sysdeps/i386/i686/multiarch/strcat.S

   1 /* Multiple versions of strcat
   2    All versions must be listed in ifunc-impl-list.c.
   3    Copyright (C) 2011-2015 Free Software Foundation, Inc.
   4    Contributed by Intel Corporation.
   5    This file is part of the GNU C Library.
   6
   7    The GNU C Library is free software; you can redistribute it and/or
   8    modify it under the terms of the GNU Lesser General Public
   9    License as published by the Free Software Foundation; either
  10    version 2.1 of the License, or (at your option) any later version.
  11
  12    The GNU C Library is distributed in the hope that it will be useful,
  13    but WITHOUT ANY WARRANTY; without even the implied warranty of
  14    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
  15    Lesser General Public License for more details.
  16
  17    You should have received a copy of the GNU Lesser General Public
  18    License along with the GNU C Library; if not, see
  19    <http://www.gnu.org/licenses/>.  */
  20
  21 #include <sysdep.h>
  22 #include <init-arch.h>
  23
  24 #ifndef USE_AS_STRNCAT
  25 # ifndef STRCAT
  26 #  define STRCAT strcat
  27 # endif
  28 #endif
  29
  30 #ifdef USE_AS_STRNCAT
  31 # define STRCAT_SSSE3   __strncat_ssse3
  32 # define STRCAT_SSE2            __strncat_sse2
  33 # define STRCAT_IA32            __strncat_ia32
  34 # define __GI_STRCAT            __GI_strncat
  35 #else
  36 # define STRCAT_SSSE3   __strcat_ssse3
  37 # define STRCAT_SSE2            __strcat_sse2
  38 # define STRCAT_IA32            __strcat_ia32
  39 # define __GI_STRCAT            __GI_strcat
  40 #endif
  41
  42
  43 /* Define multiple versions only for the definition in libc.  Don't
  44    define multiple versions for strncat in static library since we
  45    need strncat before the initialization happened.  */
  46 #if IS_IN (libc)
  47
  48 # ifdef SHARED
  49         .text
  50 ENTRY(STRCAT)
  51         .type   STRCAT, @gnu_indirect_function
  52         pushl   %ebx
  53         cfi_adjust_cfa_offset (4)
  54         cfi_rel_offset (ebx, 0)
  55         LOAD_PIC_REG(bx)
  56         cmpl    $0, KIND_OFFSET+__cpu_features@GOTOFF(%ebx)
  57         jne     1f
  58         call    __init_cpu_features
  59 1:      leal    STRCAT_IA32@GOTOFF(%ebx), %eax
  60         testl   $bit_SSE2, CPUID_OFFSET+index_SSE2+__cpu_features@GOTOFF(%ebx)
  61         jz      2f
  62         leal    STRCAT_SSE2@GOTOFF(%ebx), %eax
  63         testl   $bit_Fast_Unaligned_Load, FEATURE_OFFSET+index_Fast_Unaligned_Load+__cpu_features@GOTOFF(%ebx)
  64         jnz     2f
  65         testl   $bit_SSSE3, CPUID_OFFSET+index_SSSE3+__cpu_features@GOTOFF(%ebx)
  66         jz      2f
  67         leal    STRCAT_SSSE3@GOTOFF(%ebx), %eax
  68 2:      popl    %ebx
  69         cfi_adjust_cfa_offset (-4)
  70         cfi_restore (ebx)
  71         ret
  72 END(STRCAT)
  73 # else
  74
  75 ENTRY(STRCAT)
  76         .type   STRCAT, @gnu_indirect_function
  77         cmpl    $0, KIND_OFFSET+__cpu_features
  78         jne     1f
  79         call    __init_cpu_features
  80 1:      leal    STRCAT_IA32, %eax
  81         testl   $bit_SSE2, CPUID_OFFSET+index_SSE2+__cpu_features
  82         jz      2f
  83         leal    STRCAT_SSE2, %eax
  84         testl   $bit_Fast_Unaligned_Load, FEATURE_OFFSET+index_Fast_Unaligned_Load+__cpu_features
  85         jnz     2f
  86         testl   $bit_SSSE3, CPUID_OFFSET+index_SSSE3+__cpu_features
  87         jz      2f
  88         leal    STRCAT_SSSE3, %eax
  89 2:      ret
  90 END(STRCAT)
  91
  92 # endif
  93
  94 # undef ENTRY
  95 # define ENTRY(name) \
  96         .type STRCAT_IA32, @function; \
  97         .align 16; \
  98         .globl STRCAT_IA32; \
  99         .hidden STRCAT_IA32; \
 100         STRCAT_IA32: cfi_startproc; \
 101         CALL_MCOUNT
 102 # undef END
 103 # define END(name) \
 104         cfi_endproc; .size STRCAT_IA32, .-STRCAT_IA32
 105
 106 # ifdef SHARED
 107 #  undef libc_hidden_builtin_def
 108 /* It doesn't make sense to send libc-internal strcat calls through a PLT.
 109    The speedup we get from using SSSE3 instruction is likely eaten away
 110    by the indirect call in the PLT.  */
 111 #  define libc_hidden_builtin_def(name) \
 112         .globl __GI_STRCAT; __GI_STRCAT = STRCAT_IA32
 113 #  undef libc_hidden_def
 114 #  define libc_hidden_def(name) \
 115         .globl __GI___STRCAT; __GI___STRCAT = STRCAT_IA32
 116
 117 # endif
 118 #endif
 119
 120 #ifndef USE_AS_STRNCAT
 121 # include "../../i486/strcat.S"
 122 #endif