memcpy_aligned_sse2

[x264] / encoder / rdo.c

diff --git a/encoder/rdo.c b/encoder/rdo.c
index b204639f5d3cbee6cdb1ef07b73317fdad3d2bfe..e5c44649fd0b21f978bcb37629ce97eb22a84d32 100644 (file)
--- a/encoder/rdo.c
+++ b/encoder/rdo.c
@@ -42,9 +42,9 @@ static int cabac_prefix_size[15][128];
 /* CABAC: not exactly the same. x264_cabac_size_decision() keeps track of
  * fractional bits, but only finite precision. */
 #define x264_cabac_encode_decision(c,x,v) x264_cabac_size_decision(c,x,v)
-#define x264_cabac_encode_terminal(c,v)   x264_cabac_size_decision(c,276,v)
+#define x264_cabac_encode_terminal(c)     x264_cabac_size_decision(c,276,0)
 #define x264_cabac_encode_bypass(c,v)     ((c)->f8_bits_encoded += 256)
-#define x264_cabac_encode_flush(c)
+#define x264_cabac_encode_flush(h,c)
 #define x264_macroblock_write_cabac  x264_macroblock_size_cabac
 #define x264_cabac_mb_skip  x264_cabac_mb_size_skip_unused
 #include "cabac.c"
@@ -82,8 +82,8 @@ static int x264_rd_cost_mb( x264_t *h, int i_lambda2 )
     }
     else if( h->param.b_cabac )
     {
-        x264_cabac_t cabac_tmp = h->cabac;
-        cabac_tmp.f8_bits_encoded = 0;
+        x264_cabac_t cabac_tmp;
+        h->mc.memcpy_aligned( &cabac_tmp, &h->cabac, offsetof(x264_cabac_t,i_low) );
         x264_macroblock_size_cabac( h, &cabac_tmp );
         i_bits = ( cabac_tmp.f8_bits_encoded * i_lambda2 + 128 ) >> 8;
     }
@@ -124,8 +124,8 @@ int x264_rd_cost_part( x264_t *h, int i_lambda2, int i8, int i_pixel )
 
     if( h->param.b_cabac )
     {
-        x264_cabac_t cabac_tmp = h->cabac;
-        cabac_tmp.f8_bits_encoded = 0;
+        x264_cabac_t cabac_tmp;
+        h->mc.memcpy_aligned( &cabac_tmp, &h->cabac, offsetof(x264_cabac_t,i_low) );
         x264_partition_size_cabac( h, &cabac_tmp, i8, i_pixel );
         i_bits = ( cabac_tmp.f8_bits_encoded * i_lambda2 + 128 ) >> 8;
     }
@@ -146,8 +146,8 @@ int x264_rd_cost_i8x8( x264_t *h, int i_lambda2, int i8, int i_mode )
 
     if( h->param.b_cabac )
     {
-        x264_cabac_t cabac_tmp = h->cabac;
-        cabac_tmp.f8_bits_encoded = 0;
+        x264_cabac_t cabac_tmp;
+        h->mc.memcpy_aligned( &cabac_tmp, &h->cabac, offsetof(x264_cabac_t,i_low) );
         x264_partition_i8x8_size_cabac( h, &cabac_tmp, i8, i_mode );
         i_bits = ( cabac_tmp.f8_bits_encoded * i_lambda2 + 128 ) >> 8;
     }
@@ -168,8 +168,9 @@ int x264_rd_cost_i4x4( x264_t *h, int i_lambda2, int i4, int i_mode )
 
     if( h->param.b_cabac )
     {
-        x264_cabac_t cabac_tmp = h->cabac;
-        cabac_tmp.f8_bits_encoded = 0;
+        x264_cabac_t cabac_tmp;
+        h->mc.memcpy_aligned( &cabac_tmp, &h->cabac, offsetof(x264_cabac_t,i_low) );
+        
         x264_partition_i4x4_size_cabac( h, &cabac_tmp, i4, i_mode );
         i_bits = ( cabac_tmp.f8_bits_encoded * i_lambda2 + 128 ) >> 8;
     }
@@ -194,8 +195,8 @@ int x264_rd_cost_i8x8_chroma( x264_t *h, int i_lambda2, int i_mode, int b_dct )
 
     if( h->param.b_cabac )
     {
-        x264_cabac_t cabac_tmp = h->cabac;
-        cabac_tmp.f8_bits_encoded = 0;
+        x264_cabac_t cabac_tmp;
+        h->mc.memcpy_aligned( &cabac_tmp, &h->cabac, offsetof(x264_cabac_t,i_low) );
         x264_i8x8_chroma_size_cabac( h, &cabac_tmp );
         i_bits = ( cabac_tmp.f8_bits_encoded * i_lambda2 + 128 ) >> 8;
     }
@@ -307,7 +308,7 @@ typedef struct {
 
 static void quant_trellis_cabac( x264_t *h, int16_t *dct,
                                  const uint16_t *quant_mf, const int *unquant_mf,
-                                 const int *coef_weight, const int *zigzag,
+                                 const int *coef_weight, const uint8_t *zigzag,
                                  int i_ctxBlockCat, int i_lambda2, int b_ac, int i_coefs )
 {
     int abs_coefs[64], signs[64];
@@ -397,8 +398,8 @@ static void quant_trellis_cabac( x264_t *h, int16_t *dct,
         {
             // no need to calculate ssd of 0s: it's the same in all nodes.
             // no need to modify level_tree for ctx=0: it starts with an infinite loop of 0s.
-            const int cost_sig0 = x264_cabac_size_decision_noup( &cabac_state_sig[i], 0 )
-                                * i_lambda2 >> ( CABAC_SIZE_BITS - LAMBDA_BITS );
+            const uint32_t cost_sig0 = x264_cabac_size_decision_noup( &cabac_state_sig[i], 0 )
+                                     * (uint64_t)i_lambda2 >> ( CABAC_SIZE_BITS - LAMBDA_BITS );
             for( j = 1; j < 8; j++ )
             {
                 if( nodes_cur[j].score != TRELLIS_SCORE_MAX )