This is the mail archive of the
libc-alpha@sourceware.org
mailing list for the glibc project.
Re: PATCH: PR libc/12460: AVX audit test failures with gcc 4.6
- From: "H.J. Lu" <hongjiu dot lu at intel dot com>
- To: GNU C Library <libc-alpha at sourceware dot org>
- Date: Wed, 2 Feb 2011 17:21:01 -0800
- Subject: Re: PATCH: PR libc/12460: AVX audit test failures with gcc 4.6
- References: <20110202215226.GA25289@intel.com>
- Reply-to: "H.J. Lu" <hjl dot tools at gmail dot com>
On Wed, Feb 02, 2011 at 01:52:26PM -0800, H.J. Lu wrote:
> The vzeroupper optimization in gcc 4.6 will add vzeroupper before function
> return even if AVX isn't supported. Due to nature of special codes in AVX
> audit tests, we can't use __builtin_ia32_vzeroupper in AVX audit tests.
> This patch adds -mno-vzeroupper for those tests if needed.
>
>
> H.J.
> ---
> 2011-02-02 H.J. Lu <hongjiu.lu@intel.com>
>
> PR libc/12460
> * config.make.in (config-cflags-novzeroupper): New.
>
> * configure.in: Substitute libc_cv_cc_novzeroupper.
> * configure: Regenerated.
>
> * elf/Makefile (AVX-CFLAGS): New.
> (CFLAGS-tst-audit4.c): Replace -mavx with $(AVX-CFLAGS).
> (CFLAGS-tst-auditmod4a.c): Likewise.
> (CFLAGS-tst-auditmod4b.c): Likewise.
> (CFLAGS-tst-auditmod6b.c): Likewise.
> (CFLAGS-tst-auditmod6c.c): Likewise.
> (CFLAGS-tst-auditmod7b.c): Likewise.
>
> * sysdeps/i386/configure.in: Check -mno-vzeroupper.
> * sysdeps/i386/configure: Regenerated.
>
Here is the missing configure.in/configure change.
H.J.
---
2011-02-02 H.J. Lu <hongjiu.lu@intel.com>
PR libc/12460
* configure.in: Substitute libc_cv_cc_novzeroupper.
* configure: Regenerated.
diff --git a/configure b/configure
index 999a448..1d419cf 100755
--- a/configure
+++ b/configure
@@ -614,6 +614,7 @@ elf
ldd_rewrite_script
use_ldconfig
libc_cv_as_i686
+libc_cv_cc_novzeroupper
libc_cv_cc_avx
libc_cv_cc_sse4
libc_cv_cpp_asm_debuginfo
@@ -7813,6 +7814,7 @@ fi
+
if test $elf = yes; then
$as_echo "#define HAVE_ELF 1" >>confdefs.h
diff --git a/configure.in b/configure.in
index 668dbd1..7e46990 100644
--- a/configure.in
+++ b/configure.in
@@ -2240,6 +2240,7 @@ dnl sysdeps/CPU/configure.in checks set this via arch-specific asm tests
AC_SUBST(libc_cv_cpp_asm_debuginfo)
AC_SUBST(libc_cv_cc_sse4)
AC_SUBST(libc_cv_cc_avx)
+AC_SUBST(libc_cv_cc_novzeroupper)
AC_SUBST(libc_cv_as_i686)
AC_SUBST(use_ldconfig)