projects / thirdparty / glibc.git / blame
| Commit | Line | Data |
|---|---|---|
| 8f5ca04b RM |
1 | /* stpcpy -- copy SRC to DEST returning the address of the terminating '\0' |
| 2 | in DEST. | |
| 3 | For Intel 80x86, x>=3. | |
| cccda09f | 4 | Copyright (C) 1994, 1995, 1996 Free Software Foundation, Inc. |
| 8f5ca04b RM |
5 | Contributed by Ulrich Drepper (drepper@gnu.ai.mit.edu). |
| 6 | This file is part of the GNU C Library. | |
| 7 | ||
| 8 | The GNU C Library is free software; you can redistribute it and/or | |
| 9 | modify it under the terms of the GNU Library General Public License as | |
| 10 | published by the Free Software Foundation; either version 2 of the | |
| 11 | License, or (at your option) any later version. | |
| 12 | ||
| 13 | The GNU C Library is distributed in the hope that it will be useful, | |
| 14 | but WITHOUT ANY WARRANTY; without even the implied warranty of | |
| 15 | MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU | |
| 16 | Library General Public License for more details. | |
| 17 | ||
| 18 | You should have received a copy of the GNU Library General Public | |
| 19 | License along with the GNU C Library; see the file COPYING.LIB. If | |
| 20 | not, write to the Free Software Foundation, Inc., 59 Temple Place - Suite 330, | |
| 21 | Boston, MA 02111-1307, USA. */ | |
| 22 | ||
| 23 | /* This function is defined neither in ANSI nor POSIX standards but is | |
| 24 | also not invented here. */ | |
| 25 | ||
| 26 | #include <sysdep.h> | |
| 27 | #include "asm-syntax.h" | |
| 28 | ||
| 29 | /* | |
| 30 | INPUT PARAMETERS: | |
| 31 | dest (sp + 4) | |
| 32 | src (sp + 8) | |
| 33 | */ | |
| 34 | ||
| 35 | .text | |
| 36 | ENTRY (__stpcpy) | |
| 37 | movl 4(%esp), %eax /* load destination pointer */ | |
| 38 | movl 8(%esp), %ecx /* load source pointer */ | |
| 39 | ||
| 40 | subl %eax, %ecx /* magic: reduce number of loop variants | |
| 41 | to one using addressing mode */ | |
| 42 | ||
| 43 | /* Here we would like to write | |
| 44 | ||
| 45 | subl $4, %eax | |
| 46 | ALIGN (4) | |
| 47 | ||
| 48 | but the assembler is too smart and optimizes for the shortest | |
| 49 | form where the number only needs one byte. But if we could | |
| 50 | have the long form we would not need the alignment. */ | |
| 51 | ||
| 52 | .byte 0x81, 0xe8 /* This is `subl $0x00000004, %eax' */ | |
| 53 | .long 0x00000004 | |
| 54 | ||
| 55 | /* Four times unfolded loop with only one loop counter. This | |
| 56 | is achieved by the use of index+base adressing mode. As the | |
| 57 | loop counter we use the destination address because this is | |
| 58 | also the result. */ | |
| 59 | L1: addl $4, %eax /* increment loop counter */ | |
| 60 | ||
| 61 | movb (%eax,%ecx), %dl /* load current char */ | |
| 62 | movb %dl, (%eax) /* and store it */ | |
| 63 | testb %dl, %dl /* was it NUL? */ | |
| 64 | jz L2 /* yes, then exit */ | |
| 65 | ||
| 66 | movb 1(%eax,%ecx), %dl /* load current char */ | |
| 67 | movb %dl, 1(%eax) /* and store it */ | |
| 68 | testb %dl, %dl /* was it NUL? */ | |
| 69 | jz L3 /* yes, then exit */ | |
| 70 | ||
| 71 | movb 2(%eax,%ecx), %dl /* load current char */ | |
| 72 | movb %dl, 2(%eax) /* and store it */ | |
| 73 | testb %dl, %dl /* was it NUL? */ | |
| 74 | jz L4 /* yes, then exit */ | |
| 75 | ||
| 76 | movb 3(%eax,%ecx), %dl /* load current char */ | |
| 77 | movb %dl, 3(%eax) /* and store it */ | |
| 78 | testb %dl, %dl /* was it NUL? */ | |
| 79 | jnz L1 /* no, then continue loop */ | |
| 80 | ||
| 81 | incl %eax /* correct loop counter */ | |
| 82 | L4: incl %eax | |
| 83 | L3: incl %eax | |
| 84 | L2: | |
| 85 | ret | |
| 6ed0492f | 86 | END (__stpcpy) |
| 8f5ca04b RM |
87 | |
| 88 | weak_alias (__stpcpy, stpcpy) |