Patchwork Fix a fallout of my BIT_AND_EXPR optimization patch (PR middle-end/45876)

On Mon, 4 Oct 2010, Jakub Jelinek wrote:

> Hi!
> 
> As the testcase shows, there were some type issues in my PR31261
> patch.  Fixed thusly, bootstrapped/regtested on x86_64-linux and i686-linux,
> ok for trunk?

Ok.

Thanks,
Richard.

> 2010-10-04  Jakub Jelinek  <jakub@redhat.com>
> 
> 	PR middle-end/45876
> 	* fold-const.c (fold_binary_loc) <case BIT_AND_EXPR>: Use
> 	arg0's type or its unsigned counterpart as utype.  Convert
> 	arg1 to utype unconditionally.
> 
> 	* gcc.c-torture/compile/pr45876.c: New test.
> 
> --- gcc/fold-const.c.jj	2010-10-01 12:39:42.000000000 +0200
> +++ gcc/fold-const.c	2010-10-04 11:22:18.000000000 +0200
> @@ -11152,12 +11152,12 @@ fold_binary_loc (location_t loc,
>  		  || (TREE_CODE (arg0) != NEGATE_EXPR
>  		      && pmop[1] != TREE_OPERAND (arg0, 1)))
>  		{
> -		  tree utype = type;
> +		  tree utype = TREE_TYPE (arg0);
>  		  if (! TYPE_OVERFLOW_WRAPS (TREE_TYPE (arg0)))
>  		    {
>  		      /* Perform the operations in a type that has defined
>  			 overflow behavior.  */
> -		      utype = unsigned_type_for (type);
> +		      utype = unsigned_type_for (TREE_TYPE (arg0));
>  		      if (pmop[0] != NULL)
>  			pmop[0] = fold_convert_loc (loc, utype, pmop[0]);
>  		      if (pmop[1] != NULL)
> @@ -11184,16 +11184,9 @@ fold_binary_loc (location_t loc,
>  		    tem = fold_build2_loc (loc, MINUS_EXPR, utype,
>  					   pmop[0], pmop[1]);
>  		  /* TEM is now the new binary +, - or unary - replacement.  */
> -		  if (utype == type)
> -		    return fold_build2_loc (loc, BIT_AND_EXPR, type,
> -					    tem, arg1);
> -		  else
> -		    {
> -		      tem = fold_build2_loc (loc, BIT_AND_EXPR, utype, tem,
> -					     fold_convert_loc (loc, utype,
> -					     arg1));		
> -		      return fold_convert_loc (loc, type, tem);
> -		    }
> +		  tem = fold_build2_loc (loc, BIT_AND_EXPR, utype, tem,
> +					 fold_convert_loc (loc, utype, arg1));
> +		  return fold_convert_loc (loc, type, tem);
>  		}
>  	    }
>  	}
> --- gcc/testsuite/gcc.c-torture/compile/pr45876.c.jj	2010-10-04 11:18:00.000000000 +0200
> +++ gcc/testsuite/gcc.c-torture/compile/pr45876.c	2010-10-04 11:17:40.000000000 +0200
> @@ -0,0 +1,9 @@
> +/* PR middle-end/45876 */
> +
> +unsigned
> +foo (unsigned x)
> +{
> +  short i = 0;
> +  i = ((short) (((((unsigned) i) >> 1) & 16383) + x)) & 16383;
> +  return i;
> +}
> 
> 	Jakub
> 
>

Hi!

As the testcase shows, there were some type issues in my PR31261
patch.  Fixed thusly, bootstrapped/regtested on x86_64-linux and i686-linux,
ok for trunk?

2010-10-04  Jakub Jelinek  <jakub@redhat.com>

	PR middle-end/45876
	* fold-const.c (fold_binary_loc) <case BIT_AND_EXPR>: Use
	arg0's type or its unsigned counterpart as utype.  Convert
	arg1 to utype unconditionally.

	* gcc.c-torture/compile/pr45876.c: New test.


	Jakub