Compile AVX libm functions with -mavx -mprefer-avx128 hjl/avx128

9 files changed, 94 insertions, 11 deletions

diff --git a/ChangeLog.avx128 b/ChangeLog.avx128
new file mode 100644
index 0000000000..c0f795eb24
--- /dev/null
+++ b/ChangeLog.avx128
@@ -0,0 +1,20 @@
+2012-10-11  H.J. Lu  <hongjiu.lu@intel.com>
+
+	* config.make.in (config-cflags-avx128): New.
+	* configure.in (libc_cv_cc_avx128): AC_SUBST.
+	* sysdeps/i386/configure.in: Add -mavx -mprefer-avx128 check.
+	* sysdeps/x86_64/configure.in: Likewise.
+	* configure: Regenerated.
+	* sysdeps/i386/configure: Likewise.
+	* sysdeps/x86_64/configure: Likewise.
+	* sysdeps/x86_64/fpu/multiarch/Makefile (CFLAGS-avx): New macro.
+	(CFLAGS-e_atan2-avx.c): Set to $(CFLAGS-avx).
+	(CFLAGS-e_exp-avx.c): Likewise.
+	(CFLAGS-e_log-avx.c): Likewise.
+	(CFLAGS-mpa-avx.c): Likewise.
+	(CFLAGS-mpexp-avx.c): Likewise.
+	(CFLAGS-mplog-avx.c): Likewise.
+	(CFLAGS-s_atan-avx.c): Likewise.
+	(CFLAGS-s_sin-avx.c): Likewise.
+	(CFLAGS-slowexp-avx.c): Likewise.
+	(CFLAGS-s_tan-avx.c): Likewise.
diff --git a/config.make.in b/config.make.in
index b6d1c945dc..0474d0ce84 100644
--- a/config.make.in
+++ b/config.make.in
@@ -38,6 +38,7 @@ asflags-cpu = @libc_cv_cc_submachine@
 config-extra-cflags = @libc_extra_cflags@
 config-cflags-sse4 = @libc_cv_cc_sse4@
 config-cflags-avx = @libc_cv_cc_avx@
+config-cflags-avx128 = @libc_cv_cc_avx128@
 config-cflags-sse2avx = @libc_cv_cc_sse2avx@
 config-cflags-novzeroupper = @libc_cv_cc_novzeroupper@
 config-cflags-nofma = @libc_cv_cc_nofma@
diff --git a/configure b/configure
index 74ff690cf9..19f8f8299f 100755
--- a/configure
+++ b/configure
@@ -587,6 +587,7 @@ libc_cv_cc_nofma
 libc_cv_cc_fma4
 libc_cv_cc_novzeroupper
 libc_cv_cc_sse2avx
+libc_cv_cc_avx128
 libc_cv_cc_avx
 libc_cv_cc_sse4
 libc_cv_forced_unwind
@@ -7502,6 +7503,7 @@ fi
 
 
 
+
 { $as_echo "$as_me:${as_lineno-$LINENO}: checking whether -fPIC is default" >&5
 $as_echo_n "checking whether -fPIC is default... " >&6; }
 if ${libc_cv_pic_default+:} false; then :
diff --git a/configure.in b/configure.in
index facde5e55a..7d057be9bd 100644
--- a/configure.in
+++ b/configure.in
@@ -2148,6 +2148,7 @@ AC_SUBST(libc_cv_forced_unwind)
 dnl sysdeps/CPU/configure.in checks set this via arch-specific asm tests
 AC_SUBST(libc_cv_cc_sse4)
 AC_SUBST(libc_cv_cc_avx)
+AC_SUBST(libc_cv_cc_avx128)
 AC_SUBST(libc_cv_cc_sse2avx)
 AC_SUBST(libc_cv_cc_novzeroupper)
 AC_SUBST(libc_cv_cc_fma4)
diff --git a/sysdeps/i386/configure b/sysdeps/i386/configure
index 7b3683f600..4ea7be9193 100644
--- a/sysdeps/i386/configure
+++ b/sysdeps/i386/configure
@@ -111,6 +111,26 @@ if test $libc_cv_cc_avx = yes; then
 
 fi
 
+{ $as_echo "$as_me:${as_lineno-$LINENO}: checking -mavx -mprefer-avx128 support" >&5
+$as_echo_n "checking -mavx -mprefer-avx128 support... " >&6; }
+if ${libc_cv_cc_avx128+:} false; then :
+  $as_echo_n "(cached) " >&6
+else
+  if { ac_try='${CC-cc} -mavx -mprefer-avx128 -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_avx128=yes
+else
+  libc_cv_cc_avx128=no
+fi
+
+fi
+{ $as_echo "$as_me:${as_lineno-$LINENO}: result: $libc_cv_cc_avx128" >&5
+$as_echo "$libc_cv_cc_avx128" >&6; }
+
 { $as_echo "$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 :
diff --git a/sysdeps/i386/configure.in b/sysdeps/i386/configure.in
index 36cb3e46e0..e00d324bdd 100644
--- a/sysdeps/i386/configure.in
+++ b/sysdeps/i386/configure.in
@@ -28,6 +28,13 @@ if test $libc_cv_cc_avx = yes; then
   AC_DEFINE(HAVE_AVX_SUPPORT)
 fi
 
