]>
Commit | Line | Data |
---|---|---|
cf150d45 | 1 | /* Vector optimized 32/64 bit S/390 version of strchr. |
688903eb | 2 | Copyright (C) 2015-2018 Free Software Foundation, Inc. |
cf150d45 SL |
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 | <http://www.gnu.org/licenses/>. */ | |
18 | ||
19 | #if defined HAVE_S390_VX_ASM_SUPPORT && IS_IN (libc) | |
20 | ||
21 | # include "sysdep.h" | |
22 | # include "asm-syntax.h" | |
23 | ||
24 | .text | |
25 | ||
26 | /* char *strchr (const char *s, int c) | |
27 | Locate character in string. | |
28 | ||
29 | Register usage: | |
30 | -r1=tmp | |
31 | -r2=s | |
32 | -r3=c | |
33 | -r4=tmp | |
34 | -r5=current_len | |
35 | -v16=part of s | |
36 | -v17=index of unequal | |
37 | -v18=replicated c | |
38 | */ | |
39 | ENTRY(__strchr_vx) | |
40 | .machine "z13" | |
41 | .machinemode "zarch_nohighgprs" | |
42 | ||
43 | vlbb %v16,0(%r2),6 /* Load s until next 4k-byte boundary. */ | |
44 | lcbb %r1,0(%r2),6 /* Get bytes to 4k-byte boundary or 16. */ | |
45 | ||
46 | lghi %r5,0 /* current_len = 0. */ | |
47 | ||
48 | vlvgb %v18,%r3,0 /* Generate vector which elements are all c. | |
49 | If c > 255, c will be truncated. */ | |
50 | vrepb %v18,%v18,0 | |
51 | ||
52 | vfeezbs %v16,%v16,%v18 /* Find element equal with zero search. */ | |
53 | vlgvb %r4,%v16,7 /* Load byte index of character or zero. */ | |
54 | clrjl %r4,%r1,.Lfound /* Return if c/zero is in loaded bytes. */ | |
55 | ||
56 | /* Align s to 16 byte. */ | |
57 | risbgn %r4,%r2,60,128+63,0 /* %r3 = bits 60-63 of %r2 'and' 15. */ | |
58 | lghi %r5,16 /* current_len = 16. */ | |
59 | slr %r5,%r4 /* Compute bytes to 16bytes boundary. */ | |
60 | ||
61 | /* Find c/zero in 16 byte aligned loop */ | |
62 | .Lloop: | |
63 | vl %v16,0(%r5,%r2) /* Load s. */ | |
64 | vfeezbs %v16,%v16,%v18 /* Find element equal with zero search. */ | |
65 | jno .Lfound /* Found c/zero (cc=0|1|2). */ | |
66 | vl %v16,16(%r5,%r2) | |
67 | vfeezbs %v16,%v16,%v18 | |
68 | jno .Lfound16 | |
69 | vl %v16,32(%r5,%r2) | |
70 | vfeezbs %v16,%v16,%v18 | |
71 | jno .Lfound32 | |
72 | vl %v16,48(%r5,%r2) | |
73 | vfeezbs %v16,%v16,%v18 | |
74 | jno .Lfound48 | |
75 | ||
76 | aghi %r5,64 | |
77 | j .Lloop /* No character and no zero -> loop. */ | |
78 | ||
79 | .Lfound48: | |
80 | la %r5,16(%r5) /* Use la since aghi would clobber cc. */ | |
81 | .Lfound32: | |
82 | la %r5,16(%r5) | |
83 | .Lfound16: | |
84 | la %r5,16(%r5) | |
85 | .Lfound: | |
86 | je .Lzero /* Found zero, but no c before that zero. */ | |
87 | ||
88 | .Lcharacter: | |
89 | vlgvb %r4,%v16,7 /* Load byte index of character. */ | |
90 | algr %r5,%r4 | |
91 | la %r2,0(%r5,%r2) /* Return pointer to character. */ | |
92 | br %r14 | |
93 | ||
94 | .Lzero: | |
95 | llgcr %r3,%r3 /* char c_char = (char) c. */ | |
96 | clije %r3,0,.Lcharacter /* Found zero and c is zero. */ | |
97 | lghi %r2,0 /* Return null if character not found. */ | |
98 | br %r14 | |
99 | END(__strchr_vx) | |
100 | #endif /* HAVE_S390_VX_ASM_SUPPORT && IS_IN (libc) */ |