/* (x-1)%6 */
static const uint8_t mod6m1[8] = {5,0,1,2,3,4,5,0};
/* radius 2 hexagon. repeated entries are to avoid having to compute mod6 every time. */
-static const int hex2[8][2] = {{-1,-2}, {-2,0}, {-1,2}, {1,2}, {2,0}, {1,-2}, {-1,-2}, {-2,0}};
-static const int square1[9][2] = {{0,0}, {0,-1}, {0,1}, {-1,0}, {1,0}, {-1,-1}, {-1,1}, {1,-1}, {1,1}};
+static const int8_t hex2[8][2] = {{-1,-2}, {-2,0}, {-1,2}, {1,2}, {2,0}, {1,-2}, {-1,-2}, {-2,0}};
+static const int8_t square1[9][2] = {{0,0}, {0,-1}, {0,1}, {-1,0}, {1,0}, {-1,-1}, {-1,1}, {1,-1}, {1,1}};
static void refine_subpel( x264_t *h, x264_me_t *m, int hpel_iters, int qpel_iters, int *p_halfpel_thresh, int b_refine_qpel );
#define COST_MV_HPEL( mx, my ) \
{ \
int stride2 = 16; \
- uint8_t *src = h->mc.get_ref( pix, &stride2, m->p_fref, stride, mx, my, bw, bh, &m->weight[0] ); \
+ pixel *src = h->mc.get_ref( pix, &stride2, m->p_fref, stride, mx, my, bw, bh, &m->weight[0] ); \
int cost = h->pixf.fpelcmp[i_pixel]( p_fenc, FENC_STRIDE, src, stride2 ) \
+ p_cost_mvx[ mx ] + p_cost_mvy[ my ]; \
COPY3_IF_LT( bpred_cost, cost, bpred_mx, mx, bpred_my, my ); \
#define COST_MV_X3_DIR( m0x, m0y, m1x, m1y, m2x, m2y, costs )\
{\
- uint8_t *pix_base = p_fref_w + bmx + bmy*stride;\
+ pixel *pix_base = p_fref_w + bmx + bmy*stride;\
h->pixf.fpelcmp_x3[i_pixel]( p_fenc,\
pix_base + (m0x) + (m0y)*stride,\
pix_base + (m1x) + (m1y)*stride,\
#define COST_MV_X4_DIR( m0x, m0y, m1x, m1y, m2x, m2y, m3x, m3y, costs )\
{\
- uint8_t *pix_base = p_fref_w + bmx + bmy*stride;\
+ pixel *pix_base = p_fref_w + bmx + bmy*stride;\
h->pixf.fpelcmp_x4[i_pixel]( p_fenc,\
pix_base + (m0x) + (m0y)*stride,\
pix_base + (m1x) + (m1y)*stride,\
#define COST_MV_X4( m0x, m0y, m1x, m1y, m2x, m2y, m3x, m3y )\
{\
- uint8_t *pix_base = p_fref_w + omx + omy*stride;\
+ pixel *pix_base = p_fref_w + omx + omy*stride;\
h->pixf.fpelcmp_x4[i_pixel]( p_fenc,\
pix_base + (m0x) + (m0y)*stride,\
pix_base + (m1x) + (m1y)*stride,\
#define CROSS( start, x_max, y_max )\
{\
- i = start;\
+ int i = start;\
if( (x_max) <= X264_MIN(mv_x_max-omx, omx-mv_x_min) )\
for( ; i < (x_max)-2; i+=4 )\
COST_MV_X4( i,0, -i,0, i+2,0, -i-2,0 );\
int bmx, bmy, bcost;
int bpred_mx = 0, bpred_my = 0, bpred_cost = COST_MAX;
int omx, omy, pmx, pmy;
- uint8_t *p_fenc = m->p_fenc[0];
- uint8_t *p_fref_w = m->p_fref_w;
- ALIGNED_ARRAY_16( uint8_t, pix,[16*16] );
+ pixel *p_fenc = m->p_fenc[0];
+ pixel *p_fref_w = m->p_fref_w;
+ ALIGNED_ARRAY_16( pixel, pix,[16*16] );
- int i, j;
- int dir;
int costs[16];
int mv_x_min = h->mb.mv_min_fpel[0];
const uint16_t *p_cost_mvx = m->p_cost_mv - m->mvp[0];
const uint16_t *p_cost_mvy = m->p_cost_mv - m->mvp[1];
+ uint32_t pmv;
bmx = x264_clip3( m->mvp[0], mv_x_min_qpel, mv_x_max_qpel );
bmy = x264_clip3( m->mvp[1], mv_y_min_qpel, mv_y_max_qpel );
pmx = ( bmx + 2 ) >> 2;
/* try extra predictors if provided */
if( h->mb.i_subpel_refine >= 3 )
{
- uint32_t bmv = pack16to32_mask(bmx,bmy);
+ pmv = pack16to32_mask(bmx,bmy);
if( i_mvc )
COST_MV_HPEL( bmx, bmy );
- for( i = 0; i < i_mvc; i++ )
+ for( int i = 0; i < i_mvc; i++ )
{
- if( M32( mvc[i] ) && (bmv - M32( mvc[i] )) )
+ if( M32( mvc[i] ) && (pmv != M32( mvc[i] )) )
{
int mx = x264_clip3( mvc[i][0], mv_x_min_qpel, mv_x_max_qpel );
int my = x264_clip3( mvc[i][1], mv_y_min_qpel, mv_y_max_qpel );
else
{
/* check the MVP */
- COST_MV( pmx, pmy );
+ bmx = pmx;
+ bmy = pmy;
/* Because we are rounding the predicted motion vector to fullpel, there will be
* an extra MV cost in 15 out of 16 cases. However, when the predicted MV is
