* Foundation, Inc., 59 Temple Place - Suite 330, Boston, MA 02111, USA.
*****************************************************************************/
-#include <stdio.h>
-#include <string.h>
-
#include "common.h"
int x264_mb_predict_intra4x4_mode( x264_t *h, int idx )
int x264_mb_transform_8x8_allowed( x264_t *h )
{
- if( IS_SKIP( h->mb.i_type ) )
+ // intra and skip are disallowed
+ // large partitions are allowed
+ // direct and 8x8 are conditional
+ static const uint8_t partition_tab[X264_MBTYPE_MAX] = {
+ 0,0,0,0,1,2,0,2,1,1,1,1,1,1,1,1,1,2,0,
+ };
+ int p, i;
+
+ if( !h->pps->b_transform_8x8_mode )
return 0;
- if( h->mb.i_type == P_8x8 || h->mb.i_type == B_8x8 )
+ p = partition_tab[h->mb.i_type];
+ if( p < 2 )
+ return p;
+ else if( h->mb.i_type == B_DIRECT )
+ return h->sps->b_direct8x8_inference;
+ else if( h->mb.i_type == P_8x8 )
{
- int i;
- for( i = 0; i < 4; i++ )
- if( !IS_SUB8x8(h->mb.i_sub_partition[i])
- || ( h->mb.i_sub_partition[i] == D_DIRECT_8x8 && !h->sps->b_direct8x8_inference ) )
- {
+ if( !(h->param.analyse.inter & X264_ANALYSE_PSUB8x8) )
+ return 1;
+ for( i=0; i<4; i++ )
+ if( h->mb.i_sub_partition[i] != D_L0_8x8 )
return 0;
- }
+ return 1;
+ }
+ else // B_8x8
+ {
+ // x264 currently doesn't use sub-8x8 B partitions, so don't check for them
+ if( h->sps->b_direct8x8_inference )
+ return 1;
+ for( i=0; i<4; i++ )
+ if( h->mb.i_sub_partition[i] == D_DIRECT_8x8 )
+ return 0;
+ return 1;
}
- if( h->mb.i_type == B_DIRECT && !h->sps->b_direct8x8_inference )
- return 0;
-
- return 1;
}
void x264_mb_predict_mv( x264_t *h, int i_list, int idx, int i_width, int mvp[2] )
}
else
{
- /* the colocated ref isn't in the current list0 */
+ /* the collocated ref isn't in the current list0 */
/* FIXME: we might still be able to use direct_8x8 on some partitions */
/* FIXME: with B-pyramid + extensive ref list reordering
* (not currently used), we would also have to check
}
}
+ if( h->param.i_threads > 1 )
+ {
+ int di = b8x8 ? 4 : 1;
+ for( i4=0; i4<16; i4+=di )
+ {
+ if( h->mb.cache.mv[0][x264_scan8[i4]][1] > h->mb.mv_max_spel[1]
+ || h->mb.cache.mv[1][x264_scan8[i4]][1] > h->mb.mv_max_spel[1] )
+ {
+#if 0
+ fprintf(stderr, "direct_temporal: (%d,%d) (%d,%d) > %d \n",
+ h->mb.cache.mv[0][x264_scan8[i4]][0],
+ h->mb.cache.mv[0][x264_scan8[i4]][1],
+ h->mb.cache.mv[1][x264_scan8[i4]][0],
+ h->mb.cache.mv[1][x264_scan8[i4]][1],
+ h->mb.mv_max_spel[1]);
+#endif
+ return 0;
+ }
+ }
+ }
+
return 1;
}
if( IS_INTRA( type_col ) )
return 1;
+
+ if( h->param.i_threads > 1
+ && ( mv[0][1] > h->mb.mv_max_spel[1]
+ || mv[1][1] > h->mb.mv_max_spel[1] ) )
+ {
+#if 0
+ fprintf(stderr, "direct_spatial: (%d,%d) (%d,%d) > %d \n",
+ mv[0][0], mv[0][1], mv[1][0], mv[1][1],
+ h->mb.mv_max_spel[1]);
+#endif
+ return 0;
+ }
+
b8x8 = h->sps->b_direct8x8_inference ||
(type_col != P_8x8 && type_col != B_SKIP && type_col != B_DIRECT && type_col != B_8x8);
