Optimize BlendPlane SIMD support for odd width.

source/row_any.cc

 /*
  *  Copyright 2012 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 65 matching lines @@
 #ifdef HAS_I422TOYUY2ROW_SSE2
 ANY31(I422ToYUY2Row_Any_SSE2, I422ToYUY2Row_SSE2, 1, 1, 4, 15)
 ANY31(I422ToUYVYRow_Any_SSE2, I422ToUYVYRow_SSE2, 1, 1, 4, 15)
 #endif
 #ifdef HAS_I422TOYUY2ROW_NEON
 ANY31(I422ToYUY2Row_Any_NEON, I422ToYUY2Row_NEON, 1, 1, 4, 15)
 #endif
 #ifdef HAS_I422TOUYVYROW_NEON
 ANY31(I422ToUYVYRow_Any_NEON, I422ToUYVYRow_NEON, 1, 1, 4, 15)
 #endif
+#ifdef HAS_BLENDPLANEROW_AVX2
+ANY31(BlendPlaneRow_Any_AVX2, BlendPlaneRow_AVX2, 0, 0, 1, 31)
+#endif
+#ifdef HAS_BLENDPLANEROW_SSSE3
+ANY31(BlendPlaneRow_Any_SSSE3, BlendPlaneRow_SSSE3, 0, 0, 1, 7)
+#endif
 #undef ANY31
 
+void BlendPlaneRow_Any_AVX2(const uint8* src0, const uint8* src1,
+                            const uint8* alpha, uint8* dst, int width);
+
+
 // Note that odd width replication includes 444 due to implementation
 // on arm that subsamples 444 to 422 internally.
 // Any 3 planes to 1 with yuvconstants
 #define ANY31C(NAMEANY, ANY_SIMD, UVSHIFT, DUVSHIFT, BPP, MASK)                \
     void NAMEANY(const uint8* y_buf, const uint8* u_buf, const uint8* v_buf,   \
                  uint8* dst_ptr, const struct YuvConstants* yuvconstants,      \
                  int width) {                                                  \
       SIMD_ALIGNED(uint8 temp[64 * 4]);                                        \
       memset(temp, 0, 64 * 3);  /* for YUY2 and msan */                        \
       int r = width & MASK;                                                    \
@@ skipping 709 matching lines @@
 #endif
 #ifdef HAS_UYVYTOUVROW_NEON
 ANY12S(UYVYToUVRow_Any_NEON, UYVYToUVRow_NEON, 1, 4, 15)
 #endif
 #undef ANY12S
 
 #ifdef __cplusplus
 }  // extern "C"
 }  // namespace libyuv
 #endif