* chosen as the best predictor, it is often the case that the subpel search will
* result in a vector at or next to the predicted motion vector. Therefore, it is
- * sensible to remove the cost of the MV from the rounded MVP to avoid unfairly
+ * sensible to omit the cost of the MV from the rounded MVP to avoid unfairly
* biasing against use of the predicted motion vector. */
- bcost -= BITS_MVD( pmx, pmy );
- for( i = 0; i < i_mvc; i++ )
+ bcost = h->pixf.fpelcmp[i_pixel]( p_fenc, FENC_STRIDE, &p_fref_w[bmy*stride+bmx], stride );
+ pmv = pack16to32_mask( bmx, bmy );
+ if( i_mvc > 0 )
{
- int mx = (mvc[i][0] + 2) >> 2;
- int my = (mvc[i][1] + 2) >> 2;
- if( (mx | my) && ((mx-bmx) | (my-bmy)) )
+ ALIGNED_ARRAY_8( int16_t, mvc_fpel,[16],[2] );
+ x264_predictor_roundclip( mvc_fpel, mvc, i_mvc, mv_x_min, mv_x_max, mv_y_min, mv_y_max );
+ bcost <<= 4;
+ for( int i = 1; i <= i_mvc; i++ )
+ {
+ if( M32( mvc_fpel[i-1] ) && (pmv != M32( mvc[i-1] )) )
+ {
+ int mx = mvc_fpel[i-1][0];
+ int my = mvc_fpel[i-1][1];
+ int cost = h->pixf.fpelcmp[i_pixel]( p_fenc, FENC_STRIDE, &p_fref_w[my*stride+mx], stride ) + BITS_MVD( mx, my );
+ cost = (cost << 4) + i;
+ COPY1_IF_LT( bcost, cost );
+ }
+ }
+ if( bcost&15 )
{
- mx = x264_clip3( mx, mv_x_min, mv_x_max );
- my = x264_clip3( my, mv_y_min, mv_y_max );
- COST_MV( mx, my );
+ bmx = mvc_fpel[(bcost&15)-1][0];
+ bmy = mvc_fpel[(bcost&15)-1][1];
}
+ bcost >>= 4;
}
}
- COST_MV( 0, 0 );
+
+ if( pmv )
+ COST_MV( 0, 0 );
switch( h->mb.i_me_method )
{
- case X264_ME_DIA:
- /* diamond search, radius 1 */
- i = 0;
- bcost <<= 4;
- do
- {
- COST_MV_X4_DIR( 0,-1, 0,1, -1,0, 1,0, costs );
- COPY1_IF_LT( bcost, (costs[0]<<4)+1 );
- COPY1_IF_LT( bcost, (costs[1]<<4)+3 );
- COPY1_IF_LT( bcost, (costs[2]<<4)+4 );
- COPY1_IF_LT( bcost, (costs[3]<<4)+12 );
- if( !(bcost&15) )
- break;
- bmx -= (bcost<<28)>>30;
- bmy -= (bcost<<30)>>30;
- bcost &= ~15;
- if( !CHECK_MVRANGE(bmx, bmy) )
- break;
- } while( ++i < i_me_range );
- bcost >>= 4;
- break;
-
- case X264_ME_HEX:
-me_hex2:
- /* hexagon search, radius 2 */
-#if 0
- for( i = 0; i < i_me_range/2; i++ )
+ case X264_ME_DIA:
{
- omx = bmx; omy = bmy;
- COST_MV( omx-2, omy );
- COST_MV( omx-1, omy+2 );
- COST_MV( omx+1, omy+2 );
- COST_MV( omx+2, omy );
- COST_MV( omx+1, omy-2 );
- COST_MV( omx-1, omy-2 );
- if( bmx == omx && bmy == omy )
- break;
- if( !CHECK_MVRANGE(bmx, bmy) )
- break;
+ /* diamond search, radius 1 */
+ bcost <<= 4;
+ int i = i_me_range;
+ do
+ {
+ COST_MV_X4_DIR( 0,-1, 0,1, -1,0, 1,0, costs );
+ COPY1_IF_LT( bcost, (costs[0]<<4)+1 );
+ COPY1_IF_LT( bcost, (costs[1]<<4)+3 );
+ COPY1_IF_LT( bcost, (costs[2]<<4)+4 );
+ COPY1_IF_LT( bcost, (costs[3]<<4)+12 );
+ if( !(bcost&15) )
+ break;
+ bmx -= (bcost<<28)>>30;
+ bmy -= (bcost<<30)>>30;
+ bcost &= ~15;
+ } while( --i && CHECK_MVRANGE(bmx, bmy) );
+ bcost >>= 4;
+ break;
}
-#else
- /* equivalent to the above, but eliminates duplicate candidates */
-
- /* hexagon */
- COST_MV_X3_DIR( -2,0, -1, 2, 1, 2, costs );
- COST_MV_X3_DIR( 2,0, 1,-2, -1,-2, costs+3 );
- bcost <<= 3;
- COPY1_IF_LT( bcost, (costs[0]<<3)+2 );
- COPY1_IF_LT( bcost, (costs[1]<<3)+3 );
- COPY1_IF_LT( bcost, (costs[2]<<3)+4 );
- COPY1_IF_LT( bcost, (costs[3]<<3)+5 );
- COPY1_IF_LT( bcost, (costs[4]<<3)+6 );
- COPY1_IF_LT( bcost, (costs[5]<<3)+7 );
-
- if( bcost&7 )
- {
- dir = (bcost&7)-2;
- bmx += hex2[dir+1][0];
- bmy += hex2[dir+1][1];
- /* half hexagon, not overlapping the previous iteration */
- for( i = 1; i < i_me_range>>1 && CHECK_MVRANGE(bmx, bmy); i++ )
