1999-01-16 17:09:04 +00:00
|
|
|
/* rawmemchr (str, ch) -- Return pointer to first occurrence of CH in STR.
|
|
|
|
For Intel 80x86, x>=3.
|
2021-01-02 19:32:25 +00:00
|
|
|
Copyright (C) 1994-2021 Free Software Foundation, Inc.
|
1999-01-16 17:09:04 +00:00
|
|
|
This file is part of the GNU C Library.
|
|
|
|
Contributed by Ulrich Drepper <drepper@gnu.ai.mit.edu>
|
|
|
|
Optimised a little by Alan Modra <Alan@SPRI.Levels.UniSA.Edu.Au>
|
|
|
|
This version is developed using the same algorithm as the fast C
|
|
|
|
version which carries the following introduction:
|
|
|
|
Based on strlen implementation by Torbjorn Granlund (tege@sics.se),
|
|
|
|
with help from Dan Sahlin (dan@sics.se) and
|
|
|
|
commentary by Jim Blandy (jimb@ai.mit.edu);
|
|
|
|
adaptation to memchr suggested by Dick Karpinski (dick@cca.ucsf.edu),
|
|
|
|
and implemented by Roland McGrath (roland@ai.mit.edu).
|
|
|
|
|
|
|
|
The GNU C Library is free software; you can redistribute it and/or
|
2001-07-06 04:58:11 +00:00
|
|
|
modify it under the terms of the GNU Lesser General Public
|
|
|
|
License as published by the Free Software Foundation; either
|
|
|
|
version 2.1 of the License, or (at your option) any later version.
|
1999-01-16 17:09:04 +00:00
|
|
|
|
|
|
|
The GNU C Library is distributed in the hope that it will be useful,
|
|
|
|
but WITHOUT ANY WARRANTY; without even the implied warranty of
|
|
|
|
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
|
2001-07-06 04:58:11 +00:00
|
|
|
Lesser General Public License for more details.
|
1999-01-16 17:09:04 +00:00
|
|
|
|
2001-07-06 04:58:11 +00:00
|
|
|
You should have received a copy of the GNU Lesser General Public
|
2012-02-09 23:18:22 +00:00
|
|
|
License along with the GNU C Library; if not, see
|
Prefer https to http for gnu.org and fsf.org URLs
Also, change sources.redhat.com to sourceware.org.
This patch was automatically generated by running the following shell
script, which uses GNU sed, and which avoids modifying files imported
from upstream:
sed -ri '
s,(http|ftp)(://(.*\.)?(gnu|fsf|sourceware)\.org($|[^.]|\.[^a-z])),https\2,g
s,(http|ftp)(://(.*\.)?)sources\.redhat\.com($|[^.]|\.[^a-z]),https\2sourceware.org\4,g
' \
$(find $(git ls-files) -prune -type f \
! -name '*.po' \
! -name 'ChangeLog*' \
! -path COPYING ! -path COPYING.LIB \
! -path manual/fdl-1.3.texi ! -path manual/lgpl-2.1.texi \
! -path manual/texinfo.tex ! -path scripts/config.guess \
! -path scripts/config.sub ! -path scripts/install-sh \
! -path scripts/mkinstalldirs ! -path scripts/move-if-change \
! -path INSTALL ! -path locale/programs/charmap-kw.h \
! -path po/libc.pot ! -path sysdeps/gnu/errlist.c \
! '(' -name configure \
-execdir test -f configure.ac -o -f configure.in ';' ')' \
! '(' -name preconfigure \
-execdir test -f preconfigure.ac ';' ')' \
-print)
and then by running 'make dist-prepare' to regenerate files built
from the altered files, and then executing the following to cleanup:
