sysdeps/arm/armv6t2/strlen.S

   1 /* Copyright (C) 2010-2015 Free Software Foundation, Inc.
   2    This file is part of the GNU C Library.
   3
   4    The GNU C Library is free software; you can redistribute it and/or
   5    modify it under the terms of the GNU Lesser General Public
   6    License as published by the Free Software Foundation; either
   7    version 2.1 of the License, or (at your option) any later version.
   8
   9    The GNU C Library is distributed in the hope that it will be useful,
  10    but WITHOUT ANY WARRANTY; without even the implied warranty of
  11    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
  12    Lesser General Public License for more details.
  13
  14    You should have received a copy of the GNU Lesser General Public
  15    License along with the GNU C Library.  If not, see
  16    <http://www.gnu.org/licenses/>.  */
  17
  18 /*
  19    Assumes:
  20    ARMv6T2, AArch32
  21
  22  */
  23
  24 #include <arm-features.h>               /* This might #define NO_THUMB.  */
  25 #include <sysdep.h>
  26
  27 #ifdef __ARMEB__
  28 #define S2LO            lsl
  29 #define S2HI            lsr
  30 #else
  31 #define S2LO            lsr
  32 #define S2HI            lsl
  33 #endif
  34
  35 #ifndef NO_THUMB
  36 /* This code is best on Thumb.  */
  37         .thumb
  38 #else
  39 /* Using bne.w explicitly is desirable in Thumb mode because it helps
  40    align the following label without a nop.  In ARM mode there is no
  41    such difference.  */
  42 .macro bne.w label
  43         bne \label
  44 .endm
  45
  46 /* This clobbers the condition codes, which the real Thumb cbnz instruction
  47    does not do.  But it doesn't matter for any of the uses here.  */
  48 .macro cbnz reg, label
  49         cmp \reg, #0
  50         bne \label
  51 .endm
  52 #endif
  53
  54 /* Parameters and result.  */
  55 #define srcin           r0
  56 #define result          r0
  57
  58 /* Internal variables.  */
  59 #define src             r1
  60 #define data1a          r2
  61 #define data1b          r3
  62 #define const_m1        r12
  63 #define const_0         r4
  64 #define tmp1            r4              /* Overlaps const_0  */
  65 #define tmp2            r5
  66
  67         .text
  68         .p2align 6
  69 ENTRY(strlen)
  70         sfi_pld srcin, #0
  71         strd    r4, r5, [sp, #-8]!
  72         cfi_adjust_cfa_offset (8)
  73         cfi_rel_offset (r4, 0)
  74         cfi_rel_offset (r5, 4)
  75         cfi_remember_state
  76         bic     src, srcin, #7
  77         mvn     const_m1, #0
  78         ands    tmp1, srcin, #7         /* (8 - bytes) to alignment.  */
  79         sfi_pld src, #32
  80         bne.w   .Lmisaligned8
  81         mov     const_0, #0
  82         mov     result, #-8
  83 .Lloop_aligned:
  84         /* Bytes 0-7.  */
  85         sfi_breg src, \
  86         ldrd    data1a, data1b, [\B]
  87         sfi_pld src, #64
  88         add     result, result, #8
  89 .Lstart_realigned:
  90         uadd8   data1a, data1a, const_m1        /* Saturating GE<0:3> set.  */
  91         sel     data1a, const_0, const_m1       /* Select based on GE<0:3>.  */
  92         uadd8   data1b, data1b, const_m1
  93         sel     data1b, data1a, const_m1        /* Only used if d1a == 0.  */
  94         cbnz    data1b, .Lnull_found
  95
  96         /* Bytes 8-15.  */
  97         sfi_breg src, \
  98         ldrd    data1a, data1b, [\B, #8]
  99         uadd8   data1a, data1a, const_m1        /* Saturating GE<0:3> set.  */
 100         add     result, result, #8
 101         sel     data1a, const_0, const_m1       /* Select based on GE<0:3>.  */
 102         uadd8   data1b, data1b, const_m1
 103         sel     data1b, data1a, const_m1        /* Only used if d1a == 0.  */
 104         cbnz    data1b, .Lnull_found
 105
 106         /* Bytes 16-23.  */
 107         sfi_breg src, \
 108         ldrd    data1a, data1b, [\B, #16]
 109         uadd8   data1a, data1a, const_m1        /* Saturating GE<0:3> set.  */
 110         add     result, result, #8
 111         sel     data1a, const_0, const_m1       /* Select based on GE<0:3>.  */
 112         uadd8   data1b, data1b, const_m1
 113         sel     data1b, data1a, const_m1        /* Only used if d1a == 0.  */
 114         cbnz    data1b, .Lnull_found
 115
 116         /* Bytes 24-31.  */
 117         sfi_breg src, \
 118         ldrd    data1a, data1b, [\B, #24]
 119         add     src, src, #32
 120         uadd8   data1a, data1a, const_m1        /* Saturating GE<0:3> set.  */
 121         add     result, result, #8
 122         sel     data1a, const_0, const_m1       /* Select based on GE<0:3>.  */
 123         uadd8   data1b, data1b, const_m1
 124         sel     data1b, data1a, const_m1        /* Only used if d1a == 0.  */
 125         cmp     data1b, #0
 126         beq     .Lloop_aligned
 127
 128 .Lnull_found:
 129         cmp     data1a, #0
 130         itt     eq
 131         addeq   result, result, #4
 132         moveq   data1a, data1b
 133 #ifndef __ARMEB__
 134         rev     data1a, data1a
 135 #endif
 136         clz     data1a, data1a
 137         ldrd    r4, r5, [sp], #8
 138         cfi_adjust_cfa_offset (-8)
 139         cfi_restore (r4)
 140         cfi_restore (r5)
 141         add     result, result, data1a, lsr #3  /* Bits -> Bytes.  */
 142         DO_RET(lr)
 143
 144 .Lmisaligned8:
 145         cfi_restore_state
 146         sfi_breg src, \
 147         ldrd    data1a, data1b, [\B]
 148         and     tmp2, tmp1, #3
 149         rsb     result, tmp1, #0
 150         lsl     tmp2, tmp2, #3                  /* Bytes -> bits.  */
 151         tst     tmp1, #4
 152         sfi_pld src, #64
 153         S2HI    tmp2, const_m1, tmp2
 154 #ifdef NO_THUMB
 155         mvn     tmp1, tmp2
 156         orr     data1a, data1a, tmp1
 157         itt     ne
 158         orrne   data1b, data1b, tmp1
 159 #else
 160         orn     data1a, data1a, tmp2
 161         itt     ne
 162         ornne   data1b, data1b, tmp2
 163 #endif
 164         movne   data1a, const_m1
 165         mov     const_0, #0
 166         b       .Lstart_realigned
 167
 168 END(strlen)
 169 libc_hidden_builtin_def (strlen)