+ case X264_ME_HEX:
{
- COST_MV_X3_DIR( hex2[dir+0][0], hex2[dir+0][1],
- hex2[dir+1][0], hex2[dir+1][1],
- hex2[dir+2][0], hex2[dir+2][1],
- costs );
- bcost &= ~7;
- COPY1_IF_LT( bcost, (costs[0]<<3)+1 );
- COPY1_IF_LT( bcost, (costs[1]<<3)+2 );
- COPY1_IF_LT( bcost, (costs[2]<<3)+3 );
- if( !(bcost&7) )
+ me_hex2:
+ /* hexagon search, radius 2 */
+ #if 0
+ for( int i = 0; i < i_me_range/2; i++ )
+ {
+ omx = bmx; omy = bmy;
+ COST_MV( omx-2, omy );
+ COST_MV( omx-1, omy+2 );
+ COST_MV( omx+1, omy+2 );
+ COST_MV( omx+2, omy );
+ COST_MV( omx+1, omy-2 );
+ COST_MV( omx-1, omy-2 );
+ if( bmx == omx && bmy == omy )
+ break;
+ if( !CHECK_MVRANGE(bmx, bmy) )
break;
- dir += (bcost&7)-2;
- dir = mod6m1[dir+1];
+ }
+ #else
+ /* equivalent to the above, but eliminates duplicate candidates */
+
+ /* hexagon */
+ COST_MV_X3_DIR( -2,0, -1, 2, 1, 2, costs );
+ COST_MV_X3_DIR( 2,0, 1,-2, -1,-2, costs+3 );
+ bcost <<= 3;
+ COPY1_IF_LT( bcost, (costs[0]<<3)+2 );
+ COPY1_IF_LT( bcost, (costs[1]<<3)+3 );
+ COPY1_IF_LT( bcost, (costs[2]<<3)+4 );
+ COPY1_IF_LT( bcost, (costs[3]<<3)+5 );
+ COPY1_IF_LT( bcost, (costs[4]<<3)+6 );
+ COPY1_IF_LT( bcost, (costs[5]<<3)+7 );
+
+ if( bcost&7 )
+ {
+ int dir = (bcost&7)-2;
bmx += hex2[dir+1][0];
bmy += hex2[dir+1][1];
+
+ /* half hexagon, not overlapping the previous iteration */
+ for( int i = (i_me_range>>1) - 1; i > 0 && CHECK_MVRANGE(bmx, bmy); i-- )
+ {
+ COST_MV_X3_DIR( hex2[dir+0][0], hex2[dir+0][1],
+ hex2[dir+1][0], hex2[dir+1][1],
+ hex2[dir+2][0], hex2[dir+2][1],
+ costs );
+ bcost &= ~7;
+ COPY1_IF_LT( bcost, (costs[0]<<3)+1 );
+ COPY1_IF_LT( bcost, (costs[1]<<3)+2 );
+ COPY1_IF_LT( bcost, (costs[2]<<3)+3 );
+ if( !(bcost&7) )
+ break;
+ dir += (bcost&7)-2;
+ dir = mod6m1[dir+1];
+ bmx += hex2[dir+1][0];
+ bmy += hex2[dir+1][1];
+ }
}
+ bcost >>= 3;
+ #endif
+ /* square refine */
+ int dir = 0;
+ COST_MV_X4_DIR( 0,-1, 0,1, -1,0, 1,0, costs );
+ COPY2_IF_LT( bcost, costs[0], dir, 1 );
+ COPY2_IF_LT( bcost, costs[1], dir, 2 );
+ COPY2_IF_LT( bcost, costs[2], dir, 3 );
+ COPY2_IF_LT( bcost, costs[3], dir, 4 );
+ COST_MV_X4_DIR( -1,-1, -1,1, 1,-1, 1,1, costs );
+ COPY2_IF_LT( bcost, costs[0], dir, 5 );
+ COPY2_IF_LT( bcost, costs[1], dir, 6 );
+ COPY2_IF_LT( bcost, costs[2], dir, 7 );
+ COPY2_IF_LT( bcost, costs[3], dir, 8 );
+ bmx += square1[dir][0];
+ bmy += square1[dir][1];
+ break;
}
- bcost >>= 3;
-#endif
- /* square refine */
- dir = 0;
- COST_MV_X4_DIR( 0,-1, 0,1, -1,0, 1,0, costs );
- COPY2_IF_LT( bcost, costs[0], dir, 1 );
- COPY2_IF_LT( bcost, costs[1], dir, 2 );
- COPY2_IF_LT( bcost, costs[2], dir, 3 );
- COPY2_IF_LT( bcost, costs[3], dir, 4 );
- COST_MV_X4_DIR( -1,-1, -1,1, 1,-1, 1,1, costs );
- COPY2_IF_LT( bcost, costs[0], dir, 5 );
- COPY2_IF_LT( bcost, costs[1], dir, 6 );
- COPY2_IF_LT( bcost, costs[2], dir, 7 );
- COPY2_IF_LT( bcost, costs[3], dir, 8 );
- bmx += square1[dir][0];
- bmy += square1[dir][1];
- break;
- case X264_ME_UMH:
+ case X264_ME_UMH:
{
/* Uneven-cross Multi-Hexagon-grid Search
* as in JM, except with different early termination */
- static const int x264_pixel_size_shift[7] = { 0, 1, 1, 2, 3, 3, 4 };
+ static const uint8_t x264_pixel_size_shift[7] = { 0, 1, 1, 2, 3, 3, 4 };
int ucost1, ucost2;
int cross_start = 1;
if( pmx | pmy )
DIA1_ITER( 0, 0 );
- if(i_pixel == PIXEL_4x4)
+ if( i_pixel == PIXEL_4x4 )
goto me_hex2;
ucost2 = bcost;
/* range multipliers based on casual inspection of some statistics of
* average distance between current predictor and final mv found by ESA.
