fixed compilation issues

cmake/OpenCVCompilerOptions.cmake

@@ -227,6 +227,12 @@ if(MSVC)
   if(ENABLE_AVX AND NOT MSVC_VERSION LESS 1600)
     set(OPENCV_EXTRA_FLAGS "${OPENCV_EXTRA_FLAGS} /arch:AVX")
   endif()
+  if(ENABLE_AVX2 AND NOT MSVC_VERSION LESS 1600)
+    set(OPENCV_EXTRA_FLAGS "${OPENCV_EXTRA_FLAGS} /arch:AVX2")
+  endif()
+  if(ENABLE_FMA3 AND NOT MSVC_VERSION LESS 1600)
+    set(OPENCV_EXTRA_FLAGS "${OPENCV_EXTRA_FLAGS} /arch:FMA")
+  endif()
 
   if(ENABLE_SSE4_1 AND CV_ICC AND NOT OPENCV_EXTRA_FLAGS MATCHES "/arch:")
     set(OPENCV_EXTRA_FLAGS "${OPENCV_EXTRA_FLAGS} /arch:SSE4.1")

modules/core/src/convert.cpp

@@ -617,15 +617,18 @@ struct VMerge4<data_type>
 }
 
 MERGE2_KERNEL_TEMPLATE( uchar, __m128i, __m128i, _mm_deinterleave_epi8, si128);
+MERGE3_KERNEL_TEMPLATE( uchar, __m128i, __m128i, _mm_interleave_epi8, si128);
+MERGE4_KERNEL_TEMPLATE( uchar, __m128i, __m128i, _mm_interleave_epi8, si128);
+
 MERGE2_KERNEL_TEMPLATE(ushort, __m128i, __m128i, _mm_deinterleave_epi16, si128);
-MERGE2_KERNEL_TEMPLATE(   int,  __m128,   float, _mm_deinterleave_ps, ps);
 
-MERGE3_KERNEL_TEMPLATE( uchar, __m128i, __m128i, _mm_interleave_epi8, si128);
+#if CV_SSE4_1
 MERGE3_KERNEL_TEMPLATE(ushort, __m128i, __m128i, _mm_interleave_epi16, si128);
-MERGE3_KERNEL_TEMPLATE(   int,  __m128,   float, _mm_interleave_ps, ps);
-
-MERGE4_KERNEL_TEMPLATE( uchar, __m128i, __m128i, _mm_interleave_epi8, si128);
 MERGE4_KERNEL_TEMPLATE(ushort, __m128i, __m128i, _mm_interleave_epi16, si128);
+#endif
+
+MERGE2_KERNEL_TEMPLATE(   int,  __m128,   float, _mm_deinterleave_ps, ps);
+MERGE3_KERNEL_TEMPLATE(   int,  __m128,   float, _mm_interleave_ps, ps);
 MERGE4_KERNEL_TEMPLATE(   int,  __m128,   float, _mm_interleave_ps, ps);
 
 #endif
@@ -4328,7 +4331,7 @@ cvtScale_<short, int, float>( const short* src, size_t sstep,
         {
             __m256 scale256 = _mm256_set1_ps(scale);
             __m256 shift256 = _mm256_set1_ps(shift);
-            int shuffle = 0xD8;
+            const int shuffle = 0xD8;
 
             for ( ; x <= size.width - 16; x += 16)
             {

modules/imgproc/src/color.cpp

@@ -1445,7 +1445,7 @@ struct RGB2Gray<float>
     float32x4_t v_cb, v_cg, v_cr;
 };
 
-#elif CV_SSE2
+#elif CV_SSE4_1
 
 template <>
 struct RGB2Gray<ushort>
@@ -2106,7 +2106,7 @@ struct RGB2YCrCb_i<ushort>
     int32x4_t v_c0, v_c1, v_c2, v_c3, v_c4, v_delta, v_delta2;
 };
 
-#elif CV_SSE2
+#elif CV_SSE4_1
 
 template <>
 struct RGB2YCrCb_i<uchar>
@@ -2247,8 +2247,6 @@ struct RGB2YCrCb_i<uchar>
     __m128i v_zero;
 };
 
-#if CV_SSE4_1
-
 template <>
 struct RGB2YCrCb_i<ushort>
 {
@@ -2369,9 +2367,6 @@ struct RGB2YCrCb_i<ushort>
 
 #endif // CV_SSE4_1
 
-
-#endif
-
 template<typename _Tp> struct YCrCb2RGB_f
 {
     typedef _Tp channel_type;

modules/imgproc/src/imgwarp.cpp

@@ -2306,7 +2306,7 @@ struct ResizeAreaFastVec_SIMD_32f
 
         if (cn == 1)
         {
-            int shuffle_lo = _MM_SHUFFLE(2, 0, 2, 0), shuffle_hi = _MM_SHUFFLE(3, 1, 3, 1);
+            const int shuffle_lo = _MM_SHUFFLE(2, 0, 2, 0), shuffle_hi = _MM_SHUFFLE(3, 1, 3, 1);
             for ( ; dx <= w - 4; dx += 4, S0 += 8, S1 += 8, D += 4)
             {
                 __m128 v_row00 = _mm_loadu_ps(S0), v_row01 = _mm_loadu_ps(S0 + 4),

modules/imgproc/src/pyramids.cpp

@@ -236,7 +236,11 @@ struct PyrDownVec_32s16u
     bool haveSSE;
 };
 
-#endif
+#else
+
+typedef PyrDownNoVec<int, ushort> PyrDownVec_32s16u;
+
+#endif // CV_SSE4_1
 
 struct PyrDownVec_32s16s
 {
@@ -288,7 +292,6 @@ struct PyrDownVec_32s16s
     bool haveSSE;
 };
 
-
 struct PyrUpVec_32s8u
 {
     int operator()(int** src, uchar** dst, int, int width) const
@@ -471,7 +474,11 @@ struct PyrUpVec_32s16u
     }
 };
 
-#endif
+#else
+
+typedef PyrUpNoVec<int, ushort> PyrUpVec_32s16u;
+
+#endif // CV_SSE4_1
 
 struct PyrUpVec_32f
 {