[x264-devel] commit: 15% faster refine_bidir_satd, 10% faster refine_bidir_rd ( or less with trellis=2) (Loren Merritt )
git version control
git at videolan.org
Tue Jul 28 16:34:34 CEST 2009
x264 | branch: master | Loren Merritt <pengvado at akuvian.org> | Tue Jul 28 07:16:31 2009 +0000| [0538c56a95cdc41a42a206079276a57d5d76b5a5] | committer: Loren Merritt
15% faster refine_bidir_satd, 10% faster refine_bidir_rd (or less with trellis=2)
re-roll a loop (saves 44KB code size, which is the cause of most of this speed gain)
don't re-mc mvs that haven't changed
> http://git.videolan.org/gitweb.cgi/x264.git/?a=commit;h=0538c56a95cdc41a42a206079276a57d5d76b5a5
---
encoder/me.c | 106 ++++++++++++++++++++++++++++++----------------------------
1 files changed, 55 insertions(+), 51 deletions(-)
diff --git a/encoder/me.c b/encoder/me.c
index 1ff347c..e5952bb 100644
--- a/encoder/me.c
+++ b/encoder/me.c
@@ -858,31 +858,34 @@ static void refine_subpel( x264_t *h, x264_me_t *m, int hpel_iters, int qpel_ite
m->cost_mv = p_cost_mvx[ bmx ] + p_cost_mvy[ bmy ];
}
-#define BIME_CACHE( dx, dy ) \
+#define BIME_CACHE( dx, dy, list ) \
{ \
+ x264_me_t *m = m##list;\
int i = 4 + 3*dx + dy; \
- int mvx0 = om0x+dx, mvy0 = om0y+dy;\
- int mvx1 = om1x+dx, mvy1 = om1y+dy;\
- stride0[i] = bw;\
- stride1[i] = bw;\
- src0[i] = h->mc.get_ref( pix0[i], &stride0[i], m0->p_fref, m0->i_stride[0], mvx0, mvy0, bw, bh ); \
- src1[i] = h->mc.get_ref( pix1[i], &stride1[i], m1->p_fref, m1->i_stride[0], mvx1, mvy1, bw, bh ); \
+ int mvx = om##list##x+dx;\
+ int mvy = om##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 ); \
if( rd )\
{\
- if( h->mb.b_interlaced & ref0 )\
- mvy0 += (h->mb.i_mb_y & 1)*4 - 2;\
- if( h->mb.b_interlaced & ref1 )\
- mvy1 += (h->mb.i_mb_y & 1)*4 - 2;\
- h->mc.mc_chroma( pixu0[i], 8, m0->p_fref[4], m0->i_stride[1], mvx0, mvy0, bw>>1, bh>>1 );\
- h->mc.mc_chroma( pixu1[i], 8, m1->p_fref[4], m1->i_stride[1], mvx1, mvy1, bw>>1, bh>>1 );\
- h->mc.mc_chroma( pixv0[i], 8, m0->p_fref[5], m0->i_stride[1], mvx0, mvy0, bw>>1, bh>>1 );\
- h->mc.mc_chroma( pixv1[i], 8, m1->p_fref[5], m1->i_stride[1], mvx1, mvy1, bw>>1, bh>>1 );\
+ if( h->mb.b_interlaced & ref##list )\
+ mvy += (h->mb.i_mb_y & 1)*4 - 2;\
+ h->mc.mc_chroma( pixu_buf[list][i], 8, m->p_fref[4], m->i_stride[1], mvx, mvy, bw>>1, bh>>1 );\
+ h->mc.mc_chroma( pixv_buf[list][i], 8, m->p_fref[5], m->i_stride[1], mvx, mvy, bw>>1, bh>>1 );\
