Assuming the 17 sub-patches within patch #7 are approved, this patch adds the
necessary bits to libgcc to build the __float128 emulation functions.
Originally, it was thought that since it uses the TFmode emulation functions
from the libgcc/soft-fp directory, that we would need to submit the patches
through glibc where the master source lives.
However, since we don't do any modification of the soft-fp sources, but do it
all through support in the libgcc/config/rs6000 directory, I'm submitting this
patch to libgcc.
This patch passes bootstrap and regression testing on a little endian power8
system (previous versions of the patch also were tested on a big endian power7
system). Once the patch #7 sub-patches are checked in, is this patch
appropriate to be checked in?
2015-10-27 Michael Meissner <[email protected]>
* Makefile.in (EXTRA_CLEAN): New variable of additional files to
clean.
(FLAGS_TO_PASS): Add EXTRA_CLEAN.
(clean): Delete EXTRA_CLEAN files.
* configure.ac (powerpc*-*-linux*): If the compiler is capable of
generating __float128 code, build the __float128 emulation
libraries.
* configure: Regenerate.
* config/rs6000/extendkftf2.c: New files for IEEE 128-bit floating
point support.
* config/rs6000/cmpukf2.c: Likewise.
* config/rs6000/trunctfk2.c: Likewise.
* config/rs6000/quad-float128.h: Likewise.
* config/rs6000/t-float128: Likewise.
* config/rs6000/float128-sed: Likewise.
* config/rs6000/sfp-machine.h (_FP_W_TYPE_SIZE): Update
defintiions to use 64-bit types for 64-bit systems or 32-bit VSX
systems with IEEE 128-bit floating point support.
(_FP_W_TYPE): Likewise.
(_FP_WS_TYPE): Likewise.
(_FP_I_TYPE): Likewise.
(TItype): Add defintions for IEEE 128-bit floating point support.
(UTItype): Likewise.
(TI_BITS): Likewise.
(_FP_MUL_MEAT_Q): Use 64-bit types on 64-bit systems and for
32-bit VSX systems with IEEE 128-bit floating point.
(_FP_DIV_MEAT_Q): Likewise.
(_FP_NANFRAC_Q): Likewise.
* config/rs6000/libgcc-glibc.ver: Add requires for GCC 6.0 adding
IEEE 128-bit floating point support.
--
Michael Meissner, IBM
IBM, M/S 2506R, 550 King Street, Littleton, MA 01460-6245, USA
email: [email protected], phone: +1 (978) 899-4797
Index: libgcc/config.host
===================================================================
--- libgcc/config.host (revision 229390)
+++ libgcc/config.host (working copy)
@@ -1058,6 +1058,11 @@ powerpc*-*-linux*)
exit 1
;;
esac
+
+ if test $libgcc_cv_powerpc_float128 = yes; then
+ tmake_file="${tmake_file} rs6000/t-float128"
+ fi
+
extra_parts="$extra_parts ecrti.o ecrtn.o ncrti.o ncrtn.o"
md_unwind_header=rs6000/linux-unwind.h
;;
Index: libgcc/Makefile.in
===================================================================
--- libgcc/Makefile.in (revision 229390)
+++ libgcc/Makefile.in (working copy)
@@ -70,6 +70,8 @@ INSTALL_DATA = @INSTALL_DATA@
mkinstalldirs = $(SHELL) $(libgcc_topdir)/mkinstalldirs
INSTALL_PARTS = $(EXTRA_PARTS)
+EXTRA_CLEAN =
+
objext = .o
AR = @AR@
@@ -90,6 +92,7 @@ FLAGS_TO_PASS = \
"CFLAGS=$(CFLAGS)" \
"DESTDIR=$(DESTDIR)" \
"EXTRA_OFILES=$(EXTRA_OFILES)" \
+ "EXTRA_CLEAN=$(EXTRA_CLEAN)" \
"HDEFINES=$(HDEFINES)" \
"INSTALL=$(INSTALL)" \
"INSTALL_DATA=$(INSTALL_DATA)" \
@@ -128,7 +131,7 @@ installcheck:
clean:
-rm -f libgcc_tm.h libgcc.map
-rm -f libgcc_tm.stamp stmp-ldirs
- -rm -f *$(objext)
+ -rm -f *$(objext) $(EXTRA_CLEAN)
-rm -f *.dep
-rm -f *.a
-rm -f libunwind$(SHLIB_EXT)
Index: libgcc/configure.ac
===================================================================
--- libgcc/configure.ac (revision 229390)
+++ libgcc/configure.ac (working copy)
@@ -373,6 +373,17 @@ esac
;;
esac
+case ${host} in
+powerpc*-*-linux*)
+ AC_CACHE_CHECK([whether the PowerPC compiler can do __float128],
+ [libgcc_cv_powerpc_float128],
+ [AC_COMPILE_IFELSE(
+ [#pragma GCC target ("vsx,float128")
+ __float128 add (__float128 *a) { return *a + *(a+1); }],
+ [libgcc_cv_powerpc_float128=yes],
+ [libgcc_cv_powerpc_float128=no])])
