glibc/glibc-2.16-avx.patch

qcommit afc5ed09cbce5d6fd48b3a8c5ec427b31f996880
Author: Ulrich Drepper <drepper@gmail.com>
Date:   Thu Jan 26 07:45:14 2012 -0500

    Reset bit_AVX in __cpu_features is OS support is missing

2012-01-26  Ulrich Drepper  <drepper@gmail.com>

	[BZ #13583]
	* sysdeps/x86_64/multiarch/init-arch.h: Define bit_OSXSAVE.
	* sysdeps/x86_64/multiarch/init-arch.c (__init_cpu_features): If
	bit_AVX is set also check OSXAVE/XCR0 and reset bit_AVX if necessary.

Also contains selected changes from commits 08cf777f9e7f6d826658a99c7d77a359f73a45bf
and 56f6f6a2403cfa7267cad722597113be35ecf70d.

diff --git a/sysdeps/x86_64/multiarch/init-arch.c b/sysdeps/x86_64/multiarch/init-arch.c
index 65b0ee9..4fabbee 100644
--- a/sysdeps/x86_64/multiarch/init-arch.c
+++ b/sysdeps/x86_64/multiarch/init-arch.c
@@ -1,6 +1,6 @@
 /* Initialize CPU feature data.
    This file is part of the GNU C Library.
-   Copyright (C) 2008, 2009, 2010, 2011 Free Software Foundation, Inc.
+   Copyright (C) 2008, 2009, 2010, 2011, 2012 Free Software Foundation, Inc.
    Contributed by Ulrich Drepper <drepper@redhat.com>.
 
    The GNU C Library is free software; you can redistribute it and/or
@@ -144,6 +144,18 @@ __init_cpu_features (void)
   else
     kind = arch_kind_other;
 
+  if (__cpu_features.cpuid[COMMON_CPUID_INDEX_1].ecx & bit_AVX)
+    {
+      /* Reset the AVX bit in case OSXSAVE is disabled.  */
+      if ((__cpu_features.cpuid[COMMON_CPUID_INDEX_1].ecx & bit_OSXSAVE) == 0
+	  || ({ unsigned int xcrlow;
+	      unsigned int xcrhigh;
+	      asm ("xgetbv"
+		   : "=a" (xcrlow), "=d" (xcrhigh) : "c" (0));
+	      (xcrlow & 6) != 6; }))
+	__cpu_features.cpuid[COMMON_CPUID_INDEX_1].ecx &= ~bit_AVX;
+    }
+
   __cpu_features.family = family;
   __cpu_features.model = model;
   atomic_write_barrier ();
diff --git a/sysdeps/x86_64/multiarch/init-arch.h b/sysdeps/x86_64/multiarch/init-arch.h
index 2a1df39..408e5ae 100644
--- a/sysdeps/x86_64/multiarch/init-arch.h
+++ b/sysdeps/x86_64/multiarch/init-arch.h
@@ -1,5 +1,5 @@
 /* This file is part of the GNU C Library.
-   Copyright (C) 2008, 2009, 2010, 2011 Free Software Foundation, Inc.
+   Copyright (C) 2008, 2009, 2010, 2011, 2012 Free Software Foundation, Inc.
 
    The GNU C Library is free software; you can redistribute it and/or
    modify it under the terms of the GNU Lesser General Public
@@ -27,6 +27,7 @@
 #define bit_SSSE3	(1 << 9)
 #define bit_SSE4_1	(1 << 19)
 #define bit_SSE4_2	(1 << 20)
+#define bit_OSXSAVE	(1 << 27)
 #define bit_AVX		(1 << 28)
 #define bit_POPCOUNT	(1 << 23)
 #define bit_FMA		(1 << 12)

Index: glibc-2.15/config.h.in
===================================================================
--- glibc-2.15.orig/config.h.in
+++ glibc-2.15/config.h.in
@@ -102,7 +102,7 @@
    certain registers (CR0, MQ, CTR, LR) in asm statements.  */
 #undef	BROKEN_PPC_ASM_CR0
 
-/* Defined on SPARC if ld doesn't handle R_SPARC_WDISP22 against .hidden
+/* Defined on SPARC if ld does not handle R_SPARC_WDISP22 against .hidden
    symbol.  sysdeps/sparc/sparc32/elf/configure.  */
 #undef	BROKEN_SPARC_WDISP22
 
