try to speed-up maprect + round2i + contains

include/private/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 278 matching lines @@
 typedef SkNx<16,  uint8_t> Sk16b;
 
 typedef SkNx<4,  uint16_t> Sk4h;
 typedef SkNx<8,  uint16_t> Sk8h;
 typedef SkNx<16, uint16_t> Sk16h;
 
 typedef SkNx<4,       int> Sk4i;
 
 // Include platform specific specializations if available.
 #if !defined(SKNX_NO_SIMD) && SK_CPU_SSE_LEVEL >= SK_CPU_SSE_LEVEL_SSE2
-    #include "../opts/SkNx_sse.h"
+    #include "SkNx_sse.h"
 #elif !defined(SKNX_NO_SIMD) && defined(SK_ARM_HAS_NEON)
-    #include "../opts/SkNx_neon.h"
+    #include "SkNx_neon.h"
 #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