libvpx: Pull from upstream

source/libvpx/vpx_dsp/arm/sad_neon.c

Index: source/libvpx/vpx_dsp/arm/sad_neon.c
diff --git a/source/libvpx/vp9/encoder/arm/neon/vp9_sad_neon.c b/source/libvpx/vpx_dsp/arm/sad_neon.c
similarity index 65%
rename from source/libvpx/vp9/encoder/arm/neon/vp9_sad_neon.c
rename to source/libvpx/vpx_dsp/arm/sad_neon.c
index c4cd856804da7ec7e3f42f8df2b562f27bc62cbf..173f08ac3c3e202764a8dd01a43a9b8877d08289 100644
--- a/source/libvpx/vp9/encoder/arm/neon/vp9_sad_neon.c
+++ b/source/libvpx/vpx_dsp/arm/sad_neon.c
@@ -9,11 +9,113 @@
  */
 
 #include <arm_neon.h>
-#include "./vp9_rtcd.h"
+
 #include "./vpx_config.h"
 
 #include "vpx/vpx_integer.h"
 
+unsigned int vpx_sad8x16_neon(
+        unsigned char *src_ptr,
+        int src_stride,
+        unsigned char *ref_ptr,
+        int ref_stride) {
+    uint8x8_t d0, d8;
+    uint16x8_t q12;
+    uint32x4_t q1;
+    uint64x2_t q3;
+    uint32x2_t d5;
+    int i;
+
+    d0 = vld1_u8(src_ptr);
+    src_ptr += src_stride;
+    d8 = vld1_u8(ref_ptr);
+    ref_ptr += ref_stride;
+    q12 = vabdl_u8(d0, d8);
+
+    for (i = 0; i < 15; i++) {
+        d0 = vld1_u8(src_ptr);
+        src_ptr += src_stride;
+        d8 = vld1_u8(ref_ptr);
+        ref_ptr += ref_stride;
+        q12 = vabal_u8(q12, d0, d8);
+    }
+
+    q1 = vpaddlq_u16(q12);
+    q3 = vpaddlq_u32(q1);
+    d5 = vadd_u32(vreinterpret_u32_u64(vget_low_u64(q3)),
+                  vreinterpret_u32_u64(vget_high_u64(q3)));
+
+    return vget_lane_u32(d5, 0);
+}
+
+unsigned int vpx_sad4x4_neon(
+        unsigned char *src_ptr,
+        int src_stride,
+        unsigned char *ref_ptr,
+        int ref_stride) {
+    uint8x8_t d0, d8;
+    uint16x8_t q12;
+    uint32x2_t d1;
+    uint64x1_t d3;
+    int i;
+
+    d0 = vld1_u8(src_ptr);
+    src_ptr += src_stride;
+    d8 = vld1_u8(ref_ptr);
+    ref_ptr += ref_stride;
+    q12 = vabdl_u8(d0, d8);
+
+    for (i = 0; i < 3; i++) {
+        d0 = vld1_u8(src_ptr);
+        src_ptr += src_stride;
+        d8 = vld1_u8(ref_ptr);
+        ref_ptr += ref_stride;
+        q12 = vabal_u8(q12, d0, d8);
+    }
+
+    d1 = vpaddl_u16(vget_low_u16(q12));
+    d3 = vpaddl_u32(d1);
+
+    return vget_lane_u32(vreinterpret_u32_u64(d3), 0);
+}
+
+unsigned int vpx_sad16x8_neon(
+        unsigned char *src_ptr,
+        int src_stride,
+        unsigned char *ref_ptr,
+        int ref_stride) {
+    uint8x16_t q0, q4;
+    uint16x8_t q12, q13;
+    uint32x4_t q1;
+    uint64x2_t q3;
+    uint32x2_t d5;
+    int i;
+
+    q0 = vld1q_u8(src_ptr);
+    src_ptr += src_stride;
+    q4 = vld1q_u8(ref_ptr);
+    ref_ptr += ref_stride;
+    q12 = vabdl_u8(vget_low_u8(q0), vget_low_u8(q4));
+    q13 = vabdl_u8(vget_high_u8(q0), vget_high_u8(q4));
+
+    for (i = 0; i < 7; i++) {
+        q0 = vld1q_u8(src_ptr);
+        src_ptr += src_stride;
+        q4 = vld1q_u8(ref_ptr);
+        ref_ptr += ref_stride;
+        q12 = vabal_u8(q12, vget_low_u8(q0), vget_low_u8(q4));
+        q13 = vabal_u8(q13, vget_high_u8(q0), vget_high_u8(q4));
+    }
+
+    q12 = vaddq_u16(q12, q13);
+    q1 = vpaddlq_u16(q12);
+    q3 = vpaddlq_u32(q1);
+    d5 = vadd_u32(vreinterpret_u32_u64(vget_low_u64(q3)),
+                  vreinterpret_u32_u64(vget_high_u64(q3)));
+
+    return vget_lane_u32(d5, 0);
+}
+
 static INLINE unsigned int horizontal_long_add_16x8(const uint16x8_t vec_lo,
                                                     const uint16x8_t vec_hi) {
   const uint32x4_t vec_l_lo = vaddl_u16(vget_low_u16(vec_lo),
@@ -34,7 +136,7 @@ static INLINE unsigned int horizontal_add_16x8(const uint16x8_t vec_16x8) {
   return vget_lane_u32(c, 0);
 }
 
-unsigned int vp9_sad64x64_neon(const uint8_t *src, int src_stride,
+unsigned int vpx_sad64x64_neon(const uint8_t *src, int src_stride,
                                const uint8_t *ref, int ref_stride) {
   int i;
   uint16x8_t vec_accum_lo = vdupq_n_u16(0);
@@ -70,7 +172,7 @@ unsigned int vp9_sad64x64_neon(const uint8_t *src, int src_stride,
   return horizontal_long_add_16x8(vec_accum_lo, vec_accum_hi);
 }
 
-unsigned int vp9_sad32x32_neon(const uint8_t *src, int src_stride,
+unsigned int vpx_sad32x32_neon(const uint8_t *src, int src_stride,
                                const uint8_t *ref, int ref_stride) {
   int i;
   uint16x8_t vec_accum_lo = vdupq_n_u16(0);
@@ -95,7 +197,7 @@ unsigned int vp9_sad32x32_neon(const uint8_t *src, int src_stride,
   return horizontal_add_16x8(vaddq_u16(vec_accum_lo, vec_accum_hi));
 }
 
-unsigned int vp9_sad16x16_neon(const uint8_t *src, int src_stride,
+unsigned int vpx_sad16x16_neon(const uint8_t *src, int src_stride,
                                const uint8_t *ref, int ref_stride) {
   int i;
   uint16x8_t vec_accum_lo = vdupq_n_u16(0);
@@ -114,7 +216,7 @@ unsigned int vp9_sad16x16_neon(const uint8_t *src, int src_stride,
   return horizontal_add_16x8(vaddq_u16(vec_accum_lo, vec_accum_hi));
 }
 
-unsigned int vp9_sad8x8_neon(const uint8_t *src, int src_stride,
+unsigned int vpx_sad8x8_neon(const uint8_t *src, int src_stride,
                              const uint8_t *ref, int ref_stride) {
   int i;
   uint16x8_t vec_accum = vdupq_n_u16(0);