+esac
+
# Collect host-machine-specific information.
. ${srcdir}/config.host
Index: libgcc/config/rs6000/extendkftf2.c
===================================================================
--- libgcc/config/rs6000/extendkftf2.c (revision 0)
+++ libgcc/config/rs6000/extendkftf2.c (revision 0)
@@ -0,0 +1,59 @@
+/* Software IEEE 128-bit floating-point emulation for PowerPC.
+ Copyright (C) 2014 Free Software Foundation, Inc.
+ This file is part of the GNU C Library.
+ Contributed by Michael Meissner ([email protected])
+ Code is based on the main soft-fp library written by:
+ Richard Henderson ([email protected]) and
+ Jakub Jelinek ([email protected]).
+
+ 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.
+
+ In addition to the permissions in the GNU Lesser General Public
+ License, the Free Software Foundation gives you unlimited
+ permission to link the compiled version of this file into
+ combinations with other programs, and to distribute those
+ combinations without any restriction coming from the use of this
+ file. (The Lesser General Public License restrictions do apply in
+ other respects; for example, they cover modification of the file,
+ and distribution when not linked into a combine executable.)
+
+ 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/>. */
+
+/* Convert IEEE 128-bit floating point to IBM long double. */
+
+/* Force the use of the VSX instruction set. */
+#if defined(_ARCH_PPC) && (!defined(__VSX__) || !defined(__FLOAT128__))
+#pragma GCC target ("vsx,float128")
+#endif
+
+extern __ibm128 __extendkftf2 (__float128);
+
+__ibm128
+__extendkftf2 (__float128 value)
+{
+ double high, low;
+
+ high = (double) value;
+ if (__builtin_isnan (high) || __builtin_isinf (high))
+ low = high;
+
+ else
+ {
+ low = (double) (value - (__float128)high);
+
+ /* Use copysign to propigate the sign bit so that -0.0Q becomes -0.0L.
*/
+ low = __builtin_copysign (low, high);
+ }
+
+ return __builtin_pack_longdouble (high, low);
+}
Index: libgcc/config/rs6000/sfp-machine.h
===================================================================
--- libgcc/config/rs6000/sfp-machine.h (revision 229390)
+++ libgcc/config/rs6000/sfp-machine.h (working copy)
@@ -1,7 +1,26 @@
+/* Decide whether to use 64 or 32-bit types to do the emulation. If we are
+ doing IEEE-128 with VSX, use 64-bit emulation even if we are compiling for a
+ 32-bit target. */
+
+#if defined(_ARCH_PPC64) || defined(__VSX__) || defined(__FLOAT128__)
+#define _FP_W_TYPE_SIZE 64
+#define _FP_W_TYPE unsigned long long
+#define _FP_WS_TYPE signed long long
+#define _FP_I_TYPE long long
+
+#ifdef _ARCH_PPC64
+typedef int TItype __attribute__ ((mode (TI)));
