Use the new universal intrinsics for s/dSUM across all platforms, and generic C c/zSUM on Windows

kernel/x86_64/KERNEL

@@ -487,9 +487,10 @@ XTRSMKERNEL_RT	=  xtrsm_kernel_LT_1x1.S
 
 XGEMM3MKERNEL    =  xgemm3m_kernel_2x2.S
 
-ifeq ($(OSNAME),WINNT)
+
 SSUMKERNEL = ../arm/sum.c
 DSUMKERNEL = ../arm/sum.c
+ifeq ($(OSNAME),WINNT)
 CSUMKERNEL = ../arm/zsum.c
 ZSUMKERNEL = ../arm/zsum.c
 endif