1 /* Optimized strcmp implementation for PowerPC64.
2 Copyright (C) 2003-2014 Free Software Foundation, Inc.
3 This file is part of the GNU C Library.
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.
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.
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 <http://www.gnu.org/licenses/>. */
21 /* See strlen.s for comments on how the end-of-string testing works. */
23 /* int [r3] strncmp (const char *s1 [r3], const char *s2 [r4], size_t size [r5]) */
25 EALIGN (strncmp, 4, 0)
30 #define rSTR1 r3 /* first string arg */
31 #define rSTR2 r4 /* second string arg */
32 #define rN r5 /* max string length */
33 #define rWORD1 r6 /* current word in s1 */
34 #define rWORD2 r7 /* current word in s2 */
35 #define rFEFE r8 /* constant 0xfefefefefefefeff (-0x0101010101010101) */
36 #define r7F7F r9 /* constant 0x7f7f7f7f7f7f7f7f */
37 #define rNEG r10 /* ~(word in s1 | 0x7f7f7f7f7f7f7f7f) */
38 #define rBITDIF r11 /* bits that differ in s1 & s2 words */
45 clrldi. rTMP, rTMP, 61
49 /* We are doubleword aligned so set up for two loops. first a double word
50 loop, then fall into the byte loop if any residual. */
53 addi rFEFE, rFEFE, -0x101
54 addi r7F7F, r7F7F, 0x7f7f
58 mtctr rTMP /* Power4 wants mtctr 1st in dispatch group. */
62 insrdi r7F7F, r7F7F, 32, 0
63 add rFEFE, rFEFE, rTMP
68 bne- cr1, L(different)
70 L(g1): add rTMP, rFEFE, rWORD1
71 nor rNEG, r7F7F, rWORD1
74 cmpd cr1, rWORD1, rWORD2
77 /* OK. We've hit the end of the string. We need to be careful that
78 we don't compare two strings as different because of gunk beyond
79 the end of the strings... */
81 #ifdef __LITTLE_ENDIAN__
85 andc rTMP2, rTMP2, rTMP
86 rldimi rTMP2, rTMP2, 1, 0
87 and rWORD2, rWORD2, rTMP2 /* Mask off gunk. */
88 and rWORD1, rWORD1, rTMP2
89 cmpd cr1, rWORD1, rWORD2
91 xor rBITDIF, rWORD1, rWORD2 /* rBITDIF has bits that differ. */
93 and rNEG, rNEG, rBITDIF /* rNEG has LS bit that differs. */
94 cntlzd rNEG, rNEG /* bitcount of the bit. */
95 andi. rNEG, rNEG, 56 /* bitcount to LS byte that differs. */
96 sld rWORD1, rWORD1, rNEG /* shift left to clear MS bytes. */
97 sld rWORD2, rWORD2, rNEG
98 xor. rBITDIF, rWORD1, rWORD2
99 sub rRTN, rWORD1, rWORD2
101 sradi rRTN, rRTN, 63 /* must return an int. */
110 xor rBITDIF, rWORD1, rWORD2 /* rBITDIF has bits that differ. */
112 and rNEG, rNEG, rBITDIF /* rNEG has LS bit that differs. */
113 cntlzd rNEG, rNEG /* bitcount of the bit. */
114 andi. rNEG, rNEG, 56 /* bitcount to LS byte that differs. */
115 sld rWORD1, rWORD1, rNEG /* shift left to clear MS bytes. */
116 sld rWORD2, rWORD2, rNEG
117 xor. rBITDIF, rWORD1, rWORD2
118 sub rRTN, rWORD1, rWORD2
124 sradi rRTN, rWORD2, 63
130 and rTMP, r7F7F, rWORD1
132 add rTMP, rTMP, r7F7F
133 xor. rBITDIF, rWORD1, rWORD2
134 andc rNEG, rNEG, rTMP
136 cntlzd rBITDIF, rBITDIF
139 cmpd cr1, rNEG, rBITDIF
140 sub rRTN, rWORD1, rWORD2
142 sradi rRTN, rRTN, 63 /* must return an int. */
151 xor. rBITDIF, rWORD1, rWORD2
152 sub rRTN, rWORD1, rWORD2
158 sradi rRTN, rWORD2, 63
163 /* Oh well. In this case, we just do a byte-by-byte comparison. */
166 and. rTMP, rTMP, rNEG
167 cmpd cr1, rWORD1, rWORD2
170 bne- cr1, L(different)
174 mtctr rN /* Power4 wants mtctr 1st in dispatch group */
186 lbzu rWORD2, 1(rSTR2)
192 lbzu rWORD1, 1(rSTR1)
194 lbzu rWORD2, 1(rSTR2)
199 lbzu rWORD1, 1(rSTR1)
202 L(u2): lbzu rWORD1, -1(rSTR1)
203 L(u3): sub rRTN, rWORD1, rWORD2
206 libc_hidden_builtin_def (strncmp)