for( l = 0; l < 2; l++ )
for( i = 0; i < 4; i++ )
h->mb.cache.direct_ref[l][i] = h->mb.cache.ref[l][x264_scan8[i*4]];
- memcpy(h->mb.cache.direct_mv, h->mb.cache.mv, sizeof(h->mb.cache.mv));
+ h->mc.memcpy_aligned(h->mb.cache.direct_mv, h->mb.cache.mv, sizeof(h->mb.cache.mv));
}
return b_available;
}
}
+#define FIXED_SCALE 256
+
/* This just improves encoder performance, it's not part of the spec */
void x264_mb_predict_mv_ref16x16( x264_t *h, int i_list, int i_ref, int mvc[8][2], int *i_mvc )
{
}
if( h->mb.i_neighbour & MB_TOP )
{
- int i_mb_t = h->mb.i_mb_xy - h->mb.i_mb_stride;
+ int i_mb_t = h->mb.i_mb_top_xy;
if( !IS_SKIP( h->mb.type[i_mb_t] ) )
SET_MVP( mvr[i_mb_t] );
#undef SET_MVP
/* temporal predictors */
- if( h->fref0[0]->i_ref[0] > 0 )
+ /* FIXME temporal scaling w/ interlace */
+ if( h->fref0[0]->i_ref[0] > 0 && !h->sh.b_mbaff )
{
x264_frame_t *l0 = h->fref0[0];
- int ref_col_cur, ref_col_prev = -1;
- int scale = 0;
#define SET_TMVP(dx, dy) { \
int i_b4 = h->mb.i_b4_xy + dx*4 + dy*4*h->mb.i_b4_stride; \
int i_b8 = h->mb.i_b8_xy + dx*2 + dy*2*h->mb.i_b8_stride; \
- ref_col_cur = l0->ref[0][i_b8]; \
- if( ref_col_cur >= 0 ) \
+ int ref_col = l0->ref[0][i_b8]; \
+ if( ref_col >= 0 ) \
{ \
- /* TODO: calc once per frame and tablize? */\
- if( ref_col_cur != ref_col_prev ) \
- scale = 256 * (h->fenc->i_poc - h->fref0[i_ref]->i_poc) \
- / (l0->i_poc - l0->ref_poc[0][ref_col_cur]); \
- mvc[i][0] = l0->mv[0][i_b4][0] * scale / 256; \
- mvc[i][1] = l0->mv[0][i_b4][1] * scale / 256; \
+ int scale = (h->fdec->i_poc - h->fdec->ref_poc[0][i_ref]) * l0->inv_ref_poc[ref_col];\
+ mvc[i][0] = l0->mv[0][i_b4][0] * scale / FIXED_SCALE; \
+ mvc[i][1] = l0->mv[0][i_b4][1] * scale / FIXED_SCALE; \
i++; \
- ref_col_prev = ref_col_cur; \
} \
}
*i_mvc = i;
}
+/* Set up a lookup table for delta pocs to reduce an IDIV to an IMUL */
+static void setup_inverse_delta_pocs( x264_t *h )
+{
+ int i;
+ for( i = 0; i < h->i_ref0; i++ )
+ {
+ int delta = h->fdec->i_poc - h->fref0[i]->i_poc;
+ h->fdec->inv_ref_poc[i] = (FIXED_SCALE + delta/2) / delta;
+ }
+}
+
static inline void x264_mb_mc_0xywh( x264_t *h, int x, int y, int width, int height )
{
const int i8 = x264_scan8[0]+x+8*y;
const int i_ref = h->mb.cache.ref[0][i8];
const int mvx = x264_clip3( h->mb.cache.mv[0][i8][0], h->mb.mv_min[0], h->mb.mv_max[0] );
- const int mvy = x264_clip3( h->mb.cache.mv[0][i8][1], h->mb.mv_min[1], h->mb.mv_max[1] );
+ int mvy = x264_clip3( h->mb.cache.mv[0][i8][1], h->mb.mv_min[1], h->mb.mv_max[1] );
+
+ h->mc.mc_luma( &h->mb.pic.p_fdec[0][4*y*FDEC_STRIDE+4*x], FDEC_STRIDE,
+ h->mb.pic.p_fref[0][i_ref], h->mb.pic.i_stride[0],
+ mvx + 4*4*x, mvy + 4*4*y, 4*width, 4*height );
- h->mc.mc_luma( h->mb.pic.p_fref[0][i_ref], h->mb.pic.i_stride[0],
- &h->mb.pic.p_fdec[0][4*y*FDEC_STRIDE+4*x], FDEC_STRIDE,