chmod a+x sysdeps/unix/sysv/linux/riscv/configure
# Omit irrelevant whitespace and comment-only changes,
# perhaps from a slightly-different Autoconf version.
git checkout -f \
sysdeps/csky/configure \
sysdeps/hppa/configure \
sysdeps/riscv/configure \
sysdeps/unix/sysv/linux/csky/configure
# Omit changes that caused a pre-commit check to fail like this:
# remote: *** error: sysdeps/powerpc/powerpc64/ppc-mcount.S: trailing lines
git checkout -f \
sysdeps/powerpc/powerpc64/ppc-mcount.S \
sysdeps/unix/sysv/linux/s390/s390-64/syscall.S
# Omit change that caused a pre-commit check to fail like this:
# remote: *** error: sysdeps/sparc/sparc64/multiarch/memcpy-ultra3.S: last line does not end in newline
git checkout -f sysdeps/sparc/sparc64/multiarch/memcpy-ultra3.S
2019-09-07 05:40:42 +00:00
|
|
|
<https://www.gnu.org/licenses/>. */
|
1999-01-16 17:09:04 +00:00
|
|
|
|
|
|
|
#include <sysdep.h>
|
|
|
|
#include "asm-syntax.h"
|
|
|
|
|
2013-02-21 22:21:52 +00:00
|
|
|
#define PARMS 4+4 /* space for 1 saved reg */
|
2000-06-09 06:14:39 +00:00
|
|
|
#define RTN PARMS
|
2013-02-21 22:21:52 +00:00
|
|
|
#define STR RTN
|
|
|
|
#define CHR STR+4
|
1999-01-16 17:09:04 +00:00
|
|
|
|
|
|
|
.text
|
2013-02-21 22:21:52 +00:00
|
|
|
ENTRY (__rawmemchr)
|
2000-06-09 06:14:39 +00:00
|
|
|
|
1999-01-16 17:09:04 +00:00
|
|
|
/* Save callee-safe register used in this function. */
|
|
|
|
pushl %edi
|
2005-05-04 06:22:30 +00:00
|
|
|
cfi_adjust_cfa_offset (4)
|
|
|
|
cfi_rel_offset (edi, 0)
|
1999-01-16 17:09:04 +00:00
|
|
|
|
|
|
|
/* Load parameters into registers. */
|
2000-06-09 06:14:39 +00:00
|
|
|
movl STR(%esp), %eax
|
|
|
|
movl CHR(%esp), %edx
|
1999-01-16 17:09:04 +00:00
|
|
|
|
|
|
|
/* At the moment %edx contains C. What we need for the
|
|
|
|
algorithm is C in all bytes of the dword. Avoid
|
|
|
|
operations on 16 bit words because these require an
|
|
|
|
prefix byte (and one more cycle). */
|
|
|
|
movb %dl, %dh /* Now it is 0|0|c|c */
|
|
|
|
movl %edx, %ecx
|
|
|
|
shll $16, %edx /* Now c|c|0|0 */
|
|
|
|
movw %cx, %dx /* And finally c|c|c|c */
|
|
|
|
|
|
|
|
/* Better performance can be achieved if the word (32
|
|
|
|
bit) memory access is aligned on a four-byte-boundary.
|
|
|
|
So process first bytes one by one until boundary is
|
|
|
|
reached. Don't use a loop for better performance. */
|
|
|
|
|
|
|
|
testb $3, %al /* correctly aligned ? */
|
|
|
|
je L(1) /* yes => begin loop */
|
|
|
|
cmpb %dl, (%eax) /* compare byte */
|
|
|
|
je L(9) /* target found => return */
|
|
|
|
incl %eax /* increment source pointer */
|
|
|
|
|
|
|
|
testb $3, %al /* correctly aligned ? */
|
|
|
|
je L(1) /* yes => begin loop */
|
|
|
|
cmpb %dl, (%eax) /* compare byte */
|
|
|
|
je L(9) /* target found => return */
|
|
|
|
incl %eax /* increment source pointer */
|
|
|
|
|
|
|
|
testb $3, %al /* correctly aligned ? */
|
|
|
|
je L(1) /* yes => begin loop */
|
|
|
|
cmpb %dl, (%eax) /* compare byte */
|
|
|
|
je L(9) /* target found => return */
|
|
|
|
incl %eax /* increment source pointer */
|
|
|
|
|
|
|
|
/* We exit the loop if adding MAGIC_BITS to LONGWORD fails to
|
|
|
|
change any of the hole bits of LONGWORD.
|
|
|
|
|
|
|
|
1) Is this safe? Will it catch all the zero bytes?