* these have not been tuned much by actual encoding. */
- static const int range_mul[4][4] =
+ static const uint8_t range_mul[4][4] =
{
{ 3, 3, 4, 4 },
{ 3, 4, 4, 4 },
: mvd < 20*denom ? 1
: mvd < 40*denom ? 2 : 3;
- i_me_range = i_me_range * range_mul[mvd_ctx][sad_ctx] / 4;
+ i_me_range = i_me_range * range_mul[mvd_ctx][sad_ctx] >> 2;
}
/* FIXME if the above DIA2/OCT2/CROSS found a new mv, it has not updated omx/omy.
omx = bmx; omy = bmy;
const uint16_t *p_cost_omvx = p_cost_mvx + omx*4;
const uint16_t *p_cost_omvy = p_cost_mvy + omy*4;
- i = 1;
+ int i = 1;
do
{
- static const int hex4[16][2] = {
+ static const int8_t hex4[16][2] = {
{ 0,-4}, { 0, 4}, {-2,-3}, { 2,-3},
{-4,-2}, { 4,-2}, {-4,-1}, { 4,-1},
{-4, 0}, { 4, 0}, {-4, 1}, { 4, 1},
if( 4*i > X264_MIN4( mv_x_max-omx, omx-mv_x_min,
mv_y_max-omy, omy-mv_y_min ) )
{
- for( j = 0; j < 16; j++ )
+ for( int j = 0; j < 16; j++ )
{
int mx = omx + hex4[j][0]*i;
int my = omy + hex4[j][1]*i;
else
{
int dir = 0;
- uint8_t *pix_base = p_fref_w + omx + (omy-4*i)*stride;
+ pixel *pix_base = p_fref_w + omx + (omy-4*i)*stride;
int dy = i*stride;
#define SADS(k,x0,y0,x1,y1,x2,y2,x3,y3)\
h->pixf.fpelcmp_x4[i_pixel]( p_fenc,\
break;
}
- case X264_ME_ESA:
- case X264_ME_TESA:
+ case X264_ME_ESA:
+ case X264_ME_TESA:
{
const int min_x = X264_MAX( bmx - i_me_range, mv_x_min );
const int min_y = X264_MAX( bmy - i_me_range, mv_y_min );
const int max_y = X264_MIN( bmy + i_me_range, mv_y_max );
/* SEA is fastest in multiples of 4 */
const int width = (max_x - min_x + 3) & ~3;
- int my;
#if 0
/* plain old exhaustive search */
- int mx;
- for( my = min_y; my <= max_y; my++ )
- for( mx = min_x; mx <= max_x; mx++ )
+ for( int my = min_y; my <= max_y; my++ )
+ for( int mx = min_x; mx <= max_x; mx++ )
COST_MV( mx, my );
#else
/* successive elimination by comparing DC before a full SAD,
uint16_t *sums_base = m->integral;
/* due to a GCC bug on some platforms (win32?), zero[] may not actually be aligned.
* this is not a problem because it is not used for any SSE instructions. */
- ALIGNED_16( static uint8_t zero[8*FENC_STRIDE] );
+ ALIGNED_16( static pixel zero[8*FENC_STRIDE] );
ALIGNED_ARRAY_16( int, enc_dc,[4] );
int sad_size = i_pixel <= PIXEL_8x8 ? PIXEL_8x8 : PIXEL_4x4;
int delta = x264_pixel_size[sad_size].w;
if( h->mb.i_me_method == X264_ME_TESA )
{
// ADS threshold, then SAD threshold, then keep the best few SADs, then SATD
- mvsad_t *mvsads = (mvsad_t *)(xs + ((width+15)&~15));
+ mvsad_t *mvsads = (mvsad_t *)(xs + ((width+15)&~15) + 4);
int nmvsad = 0, limit;
int sad_thresh = i_me_range <= 16 ? 10 : i_me_range <= 24 ? 11 : 12;
int bsad = h->pixf.sad[i_pixel]( p_fenc, FENC_STRIDE, p_fref_w+bmy*stride+bmx, stride )
+ BITS_MVD( bmx, bmy );
- for( my = min_y; my <= max_y; my++ )
+ for( int my = min_y; my <= max_y; my++ )
{
+ int i;
int ycost = p_cost_mvy[my<<2];
if( bsad <= ycost )
continue;
bsad -= ycost;
xn = h->pixf.ads[i_pixel]( enc_dc, sums_base + min_x + my * stride, delta,
- cost_fpel_mvx+min_x, xs, width, bsad*17/16 );
- for( i=0; i<xn-2; i+=3 )
+ cost_fpel_mvx+min_x, xs, width, bsad * 17 >> 4 );
+ for( i = 0; i < xn-2; i += 3 )
{
- uint8_t *ref = p_fref_w+min_x+my*stride;
+ pixel *ref = p_fref_w+min_x+my*stride;
int sads[3];
h->pixf.sad_x3[i_pixel]( p_fenc, ref+xs[i], ref+xs[i+1], ref+xs[i+2], stride, sads );
- for( j=0; j<3; j++ )
+ for( int j = 0; j < 3; j++ )
{
int sad = sads[j] + cost_fpel_mvx[xs[i+j]];
if( sad < bsad*sad_thresh>>3 )
}
}
}
- for( ; i<xn; i++ )
+ for( ; i < xn; i++ )
{
int mx = min_x+xs[i];