+typedef unsigned int UTItype __attribute__ ((mode (TI)));
+
+#define TI_BITS (__CHAR_BIT__ * (int)sizeof(TItype))
+#endif
+
+#else /* 32-bits */
#define _FP_W_TYPE_SIZE 32
-#define _FP_W_TYPE unsigned long
-#define _FP_WS_TYPE signed long
-#define _FP_I_TYPE long
+#define _FP_W_TYPE unsigned int
+#define _FP_WS_TYPE signed int
+#define _FP_I_TYPE int
+#endif /* 32-bits */
/* The type of the result of a floating point comparison. This must
match `__libgcc_cmp_return__' in GCC for the target. */
@@ -12,16 +31,34 @@ typedef int __gcc_CMPtype __attribute__
_FP_MUL_MEAT_1_wide(_FP_WFRACBITS_S,R,X,Y,umul_ppmm)
#define _FP_MUL_MEAT_D(R,X,Y) \
_FP_MUL_MEAT_2_wide(_FP_WFRACBITS_D,R,X,Y,umul_ppmm)
+
+#if (_FP_W_TYPE_SIZE==64)
+#define _FP_MUL_MEAT_Q(R,X,Y) \
+ _FP_MUL_MEAT_2_wide(_FP_WFRACBITS_Q,R,X,Y,umul_ppmm)
+
+#else
#define _FP_MUL_MEAT_Q(R,X,Y) \
_FP_MUL_MEAT_4_wide(_FP_WFRACBITS_Q,R,X,Y,umul_ppmm)
+#endif
#define _FP_DIV_MEAT_S(R,X,Y) _FP_DIV_MEAT_1_loop(S,R,X,Y)
#define _FP_DIV_MEAT_D(R,X,Y) _FP_DIV_MEAT_2_udiv(D,R,X,Y)
+
+#if (_FP_W_TYPE_SIZE==64)
+#define _FP_DIV_MEAT_Q(R,X,Y) _FP_DIV_MEAT_2_udiv(Q,R,X,Y)
+#else
#define _FP_DIV_MEAT_Q(R,X,Y) _FP_DIV_MEAT_4_udiv(Q,R,X,Y)
+#endif
#define _FP_NANFRAC_S ((_FP_QNANBIT_S << 1) - 1)
#define _FP_NANFRAC_D ((_FP_QNANBIT_D << 1) - 1), -1
+
+#if (_FP_W_TYPE_SIZE==64)
+#define _FP_NANFRAC_Q ((_FP_QNANBIT_Q << 1) - 1), -1
+#else
#define _FP_NANFRAC_Q ((_FP_QNANBIT_Q << 1) - 1), -1, -1, -1
+#endif
+
#define _FP_NANSIGN_S 0
#define _FP_NANSIGN_D 0
#define _FP_NANSIGN_Q 0
Index: libgcc/config/rs6000/cmpukf2.c
===================================================================
--- libgcc/config/rs6000/cmpukf2.c (revision 0)
+++ libgcc/config/rs6000/cmpukf2.c (revision 0)
@@ -0,0 +1,87 @@
+/* Software IEEE 128-bit floating-point emulation for PowerPC.
+ Copyright (C) 2015 Free Software Foundation, Inc.
+ This file is part of the GNU C Library.
+ Contributed by Michael Meissner ([email protected])
+ Code is based on the main soft-fp library written by:
+ Richard Henderson ([email protected]) and
+ Jakub Jelinek ([email protected]).
+
+ 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.
+
+ In addition to the permissions in the GNU Lesser General Public
+ License, the Free Software Foundation gives you unlimited
+ permission to link the compiled version of this file into
+ combinations with other programs, and to distribute those
+ combinations without any restriction coming from the use of this
+ file. (The Lesser General Public License restrictions do apply in
+ other respects; for example, they cover modification of the file,
+ and distribution when not linked into a combine executable.)
