1 /* memset/bzero -- set memory area to CH/0
2 Highly optimized version for ix86, x>=5.
3 Copyright (C) 1996-2020 Free Software Foundation, Inc.
4 This file is part of the GNU C Library.
5 Contributed by Torbjorn Granlund, <tege@matematik.su.se>
7 The GNU C Library is free software; you can redistribute it and/or
8 modify it under the terms of the GNU Lesser General Public
9 License as published by the Free Software Foundation; either
10 version 2.1 of the License, or (at your option) any later version.
12 The GNU C Library is distributed in the hope that it will be useful,
13 but WITHOUT ANY WARRANTY; without even the implied warranty of
14 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
15 Lesser General Public License for more details.
17 You should have received a copy of the GNU Lesser General Public
18 License along with the GNU C Library; if not, see
19 <https://www.gnu.org/licenses/>. */
22 #include "asm-syntax.h"
24 #define PARMS 4+4 /* space for 1 saved reg */
35 #if defined SHARED && IS_IN (libc) && !defined USE_AS_BZERO
39 jb HIDDEN_JUMPTARGET (__chk_fail)
45 cfi_adjust_cfa_offset (4)
48 cfi_rel_offset (edi, 0)
51 xorl %eax, %eax /* we fill with 0 */
61 /* If less than 36 bytes to write, skip tricky code (it wouldn't work). */
63 movl %edx, %ecx /* needed when branch is taken! */
66 /* First write 0-3 bytes to make the pointer 32-bit aligned. */
67 movl %edi, %ecx /* Copy ptr to ecx... */
68 negl %ecx /* ...and negate that and... */
69 andl $3, %ecx /* ...mask to get byte count. */
70 subl %ecx, %edx /* adjust global byte count */
74 subl $32, %edx /* offset count for unrolled loop */
75 movl (%edi), %ecx /* Fetch destination cache line */
77 .align 2, 0x90 /* supply 0x90 for broken assemblers */
78 L(1): movl 28(%edi), %ecx /* allocate cache line for destination */
79 subl $32, %edx /* decr loop count */
80 movl %eax, 0(%edi) /* store words pairwise */
88 leal 32(%edi), %edi /* update destination pointer */
91 leal 32(%edx), %ecx /* reset offset count */
93 /* Write last 0-7 full 32-bit words (up to 8 words if loop was skipped). */
94 L(2): shrl $2, %ecx /* convert byte count to longword count */
98 /* Finally write the last 0-3 bytes. */
105 /* Load result (only if used as memset). */
106 movl DEST(%esp), %eax /* start address of destination is result */
109 cfi_adjust_cfa_offset (-4)
114 libc_hidden_builtin_def (memset)