Use SIMD-instructions in half-precision floats with AVX instruction set

Source/Engine/Core/Math/Half.h

@@ -6,13 +6,23 @@
 #include "Vector2.h"
 #include "Vector3.h"
 
+#if PLATFORM_SIMD_AVX
+// MSVC generates slower code without AVX
+#define USE_SSE_HALF_CONVERSION 1
+#else
+#define USE_SSE_HALF_CONVERSION 0
+#endif
+
+#if USE_SSE_HALF_CONVERSION
+#include <intrin.h>
+#include <emmintrin.h>
+#endif
+
 /// <summary>
 /// Half-precision 16 bit floating point number consisting of a sign bit, a 5 bit biased exponent, and a 10 bit mantissa
 /// </summary>
 typedef uint16 Half;
 
-#define USE_SSE_HALF_CONVERSION 0
-
 /// <summary>
 /// Utility for packing/unpacking floating point value from single precision (32 bit) to half precision (16 bit).
 /// </summary>