+
+ 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/>. */
+
+/* Force the use of the VSX instruction set. */
+#if defined(_ARCH_PPC) && (!defined(__VSX__) || !defined(__FLOAT128__))
+#pragma GCC target ("vsx,float128")
+#endif
+
+#include "soft-fp.h"
+#include "double.h"
+#include "single.h"
+#include "quad-float128.h"
+
+/* PowerPC condition register bits. */
+#define PPC_UNORDERED 0x1 /* isnan (a) || isnan (b). */
+#define PPC_EQUAL 0x2 /* a == b. */
+#define PPC_GREATER_THEN 0x4 /* a > b. */
+#define PPC_LESS_THEN 0x8 /* a < b. */
+
+/* Map FP_CMP_Q output to PowerPC condition register bits. */
+#define CMP_UNORDERED (-2) /* isnan (a) || isnan (b). */
+#define CMP_LESS_THEN (-1) /* a < b. */
+#define CMP_EQUAL 0 /* a == b. */
+#define CMP_GREATER_THEN 1 /* a < b. */
+#define CMP_INVALID 2 /* raise invalid exception. */
+
+#define CMP_LOW CMP_UNORDERED /* comparison low
value. */
+#define CMP_HIGH CMP_INVALID /* comparison high value. */
+
+static const unsigned char ppc_cr_map[] = {
+ PPC_UNORDERED, /* -2: unordered. */
+ PPC_LESS_THEN, /* -1: a < b. */
+ PPC_EQUAL, /* 0: a == b. */
+ PPC_GREATER_THEN, /* 1: a > b. */
+ PPC_UNORDERED, /* 2: invalid. */
+};
+
+/* Compare two IEEE 128-bit floating point values and return the status. We
+ return the status as a 4-bit value that can be copied into an appropriate
+ PowerPC conditon code register. */
+
+CMPtype
+__cmpukf2 (TFtype a, TFtype b)
+{
+ FP_DECL_EX;
+ FP_DECL_Q (A);
+ FP_DECL_Q (B);
+ CMPtype r;
+
+ FP_INIT_EXCEPTIONS;
+ FP_UNPACK_RAW_Q (A, a);
+ FP_UNPACK_RAW_Q (B, b);
+ FP_CMP_Q (r, A, B, 2, 2);
+ if (r == CMP_INVALID)
+ FP_SET_EXCEPTION (FP_EX_INVALID);
+ FP_HANDLE_EXCEPTIONS;
+
+ return (r < CMP_LOW || r > CMP_HIGH) ? PPC_UNORDERED : ppc_cr_map[r-CMP_LOW];
+}
Index: libgcc/config/rs6000/trunctfk2.c
===================================================================
--- libgcc/config/rs6000/trunctfk2.c (revision 0)
+++ libgcc/config/rs6000/trunctfk2.c (revision 0)
@@ -0,0 +1,65 @@
+/* Software IEEE 128-bit floating-point emulation for PowerPC.
+ Copyright (C) 2014 Free Software Foundation, Inc.
+ This file is part of the GNU C Library.
+ Contributed by Michael Meissner ([email protected])
+ Code is based on the main soft-fp library written by:
+ Richard Henderson ([email protected]) and
+ Jakub Jelinek ([email protected]).
+
+ 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.
+
+ In addition to the permissions in the GNU Lesser General Public
+ License, the Free Software Foundation gives you unlimited
+ permission to link the compiled version of this file into
+ combinations with other programs, and to distribute those
+ combinations without any restriction coming from the use of this
+ file. (The Lesser General Public License restrictions do apply in
+ other respects; for example, they cover modification of the file,
+ and distribution when not linked into a combine executable.)
+
+ 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/>. */
+
+/* Convert IBM long double to IEEE 128-bit floating point. */
+
+/* Force the use of the VSX instruction set. */
+#if defined(_ARCH_PPC) && (!defined(__VSX__) || !defined(__FLOAT128__))
+#pragma GCC target ("vsx,float128")
+#endif
+
+extern __float128 __trunctfkf2 (__ibm128);
+
+#ifdef __LITTLE_ENDIAN__
+#define HIGH_WORD 1
+#define LOW_WORD 0
+#else
+#define HIGH_WORD 0
+#define LOW_WORD 1
+#endif
+
+__float128
+__trunctfkf2 (__ibm128 value)
+{
+ double high = __builtin_unpack_longdouble (value, HIGH_WORD);
+ double low = __builtin_unpack_longdouble (value, LOW_WORD);
+
+ /* Handle the special cases of NAN and inifinity. */
+ if (__builtin_isnan (high) || __builtin_isinf (high))
+ return (__float128) high;
+
+ /* If low is 0.0, there no need to do the add. In addition, avoiding the add
+ produces the correct sign if high is -0.0. */
+ if (low == 0.0)
+ return (__float128) high;
+
+ return ((__float128)high) + ((__float128)low);
+}
Index: libgcc/config/rs6000/quad-float128.h
===================================================================
--- libgcc/config/rs6000/quad-float128.h (revision 0)
+++ libgcc/config/rs6000/quad-float128.h (revision 0)
@@ -0,0 +1,72 @@
+/* Software floating-point emulation.
