Optimized premultiplying swizzles for NEON

src/opts/SkOpts_neon.cpp

Index: src/opts/SkOpts_neon.cpp
diff --git a/src/opts/SkOpts_neon.cpp b/src/opts/SkOpts_neon.cpp
index a0388b06544fec1054bd5ae5aa6dd2324e3cd36a..8b087bddc0b167799d47857c900c59511e4e45ba 100644
--- a/src/opts/SkOpts_neon.cpp
+++ b/src/opts/SkOpts_neon.cpp
@@ -15,6 +15,7 @@
 #include "SkFloatingPoint_opts.h"
 #include "SkMatrix_opts.h"
 #include "SkMorphologyImageFilter_opts.h"
+#include "SkSwizzler_neon.h"
 #include "SkTextureCompressor_opts.h"
 #include "SkUtils_opts.h"
 #include "SkXfermode_opts.h"
@@ -47,5 +48,8 @@ namespace SkOpts {
         matrix_translate       = sk_neon::matrix_translate;
         matrix_scale_translate = sk_neon::matrix_scale_translate;
         matrix_affine          = sk_neon::matrix_affine;
+
+        premul_xxxa        = sk_neon::premul_xxxa;

[msarett, 2016/01/13 14:05:38]

This actually doesn't work.

In order to run the optimized code, I had to modify the default in SkOpts.cpp. 
I'm trying to figure out why Init_neon() seems to not be run.

+        premul_swaprb_xxxa = sk_neon::premul_swaprb_xxxa;
     }
 }