int sad = h->pixf.sad[i_pixel]( p_fenc, FENC_STRIDE, p_fref_w+mx+my*stride, stride )
bsad += ycost;
}
- limit = i_me_range / 2;
+ limit = i_me_range >> 1;
sad_thresh = bsad*sad_thresh>>3;
while( nmvsad > limit*2 && sad_thresh > bsad )
{
+ int i;
// halve the range if the domain is too large... eh, close enough
sad_thresh = (sad_thresh + bsad) >> 1;
- for( i=0; i<nmvsad && mvsads[i].sad <= sad_thresh; i++ );
- for( j=i; j<nmvsad; j++ )
+ for( i = 0; i < nmvsad && mvsads[i].sad <= sad_thresh; i++ );
+ for( int j = i; j < nmvsad; j++ )
{
uint32_t sad;
if( WORD_SIZE == 8 && sizeof(mvsad_t) == 8 )
{
uint64_t mvsad = M64( &mvsads[i] ) = M64( &mvsads[j] );
-#ifdef WORDS_BIGENDIAN
+#if WORDS_BIGENDIAN
mvsad >>= 32;
#endif
sad = mvsad;
while( nmvsad > limit )
{
int bi = 0;
- for( i=1; i<nmvsad; i++ )
+ for( int i = 1; i < nmvsad; i++ )
if( mvsads[i].sad > mvsads[bi].sad )
bi = i;
nmvsad--;
else
mvsads[bi] = mvsads[nmvsad];
}
- for( i=0; i<nmvsad; i++ )
+ for( int i = 0; i < nmvsad; i++ )
COST_MV( mvsads[i].mv[0], mvsads[i].mv[1] );
}
else
{
// just ADS and SAD
- for( my = min_y; my <= max_y; my++ )
+ for( int my = min_y; my <= max_y; my++ )
{
+ int i;
int ycost = p_cost_mvy[my<<2];
if( bcost <= ycost )
continue;
bcost -= ycost;
xn = h->pixf.ads[i_pixel]( enc_dc, sums_base + min_x + my * stride, delta,
cost_fpel_mvx+min_x, xs, width, bcost );
- for( i=0; i<xn-2; i+=3 )
+ for( i = 0; i < xn-2; i += 3 )
COST_MV_X3_ABS( min_x+xs[i],my, min_x+xs[i+1],my, min_x+xs[i+2],my );
bcost += ycost;
- for( ; i<xn; i++ )
+ for( ; i < xn; i++ )
COST_MV( min_x+xs[i], my );
}
}
#define COST_MV_SAD( mx, my ) \
{ \
int stride = 16; \
- uint8_t *src = h->mc.get_ref( pix[0], &stride, m->p_fref, m->i_stride[0], mx, my, bw, bh, &m->weight[0] ); \
+ pixel *src = h->mc.get_ref( pix, &stride, m->p_fref, m->i_stride[0], mx, my, bw, bh, &m->weight[0] ); \
int cost = h->pixf.fpelcmp[i_pixel]( m->p_fenc[0], FENC_STRIDE, src, stride ) \
+ p_cost_mvx[ mx ] + p_cost_mvy[ my ]; \
COPY3_IF_LT( bcost, cost, bmx, mx, bmy, my ); \
if( b_refine_qpel || (dir^1) != odir ) \
{ \
int stride = 16; \
- uint8_t *src = h->mc.get_ref( pix[0], &stride, m->p_fref, m->i_stride[0], mx, my, bw, bh, &m->weight[0] ); \
+ pixel *src = h->mc.get_ref( pix, &stride, m->p_fref, m->i_stride[0], mx, my, bw, bh, &m->weight[0] ); \
int cost = h->pixf.mbcmp_unaligned[i_pixel]( m->p_fenc[0], FENC_STRIDE, src, stride ) \
+ p_cost_mvx[ mx ] + p_cost_mvy[ my ]; \
if( b_chroma_me && cost < bcost ) \
{ \
- h->mc.mc_chroma( pix[0], 8, m->p_fref[4], m->i_stride[1], mx, my + mvy_offset, bw/2, bh/2 ); \
+ h->mc.mc_chroma( pix, 8, m->p_fref[4], m->i_stride[1], mx, my + mvy_offset, bw>>1, bh>>1 ); \
if( m->weight[1].weightfn ) \
- m->weight[1].weightfn[x264_pixel_size[i_pixel].w>>3]( pix[0], 8, pix[0], 8, \
+ m->weight[1].weightfn[x264_pixel_size[i_pixel].w>>3]( pix, 8, pix, 8, \
&m->weight[1], x264_pixel_size[i_pixel].h>>1 ); \
- cost += h->pixf.mbcmp[i_pixel+3]( m->p_fenc[1], FENC_STRIDE, pix[0], 8 ); \
+ cost += h->pixf.mbcmp[i_pixel+3]( m->p_fenc[1], FENC_STRIDE, pix, 8 ); \
if( cost < bcost ) \
{ \
- h->mc.mc_chroma( pix[0], 8, m->p_fref[5], m->i_stride[1], mx, my + mvy_offset, bw/2, bh/2 ); \
+ h->mc.mc_chroma( pix, 8, m->p_fref[5], m->i_stride[1], mx, my + mvy_offset, bw>>1, bh>>1 ); \
if( m->weight[2].weightfn ) \
- m->weight[2].weightfn[x264_pixel_size[i_pixel].w>>3]( pix[0], 8, pix[0], 8, \
+ m->weight[2].weightfn[x264_pixel_size[i_pixel].w>>3]( pix, 8, pix, 8, \
&m->weight[2], x264_pixel_size[i_pixel].h>>1 ); \