+ Definitions for IEEE Quad Precision on the PowerPC.
+ Copyright (C) 2015 Free Software Foundation, Inc.
+ This file is part of the GNU C Library.
+ Contributed by Michael Meissner ([email protected]).
+
+ 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.
+
+ In addition to the permissions in the GNU Lesser General Public
+ License, the Free Software Foundation gives you unlimited
+ permission to link the compiled version of this file into
+ combinations with other programs, and to distribute those
+ combinations without any restriction coming from the use of this
+ file. (The Lesser General Public License restrictions do apply in
+ other respects; for example, they cover modification of the file,
+ and distribution when not linked into a combine executable.)
+
+ 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/>. */
+
+/* quad.h defines the TFtype type by:
+ typedef float TFtype __attribute__ ((mode (TF)));
+
+ This define forces it to use KFmode (aka, ieee 128-bit floating point). */
+#define TF KF
+
+/* Force the use of the VSX instruction set. */
+#if defined(_ARCH_PPC) && (!defined(__VSX__) || !defined(__FLOAT128__))
+#pragma GCC target ("vsx,float128")
+#endif
+
+#include <quad.h>
+
+/* Add prototypes of the library functions created. In case the appropriate
+ int/long types are not declared in scope by the time quad.h is included,
+ provide our own version. */
+typedef int SItype_ppc __attribute__ ((__mode__ (__SI__)));
+typedef int DItype_ppc __attribute__ ((__mode__ (__DI__)));
+typedef unsigned USItype_ppc __attribute__ ((__mode__ (__SI__)));
+typedef unsigned UDItype_ppc __attribute__ ((__mode__ (__DI__)));
+
+TFtype __addkf3 (TFtype, TFtype);
+TFtype __subkf3 (TFtype, TFtype);
+TFtype __mulkf3 (TFtype, TFtype);
+TFtype __divkf3 (TFtype, TFtype);
+TFtype __negkf2 (TFtype);
+CMPtype __cmpukf2 (TFtype, TFtype);
+CMPtype __eqkf2 (TFtype, TFtype);
+CMPtype __gekf2 (TFtype, TFtype);
+CMPtype __lekf2 (TFtype, TFtype);
+CMPtype __unordkf2 (TFtype, TFtype);
+TFtype __extendsfkf2 (float);
+TFtype __extenddfkf2 (double);
+float __trunckfsf2 (TFtype);
+double __trunckfdf2 (TFtype);
+SItype_ppc __fixkfsi (TFtype);
+DItype_ppc __fixkfdi (TFtype);
+USItype_ppc __fixunskfsi (TFtype);
+UDItype_ppc __fixunskfdi (TFtype);
+TFtype __floatsikf (SItype_ppc);
+TFtype __floatdikf (DItype_ppc);
+TFtype __floatunsikf (USItype_ppc);
+TFtype __floatundikf (UDItype_ppc);
Index: libgcc/config/rs6000/libgcc-glibc.ver
===================================================================
--- libgcc/config/rs6000/libgcc-glibc.ver (revision 229390)
+++ libgcc/config/rs6000/libgcc-glibc.ver (working copy)
@@ -71,3 +71,81 @@ GCC_4.2.0 {
__gcc_qunord
%endif
}
+
+
+# IEEE 128-bit support (__float128) was introduced with GCC 6.0.
