Re: Irix6 long doubles implemented wrong? (27_io/ostream_inserter_arith)

[Alexandre Oliva <aoliva@redhat.com>]

[-- Attachment #1: Type: text/plain, Size: 435 bytes --]

On Jan 26, 2003, Alexandre Oliva <aoliva@redhat.com> wrote:

> 	* real.c (ibm_extended_format): Add 53 to minimum exponent.

Turned out this got encode_ibm_extended very confused when presented
with denormals, not only numbers that had to be represented as
denormal doubles, but also those that could still use a normal double
for the upper part.  This new patch fixes it, and actually passes the
numeric_limits test.  Ok to install?


[-- Warning: decoded text below may be mangled, UTF-8 assumed --]
[-- Attachment #2: gcc-mips-tfmode-ibmext-minexp.patch --]
[-- Type: text/x-patch, Size: 1748 bytes --]

Index: gcc/ChangeLog
from  Alexandre Oliva  <aoliva@redhat.com>

	* real.c (ibm_extended_format): Add 53 to minimum exponent.
	(encode_ibm_extended): Adjust.

Index: gcc/real.c
===================================================================
RCS file: /cvs/gcc/gcc/gcc/real.c,v
retrieving revision 1.105
diff -u -p -r1.105 real.c
--- gcc/real.c 17 Nov 2002 20:20:39 -0000 1.105
+++ gcc/real.c 26 Jan 2003 12:43:40 -0000
@@ -1,6 +1,6 @@
 /* real.c - software floating point emulation.
    Copyright (C) 1993, 1994, 1995, 1996, 1997, 1998,
-   1999, 2000, 2002 Free Software Foundation, Inc.
+   1999, 2000, 2002, 2003 Free Software Foundation, Inc.
    Contributed by Stephen L. Moshier (moshier@world.std.com).
    Re-written by Richard Henderson  <rth@redhat.com>
 
@@ -3254,8 +3254,23 @@ encode_ibm_extended (fmt, buf, r)
       u = *r;
       clear_significand_below (&u, SIGNIFICAND_BITS - 53);
 
-      /* v = remainder containing additional 53 bits of significand.  */
-      do_add (&v, r, &u, 1);
+      normalize (&u);
+      /* If the upper double is zero, we have a denormal double, so
+	 move it to the first double and leave the second as zero.  */
+      if (u.class == rvc_zero)
+	{
+	  v = u;
+	  u = *r;
+	  normalize (&u);
+	}
+      else
+	{
+	  /* v = remainder containing additional 53 bits of significand.  */
+	  do_add (&v, r, &u, 1);
+	  round_for_format (&ieee_double_format, &v);
+	}
+
+      round_for_format (&ieee_double_format, &u);
 
       encode_ieee_double (&ieee_double_format, &buf[0], &u);
       encode_ieee_double (&ieee_double_format, &buf[2], &v);
@@ -3292,7 +3307,7 @@ const struct real_format ibm_extended_fo
     2,
     1,
     53 + 53,
-    -1021,
+    -1021 + 53,
     1024,
     true,
     true,

[-- Attachment #3: Type: text/plain, Size: 289 bytes --]


-- 
Alexandre Oliva   Enjoy Guarana', see http://www.ic.unicamp.br/~oliva/
Red Hat GCC Developer                 aoliva@{redhat.com, gcc.gnu.org}
CS PhD student at IC-Unicamp        oliva@{lsd.ic.unicamp.br, gnu.org}
Free Software Evangelist                Professional serial bug killer