178 lines
6.3 KiB
ArmAsm
178 lines
6.3 KiB
ArmAsm
/*******************************************************************************
|
|
*
|
|
* Copyright (c) 1993 Intel Corporation
|
|
*
|
|
* Intel hereby grants you permission to copy, modify, and distribute this
|
|
* software and its documentation. Intel grants this permission provided
|
|
* that the above copyright notice appears in all copies and that both the
|
|
* copyright notice and this permission notice appear in supporting
|
|
* documentation. In addition, Intel grants this permission provided that
|
|
* you prominently mark as "not part of the original" any modifications
|
|
* made to this software or documentation, and that the name of Intel
|
|
* Corporation not be used in advertising or publicity pertaining to
|
|
* distribution of the software or the documentation without specific,
|
|
* written prior permission.
|
|
*
|
|
* Intel Corporation provides this AS IS, WITHOUT ANY WARRANTY, EXPRESS OR
|
|
* IMPLIED, INCLUDING, WITHOUT LIMITATION, ANY WARRANTY OF MERCHANTABILITY
|
|
* OR FITNESS FOR A PARTICULAR PURPOSE. Intel makes no guarantee or
|
|
* representations regarding the use of, or the results of the use of,
|
|
* the software and documentation in terms of correctness, accuracy,
|
|
* reliability, currentness, or otherwise; and you rely on the software,
|
|
* documentation and results solely at your own risk.
|
|
*
|
|
* IN NO EVENT SHALL INTEL BE LIABLE FOR ANY LOSS OF USE, LOSS OF BUSINESS,
|
|
* LOSS OF PROFITS, INDIRECT, INCIDENTAL, SPECIAL OR CONSEQUENTIAL DAMAGES
|
|
* OF ANY KIND. IN NO EVENT SHALL INTEL'S TOTAL LIABILITY EXCEED THE SUM
|
|
* PAID TO INTEL FOR THE PRODUCT LICENSED HEREUNDER.
|
|
*
|
|
******************************************************************************/
|
|
|
|
.file "strcpy.s"
|
|
#ifdef __PIC
|
|
.pic
|
|
#endif
|
|
#ifdef __PID
|
|
.pid
|
|
#endif
|
|
/*
|
|
* (c) copyright 1988,1993 Intel Corp., all rights reserved
|
|
*/
|
|
/*
|
|
procedure strcpy (optimized assembler version for the 80960K series)
|
|
procedure strcat (optimized assembler version for the 80960K series)
|
|
|
|
dest_addr = strcpy (dest_addr, src_addr)
|
|
|
|
copy the null terminated string pointed to by src_addr to
|
|
the string space pointed to by dest_addr. Return the original
|
|
dest_addr.
|
|
|
|
This routine will fail if the source and destination string
|
|
overlap (in particular, if the end of the source is overlapped
|
|
by the beginning of the destination). The behavior is undefined.
|
|
This is acceptable according to the draft C standard.
|
|
|
|
Undefined behavior will also occur if the end of the source string
|
|
(i.e. the terminating null byte) is in the last two words of the
|
|
program's allocated memory space. This is so because strcpy fetches
|
|
ahead. Disallowing the fetch ahead would impose a severe performance
|
|
penalty.
|
|
|
|
Strategy:
|
|
|
|
Fetch the source string and store the destination string by words
|
|
until the null byte is encountered. When the word with the null
|
|
byte is reached, store it by bytes up through the null byte only.
|
|
|
|
Tactics:
|
|
|
|
1) Do NOT try to fetch and store the words in a word aligned manner
|
|
because, in my judgement, the performance degradation experienced due
|
|
to non-aligned accesses does NOT outweigh the time and complexity added
|
|
by the preamble and convoluted body that would be necessary to assure
|
|
alignment. This is supported by the intuition that most source and
|
|
destination strings will be word aligned to begin with.
|
|
|
|
|
|
procedure strcat
|
|
|
|
dest_addr = strcat (dest_addr, src_addr)
|
|
|
|
Appends the string pointed to by src_addr to the string pointed
|
|
to by dest_addr. The first character of the source string is
|
|
copied to the location initially occupied by the trailing null
|
|
byte of the destination string. Thereafter, characters are copied
|
|
from the source to the destination up thru the null byte that
|
|
trails the source string.
