1 | /* Vector optimized 32/64 bit S/390 version of wcslen. |
2 | Copyright (C) 2015-2022 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 <ifunc-wcslen.h> |
20 | #if HAVE_WCSLEN_Z13 |
21 | |
22 | # include "sysdep.h" |
23 | # include "asm-syntax.h" |
24 | |
25 | .text |
26 | |
27 | /* size_t wcslen (const wchar_t *s) |
28 | Returns length of string s. |
29 | |
30 | Register usage: |
31 | -r1=bytes to 4k-byte boundary |
32 | -r2=s |
33 | -r3=tmp |
34 | -r4=tmp |
35 | -r5=current_len and return_value |
36 | -v16=part of s |
37 | */ |
38 | ENTRY(WCSLEN_Z13) |
39 | .machine "z13" |
40 | .machinemode "zarch_nohighgprs" |
41 | |
42 | vlbb %v16,0(%r2),6 /* Load s until next 4k-byte boundary. */ |
43 | lcbb %r1,0(%r2),6 /* Get bytes to 4k-byte boundary or 16. */ |
44 | |
45 | tmll %r2,3 /* Test if s is 4-byte aligned? */ |
46 | jne .Lfallback /* And use common-code variant if not. */ |
47 | |
48 | vfenezf %v16,%v16,%v16 /* Find element not equal with zero search. */ |
49 | vlgvb %r4,%v16,7 /* Load zero index or 16 if not found. */ |
50 | clr %r4,%r1 /* If found zero within loaded bytes? */ |
51 | locgrl %r2,%r4 /* Then copy return value. */ |
52 | jl .Lend /* And return. */ |
53 | |
54 | /* Align s to 16 byte. */ |
55 | risbgn %r3,%r2,60,128+63,0 /* %r3 = bits 60-63 of %r2 'and' 15. */ |
56 | lghi %r5,16 /* current_len = 16. */ |
57 | slr %r5,%r3 /* Compute bytes to 16bytes boundary. */ |
58 | |
59 | /* Find zero in 16byte aligned loop. */ |
60 | .Lloop: |
61 | vl %v16,0(%r5,%r2) /* Load s. */ |
62 | vfenezfs %v16,%v16,%v16 /* Find element not equal with zero search. */ |
63 | je .Lfound /* Jump away if zero was found. */ |
64 | vl %v16,16(%r5,%r2) |
65 | vfenezfs %v16,%v16,%v16 |
66 | je .Lfound16 |
67 | vl %v16,32(%r5,%r2) |
68 | vfenezfs %v16,%v16,%v16 |
69 | je .Lfound32 |
70 | vl %v16,48(%r5,%r2) |
71 | vfenezfs %v16,%v16,%v16 |
72 | je .Lfound48 |
73 | |
74 | aghi %r5,64 |
75 | j .Lloop /* No zero found -> loop. */ |
76 | |
77 | .Lfound48: |
78 | aghi %r5,16 |
79 | .Lfound32: |
80 | aghi %r5,16 |
81 | .Lfound16: |
82 | aghi %r5,16 |
83 | .Lfound: |
84 | vlgvb %r2,%v16,7 /* Load byte index of zero. */ |
85 | algr %r2,%r5 |
86 | .Lend: |
87 | srlg %r2,%r2,2 /* Convert byte-count to character-count. */ |
88 | br %r14 |
89 | .Lfallback: |
90 | jg WCSLEN_C |
91 | END(WCSLEN_Z13) |
92 | |
93 | # if ! HAVE_WCSLEN_IFUNC |
94 | strong_alias (WCSLEN_Z13, __wcslen) |
95 | weak_alias (__wcslen, wcslen) |
96 | # endif |
97 | #endif |
98 | |