sysdeps/alpha/bzero.S

   1 /* Copyright (C) 1996-2014 Free Software Foundation, Inc.
   2    Contributed by Richard Henderson (rth@tamu.edu)
   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 /* Fill a block of memory with zeros.  Optimized for the Alpha architecture:
  20
  21    - memory accessed as aligned quadwords only
  22    - destination memory not read unless needed for good cache behaviour
  23    - basic blocks arranged to optimize branch prediction for full-quadword
  24      aligned memory blocks.
  25    - partial head and tail quadwords constructed with byte-mask instructions
  26
  27    This is generally scheduled for the EV5 (got to look out for my own
  28    interests :-), but with EV4 needs in mind.  There *should* be no more
  29    stalls for the EV4 than there are for the EV5.
  30 */
  31
  32
  33 #include <sysdep.h>
  34
  35         .set noat
  36         .set noreorder
  37
  38         .text
  39         .type   __bzero, @function
  40         .globl  __bzero
  41         .usepv  __bzero, USEPV_PROF
  42
  43         cfi_startproc
  44
  45         /* On entry to this basic block:
  46            t3 == loop counter
  47            t4 == bytes in partial final word
  48            a0 == possibly misaligned destination pointer  */
  49
  50         .align 3
  51 bzero_loop:
  52         beq     t3, $tail       #
  53         blbc    t3, 0f          # skip single store if count even
  54
  55         stq_u   zero, 0(a0)     # e0    : store one word
  56         subq    t3, 1, t3       # .. e1 :
  57         addq    a0, 8, a0       # e0    :
  58         beq     t3, $tail       # .. e1 :
  59
  60 0:      stq_u   zero, 0(a0)     # e0    : store two words
  61         subq    t3, 2, t3       # .. e1 :
  62         stq_u   zero, 8(a0)     # e0    :
  63         addq    a0, 16, a0      # .. e1 :
  64         bne     t3, 0b          # e1    :
  65
  66 $tail:  bne     t4, 1f          # is there a tail to do?
  67         ret                     # no
  68
  69 1:      ldq_u   t0, 0(a0)       # yes, load original data
  70         mskqh   t0, t4, t0      #
  71         stq_u   t0, 0(a0)       #
  72         ret                     #
  73
  74 __bzero:
  75 #ifdef PROF
  76         ldgp    gp, 0(pv)
  77         lda     AT, _mcount
  78         jsr     AT, (AT), _mcount
  79 #endif
  80
  81         mov     a0, v0          # e0    : move return value in place
  82         beq     a1, $done       # .. e1 : early exit for zero-length store
  83         and     a0, 7, t1       # e0    :
  84         addq    a1, t1, a1      # e1    : add dest misalignment to count
  85         srl     a1, 3, t3       # e0    : loop = count >> 3
  86         and     a1, 7, t4       # .. e1 : find number of bytes in tail
  87         unop                    #       :
  88         beq     t1, bzero_loop  # e1    : aligned head, jump right in
  89
  90         ldq_u   t0, 0(a0)       # e0    : load original data to mask into
  91         cmpult  a1, 8, t2       # .. e1 : is this a sub-word set?
  92         bne     t2, $oneq       # e1    :
  93
  94         mskql   t0, a0, t0      # e0    : we span words.  finish this partial
  95         subq    t3, 1, t3       # e0    :
  96         addq    a0, 8, a0       # .. e1 :
  97         stq_u   t0, -8(a0)      # e0    :
  98         br      bzero_loop      # .. e1 :
  99
 100         .align 3
 101 $oneq:
 102         mskql   t0, a0, t2      # e0    :
 103         mskqh   t0, a1, t3      # e0    :
 104         or      t2, t3, t0      # e1    :
 105         stq_u   t0, 0(a0)       # e0    :
 106
 107 $done:  ret
 108
 109         cfi_endproc
 110 weak_alias (__bzero, bzero)