From 5c616c0a75792d2836412d6c4ed5c71e8e754992 Mon Sep 17 00:00:00 2001
From: Adhemerval Zanella <azanella@linux.vnet.ibm.com>
Date: Thu, 20 Jun 2013 19:40:55 -0500
Subject: [PATCH 41/42] Fix loop construction to functions calls

Check wheter the compiler has the option -fno-tree-loop-distribute-patterns
to inhibit loop transformation to library calls and uses it on memset
and memmove default implementation to avoid recursive calls.
(backported from commit 85c2e6110c9a01ec817c30f1b7e20549d7229987)

This backport excluded the benchmark tests from the original commit.
---
 config.h.in            |  3 +++
 configure              | 33 +++++++++++++++++++++++++++++++++
 configure.in           | 18 ++++++++++++++++++
 include/libc-symbols.h | 10 ++++++++++
 string/memmove.c       |  1 +
 string/memset.c        |  1 +
 string/test-memmove.c  |  1 +
 string/test-memset.c   |  1 +
 9 files changed, 82 insertions(+)

diff --git glibc-2.17-c758a686/config.h.in glibc-2.17-c758a686/config.h.in
index f3fe6b8..b34aac2 100644
--- glibc-2.17-c758a686/config.h.in
+++ glibc-2.17-c758a686/config.h.in
@@ -69,6 +69,9 @@
 /* Define if the compiler supports __builtin_memset.  */
 #undef	HAVE_BUILTIN_MEMSET

+/* Define if compiler accepts -ftree-loop-distribute-patterns.  */
+#undef  HAVE_CC_INHIBIT_LOOP_TO_LIBCALL
+
 /* Define if the regparm attribute shall be used for local functions
    (gcc on ix86 only).  */
 #undef	USE_REGPARMS
diff --git glibc-2.17-c758a686/configure glibc-2.17-c758a686/configure
index 8799b7d..99e85be 100755
--- glibc-2.17-c758a686/configure
+++ glibc-2.17-c758a686/configure
@@ -605,6 +605,7 @@ have_selinux
 have_libcap
 have_libaudit
 LIBGD
+libc_cv_cc_loop_to_function
 libc_cv_cc_submachine
 exceptions
 gnu89_inline
@@ -7164,6 +7165,38 @@ $as_echo "$libc_cv_cc_submachine" >&6; }
 fi


+{ $as_echo "$as_me:${as_lineno-$LINENO}: checking if $CC accepts -fno-tree-loop-distribute-patterns with \
+__attribute__ ((__optimize__))" >&5
+$as_echo_n "checking if $CC accepts -fno-tree-loop-distribute-patterns with \
+__attribute__ ((__optimize__))... " >&6; }
+if ${libc_cv_cc_loop_to_function+:} false; then :
+  $as_echo_n "(cached) " >&6
+else
+  cat > conftest.c <<EOF
+void
+__attribute__ ((__optimize__ ("-fno-tree-loop-distribute-patterns")))
+foo (void) {}
+EOF
+libc_cv_cc_loop_to_function=no
+if { ac_try='${CC-cc} $CFLAGS $CPPFLAGS -c conftest.c'
+  { { eval echo "\"\$as_me\":${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_loop_to_function=yes
+fi
+rm -f conftest*
+fi
+{ $as_echo "$as_me:${as_lineno-$LINENO}: result: $libc_cv_cc_loop_to_function" >&5
+$as_echo "$libc_cv_cc_loop_to_function" >&6; }
+if test $libc_cv_cc_loop_to_function = yes; then
+  $as_echo "#define HAVE_CC_INHIBIT_LOOP_TO_LIBCALL 1" >>confdefs.h
+
+fi
+
+
 { $as_echo "$as_me:${as_lineno-$LINENO}: checking for libgd" >&5
 $as_echo_n "checking for libgd... " >&6; }
 if test "$with_gd" != "no"; then
diff --git glibc-2.17-c758a686/configure.in glibc-2.17-c758a686/configure.in
index d369382..379e77a 100644
--- glibc-2.17-c758a686/configure.in
+++ glibc-2.17-c758a686/configure.in
@@ -1928,6 +1928,24 @@ if test -n "$submachine"; then
 fi
 AC_SUBST(libc_cv_cc_submachine)