- mvx + 4*4*x, mvy + 4*4*y, 4*width, 4*height );
+ // chroma is offset if MCing from a field of opposite parity
+ if( h->mb.b_interlaced & i_ref )
+ mvy += (h->mb.i_mb_y & 1)*4 - 2;
- h->mc.mc_chroma( &h->mb.pic.p_fref[0][i_ref][4][2*y*h->mb.pic.i_stride[1]+2*x], h->mb.pic.i_stride[1],
- &h->mb.pic.p_fdec[1][2*y*FDEC_STRIDE+2*x], FDEC_STRIDE,
- mvx, mvy, 2*width, 2*height );
+ h->mc.mc_chroma( &h->mb.pic.p_fdec[1][2*y*FDEC_STRIDE+2*x], FDEC_STRIDE,
+ &h->mb.pic.p_fref[0][i_ref][4][2*y*h->mb.pic.i_stride[1]+2*x], h->mb.pic.i_stride[1],
+ mvx, mvy, 2*width, 2*height );
- h->mc.mc_chroma( &h->mb.pic.p_fref[0][i_ref][5][2*y*h->mb.pic.i_stride[2]+2*x], h->mb.pic.i_stride[2],
- &h->mb.pic.p_fdec[2][2*y*FDEC_STRIDE+2*x], FDEC_STRIDE,
- mvx, mvy, 2*width, 2*height );
+ h->mc.mc_chroma( &h->mb.pic.p_fdec[2][2*y*FDEC_STRIDE+2*x], FDEC_STRIDE,
+ &h->mb.pic.p_fref[0][i_ref][5][2*y*h->mb.pic.i_stride[2]+2*x], h->mb.pic.i_stride[2],
+ mvx, mvy, 2*width, 2*height );
}
static inline void x264_mb_mc_1xywh( x264_t *h, int x, int y, int width, int height )
{
const int i8 = x264_scan8[0]+x+8*y;
const int i_ref = h->mb.cache.ref[1][i8];
const int mvx = x264_clip3( h->mb.cache.mv[1][i8][0], h->mb.mv_min[0], h->mb.mv_max[0] );
- const int mvy = x264_clip3( h->mb.cache.mv[1][i8][1], h->mb.mv_min[1], h->mb.mv_max[1] );
+ int mvy = x264_clip3( h->mb.cache.mv[1][i8][1], h->mb.mv_min[1], h->mb.mv_max[1] );
- h->mc.mc_luma( h->mb.pic.p_fref[1][i_ref], h->mb.pic.i_stride[0],
- &h->mb.pic.p_fdec[0][4*y*FDEC_STRIDE+4*x], FDEC_STRIDE,
- mvx + 4*4*x, mvy + 4*4*y, 4*width, 4*height );
+ h->mc.mc_luma( &h->mb.pic.p_fdec[0][4*y*FDEC_STRIDE+4*x], FDEC_STRIDE,
+ h->mb.pic.p_fref[1][i_ref], h->mb.pic.i_stride[0],
+ mvx + 4*4*x, mvy + 4*4*y, 4*width, 4*height );
- h->mc.mc_chroma( &h->mb.pic.p_fref[1][i_ref][4][2*y*h->mb.pic.i_stride[1]+2*x], h->mb.pic.i_stride[1],
- &h->mb.pic.p_fdec[1][2*y*FDEC_STRIDE+2*x], FDEC_STRIDE,
- mvx, mvy, 2*width, 2*height );
+ if( h->mb.b_interlaced & i_ref )
+ mvy += (h->mb.i_mb_y & 1)*4 - 2;
- h->mc.mc_chroma( &h->mb.pic.p_fref[1][i_ref][5][2*y*h->mb.pic.i_stride[2]+2*x], h->mb.pic.i_stride[2],
- &h->mb.pic.p_fdec[2][2*y*FDEC_STRIDE+2*x], FDEC_STRIDE,
- mvx, mvy, 2*width, 2*height );
+ h->mc.mc_chroma( &h->mb.pic.p_fdec[1][2*y*FDEC_STRIDE+2*x], FDEC_STRIDE,
+ &h->mb.pic.p_fref[1][i_ref][4][2*y*h->mb.pic.i_stride[1]+2*x], h->mb.pic.i_stride[1],
+ mvx, mvy, 2*width, 2*height );
+
+ h->mc.mc_chroma( &h->mb.pic.p_fdec[2][2*y*FDEC_STRIDE+2*x], FDEC_STRIDE,
+ &h->mb.pic.p_fref[1][i_ref][5][2*y*h->mb.pic.i_stride[2]+2*x], h->mb.pic.i_stride[2],
+ mvx, mvy, 2*width, 2*height );
}
static inline void x264_mb_mc_01xywh( x264_t *h, int x, int y, int width, int height )
const int i_ref1 = h->mb.cache.ref[1][i8];
const int mvx1 = x264_clip3( h->mb.cache.mv[1][i8][0], h->mb.mv_min[0], h->mb.mv_max[0] );