+
+%exclude {
+ __addkf3
+ __cmpukf2
+ __divkf3
+ __eqkf2
+ __extenddfkf2
+ __extendkftf2
+ __extendsfkf2
+ __fixkfdi
+ __fixkfsi
+ __fixkfti
+ __fixunskfdi
+ __fixunskfsi
+ __fixunskfti
+ __floattikf
+ __floatuntikf
+ __gekf2
+ __gtkf2
+ __lekf2
+ __ltkf2
+ __mulkf3
+ __negkf3
+ __nekf2
+ __orderedkf2
+ __sqrtkf3
+ __subkf3
+ __trunckfdf2
+ __trunckfsf2
+ __trunctfkf2
+ __uneqkf2
+ __ungekf2
+ __ungtkf2
+ __unlekf2
+ __unltkf2
+ __unordkf2
+}
+
+GCC_6.0.0 {
+ __addkf3
+ __cmpukf2
+ __divkf3
+ __eqkf2
+ __extenddfkf2
+ __extendkftf2
+ __extendsfkf2
+ __fixkfdi
+ __fixkfsi
+ __fixkfti
+ __fixunskfdi
+ __fixunskfsi
+ __fixunskfti
+ __floattikf
+ __floatuntikf
+ __gekf2
+ __gtkf2
+ __lekf2
+ __ltkf2
+ __mulkf3
+ __negkf3
+ __nekf2
+ __orderedkf2
+ __sqrtkf3
+ __subkf3
+ __trunckfdf2
+ __trunckfsf2
+ __trunctfkf2
+ __uneqkf2
+ __ungekf2
+ __ungtkf2
+ __unlekf2
+ __unltkf2
+ __unordkf2
+}
+
Index: libgcc/config/rs6000/t-float128
===================================================================
--- libgcc/config/rs6000/t-float128 (revision 0)
+++ libgcc/config/rs6000/t-float128 (revision 0)
@@ -0,0 +1,77 @@
+# Support for adding __float128 to the powerpc.
+
+# The standard 128-bit floating point support functions are TFmode. Most
+# PowerPC targets use a long double format that has a pair of doubles to give
+# you more precision, but no extra expoenent range. This long double format is
+# mostly compatible with the format used by the IBM XL compilers. Some of the
+# names used by the IBM double-double format use TF in them, so we rename
+# all of the functions provided for the new IEEE 128-bit support.
+#
+# We use the TF functions in soft-fp for 128-bit floating point support, using
+# sed to transform the names in the files from TF names to KF names.
+
+# Emulator functions from the soft-fp directory
+fp128_softfp_funcs = addkf3 subkf3 mulkf3 divkf3 \
+ negkf2 eqkf2 gekf2 lekf2 unordkf2 \
+ extendsfkf2 extenddfkf2 trunckfsf2 trunckfdf2 \
+ fixkfsi fixkfdi fixunskfsi fixunskfdi \
+ floatsikf floatdikf floatunsikf floatundikf
+
+fp128_softfp_src = $(addsuffix .c,$(fp128_softfp_funcs))
+fp128_softfp_static_obj = $(addsuffix $(objext),$(fp128_softfp_funcs))
+fp128_softfp_shared_obj = $(addsuffix _s$(objext),$(fp128_softfp_funcs))
+fp128_softfp_obj = $(fp128_softfp_static_obj) $(fp128_softfp_shared_obj)
+
+# New functions
+fp128_ppc_funcs = cmpukf2 extendkftf2 trunctfk2
+fp128_ppc_src = $(addprefix $(srcdir)/config/rs6000/,$(addsuffix
.c,$(fp128_ppc_funcs)))
+fp128_ppc_static_obj = $(addsuffix $(objext),$(fp128_ppc_funcs))
+fp128_ppc_shared_obj = $(addsuffix _s$(objext),$(fp128_ppc_funcs))
+fp128_ppc_obj = $(fp128_ppc_static_obj) $(fp128_ppc_shared_obj)
+
+# All functions
+fp128_funcs = $(fp128_softfp_funcs) $(fp128_ppc_funcs)
+fp128_src = $(fp128_softfp_src) $(fp128_ppc_src)
+fp128_obj = $(fp128_softfp_obj) $(fp128_ppc_obj)
+fp128_sed = $(srcdir)/config/rs6000/float128-sed
+fp128_dep = $(fp128_sed) $(srcdir)/config/rs6000/t-float128
+
+fp128_includes = $(srcdir)/soft-fp/double.h \