@@ -118,17 +118,20 @@
 /* Define if gcc supports AVX.  */
 #undef	HAVE_AVX_SUPPORT
 
+/* Define if gcc supports VEX encoding.  */
+#undef	HAVE_SSE2AVX_SUPPORT
+
 /* Define if gcc supports FMA4.  */
 #undef	HAVE_FMA4_SUPPORT
 
-/* Define if the compiler's exception support is based on libunwind.  */
+/* Define if the compiler\'s exception support is based on libunwind.  */
 #undef	HAVE_CC_WITH_LIBUNWIND
 
 /* Define if the access to static and hidden variables is position independent
    and does not need relocations.  */
 #undef	PI_STATIC_AND_HIDDEN
 
-/* Define this to disable the `hidden_proto' et al macros in
+/* Define this to disable the 'hidden_proto' et al macros in
    include/libc-symbols.h that avoid PLT slots in the shared objects.  */
 #undef	NO_HIDDEN
 
Index: glibc-2.15/sysdeps/i386/configure
===================================================================
--- glibc-2.15.orig/sysdeps/i386/configure
+++ glibc-2.15/sysdeps/i386/configure
@@ -756,6 +756,29 @@ if test $libc_cv_cc_avx = yes; then
 
 fi
 
+{ $as_echo "$as_me:${as_lineno-$LINENO}: checking for AVX encoding of SSE instructions" >&5
+$as_echo_n "checking for AVX encoding of SSE instructions... " >&6; }
+if ${libc_cv_cc_sse2avx+:} false; then :
+  $as_echo_n "(cached) " >&6
+else
+  if { ac_try='${CC-cc} -msse2avx -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_sse2avx=yes
+else
+  libc_cv_cc_sse2avx=no
+fi
+fi
+{ $as_echo "$as_me:${as_lineno-$LINENO}: result: $libc_cv_cc_sse2avx" >&5
+$as_echo "$libc_cv_cc_sse2avx" >&6; }
+if test $libc_cv_cc_sse2avx = yes; then
+  $as_echo "#define HAVE_SSE2AVX_SUPPORT 1" >>confdefs.h
+
+fi
+
 { $as_echo "$as_me:${as_lineno-$LINENO}: checking for FMA4 support" >&5
 $as_echo_n "checking for FMA4 support... " >&6; }
 if ${libc_cv_cc_fma4+:} false; then :
Index: glibc-2.15/sysdeps/i386/configure.in
===================================================================
--- glibc-2.15.orig/sysdeps/i386/configure.in
+++ glibc-2.15/sysdeps/i386/configure.in
@@ -67,6 +67,17 @@ if test $libc_cv_cc_avx = yes; then
   AC_DEFINE(HAVE_AVX_SUPPORT)
 fi
 
+dnl Check if -msse2avx works.
+AC_CACHE_CHECK(for AVX encoding of SSE instructions, libc_cv_cc_sse2avx, [dnl
+if AC_TRY_COMMAND([${CC-cc} -msse2avx -xc /dev/null -S -o /dev/null]); then
+  libc_cv_cc_sse2avx=yes
+else
+  libc_cv_cc_sse2avx=no
+fi])
+if test $libc_cv_cc_sse2avx = yes; then
+  AC_DEFINE(HAVE_SSE2AVX_SUPPORT)
+fi
+
 dnl Check if -mfma4 works.
 AC_CACHE_CHECK(for FMA4 support, libc_cv_cc_fma4, [dnl
 if AC_TRY_COMMAND([${CC-cc} -mfma4 -xc /dev/null -S -o /dev/null]); then
Index: glibc-2.15/sysdeps/x86_64/fpu/multiarch/Makefile
===================================================================
--- glibc-2.15.orig/sysdeps/x86_64/fpu/multiarch/Makefile
+++ glibc-2.15/sysdeps/x86_64/fpu/multiarch/Makefile
@@ -34,21 +34,21 @@ CFLAGS-s_sin-fma4.c = -mfma4
 CFLAGS-s_tan-fma4.c = -mfma4
 endif
 
-ifeq ($(config-cflags-avx),yes)
+ifeq ($(config-cflags-sse2avx),yes)
 libm-sysdep_routines += e_exp-avx e_log-avx s_atan-avx \
 			e_atan2-avx s_sin-avx s_tan-avx \
 			mplog-avx mpa-avx slowexp-avx \
 			mpexp-avx
 