- cost += h->pixf.mbcmp[i_pixel+3]( m->p_fenc[2], FENC_STRIDE, pix[0], 8 ); \
+ cost += h->pixf.mbcmp[i_pixel+3]( m->p_fenc[2], FENC_STRIDE, pix, 8 ); \
} \
} \
- if( cost < bcost ) \
- { \
- bcost = cost; \
- bmx = mx; \
- bmy = my; \
- bdir = dir; \
- } \
+ COPY4_IF_LT( bcost, cost, bmx, mx, bmy, my, bdir, dir ); \
}
static void refine_subpel( x264_t *h, x264_me_t *m, int hpel_iters, int qpel_iters, int *p_halfpel_thresh, int b_refine_qpel )
const int b_chroma_me = h->mb.b_chroma_me && i_pixel <= PIXEL_8x8;
const int mvy_offset = h->mb.b_interlaced & m->i_ref ? (h->mb.i_mb_y & 1)*4 - 2 : 0;
- ALIGNED_ARRAY_16( uint8_t, pix,[2],[32*18] ); // really 17x17, but round up for alignment
- int omx, omy;
- int i;
+ ALIGNED_ARRAY_16( pixel, pix,[64*18] ); // really 17x17x2, but round up for alignment
int bmx = m->mv[0];
int bmy = m->mv[1];
}
/* halfpel diamond search */
- for( i = hpel_iters; i > 0; i-- )
+ for( int i = hpel_iters; i > 0; i-- )
{
int omx = bmx, omy = bmy;
int costs[4];
- int stride = 32; // candidates are either all hpel or all qpel, so one stride is enough
- uint8_t *src0, *src1, *src2, *src3;
- src0 = h->mc.get_ref( pix[0], &stride, m->p_fref, m->i_stride[0], omx, omy-2, bw, bh+1, &m->weight[0] );
- src2 = h->mc.get_ref( pix[1], &stride, m->p_fref, m->i_stride[0], omx-2, omy, bw+4, bh, &m->weight[0] );
+ int stride = 64; // candidates are either all hpel or all qpel, so one stride is enough
+ pixel *src0, *src1, *src2, *src3;
+ src0 = h->mc.get_ref( pix, &stride, m->p_fref, m->i_stride[0], omx, omy-2, bw, bh+1, &m->weight[0] );
+ src2 = h->mc.get_ref( pix+32, &stride, m->p_fref, m->i_stride[0], omx-2, omy, bw+4, bh, &m->weight[0] );
src1 = src0 + stride;
src3 = src2 + 1;
h->pixf.fpelcmp_x4[i_pixel]( m->p_fenc[0], src0, src1, src2, src3, stride, costs );
break;
}
- if( !b_refine_qpel )
+ if( !b_refine_qpel && (h->pixf.mbcmp_unaligned[0] != h->pixf.fpelcmp[0] || b_chroma_me) )
{
bcost = COST_MAX;
COST_MV_SATD( bmx, bmy, -1 );
}
/* quarterpel diamond search */
- bdir = -1;
- for( i = qpel_iters; i > 0; i-- )
+ if( h->mb.i_subpel_refine != 1 )
{
- if( bmy <= h->mb.mv_min_spel[1] || bmy >= h->mb.mv_max_spel[1] || bmx <= h->mb.mv_min_spel[0] || bmx >= h->mb.mv_max_spel[0] )
- break;
- odir = bdir;
- omx = bmx;
- omy = bmy;
- COST_MV_SATD( omx, omy - 1, 0 );
- COST_MV_SATD( omx, omy + 1, 1 );
- COST_MV_SATD( omx - 1, omy, 2 );
- COST_MV_SATD( omx + 1, omy, 3 );
- if( bmx == omx && bmy == omy )
- break;
+ bdir = -1;
+ for( int i = qpel_iters; i > 0; i-- )
+ {
+ if( bmy <= h->mb.mv_min_spel[1] || bmy >= h->mb.mv_max_spel[1] || bmx <= h->mb.mv_min_spel[0] || bmx >= h->mb.mv_max_spel[0] )
+ break;
+ odir = bdir;
+ int omx = bmx, omy = bmy;
+ COST_MV_SATD( omx, omy - 1, 0 );
+ COST_MV_SATD( omx, omy + 1, 1 );
+ COST_MV_SATD( omx - 1, omy, 2 );
+ COST_MV_SATD( omx + 1, omy, 3 );
+ if( (bmx == omx) & (bmy == omy) )
+ break;
+ }
+ }
+ /* Special simplified case for subme=1 */
+ else if( bmy > h->mb.mv_min_spel[1] && bmy < h->mb.mv_max_spel[1] && bmx > h->mb.mv_min_spel[0] && bmx < h->mb.mv_max_spel[0] )
+ {
+ int costs[4];
+ int omx = bmx, omy = bmy;
+ /* We have to use mc_luma because all strides must be the same to use fpelcmp_x4 */
+ h->mc.mc_luma( pix , 64, m->p_fref, m->i_stride[0], omx, omy-1, bw, bh, &m->weight[0] );
+ h->mc.mc_luma( pix+16, 64, m->p_fref, m->i_stride[0], omx, omy+1, bw, bh, &m->weight[0] );
+ h->mc.mc_luma( pix+32, 64, m->p_fref, m->i_stride[0], omx-1, omy, bw, bh, &m->weight[0] );
+ h->mc.mc_luma( pix+48, 64, m->p_fref, m->i_stride[0], omx+1, omy, bw, bh, &m->weight[0] );
