avx2 interpolate use 8 bit

source/row_neon64.cc

 /*
  *  Copyright 2014 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 2318 matching lines @@
   : "cc", "memory", "v0", "v1", "v2", "v3", "v4", "v5", "v6", "v7", "v16"
   );
 }
 #endif  // HAS_RAWTOYROW_NEON
 
 // Bilinear filter 16x2 -> 16x1
 #ifdef HAS_INTERPOLATEROW_NEON
 void InterpolateRow_NEON(uint8* dst_ptr,
                          const uint8* src_ptr, ptrdiff_t src_stride,
                          int dst_width, int source_y_fraction) {
-  int y1_fraction = source_y_fraction >> 1;
-  int y0_fraction = 128 - y1_fraction;
+  int y1_fraction = source_y_fraction;
+  int y0_fraction = 256 - y1_fraction;
   const uint8* src_ptr1 = src_ptr + src_stride;
   asm volatile (
     "cmp        %w4, #0                        \n"
     "b.eq       100f                           \n"
-    "cmp        %w4, #64                       \n"
+    "cmp        %w4, #128                      \n"
     "b.eq       50f                            \n"
 
     "dup        v5.16b, %w4                    \n"
     "dup        v4.16b, %w5                    \n"
     // General purpose row blend.
   "1:                                          \n"
     MEMACCESS(1)
     "ld1        {v0.16b}, [%1], #16            \n"
     MEMACCESS(2)
     "ld1        {v1.16b}, [%2], #16            \n"
     "subs       %w3, %w3, #16                  \n"
     "umull      v2.8h, v0.8b,  v4.8b           \n"
     "umull2     v3.8h, v0.16b, v4.16b          \n"
     "umlal      v2.8h, v1.8b,  v5.8b           \n"
     "umlal2     v3.8h, v1.16b, v5.16b          \n"
-    "rshrn      v0.8b,  v2.8h, #7              \n"
-    "rshrn2     v0.16b, v3.8h, #7              \n"
+    "rshrn      v0.8b,  v2.8h, #8              \n"
+    "rshrn2     v0.16b, v3.8h, #8              \n"
     MEMACCESS(0)
     "st1        {v0.16b}, [%0], #16            \n"
     "b.gt       1b                             \n"
     "b          99f                            \n"
 
     // Blend 50 / 50.
   "50:                                         \n"
     MEMACCESS(1)
     "ld1        {v0.16b}, [%1], #16            \n"
     MEMACCESS(2)
@@ skipping 623 matching lines @@
   : "cc", "memory", "v0", "v1", "v2", "v3"  // Clobber List
   );
 }
 #endif  // HAS_SOBELYROW_NEON
 #endif  // !defined(LIBYUV_DISABLE_NEON) && defined(__aarch64__)
 
 #ifdef __cplusplus
 }  // extern "C"
 }  // namespace libyuv
 #endif