[PATCH 5/5] Extend generic k_rem_pio2 implementation using generics.

[PATCH 5/5] Extend generic k_rem_pio2 implementation using generics.

[Paul E. Murphy]

This is actually just an internal stub.  It is already
implemented where needed.  This ensures there is always
an empty stub for each type/format when it is not needed.

	* math/Makefile: (gen-libm-calls): Add k_rem_pio2.
	(libm-calls): Remove above.

	* math/k_rem_pio2l.c: Refactor into ...
	* math/k_rem_pio2_template.c: New file.
---
 math/Makefile              |  5 +++--
 math/k_rem_pio2_template.c |  2 ++
 math/k_rem_pio2l.c         | 15 ---------------
 3 files changed, 5 insertions(+), 17 deletions(-)
 create mode 100644 math/k_rem_pio2_template.c
 delete mode 100644 math/k_rem_pio2l.c

diff --git a/math/Makefile b/math/Makefile
index c0d7ff0..0eb904f 100644
--- a/math/Makefile
+++ b/math/Makefile
@@ -49,14 +49,15 @@ gen-libm-calls = cargF conjF cimagF crealF cabsF s_cacosF		  \
 	         s_cacoshF s_ccosF s_ccoshF s_casinF s_csinF s_casinhF	  \
 		 k_casinhF s_csinhF k_casinhF s_csinhF s_catanhF s_catanF \
 		 s_ctanF s_ctanhF s_cexpF s_clogF s_cprojF s_csqrtF	  \
-		 s_cpowF s_clog10F s_fdimF s_nextdownF s_fmaxF s_fminF
+		 s_cpowF s_clog10F s_fdimF s_nextdownF s_fmaxF s_fminF    \
+		 k_rem_pio2F
 
 libm-calls =								  \
 	e_acosF e_acoshF e_asinF e_atan2F e_atanhF e_coshF e_expF e_fmodF \
 	e_hypotF e_j0F e_j1F e_jnF e_lgammaF_r e_logF e_log10F e_powF	  \
 	e_rem_pio2F e_remainderF e_scalbF e_sinhF e_sqrtF e_gammaF_r	  \
 	e_ilogbF							  \
-	k_cosF k_rem_pio2F k_sinF k_tanF s_asinhF s_atanF s_cbrtF	  \
+	k_cosF k_sinF k_tanF s_asinhF s_atanF s_cbrtF			  \
 	s_ceilF s_cosF s_erfF s_expm1F s_fabsF				  \
 	s_floorF s_log1pF w_log1pF s_logbF				  \
 	s_nextafterF s_nexttowardF s_rintF s_scalblnF w_scalblnF	  \
diff --git a/math/k_rem_pio2_template.c b/math/k_rem_pio2_template.c
new file mode 100644
index 0000000..4bb79f6
--- /dev/null
+++ b/math/k_rem_pio2_template.c
@@ -0,0 +1,2 @@
+/* This function is not necessary for all formats/types.  If necessary, override this
+   in the format/type specific directory.  */
diff --git a/math/k_rem_pio2l.c b/math/k_rem_pio2l.c
deleted file mode 100644
index 01bf158..0000000
--- a/math/k_rem_pio2l.c
+++ /dev/null
@@ -1,15 +0,0 @@
-#include <math.h>
-#include <math_private.h>
-#include <stdio.h>
-#include <errno.h>
-
-int
-__kernel_rem_pio2l (long double *x, long double *y, int e0, int nx, int prec,
-		    const int *ipio2)
-{
-  fputs ("__kernel_rem_pio2l not implemented\n", stderr);
-  __set_errno (ENOSYS);
-  return 0.0;
-}
-
-stub_warning (__kernel_rem_pio2l)
-- 
2.4.11

[PATCH 4/5] Make common fmin implementation generic.

[Paul E. Murphy]

	 * math/Makefile (gen-libm-calls): Add s_fmin
	 (libm-calls): Remove above.

	 * math/s_fmin.c: Refactor into ...
	 * math/s_fmin_template.c: New file.

	 * math/s_fminf.c: Removed.
	 * math/s_fminl.c: Removed.
	 * sysdeps/ieee754/ldbl-opt/math-type-macros-double.h:
	 (LDOUBLE_fmin_libm_version): New macro.

	 * sysdeps/ieee754/ldbl-opt/s_fmin.c: Removed.
	 * sysdeps/ieee754/ldbl-opt/s_fminl.c: Removed.
---
 math/Makefile                                      |  4 +--
 math/s_fmin.c                                      | 32 ----------------------
 math/s_fmin_template.c                             | 32 ++++++++++++++++++++++
 math/s_fminf.c                                     | 28 -------------------
 math/s_fminl.c                                     | 28 -------------------
 sysdeps/ieee754/ldbl-opt/math-type-macros-double.h |  1 +
 sysdeps/ieee754/ldbl-opt/s_fmin.c                  |  5 ----
 sysdeps/ieee754/ldbl-opt/s_fminl.c                 |  5 ----
 8 files changed, 35 insertions(+), 100 deletions(-)
 delete mode 100644 math/s_fmin.c
 create mode 100644 math/s_fmin_template.c
 delete mode 100644 math/s_fminf.c
 delete mode 100644 math/s_fminl.c
 delete mode 100644 sysdeps/ieee754/ldbl-opt/s_fmin.c
 delete mode 100644 sysdeps/ieee754/ldbl-opt/s_fminl.c

diff --git a/math/Makefile b/math/Makefile
index 09c4c56..c0d7ff0 100644
--- a/math/Makefile
+++ b/math/Makefile
@@ -49,7 +49,7 @@ gen-libm-calls = cargF conjF cimagF crealF cabsF s_cacosF		  \
 	         s_cacoshF s_ccosF s_ccoshF s_casinF s_csinF s_casinhF	  \
 		 k_casinhF s_csinhF k_casinhF s_csinhF s_catanhF s_catanF \
 		 s_ctanF s_ctanhF s_cexpF s_clogF s_cprojF s_csqrtF	  \
-		 s_cpowF s_clog10F s_fdimF s_nextdownF s_fmaxF
+		 s_cpowF s_clog10F s_fdimF s_nextdownF s_fmaxF s_fminF
 
