Support infinite keyint (--keyint infinite).

[x264] / common / pixel.c

diff --git a/common/pixel.c b/common/pixel.c
index e144292dde02f52e1af9779395a51403ace7fd8b..8441c7ae77f6d796a0543db4a1f4ec8248ae75d4 100644 (file)
--- a/common/pixel.c
+++ b/common/pixel.c
@@ -23,16 +23,16 @@
 
 #include "common.h"
 
-#ifdef HAVE_MMX
+#if HAVE_MMX
 #   include "x86/pixel.h"
 #endif
-#ifdef ARCH_PPC
+#if ARCH_PPC
 #   include "ppc/pixel.h"
 #endif
-#ifdef ARCH_ARM
+#if ARCH_ARM
 #   include "arm/pixel.h"
 #endif
-#ifdef ARCH_UltraSparc
+#if ARCH_UltraSparc
 #   include "sparc/pixel.h"
 #endif
 
@@ -406,7 +406,7 @@ SAD_X( 8x4 )
 SAD_X( 4x8 )
 SAD_X( 4x4 )
 
-#ifdef ARCH_UltraSparc
+#if ARCH_UltraSparc
 SAD_X( 16x16_vis )
 SAD_X( 16x8_vis )
 SAD_X( 8x16_vis )
@@ -444,17 +444,50 @@ SATD_X_DECL6( cpu )\
 SATD_X( 4x4, cpu )
 
 SATD_X_DECL7()
-#ifdef HAVE_MMX
+#if HAVE_MMX
 SATD_X_DECL7( _mmxext )
 SATD_X_DECL6( _sse2 )
 SATD_X_DECL7( _ssse3 )
 SATD_X_DECL7( _sse4 )
 #endif
 
-#ifdef HAVE_ARMV6
+#if HAVE_ARMV6
 SATD_X_DECL7( _neon )
 #endif
 
+#define INTRA_MBCMP_8x8( mbcmp )\
+void x264_intra_##mbcmp##_x3_8x8( pixel *fenc, pixel edge[33], int res[3] )\
+{\
+    pixel pix[8*FDEC_STRIDE];\
+    x264_predict_8x8_v_c( pix, edge );\
+    res[0] = x264_pixel_##mbcmp##_8x8( pix, FDEC_STRIDE, fenc, FENC_STRIDE );\
+    x264_predict_8x8_h_c( pix, edge );\
+    res[1] = x264_pixel_##mbcmp##_8x8( pix, FDEC_STRIDE, fenc, FENC_STRIDE );\
+    x264_predict_8x8_dc_c( pix, edge );\
+    res[2] = x264_pixel_##mbcmp##_8x8( pix, FDEC_STRIDE, fenc, FENC_STRIDE );\
+}
+
+INTRA_MBCMP_8x8(sad)
+INTRA_MBCMP_8x8(sa8d)
+
+#define INTRA_MBCMP( mbcmp, size, pred1, pred2, pred3, chroma )\
+void x264_intra_##mbcmp##_x3_##size##x##size##chroma( pixel *fenc, pixel *fdec, int res[3] )\
+{\
+    x264_predict_##size##x##size##chroma##_##pred1##_c( fdec );\
+    res[0] = x264_pixel_##mbcmp##_##size##x##size( fdec, FDEC_STRIDE, fenc, FENC_STRIDE );\
+    x264_predict_##size##x##size##chroma##_##pred2##_c( fdec );\
+    res[1] = x264_pixel_##mbcmp##_##size##x##size( fdec, FDEC_STRIDE, fenc, FENC_STRIDE );\
+    x264_predict_##size##x##size##chroma##_##pred3##_c( fdec );\
+    res[2] = x264_pixel_##mbcmp##_##size##x##size( fdec, FDEC_STRIDE, fenc, FENC_STRIDE );\
+}
+
+INTRA_MBCMP(sad, 4, v, h, dc, )
+INTRA_MBCMP(satd, 4, v, h, dc, )
+INTRA_MBCMP(sad, 8, dc, h, v, c )
+INTRA_MBCMP(satd, 8, dc, h, v, c )
+INTRA_MBCMP(sad, 16, v, h, dc, )
+INTRA_MBCMP(satd, 16, v, h, dc, )
+
 /****************************************************************************
  * structural similarity metric
  ****************************************************************************/
@@ -514,7 +547,7 @@ float x264_pixel_ssim_wxh( x264_pixel_function_t *pf,
     int z = 0;
     float ssim = 0.0;
     int (*sum0)[4] = buf;
-    int (*sum1)[4] = sum0 + width/4+3;
+    int (*sum1)[4] = sum0 + (width >> 2) + 3;
     width >>= 2;
     height >>= 2;
     for( int y = 1; y < height; y++ )
@@ -636,7 +669,16 @@ void x264_pixel_init( int cpu, x264_pixel_function_t *pixf )
     pixf->ssim_end4 = ssim_end4;
     pixf->var2_8x8 = pixel_var2_8x8;
 
