could not resist: fast sse float <--> u16

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 330 matching lines @@
     }
 
     __m128i fVec;
 };
 
 
 template<> inline Sk4i SkNx_cast<int, float, 4>(const Sk4f& src) {
     return _mm_cvttps_epi32(src.fVec);
 }
 
+template<> inline Sk4h SkNx_cast<uint16_t, float, 4>(const Sk4f& src) {
+    auto _32 = _mm_cvttps_epi32(src.fVec);
+    return _mm_packus_epi16(_32, _32);
+}
+
 template<> inline Sk4b SkNx_cast<uint8_t, float, 4>(const Sk4f& src) {
     auto _32 = _mm_cvttps_epi32(src.fVec);
 #if SK_CPU_SSE_LEVEL >= SK_CPU_SSE_LEVEL_SSSE3
     const int _ = ~0;
     return _mm_shuffle_epi8(_32, _mm_setr_epi8(0,4,8,12, _,_,_,_, _,_,_,_, _,_,_,_));
 #else
     auto _16 = _mm_packus_epi16(_32, _32);
     return     _mm_packus_epi16(_16, _16);
 #endif
 }
 
 template<> inline Sk4f SkNx_cast<float, uint8_t, 4>(const Sk4b& src) {
 #if SK_CPU_SSE_LEVEL >= SK_CPU_SSE_LEVEL_SSSE3
     const int _ = ~0;
     auto _32 = _mm_shuffle_epi8(src.fVec, _mm_setr_epi8(0,_,_,_, 1,_,_,_, 2,_,_,_, 3,_,_,_));
 #else
     auto _16 = _mm_unpacklo_epi8(src.fVec, _mm_setzero_si128()),
          _32 = _mm_unpacklo_epi16(_16,     _mm_setzero_si128());
 #endif
     return _mm_cvtepi32_ps(_32);
 }
 
+template<> inline Sk4f SkNx_cast<float, uint16_t, 4>(const Sk4h& src) {
+    auto _32 = _mm_unpacklo_epi16(src.fVec, _mm_setzero_si128());
+    return _mm_cvtepi32_ps(_32);
+}
+
 static inline void Sk4f_ToBytes(uint8_t bytes[16],
                                 const Sk4f& a, const Sk4f& b, const Sk4f& c, const Sk4f& d) {
     _mm_storeu_si128((__m128i*)bytes,
                      _mm_packus_epi16(_mm_packus_epi16(_mm_cvttps_epi32(a.fVec),
                                                        _mm_cvttps_epi32(b.fVec)),
                                       _mm_packus_epi16(_mm_cvttps_epi32(c.fVec),
                                                        _mm_cvttps_epi32(d.fVec))));
 }
 
 template<> inline Sk4h SkNx_cast<uint16_t, uint8_t, 4>(const Sk4b& src) {
     return _mm_unpacklo_epi8(src.fVec, _mm_setzero_si128());
 }
 
 template<> inline Sk4b SkNx_cast<uint8_t, uint16_t, 4>(const Sk4h& src) {
     return _mm_packus_epi16(src.fVec, src.fVec);
 }
 
 
 }  // namespace
 
 #endif//SkNx_sse_DEFINED