libvpx: Pull from upstream

source/libvpx/third_party/libyuv/source/compare_neon64.cc

Index: source/libvpx/third_party/libyuv/source/compare_neon64.cc
diff --git a/source/libvpx/third_party/libyuv/source/compare_neon.cc b/source/libvpx/third_party/libyuv/source/compare_neon64.cc
similarity index 54%
copy from source/libvpx/third_party/libyuv/source/compare_neon.cc
copy to source/libvpx/third_party/libyuv/source/compare_neon64.cc
index 55052c0eecb707aa7c2c5c8a3d7ae658a50ea739..cc078f84cd85024a33fe996192ced97361ce23d1 100644
--- a/source/libvpx/third_party/libyuv/source/compare_neon.cc
+++ b/source/libvpx/third_party/libyuv/source/compare_neon64.cc
@@ -16,47 +16,7 @@ namespace libyuv {
 extern "C" {
 #endif
 
-#if !defined(LIBYUV_DISABLE_NEON) && defined(__ARM_NEON__)
-
-uint32 SumSquareError_NEON(const uint8* src_a, const uint8* src_b, int count) {
-  volatile uint32 sse;
-  asm volatile (
-    "vmov.u8    q8, #0                         \n"
-    "vmov.u8    q10, #0                        \n"
-    "vmov.u8    q9, #0                         \n"
-    "vmov.u8    q11, #0                        \n"
-
-    ".p2align  2                               \n"
-  "1:                                          \n"
-    MEMACCESS(0)
-    "vld1.8     {q0}, [%0]!                    \n"
-    MEMACCESS(1)
-    "vld1.8     {q1}, [%1]!                    \n"
-    "subs       %2, %2, #16                    \n"
-    "vsubl.u8   q2, d0, d2                     \n"
-    "vsubl.u8   q3, d1, d3                     \n"
-    "vmlal.s16  q8, d4, d4                     \n"
-    "vmlal.s16  q9, d6, d6                     \n"
-    "vmlal.s16  q10, d5, d5                    \n"
-    "vmlal.s16  q11, d7, d7                    \n"
-    "bgt        1b                             \n"
-
-    "vadd.u32   q8, q8, q9                     \n"
-    "vadd.u32   q10, q10, q11                  \n"
-    "vadd.u32   q11, q8, q10                   \n"
-    "vpaddl.u32 q1, q11                        \n"
-    "vadd.u64   d0, d2, d3                     \n"
-    "vmov.32    %3, d0[0]                      \n"
-    : "+r"(src_a),
-      "+r"(src_b),
-      "+r"(count),
-      "=r"(sse)
-    :
-    : "memory", "cc", "q0", "q1", "q2", "q3", "q8", "q9", "q10", "q11");
-  return sse;
-}
-
-#elif !defined(LIBYUV_DISABLE_NEON) && defined(__aarch64__)
+#if !defined(LIBYUV_DISABLE_NEON) && defined(__aarch64__)
 
 uint32 SumSquareError_NEON(const uint8* src_a, const uint8* src_b, int count) {
   volatile uint32 sse;
@@ -79,7 +39,7 @@ uint32 SumSquareError_NEON(const uint8* src_a, const uint8* src_b, int count) {
     "smlal      v17.4s, v3.4h, v3.4h           \n"
     "smlal2     v18.4s, v2.8h, v2.8h           \n"
     "smlal2     v19.4s, v3.8h, v3.8h           \n"
-    "bgt        1b                             \n"
+    "b.gt       1b                             \n"
 
     "add        v16.4s, v16.4s, v17.4s         \n"
     "add        v18.4s, v18.4s, v19.4s         \n"
@@ -95,7 +55,7 @@ uint32 SumSquareError_NEON(const uint8* src_a, const uint8* src_b, int count) {
   return sse;
 }
 
-#endif  // __ARM_NEON__
+#endif  // !defined(LIBYUV_DISABLE_NEON) && defined(__aarch64__)
 
 #ifdef __cplusplus
 }  // extern "C"