[deliverable/linux.git] / arch / arm / lib / csumpartialcopygeneric.S

/*
 *  linux/arch/arm/lib/csumpartialcopygeneric.S
 *
 *  Copyright (C) 1995-2001 Russell King
 *
 * This program is free software; you can redistribute it and/or modify
 * it under the terms of the GNU General Public License version 2 as
 * published by the Free Software Foundation.
 */
#include <asm/assembler.h>
#include <asm/export.h>

/*
 * unsigned int
 * csum_partial_copy_xxx(const char *src, char *dst, int len, int sum, )
 *  r0 = src, r1 = dst, r2 = len, r3 = sum
 *  Returns : r0 = checksum
 *
 * Note that 'tst' and 'teq' preserve the carry flag.
 */

src	.req	r0
dst	.req	r1
len	.req	r2
sum	.req	r3

.Lzero:		mov	r0, sum
		load_regs

		/*
		 * Align an unaligned destination pointer.  We know that
		 * we have >= 8 bytes here, so we don't need to check
		 * the length.  Note that the source pointer hasn't been
		 * aligned yet.
		 */
.Ldst_unaligned:
		tst	dst, #1
		beq	.Ldst_16bit

		load1b	ip
		sub	len, len, #1
		adcs	sum, sum, ip, put_byte_1	@ update checksum
		strb	ip, [dst], #1
		tst	dst, #2
		reteq	lr			@ dst is now 32bit aligned

.Ldst_16bit:	load2b	r8, ip
		sub	len, len, #2
		adcs	sum, sum, r8, put_byte_0
		strb	r8, [dst], #1
		adcs	sum, sum, ip, put_byte_1
		strb	ip, [dst], #1
		ret	lr			@ dst is now 32bit aligned

		/*
		 * Handle 0 to 7 bytes, with any alignment of source and
		 * destination pointers.  Note that when we get here, C = 0
		 */
.Lless8:	teq	len, #0			@ check for zero count
		beq	.Lzero

		/* we must have at least one byte. */
		tst	dst, #1			@ dst 16-bit aligned
		beq	.Lless8_aligned

		/* Align dst */
		load1b	ip
		sub	len, len, #1
		adcs	sum, sum, ip, put_byte_1	@ update checksum
		strb	ip, [dst], #1
		tst	len, #6
		beq	.Lless8_byteonly

1:		load2b	r8, ip
		sub	len, len, #2
		adcs	sum, sum, r8, put_byte_0
		strb	r8, [dst], #1
		adcs	sum, sum, ip, put_byte_1
		strb	ip, [dst], #1
.Lless8_aligned:
		tst	len, #6
		bne	1b
.Lless8_byteonly:
		tst	len, #1
		beq	.Ldone
		load1b	r8
		adcs	sum, sum, r8, put_byte_0	@ update checksum
		strb	r8, [dst], #1
		b	.Ldone

FN_ENTRY
		save_regs

		cmp	len, #8			@ Ensure that we have at least
		blo	.Lless8			@ 8 bytes to copy.

		adds	sum, sum, #0		@ C = 0
		tst	dst, #3			@ Test destination alignment
		blne	.Ldst_unaligned		@ align destination, return here

		/*
		 * Ok, the dst pointer is now 32bit aligned, and we know
		 * that we must have more than 4 bytes to copy.  Note
		 * that C contains the carry from the dst alignment above.
		 */

		tst	src, #3			@ Test source alignment
		bne	.Lsrc_not_aligned

		/* Routine for src & dst aligned */

		bics	ip, len, #15
		beq	2f

1:		load4l	r4, r5, r6, r7
		stmia	dst!, {r4, r5, r6, r7}
		adcs	sum, sum, r4
		adcs	sum, sum, r5
		adcs	sum, sum, r6
		adcs	sum, sum, r7
		sub	ip, ip, #16
		teq	ip, #0
		bne	1b

2:		ands	ip, len, #12
		beq	4f
		tst	ip, #8
		beq	3f
		load2l	r4, r5
		stmia	dst!, {r4, r5}
		adcs	sum, sum, r4
		adcs	sum, sum, r5
		tst	ip, #4
		beq	4f

