arm: do not fill mc_weight*_neon tabs for HIGH_BIT_DEPTH

[x264] / common / arm / mc-c.c

diff --git a/common/arm/mc-c.c b/common/arm/mc-c.c
index c31745fe054c6f31c496e2bd1f6b17c47ce4b83b..c3eec46e5696b62e6a99be266a558e348b61900a 100644 (file)
--- a/common/arm/mc-c.c
+++ b/common/arm/mc-c.c
@@ -1,9 +1,10 @@
 /*****************************************************************************
  * mc-c.c: arm motion compensation
  *****************************************************************************
- * Copyright (C) 2009-2014 x264 project
+ * Copyright (C) 2009-2015 x264 project
  *
  * Authors: David Conrad <lessen42@gmail.com>
+ *          Janne Grunau <janne-x264@jannau.net>
  *
  * This program is free software; you can redistribute it and/or modify
  * it under the terms of the GNU General Public License as published by
@@ -47,6 +48,8 @@ void x264_pixel_avg2_w8_neon ( uint8_t *, intptr_t, uint8_t *, intptr_t, uint8_t
 void x264_pixel_avg2_w16_neon( uint8_t *, intptr_t, uint8_t *, intptr_t, uint8_t *, int );
 void x264_pixel_avg2_w20_neon( uint8_t *, intptr_t, uint8_t *, intptr_t, uint8_t *, int );
 
+void x264_plane_copy_neon( pixel *dst, intptr_t i_dst,
+                           pixel *src, intptr_t i_src, int w, int h );
 void x264_plane_copy_deinterleave_neon(  pixel *dstu, intptr_t i_dstu,
                                          pixel *dstv, intptr_t i_dstv,
                                          pixel *src,  intptr_t i_src, int w, int h );
@@ -57,18 +60,21 @@ void x264_plane_copy_deinterleave_rgb_neon( pixel *dsta, intptr_t i_dsta,
 void x264_plane_copy_interleave_neon( pixel *dst,  intptr_t i_dst,
                                       pixel *srcu, intptr_t i_srcu,
                                       pixel *srcv, intptr_t i_srcv, int w, int h );
+void x264_plane_copy_swap_neon( pixel *dst, intptr_t i_dst,
+                                pixel *src, intptr_t i_src, int w, int h );
 
 void x264_store_interleave_chroma_neon( pixel *dst, intptr_t i_dst, pixel *srcu, pixel *srcv, int height );
 void x264_load_deinterleave_chroma_fdec_neon( pixel *dst, pixel *src, intptr_t i_src, int height );
 void x264_load_deinterleave_chroma_fenc_neon( pixel *dst, pixel *src, intptr_t i_src, int height );
 
+#if !HIGH_BIT_DEPTH
 #define MC_WEIGHT(func)\
 void x264_mc_weight_w20##func##_neon( uint8_t *, intptr_t, uint8_t *, intptr_t, const x264_weight_t *, int );\
 void x264_mc_weight_w16##func##_neon( uint8_t *, intptr_t, uint8_t *, intptr_t, const x264_weight_t *, int );\
 void x264_mc_weight_w8##func##_neon ( uint8_t *, intptr_t, uint8_t *, intptr_t, const x264_weight_t *, int );\
 void x264_mc_weight_w4##func##_neon ( uint8_t *, intptr_t, uint8_t *, intptr_t, const x264_weight_t *, int );\
 \
-static void (* const x264_mc##func##_wtab_neon[6])( uint8_t *, intptr_t, uint8_t *, intptr_t, const x264_weight_t *, int ) =\
+static weight_fn_t x264_mc##func##_wtab_neon[6] =\
 {\
     x264_mc_weight_w4##func##_neon,\
     x264_mc_weight_w4##func##_neon,\
@@ -82,19 +88,27 @@ MC_WEIGHT()
 MC_WEIGHT(_nodenom)
 MC_WEIGHT(_offsetadd)
 MC_WEIGHT(_offsetsub)
+#endif
 
 void x264_mc_copy_w4_neon ( uint8_t *, intptr_t, uint8_t *, intptr_t, int );
 void x264_mc_copy_w8_neon ( uint8_t *, intptr_t, uint8_t *, intptr_t, int );
 void x264_mc_copy_w16_neon( uint8_t *, intptr_t, uint8_t *, intptr_t, int );
 void x264_mc_copy_w16_aligned_neon( uint8_t *, intptr_t, uint8_t *, intptr_t, int );
 
-void x264_mc_chroma_neon( uint8_t *, intptr_t, uint8_t *, intptr_t, int, int, int, int );
+void x264_mc_chroma_neon( uint8_t *, uint8_t *, intptr_t, uint8_t *, intptr_t, int, int, int, int );
 void x264_frame_init_lowres_core_neon( uint8_t *, uint8_t *, uint8_t *, uint8_t *, uint8_t *, intptr_t, intptr_t, int, int );
 
 void x264_hpel_filter_v_neon( uint8_t *, uint8_t *, int16_t *, intptr_t, int );
 void x264_hpel_filter_c_neon( uint8_t *, int16_t *, int );
 void x264_hpel_filter_h_neon( uint8_t *, uint8_t *, int );
 
