Add swizzle for rgb8888.

src/opts/SkNx_neon.h

 /*
  * Copyright 2015 Google Inc.
  *
  * Use of this source code is governed by a BSD-style license that can be
  * found in the LICENSE file.
  */
 
 #ifndef SkNx_neon_DEFINED
 #define SkNx_neon_DEFINED
 
@@ skipping 287 matching lines @@
 
     uint16x8_t fVec;
 };
 
 template <>
 class SkNx<4, uint8_t> {
 public:
     SkNx(const uint8x8_t& vec) : fVec(vec) {}
 
     SkNx() {}
+    SkNx(uint8_t a, uint8_t b, uint8_t c, uint8_t d) {
+        fVec = (uint8x8_t){a,b,c,d, 0,0,0,0};
+    }
     static SkNx Load(const void* ptr) {
         return (uint8x8_t)vld1_dup_u32((const uint32_t*)ptr);
     }
     void store(void* ptr) const {
         return vst1_lane_u32((uint32_t*)ptr, (uint32x2_t)fVec, 0);
     }
+    uint8_t operator[](int k) const {
+        SkASSERT(0 <= k && k < 4);
+        union { uint8x8_t v; uint8_t us[8]; } pun = {fVec};
+        return pun.us[k&3];
+    }
 
     // TODO as needed
 
     uint8x8_t fVec;
 };
 
 template <>
 class SkNx<16, uint8_t> {
 public:
     SkNx(const uint8x16_t& vec) : fVec(vec) {}
@@ skipping 66 matching lines @@
 
 template<> inline Sk4h SkNx_cast<uint16_t, uint8_t>(const Sk4b& src) {
     return vget_low_u16(vmovl_u8(src.fVec));
 }
 
 template<> inline Sk4b SkNx_cast<uint8_t, uint16_t>(const Sk4h& src) {
     return vmovn_u16(vcombine_u16(src.fVec, src.fVec));
 }
 
 #endif//SkNx_neon_DEFINED