3:		load1l	r4
		str	r4, [dst], #4
		adcs	sum, sum, r4

4:		ands	len, len, #3
		beq	.Ldone
		load1l	r4
		tst	len, #2
		mov	r5, r4, get_byte_0
		beq	.Lexit
		adcs	sum, sum, r4, lspush #16
		strb	r5, [dst], #1
		mov	r5, r4, get_byte_1
		strb	r5, [dst], #1
		mov	r5, r4, get_byte_2
.Lexit:		tst	len, #1
		strneb	r5, [dst], #1
		andne	r5, r5, #255
		adcnes	sum, sum, r5, put_byte_0

		/*
		 * If the dst pointer was not 16-bit aligned, we
		 * need to rotate the checksum here to get around
		 * the inefficient byte manipulations in the
		 * architecture independent code.
		 */
.Ldone:		adc	r0, sum, #0
		ldr	sum, [sp, #0]		@ dst
		tst	sum, #1
		movne	r0, r0, ror #8
		load_regs

.Lsrc_not_aligned:
		adc	sum, sum, #0		@ include C from dst alignment
		and	ip, src, #3
		bic	src, src, #3
		load1l	r5
		cmp	ip, #2
		beq	.Lsrc2_aligned
		bhi	.Lsrc3_aligned
		mov	r4, r5, lspull #8		@ C = 0
		bics	ip, len, #15
		beq	2f
1:		load4l	r5, r6, r7, r8
		orr	r4, r4, r5, lspush #24
		mov	r5, r5, lspull #8
		orr	r5, r5, r6, lspush #24
		mov	r6, r6, lspull #8
		orr	r6, r6, r7, lspush #24
		mov	r7, r7, lspull #8
		orr	r7, r7, r8, lspush #24
		stmia	dst!, {r4, r5, r6, r7}
		adcs	sum, sum, r4
		adcs	sum, sum, r5
		adcs	sum, sum, r6
		adcs	sum, sum, r7
		mov	r4, r8, lspull #8
		sub	ip, ip, #16
		teq	ip, #0
		bne	1b
2:		ands	ip, len, #12
		beq	4f
		tst	ip, #8
		beq	3f
		load2l	r5, r6
		orr	r4, r4, r5, lspush #24
		mov	r5, r5, lspull #8
		orr	r5, r5, r6, lspush #24
		stmia	dst!, {r4, r5}
		adcs	sum, sum, r4
		adcs	sum, sum, r5
		mov	r4, r6, lspull #8
		tst	ip, #4
		beq	4f
3:		load1l	r5
		orr	r4, r4, r5, lspush #24
		str	r4, [dst], #4
		adcs	sum, sum, r4
		mov	r4, r5, lspull #8
4:		ands	len, len, #3
		beq	.Ldone
		mov	r5, r4, get_byte_0
		tst	len, #2
		beq	.Lexit
		adcs	sum, sum, r4, lspush #16
		strb	r5, [dst], #1
		mov	r5, r4, get_byte_1
		strb	r5, [dst], #1
		mov	r5, r4, get_byte_2
		b	.Lexit

.Lsrc2_aligned:	mov	r4, r5, lspull #16
		adds	sum, sum, #0
		bics	ip, len, #15
		beq	2f
1:		load4l	r5, r6, r7, r8
		orr	r4, r4, r5, lspush #16
		mov	r5, r5, lspull #16
		orr	r5, r5, r6, lspush #16
		mov	r6, r6, lspull #16
		orr	r6, r6, r7, lspush #16
		mov	r7, r7, lspull #16
		orr	r7, r7, r8, lspush #16
		stmia	dst!, {r4, r5, r6, r7}
		adcs	sum, sum, r4
		adcs	sum, sum, r5
		adcs	sum, sum, r6
		adcs	sum, sum, r7
		mov	r4, r8, lspull #16
		sub	ip, ip, #16
		teq	ip, #0
		bne	1b
2:		ands	ip, len, #12
		beq	4f
		tst	ip, #8
		beq	3f
		load2l	r5, r6
		orr	r4, r4, r5, lspush #16
		mov	r5, r5, lspull #16
		orr	r5, r5, r6, lspush #16
		stmia	dst!, {r4, r5}
		adcs	sum, sum, r4
		adcs	sum, sum, r5
		mov	r4, r6, lspull #16
		tst	ip, #4
		beq	4f
3:		load1l	r5
		orr	r4, r4, r5, lspush #16
		str	r4, [dst], #4
		adcs	sum, sum, r4
		mov	r4, r5, lspull #16
4:		ands	len, len, #3
		beq	.Ldone
		mov	r5, r4, get_byte_0
		tst	len, #2
		beq	.Lexit
		adcs	sum, sum, r4
		strb	r5, [dst], #1
		mov	r5, r4, get_byte_1
		strb	r5, [dst], #1
		tst	len, #1
		beq	.Ldone
		load1b	r5
		b	.Lexit

