This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
Patch to use soft-fp instead of fp-bit on powerpc-linux
- From: "Joseph S. Myers" <joseph at codesourcery dot com>
- To: gcc-patches at gcc dot gnu dot org
- Date: Sun, 12 Mar 2006 02:52:14 +0000 (UTC)
- Subject: Patch to use soft-fp instead of fp-bit on powerpc-linux
This patch makes powerpc*-*-linux* use soft-fp (from glibc, but under
GPL+libgcc-exception instead of LGPL, as approved by RMS) as a better
software floating-point library than fp-bit.c. For more discussion of
the issues involved and possible future projects see
<http://gcc.gnu.org/wiki/Software%20floating%20point>.
This initial patch deliberately takes a conservative approach:
* The libgcc2.c functions are not replaced with the soft-fp ones for
any target or multilib, because it would currently be difficult to
replace them for soft-float multilibs only. With toplevel libgcc it
should be easier to replace the libgcc2.c functions for soft-float
multilibs and avoid compiling in any soft-fp functions for
hard-float multilibs; at present the fp-bit functions are compiled
into hard-float multilibs but not used, and after this patch this
applies to the soft-fp functions instead.
* Only powerpc*-*-linux* targets are changed (along with powerpc64-gnu
which uses rs6000/t-linux64) instead of all PowerPC targets, so I
can have some confidence in the patches having been tested. I don't
expect other targets to be difficult to change (by people with
access to those targets to test them), the specific issues are:
+ If a target doesn't support alias attributes (as required by the
current soft-fp code), some changes will be required to use
duplicate copies of functions where aliases are used at present;
care will be needed to keep any changes to the code shared with
glibc acceptable to glibc.
+ The code in t-softfp adds to LIB2FUNCS_EXTRA rather than having
special support code in mklibgcc.in the way fp-bit does. I think
this approach is cleaner, but it does mean that other definitions
of LIB2FUNCS_EXTRA used by the target being changed need to use +=
instead of =.
+ If any target uses software floating point in 64-bit mode,
sfp-machine.h should be adjusted to allow for use in 64-bit mode.
+ Various PowerPC targets use fp-bit without using t-fprules; they
should still be able to use rs6000/t-fprules-softfp
soft-fp/t-softfp.
Much the same issues apply to conversion of non-PowerPC targets.
glibc has many sfp-machine.h files that can be used as a basis for
those in GCC (including for some targets where glibc doesn't have any
software floating point code).
The intended policy for maintaining the gcc/config/soft-fp files is
shown in the codingconventions.html patch: changes should go in glibc
first.
Tested with no regressions with cross-compilers to
powerpc-ibm-linux-gnu, --with-cpu=440. Further testing of the soft-fp
code as discussed on the wiki page: paranoia and ucbtest have been
used. Also tested on powerpc64-unknown-linux-gnu that all libgcc
multilibs build using the correct source files and with the code
properly elided for the 64-bit multilib. OK to commit to trunk?
2006-03-12 Joseph S. Myers <joseph@codesourcery.com>
* config/soft-fp: New directory. From glibc.
* config/soft-fp/t-softfp: New.
* mklibgcc.in: Handle LIB2FUNCS_EXCLUDE.
* Makefile.in (LIB2FUNCS_EXCLUDE, SFP_MACHINE): New.
(libgcc.mk): Pass LIB2FUNCS_EXCLUDE.
(LIBGCC_DEPS): Add $(SFP_MACHINE).
* config.gcc: Use rs6000/t-fprules-fpbit or
rs6000/t-fprules-softfp and soft-fp/t-softfp together with
rs6000/t-fprules.
* config/rs6000/sfp-machine.h: New. Based on glibc.
* config/rs6000/t-fprules-fpbit, config/rs6000/t-fprules-softfp:
New.
* config/rs6000/t-fprules: Remove fp-bit rules.
* config/rs6000/t-ppccomm (LIB2FUNCS_EXTRA): Use +=.
* config/rs6000/t-linux64: Likewise. Remove duplicates from
LIB2FUNCS_EXTRA. Remove fp-bit rules.
(softfp_wrap_start, softfp_wrap_end): New.
diff -rupN GCC.orig/gcc/Makefile.in GCC/gcc/Makefile.in
--- GCC.orig/gcc/Makefile.in 2006-03-09 01:20:50.000000000 +0000
+++ GCC/gcc/Makefile.in 2006-03-11 19:00:28.000000000 +0000
@@ -611,6 +611,12 @@ LIB2FUNCS_EXTRA =
# Assembler files should have names ending in `.asm'.
LIB2FUNCS_STATIC_EXTRA =
+# List of functions not to build from libgcc2.c.
+LIB2FUNCS_EXCLUDE =
+
+# Target sfp-machine.h file.
+SFP_MACHINE =
+
# Program to convert libraries.
LIBCONVERT =
@@ -1397,6 +1403,7 @@ libgcc.mk: config.status Makefile mklibg
objext='$(objext)' \
LIB1ASMFUNCS='$(LIB1ASMFUNCS)' \
LIB2FUNCS_ST='$(LIB2FUNCS_ST)' \
+ LIB2FUNCS_EXCLUDE='$(LIB2FUNCS_EXCLUDE)' \
LIBGCOV='$(LIBGCOV)' \
LIB2ADD='$(LIB2ADD)' \
LIB2ADD_ST='$(LIB2ADD_ST)' \
@@ -1448,7 +1455,7 @@ LIBGCC_DEPS = $(GCC_PASSES) $(LANGUAGES)
config/dfp-bit.h config/dfp-bit.c \
$(LIB2ADD_ST) $(LIB2ADDEH) $(LIB2ADDEHDEP) $(EXTRA_PARTS) \
$(srcdir)/config/$(LIB1ASMSRC) \
- $(srcdir)/gcov-io.h $(srcdir)/gcov-io.c gcov-iov.h
+ $(srcdir)/gcov-io.h $(srcdir)/gcov-io.c gcov-iov.h $(SFP_MACHINE)
libgcov.a: libgcc.a; @true
diff -rupN GCC.orig/gcc/config/rs6000/sfp-machine.h GCC/gcc/config/rs6000/sfp-machine.h
--- GCC.orig/gcc/config/rs6000/sfp-machine.h 1970-01-01 00:00:00.000000000 +0000
+++ GCC/gcc/config/rs6000/sfp-machine.h 2006-02-11 17:05:02.000000000 +0000
@@ -0,0 +1,63 @@
+#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_MUL_MEAT_S(R,X,Y) \
+ _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)
+#define _FP_MUL_MEAT_Q(R,X,Y) \
+ _FP_MUL_MEAT_4_wide(_FP_WFRACBITS_Q,R,X,Y,umul_ppmm)
+
+#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)
+#define _FP_DIV_MEAT_Q(R,X,Y) _FP_DIV_MEAT_4_udiv(Q,R,X,Y)
+
+#define _FP_NANFRAC_S ((_FP_QNANBIT_S << 1) - 1)
+#define _FP_NANFRAC_D ((_FP_QNANBIT_D << 1) - 1), -1
+#define _FP_NANFRAC_Q ((_FP_QNANBIT_Q << 1) - 1), -1, -1, -1
+#define _FP_NANSIGN_S 0
+#define _FP_NANSIGN_D 0
+#define _FP_NANSIGN_Q 0
+
+#define _FP_KEEPNANFRACP 1
+
+/* Someone please check this. */
+#define _FP_CHOOSENAN(fs, wc, R, X, Y, OP) \
+ do { \
+ if ((_FP_FRAC_HIGH_RAW_##fs(X) & _FP_QNANBIT_##fs) \
+ && !(_FP_FRAC_HIGH_RAW_##fs(Y) & _FP_QNANBIT_##fs)) \
+ { \
+ R##_s = Y##_s; \
+ _FP_FRAC_COPY_##wc(R,Y); \
+ } \
+ else \
+ { \
+ R##_s = X##_s; \
+ _FP_FRAC_COPY_##wc(R,X); \
+ } \
+ R##_c = FP_CLS_NAN; \
+ } while (0)
+
+#define __LITTLE_ENDIAN 1234
+#define __BIG_ENDIAN 4321
+
+#if defined __BIG_ENDIAN__ || defined _BIG_ENDIAN
+# if defined __LITTLE_ENDIAN__ || defined _LITTLE_ENDIAN
+# error "Both BIG_ENDIAN and LITTLE_ENDIAN defined!"
+# endif
+# define __BYTE_ORDER __BIG_ENDIAN
+#else
+# if defined __LITTLE_ENDIAN__ || defined _LITTLE_ENDIAN
+# define __BYTE_ORDER __LITTLE_ENDIAN
+# else
+# error "Cannot determine current byte order"
+# endif
+#endif
+
+
+/* Define ALIASNAME as a strong alias for NAME. */
+# define strong_alias(name, aliasname) _strong_alias(name, aliasname)
+# define _strong_alias(name, aliasname) \
+ extern __typeof (name) aliasname __attribute__ ((alias (#name)));
diff -rupN GCC.orig/gcc/config/rs6000/t-fprules GCC/gcc/config/rs6000/t-fprules
--- GCC.orig/gcc/config/rs6000/t-fprules 2005-11-10 22:41:15.000000000 +0000
+++ GCC/gcc/config/rs6000/t-fprules 2006-03-09 01:47:22.000000000 +0000
@@ -1,15 +1,3 @@
-# We want fine grained libraries, so use the new code to build the
-# floating point emulation libraries.
-FPBIT = fp-bit.c
-DPBIT = dp-bit.c
-
-dp-bit.c: $(srcdir)/config/fp-bit.c
- cat $(srcdir)/config/fp-bit.c > dp-bit.c
-
-fp-bit.c: $(srcdir)/config/fp-bit.c
- echo '#define FLOAT' > fp-bit.c
- cat $(srcdir)/config/fp-bit.c >> fp-bit.c
-
MULTILIB_MATCHES_FLOAT = msoft-float=mcpu?401 \
msoft-float=mcpu?403 \
msoft-float=mcpu?405 \
diff -rupN GCC.orig/gcc/config/rs6000/t-fprules-fpbit GCC/gcc/config/rs6000/t-fprules-fpbit
--- GCC.orig/gcc/config/rs6000/t-fprules-fpbit 1970-01-01 00:00:00.000000000 +0000
+++ GCC/gcc/config/rs6000/t-fprules-fpbit 2006-03-09 01:53:47.000000000 +0000
@@ -0,0 +1,11 @@
+# We want fine grained libraries, so use the new code to build the
+# floating point emulation libraries.
+FPBIT = fp-bit.c
+DPBIT = dp-bit.c
+
+dp-bit.c: $(srcdir)/config/fp-bit.c
+ cat $(srcdir)/config/fp-bit.c > dp-bit.c
+
+fp-bit.c: $(srcdir)/config/fp-bit.c
+ echo '#define FLOAT' > fp-bit.c
+ cat $(srcdir)/config/fp-bit.c >> fp-bit.c
diff -rupN GCC.orig/gcc/config/rs6000/t-fprules-softfp GCC/gcc/config/rs6000/t-fprules-softfp
--- GCC.orig/gcc/config/rs6000/t-fprules-softfp 1970-01-01 00:00:00.000000000 +0000
+++ GCC/gcc/config/rs6000/t-fprules-softfp 2006-03-11 18:35:01.000000000 +0000
@@ -0,0 +1,6 @@
+softfp_float_modes := sf df
+softfp_int_modes := si di
+softfp_extensions := sfdf
+softfp_truncations := dfsf
+softfp_machine_header := rs6000/sfp-machine.h
+softfp_exclude_libgcc2 := y
diff -rupN GCC.orig/gcc/config/rs6000/t-linux64 GCC/gcc/config/rs6000/t-linux64
--- GCC.orig/gcc/config/rs6000/t-linux64 2006-01-28 17:11:32.000000000 +0000
+++ GCC/gcc/config/rs6000/t-linux64 2006-03-11 21:55:46.000000000 +0000
@@ -1,8 +1,9 @@
#rs6000/t-linux64
-LIB2FUNCS_EXTRA = tramp.S $(srcdir)/config/rs6000/ppc64-fp.c \
+LIB2FUNCS_EXTRA += tramp.S $(srcdir)/config/rs6000/ppc64-fp.c \
$(srcdir)/config/rs6000/darwin-ldouble.c
+LIB2FUNCS_EXTRA := $(sort $(LIB2FUNCS_EXTRA))
TARGET_LIBGCC2_CFLAGS += -mno-minimal-toc
@@ -14,19 +15,5 @@ MULTILIB_EXCLUSIONS = m64/!m32/msoft
MULTILIB_OSDIRNAMES = ../lib64 ../lib nof
MULTILIB_MATCHES = $(MULTILIB_MATCHES_FLOAT)
-# We want fine grained libraries, so use the new code to build the
-# floating point emulation libraries.
-# fp-bit is only to be used by 32-bit multilibs
-FPBIT = fp-bit32.c
-DPBIT = dp-bit32.c
-
-dp-bit32.c: $(srcdir)/config/fp-bit.c
- ( echo '#ifndef __powerpc64__'; \
- cat $(srcdir)/config/fp-bit.c; \
- echo '#endif' ) > dp-bit32.c
-
-fp-bit32.c: $(srcdir)/config/fp-bit.c
- ( echo '#ifndef __powerpc64__'; \
- echo '#define FLOAT'; \
- cat $(srcdir)/config/fp-bit.c; \
- echo '#endif' ) > fp-bit32.c
+softfp_wrap_start := '\#ifndef __powerpc64__'
+softfp_wrap_end := '\#endif'
diff -rupN GCC.orig/gcc/config/rs6000/t-ppccomm GCC/gcc/config/rs6000/t-ppccomm
--- GCC.orig/gcc/config/rs6000/t-ppccomm 2006-02-08 18:02:29.000000000 +0000
+++ GCC/gcc/config/rs6000/t-ppccomm 2006-03-11 18:15:05.000000000 +0000
@@ -1,6 +1,6 @@
# Common support for PowerPC ELF targets (both EABI and SVR4).
-LIB2FUNCS_EXTRA = tramp.S $(srcdir)/config/rs6000/darwin-ldouble.c
+LIB2FUNCS_EXTRA += tramp.S $(srcdir)/config/rs6000/darwin-ldouble.c
# This one can't end up in shared libgcc
LIB2FUNCS_STATIC_EXTRA = eabi.S
diff -rupN GCC.orig/gcc/config/soft-fp/adddf3.c GCC/gcc/config/soft-fp/adddf3.c
--- GCC.orig/gcc/config/soft-fp/adddf3.c 1970-01-01 00:00:00.000000000 +0000
+++ GCC/gcc/config/soft-fp/adddf3.c 2006-02-18 02:11:04.000000000 +0000
@@ -0,0 +1,49 @@
+/* Software floating-point emulation.
+ Return a + b
+ Copyright (C) 1997,1999, 2006 Free Software Foundation, Inc.
+ This file is part of GCC.
+ Contributed by Richard Henderson (rth@cygnus.com) and
+ Jakub Jelinek (jj@ultra.linux.cz).
+
+ GCC is free software; you can redistribute it and/or modify it under
+ the terms of the GNU General Public License as published by the Free
+ Software Foundation; either version 2, or (at your option) any later
+ version.
+
+ In addition to the permissions in the GNU 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 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.)
+
+ GCC 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 General Public License
+ for more details.
+
+ You should have received a copy of the GNU General Public License
+ along with GCC; see the file COPYING. If not, write to the Free
+ Software Foundation, 51 Franklin Street, Fifth Floor, Boston, MA
+ 02110-1301, USA. */
+
+#include "soft-fp.h"
+#include "double.h"
+
+DFtype __adddf3(DFtype a, DFtype b)
+{
+ FP_DECL_EX;
+ FP_DECL_D(A); FP_DECL_D(B); FP_DECL_D(R);
+ DFtype r;
+
+ FP_INIT_ROUNDMODE;
+ FP_UNPACK_SEMIRAW_D(A, a);
+ FP_UNPACK_SEMIRAW_D(B, b);
+ FP_ADD_D(R, A, B);
+ FP_PACK_SEMIRAW_D(r, R);
+ FP_HANDLE_EXCEPTIONS;
+
+ return r;
+}
diff -rupN GCC.orig/gcc/config/soft-fp/addsf3.c GCC/gcc/config/soft-fp/addsf3.c
--- GCC.orig/gcc/config/soft-fp/addsf3.c 1970-01-01 00:00:00.000000000 +0000
+++ GCC/gcc/config/soft-fp/addsf3.c 2006-02-18 02:11:04.000000000 +0000
@@ -0,0 +1,50 @@
+/* Software floating-point emulation.
+ Return a + b
+ Copyright (C) 1997,1999,2006 Free Software Foundation, Inc.
+ This file is part of GCC.
+ Contributed by Richard Henderson (rth@cygnus.com) and
+ Jakub Jelinek (jj@ultra.linux.cz).
+
+ GCC is free software; you can redistribute it and/or modify it under
+ the terms of the GNU General Public License as published by the Free
+ Software Foundation; either version 2, or (at your option) any later
+ version.
+
+ In addition to the permissions in the GNU 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 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.)
+
+ GCC 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 General Public License
+ for more details.
+
+ You should have received a copy of the GNU General Public License
+ along with GCC; see the file COPYING. If not, write to the Free
+ Software Foundation, 51 Franklin Street, Fifth Floor, Boston, MA
+ 02110-1301, USA. */
+
+#include "soft-fp.h"
+#include "single.h"
+
+SFtype __addsf3(SFtype a, SFtype b)
+{
+ FP_DECL_EX;
+ FP_DECL_S(A); FP_DECL_S(B); FP_DECL_S(R);
+ SFtype r;
+
+ FP_INIT_ROUNDMODE;
+ FP_UNPACK_SEMIRAW_S(A, a);
+ FP_UNPACK_SEMIRAW_S(B, b);
+ FP_ADD_S(R, A, B);
+ FP_PACK_SEMIRAW_S(r, R);
+ FP_HANDLE_EXCEPTIONS;
+
+ return r;
+}
+
diff -rupN GCC.orig/gcc/config/soft-fp/addtf3.c GCC/gcc/config/soft-fp/addtf3.c
--- GCC.orig/gcc/config/soft-fp/addtf3.c 1970-01-01 00:00:00.000000000 +0000
+++ GCC/gcc/config/soft-fp/addtf3.c 2006-02-18 02:11:04.000000000 +0000
@@ -0,0 +1,49 @@
+/* Software floating-point emulation.
+ Return a + b
+ Copyright (C) 1997,1999,2006 Free Software Foundation, Inc.
+ This file is part of GCC.
+ Contributed by Richard Henderson (rth@cygnus.com) and
+ Jakub Jelinek (jj@ultra.linux.cz).
+
+ GCC is free software; you can redistribute it and/or modify it under
+ the terms of the GNU General Public License as published by the Free
+ Software Foundation; either version 2, or (at your option) any later
+ version.
+
+ In addition to the permissions in the GNU 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 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.)
+
+ GCC 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 General Public License
+ for more details.
+
+ You should have received a copy of the GNU General Public License
+ along with GCC; see the file COPYING. If not, write to the Free
+ Software Foundation, 51 Franklin Street, Fifth Floor, Boston, MA
+ 02110-1301, USA. */
+
+#include "soft-fp.h"
+#include "quad.h"
+
+TFtype __addtf3(TFtype a, TFtype b)
+{
+ FP_DECL_EX;
+ FP_DECL_Q(A); FP_DECL_Q(B); FP_DECL_Q(R);
+ TFtype r;
+
+ FP_INIT_ROUNDMODE;
+ FP_UNPACK_SEMIRAW_Q(A, a);
+ FP_UNPACK_SEMIRAW_Q(B, b);
+ FP_ADD_Q(R, A, B);
+ FP_PACK_SEMIRAW_Q(r, R);
+ FP_HANDLE_EXCEPTIONS;
+
+ return r;
+}
diff -rupN GCC.orig/gcc/config/soft-fp/divdf3.c GCC/gcc/config/soft-fp/divdf3.c
--- GCC.orig/gcc/config/soft-fp/divdf3.c 1970-01-01 00:00:00.000000000 +0000
+++ GCC/gcc/config/soft-fp/divdf3.c 2006-02-18 02:11:04.000000000 +0000
@@ -0,0 +1,49 @@
+/* Software floating-point emulation.
+ Return a / b
+ Copyright (C) 1997,1999,2006 Free Software Foundation, Inc.
+ This file is part of GCC.
+ Contributed by Richard Henderson (rth@cygnus.com) and
+ Jakub Jelinek (jj@ultra.linux.cz).
+
+ GCC is free software; you can redistribute it and/or modify it under
+ the terms of the GNU General Public License as published by the Free
+ Software Foundation; either version 2, or (at your option) any later
+ version.
+
+ In addition to the permissions in the GNU 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 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.)
+
+ GCC 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 General Public License
+ for more details.
+
+ You should have received a copy of the GNU General Public License
+ along with GCC; see the file COPYING. If not, write to the Free
+ Software Foundation, 51 Franklin Street, Fifth Floor, Boston, MA
+ 02110-1301, USA. */
+
+#include "soft-fp.h"
+#include "double.h"
+
+DFtype __divdf3(DFtype a, DFtype b)
+{
+ FP_DECL_EX;
+ FP_DECL_D(A); FP_DECL_D(B); FP_DECL_D(R);
+ DFtype r;
+
+ FP_INIT_ROUNDMODE;
+ FP_UNPACK_D(A, a);
+ FP_UNPACK_D(B, b);
+ FP_DIV_D(R, A, B);
+ FP_PACK_D(r, R);
+ FP_HANDLE_EXCEPTIONS;
+
+ return r;
+}
diff -rupN GCC.orig/gcc/config/soft-fp/divsf3.c GCC/gcc/config/soft-fp/divsf3.c
--- GCC.orig/gcc/config/soft-fp/divsf3.c 1970-01-01 00:00:00.000000000 +0000
+++ GCC/gcc/config/soft-fp/divsf3.c 2006-02-18 02:11:04.000000000 +0000
@@ -0,0 +1,49 @@
+/* Software floating-point emulation.
+ Return a / b
+ Copyright (C) 1997,1999,2006 Free Software Foundation, Inc.
+ This file is part of GCC.
+ Contributed by Richard Henderson (rth@cygnus.com) and
+ Jakub Jelinek (jj@ultra.linux.cz).
+
+ GCC is free software; you can redistribute it and/or modify it under
+ the terms of the GNU General Public License as published by the Free
+ Software Foundation; either version 2, or (at your option) any later
+ version.
+
+ In addition to the permissions in the GNU 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 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.)
+
+ GCC 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 General Public License
+ for more details.
+
+ You should have received a copy of the GNU General Public License
+ along with GCC; see the file COPYING. If not, write to the Free
+ Software Foundation, 51 Franklin Street, Fifth Floor, Boston, MA
+ 02110-1301, USA. */
+
+#include "soft-fp.h"
+#include "single.h"
+
+SFtype __divsf3(SFtype a, SFtype b)
+{
+ FP_DECL_EX;
+ FP_DECL_S(A); FP_DECL_S(B); FP_DECL_S(R);
+ SFtype r;
+
+ FP_INIT_ROUNDMODE;
+ FP_UNPACK_S(A, a);
+ FP_UNPACK_S(B, b);
+ FP_DIV_S(R, A, B);
+ FP_PACK_S(r, R);
+ FP_HANDLE_EXCEPTIONS;
+
+ return r;
+}
diff -rupN GCC.orig/gcc/config/soft-fp/divtf3.c GCC/gcc/config/soft-fp/divtf3.c
--- GCC.orig/gcc/config/soft-fp/divtf3.c 1970-01-01 00:00:00.000000000 +0000
+++ GCC/gcc/config/soft-fp/divtf3.c 2006-02-18 02:11:04.000000000 +0000
@@ -0,0 +1,49 @@
+/* Software floating-point emulation.
+ Return a / b
+ Copyright (C) 1997,1999,2006 Free Software Foundation, Inc.
+ This file is part of GCC.
+ Contributed by Richard Henderson (rth@cygnus.com) and
+ Jakub Jelinek (jj@ultra.linux.cz).
+
+ GCC is free software; you can redistribute it and/or modify it under
+ the terms of the GNU General Public License as published by the Free
+ Software Foundation; either version 2, or (at your option) any later
+ version.
+
+ In addition to the permissions in the GNU 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 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.)
+
+ GCC 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 General Public License
+ for more details.
+
+ You should have received a copy of the GNU General Public License
+ along with GCC; see the file COPYING. If not, write to the Free
+ Software Foundation, 51 Franklin Street, Fifth Floor, Boston, MA
+ 02110-1301, USA. */
+
+#include "soft-fp.h"
+#include "quad.h"
+
+TFtype __divtf3(TFtype a, TFtype b)
+{
+ FP_DECL_EX;
+ FP_DECL_Q(A); FP_DECL_Q(B); FP_DECL_Q(R);
+ TFtype r;
+
+ FP_INIT_ROUNDMODE;
+ FP_UNPACK_Q(A, a);
+ FP_UNPACK_Q(B, b);
+ FP_DIV_Q(R, A, B);
+ FP_PACK_Q(r, R);
+ FP_HANDLE_EXCEPTIONS;
+
+ return r;
+}
diff -rupN GCC.orig/gcc/config/soft-fp/double.h GCC/gcc/config/soft-fp/double.h
--- GCC.orig/gcc/config/soft-fp/double.h 1970-01-01 00:00:00.000000000 +0000
+++ GCC/gcc/config/soft-fp/double.h 2006-02-18 02:11:04.000000000 +0000
@@ -0,0 +1,264 @@
+/* Software floating-point emulation.
+ Definitions for IEEE Double Precision
+ Copyright (C) 1997,1998,1999,2006 Free Software Foundation, Inc.
+ This file is part of GCC.
+ Contributed by Richard Henderson (rth@cygnus.com),
+ Jakub Jelinek (jj@ultra.linux.cz),
+ David S. Miller (davem@redhat.com) and
+ Peter Maydell (pmaydell@chiark.greenend.org.uk).
+
+ GCC is free software; you can redistribute it and/or modify it under
+ the terms of the GNU General Public License as published by the Free
+ Software Foundation; either version 2, or (at your option) any later
+ version.
+
+ In addition to the permissions in the GNU 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 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.)
+
+ GCC 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 General Public License
+ for more details.
+
+ You should have received a copy of the GNU General Public License
+ along with GCC; see the file COPYING. If not, write to the Free
+ Software Foundation, 51 Franklin Street, Fifth Floor, Boston, MA
+ 02110-1301, USA. */
+
+#if _FP_W_TYPE_SIZE < 32
+#error "Here's a nickel kid. Go buy yourself a real computer."
+#endif
+
+#if _FP_W_TYPE_SIZE < 64
+#define _FP_FRACTBITS_D (2 * _FP_W_TYPE_SIZE)
+#else
+#define _FP_FRACTBITS_D _FP_W_TYPE_SIZE
+#endif
+
+#define _FP_FRACBITS_D 53
+#define _FP_FRACXBITS_D (_FP_FRACTBITS_D - _FP_FRACBITS_D)
+#define _FP_WFRACBITS_D (_FP_WORKBITS + _FP_FRACBITS_D)
+#define _FP_WFRACXBITS_D (_FP_FRACTBITS_D - _FP_WFRACBITS_D)
+#define _FP_EXPBITS_D 11
+#define _FP_EXPBIAS_D 1023
+#define _FP_EXPMAX_D 2047
+
+#define _FP_QNANBIT_D \
+ ((_FP_W_TYPE)1 << (_FP_FRACBITS_D-2) % _FP_W_TYPE_SIZE)
+#define _FP_QNANBIT_SH_D \
+ ((_FP_W_TYPE)1 << (_FP_FRACBITS_D-2+_FP_WORKBITS) % _FP_W_TYPE_SIZE)
+#define _FP_IMPLBIT_D \
+ ((_FP_W_TYPE)1 << (_FP_FRACBITS_D-1) % _FP_W_TYPE_SIZE)
+#define _FP_IMPLBIT_SH_D \
+ ((_FP_W_TYPE)1 << (_FP_FRACBITS_D-1+_FP_WORKBITS) % _FP_W_TYPE_SIZE)
+#define _FP_OVERFLOW_D \
+ ((_FP_W_TYPE)1 << _FP_WFRACBITS_D % _FP_W_TYPE_SIZE)
+
+typedef float DFtype __attribute__((mode(DF)));
+
+#if _FP_W_TYPE_SIZE < 64
+
+union _FP_UNION_D
+{
+ DFtype flt;
+ struct {
+#if __BYTE_ORDER == __BIG_ENDIAN
+ unsigned sign : 1;
+ unsigned exp : _FP_EXPBITS_D;
+ unsigned frac1 : _FP_FRACBITS_D - (_FP_IMPLBIT_D != 0) - _FP_W_TYPE_SIZE;
+ unsigned frac0 : _FP_W_TYPE_SIZE;
+#else
+ unsigned frac0 : _FP_W_TYPE_SIZE;
+ unsigned frac1 : _FP_FRACBITS_D - (_FP_IMPLBIT_D != 0) - _FP_W_TYPE_SIZE;
+ unsigned exp : _FP_EXPBITS_D;
+ unsigned sign : 1;
+#endif
+ } bits __attribute__((packed));
+};
+
+#define FP_DECL_D(X) _FP_DECL(2,X)
+#define FP_UNPACK_RAW_D(X,val) _FP_UNPACK_RAW_2(D,X,val)
+#define FP_UNPACK_RAW_DP(X,val) _FP_UNPACK_RAW_2_P(D,X,val)
+#define FP_PACK_RAW_D(val,X) _FP_PACK_RAW_2(D,val,X)
+#define FP_PACK_RAW_DP(val,X) \
+ do { \
+ if (!FP_INHIBIT_RESULTS) \
+ _FP_PACK_RAW_2_P(D,val,X); \
+ } while (0)
+
+#define FP_UNPACK_D(X,val) \
+ do { \
+ _FP_UNPACK_RAW_2(D,X,val); \
+ _FP_UNPACK_CANONICAL(D,2,X); \
+ } while (0)
+
+#define FP_UNPACK_DP(X,val) \
+ do { \
+ _FP_UNPACK_RAW_2_P(D,X,val); \
+ _FP_UNPACK_CANONICAL(D,2,X); \
+ } while (0)
+
+#define FP_UNPACK_SEMIRAW_D(X,val) \
+ do { \
+ _FP_UNPACK_RAW_2(D,X,val); \
+ _FP_UNPACK_SEMIRAW(D,2,X); \
+ } while (0)
+
+#define FP_UNPACK_SEMIRAW_DP(X,val) \
+ do { \
+ _FP_UNPACK_RAW_2_P(D,X,val); \
+ _FP_UNPACK_SEMIRAW(D,2,X); \
+ } while (0)
+
+#define FP_PACK_D(val,X) \
+ do { \
+ _FP_PACK_CANONICAL(D,2,X); \
+ _FP_PACK_RAW_2(D,val,X); \
+ } while (0)
+
+#define FP_PACK_DP(val,X) \
+ do { \
+ _FP_PACK_CANONICAL(D,2,X); \
+ if (!FP_INHIBIT_RESULTS) \
+ _FP_PACK_RAW_2_P(D,val,X); \
+ } while (0)
+
+#define FP_PACK_SEMIRAW_D(val,X) \
+ do { \
+ _FP_PACK_SEMIRAW(D,2,X); \
+ _FP_PACK_RAW_2(D,val,X); \
+ } while (0)
+
+#define FP_PACK_SEMIRAW_DP(val,X) \
+ do { \
+ _FP_PACK_SEMIRAW(D,2,X); \
+ if (!FP_INHIBIT_RESULTS) \
+ _FP_PACK_RAW_2_P(D,val,X); \
+ } while (0)
+
+#define FP_ISSIGNAN_D(X) _FP_ISSIGNAN(D,2,X)
+#define FP_NEG_D(R,X) _FP_NEG(D,2,R,X)
+#define FP_ADD_D(R,X,Y) _FP_ADD(D,2,R,X,Y)
+#define FP_SUB_D(R,X,Y) _FP_SUB(D,2,R,X,Y)
+#define FP_MUL_D(R,X,Y) _FP_MUL(D,2,R,X,Y)
+#define FP_DIV_D(R,X,Y) _FP_DIV(D,2,R,X,Y)
+#define FP_SQRT_D(R,X) _FP_SQRT(D,2,R,X)
+#define _FP_SQRT_MEAT_D(R,S,T,X,Q) _FP_SQRT_MEAT_2(R,S,T,X,Q)
+
+#define FP_CMP_D(r,X,Y,un) _FP_CMP(D,2,r,X,Y,un)
+#define FP_CMP_EQ_D(r,X,Y) _FP_CMP_EQ(D,2,r,X,Y)
+#define FP_CMP_UNORD_D(r,X,Y) _FP_CMP_UNORD(D,2,r,X,Y)
+
+#define FP_TO_INT_D(r,X,rsz,rsg) _FP_TO_INT(D,2,r,X,rsz,rsg)
+#define FP_FROM_INT_D(X,r,rs,rt) _FP_FROM_INT(D,2,X,r,rs,rt)
+
+#define _FP_FRAC_HIGH_D(X) _FP_FRAC_HIGH_2(X)
+#define _FP_FRAC_HIGH_RAW_D(X) _FP_FRAC_HIGH_2(X)
+
+#else
+
+union _FP_UNION_D
+{
+ DFtype flt;
+ struct {
+#if __BYTE_ORDER == __BIG_ENDIAN
+ unsigned sign : 1;
+ unsigned exp : _FP_EXPBITS_D;
+ unsigned long frac : _FP_FRACBITS_D - (_FP_IMPLBIT_D != 0);
+#else
+ unsigned long frac : _FP_FRACBITS_D - (_FP_IMPLBIT_D != 0);
+ unsigned exp : _FP_EXPBITS_D;
+ unsigned sign : 1;
+#endif
+ } bits __attribute__((packed));
+};
+
+#define FP_DECL_D(X) _FP_DECL(1,X)
+#define FP_UNPACK_RAW_D(X,val) _FP_UNPACK_RAW_1(D,X,val)
+#define FP_UNPACK_RAW_DP(X,val) _FP_UNPACK_RAW_1_P(D,X,val)
+#define FP_PACK_RAW_D(val,X) _FP_PACK_RAW_1(D,val,X)
+#define FP_PACK_RAW_DP(val,X) \
+ do { \
+ if (!FP_INHIBIT_RESULTS) \
+ _FP_PACK_RAW_1_P(D,val,X); \
+ } while (0)
+
+#define FP_UNPACK_D(X,val) \
+ do { \
+ _FP_UNPACK_RAW_1(D,X,val); \
+ _FP_UNPACK_CANONICAL(D,1,X); \
+ } while (0)
+
+#define FP_UNPACK_DP(X,val) \
+ do { \
+ _FP_UNPACK_RAW_1_P(D,X,val); \
+ _FP_UNPACK_CANONICAL(D,1,X); \
+ } while (0)
+
+#define FP_UNPACK_SEMIRAW_D(X,val) \
+ do { \
+ _FP_UNPACK_RAW_2(1,X,val); \
+ _FP_UNPACK_SEMIRAW(D,1,X); \
+ } while (0)
+
+#define FP_UNPACK_SEMIRAW_DP(X,val) \
+ do { \
+ _FP_UNPACK_RAW_2_P(1,X,val); \
+ _FP_UNPACK_SEMIRAW(D,1,X); \
+ } while (0)
+
+#define FP_PACK_D(val,X) \
+ do { \
+ _FP_PACK_CANONICAL(D,1,X); \
+ _FP_PACK_RAW_1(D,val,X); \
+ } while (0)
+
+#define FP_PACK_DP(val,X) \
+ do { \
+ _FP_PACK_CANONICAL(D,1,X); \
+ if (!FP_INHIBIT_RESULTS) \
+ _FP_PACK_RAW_1_P(D,val,X); \
+ } while (0)
+
+#define FP_PACK_SEMIRAW_D(val,X) \
+ do { \
+ _FP_PACK_SEMIRAW(D,1,X); \
+ _FP_PACK_RAW_1(D,val,X); \
+ } while (0)
+
+#define FP_PACK_SEMIRAW_DP(val,X) \
+ do { \
+ _FP_PACK_SEMIRAW(D,1,X); \
+ if (!FP_INHIBIT_RESULTS) \
+ _FP_PACK_RAW_1_P(D,val,X); \
+ } while (0)
+
+#define FP_ISSIGNAN_D(X) _FP_ISSIGNAN(D,1,X)
+#define FP_NEG_D(R,X) _FP_NEG(D,1,R,X)
+#define FP_ADD_D(R,X,Y) _FP_ADD(D,1,R,X,Y)
+#define FP_SUB_D(R,X,Y) _FP_SUB(D,1,R,X,Y)
+#define FP_MUL_D(R,X,Y) _FP_MUL(D,1,R,X,Y)
+#define FP_DIV_D(R,X,Y) _FP_DIV(D,1,R,X,Y)
+#define FP_SQRT_D(R,X) _FP_SQRT(D,1,R,X)
+#define _FP_SQRT_MEAT_D(R,S,T,X,Q) _FP_SQRT_MEAT_1(R,S,T,X,Q)
+
+/* The implementation of _FP_MUL_D and _FP_DIV_D should be chosen by
+ the target machine. */
+
+#define FP_CMP_D(r,X,Y,un) _FP_CMP(D,1,r,X,Y,un)
+#define FP_CMP_EQ_D(r,X,Y) _FP_CMP_EQ(D,1,r,X,Y)
+#define FP_CMP_UNORD_D(r,X,Y) _FP_CMP_UNORD(D,1,r,X,Y)
+
+#define FP_TO_INT_D(r,X,rsz,rsg) _FP_TO_INT(D,1,r,X,rsz,rsg)
+#define FP_FROM_INT_D(X,r,rs,rt) _FP_FROM_INT(D,1,X,r,rs,rt)
+
+#define _FP_FRAC_HIGH_D(X) _FP_FRAC_HIGH_1(X)
+#define _FP_FRAC_HIGH_RAW_D(X) _FP_FRAC_HIGH_1(X)
+
+#endif /* W_TYPE_SIZE < 64 */
diff -rupN GCC.orig/gcc/config/soft-fp/eqdf2.c GCC/gcc/config/soft-fp/eqdf2.c
--- GCC.orig/gcc/config/soft-fp/eqdf2.c 1970-01-01 00:00:00.000000000 +0000
+++ GCC/gcc/config/soft-fp/eqdf2.c 2006-02-18 02:11:04.000000000 +0000
@@ -0,0 +1,51 @@
+/* Software floating-point emulation.
+ Return 0 iff a == b, 1 otherwise
+ Copyright (C) 1997,1999,2006 Free Software Foundation, Inc.
+ This file is part of GCC.
+ Contributed by Richard Henderson (rth@cygnus.com) and
+ Jakub Jelinek (jj@ultra.linux.cz).
+
+ GCC is free software; you can redistribute it and/or modify it under
+ the terms of the GNU General Public License as published by the Free
+ Software Foundation; either version 2, or (at your option) any later
+ version.
+
+ In addition to the permissions in the GNU 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 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.)
+
+ GCC 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 General Public License
+ for more details.
+
+ You should have received a copy of the GNU General Public License
+ along with GCC; see the file COPYING. If not, write to the Free
+ Software Foundation, 51 Franklin Street, Fifth Floor, Boston, MA
+ 02110-1301, USA. */
+
+#include "soft-fp.h"
+#include "double.h"
+
+int __eqdf2(DFtype a, DFtype b)
+{
+ FP_DECL_EX;
+ FP_DECL_D(A); FP_DECL_D(B);
+ int r;
+
+ FP_UNPACK_RAW_D(A, a);
+ FP_UNPACK_RAW_D(B, b);
+ FP_CMP_EQ_D(r, A, B);
+ if (r && (FP_ISSIGNAN_D(A) || FP_ISSIGNAN_D(B)))
+ FP_SET_EXCEPTION(FP_EX_INVALID);
+ FP_HANDLE_EXCEPTIONS;
+
+ return r;
+}
+
+strong_alias(__eqdf2, __nedf2);
diff -rupN GCC.orig/gcc/config/soft-fp/eqsf2.c GCC/gcc/config/soft-fp/eqsf2.c
--- GCC.orig/gcc/config/soft-fp/eqsf2.c 1970-01-01 00:00:00.000000000 +0000
+++ GCC/gcc/config/soft-fp/eqsf2.c 2006-02-18 02:11:04.000000000 +0000
@@ -0,0 +1,51 @@
+/* Software floating-point emulation.
+ Return 0 iff a == b, 1 otherwise
+ Copyright (C) 1997,1999,2006 Free Software Foundation, Inc.
+ This file is part of GCC.
+ Contributed by Richard Henderson (rth@cygnus.com) and
+ Jakub Jelinek (jj@ultra.linux.cz).
+
+ GCC is free software; you can redistribute it and/or modify it under
+ the terms of the GNU General Public License as published by the Free
+ Software Foundation; either version 2, or (at your option) any later
+ version.
+
+ In addition to the permissions in the GNU 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 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.)
+
+ GCC 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 General Public License
+ for more details.
+
+ You should have received a copy of the GNU General Public License
+ along with GCC; see the file COPYING. If not, write to the Free
+ Software Foundation, 51 Franklin Street, Fifth Floor, Boston, MA
+ 02110-1301, USA. */
+
+#include "soft-fp.h"
+#include "single.h"
+
+int __eqsf2(SFtype a, SFtype b)
+{
+ FP_DECL_EX;
+ FP_DECL_S(A); FP_DECL_S(B);
+ int r;
+
+ FP_UNPACK_RAW_S(A, a);
+ FP_UNPACK_RAW_S(B, b);
+ FP_CMP_EQ_S(r, A, B);
+ if (r && (FP_ISSIGNAN_S(A) || FP_ISSIGNAN_S(B)))
+ FP_SET_EXCEPTION(FP_EX_INVALID);
+ FP_HANDLE_EXCEPTIONS;
+
+ return r;
+}
+
+strong_alias(__eqsf2, __nesf2);
diff -rupN GCC.orig/gcc/config/soft-fp/eqtf2.c GCC/gcc/config/soft-fp/eqtf2.c
--- GCC.orig/gcc/config/soft-fp/eqtf2.c 1970-01-01 00:00:00.000000000 +0000
+++ GCC/gcc/config/soft-fp/eqtf2.c 2006-02-18 02:11:04.000000000 +0000
@@ -0,0 +1,51 @@
+/* Software floating-point emulation.
+ Return 0 iff a == b, 1 otherwise
+ Copyright (C) 1997,1999,2006 Free Software Foundation, Inc.