+ h->pixf.fpelcmp_x4[i_pixel]( m->p_fenc[0], pix, pix+16, pix+32, pix+48, 64, costs );
+ COPY2_IF_LT( bcost, costs[0] + p_cost_mvx[omx ] + p_cost_mvy[omy-1], bmy, omy-1 );
+ COPY2_IF_LT( bcost, costs[1] + p_cost_mvx[omx ] + p_cost_mvy[omy+1], bmy, omy+1 );
+ COPY3_IF_LT( bcost, costs[2] + p_cost_mvx[omx-1] + p_cost_mvy[omy ], bmx, omx-1, bmy, omy );
+ COPY3_IF_LT( bcost, costs[3] + p_cost_mvx[omx+1] + p_cost_mvy[omy ], bmx, omx+1, bmy, omy );
}
m->cost = bcost;
m->mv[0] = bmx;
m->mv[1] = bmy;
- m->cost_mv = p_cost_mvx[ bmx ] + p_cost_mvy[ bmy ];
+ m->cost_mv = p_cost_mvx[bmx] + p_cost_mvy[bmy];
}
-#define BIME_CACHE( dx, dy, list ) \
-{ \
+#define BIME_CACHE( dx, dy, list )\
+{\
x264_me_t *m = m##list;\
- int i = 4 + 3*dx + dy; \
+ int i = 4 + 3*dx + dy;\
int mvx = bm##list##x+dx;\
int mvy = bm##list##y+dy;\
stride[list][i] = bw;\
- src[list][i] = h->mc.get_ref( pixy_buf[list][i], &stride[list][i], m->p_fref, m->i_stride[0], mvx, mvy, bw, bh, weight_none ); \
+ src[list][i] = h->mc.get_ref( pixy_buf[list][i], &stride[list][i], m->p_fref, m->i_stride[0], mvx, mvy, bw, bh, weight_none );\
if( rd )\
{\
h->mc.mc_chroma( pixu_buf[list][i], 8, m->p_fref[4], m->i_stride[1], mvx, mvy + mv##list##y_offset, bw>>1, bh>>1 );\
static void ALWAYS_INLINE x264_me_refine_bidir( x264_t *h, x264_me_t *m0, x264_me_t *m1, int i_weight, int i8, int i_lambda2, int rd )
{
- int16_t *cache0_mv = h->mb.cache.mv[0][x264_scan8[i8*4]];
- int16_t *cache1_mv = h->mb.cache.mv[1][x264_scan8[i8*4]];
+ int x = i8&1;
+ int y = i8>>1;
+ int s8 = X264_SCAN8_0 + 2*x + 16*y;
+ int16_t *cache0_mv = h->mb.cache.mv[0][s8];
+ int16_t *cache1_mv = h->mb.cache.mv[1][s8];
const int i_pixel = m0->i_pixel;
const int bw = x264_pixel_size[i_pixel].w;
const int bh = x264_pixel_size[i_pixel].h;
- ALIGNED_ARRAY_16( uint8_t, pixy_buf,[2],[9][16*16] );
- ALIGNED_ARRAY_8( uint8_t, pixu_buf,[2],[9][8*8] );
- ALIGNED_ARRAY_8( uint8_t, pixv_buf,[2],[9][8*8] );
- uint8_t *src[2][9];
- uint8_t *pix = &h->mb.pic.p_fdec[0][(i8>>1)*8*FDEC_STRIDE+(i8&1)*8];
- uint8_t *pixu = &h->mb.pic.p_fdec[1][(i8>>1)*4*FDEC_STRIDE+(i8&1)*4];
- uint8_t *pixv = &h->mb.pic.p_fdec[2][(i8>>1)*4*FDEC_STRIDE+(i8&1)*4];
- const int ref0 = h->mb.cache.ref[0][x264_scan8[i8*4]];
- const int ref1 = h->mb.cache.ref[1][x264_scan8[i8*4]];
+ ALIGNED_ARRAY_16( pixel, pixy_buf,[2],[9][16*16] );
+ ALIGNED_ARRAY_8( pixel, pixu_buf,[2],[9][8*8] );
+ ALIGNED_ARRAY_8( pixel, pixv_buf,[2],[9][8*8] );
+ pixel *src[2][9];
+ pixel *pix = &h->mb.pic.p_fdec[0][8*x + 8*y*FDEC_STRIDE];
+ pixel *pixu = &h->mb.pic.p_fdec[1][4*x + 4*y*FDEC_STRIDE];
+ pixel *pixv = &h->mb.pic.p_fdec[2][4*x + 4*y*FDEC_STRIDE];
+ int ref0 = h->mb.cache.ref[0][s8];
+ int ref1 = h->mb.cache.ref[1][s8];
const int mv0y_offset = h->mb.b_interlaced & ref0 ? (h->mb.i_mb_y & 1)*4 - 2 : 0;
const int mv1y_offset = h->mb.b_interlaced & ref1 ? (h->mb.i_mb_y & 1)*4 - 2 : 0;
int stride[2][9];
int bm1x = m1->mv[0];
int bm1y = m1->mv[1];
int bcost = COST_MAX;
- int pass = 0;
- int j;
int mc_list0 = 1, mc_list1 = 1;
uint64_t bcostrd = COST_MAX64;
uint16_t amvd;
h->mc.memzero_aligned( visited, sizeof(uint8_t[8][8][8]) );
- for( pass = 0; pass < 8; pass++ )
+ for( int pass = 0; pass < 8; pass++ )
{
int bestj = 0;
/* check all mv pairs that differ in at most 2 components from the current mvs. */
* from bidir ME are the same with and without chroma ME. */
if( mc_list0 )
- for( j = x264_iter_kludge; j < 9; j++ )
+ for( int j = x264_iter_kludge; j < 9; j++ )
