B_DIRECT, B_L0_L0, B_L0_L1, B_L0_BI, B_L1_L0, B_L1_L1,
B_L1_BI, B_BI_L0, B_BI_L1, B_BI_BI, B_8x8, B_SKIP
};
-static const uint8_t x264_mb_type_list0_table[X264_MBTYPE_MAX][2] =
-{
- {0,0}, {0,0}, {0,0}, {0,0}, /* INTRA */
- {1,1}, /* P_L0 */
- {0,0}, /* P_8x8 */
- {1,1}, /* P_SKIP */
- {0,0}, /* B_DIRECT */
- {1,1}, {1,0}, {1,1}, /* B_L0_* */
- {0,1}, {0,0}, {0,1}, /* B_L1_* */
- {1,1}, {1,0}, {1,1}, /* B_BI_* */
- {0,0}, /* B_8x8 */
- {0,0} /* B_SKIP */
-};
-static const uint8_t x264_mb_type_list1_table[X264_MBTYPE_MAX][2] =
-{
- {0,0}, {0,0}, {0,0}, {0,0}, /* INTRA */
- {0,0}, /* P_L0 */
- {0,0}, /* P_8x8 */
- {0,0}, /* P_SKIP */
- {0,0}, /* B_DIRECT */
- {0,0}, {0,1}, {0,1}, /* B_L0_* */
- {1,0}, {1,1}, {1,1}, /* B_L1_* */
- {1,0}, {1,1}, {1,1}, /* B_BI_* */
- {0,0}, /* B_8x8 */
- {0,0} /* B_SKIP */
+static const uint8_t x264_mb_type_list_table[X264_MBTYPE_MAX][2][2] =
+{
+ {{0,0},{0,0}}, {{0,0},{0,0}}, {{0,0},{0,0}}, {{0,0},{0,0}}, /* INTRA */
+ {{1,1},{0,0}}, /* P_L0 */
+ {{0,0},{0,0}}, /* P_8x8 */
+ {{1,1},{0,0}}, /* P_SKIP */
+ {{0,0},{0,0}}, /* B_DIRECT */
+ {{1,1},{0,0}}, {{1,0},{0,1}}, {{1,1},{0,1}}, /* B_L0_* */
+ {{0,1},{1,0}}, {{0,0},{1,1}}, {{0,1},{1,1}}, /* B_L1_* */
+ {{1,1},{1,0}}, {{1,0},{1,1}}, {{1,1},{1,1}}, /* B_BI_* */
+ {{0,0},{0,0}}, /* B_8x8 */
+ {{0,0},{0,0}} /* B_SKIP */
};
#define IS_SUB4x4(type) ( (type ==D_L0_4x4)||(type ==D_L1_4x4)||(type ==D_BI_4x4))
int x264_macroblock_cache_init( x264_t *h );
void x264_macroblock_slice_init( x264_t *h );
+void x264_macroblock_thread_init( x264_t *h );
void x264_macroblock_cache_load( x264_t *h, int i_mb_x, int i_mb_y );
void x264_macroblock_cache_save( x264_t *h );
void x264_macroblock_cache_end( x264_t *h );
* if b_changed != NULL, set it to whether refs or mvs differ from
* before this functioncall. */
int x264_mb_predict_mv_direct16x16( x264_t *h, int *b_changed );
-/* x264_mb_load_mv_direct8x8:
- * set h->mb.cache.mv and h->mb.cache.ref for B_DIRECT
- * must be called only after x264_mb_predict_mv_direct16x16 */
-void x264_mb_load_mv_direct8x8( x264_t *h, int idx );
/* x264_mb_predict_mv_ref16x16:
* set mvc with D_16x16 prediction.
* uses all neighbors, even those that didn't end up using this ref.