+ This file is part of GCC.
+ Contributed by Richard Henderson (rth@cygnus.com) and
+ Jakub Jelinek (jj@ultra.linux.cz).
+
+ GCC is free software; you can redistribute it and/or modify it under
+ the terms of the GNU General Public License as published by the Free
+ Software Foundation; either version 2, or (at your option) any later
+ version.
+
+ In addition to the permissions in the GNU 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 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.)
+
+ GCC 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 General Public License
+ for more details.
+
+ You should have received a copy of the GNU General Public License
+ along with GCC; see the file COPYING. If not, write to the Free
+ Software Foundation, 51 Franklin Street, Fifth Floor, Boston, MA
+ 02110-1301, USA. */
+
+#include "soft-fp.h"
+#include "quad.h"
+
+int __eqtf2(TFtype a, TFtype b)
+{
+ FP_DECL_EX;
+ FP_DECL_Q(A); FP_DECL_Q(B);
+ int r;
+
+ FP_UNPACK_RAW_Q(A, a);
+ FP_UNPACK_RAW_Q(B, b);
+ FP_CMP_EQ_Q(r, A, B);
+ if (r && (FP_ISSIGNAN_Q(A) || FP_ISSIGNAN_Q(B)))
+ FP_SET_EXCEPTION(FP_EX_INVALID);
+ FP_HANDLE_EXCEPTIONS;
+
+ return r;
+}
+
+strong_alias(__eqtf2, __netf2);
diff -rupN GCC.orig/gcc/config/soft-fp/extenddftf2.c GCC/gcc/config/soft-fp/extenddftf2.c
--- GCC.orig/gcc/config/soft-fp/extenddftf2.c 1970-01-01 00:00:00.000000000 +0000
+++ GCC/gcc/config/soft-fp/extenddftf2.c 2006-02-18 02:11:04.000000000 +0000
@@ -0,0 +1,54 @@
+/* Software floating-point emulation.
+ Return a converted to IEEE quad
+ Copyright (C) 1997,1999,2006 Free Software Foundation, Inc.
+ This file is part of GCC.
+ Contributed by Richard Henderson (rth@cygnus.com) and
+ Jakub Jelinek (jj@ultra.linux.cz).
+
+ GCC is free software; you can redistribute it and/or modify it under
+ the terms of the GNU General Public License as published by the Free
+ Software Foundation; either version 2, or (at your option) any later
+ version.
+
+ In addition to the permissions in the GNU 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 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.)
+
+ GCC 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 General Public License
+ for more details.
+
+ You should have received a copy of the GNU General Public License
+ along with GCC; see the file COPYING. If not, write to the Free
+ Software Foundation, 51 Franklin Street, Fifth Floor, Boston, MA
+ 02110-1301, USA. */
+
+#include "soft-fp.h"
+#include "double.h"
+#include "quad.h"
+
+TFtype __extenddftf2(DFtype a)
+{
+ FP_DECL_EX;
+ FP_DECL_D(A);
+ FP_DECL_Q(R);
+ TFtype r;
+
+ FP_INIT_ROUNDMODE;
+ FP_UNPACK_RAW_D(A, a);
+#if (2 * _FP_W_TYPE_SIZE) < _FP_FRACBITS_Q
+ FP_EXTEND(Q,D,4,2,R,A);
+#else
+ FP_EXTEND(Q,D,2,1,R,A);
+#endif
+ FP_PACK_RAW_Q(r, R);
+ FP_HANDLE_EXCEPTIONS;
+
+ return r;
+}
diff -rupN GCC.orig/gcc/config/soft-fp/extended.h GCC/gcc/config/soft-fp/extended.h
--- GCC.orig/gcc/config/soft-fp/extended.h 1970-01-01 00:00:00.000000000 +0000
+++ GCC/gcc/config/soft-fp/extended.h 2006-02-18 02:11:04.000000000 +0000
@@ -0,0 +1,453 @@
+/* Software floating-point emulation.
+ Definitions for IEEE Extended Precision.
+ Copyright (C) 1999,2006 Free Software Foundation, Inc.
+ This file is part of GCC.
+ Contributed by Jakub Jelinek (jj@ultra.linux.cz).
+
+ GCC is free software; you can redistribute it and/or modify it under
+ the terms of the GNU General Public License as published by the Free
+ Software Foundation; either version 2, or (at your option) any later
+ version.
+
+ In addition to the permissions in the GNU 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 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.)
+
+ GCC 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 General Public License
+ for more details.
+
+ You should have received a copy of the GNU General Public License
+ along with GCC; see the file COPYING. If not, write to the Free
+ Software Foundation, 51 Franklin Street, Fifth Floor, Boston, MA
+ 02110-1301, USA. */
+
+#if _FP_W_TYPE_SIZE < 32
+#error "Here's a nickel, kid. Go buy yourself a real computer."
+#endif
+
+#if _FP_W_TYPE_SIZE < 64
+#define _FP_FRACTBITS_E (4*_FP_W_TYPE_SIZE)
+#else
+#define _FP_FRACTBITS_E (2*_FP_W_TYPE_SIZE)
+#endif
+
+#define _FP_FRACBITS_E 64
+#define _FP_FRACXBITS_E (_FP_FRACTBITS_E - _FP_FRACBITS_E)
+#define _FP_WFRACBITS_E (_FP_WORKBITS + _FP_FRACBITS_E)
+#define _FP_WFRACXBITS_E (_FP_FRACTBITS_E - _FP_WFRACBITS_E)
+#define _FP_EXPBITS_E 15
+#define _FP_EXPBIAS_E 16383
+#define _FP_EXPMAX_E 32767
+
+#define _FP_QNANBIT_E \
+ ((_FP_W_TYPE)1 << (_FP_FRACBITS_E-2) % _FP_W_TYPE_SIZE)
+#define _FP_QNANBIT_SH_E \
+ ((_FP_W_TYPE)1 << (_FP_FRACBITS_E-2+_FP_WORKBITS) % _FP_W_TYPE_SIZE)
+#define _FP_IMPLBIT_E \
+ ((_FP_W_TYPE)1 << (_FP_FRACBITS_E-1) % _FP_W_TYPE_SIZE)
+#define _FP_IMPLBIT_SH_E \
+ ((_FP_W_TYPE)1 << (_FP_FRACBITS_E-1+_FP_WORKBITS) % _FP_W_TYPE_SIZE)
+#define _FP_OVERFLOW_E \
+ ((_FP_W_TYPE)1 << (_FP_WFRACBITS_E % _FP_W_TYPE_SIZE))
+
+typedef float XFtype __attribute__((mode(XF)));
+
+#if _FP_W_TYPE_SIZE < 64
+
+union _FP_UNION_E
+{
+ XFtype flt;
+ struct
+ {
+#if __BYTE_ORDER == __BIG_ENDIAN
+ unsigned long pad1 : _FP_W_TYPE_SIZE;
+ unsigned long pad2 : (_FP_W_TYPE_SIZE - 1 - _FP_EXPBITS_E);
+ unsigned long sign : 1;
+ unsigned long exp : _FP_EXPBITS_E;
+ unsigned long frac1 : _FP_W_TYPE_SIZE;
+ unsigned long frac0 : _FP_W_TYPE_SIZE;
+#else
+ unsigned long frac0 : _FP_W_TYPE_SIZE;
+ unsigned long frac1 : _FP_W_TYPE_SIZE;
+ unsigned exp : _FP_EXPBITS_E;
+ unsigned sign : 1;
+#endif /* not bigendian */
+ } bits __attribute__((packed));
+};
+
+
+#define FP_DECL_E(X) _FP_DECL(4,X)
+
+#define FP_UNPACK_RAW_E(X, val) \
+ do { \
+ union _FP_UNION_E _flo; _flo.flt = (val); \
+ \
+ X##_f[2] = 0; X##_f[3] = 0; \
+ X##_f[0] = _flo.bits.frac0; \
+ X##_f[1] = _flo.bits.frac1; \
+ X##_e = _flo.bits.exp; \
+ X##_s = _flo.bits.sign; \
+ if (!X##_e && (X##_f[1] || X##_f[0]) \
+ && !(X##_f[1] & _FP_IMPLBIT_E)) \
+ { \
+ X##_e++; \
+ FP_SET_EXCEPTION(FP_EX_DENORM); \
+ } \
+ } while (0)
+
+#define FP_UNPACK_RAW_EP(X, val) \
+ do { \
+ union _FP_UNION_E *_flo = \
+ (union _FP_UNION_E *)(val); \
+ \
+ X##_f[2] = 0; X##_f[3] = 0; \
+ X##_f[0] = _flo->bits.frac0; \
+ X##_f[1] = _flo->bits.frac1; \
+ X##_e = _flo->bits.exp; \
+ X##_s = _flo->bits.sign; \
+ if (!X##_e && (X##_f[1] || X##_f[0]) \
+ && !(X##_f[1] & _FP_IMPLBIT_E)) \
+ { \
+ X##_e++; \
+ FP_SET_EXCEPTION(FP_EX_DENORM); \
+ } \
+ } while (0)
+
+#define FP_PACK_RAW_E(val, X) \
+ do { \
+ union _FP_UNION_E _flo; \
+ \
+ if (X##_e) X##_f[1] |= _FP_IMPLBIT_E; \
+ else X##_f[1] &= ~(_FP_IMPLBIT_E); \
+ _flo.bits.frac0 = X##_f[0]; \
+ _flo.bits.frac1 = X##_f[1]; \
+ _flo.bits.exp = X##_e; \
+ _flo.bits.sign = X##_s; \
+ \
+ (val) = _flo.flt; \
+ } while (0)
+
+#define FP_PACK_RAW_EP(val, X) \
+ do { \
+ if (!FP_INHIBIT_RESULTS) \
+ { \
+ union _FP_UNION_E *_flo = \
+ (union _FP_UNION_E *)(val); \
+ \
+ if (X##_e) X##_f[1] |= _FP_IMPLBIT_E; \
+ else X##_f[1] &= ~(_FP_IMPLBIT_E); \
+ _flo->bits.frac0 = X##_f[0]; \
+ _flo->bits.frac1 = X##_f[1]; \
+ _flo->bits.exp = X##_e; \
+ _flo->bits.sign = X##_s; \
+ } \
+ } while (0)
+
+#define FP_UNPACK_E(X,val) \
+ do { \
+ FP_UNPACK_RAW_E(X,val); \
+ _FP_UNPACK_CANONICAL(E,4,X); \
+ } while (0)
+
+#define FP_UNPACK_EP(X,val) \
+ do { \
+ FP_UNPACK_RAW_EP(X,val); \
+ _FP_UNPACK_CANONICAL(E,4,X); \
+ } while (0)
+
+#define FP_UNPACK_SEMIRAW_E(X,val) \
+ do { \
+ _FP_UNPACK_RAW_E(X,val); \
+ _FP_UNPACK_SEMIRAW(E,4,X); \
+ } while (0)
+
+#define FP_UNPACK_SEMIRAW_EP(X,val) \
+ do { \
+ _FP_UNPACK_RAW_EP(X,val); \
+ _FP_UNPACK_SEMIRAW(E,4,X); \
+ } while (0)
+
+#define FP_PACK_E(val,X) \
+ do { \
+ _FP_PACK_CANONICAL(E,4,X); \
+ FP_PACK_RAW_E(val,X); \
+ } while (0)
+
+#define FP_PACK_EP(val,X) \
+ do { \
+ _FP_PACK_CANONICAL(E,4,X); \
+ FP_PACK_RAW_EP(val,X); \
+ } while (0)
+
+#define FP_PACK_SEMIRAW_E(val,X) \
+ do { \
+ _FP_PACK_SEMIRAW(E,4,X); \
+ _FP_PACK_RAW_E(val,X); \
+ } while (0)
+
+#define FP_PACK_SEMIRAW_EP(val,X) \
+ do { \
+ _FP_PACK_SEMIRAW(E,4,X); \
+ _FP_PACK_RAW_EP(val,X); \
+ } while (0)
+
+#define FP_ISSIGNAN_E(X) _FP_ISSIGNAN(E,4,X)
+#define FP_NEG_E(R,X) _FP_NEG(E,4,R,X)
+#define FP_ADD_E(R,X,Y) _FP_ADD(E,4,R,X,Y)
+#define FP_SUB_E(R,X,Y) _FP_SUB(E,4,R,X,Y)
+#define FP_MUL_E(R,X,Y) _FP_MUL(E,4,R,X,Y)
+#define FP_DIV_E(R,X,Y) _FP_DIV(E,4,R,X,Y)
+#define FP_SQRT_E(R,X) _FP_SQRT(E,4,R,X)
+
+/*
+ * Square root algorithms:
+ * We have just one right now, maybe Newton approximation
+ * should be added for those machines where division is fast.
+ * This has special _E version because standard _4 square
+ * root would not work (it has to start normally with the
+ * second word and not the first), but as we have to do it
+ * anyway, we optimize it by doing most of the calculations
+ * in two UWtype registers instead of four.
+ */
+
+#define _FP_SQRT_MEAT_E(R, S, T, X, q) \
+ do { \
+ q = (_FP_W_TYPE)1 << (_FP_W_TYPE_SIZE - 1); \
+ _FP_FRAC_SRL_4(X, (_FP_WORKBITS)); \
+ while (q) \
+ { \
+ T##_f[1] = S##_f[1] + q; \
+ if (T##_f[1] <= X##_f[1]) \
+ { \
+ S##_f[1] = T##_f[1] + q; \
+ X##_f[1] -= T##_f[1]; \
+ R##_f[1] += q; \
+ } \
+ _FP_FRAC_SLL_2(X, 1); \
+ q >>= 1; \
+ } \
+ q = (_FP_W_TYPE)1 << (_FP_W_TYPE_SIZE - 1); \
+ while (q) \
+ { \
+ T##_f[0] = S##_f[0] + q; \
+ T##_f[1] = S##_f[1]; \
+ if (T##_f[1] < X##_f[1] || \
+ (T##_f[1] == X##_f[1] && \
+ T##_f[0] <= X##_f[0])) \
+ { \
+ S##_f[0] = T##_f[0] + q; \
+ S##_f[1] += (T##_f[0] > S##_f[0]); \
+ _FP_FRAC_DEC_2(X, T); \
+ R##_f[0] += q; \
+ } \
+ _FP_FRAC_SLL_2(X, 1); \
+ q >>= 1; \
+ } \
+ _FP_FRAC_SLL_4(R, (_FP_WORKBITS)); \
+ if (X##_f[0] | X##_f[1]) \
+ { \
+ if (S##_f[1] < X##_f[1] || \
+ (S##_f[1] == X##_f[1] && \
+ S##_f[0] < X##_f[0])) \
+ R##_f[0] |= _FP_WORK_ROUND; \
+ R##_f[0] |= _FP_WORK_STICKY; \
+ } \
+ } while (0)
+
+#define FP_CMP_E(r,X,Y,un) _FP_CMP(E,4,r,X,Y,un)
+#define FP_CMP_EQ_E(r,X,Y) _FP_CMP_EQ(E,4,r,X,Y)
+#define FP_CMP_UNORD_E(r,X,Y) _FP_CMP_UNORD(E,4,r,X,Y)
+
+#define FP_TO_INT_E(r,X,rsz,rsg) _FP_TO_INT(E,4,r,X,rsz,rsg)
+#define FP_FROM_INT_E(X,r,rs,rt) _FP_FROM_INT(E,4,X,r,rs,rt)
+
+#define _FP_FRAC_HIGH_E(X) (X##_f[2])
+#define _FP_FRAC_HIGH_RAW_E(X) (X##_f[1])
+
+#else /* not _FP_W_TYPE_SIZE < 64 */
+union _FP_UNION_E
+{
+ XFtype flt;
+ struct {
+#if __BYTE_ORDER == __BIG_ENDIAN
+ unsigned long pad : (_FP_W_TYPE_SIZE - 1 - _FP_EXPBITS_E);
+ unsigned sign : 1;
+ unsigned exp : _FP_EXPBITS_E;
+ unsigned long frac : _FP_W_TYPE_SIZE;
+#else
+ unsigned long frac : _FP_W_TYPE_SIZE;
+ unsigned exp : _FP_EXPBITS_E;
+ unsigned sign : 1;
+#endif
+ } bits;
+};
+
+#define FP_DECL_E(X) _FP_DECL(2,X)
+
+#define FP_UNPACK_RAW_E(X, val) \
+ do { \
+ union _FP_UNION_E _flo; _flo.flt = (val); \
+ \
+ X##_f0 = _flo.bits.frac; \
+ X##_f1 = 0; \
+ X##_e = _flo.bits.exp; \
+ X##_s = _flo.bits.sign; \
+ if (!X##_e && X##_f0 && !(X##_f0 & _FP_IMPLBIT_E)) \
+ { \
+ X##_e++; \
+ FP_SET_EXCEPTION(FP_EX_DENORM); \
+ } \
+ } while (0)
+
+#define FP_UNPACK_RAW_EP(X, val) \
+ do { \
+ union _FP_UNION_E *_flo = \
+ (union _FP_UNION_E *)(val); \
+ \
+ X##_f0 = _flo->bits.frac; \
+ X##_f1 = 0; \
+ X##_e = _flo->bits.exp; \
+ X##_s = _flo->bits.sign; \
+ if (!X##_e && X##_f0 && !(X##_f0 & _FP_IMPLBIT_E)) \
+ { \
+ X##_e++; \
+ FP_SET_EXCEPTION(FP_EX_DENORM); \
+ } \
+ } while (0)
+
+#define FP_PACK_RAW_E(val, X) \
+ do { \
+ union _FP_UNION_E _flo; \
+ \
+ if (X##_e) X##_f0 |= _FP_IMPLBIT_E; \
+ else X##_f0 &= ~(_FP_IMPLBIT_E); \
+ _flo.bits.frac = X##_f0; \
+ _flo.bits.exp = X##_e; \
+ _flo.bits.sign = X##_s; \
+ \
+ (val) = _flo.flt; \
+ } while (0)
+
+#define FP_PACK_RAW_EP(fs, val, X) \
+ do { \
+ if (!FP_INHIBIT_RESULTS) \
+ { \
+ union _FP_UNION_E *_flo = \
+ (union _FP_UNION_E *)(val); \
+ \
+ if (X##_e) X##_f0 |= _FP_IMPLBIT_E; \
+ else X##_f0 &= ~(_FP_IMPLBIT_E); \
+ _flo->bits.frac = X##_f0; \
+ _flo->bits.exp = X##_e; \
+ _flo->bits.sign = X##_s; \
+ } \
+ } while (0)
+
+
+#define FP_UNPACK_E(X,val) \
+ do { \
+ FP_UNPACK_RAW_E(X,val); \
+ _FP_UNPACK_CANONICAL(E,2,X); \
+ } while (0)
+
+#define FP_UNPACK_EP(X,val) \
+ do { \
+ FP_UNPACK_RAW_EP(X,val); \
+ _FP_UNPACK_CANONICAL(E,2,X); \
+ } while (0)
+
+#define FP_UNPACK_SEMIRAW_E(X,val) \
+ do { \
+ _FP_UNPACK_RAW_E(X,val); \
+ _FP_UNPACK_SEMIRAW(E,2,X); \
+ } while (0)
+
+#define FP_UNPACK_SEMIRAW_EP(X,val) \
+ do { \
+ _FP_UNPACK_RAW_EP(X,val); \
+ _FP_UNPACK_SEMIRAW(E,2,X); \
+ } while (0)
+
+#define FP_PACK_E(val,X) \
+ do { \
+ _FP_PACK_CANONICAL(E,2,X); \
+ FP_PACK_RAW_E(val,X); \
+ } while (0)
+
+#define FP_PACK_EP(val,X) \
+ do { \
+ _FP_PACK_CANONICAL(E,2,X); \
+ FP_PACK_RAW_EP(val,X); \
+ } while (0)
+
+#define FP_PACK_SEMIRAW_E(val,X) \
+ do { \
+ _FP_PACK_SEMIRAW(E,2,X); \
+ _FP_PACK_RAW_E(val,X); \
+ } while (0)
+
+#define FP_PACK_SEMIRAW_EP(val,X) \
+ do { \
+ _FP_PACK_SEMIRAW(E,2,X); \
+ _FP_PACK_RAW_EP(val,X); \
+ } while (0)
+
+#define FP_ISSIGNAN_E(X) _FP_ISSIGNAN(E,2,X)
+#define FP_NEG_E(R,X) _FP_NEG(E,2,R,X)
+#define FP_ADD_E(R,X,Y) _FP_ADD(E,2,R,X,Y)
+#define FP_SUB_E(R,X,Y) _FP_SUB(E,2,R,X,Y)
+#define FP_MUL_E(R,X,Y) _FP_MUL(E,2,R,X,Y)
+#define FP_DIV_E(R,X,Y) _FP_DIV(E,2,R,X,Y)
+#define FP_SQRT_E(R,X) _FP_SQRT(E,2,R,X)
+
+/*
+ * Square root algorithms:
+ * We have just one right now, maybe Newton approximation
+ * should be added for those machines where division is fast.
+ * We optimize it by doing most of the calculations
+ * in one UWtype registers instead of two, although we don't
+ * have to.
+ */
+#define _FP_SQRT_MEAT_E(R, S, T, X, q) \
+ do { \
+ q = (_FP_W_TYPE)1 << (_FP_W_TYPE_SIZE - 1); \
+ _FP_FRAC_SRL_2(X, (_FP_WORKBITS)); \
+ while (q) \
+ { \
+ T##_f0 = S##_f0 + q; \
+ if (T##_f0 <= X##_f0) \
+ { \
+ S##_f0 = T##_f0 + q; \
+ X##_f0 -= T##_f0; \
+ R##_f0 += q; \
+ } \
+ _FP_FRAC_SLL_1(X, 1); \
+ q >>= 1; \
+ } \
+ _FP_FRAC_SLL_2(R, (_FP_WORKBITS)); \
+ if (X##_f0) \
+ { \
+ if (S##_f0 < X##_f0) \
+ R##_f0 |= _FP_WORK_ROUND; \
+ R##_f0 |= _FP_WORK_STICKY; \
+ } \
+ } while (0)
+
+#define FP_CMP_E(r,X,Y,un) _FP_CMP(E,2,r,X,Y,un)
+#define FP_CMP_EQ_E(r,X,Y) _FP_CMP_EQ(E,2,r,X,Y)
+#define FP_CMP_UNORD_E(r,X,Y) _FP_CMP_UNORD(E,2,r,X,Y)
+
+#define FP_TO_INT_E(r,X,rsz,rsg) _FP_TO_INT(E,2,r,X,rsz,rsg)
+#define FP_FROM_INT_E(X,r,rs,rt) _FP_FROM_INT(E,2,X,r,rs,rt)
+
+#define _FP_FRAC_HIGH_E(X) (X##_f1)
+#define _FP_FRAC_HIGH_RAW_E(X) (X##_f0)
+
+#endif /* not _FP_W_TYPE_SIZE < 64 */
diff -rupN GCC.orig/gcc/config/soft-fp/extendsfdf2.c GCC/gcc/config/soft-fp/extendsfdf2.c
--- GCC.orig/gcc/config/soft-fp/extendsfdf2.c 1970-01-01 00:00:00.000000000 +0000
+++ GCC/gcc/config/soft-fp/extendsfdf2.c 2006-02-18 02:11:04.000000000 +0000
@@ -0,0 +1,54 @@
+/* Software floating-point emulation.
+ Return a converted to IEEE double
+ Copyright (C) 1997,1999,2006 Free Software Foundation, Inc.
+ This file is part of GCC.
+ Contributed by Richard Henderson (rth@cygnus.com) and
+ Jakub Jelinek (jj@ultra.linux.cz).
+
+ GCC is free software; you can redistribute it and/or modify it under
+ the terms of the GNU General Public License as published by the Free
+ Software Foundation; either version 2, or (at your option) any later
+ version.
+
+ In addition to the permissions in the GNU 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 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.)
+
+ GCC 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 General Public License
+ for more details.
+
+ You should have received a copy of the GNU General Public License
+ along with GCC; see the file COPYING. If not, write to the Free
+ Software Foundation, 51 Franklin Street, Fifth Floor, Boston, MA
+ 02110-1301, USA. */
+
+#include "soft-fp.h"
+#include "single.h"
+#include "double.h"
+
+DFtype __extendsfdf2(SFtype a)
+{
+ FP_DECL_EX;
+ FP_DECL_S(A);
+ FP_DECL_D(R);
+ DFtype r;
+
+ FP_INIT_ROUNDMODE;
+ FP_UNPACK_RAW_S(A, a);
+#if _FP_W_TYPE_SIZE < _FP_FRACBITS_D
+ FP_EXTEND(D,S,2,1,R,A);
+#else
+ FP_EXTEND(D,S,1,1,R,A);
+#endif
+ FP_PACK_RAW_D(r, R);
+ FP_HANDLE_EXCEPTIONS;
+
+ return r;
+}
diff -rupN GCC.orig/gcc/config/soft-fp/extendsftf2.c GCC/gcc/config/soft-fp/extendsftf2.c
--- GCC.orig/gcc/config/soft-fp/extendsftf2.c 1970-01-01 00:00:00.000000000 +0000
+++ GCC/gcc/config/soft-fp/extendsftf2.c 2006-02-18 02:11:04.000000000 +0000
@@ -0,0 +1,54 @@
+/* Software floating-point emulation.
+ Return a converted to IEEE quad
+ Copyright (C) 1997,1999,2006 Free Software Foundation, Inc.
+ This file is part of GCC.
+ Contributed by Richard Henderson (rth@cygnus.com) and
+ Jakub Jelinek (jj@ultra.linux.cz).
+
+ GCC is free software; you can redistribute it and/or modify it under
+ the terms of the GNU General Public License as published by the Free
+ Software Foundation; either version 2, or (at your option) any later
+ version.
+
+ In addition to the permissions in the GNU 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 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.)
+
+ GCC 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 General Public License
+ for more details.
+
+ You should have received a copy of the GNU General Public License
+ along with GCC; see the file COPYING. If not, write to the Free
+ Software Foundation, 51 Franklin Street, Fifth Floor, Boston, MA
+ 02110-1301, USA. */
+
+#include "soft-fp.h"
+#include "single.h"
+#include "quad.h"
+
+TFtype __extendsftf2(SFtype a)
+{
+ FP_DECL_EX;
+ FP_DECL_S(A);
+ FP_DECL_Q(R);
+ TFtype r;
+
+ FP_INIT_ROUNDMODE;
+ FP_UNPACK_RAW_S(A, a);
+#if (2 * _FP_W_TYPE_SIZE) < _FP_FRACBITS_Q
+ FP_EXTEND(Q,S,4,1,R,A);
+#else
+ FP_EXTEND(Q,S,2,1,R,A);
+#endif
+ FP_PACK_RAW_Q(r, R);
+ FP_HANDLE_EXCEPTIONS;
+
+ return r;
+}
diff -rupN GCC.orig/gcc/config/soft-fp/fixdfdi.c GCC/gcc/config/soft-fp/fixdfdi.c
--- GCC.orig/gcc/config/soft-fp/fixdfdi.c 1970-01-01 00:00:00.000000000 +0000
+++ GCC/gcc/config/soft-fp/fixdfdi.c 2006-02-18 02:11:04.000000000 +0000
@@ -0,0 +1,46 @@
+/* Software floating-point emulation.
+ Convert a to 64bit signed integer
+ Copyright (C) 1997,1999,2006 Free Software Foundation, Inc.
+ This file is part of GCC.
+ Contributed by Richard Henderson (rth@cygnus.com) and
+ Jakub Jelinek (jj@ultra.linux.cz).
+
+ GCC is free software; you can redistribute it and/or modify it under
+ the terms of the GNU General Public License as published by the Free
+ Software Foundation; either version 2, or (at your option) any later
+ version.
+
+ In addition to the permissions in the GNU 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 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.)
+
+ GCC 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 General Public License
+ for more details.
+
+ You should have received a copy of the GNU General Public License
+ along with GCC; see the file COPYING. If not, write to the Free
+ Software Foundation, 51 Franklin Street, Fifth Floor, Boston, MA
+ 02110-1301, USA. */
+
+#include "soft-fp.h"
+#include "double.h"
+
+DItype __fixdfdi(DFtype a)
+{
+ FP_DECL_EX;
+ FP_DECL_D(A);
+ UDItype r;
+
+ FP_UNPACK_RAW_D(A, a);
+ FP_TO_INT_D(r, A, DI_BITS, 1);
+ FP_HANDLE_EXCEPTIONS;
+
+ return r;
+}
diff -rupN GCC.orig/gcc/config/soft-fp/fixdfsi.c GCC/gcc/config/soft-fp/fixdfsi.c
--- GCC.orig/gcc/config/soft-fp/fixdfsi.c 1970-01-01 00:00:00.000000000 +0000
+++ GCC/gcc/config/soft-fp/fixdfsi.c 2006-02-18 02:11:04.000000000 +0000
@@ -0,0 +1,46 @@
+/* Software floating-point emulation.
+ Convert a to 32bit signed integer
+ Copyright (C) 1997,1999,2006 Free Software Foundation, Inc.
+ This file is part of GCC.
+ Contributed by Richard Henderson (rth@cygnus.com) and
+ Jakub Jelinek (jj@ultra.linux.cz).
+
+ GCC is free software; you can redistribute it and/or modify it under
+ the terms of the GNU General Public License as published by the Free
+ Software Foundation; either version 2, or (at your option) any later
+ version.
+
+ In addition to the permissions in the GNU 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 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.)
+
+ GCC 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 General Public License
+ for more details.
+
+ You should have received a copy of the GNU General Public License
+ along with GCC; see the file COPYING. If not, write to the Free
+ Software Foundation, 51 Franklin Street, Fifth Floor, Boston, MA
+ 02110-1301, USA. */
+
+#include "soft-fp.h"
+#include "double.h"
+
+SItype __fixdfsi(DFtype a)
+{
+ FP_DECL_EX;
+ FP_DECL_D(A);
+ USItype r;
+
+ FP_UNPACK_RAW_D(A, a);
+ FP_TO_INT_D(r, A, SI_BITS, 1);
+ FP_HANDLE_EXCEPTIONS;
+
+ return r;
+}
diff -rupN GCC.orig/gcc/config/soft-fp/fixsfdi.c GCC/gcc/config/soft-fp/fixsfdi.c
--- GCC.orig/gcc/config/soft-fp/fixsfdi.c 1970-01-01 00:00:00.000000000 +0000
+++ GCC/gcc/config/soft-fp/fixsfdi.c 2006-02-18 02:11:04.000000000 +0000
@@ -0,0 +1,46 @@
+/* Software floating-point emulation.
+ Convert a to 64bit signed integer
+ Copyright (C) 1997,1999,2006 Free Software Foundation, Inc.
+ This file is part of GCC.
+ Contributed by Richard Henderson (rth@cygnus.com) and
+ Jakub Jelinek (jj@ultra.linux.cz).
+
+ GCC is free software; you can redistribute it and/or modify it under
+ the terms of the GNU General Public License as published by the Free
+ Software Foundation; either version 2, or (at your option) any later
+ version.
+
+ In addition to the permissions in the GNU 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 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.)
+
+ GCC 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 General Public License
+ for more details.
+
+ You should have received a copy of the GNU General Public License
+ along with GCC; see the file COPYING. If not, write to the Free
+ Software Foundation, 51 Franklin Street, Fifth Floor, Boston, MA
+ 02110-1301, USA. */
+
+#include "soft-fp.h"
+#include "single.h"
+
+DItype __fixsfdi(SFtype a)
+{
+ FP_DECL_EX;
+ FP_DECL_S(A);
+ UDItype r;
+
+ FP_UNPACK_RAW_S(A, a);
+ FP_TO_INT_S(r, A, DI_BITS, 1);
+ FP_HANDLE_EXCEPTIONS;
+
+ return r;
+}
diff -rupN GCC.orig/gcc/config/soft-fp/fixsfsi.c GCC/gcc/config/soft-fp/fixsfsi.c
--- GCC.orig/gcc/config/soft-fp/fixsfsi.c 1970-01-01 00:00:00.000000000 +0000
+++ GCC/gcc/config/soft-fp/fixsfsi.c 2006-02-18 02:11:04.000000000 +0000
@@ -0,0 +1,46 @@
+/* Software floating-point emulation.
+ Convert a to 32bit signed integer
+ Copyright (C) 1997,1999,2006 Free Software Foundation, Inc.
+ This file is part of GCC.
+ Contributed by Richard Henderson (rth@cygnus.com) and
+ Jakub Jelinek (jj@ultra.linux.cz).
+
+ GCC is free software; you can redistribute it and/or modify it under
+ the terms of the GNU General Public License as published by the Free
+ Software Foundation; either version 2, or (at your option) any later
+ version.
+
+ In addition to the permissions in the GNU 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 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.)
+
+ GCC 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 General Public License
+ for more details.
+
+ You should have received a copy of the GNU General Public License
+ along with GCC; see the file COPYING. If not, write to the Free
+ Software Foundation, 51 Franklin Street, Fifth Floor, Boston, MA
+ 02110-1301, USA. */
+
+#include "soft-fp.h"
+#include "single.h"
+
+SItype __fixsfsi(SFtype a)
+{
+ FP_DECL_EX;
+ FP_DECL_S(A);
+ USItype r;
+
+ FP_UNPACK_RAW_S(A, a);
+ FP_TO_INT_S(r, A, SI_BITS, 1);
+ FP_HANDLE_EXCEPTIONS;
+
+ return r;
+}
diff -rupN GCC.orig/gcc/config/soft-fp/fixtfdi.c GCC/gcc/config/soft-fp/fixtfdi.c
--- GCC.orig/gcc/config/soft-fp/fixtfdi.c 1970-01-01 00:00:00.000000000 +0000
+++ GCC/gcc/config/soft-fp/fixtfdi.c 2006-02-18 02:11:04.000000000 +0000
@@ -0,0 +1,46 @@
+/* Software floating-point emulation.
+ Convert a to 64bit signed integer
+ Copyright (C) 1997,1999,2006 Free Software Foundation, Inc.
+ This file is part of GCC.
+ Contributed by Richard Henderson (rth@cygnus.com) and
+ Jakub Jelinek (jj@ultra.linux.cz).
+
+ GCC is free software; you can redistribute it and/or modify it under
+ the terms of the GNU General Public License as published by the Free
+ Software Foundation; either version 2, or (at your option) any later
+ version.
+
+ In addition to the permissions in the GNU 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 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.)
+
+ GCC 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 General Public License
+ for more details.
+
+ You should have received a copy of the GNU General Public License
+ along with GCC; see the file COPYING. If not, write to the Free
+ Software Foundation, 51 Franklin Street, Fifth Floor, Boston, MA
+ 02110-1301, USA. */
+
+#include "soft-fp.h"
+#include "quad.h"
+
+DItype __fixtfdi(TFtype a)
+{
+ FP_DECL_EX;
+ FP_DECL_Q(A);
+ UDItype r;
+
+ FP_UNPACK_RAW_Q(A, a);
+ FP_TO_INT_Q(r, A, DI_BITS, 1);
+ FP_HANDLE_EXCEPTIONS;
+
+ return r;
+}
diff -rupN GCC.orig/gcc/config/soft-fp/fixtfsi.c GCC/gcc/config/soft-fp/fixtfsi.c
--- GCC.orig/gcc/config/soft-fp/fixtfsi.c 1970-01-01 00:00:00.000000000 +0000
+++ GCC/gcc/config/soft-fp/fixtfsi.c 2006-02-18 02:11:04.000000000 +0000
@@ -0,0 +1,46 @@
+/* Software floating-point emulation.
+ Convert a to 32bit signed integer
+ Copyright (C) 1997,1999,2006 Free Software Foundation, Inc.
+ This file is part of GCC.
+ Contributed by Richard Henderson (rth@cygnus.com) and
+ Jakub Jelinek (jj@ultra.linux.cz).
+
+ GCC is free software; you can redistribute it and/or modify it under
+ the terms of the GNU General Public License as published by the Free
+ Software Foundation; either version 2, or (at your option) any later
+ version.
+
+ In addition to the permissions in the GNU 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 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.)
+
+ GCC 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 General Public License
+ for more details.
+
+ You should have received a copy of the GNU General Public License
+ along with GCC; see the file COPYING. If not, write to the Free
+ Software Foundation, 51 Franklin Street, Fifth Floor, Boston, MA
+ 02110-1301, USA. */
+
+#include "soft-fp.h"
+#include "quad.h"
+
+SItype __fixtfsi(TFtype a)
+{
+ FP_DECL_EX;
+ FP_DECL_Q(A);
+ USItype r;
+
+ FP_UNPACK_RAW_Q(A, a);
+ FP_TO_INT_Q(r, A, SI_BITS, 1);
+ FP_HANDLE_EXCEPTIONS;
+
+ return r;
+}
diff -rupN GCC.orig/gcc/config/soft-fp/fixunsdfdi.c GCC/gcc/config/soft-fp/fixunsdfdi.c
--- GCC.orig/gcc/config/soft-fp/fixunsdfdi.c 1970-01-01 00:00:00.000000000 +0000
+++ GCC/gcc/config/soft-fp/fixunsdfdi.c 2006-02-18 02:11:04.000000000 +0000
@@ -0,0 +1,46 @@
+/* Software floating-point emulation.
+ Convert a to 64bit unsigned integer
+ Copyright (C) 1997,1999,2006 Free Software Foundation, Inc.
+ This file is part of GCC.
+ Contributed by Richard Henderson (rth@cygnus.com) and
+ Jakub Jelinek (jj@ultra.linux.cz).
+
+ GCC is free software; you can redistribute it and/or modify it under
+ the terms of the GNU General Public License as published by the Free
+ Software Foundation; either version 2, or (at your option) any later
+ version.
+
+ In addition to the permissions in the GNU 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 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.)
+
+ GCC 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 General Public License
+ for more details.
+
+ You should have received a copy of the GNU General Public License
+ along with GCC; see the file COPYING. If not, write to the Free
+ Software Foundation, 51 Franklin Street, Fifth Floor, Boston, MA
+ 02110-1301, USA. */
+
+#include "soft-fp.h"
+#include "double.h"
+
+UDItype __fixunsdfdi(DFtype a)
+{
+ FP_DECL_EX;
+ FP_DECL_D(A);
+ UDItype r;
+
+ FP_UNPACK_RAW_D(A, a);
+ FP_TO_INT_D(r, A, DI_BITS, 0);
+ FP_HANDLE_EXCEPTIONS;
+
+ return r;
+}
diff -rupN GCC.orig/gcc/config/soft-fp/fixunsdfsi.c GCC/gcc/config/soft-fp/fixunsdfsi.c
--- GCC.orig/gcc/config/soft-fp/fixunsdfsi.c 1970-01-01 00:00:00.000000000 +0000
+++ GCC/gcc/config/soft-fp/fixunsdfsi.c 2006-02-18 02:11:04.000000000 +0000
@@ -0,0 +1,46 @@
+/* Software floating-point emulation.
+ Convert a to 32bit unsigned integer
+ Copyright (C) 1997,1999,2006 Free Software Foundation, Inc.
+ This file is part of GCC.
+ Contributed by Richard Henderson (rth@cygnus.com) and
+ Jakub Jelinek (jj@ultra.linux.cz).
+
+ GCC is free software; you can redistribute it and/or modify it under
+ the terms of the GNU General Public License as published by the Free
+ Software Foundation; either version 2, or (at your option) any later
+ version.
+
+ In addition to the permissions in the GNU 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 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.)
+
+ GCC 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 General Public License
+ for more details.
+
+ You should have received a copy of the GNU General Public License
+ along with GCC; see the file COPYING. If not, write to the Free
+ Software Foundation, 51 Franklin Street, Fifth Floor, Boston, MA
+ 02110-1301, USA. */
+
+#include "soft-fp.h"
+#include "double.h"
+
+USItype __fixunsdfsi(DFtype a)
+{
+ FP_DECL_EX;
+ FP_DECL_D(A);
+ USItype r;
+
+ FP_UNPACK_RAW_D(A, a);
+ FP_TO_INT_D(r, A, SI_BITS, 0);
+ FP_HANDLE_EXCEPTIONS;
+
+ return r;
+}
diff -rupN GCC.orig/gcc/config/soft-fp/fixunssfdi.c GCC/gcc/config/soft-fp/fixunssfdi.c
--- GCC.orig/gcc/config/soft-fp/fixunssfdi.c 1970-01-01 00:00:00.000000000 +0000
+++ GCC/gcc/config/soft-fp/fixunssfdi.c 2006-02-18 02:11:04.000000000 +0000
@@ -0,0 +1,46 @@
+/* Software floating-point emulation.
+ Convert a to 64bit unsigned integer
+ Copyright (C) 1997,1999,2006 Free Software Foundation, Inc.
+ This file is part of GCC.
+ Contributed by Richard Henderson (rth@cygnus.com) and
+ Jakub Jelinek (jj@ultra.linux.cz).
+
+ GCC is free software; you can redistribute it and/or modify it under
+ the terms of the GNU General Public License as published by the Free
+ Software Foundation; either version 2, or (at your option) any later
+ version.
+
+ In addition to the permissions in the GNU 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 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.)
+
+ GCC 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 General Public License
+ for more details.
+
+ You should have received a copy of the GNU General Public License
+ along with GCC; see the file COPYING. If not, write to the Free
+ Software Foundation, 51 Franklin Street, Fifth Floor, Boston, MA
+ 02110-1301, USA. */
+
+#include "soft-fp.h"
+#include "single.h"
+
+UDItype __fixunssfdi(SFtype a)
+{
+ FP_DECL_EX;
+ FP_DECL_S(A);
+ UDItype r;
+
+ FP_UNPACK_RAW_S(A, a);
+ FP_TO_INT_S(r, A, DI_BITS, 0);
+ FP_HANDLE_EXCEPTIONS;
+
+ return r;
+}
diff -rupN GCC.orig/gcc/config/soft-fp/fixunssfsi.c GCC/gcc/config/soft-fp/fixunssfsi.c
--- GCC.orig/gcc/config/soft-fp/fixunssfsi.c 1970-01-01 00:00:00.000000000 +0000
+++ GCC/gcc/config/soft-fp/fixunssfsi.c 2006-02-18 02:11:04.000000000 +0000
@@ -0,0 +1,46 @@
+/* Software floating-point emulation.
+ Convert a to 32bit unsigned integer
+ Copyright (C) 1997,1999,2006 Free Software Foundation, Inc.
+ This file is part of GCC.
