libs/newlib-cygwin
libs
/
newlib-cygwin
mirror of git://sourceware.org/git/newlib-cygwin.git
4
0
Fork 0
Code Issues Projects Releases Wiki Activity
newlib-cygwin/newlib/libc/machine/i386/memset.S

100 lines
1.4 KiB
ArmAsm
Raw Normal View History

import newlib-2000-02-17 snapshot
2000-02-18 03:39:52 +08:00
/*
* ====================================================
2008-05-26 Eric Blake <ebb9@byu.net> Optimize the generic and x86 memset. * libc/string/memset.c (memset) [!__OPTIMIZE_SIZE__]: Pre-align pointer so unaligned stores aren't penalized. * libc/machine/i386/memset.S (memset): [!__OPTIMIZE_SIZE__]: Pre-align pointer so unaligned stores aren't penalized. Prefer 8-byte over 4-byte alignment. Reduce register pressure.
2008-05-27 07:23:15 +08:00
* Copyright (C) 1998, 2002, 2008 by Red Hat Inc. All rights reserved.
import newlib-2000-02-17 snapshot
2000-02-18 03:39:52 +08:00
*
* Permission to use, copy, modify, and distribute this
* software is freely granted, provided that this notice
* is preserved.
* ====================================================
*/
#include "i386mach.h"
.global SYM (memset)
2000-08-27 Werner Almesberger <Werner.Almesberger@epfl.ch> * libc/machine/i386/i386mach.h: added SOTYPE_FUNCTION to set type of global entry points if _I386MACH_NEED_SOTYPE_FUNCTION is defined; Added __CLI and __STI macros (controlled via _I386MACH_ALLOW_HW_INTERRUPTS macro). * libc/machine/i386/f_atan2.S libc/machine/i386/f_atan2f.S libc/machine/i386/f_frexp.S libc/machine/i386/f_frexpf.S libc/machine/i386/f_ldexp.S libc/machine/i386/f_ldexpf.S libc/machine/i386/f_log.S libc/machine/i386/f_log10.S libc/machine/i386/f_log10f.S libc/machine/i386/f_logf.S libc/machine/i386/f_tan.S libc/machine/i386/f_tanf.S libc/machine/i386/memchr.S libc/machine/i386/memcmp.S libc/machine/i386/memcpy.S libc/machine/i386/memmove.S libc/machine/i386/memset.S libc/machine/i386/setjmp.S libc/machine/i386/strchr.S libc/machine/i386/strlen.S: (that's libc/machine/i386/*.S) added SOTYPE_FUNCTION(symbol) for all global entry points. * libc/machine/i386/setjmp.S: removed code replicated in libc/machine/i386/i386mach.h and included i386mach.h instead; Use __CLI and __STI instead of cli and sti.
2000-08-29 01:50:06 +08:00
SOTYPE_FUNCTION(memset)
import newlib-2000-02-17 snapshot
2000-02-18 03:39:52 +08:00
SYM (memset):
pushl ebp
movl esp,ebp
pushl edi
movl 8(ebp),edi
Fix bug in previous patch. * libc/machine/i386/memset.S (memset): Mask second arg correctly.
2008-05-28 22:00:39 +08:00
movzbl 12(ebp),eax
import newlib-2000-02-17 snapshot
2000-02-18 03:39:52 +08:00
movl 16(ebp),ecx
cld
#ifndef __OPTIMIZE_SIZE__
2008-05-26 Eric Blake <ebb9@byu.net> Optimize the generic and x86 memset. * libc/string/memset.c (memset) [!__OPTIMIZE_SIZE__]: Pre-align pointer so unaligned stores aren't penalized. * libc/machine/i386/memset.S (memset): [!__OPTIMIZE_SIZE__]: Pre-align pointer so unaligned stores aren't penalized. Prefer 8-byte over 4-byte alignment. Reduce register pressure.
2008-05-27 07:23:15 +08:00
/* Less than 16 bytes won't benefit from the 'rep stosl' loop. */
import newlib-2000-02-17 snapshot
2000-02-18 03:39:52 +08:00
cmpl $16,ecx
jbe .L19
2008-05-26 Eric Blake <ebb9@byu.net> Optimize the generic and x86 memset. * libc/string/memset.c (memset) [!__OPTIMIZE_SIZE__]: Pre-align pointer so unaligned stores aren't penalized. * libc/machine/i386/memset.S (memset): [!__OPTIMIZE_SIZE__]: Pre-align pointer so unaligned stores aren't penalized. Prefer 8-byte over 4-byte alignment. Reduce register pressure.
2008-05-27 07:23:15 +08:00
testl $7,edi
je .L10
import newlib-2000-02-17 snapshot
2000-02-18 03:39:52 +08:00
2008-05-26 Eric Blake <ebb9@byu.net> Optimize the generic and x86 memset. * libc/string/memset.c (memset) [!__OPTIMIZE_SIZE__]: Pre-align pointer so unaligned stores aren't penalized. * libc/machine/i386/memset.S (memset): [!__OPTIMIZE_SIZE__]: Pre-align pointer so unaligned stores aren't penalized. Prefer 8-byte over 4-byte alignment. Reduce register pressure.
