YUV scale filter columns improved filtering accuracy

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 554 matching lines @@
 
 // TODO(Yang Zhang): Investigate less load instructions for
 // the x/dx stepping
 #define LOAD2_DATA8_LANE(n)                                    \
     "lsr        %5, %3, #16                    \n"             \
     "add        %6, %1, %5                     \n"             \
     "add        %3, %3, %4                     \n"             \
     MEMACCESS(6)                                               \
     "vld2.8     {d6["#n"], d7["#n"]}, [%6]     \n"
 
+// The NEON version mimics this formula:
+// #define BLENDER(a, b, f) (uint8)((int)(a) +
+//    ((int)(f) * ((int)(b) - (int)(a)) >> 16))
+
 void ScaleFilterCols_NEON(uint8* dst_ptr, const uint8* src_ptr,
                           int dst_width, int x, int dx) {
   int dx_offset[4] = {0, 1, 2, 3};
   int* tmp = dx_offset;
   const uint8* src_tmp = src_ptr;
   asm volatile (
     "vdup.32    q0, %3                         \n"  // x
     "vdup.32    q1, %4                         \n"  // dx
     "vld1.32    {q2}, [%5]                     \n"  // 0 1 2 3
     "vshl.i32   q3, q1, #2                     \n"  // 4 * dx
@@ skipping 423 matching lines @@
 }
 
 #undef LOAD2_DATA32_LANE
 
 #endif  // defined(__ARM_NEON__) && !defined(__aarch64__)
 
 #ifdef __cplusplus
 }  // extern "C"
 }  // namespace libyuv
 #endif