Switch to exponential interpolation between presets.

[x264] / common / pixel.c

diff --git a/common/pixel.c b/common/pixel.c
index bb1894a05b5284d3e15f5ce9d970a8698d6e443d..bb59152ead3d53169505ce452619286b4e7225a1 100644 (file)
--- a/common/pixel.c
+++ b/common/pixel.c
@@ -1,7 +1,7 @@
 /*****************************************************************************
  * pixel.c: pixel metrics
  *****************************************************************************
- * Copyright (C) 2003-2014 x264 project
+ * Copyright (C) 2003-2016 x264 project
  *
  * Authors: Loren Merritt <lorenm@u.washington.edu>
  *          Laurent Aimar <fenrir@via.ecp.fr>
@@ -42,6 +42,9 @@
 #   include "aarch64/pixel.h"
 #   include "aarch64/predict.h"
 #endif
+#if ARCH_MIPS
+#   include "mips/pixel.h"
+#endif
 
 
 /****************************************************************************
@@ -598,8 +601,8 @@ INTRA_MBCMP( sad,  4x4,   v, h, dc,  , _neon, _neon )
 INTRA_MBCMP(satd,  4x4,   v, h, dc,  , _neon, _neon )
 INTRA_MBCMP( sad,  8x8,  dc, h,  v, c, _neon, _neon )
 INTRA_MBCMP(satd,  8x8,  dc, h,  v, c, _neon, _neon )
-INTRA_MBCMP( sad,  8x16, dc, h,  v, c, _neon, _c )
-INTRA_MBCMP(satd,  8x16, dc, h,  v, c, _neon, _c )
+INTRA_MBCMP( sad,  8x16, dc, h,  v, c, _neon, _neon )
+INTRA_MBCMP(satd,  8x16, dc, h,  v, c, _neon, _neon )
 INTRA_MBCMP( sad, 16x16,  v, h, dc,  , _neon, _neon )
 INTRA_MBCMP(satd, 16x16,  v, h, dc,  , _neon, _neon )
 #endif
@@ -1372,11 +1375,14 @@ void x264_pixel_init( int cpu, x264_pixel_function_t *pixf )
         INIT4( hadamard_ac, _neon );
         pixf->sa8d[PIXEL_8x8]   = x264_pixel_sa8d_8x8_neon;
         pixf->sa8d[PIXEL_16x16] = x264_pixel_sa8d_16x16_neon;
+        pixf->sa8d_satd[PIXEL_16x16] = x264_pixel_sa8d_satd_16x16_neon;
         pixf->var[PIXEL_8x8]    = x264_pixel_var_8x8_neon;
         pixf->var[PIXEL_8x16]   = x264_pixel_var_8x16_neon;
         pixf->var[PIXEL_16x16]  = x264_pixel_var_16x16_neon;
         pixf->var2[PIXEL_8x8]   = x264_pixel_var2_8x8_neon;
         pixf->var2[PIXEL_8x16]  = x264_pixel_var2_8x16_neon;
+        pixf->vsad = x264_pixel_vsad_neon;
+        pixf->asd8 = x264_pixel_asd8_neon;
 
         pixf->intra_sad_x3_4x4    = x264_intra_sad_x3_4x4_neon;
         pixf->intra_satd_x3_4x4   = x264_intra_satd_x3_4x4_neon;
@@ -1389,6 +1395,7 @@ void x264_pixel_init( int cpu, x264_pixel_function_t *pixf )
         pixf->intra_sad_x3_16x16  = x264_intra_sad_x3_16x16_neon;
         pixf->intra_satd_x3_16x16 = x264_intra_satd_x3_16x16_neon;
 
+        pixf->ssd_nv12_core     = x264_pixel_ssd_nv12_core_neon;
         pixf->ssim_4x4x2_core   = x264_pixel_ssim_4x4x2_core_neon;
         pixf->ssim_end4         = x264_pixel_ssim_end4_neon;
 
