1 /* Optimized memchr implementation for PowerPC32/POWER7 using cmpb insn.
2 Copyright (C) 2010-2014 Free Software Foundation, Inc.
3 Contributed by Luis Machado <luisgpm@br.ibm.com>.
4 This file is part of the GNU C Library.
6 The GNU C Library is free software; you can redistribute it and/or
7 modify it under the terms of the GNU Lesser General Public
8 License as published by the Free Software Foundation; either
9 version 2.1 of the License, or (at your option) any later version.
11 The GNU C Library is distributed in the hope that it will be useful,
12 but WITHOUT ANY WARRANTY; without even the implied warranty of
13 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
14 Lesser General Public License for more details.
16 You should have received a copy of the GNU Lesser General Public
17 License along with the GNU C Library; if not, see
18 <http://www.gnu.org/licenses/>. */
22 /* int [r3] memchr (char *s [r3], int byte [r4], int size [r5]) */
28 insrwi r4,r4,8,16 /* Replicate byte to word. */
29 add r7,r3,r5 /* Calculate the last acceptable address. */
33 rlwinm r6,r3,3,27,28 /* Calculate padding. */
35 #ifdef __LITTLE_ENDIAN__
42 lwz r12,0(r8) /* Load word from memory. */
43 cmpb r3,r12,r4 /* Check for BYTEs in WORD1. */
45 clrlwi r5,r7,30 /* Byte count - 1 in last word. */
46 clrrwi r7,r7,2 /* Address of last word. */
47 cmplwi cr7,r3,0 /* If r3 == 0, no BYTEs have been found. */
51 /* Are we now aligned to a doubleword boundary? If so, skip to
52 the main loop. Otherwise, go through the alignment code. */
55 /* Handle WORD2 of pair. */
62 /* The last word we want to read in the loop below is the one
63 containing the last byte of the string, ie. the word at
64 (s + size - 1) & ~3, or r7. The first word read is at
65 r8 + 4, we read 2 * cnt words, so the last word read will
66 be at r8 + 4 + 8 * cnt - 4. Solving for cnt gives
67 cnt = (r7 - r8) / 8 */
69 srwi r6,r6,3 /* Number of loop iterations. */
70 mtctr r6 /* Setup the counter. */
72 /* Main loop to look for BYTE in the string. Since
73 it's a small loop (8 instructions), align it to 32-bytes. */
76 /* Load two words, compare and merge in a
77 single register for speed. This is an attempt
78 to speed up the byte-checking process for bigger strings. */
83 or r6,r9,r3 /* Merge everything in one word. */
88 /* We may have one more dword to read. */
100 /* OK, one (or both) of the words contains BYTE. Check
101 the first word and decrement the address in case the first
102 word really contains BYTE. */
107 /* BYTE must be in the second word. Adjust the address
108 again and move the result of cmpb to r3 so we can calculate the
114 /* r3 has the output of the cmpb instruction, that is, it contains
115 0xff in the same position as BYTE in the original
116 word from the string. Use that to calculate the pointer.
117 We need to make sure BYTE is *before* the end of the range. */
119 #ifdef __LITTLE_ENDIAN__
122 popcntw r0,r0 /* Count trailing zeros. */
124 cntlzw r0,r3 /* Count leading zeros before the match. */
126 cmplw r8,r7 /* Are we on the last word? */
127 srwi r0,r0,3 /* Convert leading/trailing zeros to bytes. */
129 cmplw cr7,r0,r5 /* If on the last dword, check byte offset. */
140 /* Deals with size <= 16. */
145 lwz r12,0(r8) /* Load word from memory. */
146 cmpb r3,r12,r4 /* Check for BYTE in DWORD1. */
149 clrlwi r5,r7,30 /* Byte count - 1 in last word. */
150 clrrwi r7,r7,2 /* Address of last word. */
151 cmplw r8,r7 /* Are we done already? */
183 weak_alias (__memchr, memchr)
184 libc_hidden_builtin_def (memchr)