+ Contributed by Richard Henderson (rth@cygnus.com) and
+ Jakub Jelinek (jj@ultra.linux.cz).
+
+ GCC is free software; you can redistribute it and/or modify it under
+ the terms of the GNU General Public License as published by the Free
+ Software Foundation; either version 2, or (at your option) any later
+ version.
+
+ In addition to the permissions in the GNU 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 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.)
+
+ GCC 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 General Public License
+ for more details.
+
+ You should have received a copy of the GNU General Public License
+ along with GCC; see the file COPYING. If not, write to the Free
+ Software Foundation, 51 Franklin Street, Fifth Floor, Boston, MA
+ 02110-1301, USA. */
+
+#include "soft-fp.h"
+#include "single.h"
+
+USItype __fixunssfsi(SFtype a)
+{
+ FP_DECL_EX;
+ FP_DECL_S(A);
+ USItype r;
+
+ FP_UNPACK_RAW_S(A, a);
+ FP_TO_INT_S(r, A, SI_BITS, 0);
+ FP_HANDLE_EXCEPTIONS;
+
+ return r;
+}
diff -rupN GCC.orig/gcc/config/soft-fp/fixunstfdi.c GCC/gcc/config/soft-fp/fixunstfdi.c
--- GCC.orig/gcc/config/soft-fp/fixunstfdi.c 1970-01-01 00:00:00.000000000 +0000
+++ GCC/gcc/config/soft-fp/fixunstfdi.c 2006-02-18 02:11:04.000000000 +0000
@@ -0,0 +1,46 @@
+/* Software floating-point emulation.
+ Convert a to 64bit unsigned integer
+ Copyright (C) 1997,1999,2006 Free Software Foundation, Inc.
+ This file is part of GCC.
+ Contributed by Richard Henderson (rth@cygnus.com) and
+ Jakub Jelinek (jj@ultra.linux.cz).
+
+ GCC is free software; you can redistribute it and/or modify it under
+ the terms of the GNU General Public License as published by the Free
+ Software Foundation; either version 2, or (at your option) any later
+ version.
+
+ In addition to the permissions in the GNU 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 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.)
+
+ GCC 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 General Public License
+ for more details.
+
+ You should have received a copy of the GNU General Public License
+ along with GCC; see the file COPYING. If not, write to the Free
+ Software Foundation, 51 Franklin Street, Fifth Floor, Boston, MA
+ 02110-1301, USA. */
+
+#include "soft-fp.h"
+#include "quad.h"
+
+UDItype __fixunstfdi(TFtype a)
+{
+ FP_DECL_EX;
+ FP_DECL_Q(A);
+ UDItype r;
+
+ FP_UNPACK_RAW_Q(A, a);
+ FP_TO_INT_Q(r, A, DI_BITS, 0);
+ FP_HANDLE_EXCEPTIONS;
+
+ return r;
+}
diff -rupN GCC.orig/gcc/config/soft-fp/fixunstfsi.c GCC/gcc/config/soft-fp/fixunstfsi.c
--- GCC.orig/gcc/config/soft-fp/fixunstfsi.c 1970-01-01 00:00:00.000000000 +0000
+++ GCC/gcc/config/soft-fp/fixunstfsi.c 2006-02-18 02:11:04.000000000 +0000
@@ -0,0 +1,46 @@
+/* Software floating-point emulation.
+ Convert a to 32bit unsigned integer
+ Copyright (C) 1997,1999,2006 Free Software Foundation, Inc.
+ This file is part of GCC.
+ Contributed by Richard Henderson (rth@cygnus.com) and
+ Jakub Jelinek (jj@ultra.linux.cz).
+
+ GCC is free software; you can redistribute it and/or modify it under
+ the terms of the GNU General Public License as published by the Free
+ Software Foundation; either version 2, or (at your option) any later
+ version.
+
+ In addition to the permissions in the GNU 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 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.)
+
+ GCC 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 General Public License
+ for more details.
+
+ You should have received a copy of the GNU General Public License
+ along with GCC; see the file COPYING. If not, write to the Free
+ Software Foundation, 51 Franklin Street, Fifth Floor, Boston, MA
+ 02110-1301, USA. */
+
+#include "soft-fp.h"
+#include "quad.h"
+
+USItype __fixunstfsi(TFtype a)
+{
+ FP_DECL_EX;
+ FP_DECL_Q(A);
+ USItype r;
+
+ FP_UNPACK_RAW_Q(A, a);
+ FP_TO_INT_Q(r, A, SI_BITS, 0);
+ FP_HANDLE_EXCEPTIONS;
+
+ return r;
+}
diff -rupN GCC.orig/gcc/config/soft-fp/floatdidf.c GCC/gcc/config/soft-fp/floatdidf.c
--- GCC.orig/gcc/config/soft-fp/floatdidf.c 1970-01-01 00:00:00.000000000 +0000
+++ GCC/gcc/config/soft-fp/floatdidf.c 2006-02-18 02:11:04.000000000 +0000
@@ -0,0 +1,46 @@
+/* Software floating-point emulation.
+ Convert a 64bit signed integer to IEEE double
+ Copyright (C) 1997,1999,2006 Free Software Foundation, Inc.
+ This file is part of GCC.
+ Contributed by Richard Henderson (rth@cygnus.com) and
+ Jakub Jelinek (jj@ultra.linux.cz).
+
+ GCC is free software; you can redistribute it and/or modify it under
+ the terms of the GNU General Public License as published by the Free
+ Software Foundation; either version 2, or (at your option) any later
+ version.
+
+ In addition to the permissions in the GNU 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 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.)
+
+ GCC 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 General Public License
+ for more details.
+
+ You should have received a copy of the GNU General Public License
+ along with GCC; see the file COPYING. If not, write to the Free
+ Software Foundation, 51 Franklin Street, Fifth Floor, Boston, MA
+ 02110-1301, USA. */
+
+#include "soft-fp.h"
+#include "double.h"
+
+DFtype __floatdidf(DItype i)
+{
+ FP_DECL_EX;
+ FP_DECL_D(A);
+ DFtype a;
+
+ FP_FROM_INT_D(A, i, DI_BITS, UDItype);
+ FP_PACK_RAW_D(a, A);
+ FP_HANDLE_EXCEPTIONS;
+
+ return a;
+}
diff -rupN GCC.orig/gcc/config/soft-fp/floatdisf.c GCC/gcc/config/soft-fp/floatdisf.c
--- GCC.orig/gcc/config/soft-fp/floatdisf.c 1970-01-01 00:00:00.000000000 +0000
+++ GCC/gcc/config/soft-fp/floatdisf.c 2006-02-18 02:11:04.000000000 +0000
@@ -0,0 +1,46 @@
+/* Software floating-point emulation.
+ Convert a 64bit signed integer to IEEE single
+ Copyright (C) 1997,1999,2006 Free Software Foundation, Inc.
+ This file is part of GCC.
+ Contributed by Richard Henderson (rth@cygnus.com) and
+ Jakub Jelinek (jj@ultra.linux.cz).
+
+ GCC is free software; you can redistribute it and/or modify it under
+ the terms of the GNU General Public License as published by the Free
+ Software Foundation; either version 2, or (at your option) any later
+ version.
+
+ In addition to the permissions in the GNU 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 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.)
+
+ GCC 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 General Public License
+ for more details.
+
+ You should have received a copy of the GNU General Public License
+ along with GCC; see the file COPYING. If not, write to the Free
+ Software Foundation, 51 Franklin Street, Fifth Floor, Boston, MA
+ 02110-1301, USA. */
+
+#include "soft-fp.h"
+#include "single.h"
+
+SFtype __floatdisf(DItype i)
+{
+ FP_DECL_EX;
+ FP_DECL_S(A);
+ SFtype a;
+
+ FP_FROM_INT_S(A, i, DI_BITS, UDItype);
+ FP_PACK_RAW_S(a, A);
+ FP_HANDLE_EXCEPTIONS;
+
+ return a;
+}
diff -rupN GCC.orig/gcc/config/soft-fp/floatditf.c GCC/gcc/config/soft-fp/floatditf.c
--- GCC.orig/gcc/config/soft-fp/floatditf.c 1970-01-01 00:00:00.000000000 +0000
+++ GCC/gcc/config/soft-fp/floatditf.c 2006-02-18 02:11:04.000000000 +0000
@@ -0,0 +1,46 @@
+/* Software floating-point emulation.
+ Convert a 64bit signed integer to IEEE quad
+ Copyright (C) 1997,1999,2006 Free Software Foundation, Inc.
+ This file is part of GCC.
+ Contributed by Richard Henderson (rth@cygnus.com) and
+ Jakub Jelinek (jj@ultra.linux.cz).
+
+ GCC is free software; you can redistribute it and/or modify it under
+ the terms of the GNU General Public License as published by the Free
+ Software Foundation; either version 2, or (at your option) any later
+ version.
+
+ In addition to the permissions in the GNU 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 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.)
+
+ GCC 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 General Public License
+ for more details.
+
+ You should have received a copy of the GNU General Public License
+ along with GCC; see the file COPYING. If not, write to the Free
+ Software Foundation, 51 Franklin Street, Fifth Floor, Boston, MA
+ 02110-1301, USA. */
+
+#include "soft-fp.h"
+#include "quad.h"
+
+TFtype __floatditf(DItype i)
+{
+ FP_DECL_EX;
+ FP_DECL_Q(A);
+ TFtype a;
+
+ FP_FROM_INT_Q(A, i, DI_BITS, UDItype);
+ FP_PACK_RAW_Q(a, A);
+ FP_HANDLE_EXCEPTIONS;
+
+ return a;
+}
diff -rupN GCC.orig/gcc/config/soft-fp/floatsidf.c GCC/gcc/config/soft-fp/floatsidf.c
--- GCC.orig/gcc/config/soft-fp/floatsidf.c 1970-01-01 00:00:00.000000000 +0000
+++ GCC/gcc/config/soft-fp/floatsidf.c 2006-02-18 02:11:04.000000000 +0000
@@ -0,0 +1,46 @@
+/* Software floating-point emulation.
+ Convert a 32bit signed integer to IEEE double
+ Copyright (C) 1997,1999,2006 Free Software Foundation, Inc.
+ This file is part of GCC.
+ Contributed by Richard Henderson (rth@cygnus.com) and
+ Jakub Jelinek (jj@ultra.linux.cz).
+
+ GCC is free software; you can redistribute it and/or modify it under
+ the terms of the GNU General Public License as published by the Free
+ Software Foundation; either version 2, or (at your option) any later
+ version.
+
+ In addition to the permissions in the GNU 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 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.)
+
+ GCC 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 General Public License
+ for more details.
+
+ You should have received a copy of the GNU General Public License
+ along with GCC; see the file COPYING. If not, write to the Free
+ Software Foundation, 51 Franklin Street, Fifth Floor, Boston, MA
+ 02110-1301, USA. */
+
+#include "soft-fp.h"
+#include "double.h"
+
+DFtype __floatsidf(SItype i)
+{
+ FP_DECL_EX;
+ FP_DECL_D(A);
+ DFtype a;
+
+ FP_FROM_INT_D(A, i, SI_BITS, USItype);
+ FP_PACK_RAW_D(a, A);
+ FP_HANDLE_EXCEPTIONS;
+
+ return a;
+}
diff -rupN GCC.orig/gcc/config/soft-fp/floatsisf.c GCC/gcc/config/soft-fp/floatsisf.c
--- GCC.orig/gcc/config/soft-fp/floatsisf.c 1970-01-01 00:00:00.000000000 +0000
+++ GCC/gcc/config/soft-fp/floatsisf.c 2006-02-18 02:11:04.000000000 +0000
@@ -0,0 +1,46 @@
+/* Software floating-point emulation.
+ Convert a 32bit signed integer to IEEE single
+ Copyright (C) 1997,1999,2006 Free Software Foundation, Inc.
+ This file is part of GCC.
+ Contributed by Richard Henderson (rth@cygnus.com) and
+ Jakub Jelinek (jj@ultra.linux.cz).
+
+ GCC is free software; you can redistribute it and/or modify it under
+ the terms of the GNU General Public License as published by the Free
+ Software Foundation; either version 2, or (at your option) any later
+ version.
+
+ In addition to the permissions in the GNU 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 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.)
+
+ GCC 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 General Public License
+ for more details.
+
+ You should have received a copy of the GNU General Public License
+ along with GCC; see the file COPYING. If not, write to the Free
+ Software Foundation, 51 Franklin Street, Fifth Floor, Boston, MA
+ 02110-1301, USA. */
+
+#include "soft-fp.h"
+#include "single.h"
+
+SFtype __floatsisf(SItype i)
+{
+ FP_DECL_EX;
+ FP_DECL_S(A);
+ SFtype a;
+
+ FP_FROM_INT_S(A, i, SI_BITS, USItype);
+ FP_PACK_RAW_S(a, A);
+ FP_HANDLE_EXCEPTIONS;
+
+ return a;
+}
diff -rupN GCC.orig/gcc/config/soft-fp/floatsitf.c GCC/gcc/config/soft-fp/floatsitf.c
--- GCC.orig/gcc/config/soft-fp/floatsitf.c 1970-01-01 00:00:00.000000000 +0000
+++ GCC/gcc/config/soft-fp/floatsitf.c 2006-02-18 02:11:04.000000000 +0000
@@ -0,0 +1,46 @@
+/* Software floating-point emulation.
+ Convert a 32bit signed integer to IEEE quad
+ Copyright (C) 1997,1999,2006 Free Software Foundation, Inc.
+ This file is part of GCC.
+ Contributed by Richard Henderson (rth@cygnus.com) and
+ Jakub Jelinek (jj@ultra.linux.cz).
+
+ GCC is free software; you can redistribute it and/or modify it under
+ the terms of the GNU General Public License as published by the Free
+ Software Foundation; either version 2, or (at your option) any later
+ version.
+
+ In addition to the permissions in the GNU 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 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.)
+
+ GCC 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 General Public License
+ for more details.
+
+ You should have received a copy of the GNU General Public License
+ along with GCC; see the file COPYING. If not, write to the Free
+ Software Foundation, 51 Franklin Street, Fifth Floor, Boston, MA
+ 02110-1301, USA. */
+
+#include "soft-fp.h"
+#include "quad.h"
+
+TFtype __floatsitf(SItype i)
+{
+ FP_DECL_EX;
+ FP_DECL_Q(A);
+ TFtype a;
+
+ FP_FROM_INT_Q(A, i, SI_BITS, USItype);
+ FP_PACK_RAW_Q(a, A);
+ FP_HANDLE_EXCEPTIONS;
+
+ return a;
+}
diff -rupN GCC.orig/gcc/config/soft-fp/floatundidf.c GCC/gcc/config/soft-fp/floatundidf.c
--- GCC.orig/gcc/config/soft-fp/floatundidf.c 1970-01-01 00:00:00.000000000 +0000
+++ GCC/gcc/config/soft-fp/floatundidf.c 2006-02-18 02:11:04.000000000 +0000
@@ -0,0 +1,47 @@
+/* Software floating-point emulation.
+ Convert a 64bit unsigned integer to IEEE double
+ Copyright (C) 1997,1999, 2006 Free Software Foundation, Inc.
+ This file is part of GCC.
+ Contributed by Richard Henderson (rth@cygnus.com) and
+ Jakub Jelinek (jj@ultra.linux.cz).
+
+ GCC is free software; you can redistribute it and/or modify it under
+ the terms of the GNU General Public License as published by the Free
+ Software Foundation; either version 2, or (at your option) any later
+ version.
+
+ In addition to the permissions in the GNU 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 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.)
+
+ GCC 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 General Public License
+ for more details.
+
+ You should have received a copy of the GNU General Public License
+ along with GCC; see the file COPYING. If not, write to the Free
+ Software Foundation, 51 Franklin Street, Fifth Floor, Boston, MA
+ 02110-1301, USA. */
+
+#include "soft-fp.h"
+#include "double.h"
+
+double
+__floatundidf(UDItype i)
+{
+ FP_DECL_EX;
+ FP_DECL_D(A);
+ DFtype a;
+
+ FP_FROM_INT_D(A, i, DI_BITS, UDItype);
+ FP_PACK_RAW_D(a, A);
+ FP_HANDLE_EXCEPTIONS;
+
+ return a;
+}
diff -rupN GCC.orig/gcc/config/soft-fp/floatundisf.c GCC/gcc/config/soft-fp/floatundisf.c
--- GCC.orig/gcc/config/soft-fp/floatundisf.c 1970-01-01 00:00:00.000000000 +0000
+++ GCC/gcc/config/soft-fp/floatundisf.c 2006-02-18 02:11:04.000000000 +0000
@@ -0,0 +1,47 @@
+/* Software floating-point emulation.
+ Convert a 64bit unsigned integer to IEEE single
+ Copyright (C) 1997,1999, 2006 Free Software Foundation, Inc.
+ This file is part of GCC.
+ Contributed by Richard Henderson (rth@cygnus.com) and
+ Jakub Jelinek (jj@ultra.linux.cz).
+
+ GCC is free software; you can redistribute it and/or modify it under
+ the terms of the GNU General Public License as published by the Free
+ Software Foundation; either version 2, or (at your option) any later
+ version.
+
+ In addition to the permissions in the GNU 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 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.)
+
+ GCC 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 General Public License
+ for more details.
+
+ You should have received a copy of the GNU General Public License
+ along with GCC; see the file COPYING. If not, write to the Free
+ Software Foundation, 51 Franklin Street, Fifth Floor, Boston, MA
+ 02110-1301, USA. */
+
+#include "soft-fp.h"
+#include "single.h"
+
+float
+__floatundisf(UDItype i)
+{
+ FP_DECL_EX;
+ FP_DECL_S(A);
+ SFtype a;
+
+ FP_FROM_INT_S(A, i, DI_BITS, UDItype);
+ FP_PACK_RAW_S(a, A);
+ FP_HANDLE_EXCEPTIONS;
+
+ return a;
+}
diff -rupN GCC.orig/gcc/config/soft-fp/floatunditf.c GCC/gcc/config/soft-fp/floatunditf.c
--- GCC.orig/gcc/config/soft-fp/floatunditf.c 1970-01-01 00:00:00.000000000 +0000
+++ GCC/gcc/config/soft-fp/floatunditf.c 2006-02-18 02:11:04.000000000 +0000
@@ -0,0 +1,47 @@
+/* Software floating-point emulation.
+ Convert a 64bit unsigned integer to IEEE quad
+ Copyright (C) 1997,1999, 2006 Free Software Foundation, Inc.
+ This file is part of GCC.
+ Contributed by Richard Henderson (rth@cygnus.com) and
+ Jakub Jelinek (jj@ultra.linux.cz).
+
+ GCC is free software; you can redistribute it and/or modify it under
+ the terms of the GNU General Public License as published by the Free
+ Software Foundation; either version 2, or (at your option) any later
+ version.
+
+ In addition to the permissions in the GNU 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 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.)
+
+ GCC 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 General Public License
+ for more details.
+
+ You should have received a copy of the GNU General Public License
+ along with GCC; see the file COPYING. If not, write to the Free
+ Software Foundation, 51 Franklin Street, Fifth Floor, Boston, MA
+ 02110-1301, USA. */
+
+#include "soft-fp.h"
+#include "quad.h"
+
+TFtype
+__floatunditf(UDItype i)
+{
+ FP_DECL_EX;
+ FP_DECL_Q(A);
+ TFtype a;
+
+ FP_FROM_INT_Q(A, i, DI_BITS, UDItype);
+ FP_PACK_RAW_Q(a, A);
+ FP_HANDLE_EXCEPTIONS;
+
+ return a;
+}
diff -rupN GCC.orig/gcc/config/soft-fp/floatunsidf.c GCC/gcc/config/soft-fp/floatunsidf.c
--- GCC.orig/gcc/config/soft-fp/floatunsidf.c 1970-01-01 00:00:00.000000000 +0000
+++ GCC/gcc/config/soft-fp/floatunsidf.c 2006-02-18 02:11:04.000000000 +0000
@@ -0,0 +1,47 @@
+/* Software floating-point emulation.
+ Convert a 32bit unsigned integer to IEEE double
+ Copyright (C) 1997,1999, 2006 Free Software Foundation, Inc.
+ This file is part of GCC.
+ Contributed by Richard Henderson (rth@cygnus.com) and
+ Jakub Jelinek (jj@ultra.linux.cz).
+
+ GCC is free software; you can redistribute it and/or modify it under
+ the terms of the GNU General Public License as published by the Free
+ Software Foundation; either version 2, or (at your option) any later
+ version.
+
+ In addition to the permissions in the GNU 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 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.)
+
+ GCC 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 General Public License
+ for more details.
+
+ You should have received a copy of the GNU General Public License
+ along with GCC; see the file COPYING. If not, write to the Free
+ Software Foundation, 51 Franklin Street, Fifth Floor, Boston, MA
+ 02110-1301, USA. */
+
+#include "soft-fp.h"
+#include "double.h"
+
+double
+__floatunsidf(USItype i)
+{
+ FP_DECL_EX;
+ FP_DECL_D(A);
+ DFtype a;
+
+ FP_FROM_INT_D(A, i, SI_BITS, USItype);
+ FP_PACK_RAW_D(a, A);
+ FP_HANDLE_EXCEPTIONS;
+
+ return a;
+}
diff -rupN GCC.orig/gcc/config/soft-fp/floatunsisf.c GCC/gcc/config/soft-fp/floatunsisf.c
--- GCC.orig/gcc/config/soft-fp/floatunsisf.c 1970-01-01 00:00:00.000000000 +0000
+++ GCC/gcc/config/soft-fp/floatunsisf.c 2006-02-18 02:11:04.000000000 +0000
@@ -0,0 +1,47 @@
+/* Software floating-point emulation.
+ Convert a 32bit unsigned integer to IEEE single
+ Copyright (C) 1997,1999, 2006 Free Software Foundation, Inc.
+ This file is part of GCC.
+ Contributed by Richard Henderson (rth@cygnus.com) and
+ Jakub Jelinek (jj@ultra.linux.cz).
+
+ GCC is free software; you can redistribute it and/or modify it under
+ the terms of the GNU General Public License as published by the Free
+ Software Foundation; either version 2, or (at your option) any later
+ version.
+
+ In addition to the permissions in the GNU 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 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.)
+
+ GCC 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 General Public License
+ for more details.
+
+ You should have received a copy of the GNU General Public License
+ along with GCC; see the file COPYING. If not, write to the Free
+ Software Foundation, 51 Franklin Street, Fifth Floor, Boston, MA
+ 02110-1301, USA. */
+
+#include "soft-fp.h"
+#include "single.h"
+
+float
+__floatunsisf(USItype i)
+{
+ FP_DECL_EX;
+ FP_DECL_S(A);
+ SFtype a;
+
+ FP_FROM_INT_S(A, i, SI_BITS, USItype);
+ FP_PACK_RAW_S(a, A);
+ FP_HANDLE_EXCEPTIONS;
+
+ return a;
+}
diff -rupN GCC.orig/gcc/config/soft-fp/floatunsitf.c GCC/gcc/config/soft-fp/floatunsitf.c
--- GCC.orig/gcc/config/soft-fp/floatunsitf.c 1970-01-01 00:00:00.000000000 +0000
+++ GCC/gcc/config/soft-fp/floatunsitf.c 2006-02-18 02:11:04.000000000 +0000
@@ -0,0 +1,47 @@
+/* Software floating-point emulation.
+ Convert a 32bit unsigned integer to IEEE quad
+ Copyright (C) 1997,1999, 2006 Free Software Foundation, Inc.
+ This file is part of GCC.
+ Contributed by Richard Henderson (rth@cygnus.com) and
+ Jakub Jelinek (jj@ultra.linux.cz).
+
+ GCC is free software; you can redistribute it and/or modify it under
+ the terms of the GNU General Public License as published by the Free
+ Software Foundation; either version 2, or (at your option) any later
+ version.
+
+ In addition to the permissions in the GNU 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 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.)
+
+ GCC 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 General Public License
+ for more details.
+
+ You should have received a copy of the GNU General Public License
+ along with GCC; see the file COPYING. If not, write to the Free
+ Software Foundation, 51 Franklin Street, Fifth Floor, Boston, MA
+ 02110-1301, USA. */
+
+#include "soft-fp.h"
+#include "quad.h"
+
+TFtype
+__floatunsitf(USItype i)
+{
+ FP_DECL_EX;
+ FP_DECL_Q(A);
+ TFtype a;
+
+ FP_FROM_INT_Q(A, i, SI_BITS, USItype);
+ FP_PACK_RAW_Q(a, A);
+ FP_HANDLE_EXCEPTIONS;
+
+ return a;
+}
diff -rupN GCC.orig/gcc/config/soft-fp/gedf2.c GCC/gcc/config/soft-fp/gedf2.c
--- GCC.orig/gcc/config/soft-fp/gedf2.c 1970-01-01 00:00:00.000000000 +0000
+++ GCC/gcc/config/soft-fp/gedf2.c 2006-02-18 02:11:04.000000000 +0000
@@ -0,0 +1,51 @@
+/* Software floating-point emulation.
+ Return 0 iff a == b, 1 iff a > b, -2 iff a ? b, -1 iff a < b
+ Copyright (C) 1997,1999,2006 Free Software Foundation, Inc.
+ This file is part of GCC.
+ Contributed by Richard Henderson (rth@cygnus.com) and
+ Jakub Jelinek (jj@ultra.linux.cz).
+
+ GCC is free software; you can redistribute it and/or modify it under
+ the terms of the GNU General Public License as published by the Free
+ Software Foundation; either version 2, or (at your option) any later
+ version.
+
+ In addition to the permissions in the GNU 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 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.)
+
+ GCC 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 General Public License
+ for more details.
+
+ You should have received a copy of the GNU General Public License
+ along with GCC; see the file COPYING. If not, write to the Free
+ Software Foundation, 51 Franklin Street, Fifth Floor, Boston, MA
+ 02110-1301, USA. */
+
+#include "soft-fp.h"
+#include "double.h"
+
+int __gedf2(DFtype a, DFtype b)
+{
+ FP_DECL_EX;
+ FP_DECL_D(A); FP_DECL_D(B);
+ int r;
+
+ FP_UNPACK_RAW_D(A, a);
+ FP_UNPACK_RAW_D(B, b);
+ FP_CMP_D(r, A, B, -2);
+ if (r == -2 && (FP_ISSIGNAN_D(A) || FP_ISSIGNAN_D(B)))
+ FP_SET_EXCEPTION(FP_EX_INVALID);
+ FP_HANDLE_EXCEPTIONS;
+
+ return r;
+}
+
+strong_alias(__gedf2, __gtdf2);
diff -rupN GCC.orig/gcc/config/soft-fp/gesf2.c GCC/gcc/config/soft-fp/gesf2.c
--- GCC.orig/gcc/config/soft-fp/gesf2.c 1970-01-01 00:00:00.000000000 +0000
+++ GCC/gcc/config/soft-fp/gesf2.c 2006-02-18 02:11:04.000000000 +0000
@@ -0,0 +1,51 @@
+/* Software floating-point emulation.
+ Return 0 iff a == b, 1 iff a > b, -2 iff a ? b, -1 iff a < b
+ Copyright (C) 1997,1999,2006 Free Software Foundation, Inc.
+ This file is part of GCC.
+ Contributed by Richard Henderson (rth@cygnus.com) and
+ Jakub Jelinek (jj@ultra.linux.cz).
+
+ GCC is free software; you can redistribute it and/or modify it under
+ the terms of the GNU General Public License as published by the Free
+ Software Foundation; either version 2, or (at your option) any later
+ version.
+
+ In addition to the permissions in the GNU 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 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.)
+
+ GCC 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 General Public License
+ for more details.
+
+ You should have received a copy of the GNU General Public License
+ along with GCC; see the file COPYING. If not, write to the Free
+ Software Foundation, 51 Franklin Street, Fifth Floor, Boston, MA
+ 02110-1301, USA. */
+
+#include "soft-fp.h"
+#include "single.h"
+
+int __gesf2(SFtype a, SFtype b)
+{
+ FP_DECL_EX;
+ FP_DECL_S(A); FP_DECL_S(B);
+ int r;
+
+ FP_UNPACK_RAW_S(A, a);
+ FP_UNPACK_RAW_S(B, b);
+ FP_CMP_S(r, A, B, -2);
+ if (r == -2 && (FP_ISSIGNAN_S(A) || FP_ISSIGNAN_S(B)))
+ FP_SET_EXCEPTION(FP_EX_INVALID);
+ FP_HANDLE_EXCEPTIONS;
+
+ return r;
+}
+
+strong_alias(__gesf2, __gtsf2);
diff -rupN GCC.orig/gcc/config/soft-fp/getf2.c GCC/gcc/config/soft-fp/getf2.c
--- GCC.orig/gcc/config/soft-fp/getf2.c 1970-01-01 00:00:00.000000000 +0000
+++ GCC/gcc/config/soft-fp/getf2.c 2006-02-18 02:11:04.000000000 +0000
@@ -0,0 +1,51 @@
+/* Software floating-point emulation.
+ Return 0 iff a == b, 1 iff a > b, -2 iff a ? b, -1 iff a < b
+ Copyright (C) 1997,1999,2006 Free Software Foundation, Inc.
+ This file is part of GCC.
+ Contributed by Richard Henderson (rth@cygnus.com) and
+ Jakub Jelinek (jj@ultra.linux.cz).
+
+ GCC is free software; you can redistribute it and/or modify it under
+ the terms of the GNU General Public License as published by the Free
+ Software Foundation; either version 2, or (at your option) any later
+ version.
+
+ In addition to the permissions in the GNU 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 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.)
+
+ GCC 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 General Public License
+ for more details.
+
+ You should have received a copy of the GNU General Public License
+ along with GCC; see the file COPYING. If not, write to the Free
+ Software Foundation, 51 Franklin Street, Fifth Floor, Boston, MA
+ 02110-1301, USA. */
+
+#include "soft-fp.h"
+#include "quad.h"
+
+int __getf2(TFtype a, TFtype b)
+{
+ FP_DECL_EX;
+ FP_DECL_Q(A); FP_DECL_Q(B);
+ int r;
+
+ FP_UNPACK_RAW_Q(A, a);
+ FP_UNPACK_RAW_Q(B, b);
+ FP_CMP_Q(r, A, B, -2);
+ if (r == -2 && (FP_ISSIGNAN_Q(A) || FP_ISSIGNAN_Q(B)))
+ FP_SET_EXCEPTION(FP_EX_INVALID);
+ FP_HANDLE_EXCEPTIONS;
+
+ return r;
+}
+
+strong_alias(__getf2, __gttf2);
diff -rupN GCC.orig/gcc/config/soft-fp/ledf2.c GCC/gcc/config/soft-fp/ledf2.c
--- GCC.orig/gcc/config/soft-fp/ledf2.c 1970-01-01 00:00:00.000000000 +0000
+++ GCC/gcc/config/soft-fp/ledf2.c 2006-02-18 02:11:04.000000000 +0000
@@ -0,0 +1,51 @@
+/* Software floating-point emulation.
+ Return 0 iff a == b, 1 iff a > b, 2 iff a ? b, -1 iff a < b
+ Copyright (C) 1997,1999,2006 Free Software Foundation, Inc.
+ This file is part of GCC.
+ Contributed by Richard Henderson (rth@cygnus.com) and
+ Jakub Jelinek (jj@ultra.linux.cz).
+
+ GCC is free software; you can redistribute it and/or modify it under
+ the terms of the GNU General Public License as published by the Free
+ Software Foundation; either version 2, or (at your option) any later
+ version.
+
+ In addition to the permissions in the GNU 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 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.)
+
+ GCC 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 General Public License
+ for more details.
+
+ You should have received a copy of the GNU General Public License
+ along with GCC; see the file COPYING. If not, write to the Free
+ Software Foundation, 51 Franklin Street, Fifth Floor, Boston, MA
+ 02110-1301, USA. */
+
+#include "soft-fp.h"
+#include "double.h"
+
+int __ledf2(DFtype a, DFtype b)
+{
+ FP_DECL_EX;
+ FP_DECL_D(A); FP_DECL_D(B);
+ int r;
+
+ FP_UNPACK_RAW_D(A, a);
+ FP_UNPACK_RAW_D(B, b);
+ FP_CMP_D(r, A, B, 2);
+ if (r == 2 && (FP_ISSIGNAN_D(A) || FP_ISSIGNAN_D(B)))
+ FP_SET_EXCEPTION(FP_EX_INVALID);
+ FP_HANDLE_EXCEPTIONS;
+
+ return r;
+}
+
+strong_alias(__ledf2, __ltdf2);
diff -rupN GCC.orig/gcc/config/soft-fp/lesf2.c GCC/gcc/config/soft-fp/lesf2.c
--- GCC.orig/gcc/config/soft-fp/lesf2.c 1970-01-01 00:00:00.000000000 +0000
+++ GCC/gcc/config/soft-fp/lesf2.c 2006-02-18 02:11:04.000000000 +0000
@@ -0,0 +1,51 @@
+/* Software floating-point emulation.
+ Return 0 iff a == b, 1 iff a > b, 2 iff a ? b, -1 iff a < b
+ Copyright (C) 1997,1999,2006 Free Software Foundation, Inc.
+ This file is part of GCC.
+ Contributed by Richard Henderson (rth@cygnus.com) and
+ Jakub Jelinek (jj@ultra.linux.cz).
+
+ GCC is free software; you can redistribute it and/or modify it under
+ the terms of the GNU General Public License as published by the Free
+ Software Foundation; either version 2, or (at your option) any later
+ version.
+
+ In addition to the permissions in the GNU 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 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.)
+
+ GCC 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 General Public License
+ for more details.
+
+ You should have received a copy of the GNU General Public License
+ along with GCC; see the file COPYING. If not, write to the Free
+ Software Foundation, 51 Franklin Street, Fifth Floor, Boston, MA
+ 02110-1301, USA. */
+
+#include "soft-fp.h"
+#include "single.h"
+
+int __lesf2(SFtype a, SFtype b)
+{
+ FP_DECL_EX;
+ FP_DECL_S(A); FP_DECL_S(B);
+ int r;
+
+ FP_UNPACK_RAW_S(A, a);
+ FP_UNPACK_RAW_S(B, b);
+ FP_CMP_S(r, A, B, 2);
+ if (r == 2 && (FP_ISSIGNAN_S(A) || FP_ISSIGNAN_S(B)))
+ FP_SET_EXCEPTION(FP_EX_INVALID);
+ FP_HANDLE_EXCEPTIONS;
+
+ return r;
+}
+
+strong_alias(__lesf2, __ltsf2);
diff -rupN GCC.orig/gcc/config/soft-fp/letf2.c GCC/gcc/config/soft-fp/letf2.c
--- GCC.orig/gcc/config/soft-fp/letf2.c 1970-01-01 00:00:00.000000000 +0000
+++ GCC/gcc/config/soft-fp/letf2.c 2006-02-18 02:11:04.000000000 +0000
@@ -0,0 +1,51 @@
+/* Software floating-point emulation.
+ Return 0 iff a == b, 1 iff a > b, 2 iff a ? b, -1 iff a < b
+ Copyright (C) 1997,1999,2006 Free Software Foundation, Inc.
+ This file is part of GCC.
+ Contributed by Richard Henderson (rth@cygnus.com) and
+ Jakub Jelinek (jj@ultra.linux.cz).
+
+ GCC is free software; you can redistribute it and/or modify it under
+ the terms of the GNU General Public License as published by the Free
+ Software Foundation; either version 2, or (at your option) any later
+ version.
+
+ In addition to the permissions in the GNU 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 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.)
+
+ GCC 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 General Public License
+ for more details.
+
+ You should have received a copy of the GNU General Public License
+ along with GCC; see the file COPYING. If not, write to the Free
+ Software Foundation, 51 Franklin Street, Fifth Floor, Boston, MA
+ 02110-1301, USA. */
+
+#include "soft-fp.h"
+#include "quad.h"
+
+int __letf2(TFtype a, TFtype b)
+{
+ FP_DECL_EX;
+ FP_DECL_Q(A); FP_DECL_Q(B);
+ int r;
+
+ FP_UNPACK_RAW_Q(A, a);
+ FP_UNPACK_RAW_Q(B, b);
+ FP_CMP_Q(r, A, B, 2);
+ if (r == 2 && (FP_ISSIGNAN_Q(A) || FP_ISSIGNAN_Q(B)))
+ FP_SET_EXCEPTION(FP_EX_INVALID);
+ FP_HANDLE_EXCEPTIONS;
+
+ return r;
+}
+
+strong_alias(__letf2, __lttf2);
diff -rupN GCC.orig/gcc/config/soft-fp/muldf3.c GCC/gcc/config/soft-fp/muldf3.c
--- GCC.orig/gcc/config/soft-fp/muldf3.c 1970-01-01 00:00:00.000000000 +0000
+++ GCC/gcc/config/soft-fp/muldf3.c 2006-02-18 02:11:04.000000000 +0000
@@ -0,0 +1,49 @@
+/* Software floating-point emulation.
+ Return a * b
+ Copyright (C) 1997,1999,2006 Free Software Foundation, Inc.
+ This file is part of GCC.
+ Contributed by Richard Henderson (rth@cygnus.com) and
+ Jakub Jelinek (jj@ultra.linux.cz).
+
+ GCC is free software; you can redistribute it and/or modify it under
+ the terms of the GNU General Public License as published by the Free
+ Software Foundation; either version 2, or (at your option) any later
+ version.
+
+ In addition to the permissions in the GNU 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 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.)
+
+ GCC 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 General Public License
+ for more details.
+
+ You should have received a copy of the GNU General Public License
+ along with GCC; see the file COPYING. If not, write to the Free
+ Software Foundation, 51 Franklin Street, Fifth Floor, Boston, MA
+ 02110-1301, USA. */
+
+#include "soft-fp.h"
+#include "double.h"
+
+DFtype __muldf3(DFtype a, DFtype b)
+{
+ FP_DECL_EX;
+ FP_DECL_D(A); FP_DECL_D(B); FP_DECL_D(R);
+ DFtype r;
+
+ FP_INIT_ROUNDMODE;
+ FP_UNPACK_D(A, a);
+ FP_UNPACK_D(B, b);
+ FP_MUL_D(R, A, B);
+ FP_PACK_D(r, R);
+ FP_HANDLE_EXCEPTIONS;
+
+ return r;
+}
diff -rupN GCC.orig/gcc/config/soft-fp/mulsf3.c GCC/gcc/config/soft-fp/mulsf3.c
--- GCC.orig/gcc/config/soft-fp/mulsf3.c 1970-01-01 00:00:00.000000000 +0000
+++ GCC/gcc/config/soft-fp/mulsf3.c 2006-02-18 02:11:04.000000000 +0000
@@ -0,0 +1,49 @@
+/* Software floating-point emulation.
+ Return a * b
+ Copyright (C) 1997,1999,2006 Free Software Foundation, Inc.
+ This file is part of GCC.
+ Contributed by Richard Henderson (rth@cygnus.com) and
+ Jakub Jelinek (jj@ultra.linux.cz).
+
+ GCC is free software; you can redistribute it and/or modify it under
+ the terms of the GNU General Public License as published by the Free
+ Software Foundation; either version 2, or (at your option) any later
+ version.
+
+ In addition to the permissions in the GNU 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 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.)
+
+ GCC 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 General Public License
+ for more details.
+
+ You should have received a copy of the GNU General Public License
+ along with GCC; see the file COPYING. If not, write to the Free
+ Software Foundation, 51 Franklin Street, Fifth Floor, Boston, MA
+ 02110-1301, USA. */
+
+#include "soft-fp.h"
+#include "single.h"
+
+SFtype __mulsf3(SFtype a, SFtype b)
+{
+ FP_DECL_EX;
+ FP_DECL_S(A); FP_DECL_S(B); FP_DECL_S(R);
+ SFtype r;
+
+ FP_INIT_ROUNDMODE;
+ FP_UNPACK_S(A, a);
+ FP_UNPACK_S(B, b);
+ FP_MUL_S(R, A, B);
+ FP_PACK_S(r, R);
+ FP_HANDLE_EXCEPTIONS;
+
+ return r;
+}
diff -rupN GCC.orig/gcc/config/soft-fp/multf3.c GCC/gcc/config/soft-fp/multf3.c
--- GCC.orig/gcc/config/soft-fp/multf3.c 1970-01-01 00:00:00.000000000 +0000
+++ GCC/gcc/config/soft-fp/multf3.c 2006-02-18 02:11:04.000000000 +0000
@@ -0,0 +1,49 @@
+/* Software floating-point emulation.
+ Return a * b
+ Copyright (C) 1997,1999,2006 Free Software Foundation, Inc.
+ This file is part of GCC.
+ Contributed by Richard Henderson (rth@cygnus.com) and
+ Jakub Jelinek (jj@ultra.linux.cz).
+
+ GCC is free software; you can redistribute it and/or modify it under
+ the terms of the GNU General Public License as published by the Free
+ Software Foundation; either version 2, or (at your option) any later
+ version.
+
+ In addition to the permissions in the GNU 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 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.)
+
+ GCC 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 General Public License
+ for more details.
+
+ You should have received a copy of the GNU General Public License
+ along with GCC; see the file COPYING. If not, write to the Free
+ Software Foundation, 51 Franklin Street, Fifth Floor, Boston, MA
+ 02110-1301, USA. */
+
+#include "soft-fp.h"
+#include "quad.h"
+
+TFtype __multf3(TFtype a, TFtype b)
+{
+ FP_DECL_EX;
+ FP_DECL_Q(A); FP_DECL_Q(B); FP_DECL_Q(R);
+ TFtype r;
+
+ FP_INIT_ROUNDMODE;
+ FP_UNPACK_Q(A, a);
+ FP_UNPACK_Q(B, b);
+ FP_MUL_Q(R, A, B);
+ FP_PACK_Q(r, R);
+ FP_HANDLE_EXCEPTIONS;
+
+ return r;
+}
diff -rupN GCC.orig/gcc/config/soft-fp/negdf2.c GCC/gcc/config/soft-fp/negdf2.c
--- GCC.orig/gcc/config/soft-fp/negdf2.c 1970-01-01 00:00:00.000000000 +0000
+++ GCC/gcc/config/soft-fp/negdf2.c 2006-02-18 02:11:04.000000000 +0000
@@ -0,0 +1,48 @@
+/* Software floating-point emulation.
+ Return -a
+ Copyright (C) 1997,1999,2006 Free Software Foundation, Inc.
+ This file is part of GCC.
+ Contributed by Richard Henderson (rth@cygnus.com) and
+ Jakub Jelinek (jj@ultra.linux.cz).