 libm-calls =								  \
 	e_acosF e_acoshF e_asinF e_atan2F e_atanhF e_coshF e_expF e_fmodF \
@@ -65,7 +65,7 @@ libm-calls =								  \
 	w_tgammaF w_hypotF w_j0F w_j1F w_jnF w_lgammaF w_lgammaF_r	  \
 	w_logF w_log10F w_powF w_remainderF w_scalbF w_sinhF w_sqrtF	  \
 	w_ilogbF							  \
-	s_fpclassifyF s_fminF s_nanF s_truncF		  	  	  \
+	s_fpclassifyF s_nanF s_truncF		  	  	          \
 	s_remquoF e_log2F e_exp2F s_roundF s_nearbyintF s_sincosF	  \
 	s_fmaF s_lrintF s_llrintF s_lroundF s_llroundF e_exp10F w_log2F	  \
 	s_issignalingF $(calls:s_%=m_%) x2y2m1F				  \
diff --git a/math/s_fmin.c b/math/s_fmin.c
deleted file mode 100644
index d22b916..0000000
--- a/math/s_fmin.c
+++ /dev/null
@@ -1,32 +0,0 @@
-/* Return minimum numeric value of X and Y.
-   Copyright (C) 1997-2016 Free Software Foundation, Inc.
-   This file is part of the GNU C Library.
-   Contributed by Ulrich Drepper <drepper@cygnus.com>, 1997.
-
-   The GNU C Library is free software; you can redistribute it and/or
-   modify it under the terms of the GNU Lesser General Public
-   License as published by the Free Software Foundation; either
-   version 2.1 of the License, or (at your option) any later version.
-
-   The GNU C Library is distributed in the hope that it will be useful,
-   but WITHOUT ANY WARRANTY; without even the implied warranty of
-   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-   Lesser General Public License for more details.
-
-   You should have received a copy of the GNU Lesser General Public
-   License along with the GNU C Library; if not, see
-   <http://www.gnu.org/licenses/>.  */
-
-#include <math.h>
-
-
-double
-__fmin (double x, double y)
-{
-  return (islessequal (x, y) || isnan (y)) ? x : y;
-}
-weak_alias (__fmin, fmin)
-#ifdef NO_LONG_DOUBLE
-strong_alias (__fmin, __fminl)
-weak_alias (__fmin, fminl)
-#endif
diff --git a/math/s_fmin_template.c b/math/s_fmin_template.c
new file mode 100644
index 0000000..b70989a
--- /dev/null
+++ b/math/s_fmin_template.c
@@ -0,0 +1,32 @@
+/* Return minimum numeric value of X and Y.
+   Copyright (C) 1997-2016 Free Software Foundation, Inc.
+   This file is part of the GNU C Library.
+   Contributed by Ulrich Drepper <drepper@cygnus.com>, 1997.
+
+   The GNU C Library is free software; you can redistribute it and/or
+   modify it under the terms of the GNU Lesser General Public
+   License as published by the Free Software Foundation; either
+   version 2.1 of the License, or (at your option) any later version.
+
+   The GNU C Library is distributed in the hope that it will be useful,
+   but WITHOUT ANY WARRANTY; without even the implied warranty of
+   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+   Lesser General Public License for more details.
+
+   You should have received a copy of the GNU Lesser General Public
+   License along with the GNU C Library; if not, see
+   <http://www.gnu.org/licenses/>.  */
+
+#include <math.h>
+
+
+FLOAT
+M_DECL_FUNC (__fmin) (FLOAT x, FLOAT y)
+{
+  return (islessequal (x, y) || isnan (y)) ? x : y;
+}
+declare_mgen_alias (__fmin, fmin);
+
+#if M_LIBM_NEED_COMPAT (fmin)
+declare_mgen_libm_compat (__fmin, fmin)
+#endif
diff --git a/math/s_fminf.c b/math/s_fminf.c
deleted file mode 100644
index 968373a..0000000
--- a/math/s_fminf.c
+++ /dev/null
@@ -1,28 +0,0 @@
-/* Return minimum numeric value of X and Y.
-   Copyright (C) 1997-2016 Free Software Foundation, Inc.
-   This file is part of the GNU C Library.
-   Contributed by Ulrich Drepper <drepper@cygnus.com>, 1997.
-
-   The GNU C Library is free software; you can redistribute it and/or
-   modify it under the terms of the GNU Lesser General Public
-   License as published by the Free Software Foundation; either
-   version 2.1 of the License, or (at your option) any later version.
-
-   The GNU C Library is distributed in the hope that it will be useful,
-   but WITHOUT ANY WARRANTY; without even the implied warranty of
-   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-   Lesser General Public License for more details.
-
-   You should have received a copy of the GNU Lesser General Public
-   License along with the GNU C Library; if not, see
-   <http://www.gnu.org/licenses/>.  */
-
-#include <math.h>
-
-
-float
-__fminf (float x, float y)
-{
-  return (islessequal (x, y) || isnan (y)) ? x : y;
-}
-weak_alias (__fminf, fminf)
diff --git a/math/s_fminl.c b/math/s_fminl.c
deleted file mode 100644
index 84e9e51..0000000
--- a/math/s_fminl.c
+++ /dev/null
@@ -1,28 +0,0 @@
-/* Return minimum numeric value of X and Y.
-   Copyright (C) 1997-2016 Free Software Foundation, Inc.
-   This file is part of the GNU C Library.
-   Contributed by Ulrich Drepper <drepper@cygnus.com>, 1997.
-
-   The GNU C Library is free software; you can redistribute it and/or
-   modify it under the terms of the GNU Lesser General Public
-   License as published by the Free Software Foundation; either
-   version 2.1 of the License, or (at your option) any later version.
-
-   The GNU C Library is distributed in the hope that it will be useful,
-   but WITHOUT ANY WARRANTY; without even the implied warranty of
-   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-   Lesser General Public License for more details.
-
-   You should have received a copy of the GNU Lesser General Public
-   License along with the GNU C Library; if not, see
-   <http://www.gnu.org/licenses/>.  */
-
-#include <math.h>
-
-
-long double
-__fminl (long double x, long double y)
-{
-  return (islessequal (x, y) || isnan (y)) ? x : y;
-}
-weak_alias (__fminl, fminl)
diff --git a/sysdeps/ieee754/ldbl-opt/math-type-macros-double.h b/sysdeps/ieee754/ldbl-opt/math-type-macros-double.h
index 5644df0..b5136b7 100644
--- a/sysdeps/ieee754/ldbl-opt/math-type-macros-double.h
+++ b/sysdeps/ieee754/ldbl-opt/math-type-macros-double.h
@@ -46,6 +46,7 @@
 #define LDOUBLE___clog10l_libm_version GLIBC_2_1
 #define LDOUBLE_fdiml_libm_version GLIBC_2_1
 #define LDOUBLE_fmaxl_libm_version GLIBC_2_1
+#define LDOUBLE_fminl_libm_version GLIBC_2_1
 
 /* Define compat symbols for long double on platforms
    where it was not always a distinct type.  */
diff --git a/sysdeps/ieee754/ldbl-opt/s_fmin.c b/sysdeps/ieee754/ldbl-opt/s_fmin.c
deleted file mode 100644
index c3fe44d..0000000
--- a/sysdeps/ieee754/ldbl-opt/s_fmin.c
+++ /dev/null
@@ -1,5 +0,0 @@
-#include <math_ldbl_opt.h>
-#include <math/s_fmin.c>
-#if LONG_DOUBLE_COMPAT(libm, GLIBC_2_1)
-compat_symbol (libm, __fmin, fminl, GLIBC_2_1);
-#endif
diff --git a/sysdeps/ieee754/ldbl-opt/s_fminl.c b/sysdeps/ieee754/ldbl-opt/s_fminl.c
deleted file mode 100644
index 9bfdc7a..0000000
--- a/sysdeps/ieee754/ldbl-opt/s_fminl.c
+++ /dev/null
@@ -1,5 +0,0 @@
-#include <math_ldbl_opt.h>
-#undef weak_alias
-#define weak_alias(n,a)
-#include <math/s_fminl.c>
-long_double_symbol (libm, __fminl, fminl);
-- 
2.4.11

[PATCH 2/5] Make common nextdown implementation generic.

[Paul E. Murphy]

With the exception of those machines using the ldbl-opt in
an Implies file, this is a trivial transformation.

nextdownl is not subject to the non-trivial versioning rules
of the other generated functions, so to keep things simple,
it is handled as a one-off case in ldbl-opt to preserve the
existing behavior.

	* math/Makefile: (gen-libm-calls): Add s_nextdown.
	(libm-calls): Remove above.

	* math/s_nextdown.c: Refactor into ...
	* math/s_nextdown_template.c: New file.
	* math/s_nextdownf.c: Removed.
	* math/s_nextdownl.c: Removed.

	* sysdeps/ieee754/ldbl-opt/s_nextdownl.c: New file.
---
 math/Makefile                          |  4 ++--
 math/s_nextdown.c                      | 33 ---------------------------------
 math/s_nextdown_template.c             | 29 +++++++++++++++++++++++++++++
 math/s_nextdownf.c                     | 29 -----------------------------
 math/s_nextdownl.c                     | 29 -----------------------------
 sysdeps/ieee754/ldbl-opt/s_nextdownl.c |  5 +++++
 6 files changed, 36 insertions(+), 93 deletions(-)
 delete mode 100644 math/s_nextdown.c
 create mode 100644 math/s_nextdown_template.c
 delete mode 100644 math/s_nextdownf.c
 delete mode 100644 math/s_nextdownl.c
 create mode 100644 sysdeps/ieee754/ldbl-opt/s_nextdownl.c

diff --git a/math/Makefile b/math/Makefile
index d4b6d98..be6b8ac 100644
--- a/math/Makefile
+++ b/math/Makefile
@@ -49,7 +49,7 @@ gen-libm-calls = cargF conjF cimagF crealF cabsF s_cacosF		  \
 	         s_cacoshF s_ccosF s_ccoshF s_casinF s_csinF s_casinhF	  \
 		 k_casinhF s_csinhF k_casinhF s_csinhF s_catanhF s_catanF \
 		 s_ctanF s_ctanhF s_cexpF s_clogF s_cprojF s_csqrtF	  \
-		 s_cpowF s_clog10F s_fdimF
+		 s_cpowF s_clog10F s_fdimF s_nextdownF
 
