1/* Vector optimized 32/64 bit S/390 version of stpcpy.
2   Copyright (C) 2015-2021 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-stpcpy.h>
20
21#if HAVE_STPCPY_Z13
22
23# include "sysdep.h"
24# include "asm-syntax.h"
25
26	.text
27
28/* char * stpcpy (const char *dest, const char *src)
29   Copy string src to dest returning a pointer to its end.
30
31   Register usage:
32   -r1=tmp
33   -r2=dest and return value
34   -r3=src
35   -r4=tmp
36   -r5=current_len
37   -v16=part of src
38   -v17=index of zero
39   -v18=part of src
40*/
41ENTRY(STPCPY_Z13)
42	.machine "z13"
43	.machinemode "zarch_nohighgprs"
44
45	vlbb	%v16,0(%r3),6	/* Load s until next 4k-byte boundary.  */
46	lcbb	%r1,0(%r3),6	/* Get bytes to 4k-byte boundary or 16.  */
47
48	vfenezb	%v17,%v16,%v16	/* Find element not equal with zero search.  */
49	vlgvb	%r5,%v17,7	/* Load zero index or 16 if not found.  */
50	clrjl	%r5,%r1,.Lfound_align /* If found zero within loaded bytes,
51					 copy bytes before and return.  */
52
53	/* Align s to 16 byte.  */
54	risbgn	%r4,%r3,60,128+63,0 /* %r3 = bits 60-63 of %r2 'and' 15.  */
55	lghi	%r5,15		/* current_len = 15.  */
56	slr	%r5,%r4		/* Compute highest index to 16byte boundary.  */
57
58	vstl	%v16,%r5,0(%r2)	/* Copy loaded characters - no zero.  */
59	ahi	%r5,1		/* Start loop at next character.  */
60
61	/* Find zero in 16byte aligned loop.  */
62.Lloop:
63	vl	%v16,0(%r5,%r3)	/* Load s.  */
64	vfenezbs %v17,%v16,%v16	/* Find element not equal with zero search.  */
65	je	.Lfound_v16_0	/* Jump away if zero was found.  */
66	vl	%v18,16(%r5,%r3) /* Load next part of s.  */
67	vst	%v16,0(%r5,%r2)	/* Store previous part without zero to dst.  */
68	vfenezbs %v17,%v18,%v18
69	je	.Lfound_v18_16
70	vl	%v16,32(%r5,%r3)
71	vst	%v18,16(%r5,%r2)
72	vfenezbs %v17,%v16,%v16
73	je	.Lfound_v16_32
74	vl	%v18,48(%r5,%r3)
75	vst	%v16,32(%r5,%r2)
76	vfenezbs %v17,%v18,%v18
77	je	.Lfound_v18_48
78	vst	%v18,48(%r5,%r2)
79
80	aghi	%r5,64
81	j	.Lloop		/* No zero found -> loop.  */
82
83.Lfound_v16_32:
84	aghi	%r5,32
85.Lfound_v16_0:
86	la	%r3,0(%r5,%r2)
87	vlgvb	%r1,%v17,7	/* Load byte index of zero.  */
88	vstl	%v16,%r1,0(%r3)	/* Copy characters including zero.  */
89	la	%r2,0(%r1,%r3)	/* Return pointer to zero.  */
90	br	%r14
91
92.Lfound_v18_48:
93	aghi	%r5,32
94.Lfound_v18_16:
95	la	%r3,16(%r5,%r2)
96	vlgvb	%r1,%v17,7	/* Load byte index of zero.  */
97	vstl	%v18,%r1,0(%r3)	/* Copy characters including zero.  */
98	la	%r2,0(%r1,%r3)	/* Return pointer to zero.  */
99	br	%r14
100
101.Lfound_align:
102	vstl	%v16,%r5,0(%r2)	/* Copy characters including zero.  */
103	la	%r2,0(%r5,%r2)	/* Return pointer to zero.  */
104	br	%r14
105END(STPCPY_Z13)
106
107# if ! HAVE_STPCPY_IFUNC
108strong_alias (STPCPY_Z13, __stpcpy)
109weak_alias (__stpcpy, stpcpy)
110# endif
111
112# if ! HAVE_STPCPY_C && defined SHARED && IS_IN (libc)
113strong_alias (STPCPY_Z13, __GI_stpcpy)
114strong_alias (STPCPY_Z13, __GI___stpcpy)
115# endif
116#endif
117