Enable AVX2 optimized memset only if -mavx2 works

	* config.h.in (HAVE_AVX2_SUPPORT): New #undef.
	* sysdeps/i386/configure.ac: Set HAVE_AVX2_SUPPORT and
	config-cflags-avx2.
	* sysdeps/x86_64/configure.ac: Likewise.
	* sysdeps/i386/configure: Regenerated.
	* sysdeps/x86_64/configure: Likewise.
	* sysdeps/x86_64/multiarch/Makefile (sysdep_routines): Add
	memset-avx2 only if config-cflags-avx2 is yes.
	* sysdeps/x86_64/multiarch/ifunc-impl-list.c (__libc_ifunc_impl_list):
	Tests for memset_chk and memset only if HAVE_AVX2_SUPPORT is
	defined.
	* sysdeps/x86_64/multiarch/memset.S: Define multiple versions
	only if HAVE_AVX2_SUPPORT is defined.
	* sysdeps/x86_64/multiarch/memset_chk.S: Likewise.

10 files changed, 111 insertions, 14 deletions

diff --git a/ChangeLog b/ChangeLog
index a05bc1e..725509a 100644
--- a/ChangeLog
+++ b/ChangeLog
@@ -1,3 +1,20 @@
+2014-07-14  H.J. Lu  <hongjiu.lu@intel.com>
+
+	* config.h.in (HAVE_AVX2_SUPPORT): New #undef.
+	* sysdeps/i386/configure.ac: Set HAVE_AVX2_SUPPORT and
+	config-cflags-avx2.
+	* sysdeps/x86_64/configure.ac: Likewise.
+	* sysdeps/i386/configure: Regenerated.
+	* sysdeps/x86_64/configure: Likewise.
+	* sysdeps/x86_64/multiarch/Makefile (sysdep_routines): Add
+	memset-avx2 only if config-cflags-avx2 is yes.
+	* sysdeps/x86_64/multiarch/ifunc-impl-list.c (__libc_ifunc_impl_list):
+	Tests for memset_chk and memset only if HAVE_AVX2_SUPPORT is
+	defined.
+	* sysdeps/x86_64/multiarch/memset.S: Define multiple versions
+	only if HAVE_AVX2_SUPPORT is defined.
+	* sysdeps/x86_64/multiarch/memset_chk.S: Likewise.
+
 2014-07-14  Alan Modra  <amodra@gmail.com>
 
 	[BZ #17153]
diff --git a/config.h.in b/config.h.in
index 2dcd135..97b5571 100644
--- a/config.h.in
+++ b/config.h.in
@@ -103,6 +103,9 @@
 /* Define if gcc supports FMA4.  */
 #undef	HAVE_FMA4_SUPPORT
 
+/* Define if gcc supports AVX2.  */
+#undef	HAVE_AVX2_SUPPORT
+
 /* Define if the compiler\'s exception support is based on libunwind.  */
 #undef	HAVE_CC_WITH_LIBUNWIND
 
diff --git a/sysdeps/i386/configure b/sysdeps/i386/configure
index f0a20e3..6e89b59 100644
--- a/sysdeps/i386/configure
+++ b/sysdeps/i386/configure
@@ -240,6 +240,32 @@ $as_echo "$libc_cv_cc_novzeroupper" >&6; }
 config_vars="$config_vars
 config-cflags-novzeroupper = $libc_cv_cc_novzeroupper"
 
+{ $as_echo "$as_me:${as_lineno-$LINENO}: checking for AVX2 support" >&5
+$as_echo_n "checking for AVX2 support... " >&6; }
+if ${libc_cv_cc_avx2+:} false; then :
+  $as_echo_n "(cached) " >&6
+else
+  if { ac_try='${CC-cc} -mavx2 -xc /dev/null -S -o /dev/null'
+  { { eval echo "\"\$as_me\":${as_lineno-$LINENO}: \"$ac_try\""; } >&5
+  (eval $ac_try) 2>&5
+  ac_status=$?
+  $as_echo "$as_me:${as_lineno-$LINENO}: \$? = $ac_status" >&5
+  test $ac_status = 0; }; }; then :
+  libc_cv_cc_avx2=yes
+else
+  libc_cv_cc_avx2=no
+fi
+
+fi
+{ $as_echo "$as_me:${as_lineno-$LINENO}: result: $libc_cv_cc_avx2" >&5
+$as_echo "$libc_cv_cc_avx2" >&6; }
+if test $libc_cv_cc_avx2 = yes; then
+  $as_echo "#define HAVE_AVX2_SUPPORT 1" >>confdefs.h
+
+fi
+config_vars="$config_vars
+config-cflags-avx2 = $libc_cv_cc_avx2"
+
 $as_echo "#define USE_REGPARMS 1" >>confdefs.h
 
 
diff --git a/sysdeps/i386/configure.ac b/sysdeps/i386/configure.ac
index dfe0b47..35c4522 100644
--- a/sysdeps/i386/configure.ac
+++ b/sysdeps/i386/configure.ac
@@ -88,6 +88,15 @@ LIBC_TRY_CC_OPTION([-mno-vzeroupper],
 ])
 LIBC_CONFIG_VAR([config-cflags-novzeroupper], [$libc_cv_cc_novzeroupper])
 