.Lsrc3_aligned:	mov	r4, r5, lspull #24
		adds	sum, sum, #0
		bics	ip, len, #15
		beq	2f
1:		load4l	r5, r6, r7, r8
		orr	r4, r4, r5, lspush #8
		mov	r5, r5, lspull #24
		orr	r5, r5, r6, lspush #8
		mov	r6, r6, lspull #24
		orr	r6, r6, r7, lspush #8
		mov	r7, r7, lspull #24
		orr	r7, r7, r8, lspush #8
		stmia	dst!, {r4, r5, r6, r7}
		adcs	sum, sum, r4
		adcs	sum, sum, r5
		adcs	sum, sum, r6
		adcs	sum, sum, r7
		mov	r4, r8, lspull #24
		sub	ip, ip, #16
		teq	ip, #0
		bne	1b
2:		ands	ip, len, #12
		beq	4f
		tst	ip, #8
		beq	3f
		load2l	r5, r6
		orr	r4, r4, r5, lspush #8
		mov	r5, r5, lspull #24
		orr	r5, r5, r6, lspush #8
		stmia	dst!, {r4, r5}
		adcs	sum, sum, r4
		adcs	sum, sum, r5
		mov	r4, r6, lspull #24
		tst	ip, #4
		beq	4f
3:		load1l	r5
		orr	r4, r4, r5, lspush #8
		str	r4, [dst], #4
		adcs	sum, sum, r4
		mov	r4, r5, lspull #24
4:		ands	len, len, #3
		beq	.Ldone
		mov	r5, r4, get_byte_0
		tst	len, #2
		beq	.Lexit
		strb	r5, [dst], #1
		adcs	sum, sum, r4
		load1l	r4
		mov	r5, r4, get_byte_0
		strb	r5, [dst], #1
		adcs	sum, sum, r4, lspush #24
		mov	r5, r4, get_byte_1
		b	.Lexit
FN_EXIT
FN_EXPORT
Commit	Line	Data
1da177e4 LT	1	/*
	2	* linux/arch/arm/lib/csumpartialcopygeneric.S
	3	*
	4	* Copyright (C) 1995-2001 Russell King
	5	*
	6	* This program is free software; you can redistribute it and/or modify
	7	* it under the terms of the GNU General Public License version 2 as
	8	* published by the Free Software Foundation.
	9	*/
6ebbf2ce	10	#include <asm/assembler.h>
4dd1837d	11	#include <asm/export.h>
1da177e4 LT	12
	13	/*
	14	* unsigned int
	15	* csum_partial_copy_xxx(const char src, char dst, int len, int sum, )
	16	* r0 = src, r1 = dst, r2 = len, r3 = sum
	17	* Returns : r0 = checksum
	18	*
	19	* Note that 'tst' and 'teq' preserve the carry flag.
	20	*/
	21
	22	src .req r0
	23	dst .req r1
	24	len .req r2
	25	sum .req r3
	26
8adbb371	27	.Lzero: mov r0, sum
90303b10	28	load_regs
1da177e4 LT	29
	30	/*
	31	* Align an unaligned destination pointer. We know that
	32	* we have >= 8 bytes here, so we don't need to check
	33	* the length. Note that the source pointer hasn't been
	34	* aligned yet.
	35	*/
8adbb371 NP	36	.Ldst_unaligned:
	37	tst dst, #1
	38	beq .Ldst_16bit
1da177e4 LT	39
	40	load1b ip
	41	sub len, len, #1
	42	adcs sum, sum, ip, put_byte_1 @ update checksum
	43	strb ip, [dst], #1
	44	tst dst, #2
6ebbf2ce	45	reteq lr @ dst is now 32bit aligned
1da177e4	46
8adbb371	47	.Ldst_16bit: load2b r8, ip
1da177e4 LT	48	sub len, len, #2
	49	adcs sum, sum, r8, put_byte_0
	50	strb r8, [dst], #1
	51	adcs sum, sum, ip, put_byte_1