 libm-calls =								  \
 	e_acosF e_acoshF e_asinF e_atan2F e_atanhF e_coshF e_expF e_fmodF \
@@ -70,7 +70,7 @@ libm-calls =								  \
 	s_fmaF s_lrintF s_llrintF s_lroundF s_llroundF e_exp10F w_log2F	  \
 	s_issignalingF $(calls:s_%=m_%) x2y2m1F				  \
 	gamma_productF lgamma_negF lgamma_productF			  \
-	s_nextupF s_nextdownF $(gen-libm-calls)
+	s_nextupF $(gen-libm-calls)
 
 libm-compat-calls-ldouble-yes = w_lgamma_compatl k_standardl
 libm-compat-calls = w_lgamma_compatf w_lgamma_compat k_standard k_standardf \
diff --git a/math/s_nextdown.c b/math/s_nextdown.c
deleted file mode 100644
index 06fd1c9..0000000
--- a/math/s_nextdown.c
+++ /dev/null
@@ -1,33 +0,0 @@
-/* Return the greatest floating-point number less than X.
-   Copyright (C) 2016 Free Software Foundation, Inc.
-   This file is part of the GNU C Library.
-
-   The GNU C Library is free software; you can redistribute it and/or
-   modify it under the terms of the GNU Lesser General Public
-   License as published by the Free Software Foundation; either
-   version 2.1 of the License, or (at your option) any later version.
-
-   The GNU C Library is distributed in the hope that it will be useful,
-   but WITHOUT ANY WARRANTY; without even the implied warranty of
-   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-   Lesser General Public License for more details.
-
-   You should have received a copy of the GNU Lesser General Public
-   License along with the GNU C Library; if not, see
-   <http://www.gnu.org/licenses/>.  */
-
-#include <math.h>
-#include <math_private.h>
-
-/* Return the greatest floating-point number less than X.  */
-double
-__nextdown (double x)
-{
-  return -__nextup (-x);
-}
-
-weak_alias (__nextdown, nextdown)
-#ifdef NO_LONG_DOUBLE
-strong_alias (__nextdown, __nextdownl)
-weak_alias (__nextdown, nextdownl)
-#endif
diff --git a/math/s_nextdown_template.c b/math/s_nextdown_template.c
new file mode 100644
index 0000000..f286dfd
--- /dev/null
+++ b/math/s_nextdown_template.c
@@ -0,0 +1,29 @@
+/* Return the greatest floating-point number less than X.
+   Copyright (C) 2016 Free Software Foundation, Inc.
+   This file is part of the GNU C Library.
+
+   The GNU C Library is free software; you can redistribute it and/or
+   modify it under the terms of the GNU Lesser General Public
+   License as published by the Free Software Foundation; either
+   version 2.1 of the License, or (at your option) any later version.
+
+   The GNU C Library is distributed in the hope that it will be useful,
+   but WITHOUT ANY WARRANTY; without even the implied warranty of
+   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+   Lesser General Public License for more details.
+
+   You should have received a copy of the GNU Lesser General Public
+   License along with the GNU C Library; if not, see
+   <http://www.gnu.org/licenses/>.  */
+
+#include <math.h>
+#include <math_private.h>
+
+/* Return the greatest floating-point number less than X.  */
+FLOAT
+M_DECL_FUNC (__nextdown) (FLOAT x)
+{
+  return -M_SUF (__nextup) (-x);
+}
+
+declare_mgen_alias (__nextdown, nextdown);
diff --git a/math/s_nextdownf.c b/math/s_nextdownf.c
deleted file mode 100644
index c0d4585..0000000
--- a/math/s_nextdownf.c
+++ /dev/null
@@ -1,29 +0,0 @@
-/* Return the greatest floating-point number less than X.
-   Copyright (C) 2016 Free Software Foundation, Inc.
-   This file is part of the GNU C Library.
-
-   The GNU C Library is free software; you can redistribute it and/or
-   modify it under the terms of the GNU Lesser General Public
-   License as published by the Free Software Foundation; either
-   version 2.1 of the License, or (at your option) any later version.
-
-   The GNU C Library is distributed in the hope that it will be useful,
-   but WITHOUT ANY WARRANTY; without even the implied warranty of
-   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-   Lesser General Public License for more details.
-
-   You should have received a copy of the GNU Lesser General Public
-   License along with the GNU C Library; if not, see
-   <http://www.gnu.org/licenses/>.  */
-
-#include <math.h>
-#include <math_private.h>
-
-/* Return the greatest floating-point number less than X.  */
-float
-__nextdownf (float x)
-{
-  return -__nextupf (-x);
-}
-
-weak_alias (__nextdownf, nextdownf)
diff --git a/math/s_nextdownl.c b/math/s_nextdownl.c
deleted file mode 100644
index e7607f5..0000000
--- a/math/s_nextdownl.c
+++ /dev/null
@@ -1,29 +0,0 @@
-/* Return the greatest floating-point number less than X.
-   Copyright (C) 2016 Free Software Foundation, Inc.
-   This file is part of the GNU C Library.
-
-   The GNU C Library is free software; you can redistribute it and/or
-   modify it under the terms of the GNU Lesser General Public
-   License as published by the Free Software Foundation; either
-   version 2.1 of the License, or (at your option) any later version.
-
-   The GNU C Library is distributed in the hope that it will be useful,
-   but WITHOUT ANY WARRANTY; without even the implied warranty of
-   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-   Lesser General Public License for more details.
-
-   You should have received a copy of the GNU Lesser General Public
-   License along with the GNU C Library; if not, see
-   <http://www.gnu.org/licenses/>.  */
-
-#include <math.h>
-#include <math_private.h>
-
-/* Return the greatest floating-point number less than X.  */
-long double
-__nextdownl (long double x)
-{
-  return -__nextupl (-x);
-}
-
-weak_alias (__nextdownl, nextdownl)
diff --git a/sysdeps/ieee754/ldbl-opt/s_nextdownl.c b/sysdeps/ieee754/ldbl-opt/s_nextdownl.c
new file mode 100644
index 0000000..c92c02e
--- /dev/null
+++ b/sysdeps/ieee754/ldbl-opt/s_nextdownl.c
@@ -0,0 +1,5 @@
+/* nextdownl is not subject to complex aliasing rules.  It was
+   added in glibc 2.24.  */
+#define declare_mgen_alias(from, to) weak_alias (M_SUF (from), M_SUF (to))
+#include <math-type-macros-ldouble.h>
+#include <s_nextdown_template.c>
-- 
2.4.11

[PATCH 3/5] Make common fmax implementation generic.

[Paul E. Murphy]

Also update aarch64 to ensure the correct s_fmin.c is included.
The include order favors including the generated copy.

	 * math/Makefile: (gen-libm-calls): Add s_fmax.
	 (libm-calls): Remove above.

	 * math/s_fmax.c: Refactor into ...
	 * math/s_fmax_template.c: New file.
	 * math/s_fmaxf.c: Removed.
	 * math/s_fmaxl.c: Removed.

	 * sysdeps/ieee754/ldbl-opt/math-type-macros-double.h:
	 (LDOUBLE_fmaxl_libm_version): New macro.

	 * sysdeps/ieee754/ldbl-opt/s_fmax.c: Removed.
	 * sysdeps/ieee754/ldbl-opt/s_fmaxl.c: Removed.

