[PATCH] SPU float format: Round towards zero

[Ulrich Weigand]

Joseph S. Myers wrote: 
> On Mon, 4 Aug 2008, Ulrich Weigand wrote:
> 
> > In addition, it turned out that the SPU back-end didn't always follow
> > round-towards-zero itself: for DImode to SFmode conversions, the default
> > libgcc routine was used (which implements round-to-nearest).  The patch
> > adds inline implementations of those conversions that round towards zero,
> > as already specified in the SPU Language Extensions specification.
> 
> If the C libgcc function is not correct for this target, you should stop 
> it being built - and to support existing objects that may refer to it, you 
> should provide an SPU-specific implementation (whether in C or assembly).

Good point, thanks.  The following patch (on top of the previous ones)
implements this suggestion.

Trevor, is this OK with you after the other patches go in?

Bye,
Ulrich


ChangeLog:

	* config/spu/float_disf.c: New file.
	* config/spu/float_unsdisf.c: New file.
	* config/spu/t-elf (LIB2FUNCS_STATIC_EXTRA): Add them.
	(LIB2FUNCS_EXCLUDE): Define.

diff -urNp -x .svn gcc-head-orig/gcc/config/spu/float_disf.c gcc-head/gcc/config/spu/float_disf.c
--- gcc-head-orig/gcc/config/spu/float_disf.c	1970-01-01 01:00:00.000000000 +0100
+++ gcc-head/gcc/config/spu/float_disf.c	2008-08-08 14:34:24.000000000 +0200
@@ -0,0 +1,30 @@
+/* Copyright (C) 2008 Free Software Foundation, Inc.
+  
+   This file 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 of the License, or (at your option)
+   any later version.
+  
+   This file 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 this file; see the file COPYING.  If not, write to the Free
+   Software Foundation, 51 Franklin Street, Fifth Floor, Boston, MA
+   02110-1301, USA.  */
+
+/* As a special exception, if you link this library with files compiled with
+   GCC to produce an executable, this does not cause the resulting executable
+   to be covered by the GNU General Public License.  The exception does not
+   however invalidate any other reasons why the executable file might be covered
+   by the GNU General Public License. */
+
+float _floatdisf (long long x)
+{
+  /* The SPU back-end now generates inline code for this conversion.
+     This file is solely used to provide the _floatdisf functions
+     for objects generated with prior versions of GCC.  */
+  return x;
+}
diff -urNp -x .svn gcc-head-orig/gcc/config/spu/float_unsdisf.c gcc-head/gcc/config/spu/float_unsdisf.c
--- gcc-head-orig/gcc/config/spu/float_unsdisf.c	1970-01-01 01:00:00.000000000 +0100
+++ gcc-head/gcc/config/spu/float_unsdisf.c	2008-08-08 18:13:57.000000000 +0200
@@ -0,0 +1,30 @@
+/* Copyright (C) 2008 Free Software Foundation, Inc.
+  
+   This file 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 of the License, or (at your option)
+   any later version.
+  
+   This file 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 this file; see the file COPYING.  If not, write to the Free
+   Software Foundation, 51 Franklin Street, Fifth Floor, Boston, MA
+   02110-1301, USA.  */
+
+/* As a special exception, if you link this library with files compiled with
+   GCC to produce an executable, this does not cause the resulting executable
+   to be covered by the GNU General Public License.  The exception does not
+   however invalidate any other reasons why the executable file might be covered
+   by the GNU General Public License. */
+
+float _floatundisf (unsigned long long x)
+{
+  /* The SPU back-end now generates inline code for this conversion.
+     This file is solely used to provide the _floatundisf function
+     for objects generated with prior versions of GCC.  */
+  return x;
+}
diff -urNp -x .svn gcc-head-orig/gcc/config/spu/t-spu-elf gcc-head/gcc/config/spu/t-spu-elf
--- gcc-head-orig/gcc/config/spu/t-spu-elf	2008-08-09 14:42:12.000000000 +0200
+++ gcc-head/gcc/config/spu/t-spu-elf	2008-08-08 18:14:09.000000000 +0200
@@ -21,8 +21,15 @@ CROSS_LIBGCC1 =
 
 TARGET_LIBGCC2_CFLAGS = -fPIC -mwarn-reloc -D__IN_LIBGCC2
 
+# We exclude those because the libgcc2.c default versions do not support
+# the SPU single-precision format (round towards zero).  We provide our
+# own versions below.
+LIB2FUNCS_EXCLUDE = _floatdisf _floatundisf
+
 LIB2FUNCS_STATIC_EXTRA = $(srcdir)/config/spu/float_unssidf.c \
 			 $(srcdir)/config/spu/float_unsdidf.c \
+			 $(srcdir)/config/spu/float_unsdisf.c \
+			 $(srcdir)/config/spu/float_disf.c \
 			 $(srcdir)/config/spu/mfc_tag_table.c \
 			 $(srcdir)/config/spu/mfc_tag_reserve.c \
 			 $(srcdir)/config/spu/mfc_tag_release.c \

-- 
  Dr. Ulrich Weigand
  GNU Toolchain for Linux on System z and Cell BE
  Ulrich.Weigand@de.ibm.com