return (a&0xFFFF) + (b<<16);
#endif
}
-static ALWAYS_INLINE void x264_macroblock_cache_rect1( void *dst, int width, int height, uint8_t val )
-{
- int dy;
- if( width == 4 )
- {
- uint32_t val2 = val * 0x01010101;
- for( dy = 0; dy < height; dy++ )
- ((uint32_t*)dst)[2*dy] = val2;
- }
- else // 2
- {
- uint32_t val2 = val * 0x0101;
- for( dy = 0; dy < height; dy++ )
- ((uint16_t*)dst)[4*dy] = val2;
- }
-}
-static ALWAYS_INLINE void x264_macroblock_cache_rect4( void *dst, int width, int height, uint32_t val )
-{
- int dy, dx;
- if( width == 1 || WORD_SIZE < 8 )
- {
- for( dy = 0; dy < height; dy++ )
- for( dx = 0; dx < width; dx++ )
- ((uint32_t*)dst)[dx+8*dy] = val;
- }
- else
- {
- uint64_t val64 = val + ((uint64_t)val<<32);
- for( dy = 0; dy < height; dy++ )
- for( dx = 0; dx < width/2; dx++ )
- ((uint64_t*)dst)[dx+4*dy] = val64;
- }
-}
-#define x264_macroblock_cache_mv_ptr(a,x,y,w,h,l,mv) x264_macroblock_cache_mv(a,x,y,w,h,l,*(uint32_t*)mv)
-static ALWAYS_INLINE void x264_macroblock_cache_mv( x264_t *h, int x, int y, int width, int height, int i_list, uint32_t mv )
-{
- x264_macroblock_cache_rect4( &h->mb.cache.mv[i_list][X264_SCAN8_0+x+8*y], width, height, mv );
-}
-static ALWAYS_INLINE void x264_macroblock_cache_mvd( x264_t *h, int x, int y, int width, int height, int i_list, uint32_t mv )
-{
- x264_macroblock_cache_rect4( &h->mb.cache.mvd[i_list][X264_SCAN8_0+x+8*y], width, height, mv );
-}
-static ALWAYS_INLINE void x264_macroblock_cache_ref( x264_t *h, int x, int y, int width, int height, int i_list, uint8_t ref )
-{
- x264_macroblock_cache_rect1( &h->mb.cache.ref[i_list][X264_SCAN8_0+x+8*y], width, height, ref );
-}
-static ALWAYS_INLINE void x264_macroblock_cache_skip( x264_t *h, int x, int y, int width, int height, int b_skip )
-{
- x264_macroblock_cache_rect1( &h->mb.cache.skip[X264_SCAN8_0+x+8*y], width, height, b_skip );
-}
-static ALWAYS_INLINE void x264_macroblock_cache_intra8x8_pred( x264_t *h, int x, int y, int i_mode )
-{
- int8_t *cache = &h->mb.cache.intra4x4_pred_mode[X264_SCAN8_0+x+8*y];
- cache[0] = cache[1] = cache[8] = cache[9] = i_mode;
-}
+
#define array_non_zero(a) array_non_zero_int(a, sizeof(a))
-#define array_non_zero_int array_non_zero_int_c
-static ALWAYS_INLINE int array_non_zero_int_c( void *v, int i_count )
+#define array_non_zero_int array_non_zero_int
+static ALWAYS_INLINE int array_non_zero_int( int16_t *v, int i_count )
{
- uint64_t *x = v;
if(i_count == 8)
- return !!x[0];
+ return !!M64( &v[0] );
else if(i_count == 16)
- return !!(x[0]|x[1]);
+ return !!(M64( &v[0] ) | M64( &v[4] ));
else if(i_count == 32)
- return !!(x[0]|x[1]|x[2]|x[3]);
+ return !!(M64( &v[0] ) | M64( &v[4] ) | M64( &v[8] ) | M64( &v[12] ));
else
{
- int i;
- i_count /= sizeof(uint64_t);
- for( i = 0; i < i_count; i++ )
- if( x[i] ) return 1;
+ for( int i = 0; i < i_count; i+=4 )
+ if( M64( &v[i] ) ) return 1;
return 0;
}
}
-/* This function and its MMX version only work on arrays of size 16 */
-static ALWAYS_INLINE int array_non_zero_count( int16_t *v )
-{
- int i;
- int i_nz;
-
- for( i = 0, i_nz = 0; i < 16; i++ )
- if( v[i] )
- i_nz++;
-
- return i_nz;
-}
-static inline int x264_mb_predict_intra4x4_mode( x264_t *h, int idx )
+static ALWAYS_INLINE int x264_mb_predict_intra4x4_mode( x264_t *h, int idx )
{
const int ma = h->mb.cache.intra4x4_pred_mode[x264_scan8[idx] - 1];
const int mb = h->mb.cache.intra4x4_pred_mode[x264_scan8[idx] - 8];
return m;
}
-static inline int x264_mb_predict_non_zero_code( x264_t *h, int idx )
+static ALWAYS_INLINE int x264_mb_predict_non_zero_code( x264_t *h, int idx )
{
const int za = h->mb.cache.non_zero_count[x264_scan8[idx] - 1];
const int zb = h->mb.cache.non_zero_count[x264_scan8[idx] - 8];
int i_ret = za + zb;
if( i_ret < 0x80 )
- {
i_ret = ( i_ret + 1 ) >> 1;
- }
return i_ret & 0x7f;
}
/* x264_mb_transform_8x8_allowed:
* check whether any partition is smaller than 8x8 (or at least
* might be, according to just partition type.)
* doesn't check for cbp */
-static inline int x264_mb_transform_8x8_allowed( x264_t *h )
+static ALWAYS_INLINE int x264_mb_transform_8x8_allowed( x264_t *h )
{
// 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,
+ 0,0,0,0,1,2,0,1,1,1,1,1,1,1,1,1,1,1,0,
};
- int p, i;
if( !h->pps->b_transform_8x8_mode )
return 0;
- 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 )
- {
- 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 != P_8x8 )
+ return partition_tab[h->mb.i_type];
+ return M32( h->mb.i_sub_partition ) == D_L0_8x8*0x01010101;
}
#endif