decide to use L4 to solve numa and pack problem for large compute.

StgSharp.Native/sn-src/m.fma_tile.x86.c

@@ -3,7 +3,8 @@
 #if SN_IS_ARCH(SN_ARCH_X86_64)
 
 #include "sn_intrinsic.h"
-#include "sn_intrincontext.x86.h"
+#include "sn_intrinsic.std.h"
+#include "sn_intrinsic.context.x86.h"
 
 #define _FMA_SELECT(i) _MM_SHUFFLE(i, i, i, i)
 
@@ -13,6 +14,8 @@
 
 SN_MK_PROC_DECL_STD(float, sse, , fma)
 {
+        __mk_param_std();
+
         int i = enumeration->m128_i32[0];
         int j = enumeration->m128_i32[1];
         int commonK = enumeration->m128_i32[2];