}\
}
-#define BIME_CACHE2(a,b) \
- BIME_CACHE(a,b) \
- BIME_CACHE(-(a),-(b))
+#define BIME_CACHE2(a,b,list) \
+ BIME_CACHE(a,b,list) \
+ BIME_CACHE(-(a),-(b),list)
+
+#define BIME_CACHE8(list) \
+{\
+ BIME_CACHE2( 1, 0, list );\
+ BIME_CACHE2( 0, 1, list );\
+ BIME_CACHE2( 1, 1, list );\
+ BIME_CACHE2( 1,-1, list );\
+}
#define SATD_THRESH 17/16
@@ -906,8 +909,8 @@ if( pass == 0 || !((visited[(m0x)&7][(m0y)&7][(m1x)&7] & (1<<((m1y)&7)))) ) \
bcost = cost; \
*(uint32_t*)cache0_mv = *(uint32_t*)cache0_mv2 = pack16to32_mask(m0x,m0y); \
*(uint32_t*)cache1_mv = *(uint32_t*)cache1_mv2 = pack16to32_mask(m1x,m1y); \
- h->mc.avg[i_pixel+3]( pixu, FDEC_STRIDE, pixu0[i0], 8, pixu1[i1], 8, i_weight );\
- h->mc.avg[i_pixel+3]( pixv, FDEC_STRIDE, pixv0[i0], 8, pixv1[i1], 8, i_weight );\
+ h->mc.avg[i_pixel+3]( pixu, FDEC_STRIDE, pixu_buf[0][i0], 8, pixu_buf[1][i1], 8, i_weight );\
+ h->mc.avg[i_pixel+3]( pixv, FDEC_STRIDE, pixv_buf[0][i0], 8, pixv_buf[1][i1], 8, i_weight );\
costrd = x264_rd_cost_part( h, i_lambda2, i8*4, m0->i_pixel ); \
if( costrd < bcostrd ) \
{\
@@ -932,16 +935,6 @@ if( pass == 0 || !((visited[(m0x)&7][(m0y)&7][(m1x)&7] & (1<<((m1y)&7)))) ) \
#define CHECK_BIDIR(a,b,c,d) \
COST_BIMV_SATD(om0x+a, om0y+b, om1x+c, om1y+d)
-#define CHECK_BIDIR2(a,b,c,d) \
- CHECK_BIDIR(a,b,c,d) \
- CHECK_BIDIR(-(a),-(b),-(c),-(d))
-
-#define CHECK_BIDIR8(a,b,c,d) \
- CHECK_BIDIR2(a,b,c,d) \
- CHECK_BIDIR2(b,c,d,a) \
- CHECK_BIDIR2(c,d,a,b) \
- CHECK_BIDIR2(d,a,b,c)
-
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 )
{
static const int pixel_mv_offs[] = { 0, 4, 4*8, 0 };
@@ -956,12 +949,9 @@ static void ALWAYS_INLINE x264_me_refine_bidir( x264_t *h, x264_me_t *m0, x264_m
const int16_t *p_cost_m0y = m0->p_cost_mv - m0->mvp[1];
const int16_t *p_cost_m1x = m1->p_cost_mv - m1->mvp[0];
const int16_t *p_cost_m1y = m1->p_cost_mv - m1->mvp[1];
- DECLARE_ALIGNED_16( uint8_t pix0[9][16*16] );
- DECLARE_ALIGNED_16( uint8_t pix1[9][16*16] );
- DECLARE_ALIGNED_8( uint8_t pixu0[9][8*8] );
- DECLARE_ALIGNED_8( uint8_t pixu1[9][8*8] );
- DECLARE_ALIGNED_8( uint8_t pixv0[9][8*8] );
- DECLARE_ALIGNED_8( uint8_t pixv1[9][8*8] );
+ DECLARE_ALIGNED_16( uint8_t pixy_buf[2][9][16*16] );
+ DECLARE_ALIGNED_8( uint8_t pixu_buf[2][9][8*8] );
+ DECLARE_ALIGNED_8( uint8_t pixv_buf[2][9][8*8] );
uint8_t *src0[9];
uint8_t *src1[9];
uint8_t *pix = &h->mb.pic.p_fdec[0][(i8>>1)*8*FDEC_STRIDE+(i8&1)*8];