- const int mvy1 = x264_clip3( h->mb.cache.mv[1][i8][1], h->mb.mv_min[1], h->mb.mv_max[1] );
- DECLARE_ALIGNED( uint8_t, tmp[16*16], 16 );
+ int mvy1 = x264_clip3( h->mb.cache.mv[1][i8][1], h->mb.mv_min[1], h->mb.mv_max[1] );
+ DECLARE_ALIGNED_16( uint8_t tmp[16*16] );
int i_mode = x264_size2pixel[height][width];
x264_mb_mc_0xywh( h, x, y, width, height );
- h->mc.mc_luma( h->mb.pic.p_fref[1][i_ref1], h->mb.pic.i_stride[0],
- tmp, 16, mvx1 + 4*4*x, mvy1 + 4*4*y, 4*width, 4*height );
+ h->mc.mc_luma( tmp, 16, h->mb.pic.p_fref[1][i_ref1], h->mb.pic.i_stride[0],
+ mvx1 + 4*4*x, mvy1 + 4*4*y, 4*width, 4*height );
+
+ if( h->mb.b_interlaced & i_ref1 )
+ mvy1 += (h->mb.i_mb_y & 1)*4 - 2;
if( h->param.analyse.b_weighted_bipred )
{
h->mc.avg_weight[i_mode]( &h->mb.pic.p_fdec[0][4*y*FDEC_STRIDE+4*x], FDEC_STRIDE, tmp, 16, weight );
- h->mc.mc_chroma( &h->mb.pic.p_fref[1][i_ref1][4][2*y*h->mb.pic.i_stride[1]+2*x], h->mb.pic.i_stride[1],
- tmp, 16, mvx1, mvy1, 2*width, 2*height );
+ h->mc.mc_chroma( tmp, 16, &h->mb.pic.p_fref[1][i_ref1][4][2*y*h->mb.pic.i_stride[1]+2*x], h->mb.pic.i_stride[1],
+ mvx1, mvy1, 2*width, 2*height );
h->mc.avg_weight[i_mode+3]( &h->mb.pic.p_fdec[1][2*y*FDEC_STRIDE+2*x], FDEC_STRIDE, tmp, 16, weight );
- h->mc.mc_chroma( &h->mb.pic.p_fref[1][i_ref1][5][2*y*h->mb.pic.i_stride[2]+2*x], h->mb.pic.i_stride[2],
- tmp, 16, mvx1, mvy1, 2*width, 2*height );
+ h->mc.mc_chroma( tmp, 16, &h->mb.pic.p_fref[1][i_ref1][5][2*y*h->mb.pic.i_stride[2]+2*x], h->mb.pic.i_stride[2],
+ mvx1, mvy1, 2*width, 2*height );
h->mc.avg_weight[i_mode+3]( &h->mb.pic.p_fdec[2][2*y*FDEC_STRIDE+2*x], FDEC_STRIDE, tmp, 16, weight );
}
else
{
h->mc.avg[i_mode]( &h->mb.pic.p_fdec[0][4*y*FDEC_STRIDE+4*x], FDEC_STRIDE, tmp, 16 );
- h->mc.mc_chroma( &h->mb.pic.p_fref[1][i_ref1][4][2*y*h->mb.pic.i_stride[1]+2*x], h->mb.pic.i_stride[1],
- tmp, 16, mvx1, mvy1, 2*width, 2*height );
+ h->mc.mc_chroma( tmp, 16, &h->mb.pic.p_fref[1][i_ref1][4][2*y*h->mb.pic.i_stride[1]+2*x], h->mb.pic.i_stride[1],
+ mvx1, mvy1, 2*width, 2*height );
h->mc.avg[i_mode+3]( &h->mb.pic.p_fdec[1][2*y*FDEC_STRIDE+2*x], FDEC_STRIDE, tmp, 16 );
- h->mc.mc_chroma( &h->mb.pic.p_fref[1][i_ref1][5][2*y*h->mb.pic.i_stride[2]+2*x], h->mb.pic.i_stride[2],
- tmp, 16, mvx1, mvy1, 2*width, 2*height );
+ h->mc.mc_chroma( tmp, 16, &h->mb.pic.p_fref[1][i_ref1][5][2*y*h->mb.pic.i_stride[2]+2*x], h->mb.pic.i_stride[2],
+ mvx1, mvy1, 2*width, 2*height );
h->mc.avg[i_mode+3]( &h->mb.pic.p_fdec[2][2*y*FDEC_STRIDE+2*x], FDEC_STRIDE, tmp, 16 );
}
}
h->mb.i_b8_stride = h->sps->i_mb_width * 2;
h->mb.i_b4_stride = h->sps->i_mb_width * 4;
+ h->mb.b_interlaced = h->param.b_interlaced;
+
CHECKED_MALLOC( h->mb.qp, i_mb_count * sizeof(int8_t) );
CHECKED_MALLOC( h->mb.cbp, i_mb_count * sizeof(int16_t) );
CHECKED_MALLOC( h->mb.skipbp, i_mb_count * sizeof(int8_t) );
/* all coeffs */
