sysdeps/alpha/alphaev67/strchr.S

   1 /* Copyright (C) 2000-2020 Free Software Foundation, Inc.
   2    Contributed by Richard Henderson <rth@tamu.edu>, 1996.
   3    EV67 optimized by Rick Gorton <rick.gorton@alpha-processor.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 /* Return the address of a given character within a null-terminated
  21    string, or null if it is not found.  */
  22
  23 #include <sysdep.h>
  24
  25         .arch ev6
  26         .set noreorder
  27         .set noat
  28
  29 ENTRY(strchr)
  30 #ifdef PROF
  31         ldgp    gp, 0(pv)
  32         lda     AT, _mcount
  33         jsr     AT, (AT), _mcount
  34         .prologue 1
  35 #else
  36         .prologue 0
  37 #endif
  38
  39         ldq_u   t0, 0(a0)       # L : load first quadword Latency=3
  40         and     a1, 0xff, t3    # E : 00000000000000ch
  41         insbl   a1, 1, t5       # U : 000000000000ch00
  42         insbl   a1, 7, a2       # U : ch00000000000000
  43
  44         insbl   t3, 6, a3       # U : 00ch000000000000
  45         or      t5, t3, a1      # E : 000000000000chch
  46         andnot  a0, 7, v0       # E : align our loop pointer
  47         lda     t4, -1          # E : build garbage mask
  48
  49         mskqh   t4, a0, t4      # U : only want relevant part of first quad
  50         or      a2, a3, a2      # E : chch000000000000
  51         inswl   a1, 2, t5       # E : 00000000chch0000
  52         inswl   a1, 4, a3       # E : 0000chch00000000
  53
  54         or      a1, a2, a1      # E : chch00000000chch
  55         or      a3, t5, t5      # E : 0000chchchch0000
  56         cmpbge  zero, t0, t2    # E : bits set iff byte == zero
  57         cmpbge  zero, t4, t4    # E : bits set iff byte is garbage
  58
  59         /* This quad is _very_ serialized.  Lots of stalling happens */
  60         or      t5, a1, a1      # E : chchchchchchchch
  61         xor     t0, a1, t1      # E : make bytes == c zero
  62         cmpbge  zero, t1, t3    # E : bits set iff byte == c
  63         or      t2, t3, t0      # E : bits set iff char match or zero match
  64
  65         andnot  t0, t4, t0      # E : clear garbage bits
  66         cttz    t0, a2          # U0 : speculative (in case we get a match)
  67         nop                     # E :
  68         bne     t0, $found      # U :
  69
  70         /*
  71          * Yuk.  This loop is going to stall like crazy waiting for the
  72          * data to be loaded.  Not much can be done about it unless it's
  73          * unrolled multiple times, which is generally unsafe.
  74          */
  75 $loop:
  76         ldq     t0, 8(v0)       # L : Latency=3
  77         addq    v0, 8, v0       # E :
  78         xor     t0, a1, t1      # E :
  79         cmpbge  zero, t0, t2    # E : bits set iff byte == 0
  80
  81         cmpbge  zero, t1, t3    # E : bits set iff byte == c
  82         or      t2, t3, t0      # E :
  83         cttz    t3, a2          # U0 : speculative (in case we get a match)
  84         beq     t0, $loop       # U :
  85
  86 $found:
  87         negq    t0, t1          # E : clear all but least set bit
  88         and     t0, t1, t0      # E :
  89         and     t0, t3, t1      # E : bit set iff byte was the char
  90         addq    v0, a2, v0      # E : Add in the bit number from above
  91
  92         cmoveq  t1, $31, v0     # E : Two mapping slots, latency = 2
  93         nop
  94         nop
  95         ret                     # L0 :
  96
  97         END(strchr)
  98
  99 weak_alias (strchr, index)
 100 libc_hidden_builtin_def (strchr)