mpl going on good process

StgSharp.Native/sn-head/sn_internal.h

@@ -2,11 +2,11 @@
 #pragma once
 #endif
 
-#include <xmmintrin.h>
-
 #ifndef SN_INTERNAL
 #define SN_INTERNAL
 
+#include <xmmintrin.h>
+
 #ifdef _MSC_VER
 #define _CRT_SECURE_NO_WARNINGS 1
 #define WIN32_LEAN_AND_MEAN

StgSharp.Native/sn-head/sn_intrinsic.h

@@ -59,7 +59,7 @@ typedef uint64_t(SN_DECL *FACTORIALROC)(int n);
 
 #pragma region matix function
 
-INTERNAL void normalize(__m128 *source, __m128 *target);
+INTERNAL void f32_normalize(VEC(float) * source, VEC(float) * target);
 
 INTERNAL void SN_DECL f32_add_sse(MAT_KERNEL(float) const *left, MAT_KERNEL(float) const *right,
                                   MAT_KERNEL(float) *restrict ans);

StgSharp.Native/sn-head/sn_matkernel.h

@@ -9,11 +9,26 @@
 #define ZMMCOUNT(T) sizeof(T) / 4
 
 #define MAT_KERNEL(T)                    \
-        union {                          \
+        union mat_kernel_##T##{          \
                 __m128 xmm[XMMCOUNT(T)]; \
                 __m256 ymm[YMMCOUNT(T)]; \
                 __m512 zmm[ZMMCOUNT(T)]; \
                 T m[4][4];               \
         }
 
+#define VEC(T)                               \
+        union vec_##T##{                     \
+                __m128 xmm[XMMCOUNT(T) / 4]; \
+                __m256 ymm[YMMCOUNT(T) / 4]; \
+                __m512 zmm[ZMMCOUNT(T) / 4]; \
+                T v[4];                      \
+        }
+
+/*
+#undef ELEMENTSIZE
+#undef XMMCOUNT
+#undef YMMCOUNT
+#undef ZMMCOUNT
+/**/
+
 #endif