	52	strb ip, [dst], #1
6ebbf2ce	53	ret lr @ dst is now 32bit aligned
1da177e4 LT	54
	55	/*
	56	* Handle 0 to 7 bytes, with any alignment of source and
	57	* destination pointers. Note that when we get here, C = 0
	58	*/
8adbb371 NP	59	.Lless8: teq len, #0 @ check for zero count
8adbb371 NP	60	beq .Lzero
1da177e4 LT	61
	62	/* we must have at least one byte. */
	63	tst dst, #1 @ dst 16-bit aligned
8adbb371	64	beq .Lless8_aligned
1da177e4 LT	65
	66	/* Align dst */
	67	load1b ip
	68	sub len, len, #1
	69	adcs sum, sum, ip, put_byte_1 @ update checksum
	70	strb ip, [dst], #1
	71	tst len, #6
8adbb371	72	beq .Lless8_byteonly
1da177e4 LT	73
	74	1: load2b r8, ip
	75	sub len, len, #2
	76	adcs sum, sum, r8, put_byte_0
	77	strb r8, [dst], #1
	78	adcs sum, sum, ip, put_byte_1
	79	strb ip, [dst], #1
8adbb371 NP	80	.Lless8_aligned:
8adbb371 NP	81	tst len, #6
1da177e4	82	bne 1b
8adbb371	83	.Lless8_byteonly:
1da177e4	84	tst len, #1
8adbb371	85	beq .Ldone
1da177e4 LT	86	load1b r8
	87	adcs sum, sum, r8, put_byte_0 @ update checksum
	88	strb r8, [dst], #1
8adbb371	89	b .Ldone
1da177e4 LT	90
1da177e4 LT	91	FN_ENTRY
1da177e4	92	save_regs
1da177e4 LT	93
1da177e4 LT	94	cmp len, #8 @ Ensure that we have at least
8adbb371	95	blo .Lless8 @ 8 bytes to copy.
1da177e4 LT	96
	97	adds sum, sum, #0 @ C = 0
	98	tst dst, #3 @ Test destination alignment
8adbb371	99	blne .Ldst_unaligned @ align destination, return here
1da177e4 LT	100
	101	/*
	102	* Ok, the dst pointer is now 32bit aligned, and we know
	103	* that we must have more than 4 bytes to copy. Note
	104	* that C contains the carry from the dst alignment above.
	105	*/
	106
	107	tst src, #3 @ Test source alignment
8adbb371	108	bne .Lsrc_not_aligned
1da177e4 LT	109
	110	/* Routine for src & dst aligned */
	111
	112	bics ip, len, #15
	113	beq 2f
	114
	115	1: load4l r4, r5, r6, r7
	116	stmia dst!, {r4, r5, r6, r7}
	117	adcs sum, sum, r4
	118	adcs sum, sum, r5
	119	adcs sum, sum, r6
	120	adcs sum, sum, r7
	121	sub ip, ip, #16
	122	teq ip, #0
	123	bne 1b
	124
	125	2: ands ip, len, #12
	126	beq 4f
	127	tst ip, #8
	128	beq 3f
	129	load2l r4, r5
	130	stmia dst!, {r4, r5}
	131	adcs sum, sum, r4
	132	adcs sum, sum, r5
	133	tst ip, #4
	134	beq 4f
	135
	136	3: load1l r4
	137	str r4, [dst], #4
	138	adcs sum, sum, r4
	139
	140	4: ands len, len, #3
8adbb371	141	beq .Ldone
1da177e4 LT	142	load1l r4
	143	tst len, #2
	144	mov r5, r4, get_byte_0
8adbb371	145	beq .Lexit
d98b90ea	146	adcs sum, sum, r4, lspush #16
1da177e4 LT	147	strb r5, [dst], #1
	148	mov r5, r4, get_byte_1
	149	strb r5, [dst], #1
	150	mov r5, r4, get_byte_2
