diff --git a/sys/amd64/amd64/support.S b/sys/amd64/amd64/support.S
index 505adcbf134c..b6e25687d748 100644
--- a/sys/amd64/amd64/support.S
+++ b/sys/amd64/amd64/support.S
@@ -41,25 +41,24 @@
 	.text
 
 /*
- * bcopy family
- * void bzero(void *buf, u_int len)
+ * void *memset(void *dest, int c, size_t len)
  */
-
-/* done */
-ENTRY(bzero)
+ENTRY(memset)
 	PUSH_FRAME_POINTER
-	movq	%rsi,%rcx
-	xorl	%eax,%eax
+	movq	%rdi,%r9
+	movq	%rdx,%rcx
+	movq	%rsi,%rax
 	shrq	$3,%rcx
 	rep
 	stosq
-	movq	%rsi,%rcx
+	movq	%rdx,%rcx
 	andq	$7,%rcx
 	rep
 	stosb
+	movq	%r9,%rax
 	POP_FRAME_POINTER
 	ret
-END(bzero)
+END(memset)
 
 /* Address: %rdi */
 ENTRY(pagezero)
@@ -116,13 +115,14 @@ ENTRY(bcmp)
 END(bcmp)
 
 /*
- * bcopy(src, dst, cnt)
+ * memmove(dst, src, cnt)
  *       rdi, rsi, rdx
+ * Original by:
  *  ws@tools.de     (Wolfgang Solfrank, TooLs GmbH) +49-228-985800
  */
-ENTRY(bcopy)
+ENTRY(memmove)
 	PUSH_FRAME_POINTER
-	xchgq	%rsi,%rdi
+	movq	%rdi,%r9
 	movq	%rdx,%rcx
 
 	movq	%rdi,%rax
@@ -137,6 +137,7 @@ ENTRY(bcopy)
 	andq	$7,%rcx				/* any bytes left? */
 	rep
 	movsb
+	movq	%r9,%rax
 	POP_FRAME_POINTER
 	ret
 
@@ -157,9 +158,10 @@ ENTRY(bcopy)
 	rep
 	movsq
 	cld
+	movq	%r9,%rax
 	POP_FRAME_POINTER
 	ret
-END(bcopy)
+END(memmove)
 
 /*
  * Note: memcpy does not support overlapping copies
diff --git a/sys/amd64/include/param.h b/sys/amd64/include/param.h
index 86f7e08d9725..e28b542c0e0d 100644
--- a/sys/amd64/include/param.h
+++ b/sys/amd64/include/param.h
@@ -158,4 +158,7 @@
 #define SC_TABLESIZE    1024                     /* Must be power of 2. */
 #endif
 
+#define ARCH_WANT_BUIILTIN_BCOPY
+#define ARCH_WANT_BUIILTIN_BZERO
+
 #endif /* !_AMD64_INCLUDE_PARAM_H_ */
diff --git a/sys/conf/files.amd64 b/sys/conf/files.amd64
index 68191238ff97..97a1ecef924e 100644
--- a/sys/conf/files.amd64
+++ b/sys/conf/files.amd64
@@ -620,8 +620,6 @@ isa/vga_isa.c			optional	vga
 kern/kern_clocksource.c		standard
 kern/link_elf_obj.c		standard
 libkern/x86/crc32_sse42.c	standard
-libkern/memmove.c		standard
-libkern/memset.c		standard
 #
 # IA32 binary support
 #
diff --git a/sys/sys/libkern.h b/sys/sys/libkern.h
index 3ec1ca76855e..e43dabaeb232 100644
--- a/sys/sys/libkern.h
+++ b/sys/sys/libkern.h
@@ -232,9 +232,11 @@ memset(void *b, int c, size_t len)
 {
 	char *bb;
 
+#ifndef ARCH_WANT_BUIILTIN_BZERO
 	if (c == 0)
 		bzero(b, len);
 	else
+#endif
 		for (bb = (char *)b; len--; )
 			*bb++ = c;
 	return (b);
diff --git a/sys/sys/systm.h b/sys/sys/systm.h
index 995e06774d64..c8834449a88a 100644
--- a/sys/sys/systm.h
+++ b/sys/sys/systm.h
@@ -41,6 +41,7 @@
 #define	_SYS_SYSTM_H_
 
 #include <sys/cdefs.h>
+#include <machine/param.h>
 #include <machine/atomic.h>
 #include <machine/cpufunc.h>
 #include <sys/callout.h>
@@ -258,13 +259,14 @@ void	hexdump(const void *ptr, int length, const char *hdr, int flags);
 #define	HD_OMIT_CHARS	(1 << 18)
 
 #define ovbcopy(f, t, l) bcopy((f), (t), (l))
+#ifdef ARCH_WANT_BUIILTIN_BCOPY
+#define	bcopy(from, to, len) __builtin_memmove(to, from, len)
+#else
 void	bcopy(const void * _Nonnull from, void * _Nonnull to, size_t len);
-#define bcopy(from, to, len) ({				\
-	if (__builtin_constant_p(len) && (len) <= 64)	\
-		__builtin_memmove((to), (from), (len));	\
-	else						\
-		bcopy((from), (to), (len));		\
-})
+#endif
+#ifdef ARCH_WANT_BUIILTIN_BZERO
+#define bzero(buf, len) __builtin_memset((buf), 0, (len))
+#else
 void	bzero(void * _Nonnull buf, size_t len);
 #define bzero(buf, len) ({				\
 	if (__builtin_constant_p(len) && (len) <= 64)	\
@@ -272,6 +274,7 @@ void	bzero(void * _Nonnull buf, size_t len);
 	else						\
 		bzero((buf), (len));			\
 })
+#endif
 void	explicit_bzero(void * _Nonnull, size_t);
 
 void	*memcpy(void * _Nonnull to, const void * _Nonnull from, size_t len);