fast sk4f <-> sk4i SSE methods

src/opts/SkNx_sse.h

Index: src/opts/SkNx_sse.h
diff --git a/src/opts/SkNx_sse.h b/src/opts/SkNx_sse.h
index 7d00ddf9ba4722ef7ff1306e266f40099824cb18..be521a1a311e666f9b67a01ca80038ec3f75c487 100644
--- a/src/opts/SkNx_sse.h
+++ b/src/opts/SkNx_sse.h
@@ -294,6 +294,13 @@ public:
     __m128i fVec;
 };
 
+template<> /*static*/ inline Sk4f SkNx_cast<float, int>(const Sk4i& src) {
+    return _mm_cvtepi32_ps(src.fVec);
+}
+
+template <> /*static*/ inline Sk4i SkNx_cast<int, float>(const Sk4f& src) {
+    return _mm_cvttps_epi32(src.fVec);
+}
 
 template<> /*static*/ inline Sk4h SkNx_cast<uint16_t, float>(const Sk4f& src) {
     auto _32 = _mm_cvttps_epi32(src.fVec);