8adbb371	151	.Lexit: tst len, #1
1da177e4 LT	152	strneb r5, [dst], #1
	153	andne r5, r5, #255
	154	adcnes sum, sum, r5, put_byte_0
	155
	156	/*
	157	* If the dst pointer was not 16-bit aligned, we
	158	* need to rotate the checksum here to get around
	159	* the inefficient byte manipulations in the
	160	* architecture independent code.
	161	*/
8adbb371	162	.Ldone: adc r0, sum, #0
1da177e4 LT	163	ldr sum, [sp, #0] @ dst
	164	tst sum, #1
	165	movne r0, r0, ror #8
90303b10	166	load_regs
1da177e4	167
8adbb371	168	.Lsrc_not_aligned:
1da177e4 LT	169	adc sum, sum, #0 @ include C from dst alignment
	170	and ip, src, #3
	171	bic src, src, #3
	172	load1l r5
	173	cmp ip, #2
8adbb371 NP	174	beq .Lsrc2_aligned
8adbb371 NP	175	bhi .Lsrc3_aligned
d98b90ea	176	mov r4, r5, lspull #8 @ C = 0
1da177e4 LT	177	bics ip, len, #15
	178	beq 2f
	179	1: load4l r5, r6, r7, r8
d98b90ea VK	180	orr r4, r4, r5, lspush #24
	181	mov r5, r5, lspull #8
	182	orr r5, r5, r6, lspush #24
	183	mov r6, r6, lspull #8
	184	orr r6, r6, r7, lspush #24
	185	mov r7, r7, lspull #8
	186	orr r7, r7, r8, lspush #24
1da177e4 LT	187	stmia dst!, {r4, r5, r6, r7}
	188	adcs sum, sum, r4
	189	adcs sum, sum, r5
	190	adcs sum, sum, r6
	191	adcs sum, sum, r7
d98b90ea	192	mov r4, r8, lspull #8
1da177e4 LT	193	sub ip, ip, #16
	194	teq ip, #0
	195	bne 1b
	196	2: ands ip, len, #12
	197	beq 4f
	198	tst ip, #8
	199	beq 3f
	200	load2l r5, r6
d98b90ea VK	201	orr r4, r4, r5, lspush #24
	202	mov r5, r5, lspull #8
	203	orr r5, r5, r6, lspush #24
1da177e4 LT	204	stmia dst!, {r4, r5}
	205	adcs sum, sum, r4
	206	adcs sum, sum, r5
d98b90ea	207	mov r4, r6, lspull #8
1da177e4 LT	208	tst ip, #4
	209	beq 4f
	210	3: load1l r5
d98b90ea	211	orr r4, r4, r5, lspush #24
1da177e4 LT	212	str r4, [dst], #4
1da177e4 LT	213	adcs sum, sum, r4
d98b90ea	214	mov r4, r5, lspull #8
1da177e4	215	4: ands len, len, #3
8adbb371	216	beq .Ldone
1da177e4 LT	217	mov r5, r4, get_byte_0
1da177e4 LT	218	tst len, #2
8adbb371	219	beq .Lexit
d98b90ea	220	adcs sum, sum, r4, lspush #16
1da177e4 LT	221	strb r5, [dst], #1
	222	mov r5, r4, get_byte_1
	223	strb r5, [dst], #1
	224	mov r5, r4, get_byte_2
8adbb371	225	b .Lexit
1da177e4	226
d98b90ea	227	.Lsrc2_aligned: mov r4, r5, lspull #16
1da177e4 LT	228	adds sum, sum, #0
	229	bics ip, len, #15
	230	beq 2f
	231	1: load4l r5, r6, r7, r8
d98b90ea VK	232	orr r4, r4, r5, lspush #16
	233	mov r5, r5, lspull #16
	234	orr r5, r5, r6, lspush #16
	235	mov r6, r6, lspull #16
	236	orr r6, r6, r7, lspush #16
	237	mov r7, r7, lspull #16
	238	orr r7, r7, r8, lspush #16
1da177e4 LT	239	stmia dst!, {r4, r5, r6, r7}
