/* Compare two strings for differences.

   For SPARC v9.

   Copyright (C) 2011-2018 Free Software Foundation, Inc.

   This file is part of the GNU C Library.

   Contributed by David S. Miller <davem@davemloft.net>

   The GNU C Library is free software; you can redistribute it and/or

   modify it under the terms of the GNU Lesser General Public

   License as published by the Free Software Foundation; either

   version 2.1 of the License, or (at your option) any later version.

   The GNU C Library is distributed in the hope that it will be useful,

   but WITHOUT ANY WARRANTY; without even the implied warranty of

   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU

   Lesser General Public License for more details.

   You should have received a copy of the GNU Lesser General Public

   License along with the GNU C Library; if not, see

   <http://www.gnu.org/licenses/>.  */

#include <sysdep.h>

#include <asm/asi.h>

#ifndef XCC

	.register	%g2, #scratch

	.register	%g3, #scratch

	.register	%g6, #scratch

#endif

#define rSTR1		%o0

#define rSTR2		%o1

#define r0101		%o2	/* 0x0101010101010101 */

#define r8080		%o3	/* 0x8080808080808080 */

#define rSTRXOR		%o4

#define rWORD1		%o5

#define rTMP1		%g1

#define rTMP2		%g2

#define rWORD2		%g3

#define rSLL		%g4

#define rSRL		%g5

#define rBARREL		%g6

	/* There are two cases, either the two pointers are aligned

	 * identically or they are not.  If they have the same

	 * alignment we can use the normal full speed loop.  Otherwise

	 * we have to use the barrel-shifter version.

	 */

	.text

	.align	32

ENTRY(strcmp)

	or	rSTR2, rSTR1, rTMP1

	sethi	%hi(0x80808080), r8080

	andcc	rTMP1, 0x7, %g0

	bne,pn	%icc, .Lmaybe_barrel_shift

	 or	r8080, %lo(0x80808080), r8080

	ldx	[rSTR1], rWORD1

	sub	rSTR2, rSTR1, rSTR2

	sllx	r8080, 32, rTMP1

	ldx	[rSTR1 + rSTR2], rWORD2

	or	r8080, rTMP1, r8080

	ba,pt	%xcc, .Laligned_loop_entry

	 srlx	r8080, 7, r0101

	.align	32

.Laligned_loop_entry:

.Laligned_loop:

	add	rSTR1, 8, rSTR1

	sub	rWORD1, r0101, rTMP2

	xorcc	rWORD1, rWORD2, rSTRXOR

	bne,pn	%xcc, .Lcommon_endstring

	 andn	r8080, rWORD1, rTMP1

	ldxa	[rSTR1] ASI_PNF, rWORD1

	andcc	rTMP1, rTMP2, %g0

	be,a,pt	%xcc, .Laligned_loop

	 ldxa	[rSTR1 + rSTR2] ASI_PNF, rWORD2

.Lcommon_equal:

	retl

	 mov	0, %o0

	/* All loops terminate here once they find an unequal word.

	 * If a zero byte appears in the word before the first unequal

	 * byte, we must report zero.  Otherwise we report '1' or '-1'

	 * depending upon whether the first mis-matching byte is larger

	 * in the first string or the second, respectively.

	 *

	 * First we compute a 64-bit mask value that has "0x01" in

	 * each byte where a zero exists in rWORD1.  rSTRXOR holds the

	 * value (rWORD1 ^ rWORD2).  Therefore, if considered as an

	 * unsigned quantity, our "0x01" mask value is "greater than"

	 * rSTRXOR then a zero terminating byte comes first and

	 * therefore we report '0'.

	 *

	 * The formula for this mask is:

	 *

	 *    mask_tmp1 = ~rWORD1 & 0x8080808080808080;

	 *    mask_tmp2 = ((rWORD1 & 0x7f7f7f7f7f7f7f7f) +

	 *                 0x7f7f7f7f7f7f7f7f);

	 *

	 *    mask = ((mask_tmp1 & ~mask_tmp2) >> 7);

	 */

.Lcommon_endstring:

	andn	rWORD1, r8080, rTMP2

	or	r8080, 1, %o1

	mov	1, %o0

	sub	rTMP2, %o1, rTMP2

	cmp	rWORD1, rWORD2

	andn	rTMP1, rTMP2, rTMP1

	movleu	%xcc, -1, %o0

	srlx	rTMP1, 7, rTMP1

	/* In order not to be influenced by bytes after the zero byte, we

	 * have to retain only the highest bit in the mask for the comparison

	 * with rSTRXOR to work properly.

