1/* memset/bzero -- set memory area to CH/0
2 Optimized version for x86-64.
3 Copyright (C) 2002-2016 Free Software Foundation, Inc.
4 This file is part of the GNU C Library.
5
6 The GNU C Library is free software; you can redistribute it and/or
7 modify it under the terms of the GNU Lesser General Public
8 License as published by the Free Software Foundation; either
9 version 2.1 of the License, or (at your option) any later version.
10
11 The GNU C Library is distributed in the hope that it will be useful,
12 but WITHOUT ANY WARRANTY; without even the implied warranty of
13 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
14 Lesser General Public License for more details.
15
16 You should have received a copy of the GNU Lesser General Public
17 License along with the GNU C Library; if not, see
18 <http://www.gnu.org/licenses/>. */
19
20#include <sysdep.h>
21
22 .text
23#if IS_IN (libc)
24ENTRY(__bzero)
25 movq %rdi, %rax /* Set return value. */
26 movq %rsi, %rdx /* Set n. */
27 pxor %xmm0, %xmm0
28 jmp L(entry_from_bzero)
29END(__bzero)
30weak_alias (__bzero, bzero)
31
32/* Like memset but takes additional parameter with return value. */
33ENTRY(__memset_tail)
34 movq %rcx, %rax /* Set return value. */
35
36 movd %esi, %xmm0
37 punpcklbw %xmm0, %xmm0
38 punpcklwd %xmm0, %xmm0
39 pshufd $0, %xmm0, %xmm0
40
41 jmp L(entry_from_bzero)
42END(__memset_tail)
43#endif
44
45#if defined PIC && IS_IN (libc)
46ENTRY_CHK (__memset_chk)
47 cmpq %rdx, %rcx
48 jb HIDDEN_JUMPTARGET (__chk_fail)
49END_CHK (__memset_chk)
50#endif
51
52ENTRY (memset)
53 movd %esi, %xmm0
54 movq %rdi, %rax
55 punpcklbw %xmm0, %xmm0
56 punpcklwd %xmm0, %xmm0
57 pshufd $0, %xmm0, %xmm0
58L(entry_from_bzero):
59 cmpq $64, %rdx
60 ja L(loop_start)
61 cmpq $16, %rdx
62 jbe L(less_16_bytes)
63 cmpq $32, %rdx
64 movdqu %xmm0, (%rdi)
65 movdqu %xmm0, -16(%rdi,%rdx)
66 ja L(between_32_64_bytes)
67L(return):
68 rep
69 ret
70 .p2align 4
71L(between_32_64_bytes):
72 movdqu %xmm0, 16(%rdi)
73 movdqu %xmm0, -32(%rdi,%rdx)
74 ret
75 .p2align 4
76L(loop_start):
77 leaq 64(%rdi), %rcx
78 movdqu %xmm0, (%rdi)
79 andq $-64, %rcx
80 movdqu %xmm0, -16(%rdi,%rdx)
81 movdqu %xmm0, 16(%rdi)
82 movdqu %xmm0, -32(%rdi,%rdx)
83 movdqu %xmm0, 32(%rdi)
84 movdqu %xmm0, -48(%rdi,%rdx)
85 movdqu %xmm0, 48(%rdi)
86 movdqu %xmm0, -64(%rdi,%rdx)
87 addq %rdi, %rdx
88 andq $-64, %rdx
89 cmpq %rdx, %rcx
90 je L(return)
91 .p2align 4
92L(loop):
93 movdqa %xmm0, (%rcx)
94 movdqa %xmm0, 16(%rcx)
95 movdqa %xmm0, 32(%rcx)
96 movdqa %xmm0, 48(%rcx)
97 addq $64, %rcx
98 cmpq %rcx, %rdx
99 jne L(loop)
100 rep
101 ret
102L(less_16_bytes):
103 movq %xmm0, %rcx
104 testb $24, %dl
105 jne L(between8_16bytes)
106 testb $4, %dl
107 jne L(between4_7bytes)
108 testb $1, %dl
109 je L(odd_byte)
110 movb %cl, (%rdi)
111L(odd_byte):
112 testb $2, %dl
113 je L(return)
114 movw %cx, -2(%rax,%rdx)
115 ret
116L(between4_7bytes):
117 movl %ecx, (%rdi)
118 movl %ecx, -4(%rdi,%rdx)
119 ret
120L(between8_16bytes):
121 movq %rcx, (%rdi)
122 movq %rcx, -8(%rdi,%rdx)
123 ret
124
125END (memset)
126libc_hidden_builtin_def (memset)
127
128#if defined PIC && IS_IN (libc) && !defined USE_MULTIARCH
129strong_alias (__memset_chk, __memset_zero_constant_len_parameter)
130 .section .gnu.warning.__memset_zero_constant_len_parameter
131 .string "memset used with constant zero length parameter; this could be due to transposed parameters"
132#endif
133