+void integral_init4h_neon( uint16_t *, uint8_t *, intptr_t );
+void integral_init4v_neon( uint16_t *, uint16_t *, intptr_t );
+void integral_init8h_neon( uint16_t *, uint8_t *, intptr_t );
+void integral_init8v_neon( uint16_t *, intptr_t );
+
+void x264_mbtree_propagate_cost_neon( int16_t *, uint16_t *, uint16_t *, uint16_t *, uint16_t *, float *, int );
+
 #if !HIGH_BIT_DEPTH
 static void x264_weight_cache_neon( x264_t *h, x264_weight_t *w )
 {
@@ -136,9 +150,6 @@ static void (* const x264_mc_copy_wtab_neon[5])( uint8_t *, intptr_t, uint8_t *,
     x264_mc_copy_w16_neon,
 };
 
-static const uint8_t hpel_ref0[16] = {0,1,1,1,0,1,1,1,2,3,3,3,0,1,1,1};
-static const uint8_t hpel_ref1[16] = {0,0,0,0,2,2,3,2,2,2,3,2,2,2,3,2};
-
 static void mc_luma_neon( uint8_t *dst,    intptr_t i_dst_stride,
                           uint8_t *src[4], intptr_t i_src_stride,
                           int mvx, int mvy,
@@ -146,13 +157,13 @@ static void mc_luma_neon( uint8_t *dst,    intptr_t i_dst_stride,
 {
     int qpel_idx = ((mvy&3)<<2) + (mvx&3);
     intptr_t offset = (mvy>>2)*i_src_stride + (mvx>>2);
-    uint8_t *src1 = src[hpel_ref0[qpel_idx]] + offset;
+    uint8_t *src1 = src[x264_hpel_ref0[qpel_idx]] + offset;
     if ( (mvy&3) == 3 )             // explict if() to force conditional add
         src1 += i_src_stride;
 
     if( qpel_idx & 5 ) /* qpel interpolation needed */
     {
-        uint8_t *src2 = src[hpel_ref1[qpel_idx]] + offset + ((mvx&3) == 3);
+        uint8_t *src2 = src[x264_hpel_ref1[qpel_idx]] + offset + ((mvx&3) == 3);
         x264_pixel_avg_wtab_neon[i_width>>2](
                 dst, i_dst_stride, src1, i_src_stride,
                 src2, i_height );
@@ -172,13 +183,13 @@ static uint8_t *get_ref_neon( uint8_t *dst,   intptr_t *i_dst_stride,
 {
     int qpel_idx = ((mvy&3)<<2) + (mvx&3);
     intptr_t offset = (mvy>>2)*i_src_stride + (mvx>>2);
-    uint8_t *src1 = src[hpel_ref0[qpel_idx]] + offset;
+    uint8_t *src1 = src[x264_hpel_ref0[qpel_idx]] + offset;
     if ( (mvy&3) == 3 )             // explict if() to force conditional add
         src1 += i_src_stride;
 
     if( qpel_idx & 5 ) /* qpel interpolation needed */
     {
-        uint8_t *src2 = src[hpel_ref1[qpel_idx]] + offset + ((mvx&3) == 3);
+        uint8_t *src2 = src[x264_hpel_ref1[qpel_idx]] + offset + ((mvx&3) == 3);
         x264_pixel_avg_wtab_neon[i_width>>2](
                 dst, *i_dst_stride, src1, i_src_stride,
                 src2, i_height );
@@ -220,6 +231,8 @@ static void hpel_filter_neon( uint8_t *dsth, uint8_t *dstv, uint8_t *dstc, uint8
 }
 #endif // !HIGH_BIT_DEPTH
 
+PROPAGATE_LIST(neon)
+
 void x264_mc_init_arm( int cpu, x264_mc_functions_t *pf )
 {
     if( !(cpu&X264_CPU_ARMV6) )
@@ -240,9 +253,11 @@ void x264_mc_init_arm( int cpu, x264_mc_functions_t *pf )
     pf->copy[PIXEL_8x8]   = x264_mc_copy_w8_neon;
     pf->copy[PIXEL_4x4]   = x264_mc_copy_w4_neon;
 
+    pf->plane_copy              = x264_plane_copy_neon;
     pf->plane_copy_deinterleave = x264_plane_copy_deinterleave_neon;
     pf->plane_copy_deinterleave_rgb = x264_plane_copy_deinterleave_rgb_neon;
     pf->plane_copy_interleave = x264_plane_copy_interleave_neon;
+    pf->plane_copy_swap = x264_plane_copy_swap_neon;
 
     pf->store_interleave_chroma = x264_store_interleave_chroma_neon;
     pf->load_deinterleave_chroma_fdec = x264_load_deinterleave_chroma_fdec_neon;
@@ -268,6 +283,14 @@ void x264_mc_init_arm( int cpu, x264_mc_functions_t *pf )
     pf->get_ref = get_ref_neon;
     pf->hpel_filter = hpel_filter_neon;
     pf->frame_init_lowres_core = x264_frame_init_lowres_core_neon;
+
+    pf->integral_init4h = integral_init4h_neon;
+    pf->integral_init8h = integral_init8h_neon;
+    pf->integral_init4v = integral_init4v_neon;
+    pf->integral_init8v = integral_init8v_neon;
+
+    pf->mbtree_propagate_cost = x264_mbtree_propagate_cost_neon;
+    pf->mbtree_propagate_list = x264_mbtree_propagate_list_neon;
 #endif // !HIGH_BIT_DEPTH
 
 // Apple's gcc stupidly cannot align stack variables, and ALIGNED_ARRAY can't work on structs