	* sysdeps/aarch64/fpu/s_fmax.c: Append fpu/ to include of
	s_fmin.c to pick the aarch64 version of the file.
	* sysdeps/aarch64/fpu/s_fmaxf.c: Likewise.
	* sysdeps/aarch64/fpu/s_fminf.c: Likewise.
---
 math/Makefile                                      |  4 +--
 math/s_fmax.c                                      | 32 ----------------------
 math/s_fmax_template.c                             | 32 ++++++++++++++++++++++
 math/s_fmaxf.c                                     | 28 -------------------
 math/s_fmaxl.c                                     | 28 -------------------
 sysdeps/aarch64/fpu/s_fmax.c                       |  2 +-
 sysdeps/aarch64/fpu/s_fmaxf.c                      |  2 +-
 sysdeps/aarch64/fpu/s_fminf.c                      |  2 +-
 sysdeps/ieee754/ldbl-opt/math-type-macros-double.h |  1 +
 sysdeps/ieee754/ldbl-opt/s_fmax.c                  |  5 ----
 sysdeps/ieee754/ldbl-opt/s_fmaxl.c                 |  5 ----
 11 files changed, 38 insertions(+), 103 deletions(-)
 delete mode 100644 math/s_fmax.c
 create mode 100644 math/s_fmax_template.c
 delete mode 100644 math/s_fmaxf.c
 delete mode 100644 math/s_fmaxl.c
 delete mode 100644 sysdeps/ieee754/ldbl-opt/s_fmax.c
 delete mode 100644 sysdeps/ieee754/ldbl-opt/s_fmaxl.c

diff --git a/math/Makefile b/math/Makefile
index be6b8ac..09c4c56 100644
--- a/math/Makefile
+++ b/math/Makefile
@@ -49,7 +49,7 @@ gen-libm-calls = cargF conjF cimagF crealF cabsF s_cacosF		  \
 	         s_cacoshF s_ccosF s_ccoshF s_casinF s_csinF s_casinhF	  \
 		 k_casinhF s_csinhF k_casinhF s_csinhF s_catanhF s_catanF \
 		 s_ctanF s_ctanhF s_cexpF s_clogF s_cprojF s_csqrtF	  \
-		 s_cpowF s_clog10F s_fdimF s_nextdownF
+		 s_cpowF s_clog10F s_fdimF s_nextdownF s_fmaxF
 
 libm-calls =								  \
 	e_acosF e_acoshF e_asinF e_atan2F e_atanhF e_coshF e_expF e_fmodF \
@@ -65,7 +65,7 @@ libm-calls =								  \
 	w_tgammaF w_hypotF w_j0F w_j1F w_jnF w_lgammaF w_lgammaF_r	  \
 	w_logF w_log10F w_powF w_remainderF w_scalbF w_sinhF w_sqrtF	  \
 	w_ilogbF							  \
-	s_fpclassifyF s_fmaxF s_fminF s_nanF s_truncF		  	  \
+	s_fpclassifyF s_fminF s_nanF s_truncF		  	  	  \
 	s_remquoF e_log2F e_exp2F s_roundF s_nearbyintF s_sincosF	  \
 	s_fmaF s_lrintF s_llrintF s_lroundF s_llroundF e_exp10F w_log2F	  \
 	s_issignalingF $(calls:s_%=m_%) x2y2m1F				  \
diff --git a/math/s_fmax.c b/math/s_fmax.c
deleted file mode 100644
index 9a8efce..0000000
--- a/math/s_fmax.c
+++ /dev/null
@@ -1,32 +0,0 @@
-/* Return maximum numeric value of X and Y.
-   Copyright (C) 1997-2016 Free Software Foundation, Inc.
-   This file is part of the GNU C Library.
-   Contributed by Ulrich Drepper <drepper@cygnus.com>, 1997.
-
-   The GNU C Library is free software; you can redistribute it and/or
-   modify it under the terms of the GNU Lesser General Public
-   License as published by the Free Software Foundation; either
-   version 2.1 of the License, or (at your option) any later version.
-
-   The GNU C Library is distributed in the hope that it will be useful,
-   but WITHOUT ANY WARRANTY; without even the implied warranty of
-   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-   Lesser General Public License for more details.
-
-   You should have received a copy of the GNU Lesser General Public
-   License along with the GNU C Library; if not, see
-   <http://www.gnu.org/licenses/>.  */
-
-#include <math.h>
-
-
-double
-__fmax (double x, double y)
-{
-  return (isgreaterequal (x, y) || isnan (y)) ? x : y;
-}
-weak_alias (__fmax, fmax)
-#ifdef NO_LONG_DOUBLE
-strong_alias (__fmax, __fmaxl)
-weak_alias (__fmax, fmaxl)
-#endif
diff --git a/math/s_fmax_template.c b/math/s_fmax_template.c
new file mode 100644
index 0000000..dea53d4
--- /dev/null
+++ b/math/s_fmax_template.c
@@ -0,0 +1,32 @@
+/* Return maximum numeric value of X and Y.
+   Copyright (C) 1997-2016 Free Software Foundation, Inc.
+   This file is part of the GNU C Library.
+   Contributed by Ulrich Drepper <drepper@cygnus.com>, 1997.
+
+   The GNU C Library is free software; you can redistribute it and/or
+   modify it under the terms of the GNU Lesser General Public
+   License as published by the Free Software Foundation; either
+   version 2.1 of the License, or (at your option) any later version.
+
+   The GNU C Library is distributed in the hope that it will be useful,
+   but WITHOUT ANY WARRANTY; without even the implied warranty of
+   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+   Lesser General Public License for more details.
+
+   You should have received a copy of the GNU Lesser General Public
+   License along with the GNU C Library; if not, see
+   <http://www.gnu.org/licenses/>.  */
+
+#include <math.h>
+
+FLOAT
+M_DECL_FUNC (__fmax) (FLOAT x, FLOAT y)
+{
+  return (isgreaterequal (x, y) || isnan (y)) ? x : y;
+}
+
+declare_mgen_alias (__fmax, fmax);
+
+#if M_LIBM_NEED_COMPAT (fmax)
+declare_mgen_libm_compat (__fmax, fmax)
+#endif
diff --git a/math/s_fmaxf.c b/math/s_fmaxf.c
deleted file mode 100644
index c241963..0000000
--- a/math/s_fmaxf.c
+++ /dev/null
@@ -1,28 +0,0 @@
-/* Return maximum numeric value of X and Y.
-   Copyright (C) 1997-2016 Free Software Foundation, Inc.
-   This file is part of the GNU C Library.
-   Contributed by Ulrich Drepper <drepper@cygnus.com>, 1997.
-
-   The GNU C Library is free software; you can redistribute it and/or
-   modify it under the terms of the GNU Lesser General Public
-   License as published by the Free Software Foundation; either
-   version 2.1 of the License, or (at your option) any later version.
-
-   The GNU C Library is distributed in the hope that it will be useful,
-   but WITHOUT ANY WARRANTY; without even the implied warranty of
-   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-   Lesser General Public License for more details.
-
-   You should have received a copy of the GNU Lesser General Public
-   License along with the GNU C Library; if not, see
-   <http://www.gnu.org/licenses/>.  */
-
-#include <math.h>
-
-
-float
-__fmaxf (float x, float y)
-{
-  return (isgreaterequal (x, y) || isnan (y)) ? x : y;
-}
-weak_alias (__fmaxf, fmaxf)
diff --git a/math/s_fmaxl.c b/math/s_fmaxl.c
deleted file mode 100644
index 3b22735..0000000
--- a/math/s_fmaxl.c
+++ /dev/null
@@ -1,28 +0,0 @@
-/* Return maximum numeric value of X and Y.
-   Copyright (C) 1997-2016 Free Software Foundation, Inc.
-   This file is part of the GNU C Library.
-   Contributed by Ulrich Drepper <drepper@cygnus.com>, 1997.
-
-   The GNU C Library is free software; you can redistribute it and/or
-   modify it under the terms of the GNU Lesser General Public
-   License as published by the Free Software Foundation; either
-   version 2.1 of the License, or (at your option) any later version.
-
-   The GNU C Library is distributed in the hope that it will be useful,
-   but WITHOUT ANY WARRANTY; without even the implied warranty of
-   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-   Lesser General Public License for more details.
-
-   You should have received a copy of the GNU Lesser General Public
-   License along with the GNU C Library; if not, see
-   <http://www.gnu.org/licenses/>.  */
-
-#include <math.h>
-
-
-long double
-__fmaxl (long double x, long double y)
-{
-  return (isgreaterequal (x, y) || isnan (y)) ? x : y;
-}
-weak_alias (__fmaxl, fmaxl)
diff --git a/sysdeps/aarch64/fpu/s_fmax.c b/sysdeps/aarch64/fpu/s_fmax.c
index bc41ec4..c90de34 100644
--- a/sysdeps/aarch64/fpu/s_fmax.c
+++ b/sysdeps/aarch64/fpu/s_fmax.c
@@ -18,4 +18,4 @@
 