BIME_CACHE( square1[j][0], square1[j][1], 0 );
if( mc_list1 )
- for( j = x264_iter_kludge; j < 9; j++ )
+ for( int j = x264_iter_kludge; j < 9; j++ )
BIME_CACHE( square1[j][0], square1[j][1], 1 );
- for( j = !!pass; j < 33; j++ )
+ for( int j = !!pass; j < 33; j++ )
{
int m0x = dia4d[j][0] + bm0x;
int m0y = dia4d[j][1] + bm0y;
if( rd )
{
- x264_macroblock_cache_mv ( h, (i8&1)*2, (i8>>1)*2, bw>>2, bh>>2, 0, pack16to32_mask(bm0x, bm0y) );
+ x264_macroblock_cache_mv ( h, 2*x, 2*y, bw>>2, bh>>2, 0, pack16to32_mask(bm0x, bm0y) );
amvd = pack8to16( X264_MIN(abs(bm0x - m0->mvp[0]),33), X264_MIN(abs(bm0y - m0->mvp[1]),33) );
- x264_macroblock_cache_mvd( h, (i8&1)*2, (i8>>1)*2, bw>>2, bh>>2, 0, amvd );
+ x264_macroblock_cache_mvd( h, 2*x, 2*y, bw>>2, bh>>2, 0, amvd );
- x264_macroblock_cache_mv ( h, (i8&1)*2, (i8>>1)*2, bw>>2, bh>>2, 1, pack16to32_mask(bm1x, bm1y) );
+ x264_macroblock_cache_mv ( h, 2*x, 2*y, bw>>2, bh>>2, 1, pack16to32_mask(bm1x, bm1y) );
amvd = pack8to16( X264_MIN(abs(bm1x - m1->mvp[0]),33), X264_MIN(abs(bm1y - m1->mvp[1]),33) );
- x264_macroblock_cache_mvd( h, (i8&1)*2, (i8>>1)*2, bw>>2, bh>>2, 1, amvd );
+ x264_macroblock_cache_mvd( h, 2*x, 2*y, bw>>2, bh>>2, 1, amvd );
}
m0->mv[0] = bm0x;
{ \
uint64_t cost; \
M32( cache_mv ) = pack16to32_mask(mx,my); \
- if( m->i_pixel <= PIXEL_8x8 )\
- {\
- h->mc.mc_chroma( pixu, FDEC_STRIDE, m->p_fref[4], m->i_stride[1], mx, my + mvy_offset, bw>>1, bh>>1 );\
- h->mc.mc_chroma( pixv, FDEC_STRIDE, m->p_fref[5], m->i_stride[1], mx, my + mvy_offset, bw>>1, bh>>1 );\
- }\
+ if( m->i_pixel <= PIXEL_8x8 ) \
+ { \
+ h->mc.mc_chroma( pixu, FDEC_STRIDE, m->p_fref[4], m->i_stride[1], mx, my + mvy_offset, bw>>1, bh>>1 ); \
+ h->mc.mc_chroma( pixv, FDEC_STRIDE, m->p_fref[5], m->i_stride[1], mx, my + mvy_offset, bw>>1, bh>>1 ); \
+ } \
cost = x264_rd_cost_part( h, i_lambda2, i4, m->i_pixel ); \
COPY4_IF_LT( bcost, cost, bmx, mx, bmy, my, dir, do_dir?mdir:dir ); \
} \
uint64_t bcost = COST_MAX64;
int bmx = m->mv[0];
int bmy = m->mv[1];
- int omx, omy, pmx, pmy, i, j;
+ int omx, omy, pmx, pmy;
unsigned bsatd;
int satd;
int dir = -2;
int i8 = i4>>2;
uint16_t amvd;
- uint8_t *pix = &h->mb.pic.p_fdec[0][block_idx_xy_fdec[i4]];
- uint8_t *pixu = &h->mb.pic.p_fdec[1][(i8>>1)*4*FDEC_STRIDE+(i8&1)*4];
- uint8_t *pixv = &h->mb.pic.p_fdec[2][(i8>>1)*4*FDEC_STRIDE+(i8&1)*4];
+ pixel *pix = &h->mb.pic.p_fdec[0][block_idx_xy_fdec[i4]];
+ pixel *pixu = &h->mb.pic.p_fdec[1][(i8>>1)*4*FDEC_STRIDE+(i8&1)*4];
+ pixel *pixv = &h->mb.pic.p_fdec[2][(i8>>1)*4*FDEC_STRIDE+(i8&1)*4];
h->mb.b_skip_mc = 1;
dir = -2;
omx = bmx;
omy = bmy;
- for( j=0; j<6; j++ )
+ for( int j = 0; j < 6; j++ )
{
COST_MV_SATD( omx + hex2[j+1][0], omy + hex2[j+1][1], satd, 1 );
COST_MV_RD ( omx + hex2[j+1][0], omy + hex2[j+1][1], satd, 1, j );
if( dir != -2 )
{
/* half hexagon, not overlapping the previous iteration */
- for( i = 1; i < 10; i++ )
+ for( int i = 1; i < 10; i++ )
{
const int odir = mod6m1[dir+1];
if( bmy < h->mb.mv_min_spel[1] + 3 ||
dir = -2;
omx = bmx;
omy = bmy;
- for( j=0; j<3; j++ )
+ for( int j = 0; j < 3; j++ )
{
COST_MV_SATD( omx + hex2[odir+j][0], omy + hex2[odir+j][1], satd, 1 );
COST_MV_RD ( omx + hex2[odir+j][0], omy + hex2[odir+j][1], satd, 1, odir-1+j );
/* square refine, same pattern as ME HEX. */
omx = bmx;
omy = bmy;
- for( i=0; i<8; i++ )
+ for( int i = 0; i < 8; i++ )
{
COST_MV_SATD( omx + square1[i+1][0], omy + square1[i+1][1], satd, 1 );
COST_MV_RD ( omx + square1[i+1][0], omy + square1[i+1][1], satd, 0, 0 );
projects / x264 / blobdiff