119 lines
4.3 KiB
ArmAsm
119 lines
4.3 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 "strcmp.s"
|
||
|
#ifdef __PIC
|
||
|
.pic
|
||
|
#endif
|
||
|
#ifdef __PID
|
||
|
.pid
|
||
|
#endif
|
||
|
/*
|
||
|
* (c) copyright 1988,1993 Intel Corp., all rights reserved
|
||
|
*/
|
||
|
/*
|
||
|
procedure strcmp (optimized assembler version for the 80960K Series)
|
||
|
|
||
|
result = strcmp (src1_addr, src2_addr)
|
||
|
|
||
|
compare the null terminated string pointed to by src1_addr to
|
||
|
the string pointed to by src2_addr. Return 0 iff the strings
|
||
|
are equal, -1 if src1_addr is lexicographically less than src2_addr,
|
||
|
and 1 if it is lexicographically greater.
|
||
|
|
||
|
Undefined behavior will occur if the end of either 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 strcmp fetches
|
||
|
ahead. Disallowing the fetch ahead would impose a severe performance
|
||
|
penalty.
|
||
|
|
||
|
Strategy:
|
||
|
|
||
|
Fetch the source strings by words and compare the words until either
|
||
|
differing words are found or the null byte is encountered. In either
|
||
|
case, move through the word until either the differing byte if found,
|
||
|
in which case return -1 or 1 appropriately; or the null byte is
|
||
|
encountered, in which case, return zero (equality).
|
||
|
|
||
|
Tactics:
|
||
|
|
||
|
1) Do NOT try to fetch 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 many source
|
||
|
strings will be word aligned to begin with.
|
||
|
*/
|
||
|
|
||
|
.globl _strcmp
|
||
|
.globl __strcmp
|
||
|
.leafproc _strcmp,__strcmp
|
||
|
.align 2
|
||
|
|
||
|
_strcmp:
|
||
|
#ifndef __PIC
|
||
|
lda .Lrett,g14
|
||
|
#else
|
||
|
lda .Lrett-(.+8)(ip),g14
|
||
|
#endif
|
||
|
__strcmp:
|
||
|
ld (g0), g5 # fetch first word of source_1
|
||
|
mov g14,g7 # preserve return address
|
||
|
ldconst 0,g14 # conform to register conventions
|
||
|
ldconst 0xff,g4 # byte extraction mask
|
||
|
.Lwloop:
|
||
|
addo 4,g0,g0 # post-increment source_1 byte ptr
|
||
|
ld (g1), g3 # fetch word of source_2
|
||
|
scanbyte 0,g5 # does word have a null byte?
|
||
|
mov g5,g2 # save a copy of the source_1 word
|
||
|
be .Lcloop # branch if null byte encountered
|
||
|
cmpo g2,g3 # are the source words the same?
|
||
|
addo 4,g1,g1 # post-increment source_2 byte ptr
|
||
|
ld (g0), g5 # fetch ahead next word of source_1
|
||
|
be .Lwloop # fall thru if words are unequal
|
||
|
|
||
|
.Lcloop: and g4,g2,g5 # extract and compare individual bytes
|
||
|
and g4,g3,g6
|
||
|
cmpobne g5,g6,.diff # if they differ, go return 1 or -1
|
||
|
cmpo 0,g6 # they are the same. Are they null?
|
||
|
shlo 8,g4,g4 # position mask for next extraction
|
||
|
bne .Lcloop # loop if null not encountered
|
||
|
|
||
|
mov 0,g0 # return equality
|
||
|
bx (g7)
|
||
|
.Lrett:
|
||
|
ret
|
||
|
.diff: bl .neg
|
||
|
mov 1,g0
|
||
|
bx (g7)
|
||
|
.neg: subi 1,0,g0
|
||
|
.Lexit:
|
||
|
bx (g7)
|