 #define FUNC fmax
 #define INSN "fmaxnm"
-#include <s_fmin.c>
+#include <fpu/s_fmin.c>
diff --git a/sysdeps/aarch64/fpu/s_fmaxf.c b/sysdeps/aarch64/fpu/s_fmaxf.c
index 6a234bb..fe9d4f3 100644
--- a/sysdeps/aarch64/fpu/s_fmaxf.c
+++ b/sysdeps/aarch64/fpu/s_fmaxf.c
@@ -20,4 +20,4 @@
 #define INSN "fmaxnm"
 #define TYPE float
 #define REGS "s"
-#include <s_fmin.c>
+#include <fpu/s_fmin.c>
diff --git a/sysdeps/aarch64/fpu/s_fminf.c b/sysdeps/aarch64/fpu/s_fminf.c
index 7860957..97c9839 100644
--- a/sysdeps/aarch64/fpu/s_fminf.c
+++ b/sysdeps/aarch64/fpu/s_fminf.c
@@ -19,4 +19,4 @@
 #define FUNC fminf
 #define TYPE float
 #define REGS "s"
-#include <s_fmin.c>
+#include <fpu/s_fmin.c>
diff --git a/sysdeps/ieee754/ldbl-opt/math-type-macros-double.h b/sysdeps/ieee754/ldbl-opt/math-type-macros-double.h
index e9b1508..5644df0 100644
--- a/sysdeps/ieee754/ldbl-opt/math-type-macros-double.h
+++ b/sysdeps/ieee754/ldbl-opt/math-type-macros-double.h
@@ -45,6 +45,7 @@
 #define LDOUBLE_clog10l_libm_version GLIBC_2_1
 #define LDOUBLE___clog10l_libm_version GLIBC_2_1
 #define LDOUBLE_fdiml_libm_version GLIBC_2_1
+#define LDOUBLE_fmaxl_libm_version GLIBC_2_1
 
 /* Define compat symbols for long double on platforms
    where it was not always a distinct type.  */
diff --git a/sysdeps/ieee754/ldbl-opt/s_fmax.c b/sysdeps/ieee754/ldbl-opt/s_fmax.c
deleted file mode 100644
index 11e7591..0000000
--- a/sysdeps/ieee754/ldbl-opt/s_fmax.c
+++ /dev/null
@@ -1,5 +0,0 @@
-#include <math_ldbl_opt.h>
-#include <math/s_fmax.c>
-#if LONG_DOUBLE_COMPAT(libm, GLIBC_2_1)
-compat_symbol (libm, __fmax, fmaxl, GLIBC_2_1);
-#endif
diff --git a/sysdeps/ieee754/ldbl-opt/s_fmaxl.c b/sysdeps/ieee754/ldbl-opt/s_fmaxl.c
deleted file mode 100644
index 98221b2..0000000
--- a/sysdeps/ieee754/ldbl-opt/s_fmaxl.c
+++ /dev/null
@@ -1,5 +0,0 @@
-#include <math_ldbl_opt.h>
-#undef weak_alias
-#define weak_alias(n,a)
-#include <math/s_fmaxl.c>
-long_double_symbol (libm, __fmaxl, fmaxl);
-- 
2.4.11

[PATCH 1/5] Make common fdim implementation generic.

[Paul E. Murphy]

The only difference is the usage of math_narrow_eval when
building s_fdiml.c.  This should be harmless for long double,
but I did observe some code generation changes on m68k, but
lack the resources to test it.

Likewise, to more easily support overriding symbol generation,
the aliasing macros are always conditionally defined on their
absence to reduce boilerplate.

I also ran builds for i486, ppc64, sparcv9, aarch64,
s390x and observed no changes to s_fdim* objects.

	 * math/Makefile (gen-libm-calls): Add s_fdim.
	 (libm-calls): Move to above.

	 * math/s_fdim.c: Renamed and refactored into to ...
	 * math/s_fdim_template.c: New file.
	 * math/s_fdiml.c: Removed.
	 * sysdeps/ieee754/ldbl-opt/s_fdim.c: Removed.
	 * sysdeps/ieee754/ldbl-opt/s_fdiml.c: Removed.

	 * sysdeps/generic/math-type-macros-double.h:
	 (declare_mgen_alias): Don't declare if already declared.
	 * sysdeps/ieee754/ldbl-opt/math-type-macros-ldouble.h:
	 Likewise.

	 * sysdeps/ieee754/ldbl-opt/math-type-macros-double.h:
	 (declare_mgen_alias): Don't declare if already declared.
	 (M_LIBM_NEED_COMPAT): Likewise.
	 (declare_mgen_libm_compat): Likewise.
	 (LDOUBLE_fdiml_libm_version): New macro.

	 * sysdeps/sparc/sparc32/sparcv9/fpu/multiarch/s_fdim.c:
	 Include math/s_fdim.c now that ldbl-opt version is not needed.
	 (declare_mgen_alias): New macro to disable aliasing.

	 * sysdeps/sparc/sparc32/sparcv9/fpu/multiarch/s_fdimf-vis3.c:
	 (declare_mgen_alias): Likewise.
	 * sysdeps/sparc/sparc32/sparcv9/fpu/multiarch/s_fdimf.c:
	 (declare_mgen_alias): Likewise.
	 * sysdeps/sparc/sparc32/sparcv9/fpu/multiarch/s_fdim-vis3.c:
	 (declare_mgen_alias): Likewise.
	 (M_LIBM_NEED_COMPAT): Likewise.
---
 math/Makefile                                      |  4 +--
 math/s_fdim.c                                      | 42 ----------------------
 math/s_fdim_template.c                             | 40 +++++++++++++++++++++
 math/s_fdimf.c                                     | 38 --------------------
 math/s_fdiml.c                                     | 35 ------------------
 sysdeps/generic/math-type-macros-double.h          |  4 +--
 sysdeps/ieee754/ldbl-opt/math-type-macros-double.h | 13 ++++---
 .../ieee754/ldbl-opt/math-type-macros-ldouble.h    |  6 ++--
 sysdeps/ieee754/ldbl-opt/s_fdim.c                  |  5 ---
 sysdeps/ieee754/ldbl-opt/s_fdiml.c                 |  5 ---
 .../sparc32/sparcv9/fpu/multiarch/s_fdim-vis3.c    |  2 ++
 .../sparc/sparc32/sparcv9/fpu/multiarch/s_fdim.c   |  3 +-
 .../sparc32/sparcv9/fpu/multiarch/s_fdimf-vis3.c   |  1 +
 .../sparc/sparc32/sparcv9/fpu/multiarch/s_fdimf.c  |  2 ++
 14 files changed, 64 insertions(+), 136 deletions(-)
 delete mode 100644 math/s_fdim.c
 create mode 100644 math/s_fdim_template.c
 delete mode 100644 math/s_fdimf.c
 delete mode 100644 math/s_fdiml.c
 delete mode 100644 sysdeps/ieee754/ldbl-opt/s_fdim.c
 delete mode 100644 sysdeps/ieee754/ldbl-opt/s_fdiml.c

diff --git a/math/Makefile b/math/Makefile
index f1b7937..d4b6d98 100644
--- a/math/Makefile
+++ b/math/Makefile
@@ -49,7 +49,7 @@ gen-libm-calls = cargF conjF cimagF crealF cabsF s_cacosF		  \
 	         s_cacoshF s_ccosF s_ccoshF s_casinF s_csinF s_casinhF	  \
 		 k_casinhF s_csinhF k_casinhF s_csinhF s_catanhF s_catanF \
 		 s_ctanF s_ctanhF s_cexpF s_clogF s_cprojF s_csqrtF	  \
-		 s_cpowF s_clog10F
+		 s_cpowF s_clog10F s_fdimF
 
