1 /* Optimized strcmp implementation for PowerPC64.
2 Copyright (C) 2003-2015 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 */
37 #define rFEFE r8 /* constant 0xfefefefefefefeff (-0x0101010101010101) */
38 #define r7F7F r9 /* constant 0x7f7f7f7f7f7f7f7f */
39 #define rNEG r10 /* ~(word in s1 | 0x7f7f7f7f7f7f7f7f) */
40 #define rBITDIF r11 /* bits that differ in s1 & s2 words */
47 clrldi. rTMP, rTMP, 61
51 /* We are doubleword aligned so set up for two loops. first a double word
52 loop, then fall into the byte loop if any residual. */
55 addi rFEFE, rFEFE, -0x101
56 addi r7F7F, r7F7F, 0x7f7f
60 mtctr rTMP /* Power4 wants mtctr 1st in dispatch group. */
64 insrdi r7F7F, r7F7F, 32, 0
65 add rFEFE, rFEFE, rTMP
70 bne- cr1, L(different)
72 L(g1): add rTMP, rFEFE, rWORD1
73 nor rNEG, r7F7F, rWORD1
76 cmpd cr1, rWORD1, rWORD2
79 /* OK. We've hit the end of the string. We need to be careful that
80 we don't compare two strings as different because of gunk beyond
81 the end of the strings... */
83 #ifdef __LITTLE_ENDIAN__
87 andc rTMP2, rTMP2, rTMP
88 rldimi rTMP2, rTMP2, 1, 0
89 and rWORD2, rWORD2, rTMP2 /* Mask off gunk. */
90 and rWORD1, rWORD1, rTMP2
91 cmpd cr1, rWORD1, rWORD2
93 xor rBITDIF, rWORD1, rWORD2 /* rBITDIF has bits that differ. */
95 and rNEG, rNEG, rBITDIF /* rNEG has LS bit that differs. */
96 cntlzd rNEG, rNEG /* bitcount of the bit. */
97 andi. rNEG, rNEG, 56 /* bitcount to LS byte that differs. */
98 sld rWORD1, rWORD1, rNEG /* shift left to clear MS bytes. */
99 sld rWORD2, rWORD2, rNEG
100 xor. rBITDIF, rWORD1, rWORD2
101 sub rRTN, rWORD1, rWORD2
103 sradi rRTN, rRTN, 63 /* must return an int. */
112 xor rBITDIF, rWORD1, rWORD2 /* rBITDIF has bits that differ. */
114 and rNEG, rNEG, rBITDIF /* rNEG has LS bit that differs. */
115 cntlzd rNEG, rNEG /* bitcount of the bit. */
116 andi. rNEG, rNEG, 56 /* bitcount to LS byte that differs. */
117 sld rWORD1, rWORD1, rNEG /* shift left to clear MS bytes. */
118 sld rWORD2, rWORD2, rNEG
119 xor. rBITDIF, rWORD1, rWORD2
120 sub rRTN, rWORD1, rWORD2
126 sradi rRTN, rWORD2, 63
132 and rTMP, r7F7F, rWORD1
134 add rTMP, rTMP, r7F7F
135 xor. rBITDIF, rWORD1, rWORD2
136 andc rNEG, rNEG, rTMP
138 cntlzd rBITDIF, rBITDIF
141 cmpd cr1, rNEG, rBITDIF
142 sub rRTN, rWORD1, rWORD2
144 sradi rRTN, rRTN, 63 /* must return an int. */
153 xor. rBITDIF, rWORD1, rWORD2
154 sub rRTN, rWORD1, rWORD2
160 sradi rRTN, rWORD2, 63
165 /* Oh well. In this case, we just do a byte-by-byte comparison. */
168 and. rTMP, rTMP, rNEG
169 cmpd cr1, rWORD1, rWORD2
172 bne- cr1, L(different)
176 mtctr rN /* Power4 wants mtctr 1st in dispatch group */
188 lbzu rWORD3, 1(rSTR1)
189 lbzu rWORD4, 1(rSTR2)
195 lbzu rWORD1, 1(rSTR1)
196 lbzu rWORD2, 1(rSTR2)
202 lbzu rWORD3, 1(rSTR1)
203 lbzu rWORD4, 1(rSTR2)
209 lbzu rWORD1, 1(rSTR1)
210 lbzu rWORD2, 1(rSTR2)
213 L(u3): sub rRTN, rWORD3, rWORD4
215 L(u4): sub rRTN, rWORD1, rWORD2
221 libc_hidden_builtin_def (strncmp)