sysdeps/powerpc/powerpc32/power7/rawmemchr.S

   1 /* Optimized rawmemchr implementation for PowerPC32/POWER7 using cmpb insn.
   2    Copyright (C) 2010-2019 Free Software Foundation, Inc.
   3    Contributed by Luis Machado <luisgpm@br.ibm.com>.
   4    This file is part of the GNU C Library.
   5
   6    The GNU C Library is free software; you can redistribute it and/or
   7    modify it under the terms of the GNU Lesser General Public
   8    License as published by the Free Software Foundation; either
   9    version 2.1 of the License, or (at your option) any later version.
  10
  11    The GNU C Library is distributed in the hope that it will be useful,
  12    but WITHOUT ANY WARRANTY; without even the implied warranty of
  13    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
  14    Lesser General Public License for more details.
  15
  16    You should have received a copy of the GNU Lesser General Public
  17    License along with the GNU C Library; if not, see
  18    <https://www.gnu.org/licenses/>.  */
  19
  20 #include <sysdep.h>
  21
  22 /* int [r3] rawmemchr (void *s [r3], int c [r4])  */
  23         .machine  power7
  24 ENTRY (__rawmemchr)
  25         CALL_MCOUNT
  26         dcbt    0,r3
  27         clrrwi  r8,r3,2       /* Align the address to word boundary.  */
  28
  29         /* Replicate byte to word.  */
  30         insrwi  r4,r4,8,16
  31         insrwi  r4,r4,16,0
  32
  33         /* Now r4 has a word of c bytes.  */
  34
  35         rlwinm  r6,r3,3,27,28 /* Calculate padding.  */
  36         lwz     r12,0(r8)     /* Load word from memory.  */
  37         cmpb    r5,r12,r4     /* Compare each byte against c byte.  */
  38 #ifdef __LITTLE_ENDIAN__
  39         srw     r5,r5,r6
  40         slw     r5,r5,r6
  41 #else
  42         slw     r5,r5,r6      /* Move left to discard ignored bits.  */
  43         srw     r5,r5,r6      /* Bring the bits back as zeros.  */
  44 #endif
  45         cmpwi   cr7,r5,0      /* If r5 == 0, no c bytes have been found.  */
  46         bne     cr7,L(done)
  47
  48         mtcrf   0x01,r8
  49
  50         /* Are we now aligned to a doubleword boundary?  If so, skip to
  51            the main loop.  Otherwise, go through the alignment code.  */
  52
  53         bt      29,L(loop)
  54
  55         /* Handle WORD2 of pair.  */
  56         lwzu    r12,4(r8)
  57         cmpb    r5,r12,r4
  58         cmpwi   cr7,r5,0
  59         bne     cr7,L(done)
  60         b       L(loop)       /* We branch here (rather than falling through)
  61                                  to skip the nops due to heavy alignment
  62                                  of the loop below.  */
  63
  64         /* Main loop to look for the end of the string.  Since it's a
  65            small loop (< 8 instructions), align it to 32-bytes.  */
  66         .p2align  5
  67 L(loop):
  68         /* Load two words, compare and merge in a
  69            single register for speed.  This is an attempt
  70            to speed up the byte-checking process for bigger strings.  */
  71         lwz     r12,4(r8)
  72         lwzu    r11,8(r8)
  73         cmpb    r5,r12,r4
  74         cmpb    r6,r11,r4
  75         or      r7,r5,r6
  76         cmpwi   cr7,r7,0
  77         beq     cr7,L(loop)
  78
  79         /* OK, one (or both) of the words contains a 'c' byte.  Check
  80            the first word and decrement the address in case the first
  81            word really contains a c byte.  */
  82
  83         cmpwi   cr6,r5,0
  84         addi    r8,r8,-4
  85         bne     cr6,L(done)
  86
  87         /* The 'c' byte must be in the second word.  Adjust the address
  88            again and move the result of cmpb to r10 so we can calculate the
  89            pointer.  */
  90         mr      r5,r6
  91         addi    r8,r8,4
  92
  93         /* r5 has the output of the cmpb instruction, that is, it contains
  94            0xff in the same position as the 'c' byte in the original
  95            word from the string.  Use that fact to find out what is
  96            the position of the byte inside the string.  */
  97 L(done):
  98 #ifdef __LITTLE_ENDIAN__
  99         addi    r0,r5,-1
 100         andc    r0,r0,r5
 101         popcntw r0,r0
 102 #else
 103         cntlzw  r0,r5         /* Count leading zeros before the match.  */
 104 #endif
 105         srwi    r0,r0,3       /* Convert leading zeros to bytes.  */
 106         add     r3,r8,r0      /* Return address of the matching char.  */
 107         blr
 108 END (__rawmemchr)
 109 weak_alias (__rawmemchr,rawmemchr)
 110 libc_hidden_builtin_def (__rawmemchr)