 libm-calls =								  \
 	e_acosF e_acoshF e_asinF e_atan2F e_atanhF e_coshF e_expF e_fmodF \
@@ -65,7 +65,7 @@ libm-calls =								  \
 	w_tgammaF w_hypotF w_j0F w_j1F w_jnF w_lgammaF w_lgammaF_r	  \
 	w_logF w_log10F w_powF w_remainderF w_scalbF w_sinhF w_sqrtF	  \
 	w_ilogbF							  \
-	s_fpclassifyF s_fmaxF s_fminF s_fdimF s_nanF s_truncF		  \
+	s_fpclassifyF s_fmaxF s_fminF s_nanF s_truncF		  	  \
 	s_remquoF e_log2F e_exp2F s_roundF s_nearbyintF s_sincosF	  \
 	s_fmaF s_lrintF s_llrintF s_lroundF s_llroundF e_exp10F w_log2F	  \
 	s_issignalingF $(calls:s_%=m_%) x2y2m1F				  \
diff --git a/math/s_fdim.c b/math/s_fdim.c
deleted file mode 100644
index 1786521..0000000
--- a/math/s_fdim.c
+++ /dev/null
@@ -1,42 +0,0 @@
-/* Return positive difference between arguments.
-   Copyright (C) 1997-2016 Free Software Foundation, Inc.
-   This file is part of the GNU C Library.
-   Contributed by Ulrich Drepper <drepper@cygnus.com>, 1997.
-
-   The GNU C Library is free software; you can redistribute it and/or
-   modify it under the terms of the GNU Lesser General Public
-   License as published by the Free Software Foundation; either
-   version 2.1 of the License, or (at your option) any later version.
-
-   The GNU C Library is distributed in the hope that it will be useful,
-   but WITHOUT ANY WARRANTY; without even the implied warranty of
-   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-   Lesser General Public License for more details.
-
-   You should have received a copy of the GNU Lesser General Public
-   License along with the GNU C Library; if not, see
-   <http://www.gnu.org/licenses/>.  */
-
-#include <errno.h>
-#include <math.h>
-#include <math_private.h>
-
-double
-__fdim (double x, double y)
-{
-  if (islessequal (x, y))
-    return 0.0;
-
-  double r = math_narrow_eval (x - y);
-  if (isinf (r) && !isinf (x) && !isinf (y))
-    __set_errno (ERANGE);
-
-  return r;
-}
-#ifndef __fdim
-weak_alias (__fdim, fdim)
-# ifdef NO_LONG_DOUBLE
-strong_alias (__fdim, __fdiml)
-weak_alias (__fdim, fdiml)
-# endif
-#endif
diff --git a/math/s_fdim_template.c b/math/s_fdim_template.c
new file mode 100644
index 0000000..d3e7379
--- /dev/null
+++ b/math/s_fdim_template.c
@@ -0,0 +1,40 @@
+/* Return positive difference between arguments.
+   Copyright (C) 1997-2016 Free Software Foundation, Inc.
+   This file is part of the GNU C Library.
+   Contributed by Ulrich Drepper <drepper@cygnus.com>, 1997.
+
+   The GNU C Library is free software; you can redistribute it and/or
+   modify it under the terms of the GNU Lesser General Public
+   License as published by the Free Software Foundation; either
+   version 2.1 of the License, or (at your option) any later version.
+
+   The GNU C Library is distributed in the hope that it will be useful,
+   but WITHOUT ANY WARRANTY; without even the implied warranty of
+   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+   Lesser General Public License for more details.
+
+   You should have received a copy of the GNU Lesser General Public
+   License along with the GNU C Library; if not, see
+   <http://www.gnu.org/licenses/>.  */
+
+#include <errno.h>
+#include <math.h>
+#include <math_private.h>
+
+FLOAT
+M_DECL_FUNC (__fdim) (FLOAT x, FLOAT y)
+{
+  if (islessequal (x, y))
+    return 0;
+
+  FLOAT r = math_narrow_eval (x - y);
+  if (isinf (r) && !isinf (x) && !isinf (y))
+    __set_errno (ERANGE);
+
+  return r;
+}
+declare_mgen_alias (__fdim, fdim);
+
+#if M_LIBM_NEED_COMPAT (fdim)
+declare_mgen_libm_compat (__fdim, fdim)
+#endif
diff --git a/math/s_fdimf.c b/math/s_fdimf.c
deleted file mode 100644
index e457f9d..0000000
--- a/math/s_fdimf.c
+++ /dev/null
@@ -1,38 +0,0 @@
-/* Return positive difference between arguments.
-   Copyright (C) 1997-2016 Free Software Foundation, Inc.
-   This file is part of the GNU C Library.
-   Contributed by Ulrich Drepper <drepper@cygnus.com>, 1997.
-
-   The GNU C Library is free software; you can redistribute it and/or
-   modify it under the terms of the GNU Lesser General Public
-   License as published by the Free Software Foundation; either
-   version 2.1 of the License, or (at your option) any later version.
-
-   The GNU C Library is distributed in the hope that it will be useful,
-   but WITHOUT ANY WARRANTY; without even the implied warranty of
-   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-   Lesser General Public License for more details.
-
-   You should have received a copy of the GNU Lesser General Public
-   License along with the GNU C Library; if not, see
-   <http://www.gnu.org/licenses/>.  */
-
-#include <errno.h>
-#include <math.h>
-#include <math_private.h>
-
-float
-__fdimf (float x, float y)
-{
-  if (islessequal (x, y))
-    return 0.0f;
-
-  float r = math_narrow_eval (x - y);
-  if (isinf (r) && !isinf (x) && !isinf (y))
-    __set_errno (ERANGE);
-
-  return r;
-}
-#ifndef __fdimf
-weak_alias (__fdimf, fdimf)
-#endif
diff --git a/math/s_fdiml.c b/math/s_fdiml.c
deleted file mode 100644
index 4a1f672..0000000
--- a/math/s_fdiml.c
+++ /dev/null
@@ -1,35 +0,0 @@
-/* Return positive difference between arguments.
-   Copyright (C) 1997-2016 Free Software Foundation, Inc.
-   This file is part of the GNU C Library.
-   Contributed by Ulrich Drepper <drepper@cygnus.com>, 1997.
-
-   The GNU C Library is free software; you can redistribute it and/or
-   modify it under the terms of the GNU Lesser General Public
-   License as published by the Free Software Foundation; either
-   version 2.1 of the License, or (at your option) any later version.
-
-   The GNU C Library is distributed in the hope that it will be useful,
-   but WITHOUT ANY WARRANTY; without even the implied warranty of
-   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-   Lesser General Public License for more details.
-
-   You should have received a copy of the GNU Lesser General Public
-   License along with the GNU C Library; if not, see
-   <http://www.gnu.org/licenses/>.  */
-
-#include <errno.h>
-#include <math.h>
-
-long double
-__fdiml (long double x, long double y)
-{
-  if (islessequal (x, y))
-    return 0.0f;
-
-  long double r = x - y;
-  if (isinf (r) && !isinf (x) && !isinf (y))
-    __set_errno (ERANGE);
-
-  return r;
-}
-weak_alias (__fdiml, fdiml)
diff --git a/sysdeps/generic/math-type-macros-double.h b/sysdeps/generic/math-type-macros-double.h
index 4149b39..154f416 100644
--- a/sysdeps/generic/math-type-macros-double.h
+++ b/sysdeps/generic/math-type-macros-double.h
@@ -29,12 +29,12 @@
 /* Machines without a distinct long double type
    alias long double functions to their double
    equivalent.  */
-#if defined NO_LONG_DOUBLE
+#if defined NO_LONG_DOUBLE && !defined declare_mgen_alias
 # define declare_mgen_alias(from, to)	    \
    weak_alias (from, to)		    \
    strong_alias (from, from ## l)	    \
    weak_alias (from, to ## l)
-#else
+#elif !defined declare_mgen_alias
 # define declare_mgen_alias(from, to)	    \
    weak_alias (M_SUF (from), M_SUF (to))
 #endif
diff --git a/sysdeps/ieee754/ldbl-opt/math-type-macros-double.h b/sysdeps/ieee754/ldbl-opt/math-type-macros-double.h
index 8cb5694..e9b1508 100644
--- a/sysdeps/ieee754/ldbl-opt/math-type-macros-double.h
+++ b/sysdeps/ieee754/ldbl-opt/math-type-macros-double.h
@@ -44,15 +44,20 @@
 #define LDOUBLE_cpowl_libm_version GLIBC_2_1
 #define LDOUBLE_clog10l_libm_version GLIBC_2_1
 #define LDOUBLE___clog10l_libm_version GLIBC_2_1
+#define LDOUBLE_fdiml_libm_version GLIBC_2_1
 
 /* Define compat symbols for long double on platforms
    where it was not always a distinct type.  */
-#define M_LIBM_NEED_COMPAT(f) \
-  LONG_DOUBLE_COMPAT (libm, LDOUBLE_ ## f ## l_libm_version)
+#if !defined M_LIBM_NEED_COMPAT
+# define M_LIBM_NEED_COMPAT(f) \
+   LONG_DOUBLE_COMPAT (libm, LDOUBLE_ ## f ## l_libm_version)
+#endif
 
-#define declare_mgen_libm_compat(from, to)	      \
-  compat_symbol (libm, from, to ## l,		      \
+#if !defined declare_mgen_libm_compat
+# define declare_mgen_libm_compat(from, to)	      \
+   compat_symbol (libm, from, to ## l,		      \
 		 LDOUBLE_ ## to ## l_libm_version);
+#endif
 