+
+ GCC is free software; you can redistribute it and/or modify it under
+ the terms of the GNU General Public License as published by the Free
+ Software Foundation; either version 2, or (at your option) any later
+ version.
+
+ In addition to the permissions in the GNU 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 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.)
+
+ GCC 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 General Public License
+ for more details.
+
+ You should have received a copy of the GNU General Public License
+ along with GCC; see the file COPYING. If not, write to the Free
+ Software Foundation, 51 Franklin Street, Fifth Floor, Boston, MA
+ 02110-1301, USA. */
+
+#include "soft-fp.h"
+#include "double.h"
+
+DFtype __negdf2(DFtype a)
+{
+ FP_DECL_EX;
+ FP_DECL_D(A); FP_DECL_D(R);
+ DFtype r;
+
+ FP_UNPACK_D(A, a);
+ FP_NEG_D(R, A);
+ FP_PACK_D(r, R);
+ FP_CLEAR_EXCEPTIONS;
+ FP_HANDLE_EXCEPTIONS;
+
+ return r;
+}
diff -rupN GCC.orig/gcc/config/soft-fp/negsf2.c GCC/gcc/config/soft-fp/negsf2.c
--- GCC.orig/gcc/config/soft-fp/negsf2.c 1970-01-01 00:00:00.000000000 +0000
+++ GCC/gcc/config/soft-fp/negsf2.c 2006-02-18 02:11:04.000000000 +0000
@@ -0,0 +1,48 @@
+/* Software floating-point emulation.
+ Return -a
+ Copyright (C) 1997,1999,2006 Free Software Foundation, Inc.
+ This file is part of GCC.
+ Contributed by Richard Henderson (rth@cygnus.com) and
+ Jakub Jelinek (jj@ultra.linux.cz).
+
+ GCC is free software; you can redistribute it and/or modify it under
+ the terms of the GNU General Public License as published by the Free
+ Software Foundation; either version 2, or (at your option) any later
+ version.
+
+ In addition to the permissions in the GNU 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 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.)
+
+ GCC 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 General Public License
+ for more details.
+
+ You should have received a copy of the GNU General Public License
+ along with GCC; see the file COPYING. If not, write to the Free
+ Software Foundation, 51 Franklin Street, Fifth Floor, Boston, MA
+ 02110-1301, USA. */
+
+#include "soft-fp.h"
+#include "single.h"
+
+SFtype __negsf2(SFtype a)
+{
+ FP_DECL_EX;
+ FP_DECL_S(A); FP_DECL_S(R);
+ SFtype r;
+
+ FP_UNPACK_S(A, a);
+ FP_NEG_S(R, A);
+ FP_PACK_S(r, R);
+ FP_CLEAR_EXCEPTIONS;
+ FP_HANDLE_EXCEPTIONS;
+
+ return r;
+}
diff -rupN GCC.orig/gcc/config/soft-fp/negtf2.c GCC/gcc/config/soft-fp/negtf2.c
--- GCC.orig/gcc/config/soft-fp/negtf2.c 1970-01-01 00:00:00.000000000 +0000
+++ GCC/gcc/config/soft-fp/negtf2.c 2006-02-18 02:11:04.000000000 +0000
@@ -0,0 +1,48 @@
+/* Software floating-point emulation.
+ Return -a
+ Copyright (C) 1997,1999,2006 Free Software Foundation, Inc.
+ This file is part of GCC.
+ Contributed by Richard Henderson (rth@cygnus.com) and
+ Jakub Jelinek (jj@ultra.linux.cz).
+
+ GCC is free software; you can redistribute it and/or modify it under
+ the terms of the GNU General Public License as published by the Free
+ Software Foundation; either version 2, or (at your option) any later
+ version.
+
+ In addition to the permissions in the GNU 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 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.)
+
+ GCC 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 General Public License
+ for more details.
+
+ You should have received a copy of the GNU General Public License
+ along with GCC; see the file COPYING. If not, write to the Free
+ Software Foundation, 51 Franklin Street, Fifth Floor, Boston, MA
+ 02110-1301, USA. */
+
+#include "soft-fp.h"
+#include "quad.h"
+
+TFtype __negtf2(TFtype a)
+{
+ FP_DECL_EX;
+ FP_DECL_Q(A); FP_DECL_Q(R);
+ TFtype r;
+
+ FP_UNPACK_Q(A, a);
+ FP_NEG_Q(R, A);
+ FP_PACK_Q(r, R);
+ FP_CLEAR_EXCEPTIONS;
+ FP_HANDLE_EXCEPTIONS;
+
+ return r;
+}
diff -rupN GCC.orig/gcc/config/soft-fp/op-1.h GCC/gcc/config/soft-fp/op-1.h
--- GCC.orig/gcc/config/soft-fp/op-1.h 1970-01-01 00:00:00.000000000 +0000
+++ GCC/gcc/config/soft-fp/op-1.h 2006-02-18 02:11:04.000000000 +0000
@@ -0,0 +1,302 @@
+/* Software floating-point emulation.
+ Basic one-word fraction declaration and manipulation.
+ Copyright (C) 1997,1998,1999,2006 Free Software Foundation, Inc.
+ This file is part of GCC.
+ Contributed by Richard Henderson (rth@cygnus.com),
+ Jakub Jelinek (jj@ultra.linux.cz),
+ David S. Miller (davem@redhat.com) and
+ Peter Maydell (pmaydell@chiark.greenend.org.uk).
+
+ GCC is free software; you can redistribute it and/or modify it under
+ the terms of the GNU General Public License as published by the Free
+ Software Foundation; either version 2, or (at your option) any later
+ version.
+
+ In addition to the permissions in the GNU 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 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.)
+
+ GCC 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 General Public License
+ for more details.
+
+ You should have received a copy of the GNU General Public License
+ along with GCC; see the file COPYING. If not, write to the Free
+ Software Foundation, 51 Franklin Street, Fifth Floor, Boston, MA
+ 02110-1301, USA. */
+
+#define _FP_FRAC_DECL_1(X) _FP_W_TYPE X##_f
+#define _FP_FRAC_COPY_1(D,S) (D##_f = S##_f)
+#define _FP_FRAC_SET_1(X,I) (X##_f = I)
+#define _FP_FRAC_HIGH_1(X) (X##_f)
+#define _FP_FRAC_LOW_1(X) (X##_f)
+#define _FP_FRAC_WORD_1(X,w) (X##_f)
+
+#define _FP_FRAC_ADDI_1(X,I) (X##_f += I)
+#define _FP_FRAC_SLL_1(X,N) \
+ do { \
+ if (__builtin_constant_p(N) && (N) == 1) \
+ X##_f += X##_f; \
+ else \
+ X##_f <<= (N); \
+ } while (0)
+#define _FP_FRAC_SRL_1(X,N) (X##_f >>= N)
+
+/* Right shift with sticky-lsb. */
+#define _FP_FRAC_SRST_1(X,S,N,sz) __FP_FRAC_SRST_1(X##_f, S, N, sz)
+#define _FP_FRAC_SRS_1(X,N,sz) __FP_FRAC_SRS_1(X##_f, N, sz)
+
+#define __FP_FRAC_SRST_1(X,S,N,sz) \
+do { \
+ S = (__builtin_constant_p(N) && (N) == 1 \
+ ? X & 1 : (X << (_FP_W_TYPE_SIZE - (N))) != 0); \
+ X = X >> (N); \
+} while (0)
+
+#define __FP_FRAC_SRS_1(X,N,sz) \
+ (X = (X >> (N) | (__builtin_constant_p(N) && (N) == 1 \
+ ? X & 1 : (X << (_FP_W_TYPE_SIZE - (N))) != 0)))
+
+#define _FP_FRAC_ADD_1(R,X,Y) (R##_f = X##_f + Y##_f)
+#define _FP_FRAC_SUB_1(R,X,Y) (R##_f = X##_f - Y##_f)
+#define _FP_FRAC_DEC_1(X,Y) (X##_f -= Y##_f)
+#define _FP_FRAC_CLZ_1(z, X) __FP_CLZ(z, X##_f)
+
+/* Predicates */
+#define _FP_FRAC_NEGP_1(X) ((_FP_WS_TYPE)X##_f < 0)
+#define _FP_FRAC_ZEROP_1(X) (X##_f == 0)
+#define _FP_FRAC_OVERP_1(fs,X) (X##_f & _FP_OVERFLOW_##fs)
+#define _FP_FRAC_CLEAR_OVERP_1(fs,X) (X##_f &= ~_FP_OVERFLOW_##fs)
+#define _FP_FRAC_EQ_1(X, Y) (X##_f == Y##_f)
+#define _FP_FRAC_GE_1(X, Y) (X##_f >= Y##_f)
+#define _FP_FRAC_GT_1(X, Y) (X##_f > Y##_f)
+
+#define _FP_ZEROFRAC_1 0
+#define _FP_MINFRAC_1 1
+#define _FP_MAXFRAC_1 (~(_FP_WS_TYPE)0)
+
+/*
+ * Unpack the raw bits of a native fp value. Do not classify or
+ * normalize the data.
+ */
+
+#define _FP_UNPACK_RAW_1(fs, X, val) \
+ do { \
+ union _FP_UNION_##fs _flo; _flo.flt = (val); \
+ \
+ X##_f = _flo.bits.frac; \
+ X##_e = _flo.bits.exp; \
+ X##_s = _flo.bits.sign; \
+ } while (0)
+
+#define _FP_UNPACK_RAW_1_P(fs, X, val) \
+ do { \
+ union _FP_UNION_##fs *_flo = \
+ (union _FP_UNION_##fs *)(val); \
+ \
+ X##_f = _flo->bits.frac; \
+ X##_e = _flo->bits.exp; \
+ X##_s = _flo->bits.sign; \
+ } while (0)
+
+/*
+ * Repack the raw bits of a native fp value.
+ */
+
+#define _FP_PACK_RAW_1(fs, val, X) \
+ do { \
+ union _FP_UNION_##fs _flo; \
+ \
+ _flo.bits.frac = X##_f; \
+ _flo.bits.exp = X##_e; \
+ _flo.bits.sign = X##_s; \
+ \
+ (val) = _flo.flt; \
+ } while (0)
+
+#define _FP_PACK_RAW_1_P(fs, val, X) \
+ do { \
+ union _FP_UNION_##fs *_flo = \
+ (union _FP_UNION_##fs *)(val); \
+ \
+ _flo->bits.frac = X##_f; \
+ _flo->bits.exp = X##_e; \
+ _flo->bits.sign = X##_s; \
+ } while (0)
+
+
+/*
+ * Multiplication algorithms:
+ */
+
+/* Basic. Assuming the host word size is >= 2*FRACBITS, we can do the
+ multiplication immediately. */
+
+#define _FP_MUL_MEAT_1_imm(wfracbits, R, X, Y) \
+ do { \
+ R##_f = X##_f * Y##_f; \
+ /* Normalize since we know where the msb of the multiplicands \
+ were (bit B), we know that the msb of the of the product is \
+ at either 2B or 2B-1. */ \
+ _FP_FRAC_SRS_1(R, wfracbits-1, 2*wfracbits); \
+ } while (0)
+
+/* Given a 1W * 1W => 2W primitive, do the extended multiplication. */
+
+#define _FP_MUL_MEAT_1_wide(wfracbits, R, X, Y, doit) \
+ do { \
+ _FP_W_TYPE _Z_f0, _Z_f1; \
+ doit(_Z_f1, _Z_f0, X##_f, Y##_f); \
+ /* Normalize since we know where the msb of the multiplicands \
+ were (bit B), we know that the msb of the of the product is \
+ at either 2B or 2B-1. */ \
+ _FP_FRAC_SRS_2(_Z, wfracbits-1, 2*wfracbits); \
+ R##_f = _Z_f0; \
+ } while (0)
+
+/* Finally, a simple widening multiply algorithm. What fun! */
+
+#define _FP_MUL_MEAT_1_hard(wfracbits, R, X, Y) \
+ do { \
+ _FP_W_TYPE _xh, _xl, _yh, _yl, _z_f0, _z_f1, _a_f0, _a_f1; \
+ \
+ /* split the words in half */ \
+ _xh = X##_f >> (_FP_W_TYPE_SIZE/2); \
+ _xl = X##_f & (((_FP_W_TYPE)1 << (_FP_W_TYPE_SIZE/2)) - 1); \
+ _yh = Y##_f >> (_FP_W_TYPE_SIZE/2); \
+ _yl = Y##_f & (((_FP_W_TYPE)1 << (_FP_W_TYPE_SIZE/2)) - 1); \
+ \
+ /* multiply the pieces */ \
+ _z_f0 = _xl * _yl; \
+ _a_f0 = _xh * _yl; \
+ _a_f1 = _xl * _yh; \
+ _z_f1 = _xh * _yh; \
+ \
+ /* reassemble into two full words */ \
+ if ((_a_f0 += _a_f1) < _a_f1) \
+ _z_f1 += (_FP_W_TYPE)1 << (_FP_W_TYPE_SIZE/2); \
+ _a_f1 = _a_f0 >> (_FP_W_TYPE_SIZE/2); \
+ _a_f0 = _a_f0 << (_FP_W_TYPE_SIZE/2); \
+ _FP_FRAC_ADD_2(_z, _z, _a); \
+ \
+ /* normalize */ \
+ _FP_FRAC_SRS_2(_z, wfracbits - 1, 2*wfracbits); \
+ R##_f = _z_f0; \
+ } while (0)
+
+
+/*
+ * Division algorithms:
+ */
+
+/* Basic. Assuming the host word size is >= 2*FRACBITS, we can do the
+ division immediately. Give this macro either _FP_DIV_HELP_imm for
+ C primitives or _FP_DIV_HELP_ldiv for the ISO function. Which you
+ choose will depend on what the compiler does with divrem4. */
+
+#define _FP_DIV_MEAT_1_imm(fs, R, X, Y, doit) \
+ do { \
+ _FP_W_TYPE _q, _r; \
+ X##_f <<= (X##_f < Y##_f \
+ ? R##_e--, _FP_WFRACBITS_##fs \
+ : _FP_WFRACBITS_##fs - 1); \
+ doit(_q, _r, X##_f, Y##_f); \
+ R##_f = _q | (_r != 0); \
+ } while (0)
+
+/* GCC's longlong.h defines a 2W / 1W => (1W,1W) primitive udiv_qrnnd
+ that may be useful in this situation. This first is for a primitive
+ that requires normalization, the second for one that does not. Look
+ for UDIV_NEEDS_NORMALIZATION to tell which your machine needs. */
+
+#define _FP_DIV_MEAT_1_udiv_norm(fs, R, X, Y) \
+ do { \
+ _FP_W_TYPE _nh, _nl, _q, _r, _y; \
+ \
+ /* Normalize Y -- i.e. make the most significant bit set. */ \
+ _y = Y##_f << _FP_WFRACXBITS_##fs; \
+ \
+ /* Shift X op correspondingly high, that is, up one full word. */ \
+ if (X##_f < Y##_f) \
+ { \
+ R##_e--; \
+ _nl = 0; \
+ _nh = X##_f; \
+ } \
+ else \
+ { \
+ _nl = X##_f << (_FP_W_TYPE_SIZE - 1); \
+ _nh = X##_f >> 1; \
+ } \
+ \
+ udiv_qrnnd(_q, _r, _nh, _nl, _y); \
+ R##_f = _q | (_r != 0); \
+ } while (0)
+
+#define _FP_DIV_MEAT_1_udiv(fs, R, X, Y) \
+ do { \
+ _FP_W_TYPE _nh, _nl, _q, _r; \
+ if (X##_f < Y##_f) \
+ { \
+ R##_e--; \
+ _nl = X##_f << _FP_WFRACBITS_##fs; \
+ _nh = X##_f >> _FP_WFRACXBITS_##fs; \
+ } \
+ else \
+ { \
+ _nl = X##_f << (_FP_WFRACBITS_##fs - 1); \
+ _nh = X##_f >> (_FP_WFRACXBITS_##fs + 1); \
+ } \
+ udiv_qrnnd(_q, _r, _nh, _nl, Y##_f); \
+ R##_f = _q | (_r != 0); \
+ } while (0)
+
+
+/*
+ * Square root algorithms:
+ * We have just one right now, maybe Newton approximation
+ * should be added for those machines where division is fast.
+ */
+
+#define _FP_SQRT_MEAT_1(R, S, T, X, q) \
+ do { \
+ while (q != _FP_WORK_ROUND) \
+ { \
+ T##_f = S##_f + q; \
+ if (T##_f <= X##_f) \
+ { \
+ S##_f = T##_f + q; \
+ X##_f -= T##_f; \
+ R##_f += q; \
+ } \
+ _FP_FRAC_SLL_1(X, 1); \
+ q >>= 1; \
+ } \
+ if (X##_f) \
+ { \
+ if (S##_f < X##_f) \
+ R##_f |= _FP_WORK_ROUND; \
+ R##_f |= _FP_WORK_STICKY; \
+ } \
+ } while (0)
+
+/*
+ * Assembly/disassembly for converting to/from integral types.
+ * No shifting or overflow handled here.
+ */
+
+#define _FP_FRAC_ASSEMBLE_1(r, X, rsize) (r = X##_f)
+#define _FP_FRAC_DISASSEMBLE_1(X, r, rsize) (X##_f = r)
+
+
+/*
+ * Convert FP values between word sizes
+ */
+
+#define _FP_FRAC_COPY_1_1(D, S) (D##_f = S##_f)
diff -rupN GCC.orig/gcc/config/soft-fp/op-2.h GCC/gcc/config/soft-fp/op-2.h
--- GCC.orig/gcc/config/soft-fp/op-2.h 1970-01-01 00:00:00.000000000 +0000
+++ GCC/gcc/config/soft-fp/op-2.h 2006-02-18 02:11:04.000000000 +0000
@@ -0,0 +1,615 @@
+/* Software floating-point emulation.
+ Basic two-word fraction declaration and manipulation.
+ Copyright (C) 1997,1998,1999,2006 Free Software Foundation, Inc.
+ This file is part of GCC.
+ Contributed by Richard Henderson (rth@cygnus.com),
+ Jakub Jelinek (jj@ultra.linux.cz),
+ David S. Miller (davem@redhat.com) and
+ Peter Maydell (pmaydell@chiark.greenend.org.uk).
+
+ GCC is free software; you can redistribute it and/or modify it under
+ the terms of the GNU General Public License as published by the Free
+ Software Foundation; either version 2, or (at your option) any later
+ version.
+
+ In addition to the permissions in the GNU 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 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.)
+
+ GCC 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 General Public License
+ for more details.
+
+ You should have received a copy of the GNU General Public License
+ along with GCC; see the file COPYING. If not, write to the Free
+ Software Foundation, 51 Franklin Street, Fifth Floor, Boston, MA
+ 02110-1301, USA. */
+
+#define _FP_FRAC_DECL_2(X) _FP_W_TYPE X##_f0, X##_f1
+#define _FP_FRAC_COPY_2(D,S) (D##_f0 = S##_f0, D##_f1 = S##_f1)
+#define _FP_FRAC_SET_2(X,I) __FP_FRAC_SET_2(X, I)
+#define _FP_FRAC_HIGH_2(X) (X##_f1)
+#define _FP_FRAC_LOW_2(X) (X##_f0)
+#define _FP_FRAC_WORD_2(X,w) (X##_f##w)
+
+#define _FP_FRAC_SLL_2(X,N) \
+(void)(((N) < _FP_W_TYPE_SIZE) \
+ ? ({ \
+ if (__builtin_constant_p(N) && (N) == 1) \
+ { \
+ X##_f1 = X##_f1 + X##_f1 + (((_FP_WS_TYPE)(X##_f0)) < 0); \
+ X##_f0 += X##_f0; \
+ } \
+ else \
+ { \
+ X##_f1 = X##_f1 << (N) | X##_f0 >> (_FP_W_TYPE_SIZE - (N)); \
+ X##_f0 <<= (N); \
+ } \
+ 0; \
+ }) \
+ : ({ \
+ X##_f1 = X##_f0 << ((N) - _FP_W_TYPE_SIZE); \
+ X##_f0 = 0; \
+ }))
+
+
+#define _FP_FRAC_SRL_2(X,N) \
+(void)(((N) < _FP_W_TYPE_SIZE) \
+ ? ({ \
+ X##_f0 = X##_f0 >> (N) | X##_f1 << (_FP_W_TYPE_SIZE - (N)); \
+ X##_f1 >>= (N); \
+ }) \
+ : ({ \
+ X##_f0 = X##_f1 >> ((N) - _FP_W_TYPE_SIZE); \
+ X##_f1 = 0; \
+ }))
+
+/* Right shift with sticky-lsb. */
+#define _FP_FRAC_SRST_2(X,S, N,sz) \
+(void)(((N) < _FP_W_TYPE_SIZE) \
+ ? ({ \
+ S = (__builtin_constant_p(N) && (N) == 1 \
+ ? X##_f0 & 1 \
+ : (X##_f0 << (_FP_W_TYPE_SIZE - (N))) != 0); \
+ X##_f0 = (X##_f1 << (_FP_W_TYPE_SIZE - (N)) | X##_f0 >> (N)); \
+ X##_f1 >>= (N); \
+ }) \
+ : ({ \
+ S = ((((N) == _FP_W_TYPE_SIZE \
+ ? 0 \
+ : (X##_f1 << (2*_FP_W_TYPE_SIZE - (N)))) \
+ | X##_f0) != 0); \
+ X##_f0 = (X##_f1 >> ((N) - _FP_W_TYPE_SIZE)); \
+ X##_f1 = 0; \
+ }))
+
+#define _FP_FRAC_SRS_2(X,N,sz) \
+(void)(((N) < _FP_W_TYPE_SIZE) \
+ ? ({ \
+ X##_f0 = (X##_f1 << (_FP_W_TYPE_SIZE - (N)) | X##_f0 >> (N) | \
+ (__builtin_constant_p(N) && (N) == 1 \
+ ? X##_f0 & 1 \
+ : (X##_f0 << (_FP_W_TYPE_SIZE - (N))) != 0)); \
+ X##_f1 >>= (N); \
+ }) \
+ : ({ \
+ X##_f0 = (X##_f1 >> ((N) - _FP_W_TYPE_SIZE) | \
+ ((((N) == _FP_W_TYPE_SIZE \
+ ? 0 \
+ : (X##_f1 << (2*_FP_W_TYPE_SIZE - (N)))) \
+ | X##_f0) != 0)); \
+ X##_f1 = 0; \
+ }))
+
+#define _FP_FRAC_ADDI_2(X,I) \
+ __FP_FRAC_ADDI_2(X##_f1, X##_f0, I)
+
+#define _FP_FRAC_ADD_2(R,X,Y) \
+ __FP_FRAC_ADD_2(R##_f1, R##_f0, X##_f1, X##_f0, Y##_f1, Y##_f0)
+
+#define _FP_FRAC_SUB_2(R,X,Y) \
+ __FP_FRAC_SUB_2(R##_f1, R##_f0, X##_f1, X##_f0, Y##_f1, Y##_f0)
+
+#define _FP_FRAC_DEC_2(X,Y) \
+ __FP_FRAC_DEC_2(X##_f1, X##_f0, Y##_f1, Y##_f0)
+
+#define _FP_FRAC_CLZ_2(R,X) \
+ do { \
+ if (X##_f1) \
+ __FP_CLZ(R,X##_f1); \
+ else \
+ { \
+ __FP_CLZ(R,X##_f0); \
+ R += _FP_W_TYPE_SIZE; \
+ } \
+ } while(0)
+
+/* Predicates */
+#define _FP_FRAC_NEGP_2(X) ((_FP_WS_TYPE)X##_f1 < 0)
+#define _FP_FRAC_ZEROP_2(X) ((X##_f1 | X##_f0) == 0)
+#define _FP_FRAC_OVERP_2(fs,X) (_FP_FRAC_HIGH_##fs(X) & _FP_OVERFLOW_##fs)
+#define _FP_FRAC_CLEAR_OVERP_2(fs,X) (_FP_FRAC_HIGH_##fs(X) &= ~_FP_OVERFLOW_##fs)
+#define _FP_FRAC_EQ_2(X, Y) (X##_f1 == Y##_f1 && X##_f0 == Y##_f0)
+#define _FP_FRAC_GT_2(X, Y) \
+ (X##_f1 > Y##_f1 || (X##_f1 == Y##_f1 && X##_f0 > Y##_f0))
+#define _FP_FRAC_GE_2(X, Y) \
+ (X##_f1 > Y##_f1 || (X##_f1 == Y##_f1 && X##_f0 >= Y##_f0))
+
+#define _FP_ZEROFRAC_2 0, 0
+#define _FP_MINFRAC_2 0, 1
+#define _FP_MAXFRAC_2 (~(_FP_WS_TYPE)0), (~(_FP_WS_TYPE)0)
+
+/*
+ * Internals
+ */
+
+#define __FP_FRAC_SET_2(X,I1,I0) (X##_f0 = I0, X##_f1 = I1)
+
+#define __FP_CLZ_2(R, xh, xl) \
+ do { \
+ if (xh) \
+ __FP_CLZ(R,xh); \
+ else \
+ { \
+ __FP_CLZ(R,xl); \
+ R += _FP_W_TYPE_SIZE; \
+ } \
+ } while(0)
+
+#if 0
+
+#ifndef __FP_FRAC_ADDI_2
+#define __FP_FRAC_ADDI_2(xh, xl, i) \
+ (xh += ((xl += i) < i))
+#endif
+#ifndef __FP_FRAC_ADD_2
+#define __FP_FRAC_ADD_2(rh, rl, xh, xl, yh, yl) \
+ (rh = xh + yh + ((rl = xl + yl) < xl))
+#endif
+#ifndef __FP_FRAC_SUB_2
+#define __FP_FRAC_SUB_2(rh, rl, xh, xl, yh, yl) \
+ (rh = xh - yh - ((rl = xl - yl) > xl))
+#endif
+#ifndef __FP_FRAC_DEC_2
+#define __FP_FRAC_DEC_2(xh, xl, yh, yl) \
+ do { \
+ UWtype _t = xl; \
+ xh -= yh + ((xl -= yl) > _t); \
+ } while (0)
+#endif
+
+#else
+
+#undef __FP_FRAC_ADDI_2
+#define __FP_FRAC_ADDI_2(xh, xl, i) add_ssaaaa(xh, xl, xh, xl, 0, i)
+#undef __FP_FRAC_ADD_2
+#define __FP_FRAC_ADD_2 add_ssaaaa
+#undef __FP_FRAC_SUB_2
+#define __FP_FRAC_SUB_2 sub_ddmmss
+#undef __FP_FRAC_DEC_2
+#define __FP_FRAC_DEC_2(xh, xl, yh, yl) sub_ddmmss(xh, xl, xh, xl, yh, yl)
+
+#endif
+
+/*
+ * Unpack the raw bits of a native fp value. Do not classify or
+ * normalize the data.
+ */
+
+#define _FP_UNPACK_RAW_2(fs, X, val) \
+ do { \
+ union _FP_UNION_##fs _flo; _flo.flt = (val); \
+ \
+ X##_f0 = _flo.bits.frac0; \
+ X##_f1 = _flo.bits.frac1; \
+ X##_e = _flo.bits.exp; \
+ X##_s = _flo.bits.sign; \
+ } while (0)
+
+#define _FP_UNPACK_RAW_2_P(fs, X, val) \
+ do { \
+ union _FP_UNION_##fs *_flo = \
+ (union _FP_UNION_##fs *)(val); \
+ \
+ X##_f0 = _flo->bits.frac0; \
+ X##_f1 = _flo->bits.frac1; \
+ X##_e = _flo->bits.exp; \
+ X##_s = _flo->bits.sign; \
+ } while (0)
+
+
+/*
+ * Repack the raw bits of a native fp value.
+ */
+
+#define _FP_PACK_RAW_2(fs, val, X) \
+ do { \
+ union _FP_UNION_##fs _flo; \
+ \
+ _flo.bits.frac0 = X##_f0; \
+ _flo.bits.frac1 = X##_f1; \
+ _flo.bits.exp = X##_e; \
+ _flo.bits.sign = X##_s; \
+ \
+ (val) = _flo.flt; \
+ } while (0)
+
+#define _FP_PACK_RAW_2_P(fs, val, X) \
+ do { \
+ union _FP_UNION_##fs *_flo = \
+ (union _FP_UNION_##fs *)(val); \
+ \
+ _flo->bits.frac0 = X##_f0; \
+ _flo->bits.frac1 = X##_f1; \
+ _flo->bits.exp = X##_e; \
+ _flo->bits.sign = X##_s; \
+ } while (0)
+
+
+/*
+ * Multiplication algorithms:
+ */
+
+/* Given a 1W * 1W => 2W primitive, do the extended multiplication. */
+
+#define _FP_MUL_MEAT_2_wide(wfracbits, R, X, Y, doit) \
+ do { \
+ _FP_FRAC_DECL_4(_z); _FP_FRAC_DECL_2(_b); _FP_FRAC_DECL_2(_c); \
+ \
+ doit(_FP_FRAC_WORD_4(_z,1), _FP_FRAC_WORD_4(_z,0), X##_f0, Y##_f0); \
+ doit(_b_f1, _b_f0, X##_f0, Y##_f1); \
+ doit(_c_f1, _c_f0, X##_f1, Y##_f0); \
+ doit(_FP_FRAC_WORD_4(_z,3), _FP_FRAC_WORD_4(_z,2), X##_f1, Y##_f1); \
+ \
+ __FP_FRAC_ADD_3(_FP_FRAC_WORD_4(_z,3),_FP_FRAC_WORD_4(_z,2), \
+ _FP_FRAC_WORD_4(_z,1), 0, _b_f1, _b_f0, \
+ _FP_FRAC_WORD_4(_z,3),_FP_FRAC_WORD_4(_z,2), \
+ _FP_FRAC_WORD_4(_z,1)); \
+ __FP_FRAC_ADD_3(_FP_FRAC_WORD_4(_z,3),_FP_FRAC_WORD_4(_z,2), \
+ _FP_FRAC_WORD_4(_z,1), 0, _c_f1, _c_f0, \
+ _FP_FRAC_WORD_4(_z,3),_FP_FRAC_WORD_4(_z,2), \
+ _FP_FRAC_WORD_4(_z,1)); \
+ \
+ /* Normalize since we know where the msb of the multiplicands \
+ were (bit B), we know that the msb of the of the product is \
+ at either 2B or 2B-1. */ \
+ _FP_FRAC_SRS_4(_z, wfracbits-1, 2*wfracbits); \
+ R##_f0 = _FP_FRAC_WORD_4(_z,0); \
+ R##_f1 = _FP_FRAC_WORD_4(_z,1); \
+ } while (0)
+
+/* Given a 1W * 1W => 2W primitive, do the extended multiplication.
+ Do only 3 multiplications instead of four. This one is for machines
+ where multiplication is much more expensive than subtraction. */
+
+#define _FP_MUL_MEAT_2_wide_3mul(wfracbits, R, X, Y, doit) \
+ do { \
+ _FP_FRAC_DECL_4(_z); _FP_FRAC_DECL_2(_b); _FP_FRAC_DECL_2(_c); \
+ _FP_W_TYPE _d; \
+ int _c1, _c2; \
+ \
+ _b_f0 = X##_f0 + X##_f1; \
+ _c1 = _b_f0 < X##_f0; \
+ _b_f1 = Y##_f0 + Y##_f1; \
+ _c2 = _b_f1 < Y##_f0; \
+ doit(_d, _FP_FRAC_WORD_4(_z,0), X##_f0, Y##_f0); \
+ doit(_FP_FRAC_WORD_4(_z,2), _FP_FRAC_WORD_4(_z,1), _b_f0, _b_f1); \
+ doit(_c_f1, _c_f0, X##_f1, Y##_f1); \
+ \
+ _b_f0 &= -_c2; \
+ _b_f1 &= -_c1; \
+ __FP_FRAC_ADD_3(_FP_FRAC_WORD_4(_z,3),_FP_FRAC_WORD_4(_z,2), \
+ _FP_FRAC_WORD_4(_z,1), (_c1 & _c2), 0, _d, \
+ 0, _FP_FRAC_WORD_4(_z,2), _FP_FRAC_WORD_4(_z,1)); \
+ __FP_FRAC_ADDI_2(_FP_FRAC_WORD_4(_z,3),_FP_FRAC_WORD_4(_z,2), \
+ _b_f0); \
+ __FP_FRAC_ADDI_2(_FP_FRAC_WORD_4(_z,3),_FP_FRAC_WORD_4(_z,2), \
+ _b_f1); \
+ __FP_FRAC_DEC_3(_FP_FRAC_WORD_4(_z,3),_FP_FRAC_WORD_4(_z,2), \
+ _FP_FRAC_WORD_4(_z,1), \
+ 0, _d, _FP_FRAC_WORD_4(_z,0)); \
+ __FP_FRAC_DEC_3(_FP_FRAC_WORD_4(_z,3),_FP_FRAC_WORD_4(_z,2), \
+ _FP_FRAC_WORD_4(_z,1), 0, _c_f1, _c_f0); \
+ __FP_FRAC_ADD_2(_FP_FRAC_WORD_4(_z,3), _FP_FRAC_WORD_4(_z,2), \
+ _c_f1, _c_f0, \
+ _FP_FRAC_WORD_4(_z,3), _FP_FRAC_WORD_4(_z,2)); \
+ \
+ /* Normalize since we know where the msb of the multiplicands \
+ were (bit B), we know that the msb of the of the product is \
+ at either 2B or 2B-1. */ \
+ _FP_FRAC_SRS_4(_z, wfracbits-1, 2*wfracbits); \
+ R##_f0 = _FP_FRAC_WORD_4(_z,0); \
+ R##_f1 = _FP_FRAC_WORD_4(_z,1); \
+ } while (0)
+
+#define _FP_MUL_MEAT_2_gmp(wfracbits, R, X, Y) \
+ do { \
+ _FP_FRAC_DECL_4(_z); \
+ _FP_W_TYPE _x[2], _y[2]; \
+ _x[0] = X##_f0; _x[1] = X##_f1; \
+ _y[0] = Y##_f0; _y[1] = Y##_f1; \
+ \
+ mpn_mul_n(_z_f, _x, _y, 2); \
+ \
+ /* Normalize since we know where the msb of the multiplicands \
+ were (bit B), we know that the msb of the of the product is \
+ at either 2B or 2B-1. */ \
+ _FP_FRAC_SRS_4(_z, wfracbits-1, 2*wfracbits); \
+ R##_f0 = _z_f[0]; \
+ R##_f1 = _z_f[1]; \
+ } while (0)
+
+/* Do at most 120x120=240 bits multiplication using double floating
+ point multiplication. This is useful if floating point
+ multiplication has much bigger throughput than integer multiply.
+ It is supposed to work for _FP_W_TYPE_SIZE 64 and wfracbits
+ between 106 and 120 only.
+ Caller guarantees that X and Y has (1LLL << (wfracbits - 1)) set.
+ SETFETZ is a macro which will disable all FPU exceptions and set rounding
+ towards zero, RESETFE should optionally reset it back. */
+
+#define _FP_MUL_MEAT_2_120_240_double(wfracbits, R, X, Y, setfetz, resetfe) \
+ do { \
+ static const double _const[] = { \
+ /* 2^-24 */ 5.9604644775390625e-08, \
+ /* 2^-48 */ 3.5527136788005009e-15, \
+ /* 2^-72 */ 2.1175823681357508e-22, \
+ /* 2^-96 */ 1.2621774483536189e-29, \
+ /* 2^28 */ 2.68435456e+08, \
+ /* 2^4 */ 1.600000e+01, \
+ /* 2^-20 */ 9.5367431640625e-07, \
+ /* 2^-44 */ 5.6843418860808015e-14, \
+ /* 2^-68 */ 3.3881317890172014e-21, \
+ /* 2^-92 */ 2.0194839173657902e-28, \
+ /* 2^-116 */ 1.2037062152420224e-35}; \
+ double _a240, _b240, _c240, _d240, _e240, _f240, \
+ _g240, _h240, _i240, _j240, _k240; \
+ union { double d; UDItype i; } _l240, _m240, _n240, _o240, \
+ _p240, _q240, _r240, _s240; \
+ UDItype _t240, _u240, _v240, _w240, _x240, _y240 = 0; \
+ \
+ if (wfracbits < 106 || wfracbits > 120) \
+ abort(); \
+ \
+ setfetz; \
+ \
+ _e240 = (double)(long)(X##_f0 & 0xffffff); \
+ _j240 = (double)(long)(Y##_f0 & 0xffffff); \
+ _d240 = (double)(long)((X##_f0 >> 24) & 0xffffff); \
+ _i240 = (double)(long)((Y##_f0 >> 24) & 0xffffff); \
+ _c240 = (double)(long)(((X##_f1 << 16) & 0xffffff) | (X##_f0 >> 48)); \
+ _h240 = (double)(long)(((Y##_f1 << 16) & 0xffffff) | (Y##_f0 >> 48)); \
+ _b240 = (double)(long)((X##_f1 >> 8) & 0xffffff); \
+ _g240 = (double)(long)((Y##_f1 >> 8) & 0xffffff); \
+ _a240 = (double)(long)(X##_f1 >> 32); \
+ _f240 = (double)(long)(Y##_f1 >> 32); \
+ _e240 *= _const[3]; \
+ _j240 *= _const[3]; \
+ _d240 *= _const[2]; \
+ _i240 *= _const[2]; \
+ _c240 *= _const[1]; \
+ _h240 *= _const[1]; \
+ _b240 *= _const[0]; \
+ _g240 *= _const[0]; \
+ _s240.d = _e240*_j240;\
+ _r240.d = _d240*_j240 + _e240*_i240;\
+ _q240.d = _c240*_j240 + _d240*_i240 + _e240*_h240;\
+ _p240.d = _b240*_j240 + _c240*_i240 + _d240*_h240 + _e240*_g240;\
+ _o240.d = _a240*_j240 + _b240*_i240 + _c240*_h240 + _d240*_g240 + _e240*_f240;\
+ _n240.d = _a240*_i240 + _b240*_h240 + _c240*_g240 + _d240*_f240; \
+ _m240.d = _a240*_h240 + _b240*_g240 + _c240*_f240; \
+ _l240.d = _a240*_g240 + _b240*_f240; \
+ _k240 = _a240*_f240; \
+ _r240.d += _s240.d; \
+ _q240.d += _r240.d; \
+ _p240.d += _q240.d; \
+ _o240.d += _p240.d; \
+ _n240.d += _o240.d; \
+ _m240.d += _n240.d; \
+ _l240.d += _m240.d; \
+ _k240 += _l240.d; \
+ _s240.d -= ((_const[10]+_s240.d)-_const[10]); \
+ _r240.d -= ((_const[9]+_r240.d)-_const[9]); \
+ _q240.d -= ((_const[8]+_q240.d)-_const[8]); \
+ _p240.d -= ((_const[7]+_p240.d)-_const[7]); \
+ _o240.d += _const[7]; \
+ _n240.d += _const[6]; \
+ _m240.d += _const[5]; \
+ _l240.d += _const[4]; \
+ if (_s240.d != 0.0) _y240 = 1; \
+ if (_r240.d != 0.0) _y240 = 1; \
+ if (_q240.d != 0.0) _y240 = 1; \
+ if (_p240.d != 0.0) _y240 = 1; \
+ _t240 = (DItype)_k240; \
+ _u240 = _l240.i; \
+ _v240 = _m240.i; \
+ _w240 = _n240.i; \
+ _x240 = _o240.i; \
+ R##_f1 = (_t240 << (128 - (wfracbits - 1))) \
+ | ((_u240 & 0xffffff) >> ((wfracbits - 1) - 104)); \
+ R##_f0 = ((_u240 & 0xffffff) << (168 - (wfracbits - 1))) \
+ | ((_v240 & 0xffffff) << (144 - (wfracbits - 1))) \
+ | ((_w240 & 0xffffff) << (120 - (wfracbits - 1))) \
+ | ((_x240 & 0xffffff) >> ((wfracbits - 1) - 96)) \
+ | _y240; \
+ resetfe; \
+ } while (0)
+
+/*
+ * Division algorithms:
+ */
+
+#define _FP_DIV_MEAT_2_udiv(fs, R, X, Y) \
+ do { \
+ _FP_W_TYPE _n_f2, _n_f1, _n_f0, _r_f1, _r_f0, _m_f1, _m_f0; \
+ if (_FP_FRAC_GT_2(X, Y)) \
+ { \
+ _n_f2 = X##_f1 >> 1; \
+ _n_f1 = X##_f1 << (_FP_W_TYPE_SIZE - 1) | X##_f0 >> 1; \
+ _n_f0 = X##_f0 << (_FP_W_TYPE_SIZE - 1); \
+ } \
+ else \
+ { \
+ R##_e--; \
+ _n_f2 = X##_f1; \
+ _n_f1 = X##_f0; \
+ _n_f0 = 0; \
+ } \
+ \
+ /* Normalize, i.e. make the most significant bit of the \
+ denominator set. */ \
+ _FP_FRAC_SLL_2(Y, _FP_WFRACXBITS_##fs); \
+ \
+ udiv_qrnnd(R##_f1, _r_f1, _n_f2, _n_f1, Y##_f1); \
+ umul_ppmm(_m_f1, _m_f0, R##_f1, Y##_f0); \
+ _r_f0 = _n_f0; \
+ if (_FP_FRAC_GT_2(_m, _r)) \
+ { \
+ R##_f1--; \
+ _FP_FRAC_ADD_2(_r, Y, _r); \
+ if (_FP_FRAC_GE_2(_r, Y) && _FP_FRAC_GT_2(_m, _r)) \
+ { \
+ R##_f1--; \
+ _FP_FRAC_ADD_2(_r, Y, _r); \
+ } \
+ } \
+ _FP_FRAC_DEC_2(_r, _m); \
+ \
+ if (_r_f1 == Y##_f1) \
+ { \
+ /* This is a special case, not an optimization \
+ (_r/Y##_f1 would not fit into UWtype). \
+ As _r is guaranteed to be < Y, R##_f0 can be either \
+ (UWtype)-1 or (UWtype)-2. But as we know what kind \
+ of bits it is (sticky, guard, round), we don't care. \
+ We also don't care what the reminder is, because the \
+ guard bit will be set anyway. -jj */ \
+ R##_f0 = -1; \
+ } \
+ else \
+ { \
+ udiv_qrnnd(R##_f0, _r_f1, _r_f1, _r_f0, Y##_f1); \
+ umul_ppmm(_m_f1, _m_f0, R##_f0, Y##_f0); \
+ _r_f0 = 0; \
+ if (_FP_FRAC_GT_2(_m, _r)) \
+ { \
+ R##_f0--; \
+ _FP_FRAC_ADD_2(_r, Y, _r); \
+ if (_FP_FRAC_GE_2(_r, Y) && _FP_FRAC_GT_2(_m, _r)) \
+ { \
+ R##_f0--; \
+ _FP_FRAC_ADD_2(_r, Y, _r); \
+ } \
+ } \
+ if (!_FP_FRAC_EQ_2(_r, _m)) \
+ R##_f0 |= _FP_WORK_STICKY; \
+ } \
+ } while (0)
+
+
+#define _FP_DIV_MEAT_2_gmp(fs, R, X, Y) \
+ do { \
+ _FP_W_TYPE _x[4], _y[2], _z[4]; \
+ _y[0] = Y##_f0; _y[1] = Y##_f1; \
+ _x[0] = _x[3] = 0; \
+ if (_FP_FRAC_GT_2(X, Y)) \
+ { \
+ R##_e++; \
+ _x[1] = (X##_f0 << (_FP_WFRACBITS_##fs-1 - _FP_W_TYPE_SIZE) | \
+ X##_f1 >> (_FP_W_TYPE_SIZE - \
+ (_FP_WFRACBITS_##fs-1 - _FP_W_TYPE_SIZE))); \
+ _x[2] = X##_f1 << (_FP_WFRACBITS_##fs-1 - _FP_W_TYPE_SIZE); \
+ } \
+ else \
+ { \
+ _x[1] = (X##_f0 << (_FP_WFRACBITS_##fs - _FP_W_TYPE_SIZE) | \
+ X##_f1 >> (_FP_W_TYPE_SIZE - \
+ (_FP_WFRACBITS_##fs - _FP_W_TYPE_SIZE))); \
+ _x[2] = X##_f1 << (_FP_WFRACBITS_##fs - _FP_W_TYPE_SIZE); \
+ } \
+ \
+ (void) mpn_divrem (_z, 0, _x, 4, _y, 2); \
+ R##_f1 = _z[1]; \
+ R##_f0 = _z[0] | ((_x[0] | _x[1]) != 0); \
+ } while (0)
+
+
+/*
+ * Square root algorithms:
+ * We have just one right now, maybe Newton approximation
+ * should be added for those machines where division is fast.
