1/* Optimized strlen implementation for PowerPC64/POWER7 using cmpb insn.
2 Copyright (C) 2010-2024 Free Software Foundation, Inc.
3 This file is part of the GNU C Library.
4
5 The GNU C Library is free software; you can redistribute it and/or
6 modify it under the terms of the GNU Lesser General Public
7 License as published by the Free Software Foundation; either
8 version 2.1 of the License, or (at your option) any later version.
9
10 The GNU C Library is distributed in the hope that it will be useful,
11 but WITHOUT ANY WARRANTY; without even the implied warranty of
12 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
13 Lesser General Public License for more details.
14
15 You should have received a copy of the GNU Lesser General Public
16 License along with the GNU C Library; if not, see
17 <https://www.gnu.org/licenses/>. */
18
19#include <sysdep.h>
20
21/* int [r3] strlen (char *s [r3]) */
22
23#ifndef STRLEN
24# define STRLEN strlen
25#endif
26 .machine power7
27ENTRY_TOCLESS (STRLEN)
28 CALL_MCOUNT 1
29 dcbt 0,r3
30 clrrdi r4,r3,3 /* Align the address to doubleword boundary. */
31 rlwinm r6,r3,3,26,28 /* Calculate padding. */
32 li r0,0 /* Doubleword with null chars to use
33 with cmpb. */
34 li r5,-1 /* MASK = 0xffffffffffffffff. */
35 ld r12,0(r4) /* Load doubleword from memory. */
36#ifdef __LITTLE_ENDIAN__
37 sld r5,r5,r6
38#else
39 srd r5,r5,r6 /* MASK = MASK >> padding. */
40#endif
41 orc r9,r12,r5 /* Mask bits that are not part of the string. */
42 cmpb r10,r9,r0 /* Check for null bytes in DWORD1. */
43 cmpdi cr7,r10,0 /* If r10 == 0, no null's have been found. */
44 bne cr7,L(done)
45
46 mtcrf 0x01,r4
47
48 /* Are we now aligned to a quadword boundary? If so, skip to
49 the main loop. Otherwise, go through the alignment code. */
50
51 bt 28,L(loop)
52
53 /* Handle DWORD2 of pair. */
54 ldu r12,8(r4)
55 cmpb r10,r12,r0
56 cmpdi cr7,r10,0
57 bne cr7,L(done)
58
59 /* Main loop to look for the end of the string. Since it's a
60 small loop (< 8 instructions), align it to 32-bytes. */
61 .p2align 5
62L(loop):
63 /* Load two doublewords, compare and merge in a
64 single register for speed. This is an attempt
65 to speed up the null-checking process for bigger strings. */
66
67 ld r12, 8(r4)
68 ldu r11, 16(r4)
69 cmpb r10,r12,r0
70 cmpb r9,r11,r0
71 or r8,r9,r10 /* Merge everything in one doubleword. */
72 cmpdi cr7,r8,0
73 beq cr7,L(loop)
74
75 /* OK, one (or both) of the doublewords contains a null byte. Check
76 the first doubleword and decrement the address in case the first
77 doubleword really contains a null byte. */
78
79 cmpdi cr6,r10,0
80 addi r4,r4,-8
81 bne cr6,L(done)
82
83 /* The null byte must be in the second doubleword. Adjust the address
84 again and move the result of cmpb to r10 so we can calculate the
85 length. */
86
87 mr r10,r9
88 addi r4,r4,8
89
90 /* r10 has the output of the cmpb instruction, that is, it contains
91 0xff in the same position as the null byte in the original
92 doubleword from the string. Use that to calculate the length. */
93L(done):
94#ifdef __LITTLE_ENDIAN__
95 addi r9, r10, -1 /* Form a mask from trailing zeros. */
96 andc r9, r9, r10
97 popcntd r0, r9 /* Count the bits in the mask. */
98#else
99 cntlzd r0,r10 /* Count leading zeros before the match. */
100#endif
101 subf r5,r3,r4
102 srdi r0,r0,3 /* Convert leading/trailing zeros to bytes. */
103 add r3,r5,r0 /* Compute final length. */
104 blr
105END (STRLEN)
106libc_hidden_builtin_def (strlen)
107

source code of glibc/sysdeps/powerpc/powerpc64/power7/strlen.S