sysdeps/powerpc/powerpc32/405/strcpy.S

   1 /* Optimized strcpy implementation for PowerPC476.
   2    Copyright (C) 2010-2015 Free Software Foundation, Inc.
   3    This file is part of the GNU C Library.
   4
   5    The GNU C Library is free software; you can redistribute it and/or
   6    modify it under the terms of the GNU Lesser General Public
   7    License as published by the Free Software Foundation; either
   8    version 2.1 of the License, or (at your option) any later version.
   9
  10    The GNU C Library is distributed in the hope that it will be useful,
  11    but WITHOUT ANY WARRANTY; without even the implied warranty of
  12    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
  13    Lesser General Public License for more details.
  14
  15    You should have received a copy of the GNU Lesser General Public
  16    License along with the GNU C Library.  If not, see
  17    <http://www.gnu.org/licenses/>.  */
  18
  19 #include <sysdep.h>
  20
  21 /* strcpy
  22
  23        Register Use
  24        r3:destination and return address
  25        r4:source address
  26        r10:temp destination address
  27
  28        Implementation description
  29        Loop by checking 2 words at a time, with dlmzb. Check if there is a null
  30        in the 2 words. If there is a null jump to end checking to determine
  31        where in the last 8 bytes it is. Copy the appropriate bytes of the last
  32        8 according to the null position. */
  33
  34 EALIGN (strcpy, 5, 0)
  35        neg     r7,r4
  36        subi    r4,r4,1
  37        clrlwi. r8,r7,29
  38        subi    r10,r3,1
  39        beq     L(pre_word8_loop)
  40        mtctr   r8
  41
  42 L(loop):
  43        lbzu    r5,0x01(r4)
  44        cmpi    cr5,r5,0x0
  45        stbu    r5,0x01(r10)
  46        beq     cr5,L(end_strcpy)
  47        bdnz    L(loop)
  48
  49 L(pre_word8_loop):
  50        subi    r4,r4,3
  51        subi    r10,r10,3
  52
  53 L(word8_loop):
  54        lwzu    r5,0x04(r4)
  55        lwzu    r6,0x04(r4)
  56        dlmzb.  r11,r5,r6
  57        bne     L(byte_copy)
  58        stwu    r5,0x04(r10)
  59        stwu    r6,0x04(r10)
  60        lwzu    r5,0x04(r4)
  61        lwzu    r6,0x04(r4)
  62        dlmzb.  r11,r5,r6
  63        bne     L(byte_copy)
  64        stwu    r5,0x04(r10)
  65        stwu    r6,0x04(r10)
  66        lwzu    r5,0x04(r4)
  67        lwzu    r6,0x04(r4)
  68        dlmzb.  r11,r5,r6
  69        bne     L(byte_copy)
  70        stwu    r5,0x04(r10)
  71        stwu    r6,0x04(r10)
  72        lwzu    r5,0x04(r4)
  73        lwzu    r6,0x04(r4)
  74        dlmzb.  r11,r5,r6
  75        bne     L(byte_copy)
  76        stwu    r5,0x04(r10)
  77        stwu    r6,0x04(r10)
  78        b       L(word8_loop)
  79
  80 L(last_bytes_copy):
  81        stwu    r5,0x04(r10)
  82        subi    r11,r11,4
  83        mtctr   r11
  84        addi    r10,r10,3
  85        subi    r4,r4,1
  86
  87 L(last_bytes_copy_loop):
  88        lbzu    r5,0x01(r4)
  89        stbu    r5,0x01(r10)
  90        bdnz    L(last_bytes_copy_loop)
  91        blr
  92
  93 L(byte_copy):
  94        blt     L(last_bytes_copy)
  95        mtctr   r11
  96        addi    r10,r10,3
  97        subi    r4,r4,5
  98
  99 L(last_bytes_copy_loop2):
 100        lbzu    r5,0x01(r4)
 101        stbu    r5,0x01(r10)
 102        bdnz    L(last_bytes_copy_loop2)
 103
 104 L(end_strcpy):
 105        blr
 106 END (strcpy)
 107 libc_hidden_builtin_def (strcpy)