+ */
+
+#define _FP_SQRT_MEAT_2(R, S, T, X, q) \
+ do { \
+ while (q) \
+ { \
+ T##_f1 = S##_f1 + q; \
+ if (T##_f1 <= X##_f1) \
+ { \
+ S##_f1 = T##_f1 + q; \
+ X##_f1 -= T##_f1; \
+ R##_f1 += q; \
+ } \
+ _FP_FRAC_SLL_2(X, 1); \
+ q >>= 1; \
+ } \
+ q = (_FP_W_TYPE)1 << (_FP_W_TYPE_SIZE - 1); \
+ while (q != _FP_WORK_ROUND) \
+ { \
+ T##_f0 = S##_f0 + q; \
+ T##_f1 = S##_f1; \
+ if (T##_f1 < X##_f1 || \
+ (T##_f1 == X##_f1 && T##_f0 <= X##_f0)) \
+ { \
+ S##_f0 = T##_f0 + q; \
+ S##_f1 += (T##_f0 > S##_f0); \
+ _FP_FRAC_DEC_2(X, T); \
+ R##_f0 += q; \
+ } \
+ _FP_FRAC_SLL_2(X, 1); \
+ q >>= 1; \
+ } \
+ if (X##_f0 | X##_f1) \
+ { \
+ if (S##_f1 < X##_f1 || \
+ (S##_f1 == X##_f1 && S##_f0 < X##_f0)) \
+ R##_f0 |= _FP_WORK_ROUND; \
+ R##_f0 |= _FP_WORK_STICKY; \
+ } \
+ } while (0)
+
+
+/*
+ * Assembly/disassembly for converting to/from integral types.
+ * No shifting or overflow handled here.
+ */
+
+#define _FP_FRAC_ASSEMBLE_2(r, X, rsize) \
+(void)((rsize <= _FP_W_TYPE_SIZE) \
+ ? ({ r = X##_f0; }) \
+ : ({ \
+ r = X##_f1; \
+ r <<= _FP_W_TYPE_SIZE; \
+ r += X##_f0; \
+ }))
+
+#define _FP_FRAC_DISASSEMBLE_2(X, r, rsize) \
+ do { \
+ X##_f0 = r; \
+ X##_f1 = (rsize <= _FP_W_TYPE_SIZE ? 0 : r >> _FP_W_TYPE_SIZE); \
+ } while (0)
+
+/*
+ * Convert FP values between word sizes
+ */
+
+#define _FP_FRAC_COPY_1_2(D, S) (D##_f = S##_f0)
+
+#define _FP_FRAC_COPY_2_1(D, S) ((D##_f0 = S##_f), (D##_f1 = 0))
diff -rupN GCC.orig/gcc/config/soft-fp/op-4.h GCC/gcc/config/soft-fp/op-4.h
--- GCC.orig/gcc/config/soft-fp/op-4.h 1970-01-01 00:00:00.000000000 +0000
+++ GCC/gcc/config/soft-fp/op-4.h 2006-02-18 02:11:04.000000000 +0000
@@ -0,0 +1,686 @@
+/* Software floating-point emulation.
+ Basic four-word fraction declaration and manipulation.
+ Copyright (C) 1997,1998,1999,2006 Free Software Foundation, Inc.
+ This file is part of GCC.
+ Contributed by Richard Henderson (rth@cygnus.com),
+ Jakub Jelinek (jj@ultra.linux.cz),
+ David S. Miller (davem@redhat.com) and
+ Peter Maydell (pmaydell@chiark.greenend.org.uk).
+
+ GCC is free software; you can redistribute it and/or modify it under
+ the terms of the GNU General Public License as published by the Free
+ Software Foundation; either version 2, or (at your option) any later
+ version.
+
+ In addition to the permissions in the GNU 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 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.)
+
+ GCC 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 General Public License
+ for more details.
+
+ You should have received a copy of the GNU General Public License
+ along with GCC; see the file COPYING. If not, write to the Free
+ Software Foundation, 51 Franklin Street, Fifth Floor, Boston, MA
+ 02110-1301, USA. */
+
+#define _FP_FRAC_DECL_4(X) _FP_W_TYPE X##_f[4]
+#define _FP_FRAC_COPY_4(D,S) \
+ (D##_f[0] = S##_f[0], D##_f[1] = S##_f[1], \
+ D##_f[2] = S##_f[2], D##_f[3] = S##_f[3])
+#define _FP_FRAC_SET_4(X,I) __FP_FRAC_SET_4(X, I)
+#define _FP_FRAC_HIGH_4(X) (X##_f[3])
+#define _FP_FRAC_LOW_4(X) (X##_f[0])
+#define _FP_FRAC_WORD_4(X,w) (X##_f[w])
+
+#define _FP_FRAC_SLL_4(X,N) \
+ do { \
+ _FP_I_TYPE _up, _down, _skip, _i; \
+ _skip = (N) / _FP_W_TYPE_SIZE; \
+ _up = (N) % _FP_W_TYPE_SIZE; \
+ _down = _FP_W_TYPE_SIZE - _up; \
+ if (!_up) \
+ for (_i = 3; _i >= _skip; --_i) \
+ X##_f[_i] = X##_f[_i-_skip]; \
+ else \
+ { \
+ for (_i = 3; _i > _skip; --_i) \
+ X##_f[_i] = X##_f[_i-_skip] << _up \
+ | X##_f[_i-_skip-1] >> _down; \
+ X##_f[_i--] = X##_f[0] << _up; \
+ } \
+ for (; _i >= 0; --_i) \
+ X##_f[_i] = 0; \
+ } while (0)
+
+/* This one was broken too */
+#define _FP_FRAC_SRL_4(X,N) \
+ do { \
+ _FP_I_TYPE _up, _down, _skip, _i; \
+ _skip = (N) / _FP_W_TYPE_SIZE; \
+ _down = (N) % _FP_W_TYPE_SIZE; \
+ _up = _FP_W_TYPE_SIZE - _down; \
+ if (!_down) \
+ for (_i = 0; _i <= 3-_skip; ++_i) \
+ X##_f[_i] = X##_f[_i+_skip]; \
+ else \
+ { \
+ for (_i = 0; _i < 3-_skip; ++_i) \
+ X##_f[_i] = X##_f[_i+_skip] >> _down \
+ | X##_f[_i+_skip+1] << _up; \
+ X##_f[_i++] = X##_f[3] >> _down; \
+ } \
+ for (; _i < 4; ++_i) \
+ X##_f[_i] = 0; \
+ } while (0)
+
+
+/* Right shift with sticky-lsb.
+ * What this actually means is that we do a standard right-shift,
+ * but that if any of the bits that fall off the right hand side
+ * were one then we always set the LSbit.
+ */
+#define _FP_FRAC_SRST_4(X,S,N,size) \
+ do { \
+ _FP_I_TYPE _up, _down, _skip, _i; \
+ _FP_W_TYPE _s; \
+ _skip = (N) / _FP_W_TYPE_SIZE; \
+ _down = (N) % _FP_W_TYPE_SIZE; \
+ _up = _FP_W_TYPE_SIZE - _down; \
+ for (_s = _i = 0; _i < _skip; ++_i) \
+ _s |= X##_f[_i]; \
+ if (!_down) \
+ for (_i = 0; _i <= 3-_skip; ++_i) \
+ X##_f[_i] = X##_f[_i+_skip]; \
+ else \
+ { \
+ _s |= X##_f[_i] << _up; \
+ for (_i = 0; _i < 3-_skip; ++_i) \
+ X##_f[_i] = X##_f[_i+_skip] >> _down \
+ | X##_f[_i+_skip+1] << _up; \
+ X##_f[_i++] = X##_f[3] >> _down; \
+ } \
+ for (; _i < 4; ++_i) \
+ X##_f[_i] = 0; \
+ S = (_s != 0); \
+ } while (0)
+
+#define _FP_FRAC_SRS_4(X,N,size) \
+ do { \
+ int _sticky; \
+ _FP_FRAC_SRST_4(X, _sticky, N, size); \
+ X##_f[0] |= _sticky; \
+ } while (0)
+
+#define _FP_FRAC_ADD_4(R,X,Y) \
+ __FP_FRAC_ADD_4(R##_f[3], R##_f[2], R##_f[1], R##_f[0], \
+ X##_f[3], X##_f[2], X##_f[1], X##_f[0], \
+ Y##_f[3], Y##_f[2], Y##_f[1], Y##_f[0])
+
+#define _FP_FRAC_SUB_4(R,X,Y) \
+ __FP_FRAC_SUB_4(R##_f[3], R##_f[2], R##_f[1], R##_f[0], \
+ X##_f[3], X##_f[2], X##_f[1], X##_f[0], \
+ Y##_f[3], Y##_f[2], Y##_f[1], Y##_f[0])
+
+#define _FP_FRAC_DEC_4(X,Y) \
+ __FP_FRAC_DEC_4(X##_f[3], X##_f[2], X##_f[1], X##_f[0], \
+ Y##_f[3], Y##_f[2], Y##_f[1], Y##_f[0])
+
+#define _FP_FRAC_ADDI_4(X,I) \
+ __FP_FRAC_ADDI_4(X##_f[3], X##_f[2], X##_f[1], X##_f[0], I)
+
+#define _FP_ZEROFRAC_4 0,0,0,0
+#define _FP_MINFRAC_4 0,0,0,1
+#define _FP_MAXFRAC_4 (~(_FP_WS_TYPE)0), (~(_FP_WS_TYPE)0), (~(_FP_WS_TYPE)0), (~(_FP_WS_TYPE)0)
+
+#define _FP_FRAC_ZEROP_4(X) ((X##_f[0] | X##_f[1] | X##_f[2] | X##_f[3]) == 0)
+#define _FP_FRAC_NEGP_4(X) ((_FP_WS_TYPE)X##_f[3] < 0)
+#define _FP_FRAC_OVERP_4(fs,X) (_FP_FRAC_HIGH_##fs(X) & _FP_OVERFLOW_##fs)
+#define _FP_FRAC_CLEAR_OVERP_4(fs,X) (_FP_FRAC_HIGH_##fs(X) &= ~_FP_OVERFLOW_##fs)
+
+#define _FP_FRAC_EQ_4(X,Y) \
+ (X##_f[0] == Y##_f[0] && X##_f[1] == Y##_f[1] \
+ && X##_f[2] == Y##_f[2] && X##_f[3] == Y##_f[3])
+
+#define _FP_FRAC_GT_4(X,Y) \
+ (X##_f[3] > Y##_f[3] || \
+ (X##_f[3] == Y##_f[3] && (X##_f[2] > Y##_f[2] || \
+ (X##_f[2] == Y##_f[2] && (X##_f[1] > Y##_f[1] || \
+ (X##_f[1] == Y##_f[1] && X##_f[0] > Y##_f[0]) \
+ )) \
+ )) \
+ )
+
+#define _FP_FRAC_GE_4(X,Y) \
+ (X##_f[3] > Y##_f[3] || \
+ (X##_f[3] == Y##_f[3] && (X##_f[2] > Y##_f[2] || \
+ (X##_f[2] == Y##_f[2] && (X##_f[1] > Y##_f[1] || \
+ (X##_f[1] == Y##_f[1] && X##_f[0] >= Y##_f[0]) \
+ )) \
+ )) \
+ )
+
+
+#define _FP_FRAC_CLZ_4(R,X) \
+ do { \
+ if (X##_f[3]) \
+ { \
+ __FP_CLZ(R,X##_f[3]); \
+ } \
+ else if (X##_f[2]) \
+ { \
+ __FP_CLZ(R,X##_f[2]); \
+ R += _FP_W_TYPE_SIZE; \
+ } \
+ else if (X##_f[1]) \
+ { \
+ __FP_CLZ(R,X##_f[1]); \
+ R += _FP_W_TYPE_SIZE*2; \
+ } \
+ else \
+ { \
+ __FP_CLZ(R,X##_f[0]); \
+ R += _FP_W_TYPE_SIZE*3; \
+ } \
+ } while(0)
+
+
+#define _FP_UNPACK_RAW_4(fs, X, val) \
+ do { \
+ union _FP_UNION_##fs _flo; _flo.flt = (val); \
+ X##_f[0] = _flo.bits.frac0; \
+ X##_f[1] = _flo.bits.frac1; \
+ X##_f[2] = _flo.bits.frac2; \
+ X##_f[3] = _flo.bits.frac3; \
+ X##_e = _flo.bits.exp; \
+ X##_s = _flo.bits.sign; \
+ } while (0)
+
+#define _FP_UNPACK_RAW_4_P(fs, X, val) \
+ do { \
+ union _FP_UNION_##fs *_flo = \
+ (union _FP_UNION_##fs *)(val); \
+ \
+ X##_f[0] = _flo->bits.frac0; \
+ X##_f[1] = _flo->bits.frac1; \
+ X##_f[2] = _flo->bits.frac2; \
+ X##_f[3] = _flo->bits.frac3; \
+ X##_e = _flo->bits.exp; \
+ X##_s = _flo->bits.sign; \
+ } while (0)
+
+#define _FP_PACK_RAW_4(fs, val, X) \
+ do { \
+ union _FP_UNION_##fs _flo; \
+ _flo.bits.frac0 = X##_f[0]; \
+ _flo.bits.frac1 = X##_f[1]; \
+ _flo.bits.frac2 = X##_f[2]; \
+ _flo.bits.frac3 = X##_f[3]; \
+ _flo.bits.exp = X##_e; \
+ _flo.bits.sign = X##_s; \
+ (val) = _flo.flt; \
+ } while (0)
+
+#define _FP_PACK_RAW_4_P(fs, val, X) \
+ do { \
+ union _FP_UNION_##fs *_flo = \
+ (union _FP_UNION_##fs *)(val); \
+ \
+ _flo->bits.frac0 = X##_f[0]; \
+ _flo->bits.frac1 = X##_f[1]; \
+ _flo->bits.frac2 = X##_f[2]; \
+ _flo->bits.frac3 = X##_f[3]; \
+ _flo->bits.exp = X##_e; \
+ _flo->bits.sign = X##_s; \
+ } while (0)
+
+/*
+ * Multiplication algorithms:
+ */
+
+/* Given a 1W * 1W => 2W primitive, do the extended multiplication. */
+
+#define _FP_MUL_MEAT_4_wide(wfracbits, R, X, Y, doit) \
+ do { \
+ _FP_FRAC_DECL_8(_z); _FP_FRAC_DECL_2(_b); _FP_FRAC_DECL_2(_c); \
+ _FP_FRAC_DECL_2(_d); _FP_FRAC_DECL_2(_e); _FP_FRAC_DECL_2(_f); \
+ \
+ doit(_FP_FRAC_WORD_8(_z,1), _FP_FRAC_WORD_8(_z,0), X##_f[0], Y##_f[0]); \
+ doit(_b_f1, _b_f0, X##_f[0], Y##_f[1]); \
+ doit(_c_f1, _c_f0, X##_f[1], Y##_f[0]); \
+ doit(_d_f1, _d_f0, X##_f[1], Y##_f[1]); \
+ doit(_e_f1, _e_f0, X##_f[0], Y##_f[2]); \
+ doit(_f_f1, _f_f0, X##_f[2], Y##_f[0]); \
+ __FP_FRAC_ADD_3(_FP_FRAC_WORD_8(_z,3),_FP_FRAC_WORD_8(_z,2), \
+ _FP_FRAC_WORD_8(_z,1), 0,_b_f1,_b_f0, \
+ 0,0,_FP_FRAC_WORD_8(_z,1)); \
+ __FP_FRAC_ADD_3(_FP_FRAC_WORD_8(_z,3),_FP_FRAC_WORD_8(_z,2), \
+ _FP_FRAC_WORD_8(_z,1), 0,_c_f1,_c_f0, \
+ _FP_FRAC_WORD_8(_z,3),_FP_FRAC_WORD_8(_z,2), \
+ _FP_FRAC_WORD_8(_z,1)); \
+ __FP_FRAC_ADD_3(_FP_FRAC_WORD_8(_z,4),_FP_FRAC_WORD_8(_z,3), \
+ _FP_FRAC_WORD_8(_z,2), 0,_d_f1,_d_f0, \
+ 0,_FP_FRAC_WORD_8(_z,3),_FP_FRAC_WORD_8(_z,2)); \
+ __FP_FRAC_ADD_3(_FP_FRAC_WORD_8(_z,4),_FP_FRAC_WORD_8(_z,3), \
+ _FP_FRAC_WORD_8(_z,2), 0,_e_f1,_e_f0, \
+ _FP_FRAC_WORD_8(_z,4),_FP_FRAC_WORD_8(_z,3), \
+ _FP_FRAC_WORD_8(_z,2)); \
+ __FP_FRAC_ADD_3(_FP_FRAC_WORD_8(_z,4),_FP_FRAC_WORD_8(_z,3), \
+ _FP_FRAC_WORD_8(_z,2), 0,_f_f1,_f_f0, \
+ _FP_FRAC_WORD_8(_z,4),_FP_FRAC_WORD_8(_z,3), \
+ _FP_FRAC_WORD_8(_z,2)); \
+ doit(_b_f1, _b_f0, X##_f[0], Y##_f[3]); \
+ doit(_c_f1, _c_f0, X##_f[3], Y##_f[0]); \
+ doit(_d_f1, _d_f0, X##_f[1], Y##_f[2]); \
+ doit(_e_f1, _e_f0, X##_f[2], Y##_f[1]); \
+ __FP_FRAC_ADD_3(_FP_FRAC_WORD_8(_z,5),_FP_FRAC_WORD_8(_z,4), \
+ _FP_FRAC_WORD_8(_z,3), 0,_b_f1,_b_f0, \
+ 0,_FP_FRAC_WORD_8(_z,4),_FP_FRAC_WORD_8(_z,3)); \
+ __FP_FRAC_ADD_3(_FP_FRAC_WORD_8(_z,5),_FP_FRAC_WORD_8(_z,4), \
+ _FP_FRAC_WORD_8(_z,3), 0,_c_f1,_c_f0, \
+ _FP_FRAC_WORD_8(_z,5),_FP_FRAC_WORD_8(_z,4), \
+ _FP_FRAC_WORD_8(_z,3)); \
+ __FP_FRAC_ADD_3(_FP_FRAC_WORD_8(_z,5),_FP_FRAC_WORD_8(_z,4), \
+ _FP_FRAC_WORD_8(_z,3), 0,_d_f1,_d_f0, \
+ _FP_FRAC_WORD_8(_z,5),_FP_FRAC_WORD_8(_z,4), \
+ _FP_FRAC_WORD_8(_z,3)); \
+ __FP_FRAC_ADD_3(_FP_FRAC_WORD_8(_z,5),_FP_FRAC_WORD_8(_z,4), \
+ _FP_FRAC_WORD_8(_z,3), 0,_e_f1,_e_f0, \
+ _FP_FRAC_WORD_8(_z,5),_FP_FRAC_WORD_8(_z,4), \
+ _FP_FRAC_WORD_8(_z,3)); \
+ doit(_b_f1, _b_f0, X##_f[2], Y##_f[2]); \
+ doit(_c_f1, _c_f0, X##_f[1], Y##_f[3]); \
+ doit(_d_f1, _d_f0, X##_f[3], Y##_f[1]); \
+ doit(_e_f1, _e_f0, X##_f[2], Y##_f[3]); \
+ doit(_f_f1, _f_f0, X##_f[3], Y##_f[2]); \
+ __FP_FRAC_ADD_3(_FP_FRAC_WORD_8(_z,6),_FP_FRAC_WORD_8(_z,5), \
+ _FP_FRAC_WORD_8(_z,4), 0,_b_f1,_b_f0, \
+ 0,_FP_FRAC_WORD_8(_z,5),_FP_FRAC_WORD_8(_z,4)); \
+ __FP_FRAC_ADD_3(_FP_FRAC_WORD_8(_z,6),_FP_FRAC_WORD_8(_z,5), \
+ _FP_FRAC_WORD_8(_z,4), 0,_c_f1,_c_f0, \
+ _FP_FRAC_WORD_8(_z,6),_FP_FRAC_WORD_8(_z,5), \
+ _FP_FRAC_WORD_8(_z,4)); \
+ __FP_FRAC_ADD_3(_FP_FRAC_WORD_8(_z,6),_FP_FRAC_WORD_8(_z,5), \
+ _FP_FRAC_WORD_8(_z,4), 0,_d_f1,_d_f0, \
+ _FP_FRAC_WORD_8(_z,6),_FP_FRAC_WORD_8(_z,5), \
+ _FP_FRAC_WORD_8(_z,4)); \
+ __FP_FRAC_ADD_3(_FP_FRAC_WORD_8(_z,7),_FP_FRAC_WORD_8(_z,6), \
+ _FP_FRAC_WORD_8(_z,5), 0,_e_f1,_e_f0, \
+ 0,_FP_FRAC_WORD_8(_z,6),_FP_FRAC_WORD_8(_z,5)); \
+ __FP_FRAC_ADD_3(_FP_FRAC_WORD_8(_z,7),_FP_FRAC_WORD_8(_z,6), \
+ _FP_FRAC_WORD_8(_z,5), 0,_f_f1,_f_f0, \
+ _FP_FRAC_WORD_8(_z,7),_FP_FRAC_WORD_8(_z,6), \
+ _FP_FRAC_WORD_8(_z,5)); \
+ doit(_b_f1, _b_f0, X##_f[3], Y##_f[3]); \
+ __FP_FRAC_ADD_2(_FP_FRAC_WORD_8(_z,7),_FP_FRAC_WORD_8(_z,6), \
+ _b_f1,_b_f0, \
+ _FP_FRAC_WORD_8(_z,7),_FP_FRAC_WORD_8(_z,6)); \
+ \
+ /* Normalize since we know where the msb of the multiplicands \
+ were (bit B), we know that the msb of the of the product is \
+ at either 2B or 2B-1. */ \
+ _FP_FRAC_SRS_8(_z, wfracbits-1, 2*wfracbits); \
+ __FP_FRAC_SET_4(R, _FP_FRAC_WORD_8(_z,3), _FP_FRAC_WORD_8(_z,2), \
+ _FP_FRAC_WORD_8(_z,1), _FP_FRAC_WORD_8(_z,0)); \
+ } while (0)
+
+#define _FP_MUL_MEAT_4_gmp(wfracbits, R, X, Y) \
+ do { \
+ _FP_FRAC_DECL_8(_z); \
+ \
+ mpn_mul_n(_z_f, _x_f, _y_f, 4); \
+ \
+ /* Normalize since we know where the msb of the multiplicands \
+ were (bit B), we know that the msb of the of the product is \
+ at either 2B or 2B-1. */ \
+ _FP_FRAC_SRS_8(_z, wfracbits-1, 2*wfracbits); \
+ __FP_FRAC_SET_4(R, _FP_FRAC_WORD_8(_z,3), _FP_FRAC_WORD_8(_z,2), \
+ _FP_FRAC_WORD_8(_z,1), _FP_FRAC_WORD_8(_z,0)); \
+ } while (0)
+
+/*
+ * Helper utility for _FP_DIV_MEAT_4_udiv:
+ * pppp = m * nnn
+ */
+#define umul_ppppmnnn(p3,p2,p1,p0,m,n2,n1,n0) \
+ do { \
+ UWtype _t; \
+ umul_ppmm(p1,p0,m,n0); \
+ umul_ppmm(p2,_t,m,n1); \
+ __FP_FRAC_ADDI_2(p2,p1,_t); \
+ umul_ppmm(p3,_t,m,n2); \
+ __FP_FRAC_ADDI_2(p3,p2,_t); \
+ } while (0)
+
+/*
+ * Division algorithms:
+ */
+
+#define _FP_DIV_MEAT_4_udiv(fs, R, X, Y) \
+ do { \
+ int _i; \
+ _FP_FRAC_DECL_4(_n); _FP_FRAC_DECL_4(_m); \
+ _FP_FRAC_SET_4(_n, _FP_ZEROFRAC_4); \
+ if (_FP_FRAC_GT_4(X, Y)) \
+ { \
+ _n_f[3] = X##_f[0] << (_FP_W_TYPE_SIZE - 1); \
+ _FP_FRAC_SRL_4(X, 1); \
+ } \
+ else \
+ R##_e--; \
+ \
+ /* Normalize, i.e. make the most significant bit of the \
+ denominator set. */ \
+ _FP_FRAC_SLL_4(Y, _FP_WFRACXBITS_##fs); \
+ \
+ for (_i = 3; ; _i--) \
+ { \
+ if (X##_f[3] == Y##_f[3]) \
+ { \
+ /* This is a special case, not an optimization \
+ (X##_f[3]/Y##_f[3] would not fit into UWtype). \
+ As X## is guaranteed to be < Y, R##_f[_i] can be either \
+ (UWtype)-1 or (UWtype)-2. */ \
+ R##_f[_i] = -1; \
+ if (!_i) \
+ break; \
+ __FP_FRAC_SUB_4(X##_f[3], X##_f[2], X##_f[1], X##_f[0], \
+ Y##_f[2], Y##_f[1], Y##_f[0], 0, \
+ X##_f[2], X##_f[1], X##_f[0], _n_f[_i]); \
+ _FP_FRAC_SUB_4(X, Y, X); \
+ if (X##_f[3] > Y##_f[3]) \
+ { \
+ R##_f[_i] = -2; \
+ _FP_FRAC_ADD_4(X, Y, X); \
+ } \
+ } \
+ else \
+ { \
+ udiv_qrnnd(R##_f[_i], X##_f[3], X##_f[3], X##_f[2], Y##_f[3]); \
+ umul_ppppmnnn(_m_f[3], _m_f[2], _m_f[1], _m_f[0], \
+ R##_f[_i], Y##_f[2], Y##_f[1], Y##_f[0]); \
+ X##_f[2] = X##_f[1]; \
+ X##_f[1] = X##_f[0]; \
+ X##_f[0] = _n_f[_i]; \
+ if (_FP_FRAC_GT_4(_m, X)) \
+ { \
+ R##_f[_i]--; \
+ _FP_FRAC_ADD_4(X, Y, X); \
+ if (_FP_FRAC_GE_4(X, Y) && _FP_FRAC_GT_4(_m, X)) \
+ { \
+ R##_f[_i]--; \
+ _FP_FRAC_ADD_4(X, Y, X); \
+ } \
+ } \
+ _FP_FRAC_DEC_4(X, _m); \
+ if (!_i) \
+ { \
+ if (!_FP_FRAC_EQ_4(X, _m)) \
+ R##_f[0] |= _FP_WORK_STICKY; \
+ break; \
+ } \
+ } \
+ } \
+ } while (0)
+
+
+/*
+ * Square root algorithms:
+ * We have just one right now, maybe Newton approximation
+ * should be added for those machines where division is fast.
+ */
+
+#define _FP_SQRT_MEAT_4(R, S, T, X, q) \
+ do { \
+ while (q) \
+ { \
+ T##_f[3] = S##_f[3] + q; \
+ if (T##_f[3] <= X##_f[3]) \
+ { \
+ S##_f[3] = T##_f[3] + q; \
+ X##_f[3] -= T##_f[3]; \
+ R##_f[3] += q; \
+ } \
+ _FP_FRAC_SLL_4(X, 1); \
+ q >>= 1; \
+ } \
+ q = (_FP_W_TYPE)1 << (_FP_W_TYPE_SIZE - 1); \
+ while (q) \
+ { \
+ T##_f[2] = S##_f[2] + q; \
+ T##_f[3] = S##_f[3]; \
+ if (T##_f[3] < X##_f[3] || \
+ (T##_f[3] == X##_f[3] && T##_f[2] <= X##_f[2])) \
+ { \
+ S##_f[2] = T##_f[2] + q; \
+ S##_f[3] += (T##_f[2] > S##_f[2]); \
+ __FP_FRAC_DEC_2(X##_f[3], X##_f[2], \
+ T##_f[3], T##_f[2]); \
+ R##_f[2] += q; \
+ } \
+ _FP_FRAC_SLL_4(X, 1); \
+ q >>= 1; \
+ } \
+ q = (_FP_W_TYPE)1 << (_FP_W_TYPE_SIZE - 1); \
+ while (q) \
+ { \
+ T##_f[1] = S##_f[1] + q; \
+ T##_f[2] = S##_f[2]; \
+ T##_f[3] = S##_f[3]; \
+ if (T##_f[3] < X##_f[3] || \
+ (T##_f[3] == X##_f[3] && (T##_f[2] < X##_f[2] || \
+ (T##_f[2] == X##_f[2] && T##_f[1] <= X##_f[1])))) \
+ { \
+ S##_f[1] = T##_f[1] + q; \
+ S##_f[2] += (T##_f[1] > S##_f[1]); \
+ S##_f[3] += (T##_f[2] > S##_f[2]); \
+ __FP_FRAC_DEC_3(X##_f[3], X##_f[2], X##_f[1], \
+ T##_f[3], T##_f[2], T##_f[1]); \
+ R##_f[1] += q; \
+ } \
+ _FP_FRAC_SLL_4(X, 1); \
+ q >>= 1; \
+ } \
+ q = (_FP_W_TYPE)1 << (_FP_W_TYPE_SIZE - 1); \
+ while (q != _FP_WORK_ROUND) \
+ { \
+ T##_f[0] = S##_f[0] + q; \
+ T##_f[1] = S##_f[1]; \
+ T##_f[2] = S##_f[2]; \
+ T##_f[3] = S##_f[3]; \
+ if (_FP_FRAC_GE_4(X,T)) \
+ { \
+ S##_f[0] = T##_f[0] + q; \
+ S##_f[1] += (T##_f[0] > S##_f[0]); \
+ S##_f[2] += (T##_f[1] > S##_f[1]); \
+ S##_f[3] += (T##_f[2] > S##_f[2]); \
+ _FP_FRAC_DEC_4(X, T); \
+ R##_f[0] += q; \
+ } \
+ _FP_FRAC_SLL_4(X, 1); \
+ q >>= 1; \
+ } \
+ if (!_FP_FRAC_ZEROP_4(X)) \
+ { \
+ if (_FP_FRAC_GT_4(X,S)) \
+ R##_f[0] |= _FP_WORK_ROUND; \
+ R##_f[0] |= _FP_WORK_STICKY; \
+ } \
+ } while (0)
+
+
+/*
+ * Internals
+ */
+
+#define __FP_FRAC_SET_4(X,I3,I2,I1,I0) \
+ (X##_f[3] = I3, X##_f[2] = I2, X##_f[1] = I1, X##_f[0] = I0)
+
+#ifndef __FP_FRAC_ADD_3
+#define __FP_FRAC_ADD_3(r2,r1,r0,x2,x1,x0,y2,y1,y0) \
+ do { \
+ _FP_W_TYPE _c1, _c2; \
+ r0 = x0 + y0; \
+ _c1 = r0 < x0; \
+ r1 = x1 + y1; \
+ _c2 = r1 < x1; \
+ r1 += _c1; \
+ _c2 |= r1 < _c1; \
+ r2 = x2 + y2 + _c2; \
+ } while (0)
+#endif
+
+#ifndef __FP_FRAC_ADD_4
+#define __FP_FRAC_ADD_4(r3,r2,r1,r0,x3,x2,x1,x0,y3,y2,y1,y0) \
+ do { \
+ _FP_W_TYPE _c1, _c2, _c3; \
+ r0 = x0 + y0; \
+ _c1 = r0 < x0; \
+ r1 = x1 + y1; \
+ _c2 = r1 < x1; \
+ r1 += _c1; \
+ _c2 |= r1 < _c1; \
+ r2 = x2 + y2; \
+ _c3 = r2 < x2; \
+ r2 += _c2; \
+ _c3 |= r2 < _c2; \
+ r3 = x3 + y3 + _c3; \
+ } while (0)
+#endif
+
+#ifndef __FP_FRAC_SUB_3
+#define __FP_FRAC_SUB_3(r2,r1,r0,x2,x1,x0,y2,y1,y0) \
+ do { \
+ _FP_W_TYPE _c1, _c2; \
+ r0 = x0 - y0; \
+ _c1 = r0 > x0; \
+ r1 = x1 - y1; \
+ _c2 = r1 > x1; \
+ r1 -= _c1; \
+ _c2 |= r1 > _c1; \
+ r2 = x2 - y2 - _c2; \
+ } while (0)
+#endif
+
+#ifndef __FP_FRAC_SUB_4
+#define __FP_FRAC_SUB_4(r3,r2,r1,r0,x3,x2,x1,x0,y3,y2,y1,y0) \
+ do { \
+ _FP_W_TYPE _c1, _c2, _c3; \
+ r0 = x0 - y0; \
+ _c1 = r0 > x0; \
+ r1 = x1 - y1; \
+ _c2 = r1 > x1; \
+ r1 -= _c1; \
+ _c2 |= r1 > _c1; \
+ r2 = x2 - y2; \
+ _c3 = r2 > x2; \
+ r2 -= _c2; \
+ _c3 |= r2 > _c2; \
+ r3 = x3 - y3 - _c3; \
+ } while (0)
+#endif
+
+#ifndef __FP_FRAC_DEC_3
+#define __FP_FRAC_DEC_3(x2,x1,x0,y2,y1,y0) \
+ do { \
+ UWtype _t0, _t1, _t2; \
+ _t0 = x0, _t1 = x1, _t2 = x2; \
+ __FP_FRAC_SUB_3 (x2, x1, x0, _t2, _t1, _t0, y2, y1, y0); \
+ } while (0)
+#endif
+
+#ifndef __FP_FRAC_DEC_4
+#define __FP_FRAC_DEC_4(x3,x2,x1,x0,y3,y2,y1,y0) \
+ do { \
+ UWtype _t0, _t1, _t2, _t3; \
+ _t0 = x0, _t1 = x1, _t2 = x2, _t3 = x3; \
+ __FP_FRAC_SUB_4 (x3,x2,x1,x0,_t3,_t2,_t1,_t0, y3,y2,y1,y0); \
+ } while (0)
+#endif
+
+#ifndef __FP_FRAC_ADDI_4
+#define __FP_FRAC_ADDI_4(x3,x2,x1,x0,i) \
+ do { \
+ UWtype _t; \
+ _t = ((x0 += i) < i); \
+ x1 += _t; _t = (x1 < _t); \
+ x2 += _t; _t = (x2 < _t); \
+ x3 += _t; \
+ } while (0)
+#endif
+
+/* Convert FP values between word sizes. This appears to be more
+ * complicated than I'd have expected it to be, so these might be
+ * wrong... These macros are in any case somewhat bogus because they
+ * use information about what various FRAC_n variables look like
+ * internally [eg, that 2 word vars are X_f0 and x_f1]. But so do
+ * the ones in op-2.h and op-1.h.
+ */
+#define _FP_FRAC_COPY_1_4(D, S) (D##_f = S##_f[0])
+
+#define _FP_FRAC_COPY_2_4(D, S) \
+do { \
+ D##_f0 = S##_f[0]; \
+ D##_f1 = S##_f[1]; \
+} while (0)
+
+/* Assembly/disassembly for converting to/from integral types.
+ * No shifting or overflow handled here.
+ */
+/* Put the FP value X into r, which is an integer of size rsize. */
+#define _FP_FRAC_ASSEMBLE_4(r, X, rsize) \
+ do { \
+ if (rsize <= _FP_W_TYPE_SIZE) \
+ r = X##_f[0]; \
+ else if (rsize <= 2*_FP_W_TYPE_SIZE) \
+ { \
+ r = X##_f[1]; \
+ r <<= _FP_W_TYPE_SIZE; \
+ r += X##_f[0]; \
+ } \
+ else \
+ { \
+ /* I'm feeling lazy so we deal with int == 3words (implausible)*/ \
+ /* and int == 4words as a single case. */ \
+ r = X##_f[3]; \
+ r <<= _FP_W_TYPE_SIZE; \
+ r += X##_f[2]; \
+ r <<= _FP_W_TYPE_SIZE; \
+ r += X##_f[1]; \
+ r <<= _FP_W_TYPE_SIZE; \
+ r += X##_f[0]; \
+ } \
+ } while (0)
+
+/* "No disassemble Number Five!" */
+/* move an integer of size rsize into X's fractional part. We rely on
+ * the _f[] array consisting of words of size _FP_W_TYPE_SIZE to avoid
+ * having to mask the values we store into it.
+ */
+#define _FP_FRAC_DISASSEMBLE_4(X, r, rsize) \
+ do { \
+ X##_f[0] = r; \
+ X##_f[1] = (rsize <= _FP_W_TYPE_SIZE ? 0 : r >> _FP_W_TYPE_SIZE); \
+ X##_f[2] = (rsize <= 2*_FP_W_TYPE_SIZE ? 0 : r >> 2*_FP_W_TYPE_SIZE); \
+ X##_f[3] = (rsize <= 3*_FP_W_TYPE_SIZE ? 0 : r >> 3*_FP_W_TYPE_SIZE); \
+ } while (0);
+
+#define _FP_FRAC_COPY_4_1(D, S) \
+do { \
+ D##_f[0] = S##_f; \
+ D##_f[1] = D##_f[2] = D##_f[3] = 0; \
+} while (0)
+
+#define _FP_FRAC_COPY_4_2(D, S) \
+do { \
+ D##_f[0] = S##_f0; \
+ D##_f[1] = S##_f1; \
+ D##_f[2] = D##_f[3] = 0; \
+} while (0)
diff -rupN GCC.orig/gcc/config/soft-fp/op-8.h GCC/gcc/config/soft-fp/op-8.h
--- GCC.orig/gcc/config/soft-fp/op-8.h 1970-01-01 00:00:00.000000000 +0000
+++ GCC/gcc/config/soft-fp/op-8.h 2006-02-18 02:11:04.000000000 +0000
@@ -0,0 +1,111 @@
+/* Software floating-point emulation.
+ Basic eight-word fraction declaration and manipulation.
+ Copyright (C) 1997,1998,1999,2006 Free Software Foundation, Inc.
+ This file is part of GCC.
+ Contributed by Richard Henderson (rth@cygnus.com),
+ Jakub Jelinek (jj@ultra.linux.cz) and
+ Peter Maydell (pmaydell@chiark.greenend.org.uk).
+
+ GCC is free software; you can redistribute it and/or modify it under
+ the terms of the GNU General Public License as published by the Free
+ Software Foundation; either version 2, or (at your option) any later
+ version.
+
+ In addition to the permissions in the GNU 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 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.)
+
+ GCC 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 General Public License
+ for more details.
+
+ You should have received a copy of the GNU General Public License
+ along with GCC; see the file COPYING. If not, write to the Free
+ Software Foundation, 51 Franklin Street, Fifth Floor, Boston, MA
+ 02110-1301, USA. */
+
+/* We need just a few things from here for op-4, if we ever need some
+ other macros, they can be added. */
+#define _FP_FRAC_DECL_8(X) _FP_W_TYPE X##_f[8]
+#define _FP_FRAC_HIGH_8(X) (X##_f[7])
+#define _FP_FRAC_LOW_8(X) (X##_f[0])
+#define _FP_FRAC_WORD_8(X,w) (X##_f[w])
+
+#define _FP_FRAC_SLL_8(X,N) \
+ do { \
+ _FP_I_TYPE _up, _down, _skip, _i; \
+ _skip = (N) / _FP_W_TYPE_SIZE; \
+ _up = (N) % _FP_W_TYPE_SIZE; \
+ _down = _FP_W_TYPE_SIZE - _up; \
+ if (!_up) \
+ for (_i = 7; _i >= _skip; --_i) \
+ X##_f[_i] = X##_f[_i-_skip]; \
+ else \
+ { \
+ for (_i = 7; _i > _skip; --_i) \
+ X##_f[_i] = X##_f[_i-_skip] << _up \
+ | X##_f[_i-_skip-1] >> _down; \
+ X##_f[_i--] = X##_f[0] << _up; \
+ } \
+ for (; _i >= 0; --_i) \
+ X##_f[_i] = 0; \
+ } while (0)
+
+#define _FP_FRAC_SRL_8(X,N) \
+ do { \
+ _FP_I_TYPE _up, _down, _skip, _i; \
+ _skip = (N) / _FP_W_TYPE_SIZE; \
+ _down = (N) % _FP_W_TYPE_SIZE; \
+ _up = _FP_W_TYPE_SIZE - _down; \
+ if (!_down) \
+ for (_i = 0; _i <= 7-_skip; ++_i) \
+ X##_f[_i] = X##_f[_i+_skip]; \
+ else \
+ { \
+ for (_i = 0; _i < 7-_skip; ++_i) \
+ X##_f[_i] = X##_f[_i+_skip] >> _down \
+ | X##_f[_i+_skip+1] << _up; \
+ X##_f[_i++] = X##_f[7] >> _down; \
+ } \
+ for (; _i < 8; ++_i) \
+ X##_f[_i] = 0; \
+ } while (0)
+
+
+/* Right shift with sticky-lsb.