CHECKED_MALLOC( h->mb.non_zero_count, i_mb_count * 24 * sizeof(uint8_t) );
+ CHECKED_MALLOC( h->mb.nnz_backup, h->sps->i_mb_width * 4 * 16 * sizeof(uint8_t) );
if( h->param.b_cabac )
{
for( i=0; i<2; i++ )
{
- int i_refs = (i ? 1 : h->param.i_frame_reference) + h->param.b_bframe_pyramid;
- for( j=0; j < i_refs && j < 16; j++ )
+ int i_refs = X264_MIN(16, (i ? 1 : h->param.i_frame_reference) + h->param.b_bframe_pyramid) << h->param.b_interlaced;
+ for( j=0; j < i_refs; j++ )
CHECKED_MALLOC( h->mb.mvr[i][j], 2 * i_mb_count * sizeof(int16_t) );
}
- /* init with not avaiable (for top right idx=7,15) */
+ for( i=0; i<=h->param.b_interlaced; i++ )
+ for( j=0; j<3; j++ )
+ {
+ CHECKED_MALLOC( h->mb.intra_border_backup[i][j], h->fdec->i_stride[j] );
+ h->mb.intra_border_backup[i][j] += 8;
+ }
+
+ /* init with not available (for top right idx=7,15) */
memset( h->mb.cache.ref[0], -2, X264_SCAN8_SIZE * sizeof( int8_t ) );
memset( h->mb.cache.ref[1], -2, X264_SCAN8_SIZE * sizeof( int8_t ) );
void x264_macroblock_cache_end( x264_t *h )
{
int i, j;
+ for( i=0; i<=h->param.b_interlaced; i++ )
+ for( j=0; j<3; j++ )
+ x264_free( h->mb.intra_border_backup[i][j] - 8 );
for( i=0; i<2; i++ )
- {
- int i_refs = i ? 1 + h->param.b_bframe_pyramid : h->param.i_frame_reference;
- for( j=0; j < i_refs; j++ )
+ for( j=0; j<32; j++ )
x264_free( h->mb.mvr[i][j] );
- }
if( h->param.b_cabac )
{
x264_free( h->mb.chroma_pred_mode );
}
x264_free( h->mb.intra4x4_pred_mode );
x264_free( h->mb.non_zero_count );
+ x264_free( h->mb.nnz_backup );
x264_free( h->mb.mb_transform_size );
x264_free( h->mb.skipbp );
x264_free( h->mb.cbp );
}
if( h->sh.i_type == SLICE_TYPE_P )
memset( h->mb.cache.skip, 0, X264_SCAN8_SIZE * sizeof( int8_t ) );
+
+ setup_inverse_delta_pocs( h );
}
+void x264_prefetch_fenc( x264_t *h, x264_frame_t *fenc, int i_mb_x, int i_mb_y )
+{
+ int stride_y = fenc->i_stride[0];
+ int stride_uv = fenc->i_stride[1];
+ int off_y = 16 * (i_mb_x + i_mb_y * stride_y);
+ int off_uv = 8 * (i_mb_x + i_mb_y * stride_uv);
+ h->mc.prefetch_fenc( fenc->plane[0]+off_y, stride_y,
+ fenc->plane[1+(i_mb_x&1)]+off_uv, stride_uv, i_mb_x );
+}
-void x264_macroblock_cache_load( x264_t *h, int i_mb_x, int i_mb_y )
+static NOINLINE void copy_column8( uint8_t *dst, uint8_t *src )
{
- const int i_mb_4x4 = 4*(i_mb_y * h->mb.i_b4_stride + i_mb_x);
- const int i_mb_8x8 = 2*(i_mb_y * h->mb.i_b8_stride + i_mb_x);
+ int i;
+ for(i=0; i<8; i++)
+ dst[i*FDEC_STRIDE] = src[i*FDEC_STRIDE];
+}
+void x264_macroblock_cache_load( x264_t *h, int i_mb_x, int i_mb_y )
+{
int i_mb_xy = i_mb_y * h->mb.i_mb_stride + i_mb_x;
- int i_top_xy = i_mb_xy - h->mb.i_mb_stride;
+ int i_mb_4x4 = 4*(i_mb_y * h->mb.i_b4_stride + i_mb_x);
+ int i_mb_8x8 = 2*(i_mb_y * h->mb.i_b8_stride + i_mb_x);
+ int i_top_y = i_mb_y - (1 << h->mb.b_interlaced);
+ int i_top_xy = i_top_y * h->mb.i_mb_stride + i_mb_x;