|
|
|
|
See the strcpy routine, above, for its caveats, as they apply here too.
|
|
|
|
Strategy:
|
|
|
|
Skip to the end (null byte) of the destination string, and then drop
|
|
into the strcpy code.
|
|
|
|
Tactics:
|
|
|
|
Skipping to the null byte is Ldone by reading the destination string
|
|
in long-words and scanbyte'ing them, then examining the bytes of the
|
|
word that contains the null byte, until the address of the null byte is
|
|
known. Then we drop into the strcpy routine. It is probable (approx.
|
|
three out of four times) that the destination string as strcpy sees
|
|
it will NOT be word aligned (i.e. that the null byte won't be the
|
|
last byte of a word). But it is not worth the complication to that
|
|
routine to force word aligned memory accesses to be gaurenteed.
|
|
*/
|
|
.globl _strcpy, _strcat
|
|
.globl __strcpy, __strcat
|
|
.leafproc _strcpy,__strcpy
|
|
.leafproc _strcat,__strcat
|
|
.align 2
|
|
_strcat:
|
|
#ifndef __PIC
|
|
lda Lrett,g14
|
|
#else
|
|
lda Lrett-(.+8)(ip),g14
|
|
#endif
|
|
__strcat:
|
|
mov g14,g13 # preserve return address
|
|
ldl (g0),g4 # fetch first two words
|
|
addo 8,g0,g2 # post-increment src word pointer
|
|
lda 0xff,g3 # byte extraction mask
|
|
|
|
Lsearch_for_word_with_null_byte:
|
|
scanbyte 0,g4 # check for null byte
|
|
mov g5,g7 # copy second word
|
|
bo.f Lsearch_for_null # branch if null found
|
|
scanbyte 0,g7 # check for null byte
|
|
ldl (g2),g4 # fetch next pair of word of src
|
|
addo 8,g2,g2 # post-increment src word pointer
|
|
bno Lsearch_for_word_with_null_byte # branch if null not found yet
|
|
|
|
subo 4,g2,g2 # back up the byte pointer
|
|
mov g7,g4 # move word with null to search word
|
|
Lsearch_for_null:
|
|
subo 9,g2,g5 # back up the byte pointer
|
|
Lsearch_for_null.a:
|
|
and g4,g3,g6 # extract byte
|
|
cmpo 0,g6 # is it null?
|
|
addo 1,g5,g5 # bump src byte ptr
|
|
shro 8,g4,g4 # shift word to position next byte
|
|
bne Lsearch_for_null.a
|
|
b Lend_of_dest_found
|
|
|
|
_strcpy:
|
|
#ifndef __PIC
|
|
lda Lrett,g14
|
|
#else
|
|
lda Lrett-(.+8)(ip),g14
|
|
#endif
|
|
__strcpy:
|
|
mov g0, g5
|
|
Lend_of_dest_found:
|
|
ld (g1), g2 # fetch first word of source
|
|
mov g14,g6 # preserve return address
|
|
lda 0xff, g3 # byte extraction mask = 0xff;
|
|
Lwloop: # word copying loop
|
|
addo 4, g1, g1 # post-increment source ptr
|
|
scanbyte 0, g2 # does source word contain null byte?
|
|
mov g2, g4 # save a copy of the source word
|
|
be Lcloop # branch if null present
|
|
ld (g1), g2 # pre-fetch next word of source
|
|
st g4, (g5) # store current word
|
|
addo 4, g5, g5 # post-increment dest ptr
|
|
b Lwloop
|
|
|
|
Lcloop: # character copying loop
|
|
and g3, g4, g14 # extract next char
|
|
shro 8, g4, g4 # position word for next byte extraction
|
|
cmpo 0, g14 # is it null?
|
|
stob g14, (g5) # store the byte
|
|
addo 1, g5, g5 # post-increment dest ptr
|
|
bne Lcloop # quit if null encountered
|
|
|
|
bx (g6) # g0 = dest string address; g14 = 0
|
|
Lrett:
|
|
ret
|