144 lines
4.7 KiB
ArmAsm
144 lines
4.7 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 "memccpy.s"
|
|
#ifdef __PIC
|
|
.pic
|
|
#endif
|
|
#ifdef __PID
|
|
.pid
|
|
#endif
|
|
/*
|
|
* (c) copyright 1989,1993 Intel Corp., all rights reserved
|
|
*/
|
|
/*
|
|
procedure memccpy (optimized assembler version for the 80960K series)
|
|
|
|
dest_addr = memccpy (dest_addr, src_addr, char, len)
|
|
|
|
copy len bytes pointed to by src_addr to the space pointed to by
|
|
dest_addr, stopping if char is copied. If char is copied,
|
|
return address of byte after char in dest string; else null.
|
|
|
|
|
|
Undefined behavior will occur if the end of the source array is in
|
|
the last two words of the program's allocated memory space. This
|
|
is so because the routine fetches ahead. Disallowing the fetch
|
|
ahead would impose a severe performance penalty.
|
|
|
|
Undefined behavior will also occur if the source and destination
|
|
strings overlap.
|
|
|
|
Strategy:
|
|
|
|
Fetch the source array by words and store them by words to the
|
|
destination array, until there are fewer than three bytes left
|
|
to copy. Then, using the last word of the source (the one that
|
|
contains the remaining 0, 1, 2, or 3 bytes to be copied), store
|
|
a byte at a time until Ldone.
|
|
|
|
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 arrays (even more true of most big source arrays) will
|
|
be word aligned to begin with.
|
|
|
|
2) Rather than decrementing len to zero,
|
|
I calculate the address of the byte after the last byte of the
|
|
destination array, and quit when the destination byte pointer passes
|
|
that.
|
|
|
|
*/
|
|
|
|
.globl _memccpy
|
|
.leafproc _memccpy, __memccpy
|
|
.align 2
|
|
_memccpy:
|
|
#ifndef __PIC
|
|
lda Lrett,g14
|
|
#else
|
|
lda Lrett-(.+8)(ip),g14
|
|
#endif
|
|
__memccpy:
|
|
mov g14, g13 # preserve return address
|
|
cmpibge 0,g3,Lexit_char_not_found
|
|
|
|
addo g3,g1,g3 # compute beyond end of src
|
|
ld (g1), g7 # fetch first word of source
|
|
lda 0xff,g5 # mask for char
|
|
and g5,g2,g2 # extract only char
|
|
shlo 8,g2,g6
|
|
or g2,g6,g6
|
|
shlo 16,g6,g4
|
|
or g6,g4,g6 # word of char
|
|
b Lwloop_b
|
|
|
|
Lwloop_a:
|
|
ld (g1), g7 # fetch ahead next word of source
|
|
st g4, (g0) # store word to dest
|
|
addo 4, g0, g0 # post-increment dest pointer
|
|
Lwloop_b: # word copying loop
|
|
addo 4, g1, g1 # pre-increment src pointer
|
|
cmpo g3, g1 # is len <= 3 ?
|
|
mov g7, g4 # keep a copy of the current word
|
|
bl Lcloop_setup # quit word loop if less than 4 bytes
|
|
scanbyte g6, g7 # check for char
|
|
bno Lwloop_a # continue word loop if char not found.
|
|
|
|
Lcloop_setup:
|
|
subo 4, g1, g1 # back down src pointer
|
|
cmpobe g1, g3, Lexit_char_not_found
|
|
|
|
Lcloop_a: # character copying loop (len < 3)
|
|
and g5,g4,g7 # check the byte against char
|
|
cmpo g7,g2
|
|
stob g7,(g0) # store the byte
|
|
addo 1, g0, g0
|
|
be Lexit_char_found
|
|
addo 1,g1,g1
|
|
cmpo g1,g3
|
|
shro 8,g4,g4 # position next byte
|
|
bne Lcloop_a
|
|
|
|
Lexit_char_not_found:
|
|
mov 0, g0
|
|
Lexit_char_found:
|
|
lda 0,g14
|
|
bx (g13) # g0 = dest array address; g14 = 0
|
|
Lrett:
|
|
ret
|
|
|
|
|
|
/* end of memccpy */
|