+ int i_top_4x4 = (4*i_top_y+3) * h->mb.i_b4_stride + 4*i_mb_x;
+ int i_top_8x8 = (2*i_top_y+1) * h->mb.i_b8_stride + 2*i_mb_x;
int i_left_xy = -1;
int i_top_type = -1; /* gcc warn */
int i_left_type= -1;
int i;
+ assert( h->mb.i_b8_stride == 2*h->mb.i_mb_stride );
+ assert( h->mb.i_b4_stride == 4*h->mb.i_mb_stride );
+
/* init index */
h->mb.i_mb_x = i_mb_x;
h->mb.i_mb_y = i_mb_y;
h->mb.i_mb_xy = i_mb_xy;
h->mb.i_b8_xy = i_mb_8x8;
h->mb.i_b4_xy = i_mb_4x4;
+ h->mb.i_mb_top_xy = i_top_xy;
h->mb.i_neighbour = 0;
- /* fdec: fenc:
- * yyyyyyy
- * yYYYY YYYY
- * yYYYY YYYY
- * yYYYY YYYY
- * yYYYY YYYY
- * uuu vvv UUVV
- * uUU vVV UUVV
- * uUU vVV
- */
- h->mb.pic.p_fenc[0] = h->mb.pic.fenc_buf;
- h->mb.pic.p_fenc[1] = h->mb.pic.fenc_buf + 16*FENC_STRIDE;
- h->mb.pic.p_fenc[2] = h->mb.pic.fenc_buf + 16*FENC_STRIDE + 8;
- h->mb.pic.p_fdec[0] = h->mb.pic.fdec_buf + 2*FDEC_STRIDE;
- h->mb.pic.p_fdec[1] = h->mb.pic.fdec_buf + 19*FDEC_STRIDE;
- h->mb.pic.p_fdec[2] = h->mb.pic.fdec_buf + 19*FDEC_STRIDE + 16;
-
- /* load picture pointers */
- for( i = 0; i < 3; i++ )
- {
- const int w = (i == 0 ? 16 : 8);
- const int i_stride = h->fdec->i_stride[i];
- const uint8_t *plane_fdec = &h->fdec->plane[i][ w * (i_mb_x + i_mb_y * i_stride) ];
- int j;
-
- h->mb.pic.i_stride[i] = i_stride;
-
- h->mc.copy[i?PIXEL_8x8:PIXEL_16x16]( h->mb.pic.p_fenc[i], FENC_STRIDE,
- &h->fenc->plane[i][ w * (i_mb_x + i_mb_y * i_stride) ], i_stride, w );
- memcpy( &h->mb.pic.p_fdec[i][-1-FDEC_STRIDE], &plane_fdec[-1-i_stride], w*3/2+1 );
- for( j = 0; j < w; j++ )
- h->mb.pic.p_fdec[i][-1+j*FDEC_STRIDE] = plane_fdec[-1+j*i_stride];
-
- for( j = 0; j < h->i_ref0; j++ )
- {
- h->mb.pic.p_fref[0][j][i==0 ? 0:i+3] = &h->fref0[j]->plane[i][ w * ( i_mb_x + i_mb_y * i_stride )];
- h->mb.pic.p_fref[0][j][i+1] = &h->fref0[j]->filtered[i+1][ 16 * ( i_mb_x + i_mb_y * h->fdec->i_stride[0] )];
- }
- for( j = 0; j < h->i_ref1; j++ )
- {
- h->mb.pic.p_fref[1][j][i==0 ? 0:i+3] = &h->fref1[j]->plane[i][ w * ( i_mb_x + i_mb_y * i_stride )];
- h->mb.pic.p_fref[1][j][i+1] = &h->fref1[j]->filtered[i+1][ 16 * ( i_mb_x + i_mb_y * h->fdec->i_stride[0] )];
- }
- }
-
- if( h->fdec->integral )
- {
- for( i = 0; i < h->i_ref0; i++ )
- h->mb.pic.p_integral[0][i] = &h->fref0[i]->integral[ 16 * ( i_mb_x + i_mb_y * h->fdec->i_stride[0] )];
- for( i = 0; i < h->i_ref1; i++ )
- h->mb.pic.p_integral[1][i] = &h->fref1[i]->integral[ 16 * ( i_mb_x + i_mb_y * h->fdec->i_stride[0] )];
- }
-
/* load cache */
- if( i_mb_xy >= h->sh.i_first_mb + h->mb.i_mb_stride )
+ if( i_top_xy >= h->sh.i_first_mb )
{
h->mb.i_mb_type_top =
i_top_type= h->mb.type[i_top_xy];
+ ( i_top_type >= 0 && h->mb.mb_transform_size[i_top_xy] );
}
+ if( h->sh.b_mbaff )
+ {
+ h->mb.pic.i_fref[0] = h->i_ref0 << h->mb.b_interlaced;
+ h->mb.pic.i_fref[1] = h->i_ref1 << h->mb.b_interlaced;
+ h->mb.cache.i_neighbour_interlaced =
