gcc-csl-2005q3: Add missing patch

1 files changed, 95 insertions, 0 deletions

diff --git a/packages/gcc/gcc-csl-arm/gcc_optab_arm.patch b/packages/gcc/gcc-csl-arm/gcc_optab_arm.patch
new file mode 100644
index 0000000000..fa21b26554
--- /dev/null
+++ b/packages/gcc/gcc-csl-arm/gcc_optab_arm.patch
@@ -0,0 +1,95 @@
+ARM is the only architecture that has a helper function that returns
+an unbiased result. This fix is trivial enough that we can show it
+doesn't effect any of the other arches. Can we consider this a
+regression fix since it used to work until the helper was added :}
+
+Tested with no regressions on x86_64-pc-linux-gnu and arm-none-eabi.
+
+Cheers,
+Carlos.
+-- 
+Carlos O'Donell
+CodeSourcery
+carlos@codesourcery.com
+(650) 331-3385 x716
+
+gcc/
+
+2006-01-27  Carlos O'Donell  <carlos@codesourcery.com>
+
+	* optabs.c (prepare_cmp_insn): If unbaised and unsigned then bias
+	the comparison routine return.
+	
+gcc/testsuite/
+
+2006-01-27  Carlos O'Donell  <carlos@codesourcery.com>
+
+	* gcc.dg/unsigned-long-compare.c: New test.
+
+Index: gcc/optabs.c
+===================================================================
+--- 1/gcc/optabs.c	(revision 110300)
++++ 2/gcc/optabs.c	(working copy)
+@@ -3711,18 +3711,24 @@
+       result = emit_library_call_value (libfunc, NULL_RTX, LCT_CONST_MAKE_BLOCK,
+ 					word_mode, 2, x, mode, y, mode);
+ 
++      /* There are two kinds of comparison routines. Biased routines
++	 return 0/1/2, and unbiased routines return -1/0/1. Other parts
++	 of gcc expect that the comparison operation is equivalent
++	 to the modified comparison. For signed comparisons compare the 
++	 result against 1 in the unbiased case, and zero in the biased
++	 case. For unsigned comparisons always compare against 1 after
++	 biasing the unbased result by adding 1. This gives us a way to
++	 represent LTU. */
+       *px = result;
+       *pmode = word_mode;
+-      if (TARGET_LIB_INT_CMP_BIASED)
+-	/* Integer comparison returns a result that must be compared
+-	   against 1, so that even if we do an unsigned compare
+-	   afterward, there is still a value that can represent the
+-	   result "less than".  */
+-	*py = const1_rtx;
+-      else
++      *py = const1_rtx;
++
++      if (!TARGET_LIB_INT_CMP_BIASED)
+ 	{
+-	  *py = const0_rtx;
+-	  *punsignedp = 1;
++	  if (*punsignedp)
++	    *px = plus_constant (result, 1);  
++	  else
++	    *py = const0_rtx;
+ 	}
+       return;
+     }
+Index: gcc/testsuite/gcc.dg/unsigned-long-compare.c
+===================================================================
+--- 1/gcc/testsuite/gcc.dg/unsigned-long-compare.c	(revision 0)
++++ 2/gcc/testsuite/gcc.dg/unsigned-long-compare.c	(revision 0)
+@@ -0,0 +1,24 @@
++/* Copyright (C) 2006 Free Software Foundation, Inc. */
++/* Contributed by Carlos O'Donell on 2006-01-27 */
++
++/* Test a division corner case where the expression simplifies
++   to a comparison, and the optab expansion is wrong. The optab 
++   expansion emits a function whose return is unbiased and needs
++   adjustment. */
++/* Origin: Carlos O'Donell <carlos@codesourcery.com> */
++/* { dg-do run { target arm-*-*eabi* } } */
++/* { dg-options "" } */
++#include <stdlib.h>
++
++#define BIG_CONSTANT 0xFFFFFFFF80000000ULL
++
++int main (void)
++{
++  unsigned long long OneULL = 1ULL;
++  unsigned long long result;
++
++  result = OneULL / BIG_CONSTANT; 
++  if (result)
++    abort ();
++  exit (0);
++}