Improve naive SkColorXform to half floats

src/core/SkNx.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_DEFINED
 #define SkNx_DEFINED
 
@@ skipping 291 matching lines @@
     #include "../opts/SkNx_neon.h"
 #else
 
 SI Sk4i Sk4f_round(const Sk4f& x) {
     return { (int) lrintf (x[0]),
              (int) lrintf (x[1]),
              (int) lrintf (x[2]),
              (int) lrintf (x[3]), };
 }
 
+// Transpose 4 Sk4h and store (256 bits total).
+SI void Sk4h_store4(void* dst, const Sk4h& r, const Sk4h& g, const Sk4h& b, const Sk4h& a) {

[mtklein, 2016/07/19 15:37:47]

How about

auto dst64 = (uint64_t*)dst;
Sk4h(r[0], g[0], b[0], a[0]).store(dst64 + 0);
Sk4h(r[1], g[1], b[1], a[1]).store(dst64 + 1);
...

[msarett, 2016/07/19 15:47:20]

SGTM

+    Sk4h rgba0 = Sk4h(r[0], g[0], b[0], a[0]);
+    Sk4h rgba1 = Sk4h(r[1], g[1], b[1], a[1]);
+    Sk4h rgba2 = Sk4h(r[2], g[2], b[2], a[2]);
+    Sk4h rgba3 = Sk4h(r[3], g[3], b[3], a[3]);
+
+    uint64_t* dst64 = (uint64_t*) dst;
+    rgba0.store(dst64 + 0);
+    rgba1.store(dst64 + 1);
+    rgba2.store(dst64 + 2);
+    rgba3.store(dst64 + 3);
+}
+
 #endif
 
 SI void Sk4f_ToBytes(uint8_t p[16], const Sk4f& a, const Sk4f& b, const Sk4f& c, const Sk4f& d) {
     SkNx_cast<uint8_t>(SkNx_join(SkNx_join(a,b), SkNx_join(c,d))).store(p);
 }
 
 #undef SI
 
 #endif//SkNx_DEFINED