+ !!(h->mb.i_neighbour & MB_LEFT)
+ + !!(h->mb.i_neighbour & MB_TOP);
+ }
+
+ /* fdec: fenc:
+ * yyyyyyy
+ * yYYYY YYYY
+ * yYYYY YYYY
+ * yYYYY YYYY
+ * yYYYY YYYY
+ * uuu vvv UUVV
+ * uUU vVV UUVV
+ * uUU vVV
+ */
+ h->mb.pic.p_fenc[0] = h->mb.pic.fenc_buf;
+ h->mb.pic.p_fenc[1] = h->mb.pic.fenc_buf + 16*FENC_STRIDE;
+ h->mb.pic.p_fenc[2] = h->mb.pic.fenc_buf + 16*FENC_STRIDE + 8;
+ h->mb.pic.p_fdec[0] = h->mb.pic.fdec_buf + 2*FDEC_STRIDE;
+ h->mb.pic.p_fdec[1] = h->mb.pic.fdec_buf + 19*FDEC_STRIDE;
+ h->mb.pic.p_fdec[2] = h->mb.pic.fdec_buf + 19*FDEC_STRIDE + 16;
+
+ if( !h->mb.b_interlaced )
+ {
+ copy_column8( h->mb.pic.p_fdec[0]-1, h->mb.pic.p_fdec[0]+15 );
+ copy_column8( h->mb.pic.p_fdec[0]-1+8*FDEC_STRIDE, h->mb.pic.p_fdec[0]+15+8*FDEC_STRIDE );
+ copy_column8( h->mb.pic.p_fdec[1]-1, h->mb.pic.p_fdec[1]+7 );
+ copy_column8( h->mb.pic.p_fdec[2]-1, h->mb.pic.p_fdec[2]+7 );
+ }
+
+ /* load picture pointers */
+ for( i = 0; i < 3; i++ )
+ {
+ const int w = (i == 0 ? 16 : 8);
+ const int i_stride = h->fdec->i_stride[i];
+ const int i_stride2 = i_stride << h->mb.b_interlaced;
+ const int i_pix_offset = h->mb.b_interlaced
+ ? w * (i_mb_x + (i_mb_y&~1) * i_stride) + (i_mb_y&1) * i_stride
+ : w * (i_mb_x + i_mb_y * i_stride);
+ int ref_pix_offset[2] = { i_pix_offset, i_pix_offset };
+ const uint8_t *intra_fdec = &h->mb.intra_border_backup[i_mb_y & h->sh.b_mbaff][i][i_mb_x*16>>!!i];
+ x264_frame_t **fref[2] = { h->fref0, h->fref1 };
+ int j, k, l;
+
+ if( h->mb.b_interlaced )
+ ref_pix_offset[1] += (1-2*(i_mb_y&1)) * i_stride;
+
+ h->mb.pic.i_stride[i] = i_stride2;
+
+ h->mc.copy[i?PIXEL_8x8:PIXEL_16x16]( h->mb.pic.p_fenc[i], FENC_STRIDE,
+ &h->fenc->plane[i][i_pix_offset], i_stride2, w );
+ memcpy( &h->mb.pic.p_fdec[i][-1-FDEC_STRIDE], intra_fdec-1, w*3/2+1 );
+ if( h->mb.b_interlaced )
+ {
+ const uint8_t *plane_fdec = &h->fdec->plane[i][i_pix_offset];
+ for( j = 0; j < w; j++ )
+ h->mb.pic.p_fdec[i][-1+j*FDEC_STRIDE] = plane_fdec[-1+j*i_stride2];
+ }
+
+ for( l=0; l<2; l++ )
+ {
+ for( j=0; j<h->mb.pic.i_fref[l]; j++ )
+ {
+ h->mb.pic.p_fref[l][j][i==0 ? 0:i+3] = &fref[l][j >> h->mb.b_interlaced]->plane[i][ref_pix_offset[j&1]];
+ if( i == 0 )
+ for( k = 1; k < 4; k++ )
+ h->mb.pic.p_fref[l][j][k] = &fref[l][j >> h->mb.b_interlaced]->filtered[k][ref_pix_offset[j&1]];
+ }
+ }
+ }
+
+ if( h->fdec->integral )
+ {
+ assert( !h->mb.b_interlaced );
+ for( i = 0; i < h->mb.pic.i_fref[0]; i++ )
+ h->mb.pic.p_integral[0][i] = &h->fref0[i]->integral[ 16 * ( i_mb_x + i_mb_y * h->fdec->i_stride[0] )];
+ for( i = 0; i < h->mb.pic.i_fref[1]; i++ )
+ h->mb.pic.p_integral[1][i] = &h->fref1[i]->integral[ 16 * ( i_mb_x + i_mb_y * h->fdec->i_stride[0] )];
+ }
+
+ x264_prefetch_fenc( h, h->fenc, i_mb_x, i_mb_y );
+
/* load ref/mv/mvd */
if( h->sh.i_type != SLICE_TYPE_I )
{
if( h->mb.i_neighbour & MB_TOPLEFT )