|
|
|
|
Suppose there is a byte with all zeros. Any carry bits
|
|
|
|
propagating from its left will fall into the hole at its
|
|
|
|
least significant bit and stop. Since there will be no
|
|
|
|
carry from its most significant bit, the LSB of the
|
|
|
|
byte to the left will be unchanged, and the zero will be
|
|
|
|
detected.
|
|
|
|
|
|
|
|
2) Is this worthwhile? Will it ignore everything except
|
|
|
|
zero bytes? Suppose every byte of LONGWORD has a bit set
|
|
|
|
somewhere. There will be a carry into bit 8. If bit 8
|
|
|
|
is set, this will carry into bit 16. If bit 8 is clear,
|
|
|
|
one of bits 9-15 must be set, so there will be a carry
|
|
|
|
into bit 16. Similarly, there will be a carry into bit
|
|
|
|
24. If one of bits 24-31 is set, there will be a carry
|
|
|
|
into bit 32 (=carry flag), so all of the hole bits will
|
|
|
|
be changed.
|
|
|
|
|
|
|
|
3) But wait! Aren't we looking for C, not zero?
|
|
|
|
Good point. So what we do is XOR LONGWORD with a longword,
|
|
|
|
each of whose bytes is C. This turns each byte that is C
|
|
|
|
into a zero. */
|
|
|
|
|
|
|
|
|
|
|
|
/* Each round the main loop processes 16 bytes. */
|
|
|
|
ALIGN (4)
|
|
|
|
|
|
|
|
L(1): movl (%eax), %ecx /* get word (= 4 bytes) in question */
|
|
|
|
movl $0xfefefeff, %edi /* magic value */
|
|
|
|
xorl %edx, %ecx /* XOR with word c|c|c|c => bytes of str == c
|
|
|
|
are now 0 */
|
|
|
|
addl %ecx, %edi /* add the magic value to the word. We get
|
|
|
|
carry bits reported for each byte which
|
|
|
|
is *not* 0 */
|
|
|
|
|
|
|
|
/* According to the algorithm we had to reverse the effect of the
|
|
|
|
XOR first and then test the overflow bits. But because the
|
|
|
|
following XOR would destroy the carry flag and it would (in a
|
|
|
|
representation with more than 32 bits) not alter then last
|
|
|
|
overflow, we can now test this condition. If no carry is signaled
|
|
|
|
no overflow must have occurred in the last byte => it was 0. */
|
|
|
|
jnc L(8)
|
|
|
|
|
|
|
|
/* We are only interested in carry bits that change due to the
|
|
|
|
previous add, so remove original bits */
|
|
|
|
xorl %ecx, %edi /* ((word^charmask)+magic)^(word^charmask) */
|
|
|
|
|
|
|
|
/* Now test for the other three overflow bits. */
|
|
|
|
orl $0xfefefeff, %edi /* set all non-carry bits */
|
|
|
|
incl %edi /* add 1: if one carry bit was *not* set
|
|
|
|
the addition will not result in 0. */
|
|
|
|
|
|
|
|
/* If at least one byte of the word is C we don't get 0 in %edi. */
|
|
|
|
jnz L(8) /* found it => return pointer */
|
|
|
|
|
|
|
|
/* This process is unfolded four times for better performance.