+ * What this actually means is that we do a standard right-shift,
+ * but that if any of the bits that fall off the right hand side
+ * were one then we always set the LSbit.
+ */
+#define _FP_FRAC_SRS_8(X,N,size) \
+ do { \
+ _FP_I_TYPE _up, _down, _skip, _i; \
+ _FP_W_TYPE _s; \
+ _skip = (N) / _FP_W_TYPE_SIZE; \
+ _down = (N) % _FP_W_TYPE_SIZE; \
+ _up = _FP_W_TYPE_SIZE - _down; \
+ for (_s = _i = 0; _i < _skip; ++_i) \
+ _s |= X##_f[_i]; \
+ if (!_down) \
+ for (_i = 0; _i <= 7-_skip; ++_i) \
+ X##_f[_i] = X##_f[_i+_skip]; \
+ else \
+ { \
+ _s |= X##_f[_i] << _up; \
+ for (_i = 0; _i < 7-_skip; ++_i) \
+ X##_f[_i] = X##_f[_i+_skip] >> _down \
+ | X##_f[_i+_skip+1] << _up; \
+ X##_f[_i++] = X##_f[7] >> _down; \
+ } \
+ for (; _i < 8; ++_i) \
+ X##_f[_i] = 0; \
+ /* don't fix the LSB until the very end when we're sure f[0] is stable */ \
+ X##_f[0] |= (_s != 0); \
+ } while (0)
+
diff -rupN GCC.orig/gcc/config/soft-fp/op-common.h GCC/gcc/config/soft-fp/op-common.h
--- GCC.orig/gcc/config/soft-fp/op-common.h 1970-01-01 00:00:00.000000000 +0000
+++ GCC/gcc/config/soft-fp/op-common.h 2006-02-18 02:11:04.000000000 +0000
@@ -0,0 +1,1330 @@
+/* Software floating-point emulation. Common operations.
+ Copyright (C) 1997,1998,1999,2006 Free Software Foundation, Inc.
+ This file is part of GCC.
+ Contributed by Richard Henderson (rth@cygnus.com),
+ Jakub Jelinek (jj@ultra.linux.cz),
+ David S. Miller (davem@redhat.com) and
+ Peter Maydell (pmaydell@chiark.greenend.org.uk).
+
+ GCC is free software; you can redistribute it and/or modify it under
+ the terms of the GNU General Public License as published by the Free
+ Software Foundation; either version 2, or (at your option) any later
+ version.
+
+ In addition to the permissions in the GNU 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 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.)
+
+ GCC 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 General Public License
+ for more details.
+
+ You should have received a copy of the GNU General Public License
+ along with GCC; see the file COPYING. If not, write to the Free
+ Software Foundation, 51 Franklin Street, Fifth Floor, Boston, MA
+ 02110-1301, USA. */
+
+#define _FP_DECL(wc, X) \
+ _FP_I_TYPE X##_c __attribute__((unused)), X##_s, X##_e; \
+ _FP_FRAC_DECL_##wc(X)
+
+/*
+ * Finish truely unpacking a native fp value by classifying the kind
+ * of fp value and normalizing both the exponent and the fraction.
+ */
+
+#define _FP_UNPACK_CANONICAL(fs, wc, X) \
+do { \
+ switch (X##_e) \
+ { \
+ default: \
+ _FP_FRAC_HIGH_RAW_##fs(X) |= _FP_IMPLBIT_##fs; \
+ _FP_FRAC_SLL_##wc(X, _FP_WORKBITS); \
+ X##_e -= _FP_EXPBIAS_##fs; \
+ X##_c = FP_CLS_NORMAL; \
+ break; \
+ \
+ case 0: \
+ if (_FP_FRAC_ZEROP_##wc(X)) \
+ X##_c = FP_CLS_ZERO; \
+ else \
+ { \
+ /* a denormalized number */ \
+ _FP_I_TYPE _shift; \
+ _FP_FRAC_CLZ_##wc(_shift, X); \
+ _shift -= _FP_FRACXBITS_##fs; \
+ _FP_FRAC_SLL_##wc(X, (_shift+_FP_WORKBITS)); \
+ X##_e -= _FP_EXPBIAS_##fs - 1 + _shift; \
+ X##_c = FP_CLS_NORMAL; \
+ FP_SET_EXCEPTION(FP_EX_DENORM); \
+ } \
+ break; \
+ \
+ case _FP_EXPMAX_##fs: \
+ if (_FP_FRAC_ZEROP_##wc(X)) \
+ X##_c = FP_CLS_INF; \
+ else \
+ { \
+ X##_c = FP_CLS_NAN; \
+ /* Check for signaling NaN */ \
+ if (!(_FP_FRAC_HIGH_RAW_##fs(X) & _FP_QNANBIT_##fs)) \
+ FP_SET_EXCEPTION(FP_EX_INVALID); \
+ } \
+ break; \
+ } \
+} while (0)
+
+/* Finish unpacking an fp value in semi-raw mode: the mantissa is
+ shifted by _FP_WORKBITS but the implicit MSB is not inserted and
+ other classification is not done. */
+#define _FP_UNPACK_SEMIRAW(fs, wc, X) _FP_FRAC_SLL_##wc(X, _FP_WORKBITS)
+
+/* A semi-raw value has overflowed to infinity. Adjust the mantissa
+ and exponent appropriately. */
+#define _FP_OVERFLOW_SEMIRAW(fs, wc, X) \
+do { \
+ if (FP_ROUNDMODE == FP_RND_NEAREST \
+ || (FP_ROUNDMODE == FP_RND_PINF && !X##_s) \
+ || (FP_ROUNDMODE == FP_RND_MINF && X##_s)) \
+ { \
+ X##_e = _FP_EXPMAX_##fs; \
+ _FP_FRAC_SET_##wc(X, _FP_ZEROFRAC_##wc); \
+ } \
+ else \
+ { \
+ X##_e = _FP_EXPMAX_##fs - 1; \
+ FP_SET_EXCEPTION(FP_EX_OVERFLOW); \
+ FP_SET_EXCEPTION(FP_EX_INEXACT); \
+ _FP_FRAC_SET_##wc(X, _FP_MAXFRAC_##wc); \
+ } \
+} while (0)
+
+/* Check for a semi-raw value being a signaling NaN and raise the
+ invalid exception if so. */
+#define _FP_CHECK_SIGNAN_SEMIRAW(fs, wc, X) \
+do { \
+ if (X##_e == _FP_EXPMAX_##fs \
+ && !_FP_FRAC_ZEROP_##wc(X) \
+ && !(_FP_FRAC_HIGH_##fs(X) & _FP_QNANBIT_SH_##fs)) \
+ FP_SET_EXCEPTION(FP_EX_INVALID); \
+} while (0)
+
+/* Choose a NaN result from an operation on two semi-raw NaN
+ values. */
+#define _FP_CHOOSENAN_SEMIRAW(fs, wc, R, X, Y, OP) \
+do { \
+ /* _FP_CHOOSENAN expects raw values, so shift as required. */ \
+ _FP_FRAC_SRL_##wc(X, _FP_WORKBITS); \
+ _FP_FRAC_SRL_##wc(Y, _FP_WORKBITS); \
+ _FP_CHOOSENAN(fs, wc, R, X, Y, OP); \
+ _FP_FRAC_SLL_##wc(R, _FP_WORKBITS); \
+} while (0)
+
+/* Test whether a biased exponent is normal (not zero or maximum). */
+#define _FP_EXP_NORMAL(fs, wc, X) (((X##_e + 1) & _FP_EXPMAX_##fs) > 1)
+
+/* Prepare to pack an fp value in semi-raw mode: the mantissa is
+ rounded and shifted right, with the rounding possibly increasing
+ the exponent (including changing a finite value to infinity). */
+#define _FP_PACK_SEMIRAW(fs, wc, X) \
+do { \
+ _FP_ROUND(wc, X); \
+ if (_FP_FRAC_HIGH_##fs(X) \
+ & (_FP_OVERFLOW_##fs >> 1)) \
+ { \
+ _FP_FRAC_HIGH_##fs(X) &= ~(_FP_OVERFLOW_##fs >> 1); \
+ X##_e++; \
+ if (X##_e == _FP_EXPMAX_##fs) \
+ _FP_OVERFLOW_SEMIRAW(fs, wc, X); \
+ } \
+ _FP_FRAC_SRL_##wc(X, _FP_WORKBITS); \
+ if (!_FP_EXP_NORMAL(fs, wc, X) && !_FP_FRAC_ZEROP_##wc(X)) \
+ { \
+ if (X##_e == 0) \
+ FP_SET_EXCEPTION(FP_EX_UNDERFLOW); \
+ else \
+ { \
+ if (!_FP_KEEPNANFRACP) \
+ { \
+ _FP_FRAC_SET_##wc(X, _FP_NANFRAC_##fs); \
+ X##_s = _FP_NANSIGN_##fs; \
+ } \
+ else \
+ _FP_FRAC_HIGH_RAW_##fs(X) |= _FP_QNANBIT_##fs; \
+ } \
+ } \
+} while (0)
+
+/*
+ * Before packing the bits back into the native fp result, take care
+ * of such mundane things as rounding and overflow. Also, for some
+ * kinds of fp values, the original parts may not have been fully
+ * extracted -- but that is ok, we can regenerate them now.
+ */
+
+#define _FP_PACK_CANONICAL(fs, wc, X) \
+do { \
+ switch (X##_c) \
+ { \
+ case FP_CLS_NORMAL: \
+ X##_e += _FP_EXPBIAS_##fs; \
+ if (X##_e > 0) \
+ { \
+ _FP_ROUND(wc, X); \
+ if (_FP_FRAC_OVERP_##wc(fs, X)) \
+ { \
+ _FP_FRAC_CLEAR_OVERP_##wc(fs, X); \
+ X##_e++; \
+ } \
+ _FP_FRAC_SRL_##wc(X, _FP_WORKBITS); \
+ if (X##_e >= _FP_EXPMAX_##fs) \
+ { \
+ /* overflow */ \
+ switch (FP_ROUNDMODE) \
+ { \
+ case FP_RND_NEAREST: \
+ X##_c = FP_CLS_INF; \
+ break; \
+ case FP_RND_PINF: \
+ if (!X##_s) X##_c = FP_CLS_INF; \
+ break; \
+ case FP_RND_MINF: \
+ if (X##_s) X##_c = FP_CLS_INF; \
+ break; \
+ } \
+ if (X##_c == FP_CLS_INF) \
+ { \
+ /* Overflow to infinity */ \
+ X##_e = _FP_EXPMAX_##fs; \
+ _FP_FRAC_SET_##wc(X, _FP_ZEROFRAC_##wc); \
+ } \
+ else \
+ { \
+ /* Overflow to maximum normal */ \
+ X##_e = _FP_EXPMAX_##fs - 1; \
+ _FP_FRAC_SET_##wc(X, _FP_MAXFRAC_##wc); \
+ } \
+ FP_SET_EXCEPTION(FP_EX_OVERFLOW); \
+ FP_SET_EXCEPTION(FP_EX_INEXACT); \
+ } \
+ } \
+ else \
+ { \
+ /* we've got a denormalized number */ \
+ X##_e = -X##_e + 1; \
+ if (X##_e <= _FP_WFRACBITS_##fs) \
+ { \
+ _FP_FRAC_SRS_##wc(X, X##_e, _FP_WFRACBITS_##fs); \
+ _FP_ROUND(wc, X); \
+ if (_FP_FRAC_HIGH_##fs(X) \
+ & (_FP_OVERFLOW_##fs >> 1)) \
+ { \
+ X##_e = 1; \
+ _FP_FRAC_SET_##wc(X, _FP_ZEROFRAC_##wc); \
+ } \
+ else \
+ { \
+ X##_e = 0; \
+ _FP_FRAC_SRL_##wc(X, _FP_WORKBITS); \
+ FP_SET_EXCEPTION(FP_EX_UNDERFLOW); \
+ } \
+ } \
+ else \
+ { \
+ /* underflow to zero */ \
+ X##_e = 0; \
+ if (!_FP_FRAC_ZEROP_##wc(X)) \
+ { \
+ _FP_FRAC_SET_##wc(X, _FP_MINFRAC_##wc); \
+ _FP_ROUND(wc, X); \
+ _FP_FRAC_LOW_##wc(X) >>= (_FP_WORKBITS); \
+ } \
+ FP_SET_EXCEPTION(FP_EX_UNDERFLOW); \
+ } \
+ } \
+ break; \
+ \
+ case FP_CLS_ZERO: \
+ X##_e = 0; \
+ _FP_FRAC_SET_##wc(X, _FP_ZEROFRAC_##wc); \
+ break; \
+ \
+ case FP_CLS_INF: \
+ X##_e = _FP_EXPMAX_##fs; \
+ _FP_FRAC_SET_##wc(X, _FP_ZEROFRAC_##wc); \
+ break; \
+ \
+ case FP_CLS_NAN: \
+ X##_e = _FP_EXPMAX_##fs; \
+ if (!_FP_KEEPNANFRACP) \
+ { \
+ _FP_FRAC_SET_##wc(X, _FP_NANFRAC_##fs); \
+ X##_s = _FP_NANSIGN_##fs; \
+ } \
+ else \
+ _FP_FRAC_HIGH_RAW_##fs(X) |= _FP_QNANBIT_##fs; \
+ break; \
+ } \
+} while (0)
+
+/* This one accepts raw argument and not cooked, returns
+ * 1 if X is a signaling NaN.
+ */
+#define _FP_ISSIGNAN(fs, wc, X) \
+({ \
+ int __ret = 0; \
+ if (X##_e == _FP_EXPMAX_##fs) \
+ { \
+ if (!_FP_FRAC_ZEROP_##wc(X) \
+ && !(_FP_FRAC_HIGH_RAW_##fs(X) & _FP_QNANBIT_##fs)) \
+ __ret = 1; \
+ } \
+ __ret; \
+})
+
+
+
+
+
+/* Addition on semi-raw values. */
+#define _FP_ADD_INTERNAL(fs, wc, R, X, Y, OP) \
+do { \
+ if (X##_s == Y##_s) \
+ { \
+ /* Addition. */ \
+ R##_s = X##_s; \
+ int ediff = X##_e - Y##_e; \
+ if (ediff > 0) \
+ { \
+ R##_e = X##_e; \
+ if (Y##_e == 0) \
+ { \
+ /* Y is zero or denormalized. */ \
+ if (_FP_FRAC_ZEROP_##wc(Y)) \
+ { \
+ _FP_CHECK_SIGNAN_SEMIRAW(fs, wc, X); \
+ _FP_FRAC_COPY_##wc(R, X); \
+ goto add_done; \
+ } \
+ else \
+ { \
+ FP_SET_EXCEPTION(FP_EX_DENORM); \
+ ediff--; \
+ if (ediff == 0) \
+ { \
+ _FP_FRAC_ADD_##wc(R, X, Y); \
+ goto add3; \
+ } \
+ if (X##_e == _FP_EXPMAX_##fs) \
+ { \
+ _FP_CHECK_SIGNAN_SEMIRAW(fs, wc, X); \
+ _FP_FRAC_COPY_##wc(R, X); \
+ goto add_done; \
+ } \
+ goto add1; \
+ } \
+ } \
+ else if (X##_e == _FP_EXPMAX_##fs) \
+ { \
+ /* X is NaN or Inf, Y is normal. */ \
+ _FP_CHECK_SIGNAN_SEMIRAW(fs, wc, X); \
+ _FP_FRAC_COPY_##wc(R, X); \
+ goto add_done; \
+ } \
+ \
+ /* Insert implicit MSB of Y. */ \
+ _FP_FRAC_HIGH_##fs(Y) |= _FP_IMPLBIT_SH_##fs; \
+ \
+ add1: \
+ /* Shift the mantissa of Y to the right EDIFF steps; \
+ remember to account later for the implicit MSB of X. */ \
+ if (ediff <= _FP_WFRACBITS_##fs) \
+ _FP_FRAC_SRS_##wc(Y, ediff, _FP_WFRACBITS_##fs); \
+ else if (!_FP_FRAC_ZEROP_##wc(Y)) \
+ _FP_FRAC_SET_##wc(Y, _FP_MINFRAC_##wc); \
+ _FP_FRAC_ADD_##wc(R, X, Y); \
+ } \
+ else if (ediff < 0) \
+ { \
+ ediff = -ediff; \
+ R##_e = Y##_e; \
+ if (X##_e == 0) \
+ { \
+ /* X is zero or denormalized. */ \
+ if (_FP_FRAC_ZEROP_##wc(X)) \
+ { \
+ _FP_CHECK_SIGNAN_SEMIRAW(fs, wc, Y); \
+ _FP_FRAC_COPY_##wc(R, Y); \
+ goto add_done; \
+ } \
+ else \
+ { \
+ FP_SET_EXCEPTION(FP_EX_DENORM); \
+ ediff--; \
+ if (ediff == 0) \
+ { \
+ _FP_FRAC_ADD_##wc(R, Y, X); \
+ goto add3; \
+ } \
+ if (Y##_e == _FP_EXPMAX_##fs) \
+ { \
+ _FP_CHECK_SIGNAN_SEMIRAW(fs, wc, Y); \
+ _FP_FRAC_COPY_##wc(R, Y); \
+ goto add_done; \
+ } \
+ goto add2; \
+ } \
+ } \
+ else if (Y##_e == _FP_EXPMAX_##fs) \
+ { \
+ /* Y is NaN or Inf, X is normal. */ \
+ _FP_CHECK_SIGNAN_SEMIRAW(fs, wc, Y); \
+ _FP_FRAC_COPY_##wc(R, Y); \
+ goto add_done; \
+ } \
+ \
+ /* Insert implicit MSB of X. */ \
+ _FP_FRAC_HIGH_##fs(X) |= _FP_IMPLBIT_SH_##fs; \
+ \
+ add2: \
+ /* Shift the mantissa of X to the right EDIFF steps; \
+ remember to account later for the implicit MSB of Y. */ \
+ if (ediff <= _FP_WFRACBITS_##fs) \
+ _FP_FRAC_SRS_##wc(X, ediff, _FP_WFRACBITS_##fs); \
+ else if (!_FP_FRAC_ZEROP_##wc(X)) \
+ _FP_FRAC_SET_##wc(X, _FP_MINFRAC_##wc); \
+ _FP_FRAC_ADD_##wc(R, Y, X); \
+ } \
+ else \
+ { \
+ /* ediff == 0. */ \
+ if (!_FP_EXP_NORMAL(fs, wc, X)) \
+ { \
+ if (X##_e == 0) \
+ { \
+ /* X and Y are zero or denormalized. */ \
+ R##_e = 0; \
+ if (_FP_FRAC_ZEROP_##wc(X)) \
+ { \
+ if (!_FP_FRAC_ZEROP_##wc(Y)) \
+ FP_SET_EXCEPTION(FP_EX_DENORM); \
+ _FP_FRAC_COPY_##wc(R, Y); \
+ goto add_done; \
+ } \
+ else if (_FP_FRAC_ZEROP_##wc(Y)) \
+ { \
+ FP_SET_EXCEPTION(FP_EX_DENORM); \
+ _FP_FRAC_COPY_##wc(R, X); \
+ goto add_done; \
+ } \
+ else \
+ { \
+ FP_SET_EXCEPTION(FP_EX_DENORM); \
+ _FP_FRAC_ADD_##wc(R, X, Y); \
+ if (_FP_FRAC_HIGH_##fs(R) & _FP_IMPLBIT_SH_##fs) \
+ { \
+ /* Normalized result. */ \
+ _FP_FRAC_HIGH_##fs(R) \
+ &= ~(_FP_W_TYPE)_FP_IMPLBIT_SH_##fs; \
+ R##_e = 1; \
+ } \
+ goto add_done; \
+ } \
+ } \
+ else \
+ { \
+ /* X and Y are NaN or Inf. */ \
+ _FP_CHECK_SIGNAN_SEMIRAW(fs, wc, X); \
+ _FP_CHECK_SIGNAN_SEMIRAW(fs, wc, Y); \
+ R##_e = _FP_EXPMAX_##fs; \
+ if (_FP_FRAC_ZEROP_##wc(X)) \
+ _FP_FRAC_COPY_##wc(R, Y); \
+ else if (_FP_FRAC_ZEROP_##wc(Y)) \
+ _FP_FRAC_COPY_##wc(R, X); \
+ else \
+ _FP_CHOOSENAN_SEMIRAW(fs, wc, R, X, Y, OP); \
+ goto add_done; \
+ } \
+ } \
+ /* The exponents of X and Y, both normal, are equal. The \
+ implicit MSBs will always add to increase the \
+ exponent. */ \
+ _FP_FRAC_ADD_##wc(R, X, Y); \
+ R##_e = X##_e + 1; \
+ _FP_FRAC_SRS_##wc(R, 1, _FP_WFRACBITS_##fs); \
+ if (R##_e == _FP_EXPMAX_##fs) \
+ /* Overflow to infinity (depending on rounding mode). */ \
+ _FP_OVERFLOW_SEMIRAW(fs, wc, R); \
+ goto add_done; \
+ } \
+ add3: \
+ if (_FP_FRAC_HIGH_##fs(R) & _FP_IMPLBIT_SH_##fs) \
+ { \
+ /* Overflow. */ \
+ _FP_FRAC_HIGH_##fs(R) &= ~(_FP_W_TYPE)_FP_IMPLBIT_SH_##fs; \
+ R##_e++; \
+ _FP_FRAC_SRS_##wc(R, 1, _FP_WFRACBITS_##fs); \
+ if (R##_e == _FP_EXPMAX_##fs) \
+ /* Overflow to infinity (depending on rounding mode). */ \
+ _FP_OVERFLOW_SEMIRAW(fs, wc, R); \
+ } \
+ add_done: ; \
+ } \
+ else \
+ { \
+ /* Subtraction. */ \
+ int ediff = X##_e - Y##_e; \
+ if (ediff > 0) \
+ { \
+ R##_e = X##_e; \
+ R##_s = X##_s; \
+ if (Y##_e == 0) \
+ { \
+ /* Y is zero or denormalized. */ \
+ if (_FP_FRAC_ZEROP_##wc(Y)) \
+ { \
+ _FP_CHECK_SIGNAN_SEMIRAW(fs, wc, X); \
+ _FP_FRAC_COPY_##wc(R, X); \
+ goto sub_done; \
+ } \
+ else \
+ { \
+ FP_SET_EXCEPTION(FP_EX_DENORM); \
+ ediff--; \
+ if (ediff == 0) \
+ { \
+ _FP_FRAC_SUB_##wc(R, X, Y); \
+ goto sub3; \
+ } \
+ if (X##_e == _FP_EXPMAX_##fs) \
+ { \
+ _FP_CHECK_SIGNAN_SEMIRAW(fs, wc, X); \
+ _FP_FRAC_COPY_##wc(R, X); \
+ goto sub_done; \
+ } \
+ goto sub1; \
+ } \
+ } \
+ else if (X##_e == _FP_EXPMAX_##fs) \
+ { \
+ /* X is NaN or Inf, Y is normal. */ \
+ _FP_CHECK_SIGNAN_SEMIRAW(fs, wc, X); \
+ _FP_FRAC_COPY_##wc(R, X); \
+ goto sub_done; \
+ } \
+ \
+ /* Insert implicit MSB of Y. */ \
+ _FP_FRAC_HIGH_##fs(Y) |= _FP_IMPLBIT_SH_##fs; \
+ \
+ sub1: \
+ /* Shift the mantissa of Y to the right EDIFF steps; \
+ remember to account later for the implicit MSB of X. */ \
+ if (ediff <= _FP_WFRACBITS_##fs) \
+ _FP_FRAC_SRS_##wc(Y, ediff, _FP_WFRACBITS_##fs); \
+ else if (!_FP_FRAC_ZEROP_##wc(Y)) \
+ _FP_FRAC_SET_##wc(Y, _FP_MINFRAC_##wc); \
+ _FP_FRAC_SUB_##wc(R, X, Y); \
+ } \
+ else if (ediff < 0) \
+ { \
+ ediff = -ediff; \
+ R##_e = Y##_e; \
+ R##_s = Y##_s; \
+ if (X##_e == 0) \
+ { \
+ /* X is zero or denormalized. */ \
+ if (_FP_FRAC_ZEROP_##wc(X)) \
+ { \
+ _FP_CHECK_SIGNAN_SEMIRAW(fs, wc, Y); \
+ _FP_FRAC_COPY_##wc(R, Y); \
+ goto sub_done; \
+ } \
+ else \
+ { \
+ FP_SET_EXCEPTION(FP_EX_DENORM); \
+ ediff--; \
+ if (ediff == 0) \
+ { \
+ _FP_FRAC_SUB_##wc(R, Y, X); \
+ goto sub3; \
+ } \
+ if (Y##_e == _FP_EXPMAX_##fs) \
+ { \
+ _FP_CHECK_SIGNAN_SEMIRAW(fs, wc, Y); \
+ _FP_FRAC_COPY_##wc(R, Y); \
+ goto sub_done; \
+ } \
+ goto sub2; \
+ } \
+ } \
+ else if (Y##_e == _FP_EXPMAX_##fs) \
+ { \
+ /* Y is NaN or Inf, X is normal. */ \
+ _FP_CHECK_SIGNAN_SEMIRAW(fs, wc, Y); \
+ _FP_FRAC_COPY_##wc(R, Y); \
+ goto sub_done; \
+ } \
+ \
+ /* Insert implicit MSB of X. */ \
+ _FP_FRAC_HIGH_##fs(X) |= _FP_IMPLBIT_SH_##fs; \
+ \
+ sub2: \
+ /* Shift the mantissa of X to the right EDIFF steps; \
+ remember to account later for the implicit MSB of Y. */ \
+ if (ediff <= _FP_WFRACBITS_##fs) \
+ _FP_FRAC_SRS_##wc(X, ediff, _FP_WFRACBITS_##fs); \
+ else if (!_FP_FRAC_ZEROP_##wc(X)) \
+ _FP_FRAC_SET_##wc(X, _FP_MINFRAC_##wc); \
+ _FP_FRAC_SUB_##wc(R, Y, X); \
+ } \
+ else \
+ { \
+ /* ediff == 0. */ \
+ if (!_FP_EXP_NORMAL(fs, wc, X)) \
+ { \
+ if (X##_e == 0) \
+ { \
+ /* X and Y are zero or denormalized. */ \
+ R##_e = 0; \
+ if (_FP_FRAC_ZEROP_##wc(X)) \
+ { \
+ _FP_FRAC_COPY_##wc(R, Y); \
+ if (_FP_FRAC_ZEROP_##wc(Y)) \
+ R##_s = (FP_ROUNDMODE == FP_RND_MINF); \
+ else \
+ { \
+ FP_SET_EXCEPTION(FP_EX_DENORM); \
+ R##_s = Y##_s; \
+ } \
+ goto sub_done; \
+ } \
+ else if (_FP_FRAC_ZEROP_##wc(Y)) \
+ { \
+ FP_SET_EXCEPTION(FP_EX_DENORM); \
+ _FP_FRAC_COPY_##wc(R, X); \
+ R##_s = X##_s; \
+ goto sub_done; \
+ } \
+ else \
+ { \
+ FP_SET_EXCEPTION(FP_EX_DENORM); \
+ _FP_FRAC_SUB_##wc(R, X, Y); \
+ R##_s = X##_s; \
+ if (_FP_FRAC_HIGH_##fs(R) & _FP_IMPLBIT_SH_##fs) \
+ { \
+ /* |X| < |Y|, negate result. */ \
+ _FP_FRAC_SUB_##wc(R, Y, X); \
+ R##_s = Y##_s; \
+ } \
+ else if (_FP_FRAC_ZEROP_##wc(R)) \
+ R##_s = (FP_ROUNDMODE == FP_RND_MINF); \
+ goto sub_done; \
+ } \
+ } \
+ else \
+ { \
+ /* X and Y are NaN or Inf, of opposite signs. */ \
+ _FP_CHECK_SIGNAN_SEMIRAW(fs, wc, X); \
+ _FP_CHECK_SIGNAN_SEMIRAW(fs, wc, Y); \
+ R##_e = _FP_EXPMAX_##fs; \
+ if (_FP_FRAC_ZEROP_##wc(X)) \
+ { \
+ if (_FP_FRAC_ZEROP_##wc(Y)) \
+ { \
+ /* Inf - Inf. */ \
+ R##_s = _FP_NANSIGN_##fs; \
+ _FP_FRAC_SET_##wc(R, _FP_NANFRAC_##fs); \
+ _FP_FRAC_SLL_##wc(R, _FP_WORKBITS); \
+ FP_SET_EXCEPTION(FP_EX_INVALID); \
+ } \
+ else \
+ { \
+ /* Inf - NaN. */ \
+ R##_s = Y##_s; \
+ _FP_FRAC_COPY_##wc(R, Y); \
+ } \
+ } \
+ else \
+ { \
+ if (_FP_FRAC_ZEROP_##wc(Y)) \
+ { \
+ /* NaN - Inf. */ \
+ R##_s = X##_s; \
+ _FP_FRAC_COPY_##wc(R, X); \
+ } \
+ else \
+ { \
+ /* NaN - NaN. */ \
+ _FP_CHOOSENAN_SEMIRAW(fs, wc, R, X, Y, OP); \
+ } \
+ } \
+ goto sub_done; \
+ } \
+ } \
+ /* The exponents of X and Y, both normal, are equal. The \
+ implicit MSBs cancel. */ \
+ R##_e = X##_e; \
+ _FP_FRAC_SUB_##wc(R, X, Y); \
+ R##_s = X##_s; \
+ if (_FP_FRAC_HIGH_##fs(R) & _FP_IMPLBIT_SH_##fs) \
+ { \
+ /* |X| < |Y|, negate result. */ \
+ _FP_FRAC_SUB_##wc(R, Y, X); \
+ R##_s = Y##_s; \
+ } \
+ else if (_FP_FRAC_ZEROP_##wc(R)) \
+ { \
+ R##_e = 0; \
+ R##_s = (FP_ROUNDMODE == FP_RND_MINF); \
+ goto sub_done; \
+ } \
+ goto norm; \
+ } \
+ sub3: \
+ if (_FP_FRAC_HIGH_##fs(R) & _FP_IMPLBIT_SH_##fs) \
+ { \
+ int diff; \
+ /* Carry into most significant bit of larger one of X and Y, \
+ canceling it; renormalize. */ \
+ _FP_FRAC_HIGH_##fs(R) &= _FP_IMPLBIT_SH_##fs - 1; \
+ norm: \
+ _FP_FRAC_CLZ_##wc(diff, R); \
+ diff -= _FP_WFRACXBITS_##fs; \
+ _FP_FRAC_SLL_##wc(R, diff); \
+ if (R##_e <= diff) \
+ { \
+ /* R is denormalized. */ \
+ diff = diff - R##_e + 1; \
+ _FP_FRAC_SRS_##wc(R, diff, _FP_WFRACBITS_##fs); \
+ R##_e = 0; \
+ } \
+ else \
+ { \
+ R##_e -= diff; \
+ _FP_FRAC_HIGH_##fs(R) &= ~(_FP_W_TYPE)_FP_IMPLBIT_SH_##fs; \
+ } \
+ } \
+ sub_done: ; \
+ } \
+} while (0)
+
+#define _FP_ADD(fs, wc, R, X, Y) _FP_ADD_INTERNAL(fs, wc, R, X, Y, '+')
+#define _FP_SUB(fs, wc, R, X, Y) \
+ do { \
+ if (!(Y##_e == _FP_EXPMAX_##fs && !_FP_FRAC_ZEROP_##wc(Y))) Y##_s ^= 1; \
+ _FP_ADD_INTERNAL(fs, wc, R, X, Y, '-'); \
+ } while (0)
+
+
+/*
+ * Main negation routine. FIXME -- when we care about setting exception
+ * bits reliably, this will not do. We should examine all of the fp classes.
+ */
+
+#define _FP_NEG(fs, wc, R, X) \
+ do { \
+ _FP_FRAC_COPY_##wc(R, X); \
+ R##_c = X##_c; \
+ R##_e = X##_e; \
+ R##_s = 1 ^ X##_s; \
+ } while (0)
+
+
+/*
+ * Main multiplication routine. The input values should be cooked.
+ */
+
+#define _FP_MUL(fs, wc, R, X, Y) \
+do { \
+ R##_s = X##_s ^ Y##_s; \
+ switch (_FP_CLS_COMBINE(X##_c, Y##_c)) \
+ { \
+ case _FP_CLS_COMBINE(FP_CLS_NORMAL,FP_CLS_NORMAL): \
+ R##_c = FP_CLS_NORMAL; \
+ R##_e = X##_e + Y##_e + 1; \
+ \
+ _FP_MUL_MEAT_##fs(R,X,Y); \
+ \
+ if (_FP_FRAC_OVERP_##wc(fs, R)) \
+ _FP_FRAC_SRS_##wc(R, 1, _FP_WFRACBITS_##fs); \
+ else \
+ R##_e--; \
+ break; \
+ \
+ case _FP_CLS_COMBINE(FP_CLS_NAN,FP_CLS_NAN): \
+ _FP_CHOOSENAN(fs, wc, R, X, Y, '*'); \
+ break; \
+ \
+ case _FP_CLS_COMBINE(FP_CLS_NAN,FP_CLS_NORMAL): \
+ case _FP_CLS_COMBINE(FP_CLS_NAN,FP_CLS_INF): \
+ case _FP_CLS_COMBINE(FP_CLS_NAN,FP_CLS_ZERO): \
+ R##_s = X##_s; \
+ \
+ case _FP_CLS_COMBINE(FP_CLS_INF,FP_CLS_INF): \
+ case _FP_CLS_COMBINE(FP_CLS_INF,FP_CLS_NORMAL): \
+ case _FP_CLS_COMBINE(FP_CLS_ZERO,FP_CLS_NORMAL): \
+ case _FP_CLS_COMBINE(FP_CLS_ZERO,FP_CLS_ZERO): \
+ _FP_FRAC_COPY_##wc(R, X); \
+ R##_c = X##_c; \
+ break; \
+ \
+ case _FP_CLS_COMBINE(FP_CLS_NORMAL,FP_CLS_NAN): \
+ case _FP_CLS_COMBINE(FP_CLS_INF,FP_CLS_NAN): \
+ case _FP_CLS_COMBINE(FP_CLS_ZERO,FP_CLS_NAN): \
+ R##_s = Y##_s; \
+ \
+ case _FP_CLS_COMBINE(FP_CLS_NORMAL,FP_CLS_INF): \
+ case _FP_CLS_COMBINE(FP_CLS_NORMAL,FP_CLS_ZERO): \
+ _FP_FRAC_COPY_##wc(R, Y); \
+ R##_c = Y##_c; \
+ break; \
+ \
+ case _FP_CLS_COMBINE(FP_CLS_INF,FP_CLS_ZERO): \
+ case _FP_CLS_COMBINE(FP_CLS_ZERO,FP_CLS_INF): \
+ R##_s = _FP_NANSIGN_##fs; \
+ R##_c = FP_CLS_NAN; \
+ _FP_FRAC_SET_##wc(R, _FP_NANFRAC_##fs); \
+ FP_SET_EXCEPTION(FP_EX_INVALID); \
+ break; \
+ \
+ default: \
+ abort(); \
+ } \
+} while (0)
+
+
+/*
+ * Main division routine. The input values should be cooked.
+ */
+
+#define _FP_DIV(fs, wc, R, X, Y) \
+do { \
+ R##_s = X##_s ^ Y##_s; \
+ switch (_FP_CLS_COMBINE(X##_c, Y##_c)) \
+ { \
+ case _FP_CLS_COMBINE(FP_CLS_NORMAL,FP_CLS_NORMAL): \
+ R##_c = FP_CLS_NORMAL; \
+ R##_e = X##_e - Y##_e; \
+ \
+ _FP_DIV_MEAT_##fs(R,X,Y); \
+ break; \
+ \
+ case _FP_CLS_COMBINE(FP_CLS_NAN,FP_CLS_NAN): \
+ _FP_CHOOSENAN(fs, wc, R, X, Y, '/'); \
+ break; \
+ \
+ case _FP_CLS_COMBINE(FP_CLS_NAN,FP_CLS_NORMAL): \
+ case _FP_CLS_COMBINE(FP_CLS_NAN,FP_CLS_INF): \
+ case _FP_CLS_COMBINE(FP_CLS_NAN,FP_CLS_ZERO): \
+ R##_s = X##_s; \
+ _FP_FRAC_COPY_##wc(R, X); \
+ R##_c = X##_c; \
+ break; \
+ \
+ case _FP_CLS_COMBINE(FP_CLS_NORMAL,FP_CLS_NAN): \
+ case _FP_CLS_COMBINE(FP_CLS_INF,FP_CLS_NAN): \
+ case _FP_CLS_COMBINE(FP_CLS_ZERO,FP_CLS_NAN): \
+ R##_s = Y##_s; \
+ _FP_FRAC_COPY_##wc(R, Y); \
+ R##_c = Y##_c; \
+ break; \
+ \
+ case _FP_CLS_COMBINE(FP_CLS_NORMAL,FP_CLS_INF): \
+ case _FP_CLS_COMBINE(FP_CLS_ZERO,FP_CLS_INF): \
+ case _FP_CLS_COMBINE(FP_CLS_ZERO,FP_CLS_NORMAL): \
+ R##_c = FP_CLS_ZERO; \
+ break; \
+ \
+ case _FP_CLS_COMBINE(FP_CLS_NORMAL,FP_CLS_ZERO): \
+ FP_SET_EXCEPTION(FP_EX_DIVZERO); \
+ case _FP_CLS_COMBINE(FP_CLS_INF,FP_CLS_ZERO): \
+ case _FP_CLS_COMBINE(FP_CLS_INF,FP_CLS_NORMAL): \
+ R##_c = FP_CLS_INF; \
+ break; \
+ \
+ case _FP_CLS_COMBINE(FP_CLS_INF,FP_CLS_INF): \
+ case _FP_CLS_COMBINE(FP_CLS_ZERO,FP_CLS_ZERO): \
+ R##_s = _FP_NANSIGN_##fs; \
+ R##_c = FP_CLS_NAN; \
+ _FP_FRAC_SET_##wc(R, _FP_NANFRAC_##fs); \
+ FP_SET_EXCEPTION(FP_EX_INVALID); \
+ break; \
+ \
+ default: \
+ abort(); \
+ } \
+} while (0)
+
+
+/*
+ * Main differential comparison routine. The inputs should be raw not
+ * cooked. The return is -1,0,1 for normal values, 2 otherwise.
+ */
+
+#define _FP_CMP(fs, wc, ret, X, Y, un) \
+ do { \
+ /* NANs are unordered */ \
+ if ((X##_e == _FP_EXPMAX_##fs && !_FP_FRAC_ZEROP_##wc(X)) \
+ || (Y##_e == _FP_EXPMAX_##fs && !_FP_FRAC_ZEROP_##wc(Y))) \
+ { \
+ ret = un; \
+ } \
+ else \
+ { \
+ int __is_zero_x; \
+ int __is_zero_y; \
+ \
+ __is_zero_x = (!X##_e && _FP_FRAC_ZEROP_##wc(X)) ? 1 : 0; \
+ __is_zero_y = (!Y##_e && _FP_FRAC_ZEROP_##wc(Y)) ? 1 : 0; \
+ \
+ if (__is_zero_x && __is_zero_y) \
+ ret = 0; \
+ else if (__is_zero_x) \
+ ret = Y##_s ? 1 : -1; \
+ else if (__is_zero_y) \
+ ret = X##_s ? -1 : 1; \
+ else if (X##_s != Y##_s) \
+ ret = X##_s ? -1 : 1; \
+ else if (X##_e > Y##_e) \
+ ret = X##_s ? -1 : 1; \
+ else if (X##_e < Y##_e) \
+ ret = X##_s ? 1 : -1; \
+ else if (_FP_FRAC_GT_##wc(X, Y)) \
+ ret = X##_s ? -1 : 1; \
+ else if (_FP_FRAC_GT_##wc(Y, X)) \
+ ret = X##_s ? 1 : -1; \
+ else \
+ ret = 0; \
+ } \
+ } while (0)
+
+
+/* Simplification for strict equality. */
+
+#define _FP_CMP_EQ(fs, wc, ret, X, Y) \
+ do { \
+ /* NANs are unordered */ \
+ if ((X##_e == _FP_EXPMAX_##fs && !_FP_FRAC_ZEROP_##wc(X)) \
+ || (Y##_e == _FP_EXPMAX_##fs && !_FP_FRAC_ZEROP_##wc(Y))) \
+ { \
+ ret = 1; \
+ } \
+ else \
+ { \
+ ret = !(X##_e == Y##_e \
+ && _FP_FRAC_EQ_##wc(X, Y) \
+ && (X##_s == Y##_s || (!X##_e && _FP_FRAC_ZEROP_##wc(X)))); \
+ } \
+ } while (0)
+
+/* Version to test unordered. */
+
+#define _FP_CMP_UNORD(fs, wc, ret, X, Y) \
+ do { \
+ ret = ((X##_e == _FP_EXPMAX_##fs && !_FP_FRAC_ZEROP_##wc(X)) \
+ || (Y##_e == _FP_EXPMAX_##fs && !_FP_FRAC_ZEROP_##wc(Y))); \
+ } while (0)
+
+/*
+ * Main square root routine. The input value should be cooked.
