1/*
2   strchrnul - find a character or nul in a string
3
4   Copyright (c) 2014, ARM Limited
5   All rights Reserved.
6
7   Redistribution and use in source and binary forms, with or without
8   modification, are permitted provided that the following conditions are met:
9       * Redistributions of source code must retain the above copyright
10         notice, this list of conditions and the following disclaimer.
11       * Redistributions in binary form must reproduce the above copyright
12         notice, this list of conditions and the following disclaimer in the
13         documentation and/or other materials provided with the distribution.
14       * Neither the name of the company nor the names of its contributors
15         may be used to endorse or promote products derived from this
16         software without specific prior written permission.
17
18   THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
19   "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
20   LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
21   A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
22   HOLDER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
23   SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
24   LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
25   DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
26   THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
27   (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
28   OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.  */
29
30/* Assumptions:
31 *
32 * ARMv8-a, AArch64
33 * Neon Available.
34 */
35
36/* Arguments and results.  */
37#define srcin		x0
38#define chrin		w1
39
40#define result		x0
41
42#define src		x2
43#define	tmp1		x3
44#define wtmp2		w4
45#define tmp3		x5
46
47#define vrepchr		v0
48#define vdata1		v1
49#define vdata2		v2
50#define vhas_nul1	v3
51#define vhas_nul2	v4
52#define vhas_chr1	v5
53#define vhas_chr2	v6
54#define vrepmask	v7
55#define vend1		v16
56
57/* Core algorithm.
58
59   For each 32-byte hunk we calculate a 64-bit syndrome value, with
60   two bits per byte (LSB is always in bits 0 and 1, for both big
61   and little-endian systems).  For each tuple, bit 0 is set iff
62   the relevant byte matched the requested character or nul.  Since the
63   bits in the syndrome reflect exactly the order in which things occur
64   in the original string a count_trailing_zeros() operation will
65   identify exactly which byte is causing the termination.  */
66
67/* Locals and temporaries.  */
68
69	.macro def_fn f p2align=0
70	.text
71	.p2align \p2align
72	.global \f
73	.type \f, %function
74\f:
75	.endm
76
77def_fn strchrnul
78	/* Magic constant 0x40100401 to allow us to identify which lane
79	   matches the termination condition.  */
80	mov	wtmp2, #0x0401
81	movk	wtmp2, #0x4010, lsl #16
82	dup	vrepchr.16b, chrin
83	bic	src, srcin, #31		/* Work with aligned 32-byte hunks.  */
84	dup	vrepmask.4s, wtmp2
85	ands	tmp1, srcin, #31
86	b.eq	.Lloop
87
88	/* Input string is not 32-byte aligned.  Rather than forcing
89	   the padding bytes to a safe value, we calculate the syndrome
90	   for all the bytes, but then mask off those bits of the
91	   syndrome that are related to the padding.  */
92	ld1	{vdata1.16b, vdata2.16b}, [src], #32
93	neg	tmp1, tmp1
94	cmeq	vhas_nul1.16b, vdata1.16b, #0
95	cmeq	vhas_chr1.16b, vdata1.16b, vrepchr.16b
96	cmeq	vhas_nul2.16b, vdata2.16b, #0
97	cmeq	vhas_chr2.16b, vdata2.16b, vrepchr.16b
98	orr	vhas_chr1.16b, vhas_chr1.16b, vhas_nul1.16b
99	orr	vhas_chr2.16b, vhas_chr2.16b, vhas_nul2.16b
100	and	vhas_chr1.16b, vhas_chr1.16b, vrepmask.16b
101	and	vhas_chr2.16b, vhas_chr2.16b, vrepmask.16b
102	lsl	tmp1, tmp1, #1
103	addp	vend1.16b, vhas_chr1.16b, vhas_chr2.16b	// 256->128
104	mov	tmp3, #~0
105	addp	vend1.16b, vend1.16b, vend1.16b		// 128->64
106	lsr	tmp1, tmp3, tmp1
107
108	mov	tmp3, vend1.d[0]
109	bic	tmp1, tmp3, tmp1	// Mask padding bits.
110	cbnz	tmp1, .Ltail
111
112.Lloop:
113	ld1	{vdata1.16b, vdata2.16b}, [src], #32
114	cmeq	vhas_nul1.16b, vdata1.16b, #0
115	cmeq	vhas_chr1.16b, vdata1.16b, vrepchr.16b
116	cmeq	vhas_nul2.16b, vdata2.16b, #0
117	cmeq	vhas_chr2.16b, vdata2.16b, vrepchr.16b
118	/* Use a fast check for the termination condition.  */
119	orr	vhas_chr1.16b, vhas_nul1.16b, vhas_chr1.16b
120	orr	vhas_chr2.16b, vhas_nul2.16b, vhas_chr2.16b
121	orr	vend1.16b, vhas_chr1.16b, vhas_chr2.16b
122	addp	vend1.2d, vend1.2d, vend1.2d
123	mov	tmp1, vend1.d[0]
124	cbz	tmp1, .Lloop
125
126	/* Termination condition found.  Now need to establish exactly why
127	   we terminated.  */
128	and	vhas_chr1.16b, vhas_chr1.16b, vrepmask.16b
129	and	vhas_chr2.16b, vhas_chr2.16b, vrepmask.16b
130	addp	vend1.16b, vhas_chr1.16b, vhas_chr2.16b		// 256->128
131	addp	vend1.16b, vend1.16b, vend1.16b		// 128->64
132
133	mov	tmp1, vend1.d[0]
134.Ltail:
135	/* Count the trailing zeros, by bit reversing...  */
136	rbit	tmp1, tmp1
137	/* Re-bias source.  */
138	sub	src, src, #32
139	clz	tmp1, tmp1	/* ... and counting the leading zeros.  */
140	/* tmp1 is twice the offset into the fragment.  */
141	add	result, src, tmp1, lsr #1
142	ret
143
144	.size	strchrnul, . - strchrnul
145