Reorganize to keep similar code together.

src/opts/SkUtils_opts.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.
  */
 
-#include "SkFloatingPoint.h"
-#include "SkOpts.h"
-#define SK_OPTS_NS neon
-#include "SkBlurImageFilter_opts.h"
-#include "SkXfermode_opts.h"
+#ifndef SkUtils_opts_DEFINED
+#define SkUtils_opts_DEFINED
 
-namespace neon {  // This helps identify methods from this file when debugging / profiling.
+namespace SK_OPTS_NS {
 
-static float rsqrt(float x) {
-    return sk_float_rsqrt(x);  // This sk_float_rsqrt copy will take the NEON compile-time path.
+#if SK_CPU_SSE_LEVEL >= SK_CPU_SSE_LEVEL_SSE2
+
+static void memset16(uint16_t* dst, uint16_t val, int n) {
+    auto dst8 = (__m128i*)dst;
+    auto val8 = _mm_set1_epi16(val);
+    for ( ; n >= 8; n -= 8) {
+        _mm_storeu_si128(dst8++, val8);
+    }
+    dst = (uint16_t*)dst8;
+    if (n & 4) {
+        _mm_storel_epi64((__m128i*)dst, val8);
+        dst += 4;
+    }
+    if (n & 2) {
+        *(uint32_t*)dst = _mm_cvtsi128_si32(val8);
+        dst += 2;
+    }
+    if (n & 1) {
+        *dst = val;
+    }
 }
 
+static void memset32(uint32_t* dst, uint32_t val, int n) {
+    auto dst4 = (__m128i*)dst;
+    auto val4 = _mm_set1_epi32(val);
+    for ( ; n >= 4; n -= 4) {
+        _mm_storeu_si128(dst4++, val4);
+    }
+    dst = (uint32_t*)dst4;
+    if (n & 2) {
+        _mm_storel_epi64((__m128i*)dst, val4);
+        dst += 2;
+    }
+    if (n & 1) {
+        *dst = val;
+    }
+}
+
+#elif defined(SK_ARM_HAS_NEON)
+
 static void memset16(uint16_t* dst, uint16_t value, int n) {
     uint16x8_t   v8  = vdupq_n_u16(value);
     uint16x8x4_t v32 = {{ v8, v8, v8, v8 }};
 
     while (n >= 32) {
         vst4q_u16(dst, v32);  // This swizzles, but we don't care: all lanes are the same, value.
         dst += 32;
         n   -= 32;
     }
     switch (n / 8) {
@@ skipping 28 matching lines @@
     }
     if (n & 2) {
         vst1_u32(dst, vget_low_u32(v4));
         dst += 2;
     }
     if (n & 1) {
         *dst = value;
     }
 }
 
-}  // namespace neon
+#else // Neither NEON nor SSE2.
 
-namespace SkOpts {
-    void Init_neon() {
-        rsqrt           = neon::rsqrt;
-        memset16        = neon::memset16;
-        memset32        = neon::memset32;
-        create_xfermode = SkCreate4pxXfermode;
+static void memset16(uint16_t* dst, uint16_t val, int n) { while (n --> 0) { *dst++ = val; } }
+static void memset32(uint32_t* dst, uint32_t val, int n) { while (n --> 0) { *dst++ = val; } }
 
-        static const auto x = neon::kX, y = neon::kY;
-        box_blur_xx = neon::box_blur<x,x>;
-        box_blur_xy = neon::box_blur<x,y>;
-        box_blur_yx = neon::box_blur<y,x>;
-    }
-}
+#endif
+
+}  // namespace SK_OPTS_NS
+
+#endif//SkUtils_opts_DEFINED