- Add an optimized page copy function for use by pmap_copy_page(). It is

roughly four times faster than bcopy() for uncached pages.
 - Sort the function prototypes in md_var.h.

sys/amd64/amd64/pmap.c

@@ -2463,7 +2463,7 @@ pmap_copy_page(vm_page_t msrc, vm_page_t mdst)
 	vm_offset_t src = PHYS_TO_DMAP(VM_PAGE_TO_PHYS(msrc));
 	vm_offset_t dst = PHYS_TO_DMAP(VM_PAGE_TO_PHYS(mdst));
 
-	bcopy((void *)src, (void *)dst, PAGE_SIZE);
+	pagecopy((void *)src, (void *)dst);
 }
 
 /*

sys/amd64/amd64/support.S

@@ -167,6 +167,32 @@ ENTRY(memcpy)
 	movsb
 	ret
 
+/*
+ * pagecopy(%rdi=from, %rsi=to)
+ */
+ENTRY(pagecopy)
+	movq	$-PAGE_SIZE,%rax
+	movq	%rax,%rdx
+	subq	%rax,%rdi
+	subq	%rax,%rsi
+1:
+	prefetchnta (%rdi,%rax)
+	addq	$64,%rax
+	jne	1b
+2:
+	movq	(%rdi,%rdx),%rax
+	movnti	%rax,(%rsi,%rdx)
+	movq	8(%rdi,%rdx),%rax
+	movnti	%rax,8(%rsi,%rdx)
+	movq	16(%rdi,%rdx),%rax
+	movnti	%rax,16(%rsi,%rdx)
+	movq	24(%rdi,%rdx),%rax
+	movnti	%rax,24(%rsi,%rdx)
+	addq	$32,%rdx
+	jne	2b
+	sfence
+	ret
+
 /* fillw(pat, base, cnt) */  
 /*       %rdi,%rsi, %rdx */
 ENTRY(fillw)

sys/amd64/include/md_var.h

@@ -65,11 +65,12 @@ void	doreti_iret(void) __asm(__STRING(doreti_iret));
 void	doreti_iret_fault(void) __asm(__STRING(doreti_iret_fault));
 void	enable_sse(void);
 void	fillw(int /*u_short*/ pat, void *base, size_t cnt);
-void	pagezero(void *addr);
+void	fpstate_drop(struct thread *td);
 int	is_physical_memory(vm_offset_t addr);
 int	isa_nmi(int cd);
+void	pagecopy(void *from, void *to);
+void	pagezero(void *addr);
 void	setidt(int idx, alias_for_inthand_t *func, int typ, int dpl, int ist);
 int	user_dbreg_trap(void);
-void	fpstate_drop(struct thread *td);
 
 #endif /* !_MACHINE_MD_VAR_H_ */