+AC_CACHE_CHECK(if $CC accepts -fno-tree-loop-distribute-patterns with \
+__attribute__ ((__optimize__)), libc_cv_cc_loop_to_function, [dnl
+cat > conftest.c <<EOF
+void
+__attribute__ ((__optimize__ ("-fno-tree-loop-distribute-patterns")))
+foo (void) {}
+EOF
+libc_cv_cc_loop_to_function=no
+if AC_TRY_COMMAND([${CC-cc} $CFLAGS $CPPFLAGS -c conftest.c])
+then
+  libc_cv_cc_loop_to_function=yes
+fi
+rm -f conftest*])
+if test $libc_cv_cc_loop_to_function = yes; then
+  AC_DEFINE(HAVE_CC_INHIBIT_LOOP_TO_LIBCALL)
+fi
+AC_SUBST(libc_cv_cc_loop_to_function)
+
 dnl Check whether we have the gd library available.
 AC_MSG_CHECKING(for libgd)
 if test "$with_gd" != "no"; then
diff --git glibc-2.17-c758a686/include/libc-symbols.h glibc-2.17-c758a686/include/libc-symbols.h
index a626d59..8f2c046 100644
--- glibc-2.17-c758a686/include/libc-symbols.h
+++ glibc-2.17-c758a686/include/libc-symbols.h
@@ -783,4 +783,14 @@ for linking")
 #define libc_ifunc_hidden_def(name) \
   libc_ifunc_hidden_def1 (__GI_##name, name)

+/* Add the compiler optimization to inhibit loop transformation to library
+   calls.  This is used to avoid recursive calls in memset and memmove
+   default implementations.  */
+#ifdef HAVE_CC_INHIBIT_LOOP_TO_LIBCALL
+# define inhibit_loop_to_libcall \
+    __attribute__ ((__optimize__ ("-fno-tree-loop-distribute-patterns")))
+#else
+# define inhibit_loop_to_libcall
+#endif
+
 #endif /* libc-symbols.h */
diff --git glibc-2.17-c758a686/string/memmove.c glibc-2.17-c758a686/string/memmove.c
index bf7dcc1..c59e7a9 100644
--- glibc-2.17-c758a686/string/memmove.c
+++ glibc-2.17-c758a686/string/memmove.c
@@ -41,6 +41,7 @@
 #endif

 rettype
+inhibit_loop_to_libcall
 MEMMOVE (a1, a2, len)
      a1const void *a1;
      a2const void *a2;
diff --git glibc-2.17-c758a686/string/memset.c glibc-2.17-c758a686/string/memset.c
index 036cb5f..eb83c1b 100644
--- glibc-2.17-c758a686/string/memset.c
+++ glibc-2.17-c758a686/string/memset.c
@@ -21,6 +21,7 @@
 #undef memset

 void *
+inhibit_loop_to_libcall
 memset (dstpp, c, len)
      void *dstpp;
      int c;
diff --git glibc-2.17-c758a686/string/test-memmove.c glibc-2.17-c758a686/string/test-memmove.c
index 13c5aff..0a969c3 100644
--- glibc-2.17-c758a686/string/test-memmove.c
+++ glibc-2.17-c758a686/string/test-memmove.c
@@ -47,6 +47,7 @@ IMPL (memmove, 1)
 #endif

 char *
+inhibit_loop_to_libcall
 simple_memmove (char *dst, const char *src, size_t n)
 {
   char *ret = dst;
diff --git glibc-2.17-c758a686/string/test-memset.c glibc-2.17-c758a686/string/test-memset.c
index af85a28..0432030 100644
--- glibc-2.17-c758a686/string/test-memset.c
+++ glibc-2.17-c758a686/string/test-memset.c
@@ -64,6 +64,7 @@ builtin_memset (char *s, int c, size_t n)
 #endif

 char *
+inhibit_loop_to_libcall
 simple_memset (char *s, int c, size_t n)
 {
   char *r = s, *end = s + n;
--
1.7.11.7