	240	adcs sum, sum, r4
	241	adcs sum, sum, r5
	242	adcs sum, sum, r6
	243	adcs sum, sum, r7
d98b90ea	244	mov r4, r8, lspull #16
1da177e4 LT	245	sub ip, ip, #16
	246	teq ip, #0
	247	bne 1b
	248	2: ands ip, len, #12
	249	beq 4f
	250	tst ip, #8
	251	beq 3f
	252	load2l r5, r6
d98b90ea VK	253	orr r4, r4, r5, lspush #16
	254	mov r5, r5, lspull #16
	255	orr r5, r5, r6, lspush #16
1da177e4 LT	256	stmia dst!, {r4, r5}
	257	adcs sum, sum, r4
	258	adcs sum, sum, r5
d98b90ea	259	mov r4, r6, lspull #16
1da177e4 LT	260	tst ip, #4
	261	beq 4f
	262	3: load1l r5
d98b90ea	263	orr r4, r4, r5, lspush #16
1da177e4 LT	264	str r4, [dst], #4
1da177e4 LT	265	adcs sum, sum, r4
d98b90ea	266	mov r4, r5, lspull #16
1da177e4	267	4: ands len, len, #3
8adbb371	268	beq .Ldone
1da177e4 LT	269	mov r5, r4, get_byte_0
1da177e4 LT	270	tst len, #2
8adbb371	271	beq .Lexit
1da177e4 LT	272	adcs sum, sum, r4
	273	strb r5, [dst], #1
	274	mov r5, r4, get_byte_1
	275	strb r5, [dst], #1
	276	tst len, #1
8adbb371	277	beq .Ldone
1da177e4	278	load1b r5
8adbb371	279	b .Lexit
1da177e4	280
d98b90ea	281	.Lsrc3_aligned: mov r4, r5, lspull #24
1da177e4 LT	282	adds sum, sum, #0
	283	bics ip, len, #15
	284	beq 2f
	285	1: load4l r5, r6, r7, r8
d98b90ea VK	286	orr r4, r4, r5, lspush #8
	287	mov r5, r5, lspull #24
	288	orr r5, r5, r6, lspush #8
	289	mov r6, r6, lspull #24
	290	orr r6, r6, r7, lspush #8
	291	mov r7, r7, lspull #24
	292	orr r7, r7, r8, lspush #8
1da177e4 LT	293	stmia dst!, {r4, r5, r6, r7}
	294	adcs sum, sum, r4
	295	adcs sum, sum, r5
	296	adcs sum, sum, r6
	297	adcs sum, sum, r7
d98b90ea	298	mov r4, r8, lspull #24
1da177e4 LT	299	sub ip, ip, #16
	300	teq ip, #0
	301	bne 1b
	302	2: ands ip, len, #12
	303	beq 4f
	304	tst ip, #8
	305	beq 3f
	306	load2l r5, r6
d98b90ea VK	307	orr r4, r4, r5, lspush #8
	308	mov r5, r5, lspull #24
	309	orr r5, r5, r6, lspush #8
1da177e4 LT	310	stmia dst!, {r4, r5}
	311	adcs sum, sum, r4
	312	adcs sum, sum, r5
d98b90ea	313	mov r4, r6, lspull #24
1da177e4 LT	314	tst ip, #4
	315	beq 4f
	316	3: load1l r5
d98b90ea	317	orr r4, r4, r5, lspush #8
1da177e4 LT	318	str r4, [dst], #4
1da177e4 LT	319	adcs sum, sum, r4
d98b90ea	320	mov r4, r5, lspull #24
1da177e4	321	4: ands len, len, #3
8adbb371	322	beq .Ldone
1da177e4 LT	323	mov r5, r4, get_byte_0
1da177e4 LT	324	tst len, #2
8adbb371	325	beq .Lexit
1da177e4 LT	326	strb r5, [dst], #1
	327	adcs sum, sum, r4
	328	load1l r4
	329	mov r5, r4, get_byte_0
	330	strb r5, [dst], #1
d98b90ea	331	adcs sum, sum, r4, lspush #24
1da177e4	332	mov r5, r4, get_byte_1
8adbb371	333	b .Lexit
93ed3970	334	FN_EXIT
4dd1837d	335	FN_EXPORT