This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
Reverse reversed_comparison_code_parts patch
- To: gcc-patches at gcc dot gnu dot org
- Subject: Reverse reversed_comparison_code_parts patch
- From: Jeffrey Oldham <oldham at codesourcery dot com>
- Date: Mon, 12 Mar 2001 11:18:52 -0800
- Cc: Jeffrey Oldham <oldham at codesourcery dot com>, jh at suse dot cz, stewart at dystopia dot lab43 dot org, rth at redhat dot com, rearnsha at arm dot com
- Organization: CodeSourcery LLC
This patch reverses a patch that has broken mainline compilation on
mips-sgi-irix6.5 and arm-linux since 01 March. See
http://gcc.gnu.org/ml/gcc-patches/2001-03/msg00660.html and
http://gcc.gnu.org/ml/gcc-patches/2001-03/msg00027.html.
2001-03-12 Jeffrey Oldham <oldham@codesourcery.com>
* jump.c (reversed_comparison_code_parts): Reverse Jan Hubricka's
2001-02-28 patch because it breaks arm-linux and mips-sgi-irix6.5.
Tested on: i686-pc-linux-gnu and mips-sgi-irix6.5
): Bootstrapping is still broken by other changes. :(
Approved by: Mark Mitchell (mark@codesourcery.com)
Thanks,
Jeffrey D. Oldham
oldham@codesourcery.com
Index: jump.c
===================================================================
RCS file: /cvs/gcc/gcc/gcc/jump.c,v
retrieving revision 1.158
diff -c -p -r1.158 jump.c
*** jump.c 2001/03/07 19:29:36 1.158
--- jump.c 2001/03/12 15:53:13
*************** reversed_comparison_code_parts (code, ar
*** 1782,1792 ****
case UNLE:
case UNGT:
case UNGE:
! /* We don't have safe way to reverse these yet - we would need
! ordered compares that may not trap. */
! if (TARGET_FLOAT_FORMAT != IEEE_FLOAT_FORMAT
! || flag_unsafe_math_optimizations)
! return reverse_condition_maybe_unordered (code);
return UNKNOWN;
default:
break;
--- 1782,1788 ----
case UNLE:
case UNGT:
case UNGE:
! /* We don't have safe way to reverse these yet. */
return UNKNOWN;
default:
break;
*************** reversed_comparison_code_parts (code, ar
*** 1844,1855 ****
return UNKNOWN;
}
}
-
- /* In case of floating point modes, we may reverse here, since
- we will be always converting an ordered compare to unordered.
- The unsafe code has been caught earlier. */
- if (FLOAT_MODE_P (mode))
- return reverse_condition_maybe_unordered (code);
/* An integer condition. */
if (GET_CODE (arg0) == CONST_INT
--- 1840,1845 ----