+dnl Check if -mavx2 works.
+AC_CACHE_CHECK(for AVX2 support, libc_cv_cc_avx2, [dnl
+LIBC_TRY_CC_OPTION([-mavx2], [libc_cv_cc_avx2=yes], [libc_cv_cc_avx2=no])
+])
+if test $libc_cv_cc_avx2 = yes; then
+  AC_DEFINE(HAVE_AVX2_SUPPORT)
+fi
+LIBC_CONFIG_VAR([config-cflags-avx2], [$libc_cv_cc_avx2])
+
 AC_DEFINE(USE_REGPARMS)
 
 dnl It is always possible to access static and hidden symbols in an
diff --git a/sysdeps/x86_64/configure b/sysdeps/x86_64/configure
index 45d868d..7d4dadd 100644
--- a/sysdeps/x86_64/configure
+++ b/sysdeps/x86_64/configure
@@ -249,6 +249,32 @@ if test $libc_cv_asm_mpx == yes; then
 
 fi
 
+{ $as_echo "$as_me:${as_lineno-$LINENO}: checking for AVX2 support" >&5
+$as_echo_n "checking for AVX2 support... " >&6; }
+if ${libc_cv_cc_avx2+:} false; then :
+  $as_echo_n "(cached) " >&6
+else
+  if { ac_try='${CC-cc} -mavx2 -xc /dev/null -S -o /dev/null'
+  { { eval echo "\"\$as_me\":${as_lineno-$LINENO}: \"$ac_try\""; } >&5
+  (eval $ac_try) 2>&5
+  ac_status=$?
+  $as_echo "$as_me:${as_lineno-$LINENO}: \$? = $ac_status" >&5
+  test $ac_status = 0; }; }; then :
+  libc_cv_cc_avx2=yes
+else
+  libc_cv_cc_avx2=no
+fi
+
+fi
+{ $as_echo "$as_me:${as_lineno-$LINENO}: result: $libc_cv_cc_avx2" >&5
+$as_echo "$libc_cv_cc_avx2" >&6; }
+if test $libc_cv_cc_avx2 = yes; then
+  $as_echo "#define HAVE_AVX2_SUPPORT 1" >>confdefs.h
+
+fi
+config_vars="$config_vars
+config-cflags-avx2 = $libc_cv_cc_avx2"
+
 $as_echo "#define PI_STATIC_AND_HIDDEN 1" >>confdefs.h
 
 # work around problem with autoconf and empty lines at the end of files
diff --git a/sysdeps/x86_64/configure.ac b/sysdeps/x86_64/configure.ac
index 9138f63..c9f9a51 100644
--- a/sysdeps/x86_64/configure.ac
+++ b/sysdeps/x86_64/configure.ac
@@ -90,6 +90,15 @@ if test $libc_cv_asm_mpx == yes; then
   AC_DEFINE(HAVE_MPX_SUPPORT)
 fi
 
+dnl Check if -mavx2 works.
+AC_CACHE_CHECK(for AVX2 support, libc_cv_cc_avx2, [dnl
+LIBC_TRY_CC_OPTION([-mavx2], [libc_cv_cc_avx2=yes], [libc_cv_cc_avx2=no])
+])
+if test $libc_cv_cc_avx2 = yes; then
+  AC_DEFINE(HAVE_AVX2_SUPPORT)
+fi
+LIBC_CONFIG_VAR([config-cflags-avx2], [$libc_cv_cc_avx2])
+
 dnl It is always possible to access static and hidden symbols in an
 dnl position independent way.
 AC_DEFINE(PI_STATIC_AND_HIDDEN)
diff --git a/sysdeps/x86_64/multiarch/Makefile b/sysdeps/x86_64/multiarch/Makefile
index 42df96f..3bb9702 100644
--- a/sysdeps/x86_64/multiarch/Makefile
+++ b/sysdeps/x86_64/multiarch/Makefile
@@ -17,8 +17,7 @@ sysdep_routines += strncat-c stpncpy-c strncpy-c strcmp-ssse3 \
 		   strcpy-sse2-unaligned strncpy-sse2-unaligned \
 		   stpcpy-sse2-unaligned stpncpy-sse2-unaligned \
 		   strcat-sse2-unaligned strncat-sse2-unaligned \
-		   strchr-sse2-no-bsf memcmp-ssse3 strstr-sse2-unaligned \
-		   memset-avx2
+		   strchr-sse2-no-bsf memcmp-ssse3 strstr-sse2-unaligned
 