-CFLAGS-e_atan2-avx.c = -mavx
-CFLAGS-e_exp-avx.c = -mavx
-CFLAGS-e_log-avx.c = -mavx
-CFLAGS-mpa-avx.c = -mavx
-CFLAGS-mpexp-avx.c = -mavx
-CFLAGS-mplog-avx.c = -mavx
-CFLAGS-s_atan-avx.c = -mavx
-CFLAGS-s_sin-avx.c = -mavx
-CFLAGS-slowexp-avx.c = -mavx
-CFLAGS-s_tan-avx.c = -mavx
+CFLAGS-e_atan2-avx.c = -msse2avx -DSSE2AVX
+CFLAGS-e_exp-avx.c = -msse2avx -DSSE2AVX
+CFLAGS-e_log-avx.c = -msse2avx -DSSE2AVX
+CFLAGS-mpa-avx.c = -msse2avx -DSSE2AVX
+CFLAGS-mpexp-avx.c = -msse2avx -DSSE2AVX
+CFLAGS-mplog-avx.c = -msse2avx -DSSE2AVX
+CFLAGS-s_atan-avx.c = -msse2avx -DSSE2AVX
+CFLAGS-s_sin-avx.c = -msse2avx -DSSE2AVX
+CFLAGS-slowexp-avx.c = -msse2avx -DSSE2AVX
+CFLAGS-s_tan-avx.c = -msse2avx -DSSE2AVX
 endif
 endif
Index: glibc-2.15/config.make.in
===================================================================
--- glibc-2.15.orig/config.make.in
+++ glibc-2.15/config.make.in
@@ -36,6 +36,7 @@ asflags-cpu = @libc_cv_cc_submachine@
 
 config-cflags-sse4 = @libc_cv_cc_sse4@
 config-cflags-avx = @libc_cv_cc_avx@
+config-cflags-sse2avx = @libc_cv_cc_sse2avx@
 config-cflags-novzeroupper = @libc_cv_cc_novzeroupper@
 config-asflags-i686 = @libc_cv_as_i686@
 
Index: glibc-2.15/configure
===================================================================
--- glibc-2.15.orig/configure
+++ glibc-2.15/configure
@@ -625,6 +625,7 @@ use_ldconfig
 libc_cv_as_i686
 libc_cv_cc_fma4
 libc_cv_cc_novzeroupper
+libc_cv_cc_sse2avx
 libc_cv_cc_avx
 libc_cv_cc_sse4
 libc_cv_cpp_asm_debuginfo
Index: glibc-2.15/configure.in
===================================================================
--- glibc-2.15.orig/configure.in
+++ glibc-2.15/configure.in
@@ -2338,6 +2338,7 @@ dnl sysdeps/CPU/configure.in checks set
 AC_SUBST(libc_cv_cpp_asm_debuginfo)
 AC_SUBST(libc_cv_cc_sse4)
 AC_SUBST(libc_cv_cc_avx)
+AC_SUBST(libc_cv_cc_sse2avx)
 AC_SUBST(libc_cv_cc_novzeroupper)
 AC_SUBST(libc_cv_cc_fma4)
 AC_SUBST(libc_cv_as_i686)


--- eglibc-2.15-fma4-orig/sysdeps/x86_64/multiarch/init-arch.c	2012-05-07 10:02:38.869334673 -0700
+++ eglibc-2.15-fma4-orig/sysdeps/x86_64/multiarch/init-arch.c	2012-05-07 10:01:16.217334959 -0700
@@ -156,6 +156,11 @@ 
 	__cpu_features.cpuid[COMMON_CPUID_INDEX_1].ecx &= ~bit_AVX;
     }
 
+  /* FMA4 depends on AVX support */
+  if (__cpu_features.cpuid[COMMON_CPUID_INDEX_80000001].ecx & bit_FMA4
+      && (__cpu_features.cpuid[COMMON_CPUID_INDEX_1].ecx & bit_AVX) == 0)
+    __cpu_features.cpuid[COMMON_CPUID_INDEX_80000001].ecx &= ~bit_FMA4;
+
   __cpu_features.family = family;
   __cpu_features.model = model;
   atomic_write_barrier ();