+ */
+
+#define _FP_SQRT(fs, wc, R, X) \
+do { \
+ _FP_FRAC_DECL_##wc(T); _FP_FRAC_DECL_##wc(S); \
+ _FP_W_TYPE q; \
+ switch (X##_c) \
+ { \
+ case FP_CLS_NAN: \
+ _FP_FRAC_COPY_##wc(R, X); \
+ R##_s = X##_s; \
+ R##_c = FP_CLS_NAN; \
+ break; \
+ case FP_CLS_INF: \
+ if (X##_s) \
+ { \
+ R##_s = _FP_NANSIGN_##fs; \
+ R##_c = FP_CLS_NAN; /* NAN */ \
+ _FP_FRAC_SET_##wc(R, _FP_NANFRAC_##fs); \
+ FP_SET_EXCEPTION(FP_EX_INVALID); \
+ } \
+ else \
+ { \
+ R##_s = 0; \
+ R##_c = FP_CLS_INF; /* sqrt(+inf) = +inf */ \
+ } \
+ break; \
+ case FP_CLS_ZERO: \
+ R##_s = X##_s; \
+ R##_c = FP_CLS_ZERO; /* sqrt(+-0) = +-0 */ \
+ break; \
+ case FP_CLS_NORMAL: \
+ R##_s = 0; \
+ if (X##_s) \
+ { \
+ R##_c = FP_CLS_NAN; /* sNAN */ \
+ R##_s = _FP_NANSIGN_##fs; \
+ _FP_FRAC_SET_##wc(R, _FP_NANFRAC_##fs); \
+ FP_SET_EXCEPTION(FP_EX_INVALID); \
+ break; \
+ } \
+ R##_c = FP_CLS_NORMAL; \
+ if (X##_e & 1) \
+ _FP_FRAC_SLL_##wc(X, 1); \
+ R##_e = X##_e >> 1; \
+ _FP_FRAC_SET_##wc(S, _FP_ZEROFRAC_##wc); \
+ _FP_FRAC_SET_##wc(R, _FP_ZEROFRAC_##wc); \
+ q = _FP_OVERFLOW_##fs >> 1; \
+ _FP_SQRT_MEAT_##wc(R, S, T, X, q); \
+ } \
+ } while (0)
+
+/*
+ * Convert from FP to integer. Input is raw.
+ */
+
+/* RSIGNED can have following values:
+ * 0: the number is required to be 0..(2^rsize)-1, if not, NV is set plus
+ * the result is either 0 or (2^rsize)-1 depending on the sign in such
+ * case.
+ * 1: the number is required to be -(2^(rsize-1))..(2^(rsize-1))-1, if not,
+ * NV is set plus the result is either -(2^(rsize-1)) or (2^(rsize-1))-1
+ * depending on the sign in such case.
+ * -1: the number is required to be -(2^(rsize-1))..(2^rsize)-1, if not, NV is
+ * set plus the result is either -(2^(rsize-1)) or (2^(rsize-1))-1
+ * depending on the sign in such case.
+ */
+#define _FP_TO_INT(fs, wc, r, X, rsize, rsigned) \
+do { \
+ if (X##_e < _FP_EXPBIAS_##fs) \
+ { \
+ r = 0; \
+ if (X##_e == 0) \
+ { \
+ if (!_FP_FRAC_ZEROP_##wc(X)) \
+ { \
+ FP_SET_EXCEPTION(FP_EX_INEXACT); \
+ FP_SET_EXCEPTION(FP_EX_DENORM); \
+ } \
+ } \
+ else \
+ FP_SET_EXCEPTION(FP_EX_INEXACT); \
+ } \
+ else if (X##_e >= _FP_EXPBIAS_##fs + rsize - (rsigned > 0 || X##_s) \
+ || (!rsigned && X##_s)) \
+ { \
+ /* Overflow or converting to the most negative integer. */ \
+ if (rsigned) \
+ { \
+ r = 1; \
+ r <<= rsize - 1; \
+ r -= 1 - X##_s; \
+ } else { \
+ r = 0; \
+ if (X##_s) \
+ r = ~r; \
+ } \
+ \
+ if (rsigned && X##_s && X##_e == _FP_EXPBIAS_##fs + rsize - 1) \
+ { \
+ /* Possibly converting to most negative integer; check the \
+ mantissa. */ \
+ int inexact = 0; \
+ (void)((_FP_FRACBITS_##fs > rsize) \
+ ? ({ _FP_FRAC_SRST_##wc(X, inexact, \
+ _FP_FRACBITS_##fs - rsize, \
+ _FP_FRACBITS_##fs); 0; }) \
+ : 0); \
+ if (!_FP_FRAC_ZEROP_##wc(X)) \
+ FP_SET_EXCEPTION(FP_EX_INVALID); \
+ else if (inexact) \
+ FP_SET_EXCEPTION(FP_EX_INEXACT); \
+ } \
+ else \
+ FP_SET_EXCEPTION(FP_EX_INVALID); \
+ } \
+ else \
+ { \
+ _FP_FRAC_HIGH_RAW_##fs(X) |= _FP_IMPLBIT_##fs; \
+ if (X##_e >= _FP_EXPBIAS_##fs + _FP_FRACBITS_##fs - 1) \
+ { \
+ _FP_FRAC_ASSEMBLE_##wc(r, X, rsize); \
+ r <<= X##_e - _FP_EXPBIAS_##fs - _FP_FRACBITS_##fs + 1; \
+ } \
+ else \
+ { \
+ int inexact; \
+ _FP_FRAC_SRST_##wc(X, inexact, \
+ (_FP_FRACBITS_##fs + _FP_EXPBIAS_##fs - 1 \
+ - X##_e), \
+ _FP_FRACBITS_##fs); \
+ if (inexact) \
+ FP_SET_EXCEPTION(FP_EX_INEXACT); \
+ _FP_FRAC_ASSEMBLE_##wc(r, X, rsize); \
+ } \
+ if (rsigned && X##_s) \
+ r = -r; \
+ } \
+} while (0)
+
+/* Convert integer to fp. Output is raw. RTYPE is unsigned even if
+ input is signed. */
+#define _FP_FROM_INT(fs, wc, X, r, rsize, rtype) \
+ do { \
+ if (r) \
+ { \
+ rtype ur_; \
+ \
+ if ((X##_s = (r < 0))) \
+ r = -(rtype)r; \
+ \
+ ur_ = (rtype) r; \
+ (void)((rsize <= _FP_W_TYPE_SIZE) \
+ ? ({ \
+ int lz_; \
+ __FP_CLZ(lz_, (_FP_W_TYPE)ur_); \
+ X##_e = _FP_EXPBIAS_##fs + _FP_W_TYPE_SIZE - 1 - lz_; \
+ }) \
+ : ((rsize <= 2 * _FP_W_TYPE_SIZE) \
+ ? ({ \
+ int lz_; \
+ __FP_CLZ_2(lz_, (_FP_W_TYPE)(ur_ >> _FP_W_TYPE_SIZE), \
+ (_FP_W_TYPE)ur_); \
+ X##_e = (_FP_EXPBIAS_##fs + 2 * _FP_W_TYPE_SIZE - 1 \
+ - lz_); \
+ }) \
+ : (abort(), 0))); \
+ \
+ if (rsize - 1 + _FP_EXPBIAS_##fs >= _FP_EXPMAX_##fs \
+ && X##_e >= _FP_EXPMAX_##fs) \
+ { \
+ /* Exponent too big; overflow to infinity. (May also \
+ happen after rounding below.) */ \
+ _FP_OVERFLOW_SEMIRAW(fs, wc, X); \
+ goto pack_semiraw; \
+ } \
+ \
+ if (rsize <= _FP_FRACBITS_##fs \
+ || X##_e < _FP_EXPBIAS_##fs + _FP_FRACBITS_##fs) \
+ { \
+ /* Exactly representable; shift left. */ \
+ _FP_FRAC_DISASSEMBLE_##wc(X, ur_, rsize); \
+ _FP_FRAC_SLL_##wc(X, (_FP_EXPBIAS_##fs \
+ + _FP_FRACBITS_##fs - 1 - X##_e)); \
+ } \
+ else \
+ { \
+ /* More bits in integer than in floating type; need to \
+ round. */ \
+ if (_FP_EXPBIAS_##fs + _FP_WFRACBITS_##fs - 1 < X##_e) \
+ ur_ = ((ur_ >> (X##_e - _FP_EXPBIAS_##fs \
+ - _FP_WFRACBITS_##fs + 1)) \
+ | ((ur_ << (rsize - (X##_e - _FP_EXPBIAS_##fs \
+ - _FP_WFRACBITS_##fs + 1))) \
+ != 0)); \
+ _FP_FRAC_DISASSEMBLE_##wc(X, ur_, rsize); \
+ if ((_FP_EXPBIAS_##fs + _FP_WFRACBITS_##fs - 1 - X##_e) > 0) \
+ _FP_FRAC_SLL_##wc(X, (_FP_EXPBIAS_##fs \
+ + _FP_WFRACBITS_##fs - 1 - X##_e)); \
+ _FP_FRAC_HIGH_##fs(X) &= ~(_FP_W_TYPE)_FP_IMPLBIT_SH_##fs; \
+ pack_semiraw: \
+ _FP_PACK_SEMIRAW(fs, wc, X); \
+ } \
+ } \
+ else \
+ { \
+ X##_s = 0; \
+ X##_e = 0; \
+ _FP_FRAC_SET_##wc(X, _FP_ZEROFRAC_##wc); \
+ } \
+ } while (0)
+
+
+/* Extend from a narrower floating-point format to a wider one. Input
+ and output are raw. */
+#define FP_EXTEND(dfs,sfs,dwc,swc,D,S) \
+do { \
+ if (_FP_FRACBITS_##dfs < _FP_FRACBITS_##sfs \
+ || (_FP_EXPMAX_##dfs - _FP_EXPBIAS_##dfs \
+ < _FP_EXPMAX_##sfs - _FP_EXPBIAS_##sfs) \
+ || _FP_EXPBIAS_##dfs < _FP_EXPBIAS_##sfs + _FP_FRACBITS_##sfs - 1) \
+ abort(); \
+ D##_s = S##_s; \
+ _FP_FRAC_COPY_##dwc##_##swc(D, S); \
+ if (_FP_EXP_NORMAL(sfs, swc, S)) \
+ { \
+ D##_e = S##_e + _FP_EXPBIAS_##dfs - _FP_EXPBIAS_##sfs; \
+ _FP_FRAC_SLL_##dwc(D, (_FP_FRACBITS_##dfs - _FP_FRACBITS_##sfs)); \
+ } \
+ else \
+ { \
+ if (S##_e == 0) \
+ { \
+ if (_FP_FRAC_ZEROP_##swc(S)) \
+ D##_e = 0; \
+ else \
+ { \
+ int _lz; \
+ FP_SET_EXCEPTION(FP_EX_DENORM); \
+ _FP_FRAC_CLZ_##swc(_lz, S); \
+ _FP_FRAC_SLL_##dwc(D, \
+ _lz + _FP_FRACBITS_##dfs \
+ - _FP_FRACTBITS_##sfs); \
+ D##_e = (_FP_EXPBIAS_##dfs - _FP_EXPBIAS_##sfs + 1 \
+ + _FP_FRACXBITS_##sfs - _lz); \
+ } \
+ } \
+ else \
+ { \
+ D##_e = _FP_EXPMAX_##dfs; \
+ if (!_FP_FRAC_ZEROP_##swc(S)) \
+ { \
+ if (!(_FP_FRAC_HIGH_RAW_##sfs(S) & _FP_QNANBIT_##sfs)) \
+ FP_SET_EXCEPTION(FP_EX_INVALID); \
+ _FP_FRAC_SLL_##dwc(D, (_FP_FRACBITS_##dfs \
+ - _FP_FRACBITS_##sfs)); \
+ } \
+ } \
+ } \
+} while (0)
+
+/* Truncate from a wider floating-point format to a narrower one.
+ Input and output are semi-raw. */
+#define FP_TRUNC(dfs,sfs,dwc,swc,D,S) \
+do { \
+ if (_FP_FRACBITS_##sfs < _FP_FRACBITS_##dfs \
+ || _FP_EXPBIAS_##sfs < _FP_EXPBIAS_##dfs + _FP_FRACBITS_##dfs - 1) \
+ abort(); \
+ D##_s = S##_s; \
+ if (_FP_EXP_NORMAL(sfs, swc, S)) \
+ { \
+ D##_e = S##_e + _FP_EXPBIAS_##dfs - _FP_EXPBIAS_##sfs; \
+ if (D##_e >= _FP_EXPMAX_##dfs) \
+ _FP_OVERFLOW_SEMIRAW(dfs, dwc, D); \
+ else \
+ { \
+ if (D##_e <= 0) \
+ { \
+ if (D##_e <= 1 - _FP_FRACBITS_##dfs) \
+ _FP_FRAC_SET_##swc(S, _FP_ZEROFRAC_##swc); \
+ else \
+ { \
+ _FP_FRAC_HIGH_##sfs(S) |= _FP_IMPLBIT_SH_##sfs; \
+ _FP_FRAC_SRS_##swc(S, (_FP_WFRACBITS_##sfs \
+ - _FP_WFRACBITS_##dfs + 1 - D##_e), \
+ _FP_WFRACBITS_##sfs); \
+ } \
+ D##_e = 0; \
+ } \
+ else \
+ _FP_FRAC_SRS_##swc(S, (_FP_WFRACBITS_##sfs \
+ - _FP_WFRACBITS_##dfs), \
+ _FP_WFRACBITS_##sfs); \
+ _FP_FRAC_COPY_##dwc##_##swc(D, S); \
+ } \
+ } \
+ else \
+ { \
+ if (S##_e == 0) \
+ { \
+ D##_e = 0; \
+ _FP_FRAC_SET_##dwc(D, _FP_ZEROFRAC_##dwc); \
+ if (!_FP_FRAC_ZEROP_##swc(S)) \
+ { \
+ FP_SET_EXCEPTION(FP_EX_DENORM); \
+ FP_SET_EXCEPTION(FP_EX_INEXACT); \
+ } \
+ } \
+ else \
+ { \
+ D##_e = _FP_EXPMAX_##dfs; \
+ if (_FP_FRAC_ZEROP_##swc(S)) \
+ _FP_FRAC_SET_##dwc(D, _FP_ZEROFRAC_##dwc); \
+ else \
+ { \
+ _FP_CHECK_SIGNAN_SEMIRAW(sfs, swc, S); \
+ _FP_FRAC_SRL_##swc(S, (_FP_WFRACBITS_##sfs \
+ - _FP_WFRACBITS_##dfs)); \
+ _FP_FRAC_COPY_##dwc##_##swc(D, S); \
+ _FP_FRAC_HIGH_##dfs(D) |= _FP_QNANBIT_SH_##dfs; \
+ } \
+ } \
+ } \
+} while (0)
+
+/*
+ * Helper primitives.
+ */
+
+/* Count leading zeros in a word. */
+
+#ifndef __FP_CLZ
+/* GCC 3.4 and later provide the builtins for us. */
+#define __FP_CLZ(r, x) \
+ do { \
+ if (sizeof (_FP_W_TYPE) == sizeof (unsigned int)) \
+ r = __builtin_clz (x); \
+ else if (sizeof (_FP_W_TYPE) == sizeof (unsigned long)) \
+ r = __builtin_clzl (x); \
+ else if (sizeof (_FP_W_TYPE) == sizeof (unsigned long long)) \
+ r = __builtin_clzll (x); \
+ else \
+ abort (); \
+ } while (0)
+#endif /* ndef __FP_CLZ */
+
+#define _FP_DIV_HELP_imm(q, r, n, d) \
+ do { \
+ q = n / d, r = n % d; \
+ } while (0)
+
+
+/* A restoring bit-by-bit division primitive. */
+
+#define _FP_DIV_MEAT_N_loop(fs, wc, R, X, Y) \
+ do { \
+ int count = _FP_WFRACBITS_##fs; \
+ _FP_FRAC_DECL_##wc (u); \
+ _FP_FRAC_DECL_##wc (v); \
+ _FP_FRAC_COPY_##wc (u, X); \
+ _FP_FRAC_COPY_##wc (v, Y); \
+ _FP_FRAC_SET_##wc (R, _FP_ZEROFRAC_##wc); \
+ /* Normalize U and V. */ \
+ _FP_FRAC_SLL_##wc (u, _FP_WFRACXBITS_##fs); \
+ _FP_FRAC_SLL_##wc (v, _FP_WFRACXBITS_##fs); \
+ /* First round. Since the operands are normalized, either the \
+ first or second bit will be set in the fraction. Produce a \
+ normalized result by checking which and adjusting the loop \
+ count and exponent accordingly. */ \
+ if (_FP_FRAC_GE_1 (u, v)) \
+ { \
+ _FP_FRAC_SUB_##wc (u, u, v); \
+ _FP_FRAC_LOW_##wc (R) |= 1; \
+ count--; \
+ } \
+ else \
+ R##_e--; \
+ /* Subsequent rounds. */ \
+ do { \
+ int msb = (_FP_WS_TYPE) _FP_FRAC_HIGH_##wc (u) < 0; \
+ _FP_FRAC_SLL_##wc (u, 1); \
+ _FP_FRAC_SLL_##wc (R, 1); \
+ if (msb || _FP_FRAC_GE_1 (u, v)) \
+ { \
+ _FP_FRAC_SUB_##wc (u, u, v); \
+ _FP_FRAC_LOW_##wc (R) |= 1; \
+ } \
+ } while (--count > 0); \
+ /* If there's anything left in U, the result is inexact. */ \
+ _FP_FRAC_LOW_##wc (R) |= !_FP_FRAC_ZEROP_##wc (u); \
+ } while (0)
+
+#define _FP_DIV_MEAT_1_loop(fs, R, X, Y) _FP_DIV_MEAT_N_loop (fs, 1, R, X, Y)
+#define _FP_DIV_MEAT_2_loop(fs, R, X, Y) _FP_DIV_MEAT_N_loop (fs, 2, R, X, Y)
+#define _FP_DIV_MEAT_4_loop(fs, R, X, Y) _FP_DIV_MEAT_N_loop (fs, 4, R, X, Y)
diff -rupN GCC.orig/gcc/config/soft-fp/quad.h GCC/gcc/config/soft-fp/quad.h
--- GCC.orig/gcc/config/soft-fp/quad.h 1970-01-01 00:00:00.000000000 +0000
+++ GCC/gcc/config/soft-fp/quad.h 2006-02-18 02:11:04.000000000 +0000
@@ -0,0 +1,271 @@
+/* Software floating-point emulation.
+ Definitions for IEEE Quad Precision.
+ Copyright (C) 1997,1998,1999,2006 Free Software Foundation, Inc.
+ This file is part of GCC.
+ Contributed by Richard Henderson (rth@cygnus.com),
+ Jakub Jelinek (jj@ultra.linux.cz),
+ David S. Miller (davem@redhat.com) and
+ Peter Maydell (pmaydell@chiark.greenend.org.uk).
+
+ GCC is free software; you can redistribute it and/or modify it under
+ the terms of the GNU General Public License as published by the Free
+ Software Foundation; either version 2, or (at your option) any later
+ version.
+
+ In addition to the permissions in the GNU 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 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.)
+
+ GCC 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 General Public License
+ for more details.
+
+ You should have received a copy of the GNU General Public License
+ along with GCC; see the file COPYING. If not, write to the Free
+ Software Foundation, 51 Franklin Street, Fifth Floor, Boston, MA
+ 02110-1301, USA. */
+
+#if _FP_W_TYPE_SIZE < 32
+#error "Here's a nickel, kid. Go buy yourself a real computer."
+#endif
+
+#if _FP_W_TYPE_SIZE < 64
+#define _FP_FRACTBITS_Q (4*_FP_W_TYPE_SIZE)
+#else
+#define _FP_FRACTBITS_Q (2*_FP_W_TYPE_SIZE)
+#endif
+
+#define _FP_FRACBITS_Q 113
+#define _FP_FRACXBITS_Q (_FP_FRACTBITS_Q - _FP_FRACBITS_Q)
+#define _FP_WFRACBITS_Q (_FP_WORKBITS + _FP_FRACBITS_Q)
+#define _FP_WFRACXBITS_Q (_FP_FRACTBITS_Q - _FP_WFRACBITS_Q)
+#define _FP_EXPBITS_Q 15
+#define _FP_EXPBIAS_Q 16383
+#define _FP_EXPMAX_Q 32767
+
+#define _FP_QNANBIT_Q \
+ ((_FP_W_TYPE)1 << (_FP_FRACBITS_Q-2) % _FP_W_TYPE_SIZE)
+#define _FP_QNANBIT_SH_Q \
+ ((_FP_W_TYPE)1 << (_FP_FRACBITS_Q-2+_FP_WORKBITS) % _FP_W_TYPE_SIZE)
+#define _FP_IMPLBIT_Q \
+ ((_FP_W_TYPE)1 << (_FP_FRACBITS_Q-1) % _FP_W_TYPE_SIZE)
+#define _FP_IMPLBIT_SH_Q \
+ ((_FP_W_TYPE)1 << (_FP_FRACBITS_Q-1+_FP_WORKBITS) % _FP_W_TYPE_SIZE)
+#define _FP_OVERFLOW_Q \
+ ((_FP_W_TYPE)1 << (_FP_WFRACBITS_Q % _FP_W_TYPE_SIZE))
+
+typedef float TFtype __attribute__((mode(TF)));
+
+#if _FP_W_TYPE_SIZE < 64
+
+union _FP_UNION_Q
+{
+ TFtype flt;
+ struct
+ {
+#if __BYTE_ORDER == __BIG_ENDIAN
+ unsigned sign : 1;
+ unsigned exp : _FP_EXPBITS_Q;
+ unsigned long frac3 : _FP_FRACBITS_Q - (_FP_IMPLBIT_Q != 0)-(_FP_W_TYPE_SIZE * 3);
+ unsigned long frac2 : _FP_W_TYPE_SIZE;
+ unsigned long frac1 : _FP_W_TYPE_SIZE;
+ unsigned long frac0 : _FP_W_TYPE_SIZE;
+#else
+ unsigned long frac0 : _FP_W_TYPE_SIZE;
+ unsigned long frac1 : _FP_W_TYPE_SIZE;
+ unsigned long frac2 : _FP_W_TYPE_SIZE;
+ unsigned long frac3 : _FP_FRACBITS_Q - (_FP_IMPLBIT_Q != 0)-(_FP_W_TYPE_SIZE * 3);
+ unsigned exp : _FP_EXPBITS_Q;
+ unsigned sign : 1;
+#endif /* not bigendian */
+ } bits __attribute__((packed));
+};
+
+
+#define FP_DECL_Q(X) _FP_DECL(4,X)
+#define FP_UNPACK_RAW_Q(X,val) _FP_UNPACK_RAW_4(Q,X,val)
+#define FP_UNPACK_RAW_QP(X,val) _FP_UNPACK_RAW_4_P(Q,X,val)
+#define FP_PACK_RAW_Q(val,X) _FP_PACK_RAW_4(Q,val,X)
+#define FP_PACK_RAW_QP(val,X) \
+ do { \
+ if (!FP_INHIBIT_RESULTS) \
+ _FP_PACK_RAW_4_P(Q,val,X); \
+ } while (0)
+
+#define FP_UNPACK_Q(X,val) \
+ do { \
+ _FP_UNPACK_RAW_4(Q,X,val); \
+ _FP_UNPACK_CANONICAL(Q,4,X); \
+ } while (0)
+
+#define FP_UNPACK_QP(X,val) \
+ do { \
+ _FP_UNPACK_RAW_4_P(Q,X,val); \
+ _FP_UNPACK_CANONICAL(Q,4,X); \
+ } while (0)
+
+#define FP_UNPACK_SEMIRAW_Q(X,val) \
+ do { \
+ _FP_UNPACK_RAW_4(Q,X,val); \
+ _FP_UNPACK_SEMIRAW(Q,4,X); \
+ } while (0)
+
+#define FP_UNPACK_SEMIRAW_QP(X,val) \
+ do { \
+ _FP_UNPACK_RAW_4_P(Q,X,val); \
+ _FP_UNPACK_SEMIRAW(Q,4,X); \
+ } while (0)
+
+#define FP_PACK_Q(val,X) \
+ do { \
+ _FP_PACK_CANONICAL(Q,4,X); \
+ _FP_PACK_RAW_4(Q,val,X); \
+ } while (0)
+
+#define FP_PACK_QP(val,X) \
+ do { \
+ _FP_PACK_CANONICAL(Q,4,X); \
+ if (!FP_INHIBIT_RESULTS) \
+ _FP_PACK_RAW_4_P(Q,val,X); \
+ } while (0)
+
+#define FP_PACK_SEMIRAW_Q(val,X) \
+ do { \
+ _FP_PACK_SEMIRAW(Q,4,X); \
+ _FP_PACK_RAW_4(Q,val,X); \
+ } while (0)
+
+#define FP_PACK_SEMIRAW_QP(val,X) \
+ do { \
+ _FP_PACK_SEMIRAW(Q,4,X); \
+ if (!FP_INHIBIT_RESULTS) \
+ _FP_PACK_RAW_4_P(Q,val,X); \
+ } while (0)
+
+#define FP_ISSIGNAN_Q(X) _FP_ISSIGNAN(Q,4,X)
+#define FP_NEG_Q(R,X) _FP_NEG(Q,4,R,X)
+#define FP_ADD_Q(R,X,Y) _FP_ADD(Q,4,R,X,Y)
+#define FP_SUB_Q(R,X,Y) _FP_SUB(Q,4,R,X,Y)
+#define FP_MUL_Q(R,X,Y) _FP_MUL(Q,4,R,X,Y)
+#define FP_DIV_Q(R,X,Y) _FP_DIV(Q,4,R,X,Y)
+#define FP_SQRT_Q(R,X) _FP_SQRT(Q,4,R,X)
+#define _FP_SQRT_MEAT_Q(R,S,T,X,Q) _FP_SQRT_MEAT_4(R,S,T,X,Q)
+
+#define FP_CMP_Q(r,X,Y,un) _FP_CMP(Q,4,r,X,Y,un)
+#define FP_CMP_EQ_Q(r,X,Y) _FP_CMP_EQ(Q,4,r,X,Y)
+#define FP_CMP_UNORD_Q(r,X,Y) _FP_CMP_UNORD(Q,4,r,X,Y)
+
+#define FP_TO_INT_Q(r,X,rsz,rsg) _FP_TO_INT(Q,4,r,X,rsz,rsg)
+#define FP_FROM_INT_Q(X,r,rs,rt) _FP_FROM_INT(Q,4,X,r,rs,rt)
+
+#define _FP_FRAC_HIGH_Q(X) _FP_FRAC_HIGH_4(X)
+#define _FP_FRAC_HIGH_RAW_Q(X) _FP_FRAC_HIGH_4(X)
+
+#else /* not _FP_W_TYPE_SIZE < 64 */
+union _FP_UNION_Q
+{
+ TFtype flt /* __attribute__((mode(TF))) */ ;
+ struct {
+ _FP_W_TYPE a, b;
+ } longs;
+ struct {
+#if __BYTE_ORDER == __BIG_ENDIAN
+ unsigned sign : 1;
+ unsigned exp : _FP_EXPBITS_Q;
+ unsigned long frac1 : _FP_FRACBITS_Q-(_FP_IMPLBIT_Q != 0)-_FP_W_TYPE_SIZE;
+ unsigned long frac0 : _FP_W_TYPE_SIZE;
+#else
+ unsigned long frac0 : _FP_W_TYPE_SIZE;
+ unsigned long frac1 : _FP_FRACBITS_Q-(_FP_IMPLBIT_Q != 0)-_FP_W_TYPE_SIZE;
+ unsigned exp : _FP_EXPBITS_Q;
+ unsigned sign : 1;
+#endif
+ } bits;
+};
+
+#define FP_DECL_Q(X) _FP_DECL(2,X)
+#define FP_UNPACK_RAW_Q(X,val) _FP_UNPACK_RAW_2(Q,X,val)
+#define FP_UNPACK_RAW_QP(X,val) _FP_UNPACK_RAW_2_P(Q,X,val)
+#define FP_PACK_RAW_Q(val,X) _FP_PACK_RAW_2(Q,val,X)
+#define FP_PACK_RAW_QP(val,X) \
+ do { \
+ if (!FP_INHIBIT_RESULTS) \
+ _FP_PACK_RAW_2_P(Q,val,X); \
+ } while (0)
+
+#define FP_UNPACK_Q(X,val) \
+ do { \
+ _FP_UNPACK_RAW_2(Q,X,val); \
+ _FP_UNPACK_CANONICAL(Q,2,X); \
+ } while (0)
+
+#define FP_UNPACK_QP(X,val) \
+ do { \
+ _FP_UNPACK_RAW_2_P(Q,X,val); \
+ _FP_UNPACK_CANONICAL(Q,2,X); \
+ } while (0)
+
+#define FP_UNPACK_SEMIRAW_Q(X,val) \
+ do { \
+ _FP_UNPACK_RAW_2(Q,X,val); \
+ _FP_UNPACK_SEMIRAW(Q,2,X); \
+ } while (0)
+
+#define FP_UNPACK_SEMIRAW_QP(X,val) \
+ do { \
+ _FP_UNPACK_RAW_2_P(Q,X,val); \
+ _FP_UNPACK_SEMIRAW(Q,2,X); \
+ } while (0)
+
+#define FP_PACK_Q(val,X) \
+ do { \
+ _FP_PACK_CANONICAL(Q,2,X); \
+ _FP_PACK_RAW_2(Q,val,X); \
+ } while (0)
+
+#define FP_PACK_QP(val,X) \
+ do { \
+ _FP_PACK_CANONICAL(Q,2,X); \
+ if (!FP_INHIBIT_RESULTS) \
+ _FP_PACK_RAW_2_P(Q,val,X); \
+ } while (0)
+
+#define FP_PACK_SEMIRAW_Q(val,X) \
+ do { \
+ _FP_PACK_SEMIRAW(Q,2,X); \
+ _FP_PACK_RAW_2(Q,val,X); \
+ } while (0)
+
+#define FP_PACK_SEMIRAW_QP(val,X) \
+ do { \
+ _FP_PACK_SEMIRAW(Q,2,X); \
+ if (!FP_INHIBIT_RESULTS) \
+ _FP_PACK_RAW_2_P(Q,val,X); \
+ } while (0)
+
+#define FP_ISSIGNAN_Q(X) _FP_ISSIGNAN(Q,2,X)
+#define FP_NEG_Q(R,X) _FP_NEG(Q,2,R,X)
+#define FP_ADD_Q(R,X,Y) _FP_ADD(Q,2,R,X,Y)
+#define FP_SUB_Q(R,X,Y) _FP_SUB(Q,2,R,X,Y)
+#define FP_MUL_Q(R,X,Y) _FP_MUL(Q,2,R,X,Y)
+#define FP_DIV_Q(R,X,Y) _FP_DIV(Q,2,R,X,Y)
+#define FP_SQRT_Q(R,X) _FP_SQRT(Q,2,R,X)
+#define _FP_SQRT_MEAT_Q(R,S,T,X,Q) _FP_SQRT_MEAT_2(R,S,T,X,Q)
+
+#define FP_CMP_Q(r,X,Y,un) _FP_CMP(Q,2,r,X,Y,un)
+#define FP_CMP_EQ_Q(r,X,Y) _FP_CMP_EQ(Q,2,r,X,Y)
+#define FP_CMP_UNORD_Q(r,X,Y) _FP_CMP_UNORD(Q,2,r,X,Y)
+
+#define FP_TO_INT_Q(r,X,rsz,rsg) _FP_TO_INT(Q,2,r,X,rsz,rsg)
+#define FP_FROM_INT_Q(X,r,rs,rt) _FP_FROM_INT(Q,2,X,r,rs,rt)
+
+#define _FP_FRAC_HIGH_Q(X) _FP_FRAC_HIGH_2(X)
+#define _FP_FRAC_HIGH_RAW_Q(X) _FP_FRAC_HIGH_2(X)
+
+#endif /* not _FP_W_TYPE_SIZE < 64 */
diff -rupN GCC.orig/gcc/config/soft-fp/single.h GCC/gcc/config/soft-fp/single.h
--- GCC.orig/gcc/config/soft-fp/single.h 1970-01-01 00:00:00.000000000 +0000
+++ GCC/gcc/config/soft-fp/single.h 2006-02-18 02:11:04.000000000 +0000
@@ -0,0 +1,151 @@
+/* Software floating-point emulation.
+ Definitions for IEEE Single Precision.
+ Copyright (C) 1997,1998,1999,2006 Free Software Foundation, Inc.
+ This file is part of GCC.
+ Contributed by Richard Henderson (rth@cygnus.com),
+ Jakub Jelinek (jj@ultra.linux.cz),
+ David S. Miller (davem@redhat.com) and
+ Peter Maydell (pmaydell@chiark.greenend.org.uk).
+
+ GCC is free software; you can redistribute it and/or modify it under
+ the terms of the GNU General Public License as published by the Free
+ Software Foundation; either version 2, or (at your option) any later
+ version.
+
+ In addition to the permissions in the GNU 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 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.)
+
+ GCC 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 General Public License
+ for more details.
+
+ You should have received a copy of the GNU General Public License
+ along with GCC; see the file COPYING. If not, write to the Free
+ Software Foundation, 51 Franklin Street, Fifth Floor, Boston, MA
+ 02110-1301, USA. */
+
+#if _FP_W_TYPE_SIZE < 32
+#error "Here's a nickel kid. Go buy yourself a real computer."
+#endif
+
+#define _FP_FRACTBITS_S _FP_W_TYPE_SIZE
+
+#define _FP_FRACBITS_S 24
+#define _FP_FRACXBITS_S (_FP_FRACTBITS_S - _FP_FRACBITS_S)
+#define _FP_WFRACBITS_S (_FP_WORKBITS + _FP_FRACBITS_S)
+#define _FP_WFRACXBITS_S (_FP_FRACTBITS_S - _FP_WFRACBITS_S)
+#define _FP_EXPBITS_S 8
+#define _FP_EXPBIAS_S 127
+#define _FP_EXPMAX_S 255
+#define _FP_QNANBIT_S ((_FP_W_TYPE)1 << (_FP_FRACBITS_S-2))
+#define _FP_QNANBIT_SH_S ((_FP_W_TYPE)1 << (_FP_FRACBITS_S-2+_FP_WORKBITS))
+#define _FP_IMPLBIT_S ((_FP_W_TYPE)1 << (_FP_FRACBITS_S-1))
+#define _FP_IMPLBIT_SH_S ((_FP_W_TYPE)1 << (_FP_FRACBITS_S-1+_FP_WORKBITS))
+#define _FP_OVERFLOW_S ((_FP_W_TYPE)1 << (_FP_WFRACBITS_S))
+
+/* The implementation of _FP_MUL_MEAT_S and _FP_DIV_MEAT_S should be
+ chosen by the target machine. */
+
+typedef float SFtype __attribute__((mode(SF)));
+
+union _FP_UNION_S
+{
+ SFtype flt;
+ struct {
+#if __BYTE_ORDER == __BIG_ENDIAN
+ unsigned sign : 1;
+ unsigned exp : _FP_EXPBITS_S;
+ unsigned frac : _FP_FRACBITS_S - (_FP_IMPLBIT_S != 0);
+#else
+ unsigned frac : _FP_FRACBITS_S - (_FP_IMPLBIT_S != 0);
+ unsigned exp : _FP_EXPBITS_S;
+ unsigned sign : 1;
+#endif
+ } bits __attribute__((packed));
+};
+
+#define FP_DECL_S(X) _FP_DECL(1,X)
+#define FP_UNPACK_RAW_S(X,val) _FP_UNPACK_RAW_1(S,X,val)
+#define FP_UNPACK_RAW_SP(X,val) _FP_UNPACK_RAW_1_P(S,X,val)
+#define FP_PACK_RAW_S(val,X) _FP_PACK_RAW_1(S,val,X)
+#define FP_PACK_RAW_SP(val,X) \
+ do { \
+ if (!FP_INHIBIT_RESULTS) \
+ _FP_PACK_RAW_1_P(S,val,X); \
+ } while (0)
+
+#define FP_UNPACK_S(X,val) \
+ do { \
+ _FP_UNPACK_RAW_1(S,X,val); \
+ _FP_UNPACK_CANONICAL(S,1,X); \
+ } while (0)
+
+#define FP_UNPACK_SP(X,val) \
+ do { \
+ _FP_UNPACK_RAW_1_P(S,X,val); \
+ _FP_UNPACK_CANONICAL(S,1,X); \
+ } while (0)
+
+#define FP_UNPACK_SEMIRAW_S(X,val) \
+ do { \
+ _FP_UNPACK_RAW_1(S,X,val); \
+ _FP_UNPACK_SEMIRAW(S,1,X); \
+ } while (0)
+
+#define FP_UNPACK_SEMIRAW_SP(X,val) \
+ do { \
+ _FP_UNPACK_RAW_1_P(S,X,val); \
+ _FP_UNPACK_SEMIRAW(S,1,X); \
+ } while (0)
+
+#define FP_PACK_S(val,X) \
+ do { \
+ _FP_PACK_CANONICAL(S,1,X); \
+ _FP_PACK_RAW_1(S,val,X); \
+ } while (0)
+
+#define FP_PACK_SP(val,X) \
+ do { \
+ _FP_PACK_CANONICAL(S,1,X); \
+ if (!FP_INHIBIT_RESULTS) \
+ _FP_PACK_RAW_1_P(S,val,X); \
+ } while (0)
+
+#define FP_PACK_SEMIRAW_S(val,X) \
+ do { \
+ _FP_PACK_SEMIRAW(S,1,X); \
+ _FP_PACK_RAW_1(S,val,X); \
+ } while (0)
+
+#define FP_PACK_SEMIRAW_SP(val,X) \
+ do { \
+ _FP_PACK_SEMIRAW(S,1,X); \
+ if (!FP_INHIBIT_RESULTS) \
+ _FP_PACK_RAW_1_P(S,val,X); \
+ } while (0)
+
+#define FP_ISSIGNAN_S(X) _FP_ISSIGNAN(S,1,X)
+#define FP_NEG_S(R,X) _FP_NEG(S,1,R,X)
+#define FP_ADD_S(R,X,Y) _FP_ADD(S,1,R,X,Y)
+#define FP_SUB_S(R,X,Y) _FP_SUB(S,1,R,X,Y)
+#define FP_MUL_S(R,X,Y) _FP_MUL(S,1,R,X,Y)
+#define FP_DIV_S(R,X,Y) _FP_DIV(S,1,R,X,Y)
+#define FP_SQRT_S(R,X) _FP_SQRT(S,1,R,X)
+#define _FP_SQRT_MEAT_S(R,S,T,X,Q) _FP_SQRT_MEAT_1(R,S,T,X,Q)
+
+#define FP_CMP_S(r,X,Y,un) _FP_CMP(S,1,r,X,Y,un)
+#define FP_CMP_EQ_S(r,X,Y) _FP_CMP_EQ(S,1,r,X,Y)
+#define FP_CMP_UNORD_S(r,X,Y) _FP_CMP_UNORD(S,1,r,X,Y)
+
+#define FP_TO_INT_S(r,X,rsz,rsg) _FP_TO_INT(S,1,r,X,rsz,rsg)
+#define FP_FROM_INT_S(X,r,rs,rt) _FP_FROM_INT(S,1,X,r,rs,rt)
+
+#define _FP_FRAC_HIGH_S(X) _FP_FRAC_HIGH_1(X)
+#define _FP_FRAC_HIGH_RAW_S(X) _FP_FRAC_HIGH_1(X)
diff -rupN GCC.orig/gcc/config/soft-fp/soft-fp.h GCC/gcc/config/soft-fp/soft-fp.h
--- GCC.orig/gcc/config/soft-fp/soft-fp.h 1970-01-01 00:00:00.000000000 +0000
+++ GCC/gcc/config/soft-fp/soft-fp.h 2006-02-18 02:11:04.000000000 +0000
@@ -0,0 +1,209 @@
+/* Software floating-point emulation.
+ Copyright (C) 1997,1998,1999,2000,2002,2003,2005,2006
+ Free Software Foundation, Inc.
+ This file is part of GCC.
+ Contributed by Richard Henderson (rth@cygnus.com),
+ Jakub Jelinek (jj@ultra.linux.cz),
+ David S. Miller (davem@redhat.com) and
+ Peter Maydell (pmaydell@chiark.greenend.org.uk).
+
+ GCC is free software; you can redistribute it and/or modify it under
+ the terms of the GNU General Public License as published by the Free
+ Software Foundation; either version 2, or (at your option) any later
+ version.
+
+ In addition to the permissions in the GNU 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 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.)
+
+ GCC 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 General Public License
+ for more details.
+
+ You should have received a copy of the GNU General Public License
+ along with GCC; see the file COPYING. If not, write to the Free
+ Software Foundation, 51 Franklin Street, Fifth Floor, Boston, MA
+ 02110-1301, USA. */
+
+#ifndef SOFT_FP_H
+#define SOFT_FP_H
+
+#ifdef _LIBC
+#include <sfp-machine.h>
+#else
+#include "sfp-machine.h"
+#endif
+
+/* Allow sfp-machine to have its own byte order definitions. */
+#ifndef __BYTE_ORDER
+#ifdef _LIBC
+#include <endian.h>
+#else
+#error "endianness not defined by sfp-machine.h"
+#endif
+#endif
+
+#define _FP_WORKBITS 3
+#define _FP_WORK_LSB ((_FP_W_TYPE)1 << 3)
+#define _FP_WORK_ROUND ((_FP_W_TYPE)1 << 2)
+#define _FP_WORK_GUARD ((_FP_W_TYPE)1 << 1)
+#define _FP_WORK_STICKY ((_FP_W_TYPE)1 << 0)
+
+#ifndef FP_RND_NEAREST
+# define FP_RND_NEAREST 0
+# define FP_RND_ZERO 1
+# define FP_RND_PINF 2
+# define FP_RND_MINF 3
+#endif
+#ifndef FP_ROUNDMODE
+# define FP_ROUNDMODE FP_RND_NEAREST
+#endif
+
+/* By default don't care about exceptions. */
+#ifndef FP_EX_INVALID
+#define FP_EX_INVALID 0
+#endif
+#ifndef FP_EX_OVERFLOW
+#define FP_EX_OVERFLOW 0
+#endif
+#ifndef FP_EX_UNDERFLOW
+#define FP_EX_UNDERFLOW 0
+#endif
+#ifndef FP_EX_DIVZERO
+#define FP_EX_DIVZERO 0
+#endif
+#ifndef FP_EX_INEXACT
+#define FP_EX_INEXACT 0
+#endif
+#ifndef FP_EX_DENORM
+#define FP_EX_DENORM 0
+#endif
+
+#ifdef _FP_DECL_EX
+#define FP_DECL_EX \
+ int _fex = 0; \
+ _FP_DECL_EX
+#else
+#define FP_DECL_EX int _fex = 0
+#endif
+
+#ifndef FP_INIT_ROUNDMODE
+#define FP_INIT_ROUNDMODE do {} while (0)
+#endif
+
+#ifndef FP_HANDLE_EXCEPTIONS
+#define FP_HANDLE_EXCEPTIONS do {} while (0)
+#endif
+
+#ifndef FP_INHIBIT_RESULTS
+/* By default we write the results always.
