patch-2.1.79 linux/arch/sparc64/lib/strncpy_from_user.S

• Lines: 154
• Date: Mon Jan 12 15:15:44 1998
• Orig file: v2.1.78/linux/arch/sparc64/lib/strncpy_from_user.S
• Orig date: Thu Sep 4 17:07:30 1997

diff -u --recursive --new-file v2.1.78/linux/arch/sparc64/lib/strncpy_from_user.S linux/arch/sparc64/lib/strncpy_from_user.S
@@ -1,4 +1,5 @@
-/* strncpy_from_user.S: Sparc64 strncpy from userspace.
+/* $Id: strncpy_from_user.S,v 1.5 1997/09/08 11:29:23 jj Exp $
+ * strncpy_from_user.S: Sparc64 strncpy from userspace.
  *
  *  Copyright (C) 1997 Jakub Jelinek (jj@sunsite.mff.cuni.cz)
  */
@@ -6,6 +7,10 @@
 #include <asm/asi.h>
 #include <asm/errno.h>
 
+	.data
+	.align	8
+0:	.xword	0x0101010101010101
+
 	.text
 	.align	4
 
@@ -14,41 +19,117 @@
 	 * -EFAULT		for an exception
 	 * count		if we hit the buffer limit
 	 * bytes copied		if we hit a null byte
+	 * (without the null byte)
+	 *
+	 * This implementation assumes:
+	 * %o1 is 8 aligned => !(%o2 & 7)
+	 * %o0 is 8 aligned (if not, it will be slooooow, but will work)
+	 *
+	 * This is optimized for the common case:
+	 * in my stats, 90% of src are 8 aligned (even on sparc32)
+	 * and average length is 18 or so.
 	 */
 
 	.globl	__strncpy_from_user
 __strncpy_from_user:
 	/* %o0=dest, %o1=src, %o2=count */
-	brlez,pn %o2, 3f
+	sethi	%hi(0b), %o5		! IEU0	Group
+	andcc	%o1, 7, %g0		! IEU1
+	bne,pn	%icc, 30f		! CTI
+	 ldx	[%o5 + %lo(0b)], %o4	! Load	Group
+60:	ldxa	[%o1] ASI_S, %g1	! Load	Group
+	add	%o1, %o2, %o1		! IEU0
+	subcc	%g0, %o2, %o3		! IEU1
+	bgeu,pn	%xcc, 10f		! CTI
+	 sllx	%o4, 7, %o5		! IEU0	Group
+	add	%o0, %o2, %o0		! IEU1
+1:	sub	%g1, %o4, %g2		! IEU0	Group
+	stx	%g1, [%o0 + %o3]	! Store
+	andcc	%g2, %o5, %g0		! IEU1	Group
+	bne,pn	%xcc, 5f		! CTI
+	 add	%o3, 8, %o3		! IEU0
+	brlz,a,pt %o3, 1b		! CTI(IEU1) Group
+61:	 ldxa	[%o1 + %o3] ASI_S, %g1	! Load
+10:	retl				! CTI	Group
+	 mov	%o2, %o0		! IEU0
+5:	srlx	%g2, 32, %g7		! IEU0	Group
+	sethi	%hi(0xff00), %g5	! IEU1
+	andcc	%g7, %o5, %g0		! IEU1	Group
+	be,pn	%icc, 2f		! CTI
+	 or	%g5, %lo(0xff00), %g5	! IEU0
+	srlx	%g1, 48, %g7		! IEU0	Group
+	andcc	%g7, %g5, %g0		! IEU1	Group
+	be,pn	%icc, 50f		! CTI
+	 andcc	%g7, 0xff, %g0		! IEU1	Group
+	be,pn	%icc, 51f		! CTI
+	 srlx	%g1, 32, %g7		! IEU0
+	andcc	%g7, %g5, %g0		! IEU1	Group
+	be,pn	%icc, 52f		! CTI
+	 andcc	%g7, 0xff, %g0		! IEU1	Group
+	be,pn	%icc, 53f		! CTI
+2:	 andcc	%g2, %o5, %g0		! IEU1	Group
+	be,pn	%icc, 2f		! CTI
+	 srl	%g1, 16, %g7		! IEU0
+	andcc	%g7, %g5, %g0		! IEU1	Group
+	be,pn	%icc, 54f		! CTI
+	 andcc	%g7, 0xff, %g0		! IEU1	Group
+	be,pn	%icc, 55f		! CTI
+	 andcc	%g1, %g5, %g0		! IEU1	Group
+	be,pn	%icc, 56f		! CTI
+	 andcc	%g1, 0xff, %g0		! IEU1	Group
+	be,a,pn	%icc, 57f		! CTI
+	 add	%o2, %o3, %o0		! IEU0
+2:	brlz,a,pt %o3, 1b		! CTI(IEU1) Group
+62:	 ldxa	[%o1 + %o3] ASI_S, %g1	! Load
+	retl				! CTI	Group
+	 mov	%o2, %o0		! IEU0
+50:	add	%o2, %o3, %o0
+	retl
+	 sub	%o0, 8, %o0
+51:	add	%o2, %o3, %o0
+	retl
+	 sub	%o0, 7, %o0
+52:	add	%o2, %o3, %o0
+	retl
+	 sub	%o0, 6, %o0
+53:	add	%o2, %o3, %o0
+	retl
+	 sub	%o0, 5, %o0
+54:	add	%o2, %o3, %o0
+	retl
+	 sub	%o0, 4, %o0
+55:	add	%o2, %o3, %o0
+	retl
+	 sub	%o0, 3, %o0
+56:	add	%o2, %o3, %o0
+	retl
+	 sub	%o0, 2, %o0
+57:	retl
+	 sub	%o0, 1, %o0
+30:	brlez,pn %o2, 3f
 	 add	%o1, %o2, %o1
 	sub	%g0, %o2, %o3
 	add	%o0, %o2, %o0
-10:
-	lduba	[%o1 + %o3] ASI_S, %o4
-1:
-	brz,pn	%o4, 2f
+63:	lduba	[%o1 + %o3] ASI_S, %o4
+1:	brz,pn	%o4, 2f
 	 stb	%o4, [%o0 + %o3]
 	addcc	%o3, 1, %o3
 	bne,pt	%xcc, 1b
-11:
-	 lduba	[%o1 + %o3] ASI_S, %o4
-	retl
+64:	 lduba	[%o1 + %o3] ASI_S, %o4
+3:	retl
 	 mov	%o2, %o0
-2:
-	add	%o3, 1, %o3
-	retl
+2:	retl
 	 add	%o2, %o3, %o0
-3:
-	retl
-	 clr	%o0
 
 	.section .fixup,#alloc,#execinstr
 	.align	4
-4:
-	retl
+4:	retl
 	 mov	-EFAULT, %o0
 
 	.section __ex_table,#alloc
 	.align	4
-	.word	10b, 4b
-	.word	11b, 4b
+	.word	60b, 4b
+	.word	61b, 4b
+	.word	62b, 4b
+	.word	63b, 4b
+	.word	64b, 4b