-#ifdef HAVE_MMX
+    pixf->intra_sad_x3_4x4    = x264_intra_sad_x3_4x4;
+    pixf->intra_satd_x3_4x4   = x264_intra_satd_x3_4x4;
+    pixf->intra_sad_x3_8x8    = x264_intra_sad_x3_8x8;
+    pixf->intra_sa8d_x3_8x8   = x264_intra_sa8d_x3_8x8;
+    pixf->intra_sad_x3_8x8c   = x264_intra_sad_x3_8x8c;
+    pixf->intra_satd_x3_8x8c  = x264_intra_satd_x3_8x8c;
+    pixf->intra_sad_x3_16x16  = x264_intra_sad_x3_16x16;
+    pixf->intra_satd_x3_16x16 = x264_intra_satd_x3_16x16;
+
+#if HAVE_MMX
     if( cpu&X264_CPU_MMX )
     {
         INIT7( ssd, _mmx );
@@ -655,7 +697,7 @@ void x264_pixel_init( int cpu, x264_pixel_function_t *pixf )
         INIT_ADS( _mmxext );
         pixf->var[PIXEL_16x16] = x264_pixel_var_16x16_mmxext;
         pixf->var[PIXEL_8x8]   = x264_pixel_var_8x8_mmxext;
-#ifdef ARCH_X86
+#if ARCH_X86
         pixf->sa8d[PIXEL_16x16] = x264_pixel_sa8d_16x16_mmxext;
         pixf->sa8d[PIXEL_8x8]   = x264_pixel_sa8d_8x8_mmxext;
         pixf->intra_sa8d_x3_8x8 = x264_intra_sa8d_x3_8x8_mmxext;
@@ -704,7 +746,7 @@ void x264_pixel_init( int cpu, x264_pixel_function_t *pixf )
         pixf->ssim_end4        = x264_pixel_ssim_end4_sse2;
         pixf->sa8d[PIXEL_16x16] = x264_pixel_sa8d_16x16_sse2;
         pixf->sa8d[PIXEL_8x8]   = x264_pixel_sa8d_8x8_sse2;
-#ifdef ARCH_X86_64
+#if ARCH_X86_64
         pixf->intra_sa8d_x3_8x8 = x264_intra_sa8d_x3_8x8_sse2;
 #endif
         pixf->var2_8x8 = x264_pixel_var2_8x8_sse2;
@@ -728,7 +770,7 @@ void x264_pixel_init( int cpu, x264_pixel_function_t *pixf )
         if( cpu&X264_CPU_CACHELINE_64 )
         {
             INIT2( ssd, _sse2); /* faster for width 16 on p4 */
-#ifdef ARCH_X86
+#if ARCH_X86
             INIT2( sad, _cache64_sse2 );
             INIT2( sad_x3, _cache64_sse2 );
             INIT2( sad_x4, _cache64_sse2 );
@@ -787,7 +829,7 @@ void x264_pixel_init( int cpu, x264_pixel_function_t *pixf )
         pixf->intra_satd_x3_8x8c  = x264_intra_satd_x3_8x8c_ssse3;
         pixf->intra_sad_x3_8x8c   = x264_intra_sad_x3_8x8c_ssse3;
         pixf->intra_satd_x3_4x4   = x264_intra_satd_x3_4x4_ssse3;
-#ifdef ARCH_X86_64
+#if ARCH_X86_64
         pixf->intra_sa8d_x3_8x8 = x264_intra_sa8d_x3_8x8_ssse3;
 #endif
         pixf->var2_8x8 = x264_pixel_var2_8x8_ssse3;
@@ -814,10 +856,13 @@ void x264_pixel_init( int cpu, x264_pixel_function_t *pixf )
         }
         pixf->sa8d[PIXEL_16x16]= x264_pixel_sa8d_16x16_sse4;
         pixf->sa8d[PIXEL_8x8]  = x264_pixel_sa8d_8x8_sse4;
+        pixf->intra_sad_x3_4x4 = x264_intra_sad_x3_4x4_sse4;
+        /* Slower on Conroe, so only enable under SSE4 */
+        pixf->intra_sad_x3_8x8  = x264_intra_sad_x3_8x8_ssse3;
     }
 #endif //HAVE_MMX
 
-#ifdef HAVE_ARMV6
+#if HAVE_ARMV6
     if( cpu&X264_CPU_ARMV6 )
     {
         pixf->sad[PIXEL_4x8] = x264_pixel_sad_4x8_armv6;
@@ -858,13 +903,13 @@ void x264_pixel_init( int cpu, x264_pixel_function_t *pixf )
         }
     }
 #endif
-#ifdef HAVE_ALTIVEC
+#if HAVE_ALTIVEC
     if( cpu&X264_CPU_ALTIVEC )
     {
         x264_pixel_altivec_init( pixf );
     }
 #endif
-#ifdef ARCH_UltraSparc
+#if ARCH_UltraSparc
     INIT4( sad, _vis );
     INIT4( sad_x3, _vis );
     INIT4( sad_x4, _vis );