1/* Multiple versions of memset
2 All versions must be listed in ifunc-impl-list.c.
3 Copyright (C) 2014-2017 Free Software Foundation, Inc.
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 <http://www.gnu.org/licenses/>. */
19
20#include <sysdep.h>
21#include <shlib-compat.h>
22#include <init-arch.h>
23
24/* Define multiple versions only for the definition in lib. */
25#if IS_IN (libc)
26ENTRY(memset)
27 .type memset, @gnu_indirect_function
28 LOAD_RTLD_GLOBAL_RO_RDX
29 lea __memset_erms(%rip), %RAX_LP
30 HAS_ARCH_FEATURE (Prefer_ERMS)
31 jnz 2f
32 lea __memset_sse2_unaligned_erms(%rip), %RAX_LP
33 HAS_CPU_FEATURE (ERMS)
34 jnz 1f
35 lea __memset_sse2_unaligned(%rip), %RAX_LP
361:
37 HAS_ARCH_FEATURE (AVX2_Usable)
38 jz 2f
39 lea __memset_avx2_unaligned_erms(%rip), %RAX_LP
40 HAS_CPU_FEATURE (ERMS)
41 jnz L(AVX512F)
42 lea __memset_avx2_unaligned(%rip), %RAX_LP
43L(AVX512F):
44 HAS_ARCH_FEATURE (AVX512F_Usable)
45 jz 2f
46 lea __memset_avx512_no_vzeroupper(%rip), %RAX_LP
47 HAS_ARCH_FEATURE (Prefer_No_VZEROUPPER)
48 jnz 2f
49 lea __memset_avx512_unaligned_erms(%rip), %RAX_LP
50 HAS_CPU_FEATURE (ERMS)
51 jnz 2f
52 lea __memset_avx512_unaligned(%rip), %RAX_LP
532: ret
54END(memset)
55#endif
56
57#if IS_IN (libc)
58# define MEMSET_SYMBOL(p,s) p##_sse2_##s
59
60# ifdef SHARED
61# undef libc_hidden_builtin_def
62/* It doesn't make sense to send libc-internal memset calls through a PLT.
63 The speedup we get from using SSE2 instructions is likely eaten away
64 by the indirect call in the PLT. */
65# define libc_hidden_builtin_def(name) \
66 .globl __GI_memset; __GI_memset = __memset_sse2_unaligned
67# endif
68
69# undef strong_alias
70# define strong_alias(original, alias)
71#endif
72
73#include "../memset.S"
74