sysdeps/powerpc/powerpc64/power7/strcasecmp.S

   1 /* Optimized strcasecmp implementation for PowerPC64.
   2    Copyright (C) 2011-2016 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 #include <locale-defines.h>
  21
  22 /* int [r3] strcasecmp (const char *s1 [r3], const char *s2 [r4] )
  23
  24    or if defined USE_IN_EXTENDED_LOCALE_MODEL:
  25
  26    int [r3] strcasecmp_l (const char *s1 [r3], const char *s2 [r4],
  27                           __locale_t loc [r5]) */
  28
  29 #ifndef STRCMP
  30 # define __STRCMP __strcasecmp
  31 # define STRCMP   strcasecmp
  32 #endif
  33
  34 ENTRY (__STRCMP)
  35 #ifndef USE_IN_EXTENDED_LOCALE_MODEL
  36         CALL_MCOUNT 2
  37 #else
  38         CALL_MCOUNT 3
  39 #endif
  40
  41 #define rRTN    r3      /* Return value */
  42 #define rSTR1   r5      /* 1st string */
  43 #define rSTR2   r4      /* 2nd string */
  44 #define rLOCARG r5      /* 3rd argument: locale_t */
  45 #define rCHAR1  r6      /* Byte read from 1st string */
  46 #define rCHAR2  r7      /* Byte read from 2nd string */
  47 #define rADDR1  r8      /* Address of tolower(rCHAR1) */
  48 #define rADDR2  r12     /* Address of tolower(rCHAR2) */
  49 #define rLWR1   r8      /* Word tolower(rCHAR1) */
  50 #define rLWR2   r12     /* Word tolower(rCHAR2) */
  51 #define rTMP    r9
  52 #define rLOC    r11     /* Default locale address */
  53
  54         cmpd    cr7, r3, r4
  55 #ifndef USE_IN_EXTENDED_LOCALE_MODEL
  56         ld      rTMP, __libc_tsd_LOCALE@got@tprel(r2)
  57         add     rLOC, rTMP, __libc_tsd_LOCALE@tls
  58         ld      rLOC, 0(rLOC)
  59 #else
  60         mr      rLOC, rLOCARG
  61 #endif
  62         ld      rLOC, LOCALE_CTYPE_TOLOWER(rLOC)
  63         mr      rSTR1, rRTN
  64         li      rRTN, 0
  65         beqlr   cr7
  66
  67
  68         /* Unrolling loop for POWER: loads are done with 'lbz' plus
  69         offset and string descriptors are only updated in the end
  70         of loop unrolling. */
  71
  72         lbz     rCHAR1, 0(rSTR1)        /* Load char from s1 */
  73         lbz     rCHAR2, 0(rSTR2)        /* Load char from s2 */
  74 L(loop):
  75         cmpdi   rCHAR1, 0               /* *s1 == '\0' ? */
  76         sldi    rADDR1, rCHAR1, 2       /* Calculate address for tolower(*s1) */
  77         sldi    rADDR2, rCHAR2, 2       /* Calculate address for tolower(*s2) */
  78         lwzx    rLWR1, rLOC, rADDR1     /* Load tolower(*s1) */
  79         lwzx    rLWR2, rLOC, rADDR2     /* Load tolower(*s2) */
  80         cmpw    cr1, rLWR1, rLWR2       /* r = tolower(*s1) == tolower(*s2) ? */
  81         crorc   4*cr1+eq,eq,4*cr1+eq    /* (*s1 != '\0') || (r == 1) */
  82         beq     cr1, L(done)
  83         lbz     rCHAR1, 1(rSTR1)
  84         lbz     rCHAR2, 1(rSTR2)
  85         cmpdi   rCHAR1, 0
  86         sldi    rADDR1, rCHAR1, 2
  87         sldi    rADDR2, rCHAR2, 2
  88         lwzx    rLWR1, rLOC, rADDR1
  89         lwzx    rLWR2, rLOC, rADDR2
  90         cmpw    cr1, rLWR1, rLWR2
  91         crorc   4*cr1+eq,eq,4*cr1+eq
  92         beq     cr1, L(done)
  93         lbz     rCHAR1, 2(rSTR1)
  94         lbz     rCHAR2, 2(rSTR2)
  95         cmpdi   rCHAR1, 0
  96         sldi    rADDR1, rCHAR1, 2
  97         sldi    rADDR2, rCHAR2, 2
  98         lwzx    rLWR1, rLOC, rADDR1
  99         lwzx    rLWR2, rLOC, rADDR2
 100         cmpw    cr1, rLWR1, rLWR2
 101         crorc   4*cr1+eq,eq,4*cr1+eq
 102         beq     cr1, L(done)
 103         lbz     rCHAR1, 3(rSTR1)
 104         lbz     rCHAR2, 3(rSTR2)
 105         cmpdi   rCHAR1, 0
 106         /* Increment both string descriptors */
 107         addi    rSTR1, rSTR1, 4
 108         addi    rSTR2, rSTR2, 4
 109         sldi    rADDR1, rCHAR1, 2
 110         sldi    rADDR2, rCHAR2, 2
 111         lwzx    rLWR1, rLOC, rADDR1
 112         lwzx    rLWR2, rLOC, rADDR2
 113         cmpw    cr1, rLWR1, rLWR2
 114         crorc   4*cr1+eq,eq,4*cr1+eq
 115         beq     cr1,L(done)
 116         lbz     rCHAR1, 0(rSTR1)        /* Load char from s1 */
 117         lbz     rCHAR2, 0(rSTR2)        /* Load char from s2 */
 118         b       L(loop)
 119 L(done):
 120         subf    r0, rLWR2, rLWR1
 121         extsw   rRTN, r0
 122         blr
 123 END (__STRCMP)
 124
 125 weak_alias (__STRCMP, STRCMP)
 126 libc_hidden_builtin_def (__STRCMP)