+dnl Check if -mavx -mprefer-avx128 works.
+AC_CACHE_CHECK(-mavx -mprefer-avx128 support, libc_cv_cc_avx128, [dnl
+LIBC_TRY_CC_OPTION([-mavx -mprefer-avx128],
+		   [libc_cv_cc_avx128=yes],
+		   [libc_cv_cc_avx128=no])
+])
+
 dnl Check if -msse2avx works.
 AC_CACHE_CHECK(for AVX encoding of SSE instructions, libc_cv_cc_sse2avx, [dnl
 LIBC_TRY_CC_OPTION([-msse2avx],
diff --git a/sysdeps/x86_64/configure b/sysdeps/x86_64/configure
index 0fa0e1ecfa..4b48c65bcc 100644
--- a/sysdeps/x86_64/configure
+++ b/sysdeps/x86_64/configure
@@ -91,6 +91,26 @@ if test $libc_cv_cc_avx = yes; then
 
 fi
 
+{ $as_echo "$as_me:${as_lineno-$LINENO}: checking -mavx -mprefer-avx128 support" >&5
+$as_echo_n "checking -mavx -mprefer-avx128 support... " >&6; }
+if ${libc_cv_cc_avx128+:} false; then :
+  $as_echo_n "(cached) " >&6
+else
+  if { ac_try='${CC-cc} -mavx -mprefer-avx128 -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_avx128=yes
+else
+  libc_cv_cc_avx128=no
+fi
+
+fi
+{ $as_echo "$as_me:${as_lineno-$LINENO}: result: $libc_cv_cc_avx128" >&5
+$as_echo "$libc_cv_cc_avx128" >&6; }
+
 { $as_echo "$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 :
diff --git a/sysdeps/x86_64/configure.in b/sysdeps/x86_64/configure.in
index 94bcd19f3c..3a38740bf0 100644
--- a/sysdeps/x86_64/configure.in
+++ b/sysdeps/x86_64/configure.in
@@ -21,6 +21,13 @@ if test $libc_cv_cc_avx = yes; then
   AC_DEFINE(HAVE_AVX_SUPPORT)
 fi
 
+dnl Check if -mavx -mprefer-avx128 works.
+AC_CACHE_CHECK(-mavx -mprefer-avx128 support, libc_cv_cc_avx128, [dnl
+LIBC_TRY_CC_OPTION([-mavx -mprefer-avx128],
+		   [libc_cv_cc_avx128=yes],
+		   [libc_cv_cc_avx128=no])
+])
+
 dnl Check if -msse2avx works.
 AC_CACHE_CHECK(for AVX encoding of SSE instructions, libc_cv_cc_sse2avx, [dnl
 LIBC_TRY_CC_OPTION([-msse2avx],
diff --git a/sysdeps/x86_64/fpu/multiarch/Makefile b/sysdeps/x86_64/fpu/multiarch/Makefile
index 12b0526e50..03bf06f20a 100644
--- a/sysdeps/x86_64/fpu/multiarch/Makefile
+++ b/sysdeps/x86_64/fpu/multiarch/Makefile
@@ -33,21 +33,26 @@ CFLAGS-s_sin-fma4.c = -mfma4
 CFLAGS-s_tan-fma4.c = -mfma4
 endif
 
-ifeq ($(config-cflags-sse2avx),yes)
+ifneq ($(config-cflags-sse2avx)$(config-cflags-avx128),nono)
 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 = -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
+ifeq ($(config-cflags-avx128),yes)
+CFLAGS-avx = -mavx -mprefer-avx128
+else
+CFLAGS-avx = -msse2avx -DSSE2AVX
+endif
+CFLAGS-e_atan2-avx.c = $(CFLAGS-avx)
+CFLAGS-e_exp-avx.c = $(CFLAGS-avx)
+CFLAGS-e_log-avx.c = $(CFLAGS-avx)
+CFLAGS-mpa-avx.c = $(CFLAGS-avx)
+CFLAGS-mpexp-avx.c = $(CFLAGS-avx)
+CFLAGS-mplog-avx.c = $(CFLAGS-avx)
+CFLAGS-s_atan-avx.c = $(CFLAGS-avx)
+CFLAGS-s_sin-avx.c = $(CFLAGS-avx)
+CFLAGS-slowexp-avx.c = $(CFLAGS-avx)
+CFLAGS-s_tan-avx.c = $(CFLAGS-avx)
 endif
 endif