Spare a vec_perm and a vec_mergeh though using a LUT of permutation vectors.

[x264] / encoder / encoder.c

diff --git a/encoder/encoder.c b/encoder/encoder.c
index 73d33f4a61c7de2ffdf0fc7788a2e9187795dba7..6aa69f12d7381febf791ae54d4560477c9a0371b 100644 (file)
--- a/encoder/encoder.c
+++ b/encoder/encoder.c
@@ -411,6 +411,7 @@ static int x264_validate_parameters( x264_t *h )
         h->param.analyse.b_fast_pskip = 0;
         h->param.analyse.i_noise_reduction = 0;
         h->param.analyse.f_psy_rd = 0;
+        h->param.i_bframe = 0;
         /* 8x8dct is not useful at all in CAVLC lossless */
         if( !h->param.b_cabac )
             h->param.analyse.b_transform_8x8 = 0;
@@ -548,8 +549,6 @@ static int x264_validate_parameters( x264_t *h )
             h->param.analyse.i_mv_range = l->mv_range >> h->param.b_interlaced;
         else
             h->param.analyse.i_mv_range = x264_clip3(h->param.analyse.i_mv_range, 32, 512 >> h->param.b_interlaced);
-        if( h->param.analyse.i_direct_8x8_inference < 0 )
-            h->param.analyse.i_direct_8x8_inference = l->direct8x8;
     }
 
     if( h->param.i_threads > 1 )
@@ -594,7 +593,6 @@ static int x264_validate_parameters( x264_t *h )
     BOOLIFY( b_deblocking_filter );
     BOOLIFY( b_interlaced );
     BOOLIFY( analyse.b_transform_8x8 );
-    BOOLIFY( analyse.i_direct_8x8_inference );
     BOOLIFY( analyse.b_chroma_me );
     BOOLIFY( analyse.b_fast_pskip );
     BOOLIFY( rc.b_stat_write );
@@ -713,6 +711,7 @@ x264_t *x264_encoder_open   ( x264_param_t *param )
           || h->param.i_bframe_adaptive
           || h->param.b_pre_scenecut );
     h->frames.b_have_lowres |= (h->param.rc.b_stat_read && h->param.rc.i_vbv_buffer_size > 0);
+    h->frames.b_have_sub8x8_esa = !!(h->param.analyse.inter & X264_ANALYSE_PSUB8x8);
 
     h->frames.i_last_idr = - h->param.i_keyint_max;
     h->frames.i_input    = 0;
@@ -728,7 +727,7 @@ x264_t *x264_encoder_open   ( x264_param_t *param )
     /* init CPU functions */
     x264_predict_16x16_init( h->param.cpu, h->predict_16x16 );
     x264_predict_8x8c_init( h->param.cpu, h->predict_8x8c );
-    x264_predict_8x8_init( h->param.cpu, h->predict_8x8 );
+    x264_predict_8x8_init( h->param.cpu, h->predict_8x8, &h->predict_8x8_filter );
     x264_predict_4x4_init( h->param.cpu, h->predict_4x4 );
     if( !h->param.b_cabac );
         x264_init_vlc_tables();
@@ -824,7 +823,9 @@ int x264_encoder_reconfig( x264_t *h, x264_param_t *param )
     COPY( analyse.intra );
     COPY( analyse.inter );
     COPY( analyse.i_direct_mv_pred );
-    COPY( analyse.i_me_range );
+    /* Scratch buffer prevents me_range from being increased for esa/tesa */
+    if( h->param.analyse.i_me_method < X264_ME_ESA || param->analyse.i_me_range < h->param.analyse.i_me_range )
+        COPY( analyse.i_me_range );
     COPY( analyse.i_noise_reduction );
     /* We can't switch out of subme=0 during encoding. */
     if( h->param.analyse.i_subpel_refine )
@@ -839,6 +840,8 @@ int x264_encoder_reconfig( x264_t *h, x264_param_t *param )
     // can only twiddle these if they were enabled to begin with:
     if( h->param.analyse.i_me_method >= X264_ME_ESA || param->analyse.i_me_method < X264_ME_ESA )
         COPY( analyse.i_me_method );
+    if( h->param.analyse.i_me_method >= X264_ME_ESA && !h->frames.b_have_sub8x8_esa )
+        h->param.analyse.inter &= ~X264_ANALYSE_PSUB8x8;
     if( h->pps->b_transform_8x8_mode )
         COPY( analyse.b_transform_8x8 );
     if( h->frames.i_max_ref1 > 1 )
@@ -1046,7 +1049,7 @@ static void x264_fdec_filter_row( x264_t *h, int mb_y )
             x264_pixel_ssim_wxh( &h->pixf,
                 h->fdec->plane[0] + 2+min_y*h->fdec->i_stride[0], h->fdec->i_stride[0],
                 h->fenc->plane[0] + 2+min_y*h->fenc->i_stride[0], h->fenc->i_stride[0],
-                h->param.i_width-2, max_y-min_y );
+                h->param.i_width-2, max_y-min_y, h->scratch_buffer );
     }
 }
 
@@ -1433,7 +1436,7 @@ int     x264_encoder_encode( x264_t *h,
             return 0;
         }
 
-        x264_slicetype_decide( h );
+        x264_stack_align( x264_slicetype_decide, h );
 
         /* 3: move some B-frames and 1 non-B to encode queue */
         while( IS_X264_TYPE_B( h->frames.next[bframes]->i_type ) )
@@ -1976,8 +1979,8 @@ void    x264_encoder_close  ( x264_t *h )
         for( i = 0; i < X264_PARTTYPE_MAX; i++ )
             for( j = 0; j < 2; j++ )
             {
-                int l0 = x264_mb_type_list0_table[i][j];
-                int l1 = x264_mb_type_list1_table[i][j];
+                int l0 = x264_mb_type_list_table[i][0][j];
+                int l1 = x264_mb_type_list_table[i][1][j];
                 if( l0 || l1 )
                     list_count[l1+l0*l1] += h->stat.i_mb_count[SLICE_TYPE_B][i] * 2;
             }