rounding for arm filter

source/scale_neon.cc

 /*
  *  Copyright 2011 The LibYuv Project Authors. All rights reserved.
  *
  *  Use of this source code is governed by a BSD-style license
  *  that can be found in the LICENSE file in the root of the source
  *  tree. An additional intellectual property rights grant can be found
  *  in the file PATENTS. All contributing project authors may
  *  be found in the AUTHORS file in the root of the source tree.
  */
 
@@ skipping 594 matching lines @@
     "vmov       q11, q2                        \n"
     "vuzp.16    q10, q11                       \n"
     "vmovl.u8   q8, d6                         \n"
     "vmovl.u8   q9, d7                         \n"
     "vsubl.s16  q11, d18, d16                  \n"
     "vsubl.s16  q12, d19, d17                  \n"
     "vmovl.u16  q13, d20                       \n"
     "vmovl.u16  q10, d21                       \n"
     "vmul.s32   q11, q11, q13                  \n"
     "vmul.s32   q12, q12, q10                  \n"
-    "vshrn.s32  d18, q11, #16                  \n"
-    "vshrn.s32  d19, q12, #16                  \n"
+    "vrshrn.s32  d18, q11, #16                 \n"
+    "vrshrn.s32  d19, q12, #16                 \n"
     "vadd.s16   q8, q8, q9                     \n"
     "vmovn.s16  d6, q8                         \n"
 
     MEMACCESS(0)
     "vst1.8     {d6}, [%0]!                    \n"  // store pixels
     "vadd.s32   q1, q1, q0                     \n"
     "vadd.s32   q2, q2, q0                     \n"
     "subs       %2, %2, #8                     \n"  // 8 processed per loop
     "bgt        1b                             \n"
   : "+r"(dst_ptr),          // %0
@@ skipping 385 matching lines @@
 }
 
 #undef LOAD2_DATA32_LANE
 
 #endif  // defined(__ARM_NEON__) && !defined(__aarch64__)
 
 #ifdef __cplusplus
 }  // extern "C"
 }  // namespace libyuv
 #endif