Re: [ PATCH ] Split out my fold-const.c change and add a test case

["Richard Guenther" <richard dot guenther at gmail dot com>]

On 1/5/07, Mark Mitchell <mark@codesourcery.com> wrote:
> Robert Kennedy wrote:
>
> > 2007-01-04  Robert Kennedy <jimbob@google.com>
> >       * fold-const.c (fold_comparison): Fold comparisons like (x *
> >       1000 < 0) to (x < 0).
> >       * fold-compare-2.c: New test case for fold_comparison.
>
> > +  /* Transform comparisons of the form X * C1 CMP 0 to X CMP 0 in the
> > +     signed arithmetic case.  That form is created by the compiler
> > +     often enough for folding it to be of value.  One example is in
> > +     computing loop trip counts after Operator Strength Reduction.  */
> > +  if (!(flag_wrapv || flag_trapv)
> > +      && !TYPE_UNSIGNED (TREE_TYPE (arg0))
> > +      && TREE_CODE (arg0) == MULT_EXPR
> > +      && (TREE_CODE (TREE_OPERAND (arg0, 1)) == INTEGER_CST
> > +          && !TREE_OVERFLOW (TREE_OPERAND (arg0, 1)))
> > +      && integer_zerop (arg1))
> > +    {
> > +      tree const1 = TREE_OPERAND (arg0, 1);
>
> Would you please add:
>
>   gcc_assert (!integer_zerop (const1))
>
> ?
>
> I'm assuming that you know at this point const1 can never be zero,
> because of something somewhere else in this function or in the compiler,
> as otherwise the optimization would of course be invalid.

Indeed.  But if this ever happened we'd have an unfolded tree which
would be another bug.  Of course silent wrong-code in this case is worse
than an ICE, so I agree with the assert.

Richard.