	 */

	mov	0, rTMP2

	andcc	rTMP1, 0x0100, %g0

	movne	%xcc, 8, rTMP2

	sllx	rTMP1, 63 - 16, %o1

	movrlz	%o1, 16, rTMP2

	sllx	rTMP1, 63 - 24, %o1

	movrlz	%o1, 24, rTMP2

	sllx	rTMP1, 63 - 32, %o1

	movrlz	%o1, 32, rTMP2

	sllx	rTMP1, 63 - 40, %o1

	movrlz	%o1, 40, rTMP2

	sllx	rTMP1, 63 - 48, %o1

	movrlz	%o1, 48, rTMP2

	sllx	rTMP1, 63 - 56, %o1

	movrlz	%o1, 56, rTMP2

	srlx	rTMP1, rTMP2, rTMP1

	sllx	rTMP1, rTMP2, rTMP1

	cmp	rTMP1, rSTRXOR

	retl

	 movgu	%xcc, 0, %o0

.Lmaybe_barrel_shift:

	sub	rSTR2, rSTR1, rSTR2

	sllx	r8080, 32, rTMP1

	or	r8080, rTMP1, r8080

	and	rSTR1, 0x7, rTMP2

	srlx	r8080, 7, r0101

	andn	rSTR1, 0x7, rSTR1

	ldxa	[rSTR1] ASI_PNF, rWORD1

	andcc	rSTR2, 0x7, rSLL

	sll	rTMP2, 3, rSTRXOR

	bne,pn	%icc, .Lneed_barrel_shift

	 mov	-1, rTMP1

	ldxa	[rSTR1 + rSTR2] ASI_PNF, rBARREL

	srlx	rTMP1, rSTRXOR, rTMP2

	orn	rWORD1, rTMP2, rWORD1

	ba,pt	%xcc, .Laligned_loop_entry

	 orn	rBARREL, rTMP2, rWORD2

.Lneed_barrel_shift:

	sllx	rSLL, 3, rSLL

	andn	rSTR2, 0x7, rSTR2

	ldxa	[rSTR1 + rSTR2] ASI_PNF, rBARREL

	mov	64, rTMP2

	sub	rTMP2, rSLL, rSRL

	srlx	rTMP1, rSTRXOR, rTMP1

	add	rSTR2, 8, rSTR2

	orn	rWORD1, rTMP1, rWORD1

	sllx	rBARREL, rSLL, rWORD2

	ldxa	[rSTR1 + rSTR2] ASI_PNF, rBARREL

	add	rSTR1, 8, rSTR1

	sub	rWORD1, r0101, rTMP2

	srlx	rBARREL, rSRL, rSTRXOR

	or	rWORD2, rSTRXOR, rWORD2

	orn	rWORD2, rTMP1, rWORD2

	ba,pt	%xcc, .Lbarrel_shift_loop_entry

	 andn	r8080, rWORD1, rTMP1

.Lbarrel_shift_loop:

	sllx	rBARREL, rSLL, rWORD2

	ldxa	[rSTR1 + rSTR2] ASI_PNF, rBARREL

	add	rSTR1, 8, rSTR1

	sub	rWORD1, r0101, rTMP2

	srlx	rBARREL, rSRL, rSTRXOR

	andn	r8080, rWORD1, rTMP1

	or	rWORD2, rSTRXOR, rWORD2

.Lbarrel_shift_loop_entry:

	xorcc	rWORD1, rWORD2, rSTRXOR

	bne,pn	%xcc, .Lcommon_endstring

	 andcc	rTMP1, rTMP2, %g0

	be,a,pt	%xcc, .Lbarrel_shift_loop

	 ldxa	[rSTR1] ASI_PNF, rWORD1

	retl

	 mov	0, %o0

END(strcmp)

libc_hidden_builtin_def (strcmp)