Add Neon implementation of vpx_highbd_sad_skip_<w>x<h> functions

Add Neon implementations of high bitdepth downsampling SAD functions
for all block sizes.

Also add corresponding unit tests.

Change-Id: I56ea656e9bb5f8b2aedfdc4637c9ab4e1951b31b

2 files changed, 43 insertions, 10 deletions

diff --git a/vpx_dsp/arm/highbd_sad_neon.c b/vpx_dsp/arm/highbd_sad_neon.c
index 813710040..b99bac66c 100644
--- a/vpx_dsp/arm/highbd_sad_neon.c
+++ b/vpx_dsp/arm/highbd_sad_neon.c
@@ -179,6 +179,36 @@ HBD_SAD_WXH_NEON(32, 64)
 HBD_SAD_WXH_NEON(64, 32)
 HBD_SAD_WXH_NEON(64, 64)
 
+#undef HBD_SAD_WXH_NEON
+
+#define HBD_SAD_SKIP_WXH_NEON(w, h)                             \
+  unsigned int vpx_highbd_sad_skip_##w##x##h##_neon(            \
+      const uint8_t *src, int src_stride, const uint8_t *ref,   \
+      int ref_stride) {                                         \
+    return 2 * highbd_sad##w##xh_neon(src, 2 * src_stride, ref, \
+                                      2 * ref_stride, (h) / 2); \
+  }
+
+HBD_SAD_SKIP_WXH_NEON(4, 4)
+HBD_SAD_SKIP_WXH_NEON(4, 8)
+
+HBD_SAD_SKIP_WXH_NEON(8, 4)
+HBD_SAD_SKIP_WXH_NEON(8, 8)
+HBD_SAD_SKIP_WXH_NEON(8, 16)
+
+HBD_SAD_SKIP_WXH_NEON(16, 8)
+HBD_SAD_SKIP_WXH_NEON(16, 16)
+HBD_SAD_SKIP_WXH_NEON(16, 32)
+
+HBD_SAD_SKIP_WXH_NEON(32, 16)
+HBD_SAD_SKIP_WXH_NEON(32, 32)
+HBD_SAD_SKIP_WXH_NEON(32, 64)
+
+HBD_SAD_SKIP_WXH_NEON(64, 32)
+HBD_SAD_SKIP_WXH_NEON(64, 64)
+
+#undef HBD_SAD_SKIP_WXH_NEON
+
 static INLINE uint32_t highbd_sad4xh_avg_neon(const uint8_t *src_ptr,
                                               int src_stride,
                                               const uint8_t *ref_ptr,
diff --git a/vpx_dsp/vpx_dsp_rtcd_defs.pl b/vpx_dsp/vpx_dsp_rtcd_defs.pl
index 05d031b8c..7bea73895 100644
--- a/vpx_dsp/vpx_dsp_rtcd_defs.pl
+++ b/vpx_dsp/vpx_dsp_rtcd_defs.pl
@@ -1068,40 +1068,43 @@ if (vpx_config("CONFIG_VP9_HIGHBITDEPTH") eq "yes") {
   specialize qw/vpx_highbd_sad4x4 neon/;
 
   add_proto qw/unsigned int vpx_highbd_sad_skip_64x64/, "const uint8_t *src_ptr, int src_stride, const uint8_t *ref_ptr, int ref_stride";
-  specialize qw/vpx_highbd_sad_skip_64x64 sse2 avx2/;
+  specialize qw/vpx_highbd_sad_skip_64x64 neon sse2 avx2/;
 
   add_proto qw/unsigned int vpx_highbd_sad_skip_64x32/, "const uint8_t *src_ptr, int src_stride, const uint8_t *ref_ptr, int ref_stride";
-  specialize qw/vpx_highbd_sad_skip_64x32 sse2 avx2/;
+  specialize qw/vpx_highbd_sad_skip_64x32 neon sse2 avx2/;
 
   add_proto qw/unsigned int vpx_highbd_sad_skip_32x64/, "const uint8_t *src_ptr, int src_stride, const uint8_t *ref_ptr, int ref_stride";
-  specialize qw/vpx_highbd_sad_skip_32x64 sse2 avx2/;
+  specialize qw/vpx_highbd_sad_skip_32x64 neon sse2 avx2/;
 
   add_proto qw/unsigned int vpx_highbd_sad_skip_32x32/, "const uint8_t *src_ptr, int src_stride, const uint8_t *ref_ptr, int ref_stride";
-  specialize qw/vpx_highbd_sad_skip_32x32 sse2 avx2/;
+  specialize qw/vpx_highbd_sad_skip_32x32 neon sse2 avx2/;
 
   add_proto qw/unsigned int vpx_highbd_sad_skip_32x16/, "const uint8_t *src_ptr, int src_stride, const uint8_t *ref_ptr, int ref_stride";
-  specialize qw/vpx_highbd_sad_skip_32x16 sse2 avx2/;
+  specialize qw/vpx_highbd_sad_skip_32x16 neon sse2 avx2/;
 
   add_proto qw/unsigned int vpx_highbd_sad_skip_16x32/, "const uint8_t *src_ptr, int src_stride, const uint8_t *ref_ptr, int ref_stride";
-  specialize qw/vpx_highbd_sad_skip_16x32 sse2 avx2/;
+  specialize qw/vpx_highbd_sad_skip_16x32 neon sse2 avx2/;
 
   add_proto qw/unsigned int vpx_highbd_sad_skip_16x16/, "const uint8_t *src_ptr, int src_stride, const uint8_t *ref_ptr, int ref_stride";
-  specialize qw/vpx_highbd_sad_skip_16x16 sse2 avx2/;
+  specialize qw/vpx_highbd_sad_skip_16x16 neon sse2 avx2/;
 
   add_proto qw/unsigned int vpx_highbd_sad_skip_16x8/, "const uint8_t *src_ptr, int src_stride, const uint8_t *ref_ptr, int ref_stride";
-  specialize qw/vpx_highbd_sad_skip_16x8 sse2 avx2/;
+  specialize qw/vpx_highbd_sad_skip_16x8 neon sse2 avx2/;
 
   add_proto qw/unsigned int vpx_highbd_sad_skip_8x16/, "const uint8_t *src_ptr, int src_stride, const uint8_t *ref_ptr, int ref_stride";
-  specialize qw/vpx_highbd_sad_skip_8x16 sse2/;
+  specialize qw/vpx_highbd_sad_skip_8x16 neon sse2/;
 
   add_proto qw/unsigned int vpx_highbd_sad_skip_8x8/, "const uint8_t *src_ptr, int src_stride, const uint8_t *ref_ptr, int ref_stride";
-  specialize qw/vpx_highbd_sad_skip_8x8 sse2/;
+  specialize qw/vpx_highbd_sad_skip_8x8 neon sse2/;
 
   add_proto qw/unsigned int vpx_highbd_sad_skip_8x4/, "const uint8_t *src_ptr, int src_stride, const uint8_t *ref_ptr, int ref_stride";
+  specialize qw/vpx_highbd_sad_skip_8x4 neon/;
 
   add_proto qw/unsigned int vpx_highbd_sad_skip_4x8/, "const uint8_t *src_ptr, int src_stride, const uint8_t *ref_ptr, int ref_stride";
+  specialize qw/vpx_highbd_sad_skip_4x8 neon/;
 
   add_proto qw/unsigned int vpx_highbd_sad_skip_4x4/, "const uint8_t *src_ptr, int src_stride, const uint8_t *ref_ptr, int ref_stride";
+  specialize qw/vpx_highbd_sad_skip_4x4 neon/;
 
   #
   # Avg