Re: [patch] Fix PR22372

[Dorit Nuzman <DORIT at il dot ibm dot com>]

(1)
> Since we are back in stage 1, could you remove all the forward
> declarations for static functions?  As a separate patch, because you'll
> be forced to also re-arrange the function bodies and that makes for a
> fairly big patch.
>
...
(2)
> >
> Use void_type_node for assignment statements.
>

I was going to do these cleanups for all the vectorizer files separately,
after the two large pending vectorizer projects from 4.2 are in mainline
(to reduce conflicts). The one (multiple data-types) was committed
yesterday, and the second (strided accesses) will probably be committed
next week. Right after that I'll go ahead and clean up the above + the
other things you pointed out in your previous reviews:

(3)
> +   /** Transform.  **/
> +
Use /* */

(4)
> +   if (!vec_stmt) /* transformation not required.  */
> +     {
>
Comment inside the if().

(5)
> +       /* Arguments are ready. Create the new vector stmt.  We are
creating
> +          two vector defs because the widened result does not fit in one
vector.
> +          The vectorized stmt can be expressed as a call to a taregt
builtin,
> +          or a using a tree-code.
> +        */
>
Closing comment should be '.  */'

(6) add missing documentation in c-tree.texi

dorit