|
|
|
|
we don't increment the source pointer each time. Instead we
|
|
|
|
use offsets and increment by 16 in each run of the loop. But
|
|
|
|
before probing for the matching byte we need some extra code
|
|
|
|
(following LL(13) below). Even the len can be compared with
|
|
|
|
constants instead of decrementing each time. */
|
|
|
|
|
|
|
|
movl 4(%eax), %ecx /* get word (= 4 bytes) in question */
|
|
|
|
movl $0xfefefeff, %edi /* magic value */
|
|
|
|
xorl %edx, %ecx /* XOR with word c|c|c|c => bytes of str == c
|
|
|
|
are now 0 */
|
|
|
|
addl %ecx, %edi /* add the magic value to the word. We get
|
|
|
|
carry bits reported for each byte which
|
|
|
|
is *not* 0 */
|
|
|
|
jnc L(7) /* highest byte is C => return pointer */
|
|
|
|
xorl %ecx, %edi /* ((word^charmask)+magic)^(word^charmask) */
|
|
|
|
orl $0xfefefeff, %edi /* set all non-carry bits */
|
|
|
|
incl %edi /* add 1: if one carry bit was *not* set
|
|
|
|
the addition will not result in 0. */
|
|
|
|
jnz L(7) /* found it => return pointer */
|
|
|
|
|
|
|
|
movl 8(%eax), %ecx /* get word (= 4 bytes) in question */
|
|
|
|
movl $0xfefefeff, %edi /* magic value */
|
|
|
|
xorl %edx, %ecx /* XOR with word c|c|c|c => bytes of str == c
|
|
|
|
are now 0 */
|
|
|
|
addl %ecx, %edi /* add the magic value to the word. We get
|
|
|
|
carry bits reported for each byte which
|
|
|
|
is *not* 0 */
|
|
|
|
jnc L(6) /* highest byte is C => return pointer */
|
|
|
|
xorl %ecx, %edi /* ((word^charmask)+magic)^(word^charmask) */
|
|
|
|
orl $0xfefefeff, %edi /* set all non-carry bits */
|
|
|
|
incl %edi /* add 1: if one carry bit was *not* set
|
|
|
|
the addition will not result in 0. */
|
|
|
|
jnz L(6) /* found it => return pointer */
|
|
|
|
|
|
|
|
movl 12(%eax), %ecx /* get word (= 4 bytes) in question */
|
|
|
|
movl $0xfefefeff, %edi /* magic value */
|
|
|
|
xorl %edx, %ecx /* XOR with word c|c|c|c => bytes of str == c
|
|
|
|
are now 0 */
|
|
|
|
addl %ecx, %edi /* add the magic value to the word. We get
|
|
|
|
carry bits reported for each byte which
|
|
|
|
is *not* 0 */
|
|
|
|
jnc L(5) /* highest byte is C => return pointer */
|
|
|
|
xorl %ecx, %edi /* ((word^charmask)+magic)^(word^charmask) */
|
|
|
|
orl $0xfefefeff, %edi /* set all non-carry bits */
|
|
|
|
incl %edi /* add 1: if one carry bit was *not* set
|
|
|
|
the addition will not result in 0. */
|
|
|
|
jnz L(5) /* found it => return pointer */
|
|
|
|
|
|
|
|
/* Adjust both counters for a full round, i.e. 16 bytes. */
|
|
|
|
addl $16, %eax
|
|
|
|
jmp L(1)
|
|
|
|
/* add missing source pointer increments */
|
|
|
|
L(5): addl $4, %eax
|
|
|
|
L(6): addl $4, %eax
|
|
|
|
L(7): addl $4, %eax
|
|
|
|
|
|
|
|
/* Test for the matching byte in the word. %ecx contains a NUL
|
|
|
|
char in the byte which originally was the byte we are looking
|
|
|
|
at. */
|
|
|
|
L(8): testb %cl, %cl /* test first byte in dword */
|
|
|
|
jz L(9) /* if zero => return pointer */
|
|
|
|
incl %eax /* increment source pointer */
|
|
|
|
|
|
|
|
testb %ch, %ch /* test second byte in dword */
|
|
|
|
jz L(9) /* if zero => return pointer */
|
|
|
|
incl %eax /* increment source pointer */
|
|
|
|
|
|
|
|
testl $0xff0000, %ecx /* test third byte in dword */
|
|
|
|
jz L(9) /* if zero => return pointer */
|
|
|
|
incl %eax /* increment source pointer */
|
|
|
|
|
|
|
|
/* No further test needed we we know it is one of the four bytes. */
|
|
|
|
|
2000-06-26 22:15:00 +00:00
|
|
|
L(9):
|
|
|
|
popl %edi /* pop saved register */
|
2005-05-04 06:22:30 +00:00
|
|
|
cfi_adjust_cfa_offset (-4)
|
|
|
|
cfi_restore (edi)
|
1999-01-16 17:09:04 +00:00
|
|
|
|
2013-02-21 22:21:52 +00:00
|
|
|
ret
|
|
|
|
END (__rawmemchr)
|
2000-06-26 22:15:00 +00:00
|
|
|
|
2013-02-21 22:21:52 +00:00
|
|
|
libc_hidden_def (__rawmemchr)
|
|
|
|
weak_alias (__rawmemchr, rawmemchr)
|