Predict 4x4_DC asm

Also remove 5-year-old unnecessary #define that reduced speed unnecessarily under MSVC-compiled builds

diff --git a/common/predict.c b/common/predict.c
index e68809944ba00d8813471a5fb8828d14fc4e6ed6..282ae807573cf995293d5f250a8a052def1b7ecd 100644 (file)
--- a/common/predict.c
+++ b/common/predict.c
@@ -27,9 +27,6 @@
 
 #include "common.h"
 
-#ifdef _MSC_VER
-#undef HAVE_MMX  /* not finished now */
-#endif
 #ifdef HAVE_MMX
 #   include "x86/predict.h"
 #endif

diff --git a/common/x86/predict-a.asm b/common/x86/predict-a.asm
index 708859c4aefe250d70bc7d51ca863bb2045080ca..0f64ca214d359eb1036b3950fdcf3741d5ea7406 100644 (file)
--- a/common/x86/predict-a.asm
+++ b/common/x86/predict-a.asm
@@ -145,6 +145,31 @@ cglobal predict_4x4_vl_mmxext, 1,1
 
     RET
 
+;-----------------------------------------------------------------------------
+; void predict_4x4_dc( uint8_t *src )
+;-----------------------------------------------------------------------------
+
+cglobal predict_4x4_dc_mmxext, 1,4
+    pxor   mm7, mm7
+    movd   mm0, [r0-FDEC_STRIDE]
+    psadbw mm0, mm7
+    movd   r3d, mm0
+    movzx  r1d, byte [r0-1]
+%assign n 1
+%rep 3
+    movzx  r2d, byte [r0+FDEC_STRIDE*n-1]
+    add    r1d, r2d
+%assign n n+1
+%endrep
+    lea    r1d, [r1+r3+4]
+    shr    r1d, 3
+    imul   r1d, 0x01010101
+    mov   [r0+FDEC_STRIDE*0], r1d
+    mov   [r0+FDEC_STRIDE*1], r1d
+    mov   [r0+FDEC_STRIDE*2], r1d
+    mov   [r0+FDEC_STRIDE*3], r1d
+    RET
+
 ;-----------------------------------------------------------------------------
 ; void predict_8x8_v_mmxext( uint8_t *src, uint8_t *edge )
 ;-----------------------------------------------------------------------------

diff --git a/common/x86/predict-c.c b/common/x86/predict-c.c
index d70c25a5d354e8a528a9193422b633527b4db326..34a98d6d6116eb3003be1ca04ee3a998408d050c 100644 (file)
--- a/common/x86/predict-c.c
+++ b/common/x86/predict-c.c
@@ -49,6 +49,7 @@ extern void predict_8x8_vl_sse2( uint8_t *src, uint8_t edge[33] );
 extern void predict_8x8_vr_core_mmxext( uint8_t *src, uint8_t edge[33] );
 extern void predict_4x4_ddl_mmxext( uint8_t *src );
 extern void predict_4x4_vl_mmxext( uint8_t *src );
+extern void predict_4x4_dc_mmxext( uint8_t *src );
 extern void predict_16x16_dc_top_sse2( uint8_t *src );
 extern void predict_16x16_dc_core_sse2( uint8_t *src, int i_dc_left );
 extern void predict_16x16_v_sse2( uint8_t *src );
@@ -555,4 +556,5 @@ void x264_predict_4x4_init_mmx( int cpu, x264_predict_t pf[12] )
         return;
     pf[I_PRED_4x4_DDL] = predict_4x4_ddl_mmxext;
     pf[I_PRED_4x4_VL]  = predict_4x4_vl_mmxext;
+    pf[I_PRED_4x4_DC]  = predict_4x4_dc_mmxext;
 }