@@ -1409,13 +1416,13 @@ void x264_pixel_init( int cpu, x264_pixel_function_t *pixf )
 #if ARCH_AARCH64
     if( cpu&X264_CPU_NEON )
     {
-        INIT7( sad, _neon );
+        INIT8( sad, _neon );
         // AArch64 has no distinct instructions for aligned load/store
-        INIT7_NAME( sad_aligned, sad, _neon );
+        INIT8_NAME( sad_aligned, sad, _neon );
         INIT7( sad_x3, _neon );
         INIT7( sad_x4, _neon );
-        INIT7( ssd, _neon );
-        INIT7( satd, _neon );
+        INIT8( ssd, _neon );
+        INIT8( satd, _neon );
         INIT7( satd_x3, _neon );
         INIT7( satd_x4, _neon );
         INIT4( hadamard_ac, _neon );
@@ -1430,6 +1437,7 @@ void x264_pixel_init( int cpu, x264_pixel_function_t *pixf )
         pixf->var2[PIXEL_8x8]   = x264_pixel_var2_8x8_neon;
         pixf->var2[PIXEL_8x16]  = x264_pixel_var2_8x16_neon;
         pixf->vsad = x264_pixel_vsad_neon;
+        pixf->asd8 = x264_pixel_asd8_neon;
 
         pixf->intra_sad_x3_4x4    = x264_intra_sad_x3_4x4_neon;
         pixf->intra_satd_x3_4x4   = x264_intra_satd_x3_4x4_neon;
@@ -1448,6 +1456,38 @@ void x264_pixel_init( int cpu, x264_pixel_function_t *pixf )
     }
 #endif // ARCH_AARCH64
 
+#if HAVE_MSA
+    if( cpu&X264_CPU_MSA )
+    {
+        INIT8( sad, _msa );
+        INIT8_NAME( sad_aligned, sad, _msa );
+        INIT8( ssd, _msa );
+        INIT7( sad_x3, _msa );
+        INIT7( sad_x4, _msa );
+        INIT8( satd, _msa );
+        INIT4( hadamard_ac, _msa );
+
+        pixf->intra_sad_x3_4x4   = x264_intra_sad_x3_4x4_msa;
+        pixf->intra_sad_x3_8x8   = x264_intra_sad_x3_8x8_msa;
+        pixf->intra_sad_x3_8x8c  = x264_intra_sad_x3_8x8c_msa;
+        pixf->intra_sad_x3_16x16 = x264_intra_sad_x3_16x16_msa;
+        pixf->intra_satd_x3_4x4   = x264_intra_satd_x3_4x4_msa;
+        pixf->intra_satd_x3_16x16 = x264_intra_satd_x3_16x16_msa;
+        pixf->intra_satd_x3_8x8c  = x264_intra_satd_x3_8x8c_msa;
+        pixf->intra_sa8d_x3_8x8   = x264_intra_sa8d_x3_8x8_msa;
+
+        pixf->ssim_4x4x2_core = x264_ssim_4x4x2_core_msa;
+
+        pixf->var[PIXEL_16x16] = x264_pixel_var_16x16_msa;
+        pixf->var[PIXEL_8x16]  = x264_pixel_var_8x16_msa;
+        pixf->var[PIXEL_8x8]   = x264_pixel_var_8x8_msa;
+        pixf->var2[PIXEL_8x16]  = x264_pixel_var2_8x16_msa;
+        pixf->var2[PIXEL_8x8]   = x264_pixel_var2_8x8_msa;
+        pixf->sa8d[PIXEL_16x16] = x264_pixel_sa8d_16x16;
+        pixf->sa8d[PIXEL_8x8]   = x264_pixel_sa8d_8x8;
+    }
+#endif // HAVE_MSA
+
 #endif // HIGH_BIT_DEPTH
 #if HAVE_ALTIVEC
     if( cpu&X264_CPU_ALTIVEC )