+ $(srcdir)/soft-fp/op-1.h \
+ $(srcdir)/soft-fp/op-4.h \
+ $(srcdir)/soft-fp/op-common.h \
+ $(srcdir)/soft-fp/single.h \
+ $(srcdir)/soft-fp/extended.h \
+ $(srcdir)/soft-fp/op-2.h \
+ $(srcdir)/soft-fp/op-8.h \
+ $(srcdir)/soft-fp/quad.h \
+ $(srcdir)/soft-fp/soft-fp.h
+
+EXTRA_CLEAN += $(fp128_softfp_src)
+
+FP128_CFLAGS = -Wno-type-limits -mvsx -mfloat128 \
+ -I$(srcdir)/soft-fp \
+ -I$(srcdir)/config/rs6000
+
+$(fp128_obj) : INTERNAL_CFLAGS += $(FP128_CFLAGS)
+$(fp128_obj) : $(fp128_includes)
+$(fp128_obj) : $(srcdir)/config/rs6000/quad-float128.h
+
+$(fp128_softfp_src) : $(srcdir)/soft-fp/$(subst kf,tf,$@) $(fp128_dep)
+ @src="$(srcdir)/soft-fp/$(subst kf,tf,$@)"; \
+ echo "Create $@"; \
+ (echo "/* file created from $$src */"; \
+ echo; \
+ sed -f $(fp128_sed) < $$src) > $@
+
+test:
+ @echo "fp128_src:"; \
+ for x in $(fp128_src); do echo " $$x"; done; \
+ echo; \
+ echo "fp128_obj:"; \
+ for x in $(fp128_obj); do echo " $$x"; done;
+
+# For now, only put it in the static library
+# LIB2ADD += $(fp128_src)
+
+LIB2ADD_ST += $(fp128_src)
Index: libgcc/config/rs6000/float128-sed
===================================================================
--- libgcc/config/rs6000/float128-sed (revision 0)
+++ libgcc/config/rs6000/float128-sed (revision 0)
@@ -0,0 +1,25 @@
+s/__addtf3/__addkf3/g
+s/__divtf3/__divkf3/g
+s/__eqtf2/__eqkf2/g
+s/__extenddftf2/__extenddfkf2/g
+s/__extendsftf2/__extendsfkf2/g
+s/__fixtfdi/__fixkfdi/g
+s/__fixtfsi/__fixkfsi/g
+s/__fixunstfdi/__fixunskfdi/g
+s/__fixunstfsi/__fixunskfsi/g
+s/__floatditf/__floatdikf/g
+s/__floatsitf/__floatsikf/g
+s/__floatunditf/__floatundikf/g
+s/__floatunsitf/__floatunsikf/g
+s/__getf2/__gekf2/g
+s/__gttf2/__gtkf2/g
+s/__letf2/__lekf2/g
+s/__lttf2/__ltkf2/g
+s/__multf3/__mulkf3/g
+s/__negtf2/__negkf2/g
+s/__netf2/__nekf2/g
+s/quad[.]h/quad-float128.h/g
+s/__subtf3/__subkf3/g
+s/__trunctfdf2/__trunckfdf2/g
+s/__trunctfsf2/__trunckfsf2/g
+s/__unordtf2/__unordkf2/g
Index: libgcc/configure
===================================================================
--- libgcc/configure (revision 229390)
+++ libgcc/configure (working copy)
@@ -4766,6 +4766,29 @@ esac
;;
esac
+case ${host} in
+powerpc*-*-linux*)
+ { $as_echo "$as_me:${as_lineno-$LINENO}: checking whether the PowerPC
compiler can do __float128" >&5
+$as_echo_n "checking whether the PowerPC compiler can do __float128... " >&6; }
+if test "${libgcc_cv_powerpc_float128+set}" = set; then :
+ $as_echo_n "(cached) " >&6
+else
+ cat confdefs.h - <<_ACEOF >conftest.$ac_ext
+/* end confdefs.h. */
+#pragma GCC target ("vsx,float128")
+ __float128 add (__float128 *a) { return *a + *(a+1); }
+_ACEOF
+if ac_fn_c_try_compile "$LINENO"; then :
+ libgcc_cv_powerpc_float128=yes
+else
+ libgcc_cv_powerpc_float128=no
+fi
+rm -f core conftest.err conftest.$ac_objext conftest.$ac_ext
+fi
+{ $as_echo "$as_me:${as_lineno-$LINENO}: result: $libgcc_cv_powerpc_float128"
>&5
+$as_echo "$libgcc_cv_powerpc_float128" >&6; }
+esac
+
# Collect host-machine-specific information.
. ${srcdir}/config.host
