1/* SPDX-License-Identifier: GPL-2.0 */
2/*
3 * memscan.S: Optimized memscan for the Sparc.
4 *
5 * Copyright (C) 1996 David S. Miller (davem@caip.rutgers.edu)
6 */
7
8#include <asm/export.h>
9
10/* In essence, this is just a fancy strlen. */
11
12#define LO_MAGIC 0x01010101
13#define HI_MAGIC 0x80808080
14
15	.text
16	.align	4
17	.globl	__memscan_zero, __memscan_generic
18	.globl	memscan
19EXPORT_SYMBOL(__memscan_zero)
20EXPORT_SYMBOL(__memscan_generic)
21__memscan_zero:
22	/* %o0 = addr, %o1 = size */
23	cmp	%o1, 0
24	bne,a	1f
25	 andcc	%o0, 3, %g0
26
27	retl
28	 nop
29
301:
31	be	mzero_scan_word
32	 sethi	%hi(HI_MAGIC), %g2
33
34	ldsb	[%o0], %g3
35mzero_still_not_word_aligned:
36	cmp	%g3, 0
37	bne	1f
38	 add	%o0, 1, %o0
39
40	retl
41	 sub	%o0, 1, %o0
42
431:
44	subcc	%o1, 1, %o1
45	bne,a	1f
46	 andcc	%o0, 3, %g0
47
48	retl
49	 nop
50
511:
52	bne,a	mzero_still_not_word_aligned
53	 ldsb	[%o0], %g3
54
55	sethi	%hi(HI_MAGIC), %g2
56mzero_scan_word:
57	or	%g2, %lo(HI_MAGIC), %o3
58	sethi	%hi(LO_MAGIC), %g3
59	or	%g3, %lo(LO_MAGIC), %o2
60mzero_next_word:
61	ld	[%o0], %g2
62mzero_next_word_preloaded:
63	sub	%g2, %o2, %g2
64mzero_next_word_preloaded_next:
65	andcc	%g2, %o3, %g0
66	bne	mzero_byte_zero
67	 add	%o0, 4, %o0
68
69mzero_check_out_of_fuel:
70	subcc	%o1, 4, %o1
71	bg,a	1f
72	 ld	[%o0], %g2
73
74	retl
75	 nop
76
771:
78	b	mzero_next_word_preloaded_next
79	 sub	%g2, %o2, %g2
80
81	/* Check every byte. */
82mzero_byte_zero:
83	ldsb	[%o0 - 4], %g2
84	cmp	%g2, 0
85	bne	mzero_byte_one
86	 sub	%o0, 4, %g3
87
88	retl
89	 mov	%g3, %o0
90
91mzero_byte_one:
92	ldsb	[%o0 - 3], %g2
93	cmp	%g2, 0
94	bne,a	mzero_byte_two_and_three
95	 ldsb	[%o0 - 2], %g2
96
97	retl
98	 sub	%o0, 3, %o0
99
100mzero_byte_two_and_three:
101	cmp	%g2, 0
102	bne,a	1f
103	 ldsb	[%o0 - 1], %g2
104
105	retl
106	 sub	%o0, 2, %o0
107
1081:
109	cmp	%g2, 0
110	bne,a	mzero_next_word_preloaded
111	 ld	[%o0], %g2
112
113	retl
114	 sub	%o0, 1, %o0
115
116mzero_found_it:
117	retl
118	 sub	%o0, 2, %o0
119
120memscan:
121__memscan_generic:
122	/* %o0 = addr, %o1 = c, %o2 = size */
123	cmp	%o2, 0
124	bne,a	0f
125	 ldub	[%o0], %g2
126
127	b,a	2f
1281:
129	ldub	[%o0], %g2
1300:
131	cmp	%g2, %o1
132	be	2f
133	 addcc	%o2, -1, %o2
134	bne	1b
135	 add	%o0, 1, %o0
1362:
137	retl
138	 nop
139