[patch] Fix float-unsigned long conversion on h8300 (revised)

[Kazu Hirata <kazu at hxi dot com>]

Hi,

Attached is a patch to fix conversion from float to unsigned long on
the h8300 port.

Without the patch, due to the -DDI=SI thing in t-h8300, a wrong
function, namely __fixunssfDI in libgcc2.c, is used as __fixunssfsi.

We would want to use __fixunssfsi in libgcc2.c, but it also has a
problem.  The return type of the function is in unsigned HImode even
though the function is supposed to convert a float to unsigned SImode.

The patch fixes this problem by defining fixunssfsi in
LIB2FUNCS_EXTRA.  To prevent the inclusion of fixunssfsi and
fixunssfdi from libgcc2.c, I added those to LIB1ASMFUNCS.

With the patch, execute/980605-1.c and execute/gofast.c pass.

Tested on the branch and the mainline.  OK to apply?

This patch obsoletes

http://gcc.gnu.org/ml/gcc-patches/2001-07/msg01530.html

p.s.
I don't know how to present a new file in the form of a patch, so I
simply appended it at the end of this message.

Thanks,

Kazu Hirata

2001-08-29  Kazu Hirata  <kazu@hxi.com>

	* config/h8300/h8300-fixunssfsi: New.
	* config/h8300/t-h8300 (LIB1ASMFUNCS): Add _fixunssfsi and
	_fixunssfdi.
	(LIB2FUNCS_EXTRA): New.

Index: t-h8300
===================================================================
RCS file: /cvs/gcc/gcc/gcc/config/h8300/t-h8300,v
retrieving revision 1.7
diff -u -r1.7 t-h8300
--- t-h8300	2001/07/09 22:05:44	1.7
+++ t-h8300	2001/08/29 14:17:32
@@ -1,9 +1,16 @@
+# The floating point functions listed in LIB1ASMFUNCS (_fixunssfsi,
+# _fixunssfdi, _floatdisf, and _fixsfdi) are used to disable the
+# inclusion of those from libgcc2.c.  They do not actually exist in
+# lib1funcs.asm.
 LIB1ASMSRC = h8300/lib1funcs.asm
 LIB1ASMFUNCS = _cmpsi2 _ucmpsi2 _divhi3 _divsi3 _mulhi3 _mulsi3 \
-  _floatdisf _fixsfdi
+  _fixunssfsi _fixunssfdi _floatdisf _fixsfdi
 
 # We do not have DF or DI types, so fake out the libgcc2 compilation.
 TARGET_LIBGCC2_CFLAGS = -DDF=SF -DDI=SI
+
+# fixunssfsi from libgcc2.c has a wrong return type, so use our own.
+LIB2FUNCS_EXTRA = $(srcdir)/config/h8300/h8300-fixunssfsi.c
 
 # We want fine grained libraries, so use the new code to build the
 # floating point emulation libraries.

gcc/config/h8300/h8300-fixunssfi.c
===================================================================
typedef int          SItype  __attribute__ ((mode (SI)));
typedef unsigned int USItype __attribute__ ((mode (SI)));
typedef float        SFtype  __attribute__ ((mode (SF)));

#define Wtype  SItype
#define UWtype USItype
#define Wtype_MAX ((Wtype) 0x7fffffffL)
#define Wtype_MIN (- Wtype_MAX - 1)

UWtype
__fixunssfsi (SFtype a)
{
  if (a >= - (SFtype) Wtype_MIN)
    return (Wtype) (a + Wtype_MIN) - Wtype_MIN;
  return (Wtype) a;
}