b_mbrd )
+ if( analysis.i_mbrd )
x264_mb_cache_fenc_satd( h );
h->mb.i_type = B_SKIP;
{
/* chance of skip is too small to bother */
}
- else if( analysis.b_mbrd )
+ else if( analysis.i_mbrd )
{
i_bskip_cost = ssd_mb( h );
/* 6 = minimum cavlc cost of a non-skipped MB */
COPY2_IF_LT( i_cost, analysis.i_cost16x16bi, i_type, B_BI_BI );
COPY2_IF_LT( i_cost, analysis.i_cost16x16direct, i_type, B_DIRECT );
- if( analysis.b_mbrd && analysis.i_cost16x16direct <= i_cost * 33/32 )
+ if( analysis.i_mbrd && analysis.i_cost16x16direct <= i_cost * 33/32 )
{
x264_mb_analyse_b_rd( h, &analysis, i_cost );
if( i_bskip_cost < analysis.i_rd16x16direct &&
}
}
- if( analysis.b_mbrd )
+ if( analysis.i_mbrd )
{
/* refine later */
}
}
}
- if( analysis.b_mbrd )
+ if( analysis.i_mbrd )
{
i_satd_inter = i_cost;
x264_mb_analyse_b_rd( h, &analysis, i_satd_inter );
x264_mb_analyse_intra( h, &analysis, i_satd_inter );
- if( analysis.b_mbrd )
+ if( analysis.i_mbrd )
{
x264_mb_analyse_transform_rd( h, &analysis, &i_satd_inter, &i_cost );
x264_intra_rd( h, &analysis, i_satd_inter * 17/16 );
h->mb.i_type = i_type;
h->mb.i_partition = i_partition;
- if( analysis.b_mbrd && h->mb.i_subpel_refine >= 7 && IS_INTRA( i_type ) && i_type != I_PCM )
+ if( analysis.i_mbrd >= 2 && IS_INTRA( i_type ) && i_type != I_PCM )
x264_intra_rd_refine( h, &analysis );
- else if( h->param.analyse.b_bidir_me )
- refine_bidir( h, &analysis );
+ if( h->mb.i_subpel_refine >= 5 )
+ x264_refine_bidir( h, &analysis );
+
+ if( analysis.i_mbrd >= 2 && i_type > B_DIRECT && i_type < B_SKIP )
+ {
+ const int i_biweight = h->mb.bipred_weight[analysis.l0.i_ref][analysis.l1.i_ref];
+ x264_analyse_update_cache( h, &analysis );
+
+ if( i_partition == D_16x16 )
+ {
+ if( i_type == B_L0_L0 )
+ x264_me_refine_qpel_rd( h, &analysis.l0.me16x16, analysis.i_lambda2, 0, 0 );
+ else if( i_type == B_L1_L1 )
+ x264_me_refine_qpel_rd( h, &analysis.l1.me16x16, analysis.i_lambda2, 0, 1 );
+ else if( i_type == B_BI_BI )
+ x264_me_refine_bidir_rd( h, &analysis.l0.me16x16, &analysis.l1.me16x16, i_biweight, 0, analysis.i_lambda2 );
+ }
+ else if( i_partition == D_16x8 )
+ {
+ for( i = 0; i < 2; i++ )
+ {
+ h->mb.i_sub_partition[i*2] = h->mb.i_sub_partition[i*2+1] = analysis.i_mb_partition16x8[i];
+ if( analysis.i_mb_partition16x8[i] == D_L0_8x8 )
+ x264_me_refine_qpel_rd( h, &analysis.l0.me16x8[i], analysis.i_lambda2, i*8, 0 );
+ else if( analysis.i_mb_partition16x8[i] == D_L1_8x8 )
+ x264_me_refine_qpel_rd( h, &analysis.l1.me16x8[i], analysis.i_lambda2, i*8, 1 );
+ else if( analysis.i_mb_partition16x8[i] == D_BI_8x8 )
+ x264_me_refine_bidir_rd( h, &analysis.l0.me16x8[i], &analysis.l1.me16x8[i], i_biweight, i*2, analysis.i_lambda2 );
+ }
+ }
+ else if( i_partition == D_8x16 )
+ {
+ for( i = 0; i < 2; i++ )
+ {
+ h->mb.i_sub_partition[i] = h->mb.i_sub_partition[i+2] = analysis.i_mb_partition8x16[i];
+ if( analysis.i_mb_partition8x16[i] == D_L0_8x8 )
+ x264_me_refine_qpel_rd( h, &analysis.l0.me8x16[i], analysis.i_lambda2, i*4, 0 );
+ else if( analysis.i_mb_partition8x16[i] == D_L1_8x8 )
+ x264_me_refine_qpel_rd( h, &analysis.l1.me8x16[i], analysis.i_lambda2, i*4, 1 );
+ else if( analysis.i_mb_partition8x16[i] == D_BI_8x8 )
+ x264_me_refine_bidir_rd( h, &analysis.l0.me8x16[i], &analysis.l1.me8x16[i], i_biweight, i, analysis.i_lambda2 );
+ }
+ }
+ else if( i_partition == D_8x8 )
+ {
+ for( i = 0; i < 4; i++ )
+ {
+ if( h->mb.i_sub_partition[i] == D_L0_8x8 )
+ x264_me_refine_qpel_rd( h, &analysis.l0.me8x8[i], analysis.i_lambda2, i*4, 0 );
+ else if( h->mb.i_sub_partition[i] == D_L1_8x8 )
+ x264_me_refine_qpel_rd( h, &analysis.l1.me8x8[i], analysis.i_lambda2, i*4, 1 );
+ else if( h->mb.i_sub_partition[i] == D_BI_8x8 )
+ x264_me_refine_bidir_rd( h, &analysis.l0.me8x8[i], &analysis.l1.me8x8[i], i_biweight, i, analysis.i_lambda2 );
+ }
+ }
+ }
}
}
x264_analyse_update_cache( h, &analysis );
- if( !analysis.b_mbrd )
+ if( !analysis.i_mbrd )
x264_mb_analyse_transform( h );
h->mb.b_trellis = h->param.analyse.i_trellis;