+ * sfp-machine may override this and e.g.
+ * check if some exceptions are unmasked
+ * and inhibit it in such a case.
+ */
+#define FP_INHIBIT_RESULTS 0
+#endif
+
+#define FP_SET_EXCEPTION(ex) \
+ _fex |= (ex)
+
+#define FP_UNSET_EXCEPTION(ex) \
+ _fex &= ~(ex)
+
+#define FP_CLEAR_EXCEPTIONS \
+ _fex = 0
+
+#define _FP_ROUND_NEAREST(wc, X) \
+do { \
+ if ((_FP_FRAC_LOW_##wc(X) & 15) != _FP_WORK_ROUND) \
+ _FP_FRAC_ADDI_##wc(X, _FP_WORK_ROUND); \
+} while (0)
+
+#define _FP_ROUND_ZERO(wc, X) (void)0
+
+#define _FP_ROUND_PINF(wc, X) \
+do { \
+ if (!X##_s && (_FP_FRAC_LOW_##wc(X) & 7)) \
+ _FP_FRAC_ADDI_##wc(X, _FP_WORK_LSB); \
+} while (0)
+
+#define _FP_ROUND_MINF(wc, X) \
+do { \
+ if (X##_s && (_FP_FRAC_LOW_##wc(X) & 7)) \
+ _FP_FRAC_ADDI_##wc(X, _FP_WORK_LSB); \
+} while (0)
+
+#define _FP_ROUND(wc, X) \
+do { \
+ if (_FP_FRAC_LOW_##wc(X) & 7) \
+ FP_SET_EXCEPTION(FP_EX_INEXACT); \
+ switch (FP_ROUNDMODE) \
+ { \
+ case FP_RND_NEAREST: \
+ _FP_ROUND_NEAREST(wc,X); \
+ break; \
+ case FP_RND_ZERO: \
+ _FP_ROUND_ZERO(wc,X); \
+ break; \
+ case FP_RND_PINF: \
+ _FP_ROUND_PINF(wc,X); \
+ break; \
+ case FP_RND_MINF: \
+ _FP_ROUND_MINF(wc,X); \
+ break; \
+ } \
+} while (0)
+
+#define FP_CLS_NORMAL 0
+#define FP_CLS_ZERO 1
+#define FP_CLS_INF 2
+#define FP_CLS_NAN 3
+
+#define _FP_CLS_COMBINE(x,y) (((x) << 2) | (y))
+
+#include "op-1.h"
+#include "op-2.h"
+#include "op-4.h"
+#include "op-8.h"
+#include "op-common.h"
+
+/* Sigh. Silly things longlong.h needs. */
+#define UWtype _FP_W_TYPE
+#define W_TYPE_SIZE _FP_W_TYPE_SIZE
+
+typedef int QItype __attribute__((mode(QI)));
+typedef int SItype __attribute__((mode(SI)));
+typedef int DItype __attribute__((mode(DI)));
+typedef unsigned int UQItype __attribute__((mode(QI)));
+typedef unsigned int USItype __attribute__((mode(SI)));
+typedef unsigned int UDItype __attribute__((mode(DI)));
+#if _FP_W_TYPE_SIZE == 32
+typedef unsigned int UHWtype __attribute__((mode(HI)));
+#elif _FP_W_TYPE_SIZE == 64
+typedef USItype UHWtype;
+#endif
+
+#define SI_BITS (__CHAR_BIT__ * (int)sizeof(SItype))
+#define DI_BITS (__CHAR_BIT__ * (int)sizeof(DItype))
+
+#ifndef umul_ppmm
+#ifdef _LIBC
+#include <stdlib/longlong.h>
+#else
+#include "longlong.h"
+#endif
+#endif
+
+#ifdef _LIBC
+#include <stdlib.h>
+#else
+extern void abort (void);
+#endif
+
+#endif
diff -rupN GCC.orig/gcc/config/soft-fp/subdf3.c GCC/gcc/config/soft-fp/subdf3.c
--- GCC.orig/gcc/config/soft-fp/subdf3.c 1970-01-01 00:00:00.000000000 +0000
+++ GCC/gcc/config/soft-fp/subdf3.c 2006-02-18 02:11:04.000000000 +0000
@@ -0,0 +1,49 @@
+/* Software floating-point emulation.
+ Return a - b
+ Copyright (C) 1997,1999,2006 Free Software Foundation, Inc.
+ This file is part of GCC.
+ Contributed by Richard Henderson (rth@cygnus.com) and
+ Jakub Jelinek (jj@ultra.linux.cz).
+
+ GCC is free software; you can redistribute it and/or modify it under
+ the terms of the GNU General Public License as published by the Free
+ Software Foundation; either version 2, or (at your option) any later
+ version.
+
+ In addition to the permissions in the GNU 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 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.)
+
+ GCC 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 General Public License
+ for more details.
+
+ You should have received a copy of the GNU General Public License
+ along with GCC; see the file COPYING. If not, write to the Free
+ Software Foundation, 51 Franklin Street, Fifth Floor, Boston, MA
+ 02110-1301, USA. */
+
+#include "soft-fp.h"
+#include "double.h"
+
+DFtype __subdf3(DFtype a, DFtype b)
+{
+ FP_DECL_EX;
+ FP_DECL_D(A); FP_DECL_D(B); FP_DECL_D(R);
+ DFtype r;
+
+ FP_INIT_ROUNDMODE;
+ FP_UNPACK_SEMIRAW_D(A, a);
+ FP_UNPACK_SEMIRAW_D(B, b);
+ FP_SUB_D(R, A, B);
+ FP_PACK_SEMIRAW_D(r, R);
+ FP_HANDLE_EXCEPTIONS;
+
+ return r;
+}
diff -rupN GCC.orig/gcc/config/soft-fp/subsf3.c GCC/gcc/config/soft-fp/subsf3.c
--- GCC.orig/gcc/config/soft-fp/subsf3.c 1970-01-01 00:00:00.000000000 +0000
+++ GCC/gcc/config/soft-fp/subsf3.c 2006-02-18 02:11:04.000000000 +0000
@@ -0,0 +1,49 @@
+/* Software floating-point emulation.
+ Return a - b
+ Copyright (C) 1997,1999,2006 Free Software Foundation, Inc.
+ This file is part of GCC.
+ Contributed by Richard Henderson (rth@cygnus.com) and
+ Jakub Jelinek (jj@ultra.linux.cz).
+
+ GCC is free software; you can redistribute it and/or modify it under
+ the terms of the GNU General Public License as published by the Free
+ Software Foundation; either version 2, or (at your option) any later
+ version.
+
+ In addition to the permissions in the GNU 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 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.)
+
+ GCC 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 General Public License
+ for more details.
+
+ You should have received a copy of the GNU General Public License
+ along with GCC; see the file COPYING. If not, write to the Free
+ Software Foundation, 51 Franklin Street, Fifth Floor, Boston, MA
+ 02110-1301, USA. */
+
+#include "soft-fp.h"
+#include "single.h"
+
+SFtype __subsf3(SFtype a, SFtype b)
+{
+ FP_DECL_EX;
+ FP_DECL_S(A); FP_DECL_S(B); FP_DECL_S(R);
+ SFtype r;
+
+ FP_INIT_ROUNDMODE;
+ FP_UNPACK_SEMIRAW_S(A, a);
+ FP_UNPACK_SEMIRAW_S(B, b);
+ FP_SUB_S(R, A, B);
+ FP_PACK_SEMIRAW_S(r, R);
+ FP_HANDLE_EXCEPTIONS;
+
+ return r;
+}
diff -rupN GCC.orig/gcc/config/soft-fp/subtf3.c GCC/gcc/config/soft-fp/subtf3.c
--- GCC.orig/gcc/config/soft-fp/subtf3.c 1970-01-01 00:00:00.000000000 +0000
+++ GCC/gcc/config/soft-fp/subtf3.c 2006-02-18 02:11:04.000000000 +0000
@@ -0,0 +1,49 @@
+/* Software floating-point emulation.
+ Return a - b
+ Copyright (C) 1997,1999,2006 Free Software Foundation, Inc.
+ This file is part of GCC.
+ Contributed by Richard Henderson (rth@cygnus.com) and
+ Jakub Jelinek (jj@ultra.linux.cz).
+
+ GCC is free software; you can redistribute it and/or modify it under
+ the terms of the GNU General Public License as published by the Free
+ Software Foundation; either version 2, or (at your option) any later
+ version.
+
+ In addition to the permissions in the GNU 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 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.)
+
+ GCC 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 General Public License
+ for more details.
+
+ You should have received a copy of the GNU General Public License
+ along with GCC; see the file COPYING. If not, write to the Free
+ Software Foundation, 51 Franklin Street, Fifth Floor, Boston, MA
+ 02110-1301, USA. */
+
+#include "soft-fp.h"
+#include "quad.h"
+
+TFtype __subtf3(TFtype a, TFtype b)
+{
+ FP_DECL_EX;
+ FP_DECL_Q(A); FP_DECL_Q(B); FP_DECL_Q(R);
+ TFtype r;
+
+ FP_INIT_ROUNDMODE;
+ FP_UNPACK_SEMIRAW_Q(A, a);
+ FP_UNPACK_SEMIRAW_Q(B, b);
+ FP_SUB_Q(R, A, B);
+ FP_PACK_SEMIRAW_Q(r, R);
+ FP_HANDLE_EXCEPTIONS;
+
+ return r;
+}
diff -rupN GCC.orig/gcc/config/soft-fp/t-softfp GCC/gcc/config/soft-fp/t-softfp
--- GCC.orig/gcc/config/soft-fp/t-softfp 1970-01-01 00:00:00.000000000 +0000
+++ GCC/gcc/config/soft-fp/t-softfp 2006-03-11 18:59:28.000000000 +0000
@@ -0,0 +1,108 @@
+# Copyright (C) 2006 Free Software Foundation, Inc.
+
+# This file is part of GCC.
+
+# GCC is free software; you can redistribute it and/or modify
+# it under the terms of the GNU General Public License as published by
+# the Free Software Foundation; either version 2, or (at your option)
+# any later version.
+
+# GCC 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 General Public License for more details.
+
+# You should have received a copy of the GNU General Public License
+# along with GCC; see the file COPYING. If not, write to
+# the Free Software Foundation, 51 Franklin Street, Fifth Floor,
+# Boston MA 02110-1301, USA.
+
+# Targets using soft-fp should define the following variables:
+#
+# softfp_float_modes: a list of soft-float floating-point modes,
+# e.g. sf df
+# softfp_int_modes: a list of integer modes for which to define conversions,
+# e.g. si di
+# softfp_extensions: a list of extensions between floating-point modes,
+# e.g. sfdf
+# softfp_truncations: a list of truncations between floating-point modes,
+# e.g. dfsf
+# softfp_machine_header: the target sfp-machine.h file (relative to config/),
+# e.g. rs6000/sfp-machine.h
+#
+# Extensions and truncations should include those where only one mode
+# is a soft-float mode; for example, sftf where sf is hard-float and
+# tf is soft-float.
+#
+# If the libgcc2.c functions should not be replaced, also define:
+#
+# softfp_exclude_libgcc2 := y
+#
+# Avoiding replacing the libgcc2.c functions is a temporary measure
+# for targets with both hard-float and soft-float multilibs, since
+# these variables apply for all multilibs. With toplevel libgcc,
+# soft-fp can be used conditionally on the multilib instead.
+#
+# If the code should not be compiled at all for some multilibs, define:
+#
+# softfp_wrap_start: text to put at the start of wrapper source files,
+# output with echo
+# e.g. '#ifndef __powerpc64__'
+# softfp_wrap_end: text to put at the end of wrapper source files,
+# e.g. '#endif'
+#
+# This is another temporary measure.
+
+softfp_float_funcs = add$(m)3 div$(m)3 eq$(m)2 ge$(m)2 le$(m)2 mul$(m)3 \
+ neg$(m)2 sub$(m)3 unord$(m)2
+softfp_floatint_funcs = fix$(m)$(i) fixuns$(m)$(i) \
+ float$(i)$(m) floatun$(i)$(m)
+
+softfp_func_list := \
+ $(foreach m,$(softfp_float_modes), \
+ $(softfp_float_funcs) \
+ $(foreach i,$(softfp_int_modes), \
+ $(softfp_floatint_funcs))) \
+ $(foreach e,$(softfp_extensions),extend$(e)2) \
+ $(foreach t,$(softfp_truncations),trunc$(t)2)
+
+ifeq ($(softfp_exclude_libgcc2),y)
+# This list is taken from mklibgcc.in and doesn't presently allow for
+# 64-bit targets where si should become di and di should become ti.
+softfp_func_list := $(filter-out floatdidf floatdisf fixunsdfsi fixunssfsi \
+ fixunsdfdi fixdfdi fixunssfdi fixsfdi fixxfdi fixunsxfdi \
+ floatdixf fixunsxfsi fixtfdi fixunstfdi floatditf \
+ floatundidf floatundisf floatundixf floatunditf,$(softfp_func_list))
+endif
+
+ifeq ($(softfp_wrap_start),)
+softfp_file_list := \
+ $(addsuffix .c,$(addprefix $(srcdir)/config/soft-fp/,$(softfp_func_list)))
+else
+softfp_file_list := $(addsuffix .c,$(softfp_func_list))
+
+$(softfp_file_list):
+ echo $(softfp_wrap_start) > $@
+ echo '#include "config/soft-fp/$@"' >> $@
+ echo $(softfp_wrap_end) >> $@
+endif
+
+LIB2FUNCS_EXTRA += $(softfp_file_list)
+
+ifneq ($(softfp_exclude_libgcc2),y)
+# Functions in libgcc2.c are excluded for each soft-float mode (a
+# target may have both soft-float and hard-float modes), for the fixed
+# list of integer modes (si and di) for which libgcc2.c defines any
+# such functions. Depending on the target, the si and di symbols may
+# in fact define di and ti functions.
+
+LIB2FUNCS_EXCLUDE += \
+ $(addprefix _,$(foreach m,$(softfp_float_modes), \
+ $(foreach i,si di, \
+ $(softfp_floatint_funcs))))
+endif
+
+SFP_MACHINE := sfp-machine.h
+
+$(SFP_MACHINE): $(srcdir)/config/$(softfp_machine_header)
+ cp $(srcdir)/config/$(softfp_machine_header) $(SFP_MACHINE)
diff -rupN GCC.orig/gcc/config/soft-fp/truncdfsf2.c GCC/gcc/config/soft-fp/truncdfsf2.c
--- GCC.orig/gcc/config/soft-fp/truncdfsf2.c 1970-01-01 00:00:00.000000000 +0000
+++ GCC/gcc/config/soft-fp/truncdfsf2.c 2006-02-18 02:11:04.000000000 +0000
@@ -0,0 +1,54 @@
+/* Software floating-point emulation.
+ Truncate IEEE double into IEEE single
+ Copyright (C) 1997,1999,2006 Free Software Foundation, Inc.
+ This file is part of GCC.
+ Contributed by Richard Henderson (rth@cygnus.com) and
+ Jakub Jelinek (jj@ultra.linux.cz).
+
+ GCC is free software; you can redistribute it and/or modify it under
+ the terms of the GNU General Public License as published by the Free
+ Software Foundation; either version 2, or (at your option) any later
+ version.
+
+ In addition to the permissions in the GNU 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 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.)
+
+ GCC 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 General Public License
+ for more details.
+
+ You should have received a copy of the GNU General Public License
+ along with GCC; see the file COPYING. If not, write to the Free
+ Software Foundation, 51 Franklin Street, Fifth Floor, Boston, MA
+ 02110-1301, USA. */
+
+#include "soft-fp.h"
+#include "single.h"
+#include "double.h"
+
+SFtype __truncdfsf2(DFtype a)
+{
+ FP_DECL_EX;
+ FP_DECL_D(A);
+ FP_DECL_S(R);
+ SFtype r;
+
+ FP_INIT_ROUNDMODE;
+ FP_UNPACK_SEMIRAW_D(A, a);
+#if _FP_W_TYPE_SIZE < _FP_FRACBITS_D
+ FP_TRUNC(S,D,1,2,R,A);
+#else
+ FP_TRUNC(S,D,1,1,R,A);
+#endif
+ FP_PACK_SEMIRAW_S(r, R);
+ FP_HANDLE_EXCEPTIONS;
+
+ return r;
+}
diff -rupN GCC.orig/gcc/config/soft-fp/trunctfdf2.c GCC/gcc/config/soft-fp/trunctfdf2.c
--- GCC.orig/gcc/config/soft-fp/trunctfdf2.c 1970-01-01 00:00:00.000000000 +0000
+++ GCC/gcc/config/soft-fp/trunctfdf2.c 2006-02-18 02:11:04.000000000 +0000
@@ -0,0 +1,54 @@
+/* Software floating-point emulation.
+ Truncate IEEE quad into IEEE double
+ Copyright (C) 1997,1999,2006 Free Software Foundation, Inc.
+ This file is part of GCC.
+ Contributed by Richard Henderson (rth@cygnus.com) and
+ Jakub Jelinek (jj@ultra.linux.cz).
+
+ GCC is free software; you can redistribute it and/or modify it under
+ the terms of the GNU General Public License as published by the Free
+ Software Foundation; either version 2, or (at your option) any later
+ version.
+
+ In addition to the permissions in the GNU 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 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.)
+
+ GCC 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 General Public License
+ for more details.
+
+ You should have received a copy of the GNU General Public License
+ along with GCC; see the file COPYING. If not, write to the Free
+ Software Foundation, 51 Franklin Street, Fifth Floor, Boston, MA
+ 02110-1301, USA. */
+
+#include "soft-fp.h"
+#include "double.h"
+#include "quad.h"
+
+DFtype __trunctfdf2(TFtype a)
+{
+ FP_DECL_EX;
+ FP_DECL_Q(A);
+ FP_DECL_D(R);
+ DFtype r;
+
+ FP_INIT_ROUNDMODE;
+ FP_UNPACK_SEMIRAW_Q(A, a);
+#if (2 * _FP_W_TYPE_SIZE) < _FP_FRACBITS_Q
+ FP_TRUNC(D,Q,2,4,R,A);
+#else
+ FP_TRUNC(D,Q,1,2,R,A);
+#endif
+ FP_PACK_SEMIRAW_D(r, R);
+ FP_HANDLE_EXCEPTIONS;
+
+ return r;
+}
diff -rupN GCC.orig/gcc/config/soft-fp/trunctfsf2.c GCC/gcc/config/soft-fp/trunctfsf2.c
--- GCC.orig/gcc/config/soft-fp/trunctfsf2.c 1970-01-01 00:00:00.000000000 +0000
+++ GCC/gcc/config/soft-fp/trunctfsf2.c 2006-02-18 02:11:04.000000000 +0000
@@ -0,0 +1,54 @@
+/* Software floating-point emulation.
+ Truncate IEEE quad into IEEE single
+ Copyright (C) 1997,1999,2006 Free Software Foundation, Inc.
+ This file is part of GCC.
+ Contributed by Richard Henderson (rth@cygnus.com) and
+ Jakub Jelinek (jj@ultra.linux.cz).
+
+ GCC is free software; you can redistribute it and/or modify it under
+ the terms of the GNU General Public License as published by the Free
+ Software Foundation; either version 2, or (at your option) any later
+ version.
+
+ In addition to the permissions in the GNU 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 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.)
+
+ GCC 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 General Public License
+ for more details.
+
+ You should have received a copy of the GNU General Public License
+ along with GCC; see the file COPYING. If not, write to the Free
+ Software Foundation, 51 Franklin Street, Fifth Floor, Boston, MA
+ 02110-1301, USA. */
+
+#include "soft-fp.h"
+#include "single.h"
+#include "quad.h"
+
+SFtype __trunctfsf2(TFtype a)
+{
+ FP_DECL_EX;
+ FP_DECL_Q(A);
+ FP_DECL_S(R);
+ SFtype r;
+
+ FP_INIT_ROUNDMODE;
+ FP_UNPACK_SEMIRAW_Q(A, a);
+#if (2 * _FP_W_TYPE_SIZE) < _FP_FRACBITS_Q
+ FP_TRUNC(S,Q,1,4,R,A);
+#else
+ FP_TRUNC(S,Q,1,2,R,A);
+#endif
+ FP_PACK_SEMIRAW_S(r, R);
+ FP_HANDLE_EXCEPTIONS;
+
+ return r;
+}
diff -rupN GCC.orig/gcc/config/soft-fp/unorddf2.c GCC/gcc/config/soft-fp/unorddf2.c
--- GCC.orig/gcc/config/soft-fp/unorddf2.c 1970-01-01 00:00:00.000000000 +0000
+++ GCC/gcc/config/soft-fp/unorddf2.c 2006-02-18 02:11:04.000000000 +0000
@@ -0,0 +1,45 @@
+/* Software floating-point emulation.
+ Return 1 iff a or b is a NaN, 0 otherwise.
+ Copyright (C) 2006 Free Software Foundation, Inc.
+ This file is part of GCC.
+ Contributed by Joseph Myers (joseph@codesourcery.com).
+
+ GCC is free software; you can redistribute it and/or modify it under
+ the terms of the GNU General Public License as published by the Free
+ Software Foundation; either version 2, or (at your option) any later
+ version.
+
+ In addition to the permissions in the GNU 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 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.)
+
+ GCC 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 General Public License
+ for more details.
+
+ You should have received a copy of the GNU General Public License
+ along with GCC; see the file COPYING. If not, write to the Free
+ Software Foundation, 51 Franklin Street, Fifth Floor, Boston, MA
+ 02110-1301, USA. */
+
+#include "soft-fp.h"
+#include "double.h"
+
+int
+__unorddf2(DFtype a, DFtype b)
+{
+ FP_DECL_D(A); FP_DECL_D(B);
+ int r;
+
+ FP_UNPACK_RAW_D(A, a);
+ FP_UNPACK_RAW_D(B, b);
+ FP_CMP_UNORD_D(r, A, B);
+
+ return r;
+}
diff -rupN GCC.orig/gcc/config/soft-fp/unordsf2.c GCC/gcc/config/soft-fp/unordsf2.c
--- GCC.orig/gcc/config/soft-fp/unordsf2.c 1970-01-01 00:00:00.000000000 +0000
+++ GCC/gcc/config/soft-fp/unordsf2.c 2006-02-18 02:11:04.000000000 +0000
@@ -0,0 +1,46 @@
+/* Software floating-point emulation.
+ Return 1 iff a or b is a NaN, 0 otherwise.
+ Copyright (C) 2006 Free Software Foundation, Inc.
+ This file is part of GCC.
+ Contributed by Joseph Myers (joseph@codesourcery.com).
+
+ GCC is free software; you can redistribute it and/or modify it under
+ the terms of the GNU General Public License as published by the Free
+ Software Foundation; either version 2, or (at your option) any later
+ version.
+
+ In addition to the permissions in the GNU 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 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.)
+
+ GCC 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 General Public License
+ for more details.
+
+ You should have received a copy of the GNU General Public License
+ along with GCC; see the file COPYING. If not, write to the Free
+ Software Foundation, 51 Franklin Street, Fifth Floor, Boston, MA
+ 02110-1301, USA. */
+
+#include "soft-fp.h"
+#include "single.h"
+
+int
+__unordsf2(SFtype a, SFtype b)
+{
+ FP_DECL_S(A);
+ FP_DECL_S(B);
+ int r;
+
+ FP_UNPACK_RAW_S(A, a);
+ FP_UNPACK_RAW_S(B, b);
+ FP_CMP_UNORD_S(r, A, B);
+
+ return r;
+}
diff -rupN GCC.orig/gcc/config/soft-fp/unordtf2.c GCC/gcc/config/soft-fp/unordtf2.c
--- GCC.orig/gcc/config/soft-fp/unordtf2.c 1970-01-01 00:00:00.000000000 +0000
+++ GCC/gcc/config/soft-fp/unordtf2.c 2006-02-18 02:11:04.000000000 +0000
@@ -0,0 +1,46 @@
+/* Software floating-point emulation.
+ Return 1 iff a or b is a NaN, 0 otherwise.
+ Copyright (C) 2006 Free Software Foundation, Inc.
+ This file is part of GCC.
+ Contributed by Joseph Myers (joseph@codesourcery.com).
+
+ GCC is free software; you can redistribute it and/or modify it under
+ the terms of the GNU General Public License as published by the Free
+ Software Foundation; either version 2, or (at your option) any later
+ version.
+
+ In addition to the permissions in the GNU 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 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.)
+
+ GCC 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 General Public License
+ for more details.
+
+ You should have received a copy of the GNU General Public License
+ along with GCC; see the file COPYING. If not, write to the Free
+ Software Foundation, 51 Franklin Street, Fifth Floor, Boston, MA
+ 02110-1301, USA. */
+
+#include "soft-fp.h"
+#include "quad.h"
+
+int
+__unordtf2(TFtype a, TFtype b)
+{
+ FP_DECL_Q(A);
+ FP_DECL_Q(B);
+ int r;
+
+ FP_UNPACK_RAW_Q(A, a);
+ FP_UNPACK_RAW_Q(B, b);
+ FP_CMP_UNORD_Q(r, A, B);
+
+ return r;
+}
diff -rupN GCC.orig/gcc/config.gcc GCC/gcc/config.gcc
--- GCC.orig/gcc/config.gcc 2006-03-07 13:59:53.000000000 +0000
+++ GCC/gcc/config.gcc 2006-03-11 21:50:17.000000000 +0000
@@ -1637,7 +1637,7 @@ pdp11-*-*)
;;
# port not yet contributed
#powerpc-*-openbsd*)
-# tmake_file="${tmake_file} rs6000/t-fprules "
+# tmake_file="${tmake_file} rs6000/t-fprules rs6000/t-fprules-fpbit "
# extra_headers=
# ;;
powerpc64-*-linux*)
@@ -1649,16 +1649,16 @@ powerpc64-*-linux*)
tm_file="rs6000/secureplt.h ${tm_file}"
fi
extra_options="${extra_options} rs6000/sysv4.opt rs6000/linux64.opt"
- tmake_file="t-dfprules rs6000/t-fprules ${tmake_file} rs6000/t-ppccomm rs6000/t-linux64"
+ tmake_file="t-dfprules rs6000/t-fprules ${tmake_file} rs6000/t-ppccomm rs6000/t-linux64 rs6000/t-fprules-softfp soft-fp/t-softfp"
;;
powerpc64-*-gnu*)
tm_file="${cpu_type}/${cpu_type}.h elfos.h svr4.h freebsd-spec.h gnu.h rs6000/sysv4.h rs6000/linux64.h rs6000/gnu.h"
extra_options="${extra_options} rs6000/sysv4.opt rs6000/linux64.opt"
- tmake_file="rs6000/t-fprules t-slibgcc-elf-ver t-gnu rs6000/t-linux64"
+ tmake_file="rs6000/t-fprules t-slibgcc-elf-ver t-gnu rs6000/t-linux64 rs6000/t-fprules-softfp soft-fp/t-softfp"
;;
powerpc-*-beos*)
tm_file="${tm_file} rs6000/aix.h rs6000/beos.h rs6000/xcoff.h"
- tmake_file="rs6000/t-fprules rs6000/t-beos"
+ tmake_file="rs6000/t-fprules rs6000/t-fprules-fpbit rs6000/t-beos"
extra_headers=
use_fixproto=yes
;;
@@ -1679,7 +1679,7 @@ powerpc-*-darwin*)
;;
powerpc*-*-freebsd*)
tm_file="${tm_file} dbxelf.h elfos.h ${fbsd_tm_file} rs6000/sysv4.h rs6000/freebsd.h"
- tmake_file="rs6000/t-fprules rs6000/t-ppcos ${tmake_file} rs6000/t-ppccomm"
+ tmake_file="rs6000/t-fprules rs6000/t-fprules-fpbit rs6000/t-ppcos ${tmake_file} rs6000/t-ppccomm"
extra_options="${extra_options} rs6000/sysv4.opt"
;;
powerpc-*-netbsd*)
@@ -1689,7 +1689,7 @@ powerpc-*-netbsd*)
;;
powerpc-*-chorusos*)
tm_file="${tm_file} dbxelf.h elfos.h svr4.h freebsd-spec.h rs6000/sysv4.h chorus.h"
- tmake_file="rs6000/t-fprules rs6000/t-ppcos rs6000/t-ppccomm"
+ tmake_file="rs6000/t-fprules rs6000/t-fprules-fpbit rs6000/t-ppcos rs6000/t-ppccomm"
extra_options="${extra_options} rs6000/sysv4.opt"
case ${enable_threads} in
yes | posix)
@@ -1706,48 +1706,48 @@ powerpc-*-eabispe*)
powerpc-*-eabisimaltivec*)
tm_file="${tm_file} dbxelf.h elfos.h svr4.h freebsd-spec.h rs6000/sysv4.h rs6000/eabi.h rs6000/eabisim.h rs6000/eabialtivec.h"
extra_options="${extra_options} rs6000/sysv4.opt"
- tmake_file="rs6000/t-fprules rs6000/t-ppcendian rs6000/t-ppccomm"
+ tmake_file="rs6000/t-fprules rs6000/t-fprules-fpbit rs6000/t-ppcendian rs6000/t-ppccomm"
;;
powerpc-*-eabisim*)
tm_file="${tm_file} dbxelf.h elfos.h svr4.h freebsd-spec.h rs6000/sysv4.h rs6000/eabi.h rs6000/eabisim.h"
extra_options="${extra_options} rs6000/sysv4.opt"
- tmake_file="rs6000/t-fprules rs6000/t-ppcgas rs6000/t-ppccomm"
+ tmake_file="rs6000/t-fprules rs6000/t-fprules-fpbit rs6000/t-ppcgas rs6000/t-ppccomm"
;;
powerpc-*-elf*)
tm_file="${tm_file} dbxelf.h elfos.h svr4.h freebsd-spec.h rs6000/sysv4.h"
extra_options="${extra_options} rs6000/sysv4.opt"
- tmake_file="rs6000/t-fprules rs6000/t-ppcgas rs6000/t-ppccomm"
+ tmake_file="rs6000/t-fprules rs6000/t-fprules-fpbit rs6000/t-ppcgas rs6000/t-ppccomm"
use_fixproto=yes
;;
powerpc-*-eabialtivec*)
tm_file="${tm_file} dbxelf.h elfos.h svr4.h freebsd-spec.h rs6000/sysv4.h rs6000/eabi.h rs6000/eabialtivec.h"
extra_options="${extra_options} rs6000/sysv4.opt"
- tmake_file="rs6000/t-fprules rs6000/t-ppcendian rs6000/t-ppccomm"
+ tmake_file="rs6000/t-fprules rs6000/t-fprules-fpbit rs6000/t-ppcendian rs6000/t-ppccomm"
;;
powerpc-*-eabi*)
tm_file="${tm_file} dbxelf.h elfos.h svr4.h freebsd-spec.h rs6000/sysv4.h rs6000/eabi.h"
extra_options="${extra_options} rs6000/sysv4.opt"
- tmake_file="rs6000/t-fprules rs6000/t-ppcgas rs6000/t-ppccomm"
+ tmake_file="rs6000/t-fprules rs6000/t-fprules-fpbit rs6000/t-ppcgas rs6000/t-ppccomm"
;;
powerpc-*-rtems*)
tm_file="${tm_file} dbxelf.h elfos.h svr4.h freebsd-spec.h rs6000/sysv4.h rs6000/eabi.h rs6000/rtems.h rtems.h"
extra_options="${extra_options} rs6000/sysv4.opt"
- tmake_file="rs6000/t-fprules rs6000/t-rtems t-rtems rs6000/t-ppccomm"
+ tmake_file="rs6000/t-fprules rs6000/t-fprules-fpbit rs6000/t-rtems t-rtems rs6000/t-ppccomm"
;;
powerpc-*-linux*altivec*)
tm_file="${tm_file} dbxelf.h elfos.h svr4.h freebsd-spec.h rs6000/sysv4.h rs6000/linux.h rs6000/linuxaltivec.h"
extra_options="${extra_options} rs6000/sysv4.opt"
- tmake_file="rs6000/t-fprules rs6000/t-ppcos ${tmake_file} rs6000/t-ppccomm"
+ tmake_file="rs6000/t-fprules rs6000/t-fprules-softfp soft-fp/t-softfp rs6000/t-ppcos ${tmake_file} rs6000/t-ppccomm"
;;
powerpc-*-linux*spe*)
tm_file="${tm_file} dbxelf.h elfos.h svr4.h freebsd-spec.h rs6000/sysv4.h rs6000/linux.h rs6000/linuxspe.h"
extra_options="${extra_options} rs6000/sysv4.opt"
- tmake_file="rs6000/t-fprules rs6000/t-ppcos ${tmake_file} rs6000/t-ppccomm"
+ tmake_file="rs6000/t-fprules rs6000/t-fprules-softfp soft-fp/t-softfp rs6000/t-ppcos ${tmake_file} rs6000/t-ppccomm"
;;
powerpc-*-linux*)
tm_file="${tm_file} dbxelf.h elfos.h svr4.h freebsd-spec.h rs6000/sysv4.h"
extra_options="${extra_options} rs6000/sysv4.opt"
- tmake_file="t-dfprules rs6000/t-fprules rs6000/t-ppcos ${tmake_file} rs6000/t-ppccomm"
+ tmake_file="t-dfprules rs6000/t-fprules rs6000/t-fprules-softfp soft-fp/t-softfp rs6000/t-ppcos ${tmake_file} rs6000/t-ppccomm"
case ${enable_targets}:${cpu_is_64bit} in
*powerpc64* | all:* | *:yes)
if test x$cpu_is_64bit = xyes; then
@@ -1768,14 +1768,14 @@ powerpc-*-linux*)
powerpc-*-gnu-gnualtivec*)
tm_file="${cpu_type}/${cpu_type}.h elfos.h svr4.h freebsd-spec.h gnu.h rs6000/sysv4.h rs6000/linux.h rs6000/linuxaltivec.h rs6000/gnu.h"
extra_options="${extra_options} rs6000/sysv4.opt"
- tmake_file="rs6000/t-fprules rs6000/t-ppcos t-slibgcc-elf-ver t-gnu rs6000/t-ppccomm"
+ tmake_file="rs6000/t-fprules rs6000/t-fprules-fpbit rs6000/t-ppcos t-slibgcc-elf-ver t-gnu rs6000/t-ppccomm"
if test x$enable_threads = xyes; then
thread_file='posix'
fi
;;
powerpc-*-gnu*)
tm_file="${cpu_type}/${cpu_type}.h elfos.h svr4.h freebsd-spec.h gnu.h rs6000/sysv4.h rs6000/linux.h rs6000/gnu.h"
- tmake_file="rs6000/t-fprules rs6000/t-ppcos t-slibgcc-elf-ver t-gnu rs6000/t-ppccomm"
+ tmake_file="rs6000/t-fprules rs6000/t-fprules-fpbit rs6000/t-ppcos t-slibgcc-elf-ver t-gnu rs6000/t-ppccomm"
extra_options="${extra_options} rs6000/sysv4.opt"
if test x$enable_threads = xyes; then
thread_file='posix'
@@ -1785,7 +1785,7 @@ powerpc-wrs-vxworks|powerpc-wrs-vxworksa
# We want vxworks.h after rs6000/sysv4.h, which unfortunately
# means we have to redo the tm_file list from scratch.
tm_file="rs6000/rs6000.h elfos.h svr4.h freebsd-spec.h rs6000/sysv4.h"
- tmake_file="${tmake_file} rs6000/t-fprules rs6000/t-ppccomm rs6000/t-vxworks"
+ tmake_file="${tmake_file} rs6000/t-fprules rs6000/t-fprules-fpbit rs6000/t-ppccomm rs6000/t-vxworks"
extra_options="${extra_options} rs6000/sysv4.opt"
extra_headers=ppc-asm.h
case ${target} in
@@ -1800,7 +1800,7 @@ powerpc-wrs-vxworks|powerpc-wrs-vxworksa
;;
powerpc-wrs-windiss*) # Instruction-level simulator for VxWorks.
tm_file="${tm_file} elfos.h svr4.h freebsd-spec.h rs6000/sysv4.h rs6000/windiss.h"
- tmake_file="rs6000/t-fprules rs6000/t-ppcgas rs6000/t-ppccomm"
+ tmake_file="rs6000/t-fprules rs6000/t-fprules-fpbit rs6000/t-ppcgas rs6000/t-ppccomm"
extra_options="${extra_options} rs6000/sysv4.opt"
thread_file=""
use_fixproto=yes
@@ -1818,39 +1818,39 @@ powerpc-*-lynxos*)
;;
powerpcle-*-sysv*)
tm_file="${tm_file} dbxelf.h elfos.h svr4.h freebsd-spec.h rs6000/sysv4.h rs6000/sysv4le.h"
- tmake_file="rs6000/t-fprules rs6000/t-ppcos rs6000/t-ppccomm"
+ tmake_file="rs6000/t-fprules rs6000/t-fprules-fpbit rs6000/t-ppcos rs6000/t-ppccomm"
extra_options="${extra_options} rs6000/sysv4.opt"
use_fixproto=yes
;;
powerpcle-*-elf*)
tm_file="${tm_file} dbxelf.h elfos.h svr4.h freebsd-spec.h rs6000/sysv4.h rs6000/sysv4le.h"
- tmake_file="rs6000/t-fprules rs6000/t-ppcgas rs6000/t-ppccomm"
+ tmake_file="rs6000/t-fprules rs6000/t-fprules-fpbit rs6000/t-ppcgas rs6000/t-ppccomm"
extra_options="${extra_options} rs6000/sysv4.opt"
use_fixproto=yes
;;
powerpcle-*-eabisim*)
tm_file="${tm_file} dbxelf.h elfos.h svr4.h freebsd-spec.h rs6000/sysv4.h rs6000/sysv4le.h rs6000/eabi.h rs6000/eabisim.h"
- tmake_file="rs6000/t-fprules rs6000/t-ppcgas rs6000/t-ppccomm"
+ tmake_file="rs6000/t-fprules rs6000/t-fprules-fpbit rs6000/t-ppcgas rs6000/t-ppccomm"
extra_options="${extra_options} rs6000/sysv4.opt"
;;
powerpcle-*-eabi*)
tm_file="${tm_file} dbxelf.h elfos.h svr4.h freebsd-spec.h rs6000/sysv4.h rs6000/sysv4le.h rs6000/eabi.h"
- tmake_file="rs6000/t-fprules rs6000/t-ppcgas rs6000/t-ppccomm"
+ tmake_file="rs6000/t-fprules rs6000/t-fprules-fpbit rs6000/t-ppcgas rs6000/t-ppccomm"
extra_options="${extra_options} rs6000/sysv4.opt"
;;
powerpc-*-kaos*)
tm_file="${tm_file} dbxelf.h elfos.h svr4.h freebsd-spec.h rs6000/sysv4.h kaos.h rs6000/kaos-ppc.h"
- tmake_file="rs6000/t-fprules rs6000/t-ppcgas rs6000/t-ppccomm"
+ tmake_file="rs6000/t-fprules rs6000/t-fprules-fpbit rs6000/t-ppcgas rs6000/t-ppccomm"
extra_options="${extra_options} rs6000/sysv4.opt"
;;
powerpcle-*-kaos*)
tm_file="${tm_file} dbxelf.h elfos.h svr4.h freebsd-spec.h rs6000/sysv4.h rs6000/sysv4le.h kaos.h rs6000/kaos-ppc.h"
- tmake_file="rs6000/t-fprules rs6000/t-ppcgas rs6000/t-ppccomm"
+ tmake_file="rs6000/t-fprules rs6000/t-fprules-fpbit rs6000/t-ppcgas rs6000/t-ppccomm"
extra_options="${extra_options} rs6000/sysv4.opt"
;;
rs6000-ibm-aix4.[12]* | powerpc-ibm-aix4.[12]*)
tm_file="${tm_file} rs6000/aix.h rs6000/aix41.h rs6000/xcoff.h"
- tmake_file="rs6000/t-fprules rs6000/t-newas"
+ tmake_file="rs6000/t-fprules rs6000/t-fprules-fpbit rs6000/t-newas"
extra_options="${extra_options} rs6000/aix41.opt"
use_collect2=yes
extra_headers=
diff -rupN GCC.orig/gcc/mklibgcc.in GCC/gcc/mklibgcc.in
--- GCC.orig/gcc/mklibgcc.in 2006-03-07 13:59:52.000000000 +0000
+++ GCC/gcc/mklibgcc.in 2006-03-11 18:31:17.000000000 +0000
@@ -11,6 +11,7 @@
# objext
# LIB1ASMFUNCS
# LIB2FUNCS_ST
+# LIB2FUNCS_EXCLUDE
# LIBGCOV
# LIB2ADD
# LIB2ADD_ST
@@ -158,8 +159,9 @@ else
fi
# Remove any objects from lib2funcs and LIB2_DIVMOD_FUNCS that are
-# defined as optimized assembly code in LIB1ASMFUNCS.
-for name in $LIB1ASMFUNCS; do
+# defined as optimized assembly code in LIB1ASMFUNCS or as C code
+# in LIB2FUNCS_EXCLUDE.
+for name in $LIB1ASMFUNCS $LIB2FUNCS_EXCLUDE; do
lib2funcs=`echo $lib2funcs | sed -e 's/^'$name' //' \
-e 's/ '$name' / /' \
-e 's/ '$name'$//'`
--- codingconventions.html.orig 2005-10-30 12:23:24.000000000 +0000
+++ codingconventions.html 2006-03-11 23:40:48.000000000 +0000
@@ -462,6 +462,12 @@
symlink-tree and ylwrap: These are copied from mainline automake, using
<code>automake --add-missing --copy --force-missing</code>.</li>
+<li>gcc/config/soft-fp: The master sources (except for t-softfp) are
+in CVS glibc, and changes should go into CVS glibc before going into
+GCC. (GCC uses these files under GPL+exception where glibc uses them
+under LGPL; this should be the only difference between the two copies
+of the files.)</li>
+
<li>fastjar: The master sources were at <a
href="http://fastjar.sourceforge.net/";>fastjar.sourceforge.net</a>.
However, the upstream source seems to be dead, so fastjar is
--
Joseph S. Myers http://www.srcf.ucam.org/~jsm28/gcc/
jsm@polyomino.org.uk (personal mail)
joseph@codesourcery.com (CodeSourcery mail)
jsm28@gcc.gnu.org (Bugzilla assignments and CCs)