restore sk4i SSE specialization

src/opts/SkNx_sse.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_sse_DEFINED
 #define SkNx_sse_DEFINED
 
@@ skipping 102 matching lines @@
 
     SkNx thenElse(const SkNx& t, const SkNx& e) const {
         return _mm_or_ps(_mm_and_ps   (fVec, t.fVec),
                          _mm_andnot_ps(fVec, e.fVec));
     }
 
     __m128 fVec;
 };
 
 template <>
+class SkNx<4, int> {
+public:
+    SkNx(const __m128i& vec) : fVec(vec) {}
+
+    SkNx() {}
+    SkNx(int val) : fVec(_mm_set1_epi32(val)) {}
+    static SkNx Load(const void* ptr) { return _mm_loadu_si128((const __m128i*)ptr); }
+    SkNx(int a, int b, int c, int d) : fVec(_mm_setr_epi32(a,b,c,d)) {}
+
+    void store(void* ptr) const { _mm_storeu_si128((__m128i*)ptr, fVec); }
+
+    SkNx operator + (const SkNx& o) const { return _mm_add_epi32(fVec, o.fVec); }
+    SkNx operator - (const SkNx& o) const { return _mm_sub_epi32(fVec, o.fVec); }
+    SkNx operator * (const SkNx& o) const {
+        __m128i mul20 = _mm_mul_epu32(fVec, o.fVec),
+                mul31 = _mm_mul_epu32(_mm_srli_si128(fVec, 4), _mm_srli_si128(o.fVec, 4));
+        return _mm_unpacklo_epi32(_mm_shuffle_epi32(mul20, _MM_SHUFFLE(0,0,2,0)),
+                                  _mm_shuffle_epi32(mul31, _MM_SHUFFLE(0,0,2,0)));
+    }
+
+    SkNx operator << (int bits) const { return _mm_slli_epi32(fVec, bits); }
+    SkNx operator >> (int bits) const { return _mm_srai_epi32(fVec, bits); }
+
+    int operator[](int k) const {
+        SkASSERT(0 <= k && k < 4);
+        union { __m128i v; int is[4]; } pun = {fVec};
+        return pun.is[k&3];
+    }
+    template <int k> int kth() const { return (*this)[k]; }
+
+    __m128i fVec;
+};
+
+template <>
 class SkNx<4, uint16_t> {
 public:
     SkNx(const __m128i& vec) : fVec(vec) {}
 
     SkNx() {}
     SkNx(uint16_t val) : fVec(_mm_set1_epi16(val)) {}
     static SkNx Load(const void* ptr) { return _mm_loadl_epi64((const __m128i*)ptr); }
     SkNx(uint16_t a, uint16_t b, uint16_t c, uint16_t d) : fVec(_mm_setr_epi16(a,b,c,d,0,0,0,0)) {}
 
     void store(void* ptr) const { _mm_storel_epi64((__m128i*)ptr, fVec); }
@@ skipping 169 matching lines @@
 
 template<> /*static*/ inline Sk4h SkNx_cast<uint16_t, uint8_t>(const Sk4b& src) {
     return _mm_unpacklo_epi8(src.fVec, _mm_setzero_si128());
 }
 
 template<> /*static*/ inline Sk4b SkNx_cast<uint8_t, uint16_t>(const Sk4h& src) {
     return _mm_packus_epi16(src.fVec, src.fVec);
 }
 
 #endif//SkNx_sse_DEFINED