2008-05-27 07:23:15 +08:00
/* It turns out that 8-byte aligned 'rep stosl' outperforms
4-byte aligned on some x86 platforms. */
movb al,(edi)
incl edi
decl ecx
testl $7,edi
je .L10
movb al,(edi)
incl edi
decl ecx
testl $7,edi
je .L10
movb al,(edi)
incl edi
decl ecx
testl $7,edi
je .L10
movb al,(edi)
incl edi
decl ecx
testl $7,edi
je .L10
import newlib-2000-02-17 snapshot
2000-02-18 03:39:52 +08:00
2008-05-26 Eric Blake <ebb9@byu.net> Optimize the generic and x86 memset. * libc/string/memset.c (memset) [!__OPTIMIZE_SIZE__]: Pre-align pointer so unaligned stores aren't penalized. * libc/machine/i386/memset.S (memset): [!__OPTIMIZE_SIZE__]: Pre-align pointer so unaligned stores aren't penalized. Prefer 8-byte over 4-byte alignment. Reduce register pressure.
2008-05-27 07:23:15 +08:00
movb al,(edi)
incl edi
decl ecx
testl $7,edi
je .L10
movb al,(edi)
incl edi
decl ecx
testl $7,edi
je .L10
movb al,(edi)
incl edi
decl ecx
/* At this point, ecx>8 and edi%8==0. */
.L10:
movb al,ah
import newlib-2000-02-17 snapshot
2000-02-18 03:39:52 +08:00
movl eax,edx
sall $16,edx
orl edx,eax
2008-05-26 Eric Blake <ebb9@byu.net> Optimize the generic and x86 memset. * libc/string/memset.c (memset) [!__OPTIMIZE_SIZE__]: Pre-align pointer so unaligned stores aren't penalized. * libc/machine/i386/memset.S (memset): [!__OPTIMIZE_SIZE__]: Pre-align pointer so unaligned stores aren't penalized. Prefer 8-byte over 4-byte alignment. Reduce register pressure.
2008-05-27 07:23:15 +08:00
movl ecx,edx
import newlib-2000-02-17 snapshot
2000-02-18 03:39:52 +08:00
shrl $2,ecx
2008-05-26 Eric Blake <ebb9@byu.net> Optimize the generic and x86 memset. * libc/string/memset.c (memset) [!__OPTIMIZE_SIZE__]: Pre-align pointer so unaligned stores aren't penalized. * libc/machine/i386/memset.S (memset): [!__OPTIMIZE_SIZE__]: Pre-align pointer so unaligned stores aren't penalized. Prefer 8-byte over 4-byte alignment. Reduce register pressure.
2008-05-27 07:23:15 +08:00
andl $3,edx
import newlib-2000-02-17 snapshot
2000-02-18 03:39:52 +08:00
rep
stosl
2008-05-26 Eric Blake <ebb9@byu.net> Optimize the generic and x86 memset. * libc/string/memset.c (memset) [!__OPTIMIZE_SIZE__]: Pre-align pointer so unaligned stores aren't penalized. * libc/machine/i386/memset.S (memset): [!__OPTIMIZE_SIZE__]: Pre-align pointer so unaligned stores aren't penalized. Prefer 8-byte over 4-byte alignment. Reduce register pressure.
2008-05-27 07:23:15 +08:00
movl edx,ecx
import newlib-2000-02-17 snapshot
2000-02-18 03:39:52 +08:00
#endif /* not __OPTIMIZE_SIZE__ */
2008-05-26 Eric Blake <ebb9@byu.net> Optimize the generic and x86 memset. * libc/string/memset.c (memset) [!__OPTIMIZE_SIZE__]: Pre-align pointer so unaligned stores aren't penalized. * libc/machine/i386/memset.S (memset): [!__OPTIMIZE_SIZE__]: Pre-align pointer so unaligned stores aren't penalized. Prefer 8-byte over 4-byte alignment. Reduce register pressure.
2008-05-27 07:23:15 +08:00
import newlib-2000-02-17 snapshot
2000-02-18 03:39:52 +08:00
.L19:
rep
stosb
movl 8(ebp),eax
2008-05-26 Eric Blake <ebb9@byu.net> Optimize the generic and x86 memset. * libc/string/memset.c (memset) [!__OPTIMIZE_SIZE__]: Pre-align pointer so unaligned stores aren't penalized. * libc/machine/i386/memset.S (memset): [!__OPTIMIZE_SIZE__]: Pre-align pointer so unaligned stores aren't penalized. Prefer 8-byte over 4-byte alignment. Reduce register pressure.
2008-05-27 07:23:15 +08:00
leal -4(ebp),esp
import newlib-2000-02-17 snapshot
2000-02-18 03:39:52 +08:00
popl edi
leave
ret