 #include_next <math-type-macros-double.h>
 #endif
diff --git a/sysdeps/ieee754/ldbl-opt/math-type-macros-ldouble.h b/sysdeps/ieee754/ldbl-opt/math-type-macros-ldouble.h
index d2af4bb..30cb2fb 100644
--- a/sysdeps/ieee754/ldbl-opt/math-type-macros-ldouble.h
+++ b/sysdeps/ieee754/ldbl-opt/math-type-macros-ldouble.h
@@ -22,8 +22,10 @@
 
 /* Use properly versioned symbols for long double on platforms where
    it was not always a distinct type.  */
-#define declare_mgen_alias(from, to) \
-  long_double_symbol (libm, from ## l, to ## l);
+#if !defined declare_mgen_alias
+# define declare_mgen_alias(from, to) \
+   long_double_symbol (libm, from ## l, to ## l);
+#endif
 
 #include_next <math-type-macros-ldouble.h>
 #endif
diff --git a/sysdeps/ieee754/ldbl-opt/s_fdim.c b/sysdeps/ieee754/ldbl-opt/s_fdim.c
deleted file mode 100644
index 02c95bf..0000000
--- a/sysdeps/ieee754/ldbl-opt/s_fdim.c
+++ /dev/null
@@ -1,5 +0,0 @@
-#include <math_ldbl_opt.h>
-#include <math/s_fdim.c>
-#if LONG_DOUBLE_COMPAT(libm, GLIBC_2_1)
-compat_symbol (libm, __fdim, fdiml, GLIBC_2_1);
-#endif
diff --git a/sysdeps/ieee754/ldbl-opt/s_fdiml.c b/sysdeps/ieee754/ldbl-opt/s_fdiml.c
deleted file mode 100644
index 06b760b..0000000
--- a/sysdeps/ieee754/ldbl-opt/s_fdiml.c
+++ /dev/null
@@ -1,5 +0,0 @@
-#include <math_ldbl_opt.h>
-#undef weak_alias
-#define weak_alias(n,a)
-#include <math/s_fdiml.c>
-long_double_symbol (libm, __fdiml, fdiml);
diff --git a/sysdeps/sparc/sparc32/sparcv9/fpu/multiarch/s_fdim-vis3.c b/sysdeps/sparc/sparc32/sparcv9/fpu/multiarch/s_fdim-vis3.c
index 2973b49..6176517 100644
--- a/sysdeps/sparc/sparc32/sparcv9/fpu/multiarch/s_fdim-vis3.c
+++ b/sysdeps/sparc/sparc32/sparcv9/fpu/multiarch/s_fdim-vis3.c
@@ -19,5 +19,7 @@
 #include <math.h>
 
 #define __fdim __fdim_vis3
+#define declare_mgen_alias(t, f)
+#define M_LIBM_NEED_COMPAT(f) 0
 
 #include <math/s_fdim.c>
diff --git a/sysdeps/sparc/sparc32/sparcv9/fpu/multiarch/s_fdim.c b/sysdeps/sparc/sparc32/sparcv9/fpu/multiarch/s_fdim.c
index 9666741..2d07f31 100644
--- a/sysdeps/sparc/sparc32/sparcv9/fpu/multiarch/s_fdim.c
+++ b/sysdeps/sparc/sparc32/sparcv9/fpu/multiarch/s_fdim.c
@@ -27,6 +27,7 @@ sparc_libm_ifunc(__fdim, hwcap & HWCAP_SPARC_VIS3 ? __fdim_vis3 : __fdim_generic
 weak_alias (__fdim, fdim)
 
 # define __fdim __fdim_generic
+# define declare_mgen_alias(t, f)
 #endif
 
-#include <ldbl-opt/s_fdim.c>
+#include <math/s_fdim.c>
diff --git a/sysdeps/sparc/sparc32/sparcv9/fpu/multiarch/s_fdimf-vis3.c b/sysdeps/sparc/sparc32/sparcv9/fpu/multiarch/s_fdimf-vis3.c
index 75997c6..3f35b0d 100644
--- a/sysdeps/sparc/sparc32/sparcv9/fpu/multiarch/s_fdimf-vis3.c
+++ b/sysdeps/sparc/sparc32/sparcv9/fpu/multiarch/s_fdimf-vis3.c
@@ -19,5 +19,6 @@
 #include <math.h>
 
 #define __fdimf __fdimf_vis3
+#define declare_mgen_alias(t, f)
 
 #include <math/s_fdimf.c>
diff --git a/sysdeps/sparc/sparc32/sparcv9/fpu/multiarch/s_fdimf.c b/sysdeps/sparc/sparc32/sparcv9/fpu/multiarch/s_fdimf.c
index 767520f..b9add1c 100644
--- a/sysdeps/sparc/sparc32/sparcv9/fpu/multiarch/s_fdimf.c
+++ b/sysdeps/sparc/sparc32/sparcv9/fpu/multiarch/s_fdimf.c
@@ -27,6 +27,8 @@ sparc_libm_ifunc(__fdimf, hwcap & HWCAP_SPARC_VIS3 ? __fdimf_vis3 : __fdimf_gene
 weak_alias (__fdimf, fdimf)
 
 # define __fdimf __fdimf_generic
+# define declare_mgen_alias(t, f)
+
 #endif
 
 #include <math/s_fdimf.c>
-- 
2.4.11

Re: [PATCH 5/5] Extend generic k_rem_pio2 implementation using generics.

[Andreas Schwab]

On Aug 31 2016, "Paul E. Murphy" <murphyp@linux.vnet.ibm.com> wrote:

> diff --git a/math/k_rem_pio2_template.c b/math/k_rem_pio2_template.c
> new file mode 100644
> index 0000000..4bb79f6
> --- /dev/null
> +++ b/math/k_rem_pio2_template.c
> @@ -0,0 +1,2 @@
> +/* This function is not necessary for all formats/types.  If necessary, override this

Overlong line.

Andreas.

-- 
Andreas Schwab, SUSE Labs, schwab@suse.de
GPG Key fingerprint = 0196 BAD8 1CE9 1970 F4BE  1748 E4D4 88E3 0EEA B9D7
"And now for something completely different."

Re: [PATCH 5/5] Extend generic k_rem_pio2 implementation using generics.

[Joseph Myers]

On Tue, 30 Aug 2016, Paul E. Murphy wrote:

> This is actually just an internal stub.  It is already
> implemented where needed.  This ensures there is always
> an empty stub for each type/format when it is not needed.

I think it would be better to list this in type-double-routines and 
type-float-routines (and remove all the k_rem_pio2l source files, not just 
the math/ one).  That is, as a function present only for float and double, 
just list it as such.

-- 
Joseph S. Myers
joseph@codesourcery.com

Re: [PATCH 1/5] Make common fdim implementation generic.

[Joseph Myers]

On Tue, 30 Aug 2016, Paul E. Murphy wrote:

>  /* Define compat symbols for long double on platforms
>     where it was not always a distinct type.  */
> -#define M_LIBM_NEED_COMPAT(f) \
> -  LONG_DOUBLE_COMPAT (libm, LDOUBLE_ ## f ## l_libm_version)
> +#if !defined M_LIBM_NEED_COMPAT
> +# define M_LIBM_NEED_COMPAT(f) \
> +   LONG_DOUBLE_COMPAT (libm, LDOUBLE_ ## f ## l_libm_version)

Two-column indentation is still correct for the continuation line; it 
should not be changed.

> -#define declare_mgen_libm_compat(from, to)	      \
> -  compat_symbol (libm, from, to ## l,		      \
> +#if !defined declare_mgen_libm_compat
> +# define declare_mgen_libm_compat(from, to)	      \
> +   compat_symbol (libm, from, to ## l,		      \
>  		 LDOUBLE_ ## to ## l_libm_version);

Likewise here (if a change were correct, the following line would have 
needed reindenting as well to stay aligned with the '(').

> -#define declare_mgen_alias(from, to) \
> -  long_double_symbol (libm, from ## l, to ## l);
> +#if !defined declare_mgen_alias
> +# define declare_mgen_alias(from, to) \
> +   long_double_symbol (libm, from ## l, to ## l);

Likewise here.

OK with those fixes.

-- 
Joseph S. Myers
joseph@codesourcery.com

Re: [PATCH 2/5] Make common nextdown implementation generic.