 ifeq (yes,$(config-cflags-sse4))
 sysdep_routines += strcspn-c strpbrk-c strspn-c varshift
@@ -27,6 +26,10 @@ CFLAGS-strcspn-c.c += -msse4
 CFLAGS-strpbrk-c.c += -msse4
 CFLAGS-strspn-c.c += -msse4
 endif
+
+ifeq (yes,$(config-cflags-avx2))
+sysdep_routines += memset-avx2
+endif
 endif
 
 ifeq ($(subdir),wcsmbs)
diff --git a/sysdeps/x86_64/multiarch/ifunc-impl-list.c b/sysdeps/x86_64/multiarch/ifunc-impl-list.c
index f1593c5..7e93e59 100644
--- a/sysdeps/x86_64/multiarch/ifunc-impl-list.c
+++ b/sysdeps/x86_64/multiarch/ifunc-impl-list.c
@@ -61,6 +61,7 @@ __libc_ifunc_impl_list (const char *name, struct libc_ifunc_impl *array,
 			      __memmove_ssse3)
 	      IFUNC_IMPL_ADD (array, i, memmove, 1, __memmove_sse2))
 
+#ifdef HAVE_AVX2_SUPPORT
   /* Support sysdeps/x86_64/multiarch/memset_chk.S.  */
   IFUNC_IMPL (i, name, __memset_chk,
 	      IFUNC_IMPL_ADD (array, i, __memset_chk, 1, __memset_chk_sse2)
@@ -71,6 +72,7 @@ __libc_ifunc_impl_list (const char *name, struct libc_ifunc_impl *array,
   IFUNC_IMPL (i, name, memset,
 	      IFUNC_IMPL_ADD (array, i, memset, 1, __memset_sse2)
 	      IFUNC_IMPL_ADD (array, i, memset, HAS_AVX2, __memset_avx2))
+#endif
 
   /* Support sysdeps/x86_64/multiarch/stpncpy.S.  */
   IFUNC_IMPL (i, name, stpncpy,
diff --git a/sysdeps/x86_64/multiarch/memset.S b/sysdeps/x86_64/multiarch/memset.S
index 3113d1c..00d46d1 100644
--- a/sysdeps/x86_64/multiarch/memset.S
+++ b/sysdeps/x86_64/multiarch/memset.S
@@ -17,12 +17,13 @@
    License along with the GNU C Library; if not, see
    <http://www.gnu.org/licenses/>.  */
 
+#ifdef HAVE_AVX2_SUPPORT
 #include <sysdep.h>
 #include <shlib-compat.h>
 #include <init-arch.h>
 
 /* Define multiple versions only for the definition in lib.  */
-#ifndef NOT_IN_libc
+# ifndef NOT_IN_libc
 ENTRY(memset)
 	.type	memset, @gnu_indirect_function
 	cmpl	$0, __cpu_features+KIND_OFFSET(%rip)
@@ -34,26 +35,27 @@ ENTRY(memset)
 	leaq	__memset_avx2(%rip), %rax
 2:	ret
 END(memset)
-#endif
+# endif
 
-#if !defined NOT_IN_libc
-# undef memset
-# define memset __memset_sse2
+# if !defined NOT_IN_libc
+#  undef memset
+#  define memset __memset_sse2
 
-# undef __memset_chk
-# define __memset_chk __memset_chk_sse2
+#  undef __memset_chk
+#  define __memset_chk __memset_chk_sse2
 
-# ifdef SHARED
+#  ifdef SHARED
 #  undef libc_hidden_builtin_def
 /* It doesn't make sense to send libc-internal memset calls through a PLT.
    The speedup we get from using GPR instruction is likely eaten away
    by the indirect call in the PLT.  */
 #  define libc_hidden_builtin_def(name) \
 	.globl __GI_memset; __GI_memset = __memset_sse2
-# endif
+#  endif
 
-# undef strong_alias
-# define strong_alias(original, alias)
+#  undef strong_alias
+#  define strong_alias(original, alias)
+# endif
 #endif
 
 #include "../memset.S"
diff --git a/sysdeps/x86_64/multiarch/memset_chk.S b/sysdeps/x86_64/multiarch/memset_chk.S
index 2182780..8a607bd 100644
--- a/sysdeps/x86_64/multiarch/memset_chk.S
+++ b/sysdeps/x86_64/multiarch/memset_chk.S
@@ -22,7 +22,7 @@
 
 /* Define multiple versions only for the definition in lib.  */
 #ifndef NOT_IN_libc
-# ifdef SHARED
+# if defined SHARED && defined HAVE_AVX2_SUPPORT
 ENTRY(__memset_chk)
 	.type	__memset_chk, @gnu_indirect_function
 	cmpl	$0, __cpu_features+KIND_OFFSET(%rip)