{
const int i8 = x264_scan8[0] - 1 - 1*8;
- const int ir = i_mb_8x8 - s8x8 - 1;
- const int iv = i_mb_4x4 - s4x4 - 1;
+ const int ir = i_top_8x8 - 1;
+ const int iv = i_top_4x4 - 1;
h->mb.cache.ref[i_list][i8] = h->mb.ref[i_list][ir];
h->mb.cache.mv[i_list][i8][0] = h->mb.mv[i_list][iv][0];
h->mb.cache.mv[i_list][i8][1] = h->mb.mv[i_list][iv][1];
if( h->mb.i_neighbour & MB_TOP )
{
const int i8 = x264_scan8[0] - 8;
- const int ir = i_mb_8x8 - s8x8;
- const int iv = i_mb_4x4 - s4x4;
-
+ const int ir = i_top_8x8;
+ const int iv = i_top_4x4;
h->mb.cache.ref[i_list][i8+0] =
h->mb.cache.ref[i_list][i8+1] = h->mb.ref[i_list][ir + 0];
h->mb.cache.ref[i_list][i8+2] =
if( h->mb.i_neighbour & MB_TOPRIGHT )
{
const int i8 = x264_scan8[0] + 4 - 1*8;
- const int ir = i_mb_8x8 - s8x8 + 2;
- const int iv = i_mb_4x4 - s4x4 + 4;
-
+ const int ir = i_top_8x8 + 2;
+ const int iv = i_top_4x4 + 4;
h->mb.cache.ref[i_list][i8] = h->mb.ref[i_list][ir];
h->mb.cache.mv[i_list][i8][0] = h->mb.mv[i_list][iv][0];
h->mb.cache.mv[i_list][i8][1] = h->mb.mv[i_list][iv][1];
const int i8 = x264_scan8[0] - 1;
const int ir = i_mb_8x8 - 1;
const int iv = i_mb_4x4 - 1;
-
h->mb.cache.ref[i_list][i8+0*8] =
h->mb.cache.ref[i_list][i8+1*8] = h->mb.ref[i_list][ir + 0*s8x8];
h->mb.cache.ref[i_list][i8+2*8] =
if( i_top_type >= 0 )
{
const int i8 = x264_scan8[0] - 8;
- const int iv = i_mb_4x4 - s4x4;
+ const int iv = i_top_4x4;
for( i = 0; i < 4; i++ )
{
h->mb.cache.mvd[i_list][i8+i][0] = h->mb.mvd[i_list][iv + i][0];
h->mb.cache.skip[x264_scan8[4] - 8] = h->mb.skipbp[i_top_xy] & 0x8;
}
}
+
+ if( h->sh.i_type == SLICE_TYPE_P )
+ x264_mb_predict_mv_pskip( h, h->mb.cache.pskip_mv );
}
h->mb.i_neighbour4[0] =
for( i = 0; i < 3; i++ )
{
int w = i ? 8 : 16;
+ int i_stride = h->fdec->i_stride[i];
+ int i_stride2 = i_stride << h->mb.b_interlaced;
+ int i_pix_offset = h->mb.b_interlaced
+ ? w * (h->mb.i_mb_x + (h->mb.i_mb_y&~1) * i_stride) + (h->mb.i_mb_y&1) * i_stride
+ : w * (h->mb.i_mb_x + h->mb.i_mb_y * i_stride);
h->mc.copy[i?PIXEL_8x8:PIXEL_16x16](
- &h->fdec->plane[i][ w * (h->mb.i_mb_x + h->mb.i_mb_y * h->fdec->i_stride[i]) ],
- h->fdec->i_stride[i],
+ &h->fdec->plane[i][i_pix_offset], i_stride2,
h->mb.pic.p_fdec[i], FDEC_STRIDE, w );
}
+ x264_prefetch_fenc( h, h->fdec, h->mb.i_mb_x, h->mb.i_mb_y );
+
h->mb.type[i_mb_xy] = i_mb_type;
if( h->mb.i_type != I_16x16 && h->mb.i_cbp_luma == 0 && h->mb.i_cbp_chroma == 0 )
h->mb.i_last_dqp = h->mb.i_qp - h->mb.i_last_qp;
h->mb.i_last_qp = h->mb.i_qp;
+ h->mb.i_mb_prev_xy = h->mb.i_mb_xy;
/* save intra4x4 */
if( i_mb_type == I_4x4 )
h->mb.bipred_weight[i_ref0][i_ref1] = 32;
}
}
+ if( h->sh.b_mbaff )
+ {
+ for( i_ref0 = 2*h->i_ref0-1; i_ref0 >= 0; i_ref0-- )
+ for( i_ref1 = 2*h->i_ref1-1; i_ref1 >= 0; i_ref1-- )
+ h->mb.bipred_weight[i_ref0][i_ref1] = h->mb.bipred_weight[i_ref0>>1][i_ref1>>1];
+ }
}