@@ -977,10 +967,22 @@ static void ALWAYS_INLINE x264_me_refine_bidir( x264_t *h, x264_me_t *m0, x264_m
int bm1y = m1->mv[1], om1y = bm1y;
int bcost = COST_MAX;
int pass = 0;
+ int j;
+ int mc_list0 = 1, mc_list1 = 1;
uint64_t bcostrd = COST_MAX64;
-
/* each byte of visited represents 8 possible m1y positions, so a 4D array isn't needed */
DECLARE_ALIGNED_16( uint8_t visited[8][8][8] );
+ /* all permutations of an offset in up to 2 of the dimensions */
+ static const int8_t dia4d[32][4] = {
+ {0,0,0,1}, {0,0,0,-1}, {0,0,1,0}, {0,0,-1,0},
+ {0,1,0,0}, {0,-1,0,0}, {1,0,0,0}, {-1,0,0,0},
+ {0,0,1,1}, {0,0,-1,-1},{0,1,1,0}, {0,-1,-1,0},
+ {1,1,0,0}, {-1,-1,0,0},{1,0,0,1}, {-1,0,0,-1},
+ {0,1,0,1}, {0,-1,0,-1},{1,0,1,0}, {-1,0,-1,0},
+ {0,0,-1,1},{0,0,1,-1}, {0,-1,1,0},{0,1,-1,0},
+ {-1,1,0,0},{1,-1,0,0}, {1,0,0,-1},{-1,0,0,1},
+ {0,-1,0,1},{0,1,0,-1}, {-1,0,1,0},{1,0,-1,0},
+ };
if( bm0y < h->mb.mv_min_spel[1] + 8 || bm1y < h->mb.mv_min_spel[1] + 8 ||
bm0y > h->mb.mv_max_spel[1] - 8 || bm1y > h->mb.mv_max_spel[1] - 8 )
@@ -988,7 +990,8 @@ static void ALWAYS_INLINE x264_me_refine_bidir( x264_t *h, x264_me_t *m0, x264_m
h->mc.memzero_aligned( visited, sizeof(visited) );
- BIME_CACHE( 0, 0 );
+ BIME_CACHE( 0, 0, 0 );
+ BIME_CACHE( 0, 0, 1 );
CHECK_BIDIR( 0, 0, 0, 0 );
for( pass = 0; pass < 8; pass++ )
@@ -997,27 +1000,28 @@ static void ALWAYS_INLINE x264_me_refine_bidir( x264_t *h, x264_me_t *m0, x264_m
/* doesn't do chroma ME. this probably doesn't matter, as the gains
* from bidir ME are the same with and without chroma ME. */
- BIME_CACHE2( 1, 0 );
- BIME_CACHE2( 0, 1 );
- BIME_CACHE2( 1, 1 );
- BIME_CACHE2( 1,-1 );
+ if( mc_list0 )
+ BIME_CACHE8( 0 );
+ if( mc_list1 )
+ BIME_CACHE8( 1 );
- CHECK_BIDIR8( 0, 0, 0, 1 );
- CHECK_BIDIR8( 0, 0, 1, 1 );
- CHECK_BIDIR2( 0, 1, 0, 1 );
- CHECK_BIDIR2( 1, 0, 1, 0 );
- CHECK_BIDIR8( 0, 0,-1, 1 );
- CHECK_BIDIR2( 0,-1, 0, 1 );
- CHECK_BIDIR2(-1, 0, 1, 0 );
+ for( j=0; j<32; j++ )
+ CHECK_BIDIR( dia4d[j][0], dia4d[j][1], dia4d[j][2], dia4d[j][3] );
- if( om0x == bm0x && om0y == bm0y && om1x == bm1x && om1y == bm1y )
+ mc_list0 = (om0x-bm0x)|(om0y-bm0y);
+ mc_list1 = (om1x-bm1x)|(om1y-bm1y);
+ if( !mc_list0 && !mc_list1 )
break;
om0x = bm0x;
om0y = bm0y;
om1x = bm1x;
om1y = bm1y;
- BIME_CACHE( 0, 0 );
+
+ if( mc_list0 )
+ BIME_CACHE( 0, 0, 0 );
+ if( mc_list1 )
+ BIME_CACHE( 0, 0, 1 );
}
m0->mv[0] = bm0x;
More information about the x264-devel
mailing list