[Joseph Myers]

On Tue, 30 Aug 2016, Paul E. Murphy wrote:

> With the exception of those machines using the ldbl-opt in
> an Implies file, this is a trivial transformation.
> 
> nextdownl is not subject to the non-trivial versioning rules
> of the other generated functions, so to keep things simple,
> it is handled as a one-off case in ldbl-opt to preserve the
> existing behavior.
> 
> 	* math/Makefile: (gen-libm-calls): Add s_nextdown.
> 	(libm-calls): Remove above.
> 
> 	* math/s_nextdown.c: Refactor into ...
> 	* math/s_nextdown_template.c: New file.
> 	* math/s_nextdownf.c: Removed.
> 	* math/s_nextdownl.c: Removed.
> 
> 	* sysdeps/ieee754/ldbl-opt/s_nextdownl.c: New file.

OK.

-- 
Joseph S. Myers
joseph@codesourcery.com

Re: [PATCH 4/5] Make common fmin implementation generic.

[Joseph Myers]

On Tue, 30 Aug 2016, Paul E. Murphy wrote:

> 	 * math/Makefile (gen-libm-calls): Add s_fmin
> 	 (libm-calls): Remove above.
> 
> 	 * math/s_fmin.c: Refactor into ...
> 	 * math/s_fmin_template.c: New file.
> 
> 	 * math/s_fminf.c: Removed.
> 	 * math/s_fminl.c: Removed.
> 	 * sysdeps/ieee754/ldbl-opt/math-type-macros-double.h:
> 	 (LDOUBLE_fmin_libm_version): New macro.
> 
> 	 * sysdeps/ieee754/ldbl-opt/s_fmin.c: Removed.
> 	 * sysdeps/ieee754/ldbl-opt/s_fminl.c: Removed.

OK.

-- 
Joseph S. Myers
joseph@codesourcery.com

Re: [PATCH 3/5] Make common fmax implementation generic.

[Joseph Myers]

On Tue, 30 Aug 2016, Paul E. Murphy wrote:

> Also update aarch64 to ensure the correct s_fmin.c is included.
> The include order favors including the generated copy.
> 
> 	 * math/Makefile: (gen-libm-calls): Add s_fmax.
> 	 (libm-calls): Remove above.
> 
> 	 * math/s_fmax.c: Refactor into ...
> 	 * math/s_fmax_template.c: New file.
> 	 * math/s_fmaxf.c: Removed.
> 	 * math/s_fmaxl.c: Removed.
> 
> 	 * sysdeps/ieee754/ldbl-opt/math-type-macros-double.h:
> 	 (LDOUBLE_fmaxl_libm_version): New macro.
> 
> 	 * sysdeps/ieee754/ldbl-opt/s_fmax.c: Removed.
> 	 * sysdeps/ieee754/ldbl-opt/s_fmaxl.c: Removed.
> 
> 	* sysdeps/aarch64/fpu/s_fmax.c: Append fpu/ to include of
> 	s_fmin.c to pick the aarch64 version of the file.
> 	* sysdeps/aarch64/fpu/s_fmaxf.c: Likewise.
> 	* sysdeps/aarch64/fpu/s_fminf.c: Likewise.

OK.

-- 
Joseph S. Myers
joseph@codesourcery.com

Re: [PATCH 1/5] Make common fdim implementation generic.

[Paul E. Murphy]

On 08/31/2016 11:46 AM, Joseph Myers wrote:
> On Tue, 30 Aug 2016, Paul E. Murphy wrote:
> 
>>  /* Define compat symbols for long double on platforms
>>     where it was not always a distinct type.  */
>> -#define M_LIBM_NEED_COMPAT(f) \
>> -  LONG_DOUBLE_COMPAT (libm, LDOUBLE_ ## f ## l_libm_version)
>> +#if !defined M_LIBM_NEED_COMPAT
>> +# define M_LIBM_NEED_COMPAT(f) \
>> +   LONG_DOUBLE_COMPAT (libm, LDOUBLE_ ## f ## l_libm_version)
> 
> Two-column indentation is still correct for the continuation line; it 
> should not be changed.
> 
>> -#define declare_mgen_libm_compat(from, to)	      \
>> -  compat_symbol (libm, from, to ## l,		      \
>> +#if !defined declare_mgen_libm_compat
>> +# define declare_mgen_libm_compat(from, to)	      \
>> +   compat_symbol (libm, from, to ## l,		      \
>>  		 LDOUBLE_ ## to ## l_libm_version);
> 
> Likewise here (if a change were correct, the following line would have 
> needed reindenting as well to stay aligned with the '(').
> 
>> -#define declare_mgen_alias(from, to) \
>> -  long_double_symbol (libm, from ## l, to ## l);
>> +#if !defined declare_mgen_alias
>> +# define declare_mgen_alias(from, to) \
>> +   long_double_symbol (libm, from ## l, to ## l);
> 
> Likewise here.
> 
> OK with those fixes.
> 

Fixes made, and committed as 7b7c394.

Re: [PATCH 3/5] Make common fmax implementation generic.

[Paul E. Murphy]

On 08/31/2016 11:55 AM, Joseph Myers wrote:
> On Tue, 30 Aug 2016, Paul E. Murphy wrote:
> 
>> Also update aarch64 to ensure the correct s_fmin.c is included.
>> The include order favors including the generated copy.
>>
>> 	 * math/Makefile: (gen-libm-calls): Add s_fmax.
>> 	 (libm-calls): Remove above.
>>
>> 	 * math/s_fmax.c: Refactor into ...
>> 	 * math/s_fmax_template.c: New file.
>> 	 * math/s_fmaxf.c: Removed.
>> 	 * math/s_fmaxl.c: Removed.
>>
>> 	 * sysdeps/ieee754/ldbl-opt/math-type-macros-double.h:
>> 	 (LDOUBLE_fmaxl_libm_version): New macro.
>>
>> 	 * sysdeps/ieee754/ldbl-opt/s_fmax.c: Removed.
>> 	 * sysdeps/ieee754/ldbl-opt/s_fmaxl.c: Removed.
>>
>> 	* sysdeps/aarch64/fpu/s_fmax.c: Append fpu/ to include of
>> 	s_fmin.c to pick the aarch64 version of the file.
>> 	* sysdeps/aarch64/fpu/s_fmaxf.c: Likewise.
>> 	* sysdeps/aarch64/fpu/s_fminf.c: Likewise.
> 
> OK.
> 

Committed as 847c916.

Re: [PATCH 4/5] Make common fmin implementation generic.

[Paul E. Murphy]

On 08/31/2016 11:55 AM, Joseph Myers wrote:
> On Tue, 30 Aug 2016, Paul E. Murphy wrote:
> 
>> 	 * math/Makefile (gen-libm-calls): Add s_fmin
>> 	 (libm-calls): Remove above.
>>
>> 	 * math/s_fmin.c: Refactor into ...
>> 	 * math/s_fmin_template.c: New file.
>>
>> 	 * math/s_fminf.c: Removed.
>> 	 * math/s_fminl.c: Removed.
>> 	 * sysdeps/ieee754/ldbl-opt/math-type-macros-double.h:
>> 	 (LDOUBLE_fmin_libm_version): New macro.
>>
>> 	 * sysdeps/ieee754/ldbl-opt/s_fmin.c: Removed.
>> 	 * sysdeps/ieee754/ldbl-opt/s_fminl.c: Removed.
> 
> OK.
> 

Committed as f306ea1.

Re: [PATCH 2/5] Make common nextdown implementation generic.

[Paul E. Murphy]

On 08/31/2016 11:54 AM, Joseph Myers wrote:
> On Tue, 30 Aug 2016, Paul E. Murphy wrote:
> 
>> With the exception of those machines using the ldbl-opt in
>> an Implies file, this is a trivial transformation.
>>
>> nextdownl is not subject to the non-trivial versioning rules
>> of the other generated functions, so to keep things simple,
>> it is handled as a one-off case in ldbl-opt to preserve the
>> existing behavior.
>>
>> 	* math/Makefile: (gen-libm-calls): Add s_nextdown.
>> 	(libm-calls): Remove above.
>>
>> 	* math/s_nextdown.c: Refactor into ...
>> 	* math/s_nextdown_template.c: New file.
>> 	* math/s_nextdownf.c: Removed.
>> 	* math/s_nextdownl.c: Removed.
>>
>> 	* sysdeps/ieee754/ldbl-opt/s_nextdownl.c: New file.
> 
> OK.
> 

Committed as ee8a490.